1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 1999 - 2018 Intel Corporation. */
3 
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5 
6 #include <linux/module.h>
7 #include <linux/types.h>
8 #include <linux/init.h>
9 #include <linux/pci.h>
10 #include <linux/vmalloc.h>
11 #include <linux/pagemap.h>
12 #include <linux/delay.h>
13 #include <linux/netdevice.h>
14 #include <linux/interrupt.h>
15 #include <linux/tcp.h>
16 #include <linux/ipv6.h>
17 #include <linux/slab.h>
18 #include <net/checksum.h>
19 #include <net/ip6_checksum.h>
20 #include <linux/ethtool.h>
21 #include <linux/if_vlan.h>
22 #include <linux/cpu.h>
23 #include <linux/smp.h>
24 #include <linux/pm_qos.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/aer.h>
27 #include <linux/prefetch.h>
28 #include <linux/suspend.h>
29 
30 #include "e1000.h"
31 
32 char e1000e_driver_name[] = "e1000e";
33 
34 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
35 static int debug = -1;
36 module_param(debug, int, 0);
37 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
38 
39 static const struct e1000_info *e1000_info_tbl[] = {
40 	[board_82571]		= &e1000_82571_info,
41 	[board_82572]		= &e1000_82572_info,
42 	[board_82573]		= &e1000_82573_info,
43 	[board_82574]		= &e1000_82574_info,
44 	[board_82583]		= &e1000_82583_info,
45 	[board_80003es2lan]	= &e1000_es2_info,
46 	[board_ich8lan]		= &e1000_ich8_info,
47 	[board_ich9lan]		= &e1000_ich9_info,
48 	[board_ich10lan]	= &e1000_ich10_info,
49 	[board_pchlan]		= &e1000_pch_info,
50 	[board_pch2lan]		= &e1000_pch2_info,
51 	[board_pch_lpt]		= &e1000_pch_lpt_info,
52 	[board_pch_spt]		= &e1000_pch_spt_info,
53 	[board_pch_cnp]		= &e1000_pch_cnp_info,
54 	[board_pch_tgp]		= &e1000_pch_tgp_info,
55 };
56 
57 struct e1000_reg_info {
58 	u32 ofs;
59 	char *name;
60 };
61 
62 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
63 	/* General Registers */
64 	{E1000_CTRL, "CTRL"},
65 	{E1000_STATUS, "STATUS"},
66 	{E1000_CTRL_EXT, "CTRL_EXT"},
67 
68 	/* Interrupt Registers */
69 	{E1000_ICR, "ICR"},
70 
71 	/* Rx Registers */
72 	{E1000_RCTL, "RCTL"},
73 	{E1000_RDLEN(0), "RDLEN"},
74 	{E1000_RDH(0), "RDH"},
75 	{E1000_RDT(0), "RDT"},
76 	{E1000_RDTR, "RDTR"},
77 	{E1000_RXDCTL(0), "RXDCTL"},
78 	{E1000_ERT, "ERT"},
79 	{E1000_RDBAL(0), "RDBAL"},
80 	{E1000_RDBAH(0), "RDBAH"},
81 	{E1000_RDFH, "RDFH"},
82 	{E1000_RDFT, "RDFT"},
83 	{E1000_RDFHS, "RDFHS"},
84 	{E1000_RDFTS, "RDFTS"},
85 	{E1000_RDFPC, "RDFPC"},
86 
87 	/* Tx Registers */
88 	{E1000_TCTL, "TCTL"},
89 	{E1000_TDBAL(0), "TDBAL"},
90 	{E1000_TDBAH(0), "TDBAH"},
91 	{E1000_TDLEN(0), "TDLEN"},
92 	{E1000_TDH(0), "TDH"},
93 	{E1000_TDT(0), "TDT"},
94 	{E1000_TIDV, "TIDV"},
95 	{E1000_TXDCTL(0), "TXDCTL"},
96 	{E1000_TADV, "TADV"},
97 	{E1000_TARC(0), "TARC"},
98 	{E1000_TDFH, "TDFH"},
99 	{E1000_TDFT, "TDFT"},
100 	{E1000_TDFHS, "TDFHS"},
101 	{E1000_TDFTS, "TDFTS"},
102 	{E1000_TDFPC, "TDFPC"},
103 
104 	/* List Terminator */
105 	{0, NULL}
106 };
107 
108 /**
109  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
110  * @hw: pointer to the HW structure
111  *
112  * When updating the MAC CSR registers, the Manageability Engine (ME) could
113  * be accessing the registers at the same time.  Normally, this is handled in
114  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
115  * accesses later than it should which could result in the register to have
116  * an incorrect value.  Workaround this by checking the FWSM register which
117  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
118  * and try again a number of times.
119  **/
__ew32_prepare(struct e1000_hw * hw)120 static void __ew32_prepare(struct e1000_hw *hw)
121 {
122 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
123 
124 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
125 		udelay(50);
126 }
127 
__ew32(struct e1000_hw * hw,unsigned long reg,u32 val)128 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
129 {
130 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
131 		__ew32_prepare(hw);
132 
133 	writel(val, hw->hw_addr + reg);
134 }
135 
136 /**
137  * e1000_regdump - register printout routine
138  * @hw: pointer to the HW structure
139  * @reginfo: pointer to the register info table
140  **/
e1000_regdump(struct e1000_hw * hw,struct e1000_reg_info * reginfo)141 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
142 {
143 	int n = 0;
144 	char rname[16];
145 	u32 regs[8];
146 
147 	switch (reginfo->ofs) {
148 	case E1000_RXDCTL(0):
149 		for (n = 0; n < 2; n++)
150 			regs[n] = __er32(hw, E1000_RXDCTL(n));
151 		break;
152 	case E1000_TXDCTL(0):
153 		for (n = 0; n < 2; n++)
154 			regs[n] = __er32(hw, E1000_TXDCTL(n));
155 		break;
156 	case E1000_TARC(0):
157 		for (n = 0; n < 2; n++)
158 			regs[n] = __er32(hw, E1000_TARC(n));
159 		break;
160 	default:
161 		pr_info("%-15s %08x\n",
162 			reginfo->name, __er32(hw, reginfo->ofs));
163 		return;
164 	}
165 
166 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
167 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
168 }
169 
e1000e_dump_ps_pages(struct e1000_adapter * adapter,struct e1000_buffer * bi)170 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
171 				 struct e1000_buffer *bi)
172 {
173 	int i;
174 	struct e1000_ps_page *ps_page;
175 
176 	for (i = 0; i < adapter->rx_ps_pages; i++) {
177 		ps_page = &bi->ps_pages[i];
178 
179 		if (ps_page->page) {
180 			pr_info("packet dump for ps_page %d:\n", i);
181 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
182 				       16, 1, page_address(ps_page->page),
183 				       PAGE_SIZE, true);
184 		}
185 	}
186 }
187 
188 /**
189  * e1000e_dump - Print registers, Tx-ring and Rx-ring
190  * @adapter: board private structure
191  **/
e1000e_dump(struct e1000_adapter * adapter)192 static void e1000e_dump(struct e1000_adapter *adapter)
193 {
194 	struct net_device *netdev = adapter->netdev;
195 	struct e1000_hw *hw = &adapter->hw;
196 	struct e1000_reg_info *reginfo;
197 	struct e1000_ring *tx_ring = adapter->tx_ring;
198 	struct e1000_tx_desc *tx_desc;
199 	struct my_u0 {
200 		__le64 a;
201 		__le64 b;
202 	} *u0;
203 	struct e1000_buffer *buffer_info;
204 	struct e1000_ring *rx_ring = adapter->rx_ring;
205 	union e1000_rx_desc_packet_split *rx_desc_ps;
206 	union e1000_rx_desc_extended *rx_desc;
207 	struct my_u1 {
208 		__le64 a;
209 		__le64 b;
210 		__le64 c;
211 		__le64 d;
212 	} *u1;
213 	u32 staterr;
214 	int i = 0;
215 
216 	if (!netif_msg_hw(adapter))
217 		return;
218 
219 	/* Print netdevice Info */
220 	if (netdev) {
221 		dev_info(&adapter->pdev->dev, "Net device Info\n");
222 		pr_info("Device Name     state            trans_start\n");
223 		pr_info("%-15s %016lX %016lX\n", netdev->name,
224 			netdev->state, dev_trans_start(netdev));
225 	}
226 
227 	/* Print Registers */
228 	dev_info(&adapter->pdev->dev, "Register Dump\n");
229 	pr_info(" Register Name   Value\n");
230 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
231 	     reginfo->name; reginfo++) {
232 		e1000_regdump(hw, reginfo);
233 	}
234 
235 	/* Print Tx Ring Summary */
236 	if (!netdev || !netif_running(netdev))
237 		return;
238 
239 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
240 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
241 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
242 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
243 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
244 		(unsigned long long)buffer_info->dma,
245 		buffer_info->length,
246 		buffer_info->next_to_watch,
247 		(unsigned long long)buffer_info->time_stamp);
248 
249 	/* Print Tx Ring */
250 	if (!netif_msg_tx_done(adapter))
251 		goto rx_ring_summary;
252 
253 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
254 
255 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
256 	 *
257 	 * Legacy Transmit Descriptor
258 	 *   +--------------------------------------------------------------+
259 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
260 	 *   +--------------------------------------------------------------+
261 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
262 	 *   +--------------------------------------------------------------+
263 	 *   63       48 47        36 35    32 31     24 23    16 15        0
264 	 *
265 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
266 	 *   63      48 47    40 39       32 31             16 15    8 7      0
267 	 *   +----------------------------------------------------------------+
268 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
269 	 *   +----------------------------------------------------------------+
270 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
271 	 *   +----------------------------------------------------------------+
272 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
273 	 *
274 	 * Extended Data Descriptor (DTYP=0x1)
275 	 *   +----------------------------------------------------------------+
276 	 * 0 |                     Buffer Address [63:0]                      |
277 	 *   +----------------------------------------------------------------+
278 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
279 	 *   +----------------------------------------------------------------+
280 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
281 	 */
282 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
283 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
284 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
285 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
286 		const char *next_desc;
287 		tx_desc = E1000_TX_DESC(*tx_ring, i);
288 		buffer_info = &tx_ring->buffer_info[i];
289 		u0 = (struct my_u0 *)tx_desc;
290 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
291 			next_desc = " NTC/U";
292 		else if (i == tx_ring->next_to_use)
293 			next_desc = " NTU";
294 		else if (i == tx_ring->next_to_clean)
295 			next_desc = " NTC";
296 		else
297 			next_desc = "";
298 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
299 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
300 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
301 			i,
302 			(unsigned long long)le64_to_cpu(u0->a),
303 			(unsigned long long)le64_to_cpu(u0->b),
304 			(unsigned long long)buffer_info->dma,
305 			buffer_info->length, buffer_info->next_to_watch,
306 			(unsigned long long)buffer_info->time_stamp,
307 			buffer_info->skb, next_desc);
308 
309 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
310 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
311 				       16, 1, buffer_info->skb->data,
312 				       buffer_info->skb->len, true);
313 	}
314 
315 	/* Print Rx Ring Summary */
316 rx_ring_summary:
317 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
318 	pr_info("Queue [NTU] [NTC]\n");
319 	pr_info(" %5d %5X %5X\n",
320 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
321 
322 	/* Print Rx Ring */
323 	if (!netif_msg_rx_status(adapter))
324 		return;
325 
326 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
327 	switch (adapter->rx_ps_pages) {
328 	case 1:
329 	case 2:
330 	case 3:
331 		/* [Extended] Packet Split Receive Descriptor Format
332 		 *
333 		 *    +-----------------------------------------------------+
334 		 *  0 |                Buffer Address 0 [63:0]              |
335 		 *    +-----------------------------------------------------+
336 		 *  8 |                Buffer Address 1 [63:0]              |
337 		 *    +-----------------------------------------------------+
338 		 * 16 |                Buffer Address 2 [63:0]              |
339 		 *    +-----------------------------------------------------+
340 		 * 24 |                Buffer Address 3 [63:0]              |
341 		 *    +-----------------------------------------------------+
342 		 */
343 		pr_info("R  [desc]      [buffer 0 63:0 ] [buffer 1 63:0 ] [buffer 2 63:0 ] [buffer 3 63:0 ] [bi->dma       ] [bi->skb] <-- Ext Pkt Split format\n");
344 		/* [Extended] Receive Descriptor (Write-Back) Format
345 		 *
346 		 *   63       48 47    32 31     13 12    8 7    4 3        0
347 		 *   +------------------------------------------------------+
348 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
349 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
350 		 *   +------------------------------------------------------+
351 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
352 		 *   +------------------------------------------------------+
353 		 *   63       48 47    32 31            20 19               0
354 		 */
355 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
356 		for (i = 0; i < rx_ring->count; i++) {
357 			const char *next_desc;
358 			buffer_info = &rx_ring->buffer_info[i];
359 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
360 			u1 = (struct my_u1 *)rx_desc_ps;
361 			staterr =
362 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
363 
364 			if (i == rx_ring->next_to_use)
365 				next_desc = " NTU";
366 			else if (i == rx_ring->next_to_clean)
367 				next_desc = " NTC";
368 			else
369 				next_desc = "";
370 
371 			if (staterr & E1000_RXD_STAT_DD) {
372 				/* Descriptor Done */
373 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
374 					"RWB", i,
375 					(unsigned long long)le64_to_cpu(u1->a),
376 					(unsigned long long)le64_to_cpu(u1->b),
377 					(unsigned long long)le64_to_cpu(u1->c),
378 					(unsigned long long)le64_to_cpu(u1->d),
379 					buffer_info->skb, next_desc);
380 			} else {
381 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
382 					"R  ", i,
383 					(unsigned long long)le64_to_cpu(u1->a),
384 					(unsigned long long)le64_to_cpu(u1->b),
385 					(unsigned long long)le64_to_cpu(u1->c),
386 					(unsigned long long)le64_to_cpu(u1->d),
387 					(unsigned long long)buffer_info->dma,
388 					buffer_info->skb, next_desc);
389 
390 				if (netif_msg_pktdata(adapter))
391 					e1000e_dump_ps_pages(adapter,
392 							     buffer_info);
393 			}
394 		}
395 		break;
396 	default:
397 	case 0:
398 		/* Extended Receive Descriptor (Read) Format
399 		 *
400 		 *   +-----------------------------------------------------+
401 		 * 0 |                Buffer Address [63:0]                |
402 		 *   +-----------------------------------------------------+
403 		 * 8 |                      Reserved                       |
404 		 *   +-----------------------------------------------------+
405 		 */
406 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
407 		/* Extended Receive Descriptor (Write-Back) Format
408 		 *
409 		 *   63       48 47    32 31    24 23            4 3        0
410 		 *   +------------------------------------------------------+
411 		 *   |     RSS Hash      |        |               |         |
412 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
413 		 *   | Packet   | IP     |        |               |  Type   |
414 		 *   | Checksum | Ident  |        |               |         |
415 		 *   +------------------------------------------------------+
416 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
417 		 *   +------------------------------------------------------+
418 		 *   63       48 47    32 31            20 19               0
419 		 */
420 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
421 
422 		for (i = 0; i < rx_ring->count; i++) {
423 			const char *next_desc;
424 
425 			buffer_info = &rx_ring->buffer_info[i];
426 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
427 			u1 = (struct my_u1 *)rx_desc;
428 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
429 
430 			if (i == rx_ring->next_to_use)
431 				next_desc = " NTU";
432 			else if (i == rx_ring->next_to_clean)
433 				next_desc = " NTC";
434 			else
435 				next_desc = "";
436 
437 			if (staterr & E1000_RXD_STAT_DD) {
438 				/* Descriptor Done */
439 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
440 					"RWB", i,
441 					(unsigned long long)le64_to_cpu(u1->a),
442 					(unsigned long long)le64_to_cpu(u1->b),
443 					buffer_info->skb, next_desc);
444 			} else {
445 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
446 					"R  ", i,
447 					(unsigned long long)le64_to_cpu(u1->a),
448 					(unsigned long long)le64_to_cpu(u1->b),
449 					(unsigned long long)buffer_info->dma,
450 					buffer_info->skb, next_desc);
451 
452 				if (netif_msg_pktdata(adapter) &&
453 				    buffer_info->skb)
454 					print_hex_dump(KERN_INFO, "",
455 						       DUMP_PREFIX_ADDRESS, 16,
456 						       1,
457 						       buffer_info->skb->data,
458 						       adapter->rx_buffer_len,
459 						       true);
460 			}
461 		}
462 	}
463 }
464 
465 /**
466  * e1000_desc_unused - calculate if we have unused descriptors
467  * @ring: pointer to ring struct to perform calculation on
468  **/
e1000_desc_unused(struct e1000_ring * ring)469 static int e1000_desc_unused(struct e1000_ring *ring)
470 {
471 	if (ring->next_to_clean > ring->next_to_use)
472 		return ring->next_to_clean - ring->next_to_use - 1;
473 
474 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
475 }
476 
477 /**
478  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
479  * @adapter: board private structure
480  * @hwtstamps: time stamp structure to update
481  * @systim: unsigned 64bit system time value.
482  *
483  * Convert the system time value stored in the RX/TXSTMP registers into a
484  * hwtstamp which can be used by the upper level time stamping functions.
485  *
486  * The 'systim_lock' spinlock is used to protect the consistency of the
487  * system time value. This is needed because reading the 64 bit time
488  * value involves reading two 32 bit registers. The first read latches the
489  * value.
490  **/
e1000e_systim_to_hwtstamp(struct e1000_adapter * adapter,struct skb_shared_hwtstamps * hwtstamps,u64 systim)491 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
492 				      struct skb_shared_hwtstamps *hwtstamps,
493 				      u64 systim)
494 {
495 	u64 ns;
496 	unsigned long flags;
497 
498 	spin_lock_irqsave(&adapter->systim_lock, flags);
499 	ns = timecounter_cyc2time(&adapter->tc, systim);
500 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
501 
502 	memset(hwtstamps, 0, sizeof(*hwtstamps));
503 	hwtstamps->hwtstamp = ns_to_ktime(ns);
504 }
505 
506 /**
507  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
508  * @adapter: board private structure
509  * @status: descriptor extended error and status field
510  * @skb: particular skb to include time stamp
511  *
512  * If the time stamp is valid, convert it into the timecounter ns value
513  * and store that result into the shhwtstamps structure which is passed
514  * up the network stack.
515  **/
e1000e_rx_hwtstamp(struct e1000_adapter * adapter,u32 status,struct sk_buff * skb)516 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
517 			       struct sk_buff *skb)
518 {
519 	struct e1000_hw *hw = &adapter->hw;
520 	u64 rxstmp;
521 
522 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
523 	    !(status & E1000_RXDEXT_STATERR_TST) ||
524 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
525 		return;
526 
527 	/* The Rx time stamp registers contain the time stamp.  No other
528 	 * received packet will be time stamped until the Rx time stamp
529 	 * registers are read.  Because only one packet can be time stamped
530 	 * at a time, the register values must belong to this packet and
531 	 * therefore none of the other additional attributes need to be
532 	 * compared.
533 	 */
534 	rxstmp = (u64)er32(RXSTMPL);
535 	rxstmp |= (u64)er32(RXSTMPH) << 32;
536 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
537 
538 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
539 }
540 
541 /**
542  * e1000_receive_skb - helper function to handle Rx indications
543  * @adapter: board private structure
544  * @netdev: pointer to netdev struct
545  * @staterr: descriptor extended error and status field as written by hardware
546  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
547  * @skb: pointer to sk_buff to be indicated to stack
548  **/
e1000_receive_skb(struct e1000_adapter * adapter,struct net_device * netdev,struct sk_buff * skb,u32 staterr,__le16 vlan)549 static void e1000_receive_skb(struct e1000_adapter *adapter,
550 			      struct net_device *netdev, struct sk_buff *skb,
551 			      u32 staterr, __le16 vlan)
552 {
553 	u16 tag = le16_to_cpu(vlan);
554 
555 	e1000e_rx_hwtstamp(adapter, staterr, skb);
556 
557 	skb->protocol = eth_type_trans(skb, netdev);
558 
559 	if (staterr & E1000_RXD_STAT_VP)
560 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
561 
562 	napi_gro_receive(&adapter->napi, skb);
563 }
564 
565 /**
566  * e1000_rx_checksum - Receive Checksum Offload
567  * @adapter: board private structure
568  * @status_err: receive descriptor status and error fields
569  * @skb: socket buffer with received data
570  **/
e1000_rx_checksum(struct e1000_adapter * adapter,u32 status_err,struct sk_buff * skb)571 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
572 			      struct sk_buff *skb)
573 {
574 	u16 status = (u16)status_err;
575 	u8 errors = (u8)(status_err >> 24);
576 
577 	skb_checksum_none_assert(skb);
578 
579 	/* Rx checksum disabled */
580 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
581 		return;
582 
583 	/* Ignore Checksum bit is set */
584 	if (status & E1000_RXD_STAT_IXSM)
585 		return;
586 
587 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
588 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
589 		/* let the stack verify checksum errors */
590 		adapter->hw_csum_err++;
591 		return;
592 	}
593 
594 	/* TCP/UDP Checksum has not been calculated */
595 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
596 		return;
597 
598 	/* It must be a TCP or UDP packet with a valid checksum */
599 	skb->ip_summed = CHECKSUM_UNNECESSARY;
600 	adapter->hw_csum_good++;
601 }
602 
e1000e_update_rdt_wa(struct e1000_ring * rx_ring,unsigned int i)603 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
604 {
605 	struct e1000_adapter *adapter = rx_ring->adapter;
606 	struct e1000_hw *hw = &adapter->hw;
607 
608 	__ew32_prepare(hw);
609 	writel(i, rx_ring->tail);
610 
611 	if (unlikely(i != readl(rx_ring->tail))) {
612 		u32 rctl = er32(RCTL);
613 
614 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
615 		e_err("ME firmware caused invalid RDT - resetting\n");
616 		schedule_work(&adapter->reset_task);
617 	}
618 }
619 
e1000e_update_tdt_wa(struct e1000_ring * tx_ring,unsigned int i)620 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
621 {
622 	struct e1000_adapter *adapter = tx_ring->adapter;
623 	struct e1000_hw *hw = &adapter->hw;
624 
625 	__ew32_prepare(hw);
626 	writel(i, tx_ring->tail);
627 
628 	if (unlikely(i != readl(tx_ring->tail))) {
629 		u32 tctl = er32(TCTL);
630 
631 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
632 		e_err("ME firmware caused invalid TDT - resetting\n");
633 		schedule_work(&adapter->reset_task);
634 	}
635 }
636 
637 /**
638  * e1000_alloc_rx_buffers - Replace used receive buffers
639  * @rx_ring: Rx descriptor ring
640  * @cleaned_count: number to reallocate
641  * @gfp: flags for allocation
642  **/
e1000_alloc_rx_buffers(struct e1000_ring * rx_ring,int cleaned_count,gfp_t gfp)643 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
644 				   int cleaned_count, gfp_t gfp)
645 {
646 	struct e1000_adapter *adapter = rx_ring->adapter;
647 	struct net_device *netdev = adapter->netdev;
648 	struct pci_dev *pdev = adapter->pdev;
649 	union e1000_rx_desc_extended *rx_desc;
650 	struct e1000_buffer *buffer_info;
651 	struct sk_buff *skb;
652 	unsigned int i;
653 	unsigned int bufsz = adapter->rx_buffer_len;
654 
655 	i = rx_ring->next_to_use;
656 	buffer_info = &rx_ring->buffer_info[i];
657 
658 	while (cleaned_count--) {
659 		skb = buffer_info->skb;
660 		if (skb) {
661 			skb_trim(skb, 0);
662 			goto map_skb;
663 		}
664 
665 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
666 		if (!skb) {
667 			/* Better luck next round */
668 			adapter->alloc_rx_buff_failed++;
669 			break;
670 		}
671 
672 		buffer_info->skb = skb;
673 map_skb:
674 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
675 						  adapter->rx_buffer_len,
676 						  DMA_FROM_DEVICE);
677 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
678 			dev_err(&pdev->dev, "Rx DMA map failed\n");
679 			adapter->rx_dma_failed++;
680 			break;
681 		}
682 
683 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
684 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
685 
686 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
687 			/* Force memory writes to complete before letting h/w
688 			 * know there are new descriptors to fetch.  (Only
689 			 * applicable for weak-ordered memory model archs,
690 			 * such as IA-64).
691 			 */
692 			wmb();
693 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
694 				e1000e_update_rdt_wa(rx_ring, i);
695 			else
696 				writel(i, rx_ring->tail);
697 		}
698 		i++;
699 		if (i == rx_ring->count)
700 			i = 0;
701 		buffer_info = &rx_ring->buffer_info[i];
702 	}
703 
704 	rx_ring->next_to_use = i;
705 }
706 
707 /**
708  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
709  * @rx_ring: Rx descriptor ring
710  * @cleaned_count: number to reallocate
711  * @gfp: flags for allocation
712  **/
e1000_alloc_rx_buffers_ps(struct e1000_ring * rx_ring,int cleaned_count,gfp_t gfp)713 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
714 				      int cleaned_count, gfp_t gfp)
715 {
716 	struct e1000_adapter *adapter = rx_ring->adapter;
717 	struct net_device *netdev = adapter->netdev;
718 	struct pci_dev *pdev = adapter->pdev;
719 	union e1000_rx_desc_packet_split *rx_desc;
720 	struct e1000_buffer *buffer_info;
721 	struct e1000_ps_page *ps_page;
722 	struct sk_buff *skb;
723 	unsigned int i, j;
724 
725 	i = rx_ring->next_to_use;
726 	buffer_info = &rx_ring->buffer_info[i];
727 
728 	while (cleaned_count--) {
729 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
730 
731 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
732 			ps_page = &buffer_info->ps_pages[j];
733 			if (j >= adapter->rx_ps_pages) {
734 				/* all unused desc entries get hw null ptr */
735 				rx_desc->read.buffer_addr[j + 1] =
736 				    ~cpu_to_le64(0);
737 				continue;
738 			}
739 			if (!ps_page->page) {
740 				ps_page->page = alloc_page(gfp);
741 				if (!ps_page->page) {
742 					adapter->alloc_rx_buff_failed++;
743 					goto no_buffers;
744 				}
745 				ps_page->dma = dma_map_page(&pdev->dev,
746 							    ps_page->page,
747 							    0, PAGE_SIZE,
748 							    DMA_FROM_DEVICE);
749 				if (dma_mapping_error(&pdev->dev,
750 						      ps_page->dma)) {
751 					dev_err(&adapter->pdev->dev,
752 						"Rx DMA page map failed\n");
753 					adapter->rx_dma_failed++;
754 					goto no_buffers;
755 				}
756 			}
757 			/* Refresh the desc even if buffer_addrs
758 			 * didn't change because each write-back
759 			 * erases this info.
760 			 */
761 			rx_desc->read.buffer_addr[j + 1] =
762 			    cpu_to_le64(ps_page->dma);
763 		}
764 
765 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
766 						  gfp);
767 
768 		if (!skb) {
769 			adapter->alloc_rx_buff_failed++;
770 			break;
771 		}
772 
773 		buffer_info->skb = skb;
774 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
775 						  adapter->rx_ps_bsize0,
776 						  DMA_FROM_DEVICE);
777 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
778 			dev_err(&pdev->dev, "Rx DMA map failed\n");
779 			adapter->rx_dma_failed++;
780 			/* cleanup skb */
781 			dev_kfree_skb_any(skb);
782 			buffer_info->skb = NULL;
783 			break;
784 		}
785 
786 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
787 
788 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
789 			/* Force memory writes to complete before letting h/w
790 			 * know there are new descriptors to fetch.  (Only
791 			 * applicable for weak-ordered memory model archs,
792 			 * such as IA-64).
793 			 */
794 			wmb();
795 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
796 				e1000e_update_rdt_wa(rx_ring, i << 1);
797 			else
798 				writel(i << 1, rx_ring->tail);
799 		}
800 
801 		i++;
802 		if (i == rx_ring->count)
803 			i = 0;
804 		buffer_info = &rx_ring->buffer_info[i];
805 	}
806 
807 no_buffers:
808 	rx_ring->next_to_use = i;
809 }
810 
811 /**
812  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
813  * @rx_ring: Rx descriptor ring
814  * @cleaned_count: number of buffers to allocate this pass
815  * @gfp: flags for allocation
816  **/
817 
e1000_alloc_jumbo_rx_buffers(struct e1000_ring * rx_ring,int cleaned_count,gfp_t gfp)818 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
819 					 int cleaned_count, gfp_t gfp)
820 {
821 	struct e1000_adapter *adapter = rx_ring->adapter;
822 	struct net_device *netdev = adapter->netdev;
823 	struct pci_dev *pdev = adapter->pdev;
824 	union e1000_rx_desc_extended *rx_desc;
825 	struct e1000_buffer *buffer_info;
826 	struct sk_buff *skb;
827 	unsigned int i;
828 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
829 
830 	i = rx_ring->next_to_use;
831 	buffer_info = &rx_ring->buffer_info[i];
832 
833 	while (cleaned_count--) {
834 		skb = buffer_info->skb;
835 		if (skb) {
836 			skb_trim(skb, 0);
837 			goto check_page;
838 		}
839 
840 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
841 		if (unlikely(!skb)) {
842 			/* Better luck next round */
843 			adapter->alloc_rx_buff_failed++;
844 			break;
845 		}
846 
847 		buffer_info->skb = skb;
848 check_page:
849 		/* allocate a new page if necessary */
850 		if (!buffer_info->page) {
851 			buffer_info->page = alloc_page(gfp);
852 			if (unlikely(!buffer_info->page)) {
853 				adapter->alloc_rx_buff_failed++;
854 				break;
855 			}
856 		}
857 
858 		if (!buffer_info->dma) {
859 			buffer_info->dma = dma_map_page(&pdev->dev,
860 							buffer_info->page, 0,
861 							PAGE_SIZE,
862 							DMA_FROM_DEVICE);
863 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
864 				adapter->alloc_rx_buff_failed++;
865 				break;
866 			}
867 		}
868 
869 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
870 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
871 
872 		if (unlikely(++i == rx_ring->count))
873 			i = 0;
874 		buffer_info = &rx_ring->buffer_info[i];
875 	}
876 
877 	if (likely(rx_ring->next_to_use != i)) {
878 		rx_ring->next_to_use = i;
879 		if (unlikely(i-- == 0))
880 			i = (rx_ring->count - 1);
881 
882 		/* Force memory writes to complete before letting h/w
883 		 * know there are new descriptors to fetch.  (Only
884 		 * applicable for weak-ordered memory model archs,
885 		 * such as IA-64).
886 		 */
887 		wmb();
888 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
889 			e1000e_update_rdt_wa(rx_ring, i);
890 		else
891 			writel(i, rx_ring->tail);
892 	}
893 }
894 
e1000_rx_hash(struct net_device * netdev,__le32 rss,struct sk_buff * skb)895 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
896 				 struct sk_buff *skb)
897 {
898 	if (netdev->features & NETIF_F_RXHASH)
899 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
900 }
901 
902 /**
903  * e1000_clean_rx_irq - Send received data up the network stack
904  * @rx_ring: Rx descriptor ring
905  * @work_done: output parameter for indicating completed work
906  * @work_to_do: how many packets we can clean
907  *
908  * the return value indicates whether actual cleaning was done, there
909  * is no guarantee that everything was cleaned
910  **/
e1000_clean_rx_irq(struct e1000_ring * rx_ring,int * work_done,int work_to_do)911 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
912 			       int work_to_do)
913 {
914 	struct e1000_adapter *adapter = rx_ring->adapter;
915 	struct net_device *netdev = adapter->netdev;
916 	struct pci_dev *pdev = adapter->pdev;
917 	struct e1000_hw *hw = &adapter->hw;
918 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
919 	struct e1000_buffer *buffer_info, *next_buffer;
920 	u32 length, staterr;
921 	unsigned int i;
922 	int cleaned_count = 0;
923 	bool cleaned = false;
924 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
925 
926 	i = rx_ring->next_to_clean;
927 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
928 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
929 	buffer_info = &rx_ring->buffer_info[i];
930 
931 	while (staterr & E1000_RXD_STAT_DD) {
932 		struct sk_buff *skb;
933 
934 		if (*work_done >= work_to_do)
935 			break;
936 		(*work_done)++;
937 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
938 
939 		skb = buffer_info->skb;
940 		buffer_info->skb = NULL;
941 
942 		prefetch(skb->data - NET_IP_ALIGN);
943 
944 		i++;
945 		if (i == rx_ring->count)
946 			i = 0;
947 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
948 		prefetch(next_rxd);
949 
950 		next_buffer = &rx_ring->buffer_info[i];
951 
952 		cleaned = true;
953 		cleaned_count++;
954 		dma_unmap_single(&pdev->dev, buffer_info->dma,
955 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
956 		buffer_info->dma = 0;
957 
958 		length = le16_to_cpu(rx_desc->wb.upper.length);
959 
960 		/* !EOP means multiple descriptors were used to store a single
961 		 * packet, if that's the case we need to toss it.  In fact, we
962 		 * need to toss every packet with the EOP bit clear and the
963 		 * next frame that _does_ have the EOP bit set, as it is by
964 		 * definition only a frame fragment
965 		 */
966 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
967 			adapter->flags2 |= FLAG2_IS_DISCARDING;
968 
969 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
970 			/* All receives must fit into a single buffer */
971 			e_dbg("Receive packet consumed multiple buffers\n");
972 			/* recycle */
973 			buffer_info->skb = skb;
974 			if (staterr & E1000_RXD_STAT_EOP)
975 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
976 			goto next_desc;
977 		}
978 
979 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
980 			     !(netdev->features & NETIF_F_RXALL))) {
981 			/* recycle */
982 			buffer_info->skb = skb;
983 			goto next_desc;
984 		}
985 
986 		/* adjust length to remove Ethernet CRC */
987 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
988 			/* If configured to store CRC, don't subtract FCS,
989 			 * but keep the FCS bytes out of the total_rx_bytes
990 			 * counter
991 			 */
992 			if (netdev->features & NETIF_F_RXFCS)
993 				total_rx_bytes -= 4;
994 			else
995 				length -= 4;
996 		}
997 
998 		total_rx_bytes += length;
999 		total_rx_packets++;
1000 
1001 		/* code added for copybreak, this should improve
1002 		 * performance for small packets with large amounts
1003 		 * of reassembly being done in the stack
1004 		 */
1005 		if (length < copybreak) {
1006 			struct sk_buff *new_skb =
1007 				napi_alloc_skb(&adapter->napi, length);
1008 			if (new_skb) {
1009 				skb_copy_to_linear_data_offset(new_skb,
1010 							       -NET_IP_ALIGN,
1011 							       (skb->data -
1012 								NET_IP_ALIGN),
1013 							       (length +
1014 								NET_IP_ALIGN));
1015 				/* save the skb in buffer_info as good */
1016 				buffer_info->skb = skb;
1017 				skb = new_skb;
1018 			}
1019 			/* else just continue with the old one */
1020 		}
1021 		/* end copybreak code */
1022 		skb_put(skb, length);
1023 
1024 		/* Receive Checksum Offload */
1025 		e1000_rx_checksum(adapter, staterr, skb);
1026 
1027 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1028 
1029 		e1000_receive_skb(adapter, netdev, skb, staterr,
1030 				  rx_desc->wb.upper.vlan);
1031 
1032 next_desc:
1033 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1034 
1035 		/* return some buffers to hardware, one at a time is too slow */
1036 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1037 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1038 					      GFP_ATOMIC);
1039 			cleaned_count = 0;
1040 		}
1041 
1042 		/* use prefetched values */
1043 		rx_desc = next_rxd;
1044 		buffer_info = next_buffer;
1045 
1046 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1047 	}
1048 	rx_ring->next_to_clean = i;
1049 
1050 	cleaned_count = e1000_desc_unused(rx_ring);
1051 	if (cleaned_count)
1052 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1053 
1054 	adapter->total_rx_bytes += total_rx_bytes;
1055 	adapter->total_rx_packets += total_rx_packets;
1056 	return cleaned;
1057 }
1058 
e1000_put_txbuf(struct e1000_ring * tx_ring,struct e1000_buffer * buffer_info,bool drop)1059 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1060 			    struct e1000_buffer *buffer_info,
1061 			    bool drop)
1062 {
1063 	struct e1000_adapter *adapter = tx_ring->adapter;
1064 
1065 	if (buffer_info->dma) {
1066 		if (buffer_info->mapped_as_page)
1067 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1068 				       buffer_info->length, DMA_TO_DEVICE);
1069 		else
1070 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1071 					 buffer_info->length, DMA_TO_DEVICE);
1072 		buffer_info->dma = 0;
1073 	}
1074 	if (buffer_info->skb) {
1075 		if (drop)
1076 			dev_kfree_skb_any(buffer_info->skb);
1077 		else
1078 			dev_consume_skb_any(buffer_info->skb);
1079 		buffer_info->skb = NULL;
1080 	}
1081 	buffer_info->time_stamp = 0;
1082 }
1083 
e1000_print_hw_hang(struct work_struct * work)1084 static void e1000_print_hw_hang(struct work_struct *work)
1085 {
1086 	struct e1000_adapter *adapter = container_of(work,
1087 						     struct e1000_adapter,
1088 						     print_hang_task);
1089 	struct net_device *netdev = adapter->netdev;
1090 	struct e1000_ring *tx_ring = adapter->tx_ring;
1091 	unsigned int i = tx_ring->next_to_clean;
1092 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1093 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1094 	struct e1000_hw *hw = &adapter->hw;
1095 	u16 phy_status, phy_1000t_status, phy_ext_status;
1096 	u16 pci_status;
1097 
1098 	if (test_bit(__E1000_DOWN, &adapter->state))
1099 		return;
1100 
1101 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1102 		/* May be block on write-back, flush and detect again
1103 		 * flush pending descriptor writebacks to memory
1104 		 */
1105 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1106 		/* execute the writes immediately */
1107 		e1e_flush();
1108 		/* Due to rare timing issues, write to TIDV again to ensure
1109 		 * the write is successful
1110 		 */
1111 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1112 		/* execute the writes immediately */
1113 		e1e_flush();
1114 		adapter->tx_hang_recheck = true;
1115 		return;
1116 	}
1117 	adapter->tx_hang_recheck = false;
1118 
1119 	if (er32(TDH(0)) == er32(TDT(0))) {
1120 		e_dbg("false hang detected, ignoring\n");
1121 		return;
1122 	}
1123 
1124 	/* Real hang detected */
1125 	netif_stop_queue(netdev);
1126 
1127 	e1e_rphy(hw, MII_BMSR, &phy_status);
1128 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1129 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1130 
1131 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1132 
1133 	/* detected Hardware unit hang */
1134 	e_err("Detected Hardware Unit Hang:\n"
1135 	      "  TDH                  <%x>\n"
1136 	      "  TDT                  <%x>\n"
1137 	      "  next_to_use          <%x>\n"
1138 	      "  next_to_clean        <%x>\n"
1139 	      "buffer_info[next_to_clean]:\n"
1140 	      "  time_stamp           <%lx>\n"
1141 	      "  next_to_watch        <%x>\n"
1142 	      "  jiffies              <%lx>\n"
1143 	      "  next_to_watch.status <%x>\n"
1144 	      "MAC Status             <%x>\n"
1145 	      "PHY Status             <%x>\n"
1146 	      "PHY 1000BASE-T Status  <%x>\n"
1147 	      "PHY Extended Status    <%x>\n"
1148 	      "PCI Status             <%x>\n",
1149 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1150 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1151 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1152 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1153 
1154 	e1000e_dump(adapter);
1155 
1156 	/* Suggest workaround for known h/w issue */
1157 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1158 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1159 }
1160 
1161 /**
1162  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1163  * @work: pointer to work struct
1164  *
1165  * This work function polls the TSYNCTXCTL valid bit to determine when a
1166  * timestamp has been taken for the current stored skb.  The timestamp must
1167  * be for this skb because only one such packet is allowed in the queue.
1168  */
e1000e_tx_hwtstamp_work(struct work_struct * work)1169 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1170 {
1171 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1172 						     tx_hwtstamp_work);
1173 	struct e1000_hw *hw = &adapter->hw;
1174 
1175 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1176 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1177 		struct skb_shared_hwtstamps shhwtstamps;
1178 		u64 txstmp;
1179 
1180 		txstmp = er32(TXSTMPL);
1181 		txstmp |= (u64)er32(TXSTMPH) << 32;
1182 
1183 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1184 
1185 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1186 		 * prior to notifying the stack of a Tx timestamp.
1187 		 */
1188 		adapter->tx_hwtstamp_skb = NULL;
1189 		wmb(); /* force write prior to skb_tstamp_tx */
1190 
1191 		skb_tstamp_tx(skb, &shhwtstamps);
1192 		dev_consume_skb_any(skb);
1193 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1194 			      + adapter->tx_timeout_factor * HZ)) {
1195 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1196 		adapter->tx_hwtstamp_skb = NULL;
1197 		adapter->tx_hwtstamp_timeouts++;
1198 		e_warn("clearing Tx timestamp hang\n");
1199 	} else {
1200 		/* reschedule to check later */
1201 		schedule_work(&adapter->tx_hwtstamp_work);
1202 	}
1203 }
1204 
1205 /**
1206  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1207  * @tx_ring: Tx descriptor ring
1208  *
1209  * the return value indicates whether actual cleaning was done, there
1210  * is no guarantee that everything was cleaned
1211  **/
e1000_clean_tx_irq(struct e1000_ring * tx_ring)1212 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1213 {
1214 	struct e1000_adapter *adapter = tx_ring->adapter;
1215 	struct net_device *netdev = adapter->netdev;
1216 	struct e1000_hw *hw = &adapter->hw;
1217 	struct e1000_tx_desc *tx_desc, *eop_desc;
1218 	struct e1000_buffer *buffer_info;
1219 	unsigned int i, eop;
1220 	unsigned int count = 0;
1221 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1222 	unsigned int bytes_compl = 0, pkts_compl = 0;
1223 
1224 	i = tx_ring->next_to_clean;
1225 	eop = tx_ring->buffer_info[i].next_to_watch;
1226 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1227 
1228 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1229 	       (count < tx_ring->count)) {
1230 		bool cleaned = false;
1231 
1232 		dma_rmb();		/* read buffer_info after eop_desc */
1233 		for (; !cleaned; count++) {
1234 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1235 			buffer_info = &tx_ring->buffer_info[i];
1236 			cleaned = (i == eop);
1237 
1238 			if (cleaned) {
1239 				total_tx_packets += buffer_info->segs;
1240 				total_tx_bytes += buffer_info->bytecount;
1241 				if (buffer_info->skb) {
1242 					bytes_compl += buffer_info->skb->len;
1243 					pkts_compl++;
1244 				}
1245 			}
1246 
1247 			e1000_put_txbuf(tx_ring, buffer_info, false);
1248 			tx_desc->upper.data = 0;
1249 
1250 			i++;
1251 			if (i == tx_ring->count)
1252 				i = 0;
1253 		}
1254 
1255 		if (i == tx_ring->next_to_use)
1256 			break;
1257 		eop = tx_ring->buffer_info[i].next_to_watch;
1258 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1259 	}
1260 
1261 	tx_ring->next_to_clean = i;
1262 
1263 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1264 
1265 #define TX_WAKE_THRESHOLD 32
1266 	if (count && netif_carrier_ok(netdev) &&
1267 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1268 		/* Make sure that anybody stopping the queue after this
1269 		 * sees the new next_to_clean.
1270 		 */
1271 		smp_mb();
1272 
1273 		if (netif_queue_stopped(netdev) &&
1274 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1275 			netif_wake_queue(netdev);
1276 			++adapter->restart_queue;
1277 		}
1278 	}
1279 
1280 	if (adapter->detect_tx_hung) {
1281 		/* Detect a transmit hang in hardware, this serializes the
1282 		 * check with the clearing of time_stamp and movement of i
1283 		 */
1284 		adapter->detect_tx_hung = false;
1285 		if (tx_ring->buffer_info[i].time_stamp &&
1286 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1287 			       + (adapter->tx_timeout_factor * HZ)) &&
1288 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1289 			schedule_work(&adapter->print_hang_task);
1290 		else
1291 			adapter->tx_hang_recheck = false;
1292 	}
1293 	adapter->total_tx_bytes += total_tx_bytes;
1294 	adapter->total_tx_packets += total_tx_packets;
1295 	return count < tx_ring->count;
1296 }
1297 
1298 /**
1299  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1300  * @rx_ring: Rx descriptor ring
1301  * @work_done: output parameter for indicating completed work
1302  * @work_to_do: how many packets we can clean
1303  *
1304  * the return value indicates whether actual cleaning was done, there
1305  * is no guarantee that everything was cleaned
1306  **/
e1000_clean_rx_irq_ps(struct e1000_ring * rx_ring,int * work_done,int work_to_do)1307 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1308 				  int work_to_do)
1309 {
1310 	struct e1000_adapter *adapter = rx_ring->adapter;
1311 	struct e1000_hw *hw = &adapter->hw;
1312 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1313 	struct net_device *netdev = adapter->netdev;
1314 	struct pci_dev *pdev = adapter->pdev;
1315 	struct e1000_buffer *buffer_info, *next_buffer;
1316 	struct e1000_ps_page *ps_page;
1317 	struct sk_buff *skb;
1318 	unsigned int i, j;
1319 	u32 length, staterr;
1320 	int cleaned_count = 0;
1321 	bool cleaned = false;
1322 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1323 
1324 	i = rx_ring->next_to_clean;
1325 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1326 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1327 	buffer_info = &rx_ring->buffer_info[i];
1328 
1329 	while (staterr & E1000_RXD_STAT_DD) {
1330 		if (*work_done >= work_to_do)
1331 			break;
1332 		(*work_done)++;
1333 		skb = buffer_info->skb;
1334 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1335 
1336 		/* in the packet split case this is header only */
1337 		prefetch(skb->data - NET_IP_ALIGN);
1338 
1339 		i++;
1340 		if (i == rx_ring->count)
1341 			i = 0;
1342 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1343 		prefetch(next_rxd);
1344 
1345 		next_buffer = &rx_ring->buffer_info[i];
1346 
1347 		cleaned = true;
1348 		cleaned_count++;
1349 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1350 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1351 		buffer_info->dma = 0;
1352 
1353 		/* see !EOP comment in other Rx routine */
1354 		if (!(staterr & E1000_RXD_STAT_EOP))
1355 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1356 
1357 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1358 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1359 			dev_kfree_skb_irq(skb);
1360 			if (staterr & E1000_RXD_STAT_EOP)
1361 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1362 			goto next_desc;
1363 		}
1364 
1365 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1366 			     !(netdev->features & NETIF_F_RXALL))) {
1367 			dev_kfree_skb_irq(skb);
1368 			goto next_desc;
1369 		}
1370 
1371 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1372 
1373 		if (!length) {
1374 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1375 			dev_kfree_skb_irq(skb);
1376 			goto next_desc;
1377 		}
1378 
1379 		/* Good Receive */
1380 		skb_put(skb, length);
1381 
1382 		{
1383 			/* this looks ugly, but it seems compiler issues make
1384 			 * it more efficient than reusing j
1385 			 */
1386 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1387 
1388 			/* page alloc/put takes too long and effects small
1389 			 * packet throughput, so unsplit small packets and
1390 			 * save the alloc/put only valid in softirq (napi)
1391 			 * context to call kmap_*
1392 			 */
1393 			if (l1 && (l1 <= copybreak) &&
1394 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1395 				u8 *vaddr;
1396 
1397 				ps_page = &buffer_info->ps_pages[0];
1398 
1399 				/* there is no documentation about how to call
1400 				 * kmap_atomic, so we can't hold the mapping
1401 				 * very long
1402 				 */
1403 				dma_sync_single_for_cpu(&pdev->dev,
1404 							ps_page->dma,
1405 							PAGE_SIZE,
1406 							DMA_FROM_DEVICE);
1407 				vaddr = kmap_atomic(ps_page->page);
1408 				memcpy(skb_tail_pointer(skb), vaddr, l1);
1409 				kunmap_atomic(vaddr);
1410 				dma_sync_single_for_device(&pdev->dev,
1411 							   ps_page->dma,
1412 							   PAGE_SIZE,
1413 							   DMA_FROM_DEVICE);
1414 
1415 				/* remove the CRC */
1416 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1417 					if (!(netdev->features & NETIF_F_RXFCS))
1418 						l1 -= 4;
1419 				}
1420 
1421 				skb_put(skb, l1);
1422 				goto copydone;
1423 			}	/* if */
1424 		}
1425 
1426 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1427 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1428 			if (!length)
1429 				break;
1430 
1431 			ps_page = &buffer_info->ps_pages[j];
1432 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1433 				       DMA_FROM_DEVICE);
1434 			ps_page->dma = 0;
1435 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1436 			ps_page->page = NULL;
1437 			skb->len += length;
1438 			skb->data_len += length;
1439 			skb->truesize += PAGE_SIZE;
1440 		}
1441 
1442 		/* strip the ethernet crc, problem is we're using pages now so
1443 		 * this whole operation can get a little cpu intensive
1444 		 */
1445 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1446 			if (!(netdev->features & NETIF_F_RXFCS))
1447 				pskb_trim(skb, skb->len - 4);
1448 		}
1449 
1450 copydone:
1451 		total_rx_bytes += skb->len;
1452 		total_rx_packets++;
1453 
1454 		e1000_rx_checksum(adapter, staterr, skb);
1455 
1456 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1457 
1458 		if (rx_desc->wb.upper.header_status &
1459 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1460 			adapter->rx_hdr_split++;
1461 
1462 		e1000_receive_skb(adapter, netdev, skb, staterr,
1463 				  rx_desc->wb.middle.vlan);
1464 
1465 next_desc:
1466 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1467 		buffer_info->skb = NULL;
1468 
1469 		/* return some buffers to hardware, one at a time is too slow */
1470 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1471 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1472 					      GFP_ATOMIC);
1473 			cleaned_count = 0;
1474 		}
1475 
1476 		/* use prefetched values */
1477 		rx_desc = next_rxd;
1478 		buffer_info = next_buffer;
1479 
1480 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1481 	}
1482 	rx_ring->next_to_clean = i;
1483 
1484 	cleaned_count = e1000_desc_unused(rx_ring);
1485 	if (cleaned_count)
1486 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1487 
1488 	adapter->total_rx_bytes += total_rx_bytes;
1489 	adapter->total_rx_packets += total_rx_packets;
1490 	return cleaned;
1491 }
1492 
e1000_consume_page(struct e1000_buffer * bi,struct sk_buff * skb,u16 length)1493 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1494 			       u16 length)
1495 {
1496 	bi->page = NULL;
1497 	skb->len += length;
1498 	skb->data_len += length;
1499 	skb->truesize += PAGE_SIZE;
1500 }
1501 
1502 /**
1503  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1504  * @rx_ring: Rx descriptor ring
1505  * @work_done: output parameter for indicating completed work
1506  * @work_to_do: how many packets we can clean
1507  *
1508  * the return value indicates whether actual cleaning was done, there
1509  * is no guarantee that everything was cleaned
1510  **/
e1000_clean_jumbo_rx_irq(struct e1000_ring * rx_ring,int * work_done,int work_to_do)1511 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1512 				     int work_to_do)
1513 {
1514 	struct e1000_adapter *adapter = rx_ring->adapter;
1515 	struct net_device *netdev = adapter->netdev;
1516 	struct pci_dev *pdev = adapter->pdev;
1517 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1518 	struct e1000_buffer *buffer_info, *next_buffer;
1519 	u32 length, staterr;
1520 	unsigned int i;
1521 	int cleaned_count = 0;
1522 	bool cleaned = false;
1523 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1524 	struct skb_shared_info *shinfo;
1525 
1526 	i = rx_ring->next_to_clean;
1527 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1528 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1529 	buffer_info = &rx_ring->buffer_info[i];
1530 
1531 	while (staterr & E1000_RXD_STAT_DD) {
1532 		struct sk_buff *skb;
1533 
1534 		if (*work_done >= work_to_do)
1535 			break;
1536 		(*work_done)++;
1537 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1538 
1539 		skb = buffer_info->skb;
1540 		buffer_info->skb = NULL;
1541 
1542 		++i;
1543 		if (i == rx_ring->count)
1544 			i = 0;
1545 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1546 		prefetch(next_rxd);
1547 
1548 		next_buffer = &rx_ring->buffer_info[i];
1549 
1550 		cleaned = true;
1551 		cleaned_count++;
1552 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1553 			       DMA_FROM_DEVICE);
1554 		buffer_info->dma = 0;
1555 
1556 		length = le16_to_cpu(rx_desc->wb.upper.length);
1557 
1558 		/* errors is only valid for DD + EOP descriptors */
1559 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1560 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1561 			      !(netdev->features & NETIF_F_RXALL)))) {
1562 			/* recycle both page and skb */
1563 			buffer_info->skb = skb;
1564 			/* an error means any chain goes out the window too */
1565 			if (rx_ring->rx_skb_top)
1566 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1567 			rx_ring->rx_skb_top = NULL;
1568 			goto next_desc;
1569 		}
1570 #define rxtop (rx_ring->rx_skb_top)
1571 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1572 			/* this descriptor is only the beginning (or middle) */
1573 			if (!rxtop) {
1574 				/* this is the beginning of a chain */
1575 				rxtop = skb;
1576 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1577 						   0, length);
1578 			} else {
1579 				/* this is the middle of a chain */
1580 				shinfo = skb_shinfo(rxtop);
1581 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1582 						   buffer_info->page, 0,
1583 						   length);
1584 				/* re-use the skb, only consumed the page */
1585 				buffer_info->skb = skb;
1586 			}
1587 			e1000_consume_page(buffer_info, rxtop, length);
1588 			goto next_desc;
1589 		} else {
1590 			if (rxtop) {
1591 				/* end of the chain */
1592 				shinfo = skb_shinfo(rxtop);
1593 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1594 						   buffer_info->page, 0,
1595 						   length);
1596 				/* re-use the current skb, we only consumed the
1597 				 * page
1598 				 */
1599 				buffer_info->skb = skb;
1600 				skb = rxtop;
1601 				rxtop = NULL;
1602 				e1000_consume_page(buffer_info, skb, length);
1603 			} else {
1604 				/* no chain, got EOP, this buf is the packet
1605 				 * copybreak to save the put_page/alloc_page
1606 				 */
1607 				if (length <= copybreak &&
1608 				    skb_tailroom(skb) >= length) {
1609 					u8 *vaddr;
1610 					vaddr = kmap_atomic(buffer_info->page);
1611 					memcpy(skb_tail_pointer(skb), vaddr,
1612 					       length);
1613 					kunmap_atomic(vaddr);
1614 					/* re-use the page, so don't erase
1615 					 * buffer_info->page
1616 					 */
1617 					skb_put(skb, length);
1618 				} else {
1619 					skb_fill_page_desc(skb, 0,
1620 							   buffer_info->page, 0,
1621 							   length);
1622 					e1000_consume_page(buffer_info, skb,
1623 							   length);
1624 				}
1625 			}
1626 		}
1627 
1628 		/* Receive Checksum Offload */
1629 		e1000_rx_checksum(adapter, staterr, skb);
1630 
1631 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1632 
1633 		/* probably a little skewed due to removing CRC */
1634 		total_rx_bytes += skb->len;
1635 		total_rx_packets++;
1636 
1637 		/* eth type trans needs skb->data to point to something */
1638 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1639 			e_err("pskb_may_pull failed.\n");
1640 			dev_kfree_skb_irq(skb);
1641 			goto next_desc;
1642 		}
1643 
1644 		e1000_receive_skb(adapter, netdev, skb, staterr,
1645 				  rx_desc->wb.upper.vlan);
1646 
1647 next_desc:
1648 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1649 
1650 		/* return some buffers to hardware, one at a time is too slow */
1651 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1652 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1653 					      GFP_ATOMIC);
1654 			cleaned_count = 0;
1655 		}
1656 
1657 		/* use prefetched values */
1658 		rx_desc = next_rxd;
1659 		buffer_info = next_buffer;
1660 
1661 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1662 	}
1663 	rx_ring->next_to_clean = i;
1664 
1665 	cleaned_count = e1000_desc_unused(rx_ring);
1666 	if (cleaned_count)
1667 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1668 
1669 	adapter->total_rx_bytes += total_rx_bytes;
1670 	adapter->total_rx_packets += total_rx_packets;
1671 	return cleaned;
1672 }
1673 
1674 /**
1675  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1676  * @rx_ring: Rx descriptor ring
1677  **/
e1000_clean_rx_ring(struct e1000_ring * rx_ring)1678 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1679 {
1680 	struct e1000_adapter *adapter = rx_ring->adapter;
1681 	struct e1000_buffer *buffer_info;
1682 	struct e1000_ps_page *ps_page;
1683 	struct pci_dev *pdev = adapter->pdev;
1684 	unsigned int i, j;
1685 
1686 	/* Free all the Rx ring sk_buffs */
1687 	for (i = 0; i < rx_ring->count; i++) {
1688 		buffer_info = &rx_ring->buffer_info[i];
1689 		if (buffer_info->dma) {
1690 			if (adapter->clean_rx == e1000_clean_rx_irq)
1691 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1692 						 adapter->rx_buffer_len,
1693 						 DMA_FROM_DEVICE);
1694 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1695 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1696 					       PAGE_SIZE, DMA_FROM_DEVICE);
1697 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1698 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1699 						 adapter->rx_ps_bsize0,
1700 						 DMA_FROM_DEVICE);
1701 			buffer_info->dma = 0;
1702 		}
1703 
1704 		if (buffer_info->page) {
1705 			put_page(buffer_info->page);
1706 			buffer_info->page = NULL;
1707 		}
1708 
1709 		if (buffer_info->skb) {
1710 			dev_kfree_skb(buffer_info->skb);
1711 			buffer_info->skb = NULL;
1712 		}
1713 
1714 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1715 			ps_page = &buffer_info->ps_pages[j];
1716 			if (!ps_page->page)
1717 				break;
1718 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1719 				       DMA_FROM_DEVICE);
1720 			ps_page->dma = 0;
1721 			put_page(ps_page->page);
1722 			ps_page->page = NULL;
1723 		}
1724 	}
1725 
1726 	/* there also may be some cached data from a chained receive */
1727 	if (rx_ring->rx_skb_top) {
1728 		dev_kfree_skb(rx_ring->rx_skb_top);
1729 		rx_ring->rx_skb_top = NULL;
1730 	}
1731 
1732 	/* Zero out the descriptor ring */
1733 	memset(rx_ring->desc, 0, rx_ring->size);
1734 
1735 	rx_ring->next_to_clean = 0;
1736 	rx_ring->next_to_use = 0;
1737 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1738 }
1739 
e1000e_downshift_workaround(struct work_struct * work)1740 static void e1000e_downshift_workaround(struct work_struct *work)
1741 {
1742 	struct e1000_adapter *adapter = container_of(work,
1743 						     struct e1000_adapter,
1744 						     downshift_task);
1745 
1746 	if (test_bit(__E1000_DOWN, &adapter->state))
1747 		return;
1748 
1749 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1750 }
1751 
1752 /**
1753  * e1000_intr_msi - Interrupt Handler
1754  * @irq: interrupt number
1755  * @data: pointer to a network interface device structure
1756  **/
e1000_intr_msi(int __always_unused irq,void * data)1757 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1758 {
1759 	struct net_device *netdev = data;
1760 	struct e1000_adapter *adapter = netdev_priv(netdev);
1761 	struct e1000_hw *hw = &adapter->hw;
1762 	u32 icr = er32(ICR);
1763 
1764 	/* read ICR disables interrupts using IAM */
1765 	if (icr & E1000_ICR_LSC) {
1766 		hw->mac.get_link_status = true;
1767 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1768 		 * disconnect (LSC) before accessing any PHY registers
1769 		 */
1770 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1771 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1772 			schedule_work(&adapter->downshift_task);
1773 
1774 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1775 		 * link down event; disable receives here in the ISR and reset
1776 		 * adapter in watchdog
1777 		 */
1778 		if (netif_carrier_ok(netdev) &&
1779 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1780 			/* disable receives */
1781 			u32 rctl = er32(RCTL);
1782 
1783 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1784 			adapter->flags |= FLAG_RESTART_NOW;
1785 		}
1786 		/* guard against interrupt when we're going down */
1787 		if (!test_bit(__E1000_DOWN, &adapter->state))
1788 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1789 	}
1790 
1791 	/* Reset on uncorrectable ECC error */
1792 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1793 		u32 pbeccsts = er32(PBECCSTS);
1794 
1795 		adapter->corr_errors +=
1796 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1797 		adapter->uncorr_errors +=
1798 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1799 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1800 
1801 		/* Do the reset outside of interrupt context */
1802 		schedule_work(&adapter->reset_task);
1803 
1804 		/* return immediately since reset is imminent */
1805 		return IRQ_HANDLED;
1806 	}
1807 
1808 	if (napi_schedule_prep(&adapter->napi)) {
1809 		adapter->total_tx_bytes = 0;
1810 		adapter->total_tx_packets = 0;
1811 		adapter->total_rx_bytes = 0;
1812 		adapter->total_rx_packets = 0;
1813 		__napi_schedule(&adapter->napi);
1814 	}
1815 
1816 	return IRQ_HANDLED;
1817 }
1818 
1819 /**
1820  * e1000_intr - Interrupt Handler
1821  * @irq: interrupt number
1822  * @data: pointer to a network interface device structure
1823  **/
e1000_intr(int __always_unused irq,void * data)1824 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1825 {
1826 	struct net_device *netdev = data;
1827 	struct e1000_adapter *adapter = netdev_priv(netdev);
1828 	struct e1000_hw *hw = &adapter->hw;
1829 	u32 rctl, icr = er32(ICR);
1830 
1831 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1832 		return IRQ_NONE;	/* Not our interrupt */
1833 
1834 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1835 	 * not set, then the adapter didn't send an interrupt
1836 	 */
1837 	if (!(icr & E1000_ICR_INT_ASSERTED))
1838 		return IRQ_NONE;
1839 
1840 	/* Interrupt Auto-Mask...upon reading ICR,
1841 	 * interrupts are masked.  No need for the
1842 	 * IMC write
1843 	 */
1844 
1845 	if (icr & E1000_ICR_LSC) {
1846 		hw->mac.get_link_status = true;
1847 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1848 		 * disconnect (LSC) before accessing any PHY registers
1849 		 */
1850 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1851 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1852 			schedule_work(&adapter->downshift_task);
1853 
1854 		/* 80003ES2LAN workaround--
1855 		 * For packet buffer work-around on link down event;
1856 		 * disable receives here in the ISR and
1857 		 * reset adapter in watchdog
1858 		 */
1859 		if (netif_carrier_ok(netdev) &&
1860 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1861 			/* disable receives */
1862 			rctl = er32(RCTL);
1863 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1864 			adapter->flags |= FLAG_RESTART_NOW;
1865 		}
1866 		/* guard against interrupt when we're going down */
1867 		if (!test_bit(__E1000_DOWN, &adapter->state))
1868 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1869 	}
1870 
1871 	/* Reset on uncorrectable ECC error */
1872 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1873 		u32 pbeccsts = er32(PBECCSTS);
1874 
1875 		adapter->corr_errors +=
1876 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1877 		adapter->uncorr_errors +=
1878 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1879 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1880 
1881 		/* Do the reset outside of interrupt context */
1882 		schedule_work(&adapter->reset_task);
1883 
1884 		/* return immediately since reset is imminent */
1885 		return IRQ_HANDLED;
1886 	}
1887 
1888 	if (napi_schedule_prep(&adapter->napi)) {
1889 		adapter->total_tx_bytes = 0;
1890 		adapter->total_tx_packets = 0;
1891 		adapter->total_rx_bytes = 0;
1892 		adapter->total_rx_packets = 0;
1893 		__napi_schedule(&adapter->napi);
1894 	}
1895 
1896 	return IRQ_HANDLED;
1897 }
1898 
e1000_msix_other(int __always_unused irq,void * data)1899 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1900 {
1901 	struct net_device *netdev = data;
1902 	struct e1000_adapter *adapter = netdev_priv(netdev);
1903 	struct e1000_hw *hw = &adapter->hw;
1904 	u32 icr = er32(ICR);
1905 
1906 	if (icr & adapter->eiac_mask)
1907 		ew32(ICS, (icr & adapter->eiac_mask));
1908 
1909 	if (icr & E1000_ICR_LSC) {
1910 		hw->mac.get_link_status = true;
1911 		/* guard against interrupt when we're going down */
1912 		if (!test_bit(__E1000_DOWN, &adapter->state))
1913 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1914 	}
1915 
1916 	if (!test_bit(__E1000_DOWN, &adapter->state))
1917 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1918 
1919 	return IRQ_HANDLED;
1920 }
1921 
e1000_intr_msix_tx(int __always_unused irq,void * data)1922 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1923 {
1924 	struct net_device *netdev = data;
1925 	struct e1000_adapter *adapter = netdev_priv(netdev);
1926 	struct e1000_hw *hw = &adapter->hw;
1927 	struct e1000_ring *tx_ring = adapter->tx_ring;
1928 
1929 	adapter->total_tx_bytes = 0;
1930 	adapter->total_tx_packets = 0;
1931 
1932 	if (!e1000_clean_tx_irq(tx_ring))
1933 		/* Ring was not completely cleaned, so fire another interrupt */
1934 		ew32(ICS, tx_ring->ims_val);
1935 
1936 	if (!test_bit(__E1000_DOWN, &adapter->state))
1937 		ew32(IMS, adapter->tx_ring->ims_val);
1938 
1939 	return IRQ_HANDLED;
1940 }
1941 
e1000_intr_msix_rx(int __always_unused irq,void * data)1942 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1943 {
1944 	struct net_device *netdev = data;
1945 	struct e1000_adapter *adapter = netdev_priv(netdev);
1946 	struct e1000_ring *rx_ring = adapter->rx_ring;
1947 
1948 	/* Write the ITR value calculated at the end of the
1949 	 * previous interrupt.
1950 	 */
1951 	if (rx_ring->set_itr) {
1952 		u32 itr = rx_ring->itr_val ?
1953 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1954 
1955 		writel(itr, rx_ring->itr_register);
1956 		rx_ring->set_itr = 0;
1957 	}
1958 
1959 	if (napi_schedule_prep(&adapter->napi)) {
1960 		adapter->total_rx_bytes = 0;
1961 		adapter->total_rx_packets = 0;
1962 		__napi_schedule(&adapter->napi);
1963 	}
1964 	return IRQ_HANDLED;
1965 }
1966 
1967 /**
1968  * e1000_configure_msix - Configure MSI-X hardware
1969  * @adapter: board private structure
1970  *
1971  * e1000_configure_msix sets up the hardware to properly
1972  * generate MSI-X interrupts.
1973  **/
e1000_configure_msix(struct e1000_adapter * adapter)1974 static void e1000_configure_msix(struct e1000_adapter *adapter)
1975 {
1976 	struct e1000_hw *hw = &adapter->hw;
1977 	struct e1000_ring *rx_ring = adapter->rx_ring;
1978 	struct e1000_ring *tx_ring = adapter->tx_ring;
1979 	int vector = 0;
1980 	u32 ctrl_ext, ivar = 0;
1981 
1982 	adapter->eiac_mask = 0;
1983 
1984 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1985 	if (hw->mac.type == e1000_82574) {
1986 		u32 rfctl = er32(RFCTL);
1987 
1988 		rfctl |= E1000_RFCTL_ACK_DIS;
1989 		ew32(RFCTL, rfctl);
1990 	}
1991 
1992 	/* Configure Rx vector */
1993 	rx_ring->ims_val = E1000_IMS_RXQ0;
1994 	adapter->eiac_mask |= rx_ring->ims_val;
1995 	if (rx_ring->itr_val)
1996 		writel(1000000000 / (rx_ring->itr_val * 256),
1997 		       rx_ring->itr_register);
1998 	else
1999 		writel(1, rx_ring->itr_register);
2000 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
2001 
2002 	/* Configure Tx vector */
2003 	tx_ring->ims_val = E1000_IMS_TXQ0;
2004 	vector++;
2005 	if (tx_ring->itr_val)
2006 		writel(1000000000 / (tx_ring->itr_val * 256),
2007 		       tx_ring->itr_register);
2008 	else
2009 		writel(1, tx_ring->itr_register);
2010 	adapter->eiac_mask |= tx_ring->ims_val;
2011 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2012 
2013 	/* set vector for Other Causes, e.g. link changes */
2014 	vector++;
2015 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2016 	if (rx_ring->itr_val)
2017 		writel(1000000000 / (rx_ring->itr_val * 256),
2018 		       hw->hw_addr + E1000_EITR_82574(vector));
2019 	else
2020 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2021 
2022 	/* Cause Tx interrupts on every write back */
2023 	ivar |= BIT(31);
2024 
2025 	ew32(IVAR, ivar);
2026 
2027 	/* enable MSI-X PBA support */
2028 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2029 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2030 	ew32(CTRL_EXT, ctrl_ext);
2031 	e1e_flush();
2032 }
2033 
e1000e_reset_interrupt_capability(struct e1000_adapter * adapter)2034 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2035 {
2036 	if (adapter->msix_entries) {
2037 		pci_disable_msix(adapter->pdev);
2038 		kfree(adapter->msix_entries);
2039 		adapter->msix_entries = NULL;
2040 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2041 		pci_disable_msi(adapter->pdev);
2042 		adapter->flags &= ~FLAG_MSI_ENABLED;
2043 	}
2044 }
2045 
2046 /**
2047  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2048  * @adapter: board private structure
2049  *
2050  * Attempt to configure interrupts using the best available
2051  * capabilities of the hardware and kernel.
2052  **/
e1000e_set_interrupt_capability(struct e1000_adapter * adapter)2053 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2054 {
2055 	int err;
2056 	int i;
2057 
2058 	switch (adapter->int_mode) {
2059 	case E1000E_INT_MODE_MSIX:
2060 		if (adapter->flags & FLAG_HAS_MSIX) {
2061 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2062 			adapter->msix_entries = kcalloc(adapter->num_vectors,
2063 							sizeof(struct
2064 							       msix_entry),
2065 							GFP_KERNEL);
2066 			if (adapter->msix_entries) {
2067 				struct e1000_adapter *a = adapter;
2068 
2069 				for (i = 0; i < adapter->num_vectors; i++)
2070 					adapter->msix_entries[i].entry = i;
2071 
2072 				err = pci_enable_msix_range(a->pdev,
2073 							    a->msix_entries,
2074 							    a->num_vectors,
2075 							    a->num_vectors);
2076 				if (err > 0)
2077 					return;
2078 			}
2079 			/* MSI-X failed, so fall through and try MSI */
2080 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2081 			e1000e_reset_interrupt_capability(adapter);
2082 		}
2083 		adapter->int_mode = E1000E_INT_MODE_MSI;
2084 		fallthrough;
2085 	case E1000E_INT_MODE_MSI:
2086 		if (!pci_enable_msi(adapter->pdev)) {
2087 			adapter->flags |= FLAG_MSI_ENABLED;
2088 		} else {
2089 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2090 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2091 		}
2092 		fallthrough;
2093 	case E1000E_INT_MODE_LEGACY:
2094 		/* Don't do anything; this is the system default */
2095 		break;
2096 	}
2097 
2098 	/* store the number of vectors being used */
2099 	adapter->num_vectors = 1;
2100 }
2101 
2102 /**
2103  * e1000_request_msix - Initialize MSI-X interrupts
2104  * @adapter: board private structure
2105  *
2106  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2107  * kernel.
2108  **/
e1000_request_msix(struct e1000_adapter * adapter)2109 static int e1000_request_msix(struct e1000_adapter *adapter)
2110 {
2111 	struct net_device *netdev = adapter->netdev;
2112 	int err = 0, vector = 0;
2113 
2114 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2115 		snprintf(adapter->rx_ring->name,
2116 			 sizeof(adapter->rx_ring->name) - 1,
2117 			 "%.14s-rx-0", netdev->name);
2118 	else
2119 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2120 	err = request_irq(adapter->msix_entries[vector].vector,
2121 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2122 			  netdev);
2123 	if (err)
2124 		return err;
2125 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2126 	    E1000_EITR_82574(vector);
2127 	adapter->rx_ring->itr_val = adapter->itr;
2128 	vector++;
2129 
2130 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2131 		snprintf(adapter->tx_ring->name,
2132 			 sizeof(adapter->tx_ring->name) - 1,
2133 			 "%.14s-tx-0", netdev->name);
2134 	else
2135 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2136 	err = request_irq(adapter->msix_entries[vector].vector,
2137 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2138 			  netdev);
2139 	if (err)
2140 		return err;
2141 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2142 	    E1000_EITR_82574(vector);
2143 	adapter->tx_ring->itr_val = adapter->itr;
2144 	vector++;
2145 
2146 	err = request_irq(adapter->msix_entries[vector].vector,
2147 			  e1000_msix_other, 0, netdev->name, netdev);
2148 	if (err)
2149 		return err;
2150 
2151 	e1000_configure_msix(adapter);
2152 
2153 	return 0;
2154 }
2155 
2156 /**
2157  * e1000_request_irq - initialize interrupts
2158  * @adapter: board private structure
2159  *
2160  * Attempts to configure interrupts using the best available
2161  * capabilities of the hardware and kernel.
2162  **/
e1000_request_irq(struct e1000_adapter * adapter)2163 static int e1000_request_irq(struct e1000_adapter *adapter)
2164 {
2165 	struct net_device *netdev = adapter->netdev;
2166 	int err;
2167 
2168 	if (adapter->msix_entries) {
2169 		err = e1000_request_msix(adapter);
2170 		if (!err)
2171 			return err;
2172 		/* fall back to MSI */
2173 		e1000e_reset_interrupt_capability(adapter);
2174 		adapter->int_mode = E1000E_INT_MODE_MSI;
2175 		e1000e_set_interrupt_capability(adapter);
2176 	}
2177 	if (adapter->flags & FLAG_MSI_ENABLED) {
2178 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2179 				  netdev->name, netdev);
2180 		if (!err)
2181 			return err;
2182 
2183 		/* fall back to legacy interrupt */
2184 		e1000e_reset_interrupt_capability(adapter);
2185 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2186 	}
2187 
2188 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2189 			  netdev->name, netdev);
2190 	if (err)
2191 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2192 
2193 	return err;
2194 }
2195 
e1000_free_irq(struct e1000_adapter * adapter)2196 static void e1000_free_irq(struct e1000_adapter *adapter)
2197 {
2198 	struct net_device *netdev = adapter->netdev;
2199 
2200 	if (adapter->msix_entries) {
2201 		int vector = 0;
2202 
2203 		free_irq(adapter->msix_entries[vector].vector, netdev);
2204 		vector++;
2205 
2206 		free_irq(adapter->msix_entries[vector].vector, netdev);
2207 		vector++;
2208 
2209 		/* Other Causes interrupt vector */
2210 		free_irq(adapter->msix_entries[vector].vector, netdev);
2211 		return;
2212 	}
2213 
2214 	free_irq(adapter->pdev->irq, netdev);
2215 }
2216 
2217 /**
2218  * e1000_irq_disable - Mask off interrupt generation on the NIC
2219  * @adapter: board private structure
2220  **/
e1000_irq_disable(struct e1000_adapter * adapter)2221 static void e1000_irq_disable(struct e1000_adapter *adapter)
2222 {
2223 	struct e1000_hw *hw = &adapter->hw;
2224 
2225 	ew32(IMC, ~0);
2226 	if (adapter->msix_entries)
2227 		ew32(EIAC_82574, 0);
2228 	e1e_flush();
2229 
2230 	if (adapter->msix_entries) {
2231 		int i;
2232 
2233 		for (i = 0; i < adapter->num_vectors; i++)
2234 			synchronize_irq(adapter->msix_entries[i].vector);
2235 	} else {
2236 		synchronize_irq(adapter->pdev->irq);
2237 	}
2238 }
2239 
2240 /**
2241  * e1000_irq_enable - Enable default interrupt generation settings
2242  * @adapter: board private structure
2243  **/
e1000_irq_enable(struct e1000_adapter * adapter)2244 static void e1000_irq_enable(struct e1000_adapter *adapter)
2245 {
2246 	struct e1000_hw *hw = &adapter->hw;
2247 
2248 	if (adapter->msix_entries) {
2249 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2250 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2251 		     IMS_OTHER_MASK);
2252 	} else if (hw->mac.type >= e1000_pch_lpt) {
2253 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2254 	} else {
2255 		ew32(IMS, IMS_ENABLE_MASK);
2256 	}
2257 	e1e_flush();
2258 }
2259 
2260 /**
2261  * e1000e_get_hw_control - get control of the h/w from f/w
2262  * @adapter: address of board private structure
2263  *
2264  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2265  * For ASF and Pass Through versions of f/w this means that
2266  * the driver is loaded. For AMT version (only with 82573)
2267  * of the f/w this means that the network i/f is open.
2268  **/
e1000e_get_hw_control(struct e1000_adapter * adapter)2269 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2270 {
2271 	struct e1000_hw *hw = &adapter->hw;
2272 	u32 ctrl_ext;
2273 	u32 swsm;
2274 
2275 	/* Let firmware know the driver has taken over */
2276 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2277 		swsm = er32(SWSM);
2278 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2279 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2280 		ctrl_ext = er32(CTRL_EXT);
2281 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2282 	}
2283 }
2284 
2285 /**
2286  * e1000e_release_hw_control - release control of the h/w to f/w
2287  * @adapter: address of board private structure
2288  *
2289  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2290  * For ASF and Pass Through versions of f/w this means that the
2291  * driver is no longer loaded. For AMT version (only with 82573) i
2292  * of the f/w this means that the network i/f is closed.
2293  *
2294  **/
e1000e_release_hw_control(struct e1000_adapter * adapter)2295 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2296 {
2297 	struct e1000_hw *hw = &adapter->hw;
2298 	u32 ctrl_ext;
2299 	u32 swsm;
2300 
2301 	/* Let firmware taken over control of h/w */
2302 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2303 		swsm = er32(SWSM);
2304 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2305 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2306 		ctrl_ext = er32(CTRL_EXT);
2307 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2308 	}
2309 }
2310 
2311 /**
2312  * e1000_alloc_ring_dma - allocate memory for a ring structure
2313  * @adapter: board private structure
2314  * @ring: ring struct for which to allocate dma
2315  **/
e1000_alloc_ring_dma(struct e1000_adapter * adapter,struct e1000_ring * ring)2316 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2317 				struct e1000_ring *ring)
2318 {
2319 	struct pci_dev *pdev = adapter->pdev;
2320 
2321 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2322 					GFP_KERNEL);
2323 	if (!ring->desc)
2324 		return -ENOMEM;
2325 
2326 	return 0;
2327 }
2328 
2329 /**
2330  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2331  * @tx_ring: Tx descriptor ring
2332  *
2333  * Return 0 on success, negative on failure
2334  **/
e1000e_setup_tx_resources(struct e1000_ring * tx_ring)2335 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2336 {
2337 	struct e1000_adapter *adapter = tx_ring->adapter;
2338 	int err = -ENOMEM, size;
2339 
2340 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2341 	tx_ring->buffer_info = vzalloc(size);
2342 	if (!tx_ring->buffer_info)
2343 		goto err;
2344 
2345 	/* round up to nearest 4K */
2346 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2347 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2348 
2349 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2350 	if (err)
2351 		goto err;
2352 
2353 	tx_ring->next_to_use = 0;
2354 	tx_ring->next_to_clean = 0;
2355 
2356 	return 0;
2357 err:
2358 	vfree(tx_ring->buffer_info);
2359 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2360 	return err;
2361 }
2362 
2363 /**
2364  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2365  * @rx_ring: Rx descriptor ring
2366  *
2367  * Returns 0 on success, negative on failure
2368  **/
e1000e_setup_rx_resources(struct e1000_ring * rx_ring)2369 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2370 {
2371 	struct e1000_adapter *adapter = rx_ring->adapter;
2372 	struct e1000_buffer *buffer_info;
2373 	int i, size, desc_len, err = -ENOMEM;
2374 
2375 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2376 	rx_ring->buffer_info = vzalloc(size);
2377 	if (!rx_ring->buffer_info)
2378 		goto err;
2379 
2380 	for (i = 0; i < rx_ring->count; i++) {
2381 		buffer_info = &rx_ring->buffer_info[i];
2382 		buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2383 						sizeof(struct e1000_ps_page),
2384 						GFP_KERNEL);
2385 		if (!buffer_info->ps_pages)
2386 			goto err_pages;
2387 	}
2388 
2389 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2390 
2391 	/* Round up to nearest 4K */
2392 	rx_ring->size = rx_ring->count * desc_len;
2393 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2394 
2395 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2396 	if (err)
2397 		goto err_pages;
2398 
2399 	rx_ring->next_to_clean = 0;
2400 	rx_ring->next_to_use = 0;
2401 	rx_ring->rx_skb_top = NULL;
2402 
2403 	return 0;
2404 
2405 err_pages:
2406 	for (i = 0; i < rx_ring->count; i++) {
2407 		buffer_info = &rx_ring->buffer_info[i];
2408 		kfree(buffer_info->ps_pages);
2409 	}
2410 err:
2411 	vfree(rx_ring->buffer_info);
2412 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2413 	return err;
2414 }
2415 
2416 /**
2417  * e1000_clean_tx_ring - Free Tx Buffers
2418  * @tx_ring: Tx descriptor ring
2419  **/
e1000_clean_tx_ring(struct e1000_ring * tx_ring)2420 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2421 {
2422 	struct e1000_adapter *adapter = tx_ring->adapter;
2423 	struct e1000_buffer *buffer_info;
2424 	unsigned long size;
2425 	unsigned int i;
2426 
2427 	for (i = 0; i < tx_ring->count; i++) {
2428 		buffer_info = &tx_ring->buffer_info[i];
2429 		e1000_put_txbuf(tx_ring, buffer_info, false);
2430 	}
2431 
2432 	netdev_reset_queue(adapter->netdev);
2433 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2434 	memset(tx_ring->buffer_info, 0, size);
2435 
2436 	memset(tx_ring->desc, 0, tx_ring->size);
2437 
2438 	tx_ring->next_to_use = 0;
2439 	tx_ring->next_to_clean = 0;
2440 }
2441 
2442 /**
2443  * e1000e_free_tx_resources - Free Tx Resources per Queue
2444  * @tx_ring: Tx descriptor ring
2445  *
2446  * Free all transmit software resources
2447  **/
e1000e_free_tx_resources(struct e1000_ring * tx_ring)2448 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2449 {
2450 	struct e1000_adapter *adapter = tx_ring->adapter;
2451 	struct pci_dev *pdev = adapter->pdev;
2452 
2453 	e1000_clean_tx_ring(tx_ring);
2454 
2455 	vfree(tx_ring->buffer_info);
2456 	tx_ring->buffer_info = NULL;
2457 
2458 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2459 			  tx_ring->dma);
2460 	tx_ring->desc = NULL;
2461 }
2462 
2463 /**
2464  * e1000e_free_rx_resources - Free Rx Resources
2465  * @rx_ring: Rx descriptor ring
2466  *
2467  * Free all receive software resources
2468  **/
e1000e_free_rx_resources(struct e1000_ring * rx_ring)2469 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2470 {
2471 	struct e1000_adapter *adapter = rx_ring->adapter;
2472 	struct pci_dev *pdev = adapter->pdev;
2473 	int i;
2474 
2475 	e1000_clean_rx_ring(rx_ring);
2476 
2477 	for (i = 0; i < rx_ring->count; i++)
2478 		kfree(rx_ring->buffer_info[i].ps_pages);
2479 
2480 	vfree(rx_ring->buffer_info);
2481 	rx_ring->buffer_info = NULL;
2482 
2483 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2484 			  rx_ring->dma);
2485 	rx_ring->desc = NULL;
2486 }
2487 
2488 /**
2489  * e1000_update_itr - update the dynamic ITR value based on statistics
2490  * @itr_setting: current adapter->itr
2491  * @packets: the number of packets during this measurement interval
2492  * @bytes: the number of bytes during this measurement interval
2493  *
2494  *      Stores a new ITR value based on packets and byte
2495  *      counts during the last interrupt.  The advantage of per interrupt
2496  *      computation is faster updates and more accurate ITR for the current
2497  *      traffic pattern.  Constants in this function were computed
2498  *      based on theoretical maximum wire speed and thresholds were set based
2499  *      on testing data as well as attempting to minimize response time
2500  *      while increasing bulk throughput.  This functionality is controlled
2501  *      by the InterruptThrottleRate module parameter.
2502  **/
e1000_update_itr(u16 itr_setting,int packets,int bytes)2503 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2504 {
2505 	unsigned int retval = itr_setting;
2506 
2507 	if (packets == 0)
2508 		return itr_setting;
2509 
2510 	switch (itr_setting) {
2511 	case lowest_latency:
2512 		/* handle TSO and jumbo frames */
2513 		if (bytes / packets > 8000)
2514 			retval = bulk_latency;
2515 		else if ((packets < 5) && (bytes > 512))
2516 			retval = low_latency;
2517 		break;
2518 	case low_latency:	/* 50 usec aka 20000 ints/s */
2519 		if (bytes > 10000) {
2520 			/* this if handles the TSO accounting */
2521 			if (bytes / packets > 8000)
2522 				retval = bulk_latency;
2523 			else if ((packets < 10) || ((bytes / packets) > 1200))
2524 				retval = bulk_latency;
2525 			else if ((packets > 35))
2526 				retval = lowest_latency;
2527 		} else if (bytes / packets > 2000) {
2528 			retval = bulk_latency;
2529 		} else if (packets <= 2 && bytes < 512) {
2530 			retval = lowest_latency;
2531 		}
2532 		break;
2533 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2534 		if (bytes > 25000) {
2535 			if (packets > 35)
2536 				retval = low_latency;
2537 		} else if (bytes < 6000) {
2538 			retval = low_latency;
2539 		}
2540 		break;
2541 	}
2542 
2543 	return retval;
2544 }
2545 
e1000_set_itr(struct e1000_adapter * adapter)2546 static void e1000_set_itr(struct e1000_adapter *adapter)
2547 {
2548 	u16 current_itr;
2549 	u32 new_itr = adapter->itr;
2550 
2551 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2552 	if (adapter->link_speed != SPEED_1000) {
2553 		current_itr = 0;
2554 		new_itr = 4000;
2555 		goto set_itr_now;
2556 	}
2557 
2558 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2559 		new_itr = 0;
2560 		goto set_itr_now;
2561 	}
2562 
2563 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2564 					   adapter->total_tx_packets,
2565 					   adapter->total_tx_bytes);
2566 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2567 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2568 		adapter->tx_itr = low_latency;
2569 
2570 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2571 					   adapter->total_rx_packets,
2572 					   adapter->total_rx_bytes);
2573 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2574 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2575 		adapter->rx_itr = low_latency;
2576 
2577 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2578 
2579 	/* counts and packets in update_itr are dependent on these numbers */
2580 	switch (current_itr) {
2581 	case lowest_latency:
2582 		new_itr = 70000;
2583 		break;
2584 	case low_latency:
2585 		new_itr = 20000;	/* aka hwitr = ~200 */
2586 		break;
2587 	case bulk_latency:
2588 		new_itr = 4000;
2589 		break;
2590 	default:
2591 		break;
2592 	}
2593 
2594 set_itr_now:
2595 	if (new_itr != adapter->itr) {
2596 		/* this attempts to bias the interrupt rate towards Bulk
2597 		 * by adding intermediate steps when interrupt rate is
2598 		 * increasing
2599 		 */
2600 		new_itr = new_itr > adapter->itr ?
2601 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2602 		adapter->itr = new_itr;
2603 		adapter->rx_ring->itr_val = new_itr;
2604 		if (adapter->msix_entries)
2605 			adapter->rx_ring->set_itr = 1;
2606 		else
2607 			e1000e_write_itr(adapter, new_itr);
2608 	}
2609 }
2610 
2611 /**
2612  * e1000e_write_itr - write the ITR value to the appropriate registers
2613  * @adapter: address of board private structure
2614  * @itr: new ITR value to program
2615  *
2616  * e1000e_write_itr determines if the adapter is in MSI-X mode
2617  * and, if so, writes the EITR registers with the ITR value.
2618  * Otherwise, it writes the ITR value into the ITR register.
2619  **/
e1000e_write_itr(struct e1000_adapter * adapter,u32 itr)2620 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2621 {
2622 	struct e1000_hw *hw = &adapter->hw;
2623 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2624 
2625 	if (adapter->msix_entries) {
2626 		int vector;
2627 
2628 		for (vector = 0; vector < adapter->num_vectors; vector++)
2629 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2630 	} else {
2631 		ew32(ITR, new_itr);
2632 	}
2633 }
2634 
2635 /**
2636  * e1000_alloc_queues - Allocate memory for all rings
2637  * @adapter: board private structure to initialize
2638  **/
e1000_alloc_queues(struct e1000_adapter * adapter)2639 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2640 {
2641 	int size = sizeof(struct e1000_ring);
2642 
2643 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2644 	if (!adapter->tx_ring)
2645 		goto err;
2646 	adapter->tx_ring->count = adapter->tx_ring_count;
2647 	adapter->tx_ring->adapter = adapter;
2648 
2649 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2650 	if (!adapter->rx_ring)
2651 		goto err;
2652 	adapter->rx_ring->count = adapter->rx_ring_count;
2653 	adapter->rx_ring->adapter = adapter;
2654 
2655 	return 0;
2656 err:
2657 	e_err("Unable to allocate memory for queues\n");
2658 	kfree(adapter->rx_ring);
2659 	kfree(adapter->tx_ring);
2660 	return -ENOMEM;
2661 }
2662 
2663 /**
2664  * e1000e_poll - NAPI Rx polling callback
2665  * @napi: struct associated with this polling callback
2666  * @budget: number of packets driver is allowed to process this poll
2667  **/
e1000e_poll(struct napi_struct * napi,int budget)2668 static int e1000e_poll(struct napi_struct *napi, int budget)
2669 {
2670 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2671 						     napi);
2672 	struct e1000_hw *hw = &adapter->hw;
2673 	struct net_device *poll_dev = adapter->netdev;
2674 	int tx_cleaned = 1, work_done = 0;
2675 
2676 	adapter = netdev_priv(poll_dev);
2677 
2678 	if (!adapter->msix_entries ||
2679 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2680 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2681 
2682 	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2683 
2684 	if (!tx_cleaned || work_done == budget)
2685 		return budget;
2686 
2687 	/* Exit the polling mode, but don't re-enable interrupts if stack might
2688 	 * poll us due to busy-polling
2689 	 */
2690 	if (likely(napi_complete_done(napi, work_done))) {
2691 		if (adapter->itr_setting & 3)
2692 			e1000_set_itr(adapter);
2693 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2694 			if (adapter->msix_entries)
2695 				ew32(IMS, adapter->rx_ring->ims_val);
2696 			else
2697 				e1000_irq_enable(adapter);
2698 		}
2699 	}
2700 
2701 	return work_done;
2702 }
2703 
e1000_vlan_rx_add_vid(struct net_device * netdev,__always_unused __be16 proto,u16 vid)2704 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2705 				 __always_unused __be16 proto, u16 vid)
2706 {
2707 	struct e1000_adapter *adapter = netdev_priv(netdev);
2708 	struct e1000_hw *hw = &adapter->hw;
2709 	u32 vfta, index;
2710 
2711 	/* don't update vlan cookie if already programmed */
2712 	if ((adapter->hw.mng_cookie.status &
2713 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2714 	    (vid == adapter->mng_vlan_id))
2715 		return 0;
2716 
2717 	/* add VID to filter table */
2718 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2719 		index = (vid >> 5) & 0x7F;
2720 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2721 		vfta |= BIT((vid & 0x1F));
2722 		hw->mac.ops.write_vfta(hw, index, vfta);
2723 	}
2724 
2725 	set_bit(vid, adapter->active_vlans);
2726 
2727 	return 0;
2728 }
2729 
e1000_vlan_rx_kill_vid(struct net_device * netdev,__always_unused __be16 proto,u16 vid)2730 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2731 				  __always_unused __be16 proto, u16 vid)
2732 {
2733 	struct e1000_adapter *adapter = netdev_priv(netdev);
2734 	struct e1000_hw *hw = &adapter->hw;
2735 	u32 vfta, index;
2736 
2737 	if ((adapter->hw.mng_cookie.status &
2738 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2739 	    (vid == adapter->mng_vlan_id)) {
2740 		/* release control to f/w */
2741 		e1000e_release_hw_control(adapter);
2742 		return 0;
2743 	}
2744 
2745 	/* remove VID from filter table */
2746 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2747 		index = (vid >> 5) & 0x7F;
2748 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2749 		vfta &= ~BIT((vid & 0x1F));
2750 		hw->mac.ops.write_vfta(hw, index, vfta);
2751 	}
2752 
2753 	clear_bit(vid, adapter->active_vlans);
2754 
2755 	return 0;
2756 }
2757 
2758 /**
2759  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2760  * @adapter: board private structure to initialize
2761  **/
e1000e_vlan_filter_disable(struct e1000_adapter * adapter)2762 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2763 {
2764 	struct net_device *netdev = adapter->netdev;
2765 	struct e1000_hw *hw = &adapter->hw;
2766 	u32 rctl;
2767 
2768 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2769 		/* disable VLAN receive filtering */
2770 		rctl = er32(RCTL);
2771 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2772 		ew32(RCTL, rctl);
2773 
2774 		if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2775 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2776 					       adapter->mng_vlan_id);
2777 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2778 		}
2779 	}
2780 }
2781 
2782 /**
2783  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2784  * @adapter: board private structure to initialize
2785  **/
e1000e_vlan_filter_enable(struct e1000_adapter * adapter)2786 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2787 {
2788 	struct e1000_hw *hw = &adapter->hw;
2789 	u32 rctl;
2790 
2791 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2792 		/* enable VLAN receive filtering */
2793 		rctl = er32(RCTL);
2794 		rctl |= E1000_RCTL_VFE;
2795 		rctl &= ~E1000_RCTL_CFIEN;
2796 		ew32(RCTL, rctl);
2797 	}
2798 }
2799 
2800 /**
2801  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2802  * @adapter: board private structure to initialize
2803  **/
e1000e_vlan_strip_disable(struct e1000_adapter * adapter)2804 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2805 {
2806 	struct e1000_hw *hw = &adapter->hw;
2807 	u32 ctrl;
2808 
2809 	/* disable VLAN tag insert/strip */
2810 	ctrl = er32(CTRL);
2811 	ctrl &= ~E1000_CTRL_VME;
2812 	ew32(CTRL, ctrl);
2813 }
2814 
2815 /**
2816  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2817  * @adapter: board private structure to initialize
2818  **/
e1000e_vlan_strip_enable(struct e1000_adapter * adapter)2819 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2820 {
2821 	struct e1000_hw *hw = &adapter->hw;
2822 	u32 ctrl;
2823 
2824 	/* enable VLAN tag insert/strip */
2825 	ctrl = er32(CTRL);
2826 	ctrl |= E1000_CTRL_VME;
2827 	ew32(CTRL, ctrl);
2828 }
2829 
e1000_update_mng_vlan(struct e1000_adapter * adapter)2830 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2831 {
2832 	struct net_device *netdev = adapter->netdev;
2833 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2834 	u16 old_vid = adapter->mng_vlan_id;
2835 
2836 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2837 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2838 		adapter->mng_vlan_id = vid;
2839 	}
2840 
2841 	if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2842 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2843 }
2844 
e1000_restore_vlan(struct e1000_adapter * adapter)2845 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2846 {
2847 	u16 vid;
2848 
2849 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2850 
2851 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2852 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2853 }
2854 
e1000_init_manageability_pt(struct e1000_adapter * adapter)2855 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2856 {
2857 	struct e1000_hw *hw = &adapter->hw;
2858 	u32 manc, manc2h, mdef, i, j;
2859 
2860 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2861 		return;
2862 
2863 	manc = er32(MANC);
2864 
2865 	/* enable receiving management packets to the host. this will probably
2866 	 * generate destination unreachable messages from the host OS, but
2867 	 * the packets will be handled on SMBUS
2868 	 */
2869 	manc |= E1000_MANC_EN_MNG2HOST;
2870 	manc2h = er32(MANC2H);
2871 
2872 	switch (hw->mac.type) {
2873 	default:
2874 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2875 		break;
2876 	case e1000_82574:
2877 	case e1000_82583:
2878 		/* Check if IPMI pass-through decision filter already exists;
2879 		 * if so, enable it.
2880 		 */
2881 		for (i = 0, j = 0; i < 8; i++) {
2882 			mdef = er32(MDEF(i));
2883 
2884 			/* Ignore filters with anything other than IPMI ports */
2885 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2886 				continue;
2887 
2888 			/* Enable this decision filter in MANC2H */
2889 			if (mdef)
2890 				manc2h |= BIT(i);
2891 
2892 			j |= mdef;
2893 		}
2894 
2895 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2896 			break;
2897 
2898 		/* Create new decision filter in an empty filter */
2899 		for (i = 0, j = 0; i < 8; i++)
2900 			if (er32(MDEF(i)) == 0) {
2901 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2902 					       E1000_MDEF_PORT_664));
2903 				manc2h |= BIT(1);
2904 				j++;
2905 				break;
2906 			}
2907 
2908 		if (!j)
2909 			e_warn("Unable to create IPMI pass-through filter\n");
2910 		break;
2911 	}
2912 
2913 	ew32(MANC2H, manc2h);
2914 	ew32(MANC, manc);
2915 }
2916 
2917 /**
2918  * e1000_configure_tx - Configure Transmit Unit after Reset
2919  * @adapter: board private structure
2920  *
2921  * Configure the Tx unit of the MAC after a reset.
2922  **/
e1000_configure_tx(struct e1000_adapter * adapter)2923 static void e1000_configure_tx(struct e1000_adapter *adapter)
2924 {
2925 	struct e1000_hw *hw = &adapter->hw;
2926 	struct e1000_ring *tx_ring = adapter->tx_ring;
2927 	u64 tdba;
2928 	u32 tdlen, tctl, tarc;
2929 
2930 	/* Setup the HW Tx Head and Tail descriptor pointers */
2931 	tdba = tx_ring->dma;
2932 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2933 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2934 	ew32(TDBAH(0), (tdba >> 32));
2935 	ew32(TDLEN(0), tdlen);
2936 	ew32(TDH(0), 0);
2937 	ew32(TDT(0), 0);
2938 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2939 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2940 
2941 	writel(0, tx_ring->head);
2942 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2943 		e1000e_update_tdt_wa(tx_ring, 0);
2944 	else
2945 		writel(0, tx_ring->tail);
2946 
2947 	/* Set the Tx Interrupt Delay register */
2948 	ew32(TIDV, adapter->tx_int_delay);
2949 	/* Tx irq moderation */
2950 	ew32(TADV, adapter->tx_abs_int_delay);
2951 
2952 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2953 		u32 txdctl = er32(TXDCTL(0));
2954 
2955 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2956 			    E1000_TXDCTL_WTHRESH);
2957 		/* set up some performance related parameters to encourage the
2958 		 * hardware to use the bus more efficiently in bursts, depends
2959 		 * on the tx_int_delay to be enabled,
2960 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2961 		 * hthresh = 1 ==> prefetch when one or more available
2962 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2963 		 * BEWARE: this seems to work but should be considered first if
2964 		 * there are Tx hangs or other Tx related bugs
2965 		 */
2966 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2967 		ew32(TXDCTL(0), txdctl);
2968 	}
2969 	/* erratum work around: set txdctl the same for both queues */
2970 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2971 
2972 	/* Program the Transmit Control Register */
2973 	tctl = er32(TCTL);
2974 	tctl &= ~E1000_TCTL_CT;
2975 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2976 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2977 
2978 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2979 		tarc = er32(TARC(0));
2980 		/* set the speed mode bit, we'll clear it if we're not at
2981 		 * gigabit link later
2982 		 */
2983 #define SPEED_MODE_BIT BIT(21)
2984 		tarc |= SPEED_MODE_BIT;
2985 		ew32(TARC(0), tarc);
2986 	}
2987 
2988 	/* errata: program both queues to unweighted RR */
2989 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2990 		tarc = er32(TARC(0));
2991 		tarc |= 1;
2992 		ew32(TARC(0), tarc);
2993 		tarc = er32(TARC(1));
2994 		tarc |= 1;
2995 		ew32(TARC(1), tarc);
2996 	}
2997 
2998 	/* Setup Transmit Descriptor Settings for eop descriptor */
2999 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
3000 
3001 	/* only set IDE if we are delaying interrupts using the timers */
3002 	if (adapter->tx_int_delay)
3003 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3004 
3005 	/* enable Report Status bit */
3006 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
3007 
3008 	ew32(TCTL, tctl);
3009 
3010 	hw->mac.ops.config_collision_dist(hw);
3011 
3012 	/* SPT and KBL Si errata workaround to avoid data corruption */
3013 	if (hw->mac.type == e1000_pch_spt) {
3014 		u32 reg_val;
3015 
3016 		reg_val = er32(IOSFPC);
3017 		reg_val |= E1000_RCTL_RDMTS_HEX;
3018 		ew32(IOSFPC, reg_val);
3019 
3020 		reg_val = er32(TARC(0));
3021 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3022 		 * Dropping the number of outstanding requests from
3023 		 * 3 to 2 in order to avoid a buffer overrun.
3024 		 */
3025 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3026 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3027 		ew32(TARC(0), reg_val);
3028 	}
3029 }
3030 
3031 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3032 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3033 
3034 /**
3035  * e1000_setup_rctl - configure the receive control registers
3036  * @adapter: Board private structure
3037  **/
e1000_setup_rctl(struct e1000_adapter * adapter)3038 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3039 {
3040 	struct e1000_hw *hw = &adapter->hw;
3041 	u32 rctl, rfctl;
3042 	u32 pages = 0;
3043 
3044 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3045 	 * If jumbo frames not set, program related MAC/PHY registers
3046 	 * to h/w defaults
3047 	 */
3048 	if (hw->mac.type >= e1000_pch2lan) {
3049 		s32 ret_val;
3050 
3051 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3052 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3053 		else
3054 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3055 
3056 		if (ret_val)
3057 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3058 	}
3059 
3060 	/* Program MC offset vector base */
3061 	rctl = er32(RCTL);
3062 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3063 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3064 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3065 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3066 
3067 	/* Do not Store bad packets */
3068 	rctl &= ~E1000_RCTL_SBP;
3069 
3070 	/* Enable Long Packet receive */
3071 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3072 		rctl &= ~E1000_RCTL_LPE;
3073 	else
3074 		rctl |= E1000_RCTL_LPE;
3075 
3076 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3077 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3078 	 * host memory when this is enabled
3079 	 */
3080 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3081 		rctl |= E1000_RCTL_SECRC;
3082 
3083 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3084 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3085 		u16 phy_data;
3086 
3087 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3088 		phy_data &= 0xfff8;
3089 		phy_data |= BIT(2);
3090 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3091 
3092 		e1e_rphy(hw, 22, &phy_data);
3093 		phy_data &= 0x0fff;
3094 		phy_data |= BIT(14);
3095 		e1e_wphy(hw, 0x10, 0x2823);
3096 		e1e_wphy(hw, 0x11, 0x0003);
3097 		e1e_wphy(hw, 22, phy_data);
3098 	}
3099 
3100 	/* Setup buffer sizes */
3101 	rctl &= ~E1000_RCTL_SZ_4096;
3102 	rctl |= E1000_RCTL_BSEX;
3103 	switch (adapter->rx_buffer_len) {
3104 	case 2048:
3105 	default:
3106 		rctl |= E1000_RCTL_SZ_2048;
3107 		rctl &= ~E1000_RCTL_BSEX;
3108 		break;
3109 	case 4096:
3110 		rctl |= E1000_RCTL_SZ_4096;
3111 		break;
3112 	case 8192:
3113 		rctl |= E1000_RCTL_SZ_8192;
3114 		break;
3115 	case 16384:
3116 		rctl |= E1000_RCTL_SZ_16384;
3117 		break;
3118 	}
3119 
3120 	/* Enable Extended Status in all Receive Descriptors */
3121 	rfctl = er32(RFCTL);
3122 	rfctl |= E1000_RFCTL_EXTEN;
3123 	ew32(RFCTL, rfctl);
3124 
3125 	/* 82571 and greater support packet-split where the protocol
3126 	 * header is placed in skb->data and the packet data is
3127 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3128 	 * In the case of a non-split, skb->data is linearly filled,
3129 	 * followed by the page buffers.  Therefore, skb->data is
3130 	 * sized to hold the largest protocol header.
3131 	 *
3132 	 * allocations using alloc_page take too long for regular MTU
3133 	 * so only enable packet split for jumbo frames
3134 	 *
3135 	 * Using pages when the page size is greater than 16k wastes
3136 	 * a lot of memory, since we allocate 3 pages at all times
3137 	 * per packet.
3138 	 */
3139 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3140 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3141 		adapter->rx_ps_pages = pages;
3142 	else
3143 		adapter->rx_ps_pages = 0;
3144 
3145 	if (adapter->rx_ps_pages) {
3146 		u32 psrctl = 0;
3147 
3148 		/* Enable Packet split descriptors */
3149 		rctl |= E1000_RCTL_DTYP_PS;
3150 
3151 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3152 
3153 		switch (adapter->rx_ps_pages) {
3154 		case 3:
3155 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3156 			fallthrough;
3157 		case 2:
3158 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3159 			fallthrough;
3160 		case 1:
3161 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3162 			break;
3163 		}
3164 
3165 		ew32(PSRCTL, psrctl);
3166 	}
3167 
3168 	/* This is useful for sniffing bad packets. */
3169 	if (adapter->netdev->features & NETIF_F_RXALL) {
3170 		/* UPE and MPE will be handled by normal PROMISC logic
3171 		 * in e1000e_set_rx_mode
3172 		 */
3173 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3174 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3175 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3176 
3177 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3178 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3179 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3180 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3181 		 * and that breaks VLANs.
3182 		 */
3183 	}
3184 
3185 	ew32(RCTL, rctl);
3186 	/* just started the receive unit, no need to restart */
3187 	adapter->flags &= ~FLAG_RESTART_NOW;
3188 }
3189 
3190 /**
3191  * e1000_configure_rx - Configure Receive Unit after Reset
3192  * @adapter: board private structure
3193  *
3194  * Configure the Rx unit of the MAC after a reset.
3195  **/
e1000_configure_rx(struct e1000_adapter * adapter)3196 static void e1000_configure_rx(struct e1000_adapter *adapter)
3197 {
3198 	struct e1000_hw *hw = &adapter->hw;
3199 	struct e1000_ring *rx_ring = adapter->rx_ring;
3200 	u64 rdba;
3201 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3202 
3203 	if (adapter->rx_ps_pages) {
3204 		/* this is a 32 byte descriptor */
3205 		rdlen = rx_ring->count *
3206 		    sizeof(union e1000_rx_desc_packet_split);
3207 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3208 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3209 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3210 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3211 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3212 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3213 	} else {
3214 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3215 		adapter->clean_rx = e1000_clean_rx_irq;
3216 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3217 	}
3218 
3219 	/* disable receives while setting up the descriptors */
3220 	rctl = er32(RCTL);
3221 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3222 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3223 	e1e_flush();
3224 	usleep_range(10000, 11000);
3225 
3226 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3227 		/* set the writeback threshold (only takes effect if the RDTR
3228 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3229 		 * enable prefetching of 0x20 Rx descriptors
3230 		 * granularity = 01
3231 		 * wthresh = 04,
3232 		 * hthresh = 04,
3233 		 * pthresh = 0x20
3234 		 */
3235 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3236 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3237 	}
3238 
3239 	/* set the Receive Delay Timer Register */
3240 	ew32(RDTR, adapter->rx_int_delay);
3241 
3242 	/* irq moderation */
3243 	ew32(RADV, adapter->rx_abs_int_delay);
3244 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3245 		e1000e_write_itr(adapter, adapter->itr);
3246 
3247 	ctrl_ext = er32(CTRL_EXT);
3248 	/* Auto-Mask interrupts upon ICR access */
3249 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3250 	ew32(IAM, 0xffffffff);
3251 	ew32(CTRL_EXT, ctrl_ext);
3252 	e1e_flush();
3253 
3254 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3255 	 * the Base and Length of the Rx Descriptor Ring
3256 	 */
3257 	rdba = rx_ring->dma;
3258 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3259 	ew32(RDBAH(0), (rdba >> 32));
3260 	ew32(RDLEN(0), rdlen);
3261 	ew32(RDH(0), 0);
3262 	ew32(RDT(0), 0);
3263 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3264 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3265 
3266 	writel(0, rx_ring->head);
3267 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3268 		e1000e_update_rdt_wa(rx_ring, 0);
3269 	else
3270 		writel(0, rx_ring->tail);
3271 
3272 	/* Enable Receive Checksum Offload for TCP and UDP */
3273 	rxcsum = er32(RXCSUM);
3274 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3275 		rxcsum |= E1000_RXCSUM_TUOFL;
3276 	else
3277 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3278 	ew32(RXCSUM, rxcsum);
3279 
3280 	/* With jumbo frames, excessive C-state transition latencies result
3281 	 * in dropped transactions.
3282 	 */
3283 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3284 		u32 lat =
3285 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3286 		     adapter->max_frame_size) * 8 / 1000;
3287 
3288 		if (adapter->flags & FLAG_IS_ICH) {
3289 			u32 rxdctl = er32(RXDCTL(0));
3290 
3291 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3292 		}
3293 
3294 		dev_info(&adapter->pdev->dev,
3295 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3296 		cpu_latency_qos_update_request(&adapter->pm_qos_req, lat);
3297 	} else {
3298 		cpu_latency_qos_update_request(&adapter->pm_qos_req,
3299 					       PM_QOS_DEFAULT_VALUE);
3300 	}
3301 
3302 	/* Enable Receives */
3303 	ew32(RCTL, rctl);
3304 }
3305 
3306 /**
3307  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3308  * @netdev: network interface device structure
3309  *
3310  * Writes multicast address list to the MTA hash table.
3311  * Returns: -ENOMEM on failure
3312  *                0 on no addresses written
3313  *                X on writing X addresses to MTA
3314  */
e1000e_write_mc_addr_list(struct net_device * netdev)3315 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3316 {
3317 	struct e1000_adapter *adapter = netdev_priv(netdev);
3318 	struct e1000_hw *hw = &adapter->hw;
3319 	struct netdev_hw_addr *ha;
3320 	u8 *mta_list;
3321 	int i;
3322 
3323 	if (netdev_mc_empty(netdev)) {
3324 		/* nothing to program, so clear mc list */
3325 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3326 		return 0;
3327 	}
3328 
3329 	mta_list = kcalloc(netdev_mc_count(netdev), ETH_ALEN, GFP_ATOMIC);
3330 	if (!mta_list)
3331 		return -ENOMEM;
3332 
3333 	/* update_mc_addr_list expects a packed array of only addresses. */
3334 	i = 0;
3335 	netdev_for_each_mc_addr(ha, netdev)
3336 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3337 
3338 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3339 	kfree(mta_list);
3340 
3341 	return netdev_mc_count(netdev);
3342 }
3343 
3344 /**
3345  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3346  * @netdev: network interface device structure
3347  *
3348  * Writes unicast address list to the RAR table.
3349  * Returns: -ENOMEM on failure/insufficient address space
3350  *                0 on no addresses written
3351  *                X on writing X addresses to the RAR table
3352  **/
e1000e_write_uc_addr_list(struct net_device * netdev)3353 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3354 {
3355 	struct e1000_adapter *adapter = netdev_priv(netdev);
3356 	struct e1000_hw *hw = &adapter->hw;
3357 	unsigned int rar_entries;
3358 	int count = 0;
3359 
3360 	rar_entries = hw->mac.ops.rar_get_count(hw);
3361 
3362 	/* save a rar entry for our hardware address */
3363 	rar_entries--;
3364 
3365 	/* save a rar entry for the LAA workaround */
3366 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3367 		rar_entries--;
3368 
3369 	/* return ENOMEM indicating insufficient memory for addresses */
3370 	if (netdev_uc_count(netdev) > rar_entries)
3371 		return -ENOMEM;
3372 
3373 	if (!netdev_uc_empty(netdev) && rar_entries) {
3374 		struct netdev_hw_addr *ha;
3375 
3376 		/* write the addresses in reverse order to avoid write
3377 		 * combining
3378 		 */
3379 		netdev_for_each_uc_addr(ha, netdev) {
3380 			int ret_val;
3381 
3382 			if (!rar_entries)
3383 				break;
3384 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3385 			if (ret_val < 0)
3386 				return -ENOMEM;
3387 			count++;
3388 		}
3389 	}
3390 
3391 	/* zero out the remaining RAR entries not used above */
3392 	for (; rar_entries > 0; rar_entries--) {
3393 		ew32(RAH(rar_entries), 0);
3394 		ew32(RAL(rar_entries), 0);
3395 	}
3396 	e1e_flush();
3397 
3398 	return count;
3399 }
3400 
3401 /**
3402  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3403  * @netdev: network interface device structure
3404  *
3405  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3406  * address list or the network interface flags are updated.  This routine is
3407  * responsible for configuring the hardware for proper unicast, multicast,
3408  * promiscuous mode, and all-multi behavior.
3409  **/
e1000e_set_rx_mode(struct net_device * netdev)3410 static void e1000e_set_rx_mode(struct net_device *netdev)
3411 {
3412 	struct e1000_adapter *adapter = netdev_priv(netdev);
3413 	struct e1000_hw *hw = &adapter->hw;
3414 	u32 rctl;
3415 
3416 	if (pm_runtime_suspended(netdev->dev.parent))
3417 		return;
3418 
3419 	/* Check for Promiscuous and All Multicast modes */
3420 	rctl = er32(RCTL);
3421 
3422 	/* clear the affected bits */
3423 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3424 
3425 	if (netdev->flags & IFF_PROMISC) {
3426 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3427 		/* Do not hardware filter VLANs in promisc mode */
3428 		e1000e_vlan_filter_disable(adapter);
3429 	} else {
3430 		int count;
3431 
3432 		if (netdev->flags & IFF_ALLMULTI) {
3433 			rctl |= E1000_RCTL_MPE;
3434 		} else {
3435 			/* Write addresses to the MTA, if the attempt fails
3436 			 * then we should just turn on promiscuous mode so
3437 			 * that we can at least receive multicast traffic
3438 			 */
3439 			count = e1000e_write_mc_addr_list(netdev);
3440 			if (count < 0)
3441 				rctl |= E1000_RCTL_MPE;
3442 		}
3443 		e1000e_vlan_filter_enable(adapter);
3444 		/* Write addresses to available RAR registers, if there is not
3445 		 * sufficient space to store all the addresses then enable
3446 		 * unicast promiscuous mode
3447 		 */
3448 		count = e1000e_write_uc_addr_list(netdev);
3449 		if (count < 0)
3450 			rctl |= E1000_RCTL_UPE;
3451 	}
3452 
3453 	ew32(RCTL, rctl);
3454 
3455 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3456 		e1000e_vlan_strip_enable(adapter);
3457 	else
3458 		e1000e_vlan_strip_disable(adapter);
3459 }
3460 
e1000e_setup_rss_hash(struct e1000_adapter * adapter)3461 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3462 {
3463 	struct e1000_hw *hw = &adapter->hw;
3464 	u32 mrqc, rxcsum;
3465 	u32 rss_key[10];
3466 	int i;
3467 
3468 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3469 	for (i = 0; i < 10; i++)
3470 		ew32(RSSRK(i), rss_key[i]);
3471 
3472 	/* Direct all traffic to queue 0 */
3473 	for (i = 0; i < 32; i++)
3474 		ew32(RETA(i), 0);
3475 
3476 	/* Disable raw packet checksumming so that RSS hash is placed in
3477 	 * descriptor on writeback.
3478 	 */
3479 	rxcsum = er32(RXCSUM);
3480 	rxcsum |= E1000_RXCSUM_PCSD;
3481 
3482 	ew32(RXCSUM, rxcsum);
3483 
3484 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3485 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3486 		E1000_MRQC_RSS_FIELD_IPV6 |
3487 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3488 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3489 
3490 	ew32(MRQC, mrqc);
3491 }
3492 
3493 /**
3494  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3495  * @adapter: board private structure
3496  * @timinca: pointer to returned time increment attributes
3497  *
3498  * Get attributes for incrementing the System Time Register SYSTIML/H at
3499  * the default base frequency, and set the cyclecounter shift value.
3500  **/
e1000e_get_base_timinca(struct e1000_adapter * adapter,u32 * timinca)3501 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3502 {
3503 	struct e1000_hw *hw = &adapter->hw;
3504 	u32 incvalue, incperiod, shift;
3505 
3506 	/* Make sure clock is enabled on I217/I218/I219  before checking
3507 	 * the frequency
3508 	 */
3509 	if ((hw->mac.type >= e1000_pch_lpt) &&
3510 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3511 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3512 		u32 fextnvm7 = er32(FEXTNVM7);
3513 
3514 		if (!(fextnvm7 & BIT(0))) {
3515 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3516 			e1e_flush();
3517 		}
3518 	}
3519 
3520 	switch (hw->mac.type) {
3521 	case e1000_pch2lan:
3522 		/* Stable 96MHz frequency */
3523 		incperiod = INCPERIOD_96MHZ;
3524 		incvalue = INCVALUE_96MHZ;
3525 		shift = INCVALUE_SHIFT_96MHZ;
3526 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3527 		break;
3528 	case e1000_pch_lpt:
3529 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3530 			/* Stable 96MHz frequency */
3531 			incperiod = INCPERIOD_96MHZ;
3532 			incvalue = INCVALUE_96MHZ;
3533 			shift = INCVALUE_SHIFT_96MHZ;
3534 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3535 		} else {
3536 			/* Stable 25MHz frequency */
3537 			incperiod = INCPERIOD_25MHZ;
3538 			incvalue = INCVALUE_25MHZ;
3539 			shift = INCVALUE_SHIFT_25MHZ;
3540 			adapter->cc.shift = shift;
3541 		}
3542 		break;
3543 	case e1000_pch_spt:
3544 		/* Stable 24MHz frequency */
3545 		incperiod = INCPERIOD_24MHZ;
3546 		incvalue = INCVALUE_24MHZ;
3547 		shift = INCVALUE_SHIFT_24MHZ;
3548 		adapter->cc.shift = shift;
3549 		break;
3550 	case e1000_pch_cnp:
3551 	case e1000_pch_tgp:
3552 	case e1000_pch_adp:
3553 	case e1000_pch_mtp:
3554 	case e1000_pch_lnp:
3555 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3556 			/* Stable 24MHz frequency */
3557 			incperiod = INCPERIOD_24MHZ;
3558 			incvalue = INCVALUE_24MHZ;
3559 			shift = INCVALUE_SHIFT_24MHZ;
3560 			adapter->cc.shift = shift;
3561 		} else {
3562 			/* Stable 38400KHz frequency */
3563 			incperiod = INCPERIOD_38400KHZ;
3564 			incvalue = INCVALUE_38400KHZ;
3565 			shift = INCVALUE_SHIFT_38400KHZ;
3566 			adapter->cc.shift = shift;
3567 		}
3568 		break;
3569 	case e1000_82574:
3570 	case e1000_82583:
3571 		/* Stable 25MHz frequency */
3572 		incperiod = INCPERIOD_25MHZ;
3573 		incvalue = INCVALUE_25MHZ;
3574 		shift = INCVALUE_SHIFT_25MHZ;
3575 		adapter->cc.shift = shift;
3576 		break;
3577 	default:
3578 		return -EINVAL;
3579 	}
3580 
3581 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3582 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3583 
3584 	return 0;
3585 }
3586 
3587 /**
3588  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3589  * @adapter: board private structure
3590  * @config: timestamp configuration
3591  *
3592  * Outgoing time stamping can be enabled and disabled. Play nice and
3593  * disable it when requested, although it shouldn't cause any overhead
3594  * when no packet needs it. At most one packet in the queue may be
3595  * marked for time stamping, otherwise it would be impossible to tell
3596  * for sure to which packet the hardware time stamp belongs.
3597  *
3598  * Incoming time stamping has to be configured via the hardware filters.
3599  * Not all combinations are supported, in particular event type has to be
3600  * specified. Matching the kind of event packet is not supported, with the
3601  * exception of "all V2 events regardless of level 2 or 4".
3602  **/
e1000e_config_hwtstamp(struct e1000_adapter * adapter,struct hwtstamp_config * config)3603 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3604 				  struct hwtstamp_config *config)
3605 {
3606 	struct e1000_hw *hw = &adapter->hw;
3607 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3608 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3609 	u32 rxmtrl = 0;
3610 	u16 rxudp = 0;
3611 	bool is_l4 = false;
3612 	bool is_l2 = false;
3613 	u32 regval;
3614 
3615 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3616 		return -EINVAL;
3617 
3618 	/* flags reserved for future extensions - must be zero */
3619 	if (config->flags)
3620 		return -EINVAL;
3621 
3622 	switch (config->tx_type) {
3623 	case HWTSTAMP_TX_OFF:
3624 		tsync_tx_ctl = 0;
3625 		break;
3626 	case HWTSTAMP_TX_ON:
3627 		break;
3628 	default:
3629 		return -ERANGE;
3630 	}
3631 
3632 	switch (config->rx_filter) {
3633 	case HWTSTAMP_FILTER_NONE:
3634 		tsync_rx_ctl = 0;
3635 		break;
3636 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3637 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3638 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3639 		is_l4 = true;
3640 		break;
3641 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3642 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3643 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3644 		is_l4 = true;
3645 		break;
3646 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3647 		/* Also time stamps V2 L2 Path Delay Request/Response */
3648 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3649 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3650 		is_l2 = true;
3651 		break;
3652 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3653 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3654 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3655 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3656 		is_l2 = true;
3657 		break;
3658 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3659 		/* Hardware cannot filter just V2 L4 Sync messages */
3660 		fallthrough;
3661 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3662 		/* Also time stamps V2 Path Delay Request/Response. */
3663 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3664 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3665 		is_l2 = true;
3666 		is_l4 = true;
3667 		break;
3668 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3669 		/* Hardware cannot filter just V2 L4 Delay Request messages */
3670 		fallthrough;
3671 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3672 		/* Also time stamps V2 Path Delay Request/Response. */
3673 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3674 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3675 		is_l2 = true;
3676 		is_l4 = true;
3677 		break;
3678 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3679 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3680 		/* Hardware cannot filter just V2 L4 or L2 Event messages */
3681 		fallthrough;
3682 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3683 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3684 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3685 		is_l2 = true;
3686 		is_l4 = true;
3687 		break;
3688 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3689 		/* For V1, the hardware can only filter Sync messages or
3690 		 * Delay Request messages but not both so fall-through to
3691 		 * time stamp all packets.
3692 		 */
3693 		fallthrough;
3694 	case HWTSTAMP_FILTER_NTP_ALL:
3695 	case HWTSTAMP_FILTER_ALL:
3696 		is_l2 = true;
3697 		is_l4 = true;
3698 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3699 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3700 		break;
3701 	default:
3702 		return -ERANGE;
3703 	}
3704 
3705 	adapter->hwtstamp_config = *config;
3706 
3707 	/* enable/disable Tx h/w time stamping */
3708 	regval = er32(TSYNCTXCTL);
3709 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3710 	regval |= tsync_tx_ctl;
3711 	ew32(TSYNCTXCTL, regval);
3712 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3713 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3714 		e_err("Timesync Tx Control register not set as expected\n");
3715 		return -EAGAIN;
3716 	}
3717 
3718 	/* enable/disable Rx h/w time stamping */
3719 	regval = er32(TSYNCRXCTL);
3720 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3721 	regval |= tsync_rx_ctl;
3722 	ew32(TSYNCRXCTL, regval);
3723 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3724 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3725 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3726 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3727 		e_err("Timesync Rx Control register not set as expected\n");
3728 		return -EAGAIN;
3729 	}
3730 
3731 	/* L2: define ethertype filter for time stamped packets */
3732 	if (is_l2)
3733 		rxmtrl |= ETH_P_1588;
3734 
3735 	/* define which PTP packets get time stamped */
3736 	ew32(RXMTRL, rxmtrl);
3737 
3738 	/* Filter by destination port */
3739 	if (is_l4) {
3740 		rxudp = PTP_EV_PORT;
3741 		cpu_to_be16s(&rxudp);
3742 	}
3743 	ew32(RXUDP, rxudp);
3744 
3745 	e1e_flush();
3746 
3747 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3748 	er32(RXSTMPH);
3749 	er32(TXSTMPH);
3750 
3751 	return 0;
3752 }
3753 
3754 /**
3755  * e1000_configure - configure the hardware for Rx and Tx
3756  * @adapter: private board structure
3757  **/
e1000_configure(struct e1000_adapter * adapter)3758 static void e1000_configure(struct e1000_adapter *adapter)
3759 {
3760 	struct e1000_ring *rx_ring = adapter->rx_ring;
3761 
3762 	e1000e_set_rx_mode(adapter->netdev);
3763 
3764 	e1000_restore_vlan(adapter);
3765 	e1000_init_manageability_pt(adapter);
3766 
3767 	e1000_configure_tx(adapter);
3768 
3769 	if (adapter->netdev->features & NETIF_F_RXHASH)
3770 		e1000e_setup_rss_hash(adapter);
3771 	e1000_setup_rctl(adapter);
3772 	e1000_configure_rx(adapter);
3773 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3774 }
3775 
3776 /**
3777  * e1000e_power_up_phy - restore link in case the phy was powered down
3778  * @adapter: address of board private structure
3779  *
3780  * The phy may be powered down to save power and turn off link when the
3781  * driver is unloaded and wake on lan is not enabled (among others)
3782  * *** this routine MUST be followed by a call to e1000e_reset ***
3783  **/
e1000e_power_up_phy(struct e1000_adapter * adapter)3784 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3785 {
3786 	if (adapter->hw.phy.ops.power_up)
3787 		adapter->hw.phy.ops.power_up(&adapter->hw);
3788 
3789 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3790 }
3791 
3792 /**
3793  * e1000_power_down_phy - Power down the PHY
3794  * @adapter: board private structure
3795  *
3796  * Power down the PHY so no link is implied when interface is down.
3797  * The PHY cannot be powered down if management or WoL is active.
3798  */
e1000_power_down_phy(struct e1000_adapter * adapter)3799 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3800 {
3801 	if (adapter->hw.phy.ops.power_down)
3802 		adapter->hw.phy.ops.power_down(&adapter->hw);
3803 }
3804 
3805 /**
3806  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3807  * @adapter: board private structure
3808  *
3809  * We want to clear all pending descriptors from the TX ring.
3810  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3811  * the data of the next descriptor. We don't care about the data we are about
3812  * to reset the HW.
3813  */
e1000_flush_tx_ring(struct e1000_adapter * adapter)3814 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3815 {
3816 	struct e1000_hw *hw = &adapter->hw;
3817 	struct e1000_ring *tx_ring = adapter->tx_ring;
3818 	struct e1000_tx_desc *tx_desc = NULL;
3819 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3820 	u16 size = 512;
3821 
3822 	tctl = er32(TCTL);
3823 	ew32(TCTL, tctl | E1000_TCTL_EN);
3824 	tdt = er32(TDT(0));
3825 	BUG_ON(tdt != tx_ring->next_to_use);
3826 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3827 	tx_desc->buffer_addr = cpu_to_le64(tx_ring->dma);
3828 
3829 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3830 	tx_desc->upper.data = 0;
3831 	/* flush descriptors to memory before notifying the HW */
3832 	wmb();
3833 	tx_ring->next_to_use++;
3834 	if (tx_ring->next_to_use == tx_ring->count)
3835 		tx_ring->next_to_use = 0;
3836 	ew32(TDT(0), tx_ring->next_to_use);
3837 	usleep_range(200, 250);
3838 }
3839 
3840 /**
3841  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3842  * @adapter: board private structure
3843  *
3844  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3845  */
e1000_flush_rx_ring(struct e1000_adapter * adapter)3846 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3847 {
3848 	u32 rctl, rxdctl;
3849 	struct e1000_hw *hw = &adapter->hw;
3850 
3851 	rctl = er32(RCTL);
3852 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3853 	e1e_flush();
3854 	usleep_range(100, 150);
3855 
3856 	rxdctl = er32(RXDCTL(0));
3857 	/* zero the lower 14 bits (prefetch and host thresholds) */
3858 	rxdctl &= 0xffffc000;
3859 
3860 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3861 	 * and make sure the granularity is "descriptors" and not "cache lines"
3862 	 */
3863 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3864 
3865 	ew32(RXDCTL(0), rxdctl);
3866 	/* momentarily enable the RX ring for the changes to take effect */
3867 	ew32(RCTL, rctl | E1000_RCTL_EN);
3868 	e1e_flush();
3869 	usleep_range(100, 150);
3870 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3871 }
3872 
3873 /**
3874  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3875  * @adapter: board private structure
3876  *
3877  * In i219, the descriptor rings must be emptied before resetting the HW
3878  * or before changing the device state to D3 during runtime (runtime PM).
3879  *
3880  * Failure to do this will cause the HW to enter a unit hang state which can
3881  * only be released by PCI reset on the device
3882  *
3883  */
3884 
e1000_flush_desc_rings(struct e1000_adapter * adapter)3885 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3886 {
3887 	u16 hang_state;
3888 	u32 fext_nvm11, tdlen;
3889 	struct e1000_hw *hw = &adapter->hw;
3890 
3891 	/* First, disable MULR fix in FEXTNVM11 */
3892 	fext_nvm11 = er32(FEXTNVM11);
3893 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3894 	ew32(FEXTNVM11, fext_nvm11);
3895 	/* do nothing if we're not in faulty state, or if the queue is empty */
3896 	tdlen = er32(TDLEN(0));
3897 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3898 			     &hang_state);
3899 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3900 		return;
3901 	e1000_flush_tx_ring(adapter);
3902 	/* recheck, maybe the fault is caused by the rx ring */
3903 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3904 			     &hang_state);
3905 	if (hang_state & FLUSH_DESC_REQUIRED)
3906 		e1000_flush_rx_ring(adapter);
3907 }
3908 
3909 /**
3910  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3911  * @adapter: board private structure
3912  *
3913  * When the MAC is reset, all hardware bits for timesync will be reset to the
3914  * default values. This function will restore the settings last in place.
3915  * Since the clock SYSTIME registers are reset, we will simply restore the
3916  * cyclecounter to the kernel real clock time.
3917  **/
e1000e_systim_reset(struct e1000_adapter * adapter)3918 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3919 {
3920 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3921 	struct e1000_hw *hw = &adapter->hw;
3922 	unsigned long flags;
3923 	u32 timinca;
3924 	s32 ret_val;
3925 
3926 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3927 		return;
3928 
3929 	if (info->adjfreq) {
3930 		/* restore the previous ptp frequency delta */
3931 		ret_val = info->adjfreq(info, adapter->ptp_delta);
3932 	} else {
3933 		/* set the default base frequency if no adjustment possible */
3934 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3935 		if (!ret_val)
3936 			ew32(TIMINCA, timinca);
3937 	}
3938 
3939 	if (ret_val) {
3940 		dev_warn(&adapter->pdev->dev,
3941 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3942 			 ret_val);
3943 		return;
3944 	}
3945 
3946 	/* reset the systim ns time counter */
3947 	spin_lock_irqsave(&adapter->systim_lock, flags);
3948 	timecounter_init(&adapter->tc, &adapter->cc,
3949 			 ktime_to_ns(ktime_get_real()));
3950 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3951 
3952 	/* restore the previous hwtstamp configuration settings */
3953 	e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config);
3954 }
3955 
3956 /**
3957  * e1000e_reset - bring the hardware into a known good state
3958  * @adapter: board private structure
3959  *
3960  * This function boots the hardware and enables some settings that
3961  * require a configuration cycle of the hardware - those cannot be
3962  * set/changed during runtime. After reset the device needs to be
3963  * properly configured for Rx, Tx etc.
3964  */
e1000e_reset(struct e1000_adapter * adapter)3965 void e1000e_reset(struct e1000_adapter *adapter)
3966 {
3967 	struct e1000_mac_info *mac = &adapter->hw.mac;
3968 	struct e1000_fc_info *fc = &adapter->hw.fc;
3969 	struct e1000_hw *hw = &adapter->hw;
3970 	u32 tx_space, min_tx_space, min_rx_space;
3971 	u32 pba = adapter->pba;
3972 	u16 hwm;
3973 
3974 	/* reset Packet Buffer Allocation to default */
3975 	ew32(PBA, pba);
3976 
3977 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3978 		/* To maintain wire speed transmits, the Tx FIFO should be
3979 		 * large enough to accommodate two full transmit packets,
3980 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3981 		 * the Rx FIFO should be large enough to accommodate at least
3982 		 * one full receive packet and is similarly rounded up and
3983 		 * expressed in KB.
3984 		 */
3985 		pba = er32(PBA);
3986 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3987 		tx_space = pba >> 16;
3988 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3989 		pba &= 0xffff;
3990 		/* the Tx fifo also stores 16 bytes of information about the Tx
3991 		 * but don't include ethernet FCS because hardware appends it
3992 		 */
3993 		min_tx_space = (adapter->max_frame_size +
3994 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3995 		min_tx_space = ALIGN(min_tx_space, 1024);
3996 		min_tx_space >>= 10;
3997 		/* software strips receive CRC, so leave room for it */
3998 		min_rx_space = adapter->max_frame_size;
3999 		min_rx_space = ALIGN(min_rx_space, 1024);
4000 		min_rx_space >>= 10;
4001 
4002 		/* If current Tx allocation is less than the min Tx FIFO size,
4003 		 * and the min Tx FIFO size is less than the current Rx FIFO
4004 		 * allocation, take space away from current Rx allocation
4005 		 */
4006 		if ((tx_space < min_tx_space) &&
4007 		    ((min_tx_space - tx_space) < pba)) {
4008 			pba -= min_tx_space - tx_space;
4009 
4010 			/* if short on Rx space, Rx wins and must trump Tx
4011 			 * adjustment
4012 			 */
4013 			if (pba < min_rx_space)
4014 				pba = min_rx_space;
4015 		}
4016 
4017 		ew32(PBA, pba);
4018 	}
4019 
4020 	/* flow control settings
4021 	 *
4022 	 * The high water mark must be low enough to fit one full frame
4023 	 * (or the size used for early receive) above it in the Rx FIFO.
4024 	 * Set it to the lower of:
4025 	 * - 90% of the Rx FIFO size, and
4026 	 * - the full Rx FIFO size minus one full frame
4027 	 */
4028 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4029 		fc->pause_time = 0xFFFF;
4030 	else
4031 		fc->pause_time = E1000_FC_PAUSE_TIME;
4032 	fc->send_xon = true;
4033 	fc->current_mode = fc->requested_mode;
4034 
4035 	switch (hw->mac.type) {
4036 	case e1000_ich9lan:
4037 	case e1000_ich10lan:
4038 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4039 			pba = 14;
4040 			ew32(PBA, pba);
4041 			fc->high_water = 0x2800;
4042 			fc->low_water = fc->high_water - 8;
4043 			break;
4044 		}
4045 		fallthrough;
4046 	default:
4047 		hwm = min(((pba << 10) * 9 / 10),
4048 			  ((pba << 10) - adapter->max_frame_size));
4049 
4050 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4051 		fc->low_water = fc->high_water - 8;
4052 		break;
4053 	case e1000_pchlan:
4054 		/* Workaround PCH LOM adapter hangs with certain network
4055 		 * loads.  If hangs persist, try disabling Tx flow control.
4056 		 */
4057 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4058 			fc->high_water = 0x3500;
4059 			fc->low_water = 0x1500;
4060 		} else {
4061 			fc->high_water = 0x5000;
4062 			fc->low_water = 0x3000;
4063 		}
4064 		fc->refresh_time = 0x1000;
4065 		break;
4066 	case e1000_pch2lan:
4067 	case e1000_pch_lpt:
4068 	case e1000_pch_spt:
4069 	case e1000_pch_cnp:
4070 	case e1000_pch_tgp:
4071 	case e1000_pch_adp:
4072 	case e1000_pch_mtp:
4073 	case e1000_pch_lnp:
4074 		fc->refresh_time = 0xFFFF;
4075 		fc->pause_time = 0xFFFF;
4076 
4077 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4078 			fc->high_water = 0x05C20;
4079 			fc->low_water = 0x05048;
4080 			break;
4081 		}
4082 
4083 		pba = 14;
4084 		ew32(PBA, pba);
4085 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4086 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4087 		break;
4088 	}
4089 
4090 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4091 	 * maximum size per Tx descriptor limited only to the transmit
4092 	 * allocation of the packet buffer minus 96 bytes with an upper
4093 	 * limit of 24KB due to receive synchronization limitations.
4094 	 */
4095 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4096 				       24 << 10);
4097 
4098 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4099 	 * fit in receive buffer.
4100 	 */
4101 	if (adapter->itr_setting & 0x3) {
4102 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4103 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4104 				dev_info(&adapter->pdev->dev,
4105 					 "Interrupt Throttle Rate off\n");
4106 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4107 				e1000e_write_itr(adapter, 0);
4108 			}
4109 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4110 			dev_info(&adapter->pdev->dev,
4111 				 "Interrupt Throttle Rate on\n");
4112 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4113 			adapter->itr = 20000;
4114 			e1000e_write_itr(adapter, adapter->itr);
4115 		}
4116 	}
4117 
4118 	if (hw->mac.type >= e1000_pch_spt)
4119 		e1000_flush_desc_rings(adapter);
4120 	/* Allow time for pending master requests to run */
4121 	mac->ops.reset_hw(hw);
4122 
4123 	/* For parts with AMT enabled, let the firmware know
4124 	 * that the network interface is in control
4125 	 */
4126 	if (adapter->flags & FLAG_HAS_AMT)
4127 		e1000e_get_hw_control(adapter);
4128 
4129 	ew32(WUC, 0);
4130 
4131 	if (mac->ops.init_hw(hw))
4132 		e_err("Hardware Error\n");
4133 
4134 	e1000_update_mng_vlan(adapter);
4135 
4136 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4137 	ew32(VET, ETH_P_8021Q);
4138 
4139 	e1000e_reset_adaptive(hw);
4140 
4141 	/* restore systim and hwtstamp settings */
4142 	e1000e_systim_reset(adapter);
4143 
4144 	/* Set EEE advertisement as appropriate */
4145 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4146 		s32 ret_val;
4147 		u16 adv_addr;
4148 
4149 		switch (hw->phy.type) {
4150 		case e1000_phy_82579:
4151 			adv_addr = I82579_EEE_ADVERTISEMENT;
4152 			break;
4153 		case e1000_phy_i217:
4154 			adv_addr = I217_EEE_ADVERTISEMENT;
4155 			break;
4156 		default:
4157 			dev_err(&adapter->pdev->dev,
4158 				"Invalid PHY type setting EEE advertisement\n");
4159 			return;
4160 		}
4161 
4162 		ret_val = hw->phy.ops.acquire(hw);
4163 		if (ret_val) {
4164 			dev_err(&adapter->pdev->dev,
4165 				"EEE advertisement - unable to acquire PHY\n");
4166 			return;
4167 		}
4168 
4169 		e1000_write_emi_reg_locked(hw, adv_addr,
4170 					   hw->dev_spec.ich8lan.eee_disable ?
4171 					   0 : adapter->eee_advert);
4172 
4173 		hw->phy.ops.release(hw);
4174 	}
4175 
4176 	if (!netif_running(adapter->netdev) &&
4177 	    !test_bit(__E1000_TESTING, &adapter->state))
4178 		e1000_power_down_phy(adapter);
4179 
4180 	e1000_get_phy_info(hw);
4181 
4182 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4183 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4184 		u16 phy_data = 0;
4185 		/* speed up time to link by disabling smart power down, ignore
4186 		 * the return value of this function because there is nothing
4187 		 * different we would do if it failed
4188 		 */
4189 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4190 		phy_data &= ~IGP02E1000_PM_SPD;
4191 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4192 	}
4193 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4194 		u32 reg;
4195 
4196 		/* Fextnvm7 @ 0xe4[2] = 1 */
4197 		reg = er32(FEXTNVM7);
4198 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4199 		ew32(FEXTNVM7, reg);
4200 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4201 		reg = er32(FEXTNVM9);
4202 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4203 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4204 		ew32(FEXTNVM9, reg);
4205 	}
4206 
4207 }
4208 
4209 /**
4210  * e1000e_trigger_lsc - trigger an LSC interrupt
4211  * @adapter:
4212  *
4213  * Fire a link status change interrupt to start the watchdog.
4214  **/
e1000e_trigger_lsc(struct e1000_adapter * adapter)4215 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4216 {
4217 	struct e1000_hw *hw = &adapter->hw;
4218 
4219 	if (adapter->msix_entries)
4220 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4221 	else
4222 		ew32(ICS, E1000_ICS_LSC);
4223 }
4224 
e1000e_up(struct e1000_adapter * adapter)4225 void e1000e_up(struct e1000_adapter *adapter)
4226 {
4227 	/* hardware has been reset, we need to reload some things */
4228 	e1000_configure(adapter);
4229 
4230 	clear_bit(__E1000_DOWN, &adapter->state);
4231 
4232 	if (adapter->msix_entries)
4233 		e1000_configure_msix(adapter);
4234 	e1000_irq_enable(adapter);
4235 
4236 	/* Tx queue started by watchdog timer when link is up */
4237 
4238 	e1000e_trigger_lsc(adapter);
4239 }
4240 
e1000e_flush_descriptors(struct e1000_adapter * adapter)4241 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4242 {
4243 	struct e1000_hw *hw = &adapter->hw;
4244 
4245 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4246 		return;
4247 
4248 	/* flush pending descriptor writebacks to memory */
4249 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4250 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4251 
4252 	/* execute the writes immediately */
4253 	e1e_flush();
4254 
4255 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4256 	 * write is successful
4257 	 */
4258 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4259 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4260 
4261 	/* execute the writes immediately */
4262 	e1e_flush();
4263 }
4264 
4265 static void e1000e_update_stats(struct e1000_adapter *adapter);
4266 
4267 /**
4268  * e1000e_down - quiesce the device and optionally reset the hardware
4269  * @adapter: board private structure
4270  * @reset: boolean flag to reset the hardware or not
4271  */
e1000e_down(struct e1000_adapter * adapter,bool reset)4272 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4273 {
4274 	struct net_device *netdev = adapter->netdev;
4275 	struct e1000_hw *hw = &adapter->hw;
4276 	u32 tctl, rctl;
4277 
4278 	/* signal that we're down so the interrupt handler does not
4279 	 * reschedule our watchdog timer
4280 	 */
4281 	set_bit(__E1000_DOWN, &adapter->state);
4282 
4283 	netif_carrier_off(netdev);
4284 
4285 	/* disable receives in the hardware */
4286 	rctl = er32(RCTL);
4287 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4288 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4289 	/* flush and sleep below */
4290 
4291 	netif_stop_queue(netdev);
4292 
4293 	/* disable transmits in the hardware */
4294 	tctl = er32(TCTL);
4295 	tctl &= ~E1000_TCTL_EN;
4296 	ew32(TCTL, tctl);
4297 
4298 	/* flush both disables and wait for them to finish */
4299 	e1e_flush();
4300 	usleep_range(10000, 11000);
4301 
4302 	e1000_irq_disable(adapter);
4303 
4304 	napi_synchronize(&adapter->napi);
4305 
4306 	del_timer_sync(&adapter->watchdog_timer);
4307 	del_timer_sync(&adapter->phy_info_timer);
4308 
4309 	spin_lock(&adapter->stats64_lock);
4310 	e1000e_update_stats(adapter);
4311 	spin_unlock(&adapter->stats64_lock);
4312 
4313 	e1000e_flush_descriptors(adapter);
4314 
4315 	adapter->link_speed = 0;
4316 	adapter->link_duplex = 0;
4317 
4318 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4319 	if ((hw->mac.type >= e1000_pch2lan) &&
4320 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4321 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4322 		e_dbg("failed to disable jumbo frame workaround mode\n");
4323 
4324 	if (!pci_channel_offline(adapter->pdev)) {
4325 		if (reset)
4326 			e1000e_reset(adapter);
4327 		else if (hw->mac.type >= e1000_pch_spt)
4328 			e1000_flush_desc_rings(adapter);
4329 	}
4330 	e1000_clean_tx_ring(adapter->tx_ring);
4331 	e1000_clean_rx_ring(adapter->rx_ring);
4332 }
4333 
e1000e_reinit_locked(struct e1000_adapter * adapter)4334 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4335 {
4336 	might_sleep();
4337 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4338 		usleep_range(1000, 1100);
4339 	e1000e_down(adapter, true);
4340 	e1000e_up(adapter);
4341 	clear_bit(__E1000_RESETTING, &adapter->state);
4342 }
4343 
4344 /**
4345  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4346  * @hw: pointer to the HW structure
4347  * @systim: PHC time value read, sanitized and returned
4348  * @sts: structure to hold system time before and after reading SYSTIML,
4349  * may be NULL
4350  *
4351  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4352  * check to see that the time is incrementing at a reasonable
4353  * rate and is a multiple of incvalue.
4354  **/
e1000e_sanitize_systim(struct e1000_hw * hw,u64 systim,struct ptp_system_timestamp * sts)4355 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4356 				  struct ptp_system_timestamp *sts)
4357 {
4358 	u64 time_delta, rem, temp;
4359 	u64 systim_next;
4360 	u32 incvalue;
4361 	int i;
4362 
4363 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4364 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4365 		/* latch SYSTIMH on read of SYSTIML */
4366 		ptp_read_system_prets(sts);
4367 		systim_next = (u64)er32(SYSTIML);
4368 		ptp_read_system_postts(sts);
4369 		systim_next |= (u64)er32(SYSTIMH) << 32;
4370 
4371 		time_delta = systim_next - systim;
4372 		temp = time_delta;
4373 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4374 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4375 
4376 		systim = systim_next;
4377 
4378 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4379 			break;
4380 	}
4381 
4382 	return systim;
4383 }
4384 
4385 /**
4386  * e1000e_read_systim - read SYSTIM register
4387  * @adapter: board private structure
4388  * @sts: structure which will contain system time before and after reading
4389  * SYSTIML, may be NULL
4390  **/
e1000e_read_systim(struct e1000_adapter * adapter,struct ptp_system_timestamp * sts)4391 u64 e1000e_read_systim(struct e1000_adapter *adapter,
4392 		       struct ptp_system_timestamp *sts)
4393 {
4394 	struct e1000_hw *hw = &adapter->hw;
4395 	u32 systimel, systimel_2, systimeh;
4396 	u64 systim;
4397 	/* SYSTIMH latching upon SYSTIML read does not work well.
4398 	 * This means that if SYSTIML overflows after we read it but before
4399 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4400 	 * will experience a huge non linear increment in the systime value
4401 	 * to fix that we test for overflow and if true, we re-read systime.
4402 	 */
4403 	ptp_read_system_prets(sts);
4404 	systimel = er32(SYSTIML);
4405 	ptp_read_system_postts(sts);
4406 	systimeh = er32(SYSTIMH);
4407 	/* Is systimel is so large that overflow is possible? */
4408 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4409 		ptp_read_system_prets(sts);
4410 		systimel_2 = er32(SYSTIML);
4411 		ptp_read_system_postts(sts);
4412 		if (systimel > systimel_2) {
4413 			/* There was an overflow, read again SYSTIMH, and use
4414 			 * systimel_2
4415 			 */
4416 			systimeh = er32(SYSTIMH);
4417 			systimel = systimel_2;
4418 		}
4419 	}
4420 	systim = (u64)systimel;
4421 	systim |= (u64)systimeh << 32;
4422 
4423 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4424 		systim = e1000e_sanitize_systim(hw, systim, sts);
4425 
4426 	return systim;
4427 }
4428 
4429 /**
4430  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4431  * @cc: cyclecounter structure
4432  **/
e1000e_cyclecounter_read(const struct cyclecounter * cc)4433 static u64 e1000e_cyclecounter_read(const struct cyclecounter *cc)
4434 {
4435 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4436 						     cc);
4437 
4438 	return e1000e_read_systim(adapter, NULL);
4439 }
4440 
4441 /**
4442  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4443  * @adapter: board private structure to initialize
4444  *
4445  * e1000_sw_init initializes the Adapter private data structure.
4446  * Fields are initialized based on PCI device information and
4447  * OS network device settings (MTU size).
4448  **/
e1000_sw_init(struct e1000_adapter * adapter)4449 static int e1000_sw_init(struct e1000_adapter *adapter)
4450 {
4451 	struct net_device *netdev = adapter->netdev;
4452 
4453 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4454 	adapter->rx_ps_bsize0 = 128;
4455 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4456 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4457 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4458 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4459 
4460 	spin_lock_init(&adapter->stats64_lock);
4461 
4462 	e1000e_set_interrupt_capability(adapter);
4463 
4464 	if (e1000_alloc_queues(adapter))
4465 		return -ENOMEM;
4466 
4467 	/* Setup hardware time stamping cyclecounter */
4468 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4469 		adapter->cc.read = e1000e_cyclecounter_read;
4470 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4471 		adapter->cc.mult = 1;
4472 		/* cc.shift set in e1000e_get_base_tininca() */
4473 
4474 		spin_lock_init(&adapter->systim_lock);
4475 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4476 	}
4477 
4478 	/* Explicitly disable IRQ since the NIC can be in any state. */
4479 	e1000_irq_disable(adapter);
4480 
4481 	set_bit(__E1000_DOWN, &adapter->state);
4482 	return 0;
4483 }
4484 
4485 /**
4486  * e1000_intr_msi_test - Interrupt Handler
4487  * @irq: interrupt number
4488  * @data: pointer to a network interface device structure
4489  **/
e1000_intr_msi_test(int __always_unused irq,void * data)4490 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4491 {
4492 	struct net_device *netdev = data;
4493 	struct e1000_adapter *adapter = netdev_priv(netdev);
4494 	struct e1000_hw *hw = &adapter->hw;
4495 	u32 icr = er32(ICR);
4496 
4497 	e_dbg("icr is %08X\n", icr);
4498 	if (icr & E1000_ICR_RXSEQ) {
4499 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4500 		/* Force memory writes to complete before acknowledging the
4501 		 * interrupt is handled.
4502 		 */
4503 		wmb();
4504 	}
4505 
4506 	return IRQ_HANDLED;
4507 }
4508 
4509 /**
4510  * e1000_test_msi_interrupt - Returns 0 for successful test
4511  * @adapter: board private struct
4512  *
4513  * code flow taken from tg3.c
4514  **/
e1000_test_msi_interrupt(struct e1000_adapter * adapter)4515 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4516 {
4517 	struct net_device *netdev = adapter->netdev;
4518 	struct e1000_hw *hw = &adapter->hw;
4519 	int err;
4520 
4521 	/* poll_enable hasn't been called yet, so don't need disable */
4522 	/* clear any pending events */
4523 	er32(ICR);
4524 
4525 	/* free the real vector and request a test handler */
4526 	e1000_free_irq(adapter);
4527 	e1000e_reset_interrupt_capability(adapter);
4528 
4529 	/* Assume that the test fails, if it succeeds then the test
4530 	 * MSI irq handler will unset this flag
4531 	 */
4532 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4533 
4534 	err = pci_enable_msi(adapter->pdev);
4535 	if (err)
4536 		goto msi_test_failed;
4537 
4538 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4539 			  netdev->name, netdev);
4540 	if (err) {
4541 		pci_disable_msi(adapter->pdev);
4542 		goto msi_test_failed;
4543 	}
4544 
4545 	/* Force memory writes to complete before enabling and firing an
4546 	 * interrupt.
4547 	 */
4548 	wmb();
4549 
4550 	e1000_irq_enable(adapter);
4551 
4552 	/* fire an unusual interrupt on the test handler */
4553 	ew32(ICS, E1000_ICS_RXSEQ);
4554 	e1e_flush();
4555 	msleep(100);
4556 
4557 	e1000_irq_disable(adapter);
4558 
4559 	rmb();			/* read flags after interrupt has been fired */
4560 
4561 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4562 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4563 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4564 	} else {
4565 		e_dbg("MSI interrupt test succeeded!\n");
4566 	}
4567 
4568 	free_irq(adapter->pdev->irq, netdev);
4569 	pci_disable_msi(adapter->pdev);
4570 
4571 msi_test_failed:
4572 	e1000e_set_interrupt_capability(adapter);
4573 	return e1000_request_irq(adapter);
4574 }
4575 
4576 /**
4577  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4578  * @adapter: board private struct
4579  *
4580  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4581  **/
e1000_test_msi(struct e1000_adapter * adapter)4582 static int e1000_test_msi(struct e1000_adapter *adapter)
4583 {
4584 	int err;
4585 	u16 pci_cmd;
4586 
4587 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4588 		return 0;
4589 
4590 	/* disable SERR in case the MSI write causes a master abort */
4591 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4592 	if (pci_cmd & PCI_COMMAND_SERR)
4593 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4594 				      pci_cmd & ~PCI_COMMAND_SERR);
4595 
4596 	err = e1000_test_msi_interrupt(adapter);
4597 
4598 	/* re-enable SERR */
4599 	if (pci_cmd & PCI_COMMAND_SERR) {
4600 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4601 		pci_cmd |= PCI_COMMAND_SERR;
4602 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4603 	}
4604 
4605 	return err;
4606 }
4607 
4608 /**
4609  * e1000e_open - Called when a network interface is made active
4610  * @netdev: network interface device structure
4611  *
4612  * Returns 0 on success, negative value on failure
4613  *
4614  * The open entry point is called when a network interface is made
4615  * active by the system (IFF_UP).  At this point all resources needed
4616  * for transmit and receive operations are allocated, the interrupt
4617  * handler is registered with the OS, the watchdog timer is started,
4618  * and the stack is notified that the interface is ready.
4619  **/
e1000e_open(struct net_device * netdev)4620 int e1000e_open(struct net_device *netdev)
4621 {
4622 	struct e1000_adapter *adapter = netdev_priv(netdev);
4623 	struct e1000_hw *hw = &adapter->hw;
4624 	struct pci_dev *pdev = adapter->pdev;
4625 	int err;
4626 
4627 	/* disallow open during test */
4628 	if (test_bit(__E1000_TESTING, &adapter->state))
4629 		return -EBUSY;
4630 
4631 	pm_runtime_get_sync(&pdev->dev);
4632 
4633 	netif_carrier_off(netdev);
4634 	netif_stop_queue(netdev);
4635 
4636 	/* allocate transmit descriptors */
4637 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4638 	if (err)
4639 		goto err_setup_tx;
4640 
4641 	/* allocate receive descriptors */
4642 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4643 	if (err)
4644 		goto err_setup_rx;
4645 
4646 	/* If AMT is enabled, let the firmware know that the network
4647 	 * interface is now open and reset the part to a known state.
4648 	 */
4649 	if (adapter->flags & FLAG_HAS_AMT) {
4650 		e1000e_get_hw_control(adapter);
4651 		e1000e_reset(adapter);
4652 	}
4653 
4654 	e1000e_power_up_phy(adapter);
4655 
4656 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4657 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4658 		e1000_update_mng_vlan(adapter);
4659 
4660 	/* DMA latency requirement to workaround jumbo issue */
4661 	cpu_latency_qos_add_request(&adapter->pm_qos_req, PM_QOS_DEFAULT_VALUE);
4662 
4663 	/* before we allocate an interrupt, we must be ready to handle it.
4664 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4665 	 * as soon as we call pci_request_irq, so we have to setup our
4666 	 * clean_rx handler before we do so.
4667 	 */
4668 	e1000_configure(adapter);
4669 
4670 	err = e1000_request_irq(adapter);
4671 	if (err)
4672 		goto err_req_irq;
4673 
4674 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4675 	 * ignore e1000e MSI messages, which means we need to test our MSI
4676 	 * interrupt now
4677 	 */
4678 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4679 		err = e1000_test_msi(adapter);
4680 		if (err) {
4681 			e_err("Interrupt allocation failed\n");
4682 			goto err_req_irq;
4683 		}
4684 	}
4685 
4686 	/* From here on the code is the same as e1000e_up() */
4687 	clear_bit(__E1000_DOWN, &adapter->state);
4688 
4689 	napi_enable(&adapter->napi);
4690 
4691 	e1000_irq_enable(adapter);
4692 
4693 	adapter->tx_hang_recheck = false;
4694 
4695 	hw->mac.get_link_status = true;
4696 	pm_runtime_put(&pdev->dev);
4697 
4698 	e1000e_trigger_lsc(adapter);
4699 
4700 	return 0;
4701 
4702 err_req_irq:
4703 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4704 	e1000e_release_hw_control(adapter);
4705 	e1000_power_down_phy(adapter);
4706 	e1000e_free_rx_resources(adapter->rx_ring);
4707 err_setup_rx:
4708 	e1000e_free_tx_resources(adapter->tx_ring);
4709 err_setup_tx:
4710 	e1000e_reset(adapter);
4711 	pm_runtime_put_sync(&pdev->dev);
4712 
4713 	return err;
4714 }
4715 
4716 /**
4717  * e1000e_close - Disables a network interface
4718  * @netdev: network interface device structure
4719  *
4720  * Returns 0, this is not allowed to fail
4721  *
4722  * The close entry point is called when an interface is de-activated
4723  * by the OS.  The hardware is still under the drivers control, but
4724  * needs to be disabled.  A global MAC reset is issued to stop the
4725  * hardware, and all transmit and receive resources are freed.
4726  **/
e1000e_close(struct net_device * netdev)4727 int e1000e_close(struct net_device *netdev)
4728 {
4729 	struct e1000_adapter *adapter = netdev_priv(netdev);
4730 	struct pci_dev *pdev = adapter->pdev;
4731 	int count = E1000_CHECK_RESET_COUNT;
4732 
4733 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4734 		usleep_range(10000, 11000);
4735 
4736 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4737 
4738 	pm_runtime_get_sync(&pdev->dev);
4739 
4740 	if (netif_device_present(netdev)) {
4741 		e1000e_down(adapter, true);
4742 		e1000_free_irq(adapter);
4743 
4744 		/* Link status message must follow this format */
4745 		netdev_info(netdev, "NIC Link is Down\n");
4746 	}
4747 
4748 	napi_disable(&adapter->napi);
4749 
4750 	e1000e_free_tx_resources(adapter->tx_ring);
4751 	e1000e_free_rx_resources(adapter->rx_ring);
4752 
4753 	/* kill manageability vlan ID if supported, but not if a vlan with
4754 	 * the same ID is registered on the host OS (let 8021q kill it)
4755 	 */
4756 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4757 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4758 				       adapter->mng_vlan_id);
4759 
4760 	/* If AMT is enabled, let the firmware know that the network
4761 	 * interface is now closed
4762 	 */
4763 	if ((adapter->flags & FLAG_HAS_AMT) &&
4764 	    !test_bit(__E1000_TESTING, &adapter->state))
4765 		e1000e_release_hw_control(adapter);
4766 
4767 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4768 
4769 	pm_runtime_put_sync(&pdev->dev);
4770 
4771 	return 0;
4772 }
4773 
4774 /**
4775  * e1000_set_mac - Change the Ethernet Address of the NIC
4776  * @netdev: network interface device structure
4777  * @p: pointer to an address structure
4778  *
4779  * Returns 0 on success, negative on failure
4780  **/
e1000_set_mac(struct net_device * netdev,void * p)4781 static int e1000_set_mac(struct net_device *netdev, void *p)
4782 {
4783 	struct e1000_adapter *adapter = netdev_priv(netdev);
4784 	struct e1000_hw *hw = &adapter->hw;
4785 	struct sockaddr *addr = p;
4786 
4787 	if (!is_valid_ether_addr(addr->sa_data))
4788 		return -EADDRNOTAVAIL;
4789 
4790 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4791 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4792 
4793 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4794 
4795 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4796 		/* activate the work around */
4797 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4798 
4799 		/* Hold a copy of the LAA in RAR[14] This is done so that
4800 		 * between the time RAR[0] gets clobbered  and the time it
4801 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4802 		 * of the RARs and no incoming packets directed to this port
4803 		 * are dropped. Eventually the LAA will be in RAR[0] and
4804 		 * RAR[14]
4805 		 */
4806 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4807 				    adapter->hw.mac.rar_entry_count - 1);
4808 	}
4809 
4810 	return 0;
4811 }
4812 
4813 /**
4814  * e1000e_update_phy_task - work thread to update phy
4815  * @work: pointer to our work struct
4816  *
4817  * this worker thread exists because we must acquire a
4818  * semaphore to read the phy, which we could msleep while
4819  * waiting for it, and we can't msleep in a timer.
4820  **/
e1000e_update_phy_task(struct work_struct * work)4821 static void e1000e_update_phy_task(struct work_struct *work)
4822 {
4823 	struct e1000_adapter *adapter = container_of(work,
4824 						     struct e1000_adapter,
4825 						     update_phy_task);
4826 	struct e1000_hw *hw = &adapter->hw;
4827 
4828 	if (test_bit(__E1000_DOWN, &adapter->state))
4829 		return;
4830 
4831 	e1000_get_phy_info(hw);
4832 
4833 	/* Enable EEE on 82579 after link up */
4834 	if (hw->phy.type >= e1000_phy_82579)
4835 		e1000_set_eee_pchlan(hw);
4836 }
4837 
4838 /**
4839  * e1000_update_phy_info - timre call-back to update PHY info
4840  * @t: pointer to timer_list containing private info adapter
4841  *
4842  * Need to wait a few seconds after link up to get diagnostic information from
4843  * the phy
4844  **/
e1000_update_phy_info(struct timer_list * t)4845 static void e1000_update_phy_info(struct timer_list *t)
4846 {
4847 	struct e1000_adapter *adapter = from_timer(adapter, t, phy_info_timer);
4848 
4849 	if (test_bit(__E1000_DOWN, &adapter->state))
4850 		return;
4851 
4852 	schedule_work(&adapter->update_phy_task);
4853 }
4854 
4855 /**
4856  * e1000e_update_phy_stats - Update the PHY statistics counters
4857  * @adapter: board private structure
4858  *
4859  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4860  **/
e1000e_update_phy_stats(struct e1000_adapter * adapter)4861 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4862 {
4863 	struct e1000_hw *hw = &adapter->hw;
4864 	s32 ret_val;
4865 	u16 phy_data;
4866 
4867 	ret_val = hw->phy.ops.acquire(hw);
4868 	if (ret_val)
4869 		return;
4870 
4871 	/* A page set is expensive so check if already on desired page.
4872 	 * If not, set to the page with the PHY status registers.
4873 	 */
4874 	hw->phy.addr = 1;
4875 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4876 					   &phy_data);
4877 	if (ret_val)
4878 		goto release;
4879 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4880 		ret_val = hw->phy.ops.set_page(hw,
4881 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4882 		if (ret_val)
4883 			goto release;
4884 	}
4885 
4886 	/* Single Collision Count */
4887 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4888 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4889 	if (!ret_val)
4890 		adapter->stats.scc += phy_data;
4891 
4892 	/* Excessive Collision Count */
4893 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4894 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4895 	if (!ret_val)
4896 		adapter->stats.ecol += phy_data;
4897 
4898 	/* Multiple Collision Count */
4899 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4900 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4901 	if (!ret_val)
4902 		adapter->stats.mcc += phy_data;
4903 
4904 	/* Late Collision Count */
4905 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4906 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4907 	if (!ret_val)
4908 		adapter->stats.latecol += phy_data;
4909 
4910 	/* Collision Count - also used for adaptive IFS */
4911 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4912 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4913 	if (!ret_val)
4914 		hw->mac.collision_delta = phy_data;
4915 
4916 	/* Defer Count */
4917 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4918 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4919 	if (!ret_val)
4920 		adapter->stats.dc += phy_data;
4921 
4922 	/* Transmit with no CRS */
4923 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4924 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4925 	if (!ret_val)
4926 		adapter->stats.tncrs += phy_data;
4927 
4928 release:
4929 	hw->phy.ops.release(hw);
4930 }
4931 
4932 /**
4933  * e1000e_update_stats - Update the board statistics counters
4934  * @adapter: board private structure
4935  **/
e1000e_update_stats(struct e1000_adapter * adapter)4936 static void e1000e_update_stats(struct e1000_adapter *adapter)
4937 {
4938 	struct net_device *netdev = adapter->netdev;
4939 	struct e1000_hw *hw = &adapter->hw;
4940 	struct pci_dev *pdev = adapter->pdev;
4941 
4942 	/* Prevent stats update while adapter is being reset, or if the pci
4943 	 * connection is down.
4944 	 */
4945 	if (adapter->link_speed == 0)
4946 		return;
4947 	if (pci_channel_offline(pdev))
4948 		return;
4949 
4950 	adapter->stats.crcerrs += er32(CRCERRS);
4951 	adapter->stats.gprc += er32(GPRC);
4952 	adapter->stats.gorc += er32(GORCL);
4953 	er32(GORCH);		/* Clear gorc */
4954 	adapter->stats.bprc += er32(BPRC);
4955 	adapter->stats.mprc += er32(MPRC);
4956 	adapter->stats.roc += er32(ROC);
4957 
4958 	adapter->stats.mpc += er32(MPC);
4959 
4960 	/* Half-duplex statistics */
4961 	if (adapter->link_duplex == HALF_DUPLEX) {
4962 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4963 			e1000e_update_phy_stats(adapter);
4964 		} else {
4965 			adapter->stats.scc += er32(SCC);
4966 			adapter->stats.ecol += er32(ECOL);
4967 			adapter->stats.mcc += er32(MCC);
4968 			adapter->stats.latecol += er32(LATECOL);
4969 			adapter->stats.dc += er32(DC);
4970 
4971 			hw->mac.collision_delta = er32(COLC);
4972 
4973 			if ((hw->mac.type != e1000_82574) &&
4974 			    (hw->mac.type != e1000_82583))
4975 				adapter->stats.tncrs += er32(TNCRS);
4976 		}
4977 		adapter->stats.colc += hw->mac.collision_delta;
4978 	}
4979 
4980 	adapter->stats.xonrxc += er32(XONRXC);
4981 	adapter->stats.xontxc += er32(XONTXC);
4982 	adapter->stats.xoffrxc += er32(XOFFRXC);
4983 	adapter->stats.xofftxc += er32(XOFFTXC);
4984 	adapter->stats.gptc += er32(GPTC);
4985 	adapter->stats.gotc += er32(GOTCL);
4986 	er32(GOTCH);		/* Clear gotc */
4987 	adapter->stats.rnbc += er32(RNBC);
4988 	adapter->stats.ruc += er32(RUC);
4989 
4990 	adapter->stats.mptc += er32(MPTC);
4991 	adapter->stats.bptc += er32(BPTC);
4992 
4993 	/* used for adaptive IFS */
4994 
4995 	hw->mac.tx_packet_delta = er32(TPT);
4996 	adapter->stats.tpt += hw->mac.tx_packet_delta;
4997 
4998 	adapter->stats.algnerrc += er32(ALGNERRC);
4999 	adapter->stats.rxerrc += er32(RXERRC);
5000 	adapter->stats.cexterr += er32(CEXTERR);
5001 	adapter->stats.tsctc += er32(TSCTC);
5002 	adapter->stats.tsctfc += er32(TSCTFC);
5003 
5004 	/* Fill out the OS statistics structure */
5005 	netdev->stats.multicast = adapter->stats.mprc;
5006 	netdev->stats.collisions = adapter->stats.colc;
5007 
5008 	/* Rx Errors */
5009 
5010 	/* RLEC on some newer hardware can be incorrect so build
5011 	 * our own version based on RUC and ROC
5012 	 */
5013 	netdev->stats.rx_errors = adapter->stats.rxerrc +
5014 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5015 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5016 	netdev->stats.rx_length_errors = adapter->stats.ruc +
5017 	    adapter->stats.roc;
5018 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5019 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5020 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5021 
5022 	/* Tx Errors */
5023 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5024 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5025 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5026 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5027 
5028 	/* Tx Dropped needs to be maintained elsewhere */
5029 
5030 	/* Management Stats */
5031 	adapter->stats.mgptc += er32(MGTPTC);
5032 	adapter->stats.mgprc += er32(MGTPRC);
5033 	adapter->stats.mgpdc += er32(MGTPDC);
5034 
5035 	/* Correctable ECC Errors */
5036 	if (hw->mac.type >= e1000_pch_lpt) {
5037 		u32 pbeccsts = er32(PBECCSTS);
5038 
5039 		adapter->corr_errors +=
5040 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5041 		adapter->uncorr_errors +=
5042 		    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
5043 		    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
5044 	}
5045 }
5046 
5047 /**
5048  * e1000_phy_read_status - Update the PHY register status snapshot
5049  * @adapter: board private structure
5050  **/
e1000_phy_read_status(struct e1000_adapter * adapter)5051 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5052 {
5053 	struct e1000_hw *hw = &adapter->hw;
5054 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5055 
5056 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5057 	    (er32(STATUS) & E1000_STATUS_LU) &&
5058 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5059 		int ret_val;
5060 
5061 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5062 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5063 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5064 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5065 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5066 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5067 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5068 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5069 		if (ret_val)
5070 			e_warn("Error reading PHY register\n");
5071 	} else {
5072 		/* Do not read PHY registers if link is not up
5073 		 * Set values to typical power-on defaults
5074 		 */
5075 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5076 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5077 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5078 			     BMSR_ERCAP);
5079 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5080 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5081 		phy->lpa = 0;
5082 		phy->expansion = EXPANSION_ENABLENPAGE;
5083 		phy->ctrl1000 = ADVERTISE_1000FULL;
5084 		phy->stat1000 = 0;
5085 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5086 	}
5087 }
5088 
e1000_print_link_info(struct e1000_adapter * adapter)5089 static void e1000_print_link_info(struct e1000_adapter *adapter)
5090 {
5091 	struct e1000_hw *hw = &adapter->hw;
5092 	u32 ctrl = er32(CTRL);
5093 
5094 	/* Link status message must follow this format for user tools */
5095 	netdev_info(adapter->netdev,
5096 		    "NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5097 		    adapter->link_speed,
5098 		    adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5099 		    (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5100 		    (ctrl & E1000_CTRL_RFCE) ? "Rx" :
5101 		    (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5102 }
5103 
e1000e_has_link(struct e1000_adapter * adapter)5104 static bool e1000e_has_link(struct e1000_adapter *adapter)
5105 {
5106 	struct e1000_hw *hw = &adapter->hw;
5107 	bool link_active = false;
5108 	s32 ret_val = 0;
5109 
5110 	/* get_link_status is set on LSC (link status) interrupt or
5111 	 * Rx sequence error interrupt.  get_link_status will stay
5112 	 * true until the check_for_link establishes link
5113 	 * for copper adapters ONLY
5114 	 */
5115 	switch (hw->phy.media_type) {
5116 	case e1000_media_type_copper:
5117 		if (hw->mac.get_link_status) {
5118 			ret_val = hw->mac.ops.check_for_link(hw);
5119 			link_active = !hw->mac.get_link_status;
5120 		} else {
5121 			link_active = true;
5122 		}
5123 		break;
5124 	case e1000_media_type_fiber:
5125 		ret_val = hw->mac.ops.check_for_link(hw);
5126 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5127 		break;
5128 	case e1000_media_type_internal_serdes:
5129 		ret_val = hw->mac.ops.check_for_link(hw);
5130 		link_active = hw->mac.serdes_has_link;
5131 		break;
5132 	default:
5133 	case e1000_media_type_unknown:
5134 		break;
5135 	}
5136 
5137 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5138 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5139 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5140 		e_info("Gigabit has been disabled, downgrading speed\n");
5141 	}
5142 
5143 	return link_active;
5144 }
5145 
e1000e_enable_receives(struct e1000_adapter * adapter)5146 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5147 {
5148 	/* make sure the receive unit is started */
5149 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5150 	    (adapter->flags & FLAG_RESTART_NOW)) {
5151 		struct e1000_hw *hw = &adapter->hw;
5152 		u32 rctl = er32(RCTL);
5153 
5154 		ew32(RCTL, rctl | E1000_RCTL_EN);
5155 		adapter->flags &= ~FLAG_RESTART_NOW;
5156 	}
5157 }
5158 
e1000e_check_82574_phy_workaround(struct e1000_adapter * adapter)5159 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5160 {
5161 	struct e1000_hw *hw = &adapter->hw;
5162 
5163 	/* With 82574 controllers, PHY needs to be checked periodically
5164 	 * for hung state and reset, if two calls return true
5165 	 */
5166 	if (e1000_check_phy_82574(hw))
5167 		adapter->phy_hang_count++;
5168 	else
5169 		adapter->phy_hang_count = 0;
5170 
5171 	if (adapter->phy_hang_count > 1) {
5172 		adapter->phy_hang_count = 0;
5173 		e_dbg("PHY appears hung - resetting\n");
5174 		schedule_work(&adapter->reset_task);
5175 	}
5176 }
5177 
5178 /**
5179  * e1000_watchdog - Timer Call-back
5180  * @t: pointer to timer_list containing private info adapter
5181  **/
e1000_watchdog(struct timer_list * t)5182 static void e1000_watchdog(struct timer_list *t)
5183 {
5184 	struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5185 
5186 	/* Do the rest outside of interrupt context */
5187 	schedule_work(&adapter->watchdog_task);
5188 
5189 	/* TODO: make this use queue_delayed_work() */
5190 }
5191 
e1000_watchdog_task(struct work_struct * work)5192 static void e1000_watchdog_task(struct work_struct *work)
5193 {
5194 	struct e1000_adapter *adapter = container_of(work,
5195 						     struct e1000_adapter,
5196 						     watchdog_task);
5197 	struct net_device *netdev = adapter->netdev;
5198 	struct e1000_mac_info *mac = &adapter->hw.mac;
5199 	struct e1000_phy_info *phy = &adapter->hw.phy;
5200 	struct e1000_ring *tx_ring = adapter->tx_ring;
5201 	u32 dmoff_exit_timeout = 100, tries = 0;
5202 	struct e1000_hw *hw = &adapter->hw;
5203 	u32 link, tctl, pcim_state;
5204 
5205 	if (test_bit(__E1000_DOWN, &adapter->state))
5206 		return;
5207 
5208 	link = e1000e_has_link(adapter);
5209 	if ((netif_carrier_ok(netdev)) && link) {
5210 		/* Cancel scheduled suspend requests. */
5211 		pm_runtime_resume(netdev->dev.parent);
5212 
5213 		e1000e_enable_receives(adapter);
5214 		goto link_up;
5215 	}
5216 
5217 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5218 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5219 		e1000_update_mng_vlan(adapter);
5220 
5221 	if (link) {
5222 		if (!netif_carrier_ok(netdev)) {
5223 			bool txb2b = true;
5224 
5225 			/* Cancel scheduled suspend requests. */
5226 			pm_runtime_resume(netdev->dev.parent);
5227 
5228 			/* Checking if MAC is in DMoff state*/
5229 			if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
5230 				pcim_state = er32(STATUS);
5231 				while (pcim_state & E1000_STATUS_PCIM_STATE) {
5232 					if (tries++ == dmoff_exit_timeout) {
5233 						e_dbg("Error in exiting dmoff\n");
5234 						break;
5235 					}
5236 					usleep_range(10000, 20000);
5237 					pcim_state = er32(STATUS);
5238 
5239 					/* Checking if MAC exited DMoff state */
5240 					if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5241 						e1000_phy_hw_reset(&adapter->hw);
5242 				}
5243 			}
5244 
5245 			/* update snapshot of PHY registers on LSC */
5246 			e1000_phy_read_status(adapter);
5247 			mac->ops.get_link_up_info(&adapter->hw,
5248 						  &adapter->link_speed,
5249 						  &adapter->link_duplex);
5250 			e1000_print_link_info(adapter);
5251 
5252 			/* check if SmartSpeed worked */
5253 			e1000e_check_downshift(hw);
5254 			if (phy->speed_downgraded)
5255 				netdev_warn(netdev,
5256 					    "Link Speed was downgraded by SmartSpeed\n");
5257 
5258 			/* On supported PHYs, check for duplex mismatch only
5259 			 * if link has autonegotiated at 10/100 half
5260 			 */
5261 			if ((hw->phy.type == e1000_phy_igp_3 ||
5262 			     hw->phy.type == e1000_phy_bm) &&
5263 			    hw->mac.autoneg &&
5264 			    (adapter->link_speed == SPEED_10 ||
5265 			     adapter->link_speed == SPEED_100) &&
5266 			    (adapter->link_duplex == HALF_DUPLEX)) {
5267 				u16 autoneg_exp;
5268 
5269 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5270 
5271 				if (!(autoneg_exp & EXPANSION_NWAY))
5272 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5273 			}
5274 
5275 			/* adjust timeout factor according to speed/duplex */
5276 			adapter->tx_timeout_factor = 1;
5277 			switch (adapter->link_speed) {
5278 			case SPEED_10:
5279 				txb2b = false;
5280 				adapter->tx_timeout_factor = 16;
5281 				break;
5282 			case SPEED_100:
5283 				txb2b = false;
5284 				adapter->tx_timeout_factor = 10;
5285 				break;
5286 			}
5287 
5288 			/* workaround: re-program speed mode bit after
5289 			 * link-up event
5290 			 */
5291 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5292 			    !txb2b) {
5293 				u32 tarc0;
5294 
5295 				tarc0 = er32(TARC(0));
5296 				tarc0 &= ~SPEED_MODE_BIT;
5297 				ew32(TARC(0), tarc0);
5298 			}
5299 
5300 			/* disable TSO for pcie and 10/100 speeds, to avoid
5301 			 * some hardware issues
5302 			 */
5303 			if (!(adapter->flags & FLAG_TSO_FORCE)) {
5304 				switch (adapter->link_speed) {
5305 				case SPEED_10:
5306 				case SPEED_100:
5307 					e_info("10/100 speed: disabling TSO\n");
5308 					netdev->features &= ~NETIF_F_TSO;
5309 					netdev->features &= ~NETIF_F_TSO6;
5310 					break;
5311 				case SPEED_1000:
5312 					netdev->features |= NETIF_F_TSO;
5313 					netdev->features |= NETIF_F_TSO6;
5314 					break;
5315 				default:
5316 					/* oops */
5317 					break;
5318 				}
5319 				if (hw->mac.type == e1000_pch_spt) {
5320 					netdev->features &= ~NETIF_F_TSO;
5321 					netdev->features &= ~NETIF_F_TSO6;
5322 				}
5323 			}
5324 
5325 			/* enable transmits in the hardware, need to do this
5326 			 * after setting TARC(0)
5327 			 */
5328 			tctl = er32(TCTL);
5329 			tctl |= E1000_TCTL_EN;
5330 			ew32(TCTL, tctl);
5331 
5332 			/* Perform any post-link-up configuration before
5333 			 * reporting link up.
5334 			 */
5335 			if (phy->ops.cfg_on_link_up)
5336 				phy->ops.cfg_on_link_up(hw);
5337 
5338 			netif_wake_queue(netdev);
5339 			netif_carrier_on(netdev);
5340 
5341 			if (!test_bit(__E1000_DOWN, &adapter->state))
5342 				mod_timer(&adapter->phy_info_timer,
5343 					  round_jiffies(jiffies + 2 * HZ));
5344 		}
5345 	} else {
5346 		if (netif_carrier_ok(netdev)) {
5347 			adapter->link_speed = 0;
5348 			adapter->link_duplex = 0;
5349 			/* Link status message must follow this format */
5350 			netdev_info(netdev, "NIC Link is Down\n");
5351 			netif_carrier_off(netdev);
5352 			netif_stop_queue(netdev);
5353 			if (!test_bit(__E1000_DOWN, &adapter->state))
5354 				mod_timer(&adapter->phy_info_timer,
5355 					  round_jiffies(jiffies + 2 * HZ));
5356 
5357 			/* 8000ES2LAN requires a Rx packet buffer work-around
5358 			 * on link down event; reset the controller to flush
5359 			 * the Rx packet buffer.
5360 			 */
5361 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5362 				adapter->flags |= FLAG_RESTART_NOW;
5363 			else
5364 				pm_schedule_suspend(netdev->dev.parent,
5365 						    LINK_TIMEOUT);
5366 		}
5367 	}
5368 
5369 link_up:
5370 	spin_lock(&adapter->stats64_lock);
5371 	e1000e_update_stats(adapter);
5372 
5373 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5374 	adapter->tpt_old = adapter->stats.tpt;
5375 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5376 	adapter->colc_old = adapter->stats.colc;
5377 
5378 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5379 	adapter->gorc_old = adapter->stats.gorc;
5380 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5381 	adapter->gotc_old = adapter->stats.gotc;
5382 	spin_unlock(&adapter->stats64_lock);
5383 
5384 	/* If the link is lost the controller stops DMA, but
5385 	 * if there is queued Tx work it cannot be done.  So
5386 	 * reset the controller to flush the Tx packet buffers.
5387 	 */
5388 	if (!netif_carrier_ok(netdev) &&
5389 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5390 		adapter->flags |= FLAG_RESTART_NOW;
5391 
5392 	/* If reset is necessary, do it outside of interrupt context. */
5393 	if (adapter->flags & FLAG_RESTART_NOW) {
5394 		schedule_work(&adapter->reset_task);
5395 		/* return immediately since reset is imminent */
5396 		return;
5397 	}
5398 
5399 	e1000e_update_adaptive(&adapter->hw);
5400 
5401 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5402 	if (adapter->itr_setting == 4) {
5403 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5404 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5405 		 * everyone else is between 2000-8000.
5406 		 */
5407 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5408 		u32 dif = (adapter->gotc > adapter->gorc ?
5409 			   adapter->gotc - adapter->gorc :
5410 			   adapter->gorc - adapter->gotc) / 10000;
5411 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5412 
5413 		e1000e_write_itr(adapter, itr);
5414 	}
5415 
5416 	/* Cause software interrupt to ensure Rx ring is cleaned */
5417 	if (adapter->msix_entries)
5418 		ew32(ICS, adapter->rx_ring->ims_val);
5419 	else
5420 		ew32(ICS, E1000_ICS_RXDMT0);
5421 
5422 	/* flush pending descriptors to memory before detecting Tx hang */
5423 	e1000e_flush_descriptors(adapter);
5424 
5425 	/* Force detection of hung controller every watchdog period */
5426 	adapter->detect_tx_hung = true;
5427 
5428 	/* With 82571 controllers, LAA may be overwritten due to controller
5429 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5430 	 */
5431 	if (e1000e_get_laa_state_82571(hw))
5432 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5433 
5434 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5435 		e1000e_check_82574_phy_workaround(adapter);
5436 
5437 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5438 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5439 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5440 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5441 			er32(RXSTMPH);
5442 			adapter->rx_hwtstamp_cleared++;
5443 		} else {
5444 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5445 		}
5446 	}
5447 
5448 	/* Reset the timer */
5449 	if (!test_bit(__E1000_DOWN, &adapter->state))
5450 		mod_timer(&adapter->watchdog_timer,
5451 			  round_jiffies(jiffies + 2 * HZ));
5452 }
5453 
5454 #define E1000_TX_FLAGS_CSUM		0x00000001
5455 #define E1000_TX_FLAGS_VLAN		0x00000002
5456 #define E1000_TX_FLAGS_TSO		0x00000004
5457 #define E1000_TX_FLAGS_IPV4		0x00000008
5458 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5459 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5460 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5461 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5462 
e1000_tso(struct e1000_ring * tx_ring,struct sk_buff * skb,__be16 protocol)5463 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5464 		     __be16 protocol)
5465 {
5466 	struct e1000_context_desc *context_desc;
5467 	struct e1000_buffer *buffer_info;
5468 	unsigned int i;
5469 	u32 cmd_length = 0;
5470 	u16 ipcse = 0, mss;
5471 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5472 	int err;
5473 
5474 	if (!skb_is_gso(skb))
5475 		return 0;
5476 
5477 	err = skb_cow_head(skb, 0);
5478 	if (err < 0)
5479 		return err;
5480 
5481 	hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5482 	mss = skb_shinfo(skb)->gso_size;
5483 	if (protocol == htons(ETH_P_IP)) {
5484 		struct iphdr *iph = ip_hdr(skb);
5485 		iph->tot_len = 0;
5486 		iph->check = 0;
5487 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5488 							 0, IPPROTO_TCP, 0);
5489 		cmd_length = E1000_TXD_CMD_IP;
5490 		ipcse = skb_transport_offset(skb) - 1;
5491 	} else if (skb_is_gso_v6(skb)) {
5492 		tcp_v6_gso_csum_prep(skb);
5493 		ipcse = 0;
5494 	}
5495 	ipcss = skb_network_offset(skb);
5496 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5497 	tucss = skb_transport_offset(skb);
5498 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5499 
5500 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5501 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5502 
5503 	i = tx_ring->next_to_use;
5504 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5505 	buffer_info = &tx_ring->buffer_info[i];
5506 
5507 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5508 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5509 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5510 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5511 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5512 	context_desc->upper_setup.tcp_fields.tucse = 0;
5513 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5514 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5515 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5516 
5517 	buffer_info->time_stamp = jiffies;
5518 	buffer_info->next_to_watch = i;
5519 
5520 	i++;
5521 	if (i == tx_ring->count)
5522 		i = 0;
5523 	tx_ring->next_to_use = i;
5524 
5525 	return 1;
5526 }
5527 
e1000_tx_csum(struct e1000_ring * tx_ring,struct sk_buff * skb,__be16 protocol)5528 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5529 			  __be16 protocol)
5530 {
5531 	struct e1000_adapter *adapter = tx_ring->adapter;
5532 	struct e1000_context_desc *context_desc;
5533 	struct e1000_buffer *buffer_info;
5534 	unsigned int i;
5535 	u8 css;
5536 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5537 
5538 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5539 		return false;
5540 
5541 	switch (protocol) {
5542 	case cpu_to_be16(ETH_P_IP):
5543 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5544 			cmd_len |= E1000_TXD_CMD_TCP;
5545 		break;
5546 	case cpu_to_be16(ETH_P_IPV6):
5547 		/* XXX not handling all IPV6 headers */
5548 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5549 			cmd_len |= E1000_TXD_CMD_TCP;
5550 		break;
5551 	default:
5552 		if (unlikely(net_ratelimit()))
5553 			e_warn("checksum_partial proto=%x!\n",
5554 			       be16_to_cpu(protocol));
5555 		break;
5556 	}
5557 
5558 	css = skb_checksum_start_offset(skb);
5559 
5560 	i = tx_ring->next_to_use;
5561 	buffer_info = &tx_ring->buffer_info[i];
5562 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5563 
5564 	context_desc->lower_setup.ip_config = 0;
5565 	context_desc->upper_setup.tcp_fields.tucss = css;
5566 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5567 	context_desc->upper_setup.tcp_fields.tucse = 0;
5568 	context_desc->tcp_seg_setup.data = 0;
5569 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5570 
5571 	buffer_info->time_stamp = jiffies;
5572 	buffer_info->next_to_watch = i;
5573 
5574 	i++;
5575 	if (i == tx_ring->count)
5576 		i = 0;
5577 	tx_ring->next_to_use = i;
5578 
5579 	return true;
5580 }
5581 
e1000_tx_map(struct e1000_ring * tx_ring,struct sk_buff * skb,unsigned int first,unsigned int max_per_txd,unsigned int nr_frags)5582 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5583 			unsigned int first, unsigned int max_per_txd,
5584 			unsigned int nr_frags)
5585 {
5586 	struct e1000_adapter *adapter = tx_ring->adapter;
5587 	struct pci_dev *pdev = adapter->pdev;
5588 	struct e1000_buffer *buffer_info;
5589 	unsigned int len = skb_headlen(skb);
5590 	unsigned int offset = 0, size, count = 0, i;
5591 	unsigned int f, bytecount, segs;
5592 
5593 	i = tx_ring->next_to_use;
5594 
5595 	while (len) {
5596 		buffer_info = &tx_ring->buffer_info[i];
5597 		size = min(len, max_per_txd);
5598 
5599 		buffer_info->length = size;
5600 		buffer_info->time_stamp = jiffies;
5601 		buffer_info->next_to_watch = i;
5602 		buffer_info->dma = dma_map_single(&pdev->dev,
5603 						  skb->data + offset,
5604 						  size, DMA_TO_DEVICE);
5605 		buffer_info->mapped_as_page = false;
5606 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5607 			goto dma_error;
5608 
5609 		len -= size;
5610 		offset += size;
5611 		count++;
5612 
5613 		if (len) {
5614 			i++;
5615 			if (i == tx_ring->count)
5616 				i = 0;
5617 		}
5618 	}
5619 
5620 	for (f = 0; f < nr_frags; f++) {
5621 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[f];
5622 
5623 		len = skb_frag_size(frag);
5624 		offset = 0;
5625 
5626 		while (len) {
5627 			i++;
5628 			if (i == tx_ring->count)
5629 				i = 0;
5630 
5631 			buffer_info = &tx_ring->buffer_info[i];
5632 			size = min(len, max_per_txd);
5633 
5634 			buffer_info->length = size;
5635 			buffer_info->time_stamp = jiffies;
5636 			buffer_info->next_to_watch = i;
5637 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5638 							    offset, size,
5639 							    DMA_TO_DEVICE);
5640 			buffer_info->mapped_as_page = true;
5641 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5642 				goto dma_error;
5643 
5644 			len -= size;
5645 			offset += size;
5646 			count++;
5647 		}
5648 	}
5649 
5650 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5651 	/* multiply data chunks by size of headers */
5652 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5653 
5654 	tx_ring->buffer_info[i].skb = skb;
5655 	tx_ring->buffer_info[i].segs = segs;
5656 	tx_ring->buffer_info[i].bytecount = bytecount;
5657 	tx_ring->buffer_info[first].next_to_watch = i;
5658 
5659 	return count;
5660 
5661 dma_error:
5662 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5663 	buffer_info->dma = 0;
5664 	if (count)
5665 		count--;
5666 
5667 	while (count--) {
5668 		if (i == 0)
5669 			i += tx_ring->count;
5670 		i--;
5671 		buffer_info = &tx_ring->buffer_info[i];
5672 		e1000_put_txbuf(tx_ring, buffer_info, true);
5673 	}
5674 
5675 	return 0;
5676 }
5677 
e1000_tx_queue(struct e1000_ring * tx_ring,int tx_flags,int count)5678 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5679 {
5680 	struct e1000_adapter *adapter = tx_ring->adapter;
5681 	struct e1000_tx_desc *tx_desc = NULL;
5682 	struct e1000_buffer *buffer_info;
5683 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5684 	unsigned int i;
5685 
5686 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5687 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5688 		    E1000_TXD_CMD_TSE;
5689 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5690 
5691 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5692 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5693 	}
5694 
5695 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5696 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5697 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5698 	}
5699 
5700 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5701 		txd_lower |= E1000_TXD_CMD_VLE;
5702 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5703 	}
5704 
5705 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5706 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5707 
5708 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5709 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5710 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5711 	}
5712 
5713 	i = tx_ring->next_to_use;
5714 
5715 	do {
5716 		buffer_info = &tx_ring->buffer_info[i];
5717 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5718 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5719 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5720 						  buffer_info->length);
5721 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5722 
5723 		i++;
5724 		if (i == tx_ring->count)
5725 			i = 0;
5726 	} while (--count > 0);
5727 
5728 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5729 
5730 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5731 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5732 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5733 
5734 	/* Force memory writes to complete before letting h/w
5735 	 * know there are new descriptors to fetch.  (Only
5736 	 * applicable for weak-ordered memory model archs,
5737 	 * such as IA-64).
5738 	 */
5739 	wmb();
5740 
5741 	tx_ring->next_to_use = i;
5742 }
5743 
5744 #define MINIMUM_DHCP_PACKET_SIZE 282
e1000_transfer_dhcp_info(struct e1000_adapter * adapter,struct sk_buff * skb)5745 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5746 				    struct sk_buff *skb)
5747 {
5748 	struct e1000_hw *hw = &adapter->hw;
5749 	u16 length, offset;
5750 
5751 	if (skb_vlan_tag_present(skb) &&
5752 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5753 	      (adapter->hw.mng_cookie.status &
5754 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5755 		return 0;
5756 
5757 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5758 		return 0;
5759 
5760 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5761 		return 0;
5762 
5763 	{
5764 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5765 		struct udphdr *udp;
5766 
5767 		if (ip->protocol != IPPROTO_UDP)
5768 			return 0;
5769 
5770 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5771 		if (ntohs(udp->dest) != 67)
5772 			return 0;
5773 
5774 		offset = (u8 *)udp + 8 - skb->data;
5775 		length = skb->len - offset;
5776 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5777 	}
5778 
5779 	return 0;
5780 }
5781 
__e1000_maybe_stop_tx(struct e1000_ring * tx_ring,int size)5782 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5783 {
5784 	struct e1000_adapter *adapter = tx_ring->adapter;
5785 
5786 	netif_stop_queue(adapter->netdev);
5787 	/* Herbert's original patch had:
5788 	 *  smp_mb__after_netif_stop_queue();
5789 	 * but since that doesn't exist yet, just open code it.
5790 	 */
5791 	smp_mb();
5792 
5793 	/* We need to check again in a case another CPU has just
5794 	 * made room available.
5795 	 */
5796 	if (e1000_desc_unused(tx_ring) < size)
5797 		return -EBUSY;
5798 
5799 	/* A reprieve! */
5800 	netif_start_queue(adapter->netdev);
5801 	++adapter->restart_queue;
5802 	return 0;
5803 }
5804 
e1000_maybe_stop_tx(struct e1000_ring * tx_ring,int size)5805 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5806 {
5807 	BUG_ON(size > tx_ring->count);
5808 
5809 	if (e1000_desc_unused(tx_ring) >= size)
5810 		return 0;
5811 	return __e1000_maybe_stop_tx(tx_ring, size);
5812 }
5813 
e1000_xmit_frame(struct sk_buff * skb,struct net_device * netdev)5814 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5815 				    struct net_device *netdev)
5816 {
5817 	struct e1000_adapter *adapter = netdev_priv(netdev);
5818 	struct e1000_ring *tx_ring = adapter->tx_ring;
5819 	unsigned int first;
5820 	unsigned int tx_flags = 0;
5821 	unsigned int len = skb_headlen(skb);
5822 	unsigned int nr_frags;
5823 	unsigned int mss;
5824 	int count = 0;
5825 	int tso;
5826 	unsigned int f;
5827 	__be16 protocol = vlan_get_protocol(skb);
5828 
5829 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5830 		dev_kfree_skb_any(skb);
5831 		return NETDEV_TX_OK;
5832 	}
5833 
5834 	if (skb->len <= 0) {
5835 		dev_kfree_skb_any(skb);
5836 		return NETDEV_TX_OK;
5837 	}
5838 
5839 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5840 	 * pad skb in order to meet this minimum size requirement
5841 	 */
5842 	if (skb_put_padto(skb, 17))
5843 		return NETDEV_TX_OK;
5844 
5845 	mss = skb_shinfo(skb)->gso_size;
5846 	if (mss) {
5847 		u8 hdr_len;
5848 
5849 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5850 		 * points to just header, pull a few bytes of payload from
5851 		 * frags into skb->data
5852 		 */
5853 		hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5854 		/* we do this workaround for ES2LAN, but it is un-necessary,
5855 		 * avoiding it could save a lot of cycles
5856 		 */
5857 		if (skb->data_len && (hdr_len == len)) {
5858 			unsigned int pull_size;
5859 
5860 			pull_size = min_t(unsigned int, 4, skb->data_len);
5861 			if (!__pskb_pull_tail(skb, pull_size)) {
5862 				e_err("__pskb_pull_tail failed.\n");
5863 				dev_kfree_skb_any(skb);
5864 				return NETDEV_TX_OK;
5865 			}
5866 			len = skb_headlen(skb);
5867 		}
5868 	}
5869 
5870 	/* reserve a descriptor for the offload context */
5871 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5872 		count++;
5873 	count++;
5874 
5875 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5876 
5877 	nr_frags = skb_shinfo(skb)->nr_frags;
5878 	for (f = 0; f < nr_frags; f++)
5879 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5880 				      adapter->tx_fifo_limit);
5881 
5882 	if (adapter->hw.mac.tx_pkt_filtering)
5883 		e1000_transfer_dhcp_info(adapter, skb);
5884 
5885 	/* need: count + 2 desc gap to keep tail from touching
5886 	 * head, otherwise try next time
5887 	 */
5888 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5889 		return NETDEV_TX_BUSY;
5890 
5891 	if (skb_vlan_tag_present(skb)) {
5892 		tx_flags |= E1000_TX_FLAGS_VLAN;
5893 		tx_flags |= (skb_vlan_tag_get(skb) <<
5894 			     E1000_TX_FLAGS_VLAN_SHIFT);
5895 	}
5896 
5897 	first = tx_ring->next_to_use;
5898 
5899 	tso = e1000_tso(tx_ring, skb, protocol);
5900 	if (tso < 0) {
5901 		dev_kfree_skb_any(skb);
5902 		return NETDEV_TX_OK;
5903 	}
5904 
5905 	if (tso)
5906 		tx_flags |= E1000_TX_FLAGS_TSO;
5907 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5908 		tx_flags |= E1000_TX_FLAGS_CSUM;
5909 
5910 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5911 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5912 	 * no longer assume, we must.
5913 	 */
5914 	if (protocol == htons(ETH_P_IP))
5915 		tx_flags |= E1000_TX_FLAGS_IPV4;
5916 
5917 	if (unlikely(skb->no_fcs))
5918 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5919 
5920 	/* if count is 0 then mapping error has occurred */
5921 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5922 			     nr_frags);
5923 	if (count) {
5924 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5925 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5926 			if (!adapter->tx_hwtstamp_skb) {
5927 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5928 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5929 				adapter->tx_hwtstamp_skb = skb_get(skb);
5930 				adapter->tx_hwtstamp_start = jiffies;
5931 				schedule_work(&adapter->tx_hwtstamp_work);
5932 			} else {
5933 				adapter->tx_hwtstamp_skipped++;
5934 			}
5935 		}
5936 
5937 		skb_tx_timestamp(skb);
5938 
5939 		netdev_sent_queue(netdev, skb->len);
5940 		e1000_tx_queue(tx_ring, tx_flags, count);
5941 		/* Make sure there is space in the ring for the next send. */
5942 		e1000_maybe_stop_tx(tx_ring,
5943 				    (MAX_SKB_FRAGS *
5944 				     DIV_ROUND_UP(PAGE_SIZE,
5945 						  adapter->tx_fifo_limit) + 2));
5946 
5947 		if (!netdev_xmit_more() ||
5948 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5949 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5950 				e1000e_update_tdt_wa(tx_ring,
5951 						     tx_ring->next_to_use);
5952 			else
5953 				writel(tx_ring->next_to_use, tx_ring->tail);
5954 		}
5955 	} else {
5956 		dev_kfree_skb_any(skb);
5957 		tx_ring->buffer_info[first].time_stamp = 0;
5958 		tx_ring->next_to_use = first;
5959 	}
5960 
5961 	return NETDEV_TX_OK;
5962 }
5963 
5964 /**
5965  * e1000_tx_timeout - Respond to a Tx Hang
5966  * @netdev: network interface device structure
5967  * @txqueue: index of the hung queue (unused)
5968  **/
e1000_tx_timeout(struct net_device * netdev,unsigned int __always_unused txqueue)5969 static void e1000_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
5970 {
5971 	struct e1000_adapter *adapter = netdev_priv(netdev);
5972 
5973 	/* Do the reset outside of interrupt context */
5974 	adapter->tx_timeout_count++;
5975 	schedule_work(&adapter->reset_task);
5976 }
5977 
e1000_reset_task(struct work_struct * work)5978 static void e1000_reset_task(struct work_struct *work)
5979 {
5980 	struct e1000_adapter *adapter;
5981 	adapter = container_of(work, struct e1000_adapter, reset_task);
5982 
5983 	rtnl_lock();
5984 	/* don't run the task if already down */
5985 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5986 		rtnl_unlock();
5987 		return;
5988 	}
5989 
5990 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5991 		e1000e_dump(adapter);
5992 		e_err("Reset adapter unexpectedly\n");
5993 	}
5994 	e1000e_reinit_locked(adapter);
5995 	rtnl_unlock();
5996 }
5997 
5998 /**
5999  * e1000e_get_stats64 - Get System Network Statistics
6000  * @netdev: network interface device structure
6001  * @stats: rtnl_link_stats64 pointer
6002  *
6003  * Returns the address of the device statistics structure.
6004  **/
e1000e_get_stats64(struct net_device * netdev,struct rtnl_link_stats64 * stats)6005 void e1000e_get_stats64(struct net_device *netdev,
6006 			struct rtnl_link_stats64 *stats)
6007 {
6008 	struct e1000_adapter *adapter = netdev_priv(netdev);
6009 
6010 	spin_lock(&adapter->stats64_lock);
6011 	e1000e_update_stats(adapter);
6012 	/* Fill out the OS statistics structure */
6013 	stats->rx_bytes = adapter->stats.gorc;
6014 	stats->rx_packets = adapter->stats.gprc;
6015 	stats->tx_bytes = adapter->stats.gotc;
6016 	stats->tx_packets = adapter->stats.gptc;
6017 	stats->multicast = adapter->stats.mprc;
6018 	stats->collisions = adapter->stats.colc;
6019 
6020 	/* Rx Errors */
6021 
6022 	/* RLEC on some newer hardware can be incorrect so build
6023 	 * our own version based on RUC and ROC
6024 	 */
6025 	stats->rx_errors = adapter->stats.rxerrc +
6026 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
6027 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
6028 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
6029 	stats->rx_crc_errors = adapter->stats.crcerrs;
6030 	stats->rx_frame_errors = adapter->stats.algnerrc;
6031 	stats->rx_missed_errors = adapter->stats.mpc;
6032 
6033 	/* Tx Errors */
6034 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
6035 	stats->tx_aborted_errors = adapter->stats.ecol;
6036 	stats->tx_window_errors = adapter->stats.latecol;
6037 	stats->tx_carrier_errors = adapter->stats.tncrs;
6038 
6039 	/* Tx Dropped needs to be maintained elsewhere */
6040 
6041 	spin_unlock(&adapter->stats64_lock);
6042 }
6043 
6044 /**
6045  * e1000_change_mtu - Change the Maximum Transfer Unit
6046  * @netdev: network interface device structure
6047  * @new_mtu: new value for maximum frame size
6048  *
6049  * Returns 0 on success, negative on failure
6050  **/
e1000_change_mtu(struct net_device * netdev,int new_mtu)6051 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6052 {
6053 	struct e1000_adapter *adapter = netdev_priv(netdev);
6054 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6055 
6056 	/* Jumbo frame support */
6057 	if ((new_mtu > ETH_DATA_LEN) &&
6058 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6059 		e_err("Jumbo Frames not supported.\n");
6060 		return -EINVAL;
6061 	}
6062 
6063 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6064 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6065 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6066 	    (new_mtu > ETH_DATA_LEN)) {
6067 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6068 		return -EINVAL;
6069 	}
6070 
6071 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6072 		usleep_range(1000, 1100);
6073 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6074 	adapter->max_frame_size = max_frame;
6075 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6076 		   netdev->mtu, new_mtu);
6077 	netdev->mtu = new_mtu;
6078 
6079 	pm_runtime_get_sync(netdev->dev.parent);
6080 
6081 	if (netif_running(netdev))
6082 		e1000e_down(adapter, true);
6083 
6084 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6085 	 * means we reserve 2 more, this pushes us to allocate from the next
6086 	 * larger slab size.
6087 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6088 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6089 	 * fragmented skbs
6090 	 */
6091 
6092 	if (max_frame <= 2048)
6093 		adapter->rx_buffer_len = 2048;
6094 	else
6095 		adapter->rx_buffer_len = 4096;
6096 
6097 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6098 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6099 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6100 
6101 	if (netif_running(netdev))
6102 		e1000e_up(adapter);
6103 	else
6104 		e1000e_reset(adapter);
6105 
6106 	pm_runtime_put_sync(netdev->dev.parent);
6107 
6108 	clear_bit(__E1000_RESETTING, &adapter->state);
6109 
6110 	return 0;
6111 }
6112 
e1000_mii_ioctl(struct net_device * netdev,struct ifreq * ifr,int cmd)6113 static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
6114 			   int cmd)
6115 {
6116 	struct e1000_adapter *adapter = netdev_priv(netdev);
6117 	struct mii_ioctl_data *data = if_mii(ifr);
6118 
6119 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6120 		return -EOPNOTSUPP;
6121 
6122 	switch (cmd) {
6123 	case SIOCGMIIPHY:
6124 		data->phy_id = adapter->hw.phy.addr;
6125 		break;
6126 	case SIOCGMIIREG:
6127 		e1000_phy_read_status(adapter);
6128 
6129 		switch (data->reg_num & 0x1F) {
6130 		case MII_BMCR:
6131 			data->val_out = adapter->phy_regs.bmcr;
6132 			break;
6133 		case MII_BMSR:
6134 			data->val_out = adapter->phy_regs.bmsr;
6135 			break;
6136 		case MII_PHYSID1:
6137 			data->val_out = (adapter->hw.phy.id >> 16);
6138 			break;
6139 		case MII_PHYSID2:
6140 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6141 			break;
6142 		case MII_ADVERTISE:
6143 			data->val_out = adapter->phy_regs.advertise;
6144 			break;
6145 		case MII_LPA:
6146 			data->val_out = adapter->phy_regs.lpa;
6147 			break;
6148 		case MII_EXPANSION:
6149 			data->val_out = adapter->phy_regs.expansion;
6150 			break;
6151 		case MII_CTRL1000:
6152 			data->val_out = adapter->phy_regs.ctrl1000;
6153 			break;
6154 		case MII_STAT1000:
6155 			data->val_out = adapter->phy_regs.stat1000;
6156 			break;
6157 		case MII_ESTATUS:
6158 			data->val_out = adapter->phy_regs.estatus;
6159 			break;
6160 		default:
6161 			return -EIO;
6162 		}
6163 		break;
6164 	case SIOCSMIIREG:
6165 	default:
6166 		return -EOPNOTSUPP;
6167 	}
6168 	return 0;
6169 }
6170 
6171 /**
6172  * e1000e_hwtstamp_set - control hardware time stamping
6173  * @netdev: network interface device structure
6174  * @ifr: interface request
6175  *
6176  * Outgoing time stamping can be enabled and disabled. Play nice and
6177  * disable it when requested, although it shouldn't cause any overhead
6178  * when no packet needs it. At most one packet in the queue may be
6179  * marked for time stamping, otherwise it would be impossible to tell
6180  * for sure to which packet the hardware time stamp belongs.
6181  *
6182  * Incoming time stamping has to be configured via the hardware filters.
6183  * Not all combinations are supported, in particular event type has to be
6184  * specified. Matching the kind of event packet is not supported, with the
6185  * exception of "all V2 events regardless of level 2 or 4".
6186  **/
e1000e_hwtstamp_set(struct net_device * netdev,struct ifreq * ifr)6187 static int e1000e_hwtstamp_set(struct net_device *netdev, struct ifreq *ifr)
6188 {
6189 	struct e1000_adapter *adapter = netdev_priv(netdev);
6190 	struct hwtstamp_config config;
6191 	int ret_val;
6192 
6193 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
6194 		return -EFAULT;
6195 
6196 	ret_val = e1000e_config_hwtstamp(adapter, &config);
6197 	if (ret_val)
6198 		return ret_val;
6199 
6200 	switch (config.rx_filter) {
6201 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6202 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6203 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6204 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6205 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6206 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6207 		/* With V2 type filters which specify a Sync or Delay Request,
6208 		 * Path Delay Request/Response messages are also time stamped
6209 		 * by hardware so notify the caller the requested packets plus
6210 		 * some others are time stamped.
6211 		 */
6212 		config.rx_filter = HWTSTAMP_FILTER_SOME;
6213 		break;
6214 	default:
6215 		break;
6216 	}
6217 
6218 	return copy_to_user(ifr->ifr_data, &config,
6219 			    sizeof(config)) ? -EFAULT : 0;
6220 }
6221 
e1000e_hwtstamp_get(struct net_device * netdev,struct ifreq * ifr)6222 static int e1000e_hwtstamp_get(struct net_device *netdev, struct ifreq *ifr)
6223 {
6224 	struct e1000_adapter *adapter = netdev_priv(netdev);
6225 
6226 	return copy_to_user(ifr->ifr_data, &adapter->hwtstamp_config,
6227 			    sizeof(adapter->hwtstamp_config)) ? -EFAULT : 0;
6228 }
6229 
e1000_ioctl(struct net_device * netdev,struct ifreq * ifr,int cmd)6230 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6231 {
6232 	switch (cmd) {
6233 	case SIOCGMIIPHY:
6234 	case SIOCGMIIREG:
6235 	case SIOCSMIIREG:
6236 		return e1000_mii_ioctl(netdev, ifr, cmd);
6237 	case SIOCSHWTSTAMP:
6238 		return e1000e_hwtstamp_set(netdev, ifr);
6239 	case SIOCGHWTSTAMP:
6240 		return e1000e_hwtstamp_get(netdev, ifr);
6241 	default:
6242 		return -EOPNOTSUPP;
6243 	}
6244 }
6245 
e1000_init_phy_wakeup(struct e1000_adapter * adapter,u32 wufc)6246 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6247 {
6248 	struct e1000_hw *hw = &adapter->hw;
6249 	u32 i, mac_reg, wuc;
6250 	u16 phy_reg, wuc_enable;
6251 	int retval;
6252 
6253 	/* copy MAC RARs to PHY RARs */
6254 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6255 
6256 	retval = hw->phy.ops.acquire(hw);
6257 	if (retval) {
6258 		e_err("Could not acquire PHY\n");
6259 		return retval;
6260 	}
6261 
6262 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6263 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6264 	if (retval)
6265 		goto release;
6266 
6267 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6268 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6269 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6270 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6271 					   (u16)(mac_reg & 0xFFFF));
6272 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6273 					   (u16)((mac_reg >> 16) & 0xFFFF));
6274 	}
6275 
6276 	/* configure PHY Rx Control register */
6277 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6278 	mac_reg = er32(RCTL);
6279 	if (mac_reg & E1000_RCTL_UPE)
6280 		phy_reg |= BM_RCTL_UPE;
6281 	if (mac_reg & E1000_RCTL_MPE)
6282 		phy_reg |= BM_RCTL_MPE;
6283 	phy_reg &= ~(BM_RCTL_MO_MASK);
6284 	if (mac_reg & E1000_RCTL_MO_3)
6285 		phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
6286 			    << BM_RCTL_MO_SHIFT);
6287 	if (mac_reg & E1000_RCTL_BAM)
6288 		phy_reg |= BM_RCTL_BAM;
6289 	if (mac_reg & E1000_RCTL_PMCF)
6290 		phy_reg |= BM_RCTL_PMCF;
6291 	mac_reg = er32(CTRL);
6292 	if (mac_reg & E1000_CTRL_RFCE)
6293 		phy_reg |= BM_RCTL_RFCE;
6294 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6295 
6296 	wuc = E1000_WUC_PME_EN;
6297 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6298 		wuc |= E1000_WUC_APME;
6299 
6300 	/* enable PHY wakeup in MAC register */
6301 	ew32(WUFC, wufc);
6302 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6303 		   E1000_WUC_PME_STATUS | wuc));
6304 
6305 	/* configure and enable PHY wakeup in PHY registers */
6306 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6307 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6308 
6309 	/* activate PHY wakeup */
6310 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6311 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6312 	if (retval)
6313 		e_err("Could not set PHY Host Wakeup bit\n");
6314 release:
6315 	hw->phy.ops.release(hw);
6316 
6317 	return retval;
6318 }
6319 
e1000e_flush_lpic(struct pci_dev * pdev)6320 static void e1000e_flush_lpic(struct pci_dev *pdev)
6321 {
6322 	struct net_device *netdev = pci_get_drvdata(pdev);
6323 	struct e1000_adapter *adapter = netdev_priv(netdev);
6324 	struct e1000_hw *hw = &adapter->hw;
6325 	u32 ret_val;
6326 
6327 	pm_runtime_get_sync(netdev->dev.parent);
6328 
6329 	ret_val = hw->phy.ops.acquire(hw);
6330 	if (ret_val)
6331 		goto fl_out;
6332 
6333 	pr_info("EEE TX LPI TIMER: %08X\n",
6334 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6335 
6336 	hw->phy.ops.release(hw);
6337 
6338 fl_out:
6339 	pm_runtime_put_sync(netdev->dev.parent);
6340 }
6341 
6342 /* S0ix implementation */
e1000e_s0ix_entry_flow(struct e1000_adapter * adapter)6343 static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6344 {
6345 	struct e1000_hw *hw = &adapter->hw;
6346 	u32 mac_data;
6347 	u16 phy_data;
6348 
6349 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
6350 		/* Request ME configure the device for S0ix */
6351 		mac_data = er32(H2ME);
6352 		mac_data |= E1000_H2ME_START_DPG;
6353 		mac_data &= ~E1000_H2ME_EXIT_DPG;
6354 		ew32(H2ME, mac_data);
6355 	} else {
6356 		/* Request driver configure the device to S0ix */
6357 		/* Disable the periodic inband message,
6358 		 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6359 		 */
6360 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6361 		phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6362 		phy_data |= BIT(10);
6363 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6364 
6365 		/* Make sure we don't exit K1 every time a new packet arrives
6366 		 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6367 		 */
6368 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6369 		phy_data |= BIT(5);
6370 		e1e_wphy(hw, I217_CGFREG, phy_data);
6371 
6372 		/* Change the MAC/PHY interface to SMBus
6373 		 * Force the SMBus in PHY page769_23[0] = 1
6374 		 * Force the SMBus in MAC CTRL_EXT[11] = 1
6375 		 */
6376 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6377 		phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6378 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6379 		mac_data = er32(CTRL_EXT);
6380 		mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6381 		ew32(CTRL_EXT, mac_data);
6382 
6383 		/* DFT control: PHY bit: page769_20[0] = 1
6384 		 * page769_20[7] - PHY PLL stop
6385 		 * page769_20[8] - PHY go to the electrical idle
6386 		 * page769_20[9] - PHY serdes disable
6387 		 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6388 		 */
6389 		e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6390 		phy_data |= BIT(0);
6391 		phy_data |= BIT(7);
6392 		phy_data |= BIT(8);
6393 		phy_data |= BIT(9);
6394 		e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6395 
6396 		mac_data = er32(EXTCNF_CTRL);
6397 		mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6398 		ew32(EXTCNF_CTRL, mac_data);
6399 
6400 		/* Enable the Dynamic Power Gating in the MAC */
6401 		mac_data = er32(FEXTNVM7);
6402 		mac_data |= BIT(22);
6403 		ew32(FEXTNVM7, mac_data);
6404 
6405 		/* Disable disconnected cable conditioning for Power Gating */
6406 		mac_data = er32(DPGFR);
6407 		mac_data |= BIT(2);
6408 		ew32(DPGFR, mac_data);
6409 
6410 		/* Don't wake from dynamic Power Gating with clock request */
6411 		mac_data = er32(FEXTNVM12);
6412 		mac_data |= BIT(12);
6413 		ew32(FEXTNVM12, mac_data);
6414 
6415 		/* Ungate PGCB clock */
6416 		mac_data = er32(FEXTNVM9);
6417 		mac_data &= ~BIT(28);
6418 		ew32(FEXTNVM9, mac_data);
6419 
6420 		/* Enable K1 off to enable mPHY Power Gating */
6421 		mac_data = er32(FEXTNVM6);
6422 		mac_data |= BIT(31);
6423 		ew32(FEXTNVM6, mac_data);
6424 
6425 		/* Enable mPHY power gating for any link and speed */
6426 		mac_data = er32(FEXTNVM8);
6427 		mac_data |= BIT(9);
6428 		ew32(FEXTNVM8, mac_data);
6429 
6430 		/* Enable the Dynamic Clock Gating in the DMA and MAC */
6431 		mac_data = er32(CTRL_EXT);
6432 		mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6433 		ew32(CTRL_EXT, mac_data);
6434 
6435 		/* No MAC DPG gating SLP_S0 in modern standby
6436 		 * Switch the logic of the lanphypc to use PMC counter
6437 		 */
6438 		mac_data = er32(FEXTNVM5);
6439 		mac_data |= BIT(7);
6440 		ew32(FEXTNVM5, mac_data);
6441 	}
6442 
6443 	/* Disable the time synchronization clock */
6444 	mac_data = er32(FEXTNVM7);
6445 	mac_data |= BIT(31);
6446 	mac_data &= ~BIT(0);
6447 	ew32(FEXTNVM7, mac_data);
6448 
6449 	/* Dynamic Power Gating Enable */
6450 	mac_data = er32(CTRL_EXT);
6451 	mac_data |= BIT(3);
6452 	ew32(CTRL_EXT, mac_data);
6453 
6454 	/* Check MAC Tx/Rx packet buffer pointers.
6455 	 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6456 	 * pending traffic indication that would prevent power gating.
6457 	 */
6458 	mac_data = er32(TDFH);
6459 	if (mac_data)
6460 		ew32(TDFH, 0);
6461 	mac_data = er32(TDFT);
6462 	if (mac_data)
6463 		ew32(TDFT, 0);
6464 	mac_data = er32(TDFHS);
6465 	if (mac_data)
6466 		ew32(TDFHS, 0);
6467 	mac_data = er32(TDFTS);
6468 	if (mac_data)
6469 		ew32(TDFTS, 0);
6470 	mac_data = er32(TDFPC);
6471 	if (mac_data)
6472 		ew32(TDFPC, 0);
6473 	mac_data = er32(RDFH);
6474 	if (mac_data)
6475 		ew32(RDFH, 0);
6476 	mac_data = er32(RDFT);
6477 	if (mac_data)
6478 		ew32(RDFT, 0);
6479 	mac_data = er32(RDFHS);
6480 	if (mac_data)
6481 		ew32(RDFHS, 0);
6482 	mac_data = er32(RDFTS);
6483 	if (mac_data)
6484 		ew32(RDFTS, 0);
6485 	mac_data = er32(RDFPC);
6486 	if (mac_data)
6487 		ew32(RDFPC, 0);
6488 }
6489 
e1000e_s0ix_exit_flow(struct e1000_adapter * adapter)6490 static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6491 {
6492 	struct e1000_hw *hw = &adapter->hw;
6493 	bool firmware_bug = false;
6494 	u32 mac_data;
6495 	u16 phy_data;
6496 	u32 i = 0;
6497 
6498 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
6499 		/* Request ME unconfigure the device from S0ix */
6500 		mac_data = er32(H2ME);
6501 		mac_data &= ~E1000_H2ME_START_DPG;
6502 		mac_data |= E1000_H2ME_EXIT_DPG;
6503 		ew32(H2ME, mac_data);
6504 
6505 		/* Poll up to 2.5 seconds for ME to unconfigure DPG.
6506 		 * If this takes more than 1 second, show a warning indicating a
6507 		 * firmware bug
6508 		 */
6509 		while (!(er32(EXFWSM) & E1000_EXFWSM_DPG_EXIT_DONE)) {
6510 			if (i > 100 && !firmware_bug)
6511 				firmware_bug = true;
6512 
6513 			if (i++ == 250) {
6514 				e_dbg("Timeout (firmware bug): %d msec\n",
6515 				      i * 10);
6516 				break;
6517 			}
6518 
6519 			usleep_range(10000, 11000);
6520 		}
6521 		if (firmware_bug)
6522 			e_warn("DPG_EXIT_DONE took %d msec. This is a firmware bug\n",
6523 			       i * 10);
6524 		else
6525 			e_dbg("DPG_EXIT_DONE cleared after %d msec\n", i * 10);
6526 	} else {
6527 		/* Request driver unconfigure the device from S0ix */
6528 
6529 		/* Disable the Dynamic Power Gating in the MAC */
6530 		mac_data = er32(FEXTNVM7);
6531 		mac_data &= 0xFFBFFFFF;
6532 		ew32(FEXTNVM7, mac_data);
6533 
6534 		/* Disable mPHY power gating for any link and speed */
6535 		mac_data = er32(FEXTNVM8);
6536 		mac_data &= ~BIT(9);
6537 		ew32(FEXTNVM8, mac_data);
6538 
6539 		/* Disable K1 off */
6540 		mac_data = er32(FEXTNVM6);
6541 		mac_data &= ~BIT(31);
6542 		ew32(FEXTNVM6, mac_data);
6543 
6544 		/* Disable Ungate PGCB clock */
6545 		mac_data = er32(FEXTNVM9);
6546 		mac_data |= BIT(28);
6547 		ew32(FEXTNVM9, mac_data);
6548 
6549 		/* Cancel not waking from dynamic
6550 		 * Power Gating with clock request
6551 		 */
6552 		mac_data = er32(FEXTNVM12);
6553 		mac_data &= ~BIT(12);
6554 		ew32(FEXTNVM12, mac_data);
6555 
6556 		/* Cancel disable disconnected cable conditioning
6557 		 * for Power Gating
6558 		 */
6559 		mac_data = er32(DPGFR);
6560 		mac_data &= ~BIT(2);
6561 		ew32(DPGFR, mac_data);
6562 
6563 		/* Disable the Dynamic Clock Gating in the DMA and MAC */
6564 		mac_data = er32(CTRL_EXT);
6565 		mac_data &= 0xFFF7FFFF;
6566 		ew32(CTRL_EXT, mac_data);
6567 
6568 		/* Revert the lanphypc logic to use the internal Gbe counter
6569 		 * and not the PMC counter
6570 		 */
6571 		mac_data = er32(FEXTNVM5);
6572 		mac_data &= 0xFFFFFF7F;
6573 		ew32(FEXTNVM5, mac_data);
6574 
6575 		/* Enable the periodic inband message,
6576 		 * Request PCIe clock in K1 page770_17[10:9] =01b
6577 		 */
6578 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6579 		phy_data &= 0xFBFF;
6580 		phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6581 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6582 
6583 		/* Return back configuration
6584 		 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6585 		 */
6586 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6587 		phy_data &= 0xFFDF;
6588 		e1e_wphy(hw, I217_CGFREG, phy_data);
6589 
6590 		/* Change the MAC/PHY interface to Kumeran
6591 		 * Unforce the SMBus in PHY page769_23[0] = 0
6592 		 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6593 		 */
6594 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6595 		phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6596 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6597 		mac_data = er32(CTRL_EXT);
6598 		mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6599 		ew32(CTRL_EXT, mac_data);
6600 	}
6601 
6602 	/* Disable Dynamic Power Gating */
6603 	mac_data = er32(CTRL_EXT);
6604 	mac_data &= 0xFFFFFFF7;
6605 	ew32(CTRL_EXT, mac_data);
6606 
6607 	/* Enable the time synchronization clock */
6608 	mac_data = er32(FEXTNVM7);
6609 	mac_data &= ~BIT(31);
6610 	mac_data |= BIT(0);
6611 	ew32(FEXTNVM7, mac_data);
6612 }
6613 
e1000e_pm_freeze(struct device * dev)6614 static int e1000e_pm_freeze(struct device *dev)
6615 {
6616 	struct net_device *netdev = dev_get_drvdata(dev);
6617 	struct e1000_adapter *adapter = netdev_priv(netdev);
6618 	bool present;
6619 
6620 	rtnl_lock();
6621 
6622 	present = netif_device_present(netdev);
6623 	netif_device_detach(netdev);
6624 
6625 	if (present && netif_running(netdev)) {
6626 		int count = E1000_CHECK_RESET_COUNT;
6627 
6628 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6629 			usleep_range(10000, 11000);
6630 
6631 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6632 
6633 		/* Quiesce the device without resetting the hardware */
6634 		e1000e_down(adapter, false);
6635 		e1000_free_irq(adapter);
6636 	}
6637 	rtnl_unlock();
6638 
6639 	e1000e_reset_interrupt_capability(adapter);
6640 
6641 	/* Allow time for pending master requests to run */
6642 	e1000e_disable_pcie_master(&adapter->hw);
6643 
6644 	return 0;
6645 }
6646 
__e1000_shutdown(struct pci_dev * pdev,bool runtime)6647 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6648 {
6649 	struct net_device *netdev = pci_get_drvdata(pdev);
6650 	struct e1000_adapter *adapter = netdev_priv(netdev);
6651 	struct e1000_hw *hw = &adapter->hw;
6652 	u32 ctrl, ctrl_ext, rctl, status, wufc;
6653 	int retval = 0;
6654 
6655 	/* Runtime suspend should only enable wakeup for link changes */
6656 	if (runtime)
6657 		wufc = E1000_WUFC_LNKC;
6658 	else if (device_may_wakeup(&pdev->dev))
6659 		wufc = adapter->wol;
6660 	else
6661 		wufc = 0;
6662 
6663 	status = er32(STATUS);
6664 	if (status & E1000_STATUS_LU)
6665 		wufc &= ~E1000_WUFC_LNKC;
6666 
6667 	if (wufc) {
6668 		e1000_setup_rctl(adapter);
6669 		e1000e_set_rx_mode(netdev);
6670 
6671 		/* turn on all-multi mode if wake on multicast is enabled */
6672 		if (wufc & E1000_WUFC_MC) {
6673 			rctl = er32(RCTL);
6674 			rctl |= E1000_RCTL_MPE;
6675 			ew32(RCTL, rctl);
6676 		}
6677 
6678 		ctrl = er32(CTRL);
6679 		ctrl |= E1000_CTRL_ADVD3WUC;
6680 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6681 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6682 		ew32(CTRL, ctrl);
6683 
6684 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6685 		    adapter->hw.phy.media_type ==
6686 		    e1000_media_type_internal_serdes) {
6687 			/* keep the laser running in D3 */
6688 			ctrl_ext = er32(CTRL_EXT);
6689 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6690 			ew32(CTRL_EXT, ctrl_ext);
6691 		}
6692 
6693 		if (!runtime)
6694 			e1000e_power_up_phy(adapter);
6695 
6696 		if (adapter->flags & FLAG_IS_ICH)
6697 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6698 
6699 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6700 			/* enable wakeup by the PHY */
6701 			retval = e1000_init_phy_wakeup(adapter, wufc);
6702 			if (retval)
6703 				return retval;
6704 		} else {
6705 			/* enable wakeup by the MAC */
6706 			ew32(WUFC, wufc);
6707 			ew32(WUC, E1000_WUC_PME_EN);
6708 		}
6709 	} else {
6710 		ew32(WUC, 0);
6711 		ew32(WUFC, 0);
6712 
6713 		e1000_power_down_phy(adapter);
6714 	}
6715 
6716 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6717 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6718 	} else if (hw->mac.type >= e1000_pch_lpt) {
6719 		if (wufc && !(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC)))
6720 			/* ULP does not support wake from unicast, multicast
6721 			 * or broadcast.
6722 			 */
6723 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6724 
6725 		if (retval)
6726 			return retval;
6727 	}
6728 
6729 	/* Ensure that the appropriate bits are set in LPI_CTRL
6730 	 * for EEE in Sx
6731 	 */
6732 	if ((hw->phy.type >= e1000_phy_i217) &&
6733 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6734 		u16 lpi_ctrl = 0;
6735 
6736 		retval = hw->phy.ops.acquire(hw);
6737 		if (!retval) {
6738 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6739 						 &lpi_ctrl);
6740 			if (!retval) {
6741 				if (adapter->eee_advert &
6742 				    hw->dev_spec.ich8lan.eee_lp_ability &
6743 				    I82579_EEE_100_SUPPORTED)
6744 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6745 				if (adapter->eee_advert &
6746 				    hw->dev_spec.ich8lan.eee_lp_ability &
6747 				    I82579_EEE_1000_SUPPORTED)
6748 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6749 
6750 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6751 							 lpi_ctrl);
6752 			}
6753 		}
6754 		hw->phy.ops.release(hw);
6755 	}
6756 
6757 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6758 	 * would have already happened in close and is redundant.
6759 	 */
6760 	e1000e_release_hw_control(adapter);
6761 
6762 	pci_clear_master(pdev);
6763 
6764 	/* The pci-e switch on some quad port adapters will report a
6765 	 * correctable error when the MAC transitions from D0 to D3.  To
6766 	 * prevent this we need to mask off the correctable errors on the
6767 	 * downstream port of the pci-e switch.
6768 	 *
6769 	 * We don't have the associated upstream bridge while assigning
6770 	 * the PCI device into guest. For example, the KVM on power is
6771 	 * one of the cases.
6772 	 */
6773 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6774 		struct pci_dev *us_dev = pdev->bus->self;
6775 		u16 devctl;
6776 
6777 		if (!us_dev)
6778 			return 0;
6779 
6780 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6781 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6782 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6783 
6784 		pci_save_state(pdev);
6785 		pci_prepare_to_sleep(pdev);
6786 
6787 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6788 	}
6789 
6790 	return 0;
6791 }
6792 
6793 /**
6794  * __e1000e_disable_aspm - Disable ASPM states
6795  * @pdev: pointer to PCI device struct
6796  * @state: bit-mask of ASPM states to disable
6797  * @locked: indication if this context holds pci_bus_sem locked.
6798  *
6799  * Some devices *must* have certain ASPM states disabled per hardware errata.
6800  **/
__e1000e_disable_aspm(struct pci_dev * pdev,u16 state,int locked)6801 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6802 {
6803 	struct pci_dev *parent = pdev->bus->self;
6804 	u16 aspm_dis_mask = 0;
6805 	u16 pdev_aspmc, parent_aspmc;
6806 
6807 	switch (state) {
6808 	case PCIE_LINK_STATE_L0S:
6809 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6810 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6811 		fallthrough; /* can't have L1 without L0s */
6812 	case PCIE_LINK_STATE_L1:
6813 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6814 		break;
6815 	default:
6816 		return;
6817 	}
6818 
6819 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6820 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6821 
6822 	if (parent) {
6823 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6824 					  &parent_aspmc);
6825 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6826 	}
6827 
6828 	/* Nothing to do if the ASPM states to be disabled already are */
6829 	if (!(pdev_aspmc & aspm_dis_mask) &&
6830 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6831 		return;
6832 
6833 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6834 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6835 		 "L0s" : "",
6836 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6837 		 "L1" : "");
6838 
6839 #ifdef CONFIG_PCIEASPM
6840 	if (locked)
6841 		pci_disable_link_state_locked(pdev, state);
6842 	else
6843 		pci_disable_link_state(pdev, state);
6844 
6845 	/* Double-check ASPM control.  If not disabled by the above, the
6846 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6847 	 * not enabled); override by writing PCI config space directly.
6848 	 */
6849 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6850 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6851 
6852 	if (!(aspm_dis_mask & pdev_aspmc))
6853 		return;
6854 #endif
6855 
6856 	/* Both device and parent should have the same ASPM setting.
6857 	 * Disable ASPM in downstream component first and then upstream.
6858 	 */
6859 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6860 
6861 	if (parent)
6862 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6863 					   aspm_dis_mask);
6864 }
6865 
6866 /**
6867  * e1000e_disable_aspm - Disable ASPM states.
6868  * @pdev: pointer to PCI device struct
6869  * @state: bit-mask of ASPM states to disable
6870  *
6871  * This function acquires the pci_bus_sem!
6872  * Some devices *must* have certain ASPM states disabled per hardware errata.
6873  **/
e1000e_disable_aspm(struct pci_dev * pdev,u16 state)6874 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6875 {
6876 	__e1000e_disable_aspm(pdev, state, 0);
6877 }
6878 
6879 /**
6880  * e1000e_disable_aspm_locked - Disable ASPM states.
6881  * @pdev: pointer to PCI device struct
6882  * @state: bit-mask of ASPM states to disable
6883  *
6884  * This function must be called with pci_bus_sem acquired!
6885  * Some devices *must* have certain ASPM states disabled per hardware errata.
6886  **/
e1000e_disable_aspm_locked(struct pci_dev * pdev,u16 state)6887 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6888 {
6889 	__e1000e_disable_aspm(pdev, state, 1);
6890 }
6891 
e1000e_pm_thaw(struct device * dev)6892 static int e1000e_pm_thaw(struct device *dev)
6893 {
6894 	struct net_device *netdev = dev_get_drvdata(dev);
6895 	struct e1000_adapter *adapter = netdev_priv(netdev);
6896 	int rc = 0;
6897 
6898 	e1000e_set_interrupt_capability(adapter);
6899 
6900 	rtnl_lock();
6901 	if (netif_running(netdev)) {
6902 		rc = e1000_request_irq(adapter);
6903 		if (rc)
6904 			goto err_irq;
6905 
6906 		e1000e_up(adapter);
6907 	}
6908 
6909 	netif_device_attach(netdev);
6910 err_irq:
6911 	rtnl_unlock();
6912 
6913 	return rc;
6914 }
6915 
__e1000_resume(struct pci_dev * pdev)6916 static int __e1000_resume(struct pci_dev *pdev)
6917 {
6918 	struct net_device *netdev = pci_get_drvdata(pdev);
6919 	struct e1000_adapter *adapter = netdev_priv(netdev);
6920 	struct e1000_hw *hw = &adapter->hw;
6921 	u16 aspm_disable_flag = 0;
6922 
6923 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6924 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6925 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6926 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6927 	if (aspm_disable_flag)
6928 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6929 
6930 	pci_set_master(pdev);
6931 
6932 	if (hw->mac.type >= e1000_pch2lan)
6933 		e1000_resume_workarounds_pchlan(&adapter->hw);
6934 
6935 	e1000e_power_up_phy(adapter);
6936 
6937 	/* report the system wakeup cause from S3/S4 */
6938 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6939 		u16 phy_data;
6940 
6941 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6942 		if (phy_data) {
6943 			e_info("PHY Wakeup cause - %s\n",
6944 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6945 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6946 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6947 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6948 			       phy_data & E1000_WUS_LNKC ?
6949 			       "Link Status Change" : "other");
6950 		}
6951 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6952 	} else {
6953 		u32 wus = er32(WUS);
6954 
6955 		if (wus) {
6956 			e_info("MAC Wakeup cause - %s\n",
6957 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6958 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6959 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6960 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6961 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6962 			       "other");
6963 		}
6964 		ew32(WUS, ~0);
6965 	}
6966 
6967 	e1000e_reset(adapter);
6968 
6969 	e1000_init_manageability_pt(adapter);
6970 
6971 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6972 	 * is up.  For all other cases, let the f/w know that the h/w is now
6973 	 * under the control of the driver.
6974 	 */
6975 	if (!(adapter->flags & FLAG_HAS_AMT))
6976 		e1000e_get_hw_control(adapter);
6977 
6978 	return 0;
6979 }
6980 
e1000e_pm_prepare(struct device * dev)6981 static __maybe_unused int e1000e_pm_prepare(struct device *dev)
6982 {
6983 	return pm_runtime_suspended(dev) &&
6984 		pm_suspend_via_firmware();
6985 }
6986 
e1000e_pm_suspend(struct device * dev)6987 static __maybe_unused int e1000e_pm_suspend(struct device *dev)
6988 {
6989 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6990 	struct e1000_adapter *adapter = netdev_priv(netdev);
6991 	struct pci_dev *pdev = to_pci_dev(dev);
6992 	int rc;
6993 
6994 	e1000e_flush_lpic(pdev);
6995 
6996 	e1000e_pm_freeze(dev);
6997 
6998 	rc = __e1000_shutdown(pdev, false);
6999 	if (rc) {
7000 		e1000e_pm_thaw(dev);
7001 	} else {
7002 		/* Introduce S0ix implementation */
7003 		if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
7004 			e1000e_s0ix_entry_flow(adapter);
7005 	}
7006 
7007 	return rc;
7008 }
7009 
e1000e_pm_resume(struct device * dev)7010 static __maybe_unused int e1000e_pm_resume(struct device *dev)
7011 {
7012 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
7013 	struct e1000_adapter *adapter = netdev_priv(netdev);
7014 	struct pci_dev *pdev = to_pci_dev(dev);
7015 	int rc;
7016 
7017 	/* Introduce S0ix implementation */
7018 	if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
7019 		e1000e_s0ix_exit_flow(adapter);
7020 
7021 	rc = __e1000_resume(pdev);
7022 	if (rc)
7023 		return rc;
7024 
7025 	return e1000e_pm_thaw(dev);
7026 }
7027 
e1000e_pm_runtime_idle(struct device * dev)7028 static __maybe_unused int e1000e_pm_runtime_idle(struct device *dev)
7029 {
7030 	struct net_device *netdev = dev_get_drvdata(dev);
7031 	struct e1000_adapter *adapter = netdev_priv(netdev);
7032 	u16 eee_lp;
7033 
7034 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
7035 
7036 	if (!e1000e_has_link(adapter)) {
7037 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
7038 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
7039 	}
7040 
7041 	return -EBUSY;
7042 }
7043 
e1000e_pm_runtime_resume(struct device * dev)7044 static __maybe_unused int e1000e_pm_runtime_resume(struct device *dev)
7045 {
7046 	struct pci_dev *pdev = to_pci_dev(dev);
7047 	struct net_device *netdev = pci_get_drvdata(pdev);
7048 	struct e1000_adapter *adapter = netdev_priv(netdev);
7049 	int rc;
7050 
7051 	rc = __e1000_resume(pdev);
7052 	if (rc)
7053 		return rc;
7054 
7055 	if (netdev->flags & IFF_UP)
7056 		e1000e_up(adapter);
7057 
7058 	return rc;
7059 }
7060 
e1000e_pm_runtime_suspend(struct device * dev)7061 static __maybe_unused int e1000e_pm_runtime_suspend(struct device *dev)
7062 {
7063 	struct pci_dev *pdev = to_pci_dev(dev);
7064 	struct net_device *netdev = pci_get_drvdata(pdev);
7065 	struct e1000_adapter *adapter = netdev_priv(netdev);
7066 
7067 	if (netdev->flags & IFF_UP) {
7068 		int count = E1000_CHECK_RESET_COUNT;
7069 
7070 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
7071 			usleep_range(10000, 11000);
7072 
7073 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
7074 
7075 		/* Down the device without resetting the hardware */
7076 		e1000e_down(adapter, false);
7077 	}
7078 
7079 	if (__e1000_shutdown(pdev, true)) {
7080 		e1000e_pm_runtime_resume(dev);
7081 		return -EBUSY;
7082 	}
7083 
7084 	return 0;
7085 }
7086 
e1000_shutdown(struct pci_dev * pdev)7087 static void e1000_shutdown(struct pci_dev *pdev)
7088 {
7089 	e1000e_flush_lpic(pdev);
7090 
7091 	e1000e_pm_freeze(&pdev->dev);
7092 
7093 	__e1000_shutdown(pdev, false);
7094 }
7095 
7096 #ifdef CONFIG_NET_POLL_CONTROLLER
7097 
e1000_intr_msix(int __always_unused irq,void * data)7098 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
7099 {
7100 	struct net_device *netdev = data;
7101 	struct e1000_adapter *adapter = netdev_priv(netdev);
7102 
7103 	if (adapter->msix_entries) {
7104 		int vector, msix_irq;
7105 
7106 		vector = 0;
7107 		msix_irq = adapter->msix_entries[vector].vector;
7108 		if (disable_hardirq(msix_irq))
7109 			e1000_intr_msix_rx(msix_irq, netdev);
7110 		enable_irq(msix_irq);
7111 
7112 		vector++;
7113 		msix_irq = adapter->msix_entries[vector].vector;
7114 		if (disable_hardirq(msix_irq))
7115 			e1000_intr_msix_tx(msix_irq, netdev);
7116 		enable_irq(msix_irq);
7117 
7118 		vector++;
7119 		msix_irq = adapter->msix_entries[vector].vector;
7120 		if (disable_hardirq(msix_irq))
7121 			e1000_msix_other(msix_irq, netdev);
7122 		enable_irq(msix_irq);
7123 	}
7124 
7125 	return IRQ_HANDLED;
7126 }
7127 
7128 /**
7129  * e1000_netpoll
7130  * @netdev: network interface device structure
7131  *
7132  * Polling 'interrupt' - used by things like netconsole to send skbs
7133  * without having to re-enable interrupts. It's not called while
7134  * the interrupt routine is executing.
7135  */
e1000_netpoll(struct net_device * netdev)7136 static void e1000_netpoll(struct net_device *netdev)
7137 {
7138 	struct e1000_adapter *adapter = netdev_priv(netdev);
7139 
7140 	switch (adapter->int_mode) {
7141 	case E1000E_INT_MODE_MSIX:
7142 		e1000_intr_msix(adapter->pdev->irq, netdev);
7143 		break;
7144 	case E1000E_INT_MODE_MSI:
7145 		if (disable_hardirq(adapter->pdev->irq))
7146 			e1000_intr_msi(adapter->pdev->irq, netdev);
7147 		enable_irq(adapter->pdev->irq);
7148 		break;
7149 	default:		/* E1000E_INT_MODE_LEGACY */
7150 		if (disable_hardirq(adapter->pdev->irq))
7151 			e1000_intr(adapter->pdev->irq, netdev);
7152 		enable_irq(adapter->pdev->irq);
7153 		break;
7154 	}
7155 }
7156 #endif
7157 
7158 /**
7159  * e1000_io_error_detected - called when PCI error is detected
7160  * @pdev: Pointer to PCI device
7161  * @state: The current pci connection state
7162  *
7163  * This function is called after a PCI bus error affecting
7164  * this device has been detected.
7165  */
e1000_io_error_detected(struct pci_dev * pdev,pci_channel_state_t state)7166 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7167 						pci_channel_state_t state)
7168 {
7169 	e1000e_pm_freeze(&pdev->dev);
7170 
7171 	if (state == pci_channel_io_perm_failure)
7172 		return PCI_ERS_RESULT_DISCONNECT;
7173 
7174 	pci_disable_device(pdev);
7175 
7176 	/* Request a slot reset. */
7177 	return PCI_ERS_RESULT_NEED_RESET;
7178 }
7179 
7180 /**
7181  * e1000_io_slot_reset - called after the pci bus has been reset.
7182  * @pdev: Pointer to PCI device
7183  *
7184  * Restart the card from scratch, as if from a cold-boot. Implementation
7185  * resembles the first-half of the e1000e_pm_resume routine.
7186  */
e1000_io_slot_reset(struct pci_dev * pdev)7187 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7188 {
7189 	struct net_device *netdev = pci_get_drvdata(pdev);
7190 	struct e1000_adapter *adapter = netdev_priv(netdev);
7191 	struct e1000_hw *hw = &adapter->hw;
7192 	u16 aspm_disable_flag = 0;
7193 	int err;
7194 	pci_ers_result_t result;
7195 
7196 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7197 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7198 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7199 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7200 	if (aspm_disable_flag)
7201 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7202 
7203 	err = pci_enable_device_mem(pdev);
7204 	if (err) {
7205 		dev_err(&pdev->dev,
7206 			"Cannot re-enable PCI device after reset.\n");
7207 		result = PCI_ERS_RESULT_DISCONNECT;
7208 	} else {
7209 		pdev->state_saved = true;
7210 		pci_restore_state(pdev);
7211 		pci_set_master(pdev);
7212 
7213 		pci_enable_wake(pdev, PCI_D3hot, 0);
7214 		pci_enable_wake(pdev, PCI_D3cold, 0);
7215 
7216 		e1000e_reset(adapter);
7217 		ew32(WUS, ~0);
7218 		result = PCI_ERS_RESULT_RECOVERED;
7219 	}
7220 
7221 	return result;
7222 }
7223 
7224 /**
7225  * e1000_io_resume - called when traffic can start flowing again.
7226  * @pdev: Pointer to PCI device
7227  *
7228  * This callback is called when the error recovery driver tells us that
7229  * its OK to resume normal operation. Implementation resembles the
7230  * second-half of the e1000e_pm_resume routine.
7231  */
e1000_io_resume(struct pci_dev * pdev)7232 static void e1000_io_resume(struct pci_dev *pdev)
7233 {
7234 	struct net_device *netdev = pci_get_drvdata(pdev);
7235 	struct e1000_adapter *adapter = netdev_priv(netdev);
7236 
7237 	e1000_init_manageability_pt(adapter);
7238 
7239 	e1000e_pm_thaw(&pdev->dev);
7240 
7241 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7242 	 * is up.  For all other cases, let the f/w know that the h/w is now
7243 	 * under the control of the driver.
7244 	 */
7245 	if (!(adapter->flags & FLAG_HAS_AMT))
7246 		e1000e_get_hw_control(adapter);
7247 }
7248 
e1000_print_device_info(struct e1000_adapter * adapter)7249 static void e1000_print_device_info(struct e1000_adapter *adapter)
7250 {
7251 	struct e1000_hw *hw = &adapter->hw;
7252 	struct net_device *netdev = adapter->netdev;
7253 	u32 ret_val;
7254 	u8 pba_str[E1000_PBANUM_LENGTH];
7255 
7256 	/* print bus type/speed/width info */
7257 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7258 	       /* bus width */
7259 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7260 		"Width x1"),
7261 	       /* MAC address */
7262 	       netdev->dev_addr);
7263 	e_info("Intel(R) PRO/%s Network Connection\n",
7264 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7265 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
7266 						E1000_PBANUM_LENGTH);
7267 	if (ret_val)
7268 		strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
7269 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7270 	       hw->mac.type, hw->phy.type, pba_str);
7271 }
7272 
e1000_eeprom_checks(struct e1000_adapter * adapter)7273 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7274 {
7275 	struct e1000_hw *hw = &adapter->hw;
7276 	int ret_val;
7277 	u16 buf = 0;
7278 
7279 	if (hw->mac.type != e1000_82573)
7280 		return;
7281 
7282 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7283 	le16_to_cpus(&buf);
7284 	if (!ret_val && (!(buf & BIT(0)))) {
7285 		/* Deep Smart Power Down (DSPD) */
7286 		dev_warn(&adapter->pdev->dev,
7287 			 "Warning: detected DSPD enabled in EEPROM\n");
7288 	}
7289 }
7290 
e1000_fix_features(struct net_device * netdev,netdev_features_t features)7291 static netdev_features_t e1000_fix_features(struct net_device *netdev,
7292 					    netdev_features_t features)
7293 {
7294 	struct e1000_adapter *adapter = netdev_priv(netdev);
7295 	struct e1000_hw *hw = &adapter->hw;
7296 
7297 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7298 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7299 		features &= ~NETIF_F_RXFCS;
7300 
7301 	/* Since there is no support for separate Rx/Tx vlan accel
7302 	 * enable/disable make sure Tx flag is always in same state as Rx.
7303 	 */
7304 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7305 		features |= NETIF_F_HW_VLAN_CTAG_TX;
7306 	else
7307 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7308 
7309 	return features;
7310 }
7311 
e1000_set_features(struct net_device * netdev,netdev_features_t features)7312 static int e1000_set_features(struct net_device *netdev,
7313 			      netdev_features_t features)
7314 {
7315 	struct e1000_adapter *adapter = netdev_priv(netdev);
7316 	netdev_features_t changed = features ^ netdev->features;
7317 
7318 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7319 		adapter->flags |= FLAG_TSO_FORCE;
7320 
7321 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7322 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7323 			 NETIF_F_RXALL)))
7324 		return 0;
7325 
7326 	if (changed & NETIF_F_RXFCS) {
7327 		if (features & NETIF_F_RXFCS) {
7328 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7329 		} else {
7330 			/* We need to take it back to defaults, which might mean
7331 			 * stripping is still disabled at the adapter level.
7332 			 */
7333 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7334 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7335 			else
7336 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7337 		}
7338 	}
7339 
7340 	netdev->features = features;
7341 
7342 	if (netif_running(netdev))
7343 		e1000e_reinit_locked(adapter);
7344 	else
7345 		e1000e_reset(adapter);
7346 
7347 	return 1;
7348 }
7349 
7350 static const struct net_device_ops e1000e_netdev_ops = {
7351 	.ndo_open		= e1000e_open,
7352 	.ndo_stop		= e1000e_close,
7353 	.ndo_start_xmit		= e1000_xmit_frame,
7354 	.ndo_get_stats64	= e1000e_get_stats64,
7355 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7356 	.ndo_set_mac_address	= e1000_set_mac,
7357 	.ndo_change_mtu		= e1000_change_mtu,
7358 	.ndo_eth_ioctl		= e1000_ioctl,
7359 	.ndo_tx_timeout		= e1000_tx_timeout,
7360 	.ndo_validate_addr	= eth_validate_addr,
7361 
7362 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7363 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7364 #ifdef CONFIG_NET_POLL_CONTROLLER
7365 	.ndo_poll_controller	= e1000_netpoll,
7366 #endif
7367 	.ndo_set_features = e1000_set_features,
7368 	.ndo_fix_features = e1000_fix_features,
7369 	.ndo_features_check	= passthru_features_check,
7370 };
7371 
7372 /**
7373  * e1000_probe - Device Initialization Routine
7374  * @pdev: PCI device information struct
7375  * @ent: entry in e1000_pci_tbl
7376  *
7377  * Returns 0 on success, negative on failure
7378  *
7379  * e1000_probe initializes an adapter identified by a pci_dev structure.
7380  * The OS initialization, configuring of the adapter private structure,
7381  * and a hardware reset occur.
7382  **/
e1000_probe(struct pci_dev * pdev,const struct pci_device_id * ent)7383 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7384 {
7385 	struct net_device *netdev;
7386 	struct e1000_adapter *adapter;
7387 	struct e1000_hw *hw;
7388 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7389 	resource_size_t mmio_start, mmio_len;
7390 	resource_size_t flash_start, flash_len;
7391 	static int cards_found;
7392 	u16 aspm_disable_flag = 0;
7393 	int bars, i, err, pci_using_dac;
7394 	u16 eeprom_data = 0;
7395 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7396 	s32 ret_val = 0;
7397 
7398 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7399 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7400 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7401 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7402 	if (aspm_disable_flag)
7403 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7404 
7405 	err = pci_enable_device_mem(pdev);
7406 	if (err)
7407 		return err;
7408 
7409 	pci_using_dac = 0;
7410 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7411 	if (!err) {
7412 		pci_using_dac = 1;
7413 	} else {
7414 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
7415 		if (err) {
7416 			dev_err(&pdev->dev,
7417 				"No usable DMA configuration, aborting\n");
7418 			goto err_dma;
7419 		}
7420 	}
7421 
7422 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7423 	err = pci_request_selected_regions_exclusive(pdev, bars,
7424 						     e1000e_driver_name);
7425 	if (err)
7426 		goto err_pci_reg;
7427 
7428 	/* AER (Advanced Error Reporting) hooks */
7429 	pci_enable_pcie_error_reporting(pdev);
7430 
7431 	pci_set_master(pdev);
7432 	/* PCI config space info */
7433 	err = pci_save_state(pdev);
7434 	if (err)
7435 		goto err_alloc_etherdev;
7436 
7437 	err = -ENOMEM;
7438 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7439 	if (!netdev)
7440 		goto err_alloc_etherdev;
7441 
7442 	SET_NETDEV_DEV(netdev, &pdev->dev);
7443 
7444 	netdev->irq = pdev->irq;
7445 
7446 	pci_set_drvdata(pdev, netdev);
7447 	adapter = netdev_priv(netdev);
7448 	hw = &adapter->hw;
7449 	adapter->netdev = netdev;
7450 	adapter->pdev = pdev;
7451 	adapter->ei = ei;
7452 	adapter->pba = ei->pba;
7453 	adapter->flags = ei->flags;
7454 	adapter->flags2 = ei->flags2;
7455 	adapter->hw.adapter = adapter;
7456 	adapter->hw.mac.type = ei->mac;
7457 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7458 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7459 
7460 	mmio_start = pci_resource_start(pdev, 0);
7461 	mmio_len = pci_resource_len(pdev, 0);
7462 
7463 	err = -EIO;
7464 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7465 	if (!adapter->hw.hw_addr)
7466 		goto err_ioremap;
7467 
7468 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7469 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7470 	    (hw->mac.type < e1000_pch_spt)) {
7471 		flash_start = pci_resource_start(pdev, 1);
7472 		flash_len = pci_resource_len(pdev, 1);
7473 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7474 		if (!adapter->hw.flash_address)
7475 			goto err_flashmap;
7476 	}
7477 
7478 	/* Set default EEE advertisement */
7479 	if (adapter->flags2 & FLAG2_HAS_EEE)
7480 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7481 
7482 	/* construct the net_device struct */
7483 	netdev->netdev_ops = &e1000e_netdev_ops;
7484 	e1000e_set_ethtool_ops(netdev);
7485 	netdev->watchdog_timeo = 5 * HZ;
7486 	netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
7487 	strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7488 
7489 	netdev->mem_start = mmio_start;
7490 	netdev->mem_end = mmio_start + mmio_len;
7491 
7492 	adapter->bd_number = cards_found++;
7493 
7494 	e1000e_check_options(adapter);
7495 
7496 	/* setup adapter struct */
7497 	err = e1000_sw_init(adapter);
7498 	if (err)
7499 		goto err_sw_init;
7500 
7501 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7502 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7503 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7504 
7505 	err = ei->get_variants(adapter);
7506 	if (err)
7507 		goto err_hw_init;
7508 
7509 	if ((adapter->flags & FLAG_IS_ICH) &&
7510 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7511 	    (hw->mac.type < e1000_pch_spt))
7512 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7513 
7514 	hw->mac.ops.get_bus_info(&adapter->hw);
7515 
7516 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7517 
7518 	/* Copper options */
7519 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7520 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7521 		adapter->hw.phy.disable_polarity_correction = 0;
7522 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7523 	}
7524 
7525 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7526 		dev_info(&pdev->dev,
7527 			 "PHY reset is blocked due to SOL/IDER session.\n");
7528 
7529 	/* Set initial default active device features */
7530 	netdev->features = (NETIF_F_SG |
7531 			    NETIF_F_HW_VLAN_CTAG_RX |
7532 			    NETIF_F_HW_VLAN_CTAG_TX |
7533 			    NETIF_F_TSO |
7534 			    NETIF_F_TSO6 |
7535 			    NETIF_F_RXHASH |
7536 			    NETIF_F_RXCSUM |
7537 			    NETIF_F_HW_CSUM);
7538 
7539 	/* Set user-changeable features (subset of all device features) */
7540 	netdev->hw_features = netdev->features;
7541 	netdev->hw_features |= NETIF_F_RXFCS;
7542 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7543 	netdev->hw_features |= NETIF_F_RXALL;
7544 
7545 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7546 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7547 
7548 	netdev->vlan_features |= (NETIF_F_SG |
7549 				  NETIF_F_TSO |
7550 				  NETIF_F_TSO6 |
7551 				  NETIF_F_HW_CSUM);
7552 
7553 	netdev->priv_flags |= IFF_UNICAST_FLT;
7554 
7555 	if (pci_using_dac) {
7556 		netdev->features |= NETIF_F_HIGHDMA;
7557 		netdev->vlan_features |= NETIF_F_HIGHDMA;
7558 	}
7559 
7560 	/* MTU range: 68 - max_hw_frame_size */
7561 	netdev->min_mtu = ETH_MIN_MTU;
7562 	netdev->max_mtu = adapter->max_hw_frame_size -
7563 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7564 
7565 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7566 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7567 
7568 	/* before reading the NVM, reset the controller to
7569 	 * put the device in a known good starting state
7570 	 */
7571 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7572 
7573 	/* systems with ASPM and others may see the checksum fail on the first
7574 	 * attempt. Let's give it a few tries
7575 	 */
7576 	for (i = 0;; i++) {
7577 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7578 			break;
7579 		if (i == 2) {
7580 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7581 			err = -EIO;
7582 			goto err_eeprom;
7583 		}
7584 	}
7585 
7586 	e1000_eeprom_checks(adapter);
7587 
7588 	/* copy the MAC address */
7589 	if (e1000e_read_mac_addr(&adapter->hw))
7590 		dev_err(&pdev->dev,
7591 			"NVM Read Error while reading MAC address\n");
7592 
7593 	memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
7594 
7595 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7596 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7597 			netdev->dev_addr);
7598 		err = -EIO;
7599 		goto err_eeprom;
7600 	}
7601 
7602 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7603 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7604 
7605 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7606 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7607 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7608 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7609 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7610 
7611 	/* Initialize link parameters. User can change them with ethtool */
7612 	adapter->hw.mac.autoneg = 1;
7613 	adapter->fc_autoneg = true;
7614 	adapter->hw.fc.requested_mode = e1000_fc_default;
7615 	adapter->hw.fc.current_mode = e1000_fc_default;
7616 	adapter->hw.phy.autoneg_advertised = 0x2f;
7617 
7618 	/* Initial Wake on LAN setting - If APM wake is enabled in
7619 	 * the EEPROM, enable the ACPI Magic Packet filter
7620 	 */
7621 	if (adapter->flags & FLAG_APME_IN_WUC) {
7622 		/* APME bit in EEPROM is mapped to WUC.APME */
7623 		eeprom_data = er32(WUC);
7624 		eeprom_apme_mask = E1000_WUC_APME;
7625 		if ((hw->mac.type > e1000_ich10lan) &&
7626 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7627 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7628 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7629 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7630 		    (adapter->hw.bus.func == 1))
7631 			ret_val = e1000_read_nvm(&adapter->hw,
7632 					      NVM_INIT_CONTROL3_PORT_B,
7633 					      1, &eeprom_data);
7634 		else
7635 			ret_val = e1000_read_nvm(&adapter->hw,
7636 					      NVM_INIT_CONTROL3_PORT_A,
7637 					      1, &eeprom_data);
7638 	}
7639 
7640 	/* fetch WoL from EEPROM */
7641 	if (ret_val)
7642 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7643 	else if (eeprom_data & eeprom_apme_mask)
7644 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7645 
7646 	/* now that we have the eeprom settings, apply the special cases
7647 	 * where the eeprom may be wrong or the board simply won't support
7648 	 * wake on lan on a particular port
7649 	 */
7650 	if (!(adapter->flags & FLAG_HAS_WOL))
7651 		adapter->eeprom_wol = 0;
7652 
7653 	/* initialize the wol settings based on the eeprom settings */
7654 	adapter->wol = adapter->eeprom_wol;
7655 
7656 	/* make sure adapter isn't asleep if manageability is enabled */
7657 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7658 	    (hw->mac.ops.check_mng_mode(hw)))
7659 		device_wakeup_enable(&pdev->dev);
7660 
7661 	/* save off EEPROM version number */
7662 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7663 
7664 	if (ret_val) {
7665 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7666 		adapter->eeprom_vers = 0;
7667 	}
7668 
7669 	/* init PTP hardware clock */
7670 	e1000e_ptp_init(adapter);
7671 
7672 	/* reset the hardware with the new settings */
7673 	e1000e_reset(adapter);
7674 
7675 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7676 	 * is up.  For all other cases, let the f/w know that the h/w is now
7677 	 * under the control of the driver.
7678 	 */
7679 	if (!(adapter->flags & FLAG_HAS_AMT))
7680 		e1000e_get_hw_control(adapter);
7681 
7682 	if (hw->mac.type >= e1000_pch_cnp)
7683 		adapter->flags2 |= FLAG2_ENABLE_S0IX_FLOWS;
7684 
7685 	strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
7686 	err = register_netdev(netdev);
7687 	if (err)
7688 		goto err_register;
7689 
7690 	/* carrier off reporting is important to ethtool even BEFORE open */
7691 	netif_carrier_off(netdev);
7692 
7693 	e1000_print_device_info(adapter);
7694 
7695 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_SMART_PREPARE);
7696 
7697 	if (pci_dev_run_wake(pdev) && hw->mac.type != e1000_pch_cnp)
7698 		pm_runtime_put_noidle(&pdev->dev);
7699 
7700 	return 0;
7701 
7702 err_register:
7703 	if (!(adapter->flags & FLAG_HAS_AMT))
7704 		e1000e_release_hw_control(adapter);
7705 err_eeprom:
7706 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7707 		e1000_phy_hw_reset(&adapter->hw);
7708 err_hw_init:
7709 	kfree(adapter->tx_ring);
7710 	kfree(adapter->rx_ring);
7711 err_sw_init:
7712 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7713 		iounmap(adapter->hw.flash_address);
7714 	e1000e_reset_interrupt_capability(adapter);
7715 err_flashmap:
7716 	iounmap(adapter->hw.hw_addr);
7717 err_ioremap:
7718 	free_netdev(netdev);
7719 err_alloc_etherdev:
7720 	pci_disable_pcie_error_reporting(pdev);
7721 	pci_release_mem_regions(pdev);
7722 err_pci_reg:
7723 err_dma:
7724 	pci_disable_device(pdev);
7725 	return err;
7726 }
7727 
7728 /**
7729  * e1000_remove - Device Removal Routine
7730  * @pdev: PCI device information struct
7731  *
7732  * e1000_remove is called by the PCI subsystem to alert the driver
7733  * that it should release a PCI device.  This could be caused by a
7734  * Hot-Plug event, or because the driver is going to be removed from
7735  * memory.
7736  **/
e1000_remove(struct pci_dev * pdev)7737 static void e1000_remove(struct pci_dev *pdev)
7738 {
7739 	struct net_device *netdev = pci_get_drvdata(pdev);
7740 	struct e1000_adapter *adapter = netdev_priv(netdev);
7741 
7742 	e1000e_ptp_remove(adapter);
7743 
7744 	/* The timers may be rescheduled, so explicitly disable them
7745 	 * from being rescheduled.
7746 	 */
7747 	set_bit(__E1000_DOWN, &adapter->state);
7748 	del_timer_sync(&adapter->watchdog_timer);
7749 	del_timer_sync(&adapter->phy_info_timer);
7750 
7751 	cancel_work_sync(&adapter->reset_task);
7752 	cancel_work_sync(&adapter->watchdog_task);
7753 	cancel_work_sync(&adapter->downshift_task);
7754 	cancel_work_sync(&adapter->update_phy_task);
7755 	cancel_work_sync(&adapter->print_hang_task);
7756 
7757 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7758 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7759 		if (adapter->tx_hwtstamp_skb) {
7760 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7761 			adapter->tx_hwtstamp_skb = NULL;
7762 		}
7763 	}
7764 
7765 	unregister_netdev(netdev);
7766 
7767 	if (pci_dev_run_wake(pdev))
7768 		pm_runtime_get_noresume(&pdev->dev);
7769 
7770 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7771 	 * would have already happened in close and is redundant.
7772 	 */
7773 	e1000e_release_hw_control(adapter);
7774 
7775 	e1000e_reset_interrupt_capability(adapter);
7776 	kfree(adapter->tx_ring);
7777 	kfree(adapter->rx_ring);
7778 
7779 	iounmap(adapter->hw.hw_addr);
7780 	if ((adapter->hw.flash_address) &&
7781 	    (adapter->hw.mac.type < e1000_pch_spt))
7782 		iounmap(adapter->hw.flash_address);
7783 	pci_release_mem_regions(pdev);
7784 
7785 	free_netdev(netdev);
7786 
7787 	/* AER disable */
7788 	pci_disable_pcie_error_reporting(pdev);
7789 
7790 	pci_disable_device(pdev);
7791 }
7792 
7793 /* PCI Error Recovery (ERS) */
7794 static const struct pci_error_handlers e1000_err_handler = {
7795 	.error_detected = e1000_io_error_detected,
7796 	.slot_reset = e1000_io_slot_reset,
7797 	.resume = e1000_io_resume,
7798 };
7799 
7800 static const struct pci_device_id e1000_pci_tbl[] = {
7801 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7802 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7803 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7804 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7805 	  board_82571 },
7806 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7807 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7808 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7809 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7810 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7811 
7812 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7813 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7814 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7815 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7816 
7817 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7818 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7819 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7820 
7821 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7822 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7823 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7824 
7825 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7826 	  board_80003es2lan },
7827 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7828 	  board_80003es2lan },
7829 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7830 	  board_80003es2lan },
7831 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7832 	  board_80003es2lan },
7833 
7834 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7835 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7836 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7837 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7838 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7839 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7840 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7841 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7842 
7843 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7844 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7845 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7846 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7847 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7848 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7849 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7850 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7851 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7852 
7853 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7854 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7855 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7856 
7857 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7858 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7859 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7860 
7861 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7862 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7863 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7864 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7865 
7866 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7867 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7868 
7869 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7870 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7871 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7872 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7873 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7874 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7875 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7876 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7877 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7878 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7879 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7880 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7881 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7882 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7883 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7884 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7885 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7886 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7887 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7888 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7889 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7890 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7891 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7892 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7893 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7894 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10), board_pch_cnp },
7895 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10), board_pch_cnp },
7896 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11), board_pch_cnp },
7897 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11), board_pch_cnp },
7898 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12), board_pch_spt },
7899 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12), board_pch_spt },
7900 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13), board_pch_tgp },
7901 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13), board_pch_tgp },
7902 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14), board_pch_tgp },
7903 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14), board_pch_tgp },
7904 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15), board_pch_tgp },
7905 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V15), board_pch_tgp },
7906 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM23), board_pch_tgp },
7907 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V23), board_pch_tgp },
7908 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM16), board_pch_tgp },
7909 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V16), board_pch_tgp },
7910 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM17), board_pch_tgp },
7911 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V17), board_pch_tgp },
7912 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM22), board_pch_tgp },
7913 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V22), board_pch_tgp },
7914 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM18), board_pch_tgp },
7915 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V18), board_pch_tgp },
7916 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM19), board_pch_tgp },
7917 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V19), board_pch_tgp },
7918 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM20), board_pch_tgp },
7919 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V20), board_pch_tgp },
7920 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM21), board_pch_tgp },
7921 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V21), board_pch_tgp },
7922 
7923 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7924 };
7925 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7926 
7927 static const struct dev_pm_ops e1000_pm_ops = {
7928 #ifdef CONFIG_PM_SLEEP
7929 	.prepare	= e1000e_pm_prepare,
7930 	.suspend	= e1000e_pm_suspend,
7931 	.resume		= e1000e_pm_resume,
7932 	.freeze		= e1000e_pm_freeze,
7933 	.thaw		= e1000e_pm_thaw,
7934 	.poweroff	= e1000e_pm_suspend,
7935 	.restore	= e1000e_pm_resume,
7936 #endif
7937 	SET_RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7938 			   e1000e_pm_runtime_idle)
7939 };
7940 
7941 /* PCI Device API Driver */
7942 static struct pci_driver e1000_driver = {
7943 	.name     = e1000e_driver_name,
7944 	.id_table = e1000_pci_tbl,
7945 	.probe    = e1000_probe,
7946 	.remove   = e1000_remove,
7947 	.driver   = {
7948 		.pm = &e1000_pm_ops,
7949 	},
7950 	.shutdown = e1000_shutdown,
7951 	.err_handler = &e1000_err_handler
7952 };
7953 
7954 /**
7955  * e1000_init_module - Driver Registration Routine
7956  *
7957  * e1000_init_module is the first routine called when the driver is
7958  * loaded. All it does is register with the PCI subsystem.
7959  **/
e1000_init_module(void)7960 static int __init e1000_init_module(void)
7961 {
7962 	pr_info("Intel(R) PRO/1000 Network Driver\n");
7963 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7964 
7965 	return pci_register_driver(&e1000_driver);
7966 }
7967 module_init(e1000_init_module);
7968 
7969 /**
7970  * e1000_exit_module - Driver Exit Cleanup Routine
7971  *
7972  * e1000_exit_module is called just before the driver is removed
7973  * from memory.
7974  **/
e1000_exit_module(void)7975 static void __exit e1000_exit_module(void)
7976 {
7977 	pci_unregister_driver(&e1000_driver);
7978 }
7979 module_exit(e1000_exit_module);
7980 
7981 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7982 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7983 MODULE_LICENSE("GPL v2");
7984 
7985 /* netdev.c */
7986