1Queue sysfs files 2================= 3 4This text file will detail the queue files that are located in the sysfs tree 5for each block device. Note that stacked devices typically do not export 6any settings, since their queue merely functions are a remapping target. 7These files are the ones found in the /sys/block/xxx/queue/ directory. 8 9Files denoted with a RO postfix are readonly and the RW postfix means 10read-write. 11 12add_random (RW) 13---------------- 14This file allows to turn off the disk entropy contribution. Default 15value of this file is '1'(on). 16 17dax (RO) 18-------- 19This file indicates whether the device supports Direct Access (DAX), 20used by CPU-addressable storage to bypass the pagecache. It shows '1' 21if true, '0' if not. 22 23discard_granularity (RO) 24----------------------- 25This shows the size of internal allocation of the device in bytes, if 26reported by the device. A value of '0' means device does not support 27the discard functionality. 28 29discard_max_hw_bytes (RO) 30---------------------- 31Devices that support discard functionality may have internal limits on 32the number of bytes that can be trimmed or unmapped in a single operation. 33The discard_max_bytes parameter is set by the device driver to the maximum 34number of bytes that can be discarded in a single operation. Discard 35requests issued to the device must not exceed this limit. A discard_max_bytes 36value of 0 means that the device does not support discard functionality. 37 38discard_max_bytes (RW) 39---------------------- 40While discard_max_hw_bytes is the hardware limit for the device, this 41setting is the software limit. Some devices exhibit large latencies when 42large discards are issued, setting this value lower will make Linux issue 43smaller discards and potentially help reduce latencies induced by large 44discard operations. 45 46hw_sector_size (RO) 47------------------- 48This is the hardware sector size of the device, in bytes. 49 50io_poll (RW) 51------------ 52When read, this file shows whether polling is enabled (1) or disabled 53(0). Writing '0' to this file will disable polling for this device. 54Writing any non-zero value will enable this feature. 55 56io_poll_delay (RW) 57------------------ 58If polling is enabled, this controls what kind of polling will be 59performed. It defaults to -1, which is classic polling. In this mode, 60the CPU will repeatedly ask for completions without giving up any time. 61If set to 0, a hybrid polling mode is used, where the kernel will attempt 62to make an educated guess at when the IO will complete. Based on this 63guess, the kernel will put the process issuing IO to sleep for an amount 64of time, before entering a classic poll loop. This mode might be a 65little slower than pure classic polling, but it will be more efficient. 66If set to a value larger than 0, the kernel will put the process issuing 67IO to sleep for this amont of microseconds before entering classic 68polling. 69 70iostats (RW) 71------------- 72This file is used to control (on/off) the iostats accounting of the 73disk. 74 75logical_block_size (RO) 76----------------------- 77This is the logical block size of the device, in bytes. 78 79max_hw_sectors_kb (RO) 80---------------------- 81This is the maximum number of kilobytes supported in a single data transfer. 82 83max_integrity_segments (RO) 84--------------------------- 85When read, this file shows the max limit of integrity segments as 86set by block layer which a hardware controller can handle. 87 88max_sectors_kb (RW) 89------------------- 90This is the maximum number of kilobytes that the block layer will allow 91for a filesystem request. Must be smaller than or equal to the maximum 92size allowed by the hardware. 93 94max_segments (RO) 95----------------- 96Maximum number of segments of the device. 97 98max_segment_size (RO) 99--------------------- 100Maximum segment size of the device. 101 102minimum_io_size (RO) 103-------------------- 104This is the smallest preferred IO size reported by the device. 105 106nomerges (RW) 107------------- 108This enables the user to disable the lookup logic involved with IO 109merging requests in the block layer. By default (0) all merges are 110enabled. When set to 1 only simple one-hit merges will be tried. When 111set to 2 no merge algorithms will be tried (including one-hit or more 112complex tree/hash lookups). 113 114nr_requests (RW) 115---------------- 116This controls how many requests may be allocated in the block layer for 117read or write requests. Note that the total allocated number may be twice 118this amount, since it applies only to reads or writes (not the accumulated 119sum). 120 121To avoid priority inversion through request starvation, a request 122queue maintains a separate request pool per each cgroup when 123CONFIG_BLK_CGROUP is enabled, and this parameter applies to each such 124per-block-cgroup request pool. IOW, if there are N block cgroups, 125each request queue may have up to N request pools, each independently 126regulated by nr_requests. 127 128optimal_io_size (RO) 129-------------------- 130This is the optimal IO size reported by the device. 131 132physical_block_size (RO) 133------------------------ 134This is the physical block size of device, in bytes. 135 136read_ahead_kb (RW) 137------------------ 138Maximum number of kilobytes to read-ahead for filesystems on this block 139device. 140 141rotational (RW) 142--------------- 143This file is used to stat if the device is of rotational type or 144non-rotational type. 145 146rq_affinity (RW) 147---------------- 148If this option is '1', the block layer will migrate request completions to the 149cpu "group" that originally submitted the request. For some workloads this 150provides a significant reduction in CPU cycles due to caching effects. 151 152For storage configurations that need to maximize distribution of completion 153processing setting this option to '2' forces the completion to run on the 154requesting cpu (bypassing the "group" aggregation logic). 155 156scheduler (RW) 157-------------- 158When read, this file will display the current and available IO schedulers 159for this block device. The currently active IO scheduler will be enclosed 160in [] brackets. Writing an IO scheduler name to this file will switch 161control of this block device to that new IO scheduler. Note that writing 162an IO scheduler name to this file will attempt to load that IO scheduler 163module, if it isn't already present in the system. 164 165write_cache (RW) 166---------------- 167When read, this file will display whether the device has write back 168caching enabled or not. It will return "write back" for the former 169case, and "write through" for the latter. Writing to this file can 170change the kernels view of the device, but it doesn't alter the 171device state. This means that it might not be safe to toggle the 172setting from "write back" to "write through", since that will also 173eliminate cache flushes issued by the kernel. 174 175write_same_max_bytes (RO) 176------------------------- 177This is the number of bytes the device can write in a single write-same 178command. A value of '0' means write-same is not supported by this 179device. 180 181wb_lat_usec (RW) 182---------------- 183If the device is registered for writeback throttling, then this file shows 184the target minimum read latency. If this latency is exceeded in a given 185window of time (see wb_window_usec), then the writeback throttling will start 186scaling back writes. Writing a value of '0' to this file disables the 187feature. Writing a value of '-1' to this file resets the value to the 188default setting. 189 190throttle_sample_time (RW) 191------------------------- 192This is the time window that blk-throttle samples data, in millisecond. 193blk-throttle makes decision based on the samplings. Lower time means cgroups 194have more smooth throughput, but higher CPU overhead. This exists only when 195CONFIG_BLK_DEV_THROTTLING_LOW is enabled. 196 197Jens Axboe <jens.axboe@oracle.com>, February 2009 198
