Documentation/blockdev/zram.txt

   1 zram: Compressed RAM based block devices
   2 ----------------------------------------
   3
   4 * Introduction
   5
   6 The zram module creates RAM based block devices named /dev/zram<id>
   7 (<id> = 0, 1, ...). Pages written to these disks are compressed and stored
   8 in memory itself. These disks allow very fast I/O and compression provides
   9 good amounts of memory savings. Some of the usecases include /tmp storage,
  10 use as swap disks, various caches under /var and maybe many more :)
  11
  12 Statistics for individual zram devices are exported through sysfs nodes at
  13 /sys/block/zram<id>/
  14
  15 * Usage
  16
  17 There are several ways to configure and manage zram device(-s):
  18 a) using zram and zram_control sysfs attributes
  19 b) using zramctl utility, provided by util-linux (util-linux@vger.kernel.org).
  20
  21 In this document we will describe only 'manual' zram configuration steps,
  22 IOW, zram and zram_control sysfs attributes.
  23
  24 In order to get a better idea about zramctl please consult util-linux
  25 documentation, zramctl man-page or `zramctl --help'. Please be informed
  26 that zram maintainers do not develop/maintain util-linux or zramctl, should
  27 you have any questions please contact util-linux@vger.kernel.org
  28
  29 Following shows a typical sequence of steps for using zram.
  30
  31 WARNING
  32 =======
  33 For the sake of simplicity we skip error checking parts in most of the
  34 examples below. However, it is your sole responsibility to handle errors.
  35
  36 zram sysfs attributes always return negative values in case of errors.
  37 The list of possible return codes:
  38 -EBUSY  -- an attempt to modify an attribute that cannot be changed once
  39 the device has been initialised. Please reset device first;
  40 -ENOMEM -- zram was not able to allocate enough memory to fulfil your
  41 needs;
  42 -EINVAL -- invalid input has been provided.
  43
  44 If you use 'echo', the returned value that is changed by 'echo' utility,
  45 and, in general case, something like:
  46
  47         echo 3 > /sys/block/zram0/max_comp_streams
  48         if [ $? -ne 0 ];
  49                 handle_error
  50         fi
  51
  52 should suffice.
  53
  54 1) Load Module:
  55         modprobe zram num_devices=4
  56         This creates 4 devices: /dev/zram{0,1,2,3}
  57
  58 num_devices parameter is optional and tells zram how many devices should be
  59 pre-created. Default: 1.
  60
  61 2) Set max number of compression streams
  62 Regardless the value passed to this attribute, ZRAM will always
  63 allocate multiple compression streams - one per online CPUs - thus
  64 allowing several concurrent compression operations. The number of
  65 allocated compression streams goes down when some of the CPUs
  66 become offline. There is no single-compression-stream mode anymore,
  67 unless you are running a UP system or has only 1 CPU online.
  68
  69 To find out how many streams are currently available:
  70         cat /sys/block/zram0/max_comp_streams
  71
  72 3) Select compression algorithm
  73 Using comp_algorithm device attribute one can see available and
  74 currently selected (shown in square brackets) compression algorithms,
  75 change selected compression algorithm (once the device is initialised
  76 there is no way to change compression algorithm).
  77
  78 Examples:
  79         #show supported compression algorithms
  80         cat /sys/block/zram0/comp_algorithm
  81         lzo [lz4]
  82
  83         #select lzo compression algorithm
  84         echo lzo > /sys/block/zram0/comp_algorithm
  85
  86 For the time being, the `comp_algorithm' content does not necessarily
  87 show every compression algorithm supported by the kernel. We keep this
  88 list primarily to simplify device configuration and one can configure
  89 a new device with a compression algorithm that is not listed in
  90 `comp_algorithm'. The thing is that, internally, ZRAM uses Crypto API
  91 and, if some of the algorithms were built as modules, it's impossible
  92 to list all of them using, for instance, /proc/crypto or any other
  93 method. This, however, has an advantage of permitting the usage of
  94 custom crypto compression modules (implementing S/W or H/W compression).
  95
  96 4) Set Disksize
  97 Set disk size by writing the value to sysfs node 'disksize'.
  98 The value can be either in bytes or you can use mem suffixes.
  99 Examples:
 100         # Initialize /dev/zram0 with 50MB disksize
 101         echo $((50*1024*1024)) > /sys/block/zram0/disksize
 102
 103         # Using mem suffixes
 104         echo 256K > /sys/block/zram0/disksize
 105         echo 512M > /sys/block/zram0/disksize
 106         echo 1G > /sys/block/zram0/disksize
 107
 108 Note:
 109 There is little point creating a zram of greater than twice the size of memory
 110 since we expect a 2:1 compression ratio. Note that zram uses about 0.1% of the
 111 size of the disk when not in use so a huge zram is wasteful.
 112
 113 5) Set memory limit: Optional
 114 Set memory limit by writing the value to sysfs node 'mem_limit'.
 115 The value can be either in bytes or you can use mem suffixes.
 116 In addition, you could change the value in runtime.
 117 Examples:
 118         # limit /dev/zram0 with 50MB memory
 119         echo $((50*1024*1024)) > /sys/block/zram0/mem_limit
 120
 121         # Using mem suffixes
 122         echo 256K > /sys/block/zram0/mem_limit
 123         echo 512M > /sys/block/zram0/mem_limit
 124         echo 1G > /sys/block/zram0/mem_limit
 125
 126         # To disable memory limit
 127         echo 0 > /sys/block/zram0/mem_limit
 128
 129 6) Activate:
 130         mkswap /dev/zram0
 131         swapon /dev/zram0
 132
 133         mkfs.ext4 /dev/zram1
 134         mount /dev/zram1 /tmp
 135
 136 7) Add/remove zram devices
 137
 138 zram provides a control interface, which enables dynamic (on-demand) device
 139 addition and removal.
 140
 141 In order to add a new /dev/zramX device, perform read operation on hot_add
 142 attribute. This will return either new device's device id (meaning that you
 143 can use /dev/zram<id>) or error code.
 144
 145 Example:
 146         cat /sys/class/zram-control/hot_add
 147         1
 148
 149 To remove the existing /dev/zramX device (where X is a device id)
 150 execute
 151         echo X > /sys/class/zram-control/hot_remove
 152
 153 8) Stats:
 154 Per-device statistics are exported as various nodes under /sys/block/zram<id>/
 155
 156 A brief description of exported device attributes. For more details please
 157 read Documentation/ABI/testing/sysfs-block-zram.
 158
 159 Name            access            description
 160 ----            ------            -----------
 161 disksize          RW    show and set the device's disk size
 162 initstate         RO    shows the initialization state of the device
 163 reset             WO    trigger device reset
 164 mem_used_max      WO    reset the `mem_used_max' counter (see later)
 165 mem_limit         WO    specifies the maximum amount of memory ZRAM can use
 166                         to store the compressed data
 167 writeback_limit   WO    specifies the maximum amount of write IO zram can
 168                         write out to backing device as 4KB unit
 169 max_comp_streams  RW    the number of possible concurrent compress operations
 170 comp_algorithm    RW    show and change the compression algorithm
 171 compact           WO    trigger memory compaction
 172 debug_stat        RO    this file is used for zram debugging purposes
 173 backing_dev       RW    set up backend storage for zram to write out
 174 idle              WO    mark allocated slot as idle
 175
 176
 177 User space is advised to use the following files to read the device statistics.
 178
 179 File /sys/block/zram<id>/stat
 180
 181 Represents block layer statistics. Read Documentation/block/stat.txt for
 182 details.
 183
 184 File /sys/block/zram<id>/io_stat
 185
 186 The stat file represents device's I/O statistics not accounted by block
 187 layer and, thus, not available in zram<id>/stat file. It consists of a
 188 single line of text and contains the following stats separated by
 189 whitespace:
 190  failed_reads     the number of failed reads
 191  failed_writes    the number of failed writes
 192  invalid_io       the number of non-page-size-aligned I/O requests
 193  notify_free      Depending on device usage scenario it may account
 194                   a) the number of pages freed because of swap slot free
 195                   notifications or b) the number of pages freed because of
 196                   REQ_OP_DISCARD requests sent by bio. The former ones are
 197                   sent to a swap block device when a swap slot is freed,
 198                   which implies that this disk is being used as a swap disk.
 199                   The latter ones are sent by filesystem mounted with
 200                   discard option, whenever some data blocks are getting
 201                   discarded.
 202
 203 File /sys/block/zram<id>/mm_stat
 204
 205 The stat file represents device's mm statistics. It consists of a single
 206 line of text and contains the following stats separated by whitespace:
 207  orig_data_size   uncompressed size of data stored in this disk.
 208                   This excludes same-element-filled pages (same_pages) since
 209                   no memory is allocated for them.
 210                   Unit: bytes
 211  compr_data_size  compressed size of data stored in this disk
 212  mem_used_total   the amount of memory allocated for this disk. This
 213                   includes allocator fragmentation and metadata overhead,
 214                   allocated for this disk. So, allocator space efficiency
 215                   can be calculated using compr_data_size and this statistic.
 216                   Unit: bytes
 217  mem_limit        the maximum amount of memory ZRAM can use to store
 218                   the compressed data
 219  mem_used_max     the maximum amount of memory zram have consumed to
 220                   store the data
 221  same_pages       the number of same element filled pages written to this disk.
 222                   No memory is allocated for such pages.
 223  pages_compacted  the number of pages freed during compaction
 224  huge_pages       the number of incompressible pages
 225
 226 File /sys/block/zram<id>/bd_stat
 227
 228 The stat file represents device's backing device statistics. It consists of
 229 a single line of text and contains the following stats separated by whitespace:
 230  bd_count       size of data written in backing device.
 231                 Unit: 4K bytes
 232  bd_reads       the number of reads from backing device
 233                 Unit: 4K bytes
 234  bd_writes      the number of writes to backing device
 235                 Unit: 4K bytes
 236
 237 9) Deactivate:
 238         swapoff /dev/zram0
 239         umount /dev/zram1
 240
 241 10) Reset:
 242         Write any positive value to 'reset' sysfs node
 243         echo 1 > /sys/block/zram0/reset
 244         echo 1 > /sys/block/zram1/reset
 245
 246         This frees all the memory allocated for the given device and
 247         resets the disksize to zero. You must set the disksize again
 248         before reusing the device.
 249
 250 * Optional Feature
 251
 252 = writeback
 253
 254 With CONFIG_ZRAM_WRITEBACK, zram can write idle/incompressible page
 255 to backing storage rather than keeping it in memory.
 256 To use the feature, admin should set up backing device via
 257
 258         "echo /dev/sda5 > /sys/block/zramX/backing_dev"
 259
 260 before disksize setting. It supports only partition at this moment.
 261 If admin want to use incompressible page writeback, they could do via
 262
 263         "echo huge > /sys/block/zramX/write"
 264
 265 To use idle page writeback, first, user need to declare zram pages
 266 as idle.
 267
 268         "echo all > /sys/block/zramX/idle"
 269
 270 From now on, any pages on zram are idle pages. The idle mark
 271 will be removed until someone request access of the block.
 272 IOW, unless there is access request, those pages are still idle pages.
 273
 274 Admin can request writeback of those idle pages at right timing via
 275
 276         "echo idle > /sys/block/zramX/writeback"
 277
 278 With the command, zram writeback idle pages from memory to the storage.
 279
 280 If there are lots of write IO with flash device, potentially, it has
 281 flash wearout problem so that admin needs to design write limitation
 282 to guarantee storage health for entire product life.
 283 To overcome the concern, zram supports "writeback_limit".
 284 The "writeback_limit"'s default value is 0 so that it doesn't limit
 285 any writeback. If admin want to measure writeback count in a certain
 286 period, he could know it via /sys/block/zram0/bd_stat's 3rd column.
 287
 288 If admin want to limit writeback as per-day 400M, he could do it
 289 like below.
 290
 291     MB_SHIFT=20
 292     4K_SHIFT=12
 293     echo $((400<<MB_SHIFT>>4K_SHIFT)) > \
 294             /sys/block/zram0/writeback_limit.
 295
 296 If admin want to allow further write again, he could do it like below
 297
 298     echo 0 > /sys/block/zram0/writeback_limit
 299
 300 If admin want to see remaining writeback budget since he set,
 301
 302     cat /sys/block/zram0/writeback_limit
 303
 304 The writeback_limit count will reset whenever you reset zram(e.g.,
 305 system reboot, echo 1 > /sys/block/zramX/reset) so keeping how many of
 306 writeback happened until you reset the zram to allocate extra writeback
 307 budget in next setting is user's job.
 308
 309 = memory tracking
 310
 311 With CONFIG_ZRAM_MEMORY_TRACKING, user can know information of the
 312 zram block. It could be useful to catch cold or incompressible
 313 pages of the process with*pagemap.
 314 If you enable the feature, you could see block state via
 315 /sys/kernel/debug/zram/zram0/block_state". The output is as follows,
 316
 317           300    75.033841 .wh.
 318           301    63.806904 s...
 319           302    63.806919 ..hi
 320
 321 First column is zram's block index.
 322 Second column is access time since the system was booted
 323 Third column is state of the block.
 324 (s: same page
 325 w: written page to backing store
 326 h: huge page
 327 i: idle page)
 328
 329 First line of above example says 300th block is accessed at 75.033841sec
 330 and the block's state is huge so it is written back to the backing
 331 storage. It's a debugging feature so anyone shouldn't rely on it to work
 332 properly.
 333
 334 Nitin Gupta
 335 ngupta@vflare.org