cachepc-linux

Fork of AMDESE/linux with modifications for CachePC side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-linux
Log | Files | Refs | README | LICENSE | sfeed.txt

usage.rst (29089B)


      1.. SPDX-License-Identifier: GPL-2.0
      2
      3===============
      4Detailed Usages
      5===============
      6
      7DAMON provides below interfaces for different users.
      8
      9- *DAMON user space tool.*
     10  `This <https://github.com/awslabs/damo>`_ is for privileged people such as
     11  system administrators who want a just-working human-friendly interface.
     12  Using this, users can use the DAMON’s major features in a human-friendly way.
     13  It may not be highly tuned for special cases, though.  It supports both
     14  virtual and physical address spaces monitoring.  For more detail, please
     15  refer to its `usage document
     16  <https://github.com/awslabs/damo/blob/next/USAGE.md>`_.
     17- *sysfs interface.*
     18  :ref:`This <sysfs_interface>` is for privileged user space programmers who
     19  want more optimized use of DAMON.  Using this, users can use DAMON’s major
     20  features by reading from and writing to special sysfs files.  Therefore,
     21  you can write and use your personalized DAMON sysfs wrapper programs that
     22  reads/writes the sysfs files instead of you.  The `DAMON user space tool
     23  <https://github.com/awslabs/damo>`_ is one example of such programs.  It
     24  supports both virtual and physical address spaces monitoring.  Note that this
     25  interface provides only simple :ref:`statistics <damos_stats>` for the
     26  monitoring results.  For detailed monitoring results, DAMON provides a
     27  :ref:`tracepoint <tracepoint>`.
     28- *debugfs interface.*
     29  :ref:`This <debugfs_interface>` is almost identical to :ref:`sysfs interface
     30  <sysfs_interface>`.  This will be removed after next LTS kernel is released,
     31  so users should move to the :ref:`sysfs interface <sysfs_interface>`.
     32- *Kernel Space Programming Interface.*
     33  :doc:`This </vm/damon/api>` is for kernel space programmers.  Using this,
     34  users can utilize every feature of DAMON most flexibly and efficiently by
     35  writing kernel space DAMON application programs for you.  You can even extend
     36  DAMON for various address spaces.  For detail, please refer to the interface
     37  :doc:`document </vm/damon/api>`.
     38
     39.. _sysfs_interface:
     40
     41sysfs Interface
     42===============
     43
     44DAMON sysfs interface is built when ``CONFIG_DAMON_SYSFS`` is defined.  It
     45creates multiple directories and files under its sysfs directory,
     46``<sysfs>/kernel/mm/damon/``.  You can control DAMON by writing to and reading
     47from the files under the directory.
     48
     49For a short example, users can monitor the virtual address space of a given
     50workload as below. ::
     51
     52    # cd /sys/kernel/mm/damon/admin/
     53    # echo 1 > kdamonds/nr && echo 1 > kdamonds/0/contexts/nr
     54    # echo vaddr > kdamonds/0/contexts/0/operations
     55    # echo 1 > kdamonds/0/contexts/0/targets/nr
     56    # echo $(pidof <workload>) > kdamonds/0/contexts/0/targets/0/pid
     57    # echo on > kdamonds/0/state
     58
     59Files Hierarchy
     60---------------
     61
     62The files hierarchy of DAMON sysfs interface is shown below.  In the below
     63figure, parents-children relations are represented with indentations, each
     64directory is having ``/`` suffix, and files in each directory are separated by
     65comma (","). ::
     66
     67    /sys/kernel/mm/damon/admin
     68    │ kdamonds/nr_kdamonds
     69    │ │ 0/state,pid
     70    │ │ │ contexts/nr_contexts
     71    │ │ │ │ 0/avail_operations,operations
     72    │ │ │ │ │ monitoring_attrs/
     73    │ │ │ │ │ │ intervals/sample_us,aggr_us,update_us
     74    │ │ │ │ │ │ nr_regions/min,max
     75    │ │ │ │ │ targets/nr_targets
     76    │ │ │ │ │ │ 0/pid_target
     77    │ │ │ │ │ │ │ regions/nr_regions
     78    │ │ │ │ │ │ │ │ 0/start,end
     79    │ │ │ │ │ │ │ │ ...
     80    │ │ │ │ │ │ ...
     81    │ │ │ │ │ schemes/nr_schemes
     82    │ │ │ │ │ │ 0/action
     83    │ │ │ │ │ │ │ access_pattern/
     84    │ │ │ │ │ │ │ │ sz/min,max
     85    │ │ │ │ │ │ │ │ nr_accesses/min,max
     86    │ │ │ │ │ │ │ │ age/min,max
     87    │ │ │ │ │ │ │ quotas/ms,bytes,reset_interval_ms
     88    │ │ │ │ │ │ │ │ weights/sz_permil,nr_accesses_permil,age_permil
     89    │ │ │ │ │ │ │ watermarks/metric,interval_us,high,mid,low
     90    │ │ │ │ │ │ │ stats/nr_tried,sz_tried,nr_applied,sz_applied,qt_exceeds
     91    │ │ │ │ │ │ ...
     92    │ │ │ │ ...
     93    │ │ ...
     94
     95Root
     96----
     97
     98The root of the DAMON sysfs interface is ``<sysfs>/kernel/mm/damon/``, and it
     99has one directory named ``admin``.  The directory contains the files for
    100privileged user space programs' control of DAMON.  User space tools or deamons
    101having the root permission could use this directory.
    102
    103kdamonds/
    104---------
    105
    106The monitoring-related information including request specifications and results
    107are called DAMON context.  DAMON executes each context with a kernel thread
    108called kdamond, and multiple kdamonds could run in parallel.
    109
    110Under the ``admin`` directory, one directory, ``kdamonds``, which has files for
    111controlling the kdamonds exist.  In the beginning, this directory has only one
    112file, ``nr_kdamonds``.  Writing a number (``N``) to the file creates the number
    113of child directories named ``0`` to ``N-1``.  Each directory represents each
    114kdamond.
    115
    116kdamonds/<N>/
    117-------------
    118
    119In each kdamond directory, two files (``state`` and ``pid``) and one directory
    120(``contexts``) exist.
    121
    122Reading ``state`` returns ``on`` if the kdamond is currently running, or
    123``off`` if it is not running.  Writing ``on`` or ``off`` makes the kdamond be
    124in the state.  Writing ``commit`` to the ``state`` file makes kdamond reads the
    125user inputs in the sysfs files except ``state`` file again.  Writing
    126``update_schemes_stats`` to ``state`` file updates the contents of stats files
    127for each DAMON-based operation scheme of the kdamond.  For details of the
    128stats, please refer to :ref:`stats section <sysfs_schemes_stats>`.
    129
    130If the state is ``on``, reading ``pid`` shows the pid of the kdamond thread.
    131
    132``contexts`` directory contains files for controlling the monitoring contexts
    133that this kdamond will execute.
    134
    135kdamonds/<N>/contexts/
    136----------------------
    137
    138In the beginning, this directory has only one file, ``nr_contexts``.  Writing a
    139number (``N``) to the file creates the number of child directories named as
    140``0`` to ``N-1``.  Each directory represents each monitoring context.  At the
    141moment, only one context per kdamond is supported, so only ``0`` or ``1`` can
    142be written to the file.
    143
    144contexts/<N>/
    145-------------
    146
    147In each context directory, two files (``avail_operations`` and ``operations``)
    148and three directories (``monitoring_attrs``, ``targets``, and ``schemes``)
    149exist.
    150
    151DAMON supports multiple types of monitoring operations, including those for
    152virtual address space and the physical address space.  You can get the list of
    153available monitoring operations set on the currently running kernel by reading
    154``avail_operations`` file.  Based on the kernel configuration, the file will
    155list some or all of below keywords.
    156
    157 - vaddr: Monitor virtual address spaces of specific processes
    158 - fvaddr: Monitor fixed virtual address ranges
    159 - paddr: Monitor the physical address space of the system
    160
    161Please refer to :ref:`regions sysfs directory <sysfs_regions>` for detailed
    162differences between the operations sets in terms of the monitoring target
    163regions.
    164
    165You can set and get what type of monitoring operations DAMON will use for the
    166context by writing one of the keywords listed in ``avail_operations`` file and
    167reading from the ``operations`` file.
    168
    169contexts/<N>/monitoring_attrs/
    170------------------------------
    171
    172Files for specifying attributes of the monitoring including required quality
    173and efficiency of the monitoring are in ``monitoring_attrs`` directory.
    174Specifically, two directories, ``intervals`` and ``nr_regions`` exist in this
    175directory.
    176
    177Under ``intervals`` directory, three files for DAMON's sampling interval
    178(``sample_us``), aggregation interval (``aggr_us``), and update interval
    179(``update_us``) exist.  You can set and get the values in micro-seconds by
    180writing to and reading from the files.
    181
    182Under ``nr_regions`` directory, two files for the lower-bound and upper-bound
    183of DAMON's monitoring regions (``min`` and ``max``, respectively), which
    184controls the monitoring overhead, exist.  You can set and get the values by
    185writing to and rading from the files.
    186
    187For more details about the intervals and monitoring regions range, please refer
    188to the Design document (:doc:`/vm/damon/design`).
    189
    190contexts/<N>/targets/
    191---------------------
    192
    193In the beginning, this directory has only one file, ``nr_targets``.  Writing a
    194number (``N``) to the file creates the number of child directories named ``0``
    195to ``N-1``.  Each directory represents each monitoring target.
    196
    197targets/<N>/
    198------------
    199
    200In each target directory, one file (``pid_target``) and one directory
    201(``regions``) exist.
    202
    203If you wrote ``vaddr`` to the ``contexts/<N>/operations``, each target should
    204be a process.  You can specify the process to DAMON by writing the pid of the
    205process to the ``pid_target`` file.
    206
    207.. _sysfs_regions:
    208
    209targets/<N>/regions
    210-------------------
    211
    212When ``vaddr`` monitoring operations set is being used (``vaddr`` is written to
    213the ``contexts/<N>/operations`` file), DAMON automatically sets and updates the
    214monitoring target regions so that entire memory mappings of target processes
    215can be covered.  However, users could want to set the initial monitoring region
    216to specific address ranges.
    217
    218In contrast, DAMON do not automatically sets and updates the monitoring target
    219regions when ``fvaddr`` or ``paddr`` monitoring operations sets are being used
    220(``fvaddr`` or ``paddr`` have written to the ``contexts/<N>/operations``).
    221Therefore, users should set the monitoring target regions by themselves in the
    222cases.
    223
    224For such cases, users can explicitly set the initial monitoring target regions
    225as they want, by writing proper values to the files under this directory.
    226
    227In the beginning, this directory has only one file, ``nr_regions``.  Writing a
    228number (``N``) to the file creates the number of child directories named ``0``
    229to ``N-1``.  Each directory represents each initial monitoring target region.
    230
    231regions/<N>/
    232------------
    233
    234In each region directory, you will find two files (``start`` and ``end``).  You
    235can set and get the start and end addresses of the initial monitoring target
    236region by writing to and reading from the files, respectively.
    237
    238contexts/<N>/schemes/
    239---------------------
    240
    241For usual DAMON-based data access aware memory management optimizations, users
    242would normally want the system to apply a memory management action to a memory
    243region of a specific access pattern.  DAMON receives such formalized operation
    244schemes from the user and applies those to the target memory regions.  Users
    245can get and set the schemes by reading from and writing to files under this
    246directory.
    247
    248In the beginning, this directory has only one file, ``nr_schemes``.  Writing a
    249number (``N``) to the file creates the number of child directories named ``0``
    250to ``N-1``.  Each directory represents each DAMON-based operation scheme.
    251
    252schemes/<N>/
    253------------
    254
    255In each scheme directory, four directories (``access_pattern``, ``quotas``,
    256``watermarks``, and ``stats``) and one file (``action``) exist.
    257
    258The ``action`` file is for setting and getting what action you want to apply to
    259memory regions having specific access pattern of the interest.  The keywords
    260that can be written to and read from the file and their meaning are as below.
    261
    262 - ``willneed``: Call ``madvise()`` for the region with ``MADV_WILLNEED``
    263 - ``cold``: Call ``madvise()`` for the region with ``MADV_COLD``
    264 - ``pageout``: Call ``madvise()`` for the region with ``MADV_PAGEOUT``
    265 - ``hugepage``: Call ``madvise()`` for the region with ``MADV_HUGEPAGE``
    266 - ``nohugepage``: Call ``madvise()`` for the region with ``MADV_NOHUGEPAGE``
    267 - ``stat``: Do nothing but count the statistics
    268
    269schemes/<N>/access_pattern/
    270---------------------------
    271
    272The target access pattern of each DAMON-based operation scheme is constructed
    273with three ranges including the size of the region in bytes, number of
    274monitored accesses per aggregate interval, and number of aggregated intervals
    275for the age of the region.
    276
    277Under the ``access_pattern`` directory, three directories (``sz``,
    278``nr_accesses``, and ``age``) each having two files (``min`` and ``max``)
    279exist.  You can set and get the access pattern for the given scheme by writing
    280to and reading from the ``min`` and ``max`` files under ``sz``,
    281``nr_accesses``, and ``age`` directories, respectively.
    282
    283schemes/<N>/quotas/
    284-------------------
    285
    286Optimal ``target access pattern`` for each ``action`` is workload dependent, so
    287not easy to find.  Worse yet, setting a scheme of some action too aggressive
    288can cause severe overhead.  To avoid such overhead, users can limit time and
    289size quota for each scheme.  In detail, users can ask DAMON to try to use only
    290up to specific time (``time quota``) for applying the action, and to apply the
    291action to only up to specific amount (``size quota``) of memory regions having
    292the target access pattern within a given time interval (``reset interval``).
    293
    294When the quota limit is expected to be exceeded, DAMON prioritizes found memory
    295regions of the ``target access pattern`` based on their size, access frequency,
    296and age.  For personalized prioritization, users can set the weights for the
    297three properties.
    298
    299Under ``quotas`` directory, three files (``ms``, ``bytes``,
    300``reset_interval_ms``) and one directory (``weights``) having three files
    301(``sz_permil``, ``nr_accesses_permil``, and ``age_permil``) in it exist.
    302
    303You can set the ``time quota`` in milliseconds, ``size quota`` in bytes, and
    304``reset interval`` in milliseconds by writing the values to the three files,
    305respectively.  You can also set the prioritization weights for size, access
    306frequency, and age in per-thousand unit by writing the values to the three
    307files under the ``weights`` directory.
    308
    309schemes/<N>/watermarks/
    310-----------------------
    311
    312To allow easy activation and deactivation of each scheme based on system
    313status, DAMON provides a feature called watermarks.  The feature receives five
    314values called ``metric``, ``interval``, ``high``, ``mid``, and ``low``.  The
    315``metric`` is the system metric such as free memory ratio that can be measured.
    316If the metric value of the system is higher than the value in ``high`` or lower
    317than ``low`` at the memoent, the scheme is deactivated.  If the value is lower
    318than ``mid``, the scheme is activated.
    319
    320Under the watermarks directory, five files (``metric``, ``interval_us``,
    321``high``, ``mid``, and ``low``) for setting each value exist.  You can set and
    322get the five values by writing to the files, respectively.
    323
    324Keywords and meanings of those that can be written to the ``metric`` file are
    325as below.
    326
    327 - none: Ignore the watermarks
    328 - free_mem_rate: System's free memory rate (per thousand)
    329
    330The ``interval`` should written in microseconds unit.
    331
    332.. _sysfs_schemes_stats:
    333
    334schemes/<N>/stats/
    335------------------
    336
    337DAMON counts the total number and bytes of regions that each scheme is tried to
    338be applied, the two numbers for the regions that each scheme is successfully
    339applied, and the total number of the quota limit exceeds.  This statistics can
    340be used for online analysis or tuning of the schemes.
    341
    342The statistics can be retrieved by reading the files under ``stats`` directory
    343(``nr_tried``, ``sz_tried``, ``nr_applied``, ``sz_applied``, and
    344``qt_exceeds``), respectively.  The files are not updated in real time, so you
    345should ask DAMON sysfs interface to updte the content of the files for the
    346stats by writing a special keyword, ``update_schemes_stats`` to the relevant
    347``kdamonds/<N>/state`` file.
    348
    349Example
    350~~~~~~~
    351
    352Below commands applies a scheme saying "If a memory region of size in [4KiB,
    3538KiB] is showing accesses per aggregate interval in [0, 5] for aggregate
    354interval in [10, 20], page out the region.  For the paging out, use only up to
    35510ms per second, and also don't page out more than 1GiB per second.  Under the
    356limitation, page out memory regions having longer age first.  Also, check the
    357free memory rate of the system every 5 seconds, start the monitoring and paging
    358out when the free memory rate becomes lower than 50%, but stop it if the free
    359memory rate becomes larger than 60%, or lower than 30%". ::
    360
    361    # cd <sysfs>/kernel/mm/damon/admin
    362    # # populate directories
    363    # echo 1 > kdamonds/nr_kdamonds; echo 1 > kdamonds/0/contexts/nr_contexts;
    364    # echo 1 > kdamonds/0/contexts/0/schemes/nr_schemes
    365    # cd kdamonds/0/contexts/0/schemes/0
    366    # # set the basic access pattern and the action
    367    # echo 4096 > access_patterns/sz/min
    368    # echo 8192 > access_patterns/sz/max
    369    # echo 0 > access_patterns/nr_accesses/min
    370    # echo 5 > access_patterns/nr_accesses/max
    371    # echo 10 > access_patterns/age/min
    372    # echo 20 > access_patterns/age/max
    373    # echo pageout > action
    374    # # set quotas
    375    # echo 10 > quotas/ms
    376    # echo $((1024*1024*1024)) > quotas/bytes
    377    # echo 1000 > quotas/reset_interval_ms
    378    # # set watermark
    379    # echo free_mem_rate > watermarks/metric
    380    # echo 5000000 > watermarks/interval_us
    381    # echo 600 > watermarks/high
    382    # echo 500 > watermarks/mid
    383    # echo 300 > watermarks/low
    384
    385Please note that it's highly recommended to use user space tools like `damo
    386<https://github.com/awslabs/damo>`_ rather than manually reading and writing
    387the files as above.  Above is only for an example.
    388
    389.. _debugfs_interface:
    390
    391debugfs Interface
    392=================
    393
    394DAMON exports eight files, ``attrs``, ``target_ids``, ``init_regions``,
    395``schemes``, ``monitor_on``, ``kdamond_pid``, ``mk_contexts`` and
    396``rm_contexts`` under its debugfs directory, ``<debugfs>/damon/``.
    397
    398
    399Attributes
    400----------
    401
    402Users can get and set the ``sampling interval``, ``aggregation interval``,
    403``update interval``, and min/max number of monitoring target regions by
    404reading from and writing to the ``attrs`` file.  To know about the monitoring
    405attributes in detail, please refer to the :doc:`/vm/damon/design`.  For
    406example, below commands set those values to 5 ms, 100 ms, 1,000 ms, 10 and
    4071000, and then check it again::
    408
    409    # cd <debugfs>/damon
    410    # echo 5000 100000 1000000 10 1000 > attrs
    411    # cat attrs
    412    5000 100000 1000000 10 1000
    413
    414
    415Target IDs
    416----------
    417
    418Some types of address spaces supports multiple monitoring target.  For example,
    419the virtual memory address spaces monitoring can have multiple processes as the
    420monitoring targets.  Users can set the targets by writing relevant id values of
    421the targets to, and get the ids of the current targets by reading from the
    422``target_ids`` file.  In case of the virtual address spaces monitoring, the
    423values should be pids of the monitoring target processes.  For example, below
    424commands set processes having pids 42 and 4242 as the monitoring targets and
    425check it again::
    426
    427    # cd <debugfs>/damon
    428    # echo 42 4242 > target_ids
    429    # cat target_ids
    430    42 4242
    431
    432Users can also monitor the physical memory address space of the system by
    433writing a special keyword, "``paddr\n``" to the file.  Because physical address
    434space monitoring doesn't support multiple targets, reading the file will show a
    435fake value, ``42``, as below::
    436
    437    # cd <debugfs>/damon
    438    # echo paddr > target_ids
    439    # cat target_ids
    440    42
    441
    442Note that setting the target ids doesn't start the monitoring.
    443
    444
    445Initial Monitoring Target Regions
    446---------------------------------
    447
    448In case of the virtual address space monitoring, DAMON automatically sets and
    449updates the monitoring target regions so that entire memory mappings of target
    450processes can be covered.  However, users can want to limit the monitoring
    451region to specific address ranges, such as the heap, the stack, or specific
    452file-mapped area.  Or, some users can know the initial access pattern of their
    453workloads and therefore want to set optimal initial regions for the 'adaptive
    454regions adjustment'.
    455
    456In contrast, DAMON do not automatically sets and updates the monitoring target
    457regions in case of physical memory monitoring.  Therefore, users should set the
    458monitoring target regions by themselves.
    459
    460In such cases, users can explicitly set the initial monitoring target regions
    461as they want, by writing proper values to the ``init_regions`` file.  Each line
    462of the input should represent one region in below form.::
    463
    464    <target idx> <start address> <end address>
    465
    466The ``target idx`` should be the index of the target in ``target_ids`` file,
    467starting from ``0``, and the regions should be passed in address order.  For
    468example, below commands will set a couple of address ranges, ``1-100`` and
    469``100-200`` as the initial monitoring target region of pid 42, which is the
    470first one (index ``0``) in ``target_ids``, and another couple of address
    471ranges, ``20-40`` and ``50-100`` as that of pid 4242, which is the second one
    472(index ``1``) in ``target_ids``.::
    473
    474    # cd <debugfs>/damon
    475    # cat target_ids
    476    42 4242
    477    # echo "0   1       100
    478            0   100     200
    479            1   20      40
    480            1   50      100" > init_regions
    481
    482Note that this sets the initial monitoring target regions only.  In case of
    483virtual memory monitoring, DAMON will automatically updates the boundary of the
    484regions after one ``update interval``.  Therefore, users should set the
    485``update interval`` large enough in this case, if they don't want the
    486update.
    487
    488
    489Schemes
    490-------
    491
    492For usual DAMON-based data access aware memory management optimizations, users
    493would simply want the system to apply a memory management action to a memory
    494region of a specific access pattern.  DAMON receives such formalized operation
    495schemes from the user and applies those to the target processes.
    496
    497Users can get and set the schemes by reading from and writing to ``schemes``
    498debugfs file.  Reading the file also shows the statistics of each scheme.  To
    499the file, each of the schemes should be represented in each line in below
    500form::
    501
    502    <target access pattern> <action> <quota> <watermarks>
    503
    504You can disable schemes by simply writing an empty string to the file.
    505
    506Target Access Pattern
    507~~~~~~~~~~~~~~~~~~~~~
    508
    509The ``<target access pattern>`` is constructed with three ranges in below
    510form::
    511
    512    min-size max-size min-acc max-acc min-age max-age
    513
    514Specifically, bytes for the size of regions (``min-size`` and ``max-size``),
    515number of monitored accesses per aggregate interval for access frequency
    516(``min-acc`` and ``max-acc``), number of aggregate intervals for the age of
    517regions (``min-age`` and ``max-age``) are specified.  Note that the ranges are
    518closed interval.
    519
    520Action
    521~~~~~~
    522
    523The ``<action>`` is a predefined integer for memory management actions, which
    524DAMON will apply to the regions having the target access pattern.  The
    525supported numbers and their meanings are as below.
    526
    527 - 0: Call ``madvise()`` for the region with ``MADV_WILLNEED``
    528 - 1: Call ``madvise()`` for the region with ``MADV_COLD``
    529 - 2: Call ``madvise()`` for the region with ``MADV_PAGEOUT``
    530 - 3: Call ``madvise()`` for the region with ``MADV_HUGEPAGE``
    531 - 4: Call ``madvise()`` for the region with ``MADV_NOHUGEPAGE``
    532 - 5: Do nothing but count the statistics
    533
    534Quota
    535~~~~~
    536
    537Optimal ``target access pattern`` for each ``action`` is workload dependent, so
    538not easy to find.  Worse yet, setting a scheme of some action too aggressive
    539can cause severe overhead.  To avoid such overhead, users can limit time and
    540size quota for the scheme via the ``<quota>`` in below form::
    541
    542    <ms> <sz> <reset interval> <priority weights>
    543
    544This makes DAMON to try to use only up to ``<ms>`` milliseconds for applying
    545the action to memory regions of the ``target access pattern`` within the
    546``<reset interval>`` milliseconds, and to apply the action to only up to
    547``<sz>`` bytes of memory regions within the ``<reset interval>``.  Setting both
    548``<ms>`` and ``<sz>`` zero disables the quota limits.
    549
    550When the quota limit is expected to be exceeded, DAMON prioritizes found memory
    551regions of the ``target access pattern`` based on their size, access frequency,
    552and age.  For personalized prioritization, users can set the weights for the
    553three properties in ``<priority weights>`` in below form::
    554
    555    <size weight> <access frequency weight> <age weight>
    556
    557Watermarks
    558~~~~~~~~~~
    559
    560Some schemes would need to run based on current value of the system's specific
    561metrics like free memory ratio.  For such cases, users can specify watermarks
    562for the condition.::
    563
    564    <metric> <check interval> <high mark> <middle mark> <low mark>
    565
    566``<metric>`` is a predefined integer for the metric to be checked.  The
    567supported numbers and their meanings are as below.
    568
    569 - 0: Ignore the watermarks
    570 - 1: System's free memory rate (per thousand)
    571
    572The value of the metric is checked every ``<check interval>`` microseconds.
    573
    574If the value is higher than ``<high mark>`` or lower than ``<low mark>``, the
    575scheme is deactivated.  If the value is lower than ``<mid mark>``, the scheme
    576is activated.
    577
    578.. _damos_stats:
    579
    580Statistics
    581~~~~~~~~~~
    582
    583It also counts the total number and bytes of regions that each scheme is tried
    584to be applied, the two numbers for the regions that each scheme is successfully
    585applied, and the total number of the quota limit exceeds.  This statistics can
    586be used for online analysis or tuning of the schemes.
    587
    588The statistics can be shown by reading the ``schemes`` file.  Reading the file
    589will show each scheme you entered in each line, and the five numbers for the
    590statistics will be added at the end of each line.
    591
    592Example
    593~~~~~~~
    594
    595Below commands applies a scheme saying "If a memory region of size in [4KiB,
    5968KiB] is showing accesses per aggregate interval in [0, 5] for aggregate
    597interval in [10, 20], page out the region.  For the paging out, use only up to
    59810ms per second, and also don't page out more than 1GiB per second.  Under the
    599limitation, page out memory regions having longer age first.  Also, check the
    600free memory rate of the system every 5 seconds, start the monitoring and paging
    601out when the free memory rate becomes lower than 50%, but stop it if the free
    602memory rate becomes larger than 60%, or lower than 30%".::
    603
    604    # cd <debugfs>/damon
    605    # scheme="4096 8192  0 5    10 20    2"  # target access pattern and action
    606    # scheme+=" 10 $((1024*1024*1024)) 1000" # quotas
    607    # scheme+=" 0 0 100"                     # prioritization weights
    608    # scheme+=" 1 5000000 600 500 300"       # watermarks
    609    # echo "$scheme" > schemes
    610
    611
    612Turning On/Off
    613--------------
    614
    615Setting the files as described above doesn't incur effect unless you explicitly
    616start the monitoring.  You can start, stop, and check the current status of the
    617monitoring by writing to and reading from the ``monitor_on`` file.  Writing
    618``on`` to the file starts the monitoring of the targets with the attributes.
    619Writing ``off`` to the file stops those.  DAMON also stops if every target
    620process is terminated.  Below example commands turn on, off, and check the
    621status of DAMON::
    622
    623    # cd <debugfs>/damon
    624    # echo on > monitor_on
    625    # echo off > monitor_on
    626    # cat monitor_on
    627    off
    628
    629Please note that you cannot write to the above-mentioned debugfs files while
    630the monitoring is turned on.  If you write to the files while DAMON is running,
    631an error code such as ``-EBUSY`` will be returned.
    632
    633
    634Monitoring Thread PID
    635---------------------
    636
    637DAMON does requested monitoring with a kernel thread called ``kdamond``.  You
    638can get the pid of the thread by reading the ``kdamond_pid`` file.  When the
    639monitoring is turned off, reading the file returns ``none``. ::
    640
    641    # cd <debugfs>/damon
    642    # cat monitor_on
    643    off
    644    # cat kdamond_pid
    645    none
    646    # echo on > monitor_on
    647    # cat kdamond_pid
    648    18594
    649
    650
    651Using Multiple Monitoring Threads
    652---------------------------------
    653
    654One ``kdamond`` thread is created for each monitoring context.  You can create
    655and remove monitoring contexts for multiple ``kdamond`` required use case using
    656the ``mk_contexts`` and ``rm_contexts`` files.
    657
    658Writing the name of the new context to the ``mk_contexts`` file creates a
    659directory of the name on the DAMON debugfs directory.  The directory will have
    660DAMON debugfs files for the context. ::
    661
    662    # cd <debugfs>/damon
    663    # ls foo
    664    # ls: cannot access 'foo': No such file or directory
    665    # echo foo > mk_contexts
    666    # ls foo
    667    # attrs  init_regions  kdamond_pid  schemes  target_ids
    668
    669If the context is not needed anymore, you can remove it and the corresponding
    670directory by putting the name of the context to the ``rm_contexts`` file. ::
    671
    672    # echo foo > rm_contexts
    673    # ls foo
    674    # ls: cannot access 'foo': No such file or directory
    675
    676Note that ``mk_contexts``, ``rm_contexts``, and ``monitor_on`` files are in the
    677root directory only.
    678
    679
    680.. _tracepoint:
    681
    682Tracepoint for Monitoring Results
    683=================================
    684
    685DAMON provides the monitoring results via a tracepoint,
    686``damon:damon_aggregated``.  While the monitoring is turned on, you could
    687record the tracepoint events and show results using tracepoint supporting tools
    688like ``perf``.  For example::
    689
    690    # echo on > monitor_on
    691    # perf record -e damon:damon_aggregated &
    692    # sleep 5
    693    # kill 9 $(pidof perf)
    694    # echo off > monitor_on
    695    # perf script