| Age | Commit message (Collapse) | Author |
|---|
|
My legal and preferred first names are SeongJae and SJ, respectively. I
was using the legal name for commits and tags, while using the preferred
name for conversations. It sometimes confuses people including myself.
Consistently use the preferred name.
Together remove copyright notes on files. Those are only confusing for
people who are not familiar with the law. Meanwhile, we can infer the
information in a better way from git logs and public information.
Link: https://lore.kernel.org/20260630013820.143366-1-sj@kernel.org
Signed-off-by: SJ Park <sj@kernel.org>
Acked-by: Lorenzo Stoakes <ljs@kernel.org>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Fix below kernel document build warning:
WARNING: ../include/linux/damon.h:909 struct member 'rnd_state' not described in 'damon_ctx'
Link: https://lore.kernel.org/20260628220808.98931-3-sj@kernel.org
Fixes: 9012c4e647df ("mm/damon: replace damon_rand() with a per-ctx lockless PRNG")
Signed-off-by: SJ Park <sj@kernel.org>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Closes: https://lore.kernel.org/4df95955-b255-4e5a-90c4-35db02f3111f@infradead.org
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The two fields of damon_ctx struct dont have their kernel-doc comments.
That causes kernel document builds to warn. Fix those.
This patch (of 2):
Fix below document build warning:
WARNING: ../include/linux/damon.h:909 struct member 'probes' not described in 'damon_ctx'
Link: https://lore.kernel.org/20260628220808.98931-1-sj@kernel.org
Link: https://lore.kernel.org/20260628220808.98931-2-sj@kernel.org
Fixes: 18c777859f28 ("mm/damon/core: embed damon_probe objects in damon_ctx")
Signed-off-by: SJ Park <sj@kernel.org>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Closes: https://lore.kernel.org/4df95955-b255-4e5a-90c4-35db02f3111f@infradead.org
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Fix three issues in damon_hot_score() that comes from wrong handling of
extreme (zero or too high) monitoring intervals user setup.
When the user sets sampling interval zero, damon_max_nr_accesses(), which
is called from damon_hot_score(), causes a divide-by-zero. Needless to
say, it is a problem.
When the user sets the aggregation interval zero, the function returns
zero. It is wrong, since the real maximum nr_acceses in the setup should
be one. Worse yet, it can cause another divide-by-zero from its caller,
damon_hot_score(), since it uses damon_max_nr_accesses() return value as a
denominator.
When the user sets the aggregation interval very high, damon_hot_score()
could return a value out of [0, DAMOS_MAX_SCORE] range. Since the return
value is used as an index to the regions_score_histogram array, which is
DAMOS_MAX_SCORE+1 size, it causes out of bounds array access.
The issues can be relatively easily reproduced like below. The sysfs
write permission is required, though.
# ./damo start --damos_action lru_prio --damos_quota_space 100M \
--damos_quota_interval 1s
# cd /sys/kernel/mm/damon/admin/kdamonds/0
# echo 0 > contexts/0/monitoring_attrs/intervals/sample_us
# echo 0 > contexts/0/monitoring_attrs/intervals/aggr_us
# echo commit > state
# dmesg
[...]
[ 131.329762] Oops: divide error: 0000 [#1] SMP NOPTI
[...]
[ 131.336089] RIP: 0010:damon_hot_score+0x27/0xd0
[...]
Fix the divide-by-zero intervals problems by explicitly handling the zero
intervals in damon_max_nr_accesses(). Fix the out-of-bound array access
by applying [0, DAMOS_MAX_SCORE] bounds before returning from
damon_hot_score().
The issue was discovered [1] by Sashiko.
Link: https://lore.kernel.org/20260623135834.67189-1-sj@kernel.org
Link: https://lore.kernel.org/20260619202459.145010-1-sj@kernel.org [1]
Fixes: 198f0f4c58b9 ("mm/damon/vaddr,paddr: support pageout prioritization")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> # 5.16.x
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: fix macro arguments and clarify quota goals doc",
v2.
This patch (of 2):
The DAMON iterator macros do not wrap their pointer arguments with
parentheses. This can cause build failures when the argument is a complex
expression due to operator precedence issues.
Add missing parentheses around the arguments in the following macros
to prevent potential build failures:
- damon_for_each_region()
- damon_for_each_region_from()
- damon_for_each_region_safe()
- damos_for_each_quota_goal()
Link: https://lore.kernel.org/20260521202020.126500-1-maksym.shcherba@lnu.edu.ua
Link: https://lore.kernel.org/20260521202020.126500-2-maksym.shcherba@lnu.edu.ua
Signed-off-by: Maksym Shcherba <maksym.shcherba@lnu.edu.ua>
Reviewed-by: SeongJae Park <sj@kernel.org>
Assisted-by: Antigravity:Gemini-3.1-Pro
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_destroy_region() is being used by only DAMON core, but exposed to
DAMON API callers. Exposing something that is not really being used by
others will only increase the maintenance cost. Hide it.
Link: https://lore.kernel.org/20260522154026.80546-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_insert_region() is being used by only DAMON core, but exposed to
DAMON API callers. Exposing something that is not really being used by
others will only increase the maintenance cost. Hide it.
Link: https://lore.kernel.org/20260522154026.80546-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_add_region() is being used by only DAMON core, but exposed to DAMON
API callers. Exposing something that is not really being used by others
will only increase the maintenance cost. Hide it.
Link: https://lore.kernel.org/20260522154026.80546-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Find and set the memcg_id for damon_filter from the user-passed memory
cgroup path when updating the DAMON input parameters.
Link: https://lore.kernel.org/20260518234119.97569-27-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Belonging memory cgoup is another data attribute that can be useful to
monitor. Introduce a new DAMON filter type, namely
DAMON_FILTER_TYPE_MEMCG, for monitoring of this attribute.
Link: https://lore.kernel.org/20260518234119.97569-23-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Extend damon_operations struct with a new callback, namely apply_probes.
The callback will be invoked for data attributes monitoring. More
specifically, the callback will apply damon_probe objects to each region
and update the per-region per-probe counters for the number of encountered
probe-positive samples.
Link: https://lore.kernel.org/20260518234119.97569-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Add an array for the per-region per-probe positive samples count. For
simple and efficient implementation, add a limit to the number of data
probes and set the array to support only the limited number of counters.
Link: https://lore.kernel.org/20260518234119.97569-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Define a data structure for constructing damon_probe's attributes check,
namely damon_filter. It is very similar to damos_filter but works only
for monitoring purposes. Also embed that into damon_probe, implement
essential handling of the link, with fundamental helpers.
Link: https://lore.kernel.org/20260518234119.97569-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Let damon_probe objects be able to be installed on a given damon_ctx, by
adding a linked list header for storing the objects. Add initialization
and cleanup of the new field with helper functions, too.
Link: https://lore.kernel.org/20260518234119.97569-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: introduce data attributes monitoring".
TL; DR
======
Extend DAMON for monitoring general data attributes other than accesses.
The short term motivation is lightweight page type (e.g., belonging
cgroup) aware monitoring. In long term, this will help extending DAMON
for multiple access events capture primitives (e.g., page faults and PMU)
and eventually pivotting DAMON to a "Data Attributes Monitoring and
Operations eNgine" in long term.
Background: High Cost of Page Level Properties Monitoring
=========================================================
DAMON is initially introduced as a Data Access MONitor. It has been
extended for not only access monitoring but also data access-aware system
operations (DAMOS). But still the monitoring part is only for data
accesses.
Data access patterns is good information, but some users need more
holistic views. Particularly, users want to show the access pattern
information together with the types of the memory. For example, users who
work for making huge pages efficiently want to know how much of
DAMON-found hot/cold regions are backed by huge pages. Users who run
multiple workloads with different cgroups want to know how much of
DAMON-found hot/cold regions belong to specific cgroups.
For the user demand, we developed a DAMOS extension for page level
properties based monitoring [1], which has landed on 6.14. Using the
feature, users can inform the page level data properties that they are
interested in, in a flexible format that uses DAMOS filters. Then, DAMON
applies the filters to each folio of the entire DAMON region and lets
users know how many bytes of memory in each DAMON region passed the given
filters.
This gives page level detailed and deterministic information to users.
But, because the operation is done at page level, the overhead is
proportional to the memory size. It was useful for test or debugging
purposes on a small number of machines. But it was obviously too heavy to
be enabled always on all machines running the real user workloads. For
real world workloads, it was recommended to use the feature with
user-space controlled sampling approaches. For example, users could do
the page level monitoring only once per hour, on randomly selected one
percent of machines of their fleet. If the runtime and the size of the
fleet is long and big enough, it should provide statistically meaningful
data.
But users are too busy to implement such controls on their own.
Data Attributes Monitoring
==========================
Extend DAMON to monitor not only data accesses, but also general data
attributes. Do the extension while keeping the main promise of DAMON, the
bounded and best-effort minimum overhead.
Allow users to specify what data attributes in addition to the data access
they want to monitor. Users can install one 'data probe' per data
attribute of their interest for this purpose. The 'data probe' should be
able to be applied to any memory, and determine if the given memory has
the appropriate data attribute. E.g., if memory of physical address 42
belongs to cgroup A. Each 'data probe' is configured with filters that
are very similar to the DAMOS filters.
When DAMON checks if each sampling address memory of each region is
accessed since the last check, it applies data probes if registered. Same
to the number of access check-positive samples accounting (nr_accesses),
it accounts the number of each data probe-positive samples in another
per-region counters array, namely 'probe_hits'. When DAMON resets
nr_accesses every aggregation interval, it resets 'probe_hits' together.
Users can read 'probe_hits' just before the values are reset. In this
way, users can know how many hot/cold memory regions have data attributes
of their interest. E.g., 30 percent of this system's hot memory is
belonging to cgroup A, and 80 percent of the cgroup A-belonging hot memory
is backed by huge pages.
Patches Sequence
================
First eight patches implement the core feature, interface and the working
support. Patch 1 introduces data probe data structure, namely
damon_probe. Patch 2 extends damon_ctx for installing data probes. Patch
3 introduces another data structure for filters of each data probe, namely
damon_filter. Patch 4 updates damon_ctx commit function to handle the
probes. Patch 5 extends damon_region for the per-region per-probe
positive samples counter, namely probe_hits. Patch 6 extends
damon_operations for applying probes on the underlying DAMON operations
implementation. Patch 7 updates kdamond_fn() to invoke the probes
applying callback. Patch 8 finally implements the probes support on paddr
ops.
Ten changes for user interface (patches 9-18) come next. Patches 9-13
implements sysfs directories and files for setting data probes, namely
probes directory, probe directory, filters directory, filter directory and
filter directory internal files, respectively. Patch 14 connects the user
inputs that are made via the sysfs files to DAMON core. Following three
patches (patches 15-17) implement sysfs directories and files for showing
the probe_hits to users, namely probes directory, probe directory and hits
files, respectively. Patch 18 introduces a new tracepoint for showing the
probe_hits via tracefs.
Patch 19 adds a selftest for the sysfs files.
Patches 20 and 21 documents the design and usage of the new feature,
respectively.
Seven additional patches (patches 22-28) for monitoring belonging memory
cgroup follow. Depending on the feedback, this part might be separated to
another series in future. Patch 22 defines the DAMON filter type for the
new attribute, namely DAMON_FILTER_TYPE_MEMCG. Patch 23 add the support
on paddr ops. Patch 24 updates the sysfs interface for setup of the
target memcg. Patch 25 move code for easy reuse of the filter target
memcg setup. Patch 26 connects the user input to the core layer.
Finally, patches 27 and 28 update the design and usage documents for the
memcg attribute monitoring support.
Discussion
==========
This allows the page properties monitoring with overhead that is low
enough to be enabled always on real world workloads. Because the sampling
time for access check is reused for data attributes check, the
upper-bounded and best-effort minimum overhead of DAMON is kept. Because
the sampling memory for access check is reused for data attributes check,
additional overhead is minimum.
Still DAMOS-based page level properties monitoring should be useful,
because it provides a deterministic page level information. When in doubt
of the sampling based information, running DAMOS-based one together and
comparing the results would be useful, for debugging and tuning.
Future Works: Mid Term
========================
This version of implementation is limiting the maximum number of data
probes to four. I will try to find a way to remove the limit in future.
I personally think it should be enough for common use cases, though, and
therefore not giving high priority at the moment.
Future Works: Long Term
=======================
There are user requests for extending DAMON with detailed access
information, for example, per-CPUs/threads/read/writes monitoring. For
that, I was working [2] on extending DAMON to use page fault events as
another access check primitives, and making the infrastructure flexible
for future use of yet another access check primitive. Actually there is
another ongoing work [3] for extending DAMON with PMU events. The
motivation of the work is reducing the overhead, though.
In my work [2], I was introducing a new interface for access sampling
primitives control. Now I think this data probe interface can be used for
that, too. That is, data access becomes just one type of data attribute.
Also, pg_idle-confirmed access, page fault-confirmed access, and PMU
event-confirmed access will be different types of data attributes.
The regions adjustment mechanism is currently working based on the access
information. That's because DAMON is designed for data access monitoring.
That is, data access information is the primary interest, and therefore
DAMON adjusts regions in a way that can best-present the information.
Once data access becomes just one of data attributes, there is no reason
to think data access that special. There might be some users not
interested in access at all but want to know the location of memory of
specific type. Data probes interface will allow doing that. Further, we
could extend the interface to let users set any data attribute as the
'primary' attribute. Then, DAMON will split and merge regions in a way
that can best-present the 'primary' attributes.
DAMOS will also be extended, to specify targets based on not only the data
access pattern, but all user-registered data attributes. From this stage,
we may be able to call DAMON as a "Data Attributes Monitoring and
Operations eNgine".
This patch (of 28):
Introduce a data structure for data attribute probe. It is just a linked
list header at this step. It will be extended in a way that it can
determine if a given memory has a specific data attribute.
Link: https://lore.kernel.org/20260518234119.97569-1-sj@kernel.org
Link: https://lore.kernel.org/20260518234119.97569-2-sj@kernel.org
Link: https://lore.kernel.org/20250106193401.109161-1-sj@kernel.org [1]
Link: https://lore.kernel.org/20251208062943.68824-1-sj@kernel.org/ [2]
Link: https://lore.kernel.org/20260423004211.7037-1-akinobu.mita@gmail.com [3]
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_rand() on the sampling_addr hot path called get_random_u32_below(),
which takes a local_lock_irqsave() around a per-CPU batched entropy pool
and periodically refills it with ChaCha20. At elevated nr_regions counts
(20k+), the lock_acquire / local_lock pair plus __get_random_u32_below()
dominate kdamond perf profiles.
Replace the helper with a lockless lfsr113 generator (struct rnd_state)
held per damon_ctx and seeded from get_random_u64() in damon_new_ctx().
kdamond is the single consumer of a given ctx, so no synchronization is
required. Range mapping uses traditional reciprocal multiplication,
similar as get_random_u32_below(); for spans larger than U32_MAX (only
reachable on 64-bit) the slow path combines two u32 outputs and uses
mul_u64_u64_shr() at 64-bit width. On 32-bit the slow path is dead code
and gets eliminated by the compiler.
The new helper takes a ctx parameter; damon_split_regions_of() and the
kunit tests that call it directly are updated accordingly.
lfsr113 is a linear PRNG and MUST NOT be used for anything
security-sensitive. DAMON's sampling_addr is not exposed to userspace and
is only consumed as a probe point for PTE accessed-bit sampling, so a
non-cryptographic PRNG is appropriate here.
Tested with paddr monitoring and max_nr_regions=20000: kdamond CPU usage
reduced from ~72% to ~50% of one core.
Link: https://lore.kernel.org/20260505145212.108644-1-jiayuan.chen@linux.dev
Link: https://lore.kernel.org/damon/20260426173346.86238-1-sj@kernel.org/T/#m4f1fd74112728f83a41511e394e8c3fef703039c
Link: https://lore.kernel.org/20260509011816.85145-1-sj@kernel.org
Signed-off-by: Jiayuan Chen <jiayuan.chen@shopee.com>
Signed-off-by: SeongJae Park <sj@kernel.org>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: Shu Anzai <shu17az@gmail.com>
Cc: Quanmin Yan <yanquanmin1@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Now nobody is using damon_set_region_biggest_system_ram_default(). Remove
it.
Link: https://lore.kernel.org/20260429041232.90257-5-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon/reclaim,lru_sort: monitor all system rams by
default".
DAMON_RECLAIM and DAMON_LRU_SORT set the biggest 'System RAM' resource of
the system as the default monitoring target address range. The main
intention behind the design is to minimize the overhead coming from
monitoring of non-System RAM areas.
This could result in an odd setup when there are multiple discrete System
RAMs of considerable sizes. For example, there are System RAMs each
having 500 GiB size. In this case, only the first 500 GiB will be set as
the monitoring region by default. This is particularly common on NUMA
systems. Hence the modules allow users to set the monitoring target
address range using the module parameters if the default setup doesn't
work for them. In other words, the current design trades ease of setup
for lower overhead.
However, because DAMON utilizes the sampling based access check and the
adaptive regions adjustment mechanisms, the overhead from the monitoring
of non-System RAM areas should be negligible in most setups. Meanwhile,
the setup complexity is causing real headaches for users who need to run
those modules on various types of systems. That is, the current tradeoff
is not a good deal.
Set the physical address range that can cover all System RAM areas of the
system as the default monitoring regions for DAMON_RECLAIM and
DAMON_LRU_SORT.
Technically speaking, this is changing documented behavior. However, it
makes no sense to believe there is a real use case that really depends on
the old weird default behavior. If the old default behavior was working
for them in the reasonable way, this change will only add a negligible
amount of monitoring overhead. If it didn't work, the users may already
be using manual monitoring regions setup, and they will not be affected by
this change.
Patches Sequence
================
Patch 1 introduces a new core function that will be used for the new
default monitoring target region setup. Patch 2 and 3 update
DAMON_RECLAIM and DAMON_LRU_SORT to use the new function instead of the
old one, respectively. Patch 4 removes the old core function that was
replaced by the new one, as there is no more user of it. Patch 5 updates
DAMON_STAT to use the new one instead of its in-house nearly-duplicate
self implementation of the functionality. Finally patches 6 and 7 update
the DAMON_RECLAIM and DAMON_LRU_SORT user documentation for the new
behaviors, respectively.
This patch (of 7):
damon_set_region_biggest_system_ram_default() sets the monitoring target
region as the caller requested. If the caller didn't specify the region,
it finds the biggest System RAM of the system and sets it as the target
region. When there are more than one considerable size of System RAM
resources in the system, the default target setup makes no sense.
Introduce a variant, namely damon_set_region_system_rams_default(). It
sets a physical address range that covers all System RAM resources as the
default target region.
Link: https://lore.kernel.org/20260429041232.90257-1-sj@kernel.org
Link: https://lore.kernel.org/20260429041232.90257-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: let DAMON be paused and resumed", v2.
DAMON utilizes a few mechanisms that enhance itself over time. Adaptive
regions adjustment, goal-based DAMOS quota auto-tuning and monitoring
intervals auto-tuning like self-training mechanisms are such examples. It
also adds access frequency stability information (age) to the monitoring
results, which makes it enhanced over time.
Sometimes users have to stop DAMON. In this case, DAMON internal state
that enhanced over the time of the last execution simply goes away.
Restarted DAMON have to train itself and enhance its output from the
scratch. This makes DAMON less useful in such cases. Introducing three
such use cases below.
Investigation of DAMON. It is best to do the investigation online,
especially when it is a production environment. DAMON therefore provides
features for such online investigations, including DAMOS stats, monitoring
result snapshot exposure, and multiple tracepoints. When those are
insufficient, and there are additional clues that could be interfered by
DAMON, users have to temporarily stop DAMON to collect the additional
clues. It is not very useful since many of DAMON internal clues are gone
when DAMON is stopped. The loss of the monitoring results that improved
over time is also problematic, especially in production environments.
Monitoring of workloads that have different user-known phases. For
example, in Android, applications are known to have very different access
patterns and behaviors when they are running on the foreground and the
background. It can therefore be useful to separate monitoring of apps
based on whether they are running on the foreground and on the background.
Having two DAMON threads per application that paused and resumed for the
apps foreground/background switches can be useful for the purpose. But
such pause/resume of the execution is not supported.
Tests of DAMON. A few DAMON selftests are using drgn to dump the internal
DAMON status. The tests show if the dumped status is the same as what the
test code expected. Because DAMON keeps running and modifying its
internal status, there are chances of data races that can cause false test
results. Stopping DAMON can avoid the race. But, since the internal
state of DAMON is dropped, the test coverage will be limited.
Let DAMON execution be paused and resumed without loss of the internal
state, to overhaul the limitations. For this, introduce a new DAMON
context parameter, namely 'pause'. API callers can update it while the
context is running, using the online parameters update functions
(damon_commit_ctx() and damon_call()). Once it is set, kdamond_fn() main
loop will do only limited works excluding the monitoring and DAMOS works,
while sleeping sampling intervals per the work. The limited works include
handling of the online parameters update. Hence users can unset the
'pause' parameter again. Once it is unset, kdamond_fn() main loop will do
all the work again (resumed). Under the paused state, it also does stop
condition checks and handling of it, so that paused DAMON can also be
stopped if needed. Expose the feature to the user space via DAMON sysfs
interface. Also, update existing drgn-based tests to test and use the
feature.
Tests
=====
I confirmed the feature functionality using real time tracing ('perf
trace' or 'trace-cmd stream') of damon:damon_aggregated DAMON tracepoint.
By pausing and resuming the DAMON execution, I was able to see the trace
stops and continued as expected. Note that the pause feature support is
added to DAMON user-space tool (damo) after v3.1.9. Users can use
'--pause_ctx' command line option of damo for that, and I actually used it
for my test. The extended drgn-based selftests are also testing a part of
the functionality.
Patches Sequence
================
Patch 1 introduces the new core API for the pause feature. Patch 2 extend
DAMON sysfs interface for the new parameter. Patches 3-5 update design,
usage and ABI documents for the new sysfs file, respectively. The
following five patches are for tests. Patch 6 implements a new kunit test
for the pause parameter online commitment. Patches 7 and 8 extend DAMON
selftest helpers to support the new feature. Patch 9 extends selftest to
test the commitment of the feature. Finally, patch 10 updates existing
selftest to be safe from the race condition using the pause/resume
feature.
This patch (of 10):
DAMON supports only start and stop of the execution. When it is stopped,
its internal data that it self-trained goes away. It will be useful if
the execution can be paused and resumed with the previous self-trained
data.
Introduce per-context API parameter, 'paused', for the purpose. The
parameter can be set and unset while DAMON is running and paused, using
the online parameters commit helper functions (damon_commit_ctx() and
damon_call()). Once 'paused' is set, the kdamond_fn() main loop does only
limited works with sampling interval sleep during the works. The limited
works include the handling of the online parameters update, so that users
can unset the 'pause' and resume the execution when they want. It also
keep checking DAMON stop conditions and handling of it, so that DAMON can
be stopped while paused if needed.
Link: https://lore.kernel.org/20260427151231.113429-1-sj@kernel.org
Link: https://lore.kernel.org/20260427151231.113429-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
This patch set introces a new action: DAMOS_COLLAPSE.
For DAMOS_HUGEPAGE and DAMOS_NOHUGEPAGE to work, khugepaged should be
working, since it relies on hugepage_madvise to add a new slot. This slot
should be picked up by khugepaged and eventually collapse (or not, if we
are using DAMOS_NOHUGEPAGE) the pages. If THP is not enabled, khugepaged
will not be working, and therefore no collapse will happen.
DAMOS_COLLAPSE eventually calls madvise_collapse, which will collapse the
address range synchronously. In cases where there is a large VMA
(databases, for example), DAMOS_COLLAPSE allows us to collapse only the
hot region, and not the entire VMA.
This new action may be required to support autotuning with hugepage
as a goal[1].
=========
Benchmarks:
=========
MySQL
=====
Tests were performed in an ARM physical server with MariaDB 10.5 and
sysbench. Read only benchmark was perform with gaussian row hitting,
which follows a normal distribution.
T n, D h: THP set to never, DAMON action set to hugepage
T m, D h: THP set to madvise, DAMON action set to hugepage
T n, D c: THP set to never, DAMON action set to collapse
Memory consumption. Lower is better.
+------------------+----------+----------+----------+
| | T n, D h | T m, D h | T n, D c |
+------------------+----------+----------+----------+
| Total memory use | 2.13 | 2.20 | 2.20 |
| Huge pages | 0 | 1.3 | 1.27 |
+------------------+----------+----------+----------+
Performance in TPS (Transactions Per Second). Higher is better.
T n, D h: 18225.58
T m, D h 18252.93
T n, D c: 18270.21
Performance counter
I got the number of L1 D/I TLB accesses and the number a D/I TLB
accesses that triggered a page walk. I divided the second by the
first to get the percentage of page walkes per TLB access. The
lower the better.
+---------------+--------------+--------------+--------------+
| | T n, D h | T m, D h | T n, D c |
+---------------+--------------+--------------+--------------+
| L1 DTLB | 127248242753 | 125431020479 | 125327001821 |
| L1 ITLB | 80332558619 | 79346759071 | 79298139590 |
| DTLB walk | 75011087 | 52800418 | 55895794 |
| ITLB walk | 71577076 | 71505137 | 67262140 |
| DTLB % misses | 0.058948623 | 0.042095183 | 0.044599961 |
| ITLB % misses | 0.089100954 | 0.090117275 | 0.084821839 |
+---------------+--------------+--------------+--------------+
Masim
=====
I used masim with the "demo" configuration, but changing the times
to 100 seconds for the initial phase and 50 seconds for the rest of
the phases.
Memory consumption:
+------------------+----------+----------+----------+
| | T n, D h | T m, D h | T n, D c |
+------------------+----------+----------+----------+
| Total memory use | 2.38 GB | 2.36 GB | 2.37 GB |
| Huge pages | 0 | 190 MB | 188 MB |
+------------------+----------+----------+----------+
Performance:
THP never, DAMOS_HUGEPAGE
initial phase: 40,491 accesses/msec, 100001 msecs run
low phase 0: 39,658 accesses/msec, 50002 msecs run
high phase 0: 41,678 accesses/msec, 50000 msecs run
low phase 1: 39,625 accesses/msec, 50003 msecs run
high phase 1: 41,658 accesses/msec, 50002 msecs run
low phase 2: 39,642 accesses/msec, 50002 msecs run
high phase 2: 41,640 accesses/msec, 50001 msecs run
THP madvise, DAMOS_HUGEPAGE
initial phase: 51,977 accesses/msec, 100000 msecs run
low phase 0: 86,953 accesses/msec, 50000 msecs run
high phase 0: 94,812 accesses/msec, 50000 msecs run
low phase 1: 101,017 accesses/msec, 50000 msecs run
high phase 1: 94,841 accesses/msec, 50000 msecs run
low phase 2: 100,993 accesses/msec, 50000 msecs run
high phase 2: 94,791 accesses/msec, 50001 msecs run
THP never, DAMOS_COLLAPSE
initial phase: 93,678 accesses/msec, 100001 msecs run
low phase 0: 101,475 accesses/msec, 50000 msecs run
high phase 0: 98,589 accesses/msec, 50000 msecs run
low phase 1: 101,531 accesses/msec, 50001 msecs run
high phase 1: 98,506 accesses/msec, 50001 msecs run
low phase 2: 101,458 accesses/msec, 50001 msecs run
high phase 2: 98,555 accesses/msec, 50000 msecs run
Memory consumption dynamic (how quickly collapses occur):
It shows in seconds how many huge pages are allocated.
+----+----------+----------+
| | T m, D h | T n, D c |
+----+----------+----------+
| 5 | 32 | 188 |
| 10 | 48 | 188 |
| 15 | 64 | 188 |
| 20 | 96 | 188 |
| 30 | 112 | 188 |
| 35 | 144 | 188 |
| 40 | 160 | 188 |
| 45 | 190 | 188 |
| 50 | 190 | 188 |
| 55 | 190 | 188 |
| 60 | 190 | 188 |
+----+----------+----------+
=========
- We can see that DAMOS "hugepage" action works only when THP is set
to madvise. "collapse" action works even when THP is set to never.
- Performance for "collapse" action is slightly lower than "hugepage"
action and THP madvise. This is due to the fact that collapases
occur synchronously. With "hugepage" they may occur during page
faults.
- Memory consumption is slighly lower for "collapse" than "hugepage"
with THP madvise. This is due to the khugepage collapses all VMAs,
while "collapse" action only collapses the VMAs in the hot region.
- There is an improvement in TLB utilization when collapse through
"hugepage" or "collapse" actions are triggered. The amount of
TLB misses is lower.
- "collapse" action is performance synchronously, which means that
page collapses happen earlier and more rapidly. This can be
useful or not, depending on the scenario.
- "hugepage" action may trigger a VMA split in some scenarios, since
it needs to change the flag of the VMA to THP enabled. This may
lead to additional overhead.
Collapse action just adds a new option to chose the correct system
balance.
Link: https://lore.kernel.org/20260426231619.107231-5-sj@kernel.org
Link: https://lore.kernel.org/damon/20260313000816.79933-1-sj@kernel.org/ [1]
Signed-off-by: Asier Gutierrez <gutierrez.asier@huawei-partners.com>
Signed-off-by: SeongJae Park <sj@kernel.org>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: Cheng-Han Wu <hank20010209@gmail.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Liew Rui Yan <aethernet65535@gmail.com>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
DAMOS quota is charged to all DAMOS action application attempted memory,
regardless of how much of the memory the action was successful and failed.
This makes understanding quota behavior without DAMOS stat but only with
end level metrics (e.g., increased amount of free memory for DAMOS_PAGEOUT
action) difficult. Also, charging action-failed memory same as
action-successful memory is somewhat unfair, as successful action
application will induce more overhead in most cases.
Introduce DAMON core API for setting the charge ratio for such
action-failed memory. It allows API callers to specify the ratio in a
flexible way, by setting the numerator and the denominator.
Link: https://lore.kernel.org/20260428013402.115171-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Background and Motivation
=========================
In heterogeneous memory systems, controlling memory distribution across
NUMA nodes is essential for performance optimization. This patch enables
system-wide page distribution with target-state goals such as "maintain
60% of scheme-eligible memory on DRAM" using PA-mode DAMON schemes.
Rather than using absolute thresholds, this metric tracks the ratio of
memory that matches each scheme's access pattern filters on a target node,
enabling the quota system to automatically adjust migration aggressiveness
to maintain the desired distribution.
What This Metric Measures
=========================
node_eligible_mem_bp:
scheme_eligible_bytes_on_node / total_scheme_eligible_bytes * 10000
Two-Scheme Setup for Hot Page Distribution
==========================================
For maintaining 60% of hot memory on DRAM (node 0) and 40% on CXL
(node 1):
PULL scheme: migrate_hot to node 0
goal: node_eligible_mem_bp, nid=0, target=6000
addr filter: node 1 address range (only migrate FROM CXL)
"Move hot pages to DRAM if less than 60% of hot data is in DRAM"
PUSH scheme: migrate_hot to node 1
goal: node_eligible_mem_bp, nid=1, target=4000
addr filter: node 0 address range (only migrate FROM DRAM)
"Move hot pages to CXL if less than 40% of hot data is in CXL"
Each scheme independently measures its own eligible memory and adjusts its
quota to achieve its target ratio. The schemes work in concert through
DAMON's unified monitoring context, with the quota autotuner balancing
their relative aggressiveness.
Implementation Details
======================
The implementation adds a new quota goal metric type
DAMOS_QUOTA_NODE_ELIGIBLE_MEM_BP to the existing DAMOS quota goal
framework. When this metric is configured for a scheme:
1. During each quota adjustment cycle, damos_get_node_eligible_mem_bp()
is called to calculate the current memory distribution.
2. The function iterates through all regions that match the scheme's
access pattern (via __damos_valid_target()) and calculates:
- Total eligible bytes across all nodes
- Eligible bytes specifically on the target node (goal->nid)
3. For each eligible region, damos_calc_eligible_bytes() walks through
the physical address range, using damon_get_folio() to look up
each folio and determine its NUMA node via folio_nid().
4. Large folios are handled by calculating the exact overlap between
the region boundaries and folio boundaries, ensuring accurate
byte counts even when regions partially span folios.
5. The ratio (node_eligible / total_eligible * 10000) is returned
as basis points, which the quota autotuner uses to adjust the
scheme's effective quota size (esz).
The implementation requires CONFIG_DAMON_PADDR since damon_get_folio()
is only available for physical address space monitoring.
Testing Results
===============
Functionally tested on a two-node heterogeneous memory system with DRAM
(node 0) and CXL memory (node 1). A PUSH+PULL scheme configuration using
migrate_hot actions was used to reach a target hot memory ratio between
the two tiers.
With the TEMPORAL tuner, the system converges quickly to the target
distribution. The tuner drives esz to maximum when under goal and to zero
once the goal is met, forming a simple on/off feedback loop that
stabilizes at the desired ratio.
With the CONSIST tuner, the scheme still converges but more slowly, as it
migrates and then throttles itself based on quota feedback. The time to
reach the goal varies depending on workload intensity.
Note: This metric works with both TEMPORAL and CONSIST goal tuners.
Link: https://lore.kernel.org/20260428030520.701-1-ravis.opensrc@gmail.com
Signed-off-by: Ravi Jonnalagadda <ravis.opensrc@gmail.com>
Suggested-by: SeongJae Park <sj@kernel.org>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: Honggyu Kim <honggyu.kim@sk.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Yunjeong Mun <yunjeong.mun@sk.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
When kdamond_fn() main loop is finished, the function cancels remaining
damos_walk() request and unset the damon_ctx->kdamond so that API callers
and API functions themselves can show the context is terminated.
damos_walk() adds the caller's request to the queue first. After that, it
shows if the kdamond of the damon_ctx is still running (damon_ctx->kdamond
is set). Only if the kdamond is running, damos_walk() starts waiting for
the kdamond's handling of the newly added request.
The damos_walk() requests registration and damon_ctx->kdamond unset are
protected by different mutexes, though. Hence, damos_walk() could race
with damon_ctx->kdamond unset, and result in deadlocks.
For example, let's suppose kdamond successfully finished the damow_walk()
request cancelling. Right after that, damos_walk() is called for the
context. It registers the new request, and shows the context is still
running, because damon_ctx->kdamond unset is not yet done. Hence the
damos_walk() caller starts waiting for the handling of the request.
However, the kdamond is already on the termination steps, so it never
handles the new request. As a result, the damos_walk() caller thread
infinitely waits.
Fix this by introducing another damon_ctx field, namely
walk_control_obsolete. It is protected by the
damon_ctx->walk_control_lock, which protects damos_walk() request
registration. Initialize (unset) it in kdamond_fn() before letting
damon_start() returns and set it just before the cancelling of the
remaining damos_walk() request is executed. damos_walk() reads the
obsolete field under the lock and avoids adding a new request.
After this change, only requests that are guaranteed to be handled or
cancelled are registered. Hence the after-registration DAMON context
termination check is no longer needed. Remove it together.
The issue is found by sashiko [1].
Link: https://lore.kernel.org/20260327233319.3528-3-sj@kernel.org
Link: https://lore.kernel.org/20260325141956.87144-1-sj@kernel.org [1]
Fixes: bf0eaba0ff9c ("mm/damon/core: implement damos_walk()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> # 6.14.x
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon/core: fix damon_call()/damos_walk() vs kdmond exit
race".
damon_call() and damos_walk() can leak memory and/or deadlock when they
race with kdamond terminations. Fix those.
This patch (of 2);
When kdamond_fn() main loop is finished, the function cancels all
remaining damon_call() requests and unset the damon_ctx->kdamond so that
API callers and API functions themselves can know the context is
terminated. damon_call() adds the caller's request to the queue first.
After that, it shows if the kdamond of the damon_ctx is still running
(damon_ctx->kdamond is set). Only if the kdamond is running, damon_call()
starts waiting for the kdamond's handling of the newly added request.
The damon_call() requests registration and damon_ctx->kdamond unset are
protected by different mutexes, though. Hence, damon_call() could race
with damon_ctx->kdamond unset, and result in deadlocks.
For example, let's suppose kdamond successfully finished the damon_call()
requests cancelling. Right after that, damon_call() is called for the
context. It registers the new request, and shows the context is still
running, because damon_ctx->kdamond unset is not yet done. Hence the
damon_call() caller starts waiting for the handling of the request.
However, the kdamond is already on the termination steps, so it never
handles the new request. As a result, the damon_call() caller threads
infinitely waits.
Fix this by introducing another damon_ctx field, namely
call_controls_obsolete. It is protected by the
damon_ctx->call_controls_lock, which protects damon_call() requests
registration. Initialize (unset) it in kdamond_fn() before letting
damon_start() returns and set it just before the cancelling of remaining
damon_call() requests is executed. damon_call() reads the obsolete field
under the lock and avoids adding a new request.
After this change, only requests that are guaranteed to be handled or
cancelled are registered. Hence the after-registration DAMON context
termination check is no longer needed. Remove it together.
Note that the deadlock will not happen when damon_call() is called for
repeat mode request. In tis case, damon_call() returns instead of waiting
for the handling when the request registration succeeds and it shows the
kdamond is running. However, if the request also has dealloc_on_cancel,
the request memory would be leaked.
The issue is found by sashiko [1].
Link: https://lore.kernel.org/20260327233319.3528-1-sj@kernel.org
Link: https://lore.kernel.org/20260327233319.3528-2-sj@kernel.org
Link: https://lore.kernel.org/20260325141956.87144-1-sj@kernel.org [1]
Fixes: 42b7491af14c ("mm/damon/core: introduce damon_call()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> # 6.14.x
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_set_region_biggest_system_ram_default()
damon_find_biggest_system_ram() was not supporting addr_unit in the past.
Hence, its caller, damon_set_region_biggest_system_ram_default(), was also
not supporting addr_unit. The previous commit has updated the inner
function to support addr_unit. There is no more reason to not support
addr_unit on damon_set_region_biggest_system_ram_default(). Rather, it
makes unnecessary inconsistency on support of addr_unit. Update it to
receive addr_unit and handle it inside.
Link: https://lkml.kernel.org/r/20260311052927.93921-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Yang yingliang <yangyingliang@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Introduce a new goal-based DAMOS quota auto-tuning algorithm, namely
DAMOS_QUOTA_GOAL_TUNER_TEMPORAL (temporal in short). The algorithm aims
to trigger the DAMOS action only for a temporal time, to achieve the goal
as soon as possible. For the temporal period, it uses as much quota as
allowed. Once the goal is achieved, it sets the quota zero, so
effectively makes the scheme be deactivated.
Link: https://lkml.kernel.org/r/20260310010529.91162-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: support multiple goal-based quota tuning
algorithms".
Aim-oriented DAMOS quota auto-tuning uses a single tuning algorithm. The
algorithm is designed to find a quota value that should be consistently
kept for achieving the aimed goal for long term. It is useful and
reliable at automatically operating systems that have dynamic environments
in the long term.
As always, however, no single algorithm fits all. When the environment
has static characteristics or there are control towers in not only the
kernel space but also the user space, the algorithm shows some
limitations. In such environments, users want kernel work in a more short
term deterministic way. Actually there were at least two reports [1,2] of
such cases.
Extend DAMOS quotas goal to support multiple quota tuning algorithms that
users can select. Keep the current algorithm as the default one, to not
break the old users. Also give it a name, "consist", as it is designed to
"consistently" apply the DAMOS action. And introduce a new tuning
algorithm, namely "temporal". It is designed to apply the DAMOS action
only temporally, in a deterministic way. In more detail, as long as the
goal is under-achieved, it uses the maximum quota available. Once the
goal is over-achieved, it sets the quota zero.
Tests
=====
I confirmed the feature is working as expected using the latest version of
DAMON user-space tool, like below.
$ # start DAMOS for reclaiming memory aiming 30% free memory
$ sudo ./damo/damo start --damos_action pageout \
--damos_quota_goal_tuner temporal \
--damos_quota_goal node_mem_free_bp 30% 0 \
--damos_quota_interval 1s \
--damos_quota_space 100M
Note that >=3.1.8 version of DAMON user-space tool supports this feature
(--damos_quota_goal_tuner). As expected, DAMOS stops reclaiming memory as
soon as the goal amount of free memory is made. When 'consist' tuner is
used, the reclamation was continued even after the goal amount of free
memory is made, resulting in more than goal amount of free memory, as
expected.
Patch Sequence
==============
First four patches implement the features. Patch 1 extends core API to
allow multiple tuners and make the current tuner as the default and only
available tuner, namely 'consist'. Patch 2 allows future tuners setting
zero effective quota. Patch 3 introduces the second tuner, namely
'temporal'. Patch 4 further extends DAMON sysfs API to let users use
that.
Three following patches (patches 5-7) update design, usage, and ABI
documents, respectively.
Final four patches (patches 8-11) are for adding tests. The eighth patch
(patch 8) extends the kunit test for online parameters commit for
validating the goal_tuner. The ninth and the tenth patches (patches 9-10)
extend the testing-purpose DAMON sysfs control helper and DAMON status
dumping tool to support the newly added feature. The final eleventh one
(patch 11) extends the existing online commit selftest to cover the new
feature.
This patch (of 11):
DAMOS quota goal feature utilizes a single feedback loop based algorithm
for automatic tuning of the effective quota. It is useful in dynamic
environments that operate systems with only kernels in the long term.
But, no one fits all. It is not very easy to control in environments
having more controlled characteristics and user-space control towers. We
actually got multiple reports [1,2] of use cases that the algorithm is not
optimal.
Introduce a new field of 'struct damos_quotas', namely 'goal_tuner'. It
specifies what tuning algorithm the given scheme should use, and allows
DAMON API callers to set it as they want. Nonetheless, this commit
introduces no new tuning algorithm but only the interface. This commit
hence makes no behavioral change. A new algorithm will be added by the
following commit.
Link: https://lkml.kernel.org/r/20260310010529.91162-2-sj@kernel.org
Link: https://lore.kernel.org/CALa+Y17__d=ZsM1yX+MXx0ozVdsXnFqF4p0g+kATEitrWyZFfg@mail.gmail.com [1]
Link: https://lore.kernel.org/20260204022537.814-1-yunjeong.mun@sk.com [2]
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes (Oracle) <ljs@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
DAMON regions are assumed to always be non-zero length. There was a
confusion [1] about it, probably due to lack of the documentation.
Document it.
Link: https://lkml.kernel.org/r/20260307195356.203753-5-sj@kernel.org
Link: https://lore.kernel.org/20251231070029.79682-1-sj@kernel.org/ [1]
Signed-off-by: SeongJae Park <sj@kernel.org>
Acked-by: wang lian <lianux.mm@gmail.com>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes (Oracle) <ljs@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_target is not used by get_scheme_score operations, nor with virtual
neither with physical addresses.
Link: https://lkml.kernel.org/r/20260213145032.1740407-1-gutierrez.asier@huawei-partners.com
Signed-off-by: Asier Gutierrez <gutierrez.asier@huawei-partners.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Quanmin Yan <yanquanmin1@huawei.com>
Cc: ze zuo <zuoze1@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
One major usage of damon_call() is online DAMON parameters update. It is
done by calling damon_commit_ctx() inside the damon_call() callback
function. damon_commit_ctx() can fail for two reasons: 1) invalid
parameters and 2) internal memory allocation failures. In case of
failures, the damon_ctx that attempted to be updated (commit destination)
can be partially updated (or, corrupted from a perspective), and therefore
shouldn't be used anymore. The function only ensures the damon_ctx object
can safely deallocated using damon_destroy_ctx().
The API callers are, however, calling damon_commit_ctx() only after
asserting the parameters are valid, to avoid damon_commit_ctx() fails due
to invalid input parameters. But it can still theoretically fail if the
internal memory allocation fails. In the case, DAMON may run with the
partially updated damon_ctx. This can result in unexpected behaviors
including even NULL pointer dereference in case of damos_commit_dests()
failure [1]. Such allocation failure is arguably too small to fail, so
the real world impact would be rare. But, given the bad consequence, this
needs to be fixed.
Avoid such partially-committed (maybe-corrupted) damon_ctx use by saving
the damon_commit_ctx() failure on the damon_ctx object. For this,
introduce damon_ctx->maybe_corrupted field. damon_commit_ctx() sets it
when it is failed. kdamond_call() checks if the field is set after each
damon_call_control->fn() is executed. If it is set, ignore remaining
callback requests and return. All kdamond_call() callers including
kdamond_fn() also check the maybe_corrupted field right after
kdamond_call() invocations. If the field is set, break the kdamond_fn()
main loop so that DAMON sill doesn't use the context that might be
corrupted.
[sj@kernel.org: let kdamond_call() with cancel regardless of maybe_corrupted]
Link: https://lkml.kernel.org/r/20260320031553.2479-1-sj@kernel.org
Link: https://sashiko.dev/#/patchset/20260319145218.86197-1-sj%40kernel.org
Link: https://lkml.kernel.org/r/20260319145218.86197-1-sj@kernel.org
Link: https://lore.kernel.org/20260319043309.97966-1-sj@kernel.org [1]
Fixes: 3301f1861d34 ("mm/damon/sysfs: handle commit command using damon_call()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> [6.15+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
'min_sz_region' field of 'struct damon_ctx' represents the minimum size of
each DAMON region for the context. 'struct damos_access_pattern' has a
field of the same name. It confuses readers and makes 'grep' less optimal
for them. Rename it to 'min_region_sz'.
Link: https://lkml.kernel.org/r/20260117175256.82826-9-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The macro is for the default minimum size of each DAMON region. There was
a case that a reader was confused if it is the minimum number of total
DAMON regions, which is set on damon_attrs->min_nr_regions. Make the name
more explicit.
Link: https://lkml.kernel.org/r/20260117175256.82826-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_call_control->dealloc_on_cancel works only when ->repeat is true.
But the behavior is not clearly documented. DAMON API callers can
understand the behavior only after reading kdamond_call() code. Document
the behavior on the kernel-doc comment of damon_call_control.
Link: https://lkml.kernel.org/r/20260117175256.82826-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: cleanup kdamond, damon_call(), damos filter and
DAMON_MIN_REGION".
Do miscellaneous code cleanups for improving readability. First three
patches cleanup kdamond termination process, by removing unused operation
set cleanup callback (patch 1) and moving damon_ctx specific resource
cleanups on kdamond termination to synchronization-easy place (patches 2
and 3). Next two patches touch damon_call() infrastructure, by
refactoring kdamond_call() function to do less and simpler locking
operations (patch 4), and documenting when dealloc_on_free does work
(patch 5). Final three patches rename things for clear uses of those.
Those rename damos_filter_out() to be more explicit about the fact that it
is only for core-handled filters (patch 6), DAMON_MIN_REGION macro to be
more explicit it is not about number of regions but size of each region
(patch 7), and damon_ctx->min_sz_region to be different from
damos_access_patern->min_sz_region (patch 8), so that those are not
confusing and easy to grep.
This patch (of 8):
damon_operations->cleanup() was added for a case that an operation set
implementation requires additional cleanups. But no such implementation
exists at the moment. Remove it.
Link: https://lkml.kernel.org/r/20260117175256.82826-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20260117175256.82826-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
There is no DAMON API caller that directly access 'kdamond' and
'kdamond_lock' fields of 'struct damon_ctx'. Keeping those exposed could
only encourage creative but error-prone usages. Hide them from DAMON API
callers by marking those as private fields.
Link: https://lkml.kernel.org/r/20260115152047.68415-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: hide kdamond and kdamond_lock from API callers".
'kdamond' and 'kdamond_lock' fields initially exposed to DAMON API callers
for flexible synchronization and use cases. As DAMON API became somewhat
complicated compared to the early days, Keeping those exposed could only
encourage the API callers to invent more creative but complicated and
difficult-to-debug use cases.
Fortunately DAMON API callers didn't invent that many creative use cases.
There exist only two use cases of 'kdamond' and 'kdamond_lock'. Finding
whether the kdamond is actively running, and getting the pid of the
kdamond. For the first use case, a dedicated API function, namely
'damon_is_running()' is provided, and all DAMON API callers are using the
function for the use case. Hence only the second use case is where the
fields are directly being used by DAMON API callers.
To prevent future invention of complicated and erroneous use cases of the
fields, hide the fields from the API callers. For that, provide new
dedicated DAMON API functions for the remaining use case, namely
damon_kdamond_pid(), migrate DAMON API callers to use the new function,
and mark the fields as private fields.
This patch (of 5):
'kdamond' and 'kdamond_lock' are directly being used by DAMON API callers
for getting the pid of the corresponding kdamond. To discourage invention
of creative but complicated and erroneous new usages of the fields that
require careful synchronization, implement a new API function that can
simply be used without the manual synchronizations.
Link: https://lkml.kernel.org/r/20260115152047.68415-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20260115152047.68415-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: advance DAMOS-based LRU sorting".
DAMOS_LRU_[DE]PRIO actions were added to DAMOS for more access-aware LRU
lists sorting. For simple usage, a specialized kernel module, namely
DAMON_LRU_SORT, has also been introduced. After the introduction of the
module, DAMON got a few important new features, including the aim-based
quota auto-tuning, age tracking, young page filter, and monitoring
intervals auto-tuning. Meanwhile, DAMOS-based LRU sorting had no direct
updates. Now we show some rooms to advance for DAMOS-based LRU sorting.
Firstly, the aim-oriented quota auto-tuning can simplify the LRU sorting
parameters tuning. But there is no good auto-tuning target metric for LRU
sorting use case. Secondly, the behavior of DAMOS_LRU_[DE]PRIO are not
very symmetric. DAMOS_LRU_DEPRIO directly moves the pages to inactive LRU
list, while DAMOS_LRU_PRIO only marks the page as accessed, so that the
page can not directly but only eventually moved to the active LRU list.
Finally, DAMON_LRU_SORT users cannot utilize the modern features that can
be useful for them, too.
Improve the situation with the following changes. First, introduce a new
DAMOS quota auto-tuning target metric for active:inactive memory size
ratio. Since LRU sorting is a kind of balancing of active and inactive
pages, the active:inactive memory size ratio can be intuitively set.
Second, update DAMOS_LRU_[DE]PRIO behaviors to be more intuitive and
symmetric, by letting them directly move the pages to [in]active LRU list.
Third, update the DAMON_LRU_SORT module user interface to be able to
fully utilize the modern features including the [in]active memory size
ratio-based quota auto-tuning, young page filter, and monitoring intervals
auto-tuning.
With these changes, for example, users can now ask DAMON to "find hot/cold
memory regions with auto-tuned monitoring intervals, do one more page
level access check for found hot/cold memory, and move pages of those to
active or inactive LRU lists accordingly, aiming X:Y active to inactive
memory ratio." For example, if they know 30% of the memory is better to be
protected from reclamation, 30:70 can be set as the target ratio.
Test Results
------------
I ran DAMON_LRU_SORT with the features introduced by this series, on a
real world server workload. For the active:inactive ratio goal, I set
50:50. I confirmed it achieves the target active:inactive ratio, without
manual tuning of the monitoring intervals and the hot/coldness thresholds.
The baseline system that was not running the DAMON_LRU_SORT was keeping
active:inactive ratio of about 1:10.
Note that the test didn't show a clear performance difference, though. I
believe that was mainly because the workload was not very memory
intensive. Also, whether the 50:50 target ratio was optimum is unclear.
Nonetheless, the positive performance impact of the basic LRU sorting idea
is already confirmed with the initial DAMON_LRU_SORT introduction patch
series. The goal of this patch series is simplifying the parameters
tuning of DAMOS-based LRU sorting, and the test confirmed the aimed goals
are achieved.
Patches Sequence
----------------
First three patches extend DAMOS quota auto-tuning to support [in]active
memory ratio target metric type. Those (patches 1-3) introduce new
metrics, implement DAMON sysfs support, and update the documentation,
respectively.
Following patch (patch 4) makes DAMOS_LRU_PRIO action to directly move
target pages to active LRU list, instead of only marking them accessed.
Following seven patches (patches 5-11) updates DAMON_LRU_SORT to support
modern DAMON features. Patch 5 makes it uses not only access frequency
but also age at under-quota regions prioritization. Patches 6-11 add the
support for young page filtering, active:inactive memory ratio based quota
auto-tuning, and monitoring intervals auto-tuning, with appropriate
document updates.
This patch (of 11):
DAMOS_LRU_[DE]PRIO are DAMOS actions for making balance of active and
inactive memory size. There is no appropriate DAMOS quota auto-tuning
target metric for the use case. Add two new DAMOS quota goal metrics for
the purpose, namely DAMOS_QUOTA_[IN]ACTIVE_MEM_BP. Those will represent
the ratio of [in]active memory to total (inactive + active) memory.
Hence, users will be able to ask DAMON to, for example, "find hot and cold
memory, and move pages of those to active and inactive LRU lists,
adjusting the hot/cold thresholds aiming 50:50 active:inactive memory
ratio."
Link: https://lkml.kernel.org/r/20260113152717.70459-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20260113152717.70459-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
DAMON was using the internal private memcg ID which is meant for tracking
kernel objects that outlive their cgroup. Switch to using the public
cgroup ID instead.
Link: https://lkml.kernel.org/r/20251225232116.294540-6-shakeel.butt@linux.dev
Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
Reviewed-by: SeongJae Park <sj@kernel.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Wei Xu <weixugc@google.com>
Cc: Yuanchu Xie <yuanchu@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
There are DAMOS use cases that require user-space centric control of its
activation and deactivation. Having the control plane on the user-space,
or using DAMOS as a way for monitoring results collection are such
examples.
DAMON parameters online commit, DAMOS quotas and watermarks can be useful
for this purpose. However, those features work only at the
sub-DAMON-snapshot level. In some use cases, the DAMON-snapshot level
control is required. For example, in DAMOS-based monitoring results
collection use case, the user online-installs a DAMOS scheme with
DAMOS_STAT action, wait it be applied to whole regions of a single
DAMON-snapshot, retrieves the stats and tried regions information, and
online-uninstall the scheme. It is efficient to ensure the lifetime of
the scheme as no more no less one snapshot consumption.
To support such use cases, introduce a new DAMOS core API per-scheme
parameter, namely max_nr_snapshots. As the name implies, it is the upper
limit of nr_snapshots, which is a DAMOS stat that represents the number of
DAMON-snapshots that the scheme has fully applied. If the limit is set
with a non-zero value and nr_snapshots reaches or exceeds the limit, the
scheme is deactivated.
Link: https://lkml.kernel.org/r/20251216080128.42991-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Commit 0e92c2ee9f45 ("mm/damon/schemes: account scheme actions that
successfully applied") has replaced ->stat_count and ->stat_sz of 'struct
damos' with ->stat. The commit mistakenly did not update the related
kernel doc comment, though. Update the comment.
Link: https://lkml.kernel.org/r/20251216080128.42991-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: introduce {,max_}nr_snapshots and tracepoint for
damos stats".
Introduce three changes for improving DAMOS stat's provided information,
deterministic control, and reading usability.
DAMOS provides stats that are important for understanding its behavior.
It lacks information about how many DAMON-generated monitoring output
snapshots it has worked on. Add a new stat, nr_snapshots, to show the
information.
Users can control DAMOS schemes in multiple ways. Using the online
parameters commit feature, they can install and uninstall DAMOS schemes
whenever they want while keeping DAMON runs. DAMOS quotas and watermarks
can be used for manually or automatically turning on/off or adjusting the
aggressiveness of the scheme. DAMOS filters can be used for applying the
scheme to specific memory entities based on their types and locations.
Some users want their DAMOS scheme to be applied to only specific number
of DAMON snapshots, for more deterministic control. One example use case
is tracepoint based snapshot reading. Add a new knob, max_nr_snapshots,
to support this. If the nr_snapshots parameter becomes same to or greater
than the value of this parameter, the scheme is deactivated.
Users can read DAMOS stats via DAMON's sysfs interface. For deep level
investigations on environments having advanced tools like perf and
bpftrace, exposing the stats via a tracepoint can be useful. Implement a
new tracepoint, namely damon:damos_stat_after_apply_interval.
First five patches (patches 1-5) of this series implement the new stat,
nr_snapshots, on the core layer (patch 1), expose on DAMON sysfs user
interface (patch 2), and update documents (patches 3-5).
Following six patches (patches 6-11) are for the new stat based DAMOS
deactivation (max_nr_snapshots). The first one (patch 6) of this group
updates a kernel-doc comment before making further changes. Then an
implementation of it on the core layer (patch 7), an introduction of a new
DAMON sysfs interface file for users of the feature (patch 8), and three
updates of the documents (patches 9-11) follow.
The final one (patch 12) introduces the new tracepoint that exposes the
DAMOS stat values for each scheme apply interval.
This patch (of 12):
DAMON generates monitoring results snapshots for every sampling interval.
DAMOS applies given schemes on the regions of the snapshots, for every
apply interval of the scheme.
DAMOS stat informs a given scheme has tried to how many memory entities
and applied, in the region and byte level. In some use cases including
user-space oriented tuning and investigations, it is useful to know that
in the DAMON-snapshot level. Introduce a new stat, namely nr_snapshots
for DAMON core API callers.
[sj@kernel.org: fix wrong list_is_last() call in damons_is_last_region()]
Link: https://lkml.kernel.org/r/20260114152049.99727-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20251216080128.42991-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20251216080128.42991-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
DAMOS filters that are handled by the ops layer are linked to
damos->ops_filters. Owing to the ops_ prefix on the name, it is easy to
understand it is for ops layer handled filters. The other types of
filters, which are handled by the core layer, are linked to
damos->filters. Because of the name, it is easy to confuse the list is
there for not only core layer handled ones but all filters. Avoid such
confusions by renaming the field to core_filters.
Link: https://lkml.kernel.org/r/20251112154114.66053-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Bill Wendling <morbo@google.com>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Justin Stitt <justinstitt@google.com>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Miguel Ojeda <ojeda@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: misc cleanups".
Yet another batch of misc cleanups and refactoring for DAMON code, tests,
and documents.
First two patches (1and 2) rename DAMOS core filters related code for
readability.
Three following patches (3-5) refactor page table walk callback functions
in DAMON, as suggested by Hugh and David, and I promised.
Next two patches (6 and 7) refactor DAMON core layer kunit test and sysfs
interface selftest to be simple and deduplicated.
Final two patches (8 and 9) fix up sphinx and grammatical errors on
documents.
This patch (of 9):
DAMOS filters handled by the core layer are called core filters, while
those handled by the ops layer are called ops filters. They share the
same type but are managed in different places since core filters are
evaluated before the ops filters. They also have different helper
functions that depend on their managed places.
The helper functions for ops filters have '_ops_' keyword on their name,
so it is easy to know they are for ops filters. Meanwhile, the helper
functions for core filters are not having the 'core' keyword on their
name. This makes it easy to be mistakenly used for ops filters. Actually
there was such a bug.
To avoid future mistakes from similar confusions, rename DAMOS core
filters helper functions to have a keyword 'core' on their names.
Link: https://lkml.kernel.org/r/20251112154114.66053-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20251112154114.66053-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Bill Wendling <morbo@google.com>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Justin Stitt <justinstitt@google.com>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Miguel Ojeda <ojeda@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: David Hildenbrand <david@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: support pin-point targets removal".
DAMON maintains the targets in a list, and allows committing only an
entire list of targets having the new parameters. Targets having same
index on the lists are treated as matching source and destination
targets. If an existing target cannot find a matching one in the
sources list, the target is removed. This means that there is no way to
remove only a specific monitoring target in the middle of the current
targets list.
Such pin-point target removal is really needed in some use cases,
though. Monitoring access patterns on virtual address spaces of
processes that spawned from the same ancestor is one example. If a
process of the group is terminated, the user may want to remove the
matching DAMON target as soon as possible, to save in-kernel memory
usage for the unnecessary target data. The user may also want to do
that without turning DAMON off or removing unnecessary targets, to keep
the current monitoring results for other active processes.
Extend DAMON kernel API and sysfs ABI to support the pin-point removal
in the following way. For API, add a new damon_target field, namely
'obsolete'. If the field on parameters commit source target is set, it
means the matching destination target is obsolete. Then the parameters
commit logic removes the destination target from the existing targets
list. For sysfs ABI, add a new file under the target directory, namely
'obsolete_target'. It is connected with the 'obsolete' field of the
commit source targets, so internally using the new API.
Also add a selftest for the new feature. The related helper scripts for
manipulating the sysfs interface and dumping in-kernel DAMON status are
also extended for this. Note that the selftest part was initially
posted as an individual RFC series [1], but now merged into this one.
Bijan Tabatabai has originally reported this issue, and participated in
this solution design on a GitHub issue [1] for DAMON user-space tool.
This patch (of 9):
DAMON's monitoring targets parameters update function,
damon_commit_targets(), is not providing a way to remove a target in the
middle of the existing targets list. Extend the API by adding a field to
struct damon_target. If the field of a damon_commit_targets() source
target is set, it indicates the matching target on the existing targets
list is obsolete. damon_commit_targets() understands that and removes
those from the list, while respecting the index based matching for other
non-obsolete targets.
Link: https://lkml.kernel.org/r/20251023012535.69625-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20251023012535.69625-2-sj@kernel.org
Link: https://github.com/damonitor/damo/issues/36 [1]
Signed-off-by: SeongJae Park <sj@kernel.org>
Reviewed-by: Bijan Tabatabai <bijan311@gmail.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
damon_set_region_biggest_system_ram_default()
Patch series "mm/damon: fixes for address alignment issues in
DAMON_LRU_SORT and DAMON_RECLAIM", v2.
In DAMON_LRU_SORT and DAMON_RECLAIM, damon_set_regions() will apply
DAMON_MIN_REGION as the core address alignment, and the monitoring target
address ranges would be aligned on DAMON_MIN_REGION * addr_unit. When
users 1) set addr_unit to a value larger than 1, and 2) set the monitoring
target address range as not aligned on DAMON_MIN_REGION * addr_unit, it
will cause DAMON_LRU_SORT and DAMON_RECLAIM to operate on unexpectedly
large physical address ranges.
For example, if the user sets the monitoring target address range to [4,
8) and addr_unit as 1024, the aimed monitoring target address range is [4
KiB, 8 KiB). Assuming DAMON_MIN_REGION is 4096, so resulting target
address range will be [0, 4096) in the DAMON core layer address system,
and [0, 4 MiB) in the physical address space, which is an unexpected
range.
To fix the issue, add a min_sz_region parameter to
damon_set_region_biggest_system_ram_default() and use it when calling
damon_set_regions(), replacing the direct use of DAMON_MIN_REGION.
This patch (of 2):
In DAMON_LRU_SORT, damon_set_regions() will apply DAMON_MIN_REGION as the
core address alignment, and the monitoring target address ranges would be
aligned on DAMON_MIN_REGION * addr_unit. When users 1) set addr_unit to a
value larger than 1, and 2) set the monitoring target address range as not
aligned on DAMON_MIN_REGION * addr_unit, it will cause DAMON_LRU_SORT to
operate on unexpectedly large physical address ranges.
For example, if the user sets the monitoring target address range to [4,
8) and addr_unit as 1024, the aimed monitoring target address range is [4
KiB, 8 KiB). Assuming DAMON_MIN_REGION is 4096, so resulting target
address range will be [0, 4096) in the DAMON core layer address system,
and [0, 4 MiB) in the physical address space, which is an unexpected
range.
To fix the issue, add a min_sz_region parameter to
damon_set_region_biggest_system_ram_default() and use it when calling
damon_set_regions(), replacing the direct use of DAMON_MIN_REGION.
Link: https://lkml.kernel.org/r/20251020130125.2875164-1-yanquanmin1@huawei.com
Link: https://lkml.kernel.org/r/20251020130125.2875164-2-yanquanmin1@huawei.com
Fixes: 2e0fe9245d6b ("mm/damon/lru_sort: support addr_unit for DAMON_LRU_SORT")
Signed-off-by: Quanmin Yan <yanquanmin1@huawei.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: ze zuo <zuoze1@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Add a variant of DAMOS_QUOTA_NODE_MEMCG_USED_BP, for the free memory
portion. The value of the metric is implemented as the entire memory of
the given NUMA node subtracted by the given cgroup's usage. So from a
perspective, "unused" could be a better term than "free". But arguably it
is not very clear what is better, so use the term "free".
Link: https://lkml.kernel.org/r/20251017212706.183502-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Define a new DAMOS quota auto-tuning target metric for per-cgroup per-node
memory usage. For specifying the cgroup of the interest, add a field,
namely memcg_id, to damos_quota_goal struct.
Note that this commit is only implementing the interface. The handling of
the interface (the metric value calculation) will be implemented in the
following commit.
Link: https://lkml.kernel.org/r/20251017212706.183502-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: allow DAMOS auto-tuned for per-memcg per-node
memory usage".
Introduce two new DAMOS quota auto-tuning target metrics for per-cgroup
per-NUMA node memory utilization. Expected use cases are cgroup level
access-aware NUMA memory managements, such as memory tiering or proactive
reclamation on cgroup-based multi-tenant NUMA systems.
Background
==========
The aim-oriented aggressiveness auto-tuning feature of DAMOS is a highly
recommended way for modern DAMOS use cases. Using it, users can specify
what system status they want to achieve with what access-aware system
operations. For example, reclaim cold memory aiming for 0.5 percent of
memory pressure (proactive reclaim), or migrate hot and cold memory
between NUMA nodes having different speed (memory tiering). Then DAMOS
automatically adjusts the aggressiveness of the system operation (e.g.,
increase/decrease reclaim target coldness threshold) based on current
status of the system.
The use case is limited by the supported system status metrics for
specifying the target system status. Two new system metrics for per-node
memory usage ratio, namely DAMOS_QUOTA_NODE_MEM_{USED,FREE}_BP, were
recently added to extend the use cases for access-aware NUMA nodes
management, such as memory tiering. Those are expected to be useful for
not only memory tiering but also general access-aware inter-NUMA node page
migration, though.
Limitation
----------
The per-node memory usage based auto-tuning can be applied only
system-wide. For cgroups-based multi-tenant systems, it could arguably
harm the fairness. For example, a cgroup may use faster NUMA node memory
more than other cgroup, depending on their access pattern. If the user of
each cgroup are promised to get the same quality and amount of the system
resource, this can arguably be an unfair situation.
DAMOS supports cgroup level system operations via DAMOS filter. But the
quota auto-tuning system is not aware of cgroups.
New DAMOS Quota Tuning Metrics for Per-Cgroup Per-NUMA Memory Usage
===================================================================
To overcome the limitation, introduce two new DAMOS quota auto-tuning goal
metrics, namely DAMOS_QUOTA_NODE_MEMCG_{USED,FREE}_BP. Those can be
thought of as a variant of DAMOS_QUOTA_NODE_MEM_{USED,FREE}_BP that
extended for cgroups.
The two metrics specifies per-cgroup, per-node amount of used and unused
memory in ratio to the total memory of the node. For example, let's
assume a system has two NUMA nodes of size 100 GiB and 50 GiB. And two
cgroups are using 40 GiB and 60 GiB of node 0, 20 GiB and 10 GiB of node
1, respectively, as illustrated by the below table.
node-0 node-1
Total memory 100 GiB 50 GiB
Cgroup A usage 40 GiB 20 GiB
Cgroup B usage 60 GiB 10 GiB
Then, DAMOS_QUOTA_NODE_MEMCG_USED_BP for the cgroups for the first node
are, 40 GiB / 100 GiB = 4,000 bp (40 percent) and 60 GiB / 100 GiB = 6,000
bp (60 percent), respectively. Those for the second node are, 20 GiB / 50
GiB = 4000 bp (40 percent) and 10 GiB / 50 GiB = 2000 bp (20 percent),
respectively.
DAMOS_QUOTA_NODE_MEMCG_FREE_BP for the four cases are, 60 GiB /100 GiB =
6000 bp, 40 GiB / 100 GiB = 4000 bp, 30 GiB / 50 GiB = 6000 bp, and 40 GiB
/ 50 GiB = 8000 bp, respectively.
DAMOS_QUOTA_NODE_MEMCG_USED_BP for cgroup A node-0: 4000 bp
DAMOS_QUOTA_NODE_MEMCG_USED_BP for cgroup B node-0: 6000 bp
DAMOS_QUOTA_NODE_MEMCG_USED_BP for cgroup A node-1: 4000 bp
DAMOS_QUOTA_NODE_MEMCG_USED_BP for cgroup B node-1: 2000 bp
DAMOS_QUOTA_NODE_MEMCG_FREE_BP for cgroup A node-0: 6000 bp
DAMOS_QUOTA_NODE_MEMCG_FREE_BP for cgroup B node-0: 4000 bp
DAMOS_QUOTA_NODE_MEMCG_FREE_BP for cgroup A node-1: 6000 bp
DAMOS_QUOTA_NODE_MEMCG_FREE_BP for cgroup B node-1: 8000 bp
Using these, users can specify how much [un]used amount of memory for
per-cgroup and per-node DAMOS should make as a result of the auto-tuning.
Example Usecase: Cgroup Level Memory Tiering
============================================
Let's suppose a typical and simple tiered memory system. The system
equips two NUMA nodes. The first node (node 0) is CPU-attached and fast.
The second node (node 1) is CPU-unattached and slow. It runs two cgroups
that desire to use about 30 percent and 70 percent of the faster node as
much as possible for their hot data, respectively. Then, the user can
implement DAMOS-based memory tiering for the system using the DAMON
user-space tool (damo), like below.
# ./damo start \
`# kdamond for node 1 (slow)` \
--numa_node 1 --monitoring_intervals_goal 4% 3 5ms 10s \
`# promotion scheme for cgroup a` \
--damos_action migrate_hot 0 --damos_access_rate 5% max \
--damos_apply_interval 1s \
--damos_filter allow memcg /workloads/a \
--damos_filter allow young \
--damos_quota_interval 1s --damos_quota_space 200MB \
--damos_quota_goal node_memcg_used_bp 29.7% 0 /workloads/a \
\
`# promotion scheme for cgroup b` \
--damos_action migrate_hot 0 --damos_access_rate 5% max \
--damos_apply_interval 1s \
--damos_filter allow memcg /workloads/b \
--damos_filter allow young \
--damos_quota_interval 1s --damos_quota_space 200MB \
--damos_quota_goal node_memcg_used_bp 69.7% 0 workloads/b \
\
`# kdamond for node 0 (fast)` \
--numa_node 0 --monitoring_intervals_goal 4% 3 5ms 10s \
`# demotion scheme for cgroup a` \
--damos_action migrate_cold 1 --damos_access_rate 0% 0% \
--damos_apply_interval 1s \
--damos_filter allow memcg /workloads/a \
--damos_filter reject young \
--damos_quota_interval 1s --damos_quota_space 200MB \
--damos_quota_goal node_memcg_free_bp 70.5% 0 \
\
`# demotion scheme for cgroup b` \
--damos_action migrate_cold 1 --damos_access_rate 0% 0% \
--damos_apply_interval 1s \
--damos_filter allow memcg /workloads/a \
--damos_filter reject young \
--damos_quota_interval 1s --damos_quota_space 200MB \
--damos_quota_goal node_memcg_free_bp 30.5% 0 \
\
--damos_nr_quota_goals 1 1 1 1 --damos_nr_filters 1 1 1 1 \
--nr_targets 1 1 --nr_schemes 2 2 --nr_ctxs 1 1
With the command, the user-space tool will ask DAMON to spawn two kernel
threads, each for monitoring accesses to node 1 (slow) and node 0 (fast),
respectively. It installs two DAMOS schemes on each thread. Let's call
them "promotion scheme for cgroup a/b", and "demotion scheme for cgroup
a/b" in the order. The promotion schemes are installed on the DAMON
thread for node 1 (slow), and demotion schemes are installed on the DAMON
thread for node 0 (fast).
Cgroup Level Hot Pages Migration (Promotion)
--------------------------------------------
Promotion schemes will find memory regions on node 1 (slow), that some
access was detected. The schemes will then migrate the found memory to
node 0 (fast), hottest pages first.
For accurate and effective migration, these schemes use two page level
filters. First, the migration will be filtered for only cgroup A and
cgroup B. That is, "promotion scheme for cgroup B" will not do the
migration if the page is for cgroup A. Secondly, the schemes will ignore
pages that having their page table's Accessed bits unset. The per-page
Accessed bit check logic will also unset the bit if it was set, for the
next check.
For controlled amounts of system resource consumption and aiming on the
target memory usage, the schemes use quotas setup. The migration is
limited to be done only up to 200 MiB per second, to limit the peak system
resource usage. And DAMOS_QUOTA_NODE_MEMCG_USED_BP target is set for
29.7% and 69.7% of node 0 (fast), respectively. The target value is lower
than the high level goal (30% and 70% system memory), to give headroom on
node 0 (fast). DAMOS will adjust the speed of the pages migration based
on the target and current per-cgroup node 0 memory usage. For example, if
cgroup A is utilizing only 10% of node 0, DAMOS will try to migrate more
of cgroup A hot pages from node 1 to node 0, up to 200 MiB per second. If
cgroup A utilizes more than 29.7% of node 0 memory, the cgroup A hot pages
migration from node 1 to node 0 will be slowed and eventually stopped.
Cgroup Level Cold Pages Migration (Demotion)
--------------------------------------------
Demotion schemes are similar to promotion schemes, but differ in filtering
setup and quota tuning setup. Those filter out pages having their page
table Accessed bits set. And set 70.5% and 30.5% of node 0 memory free
rate for the cgroup A and B, respectively. Hence, if promotion schemes or
something made cgroup A and/or B uses more than 29.5% and 69.5% of node 0,
demotion schemes will start migrating cold pages of appropriate cgroups in
node 0 to node 1, under the 200 MiB per second speed cap, while adjusting
the speed based on how much more than wanted memory is being used.
The quota target values are set to overlap with promotion targets, to keep
a minimum level of page exchanges between the nodes. This is to avoid a
case that the target memory utilization is met, and then access pattern
changes (pages in node 1 become hotter than pages in node 0) while the
memory utilization is unchanged. Without the overlap, neither promotion
of hotter pages in node 1, nor demotion of colder pages in node 0 will
happen since both goals are met. As a result, the faster and slower node
will unexpectedly serve cold and hot data.
Test: Per-cgroup Memory Tiering
===============================
I ran a simplified cgroup level memory tiering using the feature, and
confirmed it works as intended.
Setup
-----
I configured a QEMU virtual machine representing a simplified version of
the system that described on the above cgroup level memory tiering example
use case. The system equips 40 CPU cores and two NUMA nodes each having
30 GiB physical memory. The first node (node 0) represents the faster
NUMA node, and the second node (node 1) represents the slower NUMA node.
In specific, below qemu command line options are used.
[...]
-object memory-backend-ram,size=30G,id=m0 \
-object memory-backend-ram,size=30G,id=m1 \
-numa node,cpus=0-39,memdev=m0 \
-numa node,memdev=m1 \
[...]
I booted the virtual machine with a kernel that this patch series is
applied. On the virtual machine, I created two cgroups, namely workload_a
and workload_b. And ran a test program in each cgroup, resulting in one
process per cgroup. The test program allocates 10 GiB memory and evenly
split it into 10 regions. After the allocation, it repeatedly access the
first region for one minute, than the second one for one minute, and so
on. After the one minute repeated access for the 10-th region is done, it
repeats the access from the first region. So the process has 10 GiB of
data in total, but only 1 GiB of it is hot at a given moment, and the hot
data is gradually changed.
While the processes are running, run DAMON for a simple access-aware
memory tiering using below script. It migrates hot and cold data of the
cgroups into node 0 and node 1, aiming the first and the second cgroups
(workload_a and workload_b, respectively) utilizing about 9.7 percent and
19.7 percent of node 0, respectively.
Note that this setup is a simplified version of the above example use
case, for ease of test. Also note that we assigned 30 GiB physical memory
to node 0, but DAMON in this setup works for only 27 GiB of the memory.
It is due to an internal implementation detail of DAMON user-space tool
that not really important for this test.
#!/bin/bash
damo start \
--numa_node 1 \
--damos_action migrate_hot 0 --damos_access_rate 5% max \
--damos_apply_interval 1s \
--damos_filter allow memcg /workload_a \
--damos_filter allow young \
--damos_quota_interval 1s \
--damos_quota_goal node_memcg_used_bp 9.7% 0 /workload_a \
--damos_action migrate_hot 0 --damos_access_rate 5% max \
--damos_apply_interval 1s \
--damos_filter allow memcg /workload_b \
--damos_filter allow young \
--damos_quota_interval 1s \
--damos_quota_goal node_memcg_used_bp 19.7% 0 /workload_b \
--numa_node 0 \
--damos_action migrate_cold 1 --damos_access_rate 0% 0% \
--damos_apply_interval 1s \
--damos_filter allow memcg /workload_a \
--damos_filter reject young \
--damos_quota_interval 1s \
--damos_quota_goal node_memcg_free_bp 90.5% 0 /workload_a \
--damos_action migrate_cold 1 --damos_access_rate 0% 0% \
--damos_apply_interval 1s \
--damos_filter allow memcg /workload_b \
--damos_filter reject young \
--damos_quota_interval 1s \
--damos_quota_goal node_memcg_free_bp 80.5% 0 /workload_b \
--damos_nr_quota_goals 1 1 1 1 --damos_nr_filters 2 2 2 2 \
--nr_targets 1 1 --nr_schemes 2 2 --nr_ctxs 1 1
After starting DAMON, the pages continuously be migrated across nodes. A
few minutes later, the memory usage of the cgroups converges into the
aimed amounts, and keeps the level, as expected. To confirm the status is
kept in the target level as expected, I collected the memory usage stat of
the cgroups using memory.numa_stat file, after the stats are converged. I
repeat the stat collection 42 times with 5 seconds delay between each of
the collections. The results are as below:
node0_memory_usage average stdev
workload_a 2.79GiB 522.06MiB
workload_b 5.15GiB 739.10MiB
The average values are quite close to the targeted values: 27 GiB * 9.7% =
2.619 GiB for workload_a, and 27 GiB * 19.7% = 5.319 GiB. A level of
variances are expected, given the overlap of the promotion/demotion
targets, and dynamic data access pattern of the workloads. Give that, the
measured variances are at a reasonable level.
Patches Sequence
================
The first patch (patch 1) updates the kernel-doc comment of
damos_quota_goal struct to clarify usage of optional fields of the struct,
since later patches will add such optional fields.
Following four patches (patches 2-5) implement a new DAMOS quota goal
metric for per-cgroup per-node memory usage. Those extends the core layer
interface for the new metric (patch 2), implement the metric value
calculation on the core layer (patch 3), add DAMON sysfs interface file
for the target cgroup specification (patch 4), and implement support of
the new metric on DAMON sysfs interface (patch 5).
Next two patches implment the second new DAMOS quota goal metric for
per-cgroup per-node free (or, unused) memory. Those implement it in the
core layer (patch 6) and DAMON sysfs interface (patch 7), extending the
existing implementation for memory usage metric.
Final three patches update the design (patch 8), the usage (patch 9), and
the ABI (patch 10) documents for the changes that are introduced by this
patch series.
This patch (of 10):
damos_quota_goal kerneldoc comment is not explaining when @metric is used.
Update the comment for that.
Link: https://lkml.kernel.org/r/20251017212706.183502-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20251017212706.183502-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "mm/damon: define and use DAMON initialization check
function".
DAMON is initialized in subsystem initialization time, by damon_init().
If DAMON API functions are called before the initialization, the
system could crash. Actually such issues happened and were fixed [1]
in the past. For the fix, DAMON API callers have updated to check if
DAMON is initialized or not, using their own hacks. The hacks are
unnecessarily duplicated on every DAMON API callers and therefore it
would be difficult to reliably maintain in the long term.
Make it reliable and easy to maintain. For this, implement a new DAMON
core layer API function that returns if DAMON is successfully
initialized. If it returns true, it means DAMON API functions are safe
to be used. After the introduction of the new API, update DAMON API
callers to use the new function instead of their own hacks.
This patch (of 7):
If DAMON is tried to be used when it is not yet successfully initialized,
the caller could be crashed. DAMON core layer is not providing a reliable
way to see if it is successfully initialized and therefore ready to be
used, though. As a result, DAMON API callers are implementing their own
hacks to see it. The hacks simply assume DAMON should be ready on module
init time. It is not reliable as DAMON initialization can indeed fail if
KMEM_CACHE() fails, and difficult to maintain as those are duplicates.
Implement a core layer API function for better reliability and
maintainability to replace the hacks with followup commits.
Link: https://lkml.kernel.org/r/20250916033511.116366-2-sj@kernel.org
Link: https://lkml.kernel.org/r/20250916033511.116366-2-sj@kernel.org
Link: https://lore.kernel.org/20250909022238.2989-1-sj@kernel.org [1]
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
changes required by mm-stable material: hugetlb and damon.
|