| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
git://git.kernel.org/pub/scm/linux/kernel/git/crng/random
Pull random number generator updates from Jason Donenfeld:
"This adds getrandom() support to the vDSO.
First, it adds a new kind of mapping to mmap(2), MAP_DROPPABLE, which
lets the kernel zero out pages anytime under memory pressure, which
enables allocating memory that never gets swapped to disk but also
doesn't count as being mlocked.
Then, the vDSO implementation of getrandom() is introduced in a
generic manner and hooked into random.c.
Next, this is implemented on x86. (Also, though it's not ready for
this pull, somebody has begun an arm64 implementation already)
Finally, two vDSO selftests are added.
There are also two housekeeping cleanup commits"
* tag 'random-6.11-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random:
MAINTAINERS: add random.h headers to RNG subsection
random: note that RNDGETPOOL was removed in 2.6.9-rc2
selftests/vDSO: add tests for vgetrandom
x86: vdso: Wire up getrandom() vDSO implementation
random: introduce generic vDSO getrandom() implementation
mm: add MAP_DROPPABLE for designating always lazily freeable mappings
|
|
The vDSO getrandom() implementation works with a buffer allocated with a
new system call that has certain requirements:
- It shouldn't be written to core dumps.
* Easy: VM_DONTDUMP.
- It should be zeroed on fork.
* Easy: VM_WIPEONFORK.
- It shouldn't be written to swap.
* Uh-oh: mlock is rlimited.
* Uh-oh: mlock isn't inherited by forks.
- It shouldn't reserve actual memory, but it also shouldn't crash when
page faulting in memory if none is available
* Uh-oh: VM_NORESERVE means segfaults.
It turns out that the vDSO getrandom() function has three really nice
characteristics that we can exploit to solve this problem:
1) Due to being wiped during fork(), the vDSO code is already robust to
having the contents of the pages it reads zeroed out midway through
the function's execution.
2) In the absolute worst case of whatever contingency we're coding for,
we have the option to fallback to the getrandom() syscall, and
everything is fine.
3) The buffers the function uses are only ever useful for a maximum of
60 seconds -- a sort of cache, rather than a long term allocation.
These characteristics mean that we can introduce VM_DROPPABLE, which
has the following semantics:
a) It never is written out to swap.
b) Under memory pressure, mm can just drop the pages (so that they're
zero when read back again).
c) It is inherited by fork.
d) It doesn't count against the mlock budget, since nothing is locked.
e) If there's not enough memory to service a page fault, it's not fatal,
and no signal is sent.
This way, allocations used by vDSO getrandom() can use:
VM_DROPPABLE | VM_DONTDUMP | VM_WIPEONFORK | VM_NORESERVE
And there will be no problem with OOMing, crashing on overcommitment,
using memory when not in use, not wiping on fork(), coredumps, or
writing out to swap.
In order to let vDSO getrandom() use this, expose these via mmap(2) as
MAP_DROPPABLE.
Note that this involves removing the MADV_FREE special case from
sort_folio(), which according to Yu Zhao is unnecessary and will simply
result in an extra call to shrink_folio_list() in the worst case. The
chunk removed reenables the swapbacked flag, which we don't want for
VM_DROPPABLE, and we can't conditionalize it here because there isn't a
vma reference available.
Finally, the provided self test ensures that this is working as desired.
Cc: linux-mm@kvack.org
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
|
|
mem_cgroup_calculate_protection() is not stateless and should only be used
as part of a top-down tree traversal. shrink_one() traverses the per-node
memcg LRU instead of the root_mem_cgroup tree, and therefore it should not
call mem_cgroup_calculate_protection().
The existing misuse in shrink_one() can cause ineffective protection of
sub-trees that are grandchildren of root_mem_cgroup. Fix it by reusing
lru_gen_age_node(), which already traverses the root_mem_cgroup tree, to
calculate the protection.
Previously lru_gen_age_node() opportunistically skips the first pass,
i.e., when scan_control->priority is DEF_PRIORITY. On the second pass,
lruvec_is_sizable() uses appropriate scan_control->priority, set by
set_initial_priority() from lru_gen_shrink_node(), to decide whether a
memcg is too small to reclaim from.
Now lru_gen_age_node() unconditionally traverses the root_mem_cgroup tree.
So it should call set_initial_priority() upfront, to make sure
lruvec_is_sizable() uses appropriate scan_control->priority on the first
pass. Otherwise, lruvec_is_reclaimable() can return false negatives and
result in premature OOM kills when min_ttl_ms is used.
Link: https://lkml.kernel.org/r/20240712232956.1427127-1-yuzhao@google.com
Fixes: e4dde56cd208 ("mm: multi-gen LRU: per-node lru_gen_folio lists")
Signed-off-by: Yu Zhao <yuzhao@google.com>
Reported-by: T.J. Mercier <tjmercier@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
set_initial_priority() tries to jump-start global reclaim by estimating
the priority based on cold/hot LRU pages. The estimation does not account
for shrinker objects, and it cannot do so because their sizes can be in
different units other than page.
If shrinker objects are the majority, e.g., on TrueNAS SCALE 24.04.0 where
ZFS ARC can use almost all system memory, set_initial_priority() can
vastly underestimate how much memory ARC shrinker can evict and assign
extreme low values to scan_control->priority, resulting in overshoots of
shrinker objects.
To reproduce the problem, using TrueNAS SCALE 24.04.0 with 32GB DRAM, a
test ZFS pool and the following commands:
fio --name=mglru.file --numjobs=36 --ioengine=io_uring \
--directory=/root/test-zfs-pool/ --size=1024m --buffered=1 \
--rw=randread --random_distribution=random \
--time_based --runtime=1h &
for ((i = 0; i < 20; i++))
do
sleep 120
fio --name=mglru.anon --numjobs=16 --ioengine=mmap \
--filename=/dev/zero --size=1024m --fadvise_hint=0 \
--rw=randrw --random_distribution=random \
--time_based --runtime=1m
done
To fix the problem:
1. Cap scan_control->priority at or above DEF_PRIORITY/2, to prevent
the jump-start from being overly aggressive.
2. Account for the progress from mm_account_reclaimed_pages(), to
prevent kswapd_shrink_node() from raising the priority
unnecessarily.
Link: https://lkml.kernel.org/r/20240711191957.939105-2-yuzhao@google.com
Fixes: e4dde56cd208 ("mm: multi-gen LRU: per-node lru_gen_folio lists")
Signed-off-by: Yu Zhao <yuzhao@google.com>
Reported-by: Alexander Motin <mav@ixsystems.com>
Cc: Wei Xu <weixugc@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
evict_folios() uses a second pass to reclaim folios that have gone through
page writeback and become clean before it finishes the first pass, since
folio_rotate_reclaimable() cannot handle those folios due to the
isolation.
The second pass tries to avoid potential double counting by deducting
scan_control->nr_scanned. However, this can result in underflow of
nr_scanned, under a condition where shrink_folio_list() does not increment
nr_scanned, i.e., when folio_trylock() fails.
The underflow can cause the divisor, i.e., scale=scanned+reclaimed in
vmpressure_calc_level(), to become zero, resulting in the following crash:
[exception RIP: vmpressure_work_fn+101]
process_one_work at ffffffffa3313f2b
Since scan_control->nr_scanned has no established semantics, the potential
double counting has minimal risks. Therefore, fix the problem by not
deducting scan_control->nr_scanned in evict_folios().
Link: https://lkml.kernel.org/r/20240711191957.939105-1-yuzhao@google.com
Fixes: 359a5e1416ca ("mm: multi-gen LRU: retry folios written back while isolated")
Reported-by: Wei Xu <weixugc@google.com>
Signed-off-by: Yu Zhao <yuzhao@google.com>
Cc: Alexander Motin <mav@ixsystems.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The optimization of list_empty(&folio->_deferred_list) aimed to prevent
increasing the PTL duration when a large folio is partially unmapped, for
example, from subpage 0 to subpage (nr - 2).
But Ryan's commit 5ed890ce5147 ("mm: vmscan: avoid split during
shrink_folio_list()") actually splits this kind of large folios. This
makes the "optimization" useless.
Additionally, the list_empty() technically required a data_race()
annotation.
Link: https://lkml.kernel.org/r/20240629234155.53524-1-21cnbao@gmail.com
Signed-off-by: Barry Song <v-songbaohua@oppo.com>
Reviewed-by: Ryan Roberts <ryan.roberts@arm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Allow proactive reclaimers to submit an additional swappiness=<val>
argument to memory.reclaim. This overrides the global or per-memcg
swappiness setting for that reclaim attempt.
For example:
echo "2M swappiness=0" > /sys/fs/cgroup/memory.reclaim
will perform reclaim on the rootcg with a swappiness setting of 0 (no
swap) regardless of the vm.swappiness sysctl setting.
Userspace proactive reclaimers use the memory.reclaim interface to trigger
reclaim. The memory.reclaim interface does not allow for any way to
effect the balance of file vs anon during proactive reclaim. The only
approach is to adjust the vm.swappiness setting. However, there are a few
reasons we look to control the balance of file vs anon during proactive
reclaim, separately from reactive reclaim:
* Swapout should be limited to manage SSD write endurance. In near-OOM
situations we are fine with lots of swap-out to avoid OOMs. As these
are typically rare events, they have relatively little impact on write
endurance. However, proactive reclaim runs continuously and so its
impact on SSD write endurance is more significant. Therefore it is
desireable to control swap-out for proactive reclaim separately from
reactive reclaim
* Some userspace OOM killers like systemd-oomd[1] support OOM killing on
swap exhaustion. This makes sense if the swap exhaustion is triggered
due to reactive reclaim but less so if it is triggered due to proactive
reclaim (e.g. one could see OOMs when free memory is ample but anon is
just particularly cold). Therefore, it's desireable to have proactive
reclaim reduce or stop swap-out before the threshold at which OOM
killing occurs.
In the case of Meta's Senpai proactive reclaimer, we adjust vm.swappiness
before writes to memory.reclaim[2]. This has been in production for
nearly two years and has addressed our needs to control proactive vs
reactive reclaim behavior but is still not ideal for a number of reasons:
* vm.swappiness is a global setting, adjusting it can race/interfere
with other system administration that wishes to control vm.swappiness.
In our case, we need to disable Senpai before adjusting vm.swappiness.
* vm.swappiness is stateful - so a crash or restart of Senpai can leave
a misconfigured setting. This requires some additional management to
record the "desired" setting and ensure Senpai always adjusts to it.
With this patch, we avoid these downsides of adjusting vm.swappiness
globally.
[1]https://www.freedesktop.org/software/systemd/man/latest/systemd-oomd.service.html
[2]https://github.com/facebookincubator/oomd/blob/main/src/oomd/plugins/Senpai.cpp#L585-L598
Link: https://lkml.kernel.org/r/20240103164841.2800183-3-schatzberg.dan@gmail.com
Signed-off-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Suggested-by: Yosry Ahmed <yosryahmed@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Chris Li <chrisl@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Yue Zhao <findns94@gmail.com>
Cc: Zefan Li <lizefan.x@bytedance.com>
Cc: Nhat Pham <nphamcs@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "Add swappiness argument to memory.reclaim", v6.
This patch proposes augmenting the memory.reclaim interface with a
swappiness=<val> argument that overrides the swappiness value for that
instance of proactive reclaim.
Userspace proactive reclaimers use the memory.reclaim interface to trigger
reclaim. The memory.reclaim interface does not allow for any way to
effect the balance of file vs anon during proactive reclaim. The only
approach is to adjust the vm.swappiness setting. However, there are a few
reasons we look to control the balance of file vs anon during proactive
reclaim, separately from reactive reclaim:
* Swapout should be limited to manage SSD write endurance. In near-OOM
situations we are fine with lots of swap-out to avoid OOMs. As these
are typically rare events, they have relatively little impact on write
endurance. However, proactive reclaim runs continuously and so its
impact on SSD write endurance is more significant. Therefore it is
desireable to control swap-out for proactive reclaim separately from
reactive reclaim
* Some userspace OOM killers like systemd-oomd[1] support OOM killing on
swap exhaustion. This makes sense if the swap exhaustion is triggered
due to reactive reclaim but less so if it is triggered due to proactive
reclaim (e.g. one could see OOMs when free memory is ample but anon is
just particularly cold). Therefore, it's desireable to have proactive
reclaim reduce or stop swap-out before the threshold at which OOM
killing occurs.
In the case of Meta's Senpai proactive reclaimer, we adjust vm.swappiness
before writes to memory.reclaim[2]. This has been in production for
nearly two years and has addressed our needs to control proactive vs
reactive reclaim behavior but is still not ideal for a number of reasons:
* vm.swappiness is a global setting, adjusting it can race/interfere
with other system administration that wishes to control vm.swappiness.
In our case, we need to disable Senpai before adjusting vm.swappiness.
* vm.swappiness is stateful - so a crash or restart of Senpai can leave
a misconfigured setting. This requires some additional management to
record the "desired" setting and ensure Senpai always adjusts to it.
With this patch, we avoid these downsides of adjusting vm.swappiness
globally.
Previously, this exact interface addition was proposed by Yosry[3]. In
response, Roman proposed instead an interface to specify precise
file/anon/slab reclaim amounts[4]. More recently Huan also proposed this
as well[5] and others similarly questioned if this was the proper
interface.
Previous proposals sought to use this to allow proactive reclaimers to
effectively perform a custom reclaim algorithm by issuing proactive
reclaim with different settings to control file vs anon reclaim (e.g. to
only reclaim anon from some applications). Responses argued that
adjusting swappiness is a poor interface for custom reclaim.
In contrast, I argue in favor of a swappiness setting not as a way to
implement custom reclaim algorithms but rather to bias the balance of anon
vs file due to differences of proactive vs reactive reclaim. In this
context, swappiness is the existing interface for controlling this balance
and this patch simply allows for it to be configured differently for
proactive vs reactive reclaim.
Specifying explicit amounts of anon vs file pages to reclaim feels
inappropriate for this prupose. Proactive reclaimers are un-aware of the
relative age of file vs anon for a cgroup which makes it difficult to
manage proactive reclaim of different memory pools. A proactive reclaimer
would need some amount of anon reclaim attempts separate from the amount
of file reclaim attempts which seems brittle given that it's difficult to
observe the impact.
[1]https://www.freedesktop.org/software/systemd/man/latest/systemd-oomd.service.html
[2]https://github.com/facebookincubator/oomd/blob/main/src/oomd/plugins/Senpai.cpp#L585-L598
[3]https://lore.kernel.org/linux-mm/CAJD7tkbDpyoODveCsnaqBBMZEkDvshXJmNdbk51yKSNgD7aGdg@mail.gmail.com/
[4]https://lore.kernel.org/linux-mm/YoPHtHXzpK51F%2F1Z@carbon/
[5]https://lore.kernel.org/lkml/20231108065818.19932-1-link@vivo.com/
This patch (of 2):
We use the constants 0 and 200 in a few places in the mm code when
referring to the min and max swappiness. This patch adds MIN_SWAPPINESS
and MAX_SWAPPINESS #defines to improve clarity. There are no functional
changes.
Link: https://lkml.kernel.org/r/20240103164841.2800183-1-schatzberg.dan@gmail.com
Link: https://lkml.kernel.org/r/20240103164841.2800183-2-schatzberg.dan@gmail.com
Signed-off-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Chris Li <chrisl@kernel.org>
Reviewed-by: Nhat Pham <nphamcs@gmail.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Tejun Heo <tj@kernel.org>
Cc: Yosry Ahmed <yosryahmed@google.com>
Cc: Yue Zhao <findns94@gmail.com>
Cc: Zefan Li <lizefan.x@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Rename exported function related to the softlimit reclaim to have memcg1_
prefix.
Link: https://lkml.kernel.org/r/20240625005906.106920-4-roman.gushchin@linux.dev
Signed-off-by: Roman Gushchin <roman.gushchin@linux.dev>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Muchun Song <muchun.song@linux.dev>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Folios of order <= 1 are not in deferred list, the check of order is added
into folio_undo_large_rmappable() from commit 8897277acfef ("mm: support
order-1 folios in the page cache"), but there is a repeated check for
small folio (order 0) during each call of the
folio_undo_large_rmappable(), so only keep folio_order() check inside the
function.
In addition, move all the checks into header file to save a function call
for non-large-rmappable or empty deferred_list folio.
Link: https://lkml.kernel.org/r/20240521130315.46072-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Lance Yang <ioworker0@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeel.butt@linux.dev>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The alloc_demote_folio can also be used for general migration including
both demotion and promotion so it'd be better to rename it from
alloc_demote_folio to alloc_migrate_folio.
Link: https://lkml.kernel.org/r/20240614030010.751-3-honggyu.kim@sk.com
Signed-off-by: Honggyu Kim <honggyu.kim@sk.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: Gregory Price <gregory.price@memverge.com>
Cc: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Hyeongtak Ji <hyeongtak.ji@sk.com>
Cc: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Rakie Kim <rakie.kim@sk.com>
Cc: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "DAMON based tiered memory management for CXL memory", v6.
Introduction
============
With the advent of CXL/PCIe attached DRAM, which will be called simply as
CXL memory in this cover letter, some systems are becoming more
heterogeneous having memory systems with different latency and bandwidth
characteristics. They are usually handled as different NUMA nodes in
separate memory tiers and CXL memory is used as slow tiers because of its
protocol overhead compared to local DRAM.
In this kind of systems, we need to be careful placing memory pages on
proper NUMA nodes based on the memory access frequency. Otherwise, some
frequently accessed pages might reside on slow tiers and it makes
performance degradation unexpectedly. Moreover, the memory access
patterns can be changed at runtime.
To handle this problem, we need a way to monitor the memory access
patterns and migrate pages based on their access temperature. The
DAMON(Data Access MONitor) framework and its DAMOS(DAMON-based Operation
Schemes) can be useful features for monitoring and migrating pages. DAMOS
provides multiple actions based on DAMON monitoring results and it can be
used for proactive reclaim, which means swapping cold pages out with
DAMOS_PAGEOUT action, but it doesn't support migration actions such as
demotion and promotion between tiered memory nodes.
This series supports two new DAMOS actions; DAMOS_MIGRATE_HOT for
promotion from slow tiers and DAMOS_MIGRATE_COLD for demotion from fast
tiers. This prevents hot pages from being stuck on slow tiers, which
makes performance degradation and cold pages can be proactively demoted to
slow tiers so that the system can increase the chance to allocate more hot
pages to fast tiers.
The DAMON provides various tuning knobs but we found that the proactive
demotion for cold pages is especially useful when the system is running
out of memory on its fast tier nodes.
Our evaluation result shows that it reduces the performance slowdown
compared to the default memory policy from 11% to 3~5% when the system
runs under high memory pressure on its fast tier DRAM nodes.
DAMON configuration
===================
The specific DAMON configuration doesn't have to be in the scope of this
patch series, but some rough idea is better to be shared to explain the
evaluation result.
The DAMON provides many knobs for fine tuning but its configuration file
is generated by HMSDK[3]. It includes gen_config.py script that generates
a json file with the full config of DAMON knobs and it creates multiple
kdamonds for each NUMA node when the DAMON is enabled so that it can run
hot/cold based migration for tiered memory.
Evaluation Workload
===================
The performance evaluation is done with redis[4], which is a widely used
in-memory database and the memory access patterns are generated via
YCSB[5]. We have measured two different workloads with zipfian and latest
distributions but their configs are slightly modified to make memory usage
higher and execution time longer for better evaluation.
The idea of evaluation using these migrate_{hot,cold} actions covers
system-wide memory management rather than partitioning hot/cold pages of a
single workload. The default memory allocation policy creates pages to
the fast tier DRAM node first, then allocates newly created pages to the
slow tier CXL node when the DRAM node has insufficient free space. Once
the page allocation is done then those pages never move between NUMA
nodes. It's not true when using numa balancing, but it is not the scope
of this DAMON based tiered memory management support.
If the working set of redis can be fit fully into the DRAM node, then the
redis will access the fast DRAM only. Since the performance of DRAM only
is faster than partially accessing CXL memory in slow tiers, this
environment is not useful to evaluate this patch series.
To make pages of redis be distributed across fast DRAM node and slow CXL
node to evaluate our migrate_{hot,cold} actions, we pre-allocate some cold
memory externally using mmap and memset before launching redis-server. We
assumed that there are enough amount of cold memory in datacenters as
TMO[6] and TPP[7] papers mentioned.
The evaluation sequence is as follows.
1. Turn on DAMON with DAMOS_MIGRATE_COLD action for DRAM node and
DAMOS_MIGRATE_HOT action for CXL node. It demotes cold pages on DRAM
node and promotes hot pages on CXL node in a regular interval.
2. Allocate a huge block of cold memory by calling mmap and memset at
the fast tier DRAM node, then make the process sleep to make the fast
tier has insufficient space for redis-server.
3. Launch redis-server and load prebaked snapshot image, dump.rdb. The
redis-server consumes 52GB of anon pages and 33GB of file pages, but
due to the cold memory allocated at 2, it fails allocating the entire
memory of redis-server on the fast tier DRAM node so it partially
allocates the remaining on the slow tier CXL node. The ratio of
DRAM:CXL depends on the size of the pre-allocated cold memory.
4. Run YCSB to make zipfian or latest distribution of memory accesses to
redis-server, then measure its execution time when it's completed.
5. Repeat 4 over 50 times to measure the average execution time for each
run.
6. Increase the cold memory size then repeat goes to 2.
For each test at 4 took about a minute so repeating it 50 times almost
took about 1 hour for each test with a specific cold memory from 440GB to
500GB in 10GB increments for each evaluation. So it took about more than
10 hours for both zipfian and latest workloads to get the entire
evaluation results. Repeating the same test set multiple times doesn't
show much difference so I think it might be enough to make the result
reliable.
Evaluation Results
==================
All the result values are normalized to DRAM-only execution time because
the workload cannot be faster than DRAM-only unless the workload hits the
peak bandwidth but our redis test doesn't go beyond the bandwidth limit.
So the DRAM-only execution time is the ideal result without affected by
the gap between DRAM and CXL performance difference. The NUMA node
environment is as follows.
node0 - local DRAM, 512GB with a CPU socket (fast tier)
node1 - disabled
node2 - CXL DRAM, 96GB, no CPU attached (slow tier)
The following is the result of generating zipfian distribution to
redis-server and the numbers are averaged by 50 times of execution.
1. YCSB zipfian distribution read only workload
memory pressure with cold memory on node0 with 512GB of local DRAM.
====================+================================================+=========
| cold memory occupied by mmap and memset |
| 0G 440G 450G 460G 470G 480G 490G 500G |
====================+================================================+=========
Execution time normalized to DRAM-only values | GEOMEAN
--------------------+------------------------------------------------+---------
DRAM-only | 1.00 - - - - - - - | 1.00
CXL-only | 1.19 - - - - - - - | 1.19
default | - 1.00 1.05 1.08 1.12 1.14 1.18 1.18 | 1.11
DAMON tiered | - 1.03 1.03 1.03 1.03 1.03 1.07 *1.05 | 1.04
DAMON lazy | - 1.04 1.03 1.04 1.05 1.06 1.06 *1.06 | 1.05
====================+================================================+=========
CXL usage of redis-server in GB | AVERAGE
--------------------+------------------------------------------------+---------
DRAM-only | 0.0 - - - - - - - | 0.0
CXL-only | 51.4 - - - - - - - | 51.4
default | - 0.6 10.6 20.5 30.5 40.5 47.6 50.4 | 28.7
DAMON tiered | - 0.6 0.5 0.4 0.7 0.8 7.1 5.6 | 2.2
DAMON lazy | - 0.5 3.0 4.5 5.4 6.4 9.4 9.1 | 5.5
====================+================================================+=========
Each test result is based on the execution environment as follows.
DRAM-only: redis-server uses only local DRAM memory.
CXL-only: redis-server uses only CXL memory.
default: default memory policy(MPOL_DEFAULT).
numa balancing disabled.
DAMON tiered: DAMON enabled with DAMOS_MIGRATE_COLD for DRAM
nodes and DAMOS_MIGRATE_HOT for CXL nodes.
DAMON lazy: same as DAMON tiered, but turn on DAMON just
before making memory access request via YCSB.
The above result shows the "default" execution time goes up as the size of
cold memory is increased from 440G to 500G because the more cold memory
used, the more CXL memory is used for the target redis workload and this
makes the execution time increase.
However, "DAMON tiered" and other DAMON results show less slowdown because
the DAMOS_MIGRATE_COLD action at DRAM node proactively demotes
pre-allocated cold memory to CXL node and this free space at DRAM
increases more chance to allocate hot or warm pages of redis-server to
fast DRAM node. Moreover, DAMOS_MIGRATE_HOT action at CXL node also
promotes hot pages of redis-server to DRAM node actively.
As a result, it makes more memory of redis-server stay in DRAM node
compared to "default" memory policy and this makes the performance
improvement.
Please note that the result numbers of "DAMON tiered" and "DAMON lazy" at
500G are marked with * stars, which means their test results are replaced
with reproduced tests that didn't have OOM issue.
That was needed because sometimes the test processes get OOM when DRAM has
insufficient space. The DAMOS_MIGRATE_HOT doesn't kick reclaim but just
gives up migration when there is not enough space at DRAM side. The
problem happens when there is competition between normal allocation and
migration and the migration is done before normal allocation, then the
completely unrelated normal allocation can trigger reclaim, which incurs
OOM.
Because of this issue, I have also tested more cases with
"demotion_enabled" flag enabled to make such reclaim doesn't trigger OOM,
but just demote reclaimed pages. The following test results show more
tests with "kswapd" marked.
2. YCSB zipfian distribution read only workload (with demotion_enabled true)
memory pressure with cold memory on node0 with 512GB of local DRAM.
====================+================================================+=========
| cold memory occupied by mmap and memset |
| 0G 440G 450G 460G 470G 480G 490G 500G |
====================+================================================+=========
Execution time normalized to DRAM-only values | GEOMEAN
--------------------+------------------------------------------------+---------
DAMON tiered | - 1.03 1.03 1.03 1.03 1.03 1.07 1.05 | 1.04
DAMON lazy | - 1.04 1.03 1.04 1.05 1.06 1.06 1.06 | 1.05
DAMON tiered kswapd | - 1.03 1.03 1.03 1.03 1.02 1.02 1.03 | 1.03
DAMON lazy kswapd | - 1.04 1.04 1.04 1.03 1.05 1.04 1.05 | 1.04
====================+================================================+=========
CXL usage of redis-server in GB | AVERAGE
--------------------+------------------------------------------------+---------
DAMON tiered | - 0.6 0.5 0.4 0.7 0.8 7.1 5.6 | 2.2
DAMON lazy | - 0.5 3.0 4.5 5.4 6.4 9.4 9.1 | 5.5
DAMON tiered kswapd | - 0.0 0.0 0.4 0.5 0.1 0.8 1.0 | 0.4
DAMON lazy kswapd | - 4.2 4.6 5.3 1.7 6.8 8.1 5.8 | 5.2
====================+================================================+=========
Each test result is based on the exeuction environment as follows.
DAMON tiered: same as before
DAMON lazy: same as before
DAMON tiered kswapd: same as DAMON tiered, but turn on
/sys/kernel/mm/numa/demotion_enabled to make
kswapd or direct reclaim does demotion.
DAMON lazy kswapd: same as DAMON lazy, but turn on
/sys/kernel/mm/numa/demotion_enabled to make
kswapd or direct reclaim does demotion.
The "DAMON tiered kswapd" and "DAMON lazy kswapd" didn't trigger OOM at
all unlike other tests because kswapd and direct reclaim from DRAM node
can demote reclaimed pages to CXL node independently from DAMON actions
and their results are slightly better than without having
"demotion_enabled".
In summary, the evaluation results show that DAMON memory management with
DAMOS_MIGRATE_{HOT,COLD} actions reduces the performance slowdown compared
to the "default" memory policy from 11% to 3~5% when the system runs with
high memory pressure on its fast tier DRAM nodes.
Having these DAMOS_MIGRATE_HOT and DAMOS_MIGRATE_COLD actions can make
tiered memory systems run more efficiently under high memory pressures.
This patch (of 7):
The alloc_demote_folio can be used out of vmscan.c so it'd be better to
remove static keyword from it.
Link: https://lkml.kernel.org/r/20240614030010.751-1-honggyu.kim@sk.com
Link: https://lkml.kernel.org/r/20240614030010.751-2-honggyu.kim@sk.com
Signed-off-by: Honggyu Kim <honggyu.kim@sk.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Gregory Price <gregory.price@memverge.com>
Cc: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Hyeongtak Ji <hyeongtak.ji@sk.com>
Cc: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Rakie Kim <rakie.kim@sk.com>
Cc: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The commit 6be5e186fd65 ("mm: vmscan: restore incremental cgroup
iteration") added a retry reclaim heuristic to iterate all the cgroups
before returning an unsuccessful reclaim but missed to reset the
sc->priority. Let's fix it.
Link: https://lkml.kernel.org/r/20240529154911.3008025-1-shakeel.butt@linux.dev
Fixes: 6be5e186fd65 ("mm: vmscan: restore incremental cgroup iteration")
Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
Reported-by: syzbot+17416257cb95200cba44@syzkaller.appspotmail.com
Tested-by: syzbot+17416257cb95200cba44@syzkaller.appspotmail.com
Reviewed-by: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Currently, reclaim always walks the entire cgroup tree in order to ensure
fairness between groups. While overreclaim is limited in shrink_lruvec(),
many of our systems have a sizable number of active groups, and an even
bigger number of idle cgroups with cache left behind by previous jobs; the
mere act of walking all these cgroups can impose significant latency on
direct reclaimers.
In the past, we've used a save-and-restore iterator that enabled
incremental tree walks over multiple reclaim invocations. This ensured
fairness, while keeping the work of individual reclaimers small.
However, in edge cases with a lot of reclaim concurrency, individual
reclaimers would sometimes not see enough of the cgroup tree to make
forward progress and (prematurely) declare OOM. Consequently we switched
to comprehensive walks in 1ba6fc9af35b ("mm: vmscan: do not share cgroup
iteration between reclaimers").
To address the latency problem without bringing back the premature OOM
issue, reinstate the shared iteration, but with a restart condition to do
the full walk in the OOM case - similar to what we do for memory.low
enforcement and active page protection.
In the worst case, we do one more full tree walk before declaring
OOM. But the vast majority of direct reclaim scans can then finish
much quicker, while fairness across the tree is maintained:
- Before this patch, we observed that direct reclaim always takes more
than 100us and most direct reclaim time is spent in reclaim cycles
lasting between 1ms and 1 second. Almost 40% of direct reclaim time
was spent on reclaim cycles exceeding 100ms.
- With this patch, almost all page reclaim cycles last less than 10ms,
and a good amount of direct page reclaim finishes in under 100us. No
page reclaim cycles lasting over 100ms were observed anymore.
The shared iterator state is maintaned inside the target cgroup, so
fair and incremental walks are performed during both global reclaim
and cgroup limit reclaim of complex subtrees.
Link: https://lkml.kernel.org/r/20240514202641.2821494-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Rik van Riel <riel@surriel.com>
Reported-by: Rik van Riel <riel@surriel.com>
Reviewed-by: Shakeel Butt <shakeel.butt@linux.dev>
Reviewed-by: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Facebook Kernel Team <kernel-team@fb.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
The mTHP swap related counters: 'anon_swpout' and 'anon_swpout_fallback'
are confusing with an 'anon_' prefix, since the shmem can swap out
non-anonymous pages. So drop the 'anon_' prefix to keep consistent with
the old swap counter names.
This is needed in 6.10-rcX to avoid having an inconsistent ABI out in the
field.
Link: https://lkml.kernel.org/r/7a8989c13299920d7589007a30065c3e2c19f0e0.1716431702.git.baolin.wang@linux.alibaba.com
Fixes: d0f048ac39f6 ("mm: add per-order mTHP anon_swpout and anon_swpout_fallback counters")
Fixes: 42248b9d34ea ("mm: add docs for per-order mTHP counters and transhuge_page ABI")
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Suggested-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Barry Song <baohua@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Lance Yang <ioworker0@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
All reclaim_folio_list() callers are passing 'true' for
'ignore_references' parameter. In other words, the parameter is not
really being used. Simplify the code by removing the parameter.
Link: https://lkml.kernel.org/r/20240429224451.67081-5-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
All reclaim_pages() callers are setting 'ignore_references' parameter
'true'. In other words, the parameter is not really being used. Remove
the argument to make it simple.
Link: https://lkml.kernel.org/r/20240429224451.67081-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
This helps to display the fragmentation situation of the swapfile, knowing
the proportion of how much we haven't split large folios. So far, we only
support non-split swapout for anon memory, with the possibility of
expanding to shmem in the future. So, we add the "anon" prefix to the
counter names.
Link: https://lkml.kernel.org/r/20240412114858.407208-3-21cnbao@gmail.com
Signed-off-by: Barry Song <v-songbaohua@oppo.com>
Reviewed-by: Ryan Roberts <ryan.roberts@arm.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Cc: Kairui Song <kasong@tencent.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Yosry Ahmed <yosryahmed@google.com>
Cc: Yu Zhao <yuzhao@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Now that swap supports storing all mTHP sizes, avoid splitting large
folios before swap-out. This benefits performance of the swap-out path by
eliding split_folio_to_list(), which is expensive, and also sets us up for
swapping in large folios in a future series.
If the folio is partially mapped, we continue to split it since we want to
avoid the extra IO overhead and storage of writing out pages
uneccessarily.
THP_SWPOUT and THP_SWPOUT_FALLBACK counters should continue to count
events only for PMD-mappable folios to avoid user confusion. THP_SWPOUT
already has the appropriate guard. Add a guard for THP_SWPOUT_FALLBACK.
It may be appropriate to add per-size counters in future.
Link: https://lkml.kernel.org/r/20240408183946.2991168-7-ryan.roberts@arm.com
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Barry Song <v-songbaohua@oppo.com>
Cc: Barry Song <21cnbao@gmail.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Gao Xiang <xiang@kernel.org>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Lance Yang <ioworker0@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Within try_to_unmap_one(), page_vma_mapped_walk() races with other PTE
modifications preceded by pte clear. While iterating over PTEs of a large
folio, it only starts acquiring PTL from the first valid (present) PTE.
PTE modifications can temporarily set PTEs to pte_none. Consequently, the
initial PTEs of a large folio might be skipped in try_to_unmap_one().
For example, for an anon folio, if we skip PTE0, we may have PTE0 which is
still present, while PTE1 ~ PTE(nr_pages - 1) are swap entries after
try_to_unmap_one().
So folio will be still mapped, the folio fails to be reclaimed and is put
back to LRU in this round.
This also breaks up PTEs optimization such as CONT-PTE on this large folio
and may lead to accident folio_split() afterwards. And since a part of
PTEs are now swap entries, accessing those parts will introduce overhead -
do_swap_page. Although the kernel can withstand all of the above issues,
the situation still seems quite awkward and warrants making it more ideal.
The same race also occurs with small folios, but they have only one PTE,
thus, it won't be possible for them to be partially unmapped.
This patch holds PTL from PTE0, allowing us to avoid reading PTE values
that are in the process of being transformed. With stable PTE values, we
can ensure that this large folio is either completely reclaimed or that
all PTEs remain untouched in this round.
A corner case is that if we hold PTL from PTE0 and most initial PTEs have
been really unmapped before that, we may increase the duration of holding
PTL. Thus we only apply this optimization to folios which are still
entirely mapped (not in deferred_split list).
[akpm@linux-foundation.org: rewrap comment, per Matthew]
Link: https://lkml.kernel.org/r/20240306095219.71086-1-21cnbao@gmail.com
Signed-off-by: Barry Song <v-songbaohua@oppo.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Chuanhua Han <hanchuanhua@oppo.com>
Cc: Gao Xiang <xiang@kernel.org>
Cc: Huang, Ying <ying.huang@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
Patch series "make the hugetlb migration strategy consistent", v2.
As discussed in previous thread [1], there is an inconsistency when
handling hugetlb migration. When handling the migration of freed hugetlb,
it prevents fallback to other NUMA nodes in
alloc_and_dissolve_hugetlb_folio(). However, when dealing with in-use
hugetlb, it allows fallback to other NUMA nodes in
alloc_hugetlb_folio_nodemask(), which can break the per-node hugetlb pool
and might result in unexpected failures when node bound workloads doesn't
get what is asssumed available.
This patchset tries to make the hugetlb migration strategy more clear
and consistent. Please find details in each patch.
[1]
https://lore.kernel.org/all/6f26ce22d2fcd523418a085f2c588fe0776d46e7.1706794035.git.baolin.wang@linux.alibaba.com/
This patch (of 2):
To support different hugetlb allocation strategies during hugetlb
migration based on various migration reasons, record the migration reason
in the migration_target_control structure as a preparation.
Link: https://lkml.kernel.org/r/cover.1709719720.git.baolin.wang@linux.alibaba.com
Link: https://lkml.kernel.org/r/7b95d4981e07211f57139fc5b1f7ce91b920cee4.1709719720.git.baolin.wang@linux.alibaba.com
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
|
|
With cache_trim_mode on, reclaim logic doesn't bother reclaiming anon
pages. However, it should be more careful to use the mode because it's
going to prevent anon pages from being reclaimed even if there are a huge
number of anon pages that are cold and should be reclaimed. Even worse,
that leads kswapd_failures to reach MAX_RECLAIM_RETRIES and stopping
kswapd from functioning until direct reclaim eventually works to resume
kswapd.
So kswapd needs to retry its scan priority loop with cache_trim_mode off
again if the mode doesn't work for reclaim.
The problematic behavior can be reproduced by:
CONFIG_NUMA_BALANCING enabled
sysctl_numa_balancing_mode set to NUMA_BALANCING_MEMORY_TIERING
numa node0 (8GB local memory, 16 CPUs)
numa node1 (8GB slow tier memory, no CPUs)
Sequence:
1) echo 3 > /proc/sys/vm/drop_caches
2) To emulate the system with full of cold memory in local DRAM, run
the following dummy program and never touch the region:
mmap(0, 8 * 1024 * 1024 * 1024, PROT_READ | PROT_WRITE,
MAP_ANONYMOUS | MAP_PRIVATE | MAP_POPULATE, -1, 0);
3) Run any memory intensive work e.g. XSBench.
4) Check if numa balancing is working e.i. promotion/demotion.
5) Iterate 1) ~ 4) until numa balancing stops.
With this, you could see that promotion/demotion are not working because
kswapd has stopped due to ->kswapd_failures >= MAX_RECLAIM_RETRIES.
Interesting vmstat delta's differences between before and after are like:
+-----------------------+-------------------------------+
| interesting vmstat | before | after |
+-----------------------+-------------------------------+
| nr_inactive_anon | 321935 | 1664772 |
| nr_active_anon | 1780700 | 437834 |
| nr_inactive_file | 30425 | 40882 |
| nr_active_file | 14961 | 3012 |
| pgpromote_success | 356 | 1293122 |
| pgpromote_candidate | 21953245 | 1824148 |
| pgactivate | 1844523 | 3311907 |
| pgdeactivate | 50634 | 1554069 |
| pgfault | 31100294 | 6518806 |
| pgdemote_kswapd | 30856 | 2230821 |
| pgscan_kswapd | 1861981 | 7667629 |
| pgscan_anon | 1822930 | 7610583 |
| pgscan_file | 39051 | 57046 |
| pgsteal_anon | 386 | 2192033 |
| pgsteal_file | 30470 | 38788 |
| pageoutrun | 30 | 412 |
| numa_hint_faults | 27418279 | 2875955 |
| numa_pages_migrated | 356 | 1293122 |
+-----------------------+-------------------------------+
Link: https://lkml.kernel.org/r/20240304082118.20499-1-byungchul@sk.com
Signed-off-by: Byungchul Park <byungchul@sk.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
When freeing a large folio, we must remove it from the deferred split list
before we uncharge it as each memcg has its own deferred split list (with
associated lock) and removing a folio from the deferred split list while
holding the wrong lock will corrupt that list and cause various related
problems.
Link: https://lore.kernel.org/linux-mm/367a14f7-340e-4b29-90ae-bc3 |
