Instead of pre-allocating an entire array of struct binder_lru_page in
alloc->pages, install the shrinker metadata under page->private. This
ensures the memory is allocated and released as needed alongside pages.
By converting the alloc->pages[] into an array of struct page pointers,
we can access these pages directly and only reference the shrinker
metadata where it's being used (e.g. inside the shrinker's callback).
Rename struct binder_lru_page to struct binder_shrinker_mdata to better
reflect its purpose. Add convenience functions that wrap the allocation
and freeing of pages along with their shrinker metadata.
Note I've reworked this patch to avoid using page->lru and page->index
directly, as Matthew pointed out that these are being removed [1].
Link: https://lore.kernel.org/all/ZzziucEm3np6e7a0@casper.infradead.org/ [1]
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Liam R. Howlett <Liam.Howlett@oracle.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Link: https://lore.kernel.org/r/20241210143114.661252-5-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently, every list_lru has a per-node lock that protects adding,
deletion, isolation, and reparenting of all list_lru_one instances
belonging to this list_lru on this node. This lock contention is heavy
when multiple cgroups modify the same list_lru.
This lock can be split into per-cgroup scope to reduce contention.
To achieve this, we need a stable list_lru_one for every cgroup. This
commit adds a lock to each list_lru_one and introduced a helper function
lock_list_lru_of_memcg, making it possible to pin the list_lru of a memcg.
Then reworked the reparenting process.
Reparenting will switch the list_lru_one instances one by one. By locking
each instance and marking it dead using the nr_items counter, reparenting
ensures that all items in the corresponding cgroup (on-list or not,
because items have a stable cgroup, see below) will see the list_lru_one
switch synchronously.
Objcg reparent is also moved after list_lru reparent so items will have a
stable mem cgroup until all list_lru_one instances are drained.
The only caller that doesn't work the *_obj interfaces are direct calls to
list_lru_{add,del}. But it's only used by zswap and that's also based on
objcg, so it's fine.
This also changes the bahaviour of the isolation function when LRU_RETRY
or LRU_REMOVED_RETRY is returned, because now releasing the lock could
unblock reparenting and free the list_lru_one, isolation function will
have to return withoug re-lock the lru.
prepare() {
mkdir /tmp/test-fs
modprobe brd rd_nr=1 rd_size=33554432
mkfs.xfs -f /dev/ram0
mount -t xfs /dev/ram0 /tmp/test-fs
for i in $(seq 1 512); do
mkdir "/tmp/test-fs/$i"
for j in $(seq 1 10240); do
echo TEST-CONTENT > "/tmp/test-fs/$i/$j"
done &
done; wait
}
do_test() {
read_worker() {
sleep 1
tar -cv "$1" &>/dev/null
}
read_in_all() {
cd "/tmp/test-fs" && ls
for i in $(seq 1 512); do
(exec sh -c 'echo "$PPID"') > "/sys/fs/cgroup/benchmark/$i/cgroup.procs"
read_worker "$i" &
done; wait
}
for i in $(seq 1 512); do
mkdir -p "/sys/fs/cgroup/benchmark/$i"
done
echo +memory > /sys/fs/cgroup/benchmark/cgroup.subtree_control
echo 512M > /sys/fs/cgroup/benchmark/memory.max
echo 3 > /proc/sys/vm/drop_caches
time read_in_all
}
Above script simulates compression of small files in multiple cgroups
with memory pressure. Run prepare() then do_test for 6 times:
Before:
real 0m7.762s user 0m11.340s sys 3m11.224s
real 0m8.123s user 0m11.548s sys 3m2.549s
real 0m7.736s user 0m11.515s sys 3m11.171s
real 0m8.539s user 0m11.508s sys 3m7.618s
real 0m7.928s user 0m11.349s sys 3m13.063s
real 0m8.105s user 0m11.128s sys 3m14.313s
After this commit (about ~15% faster):
real 0m6.953s user 0m11.327s sys 2m42.912s
real 0m7.453s user 0m11.343s sys 2m51.942s
real 0m6.916s user 0m11.269s sys 2m43.957s
real 0m6.894s user 0m11.528s sys 2m45.346s
real 0m6.911s user 0m11.095s sys 2m43.168s
real 0m6.773s user 0m11.518s sys 2m40.774s
Link: https://lkml.kernel.org/r/20241104175257.60853-6-ryncsn@gmail.com
Signed-off-by: Kairui Song <kasong@tencent.com>
Cc: Chengming Zhou <zhouchengming@bytedance.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Waiman Long <longman@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Pull char/misc and other driver updates from Greg KH:
"Here is the big set of char/misc and other driver subsystem changes
for 6.8-rc1.
Other than lots of binder driver changes (as you can see by the merge
conflicts) included in here are:
- lots of iio driver updates and additions
- spmi driver updates
- eeprom driver updates
- firmware driver updates
- ocxl driver updates
- mhi driver updates
- w1 driver updates
- nvmem driver updates
- coresight driver updates
- platform driver remove callback api changes
- tags.sh script updates
- bus_type constant marking cleanups
- lots of other small driver updates
All of these have been in linux-next for a while with no reported
issues"
* tag 'char-misc-6.8-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: (341 commits)
android: removed duplicate linux/errno
uio: Fix use-after-free in uio_open
drivers: soc: xilinx: add check for platform
firmware: xilinx: Export function to use in other module
scripts/tags.sh: remove find_sources
scripts/tags.sh: use -n to test archinclude
scripts/tags.sh: add local annotation
scripts/tags.sh: use more portable -path instead of -wholename
scripts/tags.sh: Update comment (addition of gtags)
firmware: zynqmp: Convert to platform remove callback returning void
firmware: turris-mox-rwtm: Convert to platform remove callback returning void
firmware: stratix10-svc: Convert to platform remove callback returning void
firmware: stratix10-rsu: Convert to platform remove callback returning void
firmware: raspberrypi: Convert to platform remove callback returning void
firmware: qemu_fw_cfg: Convert to platform remove callback returning void
firmware: mtk-adsp-ipc: Convert to platform remove callback returning void
firmware: imx-dsp: Convert to platform remove callback returning void
firmware: coreboot_table: Convert to platform remove callback returning void
firmware: arm_scpi: Convert to platform remove callback returning void
firmware: arm_scmi: Convert to platform remove callback returning void
...
Patch series "workload-specific and memory pressure-driven zswap
writeback", v8.
There are currently several issues with zswap writeback:
1. There is only a single global LRU for zswap, making it impossible to
perform worload-specific shrinking - an memcg under memory pressure
cannot determine which pages in the pool it owns, and often ends up
writing pages from other memcgs. This issue has been previously
observed in practice and mitigated by simply disabling
memcg-initiated shrinking:
https://lore.kernel.org/all/20230530232435.3097106-1-nphamcs@gmail.com/T/#u
But this solution leaves a lot to be desired, as we still do not
have an avenue for an memcg to free up its own memory locked up in
the zswap pool.
2. We only shrink the zswap pool when the user-defined limit is hit.
This means that if we set the limit too high, cold data that are
unlikely to be used again will reside in the pool, wasting precious
memory. It is hard to predict how much zswap space will be needed
ahead of time, as this depends on the workload (specifically, on
factors such as memory access patterns and compressibility of the
memory pages).
This patch series solves these issues by separating the global zswap LRU
into per-memcg and per-NUMA LRUs, and performs workload-specific (i.e
memcg- and NUMA-aware) zswap writeback under memory pressure. The new
shrinker does not have any parameter that must be tuned by the user, and
can be opted in or out on a per-memcg basis.
As a proof of concept, we ran the following synthetic benchmark: build the
linux kernel in a memory-limited cgroup, and allocate some cold data in
tmpfs to see if the shrinker could write them out and improved the overall
performance. Depending on the amount of cold data generated, we observe
from 14% to 35% reduction in kernel CPU time used in the kernel builds.
This patch (of 6):
The interface of list_lru is based on the assumption that the list node
and the data it represents belong to the same allocated on the correct
node/memcg. While this assumption is valid for existing slab objects LRU
such as dentries and inodes, it is undocumented, and rather inflexible for
certain potential list_lru users (such as the upcoming zswap shrinker and
the THP shrinker). It has caused us a lot of issues during our
development.
This patch changes list_lru interface so that the caller must explicitly
specify numa node and memcg when adding and removing objects. The old
list_lru_add() and list_lru_del() are renamed to list_lru_add_obj() and
list_lru_del_obj(), respectively.
It also extends the list_lru API with a new function, list_lru_putback,
which undoes a previous list_lru_isolate call. Unlike list_lru_add, it
does not increment the LRU node count (as list_lru_isolate does not
decrement the node count). list_lru_putback also allows for explicit
memcg and NUMA node selection.
Link: https://lkml.kernel.org/r/20231130194023.4102148-1-nphamcs@gmail.com
Link: https://lkml.kernel.org/r/20231130194023.4102148-2-nphamcs@gmail.com
Signed-off-by: Nhat Pham <nphamcs@gmail.com>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Cc: Yosry Ahmed <yosryahmed@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The alloc->mutex is a highly contended lock that causes performance
issues on Android devices. When a low-priority task is given this lock
and it sleeps, it becomes difficult for the task to wake up and complete
its work. This delays other tasks that are also waiting on the mutex.
The problem gets worse when there is memory pressure in the system,
because this increases the contention on the alloc->mutex while the
shrinker reclaims binder pages.
Switching to a spinlock helps to keep the waiters running and avoids the
overhead of waking up tasks. This significantly improves the transaction
latency when the problematic scenario occurs.
The performance impact of this patchset was measured by stress-testing
the binder alloc contention. In this test, several clients of different
priorities send thousands of transactions of different sizes to a single
server. In parallel, pages get reclaimed using the shinker's debugfs.
The test was run on a Pixel 8, Pixel 6 and qemu machine. The results
were similar on all three devices:
after:
| sched | prio | average | max | min |
|--------+------+---------+-----------+---------|
| fifo | 99 | 0.135ms | 1.197ms | 0.022ms |
| fifo | 01 | 0.136ms | 5.232ms | 0.018ms |
| other | -20 | 0.180ms | 7.403ms | 0.019ms |
| other | 19 | 0.241ms | 58.094ms | 0.018ms |
before:
| sched | prio | average | max | min |
|--------+------+---------+-----------+---------|
| fifo | 99 | 0.350ms | 248.730ms | 0.020ms |
| fifo | 01 | 0.357ms | 248.817ms | 0.024ms |
| other | -20 | 0.399ms | 249.906ms | 0.020ms |
| other | 19 | 0.477ms | 297.756ms | 0.022ms |
The key metrics above are the average and max latencies (wall time).
These improvements should roughly translate to p95-p99 latencies on real
workloads. The response time is up to 200x faster in these scenarios and
there is no penalty in the regular path.
Note that it is only possible to convert this lock after a series of
changes made by previous patches. These mainly include refactoring the
sections that might_sleep() and changing the locking order with the
mmap_lock amongst others.
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Link: https://lore.kernel.org/r/20231201172212.1813387-29-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The locking order currently requires the alloc->mutex to be acquired
first followed by the mmap lock. However, the alloc->mutex is converted
into a spinlock in subsequent commits so the order needs to be reversed
to avoid nesting the sleeping mmap lock under the spinlock.
The shrinker's callback binder_alloc_free_page() is the only place that
needs to be reordered since other functions have been refactored and no
longer nest these locks.
Some minor cosmetic changes are also included in this patch.
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Link: https://lore.kernel.org/r/20231201172212.1813387-28-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The code to determine the page range for binder_lru_freelist_del() is
quite obscure. It leverages the buffer_size calculated before doing an
oversized buffer split. This is used to figure out if the last page is
being shared with another active buffer. If so, the page gets trimmed
out of the range as it has been previously removed from the freelist.
This would be equivalent to getting the start page of the next in-use
buffer explicitly. However, the code for this is much larger as we can
see in binder_free_buf_locked() routine. Instead, lets settle on
documenting the tricky step and using better names for now.
I believe an ideal solution would be to count the binder_page->users to
determine when a page should be added or removed from the freelist.
However, this is a much bigger change than what I'm willing to risk at
this time.
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Link: https://lore.kernel.org/r/20231201172212.1813387-24-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
