Patch series "mm/hugetlb: refactor sysfs/sysctl interfaces", v5.
hugetlb.c has grown significantly and become difficult to maintain. This
patch series extracts the sysfs and sysctl interface code into separate
dedicated files to improve code organization.
The refactoring includes:
- Patch 1: Extract sysfs interface into mm/hugetlb_sysfs.c
- Patch 2: Extract sysctl interface into mm/hugetlb_sysctl.c
No functional changes are introduced in this series. The code is moved
as-is, with only minor formatting adjustments for code style consistency.
This should make future maintenance and enhancements to the hugetlb
subsystem easier.
Testing: The patch series has been compile-tested and maintains the same
functionality as the original code.
This patch (of 2):
Currently, hugetlb.c contains both core management logic and sysfs
interface implementations, making it difficult to maintain. This patch
extracts the sysfs-related code into a dedicated file to improve code
organization.
The following components are moved to mm/hugetlb_sysfs.c:
- sysfs attribute definitions and handlers
- sysfs kobject management functions
- NUMA per-node hstate attribute registration
Several inline helper functions and macros are moved to
mm/hugetlb_internal.h:
- hstate_is_gigantic_no_runtime()
- next_node_allowed()
- get_valid_node_allowed()
- hstate_next_node_to_alloc()
- hstate_next_node_to_free()
- for_each_node_mask_to_alloc/to_free macros
To support code sharing, these functions are changed from static to
exported symbols:
- remove_hugetlb_folio()
- add_hugetlb_folio()
- init_new_hugetlb_folio()
- prep_and_add_allocated_folios()
- demote_pool_huge_page()
- __nr_hugepages_store_common()
The Makefile is updated to compile hugetlb_sysfs.o when CONFIG_HUGETLBFS
is enabled. This maintains all existing functionality while improving
maintainability by separating concerns.
MAINTAINERS is updated to add new file hugetlb_sysfs.c.
Link: https://lkml.kernel.org/r/cover.1762398359.git.zhuhui@kylinos.cn
Link: https://lkml.kernel.org/r/656a03dff7e2bb20e24e841ede81fdca01d21410.1762398359.git.zhuhui@kylinos.cn
Signed-off-by: Geliang Tang <geliang@kernel.org>
Signed-off-by: Hui Zhu <zhuhui@kylinos.cn>
Cc: David Hildenbrand <david@redhat.com>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "some cleanups for pageout()", v2.
Since we no longer attempt to write back filesystem folios in pageout(),
and only tmpfs/shmem folios and anonymous swapcache folios can be written
back, we can remove the redundant folio_test_private() related logic to
simplify the logic of pageout(), as tmpfs/shmem and swapcache folios do
not use the PG_private flag.
This patch (of 2):
The folio_test_private() check in pageout() was introduced by commit
ce91b575332b ("orphaned pagecache memleak fix") in 2005 (checked from a
history tree[1]). As the commit message mentioned, it was to address the
issue where reiserfs pagecache may be truncated while still pinned. To
further explain, the truncation removes the page->mapping, but the page is
still listed in the VM queues because it still has buffers.
In 2008, commit a2b345642f ("Fix dirty page accounting leak with ext3
data=journal") seems to be dealing with a similar issue, where the page
becomes dirty after truncation, and it provides a very useful call stack:
truncate_complete_page()
cancel_dirty_page() // PG_dirty cleared, decr. dirty pages
do_invalidatepage()
ext3_invalidatepage()
journal_invalidatepage()
journal_unmap_buffer()
__dispose_buffer()
__journal_unfile_buffer()
__journal_temp_unlink_buffer()
mark_buffer_dirty(); // PG_dirty set, incr. dirty pages
In this commit a2b345642f, we forcefully clear the page's dirty flag
during truncation (in truncate_complete_page()).
Now it seems this was just a peculiar usage specific to reiserfs. Maybe
reiserfs had some extra refcount on these pages, which caused them to pass
the is_page_cache_freeable() check.
With the fix provided by commit a2b345642f and reiserfs being removed
in 2024 by commit fb6f20ecb1 ("reiserfs: The last commit"), such a case
is unlikely to occur again. So let's remove the redundant
folio_test_private() checks and related buffer_head release logic, and
just leave a warning here to catch such a bug.
[akpm@linux-foundation.org: redo comment, per David]
Link: https://lkml.kernel.org/r/17d1b293-e393-4989-a357-7eea74b3c805@redhat.com
[baolin.wang@linux.alibaba.com: remove comment and WARNing, per Hugh and others]
Link: https://lkml.kernel.org/r/392a9ca3-31ac-4447-bd44-3c656d63e4ca@linux.alibaba.com
Link: https://lkml.kernel.org/r/cover.1758166683.git.baolin.wang@linux.alibaba.com
Link: https://lkml.kernel.org/r/9ef0e560dc83650bc538eb5dcd1594e112c1369f.1758166683.git.baolin.wang@linux.alibaba.com
Link: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git [1]
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Acked-by: David Hildenbrand <david@redhat.com>
Acked-by: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
damon_sysfs_test_add_targets() is assuming all dynamic memory allocation
in it will succeed. Those are indeed likely in the real use cases since
those allocations are too small to fail, but theoretically those could
fail. In the case, inappropriate memory access can happen. Fix it by
appropriately cleanup pre-allocated memory and skip the execution of the
remaining tests in the failure cases.
Link: https://lkml.kernel.org/r/20251101182021.74868-21-sj@kernel.org
Fixes: b8ee5575f7 ("mm/damon/sysfs-test: add a unit test for damon_sysfs_set_targets()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: <stable@vger.kernel.org> [6.7+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
damon_test_set_filters_default_reject() is assuming all dynamic memory
allocation in it will succeed. Those are indeed likely in the real use
cases since those allocations are too small to fail, but theoretically
those could fail. In the case, inappropriate memory access can happen.
Fix it by appropriately cleanup pre-allocated memory and skip the
execution of the remaining tests in the failure cases.
Link: https://lkml.kernel.org/r/20251101182021.74868-17-sj@kernel.org
Fixes: 094fb14913 ("mm/damon/tests/core-kunit: add a test for damos_set_filters_default_reject()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: <stable@vger.kernel.org> [6.16+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
damon_test_update_monitoring_result() is assuming all dynamic memory
allocation in it will succeed. Those are indeed likely in the real use
cases since those allocations are too small to fail, but theoretically
those could fail. In the case, inappropriate memory access can happen.
Fix it by appropriately cleanup pre-allocated memory and skip the
execution of the remaining tests in the failure cases.
Link: https://lkml.kernel.org/r/20251101182021.74868-12-sj@kernel.org
Fixes: f4c978b659 ("mm/damon/core-test: add a test for damon_update_monitoring_results()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: <stable@vger.kernel.org> [6.3+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
damon_test_ops_registration() is assuming all dynamic memory allocation in
it will succeed. Those are indeed likely in the real use cases since
those allocations are too small to fail, but theoretically those could
fail. In the case, inappropriate memory access can happen. Fix it by
appropriately cleanup pre-allocated memory and skip the execution of the
remaining tests in the failure cases.
Link: https://lkml.kernel.org/r/20251101182021.74868-10-sj@kernel.org
Fixes: 4f540f5ab4 ("mm/damon/core-test: add a kunit test case for ops registration")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: <stable@vger.kernel.org> [5.19+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "mm/damon/tests: fix memory bugs in kunit tests".
DAMON kunit tests were initially written assuming those will be run on
environments that are well controlled and therefore tolerant to transient
test failures and bugs in the test code itself. The user-mode linux based
manual run of the tests is one example of such an environment. And the
test code was written for adding more test coverage as fast as possible,
over making those safe and reliable.
As a result, the tests resulted in having a number of bugs including real
memory leaks, theoretical unhandled memory allocation failures, and unused
memory allocations. The allocation failures that are not handled well are
unlikely in the real world, since those allocations are too small to fail.
But in theory, it can happen and cause inappropriate memory access.
It is arguable if bugs in test code can really harm users. But, anyway
bugs are bugs that need to be fixed. Fix the bugs one by one. Also Cc
stable@ for the fixes of memory leak and unhandled memory allocation
failures. The unused memory allocations are only a matter of memory
efficiency, so not Cc-ing stable@.
The first patch fixes memory leaks in the test code for the DAMON core
layer.
Following fifteen, three, and one patches respectively fix unhandled
memory allocation failures in the test code for DAMON core layer, virtual
address space DAMON operation set, and DAMON sysfs interface, one by one
per test function.
Final two patches remove memory allocations that are correctly deallocated
at the end, but not really being used by any code.
This patch (of 22):
Kunit test function for damos_set_filters_default_reject() allocates two
'struct damos_filter' objects and not deallocates those, so that the
memory for the two objects are leaked for every time the test runs. Fix
this by deallocating those objects at the end of the test code.
Link: https://lkml.kernel.org/r/20251101182021.74868-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20251101182021.74868-2-sj@kernel.org
Fixes: 094fb14913 ("mm/damon/tests/core-kunit: add a test for damos_set_filters_default_reject()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendan.higgins@linux.dev>
Cc: David Gow <davidgow@google.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: <stable@vger.kernel.org> [6.16+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Poison (or ECC) errors can be very common on a large size cluster. The
kernel MM currently does not handle ECC errors / poison on a memory region
that is not backed by struct pages. If a memory region mapped using
remap_pfn_range() for example, but not added to the kernel, MM will not
have associated struct pages. Add a new mechanism to handle memory
failure on such memory.
Make kernel MM expose a function to allow modules managing the device
memory to register the device memory SPA and the address space associated
it. MM maintains this information as an interval tree. On poison, MM can
search for the range that the poisoned PFN belong and use the
address_space to determine the mapping VMA.
In this implementation, kernel MM follows the following sequence that is
largely similar to the memory_failure() handler for struct page backed
memory:
1. memory_failure() is triggered on reception of a poison error. An
absence of struct page is detected and consequently
memory_failure_pfn() is executed.
2. memory_failure_pfn() collects the processes mapped to the PFN.
3. memory_failure_pfn() sends SIGBUS to all the processes mapping the
faulty PFN using kill_procs().
Note that there is one primary difference versus the handling of the
poison on struct pages, which is to skip unmapping to the faulty PFN.
This is done to handle the huge PFNMAP support added recently [1] that
enables VM_PFNMAP vmas to map at PMD or PUD level. A poison to a PFN
mapped in such as way would need breaking the PMD/PUD mapping into PTEs
that will get mirrored into the S2. This can greatly increase the cost of
table walks and have a major performance impact.
Link: https://lore.kernel.org/all/20240826204353.2228736-1-peterx@redhat.com/ [1]
Link: https://lkml.kernel.org/r/20251102184434.2406-3-ankita@nvidia.com
Signed-off-by: Ankit Agrawal <ankita@nvidia.com>
Cc: Aniket Agashe <aniketa@nvidia.com>
Cc: Borislav Betkov <bp@alien8.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Joanthan Cameron <Jonathan.Cameron@huawei.com>
Cc: Kevin Tian <kevin.tian@intel.com>
Cc: Kirti Wankhede <kwankhede@nvidia.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matthew R. Ochs <mochs@nvidia.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Cc: Neo Jia <cjia@nvidia.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Smita Koralahalli Channabasappa <smita.koralahallichannabasappa@amd.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Tarun Gupta <targupta@nvidia.com>
Cc: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
Cc: Vikram Sethi <vsethi@nvidia.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhi Wang <zhiw@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Poison (or ECC) errors can be very common on a large size cluster. The
kernel MM currently handles ECC errors / poison only on memory page backed
by struct page. The handling is currently missing for the PFNMAP memory
that does not have struct pages. The series adds such support.
Implement a new ECC handling for memory without struct pages. Kernel MM
expose registration APIs to allow modules that are managing the device to
register its device memory region. MM then tracks such regions using
interval tree.
The mechanism is largely similar to that of ECC on pfn with struct pages.
If there is an ECC error on a pfn, all the mapping to it are identified
and a SIGBUS is sent to the user space processes owning those mappings.
Note that there is one primary difference versus the handling of the
poison on struct pages, which is to skip unmapping to the faulty PFN.
This is done to handle the huge PFNMAP support added recently [1] that
enables VM_PFNMAP vmas to map at PMD or PUD level. A poison to a PFN
mapped in such as way would need breaking the PMD/PUD mapping into PTEs
that will get mirrored into the S2. This can greatly increase the cost of
table walks and have a major performance impact.
nvgrace-gpu-vfio-pci module maps the device memory to user VA (Qemu) using
remap_pfn_range without being added to the kernel [2]. These device
memory PFNs are not backed by struct page. So make nvgrace-gpu-vfio-pci
module make use of the mechanism to get poison handling support on the
device memory.
This patch (of 3):
The GHES code allows calling of memory_failure() on the PFNs that pass the
pfn_valid() check. This contract is broken for the remapped PFNs which
fails the check and ghes_do_memory_failure() returns without triggering
memory_failure().
Update code to allow memory_failure() call on PFNs failing pfn_valid().
Link: https://lkml.kernel.org/r/20251102184434.2406-1-ankita@nvidia.com
Link: https://lkml.kernel.org/r/20251102184434.2406-2-ankita@nvidia.com
Signed-off-by: Ankit Agrawal <ankita@nvidia.com>
Reviewed-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Aniket Agashe <aniketa@nvidia.com>
Cc: Ankit Agrawal <ankita@nvidia.com>
Cc: Borislav Betkov <bp@alien8.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Joanthan Cameron <Jonathan.Cameron@huawei.com>
Cc: Kevin Tian <kevin.tian@intel.com>
Cc: Kirti Wankhede <kwankhede@nvidia.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matthew R. Ochs <mochs@nvidia.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Cc: Neo Jia <cjia@nvidia.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Smita Koralahalli Channabasappa <smita.koralahallichannabasappa@amd.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Tarun Gupta <targupta@nvidia.com>
Cc: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
Cc: Vikram Sethi <vsethi@nvidia.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhi Wang <zhiw@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Make break_ksm() receive an address range and change break_ksm_pmd_entry()
to perform a range-walk and return the address of the first ksm page
found.
This change allows break_ksm() to skip unmapped regions instead of
iterating every page address. When unmerging large sparse VMAs, this
significantly reduces runtime.
In a benchmark unmerging a 32 TiB sparse virtual address space where only
one page was populated, the runtime dropped from 9 minutes to less then 5
seconds.
Link: https://lkml.kernel.org/r/20251105184912.186329-3-pedrodemargomes@gmail.com
Signed-off-by: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Suggested-by: David Hildenbrand (Red Hat) <david@kernel.org>
Acked-by: David Hildenbrand (Red Hat) <david@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "ksm: perform a range-walk to jump over holes in break_ksm",
v4.
When unmerging an address range, unmerge_ksm_pages function walks every
page address in the specified range to locate ksm pages. This becomes
highly inefficient when scanning large virtual memory areas that contain
mostly unmapped regions, causing the process to get blocked for several
minutes.
This patch makes break_ksm, function called by unmerge_ksm_pages for every
page in an address range, perform a range walk, allowing it to skip over
entire unmapped holes in a VMA, avoiding unnecessary lookups.
As pointed out by David Hildenbrand in [1], unmerge_ksm_pages() is called
from:
* ksm_madvise() through madvise(MADV_UNMERGEABLE). There are not a lot
of users of that function.
* __ksm_del_vma() through ksm_del_vmas(). Effectively called when
disabling KSM for a process either through the sysctl or from s390x gmap
code when enabling storage keys for a VM.
Consider the following test program which creates a 32 TiB mapping in the
virtual address space but only populates a single page:
#include <unistd.h>
#include <stdio.h>
#include <sys/mman.h>
/* 32 TiB */
const size_t size = 32ul * 1024 * 1024 * 1024 * 1024;
int main() {
char *area = mmap(NULL, size, PROT_READ | PROT_WRITE,
MAP_NORESERVE | MAP_PRIVATE | MAP_ANON, -1, 0);
if (area == MAP_FAILED) {
perror("mmap() failed\n");
return -1;
}
/* Populate a single page such that we get an anon_vma. */
*area = 0;
/* Enable KSM. */
madvise(area, size, MADV_MERGEABLE);
madvise(area, size, MADV_UNMERGEABLE);
return 0;
}
Without this patch, this program takes 9 minutes to finish, while with
this patch it finishes in less then 5 seconds.
This patch (of 3):
This reverts commit e317a8d8b4 and changes
function break_ksm_pmd_entry() to use folios.
This reverts break_ksm() to use walk_page_range_vma() instead of
folio_walk_start().
Change break_ksm_pmd_entry() to call is_ksm_zero_pte() only if we know the
folio is present, and also rename variable ret to found. This will make
it easier to later modify break_ksm() to perform a proper range walk.
Link: https://lkml.kernel.org/r/20251105184912.186329-1-pedrodemargomes@gmail.com
Link: https://lkml.kernel.org/r/20251105184912.186329-2-pedrodemargomes@gmail.com
Link: https://lore.kernel.org/linux-mm/e0886fdf-d198-4130-bd9a-be276c59da37@redhat.com/ [1]
Signed-off-by: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Suggested-by: David Hildenbrand (Red Hat) <david@kernel.org>
Acked-by: David Hildenbrand (Red Hat) <david@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The state of a memory block should be restricted to values specified in
the documentation of the memory hotplug API. However, since the state
field in the memory_block struct was defined as an unsigned long, this
restriction was not enforced at compile time.
With the introduction of the enum memory_block_state, it is now possible
to incorporate the desired semantics in the field declaration and enforce
these restrictions at compile time.
[akpm@linux-foundation.org: fix whitespace, per Randy]
Link: https://lkml.kernel.org/r/20251029195617.2210700-3-linux@israelbatista.dev.br
Signed-off-by: Israel Batista <linux@israelbatista.dev.br>
Acked-by: David Hildenbrand <david@redhat.com>
Acked-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Omar Sandoval <osandov@osandov.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "mm: Convert memory block states (MEM_*) macros to enums", v2.
The MEM_* constants indicating the state of a memory block are currently
defined as macros, meaning their definitions will be omitted from the
debuginfo on most kernel builds. This makes it harder for debuggers to
correctly map the block state at runtime, which can be quite useful when
analysing errors related to memory hot plugging and unplugging with tools
such as drgn.
Converting the constants to an enum ensures the correct information is
emitted by the compiler and available for the debugger, without needing to
hard-code them into the debugger and track their changes.
This patch series aims to replace the current macros with a newly created
enum named memory_block_state, while also taking advantage of the compile
time guarantees that we get when using enums.
The first patch does the conversion of the macros to an enum, while the
2nd and 3rd patches use this enum to clean up some type declarations and
make sure that only valid values are used.
This patch (of 3):
Converting the MEM_* constants from macros to an enum ensures that their
values will be correctly emitted in the debug symbols, making it easier to
trace the meaning of each value when debugging with tools such as drgn,
without the need to hard-code the values.
Since the values are mutually exclusive and they are not exposed directly
to userspace, I also dropped the misleading pattern (1<<X) that made it
look like they were combinable flags.
Link: https://lkml.kernel.org/r/20251029195617.2210700-1-linux@israelbatista.dev.br
Link: https://lkml.kernel.org/r/20251029195617.2210700-2-linux@israelbatista.dev.br
Signed-off-by: Israel Batista <linux@israelbatista.dev.br>
Acked-by: David Hildenbrand <david@redhat.com>
Acked-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Omar Sandoval <osandov@osandov.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Swap devices are assumed to have similar accessing speed when swapon if no
priority is specified. It's unfair and doesn't make sense just because
one swap device is swapped on firstly, its priority will be higher than
the one swapped on later.
Here, set all swap devicess to have priority '-1' by default. With this
change, swap device with default priority will be selected round robin
when swapping out. This can improve the swapping efficiency a lot among
multiple swap devices with default priority.
Below are swapon output during the processes when high pressure
vm-scability test is being taken:
1) This is pre-commit a2468cc9bf, swap device is selectd one by one by
priority from high to low when one swap device is exhausted:
------------------------------------
[root@hp-dl385g10-03 ~]# swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 16G 16G -1
/dev/zram1 partition 16G 966.2M -2
/dev/zram2 partition 16G 0B -3
/dev/zram3 partition 16G 0B -4
2) This is behaviour with commit a2468cc9bf, on node, swap device
sharing the same node id is selected firstly until exhausted; while
on node no swap device sharing the node id it selects the one with
highest priority until exhaustd:
------------------------------------
[root@hp-dl385g10-03 ~]# swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 16G 15.7G -2
/dev/zram1 partition 16G 3.4G -3
/dev/zram2 partition 16G 3.4G -4
/dev/zram3 partition 16G 2.6G -5
3) After this patch applied, swap devices with default priority are selectd
round robin:
------------------------------------
[root@hp-dl385g10-03 block]# swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 16G 6.6G -1
/dev/zram1 partition 16G 6.6G -1
/dev/zram2 partition 16G 6.6G -1
/dev/zram3 partition 16G 6.6G -1
With the change, about 18% efficiency promotion relative to node based
way as below. (Surely, the pre-commit a2468cc9bf way is the worst.)
vm-scability test:
==================
Test with:
usemem --init-time -O -y -x -n 31 2G (4G memcg, zram as swap)
one by one: node based: round robin:
System time: 1087.38 s 637.92 s 526.74 s (lower is better)
Sum Throughput: 2036.55 MB/s 3546.56 MB/s 4207.56 MB/s (higher is better)
Single process Throughput: 65.69 MB/s 114.40 MB/s 135.72 MB/s (high is better)
free latency: 15769409.48 us 10138455.99 us 6810119.01 us(lower is better)
Link: https://lkml.kernel.org/r/20251028034308.929550-3-bhe@redhat.com
Signed-off-by: Baoquan He <bhe@redhat.com>
Suggested-by: Chris Li <chrisl@kernel.org>
Acked-by: Chris Li <chrisl@kernel.org>
Acked-by: Nhat Pham <nphamcs@gmail.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Kairui Song <kasong@tencent.com>
Cc: Kemeng Shi <shikemeng@huaweicloud.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "mm/swapfile.c: select swap devices of default priority round
robin", v5.
Currently, on system with multiple swap devices, swap allocation will
select one swap device according to priority. The swap device with the
highest priority will be chosen to allocate firstly.
People can specify a priority from 0 to 32767 when swapon a swap device,
or the system will set it from -2 then downwards by default. Meanwhile,
on NUMA system, the swap device with node_id will be considered first on
that NUMA node of the node_id.
In the current code, an array of plist, swap_avail_heads[nid], is used to
organize swap devices on each NUMA node. For each NUMA node, there is a
plist organizing all swap devices. The 'prio' value in the plist is the
negated value of the device's priority due to plist being sorted from low
to high. The swap device owning one node_id will be promoted to the front
position on that NUMA node, then other swap devices are put in order of
their default priority.
E.g I got a system with 8 NUMA nodes, and I setup 4 zram partition as
swap devices.
Current behaviour:
their priorities will be(note that -1 is skipped):
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 16G 0B -2
/dev/zram1 partition 16G 0B -3
/dev/zram2 partition 16G 0B -4
/dev/zram3 partition 16G 0B -5
And their positions in the 8 swap_avail_lists[nid] will be:
swap_avail_lists[0]: /* node 0's available swap device list */
zram0 -> zram1 -> zram2 -> zram3
prio:1 prio:3 prio:4 prio:5
swap_avali_lists[1]: /* node 1's available swap device list */
zram1 -> zram0 -> zram2 -> zram3
prio:1 prio:2 prio:4 prio:5
swap_avail_lists[2]: /* node 2's available swap device list */
zram2 -> zram0 -> zram1 -> zram3
prio:1 prio:2 prio:3 prio:5
swap_avail_lists[3]: /* node 3's available swap device list */
zram3 -> zram0 -> zram1 -> zram2
prio:1 prio:2 prio:3 prio:4
swap_avail_lists[4-7]: /* node 4,5,6,7's available swap device list */
zram0 -> zram1 -> zram2 -> zram3
prio:2 prio:3 prio:4 prio:5
The adjustment for swap device with node_id intended to decrease the
pressure of lock contention for one swap device by taking different swap
device on different node. The adjustment was introduced in commit
a2468cc9bf ("swap: choose swap device according to numa node").
However, the adjustment is a little coarse-grained. On the node, the swap
device sharing the node's id will always be selected firstly by node's
CPUs until exhausted, then next one. And on other nodes where no swap
device shares its node id, swap device with priority '-2' will be selected
firstly until exhausted, then next with priority '-3'.
This is the swapon output during the process high pressure vm-scability
test is being taken. It's clearly showing zram0 is heavily exploited
until exhausted.
===================================
[root@hp-dl385g10-03 ~]# swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 16G 15.7G -2
/dev/zram1 partition 16G 3.4G -3
/dev/zram2 partition 16G 3.4G -4
/dev/zram3 partition 16G 2.6G -5
The node based strategy on selecting swap device is much better then the
old way one by one selecting swap device. However it is still
unreasonable because swap devices are assumed to have similar accessing
speed if no priority is specified when swapon. It's unfair and doesn't
make sense just because one swap device is swapped on firstly, its
priority will be higher than the one swapped on later.
So in this patchset, change is made to select the swap device round robin
if default priority. In code, the plist array swap_avail_heads[nid] is
replaced with a plist swap_avail_head which reverts commit a2468cc9bf.
Meanwhile, on top of the revert, further change is taken to make any
device w/o specified priority get the same default priority '-1'. Surely,
swap device with specified priority are always put foremost, this is not
impacted. If you care about their different accessing speed, then use
'swapon -p xx' to deploy priority for your swap devices.
New behaviour:
swap_avail_list: /* one global available swap device list */
zram0 -> zram1 -> zram2 -> zram3
prio:1 prio:1 prio:1 prio:1
This is the swapon output during the process high pressure vm-scability
being taken, all is selected round robin:
=======================================
[root@hp-dl385g10-03 linux]# swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 16G 12.6G -1
/dev/zram1 partition 16G 12.6G -1
/dev/zram2 partition 16G 12.6G -1
/dev/zram3 partition 16G 12.6G -1
With the change, we can see about 18% efficiency promotion as below:
vm-scability test:
==================
Test with:
usemem --init-time -O -y -x -n 31 2G (4G memcg, zram as swap)
Before: After:
System time: 637.92 s 526.74 s (lower is better)
Sum Throughput: 3546.56 MB/s 4207.56 MB/s (higher is better)
Single process Throughput: 114.40 MB/s 135.72 MB/s (higher is better)
free latency: 10138455.99 us 6810119.01 us (low is better)
This patch (of 2):
This reverts commit a2468cc9bf ("swap: choose swap device according to
numa node").
After this patch, the behaviour will change back to pre-commit
a2468cc9bf. Means the priority will be set from -1 then downwards by
default, and when swapping, it will exhault swap device one by one
according to priority from high to low. This is preparation work for
later change.
[root@hp-dl385g10-03 ~]# swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 16G 16G -1
/dev/zram1 partition 16G 966.2M -2
/dev/zram2 partition 16G 0B -3
/dev/zram3 partition 16G 0B -4
Link: https://lkml.kernel.org/r/20251028034308.929550-1-bhe@redhat.com
Link: https://lkml.kernel.org/r/20251028034308.929550-2-bhe@redhat.com
Signed-off-by: Baoquan He <bhe@redhat.com>
Suggested-by: Chris Li <chrisl@kernel.org>
Acked-by: Chris Li <chrisl@kernel.org>
Acked-by: Nhat Pham <nphamcs@gmail.com>
Reviewed-by: Kairui Song <kasong@tencent.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Kemeng Shi <shikemeng@huaweicloud.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
