commit 8ce720d5bd upstream.
As reported, ever since commit 1013af4f58 ("mm/hugetlb: fix
huge_pmd_unshare() vs GUP-fast race") we can end up in some situations
where we perform so many IPI broadcasts when unsharing hugetlb PMD page
tables that it severely regresses some workloads.
In particular, when we fork()+exit(), or when we munmap() a large
area backed by many shared PMD tables, we perform one IPI broadcast per
unshared PMD table.
There are two optimizations to be had:
(1) When we process (unshare) multiple such PMD tables, such as during
exit(), it is sufficient to send a single IPI broadcast (as long as
we respect locking rules) instead of one per PMD table.
Locking prevents that any of these PMD tables could get reused before
we drop the lock.
(2) When we are not the last sharer (> 2 users including us), there is
no need to send the IPI broadcast. The shared PMD tables cannot
become exclusive (fully unshared) before an IPI will be broadcasted
by the last sharer.
Concurrent GUP-fast could walk into a PMD table just before we
unshared it. It could then succeed in grabbing a page from the
shared page table even after munmap() etc succeeded (and supressed
an IPI). But there is not difference compared to GUP-fast just
sleeping for a while after grabbing the page and re-enabling IRQs.
Most importantly, GUP-fast will never walk into page tables that are
no-longer shared, because the last sharer will issue an IPI
broadcast.
(if ever required, checking whether the PUD changed in GUP-fast
after grabbing the page like we do in the PTE case could handle
this)
So let's rework PMD sharing TLB flushing + IPI sync to use the mmu_gather
infrastructure so we can implement these optimizations and demystify the
code at least a bit. Extend the mmu_gather infrastructure to be able to
deal with our special hugetlb PMD table sharing implementation.
To make initialization of the mmu_gather easier when working on a single
VMA (in particular, when dealing with hugetlb), provide
tlb_gather_mmu_vma().
We'll consolidate the handling for (full) unsharing of PMD tables in
tlb_unshare_pmd_ptdesc() and tlb_flush_unshared_tables(), and track
in "struct mmu_gather" whether we had (full) unsharing of PMD tables.
Because locking is very special (concurrent unsharing+reuse must be
prevented), we disallow deferring flushing to tlb_finish_mmu() and instead
require an explicit earlier call to tlb_flush_unshared_tables().
From hugetlb code, we call huge_pmd_unshare_flush() where we make sure
that the expected lock protecting us from concurrent unsharing+reuse is
still held.
Check with a VM_WARN_ON_ONCE() in tlb_finish_mmu() that
tlb_flush_unshared_tables() was properly called earlier.
Document it all properly.
Notes about tlb_remove_table_sync_one() interaction with unsharing:
There are two fairly tricky things:
(1) tlb_remove_table_sync_one() is a NOP on architectures without
CONFIG_MMU_GATHER_RCU_TABLE_FREE.
Here, the assumption is that the previous TLB flush would send an
IPI to all relevant CPUs. Careful: some architectures like x86 only
send IPIs to all relevant CPUs when tlb->freed_tables is set.
The relevant architectures should be selecting
MMU_GATHER_RCU_TABLE_FREE, but x86 might not do that in stable
kernels and it might have been problematic before this patch.
Also, the arch flushing behavior (independent of IPIs) is different
when tlb->freed_tables is set. Do we have to enlighten them to also
take care of tlb->unshared_tables? So far we didn't care, so
hopefully we are fine. Of course, we could be setting
tlb->freed_tables as well, but that might then unnecessarily flush
too much, because the semantics of tlb->freed_tables are a bit
fuzzy.
This patch changes nothing in this regard.
(2) tlb_remove_table_sync_one() is not a NOP on architectures with
CONFIG_MMU_GATHER_RCU_TABLE_FREE that actually don't need a sync.
Take x86 as an example: in the common case (!pv, !X86_FEATURE_INVLPGB)
we still issue IPIs during TLB flushes and don't actually need the
second tlb_remove_table_sync_one().
This optimized can be implemented on top of this, by checking e.g., in
tlb_remove_table_sync_one() whether we really need IPIs. But as
described in (1), it really must honor tlb->freed_tables then to
send IPIs to all relevant CPUs.
Notes on TLB flushing changes:
(1) Flushing for non-shared PMD tables
We're converting from flush_hugetlb_tlb_range() to
tlb_remove_huge_tlb_entry(). Given that we properly initialize the
MMU gather in tlb_gather_mmu_vma() to be hugetlb aware, similar to
__unmap_hugepage_range(), that should be fine.
(2) Flushing for shared PMD tables
We're converting from various things (flush_hugetlb_tlb_range(),
tlb_flush_pmd_range(), flush_tlb_range()) to tlb_flush_pmd_range().
tlb_flush_pmd_range() achieves the same that
tlb_remove_huge_tlb_entry() would achieve in these scenarios.
Note that tlb_remove_huge_tlb_entry() also calls
__tlb_remove_tlb_entry(), however that is only implemented on
powerpc, which does not support PMD table sharing.
Similar to (1), tlb_gather_mmu_vma() should make sure that TLB
flushing keeps on working as expected.
Further, note that the ptdesc_pmd_pts_dec() in huge_pmd_share() is not a
concern, as we are holding the i_mmap_lock the whole time, preventing
concurrent unsharing. That ptdesc_pmd_pts_dec() usage will be removed
separately as a cleanup later.
There are plenty more cleanups to be had, but they have to wait until
this is fixed.
[david@kernel.org: fix kerneldoc]
Link: https://lkml.kernel.org/r/f223dd74-331c-412d-93fc-69e360a5006c@kernel.org
Link: https://lkml.kernel.org/r/20251223214037.580860-5-david@kernel.org
Fixes: 1013af4f58 ("mm/hugetlb: fix huge_pmd_unshare() vs GUP-fast race")
Signed-off-by: David Hildenbrand (Red Hat) <david@kernel.org>
Reported-by: "Uschakow, Stanislav" <suschako@amazon.de>
Closes: https://lore.kernel.org/all/4d3878531c76479d9f8ca9789dc6485d@amazon.de/
Tested-by: Laurence Oberman <loberman@redhat.com>
Acked-by: Harry Yoo <harry.yoo@oracle.com>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Lance Yang <lance.yang@linux.dev>
Cc: Liu Shixin <liushixin2@huawei.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Rik van Riel <riel@surriel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3937027cae upstream.
Ever since we stopped using the page count to detect shared PMD page
tables, these comments are outdated.
The only reason we have to flush the TLB early is because once we drop the
i_mmap_rwsem, the previously shared page table could get freed (to then
get reallocated and used for other purpose). So we really have to flush
the TLB before that could happen.
So let's simplify the comments a bit.
The "If we unshared PMDs, the TLB flush was not recorded in mmu_gather."
part introduced as in commit a4a118f2ee ("hugetlbfs: flush TLBs
correctly after huge_pmd_unshare") was confusing: sure it is recorded in
the mmu_gather, otherwise tlb_flush_mmu_tlbonly() wouldn't do anything.
So let's drop that comment while at it as well.
We'll centralize these comments in a single helper as we rework the code
next.
Link: https://lkml.kernel.org/r/20251223214037.580860-3-david@kernel.org
Fixes: 59d9094df3 ("mm: hugetlb: independent PMD page table shared count")
Signed-off-by: David Hildenbrand (Red Hat) <david@kernel.org>
Reviewed-by: Rik van Riel <riel@surriel.com>
Tested-by: Laurence Oberman <loberman@redhat.com>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Acked-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Harry Yoo <harry.yoo@oracle.com>
Cc: Liu Shixin <liushixin2@huawei.com>
Cc: Lance Yang <lance.yang@linux.dev>
Cc: "Uschakow, Stanislav" <suschako@amazon.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4795d205d7 ]
kmsan_free_page() is called by the page allocator's free_pages_prepare()
during page freeing. Its job is to poison all the memory covered by the
page. It can be called with an order-0 page, a compound high-order page
or a non-compound high-order page. But page_size() only works for order-0
and compound pages. For a non-compound high-order page it will
incorrectly return PAGE_SIZE.
The implication is that the tail pages of a high-order non-compound page
do not get poisoned at free, so any invalid access while they are free
could go unnoticed. It looks like the pages will be poisoned again at
allocation time, so that would bookend the window.
Fix this by using the order parameter to calculate the size.
Link: https://lkml.kernel.org/r/20260104134348.3544298-1-ryan.roberts@arm.com
Fixes: b073d7f8ae ("mm: kmsan: maintain KMSAN metadata for page operations")
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Alexander Potapenko <glider@google.com>
Cc: Dmitriy Vyukov <dvyukov@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Marco Elver <elver@google.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Adjust context ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 038a102535 ]
The kernel test robot has reported:
BUG: spinlock trylock failure on UP on CPU#0, kcompactd0/28
lock: 0xffff888807e35ef0, .magic: dead4ead, .owner: kcompactd0/28, .owner_cpu: 0
CPU: 0 UID: 0 PID: 28 Comm: kcompactd0 Not tainted 6.18.0-rc5-00127-ga06157804399 #1 PREEMPT 8cc09ef94dcec767faa911515ce9e609c45db470
Call Trace:
<IRQ>
__dump_stack (lib/dump_stack.c:95)
dump_stack_lvl (lib/dump_stack.c:123)
dump_stack (lib/dump_stack.c:130)
spin_dump (kernel/locking/spinlock_debug.c:71)
do_raw_spin_trylock (kernel/locking/spinlock_debug.c:?)
_raw_spin_trylock (include/linux/spinlock_api_smp.h:89 kernel/locking/spinlock.c:138)
__free_frozen_pages (mm/page_alloc.c:2973)
___free_pages (mm/page_alloc.c:5295)
__free_pages (mm/page_alloc.c:5334)
tlb_remove_table_rcu (include/linux/mm.h:? include/linux/mm.h:3122 include/asm-generic/tlb.h:220 mm/mmu_gather.c:227 mm/mmu_gather.c:290)
? __cfi_tlb_remove_table_rcu (mm/mmu_gather.c:289)
? rcu_core (kernel/rcu/tree.c:?)
rcu_core (include/linux/rcupdate.h:341 kernel/rcu/tree.c:2607 kernel/rcu/tree.c:2861)
rcu_core_si (kernel/rcu/tree.c:2879)
handle_softirqs (arch/x86/include/asm/jump_label.h:36 include/trace/events/irq.h:142 kernel/softirq.c:623)
__irq_exit_rcu (arch/x86/include/asm/jump_label.h:36 kernel/softirq.c:725)
irq_exit_rcu (kernel/softirq.c:741)
sysvec_apic_timer_interrupt (arch/x86/kernel/apic/apic.c:1052)
</IRQ>
<TASK>
RIP: 0010:_raw_spin_unlock_irqrestore (arch/x86/include/asm/preempt.h:95 include/linux/spinlock_api_smp.h:152 kernel/locking/spinlock.c:194)
free_pcppages_bulk (mm/page_alloc.c:1494)
drain_pages_zone (include/linux/spinlock.h:391 mm/page_alloc.c:2632)
__drain_all_pages (mm/page_alloc.c:2731)
drain_all_pages (mm/page_alloc.c:2747)
kcompactd (mm/compaction.c:3115)
kthread (kernel/kthread.c:465)
? __cfi_kcompactd (mm/compaction.c:3166)
? __cfi_kthread (kernel/kthread.c:412)
ret_from_fork (arch/x86/kernel/process.c:164)
? __cfi_kthread (kernel/kthread.c:412)
ret_from_fork_asm (arch/x86/entry/entry_64.S:255)
</TASK>
Matthew has analyzed the report and identified that in drain_page_zone()
we are in a section protected by spin_lock(&pcp->lock) and then get an
interrupt that attempts spin_trylock() on the same lock. The code is
designed to work this way without disabling IRQs and occasionally fail the
trylock with a fallback. However, the SMP=n spinlock implementation
assumes spin_trylock() will always succeed, and thus it's normally a
no-op. Here the enabled lock debugging catches the problem, but otherwise
it could cause a corruption of the pcp structure.
The problem has been introduced by commit 5749077415 ("mm/page_alloc:
leave IRQs enabled for per-cpu page allocations"). The pcp locking scheme
recognizes the need for disabling IRQs to prevent nesting spin_trylock()
sections on SMP=n, but the need to prevent the nesting in spin_lock() has
not been recognized. Fix it by introducing local wrappers that change the
spin_lock() to spin_lock_iqsave() with SMP=n and use them in all places
that do spin_lock(&pcp->lock).
[vbabka@suse.cz: add pcp_ prefix to the spin_lock_irqsave wrappers, per Steven]
Link: https://lkml.kernel.org/r/20260105-fix-pcp-up-v1-1-5579662d2071@suse.cz
Fixes: 5749077415 ("mm/page_alloc: leave IRQs enabled for per-cpu page allocations")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: kernel test robot <oliver.sang@intel.com>
Closes: https://lore.kernel.org/oe-lkp/202512101320.e2f2dd6f-lkp@intel.com
Analyzed-by: Matthew Wilcox <willy@infradead.org>
Link: https://lore.kernel.org/all/aUW05pyc9nZkvY-1@casper.infradead.org/
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Brendan Jackman <jackmanb@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ drop changes to decay_pcp_high() and zone_pcp_update_cacheinfo() ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 269031b15c upstream.
Commit 7ffb791423 ("x86/kaslr: Reduce KASLR entropy on most x86 systems")
is too narrow. The effect being mitigated in that commit is caused by
ZONE_DEVICE which PCI_P2PDMA has a dependency. ZONE_DEVICE, in general,
lets any physical address be added to the direct-map. I.e. not only ACPI
hotplug ranges, CXL Memory Windows, or EFI Specific Purpose Memory, but
also any PCI MMIO range for the DEVICE_PRIVATE and PCI_P2PDMA cases. Update
the mitigation, limit KASLR entropy, to apply in all ZONE_DEVICE=y cases.
Distro kernels typically have PCI_P2PDMA=y, so the practical exposure of
this problem is limited to the PCI_P2PDMA=n case.
A potential path to recover entropy would be to walk ACPI and determine the
limits for hotplug and PCI MMIO before kernel_randomize_memory(). On
smaller systems that could yield some KASLR address bits. This needs
additional investigation to determine if some limited ACPI table scanning
can happen this early without an open coded solution like
arch/x86/boot/compressed/acpi.c needs to deploy.
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Kees Cook <kees@kernel.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: "Liam R. Howlett" <Liam.Howlett@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Fixes: 7ffb791423 ("x86/kaslr: Reduce KASLR entropy on most x86 systems")
Cc: <stable@vger.kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Balbir Singh <balbirs@nvidia.com>
Tested-by: Yasunori Goto <y-goto@fujitsu.com>
Acked-by: Dave Hansen <dave.hansen@linux.intel.com>
Link: http://patch.msgid.link/692e08b2516d4_261c1100a3@dwillia2-mobl4.notmuch
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit f5548c318d upstream.
Currently, scan_get_next_rmap_item() walks every page address in a VMA to
locate mergeable pages. This becomes highly inefficient when scanning
large virtual memory areas that contain mostly unmapped regions, causing
ksmd to use large amount of cpu without deduplicating much pages.
This patch replaces the per-address lookup with a range walk using
walk_page_range(). The range walker allows KSM to skip over entire
unmapped holes in a VMA, avoiding unnecessary lookups. This problem was
previously discussed in [1].
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);
pause();
return 0;
}
$ ./ksm-sparse &
$ echo 1 > /sys/kernel/mm/ksm/run
Without this patch ksmd uses 100% of the cpu for a long time (more then 1
hour in my test machine) scanning all the 32 TiB virtual address space
that contain only one mapped page. This makes ksmd essentially deadlocked
not able to deduplicate anything of value. With this patch ksmd walks
only the one mapped page and skips the rest of the 32 TiB virtual address
space, making the scan fast using little cpu.
Link: https://lkml.kernel.org/r/20251023035841.41406-1-pedrodemargomes@gmail.com
Link: https://lkml.kernel.org/r/20251022153059.22763-1-pedrodemargomes@gmail.com
Link: https://lore.kernel.org/linux-mm/423de7a3-1c62-4e72-8e79-19a6413e420c@redhat.com/ [1]
Fixes: 31dbd01f31 ("ksm: Kernel SamePage Merging")
Signed-off-by: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Co-developed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: craftfever <craftfever@airmail.cc>
Closes: https://lkml.kernel.org/r/020cf8de6e773bb78ba7614ef250129f11a63781@murena.io
Suggested-by: David Hildenbrand <david@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Chengming Zhou <chengming.zhou@linux.dev>
Cc: xu xin <xu.xin16@zte.com.cn>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ change page to folios ]
Signed-off-by: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 590c03ca6a ]
Patch series "ksm: fix exec/fork inheritance", v2.
This series fixes exec/fork inheritance. See the detailed description of
the issue below.
This patch (of 2):
Background
==========
commit d7597f59d1 ("mm: add new api to enable ksm per process")
introduced MMF_VM_MERGE_ANY for mm->flags, and allowed user to set it by
prctl() so that the process's VMAs are forcibly scanned by ksmd.
Subsequently, the 3c6f33b727 ("mm/ksm: support fork/exec for prctl")
supported inheriting the MMF_VM_MERGE_ANY flag when a task calls execve().
Finally, commit 3a9e567ca4 ("mm/ksm: fix ksm exec support for prctl")
fixed the issue that ksmd doesn't scan the mm_struct with MMF_VM_MERGE_ANY
by adding the mm_slot to ksm_mm_head in __bprm_mm_init().
Problem
=======
In some extreme scenarios, however, this inheritance of MMF_VM_MERGE_ANY
during exec/fork can fail. For example, when the scanning frequency of
ksmd is tuned extremely high, a process carrying MMF_VM_MERGE_ANY may
still fail to pass it to the newly exec'd process. This happens because
ksm_execve() is executed too early in the do_execve flow (prematurely
adding the new mm_struct to the ksm_mm_slot list).
As a result, before do_execve completes, ksmd may have already performed a
scan and found that this new mm_struct has no VM_MERGEABLE VMAs, thus
clearing its MMF_VM_MERGE_ANY flag. Consequently, when the new program
executes, the flag MMF_VM_MERGE_ANY inheritance missed.
Root reason
===========
commit d7597f59d1 ("mm: add new api to enable ksm per process") clear
the flag MMF_VM_MERGE_ANY when ksmd found no VM_MERGEABLE VMAs.
Solution
========
Firstly, Don't clear MMF_VM_MERGE_ANY when ksmd found no VM_MERGEABLE
VMAs, because perhaps their mm_struct has just been added to ksm_mm_slot
list, and its process has not yet officially started running or has not
yet performed mmap/brk to allocate anonymous VMAS.
Secondly, recheck MMF_VM_MERGEABLE again if a process takes
MMF_VM_MERGE_ANY, and create a mm_slot and join it into ksm_scan_list
again.
Link: https://lkml.kernel.org/r/20251007182504440BJgK8VXRHh8TD7IGSUIY4@zte.com.cn
Link: https://lkml.kernel.org/r/20251007182821572h_SoFqYZXEP1mvWI4n9VL@zte.com.cn
Fixes: 3c6f33b727 ("mm/ksm: support fork/exec for prctl")
Fixes: d7597f59d1 ("mm: add new api to enable ksm per process")
Signed-off-by: xu xin <xu.xin16@zte.com.cn>
Cc: Stefan Roesch <shr@devkernel.io>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jinjiang Tu <tujinjiang@huawei.com>
Cc: Wang Yaxin <wang.yaxin@zte.com.cn>
Cc: Yang Yang <yang.yang29@zte.com.cn>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ changed mm_flags_test() and mm_flags_clear() calls to test_bit() and clear_bit() ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f2bcc99a5e ]
kmap_atomic() has been deprecated in favor of kmap_local_page().
Therefore, replace kmap_atomic() with kmap_local_page().
kmap_atomic() is implemented like a kmap_local_page() which also disables
page-faults and preemption (the latter only in !PREEMPT_RT kernels). The
kernel virtual addresses returned by these two API are only valid in the
context of the callers (i.e., they cannot be handed to other threads).
With kmap_local_page() the mappings are per thread and CPU local like in
kmap_atomic(); however, they can handle page-faults and can be called from
any context (including interrupts). The tasks that call kmap_local_page()
can be preempted and, when they are scheduled to run again, the kernel
virtual addresses are restored and are still valid.
The code blocks between the mappings and un-mappings don't rely on the
above-mentioned side effects of kmap_atomic(), so that mere replacements
of the old API with the new one is all that they require (i.e., there is
no need to explicitly call pagefault_disable() and/or preempt_disable()).
Link: https://lkml.kernel.org/r/20231120142640.7077-1-fabio.maria.de.francesco@linux.intel.com
Signed-off-by: Fabio M. De Francesco <fabio.maria.de.francesco@linux.intel.com>
Cc: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Stable-dep-of: ec33b59542 ("mm/mempool: fix poisoning order>0 pages with HIGHMEM")
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3cd1548a27 upstream.
In systemd we're trying to switch the internal credentials setup logic
to new mount API [1], and I noticed fsconfig(FSCONFIG_CMD_RECONFIGURE)
consistently fails on tmpfs with noswap option. This can be trivially
reproduced with the following:
```
int fs_fd = fsopen("tmpfs", 0);
fsconfig(fs_fd, FSCONFIG_SET_FLAG, "noswap", NULL, 0);
fsconfig(fs_fd, FSCONFIG_CMD_CREATE, NULL, NULL, 0);
fsmount(fs_fd, 0, 0);
fsconfig(fs_fd, FSCONFIG_CMD_RECONFIGURE, NULL, NULL, 0); <------ EINVAL
```
After some digging the culprit is shmem_reconfigure() rejecting
!(ctx->seen & SHMEM_SEEN_NOSWAP) && sbinfo->noswap, which is bogus
as ctx->seen serves as a mask for whether certain options are touched
at all. On top of that, noswap option doesn't use fsparam_flag_no,
hence it's not really possible to "reenable" swap to begin with.
Drop the check and redundant SHMEM_SEEN_NOSWAP flag.
[1] https://github.com/systemd/systemd/pull/39637
Fixes: 2c6efe9cf2 ("shmem: add support to ignore swap")
Signed-off-by: Mike Yuan <me@yhndnzj.com>
Link: https://patch.msgid.link/20251108190930.440685-1-me@yhndnzj.com
Cc: Luis Chamberlain <mcgrof@kernel.org>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 78ec6f9df6 upstream.
A data-race issue in memcg rstat occurs when two distinct code paths
access the same 4-byte region concurrently. KCSAN detection triggers the
following BUG as a result.
BUG: KCSAN: data-race in __count_memcg_events / mem_cgroup_css_rstat_flush
write to 0xffffe8ffff98e300 of 4 bytes by task 5274 on cpu 17:
mem_cgroup_css_rstat_flush (mm/memcontrol.c:5850)
cgroup_rstat_flush_locked (kernel/cgroup/rstat.c:243 (discriminator 7))
cgroup_rstat_flush (./include/linux/spinlock.h:401 kernel/cgroup/rstat.c:278)
mem_cgroup_flush_stats.part.0 (mm/memcontrol.c:767)
memory_numa_stat_show (mm/memcontrol.c:6911)
<snip>
read to 0xffffe8ffff98e300 of 4 bytes by task 410848 on cpu 27:
__count_memcg_events (mm/memcontrol.c:725 mm/memcontrol.c:962)
count_memcg_event_mm.part.0 (./include/linux/memcontrol.h:1097 ./include/linux/memcontrol.h:1120)
handle_mm_fault (mm/memory.c:5483 mm/memory.c:5622)
<snip>
value changed: 0x00000029 -> 0x00000000
The race occurs because two code paths access the same "stats_updates"
location. Although "stats_updates" is a per-CPU variable, it is remotely
accessed by another CPU at
cgroup_rstat_flush_locked()->mem_cgroup_css_rstat_flush(), leading to the
data race mentioned.
Considering that memcg_rstat_updated() is in the hot code path, adding a
lock to protect it may not be desirable, especially since this variable
pertains solely to statistics.
Therefore, annotating accesses to stats_updates with READ/WRITE_ONCE() can
prevent KCSAN splats and potential partial reads/writes.
Link: https://lkml.kernel.org/r/20240424125940.2410718-1-leitao@debian.org
Fixes: 9cee7e8ef3 ("mm: memcg: optimize parent iteration in memcg_rstat_updated()")
Signed-off-by: Breno Leitao <leitao@debian.org>
Suggested-by: Shakeel Butt <shakeel.butt@linux.dev>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Shakeel Butt <shakeel.butt@linux.dev>
Reviewed-by: Yosry Ahmed <yosryahmed@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9cee7e8ef3 upstream.
In memcg_rstat_updated(), we iterate the memcg being updated and its
parents to update memcg->vmstats_percpu->stats_updates in the fast path
(i.e. no atomic updates). According to my math, this is 3 memory loads
(and potentially 3 cache misses) per memcg:
- Load the address of memcg->vmstats_percpu.
- Load vmstats_percpu->stats_updates (based on some percpu calculation).
- Load the address of the parent memcg.
Avoid most of the cache misses by caching a pointer from each struct
memcg_vmstats_percpu to its parent on the corresponding CPU. In this
case, for the first memcg we have 2 memory loads (same as above):
- Load the address of memcg->vmstats_percpu.
- Load vmstats_percpu->stats_updates (based on some percpu calculation).
Then for each additional memcg, we need a single load to get the
parent's stats_updates directly. This reduces the number of loads from
O(3N) to O(2+N) -- where N is the number of memcgs we need to iterate.
Additionally, stash a pointer to memcg->vmstats in each struct
memcg_vmstats_percpu such that we can access the atomic counter that all
CPUs fold into, memcg->vmstats->stats_updates.
memcg_should_flush_stats() is changed to memcg_vmstats_needs_flush() to
accept a struct memcg_vmstats pointer accordingly.
In struct memcg_vmstats_percpu, make sure both pointers together with
stats_updates live on the same cacheline. Finally, update
mem_cgroup_alloc() to take in a parent pointer and initialize the new
cache pointers on each CPU. The percpu loop in mem_cgroup_alloc() may
look concerning, but there are multiple similar loops in the cgroup
creation path (e.g. cgroup_rstat_init()), most of which are hidden
within alloc_percpu().
According to Oliver's testing [1], this fixes multiple 30-38%
regressions in vm-scalability, will-it-scale-tlb_flush2, and
will-it-scale-fallocate1. This comes at a cost of 2 more pointers per
CPU (<2KB on a machine with 128 CPUs).
[1] https://lore.kernel.org/lkml/ZbDJsfsZt2ITyo61@xsang-OptiPlex-9020/
[yosryahmed@google.com: fix struct memcg_vmstats_percpu size and alignment]
Link: https://lkml.kernel.org/r/20240203044612.1234216-1-yosryahmed@google.com
Link: https://lkml.kernel.org/r/20240124100023.660032-1-yosryahmed@google.com
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Fixes: 8d59d2214c ("mm: memcg: make stats flushing threshold per-memcg")
Tested-by: kernel test robot <oliver.sang@intel.com>
Reported-by: kernel test robot <oliver.sang@intel.com>
Closes: https://lore.kernel.org/oe-lkp/202401221624.cb53a8ca-oliver.sang@intel.com
Acked-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit a530bbc538 upstream.
Commit 3718c02dbd ("acpi, hmat: calculate abstract distance with HMAT")
added a default_dram_perf_ref_source variable that was initialized but
never used. This causes kmemleak to report the following memory leak:
unreferenced object 0xff11000225a47b60 (size 16):
comm "swapper/0", pid 1, jiffies 4294761654
hex dump (first 16 bytes):
41 43 50 49 20 48 4d 41 54 00 c1 4b 7d b7 75 7c ACPI HMAT..K}.u|
backtrace (crc e6d0e7b2):
[<ffffffff95d5afdb>] __kmalloc_node_track_caller_noprof+0x36b/0x440
[<ffffffff95c276d6>] kstrdup+0x36/0x60
[<ffffffff95dfabfa>] mt_set_default_dram_perf+0x23a/0x2c0
[<ffffffff9ad64733>] hmat_init+0x2b3/0x660
[<ffffffff95203cec>] do_one_initcall+0x11c/0x5c0
[<ffffffff9ac9cfc4>] do_initcalls+0x1b4/0x1f0
[<ffffffff9ac9d52e>] kernel_init_freeable+0x4ae/0x520
[<ffffffff97c789cc>] kernel_init+0x1c/0x150
[<ffffffff952aecd1>] ret_from_fork+0x31/0x70
[<ffffffff9520b18a>] ret_from_fork_asm+0x1a/0x30
This reminds us that we forget to use the performance data source
information. So, use the variable in the error log message to help
identify the root cause of inconsistent performance number.
Link: https://lkml.kernel.org/r/87y13mvo0n.fsf@yhuang6-desk2.ccr.corp.intel.com
Fixes: 3718c02dbd ("acpi, hmat: calculate abstract distance with HMAT")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reported-by: Waiman Long <longman@redhat.com>
Acked-by: Waiman Long <longman@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 6f86d0534f upstream.
When a page fault occurs in a secret memory file created with
`memfd_secret(2)`, the kernel will allocate a new folio for it, mark the
underlying page as not-present in the direct map, and add it to the file
mapping.
If two tasks cause a fault in the same page concurrently, both could end
up allocating a folio and removing the page from the direct map, but only
one would succeed in adding the folio to the file mapping. The task that
failed undoes the effects of its attempt by (a) freeing the folio again
and (b) putting the page back into the direct map. However, by doing
these two operations in this order, the page becomes available to the
allocator again before it is placed back in the direct mapping.
If another task attempts to allocate the page between (a) and (b), and the
kernel tries to access it via the direct map, it would result in a
supervisor not-present page fault.
Fix the ordering to restore the direct map before the folio is freed.
Link: https://lkml.kernel.org/r/20251031120955.92116-1-lance.yang@linux.dev
Fixes: 1507f51255 ("mm: introduce memfd_secret system call to create "secret" memory areas")
Signed-off-by: Lance Yang <lance.yang@linux.dev>
Reported-by: Google Big Sleep <big-sleep-vuln-reports@google.com>
Closes: https://lore.kernel.org/linux-mm/CAEXGt5QeDpiHTu3K9tvjUTPqo+d-=wuCNYPa+6sWKrdQJ-ATdg@mail.gmail.com/
Acked-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 74207de2ba upstream.
Patch series "Fix SIGBUS semantics with large folios", v3.
Accessing memory within a VMA, but beyond i_size rounded up to the next
page size, is supposed to generate SIGBUS.
Darrick reported[1] an xfstests regression in v6.18-rc1. generic/749
failed due to missing SIGBUS. This was caused by my recent changes that
try to fault in the whole folio where possible:
19773df031 ("mm/fault: try to map the entire file folio in finish_fault()")
357b92761d ("mm/filemap: map entire large folio faultaround")
These changes did not consider i_size when setting up PTEs, leading to
xfstest breakage.
However, the problem has been present in the kernel for a long time -
since huge tmpfs was introduced in 2016. The kernel happily maps
PMD-sized folios as PMD without checking i_size. And huge=always tmpfs
allocates PMD-size folios on any writes.
I considered this corner case when I implemented a large tmpfs, and my
conclusion was that no one in their right mind should rely on receiving a
SIGBUS signal when accessing beyond i_size. I cannot imagine how it could
be useful for the workload.
But apparently filesystem folks care a lot about preserving strict SIGBUS
semantics.
Generic/749 was introduced last year with reference to POSIX, but no real
workloads were mentioned. It also acknowledged the tmpfs deviation from
the test case.
POSIX indeed says[3]:
References within the address range starting at pa and
continuing for len bytes to whole pages following the end of an
object shall result in delivery of a SIGBUS signal.
The patchset fixes the regression introduced by recent changes as well as
more subtle SIGBUS breakage due to split failure on truncation.
This patch (of 2):
Accesses within VMA, but beyond i_size rounded up to PAGE_SIZE are
supposed to generate SIGBUS.
Recent changes attempted to fault in full folio where possible. They did
not respect i_size, which led to populating PTEs beyond i_size and
breaking SIGBUS semantics.
Darrick reported generic/749 breakage because of this.
However, the problem existed before the recent changes. With huge=always
tmpfs, any write to a file leads to PMD-size allocation. Following the
fault-in of the folio will install PMD mapping regardless of i_size.
Fix filemap_map_pages() and finish_fault() to not install:
- PTEs beyond i_size;
- PMD mappings across i_size;
Make an exception for shmem/tmpfs that for long time intentionally
mapped with PMDs across i_size.
Link: https://lkml.kernel.org/r/20251027115636.82382-1-kirill@shutemov.name
Link: https://lkml.kernel.org/r/20251027115636.82382-2-kirill@shutemov.name
Signed-off-by: Kiryl Shutsemau <kas@kernel.org>
Fixes: 6795801366 ("xfs: Support large folios")
Reported-by: "Darrick J. Wong" <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Kiryl Shutsemau <kas@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 743a2753a0 upstream.
Usually the page cache does not extend beyond the size of the inode,
therefore, no PTEs are created for folios that extend beyond the size.
But with LBS support, we might extend page cache beyond the size of the
inode as we need to guarantee folios of minimum order. While doing a
read, do_fault_around() can create PTEs for pages that lie beyond the
EOF leading to incorrect error return when accessing a page beyond the
mapped file.
Cap the PTE range to be created for the page cache up to the end of
file(EOF) in filemap_map_pages() so that return error codes are consistent
with POSIX[1] for LBS configurations.
generic/749 has been created to trigger this edge case. This also fixes
generic/749 for tmpfs with huge=always on systems with 4k base page size.
[1](from mmap(2)) SIGBUS
Attempted access to a page of the buffer that lies beyond the end
of the mapped file. For an explanation of the treatment of the
bytes in the page that corresponds to the end of a mapped file
that is not a multiple of the page size, see NOTES.
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Pankaj Raghav <p.raghav@samsung.com>
Link: https://lore.kernel.org/r/20240822135018.1931258-6-kernel@pankajraghav.com
Tested-by: David Howells <dhowells@redhat.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Daniel Gomez <da.gomez@samsung.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Kiryl Shutsemau <kas@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 7d7ef0a468 ]
Stats flushing for memcg currently follows the following rules:
- Always flush the entire memcg hierarchy (i.e. flush the root).
- Only one flusher is allowed at a time. If someone else tries to flush
concurrently, they skip and return immediately.
- A periodic flusher flushes all the stats every 2 seconds.
The reason this approach is followed is because all flushes are serialized
by a global rstat spinlock. On the memcg side, flushing is invoked from
userspace reads as well as in-kernel flushers (e.g. reclaim, refault,
etc). This approach aims to avoid serializing all flushers on the global
lock, which can cause a significant performance hit under high
concurrency.
This approach has the following problems:
- Occasionally a userspace read of the stats of a non-root cgroup will
be too expensive as it has to flush the entire hierarchy [1].
- Sometimes the stats accuracy are compromised if there is an ongoing
flush, and we skip and return before the subtree of interest is
actually flushed, yielding stale stats (by up to 2s due to periodic
flushing). This is more visible when reading stats from userspace,
but can also affect in-kernel flushers.
The latter problem is particulary a concern when userspace reads stats
after an event occurs, but gets stats from before the event. Examples:
- When memory usage / pressure spikes, a userspace OOM handler may look
at the stats of different memcgs to select a victim based on various
heuristics (e.g. how much private memory will be freed by killing
this). Reading stale stats from before the usage spike in this case
may cause a wrongful OOM kill.
- A proactive reclaimer may read the stats after writing to
memory.reclaim to measure the success of the reclaim operation. Stale
stats from before reclaim may give a false negative.
- Reading the stats of a parent and a child memcg may be inconsistent
(child larger than parent), if the flush doesn't happen when the
parent is read, but happens when the child is read.
As for in-kernel flushers, they will occasionally get stale stats. No
regressions are currently known from this, but if there are regressions,
they would be very difficult to debug and link to the source of the
problem.
This patch aims to fix these problems by restoring subtree flushing, and
removing the unified/coalesced flushing logic that skips flushing if there
is an ongoing flush. This change would introduce a significant regression
with global stats flushing thresholds. With per-memcg stats flushing
thresholds, this seems to perform really well. The thresholds protect the
underlying lock from unnecessary contention.
This patch was tested in two ways to ensure the latency of flushing is
up to par, on a machine with 384 cpus:
- A synthetic test with 5000 concurrent workers in 500 cgroups doing
allocations and reclaim, as well as 1000 readers for memory.stat
(variation of [2]). No regressions were noticed in the total runtime.
Note that significant regressions in this test are observed with
global stats thresholds, but not with per-memcg thresholds.
- A synthetic stress test for concurrently reading memcg stats while
memory allocation/freeing workers are running in the background,
provided by Wei Xu [3]. With 250k threads reading the stats every
100ms in 50k cgroups, 99.9% of reads take <= 50us. Less than 0.01%
of reads take more than 1ms, and no reads take more than 100ms.
[1] https://lore.kernel.org/lkml/CABWYdi0c6__rh-K7dcM_pkf9BJdTRtAU08M43KO9ME4-dsgfoQ@mail.gmail.com/
[2] https://lore.kernel.org/lkml/CAJD7tka13M-zVZTyQJYL1iUAYvuQ1fcHbCjcOBZcz6POYTV-4g@mail.gmail.com/
[3] https://lore.kernel.org/lkml/CAAPL-u9D2b=iF5Lf_cRnKxUfkiEe0AMDTu6yhrUAzX0b6a6rDg@mail.gmail.com/
[akpm@linux-foundation.org: fix mm/zswap.c]
[yosryahmed@google.com: remove stats flushing mutex]
Link: https://lkml.kernel.org/r/CAJD7tkZgP3m-VVPn+fF_YuvXeQYK=tZZjJHj=dzD=CcSSpp2qg@mail.gmail.com
Link: https://lkml.kernel.org/r/20231129032154.3710765-6-yosryahmed@google.com
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Tested-by: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Ivan Babrou <ivan@cloudflare.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Michal Koutny <mkoutny@suse.com>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Tejun Heo <tj@kernel.org>
Cc: Waiman Long <longman@redhat.com>
Cc: Wei Xu <weixugc@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Leon Huang Fu <leon.huangfu@shopee.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit b006847222 ]
The workingset code flushes the stats in workingset_refault() to get
accurate stats of the eviction memcg. In preparation for more scoped
flushed and passing the eviction memcg to the flush call, move the call to
workingset_test_recent() where we have a pointer to the eviction memcg.
The flush call is sleepable, and cannot be made in an rcu read section.
Hence, minimize the rcu read section by also moving it into
workingset_test_recent(). Furthermore, instead of holding the rcu read
lock throughout workingset_test_recent(), only hold it briefly to get a
ref on the eviction memcg. This allows us to make the flush call after we
get the eviction memcg.
As for workingset_refault(), nothing else there appears to be protected by
rcu. The memcg of the faulted folio (which is not necessarily the same as
the eviction memcg) is protected by the folio lock, which is held from all
callsites. Add a VM_BUG_ON() to make sure this doesn't change from under
us.
No functional change intended.
Link: https://lkml.kernel.org/r/20231129032154.3710765-5-yosryahmed@google.com
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Tested-by: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Ivan Babrou <ivan@cloudflare.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Michal Koutny <mkoutny@suse.com>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Tejun Heo <tj@kernel.org>
Cc: Waiman Long <longman@redhat.com>
Cc: Wei Xu <weixugc@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Leon Huang Fu <leon.huangfu@shopee.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 8d59d2214c ]
A global counter for the magnitude of memcg stats update is maintained on
the memcg side to avoid invoking rstat flushes when the pending updates
are not significant. This avoids unnecessary flushes, which are not very
cheap even if there isn't a lot of stats to flush. It also avoids
unnecessary lock contention on the underlying global rstat lock.
Make this threshold per-memcg. The scheme is followed where percpu (now
also per-memcg) counters are incremented in the update path, and only
propagated to per-memcg atomics when they exceed a certain threshold.
This provides two benefits: (a) On large machines with a lot of memcgs,
the global threshold can be reached relatively fast, so guarding the
underlying lock becomes less effective. Making the threshold per-memcg
avoids this.
(b) Having a global threshold makes it hard to do subtree flushes, as we
cannot reset the global counter except for a full flush. Per-memcg
counters removes this as a blocker from doing subtree flushes, which helps
avoid unnecessary work when the stats of a small subtree are needed.
Nothing is free, of course. This comes at a cost: (a) A new per-cpu
counter per memcg, consuming NR_CPUS * NR_MEMCGS * 4 bytes. The extra
memory usage is insigificant.
(b) More work on the update side, although in the common case it will only
be percpu counter updates. The amount of work scales with the number of
ancestors (i.e. tree depth). This is not a new concept, adding a cgroup
to the rstat tree involves a parent loop, so is charging. Testing results
below show no significant regressions.
(c) The error margin in the stats for the system as a whole increases from
NR_CPUS * MEMCG_CHARGE_BATCH to NR_CPUS * MEMCG_CHARGE_BATCH * NR_MEMCGS.
This is probably fine because we have a similar per-memcg error in charges
coming from percpu stocks, and we have a periodic flusher that makes sure
we always flush all the stats every 2s anyway.
This patch was tested to make sure no significant regressions are
introduced on the update path as follows. The following benchmarks were
ran in a cgroup that is 2 levels deep (/sys/fs/cgroup/a/b/):
(1) Running 22 instances of netperf on a 44 cpu machine with
hyperthreading disabled. All instances are run in a level 2 cgroup, as
well as netserver:
# netserver -6
# netperf -6 -H ::1 -l 60 -t TCP_SENDFILE -- -m 10K
Averaging 20 runs, the numbers are as follows:
Base: 40198.0 mbps
Patched: 38629.7 mbps (-3.9%)
The regression is minimal, especially for 22 instances in the same
cgroup sharing all ancestors (so updating the same atomics).
(2) will-it-scale page_fault tests. These tests (specifically
per_process_ops in page_fault3 test) detected a 25.9% regression before
for a change in the stats update path [1]. These are the
numbers from 10 runs (+ is good) on a machine with 256 cpus:
LABEL | MEAN | MEDIAN | STDDEV |
------------------------------+-------------+-------------+-------------
page_fault1_per_process_ops | | | |
(A) base | 270249.164 | 265437.000 | 13451.836 |
(B) patched | 261368.709 | 255725.000 | 13394.767 |
| -3.29% | -3.66% | |
page_fault1_per_thread_ops | | | |
(A) base | 242111.345 | 239737.000 | 10026.031 |
(B) patched | 237057.109 | 235305.000 | 9769.687 |
| -2.09% | -1.85% | |
page_fault1_scalability | | |
(A) base | 0.034387 | 0.035168 | 0.0018283 |
(B) patched | 0.033988 | 0.034573 | 0.0018056 |
| -1.16% | -1.69% | |
page_fault2_per_process_ops | | |
(A) base | 203561.836 | 203301.000 | 2550.764 |
(B) patched | 197195.945 | 197746.000 | 2264.263 |
| -3.13% | -2.73% | |
page_fault2_per_thread_ops | | |
(A) base | 171046.473 | 170776.000 | 1509.679 |
(B) patched | 166626.327 | 166406.000 | 768.753 |
| -2.58% | -2.56% | |
page_fault2_scalability | | |
(A) base | 0.054026 | 0.053821 | 0.00062121 |
(B) patched | 0.053329 | 0.05306 | 0.00048394 |
| -1.29% | -1.41% | |
page_fault3_per_process_ops | | |
(A) base | 1295807.782 | 1297550.000 | 5907.585 |
(B) patched | 1275579.873 | 1273359.000 | 8759.160 |
| -1.56% | -1.86% | |
page_fault3_per_thread_ops | | |
(A) base | 391234.164 | 390860.000 | 1760.720 |
(B) patched | 377231.273 | 376369.000 | 1874.971 |
| -3.58% | -3.71% | |
page_fault3_scalability | | |
(A) base | 0.60369 | 0.60072 | 0.0083029 |
(B) patched | 0.61733 | 0.61544 | 0.009855 |
| +2.26% | +2.45% | |
All regressions seem to be minimal, and within the normal variance for the
benchmark. The fix for [1] assumes that 3% is noise -- and there were no
further practical complaints), so hopefully this means that such
variations in these microbenchmarks do not reflect on practical workloads.
(3) I also ran stress-ng in a nested cgroup and did not observe any
obvious regressions.
[1]https://lore.kernel.org/all/20190520063534.GB19312@shao2-debian/
Link: https://lkml.kernel.org/r/20231129032154.3710765-4-yosryahmed@google.com
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Tested-by: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Cc: Chris Li <chrisl@kernel.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Ivan Babrou <ivan@cloudflare.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Michal Koutny <mkoutny@suse.com>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Tejun Heo <tj@kernel.org>
Cc: Waiman Long <longman@redhat.com>
Cc: Wei Xu <weixugc@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Leon Huang Fu <leon.huangfu@shopee.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
David Hildenbrand (Red Hat)andGreg Kroah-Hartman
Jane ChuandGreg Kroah-Hartman
Pimyn GirgisandGreg Kroah-Hartman
Ryan RobertsandGreg Kroah-Hartman
Vlastimil BabkaandGreg Kroah-Hartman
SeongJae ParkandGreg Kroah-Hartman
Matthew Wilcox (Oracle)andGreg Kroah-Hartman
Aboorva DevarajanandGreg Kroah-Hartman
Dan WilliamsandGreg Kroah-Hartman
Pedro Demarchi GomesandGreg Kroah-Hartman
David HildenbrandandGreg Kroah-Hartman
Jingbo XuandGreg Kroah-Hartman
xu xinandGreg Kroah-Hartman
Maciej Wieczor-RetmanandGreg Kroah-Hartman
Fabio M. De FrancescoandGreg Kroah-Hartman
Mike YuanandGreg Kroah-Hartman
Breno LeitaoandGreg Kroah-Hartman
Yosry AhmedandGreg Kroah-Hartman
Ying HuangandGreg Kroah-Hartman
Nhat PhamandGreg Kroah-Hartman
Lance YangandGreg Kroah-Hartman
Kiryl ShutsemauandGreg Kroah-Hartman
Pankaj RaghavandGreg Kroah-Hartman