When memory.reclaim was introduced, it became the first case where
cgroup_reclaim() is true for the root cgroup. Johannes concluded [1] that
for most cases this is okay, except for one case. Historically, kswapd
would throttle reclaim on a node if a lot of pages marked for reclaim are
under writeback (aka the node is congested). This occurred by setting
LRUVEC_CONGESTED bit in lruvec->flags. The bit would be cleared when the
node is balanced.
Similarly, cgroup reclaim would set the same bit when an lruvec is
congested, and clear it on the way out of reclaim (to throttle local
reclaimers).
Before the introduction of memory.reclaim, the root memcg was the only
target of kswapd reclaim, and non-root memcgs were the only targets of
cgroup reclaim, so they would never interfere. Using the same bit for
both was fine. After memory.reclaim, it is possible for cgroup reclaim on
the root cgroup to clear the bit set by kswapd. This would result in
reclaim on the node to be unthrottled before the node is balanced.
Fix this by introducing separate bits for cgroup-level and node-level
congestion. kswapd can unthrottle an lruvec that is marked as congested
by cgroup reclaim (as the entire node should no longer be congested), but
not vice versa (to prevent premature unthrottling before the entire node
is balanced).
[1]https://lore.kernel.org/lkml/20230405200150.GA35884@cmpxchg.org/
Link: https://lkml.kernel.org/r/20230621023101.432780-1-yosryahmed@google.com
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Reported-by: Johannes Weiner <hannes@cmpxchg.org>
Closes: https://lore.kernel.org/lkml/20230405200150.GA35884@cmpxchg.org/
Cc: Michal Hocko <mhocko@suse.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
If exclusive loads are enabled for zswap, we invalidate the entry before
returning from zswap_frontswap_load(), after dropping the local reference.
However, the tree lock is dropped during decompression after the local
reference is acquired, so the entry could be invalidated before we drop
the local ref. If this happens, the entry is freed once we drop the local
ref, and zswap_invalidate_entry() tries to invalidate an already freed
entry.
Fix this by:
(a) Making sure zswap_invalidate_entry() is always called with a local
ref held, to avoid being called on a freed entry.
(b) Making sure zswap_invalidate_entry() only drops the ref if the entry
was actually on the rbtree. Otherwise, another invalidation could
have already happened, and the initial ref is already dropped.
With these changes, there is no need to check that there is no need to
make sure the entry still exists in the tree in zswap_reclaim_entry()
before invalidating it, as zswap_reclaim_entry() will make this check
internally.
Link: https://lkml.kernel.org/r/20230621093009.637544-1-yosryahmed@google.com
Fixes: b9c91c4341 ("mm: zswap: support exclusive loads")
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Reported-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Nhat Pham <nphamcs@gmail.com>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Cc: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Remove a few hidden compound_head() calls by converting the returned page
to a folio once and using the folio APIs. We also only increment the
refcount on the folio once instead of once for each page. Ideally, we
would have a for_each_sgt_folio macro, but until then this will do.
Link: https://lkml.kernel.org/r/20230621164557.3510324-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
This wrapper for sg_set_page() lets drivers add folios to a scatterlist
more easily. We could, perhaps, do better by using a different page in
the folio if offset is larger than UINT_MAX, but let's hope we get a
better data structure than this before we need to care about such large
folios.
Link: https://lkml.kernel.org/r/20230621164557.3510324-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Delete a triply out-of-date comment from add_swap_count_continuation():
1. vmalloc_to_page() changed from pte_offset_map() to pte_offset_kernel()
2. pte_offset_map() changed from using kmap_atomic() to kmap_local_page()
3. kmap_atomic() changed from using fixed FIX_KMAP addresses in 2.6.37.
Link: https://lkml.kernel.org/r/9022632b-ba9d-8cb0-c25-4be9786481b5@google.com
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
early_pfn_to_nid() is called frequently in init_reserved_page(), it
returns the node id of the PFN. These PFN are probably from the same
memory region, they have the same node id. It's not necessary to call
early_pfn_to_nid() for each PFN.
Pass nid to reserve_bootmem_region() and drop the call to
early_pfn_to_nid() in init_reserved_page(). Also, set nid on all reserved
pages before doing this, as some reserved memory regions may not be set
nid.
The most beneficial function is memmap_init_reserved_pages() if
CONFIG_DEFERRED_STRUCT_PAGE_INIT is enabled.
The following data was tested on an x86 machine with 190GB of RAM.
before:
memmap_init_reserved_pages() 67ms
after:
memmap_init_reserved_pages() 20ms
Link: https://lkml.kernel.org/r/20230619023406.424298-1-yajun.deng@linux.dev
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Reviewed-by: Mike Rapoport (IBM) <rppt@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
On some machines, the normal zone can have a large memory hole like below
memory layout, and we can see the range from 0x100000000 to 0x1800000000
is a hole. So when isolating some migratable pages, the scanner can meet
the hole and it will take more time to skip the large hole. From my
measurement, I can see the isolation scanner will take 80us ~ 100us to
skip the large hole [0x100000000 - 0x1800000000].
So adding a new helper to fast search next online memory section to skip
the large hole can help to find next suitable pageblock efficiently. With
this patch, I can see the large hole scanning only takes < 1us.
[ 0.000000] Zone ranges:
[ 0.000000] DMA [mem 0x0000000040000000-0x00000000ffffffff]
[ 0.000000] DMA32 empty
[ 0.000000] Normal [mem 0x0000000100000000-0x0000001fa7ffffff]
[ 0.000000] Movable zone start for each node
[ 0.000000] Early memory node ranges
[ 0.000000] node 0: [mem 0x0000000040000000-0x0000000fffffffff]
[ 0.000000] node 0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
[ 0.000000] node 0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
[ 0.000000] node 0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
[ 0.000000] node 0: [mem 0x0000001fa4030000-0x0000001fa40effff]
[ 0.000000] node 0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
[ 0.000000] node 0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
[ 0.000000] node 0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
[ 0.000000] node 0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
[ 0.000000] node 0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]
[baolin.wang@linux.alibaba.com: limit next_ptn to not exceed cc->free_pfn]
Link: https://lkml.kernel.org/r/a1d859c28af0c7e85e91795e7473f553eb180a9d.1686813379.git.baolin.wang@linux.alibaba.com
Link: https://lkml.kernel.org/r/75b4c8ca36bf44ad8c42bf0685ac19d272e426ec.1686705221.git.baolin.wang@linux.alibaba.com
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Suggested-by: David Hildenbrand <david@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Acked-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The recommended ways for getting DAMON monitoring results are using
tried_regions sysfs directory for partial snapshot of the results, and
DAMON tracepoint for full record of the results. However, the
tried_regions sysfs directory usage has not sufficiently updated on some
sections of the DAMON usage document. Update those.
Link: https://lkml.kernel.org/r/20230616191742.87531-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Docs/{mm,admin-guide}damon: update design and usage docs".
Update DAMON design and usage documents for outdated and unnecessarily
duplicated parts.
This patch (of 7):
The 'age' of each region in DAMON monitoring results is an important
concept for both monitoring part and DAMOS. And DAMOS section of the
design document is mentioning it. However, the age itself is not
explained in the document. Add a section for that.
Link: https://lkml.kernel.org/r/20230616191742.87531-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20230616191742.87531-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The mm_struct mm_count field is frequently updated by mmgrab/mmdrop
performed by context switch. This causes false-sharing for surrounding
mm_struct fields which are read-mostly.
This has been observed on a 2sockets/112core/224cpu Intel Sapphire Rapids
server running hackbench, and by the kernel test robot will-it-scale
testcase.
Move the mm_count field into its own cache line to prevent false-sharing
with other mm_struct fields.
Move mm_count to the first field of mm_struct to minimize the amount of
padding required: rather than adding padding before and after the mm_count
field, padding is only added after mm_count.
Note that I noticed this odd comment in mm_struct:
commit 2e3025434a ("mm: relocate 'write_protect_seq' in struct mm_struct")
/*
* With some kernel config, the current mmap_lock's offset
* inside 'mm_struct' is at 0x120, which is very optimal, as
* its two hot fields 'count' and 'owner' sit in 2 different
* cachelines, and when mmap_lock is highly contended, both
* of the 2 fields will be accessed frequently, current layout
* will help to reduce cache bouncing.
*
* So please be careful with adding new fields before
* mmap_lock, which can easily push the 2 fields into one
* cacheline.
*/
struct rw_semaphore mmap_lock;
This comment is rather odd for a few reasons:
- It requires addition/removal of mm_struct fields to carefully consider
field alignment of _other_ fields,
- It expresses the wish to keep an "optimal" alignment for a specific
kernel config.
I suspect that the author of this comment may want to revisit this topic
and perhaps introduce a split-struct approach for struct rw_semaphore,
if the need is to place various fields of this structure in different
cache lines.
Link: https://lkml.kernel.org/r/20230515143536.114960-1-mathieu.desnoyers@efficios.com
Fixes: 223baf9d17 ("sched: Fix performance regression introduced by mm_cid")
Fixes: af7f588d8f ("sched: Introduce per-memory-map concurrency ID")
Link: https://lore.kernel.org/lkml/7a0c1db1-103d-d518-ed96-1584a28fbf32@efficios.com
Reported-by: kernel test robot <yujie.liu@intel.com>
Link: https://lore.kernel.org/oe-lkp/202305151017.27581d75-yujie.liu@intel.com
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Aaron Lu <aaron.lu@intel.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Olivier Dion <odion@efficios.com>
Cc: <michael.christie@oracle.com>
Cc: Feng Tang <feng.tang@intel.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
try_get_folio() takes in a page, then chooses to do some folio operations
based on the flags (either FOLL_GET or FOLL_PIN). We can rewrite this
function to be more purpose oriented.
After calling try_get_folio(), if neither FOLL_GET nor FOLL_PIN are set,
warn and fail. If FOLL_GET is set we can return the result. If FOLL_GET
is not set then FOLL_PIN is set, so we pin the folio.
This change assists with folio conversions, and makes the function more
readable.
Link: https://lkml.kernel.org/r/20230614021312.34085-5-vishal.moola@gmail.com
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
KASAN's boot time kernel parameter 'kasan.fault=' currently supports
'report' and 'panic', which results in either only reporting bugs or also
panicking on reports.
However, some users may wish to have more control over when KASAN reports
result in a kernel panic: in particular, KASAN reported invalid _writes_
are of special interest, because they have greater potential to corrupt
random kernel memory or be more easily exploited.
To panic on invalid writes only, introduce 'kasan.fault=panic_on_write',
which allows users to choose to continue running on invalid reads, but
panic only on invalid writes.
Link: https://lkml.kernel.org/r/20230614095158.1133673-1-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Taras Madan <tarasmadan@google.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
