SeongJae Park [Wed, 2 Aug 2023 21:43:04 +0000 (21:43 +0000)]
Docs/mm/damon/design: update for address range filters
Update DAMON design document's DAMOS filters section for address range
DAMOS filters. Because address range filters are handled by the core
layer and it makes difference in schemes tried regions and schemes
statistics, clearly describe it.
SeongJae Park [Wed, 2 Aug 2023 21:43:03 +0000 (21:43 +0000)]
selftests/damon/sysfs: test address range damos filter
Add a selftest for checking existence of addr_{start,end} files under
DAMOS filter directory, and 'addr' damos filter type input of DAMON sysfs
interface.
SeongJae Park [Wed, 2 Aug 2023 21:43:02 +0000 (21:43 +0000)]
mm/damon/core-test: add a unit test for __damos_filter_out()
Implement a kunit test for the core of address range DAMOS filter
handling, namely __damos_filter_out(). The test especially focus on
regions that overlap with given filter's target address range.
SeongJae Park [Wed, 2 Aug 2023 21:43:01 +0000 (21:43 +0000)]
mm/damon/sysfs-schemes: support address range type DAMOS filter
Extend DAMON sysfs interface to support address range based DAMOS filters,
by adding a special keyword for the filter/<N>/type file, namely 'addr',
and two files under filter/<N>/ for specifying the start and the end
addresses of the range, namely 'addr_start' and 'addr_end'.
SeongJae Park [Wed, 2 Aug 2023 21:43:00 +0000 (21:43 +0000)]
mm/damon/core: introduce address range type damos filter
Patch series "Extend DAMOS filters for address ranges and DAMON monitoring
targets"
There are use cases that need to apply DAMOS schemes to specific address
ranges or DAMON monitoring targets. NUMA nodes in the physical address
space, special memory objects in the virtual address space, and monitoring
target specific efficient monitoring results snapshot retrieval could be
examples of such use cases. This patchset extends DAMOS filters feature
for such cases, by implementing two more filter types, namely address
ranges and DAMON monitoring types.
Patches sequence
----------------
The first seven patches are for the address ranges based DAMOS filter.
The first patch implements the filter feature and expose it via DAMON
kernel API. The second patch further expose the feature to users via
DAMON sysfs interface. The third and fourth patches implement unit tests
and selftests for the feature. Three patches (fifth to seventh) updating
the documents follow.
The following six patches are for the DAMON monitoring target based DAMOS
filter. The eighth patch implements the feature in the core layer and
expose it via DAMON's kernel API. The ninth patch further expose it to
users via DAMON sysfs interface. Tenth patch add a selftest, and two
patches (eleventh and twelfth) update documents.
Users can know special characteristic of specific address ranges. NUMA
nodes or special objects or buffers in virtual address space could be such
examples. For such cases, DAMOS schemes could required to be applied to
only specific address ranges. Implement yet another type of DAMOS filter
for the purpose.
Note that the existing filter types, namely anon pages and memcg DAMOS
filters needed page level type check. Because such check can be done
efficiently in the opertions set layer, those filters are handled in
operations set layer. Specifically, only paddr operations set
implementation supports these filters. Also, because statistics counting
is done in the DAMON core layer, the regions that filtered out by these
filters are counted as tried but failed to the statistics.
Unlike those, address range based filters can efficiently handled in the
core layer. Hence, do the handling in the layer, and count the regions
that filtered out by those as the scheme has not tried for the region.
This difference should clearly documented.
SeongJae Park [Wed, 2 Aug 2023 21:32:18 +0000 (21:32 +0000)]
mm/damon/sysfs: implement a command for updating only schemes tried total bytes
Using tried_regions/total_bytes file, users can efficiently retrieve the
total size of memory regions having specific access pattern. However,
DAMON sysfs interface in kernel still populates all the infomration on the
tried_regions subdirectories. That means the kernel part overhead for the
construction of tried regions directories still exists. To remove the
overhead, implement yet another command input for 'state' DAMON sysfs
file. Writing the input to the file makes DAMON sysfs interface to update
only the total_bytes file.
SeongJae Park [Wed, 2 Aug 2023 21:32:17 +0000 (21:32 +0000)]
mm/damon/sysfs-schemes: implement DAMOS tried total bytes file
Patch series "mm/damon/sysfs-schemes: implement DAMOS tried total bytes
file".
The tried_regions directory of DAMON sysfs interface is useful for
retrieving monitoring results snapshot or DAMOS debugging. However, for
common use case that need to monitor only the total size of the scheme
tried regions (e.g., monitoring working set size), the kernel overhead for
directory construction and user overhead for reading the content could be
high if the number of monitoring region is not small. This patchset
implements DAMON sysfs files for efficient support of the use case.
The first patch implements the sysfs file to reduce the user space
overhead, and the second patch implements a command for reducing the
kernel space overhead.
The third patch adds a selftest for the new file, and following two
patches update documents.
The tried_regions directory can be used for retrieving the monitoring
results snapshot for regions of specific access pattern, by setting the
scheme's action as 'stat' and the access pattern as required. While the
interface provides every detail of the monitoring results, some use cases
including working set size monitoring requires only the total size of the
regions. For such cases, users should read all the information and
calculate the total size of the regions. However, it could incur high
overhead if the number of regions is high. Add a file for retrieving only
the information, namely 'total_bytes' file. It allows users to get the
total size by reading only the file.
Kalesh Singh [Wed, 2 Aug 2023 02:56:04 +0000 (19:56 -0700)]
Multi-gen LRU: fix can_swap in lru_gen_look_around()
walk->can_swap might be invalid since it's not guaranteed to be
initialized for the particular lruvec. Instead deduce it from the folio
type (anon/file).
Kalesh Singh [Wed, 2 Aug 2023 02:56:03 +0000 (19:56 -0700)]
Multi-gen LRU: avoid race in inc_min_seq()
inc_max_seq() will try to inc_min_seq() if nr_gens == MAX_NR_GENS. This
is because the generations are reused (the last oldest now empty
generation will become the next youngest generation).
inc_min_seq() is retried until successful, dropping the lru_lock
and yielding the CPU on each failure, and retaking the lock before
trying again:
while (!inc_min_seq(lruvec, type, can_swap)) {
spin_unlock_irq(&lruvec->lru_lock);
cond_resched();
spin_lock_irq(&lruvec->lru_lock);
}
However, the initial condition that required incrementing the min_seq
(nr_gens == MAX_NR_GENS) is not retested. This can change by another
call to inc_max_seq() from run_aging() with force_scan=true from the
debugfs interface.
Since the eviction stalls when the nr_gens == MIN_NR_GENS, avoid
unnecessarily incrementing the min_seq by rechecking the number of
generations before each attempt.
This issue was uncovered in previous discussion on the list by Yu Zhao
and Aneesh Kumar [1].
Kalesh Singh [Wed, 2 Aug 2023 02:56:02 +0000 (19:56 -0700)]
Multi-gen LRU: fix per-zone reclaim
MGLRU has a LRU list for each zone for each type (anon/file) in each
generation:
long nr_pages[MAX_NR_GENS][ANON_AND_FILE][MAX_NR_ZONES];
The min_seq (oldest generation) can progress independently for each
type but the max_seq (youngest generation) is shared for both anon and
file. This is to maintain a common frame of reference.
In order for eviction to advance the min_seq of a type, all the per-zone
lists in the oldest generation of that type must be empty.
The eviction logic only considers pages from eligible zones for
eviction or promotion.
Consider the system has the movable zone configured and default 4
generations. The current state of the system is as shown below
(only illustrating one type for simplicity):
Type: ANON
Zone DMA32 Normal Movable Device
Gen 0 0 0 4GB 0
Gen 1 0 1GB 1MB 0
Gen 2 1MB 4GB 1MB 0
Gen 3 1MB 1MB 1MB 0
Now consider there is a GFP_KERNEL allocation request (eligible zone
index <= Normal), evict_folios() will return without doing any work
since there are no pages to scan in the eligible zones of the oldest
generation. Reclaim won't make progress until triggered from a ZONE_MOVABLE
allocation request; which may not happen soon if there is a lot of free
memory in the movable zone. This can lead to OOM kills, although there
is 1GB pages in the Normal zone of Gen 1 that we have not yet tried to
reclaim.
This issue is not seen in the conventional active/inactive LRU since
there are no per-zone lists.
If there are no (not enough) folios to scan in the eligible zones, move
folios from ineligible zone (zone_index > reclaim_index) to the next
generation. This allows for the progression of min_seq and reclaiming
from the next generation (Gen 1).
Qualcomm, Mediatek and raspberrypi [1] discovered this issue independently.
Efly Young [Fri, 21 Jul 2023 01:41:16 +0000 (09:41 +0800)]
mm:vmscan: fix inaccurate reclaim during proactive reclaim
Before commit f53af4285d77 ("mm: vmscan: fix extreme overreclaim and swap
floods"), proactive reclaim will extreme overreclaim sometimes. But
proactive reclaim still inaccurate and some extent overreclaim.
Problematic case is easy to construct. Allocate lots of anonymous memory
(e.g., 20G) in a memcg, then swapping by writing memory.recalim and there
is a certain probability of overreclaim. For example, request 1G by
writing memory.reclaim will eventually reclaim 1.7G or other values more
than 1G.
The reason is that reclaimer may have already reclaimed part of requested
memory in one loop, but before adjust sc->nr_to_reclaim in outer loop,
call shrink_lruvec() again will still follow the current sc->nr_to_reclaim
to work. It will eventually lead to overreclaim. In theory, the amount
of reclaimed would be in [request, 2 * request).
Reclaimer usually tends to reclaim more than request. But either direct
or kswapd reclaim have much smaller nr_to_reclaim targets, so it is less
noticeable and not have much impact.
Proactive reclaim can usually come in with a larger value, so the error is
difficult to ignore. Considering proactive reclaim is usually low
frequency, handle the batching into smaller chunks is a better approach.
SeongJae Park [Sat, 29 Jul 2023 20:37:33 +0000 (20:37 +0000)]
mm/damon/core-test: add a test for damos_new_filter()
damos_new_filter() was having a bug that not initializing ->list field of
the returning damos_filter struct, which results in access to
uninitialized memory. Add a unit test for the function.
Since commit bef8620cd8e0 ("mm: memcg: deprecate the non-hierarchical
mode"), use_hierarchy is already deprecated. And it's further removed via
commit 9d9d341df4d5 ("cgroup: remove obsoleted broken_hierarchy and
warned_broken_hierarchy"). Update corresponding comment.
When free_pages is 0, alike_pages is not used. So alike_pages calculation
can be avoided by checking free_pages early to save cpu cycles. Also fix
typo 'comparable'. It should be 'compatible' here.
Add KSM_MERGE_TIME and KSM_MERGE_TIME_HUGE_PAGES tests with
size of 100.
./run_vmtests.sh -t ksm
-----------------------------
running ./ksm_tests -H -s 100
-----------------------------
Number of normal pages: 0
Number of huge pages: 50
Total size: 100 MiB
Total time: 0.399844662 s
Average speed: 250.097 MiB/s
[PASS]
-----------------------------
running ./ksm_tests -P -s 100
-----------------------------
Total size: 100 MiB
Total time: 0.451931496 s
Average speed: 221.272 MiB/s
[PASS]
Matthew Wilcox [Thu, 27 Jul 2023 20:39:44 +0000 (21:39 +0100)]
mm: improve the comment in isolate_migratepages_block()
A recent patch shows that not everybody understands that "stabilise the
mapping" really means "prevent the mapping from being freed", so change
the wording to hopefully make that more clear.
Yang Li [Thu, 27 Jul 2023 01:55:58 +0000 (09:55 +0800)]
mm/memory.c: fix some kernel-doc comments
Add description of @mas and @tree_end, remove @mt in unmap_vmas(). to
silence the warnings:
mm/memory.c:1837: warning: Function parameter or member 'mas' not described in 'unmap_vmas'
mm/memory.c:1837: warning: Function parameter or member 'tree_end' not described in 'unmap_vmas'
mm/memory.c:1837: warning: Excess function parameter 'mt' description in 'unmap_vmas'
Miaohe Lin [Thu, 27 Jul 2023 11:59:34 +0000 (19:59 +0800)]
mm/memcg: fix obsolete function name in mem_cgroup_protection()
Commit 45c7f7e1ef17 ("mm, memcg: decouple e{low,min} state mutations from
protection checks") changed the function name but not the corresponding
comment.
Johannes Weiner [Thu, 27 Jul 2023 16:22:25 +0000 (12:22 -0400)]
mm: zswap: kill zswap_get_swap_cache_page()
The __read_swap_cache_async() interface isn't more difficult to understand
than what the helper abstracts. Save the indirection and a level of
indentation for the primary work of the writeback func.
Johannes Weiner [Thu, 27 Jul 2023 16:22:24 +0000 (12:22 -0400)]
mm: zswap: tighten up entry invalidation
Removing a zswap entry from the tree is tied to an explicit operation
that's supposed to drop the base reference: swap invalidation, exclusive
load, duplicate store. Don't silently remove the entry on final put, but
instead warn if an entry is in tree without reference.
While in that diff context, convert a BUG_ON to a WARN_ON_ONCE. No need
to crash on a refcount underflow.
Kemeng Shi [Tue, 18 Jul 2023 14:58:10 +0000 (22:58 +0800)]
mm/page_ext: add common function to get client data from page_ext
Patch series "add page_ext_data to get client data in page_ext".
Current clients get data from page_ext by adding offset which is auto
generated in page_ext core and exposes the data layout design inside
page_ext core. This series adds a page_ext_data() to hide this from
clients.
Benefits include:
1. Future clients can call page_ext_data directly instead of defining
a new function like get_page_owner to get the data.
2. There is no change to clients if the layout of page_ext data changes.
This patch (of 3):
Add common page_ext_data function to get client data. This could hide
offset which is auto generated in page_ext core and expose the desgin of
page_ext data layout.
Only convert a few easy parts of this function to use the folio passed in;
convert back to struct page for the majority of it. Removes three hidden
calls to compound_head().
Patch series "Followup folio conversions for zswap".
With frontswap killed, it's worth converting the zswap_load() and
zswap_store() functions to take a folio instead of a page pointer. They
aren't converted to support large folios, but there are a lot of
unnecessary calls to compound_head() that are removed by these patches.
This patch (of 4):
Only convert a few easy parts of this function to use the folio passed in;
convert back to struct page for the majority of it. This does remove a
few hidden calls to compound_head().
Yosry Ahmed [Tue, 20 Jun 2023 19:46:44 +0000 (19:46 +0000)]
mm: zswap: multiple zpools support
Support using multiple zpools of the same type in zswap, for concurrency
purposes. A fixed number of 32 zpools is suggested by this commit, which
was determined empirically. It can be later changed or made into a config
option if needed.
On a setup with zswap and zsmalloc, comparing a single zpool to 32 zpools
shows improvements in the zsmalloc lock contention, especially on the swap
out path.
The following shows the perf analysis of the swapout path when 10
workloads are simultaneously reclaiming and refaulting tmpfs pages. There
are some improvements on the swap in path as well, but less significant.
powerpc/book3s64/radix: add support for vmemmap optimization for radix
With 2M PMD-level mapping, we require 32 struct pages and a single vmemmap
page can contain 1024 struct pages (PAGE_SIZE/sizeof(struct page)). Hence
with 64K page size, we don't use vmemmap deduplication for PMD-level
mapping.
powerpc/book3s64/vmemmap: switch radix to use a different vmemmap handling function
This is in preparation to update radix to implement vmemmap optimization
for devdax. Below are the rules w.r.t radix vmemmap mapping
1. First try to map things using PMD (2M)
2. With altmap if altmap cross-boundary check returns true, fall back to
PAGE_SIZE
3. If we can't allocate PMD_SIZE backing memory for vmemmap, fallback to
PAGE_SIZE
On removing vmemmap mapping, check if every subsection that is using the
vmemmap area is invalid. If found to be invalid, that implies we can
safely free the vmemmap area. We don't use the PAGE_UNUSED pattern used
by x86 because with 64K page size, we need to do the above check even at
the PAGE_SIZE granularity.
This is enabled only with radix translation and 1G hugepage size. This
will be used with devdax device memory with a namespace alignment of 1G.
Anon transparent hugepage is not supported even though we do have helpers
checking pud_trans_huge(). We should never find that return true. The
only expected pte bit combination is _PAGE_PTE | _PAGE_DEVMAP.
Some of the helpers are never expected to get called on hash translation
and hence is marked to call BUG() in such a case.
mm/vmemmap optimization: split hugetlb and devdax vmemmap optimization
Arm disabled hugetlb vmemmap optimization [1] because hugetlb vmemmap
optimization includes an update of both the permissions (writeable to
read-only) and the output address (pfn) of the vmemmap ptes. That is not
supported without unmapping of pte(marking it invalid) by some
architectures.
With DAX vmemmap optimization we don't require such pte updates and
architectures can enable DAX vmemmap optimization while having hugetlb
vmemmap optimization disabled. Hence split DAX optimization support into
a different config.
s390, loongarch and riscv don't have devdax support. So the DAX config is
not enabled for them. With this change, arm64 should be able to select
DAX optimization
mm/huge pud: use transparent huge pud helpers only with CONFIG_TRANSPARENT_HUGEPAGE
pudp_set_wrprotect and move_huge_pud helpers are only used when
CONFIG_TRANSPARENT_HUGEPAGE is enabled. Similar to pmdp_set_wrprotect and
move_huge_pmd_helpers use architecture override only if
CONFIG_TRANSPARENT_HUGEPAGE is set
mm/vmemmap: allow architectures to override how vmemmap optimization works
Architectures like powerpc will like to use different page table
allocators and mapping mechanisms to implement vmemmap optimization.
Similar to vmemmap_populate allow architectures to implement
vmemap_populate_compound_pages
mm/vmemmap: improve vmemmap_can_optimize and allow architectures to override
dax vmemmap optimization requires a minimum of 2 PAGE_SIZE area within
vmemmap such that tail page mapping can point to the second PAGE_SIZE
area. Enforce that in vmemmap_can_optimize() function.
Architectures like powerpc also want to enable vmemmap optimization
conditionally (only with radix MMU translation). Hence allow architecture
override.
mm: change pudp_huge_get_and_clear_full take vm_area_struct as arg
We will use this in a later patch to do tlb flush when clearing pud
entries on powerpc. This is similar to commit 93a98695f2f9 ("mm: change
pmdp_huge_get_and_clear_full take vm_area_struct as arg")
mm/hugepage pud: allow arch-specific helper function to check huge page pud support
Patch series "Add support for DAX vmemmap optimization for ppc64", v6.
This patch series implements changes required to support DAX vmemmap
optimization for ppc64. The vmemmap optimization is only enabled with
radix MMU translation and 1GB PUD mapping with 64K page size.
The patch series also splits the hugetlb vmemmap optimization as a
separate Kconfig variable so that architectures can enable DAX vmemmap
optimization without enabling hugetlb vmemmap optimization. This should
enable architectures like arm64 to enable DAX vmemmap optimization while
they can't enable hugetlb vmemmap optimization. More details of the same
are in patch "mm/vmemmap optimization: Split hugetlb and devdax vmemmap
optimization".
With 64K page size for 16384 pages added (1G) we save 14 pages
With 4K page size for 262144 pages added (1G) we save 4094 pages
With 4K page size for 512 pages added (2M) we save 6 pages
This patch (of 13):
Architectures like powerpc would like to enable transparent huge page pud
support only with radix translation. To support that add
has_transparent_pud_hugepage() helper that architectures can override.
mm: handle faults that merely update the accessed bit under the VMA lock
Move FAULT_FLAG_VMA_LOCK check out of handle_pte_fault(). This should
have a significant performance improvement for mmaped files. Write faults
(on read-only shared pages) still take the mmap lock as we do not want to
audit all the implementations of ->pfn_mkwrite() and ->page_mkwrite().
However write-faults on private mappings are handled under the VMA lock.
mm: handle swap and NUMA PTE faults under the VMA lock
Move the FAULT_FLAG_VMA_LOCK check down in handle_pte_fault(). This is
probably not a huge win in its own right, but is a nicely separable bit
from the next patch.
The map_pages fs method should be safe to run under the VMA lock instead
of the mmap lock. This should have a measurable reduction in contention
on the mmap lock.
Push the VMA_LOCK check down from __handle_mm_fault() to
handle_pte_fault(). Once again, we refuse to call ->huge_fault() with the
VMA lock held, but we will wait for a PMD migration entry with the VMA
lock held, handle NUMA migration and set the accessed bit. We were
already doing this for anonymous VMAs, so it should be safe.
Postpone checking the VMA_LOCK flag until we've attempted to handle faults
on PUDs. There's a mild upside to this patch in that we'll allocate the
page tables while under the VMA lock rather than the mmap lock, reducing
the hold time on the mmap lock, since the retry will find the page tables
already populated. The real purpose here is to make a commit that shows
we don't call ->huge_fault under the VMA lock. We do now handle setting
the accessed bit on a PUD fault under the VMA lock, but that doesn't seem
likely to be a measurable difference.
mm: move FAULT_FLAG_VMA_LOCK check from handle_mm_fault()
Handle a little more of the page fault path outside the mmap sem. The
hugetlb path doesn't need to check whether the VMA is anonymous; the
VM_HUGETLB flag is only set on hugetlbfs VMAs. There should be no
performance change from the previous commit; this is simply a step to ease
bisection of any problems.
Remove the TCP layering violation by allowing per-VMA locks on all VMAs.
The fault path will immediately fail in handle_mm_fault(). There may be a
small performance reduction from this patch as a little unnecessary work
will be done on each page fault. See later patches for the improvement.
mm/mmap: change vma iteration order in do_vmi_align_munmap()
By delaying the setting of prev/next VMA until after the write of NULL,
the probability of the prev/next VMA already being in the CPU cache is
significantly increased, especially for larger munmap operations. It
also means that prev/next will be loaded closer to when they are used.
This requires changing the loop type when gathering the VMAs that will
be freed.
Since prev will be set later in the function, it is better to reverse
the splitting direction of the start VMA (modify the new_below argument
to __split_vma).
Using the vma_iter_prev_range() to walk back to the correct location in
the tree will, on the most part, mean walking within the CPU cache.
Usually, this is two steps vs a node reset and a tree re-walk.
mas_prealloc() may walk partially down the tree before finding that a
split or spanning store is needed. When the write occurs, relax the
logic on resetting the walk so that partial walks will not restart, but
walks that have gone too far (a store that affects beyond the current
node) should be restarted.
Relocate it and call mas_wr_extend_null() from within mas_wr_end_piv().
Extending the NULL may affect the end pivot value so call
mas_wr_endtend_null() from within mas_wr_end_piv() to keep it all
together.
Move the definition of vma_iter_clear_gfp() from mmap.c to internal.h so
it can be used in the nommu code. This will reduce node preallocations
in nommu.
maple_tree: adjust node allocation on mas_rebalance()
mas_rebalance() is called to rebalance an insufficient node into a
single node or two sufficient nodes. The preallocation estimate is
always too many in this case as the height of the tree will never grow
and there is no possibility to have a three way split in this case, so
revise the node allocation count.
maple_tree: re-introduce entry to mas_preallocate() arguments
The current preallocation strategy is to preallocate the absolute
worst-case allocation for a tree modification. The entry (or NULL) is
needed to know how many nodes are needed to write to the tree. Start by
adding the argument to the mas_preallocate() definition.
Using vma_iter_set() will reset the tree and cause a re-walk. Use
vmi_iter_config() to set the write to a sub-set of the range. Change
the file case to also use vmi_iter_config() so that the end is correctly
set.
mas_set_range() resets the node to MAS_START, which will cause a re-walk
of the tree to the range. This is unnecessary when the maple state is
already at the correct location of the write. Add a function that only
sets the range to avoid unnecessary re-walking of the tree.
If the prev does not exist, the vma iterator will be set to MAS_NONE,
which will be treated as a MAS_START when the mas_next or mas_find is
used. In this case, the next caller will be the vma iterator, which
uses mas_find() under the hood and will now do what the user expects.
mm: change do_vmi_align_munmap() tracking of VMAs to remove
The majority of the calls to munmap a vm range is within a single vma.
The maple tree is able to store a single entry at 0, with a size of 1 as
a pointer and avoid any allocations. Change do_vmi_align_munmap() to
store the VMAs being munmap()'ed into a tree indexed by the count. This
will leverage the ability to store the first entry without a node
allocation.
Storing the entries into a tree by the count and not the vma start and
end means changing the functions which iterate over the entries. Update
unmap_vmas() and free_pgtables() to take a maple state and a tree end
address to support this functionality.
Passing through the same maple state to unmap_vmas() and free_pgtables()
means the state needs to be reset between calls. This happens in the
static unmap_region() and exit_mmap().
Patch series "Reduce preallocations for maple tree", v3.
Initial work on preallocations showed no regression in performance during
testing, but recently some users (both on [1] and off [android] list) have
reported that preallocating the worst-case number of nodes has caused some
slow down. This patch set addresses the number of allocations in a few
ways.
During munmap() most munmap() operations will remove a single VMA, so
leverage the fact that the maple tree can place a single pointer at range
0 - 0 without allocating. This is done by changing the index of the VMAs
to be indexed by the count, starting at 0.
Re-introduce the entry argument to mas_preallocate() so that a more
intelligent guess of the node count can be made.
Implement the more intelligent guess of the node count, although there is
more work to be done.
During development of v2 of this patch set, I also noticed that the number
of nodes being allocated for a rebalance was beyond what could possibly be
needed. This is addressed in patch 0008.
This patch (of 15):
Add a way to test the speed of mas_for_each() to the testing code.
Despite its name, mm_drop_all_locks() does not drop _all_ locks; the mmap
lock is held write-locked by the caller, and the caller is responsible for
dropping the mmap lock at a later point (which will also release the VMA
locks).
Calling vma_end_write_all() here is dangerous because the caller might
have write-locked a VMA with the expectation that it will stay
write-locked until the mmap_lock is released, as usual.
This _almost_ becomes a problem in the following scenario:
An anonymous VMA A and an SGX VMA B are mapped adjacent to each other.
Userspace calls munmap() on a range starting at the start address of A and
ending in the middle of B.
Hypothetical call graph with additional notes in brackets:
do_vmi_align_munmap
[begin first for_each_vma_range loop]
vma_start_write [on VMA A]
vma_mark_detached [on VMA A]
__split_vma [on VMA B]
sgx_vma_open [== new->vm_ops->open]
sgx_encl_mm_add
__mmu_notifier_register [luckily THIS CAN'T ACTUALLY HAPPEN]
mm_take_all_locks
mm_drop_all_locks
vma_end_write_all [drops VMA lock taken on VMA A before]
vma_start_write [on VMA B]
vma_mark_detached [on VMA B]
[end first for_each_vma_range loop]
vma_iter_clear_gfp [removes VMAs from maple tree]
mmap_write_downgrade
unmap_region
mmap_read_unlock
In this hypothetical scenario, while do_vmi_align_munmap() thinks it still
holds a VMA write lock on VMA A, the VMA write lock has actually been
invalidated inside __split_vma().
The call from sgx_encl_mm_add() to __mmu_notifier_register() can't
actually happen here, as far as I understand, because we are duplicating
an existing SGX VMA, but sgx_encl_mm_add() only calls
__mmu_notifier_register() for the first SGX VMA created in a given
process. So this could only happen in fork(), not on munmap(). But in my
view it is just pure luck that this can't happen.
Also, we wouldn't actually have any bad consequences from this in
do_vmi_align_munmap(), because by the time the bug drops the lock on VMA
A, we've already marked VMA A as detached, which makes it completely
ineligible for any VMA-locked page faults. But again, that's just pure
luck.
So remove the vma_end_write_all(), so that VMA write locks are only ever
released on mmap_write_unlock() or mmap_write_downgrade().
Also add comments to document the locking rules established by this patch.
Nobody checks the PageError()/folio_test_error() for the page/folio in
__end_swap_bio_read/write() and sio_write_complete(). Therefore, we
don't need to set the error flag. Just drop it.
Patch series "Convert several functions in page_io.c to use a folio", v4.
Convert several functions in page_io.c to use a folio, which can remove
several implicit calls to compound_head().
This patch (of 10):
The VM_BUG_ON_FOLIO in swap_readpage() ensures that the page is already
!uptodate in __end_swap_bio_read() and sio_read_complete(). Just remove
unneeded ClearPageUptodate().
Kemeng Shi [Fri, 21 Jul 2023 15:09:56 +0000 (23:09 +0800)]
mm/compaction: correct comment of candidate pfn in fast_isolate_freepages
Patch series "Two minor cleanups for compaction", v2.
This series contains two random cleanups for compaction.
This patch (of 2):
If no preferred one was not found, we will use candidate page with maximum
pfn > min_pfn which is saved in high_pfn. Correct "minimum" to "maximum
candidate" in comment.
Miaohe Lin [Sun, 23 Jul 2023 03:31:14 +0000 (11:31 +0800)]
mm/mprotect: fix obsolete function name in change_pte_range()
Since commit 79a1971c5f14 ("mm: move the copy_one_pte() pte_present check
into the caller"), the explanation of preserving soft-dirtiness is moved
into copy_nonpresent_pte(). Update corresponding comment.
However, of this set, when launched by run_vmtests.sh, some of the
programs are invoked multiple times with different arguments. When
invoked by run_kselftest.sh, they are invoked without arguments (and as
a consequence, some fail immediately).
Some test programs are only launched by run_vmtests.sh:
test_vmalloc.sh
And some test programs and only launched by run_kselftest.sh:
Furthermore, run_vmtests.sh is invoked by run_kselftest.sh, so in this
case all the test programs invoked by both scripts are run twice!
Needless to say, this is a bit of a mess. In the absence of fully
understanding the history here, it looks to me like the best solution is
to launch ALL test programs from run_vmtests.sh, and ONLY invoke
run_vmtests.sh from run_kselftest.sh. This way, we get full control over
the parameters, each program is only invoked the intended number of
times, and regardless of which script is used, the same tests get run in
the same way.
The only drawback is that if using run_kselftest.sh, it's top-level tap
result reporting reports only a single test and it fails if any of the
contained tests fail. I don't see this as a big deal though since we
still see all the nested reporting from multiple layers. The other issue
with this is that all of run_vmtests.sh must execute within a single
kselftest timeout period, so let's increase that to something more
suitable.
In the Makefile, TEST_GEN_PROGS will compile and install the tests and
will add them to the list of tests that run_kselftest.sh will run.
TEST_GEN_FILES will compile and install the tests but will not add them
to the test list. So let's move all the programs from TEST_GEN_PROGS to
TEST_GEN_FILES so that they are built but not executed by
run_kselftest.sh. Note that run_vmtests.sh is added to TEST_PROGS, which
means it ends up in the test list. (the lack of "_GEN" means it won't be
compiled, but simply copied).
projects / linux.git / log