- 30 Nov, 2022 40 commits
-
-
Sergey Senozhatsky authored
Re-phrase writeback BIO error comment. Link: https://lkml.kernel.org/r/20221109115047.2921851-7-senozhatsky@chromium.orgReported-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by:
Minchan Kim <minchan@kernel.org> Cc: Alexey Romanov <avromanov@sberdevices.ru> Cc: Nhat Pham <nphamcs@gmail.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Suleiman Souhlal <suleiman@google.com> Signed-off-by:
-
Sergey Senozhatsky authored
Add a new flag to zram block state that shows if the page was recompressed (using alternative compression algorithm). Link: https://lkml.kernel.org/r/20221109115047.2921851-6-senozhatsky@chromium.orgSigned-off-by:
-
Sergey Senozhatsky authored
Allow zram to recompress (using secondary compression streams) pages. Re-compression algorithms (we support up to 3 at this stage) are selected via recomp_algorithm: echo "algo=zstd priority=1" > /sys/block/zramX/recomp_algorithm Please read documentation for more details. We support several recompression modes: 1) IDLE pages recompression is activated by `idle` mode echo "type=idle" > /sys/block/zram0/recompress 2) Since there may be many idle pages user-space may pass a size threshold value (in bytes) and we will recompress pages only of equal or greater size: echo "threshold=888" > /sys/block/zram0/recompress 3) HUGE pages recompression is activated by `huge` mode echo "type=huge" > /sys/block/zram0/recompress 4) HUGE_IDLE pages recompression is activated by `huge_idle` mode echo "type=huge_idle" > /sys/block/zram0/recompress [senozhatsky@chromium.org: we should always zero out err variable in recompress loop[ Link: https://lkml.kernel.org/r/20221110143423.3250790-1-senozhatsky@chromium.org Link: https://lkml.kernel.org/r/20221109115047.2921851-5-senozhatsky@chromium.orgSigned-off-by:
-
Sergey Senozhatsky authored
We will use non-WB variant in ZRAM page recompression path. Link: https://lkml.kernel.org/r/20221109115047.2921851-4-senozhatsky@chromium.orgSigned-off-by:
-
Sergey Senozhatsky authored
Introduce recomp_algorithm sysfs knob that controls secondary algorithm selection used for recompression. We will support up to 3 secondary compression algorithms which are sorted in order of their priority. To select an algorithm user has to provide its name and priority: echo "algo=zstd priority=1" > /sys/block/zramX/recomp_algorithm echo "algo=deflate priority=2" > /sys/block/zramX/recomp_algorithm During recompression zram iterates through the list of registered secondary algorithms in order of their priorities. We also have a short version for cases when there is only one secondary compression algorithm: echo "algo=zstd" > /sys/block/zramX/recomp_algorithm This will register zstd as the secondary algorithm with priority 1. Link: https://lkml.kernel.org/r/20221109115047.2921851-3-senozhatsky@chromium.orgSigned-off-by:
-
Sergey Senozhatsky authored
Patch series "zram: Support multiple compression streams", v5. This series adds support for multiple compression streams. The main idea is that different compression algorithms have different characteristics and zram may benefit when it uses a combination of algorithms: a default algorithm that is faster but have lower compression rate and a secondary algorithm that can use higher compression rate at a price of slower compression/decompression. There are several use-case for this functionality: - huge pages re-compression: zstd or deflate can successfully compress huge pages (~50% of huge pages on my synthetic ChromeOS tests), IOW pages that lzo was not able to compress. - idle pages re-compression: idle/cold pages sit in the memory and we may reduce zsmalloc memory usage if we recompress those idle pages. Userspace has a number of ways to control the behavior and impact of zram recompression: what type of pages should be recompressed, size watermarks, etc. Please refer to documentation patch. This patch (of 13): The patch turns compression streams and compressor algorithm name struct zram members into arrays, so that we can have multiple compression streams support (in the next patches). The patch uses a rather explicit API for compressor selection: - Get primary (default) compression stream zcomp_stream_get(zram->comps[ZRAM_PRIMARY_COMP]) - Get secondary compression stream zcomp_stream_get(zram->comps[ZRAM_SECONDARY_COMP]) We use similar API for compression streams put(). At this point we always have just one compression stream, since CONFIG_ZRAM_MULTI_COMP is not yet defined. Link: https://lkml.kernel.org/r/20221109115047.2921851-1-senozhatsky@chromium.org Link: https://lkml.kernel.org/r/20221109115047.2921851-2-senozhatsky@chromium.orgSigned-off-by:
-
Alexander Gordeev authored
Flag delayed_rmap of 'struct mmu_gather' is rather a private member, but it is still accessed directly. Instead, let the TLB gather code access the flag. Link: https://lkml.kernel.org/r/Y3SWCu6NRaMQ5dbD@li-4a3a4a4c-28e5-11b2-a85c-a8d192c6f089.ibm.comSigned-off-by:
Alexander Gordeev <agordeev@linux.ibm.com> Acked-by:
Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by:
-
Linus Torvalds authored
When we remove a page table entry, we are very careful to only free the page after we have flushed the TLB, because other CPUs could still be using the page through stale TLB entries until after the flush. However, we have removed the rmap entry for that page early, which means that functions like folio_mkclean() would end up not serializing with the page table lock because the page had already been made invisible to rmap. And that is a problem, because while the TLB entry exists, we could end up with the following situation: (a) one CPU could come in and clean it, never seeing our mapping of the page (b) another CPU could continue to use the stale and dirty TLB entry and continue to write to said page resulting in a page that has been dirtied, but then marked clean again, all while another CPU might have dirtied it some more. End result: possibly lost dirty data. This extends our current TLB gather infrastructure to optionally track a "should I do a delayed page_remove_rmap() for this page after flushing the TLB". It uses the newly introduced 'encoded page pointer' to do that without having to keep separate data around. Note, this is complicated by a couple of issues: - we want to delay the rmap removal, but not past the page table lock, because that simplifies the memcg accounting - only SMP configurations want to delay TLB flushing, since on UP there are obviously no remote TLBs to worry about, and the page table lock means there are no preemption issues either - s390 has its own mmu_gather model that doesn't delay TLB flushing, and as a result also does not want the delayed rmap. As such, we can treat S390 like the UP case and use a common fallback for the "no delays" case. - we can track an enormous number of pages in our mmu_gather structure, with MAX_GATHER_BATCH_COUNT batches of MAX_TABLE_BATCH pages each, all set up to be approximately 10k pending pages. We do not want to have a huge number of batched pages that we then need to check for delayed rmap handling inside the page table lock. Particularly that last point results in a noteworthy detail, where the normal page batch gathering is limited once we have delayed rmaps pending, in such a way that only the last batch (the so-called "active batch") in the mmu_gather structure can have any delayed entries. NOTE! While the "possibly lost dirty data" sounds catastrophic, for this all to happen you need to have a user thread doing either madvise() with MADV_DONTNEED or a full re-mmap() of the area concurrently with another thread continuing to use said mapping. So arguably this is about user space doing crazy things, but from a VM consistency standpoint it's better if we track the dirty bit properly even when user space goes off the rails. [akpm@linux-foundation.org: fix UP build, per Linus] Link: https://lore.kernel.org/all/B88D3073-440A-41C7-95F4-895D3F657EF2@gmail.com/ Link: https://lkml.kernel.org/r/20221109203051.1835763-4-torvalds@linux-foundation.orgSigned-off-by:Johannes Weiner <hannes@cmpxchg.org> Acked-by:
Hugh Dickins <hughd@google.com> Reported-by:
Nadav Amit <nadav.amit@gmail.com> Tested-by:
-
Linus Torvalds authored
This is purely a preparatory patch that makes all the data structures ready for encoding flags with the mmu_gather page pointers. The code currently always sets the flag to zero and doesn't use it yet, but now it's tracking the type state along. The next step will be to actually start using it. Link: https://lkml.kernel.org/r/20221109203051.1835763-3-torvalds@linux-foundation.orgSigned-off-by:
-
Linus Torvalds authored
release_pages() already could take either an array of page pointers, or an array of folio pointers. Expand it to also accept an array of encoded page pointers, which is what both the existing mlock() use and the upcoming mmu_gather use of encoded page pointers wants. Note that release_pages() won't actually use, or react to, any extra encoded bits. Instead, this is very much a case of "I have walked the array of encoded pages and done everything the extra bits tell me to do, now release it all". Also, while the "either page or folio pointers" dual use was handled with a cast of the pointer in "release_folios()", this takes a slightly different approach and uses the "transparent union" attribute to describe the set of arguments to the function: https://gcc.gnu.org/onlinedocs/gcc/Common-Type-Attributes.html which has been supported by gcc forever, but the kernel hasn't used before. That allows us to avoid using various wrappers with casts, and just use the same function regardless of use. Link: https://lkml.kernel.org/r/20221109203051.1835763-2-torvalds@linux-foundation.orgSigned-off-by:
-
Linus Torvalds authored
We already have this notion in parts of the MM code (see the mlock code with the LRU_PAGE and NEW_PAGE bits), but I'm going to introduce a new case, and I refuse to do the same thing we've done before where we just put bits in the raw pointer and say it's still a normal pointer. So this introduces a 'struct encoded_page' pointer that cannot be used for anything else than to encode a real page pointer and a couple of extra bits in the low bits. That way the compiler can trivially track the state of the pointer and you just explicitly encode and decode the extra bits. Note that this makes the alignment of 'struct page' explicit even for the case where CONFIG_HAVE_ALIGNED_STRUCT_PAGE is not set. That is entirely redundant in almost all cases, since the page structure already contains several word-sized entries. However, on m68k, the alignment of even 32-bit data is just 16 bits, and as such in theory the alignment of 'struct page' could be too. So let's just make it very very explicit that the alignment needs to be at least 32 bits, giving us a guarantee of two unused low bits in the pointer. Now, in practice, our page struct array is aligned much more than that anyway, even on m68k, and our existing code in mm/mlock.c obviously already depended on that. But since the whole point of this change is to be careful about the type system when hiding extra bits in the pointer, let's also be explicit about the assumptions we make. NOTE! This is being very careful in another way too: it has a build-time assertion that the 'flags' added to the page pointer actually fit in the two bits. That means that this helper must be inlined, and can only be used in contexts where the compiler can statically determine that the value fits in the available bits. [akpm@linux-foundation.org: kerneldoc on a forward-declared struct confuses htmldocs] Link: https://lore.kernel.org/all/Y2tKixpO4RO6DgW5@tuxmaker.boeblingen.de.ibm.com/ Link: https://lkml.kernel.org/r/20221109203051.1835763-1-torvalds@linux-foundation.orgSigned-off-by:
David Hildenbrand <david@redhat.com> Cc: Alexander Gordeev <agordeev@linux.ibm.com> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Cc: Christian Borntraeger <borntraeger@linux.ibm.com> Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com> Cc: Heiko Carstens <hca@linux.ibm.com> [s390] Cc: Nadav Amit <nadav.amit@gmail.com> Cc: Nicholas Piggin <npiggin@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Sven Schnelle <svens@linux.ibm.com> Cc: Vasily Gorbik <gor@linux.ibm.com> Cc: Will Deacon <will@kernel.org> Signed-off-by:
-
David Hildenbrand authored
Let's extend the test to cover the possible mprotect() optimization when removing write-protection. mprotect() must not allow write-access to a COW-shared page by accident. Link: https://lkml.kernel.org/r/20221108174652.198904-8-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
NUMA hinting no longer uses savedwrite, let's rip it out. ... and while at it, drop __pte_write() and __pmd_write() on ppc64. Link: https://lkml.kernel.org/r/20221108174652.198904-7-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
commit b191f9b1 ("mm: numa: preserve PTE write permissions across a NUMA hinting fault") added remembering write permissions using ordinary pte_write() for PROT_NONE mapped pages to avoid write faults when remapping the page !PROT_NONE on NUMA hinting faults. That commit noted: The patch looks hacky but the alternatives looked worse. The tidest was to rewalk the page tables after a hinting fault but it was more complex than this approach and the performance was worse. It's not generally safe to just mark the page writable during the fault if it's a write fault as it may have been read-only for COW so that approach was discarded. Later, commit 288bc549 ("mm/autonuma: let architecture override how the write bit should be stashed in a protnone pte.") introduced a family of savedwrite PTE functions that didn't necessarily improve the whole situation. One confusing thing is that nowadays, if a page is pte_protnone() and pte_savedwrite() then also pte_write() is true. Another source of confusion is that there is only a single pte_mk_savedwrite() call in the kernel. All other write-protection code seems to silently rely on pte_wrprotect(). Ever since PageAnonExclusive was introduced and we started using it in mprotect context via commit 64fe24a3 ("mm/mprotect: try avoiding write faults for exclusive anonymous pages when changing protection"), we do have machinery in place to avoid write faults when changing protection, which is exactly what we want to do here. Let's similarly do what ordinary mprotect() does nowadays when upgrading write permissions and reuse can_change_pte_writable() and can_change_pmd_writable() to detect if we can upgrade PTE permissions to be writable. For anonymous pages there should be absolutely no change: if an anonymous page is not exclusive, it could not have been mapped writable -- because only exclusive anonymous pages can be mapped writable. However, there *might* be a change for writable shared mappings that require writenotify: if they are not dirty, we cannot map them writable. While it might not matter in practice, we'd need a different way to identify whether writenotify is actually required -- and ordinary mprotect would benefit from that as well. Note that we don't optimize for the actual migration case: (1) When migration succeeds the new PTE will not be writable because the source PTE was not writable (protnone); in the future we might just optimize that case similarly by reusing can_change_pte_writable()/can_change_pmd_writable() when removing migration PTEs. (2) When migration fails, we'd have to recalculate the "writable" flag because we temporarily dropped the PT lock; for now keep it simple and set "writable=false". We'll remove all savedwrite leftovers next. Link: https://lkml.kernel.org/r/20221108174652.198904-6-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
Let's factor the check out into vma_wants_manual_pte_write_upgrade(), to be reused in NUMA hinting fault context soon. Link: https://lkml.kernel.org/r/20221108174652.198904-5-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
Let's replicate what we have for PTEs in can_change_pte_writable() also for PMDs. While this might look like a pure performance improvement, we'll us this to get rid of savedwrite handling in do_huge_pmd_numa_page() next. Place do_huge_pmd_numa_page() strategically good for that purpose. Note that MM_CP_TRY_CHANGE_WRITABLE is currently only set when we come via mprotect_fixup(). Link: https://lkml.kernel.org/r/20221108174652.198904-4-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
We want to replicate this code for handling PMDs soon. (1) No need to crash the kernel, warning and rejecting is good enough. As this will no longer get optimized out, drop the pte_write() check: no harm would be done. (2) Add a comment why PROT_NONE mapped pages are excluded. (3) Add a comment regarding MAP_SHARED handling and why we rely on the dirty bit in the PTE. Link: https://lkml.kernel.org/r/20221108174652.198904-3-david@redhat.comSigned-off-by: -
Nadav Amit authored
Patch series "mm/autonuma: replace savedwrite infrastructure", v2. As discussed in my talk at LPC, we can reuse the same mechanism for deciding whether to map a pte writable when upgrading permissions via mprotect() -- e.g., PROT_READ -> PROT_READ|PROT_WRITE -- to replace the savedwrite infrastructure used for NUMA hinting faults (e.g., PROT_NONE -> PROT_READ|PROT_WRITE). Instead of maintaining previous write permissions for a pte/pmd, we re-determine if the pte/pmd can be writable. The big benefit is that we have a common logic for deciding whether we can map a pte/pmd writable on protection changes. For private mappings, there should be no difference -- from what I understand, that is what autonuma benchmarks care about. I ran autonumabench for v1 on a system with 2 NUMA nodes, 96 GiB each via: perf stat --null --repeat 10 The numa01 benchmark is quite noisy in my environment and I failed to reduce the noise so far. numa01: mm-unstable: 146.88 +- 6.54 seconds time elapsed ( +- 4.45% ) mm-unstable++: 147.45 +- 13.39 seconds time elapsed ( +- 9.08% ) numa02: mm-unstable: 16.0300 +- 0.0624 seconds time elapsed ( +- 0.39% ) mm-unstable++: 16.1281 +- 0.0945 seconds time elapsed ( +- 0.59% ) It is worth noting that for shared writable mappings that require writenotify, we will only avoid write faults if the pte/pmd is dirty (inherited from the older mprotect logic). If we ever care about optimizing that further, we'd need a different mechanism to identify whether the FS still needs to get notified on the next write access. In any case, such an optimization will then not be autonuma-specific, but mprotect() permission upgrades would similarly benefit from it. This patch (of 7): Anonymous pages might have the dirty bit clear, but this should not prevent mprotect from making them writable if they are exclusive. Therefore, skip the test whether the page is dirty in this case. Note that there are already other ways to get a writable PTE mapping an anonymous page that is clean: for example, via MADV_FREE. In an ideal world, we'd have a different indication from the FS whether writenotify is still required. [david@redhat.com: return directly; update description] Link: https://lkml.kernel.org/r/20221108174652.198904-1-david@redhat.com Link: https://lkml.kernel.org/r/20221108174652.198904-2-david@redhat.comSigned-off-by:
Nadav Amit <namit@vmware.com> Signed-off-by:
-
Rong Tao authored
page_owner_sort was introduced since commit 48c96a36 ("mm/page_owner: keep track of page owners"), and we should ignore it. Link: https://lkml.kernel.org/r/tencent_F6CAC0ABE16839E2B2419BD07316DA65BB06@qq.comSigned-off-by:
Rong Tao <rongtao@cestc.cn> Signed-off-by:
-
Hugh Dickins authored
It's hard to add a page_add_anon_rmap() into __split_huge_pmd_locked()'s HPAGE_PMD_NR set_pte_at() loop, without wincing at the "freeze" case's HPAGE_PMD_NR page_remove_rmap() loop below it. It's just a mistake to add rmaps in the "freeze" (insert migration entries prior to splitting huge page) case: the pmd_migration case already avoids doing that, so just follow its lead. page_add_ref() versus put_page() likewise. But why is one more put_page() needed in the "freeze" case? Because it's removing the pmd rmap, already removed when pmd_migration (and freeze and pmd_migration are mutually exclusive cases). Link: https://lkml.kernel.org/r/d43748aa-fece-e0b9-c4ab-f23c9ebc9011@google.comSigned-off-by:
Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Dan Carpenter <error27@gmail.com> Cc: David Hildenbrand <david@redhat.com> Cc: James Houghton <jthoughton@google.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: John Hubbard <jhubbard@nvidia.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Miaohe Lin <linmiaohe@huawei.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Cc: Mina Almasry <almasrymina@google.com> Cc: Muchun Song <songmuchun@bytedance.com> Cc: Naoya Horiguchi <naoya.horiguchi@linux.dev> Cc: Peter Xu <peterx@redhat.com> Cc: Sidhartha Kumar <sidhartha.kumar@oracle.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Yang Shi <shy828301@gmail.com> Cc: Yu Zhao <yuzhao@google.com> Cc: Zach O'Keefe <zokeefe@google.com> Signed-off-by:
-
Hugh Dickins authored
Can the lock_compound_mapcount() bit_spin_lock apparatus be removed now? Yes. Not by atomic64_t or cmpxchg games, those get difficult on 32-bit; but if we slightly abuse subpages_mapcount by additionally demanding that one bit be set there when the compound page is PMD-mapped, then a cascade of two atomic ops is able to maintain the stats without bit_spin_lock. This is harder to reason about than when bit_spin_locked, but I believe safe; and no drift in stats detected when testing. When there are racing removes and adds, of course the sequence of operations is less well- defined; but each operation on subpages_mapcount is atomically good. What might be disastrous, is if subpages_mapcount could ever fleetingly appear negative: but the pte lock (or pmd lock) these rmap functions are called under, ensures that a last remove cannot race ahead of a first add. Continue to make an exception for hugetlb (PageHuge) pages, though that exception can be easily removed by a further commit if necessary: leave subpages_mapcount 0, don't bother with COMPOUND_MAPPED in its case, just carry on checking compound_mapcount too in folio_mapped(), page_mapped(). Evidence is that this way goes slightly faster than the previous implementation in all cases (pmds after ptes now taking around 103ms); and relieves us of worrying about contention on the bit_spin_lock. Link: https://lkml.kernel.org/r/3978f3ca-5473-55a7-4e14-efea5968d892@google.comSigned-off-by:
-
Hugh Dickins authored
Patch series "mm,thp,rmap: rework the use of subpages_mapcount", v2. This patch (of 3): Following suggestion from Linus, instead of counting every PTE map of a compound page in subpages_mapcount, just count how many of its subpages are PTE-mapped: this yields the exact number needed for NR_ANON_MAPPED and NR_FILE_MAPPED stats, without any need for a locked scan of subpages; and requires updating the count less often. This does then revert total_mapcount() and folio_mapcount() to needing a scan of subpages; but they are inherently racy, and need no locking, so Linus is right that the scans are much better done there. Plus (unlike in 6.1 and previous) subpages_mapcount lets us avoid the scan in the common case of no PTE maps. And page_mapped() and folio_mapped() remain scanless and just as efficient with the new meaning of subpages_mapcount: those are the functions which I most wanted to remove the scan from. The updated page_dup_compound_rmap() is no longer suitable for use by anon THP's __split_huge_pmd_locked(); but page_add_anon_rmap() can be used for that, so long as its VM_BUG_ON_PAGE(!PageLocked) is deleted. Evidence is that this way goes slightly faster than the previous implementation for most cases; but significantly faster in the (now scanless) pmds after ptes case, which started out at 870ms and was brought down to 495ms by the previous series, now takes around 105ms. Link: https://lkml.kernel.org/r/a5849eca-22f1-3517-bf29-95d982242742@google.com Link: https://lkml.kernel.org/r/eec17e16-4e1-7c59-f1bc-5bca90dac919@google.comSigned-off-by:
-
Joao Martins authored
Today with `hugetlb_free_vmemmap=on` the struct page memory that is freed back to page allocator is as following: for a 2M hugetlb page it will reuse the first 4K vmemmap page to remap the remaining 7 vmemmap pages, and for a 1G hugetlb it will remap the remaining 4095 vmemmap pages. Essentially, that means that it breaks the first 4K of a potentially contiguous chunk of memory of 32K (for 2M hugetlb pages) or 16M (for 1G hugetlb pages). For this reason the memory that it's free back to page allocator cannot be used for hugetlb to allocate huge pages of the same size, but rather only of a smaller huge page size: Trying to assign a 64G node to hugetlb (on a 128G 2node guest, each node having 64G): * Before allocation: Free pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10 ... Node 0, zone Normal, type Movable 340 100 32 15 1 2 0 0 0 1 15558 $ echo 32768 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages $ cat /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages 31987 * After: Node 0, zone Normal, type Movable 30893 32006 31515 7 0 0 0 0 0 0 0 Notice how the memory freed back are put back into 4K / 8K / 16K page pools. And it allocates a total of 31987 pages (63974M). To fix this behaviour rather than remapping second vmemmap page (thus breaking the contiguous block of memory backing the struct pages) repopulate the first vmemmap page with a new one. We allocate and copy from the currently mapped vmemmap page, and then remap it later on. The same algorithm works if there's a pre initialized walk::reuse_page and the head page doesn't need to be skipped and instead we remap it when the @addr being changed is the @reuse_addr. The new head page is allocated in vmemmap_remap_free() given that on restore there's no need for functional change. Note that, because right now one hugepage is remapped at a time, thus only one free 4K page at a time is needed to remap the head page. Should it fail to allocate said new page, it reuses the one that's already mapped just like before. As a result, for every 64G of contiguous hugepages it can give back 1G more of contiguous memory per 64G, while needing in total 128M new 4K pages (for 2M hugetlb) or 256k (for 1G hugetlb). After the changes, try to assign a 64G node to hugetlb (on a 128G 2node guest, each node with 64G): * Before allocation Free pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10 ... Node 0, zone Normal, type Movable 1 1 1 0 0 1 0 0 1 1 15564 $ echo 32768 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages $ cat /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages 32394 * After: Node 0, zone Normal, type Movable 0 50 97 108 96 81 70 46 18 0 0 In the example above, 407 more hugeltb 2M pages are allocated i.e. 814M out of the 32394 (64788M) allocated. So the memory freed back is indeed being used back in hugetlb and there's no massive order-0..order-2 pages accumulated unused. [joao.m.martins@oracle.com: v3] Link: https://lkml.kernel.org/r/20221109200623.96867-1-joao.m.martins@oracle.com [joao.m.martins@oracle.com: add smp_wmb() to ensure page contents are visible prior to PTE write] Link: https://lkml.kernel.org/r/20221110121214.6297-1-joao.m.martins@oracle.com Link: https://lkml.kernel.org/r/20221107153922.77094-1-joao.m.martins@oracle.comSigned-off-by:
Joao Martins <joao.m.martins@oracle.com> Reviewed-by:
Muchun Song <songmuchun@bytedance.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Signed-off-by:
-
SeongJae Park authored
There was a bug[1] that triggered by writing non-context DAMON debugfs file names to the 'rm_contexts' DAMON debugfs file. Add a selftest for the bug to avoid it happen again. [1] https://lore.kernel.org/damon/000000000000ede3ac05ec4abf8e@google.com/ Link: https://lkml.kernel.org/r/20221107165001.5717-3-sj@kernel.orgSigned-off-by:
SeongJae Park <sj@kernel.org> Signed-off-by:
-
Hugh Dickins authored
Commit ("mm,thp,rmap: lock_compound_mapcounts() on THP mapcounts") propagated the "if (compound) {lock} else if (PageCompound) {lock} else {atomic}" pattern throughout; but Linus hated the way that gives primacy to the uncommon case: switch to "if (!PageCompound) {atomic} else if (compound) {lock} else {lock}" throughout. Linus has a bigger idea for how to improve it all, but here just make that rearrangement. Link: https://lkml.kernel.org/r/fca2f694-2098-b0ef-d4e-f1d8b94d318c@google.comSigned-off-by: -
Hugh Dickins authored
Fix the races in maintaining compound_mapcount, subpages_mapcount and subpage _mapcount by using PG_locked in the first tail of any compound page for a bit_spin_lock() on such modifications; skipping the usual atomic operations on those fields in this case. Bring page_remove_file_rmap() and page_remove_anon_compound_rmap() back into page_remove_rmap() itself. Rearrange page_add_anon_rmap() and page_add_file_rmap() and page_remove_rmap() to follow the same "if (compound) {lock} else if (PageCompound) {lock} else {atomic}" pattern (with a PageTransHuge in the compound test, like before, to avoid BUG_ONs and optimize away that block when THP is not configured). Move all the stats updates outside, after the bit_spin_locked section, so that it is sure to be a leaf lock. Add page_dup_compound_rmap() to manage compound locking versus atomics in sync with the rest. In particular, hugetlb pages are still using the atomics: to avoid unnecessary interference there, and because they never have subpage mappings; but this exception can easily be changed. Conveniently, page_dup_compound_rmap() turns out to suit an anon THP's __split_huge_pmd_locked() too. bit_spin_lock() is not popular with PREEMPT_RT folks: but PREEMPT_RT sensibly excludes TRANSPARENT_HUGEPAGE already, so its only exposure is to the non-hugetlb non-THP pte-mapped compound pages (with large folios being currently dependent on TRANSPARENT_HUGEPAGE). There is never any scan of subpages in this case; but we have chosen to use PageCompound tests rather than PageTransCompound tests to gate the use of lock_compound_mapcounts(), so that page_mapped() is correct on all compound pages, whether or not TRANSPARENT_HUGEPAGE is enabled: could that be a problem for PREEMPT_RT, when there is contention on the lock - under heavy concurrent forking for example? If so, then it can be turned into a sleeping lock (like folio_lock()) when PREEMPT_RT. A simple 100 X munmap(mmap(2GB, MAP_SHARED|MAP_POPULATE, tmpfs), 2GB) took 18 seconds on small pages, and used to take 1 second on huge pages, but now takes 115 milliseconds on huge pages. Mapping by pmds a second time used to take 860ms and now takes 86ms; mapping by pmds after mapping by ptes (when the scan is needed) used to take 870ms and now takes 495ms. Mapping huge pages by ptes is largely unaffected but variable: between 5% faster and 5% slower in what I've recorded. Contention on the lock is likely to behave worse than contention on the atomics behaved. Link: https://lkml.kernel.org/r/1b42bd1a-8223-e827-602f-d466c2db7d3c@google.comSigned-off-by: -
Hugh Dickins authored
Compound page (folio) mapcount calculations have been different for anon and file (or shmem) THPs, and involved the obscure PageDoubleMap flag. And each huge mapping and unmapping of a file (or shmem) THP involved atomically incrementing and decrementing the mapcount of every subpage of that huge page, dirtying many struct page cachelines. Add subpages_mapcount field to the struct folio and first tail page, so that the total of subpage mapcounts is available in one place near the head: then page_mapcount() and total_mapcount() and page_mapped(), and their folio equivalents, are so quick that anon and file and hugetlb don't need to be optimized differently. Delete the unloved PageDoubleMap. page_add and page_remove rmap functions must now maintain the subpages_mapcount as well as the subpage _mapcount, when dealing with pte mappings of huge pages; and correct maintenance of NR_ANON_MAPPED and NR_FILE_MAPPED statistics still needs reading through the subpages, using nr_subpages_unmapped() - but only when first or last pmd mapping finds subpages_mapcount raised (double-map case, not the common case). But are those counts (used to decide when to split an anon THP, and in vmscan's pagecache_reclaimable heuristic) correctly maintained? Not quite: since page_remove_rmap() (and also split_huge_pmd()) is often called without page lock, there can be races when a subpage pte mapcount 0<->1 while compound pmd mapcount 0<->1 is scanning - races which the previous implementation had prevented. The statistics might become inaccurate, and even drift down until they underflow through 0. That is not good enough, but is better dealt with in a followup patch. Update a few comments on first and second tail page overlaid fields. hugepage_add_new_anon_rmap() has to "increment" compound_mapcount, but subpages_mapcount and compound_pincount are already correctly at 0, so delete its reinitialization of compound_pincount. A simple 100 X munmap(mmap(2GB, MAP_SHARED|MAP_POPULATE, tmpfs), 2GB) took 18 seconds on small pages, and used to take 1 second on huge pages, but now takes 119 milliseconds on huge pages. Mapping by pmds a second time used to take 860ms and now takes 92ms; mapping by pmds after mapping by ptes (when the scan is needed) used to take 870ms and now takes 495ms. But there might be some benchmarks which would show a slowdown, because tail struct pages now fall out of cache until final freeing checks them. Link: https://lkml.kernel.org/r/47ad693-717-79c8-e1ba-46c3a6602e48@google.comSigned-off-by:
-
Hugh Dickins authored
Patch series "mm,huge,rmap: unify and speed up compound mapcounts". This patch (of 3): We want to declare one more int in the first tail of a compound page: that first tail page being valuable property, since every compound page has a first tail, but perhaps no more than that. No problem on 64-bit: there is already space for it. No problem with 32-bit THPs: 5.18 commit 5232c63f ("mm: Make compound_pincount always available") kindly cleared the space for it, apparently not realizing that only 64-bit architectures enable CONFIG_THP_SWAP (whose use of tail page->private might conflict) - but make sure of that in its Kconfig. But hugetlb pages use tail page->private of the first tail page for a subpool pointer, which will conflict; and they also use page->private of the 2nd, 3rd and 4th tails. Undo "mm: add private field of first tail to struct page and struct folio"'s recent addition of private_1 to the folio tail: instead add hugetlb_subpool, hugetlb_cgroup, hugetlb_cgroup_rsvd, hugetlb_hwpoison to a second tail page of the folio: THP has long been using several fields of that tail, so make better use of it for hugetlb too. This is not how a generic folio should be declared in future, but it is an effective transitional way to make use of it. Delete the SUBPAGE_INDEX stuff, but keep __NR_USED_SUBPAGE: now 3. [hughd@google.com: prefix folio's page_1 and page_2 with double underscore, give folio's _flags_2 and _head_2 a line documentation each] Link: https://lkml.kernel.org/r/9e2cb6b-5b58-d3f2-b5ee-5f8a14e8f10@google.com Link: https://lkml.kernel.org/r/5f52de70-975-e94f-f141-543765736181@google.com Link: https://lkml.kernel.org/r/3818cc9a-9999-d064-d778-9c94c5911e6@google.comSigned-off-by:
-
Mike Kravetz authored
Commit 8ebe0a5e ("mm,madvise,hugetlb: fix unexpected data loss with MADV_DONTNEED on hugetlbfs") changed how the passed length was interpreted for hugetlb mappings. It was changed from align up to align down. The hugetlb-madvise test explicitly tests this behavior. Change test to expect new behavior. Link: https://lkml.kernel.org/r/20221104011632.357049-1-mike.kravetz@oracle.com Link: https://lore.kernel.org/oe-lkp/202211040619.2ec447d7-oliver.sang@intel.comSigned-off-by:
Mike Kravetz <mike.kravetz@oracle.com> Reported-by:
kernel test robot <oliver.sang@intel.com> Cc: David Hildenbrand <david@redhat.com> Cc: Rik van Riel <riel@surriel.com> Signed-off-by:
-
Deming Wang authored
Avoid typecasts that are needed for IS_ERR() and use IS_ERR_VALUE() instead. Link: https://lkml.kernel.org/r/20221104023818.1728-1-wangdeming@inspur.comSigned-off-by:
Deming Wang <wangdeming@inspur.com> Signed-off-by:
-
Peter Xu authored
PTE markers are ideal mechanism for things like SWP_SWAPIN_ERROR. Using a whole swap entry type for this purpose can be an overkill, especially if we already have PTE markers. Define a new bit for swapin error and replace it with pte markers. Then we can safely drop SWP_SWAPIN_ERROR and give one device slot back to swap. We used to have SWP_SWAPIN_ERROR taking the page pfn as part of the swap entry, but it's never used. Neither do I see how it can be useful because normally the swapin failure should not be caused by a bad page but bad swap device. Drop it alongside. Link: https://lkml.kernel.org/r/20221030214151.402274-3-peterx@redhat.comSigned-off-by:
Peter Xu <peterx@redhat.com> Reviewed-by:
Huang Ying <ying.huang@intel.com> Reviewed-by:
Miaohe Lin <linmiaohe@huawei.com> Acked-by:
-
Peter Xu authored
Patch series "mm: Use pte marker for swapin errors". This series uses the pte marker to replace the swapin error swap entry, then we save one more swap entry slot for swap devices. A new pte marker bit is defined. This patch (of 2): The PTE markers code is tiny and now it's enabled for most of the distributions. It's fine to keep it as-is, but to make a broader use of it (e.g. replacing read error swap entry) it needs to be there always otherwise we need special code path to take care of !PTE_MARKER case. It'll be easier just make pte marker always exist. Use this chance to extend its usage to anonymous too by simply touching up some of the old comments, because it'll be used for anonymous pages in the follow up patches. Link: https://lkml.kernel.org/r/20221030214151.402274-1-peterx@redhat.com Link: https://lkml.kernel.org/r/20221030214151.402274-2-peterx@redhat.comSigned-off-by:
-
Christophe JAILLET authored
strtobool() is the same as kstrtobool(). However, the latter is more used within the kernel. In order to remove strtobool() and slightly simplify kstrtox.h, switch to the other function name. While at it, include the corresponding header file (<linux/kstrtox.h>) Link: https://lkml.kernel.org/r/03f9401a6c8b87a1c786a2138d16b048f8d0eb53.1667336095.git.christophe.jaillet@wanadoo.frSigned-off-by:
Christophe JAILLET <christophe.jaillet@wanadoo.fr> Acked-by:
Pasha Tatashin <pasha.tatashin@soleen.com> Signed-off-by:
-
Christophe JAILLET authored
strtobool() is the same as kstrtobool(). However, the latter is more used within the kernel. In order to remove strtobool() and slightly simplify kstrtox.h, switch to the other function name. While at it, include the corresponding header file (<linux/kstrtox.h>) Link: https://lkml.kernel.org/r/ed2b46489a513988688decb53850339cc228940c.1667336095.git.christophe.jaillet@wanadoo.frSigned-off-by:
-
SeongJae Park authored
Update DAMON ABI document for the 'tried_regions' directory of DAMON sysfs interface. Link: https://lkml.kernel.org/r/20221101220328.95765-9-sj@kernel.orgSigned-off-by:
-
SeongJae Park authored
Document 'tried_regions' directory in DAMON sysfs interface usage in the administrator guide. Link: https://lkml.kernel.org/r/20221101220328.95765-8-sj@kernel.orgSigned-off-by:
-
SeongJae Park authored
Add a simple test case for ensuring tried_regions directory existence. Link: https://lkml.kernel.org/r/20221101220328.95765-7-sj@kernel.orgSigned-off-by:
-
SeongJae Park authored
When there are huge number of DAMON regions that specific scheme actions are tried to be applied, directories and files under 'tried_regions' scheme directory could waste some memory. Add another special input keyword ('clear_schemes_tried_regions') for 'state' file of each kdamond sysfs directory that can be used for cleanup of the 'tried_regions' sub-directories. [sj@kernel.org: skip regions clearing if the scheme directory was removed] Link: https://lkml.kernel.org/r/20221114182954.4745-3-sj@kernel.org Link: https://lkml.kernel.org/r/20221101220328.95765-6-sj@kernel.orgSigned-off-by: -
SeongJae Park authored
Implement the code for filling the data of 'tried_regions' DAMON sysfs directory. With this commit, DAMON sysfs interface users can write a special keyword, 'update_schemes_tried_regions' to the corresponding 'state' file of the kdamond. Then, DAMON sysfs interface will collect the tried regions information using the 'before_damos_apply()' callback for one aggregation interval and populate scheme region directories with the values. [sj@kernel.org: skip tried regions update if the scheme directory was removed] Link: https://lkml.kernel.org/r/20221114182954.4745-2-sj@kernel.org Link: https://lkml.kernel.org/r/20221101220328.95765-5-sj@kernel.orgSigned-off-by:
-
SeongJae Park authored
Implement region directories under 'tried_regions' directory of each scheme DAMON sysfs directory. This directory will provide the address range, the monitored access frequency ('nr_accesses'), and the age of each DAMON region that corresponding DAMON-based operation scheme has tried to be applied. Note that this commit doesn't implement the code for filling the data but only the sysfs directory. Link: https://lkml.kernel.org/r/20221101220328.95765-4-sj@kernel.orgSigned-off-by:
-
