From d507e2ebd2c7be9138e5cf5c0cb1931c90c42ab1 Mon Sep 17 00:00:00 2001 From: Johannes Weiner Date: Thu, 10 Aug 2017 15:23:31 -0700 Subject: mm: fix global NR_SLAB_.*CLAIMABLE counter reads As Tetsuo points out: "Commit 385386cff4c6 ("mm: vmstat: move slab statistics from zone to node counters") broke "Slab:" field of /proc/meminfo . It shows nearly 0kB" In addition to /proc/meminfo, this problem also affects the slab counters OOM/allocation failure info dumps, can cause early -ENOMEM from overcommit protection, and miscalculate image size requirements during suspend-to-disk. This is because the patch in question switched the slab counters from the zone level to the node level, but forgot to update the global accessor functions to read the aggregate node data instead of the aggregate zone data. Use global_node_page_state() to access the global slab counters. Fixes: 385386cff4c6 ("mm: vmstat: move slab statistics from zone to node counters") Link: http://lkml.kernel.org/r/20170801134256.5400-1-hannes@cmpxchg.org Signed-off-by: Johannes Weiner Reported-by: Tetsuo Handa Acked-by: Michal Hocko Cc: Josef Bacik Cc: Vladimir Davydov Cc: Stefan Agner Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/page_alloc.c | 9 +++++---- mm/util.c | 2 +- 2 files changed, 6 insertions(+), 5 deletions(-) (limited to 'mm') diff --git a/mm/page_alloc.c b/mm/page_alloc.c index fc32aa81f359..626a430e32d1 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -4458,8 +4458,9 @@ long si_mem_available(void) * Part of the reclaimable slab consists of items that are in use, * and cannot be freed. Cap this estimate at the low watermark. */ - available += global_page_state(NR_SLAB_RECLAIMABLE) - - min(global_page_state(NR_SLAB_RECLAIMABLE) / 2, wmark_low); + available += global_node_page_state(NR_SLAB_RECLAIMABLE) - + min(global_node_page_state(NR_SLAB_RECLAIMABLE) / 2, + wmark_low); if (available < 0) available = 0; @@ -4602,8 +4603,8 @@ void show_free_areas(unsigned int filter, nodemask_t *nodemask) global_node_page_state(NR_FILE_DIRTY), global_node_page_state(NR_WRITEBACK), global_node_page_state(NR_UNSTABLE_NFS), - global_page_state(NR_SLAB_RECLAIMABLE), - global_page_state(NR_SLAB_UNRECLAIMABLE), + global_node_page_state(NR_SLAB_RECLAIMABLE), + global_node_page_state(NR_SLAB_UNRECLAIMABLE), global_node_page_state(NR_FILE_MAPPED), global_node_page_state(NR_SHMEM), global_page_state(NR_PAGETABLE), diff --git a/mm/util.c b/mm/util.c index 7b07ec852e01..9ecddf568fe3 100644 --- a/mm/util.c +++ b/mm/util.c @@ -633,7 +633,7 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) * which are reclaimable, under pressure. The dentry * cache and most inode caches should fall into this */ - free += global_page_state(NR_SLAB_RECLAIMABLE); + free += global_node_page_state(NR_SLAB_RECLAIMABLE); /* * Leave reserved pages. The pages are not for anonymous pages. -- cgit v1.2.1 From 75dddef32514f7aa58930bde6a1263253bc3d4ba Mon Sep 17 00:00:00 2001 From: Jonathan Toppins Date: Thu, 10 Aug 2017 15:23:35 -0700 Subject: mm: ratelimit PFNs busy info message The RDMA subsystem can generate several thousand of these messages per second eventually leading to a kernel crash. Ratelimit these messages to prevent this crash. Doug said: "I've been carrying a version of this for several kernel versions. I don't remember when they started, but we have one (and only one) class of machines: Dell PE R730xd, that generate these errors. When it happens, without a rate limit, we get rcu timeouts and kernel oopses. With the rate limit, we just get a lot of annoying kernel messages but the machine continues on, recovers, and eventually the memory operations all succeed" And: "> Well... why are all these EBUSY's occurring? It sounds inefficient > (at least) but if it is expected, normal and unavoidable then > perhaps we should just remove that message altogether? I don't have an answer to that question. To be honest, I haven't looked real hard. We never had this at all, then it started out of the blue, but only on our Dell 730xd machines (and it hits all of them), but no other classes or brands of machines. And we have our 730xd machines loaded up with different brands and models of cards (for instance one dedicated to mlx4 hardware, one for qib, one for mlx5, an ocrdma/cxgb4 combo, etc), so the fact that it hit all of the machines meant it wasn't tied to any particular brand/model of RDMA hardware. To me, it always smelled of a hardware oddity specific to maybe the CPUs or mainboard chipsets in these machines, so given that I'm not an mm expert anyway, I never chased it down. A few other relevant details: it showed up somewhere around 4.8/4.9 or thereabouts. It never happened before, but the prinkt has been there since the 3.18 days, so possibly the test to trigger this message was changed, or something else in the allocator changed such that the situation started happening on these machines? And, like I said, it is specific to our 730xd machines (but they are all identical, so that could mean it's something like their specific ram configuration is causing the allocator to hit this on these machine but not on other machines in the cluster, I don't want to say it's necessarily the model of chipset or CPU, there are other bits of identicalness between these machines)" Link: http://lkml.kernel.org/r/499c0f6cc10d6eb829a67f2a4d75b4228a9b356e.1501695897.git.jtoppins@redhat.com Signed-off-by: Jonathan Toppins Reviewed-by: Doug Ledford Tested-by: Doug Ledford Cc: Michal Hocko Cc: Vlastimil Babka Cc: Mel Gorman Cc: Hillf Danton Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/page_alloc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'mm') diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 626a430e32d1..6d00f746c2fd 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -7669,7 +7669,7 @@ int alloc_contig_range(unsigned long start, unsigned long end, /* Make sure the range is really isolated. */ if (test_pages_isolated(outer_start, end, false)) { - pr_info("%s: [%lx, %lx) PFNs busy\n", + pr_info_ratelimited("%s: [%lx, %lx) PFNs busy\n", __func__, outer_start, end); ret = -EBUSY; goto done; -- cgit v1.2.1 From 5af10dfd0afc559bb4b0f7e3e8227a1578333995 Mon Sep 17 00:00:00 2001 From: Andrea Arcangeli Date: Thu, 10 Aug 2017 15:23:38 -0700 Subject: userfaultfd: hugetlbfs: remove superfluous page unlock in VM_SHARED case huge_add_to_page_cache->add_to_page_cache implicitly unlocks the page before returning in case of errors. The error returned was -EEXIST by running UFFDIO_COPY on a non-hole offset of a VM_SHARED hugetlbfs mapping. It was an userland bug that triggered it and the kernel must cope with it returning -EEXIST from ioctl(UFFDIO_COPY) as expected. page dumped because: VM_BUG_ON_PAGE(!PageLocked(page)) kernel BUG at mm/filemap.c:964! invalid opcode: 0000 [#1] SMP CPU: 1 PID: 22582 Comm: qemu-system-x86 Not tainted 4.11.11-300.fc26.x86_64 #1 RIP: unlock_page+0x4a/0x50 Call Trace: hugetlb_mcopy_atomic_pte+0xc0/0x320 mcopy_atomic+0x96f/0xbe0 userfaultfd_ioctl+0x218/0xe90 do_vfs_ioctl+0xa5/0x600 SyS_ioctl+0x79/0x90 entry_SYSCALL_64_fastpath+0x1a/0xa9 Link: http://lkml.kernel.org/r/20170802165145.22628-2-aarcange@redhat.com Signed-off-by: Andrea Arcangeli Tested-by: Maxime Coquelin Reviewed-by: Mike Kravetz Cc: "Dr. David Alan Gilbert" Cc: Mike Rapoport Cc: Alexey Perevalov Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/hugetlb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'mm') diff --git a/mm/hugetlb.c b/mm/hugetlb.c index a1a0ac0ad6f6..31e207cb399b 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -4062,9 +4062,9 @@ out: return ret; out_release_unlock: spin_unlock(ptl); -out_release_nounlock: if (vm_shared) unlock_page(page); +out_release_nounlock: put_page(page); goto out; } -- cgit v1.2.1 From 16af97dc5a8975371a83d9e30a64038b48f40a2d Mon Sep 17 00:00:00 2001 From: Nadav Amit Date: Thu, 10 Aug 2017 15:23:56 -0700 Subject: mm: migrate: prevent racy access to tlb_flush_pending Patch series "fixes of TLB batching races", v6. It turns out that Linux TLB batching mechanism suffers from various races. Races that are caused due to batching during reclamation were recently handled by Mel and this patch-set deals with others. The more fundamental issue is that concurrent updates of the page-tables allow for TLB flushes to be batched on one core, while another core changes the page-tables. This other core may assume a PTE change does not require a flush based on the updated PTE value, while it is unaware that TLB flushes are still pending. This behavior affects KSM (which may result in memory corruption) and MADV_FREE and MADV_DONTNEED (which may result in incorrect behavior). A proof-of-concept can easily produce the wrong behavior of MADV_DONTNEED. Memory corruption in KSM is harder to produce in practice, but was observed by hacking the kernel and adding a delay before flushing and replacing the KSM page. Finally, there is also one memory barrier missing, which may affect architectures with weak memory model. This patch (of 7): Setting and clearing mm->tlb_flush_pending can be performed by multiple threads, since mmap_sem may only be acquired for read in task_numa_work(). If this happens, tlb_flush_pending might be cleared while one of the threads still changes PTEs and batches TLB flushes. This can lead to the same race between migration and change_protection_range() that led to the introduction of tlb_flush_pending. The result of this race was data corruption, which means that this patch also addresses a theoretically possible data corruption. An actual data corruption was not observed, yet the race was was confirmed by adding assertion to check tlb_flush_pending is not set by two threads, adding artificial latency in change_protection_range() and using sysctl to reduce kernel.numa_balancing_scan_delay_ms. Link: http://lkml.kernel.org/r/20170802000818.4760-2-namit@vmware.com Fixes: 20841405940e ("mm: fix TLB flush race between migration, and change_protection_range") Signed-off-by: Nadav Amit Acked-by: Mel Gorman Acked-by: Rik van Riel Acked-by: Minchan Kim Cc: Andy Lutomirski Cc: Hugh Dickins Cc: "David S. Miller" Cc: Andrea Arcangeli Cc: Heiko Carstens Cc: Ingo Molnar Cc: Jeff Dike Cc: Martin Schwidefsky Cc: Mel Gorman Cc: Russell King Cc: Sergey Senozhatsky Cc: Tony Luck Cc: Yoshinori Sato Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/debug.c | 2 +- mm/mprotect.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'mm') diff --git a/mm/debug.c b/mm/debug.c index db1cd26d8752..d70103bb4731 100644 --- a/mm/debug.c +++ b/mm/debug.c @@ -159,7 +159,7 @@ void dump_mm(const struct mm_struct *mm) mm->numa_next_scan, mm->numa_scan_offset, mm->numa_scan_seq, #endif #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) - mm->tlb_flush_pending, + atomic_read(&mm->tlb_flush_pending), #endif mm->def_flags, &mm->def_flags ); diff --git a/mm/mprotect.c b/mm/mprotect.c index 4180ad8cc9c5..bd0f409922cb 100644 --- a/mm/mprotect.c +++ b/mm/mprotect.c @@ -244,7 +244,7 @@ static unsigned long change_protection_range(struct vm_area_struct *vma, BUG_ON(addr >= end); pgd = pgd_offset(mm, addr); flush_cache_range(vma, addr, end); - set_tlb_flush_pending(mm); + inc_tlb_flush_pending(mm); do { next = pgd_addr_end(addr, end); if (pgd_none_or_clear_bad(pgd)) @@ -256,7 +256,7 @@ static unsigned long change_protection_range(struct vm_area_struct *vma, /* Only flush the TLB if we actually modified any entries: */ if (pages) flush_tlb_range(vma, start, end); - clear_tlb_flush_pending(mm); + dec_tlb_flush_pending(mm); return pages; } -- cgit v1.2.1 From a9b802500ebbff1544519a2969323b719dac21f0 Mon Sep 17 00:00:00 2001 From: Nadav Amit Date: Thu, 10 Aug 2017 15:24:02 -0700 Subject: Revert "mm: numa: defer TLB flush for THP migration as long as possible" While deferring TLB flushes is a good practice, the reverted patch caused pending TLB flushes to be checked while the page-table lock is not taken. As a result, in architectures with weak memory model (PPC), Linux may miss a memory-barrier, miss the fact TLB flushes are pending, and cause (in theory) a memory corruption. Since the alternative of using smp_mb__after_unlock_lock() was considered a bit open-coded, and the performance impact is expected to be small, the previous patch is reverted. This reverts b0943d61b8fa ("mm: numa: defer TLB flush for THP migration as long as possible"). Link: http://lkml.kernel.org/r/20170802000818.4760-4-namit@vmware.com Signed-off-by: Nadav Amit Suggested-by: Mel Gorman Acked-by: Mel Gorman Acked-by: Rik van Riel Cc: Minchan Kim Cc: Sergey Senozhatsky Cc: Andy Lutomirski Cc: "David S. Miller" Cc: Andrea Arcangeli Cc: Heiko Carstens Cc: Hugh Dickins Cc: Ingo Molnar Cc: Jeff Dike Cc: Martin Schwidefsky Cc: Mel Gorman Cc: Nadav Amit Cc: Russell King Cc: Tony Luck Cc: Yoshinori Sato Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/huge_memory.c | 7 +++++++ mm/migrate.c | 6 ------ 2 files changed, 7 insertions(+), 6 deletions(-) (limited to 'mm') diff --git a/mm/huge_memory.c b/mm/huge_memory.c index 86975dec0ba1..216114f6ef0b 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -1495,6 +1495,13 @@ int do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd) goto clear_pmdnuma; } + /* + * The page_table_lock above provides a memory barrier + * with change_protection_range. + */ + if (mm_tlb_flush_pending(vma->vm_mm)) + flush_tlb_range(vma, haddr, haddr + HPAGE_PMD_SIZE); + /* * Migrate the THP to the requested node, returns with page unlocked * and access rights restored. diff --git a/mm/migrate.c b/mm/migrate.c index 627671551873..d68a41da6abb 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1937,12 +1937,6 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm, put_page(new_page); goto out_fail; } - /* - * We are not sure a pending tlb flush here is for a huge page - * mapping or not. Hence use the tlb range variant - */ - if (mm_tlb_flush_pending(mm)) - flush_tlb_range(vma, mmun_start, mmun_end); /* Prepare a page as a migration target */ __SetPageLocked(new_page); -- cgit v1.2.1 From 56236a59556cfd3bae7bffb7e5f438b5ef0af880 Mon Sep 17 00:00:00 2001 From: Minchan Kim Date: Thu, 10 Aug 2017 15:24:05 -0700 Subject: mm: refactor TLB gathering API This patch is a preparatory patch for solving race problems caused by TLB batch. For that, we will increase/decrease TLB flush pending count of mm_struct whenever tlb_[gather|finish]_mmu is called. Before making it simple, this patch separates architecture specific part and rename it to arch_tlb_[gather|finish]_mmu and generic part just calls it. It shouldn't change any behavior. Link: http://lkml.kernel.org/r/20170802000818.4760-5-namit@vmware.com Signed-off-by: Minchan Kim Signed-off-by: Nadav Amit Acked-by: Mel Gorman Cc: Ingo Molnar Cc: Russell King Cc: Tony Luck Cc: Martin Schwidefsky Cc: "David S. Miller" Cc: Heiko Carstens Cc: Yoshinori Sato Cc: Jeff Dike Cc: Andrea Arcangeli Cc: Andy Lutomirski Cc: Hugh Dickins Cc: Mel Gorman Cc: Nadav Amit Cc: Rik van Riel Cc: Sergey Senozhatsky Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/memory.c | 28 +++++++++++++++++++++------- 1 file changed, 21 insertions(+), 7 deletions(-) (limited to 'mm') diff --git a/mm/memory.c b/mm/memory.c index f65beaad319b..34cba5113e06 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -215,12 +215,8 @@ static bool tlb_next_batch(struct mmu_gather *tlb) return true; } -/* tlb_gather_mmu - * Called to initialize an (on-stack) mmu_gather structure for page-table - * tear-down from @mm. The @fullmm argument is used when @mm is without - * users and we're going to destroy the full address space (exit/execve). - */ -void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) +void arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) { tlb->mm = mm; @@ -275,7 +271,8 @@ void tlb_flush_mmu(struct mmu_gather *tlb) * Called at the end of the shootdown operation to free up any resources * that were required. */ -void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) +void arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) { struct mmu_gather_batch *batch, *next; @@ -398,6 +395,23 @@ void tlb_remove_table(struct mmu_gather *tlb, void *table) #endif /* CONFIG_HAVE_RCU_TABLE_FREE */ +/* tlb_gather_mmu + * Called to initialize an (on-stack) mmu_gather structure for page-table + * tear-down from @mm. The @fullmm argument is used when @mm is without + * users and we're going to destroy the full address space (exit/execve). + */ +void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) +{ + arch_tlb_gather_mmu(tlb, mm, start, end); +} + +void tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) +{ + arch_tlb_finish_mmu(tlb, start, end); +} + /* * Note: this doesn't free the actual pages themselves. That * has been handled earlier when unmapping all the memory regions. -- cgit v1.2.1 From 0a2dd266dd6b7a31503b5bbe63af05961a6b446d Mon Sep 17 00:00:00 2001 From: Minchan Kim Date: Thu, 10 Aug 2017 15:24:09 -0700 Subject: mm: make tlb_flush_pending global Currently, tlb_flush_pending is used only for CONFIG_[NUMA_BALANCING| COMPACTION] but upcoming patches to solve subtle TLB flush batching problem will use it regardless of compaction/NUMA so this patch doesn't remove the dependency. [akpm@linux-foundation.org: remove more ifdefs from world's ugliest printk statement] Link: http://lkml.kernel.org/r/20170802000818.4760-6-namit@vmware.com Signed-off-by: Minchan Kim Signed-off-by: Nadav Amit Acked-by: Mel Gorman Cc: "David S. Miller" Cc: Andrea Arcangeli Cc: Andy Lutomirski Cc: Heiko Carstens Cc: Hugh Dickins Cc: Ingo Molnar Cc: Jeff Dike Cc: Martin Schwidefsky Cc: Mel Gorman Cc: Nadav Amit Cc: Rik van Riel Cc: Russell King Cc: Sergey Senozhatsky Cc: Tony Luck Cc: Yoshinori Sato Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/debug.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'mm') diff --git a/mm/debug.c b/mm/debug.c index d70103bb4731..5715448ab0b5 100644 --- a/mm/debug.c +++ b/mm/debug.c @@ -124,9 +124,7 @@ void dump_mm(const struct mm_struct *mm) #ifdef CONFIG_NUMA_BALANCING "numa_next_scan %lu numa_scan_offset %lu numa_scan_seq %d\n" #endif -#if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) "tlb_flush_pending %d\n" -#endif "def_flags: %#lx(%pGv)\n", mm, mm->mmap, mm->vmacache_seqnum, mm->task_size, @@ -158,9 +156,7 @@ void dump_mm(const struct mm_struct *mm) #ifdef CONFIG_NUMA_BALANCING mm->numa_next_scan, mm->numa_scan_offset, mm->numa_scan_seq, #endif -#if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) atomic_read(&mm->tlb_flush_pending), -#endif mm->def_flags, &mm->def_flags ); } -- cgit v1.2.1 From 99baac21e4585f4258f919502c6e23f1e5edc98c Mon Sep 17 00:00:00 2001 From: Minchan Kim Date: Thu, 10 Aug 2017 15:24:12 -0700 Subject: mm: fix MADV_[FREE|DONTNEED] TLB flush miss problem Nadav reported parallel MADV_DONTNEED on same range has a stale TLB problem and Mel fixed it[1] and found same problem on MADV_FREE[2]. Quote from Mel Gorman: "The race in question is CPU 0 running madv_free and updating some PTEs while CPU 1 is also running madv_free and looking at the same PTEs. CPU 1 may have writable TLB entries for a page but fail the pte_dirty check (because CPU 0 has updated it already) and potentially fail to flush. Hence, when madv_free on CPU 1 returns, there are still potentially writable TLB entries and the underlying PTE is still present so that a subsequent write does not necessarily propagate the dirty bit to the underlying PTE any more. Reclaim at some unknown time at the future may then see that the PTE is still clean and discard the page even though a write has happened in the meantime. I think this is possible but I could have missed some protection in madv_free that prevents it happening." This patch aims for solving both problems all at once and is ready for other problem with KSM, MADV_FREE and soft-dirty story[3]. TLB batch API(tlb_[gather|finish]_mmu] uses [inc|dec]_tlb_flush_pending and mmu_tlb_flush_pending so that when tlb_finish_mmu is called, we can catch there are parallel threads going on. In that case, forcefully, flush TLB to prevent for user to access memory via stale TLB entry although it fail to gather page table entry. I confirmed this patch works with [4] test program Nadav gave so this patch supersedes "mm: Always flush VMA ranges affected by zap_page_range v2" in current mmotm. NOTE: This patch modifies arch-specific TLB gathering interface(x86, ia64, s390, sh, um). It seems most of architecture are straightforward but s390 need to be careful because tlb_flush_mmu works only if mm->context.flush_mm is set to non-zero which happens only a pte entry really is cleared by ptep_get_and_clear and friends. However, this problem never changes the pte entries but need to flush to prevent memory access from stale tlb. [1] http://lkml.kernel.org/r/20170725101230.5v7gvnjmcnkzzql3@techsingularity.net [2] http://lkml.kernel.org/r/20170725100722.2dxnmgypmwnrfawp@suse.de [3] http://lkml.kernel.org/r/BD3A0EBE-ECF4-41D4-87FA-C755EA9AB6BD@gmail.com [4] https://patchwork.kernel.org/patch/9861621/ [minchan@kernel.org: decrease tlb flush pending count in tlb_finish_mmu] Link: http://lkml.kernel.org/r/20170808080821.GA31730@bbox Link: http://lkml.kernel.org/r/20170802000818.4760-7-namit@vmware.com Signed-off-by: Minchan Kim Signed-off-by: Nadav Amit Reported-by: Nadav Amit Reported-by: Mel Gorman Acked-by: Mel Gorman Cc: Ingo Molnar Cc: Russell King Cc: Tony Luck Cc: Martin Schwidefsky Cc: "David S. Miller" Cc: Heiko Carstens Cc: Yoshinori Sato Cc: Jeff Dike Cc: Andrea Arcangeli Cc: Andy Lutomirski Cc: Hugh Dickins Cc: Mel Gorman Cc: Nadav Amit Cc: Rik van Riel Cc: Sergey Senozhatsky Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/memory.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) (limited to 'mm') diff --git a/mm/memory.c b/mm/memory.c index 34cba5113e06..e158f7ac6730 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -272,10 +272,13 @@ void tlb_flush_mmu(struct mmu_gather *tlb) * that were required. */ void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) + unsigned long start, unsigned long end, bool force) { struct mmu_gather_batch *batch, *next; + if (force) + __tlb_adjust_range(tlb, start, end - start); + tlb_flush_mmu(tlb); /* keep the page table cache within bounds */ @@ -404,12 +407,23 @@ void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) { arch_tlb_gather_mmu(tlb, mm, start, end); + inc_tlb_flush_pending(tlb->mm); } void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) { - arch_tlb_finish_mmu(tlb, start, end); + /* + * If there are parallel threads are doing PTE changes on same range + * under non-exclusive lock(e.g., mmap_sem read-side) but defer TLB + * flush by batching, a thread has stable TLB entry can fail to flush + * the TLB by observing pte_none|!pte_dirty, for example so flush TLB + * forcefully if we detect parallel PTE batching threads. + */ + bool force = mm_tlb_flush_nested(tlb->mm); + + arch_tlb_finish_mmu(tlb, start, end, force); + dec_tlb_flush_pending(tlb->mm); } /* -- cgit v1.2.1 From b3a81d0841a954a3d3e782059960f53e63b37066 Mon Sep 17 00:00:00 2001 From: Minchan Kim Date: Thu, 10 Aug 2017 15:24:15 -0700 Subject: mm: fix KSM data corruption Nadav reported KSM can corrupt the user data by the TLB batching race[1]. That means data user written can be lost. Quote from Nadav Amit: "For this race we need 4 CPUs: CPU0: Caches a writable and dirty PTE entry, and uses the stale value for write later. CPU1: Runs madvise_free on the range that includes the PTE. It would clear the dirty-bit. It batches TLB flushes. CPU2: Writes 4 to /proc/PID/clear_refs , clearing the PTEs soft-dirty. We care about the fact that it clears the PTE write-bit, and of course, batches TLB flushes. CPU3: Runs KSM. Our purpose is to pass the following test in write_protect_page(): if (pte_write(*pvmw.pte) || pte_dirty(*pvmw.pte) || (pte_protnone(*pvmw.pte) && pte_savedwrite(*pvmw.pte))) Since it will avoid TLB flush. And we want to do it while the PTE is stale. Later, and before replacing the page, we would be able to change the page. Note that all the operations the CPU1-3 perform canhappen in parallel since they only acquire mmap_sem for read. We start with two identical pages. Everything below regards the same page/PTE. CPU0 CPU1 CPU2 CPU3 ---- ---- ---- ---- Write the same value on page [cache PTE as dirty in TLB] MADV_FREE pte_mkclean() 4 > clear_refs pte_wrprotect() write_protect_page() [ success, no flush ] pages_indentical() [ ok ] Write to page different value [Ok, using stale PTE] replace_page() Later, CPU1, CPU2 and CPU3 would flush the TLB, but that is too late. CPU0 already wrote on the page, but KSM ignored this write, and it got lost" In above scenario, MADV_FREE is fixed by changing TLB batching API including [set|clear]_tlb_flush_pending. Remained thing is soft-dirty part. This patch changes soft-dirty uses TLB batching API instead of flush_tlb_mm and KSM checks pending TLB flush by using mm_tlb_flush_pending so that it will flush TLB to avoid data lost if there are other parallel threads pending TLB flush. [1] http://lkml.kernel.org/r/BD3A0EBE-ECF4-41D4-87FA-C755EA9AB6BD@gmail.com Link: http://lkml.kernel.org/r/20170802000818.4760-8-namit@vmware.com Signed-off-by: Minchan Kim Signed-off-by: Nadav Amit Reported-by: Nadav Amit Tested-by: Nadav Amit Reviewed-by: Andrea Arcangeli Cc: Mel Gorman Cc: Hugh Dickins Cc: "David S. Miller" Cc: Andy Lutomirski Cc: Heiko Carstens Cc: Ingo Molnar Cc: Jeff Dike Cc: Martin Schwidefsky Cc: Mel Gorman Cc: Nadav Amit Cc: Rik van Riel Cc: Russell King Cc: Sergey Senozhatsky Cc: Tony Luck Cc: Yoshinori Sato Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/ksm.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'mm') diff --git a/mm/ksm.c b/mm/ksm.c index 4dc92f138786..db20f8436bc3 100644 --- a/mm/ksm.c +++ b/mm/ksm.c @@ -1038,7 +1038,8 @@ static int write_protect_page(struct vm_area_struct *vma, struct page *page, goto out_unlock; if (pte_write(*pvmw.pte) || pte_dirty(*pvmw.pte) || - (pte_protnone(*pvmw.pte) && pte_savedwrite(*pvmw.pte))) { + (pte_protnone(*pvmw.pte) && pte_savedwrite(*pvmw.pte)) || + mm_tlb_flush_pending(mm)) { pte_t entry; swapped = PageSwapCache(page); -- cgit v1.2.1 From af54aed94bf3a1cf0b847bbbf00f9a58e278b338 Mon Sep 17 00:00:00 2001 From: Wei Wang Date: Thu, 10 Aug 2017 15:24:21 -0700 Subject: mm/balloon_compaction.c: don't zero ballooned pages Revert commit bb01b64cfab7 ("mm/balloon_compaction.c: enqueue zero page to balloon device")' Zeroing ballon pages is rather time consuming, especially when a lot of pages are in flight. E.g. 7GB worth of ballooned memory takes 2.8s with __GFP_ZERO while it takes ~491ms without it. The original commit argued that zeroing will help ksmd to merge these pages on the host but this argument is assuming that the host actually marks balloon pages for ksm which is not universally true. So we pay performance penalty for something that even might not be used in the end which is wrong. The host can zero out pages on its own when there is a need. [mhocko@kernel.org: new changelog text] Link: http://lkml.kernel.org/r/1501761557-9758-1-git-send-email-wei.w.wang@intel.com Fixes: bb01b64cfab7 ("mm/balloon_compaction.c: enqueue zero page to balloon device") Signed-off-by: Wei Wang Acked-by: Michael S. Tsirkin Acked-by: Michal Hocko Cc: zhenwei.pi Cc: David Hildenbrand Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/balloon_compaction.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'mm') diff --git a/mm/balloon_compaction.c b/mm/balloon_compaction.c index 9075aa54e955..b06d9fe23a28 100644 --- a/mm/balloon_compaction.c +++ b/mm/balloon_compaction.c @@ -24,7 +24,7 @@ struct page *balloon_page_enqueue(struct balloon_dev_info *b_dev_info) { unsigned long flags; struct page *page = alloc_page(balloon_mapping_gfp_mask() | - __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_ZERO); + __GFP_NOMEMALLOC | __GFP_NORETRY); if (!page) return NULL; -- cgit v1.2.1 From d041353dc98a6339182cd6f628b4c8f111278cb3 Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Thu, 10 Aug 2017 15:24:24 -0700 Subject: mm: fix list corruptions on shmem shrinklist We saw many list corruption warnings on shmem shrinklist: WARNING: CPU: 18 PID: 177 at lib/list_debug.c:59 __list_del_entry+0x9e/0xc0 list_del corruption. prev->next should be ffff9ae5694b82d8, but was ffff9ae5699ba960 Modules linked in: intel_rapl sb_edac edac_core x86_pkg_temp_thermal coretemp iTCO_wdt iTCO_vendor_support crct10dif_pclmul crc32_pclmul ghash_clmulni_intel raid0 dcdbas shpchp wmi hed i2c_i801 ioatdma lpc_ich i2c_smbus acpi_cpufreq tcp_diag inet_diag sch_fq_codel ipmi_si ipmi_devintf ipmi_msghandler igb ptp crc32c_intel pps_core i2c_algo_bit i2c_core dca ipv6 crc_ccitt CPU: 18 PID: 177 Comm: kswapd1 Not tainted 4.9.34-t3.el7.twitter.x86_64 #1 Hardware name: Dell Inc. PowerEdge C6220/0W6W6G, BIOS 2.2.3 11/07/2013 Call Trace: dump_stack+0x4d/0x66 __warn+0xcb/0xf0 warn_slowpath_fmt+0x4f/0x60 __list_del_entry+0x9e/0xc0 shmem_unused_huge_shrink+0xfa/0x2e0 shmem_unused_huge_scan+0x20/0x30 super_cache_scan+0x193/0x1a0 shrink_slab.part.41+0x1e3/0x3f0 shrink_slab+0x29/0x30 shrink_node+0xf9/0x2f0 kswapd+0x2d8/0x6c0 kthread+0xd7/0xf0 ret_from_fork+0x22/0x30 WARNING: CPU: 23 PID: 639 at lib/list_debug.c:33 __list_add+0x89/0xb0 list_add corruption. prev->next should be next (ffff9ae5699ba960), but was ffff9ae5694b82d8. (prev=ffff9ae5694b82d8). Modules linked in: intel_rapl sb_edac edac_core x86_pkg_temp_thermal coretemp iTCO_wdt iTCO_vendor_support crct10dif_pclmul crc32_pclmul ghash_clmulni_intel raid0 dcdbas shpchp wmi hed i2c_i801 ioatdma lpc_ich i2c_smbus acpi_cpufreq tcp_diag inet_diag sch_fq_codel ipmi_si ipmi_devintf ipmi_msghandler igb ptp crc32c_intel pps_core i2c_algo_bit i2c_core dca ipv6 crc_ccitt CPU: 23 PID: 639 Comm: systemd-udevd Tainted: G W 4.9.34-t3.el7.twitter.x86_64 #1 Hardware name: Dell Inc. PowerEdge C6220/0W6W6G, BIOS 2.2.3 11/07/2013 Call Trace: dump_stack+0x4d/0x66 __warn+0xcb/0xf0 warn_slowpath_fmt+0x4f/0x60 __list_add+0x89/0xb0 shmem_setattr+0x204/0x230 notify_change+0x2ef/0x440 do_truncate+0x5d/0x90 path_openat+0x331/0x1190 do_filp_open+0x7e/0xe0 do_sys_open+0x123/0x200 SyS_open+0x1e/0x20 do_syscall_64+0x61/0x170 entry_SYSCALL64_slow_path+0x25/0x25 The problem is that shmem_unused_huge_shrink() moves entries from the global sbinfo->shrinklist to its local lists and then releases the spinlock. However, a parallel shmem_setattr() could access one of these entries directly and add it back to the global shrinklist if it is removed, with the spinlock held. The logic itself looks solid since an entry could be either in a local list or the global list, otherwise it is removed from one of them by list_del_init(). So probably the race condition is that, one CPU is in the middle of INIT_LIST_HEAD() but the other CPU calls list_empty() which returns true too early then the following list_add_tail() sees a corrupted entry. list_empty_careful() is designed to fix this situation. [akpm@linux-foundation.org: add comments] Link: http://lkml.kernel.org/r/20170803054630.18775-1-xiyou.wangcong@gmail.com Fixes: 779750d20b93 ("shmem: split huge pages beyond i_size under memory pressure") Signed-off-by: Cong Wang Acked-by: Linus Torvalds Acked-by: Kirill A. Shutemov Cc: Hugh Dickins Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/shmem.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'mm') diff --git a/mm/shmem.c b/mm/shmem.c index b0aa6075d164..6540e5982444 100644 --- a/mm/shmem.c +++ b/mm/shmem.c @@ -1022,7 +1022,11 @@ static int shmem_setattr(struct dentry *dentry, struct iattr *attr) */ if (IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) { spin_lock(&sbinfo->shrinklist_lock); - if (list_empty(&info->shrinklist)) { + /* + * _careful to defend against unlocked access to + * ->shrink_list in shmem_unused_huge_shrink() + */ + if (list_empty_careful(&info->shrinklist)) { list_add_tail(&info->shrinklist, &sbinfo->shrinklist); sbinfo->shrinklist_len++; @@ -1817,7 +1821,11 @@ alloc_nohuge: page = shmem_alloc_and_acct_page(gfp, info, sbinfo, * to shrink under memory pressure. */ spin_lock(&sbinfo->shrinklist_lock); - if (list_empty(&info->shrinklist)) { + /* + * _careful to defend against unlocked access to + * ->shrink_list in shmem_unused_huge_shrink() + */ + if (list_empty_careful(&info->shrinklist)) { list_add_tail(&info->shrinklist, &sbinfo->shrinklist); sbinfo->shrinklist_len++; -- cgit v1.2.1 From aac2fea94f7a3df8ad1eeb477eb2643f81fd5393 Mon Sep 17 00:00:00 2001 From: "Kirill A. Shutemov" Date: Thu, 10 Aug 2017 15:24:27 -0700 Subject: rmap: do not call mmu_notifier_invalidate_page() under ptl MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit MMU notifiers can sleep, but in page_mkclean_one() we call mmu_notifier_invalidate_page() under page table lock. Let's instead use mmu_notifier_invalidate_range() outside page_vma_mapped_walk() loop. [jglisse@redhat.com: try_to_unmap_one() do not call mmu_notifier under ptl] Link: http://lkml.kernel.org/r/20170809204333.27485-1-jglisse@redhat.com Link: http://lkml.kernel.org/r/20170804134928.l4klfcnqatni7vsc@black.fi.intel.com Fixes: c7ab0d2fdc84 ("mm: convert try_to_unmap_one() to use page_vma_mapped_walk()") Signed-off-by: Kirill A. Shutemov Signed-off-by: Jérôme Glisse Reported-by: axie Cc: Alex Deucher Cc: "Writer, Tim" Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/rmap.c | 52 ++++++++++++++++++++++++++++++---------------------- 1 file changed, 30 insertions(+), 22 deletions(-) (limited to 'mm') diff --git a/mm/rmap.c b/mm/rmap.c index c8993c63eb25..c1286d47aa1f 100644 --- a/mm/rmap.c +++ b/mm/rmap.c @@ -888,10 +888,10 @@ static bool page_mkclean_one(struct page *page, struct vm_area_struct *vma, .flags = PVMW_SYNC, }; int *cleaned = arg; + bool invalidation_needed = false; while (page_vma_mapped_walk(&pvmw)) { int ret = 0; - address = pvmw.address; if (pvmw.pte) { pte_t entry; pte_t *pte = pvmw.pte; @@ -899,11 +899,11 @@ static bool page_mkclean_one(struct page *page, struct vm_area_struct *vma, if (!pte_dirty(*pte) && !pte_write(*pte)) continue; - flush_cache_page(vma, address, pte_pfn(*pte)); - entry = ptep_clear_flush(vma, address, pte); + flush_cache_page(vma, pvmw.address, pte_pfn(*pte)); + entry = ptep_clear_flush(vma, pvmw.address, pte); entry = pte_wrprotect(entry); entry = pte_mkclean(entry); - set_pte_at(vma->vm_mm, address, pte, entry); + set_pte_at(vma->vm_mm, pvmw.address, pte, entry); ret = 1; } else { #ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE @@ -913,11 +913,11 @@ static bool page_mkclean_one(struct page *page, struct vm_area_struct *vma, if (!pmd_dirty(*pmd) && !pmd_write(*pmd)) continue; - flush_cache_page(vma, address, page_to_pfn(page)); - entry = pmdp_huge_clear_flush(vma, address, pmd); + flush_cache_page(vma, pvmw.address, page_to_pfn(page)); + entry = pmdp_huge_clear_flush(vma, pvmw.address, pmd); entry = pmd_wrprotect(entry); entry = pmd_mkclean(entry); - set_pmd_at(vma->vm_mm, address, pmd, entry); + set_pmd_at(vma->vm_mm, pvmw.address, pmd, entry); ret = 1; #else /* unexpected pmd-mapped page? */ @@ -926,11 +926,16 @@ static bool page_mkclean_one(struct page *page, struct vm_area_struct *vma, } if (ret) { - mmu_notifier_invalidate_page(vma->vm_mm, address); (*cleaned)++; + invalidation_needed = true; } } + if (invalidation_needed) { + mmu_notifier_invalidate_range(vma->vm_mm, address, + address + (1UL << compound_order(page))); + } + return true; } @@ -1323,7 +1328,7 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, }; pte_t pteval; struct page *subpage; - bool ret = true; + bool ret = true, invalidation_needed = false; enum ttu_flags flags = (enum ttu_flags)arg; /* munlock has nothing to gain from examining un-locked vmas */ @@ -1363,11 +1368,9 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, VM_BUG_ON_PAGE(!pvmw.pte, page); subpage = page - page_to_pfn(page) + pte_pfn(*pvmw.pte); - address = pvmw.address; - if (!(flags & TTU_IGNORE_ACCESS)) { - if (ptep_clear_flush_young_notify(vma, address, + if (ptep_clear_flush_young_notify(vma, pvmw.address, pvmw.pte)) { ret = false; page_vma_mapped_walk_done(&pvmw); @@ -1376,7 +1379,7 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, } /* Nuke the page table entry. */ - flush_cache_page(vma, address, pte_pfn(*pvmw.pte)); + flush_cache_page(vma, pvmw.address, pte_pfn(*pvmw.pte)); if (should_defer_flush(mm, flags)) { /* * We clear the PTE but do not flush so potentially @@ -1386,11 +1389,12 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, * transition on a cached TLB entry is written through * and traps if the PTE is unmapped. */ - pteval = ptep_get_and_clear(mm, address, pvmw.pte); + pteval = ptep_get_and_clear(mm, pvmw.address, + pvmw.pte); set_tlb_ubc_flush_pending(mm, pte_dirty(pteval)); } else { - pteval = ptep_clear_flush(vma, address, pvmw.pte); + pteval = ptep_clear_flush(vma, pvmw.address, pvmw.pte); } /* Move the dirty bit to the page. Now the pte is gone. */ @@ -1405,12 +1409,12 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, if (PageHuge(page)) { int nr = 1 << compound_order(page); hugetlb_count_sub(nr, mm); - set_huge_swap_pte_at(mm, address, + set_huge_swap_pte_at(mm, pvmw.address, pvmw.pte, pteval, vma_mmu_pagesize(vma)); } else { dec_mm_counter(mm, mm_counter(page)); - set_pte_at(mm, address, pvmw.pte, pteval); + set_pte_at(mm, pvmw.address, pvmw.pte, pteval); } } else if (pte_unused(pteval)) { @@ -1434,7 +1438,7 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, swp_pte = swp_entry_to_pte(entry); if (pte_soft_dirty(pteval)) swp_pte = pte_swp_mksoft_dirty(swp_pte); - set_pte_at(mm, address, pvmw.pte, swp_pte); + set_pte_at(mm, pvmw.address, pvmw.pte, swp_pte); } else if (PageAnon(page)) { swp_entry_t entry = { .val = page_private(subpage) }; pte_t swp_pte; @@ -1460,7 +1464,7 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, * If the page was redirtied, it cannot be * discarded. Remap the page to page table. */ - set_pte_at(mm, address, pvmw.pte, pteval); + set_pte_at(mm, pvmw.address, pvmw.pte, pteval); SetPageSwapBacked(page); ret = false; page_vma_mapped_walk_done(&pvmw); @@ -1468,7 +1472,7 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, } if (swap_duplicate(entry) < 0) { - set_pte_at(mm, address, pvmw.pte, pteval); + set_pte_at(mm, pvmw.address, pvmw.pte, pteval); ret = false; page_vma_mapped_walk_done(&pvmw); break; @@ -1484,14 +1488,18 @@ static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma, swp_pte = swp_entry_to_pte(entry); if (pte_soft_dirty(pteval)) swp_pte = pte_swp_mksoft_dirty(swp_pte); - set_pte_at(mm, address, pvmw.pte, swp_pte); + set_pte_at(mm, pvmw.address, pvmw.pte, swp_pte); } else dec_mm_counter(mm, mm_counter_file(page)); discard: page_remove_rmap(subpage, PageHuge(page)); put_page(page); - mmu_notifier_invalidate_page(mm, address); + invalidation_needed = true; } + + if (invalidation_needed) + mmu_notifier_invalidate_range(mm, address, + address + (1UL << compound_order(page))); return ret; } -- cgit v1.2.1