kernel/linux.git/include/linux/mm.h, branch linux-6.5.y

mm/gup: reintroduce FOLL_NUMA as FOLL_HONOR_NUMA_FAULT

2023-08-21T20:07:20+00:00

Unfortunately commit 474098edac26 ("mm/gup: replace FOLL_NUMA by gup_can_follow_protnone()") missed that follow_page() and follow_trans_huge_pmd() never implicitly set FOLL_NUMA because they really don't want to fail on PROT_NONE-mapped pages -- either due to NUMA hinting or due to inaccessible (PROT_NONE) VMAs. As spelled out in commit 0b9d705297b2 ("mm: numa: Support NUMA hinting page faults from gup/gup_fast"): "Other follow_page callers like KSM should not use FOLL_NUMA, or they would fail to get the pages if they use follow_page instead of get_user_pages." liubo reported [1] that smaps_rollup results are imprecise, because they miss accounting of pages that are mapped PROT_NONE. Further, it's easy to reproduce that KSM no longer works on inaccessible VMAs on x86-64, because pte_protnone()/pmd_protnone() also indictaes "true" in inaccessible VMAs, and follow_page() refuses to return such pages right now. As KVM really depends on these NUMA hinting faults, removing the pte_protnone()/pmd_protnone() handling in GUP code completely is not really an option. To fix the issues at hand, let's revive FOLL_NUMA as FOLL_HONOR_NUMA_FAULT to restore the original behavior for now and add better comments. Set FOLL_HONOR_NUMA_FAULT independent of FOLL_FORCE in is_valid_gup_args(), to add that flag for all external GUP users. Note that there are three GUP-internal __get_user_pages() users that don't end up calling is_valid_gup_args() and consequently won't get FOLL_HONOR_NUMA_FAULT set. 1) get_dump_page(): we really don't want to handle NUMA hinting faults. It specifies FOLL_FORCE and wouldn't have honored NUMA hinting faults already. 2) populate_vma_page_range(): we really don't want to handle NUMA hinting faults. It specifies FOLL_FORCE on accessible VMAs, so it wouldn't have honored NUMA hinting faults already. 3) faultin_vma_page_range(): we similarly don't want to handle NUMA hinting faults. To make the combination of FOLL_FORCE and FOLL_HONOR_NUMA_FAULT work in inaccessible VMAs properly, we have to perform VMA accessibility checks in gup_can_follow_protnone(). As GUP-fast should reject such pages either way in pte_access_permitted()/pmd_access_permitted() -- for example on x86-64 and arm64 that both implement pte_protnone() -- let's just always fallback to ordinary GUP when stumbling over pte_protnone()/pmd_protnone(). As Linus notes [2], honoring NUMA faults might only make sense for selected GUP users. So we should really see if we can instead let relevant GUP callers specify it manually, and not trigger NUMA hinting faults from GUP as default. Prepare for that by making FOLL_HONOR_NUMA_FAULT an external GUP flag and adding appropriate documenation. While at it, remove a stale comment from follow_trans_huge_pmd(): That comment for pmd_protnone() was added in commit 2b4847e73004 ("mm: numa: serialise parallel get_user_page against THP migration"), which noted: THP does not unmap pages due to a lack of support for migration entries at a PMD level. This allows races with get_user_pages Nowadays, we do have PMD migration entries, so the comment no longer applies. Let's drop it. [1] https://lore.kernel.org/r/20230726073409.631838-1-liubo254@huawei.com [2] https://lore.kernel.org/r/CAHk-=wgRiP_9X0rRdZKT8nhemZGNateMtb366t37d8-x7VRs=g@mail.gmail.com Link: https://lkml.kernel.org/r/20230803143208.383663-2-david@redhat.com Fixes: 474098edac26 ("mm/gup: replace FOLL_NUMA by gup_can_follow_protnone()") Signed-off-by: David Hildenbrand Reported-by: liubo Closes: https://lore.kernel.org/r/20230726073409.631838-1-liubo254@huawei.com Reported-by: Peter Xu Closes: https://lore.kernel.org/all/ZMKJjDaqZ7FW0jfe@x1n/ Acked-by: Mel Gorman Acked-by: Peter Xu Cc: Hugh Dickins Cc: Jason Gunthorpe Cc: John Hubbard Cc: Linus Torvalds Cc: Matthew Wilcox (Oracle) Cc: Mel Gorman Cc: Paolo Bonzini Cc: Shuah Khan Cc: Signed-off-by: Andrew Morton

mm: fix memory ordering for mm_lock_seq and vm_lock_seq

2023-07-27T20:07:04+00:00

mm->mm_lock_seq effectively functions as a read/write lock; therefore it must be used with acquire/release semantics. A specific example is the interaction between userfaultfd_register() and lock_vma_under_rcu(). userfaultfd_register() does the following from the point where it changes a VMA's flags to the point where concurrent readers are permitted again (in a simple scenario where only a single private VMA is accessed and no merging/splitting is involved): userfaultfd_register userfaultfd_set_vm_flags vm_flags_reset vma_start_write down_write(&vma->vm_lock->lock) vma->vm_lock_seq = mm_lock_seq [marks VMA as busy] up_write(&vma->vm_lock->lock) vm_flags_init [sets VM_UFFD_* in __vm_flags] vma->vm_userfaultfd_ctx.ctx = ctx mmap_write_unlock vma_end_write_all WRITE_ONCE(mm->mm_lock_seq, mm->mm_lock_seq + 1) [unlocks VMA] There are no memory barriers in between the __vm_flags update and the mm->mm_lock_seq update that unlocks the VMA, so the unlock can be reordered to above the `vm_flags_init()` call, which means from the perspective of a concurrent reader, a VMA can be marked as a userfaultfd VMA while it is not VMA-locked. That's bad, we definitely need a store-release for the unlock operation. The non-atomic write to vma->vm_lock_seq in vma_start_write() is mostly fine because all accesses to vma->vm_lock_seq that matter are always protected by the VMA lock. There is a racy read in vma_start_read() though that can tolerate false-positives, so we should be using WRITE_ONCE() to keep things tidy and data-race-free (including for KCSAN). On the other side, lock_vma_under_rcu() works as follows in the relevant region for locking and userfaultfd check: lock_vma_under_rcu vma_start_read vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq) [early bailout] down_read_trylock(&vma->vm_lock->lock) vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq) [main check] userfaultfd_armed checks vma->vm_flags & __VM_UFFD_FLAGS Here, the interesting aspect is how far down the mm->mm_lock_seq read can be reordered - if this read is reordered down below the vma->vm_flags access, this could cause lock_vma_under_rcu() to partly operate on information that was read while the VMA was supposed to be locked. To prevent this kind of downwards bleeding of the mm->mm_lock_seq read, we need to read it with a load-acquire. Some of the comment wording is based on suggestions by Suren. BACKPORT WARNING: One of the functions changed by this patch (which I've written against Linus' tree) is vma_try_start_write(), but this function no longer exists in mm/mm-everything. I don't know whether the merged version of this patch will be ordered before or after the patch that removes vma_try_start_write(). If you're backporting this patch to a tree with vma_try_start_write(), make sure this patch changes that function. Link: https://lkml.kernel.org/r/20230721225107.942336-1-jannh@google.com Fixes: 5e31275cc997 ("mm: add per-VMA lock and helper functions to control it") Signed-off-by: Jann Horn Reviewed-by: Suren Baghdasaryan Cc: Signed-off-by: Andrew Morton

execve: always mark stack as growing down during early stack setup

2023-07-03T16:50:57+00:00

While our user stacks can grow either down (all common architectures) or up (parisc and the ia64 register stack), the initial stack setup when we copy the argument and environment strings to the new stack at execve() time is always done by extending the stack downwards. But it turns out that in commit 8d7071af8907 ("mm: always expand the stack with the mmap write lock held"), as part of making the stack growing code more robust, 'expand_downwards()' was now made to actually check the vma flags: if (!(vma->vm_flags & VM_GROWSDOWN)) return -EFAULT; and that meant that this execve-time stack expansion started failing on parisc, because on that architecture, the stack flags do not contain the VM_GROWSDOWN bit. At the same time the new check in expand_downwards() is clearly correct, and simplified the callers, so let's not remove it. The solution is instead to just codify the fact that yes, during execve(), the stack grows down. This not only matches reality, it ends up being particularly simple: we already have special execve-time flags for the stack (VM_STACK_INCOMPLETE_SETUP) and use those flags to avoid page migration during this setup time (see vma_is_temporary_stack() and invalid_migration_vma()). So just add VM_GROWSDOWN to that set of temporary flags, and now our stack flags automatically match reality, and the parisc stack expansion works again. Note that the VM_STACK_INCOMPLETE_SETUP bits will be cleared when the stack is finalized, so we only add the extra VM_GROWSDOWN bit on CONFIG_STACK_GROWSUP architectures (ie parisc) rather than adding it in general. Link: https://lore.kernel.org/all/612eaa53-6904-6e16-67fc-394f4faa0e16@bell.net/ Link: https://lore.kernel.org/all/5fd98a09-4792-1433-752d-029ae3545168@gmx.de/ Fixes: 8d7071af8907 ("mm: always expand the stack with the mmap write lock held") Reported-by: John David Anglin Reported-and-tested-by: Helge Deller Reported-and-tested-by: Guenter Roeck Signed-off-by: Linus Torvalds

Merge tag 'libnvdimm-for-6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm

2023-07-01T15:48:02+00:00

Pull nvdimm and DAX updates from Vishal Verma: "This is mostly small cleanups and fixes, with the biggest change being the change to the DAX fault handler allowing it to return VM_FAULT_HWPOISON. Summary: - DAX fixes and cleanups including a use after free, extra references, and device unregistration, and a redundant variable. - Allow the DAX fault handler to return VM_FAULT_HWPOISON - A few libnvdimm cleanups such as making some functions and variables static where sufficient. - Add a few missing prototypes for wrapped functions in tools/testing/nvdimm" * tag 'libnvdimm-for-6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm: dax: enable dax fault handler to report VM_FAULT_HWPOISON nvdimm: make security_show static nvdimm: make nd_class variable static dax/kmem: Pass valid argument to memory_group_register_static fsdax: remove redundant variable 'error' dax: Cleanup extra dax_region references dax: Introduce alloc_dev_dax_id() dax: Use device_unregister() in unregister_dax_mapping() dax: Fix dax_mapping_release() use after free tools/testing/nvdimm: Drop empty platform remove function libnvdimm: mark 'security_show' static again testing: nvdimm: add missing prototypes for wrapped functions dax: fix missing-prototype warnings

mm: Update do_vmi_align_munmap() return semantics

2023-07-01T15:10:56+00:00

Since do_vmi_align_munmap() will always honor the downgrade request on the success, the callers no longer have to deal with confusing return codes. Since all callers that request downgrade actually want the lock to be dropped, change the downgrade to an unlock request. Note that the lock still needs to be held in read mode during the page table clean up to avoid races with a map request. Update do_vmi_align_munmap() to return 0 for success. Clean up the callers and comments to always expect the unlock to be honored on the success path. The error path will always leave the lock untouched. As part of the cleanup, the wrapper function do_vmi_munmap() and callers to the wrapper are also updated. Suggested-by: Linus Torvalds Link: https://lore.kernel.org/linux-mm/20230629191414.1215929-1-willy@infradead.org/ Signed-off-by: Liam R. Howlett Signed-off-by: Linus Torvalds

xtensa: fix NOMMU build with lock_mm_and_find_vma() conversion

2023-07-01T04:08:05+00:00

It turns out that xtensa has a really odd configuration situation: you can do a no-MMU config, but still have the page fault code enabled. Which doesn't sound all that sensible, but it turns out that xtensa can have protection faults even without the MMU, and we have this: config PFAULT bool "Handle protection faults" if EXPERT && !MMU default y help Handle protection faults. MMU configurations must enable it. noMMU configurations may disable it if used memory map never generates protection faults or faults are always fatal. If unsure, say Y. which completely violated my expectations of the page fault handling. End result: Guenter reports that the xtensa no-MMU builds all fail with arch/xtensa/mm/fault.c: In function ‘do_page_fault’: arch/xtensa/mm/fault.c:133:8: error: implicit declaration of function ‘lock_mm_and_find_vma’ because I never exposed the new lock_mm_and_find_vma() function for the no-MMU case. Doing so is simple enough, and fixes the problem. Reported-and-tested-by: Guenter Roeck Fixes: a050ba1e7422 ("mm/fault: convert remaining simple cases to lock_mm_and_find_vma()") Signed-off-by: Linus Torvalds

Merge branch 'expand-stack'

2023-06-29T03:35:21+00:00

This modifies our user mode stack expansion code to always take the mmap_lock for writing before modifying the VM layout. It's actually something we always technically should have done, but because we didn't strictly need it, we were being lazy ("opportunistic" sounds so much better, doesn't it?) about things, and had this hack in place where we would extend the stack vma in-place without doing the proper locking. And it worked fine. We just needed to change vm_start (or, in the case of grow-up stacks, vm_end) and together with some special ad-hoc locking using the anon_vma lock and the mm->page_table_lock, it all was fairly straightforward. That is, it was all fine until Ruihan Li pointed out that now that the vma layout uses the maple tree code, we *really* don't just change vm_start and vm_end any more, and the locking really is broken. Oops. It's not actually all _that_ horrible to fix this once and for all, and do proper locking, but it's a bit painful. We have basically three different cases of stack expansion, and they all work just a bit differently: - the common and obvious case is the page fault handling. It's actually fairly simple and straightforward, except for the fact that we have something like 24 different versions of it, and you end up in a maze of twisty little passages, all alike. - the simplest case is the execve() code that creates a new stack. There are no real locking concerns because it's all in a private new VM that hasn't been exposed to anybody, but lockdep still can end up unhappy if you get it wrong. - and finally, we have GUP and page pinning, which shouldn't really be expanding the stack in the first place, but in addition to execve() we also use it for ptrace(). And debuggers do want to possibly access memory under the stack pointer and thus need to be able to expand the stack as a special case. None of these cases are exactly complicated, but the page fault case in particular is just repeated slightly differently many many times. And ia64 in particular has a fairly complicated situation where you can have both a regular grow-down stack _and_ a special grow-up stack for the register backing store. So to make this slightly more manageable, the bulk of this series is to first create a helper function for the most common page fault case, and convert all the straightforward architectures to it. Thus the new 'lock_mm_and_find_vma()' helper function, which ends up being used by x86, arm, powerpc, mips, riscv, alpha, arc, csky, hexagon, loongarch, nios2, sh, sparc32, and xtensa. So we not only convert more than half the architectures, we now have more shared code and avoid some of those twisty little passages. And largely due to this common helper function, the full diffstat of this series ends up deleting more lines than it adds. That still leaves eight architectures (ia64, m68k, microblaze, openrisc, parisc, s390, sparc64 and um) that end up doing 'expand_stack()' manually because they are doing something slightly different from the normal pattern. Along with the couple of special cases in execve() and GUP. So there's a couple of patches that first create 'locked' helper versions of the stack expansion functions, so that there's a obvious path forward in the conversion. The execve() case is then actually pretty simple, and is a nice cleanup from our old "grow-up stackls are special, because at execve time even they grow down". The #ifdef CONFIG_STACK_GROWSUP in that code just goes away, because it's just more straightforward to write out the stack expansion there manually, instead od having get_user_pages_remote() do it for us in some situations but not others and have to worry about locking rules for GUP. And the final step is then to just convert the remaining odd cases to a new world order where 'expand_stack()' is called with the mmap_lock held for reading, but where it might drop it and upgrade it to a write, only to return with it held for reading (in the success case) or with it completely dropped (in the failure case). In the process, we remove all the stack expansion from GUP (where dropping the lock wouldn't be ok without special rules anyway), and add it in manually to __access_remote_vm() for ptrace(). Thanks to Adrian Glaubitz and Frank Scheiner who tested the ia64 cases. Everything else here felt pretty straightforward, but the ia64 rules for stack expansion are really quite odd and very different from everything else. Also thanks to Vegard Nossum who caught me getting one of those odd conditions entirely the wrong way around. Anyway, I think I want to actually move all the stack expansion code to a whole new file of its own, rather than have it split up between mm/mmap.c and mm/memory.c, but since this will have to be backported to the initial maple tree vma introduction anyway, I tried to keep the patches _fairly_ minimal. Also, while I don't think it's valid to expand the stack from GUP, the final patch in here is a "warn if some crazy GUP user wants to try to expand the stack" patch. That one will be reverted before the final release, but it's left to catch any odd cases during the merge window and release candidates. Reported-by: Ruihan Li * branch 'expand-stack': gup: add warning if some caller would seem to want stack expansion mm: always expand the stack with the mmap write lock held execve: expand new process stack manually ahead of time mm: make find_extend_vma() fail if write lock not held powerpc/mm: convert coprocessor fault to lock_mm_and_find_vma() mm/fault: convert remaining simple cases to lock_mm_and_find_vma() arm/mm: Convert to using lock_mm_and_find_vma() riscv/mm: Convert to using lock_mm_and_find_vma() mips/mm: Convert to using lock_mm_and_find_vma() powerpc/mm: Convert to using lock_mm_and_find_vma() arm64/mm: Convert to using lock_mm_and_find_vma() mm: make the page fault mmap locking killable mm: introduce new 'lock_mm_and_find_vma()' page fault helper

Merge tag 'mm-nonmm-stable-2023-06-24-19-23' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

2023-06-28T17:59:38+00:00

Pull non-mm updates from Andrew Morton: - Arnd Bergmann has fixed a bunch of -Wmissing-prototypes in top-level directories - Douglas Anderson has added a new "buddy" mode to the hardlockup detector. It permits the detector to work on architectures which cannot provide the required interrupts, by having CPUs periodically perform checks on other CPUs - Zhen Lei has enhanced kexec's ability to support two crash regions - Petr Mladek has done a lot of cleanup on the hard lockup detector's Kconfig entries - And the usual bunch of singleton patches in various places * tag 'mm-nonmm-stable-2023-06-24-19-23' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (72 commits) kernel/time/posix-stubs.c: remove duplicated include ocfs2: remove redundant assignment to variable bit_off watchdog/hardlockup: fix typo in config HARDLOCKUP_DETECTOR_PREFER_BUDDY powerpc: move arch_trigger_cpumask_backtrace from nmi.h to irq.h devres: show which resource was invalid in __devm_ioremap_resource() watchdog/hardlockup: define HARDLOCKUP_DETECTOR_ARCH watchdog/sparc64: define HARDLOCKUP_DETECTOR_SPARC64 watchdog/hardlockup: make HAVE_NMI_WATCHDOG sparc64-specific watchdog/hardlockup: declare arch_touch_nmi_watchdog() only in linux/nmi.h watchdog/hardlockup: make the config checks more straightforward watchdog/hardlockup: sort hardlockup detector related config values a logical way watchdog/hardlockup: move SMP barriers from common code to buddy code watchdog/buddy: simplify the dependency for HARDLOCKUP_DETECTOR_PREFER_BUDDY watchdog/buddy: don't copy the cpumask in watchdog_next_cpu() watchdog/buddy: cleanup how watchdog_buddy_check_hardlockup() is called watchdog/hardlockup: remove softlockup comment in touch_nmi_watchdog() watchdog/hardlockup: in watchdog_hardlockup_check() use cpumask_copy() watchdog/hardlockup: don't use raw_cpu_ptr() in watchdog_hardlockup_kick() watchdog/hardlockup: HAVE_NMI_WATCHDOG must implement watchdog_hardlockup_probe() watchdog/hardlockup: keep kernel.nmi_watchdog sysctl as 0444 if probe fails ...

Merge tag 'mm-stable-2023-06-24-19-15' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

2023-06-28T17:28:11+00:00

Pull mm updates from Andrew Morton: - Yosry Ahmed brought back some cgroup v1 stats in OOM logs - Yosry has also eliminated cgroup's atomic rstat flushing - Nhat Pham adds the new cachestat() syscall. It provides userspace with the ability to query pagecache status - a similar concept to mincore() but more powerful and with improved usability - Mel Gorman provides more optimizations for compaction, reducing the prevalence of page rescanning - Lorenzo Stoakes has done some maintanance work on the get_user_pages() interface - Liam Howlett continues with cleanups and maintenance work to the maple tree code. Peng Zhang also does some work on maple tree - Johannes Weiner has done some cleanup work on the compaction code - David Hildenbrand has contributed additional selftests for get_user_pages() - Thomas Gleixner has contributed some maintenance and optimization work for the vmalloc code - Baolin Wang has provided some compaction cleanups, - SeongJae Park continues maintenance work on the DAMON code - Huang Ying has done some maintenance on the swap code's usage of device refcounting - Christoph Hellwig has some cleanups for the filemap/directio code - Ryan Roberts provides two patch series which yield some rationalization of the kernel's access to pte entries - use the provided APIs rather than open-coding accesses - Lorenzo Stoakes has some fixes to the interaction between pagecache and directio access to file mappings - John Hubbard has a series of fixes to the MM selftesting code - ZhangPeng continues the folio conversion campaign - Hugh Dickins has been working on the pagetable handling code, mainly with a view to reducing the load on the mmap_lock - Catalin Marinas has reduced the arm64 kmalloc() minimum alignment from 128 to 8 - Domenico Cerasuolo has improved the zswap reclaim mechanism by reorganizing the LRU management - Matthew Wilcox provides some fixups to make gfs2 work better with the buffer_head code - Vishal Moola also has done some folio conversion work - Matthew Wilcox has removed the remnants of the pagevec code - their functionality is migrated over to struct folio_batch * tag 'mm-stable-2023-06-24-19-15' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (380 commits) mm/hugetlb: remove hugetlb_set_page_subpool() mm: nommu: correct the range of mmap_sem_read_lock in task_mem() hugetlb: revert use of page_cache_next_miss() Revert "page cache: fix page_cache_next/prev_miss off by one" mm/vmscan: fix root proactive reclaim unthrottling unbalanced node mm: memcg: rename and document global_reclaim() mm: kill [add|del]_page_to_lru_list() mm: compaction: convert to use a folio in isolate_migratepages_block() mm: zswap: fix double invalidate with exclusive loads mm: remove unnecessary pagevec includes mm: remove references to pagevec mm: rename invalidate_mapping_pagevec to mapping_try_invalidate mm: remove struct pagevec net: convert sunrpc from pagevec to folio_batch i915: convert i915_gpu_error to use a folio_batch pagevec: rename fbatch_count() mm: remove check_move_unevictable_pages() drm: convert drm_gem_put_pages() to use a folio_batch i915: convert shmem_sg_free_table() to use a folio_batch scatterlist: add sg_set_folio() ...

mm: always expand the stack with the mmap write lock held

2023-06-27T16:41:30+00:00

This finishes the job of always holding the mmap write lock when extending the user stack vma, and removes the 'write_locked' argument from the vm helper functions again. For some cases, we just avoid expanding the stack at all: drivers and page pinning really shouldn't be extending any stacks. Let's see if any strange users really wanted that. It's worth noting that architectures that weren't converted to the new lock_mm_and_find_vma() helper function are left using the legacy "expand_stack()" function, but it has been changed to drop the mmap_lock and take it for writing while expanding the vma. This makes it fairly straightforward to convert the remaining architectures. As a result of dropping and re-taking the lock, the calling conventions for this function have also changed, since the old vma may no longer be valid. So it will now return the new vma if successful, and NULL - and the lock dropped - if the area could not be extended. Tested-by: Vegard Nossum Tested-by: John Paul Adrian Glaubitz # ia64 Tested-by: Frank Scheiner # ia64 Signed-off-by: Linus Torvalds