summaryrefslogtreecommitdiff
path: root/fs/dax.c
AgeCommit message (Collapse)Author
2018-12-31Merge tag 'dax-fix-4.21' of ↵Linus Torvalds
git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm Pull dax fix from Dan Williams: "Clean up unnecessary usage of prepare_to_wait_exclusive(). While I feel a bit silly sending a single-commit pull-request there is nothing else queued up for dax this cycle. This change has shipped in -next for multiple releases" * tag 'dax-fix-4.21' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm: dax: Use non-exclusive wait in wait_entry_unlocked()
2018-12-28mm/mmu_notifier: use structure for invalidate_range_start/end calls v2Jérôme Glisse
To avoid having to change many call sites everytime we want to add a parameter use a structure to group all parameters for the mmu_notifier invalidate_range_start/end cakks. No functional changes with this patch. [akpm@linux-foundation.org: coding style fixes] Link: http://lkml.kernel.org/r/20181205053628.3210-3-jglisse@redhat.com Signed-off-by: Jérôme Glisse <jglisse@redhat.com> Acked-by: Christian König <christian.koenig@amd.com> Acked-by: Jan Kara <jack@suse.cz> Cc: Matthew Wilcox <mawilcox@microsoft.com> Cc: Ross Zwisler <zwisler@kernel.org> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Radim Krcmar <rkrcmar@redhat.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Felix Kuehling <felix.kuehling@amd.com> Cc: Ralph Campbell <rcampbell@nvidia.com> Cc: John Hubbard <jhubbard@nvidia.com> From: Jérôme Glisse <jglisse@redhat.com> Subject: mm/mmu_notifier: use structure for invalidate_range_start/end calls v3 fix build warning in migrate.c when CONFIG_MMU_NOTIFIER=n Link: http://lkml.kernel.org/r/20181213171330.8489-3-jglisse@redhat.com Signed-off-by: Jérôme Glisse <jglisse@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-21dax: Use non-exclusive wait in wait_entry_unlocked()Dan Williams
get_unlocked_entry() uses an exclusive wait because it is guaranteed to eventually obtain the lock and follow on with an unlock+wakeup cycle. The wait_entry_unlocked() path does not have the same guarantee. Rather than open-code an extra wakeup, just switch to a non-exclusive wait. Cc: Jan Kara <jack@suse.cz> Cc: Matthew Wilcox <willy@infradead.org> Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-12-04dax: Fix unlock mismatch with updated APIMatthew Wilcox
Internal to dax_unlock_mapping_entry(), dax_unlock_entry() is used to store a replacement entry in the Xarray at the given xas-index with the DAX_LOCKED bit clear. When called, dax_unlock_entry() expects the unlocked value of the entry relative to the current Xarray state to be specified. In most contexts dax_unlock_entry() is operating in the same scope as the matched dax_lock_entry(). However, in the dax_unlock_mapping_entry() case the implementation needs to recall the original entry. In the case where the original entry is a 'pmd' entry it is possible that the pfn performed to do the lookup is misaligned to the value retrieved in the Xarray. Change the api to return the unlock cookie from dax_lock_page() and pass it to dax_unlock_page(). This fixes a bug where dax_unlock_page() was assuming that the page was PMD-aligned if the entry was a PMD entry with signatures like: WARNING: CPU: 38 PID: 1396 at fs/dax.c:340 dax_insert_entry+0x2b2/0x2d0 RIP: 0010:dax_insert_entry+0x2b2/0x2d0 [..] Call Trace: dax_iomap_pte_fault.isra.41+0x791/0xde0 ext4_dax_huge_fault+0x16f/0x1f0 ? up_read+0x1c/0xa0 __do_fault+0x1f/0x160 __handle_mm_fault+0x1033/0x1490 handle_mm_fault+0x18b/0x3d0 Link: https://lkml.kernel.org/r/20181130154902.GL10377@bombadil.infradead.org Fixes: 9f32d221301c ("dax: Convert dax_lock_mapping_entry to XArray") Reported-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Matthew Wilcox <willy@infradead.org> Tested-by: Dan Williams <dan.j.williams@intel.com> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-11-28dax: Don't access a freed inodeMatthew Wilcox
After we drop the i_pages lock, the inode can be freed at any time. The get_unlocked_entry() code has no choice but to reacquire the lock, so it can't be used here. Create a new wait_entry_unlocked() which takes care not to acquire the lock or dereference the address_space in any way. Fixes: c2a7d2a11552 ("filesystem-dax: Introduce dax_lock_mapping_entry()") Cc: <stable@vger.kernel.org> Signed-off-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-11-28dax: Check page->mapping isn't NULLMatthew Wilcox
If we race with inode destroy, it's possible for page->mapping to be NULL before we even enter this routine, as well as after having slept waiting for the dax entry to become unlocked. Fixes: c2a7d2a11552 ("filesystem-dax: Introduce dax_lock_mapping_entry()") Cc: <stable@vger.kernel.org> Reported-by: Jan Kara <jack@suse.cz> Signed-off-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-11-19dax: Avoid losing wakeup in dax_lock_mapping_entryMatthew Wilcox
After calling get_unlocked_entry(), you have to call put_unlocked_entry() to avoid subsequent waiters losing wakeups. Fixes: c2a7d2a11552 ("filesystem-dax: Introduce dax_lock_mapping_entry()") Cc: stable@vger.kernel.org Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-11-17dax: Fix huge page faultsMatthew Wilcox
Using xas_load() with a PMD-sized xa_state would work if either a PMD-sized entry was present or a PTE sized entry was present in the first 64 entries (of the 512 PTEs in a PMD on x86). If there was no PTE in the first 64 entries, grab_mapping_entry() would believe there were no entries present, allocate a PMD-sized entry and overwrite the PTE in the page cache. Use xas_find_conflict() instead which turns out to simplify both get_unlocked_entry() and grab_mapping_entry(). Also remove a WARN_ON_ONCE from grab_mapping_entry() as it will have already triggered in get_unlocked_entry(). Fixes: cfc93c6c6c96 ("dax: Convert dax_insert_pfn_mkwrite to XArray") Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-11-17dax: Fix dax_unlock_mapping_entry for PMD pagesMatthew Wilcox
Device DAX PMD pages do not set the PageHead bit for compound pages. Fix for now by retrieving the PMD bit from the entry, but eventually we will be passed the page size by the caller. Reported-by: Dan Williams <dan.j.williams@intel.com> Fixes: 9f32d221301c ("dax: Convert dax_lock_mapping_entry to XArray") Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-11-16dax: Reinstate RCU protection of inodeMatthew Wilcox
For the device-dax case, it is possible that the inode can go away underneath us. The rcu_read_lock() was there to prevent it from being freed, and not (as I thought) to protect the tree. Bring back the rcu_read_lock() protection. Also add a little kernel-doc; while this function is not exported to modules, it is used from outside dax.c Reported-by: Dan Williams <dan.j.williams@intel.com> Fixes: 9f32d221301c ("dax: Convert dax_lock_mapping_entry to XArray") Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-11-16dax: Make sure the unlocking entry isn't lockedMatthew Wilcox
I wrote the semantics in the commit message, but didn't document it in the source code. Use a BUG_ON instead (if any code does do this, it's really buggy; we can't recover and it's worth taking the machine down). Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-11-16dax: Remove optimisation from dax_lock_mapping_entryMatthew Wilcox
Skipping some of the revalidation after we sleep can lead to returning a mapping which has already been freed. Just drop this optimisation. Reported-by: Dan Williams <dan.j.williams@intel.com> Fixes: 9f32d221301c ("dax: Convert dax_lock_mapping_entry to XArray") Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Convert page fault handlers to XArrayMatthew Wilcox
This is the last part of DAX to be converted to the XArray so remove all the old helper functions. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Convert dax_lock_mapping_entry to XArrayMatthew Wilcox
Instead of always retrying when we slept, only retry if the page has moved. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Convert dax writeback to XArrayMatthew Wilcox
Use XArray iteration instead of a pagevec. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Convert __dax_invalidate_entry to XArrayMatthew Wilcox
Avoids walking the radix tree multiple times looking for tags. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Convert dax_layout_busy_page to XArrayMatthew Wilcox
Instead of using a pagevec, just use the XArray iterators. Add a conditional rescheduling point which probably should have been there in the original. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Convert dax_insert_pfn_mkwrite to XArrayMatthew Wilcox
Add some XArray-based helper functions to replace the radix tree based metaphors currently in use. The biggest change is that converted code doesn't see its own lock bit; get_unlocked_entry() always returns an entry with the lock bit clear. So we don't have to mess around loading the current entry and clearing the lock bit; we can just store the unlocked entry that we already have. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Hash on XArray instead of mappingMatthew Wilcox
Since the XArray is embedded in the struct address_space, its address contains exactly as much entropy as the address of the mapping. This patch is purely preparatory for later patches which will simplify the wait/wake interfaces. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-10-21dax: Rename some functionsMatthew Wilcox
Remove mentions of 'radix' and 'radix tree'. Simplify some names by dropping the word 'mapping'. Signed-off-by: Matthew Wilcox <willy@infradead.org>
2018-09-29xarray: Replace exceptional entriesMatthew Wilcox
Introduce xarray value entries and tagged pointers to replace radix tree exceptional entries. This is a slight change in encoding to allow the use of an extra bit (we can now store BITS_PER_LONG - 1 bits in a value entry). It is also a change in emphasis; exceptional entries are intimidating and different. As the comment explains, you can choose to store values or pointers in the xarray and they are both first-class citizens. Signed-off-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Josef Bacik <jbacik@fb.com>
2018-09-11filesystem-dax: Fix use of zero pageMatthew Wilcox
Use my_zero_pfn instead of ZERO_PAGE(), and pass the vaddr to it instead of zero so it works on MIPS and s390 who reference the vaddr to select a zero page. Cc: <stable@vger.kernel.org> Fixes: 91d25ba8a6b0 ("dax: use common 4k zero page for dax mmap reads") Signed-off-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-08-25Merge tag 'libnvdimm-for-4.19_dax-memory-failure' of ↵Linus Torvalds
gitolite.kernel.org:pub/scm/linux/kernel/git/nvdimm/nvdimm Pull libnvdimm memory-failure update from Dave Jiang: "As it stands, memory_failure() gets thoroughly confused by dev_pagemap backed mappings. The recovery code has specific enabling for several possible page states and needs new enabling to handle poison in dax mappings. In order to support reliable reverse mapping of user space addresses: 1/ Add new locking in the memory_failure() rmap path to prevent races that would typically be handled by the page lock. 2/ Since dev_pagemap pages are hidden from the page allocator and the "compound page" accounting machinery, add a mechanism to determine the size of the mapping that encompasses a given poisoned pfn. 3/ Given pmem errors can be repaired, change the speculatively accessed poison protection, mce_unmap_kpfn(), to be reversible and otherwise allow ongoing access from the kernel. A side effect of this enabling is that MADV_HWPOISON becomes usable for dax mappings, however the primary motivation is to allow the system to survive userspace consumption of hardware-poison via dax. Specifically the current behavior is: mce: Uncorrected hardware memory error in user-access at af34214200 {1}[Hardware Error]: It has been corrected by h/w and requires no further action mce: [Hardware Error]: Machine check events logged {1}[Hardware Error]: event severity: corrected Memory failure: 0xaf34214: reserved kernel page still referenced by 1 users [..] Memory failure: 0xaf34214: recovery action for reserved kernel page: Failed mce: Memory error not recovered <reboot> ...and with these changes: Injecting memory failure for pfn 0x20cb00 at process virtual address 0x7f763dd00000 Memory failure: 0x20cb00: Killing dax-pmd:5421 due to hardware memory corruption Memory failure: 0x20cb00: recovery action for dax page: Recovered Given all the cross dependencies I propose taking this through nvdimm.git with acks from Naoya, x86/core, x86/RAS, and of course dax folks" * tag 'libnvdimm-for-4.19_dax-memory-failure' of gitolite.kernel.org:pub/scm/linux/kernel/git/nvdimm/nvdimm: libnvdimm, pmem: Restore page attributes when clearing errors x86/memory_failure: Introduce {set, clear}_mce_nospec() x86/mm/pat: Prepare {reserve, free}_memtype() for "decoy" addresses mm, memory_failure: Teach memory_failure() about dev_pagemap pages filesystem-dax: Introduce dax_lock_mapping_entry() mm, memory_failure: Collect mapping size in collect_procs() mm, madvise_inject_error: Let memory_failure() optionally take a page reference mm, dev_pagemap: Do not clear ->mapping on final put mm, madvise_inject_error: Disable MADV_SOFT_OFFLINE for ZONE_DEVICE pages filesystem-dax: Set page->index device-dax: Set page->index device-dax: Enable page_mapping() device-dax: Convert to vmf_insert_mixed and vm_fault_t
2018-08-25Merge tag 'libnvdimm-for-4.19_misc' of ↵Linus Torvalds
gitolite.kernel.org:pub/scm/linux/kernel/git/nvdimm/nvdimm Pull libnvdimm updates from Dave Jiang: "Collection of misc libnvdimm patches for 4.19 submission: - Adding support to read locked nvdimm capacity. - Change test code to make DSM failure code injection an override. - Add support for calculate maximum contiguous area for namespace. - Add support for queueing a short ARS when there is on going ARS for nvdimm. - Allow NULL to be passed in to ->direct_access() for kaddr and pfn params. - Improve smart injection support for nvdimm emulation testing. - Fix test code that supports for emulating controller temperature. - Fix hang on error before devm_memremap_pages() - Fix a bug that causes user memory corruption when data returned to user for ars_status. - Maintainer updates for Ross Zwisler emails and adding Jan Kara to fsdax" * tag 'libnvdimm-for-4.19_misc' of gitolite.kernel.org:pub/scm/linux/kernel/git/nvdimm/nvdimm: libnvdimm: fix ars_status output length calculation device-dax: avoid hang on error before devm_memremap_pages() tools/testing/nvdimm: improve emulation of smart injection filesystem-dax: Do not request kaddr and pfn when not required md/dm-writecache: Don't request pointer dummy_addr when not required dax/super: Do not request a pointer kaddr when not required tools/testing/nvdimm: kaddr and pfn can be NULL to ->direct_access() s390, dcssblk: kaddr and pfn can be NULL to ->direct_access() libnvdimm, pmem: kaddr and pfn can be NULL to ->direct_access() acpi/nfit: queue issuing of ars when an uc error notification comes in libnvdimm: Export max available extent libnvdimm: Use max contiguous area for namespace size MAINTAINERS: Add Jan Kara for filesystem DAX MAINTAINERS: update Ross Zwisler's email address tools/testing/nvdimm: Fix support for emulating controller temperature tools/testing/nvdimm: Make DSM failure code injection an override acpi, nfit: Prefer _DSM over _LSR for namespace label reads libnvdimm: Introduce locked DIMM capacity support
2018-07-30filesystem-dax: Do not request kaddr and pfn when not requiredHuaisheng Ye
Some functions within fs/dax don't need to get local pointer kaddr or variable pfn from direct_access. Using NULL instead of having to pass in useless pointer or variable that caller then just throw away. Signed-off-by: Huaisheng Ye <yehs1@lenovo.com> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
2018-07-29dax: dax_layout_busy_page() warn on !exceptionalRoss Zwisler
Inodes using DAX should only ever have exceptional entries in their page caches. Make this clear by warning if the iteration in dax_layout_busy_page() ever sees a non-exceptional entry, and by adding a comment for the pagevec_release() call which only deals with struct page pointers. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jan Kara <jack@suse.cz>
2018-07-23filesystem-dax: Introduce dax_lock_mapping_entry()Dan Williams
In preparation for implementing support for memory poison (media error) handling via dax mappings, implement a lock_page() equivalent. Poison error handling requires rmap and needs guarantees that the page->mapping association is maintained / valid (inode not freed) for the duration of the lookup. In the device-dax case it is sufficient to simply hold a dev_pagemap reference. In the filesystem-dax case we need to use the entry lock. Export the entry lock via dax_lock_mapping_entry() that uses rcu_read_lock() to protect against the inode being freed, and revalidates the page->mapping association under xa_lock(). Cc: Christoph Hellwig <hch@lst.de> Cc: Matthew Wilcox <willy@infradead.org> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
2018-07-20filesystem-dax: Set page->indexDan Williams
In support of enabling memory_failure() handling for filesystem-dax mappings, set ->index to the pgoff of the page. The rmap implementation requires ->index to bound the search through the vma interval tree. The index is set and cleared at dax_associate_entry() and dax_disassociate_entry() time respectively. Cc: Christoph Hellwig <hch@lst.de> Cc: Matthew Wilcox <willy@infradead.org> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
2018-06-08Merge tag 'libnvdimm-for-4.18' of ↵Linus Torvalds
git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm Pull libnvdimm updates from Dan Williams: "This adds a user for the new 'bytes-remaining' updates to memcpy_mcsafe() that you already received through Ingo via the x86-dax- for-linus pull. Not included here, but still targeting this cycle, is support for handling memory media errors (poison) consumed via userspace dax mappings. Summary: - DAX broke a fundamental assumption of truncate of file mapped pages. The truncate path assumed that it is safe to disconnect a pinned page from a file and let the filesystem reclaim the physical block. With DAX the page is equivalent to the filesystem block. Introduce dax_layout_busy_page() to enable filesystems to wait for pinned DAX pages to be released. Without this wait a filesystem could allocate blocks under active device-DMA to a new file. - DAX arranges for the block layer to be bypassed and uses dax_direct_access() + copy_to_iter() to satisfy read(2) calls. However, the memcpy_mcsafe() facility is available through the pmem block driver. In order to safely handle media errors, via the DAX block-layer bypass, introduce copy_to_iter_mcsafe(). - Fix cache management policy relative to the ACPI NFIT Platform Capabilities Structure to properly elide cache flushes when they are not necessary. The table indicates whether CPU caches are power-fail protected. Clarify that a deep flush is always performed on REQ_{FUA,PREFLUSH} requests" * tag 'libnvdimm-for-4.18' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm: (21 commits) dax: Use dax_write_cache* helpers libnvdimm, pmem: Do not flush power-fail protected CPU caches libnvdimm, pmem: Unconditionally deep flush on *sync libnvdimm, pmem: Complete REQ_FLUSH => REQ_PREFLUSH acpi, nfit: Remove ecc_unit_size dax: dax_insert_mapping_entry always succeeds libnvdimm, e820: Register all pmem resources libnvdimm: Debug probe times linvdimm, pmem: Preserve read-only setting for pmem devices x86, nfit_test: Add unit test for memcpy_mcsafe() pmem: Switch to copy_to_iter_mcsafe() dax: Report bytes remaining in dax_iomap_actor() dax: Introduce a ->copy_to_iter dax operation uio, lib: Fix CONFIG_ARCH_HAS_UACCESS_MCSAFE compilation xfs, dax: introduce xfs_break_dax_layouts() xfs: prepare xfs_break_layouts() for another layout type xfs: prepare xfs_break_layouts() to be called with XFS_MMAPLOCK_EXCL mm, fs, dax: handle layout changes to pinned dax mappings mm: fix __gup_device_huge vs unmap mm: introduce MEMORY_DEVICE_FS_DAX and CONFIG_DEV_PAGEMAP_OPS ...
2018-06-08Merge branch 'for-4.18/mcsafe' into libnvdimm-for-nextDan Williams
2018-06-07fs/dax.c: use new return type vm_fault_tSouptick Joarder
Use new return type vm_fault_t for fault handler. For now, this is just documenting that the function returns a VM_FAULT value rather than an errno. Once all instances are converted, vm_fault_t will become a distinct type. commit 1c8f422059ae ("mm: change return type to vm_fault_t") There was an existing bug inside dax_load_hole() if vm_insert_mixed had failed to allocate a page table, we'd return VM_FAULT_NOPAGE instead of VM_FAULT_OOM. With new vmf_insert_mixed() this issue is addressed. vm_insert_mixed_mkwrite has inefficiency when it returns an error value, driver has to convert it to vm_fault_t type. With new vmf_insert_mixed_mkwrite() this limitation will be addressed. Link: http://lkml.kernel.org/r/20180510181121.GA15239@jordon-HP-15-Notebook-PC Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com> Reviewed-by: Jan Kara <jack@suse.cz> Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Michal Hocko <mhocko@suse.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-06-02dax: dax_insert_mapping_entry always succeedsMatthew Wilcox
It does not return an error, so we don't need to check the return value for IS_ERR(). Indeed, it is a bug to do so; with a sufficiently large PFN, a legitimate DAX entry may be mistaken for an error return. Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-05-22dax: Report bytes remaining in dax_iomap_actor()Dan Williams
In preparation for protecting the dax read(2) path from media errors with copy_to_iter_mcsafe() (via dax_copy_to_iter()), convert the implementation to report the bytes successfully transferred. Cc: <x86@kernel.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Tony Luck <tony.luck@intel.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-05-22dax: Introduce a ->copy_to_iter dax operationDan Williams
Similar to the ->copy_from_iter() operation, a platform may want to deploy an architecture or device specific routine for handling reads from a dax_device like /dev/pmemX. On x86 this routine will point to a machine check safe version of copy_to_iter(). For now, add the plumbing to device-mapper and the dax core. Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-05-22mm, fs, dax: handle layout changes to pinned dax mappingsDan Williams
Background: get_user_pages() in the filesystem pins file backed memory pages for access by devices performing dma. However, it only pins the memory pages not the page-to-file offset association. If a file is truncated the pages are mapped out of the file and dma may continue indefinitely into a page that is owned by a device driver. This breaks coherency of the file vs dma, but the assumption is that if userspace wants the file-space truncated it does not matter what data is inbound from the device, it is not relevant anymore. The only expectation is that dma can safely continue while the filesystem reallocates the block(s). Problem: This expectation that dma can safely continue while the filesystem changes the block map is broken by dax. With dax the target dma page *is* the filesystem block. The model of leaving the page pinned for dma, but truncating the file block out of the file, means that the filesytem is free to reallocate a block under active dma to another file and now the expected data-incoherency situation has turned into active data-corruption. Solution: Defer all filesystem operations (fallocate(), truncate()) on a dax mode file while any page/block in the file is under active dma. This solution assumes that dma is transient. Cases where dma operations are known to not be transient, like RDMA, have been explicitly disabled via commits like 5f1d43de5416 "IB/core: disable memory registration of filesystem-dax vmas". The dax_layout_busy_page() routine is called by filesystems with a lock held against mm faults (i_mmap_lock) to find pinned / busy dax pages. The process of looking up a busy page invalidates all mappings to trigger any subsequent get_user_pages() to block on i_mmap_lock. The filesystem continues to call dax_layout_busy_page() until it finally returns no more active pages. This approach assumes that the page pinning is transient, if that assumption is violated the system would have likely hung from the uncompleted I/O. Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Matthew Wilcox <mawilcox@microsoft.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Reported-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-04-16Merge branch 'mm-rst' into docs-nextJonathan Corbet
Mike Rapoport says: These patches convert files in Documentation/vm to ReST format, add an initial index and link it to the top level documentation. There are no contents changes in the documentation, except few spelling fixes. The relatively large diffstat stems from the indentation and paragraph wrapping changes. I've tried to keep the formatting as consistent as possible, but I could miss some places that needed markup and add some markup where it was not necessary. [jc: significant conflicts in vm/hmm.rst]
2018-04-16docs/vm: rename documentation files to .rstMike Rapoport
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com> Signed-off-by: Jonathan Corbet <corbet@lwn.net>
2018-04-11page cache: use xa_lockMatthew Wilcox
Remove the address_space ->tree_lock and use the xa_lock newly added to the radix_tree_root. Rename the address_space ->page_tree to ->i_pages, since we don't really care that it's a tree. [willy@infradead.org: fix nds32, fs/dax.c] Link: http://lkml.kernel.org/r/20180406145415.GB20605@bombadil.infradead.orgLink: http://lkml.kernel.org/r/20180313132639.17387-9-willy@infradead.org Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com> Acked-by: Jeff Layton <jlayton@redhat.com> Cc: Darrick J. Wong <darrick.wong@oracle.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Cc: Will Deacon <will.deacon@arm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-03fs, dax: use page->mapping to warn if truncate collides with a busy pageDan Williams
Catch cases where extent unmap operations encounter pages that are pinned / busy. Typically this is pinned pages that are under active dma. This warning is a canary for potential data corruption as truncated blocks could be allocated to a new file while the device is still performing i/o. Here is an example of a collision that this implementation catches: WARNING: CPU: 2 PID: 1286 at fs/dax.c:343 dax_disassociate_entry+0x55/0x80 [..] Call Trace: __dax_invalidate_mapping_entry+0x6c/0xf0 dax_delete_mapping_entry+0xf/0x20 truncate_exceptional_pvec_entries.part.12+0x1af/0x200 truncate_inode_pages_range+0x268/0x970 ? tlb_gather_mmu+0x10/0x20 ? up_write+0x1c/0x40 ? unmap_mapping_range+0x73/0x140 xfs_free_file_space+0x1b6/0x5b0 [xfs] ? xfs_file_fallocate+0x7f/0x320 [xfs] ? down_write_nested+0x40/0x70 ? xfs_ilock+0x21d/0x2f0 [xfs] xfs_file_fallocate+0x162/0x320 [xfs] ? rcu_read_lock_sched_held+0x3f/0x70 ? rcu_sync_lockdep_assert+0x2a/0x50 ? __sb_start_write+0xd0/0x1b0 ? vfs_fallocate+0x20c/0x270 vfs_fallocate+0x154/0x270 SyS_fallocate+0x43/0x80 entry_SYSCALL_64_fastpath+0x1f/0x96 Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Matthew Wilcox <mawilcox@microsoft.com> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Reviewed-by: Jan Kara <jack@suse.cz> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-03-30dax: store pfns in the radixDan Williams
In preparation for examining the busy state of dax pages in the truncate path, switch from sectors to pfns in the radix. Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Matthew Wilcox <mawilcox@microsoft.com> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-02-03Merge tag 'ext4_for_linus' of ↵Linus Torvalds
git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4 Pull ext4 updates from Ted Ts'o: "Only miscellaneous cleanups and bug fixes for ext4 this cycle" * tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: ext4: create ext4_kset dynamically ext4: create ext4_feat kobject dynamically ext4: release kobject/kset even when init/register fail ext4: fix incorrect indentation of if statement ext4: correct documentation for grpid mount option ext4: use 'sbi' instead of 'EXT4_SB(sb)' ext4: save error to disk in __ext4_grp_locked_error() jbd2: fix sphinx kernel-doc build warnings ext4: fix a race in the ext4 shutdown path mbcache: make sure c_entry_count is not decremented past zero ext4: no need flush workqueue before destroying it ext4: fixed alignment and minor code cleanup in ext4.h ext4: fix ENOSPC handling in DAX page fault handler dax: pass detailed error code from dax_iomap_fault() mbcache: revert "fs/mbcache.c: make count_objects() more robust" mbcache: initialize entry->e_referenced in mb_cache_entry_create() ext4: fix up remaining files with SPDX cleanups
2018-01-31mm: add unmap_mapping_pages()Matthew Wilcox
Several users of unmap_mapping_range() would prefer to express their range in pages rather than bytes. Unfortuately, on a 32-bit kernel, you have to remember to cast your page number to a 64-bit type before shifting it, and four places in the current tree didn't remember to do that. That's a sign of a bad interface. Conveniently, unmap_mapping_range() actually converts from bytes into pages, so hoist the guts of unmap_mapping_range() into a new function unmap_mapping_pages() and convert the callers which want to use pages. Link: http://lkml.kernel.org/r/20171206142627.GD32044@bombadil.infradead.org Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com> Reported-by: "zhangyi (F)" <yi.zhang@huawei.com> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-01-31fs/dax.c: release PMD lock even when there is no PMD support in DAXJan H. Schönherr
follow_pte_pmd() can theoretically return after having acquired a PMD lock, even when DAX was not compiled with CONFIG_FS_DAX_PMD. Release the PMD lock unconditionally. Link: http://lkml.kernel.org/r/20180118133839.20587-1-jschoenh@amazon.de Fixes: f729c8c9b24f ("dax: wrprotect pmd_t in dax_mapping_entry_mkclean") Signed-off-by: Jan H. Schönherr <jschoenh@amazon.de> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Reviewed-by: Andrew Morton <akpm@linux-foundation.org> Cc: Matthew Wilcox <mawilcox@microsoft.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-01-07dax: pass detailed error code from dax_iomap_fault()Jan Kara
Ext4 needs to pass through error from its iomap handler to the page fault handler so that it can properly detect ENOSPC and force transaction commit and retry the fault (and block allocation). Add argument to dax_iomap_fault() for passing such error. Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2017-12-15Revert "mm: replace p??_write with pte_access_permitted in fault + gup paths"Linus Torvalds
This reverts commits 5c9d2d5c269c, c7da82b894e9, and e7fe7b5cae90. We'll probably need to revisit this, but basically we should not complicate the get_user_pages_fast() case, and checking the actual page table protection key bits will require more care anyway, since the protection keys depend on the exact state of the VM in question. Particularly when doing a "remote" page lookup (ie in somebody elses VM, not your own), you need to be much more careful than this was. Dave Hansen says: "So, the underlying bug here is that we now a get_user_pages_remote() and then go ahead and do the p*_access_permitted() checks against the current PKRU. This was introduced recently with the addition of the new p??_access_permitted() calls. We have checks in the VMA path for the "remote" gups and we avoid consulting PKRU for them. This got missed in the pkeys selftests because I did a ptrace read, but not a *write*. I also didn't explicitly test it against something where a COW needed to be done" It's also not entirely clear that it makes sense to check the protection key bits at this level at all. But one possible eventual solution is to make the get_user_pages_fast() case just abort if it sees protection key bits set, which makes us fall back to the regular get_user_pages() case, which then has a vma and can do the check there if we want to. We'll see. Somewhat related to this all: what we _do_ want to do some day is to check the PAGE_USER bit - it should obviously always be set for user pages, but it would be a good check to have back. Because we have no generic way to test for it, we lost it as part of moving over from the architecture-specific x86 GUP implementation to the generic one in commit e585513b76f7 ("x86/mm/gup: Switch GUP to the generic get_user_page_fast() implementation"). Cc: Peter Zijlstra <peterz@infradead.org> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: "Jérôme Glisse" <jglisse@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-29mm: replace pmd_write with pmd_access_permitted in fault + gup pathsDan Williams
The 'access_permitted' helper is used in the gup-fast path and goes beyond the simple _PAGE_RW check to also: - validate that the mapping is writable from a protection keys standpoint - validate that the pte has _PAGE_USER set since all fault paths where pmd_write is must be referencing user-memory. Link: http://lkml.kernel.org/r/151043111049.2842.15241454964150083466.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: "Jérôme Glisse" <jglisse@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-17Merge tag 'libnvdimm-for-4.15' of ↵Linus Torvalds
git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm Pull libnvdimm and dax updates from Dan Williams: "Save for a few late fixes, all of these commits have shipped in -next releases since before the merge window opened, and 0day has given a build success notification. The ext4 touches came from Jan, and the xfs touches have Darrick's reviewed-by. An xfstest for the MAP_SYNC feature has been through a few round of reviews and is on track to be merged. - Introduce MAP_SYNC and MAP_SHARED_VALIDATE, a mechanism to enable 'userspace flush' of persistent memory updates via filesystem-dax mappings. It arranges for any filesystem metadata updates that may be required to satisfy a write fault to also be flushed ("on disk") before the kernel returns to userspace from the fault handler. Effectively every write-fault that dirties metadata completes an fsync() before returning from the fault handler. The new MAP_SHARED_VALIDATE mapping type guarantees that the MAP_SYNC flag is validated as supported by the filesystem's ->mmap() file operation. - Add support for the standard ACPI 6.2 label access methods that replace the NVDIMM_FAMILY_INTEL (vendor specific) label methods. This enables interoperability with environments that only implement the standardized methods. - Add support for the ACPI 6.2 NVDIMM media error injection methods. - Add support for the NVDIMM_FAMILY_INTEL v1.6 DIMM commands for latch last shutdown status, firmware update, SMART error injection, and SMART alarm threshold control. - Cleanup physical address information disclosures to be root-only. - Fix revalidation of the DIMM "locked label area" status to support dynamic unlock of the label area. - Expand unit test infrastructure to mock the ACPI 6.2 Translate SPA (system-physical-address) command and error injection commands. Acknowledgements that came after the commits were pushed to -next: - 957ac8c421ad ("dax: fix PMD faults on zero-length files"): Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> - a39e596baa07 ("xfs: support for synchronous DAX faults") and 7b565c9f965b ("xfs: Implement xfs_filemap_pfn_mkwrite() using __xfs_filemap_fault()") Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>" * tag 'libnvdimm-for-4.15' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm: (49 commits) acpi, nfit: add 'Enable Latch System Shutdown Status' command support dax: fix general protection fault in dax_alloc_inode dax: fix PMD faults on zero-length files dax: stop requiring a live device for dax_flush() brd: remove dax support dax: quiet bdev_dax_supported() fs, dax: unify IOMAP_F_DIRTY read vs write handling policy in the dax core tools/testing/nvdimm: unit test clear-error commands acpi, nfit: validate commands against the device type tools/testing/nvdimm: stricter bounds checking for error injection commands xfs: support for synchronous DAX faults xfs: Implement xfs_filemap_pfn_mkwrite() using __xfs_filemap_fault() ext4: Support for synchronous DAX faults ext4: Simplify error handling in ext4_dax_huge_fault() dax: Implement dax_finish_sync_fault() dax, iomap: Add support for synchronous faults mm: Define MAP_SYNC and VM_SYNC flags dax: Allow tuning whether dax_insert_mapping_entry() dirties entry dax: Allow dax_iomap_fault() to return pfn dax: Fix comment describing dax_iomap_fault() ...
2017-11-15mm, pagevec: remove cold parameter for pagevecsMel Gorman
Every pagevec_init user claims the pages being released are hot even in cases where it is unlikely the pages are hot. As no one cares about the hotness of pages being released to the allocator, just ditch the parameter. No performance impact is expected as the overhead is marginal. The parameter is removed simply because it is a bit stupid to have a useless parameter copied everywhere. Link: http://lkml.kernel.org/r/20171018075952.10627-6-mgorman@techsingularity.net Signed-off-by: Mel Gorman <mgorman@techsingularity.net> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Andi Kleen <ak@linux.intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Jan Kara <jack@suse.cz> Cc: Johannes Weiner <hannes@cmpxchg.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-15mm, truncate: do not check mapping for every page being truncatedMel Gorman
During truncation, the mapping has already been checked for shmem and dax so it's known that workingset_update_node is required. This patch avoids the checks on mapping for each page being truncated. In all other cases, a lookup helper is used to determine if workingset_update_node() needs to be called. The one danger is that the API is slightly harder to use as calling workingset_update_node directly without checking for dax or shmem mappings could lead to surprises. However, the API rarely needs to be used and hopefully the comment is enough to give people the hint. sparsetruncate (tiny) 4.14.0-rc4 4.14.0-rc4 oneirq-v1r1 pickhelper-v1r1 Min Time 141.00 ( 0.00%) 140.00 ( 0.71%) 1st-qrtle Time 142.00 ( 0.00%) 141.00 ( 0.70%) 2nd-qrtle Time 142.00 ( 0.00%) 142.00 ( 0.00%) 3rd-qrtle Time 143.00 ( 0.00%) 143.00 ( 0.00%) Max-90% Time 144.00 ( 0.00%) 144.00 ( 0.00%) Max-95% Time 147.00 ( 0.00%) 145.00 ( 1.36%) Max-99% Time 195.00 ( 0.00%) 191.00 ( 2.05%) Max Time 230.00 ( 0.00%) 205.00 ( 10.87%) Amean Time 144.37 ( 0.00%) 143.82 ( 0.38%) Stddev Time 10.44 ( 0.00%) 9.00 ( 13.74%) Coeff Time 7.23 ( 0.00%) 6.26 ( 13.41%) Best99%Amean Time 143.72 ( 0.00%) 143.34 ( 0.26%) Best95%Amean Time 142.37 ( 0.00%) 142.00 ( 0.26%) Best90%Amean Time 142.19 ( 0.00%) 141.85 ( 0.24%) Best75%Amean Time 141.92 ( 0.00%) 141.58 ( 0.24%) Best50%Amean Time 141.69 ( 0.00%) 141.31 ( 0.27%) Best25%Amean Time 141.38 ( 0.00%) 140.97 ( 0.29%) As you'd expect, the gain is marginal but it can be detected. The differences in bonnie are all within the noise which is not surprising given the impact on the microbenchmark. radix_tree_update_node_t is a callback for some radix operations that optionally passes in a private field. The only user of the callback is workingset_update_node and as it no longer requires a mapping, the private field is removed. Link: http://lkml.kernel.org/r/20171018075952.10627-3-mgorman@techsingularity.net Signed-off-by: Mel Gorman <mgorman@techsingularity.net> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Jan Kara <jack@suse.cz> Cc: Andi Kleen <ak@linux.intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-15mm/mmu_notifier: avoid double notification when it is uselessJérôme Glisse
This patch only affects users of mmu_notifier->invalidate_range callback which are device drivers related to ATS/PASID, CAPI, IOMMUv2, SVM ... and it is an optimization for those users. Everyone else is unaffected by it. When clearing a pte/pmd we are given a choice to notify the event under the page table lock (notify version of *_clear_flush helpers do call the mmu_notifier_invalidate_range). But that notification is not necessary in all cases. This patch removes almost all cases where it is useless to have a call to mmu_notifier_invalidate_range before mmu_notifier_invalidate_range_end. It also adds documentation in all those cases explaining why. Below is a more in depth analysis of why this is fine to do this: For secondary TLB (non CPU TLB) like IOMMU TLB or device TLB (when device use thing like ATS/PASID to get the IOMMU to walk the CPU page table to access a process virtual address space). There is only 2 cases when you need to notify those secondary TLB while holding page table lock when clearing a pte/pmd: A) page backing address is free before mmu_notifier_invalidate_range_end B) a page table entry is updated to point to a new page (COW, write fault on zero page, __replace_page(), ...) Case A is obvious you do not want to take the risk for the device to write to a page that might now be used by something completely different. Case B is more subtle. For correctness it requires the following sequence to happen: - take page table lock - clear page table entry and notify (pmd/pte_huge_clear_flush_notify()) - set page table entry to point to new page If clearing the page table entry is not followed by a notify before setting the new pte/pmd value then you can break memory model like C11 or C++11 for the device. Consider the following scenario (device use a feature similar to ATS/ PASID): Two address addrA and addrB such that |addrA - addrB| >= PAGE_SIZE we assume they are write protected for COW (other case of B apply too). [Time N] ----------------------------------------------------------------- CPU-thread-0 {try to write to addrA} CPU-thread-1 {try to write to addrB} CPU-thread-2 {} CPU-thread-3 {} DEV-thread-0 {read addrA and populate device TLB} DEV-thread-2 {read addrB and populate device TLB} [Time N+1] --------------------------------------------------------------- CPU-thread-0 {COW_step0: {mmu_notifier_invalidate_range_start(addrA)}} CPU-thread-1 {COW_step0: {mmu_notifier_invalidate_range_start(addrB)}} CPU-thread-2 {} CPU-thread-3 {} DEV-thread-0 {} DEV-thread-2 {} [Time N+2] --------------------------------------------------------------- CPU-thread-0 {COW_step1: {update page table point to new page for addrA}} CPU-thread-1 {COW_step1: {update page table point to new page for addrB}} CPU-thread-2 {} CPU-thread-3 {} DEV-thread-0 {} DEV-thread-2 {} [Time N+3] --------------------------------------------------------------- CPU-thread-0 {preempted} CPU-thread-1 {preempted} CPU-thread-2 {write to addrA which is a write to new page} CPU-thread-3 {} DEV-thread-0 {} DEV-thread-2 {} [Time N+3] --------------------------------------------------------------- CPU-thread-0 {preempted} CPU-thread-1 {preempted} CPU-thread-2 {} CPU-thread-3 {write to addrB which is a write to new page} DEV-thread-0 {} DEV-thread-2 {} [Time N+4] --------------------------------------------------------------- CPU-thread-0 {preempted} CPU-thread-1 {COW_step3: {mmu_notifier_invalidate_range_end(addrB)}} CPU-thread-2 {} CPU-thread-3 {} DEV-thread-0 {} DEV-thread-2 {} [Time N+5] --------------------------------------------------------------- CPU-thread-0 {preempted} CPU-thread-1 {} CPU-thread-2 {} CPU-thread-3 {} DEV-thread-0 {read addrA from old page} DEV-thread-2 {read addrB from new page} So here because at time N+2 the clear page table entry was not pair with a notification to invalidate the secondary TLB, the device see the new value for addrB before seing the new value for addrA. This break total memory ordering for the device. When changing a pte to write protect or to point to a new write protected page with same content (KSM) it is ok to delay invalidate_range callback to mmu_notifier_invalidate_range_end() outside the page table lock. This is true even if the thread doing page table update is preempted right after releasing page table lock before calling mmu_notifier_invalidate_range_end Thanks to Andrea for thinking of a problematic scenario for COW. [jglisse@redhat.com: v2] Link: http://lkml.kernel.org/r/20171017031003.7481-2-jglisse@redhat.com Link: http://lkml.kernel.org/r/20170901173011.10745-1-jglisse@redhat.com Signed-off-by: Jérôme Glisse <jglisse@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Nadav Amit <nadav.amit@gmail.com> Cc: Joerg Roedel <jroedel@suse.de> Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Cc: David Woodhouse <dwmw2@infradead.org> Cc: Alistair Popple <alistair@popple.id.au> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Andrew Donnellan <andrew.donnellan@au1.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>