- 04 Sep, 2024 40 commits
-
-
Lorenzo Stoakes authored
The purpose of the vmg is to thread merge state through functions and avoid egregious parameter lists. We expand this to vma_expand(), which is used for a number of merge cases. Accordingly, adjust its callers, mmap_region() and relocate_vma_down(), to use a vmg. An added purpose of this change is the ability in a future commit to perform all new VMA range merging using vma_expand(). Link: https://lkml.kernel.org/r/4bc8c9dbc9ca52452ef8e587b28fe555854ceb38.1725040657.git.lorenzo.stoakes@oracle.comSigned-off-by:
Lorenzo Stoakes <lorenzo.stoakes@oracle.com> Reviewed-by:
Liam R. Howlett <Liam.Howlett@oracle.com> Cc: Mark Brown <broonie@kernel.org> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Bert Karwatzki <spasswolf@web.de> Cc: Jeff Xu <jeffxu@chromium.org> Cc: Jiri Olsa <olsajiri@gmail.com> Cc: Kees Cook <kees@kernel.org> Cc: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: "Paul E. McKenney" <paulmck@kernel.org> Cc: Paul Moore <paul@paul-moore.com> Cc: Sidhartha Kumar <sidhartha.kumar@oracle.com> Cc: Suren Baghdasaryan <surenb@google.com> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org>
-
Lorenzo Stoakes authored
Both can_vma_merge_before() and can_vma_merge_after() are invoked after checking for compatible VMA NUMA policy, we can simply move this to is_mergeable_vma() and abstract this altogether. In mmap_region() we set vmg->policy to NULL, so the policy comparisons checked in can_vma_merge_before() and can_vma_merge_after() are exactly equivalent to !vma_policy(vmg.next) and !vma_policy(vmg.prev). Equally, in do_brk_flags(), vmg->policy is NULL, so the can_vma_merge_after() is checking !vma_policy(vma), as we set vmg.prev to vma. In vma_merge(), we compare prev and next policies with vmg->policy before checking can_vma_merge_after() and can_vma_merge_before() respectively, which this patch causes to be checked in precisely the same way. This therefore maintains precisely the same logic as before, only now abstracted into is_mergeable_vma(). Link: https://lkml.kernel.org/r/0dbff286d9c4988333bc6f4ff3734cb95dd5410a.1725040657.git.lorenzo.stoakes@oracle.comSigned-off-by:
Vlastimil Babka <vbabka@suse.cz> Reviewed-by:
-
Lorenzo Stoakes authored
Rather than passing around huge numbers of parameters to numerous helper functions, abstract them into a single struct that we thread through the operation, the vma_merge_struct ('vmg'). Adjust vma_merge() and vma_modify() to accept this parameter, as well as predicate functions can_vma_merge_before(), can_vma_merge_after(), and the vma_modify_...() helper functions. Also introduce VMG_STATE() and VMG_VMA_STATE() helper macros to allow for easy vmg declaration. We additionally remove the requirement that vma_merge() is passed a VMA object representing the candidate new VMA. Previously it used this to obtain the mm_struct, file and anon_vma properties of the proposed range (a rather confusing state of affairs), which are now provided by the vmg directly. We also remove the pgoff calculation previously performed vma_modify(), and instead calculate this in VMG_VMA_STATE() via the vma_pgoff_offset() helper. Link: https://lkml.kernel.org/r/a955aad09d81329f6fbeb636b2dd10cde7b73dab.1725040657.git.lorenzo.stoakes@oracle.comSigned-off-by: -
Lorenzo Stoakes authored
Add a variety of VMA merge unit tests to assert that the behaviour of VMA merge is correct at an abstract level and VMAs are merged or not merged as expected. These are intentionally added _before_ we start refactoring vma_merge() in order that we can continually assert correctness throughout the rest of the series. In order to reduce churn going forward, we backport the vma_merge_struct data type to the test code which we introduce and use in a future commit, and add wrappers around the merge new and existing VMA cases. Link: https://lkml.kernel.org/r/1c7a0b43cfad2c511a6b1b52f3507696478ff51a.1725040657.git.lorenzo.stoakes@oracle.comSigned-off-by:
-
Lorenzo Stoakes authored
Patch series "mm: remove vma_merge()", v3. The infamous vma_merge() function has been the cause of a great deal of pain, bugs and confusion for a very long time. It is subtle, contains many corner cases, tries to do far too much and is as a result very fragile. The fact that the function requires there to be a numbering system to cover each possible eventuality with references to each in the many branches of its implementation as to which case you are looking at speaks to all this. Some of this complexity is inherent - unfortunately there is no getting away from the need to figure out precisely how to execute the merge, whether we need to remove VMAs, whether it is safe to do so, what constitutes a mergeable VMA and so on. However, a lot of the complexity is not inherent but instead a product of the function's 'organic' development. Liam has gone to great lengths to improve the situation as a part of his maple tree implementation, greatly improving the readability of the code, and Vlastimil and myself have additionally gone to lengths to try to improve things further. However, with the availability of userland VMA testing, it now becomes possible to perform a rather more significant refactoring while maintaining confidence in its correct operation. An attempt was previously made by Vlastimil [0] to eliminate vma_merge(), however it was rather - brutal - and an astute reader might refer to the date of that patch for insight as to its intent. This series instead divides merge operations into two natural kinds - merges which occur when a NEW vma is being added to the address space, and merges which occur when a vma is being MODIFIED. Happily, the vma_expand() function introduced by Liam, which has the capacity for also deleting a subsequent VMA, covers each of the NEW vma cases. By abstracting the actual final commit of changes to a VMA to its own function, commit_merge() and writing a wrapper around vma_expand() for new VMA cases vma_merge_new_range(), we can avoid having to use vma_merge() for these instances altogether. By doing so we are also able to then de-duplicate all existing merge logic in mmap_region() and do_brk_flags() and have everything invoke this new function, so we universally take the same approach to merging new VMAs. Having done so, we can then completely rework vma_merge() into vma_merge_existing_range() and use this for the instances where a merge is proposed for a region of an existing VMA. This eliminates vma_merge() and its numbered cases and instead divides things into logical cases - merge both, merge left, merge right (the latter 2 being either partial or full merges). The code is heavily annotated with ASCII diagrams and greatly simplified in comparison to the existing vma_merge() function. Having made this change, we take the opportunity to address an issue with merging VMAs possessing a vm_ops->close() hook - commit 714965ca ("mm/mmap: start distinguishing if vma can be removed in mergeability test") and commit fc0c8f90 ("mm, mmap: fix vma_merge() case 7 with vma_ops->close") make efforts to relax how we handle these, making assumptions about which VMAs might end up deleted (and thus, if possessing a vm_ops->close() hook, cannot be). This refactor means we do not need to guess, so instead explicitly only disallow merge in instances where a VMA with a vm_ops->close() hook would be deleted (and try a smaller merge in cases where this is possible). In addition to these changes, we introduce a new vma_merge_struct abstraction to allow VMA merge state to be threaded through the operation neatly. There is heavy unit testing provided for all merge functionality, added prior to the refactoring, allowing for before/after testing. The vm_ops->close() change also introduces exhaustive testing to demonstrate that this functions as expected, and in addition to this the reproduction code from commit fc0c8f90 ("mm, mmap: fix vma_merge() case 7 with vma_ops->close") was tested and confirmed passing. [0]:https://lore.kernel.org/linux-mm/20240401192623.18575-2-vbabka@suse.cz/ This patch (of 10): Have vma.o depend on its source dependencies explicitly, as previously these were simply being ignored as existing object files were up to date. This now correctly re-triggers the build if mm/ source is changed as well as local source code. Also set clean as a phony rule. Link: https://lkml.kernel.org/r/cover.1725040657.git.lorenzo.stoakes@oracle.com Link: https://lkml.kernel.org/r/e3ea58f08364ae5432c9a074de0195a7c7e0b04a.1725040657.git.lorenzo.stoakes@oracle.comSigned-off-by:
-
Liam R. Howlett authored
The vma_munmap_struct has a hole of 4 bytes and pushes the struct to three cachelines. Relocating the three booleans upwards allows for the struct to only use two cachelines (as reported by pahole on amd64). Before: struct vma_munmap_struct { struct vma_iterator * vmi; /* 0 8 */ struct vm_area_struct * vma; /* 8 8 */ struct vm_area_struct * prev; /* 16 8 */ struct vm_area_struct * next; /* 24 8 */ struct list_head * uf; /* 32 8 */ long unsigned int start; /* 40 8 */ long unsigned int end; /* 48 8 */ long unsigned int unmap_start; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ long unsigned int unmap_end; /* 64 8 */ int vma_count; /* 72 4 */ /* XXX 4 bytes hole, try to pack */ long unsigned int nr_pages; /* 80 8 */ long unsigned int locked_vm; /* 88 8 */ long unsigned int nr_accounted; /* 96 8 */ long unsigned int exec_vm; /* 104 8 */ long unsigned int stack_vm; /* 112 8 */ long unsigned int data_vm; /* 120 8 */ /* --- cacheline 2 boundary (128 bytes) --- */ bool unlock; /* 128 1 */ bool clear_ptes; /* 129 1 */ bool closed_vm_ops; /* 130 1 */ /* size: 136, cachelines: 3, members: 19 */ /* sum members: 127, holes: 1, sum holes: 4 */ /* padding: 5 */ /* last cacheline: 8 bytes */ }; After: struct vma_munmap_struct { struct vma_iterator * vmi; /* 0 8 */ struct vm_area_struct * vma; /* 8 8 */ struct vm_area_struct * prev; /* 16 8 */ struct vm_area_struct * next; /* 24 8 */ struct list_head * uf; /* 32 8 */ long unsigned int start; /* 40 8 */ long unsigned int end; /* 48 8 */ long unsigned int unmap_start; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ long unsigned int unmap_end; /* 64 8 */ int vma_count; /* 72 4 */ bool unlock; /* 76 1 */ bool clear_ptes; /* 77 1 */ bool closed_vm_ops; /* 78 1 */ /* XXX 1 byte hole, try to pack */ long unsigned int nr_pages; /* 80 8 */ long unsigned int locked_vm; /* 88 8 */ long unsigned int nr_accounted; /* 96 8 */ long unsigned int exec_vm; /* 104 8 */ long unsigned int stack_vm; /* 112 8 */ long unsigned int data_vm; /* 120 8 */ /* size: 128, cachelines: 2, members: 19 */ /* sum members: 127, holes: 1, sum holes: 1 */ }; Link: https://lkml.kernel.org/r/20240830040101.822209-22-Liam.Howlett@oracle.comSigned-off-by: -
Liam R. Howlett authored
The comment has been outdated since 6b73cff2 ("mm: change munmap splitting order and move_vma()"). The move_vma() was altered to fix the fragile state of the accounting since then. Link: https://lkml.kernel.org/r/20240830040101.822209-21-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
The may_expand_vm() check requires the count of the pages within the munmap range. Since this is needed for accounting and obtained later, the reodering of ma_expand_vm() to later in the call stack, after the vma munmap struct (vms) is initialised and the gather stage is potentially run, will allow for a single loop over the vmas. The gather sage does not commit any work and so everything can be undone in the case of a failure. The MAP_FIXED page count is available after the vms_gather_munmap_vmas() call, so use it instead of looping over the vmas twice. Link: https://lkml.kernel.org/r/20240830040101.822209-20-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
The do_vma_munmap() wrapper existed for callers that didn't have a vma iterator and needed to check the vma mseal status prior to calling the underlying munmap(). All callers now use a vma iterator and since the mseal check has been moved to do_vmi_align_munmap() and the vmas are aligned, this function can just be called instead. do_vmi_align_munmap() can no longer be static as ipc/shm is using it and it is exported via the mm.h header. Link: https://lkml.kernel.org/r/20240830040101.822209-19-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Change from nr_pages variable to vms.nr_accounted for the charged pages calculation. This is necessary for a future patch. This also avoids checking security_vm_enough_memory_mm() if the amount of memory won't change. Link: https://lkml.kernel.org/r/20240830040101.822209-18-Liam.Howlett@oracle.comSigned-off-by:
Kees Cook <kees@kernel.org> Reviewed-by:
Suren Baghdasaryan <surenb@google.com> Acked-by: Paul Moore <paul@paul-moore.com> [LSM] Cc: Kees Cook <kees@kernel.org> Cc: Bert Karwatzki <spasswolf@web.de> Cc: Jeff Xu <jeffxu@chromium.org> Cc: Jiri Olsa <olsajiri@gmail.com> Cc: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Mark Brown <broonie@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: "Paul E. McKenney" <paulmck@kernel.org> Cc: Sidhartha Kumar <sidhartha.kumar@oracle.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by:
-
Liam R. Howlett authored
Instead of shifting the length by PAGE_SIZE, use PHYS_PFN. Also use the existing local variable everywhere instead of some of the time. Link: https://lkml.kernel.org/r/20240830040101.822209-17-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Prior to call_mmap(), the vmas that will be replaced need to clear the way for what may happen in the call_mmap(). This clean up work includes clearing the ptes and calling the close() vm_ops. Some users do more setup than can be restored by calling the vm_ops open() function. It is safer to store the gap in the vma tree in these cases. That is to say that the failure scenario that existed before the MAP_FIXED gap exposure is restored as it is safer than trying to undo a partial mapping. Since abort_munmap_vmas() is only reattaching vmas with this change, the function is renamed to reattach_vmas(). There is also a secondary failure that may occur if there is not enough memory to store the gap. In this case, the vmas are reattached and resources freed. If the system cannot complete the call_mmap() and fails to allocate with GFP_KERNEL, then the system will print a warning about the failure. [lorenzo.stoakes@oracle.com: fix off-by-one error in vms_abort_munmap_vmas()] Link: https://lkml.kernel.org/r/52ee7eb3-955c-4ade-b5f0-28fed8ba3d0b@lucifer.local Link: https://lkml.kernel.org/r/20240830040101.822209-16-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Instead of zeroing the vma tree and then overwriting the area, let the area be overwritten and then clean up the gathered vmas using vms_complete_munmap_vmas(). To ensure locking is downgraded correctly, the mm is set regardless of MAP_FIXED or not (NULL vma). If a driver is mapping over an existing vma, then clear the ptes before the call_mmap() invocation. This is done using the vms_clean_up_area() helper. If there is a close vm_ops, that must also be called to ensure any cleanup is done before mapping over the area. This also means that calling open has been added to the abort of an unmap operation, for now. Since vm_ops->open() and vm_ops->close() are not always undo each other (state cleanup may exist in ->close() that is lost forever), the code cannot be left in this way, but that change has been isolated to another commit to make this point very obvious for traceability. Temporarily keep track of the number of pages that will be removed and reduce the charged amount. This also drops the validate_mm() call in the vma_expand() function. It is necessary to drop the validate as it would fail since the mm map_count would be incorrect during a vma expansion, prior to the cleanup from vms_complete_munmap_vmas(). Clean up the error handing of the vms_gather_munmap_vmas() by calling the verification within the function. Link: https://lkml.kernel.org/r/20240830040101.822209-15-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
With the only caller to unmap_region() being the error path of mmap_region(), the argument list can be significantly reduced. Link: https://lkml.kernel.org/r/20240830040101.822209-14-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Set the start and end address for munmap when the prev and next are gathered. This is needed to avoid incorrect addresses being used during the vms_complete_munmap_vmas() function if the prev/next vma are expanded. Add a new helper vms_complete_pte_clear(), which is needed later and will avoid growing the argument list to unmap_region() beyond the 9 it already has. Link: https://lkml.kernel.org/r/20240830040101.822209-13-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Instead of moving (or leaving) the vma iterator pointing at the previous vma, leave it pointing at the insert location. Pointing the vma iterator at the insert location allows for a cleaner walk of the vma tree for MAP_FIXED and the no expansion cases. The vma_prev() call in the case of merging the previous vma is equivalent to vma_iter_prev_range(), since the vma iterator will be pointing to the location just before the previous vma. This change needs to export abort_munmap_vmas() from mm/vma. Link: https://lkml.kernel.org/r/20240830040101.822209-12-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Adding support for a NULL vma means the init_vma_munmap() can be initialized for a less error-prone process when calling vms_complete_munmap_vmas() later on. Link: https://lkml.kernel.org/r/20240830040101.822209-11-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Open code the do_vmi_align_munmap() call so that it can be broken up later in the series. This requires exposing a few more vma operations. Link: https://lkml.kernel.org/r/20240830040101.822209-10-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
mmap_region is already passed sanitized addr and len, so change the call to do_vmi_munmap() to do_vmi_align_munmap() and inline the other checks. The inlining of the function and checks is an intermediate step in the series so future patches are easier to follow. Link: https://lkml.kernel.org/r/20240830040101.822209-9-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
vma_complete() will need to be called during an unsafe time to call validate_mm(). Extract the call in all places now so that only one location can be modified in the next change. Link: https://lkml.kernel.org/r/20240830040101.822209-8-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Clean up the code by changing the munmap operation to use a structure for the accounting and munmap variables. Since remove_mt() is only called in one location and the contents will be reduced to almost nothing. The remains of the function can be added to vms_complete_munmap_vmas(). Link: https://lkml.kernel.org/r/20240830040101.822209-7-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Use a structure to pass along all the necessary information and counters involved in removing vmas from the mm_struct. Update vmi_ function names to vms_ to indicate the first argument type change. Link: https://lkml.kernel.org/r/20240830040101.822209-6-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Create vmi_gather_munmap_vmas() to handle the gathering of vmas into a detached maple tree for removal later. Part of the gathering is the splitting of vmas that span the boundary. Link: https://lkml.kernel.org/r/20240830040101.822209-5-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Extract all necessary operations that need to be completed after the vma maple tree is updated from a munmap() operation. Extracting this makes the later patch in the series easier to understand. Link: https://lkml.kernel.org/r/20240830040101.822209-4-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Extract clean up of failed munmap() operations from do_vmi_align_munmap(). This simplifies later patches in the series. It is worth noting that the mas_for_each() loop now has a different upper limit. This should not change the number of vmas visited for reattaching to the main vma tree (mm_mt), as all vmas are reattached in both scenarios. Link: https://lkml.kernel.org/r/20240830040101.822209-3-Liam.Howlett@oracle.comSigned-off-by:
-
Liam R. Howlett authored
Patch series "Avoid MAP_FIXED gap exposure", v8. It is now possible to walk the vma tree using the rcu read locks and is beneficial to do so to reduce lock contention. Doing so while a MAP_FIXED mapping is executing means that a reader may see a gap in the vma tree that should never logically exist - and does not when using the mmap lock in read mode. The temporal gap exists because mmap_region() calls munmap() prior to installing the new mapping. This patch set stops rcu readers from seeing the temporal gap by splitting up the munmap() function into two parts. The first part prepares the vma tree for modifications by doing the necessary splits and tracks the vmas marked for removal in a side tree. The second part completes the munmapping of the vmas after the vma tree has been overwritten (either by a MAP_FIXED replacement vma or by a NULL in the munmap() case). Please note that rcu walkers will still be able to see a temporary state of split vmas that may be in the process of being removed, but the temporal gap will not be exposed. vma_start_write() are called on both parts of the split vma, so this state is detectable. If existing vmas have a vm_ops->close(), then they will be called prior to mapping the new vmas (and ptes are cleared out). Without calling ->close(), hugetlbfs tests fail (hugemmap06 specifically) due to resources still being marked as 'busy'. Unfortunately, calling the corresponding ->open() may not restore the state of the vmas, so it is safer to keep the existing failure scenario where a gap is inserted and never replaced. The failure scenario is in its own patch (0015) for traceability. This patch (of 21): The vma iterator may be left pointing to the newly created vma. This happens when inserting the new vma at the end of the old vma (!new_below). The incorrect position in the vma iterator is not exposed currently since the vma iterator is repositioned in the munmap path and is not reused in any of the other paths. This has limited impact in the current code, but is required for future changes. Link: https://lkml.kernel.org/r/20240830040101.822209-2-Liam.Howlett@oracle.com Fixes: b2b3b886 ("mm: don't use __vma_adjust() in __split_vma()") Signed-off-by:
Lorenzo Stoakes <lstoakes@gmail.com> Cc: Bert Karwatzki <spasswolf@web.de> Cc: Jeff Xu <jeffxu@chromium.org> Cc: Jiri Olsa <olsajiri@gmail.com> Cc: Kees Cook <kees@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: "Paul E. McKenney" <paulmck@kernel.org> Cc: Sidhartha Kumar <sidhartha.kumar@oracle.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com> Cc: Mark Brown <broonie@kernel.org> Cc: Paul Moore <paul@paul-moore.com> Signed-off-by:
-
Mike Yuan authored
As discussed in [1], zswap-related settings natually lose their effect when zswap is disabled, specifically zswap.writeback here. Be explicit about this behavior. [1] https://lore.kernel.org/linux-kernel/CAKEwX=Mhbwhh-=xxCU-RjMXS_n=RpV3Gtznb2m_3JgL+jzz++g@mail.gmail.com/ [akpm@linux-foundation.org: fix/simplify text] Link: https://lkml.kernel.org/r/20240823162506.12117-3-me@yhndnzj.comSigned-off-by:
Mike Yuan <me@yhndnzj.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Michal Koutný <mkoutny@suse.com> Cc: Muchun Song <muchun.song@linux.dev> Cc: Nhat Pham <nphamcs@gmail.com> Cc: Roman Gushchin <roman.gushchin@linux.dev> Cc: Shakeel Butt <shakeel.butt@linux.dev> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by:
-
Mike Yuan authored
Ensure that zswap.writeback check goes up the cgroup tree, i.e. is hierarchical. Create a subcgroup which has zswap.writeback set to 1, and the upper hierarchy's restrictions shall apply. Link: https://lkml.kernel.org/r/20240823162506.12117-2-me@yhndnzj.comSigned-off-by:
-
Usama Arif authored
With an earlier commit to handle zero-filled pages in swap directly, and with only 1% of the same-filled pages being non-zero, zswap no longer needs to handle same-filled pages and can just work on compressed pages. Link: https://lkml.kernel.org/r/20240823190545.979059-3-usamaarif642@gmail.comSigned-off-by:
Usama Arif <usamaarif642@gmail.com> Reviewed-by:
Chengming Zhou <chengming.zhou@linux.dev> Acked-by:
Yosry Ahmed <yosryahmed@google.com> Reviewed-by:
Nhat Pham <nphamcs@gmail.com> Acked-by:
Johannes Weiner <hannes@cmpxchg.org> Cc: Andi Kleen <ak@linux.intel.com> Cc: David Hildenbrand <david@redhat.com> Cc: Hugh Dickins <hughd@google.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Shakeel Butt <shakeel.butt@linux.dev> Signed-off-by:
-
Usama Arif authored
Patch series "mm: store zero pages to be swapped out in a bitmap", v8. As shown in the patch series that introduced the zswap same-filled optimization [1], 10-20% of the pages stored in zswap are same-filled. This is also observed across Meta's server fleet. By using VM counters in swap_writepage (not included in this patchseries) it was found that less than 1% of the same-filled pages to be swapped out are non-zero pages. For conventional swap setup (without zswap), rather than reading/writing these pages to flash resulting in increased I/O and flash wear, a bitmap can be used to mark these pages as zero at write time, and the pages can be filled at read time if the bit corresponding to the page is set. When using zswap with swap, this also means that a zswap_entry does not need to be allocated for zero filled pages resulting in memory savings which would offset the memory used for the bitmap. A similar attempt was made earlier in [2] where zswap would only track zero-filled pages instead of same-filled. This patchseries adds zero-filled pages optimization to swap (hence it can be used even if zswap is disabled) and removes the same-filled code from zswap (as only 1% of the same-filled pages are non-zero), simplifying code. [1] https://lore.kernel.org/all/20171018104832epcms5p1b2232e2236258de3d03d1344dde9fce0@epcms5p1/ [2] https://lore.kernel.org/lkml/20240325235018.2028408-1-yosryahmed@google.com/ This patch (of 2): Approximately 10-20% of pages to be swapped out are zero pages [1]. Rather than reading/writing these pages to flash resulting in increased I/O and flash wear, a bitmap can be used to mark these pages as zero at write time, and the pages can be filled at read time if the bit corresponding to the page is set. With this patch, NVMe writes in Meta server fleet decreased by almost 10% with conventional swap setup (zswap disabled). [1] https://lore.kernel.org/all/20171018104832epcms5p1b2232e2236258de3d03d1344dde9fce0@epcms5p1/ Link: https://lkml.kernel.org/r/20240823190545.979059-1-usamaarif642@gmail.com Link: https://lkml.kernel.org/r/20240823190545.979059-2-usamaarif642@gmail.comSigned-off-by:
-
Zhongkun He authored
should_reclaim_retry() is not ALLOC_CPUSET aware and that means that it considers reclaimability of NUMA nodes which are outside of the cpuset. If other nodes have a lot of reclaimable memory then should_reclaim_retry would instruct page allocator to retry even though there is no memory reclaimable on the cpuset nodemask. This is not really a huge problem because the number of retries without any reclaim progress is bound but it could be certainly improved. This is a cold path so this shouldn't really have a measurable impact on performance on most workloads. 1.Test step and the machines. ------------ root@vm:/sys/fs/cgroup/test# numactl -H | grep size node 0 size: 9477 MB node 1 size: 10079 MB node 2 size: 10079 MB node 3 size: 10078 MB root@vm:/sys/fs/cgroup/test# cat cpuset.mems 2 root@vm:/sys/fs/cgroup/test# stress --vm 1 --vm-bytes 12g --vm-keep stress: info: [33430] dispatching hogs: 0 cpu, 0 io, 1 vm, 0 hdd stress: FAIL: [33430] (425) <-- worker 33431 got signal 9 stress: WARN: [33430] (427) now reaping child worker processes stress: FAIL: [33430] (461) failed run completed in 2s 2. reclaim_retry_zone info: We can only alloc pages from node=2, but the reclaim_retry_zone is node=0 and return true. root@vm:/sys/kernel/debug/tracing# cat trace stress-33431 [001] ..... 13223.617311: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=1 wmark_check=1 stress-33431 [001] ..... 13223.617682: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=2 wmark_check=1 stress-33431 [001] ..... 13223.618103: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=3 wmark_check=1 stress-33431 [001] ..... 13223.618454: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=4 wmark_check=1 stress-33431 [001] ..... 13223.618770: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=5 wmark_check=1 stress-33431 [001] ..... 13223.619150: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=6 wmark_check=1 stress-33431 [001] ..... 13223.619510: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=7 wmark_check=1 stress-33431 [001] ..... 13223.619850: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=8 wmark_check=1 stress-33431 [001] ..... 13223.620171: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=9 wmark_check=1 stress-33431 [001] ..... 13223.620533: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=10 wmark_check=1 stress-33431 [001] ..... 13223.620894: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=11 wmark_check=1 stress-33431 [001] ..... 13223.621224: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=12 wmark_check=1 stress-33431 [001] ..... 13223.621551: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=13 wmark_check=1 stress-33431 [001] ..... 13223.621847: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=14 wmark_check=1 stress-33431 [001] ..... 13223.622200: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=15 wmark_check=1 stress-33431 [001] ..... 13223.622580: reclaim_retry_zone: node=0 zone=Normal order=0 reclaimable=4260 available=1772019 min_wmark=5962 no_progress_loops=16 wmark_check=1 With this patch, we can check the right node and get less retry in __alloc_pages_slowpath() because there is nothing to do. Link: https://lkml.kernel.org/r/20240822092612.3209286-1-hezhongkun.hzk@bytedance.comSigned-off-by:Zhongkun He <hezhongkun.hzk@bytedance.com> Suggested-by:
Michal Hocko <mhocko@suse.com> Acked-by:
-
Johannes Weiner authored
We've been noticing a trend of significant lock contention in the swap subsystem as core counts have been increasing in our fleet. It turns out that our swapfiles on btrfs on flash were in fact using the old swap code for rotational storage. This turns out to be a detection issue in the swapon sequence: btrfs sets si->bdev during swap activation, which currently happens *after* swapon's SSD detection and cluster setup. Thus, none of the SSD optimizations and cluster lock splitting are enabled for btrfs swap. Rearrange the swapon sequence so that filesystem activation happens *before* determining swap behavior based on the backing device. Afterwards, the nonrotational drive is detected correctly: - Adding 2097148k swap on /mnt/swapfile. Priority:-3 extents:1 across:2097148k + Adding 2097148k swap on /mnt/swapfile. Priority:-3 extents:1 across:2097148k SS Link: https://lkml.kernel.org/r/20240822112707.351844-1-hannes@cmpxchg.orgSigned-off-by:
-
Yuesong Li authored
Simplify code pattern of 'folio->index + folio_nr_pages(folio)' by using the existing helper folio_next_index(). Link: https://lkml.kernel.org/r/20240821063112.4053157-1-liyuesong@vivo.comSigned-off-by:
Yuesong Li <liyuesong@vivo.com> Cc: Matthew Wilcox <willy@infradead.org> Signed-off-by:
-
David Hildenbrand authored
Currently, running the charge_reserved_hugetlb.sh selftest we can sometimes observe something like: $ ./charge_reserved_hugetlb.sh -cgroup-v2 ... write_result is 0 After write: hugetlb_usage=0 reserved_usage=10485760 killing write_to_hugetlbfs Received 2. Deleting the memory Detach failure: Invalid argument umount: /mnt/huge: target is busy. Both cases are issues in the test. While the unmount error seems to be racy, it will make the test fail: $ ./run_vmtests.sh -t hugetlb ... # [FAIL] not ok 10 charge_reserved_hugetlb.sh -cgroup-v2 # exit=32 The issue is that we are not waiting for the write_to_hugetlbfs process to quit. So it might still have a hugetlbfs file open, about which umount is not happy. Fix that by making "killall" wait for the process to quit. The other error ("Detach failure: Invalid argument") does not seem to result in a test error, but is misleading. Turns out write_to_hugetlbfs.c unconditionally tries to cleanup using shmdt(), even when we only mmap()'ed a hugetlb file. Even worse, shmaddr is never even set for the SHM case. Fix that as well. With this change it seems to work as expected. Link: https://lkml.kernel.org/r/20240821123115.2068812-1-david@redhat.com Fixes: 29750f71 ("hugetlb_cgroup: add hugetlb_cgroup reservation tests") Signed-off-by:David Hildenbrand <david@redhat.com> Reported-by:
Mario Casquero <mcasquer@redhat.com> Reviewed-by:
Mina Almasry <almasrymina@google.com> Tested-by:
-
Matthew Wilcox (Oracle) authored
Convert x86 to use PG_arch_2 instead of PG_uncached and remove PG_uncached. Link: https://lkml.kernel.org/r/20240821193445.2294269-11-willy@infradead.orgSigned-off-by:
Matthew Wilcox (Oracle) <willy@infradead.org> Signed-off-by:
-
Matthew Wilcox (Oracle) authored
This flag has similar constraints to PG_owner_priv_1 -- it is ignored by core code, and is entirely for the use of the code which allocated the folio. Since the pagecache does not use it, individual filesystems can use it. The bufferhead code does use it, so filesystems which use the buffer cache must not use it for another purpose. Link: https://lkml.kernel.org/r/20240821193445.2294269-10-willy@infradead.orgSigned-off-by:
-
Matthew Wilcox (Oracle) authored
This function has no more callers, except folio_has_private(). Combine the two functions. Link: https://lkml.kernel.org/r/20240821193445.2294269-9-willy@infradead.orgSigned-off-by:
-
Matthew Wilcox (Oracle) authored
While there are many aliases for this flag, nobody actually uses the *PageOwnerPriv1() nor folio_*_owner_priv_1() accessors. Remove them. Link: https://lkml.kernel.org/r/20240821193445.2294269-8-willy@infradead.orgSigned-off-by:
-
Matthew Wilcox (Oracle) authored
This flag is now only used on folios, so we can remove all the page accessors. Link: https://lkml.kernel.org/r/20240821193445.2294269-7-willy@infradead.orgSigned-off-by:
-
Matthew Wilcox (Oracle) authored
There is only one caller of PageUnevictable() left; convert it to call folio_test_unevictable() and remove all the page accessors. Link: https://lkml.kernel.org/r/20240821193445.2294269-6-willy@infradead.orgSigned-off-by:
-
