| Message ID | 20240125164256.4147-13-alexandru.elisei@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Add support for arm64 MTE dynamic tag storage reuse | expand |
On 1/25/24 22:12, Alexandru Elisei wrote: > arm64 uses arch_swap_restore() to restore saved tags before the page is > swapped in and it's called in atomic context (with the ptl lock held). > > Introduce arch_swap_prepare_to_restore() that will allow an architecture to > perform extra work during swap in and outside of a critical section. > This will be used by arm64 to allocate a buffer in memory where to > temporarily save tags if tag storage is not available for the page being > swapped in. Just wondering if tag storage will always be unavailable for tagged pages being swapped in ? OR there are cases where allocation might not even be required ? This prepare phase needs to be outside the critical section - only because there might be memory allocations ?
Hi, On Thu, Feb 01, 2024 at 09:00:23AM +0530, Anshuman Khandual wrote: > > > On 1/25/24 22:12, Alexandru Elisei wrote: > > arm64 uses arch_swap_restore() to restore saved tags before the page is > > swapped in and it's called in atomic context (with the ptl lock held). > > > > Introduce arch_swap_prepare_to_restore() that will allow an architecture to > > perform extra work during swap in and outside of a critical section. > > This will be used by arm64 to allocate a buffer in memory where to > > temporarily save tags if tag storage is not available for the page being > > swapped in. > > Just wondering if tag storage will always be unavailable for tagged pages > being swapped in ? OR there are cases where allocation might not even be In some (probably most) situations, tag storage will be available for the page that will be swapped in. That's because either the page will have been taken from the swap cache (which means it hasn't been freed, so it still has tag storage reserved) or it has been allocated with vma_alloc_folio() (when it's swapped back in in a VMA with VM_MTE set). I've explained a scenario where tags will be restored for a page without tag storage in patch #28 ("mte: swap: Handle tag restoring when missing tag storage") [1]. Basically, it's because tagged pages can be mapped as tagged in one VMA and untagged in another VMA; and swap tags are restored the first time a page is swapped back in, even if it's swapped in a VMA with MTE disabled. [1] https://lore.kernel.org/linux-arm-kernel/20240125164256.4147-29-alexandru.elisei@arm.com/ > required ? This prepare phase needs to be outside the critical section - > only because there might be memory allocations ? Yes, exactly. See patch above :) Thanks, Alex
diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h index 98f81ca08cbe..2d0f04042f62 100644 --- a/include/linux/pgtable.h +++ b/include/linux/pgtable.h @@ -959,6 +959,13 @@ static inline void arch_swap_invalidate_area(int type) } #endif +#ifndef __HAVE_ARCH_SWAP_PREPARE_TO_RESTORE +static inline vm_fault_t arch_swap_prepare_to_restore(swp_entry_t entry, struct folio *folio) +{ + return 0; +} +#endif + #ifndef __HAVE_ARCH_SWAP_RESTORE static inline void arch_swap_restore(swp_entry_t entry, struct folio *folio) { diff --git a/mm/memory.c b/mm/memory.c index 7e1f4849463a..8a421e168b57 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -3975,6 +3975,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf) folio_throttle_swaprate(folio, GFP_KERNEL); + ret = arch_swap_prepare_to_restore(entry, folio); + if (ret) + goto out_page; + /* * Back out if somebody else already faulted in this pte. */ diff --git a/mm/shmem.c b/mm/shmem.c index 14427e9982f9..621fabc3b8c6 100644 --- a/mm/shmem.c +++ b/mm/shmem.c @@ -1855,6 +1855,7 @@ static int shmem_swapin_folio(struct inode *inode, pgoff_t index, struct swap_info_struct *si; struct folio *folio = NULL; swp_entry_t swap; + vm_fault_t ret; int error; VM_BUG_ON(!*foliop || !xa_is_value(*foliop)); @@ -1903,6 +1904,14 @@ static int shmem_swapin_folio(struct inode *inode, pgoff_t index, } folio_wait_writeback(folio); + ret = arch_swap_prepare_to_restore(swap, folio); + if (ret) { + if (fault_type) + *fault_type = ret; + error = -EINVAL; + goto unlock; + } + /* * Some architectures may have to restore extra metadata to the * folio after reading from swap. diff --git a/mm/swapfile.c b/mm/swapfile.c index 556ff7347d5f..49425598f778 100644 --- a/mm/swapfile.c +++ b/mm/swapfile.c @@ -1785,6 +1785,11 @@ static int unuse_pte(struct vm_area_struct *vma, pmd_t *pmd, goto setpte; } + if (arch_swap_prepare_to_restore(entry, folio)) { + ret = -EINVAL; + goto out; + } + /* * Some architectures may have to restore extra metadata to the page * when reading from swap. This metadata may be indexed by swap entry
arm64 uses arch_swap_restore() to restore saved tags before the page is swapped in and it's called in atomic context (with the ptl lock held). Introduce arch_swap_prepare_to_restore() that will allow an architecture to perform extra work during swap in and outside of a critical section. This will be used by arm64 to allocate a buffer in memory where to temporarily save tags if tag storage is not available for the page being swapped in. Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com> --- include/linux/pgtable.h | 7 +++++++ mm/memory.c | 4 ++++ mm/shmem.c | 9 +++++++++ mm/swapfile.c | 5 +++++ 4 files changed, 25 insertions(+)