If hmm_range_fault() fails a folio_trylock() in do_swap_page, trying to acquire the lock of a device-private folio for migration, to ram, the function will spin until it succeeds grabbing the lock.
However, if the process holding the lock is depending on a work item to be completed, which is scheduled on the same CPU as the spinning hmm_range_fault(), that work item might be starved and we end up in a livelock / starvation situation which is never resolved. This can happen, for example if the process holding the device-private folio lock is stuck in migrate_device_unmap()->lru_add_drain_all() sinc lru_add_drain_all() requires a short work-item to be run on all online cpus to complete. A prerequisite for this to happen is: a) Both zone device and system memory folios are considered in migrate_device_unmap(), so that there is a reason to call lru_add_drain_all() for a system memory folio while a folio lock is held on a zone device folio. b) The zone device folio has an initial mapcount > 1 which causes at least one migration PTE entry insertion to be deferred to try_to_migrate(), which can happen after the call to lru_add_drain_all(). c) No or voluntary only preemption. This all seems pretty unlikely to happen, but indeed is hit by the "xe_exec_system_allocator" igt test. Resolve this by waiting for the folio to be unlocked if the folio_trylock() fails in do_swap_page(). Rename migration_entry_wait_on_locked() to softleaf_entry_wait_unlock() and update its documentation to indicate the new use-case. Future code improvements might consider moving the lru_add_drain_all() call in migrate_device_unmap() to be called *after* all pages have migration entries inserted. That would eliminate also b) above. v2: - Instead of a cond_resched() in hmm_range_fault(), eliminate the problem by waiting for the folio to be unlocked in do_swap_page() (Alistair Popple, Andrew Morton) v3: - Add a stub migration_entry_wait_on_locked() for the !CONFIG_MIGRATION case. (Kernel Test Robot) v4: - Rename migrate_entry_wait_on_locked() to softleaf_entry_wait_on_locked() and update docs (Alistair Popple) v5: - Add a WARN_ON_ONCE() for the !CONFIG_MIGRATION version of softleaf_entry_wait_on_locked(). - Modify wording around function names in the commit message (Andrew Morton) Suggested-by: Alistair Popple <[email protected]> Fixes: 1afaeb8293c9 ("mm/migrate: Trylock device page in do_swap_page") Cc: Ralph Campbell <[email protected]> Cc: Christoph Hellwig <[email protected]> Cc: Jason Gunthorpe <[email protected]> Cc: Jason Gunthorpe <[email protected]> Cc: Leon Romanovsky <[email protected]> Cc: Andrew Morton <[email protected]> Cc: Matthew Brost <[email protected]> Cc: John Hubbard <[email protected]> Cc: Alistair Popple <[email protected]> Cc: [email protected] Cc: <[email protected]> Signed-off-by: Thomas Hellström <[email protected]> Cc: <[email protected]> # v6.15+ Reviewed-by: John Hubbard <[email protected]> #v3 Reviewed-by: Alistair Popple <[email protected]> Link: https://patch.msgid.link/[email protected] (cherry picked from commit a69d1ab971a624c6f112cea61536569d579c3215) Signed-off-by: Rodrigo Vivi <[email protected]> (cherry picked from commit b570f37a2ce480be26c665345c5514686a8a0274) Signed-off-by: Thomas Hellström <[email protected]> --- include/linux/migrate.h | 8 ++++++++ mm/filemap.c | 13 +++++++++---- mm/memory.c | 3 ++- 3 files changed, 19 insertions(+), 5 deletions(-) diff --git a/include/linux/migrate.h b/include/linux/migrate.h index 1f0ac122c3bf..8061efd89041 100644 --- a/include/linux/migrate.h +++ b/include/linux/migrate.h @@ -97,6 +97,14 @@ static inline int set_movable_ops(const struct movable_operations *ops, enum pag return -ENOSYS; } +static inline void migration_entry_wait_on_locked(swp_entry_t entry, spinlock_t *ptl) + __releases(ptl) +{ + WARN_ON_ONCE(1); + + spin_unlock(ptl); +} + #endif /* CONFIG_MIGRATION */ #ifdef CONFIG_NUMA_BALANCING diff --git a/mm/filemap.c b/mm/filemap.c index 024b71da5224..8a7f4ce69aff 100644 --- a/mm/filemap.c +++ b/mm/filemap.c @@ -1386,14 +1386,16 @@ static inline int folio_wait_bit_common(struct folio *folio, int bit_nr, #ifdef CONFIG_MIGRATION /** - * migration_entry_wait_on_locked - Wait for a migration entry to be removed - * @entry: migration swap entry. + * migration_entry_wait_on_locked - Wait for a migration entry or + * device_private entry to be removed. + * @entry: migration or device_private swap entry. * @ptl: already locked ptl. This function will drop the lock. * - * Wait for a migration entry referencing the given page to be removed. This is + * Wait for a migration entry referencing the given page, or device_private + * entry referencing a dvice_private page to be unlocked. This is * equivalent to folio_put_wait_locked(folio, TASK_UNINTERRUPTIBLE) except * this can be called without taking a reference on the page. Instead this - * should be called while holding the ptl for the migration entry referencing + * should be called while holding the ptl for @entry referencing * the page. * * Returns after unlocking the ptl. @@ -1435,6 +1437,9 @@ void migration_entry_wait_on_locked(swp_entry_t entry, spinlock_t *ptl) * If a migration entry exists for the page the migration path must hold * a valid reference to the page, and it must take the ptl to remove the * migration entry. So the page is valid until the ptl is dropped. + * Similarly any path attempting to drop the last reference to a + * device-private page needs to grab the ptl to remove the device-private + * entry. */ spin_unlock(ptl); diff --git a/mm/memory.c b/mm/memory.c index 61748b762876..e43f0a4702c4 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -4642,7 +4642,8 @@ vm_fault_t do_swap_page(struct vm_fault *vmf) unlock_page(vmf->page); put_page(vmf->page); } else { - pte_unmap_unlock(vmf->pte, vmf->ptl); + pte_unmap(vmf->pte); + migration_entry_wait_on_locked(entry, vmf->ptl); } } else if (is_hwpoison_entry(entry)) { ret = VM_FAULT_HWPOISON; -- 2.53.0
