v6: move a fix from patch 3 to patch 2, more locking fixes
v5: somehow a __vma_private_lock(vma) test failed to make it from my tree into the v4 series, fix that
v4: fix unmap_vmas locking issue pointed out by Mike Kravetz, and resulting lockdep fallout
v3: fix compile error w/ lockdep and test case errors with patch 3
v2: fix the locking bug found with the libhugetlbfs tests.
Malloc libraries, like jemalloc and tcalloc, take decisions on when
to call madvise independently from the code in the main application.
This sometimes results in the application page faulting on an address,
right after the malloc library has shot down the backing memory with
MADV_DONTNEED.
Usually this is harmless, because we always have some 4kB pages
sitting around to satisfy a page fault. However, with hugetlbfs
systems often allocate only the exact number of huge pages that
the application wants.
Due to TLB batching, hugetlbfs MADV_DONTNEED will free pages outside of
any lock taken on the page fault path, which can open up the following
race condition:
CPU 1 CPU 2
MADV_DONTNEED
unmap page
shoot down TLB entry
page fault
fail to allocate a huge page
killed with SIGBUS
free page
Fix that race by extending the hugetlb_vma_lock locking scheme to also
cover private hugetlb mappings (with resv_map), and pulling the locking
from __unmap_hugepage_final_range into helper functions called from
zap_page_range_single. This ensures page faults stay locked out of
the MADV_DONTNEED VMA until the huge pages have actually been freed.
The third patch in the series is more of an RFC. Using the
invalidate_lock instead of the hugetlb_vma_lock greatly simplifies
the code, but at the cost of turning a per-VMA lock into a lock
per backing hugetlbfs file, which could slow things down when
multiple processes are mapping the same hugetlbfs file.
From: Rik van Riel <[email protected]>
Extend the locking scheme used to protect shared hugetlb mappings
from truncate vs page fault races, in order to protect private
hugetlb mappings (with resv_map) against MADV_DONTNEED.
Add a read-write semaphore to the resv_map data structure, and
use that from the hugetlb_vma_(un)lock_* functions, in preparation
for closing the race between MADV_DONTNEED and page faults.
Signed-off-by: Rik van Riel <[email protected]>
Reviewed-by: Mike Kravetz <[email protected]>
Cc: [email protected]
Fixes: 04ada095dcfc ("hugetlb: don't delete vma_lock in hugetlb MADV_DONTNEED processing")
---
include/linux/hugetlb.h | 6 ++++++
mm/hugetlb.c | 41 +++++++++++++++++++++++++++++++++++++----
2 files changed, 43 insertions(+), 4 deletions(-)
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 5b2626063f4f..694928fa06a3 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -60,6 +60,7 @@ struct resv_map {
long adds_in_progress;
struct list_head region_cache;
long region_cache_count;
+ struct rw_semaphore rw_sema;
#ifdef CONFIG_CGROUP_HUGETLB
/*
* On private mappings, the counter to uncharge reservations is stored
@@ -1231,6 +1232,11 @@ static inline bool __vma_shareable_lock(struct vm_area_struct *vma)
return (vma->vm_flags & VM_MAYSHARE) && vma->vm_private_data;
}
+static inline bool __vma_private_lock(struct vm_area_struct *vma)
+{
+ return (!(vma->vm_flags & VM_MAYSHARE)) && vma->vm_private_data;
+}
+
/*
* Safe version of huge_pte_offset() to check the locks. See comments
* above huge_pte_offset().
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index ba6d39b71cb1..ee7497f37098 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -97,6 +97,7 @@ static void hugetlb_vma_lock_alloc(struct vm_area_struct *vma);
static void __hugetlb_vma_unlock_write_free(struct vm_area_struct *vma);
static void hugetlb_unshare_pmds(struct vm_area_struct *vma,
unsigned long start, unsigned long end);
+static struct resv_map *vma_resv_map(struct vm_area_struct *vma);
static inline bool subpool_is_free(struct hugepage_subpool *spool)
{
@@ -267,6 +268,10 @@ void hugetlb_vma_lock_read(struct vm_area_struct *vma)
struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
down_read(&vma_lock->rw_sema);
+ } else if (__vma_private_lock(vma)) {
+ struct resv_map *resv_map = vma_resv_map(vma);
+
+ down_read(&resv_map->rw_sema);
}
}
@@ -276,6 +281,10 @@ void hugetlb_vma_unlock_read(struct vm_area_struct *vma)
struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
up_read(&vma_lock->rw_sema);
+ } else if (__vma_private_lock(vma)) {
+ struct resv_map *resv_map = vma_resv_map(vma);
+
+ up_read(&resv_map->rw_sema);
}
}
@@ -285,6 +294,10 @@ void hugetlb_vma_lock_write(struct vm_area_struct *vma)
struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
down_write(&vma_lock->rw_sema);
+ } else if (__vma_private_lock(vma)) {
+ struct resv_map *resv_map = vma_resv_map(vma);
+
+ down_write(&resv_map->rw_sema);
}
}
@@ -294,17 +307,27 @@ void hugetlb_vma_unlock_write(struct vm_area_struct *vma)
struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
up_write(&vma_lock->rw_sema);
+ } else if (__vma_private_lock(vma)) {
+ struct resv_map *resv_map = vma_resv_map(vma);
+
+ up_write(&resv_map->rw_sema);
}
}
int hugetlb_vma_trylock_write(struct vm_area_struct *vma)
{
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
- if (!__vma_shareable_lock(vma))
- return 1;
+ if (__vma_shareable_lock(vma)) {
+ struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
- return down_write_trylock(&vma_lock->rw_sema);
+ return down_write_trylock(&vma_lock->rw_sema);
+ } else if (__vma_private_lock(vma)) {
+ struct resv_map *resv_map = vma_resv_map(vma);
+
+ return down_write_trylock(&resv_map->rw_sema);
+ }
+
+ return 1;
}
void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
@@ -313,6 +336,10 @@ void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
lockdep_assert_held(&vma_lock->rw_sema);
+ } else if (__vma_private_lock(vma)) {
+ struct resv_map *resv_map = vma_resv_map(vma);
+
+ lockdep_assert_held(&resv_map->rw_sema);
}
}
@@ -345,6 +372,11 @@ static void __hugetlb_vma_unlock_write_free(struct vm_area_struct *vma)
struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
__hugetlb_vma_unlock_write_put(vma_lock);
+ } else if (__vma_private_lock(vma)) {
+ struct resv_map *resv_map = vma_resv_map(vma);
+
+ /* no free for anon vmas, but still need to unlock */
+ up_write(&resv_map->rw_sema);
}
}
@@ -1068,6 +1100,7 @@ struct resv_map *resv_map_alloc(void)
kref_init(&resv_map->refs);
spin_lock_init(&resv_map->lock);
INIT_LIST_HEAD(&resv_map->regions);
+ init_rwsem(&resv_map->rw_sema);
resv_map->adds_in_progress = 0;
/*
--
2.41.0
From: Rik van Riel <[email protected]>
Malloc libraries, like jemalloc and tcalloc, take decisions on when
to call madvise independently from the code in the main application.
This sometimes results in the application page faulting on an address,
right after the malloc library has shot down the backing memory with
MADV_DONTNEED.
Usually this is harmless, because we always have some 4kB pages
sitting around to satisfy a page fault. However, with hugetlbfs
systems often allocate only the exact number of huge pages that
the application wants.
Due to TLB batching, hugetlbfs MADV_DONTNEED will free pages outside of
any lock taken on the page fault path, which can open up the following
race condition:
CPU 1 CPU 2
MADV_DONTNEED
unmap page
shoot down TLB entry
page fault
fail to allocate a huge page
killed with SIGBUS
free page
Fix that race by pulling the locking from __unmap_hugepage_final_range
into helper functions called from zap_page_range_single. This ensures
page faults stay locked out of the MADV_DONTNEED VMA until the
huge pages have actually been freed.
Signed-off-by: Rik van Riel <[email protected]>
Cc: [email protected]
Fixes: 04ada095dcfc ("hugetlb: don't delete vma_lock in hugetlb MADV_DONTNEED processing")
---
include/linux/hugetlb.h | 35 +++++++++++++++++++++++++++++++++--
mm/hugetlb.c | 28 ++++++++++++++++------------
mm/memory.c | 13 ++++++++-----
3 files changed, 57 insertions(+), 19 deletions(-)
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 694928fa06a3..d9ec500cfef9 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -139,7 +139,7 @@ struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
void unmap_hugepage_range(struct vm_area_struct *,
unsigned long, unsigned long, struct page *,
zap_flags_t);
-void __unmap_hugepage_range_final(struct mmu_gather *tlb,
+void __unmap_hugepage_range(struct mmu_gather *tlb,
struct vm_area_struct *vma,
unsigned long start, unsigned long end,
struct page *ref_page, zap_flags_t zap_flags);
@@ -246,6 +246,25 @@ int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
unsigned long *start, unsigned long *end);
+extern void __hugetlb_zap_begin(struct vm_area_struct *vma,
+ unsigned long *begin, unsigned long *end);
+extern void __hugetlb_zap_end(struct vm_area_struct *vma,
+ struct zap_details *details);
+
+static inline void hugetlb_zap_begin(struct vm_area_struct *vma,
+ unsigned long *start, unsigned long *end)
+{
+ if (is_vm_hugetlb_page(vma))
+ __hugetlb_zap_begin(vma, start, end);
+}
+
+static inline void hugetlb_zap_end(struct vm_area_struct *vma,
+ struct zap_details *details)
+{
+ if (is_vm_hugetlb_page(vma))
+ __hugetlb_zap_end(vma, details);
+}
+
void hugetlb_vma_lock_read(struct vm_area_struct *vma);
void hugetlb_vma_unlock_read(struct vm_area_struct *vma);
void hugetlb_vma_lock_write(struct vm_area_struct *vma);
@@ -297,6 +316,18 @@ static inline void adjust_range_if_pmd_sharing_possible(
{
}
+static inline void hugetlb_zap_begin(
+ struct vm_area_struct *vma,
+ unsigned long *start, unsigned long *end)
+{
+}
+
+static inline void hugetlb_zap_end(
+ struct vm_area_struct *vma,
+ struct zap_details *details)
+{
+}
+
static inline struct page *hugetlb_follow_page_mask(
struct vm_area_struct *vma, unsigned long address, unsigned int flags,
unsigned int *page_mask)
@@ -442,7 +473,7 @@ static inline long hugetlb_change_protection(
return 0;
}
-static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
+static inline void __unmap_hugepage_range(struct mmu_gather *tlb,
struct vm_area_struct *vma, unsigned long start,
unsigned long end, struct page *ref_page,
zap_flags_t zap_flags)
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index ee7497f37098..424bb8da9519 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -5306,9 +5306,9 @@ int move_hugetlb_page_tables(struct vm_area_struct *vma,
return len + old_addr - old_end;
}
-static void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
- unsigned long start, unsigned long end,
- struct page *ref_page, zap_flags_t zap_flags)
+void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
+ unsigned long start, unsigned long end,
+ struct page *ref_page, zap_flags_t zap_flags)
{
struct mm_struct *mm = vma->vm_mm;
unsigned long address;
@@ -5435,16 +5435,19 @@ static void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct
tlb_flush_mmu_tlbonly(tlb);
}
-void __unmap_hugepage_range_final(struct mmu_gather *tlb,
- struct vm_area_struct *vma, unsigned long start,
- unsigned long end, struct page *ref_page,
- zap_flags_t zap_flags)
+void __hugetlb_zap_begin(struct vm_area_struct *vma,
+ unsigned long *start, unsigned long *end)
{
+ adjust_range_if_pmd_sharing_possible(vma, start, end);
hugetlb_vma_lock_write(vma);
- i_mmap_lock_write(vma->vm_file->f_mapping);
+ if (vma->vm_file)
+ i_mmap_lock_write(vma->vm_file->f_mapping);
+}
- /* mmu notification performed in caller */
- __unmap_hugepage_range(tlb, vma, start, end, ref_page, zap_flags);
+void __hugetlb_zap_end(struct vm_area_struct *vma,
+ struct zap_details *details)
+{
+ zap_flags_t zap_flags = details ? details->zap_flags : 0;
if (zap_flags & ZAP_FLAG_UNMAP) { /* final unmap */
/*
@@ -5457,11 +5460,12 @@ void __unmap_hugepage_range_final(struct mmu_gather *tlb,
* someone else.
*/
__hugetlb_vma_unlock_write_free(vma);
- i_mmap_unlock_write(vma->vm_file->f_mapping);
} else {
- i_mmap_unlock_write(vma->vm_file->f_mapping);
hugetlb_vma_unlock_write(vma);
}
+
+ if (vma->vm_file)
+ i_mmap_unlock_write(vma->vm_file->f_mapping);
}
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
diff --git a/mm/memory.c b/mm/memory.c
index 6c264d2f969c..517221f01303 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1683,7 +1683,7 @@ static void unmap_single_vma(struct mmu_gather *tlb,
if (vma->vm_file) {
zap_flags_t zap_flags = details ?
details->zap_flags : 0;
- __unmap_hugepage_range_final(tlb, vma, start, end,
+ __unmap_hugepage_range(tlb, vma, start, end,
NULL, zap_flags);
}
} else
@@ -1728,8 +1728,12 @@ void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
start_addr, end_addr);
mmu_notifier_invalidate_range_start(&range);
do {
- unmap_single_vma(tlb, vma, start_addr, end_addr, &details,
+ unsigned long start = start_addr;
+ unsigned long end = end_addr;
+ hugetlb_zap_begin(vma, &start, &end);
+ unmap_single_vma(tlb, vma, start, end, &details,
mm_wr_locked);
+ hugetlb_zap_end(vma, &details);
} while ((vma = mas_find(mas, tree_end - 1)) != NULL);
mmu_notifier_invalidate_range_end(&range);
}
@@ -1753,9 +1757,7 @@ void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
lru_add_drain();
mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma->vm_mm,
address, end);
- if (is_vm_hugetlb_page(vma))
- adjust_range_if_pmd_sharing_possible(vma, &range.start,
- &range.end);
+ hugetlb_zap_begin(vma, &range.start, &range.end);
tlb_gather_mmu(&tlb, vma->vm_mm);
update_hiwater_rss(vma->vm_mm);
mmu_notifier_invalidate_range_start(&range);
@@ -1766,6 +1768,7 @@ void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
unmap_single_vma(&tlb, vma, address, end, details, false);
mmu_notifier_invalidate_range_end(&range);
tlb_finish_mmu(&tlb);
+ hugetlb_zap_end(vma, details);
}
/**
--
2.41.0
From: Rik van Riel <[email protected]>
Replace the custom hugetlbfs VMA locking code with the recently
introduced invalidate_lock. This greatly simplifies things.
However, this is a large enough change that it should probably go in
separately from the other changes.
Another question is whether this simplification hurts scalability
for certain workloads.
Suggested-by: Matthew Wilcox <[email protected]>
Signed-off-by: Rik van Riel <[email protected]>
---
fs/hugetlbfs/inode.c | 70 ++----------
include/linux/fs.h | 6 +
include/linux/hugetlb.h | 21 +---
mm/hugetlb.c | 237 ++++------------------------------------
4 files changed, 35 insertions(+), 299 deletions(-)
diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index 316c4cebd3f3..d1aed4590b3d 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -485,7 +485,6 @@ static void hugetlb_unmap_file_folio(struct hstate *h,
struct folio *folio, pgoff_t index)
{
struct rb_root_cached *root = &mapping->i_mmap;
- struct hugetlb_vma_lock *vma_lock;
struct page *page = &folio->page;
struct vm_area_struct *vma;
unsigned long v_start;
@@ -495,9 +494,9 @@ static void hugetlb_unmap_file_folio(struct hstate *h,
start = index * pages_per_huge_page(h);
end = (index + 1) * pages_per_huge_page(h);
+ filemap_invalidate_lock(mapping);
i_mmap_lock_write(mapping);
-retry:
- vma_lock = NULL;
+
vma_interval_tree_foreach(vma, root, start, end - 1) {
v_start = vma_offset_start(vma, start);
v_end = vma_offset_end(vma, end);
@@ -505,62 +504,13 @@ static void hugetlb_unmap_file_folio(struct hstate *h,
if (!hugetlb_vma_maps_page(vma, v_start, page))
continue;
- if (!hugetlb_vma_trylock_write(vma)) {
- vma_lock = vma->vm_private_data;
- /*
- * If we can not get vma lock, we need to drop
- * immap_sema and take locks in order. First,
- * take a ref on the vma_lock structure so that
- * we can be guaranteed it will not go away when
- * dropping immap_sema.
- */
- kref_get(&vma_lock->refs);
- break;
- }
-
unmap_hugepage_range(vma, v_start, v_end, NULL,
ZAP_FLAG_DROP_MARKER);
hugetlb_vma_unlock_write(vma);
}
+ filemap_invalidate_unlock(mapping);
i_mmap_unlock_write(mapping);
-
- if (vma_lock) {
- /*
- * Wait on vma_lock. We know it is still valid as we have
- * a reference. We must 'open code' vma locking as we do
- * not know if vma_lock is still attached to vma.
- */
- down_write(&vma_lock->rw_sema);
- i_mmap_lock_write(mapping);
-
- vma = vma_lock->vma;
- if (!vma) {
- /*
- * If lock is no longer attached to vma, then just
- * unlock, drop our reference and retry looking for
- * other vmas.
- */
- up_write(&vma_lock->rw_sema);
- kref_put(&vma_lock->refs, hugetlb_vma_lock_release);
- goto retry;
- }
-
- /*
- * vma_lock is still attached to vma. Check to see if vma
- * still maps page and if so, unmap.
- */
- v_start = vma_offset_start(vma, start);
- v_end = vma_offset_end(vma, end);
- if (hugetlb_vma_maps_page(vma, v_start, page))
- unmap_hugepage_range(vma, v_start, v_end, NULL,
- ZAP_FLAG_DROP_MARKER);
-
- kref_put(&vma_lock->refs, hugetlb_vma_lock_release);
- hugetlb_vma_unlock_write(vma);
-
- goto retry;
- }
}
static void
@@ -578,20 +528,10 @@ hugetlb_vmdelete_list(struct rb_root_cached *root, pgoff_t start, pgoff_t end,
unsigned long v_start;
unsigned long v_end;
- if (!hugetlb_vma_trylock_write(vma))
- continue;
-
v_start = vma_offset_start(vma, start);
v_end = vma_offset_end(vma, end);
unmap_hugepage_range(vma, v_start, v_end, NULL, zap_flags);
-
- /*
- * Note that vma lock only exists for shared/non-private
- * vmas. Therefore, lock is not held when calling
- * unmap_hugepage_range for private vmas.
- */
- hugetlb_vma_unlock_write(vma);
}
}
@@ -725,10 +665,12 @@ static void hugetlb_vmtruncate(struct inode *inode, loff_t offset)
pgoff = offset >> PAGE_SHIFT;
i_size_write(inode, offset);
+ filemap_invalidate_lock(mapping);
i_mmap_lock_write(mapping);
if (!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root))
hugetlb_vmdelete_list(&mapping->i_mmap, pgoff, 0,
ZAP_FLAG_DROP_MARKER);
+ filemap_invalidate_unlock(mapping);
i_mmap_unlock_write(mapping);
remove_inode_hugepages(inode, offset, LLONG_MAX);
}
@@ -778,6 +720,7 @@ static long hugetlbfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
return -EPERM;
}
+ filemap_invalidate_lock(mapping);
i_mmap_lock_write(mapping);
/* If range starts before first full page, zero partial page. */
@@ -799,6 +742,7 @@ static long hugetlbfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
hole_end, offset + len);
i_mmap_unlock_write(mapping);
+ filemap_invalidate_unlock(mapping);
/* Remove full pages from the file. */
if (hole_end > hole_start)
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 4aeb3fa11927..b455a8913db4 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -847,6 +847,12 @@ static inline void filemap_invalidate_lock(struct address_space *mapping)
down_write(&mapping->invalidate_lock);
}
+static inline int filemap_invalidate_trylock(
+ struct address_space *mapping)
+{
+ return down_write_trylock(&mapping->invalidate_lock);
+}
+
static inline void filemap_invalidate_unlock(struct address_space *mapping)
{
up_write(&mapping->invalidate_lock);
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index d9ec500cfef9..2908c47e7bf2 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -60,7 +60,6 @@ struct resv_map {
long adds_in_progress;
struct list_head region_cache;
long region_cache_count;
- struct rw_semaphore rw_sema;
#ifdef CONFIG_CGROUP_HUGETLB
/*
* On private mappings, the counter to uncharge reservations is stored
@@ -107,12 +106,6 @@ struct file_region {
#endif
};
-struct hugetlb_vma_lock {
- struct kref refs;
- struct rw_semaphore rw_sema;
- struct vm_area_struct *vma;
-};
-
extern struct resv_map *resv_map_alloc(void);
void resv_map_release(struct kref *ref);
@@ -1277,17 +1270,9 @@ hugetlb_walk(struct vm_area_struct *vma, unsigned long addr, unsigned long sz)
{
#if defined(CONFIG_HUGETLB_PAGE) && \
defined(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) && defined(CONFIG_LOCKDEP)
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- /*
- * If pmd sharing possible, locking needed to safely walk the
- * hugetlb pgtables. More information can be found at the comment
- * above huge_pte_offset() in the same file.
- *
- * NOTE: lockdep_is_held() is only defined with CONFIG_LOCKDEP.
- */
- if (__vma_shareable_lock(vma))
- WARN_ON_ONCE(!lockdep_is_held(&vma_lock->rw_sema) &&
+ if (vma->vm_file)
+ WARN_ON_ONCE(!lockdep_is_held(
+ &vma->vm_file->f_mapping->invalidate_lock) &&
!lockdep_is_held(
&vma->vm_file->f_mapping->i_mmap_rwsem));
#endif
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 424bb8da9519..a942ea3579f8 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -92,9 +92,6 @@ struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);
-static void hugetlb_vma_lock_free(struct vm_area_struct *vma);
-static void hugetlb_vma_lock_alloc(struct vm_area_struct *vma);
-static void __hugetlb_vma_unlock_write_free(struct vm_area_struct *vma);
static void hugetlb_unshare_pmds(struct vm_area_struct *vma,
unsigned long start, unsigned long end);
static struct resv_map *vma_resv_map(struct vm_area_struct *vma);
@@ -264,170 +261,41 @@ static inline struct hugepage_subpool *subpool_vma(struct vm_area_struct *vma)
*/
void hugetlb_vma_lock_read(struct vm_area_struct *vma)
{
- if (__vma_shareable_lock(vma)) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- down_read(&vma_lock->rw_sema);
- } else if (__vma_private_lock(vma)) {
- struct resv_map *resv_map = vma_resv_map(vma);
-
- down_read(&resv_map->rw_sema);
- }
+ if (vma->vm_file)
+ filemap_invalidate_lock_shared(vma->vm_file->f_mapping);
}
void hugetlb_vma_unlock_read(struct vm_area_struct *vma)
{
- if (__vma_shareable_lock(vma)) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- up_read(&vma_lock->rw_sema);
- } else if (__vma_private_lock(vma)) {
- struct resv_map *resv_map = vma_resv_map(vma);
-
- up_read(&resv_map->rw_sema);
- }
+ if (vma->vm_file)
+ filemap_invalidate_unlock_shared(vma->vm_file->f_mapping);
}
void hugetlb_vma_lock_write(struct vm_area_struct *vma)
{
- if (__vma_shareable_lock(vma)) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- down_write(&vma_lock->rw_sema);
- } else if (__vma_private_lock(vma)) {
- struct resv_map *resv_map = vma_resv_map(vma);
-
- down_write(&resv_map->rw_sema);
- }
+ if (vma->vm_file)
+ filemap_invalidate_lock(vma->vm_file->f_mapping);
}
void hugetlb_vma_unlock_write(struct vm_area_struct *vma)
{
- if (__vma_shareable_lock(vma)) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- up_write(&vma_lock->rw_sema);
- } else if (__vma_private_lock(vma)) {
- struct resv_map *resv_map = vma_resv_map(vma);
-
- up_write(&resv_map->rw_sema);
- }
+ if (vma->vm_file)
+ filemap_invalidate_unlock(vma->vm_file->f_mapping);
}
int hugetlb_vma_trylock_write(struct vm_area_struct *vma)
{
- if (__vma_shareable_lock(vma)) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- return down_write_trylock(&vma_lock->rw_sema);
- } else if (__vma_private_lock(vma)) {
- struct resv_map *resv_map = vma_resv_map(vma);
-
- return down_write_trylock(&resv_map->rw_sema);
- }
+ if (vma->vm_file)
+ return filemap_invalidate_trylock(vma->vm_file->f_mapping);
return 1;
}
void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
{
- if (__vma_shareable_lock(vma)) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- lockdep_assert_held(&vma_lock->rw_sema);
- } else if (__vma_private_lock(vma)) {
- struct resv_map *resv_map = vma_resv_map(vma);
-
- lockdep_assert_held(&resv_map->rw_sema);
- }
-}
-
-void hugetlb_vma_lock_release(struct kref *kref)
-{
- struct hugetlb_vma_lock *vma_lock = container_of(kref,
- struct hugetlb_vma_lock, refs);
-
- kfree(vma_lock);
-}
-
-static void __hugetlb_vma_unlock_write_put(struct hugetlb_vma_lock *vma_lock)
-{
- struct vm_area_struct *vma = vma_lock->vma;
-
- /*
- * vma_lock structure may or not be released as a result of put,
- * it certainly will no longer be attached to vma so clear pointer.
- * Semaphore synchronizes access to vma_lock->vma field.
- */
- vma_lock->vma = NULL;
- vma->vm_private_data = NULL;
- up_write(&vma_lock->rw_sema);
- kref_put(&vma_lock->refs, hugetlb_vma_lock_release);
-}
-
-static void __hugetlb_vma_unlock_write_free(struct vm_area_struct *vma)
-{
- if (__vma_shareable_lock(vma)) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- __hugetlb_vma_unlock_write_put(vma_lock);
- } else if (__vma_private_lock(vma)) {
- struct resv_map *resv_map = vma_resv_map(vma);
-
- /* no free for anon vmas, but still need to unlock */
- up_write(&resv_map->rw_sema);
- }
-}
-
-static void hugetlb_vma_lock_free(struct vm_area_struct *vma)
-{
- /*
- * Only present in sharable vmas.
- */
- if (!vma || !__vma_shareable_lock(vma))
- return;
-
- if (vma->vm_private_data) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- down_write(&vma_lock->rw_sema);
- __hugetlb_vma_unlock_write_put(vma_lock);
- }
-}
-
-static void hugetlb_vma_lock_alloc(struct vm_area_struct *vma)
-{
- struct hugetlb_vma_lock *vma_lock;
-
- /* Only establish in (flags) sharable vmas */
- if (!vma || !(vma->vm_flags & VM_MAYSHARE))
- return;
-
- /* Should never get here with non-NULL vm_private_data */
- if (vma->vm_private_data)
- return;
-
- vma_lock = kmalloc(sizeof(*vma_lock), GFP_KERNEL);
- if (!vma_lock) {
- /*
- * If we can not allocate structure, then vma can not
- * participate in pmd sharing. This is only a possible
- * performance enhancement and memory saving issue.
- * However, the lock is also used to synchronize page
- * faults with truncation. If the lock is not present,
- * unlikely races could leave pages in a file past i_size
- * until the file is removed. Warn in the unlikely case of
- * allocation failure.
- */
- pr_warn_once("HugeTLB: unable to allocate vma specific lock\n");
- return;
- }
-
- kref_init(&vma_lock->refs);
- init_rwsem(&vma_lock->rw_sema);
- vma_lock->vma = vma;
- vma->vm_private_data = vma_lock;
+ if (vma->vm_file)
+ lockdep_assert_held(&vma->vm_file->f_mapping->invalidate_lock);
}
/* Helper that removes a struct file_region from the resv_map cache and returns
@@ -1100,7 +968,6 @@ struct resv_map *resv_map_alloc(void)
kref_init(&resv_map->refs);
spin_lock_init(&resv_map->lock);
INIT_LIST_HEAD(&resv_map->regions);
- init_rwsem(&resv_map->rw_sema);
resv_map->adds_in_progress = 0;
/*
@@ -1195,22 +1062,11 @@ void hugetlb_dup_vma_private(struct vm_area_struct *vma)
VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
/*
* Clear vm_private_data
- * - For shared mappings this is a per-vma semaphore that may be
- * allocated in a subsequent call to hugetlb_vm_op_open.
- * Before clearing, make sure pointer is not associated with vma
- * as this will leak the structure. This is the case when called
- * via clear_vma_resv_huge_pages() and hugetlb_vm_op_open has already
- * been called to allocate a new structure.
* - For MAP_PRIVATE mappings, this is the reserve map which does
* not apply to children. Faults generated by the children are
* not guaranteed to succeed, even if read-only.
*/
- if (vma->vm_flags & VM_MAYSHARE) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- if (vma_lock && vma_lock->vma != vma)
- vma->vm_private_data = NULL;
- } else
+ if (!(vma->vm_flags & VM_MAYSHARE))
vma->vm_private_data = NULL;
}
@@ -4846,25 +4702,6 @@ static void hugetlb_vm_op_open(struct vm_area_struct *vma)
resv_map_dup_hugetlb_cgroup_uncharge_info(resv);
kref_get(&resv->refs);
}
-
- /*
- * vma_lock structure for sharable mappings is vma specific.
- * Clear old pointer (if copied via vm_area_dup) and allocate
- * new structure. Before clearing, make sure vma_lock is not
- * for this vma.
- */
- if (vma->vm_flags & VM_MAYSHARE) {
- struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
-
- if (vma_lock) {
- if (vma_lock->vma != vma) {
- vma->vm_private_data = NULL;
- hugetlb_vma_lock_alloc(vma);
- } else
- pr_warn("HugeTLB: vma_lock already exists in %s.\n", __func__);
- } else
- hugetlb_vma_lock_alloc(vma);
- }
}
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
@@ -4875,8 +4712,6 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)
unsigned long reserve, start, end;
long gbl_reserve;
- hugetlb_vma_lock_free(vma);
-
resv = vma_resv_map(vma);
if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
return;
@@ -5048,16 +4883,10 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
mmu_notifier_invalidate_range_start(&range);
vma_assert_write_locked(src_vma);
raw_write_seqcount_begin(&src->write_protect_seq);
- } else {
- /*
- * For shared mappings the vma lock must be held before
- * calling hugetlb_walk() in the src vma. Otherwise, the
- * returned ptep could go away if part of a shared pmd and
- * another thread calls huge_pmd_unshare.
- */
- hugetlb_vma_lock_read(src_vma);
}
+ hugetlb_vma_lock_read(src_vma);
+
last_addr_mask = hugetlb_mask_last_page(h);
for (addr = src_vma->vm_start; addr < src_vma->vm_end; addr += sz) {
spinlock_t *src_ptl, *dst_ptl;
@@ -5209,10 +5038,10 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
if (cow) {
raw_write_seqcount_end(&src->write_protect_seq);
mmu_notifier_invalidate_range_end(&range);
- } else {
- hugetlb_vma_unlock_read(src_vma);
}
+ hugetlb_vma_unlock_read(src_vma);
+
return ret;
}
@@ -5447,25 +5276,9 @@ void __hugetlb_zap_begin(struct vm_area_struct *vma,
void __hugetlb_zap_end(struct vm_area_struct *vma,
struct zap_details *details)
{
- zap_flags_t zap_flags = details ? details->zap_flags : 0;
-
- if (zap_flags & ZAP_FLAG_UNMAP) { /* final unmap */
- /*
- * Unlock and free the vma lock before releasing i_mmap_rwsem.
- * When the vma_lock is freed, this makes the vma ineligible
- * for pmd sharing. And, i_mmap_rwsem is required to set up
- * pmd sharing. This is important as page tables for this
- * unmapped range will be asynchrously deleted. If the page
- * tables are shared, there will be issues when accessed by
- * someone else.
- */
- __hugetlb_vma_unlock_write_free(vma);
- } else {
- hugetlb_vma_unlock_write(vma);
- }
-
if (vma->vm_file)
i_mmap_unlock_write(vma->vm_file->f_mapping);
+ hugetlb_vma_unlock_write(vma);
}
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
@@ -6708,12 +6521,6 @@ bool hugetlb_reserve_pages(struct inode *inode,
return false;
}
- /*
- * vma specific semaphore used for pmd sharing and fault/truncation
- * synchronization
- */
- hugetlb_vma_lock_alloc(vma);
-
/*
* Only apply hugepage reservation if asked. At fault time, an
* attempt will be made for VM_NORESERVE to allocate a page
@@ -6836,7 +6643,6 @@ bool hugetlb_reserve_pages(struct inode *inode,
hugetlb_cgroup_uncharge_cgroup_rsvd(hstate_index(h),
chg * pages_per_huge_page(h), h_cg);
out_err:
- hugetlb_vma_lock_free(vma);
if (!vma || vma->vm_flags & VM_MAYSHARE)
/* Only call region_abort if the region_chg succeeded but the
* region_add failed or didn't run.
@@ -6906,13 +6712,10 @@ static unsigned long page_table_shareable(struct vm_area_struct *svma,
/*
* match the virtual addresses, permission and the alignment of the
* page table page.
- *
- * Also, vma_lock (vm_private_data) is required for sharing.
*/
if (pmd_index(addr) != pmd_index(saddr) ||
vm_flags != svm_flags ||
- !range_in_vma(svma, sbase, s_end) ||
- !svma->vm_private_data)
+ !range_in_vma(svma, sbase, s_end))
return 0;
return saddr;
@@ -6932,8 +6735,6 @@ bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr)
*/
if (!(vma->vm_flags & VM_MAYSHARE))
return false;
- if (!vma->vm_private_data) /* vma lock required for sharing */
- return false;
if (!range_in_vma(vma, start, end))
return false;
return true;
--
2.41.0
On Wed, 2023-10-04 at 20:19 -0700, Mike Kravetz wrote:
> On 10/03/23 23:25, [email protected] wrote:
> >
> > @@ -5457,11 +5460,12 @@ void __unmap_hugepage_range_final(struct
> > mmu_gather *tlb,
> > * someone else.
> > */
> > __hugetlb_vma_unlock_write_free(vma);
> > - i_mmap_unlock_write(vma->vm_file->f_mapping);
> > } else {
> > - i_mmap_unlock_write(vma->vm_file->f_mapping);
> > hugetlb_vma_unlock_write(vma);
> > }
> > +
> > + if (vma->vm_file)
> > + i_mmap_unlock_write(vma->vm_file->f_mapping);
> > }
>
> In the case of a mmap(hugetlbfs_file_mmap) error, the per-vma hugetlb
> lock will not be setup. The hugetlb_vma_lock/unlock routines do not
> check for this as they were previously always called after the lock
> was
> set up. So, we can now get:
Wait, the hugetlb_vma_(un)lock_{read,write} functions do
have checks for the presence of the lock:
void hugetlb_vma_lock_read(struct vm_area_struct *vma)
{
if (__vma_shareable_lock(vma)) {
struct hugetlb_vma_lock *vma_lock = vma-
>vm_private_data;
down_read(&vma_lock->rw_sema);
} else if (__vma_private_lock(vma)) {
struct resv_map *resv_map = vma_resv_map(vma);
down_read(&resv_map->rw_sema);
}
}
Both __vma_shareable_lock and __vma_private_lock check that
vma->vm_private_data points at something.
Exactly what corner case am I missing here?
What leaves vma->vm_private_data pointing at something
invalid?
>
> +++ b/mm/hugetlb.c
> @@ -5503,10 +5503,12 @@ void __unmap_hugepage_range(struct mmu_gather
> *tlb, struct vm_area_struct *vma,
> void __hugetlb_zap_begin(struct vm_area_struct *vma,
> unsigned long *start, unsigned long *end)
> {
> + if (!vma->vm_file) /* hugetlbfs_file_mmap error */
> + return;
> +
This does not seem quite correct, because the locking is needed to
avoid the race between MADV_DONTNEED and the page fault path.
> Another way to resolve would be to fix up the hugetlb_vma_lock/unlock
> routines
> to check for and handle a null lock.
I thought I had that already.
Does __vma_shareable_lock need to check for !vma->vm_file ?
--
All Rights Reversed.
On 10/03/23 23:25, [email protected] wrote:
> From: Rik van Riel <[email protected]>
>
> Malloc libraries, like jemalloc and tcalloc, take decisions on when
> to call madvise independently from the code in the main application.
>
> This sometimes results in the application page faulting on an address,
> right after the malloc library has shot down the backing memory with
> MADV_DONTNEED.
>
> Usually this is harmless, because we always have some 4kB pages
> sitting around to satisfy a page fault. However, with hugetlbfs
> systems often allocate only the exact number of huge pages that
> the application wants.
>
> Due to TLB batching, hugetlbfs MADV_DONTNEED will free pages outside of
> any lock taken on the page fault path, which can open up the following
> race condition:
>
> CPU 1 CPU 2
>
> MADV_DONTNEED
> unmap page
> shoot down TLB entry
> page fault
> fail to allocate a huge page
> killed with SIGBUS
> free page
>
> Fix that race by pulling the locking from __unmap_hugepage_final_range
> into helper functions called from zap_page_range_single. This ensures
> page faults stay locked out of the MADV_DONTNEED VMA until the
> huge pages have actually been freed.
>
> Signed-off-by: Rik van Riel <[email protected]>
> Cc: [email protected]
> Fixes: 04ada095dcfc ("hugetlb: don't delete vma_lock in hugetlb MADV_DONTNEED processing")
> ---
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index ee7497f37098..424bb8da9519 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -5435,16 +5435,19 @@ static void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct
> tlb_flush_mmu_tlbonly(tlb);
> }
>
> -void __unmap_hugepage_range_final(struct mmu_gather *tlb,
> - struct vm_area_struct *vma, unsigned long start,
> - unsigned long end, struct page *ref_page,
> - zap_flags_t zap_flags)
> +void __hugetlb_zap_begin(struct vm_area_struct *vma,
> + unsigned long *start, unsigned long *end)
> {
> + adjust_range_if_pmd_sharing_possible(vma, start, end);
> hugetlb_vma_lock_write(vma);
> - i_mmap_lock_write(vma->vm_file->f_mapping);
> + if (vma->vm_file)
> + i_mmap_lock_write(vma->vm_file->f_mapping);
> +}
>
> - /* mmu notification performed in caller */
> - __unmap_hugepage_range(tlb, vma, start, end, ref_page, zap_flags);
> +void __hugetlb_zap_end(struct vm_area_struct *vma,
> + struct zap_details *details)
> +{
> + zap_flags_t zap_flags = details ? details->zap_flags : 0;
>
> if (zap_flags & ZAP_FLAG_UNMAP) { /* final unmap */
> /*
> @@ -5457,11 +5460,12 @@ void __unmap_hugepage_range_final(struct mmu_gather *tlb,
> * someone else.
> */
> __hugetlb_vma_unlock_write_free(vma);
> - i_mmap_unlock_write(vma->vm_file->f_mapping);
> } else {
> - i_mmap_unlock_write(vma->vm_file->f_mapping);
> hugetlb_vma_unlock_write(vma);
> }
> +
> + if (vma->vm_file)
> + i_mmap_unlock_write(vma->vm_file->f_mapping);
> }
In the case of a mmap(hugetlbfs_file_mmap) error, the per-vma hugetlb
lock will not be setup. The hugetlb_vma_lock/unlock routines do not
check for this as they were previously always called after the lock was
set up. So, we can now get:
[ 47.653806] BUG: kernel NULL pointer dereference, address: 00000000000000c8
[ 47.654967] #PF: supervisor read access in kernel mode
[ 47.655900] #PF: error_code(0x0000) - not-present page
[ 47.656814] PGD 8000000307415067 P4D 8000000307415067 PUD 30587b067 PMD 0
[ 47.658005] Oops: 0000 [#1] PREEMPT SMP PTI
[ 47.658777] CPU: 3 PID: 1224 Comm: heap-overflow Tainted: G W 6.6.0-rc3-next-20230925+ #19
[ 47.660428] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.2-1.fc37 04/01/2014
[ 47.661931] RIP: 0010:__lock_acquire+0x1e6/0x2390
[ 47.662784] Code: 46 24 4c 89 e8 25 ff 1f 00 00 48 0f a3 05 f2 84 0f 02 0f 83 e9 05 00 00 48 8d 14 40 48 8d 04 90 48 c1 e0 04 48 05 a0 89 27 83 <0f> b6 98 c8 00 00 00 41 0f b7 46 20 66 25 ff 1f 0f b7 c0 48 0f a3
[ 47.665890] RSP: 0018:ffffc90004a03ac8 EFLAGS: 00010046
[ 47.667009] RAX: 0000000000000000 RBX: 000000000004138c RCX: 0000000000000000
[ 47.668321] RDX: 0000000000000002 RSI: ffffffff8246ce26 RDI: 00000000ffffffff
[ 47.669580] RBP: 0000000000000000 R08: 0000000000009ffb R09: 0000000000000001
[ 47.670825] R10: ffff888303c51ac0 R11: ffff888303c52430 R12: 0000000000000001
[ 47.672070] R13: 3b97e6ab9880538c R14: ffff888303c52458 R15: 0000000000000000
[ 47.673285] FS: 00007f3065e7a0c0(0000) GS:ffff888477d00000(0000) knlGS:0000000000000000
[ 47.675504] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 47.676646] CR2: 00000000000000c8 CR3: 000000030409a004 CR4: 0000000000370ef0
[ 47.677975] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 47.679264] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 47.680603] Call Trace:
[ 47.681196] <TASK>
[ 47.681723] ? __die+0x1f/0x70
[ 47.682440] ? page_fault_oops+0x159/0x450
[ 47.683246] ? do_user_addr_fault+0x65/0x850
[ 47.684082] ? exc_page_fault+0x6d/0x1c0
[ 47.684838] ? asm_exc_page_fault+0x22/0x30
[ 47.685611] ? __lock_acquire+0x1e6/0x2390
[ 47.686360] ? __lock_acquire+0xab9/0x2390
[ 47.687123] lock_acquire+0xd4/0x2c0
[ 47.687811] ? __hugetlb_zap_begin+0x6e/0xa0
[ 47.688595] ? mark_held_locks+0x49/0x80
[ 47.689321] down_write+0x2a/0xc0
[ 47.689976] ? __hugetlb_zap_begin+0x6e/0xa0
[ 47.690862] __hugetlb_zap_begin+0x6e/0xa0
[ 47.691707] unmap_vmas+0xb3/0x100
[ 47.692480] unmap_region.constprop.0+0xcc/0x140
[ 47.693518] ? lock_release+0x142/0x290
[ 47.694304] ? preempt_count_add+0x47/0xa0
[ 47.695109] mmap_region+0x565/0xab0
[ 47.695831] do_mmap+0x35a/0x520
[ 47.696511] vm_mmap_pgoff+0xdf/0x200
[ 47.697419] ksys_mmap_pgoff+0xdb/0x200
[ 47.698368] do_syscall_64+0x37/0x90
[ 47.699148] entry_SYSCALL_64_after_hwframe+0x6e/0xd8
[ 47.700307] RIP: 0033:0x7f3065f77086
In my environment, I added the following to this patch to resolve the
issue.
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index d25db18c9526..48370f5b70f5 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -5503,10 +5503,12 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
void __hugetlb_zap_begin(struct vm_area_struct *vma,
unsigned long *start, unsigned long *end)
{
+ if (!vma->vm_file) /* hugetlbfs_file_mmap error */
+ return;
+
adjust_range_if_pmd_sharing_possible(vma, start, end);
hugetlb_vma_lock_write(vma);
- if (vma->vm_file)
- i_mmap_lock_write(vma->vm_file->f_mapping);
+ i_mmap_lock_write(vma->vm_file->f_mapping);
}
void __hugetlb_zap_end(struct vm_area_struct *vma,
@@ -5514,6 +5516,9 @@ void __hugetlb_zap_end(struct vm_area_struct *vma,
{
zap_flags_t zap_flags = details ? details->zap_flags : 0;
+ if (!vma->vm_file) /* hugetlbfs_file_mmap mmap error */
+ return;
+
if (zap_flags & ZAP_FLAG_UNMAP) { /* final unmap */
/*
* Unlock and free the vma lock before releasing i_mmap_rwsem.
@@ -5529,8 +5534,7 @@ void __hugetlb_zap_end(struct vm_area_struct *vma,
hugetlb_vma_unlock_write(vma);
}
- if (vma->vm_file)
- i_mmap_unlock_write(vma->vm_file->f_mapping);
+ i_mmap_unlock_write(vma->vm_file->f_mapping);
}
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
Another way to resolve would be to fix up the hugetlb_vma_lock/unlock routines
to check for and handle a null lock.
--
Mike Kravetz
On 10/05/23 09:23, Rik van Riel wrote:
> On Wed, 2023-10-04 at 20:19 -0700, Mike Kravetz wrote:
> > On 10/03/23 23:25, [email protected]?wrote:
> > >
> > > @@ -5457,11 +5460,12 @@ void __unmap_hugepage_range_final(struct
> > > mmu_gather *tlb,
> > > ???????????????? * someone else.
> > > ???????????????? */
> > > ????????????????__hugetlb_vma_unlock_write_free(vma);
> > > -???????????????i_mmap_unlock_write(vma->vm_file->f_mapping);
> > > ????????} else {
> > > -???????????????i_mmap_unlock_write(vma->vm_file->f_mapping);
> > > ????????????????hugetlb_vma_unlock_write(vma);
> > > ????????}
> > > +
> > > +???????if (vma->vm_file)
> > > +???????????????i_mmap_unlock_write(vma->vm_file->f_mapping);
> > > ?}
> >
> > In the case of a mmap(hugetlbfs_file_mmap) error, the per-vma hugetlb
> > lock will not be setup.? The hugetlb_vma_lock/unlock routines do not
> > check for this as they were previously always called after the lock
> > was
> > set up.? So, we can now get:
>
> Wait, the hugetlb_vma_(un)lock_{read,write} functions do
> have checks for the presence of the lock:
>
> void hugetlb_vma_lock_read(struct vm_area_struct *vma)
> {
> if (__vma_shareable_lock(vma)) {
> struct hugetlb_vma_lock *vma_lock = vma-
> >vm_private_data;
>
> down_read(&vma_lock->rw_sema);
> } else if (__vma_private_lock(vma)) {
> struct resv_map *resv_map = vma_resv_map(vma);
>
> down_read(&resv_map->rw_sema);
> }
> }
>
> Both __vma_shareable_lock and __vma_private_lock check that
> vma->vm_private_data points at something.
>
> Exactly what corner case am I missing here?
>
> What leaves vma->vm_private_data pointing at something
> invalid?
You are correct. The checks in hugetlb_vma_(un)lock_{read,write} functions
should be sufficient.
Here is the corner case that needs to be addressed:
mmap
hugetlbfs_file_mmap
hugetlb_reserve_pages {
!VM_MAYSHARE so
resv_map = resv_map_alloc();
set_vma_resv_map(vma, resv_map);
set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
...
some error in hugetlb_reserve_pages,
goto out_err
out_error:
if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
kref_put(&resv_map->refs, resv_map_release);
return false;
Note that we free resv_map but do not clear vm_private_data. So, at
unmap time we try to use the lock in the freed resv_map.
Leaving the dangling pointer to a freed structure is BAD. However, no
code accessed the freed structure until this patch.
I would suggest incorporating this into this patch, or even as a stand
alone patch before this series.
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index d25db18c9526..a3ae13d0f8fe 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1171,8 +1171,7 @@ static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
- set_vma_private_data(vma, (get_vma_private_data(vma) &
- HPAGE_RESV_MASK) | (unsigned long)map);
+ set_vma_private_data(vma, (unsigned long)map);
}
static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
@@ -6910,8 +6909,10 @@ bool hugetlb_reserve_pages(struct inode *inode,
*/
if (chg >= 0 && add < 0)
region_abort(resv_map, from, to, regions_needed);
- if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
+ if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
kref_put(&resv_map->refs, resv_map_release);
+ set_vma_resv_map(vma, NULL);
+ }
return false;
}
At one time set_vma_resv_map must have wanted to preserve flags.
However, that is no longer the case so we can remove that code.
Of course, feel free to address some other way if you would like.
> > +++ b/mm/hugetlb.c
> > @@ -5503,10 +5503,12 @@ void __unmap_hugepage_range(struct mmu_gather
> > *tlb, struct vm_area_struct *vma,
> > ?void __hugetlb_zap_begin(struct vm_area_struct *vma,
> > ???????????????????????? unsigned long *start, unsigned long *end)
> > ?{
> > +???????if (!vma->vm_file)??????/* hugetlbfs_file_mmap error */
> > +???????????????return;
> > +
>
> This does not seem quite correct, because the locking is needed to
> avoid the race between MADV_DONTNEED and the page fault path.
>
>
Note that vma->vm_file is always set for hugetlb vmas except in the mmap
error case. The first line of code in hugetlb_fault is:
mapping = vma->vm_file->f_mapping;
So, bailing out early like this should not be an issue.
> > Another way to resolve would be to fix up the hugetlb_vma_lock/unlock
> > routines
> > to check for and handle a null lock.
>
> I thought I had that already.
>
You did. My bad! I saw the NULL deref and jumped to conclusions.
> Does __vma_shareable_lock need to check for !vma->vm_file ?
I do not think there is a need. It has the check for vma->vm_private_data
which should be sufficient. We only got in trouble in the !VM_MAYSHARE
case because of the stale/invalid pointer.
However, I do still think we could have that early return in the case of
!vma->vm_file I suggested earlier. There is no need to take the locks
in this case we will not be calling the hugetlb unmap code.
--
Mike Kravetz
On 10/03/23 23:25, [email protected] wrote:
> From: Rik van Riel <[email protected]>
>
> Replace the custom hugetlbfs VMA locking code with the recently
> introduced invalidate_lock. This greatly simplifies things.
>
> However, this is a large enough change that it should probably go in
> separately from the other changes.
>
> Another question is whether this simplification hurts scalability
> for certain workloads.
>
> Suggested-by: Matthew Wilcox <[email protected]>
> Signed-off-by: Rik van Riel <[email protected]>
> ---
> fs/hugetlbfs/inode.c | 70 ++----------
> include/linux/fs.h | 6 +
> include/linux/hugetlb.h | 21 +---
> mm/hugetlb.c | 237 ++++------------------------------------
I have not gone through the patch, but it does produce the following:
[ 49.783584] =====================================
[ 49.784570] WARNING: bad unlock balance detected!
[ 49.785589] 6.6.0-rc3-next-20230925+ #35 Not tainted
[ 49.786644] -------------------------------------
[ 49.787768] hfill2/938 is trying to release lock (mapping.invalidate_lock) at:
[ 49.789387] [<ffffffff815212e5>] remove_inode_hugepages+0x405/0x4b0
[ 49.790723] but there are no more locks to release!
[ 49.791808]
[ 49.791808] other info that might help us debug this:
[ 49.793274] 4 locks held by hfill2/938:
[ 49.794190] #0: ffff8881ff3213e8 (sb_writers#11){.+.+}-{0:0}, at: do_syscall_64+0x37/0x90
[ 49.796165] #1: ffff888181c99640 (&sb->s_type->i_mutex_key#16){+.+.}-{3:3}, at: do_truncate+0x6f/0xd0
[ 49.798188] #2: ffff888301592f98 (&hugetlb_fault_mutex_table[i]){+.+.}-{3:3}, at: remove_inode_hugepages+0x144/0x4b0
[ 49.800494] #3: ffff888181c998b0 (&hugetlbfs_i_mmap_rwsem_key){++++}-{3:3}, at: remove_inode_hugepages+0x239/0x4b0
[ 49.803599]
[ 49.803599] stack backtrace:
[ 49.804817] CPU: 0 PID: 938 Comm: hfill2 Not tainted 6.6.0-rc3-next-20230925+ #35
[ 49.806599] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.2-1.fc37 04/01/2014
[ 49.808551] Call Trace:
[ 49.809232] <TASK>
[ 49.809843] dump_stack_lvl+0x57/0x90
[ 49.810775] lock_release+0x1eb/0x290
[ 49.811692] up_write+0x17/0x1b0
[ 49.812479] remove_inode_hugepages+0x405/0x4b0
[ 49.813757] hugetlbfs_setattr+0x113/0x170
[ 49.814699] notify_change+0x228/0x4c0
[ 49.815581] ? do_truncate+0x7f/0xd0
[ 49.816413] do_truncate+0x7f/0xd0
[ 49.817220] do_sys_ftruncate+0x27d/0x2d0
[ 49.818075] do_syscall_64+0x37/0x90
[ 49.818902] entry_SYSCALL_64_after_hwframe+0x6e/0xd8
[ 49.820038] RIP: 0033:0x7f0031dfc6ab
[ 49.820870] Code: 77 05 c3 0f 1f 40 00 48 8b 15 c9 97 0c 00 f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 4d 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 05 c3 0f 1f 40 00 48 8b 15 99 97 0c 00 f7 d8
[ 49.824752] RSP: 002b:00007fffc62dbc38 EFLAGS: 00000202 ORIG_RAX: 000000000000004d
[ 49.826447] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f0031dfc6ab
[ 49.827965] RDX: 000000007d000000 RSI: 0000000000200000 RDI: 0000000000000003
[ 49.829715] RBP: 00007fffc62dbcd0 R08: 0000000000000003 R09: 0000000000000000
[ 49.831517] R10: 0000000000400468 R11: 0000000000000202 R12: 00000000004007e0
[ 49.834459] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
[ 49.836231] </TASK>
[ 49.836999] ------------[ cut here ]------------
[ 49.838264] DEBUG_RWSEMS_WARN_ON((rwsem_owner(sem) != current) && !rwsem_test_oflags(sem, RWSEM_NONSPINNABLE)): count = 0x0, magic = 0xffff888181c99770, owner = 0x1, curr 0xffff888182c51ac0, list empty
[ 49.843168] WARNING: CPU: 0 PID: 938 at kernel/locking/rwsem.c:1369 up_write+0x19a/0x1b0
[ 49.845190] Modules linked in: rfkill ip6table_filter ip6_tables sunrpc snd_hda_codec_generic snd_hda_intel snd_intel_dspcfg snd_hda_codec snd_hwdep snd_hda_core snd_seq snd_seq_device snd_pcm 9p netfs joydev snd_timer snd virtio_balloon 9pnet_virtio soundcore 9pnet virtio_blk virtio_net net_failover failover virtio_console crct10dif_pclmul crc32_pclmul crc32c_intel ghash_clmulni_intel serio_raw virtio_pci virtio virtio_pci_legacy_dev virtio_pci_modern_dev virtio_ring fuse
[ 49.854110] CPU: 0 PID: 938 Comm: hfill2 Not tainted 6.6.0-rc3-next-20230925+ #35
[ 49.855858] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.2-1.fc37 04/01/2014
[ 49.857752] RIP: 0010:up_write+0x19a/0x1b0
[ 49.858731] Code: c6 c8 35 42 82 48 c7 c7 60 35 42 82 48 39 c2 48 c7 c2 be cc 46 82 48 c7 c0 08 35 42 82 48 0f 44 c2 48 8b 13 50 e8 26 6f f7 ff <0f> 0b 5a e9 b7 fe ff ff 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00
[ 49.862672] RSP: 0000:ffffc9000475bc80 EFLAGS: 00010282
[ 49.865024] RAX: 0000000000000000 RBX: ffff888181c99770 RCX: 0000000000000000
[ 49.866759] RDX: 0000000000000002 RSI: ffffffff8246ccbe RDI: 00000000ffffffff
[ 49.868547] RBP: ffffea0008130000 R08: 0000000000009ffb R09: 00000000ffffdfff
[ 49.870316] R10: 00000000ffffdfff R11: ffffffff82676120 R12: 0000000000200000
[ 49.872040] R13: 0000000000000f30 R14: 0000000000000048 R15: 0000000000009000
[ 49.873911] FS: 00007f0031ece540(0000) GS:ffff888277c00000(0000) knlGS:0000000000000000
[ 49.876060] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 49.877535] CR2: 00007fe0de200000 CR3: 00000001f83ae006 CR4: 0000000000370ef0
[ 49.879401] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 49.881217] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 49.883037] Call Trace:
[ 49.883788] <TASK>
[ 49.884458] ? up_write+0x19a/0x1b0
[ 49.885497] ? __warn+0x81/0x170
[ 49.886433] ? up_write+0x19a/0x1b0
[ 49.887549] ? report_bug+0x18d/0x1c0
[ 49.888561] ? tick_nohz_tick_stopped+0x12/0x30
[ 49.889760] ? handle_bug+0x41/0x70
[ 49.890703] ? exc_invalid_op+0x13/0x60
[ 49.891740] ? asm_exc_invalid_op+0x16/0x20
[ 49.892799] ? up_write+0x19a/0x1b0
[ 49.893749] remove_inode_hugepages+0x405/0x4b0
[ 49.895893] hugetlbfs_setattr+0x113/0x170
[ 49.896931] notify_change+0x228/0x4c0
[ 49.897964] ? do_truncate+0x7f/0xd0
[ 49.898964] do_truncate+0x7f/0xd0
[ 49.899935] do_sys_ftruncate+0x27d/0x2d0
[ 49.900991] do_syscall_64+0x37/0x90
[ 49.902061] entry_SYSCALL_64_after_hwframe+0x6e/0xd8
[ 49.903909] RIP: 0033:0x7f0031dfc6ab
[ 49.905041] Code: 77 05 c3 0f 1f 40 00 48 8b 15 c9 97 0c 00 f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 4d 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 05 c3 0f 1f 40 00 48 8b 15 99 97 0c 00 f7 d8
[ 49.909584] RSP: 002b:00007fffc62dbc38 EFLAGS: 00000202 ORIG_RAX: 000000000000004d
[ 49.911572] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f0031dfc6ab
[ 49.913188] RDX: 000000007d000000 RSI: 0000000000200000 RDI: 0000000000000003
[ 49.914836] RBP: 00007fffc62dbcd0 R08: 0000000000000003 R09: 0000000000000000
[ 49.916489] R10: 0000000000400468 R11: 0000000000000202 R12: 00000000004007e0
[ 49.918181] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
[ 49.919813] </TASK>
[ 49.920497] irq event stamp: 15879
[ 49.921388] hardirqs last enabled at (15879): [<ffffffff81c7e560>] _raw_spin_unlock_irqrestore+0x30/0x60
[ 49.923608] hardirqs last disabled at (15878): [<ffffffff81c7e27f>] _raw_spin_lock_irqsave+0x5f/0x70
[ 49.926781] softirqs last enabled at (15588): [<ffffffff810faed1>] __irq_exit_rcu+0x91/0x100
[ 49.929092] softirqs last disabled at (15583): [<ffffffff810faed1>] __irq_exit_rcu+0x91/0x100
[ 49.931140] ---[ end trace 0000000000000000 ]---
Attached is a simple and somewhat ugly test program generating races between
truncate and page faults. Hopefully, this will allow you to recreate. You
can ignore the user space errors, the important thing is to make sure the
kernel is stable.
--
Mike Kravetz
On Thu, 2023-10-05 at 17:19 -0700, Mike Kravetz wrote:
>
> I have not gone through the patch, but it does produce the following:
>
> [ 49.783584] =====================================
> [ 49.784570] WARNING: bad unlock balance detected!
> [ 49.785589] 6.6.0-rc3-next-20230925+ #35 Not tainted
> [ 49.786644] -------------------------------------
> [ 49.787768] hfill2/938 is trying to release lock
> (mapping.invalidate_lock) at:
> [ 49.789387] [<ffffffff815212e5>]
> remove_inode_hugepages+0x405/0x4b0
> [ 49.790723] but there are no more locks to release!
> [ 49.791808]
> [ 49.791808] other info that might help us debug this:
> [ 49.793274] 4 locks held by hfill2/938:
> [ 49.794190] #0: ffff8881ff3213e8 (sb_writers#11){.+.+}-{0:0}, at:
> do_syscall_64+0x37/0x90
> [ 49.796165] #1: ffff888181c99640 (&sb->s_type-
> >i_mutex_key#16){+.+.}-{3:3}, at: do_truncate+0x6f/0xd0
> [ 49.798188] #2: ffff888301592f98
> (&hugetlb_fault_mutex_table[i]){+.+.}-{3:3}, at:
> remove_inode_hugepages+0x144/0x4b0
> [ 49.800494] #3: ffff888181c998b0
> (&hugetlbfs_i_mmap_rwsem_key){++++}-{3:3}, at:
> remove_inode_hugepages+0x239/0x4b0
Well that's a fun one. The remove_inode_hugepages function
does not take the mapping.invalidate_lock, but it calls
hugetlb_unmap_file_folio which does.
The vma_interval_tree_foreach loop has a stray
hugetlb_vma_unlock_write() left, which I should have
removed when lifting the lock outside of the loop.
Nice catch!
--
All Rights Reversed.