Since commit 7f3bfab52cab ("mm/gup: take mmap_lock in get_dump_page()"),
which landed in v5.10, core dumping doesn't enter fault handling without
holding the mmap_lock anymore. Remove the stale parts of the comments, but
leave the behavior as-is - letting core dumping block on userfault handling
would be a bad idea and could lead to deadlocks if the dumping process was
handling its own userfaults.
Signed-off-by: Jann Horn <jannh@xxxxxxxxxx>
---
fs/userfaultfd.c | 6 +-----
1 file changed, 1 insertion(+), 5 deletions(-)
diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
index 1091cb461747..56eaae9dac1a 100644
--- a/fs/userfaultfd.c
+++ b/fs/userfaultfd.c
@@ -428,15 +428,11 @@ vm_fault_t handle_userfault(struct vm_fault *vmf, unsigned long reason)
* FOLL_DUMP case, anon memory also checks for FOLL_DUMP with
* the no_page_table() helper in follow_page_mask(), but the
* shmem_vm_ops->fault method is invoked even during
- * coredumping without mmap_lock and it ends up here.
+ * coredumping and it ends up here.
*/
if (current->flags & (PF_EXITING|PF_DUMPCORE))
goto out;
- /*
- * Coredumping runs without mmap_lock so we can only check that
- * the mmap_lock is held, if PF_DUMPCORE was not set.
- */
assert_fault_locked(vmf);
ctx = vma->vm_userfaultfd_ctx.ctx;
base-commit: f8a7ed854c1a66b1984a64baf0ece09e44552fab
--
2.41.0.694.ge786442a9b-goog
