mm: never attempt async page lock if we've transferred data already

We catch the case where we enter generic_file_buffered_read() with data already transferred, but we also need to be careful not to allow an async page lock if we're looping transferring data. If not, we could be returning -EIOCBQUEUED instead of the transferred amount, and it could result in double waitqueue additions as well. Cc: [email protected] # v5.9 Fixes: 1a0a785 ("mm: support async buffered reads in generic_file_buffered_read()") Signed-off-by: Jens Axboe <[email protected]>
MonetRX19 · Nov 16, 2020 · 0abed7c · 0abed7c
1 parent 944d144
commit 0abed7c
Showing 1 changed file with 14 additions and 4 deletions.
diff --git a/mm/filemap.c b/mm/filemap.c
@@ -2347,10 +2347,15 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 page_not_up_to_date:
 		/* Get exclusive access to the page ... */
-		if (iocb->ki_flags & IOCB_WAITQ)
+		if (iocb->ki_flags & IOCB_WAITQ) {
+			if (written) {
+				put_page(page);
+				goto out;
+			}
 			error = lock_page_async(page, iocb->ki_waitq);
-		else
+		} else {
 			error = lock_page_killable(page);
+		}
 		if (unlikely(error))
 			goto readpage_error;
 
@@ -2393,10 +2398,15 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		}
 
 		if (!PageUptodate(page)) {
-			if (iocb->ki_flags & IOCB_WAITQ)
+			if (iocb->ki_flags & IOCB_WAITQ) {
+				if (written) {
+					put_page(page);
+					goto out;
+				}
 				error = lock_page_async(page, iocb->ki_waitq);
-			else
+			} else {
 				error = lock_page_killable(page);
+			}
 
 			if (unlikely(error))
 				goto readpage_error;