]> git.ipfire.org Git - thirdparty/linux.git/commitdiff
iov_iter: Fix a memory leak in iov_iter_extract_user_pages()
authorDavid Howells <dhowells@redhat.com>
Thu, 25 Jun 2026 14:06:25 +0000 (15:06 +0100)
committerChristian Brauner <brauner@kernel.org>
Wed, 1 Jul 2026 13:26:29 +0000 (15:26 +0200)
There's a potential memory leak in callers of iov_iter_extract_user_pages()
whereby if a pages array is allocated in function, it isn't freed before
returning of an error or 0.

Now, it's not a leak per se in iov_iter_extract_user_pages() as, if an
array is allocated, it's returned through *pages, so it's incumbent on the
caller to free it.  However, not all callers do.

Fix this by freeing the table and clearing *pages before returning an error
or 0.  Note that iov_iter_extract_pages() and its subfunctions are allowed
to return 0 without returning an array (for instance if the iterator count
is 0).

Fixes: 7d58fe731028 ("iov_iter: Add a function to extract a page list from an iterator")
Closes: https://sashiko.dev/#/patchset/20260616100821.2062304-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
Link: https://patch.msgid.link/20260625140640.3116900-8-dhowells@redhat.com
Reviewed-by: Christoph Hellwig <hch@lst.de>
cc: Paulo Alcantara <pc@manguebit.org>
cc: Matthew Wilcox <willy@infradead.org>
cc: Christoph Hellwig <hch@infradead.org>
cc: Jens Axboe <axboe@kernel.dk>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
lib/iov_iter.c

index 3dfad70328eb6695fa2aeaf8266efd863cf6b7c9..c2484551a4e8638b77f98f255f922dcffde9baad 100644 (file)
@@ -1756,6 +1756,7 @@ static ssize_t iov_iter_extract_user_pages(struct iov_iter *i,
        unsigned long addr;
        unsigned int gup_flags = 0;
        size_t offset;
+       bool will_alloc = !*pages;
        int res;
 
        if (i->data_source == ITER_DEST)
@@ -1772,8 +1773,14 @@ static ssize_t iov_iter_extract_user_pages(struct iov_iter *i,
        if (!maxpages)
                return -ENOMEM;
        res = pin_user_pages_fast(addr, maxpages, gup_flags, *pages);
-       if (unlikely(res <= 0))
+       if (unlikely(res <= 0)) {
+               if (will_alloc) {
+                       kvfree(*pages);
+                       *pages = NULL;
+               }
                return res;
+       }
+
        maxsize = min_t(size_t, maxsize, res * PAGE_SIZE - offset);
        iov_iter_advance(i, maxsize);
        return maxsize;