In fuse_get_user_pages(), set *nbytesp to 0 when struct page **pages allocation fails. This prevents the caller (fuse_direct_io) from making incorrect assumptions that could lead to NULL pointer dereferences when processing the request reply. Previously, *nbytesp was left unmodified on allocation failure, which could cause issues if the caller assumed pages had been added to ap->descs[] when they hadn't. Reported-by: syzbot+87b8e6ed25dbc41759f7@xxxxxxxxxxxxxxxxxxxxxxxxx Closes: https://syzkaller.appspot.com/bug?extid=87b8e6ed25dbc41759f7 Fixes: 3b97c3652d91 ("fuse: convert direct io to use folios") Signed-off-by: Bernd Schubert <bschubert@xxxxxxx> Reviewed-by: Joanne Koong <joannelkoong@xxxxxxxxx> --- Changes in v2: - Set ret in the (!pages) condition only to avoid returning ENOMEM when the while loop would not do anything - Remove the error check in fuse_copy_do(), that is a bit debatable. I had added it to prevent kernel crashes on fuse error, but then it causes 'visual clutter' (Joanne) - Link to v1: https://lore.kernel.org/r/20241202-fix-fuse_get_user_pages-v1-1-8b5cccaf5bbe@xxxxxxx --- fs/fuse/file.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/fs/fuse/file.c b/fs/fuse/file.c index 88d0946b5bc98705e0d895bc798aa4d9df080c3c..ae74d2b7ad5be14e4d157495e7c00fcf3fc40625 100644 --- a/fs/fuse/file.c +++ b/fs/fuse/file.c @@ -1541,8 +1541,10 @@ static int fuse_get_user_pages(struct fuse_args_pages *ap, struct iov_iter *ii, */ struct page **pages = kzalloc(max_pages * sizeof(struct page *), GFP_KERNEL); - if (!pages) - return -ENOMEM; + if (!pages) { + ret = -ENOMEM; + goto out; + } while (nbytes < *nbytesp && nr_pages < max_pages) { unsigned nfolios, i; @@ -1584,6 +1586,7 @@ static int fuse_get_user_pages(struct fuse_args_pages *ap, struct iov_iter *ii, else ap->args.out_pages = true; +out: *nbytesp = nbytes; return ret < 0 ? ret : 0; --- base-commit: e70140ba0d2b1a30467d4af6bcfe761327b9ec95 change-id: 20241202-fix-fuse_get_user_pages-6a920cb04184 Best regards, -- Bernd Schubert <bschubert@xxxxxxx>
