syz reported a slab-out-of-bounds Write in fuse_dev_do_write.

Using the number of bytes alone as the termination condition in a loop
can prematurely exhaust the allocated memory if the incremented byte count
is less than PAGE_SIZE.

Add a loop termination condition to prevent overruns.

Fixes: 3568a9569326 ("fuse: support large folios for retrieves")
Reported-by: syzbot+2d215d165f9354b9c4ea@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=2d215d165f9354b9c4ea
Tested-by: syzbot+2d215d165f9354b9c4ea@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
---
 fs/fuse/dev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
index e80cd8f2c049..5150aa25e64b 100644
--- a/fs/fuse/dev.c
+++ b/fs/fuse/dev.c
@@ -1893,7 +1893,7 @@ static int fuse_retrieve(struct fuse_mount *fm, struct inode *inode,
 
 	index = outarg->offset >> PAGE_SHIFT;
 
-	while (num) {
+	while (num && ap->num_folios < num_pages) {
 		struct folio *folio;
 		unsigned int folio_offset;
 		unsigned int nr_bytes;
-- 
2.43.0

On Fri, Aug 22, 2025 at 6:17 PM Edward Adam Davis <eadavis@qq.com> wrote:
>
> syz reported a slab-out-of-bounds Write in fuse_dev_do_write.
>
> Using the number of bytes alone as the termination condition in a loop
> can prematurely exhaust the allocated memory if the incremented byte count
> is less than PAGE_SIZE.

I don't think the last part of this is quite right. It's fine if the
incremented byte count is less than PAGE_SIZE (which will always be
the case if there's an offset). We only run into this issue when the
number of bytes to retrieve gets truncated by fc->max_pages as the
upper bound and there's an offset.

>
> Add a loop termination condition to prevent overruns.
>
> Fixes: 3568a9569326 ("fuse: support large folios for retrieves")
> Reported-by: syzbot+2d215d165f9354b9c4ea@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=2d215d165f9354b9c4ea
> Tested-by: syzbot+2d215d165f9354b9c4ea@syzkaller.appspotmail.com
> Signed-off-by: Edward Adam Davis <eadavis@qq.com>
> ---
>  fs/fuse/dev.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
> index e80cd8f2c049..5150aa25e64b 100644
> --- a/fs/fuse/dev.c
> +++ b/fs/fuse/dev.c
> @@ -1893,7 +1893,7 @@ static int fuse_retrieve(struct fuse_mount *fm, struct inode *inode,
>
>         index = outarg->offset >> PAGE_SHIFT;
>
> -       while (num) {
> +       while (num && ap->num_folios < num_pages) {
>                 struct folio *folio;
>                 unsigned int folio_offset;
>                 unsigned int nr_bytes;
> --
> 2.43.0

Thanks for the fix.

Reviewed-by: Joanne Koong <joannelkoong@gmail.com>

syz reported a slab-out-of-bounds Write in fuse_dev_do_write.

When the number of bytes to be retrieved is truncated to the upper limit
by fc->max_pages and there is an offset, the oob is triggered.

Add a loop termination condition to prevent overruns.

Fixes: 3568a9569326 ("fuse: support large folios for retrieves")
Reported-by: syzbot+2d215d165f9354b9c4ea@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=2d215d165f9354b9c4ea
Tested-by: syzbot+2d215d165f9354b9c4ea@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
Reviewed-by: Joanne Koong <joannelkoong@gmail.com>
---
V1 -> V2: update root cause

 fs/fuse/dev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
index e80cd8f2c049..5150aa25e64b 100644
--- a/fs/fuse/dev.c
+++ b/fs/fuse/dev.c
@@ -1893,7 +1893,7 @@ static int fuse_retrieve(struct fuse_mount *fm, struct inode *inode,
 
 	index = outarg->offset >> PAGE_SHIFT;
 
-	while (num) {
+	while (num && ap->num_folios < num_pages) {
 		struct folio *folio;
 		unsigned int folio_offset;
 		unsigned int nr_bytes;
-- 
2.43.0

On Wed, 27 Aug 2025 at 03:46, Edward Adam Davis <eadavis@qq.com> wrote:
>
> syz reported a slab-out-of-bounds Write in fuse_dev_do_write.
>
> When the number of bytes to be retrieved is truncated to the upper limit
> by fc->max_pages and there is an offset, the oob is triggered.
>
> Add a loop termination condition to prevent overruns.

Applied, thanks.

Miklos