CVE-2025-21643

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> netfs: Fix kernel async DIO<br /> <br /> Netfslib needs to be able to handle kernel-initiated asynchronous DIO that<br /> is supplied with a bio_vec[] array. Currently, because of the async flag,<br /> this gets passed to netfs_extract_user_iter() which throws a warning and<br /> fails because it only handles IOVEC and UBUF iterators. This can be<br /> triggered through a combination of cifs and a loopback blockdev with<br /> something like:<br /> <br /> mount //my/cifs/share /foo<br /> dd if=/dev/zero of=/foo/m0 bs=4K count=1K<br /> losetup --sector-size 4096 --direct-io=on /dev/loop2046 /foo/m0<br /> echo hello &gt;/dev/loop2046<br /> <br /> This causes the following to appear in syslog:<br /> <br /> WARNING: CPU: 2 PID: 109 at fs/netfs/iterator.c:50 netfs_extract_user_iter+0x170/0x250 [netfs]<br /> <br /> and the write to fail.<br /> <br /> Fix this by removing the check in netfs_unbuffered_write_iter_locked() that<br /> causes async kernel DIO writes to be handled as userspace writes. Note<br /> that this change relies on the kernel caller maintaining the existence of<br /> the bio_vec array (or kvec[] or folio_queue) until the op is complete.