[PATCH] iov_iter: validate first nonempty segment for bvec alignment

[Daehyeon Ko <4ncienth@gmail.com>]

iov_iter_extract_bvecs() checks the current IOVEC or KVEC segment against
the block device's DMA alignment before extracting pages.  Empty segments
are valid, but the check uses iter_iov_addr() and iter_iov_len() directly
while first_iovec_segment(), used by the extraction path, skips them.

An aligned empty segment followed by a misaligned nonempty segment
therefore passes the check and produces misaligned bvecs.  An unprivileged
O_DIRECT preadv() reproduced this on a regular ext4 file above brd.  The
equivalent nonempty-leading vector returned -EINVAL, while the
empty-leading vector completed 4096 bytes.  This bypasses the validation
added for CVE-2026-89584.

When the current segment is empty but the iterator still contains data, use
first_iovec_segment() to select and validate the same segment that page
extraction will use.  Keep the existing constant-time fast path for the
normal nonempty case.

Fixes: 14b007e17881 ("block: validate user space vectors during extraction")
Cc: stable@vger.kernel.org
Assisted-by: LLM
Signed-off-by: Daehyeon Ko <4ncienth@gmail.com>
---
 lib/iov_iter.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index 2072c04e99d052..893de96de8ab6a 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -1938,8 +1938,13 @@ ssize_t iov_iter_extract_bvecs(struct iov_iter *iter, struct bio_vec *bv,
 		   iter_is_iovec(iter) ||
 		   iov_iter_is_kvec(iter))) {
 		unsigned long start = (unsigned long)iter_iov_addr(iter);
+		size_t seg_size = iter_iov_len(iter);
 
-		if ((start | iter_iov_len(iter)) & mem_align_mask)
+		if (unlikely(!seg_size && iter->count)) {
+			seg_size = iter->count;
+			start = first_iovec_segment(iter, &seg_size);
+		}
+		if ((start | seg_size) & mem_align_mask)
 			return -EINVAL;
 	} else if (iov_iter_alignment(iter) & mem_align_mask) {
 		return -EINVAL;

base-commit: d6d3455d8751854e0d1820f14184682e23ac4475
-- 
2.55.0