[PATCH] md: reject v0.90 arrays with too many devices

The v0.90 superblock has room for MD_SB_DISKS device descriptors, but
super_90_load() only rejects nonpositive raid_disks values. A crafted
superblock can therefore copy a larger value into mddev->raid_disks. When
metadata is next updated, super_90_sync() walks sb->disks[] up to that
count and reads and writes beyond the 4 KiB superblock buffer.

A checksum-valid RAID1 member advertising 64 devices can be assembled by
mdadm --incremental --run. The first write to the array then produces
out-of-bounds writes from super_90_sync() under KASAN.

Reject device counts that cannot be represented by the v0.90 format while
loading the superblock.

Assisted-by: Codex:gpt-5
Signed-off-by: Jérémy Jean <[email protected]>
---
 drivers/md/md.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/md/md.c b/drivers/md/md.c
index d1465bcd86c8..873a0e5fce7d 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -1398,7 +1398,7 @@ static int super_90_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor
 		goto abort;
 	}
 
-	if (sb->raid_disks <= 0)
+	if (sb->raid_disks <= 0 || sb->raid_disks > MD_SB_DISKS)
 		goto abort;
 
 	if (md_csum_fold(calc_sb_csum(sb)) != md_csum_fold(sb->sb_csum)) {
-- 
2.47.3