| Message ID | 20260925153127.1472301-1-berto@igalia.com |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | [7.2.y] ext4: don't enable large folios on encrypted inodes | expand |
> Commit 709f0f1f1bf5 ("ext4: add checks for large folio > incompatibilities when BS > PS") disables large folios at mount time > if the "encrypt" feature is set. However, enabling that feature on a > mounted filesystem circumvents that check, and writing encrypted data > results in a crash: > > BUG: kernel NULL pointer dereference, address: 0000000000000028 > Workqueue: ext4-rsv-conversion ext4_end_io_rsv_work > RIP: 0010:ext4_finish_bio+0xf7/0x3e0 > Call Trace: > ext4_release_io_end+0x51/0x110 > ext4_end_io_end+0x4c/0xe0 > ext4_end_io_rsv_work+0xb3/0x100 > > Fix this by disabling large folios on encrypted inodes, as is > already done for inodes that use data journaling. > > This affects all kernels between v6.19 and v7.2, but it is not needed > in v7.3, where the code was significantly refactored. > > Fixes: 709f0f1f1bf5 ("ext4: add checks for large folio incompatibilities when BS > PS") > Signed-off-by: Alberto Garcia <berto@igalia.com> Sashiko has reviewed this patch and found no issues. It looks great!
On Fri, Sep 25, 2026 at 05:31:25PM +0200, Alberto Garcia wrote: > Commit 709f0f1f1bf5 ("ext4: add checks for large folio > incompatibilities when BS > PS") disables large folios at mount time > if the "encrypt" feature is set. However, enabling that feature on a > mounted filesystem circumvents that check, and writing encrypted data > results in a crash: > > BUG: kernel NULL pointer dereference, address: 0000000000000028 > Workqueue: ext4-rsv-conversion ext4_end_io_rsv_work > RIP: 0010:ext4_finish_bio+0xf7/0x3e0 > Call Trace: > ext4_release_io_end+0x51/0x110 > ext4_end_io_end+0x4c/0xe0 > ext4_end_io_rsv_work+0xb3/0x100 > > Fix this by disabling large folios on encrypted inodes, as is > already done for inodes that use data journaling. > > This affects all kernels between v6.19 and v7.2, but it is not needed > in v7.3, where the code was significantly refactored. > > Fixes: 709f0f1f1bf5 ("ext4: add checks for large folio incompatibilities when BS > PS") > Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Eric Biggers <ebiggers@kernel.org> - Eric
> This patch has no upstream equivalent: in v7.3 the code was > significantly refactored and the problem no longer happens, see commit > b4e409c18c79 ("ext4: Remove fs-layer file contents en/decryption > code"). Since this is a stable-only change, I'd like an ack from the ext4 side before queuing it. Eric's Reviewed-by is already in. Ted, could you ack this for 7.2.y? I'll queue it once you do.
> This affects all kernels between v6.19 and v7.2 but only the latter is > still maintained, so I think that it only needs to be applied there. Ted, Jan, Baokun: could one of you ack this for 7.2.y? Eric has already reviewed it. If there's no objection from the ext4 side by October 9, I'll queue it on Eric's Reviewed-by.
On 2026/9/25 23:31, Alberto Garcia wrote: > Commit 709f0f1f1bf5 ("ext4: add checks for large folio > incompatibilities when BS > PS") disables large folios at mount time > if the "encrypt" feature is set. However, enabling that feature on a > mounted filesystem circumvents that check, and writing encrypted data > results in a crash: > > BUG: kernel NULL pointer dereference, address: 0000000000000028 > Workqueue: ext4-rsv-conversion ext4_end_io_rsv_work > RIP: 0010:ext4_finish_bio+0xf7/0x3e0 > Call Trace: > ext4_release_io_end+0x51/0x110 > ext4_end_io_end+0x4c/0xe0 > ext4_end_io_rsv_work+0xb3/0x100 > > Fix this by disabling large folios on encrypted inodes, as is > already done for inodes that use data journaling. > > This affects all kernels between v6.19 and v7.2, but it is not needed > in v7.3, where the code was significantly refactored. > > Fixes: 709f0f1f1bf5 ("ext4: add checks for large folio incompatibilities when BS > PS") > Signed-off-by: Alberto Garcia <berto@igalia.com> Looks good, thanks for the fix. Reviewed-by: Baokun Li <libaokun@linux.alibaba.com> > --- > This affects all kernels between v6.19 and v7.2 but only the latter is > still maintained, so I think that it only needs to be applied there. > > This patch has no upstream equivalent: in v7.3 the code was > significantly refactored and the problem no longer happens, see commit > b4e409c18c79 ("ext4: Remove fs-layer file contents en/decryption > code"). > > To reproduce the bug: > > # mkfs.ext4 /dev/vdb > # mount /dev/vdb /mnt/ > # tune2fs -O encrypt /dev/vdb > # mkdir /mnt/foo > # fscrypt setup /mnt > # fscrypt encrypt /mnt/foo/ > # head -c 20M /dev/urandom > /mnt/foo/data > # sync > BUG: kernel NULL pointer dereference, address: 0000000000000028 > Workqueue: ext4-rsv-conversion ext4_end_io_rsv_work > [...] > > fs/ext4/inode.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c > index a33629b397b0..32f79f48a9dd 100644 > --- a/fs/ext4/inode.c > +++ b/fs/ext4/inode.c > @@ -5287,7 +5287,8 @@ void ext4_set_inode_mapping_order(struct inode *inode) > if (!min_order && !S_ISREG(inode->i_mode)) > return; > > - if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA)) > + if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) || > + ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT)) > max_order = min_order; > > mapping_set_folio_order_range(inode->i_mapping, min_order, max_order);
> Fix this by disabling large folios on encrypted inodes, as is > already done for inodes that use data journaling. Queued for 7.2, thanks.
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index a33629b397b0..32f79f48a9dd 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -5287,7 +5287,8 @@ void ext4_set_inode_mapping_order(struct inode *inode) if (!min_order && !S_ISREG(inode->i_mode)) return; - if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA)) + if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) || + ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT)) max_order = min_order; mapping_set_folio_order_range(inode->i_mapping, min_order, max_order);
Commit 709f0f1f1bf5 ("ext4: add checks for large folio incompatibilities when BS > PS") disables large folios at mount time if the "encrypt" feature is set. However, enabling that feature on a mounted filesystem circumvents that check, and writing encrypted data results in a crash: BUG: kernel NULL pointer dereference, address: 0000000000000028 Workqueue: ext4-rsv-conversion ext4_end_io_rsv_work RIP: 0010:ext4_finish_bio+0xf7/0x3e0 Call Trace: ext4_release_io_end+0x51/0x110 ext4_end_io_end+0x4c/0xe0 ext4_end_io_rsv_work+0xb3/0x100 Fix this by disabling large folios on encrypted inodes, as is already done for inodes that use data journaling. This affects all kernels between v6.19 and v7.2, but it is not needed in v7.3, where the code was significantly refactored. Fixes: 709f0f1f1bf5 ("ext4: add checks for large folio incompatibilities when BS > PS") Signed-off-by: Alberto Garcia <berto@igalia.com> --- This affects all kernels between v6.19 and v7.2 but only the latter is still maintained, so I think that it only needs to be applied there. This patch has no upstream equivalent: in v7.3 the code was significantly refactored and the problem no longer happens, see commit b4e409c18c79 ("ext4: Remove fs-layer file contents en/decryption code"). To reproduce the bug: # mkfs.ext4 /dev/vdb # mount /dev/vdb /mnt/ # tune2fs -O encrypt /dev/vdb # mkdir /mnt/foo # fscrypt setup /mnt # fscrypt encrypt /mnt/foo/ # head -c 20M /dev/urandom > /mnt/foo/data # sync BUG: kernel NULL pointer dereference, address: 0000000000000028 Workqueue: ext4-rsv-conversion ext4_end_io_rsv_work [...] fs/ext4/inode.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)