Re: [REGRESSION] ext4: oops in ext4_finish_bio() after enabling encryption on a mounted fs

[Baokun Li <libaokun@linux.alibaba.com>]

On 2026/9/24 02:01, Eric Biggers wrote:
> On Wed, Sep 23, 2026 at 09:28:54PM +0800, Baokun Li wrote:
>> On 2026/9/23 20:40, Alberto Garcia wrote:
>>> On Wed, Sep 23, 2026 at 08:12:03PM +0800, Baokun Li wrote:
>>>>> 1) Enable encryption on an ext4 filesystem with tune2fs -O encrypt
>>>> Is it valid to enable encrypt at mount time?
>>> Good question, I always understood that it was allowed and tune2fs
>>> certainly doesn't forbid it (compare with casefold):
>>>
>>> https://github.com/tytso/e2fsprogs/blob/v1.47.4/misc/tune2fs.c#L1599
>>>
>>> Berto
>>
>> If enabling encryption on a mounted ext4 fs is allowed,
>> I think the current change is fine.
>>
>> Ted, Eric - would either of you know the details here?
> Yes, it is allowed.
>
> The proposed patch looks good, even though the problem seems to be gone
> on mainline already due to the removal of the code path that used the
> non-large-folio-compatible function fscrypt_encrypt_pagecache_blocks().

Agreed, then this current modification can be made into a separate
bugfix for backporting to stable.

>
> There can be another patch that removes both that check and the
> mount-time check, if they're truly no longer needed (I don't know of any
> reason why they would be, but it needs to be properly tested).

After removing both checks, I ran

    kvm-xfstests -c ext4/32k -g encrypt

16 of the 29 tests still fail.  blk-crypto-fallback still en/decrypts one
page at a time, while the data unit size is 32k with a block size larger
than the page size.  It doesn't crash, but the writes fail with
BLK_STS_INVAL, so write() succeeds while the data never reaches the disk,
and reads fail once the page cache is dropped.

Supporting data units larger than a page is simple enough.  A rough
implementation of mine already gives 30%~50% higher buffered I/O
throughput with 32k blocks than with 4k blocks.  The patches are still
being polished, and I'll send them out in the next two days.


Thanks,
Baokun