[dm-devel] dm-flakey NOT to return -EIO on READ?
Akira Hayakawa
ruby.wktk at gmail.com
Sat Jul 30 00:26:43 UTC 2016
Mike,
> This fix should resolve the problem, I'll be staging it to go upstream
> during the 4.8-rc cycle (Akira, please test to verify your test fails as
> expected now):
I tested with the patch and it's now ok.
Akira
On 2016/07/30 2:30, Mike Snitzer wrote:
> On Fri, Jul 29 2016 at 10:35am -0400,
> Mike Snitzer <snitzer at redhat.com> wrote:
>
>> On Fri, Jul 29 2016 at 10:31am -0400,
>> Lukas Herbolt <lherbolt at redhat.com> wrote:
>>
>>> - http://people.redhat.com/~lherbolt/dm-flakey/v4.7-dm_flakey.tar.gz
>>>
>>> Lukas
>>
>> Please post an incremental patch relative to 4.7 or latest Linus
>> kernel. Don't make people suffer with tarballs of whatever you've
>> packaged above.
>
> I pulled out your incremental diff. Your proposed fix isn't correct. It
> breaks the corrupt_bio_byte feature. We cannot blindly drop reads early
> in flakey_map(), like you've proposed, because the corrupt_bio_byte
> feature relies on corrupt data being returned on read during the
> down_interval.
>
> The issue reported is: reads aren't dropped during the "down_interval".
>
> I do agree that this needs fixing, and it was commit a3998799fb4df ("dm
> flakey: add corrupt_bio_byte feature") that caused this regression.
>
> But a fix must not break the corrupt_bio_byte feature.
>
> This fix should resolve the problem, I'll be staging it to go upstream
> during the 4.8-rc cycle (Akira, please test to verify your test fails as
> expected now):
>
> From: Mike Snitzer <snitzer at redhat.com>
> Date: Fri, 29 Jul 2016 13:19:55 -0400
> Subject: [PATCH] dm flakey: error READ bios during the down_interval
>
> When the corrupt_bio_byte feature was introduced it caused READ bios to
> no longer be errored with -EIO during the down_interval. This had to do
> with the complexity of needing to submit READs if the corrupt_bio_byte
> feature was used.
>
> Fix it so READ bios are properly errored with -EIO; doing so early in
> flakey_map() as long as there isn't a match for the corrupt_bio_byte
> feature.
>
> Fixes: a3998799fb4df ("dm flakey: add corrupt_bio_byte feature")
> Reported-by: Akira Hayakawa <ruby.wktk at gmail.com>
> Signed-off-by: Mike Snitzer <snitzer at redhat.com>
> Cc: stable at vger.kernel.org
> ---
> drivers/md/dm-flakey.c | 23 +++++++++++++++--------
> 1 file changed, 15 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/md/dm-flakey.c b/drivers/md/dm-flakey.c
> index 29b99fb..19db13e 100644
> --- a/drivers/md/dm-flakey.c
> +++ b/drivers/md/dm-flakey.c
> @@ -289,10 +289,16 @@ static int flakey_map(struct dm_target *ti, struct bio *bio)
> pb->bio_submitted = true;
>
> /*
> - * Map reads as normal.
> + * Map reads as normal only if corrupt_bio_byte set.
> */
> - if (bio_data_dir(bio) == READ)
> - goto map_bio;
> + if (bio_data_dir(bio) == READ) {
> + /* If flags were specified, only corrupt those that match. */
> + if (fc->corrupt_bio_byte && (fc->corrupt_bio_rw == READ) &&
> + all_corrupt_bio_flags_match(bio, fc))
> + goto map_bio;
> + else
> + return -EIO;
> + }
>
> /*
> * Drop writes?
> @@ -330,12 +336,13 @@ static int flakey_end_io(struct dm_target *ti, struct bio *bio, int error)
>
> /*
> * Corrupt successful READs while in down state.
> - * If flags were specified, only corrupt those that match.
> */
> - if (fc->corrupt_bio_byte && !error && pb->bio_submitted &&
> - (bio_data_dir(bio) == READ) && (fc->corrupt_bio_rw == READ) &&
> - all_corrupt_bio_flags_match(bio, fc))
> - corrupt_bio_data(bio, fc);
> + if (!error && pb->bio_submitted && (bio_data_dir(bio) == READ)) {
> + if (fc->corrupt_bio_byte)
> + corrupt_bio_data(bio, fc);
> + else
> + return -EIO;
> + }
>
> return error;
> }
>
More information about the dm-devel
mailing list