[PATCH] dm-verity: skip verity_handle_error on I/O errors

[PATCH] dm-verity: skip verity_handle_error on I/O errors

[Akilesh Kailash]

If there is an I/O error and FEC correction fails, return
an error instead of calling verity_handle_error().

Suggested-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Akilesh Kailash <akailash@google.com>
---
This was discussed in [1] by Sami Tolvanen but was
never implemented.

[1] https://lore.kernel.org/dm-devel/b004e7c7-f795-77ed-19b9-983785780e92@gmail.com/T/#mec4df1ba3f3cb63846875fb2bfc1f8b3100f31f1

 drivers/md/dm-verity-target.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 22a5ac82446a..a6f7c452ee80 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -475,6 +475,7 @@ static int verity_verify_io(struct dm_verity_io *io)
 	struct bvec_iter start;
 	unsigned b;
 	struct crypto_wait wait;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
 	for (b = 0; b < io->n_blocks; b++) {
 		int r;
@@ -529,9 +530,17 @@ static int verity_verify_io(struct dm_verity_io *io)
 		else if (verity_fec_decode(v, io, DM_VERITY_BLOCK_TYPE_DATA,
 					   cur_block, NULL, &start) == 0)
 			continue;
-		else if (verity_handle_err(v, DM_VERITY_BLOCK_TYPE_DATA,
+		else {
+			if (bio->bi_status) {
+				/*
+				 * Error correction failed; Just return error
+				 */
+				return -EIO;
+			}
+			if (verity_handle_err(v, DM_VERITY_BLOCK_TYPE_DATA,
 					   cur_block))
-			return -EIO;
+				return -EIO;
+		}
 	}
 
 	return 0;
-- 
2.33.0.309.g3052b89438-goog

Re: [PATCH] dm-verity: skip verity_handle_error on I/O errors

[Sami Tolvanen]

Hi Akilesh,

On Mon, Sep 13, 2021 at 2:27 AM Akilesh Kailash <akailash@google.com> wrote:
>
> If there is an I/O error and FEC correction fails, return
> an error instead of calling verity_handle_error().
>
> Suggested-by: Sami Tolvanen <samitolvanen@google.com>
> Signed-off-by: Akilesh Kailash <akailash@google.com>
> ---
> This was discussed in [1] by Sami Tolvanen but was
> never implemented.
>
> [1] https://lore.kernel.org/dm-devel/b004e7c7-f795-77ed-19b9-983785780e92@gmail.com/T/#mec4df1ba3f3cb63846875fb2bfc1f8b3100f31f1

Great, thanks for the patch!

>  drivers/md/dm-verity-target.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
> index 22a5ac82446a..a6f7c452ee80 100644
> --- a/drivers/md/dm-verity-target.c
> +++ b/drivers/md/dm-verity-target.c
> @@ -475,6 +475,7 @@ static int verity_verify_io(struct dm_verity_io *io)
>         struct bvec_iter start;
>         unsigned b;
>         struct crypto_wait wait;
> +       struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
>
>         for (b = 0; b < io->n_blocks; b++) {
>                 int r;
> @@ -529,9 +530,17 @@ static int verity_verify_io(struct dm_verity_io *io)
>                 else if (verity_fec_decode(v, io, DM_VERITY_BLOCK_TYPE_DATA,
>                                            cur_block, NULL, &start) == 0)
>                         continue;
> -               else if (verity_handle_err(v, DM_VERITY_BLOCK_TYPE_DATA,
> +               else {
> +                       if (bio->bi_status) {
> +                               /*
> +                                * Error correction failed; Just return error
> +                                */
> +                               return -EIO;
> +                       }
> +                       if (verity_handle_err(v, DM_VERITY_BLOCK_TYPE_DATA,
>                                            cur_block))
> -                       return -EIO;
> +                               return -EIO;
> +               }
>         }
>
>         return 0;

This looks correct to me. A non-zero bio->bi_status here means that
FEC must be enabled because otherwise we wouldn't have ended up in
verity_verify_io() in the first place, and returning -EIO when
verity_fec_decode() fails after attempting to correct an I/O error is
consistent with the non-FEC behavior of dm-verity.

Reviewed-by: Sami Tolvanen <samitolvanen@google.com>

Sami