Re: [PATCH] dm-inlinecrypt: emit key type as an optional table argument
From: Eric Biggers
Date: Fri Nov 20 2026 - 15:58:30 EST
On Fri, Nov 20, 2026 at 10:06:01PM +0530, Anvik Loitongbam wrote:
> inlinecrypt_status() emits ctx->key_type as a positional argument between
> the key and iv_offset:
>
> <cipher> <key> <key_type> <iv_offset> <dev_path> <start>
>
> but inlinecrypt_ctr() parses the base table as:
>
> <cipher> <key> <iv_offset> <dev_path> <start>
>
> and keytype, is an optional feature argument parsed by
> inlinecrypt_ctr_optional(). Resulting in STATUSTYPE_TABLE output for every
> dm-inlinecrypt device to be not reloadable, the constructor reads key_type as
> iv_offset and shifts the remaining fields.
>
> Keep the base table in constructor order and emit a non-default key type
> as keytype:hw-wrapped in the optional feature list. Raw keys remain the
> default and are not emitted, preserving the existing table text for
> raw-key devices.
>
> The two dmsetup examples in the documentation acquired
> the same extra positional key-type field, so removed it there as well.
>
> Fixes: f9f57971da38 ("dm-inlinecrypt: add support for hardware-wrapped keys")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Anvik Loitongbam <permissiontogallop@xxxxxxxxx>
The original Sashiko review of this commit mentioned this bug already
(https://sashiko.dev/#/patchset/20260525121348.1321282-2-linlin.zhang%40oss.qualcomm.com).
Might be worth adding a Reported-by if you found it from there.
> + if (ctx->key_type != BLK_CRYPTO_KEY_TYPE_RAW)
> + num_feature_args++;
> if (ctx->sector_size != SECTOR_SIZE)
> num_feature_args += 2;
> if (num_feature_args != 0) {
> DMEMIT(" %d", num_feature_args);
> + if (ctx->key_type != BLK_CRYPTO_KEY_TYPE_RAW)
> + DMEMIT(" keytype:%s",
> + ctx->key_type == BLK_CRYPTO_KEY_TYPE_HW_WRAPPED ?
> + "hw-wrapped" : "raw");
The "raw" fallback case doesn't make much sense, because it was just
checked above that it is *not* raw. So if it is ever reached due to a
new enum value it will be wrong anyway. Maybe just leave that out and
simply do DMEMIT(" keytype:hw-wrapped").
Otherwise this fix looks good, thanks.
- Eric