[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <5cd121df-1f26-b484-f991-eeaaa33e767d@redhat.com>
Date: Wed, 21 Jan 2026 17:55:55 +0100 (CET)
From: Mikulas Patocka <mpatocka@...hat.com>
To: Eric Biggers <ebiggers@...nel.org>
cc: dm-devel@...ts.linux.dev, Alasdair Kergon <agk@...hat.com>,
Mike Snitzer <snitzer@...nel.org>,
Benjamin Marzinski <bmarzins@...hat.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] dm: fix excessive blk-crypto operations for invalid
keys
Accepted, thanks.
Mikulas
On Fri, 16 Jan 2026, Eric Biggers wrote:
> dm_exec_wrappedkey_op() passes through the derive_sw_secret, import_key,
> generate_key, and prepare_key blk-crypto operations to an underlying
> device.
>
> Currently, it calls the operation on every underlying device until one
> returns success.
>
> This logic is flawed when the operation is expected to fail, such as an
> invalid key being passed to derive_sw_secret. That can happen if
> userspace passes an invalid key to the FS_IOC_ADD_ENCRYPTION_KEY ioctl.
>
> When that happens on a device-mapper device that consists of many
> dm-linear targets, a lot of unnecessary key unwrapping requests get sent
> to the underlying key wrapping hardware.
>
> Fix this by considering the first device only. As already documented in
> the comment, it was already checked that all underlying devices support
> wrapped keys, so this should be fine.
>
> Fixes: e93912786e50 ("dm: pass through operations on wrapped inline crypto keys")
> Cc: stable@...r.kernel.org
> Signed-off-by: Eric Biggers <ebiggers@...nel.org>
> ---
> drivers/md/dm-table.c | 12 +++---------
> 1 file changed, 3 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
> index 0522cd700e0e..4b70872725d0 100644
> --- a/drivers/md/dm-table.c
> +++ b/drivers/md/dm-table.c
> @@ -1235,13 +1235,10 @@ static int dm_wrappedkey_op_callback(struct dm_target *ti, struct dm_dev *dev,
> struct block_device *bdev = dev->bdev;
> struct blk_crypto_profile *profile =
> bdev_get_queue(bdev)->crypto_profile;
> int err = -EOPNOTSUPP;
>
> - if (!args->err)
> - return 0;
> -
> switch (args->op) {
> case DERIVE_SW_SECRET:
> err = blk_crypto_derive_sw_secret(
> bdev,
> args->derive_sw_secret.eph_key,
> @@ -1264,13 +1261,11 @@ static int dm_wrappedkey_op_callback(struct dm_target *ti, struct dm_dev *dev,
> args->prepare_key.lt_key_size,
> args->prepare_key.eph_key);
> break;
> }
> args->err = err;
> -
> - /* Try another device in case this fails. */
> - return 0;
> + return 1; /* No need to continue the iteration. */
> }
>
> static int dm_exec_wrappedkey_op(struct blk_crypto_profile *profile,
> struct dm_wrappedkey_op_args *args)
> {
> @@ -1292,18 +1287,17 @@ static int dm_exec_wrappedkey_op(struct blk_crypto_profile *profile,
> * implementations of wrapped inline crypto keys on a single system.
> * It was already checked earlier that support for wrapped keys was
> * declared on all underlying devices. Thus, all the underlying devices
> * should support all wrapped key operations and they should behave
> * identically, i.e. work with the same keys. So, just executing the
> - * operation on the first device on which it works suffices for now.
> + * operation on the first device suffices for now.
> */
> for (i = 0; i < t->num_targets; i++) {
> ti = dm_table_get_target(t, i);
> if (!ti->type->iterate_devices)
> continue;
> - ti->type->iterate_devices(ti, dm_wrappedkey_op_callback, args);
> - if (!args->err)
> + if (ti->type->iterate_devices(ti, dm_wrappedkey_op_callback, args) != 0)
> break;
> }
> out:
> dm_put_live_table(md, srcu_idx);
> return args->err;
>
> base-commit: fb8a6c18fb9a6561f7a15b58b272442b77a242dd
> --
> 2.52.0
>
Powered by blists - more mailing lists