* [PATCH v2 0/4] Fix blk-crypto keyslot race condition
@ 2023-03-08 19:36 Eric Biggers
2023-03-08 19:36 ` [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete Eric Biggers
` (3 more replies)
0 siblings, 4 replies; 14+ messages in thread
From: Eric Biggers @ 2023-03-08 19:36 UTC (permalink / raw)
To: linux-block, Jens Axboe; +Cc: linux-fscrypt, Nathan Huckleberry
This series fixes a race condition in blk-crypto keyslot management and
makes some related cleanups. It replaces
"[PATCH] blk-crypto: make blk_crypto_evict_key() always try to evict"
(https://lore.kernel.org/r/20230226203816.207449-1-ebiggers@kernel.org),
which was a simpler fix but didn't fix the keyslot reference counting to
work as expected.
Changed in v2:
- Call blk_crypto_rq_put_keyslot() when requests are merged
- Add and use blk_crypto_rq_has_keyslot()
- Add patch "blk-crypto: drop the NULL check from blk_crypto_put_keyslot()"
Eric Biggers (4):
blk-mq: release crypto keyslot before reporting I/O complete
blk-crypto: make blk_crypto_evict_key() more robust
blk-crypto: remove blk_crypto_insert_cloned_request()
blk-crypto: drop the NULL check from blk_crypto_put_keyslot()
Documentation/block/inline-encryption.rst | 3 +-
block/blk-crypto-internal.h | 38 +++++++-------
block/blk-crypto-profile.c | 64 ++++++++---------------
block/blk-crypto.c | 47 +++++++++--------
block/blk-merge.c | 2 +
block/blk-mq.c | 17 +++++-
6 files changed, 87 insertions(+), 84 deletions(-)
base-commit: 63355b9884b3d1677de6bd1517cd2b8a9bf53978
--
2.39.2
^ permalink raw reply [flat|nested] 14+ messages in thread
* [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete
2023-03-08 19:36 [PATCH v2 0/4] Fix blk-crypto keyslot race condition Eric Biggers
@ 2023-03-08 19:36 ` Eric Biggers
2023-03-13 21:26 ` Nathan Huckleberry
2023-03-15 16:19 ` Christoph Hellwig
2023-03-08 19:36 ` [PATCH v2 2/4] blk-crypto: make blk_crypto_evict_key() more robust Eric Biggers
` (2 subsequent siblings)
3 siblings, 2 replies; 14+ messages in thread
From: Eric Biggers @ 2023-03-08 19:36 UTC (permalink / raw)
To: linux-block, Jens Axboe; +Cc: linux-fscrypt, Nathan Huckleberry, stable
From: Eric Biggers <ebiggers@google.com>
Once all I/O using a blk_crypto_key has completed, filesystems can call
blk_crypto_evict_key(). However, the block layer currently doesn't call
blk_crypto_put_keyslot() until the request is being freed, which happens
after upper layers have been told (via bio_endio()) the I/O has
completed. This causes a race condition where blk_crypto_evict_key()
can see 'slot_refs != 0' without there being an actual bug.
This makes __blk_crypto_evict_key() hit the
'WARN_ON_ONCE(atomic_read(&slot->slot_refs) != 0)' and return without
doing anything, eventually causing a use-after-free in
blk_crypto_reprogram_all_keys(). (This is a very rare bug and has only
been seen when per-file keys are being used with fscrypt.)
There are two options to fix this: either release the keyslot before
bio_endio() is called on the request's last bio, or make
__blk_crypto_evict_key() ignore slot_refs. Let's go with the first
solution, since it preserves the ability to report bugs (via
WARN_ON_ONCE) where a key is evicted while still in-use.
Fixes: a892c8d52c02 ("block: Inline encryption support for blk-mq")
Cc: stable@vger.kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
block/blk-crypto-internal.h | 25 +++++++++++++++++++++----
block/blk-crypto.c | 24 ++++++++++++------------
block/blk-merge.c | 2 ++
block/blk-mq.c | 15 ++++++++++++++-
4 files changed, 49 insertions(+), 17 deletions(-)
diff --git a/block/blk-crypto-internal.h b/block/blk-crypto-internal.h
index a8cdaf26851e..4f1de2495f0c 100644
--- a/block/blk-crypto-internal.h
+++ b/block/blk-crypto-internal.h
@@ -65,6 +65,11 @@ static inline bool blk_crypto_rq_is_encrypted(struct request *rq)
return rq->crypt_ctx;
}
+static inline bool blk_crypto_rq_has_keyslot(struct request *rq)
+{
+ return rq->crypt_keyslot;
+}
+
blk_status_t blk_crypto_get_keyslot(struct blk_crypto_profile *profile,
const struct blk_crypto_key *key,
struct blk_crypto_keyslot **slot_ptr);
@@ -119,6 +124,11 @@ static inline bool blk_crypto_rq_is_encrypted(struct request *rq)
return false;
}
+static inline bool blk_crypto_rq_has_keyslot(struct request *rq)
+{
+ return false;
+}
+
#endif /* CONFIG_BLK_INLINE_ENCRYPTION */
void __bio_crypt_advance(struct bio *bio, unsigned int bytes);
@@ -153,14 +163,21 @@ static inline bool blk_crypto_bio_prep(struct bio **bio_ptr)
return true;
}
-blk_status_t __blk_crypto_init_request(struct request *rq);
-static inline blk_status_t blk_crypto_init_request(struct request *rq)
+blk_status_t __blk_crypto_rq_get_keyslot(struct request *rq);
+static inline blk_status_t blk_crypto_rq_get_keyslot(struct request *rq)
{
if (blk_crypto_rq_is_encrypted(rq))
- return __blk_crypto_init_request(rq);
+ return __blk_crypto_rq_get_keyslot(rq);
return BLK_STS_OK;
}
+void __blk_crypto_rq_put_keyslot(struct request *rq);
+static inline void blk_crypto_rq_put_keyslot(struct request *rq)
+{
+ if (blk_crypto_rq_has_keyslot(rq))
+ __blk_crypto_rq_put_keyslot(rq);
+}
+
void __blk_crypto_free_request(struct request *rq);
static inline void blk_crypto_free_request(struct request *rq)
{
@@ -199,7 +216,7 @@ static inline blk_status_t blk_crypto_insert_cloned_request(struct request *rq)
{
if (blk_crypto_rq_is_encrypted(rq))
- return blk_crypto_init_request(rq);
+ return blk_crypto_rq_get_keyslot(rq);
return BLK_STS_OK;
}
diff --git a/block/blk-crypto.c b/block/blk-crypto.c
index 45378586151f..d0c7feb447e9 100644
--- a/block/blk-crypto.c
+++ b/block/blk-crypto.c
@@ -224,27 +224,27 @@ static bool bio_crypt_check_alignment(struct bio *bio)
return true;
}
-blk_status_t __blk_crypto_init_request(struct request *rq)
+blk_status_t __blk_crypto_rq_get_keyslot(struct request *rq)
{
return blk_crypto_get_keyslot(rq->q->crypto_profile,
rq->crypt_ctx->bc_key,
&rq->crypt_keyslot);
}
-/**
- * __blk_crypto_free_request - Uninitialize the crypto fields of a request.
- *
- * @rq: The request whose crypto fields to uninitialize.
- *
- * Completely uninitializes the crypto fields of a request. If a keyslot has
- * been programmed into some inline encryption hardware, that keyslot is
- * released. The rq->crypt_ctx is also freed.
- */
-void __blk_crypto_free_request(struct request *rq)
+void __blk_crypto_rq_put_keyslot(struct request *rq)
{
blk_crypto_put_keyslot(rq->crypt_keyslot);
+ rq->crypt_keyslot = NULL;
+}
+
+void __blk_crypto_free_request(struct request *rq)
+{
+ /* The keyslot, if one was needed, should have been released earlier. */
+ if (WARN_ON_ONCE(rq->crypt_keyslot))
+ __blk_crypto_rq_put_keyslot(rq);
+
mempool_free(rq->crypt_ctx, bio_crypt_ctx_pool);
- blk_crypto_rq_set_defaults(rq);
+ rq->crypt_ctx = NULL;
}
/**
diff --git a/block/blk-merge.c b/block/blk-merge.c
index 6460abdb2426..65e75efa9bd3 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -867,6 +867,8 @@ static struct request *attempt_merge(struct request_queue *q,
if (!blk_discard_mergable(req))
elv_merge_requests(q, req, next);
+ blk_crypto_rq_put_keyslot(next);
+
/*
* 'next' is going away, so update stats accordingly
*/
diff --git a/block/blk-mq.c b/block/blk-mq.c
index d0cb2ef18fe2..49825538d932 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -840,6 +840,12 @@ static void blk_complete_request(struct request *req)
req->q->integrity.profile->complete_fn(req, total_bytes);
#endif
+ /*
+ * Upper layers may call blk_crypto_evict_key() anytime after the last
+ * bio_endio(). Therefore, the keyslot must be released before that.
+ */
+ blk_crypto_rq_put_keyslot(req);
+
blk_account_io_completion(req, total_bytes);
do {
@@ -905,6 +911,13 @@ bool blk_update_request(struct request *req, blk_status_t error,
req->q->integrity.profile->complete_fn(req, nr_bytes);
#endif
+ /*
+ * Upper layers may call blk_crypto_evict_key() anytime after the last
+ * bio_endio(). Therefore, the keyslot must be released before that.
+ */
+ if (blk_crypto_rq_has_keyslot(req) && nr_bytes >= blk_rq_bytes(req))
+ __blk_crypto_rq_put_keyslot(req);
+
if (unlikely(error && !blk_rq_is_passthrough(req) &&
!(req->rq_flags & RQF_QUIET)) &&
!test_bit(GD_DEAD, &req->q->disk->state)) {
@@ -2967,7 +2980,7 @@ void blk_mq_submit_bio(struct bio *bio)
blk_mq_bio_to_request(rq, bio, nr_segs);
- ret = blk_crypto_init_request(rq);
+ ret = blk_crypto_rq_get_keyslot(rq);
if (ret != BLK_STS_OK) {
bio->bi_status = ret;
bio_endio(bio);
--
2.39.2
^ permalink raw reply related [flat|nested] 14+ messages in thread
* [PATCH v2 2/4] blk-crypto: make blk_crypto_evict_key() more robust
2023-03-08 19:36 [PATCH v2 0/4] Fix blk-crypto keyslot race condition Eric Biggers
2023-03-08 19:36 ` [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete Eric Biggers
@ 2023-03-08 19:36 ` Eric Biggers
2023-03-15 16:23 ` Christoph Hellwig
2023-03-08 19:36 ` [PATCH v2 3/4] blk-crypto: remove blk_crypto_insert_cloned_request() Eric Biggers
2023-03-08 19:36 ` [PATCH v2 4/4] blk-crypto: drop the NULL check from blk_crypto_put_keyslot() Eric Biggers
3 siblings, 1 reply; 14+ messages in thread
From: Eric Biggers @ 2023-03-08 19:36 UTC (permalink / raw)
To: linux-block, Jens Axboe; +Cc: linux-fscrypt, Nathan Huckleberry, stable
From: Eric Biggers <ebiggers@google.com>
If blk_crypto_evict_key() sees that the key is still in-use (due to a
bug) or that ->keyslot_evict failed, it currently just returns an error
while leaving the key linked into the keyslot management structures.
However, blk_crypto_evict_key() is only called in contexts such as inode
eviction where failure is not an option. So actually the caller
proceeds with freeing the blk_crypto_key regardless of the return value
of blk_crypto_evict_key().
These two assumptions don't match, and the result is that there can be a
use-after-free in blk_crypto_reprogram_all_keys() after one of these
errors occurs. (Note, these errors *shouldn't* happen; we're just
talking about what happens if they do anyway.)
Fix this by making blk_crypto_evict_key() unlink the key from the
keyslot management structures even on failure.
Fixes: 1b2628397058 ("block: Keyslot Manager for Inline Encryption")
Cc: stable@vger.kernel.org
Reviewed-by: Nathan Huckleberry <nhuck@google.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
block/blk-crypto-profile.c | 50 +++++++++++++++-----------------------
block/blk-crypto.c | 23 +++++++++++-------
2 files changed, 33 insertions(+), 40 deletions(-)
diff --git a/block/blk-crypto-profile.c b/block/blk-crypto-profile.c
index 0307fb0d95d3..1b20ead59f39 100644
--- a/block/blk-crypto-profile.c
+++ b/block/blk-crypto-profile.c
@@ -354,22 +354,10 @@ bool __blk_crypto_cfg_supported(struct blk_crypto_profile *profile,
return true;
}
-/**
- * __blk_crypto_evict_key() - Evict a key from a device.
- * @profile: the crypto profile of the device
- * @key: the key to evict. It must not still be used in any I/O.
- *
- * If the device has keyslots, this finds the keyslot (if any) that contains the
- * specified key and calls the driver's keyslot_evict function to evict it.
- *
- * Otherwise, this just calls the driver's keyslot_evict function if it is
- * implemented, passing just the key (without any particular keyslot). This
- * allows layered devices to evict the key from their underlying devices.
- *
- * Context: Process context. Takes and releases profile->lock.
- * Return: 0 on success or if there's no keyslot with the specified key, -EBUSY
- * if the keyslot is still in use, or another -errno value on other
- * error.
+/*
+ * This is an internal function that evicts a key from an inline encryption
+ * device that can be either a real device or the blk-crypto-fallback "device".
+ * It is used only by blk_crypto_evict_key(); see that function for details.
*/
int __blk_crypto_evict_key(struct blk_crypto_profile *profile,
const struct blk_crypto_key *key)
@@ -389,22 +377,22 @@ int __blk_crypto_evict_key(struct blk_crypto_profile *profile,
blk_crypto_hw_enter(profile);
slot = blk_crypto_find_keyslot(profile, key);
- if (!slot)
- goto out_unlock;
-
- if (WARN_ON_ONCE(atomic_read(&slot->slot_refs) != 0)) {
- err = -EBUSY;
- goto out_unlock;
+ if (slot) {
+ if (WARN_ON_ONCE(atomic_read(&slot->slot_refs) != 0)) {
+ /* BUG: key is still in use by I/O */
+ err = -EBUSY;
+ } else {
+ err = profile->ll_ops.keyslot_evict(
+ profile, key,
+ blk_crypto_keyslot_index(slot));
+ }
+ /*
+ * Callers may free the key even on error, so unlink the key
+ * from the hash table and clear slot->key even on error.
+ */
+ hlist_del(&slot->hash_node);
+ slot->key = NULL;
}
- err = profile->ll_ops.keyslot_evict(profile, key,
- blk_crypto_keyslot_index(slot));
- if (err)
- goto out_unlock;
-
- hlist_del(&slot->hash_node);
- slot->key = NULL;
- err = 0;
-out_unlock:
blk_crypto_hw_exit(profile);
return err;
}
diff --git a/block/blk-crypto.c b/block/blk-crypto.c
index d0c7feb447e9..4e26fac64199 100644
--- a/block/blk-crypto.c
+++ b/block/blk-crypto.c
@@ -399,17 +399,22 @@ int blk_crypto_start_using_key(struct block_device *bdev,
}
/**
- * blk_crypto_evict_key() - Evict a key from any inline encryption hardware
- * it may have been programmed into
- * @bdev: The block_device who's associated inline encryption hardware this key
- * might have been programmed into
- * @key: The key to evict
+ * blk_crypto_evict_key() - Evict a blk_crypto_key from a block_device
+ * @bdev: a block_device on which I/O using the key may have been done
+ * @key: the key to evict
*
- * Upper layers (filesystems) must call this function to ensure that a key is
- * evicted from any hardware that it might have been programmed into. The key
- * must not be in use by any in-flight IO when this function is called.
+ * For a given block_device, this function removes the given blk_crypto_key from
+ * the keyslot management structures and evicts it from any underlying hardware
+ * keyslot(s) or blk-crypto-fallback keyslot it may have been programmed into.
*
- * Return: 0 on success or if the key wasn't in any keyslot; -errno on error.
+ * Upper layers must call this before freeing the blk_crypto_key. It must be
+ * called for every block_device the key may have been used on. The key must no
+ * longer be in use by any I/O when this function is called.
+ *
+ * Context: May sleep.
+ * Return: 0 on success or if the key wasn't in any keyslot; -errno if the key
+ * failed to be evicted from a keyslot or is still in-use. Even on
+ * "failure", the key is removed from the keyslot management structures.
*/
int blk_crypto_evict_key(struct block_device *bdev,
const struct blk_crypto_key *key)
--
2.39.2
^ permalink raw reply related [flat|nested] 14+ messages in thread
* [PATCH v2 3/4] blk-crypto: remove blk_crypto_insert_cloned_request()
2023-03-08 19:36 [PATCH v2 0/4] Fix blk-crypto keyslot race condition Eric Biggers
2023-03-08 19:36 ` [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete Eric Biggers
2023-03-08 19:36 ` [PATCH v2 2/4] blk-crypto: make blk_crypto_evict_key() more robust Eric Biggers
@ 2023-03-08 19:36 ` Eric Biggers
2023-03-15 16:24 ` Christoph Hellwig
2023-03-08 19:36 ` [PATCH v2 4/4] blk-crypto: drop the NULL check from blk_crypto_put_keyslot() Eric Biggers
3 siblings, 1 reply; 14+ messages in thread
From: Eric Biggers @ 2023-03-08 19:36 UTC (permalink / raw)
To: linux-block, Jens Axboe; +Cc: linux-fscrypt, Nathan Huckleberry
From: Eric Biggers <ebiggers@google.com>
blk_crypto_insert_cloned_request() is the same as
blk_crypto_rq_get_keyslot(), so just use that directly.
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
Documentation/block/inline-encryption.rst | 3 +--
block/blk-crypto-internal.h | 15 ---------------
block/blk-mq.c | 2 +-
3 files changed, 2 insertions(+), 18 deletions(-)
diff --git a/Documentation/block/inline-encryption.rst b/Documentation/block/inline-encryption.rst
index f9bf18ea6509..90b733422ed4 100644
--- a/Documentation/block/inline-encryption.rst
+++ b/Documentation/block/inline-encryption.rst
@@ -270,8 +270,7 @@ Request queue based layered devices like dm-rq that wish to support inline
encryption need to create their own blk_crypto_profile for their request_queue,
and expose whatever functionality they choose. When a layered device wants to
pass a clone of that request to another request_queue, blk-crypto will
-initialize and prepare the clone as necessary; see
-``blk_crypto_insert_cloned_request()``.
+initialize and prepare the clone as necessary.
Interaction between inline encryption and blk integrity
=======================================================
diff --git a/block/blk-crypto-internal.h b/block/blk-crypto-internal.h
index 4f1de2495f0c..93a141979694 100644
--- a/block/blk-crypto-internal.h
+++ b/block/blk-crypto-internal.h
@@ -205,21 +205,6 @@ static inline int blk_crypto_rq_bio_prep(struct request *rq, struct bio *bio,
return 0;
}
-/**
- * blk_crypto_insert_cloned_request - Prepare a cloned request to be inserted
- * into a request queue.
- * @rq: the request being queued
- *
- * Return: BLK_STS_OK on success, nonzero on error.
- */
-static inline blk_status_t blk_crypto_insert_cloned_request(struct request *rq)
-{
-
- if (blk_crypto_rq_is_encrypted(rq))
- return blk_crypto_rq_get_keyslot(rq);
- return BLK_STS_OK;
-}
-
#ifdef CONFIG_BLK_INLINE_ENCRYPTION_FALLBACK
int blk_crypto_fallback_start_using_mode(enum blk_crypto_mode_num mode_num);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 49825538d932..5e819de2f5e7 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -3049,7 +3049,7 @@ blk_status_t blk_insert_cloned_request(struct request *rq)
if (q->disk && should_fail_request(q->disk->part0, blk_rq_bytes(rq)))
return BLK_STS_IOERR;
- if (blk_crypto_insert_cloned_request(rq))
+ if (blk_crypto_rq_get_keyslot(rq))
return BLK_STS_IOERR;
blk_account_io_start(rq);
--
2.39.2
^ permalink raw reply related [flat|nested] 14+ messages in thread
* [PATCH v2 4/4] blk-crypto: drop the NULL check from blk_crypto_put_keyslot()
2023-03-08 19:36 [PATCH v2 0/4] Fix blk-crypto keyslot race condition Eric Biggers
` (2 preceding siblings ...)
2023-03-08 19:36 ` [PATCH v2 3/4] blk-crypto: remove blk_crypto_insert_cloned_request() Eric Biggers
@ 2023-03-08 19:36 ` Eric Biggers
2023-03-15 16:24 ` Christoph Hellwig
3 siblings, 1 reply; 14+ messages in thread
From: Eric Biggers @ 2023-03-08 19:36 UTC (permalink / raw)
To: linux-block, Jens Axboe; +Cc: linux-fscrypt, Nathan Huckleberry
From: Eric Biggers <ebiggers@google.com>
Now that all callers of blk_crypto_put_keyslot() check for NULL before
calling it, there is no need for blk_crypto_put_keyslot() to do the NULL
check itself.
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
block/blk-crypto-profile.c | 14 ++++----------
1 file changed, 4 insertions(+), 10 deletions(-)
diff --git a/block/blk-crypto-profile.c b/block/blk-crypto-profile.c
index 1b20ead59f39..6c16edfa0dee 100644
--- a/block/blk-crypto-profile.c
+++ b/block/blk-crypto-profile.c
@@ -227,14 +227,13 @@ EXPORT_SYMBOL_GPL(blk_crypto_keyslot_index);
* @profile: the crypto profile of the device the key will be used on
* @key: the key that will be used
* @slot_ptr: If a keyslot is allocated, an opaque pointer to the keyslot struct
- * will be stored here; otherwise NULL will be stored here.
+ * will be stored here. blk_crypto_put_keyslot() must be called
+ * later to release it. Otherwise, NULL will be stored here.
*
* If the device has keyslots, this gets a keyslot that's been programmed with
* the specified key. If the key is already in a slot, this reuses it;
* otherwise this waits for a slot to become idle and programs the key into it.
*
- * This must be paired with a call to blk_crypto_put_keyslot().
- *
* Context: Process context. Takes and releases profile->lock.
* Return: BLK_STS_OK on success, meaning that either a keyslot was allocated or
* one wasn't needed; or a blk_status_t error on failure.
@@ -312,20 +311,15 @@ blk_status_t blk_crypto_get_keyslot(struct blk_crypto_profile *profile,
/**
* blk_crypto_put_keyslot() - Release a reference to a keyslot
- * @slot: The keyslot to release the reference of (may be NULL).
+ * @slot: The keyslot to release the reference of
*
* Context: Any context.
*/
void blk_crypto_put_keyslot(struct blk_crypto_keyslot *slot)
{
- struct blk_crypto_profile *profile;
+ struct blk_crypto_profile *profile = slot->profile;
unsigned long flags;
- if (!slot)
- return;
-
- profile = slot->profile;
-
if (atomic_dec_and_lock_irqsave(&slot->slot_refs,
&profile->idle_slots_lock, flags)) {
list_add_tail(&slot->idle_slot_node, &profile->idle_slots);
--
2.39.2
^ permalink raw reply related [flat|nested] 14+ messages in thread
* Re: [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete
2023-03-08 19:36 ` [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete Eric Biggers
@ 2023-03-13 21:26 ` Nathan Huckleberry
2023-03-14 18:20 ` Eric Biggers
2023-03-15 16:19 ` Christoph Hellwig
1 sibling, 1 reply; 14+ messages in thread
From: Nathan Huckleberry @ 2023-03-13 21:26 UTC (permalink / raw)
To: Eric Biggers; +Cc: linux-block, Jens Axboe, linux-fscrypt, stable
On Wed, Mar 8, 2023 at 11:39 AM Eric Biggers <ebiggers@kernel.org> wrote:
>
> From: Eric Biggers <ebiggers@google.com>
>
> Once all I/O using a blk_crypto_key has completed, filesystems can call
> blk_crypto_evict_key(). However, the block layer currently doesn't call
> blk_crypto_put_keyslot() until the request is being freed, which happens
> after upper layers have been told (via bio_endio()) the I/O has
> completed. This causes a race condition where blk_crypto_evict_key()
> can see 'slot_refs != 0' without there being an actual bug.
>
> This makes __blk_crypto_evict_key() hit the
> 'WARN_ON_ONCE(atomic_read(&slot->slot_refs) != 0)' and return without
> doing anything, eventually causing a use-after-free in
> blk_crypto_reprogram_all_keys(). (This is a very rare bug and has only
> been seen when per-file keys are being used with fscrypt.)
>
> There are two options to fix this: either release the keyslot before
> bio_endio() is called on the request's last bio, or make
> __blk_crypto_evict_key() ignore slot_refs. Let's go with the first
> solution, since it preserves the ability to report bugs (via
> WARN_ON_ONCE) where a key is evicted while still in-use.
>
> Fixes: a892c8d52c02 ("block: Inline encryption support for blk-mq")
> Cc: stable@vger.kernel.org
> Signed-off-by: Eric Biggers <ebiggers@google.com>
> ---
> block/blk-crypto-internal.h | 25 +++++++++++++++++++++----
> block/blk-crypto.c | 24 ++++++++++++------------
> block/blk-merge.c | 2 ++
> block/blk-mq.c | 15 ++++++++++++++-
> 4 files changed, 49 insertions(+), 17 deletions(-)
>
> diff --git a/block/blk-crypto-internal.h b/block/blk-crypto-internal.h
> index a8cdaf26851e..4f1de2495f0c 100644
> --- a/block/blk-crypto-internal.h
> +++ b/block/blk-crypto-internal.h
> @@ -65,6 +65,11 @@ static inline bool blk_crypto_rq_is_encrypted(struct request *rq)
> return rq->crypt_ctx;
> }
>
> +static inline bool blk_crypto_rq_has_keyslot(struct request *rq)
> +{
> + return rq->crypt_keyslot;
> +}
> +
> blk_status_t blk_crypto_get_keyslot(struct blk_crypto_profile *profile,
> const struct blk_crypto_key *key,
> struct blk_crypto_keyslot **slot_ptr);
> @@ -119,6 +124,11 @@ static inline bool blk_crypto_rq_is_encrypted(struct request *rq)
> return false;
> }
>
> +static inline bool blk_crypto_rq_has_keyslot(struct request *rq)
> +{
> + return false;
> +}
> +
> #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
>
> void __bio_crypt_advance(struct bio *bio, unsigned int bytes);
> @@ -153,14 +163,21 @@ static inline bool blk_crypto_bio_prep(struct bio **bio_ptr)
> return true;
> }
>
> -blk_status_t __blk_crypto_init_request(struct request *rq);
> -static inline blk_status_t blk_crypto_init_request(struct request *rq)
> +blk_status_t __blk_crypto_rq_get_keyslot(struct request *rq);
> +static inline blk_status_t blk_crypto_rq_get_keyslot(struct request *rq)
> {
> if (blk_crypto_rq_is_encrypted(rq))
> - return __blk_crypto_init_request(rq);
> + return __blk_crypto_rq_get_keyslot(rq);
> return BLK_STS_OK;
> }
>
> +void __blk_crypto_rq_put_keyslot(struct request *rq);
> +static inline void blk_crypto_rq_put_keyslot(struct request *rq)
> +{
> + if (blk_crypto_rq_has_keyslot(rq))
> + __blk_crypto_rq_put_keyslot(rq);
> +}
> +
> void __blk_crypto_free_request(struct request *rq);
> static inline void blk_crypto_free_request(struct request *rq)
> {
> @@ -199,7 +216,7 @@ static inline blk_status_t blk_crypto_insert_cloned_request(struct request *rq)
> {
>
> if (blk_crypto_rq_is_encrypted(rq))
> - return blk_crypto_init_request(rq);
> + return blk_crypto_rq_get_keyslot(rq);
> return BLK_STS_OK;
> }
>
> diff --git a/block/blk-crypto.c b/block/blk-crypto.c
> index 45378586151f..d0c7feb447e9 100644
> --- a/block/blk-crypto.c
> +++ b/block/blk-crypto.c
> @@ -224,27 +224,27 @@ static bool bio_crypt_check_alignment(struct bio *bio)
> return true;
> }
>
> -blk_status_t __blk_crypto_init_request(struct request *rq)
> +blk_status_t __blk_crypto_rq_get_keyslot(struct request *rq)
> {
> return blk_crypto_get_keyslot(rq->q->crypto_profile,
> rq->crypt_ctx->bc_key,
> &rq->crypt_keyslot);
> }
>
> -/**
> - * __blk_crypto_free_request - Uninitialize the crypto fields of a request.
> - *
> - * @rq: The request whose crypto fields to uninitialize.
> - *
> - * Completely uninitializes the crypto fields of a request. If a keyslot has
> - * been programmed into some inline encryption hardware, that keyslot is
> - * released. The rq->crypt_ctx is also freed.
> - */
> -void __blk_crypto_free_request(struct request *rq)
> +void __blk_crypto_rq_put_keyslot(struct request *rq)
> {
> blk_crypto_put_keyslot(rq->crypt_keyslot);
> + rq->crypt_keyslot = NULL;
> +}
> +
> +void __blk_crypto_free_request(struct request *rq)
> +{
> + /* The keyslot, if one was needed, should have been released earlier. */
> + if (WARN_ON_ONCE(rq->crypt_keyslot))
> + __blk_crypto_rq_put_keyslot(rq);
> +
> mempool_free(rq->crypt_ctx, bio_crypt_ctx_pool);
> - blk_crypto_rq_set_defaults(rq);
> + rq->crypt_ctx = NULL;
> }
>
> /**
> diff --git a/block/blk-merge.c b/block/blk-merge.c
> index 6460abdb2426..65e75efa9bd3 100644
> --- a/block/blk-merge.c
> +++ b/block/blk-merge.c
> @@ -867,6 +867,8 @@ static struct request *attempt_merge(struct request_queue *q,
> if (!blk_discard_mergable(req))
> elv_merge_requests(q, req, next);
>
> + blk_crypto_rq_put_keyslot(next);
> +
This looks good to me, but it looks like there was a pre-existing bug
in the blk-merge code. The elv_merged_request function is only called
when the request does not merge. Does anyone know if that behavior is
correct?
> /*
> * 'next' is going away, so update stats accordingly
> */
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index d0cb2ef18fe2..49825538d932 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -840,6 +840,12 @@ static void blk_complete_request(struct request *req)
> req->q->integrity.profile->complete_fn(req, total_bytes);
> #endif
>
> + /*
> + * Upper layers may call blk_crypto_evict_key() anytime after the last
> + * bio_endio(). Therefore, the keyslot must be released before that.
> + */
> + blk_crypto_rq_put_keyslot(req);
> +
> blk_account_io_completion(req, total_bytes);
>
> do {
> @@ -905,6 +911,13 @@ bool blk_update_request(struct request *req, blk_status_t error,
> req->q->integrity.profile->complete_fn(req, nr_bytes);
> #endif
>
> + /*
> + * Upper layers may call blk_crypto_evict_key() anytime after the last
> + * bio_endio(). Therefore, the keyslot must be released before that.
> + */
> + if (blk_crypto_rq_has_keyslot(req) && nr_bytes >= blk_rq_bytes(req))
> + __blk_crypto_rq_put_keyslot(req);
> +
> if (unlikely(error && !blk_rq_is_passthrough(req) &&
> !(req->rq_flags & RQF_QUIET)) &&
> !test_bit(GD_DEAD, &req->q->disk->state)) {
> @@ -2967,7 +2980,7 @@ void blk_mq_submit_bio(struct bio *bio)
>
> blk_mq_bio_to_request(rq, bio, nr_segs);
>
> - ret = blk_crypto_init_request(rq);
> + ret = blk_crypto_rq_get_keyslot(rq);
> if (ret != BLK_STS_OK) {
> bio->bi_status = ret;
> bio_endio(bio);
> --
> 2.39.2
>
This patch itself looks good to me.
Reviewed-by: Nathan Huckleberry <nhuck@google.com>
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete
2023-03-13 21:26 ` Nathan Huckleberry
@ 2023-03-14 18:20 ` Eric Biggers
2023-03-15 16:18 ` Christoph Hellwig
0 siblings, 1 reply; 14+ messages in thread
From: Eric Biggers @ 2023-03-14 18:20 UTC (permalink / raw)
To: Nathan Huckleberry
Cc: linux-block, Jens Axboe, linux-fscrypt, stable, Christoph Hellwig
On Mon, Mar 13, 2023 at 02:26:00PM -0700, Nathan Huckleberry wrote:
> > diff --git a/block/blk-merge.c b/block/blk-merge.c
> > index 6460abdb2426..65e75efa9bd3 100644
> > --- a/block/blk-merge.c
> > +++ b/block/blk-merge.c
> > @@ -867,6 +867,8 @@ static struct request *attempt_merge(struct request_queue *q,
> > if (!blk_discard_mergable(req))
> > elv_merge_requests(q, req, next);
> >
> > + blk_crypto_rq_put_keyslot(next);
> > +
>
> This looks good to me, but it looks like there was a pre-existing bug
> in the blk-merge code. The elv_merged_request function is only called
> when the request does not merge. Does anyone know if that behavior is
> correct?
That's very confusing to me too!
I did notice that attempt_merge() calls elv_merge_requests() (not to be confused
with elv_merged_request()) if it merges the requests.
So it seems there is elv_merge_requests() which means the request was merged,
and elv_merged_request() which means the request was *not* merged... I have no
idea what is going on there :-(
> This patch itself looks good to me.
>
> Reviewed-by: Nathan Huckleberry <nhuck@google.com>
Thanks.
Jens, Christoph, etc., anyone else want to take a look too?
- Eric
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete
2023-03-14 18:20 ` Eric Biggers
@ 2023-03-15 16:18 ` Christoph Hellwig
0 siblings, 0 replies; 14+ messages in thread
From: Christoph Hellwig @ 2023-03-15 16:18 UTC (permalink / raw)
To: Eric Biggers
Cc: Nathan Huckleberry, linux-block, Jens Axboe, linux-fscrypt, stable
On Tue, Mar 14, 2023 at 11:20:14AM -0700, Eric Biggers wrote:
> I did notice that attempt_merge() calls elv_merge_requests() (not to be confused
> with elv_merged_request()) if it merges the requests.
>
> So it seems there is elv_merge_requests() which means the request was merged,
> and elv_merged_request() which means the request was *not* merged... I have no
> idea what is going on there :-(
The naming looks very confusing, but that is indeed what the code
does. The elevator code is in massive need of a cleanup, as it often
is that confusing.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete
2023-03-08 19:36 ` [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete Eric Biggers
2023-03-13 21:26 ` Nathan Huckleberry
@ 2023-03-15 16:19 ` Christoph Hellwig
1 sibling, 0 replies; 14+ messages in thread
From: Christoph Hellwig @ 2023-03-15 16:19 UTC (permalink / raw)
To: Eric Biggers
Cc: linux-block, Jens Axboe, linux-fscrypt, Nathan Huckleberry, stable
Looks good:
Reviewed-by: Christoph Hellwig <hch@lst.de>
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 2/4] blk-crypto: make blk_crypto_evict_key() more robust
2023-03-08 19:36 ` [PATCH v2 2/4] blk-crypto: make blk_crypto_evict_key() more robust Eric Biggers
@ 2023-03-15 16:23 ` Christoph Hellwig
2023-03-15 18:26 ` Eric Biggers
0 siblings, 1 reply; 14+ messages in thread
From: Christoph Hellwig @ 2023-03-15 16:23 UTC (permalink / raw)
To: Eric Biggers
Cc: linux-block, Jens Axboe, linux-fscrypt, Nathan Huckleberry, stable
On Wed, Mar 08, 2023 at 11:36:43AM -0800, Eric Biggers wrote:
> const struct blk_crypto_key *key)
> @@ -389,22 +377,22 @@ int __blk_crypto_evict_key(struct blk_crypto_profile *profile,
>
> blk_crypto_hw_enter(profile);
> slot = blk_crypto_find_keyslot(profile, key);
> + if (slot) {
Isn't !slot also an error condition that we should warn on?
> + if (WARN_ON_ONCE(atomic_read(&slot->slot_refs) != 0)) {
> + /* BUG: key is still in use by I/O */
> + err = -EBUSY;
Either way two gotos to jump forward for the error case would improve
the readability of the code a fair bit I think.
> + } else {
> + err = profile->ll_ops.keyslot_evict(
> + profile, key,
> + blk_crypto_keyslot_index(slot));
> + }
> + /*
> + * Callers may free the key even on error, so unlink the key
> + * from the hash table and clear slot->key even on error.
> + */
> + hlist_del(&slot->hash_node);
> + slot->key = NULL;
> }
> blk_crypto_hw_exit(profile);
> return err;
Also given your above accurate description of the calling context,
what is the point of even returning an error here vs just logging
an error message?
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 3/4] blk-crypto: remove blk_crypto_insert_cloned_request()
2023-03-08 19:36 ` [PATCH v2 3/4] blk-crypto: remove blk_crypto_insert_cloned_request() Eric Biggers
@ 2023-03-15 16:24 ` Christoph Hellwig
2023-03-15 18:27 ` Eric Biggers
0 siblings, 1 reply; 14+ messages in thread
From: Christoph Hellwig @ 2023-03-15 16:24 UTC (permalink / raw)
To: Eric Biggers; +Cc: linux-block, Jens Axboe, linux-fscrypt, Nathan Huckleberry
> - if (blk_crypto_insert_cloned_request(rq))
> + if (blk_crypto_rq_get_keyslot(rq))
> return BLK_STS_IOERR;
Wouldn't it be better to propagate the actual error from
blk_crypto_rq_get_keyslot?
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 4/4] blk-crypto: drop the NULL check from blk_crypto_put_keyslot()
2023-03-08 19:36 ` [PATCH v2 4/4] blk-crypto: drop the NULL check from blk_crypto_put_keyslot() Eric Biggers
@ 2023-03-15 16:24 ` Christoph Hellwig
0 siblings, 0 replies; 14+ messages in thread
From: Christoph Hellwig @ 2023-03-15 16:24 UTC (permalink / raw)
To: Eric Biggers; +Cc: linux-block, Jens Axboe, linux-fscrypt, Nathan Huckleberry
Looks good:
Reviewed-by: Christoph Hellwig <hch@lst.de>
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 2/4] blk-crypto: make blk_crypto_evict_key() more robust
2023-03-15 16:23 ` Christoph Hellwig
@ 2023-03-15 18:26 ` Eric Biggers
0 siblings, 0 replies; 14+ messages in thread
From: Eric Biggers @ 2023-03-15 18:26 UTC (permalink / raw)
To: Christoph Hellwig
Cc: linux-block, Jens Axboe, linux-fscrypt, Nathan Huckleberry, stable
On Wed, Mar 15, 2023 at 09:23:12AM -0700, Christoph Hellwig wrote:
> On Wed, Mar 08, 2023 at 11:36:43AM -0800, Eric Biggers wrote:
> > const struct blk_crypto_key *key)
> > @@ -389,22 +377,22 @@ int __blk_crypto_evict_key(struct blk_crypto_profile *profile,
> >
> > blk_crypto_hw_enter(profile);
> > slot = blk_crypto_find_keyslot(profile, key);
> > + if (slot) {
>
> Isn't !slot also an error condition that we should warn on?
No, definitely not. Keys are not guaranteed to be in a keyslot. I'll add a
comment here.
>
> > + if (WARN_ON_ONCE(atomic_read(&slot->slot_refs) != 0)) {
> > + /* BUG: key is still in use by I/O */
> > + err = -EBUSY;
>
> Either way two gotos to jump forward for the error case would improve
> the readability of the code a fair bit I think.
I slightly prefer it without the gotos, but sure I'll do it that way.
> > + } else {
> > + err = profile->ll_ops.keyslot_evict(
> > + profile, key,
> > + blk_crypto_keyslot_index(slot));
> > + }
> > + /*
> > + * Callers may free the key even on error, so unlink the key
> > + * from the hash table and clear slot->key even on error.
> > + */
> > + hlist_del(&slot->hash_node);
> > + slot->key = NULL;
> > }
> > blk_crypto_hw_exit(profile);
> > return err;
>
> Also given your above accurate description of the calling context,
> what is the point of even returning an error here vs just logging
> an error message?
Yes, I was thinking about that too. I'll add a patch that makes
blk_crypto_evict_key() log errors and return void.
- Eric
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v2 3/4] blk-crypto: remove blk_crypto_insert_cloned_request()
2023-03-15 16:24 ` Christoph Hellwig
@ 2023-03-15 18:27 ` Eric Biggers
0 siblings, 0 replies; 14+ messages in thread
From: Eric Biggers @ 2023-03-15 18:27 UTC (permalink / raw)
To: Christoph Hellwig
Cc: linux-block, Jens Axboe, linux-fscrypt, Nathan Huckleberry
On Wed, Mar 15, 2023 at 09:24:04AM -0700, Christoph Hellwig wrote:
> > - if (blk_crypto_insert_cloned_request(rq))
> > + if (blk_crypto_rq_get_keyslot(rq))
> > return BLK_STS_IOERR;
>
> Wouldn't it be better to propagate the actual error from
> blk_crypto_rq_get_keyslot?
>
I'll add a patch that does that.
- Eric
^ permalink raw reply [flat|nested] 14+ messages in thread
end of thread, other threads:[~2023-03-15 18:27 UTC | newest]
Thread overview: 14+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-03-08 19:36 [PATCH v2 0/4] Fix blk-crypto keyslot race condition Eric Biggers
2023-03-08 19:36 ` [PATCH v2 1/4] blk-mq: release crypto keyslot before reporting I/O complete Eric Biggers
2023-03-13 21:26 ` Nathan Huckleberry
2023-03-14 18:20 ` Eric Biggers
2023-03-15 16:18 ` Christoph Hellwig
2023-03-15 16:19 ` Christoph Hellwig
2023-03-08 19:36 ` [PATCH v2 2/4] blk-crypto: make blk_crypto_evict_key() more robust Eric Biggers
2023-03-15 16:23 ` Christoph Hellwig
2023-03-15 18:26 ` Eric Biggers
2023-03-08 19:36 ` [PATCH v2 3/4] blk-crypto: remove blk_crypto_insert_cloned_request() Eric Biggers
2023-03-15 16:24 ` Christoph Hellwig
2023-03-15 18:27 ` Eric Biggers
2023-03-08 19:36 ` [PATCH v2 4/4] blk-crypto: drop the NULL check from blk_crypto_put_keyslot() Eric Biggers
2023-03-15 16:24 ` Christoph Hellwig
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).