Hi Eric, On Thu, Mar 2, 2023 at 11:23 PM Eric Biggers <ebiggers@xxxxxxxxxx> wrote: > > From: Eric Biggers <ebiggers@xxxxxxxxxx> > > 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@xxxxxxxxxxxxxxx > Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx> > --- > block/blk-crypto-internal.h | 15 +++++++++++---- > block/blk-crypto.c | 24 ++++++++++++------------ > block/blk-mq.c | 15 ++++++++++++++- > 3 files changed, 37 insertions(+), 17 deletions(-) > > diff --git a/block/blk-crypto-internal.h b/block/blk-crypto-internal.h > index a8cdaf26851e1..73609902349b6 100644 > --- a/block/blk-crypto-internal.h > +++ b/block/blk-crypto-internal.h > @@ -153,14 +153,21 @@ 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_is_encrypted(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 +206,7 @@ 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 45378586151f7..8e5612364c48c 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) > +{ > mempool_free(rq->crypt_ctx, bio_crypt_ctx_pool); > - blk_crypto_rq_set_defaults(rq); > + rq->crypt_ctx = NULL; > + > + /* The keyslot, if one was needed, should have been released earlier. */ > + if (WARN_ON_ONCE(rq->crypt_keyslot)) > + __blk_crypto_rq_put_keyslot(rq); > } > > /** > diff --git a/block/blk-mq.c b/block/blk-mq.c > index d3494a796ba80..738e81f518227 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_is_encrypted(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 > Thanks for the updated patchset. This patch looks good to me. Reviewed-by: Nathan Huckleberry <nhuck@xxxxxxxxxx> Thanks, Huck