[PATCH] ext4: fix deadlock allocating bio_post_read_ctx from mempool

[PATCH] ext4: fix deadlock allocating bio_post_read_ctx from mempool

[Eric Biggers]

From: Eric Biggers <ebiggers@google.com>

Without any form of coordination, any case where multiple allocations
from the same mempool are needed at a time to make forward progress can
deadlock under memory pressure.

This is the case for struct bio_post_read_ctx, as one can be allocated
to decrypt a Merkle tree page during fsverity_verify_bio(), which itself
is running from a post-read callback for a data bio which has its own
struct bio_post_read_ctx.

Fix this by freeing the first bio_post_read_ctx before calling
fsverity_verify_bio().  This works because verity (if enabled) is always
the last post-read step.

This deadlock can be reproduced by trying to read from an encrypted
verity file after reducing NUM_PREALLOC_POST_READ_CTXS to 1 and patching
mempool_alloc() to pretend that pool->alloc() always fails.

Note that since NUM_PREALLOC_POST_READ_CTXS is actually 128, to actually
hit this bug in practice would require reading from lots of encrypted
verity files at the same time.  But it's theoretically possible, as N
available objects isn't enough to guarantee forward progress when > N/2
threads each need 2 objects at a time.

Fixes: 22cfe4b48ccb ("ext4: add fs-verity read support")
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
 fs/ext4/readpage.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/fs/ext4/readpage.c b/fs/ext4/readpage.c
index fef7755300c3..410c904cf59b 100644
--- a/fs/ext4/readpage.c
+++ b/fs/ext4/readpage.c
@@ -57,6 +57,7 @@ enum bio_post_read_step {
 	STEP_INITIAL = 0,
 	STEP_DECRYPT,
 	STEP_VERITY,
+	STEP_MAX,
 };
 
 struct bio_post_read_ctx {
@@ -106,10 +107,22 @@ static void verity_work(struct work_struct *work)
 {
 	struct bio_post_read_ctx *ctx =
 		container_of(work, struct bio_post_read_ctx, work);
+	struct bio *bio = ctx->bio;
 
-	fsverity_verify_bio(ctx->bio);
+	/*
+	 * fsverity_verify_bio() may call readpages() again, and although verity
+	 * will be disabled for that, decryption may still be needed, causing
+	 * another bio_post_read_ctx to be allocated.  So to guarantee that
+	 * mempool_alloc() never deadlocks we must free the current ctx first.
+	 * This is safe because verity is the last post-read step.
+	 */
+	BUILD_BUG_ON(STEP_VERITY + 1 != STEP_MAX);
+	mempool_free(ctx, bio_post_read_ctx_pool);
+	bio->bi_private = NULL;
 
-	bio_post_read_processing(ctx);
+	fsverity_verify_bio(bio);
+
+	__read_end_io(bio);
 }
 
 static void bio_post_read_processing(struct bio_post_read_ctx *ctx)
-- 
2.24.1

Re: [PATCH] ext4: fix deadlock allocating bio_post_read_ctx from mempool

[Theodore Y. Ts'o]

On Tue, Dec 31, 2019 at 12:12:22PM -0600, Eric Biggers wrote:
> From: Eric Biggers <ebiggers@google.com>
> 
> Without any form of coordination, any case where multiple allocations
> from the same mempool are needed at a time to make forward progress can
> deadlock under memory pressure.
> 
> This is the case for struct bio_post_read_ctx, as one can be allocated
> to decrypt a Merkle tree page during fsverity_verify_bio(), which itself
> is running from a post-read callback for a data bio which has its own
> struct bio_post_read_ctx.
> 
> Fix this by freeing the first bio_post_read_ctx before calling
> fsverity_verify_bio().  This works because verity (if enabled) is always
> the last post-read step.
> 
> This deadlock can be reproduced by trying to read from an encrypted
> verity file after reducing NUM_PREALLOC_POST_READ_CTXS to 1 and patching
> mempool_alloc() to pretend that pool->alloc() always fails.
> 
> Note that since NUM_PREALLOC_POST_READ_CTXS is actually 128, to actually
> hit this bug in practice would require reading from lots of encrypted
> verity files at the same time.  But it's theoretically possible, as N
> available objects isn't enough to guarantee forward progress when > N/2
> threads each need 2 objects at a time.
> 
> Fixes: 22cfe4b48ccb ("ext4: add fs-verity read support")
> Signed-off-by: Eric Biggers <ebiggers@google.com>

Thanks, applied.

						- Ted