Skip to contentSkip to conversation

drain rsrc fix

PATCH Applied Pavel Begunkov 2 participants 5 messages Apr 27, 2021 – Apr 28, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover drain rsrc fix

1/3 is a rsrc-related follow up/fix after Hao's drain fix. Other two are simple hardening.

Pavel Begunkov (3):
io_uring: fix drain with rsrc CQEs
io_uring: dont overlap internal and user req flags
io_uring: add more build check for uapi

 fs/io_uring.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)
Signature · 1 line
2.31.1
#1 lore raw

PATCH 1/3 io_uring: fix drain with rsrc CQEs

Resource emitted CQEs are not bound to requests, so fix up counters used for DRAIN/defer logic.

Fixes b60c8dce33895 ("io_uring: preparation for rsrc tagging")
---
 fs/io_uring.c | 1 +
 1 file changed, 1 insertion(+)
fs/io_uring.c +1 -0
index 63ff70587d4f..d3b7fe6ccb0e 100644
@@ -7536,6 +7536,7 @@ static void __io_rsrc_put_work(struct io_rsrc_node *ref_node)
 			io_ring_submit_lock(ctx, lock_ring);
 			spin_lock_irqsave(&ctx->completion_lock, flags);
 			io_cqring_fill_event(ctx, prsrc->tag, 0, 0);
+			ctx->cq_extra++;
 			io_commit_cqring(ctx);
 			spin_unlock_irqrestore(&ctx->completion_lock, flags);
 			io_cqring_ev_posted(ctx);
Signature · 1 line
2.31.1
#2 lore raw

PATCH 2/3 io_uring: dont overlap internal and user req flags

In reply to Pavel Begunkov

CQE flags take one byte that we store in req->flags together with other REQ_F_* internal flags. CQE flags are copied directly into req and then verified that requires some handling on failures, e.g. to make sure that that copy doesn't set some of the internal flags.

More all internal flags to take bits after the first byte, so we don't need extra handling and make it safer overall.

---
 fs/io_uring.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)
fs/io_uring.c +3 -6
index d3b7fe6ccb0e..3419548ccaf5 100644
@@ -702,7 +702,8 @@ enum {
 	REQ_F_FORCE_ASYNC_BIT	= IOSQE_ASYNC_BIT,
 	REQ_F_BUFFER_SELECT_BIT	= IOSQE_BUFFER_SELECT_BIT,
 
-	REQ_F_FAIL_LINK_BIT,
+	/* first byte is taken by user flags, shift it to not overlap */
+	REQ_F_FAIL_LINK_BIT	= 8,
 	REQ_F_INFLIGHT_BIT,
 	REQ_F_CUR_POS_BIT,
 	REQ_F_NOWAIT_BIT,
@@ -6503,14 +6504,10 @@ static int io_init_req(struct io_ring_ctx *ctx, struct io_kiocb *req,
 	req->work.creds = NULL;
 
 	/* enforce forwards compatibility on users */
-	if (unlikely(sqe_flags & ~SQE_VALID_FLAGS)) {
-		req->flags = 0;
+	if (unlikely(sqe_flags & ~SQE_VALID_FLAGS))
 		return -EINVAL;
-	}
-
 	if (unlikely(req->opcode >= IORING_OP_LAST))
 		return -EINVAL;
-
 	if (unlikely(!io_check_restriction(ctx, req, sqe_flags)))
 		return -EACCES;
 
Signature · 1 line
2.31.1
#3 lore raw

PATCH 3/3 io_uring: add more build check for uapi

In reply to Pavel Begunkov

Add a couple of BUILD_BUG_ON() checking some rsrc uapi structs and SQE flags.

---
 fs/io_uring.c | 7 +++++++
 1 file changed, 7 insertions(+)
fs/io_uring.c +7 -0
index 3419548ccaf5..a48b88b3e289 100644
@@ -10131,6 +10131,13 @@ static int __init io_uring_init(void)
 	BUILD_BUG_SQE_ELEM(42, __u16,  personality);
 	BUILD_BUG_SQE_ELEM(44, __s32,  splice_fd_in);
 
+	BUILD_BUG_ON(sizeof(struct io_uring_files_update) !=
+		     sizeof(struct io_uring_rsrc_update));
+	BUILD_BUG_ON(sizeof(struct io_uring_rsrc_update) >
+		     sizeof(struct io_uring_rsrc_update2));
+	/* should fit into one byte */
+	BUILD_BUG_ON(SQE_VALID_FLAGS >= (1 << 8));
+
 	BUILD_BUG_ON(ARRAY_SIZE(io_op_defs) != IORING_OP_LAST);
 	BUILD_BUG_ON(__REQ_F_LAST_BIT >= 8 * sizeof(int));
 	req_cachep = KMEM_CACHE(io_kiocb, SLAB_HWCACHE_ALIGN | SLAB_PANIC |
Signature · 1 line
2.31.1
maintainer Applied
#4 lore raw
In reply to Pavel Begunkov
Quoted 2 lines · Pavel Begunkov

1/3 is a rsrc-related follow up/fix after Hao's drain fix. Other two are simple hardening.

Applied, thanks.

Signature · 1 line
Jens Axboe