io_uring: fix up O_NONBLOCK handling for sockets

We've got two issues with the non-regular file handling for non-blocking
IO:

1) We don't want to re-do a short read in full for a non-regular file,
   as we can't just read the data again.
2) For non-regular files that don't support non-blocking IO attempts,
   we need to punt to async context even if the file is opened as
   non-blocking. Otherwise the caller always gets -EAGAIN.

Add two new request flags to handle these cases. One is just a cache
of the inode S_ISREG() status, the other tells io_uring that we always
need to punt this request to async context, even if REQ_F_NOWAIT is set.

Cc: stable@vger.kernel.org
Reported-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Tested-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 39 insertions, 18 deletions

diff --git a/fs/io_uring.c b/fs/io_uring.c
index d2cb277da2f4..b7d4085d6ffd 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -322,6 +322,8 @@ struct io_kiocb {
 #define REQ_F_FAIL_LINK         256     /* fail rest of links */
 #define REQ_F_SHADOW_DRAIN      512     /* link-drain shadow req */
 #define REQ_F_TIMEOUT           1024    /* timeout request */
+#define REQ_F_ISREG             2048    /* regular file */
+#define REQ_F_MUST_PUNT         4096    /* must be punted even for NONBLOCK */
         u64                     user_data;
         u32                     result;
         u32                     sequence;
@@ -914,26 +916,26 @@ static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events,
         return ret;
 }
 
-static void kiocb_end_write(struct kiocb *kiocb)
+static void kiocb_end_write(struct io_kiocb *req)
 {
-        if (kiocb->ki_flags & IOCB_WRITE) {
-                struct inode *inode = file_inode(kiocb->ki_filp);
+        /*
+         * Tell lockdep we inherited freeze protection from submission
+         * thread.
+         */
+        if (req->flags & REQ_F_ISREG) {
+                struct inode *inode = file_inode(req->file);
 
-                /*
-                 * Tell lockdep we inherited freeze protection from submission
-                 * thread.
-                 */
-                if (S_ISREG(inode->i_mode))
-                        __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
-                file_end_write(kiocb->ki_filp);
+                __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
         }
+        file_end_write(req->file);
 }
 
 static void io_complete_rw(struct kiocb *kiocb, long res, long res2)
 {
         struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
 
-        kiocb_end_write(kiocb);
+        if (kiocb->ki_flags & IOCB_WRITE)
+                kiocb_end_write(req);
 
         if ((req->flags & REQ_F_LINK) && res != req->result)
                 req->flags |= REQ_F_FAIL_LINK;
@@ -945,7 +947,8 @@ static void io_complete_rw_iopoll(struct kiocb *kiocb, long res, long res2)
 {
         struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
 
-        kiocb_end_write(kiocb);
+        if (kiocb->ki_flags & IOCB_WRITE)
+                kiocb_end_write(req);
 
         if ((req->flags & REQ_F_LINK) && res != req->result)
                 req->flags |= REQ_F_FAIL_LINK;
@@ -1059,8 +1062,17 @@ static int io_prep_rw(struct io_kiocb *req, const struct sqe_submit *s,
         if (!req->file)
                 return -EBADF;
 
-        if (force_nonblock && !io_file_supports_async(req->file))
-                force_nonblock = false;
+        if (S_ISREG(file_inode(req->file)->i_mode))
+                req->flags |= REQ_F_ISREG;
+
+        /*
+         * If the file doesn't support async, mark it as REQ_F_MUST_PUNT so
+         * we know to async punt it even if it was opened O_NONBLOCK
+         */
+        if (force_nonblock && !io_file_supports_async(req->file)) {
+                req->flags |= REQ_F_MUST_PUNT;
+                return -EAGAIN;
+        }
 
         kiocb->ki_pos = READ_ONCE(sqe->off);
         kiocb->ki_flags = iocb_flags(kiocb->ki_filp);
@@ -1081,7 +1093,8 @@ static int io_prep_rw(struct io_kiocb *req, const struct sqe_submit *s,
                 return ret;
 
         /* don't allow async punt if RWF_NOWAIT was requested */
-        if (kiocb->ki_flags & IOCB_NOWAIT)
+        if ((kiocb->ki_flags & IOCB_NOWAIT) ||
+            (req->file->f_flags & O_NONBLOCK))
                 req->flags |= REQ_F_NOWAIT;
 
         if (force_nonblock)
@@ -1382,7 +1395,9 @@ static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
                  * need async punt anyway, so it's more efficient to do it
                  * here.
                  */
-                if (force_nonblock && ret2 > 0 && ret2 < read_size)
+                if (force_nonblock && !(req->flags & REQ_F_NOWAIT) &&
+                    (req->flags & REQ_F_ISREG) &&
+                    ret2 > 0 && ret2 < read_size)
                         ret2 = -EAGAIN;
                 /* Catch -EAGAIN return for forced non-blocking submission */
                 if (!force_nonblock || ret2 != -EAGAIN) {
@@ -1447,7 +1462,7 @@ static int io_write(struct io_kiocb *req, const struct sqe_submit *s,
                  * released so that it doesn't complain about the held lock when
                  * we return to userspace.
                  */
-                if (S_ISREG(file_inode(file)->i_mode)) {
+                if (req->flags & REQ_F_ISREG) {
                         __sb_start_write(file_inode(file)->i_sb,
                                                 SB_FREEZE_WRITE, true);
                         __sb_writers_release(file_inode(file)->i_sb,
@@ -2282,7 +2297,13 @@ static int __io_queue_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
         int ret;
 
         ret = __io_submit_sqe(ctx, req, s, force_nonblock);
-        if (ret == -EAGAIN && !(req->flags & REQ_F_NOWAIT)) {
+
+        /*
+         * We async punt it if the file wasn't marked NOWAIT, or if the file
+         * doesn't support non-blocking read/write attempts
+         */
+        if (ret == -EAGAIN && (!(req->flags & REQ_F_NOWAIT) ||
+            (req->flags & REQ_F_MUST_PUNT))) {
                 struct io_uring_sqe *sqe_copy;
 
                 sqe_copy = kmemdup(s->sqe, sizeof(*sqe_copy), GFP_KERNEL);