fuse: cleanup fuse_dev_do_read()
authorMiklos Szeredi <mszeredi@suse.cz>
Wed, 1 Jul 2015 14:26:05 +0000 (16:26 +0200)
committerMiklos Szeredi <mszeredi@suse.cz>
Wed, 1 Jul 2015 14:26:05 +0000 (16:26 +0200)
 - locked list_add() + list_del_init() cancel out

 - common handling of case when request is ended here in the read phase

Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Reviewed-by: Ashish Samant <ashish.samant@oracle.com>
fs/fuse/dev.c

index 98568bd2d81bde6902789fc73d2cb82e72ca5a80..1ad75e4ceba52840c9da2a0c2150abd7b3cd5202 100644 (file)
@@ -1237,7 +1237,7 @@ __releases(fiq->waitq.lock)
 static ssize_t fuse_dev_do_read(struct fuse_conn *fc, struct file *file,
                                struct fuse_copy_state *cs, size_t nbytes)
 {
-       int err;
+       ssize_t err;
        struct fuse_iqueue *fiq = &fc->iq;
        struct fuse_pqueue *fpq = &fc->pq;
        struct fuse_req *req;
@@ -1280,8 +1280,6 @@ static ssize_t fuse_dev_do_read(struct fuse_conn *fc, struct file *file,
        spin_unlock(&fiq->waitq.lock);
 
        spin_lock(&fc->lock);
-       list_add(&req->list, &fpq->io);
-
        in = &req->in;
        reqsize = in->h.len;
        /* If request is too large, reply with an error and restart the read */
@@ -1290,10 +1288,10 @@ static ssize_t fuse_dev_do_read(struct fuse_conn *fc, struct file *file,
                /* SETXATTR is special, since it may contain too large data */
                if (in->h.opcode == FUSE_SETXATTR)
                        req->out.h.error = -E2BIG;
-               list_del_init(&req->list);
                request_end(fc, req);
                goto restart;
        }
+       list_add(&req->list, &fpq->io);
        spin_unlock(&fc->lock);
        cs->req = req;
        err = fuse_copy_one(cs, &in->h, sizeof(in->h));
@@ -1304,30 +1302,32 @@ static ssize_t fuse_dev_do_read(struct fuse_conn *fc, struct file *file,
        spin_lock(&fc->lock);
        clear_bit(FR_LOCKED, &req->flags);
        if (!fpq->connected) {
-               list_del_init(&req->list);
-               request_end(fc, req);
-               return -ENODEV;
+               err = -ENODEV;
+               goto out_end;
        }
        if (err) {
                req->out.h.error = -EIO;
-               list_del_init(&req->list);
-               request_end(fc, req);
-               return err;
+               goto out_end;
        }
        if (!test_bit(FR_ISREPLY, &req->flags)) {
-               list_del_init(&req->list);
-               request_end(fc, req);
-       } else {
-               list_move_tail(&req->list, &fpq->processing);
-               set_bit(FR_SENT, &req->flags);
-               /* matches barrier in request_wait_answer() */
-               smp_mb__after_atomic();
-               if (test_bit(FR_INTERRUPTED, &req->flags))
-                       queue_interrupt(fiq, req);
-               spin_unlock(&fc->lock);
+               err = reqsize;
+               goto out_end;
        }
+       list_move_tail(&req->list, &fpq->processing);
+       set_bit(FR_SENT, &req->flags);
+       /* matches barrier in request_wait_answer() */
+       smp_mb__after_atomic();
+       if (test_bit(FR_INTERRUPTED, &req->flags))
+               queue_interrupt(fiq, req);
+       spin_unlock(&fc->lock);
+
        return reqsize;
 
+out_end:
+       list_del_init(&req->list);
+       request_end(fc, req);
+       return err;
+
  err_unlock:
        spin_unlock(&fiq->waitq.lock);
        return err;