On Mon, 2022-10-31 at 10:04 -0600, Jens Axboe wrote: > On 10/31/22 7:41 AM, Dylan Yudaken wrote: > > @@ -1849,17 +1866,14 @@ inline struct file > > *io_file_get_fixed(struct io_kiocb *req, int fd, > > unsigned long file_ptr; > > > > io_ring_submit_lock(ctx, issue_flags); > > - > > - if (unlikely((unsigned int)fd >= ctx->nr_user_files)) > > - goto out; > > - fd = array_index_nospec(fd, ctx->nr_user_files); > > - file_ptr = io_fixed_file_slot(&ctx->file_table, fd)- > > >file_ptr; > > + file_ptr = __io_file_peek_fixed(req, fd); > > file = (struct file *) (file_ptr & FFS_MASK); > > file_ptr &= ~FFS_MASK; > > /* mask in overlapping REQ_F and FFS bits */ > > req->flags |= (file_ptr << REQ_F_SUPPORT_NOWAIT_BIT); > > io_req_set_rsrc_node(req, ctx, 0); > > -out: > > + WARN_ON_ONCE(file && !test_bit(fd, ctx- > > >file_table.bitmap)); > > + > > io_ring_submit_unlock(ctx, issue_flags); > > return file; > > } > > Is this WARN_ON_ONCE() a leftover from being originally based on a > tree > before: > > commit 4d5059512d283dab7372d282c2fbd43c7f5a2456 > Author: Pavel Begunkov <asml.silence@xxxxxxxxx> > Date: Sun Oct 16 21:30:49 2022 +0100 > > io_uring: kill hot path fixed file bitmap debug checks > > got added? Seems not related to the changes here otherwise. > Ah yes. That was a bad merge in that case with the "out:" label. I'll fix that in v2. I am assuming there will be some more comments. Thanks for pointing this out. Dylan
