[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240403224037.GM538574@ZenIV>
Date: Wed, 3 Apr 2024 23:40:37 +0100
From: Al Viro <viro@...iv.linux.org.uk>
To: Jens Axboe <axboe@...nel.dk>
Cc: linux-fsdevel@...r.kernel.org, brauner@...nel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/3] timerfd: convert to ->read_iter()
On Wed, Apr 03, 2024 at 08:02:52AM -0600, Jens Axboe wrote:
> - res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
> + res = copy_to_iter(&ticks, sizeof(ticks), to);
Umm... That's not an equivalent transformation - different behaviour on
short copy; try to call it via read(fd, unmapped_buffer, 8) and see what
happens.
copy_to_iter() returns the amount copied; no data copied => return 0, not -EFAULT.
> + ufd = get_unused_fd_flags(O_RDWR | (flags & TFD_SHARED_FCNTL_FLAGS));
You do realize that get_unused_fd_flags() ignores O_RDWR (or O_NDELAY), right?
Mixing those with O_CLOEXEC makes sense for anon_inode_getfd(), but here you
have separate calls of get_unused_fd_flags() and anon_inode_getfile(), so...
Powered by blists - more mailing lists