[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAJfpegsTfUQ53hmnm7192-4ywLmXDLLwjV01tjCK7PVEqtE=yw@mail.gmail.com>
Date: Fri, 9 May 2025 16:59:34 +0200
From: Miklos Szeredi <miklos@...redi.hu>
To: Chen Linxuan <chenlinxuan@...ontech.com>
Cc: Amir Goldstein <amir73il@...il.com>, linux-fsdevel@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 2/3] fs: fuse: add backing_files control file
On Fri, 9 May 2025 at 16:43, Chen Linxuan <chenlinxuan@...ontech.com> wrote:
> Note that io_uring has already exposed information about fixed files
> in its fdinfo.
And that has limitations, since fdinfo lacks a hierarchy. E.g.
recursively retrieving "hidden files" info is impractical.
> For io_uring, the path could be the corresponding /proc/PID/fd/* of io_uring.
> For FUSE, the path could be /sys/fs/fuse/connections/*.
Since lsof is ultimately interested in the mapping between open files
and processes, I think using /proc/PID/fd/* for fuse is also useful.
> But for SCM_RIGHTS, what would the corresponding path be? Could it be
> the fd under procfs of unix domain socket?
Right. But I'm not asking you to implement this, just thinking about
an interface that is generic enough to cover past and possible future
cases.
> I am also uncertain whether there might be similar situations in the future.
> Would we really be able to find a suitable path for all of them?
Open file descriptors are generic enough to represent most kernel
objects that need to be represented in userspace, so I think this is
not a worry.
Thanks,
Miklos
Powered by blists - more mailing lists