[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAGudoHEBNRE+78n=WEY=Z0ZCnLmDFadisR-K2ah4SUO6uSm4TA@mail.gmail.com>
Date: Mon, 29 Jul 2024 13:06:28 +0200
From: Mateusz Guzik <mjguzik@...il.com>
To: Florian Weimer <fweimer@...hat.com>
Cc: linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-api@...r.kernel.org, Dave Chinner <dchinner@...hat.com>,
Christian Brauner <brauner@...nel.org>
Subject: Re: Testing if two open descriptors refer to the same inode
On Mon, Jul 29, 2024 at 12:57 PM Florian Weimer <fweimer@...hat.com> wrote:
>
> * Mateusz Guzik:
>
> > On Mon, Jul 29, 2024 at 12:40:35PM +0200, Florian Weimer wrote:
> >> * Mateusz Guzik:
> >>
> >> > On Mon, Jul 29, 2024 at 08:55:46AM +0200, Florian Weimer wrote:
> >> >> It was pointed out to me that inode numbers on Linux are no longer
> >> >> expected to be unique per file system, even for local file systems.
> >> >
> >> > I don't know if I'm parsing this correctly.
> >> >
> >> > Are you claiming on-disk inode numbers are not guaranteed unique per
> >> > filesystem? It sounds like utter breakage, with capital 'f'.
> >>
> >> Yes, POSIX semantics and traditional Linux semantics for POSIX-like
> >> local file systems are different.
> >
> > Can you link me some threads about this?
>
> Sorry, it was an internal thread. It's supposed to be common knowledge
> among Linux file system developers. Aleksa referenced LSF/MM
> discussions.
>
So much for open development :-P
> > I had this in mind (untested modulo compilation):
> >
> > diff --git a/fs/fcntl.c b/fs/fcntl.c
> > index 300e5d9ad913..5723c3e82eac 100644
> > --- a/fs/fcntl.c
> > +++ b/fs/fcntl.c
> > @@ -343,6 +343,13 @@ static long f_dupfd_query(int fd, struct file *filp)
> > return f.file == filp;
> > }
> >
> > +static long f_dupfd_query_inode(int fd, struct file *filp)
> > +{
> > + CLASS(fd_raw, f)(fd);
> > +
> > + return f.file->f_inode == filp->f_inode;
> > +}
> > +
> > static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
> > struct file *filp)
> > {
> > @@ -361,6 +368,9 @@ static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
> > case F_DUPFD_QUERY:
> > err = f_dupfd_query(argi, filp);
> > break;
> > + case F_DUPFD_QUERY_INODE:
> > + err = f_dupfd_query_inode(argi, filp);
> > + break;
> > case F_GETFD:
> > err = get_close_on_exec(fd) ? FD_CLOEXEC : 0;
> > break;
> > diff --git a/include/uapi/linux/fcntl.h b/include/uapi/linux/fcntl.h
> > index c0bcc185fa48..2e93dbdd8fd2 100644
> > --- a/include/uapi/linux/fcntl.h
> > +++ b/include/uapi/linux/fcntl.h
> > @@ -16,6 +16,8 @@
> >
> > #define F_DUPFD_QUERY (F_LINUX_SPECIFIC_BASE + 3)
> >
> > +#define F_DUPFD_QUERY_INODE (F_LINUX_SPECIFIC_BASE + 4)
> > +
> > /*
> > * Cancel a blocking posix lock; internal use only until we expose an
> > * asynchronous lock api to userspace:
>
> It's certainly much easier to use than name_to_handle_at, so it looks
> like a useful option to have.
>
> Could we return a three-way comparison result for sorting? Or would
> that expose too much about kernel pointer values?
>
As is this would sort by inode *address* which I don't believe is of
any use -- the order has to be assumed arbitrary.
Perhaps there is something which is reliably the same and can be
combined with something else to be unique system-wide (the magic
handle thing?).
But even then you would need to justify trying to sort by fcntl calls,
which sounds pretty dodgey to me.
Given that thing I *suspect* statx() may want to get extended with
some guaranteed unique identifier. Then you can sort in userspace all
you want.
Based on your opening mail I assumed you only need to check 2 files,
for which the proposed fcntl does the trick.
Or to put it differently: there seems to be more to the picture than
in the opening mail, so perhaps you could outline what you are looking
for.
--
Mateusz Guzik <mjguzik gmail.com>
Powered by blists - more mailing lists