Hi Manfred,
On Sat, 24 Jun 2000, Manfred Spraul wrote:
> Hi Tigran,
>
> From: "Tigran Aivazian" <tigran@veritas.com>
> >
> > the problem is not a deadlock but the fact that locking algorithms refer
> > to 'current' explicitly - see the problem now?
> >
> Where?
> The locking algorithm is:
~~~~~~~
we are talking about different meanings of the word "locking". You are
talking about locking in general (spinlocks etc.). I am talking about
userspace POSIX fcntl file region locking. Those "locking algorithms"
(look very deep down the stack from do_close()) refer to "current"
explicitly which makes the scheme you propose below unusable except in
current's context.
Regards,
Tigran
>
> * if you want to _change_ current->{fs,files}, you must be "current" _and_
> you must acquire the task_lock spinlock. See linux/kernel/exit.c,
> __exit_files()
> * if you want to read current->fs, you must either acquire the spinlock, or
> be "current".
>
> >
> > yes, it can from userspace. when you move to kernel space and to a
> > different context you must lock files->file_lock (which is what
> > put_unused_fd() or do_close() do and what the closefd.o module didn't),
> > otherwise you are broken :)
> >
> Obviously you must lock files_lock, but all necessary spinlocks exist:
>
> read_lock(&tasklist_lock);
> tsk = find_task_by_pid();
> get_task_struct(tsk);
> read_unlock(&tasklist_lock);
> task_lock(&tsk);
> files = tsk->files;
> atomic_inc(&files->count);
> task_unlock(&tsk);
> free_task_struct(tsk);
> __do_close(files, fd, release); /* do_close, but on files instead of
> * current->files */
> put_files_struct(files);
>
> If there is a flaw in the locking, we must fix it asap: proc uses that
> locking and several distros poll /proc/<pid>/* during shutdown. We'll see
> crashes if the locking is wrong.
>
> --
> Manfred
>
>
>
>
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jun 26 2000 - 21:00:04 EST