Hi Viro,
Thank you for your replay.
> > Stack traceback for pid 2130
> > 0xf717f1b0 2130 1 1 0 R
> 0xf717f400 *irqbalance
> > ESP EIP Function (args)
> > 0xf75bfe38 0xc02d04b2 _spin_lock+0x2e (0xf7441a80)
> > 0xf75bff34 0xc015667c file_move+0x14 (0xf63080e4,
> 0xf75bff58, 0x0, 0xf74bf000)
> > 0xf75bff40 0xc0154e37 dentry_open+0xb9 (0xf63080e4,
> 0xf7f5ad80, 0xc02d00e6, 0x100100, 0x246)
> > 0xf75bff58 0xc0154d78 filp_open+0x36
> > 0xf75bffb4 0xc0155079 sys_open+0x31
> > 0xf75bffc4 0xc02d196f syscall_call+0x7
> >
> > The patch was made. Is this patch right?
> >
> > diff -urN linux-2.6.12-rc1.orig/fs/file_table.c
> linux-2.6.12-rc1/fs/file_table.c
> > --- linux-2.6.12-rc1.orig/fs/file_table.c 2005-03-02
> 16:37:47.000000000 +0900
> > +++ linux-2.6.12-rc1/fs/file_table.c 2005-03-31
> 17:50:46.323999320 +0900
> > @@ -209,11 +209,11 @@
> >
> > void file_kill(struct file *file)
> > {
> > + file_list_lock();
> > if (!list_empty(&file->f_list)) {
> > - file_list_lock();
> > list_del_init(&file->f_list);
> > - file_list_unlock();
> > }
> > + file_list_unlock();
> > }
>
> This is absolutely useless. What are you trying to protect
> and how the
> hell could keeping a lock around that check prevent any sort
> of deadlock?
I think that it is true not to be able to acquire file_list_lock()
that is called from file_move().
> Besides, who could possibly call fput() on struct file allocated by
> dentry_open() and do that before the latter returns a reference to
> that struct file?
Indeed, Is there a good method of debugging this issue?
In the check on the source, a doubtful place was not found except file_kill().
I might call not race of fput() and the dentry_open() but
the deadlock of file_list_lock().
--
Haruo
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
