Re: [PATCH] [8/18] BKL-removal: Remove BKL from remote_llseek

[Steve French]

If two seeks overlap, can't you end up with an f_pos value that is
different than what either thread seeked to? or if you have a seek and
a read overlap can't you end up with the read occurring in the midst
of an update of f_pos (which takes more than one instruction on
various architectures), e.g. reading an f_pos, which has only the
lower half of a 64 bit field updated?   I agree that you shouldn't
have seeks racing in parallel but I think it is preferable to get
either the updated f_pos or the earlier f_pos not something 1/2
updated.

On Jan 27, 2008 11:56 AM, Trond Myklebust <Trond.Myklebust@xxxxxxxxxx> wrote:
>
> On Sun, 2008-01-27 at 10:57 -0600, Steve French wrote:
> > Don't you need to a spinlock/spinunlock(i_lock) or something similar
> > (there isn't a spinlock in the file struct unfortunately) around the
> > reads and writes from f_pos in fs/read_write.c in remote_llseek with
> > your patch since the reads/writes from that field are not necessarily
> > atomic and threads could be racing in seek on the same file struct?
>
> Where does is state in POSIX or SUS that we need to cater to that kind
> of application?
> In any case, the current behaviour of f_pos if two threads are sharing
> the file struct is undefined no matter whether you spinlock or not,
> since there is no special locking around sys_read() or sys_write().
>
> Trond
>



-- 
Thanks,

Steve
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/