Re: Race between vmtruncate and mapped areas?

Daniel McNeil (daniel@osdl.org)
15 May 2003 09:38:26 -0700


This is a MIME-formatted message. If you see this text it means that your
E-mail software does not support MIME-formatted messages.

--=_courier-9137-1053016758-0001-2
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 7bit

On Thu, 2003-05-15 at 02:40, Andrea Arcangeli wrote:
> On Thu, May 15, 2003 at 02:20:00AM -0700, Andrew Morton wrote:
> > Andrea Arcangeli <andrea@suse.de> wrote:
> > >
> > > and it's still racy
> >
> > damn, and it just booted ;)
> >
> > I'm just a little bit concerned over the ever-expanding inode. Do you
> > think the dual sequence numbers can be replaced by a single generation
> > counter?
>
> yes, I wrote it as a single counter first, but was unreadable and it had
> more branches, so I added the other sequence number to make it cleaner.
> I don't mind another 4 bytes, that cacheline should be hot anyways.

You could use the seqlock.h sequence locking. It only uses 1 sequence
counter. The 2.5 isize patch 1 has a sequence lock without the spinlock
so it only uses 4 bytes and it is somewhat more readable. I don't
think it has more branches.

I've attached the isize seqlock.h patch.

-- 
Daniel McNeil <daniel@osdl.org>

--=_courier-9137-1053016758-0001-2 Content-Type: text/x-patch; name="patch-2.5.69-isize.1"; charset=iso-8859-1 Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename=patch-2.5.69-isize.1

diff -rupN -X /home/daniel/dontdiff linux-2.5.69/include/linux/seqlock.h linux-2.5.69-isize/include/linux/seqlock.h --- linux-2.5.69/include/linux/seqlock.h Sun May 4 16:53:14 2003 +++ linux-2.5.69-isize/include/linux/seqlock.h Wed May 7 16:00:25 2003 @@ -94,6 +94,57 @@ static inline int read_seqretry(const se return (iv & 1) | (sl->sequence ^ iv); } + +/* + * Version using sequence counter only. + * This can be used when code has its own mutex protecting the + * updating starting before the write_seqcntbeqin() and ending + * after the write_seqcntend(). + */ + +typedef struct seqcnt { + unsigned sequence; +} seqcnt_t; + +#define SEQCNT_ZERO { 0 } +#define seqcnt_init(x) do { *(x) = (seqcnt_t) SEQCNT_ZERO; } while (0) + +/* Start of read using pointer to a sequence counter only. */ +static inline unsigned read_seqcntbegin(const seqcnt_t *s) +{ + unsigned ret = s->sequence; + smp_rmb(); + return ret; +} + +/* Test if reader processed invalid data. + * Equivalent to: iv is odd or sequence number has changed. + * (iv & 1) || (*s != iv) + * Using xor saves one conditional branch. + */ +static inline int read_seqcntretry(const seqcnt_t *s, unsigned iv) +{ + smp_rmb(); + return (iv & 1) | (s->sequence ^ iv); +} + + +/* + * Sequence counter only version assumes that callers are using their + * own mutexing. + */ +static inline void write_seqcntbegin(seqcnt_t *s) +{ + s->sequence++; + smp_wmb(); +} + +static inline void write_seqcntend(seqcnt_t *s) +{ + smp_wmb(); + s->sequence++; +} + /* * Possible sw/hw IRQ protected versions of the interfaces. */

--=_courier-9137-1053016758-0001-2--