| To: | Andi Kleen <ak@xxxxxxx> |
|---|---|
| Subject: | Re: TAKE 950027 - xfs_icsb_lock_all_counters fails with CONFIG_PREEMPT and >=256p |
| From: | David Chinner <dgc@xxxxxxx> |
| Date: | Fri, 3 Mar 2006 07:43:40 +1100 |
| Cc: | David Chinner <dgc@xxxxxxx>, linux-xfs@xxxxxxxxxxx, mingo@xxxxxxx, torvalds@xxxxxxxx, tony.luck@xxxxxxxxx |
| In-reply-to: | <200603021309.46495.ak@suse.de> |
| References: | <20060301125320.20FDA49F1681@chook.melbourne.sgi.com> <p73fym1zbqo.fsf@verdi.suse.de> <20060302065807.GG1173973@melbourne.sgi.com> <200603021309.46495.ak@suse.de> |
| Sender: | linux-xfs-bounce@xxxxxxxxxxx |
| User-agent: | Mutt/1.4.2.1i |
On Thu, Mar 02, 2006 at 01:09:45PM +0100, Andi Kleen wrote: > > Ingo, Linus, Tony, what do you think? XFS is running into trouble > on preemptive kernels on >256CPU systems because there are > cases where one thread can hold 2*NR_CPUS spinlocks > and that overflows the current 8 bit preempt count. 2 * NR_CPU spinlocks held by a single thread was a purely hypothetical question. THe code does not do that. It requires a barrier per CPU, and as I've said before, spinlocks were a bad implementation choice. Cheers, Dave. -- Dave Chinner R&D Software Enginner SGI Australian Software Group |
| Previous by Date: | Re: TAKE 950027 - xfs_icsb_lock_all_counters fails with CONFIG_PREEMPT and >=256p, David Chinner |
|---|---|
| Next by Date: | Re: TAKE 950027 - xfs_icsb_lock_all_counters fails with CONFIG_PREEMPT and >=256p, David Chinner |
| Previous by Thread: | Re: TAKE 950027 - xfs_icsb_lock_all_counters fails with CONFIG_PREEMPT and >=256p, David Chinner |
| Next by Thread: | RE: TAKE 950027 - xfs_icsb_lock_all_counters fails with CONFIG_PREEMPT and >=256p, Luck, Tony |
| Indexes: | [Date] [Thread] [Top] [All Lists] |