Messages in this thread | | | Date | Wed, 10 Nov 2004 16:33:00 +0100 | From | Ingo Molnar <> | Subject | Re: [patch] Real-Time Preemption, -RT-2.6.10-rc1-mm3-V0.7.23 |
| |
* Karsten Wiese <annabellesgarden@yahoo.de> wrote:
> Yes, it always happens, when callling ./cvscompile script of a > project, that is mounted via nfs. Haven't tried to do that > ./cvscompile locally, should I?
very interesting, nfs is indeed one of the frequent BKL users.
a 'BKL leak' is an unbalanced lock, e.g.:
lock_kernel(); ... do stuff ... if (error) return; ... do more stuff ... unlock_kernel();
the BKL is a very special type of lock which fact has the side-effect that in the stock kernel a 'BKL leak' can go unnoticed very easily: it causes no problems other than hard-to-debug (but severe) scalability regressions. The moment the BKL count leaked from the NFS code that process has been 'scalability-poisoned' and will be handicapped until it exits.
In the PREEMPT_RT patchset i added a strict locking checker to do_exit() that found this apparent NFS bug. Unfortunately the deadlock detector only reported a pretty common place where the BKL gets dropped/reacquired frequently so we dont know where the NFS code has the lock/unlock imbalance. Could you report this bug to the NFS maintainers (along with the above and the previous analysis)?
Ingo - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |