Messages in this thread | | | Subject | Re: [patch] voluntary-preempt-2.6.9-rc1-bk4-R0 | From | Lee Revell <> | Date | Thu, 02 Sep 2004 18:06:53 -0400 |
| |
On Thu, 2004-09-02 at 17:57, Ingo Molnar wrote: > i've released the -R0 patch: > > http://redhat.com/~mingo/voluntary-preempt/voluntary-preempt-2.6.9-rc1-bk4-R0 > > ontop of: > > http://redhat.com/~mingo/voluntary-preempt/diff-bk-040828-2.6.8.1.bz2 > > i've given up on the netdev_backlog_granularity approach, and as a > replacement i've modified specific network drivers to return at a safe > point if softirq preemption is requested.
Makes sense, netdev_max_backlog never made a difference on my system (via-rhine driver).
If you read that Microsoft paper I posted a link to earlier, they describe all of the main categories of latencies we have dealt with. They give special mention to network driver DPCs/softirqs.
The worst offender was a driver that used a DPC/softirq to reset the card if there was no traffic in 10 seconds, to work around a hardware lockup bug in some versions. Since the reset path was never designed to be fast this caused problems. They also mention a similar problem to the one that still exists with the via-rhine driver (fixed in -mm) where it uses a DPC to poll for link status.
Lee
- 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/
| |