On Sun, Oct 22, 2006 at 12:33:28PM -0700, Jason Thorpe wrote:
> 
> Why would LFS be at splhigh()?  Have you confirmed that in the code  
> paths in question?

You're right.  It's splbio() everywhere.  There were splhigh() calls
in there years ago, and when I saw this misbehavior, I assumed that
there still were.

> More likely, I think, is that your network interface and your RAID  
> controller are sharing an IRQ, or that the RAID controller is  
> consuming a lot of PCI bandwidth during these bursts of activity.

bge1 at pci2 dev 9 function 1: Broadcom BCM5704C Dual Gigabit Ethernet
bge1: interrupting at ioapic1 pin 1 (irq 15)

amr0 at pci3 dev 3 function 0: AMI RAID <MegaRAID SATA 150-4D>
amr0: interrupting at ioapic2 pin 0 (irq 11)

The controllers are on separate PCI-X buses (which are not bridged
to the same "primary" bus; they're on an AMD HT<->PCI-X tunnel).
60MB/sec is the most I've seen to the disk, anyway.

What do you think is going on?

-- 
  Thor Lancelot Simon	                                     tls@rek.tjls.com

  "We cannot usually in social life pursue a single value or a single moral
   aim, untroubled by the need to compromise with others."      - H.L.A. Hart