Just PCI device interrupts are level triggered, interesting would be to get edge triggered cases, esp MSI capable devices. This can locate if fixup_irqs() need any special handling. Also unsolicited intr, from network are great loads, since during cpu remove we bring system to Knees, nothing originates locally, but we cant stop outside world talking to us, these are great cases to find missing intr as well. Also Shaohua (cced) found some corruption problem in large cache cases, iam checking if he can send us test specifics, so we could incorporate them as well. Cheers, ashok raj - Open Source Technology Center >-----Original Message----- >From: hotplug_sig-bounces@xxxxxxxxxxxxxx [mailto:hotplug_sig-bounces@xxxxxxxxxxxxxx] On Behalf Of Joel >Schopp >Sent: Friday, June 10, 2005 9:35 AM >To: Craig Thomas >Cc: hotplug_sig@xxxxxxxxxxxxxx >Subject: Re: [Hotplug_sig] MINUTES for Hotplug SIG Con Call 03/26 > >> One of the background loads needs to consist of several >> threaded applications running concurrently. Threads really mess >> up the scheduler and CPU usage. Java programs and just plain >> C programs that are threaded will make good background loads. > >Yes, and loads which generate lots of interrupts. Most SCSI and >ethernet traffic generates interrupts at a frantic pace. So any disk or >network benchmark would do wonders at causing interrupts at the wrong >place or at the wrong time. >
