One system I have is configured thusly: AMD Duron 1200 (I have frequency restrictions in this application, and must stay below 1.3GHz or must be above 2.0GHz, and the 1200 fits nicely in our interference plan; besides, I was given the chip, and it's fast enough for the application). Soyo DRAGON KT400 motherboard, 512MB DDR RAM (DDR266). Silicon Image SATA interface card.
Machine boots CentOSPlus 4.1 kernel (2.6.9-11.106.unsupported) perfectly fine UNTIL I plug in a Maxtor SATA drive into the drive slide I have in the machine. Drive and slide work perfectly fine at home with a Gigabyte KT600-based motherboard with VIA SATA interface. After plugging in the SATA drive and rebooting (since libata doesn't yet do hotplug), I get a hang after seeing an 'irq 10: nobody cared! (screaming interrupt?)' error.
I have done the kernel command line 'pci=biosirq' and 'acpi=off' suggestions, I have disabled USB, disabled everything that the BIOS will allow. Nothing at this point is sharing IRQ 10 with the Sil 3112 controller. I think it is the latest BIOS version for that motherboard. No matter what I do, it issues the screaming interrupt error until I remove the SATA drive.
The DRAGON motherboard has HPT372 software RAID on board (I'm using it as just another two channel ATA host adapter) as well as audio, network, etc). LAN is enabled; both COMs and the LPT are disabled, USB is disabled, audio is disabled.
And the motherboard works and has worked perfectly fine for months without the SATA drive in the slide (and with all the peripherals enabled!).
Red Hat Bugzilla unresponsive; maybe someone here has seen the problem and knows of some other thing I can tweak. Odd that it works fine with the card in with no drive, but plugging the drive in kills it. With the drive not installed, the card still uses the IRQ, still loads its BIOS, and still shows up in dmesg (this produced with a default boot, non-quiet): [snip] ACPI: (supports S0 S1 S4 S5) ACPI wakeup devices: SLPB PCI0 USB0 USB1 USB2 USB6 USB7 USB8 USB9 LAN0 Freeing unused kernel memory: 148k freed SCSI subsystem initialized libata version 1.10 loaded. sata_sil version 0.8 ACPI: PCI interrupt 0000:00:0b.0[A] -> GSI 10 (level, low) -> IRQ 10 ata1: SATA max UDMA/100 cmd 0xE0832080 ctl 0xE083208A bmdma 0xE0832000 irq 10 ata2: SATA max UDMA/100 cmd 0xE08320C0 ctl 0xE08320CA bmdma 0xE0832008 irq 10 ata1: no device found (phy stat 00000000) scsi0 : sata_sil ata2: no device found (phy stat 00000000) scsi1 : sata_sil device-mapper: 4.4.0-ioctl (2005-01-12) initialised: dm@uk.sistina.com [snip]
It's in the area after the sata_sil driver gets loaded that, when the drive is plugged in, the screaming interrupt occurs, and the system hangs. It DOES detect the drive, though. Oh, and let me head off the obvious thing at the pass: I'm not using any pseudo-RAID features on either the HPT372 or the Sil 3112. Both chips work fine; the Sil 3112 card was in a Windows box previously and has worked with this drive. The HPT372 is working FINE right now as two independent ATA channels.
Just looking for some ideas of things to look at (other than 'don't use a FRAID card' 'try another motherboard' and the like; those pseudoanswers will be ignored; please only answer if you have an idea, don't waste bandwidth telling me you don't know, but that SOYO motherboards are junk, etc. Too much of that around here, unfortunately, and it is annoying).