Re: [ACPI-sppt] Kernel 2.4.22 ACPI USB IRQ init sequence broken
Ole Myren Rohne <[email protected]>
| Newsgroups | gmane.linux.usb.user,gmane.linux.acpi.support |
|---|---|
| Message-ID | <[email protected]> |
I've seen the same symptoms (plugging a usb device hardlocks the p2120 on recent kernels), I just want to add an observation of the rather peculiar usb hookup on this machine: Wan Tat Chee <[email protected]> writes: > What happened to the ACPI IRQ init code?? I've noted that the p2120 has two USB chips: One is integrated on the ALi chipset (00:02.0) and is not even connected, then there is a NEC chip providing 2 x ohci (00:09.0, 00:09.1) plus ehci (00:09.2). The pre-2.4.22 ACPI IRQ routing is broken in a benign way, it puts all its interrupts (acpi, 2 x ohci, ehci, fw, audio, eth, wifi) on one line (IRQ 9 on my machine). The new code (2.4.22 as well as 2.6.0-preN) seems much smarter at distributing the available lines; unfortunately, one of the live NEC ohci's end up sharing with the defunct ALi ohci. I'm guessing that this is the reason why the usb-ohci driver locks up. Supposing I'm correct, one solution would be to have the usb-ohci driver fixed to handle the somewhat funny hardware hookup of the p2120 (a unused hcd of one type sharing an IRQ with a different type hcd). Another way would be to have the ACPI system never share IRQ with a non-ACPI device. One could also imagine a solution that specify IRQ lines by hand or expressively forbids the driver to connect to the unused device... Ole ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ [email protected] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-users