Bug 3762
Summary: | via-usb2: can | ||
---|---|---|---|
Product: | ACPI | Reporter: | Steffen Schwientek (schwientek) |
Component: | Config-Interrupts | Assignee: | Len Brown (lenb) |
Status: | CLOSED CODE_FIX | ||
Severity: | normal | CC: | acpi-bugzilla, akpm, greg |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.9 | Subsystem: | |
Regression: | --- | Bisected commit-id: | |
Attachments: |
output of lspci
the interrupts dmesg output patch applied to acpi test tree |
Description
Steffen Schwientek
2004-11-17 16:02:58 UTC
I'll add this to my list of things-to-fix-for-2.6.10. Please let me know if it gets fixed or is deemed not-a-bug, thanks. Nothing changed recently in how usbcore or EHCI manage IRQs, so this sure looks like an IRQ assignment issue ... if EHCI actually got the IRQ it requested, there'd be no problem. There are way too many options about how IRQs get assigned. I'd hope one of them would resolve this issue. See what the Documentation/* files have to suggest. Seems like not-a-USB-bug ... I'll reassign it unless evidence shows up that the root cause is USB. But the report should also get updated to report behavior with the 2.6.10-rc2 tree (or more recent bk snapshot). Also basics like "lspci -vv" and "cat /proc/interrupts", and the boot logs showing how (and why) specific IRQs got assigned. Erm, updates please? Does 2.6.10-rc2-bk have this problem still? I'm re-categorizing this IRQ routing issue to ACPI. Created attachment 4157 [details]
output of lspci
Created attachment 4158 [details]
the interrupts
output od the interrups
Created attachment 4159 [details]
dmesg output
The dmesg output before inserting the burner...
I I plug in the burner, dmesg adds
usb 1-1: new high speed USB device using ehci_hcd and address 5
ehci_hcd 0000:00:10.3: Unlink after no-IRQ? Different ACPI or APIC settings
may help.
usb 1-1: khubd timed out on ep0i
If I detach the burner, sometimes my system freeze completely.
Yes, indead. It's an ACPI bug. One patch from mine updated a big value for legacy IRQ penalty. Changing to a small value for IRQ penalty would fix it. Please use below workaround till we get a real fix. in 2.6/drivers/acpi/pci_link.c void acpi_penalize_isa_irq(int irq) { - acpi_irq_penalty[irq] += PIRQ_PENALTY_ISA_USED; + acpi_irq_penalty[irq] += 0; } Gents, did the workaround fix things, and are we likely to have a permanent fix in place for 2.6.10? Thanks. Created attachment 4181 [details]
patch applied to acpi test tree
please confirm that this patch fixes the problem.
shipped in 2.6.10-rc3 closing. |