RE: [ACPI] Re: Fw: ACPI bug causes cd-rom lock-ups (2.6.10-rc2)

RE: [ACPI] Re: Fw: ACPI bug causes cd-rom lock-ups (2.6.10-rc2)

[Li, Shaohua]

>Len Brown wrote:
>> CONFIG_PNP_ACPI=n should workaround it too then, I expect.
>Yes. Log attached.
>
>> Please apply this debug patch to the failing kernel
>> and send along the dmesg.
>Done. Attached are 2 logs: one of
>the functional kernel due to disabled
>PNP_ACPI, another one of a broken.
>Both are the -rc2-mm2 with your patch.
Thanks the message. Looks like the system claims IRQ 9, 10, 11 are
possible legacy IRQs. Selecting a small value for ISA IRQ penalty will
solve the issue.
Will discuss with Len which parameter should be selected.

Thanks,
Shaohua

Re: [ACPI] Re: Fw: ACPI bug causes cd-rom lock-ups (2.6.10-rc2)

[Len Brown]

[-- Attachment #1: Type: text/plain, Size: 134 bytes --]

Thanks for running the tests.
Please confirm that this patch make the problem go away in the 
CONFIG_PNP_ACPI=y configuration.

-Len


[-- Attachment #2: pnp.patch --]
[-- Type: text/plain, Size: 612 bytes --]

===== drivers/acpi/pci_link.c 1.35 vs edited =====
--- 1.35/drivers/acpi/pci_link.c	2004-11-09 03:08:36 -05:00
+++ edited/drivers/acpi/pci_link.c	2004-12-01 00:38:18 -05:00
@@ -791,9 +791,16 @@
 	return 1;
 }
 
+/*
+ * We'd like PNP to call this routine for the
+ * single ISA_USED value for each legacy device.
+ * But instead it calls us with each POSSIBLE setting.
+ * There is no ISA_POSSIBLE weight, so we simply use
+ * the (small) PCI_USING penalty.
+ */
 void acpi_penalize_isa_irq(int irq)
 {
-	acpi_irq_penalty[irq] += PIRQ_PENALTY_ISA_USED;
+	acpi_irq_penalty[irq] += PIRQ_PENALTY_PCI_USING;
 }
 
 /*

Re: [ACPI] Re: Fw: ACPI bug causes cd-rom lock-ups (2.6.10-rc2)

[Len Brown]

CONFIG_PNPACPI=y that is...

On Wed, 2004-12-01 at 00:41, Len Brown wrote:
> Thanks for running the tests.
> Please confirm that this patch make the problem go away in the 
> CONFIG_PNP_ACPI=y configuration.
> 
> -Len
> 

[ACPI] Re: Fw: ACPI bug causes cd-rom lock-ups (2.6.10-rc2)

[matthieu castet]

 >+/*
 >+ * We'd like PNP to call this routine for the
 >+ * single ISA_USED value for each legacy device.
 >+ * But instead it calls us with each POSSIBLE setting.
 >+ * There is no ISA_POSSIBLE weight, so we simply use
 >+ * the (small) PCI_USING penalty.
 >+ */

Couldn't be better to change the pnp core behaviour ?

In drivers/pnp/resource.c, pnp_register_irq_resource, instead of calling 
pcibios_penalize_isa_irq couldn't we call something like 
pcibios_penalize_possible_isa_irq ?

The pnp implemations already use pcibios_penalize_isa_irq [1] for the 
irq that are used.
So it seem it is call 2 times for the allocated resources...
Also don't we need to depenalize the irq, if we change the irq or 
disable the device ?

Matthieu CASTET


[1]
$grep -r penalize */*
pnpacpi/rsparser.c: 
pcibios_penalize_isa_irq(res->data.irq.interrupts[0]);
pnpacpi/rsparser.c: 
pcibios_penalize_isa_irq(res->data.extended_irq.interrupts[0]);
pnpbios/rsparser.c:             pcibios_penalize_isa_irq(irq);

Re: [ACPI] Re: Fw: ACPI bug causes cd-rom lock-ups (2.6.10-rc2)

[Stas Sergeev]

Hello.

Len Brown wrote:
> Please confirm that this patch make the problem go away in the 
Confirmed, thanks!

(Andrew, sorry for failing to reply to
you query, and now you know the latest
status anyway. I'll start posting the
-rc2-mm4 problems soon.)