[Qemu-devel] [RFC][PATCH] disable special ioapic inti0 routing

[Qemu-devel] [RFC][PATCH] disable special ioapic inti0 routing

[Jan Kiszka]

This patch comes from the kvm tree and fixes the timer IRQ routing for
me which is broken from the POV of certain Linux guest kernels. As I'm
not up-to-date with the development around that problematic hunk, I'm
leaving it to someone more deeply involved to sign this off. But please
commit some fix.

Thanks,
Jan

------->

From: Avi Kivity <avi@redhat.com>

we don't support it yet (need bios support, and modifications to kernel
irq routing).

---
 qemu/hw/apic.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/qemu/hw/apic.c b/qemu/hw/apic.c
index df80444..f9ef995 100644
--- a/qemu/hw/apic.c
+++ b/qemu/hw/apic.c
@@ -1055,12 +1055,14 @@ void ioapic_set_irq(void *opaque, int vector, int level)
 {
     IOAPICState *s = opaque;
 
+#if 0
     /* ISA IRQs map to GSI 1-1 except for IRQ0 which maps
      * to GSI 2.  GSI maps to ioapic 1-1.  This is not
      * the cleanest way of doing it but it should work. */
 
     if (vector == 0)
         vector = 2;
+#endif
 
     if (vector >= 0 && vector < IOAPIC_NUM_PINS) {
         uint32_t mask = 1 << vector;

Re: [Qemu-devel] [RFC][PATCH] disable special ioapic inti0 routing

[Anthony Liguori]

Jan Kiszka wrote:
> This patch comes from the kvm tree and fixes the timer IRQ routing for
> me which is broken from the POV of certain Linux guest kernels. As I'm
> not up-to-date with the development around that problematic hunk, I'm
> leaving it to someone more deeply involved to sign this off. But please
> commit some fix.
>   

Are you using an old bios by chance?  The reason this fixes KVM has 
nothing to do with QEMU FWIW so if this fixes something for you, it's 
just a coincidence.

Regards,

Anthony Liguroi

> Thanks,
> Jan
>
> ------->
>
> From: Avi Kivity <avi@redhat.com>
>
> we don't support it yet (need bios support, and modifications to kernel
> irq routing).
>
> ---
>  qemu/hw/apic.c |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)
>
> diff --git a/qemu/hw/apic.c b/qemu/hw/apic.c
> index df80444..f9ef995 100644
> --- a/qemu/hw/apic.c
> +++ b/qemu/hw/apic.c
> @@ -1055,12 +1055,14 @@ void ioapic_set_irq(void *opaque, int vector, int level)
>  {
>      IOAPICState *s = opaque;
>  
> +#if 0
>      /* ISA IRQs map to GSI 1-1 except for IRQ0 which maps
>       * to GSI 2.  GSI maps to ioapic 1-1.  This is not
>       * the cleanest way of doing it but it should work. */
>  
>      if (vector == 0)
>          vector = 2;
> +#endif
>  
>      if (vector >= 0 && vector < IOAPIC_NUM_PINS) {
>          uint32_t mask = 1 << vector;
>
>
>
>   

[Qemu-devel] Re: [RFC][PATCH] disable special ioapic inti0 routing

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1638 bytes --]

Anthony Liguori wrote:
> Jan Kiszka wrote:
>> This patch comes from the kvm tree and fixes the timer IRQ routing for
>> me which is broken from the POV of certain Linux guest kernels. As I'm
>> not up-to-date with the development around that problematic hunk, I'm
>> leaving it to someone more deeply involved to sign this off. But please
>> commit some fix.
>>   
> 
> Are you using an old bios by chance?  The reason this fixes KVM has
> nothing to do with QEMU FWIW so if this fixes something for you, it's
> just a coincidence.

My first thought as well, but the binary bios from latest svn made no
difference.

Jan

> 
> Regards,
> 
> Anthony Liguroi
> 
>> Thanks,
>> Jan
>>
>> ------->
>>
>> From: Avi Kivity <avi@redhat.com>
>>
>> we don't support it yet (need bios support, and modifications to kernel
>> irq routing).
>>
>> ---
>>  qemu/hw/apic.c |    2 ++
>>  1 files changed, 2 insertions(+), 0 deletions(-)
>>
>> diff --git a/qemu/hw/apic.c b/qemu/hw/apic.c
>> index df80444..f9ef995 100644
>> --- a/qemu/hw/apic.c
>> +++ b/qemu/hw/apic.c
>> @@ -1055,12 +1055,14 @@ void ioapic_set_irq(void *opaque, int vector,
>> int level)
>>  {
>>      IOAPICState *s = opaque;
>>  
>> +#if 0
>>      /* ISA IRQs map to GSI 1-1 except for IRQ0 which maps
>>       * to GSI 2.  GSI maps to ioapic 1-1.  This is not
>>       * the cleanest way of doing it but it should work. */
>>  
>>      if (vector == 0)
>>          vector = 2;
>> +#endif
>>  
>>      if (vector >= 0 && vector < IOAPIC_NUM_PINS) {
>>          uint32_t mask = 1 << vector;
>>
>>
>>
>>   
> 
> 
> 
> 



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]

[Qemu-devel] Re: [RFC][PATCH] disable special ioapic inti0 routing

[Anthony Liguori]

Jan Kiszka wrote:
> Anthony Liguori wrote:
>   
>> Jan Kiszka wrote:
>>     
>>> This patch comes from the kvm tree and fixes the timer IRQ routing for
>>> me which is broken from the POV of certain Linux guest kernels. As I'm
>>> not up-to-date with the development around that problematic hunk, I'm
>>> leaving it to someone more deeply involved to sign this off. But please
>>> commit some fix.
>>>   
>>>       
>> Are you using an old bios by chance?  The reason this fixes KVM has
>> nothing to do with QEMU FWIW so if this fixes something for you, it's
>> just a coincidence.
>>     
>
> My first thought as well, but the binary bios from latest svn made no
> difference.
>   

What linux guests broke?

Regards,

Anthony Liguori

> Jan
>
>   
>> Regards,
>>
>> Anthony Liguroi
>>
>>     
>>> Thanks,
>>> Jan
>>>
>>> ------->
>>>
>>> From: Avi Kivity <avi@redhat.com>
>>>
>>> we don't support it yet (need bios support, and modifications to kernel
>>> irq routing).
>>>
>>> ---
>>>  qemu/hw/apic.c |    2 ++
>>>  1 files changed, 2 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/qemu/hw/apic.c b/qemu/hw/apic.c
>>> index df80444..f9ef995 100644
>>> --- a/qemu/hw/apic.c
>>> +++ b/qemu/hw/apic.c
>>> @@ -1055,12 +1055,14 @@ void ioapic_set_irq(void *opaque, int vector,
>>> int level)
>>>  {
>>>      IOAPICState *s = opaque;
>>>  
>>> +#if 0
>>>      /* ISA IRQs map to GSI 1-1 except for IRQ0 which maps
>>>       * to GSI 2.  GSI maps to ioapic 1-1.  This is not
>>>       * the cleanest way of doing it but it should work. */
>>>  
>>>      if (vector == 0)
>>>          vector = 2;
>>> +#endif
>>>  
>>>      if (vector >= 0 && vector < IOAPIC_NUM_PINS) {
>>>          uint32_t mask = 1 << vector;
>>>
>>>
>>>
>>>   
>>>       
>>
>>
>>     
>
>
>   

[Qemu-devel] Re: [RFC][PATCH] disable special ioapic inti0 routing

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1108 bytes --]

Anthony Liguori wrote:
> Jan Kiszka wrote:
>> Anthony Liguori wrote:
>>  
>>> Jan Kiszka wrote:
>>>    
>>>> This patch comes from the kvm tree and fixes the timer IRQ routing for
>>>> me which is broken from the POV of certain Linux guest kernels. As I'm
>>>> not up-to-date with the development around that problematic hunk, I'm
>>>> leaving it to someone more deeply involved to sign this off. But please
>>>> commit some fix.
>>>>         
>>> Are you using an old bios by chance?  The reason this fixes KVM has
>>> nothing to do with QEMU FWIW so if this fixes something for you, it's
>>> just a coincidence.
>>>     
>>
>> My first thought as well, but the binary bios from latest svn made no
>> difference.
>>   
> 
> What linux guests broke?

An old 32-bit customized one. And it turned out that it had, for some
reasons I don't recall anymore, ACPI disabled. So it missed the timer
IRQ pin moving QEMU is doing. At the same time, some very similar kernel
with ACPI enabled boots fine. Think I should delete the former from my
test image...

Sorry for the noise,
Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]