2015-11-17 15:36:26

by Boris Ostrovsky

Subject: [PATCH] xen/events: Always allocate legacy interrupts on PV guests

After commit 8c058b0b9c34 ("x86/irq: Probe for PIC presence before
allocating descs for legacy IRQs") early_irq_init() will no longer
preallocate descriptors for legacy interrupts if PIT does not
exist.

Therefore we need to allocate those descriptors for PV guests
ourselves.

Signed-off-by: Boris Ostrovsky <[email protected]>
Suggested-by: Thomas Gleixner <[email protected]>
---
drivers/xen/events/events_base.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
index 849500e..a2bb333 100644
--- a/drivers/xen/events/events_base.c
+++ b/drivers/xen/events/events_base.c
@@ -419,8 +419,8 @@ static int __must_check xen_allocate_irq_gsi(unsigned gsi)
if (xen_pv_domain() && !xen_initial_domain())
return xen_allocate_irq_dynamic();

- /* Legacy IRQ descriptors are already allocated by the arch. */
- if (gsi < NR_IRQS_LEGACY)
+ /* On HVM legacy IRQ descriptors are already allocated by the arch. */
+ if (xen_hvm_domain() && gsi < NR_IRQS_LEGACY)
irq = gsi;
else
irq = irq_alloc_desc_at(gsi, -1);
@@ -445,8 +445,8 @@ static void xen_free_irq(unsigned irq)

kfree(info);

- /* Legacy IRQ descriptors are managed by the arch. */
- if (irq < NR_IRQS_LEGACY)
+ /* On HVM legacy IRQ descriptors are managed by the arch. */
+ if (xen_hvm_domain() && irq < NR_IRQS_LEGACY)
return;

irq_free_desc(irq);
--
1.9.3

2015-11-17 15:38:25

by David Vrabel

Subject: Re: [PATCH] xen/events: Always allocate legacy interrupts on PV guests

On 17/11/15 15:36, Boris Ostrovsky wrote:
> After commit 8c058b0b9c34 ("x86/irq: Probe for PIC presence before
> allocating descs for legacy IRQs") early_irq_init() will no longer
> preallocate descriptors for legacy interrupts if PIT does not
> exist.
>
> Therefore we need to allocate those descriptors for PV guests
> ourselves.
>
> Signed-off-by: Boris Ostrovsky <[email protected]>
> Suggested-by: Thomas Gleixner <[email protected]>

Reviewed-by: David Vrabel <[email protected]>

David

2015-11-17 16:24:35

by Juergen Gross

Subject: Re: [PATCH] xen/events: Always allocate legacy interrupts on PV guests

On 17/11/15 16:36, Boris Ostrovsky wrote:
> After commit 8c058b0b9c34 ("x86/irq: Probe for PIC presence before
> allocating descs for legacy IRQs") early_irq_init() will no longer
> preallocate descriptors for legacy interrupts if PIT does not
> exist.
>
> Therefore we need to allocate those descriptors for PV guests
> ourselves.
>
> Signed-off-by: Boris Ostrovsky <[email protected]>
> Suggested-by: Thomas Gleixner <[email protected]>

Tested-by: Juergen Gross <[email protected]>


Juergen

2015-11-18 11:16:34

by Vitaly Kuznetsov

Subject: Re: [Xen-devel] [PATCH] xen/events: Always allocate legacy interrupts on PV guests

Boris Ostrovsky <[email protected]> writes:

> After commit 8c058b0b9c34 ("x86/irq: Probe for PIC presence before
> allocating descs for legacy IRQs") early_irq_init() will no longer
> preallocate descriptors for legacy interrupts if PIT does not
> exist.

PIC?

>
> Therefore we need to allocate those descriptors for PV guests
> ourselves.
>
> Signed-off-by: Boris Ostrovsky <[email protected]>
> Suggested-by: Thomas Gleixner <[email protected]>
> ---
> drivers/xen/events/events_base.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
> index 849500e..a2bb333 100644
> --- a/drivers/xen/events/events_base.c
> +++ b/drivers/xen/events/events_base.c
> @@ -419,8 +419,8 @@ static int __must_check xen_allocate_irq_gsi(unsigned gsi)
> if (xen_pv_domain() && !xen_initial_domain())
> return xen_allocate_irq_dynamic();
>
> - /* Legacy IRQ descriptors are already allocated by the arch. */
> - if (gsi < NR_IRQS_LEGACY)
> + /* On HVM legacy IRQ descriptors are already allocated by the arch. */
> + if (xen_hvm_domain() && gsi < NR_IRQS_LEGACY)
> irq = gsi;

Wouldn't it be better to write it as
if (gsi < nr_legacy_irqs()) ?

I think it's possible to have PIC-less HVM guests in future (btw, what
about HVMlite?). I see nr_legacy_irqs() is x86-only but it can easily be
defined to NR_IRQS_LEGACY on other arches.

> else
> irq = irq_alloc_desc_at(gsi, -1);
> @@ -445,8 +445,8 @@ static void xen_free_irq(unsigned irq)
>
> kfree(info);
>
> - /* Legacy IRQ descriptors are managed by the arch. */
> - if (irq < NR_IRQS_LEGACY)
> + /* On HVM legacy IRQ descriptors are managed by the arch. */
> + if (xen_hvm_domain() && irq < NR_IRQS_LEGACY)
> return;
>
> irq_free_desc(irq);

--
Vitaly

2015-11-18 14:03:51

by Boris Ostrovsky

Subject: Re: [Xen-devel] [PATCH] xen/events: Always allocate legacy interrupts on PV guests

On 11/18/2015 06:16 AM, Vitaly Kuznetsov wrote:
> Boris Ostrovsky <[email protected]> writes:
>
>> After commit 8c058b0b9c34 ("x86/irq: Probe for PIC presence before
>> allocating descs for legacy IRQs") early_irq_init() will no longer
>> preallocate descriptors for legacy interrupts if PIT does not
>> exist.
> PIC?

Right. David, can you fix this before committing?

>
>> Therefore we need to allocate those descriptors for PV guests
>> ourselves.
>>
>> Signed-off-by: Boris Ostrovsky <[email protected]>
>> Suggested-by: Thomas Gleixner <[email protected]>
>> ---
>> drivers/xen/events/events_base.c | 8 ++++----
>> 1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
>> index 849500e..a2bb333 100644
>> --- a/drivers/xen/events/events_base.c
>> +++ b/drivers/xen/events/events_base.c
>> @@ -419,8 +419,8 @@ static int __must_check xen_allocate_irq_gsi(unsigned gsi)
>> if (xen_pv_domain() && !xen_initial_domain())
>> return xen_allocate_irq_dynamic();
>>
>> - /* Legacy IRQ descriptors are already allocated by the arch. */
>> - if (gsi < NR_IRQS_LEGACY)
>> + /* On HVM legacy IRQ descriptors are already allocated by the arch. */
>> + if (xen_hvm_domain() && gsi < NR_IRQS_LEGACY)
>> irq = gsi;
> Wouldn't it be better to write it as
> if (gsi < nr_legacy_irqs()) ?

I don't think so: on PV we end up setting legacy_pic to null_legacy_pic
in probe_8259A() and that will make nr_legacy_irqs() return 0.

>
> I think it's possible to have PIC-less HVM guests in future (btw, what
> about HVMlite?). I see nr_legacy_irqs() is x86-only but it can easily be
> defined to NR_IRQS_LEGACY on other arches.

Yes, HVMlite (or however we will end up calling it) will allow these
sorts of thing. But HVMlite is not supported yet at all. And it's x86 only.

-boris

>
>> else
>> irq = irq_alloc_desc_at(gsi, -1);
>> @@ -445,8 +445,8 @@ static void xen_free_irq(unsigned irq)
>>
>> kfree(info);
>>
>> - /* Legacy IRQ descriptors are managed by the arch. */
>> - if (irq < NR_IRQS_LEGACY)
>> + /* On HVM legacy IRQ descriptors are managed by the arch. */
>> + if (xen_hvm_domain() && irq < NR_IRQS_LEGACY)
>> return;
>>
>> irq_free_desc(irq);

2015-11-18 14:28:58

by Vitaly Kuznetsov

Subject: Re: [Xen-devel] [PATCH] xen/events: Always allocate legacy interrupts on PV guests

Boris Ostrovsky <[email protected]> writes:

> On 11/18/2015 06:16 AM, Vitaly Kuznetsov wrote:
>> Boris Ostrovsky <[email protected]> writes:
>>
>>> After commit 8c058b0b9c34 ("x86/irq: Probe for PIC presence before
>>> allocating descs for legacy IRQs") early_irq_init() will no longer
>>> preallocate descriptors for legacy interrupts if PIT does not
>>> exist.
>> PIC?
>
> Right. David, can you fix this before committing?
>
>>
>>> Therefore we need to allocate those descriptors for PV guests
>>> ourselves.
>>>
>>> Signed-off-by: Boris Ostrovsky <[email protected]>
>>> Suggested-by: Thomas Gleixner <[email protected]>
>>> ---
>>> drivers/xen/events/events_base.c | 8 ++++----
>>> 1 file changed, 4 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
>>> index 849500e..a2bb333 100644
>>> --- a/drivers/xen/events/events_base.c
>>> +++ b/drivers/xen/events/events_base.c
>>> @@ -419,8 +419,8 @@ static int __must_check xen_allocate_irq_gsi(unsigned gsi)
>>> if (xen_pv_domain() && !xen_initial_domain())
>>> return xen_allocate_irq_dynamic();
>>>
>>> - /* Legacy IRQ descriptors are already allocated by the arch. */
>>> - if (gsi < NR_IRQS_LEGACY)
>>> + /* On HVM legacy IRQ descriptors are already allocated by the arch. */
>>> + if (xen_hvm_domain() && gsi < NR_IRQS_LEGACY)
>>> irq = gsi;
>> Wouldn't it be better to write it as
>> if (gsi < nr_legacy_irqs()) ?
>
> I don't think so: on PV we end up setting legacy_pic to
> null_legacy_pic in probe_8259A() and that will make nr_legacy_irqs()
> return 0.

Yes, so the condition will always be false for PV and it equals to
xen_hvm_domain() or am I missng something?

>
>>
>> I think it's possible to have PIC-less HVM guests in future (btw, what
>> about HVMlite?). I see nr_legacy_irqs() is x86-only but it can easily be
>> defined to NR_IRQS_LEGACY on other arches.
>
> Yes, HVMlite (or however we will end up calling it) will allow these
> sorts of thing. But HVMlite is not supported yet at all. And it's x86
> only.
>
> -boris
>
>>
>>> else
>>> irq = irq_alloc_desc_at(gsi, -1);
>>> @@ -445,8 +445,8 @@ static void xen_free_irq(unsigned irq)
>>>
>>> kfree(info);
>>>
>>> - /* Legacy IRQ descriptors are managed by the arch. */
>>> - if (irq < NR_IRQS_LEGACY)
>>> + /* On HVM legacy IRQ descriptors are managed by the arch. */
>>> + if (xen_hvm_domain() && irq < NR_IRQS_LEGACY)
>>> return;
>>>
>>> irq_free_desc(irq);

--
Vitaly

2015-11-18 15:01:04

by Boris Ostrovsky

Subject: Re: [Xen-devel] [PATCH] xen/events: Always allocate legacy interrupts on PV guests

On 11/18/2015 09:28 AM, Vitaly Kuznetsov wrote:
> Boris Ostrovsky <[email protected]> writes:
>
>> On 11/18/2015 06:16 AM, Vitaly Kuznetsov wrote:
>>> Boris Ostrovsky <[email protected]> writes:
>>>
>>>> After commit 8c058b0b9c34 ("x86/irq: Probe for PIC presence before
>>>> allocating descs for legacy IRQs") early_irq_init() will no longer
>>>> preallocate descriptors for legacy interrupts if PIT does not
>>>> exist.
>>> PIC?
>> Right. David, can you fix this before committing?
>>
>>>> Therefore we need to allocate those descriptors for PV guests
>>>> ourselves.
>>>>
>>>> Signed-off-by: Boris Ostrovsky <[email protected]>
>>>> Suggested-by: Thomas Gleixner <[email protected]>
>>>> ---
>>>> drivers/xen/events/events_base.c | 8 ++++----
>>>> 1 file changed, 4 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
>>>> index 849500e..a2bb333 100644
>>>> --- a/drivers/xen/events/events_base.c
>>>> +++ b/drivers/xen/events/events_base.c
>>>> @@ -419,8 +419,8 @@ static int __must_check xen_allocate_irq_gsi(unsigned gsi)
>>>> if (xen_pv_domain() && !xen_initial_domain())
>>>> return xen_allocate_irq_dynamic();
>>>>
>>>> - /* Legacy IRQ descriptors are already allocated by the arch. */
>>>> - if (gsi < NR_IRQS_LEGACY)
>>>> + /* On HVM legacy IRQ descriptors are already allocated by the arch. */
>>>> + if (xen_hvm_domain() && gsi < NR_IRQS_LEGACY)
>>>> irq = gsi;
>>> Wouldn't it be better to write it as
>>> if (gsi < nr_legacy_irqs()) ?
>> I don't think so: on PV we end up setting legacy_pic to
>> null_legacy_pic in probe_8259A() and that will make nr_legacy_irqs()
>> return 0.
> Yes, so the condition will always be false for PV and it equals to
> xen_hvm_domain() or am I missng something?

Oh, I see what you are saying. Yes, it would be cleaner.

-boris



>
>>> I think it's possible to have PIC-less HVM guests in future (btw, what
>>> about HVMlite?). I see nr_legacy_irqs() is x86-only but it can easily be
>>> defined to NR_IRQS_LEGACY on other arches.
>>>