2015-07-27 11:36:26

by Julien Grall

[permalink] [raw]
Subject: [PATCH v2] xen/events: Support event channel rebind on ARM

Currently, the event channel rebind code is gated with the presence of
the vector callback.

The virtual interrupt controller on ARM has the concept of per-CPU
interrupt (PPI) which allow us to support per-VCPU event channel.
Therefore there is no need of vector callback for ARM.

Xen is already using a free PPI to notify the guest VCPU of an event.
Furthermore, the xen code initialization in Linux (see
arch/arm/xen/enlighten.c) is requesting correctly a per-CPU IRQ.

Introduce new helper xen_support_evtchn_rebind to allow architecture
decide whether rebind an event is support or not. It will always return
1 on ARM and keep the same behavior on x86.

This is also allow us to drop the usage of xen_have_vector_callback
entirely in the ARM code.

Signed-off-by: Julien Grall <[email protected]>
Reviewed-by: David Vrabel <[email protected]>
Cc: Stefano Stabellini <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Konrad Rzeszutek Wilk <[email protected]>
Cc: Boris Ostrovsky <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Cc: [email protected]
---
Tested by setting the affinity manually via /proc on dom0 ARM64

Changes in v2:
- Use a static inline rather than a macro
- Add David's reviewed-by
---
arch/arm/include/asm/xen/events.h | 6 ++++++
arch/arm/xen/enlighten.c | 4 ----
arch/arm64/include/asm/xen/events.h | 6 ++++++
arch/x86/include/asm/xen/events.h | 14 ++++++++++++++
drivers/xen/events/events_base.c | 6 +-----
include/xen/events.h | 1 -
6 files changed, 27 insertions(+), 10 deletions(-)

diff --git a/arch/arm/include/asm/xen/events.h b/arch/arm/include/asm/xen/events.h
index 8b1f37b..2123aaa 100644
--- a/arch/arm/include/asm/xen/events.h
+++ b/arch/arm/include/asm/xen/events.h
@@ -20,4 +20,10 @@ static inline int xen_irqs_disabled(struct pt_regs *regs)
atomic64_t, \
counter), (val))

+/* Rebind event channel is supported by default */
+static inline bool xen_support_evtchn_rebind(void)
+{
+ return 1;
+}
+
#endif /* _ASM_ARM_XEN_EVENTS_H */
diff --git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
index 9055c92..c50c8d3 100644
--- a/arch/arm/xen/enlighten.c
+++ b/arch/arm/xen/enlighten.c
@@ -45,10 +45,6 @@ static struct vcpu_info __percpu *xen_vcpu_info;
unsigned long xen_released_pages;
struct xen_memory_region xen_extra_mem[XEN_EXTRA_MEM_MAX_REGIONS] __initdata;

-/* TODO: to be removed */
-__read_mostly int xen_have_vector_callback;
-EXPORT_SYMBOL_GPL(xen_have_vector_callback);
-
static __read_mostly unsigned int xen_events_irq;

static __initdata struct device_node *xen_node;
diff --git a/arch/arm64/include/asm/xen/events.h b/arch/arm64/include/asm/xen/events.h
index 8655321..ed4aa90 100644
--- a/arch/arm64/include/asm/xen/events.h
+++ b/arch/arm64/include/asm/xen/events.h
@@ -18,4 +18,10 @@ static inline int xen_irqs_disabled(struct pt_regs *regs)

#define xchg_xen_ulong(ptr, val) xchg((ptr), (val))

+/* Rebind event channel is supported by default */
+static inline bool xen_support_evtchn_rebind(void)
+{
+ return 1;
+}
+
#endif /* _ASM_ARM64_XEN_EVENTS_H */
diff --git a/arch/x86/include/asm/xen/events.h b/arch/x86/include/asm/xen/events.h
index 608a79d..37bc720 100644
--- a/arch/x86/include/asm/xen/events.h
+++ b/arch/x86/include/asm/xen/events.h
@@ -20,4 +20,18 @@ static inline int xen_irqs_disabled(struct pt_regs *regs)
/* No need for a barrier -- XCHG is a barrier on x86. */
#define xchg_xen_ulong(ptr, val) xchg((ptr), (val))

+/*
+ * Events delivered via platform PCI interrupts are always
+ * routed to vcpu 0 and hence cannot be rebound.
+ */
+static inline bool xen_support_evtchn_rebind(void)
+{
+ return (!xen_hvm_domain() || xen_have_vector_callback);
+}
+
+#define xen_support_evtchn_rebind() \
+ (!xen_hvm_domain() || xen_have_vector_callback)
+
+extern int xen_have_vector_callback;
+
#endif /* _ASM_X86_XEN_EVENTS_H */
diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
index 96093ae..ed620e5 100644
--- a/drivers/xen/events/events_base.c
+++ b/drivers/xen/events/events_base.c
@@ -1301,11 +1301,7 @@ static int rebind_irq_to_cpu(unsigned irq, unsigned tcpu)
if (!VALID_EVTCHN(evtchn))
return -1;

- /*
- * Events delivered via platform PCI interrupts are always
- * routed to vcpu 0 and hence cannot be rebound.
- */
- if (xen_hvm_domain() && !xen_have_vector_callback)
+ if (!xen_support_evtchn_rebind())
return -1;

/* Send future instances of this interrupt to other vcpu. */
diff --git a/include/xen/events.h b/include/xen/events.h
index 7d95fdf..88da2ab 100644
--- a/include/xen/events.h
+++ b/include/xen/events.h
@@ -92,7 +92,6 @@ void xen_hvm_callback_vector(void);
#ifdef CONFIG_TRACING
#define trace_xen_hvm_callback_vector xen_hvm_callback_vector
#endif
-extern int xen_have_vector_callback;
int xen_set_callback_via(uint64_t via);
void xen_evtchn_do_upcall(struct pt_regs *regs);
void xen_hvm_evtchn_do_upcall(void);
--
2.1.4


2015-07-27 12:35:26

by David Vrabel

[permalink] [raw]
Subject: Re: [PATCH v2] xen/events: Support event channel rebind on ARM

On 27/07/15 12:35, Julien Grall wrote:
> Currently, the event channel rebind code is gated with the presence of
> the vector callback.
>
> The virtual interrupt controller on ARM has the concept of per-CPU
> interrupt (PPI) which allow us to support per-VCPU event channel.
> Therefore there is no need of vector callback for ARM.
>
> Xen is already using a free PPI to notify the guest VCPU of an event.
> Furthermore, the xen code initialization in Linux (see
> arch/arm/xen/enlighten.c) is requesting correctly a per-CPU IRQ.
>
> Introduce new helper xen_support_evtchn_rebind to allow architecture
> decide whether rebind an event is support or not. It will always return
> 1 on ARM and keep the same behavior on x86.
>
> This is also allow us to drop the usage of xen_have_vector_callback
> entirely in the ARM code.

This did not apply cleanly. Please always base patches on Linus's
master branch.

This also breaks the x86 build.

/local/davidvr/work/k.org/tip/arch/x86/include/asm/xen/events.h: In
function ?xen_support_evtchn_rebind?:
/local/davidvr/work/k.org/tip/arch/x86/include/asm/xen/events.h:29:85:
error: ?xen_have_vector_callback? undeclared (first use in this function)
return (!xen_hvm_domain() || xen_have_vector_callback);

> --- a/arch/arm/include/asm/xen/events.h
> +++ b/arch/arm/include/asm/xen/events.h
> @@ -20,4 +20,10 @@ static inline int xen_irqs_disabled(struct pt_regs *regs)
> atomic64_t, \
> counter), (val))
>
> +/* Rebind event channel is supported by default */
> +static inline bool xen_support_evtchn_rebind(void)
> +{
> + return 1;

This should be true (similarly for arm64).

David

2015-07-27 12:44:39

by Julien Grall

[permalink] [raw]
Subject: Re: [Xen-devel] [PATCH v2] xen/events: Support event channel rebind on ARM

On 27/07/15 13:35, David Vrabel wrote:
> On 27/07/15 12:35, Julien Grall wrote:
>> Currently, the event channel rebind code is gated with the presence of
>> the vector callback.
>>
>> The virtual interrupt controller on ARM has the concept of per-CPU
>> interrupt (PPI) which allow us to support per-VCPU event channel.
>> Therefore there is no need of vector callback for ARM.
>>
>> Xen is already using a free PPI to notify the guest VCPU of an event.
>> Furthermore, the xen code initialization in Linux (see
>> arch/arm/xen/enlighten.c) is requesting correctly a per-CPU IRQ.
>>
>> Introduce new helper xen_support_evtchn_rebind to allow architecture
>> decide whether rebind an event is support or not. It will always return
>> 1 on ARM and keep the same behavior on x86.
>>
>> This is also allow us to drop the usage of xen_have_vector_callback
>> entirely in the ARM code.
>
> This did not apply cleanly. Please always base patches on Linus's
> master branch.

My patch is based on Linus's master branch. Although, it has a
pre-requisite of [1] which I sent a few seconds before it. Though it's
only for applying because the 2 patches are not related.

I should have mentioned it in the mail, sorry.

> This also breaks the x86 build.

I forgot to test this new version when I turned the macro into a static
inline.

> /local/davidvr/work/k.org/tip/arch/x86/include/asm/xen/events.h: In
> function ?xen_support_evtchn_rebind?:
> /local/davidvr/work/k.org/tip/arch/x86/include/asm/xen/events.h:29:85:
> error: ?xen_have_vector_callback? undeclared (first use in this function)
> return (!xen_hvm_domain() || xen_have_vector_callback);
>
>> --- a/arch/arm/include/asm/xen/events.h
>> +++ b/arch/arm/include/asm/xen/events.h
>> @@ -20,4 +20,10 @@ static inline int xen_irqs_disabled(struct pt_regs *regs)
>> atomic64_t, \
>> counter), (val))
>>
>> +/* Rebind event channel is supported by default */
>> +static inline bool xen_support_evtchn_rebind(void)
>> +{
>> + return 1;
>
> This should be true (similarly for arm64).

Will resend a new version, and remove the dependency on [1].

Regards,

[1] http://permalink.gmane.org/gmane.linux.kernel/2004690

--
Julien Grall