Message ID | 20090629100837.GW20289@redhat.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
* Gleb Natapov <gleb@redhat.com> wrote: > KVM would like to provide x2APIC interface to a guest without emulating > interrupt remapping device. The reason KVM prefers guest to use x2APIC > is that x2APIC interface is better virtualizable and provides better > performance than mmio xAPIC interface: > > - msr exits are faster than mmio (no page table walk, emulation) > - no need to read back ICR to look at the busy bit > - one 64 bit ICR write instead of two 32 bit writes > - shared code with the Hyper-V paravirt interface > > Included patch changes x2APIC enabling logic to enable it even if IR > initialization failed, but kernel runs under KVM and no apic id is > greater than 255 (if there is one spec requires BIOS to move to x2apic > mode before starting an OS). > > Signed-off-by: Gleb Natapov <gleb@redhat.com> > --- > > v2: > Add help message to Kconfig > Force physical mode if IR is not available. > > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig > index d1430ef..3e5b6ea 100644 > --- a/arch/x86/Kconfig > +++ b/arch/x86/Kconfig > @@ -260,12 +260,15 @@ config SMP > > config X86_X2APIC > bool "Support x2apic" > - depends on X86_LOCAL_APIC && X86_64 && INTR_REMAP > + depends on X86_LOCAL_APIC && X86_64 Have you tried to build this with X86_X2APIC set but INTR_REMAP disabled? Ingo -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Jun 29, 2009 at 12:22:35PM +0200, Ingo Molnar wrote: > > * Gleb Natapov <gleb@redhat.com> wrote: > > > KVM would like to provide x2APIC interface to a guest without emulating > > interrupt remapping device. The reason KVM prefers guest to use x2APIC > > is that x2APIC interface is better virtualizable and provides better > > performance than mmio xAPIC interface: > > > > - msr exits are faster than mmio (no page table walk, emulation) > > - no need to read back ICR to look at the busy bit > > - one 64 bit ICR write instead of two 32 bit writes > > - shared code with the Hyper-V paravirt interface > > > > Included patch changes x2APIC enabling logic to enable it even if IR > > initialization failed, but kernel runs under KVM and no apic id is > > greater than 255 (if there is one spec requires BIOS to move to x2apic > > mode before starting an OS). > > > > Signed-off-by: Gleb Natapov <gleb@redhat.com> > > --- > > > > v2: > > Add help message to Kconfig > > Force physical mode if IR is not available. > > > > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig > > index d1430ef..3e5b6ea 100644 > > --- a/arch/x86/Kconfig > > +++ b/arch/x86/Kconfig > > @@ -260,12 +260,15 @@ config SMP > > > > config X86_X2APIC > > bool "Support x2apic" > > - depends on X86_LOCAL_APIC && X86_64 && INTR_REMAP > > + depends on X86_LOCAL_APIC && X86_64 > > Have you tried to build this with X86_X2APIC set but INTR_REMAP > disabled? > Yes. I notice that X86_X2APIC depends on INTR_REMAP when I did it. -- Gleb. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, 2009-06-29 at 03:08 -0700, Gleb Natapov wrote: > > - local_irq_save(flags); > mask_IO_APIC_setup(ioapic_entries); > - mask_8259A(); Is there a reason why the 8259 mask/unmask operations are separated from io-apic mask/unmask operations. Can we keep it together so that it will be easy to read and understand that we first do the interrupt subsystem mask, try enabling IR and x2apic and unmask the interrupt subsystem. Otherwise I am ok with this change. thanks, suresh -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Jun 29, 2009 at 03:46:56AM -0700, Suresh Siddha wrote: > On Mon, 2009-06-29 at 03:08 -0700, Gleb Natapov wrote: > > > > - local_irq_save(flags); > > mask_IO_APIC_setup(ioapic_entries); > > - mask_8259A(); > > Is there a reason why the 8259 mask/unmask operations are separated from > io-apic mask/unmask operations. > I left interrupt masking in enable_IR_x2apic() because interrupt should be masked during transition to x2apic mode, so it can't be moved to enable_IR(). I moved io-apic into enable_IR() because the state of io-apic depend on whether IR was enabled or not, so I left it close to IR enabling logic. Also io-apic masking can fail, but we still want to enable x2apic on KVM if possible, and moving io-apic masking to enable_IR_x2apic() will complicate the logic. > Can we keep it together so that it will be easy to read and understand > that we first do the interrupt subsystem mask, try enabling IR and > x2apic and unmask the interrupt subsystem. > For io-apic this is not 100% symmetric. We mask io-apic, enable IR and if this succeeds we do not unmask io-apic. > Otherwise I am ok with this change. > > thanks, > suresh -- Gleb. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, 2009-06-29 at 04:25 -0700, Gleb Natapov wrote: > I left interrupt masking in enable_IR_x2apic() because interrupt should > be masked during transition to x2apic mode, so it can't be moved to > enable_IR(). I moved io-apic into enable_IR() because the state of > io-apic depend on whether IR was enabled or not, so I left it close to > IR enabling logic. This can be based on the outcome of enable_IR(). > Also io-apic masking can fail, but we still want to > enable x2apic on KVM if possible, and moving io-apic masking to > enable_IR_x2apic() will complicate the logic. io-apic masking can fail because of memory allocation failures, right? If we want a simple solution, we can skip both x2apic and IR. This is not going to be common scenario anyhow. If we really hit this, we will worry about more things than x2apic :) > > > Can we keep it together so that it will be easy to read and understand > > that we first do the interrupt subsystem mask, try enabling IR and > > x2apic and unmask the interrupt subsystem. > > > For io-apic this is not 100% symmetric. We mask io-apic, enable IR and > if this succeeds we do not unmask io-apic. Correct. Can we do the same, otherwise we might have a situation (under kvm atleast) where io-apic will be active while we will be changing the cpu mode. Just trying to be defensive as these might lead to hard to debug corner case conditions. thanks, suresh -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index d1430ef..3e5b6ea 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -260,12 +260,15 @@ config SMP config X86_X2APIC bool "Support x2apic" - depends on X86_LOCAL_APIC && X86_64 && INTR_REMAP + depends on X86_LOCAL_APIC && X86_64 ---help--- This enables x2apic support on CPUs that have this feature. This allows 32-bit apic IDs (so it can support very large systems), - and accesses the local apic via MSRs not via mmio. + and accesses the local apic via MSRs not via mmio. CONFIG_INTR_REMAP + is required for x2apic to take affect when running on physical + machine. If you intend to run the kernel as KVM guest x2apic can + be used without interrupt remapping. If you don't know what to do here, say N. diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c index 8c7c042..19d97ae 100644 --- a/arch/x86/kernel/apic/apic.c +++ b/arch/x86/kernel/apic/apic.c @@ -49,6 +49,7 @@ #include <asm/mtrr.h> #include <asm/smp.h> #include <asm/mce.h> +#include <asm/kvm_para.h> unsigned int num_processors; @@ -1363,11 +1364,10 @@ void enable_x2apic(void) } #endif /* CONFIG_X86_X2APIC */ -void __init enable_IR_x2apic(void) +int __init enable_IR(void) { #ifdef CONFIG_INTR_REMAP int ret; - unsigned long flags; struct IO_APIC_route_entry **ioapic_entries = NULL; ret = dmar_table_init(); @@ -1381,11 +1381,10 @@ void __init enable_IR_x2apic(void) goto ir_failed; } - if (!x2apic_preenabled && skip_ioapic_setup) { pr_info("Skipped enabling intr-remap because of skipping " "io-apic setup\n"); - return; + goto ir_failed; } ioapic_entries = alloc_ioapic_entries(); @@ -1400,9 +1399,7 @@ void __init enable_IR_x2apic(void) goto end; } - local_irq_save(flags); mask_IO_APIC_setup(ioapic_entries); - mask_8259A(); ret = enable_intr_remapping(x2apic_supported()); if (ret) @@ -1410,12 +1407,7 @@ void __init enable_IR_x2apic(void) pr_info("Enabled Interrupt-remapping\n"); - if (x2apic_supported() && !x2apic_mode) { - x2apic_mode = 1; - enable_x2apic(); - pr_info("Enabled x2apic\n"); - } - + return 1; end_restore: if (ret) /* @@ -1423,30 +1415,57 @@ end_restore: */ restore_IO_APIC_setup(ioapic_entries); - unmask_8259A(); - local_irq_restore(flags); - end: if (ioapic_entries) free_ioapic_entries(ioapic_entries); - if (!ret) - return; - ir_failed: - if (x2apic_preenabled) +#endif + return 0; +} + +void __init enable_IR_x2apic(void) +{ + unsigned long flags; + + local_irq_save(flags); + mask_8259A(); + + if (!enable_IR()) { + /* IR is required if x2apic is enabled by BIOS even + * when running in kvm since this indicates present + * of APIC ID > 255 */ + if (x2apic_preenabled || !kvm_para_available()) + goto nox2apic; + else + /* without IR all CPUs can be addressed by IOAPIC/MSI + * only in physical mode */ + x2apic_phys = 1; + } + + if (x2apic_supported() && !x2apic_mode) { + x2apic_mode = 1; + enable_x2apic(); + pr_info("Enabled x2apic\n"); + } + + unmask_8259A(); + local_irq_restore(flags); + return; + +nox2apic: + unmask_8259A(); + local_irq_restore(flags); + + if (x2apic_preenabled) { +#ifdef CONFIG_INTR_REMAP panic("x2apic enabled by bios. But IR enabling failed"); - else if (cpu_has_x2apic) - pr_info("Not enabling x2apic,Intr-remapping\n"); #else - if (!cpu_has_x2apic) - return; - - if (x2apic_preenabled) panic("x2apic enabled prior OS handover," " enable CONFIG_X86_X2APIC, CONFIG_INTR_REMAP"); #endif - + } else if (cpu_has_x2apic) + pr_info("Not enabling x2apic,Intr-remapping\n"); return; }
KVM would like to provide x2APIC interface to a guest without emulating interrupt remapping device. The reason KVM prefers guest to use x2APIC is that x2APIC interface is better virtualizable and provides better performance than mmio xAPIC interface: - msr exits are faster than mmio (no page table walk, emulation) - no need to read back ICR to look at the busy bit - one 64 bit ICR write instead of two 32 bit writes - shared code with the Hyper-V paravirt interface Included patch changes x2APIC enabling logic to enable it even if IR initialization failed, but kernel runs under KVM and no apic id is greater than 255 (if there is one spec requires BIOS to move to x2apic mode before starting an OS). Signed-off-by: Gleb Natapov <gleb@redhat.com> --- v2: Add help message to Kconfig Force physical mode if IR is not available. -- Gleb. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html