* Re: [PATCH 08/23] Add interrupt handling code
2009-07-07 14:17 [PATCH 08/23] Add interrupt handling code Alexander Graf
@ 2009-07-08 5:24 ` Benjamin Herrenschmidt
2009-07-08 7:27 ` Alexander Graf
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Benjamin Herrenschmidt @ 2009-07-08 5:24 UTC (permalink / raw)
To: kvm-ppc
On Tue, 2009-07-07 at 16:17 +0200, Alexander Graf wrote:
> Getting from host state to the guest is only half the story. We also need
> to return to our host context and handle whatever happened to get us out of
> the guest.
>
> On PowerPC every guest exit is an interrupt. So all we need to do is trap
> the host's interrupt handlers and get into our #VMEXIT code to handle it.
>
> PowerPCs also have a register that can add an offset to the interrupt handlers'
> adresses which is what the booke KVM code uses. Unfortunately that is a
> hypervisor ressource and we also want to be able to run KVM when we're running
> in an LPAR. So we have to hook into the Linux interrupt handlers.
Right, besides it doesn't always work :-)
> Signed-off-by: Alexander Graf <agraf@suse.de>
> ---
> arch/powerpc/kvm/970_rmhandlers.S | 128 +++++++++++++++++++++++++++++++++++++
> 1 files changed, 128 insertions(+), 0 deletions(-)
> create mode 100644 arch/powerpc/kvm/970_rmhandlers.S
>
> diff --git a/arch/powerpc/kvm/970_rmhandlers.S b/arch/powerpc/kvm/970_rmhandlers.S
> new file mode 100644
> index 0000000..91dacc4
> --- /dev/null
> +++ b/arch/powerpc/kvm/970_rmhandlers.S
> @@ -0,0 +1,128 @@
> +/*
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License, version 2, as
> + * published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program; if not, write to the Free Software
> + * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
> + *
> + * Copyright SUSE Linux Products GmbH 2009
> + *
> + * Authors: Alexander Graf <agraf@suse.de>
> + */
> +
> +#include <asm/kvm_970_asm.h>
> +
> +/*****************************************************************************
> + * *
> + * Real Mode handlers that need to be in low physical memory *
> + * *
> + ****************************************************************************/
> +
> +
> +.macro INTERRUPT_TRAMPOLINE intno
> +
> +.global kvmppc_trampoline_\intno
> +kvmppc_trampoline_\intno:
> + /* We're replacing this instruction in the real handler */
> + mtspr SPRN_SPRG1, r13 /* Save r13 */
> + /*
> + * First thing to do is to find out if we're coming
> + * from a KVM guest or a Linux process.
> + *
> + * To distinguish, we only need to check SPRG3 (PACA),
> + * because our entry code sets SPRG3 = PACA | 1, while
> + * Linux has the PACA on at least word boundary.
> + */
> + mfcr r13
> + mtspr SPRN_SPRG2, r13 /* Save CR */
> + mfspr r13, SPRN_SPRG3 /* r13 = PACA | 1 */
> + clrldi. r13, r13, 63 /* CR = ((r13 & 1) = 0) */
> + bne ..kvmppc_handler_hasmagic_\intno
> + /* No KVM guest? Then jump back to the Linux handler! */
> + mfspr r13, SPRN_SPRG2
> + mtcr r13
> + mfspr r13, SPRN_SPRG1 /* r13 = original r13 */
> + b kvmppc_resume_\intno /* Get back original handler */
> +
> + /* Now we know we're handling a KVM guest */
> +..kvmppc_handler_hasmagic_\intno:
> + /* Unset SPRG3 */
> + mfspr r13, SPRN_SPRG3 /* r13 = PACA */
> + clrrdi r13, r13, 1 /* r13 &= ~1 */
> + mtspr SPRN_SPRG3, r13 /* SPRG3 = PACA & ~1 */
What about instead, we make the generic interrupt entry test the PACA
low bit and branch off to a KVM supplied function table ? That would
lower the overhead and avoid having to do any kind of patching no ? I'm
happy to have that logic permanently in there, maybe controlled by a
CONFIG option in case somebody wants to do kernel that never do KVM, if
we can make it such that it doesn't bloat things.
In fact we could even have it there, NOP it out when KVM isn't loaded
and put the instructions back when KVM is.
Cheers,
Ben.
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH 08/23] Add interrupt handling code
2009-07-07 14:17 [PATCH 08/23] Add interrupt handling code Alexander Graf
2009-07-08 5:24 ` Benjamin Herrenschmidt
@ 2009-07-08 7:27 ` Alexander Graf
2009-07-08 7:54 ` Benjamin Herrenschmidt
2009-07-16 13:30 ` Alexander Graf
3 siblings, 0 replies; 5+ messages in thread
From: Alexander Graf @ 2009-07-08 7:27 UTC (permalink / raw)
To: kvm-ppc
On 08.07.2009, at 07:24, Benjamin Herrenschmidt wrote:
> On Tue, 2009-07-07 at 16:17 +0200, Alexander Graf wrote:
>> Getting from host state to the guest is only half the story. We
>> also need
>> to return to our host context and handle whatever happened to get
>> us out of
>> the guest.
>>
>> On PowerPC every guest exit is an interrupt. So all we need to do
>> is trap
>> the host's interrupt handlers and get into our #VMEXIT code to
>> handle it.
>>
>> PowerPCs also have a register that can add an offset to the
>> interrupt handlers'
>> adresses which is what the booke KVM code uses. Unfortunately that
>> is a
>> hypervisor ressource and we also want to be able to run KVM when
>> we're running
>> in an LPAR. So we have to hook into the Linux interrupt handlers.
>
> Right, besides it doesn't always work :-)
>
>> Signed-off-by: Alexander Graf <agraf@suse.de>
>> ---
>> arch/powerpc/kvm/970_rmhandlers.S | 128 +++++++++++++++++++++++++++
>> ++++++++++
>> 1 files changed, 128 insertions(+), 0 deletions(-)
>> create mode 100644 arch/powerpc/kvm/970_rmhandlers.S
>>
>> diff --git a/arch/powerpc/kvm/970_rmhandlers.S b/arch/powerpc/kvm/
>> 970_rmhandlers.S
>> new file mode 100644
>> index 0000000..91dacc4
>> --- /dev/null
>> +++ b/arch/powerpc/kvm/970_rmhandlers.S
>> @@ -0,0 +1,128 @@
>> +/*
>> + * This program is free software; you can redistribute it and/or
>> modify
>> + * it under the terms of the GNU General Public License, version
>> 2, as
>> + * published by the Free Software Foundation.
>> + *
>> + * This program is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
>> + * GNU General Public License for more details.
>> + *
>> + * You should have received a copy of the GNU General Public License
>> + * along with this program; if not, write to the Free Software
>> + * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
>> 02110-1301, USA.
>> + *
>> + * Copyright SUSE Linux Products GmbH 2009
>> + *
>> + * Authors: Alexander Graf <agraf@suse.de>
>> + */
>> +
>> +#include <asm/kvm_970_asm.h>
>> +
>> +/
>> *****************************************************************************
>> +
>> * *
>> + * Real Mode handlers that need to be in low physical
>> memory *
>> +
>> * *
>> +
>> ****************************************************************************/
>> +
>> +
>> +.macro INTERRUPT_TRAMPOLINE intno
>> +
>> +.global kvmppc_trampoline_\intno
>> +kvmppc_trampoline_\intno:
>> + /* We're replacing this instruction in the real handler */
>> + mtspr SPRN_SPRG1, r13 /* Save r13 */
>> + /*
>> + * First thing to do is to find out if we're coming
>> + * from a KVM guest or a Linux process.
>> + *
>> + * To distinguish, we only need to check SPRG3 (PACA),
>> + * because our entry code sets SPRG3 = PACA | 1, while
>> + * Linux has the PACA on at least word boundary.
>> + */
>> + mfcr r13
>> + mtspr SPRN_SPRG2, r13 /* Save CR */
>> + mfspr r13, SPRN_SPRG3 /* r13 = PACA | 1 */
>> + clrldi. r13, r13, 63 /* CR = ((r13 & 1) = 0) */
>> + bne ..kvmppc_handler_hasmagic_\intno
>> + /* No KVM guest? Then jump back to the Linux handler! */
>> + mfspr r13, SPRN_SPRG2
>> + mtcr r13
>> + mfspr r13, SPRN_SPRG1 /* r13 = original r13 */
>> + b kvmppc_resume_\intno /* Get back original handler */
>> +
>> + /* Now we know we're handling a KVM guest */
>> +..kvmppc_handler_hasmagic_\intno:
>> + /* Unset SPRG3 */
>> + mfspr r13, SPRN_SPRG3 /* r13 = PACA */
>> + clrrdi r13, r13, 1 /* r13 &= ~1 */
>> + mtspr SPRN_SPRG3, r13 /* SPRG3 = PACA & ~1 */
>
> What about instead, we make the generic interrupt entry test the PACA
> low bit and branch off to a KVM supplied function table ? That would
> lower the overhead and avoid having to do any kind of patching no ?
> I'm
> happy to have that logic permanently in there, maybe controlled by a
> CONFIG option in case somebody wants to do kernel that never do KVM,
> if
> we can make it such that it doesn't bloat things.
Hum, something like a generic interrupt hooking mechanism? Sounds
cool :-). But not all interrupt handlers use generic code, so we'd
have to duplicate some bits quite some times.
Alex
>
> In fact we could even have it there, NOP it out when KVM isn't loaded
> and put the instructions back when KVM is.
>
> Cheers,
> Ben.
>
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH 08/23] Add interrupt handling code
2009-07-07 14:17 [PATCH 08/23] Add interrupt handling code Alexander Graf
2009-07-08 5:24 ` Benjamin Herrenschmidt
2009-07-08 7:27 ` Alexander Graf
@ 2009-07-08 7:54 ` Benjamin Herrenschmidt
2009-07-16 13:30 ` Alexander Graf
3 siblings, 0 replies; 5+ messages in thread
From: Benjamin Herrenschmidt @ 2009-07-08 7:54 UTC (permalink / raw)
To: kvm-ppc
On Wed, 2009-07-08 at 09:27 +0200, Alexander Graf wrote:
> Hum, something like a generic interrupt hooking mechanism? Sounds
> cool :-). But not all interrupt handlers use generic code, so we'd
> have to duplicate some bits quite some times.
Right, some of them are subtly different. ..
One cool thing however is that they all have about 0x100 bytes of space
and only use a portion of that so we should be able to do something
about it... maybe having 2 variants of the prolog, with and without the
test to jump to KVM for example and "replace" them when KVM is loaded.
I need to think about it a bit more. One thing we may want to do is
instead to set SPRG3 bit 0, instead, is to set a bit in the PACA,
that would make things a lot easier. We could run through the base
prolog and just add three instructions (that can be noped out easily) to
load that flag, test it and branch out of line to a special if set.
In the standard prolog, we can clobber r10, r11 and r12 right after
EXCEPTION_PROLOG_1 and we should be able to clobber CR0 as well since we
just saved CR into r9, so we could probably do something like lbz the
"KVM" flag into r11 before loading PACAKBASE into r12, then compare it
to 0 and branch conditional after loading PACAKMSR into r10 or something
like that (to keep a gap between load and use to avoid back to back
here).
In fact, more fun: You could hack PACAKBASE and PACAMSR :-) But that's a
bit harder to get the offsets right. Would probably work to. Make it
point to a page where the handlers are at the right offset (we could
make them all be at 0x0f0 or so from the base of the exception easily
instead of right after so you don't have to do black magic to find where
to put them or you could just fill the whole 0x100 bytes with copies).
That way, depending on whether you are in KVM or not on that CPU, the
existing code would branch to your secondary handlers with the MSR of
your choice (typically still in real mode) without adding a conditional
branch to the exception entry code.
We can carve out a page down there in the RMA for use by KVM easily, in
fact we probably have some unused space already.
Cheers,
Ben.
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 08/23] Add interrupt handling code
2009-07-07 14:17 [PATCH 08/23] Add interrupt handling code Alexander Graf
` (2 preceding siblings ...)
2009-07-08 7:54 ` Benjamin Herrenschmidt
@ 2009-07-16 13:30 ` Alexander Graf
3 siblings, 0 replies; 5+ messages in thread
From: Alexander Graf @ 2009-07-16 13:30 UTC (permalink / raw)
To: kvm-ppc
Getting from host state to the guest is only half the story. We also need
to return to our host context and handle whatever happened to get us out of
the guest.
On PowerPC every guest exit is an interrupt. So all we need to do is trap
the host's interrupt handlers and get into our #VMEXIT code to handle it.
PowerPCs also have a register that can add an offset to the interrupt handlers'
adresses which is what the booke KVM code uses. Unfortunately that is a
hypervisor ressource and we also want to be able to run KVM when we're running
in an LPAR. So we have to hook into the Linux interrupt handlers.
Signed-off-by: Alexander Graf <agraf@suse.de>
---
arch/powerpc/kvm/book3s_64_rmhandlers.S | 133 +++++++++++++++++++++++++++++++
1 files changed, 133 insertions(+), 0 deletions(-)
create mode 100644 arch/powerpc/kvm/book3s_64_rmhandlers.S
diff --git a/arch/powerpc/kvm/book3s_64_rmhandlers.S b/arch/powerpc/kvm/book3s_64_rmhandlers.S
new file mode 100644
index 0000000..cb70dfd
--- /dev/null
+++ b/arch/powerpc/kvm/book3s_64_rmhandlers.S
@@ -0,0 +1,133 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ *
+ * Copyright SUSE Linux Products GmbH 2009
+ *
+ * Authors: Alexander Graf <agraf@suse.de>
+ */
+
+#include <asm/ppc_asm.h>
+#include <asm/kvm_asm.h>
+#include <asm/reg.h>
+#include <asm/page.h>
+#include <asm/asm-offsets.h>
+#include <asm/exception.h>
+
+/*****************************************************************************
+ * *
+ * Real Mode handlers that need to be in low physical memory *
+ * *
+ ****************************************************************************/
+
+
+.macro INTERRUPT_TRAMPOLINE intno
+
+.global kvmppc_trampoline_\intno
+kvmppc_trampoline_\intno:
+ /* We're replacing this instruction in the real handler */
+ mtspr SPRN_SPRG1, r13 /* Save r13 */
+ /*
+ * First thing to do is to find out if we're coming
+ * from a KVM guest or a Linux process.
+ *
+ * To distinguish, we only need to check SPRG3 (PACA),
+ * because our entry code sets SPRG3 = PACA | 1, while
+ * Linux has the PACA on at least word boundary.
+ */
+ mfcr r13
+ mtspr SPRN_SPRG2, r13 /* Save CR */
+ mfspr r13, SPRN_SPRG3 /* r13 = PACA | 1 */
+ clrldi. r13, r13, 63 /* CR = ((r13 & 1) = 0) */
+ bne ..kvmppc_handler_hasmagic_\intno
+ /* No KVM guest? Then jump back to the Linux handler! */
+ mfspr r13, SPRN_SPRG2
+ mtcr r13
+ mfspr r13, SPRN_SPRG1 /* r13 = original r13 */
+ b kvmppc_resume_\intno /* Get back original handler */
+
+ /* Now we know we're handling a KVM guest */
+..kvmppc_handler_hasmagic_\intno:
+ /* Unset SPRG3 */
+ mfspr r13, SPRN_SPRG3 /* r13 = PACA */
+ clrrdi r13, r13, 1 /* r13 &= ~1 */
+ mtspr SPRN_SPRG3, r13 /* SPRG3 = PACA & ~1 */
+
+ /* As of here PACA is in r13 and we're safe to do something */
+
+ std r1, (PACA_EXMC+EX_R9)(r13)
+ std r10, (PACA_EXMC+EX_R10)(r13)
+ std r11, (PACA_EXMC+EX_R11)(r13)
+ std r12, (PACA_EXMC+EX_R12)(r13)
+ std r2, (PACA_EXMC+EX_R13)(r13)
+
+ mfsrr0 r10
+ mfsrr1 r11
+
+ /* Restore R1/R2 so we can handle faults */
+ ld r1, PACAR1(r13)
+ ld r2, (PACA_EXMC+EX_SRR0)(r13)
+
+ /* Let's store which interrupt we're handling */
+ li r12, \intno
+
+ /* Jump into the SLB exit code that goes to the highmem handler */
+ b kvmppc_handler_trampoline_exit
+
+.endm
+
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_SYSTEM_RESET
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_MACHINE_CHECK
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_DATA_STORAGE
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_DATA_SEGMENT
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_INST_STORAGE
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_INST_SEGMENT
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_EXTERNAL
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_ALIGNMENT
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_PROGRAM
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_FP_UNAVAIL
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_DECREMENTER
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_SYSCALL
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_TRACE
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_PERFMON
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_ALTIVEC
+INTERRUPT_TRAMPOLINE BOOK3S_INTERRUPT_VSX
+
+/*
+ * This trampoline brings us back to a real mode handler
+ *
+ * Input Registers:
+ *
+ * R6 = SRR0
+ * R7 = SRR1
+ * LR = real-mode IP
+ *
+ */
+.global kvmppc_handler_lowmem_trampoline
+kvmppc_handler_lowmem_trampoline:
+
+ mtsrr0 r6
+ mtsrr1 r7
+ blr
+kvmppc_handler_lowmem_trampoline_end:
+
+.global kvmppc_trampoline_lowmem
+kvmppc_trampoline_lowmem:
+ .long kvmppc_handler_lowmem_trampoline - _stext
+
+.global kvmppc_trampoline_enter
+kvmppc_trampoline_enter:
+ .long kvmppc_handler_trampoline_enter - _stext
+
+#include "book3s_64_slb.S"
+
--
1.6.0.2
^ permalink raw reply related [flat|nested] 5+ messages in thread