All of lore.kernel.org
 help / color / mirror / Atom feed
From: David Gibson <dgibson@redhat.com>
To: Laurent Vivier <lvivier@redhat.com>
Cc: kvm@vger.kernel.org, kvm-ppc@vger.kernel.org, drjones@redhat.com,
	thuth@redhat.com, pbonzini@redhat.com
Subject: Re: [kvm-unit-tests PATCH 2/2] powerpc: restore TOC pointer
Date: Wed, 20 Apr 2016 05:59:42 +0000	[thread overview]
Message-ID: <20160420155942.3123963e@voom.fritz.box> (raw)
In-Reply-To: <1461086788-3102-3-git-send-email-lvivier@redhat.com>

[-- Attachment #1: Type: text/plain, Size: 2789 bytes --]

On Tue, 19 Apr 2016 19:26:28 +0200
Laurent Vivier <lvivier@redhat.com> wrote:

> As the TOC pointer can be corrupted by the main program,
> we must restore it in the exception handler.
> 
> As we know where we are loaded, we can now compute it easily.
> 
> To compute it only in the common part of the exception handler
> (call_handler), store the address of call_handler at an absolute
> address in memory to be able to call the handler from the exception
> table (as SLOF does).
> 
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: Laurent Vivier <lvivier@redhat.com>

So, this looks ok as long as the unit tests are built with a single
TOC.  The more stricly correct approach would be to actually load the
correct TOC pointer for the specific handler function you're going to
jump to.

That would need different paths for ABIv1 and v2, though (in practice
ppc64 vs. ppc64le).  With ABIv1 (ppc64) you'd need to load the TOC
pointer from the procedure descriptor into r2 before jumping to the
code address.  With ABIv2 (ppc64le), I believe there's both a "far" and
"near" entry point to the function - the "far" one has a few
instructions to load the correct TOC before continuing onto the "near"
version.  I forget which entry point the plain symbol references.  If
it's the "far" one, the current code is probably already correct.

> ---
>  powerpc/cstart64.S | 17 ++++++++++++++++-
>  1 file changed, 16 insertions(+), 1 deletion(-)
> 
> diff --git a/powerpc/cstart64.S b/powerpc/cstart64.S
> index c47b67d..1ddfa13 100644
> --- a/powerpc/cstart64.S
> +++ b/powerpc/cstart64.S
> @@ -13,6 +13,8 @@
>  
>  #include "spapr.h"
>  
> +#define P_HANDLER	0x2ff8
> +
>  .section .init
>  
>  /*
> @@ -46,6 +48,11 @@ start:
>  	add	r4, r4, r31
>  	bl	relocate
>  
> +	/* compute address of call_handler */
> +
> +	LOAD_REG_ADDR(r4, call_handler)
> +	std	r4, P_HANDLER(0)
> +
>  	/* relocate vector table to base address 0x0 (MSR_IP = 0) */
>  
>  	/* source: r4, dest end: r5, destination: r6 */
> @@ -166,6 +173,12 @@ call_handler:
>  	mfsrr1	r0
>  	std	r0, _MSR(r1)
>  
> +	/* restore TOC pointer */
> +
> +	LOAD_REG_IMMEDIATE(r31, SPAPR_KERNEL_LOAD_ADDR)
> +	ld	r2, (p_toc - start)(r31)
> +	add	r2, r2, r31
> +
>  	/* FIXME: build stack frame */
>  
>  	/* call generic handler */
> @@ -221,7 +234,7 @@ call_handler:
>  	mfctr	r0
>  	std	r0,_CTR(r1)
>  
> -	LOAD_REG_ADDR(r0, call_handler)
> +	ld	r0, P_HANDLER(0)
>  	mtctr	r0
>  
>  	li	r0,\vec
> @@ -245,3 +258,5 @@ VECTOR(0x900)
>  	.align 7
>  	.globl __end_interrupts
>  __end_interrupts:
> +	.org	P_HANDLER
> +	.llong	0
> -- 
> 2.5.5
> 


-- 
David Gibson <dgibson@redhat.com>
Senior Software Engineer, Virtualization, Red Hat

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

WARNING: multiple messages have this Message-ID (diff)
From: David Gibson <dgibson@redhat.com>
To: Laurent Vivier <lvivier@redhat.com>
Cc: kvm@vger.kernel.org, kvm-ppc@vger.kernel.org, drjones@redhat.com,
	thuth@redhat.com, pbonzini@redhat.com
Subject: Re: [kvm-unit-tests PATCH 2/2] powerpc: restore TOC pointer
Date: Wed, 20 Apr 2016 15:59:42 +1000	[thread overview]
Message-ID: <20160420155942.3123963e@voom.fritz.box> (raw)
In-Reply-To: <1461086788-3102-3-git-send-email-lvivier@redhat.com>

[-- Attachment #1: Type: text/plain, Size: 2789 bytes --]

On Tue, 19 Apr 2016 19:26:28 +0200
Laurent Vivier <lvivier@redhat.com> wrote:

> As the TOC pointer can be corrupted by the main program,
> we must restore it in the exception handler.
> 
> As we know where we are loaded, we can now compute it easily.
> 
> To compute it only in the common part of the exception handler
> (call_handler), store the address of call_handler at an absolute
> address in memory to be able to call the handler from the exception
> table (as SLOF does).
> 
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: Laurent Vivier <lvivier@redhat.com>

So, this looks ok as long as the unit tests are built with a single
TOC.  The more stricly correct approach would be to actually load the
correct TOC pointer for the specific handler function you're going to
jump to.

That would need different paths for ABIv1 and v2, though (in practice
ppc64 vs. ppc64le).  With ABIv1 (ppc64) you'd need to load the TOC
pointer from the procedure descriptor into r2 before jumping to the
code address.  With ABIv2 (ppc64le), I believe there's both a "far" and
"near" entry point to the function - the "far" one has a few
instructions to load the correct TOC before continuing onto the "near"
version.  I forget which entry point the plain symbol references.  If
it's the "far" one, the current code is probably already correct.

> ---
>  powerpc/cstart64.S | 17 ++++++++++++++++-
>  1 file changed, 16 insertions(+), 1 deletion(-)
> 
> diff --git a/powerpc/cstart64.S b/powerpc/cstart64.S
> index c47b67d..1ddfa13 100644
> --- a/powerpc/cstart64.S
> +++ b/powerpc/cstart64.S
> @@ -13,6 +13,8 @@
>  
>  #include "spapr.h"
>  
> +#define P_HANDLER	0x2ff8
> +
>  .section .init
>  
>  /*
> @@ -46,6 +48,11 @@ start:
>  	add	r4, r4, r31
>  	bl	relocate
>  
> +	/* compute address of call_handler */
> +
> +	LOAD_REG_ADDR(r4, call_handler)
> +	std	r4, P_HANDLER(0)
> +
>  	/* relocate vector table to base address 0x0 (MSR_IP = 0) */
>  
>  	/* source: r4, dest end: r5, destination: r6 */
> @@ -166,6 +173,12 @@ call_handler:
>  	mfsrr1	r0
>  	std	r0, _MSR(r1)
>  
> +	/* restore TOC pointer */
> +
> +	LOAD_REG_IMMEDIATE(r31, SPAPR_KERNEL_LOAD_ADDR)
> +	ld	r2, (p_toc - start)(r31)
> +	add	r2, r2, r31
> +
>  	/* FIXME: build stack frame */
>  
>  	/* call generic handler */
> @@ -221,7 +234,7 @@ call_handler:
>  	mfctr	r0
>  	std	r0,_CTR(r1)
>  
> -	LOAD_REG_ADDR(r0, call_handler)
> +	ld	r0, P_HANDLER(0)
>  	mtctr	r0
>  
>  	li	r0,\vec
> @@ -245,3 +258,5 @@ VECTOR(0x900)
>  	.align 7
>  	.globl __end_interrupts
>  __end_interrupts:
> +	.org	P_HANDLER
> +	.llong	0
> -- 
> 2.5.5
> 


-- 
David Gibson <dgibson@redhat.com>
Senior Software Engineer, Virtualization, Red Hat

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

  reply	other threads:[~2016-04-20  5:59 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-04-19 17:26 [kvm-unit-tests PATCH 0/2] Fix TOC value in exception handler Laurent Vivier
2016-04-19 17:26 ` Laurent Vivier
2016-04-19 17:26 ` [kvm-unit-tests PATCH 1/2] powerpc: use well known kernel start address Laurent Vivier
2016-04-19 17:26   ` Laurent Vivier
2016-04-20  2:01   ` David Gibson
2016-04-20  2:01     ` David Gibson
2016-04-20  5:48     ` Laurent Vivier
2016-04-20  5:48       ` Laurent Vivier
2016-04-20  5:55       ` David Gibson
2016-04-20  5:55         ` David Gibson
2016-04-20 10:15   ` Thomas Huth
2016-04-20 10:15     ` Thomas Huth
2016-04-19 17:26 ` [kvm-unit-tests PATCH 2/2] powerpc: restore TOC pointer Laurent Vivier
2016-04-19 17:26   ` Laurent Vivier
2016-04-20  5:59   ` David Gibson [this message]
2016-04-20  5:59     ` David Gibson
2016-04-20 10:28     ` Thomas Huth
2016-04-20 10:28       ` Thomas Huth

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20160420155942.3123963e@voom.fritz.box \
    --to=dgibson@redhat.com \
    --cc=drjones@redhat.com \
    --cc=kvm-ppc@vger.kernel.org \
    --cc=kvm@vger.kernel.org \
    --cc=lvivier@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=thuth@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.