[PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region from Trigger resources

[PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region from Trigger resources

[Yazen Ghannam]

From: Yazen Ghannam <yazen.ghannam@amd.com>

ACPI defines a number of instructions to use for triggering errors. However
we are currently removing the address resources from the trigger resources
for only the WRITE_REGISTER_VALUE instruction. This leads to a resource
conflict for any other valid instruction.

Remove the WRITE_REGISTER_VALUE instruction check to fix the resource
conflict with other instructions.

Fixes: b4e008dc53a3 ("ACPI, APEI, EINJ, Refine the fix of resource conflict")
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
---
 drivers/acpi/apei/einj.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/acpi/apei/einj.c b/drivers/acpi/apei/einj.c
index ec50c32ea3da..e0f201ea161d 100644
--- a/drivers/acpi/apei/einj.c
+++ b/drivers/acpi/apei/einj.c
@@ -281,7 +281,6 @@ static struct acpi_generic_address *einj_get_trigger_parameter_region(
 		((char *)trigger_tab + sizeof(struct acpi_einj_trigger));
 	for (i = 0; i < trigger_tab->entry_count; i++) {
 		if (entry->action == ACPI_EINJ_TRIGGER_ERROR &&
-		entry->instruction == ACPI_EINJ_WRITE_REGISTER_VALUE &&
 		entry->register_region.space_id ==
 			ACPI_ADR_SPACE_SYSTEM_MEMORY &&
 		(entry->register_region.address & param2) == (param1 & param2))
-- 
2.7.4

RE: [PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region from Trigger resources

[Ghannam, Yazen]

+EDAC, Boris and Tony for RAS comments.

Any comments from ACPI folks?

Thanks,
Yazen

> -----Original Message-----
> From: Ghannam, Yazen
> Sent: Thursday, August 10, 2017 1:58 PM
> To: linux-acpi@vger.kernel.org
> Cc: linux-kernel@vger.kernel.org; rjw@rjwysocki.net; lenb@kernel.org;
> Ghannam, Yazen <Yazen.Ghannam@amd.com>
> Subject: [PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region from
> Trigger resources
> 
> From: Yazen Ghannam <yazen.ghannam@amd.com>
> 
> ACPI defines a number of instructions to use for triggering errors. However
> we are currently removing the address resources from the trigger resources
> for only the WRITE_REGISTER_VALUE instruction. This leads to a resource
> conflict for any other valid instruction.
> 
> Remove the WRITE_REGISTER_VALUE instruction check to fix the resource
> conflict with other instructions.
> 
> Fixes: b4e008dc53a3 ("ACPI, APEI, EINJ, Refine the fix of resource conflict")
> Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
> ---
>  drivers/acpi/apei/einj.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/drivers/acpi/apei/einj.c b/drivers/acpi/apei/einj.c
> index ec50c32ea3da..e0f201ea161d 100644
> --- a/drivers/acpi/apei/einj.c
> +++ b/drivers/acpi/apei/einj.c
> @@ -281,7 +281,6 @@ static struct acpi_generic_address
> *einj_get_trigger_parameter_region(
>  		((char *)trigger_tab + sizeof(struct acpi_einj_trigger));
>  	for (i = 0; i < trigger_tab->entry_count; i++) {
>  		if (entry->action == ACPI_EINJ_TRIGGER_ERROR &&
> -		entry->instruction == ACPI_EINJ_WRITE_REGISTER_VALUE &&
>  		entry->register_region.space_id ==
>  			ACPI_ADR_SPACE_SYSTEM_MEMORY &&
>  		(entry->register_region.address & param2) == (param1 &
> param2))
> --
> 2.7.4

Re: [PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region from Trigger resources

[Luck, Tony]

On Tue, Aug 22, 2017 at 02:04:27PM +0000, Ghannam, Yazen wrote:
> > *einj_get_trigger_parameter_region(
> >  		((char *)trigger_tab + sizeof(struct acpi_einj_trigger));
> >  	for (i = 0; i < trigger_tab->entry_count; i++) {
> >  		if (entry->action == ACPI_EINJ_TRIGGER_ERROR &&
> > -		entry->instruction == ACPI_EINJ_WRITE_REGISTER_VALUE &&
> >  		entry->register_region.space_id ==
> >  			ACPI_ADR_SPACE_SYSTEM_MEMORY &&
> >  		(entry->register_region.address & param2) == (param1 &
> > param2))

Maybe you should add the other instruction types that make sense
here (ACPI_EINJ_READ_REGISTER, ACPI_EINJ_READ_REGISTER_VALUE and
ACPI_EINJ_WRITE_REGISTER) as a sanity check against the BIOS
giving you an action in the trigger table that doesn't involve
a memory access?

-Tony

RE: [PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region from Trigger resources

[Ghannam, Yazen]

> -----Original Message-----
> From: Luck, Tony [mailto:tony.luck@intel.com]
> Sent: Tuesday, August 22, 2017 12:03 PM
> To: Ghannam, Yazen <Yazen.Ghannam@amd.com>
> Cc: linux-acpi@vger.kernel.org; linux-kernel@vger.kernel.org;
> rjw@rjwysocki.net; lenb@kernel.org; linux-edac@vger.kernel.org; Borislav
> Petkov <bp@suse.de>
> Subject: Re: [PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region
> from Trigger resources
> 
> On Tue, Aug 22, 2017 at 02:04:27PM +0000, Ghannam, Yazen wrote:
> > > *einj_get_trigger_parameter_region(
> > >  		((char *)trigger_tab + sizeof(struct acpi_einj_trigger));
> > >  	for (i = 0; i < trigger_tab->entry_count; i++) {
> > >  		if (entry->action == ACPI_EINJ_TRIGGER_ERROR &&
> > > -		entry->instruction == ACPI_EINJ_WRITE_REGISTER_VALUE &&
> > >  		entry->register_region.space_id ==
> > >  			ACPI_ADR_SPACE_SYSTEM_MEMORY &&
> > >  		(entry->register_region.address & param2) == (param1 &
> > > param2))
> 
> Maybe you should add the other instruction types that make sense here
> (ACPI_EINJ_READ_REGISTER, ACPI_EINJ_READ_REGISTER_VALUE and
> ACPI_EINJ_WRITE_REGISTER) as a sanity check against the BIOS giving you an
> action in the trigger table that doesn't involve a memory access?
> 

Yes, in that case we can do:
entry->instruction <= ACPI_EINJ_WRITE_REGISTER_VALUE

That would cover all the memory access instructions and protect against invalid
instructions.

How does that sound?

Thanks,
Yazen

Re: [PATCH] ACPI, APEI, EINJ: Subtract any matching Register Region from Trigger resources

[Luck, Tony]

On Wed, Aug 23, 2017 at 04:37:43PM +0000, Ghannam, Yazen wrote:
> > -----Original Message-----
> > From: Luck, Tony [mailto:tony.luck@intel.com]
> > Maybe you should add the other instruction types that make sense here
> > (ACPI_EINJ_READ_REGISTER, ACPI_EINJ_READ_REGISTER_VALUE and
> > ACPI_EINJ_WRITE_REGISTER) as a sanity check against the BIOS giving you an
> > action in the trigger table that doesn't involve a memory access?
> > 
> 
> Yes, in that case we can do:
> entry->instruction <= ACPI_EINJ_WRITE_REGISTER_VALUE
> 
> That would cover all the memory access instructions and protect against invalid
> instructions.
> 
> How does that sound?

Perfect.

Acked-by: Tony Luck <tony.luck@intel.com>

-Tony