cacheinfo: Correctly handle new acpi_get_cache_info() prototype

[PATCH -next] cacheinfo: Correctly handle new acpi_get_cache_info() prototype

Posted by Pierre Gondois 2 years, 7 months ago

commit bd500361a937 ("ACPI: PPTT: Update acpi_find_last_cache_level()
to acpi_get_cache_info()")
updates the function acpi_get_cache_info().

If CONFIG_ACPI_PPTT is not defined, acpi_get_cache_info() doesn't
update its *levels and *split_levels parameters and returns 0.
This can lead to a faulty behaviour.

Make acpi_get_cache_info() return an error code if CONFIG_ACPI_PPTT
is not defined. Initialize levels and split_levels before passing
their address to acpi_get_cache_info().

Also, in init_cache_level():
- commit e75d18cecbb3 ("arm64: cacheinfo: Fix incorrect
  assignment of signed error value to unsigned fw_level")
  checks the fw_level value in init_cache_level() in case
  the value is negative. Remove this check as the error code
  is not returned through fw_level anymore.
- if no PPTT is present or CONFIG_ACPI_PPTT is not defined,
  it is still possible to use the cache information from clidr_el1.
  Instead of aborting if acpi_get_cache_info() returns an error
  code, just continue.

Reported-by: Dan Carpenter <error27@gmail.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Pierre Gondois <pierre.gondois@arm.com>
---
 arch/arm64/kernel/cacheinfo.c | 5 +----
 drivers/base/cacheinfo.c      | 2 +-
 include/linux/cacheinfo.h     | 2 +-
 3 files changed, 3 insertions(+), 6 deletions(-)

diff --git a/arch/arm64/kernel/cacheinfo.c b/arch/arm64/kernel/cacheinfo.c
index 36c3b07cdf2d..3ba70985e3a2 100644
--- a/arch/arm64/kernel/cacheinfo.c
+++ b/arch/arm64/kernel/cacheinfo.c
@@ -64,12 +64,9 @@ int init_cache_level(unsigned int cpu)
 	} else {
 		ret = acpi_get_cache_info(cpu, &fw_level, NULL);
 		if (ret < 0)
-			return ret;
+			fw_level = 0;
 	}
 
-	if (fw_level < 0)
-		return fw_level;
-
 	if (level < fw_level) {
 		/*
 		 * some external caches not specified in CLIDR_EL1
diff --git a/drivers/base/cacheinfo.c b/drivers/base/cacheinfo.c
index b57fbd0d7114..f184ef7dc1d2 100644
--- a/drivers/base/cacheinfo.c
+++ b/drivers/base/cacheinfo.c
@@ -414,7 +414,7 @@ int allocate_cache_info(int cpu)
 int fetch_cache_info(unsigned int cpu)
 {
 	struct cpu_cacheinfo *this_cpu_ci;
-	unsigned int levels, split_levels;
+	unsigned int levels = 0, split_levels = 0;
 	int ret;
 
 	if (acpi_disabled) {
diff --git a/include/linux/cacheinfo.h b/include/linux/cacheinfo.h
index dfef57077cd0..908e19d17f49 100644
--- a/include/linux/cacheinfo.h
+++ b/include/linux/cacheinfo.h
@@ -100,7 +100,7 @@ static inline
 int acpi_get_cache_info(unsigned int cpu,
 			unsigned int *levels, unsigned int *split_levels)
 {
-	return 0;
+	return -ENOENT;
 }
 #else
 int acpi_get_cache_info(unsigned int cpu,
-- 
2.25.1

Re: [PATCH -next] cacheinfo: Correctly handle new acpi_get_cache_info() prototype

Posted by Conor Dooley 2 years, 7 months ago

Hey!

On Tue, Jan 24, 2023 at 01:34:46PM +0100, Pierre Gondois wrote:
> commit bd500361a937 ("ACPI: PPTT: Update acpi_find_last_cache_level()
> to acpi_get_cache_info()")
> updates the function acpi_get_cache_info().
> 
> If CONFIG_ACPI_PPTT is not defined, acpi_get_cache_info() doesn't
> update its *levels and *split_levels parameters and returns 0.
> This can lead to a faulty behaviour.
> 
> Make acpi_get_cache_info() return an error code if CONFIG_ACPI_PPTT
> is not defined. Initialize levels and split_levels before passing
> their address to acpi_get_cache_info().
> 
> Also, in init_cache_level():

Hmm...

> - commit e75d18cecbb3 ("arm64: cacheinfo: Fix incorrect
>   assignment of signed error value to unsigned fw_level")
>   checks the fw_level value in init_cache_level() in case
>   the value is negative. Remove this check as the error code
>   is not returned through fw_level anymore.
> - if no PPTT is present or CONFIG_ACPI_PPTT is not defined,
>   it is still possible to use the cache information from clidr_el1.
>   Instead of aborting if acpi_get_cache_info() returns an error
>   code, just continue.

To be honest, these feel like entirely separate things that should be
in different patches. You've got:
- Dan's smatch fixes
- a redundant check being removed
- a behaviour change for if acpi_get_cache_info() returns an error

> Reported-by: Dan Carpenter <error27@gmail.com>
> Reported-by: kernel test robot <lkp@intel.com>

How about Link: to the LKP/Dan's report?
Link: https://lore.kernel.org/all/Y86iruJPuwNN7rZw@kili/

I did a quick check but didn't don't see the LKP report...

Also a Fixes: tag too, no?

Thanks,
Conor.

Re: [PATCH -next] cacheinfo: Correctly handle new acpi_get_cache_info() prototype

Posted by Sudeep Holla 2 years, 7 months ago

On Tue, Jan 24, 2023 at 01:31:06PM +0000, Conor Dooley wrote:
> Hey!
> 
> On Tue, Jan 24, 2023 at 01:34:46PM +0100, Pierre Gondois wrote:
> > commit bd500361a937 ("ACPI: PPTT: Update acpi_find_last_cache_level()
> > to acpi_get_cache_info()")
> > updates the function acpi_get_cache_info().
> > 
> > If CONFIG_ACPI_PPTT is not defined, acpi_get_cache_info() doesn't
> > update its *levels and *split_levels parameters and returns 0.
> > This can lead to a faulty behaviour.
> > 
> > Make acpi_get_cache_info() return an error code if CONFIG_ACPI_PPTT
> > is not defined. Initialize levels and split_levels before passing
> > their address to acpi_get_cache_info().
> > 
> > Also, in init_cache_level():
> 
> Hmm...
> 
> > - commit e75d18cecbb3 ("arm64: cacheinfo: Fix incorrect
> >   assignment of signed error value to unsigned fw_level")
> >   checks the fw_level value in init_cache_level() in case
> >   the value is negative. Remove this check as the error code
> >   is not returned through fw_level anymore.
> > - if no PPTT is present or CONFIG_ACPI_PPTT is not defined,
> >   it is still possible to use the cache information from clidr_el1.
> >   Instead of aborting if acpi_get_cache_info() returns an error
> >   code, just continue.
> 
> To be honest, these feel like entirely separate things that should be
> in different patches. You've got:
> - Dan's smatch fixes
> - a redundant check being removed
> - a behaviour change for if acpi_get_cache_info() returns an error
>

I am not too fussy about it, but for sure it would be cleaner for sure.

> > Reported-by: Dan Carpenter <error27@gmail.com>
> > Reported-by: kernel test robot <lkp@intel.com>
>
> How about Link: to the LKP/Dan's report?
> Link: https://lore.kernel.org/all/Y86iruJPuwNN7rZw@kili/
>
> I did a quick check but didn't don't see the LKP report...
>

Yes, LKP dropped all the cc when reported, even I saw after merging the
changes. I think this is the one:

https://lore.kernel.org/all/202301052307.JYt1GWaJ-lkp@intel.com/

> Also a Fixes: tag too, no?
>

+1, if you split make sure you tag fixes to the right one(mainly one
that changes return from acpi_get_cache_info())

> Thanks,
> Conor.



-- 
Regards,
Sudeep