Re: parisc: adjust L1_CACHE_BYTES to 128 bytes on PA8800 and PA8900 CPUs

Re: parisc: adjust L1_CACHE_BYTES to 128 bytes on PA8800 and PA8900 CPUs

[Andreas Ziegler]

Hi Helge,

today's linux-next tree (next-20150903) contains commit 20f924902ff6
("parisc: adjust L1_CACHE_BYTES to 128 bytes on PA8800 and PA8900 CPUs")
which you authored.

I noticed it because we[0] are running a daily analysis on all commits
in linux-next as part of our research and our tools reported it.

In the patch, you create the following #if defined() structure in
arch/parisc/include/asm/cache.h (lines 16 and following):

 #if defined(CONFIG_PA8X00)
  ...
 #elif defined(CONFIG_PA20)
  ...
 #else
  ...
 #endif

In Kconfig, CONFIG_PA20 is defined as the following
(arch/parisc/Kconfig, line 163):

config PA20
  def_bool y
  depends on PA8X00

This means that CONFIG_PA20 can and will only be enabled if
CONFIG_PA8X00 has already been enabled, which means that the contents of
the "#elif defined(CONFIG_PA20)" block can never be reached: its
condition is only evaluated if CONFIG_PA8X00 is disabled, but then
CONFIG_PA20 can never be enabled either.

Best regards,

Andreas

[0] https://cados.cs.fau.de

Re: parisc: adjust L1_CACHE_BYTES to 128 bytes on PA8800 and PA8900 CPUs

[Helge Deller]

Hi Andreas,

On 03.09.2015 11:23, Andreas Ziegler wrote:
> today's linux-next tree (next-20150903) contains commit 20f924902ff6
> ("parisc: adjust L1_CACHE_BYTES to 128 bytes on PA8800 and PA8900 CPUs")
> which you authored.
> 
> I noticed it because we[0] are running a daily analysis on all commits
> in linux-next as part of our research and our tools reported it.
> 
> In the patch, you create the following #if defined() structure in
> arch/parisc/include/asm/cache.h (lines 16 and following):
> 
>  #if defined(CONFIG_PA8X00)
>   ...
>  #elif defined(CONFIG_PA20)
>   ...
>  #else
>   ...
>  #endif
> 
> In Kconfig, CONFIG_PA20 is defined as the following
> (arch/parisc/Kconfig, line 163):
> 
> config PA20
>   def_bool y
>   depends on PA8X00
> 
> This means that CONFIG_PA20 can and will only be enabled if
> CONFIG_PA8X00 has already been enabled, which means that the contents of
> the "#elif defined(CONFIG_PA20)" block can never be reached: its
> condition is only evaluated if CONFIG_PA8X00 is disabled, but then
> CONFIG_PA20 can never be enabled either.

That's absolutely true.
I'll fix it with the next iteration of this patch.

Thanks,
Helge