Re: [0/2] powerpc/powernv/vas: Adjustments for two function implementations

[Christophe Leroy]

Le 16/04/2024 à 13:11, Michael Ellerman a écrit :
> Markus Elfring <Markus.Elfring@xxxxxx> writes:
>>> A few update suggestions were taken into account
>>> from static source code analysis.
>>>
>>> Markus Elfring (2):
>>
>> I would appreciate a bit more information about the reasons
>> why this patch series was rejected.
>>
>>
>>>    One function call less in vas_window_alloc() after error detection
>>
>> https://patchwork.ozlabs.org/project/linuxppc-dev/patch/1f1c21cf-c34c-418c-b00c-8e6474f12612@xxxxxx/
> 
> It introduced a new goto and label to avoid a kfree(NULL) call, but
> kfree() explicitly accepts NULL and handles it. So it complicates the
> source code for no gain.

This is explicit in Kernel documentation:

/**
  * kfree - free previously allocated memory
  * @object: pointer returned by kmalloc() or kmem_cache_alloc()
  *
  * If @object is NULL, no operation is performed.
  */

That's exactly the same behaviour as free() in libc.

So Coccinelle should be fixed if it reports an error for that.

> 
>>>    Return directly after a failed kasprintf() in map_paste_region()
>>
>> https://patchwork.ozlabs.org/project/linuxppc-dev/patch/f46f04bc-613c-4e98-b602-4c5120556b09@xxxxxx/
> 
> Basically the same reasoning. And it also changes the function from
> having two return paths (success and error), to three.
> 

Looking at that function, I however see a missing region release when 
ioremap_cache() fails.

Christophe