Re: [PATCH 2/2] MIPS: Optimize TLB handlers for Octeon CPUs

[David Daney <ddaney@caviumnetworks.com>]

On 01/19/2011 11:46 AM, Jonas Gorski wrote:
> On 19 January 2011 20:41, David Daney<ddaney@caviumnetworks.com>  wrote:
>> On 01/19/2011 11:35 AM, Jonas Gorski wrote:
>>>
>>> On 28/12/2010, David Daney<ddaney@caviumnetworks.com>    wrote:
>>>>
>>>> +#if defined(CONFIG_CAVIUM_OCTEON_CVMSEG_SIZE)&&    \
>>>> +    CONFIG_CAVIUM_OCTEON_CVMSEG_SIZE>    0
>>>> (...)
>>>> +#else
>>>> +static bool scratchpad_available(void)
>>>> +{
>>>> +       return false;
>>>> +}
>>>> +static int scratchpad_offset(int i)
>>>> +{
>>>> +       BUG();
>>>> +}
>>>> +#endif
>>>
>>> This seems to have broken the build for any non-octeon mips build:
>>>
>>>    CC      arch/mips/mm/tlbex.o
>>> cc1: warnings being treated as errors
>>> arch/mips/mm/tlbex.c: In function 'scratchpad_offset':
>>> arch/mips/mm/tlbex.c:112: error: no return statement in function
>>> returning non-void
>>>
>>
>> Can you tell me which version of GCC you are using?
>>
>> I tested it with gcc-4.5.x, BUG() may have problems if builtin_unreachable
>> is not available.
>
> That's probably it, It's a 4.3.3 (with code sourcery extensions, the
> OpenWrt default one).
>

It is a bug in GCC-4.3.  The proper fix is to add 'return 0;' after that 
BUG() statement.

I will prepare a patch.

David Daney