Re: CONFIG_XPS depends on L1_CACHE_BYTES being greater than sizeof(struct xps_map)

[Helge Deller <deller@gmx.de>]

On 22.10.2015 23:37, Tom Herbert wrote:
> On Thu, Oct 22, 2015 at 1:00 PM, Helge Deller <deller@gmx.de> wrote:
>> Hi Tom & David,
>>
>> I've queued-up a patch for the parisc architecture which reduces L1_=
CACHE_BYTES from 32 to 16:
>>   https://patchwork.kernel.org/patch/7399291/
>>
>> But this change will break the kernel build like this:
>>
>> In file included from net/core/dev.c:92:0:
>> net/core/dev.c: In function =E2=80=98expand_xps_map=E2=80=99:
>> include/linux/netdevice.h:721:27: warning: overflow in implicit cons=
tant conversion [-Woverflow]
>>    #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_ma=
p)) \
>> net/core/dev.c:1972:18: note: in expansion of macro =E2=80=98XPS_MIN=
_MAP_ALLOC=E2=80=99
>>    int alloc_len =3D XPS_MIN_MAP_ALLOC;
>>
>> Do you see an easy way to fix this ?
>>
> How about
>=20
>  #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - ((sizeof(struct xps_map=
)
> % L1_CACHE_BYTES)) \

The full line would then be:
#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - (sizeof(struct xps_map) % =
L1_CACHE_BYTES)) / sizeof(u16))

The only problem I see with this is, that XPS_MIN_MAP_ALLOC might becom=
e zero.
In that case the call to kzalloc_node() in expand_xps_map() doesn't all=
ocate any memory for the queues.

Helge
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc"=
 in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html