From: David Laight <David.Laight@xxxxxxxxxx>
Date: Tue, 1 Jul 2014 13:14:32 +0000
> From: David Laight
>> Even if memset() is inlined (as on x86) using it to zero the union
>> generates a memory word write of zero, followed by a write of the
>> smaller field, and then a read of the word.
>> As well as being a lot of instructions the sequence is unlikely to
>> be optimised by the store-load forward hardware so will be slow.
>>
>> Instead allocate a field of the union that is the same size as the
>> entire union and write a zero value to it. The compiler will then
>> generate the required value in a register.
>>
>> Signed-off-by: David Laight <david.laight@xxxxxxxxxx>
>> ---
>> include/net/sctp/command.h | 6 +++---
>> 1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/include/net/sctp/command.h b/include/net/sctp/command.h
>> index 020eb95..5fcd1a7 100644
>> --- a/include/net/sctp/command.h
>> +++ b/include/net/sctp/command.h
> ...
>> @@ -212,7 +213,6 @@ typedef struct {
>> */
>> static inline int sctp_init_cmd_seq(sctp_cmd_seq_t *seq)
>> {
>> - memset(seq, 0, sizeof(sctp_cmd_seq_t));
>> return 1; /* We always succeed. */
>> }
>
> I've realised that this version of this chunk leaves seq->next_free_slot
> and seq->next_cmd uninitialized (they need to be zero).
> The next patch in the series contains different initialisers - which are correct.
> I guess this needs fixing for 'git bisect' ?
Yes, it does.
--
To unsubscribe from this list: send the line "unsubscribe linux-sctp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
