Re: [PATCH 1/2] taskstats: factor out reply assembling

[Shailabh Nagar <nagar@watson.ibm.com>]

Thomas Graf wrote:
> * Oleg Nesterov <oleg@tv-sign.ru> 2006-11-01 21:26
>> Introduce mk_reply() helper which does all nla_put()s on reply.
>>
>> Saves 453 bytes and a preparation for the next patch.
>>
>> Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
>>
>>  taskstats.c |   55 ++++++++++++++++++++++++++++---------------------------
>>  1 files changed, 28 insertions(+), 27 deletions(-)
>>
>> --- STATS/kernel/taskstats.c~5_factor	2006-10-31 16:33:56.000000000 +0300
>> +++ STATS/kernel/taskstats.c	2006-11-01 14:00:03.000000000 +0300
>> @@ -348,6 +348,25 @@ static int parse(struct nlattr *na, cpum
>>  	return ret;
>>  }
>>  
>> +static int mk_reply(struct sk_buff *skb, int type, u32 pid, struct taskstats *stats)
>> +{
>> +	struct nlattr *na;
>> +	int aggr;
>> +
>> +	aggr = TASKSTATS_TYPE_AGGR_TGID;
>> +	if (type == TASKSTATS_TYPE_PID)
>> +		aggr = TASKSTATS_TYPE_AGGR_PID;
>> +
>> +	na = nla_nest_start(skb, aggr);

	if (!na)
		goto nla_put_failure;

>> +	NLA_PUT_U32(skb, type, pid);
>> +	NLA_PUT_TYPE(skb, struct taskstats, TASKSTATS_TYPE_STATS, *stats);
>> +	nla_nest_end(skb, na);
>> +
>> +	return 0;
>> +nla_put_failure:
>> +	return -1;
>> +}
> 
> nla_nest_start() may return NULL, either rely on prepare_reply() to be
> correct and BUG() on failure or do proper error handling for all
> functions.

Thanks for catching that bug which was present in the unrefactored code.
The error path from prepare_reply() which generates the skb is being
handled already.

But as you point out, nla_nest_start failure, unlike those of the NLA_PUT_*
macros, isn't handled by having a nla_put_failure label alone. So we'll
need to add something like the above since we certainly don't want a BUG()
on failure !

Thanks,
Shailabh