Re: [PATCH 3/4] ipc/sem.c: convert sem_array.sem_pending to struct list_head

[Manfred Spraul <manfred@colorfullife.com>]

Nadia Derbey wrote:
> Manfred Spraul wrote:
>> sem_array.sem_pending is a double linked list, the attached
>> patch converts it to struct list_head.
>>
>> Signed-Off-By: Manfred Spraul <manfred@colorfullife.com>
>
> Reviewed-By: Nadia Derbey <Nadia.Derbey@bull.net>
>
>>
>> @@ -438,16 +405,15 @@ static void update_queue (struct sem_array * sma)
>>      int error;
>>      struct sem_queue * q;
>>  
>> -    q = sma->sem_pending;
>> -    while(q) {
>> +    q = list_entry(sma->sem_pending.next, struct sem_queue, list);
>> +    while(&q->list != &sma->sem_pending) {
>
> I guess here you are not using list_first_entry() because the pending 
> requests might be empty?
>
Actually - no. I wasn't aware of list_first_entry().
But:
- Looking at list.h: list_first_entry() shouldn't be used on empty lists.
- the loop in update_queue() is complicated enough, some open-coding 
doesn't hurt.

>> @@ -1194,7 +1171,6 @@ asmlinkage long sys_semtimedop(int semid, 
>> struct sembuf __user *tsops,
>>  
>>      sma = sem_lock(ns, semid);
>>      if (IS_ERR(sma)) {
>> -        BUG_ON(queue.prev != NULL);
>
> Instead of removing it why not replacing the bug_ON() by a check on 
> the queue still being linked?
>
The list_del() poisoning is IMHO efficient enough.

--
    Manfred