[SSI-devel] Re: [ ssic-linux-Bugs-986537 ] semundo doesn't happen upon nodedown

[Andreas <ro...@co...>]

Keerthi Bhushan K N wrote:
> Hi Andreas,
> 
> Andreas wrote:
> 
>>> Hello
>>>
>>> I am using the new sources out of the CVS an now there is a problem 
>>> with the semaphores. It is possible to create a semaphore.
>>>
>>> semget (4000,1,IPC_CREAT | IPC_EXCL | U_R | U_W)
>>>
>>> returns a valid id. The second call tells that it allready exists, 
>>> what is right But every call of
>>>
>>> semget (4000,0,0)
>>>
>>> from another node as the one where the semaphore is created returns 
>>> EINVAL (22).
>>
>>
>>
>> In line 271 of sem.c is tested, if nsems is 0. But I think that should 
>> be changed?
>>
>> The test should be
>>
>>
>> if (!nsems && semflg) {
>>         retval = -EINVAL;
>>         goto out_up;
>> }
>>
>>
>> Isn't that correct? If you do not give any flags and set the number of 
>> the semaphores allthough to zero you intend to just open the 
>> semaphore. In nearly every samplecode semaphores are opened with 
>> nsems=0 and no flags. And as it works if the semaphore node is the 
>> lokal node I think it should be changed.
> 
> 
> 
> This is a bug which I have fixed. Please try out the latest sources.
> Basically we do the (correct) check in sys_msgget() before calling 
> ssi_msgget(). So the piece of code u have mentioned above (sem.c:271) is 
> actually redundant.

Thank you for you answere. I am using the code out of the stable branch 
now and it works fine


Andreas

> 
> Keerthi
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by BEA Weblogic Workshop
> FREE Java Enterprise J2EE developer tools!
> Get your free copy of BEA WebLogic Workshop 8.1 today.
> http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click