Re: [fuse-devel] does FUSE support concurrent write ?

[Xiangyong (S. O. <ou...@cs...>]

Hi Goswin,

Thanks for your reply!  It turned out that I should use pthread_self()
to show thread id, instead of getpid():  getpid() always returns same
value for threads created in a same process.  Thanks!


--Shangyong

On Wed, Apr 18, 2012 at 1:22 AM, Goswin von Brederlow <gos...@we...> wrote:
> "Xiangyong (Shangyong) Ouyang" <ou...@cs...> writes:
>
>> Hi all,
>>
>> I let 8 processes perform concurrent pwrite() to an example FUSE-based
>> filesystem (http://www.cs.nmsu.edu/~pfeiffer/fuse-tutorial/).  In
>> fuse_write() , I do a getpid() to show the current process id
>> that conducts the write.  I always see the SAME pid, regardless of how
>> many user processes are writing concurrently.
>>
>> I'm under the impression that fuse library internally has multiple
>> worker-threads that fetch requests from /dev/fuse.  Therefore I'm
>> expecting to see different pids in fuse_write().  Am I missing some
>> flags to enable fuse internal multi-threads? I didn't pass any special
>> command line arguments to fuse except for "-f".
>>
>> btw,  I'm running 8-core RHEL machine,  fuse 2.8.1,  kernel 2.6.30.10.  Thanks!
>>
>
> Fuse supports both a single threaded and multi threaded loop. Maybe your
> filesystem is calling the single threaded one or adds the single
> threaded option before calling libfuse.
>
> MfG
>        Goswin