On Tue, Feb 13, 2001 at 01:10:02PM -0700, K. Leif D. Van Horn wrote:
> On Tue, Feb 13, 2001 at 01:01:00PM -0500, Jan Harkes wrote:
> > On Tue, Feb 13, 2001 at 10:23:53AM -0700, K. Leif D. Van Horn wrote:
> > > This concerns Linux 2.4.1 coda and podfuk/uservfs (which uses /dev/cfs0)
> > > to "mount" Zope ftp server contents.  Everything seems to work with
> > > the filesystem as far as lookups, reads, and writes.  The close system
> > > call seems to not propagate error state to my applications.
> > > 
> > > The way I see it, I should be able to write a message to the
> > > coda device (/dev/cfs0) at the end of the upcall.  This message
> > > contains the outputArgs union with result set to -1.  The application
> > > never sees close return a error status.  Here is a example
> > > perl script:
> > 
> > The Coda kernel module expects userspace to return positive error
> > numbers:
> > 
> > linux/fs/coda/upcall.c:740
> >     /* Op went through, interrupt or not... */
> >     if (req->uc_flags & REQ_WRITE) {
> > 	    out = (union outputArgs *)req->uc_data;
> > 	    /* here we map positive Venus errors to * kernel errors */
> > 	    error = -out->oh.result;
> > 	    CDEBUG(D_UPCALL, "upcall: (u,o,r) (%ld, %ld, %ld) out at %p\n", 
> > 		   out->oh.unique, out->oh.opcode, out->oh.result, out);
> > 	    *outSize = req->uc_outSize;
> > 	    goto exit;
> >     }
> > 
> > The error is expected to be something like EINVAL. When we get back into the
> > VFS with such a negative error, errno is set to this value and close will
> > return -1.
> > 
> > So you would have to return something like EPARSE instead of -EPARSE.
> > 
> > Jan
> > 
> 
> Whoops.  I was not looking closely enough at my code.  The value
> written in the union is EBADF which is not quite what I wanted
> but is still a error.  At least is something that my close()
> man page refers to.
> 
> Thanks,
> -Leif

It seems that the kernel drops the return value of release()/close().
Is there any possibility of coda implementing a flush() method for
the file_operations struct?

Thanks,
-Leif