Write failures and proper error code return to clients

[Jeremy Allison]

On Sat, Jul 22, 2006 at 06:53:30AM +0300, Shlomi Yaakobovich wrote:
> Hi,
>  
> We've recently ran into situations when the pwrite system call fails. I've noticed that on -1 return code, a "disk full" error is returned to the user. For example, in reply_write_and_X:
>  
>  if (nwritten < (ssize_t)numtowrite) {
> SCVAL(outbuf,smb_rcls,ERRHRD);
> SSVAL(outbuf,smb_err,ERRdiskfull);      
> }
> 
> I saw no attempt to look at the exact errno, and attempt to return a more valid error code.

Ok, I've now had time to look at this.
You missed this code in reply_write_and_X() :

        if(((nwritten == 0) && (numtowrite != 0))||(nwritten < 0)) {
                END_PROFILE(SMBwriteX);
                return(UNIXERROR(ERRHRD,ERRdiskfull));
        }

UNIXERROR looks at errno. What is errno set to in your
particular case ? The default of ERRdiskfull is only
returned if errno == 0.

The code you complain about is a fallback case just in case
there are some systems where write returns less than was
requested on ENOSPC. It should never be true on a working POSIX
system.

> Furthermore, perhaps it is worth while thinking of retrying the call (e.g. if errno is EINTR) instead of returning an error to the client ?  We've noted that some applications don't behave well to "disk full" error, and abort their operations.

As noted before, this can't happen on disk writes except
for soft mounts.

Jeremy.