> On Wed, 2 Jun 1999 13:17:37 GMT, David S. Miller <davem@redhat.com> wrote:
> > Date: Sat, 22 May 1999 08:17:24 -0400 (EDT)
> > From: HighWind Software Information <info@highwind.com>
> >
> > In any case, ::writev() does sometimes perform a partial write
> > on a blocking socket, that is a BUG.
> >
> >No, it is by no means a bug, it is perfectly fine behavior, and your
> >application must check for and deal with this case.
> >
> >It is not getting fixed, because there is nothing to fix.
>
> Uhm, are you sure? I have never seen that behaviour on any
> other Unix. If a socket is in blocking mode, writev (or send)
> is supposed to sleep until all is send or an error occurs. I
> would believe a lot software would break if write returned partial
> results on blocking software.
[SNIPPED]
With send(), on a BLOCKING socket, the socket should block until the
data can be sent. If the length of the data is too great to send
"atomically", send returns -1 with errno set to EMSGSIZE. This
seems to mean that no partial messages are possible with a blocking
send().
If Linux allows partial transmissions, it it not BSD socket behavior.
However, it seems to be a good idea (if it didn't break existing
code).
The original report was about writev(), which should write multiple
buffers. Implicit is that it will take a big buffer and chop it up
to lengths that can be handled by send(). writev() returns the number
of bytes written or -1 with errno set. I can't find any information
anywhere if it is supposed to return less than the requested bytes.
However, when writting to a nearly full disk, where there are only
N bytes available, the first write returns N which is less than the
requested amount, subsequent writes return 0 which doesn't seem to be
correct. I would expect it to return -1 with errno set to something
that would allow a program to know that the disk is full (like ENOSPC).
This is probably a libc issue, though.
Cheers,
Dick Johnson
***** FILE SYSTEM WAS MODIFIED *****
Penguin : Linux version 2.2.6 on an i686 machine (400.59 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/