From: Burkhard K. <bu...@bu...> - 2000-09-25 18:55:05
|
damcha wrote: > Here is the patch which includes both Burkhard's and mine. > It now compile on 2.4.0-test9. Please report comments in the ML. > I am afraid, but I have to correct my own patch. Although it now runs, I made an ugly typo and did not get the while-loop right in in mlc_recvmsg(). Damcha, would you incorporate my corrections into your patch? Is it o.k. for you to maintain the kernel module part of the project? I will send you my corrections with a separate mail. I am still concerned about the msghdr flag issue, which I don't fully understand. With the 2.4 kernel we see an MSG_EOR flag set. Does that mean we have to hand this flag further on to the next layer or does it mean we are expected to perform some additional action. As I noticed from Alan Cox comments in socket.[ch] not all flags which are defined in IEEE1003.1 are implemented, but later kernel versions might change that. Some flags obviously correspond to some service or type of service. If these flags are set and we have not implemented this type of service, what is the correct behaviour.? Just return with -EOPNOTSUPP? Or might it be appropriate to silently ignore some flags? O I have to admit, that I don't know much about sockets and their represation within the kernel. Furthermore I am wondering about the correct handling of the DONTWAIT flag. nonblock is set to 0 unconditionally, but obviously we should accept this flag. On the other hand what sense does it make to accept a flag and ignore it? On the other hand I have never seen the DONTWAIT flag being set, does it mean, we can savely reject it? Burkhard -- Burkhard Kohl Tel/FAX: +49 30 698 15 905 Melchiorstr. 8 bu...@bu... 10179 Berlin |