On Thu, Aug 17, 2017 at 01:21:57PM +0000, David Laight wrote: > > > How do we stop our customers insisting on using SCTP in RHEL 5.7 :-) > > > > > Release a kernel update with CONFIG_SCTP=n? > > Go for it :-) > > > Sorry, wish there was a better way, but people keep paying for it, so we keep > > providing it. > > Yes, at least we've managed to lose all the people who wanted 2.4.x support. > > In this case I actually suspect the problem is that our code now puts port==0 > in bindx calls - which is only allowed after the fixes that went into 2.6.22 (ish). > But our support group are trying to get them to upgrade our software to > the 'supported' version - which won't make a blind bit of difference. > We're also trying to get them to upgrade to something later than RHEL 5.7 > but that is probably even harder! > > I wish I had a list of the very bad SCTP bugs and the kernel versions > that fix them. > > We only use it for M3UA - so there are very few connections that are > established very infrequently. > I suspect the error paths don't get stressed. Seems the new features (instead of the bugs) won't help then either, would them? Like partial reliability or stream reconf. Also I'm about to post patches for ndata stream schedulers. Just need to work around an issue with memory allocation (extended struct sctp_stream_out is too big now) Though the schedulers are probably only useful if your socket gets tx buffered. > > I don't want to start testing SCTP - I might have to fix the bugs in > the copy we 'stole' from FreeBSD and hacked into the windows kernel. 8-) Marcelo -- To unsubscribe from this list: send the line "unsubscribe linux-sctp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
