netdev - RE: sctp: num_ostreams and max

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <868bd24b536345e6a5596f856a0ebe90@AcuMS.aculab.com>
Date:   Fri, 14 Aug 2020 16:18:05 +0000
From:   David Laight <David.Laight@...LAB.COM>
To:     "'linux-sctp@...r.kernel.org'" <linux-sctp@...r.kernel.org>,
        'Neil Horman' <nhorman@...driver.com>,
        "'kent.overstreet@...il.com'" <kent.overstreet@...il.com>,
        Andrew Morton <akpm@...ux-foundation.org>
CC:     "'netdev@...r.kernel.org'" <netdev@...r.kernel.org>
Subject: RE: sctp: num_ostreams and max_instreams negotiation

> > > At some point the negotiation of the number of SCTP streams
> > > seems to have got broken.
> > > I've definitely tested it in the past (probably 10 years ago!)
> > > but on a 5.8.0 kernel getsockopt(SCTP_INFO) seems to be
> > > returning the 'num_ostreams' set by setsockopt(SCTP_INIT)
> > > rather than the smaller of that value and that configured
> > > at the other end of the connection.
> > >
> > > I'll do a bit of digging.
> >
> > I can't find the code that processes the init_ack.
> > But when sctp_procss_int() saves the smaller value
> > in asoc->c.sinint_max_ostreams.
> >
> > But afe899962ee079 (if I've typed it right) changed
> > the values SCTP_INFO reported.
> > Apparantly adding 'sctp reconfig' had changed things.
> >
> > So I suspect this has all been broken for over 3 years.
> 
> It looks like the changes that broke it went into 4.11.
> I've just checked a 3.8 kernel and that negotiates the
> values down in both directions.
> 
> I don't have any kernels lurking between 3.8 and 4.15.
> (Yes, I could build one, but it doesn't really help.)

Ok, bug located - pretty obvious really.
net/sctp/stream. has the following code:

static int sctp_stream_alloc_out(struct sctp_stream *stream, __u16 outcnt,
				 gfp_t gfp)
{
	int ret;

	if (outcnt <= stream->outcnt)
		return 0;

	ret = genradix_prealloc(&stream->out, outcnt, gfp);
	if (ret)
		return ret;

	stream->outcnt = outcnt;
	return 0;
}

sctp_stream_alloc_in() is the same.

This is called to reduce the number of streams.
But in that case it does nothing at all.

Which means that the 'convert to genradix' change broke it.
Tag 2075e50caf5ea.

I don't know what 'genradix' arrays or the earlier 'flex_array'
actually look like.
But if 'genradix' is some kind of radix-tree it is probably the
wrong beast for SCTP streams.
Lots of code loops through all of them.

This does mean that it has only been broken since the 5.1
merge window.

While just assigning to stream->outcnt when the value
is reduced will fix the negotiation, I've no idea
what side-effects that has.

	David


	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)