| Message ID | 46079a126ad542d380add5f9ba6ffa85@AcuMS.aculab.com |
|---|---|
| State | Changes Requested |
| Delegated to: | David Miller |
| Headers | show |
| Series | net: sctp: Fix negotiation of the number of data streams. | expand |
From: David Laight <David.Laight@ACULAB.COM> Date: Tue, 18 Aug 2020 14:36:58 +0000 > Fixes 2075e50caf5ea. > > Signed-off-by: David Laight <david.laight@aculab.com> This is not the correct format for a Fixes tag, also it should not be separated by empty lines from other tags such as Signed-off-by and Acked-by. As someone who reads netdev frequently, these patterns should be etched into your brain by now. :-)
From: David Miller > Sent: 18 August 2020 20:58 > > From: David Laight <David.Laight@ACULAB.COM> > Date: Tue, 18 Aug 2020 14:36:58 +0000 > > > Fixes 2075e50caf5ea. > > > > Signed-off-by: David Laight <david.laight@aculab.com> > > This is not the correct format for a Fixes tag, also it should not > be separated by empty lines from other tags such as Signed-off-by > and Acked-by. > > As someone who reads netdev frequently, these patterns should be > etched into your brain by now. :-) Sorry I don't often write patches - I'm only paid to support our drivers, not fix the Linux kernel :-) (We don't ship a kernel, our drivers have to work in disto kernels.) David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
On Tue, Aug 18, 2020 at 02:36:58PM +0000, David Laight wrote: > The number of streams offered by the remote system was being ignored. > Any data sent on those streams would get discarded by the remote system. That's quite brief and not accurate: it was only ignored if 'Xcnt <= stream->Xcnt'. Other than this and the Fixes tag, LGTM. Passes the tests here. I'll ack the v2 then. > > Fixes 2075e50caf5ea. > > Signed-off-by: David Laight <david.laight@aculab.com> > --- > net/sctp/stream.c | 6 ++++-- > 1 file changed, 4 insertions(+), 2 deletions(-) > > This needs backporting to 5.1 and all later kernels. A 'net' tree tag in patch tags is welcomed. > > diff --git a/net/sctp/stream.c b/net/sctp/stream.c > index bda2536dd740..6dc95dcc0ff4 100644 > --- a/net/sctp/stream.c > +++ b/net/sctp/stream.c > @@ -88,12 +88,13 @@ static int sctp_stream_alloc_out(struct sctp_stream *stream, __u16 outcnt, > int ret; > > if (outcnt <= stream->outcnt) > - return 0; > + goto out; > > ret = genradix_prealloc(&stream->out, outcnt, gfp); > if (ret) > return ret; > > +out: > stream->outcnt = outcnt; > return 0; > } > @@ -104,12 +105,13 @@ static int sctp_stream_alloc_in(struct sctp_stream *stream, __u16 incnt, > int ret; > > if (incnt <= stream->incnt) > - return 0; > + goto out; > > ret = genradix_prealloc(&stream->in, incnt, gfp); > if (ret) > return ret; > > +out: > stream->incnt = incnt; > return 0; > } > -- > 2.25.1 > > - > Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK > Registration No: 1397386 (Wales) >
From: 'Marcelo Ricardo Leitner' > Sent: 18 August 2020 22:47 > > On Tue, Aug 18, 2020 at 02:36:58PM +0000, David Laight wrote: > > The number of streams offered by the remote system was being ignored. > > Any data sent on those streams would get discarded by the remote system. > > That's quite brief and not accurate: it was only ignored if 'Xcnt <= > stream->Xcnt'. The number of streams (esp out ones) received from the remote system in an INIT or INIT_ACK was ignored. So it would always send data chunks using the number of streams requested by the local user. I managed to tweak our M3UA config to get invalid stream numbers sent on both inwards and outwards connections. I only noticed because of testing a (slightly horrid) workaround for no longer being able to use kernel_getsockopt() to retrieve the number of ostreams. The number of ostreams was about the only thing we didn't trace :-( At least my code can now obtain the correct value even for buggy kernels. > Other than this and the Fixes tag, LGTM. Passes the tests here. I'll > ack the v2 then. I wasn't sure whether DM actually wanted a V2 with the fixes tag fixed. I can send one tomorrow. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
diff --git a/net/sctp/stream.c b/net/sctp/stream.c index bda2536dd740..6dc95dcc0ff4 100644 --- a/net/sctp/stream.c +++ b/net/sctp/stream.c @@ -88,12 +88,13 @@ static int sctp_stream_alloc_out(struct sctp_stream *stream, __u16 outcnt, int ret; if (outcnt <= stream->outcnt) - return 0; + goto out; ret = genradix_prealloc(&stream->out, outcnt, gfp); if (ret) return ret; +out: stream->outcnt = outcnt; return 0; } @@ -104,12 +105,13 @@ static int sctp_stream_alloc_in(struct sctp_stream *stream, __u16 incnt, int ret; if (incnt <= stream->incnt) - return 0; + goto out; ret = genradix_prealloc(&stream->in, incnt, gfp); if (ret) return ret; +out: stream->incnt = incnt; return 0; }
The number of streams offered by the remote system was being ignored. Any data sent on those streams would get discarded by the remote system. Fixes 2075e50caf5ea. Signed-off-by: David Laight <david.laight@aculab.com> --- net/sctp/stream.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) This needs backporting to 5.1 and all later kernels.