Hi Marcelo > On Oct 11, 2021, at 22:15, Marcelo Ricardo Leitner <marcelo.leitner@xxxxxxxxx> wrote: > > ... > > So if stream_num was originally 1, stream_len would be 2, and with > padding, 4. Here, nums would be 2 then, and not 1. The padding gets > accounted as if it was payload. > > IOW, the patch is making the padding part of the parameter data by > adding it to the header as well. SCTP padding works by having it in > between them, and not inside them. > > This other approach avoids this issue by adding the padding only when > allocating the packet. It (ab)uses the fact that inreq and outreq are > already aligned to 4 bytes. Eiichi, can you please give it a go? > > Thanks, I understood. I’ve tested your diff with my reproducer and it certainly works. Your diff looks good to me. > > ---8<--- > > diff --git a/net/sctp/sm_make_chunk.c b/net/sctp/sm_make_chunk.c > index b8fa8f1a7277..c7503fd64915 100644 > --- a/net/sctp/sm_make_chunk.c > +++ b/net/sctp/sm_make_chunk.c > @@ -3697,7 +3697,7 @@ struct sctp_chunk *sctp_make_strreset_req( > outlen = (sizeof(outreq) + stream_len) * out; > inlen = (sizeof(inreq) + stream_len) * in; > > - retval = sctp_make_reconf(asoc, outlen + inlen); > + retval = sctp_make_reconf(asoc, SCTP_PAD4(outlen) + SCTP_PAD4(inlen)); > if (!retval) > return NULL; Regards, Eiichi
