Re: rtwn0: rtwn_rx_copy_to_mbuf: could not allocate RX mbuf

Oh, and would you mind filing a bug at bugs.freebsd.org ? I'd like to track
this as a normal honest to god bug.



-adrian


On Sat, 26 Apr 2025 at 08:33, Nuno Teixeira <eduardo@freebsd.org> wrote:

> (...)
>
> Found stability and no warnings/errors to report on usb nics:
>
> RTL8821AU 11ac @aarch64
> RTL8811AU 11ac @amd64
>
> Cheers!
>
> Nuno Teixeira <eduardo@freebsd.org> escreveu (sábado, 26/04/2025 à(s)
> 08:58):
> >
> > Hello!
> >
> > > Ok, that is interesting. We're not keeping counters in m_get2() for
> when the buffer requested is too large; it only returns NULL.
> > > I guess it's bailing if it's above a page size (4kbytes.)
> > >
> > > Try changing the m_get2() call there to m_get3(); let's see if that
> improves things.
> > > (m_get3() will fail above 16kbytes; and if it can't allocate a buffer
> < 16kbytes then we should get some UMA error counted somehwere.)
> >
> > iperf3 -R is happy, and git pull --rebase completed OK and fast:
> >
> > Connecting to host hp, port 5201
> > Reverse mode, remote host hp is sending
> > [  5] local 192.168.1.82 port 39601 connected to 192.168.1.100 port 5201
> > [ ID] Interval           Transfer     Bitrate
> > [  5]   0.00-1.06   sec  6.00 MBytes  47.4 Mbits/sec
> > [  5]   1.06-2.06   sec  15.9 MBytes   134 Mbits/sec
> > [  5]   2.06-3.06   sec  16.1 MBytes   135 Mbits/sec
> > [  5]   3.06-4.06   sec  16.4 MBytes   137 Mbits/sec
> > [  5]   4.06-5.00   sec  15.0 MBytes   134 Mbits/sec
> > [  5]   5.00-6.06   sec  16.4 MBytes   129 Mbits/sec
> > [  5]   6.06-7.06   sec  15.2 MBytes   128 Mbits/sec
> > [  5]   7.06-8.06   sec  14.8 MBytes   124 Mbits/sec
> > [  5]   8.06-9.06   sec  15.8 MBytes   132 Mbits/sec
> > [  5]   9.06-10.06  sec  16.1 MBytes   135 Mbits/sec
> > - - - - - - - - - - - - - - - - - - - - - - - - -
> > [ ID] Interval           Transfer     Bitrate         Retr
> > [  5]   0.00-10.08  sec   148 MBytes   123 Mbits/sec    0
> sender
> > [  5]   0.00-10.06  sec   148 MBytes   123 Mbits/sec
> receiver
> >
> > iperf Done.
> >
> > I will use this dongle on amd64 for further testing, I will patch
> aarch64 soon.
> >
> > ```
> > diff --git a/sys/dev/rtwn/usb/rtwn_usb_rx.c
> b/sys/dev/rtwn/usb/rtwn_usb_rx.c
> > index 657d6bdeb9e4..54869fe336bd 100644
> > --- a/sys/dev/rtwn/usb/rtwn_usb_rx.c
> > +++ b/sys/dev/rtwn/usb/rtwn_usb_rx.c
> > @@ -124,10 +124,10 @@ rtwn_rx_copy_to_mbuf(struct rtwn_softc *sc,
> > struct rtwn_rx_stat_common *stat,
> >         if (rtwn_rx_check_pre_alloc(sc, stat) != 0)
> >                 goto fail;
> >
> > -       m = m_get2(totlen, M_NOWAIT, MT_DATA, M_PKTHDR);
> > +       m = m_get3(totlen, M_NOWAIT, MT_DATA, M_PKTHDR);
> >         if (__predict_false(m == NULL)) {
> > -               device_printf(sc->sc_dev, "%s: could not allocate RX
> mbuf\n",
> > -                   __func__);
> > +               device_printf(sc->sc_dev, "%s: could not allocate RX
> > mbuf (%d bytes)\n",
> > +                   __func__, totlen);
> >                 goto fail;
> >         }
> > ```
> >
> > Thanks!
> >
> >
> > --
> > Nuno Teixeira
> > FreeBSD UNIX:  <eduardo@FreeBSD.org>   Web:  https://FreeBSD.org
>
>
>
> --
> Nuno Teixeira
> FreeBSD UNIX:  <eduardo@FreeBSD.org>   Web:  https://FreeBSD.org
>