Date: Sun, 19 May 2013 10:05:13 -0700 From: Adrian Chadd <adrian@freebsd.org> To: lev@freebsd.org Cc: freebsd-wireless@freebsd.org Subject: Re: [rft] please test -HEAD ath; lots of TX changes Message-ID: <CAJ-Vmon3UXXEf-fJ6Cc29qRF1U31UJegP-fR3NQ5_zGvv1tRsg@mail.gmail.com> In-Reply-To: <1377052407.20130519195416@serebryakov.spb.ru> References: <CAJ-VmongHgfCBPFpnMLi_5ppVbz%2BtOXVSOzRmCF=yL8FtnMBHQ@mail.gmail.com> <372806514.20130519141024@serebryakov.spb.ru> <CAJ-Vmo=Ak=Af7mdc2LP0niSrGbHsVAFOAen16wCLaEXrj2A82A@mail.gmail.com> <1106213329.20130519193856@serebryakov.spb.ru> <CAJ-VmonobPnfd4u9QnA%2BaCWDWJZYxWFV_N4PwusLyZxiK%2BuHoA@mail.gmail.com> <1377052407.20130519195416@serebryakov.spb.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
No, the driver drops frames only on error and otherwise it sends ENOBUFS up to the nrt layer. This stalls the sender. Check athratestats in tools/ath. See what the stats are during transmit. Adrian Adrian On May 19, 2013 11:54 AM, "Lev Serebryakov" <lev@freebsd.org> wrote: > Hello, Adrian. > You wrote 19 =D0=BC=D0=B0=D1=8F 2013 =D0=B3., 19:49:48: > > AC> Ok. So the hardware queue isnt hung. Good! > > AC> The 30mbit is the transmit rate, not throughput. No idea why is isnt > AC> downgrading though. > 300! It doesn't downgrading, because it is UDP and it is FreeBSD -- > Linux blocks sendto() on UDP socket when buffers/queue is full, and > FreeBSD simply discard data and returns. FreeBSD behaves more > correctly from POSIX point of view, but Linux is more "expectable". > > AC> So lets do more testing to aee if the transmit queue stalls. Also, we > can > AC> diagnose the disassociate at some point. Then after that, rate contro= l. > Logs sent to you should show, that client deassociate in middle of > process, and it was unexpected :) > > > -- > // Black Lion AKA Lev Serebryakov <lev@FreeBSD.org> > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-Vmon3UXXEf-fJ6Cc29qRF1U31UJegP-fR3NQ5_zGvv1tRsg>