Date: Wed, 26 Nov 2008 13:46:16 +0300 From: Vladimir Grebenschikov <vova@fbsd.ru> To: Nick Hibma <nick@van-laarhoven.org> Cc: freebsd-current@freebsd.org Subject: Re: u3g and ubsa Message-ID: <1227696376.1755.1.camel@localhost> In-Reply-To: <200811261029.05178.nick@van-laarhoven.org> References: <492A62EF.7000803@bsdforen.de> <200811251545.10807.nick@van-laarhoven.org> <1227685292.1852.12.camel@localhost> <200811261029.05178.nick@van-laarhoven.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 2008-11-26 at 10:29 +0100, Nick Hibma wrote: > Could you try the following: > > Change the ibufsize and obufsize values to a power of 2 in u3g_attach: > > ucom->sc_ibufsize = 4096; > ucom->sc_ibufsizepad = ucom->sc_ibufsize; > ucom->sc_obufsize = 4096; > > Perhaps the USB stack does not like arbitrarily sized buffers due to > transfer rollovers at the end of the buffer / memory page or something. > > If that does not work could you lower it to 512, and if that does not work > change it to: > > ucom->sc_ibufsize = UGETW(ed->wMaxPacketSize); > ucom->sc_ibufsizepad = ucom->sc_ibufsize; > ucom->sc_obufsize = UGETW(ed->wMaxPacketSize); > > That last case is worst case as it would considerably increase overhead, > especially in USB 1.0 cases where the max. wMaxPacketSize is 64 bytes. All cases makes same panic on connecting to port: ohci.c:545 ohci_alloc_std_chain: XXX need to copy > If it works I'd be interested to hear about performance (download from SUN > is always a good one), and any messages in dmesg or in a ppp logfile for > dropped chars, and invalid FCS. > > Cheers, > > Nick -- Vladimir B. Grebenschikov vova@fbsd.ru
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1227696376.1755.1.camel>