Return-path: Received: from fg-out-1718.google.com ([72.14.220.159]:31905 "EHLO fg-out-1718.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751718Ab0DKNfb convert rfc822-to-8bit (ORCPT ); Sun, 11 Apr 2010 09:35:31 -0400 Received: by fg-out-1718.google.com with SMTP id 19so893294fgg.1 for ; Sun, 11 Apr 2010 06:35:30 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: References: <4BB7660A.3030900@lwfinger.net> <4BB773A9.2070703@lwfinger.net> From: =?ISO-8859-1?Q?G=E1bor_Stefanik?= Date: Sun, 11 Apr 2010 15:35:10 +0200 Message-ID: Subject: Re: RTL 8187b - naughty behaviour To: Rogerio Luz Coelho Cc: Larry Finger , linux-wireless@vger.kernel.org, Herton Ronaldo Krzesinski Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: 2010/4/10 Rogerio Luz Coelho : > 2010/4/10 Gábor Stefanik : >> On Sat, Apr 10, 2010 at 2:07 PM, Rogerio Luz Coelho >> wrote: >>> 2010/4/3 Rogerio Luz Coelho : >>>> 2010/4/3 Larry Finger : >>>>> On 04/03/2010 11:49 AM, Rogerio Luz Coelho wrote: >>>>>> >>>>>> I have a MS-Win7 instalation on this Laptop and it works 100% even at >>>>>> longer distances than you report,even with walls in between. >>>>>> >>>>>> Will do the tcpperf test as soon as I get home. There are other with >>>>>> the same problem with my exact chip here in Brasil , that is because >>>>>> one of our leading Laptop vendors uses this config. >>>>> >>>>> What is the exact make and model of the device? It might be difficult >>>>> for me to get one in the US, but one of the rtl8187 developers lives in >>>>> Brazil. He should be able to find one. >>>>> >>>> >>>> It愀 a POSITIVO R430L - Core 2 Duo 4GB RAM >>>> >>>> >>>>>> Does it matter that the USB port used in this setting is a connected >>>>>> to a USB HUB and then to the mother board? >>>>> >>>>> Is the hub passive, or is it powered? If the hub is good and not >>>>> underpowered, it should be OK. >>>>> >>>>>> More details and tests will come your way in a day or two ... will >>>>>> reinstall Mandriva 2010 and do some tests with older kernels. >>>>> >>>>> If you can download an openSUSE 11.2 Live CD and burn it, you would not >>>>> even have to do an installation. That one has a 2.6.31 kernel. >>>>> >>>>> Larry >>> >>> >>> Downloaded and installed: >>> >>> OpenSuse 11.2 - LiveCD  result: Gnome Network-Manager tells me there >>> is connection but no ping resposes, Firefox complains about conections >>> errors. >>> >>> Mandriva 2010.0 (kernel 2.6.31-12) result: concection comes and goes >>> ... generally the first boot of the day connects, but it lasts less >>> than 5min. The error code is the "deauthentication by local reason >>> (reason=3) again. >>> >>> MS-Win7 : connects and is as fast as ever. >>> >>> -- Any other suggestions? >>> >>> Rogerio >>> >>> PS: Will give OpenSuse a HD install atempt and will try to debug the >>> Mandriva for some time, future tries involve Debian Stable (old >>> kernel) and some Fedora (any suggestions?) and the NDISwrapper drive >>> in a Debian Testing (the Distro I need working) >> >> Hi, >> >> Please also test Realtek's official r8187b driver. If possible, please >> do an mmiotrace on that driver. > > Ok ... The official Realtek is only for 32bits systems right ? ( At > least the guys at Realtek support told me so... ) and this mmiotrace > can be accomplished how? Is it a program, a command, a debug message > in some form? Sorry, as I said earlier I am NOT a kernel programer. > > Rogerio > You need to enable MMIO tracing support in your kernel config, and recompile your kernel (assuming it is not enabled already - distros usually disable the entire tracing framework by default). Then: -Mount debugfs to /debug. (Usually it is mounted by default, but some distros like to mount it @ /sys/kernel/debug instead - remount it to /debug.) -"echo mmiotrace > /debug/tracing/current_tracer" -"cat /debug/tracing/trace_pipe > trace.txt &" - at this point, tracing begins. -Load r8187b and connect. -"echo "Driver loaded!" > /debug/tracing/trace_marker" -"echo nop > /debug/tracing/current_tracer" - tracing ends here. Do this for rtl8187 as well. After each trace, check if "grep -i lost trace.txt" returns anything. If it does, enlarge the trace buffer and retry: -"cat /debug/tracing/trace_buffer_kb" -"echo [BUFSIZE] > /debug/tracing/trace_buffer_kb", where [BUFSIZE] is recommended to be about twice the number returned by the previous command. Repeat these steps until the "lost events" warning disappears. Once you find the right BUFSIZE for one driver, you can use the same size for the other one too. Hth, Gábor -- Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)