Wireless problems on 8.0-RELEASE

[Paul B Mahol]

On Tue, Jan 19, 2010 at 12:54 PM, Alex <alex323 at gmail.com> wrote:
> On Tue, 19 Jan 2010 12:17:49 +0000
> Paul B Mahol <onemda at gmail.com> wrote:
>
>> On 1/19/10, Alex <alex323 at gmail.com> wrote:
>> > On Mon, 18 Jan 2010 21:34:27 +0000
>> > Paul B Mahol <onemda at gmail.com> wrote:
>> >
>> >> On 1/18/10, Alex <alex323 at gmail.com> wrote:
>> >> > On Mon, 18 Jan 2010 16:23:56 +0000
>> >> > Paul B Mahol <onemda at gmail.com> wrote:
>> >> >
>> >> >> On 1/18/10, Alex <alex323 at gmail.com> wrote:
>> >> >> > The module is loaded at boot time in loader.conf, but I think
>> >> >> > this is the message you were looking for:
>> >> >> >
>> >> >> > ntoskrnl dummy called...
>> >> >>
>> >> >> I was looking for:
>> >> >>
>> >> >> no match for IoWMIQueryAllData
>> >> >> no match for IoWMIOpenBlock
>> >> >>
>> >> >> Does message appears only when uploading
>> >> >> or just always, after boot is completed
>> >> >> but before interface is turned up (via ifconfig)
>> >> >>
>> >> >
>> >> > I've never seen those messages in the dmesg output. Strangely,
>> >> > the only method by which I can successfully upload large files
>> >> > is scp.
>> >>
>> >> Here is my test results (another driver/card combination):
>> >> With opera and gmail, upload "stuck" at ~32kb but rtorrent and sftp
>> >> uploads just fine. Upload of 1.3M file in opera stuck at few kb but
>> >> it was complete after
>> >> few seconds.
>> >>
>> >> >> You said that 7.2 doesnt have this problem,
>> >> >> are you using same SYS and INF files with 8.0 ?
>> >> >>
>> >> >
>> >> > No, because Lenovo released updates for the driver. Both the
>> >> > older version of the driver (the one I used on 7.2) and the
>> >> > newest version of the driver produce the same problems on 8.0.
>> >> > If you want, I can upload the older version of the driver too.
>> >>
>> >> Try to remove line with OID_GEN_CURRENT_LOOKAHEAD
>> >> in /sys/dev/if_ndis.c, rebuild and reinstall if_ndis module and see
>> >> if that helps.
>> >>
>> >
>> > /sys/dev/if_ndis.c does not exist,
>> > however /sys/dev/if_ndis/if_ndis.c does exist. I have commented out
>> > the line and executed:
>> >
>> > make buildkernel
>> > make installkernel
>> >
>> > but it did not seem to fix the problem (I don't use modules -- I
>> > only compile in what I need).
>> >
>> >> Why are you have tso disabled, does it works with tso enabled?
>> >> Are there any other custom sysctl options?
>> >>
>> >
>> > If I recall correctly, someone on IRC recommended I try disabling
>> > that. Changing this setting had no effect -- the problem existed
>> > either way. It is enabled now. No other sysctl options are in
>> > effect.
>> >
>> >> What is signal strength of the AP? And current link speed? (this
>> >> one is available only on CURRENT and from my git repo code)
>> >
>> > SSID/MESH ID    BSSID              CHAN RATE   S:N     INT CAPS
>> > skynet          00:28:10:00:28:10    6   54M -71:-96  100 EP
>> >
>> > media: IEEE 802.11 Wireless Ethernet OFDM/36Mbps mode 11g
>> That one is just after association, code from CURRENT have fixed this.
>>
>> Does lowering mtu of wlan0 interface to 1400 helps?
>>
>> Speaking about "ntoskrnl dummy called..."
>>
>> Try to change return value of IoWMIRegistrationControl, in
>> /sys/compat/ndis/subr_ntoskrnl.c , from STATUS_SUCCESS to
>> STATUS_NOT_IMPLEMENTED and recompile/reinstall kernel.
>>
>
> I have changed the return value, but the ntoskrnl messages are still
> being displayed at a rate of about one per second.

And on 7.2 message where not displayed? That's strange ...
>
> I have changed the MTU and am now able to log in to gmail, however when
> I do the following message appears:
>
> "Some Gmail features have failed to load due to an Internet
> connectivity problem."
>
> Moreover, the user interface is unresponsive.

Maybe miniport driver modify/corrupts ethernet packets somehow. Try to
pick packets transferred in air (via another card with working monitor
mode).
Tcpdump on wlan(0) can not help because that is what FreeBSD sends to
ndis and not what driver sends, so with your already reported output
we can only see outcome of bug, not real source of it.

Try code from gitorious, if something come up with bootverbose I could
add bunch of DEBUG
printfs.

-- 
Paul B Mahol