[ath9k-devel] ar9281

[ath9k-devel] ar9281

[Pat O'Neil]

Greetings all,

I've got a recent toshiba qosmio laptop, 64bit that has an ar9281
chipset that was "almost" working under F9, as after downloading the
wireless git and compiling I could get it to come up and associate, but
the performance was extremely slow, with ping times of 10's of seconds,
etc.  I also sometimes got the aforementioned "No ProbeResp from current
AP ... - assume out of range" error.

Since then I've loaded Fedora 10, and have hit a new snag with its
bundled version of ath9k, as in the dmesg log I see the following:

...snip...

ath9k: 0.1
ath_pci: software IOTLB in use, aborting.

...snip...

and no wireless interface shows up at all.  Any ideas?  

Thanks for all the work on atheros drivers!

Pat O'Neil

[ath9k-devel] ar9281

[Luis R. Rodriguez]

On Mon, Dec 1, 2008 at 9:13 AM, Pat O'Neil <Pat.ONeil@fedex.com> wrote:
> Greetings all,
>
> I've got a recent toshiba qosmio laptop, 64bit that has an ar9281
> chipset that was "almost" working under F9, as after downloading the
> wireless git and compiling I could get it to come up and associate, but
> the performance was extremely slow, with ping times of 10's of seconds,
> etc.  I also sometimes got the aforementioned "No ProbeResp from current
> AP ... - assume out of range" error.
>
> Since then I've loaded Fedora 10, and have hit a new snag with its
> bundled version of ath9k, as in the dmesg log I see the following:
>
> ...snip...
>
> ath9k: 0.1
> ath_pci: software IOTLB in use, aborting.
>
> ...snip...

Fedora 10 kernel is broken as a workaround to a kernel bug in ath9k
they decided to disable ath9k when the swiotbl is used due to a bug
that *we actually did fix*. They should fix their kernel. They may
have an updated kernel you may want to check.

http://bugzilla.kernel.org/show_bug.cgi?id=11811

  Luis

[ath9k-devel] ar9281

[John W. Linville]

On Mon, Dec 01, 2008 at 09:25:33AM -0800, Luis R. Rodriguez wrote:
> On Mon, Dec 1, 2008 at 9:13 AM, Pat O'Neil <Pat.ONeil@fedex.com> wrote:
> > Greetings all,
> >
> > I've got a recent toshiba qosmio laptop, 64bit that has an ar9281
> > chipset that was "almost" working under F9, as after downloading the
> > wireless git and compiling I could get it to come up and associate, but
> > the performance was extremely slow, with ping times of 10's of seconds,
> > etc.  I also sometimes got the aforementioned "No ProbeResp from current
> > AP ... - assume out of range" error.
> >
> > Since then I've loaded Fedora 10, and have hit a new snag with its
> > bundled version of ath9k, as in the dmesg log I see the following:
> >
> > ...snip...
> >
> > ath9k: 0.1
> > ath_pci: software IOTLB in use, aborting.
> >
> > ...snip...
> 
> Fedora 10 kernel is broken as a workaround to a kernel bug in ath9k
> they decided to disable ath9k when the swiotbl is used due to a bug
> that *we actually did fix*. They should fix their kernel. They may
> have an updated kernel you may want to check.
> 
> http://bugzilla.kernel.org/show_bug.cgi?id=11811

Not quite that simple.  Those patches as written don't apply to 2.6.27,
and F10 won't be on 2.6.28 for quite some time (since 2.6.28 has yet
to be released).

Now, if you know of 2.6.27 versions of those patches... :-)

John
-- 
John W. Linville		Linux should be at the core
linville at tuxdriver.com			of your literate lifestyle.

[ath9k-devel] ar9281

[Luis R. Rodriguez]

On Tue, Dec 2, 2008 at 12:22 PM, John W. Linville
<linville@tuxdriver.com> wrote:
> On Mon, Dec 01, 2008 at 09:25:33AM -0800, Luis R. Rodriguez wrote:
>> On Mon, Dec 1, 2008 at 9:13 AM, Pat O'Neil <Pat.ONeil@fedex.com> wrote:
>> > Greetings all,
>> >
>> > I've got a recent toshiba qosmio laptop, 64bit that has an ar9281
>> > chipset that was "almost" working under F9, as after downloading the
>> > wireless git and compiling I could get it to come up and associate, but
>> > the performance was extremely slow, with ping times of 10's of seconds,
>> > etc.  I also sometimes got the aforementioned "No ProbeResp from current
>> > AP ... - assume out of range" error.
>> >
>> > Since then I've loaded Fedora 10, and have hit a new snag with its
>> > bundled version of ath9k, as in the dmesg log I see the following:
>> >
>> > ...snip...
>> >
>> > ath9k: 0.1
>> > ath_pci: software IOTLB in use, aborting.
>> >
>> > ...snip...
>>
>> Fedora 10 kernel is broken as a workaround to a kernel bug in ath9k
>> they decided to disable ath9k when the swiotbl is used due to a bug
>> that *we actually did fix*. They should fix their kernel. They may
>> have an updated kernel you may want to check.
>>
>> http://bugzilla.kernel.org/show_bug.cgi?id=11811
>
> Not quite that simple.  Those patches as written don't apply to 2.6.27,
> and F10 won't be on 2.6.28 for quite some time (since 2.6.28 has yet
> to be released).
>
> Now, if you know of 2.6.27 versions of those patches... :-)

I provided a port of the required 3 patches to 2.6.27 in the bug
report, here's the linky:

http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2008-11-22/27-IOMMU-01/

Let me know if there is anything else I can do to help.

  Luis

[ath9k-devel] ar9281

[Christoph Thiel]

On Tue, Dec 02, 2008 at 03:22:22PM -0500, John W. Linville wrote:
> On Mon, Dec 01, 2008 at 09:25:33AM -0800, Luis R. Rodriguez wrote:
> > On Mon, Dec 1, 2008 at 9:13 AM, Pat O'Neil <Pat.ONeil@fedex.com> wrote:
> > > Greetings all,
> > >
> > > I've got a recent toshiba qosmio laptop, 64bit that has an ar9281
> > > chipset that was "almost" working under F9, as after downloading the
> > > wireless git and compiling I could get it to come up and associate, but
> > > the performance was extremely slow, with ping times of 10's of seconds,
> > > etc.  I also sometimes got the aforementioned "No ProbeResp from current
> > > AP ... - assume out of range" error.
> > >
> > > Since then I've loaded Fedora 10, and have hit a new snag with its
> > > bundled version of ath9k, as in the dmesg log I see the following:
> > >
> > > ...snip...
> > >
> > > ath9k: 0.1
> > > ath_pci: software IOTLB in use, aborting.
> > >
> > > ...snip...
> > 
> > Fedora 10 kernel is broken as a workaround to a kernel bug in ath9k
> > they decided to disable ath9k when the swiotbl is used due to a bug
> > that *we actually did fix*. They should fix their kernel. They may
> > have an updated kernel you may want to check.
> > 
> > http://bugzilla.kernel.org/show_bug.cgi?id=11811
> 
> Not quite that simple.  Those patches as written don't apply to 2.6.27,
> and F10 won't be on 2.6.28 for quite some time (since 2.6.28 has yet
> to be released).
> 
> Now, if you know of 2.6.27 versions of those patches... :-)

Check out
http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2008-11-22/27-IOMMU-01/

I hope we'll get them included in one of the next -stable kernels.



Best,
Christoph

[ath9k-devel] ar9281

[John W. Linville]

On Tue, Dec 02, 2008 at 12:36:36PM -0800, Luis R. Rodriguez wrote:
> On Tue, Dec 2, 2008 at 12:22 PM, John W. Linville
> <linville@tuxdriver.com> wrote:
> > On Mon, Dec 01, 2008 at 09:25:33AM -0800, Luis R. Rodriguez wrote:
> >> On Mon, Dec 1, 2008 at 9:13 AM, Pat O'Neil <Pat.ONeil@fedex.com> wrote:
> >> > Greetings all,
> >> >
> >> > I've got a recent toshiba qosmio laptop, 64bit that has an ar9281
> >> > chipset that was "almost" working under F9, as after downloading the
> >> > wireless git and compiling I could get it to come up and associate, but
> >> > the performance was extremely slow, with ping times of 10's of seconds,
> >> > etc.  I also sometimes got the aforementioned "No ProbeResp from current
> >> > AP ... - assume out of range" error.
> >> >
> >> > Since then I've loaded Fedora 10, and have hit a new snag with its
> >> > bundled version of ath9k, as in the dmesg log I see the following:
> >> >
> >> > ...snip...
> >> >
> >> > ath9k: 0.1
> >> > ath_pci: software IOTLB in use, aborting.
> >> >
> >> > ...snip...
> >>
> >> Fedora 10 kernel is broken as a workaround to a kernel bug in ath9k
> >> they decided to disable ath9k when the swiotbl is used due to a bug
> >> that *we actually did fix*. They should fix their kernel. They may
> >> have an updated kernel you may want to check.
> >>
> >> http://bugzilla.kernel.org/show_bug.cgi?id=11811
> >
> > Not quite that simple.  Those patches as written don't apply to 2.6.27,
> > and F10 won't be on 2.6.28 for quite some time (since 2.6.28 has yet
> > to be released).
> >
> > Now, if you know of 2.6.27 versions of those patches... :-)
> 
> I provided a port of the required 3 patches to 2.6.27 in the bug
> report, here's the linky:
> 
> http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2008-11-22/27-IOMMU-01/
> 
> Let me know if there is anything else I can do to help.

Ah, sorry I overlooked that.  BTW, did you ever send the third one
to linux-wireless?  I don't remember seeing it there.

John
-- 
John W. Linville		Linux should be at the core
linville at tuxdriver.com			of your literate lifestyle.

[ath9k-devel] ar9281

[Luis R. Rodriguez]

On Tue, Dec 2, 2008 at 12:49 PM, John W. Linville
<linville@tuxdriver.com> wrote:
> On Tue, Dec 02, 2008 at 12:36:36PM -0800, Luis R. Rodriguez wrote:
>> On Tue, Dec 2, 2008 at 12:22 PM, John W. Linville
>> <linville@tuxdriver.com> wrote:
>> > On Mon, Dec 01, 2008 at 09:25:33AM -0800, Luis R. Rodriguez wrote:
>> >> On Mon, Dec 1, 2008 at 9:13 AM, Pat O'Neil <Pat.ONeil@fedex.com> wrote:
>> >> > Greetings all,
>> >> >
>> >> > I've got a recent toshiba qosmio laptop, 64bit that has an ar9281
>> >> > chipset that was "almost" working under F9, as after downloading the
>> >> > wireless git and compiling I could get it to come up and associate, but
>> >> > the performance was extremely slow, with ping times of 10's of seconds,
>> >> > etc.  I also sometimes got the aforementioned "No ProbeResp from current
>> >> > AP ... - assume out of range" error.
>> >> >
>> >> > Since then I've loaded Fedora 10, and have hit a new snag with its
>> >> > bundled version of ath9k, as in the dmesg log I see the following:
>> >> >
>> >> > ...snip...
>> >> >
>> >> > ath9k: 0.1
>> >> > ath_pci: software IOTLB in use, aborting.
>> >> >
>> >> > ...snip...
>> >>
>> >> Fedora 10 kernel is broken as a workaround to a kernel bug in ath9k
>> >> they decided to disable ath9k when the swiotbl is used due to a bug
>> >> that *we actually did fix*. They should fix their kernel. They may
>> >> have an updated kernel you may want to check.
>> >>
>> >> http://bugzilla.kernel.org/show_bug.cgi?id=11811
>> >
>> > Not quite that simple.  Those patches as written don't apply to 2.6.27,
>> > and F10 won't be on 2.6.28 for quite some time (since 2.6.28 has yet
>> > to be released).
>> >
>> > Now, if you know of 2.6.27 versions of those patches... :-)
>>
>> I provided a port of the required 3 patches to 2.6.27 in the bug
>> report, here's the linky:
>>
>> http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2008-11-22/27-IOMMU-01/
>>
>> Let me know if there is anything else I can do to help.
>
> Ah, sorry I overlooked that.  BTW, did you ever send the third one
> to linux-wireless?  I don't remember seeing it there.

Yup, you have applied all 3:

commit 44ace58df6be47e5c63f152cde87c8d83f2dc93d
commit 49763f277cd0aa34744394ae8e31374a00eb479a
commit 352c2843429b572279d208caa1e74b8201b395cb

The one that is missing is the fix to remove the existence of the
emergency DMA bounce buffers in the swiotlb as that is what caused the
hard drive issues. But I didn't write that patch and Maciej was going
to post them.

Applying these 3 patches mean the issue won't pop up again though for
ath9k users as means we are no longer hogging up DMA. Maciej's patch
will be needed for any other buggy DMA drivers hogging up bounce
buffers.

Please let me know if you need anything else.

 Luis

[ath9k-devel] ar9281

[John W. Linville]

On Tue, Dec 02, 2008 at 01:11:54PM -0800, Luis R. Rodriguez wrote:
> On Tue, Dec 2, 2008 at 12:49 PM, John W. Linville
> <linville@tuxdriver.com> wrote:

> > Ah, sorry I overlooked that.  BTW, did you ever send the third one
> > to linux-wireless?  I don't remember seeing it there.
> 
> Yup, you have applied all 3:
> 
> commit 44ace58df6be47e5c63f152cde87c8d83f2dc93d
> commit 49763f277cd0aa34744394ae8e31374a00eb479a
> commit 352c2843429b572279d208caa1e74b8201b395cb

Ah, well clearly I'm having a delayed Monday... :-)  FWIW, I'm still
suffering from a bit of a cold...

Thanks!

John
-- 
John W. Linville		Linux should be at the core
linville at tuxdriver.com			of your literate lifestyle.