Request for bisecting DMA partial fix

Request for bisecting DMA partial fix

[Rafał Miłecki]

HI,

We are quite sure that some patch than gone into mainline between
2.6.34 and 2.6.35 partially fixed DMA errors. It was not b43-specific
fix and it probably wasn't commited via wireless-testing.

The problem is this fix was only partial, under high transfer this is
still noticeable. I think the shortest way to understand this and fix
it totally is to find commit that partially fixed DMA errors.

So we need someone with:
1) Machine that throws DMA errors on 2.6.34 and works noticeably
better with 2.6.35
2) Time to bisect commit
3) Basic kernel building skills (however I believe we can help in case
of problems)

I believe bisecting has to be done on Linus's tree, between mentioned kernels.

So is there anyone around who could handle this task?

-- 
Rafa?

Request for bisecting DMA partial fix

[Gábor Stefanik]

2010/11/8 Rafa? Mi?ecki <zajec5@gmail.com>:
> HI,
>
> We are quite sure that some patch than gone into mainline between
> 2.6.34 and 2.6.35 partially fixed DMA errors. It was not b43-specific
> fix and it probably wasn't commited via wireless-testing.
>
> The problem is this fix was only partial, under high transfer this is
> still noticeable. I think the shortest way to understand this and fix
> it totally is to find commit that partially fixed DMA errors.
>
> So we need someone with:
> 1) Machine that throws DMA errors on 2.6.34 and works noticeably
> better with 2.6.35
> 2) Time to bisect commit
> 3) Basic kernel building skills (however I believe we can help in case
> of problems)
>
> I believe bisecting has to be done on Linus's tree, between mentioned kernels.
>
> So is there anyone around who could handle this task?

Well, my functional programming teacher at the university is
experiencing this bug (& resorted to using wl in the meantime -
machine is a Lenovo S10e), but I don't know how well he understands
the Linux kernel (given that it's not written in Haskell). But I'll
write back if I find out anything. In the meantime, keep on searching.

>
> --
> Rafa?
>
> _______________________________________________
> b43-dev mailing list
> b43-dev at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/b43-dev
>



-- 
Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)

Request for bisecting DMA partial fix

[Rafał Miłecki]

W dniu 8 listopada 2010 20:42 u?ytkownik Rafa? Mi?ecki
<zajec5@gmail.com> napisa?:
> We are quite sure that some patch than gone into mainline between
> 2.6.34 and 2.6.35 partially fixed DMA errors. It was not b43-specific
> fix and it probably wasn't commited via wireless-testing.
>
> The problem is this fix was only partial, under high transfer this is
> still noticeable. I think the shortest way to understand this and fix
> it totally is to find commit that partially fixed DMA errors.
>
> So we need someone with:
> 1) Machine that throws DMA errors on 2.6.34 and works noticeably
> better with 2.6.35
> 2) Time to bisect commit
> 3) Basic kernel building skills (however I believe we can help in case
> of problems)
>
> I believe bisecting has to be done on Linus's tree, between mentioned kernels.
>
> So is there anyone around who could handle this task?

Anyone? Please?

-- 
Rafa?

Request for bisecting DMA partial fix

[Eric Garrido]

On 2010-11-25 10:22am, Rafa? Mi?ecki wrote:
> W dniu 8 listopada 2010 20:42 u?ytkownik Rafa? Mi?ecki
> <zajec5@gmail.com> napisa?:
> > We are quite sure that some patch than gone into mainline between
> > 2.6.34 and 2.6.35 partially fixed DMA errors. It was not b43-specific
> > fix and it probably wasn't commited via wireless-testing.
> >
> > The problem is this fix was only partial, under high transfer this is
> > still noticeable. I think the shortest way to understand this and fix
> > it totally is to find commit that partially fixed DMA errors.
> >
> > So we need someone with:
> > 1) Machine that throws DMA errors on 2.6.34 and works noticeably
> > better with 2.6.35
> > 2) Time to bisect commit
> > 3) Basic kernel building skills (however I believe we can help in case
> > of problems)
> >
> > I believe bisecting has to be done on Linus's tree, between mentioned kernels.
> >
> > So is there anyone around who could handle this task?
> 
> Anyone? Please?

I have 2) and 3), plus have a Lenovo S10 that I haven't used in months
that appears (based on a GP and some research) to be susceptible to
this issue. In fact, I think I hit this issue and started using wl
oh so long ago (sorry!).

I'm away for the NA holiday, but I can try to verify I have 1) and
then poke at this on Sunday. I'll let you know.

Request for bisecting DMA partial fix

[Rafał Miłecki]

W dniu 26 listopada 2010 18:39 u?ytkownik Eric Garrido
<eric@ericgar.com> napisa?:
> On 2010-11-25 10:22am, Rafa? Mi?ecki wrote:
>> W dniu 8 listopada 2010 20:42 u?ytkownik Rafa? Mi?ecki
>> <zajec5@gmail.com> napisa?:
>> > We are quite sure that some patch than gone into mainline between
>> > 2.6.34 and 2.6.35 partially fixed DMA errors. It was not b43-specific
>> > fix and it probably wasn't commited via wireless-testing.
>> >
>> > The problem is this fix was only partial, under high transfer this is
>> > still noticeable. I think the shortest way to understand this and fix
>> > it totally is to find commit that partially fixed DMA errors.
>> >
>> > So we need someone with:
>> > 1) Machine that throws DMA errors on 2.6.34 and works noticeably
>> > better with 2.6.35
>> > 2) Time to bisect commit
>> > 3) Basic kernel building skills (however I believe we can help in case
>> > of problems)
>> >
>> > I believe bisecting has to be done on Linus's tree, between mentioned kernels.
>> >
>> > So is there anyone around who could handle this task?
>>
>> Anyone? Please?
>
> I have 2) and 3), plus have a Lenovo S10 that I haven't used in months
> that appears (based on a GP and some research) to be susceptible to
> this issue. In fact, I think I hit this issue and started using wl
> oh so long ago (sorry!).
>
> I'm away for the NA holiday, but I can try to verify I have 1) and
> then poke at this on Sunday. I'll let you know.

Ping? :)

-- 
Rafa?

Request for bisecting DMA partial fix

[Eric Garrido]

On 2010-12-01  8:55pm, Rafa? Mi?ecki wrote:
> > I'm away for the NA holiday, but I can try to verify I have 1) and
> > then poke at this on Sunday. I'll let you know.
> 
> Ping? :)

Ack. My apologies. Life got in the way (giving notice to my employer,
when one would think I'd work *less* afterward, but the opposite has
proven to be true). I should be *actually* able to do this tomorrow,
and if need be, on Sunday.

Request for bisecting DMA partial fix

[Rafał Miłecki]

W dniu 11 grudnia 2010 09:51 u?ytkownik Eric Garrido <eric@ericgar.com> napisa?:
> On 2010-12-01 ?8:55pm, Rafa? Mi?ecki wrote:
>> > I'm away for the NA holiday, but I can try to verify I have 1) and
>> > then poke at this on Sunday. I'll let you know.
>>
>> Ping? :)
>
> Ack. My apologies. Life got in the way (giving notice to my employer,
> when one would think I'd work *less* afterward, but the opposite has
> proven to be true). I should be *actually* able to do this tomorrow,
> and if need be, on Sunday.

Great, thanks :)

Larry: when you noticed this fix on your machine, did you use same
config for building new kernel as was used for earlier (distro?) one?

-- 
Rafa?

Request for bisecting DMA partial fix

[Larry Finger]

On 12/11/2010 11:16 AM, Rafa? Mi?ecki wrote:
> W dniu 11 grudnia 2010 09:51 u?ytkownik Eric Garrido <eric@ericgar.com> napisa?:
>> On 2010-12-01  8:55pm, Rafa? Mi?ecki wrote:
>>>> I'm away for the NA holiday, but I can try to verify I have 1) and
>>>> then poke at this on Sunday. I'll let you know.
>>>
>>> Ping? :)
>>
>> Ack. My apologies. Life got in the way (giving notice to my employer,
>> when one would think I'd work *less* afterward, but the opposite has
>> proven to be true). I should be *actually* able to do this tomorrow,
>> and if need be, on Sunday.
> 
> Great, thanks :)
> 
> Larry: when you noticed this fix on your machine, did you use same
> config for building new kernel as was used for earlier (distro?) one?

No. I always trim the module list to match the hardware on the machine, and add
some module hacking options that they leave off.

I need to go back to 2.6.34 to verify that I do get DMA errors with that kernel.

Larry

Request for bisecting DMA partial fix

[Rafał Miłecki]

W dniu 11 grudnia 2010 18:22 u?ytkownik Larry Finger
<Larry.Finger@lwfinger.net> napisa?:
> On 12/11/2010 11:16 AM, Rafa? Mi?ecki wrote:
>> W dniu 11 grudnia 2010 09:51 u?ytkownik Eric Garrido <eric@ericgar.com> napisa?:
>>> On 2010-12-01 ?8:55pm, Rafa? Mi?ecki wrote:
>>>>> I'm away for the NA holiday, but I can try to verify I have 1) and
>>>>> then poke at this on Sunday. I'll let you know.
>>>>
>>>> Ping? :)
>>>
>>> Ack. My apologies. Life got in the way (giving notice to my employer,
>>> when one would think I'd work *less* afterward, but the opposite has
>>> proven to be true). I should be *actually* able to do this tomorrow,
>>> and if need be, on Sunday.
>>
>> Great, thanks :)
>>
>> Larry: when you noticed this fix on your machine, did you use same
>> config for building new kernel as was used for earlier (distro?) one?
>
> No. I always trim the module list to match the hardware on the machine, and add
> some module hacking options that they leave off.
>
> I need to go back to 2.6.34 to verify that I do get DMA errors with that kernel.

I don't want to try be smarter than you... ;) but is that possible you
used distro's 2.6.34, and then directly switched to self-compiled
2.6.35? Then difference in frequency of DMA errors could be matter of
kernel configuration.

Of course, as you said, compiling 2.6.34 with your standard
configuration would explain that.

-- 
Rafa?

Request for bisecting DMA partial fix

[Eric Garrido]

On 2010-12-11  6:16pm, Rafa? Mi?ecki wrote:
> W dniu 11 grudnia 2010 09:51 u?ytkownik Eric Garrido <eric@ericgar.com> napisa?:
> > On 2010-12-01 ?8:55pm, Rafa? Mi?ecki wrote:
> >> > I'm away for the NA holiday, but I can try to verify I have 1) and
> >> > then poke at this on Sunday. I'll let you know.
> >>
> >> Ping? :)
> >
> > Ack. My apologies. Life got in the way (giving notice to my employer,
> > when one would think I'd work *less* afterward, but the opposite has
> > proven to be true). I should be *actually* able to do this tomorrow,
> > and if need be, on Sunday.
> 
> Great, thanks :)

I'm still hacking on this and will report any findings.

Request for bisecting DMA partial fix

[Rafał Miłecki]

Guys, is there anyone experiencing DMA errors on high transfer? I'd be
even interested in getting SSH access to affected machine, if noone
has time to debug this.

-- 
Rafa?

Request for bisecting DMA partial fix

[Juan Carlos Romero]

I have a Lenovo S10e with this problem since one year and a half. Can I help
somehow ?

2011/4/14 Rafa? Mi?ecki <zajec5@gmail.com>

> Guys, is there anyone experiencing DMA errors on high transfer? I'd be
> even interested in getting SSH access to affected machine, if noone
> has time to debug this.
>
> --
> Rafa?
>
> _______________________________________________
> b43-dev mailing list
> b43-dev at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/b43-dev
>



-- 
J. Carlos
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/b43-dev/attachments/20110414/bd0aef8a/attachment.html>

Request for bisecting DMA partial fix

[Rafał Miłecki]

W dniu 14 kwietnia 2011 22:37 u?ytkownik Juan Carlos Romero
<juancarlos.romero@gmail.com> napisa?:
> I have a Lenovo S10e with this problem since one year and a half. Can I help
> somehow ?

Of course! First questions:
1) On which kernel do you experience DMA errors
2) How easy is this for you to see DMA errors (normal ping, or torrent
transfer is needed?)
3) Did you try kernel 2.6.35 or newer? Do you still experience DMA
errors with it?

-- 
Rafa?

Request for bisecting DMA partial fix

[Juan Carlos Romero]

I have experienced DMA errors always. I have used all kernels from 2.6.32 to
2.6.38 (Debian official kernels)

I see DMA errors always if I boot with b43 module. I always boot with wl
module, rmmod it and moprobe b43. Then I  get no errors.

Can we chat in #bcm-users channel ?

2011/4/14 Rafa? Mi?ecki <zajec5@gmail.com>

> W dniu 14 kwietnia 2011 22:37 u?ytkownik Juan Carlos Romero
> <juancarlos.romero@gmail.com> napisa?:
> > I have a Lenovo S10e with this problem since one year and a half. Can I
> help
> > somehow ?
>
> Of course! First questions:
> 1) On which kernel do you experience DMA errors
> 2) How easy is this for you to see DMA errors (normal ping, or torrent
> transfer is needed?)
> 3) Did you try kernel 2.6.35 or newer? Do you still experience DMA
> errors with it?
>
> --
> Rafa?
>



-- 
J. Carlos
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/b43-dev/attachments/20110414/46f79454/attachment.html>

Request for bisecting DMA partial fix

[Larry Finger]

On 04/14/2011 03:47 PM, Juan Carlos Romero wrote:
> I have experienced DMA errors always. I have used all kernels from 2.6.32 to
> 2.6.38 (Debian official kernels)
>
> I see DMA errors always if I boot with b43 module. I always boot with wl module,
> rmmod it and moprobe b43. Then I  get no errors.
>
> Can we chat in #bcm-users channel ?

That is the BCM4312 (14e4:4315) PCI initialization problem, not a heavy traffic 
problem.

Larry

Request for bisecting DMA partial fix

[Larry Finger]

On 04/14/2011 01:39 PM, Rafa? Mi?ecki wrote:
> Guys, is there anyone experiencing DMA errors on high transfer? I'd be
> even interested in getting SSH access to affected machine, if noone
> has time to debug this.

I have a netbook that gets out-of-order TX descriptors on a 14e4:4315 under 
heavy loads. I am not sure if this is a regression, a problem with modern 
firmware, or what. This is happening with 2.6.38. I'm currently testing to see 
if it happens with earlier versions.

Larry