RFT bcm43xx: Really _GOOD_ news

RFT bcm43xx: Really _GOOD_ news

[Larry Finger]

I have really _GOOD_ news!!!!! While reviewing the latest changes in the V4 specs, I found an
interchange of the PHY version and PHY revision fields, relative to the current V3 specs. Of the
three cards that I have, the 4306 had the same value for the PHY version and revision. In addition,
it was the only card of the 3 thaw would work at 11Mbs. A light went on!

After installing the patch listed below, both my 4318 and my 4311 will now run at 11Mbs. From tests
with iperf, I now get rates of 6.1 - 6.5 Mbs from all three.

Larry


==========

Index: linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c
===================================================================
--- linux-2.6.orig/drivers/net/wireless/bcm43xx/bcm43xx_main.c
+++ linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c
@@ -3707,9 +3707,9 @@ static int bcm43xx_read_phyinfo(struct b

 	value = bcm43xx_read16(bcm, BCM43xx_MMIO_PHY_VER);

-	phy_version = (value & 0xF000) >> 12;
+	phy_rev = (value & 0xF000) >> 12;
 	phy_type = (value & 0x0F00) >> 8;
-	phy_rev = (value & 0x000F);
+	phy_version = (value & 0x000F);

 	dprintk(KERN_INFO PFX "Detected PHY: Version: %x, Type %x, Revision %x\n",
 		phy_version, phy_type, phy_rev);

Re: RFT bcm43xx: Really _GOOD_ news

[Gene Heskett]

On Tuesday 06 February 2007 15:34, Larry Finger wrote:
>I have really _GOOD_ news!!!!! While reviewing the latest changes in the
> V4 specs, I found an interchange of the PHY version and PHY revision
> fields, relative to the current V3 specs. Of the three cards that I
> have, the 4306 had the same value for the PHY version and revision. In
> addition, it was the only card of the 3 thaw would work at 11Mbs. A
> light went on!
>
>After installing the patch listed below, both my 4318 and my 4311 will
> now run at 11Mbs. From tests with iperf, I now get rates of 6.1 - 6.5
> Mbs from all three.
>
>Larry
>

That's great Larry.  Now how long does it take to get this into a distro's 
update kernel queue?  Or do I have to build a kernel for my lappy, 
currently on FC5?

>==========
>
>Index: linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c
>===================================================================
>--- linux-2.6.orig/drivers/net/wireless/bcm43xx/bcm43xx_main.c
>+++ linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c
>@@ -3707,9 +3707,9 @@ static int bcm43xx_read_phyinfo(struct b
>
> 	value = bcm43xx_read16(bcm, BCM43xx_MMIO_PHY_VER);
>
>-	phy_version = (value & 0xF000) >> 12;
>+	phy_rev = (value & 0xF000) >> 12;
> 	phy_type = (value & 0x0F00) >> 8;
>-	phy_rev = (value & 0x000F);
>+	phy_version = (value & 0x000F);
>
> 	dprintk(KERN_INFO PFX "Detected PHY: Version: %x, Type %x, Revision
> %x\n", phy_version, phy_type, phy_rev);
>
>
>_______________________________________________
>Bcm43xx-dev mailing list
>Bcm43xx-dev@lists.berlios.de
>https://lists.berlios.de/mailman/listinfo/bcm43xx-dev

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Yahoo.com and AOL/TW attorneys please note, additions to the above
message by Gene Heskett are:
Copyright 2007 by Maurice Eugene Heskett, all rights reserved.

Re: RFT bcm43xx: Really _GOOD_ news

[Pavel Roskin]

On Tue, 2007-02-06 at 16:37 -0500, Gene Heskett wrote:

> That's great Larry.  Now how long does it take to get this into a
> distro's 
> update kernel queue?  Or do I have to build a kernel for my lappy, 
> currently on FC5?

Gene, I think it would be better for all of us if we don't bother Larry
with such questions and let him discover the secrets of the 802.11g
rates :)

I think that patch has a good chance to get into Fedora 7 from the
beginning.  We may see it in Fedora Core 6 if it goes to Linux 2.6.20
and this patch goes to the 2.6.20.x branch before that.

I think the best we can do now would be to test the driver and report
the results.  This would speed up the process of getting the patch to
the kernel.  Yes, that means recompiling the kernel.

-- 
Regards,
Pavel Roskin

Re: RFT bcm43xx: Really _GOOD_ news

[Gene Heskett]

On Tuesday 06 February 2007 17:29, Pavel Roskin wrote:
>On Tue, 2007-02-06 at 16:37 -0500, Gene Heskett wrote:
>> That's great Larry.  Now how long does it take to get this into a
>> distro's
>> update kernel queue?  Or do I have to build a kernel for my lappy,
>> currently on FC5?
>
>Gene, I think it would be better for all of us if we don't bother Larry
>with such questions and let him discover the secrets of the 802.11g
>rates :)
>
Probably, Larry seems to be the only one with the urge that's for sure.  
And, darnit, I keep forgetting to say thank you, Larry.

>I think that patch has a good chance to get into Fedora 7 from the
>beginning.  We may see it in Fedora Core 6 if it goes to Linux 2.6.20
>and this patch goes to the 2.6.20.x branch before that.

2.6.20 is out now, I'm running it on this box for the last 50+ hours.  So 
make it for 2.6.21 maybe.

>I think the best we can do now would be to test the driver and report
>the results.  This would speed up the process of getting the patch to
>the kernel.  Yes, that means recompiling the kernel.

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Yahoo.com and AOL/TW attorneys please note, additions to the above
message by Gene Heskett are:
Copyright 2007 by Maurice Eugene Heskett, all rights reserved.

Re: RFT bcm43xx: Really _GOOD_ news

[Larry Finger]

I was a bit too enthusiastic. The patch I sent earlier actually breaks the 4318. It does make 4311
and 4312's work for the wrong reason. If you have that card, you can use it in the meantime while we
find out just why it helps there.

Sorry for the false alarm, but we are making progress.

Larry

Re: RFT bcm43xx: Really _GOOD_ news

[Gene Heskett]

On Tuesday 06 February 2007 23:40, Larry Finger wrote:
>I was a bit too enthusiastic. The patch I sent earlier actually breaks
> the 4318. It does make 4311 and 4312's work for the wrong reason. If
> you have that card, you can use it in the meantime while we find out
> just why it helps there.
>
>Sorry for the false alarm, but we are making progress.
>
>Larry

And the 4318 is whats in my lappy.  What does it break?

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Yahoo.com and AOL/TW attorneys please note, additions to the above
message by Gene Heskett are:
Copyright 2007 by Maurice Eugene Heskett, all rights reserved.

Re: RFT bcm43xx: Really _GOOD_ news

[Pavel Roskin]

On Wed, 2007-02-07 at 03:03 -0500, Gene Heskett wrote:

> And the 4318 is whats in my lappy.  What does it break?

Oh, please, let's stop such questions.  Does it really matter what some
knowingly wrong change does to particular hardware?  Just don't try the
old patch.  Just forget that it existed.  It's not going to the kernel.

Please try the new patch.  Ideally, measure speeds for various rates,
because it's the speed that the patch is supposed to improve.

If you have more questions of this kind, please make sure my personal
e-mail is not in CC.

-- 
Regards,
Pavel Roskin

Re: RFT bcm43xx: Really _GOOD_ news

[Daniel Gryniewicz]

On Wed, 2007-02-07 at 03:17 -0500, Pavel Roskin wrote:
> On Wed, 2007-02-07 at 03:03 -0500, Gene Heskett wrote:
> 
> > And the 4318 is whats in my lappy.  What does it break?
> 
> Oh, please, let's stop such questions.  Does it really matter what some
> knowingly wrong change does to particular hardware?  Just don't try the
> old patch.  Just forget that it existed.  It's not going to the kernel.
> 
> Please try the new patch.  Ideally, measure speeds for various rates,
> because it's the speed that the patch is supposed to improve.
> 
> If you have more questions of this kind, please make sure my personal
> e-mail is not in CC.
> 

Except that the new patch changes nothing for us poor 4318 users, so
what's the point of trying it?

That said, I'm running the broken patch right now, and my 4318 is
working better than ever before, so it changed *something* for my card.

Daniel

Re: RFT bcm43xx: Really _GOOD_ news

[Larry Finger]

Daniel Gryniewicz wrote:
> On Wed, 2007-02-07 at 03:17 -0500, Pavel Roskin wrote:
>> On Wed, 2007-02-07 at 03:03 -0500, Gene Heskett wrote:
>>
>>> And the 4318 is whats in my lappy.  What does it break?
>> Oh, please, let's stop such questions.  Does it really matter what some
>> knowingly wrong change does to particular hardware?  Just don't try the
>> old patch.  Just forget that it existed.  It's not going to the kernel.
>>
>> Please try the new patch.  Ideally, measure speeds for various rates,
>> because it's the speed that the patch is supposed to improve.
>>
>> If you have more questions of this kind, please make sure my personal
>> e-mail is not in CC.
>>esg 
> 
> Except that the new patch changes nothing for us poor 4318 users, so
> what's the point of trying it?
> 
> That said, I'm running the broken patch right now, and my 4318 is
> working better than ever before, so it changed *something* for my card.

Intriguing.  Please send me the output of 'dmesg | grep bcm43xx'.  Thanks.

Larry

Re: RFT bcm43xx: Really _GOOD_ news

[Pavel Roskin]

On Tue, 2007-02-06 at 14:34 -0600, Larry Finger wrote:
> I have really _GOOD_ news!!!!! While reviewing the latest changes in the V4 specs, I found an
> interchange of the PHY version and PHY revision fields, relative to the current V3 specs. Of the
> three cards that I have, the 4306 had the same value for the PHY version and revision. In addition,
> it was the only card of the 3 thaw would work at 11Mbs. A light went on!
> 
> After installing the patch listed below, both my 4318 and my 4311 will now run at 11Mbs. From tests
> with iperf, I now get rates of 6.1 - 6.5 Mbs from all three.

I'm getting a sustained 1.7 megabytes per second over FTP with 4312
(PCIe card in Dell Latitude).  The connection to a 802.11g AP was
established instantaneously.

I have never had much success with the softmac driver at all.  Now it's
working better than the d80211 driver, with waits for minutes before it
starts talking to the same 802.11g AP.

I think it's a 2.6.20.1 material, considering its gain/changes ratio.
I'm actually going to compile this kernel without debug and keep it as a
"production" kernel.

Congratulations and many many thanks!

-- 
Regards,
Pavel Roskin

Re: RFT bcm43xx: Really _GOOD_ news

[Rafael J. Wysocki]

On Tuesday, 6 February 2007 23:18, Pavel Roskin wrote:
> On Tue, 2007-02-06 at 14:34 -0600, Larry Finger wrote:
> > I have really _GOOD_ news!!!!! While reviewing the latest changes in the V4 specs, I found an
> > interchange of the PHY version and PHY revision fields, relative to the current V3 specs. Of the
> > three cards that I have, the 4306 had the same value for the PHY version and revision. In addition,
> > it was the only card of the 3 thaw would work at 11Mbs. A light went on!
> > 
> > After installing the patch listed below, both my 4318 and my 4311 will now run at 11Mbs. From tests
> > with iperf, I now get rates of 6.1 - 6.5 Mbs from all three.
> 
> I'm getting a sustained 1.7 megabytes per second over FTP with 4312
> (PCIe card in Dell Latitude).  The connection to a 802.11g AP was
> established instantaneously.
> 
> I have never had much success with the softmac driver at all.  Now it's
> working better than the d80211 driver, with waits for minutes before it
> starts talking to the same 802.11g AP.
> 
> I think it's a 2.6.20.1 material, considering its gain/changes ratio.
> I'm actually going to compile this kernel without debug and keep it as a
> "production" kernel.
> 
> Congratulations and many many thanks!

I can confirm the driver works great with the patch!

Many thanks, Larry!

Greetings,
Rafael

Re: RFT bcm43xx: Really _GOOD_ news

[Larry Finger]

Rafael J. Wysocki wrote:
> On Tuesday, 6 February 2007 23:18, Pavel Roskin wrote:
>> On Tue, 2007-02-06 at 14:34 -0600, Larry Finger wrote:
>>> I have really _GOOD_ news!!!!! While reviewing the latest changes in the V4 specs, I found an
>>> interchange of the PHY version and PHY revision fields, relative to the current V3 specs. Of the
>>> three cards that I have, the 4306 had the same value for the PHY version and revision. In addition,
>>> it was the only card of the 3 thaw would work at 11Mbs. A light went on!
>>>
>>> After installing the patch listed below, both my 4318 and my 4311 will now run at 11Mbs. From tests
>>> with iperf, I now get rates of 6.1 - 6.5 Mbs from all three.
>> I'm getting a sustained 1.7 megabytes per second over FTP with 4312
>> (PCIe card in Dell Latitude).  The connection to a 802.11g AP was
>> established instantaneously.
>>
>> I have never had much success with the softmac driver at all.  Now it's
>> working better than the d80211 driver, with waits for minutes before it
>> starts talking to the same 802.11g AP.
>>
>> I think it's a 2.6.20.1 material, considering its gain/changes ratio.
>> I'm actually going to compile this kernel without debug and keep it as a
>> "production" kernel.
>>
>> Congratulations and many many thanks!
> 
> I can confirm the driver works great with the patch!
> 
> Many thanks, Larry!

You are all very welcome.

Now that others have verified that the patch makes their system work, I will be submitting it for
wireless-2.6 that will get it into 2.6.21. I will also submit it to stable so it should be included
in 2.6.20.1 and 2.6.19.4.

Larry

Re: RFT bcm43xx: Really _GOOD_ news

[Jory A. Pratt]

Larry Finger wrote:
> I have really _GOOD_ news!!!!! While reviewing the latest changes in the V4 specs, I found an
> interchange of the PHY version and PHY revision fields, relative to the current V3 specs. Of the
> three cards that I have, the 4306 had the same value for the PHY version and revision. In addition,
> it was the only card of the 3 thaw would work at 11Mbs. A light went on!
>
> After installing the patch listed below, both my 4318 and my 4311 will now run at 11Mbs. From tests
> with iperf, I now get rates of 6.1 - 6.5 Mbs from all three.
>
> Larry
>
>
> ==========
>
> Index: linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c
> ===================================================================
> --- linux-2.6.orig/drivers/net/wireless/bcm43xx/bcm43xx_main.c
> +++ linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c
> @@ -3707,9 +3707,9 @@ static int bcm43xx_read_phyinfo(struct b
>
>  	value = bcm43xx_read16(bcm, BCM43xx_MMIO_PHY_VER);
>
> -	phy_version = (value & 0xF000) >> 12;
> +	phy_rev = (value & 0xF000) >> 12;
>  	phy_type = (value & 0x0F00) >> 8;
> -	phy_rev = (value & 0x000F);
> +	phy_version = (value & 0x000F);
>
>  	dprintk(KERN_INFO PFX "Detected PHY: Version: %x, Type %x, Revision %x\n",
>  		phy_version, phy_type, phy_rev);
>
>   
I have tested this until I am blue in the face on a 4318. It killed me,
I was getting less then 200k/b a sec on a 5M pipe :(. I reverted the
patch and all is back to working even tho transfering files on the
internal network is still way out there.

Re: RFT bcm43xx: Really _GOOD_ news

[John W. Linville]

On Thu, Feb 08, 2007 at 12:20:30PM -0600, Jory A. Pratt wrote:
> Larry Finger wrote:
> > I have really _GOOD_ news!!!!! While reviewing the latest changes in the V4 specs, I found an
> > interchange of the PHY version and PHY revision fields, relative to the current V3 specs. Of the
> > three cards that I have, the 4306 had the same value for the PHY version and revision. In addition,
> > it was the only card of the 3 thaw would work at 11Mbs. A light went on!
> >
> > After installing the patch listed below, both my 4318 and my 4311 will now run at 11Mbs. From tests
> > with iperf, I now get rates of 6.1 - 6.5 Mbs from all three.

> I have tested this until I am blue in the face on a 4318. It killed me,
> I was getting less then 200k/b a sec on a 5M pipe :(. I reverted the
> patch and all is back to working even tho transfering files on the
> internal network is still way out there.

Jory,

Have you tried the patch Larry posted yesterday w/ subject "RFT:
The real fix for BCM4311 and BCM4312"?

Thanks,

John
-- 
John W. Linville
linville@tuxdriver.com