[PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Larry Finger]

John,

PLease send this upstream for inclusion in 2.6.18, if possible. This patch will not work for
wireless-2.6. That patch will be sent to you soon.

Larry

=============================================


This patch fixes a bug in the bcm43xx driver in 2.6.18-rcX that hangs the machine due to improper
locking. Between 2.6.17 and .18, longer portions of certain periodic work were made preemptible to
improve latency, which is how this bug was introduced. It happens relatively infrequently - every 6
- 10 hours, but when it does, the power button is the only possible recovery.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>

==================================


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
@@ -3182,19 +3182,21 @@ static void bcm43xx_periodic_work_handle
  		/* Periodic work will take a long time, so we want it to
  		 * be preemtible.
  		 */
-		bcm43xx_lock_irqonly(bcm, flags);
+		bcm43xx_lock_noirq(bcm);
  		netif_stop_queue(bcm->net_dev);
+		bcm43xx_lock_irqonly(bcm, flags);
+		bcm43xx_mac_suspend(bcm);
  		if (bcm43xx_using_pio(bcm))
  			bcm43xx_pio_freeze_txqueues(bcm);
  		savedirqs = bcm43xx_interrupt_disable(bcm, BCM43xx_IRQ_ALL);
  		bcm43xx_unlock_irqonly(bcm, flags);
-		bcm43xx_lock_noirq(bcm);
  		bcm43xx_synchronize_irq(bcm);
  	} else {
  		/* Periodic work should take short time, so we want low
  		 * locking overhead.
  		 */
-		bcm43xx_lock_irqsafe(bcm, flags);
+		bcm43xx_lock_noirq(bcm);
+		bcm43xx_lock_irqonly(bcm, flags);
  	}

  	do_periodic_work(bcm);
@@ -3206,6 +3208,7 @@ static void bcm43xx_periodic_work_handle
  			bcm43xx_interrupt_enable(bcm, savedirqs);
  			if (bcm43xx_using_pio(bcm))
  				bcm43xx_pio_thaw_txqueues(bcm);
+			bcm43xx_mac_enable(bcm);
  		}
  		netif_wake_queue(bcm->net_dev);
  		mmiowb();
@@ -3213,7 +3216,8 @@ static void bcm43xx_periodic_work_handle
  		bcm43xx_unlock_noirq(bcm);
  	} else {
  		mmiowb();
-		bcm43xx_unlock_irqsafe(bcm, flags);
+		bcm43xx_unlock_irqonly(bcm, flags);
+		bcm43xx_unlock_noirq(bcm);
  	}
  }

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[John W. Linville]

On Fri, Sep 08, 2006 at 08:47:54PM -0500, Larry Finger wrote:

> PLease send this upstream for inclusion in 2.6.18, if possible. This patch 
> will not work for
> wireless-2.6. That patch will be sent to you soon.

Are you saying this will break the upstream branch of wireless-2.6?
I'm not too excited about that.  Under the normal process, that would
mean 2.6.19 would be broken until we get the next patch.

When will the patch for the upstream branch be available?  Can you make
that patch based on this one (i.e. apply this one before generating
the next patch)?

John
-- 
John W. Linville
linville-2XuSBdqkA4R54TAoqtyWWQ@public.gmane.org

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Larry Finger]

John W. Linville wrote:
> On Fri, Sep 08, 2006 at 08:47:54PM -0500, Larry Finger wrote:
> 
>> PLease send this upstream for inclusion in 2.6.18, if possible. This patch 
>> will not work for
>> wireless-2.6. That patch will be sent to you soon.
> 
> Are you saying this will break the upstream branch of wireless-2.6?
> I'm not too excited about that.  Under the normal process, that would
> mean 2.6.19 would be broken until we get the next patch.
> 
> When will the patch for the upstream branch be available?  Can you make
> that patch based on this one (i.e. apply this one before generating
> the next patch)?

It will not break wireless-2.6. I sent the equivalent of this patch for wireless-2.6 to you under 
the subject "[PATCH] bcm43xx-softmac: Init, shutdown and restart fixes" on August 21, just after I 
took over maintenance of bcm43xx-softmac. If this is not currently queued, please let me know. There 
is still a problem with one section of the code touched by that patch that is related to a 
synchronize_net() call. Michael and I are trying to get a network guru to answer some questions 
regarding that function. So far no response to our pleas.

I would like to get a listing of patches for bcm43xx-softmac that are queued but not yet applied, 
and the order in which they will be applied. I want to make sure nothing has fallen through the 
cracks, and that the patches will apply cleanly.

Thanks,

Larry

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Daniel Drake]

John,

Larry Finger wrote:
> I would like to get a listing of patches for bcm43xx-softmac that are 
> queued but not yet applied, and the order in which they will be applied. 
> I want to make sure nothing has fallen through the cracks, and that the 
> patches will apply cleanly.

I know you have various other commitments but I would also find it 
beneficial if you could reduce the amount of time it takes between the 
patch arriving in your inbox and the commit appearing in 
wireless-2.6.git on kernel.org.

I'm currently waiting on the last few ZD1211 patches to appear in the 
git tree before sending the next few, since otherwise I have to 
backtrack and find out which patches I have already sent before 
producing the next batch.

I'm probably just being lazy -- the current scheme is workable, it is 
just a simple matter of convenience to have the patches appearing in git 
relatively quickly (even if there is still a delay before they head 
Jeffwards).

Thanks!
Daniel

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[John W. Linville]

On Sat, Sep 09, 2006 at 11:04:59AM -0500, Larry Finger wrote:
> John W. Linville wrote:
> >On Fri, Sep 08, 2006 at 08:47:54PM -0500, Larry Finger wrote:
> >
> >>PLease send this upstream for inclusion in 2.6.18, if possible. This 
> >>patch will not work for
> >>wireless-2.6. That patch will be sent to you soon.
> >
> >Are you saying this will break the upstream branch of wireless-2.6?
> >I'm not too excited about that.  Under the normal process, that would
> >mean 2.6.19 would be broken until we get the next patch.
> >
> >When will the patch for the upstream branch be available?  Can you make
> >that patch based on this one (i.e. apply this one before generating
> >the next patch)?
> 
> It will not break wireless-2.6. I sent the equivalent of this patch for 
> wireless-2.6 to you under the subject "[PATCH] bcm43xx-softmac: Init, 
> shutdown and restart fixes" on August 21, just after I took over 
> maintenance of bcm43xx-softmac. If this is not currently queued, please let 
> me know.

OK, I see now.  The patch won't apply to the upstream branch at all
due to the open-coded locking patch applied previously.

The problem now is that if I send this up for 2.6.18 then it creates
some messiness in the git repositories.  Plus, it still might not
make 2.6.18 because it is already very late.

I think I will have to ask you to hold-on to this until Linus
officially releases 2.6.18.  Then you can/should send this patch to
the stable team (stable-u79uwXL29TY76Z2rM5mHXA@public.gmane.org) for inclusion in 2.6.18.1
or later.

> I would like to get a listing of patches for bcm43xx-softmac that are 
> queued but not yet applied, and the order in which they will be applied. I 
> want to make sure nothing has fallen through the cracks, and that the 
> patches will apply cleanly.

I have these:

Date	Title
----	-----
8/21	bcm43xx-softmac: Init, shutdown and restart fixes
8/30	WE-21 for bcm43xx
9/3	remove dead code from bcm43xx-softmac
9/4	add firmware version printout
9/6	ucode debug status via sysfs
9/7	remove dead code in bcm43xx_sysfs.c

The first has not been merged because I haven't seen a reply to
Michael's last comments in that thread.  The WE-21 patch has not been
merged because I haven't merged the WE-21 stuff yet.  The others are
all from last week, which I just haven't processed yet.

Will you be responding to Michael's comments on the "Init, shutdown
and restart fixes" patch?

Thanks,

John

P.S.  For some reason, I often have trouble applying your patches.
I'm not sure what process you are using to generate your patches, but
please make sure that your patches can be successfully applied when
your email is fed to 'git applymbox'.  If you use git, the output of
'git format-patch' should be usable.

Also, if you could just use "---" instead of a long line of "======",
I think I would have to do less manual processing...thanks!
-- 
John W. Linville
linville-2XuSBdqkA4R54TAoqtyWWQ@public.gmane.org

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Larry Finger]

John W. Linville wrote:
> On Sat, Sep 09, 2006 at 11:04:59AM -0500, Larry Finger wrote:
>> John W. Linville wrote:
> 
> OK, I see now.  The patch won't apply to the upstream branch at all
> due to the open-coded locking patch applied previously.
> 
> The problem now is that if I send this up for 2.6.18 then it creates
> some messiness in the git repositories.  Plus, it still might not
> make 2.6.18 because it is already very late.
> 
> I think I will have to ask you to hold-on to this until Linus
> officially releases 2.6.18.  Then you can/should send this patch to
> the stable team (stable@vger.kernel.org) for inclusion in 2.6.18.1
> or later.

I will do that as soon as 18 is released. I don't want that bug in the 
wild too long.
> 
> I have these:
> 
> Date	Title
> ----	-----
> 8/21	bcm43xx-softmac: Init, shutdown and restart fixes
> 8/30	WE-21 for bcm43xx
> 9/3	remove dead code from bcm43xx-softmac
> 9/4	add firmware version printout
> 9/6	ucode debug status via sysfs
> 9/7	remove dead code in bcm43xx_sysfs.c
> 
> The first has not been merged because I haven't seen a reply to
> Michael's last comments in that thread.  The WE-21 patch has not been
> merged because I haven't merged the WE-21 stuff yet.  The others are
> all from last week, which I just haven't processed yet.
> 
> Will you be responding to Michael's comments on the "Init, shutdown
> and restart fixes" patch?

That has already been sent. As I said in that message, it looks as if 
the first patch has already been merged; therefore, I sent only the 
patch to move the step that Michael noted. If that is not true, let me 
know.

> P.S.  For some reason, I often have trouble applying your patches.
> I'm not sure what process you are using to generate your patches, but
> please make sure that your patches can be successfully applied when
> your email is fed to 'git applymbox'.  If you use git, the output of
> 'git format-patch' should be usable.
> 
> Also, if you could just use "---" instead of a long line of "======",
> I think I would have to do less manual processing...thanks!

Sorry about the === rather than ---. Is that documented somewhere?

I use quilt to generate the patches, but will check them with git in 
the future.

In addition to the items listed above, my list for wireless-2.6 includes:

8/15  bcm43xx: >1G and 64bit DMA support
8/15  bcm43xx: re-add bcm43xx_rng_init() call
8/15  Add Larry Finger to bcm43xx MAINTAINERS
8/16  bcm43xx-softmac: optimization of DMA bitfields.
8/19  bcm43xx: return correct hard_start_xmit error code]
9/10  Change PHY initialization to match updated clean-room specs

The first 3 were submitted by Michael before I took over.

I hope I'm not causing too much trouble while I learn the ropes.

Larry

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[John W. Linville]

On Mon, Sep 11, 2006 at 02:59:01PM -0500, Larry Finger wrote:
> John W. Linville wrote:

> >Also, if you could just use "---" instead of a long line of "======",
> >I think I would have to do less manual processing...thanks!
> 
> Sorry about the === rather than ---. Is that documented somewhere?

Not too explicitly, but I think Andrew mentions it in "The Perfect Patch":

	http://www.zip.com.au/~akpm/linux/patches/stuff/tpp.txt

While you are looking, you might review Jeff's version of the same idea:

	http://linux.yyz.us/patch-format.html
 
> I use quilt to generate the patches, but will check them with git in 
> the future.

It's not terribly critical, but it does slow me down if I have to
regenerate patches, edit comments, change subject lines, etc.

> In addition to the items listed above, my list for wireless-2.6 includes:
> 
> 8/15  bcm43xx: >1G and 64bit DMA support
> 8/15  bcm43xx: re-add bcm43xx_rng_init() call
> 8/15  Add Larry Finger to bcm43xx MAINTAINERS
> 8/16  bcm43xx-softmac: optimization of DMA bitfields.
> 8/19  bcm43xx: return correct hard_start_xmit error code]

I have all of these in the upstream branch of wireless-2.6.  BTW,
I have _not_ applied the Init patch yet. But I still have it, so I'll
just combine that w/ the new patch you posted this afternoon.

> 9/10  Change PHY initialization to match updated clean-room specs

The archives show this as being posted today at ~2:00pm EDT, but I
seem to have missed it?  Could you resend it just to me?

> I hope I'm not causing too much trouble while I learn the ropes.

We all go through it -- if you start feeding another maintainer,
he/she will probably have other quirks too... :-)

Thanks,

John
-- 
John W. Linville
linville-2XuSBdqkA4R54TAoqtyWWQ@public.gmane.org

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Jeff Garzik]

John W. Linville wrote:
> On Mon, Sep 11, 2006 at 02:59:01PM -0500, Larry Finger wrote:
>> John W. Linville wrote:
> 
>>> Also, if you could just use "---" instead of a long line of "======",
>>> I think I would have to do less manual processing...thanks!
>> Sorry about the === rather than ---. Is that documented somewhere?
> 
> Not too explicitly, but I think Andrew mentions it in "The Perfect Patch":
> 
> 	http://www.zip.com.au/~akpm/linux/patches/stuff/tpp.txt
> 
> While you are looking, you might review Jeff's version of the same idea:
> 
> 	http://linux.yyz.us/patch-format.html

And also Documentation/SubmittingPatches, which ATM seems to have more 
detail than either page.

	Jeff

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Jarek Poplawski]

On 12-09-2006 00:04, Jeff Garzik wrote:
> John W. Linville wrote:
...
>> Not too explicitly, but I think Andrew mentions it in "The Perfect 
>> Patch":
>>
>>     http://www.zip.com.au/~akpm/linux/patches/stuff/tpp.txt
>>
>> While you are looking, you might review Jeff's version of the same idea:
>>
>>     http://linux.yyz.us/patch-format.html
> 
> And also Documentation/SubmittingPatches, which ATM seems to have more 
> detail than either page.
> 
>     Jeff

"Attachments are discouraged, but some corporate mail systems 
provide no other way to send patches."

I thought they didn't read this but now I understand for whom 
Mozilla Firefox is breaking all those lines with no mercy!

Jarek P.

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Jarek Poplawski]

On Thu, Sep 14, 2006 at 10:25:32AM +0200, Jarek Poplawski wrote:
...
> "Attachments are discouraged, but some corporate mail systems 
> provide no other way to send patches."
> 
> I thought they didn't read this but now I understand for whom 
> Mozilla Firefox is breaking all those lines with no mercy!
 
Mozilla Thunderbird. Sorry.

Jarek P.

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Randy Dunlap]

On Thu, 14 Sep 2006 10:29:30 +0200 Jarek Poplawski wrote:

> On Thu, Sep 14, 2006 at 10:25:32AM +0200, Jarek Poplawski wrote:
> ...
> > "Attachments are discouraged, but some corporate mail systems 
> > provide no other way to send patches."
> > 
> > I thought they didn't read this but now I understand for whom 
> > Mozilla Firefox is breaking all those lines with no mercy!
>  
> Mozilla Thunderbird. Sorry.

see http://mbligh.org/linuxdocs/Email/Clients/Thunderbird

---
~Randy

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Jarek Poplawski]

On Thu, Oct 05, 2006 at 01:52:19PM -0700, Randy Dunlap wrote:
> On Thu, 14 Sep 2006 10:29:30 +0200 Jarek Poplawski wrote:
> 
> > On Thu, Sep 14, 2006 at 10:25:32AM +0200, Jarek Poplawski wrote:
> > ...
> > > "Attachments are discouraged, but some corporate mail systems 
> > > provide no other way to send patches."
> > > 
> > > I thought they didn't read this but now I understand for whom 
> > > Mozilla Firefox is breaking all those lines with no mercy!
> >  
> > Mozilla Thunderbird. Sorry.
> 
> see http://mbligh.org/linuxdocs/Email/Clients/Thunderbird

As a matter of fact I didn't have any doubts it can be
done in Thunderbird with some tweacking. But I don't
intend to try this. In my opinion there is even more
dangerous feature of Firebird - you never could be sure
how a message will look because it is reformated
after send command. So it isn't "you send what you see"
definitely (I mean text messages here). 

Jarek P.

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Jarek Poplawski]

On Fri, Oct 06, 2006 at 07:57:32AM +0200, Jarek Poplawski wrote:
> On Thu, Oct 05, 2006 at 01:52:19PM -0700, Randy Dunlap wrote:
> > On Thu, 14 Sep 2006 10:29:30 +0200 Jarek Poplawski wrote:
> > 
> > > On Thu, Sep 14, 2006 at 10:25:32AM +0200, Jarek Poplawski wrote:
...
> > > > Mozilla Firefox is breaking all those lines with no mercy!
> > >  
> > > Mozilla Thunderbird. Sorry.
...
> dangerous feature of Firebird - you never could be sure

Thunderbird. Sorry.
 
Jarek P.

PS: Once more and Firebird should kill me!

Re: [PATCH] fix for system lockups in 2.6.18-rcX caused by bcm43xx

[Jarek Poplawski]

On Fri, Oct 06, 2006 at 07:57:32AM +0200, Jarek Poplawski wrote:
> On Thu, Oct 05, 2006 at 01:52:19PM -0700, Randy Dunlap wrote:
...
> > see http://mbligh.org/linuxdocs/Email/Clients/Thunderbird
...

dangerous feature of Thunderbird - you never could be sure
> how a message will look because it is reformated
> after send command. So it isn't "you send what you see"
> definitely (I mean text messages here). 

I see from this link it was told in the discussion,
so I'll only add that with a feature like this even
if you do everything correctly you are never sure.
You always have to trust yourself you didn't miss
any step. 

Jarek P.