A problem with shared interrupts

A problem with shared interrupts

[Larry Finger]

Michael,

When I try the -mb tree on my old i386 notebook, I get an "irq 11: nobody cared"
message and interrupts for the bcm43xx-mac80211 device are not initialized. The
code in Linville's tree works. The only difference that seems to be important
is the code returned when a shared interrupt not intended for us is received.
By making the changes shown below, both trees now work.

I recall you having a discussion with someone over some aspect of shared IRQ's,
but I don't remember the details and I'm too busy (lazy) to go back and look.

Thanks,

Larry
----

Index: wireless-mb/drivers/net/wireless/mac80211/bcm43xx/bcm43xx_main.c
===================================================================
--- wireless-mb.orig/drivers/net/wireless/mac80211/bcm43xx/bcm43xx_main.c
+++ wireless-mb/drivers/net/wireless/mac80211/bcm43xx/bcm43xx_main.c
@@ -1503,7 +1503,7 @@ static void bcm43xx_interrupt_ack(struct
 /* Interrupt handler top-half */
 static irqreturn_t bcm43xx_interrupt_handler(int irq, void *dev_id)
 {
-	irqreturn_t ret = IRQ_NONE;
+	irqreturn_t ret = IRQ_HANDLED;
 	struct bcm43xx_wldev *dev = dev_id;
 	u32 reason;
 
@@ -1517,7 +1517,6 @@ static irqreturn_t bcm43xx_interrupt_han
 	reason = bcm43xx_read32(dev, BCM43xx_MMIO_GEN_IRQ_REASON);
 	if (reason == 0xffffffff) /* shared IRQ */
 		goto out;
-	ret = IRQ_HANDLED;
 	reason &= bcm43xx_read32(dev, BCM43xx_MMIO_GEN_IRQ_MASK);
 	if (!reason)
 		goto out;

Re: A problem with shared interrupts

[Michael Buesch]

On Thursday 26 July 2007 22:13:54 Larry Finger wrote:
> Michael,
> 
> When I try the -mb tree on my old i386 notebook, I get an "irq 11: nobody cared"
> message and interrupts for the bcm43xx-mac80211 device are not initialized. The
> code in Linville's tree works. The only difference that seems to be important
> is the code returned when a shared interrupt not intended for us is received.
> By making the changes shown below, both trees now work.
> 
> I recall you having a discussion with someone over some aspect of shared IRQ's,
> but I don't remember the details and I'm too busy (lazy) to go back and look.
> 
> Thanks,
> 
> Larry
> ----
> 
> Index: wireless-mb/drivers/net/wireless/mac80211/bcm43xx/bcm43xx_main.c
> ===================================================================
> --- wireless-mb.orig/drivers/net/wireless/mac80211/bcm43xx/bcm43xx_main.c
> +++ wireless-mb/drivers/net/wireless/mac80211/bcm43xx/bcm43xx_main.c
> @@ -1503,7 +1503,7 @@ static void bcm43xx_interrupt_ack(struct
>  /* Interrupt handler top-half */
>  static irqreturn_t bcm43xx_interrupt_handler(int irq, void *dev_id)
>  {
> -	irqreturn_t ret = IRQ_NONE;
> +	irqreturn_t ret = IRQ_HANDLED;
>  	struct bcm43xx_wldev *dev = dev_id;
>  	u32 reason;
>  
> @@ -1517,7 +1517,6 @@ static irqreturn_t bcm43xx_interrupt_han
>  	reason = bcm43xx_read32(dev, BCM43xx_MMIO_GEN_IRQ_REASON);
>  	if (reason == 0xffffffff) /* shared IRQ */
>  		goto out;
> -	ret = IRQ_HANDLED;
>  	reason &= bcm43xx_read32(dev, BCM43xx_MMIO_GEN_IRQ_MASK);
>  	if (!reason)
>  		goto out;
> 
> 

I neither think this is the correct solution, nor do I think that this
is the way bcm43xx-softmac does it. This would always return HANDLED, right?
regardless if the IRQ was for bcm43xx or not. I _do_ think that the bug is
in the driver sharing the IRQ with bcm43xx.

-- 
Greetings Michael.

Re: A problem with shared interrupts

[Larry Finger]

Michael Buesch wrote:
> 
> I neither think this is the correct solution, nor do I think that this
> is the way bcm43xx-softmac does it. This would always return HANDLED, right?
> regardless if the IRQ was for bcm43xx or not. I _do_ think that the bug is
> in the driver sharing the IRQ with bcm43xx.
> 

I misread the bcm43xx-mac80211 code from Linville's tree. clearly my patch is wrong. I'll have to 
find the differences between the two trees in drivers that share that IRQ. Fortunately, it is only 
yenta and bcm4301.

Larry

Re: A problem with shared interrupts

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 337 bytes --]

On Thu, 2007-07-26 at 15:13 -0500, Larry Finger wrote:
> Michael,
> 
> When I try the -mb tree on my old i386 notebook, I get an "irq 11: nobody cared"
> message and interrupts for the bcm43xx-mac80211 device are not initialized.

Hm. Just a thought about something related: do you have 'debug shared
interrupts' on?

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: A problem with shared interrupts

[Larry Finger]

Johannes Berg wrote:
> On Thu, 2007-07-26 at 15:13 -0500, Larry Finger wrote:
>> Michael,
>>
>> When I try the -mb tree on my old i386 notebook, I get an "irq 11: nobody cared"
>> message and interrupts for the bcm43xx-mac80211 device are not initialized.
> 
> Hm. Just a thought about something related: do you have 'debug shared
> interrupts' on?

No I didn't. Thanks for the suggestion.

Larry