2.6.19-rc1: forcedeth, nobody cared

2.6.19-rc1: forcedeth, nobody cared

[Prakash Punnoor]

[-- Attachment #1: Type: text/plain, Size: 1036 bytes --]

Hi,

subjects say it all. Without irqpoll my nic doesn't work anymore. I added Ingo 
to cc, as my IRQs look different, so it may be a prob of APIC routing or the 
like.

The kernel is patched with reiser4 and acpi_skip_timer_override quirk is 
deactivated (see last link why).

I tried different combinations (dmesg + .config). Differences are mostly pci 
mt init, irqpoll, nforce eth napi, pata/ide amd driver. Last is current 
(working, but with irqpoll)

http://www.prakash.gmxhome.de/linux/2.6.19-rc1-1.txt.bz2
http://www.prakash.gmxhome.de/linux/2.6.19-rc1-2.txt.bz2
http://www.prakash.gmxhome.de/linux/2.6.19-rc1-3.txt.bz2
http://www.prakash.gmxhome.de/linux/2.6.19-rc1-4.txt.bz2

cat /proc/interrupts for 2.6.18 resp .19-rc1
http://www.prakash.gmxhome.de/linux/irqs18.txt
http://www.prakash.gmxhome.de/linux/irqs19.txt

lspci can be found here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=115545986619977&w=2

Cheers,
-- 
(°=                 =°)
//\ Prakash Punnoor /\\
V_/                 \_V

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.6.19-rc1: forcedeth, nobody cared

[Prakash Punnoor]

[-- Attachment #1: Type: text/plain, Size: 374 bytes --]

Am Donnerstag 05 Oktober 2006 09:38 schrieb Prakash Punnoor:
> subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> Ingo to cc, as my IRQs look different, so it may be a prob of APIC routing
> or the like.

BTW, it also happens w/o nvidia binary loaded.

Cheers,
-- 
(°=                 =°)
//\ Prakash Punnoor /\\
V_/                 \_V

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.6.19-rc1: forcedeth, nobody cared

[Gene Heskett]

On Thursday 05 October 2006 04:03, Prakash Punnoor wrote:
>Am Donnerstag 05 Oktober 2006 09:38 schrieb Prakash Punnoor:
>> subjects say it all. Without irqpoll my nic doesn't work anymore. I
>> added Ingo to cc, as my IRQs look different, so it may be a prob of
>> APIC routing or the like.
>
>BTW, it also happens w/o nvidia binary loaded.
>
>Cheers,

It might be something local to your system, Prakash, its working normally 
here, with no unusual messages in /var/log/dmesg.  But I'm an ati user, 
in-kernel radeon driver too.

-- 
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 2006 by Maurice Eugene Heskett, all rights reserved.

Re: 2.6.19-rc1: forcedeth, nobody cared

[Fatih Aşıcı]

2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> Hi,
>
> subjects say it all. Without irqpoll my nic doesn't work anymore. I added Ingo
> to cc, as my IRQs look different, so it may be a prob of APIC routing or the
> like.
>
> The kernel is patched with reiser4 and acpi_skip_timer_override quirk is
> deactivated (see last link why).
>
> I tried different combinations (dmesg + .config). Differences are mostly pci
> mt init, irqpoll, nforce eth napi, pata/ide amd driver. Last is current
> (working, but with irqpoll)
>
Can you try booting with pci=nomsi ? I have a similar problem with my sound
device:

https://bugtrack.alsa-project.org/alsa-bug/view.php?id=2503

Re: 2.6.19-rc1: forcedeth, nobody cared

[Prakash Punnoor]

[-- Attachment #1: Type: text/plain, Size: 1568 bytes --]

Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih Aşıcı:
> 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > Hi,
> >
> > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > Ingo
> > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > the
> > like.

> > Can you try booting with pci=nomsi ? I have a similar problem with my

I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
nic works w/o problems. So it was the audio driver causing havoc on the nic. 
But (I noticed this with the kernel booted with irqpoll, as well) IRQ 
balancing seems not to work anymore:

           CPU0       CPU1
  0:      53826          0   IO-APIC-edge     timer
  1:         66          0   IO-APIC-edge     i8042
  7:          0          0   IO-APIC-edge     parport0
  8:          0          0   IO-APIC-edge     rtc
  9:          0          0   IO-APIC-fasteoi  acpi
 12:       1650          0   IO-APIC-edge     i8042
 14:         63          0   IO-APIC-edge     ide0
 16:       2095          0   IO-APIC-fasteoi  nvidia
 19:          6          0   IO-APIC-fasteoi  ohci1394
 20:          0          0   IO-APIC-fasteoi  ohci_hcd:usb1
 21:         66          0   IO-APIC-fasteoi  ehci_hcd:usb2
 22:       5652          0   IO-APIC-fasteoi  libata
 23:       5105          0   IO-APIC-fasteoi  HDA Intel, eth0
NMI:        155         75
LOC:      53715      53649
ERR:          0


-- 
(°=                 =°)
//\ Prakash Punnoor /\\
V_/                 \_V

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

[RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Adrian Bunk]

On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > Hi,
> > >
> > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > Ingo
> > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > the
> > > like.
> 
> > > Can you try booting with pci=nomsi ? I have a similar problem with my
> 
> I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> nic works w/o problems. So it was the audio driver causing havoc on the nic. 
>...

Unless someone finds and fixes what causes such problems, I'd therefore 
suggest the patch below to let MSI support to be turned off by default.

cu
Adrian


<--  snip  -->


As reported in http://lkml.org/lkml/2006/10/7/164, MSI support in 
snd-hda-intel seems to break some previously working setups.

Signed-off-by: Adrian Bunk <bunk@stusta.de>

---

 Documentation/sound/alsa/ALSA-Configuration.txt |    2 +-
 sound/pci/hda/hda_intel.c                       |   14 +++++++-------
 2 files changed, 8 insertions(+), 8 deletions(-)

--- linux-2.6/Documentation/sound/alsa/ALSA-Configuration.txt.old	2006-10-17 18:11:48.000000000 +0200
+++ linux-2.6/Documentation/sound/alsa/ALSA-Configuration.txt	2006-10-17 18:12:54.000000000 +0200
@@ -753,7 +753,7 @@
     position_fix - Fix DMA pointer (0 = auto, 1 = none, 2 = POSBUF, 3 = FIFO size)
     single_cmd  - Use single immediate commands to communicate with
 		codecs (for debugging only)
-    disable_msi - Disable Message Signaled Interrupt (MSI)
+    enable_msi - Enable Message Signaled Interrupt (MSI)
 
     This module supports one card and autoprobe.
 
--- linux-2.6/sound/pci/hda/hda_intel.c.old	2006-10-17 18:10:05.000000000 +0200
+++ linux-2.6/sound/pci/hda/hda_intel.c	2006-10-17 18:10:56.000000000 +0200
@@ -55,7 +55,7 @@
 static int position_fix;
 static int probe_mask = -1;
 static int single_cmd;
-static int disable_msi;
+static int enable_msi;
 
 module_param(index, int, 0444);
 MODULE_PARM_DESC(index, "Index value for Intel HD audio interface.");
@@ -69,8 +69,8 @@
 MODULE_PARM_DESC(probe_mask, "Bitmask to probe codecs (default = -1).");
 module_param(single_cmd, bool, 0444);
 MODULE_PARM_DESC(single_cmd, "Use single command to communicate with codecs (for debugging only).");
-module_param(disable_msi, int, 0);
-MODULE_PARM_DESC(disable_msi, "Disable Message Signaled Interrupt (MSI)");
+module_param(enable_msi, int, 0);
+MODULE_PARM_DESC(enable_msi, "Enable Message Signaled Interrupt (MSI)");
 
 
 /* just for backward compatibility */
@@ -1381,7 +1381,7 @@
 	azx_free_cmd_io(chip);
 	if (chip->irq >= 0)
 		free_irq(chip->irq, chip);
-	if (!disable_msi)
+	if (enable_msi)
 		pci_disable_msi(chip->pci);
 	pci_disable_device(pci);
 	pci_save_state(pci);
@@ -1395,7 +1395,7 @@
 
 	pci_restore_state(pci);
 	pci_enable_device(pci);
-	if (!disable_msi)
+	if (enable_msi)
 		pci_enable_msi(pci);
 	/* FIXME: need proper error handling */
 	request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
@@ -1437,7 +1437,7 @@
 
 	if (chip->irq >= 0) {
 		free_irq(chip->irq, (void*)chip);
-		if (!disable_msi)
+		if (enable_msi)
 			pci_disable_msi(chip->pci);
 	}
 	if (chip->remap_addr)
@@ -1523,7 +1523,7 @@
 		goto errout;
 	}
 
-	if (!disable_msi)
+	if (enable_msi)
 		pci_enable_msi(pci);
 
 	if (request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,

Re: [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Randy Dunlap]

On Tue, 17 Oct 2006 23:13:01 +0200 Adrian Bunk wrote:

> On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > Hi,
> > > >
> > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > Ingo
> > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > the
> > > > like.
> > 
> > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > 
> > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> >...
> 
> Unless someone finds and fixes what causes such problems, I'd therefore 
> suggest the patch below to let MSI support to be turned off by default.

MSI works with my sound, ethernet, and SATA:

 66:    1131814          0         PCI-MSI  libata
 98:      89556          0         PCI-MSI  HDA Intel
106:     943530          0         PCI-MSI  eth0

so maybe this should be done for nforce chipsets or
whatever is seeing the problem?


> <--  snip  -->
> 
> 
> As reported in http://lkml.org/lkml/2006/10/7/164, MSI support in 
> snd-hda-intel seems to break some previously working setups.
> 
> Signed-off-by: Adrian Bunk <bunk@stusta.de>
> 
> ---
> 
>  Documentation/sound/alsa/ALSA-Configuration.txt |    2 +-
>  sound/pci/hda/hda_intel.c                       |   14 +++++++-------
>  2 files changed, 8 insertions(+), 8 deletions(-)
> 
> --- linux-2.6/Documentation/sound/alsa/ALSA-Configuration.txt.old	2006-10-17 18:11:48.000000000 +0200
> +++ linux-2.6/Documentation/sound/alsa/ALSA-Configuration.txt	2006-10-17 18:12:54.000000000 +0200
> @@ -753,7 +753,7 @@
>      position_fix - Fix DMA pointer (0 = auto, 1 = none, 2 = POSBUF, 3 = FIFO size)
>      single_cmd  - Use single immediate commands to communicate with
>  		codecs (for debugging only)
> -    disable_msi - Disable Message Signaled Interrupt (MSI)
> +    enable_msi - Enable Message Signaled Interrupt (MSI)
>  
>      This module supports one card and autoprobe.
>  
> --- linux-2.6/sound/pci/hda/hda_intel.c.old	2006-10-17 18:10:05.000000000 +0200
> +++ linux-2.6/sound/pci/hda/hda_intel.c	2006-10-17 18:10:56.000000000 +0200
> @@ -55,7 +55,7 @@
>  static int position_fix;
>  static int probe_mask = -1;
>  static int single_cmd;
> -static int disable_msi;
> +static int enable_msi;
>  
>  module_param(index, int, 0444);
>  MODULE_PARM_DESC(index, "Index value for Intel HD audio interface.");
> @@ -69,8 +69,8 @@
>  MODULE_PARM_DESC(probe_mask, "Bitmask to probe codecs (default = -1).");
>  module_param(single_cmd, bool, 0444);
>  MODULE_PARM_DESC(single_cmd, "Use single command to communicate with codecs (for debugging only).");
> -module_param(disable_msi, int, 0);
> -MODULE_PARM_DESC(disable_msi, "Disable Message Signaled Interrupt (MSI)");
> +module_param(enable_msi, int, 0);
> +MODULE_PARM_DESC(enable_msi, "Enable Message Signaled Interrupt (MSI)");
>  
>  
>  /* just for backward compatibility */
> @@ -1381,7 +1381,7 @@
>  	azx_free_cmd_io(chip);
>  	if (chip->irq >= 0)
>  		free_irq(chip->irq, chip);
> -	if (!disable_msi)
> +	if (enable_msi)
>  		pci_disable_msi(chip->pci);
>  	pci_disable_device(pci);
>  	pci_save_state(pci);
> @@ -1395,7 +1395,7 @@
>  
>  	pci_restore_state(pci);
>  	pci_enable_device(pci);
> -	if (!disable_msi)
> +	if (enable_msi)
>  		pci_enable_msi(pci);
>  	/* FIXME: need proper error handling */
>  	request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
> @@ -1437,7 +1437,7 @@
>  
>  	if (chip->irq >= 0) {
>  		free_irq(chip->irq, (void*)chip);
> -		if (!disable_msi)
> +		if (enable_msi)
>  			pci_disable_msi(chip->pci);
>  	}
>  	if (chip->remap_addr)
> @@ -1523,7 +1523,7 @@
>  		goto errout;
>  	}
>  
> -	if (!disable_msi)
> +	if (enable_msi)
>  		pci_enable_msi(pci);
>  
>  	if (request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 


---
~Randy

Re: [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Takashi Iwai]

At Tue, 17 Oct 2006 14:29:57 -0700,
Randy Dunlap wrote:
> 
> On Tue, 17 Oct 2006 23:13:01 +0200 Adrian Bunk wrote:
> 
> > On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > > Hi,
> > > > >
> > > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > > Ingo
> > > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > > the
> > > > > like.
> > > 
> > > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > > 
> > > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> > >...
> > 
> > Unless someone finds and fixes what causes such problems, I'd therefore 
> > suggest the patch below to let MSI support to be turned off by default.
> 
> MSI works with my sound, ethernet, and SATA:
> 
>  66:    1131814          0         PCI-MSI  libata
>  98:      89556          0         PCI-MSI  HDA Intel
> 106:     943530          0         PCI-MSI  eth0
> 
> so maybe this should be done for nforce chipsets or
> whatever is seeing the problem?

I agree that this is pretty much depending on the hardware, so
disabling all might be exaggerated.
The sound with MSI works on all of my test machines with Intel
chipset, so far.


Takashi

Re: [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Stephen Hemminger]

On Tue, 17 Oct 2006 23:13:01 +0200
Adrian Bunk <bunk@stusta.de> wrote:

> On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > Hi,
> > > >
> > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > Ingo
> > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > the
> > > > like.
> > 
> > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > 
> > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> >...
> 
> Unless someone finds and fixes what causes such problems, I'd therefore 
> suggest the patch below to let MSI support to be turned off by default.
> 
> cu
> Adrian
> 

It shouldn't be that hard to write a small bit of code to force an interrupt
and catch it, that's what other drivers do to workaround the BIOS braindamage
that seems to be rampant (until M$ Vista comes out and supports MSI).

Re: [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Prakash Punnoor]

[-- Attachment #1: Type: text/plain, Size: 1202 bytes --]

Am Dienstag 17 Oktober 2006 23:40 schrieb Stephen Hemminger:
> On Tue, 17 Oct 2006 23:13:01 +0200
>
> Adrian Bunk <bunk@stusta.de> wrote:
> > On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > > Hi,
> > > > >
> > > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I
> > > > > added Ingo
> > > > > to cc, as my IRQs look different, so it may be a prob of APIC
> > > > > routing or the
> > > > > like.
> > > > >
> > > > > Can you try booting with pci=nomsi ? I have a similar problem with
> > > > > my
> > >
> > > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the
> > > nforce nic works w/o problems. So it was the audio driver causing havoc
> > > on the nic. ...
> >
> > Unless someone finds and fixes what causes such problems, I'd therefore
> > suggest the patch below to let MSI support to be turned off by default.

Could it perhaps be that the forcedeth driver isn't MSI aware and this causes 
breakage?

-- 
(°=                 =°)
//\ Prakash Punnoor /\\
V_/                 \_V

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Takashi Iwai]

At Tue, 17 Oct 2006 14:40:53 -0700,
Stephen Hemminger wrote:
> 
> On Tue, 17 Oct 2006 23:13:01 +0200
> Adrian Bunk <bunk@stusta.de> wrote:
> 
> > On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > > Hi,
> > > > >
> > > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > > Ingo
> > > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > > the
> > > > > like.
> > > 
> > > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > > 
> > > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> > >...
> > 
> > Unless someone finds and fixes what causes such problems, I'd therefore 
> > suggest the patch below to let MSI support to be turned off by default.
> > 
> > cu
> > Adrian
> > 
> 
> It shouldn't be that hard to write a small bit of code to force an interrupt
> and catch it, that's what other drivers do to workaround the BIOS braindamage
> that seems to be rampant (until M$ Vista comes out and supports MSI).

OK, what about a patch like below?

It's against the latest ALSA tree, so please pull alsa.git from
	git://kernel.org/pub/scm/linux/kernel/git/perex/alsa.git
before applying it (this tree includes only patches to be pushed to
2.6.19).


Takashi

diff -r 0047a372140b pci/hda/hda_intel.c
--- a/pci/hda/hda_intel.c	Tue Oct 17 20:41:38 2006 +0200
+++ b/pci/hda/hda_intel.c	Wed Oct 18 11:12:01 2006 +0200
@@ -337,6 +337,7 @@ struct azx {
 	unsigned int initialized :1;
 	unsigned int single_cmd :1;
 	unsigned int polling_mode :1;
+	unsigned int msi :1;
 };
 
 /* driver types */
@@ -397,6 +398,7 @@ static char *driver_short_names[] __devi
  */
 #define upper_32bit(addr) (sizeof(addr) > 4 ? (u32)((addr) >> 32) : (u32)0)
 
+static int azx_acquire_irq(struct azx *chip);
 
 /*
  * Interface for HD codec
@@ -535,6 +537,18 @@ static unsigned int azx_rirb_get_respons
 			return chip->rirb.res; /* the last value */
 		schedule_timeout_interruptible(1);
 	} while (time_after_eq(timeout, jiffies));
+
+	if (chip->msi) {
+		snd_printk(KERN_WARNING "hda_intel: No response from codec, "
+			   "disabling MSI...\n");
+		free_irq(chip->irq, chip);
+		chip->irq = -1;
+		pci_disable_msi(chip->pci);
+		chip->msi = 0;
+		if (azx_acquire_irq(chip) < 0)
+			return -1;
+		goto again;
+	}
 
 	if (!chip->polling_mode) {
 		snd_printk(KERN_WARNING "hda_intel: azx_get_response timeout, "
@@ -1364,6 +1378,19 @@ static int __devinit azx_init_stream(str
 	return 0;
 }
 
+static int azx_acquire_irq(struct azx *chip)
+{
+	if (request_irq(chip->pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
+			"HDA Intel", chip)) {
+		printk(KERN_ERR "hda-intel: unable to grab IRQ %d, "
+		       "disabling device\n", chip->pci->irq);
+		snd_card_disconnect(chip->card);
+		return -1;
+	}
+	chip->irq = chip->pci->irq;
+	return 0;
+}
+
 
 #ifdef CONFIG_PM
 /*
@@ -1385,7 +1412,7 @@ static int azx_suspend(struct pci_dev *p
 		free_irq(chip->irq, chip);
 		chip->irq = -1;
 	}
-	if (!disable_msi)
+	if (chip->msi)
 		pci_disable_msi(chip->pci);
 	pci_disable_device(pci);
 	pci_save_state(pci);
@@ -1407,16 +1434,10 @@ static int azx_resume(struct pci_dev *pc
 		return -EIO;
 	}
 	pci_set_master(pci);
-	if (!disable_msi)
+	if (chip->msi)
 		pci_enable_msi(pci);
-	if (request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
-			"HDA Intel", chip)) {
-		printk(KERN_ERR "hda-intel: unable to grab IRQ %d, "
-		       "disabling device\n", pci->irq);
-		snd_card_disconnect(card);
+	if (azx_acquire_irq(chip) < 0)
 		return -EIO;
-	}
-	chip->irq = pci->irq;
 	azx_init_chip(chip);
 	snd_hda_resume(chip->bus);
 	snd_power_change_state(card, SNDRV_CTL_POWER_D0);
@@ -1452,7 +1473,7 @@ static int azx_free(struct azx *chip)
 		synchronize_irq(chip->irq);
 		free_irq(chip->irq, (void*)chip);
 	}
-	if (!disable_msi)
+	if (chip->msi)
 		pci_disable_msi(chip->pci);
 	if (chip->remap_addr)
 		iounmap(chip->remap_addr);
@@ -1508,6 +1529,7 @@ static int __devinit azx_create(struct s
 	chip->pci = pci;
 	chip->irq = -1;
 	chip->driver_type = driver_type;
+	chip->msi = !disable_msi;
 
 	chip->position_fix = position_fix;
 	chip->single_cmd = single_cmd;
@@ -1537,7 +1559,7 @@ static int __devinit azx_create(struct s
 		goto errout;
 	}
 
-	if (!disable_msi)
+	if (chip->msi)
 		pci_enable_msi(pci);
 
 	if (request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Stephen Hemminger]

On Wed, 18 Oct 2006 12:24:54 +0200
Takashi Iwai <tiwai@suse.de> wrote:

> At Tue, 17 Oct 2006 14:40:53 -0700,
> Stephen Hemminger wrote:
> > 
> > On Tue, 17 Oct 2006 23:13:01 +0200
> > Adrian Bunk <bunk@stusta.de> wrote:
> > 
> > > On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > > > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > > > Hi,
> > > > > >
> > > > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > > > Ingo
> > > > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > > > the
> > > > > > like.
> > > > 
> > > > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > > > 
> > > > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > > > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> > > >...
> > > 
> > > Unless someone finds and fixes what causes such problems, I'd therefore 
> > > suggest the patch below to let MSI support to be turned off by default.
> > > 
> > > cu
> > > Adrian
> > > 
> > 
> > It shouldn't be that hard to write a small bit of code to force an interrupt
> > and catch it, that's what other drivers do to workaround the BIOS braindamage
> > that seems to be rampant (until M$ Vista comes out and supports MSI).
> 
> OK, what about a patch like below?
> 
> It's against the latest ALSA tree, so please pull alsa.git from
> 	git://kernel.org/pub/scm/linux/kernel/git/perex/alsa.git
> before applying it (this tree includes only patches to be pushed to
> 2.6.19).
>

Looks good, I'll try it

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Stephen Hemminger]

On Wed, 18 Oct 2006 12:24:54 +0200
Takashi Iwai <tiwai@suse.de> wrote:

> At Tue, 17 Oct 2006 14:40:53 -0700,
> Stephen Hemminger wrote:
> > 
> > On Tue, 17 Oct 2006 23:13:01 +0200
> > Adrian Bunk <bunk@stusta.de> wrote:
> > 
> > > On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > > > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > > > Hi,
> > > > > >
> > > > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > > > Ingo
> > > > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > > > the
> > > > > > like.
> > > > 
> > > > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > > > 
> > > > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > > > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> > > >...
> > > 
> > > Unless someone finds and fixes what causes such problems, I'd therefore 
> > > suggest the patch below to let MSI support to be turned off by default.
> > > 
> > > cu
> > > Adrian
> > > 
> > 
> > It shouldn't be that hard to write a small bit of code to force an interrupt
> > and catch it, that's what other drivers do to workaround the BIOS braindamage
> > that seems to be rampant (until M$ Vista comes out and supports MSI).
> 
> OK, what about a patch like below?
> 
> It's against the latest ALSA tree, so please pull alsa.git from
> 	git://kernel.org/pub/scm/linux/kernel/git/perex/alsa.git
> before applying it (this tree includes only patches to be pushed to
> 2.6.19).
> 
> 
> Takashi
> 

Don't you need to check return value from pci_enable_msi()?

-- 
Stephen Hemminger <shemminger@osdl.org>

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Takashi Iwai]

At Wed, 18 Oct 2006 10:02:19 -0700,
Stephen Hemminger wrote:
> 
> On Wed, 18 Oct 2006 12:24:54 +0200
> Takashi Iwai <tiwai@suse.de> wrote:
> 
> > At Tue, 17 Oct 2006 14:40:53 -0700,
> > Stephen Hemminger wrote:
> > > 
> > > On Tue, 17 Oct 2006 23:13:01 +0200
> > > Adrian Bunk <bunk@stusta.de> wrote:
> > > 
> > > > On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > > > > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > > > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > > > > Hi,
> > > > > > >
> > > > > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > > > > Ingo
> > > > > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > > > > the
> > > > > > > like.
> > > > > 
> > > > > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > > > > 
> > > > > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > > > > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> > > > >...
> > > > 
> > > > Unless someone finds and fixes what causes such problems, I'd therefore 
> > > > suggest the patch below to let MSI support to be turned off by default.
> > > > 
> > > > cu
> > > > Adrian
> > > > 
> > > 
> > > It shouldn't be that hard to write a small bit of code to force an interrupt
> > > and catch it, that's what other drivers do to workaround the BIOS braindamage
> > > that seems to be rampant (until M$ Vista comes out and supports MSI).
> > 
> > OK, what about a patch like below?
> > 
> > It's against the latest ALSA tree, so please pull alsa.git from
> > 	git://kernel.org/pub/scm/linux/kernel/git/perex/alsa.git
> > before applying it (this tree includes only patches to be pushed to
> > 2.6.19).
> > 
> > 
> > Takashi
> > 
> 
> Don't you need to check return value from pci_enable_msi()?

Yes, it would be better to check the value and reset chip->msi if
not successful.  But it's not a fatal error, so the current code
should work.


Takashi

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Takashi Iwai]

At Wed, 18 Oct 2006 19:12:50 +0200,
I wrote:
> 
> At Wed, 18 Oct 2006 10:02:19 -0700,
> Stephen Hemminger wrote:
> > 
> > On Wed, 18 Oct 2006 12:24:54 +0200
> > Takashi Iwai <tiwai@suse.de> wrote:
> > 
> > > At Tue, 17 Oct 2006 14:40:53 -0700,
> > > Stephen Hemminger wrote:
> > > > 
> > > > On Tue, 17 Oct 2006 23:13:01 +0200
> > > > Adrian Bunk <bunk@stusta.de> wrote:
> > > > 
> > > > > On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > > > > > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > > > > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > > > > > Hi,
> > > > > > > >
> > > > > > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > > > > > Ingo
> > > > > > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > > > > > the
> > > > > > > > like.
> > > > > > 
> > > > > > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> > > > > > 
> > > > > > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce 
> > > > > > nic works w/o problems. So it was the audio driver causing havoc on the nic. 
> > > > > >...
> > > > > 
> > > > > Unless someone finds and fixes what causes such problems, I'd therefore 
> > > > > suggest the patch below to let MSI support to be turned off by default.
> > > > > 
> > > > > cu
> > > > > Adrian
> > > > > 
> > > > 
> > > > It shouldn't be that hard to write a small bit of code to force an interrupt
> > > > and catch it, that's what other drivers do to workaround the BIOS braindamage
> > > > that seems to be rampant (until M$ Vista comes out and supports MSI).
> > > 
> > > OK, what about a patch like below?
> > > 
> > > It's against the latest ALSA tree, so please pull alsa.git from
> > > 	git://kernel.org/pub/scm/linux/kernel/git/perex/alsa.git
> > > before applying it (this tree includes only patches to be pushed to
> > > 2.6.19).
> > > 
> > > 
> > > Takashi
> > > 
> > 
> > Don't you need to check return value from pci_enable_msi()?
> 
> Yes, it would be better to check the value and reset chip->msi if
> not successful.  But it's not a fatal error, so the current code
> should work.

The below is the revised patch.


Takashi

diff -r 74a0618b5f3e sound/pci/hda/hda_intel.c
--- a/sound/pci/hda/hda_intel.c	Wed Oct 18 19:09:46 2006 +0200
+++ b/sound/pci/hda/hda_intel.c	Wed Oct 18 19:17:48 2006 +0200
@@ -337,6 +337,7 @@ struct azx {
 	unsigned int initialized :1;
 	unsigned int single_cmd :1;
 	unsigned int polling_mode :1;
+	unsigned int msi :1;
 };
 
 /* driver types */
@@ -397,6 +398,7 @@ static char *driver_short_names[] __devi
  */
 #define upper_32bit(addr) (sizeof(addr) > 4 ? (u32)((addr) >> 32) : (u32)0)
 
+static int azx_acquire_irq(struct azx *chip, int do_disconnect);
 
 /*
  * Interface for HD codec
@@ -535,6 +537,18 @@ static unsigned int azx_rirb_get_respons
 			return chip->rirb.res; /* the last value */
 		schedule_timeout_interruptible(1);
 	} while (time_after_eq(timeout, jiffies));
+
+	if (chip->msi) {
+		snd_printk(KERN_WARNING "hda_intel: No response from codec, "
+			   "disabling MSI...\n");
+		free_irq(chip->irq, chip);
+		chip->irq = -1;
+		pci_disable_msi(chip->pci);
+		chip->msi = 0;
+		if (azx_acquire_irq(chip, 1) < 0)
+			return -1;
+		goto again;
+	}
 
 	if (!chip->polling_mode) {
 		snd_printk(KERN_WARNING "hda_intel: azx_get_response timeout, "
@@ -1364,6 +1378,20 @@ static int __devinit azx_init_stream(str
 	return 0;
 }
 
+static int azx_acquire_irq(struct azx *chip, int do_disconnect)
+{
+	if (request_irq(chip->pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
+			"HDA Intel", chip)) {
+		printk(KERN_ERR "hda-intel: unable to grab IRQ %d, "
+		       "disabling device\n", chip->pci->irq);
+		if (do_disconnect)
+			snd_card_disconnect(chip->card);
+		return -1;
+	}
+	chip->irq = chip->pci->irq;
+	return 0;
+}
+
 
 #ifdef CONFIG_PM
 /*
@@ -1385,7 +1413,7 @@ static int azx_suspend(struct pci_dev *p
 		free_irq(chip->irq, chip);
 		chip->irq = -1;
 	}
-	if (!disable_msi)
+	if (chip->msi)
 		pci_disable_msi(chip->pci);
 	pci_disable_device(pci);
 	pci_save_state(pci);
@@ -1407,16 +1435,11 @@ static int azx_resume(struct pci_dev *pc
 		return -EIO;
 	}
 	pci_set_master(pci);
-	if (!disable_msi)
-		pci_enable_msi(pci);
-	if (request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
-			"HDA Intel", chip)) {
-		printk(KERN_ERR "hda-intel: unable to grab IRQ %d, "
-		       "disabling device\n", pci->irq);
-		snd_card_disconnect(card);
+	if (chip->msi)
+		if (pci_enable_msi(pci) < 0)
+			chip->msi = 0;
+	if (azx_acquire_irq(chip, 1) < 0)
 		return -EIO;
-	}
-	chip->irq = pci->irq;
 	azx_init_chip(chip);
 	snd_hda_resume(chip->bus);
 	snd_power_change_state(card, SNDRV_CTL_POWER_D0);
@@ -1452,7 +1475,7 @@ static int azx_free(struct azx *chip)
 		synchronize_irq(chip->irq);
 		free_irq(chip->irq, (void*)chip);
 	}
-	if (!disable_msi)
+	if (chip->msi)
 		pci_disable_msi(chip->pci);
 	if (chip->remap_addr)
 		iounmap(chip->remap_addr);
@@ -1508,6 +1531,7 @@ static int __devinit azx_create(struct s
 	chip->pci = pci;
 	chip->irq = -1;
 	chip->driver_type = driver_type;
+	chip->msi = !disable_msi;
 
 	chip->position_fix = position_fix;
 	chip->single_cmd = single_cmd;
@@ -1537,16 +1561,14 @@ static int __devinit azx_create(struct s
 		goto errout;
 	}
 
-	if (!disable_msi)
-		pci_enable_msi(pci);
-
-	if (request_irq(pci->irq, azx_interrupt, IRQF_DISABLED|IRQF_SHARED,
-			"HDA Intel", (void*)chip)) {
-		snd_printk(KERN_ERR SFX "unable to grab IRQ %d\n", pci->irq);
+	if (chip->msi)
+		if (pci_enable_msi(pci) < 0)
+			chip->msi = 0;
+
+	if (azx_acquire_irq(chip, 0) < 0) {
 		err = -EBUSY;
 		goto errout;
 	}
-	chip->irq = pci->irq;
 
 	pci_set_master(pci);
 	synchronize_irq(chip->irq);

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Prakash Punnoor]

[-- Attachment #1: Type: text/plain, Size: 443 bytes --]

Am Mittwoch 18 Oktober 2006 19:21 schrieb Takashi Iwai:

> > Yes, it would be better to check the value and reset chip->msi if
> > not successful.  But it's not a fatal error, so the current code
> > should work.
>
> The below is the revised patch.

I tried it and it works fine for me now (with the driver not using msi 
automatically now).

Cheers,
-- 
(°=                 =°)
//\ Prakash Punnoor /\\
V_/                 \_V

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Takashi Iwai]

At Sun, 22 Oct 2006 22:29:13 +0200,
Prakash Punnoor wrote:
> 
> Am Mittwoch 18 Oktober 2006 19:21 schrieb Takashi Iwai:
> 
> > > Yes, it would be better to check the value and reset chip->msi if
> > > not successful.  But it's not a fatal error, so the current code
> > > should work.
> >
> > The below is the revised patch.
> 
> I tried it and it works fine for me now (with the driver not using msi 
> automatically now).

Thanks for checking.  I applied the patch to ALSA tree for the next
push round.


Takashi

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Fatih Asici]

On 10/23/06, Takashi Iwai <tiwai@suse.de> wrote:
> At Sun, 22 Oct 2006 22:29:13 +0200,
> Prakash Punnoor wrote:
> >
> > Am Mittwoch 18 Oktober 2006 19:21 schrieb Takashi Iwai:
> >
> > > > Yes, it would be better to check the value and reset chip->msi if
> > > > not successful.  But it's not a fatal error, so the current code
> > > > should work.
> > >
> > > The below is the revised patch.
> >
> > I tried it and it works fine for me now (with the driver not using msi
> > automatically now).
>
> Thanks for checking.  I applied the patch to ALSA tree for the next
> push round.
>

It does not solve my problem. I still need to boot with pci=nomsi option.

Prakash: did you use Takashi's patch or Andrian's patch?

Re: [Alsa-devel] [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Prakash Punnoor]

[-- Attachment #1: Type: text/plain, Size: 1462 bytes --]

Am Dienstag 31 Oktober 2006 07:56 schrieb Fatih Asici:
> On 10/23/06, Takashi Iwai <tiwai@suse.de> wrote:
> > At Sun, 22 Oct 2006 22:29:13 +0200,
> >
> > Prakash Punnoor wrote:
> > > Am Mittwoch 18 Oktober 2006 19:21 schrieb Takashi Iwai:
> > > > > Yes, it would be better to check the value and reset chip->msi if
> > > > > not successful.  But it's not a fatal error, so the current code
> > > > > should work.
> > > >
> > > > The below is the revised patch.
> > >
> > > I tried it and it works fine for me now (with the driver not using msi
> > > automatically now).
> >
> > Thanks for checking.  I applied the patch to ALSA tree for the next
> > push round.
>
> It does not solve my problem. I still need to boot with pci=nomsi option.
>
> Prakash: did you use Takashi's patch or Andrian's patch?

I used Takashi's revised patch, but rc3 works w/o patching:

dvanced Linux Sound Architecture Driver Version 1.0.13 (Sun Oct 22 08:56:16 
2006 UTC).
ACPI: PCI Interrupt Link [AAZA] enabled at IRQ 23
ACPI: PCI Interrupt 0000:00:10.1[B] -> Link [AAZA] -> GSI 23 (level, low) -> 
IRQ 23
PCI: Setting latency timer of device 0000:00:10.1 to 64
input: ImPS/2 Generic Wheel Mouse as /class/input/input1
hda_intel: No response from codec, disabling MSI...
ALSA device list:
  #0: HDA NVidia at 0xfe028000 irq 316

And it disables MSI automatically.
-- 
(°=                 =°)
//\ Prakash Punnoor /\\
V_/                 \_V

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [RFC: 2.6.19 patch] snd-hda-intel: default MSI to off

[Fatih Asici]

On 10/18/06, Adrian Bunk <bunk@stusta.de> wrote:
> On Thu, Oct 05, 2006 at 11:08:57PM +0200, Prakash Punnoor wrote:
> > Am Donnerstag 05 Oktober 2006 19:30 schrieb Fatih A????c??:
> > > 2006/10/5, Prakash Punnoor <prakash@punnoor.de>:
> > > > Hi,
> > > >
> > > > subjects say it all. Without irqpoll my nic doesn't work anymore. I added
> > > > Ingo
> > > > to cc, as my IRQs look different, so it may be a prob of APIC routing or
> > > > the
> > > > like.
> >
> > > > Can you try booting with pci=nomsi ? I have a similar problem with my
> >
> > I used snd-hda-intel.disable_msi=1 and this actually helped! Now the nforce
> > nic works w/o problems. So it was the audio driver causing havoc on the nic.
> >...
>
> Unless someone finds and fixes what causes such problems, I'd therefore
> suggest the patch below to let MSI support to be turned off by default.
>

What about putting MCP51 to the msi quirk list?


diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
index 23b599d..75311df 100644
--- a/drivers/pci/quirks.c
+++ b/drivers/pci/quirks.c
@@ -1716,6 +1716,7 @@ static void __devinit quirk_disable_msi(
       }
 }
 DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD,
PCI_DEVICE_ID_AMD_8131_BRIDGE, quirk_disable_msi);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_NVIDIA,
PCI_DEVICE_ID_NVIDIA_NFORCE_MCP51_BRIDGE, quirk_disable_msi);

 /* Go through the list of Hypertransport capabilities and
 * return 1 if a HT MSI capability is found and enabled */
diff --git a/include/linux/pci_ids.h b/include/linux/pci_ids.h
index f069df2..03cb469 100644
--- a/include/linux/pci_ids.h
+++ b/include/linux/pci_ids.h
@@ -1146,6 +1146,7 @@ #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP5
 #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP51_IDE  0x0265
 #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP51_SATA 0x0266
 #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP51_SATA2        0x0267
+#define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP51_BRIDGE       0x026F
 #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP55_SMBUS        0x0368
 #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP55_IDE  0x036E
 #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP55_SATA 0x037E