Skip to contentSkip to conversation

PROBLEM: snd-au8830: dead after software suspend

DISCUSSION Gert Robben 4 participants 8 messages Aug 3, 2007 – Aug 6, 2007

lore
Share & tools
lore
#0 lore raw

After a suspend-resume cycle (using Suspend2), my sound card doesn't work anymore. It works again after reloading the module.

dmesg, initial boot:

Quoted 1 line

Linux version 2.6.22-ck1

----8<----

Quoted 1 line

PCI driver au8830 lacks driver specific resume support.

dmesg, after resume, trying to use the card:

Quoted 1 line

vortex: ac97 codec stuck busy

Thanks for any help!

Gert Robben

#1 lore raw
Quoted 1 line · Robert Hancock

Have you reported this to the ALSA people?

No, I thought this might as well be something for PCI or PM people, and I expected some ALSA people are also reading this list. If not, where else should I report this exactly?

Gert Robben

#2 lore raw
Quoted 6 lines · Gert Robben
Quoted 1 line · Robert Hancock

Have you reported this to the ALSA people?

No, I thought this might as well be something for PCI or PM people, and I expected some ALSA people are also reading this list.
If not, where else should I report this exactly?

alsa-devel would seem to be a good starting point - see
http://www.alsa-project.org/mailing-lists.php
Also, the author of the driver, Manuel Jander <[email protected]>, would probably be a good person to add to Cc.
And keeping LKML on Cc while sending to above people usually never hurts :-)

Signature · 3 lines
Jesper Juhl <[email protected]>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html
#3 lore raw
In reply to Gert Robben
Quoted 4 lines
Quoted 3 lines
Quoted 1 line

Have you reported this to the ALSA people?

No, I thought this might as well be something for PCI or PM people, and I expected some ALSA people are also reading this list. If not, where else should I report this exactly?

alsa-devel [...] the author of the driver

Done (original message below). Thanks again :)

Quoted 13 lines · Gert Robben

After a suspend-resume cycle (using Suspend2), my sound card doesn't work anymore. It works again after reloading the module.

dmesg, initial boot:

Quoted 1 line

Linux version 2.6.22-ck1

----8<----

Quoted 1 line

PCI driver au8830 lacks driver specific resume support.

dmesg, after resume, trying to use the card:

Quoted 1 line

vortex: ac97 codec stuck busy

Thanks for any help!

Gert Robben

#4 lore raw
Quoted 6 lines · Gert Robben
Quoted 5 lines
Quoted 4 lines
Quoted 1 line

Have you reported this to the ALSA people?

No, I thought this might as well be something for PCI or PM people, and I expected some ALSA people are also reading this list.
If not, where else should I report this exactly?

alsa-devel [...] the author of the driver

Done (original message below). Thanks again :)

You also may want to file a bug at http://bugzilla.kernel.org .

If you decide to do that, please file it as
"Power Management"->Hibernation/Suspend bug, I'll redirect it to the ALSA people.

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#5 lore raw
In reply to Gert Robben

At Sat, 04 Aug 2007 02:51:44 +0200,

Quoted 23 lines · Gert Robben
Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

Have you reported this to the ALSA people?

No, I thought this might as well be something for PCI or PM people, and I expected some ALSA people are also reading this list.
If not, where else should I report this exactly?

alsa-devel [...] the author of the driver

Done (original message below). Thanks again :)

Quoted 14 lines · Gert Robben

After a suspend-resume cycle (using Suspend2), my sound card doesn't work anymore. It works again after reloading the module.

dmesg, initial boot:

Quoted 1 line

Linux version 2.6.22-ck1

----8<----

Quoted 1 line

PCI driver au8830 lacks driver specific resume support.

dmesg, after resume, trying to use the card:

Quoted 1 line

vortex: ac97 codec stuck busy

Thanks for any help!

Gert Robben

The below is an untested fix. Give it a try.

Takashi

pci/au88x0/au88x0.c +53 -3
diff -r 23fc708178e1 pci/au88x0/au88x0.c
@@ -78,7 +78,7 @@ static void vortex_fix_agp_bridge(struct
 	}
 }
 
-static void __devinit snd_vortex_workaround(struct pci_dev *vortex, int fix)
+static void snd_vortex_workaround(struct pci_dev *vortex, int fix)
 {
 	struct pci_dev *via = NULL;
 
@@ -117,6 +117,50 @@ static void __devinit snd_vortex_workaro
 	pci_dev_put(via);
 }
 
+/*
+ * Power management code
+ */
+#ifdef CONFIG_PM
+static int snd_vortex_suspend(struct pci_dev *pci, pm_message_t state)
+{
+	struct snd_card *card = pci_get_drvdata(pci);
+	struct snd_vortex *chip = card->private_data;
+	int i;
+
+	snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
+	for (i = 0; i < VORTEX_PCM_LAST; i++)
+		snd_pcm_suspend_all(chip->pcm[i]);
+	snd_ac97_suspend(chip->codec);
+	vortex_core_shutdown(chip);
+	pci_disable_device(pci);
+	pci_save_state(pci);
+	return 0;
+}
+
+static int snd_vortex_resume(struct pci_dev *pci)
+{
+	struct snd_card *card = pci_get_drvdata(pci);
+	struct snd_vortex *chip = card->private_data;
+
+	pci_set_power_state(pci, PCI_D0);
+	pci_restore_state(pci);
+	if (pci_enable_device(pci) < 0) {
+		printk(KERN_ERR "au88x0: pci_enable_device failed, "
+		       "disabling device\n");
+		snd_card_disconnect(card);
+		return -EIO;
+	}
+	pci_set_master(pci);
+	vortex_core_init(chip, 1);
+	snd_vortex_workaround(pci, chip->pcifix);
+	snd_ac97_resume(chip->codec);
+	vortex_connect_default(chip, 1);
+	vortex_enable_int(chip);
+	snd_power_change_state(card, SNDRV_CTL_POWER_D0);
+	return 0;
+}
+#endif /* CONFIG_PM */
+
 // component-destructor
 // (see "Management of Cards and Components")
 static int snd_vortex_dev_free(struct snd_device *device)
@@ -192,7 +236,7 @@ snd_vortex_create(struct snd_card *card,
 	/* Init audio core.
 	 * This must be done before we do request_irq otherwise we can get spurious
 	 * interupts that we do not handle properly and make a mess of things */
-	if ((err = vortex_core_init(chip)) != 0) {
+	if ((err = vortex_core_init(chip, 0)) != 0) {
 		printk(KERN_ERR "hw core init failed\n");
 		goto core_out;
 	}
@@ -262,7 +306,9 @@ snd_vortex_probe(struct pci_dev *pci, co
 		snd_card_free(card);
 		return err;
 	}
-	snd_vortex_workaround(pci, pcifix[dev]);
+	card->private_data = chip;
+	chip->pcifix = pcifix[dev];
+	snd_vortex_workaround(pci, chip->pcifix);
 
 	// Card details needed in snd_vortex_midi
 	strcpy(card->driver, CARD_NAME_SHORT);
@@ -382,6 +428,10 @@ static struct pci_driver driver = {
 	.id_table = snd_vortex_ids,
 	.probe = snd_vortex_probe,
 	.remove = __devexit_p(snd_vortex_remove),
+#ifdef CONFIG_PM
+	.suspend = snd_vortex_suspend,
+	.resume = snd_vortex_resume,
+#endif
 };
 
 // initialization of the module
pci/au88x0/au88x0.h +3 -2
diff -r 23fc708178e1 pci/au88x0/au88x0.h
@@ -178,7 +178,7 @@ struct snd_vortex {
 	/* PCI hardware resources */
 	unsigned long io;
 	void __iomem *mmio;
-	unsigned int irq;
+	int irq;
 	spinlock_t lock;
 
 	/* PCI device */
@@ -186,6 +186,7 @@ struct snd_vortex {
 	u16 vendor;
 	u16 device;
 	u8 rev;
+	int pcifix;
 };
 
 /* Functions. */
@@ -233,7 +234,7 @@ static unsigned short vortex_codec_read(
 static unsigned short vortex_codec_read(struct snd_ac97 * codec, unsigned short addr);
 static void vortex_spdif_init(vortex_t * vortex, int spdif_sr, int spdif_mode);
 
-static int vortex_core_init(vortex_t * card);
+static int vortex_core_init(vortex_t * card, int do_resume);
 static int vortex_core_shutdown(vortex_t * card);
 static void vortex_enable_int(vortex_t * card);
 static irqreturn_t vortex_interrupt(int irq, void *dev_id);
pci/au88x0/au88x0_a3d.c +1 -1
diff -r 23fc708178e1 pci/au88x0/au88x0_a3d.c
@@ -593,7 +593,7 @@ static int vortex_a3d_register_controls(
 static int vortex_a3d_register_controls(vortex_t * vortex);
 static void vortex_a3d_unregister_controls(vortex_t * vortex);
 /* A3D base support init/shudown */
-static void __devinit vortex_Vort3D_enable(vortex_t * v)
+static void vortex_Vort3D_enable(vortex_t * v)
 {
 	int i;
 
pci/au88x0/au88x0_core.c +3 -2
diff -r 23fc708178e1 pci/au88x0/au88x0_core.c
@@ -2658,7 +2658,7 @@ static void vortex_spdif_init(vortex_t *
 
 /* Initialization */
 
-static int __devinit vortex_core_init(vortex_t * vortex)
+static int vortex_core_init(vortex_t * vortex, int do_resume)
 {
 
 	printk(KERN_INFO "Vortex: init.... ");
@@ -2688,7 +2688,8 @@ static int __devinit vortex_core_init(vo
 	vortex_mixer_init(vortex);
 	vortex_srcblock_init(vortex);
 #ifndef CHIP_AU8820
-	vortex_eq_init(vortex);
+	if (!do_resume)
+		vortex_eq_init(vortex);
 	vortex_spdif_init(vortex, 48000, 1);
 	vortex_Vort3D_enable(vortex);
 #endif
#6 lore raw
Quoted 1 line · Takashi Iwai

The below is an untested fix. Give it a try.

With the patch, things seem to be much better (but I'm no expert).

The things remaining seem to be:
- sound can't be heard
- MPlayer gives an error, instead of doing nothing.

I've enabled debugging, and did the following, without and with the patch:
boot, mplayer start+pause, suspend, resume, mplayer unpause+exit+start+exit, rmmod.

Below is an edited diff of the dmesges (printk's begin with a "k"),
without and with the patch, with my comments added. I hope this is enough info. If you need the full dmesges, let me know.

Anyway, thanks for you work :)

Gert Robben

--- without patch
+++ with patch
initial boot:
k ACPI: PCI Interrupt 0000:00:11.0[A] -> GSI 19 (level, low) -> IRQ 19 k Vortex: init.... <3>ALSA sound/core/control.c:331: control 3:0:0:Play k back PCM advanced processing:0 is already present
k done.
-k PCI driver au8830 lacks driver specific resume support.
during suspend:
+k Vortex: shutdown...<6>done.
+k ACPI: PCI interrupt for device 0000:00:11.0 disabled
during resume:
+k PM: Writing back config space on device 0000:00:11.0 at offset 1 (was
k 2900007, writing 2900003)
k ACPI: PCI Interrupt 0000:00:11.0[A] -> GSI 19 (level, low) -> IRQ 19
+k Vortex: init.... <6>ata1.00: configured for UDMA/33
+k ALSA sound/core/control.c:331: control 3:0:0:Playback PCM advanced pr
k ocessing:0 is already present
+k done.
mplayer unpause:
+- gives mplayer error: alsa-lib: pcm_hw.c:612(snd_pcm_hw_pause) SNDRV_P
CM_IOCTL_PAUSE failed: File descriptor in bad state
- nothing else happens
mplayer exit:
+k vortex_adb_delroutes: route not found! 0x40
+k vortex_adb_delroutes: route not found! 0x2060
k src alarm
+k vortex_adb_delroutes: route not found! 0x41
+k vortex_adb_delroutes: route not found! 0x2161
k src alarm
mplayer start:
-- nothing happens
+- mplayer seems to play fine, but no sound can be heard
rmmod:
-k vortex: ac97 codec stuck busy
-k vortex: ac97 codec stuck busy
-k vortex: ac97 codec stuck busy
-k vortex: ac97 codec stuck busy
k analog.c: 0 out of 0 reads (0%) on pci0000:00:11.0/gameport0 failed
-k Vortex: shutdown...<3>mix ALARM 0
-k mix ALARM 0
-k mix ALARM 0
-k mix ALARM 0
-k mix ALARM 0
-k mix ALARM 0
-k done.
+k Vortex: shutdown...<6>done.
k ACPI: PCI interrupt for device 0000:00:11.0 disabled

#7 lore raw

At Mon, 06 Aug 2007 18:15:27 +0200,

Quoted 9 lines · Gert Robben
Quoted 1 line · Takashi Iwai

The below is an untested fix. Give it a try.

With the patch, things seem to be much better (but I'm no expert).

The things remaining seem to be:
- sound can't be heard
- MPlayer gives an error, instead of doing nothing.

Quoted 8 lines

I've enabled debugging, and did the following, without and with the patch:
boot, mplayer start+pause, suspend, resume, mplayer unpause+exit+start+exit, rmmod.

Below is an edited diff of the dmesges (printk's begin with a "k"), without and with the patch, with my comments added. I hope this is enough info. If you need the full dmesges, let me know.

Hm, the routing inside the driver seems more (too) complicated than I first took a glance. Sorry, I have no time for further digging right now. Hopefully Manuel or any other guy will take over...

Takashi