Re: [BUG 2.6.15-git5] new alsa power management completly broken

Re: [BUG 2.6.15-git5] new alsa power management completly broken

[Pavel Machek]

On St 11-01-06 23:48:10, Andrew Morton wrote:
> 
> swsusp problems...

Where? :-)

> sorry for delay, however i tested it and works perfectly.
> 
> just 2 things to notice:
> 1) [unrelated to this bug] swsusp ram pages write and read is really
> slower than 2.6.14 one. i didn't follow changes, so don't know what
> happened

That's a feature. echo 0 > /sys/power/image_size should make it behave
like 2.6.14.

> 2) when i suspend i continue to see errors 0x660 in my tty. The strange
> thing is that after resume they are no more in dmesg!
> strange. however, as i wrote before, it works.

That's not so strange. Messages that happen after memory snapshot are
lost (of course).
								Pavel
-- 
Thanks, Sharp!

Re: [BUG 2.6.15-git5] new alsa power management completly broken

[Patrizio Bassi]

Pavel Machek ha scritto:

>On St 11-01-06 23:48:10, Andrew Morton wrote:
>  
>
>>swsusp problems...
>>    
>>
>
>Where? :-)
>  
>
nowhere now :)
i still have problems on my noteboot about irq 14 not cared...(since ages!)
i'll try to do something, if i can't fix i'll report.
(stopping here cause going OT)

>  
>
>>sorry for delay, however i tested it and works perfectly.
>>
>>just 2 things to notice:
>>1) [unrelated to this bug] swsusp ram pages write and read is really
>>slower than 2.6.14 one. i didn't follow changes, so don't know what
>>happened
>>    
>>
>
>That's a feature. echo 0 > /sys/power/image_size should make it behave
>like 2.6.14.
>
>  
>
ok, so i think i should check docs (i hope they are updated)

>>2) when i suspend i continue to see errors 0x660 in my tty. The strange
>>thing is that after resume they are no more in dmesg!
>>strange. however, as i wrote before, it works.
>>    
>>
>
>That's not so strange. Messages that happen after memory snapshot are
>lost (of course).
>								Pavel
>  
>
ok, sure, but before this patch i had them in my dmesg.
the patch shouldn't have changed the saving sequence, so..why this happens?
and, however some errors still occour.

alsa guys should judge if they are revelant to trace and fix the source
of 0x660.

Thanks for support

--
Patrizio Bassi
www.patriziobassi.it

[BUG 2.6.15-git5] new alsa power management completly broken

[Patrizio Bassi]

With 2.6.15: suspend works but on resume mixer volumes are not restored
(registers not saved), so you need to restart your alsa script.
boring but working.

with 2.6.15-git5 i saw a new power management part. lt's test it:

suspending...

Jan 10 08:44:51 blight kernel: Stopping tasks: ===================|
Jan 10 08:44:51 blight kernel: Shrinking memory...  ^H-^Hdone (0 pages
freed)
Jan 10 08:44:51 blight kernel: pnp: Device 00:0b disabled.
Jan 10 08:44:51 blight kernel: pnp: Device 00:07 disabled.
Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
0000:00:0b.0 disabled
Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
0000:00:0a.2 disabled
Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
0000:00:0a.1 disabled
Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
0000:00:0a.0 disabled
Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
0000:00:09.0 disabled
Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
0000:00:04.2 disabled
Jan 10 08:44:51 blight kernel: swsusp: Need to copy 15888 pages
Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:04.2[D] ->
Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:09.0[A] ->
Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:09.0[A] ->
Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0a.0[A] ->
Link [LNKC] -> GSI 5 (level, low) -> IRQ 5
Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0a.1[B] ->
Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0a.2[C] ->
Link [LNKA] -> GSI 11 (level, low) -> IRQ 11
Jan 10 08:44:51 blight kernel: usb usb1: root hub lost power or was reset
Jan 10 08:44:51 blight kernel: ehci_hcd 0000:00:0a.2: USB 2.0 started,
EHCI 0.95, driver 10 Dec 2004
Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0b.0[A] ->
Link [LNKB] -> GSI 10 (level, low) -> IRQ 10
Jan 10 08:44:51 blight kernel: codec write timeout, status = 0x660
Jan 10 08:44:51 blight last message repeated 25 times

25 times....about half on suspending and about half on restarting.


now restarting..

Jan 10 08:44:51 blight kernel: pnp: Device 00:07 activated.
Jan 10 08:44:51 blight kernel: pnp: Failed to activate device 00:0a.
Jan 10 08:44:51 blight kernel: pnp: Device 00:0b activated.
Jan 10 08:44:51 blight kernel: Restarting tasks... done

device can't play any sound, as before. so i tried the usual solution
(restart the alsasound script in gentoo)

Jan 10 08:45:45 blight rc-scripts: WARNING:  you are stopping a boot
service.
Jan 10 08:45:47 blight kernel: codec write timeout, status = 0x660
Jan 10 08:46:33 blight last message repeated 64 times
Jan 10 08:46:57 blight last message repeated 199 times

as you can see it flood the syslog with 0x660 errors.
infact if i try to adjust the volume via alsamixer the volume level
moves soooo slow (it's flooding syslog)

so actually power resume is completly broken. the only way is restarting
the kernel via a reboot (built-in alsa).


Audio card: Creative Sound Blaster PCI 128
00:0b.0 Multimedia audio controller: Ensoniq ES1370 [AudioPCI] (rev 01)

Ready to test patches.

Thanks.

--
Patrizio Bassi

www.patriziobassi.it

Re: [BUG 2.6.15-git5] new alsa power management completly broken

[Takashi Iwai]

At Tue, 10 Jan 2006 09:06:37 +0100,
Patrizio Bassi wrote:
> 
> With 2.6.15: suspend works but on resume mixer volumes are not restored
> (registers not saved), so you need to restart your alsa script.
> boring but working.

The driver didn't support PM at all on 2.6.15.  It worked casually in
your case with a pure luck :)

> with 2.6.15-git5 i saw a new power management part. lt's test it:
> 
> suspending...
> 
> Jan 10 08:44:51 blight kernel: Stopping tasks: ===================|
> Jan 10 08:44:51 blight kernel: Shrinking memory...  ^H-^Hdone (0 pages
> freed)
> Jan 10 08:44:51 blight kernel: pnp: Device 00:0b disabled.
> Jan 10 08:44:51 blight kernel: pnp: Device 00:07 disabled.
> Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
> 0000:00:0b.0 disabled
> Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
> 0000:00:0a.2 disabled
> Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
> 0000:00:0a.1 disabled
> Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
> 0000:00:0a.0 disabled
> Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
> 0000:00:09.0 disabled
> Jan 10 08:44:51 blight kernel: ACPI: PCI interrupt for device
> 0000:00:04.2 disabled
> Jan 10 08:44:51 blight kernel: swsusp: Need to copy 15888 pages
> Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:04.2[D] ->
> Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
> Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:09.0[A] ->
> Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
> Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:09.0[A] ->
> Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
> Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0a.0[A] ->
> Link [LNKC] -> GSI 5 (level, low) -> IRQ 5
> Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0a.1[B] ->
> Link [LNKD] -> GSI 9 (level, low) -> IRQ 9
> Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0a.2[C] ->
> Link [LNKA] -> GSI 11 (level, low) -> IRQ 11
> Jan 10 08:44:51 blight kernel: usb usb1: root hub lost power or was reset
> Jan 10 08:44:51 blight kernel: ehci_hcd 0000:00:0a.2: USB 2.0 started,
> EHCI 0.95, driver 10 Dec 2004
> Jan 10 08:44:51 blight kernel: ACPI: PCI Interrupt 0000:00:0b.0[A] ->
> Link [LNKB] -> GSI 10 (level, low) -> IRQ 10
> Jan 10 08:44:51 blight kernel: codec write timeout, status = 0x660
> Jan 10 08:44:51 blight last message repeated 25 times
> 
> 25 times....about half on suspending and about half on restarting.

It's the same symptom as resume.


> now restarting..
> 
> Jan 10 08:44:51 blight kernel: pnp: Device 00:07 activated.
> Jan 10 08:44:51 blight kernel: pnp: Failed to activate device 00:0a.
> Jan 10 08:44:51 blight kernel: pnp: Device 00:0b activated.
> Jan 10 08:44:51 blight kernel: Restarting tasks... done
> 
> device can't play any sound, as before. so i tried the usual solution
> (restart the alsasound script in gentoo)
> 
> Jan 10 08:45:45 blight rc-scripts: WARNING:  you are stopping a boot
> service.
> Jan 10 08:45:47 blight kernel: codec write timeout, status = 0x660
> Jan 10 08:46:33 blight last message repeated 64 times
> Jan 10 08:46:57 blight last message repeated 199 times
> 
> as you can see it flood the syslog with 0x660 errors.
> infact if i try to adjust the volume via alsamixer the volume level
> moves soooo slow (it's flooding syslog)
> 
> so actually power resume is completly broken. the only way is restarting
> the kernel via a reboot (built-in alsa).
> 
> 
> Audio card: Creative Sound Blaster PCI 128
> 00:0b.0 Multimedia audio controller: Ensoniq ES1370 [AudioPCI] (rev 01)
> 
> Ready to test patches.

The symptom implies that the initialization of codec chip failed
somehow.   Please build the driver as a module and reload it after
resume to check whether it works without reboot.


Takashi

Re: [BUG 2.6.15-git5] new alsa power management completly broken

[Patrizio Bassi]

Takashi Iwai ha scritto:

>At Tue, 10 Jan 2006 13:02:20 +0100,
>Patrizio Bassi wrote:
>  
>
>>ok i'll test asap. (i think tomorrow).
>>
>>i'll do:
>>1) start pc
>>2) load module and play a wav
>>3) suspend without unloading
>>4) resume (and it shouldn't work)
>>5) unload
>>6) reload and try if that works again (should)
>>
>><joke> blame you, after a report of course. :PPP </joke>
>>    
>>
>
>Any news about blaming or blessing? :)
>
>  
>
>>do you want me to add any debug printk? if so, where? (or post me a patch)
>>    
>>
>
>The patch below might help.  Give it a shot.
>
>
>Takashi
>
>--- a/sound/pci/ens1370.c	7 Dec 2005 11:13:55 -0000	1.91
>+++ b/sound/pci/ens1370.c	10 Jan 2006 16:41:08 -0000
>@@ -2061,6 +2061,13 @@
> #ifdef CHIP1371	
> 	snd_ac97_suspend(ensoniq->u.es1371.ac97);
> #else
>+	/* try to reset AK4531 */
>+	outw(ES_1370_CODEC_WRITE(AK4531_RESET, 0x02), ES_REG(ensoniq, 1370_CODEC));
>+	inw(ES_REG(ensoniq, 1370_CODEC));
>+	udelay(100);
>+	outw(ES_1370_CODEC_WRITE(AK4531_RESET, 0x03), ES_REG(ensoniq, 1370_CODEC));
>+	inw(ES_REG(ensoniq, 1370_CODEC));
>+	udelay(100);
> 	snd_ak4531_suspend(ensoniq->u.es1370.ak4531);
> #endif	
> 	pci_set_power_state(pci, PCI_D3hot);
>
>  
>

sorry for delay, however i tested it and works perfectly.

just 2 things to notice:
1) [unrelated to this bug] swsusp ram pages write and read is really
slower than 2.6.14 one. i didn't follow changes, so don't know what happened

2) when i suspend i continue to see errors 0x660 in my tty. The strange
thing is that after resume they are no more in dmesg!
strange. however, as i wrote before, it works.

ps. i tested with module only, i'll switch back to built-in solution but
i think there won't be any problems.

big thanks to takashi for patch, please apply.

--
Patrizio Bassi
www.patriziobassi.it