[Alsa-user] Regarding Bug ID 0002475 (Power off shows "acpi_power_off called" and hangs)

[Khelben B. <eye...@gm...>]

Hello.

I post to ask if there is any progress on this bug ?
I saw the bug some time ago but it still exists on 2.6.24-rc2 and it is
annoying to run "rmmod snd-hda-intel" before i need to power off the machine.

My motherboard is not the same as the original poster's.
Mine is the Asus P5AD2 Premium and my codec is the Cmedia CMI9880
I run slackware too like the original poster but it is coincidence
since i reproduced it in debian 32bit and sabayon 64bit too.

I guess not many people have this issue that is why it hasn't been resolved
till now. So i tried to bisect the kernel to see which commit causes it.

I remember that the last working kernel was 2.6.13.
I had the impression that 2.6.14-rc1 introduced the bug so i bisected
the changes from 2.6.13 to 2.6.14-rc1 with no success of course because
the bug was actually introduced in 2.6.14-rc2 (i guess i remembered wrong).

I bisected the changes from 2.6.14-rc1 to 2.6.14-rc2 and git showed the
patch i include as "alsa-faulty-2.6.14-rc1-rc2.diff" to be the cause.
I compiled two kernels. One the vanilla 2.6.14-rc1 and one with this
patch applied and indeed it is the patch that causes the problem.

Then, i tried to remove the changes from the 2.6.24-rc2 i run now.
Some hunks of code didn't exist in 2.6.24-rc2, so i tried what existed.
I tried all the hunks one by one to see if maybe only a part of the
patch breaks the driver and indeed only one hunk is the problematic.

--- hda_intel.c.orig	2007-11-26 11:49:00.000000000 +0200
+++ hda_intel.c	2007-11-26 15:00:37.000000000 +0200
@@ -1004,10 +1004,11 @@
 	/* upper BDL address */
 	azx_sd_writel(azx_dev, SD_BDLPU, upper_32bit(azx_dev->bdl_addr));
 
-	/* enable the position buffer */
-	if (!(azx_readl(chip, DPLBASE) & ICH6_DPLBASE_ENABLE))
-		azx_writel(chip, DPLBASE,
-			   (u32)chip->posbuf.addr |ICH6_DPLBASE_ENABLE);
+        if (chip->position_fix == POS_FIX_POSBUF) {
+	        /* enable the position buffer */
+		if (! (azx_readl(chip, DPLBASE) & ICH6_DPLBASE_ENABLE))
+		        azx_writel(chip, DPLBASE, (u32)chip->posbuf.addr | ICH6_DPLBASE_ENABLE);
+	}
 
 	/* set the interrupt enable bits in the descriptor control register */
 	azx_sd_writel(azx_dev, SD_CTL,

The above is the patch to make poweroff work again in 2.6.24-rc2.
I don't understand what the azx_readl,azx_writel functions do, but
with this patch applied everything works ok.

Can Mr. Takashi Iwai or anyone else that has knowledge about this,
see if this patch i include is correct or not and if it is correct
to include it in the kernel ?

Thank you for your time.