Test patch for sleep on Aluminium PowerBooks

Test patch for sleep on Aluminium PowerBooks

[Benjamin Herrenschmidt]

[-- Attachment #1: Type: text/plain, Size: 413 bytes --]

Hi !

This is an experimental patch against 2.6.9. It concerns the ATI based
Aluminium PowerBook. The nVidia based ones aren't concerned at all, and
the iBook G4 will have to wait a bit more for me to get the video wakeup
code right.

It will add sleep (suspend-to-ram) support for these machines.

It's based on various tested bits but is untested on these machines in
it's current form, so let me know.

Ben.



[-- Attachment #2: albook-sleep.diff.bz2 --]
[-- Type: application/x-bzip, Size: 22275 bytes --]

Re: Test patch for sleep on Aluminium PowerBooks

[David Woodhouse]

[-- Attachment #1: Type: text/plain, Size: 571 bytes --]

On Wed, 2004-10-27 at 19:50 +1000, Benjamin Herrenschmidt wrote:
> Hi !
> 
> This is an experimental patch against 2.6.9. It concerns the ATI based
> Aluminium PowerBook. The nVidia based ones aren't concerned at all, and
> the iBook G4 will have to wait a bit more for me to get the video wakeup
> code right.

Works for me, thanks. Works better if I have the attached patch too. 

Xorg doesn't seem to receive mouse events from the USB mouse after
resume, but 'od -t x1 /dev/input/mice' does continue to work after
resume. I'll investigate that one further.

-- 
dwmw2

[-- Attachment #2: linux-2.6.9-usbsuspend.patch --]
[-- Type: text/x-patch, Size: 618 bytes --]

--- linux-2.6.9/drivers/usb/core/hub.c.susp	2004-10-27 17:30:39.796999176 +0100
+++ linux-2.6.9/drivers/usb/core/hub.c	2004-10-27 17:31:49.991979112 +0100
@@ -1527,12 +1527,12 @@ static int __usb_suspend_device (struct 
 			 * even for drivers that can't suspend.
 			 */
 			if (!driver->suspend || state > PM_SUSPEND_MEM) {
-#if 1
+#if 0
 				dev_warn(&intf->dev, "resume is unsafe!\n");
 #else
-				down_write(&usb_bus_type.rwsem);
+				down_write(&usb_bus_type.subsys.rwsem);
 				device_release_driver(&intf->dev);
-				up_write(&usb_bus_type.rwsem);
+				up_write(&usb_bus_type.subsys.rwsem);
 #endif
 			}
 		}

Re: Test patch for sleep on Aluminium PowerBooks

[Armando Di Cianno]

[-- Attachment #1.1: Type: text/plain, Size: 1475 bytes --]

On 2004-10-27 05:50:36 -0400 Benjamin Herrenschmidt 
<benh@kernel.crashing.org> wrote:
> This is an experimental patch against 2.6.9. It concerns the ATI based
> Aluminium PowerBook. The nVidia based ones aren't concerned at all, 
> and
> the iBook G4 will have to wait a bit more for me to get the video 
> wakeup
> code right.

Hello all; a little late in responding to this, but I wanted to ask 
some questions concerning possible issues I've been having.  I'm 
running on an AluBook 1.5Ghz, Linux kernel 2.6.9.  I suppose this is 
mainly addressed to Ben H.

Does this patch do, or should have to do, anything special to 
"normalize" the mounted filesystems before sleep?  When I patched the 
kernel first, I was running XFS, and of course that is notorious for 
file corruptions with power abnormalities on a system.  I started to 
get really odd file corruptions, where 'ls file' would should the 
file, but 'ls -l file' would not.

I switched to ReiserFS (v3) now, and have seen symptoms like this 
problem caused appear once more (but not to the extent that XFS showed 
[where I thought my system was disintegrating before my eyes]).

Is there something that was overlooked in the patch, or something that 
I'm overlooking that needs to be done?

If filesystems do not need to be "normalized" at sleep, what 
filesystems are known to work?  (Hesistant about ext3, as it's quite 
slow, imho, but I'll go that route if needed, I suppose).

Thanks!

__Armando Di Cianno

[-- Attachment #1.2: RFC3156.txt --]
[-- Type: text/plain, Size: 65 bytes --]

RFC3156 defines security multipart formats for MIME with OpenPGP.

[-- Attachment #2: Type: application/pgp-signature, Size: 230 bytes --]

Re: Test patch for sleep on Aluminium PowerBooks

[Benjamin Herrenschmidt]

> Does this patch do, or should have to do, anything special to 
> "normalize" the mounted filesystems before sleep?  When I patched the 
> kernel first, I was running XFS, and of course that is notorious for 
> file corruptions with power abnormalities on a system.  I started to 
> get really odd file corruptions, where 'ls file' would should the 
> file, but 'ls -l file' would not.

I though xfs was supposed to be journaled ? Yes, the sleep code will
sync all disks before sleep. You should update to patch #4 though, there
have been number of problems fixes since then.

> I switched to ReiserFS (v3) now, and have seen symptoms like this 
> problem caused appear once more (but not to the extent that XFS showed 
> [where I thought my system was disintegrating before my eyes]).

Hrm... reiserfs is evil, ask any sane kernel hacker :) Why not ext3 ?

> Is there something that was overlooked in the patch, or something that 
> I'm overlooking that needs to be done?
> 
> If filesystems do not need to be "normalized" at sleep, what 
> filesystems are known to work?  (Hesistant about ext3, as it's quite 
> slow, imho, but I'll go that route if needed, I suppose).

It's not that slow ...

Ben.

Re: Test patch for sleep on Aluminium PowerBooks

[Colin Leroy]

On 23 Nov 2004 at 09h11, Benjamin Herrenschmidt wrote:

Hi, 

> > If filesystems do not need to be "normalized" at sleep, what 
> > filesystems are known to work?  (Hesistant about ext3, as it's quite
> > slow, imho, but I'll go that route if needed, I suppose).
> 
> It's not that slow ...

It also has the advantages of being natively big-endian, giving us ppc
users a bit more speed :)

-- 
 \|/ ____ \|/   Colin
 "@'/ ,. \`@"   http://www.geekounet.org/
 /_| \__/ |_\
    \__U_/