Skip to contentSkip to conversation

Kernel 2.6 pm_send_all() issues.

Swsusp-devel Michael Frank 4 participants 4 messages Feb 10, 2004 – Feb 12, 2004

lore
Share & tools
lore
#0 lore raw

This question is probably better suited for LKML.

---------- Forwarded Message ----------

Subject: [Swsusp-devel] Kernel 2.6 pm_send_all() issues.
Date: Tuesday 10 February 2004 04:06
From: Wolfgang Glas <[email protected]>
To: [email protected]

Coming back to an earlier post, which states, that NVidia suspend/resume under 2.6.x could be resurrected, if drivers_suspend() and drivers_resume() in swsusp.c are tuned in a way, that pm_send_all() is called, I want to direct the following question to more eligible persons in this list:

I inspected NVidia's driver and I found out, that it's implemented as a
character device and hence does not implement the suspend/resume kernel driver interface, which has been introduced in linux-2.6. But nevertheless this driver has the ability to register its own power management handler through pm_register(), an interface already present in linux-2.4 and still used in some 20 drivers present in the linux-2.6.1 code base, most notable some audio driver and some irda drivers.

Diving further into the code, I recognized, that the kernel function
ppm_send_all(), which calls all handlers registered through pm_register() is never called inside of swsusp2.c

So, what I want to ask is, whether the pm_register/pm_send/pm_send_all
interface is simply deprecated and the NVidia driver (and other drivers, which use pm_register) should be reeingineered in order to use the resume/suspend interface of linux-2.6. Or should we try to modify swsusp2.c in a way, that it additionally calls pm_send_all for these drivers?

TIA,

Wolfgang

-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
swsusp-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/swsusp-devel

-------------------------------------------------------

#1 lore raw
Quoted 1 line · Michael Frank

This question is probably better suited for LKML.

Get nVidia to fix their drivers. pm_send_all() stuff is deprecated and isn't called on purpose. Some of the handlers for these old-style calls were doing weird stuff I'd rather not do again ;)

Ben.

#2 lore raw
In reply to Michael Frank
Quoted 17 lines · Michael Frank

This question is probably better suited for LKML.

---------- Forwarded Message ----------

Subject: [Swsusp-devel] Kernel 2.6 pm_send_all() issues.
Date: Tuesday 10 February 2004 04:06
From: Wolfgang Glas <[email protected]>
To: [email protected]

Coming back to an earlier post, which states, that NVidia suspend/resume under 2.6.x could be resurrected, if drivers_suspend() and drivers_resume() in swsusp.c are tuned in a way, that pm_send_all() is called, I want to direct the following question to more eligible persons in this list:

I inspected NVidia's driver and I found out, that it's implemented as a
character device and hence does not implement the suspend/resume kernel driver interface, which has been introduced in linux-2.6. But nevertheless this driver has the ability to register its own power management handler through pm_register(), an interface already present in linux-2.4 and still used in some 20 drivers present in the linux-2.6.1 code base, most notable some audio driver and some irda drivers.

Diving further into the code, I recognized, that the kernel function
ppm_send_all(), which calls all handlers registered through pm_register() is never called inside of swsusp2.c

Quoted 6 lines

So, what I want to ask is, whether the pm_register/pm_send/pm_send_all
interface is simply deprecated and the NVidia driver (and other drivers, which use pm_register) should be reeingineered in order to use the resume/suspend interface of linux-2.6. Or should we try to modify swsusp2.c in a way, that it additionally calls pm_send_all for these drivers?

TIA,

Wolfgang

I sent them a patch that converts the PCI device to the device model and uses suspend/resume hooks. But they didn't seem to put it in the current version

#3 lore raw
Quoted 2 lines · Stephen Hemminger

I sent them a patch that converts the PCI device to the device model and uses suspend/resume hooks. But they didn't seem to put it in the current version

Could you make the patch publicly available somewhere?

I'm currently using swsusp on 2.6.2 with the XFree86 4.3.0 nv driver which sort of works, but it lacks 3D of course.

Signature · 4 lines
                                     ______
                                    (Muuuhh)
Global Village Sau  ==>        ^..^ |/¯¯¯¯¯
(Kann Fremdsprache) ==>        (oo)