[ltp] Mails from ATI about the Suspend-To-Ram problem with fglrx drivers

Sat, 09 Apr 2005 02:23:38 -0400

Nils wrote:

>I've removed the headers and you can read mails between me and ATI.
>You can read this long mail, but you won't find a solution except waiting :-)
>
>
>1.) To ATI
>Linux Driver Feedback
>Details:  I've seen that it is a problem to resume a linux computer from 
>suspend mode with ATI fglrx drivers. (FireGL T2) The problem is that I wait 
>over a year that ati solves this really very big problem, because who needs a 
>driver for a notebook which can't resume ??? Everbody can make mistakes, but 
>this no joke anymore. Please release better driver or many people are so 
>angry, that they'll never buy ATI products in the future and I am one of 
>them.
>I hope I'll get an answer to solve this issue,
>Nils
>
>[Or release the driver under GPL and it would be the best graphic driver of 
>the world in some weeks :-) ]
>
>
>2.) From ATI
>Sleep mode settings is not controlled by display driver - it is controlled by 
>the Linux configuration tools available in your particular distro.
>
>This is a laptop as well, you should set up the fglrxconfig using laptop mode.
>
>Regards,
>
>
>3.) To ATI
>I think my describtion of the problem was not clear enough. My laptop works 
>fine with suspend to ram (or disk) if I use the opensource driver radeon 
>(Xfree / Xorg), but these drivers don't support 3D acceleration.
>That is the reason why I want to use the firegl driver from ATI. The laptop 
>also works fine, it suspends without problems, but then I wake it up and my 
>screen is garbled and I've to reboot it. (really really bad ...)
>
>On the ATI support pages I found this: (for over a year!)
>( http://www2.ati.com/drivers/linux/linux_8.10.19.html
>Attempting to resume from system suspension results in the system failing to 
>respond. )
>
>Is ATI working to fix this issue or should it be fixed and working ? But I 
>don't know anybody of some hundred linux users.
>
>
>4.) From ATI
>Laptops hang on suspend/resume mode
>
>This issue affects the following system configurations:
>
>ATI Proprietary Linux driver
>Laptop system with ATI display adapter
>
>Suspending the system works correctly.  Resuming from system suspension 
>results in a system hang.
>
>ATI engineering is aware of this issue and is investigating.
>
>This issue is under investigation.  There is no solution available at this 
>time.
>
>I still believe that changing some of the settings on the fglrxconfig may help 
>you fix this issue.
>
>Have tried the external agpgart?
>
>Regards,
>
>
>5.) To ATI
>Hi,
>I tried external agp-gart.
>
>lsmod | grep agp
>intel_agp 20060 1
>agpgart 28840 1 intel_agp
>
>In XF86Config.4 :
>(...)
>Option "UseInternalAGPGART" "no"
>(...)
>
>But I have no success with it. After resuming the laptop I see only the 
>garbled screen and I have to reboot it.
>If I deactivate DRI it is perfectly working, but then I have no 3D 
>acceleration. Then it makes no sense to use the ati driver!
>I can't debug the driver, because it is closed source. I think ATI is big 
>enough to check the driver on any IBM Thinkpad with a fireGL card.
>It seems to be a problem, because every thinkpad user I know has the same 
>problem.
>
>I still hope for fix,
>
>
>6.) From ATI
>It could be a Bios issue like the one in the ACER,
>
>http://localhost.ruhr.de/~stefan/acerTM292/
>
>In this article he also presentes other solutions that may work for you, like 
>the following,
>
>"Software Suspend 2" is a kernel patch that allows you to suspend your system 
>by saving the memory content into a swap partition and powering down. When 
>booting the system again, the kernel recognizes the frozen image, und 
>restores he former state of the notebook. The use of Software Suspend 
>requires some prerequisites:
>
>    * a running kernel with "Software Suspend 2" patched in, e.g. the 
>kernel-image-2.6.10-susp-tm292 available here (future packages may not carry 
>the additional "-susp" suffix)
>    * a swap partition (swap files do not work) large enough to fit your RAM 
>content into
>    * the debian package hibernate, which supplies a script that takes care of 
>nasty hard- and software during suspension. 
>
>Once the kernel is running, you have to modify your bootloader (lilo or grub, 
>preferred the later) to include the additional commandline statement 
>"resume2=swap:/dev/hda6", where /dev/hda6 is the swap partition you wish to 
>resume to. However, the ATI drivers are breaking Software Suspend, so you 
>have to quit X11 before hibernating your notebook: My hibernate.conf does 
>this automatically by shutting down gdm, it as well restores acpi and acerhk 
>drivers after resume. Just copy it to /etc/hibernate/. If you want to suspend 
>the system, just su to root and enter "hibernate".
>
>If you do not need DRI and OpenGL acceleration, you can use the open source 
>"radeon" driver. Software Suspend 2 works fine with this driver, even without 
>quitting X11. Just replace the driver "fglrx" with "radeon" in you 
>XF86Config-4, and resume "gdm" from the "RestartServices" line of 
>"hibernate.conf". If you want to switch between the two configurations, use 
>the "switch-xconfig.sh" script.
>
>Regards,
>
>
>7.) To ATI
>Thanks for the detailed answer!
>
>I have tried several thinks:
>working:
>-> Suspend-To-Disk is working (with software suspend / bios suspend)
>-> Suspend-To-Ram is working if I close the X Server before.
>
>not working:
>-> Suspend-To-Ram with DRI and a running X Server and the ATI Driver. (garbled 
>screen)
>(I like Suspend-To-Ram because the laptop needs only 2 seconds and I can work)
>
>I have only one more question, where is the problem that the last point is not 
>working. Is it kernel related or a problem with the fglrx binary or the 
>opengl / dri implementation?
>
>Regards,
>
>
>8.) From ATI
>I do believe that must be Kernel or Hardware related.
>
>If the problem was the driver it would happen always in every machine.
>
>Look at the following article, it is great,
>
>http://www.stanford.edu/~sanjiv/thinkpad/thinkpad.html
>
>Regards,
>
>
>9.) To ATI
>thanks for the link
>http://www.stanford.edu/~sanjiv/thinkpad/thinkpad.html
>Okay he has no garbled screen, but he has to reboot his X-Server, because of a 
>slowdown. With XOrg he doesn't use the ati proprietary driver. (the same 
>problem, no 3D ...)
>
>You think it's a problem with the kernel or hardware. That's the reason why I 
>am searching for people who has a working configuration with DRI/ATI 
>proprietary driver/XFree or Xorg/Suspend-To-Ram/running X-Server. I have 
>spooken to more than 20 people (okay, it's not much enough) and it seems that 
>nobody has a working configuration. The only part I can't check is the driver 
>and I think there is a memory problem in it. (Simple refresh ?)
>You can help me very much, if you say that you know somebody with such a 
>working configuration. (ATI card / model / config)
>(I know that there exists work-arounds - e.g. restarting the X-Server is no 
>solution for me.)
>
>Is there any chance that ati can help to support the following project ? 
>(little hints / a little bit documentation / code fragments)
>http://r300.sourceforge.net/
>
>Regards,
>
>
>10.) From ATI
>That is all the help I can give you. The Linux team is always working on 
>improving our driver.
>
>You have to Understant that with So many Kernel versions being released and 
>used it is very difficult to keep up. And that applies to all hardware 
>companies.
>
>That is why I just use the Stable kernels.
>
>As for the Mesa and Gatos projects ATI does support such a projects.
>
>Dear ---
>
>I do believe that must be Kernel or Hardware related.
>
>If the problem was the driver it would happen always in every machine.
>
>Look at the following article, it is great,
>
>http://www.stanford.edu/~sanjiv/thinkpad/thinkpad.html
>
>Regards,
>
>
>11.) To ATI
>
>thanks for your help. I will waiting for the next versions of the fglrx driver 
>and xserver improvements.
>
>
>This page is interesting: (and exactly correct !!)
>http://www.doesi.gmxhome.de/linux/tm800s3/s3.html
>(see -> Informal: ATI's binary driver / Old Alternative 1-3)
>The kernel developer says that the drivers have to reinitialize the devices. 
>That's no problem with drivers which are included in the kernel, but external 
>drivers have to do the same, but they can't fix this.
>
>
>
>----
>I hope they will fix their drivers ...
>
>  
>
What has been your makeshift solution in the meanwhile.  I am switching 
to ACPI from APM because I was getting random crashing and didn't like 
the power usage from APM when I knew it should a lot better.  So far 
today in searching these forums, the only solution I see that still uses 
fglrx is to kill the X server prior to suspending, followed by starting 
it right after resuming, is this the case, or do you have a different 
makeshift solution?

Mario