|
Bugzilla – Full Text Bug Listing |
| Summary: | Suspend to disk is not working with NVidia due to AGPGART compiled into kernel | ||
|---|---|---|---|
| Product: | [openSUSE] SUSE LINUX 10.0 | Reporter: | Andras Mantia <amantia> |
| Component: | Kernel | Assignee: | Andreas Kleen <ak> |
| Status: | RESOLVED FIXED | QA Contact: | E-mail List <qa-bugs> |
| Severity: | Normal | ||
| Priority: | P5 - None | CC: | aritger, eich, sndirsch |
| Version: | Final | ||
| Target Milestone: | --- | ||
| Hardware: | x86-64 | ||
| OS: | SuSE Linux 10.0 | ||
| Whiteboard: | |||
| Found By: | Customer | Services Priority: | |
| Business Priority: | Blocker: | --- | |
| Marketing QA Status: | --- | IT Deployment: | --- |
|
Description
Andras Mantia
2006-01-09 09:04:25 UTC
Assigning to the X11 folks. I guess there's little we can do - we need CONFIG_GART_IOMMU in our shipped kernels. I didn't know it is needed, I thought the modules can replace it. The problem with self compiled kernels is when a security update appears, this is why it would be nice to have it working by default. Yes, I know that the real fix should be in the nvidia module, and I sent a mail to them as well, but I don't think they will do it soon (or ever?). The root of this particular problem might be that the kernel agp driver does not have suspend support. Maybe this is a feature that should be added. (Adding Andi for comment). AMD K8 suspend support went into mainline just a few days ago. Probably next release. Is it in some officially released kernel or only in the development tree? Just that I would test if it was officially released. Andy, could you close this one as FIXED once it's done for an official SUSE kernel? Thanks. Hmm, it's actually not in -rc1. Maybe we need to add the patch ourselves. According to the maintainer it's going to be merged soon, just waiting for Linus to come back from New Zealand. Should be fixed in kotd now (= for beta4) Just for the record:it was not fixed out of the box even in beta9. But the new NVidia driver came out, and that one works with beta9. So you need at least the 8756 binary driver with NVAGP enabled. Aargh, it worked that way, but I realized that AGP is disabled due to the same reasons as I wrote in the original report. I switched to use the system AGPGART, but this was suspend2disk does not work. I played with removing IOMMU from the kernel and using NvAGP=1, and it did not worked. Now it doesn't even work with IOMMU in the kernel and a disabled AGP. Certainly, the bug doesn't seem to be fixed in the recent kernel. So, until now I found: - NvAGP="1": suspend 2 disk works, AGP becomes disabled as AGPGART is already loaded - NvAGP="2": suspend 2 disk does not work (after resume, the monitor is switched off, if I suspend in text mode, it resumes, but the monitor is switched off as soon as I switch to the graphical session), AGP works via AGPGART. After resume, AGP becomes disabled Tomorrow I will try NVAGP="1" without IOMMU compiled in the kernel, maybe what I describe below will help in that case as well. Note: Resume fails if I don't specify SUSPEND2DISK_RESUME_DEVICE="/dev/hda6" in /etc/sysconfig/powersave/sleep, even if resume=/dev/hda6 is passed to the kernel. The kernel prints "Attempting manual resume" and boots without resuming and tries to fix the file system. After a while it hangs. (In reply to comment #13) > Note: Resume fails if I don't specify SUSPEND2DISK_RESUME_DEVICE="/dev/hda6" > in /etc/sysconfig/powersave/sleep, even if resume=/dev/hda6 is passed to the > kernel. The kernel prints "Attempting manual resume" and boots without resuming > and tries to fix the file system. After a while it hangs. This is strange, but another bug, so please file a separate bug against 10.1beta9, component "Mobile devices". |