Bug 55587
| Summary: | Neomagic problems causing by configuration | ||
|---|---|---|---|
| Product: | [Retired] Red Hat Linux | Reporter: | Alan Cox <alan> |
| Component: | Xconfigurator | Assignee: | Mike A. Harris <mharris> |
| Status: | CLOSED RAWHIDE | QA Contact: | David Lawrence <dkl> |
| Severity: | high | Docs Contact: | |
| Priority: | medium | ||
| Version: | 7.2 | CC: | hobbit, notting, sunnychan |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2003-01-24 16:44:36 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
Does this hang require a complete poweroff to properly reset the card? (i.e., does starting the X server again without a cold boot also hang?) Just trying to correlate this with hangs on other neomagic chipsets. Restarting an Xserver until a cold power off also hangs yes "Mid-air collision detected!" Alan beat me to it. On *my* bug. Yes. It does. I have to switch off completely before I reboot. I have added the two magic lines now. Btw, it's much more variable than "every two or three days". Sometimes it's within half an hour of the last one.Sometimes it stays up for nearly a week. OK, so this appears to apply to other Neomagic chipsets, such as: 01:00.0 VGA compatible controller: Neomagic Corporation NM2360 [MagicMedia 256ZX] I have a similar problem with my laptop - It is a IBM Thinkpad 600X, using Neomagic 2360 chipset. The screen is corrupted when I use X under 1024x768x24/32bit - the windows and characters will disintegrate. However with the above two lines it seems to fix the problem. These options are now the default for all neomagic variants. I am removing these options from the hardware database for 2 purposes: 1) To test and see if the bugs are fixed 2) To implement an alternate more robust solution if it is not fixed yet, by either fixing it, getting someone to fix it, or disabling the code directly in the driver source. This way nobody can enable the broken code, and upgrades also work from distro to distro without relying on whatever a user has in their config file before upgrade. Please test the current CVS XFree86 4.2.99.2 driver if at all possible, and remove or comment out all swcursor options, and "XaaNoxxxxxx" options as well as noaccel et al. If a problem occurs, please open a new bug report against rawhide, and perhaps point out this bug as reference also. Just attempting to clean up some of the prior dirty bandaids with cleaner bandaids. Thanks The XFree86 documentation still explicitly states that this is neccessary in 4.2.99 from 8.0.90 Reopened since either the bug or the docs need fixing. Still needs testing to see which I think you misunderstood me a bit... The prior solution was a rather poor bandaid, which has ended up causing some other unexpected problems when people upgrade from older releases to 8.0. What I want to do is remove the bandaid, and replace it with either a real fix, or a better bandaid directly in the driver itself. Basically, if the cursor can't be fixed, I want to disable the section of code for hwcursor in the driver, then we can flush out the various adhoc Cards database hacks. The problem is when someone was using an older driver and not experiencing the problem, then upgrades to 8.0, their config file is not changed, and bang - incoming bug reports. This is a regression for them, and so I'd like to fix the broken code, or if not, disable it, so that users don't have to twiddle config options, or get hurt on upgrades. Sorry for confusion. ;o) Now I know i've got a beta tester though. ;o) <evil grin> Its nothing to do with the cursor, its a hardware bug. Ask Bill for the documentation for the neomagic and the errata docs The docs Bill has are under NDA, and he has not gotten clearance from neomagic to do anything useful with them. I don't want docs under NDA unless I can do something with them personally for open source. So unless they open up the docs, or they fall off the back of a truck, I don't want Bill's copies as they are accounted for. I will go inside the driver guts and disable these two options on the 2160. I'll put that into CVS, and make a patch for 4.2.1 erratum also. Also... Did any previous version of XFree86 4.x work on this chip without disabling acceleration components? If so, I can try to find the breakage with rdiff sunnychan.com) What version of XFree86 are the above options needed for your particular chip? I just found a small bug in the neomagic driver source, which might explain this for the 2160, but not for the 2360. Right now, I'm working on the CVS X driver, but if it is impossible for one of you to test XFree86 CVS X for me, let me know and I will hack on earlier X code also, and provide test driver for 7.[123] and 8.0 stock XFree86 if needed. Neomagic configuration defaults, etc. are much different now than in the past. I also fixed a few driver bugs found while driver code surfing with lots of coffee. Been in rawhide for quite a while. If any new neomagic problems come up, please file new reports, and we should be able to find/fix them rather quickly. |
One some laptops XFree86 4.x hangs erratically (often every two or three days) on some neomagic chips. The documentation says On some laptops using the 2160 chipset (MagicGraph 128XD) the following options are needed to avoid a lock-up of the graphic engine: Option "XaaNoScanlineImageWriteRect" Option "XaaNoScanlineCPUToScreenColorExpandFill" These options cause nothing more than a tiny slowdown on other 2160 using hosts so I feel Xconfigurator should write those lines in for the affected chipset regardless. End users are not going to be able to follow man neomagic and what is needed. 3.3.x also breaks with the Xconfigurator complaining that XF86_None is not installed when you use --preferxf3 This shows up on Telsa's laptop so fixing it might make my life a lot easier 8)