Description of problem: Display corruption occurs in X11 with sisimedia driver. This problem also occurred in Mageia 5 -- but interestingly in an opposite way: everything worked nice expect Libreoffice, which would severely corrupt the display upon scrolling. At the time, there was some doubt where was the bug -- whether in sisimedia or in Libreoffice itself. A bug existed: https://bugs.documentfoundation.org/show_bug.cgi?id=58358 ...and a workaround was found: disabling acceleration -- not in Libreoffice (there's an option for that) but in xorg.conf. Fast forward to present day, I am testing aMageia 6 and decide to try to reproduce (and report) the bug. To my surprise, Libreoffice worked right out of the box (i.e., on a recently installed Mageia 6 sta1 with Libreoffice added). Unfortunately, now, the opposite effect was observed and display corruption started to happen outside Libreoffice (when running mc, for instance). A screenshot of the problem will be added. Version-Release number of selected component (if applicable): How reproducible: Tried a few times and it was reproduceable. Steps to Reproduce: 1.Install Mageia 6 on a computer which requires sismedia (config details ahead). 2.Open a terminal e run mc (install if not available). 3.Move from line to line (e.g. with the down arrow). 4. Edit Mageia's auto-created xorg.conf. 5. Add Option "NoAcell" "True" to device section. 6. Reboot and see problems vanish. Hardware: Notebook with: - Pentium T3400 (fairly reasonable 64-bit CPU) - 2 GB RAM - 14" 1280x800 LCD controlled by a SiS 771/671 chip <-- the problem! I'm marking it with a minor severity, because disabling acceleration with the graphics hardware in question has not much of a drastic effect. Video performance is bad with or without acceleration. 480p media play well, though.
Created attachment 8574 [details] Display corruption.
Sorry "expect" should be "expect"...
Any further info I may have forgotten to say, please ask.
Assigning to the kernel and drivers maintainers
CC: (none) => marja11Assignee: bugsquad => kernel
Created attachment 8579 [details] Notebook hardware summary
Created attachment 8580 [details] Connected/internal usb devices
Created attachment 8581 [details] CPU specs
Added some info about the hardware. BTW, I wish to congratulate those in Mageia who make it happen. Not all distributions offer drivers for these older, rarer graphic cards. Regarding sisimedia, for instance, I know 2. Yes, for all the talk about hundreds of Linux distributions, sometimes the viable options are extremely reduced. It's great that you keep the old Mandrake spirit! Thank you very much.
In light of some recent developments reported in bug 19849, the suggested workaround started to prevent the computer from automatically logging a user. Since problems with display corruption were more or less constrained to the use of midnight commander, a possible course of action is to use another program or use "mc -s" (slow), which reduces the occurrence of corruption to a minimum.
The thing is this driver is totally unmaintained: http://www.linuxconsulting.ro/xorg-drivers/ The upstream project stopped caring since 2009, aka there's no support for 7 years so I'm wondering if it wouldn't be better to default to some generic driver like fbdev or vesa... Can you update the "source RPM" field with the result of "rpm -q x11-driver-video-sisimedia" ? Does the official sis driver works for this card? Thanks
CC: (none) => thierry.vignaudSource RPM: (none) => x11-driver-video-sisimedia
Keywords: (none) => NEEDINFO
Thanks, Thierry, for the info & insights. Hmm, I thought the last guy to work on this was that the one cited in the man page, Thomas Winischhofer. Alas, all given online addresses are now 404ed. The answer to the command you mentioned is: [zz@localhost ~]$ rpm -q x11-driver-video-sisimedia x11-driver-video-sisimedia-0.9.1-9.20091203.25.mga6 [zz@localhost ~]$ I guess one has to balance the benefits -- reusing old but functional hardware -- and costs -- developer time is scarce. A lot of folks get only computers which were donated or bought for symbolic prices. Old machines sometimes are no longer fit for modern uses. Though Netscape 4.7 would run fast on them, Firefox 50 may not even start. From my own personal history, things get creative when people have smaller budgets. This notebook, for instance, could be paired with another old machine with good graphics but a weak processor (by using xdmcp). This certainly has drawbacks (energy for instance), but maybe the only way for a poor family to be able to do internet banking, learn online etc. VESA works (and even at 1280x800) but the generated image is imperfect with thin vertical stripes regularly spaced. It is enough for using the computer as a server, though. Performance might be a tad bit worse -- nothing meaningful. I've read everywhere that the official SiS driver won't cover the 771/671 chip, but I must try it myself. Maybe it can be made to work with inferior results like VESA, which might be an alternative in the future. I'll try with drakx11 if I remember. I'm typing this on that very notebook, so I have no way to test it now. BTW, it works quite well... I found this: http://askubuntu.com/questions/691600/sis671-video-driver and there's the Arch help (which mentions the SiS 671 chip): https://wiki.archlinux.org/index.php/SiS The graphics card, according to https://en.wikipedia.org/wiki/Silicon_Integrated_Systems, was released in 2003, but this notebook was made many years later... probably the project was from circa 2007~2008 and this particular one must have been acquired around 2009 (with WXP, as I've been told). Officially, SiS only provides drives for W7 (I checked some months ago). Please feel free to ask for info you think that might be useful -- sometimes I run out of ideas to test...
Source RPM: x11-driver-video-sisimedia => x11-driver-video-sisimedia-0.9.1-9.20091203.25.mga6
Just typed: [zz@localhost ~]$ rpm -qi x11-driver-video-sisimedia Name : x11-driver-video-sisimedia Version : 0.9.1 Release : 9.20091203.25.mga6 Architecture: i586 Install Date: Sáb 26 Nov 2016 15:01:52 BRST Group : System/X11 Size : 648929 License : MIT Signature : RSA/SHA1, Qui 17 Nov 2016 14:08:04 BRST, Key ID b742fa8b80420f66 Source RPM : x11-driver-video-sisimedia-0.9.1-9.20091203.25.mga6.src.rpm Build Date : Qui 17 Nov 2016 13:54:29 BRST Build Host : rabbit.mageia.org Relocations : (not relocatable) Packager : tv <tv> Vendor : Mageia.Org URL : http://www.linuxconsulting.ro/xorg-drivers/ Summary : Video driver for SiS 670 / 671 cards ================================================= important part! Description : x11-driver-video-sisimedia is the video driver for SiS 670 / 671 cards. These are not supported by the X.org 'sis' driver. This code is very different, so the two cannot be easily merged. ============================================================= [zz@localhost ~]$ Also, right now, there was a problem with the terminal when it was brought to front -- image corrupted and fixed by shading (and showing again) the terminal window. Maybe some corruption problems are related to some kind of double buffering... I'll check that later, too.
(In reply to Renato Dali from comment #11) I know about old hardware. I'm not saying we should ignore them. I'm just pointing this driver was a quick hack that has been abandoned for 7 years and barely compiles b/c I blindly patch it to work with latest xservers w/o any HW to test And we cannot do anything about reported bugs as there's no upstream to look about them. Ideally we would need someone to analyze what were the interesting bits in the *huge* diff between sis & sisimedia and then fast forward it to current sis driver... Which is what the original hacked should have tried to attempt instead of forking a quick hack (which was valuable for users in the short term, but not on the long term) In the meantime, we can disable accel by default for sisimedia I'll suggest a patch (In reply to Renato Dali from comment #12) I know as I'm keeping this driver to compile since xserver 1.10 up to 1.19. I was just wondering if anyone has tested it with the real sis driver just in case... There does be references to sis 67[01] in its code
Please try x11-driver-video-sisimedia-0.9.1-9.20091203.26.mga6 once it lands on your favorite mirror (compile tested only)
Also, one could try if modesetting driver works
CC: (none) => tmb
Without a prorper drm driver :-) ?
heh, yeah... seems my head is not where it's supposed to be :)
Sorry for my lengthy discourse on the subject of reusing computers. I'm revealing my age here, but I'm from a time when RAM was prohibitively expensive. One dreamed of going from 4Kbytes to 16Kbytes and then to 64Kbytes... sorry if I look obsessed. I will try it -- it already appeared. Just have in mind disabling acceleration cause a problem with this hacked driver. > I know as I'm keeping this driver to compile since xserver 1.10 up to 1.19. No, it was done by a tv guy. Oh, wait a minute... duh! This issue about unmaintained upstream code is somewhat like the recent problems with libav codecs. In that case, because vulnerabilities were involved, Mozilla decided no play until people sort things out (which happened recently and certain videos came back to life in FF). If sisimedia caused some kind of vulnerability or danger that would be arguable, too, IMHO. I'm going to test the new driver and post the results afterward. Thanks for your kind attention.
err. I'm confused: What do you mean? You put "workaround: NoAccel" in the summary and now you're saying "disabling acceleration cause a problem with this hacked driver"? Which one is true?
Both. NoAccel solves the screen corruption problem. Someone must really want that and consider the loss of that feature (acceleration) acceptable. There are scenarios where display artifacts simply are not tolerated (e.g. for a Point-of-sale in a store). NoAccel is causing autologin not to function (I cannot fathom why). This is important for me. I want the computer ready to use by the time I get back with my coffee cup. But I failed to update the title to reflect what I discovered in bug 19849 (in essence, summarized above).
Summary: Display corruption with sisimedia (workaround: NoAccel). => Display corruption with sisimedia (NoAccel solves that but prevents autologin).
Ok, installed the new driver and it works as expected: no display corruption at all but no more autologin... It is also not possible to turn Acceleration on, apparently. Option AccelMethod is not used, it is in the log, but AccelMethod sna made the computer show the "Good luck" message. Go figure! I believe that, from the point of view of most users and Mageia as a whole, the current version is better -- display corruption essentially sucks. And I'll investigate the text mode autologin (what used to be called init level 3...). tl;dr: The current version avoids the worse of two evils -- better than the former, IMHO.
I had to boot Mageia 5 (*) because I tried Defaultdepth 16 and that made the notebook stop booting... Then I took the opportunity to get some files from it (Xorg.0.log and xorg.conf). In Mageia 6, it was: x11-driver-video-sisimedia-0.9.1-9.20091203.25.mga6 [zz@localhost ~]$ rpm -q x11-driver-video-sisimedia x11-driver-video-sisimedia-0.9.1-9.20091203.26.mga6 [zz@localhost ~]$ In Mageia 5: x11-driver-video-sisimedia-0.9.1-6.20091203.20.mga5 (*) I downloaded Mageia 5.1 and dumped it on a USB drive, but it is written Mageia 5 everywhere... but I even have the downloaded image (on this netbook, I use Mageia 6 to create it). KDE4 works nicely -- much better in fact than Xfce. The terminal window can be moved while showing its contents -- in KDE I must reduce it to a frame while moving for performance reasons... Anyway, maybe those files can be useful to see things like what was working in version 5 etc. Thanks.
Created attachment 8779 [details] X11 log when running Mageia 5.1 KDE live on the same hardware.
Created attachment 8780 [details] xorg.conf used by Mageia 5 KDE live
Created attachment 8781 [details] X11 log with Mageia 6 for comparison
Created attachment 8782 [details] Log diffs for convenience... Maybe this can be better done...
The diff surely can be better done... I have had trouble when comparing configs, for instance. Does anyone have a favorite diff tool with interesting features?
(In reply to Renato Dali from comment #21) > Option AccelMethod is not used, it is in the log, but AccelMethod sna made > the computer show the "Good luck" message. Go figure! "sna" is an *intel* specific option, it won't work with any other driver...
Created attachment 8783 [details] X.0.log diff between mga[56]
Attachment 8782 is obsolete: 0 => 1
The diff got much better, thanks. If nobody takes on the task of maintaining this driver -- and I unfortunately can't -- it follows that it will eventually break. From then on, VESA can be used... or xdmcp. Come to think, this would be ideal work for a DriverJanitors group, much like what the KernelJanitors do for the kernel... or as an exercise for undergraduate Programming students. For the moment, things are not bad at all: the driver works well enough both in Mageia 5.1 and in 6. Systemd-related issues are much more important, me thinks -- and might be playing a role in this case, too. Thanks for your hard work.
As the display corruption is effectively prevented from happening, closing as FIXED.
Status: NEW => RESOLVEDResolution: (none) => FIXED