User avatar
diereinegier
Posts: 166
Joined: Sun Dec 30, 2012 5:45 pm
Location: Bonn, Germany
Contact: Website

Re: Accelerated X driver testing

Sat Jan 05, 2013 10:49 pm

Where do I get an image containing kernel > 3.6.1 ?
Download my repositories at https://github.com/GeorgBisseling

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sat Jan 05, 2013 11:48 pm

bleep42 wrote:snip
Cool, I'm try and reproduce that corruption. Can you please try disabling the VpuOffload option, restart X and try again? Change the option to false or simply remove that line.
http://elinux.org/RPi_Xorg_rpi_Driver#C ... the_driver
bleep42 wrote:I did get a message

Code: Select all

Obt-Message: Xinerama extension is not present on the server
The xinerama thing is fine, that's for multiple displays attached to one GPU - the Raspberry Pi doesn't support that.
when I ran 'startlxde' and I am getting in gdb

Code: Select all

(WW) FBDEV_RPI(0): composite would have overread in X by 1108 pixels
--repeated 10 times--
(WW) FBDEV_RPI(0): composite would have overread in X by 1108 pixels
This is bad application behaviour! The program you're running here (LXterminal does it for example) is trying to write outside of its window. I'm catching that here, and neutralising the call before it crashes the system. What were you doing?
As an aside, do I need to do the procedure under "Kernel module install and set-up" after every reboot, or only once. What is the recommended way of exiting cleanly to shutdown?
The kernel module must always be loaded in memory when the driver is run. You can get the OS to auto-load that module on boot-up and that should do the trick (don't ask me how!). The character device you make with mknod needs to match the major/minor numbers of the kernel module though, and they only vary when modules are loaded in a different order. So if you have a module load order which never changes then you only need to do mknod once.

Just for clarity: the module does not need to be loaded/unloaded around every time you run X in a given session. Just load the module once and never bother unload it.

And when you want to turn it off? I just ctrl-c the X server and sudo shutdown.

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sat Jan 05, 2013 11:49 pm

diereinegier wrote:Where do I get an image containing kernel > 3.6.1 ?
Have a look here for the pre-install set-up.
http://elinux.org/RPi_Xorg_rpi_Driver#P ... all_set-up

The first command will retrieve the bleeding edge kernel and firmware.

User avatar
diereinegier
Posts: 166
Joined: Sun Dec 30, 2012 5:45 pm
Location: Bonn, Germany
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 12:11 am

Thanks!

But I saw that rpi-update gave me new kernel (3.6.x) and firmware greater than required in your instructions.

Now I was right at the point before driver installation and observed a strange issue:

In light blue shades I had magenta noise in exactly vertical 1-pixel wide bands and in middle blue shades I had green noise in more horizontal bands.

This did not go away by playing with config options (config_hdmi_boost etc.). I saw the similar effects with raspbmc btw. : the noise appeared at the border between different shades of blue in the background wallpaper: white-bluish bubbles on blue-foggy background - kind off.

To check if this was an issue with my monitor I flashed a card with the official wheezy image as of 2012-12-16. No issues, crisp sharp image.

AFAIK rpi-update changes firmware, kernel and modules. So somehwere in there is the problem.

Next thing I will try will be installation of the driver. But not tonight: it is after one o'clock in the morning already ;-)
Or should I instead try the image you mentioned above?

cheers!
Download my repositories at https://github.com/GeorgBisseling

User avatar
diereinegier
Posts: 166
Joined: Sun Dec 30, 2012 5:45 pm
Location: Bonn, Germany
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 12:26 am

Another sample point of a relatively new but crisp image (pun intended):

Based on wheezy 2012-12-16 and rpi-update'd to
firmware version 360331 (release)
and kernel
Linux crumb 3.6.11+ #348 PREEMPT Tue Jan 1 16:33:22 GMT 2013 armv6l GNU/Linux
Download my repositories at https://github.com/GeorgBisseling

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 8:47 am

bleep42 wrote:Hi Simon,
Using Midori, I am getting screen corruption on the Raspberry Pi banner across the top of the forum pages, it happens as I use the scroll wheel to scroll up or down, not continuously, mainly it's ok but say 1 in 3 clicks of the scroll wheel it'll corrupt, picture here.
https://picasaweb.google.com/knmoore/Pi ... 2657895266
Hello again

I haven't managed to reproduce this. Just to confirm - this is in Midori 0.4.3 in the Raspbian image, and you've not disabled the preload for my memcpy/memset replacement? The site is http://www.raspberrypi.org/phpBB3/ and you're not logged on or anything? Has Midori had any options changed?
Also you mention scrolling - the thing in your image which is corrupt does not scroll for me (it's a fixed banner). Does it scroll for you?

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 8:52 am

diereinegier wrote:Thanks!

But I saw that rpi-update gave me new kernel (3.6.x) and firmware greater than required in your instructions.
Yes, that's fine!

That's why I put on the wiki page,

Code: Select all

uname -a should give something like > 3.6.11 
vcgencmd version should give something like > 359004
The > being greater. I'll clarify this on the page.
Now I was right at the point before driver installation and observed a strange issue:
This sounds a bit odd (and something I can't control I'm afraid). If you still have the image it might be worth reporting it in this thread in case others see it: http://www.raspberrypi.org/phpBB3/viewt ... 29&t=19334

EDIT FOR bleep42: yeah I still can't reproduce that Midori error. I notice that you're also running Chromium and some other program (what is that fractal program?). It's likely that the memory allocated to X is running low.

In addition to trying to disable the VpuOffload option, can you also try increasing the BlockSize option? You should be able to get to 32 MB 33554432 easily.

juppiter89
Posts: 91
Joined: Fri Jan 04, 2013 10:50 pm

Re: Accelerated X driver testing

Sun Jan 06, 2013 10:22 am

I also noticed some graphic corruptions in midori, chromium and freecell game (my girlfriend plays with it :lol: ). I can't reproduce them though, just closing and restarting the application and artifacts disappear, they seem to be purely casual. I'll try later your last hints ;)

juppiter89
Posts: 91
Joined: Fri Jan 04, 2013 10:50 pm

Re: Accelerated X driver testing

Sun Jan 06, 2013 2:07 pm

Testing continues... I got this in Xorg log during Midori browsing, Raspberry hung up.

Code: Select all

(WW) FBDEV_RPI(0): composite would have overread in Y by 7560 pixels

Program received signal SIGSEGV, Segmentation fault.
Op (width=<optimized out>, height=<optimized out>, source=..., dest=..., 
    mask=..., coord=..., width=<optimized out>, height=<optimized out>)
    at /home/simon/Desktop/x/src/driver/xf86-video-fbdev_exa/src/./composite_ops.inl:876
876	/home/simon/Desktop/x/src/driver/xf86-video-fbdev_exa/src/./composite_ops.inl: File or directory not existing.

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 3:52 pm

Niiiice! Is it still crashed? Do you have a core file?
If not, if ever you get something like that can try these instructions and make me a core file?

http://elinux.org/RPi_Xorg_rpi_Driver#S ... or_SIGTRAP
I can then analyse the crash.

Btw did the messages about over-reading occur at the same time as the crash?
EDIT did you also try those two options from earlier, VpuComposite and the BlockSize option?

juppiter89
Posts: 91
Joined: Fri Jan 04, 2013 10:50 pm

Re: Accelerated X driver testing

Sun Jan 06, 2013 4:36 pm

Well, I would like to help you, but I can't start Xorg with gdm: when I run startxfce4 from other terminal, gdm seems to hang up, I noticed that it reach almost 100% CPU load and I have to restart the Raspberry :? :(

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 5:30 pm

Ok. Just to clarify, gdm or gdb?

juppiter89
Posts: 91
Joined: Fri Jan 04, 2013 10:50 pm

Re: Accelerated X driver testing

Sun Jan 06, 2013 5:32 pm

teh_orph wrote:Ok. Just to clarify, gdm or gdb?
gdb

EDIT : now I noticed that I wrote gdm, of course I meant gdb ;)

User avatar
bleep42
Posts: 156
Joined: Wed Mar 07, 2012 12:43 pm
Location: Sussex

Re: Accelerated X driver testing

Sun Jan 06, 2013 6:56 pm

teh_orph wrote:
bleep42 wrote:Hi Simon,
Using Midori, I am getting screen corruption on the Raspberry Pi banner across the top of the forum pages, it happens as I use the scroll wheel to scroll up or down, not continuously, mainly it's ok but say 1 in 3 clicks of the scroll wheel it'll corrupt, picture here.
https://picasaweb.google.com/knmoore/Pi ... 2657895266
Hello again
I haven't managed to reproduce this. Just to confirm - this is in Midori 0.4.3 in the Raspbian image, and you've not disabled the preload for my memcpy/memset replacement? The site is http://www.raspberrypi.org/phpBB3/ and you're not logged on or anything? Has Midori had any options changed?
Also you mention scrolling - the thing in your image which is corrupt does not scroll for me (it's a fixed banner). Does it scroll for you?
Hi Simon,
Yes just checked and it is Midori 0.4.3, completely standard as installed from the raspbian image. As far as I know I haven't disabled any of your drivers, how can I check?. It is a straight image of 2012-12-16-wheezy-raspbian.zip then updated as you specify in your http://elinux.org/RPi_Xorg_rpi_Driver page, I didn't have to enable your drivers did I and I missed that bit? The only change is I've got /boot on a 8Gb CF card plugged in via a USB adapter. I'm using a WiFi USB dongle as well.

Code: Select all

[email protected] ~ $ uname -a
Linux raspberrypi 3.6.11+ #334 PREEMPT Wed Dec 19 23:02:43 GMT 2012 armv6l GNU/Linux
[email protected] ~ $ vcgencmd version
Dec 19 2012 23:13:22
Copyright (c) 2012 Broadcom
version 359004 (release)
[email protected] ~ $ fbset
mode "1920x1200"
    geometry 1920 1200 1920 1200 32
    timings 0 0 0 0 0 0 0
    rgba 8/0,8/8,8/16,8/24
endmode
What I meant was scrolling the page, yes the Raspberry Pi banner stays stationary and the page disappears under it. Personally I think it's a pain staying there, but that's how they have designed the site. I have re loaded from cold at least 4 times now and always get the same.
I have also now set 'VpuOffload' to 'false' in file /usr/share/X11/xorg.conf.d/01-fbdev-rpi.conf but I still get the same corruption.
Have just found that by changing the window width of Midori, I can sometimes make it stop corrupting, so it might be worth changing your window size; if I make it so narrow that the banner does not stay at the top of the page any more, that also stops any corruption.
Let me know if you have any more queries, suggestions or things you'd like me to try.
Regards, Kevin.

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 9:12 pm

Cool thanks very much for the detail. I think I've uncovered a trouble with a VPU function, I'll see what's going on.
If you can also see a trouble with VpuComposite off, can you also set SelfManagedOffscreen to false?

(and just in case I'd not mentioned - any change in that config file requires Xorg being restarted for it to take effect)

User avatar
diereinegier
Posts: 166
Joined: Sun Dec 30, 2012 5:45 pm
Location: Bonn, Germany
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 10:04 pm

I observed the following drawing errors:

When using iceweasel on raspberry.org some of the badges under "BUY A PI" seem to have red and blue interchanged. However the colors are dsplayed correctly i the Print Preview.

When using chromium on raspberrypi.org and hovering around over the "Raspberry Pi (R)" in the top left corner, leaving that sometimes paints some pixel noise under the letters. this seems to happen only when this logo and the tip off the day are laid out overlapping (narrow window).

More to come.
Download my repositories at https://github.com/GeorgBisseling

User avatar
diereinegier
Posts: 166
Joined: Sun Dec 30, 2012 5:45 pm
Location: Bonn, Germany
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 10:17 pm

Using iceweasel on http://www.heise.de/?seite=1 and scrolling down the list of articles I see parts of one article thumbnail inside the thumbnail of the article above.
Download my repositories at https://github.com/GeorgBisseling

User avatar
diereinegier
Posts: 166
Joined: Sun Dec 30, 2012 5:45 pm
Location: Bonn, Germany
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 10:32 pm

Iceweasel actually tries to play the video on raspberrypi.org!
Framerate is absent, maybe 0.01 fps. Sound is noise only, but hey: there are frames.
Reload: total freeze, even on the ssh terminals. No info in gdb.
Try to login. Nope.
Doing a powercycle...

BTW: I am using a Model B with 512 Mb and 4kB pages with CMA.
Download my repositories at https://github.com/GeorgBisseling

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 10:59 pm

Iceweasel frequently hangs my Pi regardless of which driver I use (due to CPU load), and also has R/B swapped on many images with the stock driver too. It's been like that for months...but I'm not sure why to report it to!
Any chance of a screenshot of the heise site thing? Does it do this with the original driver?

I've managed to find a failure in one of the pixel format permutations for the VPU assembly. It's now just a case of tracking down why it's like that! Having no debugger for it makes this quite a chore.

juppiter89
Posts: 91
Joined: Fri Jan 04, 2013 10:50 pm

Re: Accelerated X driver testing

Sun Jan 06, 2013 11:09 pm

It seems that I'm the only one that can't start Xorg by gdb, quite frustrating...
So I imagine that there is no other way to get a core file dump, true? :roll:

User avatar
diereinegier
Posts: 166
Joined: Sun Dec 30, 2012 5:45 pm
Location: Bonn, Germany
Contact: Website

Re: Accelerated X driver testing

Sun Jan 06, 2013 11:41 pm

The original driver displays the site correctly.

Attached you'll find the screenshot.

Enough for tonight: got to get to work tomorrow.
Attachments
2013-01-06-233225_1280x1024_scrot_cropped.png
i had to crop because of the file size limit.
2013-01-06-233225_1280x1024_scrot_cropped.png (51.14 KiB) Viewed 3552 times
Download my repositories at https://github.com/GeorgBisseling

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Mon Jan 07, 2013 8:36 am

diereinegier wrote:The original driver displays the site correctly.

Attached you'll find the screenshot.

Enough for tonight: got to get to work tomorrow.
That's helpful thanks. At this end I'm tracking a bug which should cause general corruption with a specific pixel format input combination but your picture looks like it's sort of sheared... Does it do this *every* time the page is rendered?

I'm also unsure of why other instances of corruption are not repeatable. There may be another issue...
juppiter89 wrote:It seems that I'm the only one that can't start Xorg by gdb, quite frustrating...
So I imagine that there is no other way to get a core file dump, true? :roll:
Yeah this is actually really annoying - you're the only person other that me who's been able to crash it! (ever!) You are typing 'run' into gdb, right? Running in gdb should make no difference at all. Anyway it's too late now, we can't retrospectively get a core file :(

User avatar
bleep42
Posts: 156
Joined: Wed Mar 07, 2012 12:43 pm
Location: Sussex

Re: Accelerated X driver testing

Mon Jan 07, 2013 9:26 am

teh_orph wrote:Cool thanks very much for the detail. I think I've uncovered a trouble with a VPU function, I'll see what's going on.
If you can also see a trouble with VpuComposite off, can you also set SelfManagedOffscreen to false?

(and just in case I'd not mentioned - any change in that config file requires Xorg being restarted for it to take effect)
Hi Simon,
When I’m doing testing and changing options I almost always do a complete reboot, just to be sure. :-)
Sorry I missed this comment from above "In addition to trying to disable the VpuOffload option, can you also try increasing the BlockSize option? You should be able to get to 32 MB 33554432 easily." I'll give that a go later when I get home.
In the screen shot I gave you I did have a fractal program going as well (xaos), but with all subsequent testing I’ve only had Midori and maybe a terminal on screen at any time.
I now think that the “(WW) FBDEV_RPI(0): composite would have overread in X by 1108 pixels” was caused by xaos as I’ve not seen it again.
I’ll also give the option ‘SelfManagedOffscreen’ to ‘false’ a go, should I leave ‘VpuComposite’ set ‘false’ as well?
For your information, I have had it crash once, but it was a complete wipe out, the first I knew was when the SSH connections all dropped out, when I went back to the Pi desk top it was blank, I had to pull the plug :-(
I’ll try to let you know what I find later tonight, if I’m allowed ;-)
Regards, Kevin.

User avatar
teh_orph
Posts: 346
Joined: Mon Jan 30, 2012 2:09 pm
Location: London
Contact: Website

Re: Accelerated X driver testing

Mon Jan 07, 2013 9:51 am

Thanks again.
Basically the two things which have had the least amount of testing are the VpuOffload option (about a month's worth) and the SelfManagedOffscreen (about two weeks worth). With those options turned off you lose a lot of performance but is by far the most tested configuration (many many months worth).

Is there any chance that you have a USB serial port adaptor which you can connect to the UART on the board? I get a lot of kernel oopses/panics from the Ethernet driver when the X driver is going at full bore. Unfortunately the only way that you can see this happening 'live' is to watch the kernel output over a serial port. Once the Ethernet driver fails completely you can't watch it over SSH.

Just for lols, can you try: http://elinux.org/RPi_Xorg_rpi_Driver#T ... ps_working
(try the smsc95xx.turbo_mode=N and vm.min_free_kbytes things together, ignore the other suggestions)

And finally when running X can you open another SSH and just leave it as tail -f /var/log/kern.log just in case you can see an oops? (I'm not at my rpi so can't double-check that's the path to the log file)

juppiter89
Posts: 91
Joined: Fri Jan 04, 2013 10:50 pm

Re: Accelerated X driver testing

Mon Jan 07, 2013 12:21 pm

teh_orph wrote:
diereinegier wrote:The original driver displays the site correctly.

Attached you'll find the screenshot.

Enough for tonight: got to get to work tomorrow.
That's helpful thanks. At this end I'm tracking a bug which should cause general corruption with a specific pixel format input combination but your picture looks like it's sort of sheared... Does it do this *every* time the page is rendered?

I'm also unsure of why other instances of corruption are not repeatable. There may be another issue...
juppiter89 wrote:It seems that I'm the only one that can't start Xorg by gdb, quite frustrating...
So I imagine that there is no other way to get a core file dump, true? :roll:
Yeah this is actually really annoying - you're the only person other that me who's been able to crash it! (ever!) You are typing 'run' into gdb, right? Running in gdb should make no difference at all. Anyway it's too late now, we can't retrospectively get a core file :(
Well, now I explain exactly what i do...
I open first one SSH terminal, and I run in the order:

Code: Select all

sudo modprobe dmaer_master
major=$(awk '$2=="dmaer" {print $1}' /proc/devices)
sudo mknod /dev/dmaer_4k c $major 0
then

Code: Select all

sudo gdb Xorg
and in gdb

Code: Select all

set args -verbose -keeptty
handle SIGPIPE nostop
handle SIGUSR1 nostop
run
So far seems to run correctly... then I run in another SSH terminal:

Code: Select all

export DISPLAY=:0
startxfce4
Then i get in terminal only "/usr/bin/startxfce4: X server already running on display :0"; everything seem to freeze and gdb stably reach almost 100% CPU load :?

Return to “General discussion”