R600/r700 kms + 3d dri1/dri2

[tettamanti]

3) it's really really fragile: whenever I try to start certain apps (I'll see which they are) X crashes and reloads gdm from what I saw it's related to exa sofar ...

at least it's stable enough to provide basic X with blackbox and firefox-bin (32bit firefox-binary from mozilla.org) that's the environment I'm posting from right now

error message was:

Backtrace:
0: /usr/X11R6/bin/X(xorg_backtrace+0x3d) [0x7f419700c09d]
1: /usr/X11R6/bin/X(xf86SigHandler+0xfd) [0x7f4196f8b27d]
2: /lib/libc.so.6 [0x7f4194d7a470]
3: /usr/lib64/xorg/modules/drivers//radeon_drv.so [0x7f419333a546]
4: /usr/lib64/xorg/modules//libexa.so [0x7f4192c16b93]
5: /usr/lib64/xorg/modules//libexa.so [0x7f4192c175ce]
6: /usr/X11R6/bin/X [0x7f419706d9e4]
7: /usr/lib64/xorg/modules//libexa.so(exaGlyphs+0x2a9) [0x7f4192c18999]
8: /usr/X11R6/bin/X [0x7f419706cea5]
9: /usr/X11R6/bin/X [0x7f41970633b7]
10: /usr/X11R6/bin/X(Dispatch+0x34d) [0x7f4196f3597d]
11: /usr/X11R6/bin/X(main+0x43a) [0x7f4196f13e7a]
12: /lib/libc.so.6(__libc_start_main+0xfd) [0x7f4194d65a0d]
13: /usr/X11R6/bin/X [0x7f4196f13169]

Same here; it's easily reproducible with krunner. gdb indicates that it's crashing inside R600Composite. I've created a bug entry here.

[kernelOfTruth]

No, it works built-in. I gave it a try by including the firmware blobs in the kernel binary and it boots (and X starts).

@kernelOfTruth: those crashes are there for me too. They happen with fluxbox and vboxgtk. (edit) More specifically, it happens in vboxgtk when I move the mouse cursor over the GtkTreeView with the VMs. Weird.

for me it crashes with fluxbox too, so yeah - something gtk-related / heavy drawing might trigger this crash;

strange thing is, when launching thunar, mousepad (lightweight gtk apps) it doesn't crash, but when trying to launch nautilus, (if I recall right) gedit it produces the mentioned crash

[bridgman]

There was mention of an EXA bug on #radeon earlier today, might be the same thing.

[nanonyme]

I also sometimes get display turnoff's immediately followed by a lockup with KMS.

[tettamanti]

Same here; it's easily reproducible with krunner. gdb indicates that it's crashing inside R600Composite. I've created a bug entry here.

Fixed for me in git current (commit 0bb0ff0e).

[Fran]

Fixed for me in git current (commit 0bb0ff0e).

Yeah, it was solved when I synced this afternoon.

Now the only problem for me is the lack of suspend. Other than that it works well: stable and fast 2D and 3D.

[kernelOfTruth]

Yeah, it was solved when I synced this afternoon.

Now the only problem for me is the lack of suspend. Other than that it works well: stable and fast 2D and 3D.

corruption-free text-input or graphics with compiz / kwin too ?

[bridgman]

IIRC agd5f mentioned that the compiz text-input corruption does not happen with dri2.

[pingufunkybeat]

Does this mean that KMS+DRI2 is at a similar state as DRI1 on r600 now? Or am I reading this too optimistically?

Because that was mighty fast.

[bridgman]

Probably too optimistically, but it does seem to be stabilizing very quickly. Then again, if you count the lack of compiz corruption... hmmm.

The 3D code is mostly common between DRI1 and DRI2 -- this is the upside from porting the 3D driver over to the radeon-rewrite code base a few months ago. Doing that port delayed the 3D driver by a couple of months, but it meant that the move to KMS and DRI2 would be simplified.

That said, I am *still* surprised how well it all came together.