Problem with graphics drm modules

[andyj]

I updated hsetroot in CorePure64 Xlibs - it says it fixes a seg fault when the display is not present, maybe that will help?

The segfault for hsetroot in dmesg does look to be fixed, but X still segfaults. A step in the right direction anyway.

[Juanito]

Xlibs also updated in the x86 repo

Did you submit a bug report for xf86-video-vmware?

[andyj]

I created the following VMs:

OpenSUSE Tumbleweed, xf86-video-vmware 13.3.0 and Xorg 1.19.6, works.
Fedora Rawhide xf86-video-vmware 13.2.1 with Xorg 1.20.0 works.
Debian Sid, xf86-video-vmware 13.3.0 and Xorg 1.20.0, works.
Slackware-current, xf86-video-vmware 13.3.0 and Xorg 1.20.0 works.

I'm not convinced that it's a problem in xf86-video-vmware. It seems more likely that there is some other compile / library mismatch in TC.

[Juanito]

Hardware-wise I can only test the modesetting, intel and vesa modules. I've also tested with qemu and virtualbox and I cannot make x fail with any of them.

If I understand correctly, there's no free version of vmware with which to test?

[andyj]

"VMware Player" is the free (lite) version of VMware Workstation and is what I am using. ESXi is the free bare-metal hypervisor for servers.

[Juanito]

so, using VMware Workstation 14.1.2 Player with CorePure64

Xorg-7.7 works
+ graphics-KERNEL works
+ Xorg-7.7-3d works
+ xf86-input-vmmouse xf86-video-vmware hangs

It looks like there's either a problem with a lack of suitable gallium driver or that vmmouse conflicts with evdev's choice of mouse and x sits there waiting for an input device - see attached

The above being said, I never looked at a vmware log, so maybe this is normal..

[andyj]

In settings, click on Display and enable 3D if you haven't already. Then in dmesg you should have this to tell you that the kernel DRM has initialized:

Code: [Select]

[drm] DX: yes.3D is detected.

Code: [Select]

[drm] Initialized vmwgfx 2.14.0 20170612 for 0000:00:0f.0 on minor 0DRM is working.

I've been booting into text mode, starting sshd, and connecting from a host konsole window so I can startx manually in the VM console and see the system after X hangs.

In /var/log/Xorg.0.log:

Code: [Select]

[    54.792] (II) vmware(0): Initialized VMWARE_CTRL extension version 0.2
[    54.804] (II) vmware(0): Gallium3D XA version: 2.3.0.
[    54.804] (WW) Gallium3D XA version insufficient for dri3.
[    54.804] (II) vmware(0): Path of drm device is "/dev/dri/card0".
[    54.804] (II) vmware(0): [DRI2] Setup complete
[    54.804] (II) vmware(0): [DRI2]   DRI driver: vmwgfx
[    54.804] (--) vmware(0): Render acceleration is enabled.
[    54.804] (==) vmware(0): Rendercheck mode is disabled.
[    54.804] (--) vmware(0): Direct rendering (DRI2 3D) is enabled.
[    54.804] (--) vmware(0): Direct rendering (DRI3 3D) is disabled.
Sid and Tumbleweed have the same version of Gallium, and the logs look the same, so it's apparently not a serious problem. Rawhide also has Gallium 2.3.0, but it's not trying to load dri3 so it doesn't complain about Gallium.

Even when I don't load vmmouse it still segfaults, so I don't think that's it either.

You can create /etc/vmware-tools/tools.conf to turn on some debugging:

Code: [Select]

[logging]
log = true

# Enable VMware Tools service logging to a file.
vmtoolsd.level = debug
vmtoolsd.handler = file
vmtoolsd.data = /var/log/vmtoold.log

# Enable "vmsvc" service logging to a file.
vmsvc.level = debug
vmsvc.handler = file
vmsvc.data = /var/log/vmsvc.log

# Enable new "vmusr" service logging to a file.
vmusr.level = debug
vmusr.handler = file
vmusr.data = /var/log/vmusr.${USER}.log

# Enable the "vmvss" snapshot service logging to a file.
vmvss.level = debug
vmvss.handler = file
vmvss.data = /var/log/vmvss.log

# Enable VMwareResolutinSet logging to a file
vmresset.level = debug
vmresset.handler = file
vmresset.data = /var/log/vmresset.log
But so far I haven't found that it has anything that could give some direction.

[Juanito]

I wasn't aware that 3d graphics needed to be switched on in vmware - yes, it seems to work, even with xf86-input-vmmouse, but without xf86-video-vmware.

As soon as xf86-video-vmware is added:

Code: [Select]

[   739.736] (EE) Backtrace:
[   739.742] (EE) 0: /usr/local/lib/xorg/Xorg (xorg_backtrace+0x49) [0x51b05c]
[   739.742] (EE) 1: /usr/local/lib/xorg/Xorg (0x400000+0x13bdf2) [0x53bdf2]
[   739.742] (EE) 2: /lib/libpthread.so.0 (0x7f39b4fe8000+0x10030) [0x7f39b4ff8030]
[   739.743] (EE) 3: /usr/local/lib/xorg/Xorg (0x400000+0x12d4a2) [0x52d4a2]
[   739.743] (EE) 4: /usr/local/lib/xorg/Xorg (RROutputIsLeased+0x14) [0x4c5c4e]
[   739.743] (EE) 5: /usr/local/lib/xorg/Xorg (xf86DPMSSet+0x54) [0x4cd2f8]
[   739.743] (EE) 6: /usr/local/lib/xorg/modules/drivers/vmware_drv.so (0x7f39b1876000+0x9b98) [0x7f39b187fb98]
[   739.743] (EE) 7: /usr/local/lib/xorg/modules/drivers/vmware_drv.so (0x7f39b1876000+0x1189c) [0x7f39b188789c]
[   739.743] (EE) 8: /usr/local/lib/xorg/modules/drivers/vmware_drv.so (0x7f39b1876000+0x9c2e) [0x7f39b187fc2e]
[   739.743] (EE) 9: /usr/local/lib/xorg/Xorg (0x400000+0xcdce1) [0x4cdce1]
[   739.743] (EE) 10: /usr/local/lib/xorg/Xorg (0x400000+0x108bfc) [0x508bfc]
[   739.743] (EE) 11: /usr/local/lib/xorg/Xorg (0x400000+0x85631) [0x485631]
[   739.743] (EE) 12: /usr/local/lib/xorg/Xorg (0x400000+0x2de01) [0x42de01]
[   739.743] (EE) 13: /lib/libc.so.6 (__libc_start_main+0x15a) [0x7f39b414509d]
[   739.743] (EE) 14: /usr/local/lib/xorg/Xorg (_start+0x2a) [0x42defa]
[   739.743] (EE)
[   739.743] (EE) Segmentation fault at address 0x1000000b1
[   739.743] (EE)
Fatal server error:
[   739.743] (EE) Caught signal 11 (Segmentation fault). Server aborting
[   739.743] (EE)

Now I guess the trick is to add gdb to the mix somehow, but it would seem to merit an xf86-video-vmware bug report...

[andyj]

Might be a bug, but my guess is the developers of X and all the other libraries are using a distro other than TC. I was unable to reproduce the problem in four other distros, i.e., it works as expected, so it might be a tough sell. Are we sure all the parts are compiled to the same ABI?

[Juanito]

If you mean the new xorg-server abi, then only xf86-input* and xf86-video* need recompiling against it and Xorg.0.log gives an error for any drivers compiled against the wrong abi.

[Juanito]

I managed to get something with gdb, but I'm not sure whether this is a problem with xorg-server or xf86-video-vmware...

Code: [Select]

0x000000000050fbfe in dixGetPrivate (privates=0x1b740d0,
    key=0x8d2840 <rrPrivKeyRec>) at ../include/privates.h:136
No locals.
#1  0x000000000050fc46 in dixLookupPrivate (privates=0x1b740d0,
    key=0x8d2840 <rrPrivKeyRec>) at ../include/privates.h:166
No locals.
#2  0x000000000050ffa0 in RROutputIsLeased (output=0x1b73720) at rrlease.c:124
        screen = 0x1b73d00
        scr_priv = 0x1b6eeb0
        lease = 0x1b6f010
        o = 3
#3  0x0000000000585de8 in xf86DPMSSet (scrn=0x1aee340, mode=3, flags=0)
    at xf86Crtc.c:2967
        output = 0x1af8bc0
        config = 0x1aef650
        i = 5
#4  0x00007f4b70b812cb in vmwgfx_disable_scanout (pScrn=0x1aee340)
    at vmwgfx_crtc.c:115
        i = 0
        save_enabled = 0
        crtc = 0x0
        config = 0x1aef650
#5  0x00007f4b70b80f35 in drv_leave_vt (arg=0x1aee340) at vmwgfx_driver.c:1271
        pScrn = 0x1aee340
        ms = 0x1aeea60
#6  0x00007f4b70b81111 in drv_close_screen (pScreen=0x1aed440)
    at vmwgfx_driver.c:1327
        pScrn = 0x1aee340
        ms = 0x1aeea60
#7  0x000000000052d93f in RRCloseScreen (pScreen=0x1aed440) at randr.c:108
        pScrPriv = 0x1b6eeb0
        j = -1
#8  0x0000000000580a44 in xf86CrtcCloseScreen (screen=0x1aed440)
    at xf86Crtc.c:743
        scrn = 0x1aee340
        config = 0x1aef650
        o = 0
        c = 0
#9  0x00000000005d4a07 in DGACloseScreen (pScreen=0x1aed440) at xf86DGA.c:288
        pScreenPriv = 0x1b761c0
#10 0x00000000005bcbd1 in CMapCloseScreen (pScreen=0x1aed440) at xf86cmap.c:250
No locals.
#11 0x0000000000551644 in XvCloseScreen (pScreen=0x1aed440) at xvmain.c:309
        pxvs = 0x1b75bd0
#12 0x00000000005a7573 in xf86XVCloseScreen (pScreen=0x1aed440)
    at xf86xv.c:1168
        pScrn = 0x1aee340
        pxvs = 0x1b75bd0
        ScreenPriv = 0x1afbee0
        pa = 0x1b85290
        c = 2
#13 0x000000000046fdd6 in SyncCloseScreen (pScreen=0x1aed440) at misync.c:156
        pScreenPriv = 0x1c47758
#14 0x000000000055e809 in CursorCloseScreen (pScreen=0x1aed440) at cursor.c:205
        cs = 0x1c0ce80
        ret = 32766
        close_proc = 0x55e77d <CursorCloseScreen>
        display_proc = 0x55e599 <CursorDisplayCursor>
#15 0x00000000004f226d in AnimCurCloseScreen (pScreen=0x1aed440)
    at animcur.c:100
        as = 0x1c47778
        ret = 0
#16 0x0000000000559660 in compCloseScreen (pScreen=0x1aed440) at compinit.c:86
        cs = 0x1b85020
        ret = 0
#17 0x00000000004f4f04 in present_close_screen (screen=0x1aed440)
    at present_screen.c:70
        screen_priv = 0x1c0db60
#18 0x00007f4b72fe0e76 in glxCloseScreen (pScreen=0x1aed440)
    at glxscreens.c:171
        pGlxScreen = 0x1c0e030
#19 0x00000000006176a5 in dix_main (argc=3, argv=0x7ffe301caa28,
    envp=0x7ffe301caa48) at main.c:325
        i = 0
        alwaysCheckForInput = {0, 1}
#20 0x000000000061e67a in main (argc=3, argv=0x7ffe301caa28,
    envp=0x7ffe301caa48) at stubmain.c:34
No locals.
(EE)
(EE) Backtrace:
(EE) 0: /usr/local/lib/xorg/Xorg (xorg_backtrace+0xb0) [0x431177]
(EE) 1: /usr/local/lib/xorg/Xorg (0x400000+0x35f40) [0x435f40]
(EE) 2: /lib/libpthread.so.0 (0x7f4b742ff000+0x10030) [0x7f4b7430f030]
(EE) 3: /usr/local/lib/xorg/Xorg (0x400000+0x10fbfe) [0x50fbfe]
(EE) 4: /usr/local/lib/xorg/Xorg (0x400000+0x10fc46) [0x50fc46]
(EE) 5: /usr/local/lib/xorg/Xorg (RROutputIsLeased+0x35) [0x50ffa0]
(EE) 6: /usr/local/lib/xorg/Xorg (xf86DPMSSet+0x86) [0x585de8]
(EE) 7: /usr/local/lib/xorg/modules/drivers/vmware_drv.so (0x7f4b70b5e000+0x232cb) [0x7f4b70b812cb]
(EE) 8: /usr/local/lib/xorg/modules/drivers/vmware_drv.so (0x7f4b70b5e000+0x22f35) [0x7f4b70b80f35]
(EE) 9: /usr/local/lib/xorg/modules/drivers/vmware_drv.so (0x7f4b70b5e000+0x23111) [0x7f4b70b81111]
(EE) 10: /usr/local/lib/xorg/Xorg (0x400000+0x12d93f) [0x52d93f]
(EE) 11: /usr/local/lib/xorg/Xorg (0x400000+0x180a44) [0x580a44]
(EE) 12: /usr/local/lib/xorg/Xorg (0x400000+0x1d4a07) [0x5d4a07]
(EE) 13: /usr/local/lib/xorg/Xorg (0x400000+0x1bcbd1) [0x5bcbd1]
(EE) 14: /usr/local/lib/xorg/Xorg (0x400000+0x151644) [0x551644]
(EE) 15: /usr/local/lib/xorg/Xorg (0x400000+0x1a7573) [0x5a7573]
(EE) 16: /usr/local/lib/xorg/Xorg (0x400000+0x6fdd6) [0x46fdd6]
(EE) 17: /usr/local/lib/xorg/Xorg (0x400000+0x15e809) [0x55e809]
(EE) 18: /usr/local/lib/xorg/Xorg (0x400000+0xf226d) [0x4f226d]
(EE) 19: /usr/local/lib/xorg/Xorg (0x400000+0x159660) [0x559660]
(EE) 20: /usr/local/lib/xorg/Xorg (0x400000+0xf4f04) [0x4f4f04]
(EE) 21: /usr/local/lib/xorg/modules/extensions/libglx.so (0x7f4b72fcf000+0x11e76) [0x7f4b72fe0e76]
(EE) 22: /usr/local/lib/xorg/Xorg (0x400000+0x2176a5) [0x6176a5]
(EE) 23: /usr/local/lib/xorg/Xorg (0x400000+0x21e67a) [0x61e67a]
(EE) 24: /lib/libc.so.6 (__libc_start_main+0x15a) [0x7f4b7345c09d]
(EE) 25: /usr/local/lib/xorg/Xorg (_start+0x2a) [0x420c3a]
(EE)
(EE) Segmentation fault at address 0x1000000b1
(EE)
Fatal server error:
(EE) Caught signal 11 (Segmentation fault). Server aborting

[andyj]

I noticed that when I compared the old and new vmware_drv.so, libudev.so.0 is a new dependency. It seems to me that either it's trying to do something it wasn't doing before that is available on other distros but not TC, or it's doing something it was doing before but is doing it differently now. Or, the udev dependency shouldn't be there at all and a new configure option needs to be turned on or off when building? When I compare the ldd results to other distros I see other slight differences too, but nothing that jumps out for attention.

[curaga]

This looks similar:
https://bugs.freedesktop.org/show_bug.cgi?id=106772

[Juanito]

Thanks for the suggestion, but it still segfaults after recompiling xorg-server with the patch

[Juanito]

accepted as a bug by freedesktop and assigned