VirtualBox

Ticket #20055 (new defect)

Opened 6 months ago

Last modified 2 weeks ago

Kernel modules do not build with linux mainline kernel (5.10-rc4)

Reported by: jekb Owned by:
Component: other Version: VirtualBox 6.1.16
Keywords: Cc:
Guest type: Linux Host type: Linux

Description

This source no longer builds in Linux mainline, due to the symbol alloc_vm_area being undefined in memobj-r0drv-linux.c.

alloc_vm_area was removed from the kernel:  https://lkml.org/lkml/2020/9/30/1124

Attachments

make.log Download (6.9 KB) - added by jekb 6 months ago.
linux-5.10-r0drv-memobj-fix-r0.patch Download (3.2 KB) - added by aeichner 5 months ago.
Fix to get the kernel module compiled and working on 5.10
linux-5.10-address-space-fixes.patch Download (731 bytes) - added by frispete 5 months ago.
This fixes the shared folder driver
linux-5.10-framebuffer-fixes.patch Download (1.5 KB) - added by frispete 5 months ago.
The beginning of the framebuffer fixes
.config Download (224.8 KB) - added by Pok 5 months ago.
Working kernel config
vbox-setup.log.1 Download (213.2 KB) - added by Thomas-H 3 months ago.
ubunut 20.10 5.10.12 setup log

Change History

Changed 6 months ago by jekb

comment:1 Changed 6 months ago by sergiomb

+1

comment:3 Changed 5 months ago by microcris

One more on Linux CRIS-DELL 5.10.0-1005-oem #6-Ubuntu SMP Wed Nov 25 15:20:02 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux

/var/lib/dkms/virtualbox/6.1.16/build/make.log:

  /var/lib/dkms/virtualbox/6.1.16/build/vboxdrv/r0drv/linux/memobj-r0drv-linux.c: In function ‘rtR0MemObjLinuxVMap’:
/var/lib/dkms/virtualbox/6.1.16/build/vboxdrv/r0drv/linux/memobj-r0drv-linux.c:560:34: error: implicit declaration of function ‘alloc_vm_area’ [-Werror=implicit-function-declaration]
  560 |                 pMemLnx->pArea = alloc_vm_area(pMemLnx->Core.cb, papPtes); /* Note! pArea->nr_pages is not set. */
      |                                  ^~~~~~~~~~~~~
/var/lib/dkms/virtualbox/6.1.16/build/vboxdrv/r0drv/linux/memobj-r0drv-linux.c:560:32: warning: assignment to ‘struct vm_struct *’ from ‘int’ makes pointer from integer without a cast [-Wint-conversion]
  560 |                 pMemLnx->pArea = alloc_vm_area(pMemLnx->Core.cb, papPtes); /* Note! pArea->nr_pages is not set. */

Last edited 5 months ago by microcris (previous) (diff)

Changed 5 months ago by aeichner

Fix to get the kernel module compiled and working on 5.10

comment:4 follow-up: ↓ 5 Changed 5 months ago by aeichner

Could you please try the attached patch? It should fix the issue.

comment:5 in reply to: ↑ 4 Changed 5 months ago by microcris

Replying to aeichner:

Could you please try the attached patch? It should fix the issue.

It is working. Tested with a Windows 7 virtual machine. Thank you.

cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ sudo patch -p3 < ./linux-5.10-r0drv-memobj-fix-r0.patch 
patching file r0drv/linux/memobj-r0drv-linux.c
cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ 
cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ 
cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ 
cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ 
cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ 
cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ 
cris@CRIS-DELL:/usr/src/virtualbox-6.1.16$ sudo sudo apt-get install virtualbox-dkms 
Reading package lists... Done
Building dependency tree       
Reading state information... Done
virtualbox-dkms is already the newest version (6.1.16-dfsg-4build1).
virtualbox-dkms set to manually installed.
0 to upgrade, 0 to newly install, 0 to remove and 3 not to upgrade.
4 not fully installed or removed.
After this operation, 0 B of additional disk space will be used.
Do you want to continue? [Y/n] y
Setting up virtualbox-dkms (6.1.16-dfsg-4build1) ...
Removing old virtualbox-6.1.16 DKMS files...

------------------------------
Deleting module version: 6.1.16
completely from the DKMS tree.
------------------------------
Done.
Loading new virtualbox-6.1.16 DKMS files...
Building for 5.10.0-1006-oem
Building initial module for 5.10.0-1006-oem
Secure Boot not enabled on this system.
Done.

vboxdrv.ko:
Running module version sanity check.
 - Original module
   - No original module exists within this kernel
 - Installation
   - Installing to /lib/modules/5.10.0-1006-oem/updates/dkms/

vboxnetadp.ko:
Running module version sanity check.
 - Original module
   - No original module exists within this kernel
 - Installation
   - Installing to /lib/modules/5.10.0-1006-oem/updates/dkms/

vboxnetflt.ko:
Running module version sanity check.
 - Original module
   - No original module exists within this kernel
 - Installation
   - Installing to /lib/modules/5.10.0-1006-oem/updates/dkms/

depmod...

DKMS: install completed.
Setting up virtualbox (6.1.16-dfsg-4build1) ...
Job for vboxweb.service failed because the service did not take the steps required by its unit configuration.
See "systemctl status vboxweb.service" and "journalctl -xe" for details.
Setting up virtualbox-ext-pack (6.1.16-1) ...
virtualbox-ext-pack: downloading: http://download.virtualbox.org/virtualbox/6.1.16/Oracle_VM_VirtualBox_Extension_Pack-6.1.16.vbox-extpack
The file will be downloaded into /usr/share/virtualbox-ext-pack
License accepted.
0%...10%...20%...30%...40%...50%...60%...70%...80%...90%...100%
Successfully installed "Oracle VM VirtualBox Extension Pack".
Setting up virtualbox-qt (6.1.16-dfsg-4build1) ...

comment:6 Changed 5 months ago by frispete

Thanks, but no cigar here.

These changes seems to be not enough.

Attaching the shared folder fixes, and the beginning of framebuffer fixes. ttm initialization is changed significantly as well (which is, where I stopped now.)

Changed 5 months ago by frispete

This fixes the shared folder driver

Changed 5 months ago by frispete

The beginning of the framebuffer fixes

comment:8 follow-ups: ↓ 9 ↓ 12 Changed 5 months ago by klaus

The host side fixes for supporting Linux 5.10 are in test builds starting with r141850.

The guest side fix isn't yet in a test build.

comment:9 in reply to: ↑ 8 Changed 5 months ago by Didier

Replying to klaus:

The host side fixes for supporting Linux 5.10 are in test builds starting with r141850.

Could you make this test build available for everybody ?

comment:10 Changed 5 months ago by Oleksii

In my case any attempt to start any VM ends with crash: Kernel 5.10.0/5.10.1

[   33.987753] SUPR0GipMap: fGetGipCpu=0x1b
[   34.708660] BUG: kernel NULL pointer dereference, address: 0000000000000020
[   34.708665] #PF: supervisor read access in kernel mode
[   34.708667] #PF: error_code(0x0000) - not-present page
[   34.708668] PGD 0 P4D 0 
[   34.708673] Oops: 0000 [#1] PREEMPT SMP NOPTI
[   34.708676] CPU: 4 PID: 2796 Comm: EMT Tainted: G S   U     O      5.10.0 #3
[   34.708678] Hardware name: LENOVO 20NXCTO1WW/20NXCTO1WW, BIOS N2JET92W (1.70 ) 08/31/2020
[   34.708683] RIP: 0010:__lock_acquire.constprop.0+0x30f/0x530
[   34.708686] Code: c0 74 10 44 8b 25 99 61 78 01 45 85 e4 0f 84 be 55 97 00 45 31 e4 48 83 c4 18 44 89 e0 5b 5d 41 5c 41 5d 41 5e 41 5f c3 89 f0 <48> 8b 44 c7 08 48 85 c0 0f 85 41 fd ff ff e9 1a fd ff ff 44 8b 54
[   34.708688] RSP: 0018:ffffb2c4c2097c98 EFLAGS: 00010097
[   34.708690] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
[   34.708691] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000018
[   34.708693] RBP: 0000000000000018 R08: 0000000000000000 R09: 0000000000000000
[   34.708694] R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000000000
[   34.708696] R13: ffff900135440000 R14: 0000000000000000 R15: 0000000000000246
[   34.708698] FS:  00007faac04e7640(0000) GS:ffff900490500000(0000) knlGS:0000000000000000
[   34.708700] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   34.708701] CR2: 0000000000000020 CR3: 0000000197c9c002 CR4: 00000000003706e0
[   34.708703] Call Trace:
[   34.708709]  lock_acquire+0x11f/0x160
[   34.708712]  ? __apply_to_page_range+0x2e4/0x6a0
[   34.708726]  ? rtR0TermNative+0xf0/0x250 [vboxdrv]
[   34.708730]  _raw_spin_lock+0x2c/0x40
[   34.708732]  ? __apply_to_page_range+0x2e4/0x6a0
[   34.708735]  __apply_to_page_range+0x2e4/0x6a0
[   34.708744]  ? rtR0TermNative+0xf0/0x250 [vboxdrv]
[   34.708754]  rtR0MemObjNativeProtect+0x80/0x100 [vboxdrv]
[   34.708763]  VBoxHost_RTR0MemObjProtect+0x86/0xc0 [vboxdrv]
[   34.708771]  supdrvIOCtl+0x31bc/0x35a0 [vboxdrv]
[   34.708779]  SUPR0Printf+0x1ee/0x300 [vboxdrv]
[   34.708783]  __x64_sys_ioctl+0x80/0xb0
[   34.708787]  do_syscall_64+0x2d/0x40
[   34.708790]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[   34.708792] RIP: 0033:0x7faadc1d0a67
[   34.708795] Code: d8 49 8d 3c 1c 49 39 c4 72 b9 e8 24 ff ff ff 85 c0 78 be 5b 5d 4c 89 e0 41 5c c3 0f 1f 84 00 00 00 00 00 b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d c9 63 0c 00 f7 d8 64 89 01 48
[   34.708796] RSP: 002b:00007faac04e5bd8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[   34.708799] RAX: ffffffffffffffda RBX: 00007faac0001010 RCX: 00007faadc1d0a67
[   34.708800] RDX: 00007faac0001010 RSI: 0000000000005684 RDI: 0000000000000007
[   34.708802] RBP: 00007faac04e5be0 R08: 0000000000000000 R09: 00000000fffffffc
[   34.708803] R10: 0000000000000000 R11: 0000000000000246 R12: 00007faac88eb2ff
[   34.708805] R13: 0000000000000000 R14: 00007faac04e5e90 R15: 0000000000000004
[   34.708809] Modules linked in: xt_MASQUERADE xt_state xt_conntrack ipt_REJECT nf_reject_ipv4 iptable_filter xt_tcpudp bpfilter nf_nat_h323 nf_conntrack_h323 nf_nat_pptp nf_conntrack_pptp nf_nat_tftp nf_conntrack_tftp nf_nat_sip nf_conntrack_sip nf_nat_irc nf_conntrack_irc nf_nat_ftp nf_conntrack_ftp iptable_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 af_packet bridge stp llc xfs rfcomm algif_hash algif_skcipher af_alg bnep hid_logitech_hidpp hid_logitech_dj snd_usb_audio snd_usbmidi_lib snd_rawmidi snd_seq_device snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic btusb btrtl btbcm btintel hid_generic bluetooth uvcvideo videobuf2_vmalloc videobuf2_memops videobuf2_v4l2 videobuf2_common videodev mc usbhid sd_mod joydev mousedev uas usb_storage scsi_mod elan_i2c iwlmvm intel_pmc_core_pltdrv iTCO_wdt intel_pmc_core iTCO_vendor_support mac80211 kvm_intel intel_wmi_thunderbolt wmi_bmof kvm sdhci_pci psmouse snd_hda_intel cqhci irqbypass crct10dif_pclmul iwlwifi crc32_pclmul
[   34.708873]  sdhci ghash_clmulni_intel snd_intel_dspcfg intel_rapl_msr atkbd snd_hda_codec snd_hwdep mmc_core snd_hda_core i915 snd_pcm i2c_i801 snd_timer thunderbolt i2c_smbus i2c_algo_bit cfg80211 drm_kms_helper mei_hdcp syscopyarea sysfillrect sysimgblt fb_sys_fops intel_gtt thinkpad_acpi wmi snd soundcore vboxnetadp(O) vboxnetflt(O) drm auth_rpcgss vboxdrv(O) i2c_core agpgart sunrpc bpf_preload ip_tables x_tables
[   34.708906] CR2: 0000000000000020
[   34.708909] ---[ end trace ae2da90b3349db38 ]---
[   34.708912] RIP: 0010:__lock_acquire.constprop.0+0x30f/0x530
[   34.708914] Code: c0 74 10 44 8b 25 99 61 78 01 45 85 e4 0f 84 be 55 97 00 45 31 e4 48 83 c4 18 44 89 e0 5b 5d 41 5c 41 5d 41 5e 41 5f c3 89 f0 <48> 8b 44 c7 08 48 85 c0 0f 85 41 fd ff ff e9 1a fd ff ff 44 8b 54
[   34.708915] RSP: 0018:ffffb2c4c2097c98 EFLAGS: 00010097
[   34.708917] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
[   34.708919] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000018
[   34.708920] RBP: 0000000000000018 R08: 0000000000000000 R09: 0000000000000000
[   34.708922] R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000000000
[   34.708923] R13: ffff900135440000 R14: 0000000000000000 R15: 0000000000000246
[   34.708925] FS:  00007faac04e7640(0000) GS:ffff900490500000(0000) knlGS:0000000000000000
[   34.708926] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   34.708928] CR2: 0000000000000020 CR3: 0000000197c9c002 CR4: 00000000003706e0
[   34.708930] note: EMT[2796] exited with preempt_count 1

It looks like it crashes at apply_to_page_range itself, not at helper function.

Last edited 5 months ago by Oleksii (previous) (diff)

comment:11 Changed 5 months ago by klaus

The latest test builds have all those fixes (if you want the CentOS 8.4 fixes you need to download the separate VBoxGuestAdditions ISO file, because the packages will not have the latest one included).

comment:12 in reply to: ↑ 8 Changed 5 months ago by sergiomb

Last edited 5 months ago by sergiomb (previous) (diff)

comment:13 Changed 5 months ago by Oleksii

It looks like apply_to_page_range which is used in the patch doesn't work as expected with some kernel configurations. Maybe related to split page table lock, maybe not. In case this patch (memobj-r0drv-linux) works for you, maybe you can share your current kernel config to compare options?

Last edited 5 months ago by Oleksii (previous) (diff)

Changed 5 months ago by Pok

Working kernel config

comment:14 follow-up: ↓ 15 Changed 5 months ago by Pok

Not sure if it helps, but I have applied patch linux-5.10-r0drv-memobj-fix-r0.patch to a binary installation of 6.1.16 (on Slackware64 current) and now the driver compiles and VMs thus far seem to be running OK. The kernel config (5.10.1) is the one installed by the distro, attached. Switching to kernel config 5.10.2 is still working.

Last edited 5 months ago by Pok (previous) (diff)

comment:15 in reply to: ↑ 14 Changed 5 months ago by sergiomb

Replying to Pok:

Not sure if it helps, but I have applied patch linux-5.10-r0drv-memobj-fix-r0.patch to a binary installation of 6.1.16 (on Slackware64 current) and now the driver compiles and VMs thus far seem to be running OK. The kernel config (5.10.1) is the one installed by the distro, attached. Switching to kernel config 5.10.2 is still working.

Correct, when we enable CONFIG_DRM_VBOXVIDEO=m, CONFIG_VBOXGUEST=m and CONFIG_VBOXSF_FS=m in kernel, we don't need drm and shared folder patches

comment:16 follow-up: ↓ 28 Changed 5 months ago by Oleksii

Thank you very much. It looks like disabling CONFIG_DEBUG_*LOCK* helps. Not sure it has something to do to VirtualBox code

comment:17 Changed 5 months ago by GoofyXL

So, when should we expect a newer version that supports 5.10?

comment:18 Changed 5 months ago by MasterCATZ

thanks for the patches VB 6.1.16 r140961 seems to be running with Kernel 5.10 now

any idea what is causing this

depmod: ERROR: /etc/depmod.d/ubuntu.conf:1: ignoring bad line starting with 'extra'

comment:19 Changed 5 months ago by jarv

There is a bug in the linux-5.10-framebuffer-fixes.patch, there should be curly braces around the added code in the #ifdef blocks because of the if statement, the old code had only line whereas the new code has two lines added so the second line will always be executed. Therefore will always default to SYSTEM and never VRAM placement

comment:20 Changed 5 months ago by drankinatty

Here is another bit of info on 5.10.3 on Archlinux with testbuild 6.1.17-141968:

Archlinux guest, running on 64-bit Archlinux host, after upgrading both host and guest to Linux 5.10.3 and using the virtualbox testbuild 6.1.17-141968.

Kernel modules on the host built without a problem, the guest launched without a problem running the 5.9.14 kernel. I upgraded the guest kernel to 5.10.3 and rebuilt the guest-addition modules from the 6.1.17-141968 guest iso. Modules for 5.10.3 built fine.

On reboot, the guest had a kernel panic and locked. "Kenel panic - not syncing: Fatal exception in interrupt". Since the guest is dead I cannot get any logs from it, but I do have a screenshot of the end of the kernel-panic if that would be helpful.

comment:21 follow-up: ↓ 24 Changed 4 months ago by iceviking

Maybe this has been answered elsewhere, but why can't the guest additions just directly use the guest/sf drivers included in the 5.10.x kernels? Apparently those are maintained by mainline kernel folks.

comment:22 Changed 4 months ago by drankinatty

Updated Archlinux Host to 6.1.18. (took snapshot of guest this time) and proceeded to update Archlinux guest from Linux 5.9 to 5.10.9. After install, all Linux_Guest_Additions were built from the 6.1.18 host .iso for the newly installed 5.10.9 kernel. All modules built fine, no error.

On reboot, Kernel Panic again. Just the same as it was with the 6.1.17 testbuild in my last post above. I have accessing the guest headless over the network via rdesktop as I have for the last 4 years, so the kernel panic occurs before I can capture anything. I did take a screenshot of the kernel panic in the rdesltop window. Not sure if it will help, but I'll attach it here. I had hope this problem would be fixed in 6.1.18 -- but it definitely is not fixed.

http://www.3111skyline.com/dl/bugs/vbox/5.10.9_kernel_panic.png

comment:23 Changed 4 months ago by KlCZ

Observed the same behaviour with Fedora hosts/guests there. Firstly I upgraded host (Fedora32) to kernel 5.10.8 and VirtualBox 6.1.18 (with appropriate ExtensionPack) - everything working smoothly. Then I upgraded guest (F32, kernel 5.10.8, VBox 6.1.18 as well) and I noticed kernel panic after reboot of the guest. Downgrading the guest into 5.9 kernel series solved this crash.

What is interesting - on the RHEL/CentOS hosts with older kernels (RHEL7 = 3.10.0-1160.11.1, RHEL8 = 4.18.0-240.10.1) the same upgrading procedure into VirtualBox 6.1.18 (both hosts/guests sides) works fine, without booting troubles, even if the guests are the same "Fedora based on kernel 5.10".

comment:24 in reply to: ↑ 21 ; follow-up: ↓ 25 Changed 4 months ago by sergiomb

Replying to iceviking:

Maybe this has been answered elsewhere, but why can't the guest additions just directly use the guest/sf drivers included in the 5.10.x kernels? Apparently those are maintained by mainline kernel folks.

drives included in the 5.10.x kernel, are not officially supported by virualbox and I think that vboxvideo aren't sync with last Virtualbox release , and is breaking things .

comment:25 in reply to: ↑ 24 ; follow-up: ↓ 26 Changed 4 months ago by AlexKemp

Replying to sergiomb:

Replying to iceviking:

Maybe this has been answered elsewhere, but why can't the guest additions just directly use the guest/sf drivers included in the 5.10.x kernels? Apparently those are maintained by mainline kernel folks.

drives included in the 5.10.x kernel, are not officially supported by virualbox and I think that vboxvideo aren't sync with last Virtualbox release , and is breaking things .

The drivers included in the 5.10.x kernel don't support bi-directional copy/paste.

comment:26 in reply to: ↑ 25 ; follow-up: ↓ 30 Changed 4 months ago by sergiomb

Replying to AlexKemp:

Replying to sergiomb:

Replying to iceviking:

Maybe this has been answered elsewhere, but why can't the guest additions just directly use the guest/sf drivers included in the 5.10.x kernels? Apparently those are maintained by mainline kernel folks.

drives included in the 5.10.x kernel, are not officially supported by virualbox and I think that vboxvideo aren't sync with last Virtualbox release , and is breaking things .

The drivers included in the 5.10.x kernel don't support bi-directional copy/paste.

bi-directional copy/paste works for me with 5.10.x kernel and vboxvideo crash is an issue with recent KDE update AFAIK , not kernel related and only when 3D acceleration setting is enabled .

Sometimes copy/paste stops to work after suspend and resume vm , please try systemctl restart vboxservice

comment:27 Changed 4 months ago by klaus

The copy/paste functionality (guest side) is in the userland part of the Guest Additions (and thus restarting the service may help, as pointed out in previous comment).

It uses the vboxguest.ko driver functionality, but that's very generic data transport which should not introduce any additional incompatibilities. The transport may break (IIRC shouldn't on suspend/resume, but it does on save state/restart) and that's why the userland code contains logic to re-connect. VirtualBox 6.1.18 contains improvements to X11 clipboard handling (making it behave more consistently and probably also improving reliability). Not sure if you're using this new code already.

comment:28 in reply to: ↑ 16 Changed 4 months ago by ozugzug

Replying to Oleksii:

Thank you very much. It looks like disabling CONFIG_DEBUG_*LOCK* helps. Not sure it has something to do to VirtualBox code

Thank you for that! Looks like CONFIG_DEBUG_SPINLOCK=y is causing the kernel panic (and =n doesn't). Made a new ticket as https://www.virtualbox.org/ticket/20163

comment:29 Changed 3 months ago by Thomas-H

virtualbox 6.1.18-142142 stopped working in kernel 5.10.12 and 5.10.14n ubuntu 20.10. it still works fine in kernel 5.10.10 (from  https://kernel.ubuntu.com/~kernel-ppa/mainline/). I have attached the setup-log from /sbin/vboxconfig.

as I'm not familiar with the vbox build process, please let me know how I can provide more information in case needed.

Last edited 3 months ago by Thomas-H (previous) (diff)

Changed 3 months ago by Thomas-H

ubunut 20.10 5.10.12 setup log

comment:30 in reply to: ↑ 26 Changed 3 months ago by drankinatty

Replying to sergiomb:

Replying to AlexKemp:

Replying to sergiomb:

Replying to iceviking:

Maybe this has been answered elsewhere, but why can't the guest additions just directly use the guest/sf drivers included in the 5.10.x kernels? Apparently those are maintained by mainline kernel folks.

drives included in the 5.10.x kernel, are not officially supported by virualbox and I think that vboxvideo aren't sync with last Virtualbox release , and is breaking things .

The drivers included in the 5.10.x kernel don't support bi-directional copy/paste.

bi-directional copy/paste works for me with 5.10.x kernel and vboxvideo crash is an issue with recent KDE update AFAIK , not kernel related and only when 3D acceleration setting is enabled .

Note, the guest crash on 5.10 I am experiencing has nothing to do with KDE. I run the host booting to multi-user.target -- there is no desktop running on the host. The guests are started headless and accessed via LAN. The Archlinux guests also boot to multi-user.target and launch Fluxbox via startx. So there is no KDE even installed on either system.

Something is simply not syncing causing the kernel panic on the guest when they are upgraded to 5.10.x.

Right now the host is running 5.10.15 without any problems. The guests are stuck with the 5.9.14 kernel and any attempt to update the guest kernel results in the same kernel panic.

comment:31 Changed 2 months ago by AlexKemp

Using the 5.10.15 kernel on host/guest

If I have Guest Additions installed and I choose the Graphics Controller VBOXVGA the Guest gets a segmentation fault and fails to boot. If I remove Guest Additions the guest boots successfully and I get Shared Folders and Auto-resize Guest Display. Bi-directional Copy/Paste doesn't work.

If I have Guest Additions installed and I choose the Graphics Controller VMSVGA the Guest boots successfully. I get Shared Folders, Auto-resize Guest Display and Bi-directional Copy/Paste.

comment:32 Changed 2 months ago by sergiomb

We (Fedora people) found a big regression on mesa-20.3.x and mesa-21.0.x , as VirtualBox use vmware video drive by default , also applies to VirtualBox guests vms on Linux .

About Bi-directional Copy/Paste also I'm seeing problems but I don't know what is it .

 https://src.fedoraproject.org/rpms/mesa/c/5e07467adaf8ecdc4c2b783171f4c357c99dd8db?branch=f33

 https://bodhi.fedoraproject.org/updates/?packages=mesa

  • Mon Feb 22 2021 Dave Airlie <airlied@…> - 21.0.0~rc5-2
  • fix sddm/vmware regression

comment:33 Changed 3 weeks ago by drankinatty

I can confirm this bug is mute now. Installed 6.20 for host on Archlinux. Fired up the Archlinux guest (still running 5.9.14 due to this bug), did a full upgrade to 5.11.16 on guest and installed guest additions for 6.1.20 - rebooted and No Kernel Panic.

So, at least for Archlinux, this issue for Linux guests with the 5.10 - 5.11 kernel is fixed (of course Linus released 5.12 day before yesterday...)

comment:34 Changed 2 weeks ago by sergiomb

IIRC we don't need any patch for kernel 5.12

Note: See TracTickets for help on using tickets.

www.oracle.com
ContactPrivacy policyTerms of Use