Re: [vfio-users] GPU passthrough errors with linux 5.1 and newer
On Sun, 21 Jul 2019 at 21:59, Zoltán Kővágó wrote: > > Hi, > > Recently my previously perfectly working GPU passthrough setup (with a > win8.1 x64 guest with OVMF) started to malfunction in various ways: > screen randomly turned off for a few seconds, BSOD with > VIDEO_TDR_FAILURE, 3d apps randomly crashing, not drawing the windows' > content, and graphical glitches (for example in furmark the OSD text > flickers). > > After fiddling around with various qemu versions, nvidia driver versions > on the guest, I figured out that with a linux 5.0 kernel it works fine, > but with 5.1 it randomly fails. I bisected it and it looks like the > culprit is the commit 4e103134b862 "KVM: x86/mmu: Zap only the relevant > pages when removing a memslot"[1]. I tried to revert in on top of 5.2.1 > but too many things changed in the meantime. Anyway, if I replace the > body of kvm_mmu_invalidate_zap_pages_in_memslot with > kvm_mmu_zap_all(kvm); it works again (probably with horrible performance > degradation). > > Did anyone experience anything like this? I'm using Alex's ACS override > patch, maybe it violates some assumption that the new code has? Hi, I noticed some changes that made 5.0 not working well when detecting screen speakers through hdmi, but this I didn't see anytime. My problem flew away with 5.1.15(the one I currently use), and no other spread. I never needed the ACS override patch in my setup, what happen if you try without it, does your groups comes wrong in any ways? Best regards. José. ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
Re: [vfio-users] remote display + vfio gpu?
On Sun, 30 Jun 2019 at 23:15, James Courtier-Dutton wrote: > Hi, > > Have you looked into the feasibility of what you are asking. > I see the main problem being the massive bandwidth that is needed. > Taking the output frames and passing them via shared memory is reasonable, > and that is what "looking glass" does. > Passing these same frames over the network is impractical. > There is just too much data to pass over the network. > > This is why various tools like QXL, Spice and VNC exist. Sorry if I didn't express myself correctly, but basically I'm trying to access one or 2 vms over spice, the thing is that I want to use a normal graphic card for 3d instead of the qxl device. If the performance is bad or good, as it's for testing I simply can live with bad performance. Best regards. José. ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
Re: [vfio-users] remote display + vfio gpu?
On Sun, 30 Jun 2019 at 18:55, Patrick O'Callaghan wrote: > > On Sun, 2019-06-30 at 16:35 +0300, José Ramón Muñoz Pekkarinen wrote: > > Hi vfio users, > > > > I've been taking a look recently to the looking > > glass project[1], as well as spice display capabilities of > > qemu, I was wondering if it's possible to have a linux > > guest with only remote display and use a gpu passed > > through for 3d rendering either through qxl driver with > > dri prime, or just removing the qxl device after install. > > > > I did a try with qxl + dri prime, though glxinfo came > > from qxl no matter of the env variable. > > > > [1] https://looking-glass.hostfission.com/ > > According to that page that only Windows 10 guests are supported at > present. Oh, sure, my interest is to implement this kind of vm in linux and windows 10, but I'm looking for alternative ways rather than using looking glass, as it seems it enforces to use shared memory. I see that using egl-headless I could get some sort of capabilities, but not all of the gpu, and using spice only I don't find a way to redirect the vga output to the remote display. Best regards. José. ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
[vfio-users] remote display + vfio gpu?
Hi vfio users, I've been taking a look recently to the looking glass project[1], as well as spice display capabilities of qemu, I was wondering if it's possible to have a linux guest with only remote display and use a gpu passed through for 3d rendering either through qxl driver with dri prime, or just removing the qxl device after install. I did a try with qxl + dri prime, though glxinfo came from qxl no matter of the env variable. [1] https://looking-glass.hostfission.com/ Thanks! José Pekkarinen. ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
Re: [vfio-users] issues about igd-passthrough using vfio-pci
On Tuesday 13 September 2016 21:38:01 Alex Williamson wrote: > [adding the list back] > > On Tue, Sep 13, 2016 at 9:17 PM, fulaiyang wrote: > > hello, > > > > my kernel does include simplefb:CONFIG_FB_SIMPLE=y.QEMU does not > > > > print any messages when started but the top tree show that qemu'cpu usage > > is always about 100%.I have confirmed that the windows 7 os does not > > boot.I > > don't know how I can get other qemu information,could you tell me? thanks. > > > >PID USER PR NIVIRTRESSHR S *%CPU* %MEM TIME+ > > > > COMMAND > > > > 9542 root 20 0 2768476 2.028g 11848 S * 99.7* 53.3 3:48.82 > > > > qemu-system-x86 > > Try removing the modprobe.blacklist and video options from your kernel > command line (the unsafe interrupts thing isn't necessary on your system > either). After boot, IGD should be bound to i915. Unbind it, bind to > vfio-pci, and try QEMU again. Since you have simplefb in your kernel, I > don't trust that it's not claiming device resources as you're using it now. Hi Alex, Yang. I've been toying around with this case again, and it seems I got some output that can be of any help. I think it points that i915 is reluctant to free the resources, but it can highlight something else to your eyes. [ 225.155202] [ cut here ] [ 225.155217] WARNING: CPU: 2 PID: 7101 at drivers/gpu/drm/drm_crtc.c:5939 drm_mode_config_cleanup+0x20f/0x230 [drm] [ 225.155218] Modules linked in: vfio_pci vfio_iommu_type1 vfio_virqfd vfio drbg ansi_cprng ctr ccm bridge stp llc af_packet ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat nf_nat_ipv4 nf_nat nf_conntrack_ipv4 nf_defrag_ipv4 iptable_filter ip_tables nf_conntrack_ipv6 nf_defrag_ipv6 xt_conntrack nf_conntrack ip6table_filter ip6_tables snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic binfmt_misc i915 arc4 iwlmvm mac80211 loop i2c_algo_bit drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops drm x86_pkg_temp_thermal intel_powerclamp iwlwifi coretemp kvm_intel kvm joydev snd_hda_intel uvcvideo mousedev snd_hda_codec btusb cfg80211 btrtl videobuf2_vmalloc videobuf2_memops btbcm btintel snd_hwdep videobuf2_v4l2 rtsx_pci_sdmmc videobuf2_core bluetooth mmc_core videodev snd_hda_core [ 225.155245] snd_pcsp rtsx_pci_ms snd_pcm memstick media irqbypass crc32c_intel snd_timer psmouse ghash_clmulni_intel snd i2c_hid efi_pstore evdev wmi video input_leds efivars i2c_i801 serio_raw rtsx_pci i2c_core backlight intel_lpss_acpi intel_lpss thermal tpm_crb soundcore mfd_core button battery ac acpi_pad efivarfs unix dm_zero dm_thin_pool dm_persistent_data dm_bio_prison dm_service_time dm_round_robin dm_queue_length dm_multipath dm_log_userspace cn dm_flakey dm_delay xts aesni_intel glue_helper lrw gf128mul ablk_helper cryptd aes_x86_64 cbc sha256_generic scsi_transport_iscsi r8169 mii fuse nfs lockd grace sunrpc fscache ext4 jbd2 mbcache multipath linear raid10 raid1 raid0 dm_raid raid456 libcrc32c md_mod async_raid6_recov async_memcpy async_pq async_xor xor async_tx raid6_pq dm_snapshot [ 225.155275] dm_bufio dm_crypt dm_mirror dm_region_hash dm_log dm_mod hid_generic usbhid xhci_pci xhci_hcd ohci_hcd uhci_hcd usb_storage ehci_pci ehci_hcd usbcore usb_common scsi_transport_fc sr_mod cdrom sg sd_mod ata_piix ahci libahci sata_sx4 pata_oldpiix [ 225.155286] CPU: 2 PID: 7101 Comm: vfio-bind Not tainted 4.7.2 #5 [ 225.155287] Hardware name: PC Specialist Limited N24_25JU/N24_25JU, BIOS 5.11 12/14/2015 [ 225.155288] 88035a43fc08 812d18c2 [ 225.155290] 88035a43fc48 8105b211 1733a0928530 [ 225.155292] 88035bfbe498 88035bfbe000 88035bfbe340 88035e65bb80 [ 225.155293] Call Trace: [ 225.155297] [] dump_stack+0x67/0x95 [ 225.155299] [] __warn+0xd1/0xf0 [ 225.155301] [] warn_slowpath_null+0x1d/0x20 [ 225.155310] [] drm_mode_config_cleanup+0x20f/0x230 [drm] [ 225.155331] [] intel_modeset_cleanup+0x80/0xa0 [i915] [ 225.155347] [] i915_driver_unload+0x74/0x1d0 [i915] [ 225.155354] [] drm_dev_unregister+0x29/0xb0 [drm] [ 225.155361] [] drm_put_dev+0x23/0x60 [drm] [ 225.155370] [] i915_pci_remove+0x15/0x20 [i915] [ 225.155372] [] pci_device_remove+0x39/0xc0 [ 225.155375] [] __device_release_driver+0x9a/0x140 [ 225.155376] [] device_release_driver+0x23/0x30 [ 225.155377] [] unbind_store+0xe7/0x140 [ 225.155379] [] drv_attr_store+0x25/0x30 [ 225.155381] [] sysfs_kf_write+0x37/0x40 [ 225.155382] [] kernfs_fop_write+0x118/0x190 [ 225.155384] [] __vfs_write+0x28/0x120 [ 225.155386] [] ? security_file_permission+0x3d/0xc0 [ 225.155388] [] ? percpu_down_read+0x12/0x60 [ 225.155390] [] vfs_write+0xb8/0x1a0 [ 225.155391] [] SyS_write+0x46/0xb0 [ 225.155393] [] do_syscall_64+0x61/0x110 [ 225.155395] [] entry_SYSCALL64_slow_path+0x25/0x25 [ 225.155396] ---[ end trace 4631a3679da5de09 ]--- [ 225.157
Re: [vfio-users] IGD Skylake rom?
> This means there's another (non-pci) driver claiming resources from the > device, that's going to be a problem. If you prevent i915 from binding to > igd then you also need to prevent the other framebuffers from getting > resources. Use video=efifb:off,vesafb:off to disable both efi and non-efi > framebuffer drivers. Do you know if it's possible to disable simplefb this way? video=efifb:off,vesafb:off,simplefb:off or video=efifb:off,vesafb:off,simple- framebuffer:off doesn't seem to work. Thanks! José. ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
Re: [vfio-users] IGD Skylake rom?
On Saturday 10 September 2016 17:02:44 José Ramón Muñoz Pekkarinen wrote: > Yep, there is no other graphic card on this machine, might this be a > problem of efi fb or sime fb jumping on the igd? > > Thanks! > > José. > > On 10 Sep 2016 18:52, "Alex Williamson" wrote: > > On Sat, Sep 10, 2016 at 9:51 AM, José Ramón Muñoz Pekkarinen < > > > > koali...@gmail.com> wrote: > >> Hi Alex, > >> > >> Sorry to say I'm using 4.7.2, so that might not be the problem, any other > >> idea? > > > > Is IGD the host boot graphics? It needs to be. Some output for perusal: qemu call: qemu-system-x86_64 -enable-kvm -vga none -cpu host -smp 3,cores=3,threads=1 \ -m 7G -mem-path /mnt/hugepages \ -device vfio-pci,host=00:02.0,addr=02.0 \ -netdev user,id=user.0,smb=/home/pekkari/share -device e1000,netdev=user.0 \ -device ich9-usb-uhci3,id=uhci -device usb-ehci,id=ehci -device nec- usb-xhci,id=xhci \ -object input-linux,id=kbd,evdev=/dev/input/by-path/platform-i8042- serio-0-event-kbd,grab_all=on \ -object input-linux,id=mouse,evdev=/dev/input/by-path/platform-i8042- serio-2-event-mouse \ -drive file=win7.img,format=raw,cache=none,aio=native,if=virtio output: qemu-system-x86_64: -device vfio-pci,host=00:02.0,addr=02.0: Failed to mmap :00:02.0 BAR 2. Performance may be slow qemu-system-x86_64: -device vfio-pci,host=00:02.0,addr=02.0: IGD device :00:02.0 has no ROM, legacy mode disabled Could not initialize SDL(No available video device) - exiting lspci -k: 00:02.0 VGA compatible controller: Intel Corporation Sky Lake Integrated Graphics (rev 07) Subsystem: CLEVO/KAPOK Computer Sky Lake Integrated Graphics Kernel driver in use: vfio-pci Kernel modules: i915 uname -a: Linux ant 4.7.2 #1 SMP Fri Sep 2 08:54:30 BST 2016 x86_64 Intel(R) Core(TM) i7-6500U CPU @ 2.50GHz GenuineIntel GNU/Linux GRUB_CMDLINE_LINUX: GRUB_CMDLINE_LINUX="dolvm crypt_root=/dev/sda3 real_root=/dev/ant/root resume=/dev/ant/swap default_hugepagesz=1GB hugepagesz=1GB hugepages=8 vfio- pci.ids=8086:1916" dmesg | grep -e DMAR -e IOMMU [0.00] ACPI: DMAR 0x87B81888 A8 (v01 INTEL SKL 0001 INTL 0001) [0.037092] DMAR: Host address width 39 [0.037094] DMAR: DRHD base: 0x00fed9 flags: 0x0 [0.037101] DMAR: dmar0: reg_base_addr fed9 ver 1:0 cap 1cc40660462 ecap 7e3ff0505e [0.037105] DMAR: DRHD base: 0x00fed91000 flags: 0x1 [0.037109] DMAR: dmar1: reg_base_addr fed91000 ver 1:0 cap d2008c40660462 ecap f050da [0.037112] DMAR: RMRR base: 0x00874ff000 end: 0x008751efff [0.037115] DMAR: RMRR base: 0x008880 end: 0x008aff [0.037117] DMAR-IR: IOAPIC id 2 under DRHD base 0xfed91000 IOMMU 1 [0.037120] DMAR-IR: HPET id 0 under DRHD base 0xfed91000 [0.038634] DMAR-IR: Enabled IRQ remapping in xapic mode [1.968949] DMAR: No ATSR found [1.969158] DMAR: dmar0: Using Queued invalidat
Re: [vfio-users] IGD Skylake rom?
Yep, there is no other graphic card on this machine, might this be a problem of efi fb or sime fb jumping on the igd? Thanks! José. On 10 Sep 2016 18:52, "Alex Williamson" wrote: > On Sat, Sep 10, 2016 at 9:51 AM, José Ramón Muñoz Pekkarinen < > koali...@gmail.com> wrote: > >> Hi Alex, >> >> Sorry to say I'm using 4.7.2, so that might not be the problem, any other >> idea? >> > > Is IGD the host boot graphics? It needs to be. > ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
Re: [vfio-users] IGD Skylake rom?
Hi Alex, Sorry to say I'm using 4.7.2, so that might not be the problem, any other idea? Thanks! José. On 10 Sep 2016 17:58, "Alex Williamson" wrote: > On Sat, Sep 10, 2016 at 7:53 AM, José Ramón Muñoz Pekkarinen < > koali...@gmail.com> wrote: > >> >> >> Hi, >> >> >> >> I got qemu-2.7 last weekend, and I wanted to try the IGD passthorugh in >> my laptop. Unfortunately it seems my vgadoesn't use any rom, so I wonder if >> someone could provide any for the same cpu/vga, and inject it to the qemu >> command line. Model is the following: >> > > Sounds like you just don't have a new enough host kernel. If you need the > ROM (ie. "legacy mode"), you need a v4.6 or newer kernel. > ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users
[vfio-users] IGD Skylake rom?
Hi, I got qemu-2.7 last weekend, and I wanted to try the IGD passthorugh in my laptop. Unfortunately it seems my vgadoesn't use any rom, so I wonder if someone could provide any for the same cpu/vga, and inject it to the qemu command line. Model is the following: 00:02.0 VGA compatible controller: Intel Corporation Sky Lake Integrated Graphics (rev 07) (prog-if 00 [VGA controller]) lscpu If there is any other possibility, I'll be happy to try it. Thanks! José. ___ vfio-users mailing list vfio-users@redhat.com https://www.redhat.com/mailman/listinfo/vfio-users