Created attachment 129834 [details] xrandr --verbose Bug description: My laptop locks up hard (Alt-SysRq-B does not work) when I enable video modes 3840x2160 on both monitors and desktop is extended, not mirrored. If I use 3840x2160 on one monitor and 1920x1080 on the other, it seems stable. When I tried 2560x1440 on the 2nd monitor (the native resolution of my built-in laptop display) it also locked up. It locked up when I tried to use only one external monitor in 4k mode with my laptop screen in native resolution (2560x1440), but I did not experiment with that mode much, so I don't have as much specific information about it. The crash is very reliable and immediate. I tried some 4.10 kernel flavors, but I failed when trying to install drm-tip for testing with latest code. System environment: -- chipset: Intel(R) Core(TM) i7-6600U CPU @ 2.60GHz -- system architecture: 64-bit -- xf86-video-intel: -- xserver: xorg-server 2:1.18.4-1ubuntu6 -- mesa: -- libdrm: 2.4.70-1 -- kernel: 4.8.0-37-generic -- Linux distribution: Ubuntu 16.10 -- Machine or mobo model: Lenovo Carbon X1 -- Display connector: HDMI on OnePort dock Reproducing steps: Enable extended desktop on two monitors in 3840x2160 resolution. Additional info: $ uname -m ; uname -r x86_64 4.8.0-37-generic Lenovo Thinkpad Carbon X1 description: Notebook product: 20FB005LUS (LENOVO_MT_20FB_BU_Think_FM_ThinkPad X1 Carbon 4th) vendor: LENOVO version: ThinkPad X1 Carbon 4th width: 64 bits capabilities: smbios-2.8 dmi-2.8 smp vsyscall32 description: Motherboard product: 20FB005LUS vendor: LENOVO physical id: 0 version: SDK0J40697 WIN
Created attachment 129835 [details] intel_reg dump --all
Created attachment 129836 [details] vbios.dump
Correction: My monitors are connected via DP (version 1.2) through my ThinkPad OneLink+ dock, not HDMI.
A coworker with the same problem but on a different laptop and Intel video card has discovered that 2x4K monitors are stable in 30Hz refresh mode. I haven't had time to try this myself.
Confirmed that setting the mode to 30Hz on both monitors allows me to use 4k on both successfully. I'll try to generate and gather crash logs later this week.
Phil - any change for update information? See also instructions related to attachments from https://01.org/linuxgraphics/documentation/how-report-bugs.
Now with 2 months of experience, I can say: This crash _never_ occurs in 30hz refresh rate modes. This crash _always_ occurs in 60hz modes when both monitors are in 4k resolution. Today I tried to capture the failure reported in syslog but unfortunately syslog doesn't even flush when this hits. I don't have a UART console available for testing this at work where my 4k monitors live; maybe I can cobble something together with a raspi module. I initially tried to collect more detailed and up-to-date information for this bug report by following the build guides at https://01.org/linuxgraphics/documentation/build-guide-0. Somewhere during that process I hosed my boot config and could no longer boot at all. I may have some time coming up and can give it another go, this time on a secondary partition. I am happy to provide more info if there's something specific that would help you.
Please try current drm-tip branch of https://cgit.freedesktop.org/drm/drm-tip
Phil - please note previous comment, and mark status as REOPENED if problem still persist. Otherwise mark status as RESOLVED.
The problem worsened in Ubuntu 17.04 so much that I reverted to 16.10. My previous attempts to build drm-tip as requested were so far unsuccessful. I did attempt to build the drm-tip version of the driver against last month, but did not succeed. For safety I tried to build in a clean Ubuntu installation. The build ran for a long time before I had to give up and get back to my real work. I later tried to finish the build in a VM so I could carry on working in parallel but ran into too many problems there. I have been an embedded Linux developer for 9+ years with a bit of experience in kernel drivers and build systems. I don't usually have to deal with video displays so many of these external packages (mesa, glx, X, etc.) are foreign to me. But I find the instructions at https://01.org/linuxgraphics/documentation/build-guide-0 to be frustratingly vague in some places and broken in others. I kept a log of errors for a while hoping I could submit a patch to someone, but ultimately I could not confidently identify the someone to report it to. I managed to muddle through most of the build steps, but it's not clear to me if all of them are needed or some subset, nor how to proceed with testing afterward, should I ever reach the end. Do I really need to build a replacement kernel, i915 driver, mesa driver, X server, libdrm middleware, and so on -- the complete stack -- to properly test? I suppose so, but this just makes the task that much more daunting. With just a simple request to "try building from drm-tip" I am left with too many questions. On two different attempts to build this system following this guide, I got so mired in the mud that I gave up. The .config differences were significant enough in drm-tip to make me tune out and take all the defaults. Does it matter? I may never know. In the meantime, I did try a newer release with Ubuntu 17.04 last month. The problems there had worsened such that I was unable to start at all with two monitors connected, even at 30Hz. I went back to trying to build drm-tip so I could get useful info for this bug report, but again the build took hours and lots of babysitting.
(In reply to Phil Hord from comment #10) > The problem worsened in Ubuntu 17.04 so much that I reverted to 16.10. My > previous attempts to build drm-tip as requested were so far unsuccessful. > ... Hello Phil, Sorry for the delay. If you are having troubles building the drm-tip, try with the latest kernel 4.12. You can find aid on youtube tutorials like this: https://www.youtube.com/watch?v=oN5PdIKc0tM (this is just an example, you can search for a better one), and in the community forums: https://01.org/linuxgraphics/community. > ... The build ran for a long time before I had to give up and get > back to my real work. > You can disable CONFIG_DEBUG_INFO in order to build the kernel faster, but if later is needed, you may need to build again with this feature enable. > ... But I find the instructions at > https://01.org/linuxgraphics/documentation/build-guide-0 to be frustratingly > vague in some places and broken in others. I kept a log of errors for a > while hoping I could submit a patch to someone, but ultimately I could not > confidently identify the someone to report it to. > The documentation on the build guide is intended for giving a simple aid for users to do the basic functions, not being too specific, since in every new linux version the steps for this process can change, so is better to research on specifics steps per version. Though as you mention, the documents aren't up-to-date now, so we'll work on it to update the information to a more valid standar procedure. >... > Do I really need to build a replacement kernel, i915 driver, mesa driver, X > server, libdrm middleware, and so on -- the complete stack -- to properly > test? ... In this case, the problem is assumed to be in the kernel, so only building the kernel would be needed by the moment. Hope this help. Thank you.
Really? You think the bug is in the linux kernel, not the i915 driver? So I should build just the code in drm-tip and load that kernel and then I don't need to bother with libmesa, libdrm, xserver, etc.? I'll give that a try.
(In reply to Phil Hord from comment #12) > Really? You think the bug is in the linux kernel, not the i915 driver? So > I should build just the code in drm-tip and load that kernel and then I > don't need to bother with libmesa, libdrm, xserver, etc.? > > I'll give that a try. Me coming up to speed: the i915 driver is part of the kernel build. I built the bare kernel from drm-tip but the defaults wouldn't boot from my btrfs root. I added the btrfs support and it still wouldn't boot; not sure why this time. Now I'm trying with my stock .config but it's taking 10x longer, of course. Will report what I find.
I built the kernel from drm-tip: drm-intel-nightly @ 3cbfafbde5bb92b (4.13.0+) It is more stable for me. * I do not see any errors reported yet about FIFOs in syslog; * Attempting to set the mode for 2 monitors to 4k@60hz does not crash, but neither does it succeed. * I am able to configure both monitors for 4k@30hz, and this does seem to work sometimes. I often get some error reported from xrandr about being unable to apply this configuration. It's possible some Ubuntu configurator is responding to hotplug events and fighting with me. I can't quite tell; but I do get an error message on the GUI sometimes about being unable to apply this requested mode. After fighting with it via xrandr (for about 2 hours) I can eventually get 4k@30Hz on two monitors at the same time. So the display driver seems to be doing the right thing now. I don't know why 60Hz doesn't work, though. Should it work? Is there some video card limitation that prevents it? Thanks for your help, and thanks for updating the build guide.
(In reply to Phil Hord from comment #14) > I built the kernel from drm-tip: > drm-intel-nightly @ 3cbfafbde5bb92b (4.13.0+) > > It is more stable for me. > * I do not see any errors reported yet about FIFOs in syslog; > * Attempting to set the mode for 2 monitors to 4k@60hz does not crash, but > neither does it succeed. > * I am able to configure both monitors for 4k@30hz, and this does seem to > work sometimes. > > I often get some error reported from xrandr about being unable to apply this > configuration. It's possible some Ubuntu configurator is responding to > hotplug events and fighting with me. I can't quite tell; but I do get an > error message on the GUI sometimes about being unable to apply this > requested mode. > > After fighting with it via xrandr (for about 2 hours) I can eventually get > 4k@30Hz on two monitors at the same time. So the display driver seems to be > doing the right thing now. > > I don't know why 60Hz doesn't work, though. Should it work? Is there some > video card limitation that prevents it? > > Thanks for your help, and thanks for updating the build guide. There seems to be an actual thing with 60Hz as stated by https://bugs.freedesktop.org/show_bug.cgi?id=101390#c12. Good to hear is working now. And thanks for helping to improve the build guide. Regards.
Spoke too soon, maybe. I'm trying to find a reliable way to enable my 2 4k monitors in 30hz mode. While trying I set one to a much smaller resolution (1680x1050) and then tried connecting the 2nd one again. My laptop hung hard and video memory was visibly corrupted a bit. I don't see any kernel panic or core file. Is there something I can do to help gather more details?
Please add drm.debug=14 module parameter, and attach dmesg all the way from boot to trying to enable the displays and reproducing the problem.
I haven't been able to repro this display corruption issue again, but I did get this oops today while Ubuntu's hotplug-handler was fighting with the drm. Not sure if it's related to this bug: Sep 25 09:33:11 phord-x1 kernel: [ 601.970328] BUG: unable to handle kernel NULL pointer dereference at 0000000000000038 Sep 25 09:33:11 phord-x1 kernel: [ 601.970368] IP: process_single_tx_qlock+0x67/0x610 [drm_kms_helper] Sep 25 09:33:11 phord-x1 kernel: [ 601.970374] PGD 0 Sep 25 09:33:11 phord-x1 kernel: [ 601.970375] P4D 0 Sep 25 09:33:11 phord-x1 kernel: [ 601.970379] Sep 25 09:33:11 phord-x1 kernel: [ 601.970387] Oops: 0000 [#1] SMP Sep 25 09:33:11 phord-x1 kernel: [ 601.970393] Modules linked in: ccm ipt_MASQUERADE nf_nat_masquerade_ipv4 nf_conntrack_netlink nfnetlink xfrm_user xfrm_algo iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 xt_addrtype iptable_filter x t_conntrack nf_nat nf_conntrack libcrc32c br_netfilter bridge stp llc bnep btusb btrtl btbcm btintel bluetooth snd_usb_audio hid_generic snd_usbmidi_lib uvcvideo videobuf2_vmalloc videobuf2_memops videobuf2_v4l2 videobuf2_core videodev media snd _hda_codec_hdmi ecdh_generic snd_hda_codec_conexant snd_hda_codec_generic arc4 cdc_ether usbnet mii intel_rapl x86_pkg_temp_thermal intel_powerclamp coretemp iwlmvm binfmt_misc kvm_intel kvm mac80211 snd_soc_skl irqbypass snd_soc_skl_ipc crct10d if_pclmul snd_soc_sst_ipc crc32_pclmul snd_soc_sst_dsp ghash_clmulni_intel snd_hda_ext_core pcbc snd_soc_sst_match Sep 25 09:33:11 phord-x1 kernel: [ 601.970503] snd_soc_core aesni_intel snd_compress snd_pcm_dmaengine aes_x86_64 ac97_bus snd_hda_intel crypto_simd snd_hda_codec glue_helper cryptd snd_hda_core snd_hwdep iwlwifi snd_pcm thinkpad_acpi nvram sn d_seq_midi snd_seq_midi_event snd_rawmidi intel_cstate intel_rapl_perf snd_seq snd_seq_device cfg80211 snd_timer joydev input_leds sparse_keymap serio_raw wmi_bmof rtsx_pci_ms snd memstick hid_sensor_accel_3d hid_sensor_trigger industrialio_trig gered_buffer kfifo_buf hid_sensor_iio_common industrialio soundcore mei_me mei tpm_crb shpchp intel_pch_thermal mac_hid parport_pc ppdev lp parport ip_tables x_tables autofs4 btrfs xor hid_logitech_hidpp hid_logitech_dj usbhid raid6_pq hid_senso r_custom hid_sensor_hub intel_ishtp_hid hid i915 i2c_algo_bit drm_kms_helper syscopyarea rtsx_pci_sdmmc sysfillrect Sep 25 09:33:11 phord-x1 kernel: [ 601.970612] sysimgblt fb_sys_fops drm psmouse nvme nvme_core e1000e ptp pps_core intel_ish_ipc video intel_ishtp rtsx_pci wmi Sep 25 09:33:11 phord-x1 kernel: [ 601.970641] CPU: 3 PID: 281 Comm: kworker/u8:4 Not tainted 4.13.1+ #4 Sep 25 09:33:11 phord-x1 kernel: [ 601.970647] Hardware name: LENOVO 20FB005LUS/20FB005LUS, BIOS N1FET43W (1.17 ) 08/02/2016 Sep 25 09:33:11 phord-x1 kernel: [ 601.970742] Workqueue: i915-dp i915_digport_work_func [i915] Sep 25 09:33:11 phord-x1 kernel: [ 601.970748] task: ffff884211abdf00 task.stack: ffffa66bc2264000 Sep 25 09:33:11 phord-x1 kernel: [ 601.970772] RIP: 0010:process_single_tx_qlock+0x67/0x610 [drm_kms_helper] Sep 25 09:33:11 phord-x1 kernel: [ 601.970777] RSP: 0018:ffffa66bc2267b40 EFLAGS: 00010246 Sep 25 09:33:11 phord-x1 kernel: [ 601.970783] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000 Sep 25 09:33:11 phord-x1 kernel: [ 601.970789] RDX: 0000000000000001 RSI: ffff884246a21000 RDI: 0000000000000000 Sep 25 09:33:11 phord-x1 kernel: [ 601.970794] RBP: ffffa66bc2267bd8 R08: ffff88428159f500 R09: ffff884246a21000 Sep 25 09:33:11 phord-x1 kernel: [ 601.970799] R10: ffff884246a21000 R11: ffff8842161ba988 R12: ffff884214b12a40 Sep 25 09:33:11 phord-x1 kernel: [ 601.970804] R13: ffff884214b12720 R14: ffff884214b12720 R15: 0000000000000000 Sep 25 09:33:11 phord-x1 kernel: [ 601.970811] FS: 0000000000000000(0000) GS:ffff884281580000(0000) knlGS:0000000000000000 Sep 25 09:33:11 phord-x1 kernel: [ 601.970817] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 Sep 25 09:33:11 phord-x1 kernel: [ 601.970822] CR2: 0000000000000038 CR3: 000000038f009000 CR4: 00000000003406e0 Sep 25 09:33:11 phord-x1 kernel: [ 601.970827] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 Sep 25 09:33:11 phord-x1 kernel: [ 601.970832] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Sep 25 09:33:11 phord-x1 kernel: [ 601.970836] Call Trace: Sep 25 09:33:11 phord-x1 kernel: [ 601.970862] ? drm_dp_sideband_msg_build+0x230/0x3d0 [drm_kms_helper] Sep 25 09:33:11 phord-x1 kernel: [ 601.970878] ? kmem_cache_alloc_trace+0xd7/0x190 Sep 25 09:33:11 phord-x1 kernel: [ 601.970898] drm_dp_send_up_ack_reply.constprop.23+0x7b/0xe0 [drm_kms_helper] Sep 25 09:33:11 phord-x1 kernel: [ 601.970917] drm_dp_mst_handle_up_req+0x412/0x5b0 [drm_kms_helper] Sep 25 09:33:11 phord-x1 kernel: [ 601.970938] drm_dp_mst_hpd_irq+0x60/0x8d0 [drm_kms_helper] Sep 25 09:33:11 phord-x1 kernel: [ 601.970955] ? drm_dp_mst_hpd_irq+0x60/0x8d0 [drm_kms_helper] Sep 25 09:33:11 phord-x1 kernel: [ 601.971050] intel_dp_check_mst_status+0x127/0x1f0 [i915] Sep 25 09:33:11 phord-x1 kernel: [ 601.971135] ? intel_dp_check_mst_status+0x127/0x1f0 [i915] Sep 25 09:33:11 phord-x1 kernel: [ 601.971218] intel_dp_hpd_pulse+0x1fd/0x310 [i915] Sep 25 09:33:11 phord-x1 kernel: [ 601.971300] i915_digport_work_func+0x95/0x110 [i915] Sep 25 09:33:11 phord-x1 kernel: [ 601.971311] process_one_work+0x1e9/0x410 Sep 25 09:33:11 phord-x1 kernel: [ 601.971319] worker_thread+0x4b/0x410 Sep 25 09:33:11 phord-x1 kernel: [ 601.971327] kthread+0x109/0x140 Sep 25 09:33:11 phord-x1 kernel: [ 601.971335] ? process_one_work+0x410/0x410 Sep 25 09:33:11 phord-x1 kernel: [ 601.971346] ? kthread_create_on_node+0x70/0x70 Sep 25 09:33:11 phord-x1 kernel: [ 601.971355] ret_from_fork+0x25/0x30 Sep 25 09:33:11 phord-x1 kernel: [ 601.971361] Code: 00 00 00 00 85 c0 48 c7 45 98 00 00 00 00 75 55 48 8b be 38 01 00 00 c7 86 50 01 00 00 ff ff ff ff c7 86 54 01 00 00 01 00 00 00 <48> 83 7f 38 00 0f 84 01 05 00 00 48 83 7f 40 00 0f 85 3d 05 00 Sep 25 09:33:11 phord-x1 kernel: [ 601.971481] RIP: process_single_tx_qlock+0x67/0x610 [drm_kms_helper] RSP: ffffa66bc2267b40 Sep 25 09:33:11 phord-x1 kernel: [ 601.971485] CR2: 0000000000000038 Sep 25 09:33:11 phord-x1 kernel: [ 601.971492] ---[ end trace c1d9d8456eed809b ]--- Bear in mind this is on a kernel I compiled myself from tip by merging with Ubuntu's tip a month ago. There were some conflicts, so this could be only my bug. I hope it's helpful.
Created attachment 134468 [details] dmesg with drm.debug=14 dmesg with drm.debug=14 as requested.
To clarify: that dmesg trace shows my starting up and trying to get one of my monitors to sync. It never did. But I didn't see the video memory corruption this time.
Hit another similar video-corruption + hang-cpu event today. I'm now on the stock Ubuntu 17.10 kernel, 4.13.1. I have an xorg.log and a syslog I can send. The complete syslog is many MB because my machine hadn't rebooted in over a week and I have drm.debug=14, so about 1.4 million extra "drm:drm_mode_addfb2" messages. I see this in the syslog around the moment of hang: Nov 20 11:44:36 phord-x1 kernel: [189354.174270] [drm:pipe_config_err [i915]] *ERROR* mismatch in pixel_rate (expected 297000, found 593999) Nov 20 11:44:36 phord-x1 kernel: [189354.174286] [drm:pipe_config_err [i915]] *ERROR* mismatch in shared_dpll (expected ffff98bbd3b0c618, found ffff98bbd3b0c580) Nov 20 11:44:36 phord-x1 kernel: [189354.174301] [drm:pipe_config_err [i915]] *ERROR* mismatch in dpll_hw_state.ctrl1 (expected 0x00000003, found 0x00000001) Nov 20 11:44:36 phord-x1 kernel: [189354.174315] [drm:pipe_config_err [i915]] *ERROR* mismatch in base.adjusted_mode.crtc_clock (expected 297000, found 593999) Nov 20 11:44:36 phord-x1 kernel: [189354.174328] [drm:pipe_config_err [i915]] *ERROR* mismatch in port_clock (expected 270000, found 540000) Nov 20 11:44:36 phord-x1 kernel: [189354.174329] pipe state doesn't match! Nov 20 11:44:36 phord-x1 kernel: [189354.174362] ------------[ cut here ]------------ ... I'll upload a more complete snippet around the failure time. At the time it hung I had one 4k monitor on in 4k@30hz; I powered on the other monitor; the GUI did some reconfiguring and then the system was hung. The display had mostly correct imaging on it, but it had 100 or so "lines" of corrupted noise scattered evenly across the display. I haven't tried to rebuild the drm-tip again for this kernel since the system has been much more stable with this recent distribution version. But I can try to build it again if it will help. Let me know if you need the more complete week-long syslog.
Created attachment 135617 [details] 2-minute syslog around a corruption+hang on Nov-20
Created attachment 135618 [details] Xorg.0.log which presumably was "active" during the Nov-20 failure Syslog says this: /var/log/syslog:Nov 20 11:56:42 phord-x1 /usr/lib/gdm3/gdm-x-session[1327]: (--) Log file renamed from "/var/lib/gdm3/.local/share/xorg/Xorg.pid-1329.log" to "/var/lib/gdm3/.local/share/xorg/Xorg.0.log" The previous such line was from a week earlier, so I think that one is unrelated.
First of all. Sorry about spam. This is mass update for our bugs. Sorry if you feel this annoying but with this trying to understand if bug still valid or not. If bug investigation still in progress, please ignore this and I apologize! If you think this is not anymore valid, please comment to the bug that can be closed. If you haven't tested with our latest pre-upstream tree(drm-tip), can you do that also to see if issue is valid there still and if you cannot see issue there, please comment to the bug.
Closing, please re-open if still occurs.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.