[Intel-gfx] REGRESSION 3.14 i915 warning & mouse cursor vanishing

[Imre Deak]

On Tue, 2014-06-10 at 12:35 -0700, Steven Noonan wrote:
> On Wed, Apr 16, 2014 at 3:03 PM, Steven Noonan <steven at uplinklabs.net> wrote:
> > On Wed, Apr 16, 2014 at 2:46 PM, Jani Nikula
> > <jani.nikula at linux.intel.com> wrote:
> >> On Tue, 15 Apr 2014, Imre Deak <imre.deak at intel.com> wrote:
> >>> On Tue, 2014-04-15 at 21:43 +0200, Daniel Vetter wrote:
> >>>> On Mon, Apr 14, 2014 at 11:56:03AM -0700, Steven Noonan wrote:
> >>>> > On Mon, Apr 14, 2014 at 11:35:05AM -0700, Keith Packard wrote:
> >>>> > > Steven Noonan <steven at uplinklabs.net> writes:
> >>>> > >
> >>>> > > > Was using my machine normally, then my mouse cursor vanished. After switching
> >>>> > > > to a VT and back to X11, my cursor came back. But I did notice a nasty trace in
> >>>> > > > dmesg (below).
> >>>> > >
> >>>> > > I don't think the trace below is related to the cursor disappearing.
> >>>> >
> >>>> > Any idea what the trace is all about then? Seems it has something to do
> >>>> > with runtime power management (maybe my aggressive kernel command-line
> >>>> > options are triggering it).
> >>>>
> >>>> Please test without them. Currently runtime pm should be disabled still on
> >>>> vlv (since it's incomplete in 3.14). If you've force-enabled that then you
> >>>> get to keep all pieces ;-)
> >>>>
> >>>> In general don't set any i915 options if you're not a developer or someone
> >>>> else who _really_ knows what's going on.
> >>>
> >>> Note that the lspci output and the
> >>>
> >>> [ 1795.275026] [drm:hsw_unclaimed_reg_clear] *ERROR* Unknown unclaimed
> >>> register before writing to 70084
> >>>
> >>> line suggests HSW and the specs for ThinkPad Yoga suggests the same. But
> >>> I don't know how the vlv_* functions can possible end up in those traces
> >>> then, perhaps just a coincidence, random data on stack?
> >>
> >> I'm wondering the same. Perhaps double check your kernel build and
> >> modules are all right and matching?
> >>
> >
> > It was a clean build (built in a clean chroot, no ccache or anything
> > fancy), so those stack traces are as "right" as they could be under
> > those conditions.
> >
> > The "good" news (or perhaps scary news) is that I've been running
> > 3.14.1 for the past 36 hours and haven't been able to reproduce either
> > problem since then (warnings or ninja mouse cursor). Nothing in the
> > changelog for v3.14..v3.14.1 really stands out as a clear fix though.
> > The only changes that appear to directly affect my configuration would
> > be the futex changes, iwlwifi change, efi change, and ipv6 change.
> 
> This issue is haunting me again. This time I'm running 3.14.6. My mouse
> cursor vanished, and I have a bunch of warnings in dmesg:
> 
> [ 5622.922652] ------------[ cut here ]------------
> [ 5622.922707] WARNING: CPU: 0 PID: 312 at drivers/gpu/drm/i915/intel_uncore.c:455 vlv_flisdsi_write+0x1d69/0x2cc0 [i915]()
> [ 5622.922710] Device suspended
> [ 5622.922714] Modules linked in: cpufreq_stats fuse ctr ccm hid_generic hid_sensor_magn_3d hid_sensor_als hid_sensor_gyro_3d hid_sensor_accel_3d hid_sensor_trigger industrialio_triggered_buffer kfifo_buf hid_sensor_iio_common industrialio hid_sensor_hub hid_multitouch usbhid wacom hid btusb joydev bnep bluetooth 6lowpan_iphc tun arc4 nls_cp437 vfat fat iwlmvm mac80211 x86_pkg_temp_thermal intel_powerclamp coretemp snd_hda_codec_hdmi iTCO_wdt kvm_intel iTCO_vendor_support kvm crct10dif_pclmul crc32_pclmul ghash_clmulni_intel aesni_intel aes_x86_64 lrw gf128mul glue_helper iwlwifi ablk_helper cryptd microcode psmouse cfg80211 serio_raw snd_hda_codec_conexant i2c_i801 lpc_ich snd_hda_codec_generic mfd_core tpm_tis snd_hda_intel thermal tpm wmi thinkpad_acpi snd_hda_codec snd_hwdep nvram battery snd_pcm
> [ 5622.922796]  ac evdev snd_timer snd soundcore intel_smartconnect acpi_cpufreq processor vmnet(O) vmblock(O) vmci(O) vmmon(O) msr ax88179_178a(O) usbnet mii efivarfs ext4 crc16 jbd2 mbcache sd_mod crc_t10dif crct10dif_common crc32c_intel ahci libahci libata xhci_hcd scsi_mod usbcore usb_common i915 video intel_gtt i2c_algo_bit drm_kms_helper drm i2c_core e1000e ptp pps_core ipmi_poweroff ipmi_msghandler button
> [ 5622.922853] CPU: 0 PID: 312 Comm: X Tainted: G           O 3.14.6-1-ec2 #1
> [ 5622.922856] Hardware name: LENOVO 20CDCTO1WW/20CDCTO1WW, BIOS GQET35WW (1.15 ) 04/21/2014
> [ 5622.922860]  0000000000000009 ffff880212319c08 ffffffff814fd4c3 ffff880212319c50
> [ 5622.922867]  ffff880212319c40 ffffffff810664ad 0000000000070088 ffff8802119b8000
> [ 5622.922873]  ffff8802119b8028 ffff8802119b8020 0000000003bb0611 ffff880212319ca0
> [ 5622.922880] Call Trace:
> [ 5622.922891]  [<ffffffff814fd4c3>] dump_stack+0x45/0x56
> [ 5622.922900]  [<ffffffff810664ad>] warn_slowpath_common+0x7d/0xa0
> [ 5622.922906]  [<ffffffff8106651c>] warn_slowpath_fmt+0x4c/0x50
> [ 5622.922922]  [<ffffffff811dcaf0>] ? poll_select_copy_remaining+0x130/0x130
> [ 5622.922942]  [<ffffffffa013a849>] vlv_flisdsi_write+0x1d69/0x2cc0 [i915]
> [ 5622.922963]  [<ffffffffa013af46>] vlv_flisdsi_write+0x2466/0x2cc0 [i915]
> [ 5622.922986]  [<ffffffffa00f364b>] i915_restore_display_reg+0x198b/0x8430 [i915]
> [ 5622.922995]  [<ffffffff81501062>] ? mutex_lock+0x12/0x30
> [ 5622.923017]  [<ffffffffa00f390f>] i915_restore_display_reg+0x1c4f/0x8430 [i915]
> [ 5622.923044]  [<ffffffffa006d060>] drm_mode_set_config_internal+0x180/0x260 [drm]
> [ 5622.923053]  [<ffffffff81075ed7>] ? __dequeue_signal+0x117/0x200

As before, the stack trace doesn't make sense to me above this line on
a HSW machine, all I can think of is a problem in your build (you ruled
it out), or a problem in the kernel's stack unwinding. In any case the
below drm_mode_cursor_ioctl line and the unclaimed register write
messages match and it seems that the problem happened already earlier
since the unclaimed flag was already set when we got here.

I think the best way forward would be to open a bug at bugs.freedesktop.org
and provide a dmesg log when the problem happens with drm.debug=14
kernel parameter set.

> [ 5622.923083]  [<ffffffffa0070d21>] drm_mode_cursor_ioctl+0x41/0x50 [drm]
> [ 5622.923102]  [<ffffffffa0061bb2>] drm_ioctl+0x4c2/0x700 [drm]
> [ 5622.923131]  [<ffffffffa0070ce0>] ? drm_mode_setcrtc+0x5e0/0x5e0 [drm]
> [ 5622.923140]  [<ffffffff811dbf28>] do_vfs_ioctl+0x2c8/0x4a0
> [ 5622.923149]  [<ffffffff811c90b7>] ? vfs_read+0x97/0x160
> [ 5622.923156]  [<ffffffff811dc181>] SyS_ioctl+0x81/0xa0
> [ 5622.923167]  [<ffffffff811002d6>] ? __audit_syscall_exit+0x236/0x2e0
> [ 5622.923175]  [<ffffffff8150bfad>] system_call_fastpath+0x1a/0x1f
> [ 5622.923179] ---[ end trace 5e18fb54489137d5 ]---
> [ 5622.923185] [drm:hsw_unclaimed_reg_clear] *ERROR* Unknown unclaimed register before writing to 70088
> [ 5622.923190] [drm:hsw_unclaimed_reg_check] *ERROR* Unclaimed write to 70088

--Imre