[Intel-gfx] New " Unclaimed read from register 0x1f0034" in 4.12-rc1

[Hans de Goede]

Hi,

So while trying to reproduce my previously reported
"RPM wakelock ref not held during HW access" in 4.12-rc1 ?"

Bug, to test the possible fixes I encountered a new
"RPM wakelock ref not held during HW access" error combined
with a "Unclaimed read from register 0x1f0034" error:

Which I believe is a new issue in 4.12-rc1 I've tried adding
i915.mmio_debug=99 to the kernel cmdline and then I get the
same unclaimed read error 3 times for a single resume.

[   69.462566] Suspended for 3.099 seconds
[   69.463092] PM: resume from suspend-to-idle
[   69.489870] PM: noirq resume of devices complete after 26.368 msecs
[   69.490177] RPM wakelock ref not held during HW access
[   69.490355] ------------[ cut here ]------------
[   69.490441] WARNING: CPU: 1 PID: 2003 at drivers/gpu/drm/i915/intel_drv.h:1780 intel_init_vebox_ring_buffer+0x3753/0x4cb0 [i915]
...
[   69.491022] RIP: 0010:intel_init_vebox_ring_buffer+0x3753/0x4cb0 [i915]
...
[   69.491089] Call Trace:
[   69.491152]  intel_gvt_init_cmd_parser+0x25a/0x360 [i915]
[   69.491172]  irq_enable+0x3a/0x50
[   69.491185]  __enable_irq+0x34/0x60
[   69.491197]  resume_irqs+0x8f/0xd0
[   69.491208]  resume_device_irqs+0x10/0x20
[   69.491223]  dpm_resume_noirq+0x23e/0x310
[   69.491240]  suspend_devices_and_enter+0x2f3/0x7f0
[   69.491254]  ? finish_wait+0x80/0x80
[   69.491266]  pm_suspend+0x335/0x3a0
[   69.491277]  state_store+0x85/0xf0
[   69.491294]  kobj_attr_store+0xf/0x20
[   69.491308]  sysfs_kf_write+0x37/0x40
[   69.491317]  kernfs_fop_write+0x11c/0x1a0
[   69.491332]  __vfs_write+0x37/0x160
[   69.491346]  ? selinux_file_permission+0xfb/0x120
[   69.491359]  ? security_file_permission+0x3b/0xc0
[   69.491370]  vfs_write+0xb1/0x1a0
[   69.491381]  SyS_write+0x55/0xc0
[   69.491398]  entry_SYSCALL_64_fastpath+0x1a/0xa5
[   69.491409] RIP: 0033:0x7f2d526bbc20
...
[   69.491643] ---[ end trace 18cf4afd61c66532 ]---
[   70.561773] PM: early resume of devices complete after 1069.774 msecs
[   70.562713] pcieport 0000:00:1c.0: System wakeup disabled by ACPI
[   70.563597] rtc_cmos 00:04: System wakeup disabled by ACPI
[   70.576148] Unclaimed read from register 0x1f0034
[   70.576244] ------------[ cut here ]------------
[   70.576278] WARNING: CPU: 2 PID: 5 at drivers/gpu/drm/i915/intel_uncore.c:801 intel_init_vebox_ring_buffer+0x87e/0x4cb0 [i915]
...
[   70.576470] RIP: 0010:intel_init_vebox_ring_buffer+0x87e/0x4cb0 [i915]
...
[   70.576500] Call Trace:
[   70.576529]  intel_init_vebox_ring_buffer+0x47da/0x4cb0 [i915]
[   70.576550]  i915_gpu_lower+0x9d4/0xd70 [i915]
[   70.576573]  ? intel_hdmi_compute_config+0xe07/0x2710 [i915]
[   70.576592]  i915_gpu_lower+0xcb5/0xd70 [i915]
[   70.576617]  hsw_disable_ips+0xc42/0x1e60 [i915]
[   70.576624]  ? tracing_record_cmdline+0x32/0x120
[   70.576630]  ? __schedule+0x23e/0x860
[   70.576655]  hsw_disable_ips+0x1d29/0x1e60 [i915]
[   70.576691]  ? drm_atomic_check_only+0x37f/0x540 [drm]
[   70.576721]  drm_atomic_commit+0x4b/0x50 [drm]
[   70.576740]  drm_atomic_helper_commit_duplicated_state+0xc2/0xd0 [drm_kms_helper]
[   70.576765]  i915_redisable_vga+0xc5/0x100 [i915]
[   70.576790]  intel_display_resume+0xf7/0x120 [i915]
[   70.576812]  0xffffffffc053b341
[   70.576828]  0xffffffffc053b3fe
[   70.576834]  pci_pm_resume+0x65/0xa0
[   70.576841]  dpm_run_callback+0x57/0x140
[   70.576845]  ? pci_pm_thaw+0x90/0x90
[   70.576848]  device_resume+0xe1/0x200
[   70.576852]  async_resume+0x1d/0x50
[   70.576856]  async_run_entry_fn+0x39/0x170
[   70.576863]  process_one_work+0x193/0x3c0
[   70.576867]  worker_thread+0x4a/0x3a0
[   70.576871]  kthread+0x125/0x140
[   70.576874]  ? process_one_work+0x3c0/0x3c0
[   70.576877]  ? kthread_park+0x60/0x60
[   70.576884]  ret_from_fork+0x2c/0x40
[   70.576887] Code: 5b 41 5c 41 5d 5d c3 45 84 ed 48 c7 c0 a6 1f 64 c0 48 c7 c6 9c 1f 64 c0 48 0f 44 f0 44 89 e2 48 c7 c7 af 1f 64 c0 e8 73 ec c2 c5 <0f> ff 83 2d 95 31 0e 00 01 5b 41 5c 41 5d 5d c3 66 90 0f 1f 44
[   70.576961] ---[ end trace 18cf4afd61c66533 ]---

And then 2 more Unclaimed read from register 0x1f0034 messages with the same backtrace.

This seems to be consistently reproducible.

Regards,

Hans