| Message ID | 1510320714-19904-1-git-send-email-mika.kahola@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Quoting Mika Kahola (2017-11-10 13:31:54) > At least in Coffee Lake it happens that we start initiliazing audio when > no display is connected. This was discovered by CI when running IGT test > case > > drv_module_reload --r basic-no-display > > The issue here is that the 'intel_device_info_runtime_init()' sets > num_pipes to 0 but before this happens the audio part has already started > to initialize itself. Later on the num_pipes is updated to 0 in > intel_device_info_runtime_init() runtime_init happens twice? (I am confused by this pair of sentences.) and we hit that when audio part is digging > out ELD. This causes a warning in dmesg. To fix this issue, let's register > the audio driver only in a case when display is enabled. > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103206 > Signed-off-by: Mika Kahola <mika.kahola@intel.com> > --- > drivers/gpu/drm/i915/i915_drv.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c > index e7e9e06..f3cee1b 100644 > --- a/drivers/gpu/drm/i915/i915_drv.c > +++ b/drivers/gpu/drm/i915/i915_drv.c > @@ -1243,7 +1243,8 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) > if (IS_GEN5(dev_priv)) > intel_gpu_ips_init(dev_priv); > > - intel_audio_init(dev_priv); > + if (!i915_modparams.disable_display) INTEL_INFO()->num_ports == 0 is the derived value that includes cases where the display is fused off as well. -Chris
On Fri, 2017-11-10 at 13:37 +0000, Chris Wilson wrote: > Quoting Mika Kahola (2017-11-10 13:31:54) > > > > At least in Coffee Lake it happens that we start initiliazing audio > > when > > no display is connected. This was discovered by CI when running IGT > > test > > case > > > > drv_module_reload --r basic-no-display > > > > The issue here is that the 'intel_device_info_runtime_init()' sets > > num_pipes to 0 but before this happens the audio part has already > > started > > to initialize itself. Later on the num_pipes is updated to 0 in > > intel_device_info_runtime_init() > runtime_init happens twice? (I am confused by this pair of > sentences.) > > and we hit that when audio part is digging This is tricky. i915_audio_component_get_eld() gets called from a sound driver. Even though, I disable intel_audio_init() call, these num_pipes warnings keep popping up in dmesg like this. [ 475.875800] Setting dangerous option reset - tainting kernel [ 476.966234] WARN_ON(pipe >= intel_info((dev_priv))->num_pipes) [ 476.966247] ------------[ cut here ]------------ [ 476.966264] WARNING: CPU: 3 PID: 4697 at drivers/gpu/drm/i915/intel_audio.c:782 get_saved_enc+0x6f/0x90 [i915] [ 476.966266] Modules linked in: snd_hda_codec_hdmi(+) snd_hda_codec_realtek snd_hda_codec_generic snd_hda_intel snd_hda_codec snd_hwdep snd_hda_core snd_pcm i915 vgem ax88179_178a usbnet x86_pkg_temp_thermal intel_powerclamp mii coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel e1000e ptp prime_numbers pps_core i2c_hid [last unloaded: i915] [ 476.966302] CPU: 3 PID: 4697 Comm: modprobe Tainted: G U 4.14.0-rc8-CI-Trybot_1397+ #1 [ 476.966303] Hardware name: Intel Corporation CoffeeLake Client Platform/CoffeeLake S UDIMM RVP, BIOS CNLSFWR1.R00.X104.B11.1710091318 10/09/2017 [ 476.966305] task: ffff8803f0622880 task.stack: ffffc90000718000 [ 476.966320] RIP: 0010:get_saved_enc+0x6f/0x90 [i915] [ 476.966322] RSP: 0018:ffffc9000071ba58 EFLAGS: 00010292 [ 476.966325] RAX: 0000000000000032 RBX: ffff8803f7b00000 RCX: 0000000000000006 [ 476.966326] RDX: 000000000000203a RSI: ffffffff81d11254 RDI: ffffffff81cc3d2e [ 476.966328] RBP: ffffc9000071ba58 R08: ffff8803f06231e0 R09: 0000000000000000 [ 476.966329] R10: ffffc9000071ba58 R11: 0000000000000000 R12: ffff8803f7b07738 [ 476.966331] R13: ffff880452530278 R14: 0000000000000001 R15: ffff880452530270 [ 476.966332] FS: 00007fd67cba0700(0000) GS:ffff88045b2c0000(0000) knlGS:0000000000000000 [ 476.966334] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 476.966335] CR2: 000055cf58ac3050 CR3: 000000044e615003 CR4: 00000000003606e0 [ 476.966337] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 476.966338] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 476.966340] Call Trace: [ 476.966355] i915_audio_component_get_eld+0x4c/0x160 [i915] [ 476.966361] snd_hdac_acomp_get_eld+0x64/0x80 [snd_hda_core] [ 476.966365] hdmi_present_sense+0xaa/0x390 [snd_hda_codec_hdmi] [ 476.966369] generic_hdmi_build_controls+0x15b/0x200 [snd_hda_codec_hdmi] [ 476.966375] snd_hda_codec_build_controls+0x188/0x1d0 [snd_hda_codec] [ 476.966379] hda_codec_driver_probe+0x80/0x110 [snd_hda_codec] [ 476.966383] driver_probe_device+0x29c/0x450 [ 476.966386] __driver_attach+0xe3/0xf0 [ 476.966389] ? driver_probe_device+0x450/0x450 [ 476.966391] bus_for_each_dev+0x62/0xa0 [ 476.966395] driver_attach+0x1e/0x20 [ 476.966397] bus_add_driver+0x173/0x270 [ 476.966400] driver_register+0x60/0xe0 [ 476.966401] ? 0xffffffffa012a000 [ 476.966405] __hda_codec_driver_register+0x5a/0x60 [snd_hda_codec] [ 476.966408] hdmi_driver_init+0x1e/0x1000 [snd_hda_codec_hdmi] [ 476.966410] do_one_initcall+0x43/0x170 [ 476.966413] ? rcu_read_lock_sched_held+0x7a/0x90 [ 476.966415] ? kmem_cache_alloc_trace+0x270/0x2d0 [ 476.966419] do_init_module+0x5f/0x206 [ 476.966422] load_module+0x2581/0x2dd0 [ 476.966425] ? show_coresize+0x30/0x30 [ 476.966428] ? kernel_read+0x31/0x50 [ 476.966433] SyS_finit_module+0xc1/0x100 [ 476.966435] ? SyS_finit_module+0xc1/0x100 [ 476.966441] entry_SYSCALL_64_fastpath+0x1c/0xb1 [ 476.966442] RIP: 0033:0x7fd67c6d69f9 [ 476.966444] RSP: 002b:00007fff8bd21c98 EFLAGS: 00000246 ORIG_RAX: 0000000000000139 [ 476.966447] RAX: ffffffffffffffda RBX: ffffffff81492003 RCX: 00007fd67c6d69f9 [ 476.966448] RDX: 0000000000000000 RSI: 000055cf57227f8b RDI: 0000000000000000 [ 476.966450] RBP: ffffc9000071bf88 R08: 0000000000000000 R09: 0000000000000000 [ 476.966451] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000040000 [ 476.966453] R13: 000055cf58abd720 R14: 0000000000000000 R15: 000055cf58ac29f0 [ 476.966455] ? __this_cpu_preempt_check+0x13/0x20 [ 476.966459] Code: 70 0b 74 05 3b 70 74 74 27 48 83 c2 01 39 d1 7f e0 31 c0 c3 55 48 c7 c6 08 19 40 a0 48 c7 c7 fa 81 3e a0 48 89 e5 e8 32 9a db e0 <0f> ff 5d 31 c0 f3 c3 83 78 70 0b 74 f8 31 c0 85 d2 74 ad eb f0 [ 476.966542] ---[ end trace 0b809f2b0052649f ]--- > > > > out ELD. This causes a warning in dmesg. To fix this issue, let's > > register > > the audio driver only in a case when display is enabled. > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103206 > > Signed-off-by: Mika Kahola <mika.kahola@intel.com> > > --- > > drivers/gpu/drm/i915/i915_drv.c | 3 ++- > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > diff --git a/drivers/gpu/drm/i915/i915_drv.c > > b/drivers/gpu/drm/i915/i915_drv.c > > index e7e9e06..f3cee1b 100644 > > --- a/drivers/gpu/drm/i915/i915_drv.c > > +++ b/drivers/gpu/drm/i915/i915_drv.c > > @@ -1243,7 +1243,8 @@ static void i915_driver_register(struct > > drm_i915_private *dev_priv) > > if (IS_GEN5(dev_priv)) > > intel_gpu_ips_init(dev_priv); > > > > - intel_audio_init(dev_priv); > > + if (!i915_modparams.disable_display) > INTEL_INFO()->num_ports == 0 is the derived value that includes cases > where the display is fused off as well. > -Chris
Quoting Mika Kahola (2017-11-10 14:06:38) > On Fri, 2017-11-10 at 13:37 +0000, Chris Wilson wrote: > > Quoting Mika Kahola (2017-11-10 13:31:54) > > > > > > At least in Coffee Lake it happens that we start initiliazing audio > > > when > > > no display is connected. This was discovered by CI when running IGT > > > test > > > case > > > > > > drv_module_reload --r basic-no-display > > > > > > The issue here is that the 'intel_device_info_runtime_init()' sets > > > num_pipes to 0 but before this happens the audio part has already > > > started > > > to initialize itself. Later on the num_pipes is updated to 0 in > > > intel_device_info_runtime_init() > > runtime_init happens twice? (I am confused by this pair of > > sentences.) > > > > and we hit that when audio part is digging > This is tricky. i915_audio_component_get_eld() gets called from a sound > driver. Even though, I disable intel_audio_init() call, these num_pipes > warnings keep popping up in dmesg like this. Ask for a kasan run. I don't see where i915 is unbound during cleanup. component_del() only unbinds the master, and not itself. As I read the code it seems like the old i915 device is not uncoupled. First time I dipped my toe into the component code so take that with a pinch of salt. -Chris
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index e7e9e06..f3cee1b 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1243,7 +1243,8 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) if (IS_GEN5(dev_priv)) intel_gpu_ips_init(dev_priv); - intel_audio_init(dev_priv); + if (!i915_modparams.disable_display) + intel_audio_init(dev_priv); /* * Some ports require correctly set-up hpd registers for detection to
At least in Coffee Lake it happens that we start initiliazing audio when no display is connected. This was discovered by CI when running IGT test case drv_module_reload --r basic-no-display The issue here is that the 'intel_device_info_runtime_init()' sets num_pipes to 0 but before this happens the audio part has already started to initialize itself. Later on the num_pipes is updated to 0 in intel_device_info_runtime_init() and we hit that when audio part is digging out ELD. This causes a warning in dmesg. To fix this issue, let's register the audio driver only in a case when display is enabled. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103206 Signed-off-by: Mika Kahola <mika.kahola@intel.com> --- drivers/gpu/drm/i915/i915_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)