[Intel-gfx] [PATCH 11/20] drm/i915: Grab a runtime pm reference before run live selftests
José Roberto de Souza
jose.souza at intel.com
Thu Aug 9 00:15:57 UTC 2018
Now that modeset stuff is not being initialized when display is
disabled nothing is holding a power well or runtime pm reference
when running live selftests. And IGT runs live selftests with
display on and off, causing the below warning:
[ 473.586533] Setting dangerous option live_selftests - tainting kernel
[ 484.219068] ------------[ cut here ]------------
[ 484.219070] RPM wakelock ref not held during HW access
[ 484.219126] WARNING: CPU: 3 PID: 4659 at drivers/gpu/drm/i915/intel_drv.h:1986 intel_runtime_pm_get_noresume+0x6c/0x70 [i915]
[ 484.219128] Modules linked in: i915(+) amdgpu chash gpu_sched ttm vgem snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic btusb btrtl btbcm x86_pkg_temp_thermal btintel coretemp snd_hda_codec asix crct10dif_pclmul bluetooth crc32_pclmul usbnet snd_hwdep mii snd_hda_core ghash_clmulni_intel e1000e snd_pcm ecdh_generic mei_me mei prime_numbers pinctrl_sunrisepoint pinctrl_intel [last unloaded: i915]
[ 484.219173] CPU: 3 PID: 4659 Comm: drv_selftest Tainted: G U 4.18.0-rc8-CI-Trybot_2696+ #1
[ 484.219174] Hardware name: Intel Corporation Kabylake Client platform/Kabylake R DDR4 RVP, BIOS KBLSE2R1.R00.X078.P02.1703030515 03/03/2017
[ 484.219205] RIP: 0010:intel_runtime_pm_get_noresume+0x6c/0x70 [i915]
[ 484.219206] Code: 94 77 20 00 01 e8 04 bf 53 e0 0f 0b eb c5 80 3d 83 77 20 00 00 75 c6 48 c7 c7 50 b4 ca a0 c6 05 73 77 20 00 01 e8 e4 be 53 e0 <0f> 0b eb af 41 54 55 53 80 bf e4 ab 00 00 00 48 89 fb 48 8b af e0
[ 484.219277] RSP: 0018:ffffc900003bfa10 EFLAGS: 00010286
[ 484.219279] RAX: 0000000000000000 RBX: ffff88023c160000 RCX: 0000000000000001
[ 484.219281] RDX: 0000000080000001 RSI: ffffffff820c708c RDI: 00000000ffffffff
[ 484.219282] RBP: ffff8802b4ceb3f8 R08: 00000000afcc94fa R09: 0000000000000000
[ 484.219284] R10: ffff8802b5820358 R11: 0000000000000000 R12: ffff88023c160068
[ 484.219285] R13: ffffffffa0c80e30 R14: ffffffffa0cce2a0 R15: ffff88023c160000
[ 484.219287] FS: 00007f4d95a83980(0000) GS:ffff8802becc0000(0000) knlGS:0000000000000000
[ 484.219288] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 484.219290] CR2: 00007f4d953257a0 CR3: 00000001fa1d8001 CR4: 00000000003606e0
[ 484.219291] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 484.219292] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 484.219294] Call Trace:
[ 484.219328] i915_gem_unpark+0xb9/0x170 [i915]
[ 484.219361] igt_mmap_offset_exhaustion+0x365/0x6e0 [i915]
[ 484.219400] ? __i915_subtests+0x39/0xf0 [i915]
[ 484.219403] ? ring_buffer_unlock_commit+0x20/0xd0
[ 484.219406] ? trace_vbprintk+0x171/0x220
[ 484.219411] ? __trace_bprintk+0x57/0x80
[ 484.219450] __i915_subtests+0x5e/0xf0 [i915]
[ 484.219487] __run_selftests+0x10b/0x190 [i915]
[ 484.219522] i915_live_selftests+0x2c/0x60 [i915]
[ 484.219552] i915_pci_probe+0x50/0xa0 [i915]
[ 484.219556] pci_device_probe+0xa1/0x130
[ 484.219560] driver_probe_device+0x2f5/0x470
So now grabbing and releasing a runtime pm reference around
i915_live_selftests() call.
Signed-off-by: José Roberto de Souza <jose.souza at intel.com>
---
drivers/gpu/drm/i915/i915_pci.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/i915/i915_pci.c b/drivers/gpu/drm/i915/i915_pci.c
index e931b48369dd..d43b950c2798 100644
--- a/drivers/gpu/drm/i915/i915_pci.c
+++ b/drivers/gpu/drm/i915/i915_pci.c
@@ -695,6 +695,7 @@ static void i915_pci_remove(struct pci_dev *pdev)
static int i915_pci_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
{
+ struct drm_i915_private *dev_priv;
struct intel_device_info *intel_info =
(struct intel_device_info *) ent->driver_data;
int err;
@@ -730,7 +731,10 @@ static int i915_pci_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
return -ENODEV;
}
+ dev_priv = to_i915(pci_get_drvdata(pdev));
+ intel_runtime_pm_get(dev_priv);
err = i915_live_selftests(pdev);
+ intel_runtime_pm_put(dev_priv);
if (err) {
i915_pci_remove(pdev);
return err > 0 ? -ENOTTY : err;
--
2.18.0
More information about the Intel-gfx
mailing list