[Intel-gfx] i915 warning: WARN_ON_ONCE(!intel_dp->lane_count)

[Ville Syrjälä]

On Tue, Dec 13, 2016 at 05:29:54PM -0200, Paulo Zanoni wrote:
> Em Ter, 2016-12-13 às 21:17 +0200, Ville Syrjälä escreveu:
> > On Tue, Dec 13, 2016 at 11:00:50AM -0800, Linus Torvalds wrote:
> > > 
> > > This isn't new, but I thought I'd report it since it doesn't seem
> > > to
> > > get fixed on its own..
> 
> For me this is new. Ever since September, my SKL SDP booted 100% fine
> without any sort of error message. In the last few weeks I also noticed
> this message started appearing. I guess I just assumed we'd end up
> fixing it since I saw DP patches floating all the time.
> 
> We should really try to bisect it, especially since it's easy to
> reproduce this with our SDPs...

No need to bisect since it's been known all along. The actual
bug was uncovered by dd7880908223 ("drm/i915: Warn about aux
msg buffer vs. size mismatch") which resulted in tons of
WARNs, and we put in d4cb3fd9b548 ("drm/i915/dp: add lane_count
check in intel_dp_check_link_status") to reduce it to just
this single WARN instead.

> Also, shouldn't our CI have caught this? Why didn't it?

Unfortunately CI doesn't catch anything that happens at boot.
And we turn off the outputs at driver unload, so the reload
test won't catch it either. We should really prevent the driver
from loading until we start the tests.

> 
> > > 
> > >   [drm] Initialized i915 1.6.0 20161121 for 0000:00:02.0 on minor 0
> > >   ------------[ cut here ]------------
> > >   WARNING: CPU: 1 PID: 130 at drivers/gpu/drm/i915/intel_dp.c:4018
> > > intel_dp_check_link_status+0x1db$
> > >   WARN_ON_ONCE(!intel_dp->lane_count)
> > >   Modules linked in:
> > >    rtsx_pci_sdmmc mmc_core crct10dif_pclmul crc32_pclmul
> > > crc32c_intel
> > > i915 ghash_clmulni_intel i2c_$
> > >   CPU: 1 PID: 130 Comm: kworker/u8:7 Not tainted 4.9.0-04822-
> > > g9439b3710df6 #6
> > >   Hardware name: Dell Inc. XPS 13 9350/09JHRY, BIOS 1.4.4
> > > 06/14/2016
> > >   Workqueue: events_unbound async_run_entry_fn
> > >   Call Trace:
> > >    intel_dp_check_link_status+0x1db/0x200 [i915]
> > >    intel_dp_detect+0x697/0xa40 [i915]
> > >    drm_helper_probe_single_connector_modes+0x2a3/0x500
> > > [drm_kms_helper]
> > >    drm_setup_crtcs+0x7b/0x9c0 [drm_kms_helper]
> > >    drm_fb_helper_initial_config+0x79/0x3e0 [drm_kms_helper]
> > >    intel_fbdev_initial_config+0x18/0x30 [i915]
> > >    async_run_entry_fn+0x37/0x150
> > >    process_one_work+0x1f1/0x480
> > >    worker_thread+0x48/0x4d0
> > >    kthread+0x101/0x140
> > >    ret_from_fork+0x22/0x30
> > >   ---[ end trace 10162024459bbe32 ]---
> > > 
> > > That's obviously my XPS13, and it has nothing attached to it, so
> > > just
> > > the internal laptop display.
> > > 
> > > This is all intel, with the Skylake Iris Graphics 540 (i7-6560U).
> > > 
> > > Everything seems to work fine, it's just the ugly warning.
> > 
> > Yeah, it shouldn't really matter as long as the DP link doesn't fall
> > over during boot. If that were to happen we wouldn't be able to
> > retrain the link since we didn't copy the link parameters (eg. the
> > lane count) to the right place.
> > 
> > I posted a patch a while back that should fix it, but Daniel didn't
> > like it, and it's looking like no one is volunteering to work on it
> > further.
> > 

-- 
Ville Syrjälä
Intel OTC