[BUG] drm/vc4: *ERROR* [PLANE:67:plane-20] flip_done timed out

[Stefan Wahren]

Hi Eric,

> Eric Anholt <eric at anholt.net> hat am 10. Februar 2018 um 18:24 geschrieben:
> 
> 
> Stefan Wahren <stefan.wahren at i2se.com> writes:
> 
> > Hi, i'm having problems while booting Raspberry Pi 1 B (Linux 4.15,
> > bcm2835_defconfig) on my display, which is connected via HDMI. Mostly
> > the screen stays black and i'm getting the following warnings via
> > debug UART. Sometimes after disconnecting HDMI, i'm getting the
> > expected screen.
> >
> > [    3.253930] vc4-drm soc:gpu: bound 20902000.hdmi (ops vc4_hdmi_ops)
> > [    3.283413] vc4-drm soc:gpu: bound 20806000.vec (ops vc4_vec_ops)
> > [    3.310306] vc4-drm soc:gpu: bound 20400000.hvs (ops vc4_hvs_ops)
> > [    3.339131] vc4-drm soc:gpu: bound 20206000.pixelvalve (ops vc4_crtc_ops)
> > [    3.368727] vc4-drm soc:gpu: bound 20207000.pixelvalve (ops vc4_crtc_ops)
> > [    3.397935] vc4-drm soc:gpu: bound 20807000.pixelvalve (ops vc4_crtc_ops)
> > [    3.450306] vc4-drm soc:gpu: bound 20c00000.v3d (ops vc4_v3d_ops)
> > [    3.475686] checking generic (e330000 8ca000) vs hw (0 ffffffff)
> > [    3.475700] fb: switching to vc4drmfb from simple
> > [    3.499714] Console: switching to colour dummy device 80x30
> > [    3.520255] [drm] Initialized vc4 0.0.0 20140616 for soc:gpu on minor 0
> > [    3.527030] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
> > [    3.533779] [drm] Driver supports precise vblank timestamp query.
> > [    3.647269] ------------[ cut here ]------------
> > [    3.647320] WARNING: CPU: 0 PID: 10 at drivers/gpu/drm/drm_atomic_helper.c:1349 drm_atomic_helper_wait_for_vblanks+0x1fc/0x21c
> > [    3.647328] [CRTC:68:crtc-2] vblank wait timed out
> > [    3.647333] Modules linked in:
> > [    3.647354] CPU: 0 PID: 10 Comm: kworker/0:1 Not tainted 4.15.0+ #2
> > [    3.647358] Hardware name: BCM2835
> > [    3.647383] Workqueue: events deferred_probe_work_func
> > [    3.647429] [<c010fd80>] (unwind_backtrace) from [<c010d0dc>] (show_stack+0x20/0x24)
> > [    3.647459] [<c010d0dc>] (show_stack) from [<c07914f4>] (dump_stack+0x20/0x28)
> > [    3.647488] [<c07914f4>] (dump_stack) from [<c011f620>] (__warn+0xec/0x104)
> > [    3.647509] [<c011f620>] (__warn) from [<c011f680>] (warn_slowpath_fmt+0x48/0x50)
> > [    3.647532] [<c011f680>] (warn_slowpath_fmt) from [<c044b564>] (drm_atomic_helper_wait_for_vblanks+0x1fc/0x21c)
> > [    3.647556] [<c044b564>] (drm_atomic_helper_wait_for_vblanks) from [<c04815e4>] (vc4_atomic_complete_commit+0x80/0xb8)
> 
> Boris's clk series (https://patchwork.kernel.org/patch/10207161/) from
> Thursday was trying to fix some of this type of failure.
> 
> > Additionally after 10 minutes there are memleaks like this:
> >
> > unreferenced object 0xc8cb5c00 (size 64):
> >   comm "kworker/0:1", pid 10, jiffies 4294937654 (age 881.050s)
> >   hex dump (first 32 bytes):
> >     10 e0 ca c8 01 00 00 00 00 00 00 00 0c 5c cb c8  .............\..
> >     0c 5c cb c8 ff ff ff ff 18 5c cb c8 18 5c cb c8  .\.......\...\..
> >   backtrace:
> >     [<7c3bc269>] kmemleak_alloc+0x78/0xc0
> >     [<c58d2bd1>] kmem_cache_alloc_trace+0xfc/0x1e4
> >     [<6ecb9de5>] drm_atomic_helper_setup_commit+0xa4/0x3b8
> >     [<e4fdc746>] vc4_atomic_commit+0x30/0x11c
> >     [<ab19bb72>] drm_atomic_commit+0x54/0x68
> 
> I expect this is just waiting for this commit to trickle in:
> 
> commit 1c6ceeee6ebbcdf9839dbce00719162591180524
> Author: Leo (Sunpeng) Li <sunpeng.li at amd.com>
> Date:   Wed Jan 17 12:51:08 2018 +0100
> 
>     drm/atomic: Fix memleak on ERESTARTSYS during non-blocking commits

i reverted Boris' patches and applied this one [1], but the leaks still persists.

Stefan

[1] - https://patchwork.kernel.org/patch/10169179/