[Intel-gfx] [PATCH 53/59] drm/arc: Move to drm/tiny

[Eugeniy Paltsev]

I've tested your change and one issue gone.

However I still see kernel crash (due to invalid read in kernel mode by 0x0 address) on weston stop:
----------------------------------->8-------------------------------------------
Oops
Path: (null)
CPU: 0 PID: 12 Comm: kworker/0:1 Not tainted 5.7.0-rc6-01594-g4ceda91a4176-dirty #6
Workqueue: events drm_mode_rmfb_work_fn
Invalid Read @ 0x00000000 by insn @ drm_gem_fb_destroy+0x32/0x130
ECR: 0x00050100 EFA: 0x00000000 ERET: 0x813b9a76
STAT32: 0x80080602 [IE K     ]  BTA: 0x813b9a72
BLK: drm_gem_fb_destroy+0xc0/0x130
 SP: 0x9f055ea4  FP: 0x00000000
LPS: 0x813560ec LPE: 0x813560f0 LPC: 0x00000000
r00: 0x00000000 r01: 0x9f6a6100 r02: 0x00000001
r03: 0x9fd5dde8 r04: 0x810f5de8 r05: 0x00000000
r06: 0x00000000 r07: 0x00000000 r08: 0x000000e1
r09: 0x00000000 r10: 0x00000000 r11: 0x000000e1
r12: 0x813b9b04

Stack Trace:
  drm_gem_fb_destroy+0x32/0x130
  drm_framebuffer_remove+0x1d2/0x358
  drm_mode_rmfb_work_fn+0x28/0x38
  process_one_work+0x19a/0x358
  worker_thread+0x2c4/0x494
  kthread+0xec/0x100
  ret_from_fork+0x18/0x1c
----------------------------------->8-------------------------------------------


The stack traces may vary but always end in drm_gem_fb_destroy:
----------------------------------->8-------------------------------------------
Stack Trace:
  drm_gem_fb_destroy+0x32/0x130
  drm_mode_rmfb+0x10e/0x148
  drm_ioctl_kernel+0x70/0xa0
  drm_ioctl+0x284/0x410
  ksys_ioctl+0xea/0xa3c
  EV_Trap+0xcc/0xd0
----------------------------------->8-------------------------------------------
Stack Trace:
  drm_gem_fb_destroy+0x32/0x130
  drm_fb_release+0x66/0xb0
  drm_file_free.part.11+0x112/0x1bc
  drm_release+0x80/0x120
  __fput+0x98/0x1bc
  task_work_run+0x6e/0xa8
  do_exit+0x2b4/0x7fc
  do_group_exit+0x2a/0x8c
  get_signal+0x9a/0x5f0
  do_signal+0x86/0x23c
  resume_user_mode_begin+0x88/0xd0
----------------------------------->8-------------------------------------------


---
 Eugeniy Paltsev


________________________________________
From: Daniel Vetter <daniel at ffwll.ch>
Sent: Thursday, June 4, 2020 14:19
To: Eugeniy Paltsev
Cc: Intel Graphics Development; DRI Development; Daniel Vetter; Sam Ravnborg; Alexey Brodkin
Subject: Re: [PATCH 53/59] drm/arc: Move to drm/tiny

Hi Eugeniy,

Apologies, somehow I missed your mail. I looked at the code again, and I
think I fumbled something. Does the below diff help to prevent the issues?

Thanks, Daniel


diff --git a/drivers/gpu/drm/tiny/arcpgu.c b/drivers/gpu/drm/tiny/arcpgu.c
index 857812f25bec..33d812a5ad7f 100644
--- a/drivers/gpu/drm/tiny/arcpgu.c
+++ b/drivers/gpu/drm/tiny/arcpgu.c
@@ -228,6 +228,9 @@ static void arc_pgu_update(struct drm_simple_display_pipe *pipe,
        struct arcpgu_drm_private *arcpgu;
        struct drm_gem_cma_object *gem;

+       if (!pipe->plane.state->fb)
+               return;
+
        arcpgu = pipe_to_arcpgu_priv(pipe);
        gem = drm_fb_cma_get_gem_obj(pipe->plane.state->fb, 0);
        arc_pgu_write(arcpgu, ARCPGU_REG_BUF0_ADDR, gem->paddr);
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - https://urldefense.com/v3/__http://blog.ffwll.ch__;!!A4F2R9G_pg!P0EvyJfMuDwqbeZmHZM5S9po30QWr4KgGrggRirNfgo7wrRXfnUO-8iq0AA4fQCW2WGPlDc$