rk3399: Power Domain infinite loop 'pd_gpu'

Commit Message

Chris Ruehl Nov. 24, 2020, 4:34 a.m. UTC

Hi All.

Just have a board boot from sd-card and hang after time out on
the ack for the power domain of the gpu subsystem

I looked into the pm_domain.c of the rk3399 where the error message
is reported and realized that the return code isn't handled at all
(vanilla v5.4.77).

this is just a preliminary reach out to you before I go and file
a bug and send official patch to the kernel guys.
The boot log and patch below.

[--shnip--]
[    1.840491] 005: dwhdmi-rockchip ff940000.hdmi: Detected HDMI TX controller 
v2.11a with HDCP (DWC HDMI 2.0 TX PHY)
[    1.841127] 005: rockchip-drm display-subsystem: bound ff940000.hdmi (ops 
dw_hdmi_rockchip_ops)
[    1.841133] 005: [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
[    1.841135] 005: [drm] No driver support for vblank timestamp query.
[    1.841160] 005: [drm] Cannot find any crtc or sizes
[    1.841455] 005: [drm] Initialized rockchip 1.0.0 20140818 for 
display-subsystem on minor 0
[    1.851607] 005: rockchip-pm-domain 
ff310000.power-management:power-controller: failed to get ack on domain 
'pd_gpu', val=0x800789
[--shnap--]


Patch to escalate the error upstream.
not call rockchip_pmu_restore_qos() if the ACK timed out.

Patch

--- drivers/soc/rockchip/pm_domains.c.orig      2020-11-24 11:48:42.216855194 +0800
+++ drivers/soc/rockchip/pm_domains.c   2020-11-24 12:05:27.242044445 +0800
@@ -291,7 +291,7 @@ 
  static int rockchip_pd_power(struct rockchip_pm_domain *pd, bool power_on)
  {
         struct rockchip_pmu *pmu = pd->pmu;
-       int ret;
+       int ret = 0;

         mutex_lock(&pmu->mutex);

@@ -307,23 +307,24 @@ 
                         rockchip_pmu_save_qos(pd);

                         /* if powering down, idle request to NIU first */
-                       rockchip_pmu_set_idle_request(pd, true);
+                       ret = rockchip_pmu_set_idle_request(pd, true);
                 }

                 rockchip_do_pmu_set_power_domain(pd, power_on);

                 if (power_on) {
                         /* if powering up, leave idle mode */
-                       rockchip_pmu_set_idle_request(pd, false);
+                       ret = rockchip_pmu_set_idle_request(pd, false);

-                       rockchip_pmu_restore_qos(pd);
+                       if (ret==0)
+                               rockchip_pmu_restore_qos(pd);
                 }

                 clk_bulk_disable(pd->num_clks, pd->clks);
         }

         mutex_unlock(&pmu->mutex);
-       return 0;
+       return ret;
  }

  static int rockchip_pd_power_on(struct generic_pm_domain *domain)