2014-07-04 10:36:56

by Tushar Behera

[permalink] [raw]
Subject: [PATCH] drm/exynos: fimd: Keep power enabled during fimd_bind

Under some conditions (when IOMMU is enabled), fimd_bind() accesses
hardware registers and power-domain should be enabled during that time.

fimd_bind --> fimd_mgr_initialize --> fimd_clear_channel

If the power-domain is disabled by that time, we get a boot-time crash.
It would be better to keep power-domain enabled explicitly.

Unhandled fault: external abort on non-linefetch (0x1008) at 0xf0180034
Internal error: : 1008 [#1] PREEMPT SMP ARM
...
PC is at fimd_bind+0x84/0x134
LR is at component_bind_all+0xb4/0x1d8

Signed-off-by: Tushar Behera <[email protected]>
---
drivers/gpu/drm/exynos/exynos_drm_fimd.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c
index 33161ad..34275fb 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c
@@ -888,11 +888,15 @@ static int fimd_bind(struct device *dev, struct device *master, void *data)
struct fimd_context *ctx = fimd_manager.ctx;
struct drm_device *drm_dev = data;

+ pm_runtime_get_sync(dev);
+
fimd_mgr_initialize(&fimd_manager, drm_dev);
exynos_drm_crtc_create(&fimd_manager);
if (ctx->display)
exynos_drm_create_enc_conn(drm_dev, ctx->display);

+ pm_runtime_put_sync(dev);
+
return 0;

}
--
1.7.9.5


Subject: Re: [PATCH] drm/exynos: fimd: Keep power enabled during fimd_bind

Thanks for patch.

2014-07-04 19:33 GMT+09:00 Tushar Behera <[email protected]>:
> Under some conditions (when IOMMU is enabled), fimd_bind() accesses
> hardware registers and power-domain should be enabled during that time.
>
> fimd_bind --> fimd_mgr_initialize --> fimd_clear_channel
>
> If the power-domain is disabled by that time, we get a boot-time crash.
> It would be better to keep power-domain enabled explicitly.
>
> Unhandled fault: external abort on non-linefetch (0x1008) at 0xf0180034
> Internal error: : 1008 [#1] PREEMPT SMP ARM
> ...
> PC is at fimd_bind+0x84/0x134
> LR is at component_bind_all+0xb4/0x1d8
>
> Signed-off-by: Tushar Behera <[email protected]>
> ---
> drivers/gpu/drm/exynos/exynos_drm_fimd.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c
> index 33161ad..34275fb 100644
> --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c
> +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c
> @@ -888,11 +888,15 @@ static int fimd_bind(struct device *dev, struct device *master, void *data)
> struct fimd_context *ctx = fimd_manager.ctx;
> struct drm_device *drm_dev = data;
>
> + pm_runtime_get_sync(dev);
> +

fimd driver has no runtime pm interfaces so pm_runtime_get_sync call
will enable lcd0 power domain. So shouldn't fimd clock also be
enabled?

> fimd_mgr_initialize(&fimd_manager, drm_dev);

And here would be good to call clk_disable_unprepare() and
pm_runtime_put_sync(). Below codes don't access any fimd register. Or
move these function calls into fimd_mgr_initialize().

Thanks,
Inki Dae

> exynos_drm_crtc_create(&fimd_manager);
> if (ctx->display)
> exynos_drm_create_enc_conn(drm_dev, ctx->display);
>
> + pm_runtime_put_sync(dev);
> +
> return 0;
>
> }
> --
> 1.7.9.5
>
> _______________________________________________
> dri-devel mailing list
> [email protected]
> http://lists.freedesktop.org/mailman/listinfo/dri-devel