Hi All,
This is an assortment of fixes which solve issues seen using an external RGB converter and makes sure that pixel clock is only on when really required.
Meng, could you test that on a LS1021a?
-- Stefan
Changes since v2: - Don't enable TCON bypass in case TCON is not available
Changes since v1: - add patch to no not transfer registers in mode_set_nofb - add patch which only init fbdev if required - remove disable unprepare pixel clock on module remove (already disabled in CRTC disable callback). - remove unused label
Stefan Agner (5): drm/fsl-dcu: enable TCON bypass mode by default drm/fsl-dcu: do not transfer registers on plane init drm/fsl-dcu: do not transfer registers in mode_set_nofb drm/fsl-dcu: enable pixel clock when enabling CRTC drm/fsl-dcu: only init fbdev if required
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c | 4 +-- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 26 ++++--------------- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c | 5 ---- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c | 39 ++++------------------------- 4 files changed, 12 insertions(+), 62 deletions(-)
Do not use encoder disable/enable callbacks to control bypass mode as this seems to mess with the signals not liked by displays. This also makes more sense since the encoder is already defined to be parallel RGB/LVDS at creation time.
Signed-off-by: Stefan Agner stefan@agner.ch --- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 2 ++ drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c | 39 ++++--------------------------- 2 files changed, 7 insertions(+), 34 deletions(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c index 0884c45..3897f56 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -273,6 +273,8 @@ static int fsl_dcu_drm_pm_resume(struct device *dev) goto disable_dcu_clk; }
+ if (fsl_dev->tcon) + fsl_tcon_bypass_enable(fsl_dev->tcon); fsl_dcu_drm_init_planes(fsl_dev->drm); drm_atomic_helper_resume(fsl_dev->drm, fsl_dev->state);
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c index 26edcc8..e1dd75b 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c @@ -20,38 +20,6 @@ #include "fsl_dcu_drm_drv.h" #include "fsl_tcon.h"
-static int -fsl_dcu_drm_encoder_atomic_check(struct drm_encoder *encoder, - struct drm_crtc_state *crtc_state, - struct drm_connector_state *conn_state) -{ - return 0; -} - -static void fsl_dcu_drm_encoder_disable(struct drm_encoder *encoder) -{ - struct drm_device *dev = encoder->dev; - struct fsl_dcu_drm_device *fsl_dev = dev->dev_private; - - if (fsl_dev->tcon) - fsl_tcon_bypass_disable(fsl_dev->tcon); -} - -static void fsl_dcu_drm_encoder_enable(struct drm_encoder *encoder) -{ - struct drm_device *dev = encoder->dev; - struct fsl_dcu_drm_device *fsl_dev = dev->dev_private; - - if (fsl_dev->tcon) - fsl_tcon_bypass_enable(fsl_dev->tcon); -} - -static const struct drm_encoder_helper_funcs encoder_helper_funcs = { - .atomic_check = fsl_dcu_drm_encoder_atomic_check, - .disable = fsl_dcu_drm_encoder_disable, - .enable = fsl_dcu_drm_encoder_enable, -}; - static void fsl_dcu_drm_encoder_destroy(struct drm_encoder *encoder) { drm_encoder_cleanup(encoder); @@ -68,13 +36,16 @@ int fsl_dcu_drm_encoder_create(struct fsl_dcu_drm_device *fsl_dev, int ret;
encoder->possible_crtcs = 1; + + /* Use bypass mode for parallel RGB/LVDS encoder */ + if (fsl_dev->tcon) + fsl_tcon_bypass_enable(fsl_dev->tcon); + ret = drm_encoder_init(fsl_dev->drm, encoder, &encoder_funcs, DRM_MODE_ENCODER_LVDS, NULL); if (ret < 0) return ret;
- drm_encoder_helper_add(encoder, &encoder_helper_funcs); - return 0; }
Do not use encoder disable/enable callbacks to control bypass mode as this seems to mess with the signals not liked by displays. This also makes more sense since the encoder is already defined to be parallel RGB/LVDS at creation time.
Signed-off-by: Stefan Agner stefan@agner.ch
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 2 ++ drivers/gpu/drm/fsl- dcu/fsl_dcu_drm_rgb.c | 39 ++++--------------------------- 2 files changed, 7 insertions(+), 34 deletions(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl- dcu/fsl_dcu_drm_drv.c index 0884c45..3897f56 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -273,6 +273,8 @@ static int fsl_dcu_drm_pm_resume(struct device *dev) goto disable_dcu_clk; }
- if (fsl_dev->tcon)
fsl_tcon_bypass_enable(fsl_dev->tcon);diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c b/drivers/gpu/drm/fsl- dcu/fsl_dcu_drm_rgb.c index 26edcc8..e1dd75b 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c @@ -20,38 +20,6 @@ #include "fsl_dcu_drm_drv.h" #include "fsl_tcon.h"
-static int -fsl_dcu_drm_encoder_atomic_check(struct drm_encoder *encoder,
struct drm_crtc_state *crtc_state,struct drm_connector_state *conn_state)-{
- return 0;
-}
-static void fsl_dcu_drm_encoder_disable(struct drm_encoder *encoder) -{
- struct drm_device *dev = encoder->dev;
- struct fsl_dcu_drm_device *fsl_dev = dev->dev_private;
- if (fsl_dev->tcon)
fsl_tcon_bypass_disable(fsl_dev->tcon);-}
-static void fsl_dcu_drm_encoder_enable(struct drm_encoder *encoder) -{
- struct drm_device *dev = encoder->dev;
- struct fsl_dcu_drm_device *fsl_dev = dev->dev_private;
- if (fsl_dev->tcon)
fsl_tcon_bypass_enable(fsl_dev->tcon);-}
-static const struct drm_encoder_helper_funcs encoder_helper_funcs = {
- .atomic_check = fsl_dcu_drm_encoder_atomic_check,
- .disable = fsl_dcu_drm_encoder_disable,
- .enable = fsl_dcu_drm_encoder_enable,
-};
static void fsl_dcu_drm_encoder_destroy(struct drm_encoder *encoder) { drm_encoder_cleanup(encoder); @@ -68,13 +36,16 @@ int fsl_dcu_drm_encoder_create(struct fsl_dcu_drm_device *fsl_dev, int ret;
encoder->possible_crtcs = 1;
- /* Use bypass mode for parallel RGB/LVDS encoder */
- if (fsl_dev->tcon)
fsl_tcon_bypass_enable(fsl_dev->tcon);- ret = drm_encoder_init(fsl_dev->drm, encoder, &encoder_funcs, DRM_MODE_ENCODER_LVDS, NULL); if (ret < 0) return ret;
- drm_encoder_helper_add(encoder, &encoder_helper_funcs);
- return 0;
}
-- 2.10.0
Tested-By: Meng Yi meng.yi@nxp.com
Tested those 5 patches on LS1021a-twr, and everything is fine.
Meng
On 2016-10-19 02:35, Meng Yi wrote:
Do not use encoder disable/enable callbacks to control bypass mode as this seems to mess with the signals not liked by displays. This also makes more sense since the encoder is already defined to be parallel RGB/LVDS at creation time.
Signed-off-by: Stefan Agner stefan@agner.ch
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 2 ++ drivers/gpu/drm/fsl- dcu/fsl_dcu_drm_rgb.c | 39 ++++--------------------------- 2 files changed, 7 insertions(+), 34 deletions(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl- dcu/fsl_dcu_drm_drv.c index 0884c45..3897f56 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -273,6 +273,8 @@ static int fsl_dcu_drm_pm_resume(struct device *dev) goto disable_dcu_clk; }
- if (fsl_dev->tcon)
fsl_tcon_bypass_enable(fsl_dev->tcon);diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c b/drivers/gpu/drm/fsl- dcu/fsl_dcu_drm_rgb.c index 26edcc8..e1dd75b 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c @@ -20,38 +20,6 @@ #include "fsl_dcu_drm_drv.h" #include "fsl_tcon.h"
-static int -fsl_dcu_drm_encoder_atomic_check(struct drm_encoder *encoder,
struct drm_crtc_state *crtc_state,struct drm_connector_state *conn_state)-{
- return 0;
-}
-static void fsl_dcu_drm_encoder_disable(struct drm_encoder *encoder) -{
- struct drm_device *dev = encoder->dev;
- struct fsl_dcu_drm_device *fsl_dev = dev->dev_private;
- if (fsl_dev->tcon)
fsl_tcon_bypass_disable(fsl_dev->tcon);-}
-static void fsl_dcu_drm_encoder_enable(struct drm_encoder *encoder) -{
- struct drm_device *dev = encoder->dev;
- struct fsl_dcu_drm_device *fsl_dev = dev->dev_private;
- if (fsl_dev->tcon)
fsl_tcon_bypass_enable(fsl_dev->tcon);-}
-static const struct drm_encoder_helper_funcs encoder_helper_funcs = {
- .atomic_check = fsl_dcu_drm_encoder_atomic_check,
- .disable = fsl_dcu_drm_encoder_disable,
- .enable = fsl_dcu_drm_encoder_enable,
-};
static void fsl_dcu_drm_encoder_destroy(struct drm_encoder *encoder) { drm_encoder_cleanup(encoder); @@ -68,13 +36,16 @@ int fsl_dcu_drm_encoder_create(struct fsl_dcu_drm_device *fsl_dev, int ret;
encoder->possible_crtcs = 1;
- /* Use bypass mode for parallel RGB/LVDS encoder */
- if (fsl_dev->tcon)
fsl_tcon_bypass_enable(fsl_dev->tcon);- ret = drm_encoder_init(fsl_dev->drm, encoder, &encoder_funcs, DRM_MODE_ENCODER_LVDS, NULL); if (ret < 0) return ret;
- drm_encoder_helper_add(encoder, &encoder_helper_funcs);
- return 0;
}
-- 2.10.0
Tested-By: Meng Yi meng.yi@nxp.com
Tested those 5 patches on LS1021a-twr, and everything is fine.
Thanks.
Applied 1~4.
-- Stefan
There is no need to explicitly initiate a register transfer and turn off the DCU after initializing the plane registers. In fact, this is harmful and leads to unnecessary flickers if the DCU has been left on by the bootloader.
Signed-off-by: Stefan Agner stefan@agner.ch --- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c | 5 ----- 1 file changed, 5 deletions(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c index a7e5486..9e6f7d8 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c @@ -211,11 +211,6 @@ void fsl_dcu_drm_init_planes(struct drm_device *dev) for (j = 1; j <= fsl_dev->soc->layer_regs; j++) regmap_write(fsl_dev->regmap, DCU_CTRLDESCLN(i, j), 0); } - regmap_update_bits(fsl_dev->regmap, DCU_DCU_MODE, - DCU_MODE_DCU_MODE_MASK, - DCU_MODE_DCU_MODE(DCU_MODE_OFF)); - regmap_write(fsl_dev->regmap, DCU_UPDATE_MODE, - DCU_UPDATE_MODE_READREG); }
struct drm_plane *fsl_dcu_drm_primary_create_plane(struct drm_device *dev)
Do not schedule a transfer of mode settings early. Modes should get applied on on CRTC enable where we also enable the pixel clock.
Signed-off-by: Stefan Agner stefan@agner.ch --- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c | 2 -- 1 file changed, 2 deletions(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c index 3371635..5ad1d68 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c @@ -116,8 +116,6 @@ static void fsl_dcu_drm_crtc_mode_set_nofb(struct drm_crtc *crtc) DCU_THRESHOLD_LS_BF_VS(BF_VS_VAL) | DCU_THRESHOLD_OUT_BUF_HIGH(BUF_MAX_VAL) | DCU_THRESHOLD_OUT_BUF_LOW(BUF_MIN_VAL)); - regmap_write(fsl_dev->regmap, DCU_UPDATE_MODE, - DCU_UPDATE_MODE_READREG); return; }
The pixel clock should not be on if the CRTC is not in use, hence move clock enable/disable calls into CRTC callbacks.
Signed-off-by: Stefan Agner stefan@agner.ch --- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c | 2 ++ drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 21 +-------------------- 2 files changed, 3 insertions(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c index 5ad1d68..b2d5e18 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_crtc.c @@ -51,6 +51,7 @@ static void fsl_dcu_drm_disable_crtc(struct drm_crtc *crtc) DCU_MODE_DCU_MODE(DCU_MODE_OFF)); regmap_write(fsl_dev->regmap, DCU_UPDATE_MODE, DCU_UPDATE_MODE_READREG); + clk_disable_unprepare(fsl_dev->pix_clk); }
static void fsl_dcu_drm_crtc_enable(struct drm_crtc *crtc) @@ -58,6 +59,7 @@ static void fsl_dcu_drm_crtc_enable(struct drm_crtc *crtc) struct drm_device *dev = crtc->dev; struct fsl_dcu_drm_device *fsl_dev = dev->dev_private;
+ clk_prepare_enable(fsl_dev->pix_clk); regmap_update_bits(fsl_dev->regmap, DCU_DCU_MODE, DCU_MODE_DCU_MODE_MASK, DCU_MODE_DCU_MODE(DCU_MODE_NORMAL)); diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c index 3897f56..e04efbe 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -267,12 +267,6 @@ static int fsl_dcu_drm_pm_resume(struct device *dev) return ret; }
- ret = clk_prepare_enable(fsl_dev->pix_clk); - if (ret < 0) { - dev_err(dev, "failed to enable pix clk\n"); - goto disable_dcu_clk; - } - if (fsl_dev->tcon) fsl_tcon_bypass_enable(fsl_dev->tcon); fsl_dcu_drm_init_planes(fsl_dev->drm); @@ -286,10 +280,6 @@ static int fsl_dcu_drm_pm_resume(struct device *dev) enable_irq(fsl_dev->irq);
return 0; - -disable_dcu_clk: - clk_disable_unprepare(fsl_dev->clk); - return ret; } #endif
@@ -403,18 +393,12 @@ static int fsl_dcu_drm_probe(struct platform_device *pdev) goto disable_clk; }
- ret = clk_prepare_enable(fsl_dev->pix_clk); - if (ret < 0) { - dev_err(dev, "failed to enable pix clk\n"); - goto unregister_pix_clk; - } - fsl_dev->tcon = fsl_tcon_init(dev);
drm = drm_dev_alloc(driver, dev); if (IS_ERR(drm)) { ret = PTR_ERR(drm); - goto disable_pix_clk; + goto unregister_pix_clk; }
fsl_dev->dev = dev; @@ -435,8 +419,6 @@ static int fsl_dcu_drm_probe(struct platform_device *pdev)
unref: drm_dev_unref(drm); -disable_pix_clk: - clk_disable_unprepare(fsl_dev->pix_clk); unregister_pix_clk: clk_unregister(fsl_dev->pix_clk); disable_clk: @@ -449,7 +431,6 @@ static int fsl_dcu_drm_remove(struct platform_device *pdev) struct fsl_dcu_drm_device *fsl_dev = platform_get_drvdata(pdev);
clk_disable_unprepare(fsl_dev->clk); - clk_disable_unprepare(fsl_dev->pix_clk); clk_unregister(fsl_dev->pix_clk); drm_put_dev(fsl_dev->drm);
There is no need to request a CMA backed framebuffer if fbdev emulation is not enabled.
Signed-off-by: Stefan Agner stefan@agner.ch --- drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c index e04efbe..3a5880c 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -87,7 +87,8 @@ static int fsl_dcu_load(struct drm_device *dev, unsigned long flags) goto done; dev->irq_enabled = true;
- fsl_dcu_fbdev_init(dev); + if (IS_ENABLED(CONFIG_DRM_FBDEV_EMULATION)) + fsl_dcu_fbdev_init(dev);
return 0; done:
On Mon, Oct 17, 2016 at 02:33:21PM -0700, Stefan Agner wrote:
There is no need to request a CMA backed framebuffer if fbdev emulation is not enabled.
Signed-off-by: Stefan Agner stefan@agner.ch
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c index e04efbe..3a5880c 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -87,7 +87,8 @@ static int fsl_dcu_load(struct drm_device *dev, unsigned long flags) goto done; dev->irq_enabled = true;
- fsl_dcu_fbdev_init(dev);
- if (IS_ENABLED(CONFIG_DRM_FBDEV_EMULATION))
fsl_dcu_fbdev_init(dev);
Totally not required, since this will all no-op out. Also, please nuke that fsl_dcu_fbdv_init wrapper seems like pointless indirection.
And if there really is an issue with the cma helpers allocating an fb when they should, then the correct fix is to fix that in the helpers, not in the drivers.
Nack. -Daniel
return 0; done: -- 2.10.0
dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
On 2016-10-18 00:44, Daniel Vetter wrote:
On Mon, Oct 17, 2016 at 02:33:21PM -0700, Stefan Agner wrote:
There is no need to request a CMA backed framebuffer if fbdev emulation is not enabled.
Signed-off-by: Stefan Agner stefan@agner.ch
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c index e04efbe..3a5880c 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -87,7 +87,8 @@ static int fsl_dcu_load(struct drm_device *dev, unsigned long flags) goto done; dev->irq_enabled = true;
- fsl_dcu_fbdev_init(dev);
- if (IS_ENABLED(CONFIG_DRM_FBDEV_EMULATION))
fsl_dcu_fbdev_init(dev);Totally not required, since this will all no-op out. Also, please nuke that fsl_dcu_fbdv_init wrapper seems like pointless indirection.
Regarding fsl_dcu_fbdev_init wrapper: Fully agreed. That thing was there since inception of that driver, and I always was on the fence of doing it. Will do it for the next merge window!
I somehow remembered there was something more to it than just "no need", but I somehow failed to document it in the patch description... So I went back and tested some things without the patch, here is when it blows:
Unable to handle kernel NULL pointer dereference at virtual address 000002f0 pgd = cc24c000 [000002f0] *pgd=8c0df831, *pte=00000000, *ppte=00000000 Internal error: Oops: 17 [#1] SMP ARM Modules linked in: CPU: 0 PID: 536 Comm: sh Not tainted 4.9.0-rc1-00001-g4b1532a-dirty #568 Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree) task: cc213000 task.stack: cc23e000 PC is at drm_fbdev_cma_fini+0x14/0x5c LR is at fsl_dcu_unload+0x28/0x4c pc : [<c04cfb20>] lr : [<c04fad74>] psr: a0000013 sp : cc23fd80 ip : cc23fd98 fp : cc23fd94 r10: cc1d1e4c r9 : cc23e000 r8 : 00000000 r7 : c0e34888 r6 : 0000000d r5 : 00000000 r4 : ce6ff100 r3 : c0e13b18 r2 : c0e13b18 r1 : 00000110 r0 : ce6ff100 Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none Control: 10c5387d Table: 8c24c04a DAC: 00000051 Process sh (pid: 536, stack limit = 0xcc23e210) Stack: (0xcc23fd80 to 0xcc240000) ... Backtrace: [<c04cfb0c>] (drm_fbdev_cma_fini) from [<c04fad74>] (fsl_dcu_unload+0x28/0x4c) r5:ce61e810[ 372.213609] r4:ce61f000
[<c04fad4c>] (fsl_dcu_unload) from [<c04d9550>] (drm_dev_unregister+0x38/0xbc) r5:ce61f000[ 372.228535] r4:ce61f000 ...
And if there really is an issue with the cma helpers allocating an fb when they should, then the correct fix is to fix that in the helpers, not in the drivers.
Hm, to safe memory, it would probably be best to do something like:
--- a/drivers/gpu/drm/drm_fb_cma_helper.c +++ b/drivers/gpu/drm/drm_fb_cma_helper.c @@ -492,6 +492,9 @@ struct drm_fbdev_cma *drm_fbdev_cma_init_with_funcs(struct drm_device *dev, struct drm_fb_helper *helper; int ret;
+ if (!drm_fbdev_emulation) + return NULL; + fbdev_cma = kzalloc(sizeof(*fbdev_cma), GFP_KERNEL); if (!fbdev_cma) { dev_err(dev->dev, "Failed to allocate drm fbdev.\n");
But we don't have drm_fbdev_emulation emulation there.
Maybe just add some NULL check in drm_fbdev_cma_fini?
-- Stefan
Nack. -Daniel
return 0; done: -- 2.10.0
dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
On Tue, Oct 18, 2016 at 10:42:46AM -0700, Stefan Agner wrote:
On 2016-10-18 00:44, Daniel Vetter wrote:
On Mon, Oct 17, 2016 at 02:33:21PM -0700, Stefan Agner wrote:
There is no need to request a CMA backed framebuffer if fbdev emulation is not enabled.
Signed-off-by: Stefan Agner stefan@agner.ch
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c index e04efbe..3a5880c 100644 --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c @@ -87,7 +87,8 @@ static int fsl_dcu_load(struct drm_device *dev, unsigned long flags) goto done; dev->irq_enabled = true;
- fsl_dcu_fbdev_init(dev);
- if (IS_ENABLED(CONFIG_DRM_FBDEV_EMULATION))
fsl_dcu_fbdev_init(dev);Totally not required, since this will all no-op out. Also, please nuke that fsl_dcu_fbdv_init wrapper seems like pointless indirection.
Regarding fsl_dcu_fbdev_init wrapper: Fully agreed. That thing was there since inception of that driver, and I always was on the fence of doing it. Will do it for the next merge window!
I somehow remembered there was something more to it than just "no need", but I somehow failed to document it in the patch description... So I went back and tested some things without the patch, here is when it blows:
Unable to handle kernel NULL pointer dereference at virtual address 000002f0 pgd = cc24c000 [000002f0] *pgd=8c0df831, *pte=00000000, *ppte=00000000 Internal error: Oops: 17 [#1] SMP ARM Modules linked in: CPU: 0 PID: 536 Comm: sh Not tainted 4.9.0-rc1-00001-g4b1532a-dirty #568 Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree) task: cc213000 task.stack: cc23e000 PC is at drm_fbdev_cma_fini+0x14/0x5c LR is at fsl_dcu_unload+0x28/0x4c pc : [<c04cfb20>] lr : [<c04fad74>] psr: a0000013 sp : cc23fd80 ip : cc23fd98 fp : cc23fd94 r10: cc1d1e4c r9 : cc23e000 r8 : 00000000 r7 : c0e34888 r6 : 0000000d r5 : 00000000 r4 : ce6ff100 r3 : c0e13b18 r2 : c0e13b18 r1 : 00000110 r0 : ce6ff100 Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none Control: 10c5387d Table: 8c24c04a DAC: 00000051 Process sh (pid: 536, stack limit = 0xcc23e210) Stack: (0xcc23fd80 to 0xcc240000) ... Backtrace: [<c04cfb0c>] (drm_fbdev_cma_fini) from [<c04fad74>] (fsl_dcu_unload+0x28/0x4c) r5:ce61e810[ 372.213609] r4:ce61f000
[<c04fad4c>] (fsl_dcu_unload) from [<c04d9550>] (drm_dev_unregister+0x38/0xbc) r5:ce61f000[ 372.228535] r4:ce61f000 ...
And if there really is an issue with the cma helpers allocating an fb when they should, then the correct fix is to fix that in the helpers, not in the drivers.
Hm, to safe memory, it would probably be best to do something like:
--- a/drivers/gpu/drm/drm_fb_cma_helper.c +++ b/drivers/gpu/drm/drm_fb_cma_helper.c @@ -492,6 +492,9 @@ struct drm_fbdev_cma *drm_fbdev_cma_init_with_funcs(struct drm_device *dev, struct drm_fb_helper *helper; int ret;
if (!drm_fbdev_emulation)return NULL;fbdev_cma = kzalloc(sizeof(*fbdev_cma), GFP_KERNEL); if (!fbdev_cma) { dev_err(dev->dev, "Failed to allocate drm fbdev.\n");But we don't have drm_fbdev_emulation emulation there.
Not just that, but you'd leak memory since cma_init always does the kmalloc of the fbdev_cma struct.
Maybe just add some NULL check in drm_fbdev_cma_fini?
Probably. I can't read arm-oopses, so no idea what exactly blew up. On a hunch it's probably drm_fbdev_cma_defio_fini that got things wrong and needs to check for !fbi || !fbi->defio. -Daniel
dri-devel@lists.freedesktop.org
-
Daniel Vetter -
Meng Yi -
Stefan Agner