diff mbox series

[v1,1/4] media: camss: vfe: Don't read hardware version needlessly

Message ID 20210726080113.8378-1-robert.foss@linaro.org (mailing list archive)
State New, archived
Headers show
Series [v1,1/4] media: camss: vfe: Don't read hardware version needlessly | expand

Commit Message

Robert Foss July 26, 2021, 8:01 a.m. UTC
vfe_set_power() is invoced a few times, which causes a lot of
VFE HW version prints.

Secondly there is no real reason for calling hw_version_read() from
this location specifically, since it does nothing more than print
the VFE HW version.

A better location for the hw_version_read() call would be somewhere
which is only executed once per VFE unit.

Signed-off-by: Robert Foss <robert.foss@linaro.org>
---
 drivers/media/platform/qcom/camss/camss-vfe.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

Comments

Marek Szyprowski Aug. 10, 2021, 9:46 p.m. UTC | #1
Hi,

On 26.07.2021 10:01, Robert Foss wrote:
> vfe_set_power() is invoced a few times, which causes a lot of
> VFE HW version prints.
>
> Secondly there is no real reason for calling hw_version_read() from
> this location specifically, since it does nothing more than print
> the VFE HW version.
>
> A better location for the hw_version_read() call would be somewhere
> which is only executed once per VFE unit.
>
> Signed-off-by: Robert Foss <robert.foss@linaro.org>

This patch landed recently in linux-next as commit b10b5334528a ("media: 
camss: vfe: Don't read hardware version needlessly"). Sadly it causes a 
following regression on DragonBoard410c 
(arch/arm64/boot/dts/qcom/apq8016-sbc.dts):

Unable to handle kernel NULL pointer dereference at virtual address 
0000000000000000
Mem abort info:
   ESR = 0x96000006
   EC = 0x25: DABT (current EL), IL = 32 bits
   SET = 0, FnV = 0
   EA = 0, S1PTW = 0
   FSC = 0x06: level 2 translation fault
input: pm8941_resin as 
/devices/platform/soc/200f000.spmi/spmi-0/0-00/200f000.spmi:pmic@0:pon@800/200f000.spmi:pmic@0:pon@800:resin/input/input2
Data abort info:
   ISV = 0, ISS = 0x00000006
   CM = 0, WnR = 0
user pgtable: 4k pages, 48-bit VAs, pgdp=0000000087ca0000
[0000000000000000] pgd=0800000084c4c003, p4d=0800000084c4c003, 
pud=08000000859c2003, pmd=0000000000000000
Internal error: Oops: 96000006 [#1] PREEMPT SMP
Modules linked in: msm(+) qcom_pon rtc_pm8xxx qcom_spmi_vadc 
qcom_vadc_common qcom_camss(+) snd_soc_msm8916_analog 
qcom_spmi_temp_alarm venus_core(+) videobuf2_dma_sg v4l2_fwnode 
snd_soc_lpass_apq8016 v4l2_async videobuf2_memops snd_soc_lpass_cpu 
v4l2_mem2mem snd_soc_apq8016_sbc snd_soc_lpass_platform 
snd_soc_msm8916_digital videobuf2_v4l2 snd_soc_qcom_common 
videobuf2_common qnoc_msm8916 qcom_rng videodev i2c_qcom_cci mdt_loader 
mc icc_smd_rpm crct10dif_ce socinfo rmtfs_mem display_connector
adreno 1c00000.gpu: Adding to iommu group 2
CPU: 1 PID: 286 Comm: systemd-udevd Not tainted 5.14.0-rc4+ #3778
Hardware name: Qualcomm Technologies, Inc. APQ 8016 SBC (DT)
pstate: 80000005 (Nzcv daif -PAN -UAO -TCO BTYPE=--)
pc : vfe_hw_version_read+0x4/0x10 [qcom_camss]
platform 1a98000.dsi: Fixing up cyclic dependency with 3-0039
lr : msm_vfe_subdev_init+0x84/0x558 [qcom_camss]
platform 1a98000.dsi: Fixing up cyclic dependency with 1a01000.mdp
sp : ffff8000148337f0
x29: ffff8000148337f0 x28: ffff000006c1d800 x27: 0000000000000001
x26: 0000000000000002 x25: ffff8000091976d0 x24: ffff000006a98080
x23: 0000000000000000 x22: ffff000006c1d800 x21: ffff000000298010
x20: ffff800009197028 x19: ffff000000298000 x18: 0000000000000000
x17: 2e726f74616c7567 x16: 0000000000000028 x15: 0000b5f01da86af8
x14: 0000000000000001 x13: 000000000000001f x12: 0000000000000001
x11: 0000000000000010 x10: 0101010101010101 x9 : ffff8000091922f0
x8 : ffff8000091949f0 x7 : ffff800009194a08 x6 : ffff8000091925c8
x5 : ffff800009191f00 x4 : ffff000006a9ca00 x3 : ffff000006a9c080
x2 : ffff80000918d3a0 x1 : ffff000000298010 x0 : 0000000000000000
msm_mdp 1a01000.mdp: Adding to iommu group 3
Call trace:
  vfe_hw_version_read+0x4/0x10 [qcom_camss]
  camss_probe+0x370/0xdd0 [qcom_camss]
  platform_probe+0x90/0xd8
  really_probe+0x138/0x2d0
  __driver_probe_device+0x78/0xd8
  driver_probe_device+0x40/0x110
  __driver_attach+0xcc/0x118
  bus_for_each_dev+0x68/0xc8
  driver_attach+0x20/0x28
  bus_add_driver+0x168/0x1f8
  driver_register+0x60/0x110
  __platform_driver_register+0x24/0x30
  qcom_camss_driver_init+0x1c/0x1000 [qcom_camss]
  do_one_initcall+0x84/0x450
  do_init_module+0x58/0x200
  load_module+0x1f54/0x26a0
  __do_sys_finit_module+0xe4/0x100
  __arm64_sys_finit_module+0x18/0x20
  invoke_syscall+0x40/0xf8
  el0_svc_common+0x60/0x100
  do_el0_svc_compat+0x1c/0x48
  el0_svc_compat+0x20/0x30
  el0t_32_sync_handler+0xec/0x140
  el0t_32_sync+0x168/0x16c
Code: 88dffc63 88dffc63 00000000 f9400800 (b9400000)
---[ end trace e1e3b303eae40654 ]---

Let me know if I can help somehow to fix this issue.

> ---
>   drivers/media/platform/qcom/camss/camss-vfe.c | 3 +--
>   1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/drivers/media/platform/qcom/camss/camss-vfe.c b/drivers/media/platform/qcom/camss/camss-vfe.c
> index e0f3a36f3f3f..f6e889c1598e 100644
> --- a/drivers/media/platform/qcom/camss/camss-vfe.c
> +++ b/drivers/media/platform/qcom/camss/camss-vfe.c
> @@ -713,8 +713,6 @@ static int vfe_set_power(struct v4l2_subdev *sd, int on)
>   		ret = vfe_get(vfe);
>   		if (ret < 0)
>   			return ret;
> -
> -		vfe->ops->hw_version_read(vfe, vfe->camss->dev);
>   	} else {
>   		vfe_put(vfe);
>   	}
> @@ -1301,6 +1299,7 @@ int msm_vfe_subdev_init(struct camss *camss, struct vfe_device *vfe,
>   		return -EINVAL;
>   	}
>   	vfe->ops->subdev_init(dev, vfe);
> +	vfe->ops->hw_version_read(vfe, dev);
>   
>   	/* Memory */
>   

Best regards
Robert Foss Aug. 11, 2021, 7:32 a.m. UTC | #2
Hey Marek,



On Tue, 10 Aug 2021 at 23:46, Marek Szyprowski <m.szyprowski@samsung.com> wrote:
>
> Hi,
>
> On 26.07.2021 10:01, Robert Foss wrote:
> > vfe_set_power() is invoced a few times, which causes a lot of
> > VFE HW version prints.
> >
> > Secondly there is no real reason for calling hw_version_read() from
> > this location specifically, since it does nothing more than print
> > the VFE HW version.
> >
> > A better location for the hw_version_read() call would be somewhere
> > which is only executed once per VFE unit.
> >
> > Signed-off-by: Robert Foss <robert.foss@linaro.org>
>
> This patch landed recently in linux-next as commit b10b5334528a ("media:
> camss: vfe: Don't read hardware version needlessly"). Sadly it causes a
> following regression on DragonBoard410c
> (arch/arm64/boot/dts/qcom/apq8016-sbc.dts):
>
> Unable to handle kernel NULL pointer dereference at virtual address
> 0000000000000000
> Mem abort info:
>    ESR = 0x96000006
>    EC = 0x25: DABT (current EL), IL = 32 bits
>    SET = 0, FnV = 0
>    EA = 0, S1PTW = 0
>    FSC = 0x06: level 2 translation fault
> input: pm8941_resin as
> /devices/platform/soc/200f000.spmi/spmi-0/0-00/200f000.spmi:pmic@0:pon@800/200f000.spmi:pmic@0:pon@800:resin/input/input2
> Data abort info:
>    ISV = 0, ISS = 0x00000006
>    CM = 0, WnR = 0
> user pgtable: 4k pages, 48-bit VAs, pgdp=0000000087ca0000
> [0000000000000000] pgd=0800000084c4c003, p4d=0800000084c4c003,
> pud=08000000859c2003, pmd=0000000000000000
> Internal error: Oops: 96000006 [#1] PREEMPT SMP
> Modules linked in: msm(+) qcom_pon rtc_pm8xxx qcom_spmi_vadc
> qcom_vadc_common qcom_camss(+) snd_soc_msm8916_analog
> qcom_spmi_temp_alarm venus_core(+) videobuf2_dma_sg v4l2_fwnode
> snd_soc_lpass_apq8016 v4l2_async videobuf2_memops snd_soc_lpass_cpu
> v4l2_mem2mem snd_soc_apq8016_sbc snd_soc_lpass_platform
> snd_soc_msm8916_digital videobuf2_v4l2 snd_soc_qcom_common
> videobuf2_common qnoc_msm8916 qcom_rng videodev i2c_qcom_cci mdt_loader
> mc icc_smd_rpm crct10dif_ce socinfo rmtfs_mem display_connector
> adreno 1c00000.gpu: Adding to iommu group 2
> CPU: 1 PID: 286 Comm: systemd-udevd Not tainted 5.14.0-rc4+ #3778
> Hardware name: Qualcomm Technologies, Inc. APQ 8016 SBC (DT)
> pstate: 80000005 (Nzcv daif -PAN -UAO -TCO BTYPE=--)
> pc : vfe_hw_version_read+0x4/0x10 [qcom_camss]
> platform 1a98000.dsi: Fixing up cyclic dependency with 3-0039
> lr : msm_vfe_subdev_init+0x84/0x558 [qcom_camss]
> platform 1a98000.dsi: Fixing up cyclic dependency with 1a01000.mdp
> sp : ffff8000148337f0
> x29: ffff8000148337f0 x28: ffff000006c1d800 x27: 0000000000000001
> x26: 0000000000000002 x25: ffff8000091976d0 x24: ffff000006a98080
> x23: 0000000000000000 x22: ffff000006c1d800 x21: ffff000000298010
> x20: ffff800009197028 x19: ffff000000298000 x18: 0000000000000000
> x17: 2e726f74616c7567 x16: 0000000000000028 x15: 0000b5f01da86af8
> x14: 0000000000000001 x13: 000000000000001f x12: 0000000000000001
> x11: 0000000000000010 x10: 0101010101010101 x9 : ffff8000091922f0
> x8 : ffff8000091949f0 x7 : ffff800009194a08 x6 : ffff8000091925c8
> x5 : ffff800009191f00 x4 : ffff000006a9ca00 x3 : ffff000006a9c080
> x2 : ffff80000918d3a0 x1 : ffff000000298010 x0 : 0000000000000000
> msm_mdp 1a01000.mdp: Adding to iommu group 3
> Call trace:
>   vfe_hw_version_read+0x4/0x10 [qcom_camss]
>   camss_probe+0x370/0xdd0 [qcom_camss]
>   platform_probe+0x90/0xd8
>   really_probe+0x138/0x2d0
>   __driver_probe_device+0x78/0xd8
>   driver_probe_device+0x40/0x110
>   __driver_attach+0xcc/0x118
>   bus_for_each_dev+0x68/0xc8
>   driver_attach+0x20/0x28
>   bus_add_driver+0x168/0x1f8
>   driver_register+0x60/0x110
>   __platform_driver_register+0x24/0x30
>   qcom_camss_driver_init+0x1c/0x1000 [qcom_camss]
>   do_one_initcall+0x84/0x450
>   do_init_module+0x58/0x200
>   load_module+0x1f54/0x26a0
>   __do_sys_finit_module+0xe4/0x100
>   __arm64_sys_finit_module+0x18/0x20
>   invoke_syscall+0x40/0xf8
>   el0_svc_common+0x60/0x100
>   do_el0_svc_compat+0x1c/0x48
>   el0_svc_compat+0x20/0x30
>   el0t_32_sync_handler+0xec/0x140
>   el0t_32_sync+0x168/0x16c
> Code: 88dffc63 88dffc63 00000000 f9400800 (b9400000)
> ---[ end trace e1e3b303eae40654 ]---
>
> Let me know if I can help somehow to fix this issue.
>
> > ---
> >   drivers/media/platform/qcom/camss/camss-vfe.c | 3 +--
> >   1 file changed, 1 insertion(+), 2 deletions(-)
> >
> > diff --git a/drivers/media/platform/qcom/camss/camss-vfe.c b/drivers/media/platform/qcom/camss/camss-vfe.c
> > index e0f3a36f3f3f..f6e889c1598e 100644
> > --- a/drivers/media/platform/qcom/camss/camss-vfe.c
> > +++ b/drivers/media/platform/qcom/camss/camss-vfe.c
> > @@ -713,8 +713,6 @@ static int vfe_set_power(struct v4l2_subdev *sd, int on)
> >               ret = vfe_get(vfe);
> >               if (ret < 0)
> >                       return ret;
> > -
> > -             vfe->ops->hw_version_read(vfe, vfe->camss->dev);
> >       } else {
> >               vfe_put(vfe);
> >       }
> > @@ -1301,6 +1299,7 @@ int msm_vfe_subdev_init(struct camss *camss, struct vfe_device *vfe,
> >               return -EINVAL;
> >       }
> >       vfe->ops->subdev_init(dev, vfe);
> > +     vfe->ops->hw_version_read(vfe, dev);
> >
> >       /* Memory */
> >
>

Thanks for reporting this.

I submitted a patch[1] for this yesterday, if you have time to test it
and supply a Tested-by/Reviewed-by, that would be really helpful.

[1] https://lore.kernel.org/linux-arm-msm/20210810103336.114077-1-robert.foss@linaro.org/
diff mbox series

Patch

diff --git a/drivers/media/platform/qcom/camss/camss-vfe.c b/drivers/media/platform/qcom/camss/camss-vfe.c
index e0f3a36f3f3f..f6e889c1598e 100644
--- a/drivers/media/platform/qcom/camss/camss-vfe.c
+++ b/drivers/media/platform/qcom/camss/camss-vfe.c
@@ -713,8 +713,6 @@  static int vfe_set_power(struct v4l2_subdev *sd, int on)
 		ret = vfe_get(vfe);
 		if (ret < 0)
 			return ret;
-
-		vfe->ops->hw_version_read(vfe, vfe->camss->dev);
 	} else {
 		vfe_put(vfe);
 	}
@@ -1301,6 +1299,7 @@  int msm_vfe_subdev_init(struct camss *camss, struct vfe_device *vfe,
 		return -EINVAL;
 	}
 	vfe->ops->subdev_init(dev, vfe);
+	vfe->ops->hw_version_read(vfe, dev);
 
 	/* Memory */