Right now we use fixed indexes to assign power-domains, with a
requirement for the TOP GDSC to come last in the list.
Adding support for named power-domains means the declaration in the dtsi
can come in any order.
After this change we continue to support the old indexing - if a SoC
resource declaration or the in-use dtb doesn't declare power-domain names
we fall back to the default legacy indexing.
From this point on though new SoC additions should contain named
power-domains, eventually we will drop support for legacy indexing.
Tested-by: Matti Lehtimäki <[email protected]>
Signed-off-by: Bryan O'Donoghue <[email protected]>
---
drivers/media/platform/qcom/camss/camss-vfe.c | 24 +++++++++++++++++++++++-
drivers/media/platform/qcom/camss/camss.c | 26 +++++++++++++++++++++-----
drivers/media/platform/qcom/camss/camss.h | 2 ++
3 files changed, 46 insertions(+), 6 deletions(-)
diff --git a/drivers/media/platform/qcom/camss/camss-vfe.c b/drivers/media/platform/qcom/camss/camss-vfe.c
index defff24f07ce3..123e5ead7602d 100644
--- a/drivers/media/platform/qcom/camss/camss-vfe.c
+++ b/drivers/media/platform/qcom/camss/camss-vfe.c
@@ -1382,7 +1382,29 @@ int msm_vfe_subdev_init(struct camss *camss, struct vfe_device *vfe,
if (!res->line_num)
return -EINVAL;
- if (res->has_pd) {
+ /* Power domain */
+
+ if (res->pd_name) {
+ vfe->genpd = dev_pm_domain_attach_by_name(camss->dev,
+ res->pd_name);
+ if (IS_ERR(vfe->genpd)) {
+ ret = PTR_ERR(vfe->genpd);
+ return ret;
+ }
+ }
+
+ if (!vfe->genpd && res->has_pd) {
+ /*
+ * Legacy magic index.
+ * Requires
+ * power-domain = <VFE_X>,
+ * <VFE_Y>,
+ * <TITAN_TOP>
+ * id must correspondng to the index of the VFE which must
+ * come before the TOP GDSC. VFE Lite has no individually
+ * collapasible domain which is why id < vfe_num is a valid
+ * check.
+ */
vfe->genpd = dev_pm_domain_attach_by_id(camss->dev, id);
if (IS_ERR(vfe->genpd)) {
ret = PTR_ERR(vfe->genpd);
diff --git a/drivers/media/platform/qcom/camss/camss.c b/drivers/media/platform/qcom/camss/camss.c
index f10a47f1aee7d..d565260fee0d6 100644
--- a/drivers/media/platform/qcom/camss/camss.c
+++ b/drivers/media/platform/qcom/camss/camss.c
@@ -1523,12 +1523,28 @@ static int camss_configure_pd(struct camss *camss)
return 0;
/*
- * VFE power domains are in the beginning of the list, and while all
- * power domains should be attached, only if TITAN_TOP power domain is
- * found in the list, it should be linked over here.
+ * If a power-domain name is defined try to use it.
+ * It is possible we are running a new kernel with an old dtb so
+ * fallback to indexes even if a pd_name is defined but not found.
*/
- camss->genpd = dev_pm_domain_attach_by_id(camss->dev, camss->genpd_num - 1);
- if (IS_ERR(camss->genpd)) {
+ if (camss->res->pd_name) {
+ camss->genpd = dev_pm_domain_attach_by_name(camss->dev,
+ camss->res->pd_name);
+ if (IS_ERR(camss->genpd)) {
+ ret = PTR_ERR(camss->genpd);
+ goto fail_pm;
+ }
+ }
+
+ if (!camss->genpd) {
+ /*
+ * Legacy magic index. TITAN_TOP GDSC must be the last
+ * item in the power-domain list.
+ */
+ camss->genpd = dev_pm_domain_attach_by_id(camss->dev,
+ camss->genpd_num - 1);
+ }
+ if (IS_ERR_OR_NULL(camss->genpd)) {
ret = PTR_ERR(camss->genpd);
goto fail_pm;
}
diff --git a/drivers/media/platform/qcom/camss/camss.h b/drivers/media/platform/qcom/camss/camss.h
index 1ba824a2cb76c..cd8186fe1797b 100644
--- a/drivers/media/platform/qcom/camss/camss.h
+++ b/drivers/media/platform/qcom/camss/camss.h
@@ -48,6 +48,7 @@ struct camss_subdev_resources {
u32 clock_rate[CAMSS_RES_MAX][CAMSS_RES_MAX];
char *reg[CAMSS_RES_MAX];
char *interrupt[CAMSS_RES_MAX];
+ char *pd_name;
u8 line_num;
bool has_pd;
const void *ops;
@@ -84,6 +85,7 @@ enum icc_count {
struct camss_resources {
enum camss_version version;
+ const char *pd_name;
const struct camss_subdev_resources *csiphy_res;
const struct camss_subdev_resources *csid_res;
const struct camss_subdev_resources *ispif_res;
--
2.42.0
On 3.11.2023 13:29, Bryan O'Donoghue wrote:
> Right now we use fixed indexes to assign power-domains, with a
> requirement for the TOP GDSC to come last in the list.
>
> Adding support for named power-domains means the declaration in the dtsi
> can come in any order.
>
> After this change we continue to support the old indexing - if a SoC
> resource declaration or the in-use dtb doesn't declare power-domain names
> we fall back to the default legacy indexing.
>
> From this point on though new SoC additions should contain named
> power-domains, eventually we will drop support for legacy indexing.
>
> Tested-by: Matti Lehtimäki <[email protected]>
> Signed-off-by: Bryan O'Donoghue <[email protected]>
> ---
rg -l "&cc.* \{" arch/arm64/boot/dts/qcom # (the bus is named cci or cciN)
arch/arm64/boot/dts/qcom/sm7225-fairphone-fp4.dts
arch/arm64/boot/dts/qcom/sdm845-db845c-navigation-mezzanine.dts
arch/arm64/boot/dts/qcom/apq8016-sbc-d3-camera-mezzanine.dts
arch/arm64/boot/dts/qcom/qrb5165-rb5-vision-mezzanine.dts
the FP4 only has bus enablements (i.e. no attached sensors)
the number of working DB845c/RB3s is probably single-digit today
410c and rb5 are probably legit users.. part of me would like to
get rid of this messy old binding ASAP.. but then are there really
non-Bryans with these boards and the camera mezzs?
Konrad
On 18/11/2023 00:38, Konrad Dybcio wrote:
> On 3.11.2023 13:29, Bryan O'Donoghue wrote:
>> Right now we use fixed indexes to assign power-domains, with a
>> requirement for the TOP GDSC to come last in the list.
>>
>> Adding support for named power-domains means the declaration in the dtsi
>> can come in any order.
>>
>> After this change we continue to support the old indexing - if a SoC
>> resource declaration or the in-use dtb doesn't declare power-domain names
>> we fall back to the default legacy indexing.
>>
>> From this point on though new SoC additions should contain named
>> power-domains, eventually we will drop support for legacy indexing.
>>
>> Tested-by: Matti Lehtimäki <[email protected]>
>> Signed-off-by: Bryan O'Donoghue <[email protected]>
>> ---
> rg -l "&cc.* \{" arch/arm64/boot/dts/qcom # (the bus is named cci or cciN)
>
> arch/arm64/boot/dts/qcom/sm7225-fairphone-fp4.dts
> arch/arm64/boot/dts/qcom/sdm845-db845c-navigation-mezzanine.dts
> arch/arm64/boot/dts/qcom/apq8016-sbc-d3-camera-mezzanine.dts
> arch/arm64/boot/dts/qcom/qrb5165-rb5-vision-mezzanine.dts
>
> the FP4 only has bus enablements (i.e. no attached sensors)
> the number of working DB845c/RB3s is probably single-digit today
>
> 410c and rb5 are probably legit users.. part of me would like to
> get rid of this messy old binding ASAP.. but then are there really
> non-Bryans with these boards and the camera mezzs?
>
> Konrad
The plan here is to follow the procedures as I understand them.
- Deprecate
- Wait a reasonable amount of time ~ 1 year or so
- Drop
We could go faster but, there's no need. Drawing a line on new SoCs and
deprecating the old way will do fine.
---
bod