[PATCH v3 0/9] The great interconnecification fixation

[PATCH v3 0/9] The great interconnecification fixation

[Konrad Dybcio]

Hi!

This series grew quite a bit bigger than the previous [1] attempt, so
I decided to also add a cover letter.

Link to v2: [2]

It addresses a few things that were not quite right:

- Setting QoS params before a "real" (non-zero) bandwidth request
  makes little sense (since there's no data supposed to flow through
  the bus, why would the QoS matter) and (at least newer) downstream
  prevents that from happening. Do the same in Patch 1.

- QNoC type buses expect to always have their QoS registers set as long
  as there's a non-INVALID QoS mode set; ap_owned is not really a thing
  on these anymore, Patch 3 handles that.

- The recent MSM8996 boot fix was done quickly and not quite properly,
  leading to possibly setting the aggregate bus rate on "normal"
  hardware interface clocks; this series handles that by limiting the
  number of bus_clocks to 2 (which is the maximum that makes sense,
  anyway) and handling the rest as "intf_clocks", which are required
  to access the   hardware at the other end. Patches 5-8 take care of
  that and Patch 10 reverts the _optional moniker in clk_get_ to make
  sure we always have the bus scaling clocks, as they're well, kind
  of important :)

- Similarly to QNoC, BIMC on "newer" (which can be loosely approximated
  by "new enough" == "has only BIMC and QNoC hosts") SoCs expects to
  always receive QoS programming, whereas BIMC on "older" SoCs cries
  like a wild boar and crashes the platform when trying to do so
  unconditionally. Patch 9 adds a way to take care of that for newer
  SoCs (like SM6375)

- QoS mode INVALID was assumed by developers before to be the default
  ("I didn't specify any QoS settings, so the driver can't assume I
  did.. right? right!?" - wrong, partial struct initialization led to
  0 being set and 0 corresponded to QoS mode FIXED). Make it so, as
  that's the logical choice. This allows the "Always set QoS params
  on QNoC" patch to work without setting tons of what-should-
  -obviously-be-the-default values everywhere, as well as fixes older
  drivers that set ap_owned = true but left the QoS mode field unset.
  Patch 2 cleans that up.

- Some nodes are physically connected over more than one channel
  (usually DDR or other high-throughput paths). Patch 4 allows that
  to be reflected in calculations. This will be required for at least
  MSM8998 and SM6375 (which will be submitted soon after this lands)

v1 -> v2 changelog:
- reorder "make QoS INVALID default", makes more sense to have it
  before "Always set QoS params on QNoC"

- Limit ap_owned-independent QoS setting to QNoC only

- Add new patches for handling the 8996-and-friends clocks situation
  and optional BIMC regardless-of-ap_owned QoS programming

v2 -> v3 changelog:
- Drop "Don't set QoS params before non-zero bw is requested"

- Rebase on next

- [1/9] ("..make QoS INVALID default.."): remove unused define for
  MODE_INVALID_VAL

- Pick up tags

[1] https://lore.kernel.org/linux-arm-msm/14e06574-f95e-8960-0243-8c95a1c294e9@linaro.org/T/#m056692bea71d4c272968d5e07afbd9eb07a88123
[2] https://lore.kernel.org/linux-arm-msm/20230110132202.956619-1-konrad.dybcio@linaro.org/


Konrad Dybcio (9):
  interconnect: qcom: rpm: make QoS INVALID default, separate out driver
    data
  interconnect: qcom: rpm: Always set QoS params on QNoC
  interconnect: qcom: rpm: Add support for specifying channel num
  interconnect: qcom: Sort kerneldoc entries
  interconnect: qcom: rpm: Rename icc desc clocks to bus_blocks
  interconnect: qcom: rpm: Rename icc provider num_clocks to
    num_bus_clocks
  interconnect: qcom: rpm: Handle interface clocks
  interconnect: qcom: rpm: Add a way to always set QoS registers
  interconnect: qcom: rpm: Don't use clk_get_optional for bus clocks
    anymore

 drivers/interconnect/qcom/icc-rpm.c | 111 +++++++++++++++++++---------
 drivers/interconnect/qcom/icc-rpm.h |  36 ++++++---
 drivers/interconnect/qcom/msm8996.c |  22 +++---
 drivers/interconnect/qcom/sdm660.c  |  16 ++--
 4 files changed, 117 insertions(+), 68 deletions(-)

-- 
2.39.0

[PATCH v3 1/9] interconnect: qcom: rpm: make QoS INVALID default, separate out driver data

[Konrad Dybcio]

Currently, NOC_QOS_MODE_FIXED is defined as 0x0, which makes it the
"default" option (as that's what uninitialized members of partially
initialized structs are set to), which should really have been
NOC_QOS_MODE_INVALID, and that's what people (wrongly) assumed was
the case when .qos.qos_mode was not defined and what really makes
the most sense..

That resulted in {port 0, prio 0, areq_prio 0, urg_fwd = false, rpm-voted}
QoS being always voted for, because the code flow assumed "hey, it's fixed
QoS, so let's just roll with whatever parameters are set" [again, set by
partial struct initialization, as these fields were left unfilled by the
developers]. That is of course incorrect, and on many of these platforms
port 0 is MAS_APPS_PROC, which 9/10 times is supposed to be handled by
the ap_owned path, not to mention the rest of the parameters may differ.
Arguably, the APPS node is the most important one, next to EBI0..

The modes are defined as preprocessor constants. They are not used
anywhere outside the driver or sent to any remote processor outside
qcom_icc_set_noc_qos(), which is easily worked around.
Separate the type specified in driver data from the value sent to msmbus.
Make the former an enum for better mainainability.

This is an implicit fix for every SMD RPM ICC driver that didn't
explicitly specify NOC_QOS_MODE_INVALID on non-AP_owned nodes that
don't have QoS settings.

Fixes: 30c8fa3ec61a ("interconnect: qcom: Add MSM8916 interconnect provider driver")
Fixes: 6c6fe5d3dc5e ("interconnect: qcom: Add MSM8939 interconnect provider driver")
Fixes: 4e60a9568dc6 ("interconnect: qcom: add msm8974 driver")
Fixes: 7add937f5222 ("interconnect: qcom: Add MSM8996 interconnect provider driver")
Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c | 22 ++++++++++++----------
 drivers/interconnect/qcom/icc-rpm.h | 10 ++++++----
 2 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index df3196f72536..385a67c20956 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -48,6 +48,9 @@
 #define NOC_QOS_MODEn_ADDR(n)		(0xc + (n * 0x1000))
 #define NOC_QOS_MODEn_MASK		0x3
 
+#define NOC_QOS_MODE_FIXED_VAL		0x0
+#define NOC_QOS_MODE_BYPASS_VAL		0x2
+
 static int qcom_icc_set_qnoc_qos(struct icc_node *src, u64 max_bw)
 {
 	struct icc_provider *provider = src->provider;
@@ -153,7 +156,7 @@ static int qcom_icc_set_noc_qos(struct icc_node *src, u64 max_bw)
 	struct qcom_icc_provider *qp;
 	struct qcom_icc_node *qn;
 	struct icc_provider *provider;
-	u32 mode = NOC_QOS_MODE_BYPASS;
+	u32 mode = NOC_QOS_MODE_BYPASS_VAL;
 	int rc = 0;
 
 	qn = src->data;
@@ -167,18 +170,17 @@ static int qcom_icc_set_noc_qos(struct icc_node *src, u64 max_bw)
 		return 0;
 	}
 
-	if (qn->qos.qos_mode != NOC_QOS_MODE_INVALID)
-		mode = qn->qos.qos_mode;
-
-	if (mode == NOC_QOS_MODE_FIXED) {
-		dev_dbg(src->provider->dev, "NoC QoS: %s: Set Fixed mode\n",
-			qn->name);
+	if (qn->qos.qos_mode == NOC_QOS_MODE_FIXED) {
+		dev_dbg(src->provider->dev, "NoC QoS: %s: Set Fixed mode\n", qn->name);
+		mode = NOC_QOS_MODE_FIXED_VAL;
 		rc = qcom_icc_noc_set_qos_priority(qp, &qn->qos);
 		if (rc)
 			return rc;
-	} else if (mode == NOC_QOS_MODE_BYPASS) {
-		dev_dbg(src->provider->dev, "NoC QoS: %s: Set Bypass mode\n",
-			qn->name);
+	} else if (qn->qos.qos_mode == NOC_QOS_MODE_BYPASS) {
+		dev_dbg(src->provider->dev, "NoC QoS: %s: Set Bypass mode\n", qn->name);
+		mode = NOC_QOS_MODE_BYPASS_VAL;
+	} else {
+		/* How did we get here? */
 	}
 
 	return regmap_update_bits(qp->regmap,
diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
index a49af844ab13..8ba1918d7997 100644
--- a/drivers/interconnect/qcom/icc-rpm.h
+++ b/drivers/interconnect/qcom/icc-rpm.h
@@ -97,10 +97,12 @@ struct qcom_icc_desc {
 	unsigned int qos_offset;
 };
 
-/* Valid for both NoC and BIMC */
-#define NOC_QOS_MODE_INVALID		-1
-#define NOC_QOS_MODE_FIXED		0x0
-#define NOC_QOS_MODE_BYPASS		0x2
+/* Valid for all bus types */
+enum qos_mode {
+	NOC_QOS_MODE_INVALID = 0,
+	NOC_QOS_MODE_FIXED,
+	NOC_QOS_MODE_BYPASS,
+};
 
 int qnoc_probe(struct platform_device *pdev);
 int qnoc_remove(struct platform_device *pdev);
-- 
2.39.0

[PATCH v3 2/9] interconnect: qcom: rpm: Always set QoS params on QNoC

[Konrad Dybcio]

On newer SoCs, QoS parameters and RPM bandwidth requests are wholly
separate. Setting one should only depend on the description of the
interconnect node and not whether the other is present. If we don't
vote through RPM, QoS parameters should be set regardless, as we're
requesting additional bandwidth by setting the interconnect clock
rates.

With NoC (the old-SoC bus type), this is not the case and they are
mutually exclusive (so, the current upstream logic is correct).

For BIMC however, newer SoCs expect QoS params to be always set
(like QNoC) whereas older ones (like MSM8998) hang up completely when
doing so, hence this will be addressed in the next commit.

The Fixes tag references the commit in which this logic was added, it
has since been shuffled around to a different file, but it's the one
where it originates from.

Fixes: f80a1d414328 ("interconnect: qcom: Add SDM660 interconnect provider driver")
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index 385a67c20956..b1337f90c975 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -239,15 +239,27 @@ static int qcom_icc_rpm_set(int mas_rpm_id, int slv_rpm_id, u64 sum_bw)
 static int __qcom_icc_set(struct icc_node *n, struct qcom_icc_node *qn,
 			  u64 sum_bw)
 {
+	struct qcom_icc_provider *qp = to_qcom_provider(n->provider);
+	bool vote_ap, vote_rpm;
 	int ret;
 
-	if (!qn->qos.ap_owned) {
-		/* send bandwidth request message to the RPM processor */
+	if (qp->type == QCOM_ICC_QNOC) {
+		vote_ap = true;
+		vote_rpm = true;
+	} else {
+		vote_ap = qn->qos.ap_owned;
+		vote_rpm = !vote_ap;
+	}
+
+	if (vote_rpm) {
+		/* Send bandwidth request message to the RPM processor */
 		ret = qcom_icc_rpm_set(qn->mas_rpm_id, qn->slv_rpm_id, sum_bw);
 		if (ret)
 			return ret;
-	} else if (qn->qos.qos_mode != -1) {
-		/* set bandwidth directly from the AP */
+	}
+
+	if (vote_ap && qn->qos.qos_mode != NOC_QOS_MODE_INVALID) {
+		/* Set QoS params from the AP */
 		ret = qcom_icc_qos_set(n, sum_bw);
 		if (ret)
 			return ret;
-- 
2.39.0

[PATCH v3 3/9] interconnect: qcom: rpm: Add support for specifying channel num

[Konrad Dybcio]

Some nodes, like EBI0 (DDR) or L3/LLCC, may be connected over more than
one channel. This should be taken into account in bandwidth calcualtion,
as we're supposed to feed msmbus with the per-channel bandwidth. Add
support for specifying that and use it during bandwidth aggregation.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c | 7 ++++++-
 drivers/interconnect/qcom/icc-rpm.h | 2 ++
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index b1337f90c975..41a6dd40c38a 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -329,6 +329,7 @@ static void qcom_icc_bus_aggregate(struct icc_provider *provider,
 {
 	struct icc_node *node;
 	struct qcom_icc_node *qn;
+	u64 sum_avg[QCOM_ICC_NUM_BUCKETS];
 	int i;
 
 	/* Initialise aggregate values */
@@ -346,7 +347,11 @@ static void qcom_icc_bus_aggregate(struct icc_provider *provider,
 	list_for_each_entry(node, &provider->nodes, node_list) {
 		qn = node->data;
 		for (i = 0; i < QCOM_ICC_NUM_BUCKETS; i++) {
-			agg_avg[i] += qn->sum_avg[i];
+			if (qn->channels)
+				sum_avg[i] = div_u64(qn->sum_avg[i], qn->channels);
+			else
+				sum_avg[i] = qn->sum_avg[i];
+			agg_avg[i] += sum_avg[i];
 			agg_peak[i] = max_t(u64, agg_peak[i], qn->max_peak[i]);
 		}
 	}
diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
index 8ba1918d7997..8aed5400afda 100644
--- a/drivers/interconnect/qcom/icc-rpm.h
+++ b/drivers/interconnect/qcom/icc-rpm.h
@@ -66,6 +66,7 @@ struct qcom_icc_qos {
  * @id: a unique node identifier
  * @links: an array of nodes where we can go next while traversing
  * @num_links: the total number of @links
+ * @channels: number of channels at this node (e.g. DDR channels)
  * @buswidth: width of the interconnect between a node and the bus (bytes)
  * @sum_avg: current sum aggregate value of all avg bw requests
  * @max_peak: current max aggregate value of all peak bw requests
@@ -78,6 +79,7 @@ struct qcom_icc_node {
 	u16 id;
 	const u16 *links;
 	u16 num_links;
+	u16 channels;
 	u16 buswidth;
 	u64 sum_avg[QCOM_ICC_NUM_BUCKETS];
 	u64 max_peak[QCOM_ICC_NUM_BUCKETS];
-- 
2.39.0

[PATCH v3 4/9] interconnect: qcom: Sort kerneldoc entries

[Konrad Dybcio]

Sort the kerneldoc entries the same way the struct members are
sorted.

Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
index 8aed5400afda..21f440beda86 100644
--- a/drivers/interconnect/qcom/icc-rpm.h
+++ b/drivers/interconnect/qcom/icc-rpm.h
@@ -23,12 +23,12 @@ enum qcom_icc_type {
 /**
  * struct qcom_icc_provider - Qualcomm specific interconnect provider
  * @provider: generic interconnect provider
- * @bus_clks: the clk_bulk_data table of bus clocks
  * @num_clks: the total number of clk_bulk_data entries
  * @type: the ICC provider type
- * @qos_offset: offset to QoS registers
  * @regmap: regmap for QoS registers read/write access
+ * @qos_offset: offset to QoS registers
  * @bus_clk_rate: bus clock rate in Hz
+ * @bus_clks: the clk_bulk_data table of bus clocks
  */
 struct qcom_icc_provider {
 	struct icc_provider provider;
-- 
2.39.0

[PATCH v3 5/9] interconnect: qcom: rpm: Rename icc desc clocks to bus_blocks

[Konrad Dybcio]

Rename the "clocks" (and _names) fields of qcom_icc_desc to
"bus_clocks" in preparation for introducing handling of clocks that
need to be enabled but not voted on with aggregate frequency.

Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c |  6 +++---
 drivers/interconnect/qcom/icc-rpm.h |  4 ++--
 drivers/interconnect/qcom/msm8996.c | 12 ++++++------
 drivers/interconnect/qcom/sdm660.c  |  8 ++++----
 4 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index 41a6dd40c38a..31f6f09da4ae 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -453,9 +453,9 @@ int qnoc_probe(struct platform_device *pdev)
 	qnodes = desc->nodes;
 	num_nodes = desc->num_nodes;
 
-	if (desc->num_clocks) {
-		cds = desc->clocks;
-		cd_num = desc->num_clocks;
+	if (desc->num_bus_clocks) {
+		cds = desc->bus_clocks;
+		cd_num = desc->num_bus_clocks;
 	} else {
 		cds = bus_clocks;
 		cd_num = ARRAY_SIZE(bus_clocks);
diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
index 21f440beda86..d6b4c56bf02c 100644
--- a/drivers/interconnect/qcom/icc-rpm.h
+++ b/drivers/interconnect/qcom/icc-rpm.h
@@ -91,8 +91,8 @@ struct qcom_icc_node {
 struct qcom_icc_desc {
 	struct qcom_icc_node * const *nodes;
 	size_t num_nodes;
-	const char * const *clocks;
-	size_t num_clocks;
+	const char * const *bus_clocks;
+	size_t num_bus_clocks;
 	bool has_bus_pd;
 	enum qcom_icc_type type;
 	const struct regmap_config *regmap_cfg;
diff --git a/drivers/interconnect/qcom/msm8996.c b/drivers/interconnect/qcom/msm8996.c
index 25a1a32bc611..69fc50a6fa5c 100644
--- a/drivers/interconnect/qcom/msm8996.c
+++ b/drivers/interconnect/qcom/msm8996.c
@@ -1821,8 +1821,8 @@ static const struct qcom_icc_desc msm8996_a0noc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = a0noc_nodes,
 	.num_nodes = ARRAY_SIZE(a0noc_nodes),
-	.clocks = bus_a0noc_clocks,
-	.num_clocks = ARRAY_SIZE(bus_a0noc_clocks),
+	.bus_clocks = bus_a0noc_clocks,
+	.num_bus_clocks = ARRAY_SIZE(bus_a0noc_clocks),
 	.has_bus_pd = true,
 	.regmap_cfg = &msm8996_a0noc_regmap_config
 };
@@ -1866,8 +1866,8 @@ static const struct qcom_icc_desc msm8996_a2noc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = a2noc_nodes,
 	.num_nodes = ARRAY_SIZE(a2noc_nodes),
-	.clocks = bus_a2noc_clocks,
-	.num_clocks = ARRAY_SIZE(bus_a2noc_clocks),
+	.bus_clocks = bus_a2noc_clocks,
+	.num_bus_clocks = ARRAY_SIZE(bus_a2noc_clocks),
 	.regmap_cfg = &msm8996_a2noc_regmap_config
 };
 
@@ -2005,8 +2005,8 @@ static const struct qcom_icc_desc msm8996_mnoc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = mnoc_nodes,
 	.num_nodes = ARRAY_SIZE(mnoc_nodes),
-	.clocks = bus_mm_clocks,
-	.num_clocks = ARRAY_SIZE(bus_mm_clocks),
+	.bus_clocks = bus_mm_clocks,
+	.num_bus_clocks = ARRAY_SIZE(bus_mm_clocks),
 	.regmap_cfg = &msm8996_mnoc_regmap_config
 };
 
diff --git a/drivers/interconnect/qcom/sdm660.c b/drivers/interconnect/qcom/sdm660.c
index 8d879b0bcabc..a22ba821efbf 100644
--- a/drivers/interconnect/qcom/sdm660.c
+++ b/drivers/interconnect/qcom/sdm660.c
@@ -1516,8 +1516,8 @@ static const struct qcom_icc_desc sdm660_a2noc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = sdm660_a2noc_nodes,
 	.num_nodes = ARRAY_SIZE(sdm660_a2noc_nodes),
-	.clocks = bus_a2noc_clocks,
-	.num_clocks = ARRAY_SIZE(bus_a2noc_clocks),
+	.bus_clocks = bus_a2noc_clocks,
+	.num_bus_clocks = ARRAY_SIZE(bus_a2noc_clocks),
 	.regmap_cfg = &sdm660_a2noc_regmap_config,
 };
 
@@ -1659,8 +1659,8 @@ static const struct qcom_icc_desc sdm660_mnoc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = sdm660_mnoc_nodes,
 	.num_nodes = ARRAY_SIZE(sdm660_mnoc_nodes),
-	.clocks = bus_mm_clocks,
-	.num_clocks = ARRAY_SIZE(bus_mm_clocks),
+	.bus_clocks = bus_mm_clocks,
+	.num_bus_clocks = ARRAY_SIZE(bus_mm_clocks),
 	.regmap_cfg = &sdm660_mnoc_regmap_config,
 };
 
-- 
2.39.0

[PATCH v3 6/9] interconnect: qcom: rpm: Rename icc provider num_clocks to num_bus_clocks

[Konrad Dybcio]

In preparation for handling non-scaling clocks that we still have to
enable, rename num_clocks to more descriptive num_bus_clocks.

Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c | 14 +++++++-------
 drivers/interconnect/qcom/icc-rpm.h |  4 ++--
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index 31f6f09da4ae..5888241c0477 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -392,7 +392,7 @@ static int qcom_icc_set(struct icc_node *src, struct icc_node *dst)
 			return ret;
 	}
 
-	for (i = 0; i < qp->num_clks; i++) {
+	for (i = 0; i < qp->num_bus_clks; i++) {
 		/*
 		 * Use WAKE bucket for active clock, otherwise, use SLEEP bucket
 		 * for other clocks.  If a platform doesn't set interconnect
@@ -477,7 +477,7 @@ int qnoc_probe(struct platform_device *pdev)
 
 	for (i = 0; i < cd_num; i++)
 		qp->bus_clks[i].id = cds[i];
-	qp->num_clks = cd_num;
+	qp->num_bus_clks = cd_num;
 
 	qp->type = desc->type;
 	qp->qos_offset = desc->qos_offset;
@@ -507,11 +507,11 @@ int qnoc_probe(struct platform_device *pdev)
 	}
 
 regmap_done:
-	ret = devm_clk_bulk_get_optional(dev, qp->num_clks, qp->bus_clks);
+	ret = devm_clk_bulk_get_optional(dev, qp->num_bus_clks, qp->bus_clks);
 	if (ret)
 		return ret;
 
-	ret = clk_bulk_prepare_enable(qp->num_clks, qp->bus_clks);
+	ret = clk_bulk_prepare_enable(qp->num_bus_clks, qp->bus_clks);
 	if (ret)
 		return ret;
 
@@ -533,7 +533,7 @@ int qnoc_probe(struct platform_device *pdev)
 	ret = icc_provider_add(provider);
 	if (ret) {
 		dev_err(dev, "error adding interconnect provider: %d\n", ret);
-		clk_bulk_disable_unprepare(qp->num_clks, qp->bus_clks);
+		clk_bulk_disable_unprepare(qp->num_bus_clks, qp->bus_clks);
 		return ret;
 	}
 
@@ -566,7 +566,7 @@ int qnoc_probe(struct platform_device *pdev)
 	return 0;
 err:
 	icc_nodes_remove(provider);
-	clk_bulk_disable_unprepare(qp->num_clks, qp->bus_clks);
+	clk_bulk_disable_unprepare(qp->num_bus_clks, qp->bus_clks);
 	icc_provider_del(provider);
 
 	return ret;
@@ -578,7 +578,7 @@ int qnoc_remove(struct platform_device *pdev)
 	struct qcom_icc_provider *qp = platform_get_drvdata(pdev);
 
 	icc_nodes_remove(&qp->provider);
-	clk_bulk_disable_unprepare(qp->num_clks, qp->bus_clks);
+	clk_bulk_disable_unprepare(qp->num_bus_clks, qp->bus_clks);
 	icc_provider_del(&qp->provider);
 
 	return 0;
diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
index d6b4c56bf02c..d4401f35f6d2 100644
--- a/drivers/interconnect/qcom/icc-rpm.h
+++ b/drivers/interconnect/qcom/icc-rpm.h
@@ -23,7 +23,7 @@ enum qcom_icc_type {
 /**
  * struct qcom_icc_provider - Qualcomm specific interconnect provider
  * @provider: generic interconnect provider
- * @num_clks: the total number of clk_bulk_data entries
+ * @num_bus_clks: the total number of bus_clks clk_bulk_data entries
  * @type: the ICC provider type
  * @regmap: regmap for QoS registers read/write access
  * @qos_offset: offset to QoS registers
@@ -32,7 +32,7 @@ enum qcom_icc_type {
  */
 struct qcom_icc_provider {
 	struct icc_provider provider;
-	int num_clks;
+	int num_bus_clks;
 	enum qcom_icc_type type;
 	struct regmap *regmap;
 	unsigned int qos_offset;
-- 
2.39.0

[PATCH v3 7/9] interconnect: qcom: rpm: Handle interface clocks

[Konrad Dybcio]

Some (but not all) providers (or their specific nodes) require
specific clocks to be turned on before they can be accessed. Failure
to ensure that results in a seemingly random system crash (which
would usually happen at boot with the interconnect driver built-in),
resulting in the platform not booting up properly.

Limit the number of bus_clocks to 2 (which is the maximum that SMD
RPM interconnect supports anyway) and handle non-scaling clocks
separately. Update MSM8996 and SDM660 drivers to make sure they do
not regress with this change.

This unfortunately has to be done in one patch to prevent either
compile errors or broken bisect.

Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c | 48 ++++++++++++++++++++++-------
 drivers/interconnect/qcom/icc-rpm.h | 10 ++++--
 drivers/interconnect/qcom/msm8996.c | 22 ++++++-------
 drivers/interconnect/qcom/sdm660.c  | 16 ++++------
 4 files changed, 60 insertions(+), 36 deletions(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index 5888241c0477..509cadf398e9 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -453,28 +453,43 @@ int qnoc_probe(struct platform_device *pdev)
 	qnodes = desc->nodes;
 	num_nodes = desc->num_nodes;
 
-	if (desc->num_bus_clocks) {
-		cds = desc->bus_clocks;
-		cd_num = desc->num_bus_clocks;
+	if (desc->num_intf_clocks) {
+		cds = desc->intf_clocks;
+		cd_num = desc->num_intf_clocks;
 	} else {
-		cds = bus_clocks;
-		cd_num = ARRAY_SIZE(bus_clocks);
+		/* 0 intf clocks is perfectly fine */
+		cd_num = 0;
 	}
 
-	qp = devm_kzalloc(dev, struct_size(qp, bus_clks, cd_num), GFP_KERNEL);
+	qp = devm_kzalloc(dev, struct_size(qp, intf_clks, cd_num), GFP_KERNEL);
 	if (!qp)
 		return -ENOMEM;
 
-	qp->bus_clk_rate = devm_kcalloc(dev, cd_num, sizeof(*qp->bus_clk_rate),
-					GFP_KERNEL);
-	if (!qp->bus_clk_rate)
-		return -ENOMEM;
-
 	data = devm_kzalloc(dev, struct_size(data, nodes, num_nodes),
 			    GFP_KERNEL);
 	if (!data)
 		return -ENOMEM;
 
+	for (i = 0; i < cd_num; i++)
+		qp->intf_clks[i].id = cds[i];
+	qp->num_intf_clks = cd_num;
+
+	if (desc->num_bus_clocks) {
+		cds = desc->bus_clocks;
+		cd_num = desc->num_bus_clocks;
+	} else {
+		cds = bus_clocks;
+		cd_num = ARRAY_SIZE(bus_clocks);
+	}
+
+	/*
+	 * This is not realistic, scaling is only possible with an
+	 * always-active and an active-only clock, or at least one
+	 * of them in some very bizzare cases.
+	 */
+	if (WARN_ON(cd_num > 2))
+		cd_num = 2;
+
 	for (i = 0; i < cd_num; i++)
 		qp->bus_clks[i].id = cds[i];
 	qp->num_bus_clks = cd_num;
@@ -515,6 +530,14 @@ int qnoc_probe(struct platform_device *pdev)
 	if (ret)
 		return ret;
 
+	ret = devm_clk_bulk_get(dev, qp->num_intf_clks, qp->intf_clks);
+	if (ret)
+		return ret;
+
+	ret = clk_bulk_prepare_enable(qp->num_intf_clks, qp->intf_clks);
+	if (ret)
+		return ret;
+
 	if (desc->has_bus_pd) {
 		ret = dev_pm_domain_attach(dev, true);
 		if (ret)
@@ -533,6 +556,7 @@ int qnoc_probe(struct platform_device *pdev)
 	ret = icc_provider_add(provider);
 	if (ret) {
 		dev_err(dev, "error adding interconnect provider: %d\n", ret);
+		clk_bulk_disable_unprepare(qp->num_intf_clks, qp->intf_clks);
 		clk_bulk_disable_unprepare(qp->num_bus_clks, qp->bus_clks);
 		return ret;
 	}
@@ -566,6 +590,7 @@ int qnoc_probe(struct platform_device *pdev)
 	return 0;
 err:
 	icc_nodes_remove(provider);
+	clk_bulk_disable_unprepare(qp->num_intf_clks, qp->intf_clks);
 	clk_bulk_disable_unprepare(qp->num_bus_clks, qp->bus_clks);
 	icc_provider_del(provider);
 
@@ -578,6 +603,7 @@ int qnoc_remove(struct platform_device *pdev)
 	struct qcom_icc_provider *qp = platform_get_drvdata(pdev);
 
 	icc_nodes_remove(&qp->provider);
+	clk_bulk_disable_unprepare(qp->num_intf_clks, qp->intf_clks);
 	clk_bulk_disable_unprepare(qp->num_bus_clks, qp->bus_clks);
 	icc_provider_del(&qp->provider);
 
diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
index d4401f35f6d2..729573f0d9fe 100644
--- a/drivers/interconnect/qcom/icc-rpm.h
+++ b/drivers/interconnect/qcom/icc-rpm.h
@@ -24,20 +24,24 @@ enum qcom_icc_type {
  * struct qcom_icc_provider - Qualcomm specific interconnect provider
  * @provider: generic interconnect provider
  * @num_bus_clks: the total number of bus_clks clk_bulk_data entries
+ * @num_intf_clks: the total number of intf_clks clk_bulk_data entries
  * @type: the ICC provider type
  * @regmap: regmap for QoS registers read/write access
  * @qos_offset: offset to QoS registers
  * @bus_clk_rate: bus clock rate in Hz
  * @bus_clks: the clk_bulk_data table of bus clocks
+ * @intf_clks: the clk_bulk_data table of interface clocks
  */
 struct qcom_icc_provider {
 	struct icc_provider provider;
 	int num_bus_clks;
+	int num_intf_clks;
 	enum qcom_icc_type type;
 	struct regmap *regmap;
 	unsigned int qos_offset;
-	u64 *bus_clk_rate;
-	struct clk_bulk_data bus_clks[];
+	u64 bus_clk_rate[2];
+	struct clk_bulk_data bus_clks[2];
+	struct clk_bulk_data intf_clks[];
 };
 
 /**
@@ -93,6 +97,8 @@ struct qcom_icc_desc {
 	size_t num_nodes;
 	const char * const *bus_clocks;
 	size_t num_bus_clocks;
+	const char * const *intf_clocks;
+	size_t num_intf_clocks;
 	bool has_bus_pd;
 	enum qcom_icc_type type;
 	const struct regmap_config *regmap_cfg;
diff --git a/drivers/interconnect/qcom/msm8996.c b/drivers/interconnect/qcom/msm8996.c
index 69fc50a6fa5c..1a5e0ad36cc4 100644
--- a/drivers/interconnect/qcom/msm8996.c
+++ b/drivers/interconnect/qcom/msm8996.c
@@ -21,21 +21,17 @@
 #include "smd-rpm.h"
 #include "msm8996.h"
 
-static const char * const bus_mm_clocks[] = {
-	"bus",
-	"bus_a",
+static const char * const mm_intf_clocks[] = {
 	"iface"
 };
 
-static const char * const bus_a0noc_clocks[] = {
+static const char * const a0noc_intf_clocks[] = {
 	"aggre0_snoc_axi",
 	"aggre0_cnoc_ahb",
 	"aggre0_noc_mpu_cfg"
 };
 
-static const char * const bus_a2noc_clocks[] = {
-	"bus",
-	"bus_a",
+static const char * const a2noc_intf_clocks[] = {
 	"aggre2_ufs_axi",
 	"ufs_axi"
 };
@@ -1821,8 +1817,8 @@ static const struct qcom_icc_desc msm8996_a0noc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = a0noc_nodes,
 	.num_nodes = ARRAY_SIZE(a0noc_nodes),
-	.bus_clocks = bus_a0noc_clocks,
-	.num_bus_clocks = ARRAY_SIZE(bus_a0noc_clocks),
+	.intf_clocks = a0noc_intf_clocks,
+	.num_intf_clocks = ARRAY_SIZE(a0noc_intf_clocks),
 	.has_bus_pd = true,
 	.regmap_cfg = &msm8996_a0noc_regmap_config
 };
@@ -1866,8 +1862,8 @@ static const struct qcom_icc_desc msm8996_a2noc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = a2noc_nodes,
 	.num_nodes = ARRAY_SIZE(a2noc_nodes),
-	.bus_clocks = bus_a2noc_clocks,
-	.num_bus_clocks = ARRAY_SIZE(bus_a2noc_clocks),
+	.intf_clocks = a2noc_intf_clocks,
+	.num_intf_clocks = ARRAY_SIZE(a2noc_intf_clocks),
 	.regmap_cfg = &msm8996_a2noc_regmap_config
 };
 
@@ -2005,8 +2001,8 @@ static const struct qcom_icc_desc msm8996_mnoc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = mnoc_nodes,
 	.num_nodes = ARRAY_SIZE(mnoc_nodes),
-	.bus_clocks = bus_mm_clocks,
-	.num_bus_clocks = ARRAY_SIZE(bus_mm_clocks),
+	.intf_clocks = mm_intf_clocks,
+	.num_intf_clocks = ARRAY_SIZE(mm_intf_clocks),
 	.regmap_cfg = &msm8996_mnoc_regmap_config
 };
 
diff --git a/drivers/interconnect/qcom/sdm660.c b/drivers/interconnect/qcom/sdm660.c
index a22ba821efbf..0e8a96f4ce90 100644
--- a/drivers/interconnect/qcom/sdm660.c
+++ b/drivers/interconnect/qcom/sdm660.c
@@ -127,15 +127,11 @@ enum {
 	SDM660_SNOC,
 };
 
-static const char * const bus_mm_clocks[] = {
-	"bus",
-	"bus_a",
+static const char * const mm_intf_clocks[] = {
 	"iface",
 };
 
-static const char * const bus_a2noc_clocks[] = {
-	"bus",
-	"bus_a",
+static const char * const a2noc_intf_clocks[] = {
 	"ipa",
 	"ufs_axi",
 	"aggre2_ufs_axi",
@@ -1516,8 +1512,8 @@ static const struct qcom_icc_desc sdm660_a2noc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = sdm660_a2noc_nodes,
 	.num_nodes = ARRAY_SIZE(sdm660_a2noc_nodes),
-	.bus_clocks = bus_a2noc_clocks,
-	.num_bus_clocks = ARRAY_SIZE(bus_a2noc_clocks),
+	.intf_clocks = a2noc_intf_clocks,
+	.num_intf_clocks = ARRAY_SIZE(a2noc_intf_clocks),
 	.regmap_cfg = &sdm660_a2noc_regmap_config,
 };
 
@@ -1659,8 +1655,8 @@ static const struct qcom_icc_desc sdm660_mnoc = {
 	.type = QCOM_ICC_NOC,
 	.nodes = sdm660_mnoc_nodes,
 	.num_nodes = ARRAY_SIZE(sdm660_mnoc_nodes),
-	.bus_clocks = bus_mm_clocks,
-	.num_bus_clocks = ARRAY_SIZE(bus_mm_clocks),
+	.intf_clocks = mm_intf_clocks,
+	.num_intf_clocks = ARRAY_SIZE(mm_intf_clocks),
 	.regmap_cfg = &sdm660_mnoc_regmap_config,
 };
 
-- 
2.39.0

[PATCH v3 8/9] interconnect: qcom: rpm: Add a way to always set QoS registers

[Konrad Dybcio]

On newer SoCs (there's no clear boundary, but probably "new enough"
means every interconnect provider is either BIMC or QNoC and there
are no old-style NoC hosts) we're expected to set QoS registers
regardless of the ap_owned param. Add a bool in the qcom_icc_provider
and make the logic assume it's okay to set the registers when it's
set.

Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c | 2 +-
 drivers/interconnect/qcom/icc-rpm.h | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index 509cadf398e9..343e6021a93a 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -243,7 +243,7 @@ static int __qcom_icc_set(struct icc_node *n, struct qcom_icc_node *qn,
 	bool vote_ap, vote_rpm;
 	int ret;
 
-	if (qp->type == QCOM_ICC_QNOC) {
+	if (qp->type == QCOM_ICC_QNOC || qp->always_set_qos) {
 		vote_ap = true;
 		vote_rpm = true;
 	} else {
diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
index 729573f0d9fe..3c8888482a39 100644
--- a/drivers/interconnect/qcom/icc-rpm.h
+++ b/drivers/interconnect/qcom/icc-rpm.h
@@ -28,6 +28,7 @@ enum qcom_icc_type {
  * @type: the ICC provider type
  * @regmap: regmap for QoS registers read/write access
  * @qos_offset: offset to QoS registers
+ * @always_set_qos: whether to always set QoS registers regardless of bus type
  * @bus_clk_rate: bus clock rate in Hz
  * @bus_clks: the clk_bulk_data table of bus clocks
  * @intf_clks: the clk_bulk_data table of interface clocks
@@ -39,6 +40,7 @@ struct qcom_icc_provider {
 	enum qcom_icc_type type;
 	struct regmap *regmap;
 	unsigned int qos_offset;
+	bool always_set_qos;
 	u64 bus_clk_rate[2];
 	struct clk_bulk_data bus_clks[2];
 	struct clk_bulk_data intf_clks[];
-- 
2.39.0

[PATCH v3 9/9] interconnect: qcom: rpm: Don't use clk_get_optional for bus clocks anymore

[Konrad Dybcio]

Commit dd42ec8ea5b9 ("interconnect: qcom: rpm: Use _optional func for provider clocks")
relaxed the requirements around probing bus clocks. This was a decent
solution for making sure MSM8996 would still boot with old DTs, but
now that there's a proper fix in place that both old and new DTs
will be happy about, revert back to the safer variant of the
function.

Fixes: dd42ec8ea5b9 ("interconnect: qcom: rpm: Use _optional func for provider clocks")
Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
---
 drivers/interconnect/qcom/icc-rpm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
index 343e6021a93a..8cff724661f2 100644
--- a/drivers/interconnect/qcom/icc-rpm.c
+++ b/drivers/interconnect/qcom/icc-rpm.c
@@ -522,7 +522,7 @@ int qnoc_probe(struct platform_device *pdev)
 	}
 
 regmap_done:
-	ret = devm_clk_bulk_get_optional(dev, qp->num_bus_clks, qp->bus_clks);
+	ret = devm_clk_bulk_get(dev, qp->num_bus_clks, qp->bus_clks);
 	if (ret)
 		return ret;
 
-- 
2.39.0

Re: [PATCH v3 8/9] interconnect: qcom: rpm: Add a way to always set QoS registers

[Bryan O'Donoghue]

On 16/01/2023 13:21, Konrad Dybcio wrote:
> On newer SoCs (there's no clear boundary, but probably "new enough"
> means every interconnect provider is either BIMC or QNoC and there
> are no old-style NoC hosts) we're expected to set QoS registers
> regardless of the ap_owned param. Add a bool in the qcom_icc_provider
> and make the logic assume it's okay to set the registers when it's
> set.
> 
> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
> ---
>   drivers/interconnect/qcom/icc-rpm.c | 2 +-
>   drivers/interconnect/qcom/icc-rpm.h | 2 ++
>   2 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
> index 509cadf398e9..343e6021a93a 100644

On next-20230116,

git am < 
~/Downloads/v3-1-9-interconnect-qcom-rpm-make-QoS-INVALID-default-separate-out-driver-data.patch
Applying: interconnect: qcom: rpm: make QoS INVALID default, separate 
out driver data
Applying: interconnect: qcom: rpm: Add support for specifying channel num
Applying: interconnect: qcom: rpm: Rename icc desc clocks to bus_blocks
Applying: interconnect: qcom: rpm: Add a way to always set QoS registers

error: patch failed: drivers/interconnect/qcom/icc-rpm.c:243
error: drivers/interconnect/qcom/icc-rpm.c: patch does not apply
error: patch failed: drivers/interconnect/qcom/icc-rpm.h:28
error: drivers/interconnect/qcom/icc-rpm.h: patch does not apply
Patch failed at 0004 interconnect: qcom: rpm: Add a way to always set 
QoS registers
hint: Use 'git am --show-current-patch=diff' to see the failed patch

Should I pick up another series from elsewhere to apply this ?

---
bod

Re: [PATCH v3 8/9] interconnect: qcom: rpm: Add a way to always set QoS registers

[Konrad Dybcio]

On 17.01.2023 14:27, Bryan O'Donoghue wrote:
> On 16/01/2023 13:21, Konrad Dybcio wrote:
>> On newer SoCs (there's no clear boundary, but probably "new enough"
>> means every interconnect provider is either BIMC or QNoC and there
>> are no old-style NoC hosts) we're expected to set QoS registers
>> regardless of the ap_owned param. Add a bool in the qcom_icc_provider
>> and make the logic assume it's okay to set the registers when it's
>> set.
>>
>> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
>> ---
>>   drivers/interconnect/qcom/icc-rpm.c | 2 +-
>>   drivers/interconnect/qcom/icc-rpm.h | 2 ++
>>   2 files changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
>> index 509cadf398e9..343e6021a93a 100644
> 
> On next-20230116,
g fetch linux-next next-20230116
gco FETCH_HEAD
b4 am 20230116132152.405535-1-konrad.dybcio@linaro.org
g am ./v3_20230116_konrad_dybcio_the_great_interconnecification_fixation.mbx

Stosowanie: interconnect: qcom: rpm: make QoS INVALID default, separate out driver data
Stosowanie: interconnect: qcom: rpm: Always set QoS params on QNoC
Stosowanie: interconnect: qcom: rpm: Add support for specifying channel num
Stosowanie: interconnect: qcom: Sort kerneldoc entries
Stosowanie: interconnect: qcom: rpm: Rename icc desc clocks to bus_blocks
Stosowanie: interconnect: qcom: rpm: Rename icc provider num_clocks to num_bus_clocks
Stosowanie: interconnect: qcom: rpm: Handle interface clocks
Stosowanie: interconnect: qcom: rpm: Add a way to always set QoS registers
Stosowanie: interconnect: qcom: rpm: Don't use clk_get_optional for bus clocks anymore
(no errors)

Konrad
> 
> git am < ~/Downloads/v3-1-9-interconnect-qcom-rpm-make-QoS-INVALID-default-separate-out-driver-data.patch
> Applying: interconnect: qcom: rpm: make QoS INVALID default, separate out driver data
> Applying: interconnect: qcom: rpm: Add support for specifying channel num
> Applying: interconnect: qcom: rpm: Rename icc desc clocks to bus_blocks
> Applying: interconnect: qcom: rpm: Add a way to always set QoS registers
> 
> error: patch failed: drivers/interconnect/qcom/icc-rpm.c:243
> error: drivers/interconnect/qcom/icc-rpm.c: patch does not apply
> error: patch failed: drivers/interconnect/qcom/icc-rpm.h:28
> error: drivers/interconnect/qcom/icc-rpm.h: patch does not apply
> Patch failed at 0004 interconnect: qcom: rpm: Add a way to always set QoS registers
> hint: Use 'git am --show-current-patch=diff' to see the failed patch
> 
> Should I pick up another series from elsewhere to apply this ?
> 
> ---
> bod

Re: [PATCH v3 8/9] interconnect: qcom: rpm: Add a way to always set QoS registers

[Konrad Dybcio]

On 16.01.2023 14:21, Konrad Dybcio wrote:
> On newer SoCs (there's no clear boundary, but probably "new enough"
> means every interconnect provider is either BIMC or QNoC and there
> are no old-style NoC hosts) we're expected to set QoS registers
> regardless of the ap_owned param. Add a bool in the qcom_icc_provider
> and make the logic assume it's okay to set the registers when it's
> set.
> 
> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
> ---
Argh, I only noticed now that this patch is incomplete,
missing setting the value in qnoc_probe() and adding a
similar struct member to qcom_icc_desc..

Konrad
>  drivers/interconnect/qcom/icc-rpm.c | 2 +-
>  drivers/interconnect/qcom/icc-rpm.h | 2 ++
>  2 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
> index 509cadf398e9..343e6021a93a 100644
> --- a/drivers/interconnect/qcom/icc-rpm.c
> +++ b/drivers/interconnect/qcom/icc-rpm.c
> @@ -243,7 +243,7 @@ static int __qcom_icc_set(struct icc_node *n, struct qcom_icc_node *qn,
>  	bool vote_ap, vote_rpm;
>  	int ret;
>  
> -	if (qp->type == QCOM_ICC_QNOC) {
> +	if (qp->type == QCOM_ICC_QNOC || qp->always_set_qos) {
>  		vote_ap = true;
>  		vote_rpm = true;
>  	} else {
> diff --git a/drivers/interconnect/qcom/icc-rpm.h b/drivers/interconnect/qcom/icc-rpm.h
> index 729573f0d9fe..3c8888482a39 100644
> --- a/drivers/interconnect/qcom/icc-rpm.h
> +++ b/drivers/interconnect/qcom/icc-rpm.h
> @@ -28,6 +28,7 @@ enum qcom_icc_type {
>   * @type: the ICC provider type
>   * @regmap: regmap for QoS registers read/write access
>   * @qos_offset: offset to QoS registers
> + * @always_set_qos: whether to always set QoS registers regardless of bus type
>   * @bus_clk_rate: bus clock rate in Hz
>   * @bus_clks: the clk_bulk_data table of bus clocks
>   * @intf_clks: the clk_bulk_data table of interface clocks
> @@ -39,6 +40,7 @@ struct qcom_icc_provider {
>  	enum qcom_icc_type type;
>  	struct regmap *regmap;
>  	unsigned int qos_offset;
> +	bool always_set_qos;
>  	u64 bus_clk_rate[2];
>  	struct clk_bulk_data bus_clks[2];
>  	struct clk_bulk_data intf_clks[];

Re: [PATCH v3 2/9] interconnect: qcom: rpm: Always set QoS params on QNoC

[Konrad Dybcio]

On 16.01.2023 14:21, Konrad Dybcio wrote:
> On newer SoCs, QoS parameters and RPM bandwidth requests are wholly
> separate. Setting one should only depend on the description of the
> interconnect node and not whether the other is present. If we don't
> vote through RPM, QoS parameters should be set regardless, as we're
> requesting additional bandwidth by setting the interconnect clock
> rates.
> 
> With NoC (the old-SoC bus type), this is not the case and they are
> mutually exclusive (so, the current upstream logic is correct).
> 
> For BIMC however, newer SoCs expect QoS params to be always set
> (like QNoC) whereas older ones (like MSM8998) hang up completely when
> doing so, hence this will be addressed in the next commit.
> 
> The Fixes tag references the commit in which this logic was added, it
> has since been shuffled around to a different file, but it's the one
> where it originates from.
> 
> Fixes: f80a1d414328 ("interconnect: qcom: Add SDM660 interconnect provider driver")
> Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
> ---
Would be very nice if somebody could test this one in particular
on QCM2290 to make sure it does not regress that SoC..

Shawn, Loic?

Konrad
>  drivers/interconnect/qcom/icc-rpm.c | 20 ++++++++++++++++----
>  1 file changed, 16 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
> index 385a67c20956..b1337f90c975 100644
> --- a/drivers/interconnect/qcom/icc-rpm.c
> +++ b/drivers/interconnect/qcom/icc-rpm.c
> @@ -239,15 +239,27 @@ static int qcom_icc_rpm_set(int mas_rpm_id, int slv_rpm_id, u64 sum_bw)
>  static int __qcom_icc_set(struct icc_node *n, struct qcom_icc_node *qn,
>  			  u64 sum_bw)
>  {
> +	struct qcom_icc_provider *qp = to_qcom_provider(n->provider);
> +	bool vote_ap, vote_rpm;
>  	int ret;
>  
> -	if (!qn->qos.ap_owned) {
> -		/* send bandwidth request message to the RPM processor */
> +	if (qp->type == QCOM_ICC_QNOC) {
> +		vote_ap = true;
> +		vote_rpm = true;
> +	} else {
> +		vote_ap = qn->qos.ap_owned;
> +		vote_rpm = !vote_ap;
> +	}
> +
> +	if (vote_rpm) {
> +		/* Send bandwidth request message to the RPM processor */
>  		ret = qcom_icc_rpm_set(qn->mas_rpm_id, qn->slv_rpm_id, sum_bw);
>  		if (ret)
>  			return ret;
> -	} else if (qn->qos.qos_mode != -1) {
> -		/* set bandwidth directly from the AP */
> +	}
> +
> +	if (vote_ap && qn->qos.qos_mode != NOC_QOS_MODE_INVALID) {
> +		/* Set QoS params from the AP */
>  		ret = qcom_icc_qos_set(n, sum_bw);
>  		if (ret)
>  			return ret;

Re: [PATCH v3 2/9] interconnect: qcom: rpm: Always set QoS params on QNoC

[Bryan O'Donoghue]

On 18/01/2023 20:36, Konrad Dybcio wrote:
> Would be very nice if somebody could test this one in particular
> on QCM2290 to make sure it does not regress that SoC..
> 
> Shawn, Loic?
> 
> Konrad

I have a 2290.

Since I'm already helping with review, I'm happy to help with test too.

---
bod

Re: [PATCH v3 9/9] interconnect: qcom: rpm: Don't use clk_get_optional for bus clocks anymore

[Arnaud Vrac]

Hi Konrad,

With this series on a db820c, I get an error on boot when probing
a0noc because no "bus" clock is defined. If I revert this patch it
works again.

-Arnaud

Le lun. 16 janv. 2023 à 14:31, Konrad Dybcio
<konrad.dybcio@linaro.org> a écrit :
>
> Commit dd42ec8ea5b9 ("interconnect: qcom: rpm: Use _optional func for provider clocks")
> relaxed the requirements around probing bus clocks. This was a decent
> solution for making sure MSM8996 would still boot with old DTs, but
> now that there's a proper fix in place that both old and new DTs
> will be happy about, revert back to the safer variant of the
> function.
>
> Fixes: dd42ec8ea5b9 ("interconnect: qcom: rpm: Use _optional func for provider clocks")
> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
> ---
>  drivers/interconnect/qcom/icc-rpm.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
> index 343e6021a93a..8cff724661f2 100644
> --- a/drivers/interconnect/qcom/icc-rpm.c
> +++ b/drivers/interconnect/qcom/icc-rpm.c
> @@ -522,7 +522,7 @@ int qnoc_probe(struct platform_device *pdev)
>         }
>
>  regmap_done:
> -       ret = devm_clk_bulk_get_optional(dev, qp->num_bus_clks, qp->bus_clks);
> +       ret = devm_clk_bulk_get(dev, qp->num_bus_clks, qp->bus_clks);
>         if (ret)
>                 return ret;
>
> --
> 2.39.0
>

Re: [PATCH v3 9/9] interconnect: qcom: rpm: Don't use clk_get_optional for bus clocks anymore

[Konrad Dybcio]

On 23.01.2023 19:41, Arnaud Vrac wrote:
> Hi Konrad,
> 
> With this series on a db820c, I get an error on boot when probing
> a0noc because no "bus" clock is defined. If I revert this patch it
> works again.
> 
> -Arnaud
Ouch! Thanks for testing and catching this, definitely an edge
case, as we usually expect there's an interconnect bus and bus_a
clock.. but that can be worked around!

Konrad
> 
> Le lun. 16 janv. 2023 à 14:31, Konrad Dybcio
> <konrad.dybcio@linaro.org> a écrit :
>>
>> Commit dd42ec8ea5b9 ("interconnect: qcom: rpm: Use _optional func for provider clocks")
>> relaxed the requirements around probing bus clocks. This was a decent
>> solution for making sure MSM8996 would still boot with old DTs, but
>> now that there's a proper fix in place that both old and new DTs
>> will be happy about, revert back to the safer variant of the
>> function.
>>
>> Fixes: dd42ec8ea5b9 ("interconnect: qcom: rpm: Use _optional func for provider clocks")
>> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
>> ---
>>  drivers/interconnect/qcom/icc-rpm.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/interconnect/qcom/icc-rpm.c b/drivers/interconnect/qcom/icc-rpm.c
>> index 343e6021a93a..8cff724661f2 100644
>> --- a/drivers/interconnect/qcom/icc-rpm.c
>> +++ b/drivers/interconnect/qcom/icc-rpm.c
>> @@ -522,7 +522,7 @@ int qnoc_probe(struct platform_device *pdev)
>>         }
>>
>>  regmap_done:
>> -       ret = devm_clk_bulk_get_optional(dev, qp->num_bus_clks, qp->bus_clks);
>> +       ret = devm_clk_bulk_get(dev, qp->num_bus_clks, qp->bus_clks);
>>         if (ret)
>>                 return ret;
>>
>> --
>> 2.39.0
>>

Re: [PATCH v3 2/9] interconnect: qcom: rpm: Always set QoS params on QNoC

[Konrad Dybcio]

On 19.01.2023 01:27, Bryan O'Donoghue wrote:
> On 18/01/2023 20:36, Konrad Dybcio wrote:
>> Would be very nice if somebody could test this one in particular
>> on QCM2290 to make sure it does not regress that SoC..
>>
>> Shawn, Loic?
>>
>> Konrad
> 
> I have a 2290.
> 
> Since I'm already helping with review, I'm happy to help with test too.
Thanks!

I'm afraid it won't be necessary though..

This patch gives more trouble than it solves. Here's the thing:

It was working perfectly on my SM6375 for quite some time and then
I rebased my branch on a newer -next and suddenly it makes things go
boom boom.. I think I'll drop it in the next revision, even if that's
what downstream does.. At least until I pretend to be an archaeologist
surgeon again and debug this..

Konrad

> 
> ---
> bod