Re: [PATCH v1] clk: qcom: clk-rpmh: Add IPA clock support

From: David Dai
Date: Fri Jan 11 2019 - 19:56:21 EST



On 1/9/2019 11:28 AM, Stephen Boyd wrote:
Quoting David Dai (2018-12-13 18:35:04)
The current clk-rpmh driver only supports on and off RPMh clock resources,
let's extend the current driver by add support for clocks that are managed
by a different type of RPMh resource known as Bus Clock Manager(BCM).
The BCM is a configurable shared resource aggregator that scales performance
based on a preset of frequency points. The Qualcomm IP Accelerator (IPA) clock
is an example of a resource that is managed by the BCM and this a requirement
from the IPA driver in order to scale its core clock.
Please use this commit text instead:

The clk-rpmh driver only supports on and off RPMh clock resources. Let's
extend the driver by add support for clocks that are managed by a
different type of RPMh resource known as Bus Clock Manager(BCM). The BCM
is a configurable shared resource aggregator that scales performance
based on a set of frequency points. The Qualcomm IP Accelerator (IPA)
clock is an example of a resource that is managed by the BCM and this a
requirement from the IPA driver in order to scale its core clock.
Ok.
Signed-off-by: David Dai <daidavid1@xxxxxxxxxxxxxx>
---
drivers/clk/qcom/clk-rpmh.c | 141 ++++++++++++++++++++++++++++++++++
include/dt-bindings/clock/qcom,rpmh.h | 1 +
2 files changed, 142 insertions(+)

diff --git a/drivers/clk/qcom/clk-rpmh.c b/drivers/clk/qcom/clk-rpmh.c
index 9f4fc77..ee78c4b 100644
--- a/drivers/clk/qcom/clk-rpmh.c
+++ b/drivers/clk/qcom/clk-rpmh.c
@@ -18,6 +18,31 @@
#define CLK_RPMH_ARC_EN_OFFSET 0
#define CLK_RPMH_VRM_EN_OFFSET 4
+#define BCM_TCS_CMD_COMMIT_MASK 0x40000000
+#define BCM_TCS_CMD_VALID_SHIFT 29
+#define BCM_TCS_CMD_VOTE_MASK 0x3fff
+#define BCM_TCS_CMD_VOTE_SHIFT 0
+
+#define BCM_TCS_CMD(valid, vote) \
+ (BCM_TCS_CMD_COMMIT_MASK | \
+ ((valid) << BCM_TCS_CMD_VALID_SHIFT) | \
+ ((cpu_to_le32(vote) & \
Why?
Sorry, could you clarify this question? If you're referring to the cpu_to_le32, shouldn't we be explicit about converting endianness even if it might be redundant for this particular qcom platform?

+ BCM_TCS_CMD_VOTE_MASK) << BCM_TCS_CMD_VOTE_SHIFT))
+
+/**
+ * struct bcm_db - Auxiliary data pertaining to each Bus Clock Manager(BCM)
+ * @unit: divisor used to convert Hz value to an RPMh msg
+ * @width: multiplier used to convert Hz value to an RPMh msg
+ * @vcd: virtual clock domain that this bcm belongs to
+ * @reserved: reserved to pad the struct
+ */
+struct bcm_db {
+ __le32 unit;
+ __le16 width;
+ u8 vcd;
+ u8 reserved;
+};
+
/**
* struct clk_rpmh - individual rpmh clock data structure
* @hw: handle between common and hardware-specific interfaces
@@ -29,6 +54,7 @@
* @aggr_state: rpmh clock aggregated state
* @last_sent_aggr_state: rpmh clock last aggr state sent to RPMh
* @valid_state_mask: mask to determine the state of the rpmh clock
+ * @aux_data: data specific to the bcm rpmh resource
But there isn't aux_data. Is this supposed to be unit? And what is unit
going to be? 1000?
Supposed to be unit, the value is on the order of Khz.
* @dev: device to which it is attached
* @peer: pointer to the clock rpmh sibling
*/
@@ -42,6 +68,7 @@ struct clk_rpmh {
u32 aggr_state;
u32 last_sent_aggr_state;
u32 valid_state_mask;
+ u32 unit;
struct device *dev;
struct clk_rpmh *peer;
};
@@ -210,6 +248,91 @@ static unsigned long clk_rpmh_recalc_rate(struct clk_hw *hw,
.recalc_rate = clk_rpmh_recalc_rate,
};
+static int clk_rpmh_bcm_send_cmd(struct clk_rpmh *c, bool enable)
+{
+ struct tcs_cmd cmd = { 0 };
+ u32 cmd_state;
+ int ret;
+
+ mutex_lock(&rpmh_clk_lock);
+
+ cmd_state = enable ? (c->aggr_state ? c->aggr_state : 1) : 0;
Make this into an if statement instead of double ternary?
Ok.
cmd_state = 0;
if (enable) {
cmd_state = 1;
if (c->aggr_state)
cmd_state = c->aggr_state;

}

+
+ if (c->last_sent_aggr_state == cmd_state)
+ return 0;
We just returned with a mutex held. Oops!
Oops, will fix.
+
+ cmd.addr = c->res_addr;
+ cmd.data = BCM_TCS_CMD(enable, cmd_state);
+
+ ret = rpmh_write_async(c->dev, RPMH_ACTIVE_ONLY_STATE, &cmd, 1);
+ if (ret) {
+ dev_err(c->dev, "set active state of %s failed: (%d)\n",
+ c->res_name, ret);
+ return ret;
Again!
+ }
+
+ c->last_sent_aggr_state = cmd_state;
+
+ mutex_unlock(&rpmh_clk_lock);
+
+ return 0;
+}
+
+static int clk_rpmh_bcm_prepare(struct clk_hw *hw)
+{
+ struct clk_rpmh *c = to_clk_rpmh(hw);
+ int ret;
+
+ ret = clk_rpmh_bcm_send_cmd(c, true);
+
+ return ret;
Just write return clk_rpmh_bcm_send_cmd(...)
Ok.

+};
+
+static void clk_rpmh_bcm_unprepare(struct clk_hw *hw)
+{
+ struct clk_rpmh *c = to_clk_rpmh(hw);
+
+ clk_rpmh_bcm_send_cmd(c, false);
+};
+
+static int clk_rpmh_bcm_set_rate(struct clk_hw *hw, unsigned long rate,
+ unsigned long parent_rate)
+{
+ struct clk_rpmh *c = to_clk_rpmh(hw);
+
+ c->aggr_state = rate / c->unit;
+ /*
+ * Since any non-zero value sent to hw would result in enabling the
+ * clock, only send the value if the clock has already been prepared.
+ */
+ if (clk_hw_is_prepared(hw))
This is sad, but OK.

+ clk_rpmh_bcm_send_cmd(c, true);
+
+ return 0;
+};
+
[...]
@@ -217,6 +340,7 @@ static unsigned long clk_rpmh_recalc_rate(struct clk_hw *hw,
DEFINE_CLK_RPMH_VRM(sdm845, rf_clk1, rf_clk1_ao, "rfclka1", 1);
DEFINE_CLK_RPMH_VRM(sdm845, rf_clk2, rf_clk2_ao, "rfclka2", 1);
DEFINE_CLK_RPMH_VRM(sdm845, rf_clk3, rf_clk3_ao, "rfclka3", 1);
+DEFINE_CLK_RPMH_BCM(sdm845, ipa, "IP0");
It's really IP0?! What was wrong with IPA?
Yeah... convention seems to be 2 letters and then a number for most BCM resources.
static struct clk_hw *sdm845_rpmh_clocks[] = {
[RPMH_CXO_CLK] = &sdm845_bi_tcxo.hw,
@@ -275,6 +402,20 @@ static int clk_rpmh_probe(struct platform_device *pdev)
rpmh_clk->res_name);
return -ENODEV;
}
+
+ data = cmd_db_read_aux_data(rpmh_clk->res_name, &aux_data_len);
+ if (IS_ERR(data)) {
+ ret = PTR_ERR(data);
+ dev_err(&pdev->dev,
+ "error reading RPMh aux data for %s (%d)\n",
+ rpmh_clk->res_name, ret);
+ return ret;
Weird indent here.
will fix.

--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project