Skip to contentSkip to conversation

mmc: sdhci-msm: Add clk-rates, DDR, HS400 support

PATCH Superseded Ritesh Harjani 7 participants 32 messages Nov 14, 2016 – Nov 16, 2016 4 messages not in the archive

Subsystems (7)

v7 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v7 cover mmc: sdhci-msm: Add clk-rates, DDR, HS400 support

Hi,

This is v7 version of the patch series which adds support for MSM8996. Adds HS400 driver support as well.
These are tested on internal msm8996 & db410c HW.

Below are the changes in v7.

Changes from v6 -> v7 :-
1. Removed patch "clk: Add clk_hw_get_clk() helper API to be used by clk providers" in v7 as it was not required.
2. Addressed Stephen review comments on -
"clk: qcom: Add rcg ops to return floor value closest to the requested rate"
3. Addressed comments from Stephen to add xo_clock entry in the sdhc clock node. Using the same xo_clock entry from DT to get the clk_rate of xo_clock used in sdhci-msm driver. Patch 04 adds this entry into DT.
Patch 05 adds the driver support for xo_clock mentioned above.
Hence there is a minor change in Patch05, which can be reviewed and taken into the tree.

IMHO, almost all patches are almost done and are ready to be accepted. Will below process work out?
Patches 001 & 002 :- (clock changes) - Can go via Stephen's Boyd Tree. Patches 004 & 010 :- (DTS changes) - Can go via Andy Gross. Patches 003, 005-009 & 011-014 :- (sdhci-msm changes) - Adrian's tree.

Please let me know in case if anything else is required on above.

Changes from v5 -> v6 :-
1. Earlier in v5 series DT node was added to get the clk-rates table needed for sdhci-msm driver. But this is removed in this(v6) patch series and instead the clk changes are done in the clk driver as per Rob H comment.

2. Added clk driver changes(patch 1-3) to provide floor rate values of requested clock for sdhc client.
For following boards- apq8084, msm8996, msm8916, msm8974.

3. Other minor patch comments were addressed.

Changes from v4 -> v5 :-
1. Added HS400 sdhci-msm controller specific changes:- (Patch 10, 11, 12)
2. Addressed comment from Adrian on Patch 07 @[3].
3. Addressed comment from Arnd on Patch 03, to directly add clk_table into sdhci_msm_host. [4]
4. Addressed comment from Bjorn to not enforce having clk-rates property in DT for older targets based on discussion at [5]
5. Retained Acks from Adrian on patches (01 & 02 & 06) where there were no changes made while addressing above comments.

Older history:-
This is v4 version of the patch series.
Patches 01, 02, 05 & 06 were Acked-by Adrian.

Changes from v3 -> v4 :-
1. Addressed comments from Adrian on Patch 03, 07, 08.
2. Addressed comments from Bjorn on Patch 03.
3. Added clk-rate support for sdhc DT nodes to all MSM platforms. in Pacth 04.
4. Rebased on next branch of Ulf.

Changes from v2 -> v3 :-
1. Addded Patch 01 based on Bjorn comment[2] -
This fixes/unrolls the poor coding style of read/writes of registers from base sdhci-msm driver.

2. Fixed/unrolled poor style of reads/writes of registers in Patch 02, based on Bjorn comment[2]. Also changed name of flag from use_updated_dll_reset -> use_14lpp_dll_reset.

Changes from v1->v2 :-
1. Removed patch 06 & 08 from v1 patch series[1] (which were introducing unnecessary quirks). Instead have implemented __sdhci_msm_set_clock version of sdhci_set_clock in sdhci_msm driver itself in patch 07 of this patch series.
2. Enabled extra quirk (SDHCI_QUIRK2_PRESET_VALUE_BROKEN) in patch 05 of this patch series.

Description of patches :-
This patchset adds clk-rates & other required changes to upstream sdhci-msm driver from codeaurora tree.
It has been tested on a db410c Dragonboard and msm8996 based platform.

Patch 0001-0003- Adds support in qcom clk driver to return floor value of requested clock rate instead of ceil rate for sdhc clients.

Patch 0004- Adds updated dll sequence for newer controllers which has minor_version >= 0x42. This is required for msm8996.

MSM controller HW recommendation is to use the base MCI clock and directly control this MCI clock at GCC in order to change the clk-rate.
Patches 06-08 bring in required change for this to sdhci-msm.

MSM controller would require 2x clock rate from source for DDR bus speed modes. Patch 09 adds this support.

Patch 0010- adds DDR support in DT for sdhc1 of msm8916.

Patches 0011-0014- Adds HS400 support to sdhci-msm.

[1]:- http://www.spinics.net/lists/linux-mmc/msg38467.html
[2]:- http://www.spinics.net/lists/linux-mmc/msg38578.html
[3]:- https://patchwork.kernel.org/patch/9289345/
[4]:- https://www.spinics.net/lists/linux-mmc/msg39107.html
[5]:- http://www.spinics.net/lists/linux-mmc/msg38749.html
[6]:- https://patchwork.kernel.org/patch/9297381/

Rajendra Nayak (2):
clk: qcom: Add rcg ops to return floor value closest to the requested
rate
clk: qcom: Move all sdcc rcgs to use clk_rcg2_floor_ops

Ritesh Harjani (10):
mmc: sdhci-msm: Change poor style writel/readl of registers
ARM: dts: Add xo_clock to sdhc nodes on qcom platforms
mmc: sdhci-msm: Add get_min_clock() and get_max_clock() callback
mmc: sdhci-msm: Enable few quirks
mmc: sdhci-msm: Implement set_clock callback for sdhci-msm
mmc: sdhci-msm: Add clock changes for DDR mode.
arm64: dts: qcom: msm8916: Add ddr support to sdhc1
mmc: sdhci-msm: Save the calculated tuning phase
mmc: sdhci-msm: Add calibration tuning for CDCLP533 circuit
sdhci: sdhci-msm: update dll configuration

Venkat Gopalakrishnan (2):
mmc: sdhci-msm: Update DLL reset sequence
mmc: sdhci-msm: Add HS400 platform support

 arch/arm/boot/dts/qcom-apq8084.dtsi   |  14 +-
 arch/arm/boot/dts/qcom-msm8974.dtsi   |  14 +-
 arch/arm64/boot/dts/qcom/msm8916.dtsi |  11 +-
 arch/arm64/boot/dts/qcom/msm8996.dtsi |   9 +-
 drivers/clk/qcom/clk-rcg.h            |   1 +
 drivers/clk/qcom/clk-rcg2.c           |  76 +++-
 drivers/clk/qcom/common.c             |  16 +
 drivers/clk/qcom/common.h             |   2 +
 drivers/clk/qcom/gcc-apq8084.c        |   8 +-
 drivers/clk/qcom/gcc-msm8916.c        |   4 +-
 drivers/clk/qcom/gcc-msm8974.c        |   8 +-
 drivers/clk/qcom/gcc-msm8996.c        |   8 +-
 drivers/mmc/host/sdhci-msm.c          | 664 ++++++++++++++++++++++++++++++++--
 13 files changed, 758 insertions(+), 77 deletions(-)
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.

A message that is not in the archive (sent off-list or not yet received): 1479103248-9491-1-git-send-email-riteshh-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org

#2 lore raw

PATCH v7 1/14 clk: qcom: Add rcg ops to return floor value closest to the requested rate

From: Rajendra Nayak <[email protected]>

The default behaviour with clk_rcg2_ops is for the clk_round_rate()/clk_set_rate() to return/set a ceil clock rate closest to the requested rate by looking up the corresponding frequency table.
However, we do have some instances (mainly sdcc on various platforms) of clients expecting a clk_set_rate() to set a floor value instead. Add a new clk_rcg2_floor_ops to handle this for such specific rcg instances

---
 drivers/clk/qcom/clk-rcg.h  |  1 +
 drivers/clk/qcom/clk-rcg2.c | 76 +++++++++++++++++++++++++++++++++++++++------
 drivers/clk/qcom/common.c   | 16 ++++++++++
 drivers/clk/qcom/common.h   |  2 ++
 4 files changed, 86 insertions(+), 9 deletions(-)
drivers/clk/qcom/clk-rcg.h +1 -0
index b904c33..1b3e8d2 100644
@@ -173,6 +173,7 @@ struct clk_rcg2 {
 #define to_clk_rcg2(_hw) container_of(to_clk_regmap(_hw), struct clk_rcg2, clkr)
 
 extern const struct clk_ops clk_rcg2_ops;
+extern const struct clk_ops clk_rcg2_floor_ops;
 extern const struct clk_ops clk_rcg2_shared_ops;
 extern const struct clk_ops clk_edp_pixel_ops;
 extern const struct clk_ops clk_byte_ops;
drivers/clk/qcom/clk-rcg2.c +67 -9

133 lines not on this page · raw message

drivers/clk/qcom/common.c +16 -0
index fffcbaf..d56dd14 100644
@@ -46,6 +46,22 @@ struct freq_tbl *qcom_find_freq(const struct freq_tbl *f, unsigned long rate)
 }
 EXPORT_SYMBOL_GPL(qcom_find_freq);
 
+const struct freq_tbl *qcom_find_freq_floor(const struct freq_tbl *f,
+				      unsigned long rate)
+{
+	const struct freq_tbl *best = NULL;
+
+	for ( ; f->freq; f++) {
+		if (rate >= f->freq)
+			best = f;
+		else
+			break;
+	}
+
+	return best;
+}
+EXPORT_SYMBOL_GPL(qcom_find_freq_floor);
+
 int qcom_find_src_index(struct clk_hw *hw, const struct parent_map *map, u8 src)
 {
 	int i, num_parents = clk_hw_get_num_parents(hw);
drivers/clk/qcom/common.h +2 -0
index ae9bdeb..76886a1 100644
@@ -34,6 +34,8 @@ struct qcom_cc_desc {
 
 extern const struct freq_tbl *qcom_find_freq(const struct freq_tbl *f,
 					     unsigned long rate);
+extern const struct freq_tbl *qcom_find_freq_floor(const struct freq_tbl *f,
+						   unsigned long rate);
 extern int qcom_find_src_index(struct clk_hw *hw, const struct parent_map *map,
 			       u8 src);
 
Signature · 7 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#3 lore raw

PATCH v7 7/14 mmc: sdhci-msm: Enable few quirks

In reply to a message not in the archive

sdhc-msm controller needs this SDHCI_QUIRK_CAP_CLOCK_BASE_BROKEN & SDHCI_QUIRK2_PRESET_VALUE_BROKEN to be set. Hence setting it.

---
 drivers/mmc/host/sdhci-msm.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-msm.c +3 -1
index f63ffb0..268424c 100644
@@ -598,7 +598,9 @@ static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host)
 static const struct sdhci_pltfm_data sdhci_msm_pdata = {
 	.quirks = SDHCI_QUIRK_BROKEN_CARD_DETECTION |
 		  SDHCI_QUIRK_NO_CARD_NO_RESET |
-		  SDHCI_QUIRK_SINGLE_POWER_WRITE,
+		  SDHCI_QUIRK_SINGLE_POWER_WRITE |
+		  SDHCI_QUIRK_CAP_CLOCK_BASE_BROKEN,
+	.quirks2 = SDHCI_QUIRK2_PRESET_VALUE_BROKEN,
 	.ops = &sdhci_msm_ops,
 };
 
Signature · 7 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#4 lore raw

PATCH v7 9/14 mmc: sdhci-msm: Add clock changes for DDR mode.

In reply to a message not in the archive

SDHC MSM controller need 2x clock for MCLK at GCC. Hence make required changes to have 2x clock for DDR timing modes.

---
 drivers/mmc/host/sdhci-msm.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)
drivers/mmc/host/sdhci-msm.c +15 -2
index b96a4a7..41a4ea7 100644
@@ -637,6 +637,7 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 {
 	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
 	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+	struct mmc_ios curr_ios = host->mmc->ios;
 	int rc;
 
 	if (!clock) {
@@ -645,11 +646,23 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 	}
 
 	spin_unlock_irq(&host->lock);
+	/*
+	 * The SDHC requires internal clock frequency to be double the
+	 * actual clock that will be set for DDR mode. The controller
+	 * uses the faster clock(100/400MHz) for some of its parts and
+	 * send the actual required clock (50/200MHz) to the card.
+	 */
+	if ((curr_ios.timing == MMC_TIMING_UHS_DDR50) ||
+	    (curr_ios.timing == MMC_TIMING_MMC_DDR52) ||
+	    (curr_ios.timing == MMC_TIMING_MMC_HS400))
+		clock *= 2;
+
 	if (clock != msm_host->clk_rate) {
 		rc = clk_set_rate(msm_host->clk, clock);
 		if (rc) {
-			pr_err("%s: Failed to set clock at rate %u\n",
-			       mmc_hostname(host->mmc), clock);
+			pr_err("%s: Failed to set clock at rate %u at timing %d\n",
+			       mmc_hostname(host->mmc), clock,
+			       curr_ios.timing);
 			spin_lock_irq(&host->lock);
 			goto out;
 		}
Signature · 7 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#5 lore raw

PATCH v7 10/14 arm64: dts: qcom: msm8916: Add ddr support to sdhc1

In reply to a message not in the archive

This adds mmc-ddr-1_8v support to DT for sdhc1 of msm8916.

---
 arch/arm64/boot/dts/qcom/msm8916.dtsi | 1 +
 1 file changed, 1 insertion(+)
arch/arm64/boot/dts/qcom/msm8916.dtsi +1 -0
index 6265822..5bc187c 100644
@@ -515,6 +515,7 @@
 				 <&gcc GCC_SDCC1_AHB_CLK>,
 				 <&xo_board 0>;
 			clock-names = "core", "iface", "xo_clock";
+			mmc-ddr-1_8v;
 			bus-width = <8>;
 			non-removable;
 			status = "disabled";
Signature · 7 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#6 lore raw

PATCH v7 2/14 clk: qcom: Move all sdcc rcgs to use clk_rcg2_floor_ops

In reply to Ritesh Harjani

From: Rajendra Nayak <[email protected]>

The sdcc driver for msm8996/msm8916/msm8974 and apq8084 expects a clk_set_rate() on the sdcc rcg clk to set a floor value of supported clk rate closest to the requested rate, by looking up the frequency table.
So move all the sdcc rcgs on all these platforms to use the newly introduced clk_rcg2_floor_ops

---
 drivers/clk/qcom/gcc-apq8084.c | 8 ++++----
 drivers/clk/qcom/gcc-msm8916.c | 4 ++--
 drivers/clk/qcom/gcc-msm8974.c | 8 ++++----
 drivers/clk/qcom/gcc-msm8996.c | 8 ++++----
 4 files changed, 14 insertions(+), 14 deletions(-)
drivers/clk/qcom/gcc-apq8084.c +4 -4
index 070037a..486d961 100644
@@ -1142,7 +1142,7 @@ enum {
 		.name = "sdcc1_apps_clk_src",
 		.parent_names = gcc_xo_gpll0_gpll4,
 		.num_parents = 3,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -1156,7 +1156,7 @@ enum {
 		.name = "sdcc2_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -1170,7 +1170,7 @@ enum {
 		.name = "sdcc3_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -1184,7 +1184,7 @@ enum {
 		.name = "sdcc4_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
drivers/clk/qcom/gcc-msm8916.c +2 -2
index 5c4e193..628e6ca 100644
@@ -1107,7 +1107,7 @@ enum {
 		.name = "sdcc1_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -1132,7 +1132,7 @@ enum {
 		.name = "sdcc2_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
drivers/clk/qcom/gcc-msm8974.c +4 -4
index 0091520..348e30d 100644
@@ -872,7 +872,7 @@ enum {
 	.name = "sdcc1_apps_clk_src",
 	.parent_names = gcc_xo_gpll0,
 	.num_parents = 2,
-	.ops = &clk_rcg2_ops,
+	.ops = &clk_rcg2_floor_ops,
 };
 
 static struct clk_rcg2 sdcc1_apps_clk_src = {
@@ -894,7 +894,7 @@ enum {
 		.name = "sdcc2_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -908,7 +908,7 @@ enum {
 		.name = "sdcc3_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -922,7 +922,7 @@ enum {
 		.name = "sdcc4_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
drivers/clk/qcom/gcc-msm8996.c +4 -4
index fe03e6f..5adb163 100644
@@ -460,7 +460,7 @@ enum {
 		.name = "sdcc1_apps_clk_src",
 		.parent_names = gcc_xo_gpll0_gpll4_gpll0_early_div,
 		.num_parents = 4,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -497,7 +497,7 @@ enum {
 		.name = "sdcc2_apps_clk_src",
 		.parent_names = gcc_xo_gpll0_gpll4,
 		.num_parents = 3,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -511,7 +511,7 @@ enum {
 		.name = "sdcc3_apps_clk_src",
 		.parent_names = gcc_xo_gpll0_gpll4,
 		.num_parents = 3,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
@@ -535,7 +535,7 @@ enum {
 		.name = "sdcc4_apps_clk_src",
 		.parent_names = gcc_xo_gpll0,
 		.num_parents = 2,
-		.ops = &clk_rcg2_ops,
+		.ops = &clk_rcg2_floor_ops,
 	},
 };
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#7 lore raw

Re: [PATCH v7 02/14] clk: qcom: Move all sdcc rcgs to use clk_rcg2_floor_ops

Quoted 9 lines · Ritesh Harjani

From: Rajendra Nayak <[email protected]>

The sdcc driver for msm8996/msm8916/msm8974 and apq8084 expects a clk_set_rate() on the sdcc rcg clk to set a floor value of supported clk rate closest to the requested rate, by looking up the frequency table.
So move all the sdcc rcgs on all these platforms to use the newly introduced clk_rcg2_floor_ops

Quoted 7 lines
---
 drivers/clk/qcom/gcc-apq8084.c | 8 ++++----
 drivers/clk/qcom/gcc-msm8916.c | 4 ++--
 drivers/clk/qcom/gcc-msm8974.c | 8 ++++----
 drivers/clk/qcom/gcc-msm8996.c | 8 ++++----

Can you please include the same changes for gcc-msm8994.c as it should be in sboyd's tree?

If not I will send a patch directly, either way works for me.

-jeremy

#8 lore raw

PATCH v7 3/14 mmc: sdhci-msm: Change poor style writel/readl of registers

In reply to Ritesh Harjani

This patch changes the poor style of writel/readl registers into more readable format. Also to avoid mixed style format of readl/writel in sdhci-msm driver.

---
 drivers/mmc/host/sdhci-msm.c | 54 ++++++++++++++++++++++++++------------------
 1 file changed, 32 insertions(+), 22 deletions(-)
drivers/mmc/host/sdhci-msm.c +32 -22
index 90ed2e1..6ea89c2 100644
@@ -137,8 +137,9 @@ static int msm_config_cm_dll_phase(struct sdhci_host *host, u8 phase)
 	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
 	/* Set CK_OUT_EN bit of DLL_CONFIG register to 1. */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_DLL_CONFIG)
-			| CORE_CK_OUT_EN), host->ioaddr + CORE_DLL_CONFIG);
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config |= CORE_CK_OUT_EN;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
 	/* Wait until CK_OUT_EN bit of DLL_CONFIG register becomes '1' */
 	rc = msm_dll_poll_ck_out_en(host, 1);
@@ -305,6 +306,7 @@ static int msm_init_cm_dll(struct sdhci_host *host)
 	struct mmc_host *mmc = host->mmc;
 	int wait_cnt = 50;
 	unsigned long flags;
+	u32 config;
 
 	spin_lock_irqsave(&host->lock, flags);
 
@@ -313,33 +315,40 @@ static int msm_init_cm_dll(struct sdhci_host *host)
 	 * tuning is in progress. Keeping PWRSAVE ON may
 	 * turn off the clock.
 	 */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC)
-			& ~CORE_CLK_PWRSAVE), host->ioaddr + CORE_VENDOR_SPEC);
+	config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
+	config &= ~CORE_CLK_PWRSAVE;
+	writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
 
 	/* Write 1 to DLL_RST bit of DLL_CONFIG register */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_DLL_CONFIG)
-			| CORE_DLL_RST), host->ioaddr + CORE_DLL_CONFIG);
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config |= CORE_DLL_RST;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
 	/* Write 1 to DLL_PDN bit of DLL_CONFIG register */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_DLL_CONFIG)
-			| CORE_DLL_PDN), host->ioaddr + CORE_DLL_CONFIG);
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config |= CORE_DLL_PDN;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 	msm_cm_dll_set_freq(host);
 
 	/* Write 0 to DLL_RST bit of DLL_CONFIG register */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_DLL_CONFIG)
-			& ~CORE_DLL_RST), host->ioaddr + CORE_DLL_CONFIG);
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config &= ~CORE_DLL_RST;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
 	/* Write 0 to DLL_PDN bit of DLL_CONFIG register */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_DLL_CONFIG)
-			& ~CORE_DLL_PDN), host->ioaddr + CORE_DLL_CONFIG);
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config &= ~CORE_DLL_PDN;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
 	/* Set DLL_EN bit to 1. */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_DLL_CONFIG)
-			| CORE_DLL_EN), host->ioaddr + CORE_DLL_CONFIG);
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config |= CORE_DLL_EN;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
 	/* Set CK_OUT_EN bit to 1. */
-	writel_relaxed((readl_relaxed(host->ioaddr + CORE_DLL_CONFIG)
-			| CORE_CK_OUT_EN), host->ioaddr + CORE_DLL_CONFIG);
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config |= CORE_CK_OUT_EN;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
 	/* Wait until DLL_LOCK bit of DLL_STATUS register becomes '1' */
 	while (!(readl_relaxed(host->ioaddr + CORE_DLL_STATUS) &
@@ -536,7 +545,7 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 	struct resource *core_memres;
 	int ret;
 	u16 host_version, core_minor;
-	u32 core_version, caps;
+	u32 core_version, config;
 	u8 core_major;
 
 	host = sdhci_pltfm_init(pdev, &sdhci_msm_pdata, sizeof(*msm_host));
@@ -605,8 +614,9 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 	}
 
 	/* Reset the core and Enable SDHC mode */
-	writel_relaxed(readl_relaxed(msm_host->core_mem + CORE_POWER) |
-		       CORE_SW_RST, msm_host->core_mem + CORE_POWER);
+	config = readl_relaxed(msm_host->core_mem + CORE_POWER);
+	config |= CORE_SW_RST;
+	writel_relaxed(config, msm_host->core_mem + CORE_POWER);
 
 	/* SW reset can take upto 10HCLK + 15MCLK cycles. (min 40us) */
 	usleep_range(1000, 5000);
@@ -636,9 +646,9 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 	 * controller versions and must be explicitly enabled.
 	 */
 	if (core_major >= 1 && core_minor != 0x11 && core_minor != 0x12) {
-		caps = readl_relaxed(host->ioaddr + SDHCI_CAPABILITIES);
-		caps |= SDHCI_CAN_VDD_300 | SDHCI_CAN_DO_8BIT;
-		writel_relaxed(caps, host->ioaddr +
+		config = readl_relaxed(host->ioaddr + SDHCI_CAPABILITIES);
+		config |= SDHCI_CAN_VDD_300 | SDHCI_CAN_DO_8BIT;
+		writel_relaxed(config, host->ioaddr +
 			       CORE_VENDOR_SPEC_CAPABILITIES0);
 	}
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#9 lore raw

PATCH v7 4/14 ARM: dts: Add xo_clock to sdhc nodes on qcom platforms

In reply to Ritesh Harjani

Add xo_clock to sdhc clock node on all qcom platforms.

---
 arch/arm/boot/dts/qcom-apq8084.dtsi   | 14 ++++++++------
 arch/arm/boot/dts/qcom-msm8974.dtsi   | 14 ++++++++------
 arch/arm64/boot/dts/qcom/msm8916.dtsi | 10 ++++++----
 arch/arm64/boot/dts/qcom/msm8996.dtsi |  9 +++++----
 4 files changed, 27 insertions(+), 20 deletions(-)
arch/arm/boot/dts/qcom-apq8084.dtsi +8 -6
index 39eb7a4..d421ef3 100644
@@ -182,13 +182,13 @@
 	};
 
 	clocks {
-		xo_board {
+		xo_board: xo_board {
 			compatible = "fixed-clock";
 			#clock-cells = <0>;
 			clock-frequency = <19200000>;
 		};
 
-		sleep_clk {
+		sleep_clk: sleep_clk {
 			compatible = "fixed-clock";
 			#clock-cells = <0>;
 			clock-frequency = <32768>;
@@ -416,8 +416,9 @@
 			reg-names = "hc_mem", "core_mem";
 			interrupts = <0 123 0>, <0 138 0>;
 			interrupt-names = "hc_irq", "pwr_irq";
-			clocks = <&gcc GCC_SDCC1_APPS_CLK>, <&gcc GCC_SDCC1_AHB_CLK>;
-			clock-names = "core", "iface";
+			clocks = <&gcc GCC_SDCC1_APPS_CLK>, <&gcc GCC_SDCC1_AHB_CLK>,
+				 <&xo_board 0>;
+			clock-names = "core", "iface", "xo_clock";
 			status = "disabled";
 		};
 
@@ -427,8 +428,9 @@
 			reg-names = "hc_mem", "core_mem";
 			interrupts = <0 125 0>, <0 221 0>;
 			interrupt-names = "hc_irq", "pwr_irq";
-			clocks = <&gcc GCC_SDCC2_APPS_CLK>, <&gcc GCC_SDCC2_AHB_CLK>;
-			clock-names = "core", "iface";
+			clocks = <&gcc GCC_SDCC2_APPS_CLK>, <&gcc GCC_SDCC2_AHB_CLK>,
+				 <&xo_board 0>;
+			clock-names = "core", "iface", "xo_clock";
 			status = "disabled";
 		};
 
arch/arm/boot/dts/qcom-msm8974.dtsi +8 -6
index d210947..bec0ebc 100644
@@ -220,13 +220,13 @@
 	};
 
 	clocks {
-		xo_board {
+		xo_board: xo_board {
 			compatible = "fixed-clock";
 			#clock-cells = <0>;
 			clock-frequency = <19200000>;
 		};
 
-		sleep_clk {
+		sleep_clk: sleep_clk {
 			compatible = "fixed-clock";
 			#clock-cells = <0>;
 			clock-frequency = <32768>;
@@ -558,8 +558,9 @@
 			reg-names = "hc_mem", "core_mem";
 			interrupts = <0 123 0>, <0 138 0>;
 			interrupt-names = "hc_irq", "pwr_irq";
-			clocks = <&gcc GCC_SDCC1_APPS_CLK>, <&gcc GCC_SDCC1_AHB_CLK>;
-			clock-names = "core", "iface";
+			clocks = <&gcc GCC_SDCC1_APPS_CLK>, <&gcc GCC_SDCC1_AHB_CLK>,
+				 <&xo_board 0>;
+			clock-names = "core", "iface", "xo_clock";
 			status = "disabled";
 		};
 
@@ -569,8 +570,9 @@
 			reg-names = "hc_mem", "core_mem";
 			interrupts = <0 125 0>, <0 221 0>;
 			interrupt-names = "hc_irq", "pwr_irq";
-			clocks = <&gcc GCC_SDCC2_APPS_CLK>, <&gcc GCC_SDCC2_AHB_CLK>;
-			clock-names = "core", "iface";
+			clocks = <&gcc GCC_SDCC2_APPS_CLK>, <&gcc GCC_SDCC2_AHB_CLK>,
+				 <&xo_board 0>;
+			clock-names = "core", "iface", "xo_clock";
 			status = "disabled";
 		};
 
arch/arm64/boot/dts/qcom/msm8916.dtsi +6 -4
index 466ca57..6265822 100644
@@ -512,8 +512,9 @@
 			interrupts = <0 123 0>, <0 138 0>;
 			interrupt-names = "hc_irq", "pwr_irq";
 			clocks = <&gcc GCC_SDCC1_APPS_CLK>,
-				 <&gcc GCC_SDCC1_AHB_CLK>;
-			clock-names = "core", "iface";
+				 <&gcc GCC_SDCC1_AHB_CLK>,
+				 <&xo_board 0>;
+			clock-names = "core", "iface", "xo_clock";
 			bus-width = <8>;
 			non-removable;
 			status = "disabled";
@@ -527,8 +528,9 @@
 			interrupts = <0 125 0>, <0 221 0>;
 			interrupt-names = "hc_irq", "pwr_irq";
 			clocks = <&gcc GCC_SDCC2_APPS_CLK>,
-				 <&gcc GCC_SDCC2_AHB_CLK>;
-			clock-names = "core", "iface";
+				 <&gcc GCC_SDCC2_AHB_CLK>,
+				 <&xo_board 0>;
+			clock-names = "core", "iface", "xo_clock";
 			bus-width = <4>;
 			status = "disabled";
 		};
arch/arm64/boot/dts/qcom/msm8996.dtsi +5 -4
index 338f82a..73178fd 100644
@@ -192,14 +192,14 @@
 	};
 
 	clocks {
-		xo_board {
+		xo_board: xo_board {
 			compatible = "fixed-clock";
 			#clock-cells = <0>;
 			clock-frequency = <19200000>;
 			clock-output-names = "xo_board";
 		};
 
-		sleep_clk {
+		sleep_clk: sleep_clk {
 			compatible = "fixed-clock";
 			#clock-cells = <0>;
 			clock-frequency = <32764>;
@@ -347,9 +347,10 @@
 			 interrupts = <0 125 0>, <0 221 0>;
 			 interrupt-names = "hc_irq", "pwr_irq";
 
-			 clock-names = "iface", "core";
+			 clock-names = "iface", "core", "xo_clock";
 			 clocks = <&gcc GCC_SDCC2_AHB_CLK>,
-			 <&gcc GCC_SDCC2_APPS_CLK>;
+			 <&gcc GCC_SDCC2_APPS_CLK>,
+			 <&xo_board 0>;
 			 bus-width = <4>;
 		 };
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#10 lore raw

Re: [PATCH v7 04/14] ARM: dts: Add xo_clock to sdhc nodes on qcom platforms

Quoted 10 lines · Ritesh Harjani

Add xo_clock to sdhc clock node on all qcom platforms.

---
 arch/arm/boot/dts/qcom-apq8084.dtsi   | 14 ++++++++------
 arch/arm/boot/dts/qcom-msm8974.dtsi   | 14 ++++++++------
 arch/arm64/boot/dts/qcom/msm8916.dtsi | 10 ++++++----
 arch/arm64/boot/dts/qcom/msm8996.dtsi |  9 +++++----
 4 files changed, 27 insertions(+), 20 deletions(-)

Is there an update to Documentation/devicetree/bindings/mmc/sdhci-msm.txt as well?

Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

A message that is not in the archive (sent off-list or not yet received): [email protected]

#12 lore raw

Re: [PATCH v7 04/14] ARM: dts: Add xo_clock to sdhc nodes on qcom platforms

Quoted 12 lines · Stephen Boyd
Quoted 10 lines · Ritesh Harjani

Add xo_clock to sdhc clock node on all qcom platforms.

---
 arch/arm/boot/dts/qcom-apq8084.dtsi   | 14 ++++++++------
 arch/arm/boot/dts/qcom-msm8974.dtsi   | 14 ++++++++------
 arch/arm64/boot/dts/qcom/msm8916.dtsi | 10 ++++++----
 arch/arm64/boot/dts/qcom/msm8996.dtsi |  9 +++++----
 4 files changed, 27 insertions(+), 20 deletions(-)
Quoted 2 lines

Is there an update to Documentation/devicetree/bindings/mmc/sdhci-msm.txt as well?

Sure, I will update.

Quoted 1 line
Signature · 6 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#13 lore raw

PATCH v7 5/14 mmc: sdhci-msm: Update DLL reset sequence

In reply to Ritesh Harjani

From: Venkat Gopalakrishnan <[email protected]>

SDCC core with minor version >= 0x42 introduced new 14lpp DLL. This has additional requirements in the reset sequence for DLL tuning. Make necessary changes as needed.

Without this patch we see below errors on such SDHC controllers
sdhci_msm 7464900.sdhci: mmc0: DLL failed to LOCK
mmc0: tuning execution failed: -110

---
 drivers/mmc/host/sdhci-msm.c | 61 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)
drivers/mmc/host/sdhci-msm.c +61 -0
index 6ea89c2..643956e 100644
@@ -58,6 +58,10 @@
 #define CORE_DLL_CONFIG		0x100
 #define CORE_DLL_STATUS		0x108
 
+#define CORE_DLL_CONFIG_2	0x1b4
+#define CORE_FLL_CYCLE_CNT	BIT(18)
+#define CORE_DLL_CLOCK_DISABLE	BIT(21)
+
 #define CORE_VENDOR_SPEC	0x10c
 #define CORE_CLK_PWRSAVE	BIT(1)
 
@@ -75,7 +79,9 @@ struct sdhci_msm_host {
 	struct clk *clk;	/* main SD/MMC bus clock */
 	struct clk *pclk;	/* SDHC peripheral bus clock */
 	struct clk *bus_clk;	/* SDHC bus voter clock */
+	struct clk *xo_clk;	/* TCXO clk needed for FLL feature of cm_dll*/
 	struct mmc_host *mmc;
+	bool use_14lpp_dll_reset;
 };
 
 /* Platform specific tuning */
@@ -304,6 +310,8 @@ static inline void msm_cm_dll_set_freq(struct sdhci_host *host)
 static int msm_init_cm_dll(struct sdhci_host *host)
 {
 	struct mmc_host *mmc = host->mmc;
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
 	int wait_cnt = 50;
 	unsigned long flags;
 	u32 config;
@@ -319,6 +327,16 @@ static int msm_init_cm_dll(struct sdhci_host *host)
 	config &= ~CORE_CLK_PWRSAVE;
 	writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
 
+	if (msm_host->use_14lpp_dll_reset) {
+		config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+		config &= ~CORE_CK_OUT_EN;
+		writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
+
+		config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG_2);
+		config |= CORE_DLL_CLOCK_DISABLE;
+		writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG_2);
+	}
+
 	/* Write 1 to DLL_RST bit of DLL_CONFIG register */
 	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
 	config |= CORE_DLL_RST;
@@ -330,6 +348,28 @@ static int msm_init_cm_dll(struct sdhci_host *host)
 	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 	msm_cm_dll_set_freq(host);
 
+	if (msm_host->use_14lpp_dll_reset &&
+	    !IS_ERR_OR_NULL(msm_host->xo_clk)) {
+		u32 mclk_freq = 0;
+
+		config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG_2);
+		config &= CORE_FLL_CYCLE_CNT;
+		if (config)
+			mclk_freq = DIV_ROUND_CLOSEST_ULL((host->clock * 8),
+							  clk_get_rate(msm_host->xo_clk));
+		else
+			mclk_freq = DIV_ROUND_CLOSEST_ULL((host->clock * 4),
+							  clk_get_rate(msm_host->xo_clk));
+
+		config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG_2);
+		config &= ~(0xFF << 10);
+		config |= mclk_freq << 10;
+
+		writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG_2);
+		/* wait for 5us before enabling DLL clock */
+		udelay(5);
+	}
+
 	/* Write 0 to DLL_RST bit of DLL_CONFIG register */
 	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
 	config &= ~CORE_DLL_RST;
@@ -340,6 +380,14 @@ static int msm_init_cm_dll(struct sdhci_host *host)
 	config &= ~CORE_DLL_PDN;
 	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
 
+	if (msm_host->use_14lpp_dll_reset) {
+		msm_cm_dll_set_freq(host);
+		/* Enable the DLL clock */
+		config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG_2);
+		config &= ~CORE_DLL_CLOCK_DISABLE;
+		writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG_2);
+	}
+
 	/* Set DLL_EN bit to 1. */
 	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
 	config |= CORE_DLL_EN;
@@ -595,6 +643,16 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 		goto pclk_disable;
 	}
 
+	/*
+	 * xo_clock is needed for FLL feature of cm_dll.
+	 * In case if xo_clock is not mentioned in DT, warn and proceed.
+	 */
+	msm_host->xo_clk = devm_clk_get(&pdev->dev, "xo_clock");
+	if (IS_ERR(msm_host->xo_clk)) {
+		ret = PTR_ERR(msm_host->xo_clk);
+		dev_warn(&pdev->dev, "TCXO clk not present (%d)\n", ret);
+	}
+
 	/* Vote for maximum clock rate for maximum performance */
 	ret = clk_set_rate(msm_host->clk, INT_MAX);
 	if (ret)
@@ -641,6 +699,9 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 	dev_dbg(&pdev->dev, "MCI Version: 0x%08x, major: 0x%04x, minor: 0x%02x\n",
 		core_version, core_major, core_minor);
 
+	if (core_major == 1 && core_minor >= 0x42)
+		msm_host->use_14lpp_dll_reset = true;
+
 	/*
 	 * Support for some capabilities is not advertised by newer
 	 * controller versions and must be explicitly enabled.
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#14 lore raw

PATCH v7 6/14 mmc: sdhci-msm: Add get_min_clock() and get_max_clock() callback

In reply to Ritesh Harjani

This add get_min_clock() and get_max_clock() callback for sdhci-msm. sdhci-msm min/max clocks may be different hence implement these callbacks.

---
 drivers/mmc/host/sdhci-msm.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
drivers/mmc/host/sdhci-msm.c +17 -0
index 643956e..f63ffb0 100644
@@ -67,6 +67,8 @@
 
 #define CORE_VENDOR_SPEC_CAPABILITIES0	0x11c
 
+#define SDHCI_MSM_MIN_CLOCK	400000
+
 #define CDR_SELEXT_SHIFT	20
 #define CDR_SELEXT_MASK		(0xf << CDR_SELEXT_SHIFT)
 #define CMUX_SHIFT_PHASE_SHIFT	24
@@ -562,6 +564,19 @@ static irqreturn_t sdhci_msm_pwr_irq(int irq, void *data)
 	return IRQ_HANDLED;
 }
 
+static unsigned int sdhci_msm_get_max_clock(struct sdhci_host *host)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+
+	return clk_round_rate(msm_host->clk, ULONG_MAX);
+}
+
+static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host)
+{
+	return SDHCI_MSM_MIN_CLOCK;
+}
+
 static const struct of_device_id sdhci_msm_dt_match[] = {
 	{ .compatible = "qcom,sdhci-msm-v4" },
 	{},
@@ -573,6 +588,8 @@ static irqreturn_t sdhci_msm_pwr_irq(int irq, void *data)
 	.platform_execute_tuning = sdhci_msm_execute_tuning,
 	.reset = sdhci_reset,
 	.set_clock = sdhci_set_clock,
+	.get_min_clock = sdhci_msm_get_min_clock,
+	.get_max_clock = sdhci_msm_get_max_clock,
 	.set_bus_width = sdhci_set_bus_width,
 	.set_uhs_signaling = sdhci_msm_set_uhs_signaling,
 	.voltage_switch = sdhci_msm_voltage_switch,
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#15 lore raw

PATCH v7 8/14 mmc: sdhci-msm: Implement set_clock callback for sdhci-msm

In reply to Ritesh Harjani

sdhci-msm controller may have different clk-rates for each bus speed mode. Thus implement set_clock callback for sdhci-msm driver.

---
 drivers/mmc/host/sdhci-msm.c | 87 +++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 86 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-msm.c +86 -1
index 268424c..b96a4a7 100644
@@ -82,6 +82,7 @@ struct sdhci_msm_host {
 	struct clk *pclk;	/* SDHC peripheral bus clock */
 	struct clk *bus_clk;	/* SDHC bus voter clock */
 	struct clk *xo_clk;	/* TCXO clk needed for FLL feature of cm_dll*/
+	unsigned long clk_rate;
 	struct mmc_host *mmc;
 	bool use_14lpp_dll_reset;
 };
@@ -577,6 +578,90 @@ static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host)
 	return SDHCI_MSM_MIN_CLOCK;
 }
 
+/**
+ * __sdhci_msm_set_clock - sdhci_msm clock control.
+ *
+ * Description:
+ * Implement MSM version of sdhci_set_clock.
+ * This is required since MSM controller does not
+ * use internal divider and instead directly control
+ * the GCC clock as per HW recommendation.
+ **/
+void __sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+	u16 clk;
+	unsigned long timeout;
+
+	/*
+	 * Keep actual_clock as zero -
+	 * - since there is no divider used so no need of having actual_clock.
+	 * - MSM controller uses SDCLK for data timeout calculation. If
+	 *   actual_clock is zero, host->clock is taken for calculation.
+	 */
+	host->mmc->actual_clock = 0;
+
+	sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+	if (clock == 0)
+		return;
+
+	/*
+	 * MSM controller do not use clock divider.
+	 * Thus read SDHCI_CLOCK_CONTROL and only enable
+	 * clock with no divider value programmed.
+	 */
+	clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+
+	clk |= SDHCI_CLOCK_INT_EN;
+	sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
+
+	/* Wait max 20 ms */
+	timeout = 20;
+	while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
+		& SDHCI_CLOCK_INT_STABLE)) {
+		if (timeout == 0) {
+			pr_err("%s: Internal clock never stabilised\n",
+			       mmc_hostname(host->mmc));
+			return;
+		}
+		timeout--;
+		mdelay(1);
+	}
+
+	clk |= SDHCI_CLOCK_CARD_EN;
+	sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
+}
+
+/* sdhci_msm_set_clock - Called with (host->lock) spinlock held. */
+static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+	int rc;
+
+	if (!clock) {
+		msm_host->clk_rate = clock;
+		goto out;
+	}
+
+	spin_unlock_irq(&host->lock);
+	if (clock != msm_host->clk_rate) {
+		rc = clk_set_rate(msm_host->clk, clock);
+		if (rc) {
+			pr_err("%s: Failed to set clock at rate %u\n",
+			       mmc_hostname(host->mmc), clock);
+			spin_lock_irq(&host->lock);
+			goto out;
+		}
+		msm_host->clk_rate = clock;
+		pr_debug("%s: Setting clock at rate %lu\n",
+			 mmc_hostname(host->mmc), clk_get_rate(msm_host->clk));
+	}
+	spin_lock_irq(&host->lock);
+out:
+	__sdhci_msm_set_clock(host, clock);
+}
+
 static const struct of_device_id sdhci_msm_dt_match[] = {
 	{ .compatible = "qcom,sdhci-msm-v4" },
 	{},
@@ -587,7 +672,7 @@ static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host)
 static const struct sdhci_ops sdhci_msm_ops = {
 	.platform_execute_tuning = sdhci_msm_execute_tuning,
 	.reset = sdhci_reset,
-	.set_clock = sdhci_set_clock,
+	.set_clock = sdhci_msm_set_clock,
 	.get_min_clock = sdhci_msm_get_min_clock,
 	.get_max_clock = sdhci_msm_get_max_clock,
 	.set_bus_width = sdhci_set_bus_width,
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#16 lore raw

Re: [PATCH v7 08/14] mmc: sdhci-msm: Implement set_clock callback for sdhci-msm

Quoted 46 lines · Ritesh Harjani
drivers/mmc/host/sdhci-msm.c +42 -0
@@ -577,6 +578,90 @@ static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host)
 	return SDHCI_MSM_MIN_CLOCK;
 }
 
+/**
+ * __sdhci_msm_set_clock - sdhci_msm clock control.
+ *
+ * Description:
+ * Implement MSM version of sdhci_set_clock.
+ * This is required since MSM controller does not
+ * use internal divider and instead directly control
+ * the GCC clock as per HW recommendation.
+ **/
+void __sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+	u16 clk;
+	unsigned long timeout;
+
+	/*
+	 * Keep actual_clock as zero -
+	 * - since there is no divider used so no need of having actual_clock.
+	 * - MSM controller uses SDCLK for data timeout calculation. If
+	 *   actual_clock is zero, host->clock is taken for calculation.
+	 */
+	host->mmc->actual_clock = 0;
+
+	sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+	if (clock == 0)
+		return;
+
+	/*
+	 * MSM controller do not use clock divider.
+	 * Thus read SDHCI_CLOCK_CONTROL and only enable
+	 * clock with no divider value programmed.
+	 */
+	clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+
+	clk |= SDHCI_CLOCK_INT_EN;
+	sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
+
+	/* Wait max 20 ms */
+	timeout = 20;
+	while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
+		& SDHCI_CLOCK_INT_STABLE)) {
+		if (timeout == 0) {

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 632

Quoted 10 lines
drivers/mmc/host/sdhci-msm.c +10 -0
+			pr_err("%s: Internal clock never stabilised\n",
+			       mmc_hostname(host->mmc));
+			return;
+		}
+		timeout--;
+		mdelay(1);
+	}
+
+	clk |= SDHCI_CLOCK_CARD_EN;
+	sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);

This is almost a copy/paste of sdhci_set_clock(). Can we make sdhci_set_clock() call a __sdhci_set_clock() function that takes unsigned int clock, and also a flag indicating if we want to set the internal clock divider or not? Then we can call __sdhci_set_clock() from sdhci_set_clock() with (clock, true) as arguments and (clock, false).

Quoted 6 lines
drivers/mmc/host/sdhci-msm.c +6 -0
+}
+
+/* sdhci_msm_set_clock - Called with (host->lock) spinlock held. */
+static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 648

Quoted 10 lines
drivers/mmc/host/sdhci-msm.c +10 -0
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+	int rc;
+
+	if (!clock) {
+		msm_host->clk_rate = clock;
+		goto out;
+	}
+
+	spin_unlock_irq(&host->lock);
+	if (clock != msm_host->clk_rate) {

Why do we need to check here? Can't we call clk_set_rate() Unconditionally?

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 654

Quoted 6 lines
drivers/mmc/host/sdhci-msm.c +6 -0
+		rc = clk_set_rate(msm_host->clk, clock);
+		if (rc) {
+			pr_err("%s: Failed to set clock at rate %u\n",
+			       mmc_hostname(host->mmc), clock);
+			spin_lock_irq(&host->lock);
+			goto out;

Or replace the above two lines with goto err;

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 659

Quoted 5 lines
drivers/mmc/host/sdhci-msm.c +5 -0
+		}
+		msm_host->clk_rate = clock;
+		pr_debug("%s: Setting clock at rate %lu\n",
+			 mmc_hostname(host->mmc), clk_get_rate(msm_host->clk));
+	}

And put an err label here.

Quoted 8 lines
drivers/mmc/host/sdhci-msm.c +5 -0
+	spin_lock_irq(&host->lock);
+out:
+	__sdhci_msm_set_clock(host, clock);
+}
+
 static const struct of_device_id sdhci_msm_dt_match[] = {
 	{ .compatible = "qcom,sdhci-msm-v4" },
 	{},
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#17 lore raw

Re: [PATCH v7 08/14] mmc: sdhci-msm: Implement set_clock callback for sdhci-msm

Hi Stephen/Adrian,

Quoted 56 lines · Stephen Boyd
Quoted 55 lines · Ritesh Harjani
drivers/mmc/host/sdhci-msm.c +51 -0
@@ -577,6 +578,90 @@ static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host)
 	return SDHCI_MSM_MIN_CLOCK;
 }
 
+/**
+ * __sdhci_msm_set_clock - sdhci_msm clock control.
+ *
+ * Description:
+ * Implement MSM version of sdhci_set_clock.
+ * This is required since MSM controller does not
+ * use internal divider and instead directly control
+ * the GCC clock as per HW recommendation.
+ **/
+void __sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+	u16 clk;
+	unsigned long timeout;
+
+	/*
+	 * Keep actual_clock as zero -
+	 * - since there is no divider used so no need of having actual_clock.
+	 * - MSM controller uses SDCLK for data timeout calculation. If
+	 *   actual_clock is zero, host->clock is taken for calculation.
+	 */
+	host->mmc->actual_clock = 0;
+
+	sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+	if (clock == 0)
+		return;
+
+	/*
+	 * MSM controller do not use clock divider.
+	 * Thus read SDHCI_CLOCK_CONTROL and only enable
+	 * clock with no divider value programmed.
+	 */
+	clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+
+	clk |= SDHCI_CLOCK_INT_EN;
+	sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
+
+	/* Wait max 20 ms */
+	timeout = 20;
+	while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
+		& SDHCI_CLOCK_INT_STABLE)) {
+		if (timeout == 0) {
+			pr_err("%s: Internal clock never stabilised\n",
+			       mmc_hostname(host->mmc));
+			return;
+		}
+		timeout--;
+		mdelay(1);
+	}
+
+	clk |= SDHCI_CLOCK_CARD_EN;
Quoted 8 lines

Review static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 632

Quoted 1 line
(quoted diff) +1 -0
+	sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);

This is almost a copy/paste of sdhci_set_clock(). Can we make sdhci_set_clock() call a __sdhci_set_clock() function that takes unsigned int clock, and also a flag indicating if we want to set the internal clock divider or not? Then we can call __sdhci_set_clock() from sdhci_set_clock() with (clock, true) as arguments and (clock, false).

Adrian,
Could you please comment here ?

Quoted 12 lines
Quoted 12 lines
drivers/mmc/host/sdhci-msm.c +11 -0
+}
+
+/* sdhci_msm_set_clock - Called with (host->lock) spinlock held. */
+static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+	int rc;
+
+	if (!clock) {
+		msm_host->clk_rate = clock;
Quoted 8 lines

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 648

Quoted 5 lines
drivers/mmc/host/sdhci-msm.c +5 -0
+		goto out;
+	}
+
+	spin_unlock_irq(&host->lock);
+	if (clock != msm_host->clk_rate) {

Why do we need to check here? Can't we call clk_set_rate() Unconditionally?

Since it may so happen that above layers may call for ->set_clock
function with same requested clock more than once, hence we cache the host->clock here.
Also, since requested clock (host->clock) can be say 400Mhz but the actual pltfm supported clock would be say 384MHz.

Quoted 9 lines

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 654

Quoted 7 lines
drivers/mmc/host/sdhci-msm.c +6 -0
+		rc = clk_set_rate(msm_host->clk, clock);
+		if (rc) {
+			pr_err("%s: Failed to set clock at rate %u\n",
+			       mmc_hostname(host->mmc), clock);
+			spin_lock_irq(&host->lock);
+			goto out;

Or replace the above two lines with goto err;

Ok, I will have another label out_lock instead of err.

Quoted 8 lines

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 659

Quoted 6 lines
drivers/mmc/host/sdhci-msm.c +5 -0
+		}
+		msm_host->clk_rate = clock;
+		pr_debug("%s: Setting clock at rate %lu\n",
+			 mmc_hostname(host->mmc), clk_get_rate(msm_host->clk));
+	}

And put an err label here.

will put the label here as out_lock;

Quoted 10 lines
Quoted 9 lines
drivers/mmc/host/sdhci-msm.c +5 -0
+	spin_lock_irq(&host->lock);
+out:
+	__sdhci_msm_set_clock(host, clock);
+}
+
 static const struct of_device_id sdhci_msm_dt_match[] = {
 	{ .compatible = "qcom,sdhci-msm-v4" },
 	{},
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#18 lore raw

Re: [PATCH v7 08/14] mmc: sdhci-msm: Implement set_clock callback for sdhci-msm

Quoted 22 lines · Ritesh Harjani
Quoted 13 lines · Stephen Boyd
Quoted 11 lines · Ritesh Harjani
drivers/mmc/host/sdhci-msm.c +11 -0
+}
+
+/* sdhci_msm_set_clock - Called with (host->lock) spinlock held. */
+static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+	int rc;
+
+	if (!clock) {
+		msm_host->clk_rate = clock;
Quoted 8 lines

drivers/mmc/host/sdhci-msm.c static unsigned int sdhci_msm_get_min_clock(struct sdhci_host *host) · line 648

Quoted 5 lines
drivers/mmc/host/sdhci-msm.c +5 -0
+		goto out;
+	}
+
+	spin_unlock_irq(&host->lock);
+	if (clock != msm_host->clk_rate) {

Why do we need to check here? Can't we call clk_set_rate() Unconditionally?

Quoted 5 lines

Since it may so happen that above layers may call for ->set_clock function with same requested clock more than once, hence we cache the host->clock here.
Also, since requested clock (host->clock) can be say 400Mhz but the actual pltfm supported clock would be say 384MHz.

clk_set_rate() detects the same rate being set even after it internally rounds the rate. We're not going to touch the clk hardware if 400 is requested once but 384 is what's set and then 400 is requested again. Caching the rate here in the driver can lead to problems too if the driver is out of sync with the clk hardware state, so it's best to avoid doing anything fancy here and just let the framework handle duplicates.

Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#19 lore raw

Re: [PATCH v7 08/14] mmc: sdhci-msm: Implement set_clock callback for sdhci-msm

In reply to Ritesh Harjani

Hi,

Quoted 69 lines · Ritesh Harjani

Hi Stephen/Adrian,

Quoted 66 lines · Stephen Boyd
Quoted 48 lines · Ritesh Harjani
(quoted diff) +17 -0
@@ -577,6 +578,90 @@ static unsigned int
 sdhci_msm_get_min_clock(struct sdhci_host *host)
     return SDHCI_MSM_MIN_CLOCK;
 }
 
+/**
+ * __sdhci_msm_set_clock - sdhci_msm clock control.
+ *
+ * Description:
+ * Implement MSM version of sdhci_set_clock.
+ * This is required since MSM controller does not
+ * use internal divider and instead directly control
+ * the GCC clock as per HW recommendation.
+ **/
+void __sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+    u16 clk;
+    unsigned long timeout;
+
+    /*
+     * Keep actual_clock as zero -
+     * - since there is no divider used so no need of having

actual_clock.

(quoted diff) +25 -0
+     * - MSM controller uses SDCLK for data timeout calculation. If
+     *   actual_clock is zero, host->clock is taken for calculation.
+     */
+    host->mmc->actual_clock = 0;
+
+    sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+    if (clock == 0)
+        return;
+
+    /*
+     * MSM controller do not use clock divider.
+     * Thus read SDHCI_CLOCK_CONTROL and only enable
+     * clock with no divider value programmed.
+     */
+    clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+
+    clk |= SDHCI_CLOCK_INT_EN;
+    sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
+
+    /* Wait max 20 ms */
+    timeout = 20;
+    while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
+        & SDHCI_CLOCK_INT_STABLE)) {
+        if (timeout == 0) {

Review

Quoted 10 lines
(quoted diff) +10 -0
+            pr_err("%s: Internal clock never stabilised\n",
+                   mmc_hostname(host->mmc));
+            return;
+        }
+        timeout--;
+        mdelay(1);
+    }
+
+    clk |= SDHCI_CLOCK_CARD_EN;
+    sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);

This is almost a copy/paste of sdhci_set_clock(). Can we make sdhci_set_clock() call a __sdhci_set_clock() function that takes unsigned int clock, and also a flag indicating if we want to set the internal clock divider or not? Then we can call __sdhci_set_clock() from sdhci_set_clock() with (clock, true) as arguments and (clock, false).

Actually what you may be referring here is some sort of quirks which is not entertained any more for sdhci driver. sdhci is tending towards becoming a library and hence such changes are restricted.

But I think we may do below changes to avoid duplication of code which enables the sdhci internal clock and waits for internal clock to be stable.

Adrian, could you please tell if this should be ok?
Then we may be able to call for sdhci_set_clock_enable function from sdhci_msm_set_clock.

drivers/mmc/host/sdhci.c +19 -12
index 42ef3eb..28e605c 100644
@@ -1343,19 +1343,8 @@ u16 sdhci_calc_clk(struct sdhci_host *host, unsigned int clock,
 }
 EXPORT_SYMBOL_GPL(sdhci_calc_clk);
 
-void sdhci_set_clock(struct sdhci_host *host, unsigned int clock)
+void sdhci_set_clock_enable(struct sdhci_host *host, unsigned short clk)
 {
-       u16 clk;
-       unsigned long timeout;
-
-       host->mmc->actual_clock = 0;
-
-       sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
-
-       if (clock == 0)
-               return;
-
-       clk = sdhci_calc_clk(host, clock, &host->mmc->actual_clock);
 
        clk |= SDHCI_CLOCK_INT_EN;
        sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
@@ -1377,6 +1366,24 @@ void sdhci_set_clock(struct sdhci_host *host, unsigned int clock)
        clk |= SDHCI_CLOCK_CARD_EN;
        sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
 }
+EXPORT_SYMBOL_GPL(sdhci_set_clock_enable);
+
+void sdhci_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+       u16 clk;
+       unsigned long timeout;
+
+       host->mmc->actual_clock = 0;
+
+       sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+       if (clock == 0)
+               return;
+
+       clk = sdhci_calc_clk(host, clock, &host->mmc->actual_clock);
+
+       sdhci_set_clock_enable(host, clk);
+}
 EXPORT_SYMBOL_GPL(sdhci_set_clock);
 
 static void sdhci_set_power_reg(struct sdhci_host *host, unsigned char 

mode,

drivers/mmc/host/sdhci.h +1 -0
index 766df17..43382e1 100644
@@ -681,6 +681,7 @@ static inline bool sdhci_sdio_irq_enabled(struct sdhci_host *host)
 u16 sdhci_calc_clk(struct sdhci_host *host, unsigned int clock,
                   unsigned int *actual_clock);
 void sdhci_set_clock(struct sdhci_host *host, unsigned int clock);
+void sdhci_set_clock_enable(struct sdhci_host *host, unsigned short clk);
 void sdhci_set_power(struct sdhci_host *host, unsigned char mode,
                     unsigned short vdd);
 void sdhci_set_power_noreg(struct sdhci_host *host, unsigned char mode,
Quoted 61 lines

Adrian,
Could you please comment here ?

Quoted 13 lines
Quoted 13 lines
(quoted diff) +4 -0
+}
+
+/* sdhci_msm_set_clock - Called with (host->lock) spinlock held. */
+static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned

int clock)

(quoted diff) +7 -0
+{
+    struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+    struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+    int rc;
+
+    if (!clock) {
+        msm_host->clk_rate = clock;
Quoted 8 lines

Review

Quoted 5 lines
(quoted diff) +5 -0
+        goto out;
+    }
+
+    spin_unlock_irq(&host->lock);
+    if (clock != msm_host->clk_rate) {

Why do we need to check here? Can't we call clk_set_rate() Unconditionally?

Since it may so happen that above layers may call for ->set_clock function with same requested clock more than once, hence we cache the host->clock here.
Also, since requested clock (host->clock) can be say 400Mhz but the actual pltfm supported clock would be say 384MHz.

Quoted 9 lines

Review

Quoted 7 lines
(quoted diff) +6 -0
+        rc = clk_set_rate(msm_host->clk, clock);
+        if (rc) {
+            pr_err("%s: Failed to set clock at rate %u\n",
+                   mmc_hostname(host->mmc), clock);
+            spin_lock_irq(&host->lock);
+            goto out;

Or replace the above two lines with goto err;

Ok, I will have another label out_lock instead of err.

Quoted 8 lines

Review

Quoted 6 lines
(quoted diff) +5 -0
+        }
+        msm_host->clk_rate = clock;
+        pr_debug("%s: Setting clock at rate %lu\n",
+             mmc_hostname(host->mmc), clk_get_rate(msm_host->clk));
+    }

And put an err label here.

will put the label here as out_lock;

Quoted 10 lines
Quoted 9 lines
(quoted diff) +5 -0
+    spin_lock_irq(&host->lock);
+out:
+    __sdhci_msm_set_clock(host, clock);
+}
+
 static const struct of_device_id sdhci_msm_dt_match[] = {
     { .compatible = "qcom,sdhci-msm-v4" },
     {},
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
maintainer
#20 lore raw

Re: [PATCH v7 08/14] mmc: sdhci-msm: Implement set_clock callback for sdhci-msm

Quoted 77 lines · Ritesh Harjani

Hi,

Quoted 69 lines · Ritesh Harjani

Hi Stephen/Adrian,

Quoted 66 lines · Stephen Boyd
Quoted 48 lines · Ritesh Harjani
(quoted diff) +17 -0
@@ -577,6 +578,90 @@ static unsigned int
 sdhci_msm_get_min_clock(struct sdhci_host *host)
     return SDHCI_MSM_MIN_CLOCK;
 }
 
+/**
+ * __sdhci_msm_set_clock - sdhci_msm clock control.
+ *
+ * Description:
+ * Implement MSM version of sdhci_set_clock.
+ * This is required since MSM controller does not
+ * use internal divider and instead directly control
+ * the GCC clock as per HW recommendation.
+ **/
+void __sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+    u16 clk;
+    unsigned long timeout;
+
+    /*
+     * Keep actual_clock as zero -
+     * - since there is no divider used so no need of having

actual_clock.

(quoted diff) +25 -0
+     * - MSM controller uses SDCLK for data timeout calculation. If
+     *   actual_clock is zero, host->clock is taken for calculation.
+     */
+    host->mmc->actual_clock = 0;
+
+    sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+    if (clock == 0)
+        return;
+
+    /*
+     * MSM controller do not use clock divider.
+     * Thus read SDHCI_CLOCK_CONTROL and only enable
+     * clock with no divider value programmed.
+     */
+    clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+
+    clk |= SDHCI_CLOCK_INT_EN;
+    sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
+
+    /* Wait max 20 ms */
+    timeout = 20;
+    while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
+        & SDHCI_CLOCK_INT_STABLE)) {
+        if (timeout == 0) {

Review

Quoted 10 lines
(quoted diff) +10 -0
+            pr_err("%s: Internal clock never stabilised\n",
+                   mmc_hostname(host->mmc));
+            return;
+        }
+        timeout--;
+        mdelay(1);
+    }
+
+    clk |= SDHCI_CLOCK_CARD_EN;
+    sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);

This is almost a copy/paste of sdhci_set_clock(). Can we make sdhci_set_clock() call a __sdhci_set_clock() function that takes unsigned int clock, and also a flag indicating if we want to set the internal clock divider or not? Then we can call __sdhci_set_clock() from sdhci_set_clock() with (clock, true) as arguments and (clock, false).

Actually what you may be referring here is some sort of quirks which is not entertained any more for sdhci driver.
sdhci is tending towards becoming a library and hence such changes are restricted.

Quoted 4 lines

But I think we may do below changes to avoid duplication of code which enables the sdhci internal clock and waits for internal clock to be stable.

Adrian, could you please tell if this should be ok?

That seems fine, but the name seems too long - how about changing sdhci_set_clock_enable to sdhci_enable_clk.

Quoted 134 lines

Then we may be able to call for sdhci_set_clock_enable function from sdhci_msm_set_clock.

drivers/mmc/host/sdhci.c +1 -12
index 42ef3eb..28e605c 100644
@@ -1343,19 +1343,8 @@ u16 sdhci_calc_clk(struct sdhci_host *host, unsigned
 int clock,
 }
 EXPORT_SYMBOL_GPL(sdhci_calc_clk);
 
-void sdhci_set_clock(struct sdhci_host *host, unsigned int clock)
+void sdhci_set_clock_enable(struct sdhci_host *host, unsigned short clk)
 {
-       u16 clk;
-       unsigned long timeout;
-
-       host->mmc->actual_clock = 0;
-
-       sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
-
-       if (clock == 0)
-               return;
-
-       clk = sdhci_calc_clk(host, clock, &host->mmc->actual_clock);
 
        clk |= SDHCI_CLOCK_INT_EN;
        sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
drivers/mmc/host/sdhci.c +18 -0
@@ -1377,6 +1366,24 @@ void sdhci_set_clock(struct sdhci_host *host,
 unsigned int clock)
        clk |= SDHCI_CLOCK_CARD_EN;
        sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
 }
+EXPORT_SYMBOL_GPL(sdhci_set_clock_enable);
+
+void sdhci_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+       u16 clk;
+       unsigned long timeout;
+
+       host->mmc->actual_clock = 0;
+
+       sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+       if (clock == 0)
+               return;
+
+       clk = sdhci_calc_clk(host, clock, &host->mmc->actual_clock);
+
+       sdhci_set_clock_enable(host, clk);
+}
 EXPORT_SYMBOL_GPL(sdhci_set_clock);
 

static void sdhci_set_power_reg(struct sdhci_host *host, unsigned char mode,

drivers/mmc/host/sdhci.h +1 -0
index 766df17..43382e1 100644
@@ -681,6 +681,7 @@ static inline bool sdhci_sdio_irq_enabled(struct
 sdhci_host *host)
 u16 sdhci_calc_clk(struct sdhci_host *host, unsigned int clock,
                   unsigned int *actual_clock);
 void sdhci_set_clock(struct sdhci_host *host, unsigned int clock);
+void sdhci_set_clock_enable(struct sdhci_host *host, unsigned short clk);
 void sdhci_set_power(struct sdhci_host *host, unsigned char mode,
                     unsigned short vdd);

void sdhci_set_power_noreg(struct sdhci_host *host, unsigned char mode,

Quoted 61 lines

Adrian,
Could you please comment here ?

Quoted 13 lines
Quoted 13 lines
(quoted diff) +4 -0
+}
+
+/* sdhci_msm_set_clock - Called with (host->lock) spinlock held. */
+static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned

int clock)

(quoted diff) +7 -0
+{
+    struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+    struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+    int rc;
+
+    if (!clock) {
+        msm_host->clk_rate = clock;
Quoted 8 lines

Review

Quoted 5 lines
(quoted diff) +5 -0
+        goto out;
+    }
+
+    spin_unlock_irq(&host->lock);
+    if (clock != msm_host->clk_rate) {

Why do we need to check here? Can't we call clk_set_rate() Unconditionally?

Since it may so happen that above layers may call for ->set_clock function with same requested clock more than once, hence we cache the host->clock here.
Also, since requested clock (host->clock) can be say 400Mhz but the actual pltfm supported clock would be say 384MHz.

Quoted 9 lines

Review

Quoted 7 lines
(quoted diff) +6 -0
+        rc = clk_set_rate(msm_host->clk, clock);
+        if (rc) {
+            pr_err("%s: Failed to set clock at rate %u\n",
+                   mmc_hostname(host->mmc), clock);
+            spin_lock_irq(&host->lock);
+            goto out;

Or replace the above two lines with goto err;

Ok, I will have another label out_lock instead of err.

Quoted 8 lines

Review

Quoted 6 lines
(quoted diff) +5 -0
+        }
+        msm_host->clk_rate = clock;
+        pr_debug("%s: Setting clock at rate %lu\n",
+             mmc_hostname(host->mmc), clk_get_rate(msm_host->clk));
+    }

And put an err label here.

will put the label here as out_lock;

Quoted 10 lines
Quoted 9 lines
(quoted diff) +5 -0
+    spin_lock_irq(&host->lock);
+out:
+    __sdhci_msm_set_clock(host, clock);
+}
+
 static const struct of_device_id sdhci_msm_dt_match[] = {
     { .compatible = "qcom,sdhci-msm-v4" },
     {},
#21 lore raw

Re: [PATCH v7 08/14] mmc: sdhci-msm: Implement set_clock callback for sdhci-msm

Quoted 78 lines · Adrian Hunter
Quoted 77 lines · Ritesh Harjani

Hi,

Quoted 69 lines · Ritesh Harjani

Hi Stephen/Adrian,

Quoted 66 lines · Stephen Boyd
Quoted 48 lines · Ritesh Harjani
(quoted diff) +17 -0
@@ -577,6 +578,90 @@ static unsigned int
 sdhci_msm_get_min_clock(struct sdhci_host *host)
     return SDHCI_MSM_MIN_CLOCK;
 }
 
+/**
+ * __sdhci_msm_set_clock - sdhci_msm clock control.
+ *
+ * Description:
+ * Implement MSM version of sdhci_set_clock.
+ * This is required since MSM controller does not
+ * use internal divider and instead directly control
+ * the GCC clock as per HW recommendation.
+ **/
+void __sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+    u16 clk;
+    unsigned long timeout;
+
+    /*
+     * Keep actual_clock as zero -
+     * - since there is no divider used so no need of having

actual_clock.

(quoted diff) +25 -0
+     * - MSM controller uses SDCLK for data timeout calculation. If
+     *   actual_clock is zero, host->clock is taken for calculation.
+     */
+    host->mmc->actual_clock = 0;
+
+    sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+    if (clock == 0)
+        return;
+
+    /*
+     * MSM controller do not use clock divider.
+     * Thus read SDHCI_CLOCK_CONTROL and only enable
+     * clock with no divider value programmed.
+     */
+    clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+
+    clk |= SDHCI_CLOCK_INT_EN;
+    sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
+
+    /* Wait max 20 ms */
+    timeout = 20;
+    while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
+        & SDHCI_CLOCK_INT_STABLE)) {
+        if (timeout == 0) {

Review

Quoted 10 lines
(quoted diff) +10 -0
+            pr_err("%s: Internal clock never stabilised\n",
+                   mmc_hostname(host->mmc));
+            return;
+        }
+        timeout--;
+        mdelay(1);
+    }
+
+    clk |= SDHCI_CLOCK_CARD_EN;
+    sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);

This is almost a copy/paste of sdhci_set_clock(). Can we make sdhci_set_clock() call a __sdhci_set_clock() function that takes unsigned int clock, and also a flag indicating if we want to set the internal clock divider or not? Then we can call __sdhci_set_clock() from sdhci_set_clock() with (clock, true) as arguments and (clock, false).

Actually what you may be referring here is some sort of quirks which is not entertained any more for sdhci driver.
sdhci is tending towards becoming a library and hence such changes are restricted.

Quoted 7 lines
Quoted 4 lines

But I think we may do below changes to avoid duplication of code which enables the sdhci internal clock and waits for internal clock to be stable.

Adrian, could you please tell if this should be ok?

That seems fine, but the name seems too long - how about changing sdhci_set_clock_enable to sdhci_enable_clk.

Sure. Will make the changes then.

Quoted 136 lines
Quoted 135 lines

Then we may be able to call for sdhci_set_clock_enable function from sdhci_msm_set_clock.

drivers/mmc/host/sdhci.c +1 -12
index 42ef3eb..28e605c 100644
@@ -1343,19 +1343,8 @@ u16 sdhci_calc_clk(struct sdhci_host *host, unsigned
 int clock,
 }
 EXPORT_SYMBOL_GPL(sdhci_calc_clk);
 
-void sdhci_set_clock(struct sdhci_host *host, unsigned int clock)
+void sdhci_set_clock_enable(struct sdhci_host *host, unsigned short clk)
 {
-       u16 clk;
-       unsigned long timeout;
-
-       host->mmc->actual_clock = 0;
-
-       sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
-
-       if (clock == 0)
-               return;
-
-       clk = sdhci_calc_clk(host, clock, &host->mmc->actual_clock);
 
        clk |= SDHCI_CLOCK_INT_EN;
        sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
drivers/mmc/host/sdhci.c +18 -0
@@ -1377,6 +1366,24 @@ void sdhci_set_clock(struct sdhci_host *host,
 unsigned int clock)
        clk |= SDHCI_CLOCK_CARD_EN;
        sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
 }
+EXPORT_SYMBOL_GPL(sdhci_set_clock_enable);
+
+void sdhci_set_clock(struct sdhci_host *host, unsigned int clock)
+{
+       u16 clk;
+       unsigned long timeout;
+
+       host->mmc->actual_clock = 0;
+
+       sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
+
+       if (clock == 0)
+               return;
+
+       clk = sdhci_calc_clk(host, clock, &host->mmc->actual_clock);
+
+       sdhci_set_clock_enable(host, clk);
+}
 EXPORT_SYMBOL_GPL(sdhci_set_clock);
 

static void sdhci_set_power_reg(struct sdhci_host *host, unsigned char mode,

drivers/mmc/host/sdhci.h +1 -0
index 766df17..43382e1 100644
@@ -681,6 +681,7 @@ static inline bool sdhci_sdio_irq_enabled(struct
 sdhci_host *host)
 u16 sdhci_calc_clk(struct sdhci_host *host, unsigned int clock,
                   unsigned int *actual_clock);
 void sdhci_set_clock(struct sdhci_host *host, unsigned int clock);
+void sdhci_set_clock_enable(struct sdhci_host *host, unsigned short clk);
 void sdhci_set_power(struct sdhci_host *host, unsigned char mode,
                     unsigned short vdd);

void sdhci_set_power_noreg(struct sdhci_host *host, unsigned char mode,

Quoted 61 lines

Adrian,
Could you please comment here ?

Quoted 13 lines
Quoted 13 lines
(quoted diff) +4 -0
+}
+
+/* sdhci_msm_set_clock - Called with (host->lock) spinlock held. */
+static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned

int clock)

(quoted diff) +7 -0
+{
+    struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+    struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+    int rc;
+
+    if (!clock) {
+        msm_host->clk_rate = clock;
Quoted 8 lines

Review

Quoted 5 lines
(quoted diff) +5 -0
+        goto out;
+    }
+
+    spin_unlock_irq(&host->lock);
+    if (clock != msm_host->clk_rate) {

Why do we need to check here? Can't we call clk_set_rate() Unconditionally?

Since it may so happen that above layers may call for ->set_clock function with same requested clock more than once, hence we cache the host->clock here.
Also, since requested clock (host->clock) can be say 400Mhz but the actual pltfm supported clock would be say 384MHz.

Quoted 9 lines

Review

Quoted 7 lines
(quoted diff) +6 -0
+        rc = clk_set_rate(msm_host->clk, clock);
+        if (rc) {
+            pr_err("%s: Failed to set clock at rate %u\n",
+                   mmc_hostname(host->mmc), clock);
+            spin_lock_irq(&host->lock);
+            goto out;

Or replace the above two lines with goto err;

Ok, I will have another label out_lock instead of err.

Quoted 8 lines

Review

Quoted 6 lines
(quoted diff) +5 -0
+        }
+        msm_host->clk_rate = clock;
+        pr_debug("%s: Setting clock at rate %lu\n",
+             mmc_hostname(host->mmc), clk_get_rate(msm_host->clk));
+    }

And put an err label here.

will put the label here as out_lock;

Quoted 10 lines
Quoted 9 lines
(quoted diff) +5 -0
+    spin_lock_irq(&host->lock);
+out:
+    __sdhci_msm_set_clock(host, clock);
+}
+
 static const struct of_device_id sdhci_msm_dt_match[] = {
     { .compatible = "qcom,sdhci-msm-v4" },
     {},
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#22 lore raw

PATCH v7 11/14 mmc: sdhci-msm: Add HS400 platform support

In reply to Ritesh Harjani

From: Venkat Gopalakrishnan <[email protected]>

The following msm platform specific changes are added to support HS400.
- Allow tuning for HS400 mode.
- Configure HS400 timing mode using the VENDOR_SPECIFIC_FUNC register.

---
 drivers/mmc/host/sdhci-msm.c | 127 +++++++++++++++++++++++++++++++++++++++----
 1 file changed, 117 insertions(+), 10 deletions(-)
drivers/mmc/host/sdhci-msm.c +117 -10
index 41a4ea7..7618bad 100644
@@ -31,6 +31,7 @@
 #define HC_MODE_EN		0x1
 #define CORE_POWER		0x0
 #define CORE_SW_RST		BIT(7)
+#define FF_CLK_SW_RST_DIS	BIT(13)
 
 #define CORE_PWRCTL_STATUS	0xdc
 #define CORE_PWRCTL_MASK	0xe0
@@ -64,10 +65,17 @@
 
 #define CORE_VENDOR_SPEC	0x10c
 #define CORE_CLK_PWRSAVE	BIT(1)
+#define CORE_HC_MCLK_SEL_DFLT	(2 << 8)
+#define CORE_HC_MCLK_SEL_HS400	(3 << 8)
+#define CORE_HC_MCLK_SEL_MASK	(3 << 8)
+#define CORE_HC_SELECT_IN_EN	BIT(18)
+#define CORE_HC_SELECT_IN_HS400	(6 << 19)
+#define CORE_HC_SELECT_IN_MASK	(7 << 19)
 
 #define CORE_VENDOR_SPEC_CAPABILITIES0	0x11c
 
 #define SDHCI_MSM_MIN_CLOCK	400000
+#define CORE_FREQ_100MHZ	(100 * 1000 * 1000)
 
 #define CDR_SELEXT_SHIFT	20
 #define CDR_SELEXT_MASK		(0xf << CDR_SELEXT_SHIFT)
@@ -85,6 +93,8 @@ struct sdhci_msm_host {
 	unsigned long clk_rate;
 	struct mmc_host *mmc;
 	bool use_14lpp_dll_reset;
+	bool tuning_done;
+	bool calibration_done;
 };
 
 /* Platform specific tuning */
@@ -173,8 +183,8 @@ static int msm_config_cm_dll_phase(struct sdhci_host *host, u8 phase)
  * Find out the greatest range of consecuitive selected
  * DLL clock output phases that can be used as sampling
  * setting for SD3.0 UHS-I card read operation (in SDR104
- * timing mode) or for eMMC4.5 card read operation (in HS200
- * timing mode).
+ * timing mode) or for eMMC4.5 card read operation (in
+ * HS400/HS200 timing mode).
  * Select the 3/4 of the range and configure the DLL with the
  * selected DLL clock output phase.
  */
@@ -430,9 +440,10 @@ static int sdhci_msm_execute_tuning(struct sdhci_host *host, u32 opcode)
 	 * Tuning is required for SDR104, HS200 and HS400 cards and
 	 * if clock frequency is greater than 100MHz in these modes.
 	 */
-	if (host->clock <= 100 * 1000 * 1000 ||
-	    !((ios.timing == MMC_TIMING_MMC_HS200) ||
-	      (ios.timing == MMC_TIMING_UHS_SDR104)))
+	if (host->clock <= CORE_FREQ_100MHZ ||
+	    !(ios.timing == MMC_TIMING_MMC_HS400 ||
+	    ios.timing == MMC_TIMING_MMC_HS200 ||
+	    ios.timing == MMC_TIMING_UHS_SDR104))
 		return 0;
 
 retry:
@@ -483,6 +494,8 @@ static int sdhci_msm_execute_tuning(struct sdhci_host *host, u32 opcode)
 		rc = -EIO;
 	}
 
+	if (!rc)
+		msm_host->tuning_done = true;
 	return rc;
 }
 
@@ -490,7 +503,10 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
 					unsigned int uhs)
 {
 	struct mmc_host *mmc = host->mmc;
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
 	u16 ctrl_2;
+	u32 config;
 
 	ctrl_2 = sdhci_readw(host, SDHCI_HOST_CONTROL2);
 	/* Select Bus Speed Mode for host */
@@ -505,6 +521,7 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
 	case MMC_TIMING_UHS_SDR50:
 		ctrl_2 |= SDHCI_CTRL_UHS_SDR50;
 		break;
+	case MMC_TIMING_MMC_HS400:
 	case MMC_TIMING_MMC_HS200:
 	case MMC_TIMING_UHS_SDR104:
 		ctrl_2 |= SDHCI_CTRL_UHS_SDR104;
@@ -521,11 +538,31 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
 	 * provide feedback clock, the mode selection can be any value less
 	 * than 3'b011 in bits [2:0] of HOST CONTROL2 register.
 	 */
-	if (host->clock <= 100000000 &&
-	    (uhs == MMC_TIMING_MMC_HS400 ||
-	     uhs == MMC_TIMING_MMC_HS200 ||
-	     uhs == MMC_TIMING_UHS_SDR104))
-		ctrl_2 &= ~SDHCI_CTRL_UHS_MASK;
+	if (host->clock <= CORE_FREQ_100MHZ) {
+		if ((uhs == MMC_TIMING_MMC_HS400) ||
+		    (uhs == MMC_TIMING_MMC_HS200) ||
+		    (uhs == MMC_TIMING_UHS_SDR104))
+			ctrl_2 &= ~SDHCI_CTRL_UHS_MASK;
+		/*
+		 * Make sure DLL is disabled when not required
+		 *
+		 * Write 1 to DLL_RST bit of DLL_CONFIG register
+		 */
+		config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+		config |= CORE_DLL_RST;
+		writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
+
+		/* Write 1 to DLL_PDN bit of DLL_CONFIG register */
+		config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+		config |= CORE_DLL_PDN;
+		writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
+
+		/*
+		 * The DLL needs to be restored and CDCLP533 recalibrated
+		 * when the clock frequency is set back to 400MHz.
+		 */
+		msm_host->calibration_done = false;
+	}
 
 	dev_dbg(mmc_dev(mmc), "%s: clock=%u uhs=%u ctrl_2=0x%x\n",
 		mmc_hostname(host->mmc), host->clock, uhs, ctrl_2);
@@ -638,6 +675,7 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
 	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
 	struct mmc_ios curr_ios = host->mmc->ios;
+	u32 config;
 	int rc;
 
 	if (!clock) {
@@ -656,6 +694,70 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 	    (curr_ios.timing == MMC_TIMING_MMC_DDR52) ||
 	    (curr_ios.timing == MMC_TIMING_MMC_HS400))
 		clock *= 2;
+	/*
+	 * In general all timing modes are controlled via UHS mode select in
+	 * Host Control2 register. eMMC specific HS200/HS400 doesn't have
+	 * their respective modes defined here, hence we use these values.
+	 *
+	 * HS200 - SDR104 (Since they both are equivalent in functionality)
+	 * HS400 - This involves multiple configurations
+	 *		Initially SDR104 - when tuning is required as HS200
+	 *		Then when switching to DDR @ 400MHz (HS400) we use
+	 *		the vendor specific HC_SELECT_IN to control the mode.
+	 *
+	 * In addition to controlling the modes we also need to select the
+	 * correct input clock for DLL depending on the mode.
+	 *
+	 * HS400 - divided clock (free running MCLK/2)
+	 * All other modes - default (free running MCLK)
+	 */
+	if (curr_ios.timing == MMC_TIMING_MMC_HS400) {
+		/* Select the divided clock (free running MCLK/2) */
+		config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
+		config &= ~CORE_HC_MCLK_SEL_MASK;
+		config |= CORE_HC_MCLK_SEL_HS400;
+
+		writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
+		/*
+		 * Select HS400 mode using the HC_SELECT_IN from VENDOR SPEC
+		 * register
+		 */
+		if (msm_host->tuning_done && !msm_host->calibration_done) {
+			/*
+			 * Write 0x6 to HC_SELECT_IN and 1 to HC_SELECT_IN_EN
+			 * field in VENDOR_SPEC_FUNC
+			 */
+			config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
+			config |= CORE_HC_SELECT_IN_HS400;
+			config |= CORE_HC_SELECT_IN_EN;
+			writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
+		}
+	} else {
+		/* Select the default clock (free running MCLK) */
+		config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
+		config &= ~CORE_HC_MCLK_SEL_MASK;
+		config |= CORE_HC_MCLK_SEL_DFLT;
+		writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
+
+		/*
+		 * Disable HC_SELECT_IN to be able to use the UHS mode select
+		 * configuration from Host Control2 register for all other
+		 * modes.
+		 *
+		 * Write 0 to HC_SELECT_IN and HC_SELECT_IN_EN field
+		 * in VENDOR_SPEC_FUNC
+		 */
+		config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
+		config &= ~CORE_HC_SELECT_IN_EN;
+		config &= ~CORE_HC_SELECT_IN_MASK;
+		writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
+	}
+
+	/*
+	 * Make sure above writes impacting free running MCLK are completed
+	 * before changing the clk_rate at GCC.
+	 */
+	wmb();
 
 	if (clock != msm_host->clk_rate) {
 		rc = clk_set_rate(msm_host->clk, clock);
@@ -804,6 +906,11 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 	/* Set HC_MODE_EN bit in HC_MODE register */
 	writel_relaxed(HC_MODE_EN, (msm_host->core_mem + CORE_HC_MODE));
 
+	/* Set FF_CLK_SW_RST_DIS bit in HC_MODE register */
+	config = readl_relaxed(msm_host->core_mem + CORE_HC_MODE);
+	config |= FF_CLK_SW_RST_DIS;
+	writel_relaxed(config, msm_host->core_mem + CORE_HC_MODE);
+
 	host_version = readw_relaxed((host->ioaddr + SDHCI_HOST_VERSION));
 	dev_dbg(&pdev->dev, "Host Version: 0x%x Vendor Version 0x%x\n",
 		host_version, ((host_version & SDHCI_VENDOR_VER_MASK) >>
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.

A message that is not in the archive (sent off-list or not yet received): 1479103248-9491-12-git-send-email-riteshh-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org

#24 lore raw

Re: [PATCH v7 11/14] mmc: sdhci-msm: Add HS400 platform support

reproduce: wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross # save the attached .config to linux build tree…

Hi Venkat,

[auto build test ERROR on ulf.hansson-mmc/next]
[also build test ERROR on v4.9-rc5]
[cannot apply to next-20161114]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Ritesh-Harjani/mmc-sdhci-msm-Add-clk-rates-DDR-HS400-support/20161114-142815
base:   https://git.linaro.org/people/ulf.hansson/mmc next
config: arm-defconfig (attached as .config)
compiler: arm-linux-gnueabi-gcc (Debian 6.1.1-9) 6.1.1 20160705

reproduce:
wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
make.cross ARCH=arm

Note: the linux-review/Ritesh-Harjani/mmc-sdhci-msm-Add-clk-rates-DDR-HS400-support/20161114-142815 HEAD baef00575b049e246cebd910c417f34cada20ee0 builds fine.
It only hurts bisectibility.

All errors (new ones prefixed by >>):

drivers/mmc/host/sdhci-msm.c: In function 'sdhci_msm_execute_tuning':

>> drivers/mmc/host/sdhci-msm.c:498:3: error: 'msm_host' undeclared (first use in this function)
      msm_host->tuning_done = true;
      ^~~~~~~~

drivers/mmc/host/sdhci-msm.c:498:3: note: each undeclared identifier is reported only once for each function it appears in

vim +/msm_host +498 drivers/mmc/host/sdhci-msm.c

492 dev_dbg(mmc_dev(mmc), "%s: No tuning point found\n",
493 mmc_hostname(mmc));
494 rc = -EIO;
495 }
496
497 if (!rc)

 > 498			msm_host->tuning_done = true;

499 return rc;
500 }
501

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation

#25 lore raw

Re: [PATCH v7 11/14] mmc: sdhci-msm: Add HS400 platform support

Hi,

Quoted 6 lines · kbuild test robot

Hi Venkat,

[auto build test ERROR on ulf.hansson-mmc/next]
[also build test ERROR on v4.9-rc5]
[cannot apply to next-20161114]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

Here we go...

Quoted 3 lines
url:    https://github.com/0day-ci/linux/commits/Ritesh-Harjani/mmc-sdhci-msm-Add-clk-rates-DDR-HS400-support/20161114-142815
base:   https://git.linaro.org/people/ulf.hansson/mmc next

This above is the old tree, here's the new:
git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc.git

Could you please update the path in the build system?

[...]

Kind regards
Ulf Hansson

A message that is not in the archive (sent off-list or not yet received): CAPDyKFoUiGgXhLtW9-+iAxdV6sy4+wgQfW8P5+VwsqHc3QwkqA-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#27 lore raw

Re: [PATCH v7 11/14] mmc: sdhci-msm: Add HS400 platform support

Quoted 13 lines · Ulf Hansson

Hi,

Quoted 6 lines · kbuild test robot

Hi Venkat,

[auto build test ERROR on ulf.hansson-mmc/next]
[also build test ERROR on v4.9-rc5]
[cannot apply to next-20161114]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

Here we go...

Quoted 1 line
Quoted 7 lines
Quoted 2 lines
url:    https://github.com/0day-ci/linux/commits/Ritesh-Harjani/mmc-sdhci-msm-Add-clk-rates-DDR-HS400-support/20161114-142815
base:   https://git.linaro.org/people/ulf.hansson/mmc next

This above is the old tree, here's the new:
git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc.git

Could you please update the path in the build system?

OK, done.

Regards,
Fengguang
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#28 lore raw

PATCH v7 12/14 mmc: sdhci-msm: Save the calculated tuning phase

In reply to Ritesh Harjani

Save the tuning phase once the tuning is performed. This phase value will be used while calibrating DLL for HS400 mode.

---
 drivers/mmc/host/sdhci-msm.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
drivers/mmc/host/sdhci-msm.c +10 -0
index 7618bad..95b47aa 100644
@@ -74,6 +74,7 @@
 
 #define CORE_VENDOR_SPEC_CAPABILITIES0	0x11c
 
+#define INVALID_TUNING_PHASE	-1
 #define SDHCI_MSM_MIN_CLOCK	400000
 #define CORE_FREQ_100MHZ	(100 * 1000 * 1000)
 
@@ -95,6 +96,7 @@ struct sdhci_msm_host {
 	bool use_14lpp_dll_reset;
 	bool tuning_done;
 	bool calibration_done;
+	u8 saved_tuning_phase;
 };
 
 /* Platform specific tuning */
@@ -134,6 +136,9 @@ static int msm_config_cm_dll_phase(struct sdhci_host *host, u8 phase)
 	u32 config;
 	struct mmc_host *mmc = host->mmc;
 
+	if (phase > 0xf)
+		return -EINVAL;
+
 	spin_lock_irqsave(&host->lock, flags);
 
 	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
@@ -435,6 +440,8 @@ static int sdhci_msm_execute_tuning(struct sdhci_host *host, u32 opcode)
 	int rc;
 	struct mmc_host *mmc = host->mmc;
 	struct mmc_ios ios = host->mmc->ios;
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
 
 	/*
 	 * Tuning is required for SDR104, HS200 and HS400 cards and
@@ -459,6 +466,7 @@ static int sdhci_msm_execute_tuning(struct sdhci_host *host, u32 opcode)
 		if (rc)
 			return rc;
 
+		msm_host->saved_tuning_phase = phase;
 		rc = mmc_send_tuning(mmc, opcode, NULL);
 		if (!rc) {
 			/* Tuning is successful at this tuning point */
@@ -830,6 +838,8 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 
 	sdhci_get_of_property(pdev);
 
+	msm_host->saved_tuning_phase = INVALID_TUNING_PHASE;
+
 	/* Setup SDCC bus voter clock. */
 	msm_host->bus_clk = devm_clk_get(&pdev->dev, "bus");
 	if (!IS_ERR(msm_host->bus_clk)) {
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#29 lore raw

PATCH v7 13/14 mmc: sdhci-msm: Add calibration tuning for CDCLP533 circuit

In reply to Ritesh Harjani

In HS400 mode a new RCLK is introduced on the interface for read data transfers. The eMMC5.0 device transmits the read data to the host with respect to rising and falling edges of RCLK. In order to ensure correct operation of read data transfers in HS400 mode, the incoming RX data needs to be sampled by delayed version of RCLK.

The CDCLP533 delay circuit shifts the RCLK by T/4. It needs to be initialized, configured and enabled once during HS400 mode switch and when operational voltage/clock is changed.

---
 drivers/mmc/host/sdhci-msm.c | 163 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 163 insertions(+)
drivers/mmc/host/sdhci-msm.c +163 -0
index 95b47aa..9f8f070 100644
@@ -19,6 +19,7 @@
 #include <linux/delay.h>
 #include <linux/mmc/mmc.h>
 #include <linux/slab.h>
+#include <linux/iopoll.h>
 
 #include "sdhci-pltfm.h"
 
@@ -57,6 +58,7 @@
 #define CORE_DLL_PDN		BIT(29)
 #define CORE_DLL_RST		BIT(30)
 #define CORE_DLL_CONFIG		0x100
+#define CORE_CMD_DAT_TRACK_SEL	BIT(0)
 #define CORE_DLL_STATUS		0x108
 
 #define CORE_DLL_CONFIG_2	0x1b4
@@ -72,6 +74,33 @@
 #define CORE_HC_SELECT_IN_HS400	(6 << 19)
 #define CORE_HC_SELECT_IN_MASK	(7 << 19)
 
+#define CORE_CSR_CDC_CTLR_CFG0		0x130
+#define CORE_SW_TRIG_FULL_CALIB		BIT(16)
+#define CORE_HW_AUTOCAL_ENA		BIT(17)
+
+#define CORE_CSR_CDC_CTLR_CFG1		0x134
+#define CORE_CSR_CDC_CAL_TIMER_CFG0	0x138
+#define CORE_TIMER_ENA			BIT(16)
+
+#define CORE_CSR_CDC_CAL_TIMER_CFG1	0x13C
+#define CORE_CSR_CDC_REFCOUNT_CFG	0x140
+#define CORE_CSR_CDC_COARSE_CAL_CFG	0x144
+#define CORE_CDC_OFFSET_CFG		0x14C
+#define CORE_CSR_CDC_DELAY_CFG		0x150
+#define CORE_CDC_SLAVE_DDA_CFG		0x160
+#define CORE_CSR_CDC_STATUS0		0x164
+#define CORE_CALIBRATION_DONE		BIT(0)
+
+#define CORE_CDC_ERROR_CODE_MASK	0x7000000
+
+#define CORE_CSR_CDC_GEN_CFG		0x178
+#define CORE_CDC_SWITCH_BYPASS_OFF	BIT(0)
+#define CORE_CDC_SWITCH_RC_EN		BIT(1)
+
+#define CORE_DDR_200_CFG		0x184
+#define CORE_CDC_T4_DLY_SEL		BIT(0)
+#define CORE_START_CDC_TRAFFIC		BIT(6)
+
 #define CORE_VENDOR_SPEC_CAPABILITIES0	0x11c
 
 #define INVALID_TUNING_PHASE	-1
@@ -433,6 +462,131 @@ static int msm_init_cm_dll(struct sdhci_host *host)
 	return 0;
 }
 
+static int sdhci_msm_cdclp533_calibration(struct sdhci_host *host)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+	u32 config, calib_done;
+	int ret;
+
+	pr_debug("%s: %s: Enter\n", mmc_hostname(host->mmc), __func__);
+
+	/*
+	 * Retuning in HS400 (DDR mode) will fail, just reset the
+	 * tuning block and restore the saved tuning phase.
+	 */
+	ret = msm_init_cm_dll(host);
+	if (ret)
+		goto out;
+
+	/* Set the selected phase in delay line hw block */
+	ret = msm_config_cm_dll_phase(host, msm_host->saved_tuning_phase);
+	if (ret)
+		goto out;
+
+	/* Write 1 to CMD_DAT_TRACK_SEL field in DLL_CONFIG */
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config |= CORE_CMD_DAT_TRACK_SEL;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
+
+	/* Write 0 to CDC_T4_DLY_SEL field in VENDOR_SPEC_DDR200_CFG */
+	config = readl_relaxed(host->ioaddr + CORE_DDR_200_CFG);
+	config &= ~CORE_CDC_T4_DLY_SEL;
+	writel_relaxed(config, host->ioaddr + CORE_DDR_200_CFG);
+
+	/* Write 0 to CDC_SWITCH_BYPASS_OFF field in CORE_CSR_CDC_GEN_CFG */
+	config = readl_relaxed(host->ioaddr + CORE_CSR_CDC_GEN_CFG);
+	config &= ~CORE_CDC_SWITCH_BYPASS_OFF;
+	writel_relaxed(config, host->ioaddr + CORE_CSR_CDC_GEN_CFG);
+
+	/* Write 1 to CDC_SWITCH_RC_EN field in CORE_CSR_CDC_GEN_CFG */
+	config = readl_relaxed(host->ioaddr + CORE_CSR_CDC_GEN_CFG);
+	config |= CORE_CDC_SWITCH_RC_EN;
+	writel_relaxed(config, host->ioaddr + CORE_CSR_CDC_GEN_CFG);
+
+	/* Write 0 to START_CDC_TRAFFIC field in CORE_DDR200_CFG */
+	config = readl_relaxed(host->ioaddr + CORE_DDR_200_CFG);
+	config &= ~CORE_START_CDC_TRAFFIC;
+	writel_relaxed(config, host->ioaddr + CORE_DDR_200_CFG);
+
+	/*
+	 * Perform CDC Register Initialization Sequence
+	 *
+	 * CORE_CSR_CDC_CTLR_CFG0	0x11800EC
+	 * CORE_CSR_CDC_CTLR_CFG1	0x3011111
+	 * CORE_CSR_CDC_CAL_TIMER_CFG0	0x1201000
+	 * CORE_CSR_CDC_CAL_TIMER_CFG1	0x4
+	 * CORE_CSR_CDC_REFCOUNT_CFG	0xCB732020
+	 * CORE_CSR_CDC_COARSE_CAL_CFG	0xB19
+	 * CORE_CSR_CDC_DELAY_CFG	0x3AC
+	 * CORE_CDC_OFFSET_CFG		0x0
+	 * CORE_CDC_SLAVE_DDA_CFG	0x16334
+	 */
+
+	writel_relaxed(0x11800EC, host->ioaddr + CORE_CSR_CDC_CTLR_CFG0);
+	writel_relaxed(0x3011111, host->ioaddr + CORE_CSR_CDC_CTLR_CFG1);
+	writel_relaxed(0x1201000, host->ioaddr + CORE_CSR_CDC_CAL_TIMER_CFG0);
+	writel_relaxed(0x4, host->ioaddr + CORE_CSR_CDC_CAL_TIMER_CFG1);
+	writel_relaxed(0xCB732020, host->ioaddr + CORE_CSR_CDC_REFCOUNT_CFG);
+	writel_relaxed(0xB19, host->ioaddr + CORE_CSR_CDC_COARSE_CAL_CFG);
+	writel_relaxed(0x3AC, host->ioaddr + CORE_CSR_CDC_DELAY_CFG);
+	writel_relaxed(0x0, host->ioaddr + CORE_CDC_OFFSET_CFG);
+	writel_relaxed(0x16334, host->ioaddr + CORE_CDC_SLAVE_DDA_CFG);
+
+	/* CDC HW Calibration */
+
+	/* Write 1 to SW_TRIG_FULL_CALIB field in CORE_CSR_CDC_CTLR_CFG0 */
+	config = readl_relaxed(host->ioaddr + CORE_CSR_CDC_CTLR_CFG0);
+	config |= CORE_SW_TRIG_FULL_CALIB;
+	writel_relaxed(config, host->ioaddr + CORE_CSR_CDC_CTLR_CFG0);
+
+	/* Write 0 to SW_TRIG_FULL_CALIB field in CORE_CSR_CDC_CTLR_CFG0 */
+	config = readl_relaxed(host->ioaddr + CORE_CSR_CDC_CTLR_CFG0);
+	config &= ~CORE_SW_TRIG_FULL_CALIB;
+	writel_relaxed(config, host->ioaddr + CORE_CSR_CDC_CTLR_CFG0);
+
+	/* Write 1 to HW_AUTOCAL_ENA field in CORE_CSR_CDC_CTLR_CFG0 */
+	config = readl_relaxed(host->ioaddr + CORE_CSR_CDC_CTLR_CFG0);
+	config |= CORE_HW_AUTOCAL_ENA;
+	writel_relaxed(config, host->ioaddr + CORE_CSR_CDC_CTLR_CFG0);
+
+	/* Write 1 to TIMER_ENA field in CORE_CSR_CDC_CAL_TIMER_CFG0 */
+	config = readl_relaxed(host->ioaddr + CORE_CSR_CDC_CAL_TIMER_CFG0);
+	config |= CORE_TIMER_ENA;
+	writel_relaxed(config, host->ioaddr + CORE_CSR_CDC_CAL_TIMER_CFG0);
+
+	/* Poll on CALIBRATION_DONE field in CORE_CSR_CDC_STATUS0 to be 1 */
+	ret = readl_relaxed_poll_timeout(host->ioaddr + CORE_CSR_CDC_STATUS0,
+					 calib_done,
+					 (calib_done & CORE_CALIBRATION_DONE),
+					 1, 50);
+
+	if (ret == -ETIMEDOUT) {
+		pr_err("%s: %s: CDC calibration was not completed\n",
+		       mmc_hostname(host->mmc), __func__);
+		goto out;
+	}
+
+	/* Verify CDC_ERROR_CODE field in CORE_CSR_CDC_STATUS0 is 0 */
+	ret = readl_relaxed(host->ioaddr + CORE_CSR_CDC_STATUS0)
+			& CORE_CDC_ERROR_CODE_MASK;
+	if (ret) {
+		pr_err("%s: %s: CDC error code %d\n",
+		       mmc_hostname(host->mmc), __func__, ret);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* Write 1 to START_CDC_TRAFFIC field in CORE_DDR200_CFG */
+	config = readl_relaxed(host->ioaddr + CORE_DDR_200_CFG);
+	config |= CORE_START_CDC_TRAFFIC;
+	writel_relaxed(config, host->ioaddr + CORE_DDR_200_CFG);
+out:
+	pr_debug("%s: %s: Exit, ret %d\n", mmc_hostname(host->mmc),
+		 __func__, ret);
+	return ret;
+}
+
 static int sdhci_msm_execute_tuning(struct sdhci_host *host, u32 opcode)
 {
 	int tuning_seq_cnt = 3;
@@ -575,6 +729,15 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
 	dev_dbg(mmc_dev(mmc), "%s: clock=%u uhs=%u ctrl_2=0x%x\n",
 		mmc_hostname(host->mmc), host->clock, uhs, ctrl_2);
 	sdhci_writew(host, ctrl_2, SDHCI_HOST_CONTROL2);
+
+	spin_unlock_irq(&host->lock);
+	/* CDCLP533 HW calibration is only required for HS400 mode*/
+	if (host->clock > CORE_FREQ_100MHZ &&
+	    msm_host->tuning_done && !msm_host->calibration_done &&
+	    (mmc->ios.timing == MMC_TIMING_MMC_HS400))
+		if (!sdhci_msm_cdclp533_calibration(host))
+			msm_host->calibration_done = true;
+	spin_lock_irq(&host->lock);
 }
 
 static void sdhci_msm_voltage_switch(struct sdhci_host *host)
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#30 lore raw

Re: [PATCH v7 13/14] mmc: sdhci-msm: Add calibration tuning for CDCLP533 circuit

drivers/mmc/host/sdhci-msm.c static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host, · line 737

Quoted 10 lines · Ritesh Harjani
drivers/mmc/host/sdhci-msm.c +6 -0
@@ -575,6 +729,15 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
 	dev_dbg(mmc_dev(mmc), "%s: clock=%u uhs=%u ctrl_2=0x%x\n",
 		mmc_hostname(host->mmc), host->clock, uhs, ctrl_2);
 	sdhci_writew(host, ctrl_2, SDHCI_HOST_CONTROL2);
+
+	spin_unlock_irq(&host->lock);
+	/* CDCLP533 HW calibration is only required for HS400 mode*/
+	if (host->clock > CORE_FREQ_100MHZ &&
+	    msm_host->tuning_done && !msm_host->calibration_done &&
+	    (mmc->ios.timing == MMC_TIMING_MMC_HS400))

Drop useless parenthesis.

Quoted 4 lines
drivers/mmc/host/sdhci-msm.c +3 -0
+		if (!sdhci_msm_cdclp533_calibration(host))
+			msm_host->calibration_done = true;
+	spin_lock_irq(&host->lock);
 }
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#31 lore raw

Re: [PATCH v7 13/14] mmc: sdhci-msm: Add calibration tuning for CDCLP533 circuit

Quoted 13 lines · Stephen Boyd

drivers/mmc/host/sdhci-msm.c static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host, · line 737

Quoted 10 lines · Ritesh Harjani
drivers/mmc/host/sdhci-msm.c +6 -0
@@ -575,6 +729,15 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
 	dev_dbg(mmc_dev(mmc), "%s: clock=%u uhs=%u ctrl_2=0x%x\n",
 		mmc_hostname(host->mmc), host->clock, uhs, ctrl_2);
 	sdhci_writew(host, ctrl_2, SDHCI_HOST_CONTROL2);
+
+	spin_unlock_irq(&host->lock);
+	/* CDCLP533 HW calibration is only required for HS400 mode*/
+	if (host->clock > CORE_FREQ_100MHZ &&
+	    msm_host->tuning_done && !msm_host->calibration_done &&
+	    (mmc->ios.timing == MMC_TIMING_MMC_HS400))

Drop useless parenthesis.

Ok, sure.

Quoted 6 lines
Quoted 5 lines
drivers/mmc/host/sdhci-msm.c +3 -0
+		if (!sdhci_msm_cdclp533_calibration(host))
+			msm_host->calibration_done = true;
+	spin_lock_irq(&host->lock);
 }
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#32 lore raw

PATCH v7 14/14 sdhci: sdhci-msm: update dll configuration

In reply to Ritesh Harjani

The newer msm sdhci's cores use a different DLL hardware for HS400. Update the configuration and calibration of the newer DLL block.

The HS400 DLL block used previously is CDC LP 533 and requires programming multiple registers and waiting for configuration to complete and then enable it. It has about 18 register writes and two register reads.

The newer HS400 DLL block is SDC4 DLL and requires two register writes for configuration and one register read to confirm that it is initialized. There is an additional register write to enable the power save mode for SDC4 DLL block.

---
 drivers/mmc/host/sdhci-msm.c | 132 ++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 130 insertions(+), 2 deletions(-)
drivers/mmc/host/sdhci-msm.c +130 -2
index 9f8f070..4711ee6 100644
@@ -51,6 +51,7 @@
 #define INT_MASK		0xf
 #define MAX_PHASES		16
 #define CORE_DLL_LOCK		BIT(7)
+#define CORE_DDR_DLL_LOCK	BIT(11)
 #define CORE_DLL_EN		BIT(16)
 #define CORE_CDR_EN		BIT(17)
 #define CORE_CK_OUT_EN		BIT(18)
@@ -62,6 +63,7 @@
 #define CORE_DLL_STATUS		0x108
 
 #define CORE_DLL_CONFIG_2	0x1b4
+#define CORE_DDR_CAL_EN		BIT(0)
 #define CORE_FLL_CYCLE_CNT	BIT(18)
 #define CORE_DLL_CLOCK_DISABLE	BIT(21)
 
@@ -100,6 +102,11 @@
 #define CORE_DDR_200_CFG		0x184
 #define CORE_CDC_T4_DLY_SEL		BIT(0)
 #define CORE_START_CDC_TRAFFIC		BIT(6)
+#define CORE_VENDOR_SPEC3	0x1b0
+#define CORE_PWRSAVE_DLL	BIT(3)
+
+#define CORE_DDR_CONFIG		0x1b8
+#define DDR_CONFIG_POR_VAL	0x80040853
 
 #define CORE_VENDOR_SPEC_CAPABILITIES0	0x11c
 
@@ -126,6 +133,7 @@ struct sdhci_msm_host {
 	bool tuning_done;
 	bool calibration_done;
 	u8 saved_tuning_phase;
+	bool use_cdclp533;
 };
 
 /* Platform specific tuning */
@@ -587,6 +595,93 @@ static int sdhci_msm_cdclp533_calibration(struct sdhci_host *host)
 	return ret;
 }
 
+static int sdhci_msm_cm_dll_sdc4_calibration(struct sdhci_host *host)
+{
+	u32 dll_status, config;
+	int ret;
+
+	pr_debug("%s: %s: Enter\n", mmc_hostname(host->mmc), __func__);
+
+	/*
+	 * Currently the CORE_DDR_CONFIG register defaults to desired
+	 * configuration on reset. Currently reprogramming the power on
+	 * reset (POR) value in case it might have been modified by
+	 * bootloaders. In the future, if this changes, then the desired
+	 * values will need to be programmed appropriately.
+	 */
+	writel_relaxed(DDR_CONFIG_POR_VAL, host->ioaddr + CORE_DDR_CONFIG);
+
+	/* Write 1 to DDR_CAL_EN field in CORE_DLL_CONFIG_2 */
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG_2);
+	config |= CORE_DDR_CAL_EN;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG_2);
+
+	/* Poll on DDR_DLL_LOCK bit in CORE_DLL_STATUS to be set */
+	ret = readl_relaxed_poll_timeout(host->ioaddr + CORE_DLL_STATUS,
+					 dll_status,
+					 (dll_status & CORE_DDR_DLL_LOCK),
+					 10, 1000);
+
+	if (ret == -ETIMEDOUT) {
+		pr_err("%s: %s: CM_DLL_SDC4 calibration was not completed\n",
+		       mmc_hostname(host->mmc), __func__);
+		goto out;
+	}
+
+	/* set CORE_PWRSAVE_DLL bit in CORE_VENDOR_SPEC3 */
+	config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC3);
+	config |= CORE_PWRSAVE_DLL;
+	writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC3);
+
+	/*
+	 * Drain writebuffer to ensure above DLL calibration
+	 * and PWRSAVE DLL is enabled.
+	 */
+	wmb();
+out:
+	pr_debug("%s: %s: Exit, ret %d\n", mmc_hostname(host->mmc),
+		 __func__, ret);
+	return ret;
+}
+
+static int sdhci_msm_hs400_dll_calibration(struct sdhci_host *host)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+	int ret;
+	u32 config;
+
+	pr_debug("%s: %s: Enter\n", mmc_hostname(host->mmc), __func__);
+
+	/*
+	 * Retuning in HS400 (DDR mode) will fail, just reset the
+	 * tuning block and restore the saved tuning phase.
+	 */
+	ret = msm_init_cm_dll(host);
+	if (ret)
+		goto out;
+
+	/* Set the selected phase in delay line hw block */
+	ret = msm_config_cm_dll_phase(host, msm_host->saved_tuning_phase);
+	if (ret)
+		goto out;
+
+	/* Write 1 to CMD_DAT_TRACK_SEL field in DLL_CONFIG */
+	config = readl_relaxed(host->ioaddr + CORE_DLL_CONFIG);
+	config |= CORE_CMD_DAT_TRACK_SEL;
+	writel_relaxed(config, host->ioaddr + CORE_DLL_CONFIG);
+	if (msm_host->use_cdclp533)
+		/* Calibrate CDCLP533 DLL HW */
+		ret = sdhci_msm_cdclp533_calibration(host);
+	else
+		/* Calibrate CM_DLL_SDC4 HW */
+		ret = sdhci_msm_cm_dll_sdc4_calibration(host);
+out:
+	pr_debug("%s: %s: Exit, ret %d\n", mmc_hostname(host->mmc),
+		 __func__, ret);
+	return ret;
+}
+
 static int sdhci_msm_execute_tuning(struct sdhci_host *host, u32 opcode)
 {
 	int tuning_seq_cnt = 3;
@@ -735,7 +830,7 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
 	if (host->clock > CORE_FREQ_100MHZ &&
 	    msm_host->tuning_done && !msm_host->calibration_done &&
 	    (mmc->ios.timing == MMC_TIMING_MMC_HS400))
-		if (!sdhci_msm_cdclp533_calibration(host))
+		if (!sdhci_msm_hs400_dll_calibration(host))
 			msm_host->calibration_done = true;
 	spin_lock_irq(&host->lock);
 }
@@ -846,7 +941,7 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
 	struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
 	struct mmc_ios curr_ios = host->mmc->ios;
-	u32 config;
+	u32 config, dll_lock;
 	int rc;
 
 	if (!clock) {
@@ -903,7 +998,33 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 			config |= CORE_HC_SELECT_IN_EN;
 			writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
 		}
+		if (!msm_host->clk_rate && !msm_host->use_cdclp533) {
+			/*
+			 * Poll on DLL_LOCK or DDR_DLL_LOCK bits in
+			 * CORE_DLL_STATUS to be set.  This should get set
+			 * within 15 us at 200 MHz.
+			 */
+			rc = readl_relaxed_poll_timeout(host->ioaddr +
+							CORE_DLL_STATUS,
+							dll_lock,
+							(dll_lock &
+							(CORE_DLL_LOCK |
+							CORE_DDR_DLL_LOCK)), 10,
+							1000);
+			if (rc == -ETIMEDOUT)
+				pr_err("%s: Unable to get DLL_LOCK/DDR_DLL_LOCK, dll_status: 0x%08x\n",
+				       mmc_hostname(host->mmc), dll_lock);
+		}
 	} else {
+		if (!msm_host->use_cdclp533) {
+			/* set CORE_PWRSAVE_DLL bit in CORE_VENDOR_SPEC3 */
+			config = readl_relaxed(host->ioaddr +
+					CORE_VENDOR_SPEC3);
+			config &= ~CORE_PWRSAVE_DLL;
+			writel_relaxed(config, host->ioaddr +
+					CORE_VENDOR_SPEC3);
+		}
+
 		/* Select the default clock (free running MCLK) */
 		config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
 		config &= ~CORE_HC_MCLK_SEL_MASK;
@@ -1100,6 +1221,13 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 		msm_host->use_14lpp_dll_reset = true;
 
 	/*
+	 * SDCC 5 controller with major version 1, minor version 0x34 and later
+	 * with HS 400 mode support will use CM DLL instead of CDC LP 533 DLL.
+	 */
+	if ((core_major == 1) && (core_minor < 0x34))
+		msm_host->use_cdclp533 = true;
+
+	/*
 	 * Support for some capabilities is not advertised by newer
 	 * controller versions and must be explicitly enabled.
 	 */
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
a Linux Foundation Collaborative Project.
#33 lore raw

Re: [PATCH v7 14/14] sdhci: sdhci-msm: update dll configuration

Quoted 14 lines · Ritesh Harjani
drivers/mmc/host/sdhci-msm.c +10 -0
@@ -903,7 +998,33 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 			config |= CORE_HC_SELECT_IN_EN;
 			writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
 		}
+		if (!msm_host->clk_rate && !msm_host->use_cdclp533) {
+			/*
+			 * Poll on DLL_LOCK or DDR_DLL_LOCK bits in
+			 * CORE_DLL_STATUS to be set.  This should get set
+			 * within 15 us at 200 MHz.
+			 */
+			rc = readl_relaxed_poll_timeout(host->ioaddr +
+							CORE_DLL_STATUS,
+							dll_lock,
+							(dll_lock &

drivers/mmc/host/sdhci-msm.c static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock) · line 1020

Quoted 10 lines
drivers/mmc/host/sdhci-msm.c +9 -0
+							(CORE_DLL_LOCK |
+							CORE_DDR_DLL_LOCK)), 10,
+							1000);
+			if (rc == -ETIMEDOUT)
+				pr_err("%s: Unable to get DLL_LOCK/DDR_DLL_LOCK, dll_status: 0x%08x\n",
+				       mmc_hostname(host->mmc), dll_lock);
+		}
 	} else {
+		if (!msm_host->use_cdclp533) {
+			/* set CORE_PWRSAVE_DLL bit in CORE_VENDOR_SPEC3 */

These types of comments are totally useless. The code says exactly what is being done, and the comment is actually wrong in this case. Please remove all these "set/clear bit X in register Y" comments.

Quoted 11 lines
drivers/mmc/host/sdhci-msm.c +7 -0
+			config = readl_relaxed(host->ioaddr +
+					CORE_VENDOR_SPEC3);
+			config &= ~CORE_PWRSAVE_DLL;
+			writel_relaxed(config, host->ioaddr +
+					CORE_VENDOR_SPEC3);
+		}
+
 		/* Select the default clock (free running MCLK) */
 		config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
 		config &= ~CORE_HC_MCLK_SEL_MASK;

drivers/mmc/host/sdhci-msm.c static int sdhci_msm_probe(struct platform_device *pdev) · line 1227

Quoted 7 lines
drivers/mmc/host/sdhci-msm.c +4 -0
@@ -1100,6 +1221,13 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 		msm_host->use_14lpp_dll_reset = true;
 
 	/*
+	 * SDCC 5 controller with major version 1, minor version 0x34 and later
+	 * with HS 400 mode support will use CM DLL instead of CDC LP 533 DLL.
+	 */
+	if ((core_major == 1) && (core_minor < 0x34))

Drop useless parenthesis please.

Quoted 3 lines
drivers/mmc/host/sdhci-msm.c +3 -0
+		msm_host->use_cdclp533 = true;
+
+	/*
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#34 lore raw

Re: [PATCH v7 14/14] sdhci: sdhci-msm: update dll configuration

Quoted 22 lines · Stephen Boyd
Quoted 21 lines · Ritesh Harjani
drivers/mmc/host/sdhci-msm.c +17 -0
@@ -903,7 +998,33 @@ static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock)
 			config |= CORE_HC_SELECT_IN_EN;
 			writel_relaxed(config, host->ioaddr + CORE_VENDOR_SPEC);
 		}
+		if (!msm_host->clk_rate && !msm_host->use_cdclp533) {
+			/*
+			 * Poll on DLL_LOCK or DDR_DLL_LOCK bits in
+			 * CORE_DLL_STATUS to be set.  This should get set
+			 * within 15 us at 200 MHz.
+			 */
+			rc = readl_relaxed_poll_timeout(host->ioaddr +
+							CORE_DLL_STATUS,
+							dll_lock,
+							(dll_lock &
+							(CORE_DLL_LOCK |
+							CORE_DDR_DLL_LOCK)), 10,
+							1000);
+			if (rc == -ETIMEDOUT)
+				pr_err("%s: Unable to get DLL_LOCK/DDR_DLL_LOCK, dll_status: 0x%08x\n",
+				       mmc_hostname(host->mmc), dll_lock);
+		}
Quoted 8 lines

drivers/mmc/host/sdhci-msm.c static void sdhci_msm_set_clock(struct sdhci_host *host, unsigned int clock) · line 1020

Quoted 3 lines
drivers/mmc/host/sdhci-msm.c +2 -0
 	} else {
+		if (!msm_host->use_cdclp533) {
+			/* set CORE_PWRSAVE_DLL bit in CORE_VENDOR_SPEC3 */

These types of comments are totally useless. The code says exactly what is being done, and the comment is actually wrong in this case. Please remove all these "set/clear bit X in register Y" comments.

Ok, done.

Quoted 14 lines
Quoted 14 lines
drivers/mmc/host/sdhci-msm.c +7 -0
+			config = readl_relaxed(host->ioaddr +
+					CORE_VENDOR_SPEC3);
+			config &= ~CORE_PWRSAVE_DLL;
+			writel_relaxed(config, host->ioaddr +
+					CORE_VENDOR_SPEC3);
+		}
+
 		/* Select the default clock (free running MCLK) */
 		config = readl_relaxed(host->ioaddr + CORE_VENDOR_SPEC);
 		config &= ~CORE_HC_MCLK_SEL_MASK;
@@ -1100,6 +1221,13 @@ static int sdhci_msm_probe(struct platform_device *pdev)
 		msm_host->use_14lpp_dll_reset = true;
 
Quoted 7 lines

drivers/mmc/host/sdhci-msm.c static int sdhci_msm_probe(struct platform_device *pdev) · line 1227

Quoted 5 lines
drivers/mmc/host/sdhci-msm.c +4 -0
 	/*
+	 * SDCC 5 controller with major version 1, minor version 0x34 and later
+	 * with HS 400 mode support will use CM DLL instead of CDC LP 533 DLL.
+	 */
+	if ((core_major == 1) && (core_minor < 0x34))

Drop useless parenthesis please.

Done.

Quoted 5 lines
Quoted 4 lines
drivers/mmc/host/sdhci-msm.c +3 -0
+		msm_host->use_cdclp533 = true;
+
+	/*
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
Tested-by on the v7 series
#35 lore raw
In reply to Ritesh Harjani
Quoted 8 lines · Ritesh Harjani

Hi,

This is v7 version of the patch series which adds support for MSM8996. Adds HS400 driver support as well.
These are tested on internal msm8996 & db410c HW.

Below are the changes in v7.

After adding the associated changes to drivers/clk/qcom/gcc-msm8994.c (2/14) this version of the series work fine on my Nexus 5X (msm8992). All partitions have been found and can be mounted as expected. Reading files seems to work as well, I didn't try writing anything as I didn't want to mess up my phone at this point.

So,

-jeremy