Skip to contentSkip to conversation

ARM: shmobile: R-Mobile: Add SYSC restart handler

PATCH New Geert Uytterhoeven 4 participants 11 messages Dec 3, 2014 – Dec 4, 2014

Subsystems (4)
lore
Share & tools
lore

Geert Uytterhoeven

author maintainer
#0 lore raw

cover ARM: shmobile: R-Mobile: Add SYSC restart handler

Hi Simon, Magnus, Arnd,

This patch series extends the PM domain platform driver for the R-Mobile System Controller (SYSC) to register a restart handler, and removes the restart handler hacks on DT-based R-Mobile A1 (r8a7740) and SH-Mobile AG5 (sh73a0) platforms.

Note that this supports DT-based platforms only. Legacy platforms keep on using the restart callback in struct machine_desc.

This was tested on r8a7740/armadillo-multiplatform, sh73a0/kzm9g-reference, and sh73a0/kzm9g-multiplatform.

Arnd: is this what you had in mind in your response to "[GIT PULL] Third Round of Renesas ARM Based Soc Updates for v3.19"
(http://www.spinics.net/lists/arm-kernel/msg379348.html)?

Dependencies:
- The first two patches depend on the series "[PATCH v6 0/7] ARM:
shmobile: R-Mobile: DT PM domain support", as that series adds the SYSC PM domain platform driver for DT-based platforms and adds the SYSC device node to r8a7740.dtsi,
- The last two patches also depend on the series "[PATCH RFC 0/7] ARM:
shmobile: sh73a0: DT PM domain support", as that series adds the SYSC device node to sh73a0.dtsi.

Thanks!

Geert Uytterhoeven (4):
ARM: shmobile: R-Mobile: Add SYSC restart handler
ARM: shmobile: r8a7740: Remove restart callback
ARM: shmobile: sh73a0: Remove restart callback
ARM: shmobile: kzm9g-reference: Remove restart callback

 arch/arm/mach-shmobile/board-kzm9g-reference.c |  8 -----
 arch/arm/mach-shmobile/pm-rmobile.c            | 48 +++++++++++++++++++++++++-
 arch/arm/mach-shmobile/setup-r8a7740.c         |  8 -----
 arch/arm/mach-shmobile/setup-sh73a0.c          |  8 -----
 4 files changed, 47 insertions(+), 25 deletions(-)
Signature · 12 lines
1.9.1

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds
author maintainer
#1 lore raw

PATCH 1/4 ARM: shmobile: R-Mobile: Add SYSC restart handler

Extend the PM domain platform driver for the R-Mobile System Controller (SYSC) to register a restart handler, to be used on various Renesas ARM SoCs (e.g. R-Mobile A1 (r8a7740) and APE6 (r8a73a4), and SH-Mobile AP4 (sh7372) and AG5 (sh73a0)).

Note that this supports DT-based platforms only.

---
 arch/arm/mach-shmobile/pm-rmobile.c | 48 ++++++++++++++++++++++++++++++++++++-
 1 file changed, 47 insertions(+), 1 deletion(-)
arch/arm/mach-shmobile/pm-rmobile.c +47 -1
index f540096c05f6eabd..b82da04c7210a98d 100644
@@ -23,14 +23,22 @@
 #include <linux/slab.h>
 
 #include <asm/io.h>
+#include <asm/system_misc.h>
 
 #include "pm-rmobile.h"
 
-/* SYSC */
+/* SYSC Register Bank 1 */
 #define SPDCR		0x08	/* SYS Power Down Control Register */
 #define SWUCR		0x14	/* SYS Wakeup Control Register */
 #define PSTR		0x80	/* Power Status Register */
 
+/* SYSC Register Bank 2 */
+#define RESCNT2		0x20	/* Reset Control Register 2 */
+
+/* Reset Control Register 2 */
+#define RESCNT2_PRES	0x80000000	/* Soft power-on reset */
+
+
 #define PSTR_RETRIES	100
 #define PSTR_DELAY_US	10
 
@@ -400,3 +408,41 @@ static int __init rmobile_init_pm_domains(void)
 core_initcall(rmobile_init_pm_domains);
 
 #endif /* !CONFIG_ARCH_SHMOBILE_LEGACY */
+
+
+static void __iomem *sysc_base2;
+
+static void rmobile_sysc_restart(enum reboot_mode reboot_mode, const char *cmd)
+{
+	pr_debug("%s %u/%s\n", __func__, reboot_mode, cmd);
+
+	writel(RESCNT2_PRES, sysc_base2 + RESCNT2);
+}
+
+static int rmobile_sysc_probe(struct platform_device *pdev)
+{
+	dev_dbg(&pdev->dev, "%s\n", __func__);
+
+	sysc_base2 = of_iomap(pdev->dev.of_node, 1);
+	if (!sysc_base2)
+		return -ENODEV;
+
+	arm_pm_restart = rmobile_sysc_restart;
+
+	return 0;
+}
+
+static const struct of_device_id rmobile_sysc_of_match[] = {
+	{ .compatible = "renesas,sysc-rmobile", },
+	{ /* sentinel */ }
+};
+
+static struct platform_driver rmobile_sysc_driver = {
+	.probe = rmobile_sysc_probe,
+	.driver = {
+		.name = "rmobile_sysc",
+		.of_match_table = rmobile_sysc_of_match,
+	},
+};
+
+module_platform_driver(rmobile_sysc_driver);
Signature · 1 line
1.9.1
#2 lore raw

PATCH 1/4 ARM: shmobile: R-Mobile: Add SYSC restart handler

Quoted 8 lines · On Wednesday 03 December 2014 15:17:04 Geert Uytterhoeven wrote:

Extend the PM domain platform driver for the R-Mobile System Controller (SYSC) to register a restart handler, to be used on various Renesas ARM SoCs (e.g. R-Mobile A1 (r8a7740) and APE6 (r8a73a4), and SH-Mobile AP4 (sh7372) and AG5 (sh73a0)).

Note that this supports DT-based platforms only.

Is this one of the typical system controllers that contains lots of random registers? Maybe you could mark it as "syscon" in DT and just use the existing drivers/power/reset/syscon-reboot.c driver?

Quoted 26 lines
arch/arm/mach-shmobile/pm-rmobile.c +26 -0
+static void __iomem *sysc_base2;
+
+static void rmobile_sysc_restart(enum reboot_mode reboot_mode, const char *cmd)
+{
+	pr_debug("%s %u/%s\n", __func__, reboot_mode, cmd);
+
+	writel(RESCNT2_PRES, sysc_base2 + RESCNT2);
+}
+
+static int rmobile_sysc_probe(struct platform_device *pdev)
+{
+	dev_dbg(&pdev->dev, "%s\n", __func__);
+
+	sysc_base2 = of_iomap(pdev->dev.of_node, 1);
+	if (!sysc_base2)
+		return -ENODEV;
+
+	arm_pm_restart = rmobile_sysc_restart;
+
+	return 0;
+}
+
+static const struct of_device_id rmobile_sysc_of_match[] = {
+	{ .compatible = "renesas,sysc-rmobile", },
+	{ /* sentinel */ }
+};

arch/arm/mach-shmobile/pm-rmobile.c static int __init rmobile_init_pm_domains(void) · line 448

Quoted 10 lines
arch/arm/mach-shmobile/pm-rmobile.c +10 -0
+
+static struct platform_driver rmobile_sysc_driver = {
+	.probe = rmobile_sysc_probe,
+	.driver = {
+		.name = "rmobile_sysc",
+		.of_match_table = rmobile_sysc_of_match,
+	},
+};
+
+module_platform_driver(rmobile_sysc_driver);

This clearly looks much better than the previous implementation, so that is great. If you can decouple it from the rest of this file, either through syscon or by asserting that sysc_base2 is not going to be needed for anything else, how about moving the new code to a standalon driver in drivers/power/reset?

Arnd

#3 lore raw

PATCH 1/4 ARM: shmobile: R-Mobile: Add SYSC restart handler

Hi Arnd,

Quoted 6 lines · Arnd Bergmann
Quoted 5 lines · On Wednesday 03 December 2014 15:17:04 Geert Uytterhoeven wrote:

Extend the PM domain platform driver for the R-Mobile System Controller (SYSC) to register a restart handler, to be used on various Renesas ARM SoCs (e.g. R-Mobile A1 (r8a7740) and APE6 (r8a73a4), and SH-Mobile AP4 (sh7372) and AG5 (sh73a0)).

Quoted 7 lines
Quoted 3 lines

Note that this supports DT-based platforms only.

Is this one of the typical system controllers that contains lots of random registers? Maybe you could mark it as "syscon" in DT and just use the existing drivers/power/reset/syscon-reboot.c driver?

The hardware block has two register ranges.
The first one handles boot mode and power management,
the second handles reset control and allows to read the mode strap pins.

The PM domain driver uses the first range, the reset handler uses the second range.

I tried using syscon and syscon-reboot, but that needs some changes to the syscon framework, as it currently provides access to the first register range only. Providing access to the second region requires either:

1. Extending the regmap to cover all ranges.
Does regmap support multiple ranges? It seems regmap_range_cfg handles only virtual ranges (switchable by writing to a register), not always-mapped discontiguous ranges?

2. Providing multiple regmaps, one for each register range. This needs modifications to look up ranges using a phandle and an (optional for backwards compatibility) index, and updates to drivers and bindings.

Am I missing an option?

Quoted 34 lines
Quoted 34 lines
arch/arm/mach-shmobile/pm-rmobile.c +34 -0
+static void __iomem *sysc_base2;
+
+static void rmobile_sysc_restart(enum reboot_mode reboot_mode, const char *cmd)
+{
+     pr_debug("%s %u/%s\n", __func__, reboot_mode, cmd);
+
+     writel(RESCNT2_PRES, sysc_base2 + RESCNT2);
+}
+
+static int rmobile_sysc_probe(struct platform_device *pdev)
+{
+     dev_dbg(&pdev->dev, "%s\n", __func__);
+
+     sysc_base2 = of_iomap(pdev->dev.of_node, 1);
+     if (!sysc_base2)
+             return -ENODEV;
+
+     arm_pm_restart = rmobile_sysc_restart;
+
+     return 0;
+}
+
+static const struct of_device_id rmobile_sysc_of_match[] = {
+     { .compatible = "renesas,sysc-rmobile", },
+     { /* sentinel */ }
+};
+
+static struct platform_driver rmobile_sysc_driver = {
+     .probe = rmobile_sysc_probe,
+     .driver = {
+             .name = "rmobile_sysc",
+             .of_match_table = rmobile_sysc_of_match,
+     },
+};
Quoted 8 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+module_platform_driver(rmobile_sysc_driver);

This clearly looks much better than the previous implementation, so that is great. If you can decouple it from the rest of this file, either through syscon or by asserting that sysc_base2 is not going to be needed for anything else, how about moving the new code to a standalon driver in drivers/power/reset?

Yes, sysc_base2 is used for reset control only (I don't think we need to read the mode strap pins).

So unless there's a simple solution to the regmap problem, I think moving this driver part to drivers/power/reset is the most viable option.

Anyway, thanks for the syscon hint! I think it'll be useful for reset on R-Car Gen2.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

#4 lore raw

PATCH 1/4 ARM: shmobile: R-Mobile: Add SYSC restart handler

Hi Arnd,

Quoted 11 lines · Geert Uytterhoeven
Quoted 5 lines

This clearly looks much better than the previous implementation, so that is great. If you can decouple it from the rest of this file, either through syscon or by asserting that sysc_base2 is not going to be needed for anything else, how about moving the new code to a standalon driver in drivers/power/reset?

Yes, sysc_base2 is used for reset control only (I don't think we need to read the mode strap pins).

So unless there's a simple solution to the regmap problem, I think moving this driver part to drivers/power/reset is the most viable option.

There's another reasone for having a separate driver: if both the SH core and the ARM core are in use, the second register bank can only be accessed after acquiring the HPB (Peripheral Bus Bridge) semaphore.
While we don't actively support running anything on the SH core, we don't want to prevent users from doing that.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#5 lore raw

PATCH 2/4 ARM: shmobile: r8a7740: Remove restart callback

In reply to Geert Uytterhoeven

Remove the restart handling hack from the r8a7740 generic multiplatform case.

Restart on DT-based r8a7740 platforms is now handled through the restart handler registered by the R-Mobile System Controller (SYSC) driver.

This reverts commit 1174c712afa2779f ("ARM: shmobile: r8a7740: Add restart callback").

---
 arch/arm/mach-shmobile/setup-r8a7740.c | 8 --------
 1 file changed, 8 deletions(-)
arch/arm/mach-shmobile/setup-r8a7740.c +0 -8
index 79ad93dfdae4ee70..cce7cdab0524d4a0 100644
@@ -835,13 +835,6 @@ static void __init r8a7740_generic_init(void)
 	of_platform_populate(NULL, of_default_bus_match_table, NULL, NULL);
 }
 
-#define RESCNT2 IOMEM(0xe6188020)
-static void r8a7740_restart(enum reboot_mode mode, const char *cmd)
-{
-	/* Do soft power on reset */
-	writel(1 << 31, RESCNT2);
-}
-
 static const char *r8a7740_boards_compat_dt[] __initdata = {
 	"renesas,r8a7740",
 	NULL,
@@ -854,7 +847,6 @@ DT_MACHINE_START(R8A7740_DT, "Generic R8A7740 (Flattened Device Tree)")
 	.init_machine	= r8a7740_generic_init,
 	.init_late	= shmobile_init_late,
 	.dt_compat	= r8a7740_boards_compat_dt,
-	.restart	= r8a7740_restart,
 MACHINE_END
 
 #endif /* CONFIG_USE_OF */
Signature · 1 line
1.9.1
author maintainer
#6 lore raw

PATCH 3/4 ARM: shmobile: sh73a0: Remove restart callback

In reply to Geert Uytterhoeven

Remove the restart handling hack from the sh73a0 generic multiplatform case.

Restart on DT-based sh73a0 platforms is now handled through the restart handler registered by the R-Mobile System Controller (SYSC) driver.

This reverts commit cad900819fba0176 ("ARM: shmobile: sh73a0: Add restart callback").

---
 arch/arm/mach-shmobile/setup-sh73a0.c | 8 --------
 1 file changed, 8 deletions(-)
arch/arm/mach-shmobile/setup-sh73a0.c +0 -8
index 2ec702c51dd25050..9e5e5ebbd1af641d 100644
@@ -818,13 +818,6 @@ void __init sh73a0_add_standard_devices_dt(void)
 	of_platform_populate(NULL, of_default_bus_match_table, NULL, NULL);
 }
 
-#define RESCNT2 IOMEM(0xe6188020)
-static void sh73a0_restart(enum reboot_mode mode, const char *cmd)
-{
-	/* Do soft power on reset */
-	writel((1 << 31), RESCNT2);
-}
-
 static const char *sh73a0_boards_compat_dt[] __initdata = {
 	"renesas,sh73a0",
 	NULL,
@@ -836,7 +829,6 @@ DT_MACHINE_START(SH73A0_DT, "Generic SH73A0 (Flattened Device Tree)")
 	.init_early	= shmobile_init_delay,
 	.init_machine	= sh73a0_add_standard_devices_dt,
 	.init_late	= shmobile_init_late,
-	.restart	= sh73a0_restart,
 	.dt_compat	= sh73a0_boards_compat_dt,
 MACHINE_END
 #endif /* CONFIG_USE_OF */
Signature · 1 line
1.9.1
author maintainer
#7 lore raw

PATCH 4/4 ARM: shmobile: kzm9g-reference: Remove restart callback

In reply to Geert Uytterhoeven

Remove the restart handling hack from the kzm9g-reference case.

Restart on DT-based sh73a0 platforms is now handled through the restart handler registered by the R-Mobile System Controller (SYSC) driver.

This reverts commit 534547c036afad3f ("ARM: shmobile: kzm9g-reference:
Add restart callback").

---
 arch/arm/mach-shmobile/board-kzm9g-reference.c | 8 --------
 1 file changed, 8 deletions(-)
arch/arm/mach-shmobile/board-kzm9g-reference.c +0 -8
index 2e82e44ab85258b7..f2ef759b6e969cc8 100644
@@ -39,13 +39,6 @@ static void __init kzm_init(void)
 #endif
 }
 
-#define RESCNT2 IOMEM(0xe6188020)
-static void kzm9g_restart(enum reboot_mode mode, const char *cmd)
-{
-	/* Do soft power on reset */
-	writel((1 << 31), RESCNT2);
-}
-
 static const char *kzm9g_boards_compat_dt[] __initdata = {
 	"renesas,kzm9g-reference",
 	NULL,
@@ -57,6 +50,5 @@ DT_MACHINE_START(KZM9G_DT, "kzm9g-reference")
 	.init_early	= shmobile_init_delay,
 	.init_machine	= kzm_init,
 	.init_late	= shmobile_init_late,
-	.restart	= kzm9g_restart,
 	.dt_compat	= kzm9g_boards_compat_dt,
 MACHINE_END
Signature · 1 line
1.9.1
#8 lore raw

cover ARM: shmobile: R-Mobile: Add SYSC restart handler

In reply to Geert Uytterhoeven
Quoted 17 lines · Geert Uytterhoeven

Hi Simon, Magnus, Arnd,

This patch series extends the PM domain platform driver for the R-Mobile System Controller (SYSC) to register a restart handler, and removes the restart handler hacks on DT-based R-Mobile A1 (r8a7740) and SH-Mobile AG5 (sh73a0) platforms.

Note that this supports DT-based platforms only. Legacy platforms keep on using the restart callback in struct machine_desc.

This was tested on r8a7740/armadillo-multiplatform, sh73a0/kzm9g-reference, and sh73a0/kzm9g-multiplatform.

Arnd: is this what you had in mind in your response to "[GIT PULL] Third Round of Renesas ARM Based Soc Updates for v3.19"
(http://www.spinics.net/lists/arm-kernel/msg379348.html)?

Quoted 5 lines

Dependencies:
- The first two patches depend on the series "[PATCH v6 0/7] ARM:
shmobile: R-Mobile: DT PM domain support", as that series adds the SYSC PM domain platform driver for DT-based platforms and adds the SYSC device node to r8a7740.dtsi,

As I have queued up that series I believe I could also queue up the first two patches of this series. Is that worthwhile?

Quoted 32 lines

- The last two patches also depend on the series "[PATCH RFC 0/7] ARM:
shmobile: sh73a0: DT PM domain support", as that series adds the SYSC device node to sh73a0.dtsi.

Thanks!

Geert Uytterhoeven (4):
ARM: shmobile: R-Mobile: Add SYSC restart handler
ARM: shmobile: r8a7740: Remove restart callback
ARM: shmobile: sh73a0: Remove restart callback
ARM: shmobile: kzm9g-reference: Remove restart callback

 arch/arm/mach-shmobile/board-kzm9g-reference.c |  8 -----
 arch/arm/mach-shmobile/pm-rmobile.c            | 48 +++++++++++++++++++++++++-
 arch/arm/mach-shmobile/setup-r8a7740.c         |  8 -----
 arch/arm/mach-shmobile/setup-sh73a0.c          |  8 -----
 4 files changed, 47 insertions(+), 25 deletions(-)
Signature · 12 lines
1.9.1

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds
#9 lore raw

cover ARM: shmobile: R-Mobile: Add SYSC restart handler

Hi Simon,

Quoted 16 lines · Simon Horman
Quoted 15 lines · Geert Uytterhoeven

This patch series extends the PM domain platform driver for the R-Mobile System Controller (SYSC) to register a restart handler, and removes the restart handler hacks on DT-based R-Mobile A1 (r8a7740) and SH-Mobile AG5 (sh73a0) platforms.

Note that this supports DT-based platforms only. Legacy platforms keep on using the restart callback in struct machine_desc.

This was tested on r8a7740/armadillo-multiplatform, sh73a0/kzm9g-reference, and sh73a0/kzm9g-multiplatform.

Arnd: is this what you had in mind in your response to "[GIT PULL] Third Round of Renesas ARM Based Soc Updates for v3.19"
(http://www.spinics.net/lists/arm-kernel/msg379348.html)?

Quoted 8 lines
Quoted 5 lines

Dependencies:
- The first two patches depend on the series "[PATCH v6 0/7] ARM:
shmobile: R-Mobile: DT PM domain support", as that series adds the SYSC PM domain platform driver for DT-based platforms and adds the SYSC device node to r8a7740.dtsi,

As I have queued up that series I believe I could also queue up the first two patches of this series. Is that worthwhile?

Yes you can, if Arnd doesn't object against the driver.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

#10 lore raw

cover ARM: shmobile: R-Mobile: Add SYSC restart handler

Quoted 25 lines · Geert Uytterhoeven

Hi Simon,

Quoted 22 lines · Simon Horman
Quoted 20 lines · Geert Uytterhoeven

This patch series extends the PM domain platform driver for the R-Mobile System Controller (SYSC) to register a restart handler, and removes the restart handler hacks on DT-based R-Mobile A1 (r8a7740) and SH-Mobile AG5 (sh73a0) platforms.

Note that this supports DT-based platforms only. Legacy platforms keep on using the restart callback in struct machine_desc.

This was tested on r8a7740/armadillo-multiplatform, sh73a0/kzm9g-reference, and sh73a0/kzm9g-multiplatform.

Arnd: is this what you had in mind in your response to "[GIT PULL] Third Round of Renesas ARM Based Soc Updates for v3.19"
(http://www.spinics.net/lists/arm-kernel/msg379348.html)?

Dependencies:
- The first two patches depend on the series "[PATCH v6 0/7] ARM:
shmobile: R-Mobile: DT PM domain support", as that series adds the SYSC PM domain platform driver for DT-based platforms and adds the SYSC device node to r8a7740.dtsi,

Quoted 4 lines
Quoted 2 lines

As I have queued up that series I believe I could also queue up the first two patches of this series. Is that worthwhile?

Yes you can, if Arnd doesn't object against the driver.

Thanks, I'll wait for a resolution to that discussion.