ARM: rockchip: some soc-level fixes for 4.1

GIT PULL New Heiko Stübner 3 participants 7 messages Apr 25, 2015 – May 13, 2015 1 message not in the archive

Subsystems (3)

Arm (32-bit) Arm sub-architectures Arm Rockchip

lore

Share & tools

lore

Heiko Stübner

author maintainer

2015-04-25 00:11 UTC #0 lore raw

Hi Arnd, Olof,

I've collected some fixes which might be appropriate for 4.1-rc, so if they're ok, please pull.

Thanks
Heiko

The following changes since commit a0307d186f5015ee3d77e209e5c06190cbf02478:

ARM: rockchip: disable watchdog during suspend (2015-03-11 22:41:02 +0100)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip.git tags/v4.1-rockchip-socfixes1

for you to fetch changes up to 2a9fe3ca84afff6259820c4f62e579f41476becc:

rockchip: make sure timer7 is enabled on rk3288 platforms (2015-04-16 21:22:17 +0200)

----------------------------------------------------------------
Two fixes from Chris Zhong, fixing some suspend oddities. And I've given up on the timer7 issue. While I initially thought devices would either have both the grave mmu issue requiring a uboot update and the timer7 issue or none, it looks like in all units in the field the mmu issue got fixed while the timer7 issue stayed on. So instead of making everybody wanting to use mainline jump through a hoop just make sure timer7 is on on boot before we init the arch-timer.

----------------------------------------------------------------
Chris Zhong (2):
ARM: rockchip: disable dapswjdp during suspend
ARM: rockchip: fix undefined instruction of reset_ctrl_regs

Heiko Stuebner (1):
rockchip: make sure timer7 is enabled on rk3288 platforms

 arch/arm/mach-rockchip/pm.c       | 33 +++++++++++++++++++++++++++++++++
 arch/arm/mach-rockchip/pm.h       |  8 ++++++++
 arch/arm/mach-rockchip/rockchip.c | 19 +++++++++++++++++++
 3 files changed, 60 insertions(+)

Heiko Stübner

author maintainer

2015-04-25 00:13 UTC #1 lore raw

PATCH 1/2 ARM: rockchip: fix undefined instruction of reset_ctrl_regs

Sometimes the debug module may not work well after resume, since it has not been correctly reset when wakeup from suspend. That cause system crash during reusme, and a 'undefined instruction' is displayed on the console. Set the GRF_FORCE_JTAG bit of RK3288_GRF_SOC_CON0 can ensure that debug modul is reset. And we can change the value of RK3288_GRF_SOC_CON0 back when system resume.

Signed-off-by Chris Zhong <[email protected]>

Tested-by Caesar Wang <[email protected]>

Reviewed-by Douglas Anderson <[email protected]>

According to discussions, there does not seem a better solution available. Please also see the potential security implication described in the comment inline in the code.

Signed-off-by Heiko Stuebner <[email protected]>

---
 arch/arm/mach-rockchip/pm.c | 26 ++++++++++++++++++++++++++
 arch/arm/mach-rockchip/pm.h |  4 ++++
 2 files changed, 30 insertions(+)

arch/arm/mach-rockchip/pm.c +26 -0

index b0dcbe2..22812fe 100644
@@ -44,9 +44,11 @@ static void __iomem *rk3288_bootram_base;
 static phys_addr_t rk3288_bootram_phy;
 
 static struct regmap *pmu_regmap;
+static struct regmap *grf_regmap;
 static struct regmap *sgrf_regmap;
 
 static u32 rk3288_pmu_pwr_mode_con;
+static u32 rk3288_grf_soc_con0;
 static u32 rk3288_sgrf_soc_con0;
 
 static inline u32 rk3288_l2_config(void)

@@ -70,12 +72,26 @@ static void rk3288_slp_mode_set(int level)
 {
 	u32 mode_set, mode_set1;
 
+	regmap_read(grf_regmap, RK3288_GRF_SOC_CON0, &rk3288_grf_soc_con0);
+
 	regmap_read(sgrf_regmap, RK3288_SGRF_SOC_CON0, &rk3288_sgrf_soc_con0);
 
 	regmap_read(pmu_regmap, RK3288_PMU_PWRMODE_CON,
 		    &rk3288_pmu_pwr_mode_con);
 
 	/*
+	 * We need set this bit GRF_FORCE_JTAG here, for the debug module,
+	 * otherwise, it may become inaccessible after resume.
+	 * This creates a potential security issue, as the sdmmc pins may
+	 * accept jtag data for a short time during resume if no card is
+	 * inserted.
+	 * But this is of course also true for the regular boot, before we
+	 * turn of the jtag/sdmmc autodetect.
+	 */
+	regmap_write(grf_regmap, RK3288_GRF_SOC_CON0, GRF_FORCE_JTAG |
+		     GRF_FORCE_JTAG_WRITE);
+
+	/*
 	 * SGRF_FAST_BOOT_EN - system to boot from FAST_BOOT_ADDR
 	 * PCLK_WDT_GATE - disable WDT during suspend.
 	 */

@@ -135,6 +151,9 @@ static void rk3288_slp_mode_set_resume(void)
 	regmap_write(sgrf_regmap, RK3288_SGRF_SOC_CON0,
 		     rk3288_sgrf_soc_con0 | SGRF_PCLK_WDT_GATE_WRITE
 		     | SGRF_FAST_BOOT_EN_WRITE);
+
+	regmap_write(grf_regmap, RK3288_GRF_SOC_CON0, rk3288_grf_soc_con0 |
+		     GRF_FORCE_JTAG_WRITE);
 }
 
 static int rockchip_lpmode_enter(unsigned long arg)

@@ -193,6 +212,13 @@ static int rk3288_suspend_init(struct device_node *np)
 		return PTR_ERR(pmu_regmap);
 	}
 
+	grf_regmap = syscon_regmap_lookup_by_compatible(
+				"rockchip,rk3288-grf");
+	if (IS_ERR(grf_regmap)) {
+		pr_err("%s: could not find grf regmap\n", __func__);
+		return PTR_ERR(pmu_regmap);
+	}
+
 	sram_np = of_find_compatible_node(NULL, NULL,
 					  "rockchip,rk3288-pmu-sram");
 	if (!sram_np) {

arch/arm/mach-rockchip/pm.h +4 -0

index 3e8d39c..f8a747b 100644
@@ -48,6 +48,10 @@ static inline void rockchip_suspend_init(void)
 #define RK3288_PMU_WAKEUP_RST_CLR_CNT	0x44
 #define RK3288_PMU_PWRMODE_CON1		0x90
 
+#define RK3288_GRF_SOC_CON0		0x244
+#define GRF_FORCE_JTAG			BIT(12)
+#define GRF_FORCE_JTAG_WRITE		BIT(28)
+
 #define RK3288_SGRF_SOC_CON0		(0x0000)
 #define RK3288_SGRF_FAST_BOOT_ADDR	(0x0120)
 #define SGRF_PCLK_WDT_GATE		BIT(6)

Signature · 1 line

2.1.4

Heiko Stübner

author maintainer

2015-04-25 00:13 UTC #2 lore raw

PATCH 2/2 rockchip: make sure timer7 is enabled on rk3288 platforms

timer7 supplies the architected timer and thus as has to run when the system clocksource and clockevents drivers are registered.

While it should be the responsibility of the bootloader to do this, and there exists a fix in a community u-boot, all u-boot based systems that actually shipped have the mentioned issue.

Therefore to not require every developer to update their u-boot, add a snippet for this, enabling the timer early in the kernel.

Signed-off-by Heiko Stuebner <[email protected]>

---
 arch/arm/mach-rockchip/rockchip.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

arch/arm/mach-rockchip/rockchip.c +19 -0

index d360ec0..b6cf3b4 100644
@@ -30,11 +30,30 @@
 #include "pm.h"
 
 #define RK3288_GRF_SOC_CON0 0x244
+#define RK3288_TIMER6_7_PHYS 0xff810000
 
 static void __init rockchip_timer_init(void)
 {
 	if (of_machine_is_compatible("rockchip,rk3288")) {
 		struct regmap *grf;
+		void __iomem *reg_base;
+
+		/*
+		 * Most/all uboot versions for rk3288 don't enable timer7
+		 * which is needed for the architected timer to work.
+		 * So make sure it is running during early boot.
+		 */
+		reg_base = ioremap(RK3288_TIMER6_7_PHYS, SZ_16K);
+		if (reg_base) {
+			writel(0, reg_base + 0x30);
+			writel(0xffffffff, reg_base + 0x20);
+			writel(0xffffffff, reg_base + 0x24);
+			writel(1, reg_base + 0x30);
+			dsb();
+			iounmap(reg_base);
+		} else {
+			pr_err("rockchip: could not map timer7 registers\n");
+		}
 
 		/*
 		 * Disable auto jtag/sdmmc switching that causes issues

Signature · 1 line

2.1.4

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

Heiko Stuebner

author maintainer

2015-05-12 22:48 UTC #4 lore raw

PATCH 1/2 ARM: rockchip: fix undefined instruction of reset_ctrl_regs

Hi Chris,

Am Dienstag, 12. Mai 2015, 12:00:00 schrieben Sie:

Quoted 4 lines

We do not need this patch, since we have this one:
<https://patchwork.kernel.org/patch/6218791/>
these 2 patches fixed a same issue.

thanks for pointing that out - looks like I didn't need to resurrect the old patch then. Just to make sure I don't mess up when removing the obsolete one:

both patches
b403125 ARM: rockchip: fix undefined instruction of reset_ctrl_regs 0ea001d ARM: rockchip: disable dapswjdp during suspend
actually fix the same issue and
b403125 ARM: rockchip: fix undefined instruction of reset_ctrl_regs
is the one that should be removed, right?

Thanks
Heiko

[I've readded the lists to give this a bit publicity, so people can see why we remove one of the patches again]

Quoted 47 lines · Heiko St?bner

Quoted 46 lines
Sometimes the debug module may not work well after resume, since it has not been correctly reset when wakeup from suspend. That cause system crash during reusme, and a 'undefined instruction' is displayed on the console. Set the GRF_FORCE_JTAG bit of RK3288_GRF_SOC_CON0 can ensure that debug modul is reset. And we can change the value of RK3288_GRF_SOC_CON0 back when system resume.
Signed-off-by Chris Zhong <[email protected]>
Tested-by Caesar Wang <[email protected]>
Reviewed-by Douglas Anderson <[email protected]>
According to discussions, there does not seem a better solution available. Please also see the potential security implication described in the comment inline in the code.
Signed-off-by Heiko Stuebner <[email protected]>
---
  arch/arm/mach-rockchip/pm.c | 26 ++++++++++++++++++++++++++
  arch/arm/mach-rockchip/pm.h |  4 ++++
  2 files changed, 30 insertions(+)
arch/arm/mach-rockchip/pm.c +2 -0
index b0dcbe2..22812fe 100644
@@ -44,9 +44,11 @@ static void __iomem *rk3288_bootram_base;
 
  static phys_addr_t rk3288_bootram_phy;
 
  static struct regmap *pmu_regmap;
 
+static struct regmap *grf_regmap;
 
  static struct regmap *sgrf_regmap;
 
  static u32 rk3288_pmu_pwr_mode_con;
 
+static u32 rk3288_grf_soc_con0;
 
  static u32 rk3288_sgrf_soc_con0;
 
  static inline u32 rk3288_l2_config(void)
@@ -70,12 +72,26 @@ static void rk3288_slp_mode_set(int level)
 

Quoted 8 lines

arch/arm/mach-rockchip/pm.c static void rk3288_slp_mode_set(int level)
Quoted 8 lines
arch/arm/mach-rockchip/pm.c +2 -0
  {
 
  	u32 mode_set, mode_set1;
 
+	regmap_read(grf_regmap, RK3288_GRF_SOC_CON0, &rk3288_grf_soc_con0);
+
 
  	regmap_read(sgrf_regmap, RK3288_SGRF_SOC_CON0, 

&rk3288_sgrf_soc_con0);

Quoted 78 lines

arch/arm/mach-rockchip/pm.c +22 -0

  	regmap_read(pmu_regmap, RK3288_PMU_PWRMODE_CON,
 
  		    &rk3288_pmu_pwr_mode_con);
 
  	/*
 
+	 * We need set this bit GRF_FORCE_JTAG here, for the debug module,
+	 * otherwise, it may become inaccessible after resume.
+	 * This creates a potential security issue, as the sdmmc pins may
+	 * accept jtag data for a short time during resume if no card is
+	 * inserted.
+	 * But this is of course also true for the regular boot, before we
+	 * turn of the jtag/sdmmc autodetect.
+	 */
+	regmap_write(grf_regmap, RK3288_GRF_SOC_CON0, GRF_FORCE_JTAG |
+		     GRF_FORCE_JTAG_WRITE);
+
+	/*
 
  	 * SGRF_FAST_BOOT_EN - system to boot from FAST_BOOT_ADDR
  	 * PCLK_WDT_GATE - disable WDT during suspend.
  	 */

@@ -135,6 +151,9 @@ static void rk3288_slp_mode_set_resume(void)
 
  	regmap_write(sgrf_regmap, RK3288_SGRF_SOC_CON0,
 
  		     rk3288_sgrf_soc_con0 | SGRF_PCLK_WDT_GATE_WRITE
 
  		     | SGRF_FAST_BOOT_EN_WRITE);
 
+
+	regmap_write(grf_regmap, RK3288_GRF_SOC_CON0, rk3288_grf_soc_con0 |
+		     GRF_FORCE_JTAG_WRITE);
 
  }
 
  static int rockchip_lpmode_enter(unsigned long arg)

@@ -193,6 +212,13 @@ static int rk3288_suspend_init(struct device_node
 *np)
 
  		return PTR_ERR(pmu_regmap);
 
  	}
 
+	grf_regmap = syscon_regmap_lookup_by_compatible(
+				"rockchip,rk3288-grf");
+	if (IS_ERR(grf_regmap)) {
+		pr_err("%s: could not find grf regmap\n", __func__);
+		return PTR_ERR(pmu_regmap);
+	}
+

sram_np = of_find_compatible_node(NULL, NULL,

"rockchip,rk3288-pmu-sram");

  	if (!sram_np) {

arch/arm/mach-rockchip/pm.h +4 -0

index 3e8d39c..f8a747b 100644
@@ -48,6 +48,10 @@ static inline void rockchip_suspend_init(void)
 
  #define RK3288_PMU_WAKEUP_RST_CLR_CNT	0x44
  #define RK3288_PMU_PWRMODE_CON1		0x90
 
+#define RK3288_GRF_SOC_CON0		0x244
+#define GRF_FORCE_JTAG			BIT(12)
+#define GRF_FORCE_JTAG_WRITE		BIT(28)
+
 
  #define RK3288_SGRF_SOC_CON0		(0x0000)

  #define RK3288_SGRF_FAST_BOOT_ADDR	(0x0120)
  #define SGRF_PCLK_WDT_GATE		BIT(6)

Chris Zhong

2015-05-13 00:53 UTC #5 lore raw

PATCH 1/2 ARM: rockchip: fix undefined instruction of reset_ctrl_regs

Hi Heiko

Quoted 10 lines · Heiko Stuebner

Hi Chris,
Am Dienstag, 12. Mai 2015, 12:00:00 schrieben Sie:
Quoted 4 lines
We do not need this patch, since we have this one:
<https://patchwork.kernel.org/patch/6218791/>
these 2 patches fixed a same issue.
thanks for pointing that out - looks like I didn't need to resurrect the old patch then. Just to make sure I don't mess up when removing the obsolete one:

Quoted 6 lines

both patches
b403125 ARM: rockchip: fix undefined instruction of reset_ctrl_regs 0ea001d ARM: rockchip: disable dapswjdp during suspend
actually fix the same issue and
b403125 ARM: rockchip: fix undefined instruction of reset_ctrl_regs
is the one that should be removed, right?

Yes, this one should be removed.

Quoted 146 lines

Thanks
Heiko

[I've readded the lists to give this a bit publicity, so people can see why we remove one of the patches again]

Quoted 47 lines · Heiko St?bner

Quoted 46 lines
Sometimes the debug module may not work well after resume, since it has not been correctly reset when wakeup from suspend. That cause system crash during reusme, and a 'undefined instruction' is displayed on the console. Set the GRF_FORCE_JTAG bit of RK3288_GRF_SOC_CON0 can ensure that debug modul is reset. And we can change the value of RK3288_GRF_SOC_CON0 back when system resume.
Signed-off-by Chris Zhong <[email protected]>
Tested-by Caesar Wang <[email protected]>
Reviewed-by Douglas Anderson <[email protected]>
According to discussions, there does not seem a better solution available. Please also see the potential security implication described in the comment inline in the code.
Signed-off-by Heiko Stuebner <[email protected]>
---
   arch/arm/mach-rockchip/pm.c | 26 ++++++++++++++++++++++++++
   arch/arm/mach-rockchip/pm.h |  4 ++++
   2 files changed, 30 insertions(+)
arch/arm/mach-rockchip/pm.c +2 -0
index b0dcbe2..22812fe 100644
@@ -44,9 +44,11 @@ static void __iomem *rk3288_bootram_base;
 
   static phys_addr_t rk3288_bootram_phy;
 
   static struct regmap *pmu_regmap;
 
+static struct regmap *grf_regmap;
 
   static struct regmap *sgrf_regmap;
 
   static u32 rk3288_pmu_pwr_mode_con;
 
+static u32 rk3288_grf_soc_con0;
 
   static u32 rk3288_sgrf_soc_con0;
 
   static inline u32 rk3288_l2_config(void)
@@ -70,12 +72,26 @@ static void rk3288_slp_mode_set(int level)
 

Quoted 8 lines

arch/arm/mach-rockchip/pm.c static void rk3288_slp_mode_set(int level)
Quoted 8 lines
arch/arm/mach-rockchip/pm.c +2 -0
   {
 
   	u32 mode_set, mode_set1;
 
+	regmap_read(grf_regmap, RK3288_GRF_SOC_CON0, &rk3288_grf_soc_con0);
+
 
   	regmap_read(sgrf_regmap, RK3288_SGRF_SOC_CON0,

&rk3288_sgrf_soc_con0);

Quoted 78 lines

arch/arm/mach-rockchip/pm.c +22 -0

   	regmap_read(pmu_regmap, RK3288_PMU_PWRMODE_CON,
 
   		    &rk3288_pmu_pwr_mode_con);
 
   	/*
 
+	 * We need set this bit GRF_FORCE_JTAG here, for the debug module,
+	 * otherwise, it may become inaccessible after resume.
+	 * This creates a potential security issue, as the sdmmc pins may
+	 * accept jtag data for a short time during resume if no card is
+	 * inserted.
+	 * But this is of course also true for the regular boot, before we
+	 * turn of the jtag/sdmmc autodetect.
+	 */
+	regmap_write(grf_regmap, RK3288_GRF_SOC_CON0, GRF_FORCE_JTAG |
+		     GRF_FORCE_JTAG_WRITE);
+
+	/*
 
   	 * SGRF_FAST_BOOT_EN - system to boot from FAST_BOOT_ADDR
   	 * PCLK_WDT_GATE - disable WDT during suspend.
   	 */

@@ -135,6 +151,9 @@ static void rk3288_slp_mode_set_resume(void)
 
   	regmap_write(sgrf_regmap, RK3288_SGRF_SOC_CON0,
 
   		     rk3288_sgrf_soc_con0 | SGRF_PCLK_WDT_GATE_WRITE
 
   		     | SGRF_FAST_BOOT_EN_WRITE);
 
+
+	regmap_write(grf_regmap, RK3288_GRF_SOC_CON0, rk3288_grf_soc_con0 |
+		     GRF_FORCE_JTAG_WRITE);
 
   }
 
   static int rockchip_lpmode_enter(unsigned long arg)

@@ -193,6 +212,13 @@ static int rk3288_suspend_init(struct device_node
 *np)
 
   		return PTR_ERR(pmu_regmap);
 
   	}
 
+	grf_regmap = syscon_regmap_lookup_by_compatible(
+				"rockchip,rk3288-grf");
+	if (IS_ERR(grf_regmap)) {
+		pr_err("%s: could not find grf regmap\n", __func__);
+		return PTR_ERR(pmu_regmap);
+	}
+

   	sram_np = of_find_compatible_node(NULL, NULL,
   	
   					  "rockchip,rk3288-pmu-sram");
   	
   	if (!sram_np) {

arch/arm/mach-rockchip/pm.h +4 -0

index 3e8d39c..f8a747b 100644
@@ -48,6 +48,10 @@ static inline void rockchip_suspend_init(void)
 
   #define RK3288_PMU_WAKEUP_RST_CLR_CNT	0x44
   #define RK3288_PMU_PWRMODE_CON1		0x90
 
+#define RK3288_GRF_SOC_CON0		0x244
+#define GRF_FORCE_JTAG			BIT(12)
+#define GRF_FORCE_JTAG_WRITE		BIT(28)
+
 
   #define RK3288_SGRF_SOC_CON0		(0x0000)

   #define RK3288_SGRF_FAST_BOOT_ADDR	(0x0120)
   #define SGRF_PCLK_WDT_GATE		BIT(6)

Heiko Stübner

author maintainer

2015-05-06 11:31 UTC #6 lore raw

In reply to Heiko Stübner

Hi Arnd, Olof,

Am Samstag, 25. April 2015, 02:11:14 schrieb Heiko St?bner:

Quoted 2 lines

I've collected some fixes which might be appropriate for 4.1-rc, so if they're ok, please pull.

is somebody doing arm-soc fixes for 4.1?
I currently don't see any activity at all and this pull is open for two weeks now?

Heiko

Quoted 40 lines

Thanks
Heiko
The following changes since commit a0307d186f5015ee3d77e209e5c06190cbf02478:
ARM: rockchip: disable watchdog during suspend (2015-03-11 22:41:02 +0100)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip.git
tags/v4.1-rockchip-socfixes1
for you to fetch changes up to 2a9fe3ca84afff6259820c4f62e579f41476becc:
rockchip: make sure timer7 is enabled on rk3288 platforms (2015-04-16
21:22:17 +0200)
----------------------------------------------------------------
Two fixes from Chris Zhong, fixing some suspend oddities. And I've given up on the timer7 issue. While I initially thought devices would either have both the grave mmu issue requiring a uboot update and the timer7 issue or none, it looks like in all units in the field the mmu issue got fixed while the timer7 issue stayed on. So instead of making everybody wanting to use mainline jump through a hoop just make sure timer7 is on on boot before we init the arch-timer.
----------------------------------------------------------------
Chris Zhong (2):
ARM: rockchip: disable dapswjdp during suspend
ARM: rockchip: fix undefined instruction of reset_ctrl_regs
Heiko Stuebner (1):
rockchip: make sure timer7 is enabled on rk3288 platforms
 arch/arm/mach-rockchip/pm.c       | 33 +++++++++++++++++++++++++++++++++
 arch/arm/mach-rockchip/pm.h       |  8 ++++++++
 arch/arm/mach-rockchip/rockchip.c | 19 +++++++++++++++++++
 3 files changed, 60 insertions(+)

Arnd Bergmann

2015-05-07 16:20 UTC #7 lore raw

In reply to Heiko Stübner

Quoted 4 lines · Heiko St?bner

Hi Arnd, Olof,
I've collected some fixes which might be appropriate for 4.1-rc, so if they're ok, please pull.

Pulled into fixes branch, thanks and sorry for the delay.

Arnd