Skip to contentSkip to conversation

2nd Round Samsung mach-exynos for v3.12

GIT PULL New Kukjin Kim 6 participants 15 messages Aug 26, 2013 – Aug 29, 2013

Subsystems (3)
lore
Share & tools
lore

Kukjin Kim

author
#0 lore raw

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

Bartlomiej Zolnierkiewicz (1):
ARM: EXYNOS: always enable PM domains support for EXYNOS4X12

 arch/arm/mach-exynos/Kconfig   |    7 +++++++
 arch/arm/mach-exynos/cpuidle.c |    2 +-
 2 files changed, 8 insertions(+), 1 deletion(-)
#1 lore raw

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

To disable C1 state on EXYNOS5440 something like:

	static int __init exynos4_init_cpuidle(void)
	{
	...
		if (soc_is_exynos5440())
			exynos4_idle_driver.state_count = 1;
	...
	}

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

#2 lore raw

Hi,

Quoted 47 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

Quoted 7 lines

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

I looked at your patch series and it seems reasonable. I will repost this patch on top of yours.
But I suggest to keep this patch temporary till your patch series gets merged.

Thanks,
Amit Daniel

Quoted 25 lines

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Quoted 79 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#3 lore raw
Quoted 55 lines · amit daniel kachhap

Hi,

Quoted 51 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

Quoted 5 lines
Quoted 3 lines

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

I looked at your patch series and it seems reasonable. I will repost this patch on top of yours.

If you correctly use driver's state_count (instead of device's) there will be no dependency on my patch series and the new patch can be applied immediately.

Quoted 1 line

But I suggest to keep this patch temporary till your patch series gets merged.

The current patch (the one Kukjin merged) is incorrect as it just doesn't do what it advertises. I see no reason to keep it.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

Quoted 110 lines

Thanks,
Amit Daniel

Quoted 25 lines

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Quoted 79 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#4 lore raw

Bartlomiej Zolnierkiewicz wrote:

[...]

Quoted 23 lines
Quoted 3 lines
Quoted 3 lines

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-

July/186355.html

Quoted 4 lines
Quoted 2 lines

[ Because of the deficiency in the core cpuidle core (device-

state_count

Quoted 1 line

not being used by governors' code) only sysfs entries for C1 state

will be

Quoted 1 line
Quoted 1 line

disabled and EXYNOS cpuidle driver will still attempt to use C1

state.

Quoted 6 lines
Quoted 4 lines

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

I looked at your patch series and it seems reasonable. I will repost this patch on top of yours.

If you correctly use driver's state_count (instead of device's) there will be
no dependency on my patch series and the new patch can be applied immediately.

Quoted 6 lines
Quoted 1 line

But I suggest to keep this patch temporary till your patch series gets

merged.

The current patch (the one Kukjin merged) is incorrect as it just doesn't do what it advertises. I see no reason to keep it.

Well, I don't think so, because if the patch is missing, following kernel panic happens on exynos5440 platform.

Unable to handle kernel paging request at virtual address f8180608
pgd = c0003000
[f8180608] *pgd=80000080007003, *pmd=af7fb003, *pte=00000000
Internal error: Oops: a07 [#1] PREEMPT ARM
Modules linked in:
CPU: 0 PID: 0 Comm: swapper Not tainted 3.11.0-0-generic

#1~d20130819t044008~3372c79
task: c0529d80 ti: c051e000 task.ti: c051e000
PC is at exynos4_enter_lowpower+0x18/0x130
LR is at cpuidle_enter_state+0x3c/0xe8

pc : []    lr : []    psr: 200f0093

sp : c051ff68 ip : 00000018 fp : 00000000
r10: 00000000 r9 : 412fc0f3 r8 : 00000000
r7 : c052c9cc r6 : 00000001 r5 : 00000000 r4 : d5c3cc94 r3 : f8180000 r2 : 0000ff3e r1 : c052c980 r0 : c052cc98
Flags: nzCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment kernel
Control: 30c5387d Table: af139b00 DAC: fffffffd
Process swapper (pid: 0, stack limit = 0xc051e230)

Stack: (0xc051ff68 to 0xc0520000)

ff60: d5c3cc94 00000000 c052cc98 c02c1310 d5c3cc94
00000000
ff80: c0550288 c052c980 c058898c c052cc98 00000001 c052c980 c058898c c02c1458
ffa0: c051e000 c0550288 c0550288 00000001 c05260dc c001b2cc 000001fe c00573e4
ffc0: ffffffff c04f07d8 ffffffff ffffffff c04f0290 00000000 00000000 c05134d8
ffe0: 30c7387d c0526064 c05134d4 c052b5b4 80007000 80008080 00000000 00000000
[] (exynos4_enter_lowpower+0x18/0x130) from [] (cpuidle_enter_state+0x3c/0xe8)
[] (cpuidle_enter_state+0x3c/0xe8) from [] (cpuidle_idle_call+0x9c/0x140) [] (cpuidle_idle_call+0x9c/0x140) from [] (arch_cpu_idle+0x8/0x38) [] (arch_cpu_idle+0x8/0x38) from [] (cpu_startup_entry+0x4c/0x114) [] (cpu_startup_entry+0x4c/0x114) from [] (start_kernel+0x324/0x37c)

Code: 0a000043 e3a03000 e30f2f3e e34f3818 (e5832608) 
---[ end trace 617b9e1a4ff91d2f ]---
Kernel panic - not syncing: Attempted to kill the idle task!

In addition, I didn't see your patches in linux-next for upcoming merge window. So I think, if any fixup is required, it should be done on top of current shape.

- Kukjin

#5 lore raw
Quoted 71 lines · Kukjin Kim

Bartlomiej Zolnierkiewicz wrote:

[...]

Quoted 23 lines
Quoted 3 lines
Quoted 3 lines

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-

July/186355.html

Quoted 4 lines
Quoted 2 lines

[ Because of the deficiency in the core cpuidle core (device-

state_count

Quoted 1 line

not being used by governors' code) only sysfs entries for C1 state

will be

Quoted 1 line
Quoted 1 line

disabled and EXYNOS cpuidle driver will still attempt to use C1

state.

Quoted 6 lines
Quoted 4 lines

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

I looked at your patch series and it seems reasonable. I will repost this patch on top of yours.

If you correctly use driver's state_count (instead of device's) there will be
no dependency on my patch series and the new patch can be applied immediately.

Quoted 6 lines
Quoted 1 line

But I suggest to keep this patch temporary till your patch series gets

merged.

The current patch (the one Kukjin merged) is incorrect as it just doesn't do what it advertises. I see no reason to keep it.

Well, I don't think so, because if the patch is missing, following kernel panic happens on exynos5440 platform.

Unable to handle kernel paging request at virtual address f8180608
pgd = c0003000
[f8180608] *pgd=80000080007003, *pmd=af7fb003, *pte=00000000
Internal error: Oops: a07 [#1] PREEMPT ARM
Modules linked in:
CPU: 0 PID: 0 Comm: swapper Not tainted 3.11.0-0-generic

#1~d20130819t044008~3372c79
task: c0529d80 ti: c051e000 task.ti: c051e000
PC is at exynos4_enter_lowpower+0x18/0x130
LR is at cpuidle_enter_state+0x3c/0xe8

pc : []    lr : []    psr: 200f0093

sp : c051ff68 ip : 00000018 fp : 00000000
r10: 00000000 r9 : 412fc0f3 r8 : 00000000
r7 : c052c9cc r6 : 00000001 r5 : 00000000 r4 : d5c3cc94 r3 : f8180000 r2 : 0000ff3e r1 : c052c980 r0 : c052cc98
Flags: nzCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment kernel
Control: 30c5387d Table: af139b00 DAC: fffffffd
Process swapper (pid: 0, stack limit = 0xc051e230)

Stack: (0xc051ff68 to 0xc0520000)

ff60: d5c3cc94 00000000 c052cc98 c02c1310 d5c3cc94
00000000
ff80: c0550288 c052c980 c058898c c052cc98 00000001 c052c980 c058898c c02c1458
ffa0: c051e000 c0550288 c0550288 00000001 c05260dc c001b2cc 000001fe c00573e4
ffc0: ffffffff c04f07d8 ffffffff ffffffff c04f0290 00000000 00000000 c05134d8
ffe0: 30c7387d c0526064 c05134d4 c052b5b4 80007000 80008080 00000000 00000000
[] (exynos4_enter_lowpower+0x18/0x130) from [] (cpuidle_enter_state+0x3c/0xe8)
[] (cpuidle_enter_state+0x3c/0xe8) from [] (cpuidle_idle_call+0x9c/0x140) [] (cpuidle_idle_call+0x9c/0x140) from [] (arch_cpu_idle+0x8/0x38) [] (arch_cpu_idle+0x8/0x38) from [] (cpu_startup_entry+0x4c/0x114) [] (cpu_startup_entry+0x4c/0x114) from [] (start_kernel+0x324/0x37c)

Quoted 3 lines
Code: 0a000043 e3a03000 e30f2f3e e34f3818 (e5832608) 
---[ end trace 617b9e1a4ff91d2f ]---
Kernel panic - not syncing: Attempted to kill the idle task!

1) samsung-mach-exynos kernel (from your pull request) doesn't have neither exynos5440_defconfig nor EXYNOS5440 enabled in exynos_defconfig so could you please tell me what kernel version and kernel config are you using to get the above panic?

Could you also see what does exynos4_enter_lowpower+0x18 map to in your source code?

2) dev->state_count is used only for managing sysfs entries (and clearing dev->states_usage) in the cpuidle core.

$ git grep state_count drivers/cpuidle/

drivers/cpuidle/cpuidle-calxeda.c: .state_count = 2,
drivers/cpuidle/cpuidle-kirkwood.c: .state_count = KIRKWOOD_MAX_STATES,

drivers/cpuidle/cpuidle-zynq.c: .state_count = ZYNQ_MAX_STATES,
drivers/cpuidle/cpuidle.c:      for (i = drv->state_count - 1; i >= CPUIDLE_DRIVER_STATE_START; i--)
drivers/cpuidle/cpuidle.c:      if (!dev->state_count)
drivers/cpuidle/cpuidle.c:              dev->state_count = drv->state_count;
drivers/cpuidle/cpuidle.c:      for (i = 0; i < dev->state_count; i++) {
drivers/cpuidle/driver.c:       for (i = drv->state_count - 1; i >= 0 ; i--) {
drivers/cpuidle/driver.c:       if (!drv || !drv->state_count)
drivers/cpuidle/governors/ladder.c:     if (last_idx < drv->state_count - 1 &&
drivers/cpuidle/governors/ladder.c:     for (i = 0; i < drv->state_count; i++) {
drivers/cpuidle/governors/ladder.c:             if (i < drv->state_count - 1)
drivers/cpuidle/governors/menu.c:       for (i = CPUIDLE_DRIVER_STATE_START; i < drv->state_count; i++) {
drivers/cpuidle/sysfs.c:        for (i = 0; i < device->state_count; i++) {
drivers/cpuidle/sysfs.c:        for (i = 0; i < device->state_count; i++)

With the current code I fail to see how it is possible that dev->state_count smaller than drv->state_count affects anything besides sysfs cpuidle entries in the practice.

IOW I worry that the current patch may be just masking some other issue.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

#6 lore raw

PATCH v2 1/1 ARM: EXYNOS: cpuidle: Skip C1 cpuidle state for exynos5440

In reply to Bartlomiej Zolnierkiewicz

This patch skips the deep C1(AFTR -Arm off top running) state for exynos5440 soc as this soc does not support this state. The cpu's only allows the basic C0 state. The C1 state is filtered by re-initialising the driver state_count value to 1.

Cc Kukjin Kim <[email protected]>
Cc Bartlomiej Zolnierkiewicz <[email protected]>

---

Changes in V2:
* Used driver state_count variable to filter the C1 state instead of device state count variable. Although this V2 patch is also not complete as there are suggestions to make this driver platform driver.

 arch/arm/mach-exynos/cpuidle.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)
arch/arm/mach-exynos/cpuidle.c +3 -0
index 225ee84..ac13922 100644
@@ -200,6 +200,9 @@ static int __init exynos4_init_cpuidle(void)
 	if (soc_is_exynos5250())
 		exynos5_core_down_clk();
 
+	if (soc_is_exynos5440())
+		exynos4_idle_driver.state_count = 1;
+
 	ret = cpuidle_register_driver(&exynos4_idle_driver);
 	if (ret) {
 		printk(KERN_ERR "CPUidle failed to register driver\n");
Signature · 1 line
1.7.1
#7 lore raw

PATCH v2 1/1 ARM: EXYNOS: cpuidle: Skip C1 cpuidle state for exynos5440

Quoted 7 lines · Amit Daniel Kachhap

This patch skips the deep C1(AFTR -Arm off top running) state for exynos5440 soc as this soc does not support this state. The cpu's only allows the basic C0 state. The C1 state is filtered by re-initialising the driver state_count value to 1.

Cc Kukjin Kim <[email protected]>
Cc Bartlomiej Zolnierkiewicz <[email protected]>

Thanks for fixing this.

Quoted 7 lines

---

Changes in V2:
* Used driver state_count variable to filter the C1 state instead of device state count variable. Although this V2 patch is also not complete as there are suggestions to make this driver platform driver.

I'm going to look into making it a platform driver later this week.

Quoted 8 lines
 arch/arm/mach-exynos/cpuidle.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)
arch/arm/mach-exynos/cpuidle.c +0 -0
index 225ee84..ac13922 100644

arch/arm/mach-exynos/cpuidle.c static int __init exynos4_init_cpuidle(void) · line 208

Quoted 9 lines
arch/arm/mach-exynos/cpuidle.c +3 -0
@@ -200,6 +200,9 @@ static int __init exynos4_init_cpuidle(void)
 	if (soc_is_exynos5250())
 		exynos5_core_down_clk();
 
+	if (soc_is_exynos5440())
+		exynos4_idle_driver.state_count = 1;
+
 	ret = cpuidle_register_driver(&exynos4_idle_driver);
 	if (ret) {
 		printk(KERN_ERR "CPUidle failed to register driver\n");

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

#8 lore raw
In reply to Bartlomiej Zolnierkiewicz

Submitted the V2 version of this patch with your suggestion.

Thanks
Amit

Quoted 192 lines · Bartlomiej Zolnierkiewicz
Quoted 55 lines · amit daniel kachhap

Hi,

Quoted 51 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

Quoted 5 lines
Quoted 3 lines

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

I looked at your patch series and it seems reasonable. I will repost this patch on top of yours.

If you correctly use driver's state_count (instead of device's) there will be no dependency on my patch series and the new patch can be applied immediately.

Quoted 1 line

But I suggest to keep this patch temporary till your patch series gets merged.

The current patch (the one Kukjin merged) is incorrect as it just doesn't do what it advertises. I see no reason to keep it.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

Quoted 106 lines

Thanks,
Amit Daniel

Quoted 25 lines

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Quoted 75 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

Quoted 4 lines
Quoted 4 lines

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#9 lore raw
Quoted 1 line · amit daniel kachhap

Submitted the V2 version of this patch with your suggestion.

So will there be an updated branch (and pull request) with these changes?

Kevin

Quoted 194 lines
Quoted 192 lines · Bartlomiej Zolnierkiewicz
Quoted 55 lines · amit daniel kachhap

Hi,

Quoted 51 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

Quoted 5 lines
Quoted 3 lines

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

I looked at your patch series and it seems reasonable. I will repost this patch on top of yours.

If you correctly use driver's state_count (instead of device's) there will be no dependency on my patch series and the new patch can be applied immediately.

Quoted 1 line

But I suggest to keep this patch temporary till your patch series gets merged.

The current patch (the one Kukjin merged) is incorrect as it just doesn't do what it advertises. I see no reason to keep it.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

Quoted 106 lines

Thanks,
Amit Daniel

Quoted 25 lines

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Quoted 75 lines · Bartlomiej Zolnierkiewicz

Hi,

Quoted 20 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos

for you to fetch changes up to f52616f4233d71d0fb00f06f86d046d18d2b7f3b:

ARM: EXYNOS: always enable PM domains support for EXYNOS4X12 (2013-08-19
05:05:16 +0900)

----------------------------------------------------------------
update mach-exynos
- enable ARCH_HAS_BANDGAP for exynos SoCs
- skip C1 cpuidle for exynos5440 because non-supporting
- always enable PM domains for exynos4x12

Quoted 4 lines

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

The patch "ARM: EXYNOS: Skip C1 cpuidle state for exynos5440":

arch/arm/mach-exynos/cpuidle.c +1 -1
@@ -210,7 +210,7 @@ static int __init exynos4_init_cpuidle(void)
                device->cpu = cpu_id;
 
                /* Support IDLE only */
-               if (cpu_id != 0)
+               if (soc_is_exynos5440() || cpu_id != 0)
                        device->state_count = 1;
 
                ret = cpuidle_register_device(device);

is incorrect as noted a month ago in:

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-July/186355.html

[ Because of the deficiency in the core cpuidle core (device->state_count
not being used by governors' code) only sysfs entries for C1 state will be disabled and EXYNOS cpuidle driver will still attempt to use C1 state.

also non-working device->state_count is planned to be removed by:

http://permalink.gmane.org/gmane.linux.power-management.general/37390

To disable C1 state on EXYNOS5440 something like:

        static int __init exynos4_init_cpuidle(void)
        {
        ...
                if (soc_is_exynos5440())
                        exynos4_idle_driver.state_count = 1;
        ...
        }

should be done instead.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

Quoted 4 lines
Quoted 4 lines

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

author
#10 lore raw
Quoted 6 lines · Kevin Hilman
Quoted 1 line · amit daniel kachhap

Submitted the V2 version of this patch with your suggestion.

So will there be an updated branch (and pull request) with these changes?

OK, Bart's concern makes sense so I replaced with Amit's v2 patch.

Kevin, here is updated branch so please pull following.

If any problems, please kindly let me know.

Thanks,
Kukjin

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos-v2

for you to fetch changes up to 1e9fec0e40deff9d18e4197b5f5f99f85f5d36a5:

ARM: EXYNOS: Skip C1 cpuidle state for exynos5440 (2013-08-28
00:51:21 +0900)

----------------------------------------------------------------
update mach-exynos v2 for v3.12
- enable ARCH_HAS_BANDGAP for exynos SoCs
- always enable PM domains for exynos4x12
- skip C1 cpuidle state for exynos5440

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

Bartlomiej Zolnierkiewicz (1):
ARM: EXYNOS: always enable PM domains support for EXYNOS4X12

  arch/arm/mach-exynos/Kconfig   | 7 +++++++
  arch/arm/mach-exynos/cpuidle.c | 3 +++
  2 files changed, 10 insertions(+)
#11 lore raw
Quoted 5 lines · Kukjin Kim
Quoted 4 lines · Kevin Hilman
Quoted 1 line · amit daniel kachhap

Submitted the V2 version of this patch with your suggestion.

Quoted 7 lines
Quoted 2 lines

So will there be an updated branch (and pull request) with these changes?

OK, Bart's concern makes sense so I replaced with Amit's v2 patch.

Kevin, here is updated branch so please pull following.

If any problems, please kindly let me know.

Please don't post pull requests in the middle of a comment thread; start a new thread or at least do a new subject for them -- they are much too easy to miss otherwise.

-Olof

author
#12 lore raw
Quoted 14 lines · Olof Johansson
Quoted 6 lines · Kukjin Kim
Quoted 5 lines · Kevin Hilman
Quoted 1 line · amit daniel kachhap

Submitted the V2 version of this patch with your suggestion.

So will there be an updated branch (and pull request) with these

changes?

Quoted 5 lines
Quoted 1 line

OK, Bart's concern makes sense so I replaced with Amit's v2 patch.

Kevin, here is updated branch so please pull following.

Quoted 8 lines
Quoted 1 line

If any problems, please kindly let me know.

Please don't post pull requests in the middle of a comment thread; start a new
thread or at least do a new subject for them -- they are much too easy to miss
otherwise.

Oh, OK. I see.

Thanks,
Kukjin

author
#13 lore raw

[GIT PULL V2 3/3] 2nd Round Samsung mach-exynos-v2 for v3.12

In reply to Kevin Hilman

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos-v2

for you to fetch changes up to 1e9fec0e40deff9d18e4197b5f5f99f85f5d36a5:

ARM: EXYNOS: Skip C1 cpuidle state for exynos5440 (2013-08-28
00:51:21 +0900)

----------------------------------------------------------------
update mach-exynos v2 for v3.12
- enable ARCH_HAS_BANDGAP for exynos SoCs
- always enable PM domains for exynos4x12
- skip C1 cpuidle state for exynos5440

----------------------------------------------------------------
Amit Daniel Kachhap (2):
ARM: EXYNOS: enable ARCH_HAS_BANDGAP
ARM: EXYNOS: Skip C1 cpuidle state for exynos5440

Bartlomiej Zolnierkiewicz (1):
ARM: EXYNOS: always enable PM domains support for EXYNOS4X12

  arch/arm/mach-exynos/Kconfig   | 7 +++++++
  arch/arm/mach-exynos/cpuidle.c | 3 +++
  2 files changed, 10 insertions(+)
#14 lore raw

[GIT PULL V2 3/3] 2nd Round Samsung mach-exynos-v2 for v3.12

Quoted 9 lines · Kukjin Kim

The following changes since commit ad81f0545ef01ea651886dddac4bef6cec930092:

   Linux 3.11-rc1 (2013-07-14 15:18:27 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung.git
tags/samsung-mach-exynos-v2

Pulled into late/all.

-Olof