Skip to contentSkip to conversation

ARM: pxa: support 806MHz operating points for PXA31x processors A2 stepping

PATCH In review Axel Lin 4 participants 8 messages Feb 15, 2011 – Feb 23, 2011

Subsystems (3)
Arm (32-bit) Arm sub-architectures PXA2xx/PXA3xx support
lore
Share & tools
lore

Axel Lin

author
#0 lore raw

PATCH 1/1 ARM: pxa: support 806MHz operating points for PXA31x processors A2 stepping

PXA3xx Specification Update document states that the 806MHz operating points were added for PXA31x processors A2 stepping.

This patch adds 806MHz operating points support for PXA31x A2 stepping.

Signed-off-by Axel Lin <[email protected]>

---

I cannot find any details for PXA31x stepping B1 and B2, thus I made this patch for PXA31x stepping A2 only.

Axel

 arch/arm/mach-pxa/cpufreq-pxa3xx.c |   23 +++++++++++++++++++----
 1 files changed, 19 insertions(+), 4 deletions(-)
arch/arm/mach-pxa/cpufreq-pxa3xx.c +19 -4
index 88fbec0..f4f0c2e 100644
@@ -210,15 +210,30 @@ static int pxa3xx_cpufreq_init(struct cpufreq_policy *policy)
 
 	/* set default policy and cpuinfo */
 	policy->cpuinfo.min_freq = 104000;
-	policy->cpuinfo.max_freq = (cpu_is_pxa320()) ? 806000 : 624000;
+
+	if (cpu_is_pxa300())
+		policy->cpuinfo.max_freq = 624000;
+	else if (cpu_is_pxa310())
+		policy->cpuinfo.max_freq = ((read_cpuid_id() & 0xf) == 0x2) ?
+						806000 : 624000;
+	else if (cpu_is_pxa320())
+		policy->cpuinfo.max_freq = 806000;
+
 	policy->cpuinfo.transition_latency = 1000; /* FIXME: 1 ms, assumed */
 	policy->max = pxa3xx_get_clk_frequency_khz(0);
 	policy->cur = policy->min = policy->max;
 
-	if (cpu_is_pxa300() || cpu_is_pxa310())
+	if (cpu_is_pxa300())
 		ret = setup_freqs_table(policy, ARRAY_AND_SIZE(pxa300_freqs));
-
-	if (cpu_is_pxa320())
+	else if (cpu_is_pxa310())
+		/* PXA310 A2 STEPPING supports 806Mhz operating points */
+		if ((read_cpuid_id() & 0xf) == 0x2)
+			ret = setup_freqs_table(policy,
+						ARRAY_AND_SIZE(pxa320_freqs));
+		else
+			ret = setup_freqs_table(policy,
+						ARRAY_AND_SIZE(pxa300_freqs));
+	else if (cpu_is_pxa320())
 		ret = setup_freqs_table(policy, ARRAY_AND_SIZE(pxa320_freqs));
 
 	if (ret) {
Signature · 1 line
1.7.2
#1 lore raw

Hi,

Quoted 45 lines · Axel Lin

PXA3xx Specification Update document states that the 806MHz operating points were added for PXA31x processors A2 stepping.

This patch adds 806MHz operating points support for PXA31x A2 stepping.

Signed-off-by Axel Lin <[email protected]>

---

I cannot find any details for PXA31x stepping B1 and B2, thus I made this patch for PXA31x stepping A2 only.

Axel

 arch/arm/mach-pxa/cpufreq-pxa3xx.c |   23 +++++++++++++++++++----
 1 files changed, 19 insertions(+), 4 deletions(-)
arch/arm/mach-pxa/cpufreq-pxa3xx.c +12 -4
index 88fbec0..f4f0c2e 100644
@@ -210,15 +210,30 @@ static int pxa3xx_cpufreq_init(struct cpufreq_policy *policy)
 
 	/* set default policy and cpuinfo */
 	policy->cpuinfo.min_freq = 104000;
-	policy->cpuinfo.max_freq = (cpu_is_pxa320()) ? 806000 : 624000;
+
+	if (cpu_is_pxa300())
+		policy->cpuinfo.max_freq = 624000;
+	else if (cpu_is_pxa310())
+		policy->cpuinfo.max_freq = ((read_cpuid_id() & 0xf) == 0x2) ?
+						806000 : 624000;
+	else if (cpu_is_pxa320())
+		policy->cpuinfo.max_freq = 806000;
+
 	policy->cpuinfo.transition_latency = 1000; /* FIXME: 1 ms, assumed */
 	policy->max = pxa3xx_get_clk_frequency_khz(0);
 	policy->cur = policy->min = policy->max;
 
-	if (cpu_is_pxa300() || cpu_is_pxa310())
+	if (cpu_is_pxa300())
 		ret = setup_freqs_table(policy, ARRAY_AND_SIZE(pxa300_freqs));
-
-	if (cpu_is_pxa320())
+	else if (cpu_is_pxa310())
+		/* PXA310 A2 STEPPING supports 806Mhz operating points */

arch/arm/mach-pxa/cpufreq-pxa3xx.c static int pxa3xx_cpufreq_init(struct cpufreq_policy *policy) · line 239

Quoted 10 lines
arch/arm/mach-pxa/cpufreq-pxa3xx.c +7 -0
+		if ((read_cpuid_id() & 0xf) == 0x2)
+			ret = setup_freqs_table(policy,
+						ARRAY_AND_SIZE(pxa320_freqs));
+		else
+			ret = setup_freqs_table(policy,
+						ARRAY_AND_SIZE(pxa300_freqs));
+	else if (cpu_is_pxa320())
 		ret = setup_freqs_table(policy, ARRAY_AND_SIZE(pxa320_freqs));
 
 	if (ret) {

Yeah, I've proposed something like that about half a year ago:
http://www.spinics.net/lists/arm-kernel/msg92899.html

Eric was not so happy about it back then, may be this time? Eric, what do you think?

Also, there is one thing that is not related to this patch, but if you have the hardware, you need to be concerned with, is that the vcc_core voltage has to be 1.4V, otherwise your hardware will be unstable.

Signature · 2 lines
Regards,
Igor.
author
#2 lore raw

Hi Igor,

Quoted 4 lines

Also, there is one thing that is not related to this patch, but if you have the hardware, you need to be concerned with, is that the vcc_core voltage has to be 1.4V, otherwise your hardware will be unstable.

I don't see any unstable issue on my device.
But it seems the vcc_core and vcc_sram fields of pxa3xx_freq_info are not used now.

Regards,
Axel

#3 lore raw

Hi,

Quoted 9 lines · Axel Lin

Hi Igor,

Quoted 4 lines

Also, there is one thing that is not related to this patch, but if you have the hardware, you need to be concerned with, is that the vcc_core voltage has to be 1.4V, otherwise your hardware will be unstable.

I don't see any unstable issue on my device.
But it seems the vcc_core and vcc_sram fields of pxa3xx_freq_info are not used now.

Indeed, they are not used, because we don't have any DVFS driver for PXA3xx. But, the hardware is running regardless of you have the driver or not ;) There are other means to make sure the vcc_core voltage is set appropriately:
1) could be already set (PMIC default)
2) could be set by "Hardware-Controlled Voltage-Change Sequencer" automatically (PXA3xx_DM_Vol_I chapter 6.8.3) if your PMIC is 100% compatible with PXA3xx and PVCR[FVE/PVE] bits set accordingly.
3) If your PMIC is for some reason not 100% compatible, then the voltage can be set by regulator API or even by the bootloader.

Anyway, to make sure it is set to the appropriate level, you need to measure it. Usually, this line has some capacitors on it, so it shouldn't be a hard task

Now, regarding hardware instability:
I've seen instabilities related to the vcc_core set improperly, but it has never showed up that quickly, sometimes it took a week of testing for it to show up. But when it shows up, you can see strange things happening, like your DRAM is messed or buses fail or the cpu just freezes.

Another thing (which could be related to the patch) is the package marking. We have PXA3xx SoCs marked A2 stepping, but C624, which means that the highest running frequency should not exceed 624MHz.
What is your PXA3xx A2 marking in respect to Cxxx option?

Signature · 2 lines
Regards,
Igor.
author
#4 lore raw

2011/2/22 Igor Grinberg <[email protected]>:

Quoted 24 lines

Hi,

Quoted 9 lines · Axel Lin

Hi Igor,

Quoted 4 lines

Also, there is one thing that is not related to this patch, but if you have the hardware, you need to be concerned with, is that the vcc_core voltage has to be 1.4V, otherwise your hardware will be unstable.

I don't see any unstable issue on my device.
But it seems the vcc_core and vcc_sram fields of pxa3xx_freq_info are not used now.

Indeed, they are not used, because we don't have any DVFS driver for PXA3xx. But, the hardware is running regardless of you have the driver or not ;) There are other means to make sure the vcc_core voltage is set appropriately:
1) could be already set (PMIC default)
2) could be set by "Hardware-Controlled Voltage-Change Sequencer" automatically ? (PXA3xx_DM_Vol_I chapter 6.8.3) if your PMIC is 100% compatible with PXA3xx ? and PVCR[FVE/PVE] bits set accordingly.
3) If your PMIC is for some reason not 100% compatible, then the voltage can be ? set by regulator API or even by the bootloader.

Quoted 3 lines

Anyway, to make sure it is set to the appropriate level, you need to measure it. Usually, this line has some capacitors on it, so it shouldn't be a hard task

Thanks for the information. I'll check my settings.

Quoted 10 lines

Now, regarding hardware instability:
I've seen instabilities related to the vcc_core set improperly, but it has never showed up that quickly, sometimes it took a week of testing for it to show up. But when it shows up, you can see strange things happening, like your DRAM is messed or buses fail or the cpu just freezes.

Another thing (which could be related to the patch) is the package marking. We have PXA3xx SoCs marked A2 stepping, but C624, which means that the highest running frequency should not exceed 624MHz.

This means not all PXA310 A2 stepping supports 806Mhz. Can we differentiate the Cxxx option in software?

Quoted 1 line

What is your PXA3xx A2 marking in respect to Cxxx option?

Mine is 88AP310-A2-BGK2C806.

Regards,
Axel

#5 lore raw
Quoted 11 lines
Quoted 3 lines

Another thing (which could be related to the patch) is the package marking. We have PXA3xx SoCs marked A2 stepping, but C624, which means that the highest running frequency should not exceed 624MHz.

This means not all PXA310 A2 stepping supports 806Mhz. Can we differentiate the Cxxx option in software?

Quoted 1 line

What is your PXA3xx A2 marking in respect to Cxxx option?

Mine is 88AP310-A2-BGK2C806.

Haojian,

Can you help check on the stepping/markings on these processors to see which of them supports 806MHz, reliably?

- eric

#7 lore raw
In reply to Axel Lin
Quoted 6 lines · Axel Lin

2011/2/22 Igor Grinberg <[email protected]>:

Quoted 3 lines

Another thing (which could be related to the patch) is the package marking. We have PXA3xx SoCs marked A2 stepping, but C624, which means that the highest running frequency should not exceed 624MHz.

This means not all PXA310 A2 stepping supports 806Mhz. Can we differentiate the Cxxx option in software?

Some time ago I spoke with our local Marvell representative... and he said that this information cannot be retrieved by software.

So this makes me think, that we may be need an option for the platform data passed from the board (assuming that board should know what SoC is installed).

Or a kernel command line parameter... something much like Eric proposed:
http://www.spinics.net/lists/arm-kernel/msg92900.html

Quoted 2 lines
Quoted 1 line

What is your PXA3xx A2 marking in respect to Cxxx option?

Mine is 88AP310-A2-BGK2C806.

nice! :)

Signature · 2 lines
Regards,
Igor.
#6 lore raw
In reply to Eric Miao
Quoted 25 lines

-----Original Message-----
From: Eric Miao [mailto:eric.y.miao at gmail.com]
Sent: 2011?2?22? 9:19 PM
To: Axel Lin; Haojian Zhuang
Cc: Igor Grinberg; linux-kernel at vger.kernel.org; Russell King; linux- arm-kernel at lists.infradead.org
Subject: Re: [PATCH] ARM: pxa: support 806MHz operating points for PXA31x processors A2 stepping

Quoted 1 line
Quoted 1 line

Another thing (which could be related to the patch) is the package

marking.

Quoted 1 line
Quoted 1 line

We have PXA3xx SoCs marked A2 stepping, but C624, which means that

the

Quoted 9 lines
Quoted 1 line

highest running frequency should not exceed 624MHz.

This means not all PXA310 A2 stepping supports 806Mhz. Can we differentiate the Cxxx option in software?

Quoted 1 line

What is your PXA3xx A2 marking in respect to Cxxx option?

Mine is 88AP310-A2-BGK2C806.

Quoted 6 lines

Haojian,

Can you help check on the stepping/markings on these processors to see which of them supports 806MHz, reliably?

- eric

Hi Eric,

There's no way to recognize 806MHz on PXA310 A2 since not all A2 chips can support 806MHz. The only way to recognize 806MHz is to read marks on the top of chip.

Thanks
Haojian