[PATCH v5] coresight: etm4x: Modify core-commit of cpu to avoid the overflow of HiSilicon ETM

[Suzuki K Poulose]

On 12/7/20 2:08 AM, Qi Liu wrote:
> Hi Mathieu,
> 
> On 2020/12/5 2:55, Mathieu Poirier wrote:
>> On Thu, Nov 26, 2020 at 09:34:30PM +0800, Qi Liu wrote:
>>> The ETM device can't keep up with the core pipeline when cpu core
>>> is at full speed. This may cause overflow within core and its ETM.
>>> This is a common phenomenon on ETM devices.
>>>
>>> On HiSilicon Hip08 platform, a specific feature is added to set
>>> core pipeline. So commit rate can be reduced manually to avoid ETM
>>> overflow.
>>>
>>> Signed-off-by: Qi Liu <liuqi115 at huawei.com>
>>> ---
>>> Change since v1:
>>> - add CONFIG_ETM4X_IMPDEF_FEATURE and CONFIG_ETM4X_IMPDEF_HISILICON
>>>    to keep specific feature off platforms which don't use it.
>>> Change since v2:
>>> - remove some unused variable.
>>> Change since v3:
>>> - use read/write_sysreg_s() to access register.
>>> Change since v4:
>>> - rename the call back function to a more generic name, and fix some
>>>    compile warnings.
>>>
>>>   drivers/hwtracing/coresight/Kconfig                |  9 +++
>>>   drivers/hwtracing/coresight/coresight-etm4x-core.c | 88 ++++++++++++++++++++++
>>>   drivers/hwtracing/coresight/coresight-etm4x.h      |  8 ++
>>>   3 files changed, 105 insertions(+)
>>>
>>> diff --git a/drivers/hwtracing/coresight/Kconfig b/drivers/hwtracing/coresight/Kconfig
>>> index c119824..1cc3601 100644
>>> --- a/drivers/hwtracing/coresight/Kconfig
>>> +++ b/drivers/hwtracing/coresight/Kconfig
>>> @@ -110,6 +110,15 @@ config CORESIGHT_SOURCE_ETM4X
>>>   	  To compile this driver as a module, choose M here: the
>>>   	  module will be called coresight-etm4x.
>>>
>>> +config ETM4X_IMPDEF_FEATURE
>>> +	bool "Control overflow impdef support in CoreSight ETM 4.x driver "
>>> +	depends on CORESIGHT_SOURCE_ETM4X
>>> +	help
>>> +	  This control provides overflow implement define for CoreSight
>>> +	  ETM 4.x tracer module which could not reduce commit race
>>> +	  automatically, and could avoid overflow within ETM tracer module
>>> +	  and its cpu core.
>>> +
>>>   config CORESIGHT_STM
>>>   	tristate "CoreSight System Trace Macrocell driver"
>>>   	depends on (ARM && !(CPU_32v3 || CPU_32v4 || CPU_32v4T)) || ARM64
>>> diff --git a/drivers/hwtracing/coresight/coresight-etm4x-core.c b/drivers/hwtracing/coresight/coresight-etm4x-core.c
>>> index abd706b..fcee27a 100644
>>> --- a/drivers/hwtracing/coresight/coresight-etm4x-core.c
>>> +++ b/drivers/hwtracing/coresight/coresight-etm4x-core.c
>>> @@ -3,6 +3,7 @@
>>>    * Copyright (c) 2014, The Linux Foundation. All rights reserved.
>>>    */
>>>
>>> +#include <linux/bitops.h>
>>>   #include <linux/kernel.h>
>>>   #include <linux/moduleparam.h>
>>>   #include <linux/init.h>
>>> @@ -28,7 +29,9 @@
>>>   #include <linux/perf_event.h>
>>>   #include <linux/pm_runtime.h>
>>>   #include <linux/property.h>
>>> +
>>>   #include <asm/sections.h>
>>> +#include <asm/sysreg.h>
>>>   #include <asm/local.h>
>>>   #include <asm/virt.h>
>>>
>>> @@ -103,6 +106,87 @@ struct etm4_enable_arg {
>>>   	int rc;
>>>   };
>>>
>>> +#ifdef CONFIG_ETM4X_IMPDEF_FEATURE
>>> +
>>> +#define HISI_HIP08_AMBA_ID		0x000b6d01
>>> +#define ETM4_AMBA_MASK			0xfffff
>>> +#define HISI_HIP08_CORE_COMMIT_CLEAR	0x3000
>>
>> Here bit 12 and 13 are cleared but in etm4_hisi_config_core_commit() only bit 12
>> is set - is this intentional?  What is bit 13 for?
>>
> bit 12 and 13 are used together to set core-commit, 2'b00 means cpu is at full speed,
> 2'b01, 2'b10, 2'b11 means reduce the speed of cpu pipeline, and 2'b01 means speed is
> reduced to minimum value. So bit 12 and 13 should be cleared together in
> etm4_hisi_config_core_commit().

Please could you document this in the function.

> 
> Qi
> 
>>> +#define HISI_HIP08_CORE_COMMIT_SHIFT	12
>>> +#define HISI_HIP08_CORE_COMMIT_REG	sys_reg(3, 1, 15, 2, 5)
>>> +
>>> +struct etm4_arch_features {
>>> +	void (*arch_callback)(bool enable);
>>> +};
>>> +
>>> +static bool etm4_hisi_match_pid(unsigned int id)
>>> +{
>>> +	return (id & ETM4_AMBA_MASK) == HISI_HIP08_AMBA_ID;
>>> +}
>>> +
>>> +static void etm4_hisi_config_core_commit(bool enable)
>>> +{
>>> +	u64 val;
>>> +
>>> +	val = read_sysreg_s(HISI_HIP08_CORE_COMMIT_REG);
>>> +	val &= ~HISI_HIP08_CORE_COMMIT_CLEAR;
>>> +	val |= enable << HISI_HIP08_CORE_COMMIT_SHIFT;

I would use the explicitly masked values when you update
a register.

With the above:

Reviewed-by: Suzuki K Poulose <suzuki.poulose at arm.com>