Re: [PATCH V3 1/3] coresight: etm: Override TRCIDR3.CCITMIN on errata affected cpus

[Suzuki K Poulose <suzuki.poulose@arm.com>]

On 11/08/2023 10:03, Mike Leach wrote:
> Hi Anshuman,
> 
> A few minor points.
> 
> On Fri, 11 Aug 2023 at 04:46, Anshuman Khandual
> <anshuman.khandual@arm.com> wrote:
>>
>> This work arounds errata 1490853 on Cortex-A76, and Neoverse-N1, errata
>> 1491015 on Cortex-A77, errata 1502854 on Cortex-X1, and errata 1619801 on
>> Neoverse-V1, based affected cpus, where software read for TRCIDR3.CCITMIN
>> field in ETM gets an wrong value.
>>
>> If software uses the value returned by the TRCIDR3.CCITMIN register field,
>> then it will limit the range which could be used for programming the ETM.
>> In reality, the ETM could be programmed with a much smaller value than what
>> is indicated by the TRCIDR3.CCITMIN field and still function correctly.
>>
>> If software reads the TRCIDR3.CCITMIN register field, corresponding to the
>> instruction trace counting minimum threshold, observe the value 0x100 or a
>> minimum cycle count threshold of 256. The correct value should be 0x4 or a
>> minimum cycle count threshold of 4.
>>
>> This work arounds the problem via storing 4 in drvdata->ccitmin on affected
>> systems where the TRCIDR3.CCITMIN has been 256, thus preserving cycle count
>> threshold granularity.
>>
>> These errata information has been updated in arch/arm64/silicon-errata.rst,
>> but without their corresponding configs because these have been implemented
>> directly in the driver.
>>
>> Cc: Catalin Marinas <catalin.marinas@arm.com>
>> Cc: Will Deacon <will@kernel.org>
>> Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
>> Cc: Mike Leach <mike.leach@linaro.org>
>> Cc: James Clark <james.clark@arm.com>
>> Cc: Jonathan Corbet <corbet@lwn.net>
>> Cc: linux-doc@vger.kernel.org
>> Cc: coresight@lists.linaro.org
>> Cc: linux-arm-kernel@lists.infradead.org
>> Cc: linux-kernel@vger.kernel.org
>> Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
>> ---
>>   Documentation/arch/arm64/silicon-errata.rst   | 10 +++++
>>   .../coresight/coresight-etm4x-core.c          | 37 +++++++++++++++++++
>>   2 files changed, 47 insertions(+)
>>
>> diff --git a/Documentation/arch/arm64/silicon-errata.rst b/Documentation/arch/arm64/silicon-errata.rst
>> index bedd3a1d7b42..b08f33eda5f1 100644
>> --- a/Documentation/arch/arm64/silicon-errata.rst
>> +++ b/Documentation/arch/arm64/silicon-errata.rst
>> @@ -107,6 +107,10 @@ stable kernels.
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Cortex-A76      | #1463225        | ARM64_ERRATUM_1463225       |
>>   +----------------+-----------------+-----------------+-----------------------------+
>> +| ARM            | Cortex-A76      | #1490853        | N/A                         |
>> ++----------------+-----------------+-----------------+-----------------------------+
>> +| ARM            | Cortex-A77      | #1491015        | N/A                         |
>> ++----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Cortex-A77      | #1508412        | ARM64_ERRATUM_1508412       |
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Cortex-A510     | #2051678        | ARM64_ERRATUM_2051678       |
>> @@ -125,6 +129,8 @@ stable kernels.
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Cortex-A715     | #2645198        | ARM64_ERRATUM_2645198       |
>>   +----------------+-----------------+-----------------+-----------------------------+
>> +| ARM            | Cortex-X1       | #1502854        | N/A                         |
>> ++----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Cortex-X2       | #2119858        | ARM64_ERRATUM_2119858       |
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Cortex-X2       | #2224489        | ARM64_ERRATUM_2224489       |
>> @@ -133,6 +139,8 @@ stable kernels.
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Neoverse-N1     | #1349291        | N/A                         |
>>   +----------------+-----------------+-----------------+-----------------------------+
>> +| ARM            | Neoverse-N1     | #1490853        | N/A                         |
>> ++----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Neoverse-N1     | #1542419        | ARM64_ERRATUM_1542419       |
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Neoverse-N2     | #2139208        | ARM64_ERRATUM_2139208       |
>> @@ -141,6 +149,8 @@ stable kernels.
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | Neoverse-N2     | #2253138        | ARM64_ERRATUM_2253138       |
>>   +----------------+-----------------+-----------------+-----------------------------+
>> +| ARM            | Neoverse-V1     | #1619801        | N/A                         |
>> ++----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | MMU-500         | #841119,826419  | N/A                         |
>>   +----------------+-----------------+-----------------+-----------------------------+
>>   | ARM            | MMU-600         | #1076982,1209401| N/A                         |
> 
> Could these doc changes not go in patch 3?

I believe no, because this patch works around the erratum and it will 
help anybody looking for the workaround using the list above + the patch
3 is not related to the erratum. It is simply documenting the new perf
option.

> 
>> diff --git a/drivers/hwtracing/coresight/coresight-etm4x-core.c b/drivers/hwtracing/coresight/coresight-etm4x-core.c
>> index 703b6fcbb6a5..1f3d29a639ff 100644
>> --- a/drivers/hwtracing/coresight/coresight-etm4x-core.c
>> +++ b/drivers/hwtracing/coresight/coresight-etm4x-core.c
>> @@ -1150,6 +1150,31 @@ static void cpu_detect_trace_filtering(struct etmv4_drvdata *drvdata)
>>          drvdata->trfcr = trfcr;
>>   }
>>
>> +/*
>> + * The following errata on applicable cpu rangess affect the CCITMIN filed
> 
> s/rangess/ranges
> 
>> + * in TCRIDR3 register. Software read for the field returns 0x100 limiting
>> + * the cycle threshold granularity, where as the right value should have
>> + * been 0x4, which is well supported in the hardware.
>> + */
>> +static struct midr_range etm_wrong_ccitmin_cpus[] = {
>> +       /* Erratum #1490853 - Cortex-A76 */
>> +       MIDR_RANGE(MIDR_CORTEX_A76, 0, 0, 4, 0),
>> +       /* Erratum #1490853 - Neoverse-N1 */
>> +       MIDR_RANGE(MIDR_NEOVERSE_N1, 0, 0, 4, 0),
>> +       /* Erratum #1491015 - Cortex-A77 */
>> +       MIDR_RANGE(MIDR_CORTEX_A77, 0, 0, 1, 0),
>> +       /* Erratum #1502854 - Cortex-X1 */
>> +       MIDR_REV(MIDR_CORTEX_X1, 0, 0),
>> +       /* Erratum #1619801 - Neoverse-V1 */
>> +       MIDR_REV(MIDR_NEOVERSE_V1, 0, 0),
>> +       {},
>> +};
>> +
>> +static bool etm4_work_around_wrong_ccitmin(void)
> 
> This is not the actual work around - perhaps this should be
> etm4_core_reads_wrong_ccitmin()?
> 
>> +{
>> +       return is_midr_in_range_list(read_cpuid_id(), etm_wrong_ccitmin_cpus);
>> +}
>> +
>>   static void etm4_init_arch_data(void *info)
>>   {
>>          u32 etmidr0;
>> @@ -1214,6 +1239,18 @@ static void etm4_init_arch_data(void *info)
>>          etmidr3 = etm4x_relaxed_read32(csa, TRCIDR3);
>>          /* CCITMIN, bits[11:0] minimum threshold value that can be programmed */
>>          drvdata->ccitmin = FIELD_GET(TRCIDR3_CCITMIN_MASK, etmidr3);
>> +       if (etm4_work_around_wrong_ccitmin()) {
>> +               /*
>> +                * Erratum affected cpus will read 256 as the minimum
>> +                * instruction trace cycle counting threshold where as
>> +                * the correct value should be 4 instead. Override the
>> +                * recorded value for 'drvdata->ccitmin' to workaround
>> +                * this problem.
>> +                */
>> +               if (drvdata->ccitmin == 256) >
> Not sure this check matters - fixed cores will be 4, non fixed cores
> as identified by the list need to be 4, we don't care what the read
> value is if the core is on the list.

To be on the safer side, we may allow a core to have ccitmin more than 
0x4 if we only fixup what is documented per erratum. So this looks fine
to me.


Suzuki

> 
>> +                       drvdata->ccitmin = 4;
>> +       }
>> +
>>          /* EXLEVEL_S, bits[19:16] Secure state instruction tracing */
>>          drvdata->s_ex_level = FIELD_GET(TRCIDR3_EXLEVEL_S_MASK, etmidr3);
>>          drvdata->config.s_ex_level = drvdata->s_ex_level;
>> --
>> 2.25.1
>>
> 
> Regards
> 
> Mike