linux-kernel - Re: [PATCH v2] coresight: etm4x: Fix crash observed on Qcom ETM parts with 'Low power override'

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 14 Sep 2022 15:43:53 +0100
From:   Suzuki K Poulose <suzuki.poulose@....com>
To:     Leo Yan <leo.yan@...aro.org>,
        Bhupesh Sharma <bhupesh.sharma@...aro.org>
Cc:     coresight@...ts.linaro.org, linux-arm-kernel@...ts.infradead.org,
        mathieu.poirier@...aro.org, mike.leach@...aro.org,
        alexander.shishkin@...ux.intel.com, linux-kernel@...r.kernel.org,
        bhupesh.linux@...il.com,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Subject: Re: [PATCH v2] coresight: etm4x: Fix crash observed on Qcom ETM parts
 with 'Low power override'

On 13/09/2022 11:52, Leo Yan wrote:
> Hi Bhupesh,
> 
> On Thu, Aug 25, 2022 at 10:52:32AM +0530, Bhupesh Sharma wrote:
>> Some Qualcomm ETM implementations require skipping powering up
>> the trace unit, as the ETMs are in the same power domain as
>> their CPU cores.
>>
>> Via commit 5214b563588e ("coresight: etm4x: Add support for
>> sysreg only devices"), the setting of 'skip_power_up' flag was
>> moved after the 'etm4_init_arch_data' function is called, whereas
>> the flag value is itself used inside the function. This causes
>> a crash when ETM mode 'Low-power state behavior override' is set
>> on some Qualcomm parts.
>>
>> Fix the same.
>>
>> Fixes: 5214b563588e ("coresight: etm4x: Add support for sysreg only devices")
>> Cc: Mike Leach <mike.leach@...aro.org>
>> Cc: Suzuki K Poulose <suzuki.poulose@....com>
>> Cc: Mathieu Poirier <mathieu.poirier@...aro.org>
>> Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
>> Signed-off-by: Bhupesh Sharma <bhupesh.sharma@...aro.org>
>> ---
>>   - v1 can be seen here: https://lore.kernel.org/lkml/20220803191236.3037591-1-bhupesh.sharma@linaro.org/
>>   - Addressed the review comments from Suzuki.
>>   - Rebased on linux-next.
>>
>>   drivers/hwtracing/coresight/coresight-etm4x-core.c | 13 +++++++++++--
>>   1 file changed, 11 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/hwtracing/coresight/coresight-etm4x-core.c b/drivers/hwtracing/coresight/coresight-etm4x-core.c
>> index d39660a3e50c..14c1c7869795 100644
>> --- a/drivers/hwtracing/coresight/coresight-etm4x-core.c
>> +++ b/drivers/hwtracing/coresight/coresight-etm4x-core.c
>> @@ -977,6 +977,16 @@ static bool etm4_init_sysreg_access(struct etmv4_drvdata *drvdata,
>>   	if (!cpu_supports_sysreg_trace())
>>   		return false;
>>   
>> +	/*
>> +	 * Some Qualcomm implementations require skipping powering up the trace unit,
>> +	 * as the ETMs are in the same power domain as their CPU cores.
>> +	 *
>> +	 * Since the 'skip_power_up' flag is used inside 'etm4_init_arch_data' function,
>> +	 * initialize it before the function is called.
>> +	 */
>> +	if (fwnode_property_present(dev_fwnode(dev), "qcom,skip-power-up"))
>> +		drvdata->skip_power_up = true;
>> +
> 
> I personally think this sentence should be placed in the function
> etm4_probe(), you need to move it just before smp call
> etm4_init_arch_data(), this can allow DT property "qcom,skip-power-up"
> to be respected.

Or we could move this to init_iomem_access() and explicitly set the flag
to true in init_sysreg_access().

> 
>>   	/*
>>   	 * ETMs implementing sysreg access must implement TRCDEVARCH.
>>   	 */
>> @@ -1951,8 +1961,7 @@ static int etm4_probe(struct device *dev, void __iomem *base, u32 etm_pid)
>>   		return -EINVAL;
>>   
>>   	/* TRCPDCR is not accessible with system instructions. */
>> -	if (!desc.access.io_mem ||
>> -	    fwnode_property_present(dev_fwnode(dev), "qcom,skip-power-up"))
>> +	if (!desc.access.io_mem)
>>   		drvdata->skip_power_up = true;
> 
> I prefer to move the condition checking for "desc.access.io_mem" to
> etm4_init_sysreg_access(), this can make sure the flag skip_power_up
> is set correctly based on property of system register access.

+1

> 
> A side topic, in the mainline kernel I found the value
> "desc.access.io_mem" is always zero (see the initialized value in
> etm4_probe() and etm4_init_sysreg_access()).  Should we initialize
> desc.access.io_mem to true in etm4_probe()?

Thats not true. It is initialised at :

etm4_init_iomem_access():

	*csa = CSDEV_ACCESS_IOMEM(drvdata->base);

Where

#define CSDEV_ACCESS_IOMEM(_addr)               \
         ((struct csdev_access)  {               \
                 .io_mem         = true,         \
                 .base           = (_addr),      \
         })



Suzuki