Re: [PATCH V4 03/11] coresight: tmc: Extract device properties from AMBA pid based table lookup

[Anshuman Khandual <anshuman.khandual@arm.com>]

On 2/13/24 16:02, Suzuki K Poulose wrote:
> On 13/02/2024 03:13, Anshuman Khandual wrote:
>>
>>
>> On 2/12/24 17:43, Suzuki K Poulose wrote:
>>> On 23/01/2024 05:46, Anshuman Khandual wrote:
>>>> This extracts device properties from AMBA pid based table lookup. This also
>>>> defers tmc_etr_setup_caps() after the coresight device has been initialized
>>>> so that PID value can be read.
>>>>
>>>> Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
>>>> Cc: Mike Leach <mike.leach@linaro.org>
>>>> Cc: James Clark <james.clark@arm.com>
>>>> 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>
>>>> ---
>>>>    .../hwtracing/coresight/coresight-tmc-core.c  | 19 +++++++++++++------
>>>>    1 file changed, 13 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/drivers/hwtracing/coresight/coresight-tmc-core.c b/drivers/hwtracing/coresight/coresight-tmc-core.c
>>>> index 7ec5365e2b64..e71db3099a29 100644
>>>> --- a/drivers/hwtracing/coresight/coresight-tmc-core.c
>>>> +++ b/drivers/hwtracing/coresight/coresight-tmc-core.c
>>>> @@ -370,16 +370,24 @@ static inline bool tmc_etr_has_non_secure_access(struct tmc_drvdata *drvdata)
>>>>        return (auth & TMC_AUTH_NSID_MASK) == 0x3;
>>>>    }
>>>>    +#define TMC_AMBA_MASK 0xfffff
>>>> +
>>>> +static const struct amba_id tmc_ids[];
>>>> +
>>>>    /* Detect and initialise the capabilities of a TMC ETR */
>>>> -static int tmc_etr_setup_caps(struct device *parent, u32 devid, void *dev_caps)
>>>> +static int tmc_etr_setup_caps(struct device *parent, u32 devid)
>>>>    {
>>>>        int rc;
>>>> -    u32 dma_mask = 0;
>>>> +    u32 tmc_pid, dma_mask = 0;
>>>>        struct tmc_drvdata *drvdata = dev_get_drvdata(parent);
>>>> +    void *dev_caps;
>>>>          if (!tmc_etr_has_non_secure_access(drvdata))
>>>>            return -EACCES;
>>>>    +    tmc_pid = coresight_get_pid(&drvdata->csdev->access) & TMC_AMBA_MASK;
>>>> +    dev_caps = coresight_get_uci_data_from_amba(tmc_ids, tmc_pid);
>>>> +
>>>>        /* Set the unadvertised capabilities */
>>>>        tmc_etr_init_caps(drvdata, (u32)(unsigned long)dev_caps);
>>>>    @@ -497,10 +505,6 @@ static int tmc_probe(struct amba_device *adev, const struct amba_id *id)
>>>>            desc.type = CORESIGHT_DEV_TYPE_SINK;
>>>>            desc.subtype.sink_subtype = CORESIGHT_DEV_SUBTYPE_SINK_SYSMEM;
>>>>            desc.ops = &tmc_etr_cs_ops;
>>>> -        ret = tmc_etr_setup_caps(dev, devid,
>>>> -                     coresight_get_uci_data(id));
>>>> -        if (ret)
>>>> -            goto out;
>>>>            idr_init(&drvdata->idr);
>>>>            mutex_init(&drvdata->idr_mutex);
>>>>            dev_list = &etr_devs;
>>>> @@ -539,6 +543,9 @@ static int tmc_probe(struct amba_device *adev, const struct amba_id *id)
>>>>            goto out;
>>>>        }
>>>>    +    if (drvdata->config_type == TMC_CONFIG_TYPE_ETR)
>>>> +        ret = tmc_etr_setup_caps(dev, devid);
>>>> +
>>>
>>> With this change, we silently accept an ETR that may only have "SECURE" access only and crash later while we try to enable tracing. You could
>>> pass in the "access" (which is already in 'desc.access' in the original
>>> call site and deal with it ?
>>
>> Just wondering, if something like the following will help ? A failed tmc_etr_setup_caps()
>> because of failed tmc_etr_has_non_secure_access(), will unregister the coresight device
>> before returning.
>>
>> --- a/drivers/hwtracing/coresight/coresight-tmc-core.c
>> +++ b/drivers/hwtracing/coresight/coresight-tmc-core.c
>> @@ -538,8 +538,13 @@ static int __tmc_probe(struct device *dev, struct resource *res)
>>                  goto out;
>>          }
>>   -       if (drvdata->config_type == TMC_CONFIG_TYPE_ETR)
>> +       if (drvdata->config_type == TMC_CONFIG_TYPE_ETR) {
>>                  ret = tmc_etr_setup_caps(dev, devid);
>> +               if (ret) {
>> +                       coresight_unregister(drvdata->csdev);
>> +                       goto out;
>> +               }
>> +       }
> 
> Why do we move the tmc_etr_setup_caps() in the first place ? We could retain where that was and pass "desc.access" parameter rather than registering the csdev and then relying csdev->access ?

Agreed, and after implementing the changes suggested above, the entire patch
will look something like the following. Please do confirm if this looks good
enough.

--- a/drivers/hwtracing/coresight/coresight-tmc-core.c
+++ b/drivers/hwtracing/coresight/coresight-tmc-core.c
@@ -370,16 +370,25 @@ static inline bool tmc_etr_has_non_secure_access(struct tmc_drvdata *drvdata)
        return (auth & TMC_AUTH_NSID_MASK) == 0x3;
 }
 
+#define TMC_AMBA_MASK 0xfffff
+
+static const struct amba_id tmc_ids[];
+
 /* Detect and initialise the capabilities of a TMC ETR */
-static int tmc_etr_setup_caps(struct device *parent, u32 devid, void *dev_caps)
+static int tmc_etr_setup_caps(struct device *parent, u32 devid,
+                             struct csdev_access *access)
 {
        int rc;
-       u32 dma_mask = 0;
+       u32 tmc_pid, dma_mask = 0;
        struct tmc_drvdata *drvdata = dev_get_drvdata(parent);
+       void *dev_caps;
 
        if (!tmc_etr_has_non_secure_access(drvdata))
                return -EACCES;
 
+       tmc_pid = coresight_get_pid(access) & TMC_AMBA_MASK;
+       dev_caps = coresight_get_uci_data_from_amba(tmc_ids, tmc_pid);
+
        /* Set the unadvertised capabilities */
        tmc_etr_init_caps(drvdata, (u32)(unsigned long)dev_caps);
 
@@ -497,8 +506,7 @@ static int tmc_probe(struct amba_device *adev, const struct amba_id *id)
                desc.type = CORESIGHT_DEV_TYPE_SINK;
                desc.subtype.sink_subtype = CORESIGHT_DEV_SUBTYPE_SINK_SYSMEM;
                desc.ops = &tmc_etr_cs_ops;
-               ret = tmc_etr_setup_caps(dev, devid,
-                                        coresight_get_uci_data(id));
+               ret = tmc_etr_setup_caps(dev, devid, &desc.access);
                if (ret)
                        goto out;
                idr_init(&drvdata->idr);