[PATCH V4 03/11] coresight: tmc: Extract device properties from AMBA pid based table lookup
Suzuki K Poulose
suzuki.poulose at arm.com
Wed Feb 14 02:16:56 PST 2024
On 14/02/2024 03:35, Anshuman Khandual wrote:
>
>
> 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 at arm.com>
>>>>> Cc: Mike Leach <mike.leach at linaro.org>
>>>>> Cc: James Clark <james.clark at arm.com>
>>>>> Cc: coresight at lists.linaro.org
>>>>> Cc: linux-arm-kernel at lists.infradead.org
>>>>> Cc: linux-kernel at vger.kernel.org
>>>>> Signed-off-by: Anshuman Khandual <anshuman.khandual at 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);
>
That looks much better. Thanks
Suzuki
More information about the linux-arm-kernel
mailing list