[PATCH net-next v4] net: ti: icssg_prueth: add TAPRIO offload support

[MD Danish Anwar]

On 18/01/24 6:57 am, Vladimir Oltean wrote:
> On Mon, Jan 15, 2024 at 12:24:12PM +0530, MD Danish Anwar wrote:
>>> I believe the intention is for this code to be run before any taprio
>>> offload is added, correct? But it is possible for the user to add an
>>
>> Yes, the intention here is to run this code before any taprio offload is
>> added.
> 
> Then it is misplaced?
> 

Where should I move it then? Perhaps to end of prueth_probe()?
If this is moved to prueth_probe() then it will mean it's always called.
If user adds an offloaded Qdisc even while the netdev has not yet been
brought up, it will not result into any error

>>> offloaded Qdisc even while the netdev has not yet been brought up.
>>> Is that case handled correctly, or will it simply result in NULL pointer
>>> dereferences (tas->config_list)?
>>>
>>
>> In that case, it will eventually result in NULL pointer dereference as
>> tas->config_list will be pointing to NULL. To handle this correctly we
>> can add the below check in emac_taprio_replace().
>>
>> if (!ndev_running(ndev)) {
>> 	netdev_err(ndev, "Device is not running");
>> 	return -EINVAL;
>> }
> 
> What is the reason for which the device has to be running, other than
> your placement of icssg_qos_tas_init()?
> 

Yes, I only suggested this check for that. If icssg_qos_tas_init() is
handled correctly and called before user adds qdisc, this cheeck will no
longer be needed.

>>>> +
>>>> +	cycle_time = admin_list->cycle_time - 4; /* -4ns to compensate for IEP wraparound time */
>>>
>>> Details? Doesn't this make the phase alignment of the schedule diverge
>>> from what the user expects?
>>
>> 4ns is needed to compensate for IEP wraparound time. IEP is the clock
>> used by ICSSG driver. IEP tick is 4ns and we adjust this 4ns whenever
>> calculating cycle_time. You may refer to [1] for details on IEP driver.
> 
> What is understood by "IEP wraparound time"? Its time wraps around what?

IEP clock runs at 250 MHz, 1 tick of IEP clock = NSEC_PER_SEC /
iep->refclk_freq i.e. 1000000000 / 250000000 = 4ns.

Thus 1 tick of IEP clock is 4ns.

> It wraps around exactly once every taprio cycle of each port and that's
> why the cycle-time is compensated, or how does that work?
> 

Yes, it wraps around exactly once every taprio cycle and to compensate
for that we adjust 4ns. Instead of hardcoding I will use a varaible here.

It is a hardware errata but it is not public yet.


>>>
>>>> +	base_time = admin_list->base_time;
>>>> +	cur_time = prueth_iep_gettime(emac, &sts);
>>>> +
>>>> +	if (base_time > cur_time)
>>>> +		change_cycle_count = DIV_ROUND_UP_ULL(base_time - cur_time, cycle_time);
>>>> +	else
>>>> +		change_cycle_count = 1;
>>>
>>> I see that the base_time is only used to calculate the number of cycles
>>> relative to cur_time. Taprio users want to specify a basetime value
>>> which indicates the phase alignment of the schedule. This is important
>>> when the device is synchronized over PTP with other switches in the
>>> network. Can you explain how is the basetime taken into consideration in
>>> your implementation?
>>>
>>
>> In this implementation base_time is used only to obtain the
>> change_cycle_count and to write it to TAS_CONFIG_CHANGE_CYCLE_COUNT. In
>> this implementation base_time is not used for anything else.
> 
> So there is zero granularity in the base-time beyond the number of cycles?
> That is very bad, because it means the hardware cannot be used in a
> practical TSN network where schedules are offset in phase to each other.
> It needs to be able to be told when the schedule begins, with precision.
> Not just how many cycles from now (what does 'now' even mean?).
> 

Currently base_time is only used for calculating number of cycles.

>>> Better to say what's the hardware maximum, than to report back num_entries
>>> as being not supported.
>>>
>>
>> Sure, I'll change it to below,
>>
>> if (taprio->num_entries > TAS_MAX_CMD_LISTS) {
>> 	NL_SET_ERR_MSG_FMT_MOD(taprio->extack, "num_entries %ld is more than maximum supported entries %ld in taprio config\n",
>> 				       taprio->num_entries, TAS_MAX_CMD_LISTS);
>> 	return -EINVAL;
>> }
> 
> Keep in mind that NETLINK_MAX_FMTMSG_LEN is only 80 characters. Also, \n
> is not needed in netlink extack messages. And indentation also looks off.
> 

Sure.

>>>> +
>>>> +	emac_cp_taprio(taprio, est_new);
>>>> +	emac->qos.tas.taprio_admin = est_new;
>>>> +	ret = tas_update_oper_list(emac);
>>>> +	if (ret)
>>>> +		return ret;
>>>> +
>>>> +	ret =  tas_set_state(emac, TAS_STATE_ENABLE);
>>>> +	if (ret)
>>>> +		devm_kfree(&ndev->dev, est_new);
>>>> +
>>>> +	return ret;
>>>> +}
>>
>> Below is how the code will look like.
>>
>> 	emac->qos.tas.taprio_admin = taprio_offload_get(taprio);
> 
> emac->qos.tas.taprio_admin can also hold an old offload, which is leaked
> here when assigning the new one ("tc qdisc replace dev eth0 root taprio").
> 
>> 	ret = tas_update_oper_list(emac);
>> 	if (ret)
>> 		return ret;
>>
>> 	ret = tas_set_state(emac, TAS_STATE_ENABLE);
>> 	if (ret) {
>> 		emac->qos.tas.taprio_admin = NULL;
>> 		taprio_offload_free(taprio);
>> 	}
>>
>> 	return ret;
>>
>> Please let me know if all of these changes looks ok, I'll resend the
>> patch once you confirm. Thanks for reviewing.
> 
> Hard to say from this snippet. taprio_offload_free() will be needed from
> emac_taprio_destroy() as well.

Sure. I will do that too. I will be sharing a next revision soon. Thanks
for reviewing.

-- 
Thanks and Regards,
Danish