[FEATURE] Implement SOF_IPC4_FW_PARAM_SYSTEM_TIME to sync host and FW timestamps

Is your feature request related to a problem? Please describe.
Host is sending SOF_IPC4_FW_PARAM_SYSTEM_TIME to synchronize the host and DSP timestamps so that timestamps in logs can be compared directly.

Describe the solution you'd like
Implement IPC4 handle to set the timestamp epoch and implement a custom timestamp generator and set this using Zephyr log_set_timestamp_func() (in order for logging subsystem to use this timestamp).

Describe alternatives you've considered
Setting the Zephyr kernel time is one option, but there is currently no API to do this from applications and this could create many issues (i.e. time going back for kernel code).

Additional context
Add any other context or screenshots about the feature request here.

@btian1

@btian1 this is one potential, please assign to yourself if you have bandwidth

@ujfalusi

@ujfalusi and I already discussed this, I can't recall what the outcome was and where this discussion took place.

@plbossart

@plbossart , @kv2019i, the kernel is sending the time when the mtrace is enabled.
The message setup starts here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/sof/ipc4-mtrace.c#n433
and the send is
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/sof/ipc4-mtrace.c#n444

The epoch is adjusted as per specification, but I'm no longer sure if we should do that and just send the time directly.

This is the SOF/Zephyr side handling of the message, right?

@kv2019i

@kv2019i IIUC, driver is sending the message but we need to make sure we get teh message to Zephyr. @abonislawski @mwasko @nashif we will need this for PSS.

@alex-cri

@alex-cri @lgirdwood @mengdonglin proposing to move to v2.8 . work not yet started, so this did not make the v2.7 cut

@plbossart

@plbossart , @kv2019i, the kernel is sending the time when the mtrace is enabled. The message setup starts here: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/sof/ipc4-mtrace.c#n433 and the send is https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/sof/ipc4-mtrace.c#n444

The epoch is adjusted as per specification, but I'm no longer sure if we should do that and just send the time directly.

@ujfalusi we need the kernel and FW work in the same time domain. In kernel log, the time is increased from the OS boot up, but the time sent for fw is UTC, epoch is 1601-01-01 00:00:00, so the time domain is still different, can we change the time to kernel log so that easy for us to check the time between kernel log and fw log ?

[    6.366234] sof-audio-pci-intel-tgl 0000:00:1f.3: ipc tx      : MOD_LARGE_CONFIG_SET [data size: 76]
[    6.366559] sof-audio-pci-intel-tgl 0000:00:1f.3: ipc tx reply: MOD_LARGE_CONFIG_SET
[    6.366593] sof-audio-pci-intel-tgl 0000:00:1f.3: ipc tx done : MOD_LARGE_CONFIG_SET [data size: 76]

The IPC4 specification is explicitly tells that the epoch for the set time is 1601-01-01 00:00:00, this is why the kernel driver converts it.

We can spell this out with some comments if needed. TBH using an epoch in FW is a waste of cycles, we shoulds use native timer value and do any epoch conversion in host.

No consensus in the review this week, pushing to v2.9, but let's try to get to main this month in any case.

All merged, close?

Navigation Menu

Repository navigation

[FEATURE] Implement SOF_IPC4_FW_PARAM_SYSTEM_TIME to sync host and FW timestamps #8016

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[FEATURE] Implement SOF_IPC4_FW_PARAM_SYSTEM_TIME to sync host and FW timestamps #8016

Description

Activity

kv2019i commented on Aug 8, 2023

plbossart commented on Aug 8, 2023

ujfalusi commented on Aug 9, 2023

ujfalusi commented on Aug 9, 2023

lgirdwood commented on Aug 16, 2023

kv2019i commented on Sep 6, 2023

RanderWang commented on Nov 7, 2023

ujfalusi commented on Nov 7, 2023

lgirdwood commented on Nov 10, 2023

kv2019i commented on Dec 1, 2023

marc-hb commented on Jan 12, 2024

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions