ipc: add DAI_HW_FREE IPC command #4262
Conversation
So that we can stop ssp clock in dai_reset. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com> (cherry picked from commit 7feb456)
There is an STREAM_PCM_FREE IPC command which is sent to a SOF pipeline in trigger stop. Here we add an new DAI_HW_FREE command which will be sent to SOF DAI component. The command is sent from BE DAI link so codec driver has chance to run code in the mute_stream() callback before the DAI_HW_FREE IPC command is sent. Signed-off-by: Brent Lu <brent.lu@intel.com>
We will enable/disable ssp clock in the ops. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com> (cherry picked from commit 3ffe417)
brentlu
requested review from
a team,
bardliao,
dbaluta,
lbetlej,
lgirdwood,
mmaka1 and
plbossart
as code owners
|
Although this is backwards compatible, this needs to follow ABI change process: |
| //return ipc_comp_set_value(header, COMP_CMD_LOOPBACK); | ||
| return -EINVAL; | ||
| case SOF_IPC_DAI_HW_FREE: | ||
| return ipc_msg_dai_hw_free(header); |
There was a problem hiding this comment.
For two reasons, first the PCM_FREE is 'usually' sent in trigger stop now. Second, we need to send something in BE DAI's hw_free instead of FE DAI or the SOF will still stop the bclk too early.
DROP:
sof's trigger()
=> send STREAM_TRIG_STOP
=> send STREAM_PCM_FREE
HW_FREE:
sof's hw_free() (FE)
=> too early, codec could do nothing
codec's mute_stream()
=> codec could handle it's PLL here
sof's hw_free() (BE)
=> send DAI_HW_FREE to SOF to stop bclk
codec's hw_free()
There was a problem hiding this comment.
got the issue now @brentlu
IMO we can still use the existed _IPC_STREAM_PCM_FREE but change the point of sending it, e.g. send it later, at the point where we want the clocks to be freed.
There was a problem hiding this comment.
It seems weird for me to move it from FE to BE DAI Link since this command deals with PCM pipe...
|
@brentlu lets get input from @plbossart here as this is probably more complex for the kernel than the FW. |
|
I don't understand any of this @brentlu @lgirdwood @keyonjie a) "the PCM_FREE is 'usually' sent in trigger stop now." Why is this the case? This seems like a major bug to me. |
Sorry I forgot to mention kernel PR is this one: |
|
we are not adding extra ipc command now. |
6 participants
This PR is for discussion in #4219 to solver the hw_free issue, see comment #4219 (comment)