Repository navigation
volume: fix cache invalidation in volume_set_config() - #6274
Conversation
With IPC4 volume_set_config() is called from module_set_large_config() with the host mailbox used as a configuration source. Data there is written by the host, therefore any DSP cache contents for that area is invalid. volume_set_config() does invalidate local cache for that area but it's doing it too late - after calling module_set_configuration() which already reads the data. This leads to sporadic wrong configuration readings. Move cache invalidation before module_set_configuration() to fix that. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
lgirdwood
left a comment
lgirdwood
left a comment
There was a problem hiding this comment.
Can we do all set/get large config cache mgmt in the IPC layer ?
|
CI: licencing server connection failed in https://sof-ci.01.org/sofpr/PR6274/build1517/build/cnl_xcc_error.txt |
@lgirdwood that was also my preference and I tried to do that, I really did. But it turned out to be more complicated than I expected and potentially more risky. To move this cache management up into the caller would be a larger change because of code reuse. Just to explain a bit - |
juimonen
left a comment
juimonen
left a comment
There was a problem hiding this comment.
so I vote to merge this now and do bigger changes in later pr.
|
@marc-hb @keqiaozhang fyi - we are still building for older platforms with xcc and gcc. See https://sof-ci.01.org/sofpr/PR6274/build1517/build/ |
|
With IPC4 volume_set_config() is called from module_set_large_config() with the host mailbox used as a configuration source. Data there is written by the host, therefore any DSP cache contents for that area is invalid. volume_set_config() does invalidate local cache for that area but it's doing it too late - after calling module_set_configuration() which already reads the data. This leads to sporadic wrong configuration readings. Move cache invalidation before module_set_configuration() to fix that.