On 5/8/19 11:51 AM, Evan Green wrote:
On Tue, May 7, 2019 at 3:31 PM Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com wrote:
On 5/7/19 4:53 PM, Evan Green wrote:
Add PCI IDs for Intel CometLake platforms, which from a software point of view are extremely similar to Cannonlake platforms.
Humm, I have mixed feelings here.
Yes the hardware is nearly identical, with the exception of one detail that's not visible to the kernel, but there is no support for DMICs with the Skylake driver w/ HDaudio, and Chrome platforms are only going with SOF, so is it wise to add these two CometLake platforms to the default SND_SOC_INTEL_SKYLAKE selector, which is used by a number of distributions?
I don't mind if we add those PCI IDs and people use this driver if they wish to, but it may be time for an explicit opt-in? The SND_SOC_INTEL_SKYLAKE definition should even be pruned to mean SKL, APL, KBL and GLK, and we can add DMI-based quirks for e.g. the Up2 board and GLK Chromebooks which work with SOF.
I don't have the context here, so feel free to ignore me. But it seems like such a tiny amount of extra bits to add to make Cometlake work, and then there's no hassle for the distributions when Cometlake devices start showing up in the wild. So while things are more or less the same, why not continue piggypacking off the default? Or are you saying that the lack of DMIC support means the default should be to use a different driver?
Yes, it's the latter case, SOF will be the only driver supporting DMICs on CometLake, so it'd be better to avoid creating a conflict with SOF or enabling a configuration by default which is known to have restrictions. It's fine to add the CML IDs, but better avoid adding CML under the SKYLAKE all-you-can-eat selector.