Andy Shevchenko

@htot The BAR is 2M, but the rest of the PCI resources are split from that BAR IIUC. * **ALL** (remapped by firmware?): Memory at 05e00000 (32-bit, non-prefetchable) [size=2M] *...

The **0xff2xxxxx** addresses are reserved up to Audio IPs start with the offset **0xff290000**. The _pci-tng.c_ is so broken, that simple fix as proposed by @htot does *not* seem correct....

@plbossart Reverting is not the fix. The entire code assumes **wrong** start addresses, i.e. mapping of 2M starting at 0xff200000 is incorrect, and I think somebody who knows better what...

@plbossart Ah, I just read your message before the revert. The BAR I guess should be changed to 0 and then simple `pcim_iomap_bar()` or equivalent should suffice. As for HDA,...

@plbossart OK, that's what I assumed when said that _Intel never tested this_. @htot It looks like your setup is closest to the real life, I can mock up a...

> @htot @andy-shev it's unlikely that anyone in our team will have time to look into this, my only recommendation at this time is to revert the commit 5947b27 with...

> @andy-shev would you happen to know if there's a way to ignore one of the two addresses in ACPI? see above the DSDT contains two addresses for the same...

> @andy-shev we've never seen a form-factor device with e.g. two headset codecs and jacks. This sort of low-cost devices only rely on one codec, that's a safe assumption. Put...

@jwrdegoede, read above, very interesting case! It also might affect serial-multi-instantiate.c.

@jwrdegoede, thanks for your ideas! From hw perspective my understanding is that the floating pin can take any state at any time (especially taking into account temperature differences as cold...