On 29/12/2020 12:01, BALATON Zoltan via wrote:
Fortunately with PCI_CLASS_PROG at 0x8a Linux will keep the VIA IDE in compatible
mode and not attempt to switch to native mode: therefore if you keep this as-is and
add the legacy IDE ioports back, that just leaves the problem with BAR4 (BMDMA). In
effect your patch isn't setting compatible mode anymore, it is just disabling BMDMA.
It's actually Guenter's patch so you're now bashing him not me...
This is a deliberately misleading comment, and not a good introduction for someone
external becoming involved with the project. Guenter's patch was a PoC demonstrating
how to fix the fuloong2e machine, which is really appreciated since it clearly
locates the problems to allow a fix to be applied upstream.
(But I also think your time could be better spent than getting rid of this hack when
there are much more hacks or missing functionalities to get rid of in the part you
maintain.)
And comments like this are not appropriate for a technical mailing list either. I've
done my best to review your patch in good faith (including reading related
specifications and testing your pegasos2 model) and explain why it isn't reporting
the correct information to the guest.
Phil - I hope you find that found my review comments useful and that they explain why
the patchset is wrong by always claiming legacy IDE ioports exist but not providing
them unless the new option is set (and indeed indicating some of the shortcomings of
QEMU related to PCI BARs in this area that can be improved in future). As I feel
comments in this thread have become directed at me personally, I am going to take a
step back from this.
ATB,
Mark.
