No longer applies.

It had been committed for quite some time.

In D7693#160050, @mav wrote:

One more downside of this CAM modification approach is that if previously user not loading hyperv drivers automatically got legacy emulation ATA, after this change -- he will get no disks at all.

Split intr_table_lock into two. One for intr sources, which is now sx. One for PIC, which stays mutex. Suggested-by: jhb

In D7784#162182, @jhb wrote:

My worry is that this has a negative impact on suspend/resume. (In general suspend/resume can't sleep).) One option perhaps would be to split the lock into a mutex protecting list of PICs (that is what suspend/resume needs) and the intr_table_lock would only protect interrupt_sources[].

If no objection comes, it will be committed early next week.

In D7715#160249, @mav wrote:

I can't say that I like this. All this area looks uglier and uglier. :(

Just curious, how is it handled in Windows guest? How Hyper-V designers supposed this functionality to be used?

I plan to commit this next Monday, if no objection comes.

In D7693#160050, @mav wrote:

One more downside of this CAM modification approach is that if previously user not loading hyperv drivers automatically got legacy emulation ATA, after this change -- he will get no disks at all.

In D7693#160046, @mav wrote:

In D7693#159890, @sepherosa_gmail.com wrote:

Pass-through support is a good argument. Hmm, how about this? We move further on the probe path to PROBE_IDENTIFY, so we can have the disk model to filter, the virtual disk is <Msft blah blah>, we just ignore those ones.

This could probably look better. There are concepts of model-based quirks on both transport and peripheral layers of CAM. There could probably be added new quirk to either completely ignore device, aborting its probe on transport layer, or only block ada driver from attachment, if there is any reason to keep ATA commands pass-through to it.

In D7693#159716, @mav wrote:

In D7693#159711, @honzhan_microsoft.com wrote:

Hyper-V manager provides 2 IDE controllers, and each controller has 2 luns. The default locations for CD/DVD and OS disk work fine. But if you change the IDE controller of CD/DVD and OS disk, or merge them to locate in the same IDE controller.

OK, so one thing that I was told never happen did happen.

Then let me remind you one more of my arguments that I was told to never happen: your patch completely disables ATA disks, not looking where they are connected, so it can easily be real ATA disk, connected to passed through PCIe SATA controller, that as I understand is going to be supported in 2016 version.

In D7691#159675, @hselasky wrote:

For example see here:
http://code.metager.de/source/xref/NetBSD/src/sys/dev/usb/if_urndis.c#589

I'd like these structures kept somewhere in case they will be needed in the future. Refer to MS USB RNDIS documents.

--HPS

In D7681#159619, @hselasky wrote:

Why not keep RNDIS instead of NDIS, like the rest of the header file does?