Re: [Qemu-devel] [PATCH for-2.3 0/3] Contain drive_get() misuse

[Markus Armbruster <armbru@redhat.com>]

Adding Andreas because he's the odd fixer for pc87312.

Paolo Bonzini <pbonzini@redhat.com> writes:

> On 23/03/2015 20:09, Markus Armbruster wrote:
>> Drives defined with if!=none are for board initialization to wire up.
>> Board code calls drive_get() or similar to find them, and creates
>> devices with their qdev drive properties set accordingly.
>> 
>> Except a few devices go on a fishing expedition for a suitable backend
>> instead of exposing a drive property for board code to set: they call
>> driver_get() or drive_get_next() in their realize() or init() method.
>> Wrong.
>> 
>> We can't fix this in time for the release, so do the next best thing:
>> contain the mistakes as far as possible so they don't become ABI:
>> 
>> * Mark them all with suitable FIXME comments [PATCH 1]
>> 
>> * sdhci-pci is new, set cannot_instantiate_with_device_add_yet to make
>>   it unavailable with -device [PATCH 2]
>> 
>> * A few more aren't currently available with -device, set
>>   cannot_instantiate_with_device_add_yet to ensure they stay
>>   unavailable [PATCH 3]
>> 
>> * Left alone: m25p80-generic and its derivatives, ssi-sd, pc87312
>
> Maybe worth documenting as future incompatible changes?  These machines
> are not versioned, so it's not the end of the world to make things saner
> if somebody comes and qdevifies the SD card.

Two questions: what exactly is going to change, and where do we want to
document it?

On the former:

Use of -drive if=floppy with onboard pc87312 (machine "prep") shouldn't
be affected.  Likewise for connecting onboard m25p80-generic derivatives
with if=mtd drives, or onboard ssi-sd with if=sd.

Weird usage similar to the one you caught in time for sdhci-pci (--drive
if=sd --device sdhci-pci) would break.  It's possible when the target
has the device, and the machine type has a suitable bus.

* pc87312

  Depends on CONFIG_PC87312, set in {ppc,ppc64}-softmmu.mak.

  Requires an ISA bus.  I believe "prep" is the only machine providing
  one.  Adding a second one with --device seems unlikely to work (I
  didn't try).  If that's correct, there's nothing to document.

  If Andreas agrees, I can set cannot_instantiate_with_device_add_yet
  for pc87312 now.

* ssi-sd

  Depends on CONFIG_SSI_SD, set in arm-softmmu.mak.

  Requires an SSI bus.  At least the following targets provide one:
  collie, highbank, lm3s6965evb, lm3s811evb, midway, xilinx-zynq-a9.
  Can't exclude use of --device ssi-sd.

  I guess we want to document that --device ssi-sd will at some point
  cease to auto-connect to the next available if=sd drive and require
  the usual drive property instead.  Okay?

* m25p80-generic
 
  Depends on CONFIG_SSI_M25P80, set in
  {arm,microblaze,microblazeel}-softmmu.mak.

  Requires an SSI bus.  At least the ARM targets above provide one.
  Can't exclude use with --device.

  Document just like ssi-sd.

[...]
> Acked-by: Paolo Bonzini <pbonzini@redhat.com>

Thanks!