Re: [Qemu-block] [Qemu-devel] [RFC v4 08/13] ide: enumerate_slots implementation
On 08/18/2017 12:57 PM, Eduardo Habkost wrote: > On Wed, Aug 16, 2017 at 05:46:18PM -0400, John Snow wrote: >> >> >> On 08/14/2017 05:57 PM, Eduardo Habkost wrote: >>> Example output when using "-machine q35": >>> >>> { >>> "available": true, >>> "count": 1, >>> "device-types": [ >>> "ide-device" >>> ], >>> "hotpluggable": false, >>> "opts": [ >>> { "option": "unit", "values": 0 }, >>> { "option": "bus", "values": "ide.2" } >>> ], >>> "opts-complete": true >>> } >>> { >>> "available": false, >>> "count": 1, >>> "device": "/machine/unattached/device[19]", >>> "device-types": [ >>> "ide-device" >>> ], >>> "hotpluggable": false, >>> "opts": [ >>> { "option": "unit", "values": 1 }, >>> { "option": "bus", "values": "ide.2" } ], >>> "opts-complete": true >>> } >>> { >>> "available": true, >>> "count": 10, >>> "device-types": [ >>> "ide-device" >>> ], >>> "hotpluggable": false, >>> "opts": [ >>> { "option": "unit", "values": [ [ 0, 1 ] ] }, >> >> Hm, these unit values aren't really correct -- we do not support >> primary/secondary semantics for IDE buses on the AHCI device. I guess >> they technically exist, but you cannot use them for anything. >> >> Should I do something to "disable" or otherwise hide the unusable >> secondary unit slots for AHCI devices? > > If the device is already rejecting -device ...,unit=1, then the > bug is in my implementation of enumerate_devices. Maybe it > should just look at IDEBus::max_units to find that out? > jhuston@probe (master) ~/s/q/b/git> ./x86_64-softmmu/qemu-system-x86_64 \ -M q35 \ -nodefaults \ -device ich9-ahci,id=ahci0 \ -drive id=bar,file=/media/ext/img/f25.qcow2,if=none \ -device ide-hd,id=foo,bus=ahci0.0,drive=bar,unit=1 \ -vga std qemu-system-x86_64: -device ide-hd,id=foo,bus=ahci0.0,drive=bar,unit=1: Can't create IDE unit 1, bus supports only 1 units qemu-system-x86_64: -device ide-hd,id=foo,bus=ahci0.0,drive=bar,unit=1: Device initialization failed. based on; IDEBus.max_units as seen in: hw/ide/qdev.c line 93. Seems a bit like a hack on IDE's end -- but if your enumerate devices code has a call-in inside of the IDE layer, please do check the IDEBus::max_units property. --js
Re: [Qemu-block] [Qemu-devel] [RFC v4 08/13] ide: enumerate_slots implementation
On 08/17/2017 12:54 AM, Markus Armbruster wrote: > John Snow writes: > >> On 08/14/2017 05:57 PM, Eduardo Habkost wrote: >>> Example output when using "-machine q35": >>> >>> { >>> "available": true, >>> "count": 1, >>> "device-types": [ >>> "ide-device" >>> ], >>> "hotpluggable": false, >>> "opts": [ >>> { "option": "unit", "values": 0 }, >>> { "option": "bus", "values": "ide.2" } >>> ], >>> "opts-complete": true >>> } >>> { >>> "available": false, >>> "count": 1, >>> "device": "/machine/unattached/device[19]", >>> "device-types": [ >>> "ide-device" >>> ], >>> "hotpluggable": false, >>> "opts": [ >>> { "option": "unit", "values": 1 }, >>> { "option": "bus", "values": "ide.2" } ], >>> "opts-complete": true >>> } >>> { >>> "available": true, >>> "count": 10, >>> "device-types": [ >>> "ide-device" >>> ], >>> "hotpluggable": false, >>> "opts": [ >>> { "option": "unit", "values": [ [ 0, 1 ] ] }, >> >> Hm, these unit values aren't really correct -- we do not support >> primary/secondary semantics for IDE buses on the AHCI device. I guess >> they technically exist, but you cannot use them for anything. >> >> Should I do something to "disable" or otherwise hide the unusable >> secondary unit slots for AHCI devices? > > I think that would make this introspection more useful. > OK, let me look into how to meaningfully disable them. I didn't think they were addressable in any tangible way currently, you certainly can't put devices on them. I'm not sure where these values are coming from -- I'll play with the patches. The only reason they even technically exist is because IDEBus comes indelibly with two IDEStates baked in -- it can't be changed currently, because everything's all wonky in IDE land. I really ought to just rip off the bandaid and do an aggressive refactoring .. :|
Re: [Qemu-block] [Qemu-devel] [RFC v4 08/13] ide: enumerate_slots implementation
John Snow writes: > On 08/14/2017 05:57 PM, Eduardo Habkost wrote: >> Example output when using "-machine q35": >> >> { >> "available": true, >> "count": 1, >> "device-types": [ >> "ide-device" >> ], >> "hotpluggable": false, >> "opts": [ >> { "option": "unit", "values": 0 }, >> { "option": "bus", "values": "ide.2" } >> ], >> "opts-complete": true >> } >> { >> "available": false, >> "count": 1, >> "device": "/machine/unattached/device[19]", >> "device-types": [ >> "ide-device" >> ], >> "hotpluggable": false, >> "opts": [ >> { "option": "unit", "values": 1 }, >> { "option": "bus", "values": "ide.2" } ], >> "opts-complete": true >> } >> { >> "available": true, >> "count": 10, >> "device-types": [ >> "ide-device" >> ], >> "hotpluggable": false, >> "opts": [ >> { "option": "unit", "values": [ [ 0, 1 ] ] }, > > Hm, these unit values aren't really correct -- we do not support > primary/secondary semantics for IDE buses on the AHCI device. I guess > they technically exist, but you cannot use them for anything. > > Should I do something to "disable" or otherwise hide the unusable > secondary unit slots for AHCI devices? I think that would make this introspection more useful.