[Qemu-devel] Bug in option parsing

[Qemu-devel] Bug in option parsing

[Martin Kletzander]

Hi everybody,

while coding the support for Jason's dump-guest-core option I realized
there is (probably) a problem with the way QEMU parses additional
machine options ('dump-guest-core', 'kvm_shadow_mem' etc.). Running QEMU
with option to -machine works ok, but using '-M' (as libvirt does) works
only w/o additional options, otherwise it ends in an error (the whole
string is probably parsed as a machine name).

Is '-M' so obsolete it shouldn't be used at all or is it just an bug? We
still use '-M' everywhere I know and '-machine' isn't compatible with
older versions and different builds of QEMU.

Should I file a bug or do we have to drop '-M' for this situations?

Have a nice day,
Martin

Re: [Qemu-devel] Bug in option parsing

[Jan Kiszka]

On 2012-08-22 17:05, Martin Kletzander wrote:
> Hi everybody,
> 
> while coding the support for Jason's dump-guest-core option I realized
> there is (probably) a problem with the way QEMU parses additional
> machine options ('dump-guest-core', 'kvm_shadow_mem' etc.). Running QEMU
> with option to -machine works ok, but using '-M' (as libvirt does) works
> only w/o additional options, otherwise it ends in an error (the whole
> string is probably parsed as a machine name).
> 
> Is '-M' so obsolete it shouldn't be used at all or is it just an bug? We
> still use '-M' everywhere I know and '-machine' isn't compatible with
> older versions and different builds of QEMU.
> 
> Should I file a bug or do we have to drop '-M' for this situations?

If libvirt uses -M just like -machine, i.e. with more than a machine
name, that would be a libvirt bug (but that would have been noticed much
earlier - did you patch something?). QEMU only keeps -M <machine-name>
around to please existing users that didn't switch yet. It is NOT an
alias for -machine.

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux

Re: [Qemu-devel] Bug in option parsing

[Martin Kletzander]

On 08/22/2012 05:15 PM, Jan Kiszka wrote:
> On 2012-08-22 17:05, Martin Kletzander wrote:
>> Hi everybody,
>>
>> while coding the support for Jason's dump-guest-core option I realized
>> there is (probably) a problem with the way QEMU parses additional
>> machine options ('dump-guest-core', 'kvm_shadow_mem' etc.). Running QEMU
>> with option to -machine works ok, but using '-M' (as libvirt does) works
>> only w/o additional options, otherwise it ends in an error (the whole
>> string is probably parsed as a machine name).
>>
>> Is '-M' so obsolete it shouldn't be used at all or is it just an bug? We
>> still use '-M' everywhere I know and '-machine' isn't compatible with
>> older versions and different builds of QEMU.
>>
>> Should I file a bug or do we have to drop '-M' for this situations?
> 
> If libvirt uses -M just like -machine, i.e. with more than a machine
> name, that would be a libvirt bug (but that would have been noticed much
> earlier - did you patch something?). QEMU only keeps -M <machine-name>
> around to please existing users that didn't switch yet. It is NOT an
> alias for -machine.
> 

There is no option *yet* that libvirt would like to use as an additional
option for machine type, but I'm working on the first one and I've hit this.

So your suggestion would be to use -machine wherever it's possible and
fallback to -M only when -machine is not supported?

Martin

Re: [Qemu-devel] Bug in option parsing

[Jan Kiszka]

On 2012-08-22 17:58, Martin Kletzander wrote:
> On 08/22/2012 05:15 PM, Jan Kiszka wrote:
>> On 2012-08-22 17:05, Martin Kletzander wrote:
>>> Hi everybody,
>>>
>>> while coding the support for Jason's dump-guest-core option I realized
>>> there is (probably) a problem with the way QEMU parses additional
>>> machine options ('dump-guest-core', 'kvm_shadow_mem' etc.). Running QEMU
>>> with option to -machine works ok, but using '-M' (as libvirt does) works
>>> only w/o additional options, otherwise it ends in an error (the whole
>>> string is probably parsed as a machine name).
>>>
>>> Is '-M' so obsolete it shouldn't be used at all or is it just an bug? We
>>> still use '-M' everywhere I know and '-machine' isn't compatible with
>>> older versions and different builds of QEMU.
>>>
>>> Should I file a bug or do we have to drop '-M' for this situations?
>>
>> If libvirt uses -M just like -machine, i.e. with more than a machine
>> name, that would be a libvirt bug (but that would have been noticed much
>> earlier - did you patch something?). QEMU only keeps -M <machine-name>
>> around to please existing users that didn't switch yet. It is NOT an
>> alias for -machine.
>>
> 
> There is no option *yet* that libvirt would like to use as an additional
> option for machine type, but I'm working on the first one and I've hit this.
> 
> So your suggestion would be to use -machine wherever it's possible and
> fallback to -M only when -machine is not supported?

Yep.

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux