Re: [Qemu-devel] [REBASE PATCH v5 1/2] machine: add default_ram_size to machine class

[Nikunj A Dadhania <nikunj@linux.vnet.ibm.com>]

Alexander Graf <agraf@suse.de> writes:
>> Am 30.04.2015 um 11:40 schrieb Thomas Huth <thuth@redhat.com>:
>> On Thu, 30 Apr 2015 11:18:05 +0200
>> Alexander Graf <agraf@suse.de> wrote:
>>>> On 30.04.15 06:41, Nikunj A Dadhania wrote:
>>>> 
>>>> Hi Paolo,
>>>> 
>>>> Paolo Bonzini <pbonzini@redhat.com> writes:
>>>>> On 29/04/2015 11:06, Nikunj A Dadhania wrote:
>>>>>>> so David can push both patches.
>>>>>>> 
>>>>>>> But isn't 1G a bit too much?  At least on x86 you can easily boot with 512M.
>>>>>> 
>>>>>> I understood this number as not the _minimum memory_ to boot the
>>>>>> VM. And this will only come in picture when the user has not specified
>>>>>> any memory.
>>>>> 
>>>>> This in turn will basically only happen for QEMU developers.  So keeping
>>>>> the default on the low side would make sense.
>>>>> 
>>>>> On my (4G memory) laptop I might not even be able to boot a PPC64 VM
>>>>> with 1G and TCG, but I can do that nicely with 256M.
>>>> 
>>>> That will be fine with me as well, i.e. 256M
>>>> 
>>>> David/Alex, Do you have comments on this before we change it?
>>> 
>>> I've seen RAM size combinations that seemed to work ok, but then failed
>>> during grub2 execution for example. Please verify with all reasonably
>>> realistically executed distributions that 256MB is enough.
>> 
>> Since this default value will likely be there for the next couple of
>> years, it's maybe better to use a slightly higher value than one that
>> is too low - the amount of RAM that a guest requires likely rather
>> increases in the next years instead of going down again. So I think
>> using 512 MB instead is maybe a good compromise?
>
> Again, even with 512, please verify a few different distros and check that they run.

Verified the few distro images available in my virt-test setup, we boot
fine with 512MB memory. Will send an updated patch. 

Regards
Nikunj