qemu-system-x86_64 defaults to a memory size of 128 MB for as long as I
can remember.  This is however not even enough to load a current Debian
shipped Linux kernel.  It then fails with weird error messages, but no
clear indication what is wrong.  For example with Linux:

| EFI stub: ERROR: Failed to decompress kernel
| EFI stub: ERROR: efi_stub_entry() failed!
| BdsDxe: failed to load Boot0001 "UEFI Non-Block Boot Device" from VenMedia(1428F772-B64A-441E-B8C3-9EBDD7F893C7): Not Found

Don't you think it would be useful to raise this value somewhat, so at
least we can get more useful error indications?

Control: tag -1 + upstream

Hi!

I sort of agree, yes.  The main problem here, IMHO, is not the too small
memory size which qemu provides by default, but a lack of error reporting
in the kernel.  As far as I remember, it's always been this way, and this
very issue has always been difficult to debug - including real HW way
before qemu.

Speaking of qemu, - for a very long time, years if not decades, every
guide out there says to specify amount of memory you want to allocate
for your guest.  Maybe we can raise default RAM size for new machine
types, but I guess it will be a kind of mis-service because users will
start forgetting to specify -m again :)

And this definitely needs to be raised upstream, - this kind of default
should not be modified downstream, making debian build different from
everything else.  This should be changed in a way to be migration-
compatible between different qemu versions (though I guess this is not
very important in this context since for a migratable VM you always
should specify the ram size with most accuracy, the same at both ends).

Thanks,

/mjt

And this is a sign of not working default.

Why do you want users to be forced to set this one specific value?
Okay, you also want to specify accel=kvm, but it does not break without,
just be slow.

qemu adds new revisions of the machine models.  So sure, use this
mechanism to change defaults.  Do we then also get q35 as default for
x86?

Also you need to specify the same machine model, or?  And if this
default change is part of a new revision, this is already enforced.

Bastian