I don't like to have that 'maxmemory' restriction. If hotplug is enabled, I
want
to be able to
add unlimited amount of memory?
Yes. (I don't have check qemu code limit, but maxmemory=2TB is working)
Not sure if we really need 2TB with hotplug? I mean, if someone really starts
a VM with
(With my last patch, I fix the static memory to 512MB(or 1GB) if you want,
so all upper memory is dimm devices.
Note that 4TB is the physical limit for 64bits system. (and for qemu)
A real 64bit system is not limited to 4TB, because 2^64 is 18446744073709551616.
But I guess current
On Sun, 25 Jan 2015 19:20:12 +0100 (CET)
Dietmar Maurer diet...@proxmox.com wrote:
A real 64bit system is not limited to 4TB, because 2^64 is
18446744073709551616.
But I guess current processors do not use 64 address lines ;-)
Current processors use 48 bits for addresses, so 2^48 -1 =
Not sure if we really need 2TB with hotplug? I mean, if someone really starts
a VM with 2TB memory, he can also use fixed assigment for that. Most VMs will
use less than 2TB ;-)
(With my last patch, I fix the static memory to 512MB(or 1GB) if you want,
so all upper memory is dimm devices.
I don't like to have that 'maxmemory' restriction. If hotplug is enabled, I
want
to be able to
add unlimited amount of memory?
Yes. (I don't have check qemu code limit, but maxmemory=2TB is working)
- Mail original -
De: dietmar diet...@proxmox.com
À: aderumier aderum...@odiso.com,
This patch allow to hotplug memory dimm modules
though a new option : dimm_memory
The dimm modules are generated from a map
dimmid size dimm_memory
dimm0128128 100.00 0
dimm1128256 50.00 1
dimm2128384 33.33
Ok,I find I have found the problem.
Windows have no problem with any memory static + dimm_memory,
but Linux want alignment on 128MB for dimm_memory.
But that's also mean that static memory need to be aligned on 128MB,
because dimm_memory start on first memory address available.
Also, without
I am also thinking about changing the 'hotplug' config like this:
hotplug: [mem|cpu|net|disk][,mem|cpu|net|disk]*
so that we can selectively enable them.
Yes, I think it could be great.
Not all guest os/kernel version support memory and cpu hotplug.
maybe to not break current config:
1:
Also, from my point of view we can require a limit of 1GB mem. I guess this
would be
safer to boot.
I have resent a patch, with 512MB min. (this works from with 4numa nodes).
But with 1GB, it could work too with 8 numa nodes.
(We need 128MB alignment for linux, details are in my patch email)
Yes. (I don't have check qemu code limit, but maxmemory=2TB is working)
But, I don't known with current mapping how we can reach any maxmemory size
with 255 slots.
maybe if use can specify maxmemory, we can do a different mapping related to
the maxmemory.
- Mail original -
De:
10 matches
Mail list logo