Robert Garrett wrote:
I would like to add that even with the sysctl, he had issues with it working
intermttently.
RG
It turns out that the MADT patch did not have anything to do with my
issue. The crash-on-boot for x86_64 appeared in the 16th September
snapshot, 9 days before this patch was committed.
The cause was that the minimum memory requirement for booting x86_64
jumped to 640MB of ram, and my virtualbox was only set to 384MB, which
was fine before.
Matt tracked down the cause and added these commits last night to
address it:
c8b476401ffec29b65d73f504ca744106d8bca4b: kernel - Increase MSGBUF_SIZE
from 64K to 128K
906c754ce7d9e9b435f77dba229ad08f65d2ea4a: kernel - Move action_list out
of vm_page, and change flags from 16->32 bits
722871d34e51e1c7f2e4faec1ae2bca0e3a8f18a: kernel - Greatly reduce kernel
memory use for x86_64
The last commit message:
* Remove the kmemusage array entirely and instead store ku_pagecnt
in the vm_page structure.
This allows x86_64 to boot with 128G of KVM without eating tons of
real memory, in addition to another recent commit which reduced
real memory use w/ large KVM spaces.
The summary is that if your kernel is from ~15th sept to 28th Sept,
you'll need at least 640MB ram to boot an x86_64 system. After that,
the memory requirements are much lower. I will test my system later
today, but I expect it will boot fine with 384MB ram again.
Thanks for addressing this quickly, Matt.
Regards,
John
