On 09/23/2015 01:37 PM, Peter Maydell wrote:
On 23 September 2015 at 13:00, Richard Henderson <r...@twiddle.net> wrote:
I've wondered about over-allocating on the mmap path, so that we can choose the
hugepage aligned subregion. But as far as I can tell, my kernel doesn't
allocate hugepages at all, no matter what we do. So it seems a little silly to
go so far out of the way to get an aligned buffer.
This raises the converse question of "why are we bothering with
MADV_HUGEPAGE at all?" :-)
I beg your pardon -- I was merely looking in the wrong place for the info.
/proc/pid/smap does show that nearly all of the area is using huge pages:
Main memory:
7fc130000000-7fc1b0000000 rw-p 00000000 00:00 0
Size: 2097152 kB
Rss: 88064 kB
Pss: 88064 kB
Shared_Clean: 0 kB
Shared_Dirty: 0 kB
Private_Clean: 0 kB
Private_Dirty: 88064 kB
Referenced: 88064 kB
Anonymous: 88064 kB
AnonHugePages: 88064 kB
Swap: 0 kB
KernelPageSize: 4 kB
MMUPageSize: 4 kB
Locked: 0 kB
code_gen_buffer:
7fc1d76e6000-7fc1f76e6000 rwxp 00000000 00:00 0
Size: 524288 kB
Rss: 58472 kB
Pss: 58472 kB
Shared_Clean: 0 kB
Shared_Dirty: 0 kB
Private_Clean: 0 kB
Private_Dirty: 58472 kB
Referenced: 58472 kB
Anonymous: 58472 kB
AnonHugePages: 57344 kB
Swap: 0 kB
KernelPageSize: 4 kB
MMUPageSize: 4 kB
Locked: 0 kB
r~
