On 18/11/2024 10.08, Yunsheng Lin wrote:
On 2024/11/12 22:19, Jesper Dangaard Brouer wrote:
Yes, there seems to be many MM system internals, like the CONFIG_SPARSEMEM*
config, memory offline/online and other MM specific optimization that it
is hard to tell it is feasible.
It would be good if MM experts can clarify on this.
Yes, please. Can Alex Duyck or MM-experts point me at some code walking
entire system page table?
Then I'll write some kernel code (maybe module) that I can benchmark how
long it takes on my machine with 384GiB. I do like Alex'es suggestion,
but I want to assess the overhead of doing this on modern hardware.
After looking more closely into MM subsystem, it seems there is some existing
pattern or API to walk the entire pages from the buddy allocator subsystem,
see the kmemleak_scan() in mm/kmemleak.c:
https://elixir.bootlin.com/linux/v6.12/source/mm/kmemleak.c#L1680
I used that to walk the pages in a arm64 system with over 300GB memory,
it took about 1.3 sec to do the walking, which seems acceptable?
Yes, that seems acceptable to me.
I'll also do a test on one of my 384 GB systems.
- It took approx 0.391661 seconds.
I just deref page->pp_magic and counted the pages, not many page were
in-use (page_count(page) > 0) as machine has just been rebooted into
this kernel:
- pages=100592572 in-use:2079607
--Jesper
