:
:just for curious: what is recursive pml4 mapping, found from freebsd pmap.h
:
:also, freebsd only have a page of PML4 entries, about 4K/64=64
:entries, not all 512 (2**9).
I'm not sure but I think what they are doing is mapping the page table
itself with a page table entry.
For example, under i386 if you take the page directory address and
program that into one of the page directory array slots, what you
get at that address is a representation of the page table.
[pde] -> [page_table0] -------> [normal-page-in-memory0]
[page_table1] [normal-page-in-memory1]
[page_table2] [normal-page-in-memory2]
[page_table3]
[page_table4]
[pde] -----------> [page_table-page-in-memory0]
[page_table6] [page_table-page-in-memory1]
... [page_table-page-in-memory2]
...
This allows the system to look up a page table entry directly in
memory by indexing relative to some base address.
I do not think we need to map page tables into memory because we
will have a direct physical map available and can do the translations
manually without having to load down the TLB more then it is already.
-Matt| debian developer | Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 |
| Eric W. Biederman | [PATCH 02/10] sysfs: Support for preventing unmounts. |
| Greg KH | [GIT PATCH] driver core patches against 2.6.24 |
| Linus Torvalds | Re: LSM conversion to static interface |
git: | |
| Antonio Almeida | HTB accuracy for high speed |
| David Miller | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Gerrit Renker | [PATCH 18/37] dccp: Support for Mandatory options |
| Timo Teräs | Re: xfrm_state locking regression... |
