lkml.org 
[lkml]   [2008]   [Feb]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Subject[PATCH 0 of 4] x86: cleanups from pmd lifetime series
Date
From
Hi Ingo,

Here's a followup set from that last batch of patches:
1. fix up the pgd_ctor merge, so that non-PAE will end up getting
kernel mappings
2. revert "optimise-pud_clear-cr3-reload"
3. only do a cr3 reload if pud_clear is being used on the active pagetable
4. update documentation about PAE tlb flushing.

The third of these makes pud_clear more robust, since it doesn't rely
on it being followed by the right kind of TLB flush. In practice it
shouldn't make any performance difference, since the only performance
critical paths pud_clear is used on are exit and execve, and they both
operate on some other pagetable at the time the old pagetable is being
pulled down.

It will generate TLB flushes in the case of a usermode process
munmapping a 1+G chunk of its address space, or something to do with
unsharing a hugetlbfs mapping. I don't think either of these are
performance critical.

Thanks,
J




\
 
 \ /
  Last update: 2008-02-01 17:43    [W:0.473 / U:0.092 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site