lkml.org 
[lkml]   [2010]   [Oct]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] x86: use pgd accessors when cloning a pgd range.
 On 10/27/2010 03:40 AM, Borislav Petkov wrote:
> On Wed, Oct 27, 2010 at 09:50:13AM +0100, Ian Campbell wrote:
>> Page tables should always be updated using the proper accessor
>> methods. Not doing so bypasses the paravirt infrastructure.
>>
>> In this case the failure to do so was exposed under Xen by
>> b40827fa7268 "x86-32, mm: Add an initial page table for core
>> bootstrapping".
>>
>> Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
>> Cc: Borislav Petkov <bp@alien8.de>
>> Cc: H. Peter Anvin <hpa@linux.intel.com>
>> Cc: Jeremy Fitzhardinge <jeremy@goop.org>
>> ---
>> arch/x86/include/asm/pgtable.h | 5 ++++-
>> 1 files changed, 4 insertions(+), 1 deletions(-)
>>
>> diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
>> index ada823a..0b4c514 100644
>> --- a/arch/x86/include/asm/pgtable.h
>> +++ b/arch/x86/include/asm/pgtable.h
>> @@ -619,7 +619,10 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm,
>> */
>> static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
>> {
>> - memcpy(dst, src, count * sizeof(pgd_t));
>> + int i;
>> +
>> + for (i=0; i<count; i++)
>> + set_pgd(&dst[i], src[i]);
> Hmm, this slows down clone_pgd_range(). It is called at 3 sites total,
> two of which happen only on boot in setup_arch() so they can be ignored
> but the callchain
>
> copy_process()
> ...
> mm_init()
> |->mm_alloc_pgd()
> |->pgd_alloc()
> |->pgd_ctor()
> |->clone_pgd_range()
>
> could become noticeable. To be on the safe side, I'd make
> clone_pgd_range() a macro calling either the native or the xen version..

Frankly I'd want to see some numbers before getting too worried about
it; if it were a problem we could make the native set_pgd inlined into
the callside (it is just a memory write after all), which will be very
similar to memcpy in performance.

I am, however, more concerned about the effect on performance under
Xen. xen_set_pgd will avoid doing a hypercall in this case (the
pagetable isn't yet pinned), but it has to do a moderate amount of work
to avoid doing the hypercall, and could really add some measurable
latency to process creation (which is not something we need right now).
For that a clone_pgd_range() hypercall is the most straightforward
answer, but I'm loathe to propose that right now.

This never used to be a problem. Perhaps we can change how
clone_pgd_range is used at boot time to avoid it in the Xen case (since
we don't care about the secondary pagetable)?

J


\
 
 \ /
  Last update: 2010-10-27 18:53    [W:0.154 / U:0.016 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site