lkml.org 
[lkml]   [2014]   [Apr]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/1] doc, mempolicy: Fix wrong document in numa_memory_policy.txt

Hi Randy,

On 04/11/2014 07:23 AM, Randy Dunlap wrote:
> On 04/01/2014 08:53 PM, Tang Chen wrote:
>> In document numa_memory_policy.txt, the following examples for flag
>> MPOL_F_RELATIVE_NODES are incorrect.
>>
>> For example, consider a task that is attached to a cpuset with
>> mems 2-5 that sets an Interleave policy over the same set with
>> MPOL_F_RELATIVE_NODES. If the cpuset's mems change to 3-7, the
>> interleave now occurs over nodes 3,5-6. If the cpuset's mems
>> then change to 0,2-3,5, then the interleave occurs over nodes
>> 0,3,5.
>>
>> According to the comment of the patch adding flag MPOL_F_RELATIVE_NODES,
>> the nodemasks the user specifies should be considered relative to the
>> current task's mems_allowed.
>> (https://lkml.org/lkml/2008/2/29/428)
>>
>> And according to numa_memory_policy.txt, if the user's nodemask includes
>> nodes that are outside the range of the new set of allowed nodes, then
>> the remap wraps around to the beginning of the nodemask and, if not already
>> set, sets the node in the mempolicy nodemask.
>>
>> So in the example, if the user specifies 2-5, for a task whose mems_allowed
>> is 3-7, the nodemasks should be remapped the third, fourth, fifth, sixth
>> node in mems_allowed. like the following:
>>
>> mems_allowed: 3 4 5 6 7
>>
>> relative index: 0 1 2 3 4
>> 5
>>
>> So the nodemasks should be remapped to 3,5-7, but not 3,5-6.
>>
>> And for a task whose mems_allowed is 0,2-3,5, the nodemasks should be
>> remapped to 0,2-3,5, but not 0,3,5.
>>
>> mems_allowed: 0 2 3 5
>>
>> relative index: 0 1 2 3
>> 4 5
>>
>>
>> Signed-off-by: Tang Chen<tangchen@cn.fujitsu.com>
>
> Wow. This was not an April fools joke, right?
>
> Have there been any acks of this? I haven't seen any responses to it.

Thanks for the reply. I found this problem when I was reading the doc.
I think it is wrong. And according to the original patch:

https://lkml.org/lkml/2008/2/29/428

I think it should be fixed in the above way. But if I was wrong, please
let me know, and I think we can at least improve the doc since it is
not that easy to understand.

Thanks. :)

>
> Andrew, do you want to merge it?
>
>
>> ---
>> Documentation/vm/numa_memory_policy.txt | 5 ++---
>> 1 file changed, 2 insertions(+), 3 deletions(-)
>>
>> diff --git a/Documentation/vm/numa_memory_policy.txt b/Documentation/vm/numa_memory_policy.txt
>> index 4e7da65..badb050 100644
>> --- a/Documentation/vm/numa_memory_policy.txt
>> +++ b/Documentation/vm/numa_memory_policy.txt
>> @@ -174,7 +174,6 @@ Components of Memory Policies
>> allocation fails, the kernel will search other nodes, in order of
>> increasing distance from the preferred node based on information
>> provided by the platform firmware.
>> - containing the cpu where the allocation takes place.
>>
>> Internally, the Preferred policy uses a single node--the
>> preferred_node member of struct mempolicy. When the internal
>> @@ -275,9 +274,9 @@ Components of Memory Policies
>> For example, consider a task that is attached to a cpuset with
>> mems 2-5 that sets an Interleave policy over the same set with
>> MPOL_F_RELATIVE_NODES. If the cpuset's mems change to 3-7, the
>> - interleave now occurs over nodes 3,5-6. If the cpuset's mems
>> + interleave now occurs over nodes 3,5-7. If the cpuset's mems
>> then change to 0,2-3,5, then the interleave occurs over nodes
>> - 0,3,5.
>> + 0,2-3,5.
>>
>> Thanks to the consistent remapping, applications preparing
>> nodemasks to specify memory policies using this flag should
>>
>
>


\
 
 \ /
  Last update: 2014-04-11 11:01    [W:0.103 / U:0.140 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site