lkml.org 
[lkml]   [2015]   [Dec]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [lustre-devel] [PATCH] staging: lustre: Handle nodemask on UMP machines
Date
On 2015/12/21, 15:08, "lustre-devel on behalf of Greg Kroah-Hartman"
<lustre-devel-bounces@lists.lustre.org on behalf of
gregkh@linuxfoundation.org> wrote:

>On Sun, Nov 08, 2015 at 11:34:55AM -0500, James Simmons wrote:
>> For UMP and SMP machines the struct cfs_cpt_table are
>> defined differently. In the case handled by this patch
>> nodemask is defined as a integer for the UMP case and
>> as a pointer for the SMP case. This will cause a problem
>> for ost_setup which reads the nodemask directly. Instead
>> we create a UMP version of cfs_cpt_nodemask and use that
>> in ost_setup.
>>
>> Signed-off-by: James Simmons <uja.ornl@gmail.com>
>> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4199
>> Reviewed-on: http://review.whamcloud.com/9219
>> Reviewed-by: Liang Zhen <liang.zhen@intel.com>
>> Reviewed-by: Li Xi <pkuelelixi@gmail.com>
>> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
>>
>> Starting in 3.14 kernels nodemask_t was changed from a
>> a unsigned long to a linux bitmap so more than 32 cores
>> could be supported. Using set_bit in cfs_cpt_table_alloc
>> no longer compiles so this patch backports bits of the
>> node management function that use a linux bitmap back
>> end. Cleaned up libcfs bitmap.h to use the libcfs layers
>> memory allocation function. This was pulling in lustre
>> related code that was not defined.
>>
>> Signed-off-by: James Simmons <uja.ornl@gmail.com>
>> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4993
>> Reviewed-on: http://review.whamcloud.com/10332
>> Reviewed-by: Liang Zhen <liang.zhen@intel.com>
>> Reviewed-by: Bob Glossman <bob.glossman@intel.com>
>> Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
>
>What is with this crazy two sections of signed-off-by? If this was 2
>patches, make it two patches.
>
>If not, then don't do this.
>
>Also, this whole series had no numbering, so I don't know how to apply
>them, please fix and resend it.

I suspect that this is merging two separate patches so that they do not
introduce a regression when landed to master. In the past you've said
you wanted fix patches merged into the original patch for this reason.

I guess the right thing to do is to merge the Signed-off-by: lines at
the end of the combined patch, rather than just mashing the commit
messages together.

Cheers, Andreas
--
Andreas Dilger

Lustre Principal Architect
Intel High Performance Data Division




\
 
 \ /
  Last update: 2015-12-22 02:21    [W:0.059 / U:0.100 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site