Skip to content

hugetlb: V9 numa control of persistent huge pages alloc/free

Lee Schermerhorn linux-mm linux-numa 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest11 patches 15 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/11hugetlb: rework hstate_next_node_* functionsmm/hugetlb.c+45 -250 replies
2/11hugetlb: add nodemask arg to huge page alloc, free and surplus adjust fcnsmm/hugetlb.c+75 -523 replies
3/11hugetlb: factor init_nodemask_of_nodeinclude/linux/nodemask.h+7 -21 reply
4/11hugetlb: derive huge pages nodes allowed from task mempolicyinclude/linux/hugetlb.h include/linux/mempolicy.h kernel/sysctl.c +2 more+154 -153 replies
5/11hugetlb: accomodate reworked NODEMASK_ALLOCmm/hugetlb.c+2 -20 replies
6/11hugetlb: add generic definition of NUMA_NO_NODEarch/ia64/include/asm/numa.h arch/x86/include/asm/topology.h include/linux/numa.h+4 -51 reply
7/11hugetlb: add per node hstate attributesdrivers/base/node.c include/linux/node.h mm/hugetlb.c+295 -251 reply
8/11hugetlb: update hugetlb documentation for NUMA controls.Documentation/vm/hugetlbpage.txt+179 -880 replies
9/11hugetlb: use only nodes with memory for huge pagesDocumentation/vm/hugetlbpage.txt mm/hugetlb.c+24 -230 replies
10/11hugetlb: handle memory hot-plug eventsDocumentation/vm/hugetlbpage.txt drivers/base/node.c+50 -61 reply
11/11hugetlb: offload per node attribute registrationsdrivers/base/node.c include/linux/node.h+47 -98 replies

Reviews trailers per patch

8 of 11 patches have a Reviewed-by or Acked-by.

PatchMel GormanDavid Rientjes
1/11R-b A-b
2/11R-b A-b
3/11A-b
4/11A-b
5/11
6/11A-b A-b
7/11A-b
8/11A-b A-b
9/11
10/11A-b
11/11

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v1 thread

Cover letter [PATCH 0/11] hugetlb: V9 numa control of persistent huge pages alloc/free

PATCH 0/11 hugetlb: numa control of persistent huge pages alloc/free

Against: 2.6.31-mmotm-090925-1435 plus David Rientjes' "nodemask: make NODEMASK_ALLOC more general" patch applied

This is V9 of a series of patches to provide control over the location of the allocation and freeing of persistent huge pages on a NUMA
platform. Please consider for merging into mmotm.

This series uses two mechanisms to constrain the nodes from which persistent huge pages are allocated: 1) the task NUMA mempolicy of the task modifying a new sysctl "nr_hugepages_mempolicy", based on a suggestion by Mel Gorman; and 2) a subset of the hugepages hstate sysfs attributes have been added [in V4] to each node system device under:

	/sys/devices/node/node[0-9]*/hugepages.

The per node attibutes allow direct assignment of a huge page count on a specific node, regardless of the task's mempolicy or cpuset constraints.

V5 addressed review comments -- changes described in patch descriptions.

V6 addressed more review comments, described in the patches.

V6 also included a 3 patch series that implements an enhancement suggested
by David Rientjes: the default huge page nodes allowed mask will be the
nodes with memory rather than all on-line nodes and we will allocate per node hstate attributes only for nodes with memory. This requires that we register a memory on/off-line notifier and [un]register the attributes on transitions to/from memoryless state.

V7 addressed review comments, described in the patches, and included a new patch, originally from Mel Gorman, to define a new vm sysctl and sysfs global hugepages attribute "nr_hugepages_mempolicy" rather than apply mempolicy contraints to pool adujstments via the pre-existing "nr_hugepages". The 3 patches to restrict hugetlb to visiting only nodes with memory and to add/remove per node hstate attributes on memory hotplug completed V7.

V8 reorganized the sysctl and sysfs attribute handlers to default the nodes to default or define the nodes_allowed mask up in the handlers and pass nodes_allowed [pointer] to set_max_huge_pages(). This cleanup was suggested by David Rientjes. V8 also merged Mel Gorman's "nr_hugepages_mempolicy" back into the patch to compute nodes_allowed from mempolicy.

V8 turned out to be too large a reorg to pull off without botching something. V9 fixes these. In the meantime, David Rientjes has posted a patch to generalize NODEMASK_ALLOC. This causes a build error in my series. David provided a patch to fix the build failure. I have included David's fixup as patch NN. This causes V9 to depend on David's patch.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>