)]}'
{
  "commit": "708c1bbc9d0c3e57f40501794d9b0eed29d10fce",
  "tree": "4b19caf68c420b32abdf05af3d413b1320f04fbe",
  "parents": [
    "971ada0f6659488c3f36aed4c6f7670ff5ce4368"
  ],
  "author": {
    "name": "Miao Xie",
    "email": "miaox@cn.fujitsu.com",
    "time": "Mon May 24 14:32:07 2010 -0700"
  },
  "committer": {
    "name": "Linus Torvalds",
    "email": "torvalds@linux-foundation.org",
    "time": "Tue May 25 08:06:57 2010 -0700"
  },
  "message": "mempolicy: restructure rebinding-mempolicy functions\n\nNick Piggin reported that the allocator may see an empty nodemask when\nchanging cpuset\u0027s mems[1].  It happens only on the kernel that do not do\natomic nodemask_t stores.  (MAX_NUMNODES \u003e BITS_PER_LONG)\n\nBut I found that there is also a problem on the kernel that can do atomic\nnodemask_t stores.  The problem is that the allocator can\u0027t find a node to\nalloc page when changing cpuset\u0027s mems though there is a lot of free\nmemory.  The reason is like this:\n\n(mpol: mempolicy)\n\ttask1\t\t\ttask1\u0027s mpol\ttask2\n\talloc page\t\t1\n\t  alloc on node0? NO\t1\n\t\t\t\t1\t\tchange mems from 1 to 0\n\t\t\t\t1\t\trebind task1\u0027s mpol\n\t\t\t\t0-1\t\t  set new bits\n\t\t\t\t0\t  \t  clear disallowed bits\n\t  alloc on node1? NO\t0\n\t  ...\n\tcan\u0027t alloc page\n\t  goto oom\n\nI can use the attached program reproduce it by the following step:\n\n# mkdir /dev/cpuset\n# mount -t cpuset cpuset /dev/cpuset\n# mkdir /dev/cpuset/1\n# echo `cat /dev/cpuset/cpus` \u003e /dev/cpuset/1/cpus\n# echo `cat /dev/cpuset/mems` \u003e /dev/cpuset/1/mems\n# echo $$ \u003e /dev/cpuset/1/tasks\n# numactl --membind\u003d`cat /dev/cpuset/mems` ./cpuset_mem_hog \u003cnr_tasks\u003e \u0026\n   \u003cnr_tasks\u003e \u003d max(nr_cpus - 1, 1)\n# killall -s SIGUSR1 cpuset_mem_hog\n# ./change_mems.sh\n\nseveral hours later, oom will happen though there is a lot of free memory.\n\nThis patchset fixes this problem by expanding the nodes range first(set\nnewly allowed bits) and shrink it lazily(clear newly disallowed bits).  So\nwe use a variable to tell the write-side task that read-side task is\nreading nodemask, and the write-side task clears newly disallowed nodes\nafter read-side task ends the current memory allocation.\n\nThis patch:\n\nIn order to fix no node to alloc memory, when we want to update mempolicy\nand mems_allowed, we expand the set of nodes first (set all the newly\nnodes) and shrink the set of nodes lazily(clean disallowed nodes), But the\nmempolicy\u0027s rebind functions may breaks the expanding.\n\nSo we restructure the mempolicy\u0027s rebind functions and split the rebind\nwork to two steps, just like the update of cpuset\u0027s mems: The 1st step:\nexpand the set of the mempolicy\u0027s nodes.  The 2nd step: shrink the set of\nthe mempolicy\u0027s nodes.  It is used when there is no real lock to protect\nthe mempolicy in the read-side.  Otherwise we can do rebind work at once.\n\nIn order to implement it, we define\n\n\tenum mpol_rebind_step {\n\t\tMPOL_REBIND_ONCE,\n\t\tMPOL_REBIND_STEP1,\n\t\tMPOL_REBIND_STEP2,\n\t\tMPOL_REBIND_NSTEP,\n\t};\n\nIf the mempolicy needn\u0027t be updated by two steps, we can pass\nMPOL_REBIND_ONCE to the rebind functions.  Or we can pass\nMPOL_REBIND_STEP1 to do the first step of the rebind work and pass\nMPOL_REBIND_STEP2 to do the second step work.\n\nBesides that, it maybe long time between these two step and we have to\nrelease the lock that protects mempolicy and mems_allowed.  If we hold the\nlock once again, we must check whether the current mempolicy is under the\nrebinding (the first step has been done) or not, because the task may\nalloc a new mempolicy when we don\u0027t hold the lock.  So we defined the\nfollowing flag to identify it:\n\n#define MPOL_F_REBINDING (1 \u003c\u003c 2)\n\nThe new functions will be used in the next patch.\n\nSigned-off-by: Miao Xie \u003cmiaox@cn.fujitsu.com\u003e\nCc: David Rientjes \u003crientjes@google.com\u003e\nCc: Nick Piggin \u003cnpiggin@suse.de\u003e\nCc: Paul Menage \u003cmenage@google.com\u003e\nCc: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nCc: Hugh Dickins \u003chugh.dickins@tiscali.co.uk\u003e\nCc: Ravikiran Thirumalai \u003ckiran@scalex86.org\u003e\nCc: KOSAKI Motohiro \u003ckosaki.motohiro@jp.fujitsu.com\u003e\nCc: Christoph Lameter \u003ccl@linux-foundation.org\u003e\nCc: Andi Kleen \u003candi@firstfloor.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "1cc966cd3e5fff90b3cf893a2f91a6ec8747b0f7",
      "old_mode": 33188,
      "old_path": "include/linux/mempolicy.h",
      "new_id": "7b9ef6bf45aa3c6648bf344f2fff53335a976820",
      "new_mode": 33188,
      "new_path": "include/linux/mempolicy.h"
    },
    {
      "type": "modify",
      "old_id": "9a50c5f6e727f3f77ec5dcf2d5f993fc62717ce9",
      "old_mode": 33188,
      "old_path": "kernel/cpuset.c",
      "new_id": "db0990ac3fac98970fe84f143c120716ef22f5d9",
      "new_mode": 33188,
      "new_path": "kernel/cpuset.c"
    },
    {
      "type": "modify",
      "old_id": "0c73c8b814cdbc0815d6eb690a82dda2ed3f3d77",
      "old_mode": 33188,
      "old_path": "mm/mempolicy.c",
      "new_id": "8a993db8802999042a7c76a9e114f70dab5f5f83",
      "new_mode": 33188,
      "new_path": "mm/mempolicy.c"
    }
  ]
}