)]}'
{
  "commit": "6d2144d355d2a532e5cc3fc12a6ba2a8d4ef15e4",
  "tree": "1c27eecfce86a166ca827bfbd2fe7a46888a7bc0",
  "parents": [
    "b6210386787728b84db25adc4f1eba70440a4c73"
  ],
  "author": {
    "name": "Joe Korty",
    "email": "joe.korty@ccur.com",
    "time": "Wed Mar 05 15:04:59 2008 -0800"
  },
  "committer": {
    "name": "Christoph Lameter",
    "email": "clameter@sgi.com",
    "time": "Thu Mar 06 16:21:50 2008 -0800"
  },
  "message": "slab: NUMA slab allocator migration bugfix\n\nNUMA slab allocator cpu migration bugfix\n\nThe NUMA slab allocator (specifically, cache_alloc_refill)\nis not refreshing its local copies of what cpu and what\nnuma node it is on, when it drops and reacquires the irq\nblock that it inherited from its caller.  As a result\nthose values become invalid if an attempt to migrate the\nprocess to another numa node occured while the irq block\nhad been dropped.\n\nThe solution is to make cache_alloc_refill reload these\nvariables whenever it drops and reacquires the irq block.\n\nThe error is very difficult to hit.  When it does occur,\none gets the following oops + stack traceback bits in\ncheck_spinlock_acquired:\n\n\tkernel BUG at mm/slab.c:2417\n\tcache_alloc_refill+0xe6\n\tkmem_cache_alloc+0xd0\n\t...\n\nThis patch was developed against 2.6.23, ported to and\ncompiled-tested only against 2.6.25-rc4.\n\nSigned-off-by: Joe Korty \u003cjoe.korty@ccur.com\u003e\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "f7faff72cf56e73a1019e9453c634a01ea7e3165",
      "old_mode": 33188,
      "old_path": "mm/slab.c",
      "new_id": "e6c698f55674c9599de52137f2597e0f59b631ce",
      "new_mode": 33188,
      "new_path": "mm/slab.c"
    }
  ]
}