)]}'
{
  "commit": "91c00924846a0034020451c280c76baa4299f9dc",
  "tree": "7124ed6706937b793a10c37a861c5fc0f2e5b348",
  "parents": [
    "45b4233caac05da0118b608a9fc2a40a9fc580cd"
  ],
  "author": {
    "name": "Dan Williams",
    "email": "dan.j.williams@intel.com",
    "time": "Tue Jan 02 13:52:30 2007 -0700"
  },
  "committer": {
    "name": "Dan Williams",
    "email": "dan.j.williams@intel.com",
    "time": "Fri Jul 13 08:06:15 2007 -0700"
  },
  "message": "md: raid5_run_ops - run stripe operations outside sh-\u003elock\n\nWhen the raid acceleration work was proposed, Neil laid out the following\nattack plan:\n\n1/ move the xor and copy operations outside spin_lock(\u0026sh-\u003elock)\n2/ find/implement an asynchronous offload api\n\nThe raid5_run_ops routine uses the asynchronous offload api (async_tx) and\nthe stripe_operations member of a stripe_head to carry out xor+copy\noperations asynchronously, outside the lock.\n\nTo perform operations outside the lock a new set of state flags is needed\nto track new requests, in-flight requests, and completed requests.  In this\nnew model handle_stripe is tasked with scanning the stripe_head for work,\nupdating the stripe_operations structure, and finally dropping the lock and\ncalling raid5_run_ops for processing.  The following flags outline the\nrequests that handle_stripe can make of raid5_run_ops:\n\nSTRIPE_OP_BIOFILL\n - copy data into request buffers to satisfy a read request\nSTRIPE_OP_COMPUTE_BLK\n - generate a missing block in the cache from the other blocks\nSTRIPE_OP_PREXOR\n - subtract existing data as part of the read-modify-write process\nSTRIPE_OP_BIODRAIN\n - copy data out of request buffers to satisfy a write request\nSTRIPE_OP_POSTXOR\n - recalculate parity for new data that has entered the cache\nSTRIPE_OP_CHECK\n - verify that the parity is correct\nSTRIPE_OP_IO\n - submit i/o to the member disks (note this was already performed outside\n   the stripe lock, but it made sense to add it as an operation type\n\nThe flow is:\n1/ handle_stripe sets STRIPE_OP_* in sh-\u003eops.pending\n2/ raid5_run_ops reads sh-\u003eops.pending, sets sh-\u003eops.ack, and submits the\n   operation to the async_tx api\n3/ async_tx triggers the completion callback routine to set\n   sh-\u003eops.complete and release the stripe\n4/ handle_stripe runs again to finish the operation and optionally submit\n   new operations that were previously blocked\n\nNote this patch just defines raid5_run_ops, subsequent commits (one per\nmajor operation type) modify handle_stripe to take advantage of this\nroutine.\n\nChangelog:\n* removed ops_complete_biodrain in favor of ops_complete_postxor and\n  ops_complete_write.\n* removed the raid5_run_ops workqueue\n* call bi_end_io for reads in ops_complete_biofill, saves a call to\n  handle_stripe\n* explicitly handle the 2-disk raid5 case (xor becomes memcpy), Neil Brown\n* fix race between async engines and bi_end_io call for reads, Neil Brown\n* remove unnecessary spin_lock from ops_complete_biofill\n* remove test_and_set/test_and_clear BUG_ONs, Neil Brown\n* remove explicit interrupt handling for channel switching, this feature\n  was absorbed (i.e. it is now implicit) by the async_tx api\n* use return_io in ops_complete_biofill\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "e372e57687ee3a3388fe9ca292ad19c61935fad2",
      "old_mode": 33188,
      "old_path": "drivers/md/raid5.c",
      "new_id": "0b7002479655008e028cb1f80ab354b424460125",
      "new_mode": 33188,
      "new_path": "drivers/md/raid5.c"
    },
    {
      "type": "modify",
      "old_id": "b99d354f6128562e2fd43f5b36049c5c28e757bb",
      "old_mode": 33188,
      "old_path": "include/linux/raid/raid5.h",
      "new_id": "6fb9d94e6f2e24bac6cef9889a2a91529b44e3da",
      "new_mode": 33188,
      "new_path": "include/linux/raid/raid5.h"
    }
  ]
}