)]}'
{
  "commit": "f8b12e513b953aebf30f8ff7d2de9be7e024dbbe",
  "tree": "ec261949b674283b8ba214fd2715f3a7674da11c",
  "parents": [
    "2fdc246aaf9a7fa088451ad2a72e9119b5f7f029"
  ],
  "author": {
    "name": "Christoph Hellwig",
    "email": "hch@lst.de",
    "time": "Fri Sep 04 22:44:42 2009 +0200"
  },
  "committer": {
    "name": "Rusty Russell",
    "email": "rusty@rustcorp.com.au",
    "time": "Thu Oct 22 16:39:26 2009 +1030"
  },
  "message": "virtio_blk: revert QUEUE_FLAG_VIRT addition\n\nIt seems like the addition of QUEUE_FLAG_VIRT caueses major performance\nregressions for Fedora users:\n\n\thttps://bugzilla.redhat.com/show_bug.cgi?id\u003d509383\n\thttps://bugzilla.redhat.com/show_bug.cgi?id\u003d505695\n\nwhile I can\u0027t reproduce those extreme regressions myself I think the flag\nis wrong.\n\nRationale:\n\n  QUEUE_FLAG_VIRT expands to QUEUE_FLAG_NONROT which casus the queue\n  unplugged immediately.  This is not a good behaviour for at least\n  qemu and kvm where we do have significant overhead for every\n  I/O operations.  Even with all the latested speeups (native AIO,\n  MSI support, zero copy) we can only get native speed for up to 128kb\n  I/O requests we already are down to 66% of native performance for 4kb\n  requests even on my laptop running the Intel X25-M SSD for which the\n  QUEUE_FLAG_NONROT was designed.\n  If we ever get virtio-blk overhead low enough that this flag makes\n  sense it should only be set based on a feature flag set by the host.\n\nSigned-off-by: Christoph Hellwig \u003chch@lst.de\u003e\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "43f19389647a99b4efb7e78ce556e94153189197",
      "old_mode": 33188,
      "old_path": "drivers/block/virtio_blk.c",
      "new_id": "348befaaec73466bcb037951a3c21b3d5cd77e14",
      "new_mode": 33188,
      "new_path": "drivers/block/virtio_blk.c"
    }
  ]
}