)]}'
{
  "commit": "8a3360f06cd4d3b1d57f73e029f2f8b6057fdcba",
  "tree": "299eee7ef2848e29cec8894fd9e13c7151771c21",
  "parents": [
    "2224661494278bfc1c35b392cf6ee6f58e1d5e64"
  ],
  "author": {
    "name": "Bernhard Walle",
    "email": "bwalle@suse.de",
    "time": "Wed May 14 15:40:40 2008 -0700"
  },
  "committer": {
    "name": "Tony Luck",
    "email": "tony.luck@intel.com",
    "time": "Wed May 14 15:40:40 2008 -0700"
  },
  "message": "[IA64] Don\u0027t reserve crashkernel memory \u003e 4 GB\n\nSome IA64 machines map all cell-local memory above 4 GB (32 bit limit).\nHowever, in most cases, the kernel needs some memory below that limit that is\nDMA-capable. So in this machine configuration, the crashkernel will be reserved\nabove 4 GB.\n\nFor machines that use SWIOTLB implementation because they lack an I/O MMU\nthe low memory is required by the SWIOTLB implementation. In that case,\nit doesn\u0027t make sense to reserve the crashkernel at all because it\u0027s unusable\nfor kdump.\n\nA special case is the \"hpzx1\" machine vector. In theory, it has a I/O MMU, so\nit can be booted above 4 GB. However, in the kdump case that is not possible\nbecause of changeset 51b58e3e26ebfb8cd56825c4b396ed251f51dec9:\n\n    On HP zx1 machines, the \u0027machvec\u003ddig\u0027 parameter is needed for the kdump\n    kernel to avoid problems with the HP sba iommu.  The problem is that during\n    the boot of the kdump kernel, the iommu is re-initialized, so in-flight DMA\n    from improperly shutdown drivers causes an IOTLB miss which leads to an\n    MCA.  With kdump, the idea is to get into the kdump kernel with as little\n    code as we can, so shutting down drivers properly is not an option.\n\n    The workaround is to add \u0027machvec\u003ddig\u0027 to the kdump kernel boot parameters.\n    This makes the kdump kernel avoid using the sba iommu altogether, leaving\n    the IOTLB intact.  Any ongoing DMA falls harmlessly outside the kdump\n    kernel.  After the kdump kernel reboots, all devices will have been\n    shutdown properly and DMA stopped.\n\nThis patch pushes that functionality into the sba iommu initialization\ncode, so that users won\u0027t have to find the obscure documentation telling\nthem about \u0027machvec\u003ddig\u0027.\n\nThis means that also for hpzx1 it\u0027s not possible to boot when all\nmemory is above the 4 GB limit. So the only machine vectors that can handle\nthis case are \"sn2\" and \"uv\".\n\nSigned-off-by: Bernhard Walle \u003cbwalle@suse.de\u003e\nSigned-off-by: Tony Luck \u003ctony.luck@intel.com\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "5015ca1275ca0520f4e0530e94fd57eae9b7cbef",
      "old_mode": 33188,
      "old_path": "arch/ia64/kernel/setup.c",
      "new_id": "e9596cd0cdab21595f7c5f85d7da7d2952beb5d6",
      "new_mode": 33188,
      "new_path": "arch/ia64/kernel/setup.c"
    }
  ]
}