Permalink
Commits on Sep 26, 2013
  1. Merge pull request #1505 from johnelse/revert-for-hotfix

    johnelse committed Sep 26, 2013
    Revert changes since the tech preview hotfix build
  2. Revert "Rename some vGPU release parameter variables to avoid future …

    johnelse committed Sep 26, 2013
    …confusion"
    
    This reverts commit c51823d.
Commits on Sep 12, 2013
  1. Merge pull request #1494 from johnelse/ca-115031

    johnelse committed Sep 12, 2013
    Allow emulated vGPUs if IOMMU is disabled
  2. CA-115031: Allow emulated vGPUs if IOMMU is disabled

    johnelse committed Sep 9, 2013
    IOMMU is only required for passthrough vGPUs.
    
    Signed-off-by: John Else <john.else@citrix.com>
  3. Merge pull request #1493 from johnelse/vgpu-release-rename

    johnelse committed Sep 12, 2013
    Rename some vGPU release parameter variables to avoid future confusion
  4. Rename some vGPU release parameter variables to avoid future confusion

    johnelse committed Sep 9, 2013
    Signed-off-by: John Else <john.else@citrix.com>
Commits on Sep 6, 2013
  1. Merge pull request #1485 from johnelse/ca-109035-revert

    johnelse committed Sep 6, 2013
    Revert "CA-109035: Remove the correct xenstore key when PCI unplugging."
  2. Revert "CA-109035: Remove the correct xenstore key when PCI unplugging."

    johnelse committed Sep 6, 2013
    This reverts commit 1fd7e5a.
    
    Removing this patch for the tech preview since it hasn't been QA'd yet.
    
    Signed-off-by: John Else <john.else@citrix.com>
Commits on Sep 5, 2013
  1. Merge pull request #1484 from johnelse/ca-109035-pr-1061

    johnelse committed Sep 5, 2013
    Remove the correct xenstore key when PCI unplugging.
  2. CA-109035: Remove the correct xenstore key when PCI unplugging.

    jonludlam authored and johnelse committed Jun 17, 2013
    We were always attempting to remove dev-0 instead of dev-n. This
    led to multiple attempts to e.g. deassign_device for all but the
    first passed through device. It also led to a cancelled shutdown
    attempting to unplug everything again without any exception handler,
    leading to an infinite loop.
    
    Signed-off-by: Jon Ludlam <jonathan.ludlam@eu.citrix.com>
Commits on Aug 30, 2013
  1. Merge pull request #1461 from simonjbeaumont/depantsify-import-error

    johnelse committed Aug 30, 2013
    Give message to error when VGPU_type import fails
  2. Merge pull request #1470 from johnelse/ca-114551

    johnelse committed Aug 30, 2013
    Update VGPU_type objects on xapi startup if the VGPU config files change
  3. CA-114551: Add unit test

    johnelse committed Aug 30, 2013
    find_or_create-ing a VGPU type, then find_or_create-ing the same type
    but with modified framebuffer_size and/or max_heads and/or size, should
    result in the original VGPU type being updated with the new values.
    
    Signed-off-by: John Else <john.else@citrix.com>
  4. CA-114551: Allow framebuffer_size to change

    johnelse committed Aug 29, 2013
    framebuffer_size used to be one of the invariants which we used to look
    up existing VGPU types - now we update the existing VGPU type with the
    new framebuffer_size.
    
    Signed-off-by: John Else <john.else@citrix.com>
  5. CA-114551: Update VGPU_type if the config has changed

    johnelse committed Aug 29, 2013
    max_instances or max_heads may have changed since we last read the
    config file. In future, we may also want to update internal_config at
    this point.
    
    Signed-off-by: John Else <john.else@citrix.com>
Commits on Aug 28, 2013
  1. Merge pull request #1467 from johnelse/ca-114501

    johnelse committed Aug 28, 2013
    Enable newly-supported VGPU types
  2. CA-114501: Enable newly-supported VGPU types

    johnelse committed Aug 28, 2013
    Newly-supported VGPU types will be automatically enabled on the PGPUs on
    which they are supported.
    
    VGPU types which were already supported will not have their enabled
    status changed.
    
    Signed-off-by: John Else <john.else@citrix.com>
Commits on Aug 27, 2013
  1. Merge pull request #1464 from simonjbeaumont/cp-5923-vm-helpers

    johnelse committed Aug 27, 2013
    Fix host choice for vGPU allocation (and unit tests)
  2. Unit tests for Xapi_vm_helpers (vGPU related)

    simonjbeaumont committed Aug 27, 2013
    Tests for assert_gpus_available and for group_hosts_by_best_pgpu_in_group.
    
    Signed-off-by: Si Beaumont <simon.beaumont@citrix.com>
  3. Fix host choice for vGPU allocation

    simonjbeaumont committed Aug 27, 2013
    The ordering for the group_by function when grouping hosts based on "best" PGPU
    was not specified. Now the hosts should be ordered and grouped correctly
    according to GPU_group.allocation_algorithm.
    
    Signed-off-by: Si Beaumont <simon.beaumont@citrix.com>
  4. Factor out grouping of hosts by optimal PGPU

    simonjbeaumont committed Aug 23, 2013
    ...for unit testing.
    
    Signed-off-by: Si Beaumont <simon.beaumont@citrix.com>
  5. Merge pull request #1460 from simonjbeaumont/fix-passthru-allocation

    johnelse committed Aug 27, 2013
    Check PGPU not in use before using for passthrough
Commits on Aug 23, 2013
  1. Give message to error when VGPU_type import fails

    simonjbeaumont committed Aug 23, 2013
    This was raising an error with an empty string. Now it follows the same pattern
    as the error paths for importing other types (see GPU group import).
    
    Signed-off-by: Si Beaumont <simon.beaumont@citrix.com>
  2. Log clearly when creating GPU-passthough and vGPU

    simonjbeaumont committed Aug 23, 2013
    Signed-off-by: Si Beaumont <simon.beaumont@citrix.com>
  3. Remove redundant check for allowed VGPU type

    simonjbeaumont committed Aug 23, 2013
    The call to assert_capacity_exists_for_VGPU_type will make this call. Note that
    assert_capacity_exists_for_VGPU_type returns either VGPU_TYPE_NOT_ALLOWED or
    PGPU_INSUFFICIENT_CAPACITY_FOR_VGPU but we are currently ignoring this and
    raising VM_REQUIRES_VGPU if the VGPU cannot be allocated on any PGPU on the
    host.
    
    Signed-off-by: Si Beaumont <simon.beaumont@citrix.com>
  4. Check PGPU not in use before using for passthrough

    simonjbeaumont committed Aug 23, 2013
    This is done with the same mutex used for allocating vGPUs.
    
    Signed-off-by: Si Beaumont <simon.beaumont@citrix.com>
  5. Merge pull request #1459 from johnelse/pgpu-helpers-more-tests

    johnelse committed Aug 23, 2013
    More unit tests for Xapi_pgpu_helpers.ml
  6. CP-5923: Add more tests of the PGPU capacity checks

    johnelse committed Aug 23, 2013
    Signed-off-by: John Else <john.else@citrix.com>
  7. Simplify the pgpu_helpers unit tests

    johnelse committed Aug 23, 2013
    Signed-off-by: John Else <john.else@citrix.com>
  8. CP-5923: Add passthrough VGPU types to PGPU capacity checks

    johnelse committed Aug 23, 2013
    Signed-off-by: John Else <john.else@citrix.com>
  9. CP-5923: More unit tests

    johnelse committed Aug 21, 2013
    Signed-off-by: John Else <john.else@citrix.com>
  10. Improve unit tests for passthrough VGPUs

    johnelse committed Aug 23, 2013
    When creating a passthrough VGPU on a PGPU, create a VM and mark it as
    attached to the PGPU's PCI device.
    
    Signed-off-by: John Else <john.else@citrix.com>
  11. Rename a group of tests

    johnelse committed Aug 23, 2013
    Signed-off-by: John Else <john.else@citrix.com>
  12. Merge pull request #1454 from johnelse/pgpu-error-handling

    johnelse committed Aug 23, 2013
    Improve error handling for PGPU/GPU_group capacity tracking