]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commit
drm/i915/gvt: handle workload lifecycle properly
authorChuanxiao Dong <chuanxiao.dong@intel.com>
Mon, 6 Mar 2017 05:05:24 +0000 (13:05 +0800)
committerZhenyu Wang <zhenyuw@linux.intel.com>
Mon, 6 Mar 2017 09:28:52 +0000 (17:28 +0800)
commit8f1117abb408808af9cc4c948925c726bec4755a
treef2d51be82697165db1560bbb43288a60a83b8470
parent4ec3dd89052a437304e1451733c989b8cec681af
drm/i915/gvt: handle workload lifecycle properly

Currently i915 has a request replay mechanism which can make sure
the request can be replayed after a GPU reset. With this mechanism,
gvt should wait until the GVT request seqno passed before complete
the current workload. So that there should be a context switch interrupt
come before gvt free the workload. In this way, workload lifecylce
matches with the i915 request lifecycle. The workload can only be freed
after the request is completed.

v2: use gvt_dbg_sched instead of gvt_err to print when wait again

Signed-off-by: Chuanxiao Dong <chuanxiao.dong@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
drivers/gpu/drm/i915/gvt/scheduler.c