mirror of
git://nv-tegra.nvidia.com/linux-nvgpu.git
synced 2025-12-22 09:12:24 +03:00
144f548552
There are three issues with shadow domain submission: 1. runlist mem is not being swapped with mem_hw for shadow domain when there is non-shadow domain being bound to tsg which does not allow runlist to have all the tsgs. To fix this nvgpu_runlist_swap_mem() is being called for shadow domain as well. 2. tsg num_active_channels is being set as part of non-shadow domain which does happen after shadow domain. Due to this, runlist tsg length is not being set as part of runlist reconstruct and leaving tsg length 0 for last tsg. To fix this, tsg num_active_channels always get set for shadow domain as it gets configured first. 3. NV_BUILD_CONFIGURATION_VARIANT_IS_EMBEDDED is not solving the purpose to differentiate l4t and embedded_linux builds so using NV_BUILD_SYSTEM_TYPE in place of this to find out the build type. L4t is using round robin scheduling and this issue coming with manual mode scheduling so adding the fix only for manual mode scheduling support. Bug 3884011 Change-Id: Ic55da8f75294eb32c8df6e35fb1fa47df78db8f8 Signed-off-by: prsethi <prsethi@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2833880 Tested-by: mobile promotions <svcmobile_promotions@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>