mirror of
https://github.com/NVIDIA/open-gpu-kernel-modules.git
synced 2026-01-30 21:19:49 +00:00
580.65.06
This commit is contained in:
@@ -135,6 +135,8 @@ void uvm_pte_batch_write_ptes(uvm_pte_batch_t *batch, uvm_gpu_phys_address_t fir
|
||||
|
||||
// Updating PTEs in sysmem requires a sysmembar after writing them and
|
||||
// before any TLB invalidates.
|
||||
// PTEs should never be in a location using UVM_APERTURE_SYS_NON_COHERENT.
|
||||
UVM_ASSERT(first_pte.aperture != UVM_APERTURE_SYS_NON_COHERENT);
|
||||
if (first_pte.aperture == UVM_APERTURE_SYS)
|
||||
batch->membar = UVM_MEMBAR_SYS;
|
||||
|
||||
@@ -165,6 +167,8 @@ void uvm_pte_batch_write_pte(uvm_pte_batch_t *batch, uvm_gpu_phys_address_t pte,
|
||||
|
||||
// Updating PTEs in sysmem requires a sysmembar after writing them and
|
||||
// before any TLB invalidates.
|
||||
// PTEs should never be in a location using UVM_APERTURE_SYS_NON_COHERENT.
|
||||
UVM_ASSERT(pte.aperture != UVM_APERTURE_SYS_NON_COHERENT);
|
||||
if (pte.aperture == UVM_APERTURE_SYS)
|
||||
batch->membar = UVM_MEMBAR_SYS;
|
||||
|
||||
@@ -205,6 +209,8 @@ void uvm_pte_batch_clear_ptes(uvm_pte_batch_t *batch, uvm_gpu_phys_address_t fir
|
||||
empty_pte_bits,
|
||||
entry_size * entry_count);
|
||||
|
||||
// PTEs should never be in a location using UVM_APERTURE_SYS_NON_COHERENT.
|
||||
UVM_ASSERT(first_pte.aperture != UVM_APERTURE_SYS_NON_COHERENT);
|
||||
if (first_pte.aperture == UVM_APERTURE_SYS)
|
||||
batch->membar = UVM_MEMBAR_SYS;
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user