leandrof/linux
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

drm/amdgpu: use leaf iterator for filling PTs

Browse Source 
Less overhead and is the starting point for further cleanups and
improvements.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Christian König
2018-09-03 14:34:51 +02:00
committed by Alex Deucher
parent d4085ea9bc
commit dfa70550f5
1 changed files with 15 additions and 47 deletions
Download Patch File Download Diff File Expand all files Collapse all files

62
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
View File

@@ -1487,36 +1487,6 @@ error:
return r; return r;
} }
/**
* amdgpu_vm_find_entry - find the entry for an address
*
* @p: see amdgpu_pte_update_params definition
* @addr: virtual address in question
* @entry: resulting entry or NULL
* @parent: parent entry
*
* Find the vm_pt entry and it's parent for the given address.
*/
void amdgpu_vm_get_entry(struct amdgpu_pte_update_params *p, uint64_t addr,
struct amdgpu_vm_pt **entry,
struct amdgpu_vm_pt **parent)
{
unsigned level = p->adev->vm_manager.root_level;
*parent = NULL;
*entry = &p->vm->root;
while ((*entry)->entries) {
unsigned shift = amdgpu_vm_level_shift(p->adev, level++);
*parent = *entry;
*entry = &(*entry)->entries[addr >> shift];
addr &= (1ULL << shift) - 1;
}
if (level != AMDGPU_VM_PTB)
*entry = NULL;
}
/** /**
* amdgpu_vm_handle_huge_pages - handle updating the PD with huge pages * amdgpu_vm_handle_huge_pages - handle updating the PD with huge pages
* *
@@ -1580,36 +1550,34 @@ static int amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params,
{ {
struct amdgpu_device *adev = params->adev; struct amdgpu_device *adev = params->adev;
const uint64_t mask = AMDGPU_VM_PTE_COUNT(adev) - 1; const uint64_t mask = AMDGPU_VM_PTE_COUNT(adev) - 1;
struct amdgpu_vm_pt_cursor cursor;
uint64_t addr, pe_start;
struct amdgpu_bo *pt;
unsigned nptes;
/* walk over the address space and update the page tables */ /* walk over the address space and update the page tables */
for (addr = start; addr < end; addr += nptes, for_each_amdgpu_vm_pt_leaf(adev, params->vm, start, end - 1, cursor) {
dst += nptes * AMDGPU_GPU_PAGE_SIZE) { struct amdgpu_bo *pt = cursor.entry->base.bo;
struct amdgpu_vm_pt *entry, *parent; uint64_t pe_start;
unsigned nptes;
amdgpu_vm_get_entry(params, addr, &entry, &parent); if (!pt || cursor.level != AMDGPU_VM_PTB)
if (!entry)
return -ENOENT; return -ENOENT;
if ((addr & ~mask) == (end & ~mask)) if ((cursor.pfn & ~mask) == (end & ~mask))
nptes = end - addr; nptes = end - cursor.pfn;
else else
nptes = AMDGPU_VM_PTE_COUNT(adev) - (addr & mask); nptes = AMDGPU_VM_PTE_COUNT(adev) - (cursor.pfn & mask);
amdgpu_vm_handle_huge_pages(params, entry, parent, amdgpu_vm_handle_huge_pages(params, cursor.entry, cursor.parent,
nptes, dst, flags); nptes, dst, flags);
/* We don't need to update PTEs for huge pages */ /* We don't need to update PTEs for huge pages */
if (entry->huge) if (cursor.entry->huge) {
dst += nptes * AMDGPU_GPU_PAGE_SIZE;
continue; continue;
}
pt = entry->base.bo; pe_start = (cursor.pfn & mask) * 8;
pe_start = (addr & mask) * 8;
amdgpu_vm_update_func(params, pt, pe_start, dst, nptes, amdgpu_vm_update_func(params, pt, pe_start, dst, nptes,
AMDGPU_GPU_PAGE_SIZE, flags); AMDGPU_GPU_PAGE_SIZE, flags);
dst += nptes * AMDGPU_GPU_PAGE_SIZE;
} }
return 0; return 0;