drm/radeon/kms: enable writeback (v2)
When writeback is enabled, the GPU shadows writes to certain
registers into a buffer in memory. The driver can then read
the values from the shadow rather than reading back from the
register across the bus. Writeback can be disabled by setting
the no_wb module param to 1.
On r6xx/r7xx/evergreen, the following registers are shadowed:
- CP scratch registers
- CP read pointer
- IH write pointer
On r1xx-rr5xx, the following registers are shadowed:
- CP scratch registers
- CP read pointer
v2:
- Combine wb patches for r6xx-evergreen and r1xx-r5xx
- Writeback is disabled on AGP boards since it tends to be
unreliable on AGP using the gart.
- Check radeon_wb_init return values properly.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c
index 927509f..fbce58b 100644
--- a/drivers/gpu/drm/radeon/r600.c
+++ b/drivers/gpu/drm/radeon/r600.c
@@ -1918,6 +1918,7 @@
void r600_cp_stop(struct radeon_device *rdev)
{
WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1));
+ WREG32(SCRATCH_UMSK, 0);
}
int r600_init_microcode(struct radeon_device *rdev)
@@ -2150,7 +2151,7 @@
/* Set ring buffer size */
rb_bufsz = drm_order(rdev->cp.ring_size / 8);
- tmp = RB_NO_UPDATE | (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
+ tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
#ifdef __BIG_ENDIAN
tmp |= BUF_SWAP_32BIT;
#endif
@@ -2164,8 +2165,19 @@
WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA);
WREG32(CP_RB_RPTR_WR, 0);
WREG32(CP_RB_WPTR, 0);
- WREG32(CP_RB_RPTR_ADDR, rdev->cp.gpu_addr & 0xFFFFFFFF);
- WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->cp.gpu_addr));
+
+ /* set the wb address whether it's enabled or not */
+ WREG32(CP_RB_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC);
+ WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF);
+ WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF);
+
+ if (rdev->wb.enabled)
+ WREG32(SCRATCH_UMSK, 0xff);
+ else {
+ tmp |= RB_NO_UPDATE;
+ WREG32(SCRATCH_UMSK, 0);
+ }
+
mdelay(1);
WREG32(CP_RB_CNTL, tmp);
@@ -2217,9 +2229,10 @@
int i;
rdev->scratch.num_reg = 7;
+ rdev->scratch.reg_base = SCRATCH_REG0;
for (i = 0; i < rdev->scratch.num_reg; i++) {
rdev->scratch.free[i] = true;
- rdev->scratch.reg[i] = SCRATCH_REG0 + (i * 4);
+ rdev->scratch.reg[i] = rdev->scratch.reg_base + (i * 4);
}
}
@@ -2263,70 +2276,6 @@
return r;
}
-void r600_wb_disable(struct radeon_device *rdev)
-{
- int r;
-
- WREG32(SCRATCH_UMSK, 0);
- if (rdev->wb.wb_obj) {
- r = radeon_bo_reserve(rdev->wb.wb_obj, false);
- if (unlikely(r != 0))
- return;
- radeon_bo_kunmap(rdev->wb.wb_obj);
- radeon_bo_unpin(rdev->wb.wb_obj);
- radeon_bo_unreserve(rdev->wb.wb_obj);
- }
-}
-
-void r600_wb_fini(struct radeon_device *rdev)
-{
- r600_wb_disable(rdev);
- if (rdev->wb.wb_obj) {
- radeon_bo_unref(&rdev->wb.wb_obj);
- rdev->wb.wb = NULL;
- rdev->wb.wb_obj = NULL;
- }
-}
-
-int r600_wb_enable(struct radeon_device *rdev)
-{
- int r;
-
- if (rdev->wb.wb_obj == NULL) {
- r = radeon_bo_create(rdev, NULL, RADEON_GPU_PAGE_SIZE, true,
- RADEON_GEM_DOMAIN_GTT, &rdev->wb.wb_obj);
- if (r) {
- dev_warn(rdev->dev, "(%d) create WB bo failed\n", r);
- return r;
- }
- r = radeon_bo_reserve(rdev->wb.wb_obj, false);
- if (unlikely(r != 0)) {
- r600_wb_fini(rdev);
- return r;
- }
- r = radeon_bo_pin(rdev->wb.wb_obj, RADEON_GEM_DOMAIN_GTT,
- &rdev->wb.gpu_addr);
- if (r) {
- radeon_bo_unreserve(rdev->wb.wb_obj);
- dev_warn(rdev->dev, "(%d) pin WB bo failed\n", r);
- r600_wb_fini(rdev);
- return r;
- }
- r = radeon_bo_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb);
- radeon_bo_unreserve(rdev->wb.wb_obj);
- if (r) {
- dev_warn(rdev->dev, "(%d) map WB bo failed\n", r);
- r600_wb_fini(rdev);
- return r;
- }
- }
- WREG32(SCRATCH_ADDR, (rdev->wb.gpu_addr >> 8) & 0xFFFFFFFF);
- WREG32(CP_RB_RPTR_ADDR, (rdev->wb.gpu_addr + 1024) & 0xFFFFFFFC);
- WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + 1024) & 0xFF);
- WREG32(SCRATCH_UMSK, 0xff);
- return 0;
-}
-
void r600_fence_ring_emit(struct radeon_device *rdev,
struct radeon_fence *fence)
{
@@ -2427,6 +2376,11 @@
dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
}
+ /* allocate wb buffer */
+ r = radeon_wb_init(rdev);
+ if (r)
+ return r;
+
/* Enable IRQ */
r = r600_irq_init(rdev);
if (r) {
@@ -2445,8 +2399,7 @@
r = r600_cp_resume(rdev);
if (r)
return r;
- /* write back buffer are not vital so don't worry about failure */
- r600_wb_enable(rdev);
+
return 0;
}
@@ -2505,7 +2458,7 @@
r600_cp_stop(rdev);
rdev->cp.ready = false;
r600_irq_suspend(rdev);
- r600_wb_disable(rdev);
+ radeon_wb_disable(rdev);
r600_pcie_gart_disable(rdev);
/* unpin shaders bo */
if (rdev->r600_blit.shader_obj) {
@@ -2602,8 +2555,8 @@
if (r) {
dev_err(rdev->dev, "disabling GPU acceleration\n");
r600_cp_fini(rdev);
- r600_wb_fini(rdev);
r600_irq_fini(rdev);
+ radeon_wb_fini(rdev);
radeon_irq_kms_fini(rdev);
r600_pcie_gart_fini(rdev);
rdev->accel_working = false;
@@ -2633,8 +2586,8 @@
r600_audio_fini(rdev);
r600_blit_fini(rdev);
r600_cp_fini(rdev);
- r600_wb_fini(rdev);
r600_irq_fini(rdev);
+ radeon_wb_fini(rdev);
radeon_irq_kms_fini(rdev);
r600_pcie_gart_fini(rdev);
radeon_agp_fini(rdev);
@@ -2969,10 +2922,13 @@
ih_rb_cntl = (IH_WPTR_OVERFLOW_ENABLE |
IH_WPTR_OVERFLOW_CLEAR |
(rb_bufsz << 1));
- /* WPTR writeback, not yet */
- /*ih_rb_cntl |= IH_WPTR_WRITEBACK_ENABLE;*/
- WREG32(IH_RB_WPTR_ADDR_LO, 0);
- WREG32(IH_RB_WPTR_ADDR_HI, 0);
+
+ if (rdev->wb.enabled)
+ ih_rb_cntl |= IH_WPTR_WRITEBACK_ENABLE;
+
+ /* set the writeback address whether it's enabled or not */
+ WREG32(IH_RB_WPTR_ADDR_LO, (rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFFFFFFFC);
+ WREG32(IH_RB_WPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFF);
WREG32(IH_RB_CNTL, ih_rb_cntl);
@@ -3230,8 +3186,10 @@
{
u32 wptr, tmp;
- /* XXX use writeback */
- wptr = RREG32(IH_RB_WPTR);
+ if (rdev->wb.enabled)
+ wptr = rdev->wb.wb[R600_WB_IH_WPTR_OFFSET/4];
+ else
+ wptr = RREG32(IH_RB_WPTR);
if (wptr & RB_OVERFLOW) {
/* When a ring buffer overflow happen start parsing interrupt