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

drm/radeon: make all functions work with multiple rings.

Browse Source 
Give all asic and radeon_ring_* functions a
radeon_cp parameter, so they know the ring to work with.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This commit is contained in:
Christian König 2011-09-23 15:11:23 +02:00 committed by Dave Airlie
parent 15d3332f31
commit 7b1f2485db
20 changed files with 809 additions and 742 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

123
drivers/gpu/drm/radeon/evergreen.c
View File

@ -1311,18 +1311,20 @@ void evergreen_mc_program(struct radeon_device *rdev)
*/
void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
{
struct radeon_cp *cp = &rdev->cp;
/* set to DX10/11 mode */
radeon_ring_write(rdev, PACKET3(PACKET3_MODE_CONTROL, 0));
radeon_ring_write(rdev, 1);
radeon_ring_write(cp, PACKET3(PACKET3_MODE_CONTROL, 0));
radeon_ring_write(cp, 1);
/* FIXME: implement */
radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(cp,
#ifdef __BIG_ENDIAN
(2 << 0) |
#endif
(ib->gpu_addr & 0xFFFFFFFC));
radeon_ring_write(rdev, upper_32_bits(ib->gpu_addr) & 0xFF);
radeon_ring_write(rdev, ib->length_dw);
radeon_ring_write(cp, upper_32_bits(ib->gpu_addr) & 0xFF);
radeon_ring_write(cp, ib->length_dw);
}
@ -1360,71 +1362,73 @@ static int evergreen_cp_load_microcode(struct radeon_device *rdev)
static int evergreen_cp_start(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r, i;
uint32_t cp_me;
r = radeon_ring_lock(rdev, 7);
r = radeon_ring_lock(rdev, cp, 7);
if (r) {
DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
return r;
}
radeon_ring_write(rdev, PACKET3(PACKET3_ME_INITIALIZE, 5));
radeon_ring_write(rdev, 0x1);
radeon_ring_write(rdev, 0x0);
radeon_ring_write(rdev, rdev->config.evergreen.max_hw_contexts - 1);
radeon_ring_write(rdev, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, PACKET3(PACKET3_ME_INITIALIZE, 5));
radeon_ring_write(cp, 0x1);
radeon_ring_write(cp, 0x0);
radeon_ring_write(cp, rdev->config.evergreen.max_hw_contexts - 1);
radeon_ring_write(cp, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_unlock_commit(rdev, cp);
cp_me = 0xff;
WREG32(CP_ME_CNTL, cp_me);
r = radeon_ring_lock(rdev, evergreen_default_size + 19);
r = radeon_ring_lock(rdev, cp, evergreen_default_size + 19);
if (r) {
DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
return r;
}
/* setup clear context state */
radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(rdev, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
radeon_ring_write(cp, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(cp, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
for (i = 0; i < evergreen_default_size; i++)
radeon_ring_write(rdev, evergreen_default_state[i]);
radeon_ring_write(cp, evergreen_default_state[i]);
radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(rdev, PACKET3_PREAMBLE_END_CLEAR_STATE);
radeon_ring_write(cp, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(cp, PACKET3_PREAMBLE_END_CLEAR_STATE);
/* set clear context state */
radeon_ring_write(rdev, PACKET3(PACKET3_CLEAR_STATE, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_CLEAR_STATE, 0));
radeon_ring_write(cp, 0);
/* SQ_VTX_BASE_VTX_LOC */
radeon_ring_write(rdev, 0xc0026f00);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(cp, 0xc0026f00);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000000);
/* Clear consts */
radeon_ring_write(rdev, 0xc0036f00);
radeon_ring_write(rdev, 0x00000bc4);
radeon_ring_write(rdev, 0xffffffff);
radeon_ring_write(rdev, 0xffffffff);
radeon_ring_write(rdev, 0xffffffff);
radeon_ring_write(cp, 0xc0036f00);
radeon_ring_write(cp, 0x00000bc4);
radeon_ring_write(cp, 0xffffffff);
radeon_ring_write(cp, 0xffffffff);
radeon_ring_write(cp, 0xffffffff);
radeon_ring_write(rdev, 0xc0026900);
radeon_ring_write(rdev, 0x00000316);
radeon_ring_write(rdev, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
radeon_ring_write(rdev, 0x00000010); /* */
radeon_ring_write(cp, 0xc0026900);
radeon_ring_write(cp, 0x00000316);
radeon_ring_write(cp, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
radeon_ring_write(cp, 0x00000010); /* */
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, cp);
return 0;
}
int evergreen_cp_resume(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
u32 tmp;
u32 rb_bufsz;
int r;
@ -1442,7 +1446,7 @@ int evergreen_cp_resume(struct radeon_device *rdev)
RREG32(GRBM_SOFT_RESET);
/* Set ring buffer size */
rb_bufsz = drm_order(rdev->cp.ring_size / 8);
rb_bufsz = drm_order(cp->ring_size / 8);
tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
#ifdef __BIG_ENDIAN
tmp |= BUF_SWAP_32BIT;
@ -1456,8 +1460,8 @@ int evergreen_cp_resume(struct radeon_device *rdev)
/* Initialize the ring buffer's read and write pointers */
WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA);
WREG32(CP_RB_RPTR_WR, 0);
rdev->cp.wptr = 0;
WREG32(CP_RB_WPTR, rdev->cp.wptr);
cp->wptr = 0;
WREG32(CP_RB_WPTR, cp->wptr);
/* set the wb address wether it's enabled or not */
WREG32(CP_RB_RPTR_ADDR,
@ -1475,16 +1479,16 @@ int evergreen_cp_resume(struct radeon_device *rdev)
mdelay(1);
WREG32(CP_RB_CNTL, tmp);
WREG32(CP_RB_BASE, rdev->cp.gpu_addr >> 8);
WREG32(CP_RB_BASE, cp->gpu_addr >> 8);
WREG32(CP_DEBUG, (1 << 27) | (1 << 28));
rdev->cp.rptr = RREG32(CP_RB_RPTR);
cp->rptr = RREG32(CP_RB_RPTR);
evergreen_cp_start(rdev);
rdev->cp.ready = true;
r = radeon_ring_test(rdev);
cp->ready = true;
r = radeon_ring_test(rdev, cp);
if (r) {
rdev->cp.ready = false;
cp->ready = false;
return r;
}
return 0;
@ -2353,7 +2357,7 @@ int evergreen_mc_init(struct radeon_device *rdev)
return 0;
}
bool evergreen_gpu_is_lockup(struct radeon_device *rdev)
bool evergreen_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp)
{
u32 srbm_status;
u32 grbm_status;
@ -2366,19 +2370,19 @@ bool evergreen_gpu_is_lockup(struct radeon_device *rdev)
grbm_status_se0 = RREG32(GRBM_STATUS_SE0);
grbm_status_se1 = RREG32(GRBM_STATUS_SE1);
if (!(grbm_status & GUI_ACTIVE)) {
r100_gpu_lockup_update(lockup, &rdev->cp);
r100_gpu_lockup_update(lockup, cp);
return false;
}
/* force CP activities */
r = radeon_ring_lock(rdev, 2);
r = radeon_ring_lock(rdev, cp, 2);
if (!r) {
/* PACKET2 NOP */
radeon_ring_write(rdev, 0x80000000);
radeon_ring_write(rdev, 0x80000000);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, 0x80000000);
radeon_ring_write(cp, 0x80000000);
radeon_ring_unlock_commit(rdev, cp);
}
rdev->cp.rptr = RREG32(CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, lockup, &rdev->cp);
cp->rptr = RREG32(CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, lockup, cp);
}
static int evergreen_gpu_soft_reset(struct radeon_device *rdev)
@ -3052,6 +3056,7 @@ restart_ih:
static int evergreen_startup(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
/* enable pcie gen2 link */
@ -3115,7 +3120,7 @@ static int evergreen_startup(struct radeon_device *rdev)
}
evergreen_irq_set(rdev);
r = radeon_ring_init(rdev, rdev->cp.ring_size);
r = radeon_ring_init(rdev, cp, cp->ring_size);
if (r)
return r;
r = evergreen_cp_load_microcode(rdev);
@ -3150,7 +3155,7 @@ int evergreen_resume(struct radeon_device *rdev)
return r;
}
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("radeon: failed testing IB (%d).\n", r);
return r;
@ -3162,9 +3167,11 @@ int evergreen_resume(struct radeon_device *rdev)
int evergreen_suspend(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
/* FIXME: we should wait for ring to be empty */
r700_cp_stop(rdev);
rdev->cp.ready = false;
cp->ready = false;
evergreen_irq_suspend(rdev);
radeon_wb_disable(rdev);
evergreen_pcie_gart_disable(rdev);
@ -3244,7 +3251,7 @@ int evergreen_init(struct radeon_device *rdev)
return r;
rdev->cp.ring_obj = NULL;
r600_ring_init(rdev, 1024 * 1024);
r600_ring_init(rdev, &rdev->cp, 1024 * 1024);
rdev->ih.ring_obj = NULL;
r600_ih_ring_init(rdev, 64 * 1024);
@ -3270,7 +3277,7 @@ int evergreen_init(struct radeon_device *rdev)
DRM_ERROR("radeon: failed initializing IB pool (%d).\n", r);
rdev->accel_working = false;
}
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("radeon: failed testing IB (%d).\n", r);
rdev->accel_working = false;

242
drivers/gpu/drm/radeon/evergreen_blit_kms.c
View File

@ -49,6 +49,7 @@ static void
set_render_target(struct radeon_device *rdev, int format,
int w, int h, u64 gpu_addr)
{
struct radeon_cp *cp = &rdev->cp;
u32 cb_color_info;
int pitch, slice;
@ -62,23 +63,23 @@ set_render_target(struct radeon_device *rdev, int format,
pitch = (w / 8) - 1;
slice = ((w * h) / 64) - 1;
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 15));
radeon_ring_write(rdev, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(rdev, pitch);
radeon_ring_write(rdev, slice);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, cb_color_info);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, (w - 1) | ((h - 1) << 16));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 15));
radeon_ring_write(cp, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(cp, pitch);
radeon_ring_write(cp, slice);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, cb_color_info);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, (w - 1) | ((h - 1) << 16));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
}
/* emits 5dw */
@ -87,6 +88,7 @@ cp_set_surface_sync(struct radeon_device *rdev,
u32 sync_type, u32 size,
u64 mc_addr)
{
struct radeon_cp *cp = &rdev->cp;
u32 cp_coher_size;
if (size == 0xffffffff)
@ -99,39 +101,40 @@ cp_set_surface_sync(struct radeon_device *rdev,
* to the RB directly. For IBs, the CP programs this as part of the
* surface_sync packet.
*/
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, (0x85e8 - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(rdev, 0); /* CP_COHER_CNTL2 */
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, (0x85e8 - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(cp, 0); /* CP_COHER_CNTL2 */
}
radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(rdev, sync_type);
radeon_ring_write(rdev, cp_coher_size);
radeon_ring_write(rdev, mc_addr >> 8);
radeon_ring_write(rdev, 10); /* poll interval */
radeon_ring_write(cp, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(cp, sync_type);
radeon_ring_write(cp, cp_coher_size);
radeon_ring_write(cp, mc_addr >> 8);
radeon_ring_write(cp, 10); /* poll interval */
}
/* emits 11dw + 1 surface sync = 16dw */
static void
set_shaders(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
u64 gpu_addr;
/* VS */
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 3));
radeon_ring_write(rdev, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(rdev, 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 3));
radeon_ring_write(cp, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(cp, 2);
radeon_ring_write(cp, 0);
/* PS */
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset;
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 4));
radeon_ring_write(rdev, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(rdev, 1);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 2);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 4));
radeon_ring_write(cp, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(cp, 1);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 2);
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr);
@ -141,6 +144,7 @@ set_shaders(struct radeon_device *rdev)
static void
set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
{
struct radeon_cp *cp = &rdev->cp;
u32 sq_vtx_constant_word2, sq_vtx_constant_word3;
/* high addr, stride */
@ -155,16 +159,16 @@ set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
SQ_VTCX_SEL_Z(SQ_SEL_Z) |
SQ_VTCX_SEL_W(SQ_SEL_W);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 8));
radeon_ring_write(rdev, 0x580);
radeon_ring_write(rdev, gpu_addr & 0xffffffff);
radeon_ring_write(rdev, 48 - 1); /* size */
radeon_ring_write(rdev, sq_vtx_constant_word2);
radeon_ring_write(rdev, sq_vtx_constant_word3);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, S__SQ_CONSTANT_TYPE(SQ_TEX_VTX_VALID_BUFFER));
radeon_ring_write(cp, PACKET3(PACKET3_SET_RESOURCE, 8));
radeon_ring_write(cp, 0x580);
radeon_ring_write(cp, gpu_addr & 0xffffffff);
radeon_ring_write(cp, 48 - 1); /* size */
radeon_ring_write(cp, sq_vtx_constant_word2);
radeon_ring_write(cp, sq_vtx_constant_word3);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, S__SQ_CONSTANT_TYPE(SQ_TEX_VTX_VALID_BUFFER));
if ((rdev->family == CHIP_CEDAR) ||
(rdev->family == CHIP_PALM) ||
@ -185,6 +189,7 @@ set_tex_resource(struct radeon_device *rdev,
int format, int w, int h, int pitch,
u64 gpu_addr, u32 size)
{
struct radeon_cp *cp = &rdev->cp;
u32 sq_tex_resource_word0, sq_tex_resource_word1;
u32 sq_tex_resource_word4, sq_tex_resource_word7;
@ -208,16 +213,16 @@ set_tex_resource(struct radeon_device *rdev,
cp_set_surface_sync(rdev,
PACKET3_TC_ACTION_ENA, size, gpu_addr);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 8));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, sq_tex_resource_word0);
radeon_ring_write(rdev, sq_tex_resource_word1);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(rdev, sq_tex_resource_word4);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, sq_tex_resource_word7);
radeon_ring_write(cp, PACKET3(PACKET3_SET_RESOURCE, 8));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, sq_tex_resource_word0);
radeon_ring_write(cp, sq_tex_resource_word1);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(cp, sq_tex_resource_word4);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, sq_tex_resource_word7);
}
/* emits 12 */
@ -225,6 +230,7 @@ static void
set_scissors(struct radeon_device *rdev, int x1, int y1,
int x2, int y2)
{
struct radeon_cp *cp = &rdev->cp;
/* workaround some hw bugs */
if (x2 == 0)
x1 = 1;
@ -235,43 +241,44 @@ set_scissors(struct radeon_device *rdev, int x1, int y1,
x2 = 2;
}
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(rdev, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(rdev, (x1 << 0) | (y1 << 16));
radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(cp, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(cp, (x1 << 0) | (y1 << 16));
radeon_ring_write(cp, (x2 << 0) | (y2 << 16));
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(rdev, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(cp, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(cp, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(cp, (x2 << 0) | (y2 << 16));
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(rdev, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(cp, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
radeon_ring_write(cp, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(cp, (x2 << 0) | (y2 << 16));
}
/* emits 10 */
static void
draw_auto(struct radeon_device *rdev)
{
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(rdev, DI_PT_RECTLIST);
struct radeon_cp *cp = &rdev->cp;
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(cp, DI_PT_RECTLIST);
radeon_ring_write(rdev, PACKET3(PACKET3_INDEX_TYPE, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET3(PACKET3_INDEX_TYPE, 0));
radeon_ring_write(cp,
#ifdef __BIG_ENDIAN
(2 << 2) |
#endif
DI_INDEX_SIZE_16_BIT);
radeon_ring_write(rdev, PACKET3(PACKET3_NUM_INSTANCES, 0));
radeon_ring_write(rdev, 1);
radeon_ring_write(cp, PACKET3(PACKET3_NUM_INSTANCES, 0));
radeon_ring_write(cp, 1);
radeon_ring_write(rdev, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
radeon_ring_write(rdev, 3);
radeon_ring_write(rdev, DI_SRC_SEL_AUTO_INDEX);
radeon_ring_write(cp, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
radeon_ring_write(cp, 3);
radeon_ring_write(cp, DI_SRC_SEL_AUTO_INDEX);
}
@ -279,6 +286,7 @@ draw_auto(struct radeon_device *rdev)
static void
set_default_state(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2, sq_gpr_resource_mgmt_3;
u32 sq_thread_resource_mgmt, sq_thread_resource_mgmt_2;
u32 sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2, sq_stack_resource_mgmt_3;
@ -292,8 +300,8 @@ set_default_state(struct radeon_device *rdev)
int dwords;
/* set clear context state */
radeon_ring_write(rdev, PACKET3(PACKET3_CLEAR_STATE, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_CLEAR_STATE, 0));
radeon_ring_write(cp, 0);
if (rdev->family < CHIP_CAYMAN) {
switch (rdev->family) {
@ -550,60 +558,60 @@ set_default_state(struct radeon_device *rdev)
NUM_LS_STACK_ENTRIES(num_ls_stack_entries));
/* disable dyn gprs */
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, (SQ_DYN_GPR_CNTL_PS_FLUSH_REQ - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, (SQ_DYN_GPR_CNTL_PS_FLUSH_REQ - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(cp, 0);
/* setup LDS */
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, (SQ_LDS_RESOURCE_MGMT - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(rdev, 0x10001000);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, (SQ_LDS_RESOURCE_MGMT - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(cp, 0x10001000);
/* SQ config */
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 11));
radeon_ring_write(rdev, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(rdev, sq_config);
radeon_ring_write(rdev, sq_gpr_resource_mgmt_1);
radeon_ring_write(rdev, sq_gpr_resource_mgmt_2);
radeon_ring_write(rdev, sq_gpr_resource_mgmt_3);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, sq_thread_resource_mgmt);
radeon_ring_write(rdev, sq_thread_resource_mgmt_2);
radeon_ring_write(rdev, sq_stack_resource_mgmt_1);
radeon_ring_write(rdev, sq_stack_resource_mgmt_2);
radeon_ring_write(rdev, sq_stack_resource_mgmt_3);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 11));
radeon_ring_write(cp, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_START) >> 2);
radeon_ring_write(cp, sq_config);
radeon_ring_write(cp, sq_gpr_resource_mgmt_1);
radeon_ring_write(cp, sq_gpr_resource_mgmt_2);
radeon_ring_write(cp, sq_gpr_resource_mgmt_3);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, sq_thread_resource_mgmt);
radeon_ring_write(cp, sq_thread_resource_mgmt_2);
radeon_ring_write(cp, sq_stack_resource_mgmt_1);
radeon_ring_write(cp, sq_stack_resource_mgmt_2);
radeon_ring_write(cp, sq_stack_resource_mgmt_3);
}
/* CONTEXT_CONTROL */
radeon_ring_write(rdev, 0xc0012800);
radeon_ring_write(rdev, 0x80000000);
radeon_ring_write(rdev, 0x80000000);
radeon_ring_write(cp, 0xc0012800);
radeon_ring_write(cp, 0x80000000);
radeon_ring_write(cp, 0x80000000);
/* SQ_VTX_BASE_VTX_LOC */
radeon_ring_write(rdev, 0xc0026f00);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(cp, 0xc0026f00);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000000);
/* SET_SAMPLER */
radeon_ring_write(rdev, 0xc0036e00);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000012);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(cp, 0xc0036e00);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000012);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000000);
/* set to DX10/11 mode */
radeon_ring_write(rdev, PACKET3(PACKET3_MODE_CONTROL, 0));
radeon_ring_write(rdev, 1);
radeon_ring_write(cp, PACKET3(PACKET3_MODE_CONTROL, 0));
radeon_ring_write(cp, 1);
/* emit an IB pointing at default state */
dwords = ALIGN(rdev->r600_blit.state_len, 0x10);
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset;
radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(rdev, gpu_addr & 0xFFFFFFFC);
radeon_ring_write(rdev, upper_32_bits(gpu_addr) & 0xFF);
radeon_ring_write(rdev, dwords);
radeon_ring_write(cp, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(cp, gpu_addr & 0xFFFFFFFC);
radeon_ring_write(cp, upper_32_bits(gpu_addr) & 0xFF);
radeon_ring_write(cp, dwords);
}

129
drivers/gpu/drm/radeon/ni.c
View File

@ -1049,63 +1049,64 @@ static int cayman_cp_load_microcode(struct radeon_device *rdev)
static int cayman_cp_start(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r, i;
r = radeon_ring_lock(rdev, 7);
r = radeon_ring_lock(rdev, cp, 7);
if (r) {
DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
return r;
}
radeon_ring_write(rdev, PACKET3(PACKET3_ME_INITIALIZE, 5));
radeon_ring_write(rdev, 0x1);
radeon_ring_write(rdev, 0x0);
radeon_ring_write(rdev, rdev->config.cayman.max_hw_contexts - 1);
radeon_ring_write(rdev, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, PACKET3(PACKET3_ME_INITIALIZE, 5));
radeon_ring_write(cp, 0x1);
radeon_ring_write(cp, 0x0);
radeon_ring_write(cp, rdev->config.cayman.max_hw_contexts - 1);
radeon_ring_write(cp, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_unlock_commit(rdev, cp);
cayman_cp_enable(rdev, true);
r = radeon_ring_lock(rdev, cayman_default_size + 19);
r = radeon_ring_lock(rdev, cp, cayman_default_size + 19);
if (r) {
DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
return r;
}
/* setup clear context state */
radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(rdev, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
radeon_ring_write(cp, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(cp, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
for (i = 0; i < cayman_default_size; i++)
radeon_ring_write(rdev, cayman_default_state[i]);
radeon_ring_write(cp, cayman_default_state[i]);
radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(rdev, PACKET3_PREAMBLE_END_CLEAR_STATE);
radeon_ring_write(cp, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
radeon_ring_write(cp, PACKET3_PREAMBLE_END_CLEAR_STATE);
/* set clear context state */
radeon_ring_write(rdev, PACKET3(PACKET3_CLEAR_STATE, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_CLEAR_STATE, 0));
radeon_ring_write(cp, 0);
/* SQ_VTX_BASE_VTX_LOC */
radeon_ring_write(rdev, 0xc0026f00);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(rdev, 0x00000000);
radeon_ring_write(cp, 0xc0026f00);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000000);
radeon_ring_write(cp, 0x00000000);
/* Clear consts */
radeon_ring_write(rdev, 0xc0036f00);
radeon_ring_write(rdev, 0x00000bc4);
radeon_ring_write(rdev, 0xffffffff);
radeon_ring_write(rdev, 0xffffffff);
radeon_ring_write(rdev, 0xffffffff);
radeon_ring_write(cp, 0xc0036f00);
radeon_ring_write(cp, 0x00000bc4);
radeon_ring_write(cp, 0xffffffff);
radeon_ring_write(cp, 0xffffffff);
radeon_ring_write(cp, 0xffffffff);
radeon_ring_write(rdev, 0xc0026900);
radeon_ring_write(rdev, 0x00000316);
radeon_ring_write(rdev, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
radeon_ring_write(rdev, 0x00000010); /* */
radeon_ring_write(cp, 0xc0026900);
radeon_ring_write(cp, 0x00000316);
radeon_ring_write(cp, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
radeon_ring_write(cp, 0x00000010); /* */
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, cp);
/* XXX init other rings */
@ -1115,11 +1116,12 @@ static int cayman_cp_start(struct radeon_device *rdev)
static void cayman_cp_fini(struct radeon_device *rdev)
{
cayman_cp_enable(rdev, false);
radeon_ring_fini(rdev);
radeon_ring_fini(rdev, &rdev->cp);
}
int cayman_cp_resume(struct radeon_device *rdev)
{
struct radeon_cp *cp;
u32 tmp;
u32 rb_bufsz;
int r;
@ -1145,7 +1147,8 @@ int cayman_cp_resume(struct radeon_device *rdev)
/* ring 0 - compute and gfx */
/* Set ring buffer size */
rb_bufsz = drm_order(rdev->cp.ring_size / 8);
cp = &rdev->cp;
rb_bufsz = drm_order(cp->ring_size / 8);
tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
#ifdef __BIG_ENDIAN
tmp |= BUF_SWAP_32BIT;
@ -1154,8 +1157,8 @@ int cayman_cp_resume(struct radeon_device *rdev)
/* Initialize the ring buffer's read and write pointers */
WREG32(CP_RB0_CNTL, tmp | RB_RPTR_WR_ENA);
rdev->cp.wptr = 0;
WREG32(CP_RB0_WPTR, rdev->cp.wptr);
cp->wptr = 0;
WREG32(CP_RB0_WPTR, cp->wptr);
/* set the wb address wether it's enabled or not */
WREG32(CP_RB0_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC);
@ -1172,13 +1175,14 @@ int cayman_cp_resume(struct radeon_device *rdev)
mdelay(1);
WREG32(CP_RB0_CNTL, tmp);
WREG32(CP_RB0_BASE, rdev->cp.gpu_addr >> 8);
WREG32(CP_RB0_BASE, cp->gpu_addr >> 8);
rdev->cp.rptr = RREG32(CP_RB0_RPTR);
cp->rptr = RREG32(CP_RB0_RPTR);
/* ring1 - compute only */
/* Set ring buffer size */
rb_bufsz = drm_order(rdev->cp1.ring_size / 8);
cp = &rdev->cp1;
rb_bufsz = drm_order(cp->ring_size / 8);
tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
#ifdef __BIG_ENDIAN
tmp |= BUF_SWAP_32BIT;
@ -1187,8 +1191,8 @@ int cayman_cp_resume(struct radeon_device *rdev)
/* Initialize the ring buffer's read and write pointers */
WREG32(CP_RB1_CNTL, tmp | RB_RPTR_WR_ENA);
rdev->cp1.wptr = 0;
WREG32(CP_RB1_WPTR, rdev->cp1.wptr);
cp->wptr = 0;
WREG32(CP_RB1_WPTR, cp->wptr);
/* set the wb address wether it's enabled or not */
WREG32(CP_RB1_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP1_RPTR_OFFSET) & 0xFFFFFFFC);
@ -1197,13 +1201,14 @@ int cayman_cp_resume(struct radeon_device *rdev)
mdelay(1);
WREG32(CP_RB1_CNTL, tmp);
WREG32(CP_RB1_BASE, rdev->cp1.gpu_addr >> 8);
WREG32(CP_RB1_BASE, cp->gpu_addr >> 8);
rdev->cp1.rptr = RREG32(CP_RB1_RPTR);
cp->rptr = RREG32(CP_RB1_RPTR);
/* ring2 - compute only */
/* Set ring buffer size */
rb_bufsz = drm_order(rdev->cp2.ring_size / 8);
cp = &rdev->cp2;
rb_bufsz = drm_order(cp->ring_size / 8);
tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
#ifdef __BIG_ENDIAN
tmp |= BUF_SWAP_32BIT;
@ -1212,8 +1217,8 @@ int cayman_cp_resume(struct radeon_device *rdev)
/* Initialize the ring buffer's read and write pointers */
WREG32(CP_RB2_CNTL, tmp | RB_RPTR_WR_ENA);
rdev->cp2.wptr = 0;
WREG32(CP_RB2_WPTR, rdev->cp2.wptr);
cp->wptr = 0;
WREG32(CP_RB2_WPTR, cp->wptr);
/* set the wb address wether it's enabled or not */
WREG32(CP_RB2_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP2_RPTR_OFFSET) & 0xFFFFFFFC);
@ -1222,9 +1227,9 @@ int cayman_cp_resume(struct radeon_device *rdev)
mdelay(1);
WREG32(CP_RB2_CNTL, tmp);
WREG32(CP_RB2_BASE, rdev->cp2.gpu_addr >> 8);
WREG32(CP_RB2_BASE, cp->gpu_addr >> 8);
rdev->cp2.rptr = RREG32(CP_RB2_RPTR);
cp->rptr = RREG32(CP_RB2_RPTR);
/* start the rings */
cayman_cp_start(rdev);
@ -1232,7 +1237,7 @@ int cayman_cp_resume(struct radeon_device *rdev)
rdev->cp1.ready = true;
rdev->cp2.ready = true;
/* this only test cp0 */
r = radeon_ring_test(rdev);
r = radeon_ring_test(rdev, &rdev->cp);
if (r) {
rdev->cp.ready = false;
rdev->cp1.ready = false;
@ -1243,7 +1248,7 @@ int cayman_cp_resume(struct radeon_device *rdev)
return 0;
}
bool cayman_gpu_is_lockup(struct radeon_device *rdev)
bool cayman_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp)
{
u32 srbm_status;
u32 grbm_status;
@ -1256,20 +1261,20 @@ bool cayman_gpu_is_lockup(struct radeon_device *rdev)
grbm_status_se0 = RREG32(GRBM_STATUS_SE0);
grbm_status_se1 = RREG32(GRBM_STATUS_SE1);
if (!(grbm_status & GUI_ACTIVE)) {
r100_gpu_lockup_update(lockup, &rdev->cp);
r100_gpu_lockup_update(lockup, cp);
return false;
}
/* force CP activities */
r = radeon_ring_lock(rdev, 2);
r = radeon_ring_lock(rdev, cp, 2);
if (!r) {
/* PACKET2 NOP */
radeon_ring_write(rdev, 0x80000000);
radeon_ring_write(rdev, 0x80000000);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, 0x80000000);
radeon_ring_write(cp, 0x80000000);
radeon_ring_unlock_commit(rdev, cp);
}
/* XXX deal with CP0,1,2 */
rdev->cp.rptr = RREG32(CP_RB0_RPTR);
return r100_gpu_cp_is_lockup(rdev, lockup, &rdev->cp);
cp->rptr = RREG32(CP_RB0_RPTR);
return r100_gpu_cp_is_lockup(rdev, lockup, cp);
}
static int cayman_gpu_soft_reset(struct radeon_device *rdev)
@ -1338,6 +1343,7 @@ int cayman_asic_reset(struct radeon_device *rdev)
static int cayman_startup(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
/* enable pcie gen2 link */
@ -1387,7 +1393,7 @@ static int cayman_startup(struct radeon_device *rdev)
}
evergreen_irq_set(rdev);
r = radeon_ring_init(rdev, rdev->cp.ring_size);
r = radeon_ring_init(rdev, cp, cp->ring_size);
if (r)
return r;
r = cayman_cp_load_microcode(rdev);
@ -1417,7 +1423,7 @@ int cayman_resume(struct radeon_device *rdev)
return r;
}
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("radeon: failled testing IB (%d).\n", r);
return r;
@ -1448,6 +1454,7 @@ int cayman_suspend(struct radeon_device *rdev)
*/
int cayman_init(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
/* This don't do much */
@ -1500,8 +1507,8 @@ int cayman_init(struct radeon_device *rdev)
if (r)
return r;
rdev->cp.ring_obj = NULL;
r600_ring_init(rdev, 1024 * 1024);
cp->ring_obj = NULL;
r600_ring_init(rdev, cp, 1024 * 1024);
rdev->ih.ring_obj = NULL;
r600_ih_ring_init(rdev, 64 * 1024);
@ -1527,7 +1534,7 @@ int cayman_init(struct radeon_device *rdev)
DRM_ERROR("radeon: failed initializing IB pool (%d).\n", r);
rdev->accel_working = false;
}
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("radeon: failed testing IB (%d).\n", r);
rdev->accel_working = false;

147
drivers/gpu/drm/radeon/r100.c
View File

@ -811,30 +811,33 @@ u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc)
void r100_fence_ring_emit(struct radeon_device *rdev,
struct radeon_fence *fence)
{
struct radeon_cp *cp = &rdev->cp;
/* We have to make sure that caches are flushed before
* CPU might read something from VRAM. */
radeon_ring_write(rdev, PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, RADEON_RB3D_DC_FLUSH_ALL);
radeon_ring_write(rdev, PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, RADEON_RB3D_ZC_FLUSH_ALL);
radeon_ring_write(cp, PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, RADEON_RB3D_DC_FLUSH_ALL);
radeon_ring_write(cp, PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(cp, RADEON_RB3D_ZC_FLUSH_ALL);
/* Wait until IDLE & CLEAN */
radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(rdev, RADEON_WAIT_2D_IDLECLEAN | RADEON_WAIT_3D_IDLECLEAN);
radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(rdev, rdev->config.r100.hdp_cntl |
radeon_ring_write(cp, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(cp, RADEON_WAIT_2D_IDLECLEAN | RADEON_WAIT_3D_IDLECLEAN);
radeon_ring_write(cp, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(cp, rdev->config.r100.hdp_cntl |
RADEON_HDP_READ_BUFFER_INVALIDATE);
radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(rdev, rdev->config.r100.hdp_cntl);
radeon_ring_write(cp, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(cp, rdev->config.r100.hdp_cntl);
/* Emit fence sequence & fire IRQ */
radeon_ring_write(rdev, PACKET0(rdev->fence_drv[fence->ring].scratch_reg, 0));
radeon_ring_write(rdev, fence->seq);
radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0));
radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
radeon_ring_write(cp, PACKET0(rdev->fence_drv[fence->ring].scratch_reg, 0));
radeon_ring_write(cp, fence->seq);
radeon_ring_write(cp, PACKET0(RADEON_GEN_INT_STATUS, 0));
radeon_ring_write(cp, RADEON_SW_INT_FIRE);
}
void r100_semaphore_ring_emit(struct radeon_device *rdev,
struct radeon_cp *cp,
struct radeon_semaphore *semaphore,
unsigned ring, bool emit_wait)
bool emit_wait)
{
/* Unused on older asics, since we don't have semaphores or multiple rings */
BUG();
@ -846,6 +849,7 @@ int r100_copy_blit(struct radeon_device *rdev,
unsigned num_gpu_pages,
struct radeon_fence *fence)
{
struct radeon_cp *cp = &rdev->cp;
uint32_t cur_pages;
uint32_t stride_bytes = RADEON_GPU_PAGE_SIZE;
uint32_t pitch;
@ -863,7 +867,7 @@ int r100_copy_blit(struct radeon_device *rdev,
/* Ask for enough room for blit + flush + fence */
ndw = 64 + (10 * num_loops);
r = radeon_ring_lock(rdev, ndw);
r = radeon_ring_lock(rdev, cp, ndw);
if (r) {
DRM_ERROR("radeon: moving bo (%d) asking for %u dw.\n", r, ndw);
return -EINVAL;
@ -877,8 +881,8 @@ int r100_copy_blit(struct radeon_device *rdev,
/* pages are in Y direction - height
page width in X direction - width */
radeon_ring_write(rdev, PACKET3(PACKET3_BITBLT_MULTI, 8));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET3(PACKET3_BITBLT_MULTI, 8));
radeon_ring_write(cp,
RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
RADEON_GMC_DST_PITCH_OFFSET_CNTL |
RADEON_GMC_SRC_CLIPPING |
@ -890,26 +894,26 @@ int r100_copy_blit(struct radeon_device *rdev,
RADEON_DP_SRC_SOURCE_MEMORY |
RADEON_GMC_CLR_CMP_CNTL_DIS |
RADEON_GMC_WR_MSK_DIS);
radeon_ring_write(rdev, (pitch << 22) | (src_offset >> 10));
radeon_ring_write(rdev, (pitch << 22) | (dst_offset >> 10));
radeon_ring_write(rdev, (0x1fff) | (0x1fff << 16));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, (0x1fff) | (0x1fff << 16));
radeon_ring_write(rdev, num_gpu_pages);
radeon_ring_write(rdev, num_gpu_pages);
radeon_ring_write(rdev, cur_pages | (stride_pixels << 16));
radeon_ring_write(cp, (pitch << 22) | (src_offset >> 10));
radeon_ring_write(cp, (pitch << 22) | (dst_offset >> 10));
radeon_ring_write(cp, (0x1fff) | (0x1fff << 16));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, (0x1fff) | (0x1fff << 16));
radeon_ring_write(cp, num_gpu_pages);
radeon_ring_write(cp, num_gpu_pages);
radeon_ring_write(cp, cur_pages | (stride_pixels << 16));
}
radeon_ring_write(rdev, PACKET0(RADEON_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, RADEON_RB2D_DC_FLUSH_ALL);
radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(RADEON_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, RADEON_RB2D_DC_FLUSH_ALL);
radeon_ring_write(cp, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(cp,
RADEON_WAIT_2D_IDLECLEAN |
RADEON_WAIT_HOST_IDLECLEAN |
RADEON_WAIT_DMA_GUI_IDLE);
if (fence) {
r = radeon_fence_emit(rdev, fence);
}
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, cp);
return r;
}
@ -930,19 +934,20 @@ static int r100_cp_wait_for_idle(struct radeon_device *rdev)
void r100_ring_start(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
r = radeon_ring_lock(rdev, 2);
r = radeon_ring_lock(rdev, cp, 2);
if (r) {
return;
}
radeon_ring_write(rdev, PACKET0(RADEON_ISYNC_CNTL, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(RADEON_ISYNC_CNTL, 0));
radeon_ring_write(cp,
RADEON_ISYNC_ANY2D_IDLE3D |
RADEON_ISYNC_ANY3D_IDLE2D |
RADEON_ISYNC_WAIT_IDLEGUI |
RADEON_ISYNC_CPSCRATCH_IDLEGUI);
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, cp);
}
@ -1043,6 +1048,7 @@ static void r100_cp_load_microcode(struct radeon_device *rdev)
int r100_cp_init(struct radeon_device *rdev, unsigned ring_size)
{
struct radeon_cp *cp = &rdev->cp;
unsigned rb_bufsz;
unsigned rb_blksz;
unsigned max_fetch;
@ -1068,7 +1074,7 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size)
rb_bufsz = drm_order(ring_size / 8);
ring_size = (1 << (rb_bufsz + 1)) * 4;
r100_cp_load_microcode(rdev);
r = radeon_ring_init(rdev, ring_size);
r = radeon_ring_init(rdev, cp, ring_size);
if (r) {
return r;
}
@ -1077,7 +1083,7 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size)
rb_blksz = 9;
/* cp will read 128bytes at a time (4 dwords) */
max_fetch = 1;
rdev->cp.align_mask = 16 - 1;
cp->align_mask = 16 - 1;
/* Write to CP_RB_WPTR will be delayed for pre_write_timer clocks */
pre_write_timer = 64;
/* Force CP_RB_WPTR write if written more than one time before the
@ -1107,13 +1113,13 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size)
WREG32(RADEON_CP_RB_CNTL, tmp | RADEON_RB_NO_UPDATE);
/* Set ring address */
DRM_INFO("radeon: ring at 0x%016lX\n", (unsigned long)rdev->cp.gpu_addr);
WREG32(RADEON_CP_RB_BASE, rdev->cp.gpu_addr);
DRM_INFO("radeon: ring at 0x%016lX\n", (unsigned long)cp->gpu_addr);
WREG32(RADEON_CP_RB_BASE, cp->gpu_addr);
/* Force read & write ptr to 0 */
WREG32(RADEON_CP_RB_CNTL, tmp | RADEON_RB_RPTR_WR_ENA | RADEON_RB_NO_UPDATE);
WREG32(RADEON_CP_RB_RPTR_WR, 0);
rdev->cp.wptr = 0;
WREG32(RADEON_CP_RB_WPTR, rdev->cp.wptr);
cp->wptr = 0;
WREG32(RADEON_CP_RB_WPTR, cp->wptr);
/* set the wb address whether it's enabled or not */
WREG32(R_00070C_CP_RB_RPTR_ADDR,
@ -1129,7 +1135,7 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size)
WREG32(RADEON_CP_RB_CNTL, tmp);
udelay(10);
rdev->cp.rptr = RREG32(RADEON_CP_RB_RPTR);
cp->rptr = RREG32(RADEON_CP_RB_RPTR);
/* Set cp mode to bus mastering & enable cp*/
WREG32(RADEON_CP_CSQ_MODE,
REG_SET(RADEON_INDIRECT2_START, indirect2_start) |
@ -1138,12 +1144,12 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size)
WREG32(RADEON_CP_CSQ_MODE, 0x00004D4D);
WREG32(RADEON_CP_CSQ_CNTL, RADEON_CSQ_PRIBM_INDBM);
radeon_ring_start(rdev);
r = radeon_ring_test(rdev);
r = radeon_ring_test(rdev, cp);
if (r) {
DRM_ERROR("radeon: cp isn't working (%d).\n", r);
return r;
}
rdev->cp.ready = true;
cp->ready = true;
radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size);
return 0;
}
@ -1155,7 +1161,7 @@ void r100_cp_fini(struct radeon_device *rdev)
}
/* Disable ring */
r100_cp_disable(rdev);
radeon_ring_fini(rdev);
radeon_ring_fini(rdev, &rdev->cp);
DRM_INFO("radeon: cp finalized\n");
}
@ -1173,9 +1179,9 @@ void r100_cp_disable(struct radeon_device *rdev)
}
}
void r100_cp_commit(struct radeon_device *rdev)
void r100_cp_commit(struct radeon_device *rdev, struct radeon_cp *cp)
{
WREG32(RADEON_CP_RB_WPTR, rdev->cp.wptr);
WREG32(RADEON_CP_RB_WPTR, cp->wptr);
(void)RREG32(RADEON_CP_RB_WPTR);
}
@ -2160,26 +2166,26 @@ bool r100_gpu_cp_is_lockup(struct radeon_device *rdev, struct r100_gpu_lockup *l
return false;
}
bool r100_gpu_is_lockup(struct radeon_device *rdev)
bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp)
{
u32 rbbm_status;
int r;
rbbm_status = RREG32(R_000E40_RBBM_STATUS);
if (!G_000E40_GUI_ACTIVE(rbbm_status)) {
r100_gpu_lockup_update(&rdev->config.r100.lockup, &rdev->cp);
r100_gpu_lockup_update(&rdev->config.r100.lockup, cp);
return false;
}
/* force CP activities */
r = radeon_ring_lock(rdev, 2);
r = radeon_ring_lock(rdev, cp, 2);
if (!r) {
/* PACKET2 NOP */
radeon_ring_write(rdev, 0x80000000);
radeon_ring_write(rdev, 0x80000000);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, 0x80000000);
radeon_ring_write(cp, 0x80000000);
radeon_ring_unlock_commit(rdev, cp);
}
rdev->cp.rptr = RREG32(RADEON_CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, &rdev->config.r100.lockup, &rdev->cp);
cp->rptr = RREG32(RADEON_CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, &rdev->config.r100.lockup, cp);
}
void r100_bm_disable(struct radeon_device *rdev)
@ -2587,21 +2593,22 @@ static int r100_debugfs_cp_ring_info(struct seq_file *m, void *data)
struct drm_info_node *node = (struct drm_info_node *) m->private;
struct drm_device *dev = node->minor->dev;
struct radeon_device *rdev = dev->dev_private;
struct radeon_cp *cp = &rdev->cp;
uint32_t rdp, wdp;
unsigned count, i, j;
radeon_ring_free_size(rdev);
radeon_ring_free_size(rdev, cp);
rdp = RREG32(RADEON_CP_RB_RPTR);
wdp = RREG32(RADEON_CP_RB_WPTR);
count = (rdp + rdev->cp.ring_size - wdp) & rdev->cp.ptr_mask;
count = (rdp + cp->ring_size - wdp) & cp->ptr_mask;
seq_printf(m, "CP_STAT 0x%08x\n", RREG32(RADEON_CP_STAT));
seq_printf(m, "CP_RB_WPTR 0x%08x\n", wdp);
seq_printf(m, "CP_RB_RPTR 0x%08x\n", rdp);
seq_printf(m, "%u free dwords in ring\n", rdev->cp.ring_free_dw);
seq_printf(m, "%u free dwords in ring\n", cp->ring_free_dw);
seq_printf(m, "%u dwords in ring\n", count);
for (j = 0; j <= count; j++) {
i = (rdp + j) & rdev->cp.ptr_mask;
seq_printf(m, "r[%04d]=0x%08x\n", i, rdev->cp.ring[i]);
i = (rdp + j) & cp->ptr_mask;
seq_printf(m, "r[%04d]=0x%08x\n", i, cp->ring[i]);
}
return 0;
}
@ -3643,7 +3650,7 @@ void r100_cs_track_clear(struct radeon_device *rdev, struct r100_cs_track *track
}
}
int r100_ring_test(struct radeon_device *rdev)
int r100_ring_test(struct radeon_device *rdev, struct radeon_cp *cp)
{
uint32_t scratch;
uint32_t tmp = 0;
@ -3656,15 +3663,15 @@ int r100_ring_test(struct radeon_device *rdev)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = radeon_ring_lock(rdev, 2);
r = radeon_ring_lock(rdev, cp, 2);
if (r) {
DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
radeon_scratch_free(rdev, scratch);
return r;
}
radeon_ring_write(rdev, PACKET0(scratch, 0));
radeon_ring_write(rdev, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, PACKET0(scratch, 0));
radeon_ring_write(cp, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev, cp);
for (i = 0; i < rdev->usec_timeout; i++) {
tmp = RREG32(scratch);
if (tmp == 0xDEADBEEF) {
@ -3685,9 +3692,11 @@ int r100_ring_test(struct radeon_device *rdev)
void r100_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
{
radeon_ring_write(rdev, PACKET0(RADEON_CP_IB_BASE, 1));
radeon_ring_write(rdev, ib->gpu_addr);
radeon_ring_write(rdev, ib->length_dw);
struct radeon_cp *cp = &rdev->cp;
radeon_ring_write(cp, PACKET0(RADEON_CP_IB_BASE, 1));
radeon_ring_write(cp, ib->gpu_addr);
radeon_ring_write(cp, ib->length_dw);
}
int r100_ib_test(struct radeon_device *rdev)
@ -3704,7 +3713,7 @@ int r100_ib_test(struct radeon_device *rdev)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = radeon_ib_get(rdev, &ib);
r = radeon_ib_get(rdev, RADEON_RING_TYPE_GFX_INDEX, &ib);
if (r) {
return r;
}

21
drivers/gpu/drm/radeon/r200.c
View File

@ -87,6 +87,7 @@ int r200_copy_dma(struct radeon_device *rdev,
unsigned num_gpu_pages,
struct radeon_fence *fence)
{
struct radeon_cp *cp = &rdev->cp;
uint32_t size;
uint32_t cur_size;
int i, num_loops;
@ -95,33 +96,33 @@ int r200_copy_dma(struct radeon_device *rdev,
/* radeon pitch is /64 */
size = num_gpu_pages << RADEON_GPU_PAGE_SHIFT;
num_loops = DIV_ROUND_UP(size, 0x1FFFFF);
r = radeon_ring_lock(rdev, num_loops * 4 + 64);
r = radeon_ring_lock(rdev, cp, num_loops * 4 + 64);
if (r) {
DRM_ERROR("radeon: moving bo (%d).\n", r);
return r;
}
/* Must wait for 2D idle & clean before DMA or hangs might happen */
radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(rdev, (1 << 16));
radeon_ring_write(cp, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(cp, (1 << 16));
for (i = 0; i < num_loops; i++) {
cur_size = size;
if (cur_size > 0x1FFFFF) {
cur_size = 0x1FFFFF;
}
size -= cur_size;
radeon_ring_write(rdev, PACKET0(0x720, 2));
radeon_ring_write(rdev, src_offset);
radeon_ring_write(rdev, dst_offset);
radeon_ring_write(rdev, cur_size | (1 << 31) | (1 << 30));
radeon_ring_write(cp, PACKET0(0x720, 2));
radeon_ring_write(cp, src_offset);
radeon_ring_write(cp, dst_offset);
radeon_ring_write(cp, cur_size | (1 << 31) | (1 << 30));
src_offset += cur_size;
dst_offset += cur_size;
}
radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(rdev, RADEON_WAIT_DMA_GUI_IDLE);
radeon_ring_write(cp, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(cp, RADEON_WAIT_DMA_GUI_IDLE);
if (fence) {
r = radeon_fence_emit(rdev, fence);
}
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, cp);
return r;
}

127
drivers/gpu/drm/radeon/r300.c
View File

@ -175,37 +175,40 @@ void rv370_pcie_gart_fini(struct radeon_device *rdev)
void r300_fence_ring_emit(struct radeon_device *rdev,
struct radeon_fence *fence)
{
struct radeon_cp *cp = &rdev->cp;
/* Who ever call radeon_fence_emit should call ring_lock and ask
* for enough space (today caller are ib schedule and buffer move) */
/* Write SC register so SC & US assert idle */
radeon_ring_write(rdev, PACKET0(R300_RE_SCISSORS_TL, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(R300_RE_SCISSORS_BR, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET0(R300_RE_SCISSORS_TL, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(R300_RE_SCISSORS_BR, 0));
radeon_ring_write(cp, 0);
/* Flush 3D cache */
radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, R300_RB3D_DC_FLUSH);
radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, R300_ZC_FLUSH);
radeon_ring_write(cp, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, R300_RB3D_DC_FLUSH);
radeon_ring_write(cp, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(cp, R300_ZC_FLUSH);
/* Wait until IDLE & CLEAN */
radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(rdev, (RADEON_WAIT_3D_IDLECLEAN |
radeon_ring_write(cp, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(cp, (RADEON_WAIT_3D_IDLECLEAN |
RADEON_WAIT_2D_IDLECLEAN |
RADEON_WAIT_DMA_GUI_IDLE));
radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(rdev, rdev->config.r300.hdp_cntl |
radeon_ring_write(cp, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(cp, rdev->config.r300.hdp_cntl |
RADEON_HDP_READ_BUFFER_INVALIDATE);
radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(rdev, rdev->config.r300.hdp_cntl);
radeon_ring_write(cp, PACKET0(RADEON_HOST_PATH_CNTL, 0));
radeon_ring_write(cp, rdev->config.r300.hdp_cntl);
/* Emit fence sequence & fire IRQ */
radeon_ring_write(rdev, PACKET0(rdev->fence_drv[fence->ring].scratch_reg, 0));
radeon_ring_write(rdev, fence->seq);
radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0));
radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
radeon_ring_write(cp, PACKET0(rdev->fence_drv[fence->ring].scratch_reg, 0));
radeon_ring_write(cp, fence->seq);
radeon_ring_write(cp, PACKET0(RADEON_GEN_INT_STATUS, 0));
radeon_ring_write(cp, RADEON_SW_INT_FIRE);
}
void r300_ring_start(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
unsigned gb_tile_config;
int r;
@ -227,44 +230,44 @@ void r300_ring_start(struct radeon_device *rdev)
break;
}
r = radeon_ring_lock(rdev, 64);
r = radeon_ring_lock(rdev, cp, 64);
if (r) {
return;
}
radeon_ring_write(rdev, PACKET0(RADEON_ISYNC_CNTL, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(RADEON_ISYNC_CNTL, 0));
radeon_ring_write(cp,
RADEON_ISYNC_ANY2D_IDLE3D |
RADEON_ISYNC_ANY3D_IDLE2D |
RADEON_ISYNC_WAIT_IDLEGUI |
RADEON_ISYNC_CPSCRATCH_IDLEGUI);
radeon_ring_write(rdev, PACKET0(R300_GB_TILE_CONFIG, 0));
radeon_ring_write(rdev, gb_tile_config);
radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(R300_GB_TILE_CONFIG, 0));
radeon_ring_write(cp, gb_tile_config);
radeon_ring_write(cp, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(cp,
RADEON_WAIT_2D_IDLECLEAN |
RADEON_WAIT_3D_IDLECLEAN);
radeon_ring_write(rdev, PACKET0(R300_DST_PIPE_CONFIG, 0));
radeon_ring_write(rdev, R300_PIPE_AUTO_CONFIG);
radeon_ring_write(rdev, PACKET0(R300_GB_SELECT, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(R300_GB_ENABLE, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(R300_DST_PIPE_CONFIG, 0));
radeon_ring_write(cp, R300_PIPE_AUTO_CONFIG);
radeon_ring_write(cp, PACKET0(R300_GB_SELECT, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(R300_GB_ENABLE, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
radeon_ring_write(cp, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(cp, R300_ZC_FLUSH | R300_ZC_FREE);
radeon_ring_write(cp, PACKET0(RADEON_WAIT_UNTIL, 0));
radeon_ring_write(cp,
RADEON_WAIT_2D_IDLECLEAN |
RADEON_WAIT_3D_IDLECLEAN);
radeon_ring_write(rdev, PACKET0(R300_GB_AA_CONFIG, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS0, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(R300_GB_AA_CONFIG, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
radeon_ring_write(cp, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
radeon_ring_write(cp, R300_ZC_FLUSH | R300_ZC_FREE);
radeon_ring_write(cp, PACKET0(R300_GB_MSPOS0, 0));
radeon_ring_write(cp,
((6 << R300_MS_X0_SHIFT) |
(6 << R300_MS_Y0_SHIFT) |
(6 << R300_MS_X1_SHIFT) |
@ -273,8 +276,8 @@ void r300_ring_start(struct radeon_device *rdev)
(6 << R300_MS_Y2_SHIFT) |
(6 << R300_MSBD0_Y_SHIFT) |
(6 << R300_MSBD0_X_SHIFT)));
radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS1, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(R300_GB_MSPOS1, 0));
radeon_ring_write(cp,
((6 << R300_MS_X3_SHIFT) |
(6 << R300_MS_Y3_SHIFT) |
(6 << R300_MS_X4_SHIFT) |
@ -282,16 +285,16 @@ void r300_ring_start(struct radeon_device *rdev)
(6 << R300_MS_X5_SHIFT) |
(6 << R300_MS_Y5_SHIFT) |
(6 << R300_MSBD1_SHIFT)));
radeon_ring_write(rdev, PACKET0(R300_GA_ENHANCE, 0));
radeon_ring_write(rdev, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL);
radeon_ring_write(rdev, PACKET0(R300_GA_POLY_MODE, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(R300_GA_ENHANCE, 0));
radeon_ring_write(cp, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL);
radeon_ring_write(cp, PACKET0(R300_GA_POLY_MODE, 0));
radeon_ring_write(cp,
R300_FRONT_PTYPE_TRIANGE | R300_BACK_PTYPE_TRIANGE);
radeon_ring_write(rdev, PACKET0(R300_GA_ROUND_MODE, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(R300_GA_ROUND_MODE, 0));
radeon_ring_write(cp,
R300_GEOMETRY_ROUND_NEAREST |
R300_COLOR_ROUND_NEAREST);
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, cp);
}
void r300_errata(struct radeon_device *rdev)
@ -375,26 +378,26 @@ void r300_gpu_init(struct radeon_device *rdev)
rdev->num_gb_pipes, rdev->num_z_pipes);
}
bool r300_gpu_is_lockup(struct radeon_device *rdev)
bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp)
{
u32 rbbm_status;
int r;
rbbm_status = RREG32(R_000E40_RBBM_STATUS);
if (!G_000E40_GUI_ACTIVE(rbbm_status)) {
r100_gpu_lockup_update(&rdev->config.r300.lockup, &rdev->cp);
r100_gpu_lockup_update(&rdev->config.r300.lockup, cp);
return false;
}
/* force CP activities */
r = radeon_ring_lock(rdev, 2);
r = radeon_ring_lock(rdev, cp, 2);
if (!r) {
/* PACKET2 NOP */
radeon_ring_write(rdev, 0x80000000);
radeon_ring_write(rdev, 0x80000000);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, 0x80000000);
radeon_ring_write(cp, 0x80000000);
radeon_ring_unlock_commit(rdev, cp);
}
rdev->cp.rptr = RREG32(RADEON_CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, &rdev->config.r300.lockup, &rdev->cp);
cp->rptr = RREG32(RADEON_CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, &rdev->config.r300.lockup, cp);
}
int r300_asic_reset(struct radeon_device *rdev)

22
drivers/gpu/drm/radeon/r420.c
View File

@ -199,6 +199,8 @@ static void r420_clock_resume(struct radeon_device *rdev)
static void r420_cp_errata_init(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
/* RV410 and R420 can lock up if CP DMA to host memory happens
* while the 2D engine is busy.
*
@ -206,22 +208,24 @@ static void r420_cp_errata_init(struct radeon_device *rdev)
* of the CP init, apparently.
*/
radeon_scratch_get(rdev, &rdev->config.r300.resync_scratch);
radeon_ring_lock(rdev, 8);
radeon_ring_write(rdev, PACKET0(R300_CP_RESYNC_ADDR, 1));
radeon_ring_write(rdev, rdev->config.r300.resync_scratch);
radeon_ring_write(rdev, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev);
radeon_ring_lock(rdev, cp, 8);
radeon_ring_write(cp, PACKET0(R300_CP_RESYNC_ADDR, 1));
radeon_ring_write(cp, rdev->config.r300.resync_scratch);
radeon_ring_write(cp, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev, cp);
}
static void r420_cp_errata_fini(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
/* Catch the RESYNC we dispatched all the way back,
* at the very beginning of the CP init.
*/
radeon_ring_lock(rdev, 8);
radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, R300_RB3D_DC_FINISH);
radeon_ring_unlock_commit(rdev);
radeon_ring_lock(rdev, cp, 8);
radeon_ring_write(cp, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, R300_RB3D_DC_FINISH);
radeon_ring_unlock_commit(rdev, cp);
radeon_scratch_free(rdev, rdev->config.r300.resync_scratch);
}

201
drivers/gpu/drm/radeon/r600.c
View File

@ -1344,7 +1344,7 @@ int r600_gpu_soft_reset(struct radeon_device *rdev)
return 0;
}
bool r600_gpu_is_lockup(struct radeon_device *rdev)
bool r600_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp)
{
u32 srbm_status;
u32 grbm_status;
@ -1361,19 +1361,19 @@ bool r600_gpu_is_lockup(struct radeon_device *rdev)
grbm_status = RREG32(R_008010_GRBM_STATUS);
grbm_status2 = RREG32(R_008014_GRBM_STATUS2);
if (!G_008010_GUI_ACTIVE(grbm_status)) {
r100_gpu_lockup_update(lockup, &rdev->cp);
r100_gpu_lockup_update(lockup, cp);
return false;
}
/* force CP activities */
r = radeon_ring_lock(rdev, 2);
r = radeon_ring_lock(rdev, cp, 2);
if (!r) {
/* PACKET2 NOP */
radeon_ring_write(rdev, 0x80000000);
radeon_ring_write(rdev, 0x80000000);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, 0x80000000);
radeon_ring_write(cp, 0x80000000);
radeon_ring_unlock_commit(rdev, cp);
}
rdev->cp.rptr = RREG32(R600_CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, lockup, &rdev->cp);
cp->rptr = RREG32(R600_CP_RB_RPTR);
return r100_gpu_cp_is_lockup(rdev, lockup, cp);
}
int r600_asic_reset(struct radeon_device *rdev)
@ -2144,27 +2144,28 @@ static int r600_cp_load_microcode(struct radeon_device *rdev)
int r600_cp_start(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
uint32_t cp_me;
r = radeon_ring_lock(rdev, 7);
r = radeon_ring_lock(rdev, cp, 7);
if (r) {
DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
return r;
}
radeon_ring_write(rdev, PACKET3(PACKET3_ME_INITIALIZE, 5));
radeon_ring_write(rdev, 0x1);
radeon_ring_write(cp, PACKET3(PACKET3_ME_INITIALIZE, 5));
radeon_ring_write(cp, 0x1);
if (rdev->family >= CHIP_RV770) {
radeon_ring_write(rdev, 0x0);
radeon_ring_write(rdev, rdev->config.rv770.max_hw_contexts - 1);
radeon_ring_write(cp, 0x0);
radeon_ring_write(cp, rdev->config.rv770.max_hw_contexts - 1);
} else {
radeon_ring_write(rdev, 0x3);
radeon_ring_write(rdev, rdev->config.r600.max_hw_contexts - 1);
radeon_ring_write(cp, 0x3);
radeon_ring_write(cp, rdev->config.r600.max_hw_contexts - 1);
}
radeon_ring_write(rdev, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_unlock_commit(rdev, cp);
cp_me = 0xff;
WREG32(R_0086D8_CP_ME_CNTL, cp_me);
@ -2173,6 +2174,7 @@ int r600_cp_start(struct radeon_device *rdev)
int r600_cp_resume(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
u32 tmp;
u32 rb_bufsz;
int r;
@ -2184,7 +2186,7 @@ int r600_cp_resume(struct radeon_device *rdev)
WREG32(GRBM_SOFT_RESET, 0);
/* Set ring buffer size */
rb_bufsz = drm_order(rdev->cp.ring_size / 8);
rb_bufsz = drm_order(cp->ring_size / 8);
tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
#ifdef __BIG_ENDIAN
tmp |= BUF_SWAP_32BIT;
@ -2198,8 +2200,8 @@ int r600_cp_resume(struct radeon_device *rdev)
/* Initialize the ring buffer's read and write pointers */
WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA);
WREG32(CP_RB_RPTR_WR, 0);
rdev->cp.wptr = 0;
WREG32(CP_RB_WPTR, rdev->cp.wptr);
cp->wptr = 0;
WREG32(CP_RB_WPTR, cp->wptr);
/* set the wb address whether it's enabled or not */
WREG32(CP_RB_RPTR_ADDR,
@ -2217,42 +2219,42 @@ int r600_cp_resume(struct radeon_device *rdev)
mdelay(1);
WREG32(CP_RB_CNTL, tmp);
WREG32(CP_RB_BASE, rdev->cp.gpu_addr >> 8);
WREG32(CP_RB_BASE, cp->gpu_addr >> 8);
WREG32(CP_DEBUG, (1 << 27) | (1 << 28));
rdev->cp.rptr = RREG32(CP_RB_RPTR);
cp->rptr = RREG32(CP_RB_RPTR);
r600_cp_start(rdev);
rdev->cp.ready = true;
r = radeon_ring_test(rdev);
cp->ready = true;
r = radeon_ring_test(rdev, cp);
if (r) {
rdev->cp.ready = false;
cp->ready = false;
return r;
}
return 0;
}
void r600_cp_commit(struct radeon_device *rdev)
void r600_cp_commit(struct radeon_device *rdev, struct radeon_cp *cp)
{
WREG32(CP_RB_WPTR, rdev->cp.wptr);
WREG32(CP_RB_WPTR, cp->wptr);
(void)RREG32(CP_RB_WPTR);
}
void r600_ring_init(struct radeon_device *rdev, unsigned ring_size)
void r600_ring_init(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ring_size)
{
u32 rb_bufsz;
/* Align ring size */
rb_bufsz = drm_order(ring_size / 8);
ring_size = (1 << (rb_bufsz + 1)) * 4;
rdev->cp.ring_size = ring_size;
rdev->cp.align_mask = 16 - 1;
cp->ring_size = ring_size;
cp->align_mask = 16 - 1;
}
void r600_cp_fini(struct radeon_device *rdev)
{
r600_cp_stop(rdev);
radeon_ring_fini(rdev);
radeon_ring_fini(rdev, &rdev->cp);
}
@ -2271,7 +2273,7 @@ void r600_scratch_init(struct radeon_device *rdev)
}
}
int r600_ring_test(struct radeon_device *rdev)
int r600_ring_test(struct radeon_device *rdev, struct radeon_cp *cp)
{
uint32_t scratch;
uint32_t tmp = 0;
@ -2284,16 +2286,16 @@ int r600_ring_test(struct radeon_device *rdev)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = radeon_ring_lock(rdev, 3);
r = radeon_ring_lock(rdev, cp, 3);
if (r) {
DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
DRM_ERROR("radeon: cp failed to lock ring %p (%d).\n", cp, r);
radeon_scratch_free(rdev, scratch);
return r;
}
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, ((scratch - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
radeon_ring_write(rdev, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, ((scratch - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
radeon_ring_write(cp, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev, cp);
for (i = 0; i < rdev->usec_timeout; i++) {
tmp = RREG32(scratch);
if (tmp == 0xDEADBEEF)
@ -2301,10 +2303,10 @@ int r600_ring_test(struct radeon_device *rdev)
DRM_UDELAY(1);
}
if (i < rdev->usec_timeout) {
DRM_INFO("ring test succeeded in %d usecs\n", i);
DRM_INFO("ring test on %p succeeded in %d usecs\n", cp, i);
} else {
DRM_ERROR("radeon: ring test failed (scratch(0x%04X)=0x%08X)\n",
scratch, tmp);
DRM_ERROR("radeon: ring %p test failed (scratch(0x%04X)=0x%08X)\n",
cp, scratch, tmp);
r = -EINVAL;
}
radeon_scratch_free(rdev, scratch);
@ -2314,59 +2316,62 @@ int r600_ring_test(struct radeon_device *rdev)
void r600_fence_ring_emit(struct radeon_device *rdev,
struct radeon_fence *fence)
{
struct radeon_cp *cp = &rdev->cp;
if (rdev->wb.use_event) {
u64 addr = rdev->wb.gpu_addr + R600_WB_EVENT_OFFSET +
(u64)(rdev->fence_drv[fence->ring].scratch_reg - rdev->scratch.reg_base);
/* flush read cache over gart */
radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(rdev, PACKET3_TC_ACTION_ENA |
PACKET3_VC_ACTION_ENA |
PACKET3_SH_ACTION_ENA);
radeon_ring_write(rdev, 0xFFFFFFFF);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 10); /* poll interval */
radeon_ring_write(cp, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(cp, PACKET3_TC_ACTION_ENA |
PACKET3_VC_ACTION_ENA |
PACKET3_SH_ACTION_ENA);
radeon_ring_write(cp, 0xFFFFFFFF);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 10); /* poll interval */
/* EVENT_WRITE_EOP - flush caches, send int */
radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
radeon_ring_write(rdev, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT_TS) | EVENT_INDEX(5));
radeon_ring_write(rdev, addr & 0xffffffff);
radeon_ring_write(rdev, (upper_32_bits(addr) & 0xff) | DATA_SEL(1) | INT_SEL(2));
radeon_ring_write(rdev, fence->seq);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
radeon_ring_write(cp, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT_TS) | EVENT_INDEX(5));
radeon_ring_write(cp, addr & 0xffffffff);
radeon_ring_write(cp, (upper_32_bits(addr) & 0xff) | DATA_SEL(1) | INT_SEL(2));
radeon_ring_write(cp, fence->seq);
radeon_ring_write(cp, 0);
} else {
/* flush read cache over gart */
radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(rdev, PACKET3_TC_ACTION_ENA |
PACKET3_VC_ACTION_ENA |
PACKET3_SH_ACTION_ENA);
radeon_ring_write(rdev, 0xFFFFFFFF);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 10); /* poll interval */
radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0));
radeon_ring_write(rdev, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT) | EVENT_INDEX(0));
radeon_ring_write(cp, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(cp, PACKET3_TC_ACTION_ENA |
PACKET3_VC_ACTION_ENA |
PACKET3_SH_ACTION_ENA);
radeon_ring_write(cp, 0xFFFFFFFF);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 10); /* poll interval */
radeon_ring_write(cp, PACKET3(PACKET3_EVENT_WRITE, 0));
radeon_ring_write(cp, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT) | EVENT_INDEX(0));
/* wait for 3D idle clean */
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
radeon_ring_write(rdev, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
radeon_ring_write(cp, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
/* Emit fence sequence & fire IRQ */
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, ((rdev->fence_drv[fence->ring].scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
radeon_ring_write(rdev, fence->seq);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, ((rdev->fence_drv[fence->ring].scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
radeon_ring_write(cp, fence->seq);
/* CP_INTERRUPT packet 3 no longer exists, use packet 0 */
radeon_ring_write(rdev, PACKET0(CP_INT_STATUS, 0));
radeon_ring_write(rdev, RB_INT_STAT);
radeon_ring_write(cp, PACKET0(CP_INT_STATUS, 0));
radeon_ring_write(cp, RB_INT_STAT);
}
}
void r600_semaphore_ring_emit(struct radeon_device *rdev,
struct radeon_cp *cp,
struct radeon_semaphore *semaphore,
unsigned ring, bool emit_wait)
bool emit_wait)
{
uint64_t addr = semaphore->gpu_addr;
unsigned sel = emit_wait ? PACKET3_SEM_SEL_WAIT : PACKET3_SEM_SEL_SIGNAL;
radeon_ring_write(rdev, PACKET3(PACKET3_MEM_SEMAPHORE, 1));
radeon_ring_write(rdev, addr & 0xffffffff);
radeon_ring_write(rdev, (upper_32_bits(addr) & 0xff) | sel);
radeon_ring_write(cp, PACKET3(PACKET3_MEM_SEMAPHORE, 1));
radeon_ring_write(cp, addr & 0xffffffff);
radeon_ring_write(cp, (upper_32_bits(addr) & 0xff) | sel);
}
int r600_copy_blit(struct radeon_device *rdev,
@ -2421,6 +2426,7 @@ void r600_clear_surface_reg(struct radeon_device *rdev, int reg)
int r600_startup(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
/* enable pcie gen2 link */
@ -2468,7 +2474,7 @@ int r600_startup(struct radeon_device *rdev)
}
r600_irq_set(rdev);
r = radeon_ring_init(rdev, rdev->cp.ring_size);
r = radeon_ring_init(rdev, cp, cp->ring_size);
if (r)
return r;
r = r600_cp_load_microcode(rdev);
@ -2512,7 +2518,7 @@ int r600_resume(struct radeon_device *rdev)
return r;
}
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("radeon: failed testing IB (%d).\n", r);
return r;
@ -2608,7 +2614,7 @@ int r600_init(struct radeon_device *rdev)
return r;
rdev->cp.ring_obj = NULL;
r600_ring_init(rdev, 1024 * 1024);
r600_ring_init(rdev, &rdev->cp, 1024 * 1024);
rdev->ih.ring_obj = NULL;
r600_ih_ring_init(rdev, 64 * 1024);
@ -2634,7 +2640,7 @@ int r600_init(struct radeon_device *rdev)
dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
rdev->accel_working = false;
} else {
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
dev_err(rdev->dev, "IB test failed (%d).\n", r);
rdev->accel_working = false;
@ -2675,18 +2681,20 @@ void r600_fini(struct radeon_device *rdev)
*/
void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
{
struct radeon_cp *cp = &rdev->cp;
/* FIXME: implement */
radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(cp,
#ifdef __BIG_ENDIAN
(2 << 0) |
#endif
(ib->gpu_addr & 0xFFFFFFFC));
radeon_ring_write(rdev, upper_32_bits(ib->gpu_addr) & 0xFF);
radeon_ring_write(rdev, ib->length_dw);
radeon_ring_write(cp, upper_32_bits(ib->gpu_addr) & 0xFF);
radeon_ring_write(cp, ib->length_dw);
}
int r600_ib_test(struct radeon_device *rdev)
int r600_ib_test(struct radeon_device *rdev, int ring)
{
struct radeon_ib *ib;
uint32_t scratch;
@ -2700,7 +2708,7 @@ int r600_ib_test(struct radeon_device *rdev)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = radeon_ib_get(rdev, &ib);
r = radeon_ib_get(rdev, ring, &ib);
if (r) {
DRM_ERROR("radeon: failed to get ib (%d).\n", r);
return r;
@ -2741,7 +2749,7 @@ int r600_ib_test(struct radeon_device *rdev)
DRM_UDELAY(1);
}
if (i < rdev->usec_timeout) {
DRM_INFO("ib test succeeded in %u usecs\n", i);
DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib->fence->ring, i);
} else {
DRM_ERROR("radeon: ib test failed (scratch(0x%04X)=0x%08X)\n",
scratch, tmp);
@ -3514,21 +3522,22 @@ static int r600_debugfs_cp_ring_info(struct seq_file *m, void *data)
struct drm_info_node *node = (struct drm_info_node *) m->private;
struct drm_device *dev = node->minor->dev;
struct radeon_device *rdev = dev->dev_private;
struct radeon_cp *cp = &rdev->cp;
unsigned count, i, j;
radeon_ring_free_size(rdev);
count = (rdev->cp.ring_size / 4) - rdev->cp.ring_free_dw;
radeon_ring_free_size(rdev, cp);
count = (cp->ring_size / 4) - cp->ring_free_dw;
seq_printf(m, "CP_STAT 0x%08x\n", RREG32(CP_STAT));
seq_printf(m, "CP_RB_WPTR 0x%08x\n", RREG32(CP_RB_WPTR));
seq_printf(m, "CP_RB_RPTR 0x%08x\n", RREG32(CP_RB_RPTR));
seq_printf(m, "driver's copy of the CP_RB_WPTR 0x%08x\n", rdev->cp.wptr);
seq_printf(m, "driver's copy of the CP_RB_RPTR 0x%08x\n", rdev->cp.rptr);
seq_printf(m, "%u free dwords in ring\n", rdev->cp.ring_free_dw);
seq_printf(m, "driver's copy of the CP_RB_WPTR 0x%08x\n", cp->wptr);
seq_printf(m, "driver's copy of the CP_RB_RPTR 0x%08x\n", cp->rptr);
seq_printf(m, "%u free dwords in ring\n", cp->ring_free_dw);
seq_printf(m, "%u dwords in ring\n", count);
i = rdev->cp.rptr;
i = cp->rptr;
for (j = 0; j <= count; j++) {
seq_printf(m, "r[%04d]=0x%08x\n", i, rdev->cp.ring[i]);
i = (i + 1) & rdev->cp.ptr_mask;
seq_printf(m, "r[%04d]=0x%08x\n", i, cp->ring[i]);
i = (i + 1) & cp->ptr_mask;
}
return 0;
}

217
drivers/gpu/drm/radeon/r600_blit_kms.c
View File

@ -50,6 +50,7 @@ static void
set_render_target(struct radeon_device *rdev, int format,
int w, int h, u64 gpu_addr)
{
struct radeon_cp *cp = &rdev->cp;
u32 cb_color_info;
int pitch, slice;
@ -63,38 +64,38 @@ set_render_target(struct radeon_device *rdev, int format,
pitch = (w / 8) - 1;
slice = ((w * h) / 64) - 1;
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, gpu_addr >> 8);
if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770) {
radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_BASE_UPDATE, 0));
radeon_ring_write(rdev, 2 << 0);
radeon_ring_write(cp, PACKET3(PACKET3_SURFACE_BASE_UPDATE, 0));
radeon_ring_write(cp, 2 << 0);
}
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (CB_COLOR0_SIZE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, (pitch << 0) | (slice << 10));
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (CB_COLOR0_SIZE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, (pitch << 0) | (slice << 10));
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (CB_COLOR0_VIEW - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (CB_COLOR0_VIEW - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, 0);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (CB_COLOR0_INFO - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, cb_color_info);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (CB_COLOR0_INFO - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, cb_color_info);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (CB_COLOR0_TILE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (CB_COLOR0_TILE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, 0);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (CB_COLOR0_FRAG - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (CB_COLOR0_FRAG - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, 0);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (CB_COLOR0_MASK - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (CB_COLOR0_MASK - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, 0);
}
/* emits 5dw */
@ -103,6 +104,7 @@ cp_set_surface_sync(struct radeon_device *rdev,
u32 sync_type, u32 size,
u64 mc_addr)
{
struct radeon_cp *cp = &rdev->cp;
u32 cp_coher_size;
if (size == 0xffffffff)
@ -110,17 +112,18 @@ cp_set_surface_sync(struct radeon_device *rdev,
else
cp_coher_size = ((size + 255) >> 8);
radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(rdev, sync_type);
radeon_ring_write(rdev, cp_coher_size);
radeon_ring_write(rdev, mc_addr >> 8);
radeon_ring_write(rdev, 10); /* poll interval */
radeon_ring_write(cp, PACKET3(PACKET3_SURFACE_SYNC, 3));
radeon_ring_write(cp, sync_type);
radeon_ring_write(cp, cp_coher_size);
radeon_ring_write(cp, mc_addr >> 8);
radeon_ring_write(cp, 10); /* poll interval */
}
/* emits 21dw + 1 surface sync = 26dw */
static void
set_shaders(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
u64 gpu_addr;
u32 sq_pgm_resources;
@ -129,35 +132,35 @@ set_shaders(struct radeon_device *rdev)
/* VS */
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (SQ_PGM_RESOURCES_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, sq_pgm_resources);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (SQ_PGM_RESOURCES_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, sq_pgm_resources);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (SQ_PGM_CF_OFFSET_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (SQ_PGM_CF_OFFSET_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, 0);
/* PS */
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset;
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (SQ_PGM_RESOURCES_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, sq_pgm_resources | (1 << 28));
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (SQ_PGM_RESOURCES_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, sq_pgm_resources | (1 << 28));
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (SQ_PGM_EXPORTS_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, 2);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (SQ_PGM_EXPORTS_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, 2);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(rdev, (SQ_PGM_CF_OFFSET_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, 0);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
radeon_ring_write(cp, (SQ_PGM_CF_OFFSET_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, 0);
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr);
@ -167,6 +170,7 @@ set_shaders(struct radeon_device *rdev)
static void
set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
{
struct radeon_cp *cp = &rdev->cp;
u32 sq_vtx_constant_word2;
sq_vtx_constant_word2 = SQ_VTXC_BASE_ADDR_HI(upper_32_bits(gpu_addr) & 0xff) |
@ -175,15 +179,15 @@ set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
sq_vtx_constant_word2 |= SQ_VTXC_ENDIAN_SWAP(SQ_ENDIAN_8IN32);
#endif
radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 7));
radeon_ring_write(rdev, 0x460);
radeon_ring_write(rdev, gpu_addr & 0xffffffff);
radeon_ring_write(rdev, 48 - 1);
radeon_ring_write(rdev, sq_vtx_constant_word2);
radeon_ring_write(rdev, 1 << 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, SQ_TEX_VTX_VALID_BUFFER << 30);
radeon_ring_write(cp, PACKET3(PACKET3_SET_RESOURCE, 7));
radeon_ring_write(cp, 0x460);
radeon_ring_write(cp, gpu_addr & 0xffffffff);
radeon_ring_write(cp, 48 - 1);
radeon_ring_write(cp, sq_vtx_constant_word2);
radeon_ring_write(cp, 1 << 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, SQ_TEX_VTX_VALID_BUFFER << 30);
if ((rdev->family == CHIP_RV610) ||
(rdev->family == CHIP_RV620) ||
@ -203,6 +207,7 @@ set_tex_resource(struct radeon_device *rdev,
int format, int w, int h, int pitch,
u64 gpu_addr, u32 size)
{
struct radeon_cp *cp = &rdev->cp;
uint32_t sq_tex_resource_word0, sq_tex_resource_word1, sq_tex_resource_word4;
if (h < 1)
@ -225,15 +230,15 @@ set_tex_resource(struct radeon_device *rdev,
cp_set_surface_sync(rdev,
PACKET3_TC_ACTION_ENA, size, gpu_addr);
radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 7));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, sq_tex_resource_word0);
radeon_ring_write(rdev, sq_tex_resource_word1);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(rdev, gpu_addr >> 8);
radeon_ring_write(rdev, sq_tex_resource_word4);
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, SQ_TEX_VTX_VALID_TEXTURE << 30);
radeon_ring_write(cp, PACKET3(PACKET3_SET_RESOURCE, 7));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, sq_tex_resource_word0);
radeon_ring_write(cp, sq_tex_resource_word1);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(cp, gpu_addr >> 8);
radeon_ring_write(cp, sq_tex_resource_word4);
radeon_ring_write(cp, 0);
radeon_ring_write(cp, SQ_TEX_VTX_VALID_TEXTURE << 30);
}
/* emits 12 */
@ -241,43 +246,45 @@ static void
set_scissors(struct radeon_device *rdev, int x1, int y1,
int x2, int y2)
{
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(rdev, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, (x1 << 0) | (y1 << 16));
radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
struct radeon_cp *cp = &rdev->cp;
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(cp, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, (x1 << 0) | (y1 << 16));
radeon_ring_write(cp, (x2 << 0) | (y2 << 16));
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(rdev, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(cp, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(cp, (x2 << 0) | (y2 << 16));
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(rdev, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
radeon_ring_write(cp, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
radeon_ring_write(cp, (x1 << 0) | (y1 << 16) | (1 << 31));
radeon_ring_write(cp, (x2 << 0) | (y2 << 16));
}
/* emits 10 */
static void
draw_auto(struct radeon_device *rdev)
{
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(rdev, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
radeon_ring_write(rdev, DI_PT_RECTLIST);
struct radeon_cp *cp = &rdev->cp;
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(cp, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
radeon_ring_write(cp, DI_PT_RECTLIST);
radeon_ring_write(rdev, PACKET3(PACKET3_INDEX_TYPE, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET3(PACKET3_INDEX_TYPE, 0));
radeon_ring_write(cp,
#ifdef __BIG_ENDIAN
(2 << 2) |
#endif
DI_INDEX_SIZE_16_BIT);
radeon_ring_write(rdev, PACKET3(PACKET3_NUM_INSTANCES, 0));
radeon_ring_write(rdev, 1);
radeon_ring_write(cp, PACKET3(PACKET3_NUM_INSTANCES, 0));
radeon_ring_write(cp, 1);
radeon_ring_write(rdev, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
radeon_ring_write(rdev, 3);
radeon_ring_write(rdev, DI_SRC_SEL_AUTO_INDEX);
radeon_ring_write(cp, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
radeon_ring_write(cp, 3);
radeon_ring_write(cp, DI_SRC_SEL_AUTO_INDEX);
}
@ -285,6 +292,7 @@ draw_auto(struct radeon_device *rdev)
static void
set_default_state(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2;
u32 sq_thread_resource_mgmt, sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2;
int num_ps_gprs, num_vs_gprs, num_temp_gprs, num_gs_gprs, num_es_gprs;
@ -440,24 +448,24 @@ set_default_state(struct radeon_device *rdev)
/* emit an IB pointing at default state */
dwords = ALIGN(rdev->r600_blit.state_len, 0x10);
gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset;
radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
radeon_ring_write(cp,
#ifdef __BIG_ENDIAN
(2 << 0) |
#endif
(gpu_addr & 0xFFFFFFFC));
radeon_ring_write(rdev, upper_32_bits(gpu_addr) & 0xFF);
radeon_ring_write(rdev, dwords);
radeon_ring_write(cp, upper_32_bits(gpu_addr) & 0xFF);
radeon_ring_write(cp, dwords);
/* SQ config */
radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 6));
radeon_ring_write(rdev, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
radeon_ring_write(rdev, sq_config);
radeon_ring_write(rdev, sq_gpr_resource_mgmt_1);
radeon_ring_write(rdev, sq_gpr_resource_mgmt_2);
radeon_ring_write(rdev, sq_thread_resource_mgmt);
radeon_ring_write(rdev, sq_stack_resource_mgmt_1);
radeon_ring_write(rdev, sq_stack_resource_mgmt_2);
radeon_ring_write(cp, PACKET3(PACKET3_SET_CONFIG_REG, 6));
radeon_ring_write(cp, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
radeon_ring_write(cp, sq_config);
radeon_ring_write(cp, sq_gpr_resource_mgmt_1);
radeon_ring_write(cp, sq_gpr_resource_mgmt_2);
radeon_ring_write(cp, sq_thread_resource_mgmt);
radeon_ring_write(cp, sq_stack_resource_mgmt_1);
radeon_ring_write(cp, sq_stack_resource_mgmt_2);
}
static uint32_t i2f(uint32_t input)
@ -614,7 +622,7 @@ void r600_blit_fini(struct radeon_device *rdev)
static int r600_vb_ib_get(struct radeon_device *rdev)
{
int r;
r = radeon_ib_get(rdev, &rdev->r600_blit.vb_ib);
r = radeon_ib_get(rdev, RADEON_RING_TYPE_GFX_INDEX, &rdev->r600_blit.vb_ib);
if (r) {
DRM_ERROR("failed to get IB for vertex buffer\n");
return r;
@ -679,6 +687,7 @@ static unsigned r600_blit_create_rect(unsigned num_gpu_pages,
int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_pages)
{
struct radeon_cp *cp = &rdev->cp;
int r;
int ring_size;
int num_loops = 0;
@ -699,7 +708,7 @@ int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_pages)
/* calculate number of loops correctly */
ring_size = num_loops * dwords_per_loop;
ring_size += rdev->r600_blit.ring_size_common;
r = radeon_ring_lock(rdev, ring_size);
r = radeon_ring_lock(rdev, cp, ring_size);
if (r)
return r;
@ -718,7 +727,7 @@ void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence *fence)
if (fence)
r = radeon_fence_emit(rdev, fence);
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, &rdev->cp);
}
void r600_kms_blit_copy(struct radeon_device *rdev,

53
drivers/gpu/drm/radeon/radeon.h
View File

@ -230,6 +230,8 @@ void radeon_fence_unref(struct radeon_fence **fence);
/*
* Semaphores.
*/
struct radeon_cp;
struct radeon_semaphore_driver {
rwlock_t lock;
struct list_head free;
@ -585,7 +587,7 @@ struct r600_blit {
void r600_blit_suspend(struct radeon_device *rdev);
int radeon_ib_get(struct radeon_device *rdev, struct radeon_ib **ib);
int radeon_ib_get(struct radeon_device *rdev, int ring, struct radeon_ib **ib);
void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib **ib);
int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib);
int radeon_ib_pool_init(struct radeon_device *rdev);
@ -593,15 +595,15 @@ void radeon_ib_pool_fini(struct radeon_device *rdev);
int radeon_ib_test(struct radeon_device *rdev);
extern void radeon_ib_bogus_add(struct radeon_device *rdev, struct radeon_ib *ib);
/* Ring access between begin & end cannot sleep */
void radeon_ring_free_size(struct radeon_device *rdev);
int radeon_ring_alloc(struct radeon_device *rdev, unsigned ndw);
int radeon_ring_lock(struct radeon_device *rdev, unsigned ndw);
void radeon_ring_commit(struct radeon_device *rdev);
void radeon_ring_unlock_commit(struct radeon_device *rdev);
void radeon_ring_unlock_undo(struct radeon_device *rdev);
int radeon_ring_test(struct radeon_device *rdev);
int radeon_ring_init(struct radeon_device *rdev, unsigned ring_size);
void radeon_ring_fini(struct radeon_device *rdev);
void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_cp *cp);
int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ndw);
int radeon_ring_lock(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ndw);
void radeon_ring_commit(struct radeon_device *rdev, struct radeon_cp *cp);
void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_cp *cp);
void radeon_ring_unlock_undo(struct radeon_device *rdev, struct radeon_cp *cp);
int radeon_ring_test(struct radeon_device *rdev, struct radeon_cp *cp);
int radeon_ring_init(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ring_size);
void radeon_ring_fini(struct radeon_device *rdev, struct radeon_cp *cp);
/*
@ -930,24 +932,25 @@ struct radeon_asic {
int (*resume)(struct radeon_device *rdev);
int (*suspend)(struct radeon_device *rdev);
void (*vga_set_state)(struct radeon_device *rdev, bool state);
bool (*gpu_is_lockup)(struct radeon_device *rdev);
bool (*gpu_is_lockup)(struct radeon_device *rdev, struct radeon_cp *cp);
int (*asic_reset)(struct radeon_device *rdev);
void (*gart_tlb_flush)(struct radeon_device *rdev);
int (*gart_set_page)(struct radeon_device *rdev, int i, uint64_t addr);
int (*cp_init)(struct radeon_device *rdev, unsigned ring_size);
void (*cp_fini)(struct radeon_device *rdev);
void (*cp_disable)(struct radeon_device *rdev);
void (*cp_commit)(struct radeon_device *rdev);
void (*cp_commit)(struct radeon_device *rdev, struct radeon_cp *cp);
void (*ring_start)(struct radeon_device *rdev);
int (*ring_test)(struct radeon_device *rdev);
int (*ring_test)(struct radeon_device *rdev, struct radeon_cp *cp);
void (*ring_ib_execute)(struct radeon_device *rdev, struct radeon_ib *ib);
int (*irq_set)(struct radeon_device *rdev);
int (*irq_process)(struct radeon_device *rdev);
u32 (*get_vblank_counter)(struct radeon_device *rdev, int crtc);
void (*fence_ring_emit)(struct radeon_device *rdev, struct radeon_fence *fence);
void (*semaphore_ring_emit)(struct radeon_device *rdev,
struct radeon_cp *cp,
struct radeon_semaphore *semaphore,
unsigned ring, bool emit_wait);
bool emit_wait);
int (*cs_parse)(struct radeon_cs_parser *p);
int (*copy_blit)(struct radeon_device *rdev,
uint64_t src_offset,
@ -1279,7 +1282,6 @@ struct radeon_device {
struct radeon_fence_driver fence_drv[RADEON_NUM_RINGS];
struct radeon_semaphore_driver semaphore_drv;
struct radeon_cp cp;
/* cayman compute rings */
struct radeon_cp cp1;
struct radeon_cp cp2;
struct radeon_ib_pool ib_pool;
@ -1463,18 +1465,17 @@ void radeon_atombios_fini(struct radeon_device *rdev);
/*
* RING helpers.
*/
#if DRM_DEBUG_CODE == 0
static inline void radeon_ring_write(struct radeon_device *rdev, uint32_t v)
static inline void radeon_ring_write(struct radeon_cp *cp, uint32_t v)
{
rdev->cp.ring[rdev->cp.wptr++] = v;
rdev->cp.wptr &= rdev->cp.ptr_mask;
rdev->cp.count_dw--;
rdev->cp.ring_free_dw--;
cp->ring[cp->wptr++] = v;
cp->wptr &= cp->ptr_mask;
cp->count_dw--;
cp->ring_free_dw--;
}
#else
/* With debugging this is just too big to inline */
void radeon_ring_write(struct radeon_device *rdev, uint32_t v);
void radeon_ring_write(struct radeon_cp *cp, uint32_t v);
#endif
/*
@ -1486,19 +1487,19 @@ void radeon_ring_write(struct radeon_device *rdev, uint32_t v);
#define radeon_suspend(rdev) (rdev)->asic->suspend((rdev))
#define radeon_cs_parse(p) rdev->asic->cs_parse((p))
#define radeon_vga_set_state(rdev, state) (rdev)->asic->vga_set_state((rdev), (state))
#define radeon_gpu_is_lockup(rdev) (rdev)->asic->gpu_is_lockup((rdev))
#define radeon_gpu_is_lockup(rdev, cp) (rdev)->asic->gpu_is_lockup((rdev), (cp))
#define radeon_asic_reset(rdev) (rdev)->asic->asic_reset((rdev))
#define radeon_gart_tlb_flush(rdev) (rdev)->asic->gart_tlb_flush((rdev))
#define radeon_gart_set_page(rdev, i, p) (rdev)->asic->gart_set_page((rdev), (i), (p))
#define radeon_cp_commit(rdev) (rdev)->asic->cp_commit((rdev))
#define radeon_cp_commit(rdev, cp) (rdev)->asic->cp_commit((rdev), (cp))
#define radeon_ring_start(rdev) (rdev)->asic->ring_start((rdev))
#define radeon_ring_test(rdev) (rdev)->asic->ring_test((rdev))
#define radeon_ring_test(rdev, cp) (rdev)->asic->ring_test((rdev), (cp))
#define radeon_ring_ib_execute(rdev, ib) (rdev)->asic->ring_ib_execute((rdev), (ib))
#define radeon_irq_set(rdev) (rdev)->asic->irq_set((rdev))
#define radeon_irq_process(rdev) (rdev)->asic->irq_process((rdev))
#define radeon_get_vblank_counter(rdev, crtc) (rdev)->asic->get_vblank_counter((rdev), (crtc))
#define radeon_fence_ring_emit(rdev, fence) (rdev)->asic->fence_ring_emit((rdev), (fence))
#define radeon_semaphore_ring_emit(rdev, semaphore, ring, emit_wait) (rdev)->asic->semaphore_ring_emit((rdev), (semaphore), (ring), (emit_wait))
#define radeon_semaphore_ring_emit(rdev, cp, semaphore, emit_wait) (rdev)->asic->semaphore_ring_emit((rdev), (cp), (semaphore), (emit_wait))
#define radeon_copy_blit(rdev, s, d, np, f) (rdev)->asic->copy_blit((rdev), (s), (d), (np), (f))
#define radeon_copy_dma(rdev, s, d, np, f) (rdev)->asic->copy_dma((rdev), (s), (d), (np), (f))
#define radeon_copy(rdev, s, d, np, f) (rdev)->asic->copy((rdev), (s), (d), (np), (f))

28
drivers/gpu/drm/radeon/radeon_asic.h
View File

@ -58,20 +58,21 @@ void r100_fini(struct radeon_device *rdev);
int r100_suspend(struct radeon_device *rdev);
int r100_resume(struct radeon_device *rdev);
void r100_vga_set_state(struct radeon_device *rdev, bool state);
bool r100_gpu_is_lockup(struct radeon_device *rdev);
bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp);
int r100_asic_reset(struct radeon_device *rdev);
u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc);
void r100_pci_gart_tlb_flush(struct radeon_device *rdev);
int r100_pci_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr);
void r100_cp_commit(struct radeon_device *rdev);
void r100_cp_commit(struct radeon_device *rdev, struct radeon_cp *cp);
void r100_ring_start(struct radeon_device *rdev);
int r100_irq_set(struct radeon_device *rdev);
int r100_irq_process(struct radeon_device *rdev);
void r100_fence_ring_emit(struct radeon_device *rdev,
struct radeon_fence *fence);
void r100_semaphore_ring_emit(struct radeon_device *rdev,
struct radeon_cp *cp,
struct radeon_semaphore *semaphore,
unsigned ring, bool emit_wait);
bool emit_wait);
int r100_cs_parse(struct radeon_cs_parser *p);
void r100_pll_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
uint32_t r100_pll_rreg(struct radeon_device *rdev, uint32_t reg);
@ -86,7 +87,7 @@ int r100_set_surface_reg(struct radeon_device *rdev, int reg,
void r100_clear_surface_reg(struct radeon_device *rdev, int reg);
void r100_bandwidth_update(struct radeon_device *rdev);
void r100_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
int r100_ring_test(struct radeon_device *rdev);
int r100_ring_test(struct radeon_device *rdev, struct radeon_cp *cp);
void r100_hpd_init(struct radeon_device *rdev);
void r100_hpd_fini(struct radeon_device *rdev);
bool r100_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd);
@ -157,7 +158,7 @@ extern int r300_init(struct radeon_device *rdev);
extern void r300_fini(struct radeon_device *rdev);
extern int r300_suspend(struct radeon_device *rdev);
extern int r300_resume(struct radeon_device *rdev);
extern bool r300_gpu_is_lockup(struct radeon_device *rdev);
extern bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp);
extern int r300_asic_reset(struct radeon_device *rdev);
extern void r300_ring_start(struct radeon_device *rdev);
extern void r300_fence_ring_emit(struct radeon_device *rdev,
@ -296,7 +297,7 @@ int r600_resume(struct radeon_device *rdev);
void r600_vga_set_state(struct radeon_device *rdev, bool state);
int r600_wb_init(struct radeon_device *rdev);
void r600_wb_fini(struct radeon_device *rdev);
void r600_cp_commit(struct radeon_device *rdev);
void r600_cp_commit(struct radeon_device *rdev, struct radeon_cp *cp);
void r600_pcie_gart_tlb_flush(struct radeon_device *rdev);
uint32_t r600_pciep_rreg(struct radeon_device *rdev, uint32_t reg);
void r600_pciep_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
@ -304,17 +305,18 @@ int r600_cs_parse(struct radeon_cs_parser *p);
void r600_fence_ring_emit(struct radeon_device *rdev,
struct radeon_fence *fence);
void r600_semaphore_ring_emit(struct radeon_device *rdev,
struct radeon_cp *cp,
struct radeon_semaphore *semaphore,
unsigned ring, bool emit_wait);
bool r600_gpu_is_lockup(struct radeon_device *rdev);
bool emit_wait);
bool r600_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp);
int r600_asic_reset(struct radeon_device *rdev);
int r600_set_surface_reg(struct radeon_device *rdev, int reg,
uint32_t tiling_flags, uint32_t pitch,
uint32_t offset, uint32_t obj_size);
void r600_clear_surface_reg(struct radeon_device *rdev, int reg);
int r600_ib_test(struct radeon_device *rdev);
int r600_ib_test(struct radeon_device *rdev, int ring);
void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
int r600_ring_test(struct radeon_device *rdev);
int r600_ring_test(struct radeon_device *rdev, struct radeon_cp *cp);
int r600_copy_blit(struct radeon_device *rdev,
uint64_t src_offset, uint64_t dst_offset,
unsigned num_gpu_pages, struct radeon_fence *fence);
@ -334,7 +336,7 @@ extern int r600_get_pcie_lanes(struct radeon_device *rdev);
bool r600_card_posted(struct radeon_device *rdev);
void r600_cp_stop(struct radeon_device *rdev);
int r600_cp_start(struct radeon_device *rdev);
void r600_ring_init(struct radeon_device *rdev, unsigned ring_size);
void r600_ring_init(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ring_size);
int r600_cp_resume(struct radeon_device *rdev);
void r600_cp_fini(struct radeon_device *rdev);
int r600_count_pipe_bits(uint32_t val);
@ -403,7 +405,7 @@ int evergreen_init(struct radeon_device *rdev);
void evergreen_fini(struct radeon_device *rdev);
int evergreen_suspend(struct radeon_device *rdev);
int evergreen_resume(struct radeon_device *rdev);
bool evergreen_gpu_is_lockup(struct radeon_device *rdev);
bool evergreen_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp);
int evergreen_asic_reset(struct radeon_device *rdev);
void evergreen_bandwidth_update(struct radeon_device *rdev);
void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
@ -434,7 +436,7 @@ int cayman_init(struct radeon_device *rdev);
void cayman_fini(struct radeon_device *rdev);
int cayman_suspend(struct radeon_device *rdev);
int cayman_resume(struct radeon_device *rdev);
bool cayman_gpu_is_lockup(struct radeon_device *rdev);
bool cayman_gpu_is_lockup(struct radeon_device *rdev, struct radeon_cp *cp);
int cayman_asic_reset(struct radeon_device *rdev);
#endif

2
drivers/gpu/drm/radeon/radeon_cs.c
View File

@ -246,7 +246,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
radeon_mutex_unlock(&rdev->cs_mutex);
return r;
}
r = radeon_ib_get(rdev, &parser.ib);
r = radeon_ib_get(rdev, RADEON_RING_TYPE_GFX_INDEX, &parser.ib);
if (r) {
DRM_ERROR("Failed to get ib !\n");
radeon_cs_parser_fini(&parser, r);

2
drivers/gpu/drm/radeon/radeon_fence.c
View File

@ -269,7 +269,7 @@ retry:
* if we experiencing a lockup the value doesn't change
*/
if (seq == rdev->fence_drv[fence->ring].last_seq &&
radeon_gpu_is_lockup(rdev)) {
radeon_gpu_is_lockup(rdev, &rdev->cp)) {
/* good news we believe it's a lockup */
printk(KERN_WARNING "GPU lockup (waiting for 0x%08X last fence id 0x%08X)\n",
fence->seq, seq);

4
drivers/gpu/drm/radeon/radeon_gem.c
View File

@ -160,8 +160,8 @@ int radeon_gem_info_ioctl(struct drm_device *dev, void *data,
if (rdev->stollen_vga_memory)
args->vram_visible -= radeon_bo_size(rdev->stollen_vga_memory);
args->vram_visible -= radeon_fbdev_total_size(rdev);
args->gart_size = rdev->mc.gtt_size - rdev->cp.ring_size - 4096 -
RADEON_IB_POOL_SIZE*64*1024;
args->gart_size = rdev->mc.gtt_size - 4096 - RADEON_IB_POOL_SIZE*64*1024;
args->gart_size -= rdev->cp.ring_size;
return 0;
}

13
drivers/gpu/drm/radeon/radeon_pm.c
View File

@ -252,7 +252,8 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
mutex_lock(&rdev->ddev->struct_mutex);
mutex_lock(&rdev->vram_mutex);
mutex_lock(&rdev->cp.mutex);
if (rdev->cp.ring_obj)
mutex_lock(&rdev->cp.mutex);
/* gui idle int has issues on older chips it seems */
if (rdev->family >= CHIP_R600) {
@ -268,12 +269,13 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
radeon_irq_set(rdev);
}
} else {
if (rdev->cp.ready) {
struct radeon_cp *cp = &rdev->cp;
if (cp->ready) {
struct radeon_fence *fence;
radeon_ring_alloc(rdev, 64);
radeon_ring_alloc(rdev, cp, 64);
radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
radeon_fence_emit(rdev, fence);
radeon_ring_commit(rdev);
radeon_ring_commit(rdev, cp);
radeon_fence_wait(fence, false);
radeon_fence_unref(&fence);
}
@ -307,7 +309,8 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
rdev->pm.dynpm_planned_action = DYNPM_ACTION_NONE;
mutex_unlock(&rdev->cp.mutex);
if (rdev->cp.ring_obj)
mutex_unlock(&rdev->cp.mutex);
mutex_unlock(&rdev->vram_mutex);
mutex_unlock(&rdev->ddev->struct_mutex);
}

121
drivers/gpu/drm/radeon/radeon_ring.c
View File

@ -60,17 +60,17 @@ u32 radeon_get_ib_value(struct radeon_cs_parser *p, int idx)
return idx_value;
}
void radeon_ring_write(struct radeon_device *rdev, uint32_t v)
void radeon_ring_write(struct radeon_cp *cp, uint32_t v)
{
#if DRM_DEBUG_CODE
if (rdev->cp.count_dw <= 0) {
if (cp->count_dw <= 0) {
DRM_ERROR("radeon: writting more dword to ring than expected !\n");
}
#endif
rdev->cp.ring[rdev->cp.wptr++] = v;
rdev->cp.wptr &= rdev->cp.ptr_mask;
rdev->cp.count_dw--;
rdev->cp.ring_free_dw--;
cp->ring[cp->wptr++] = v;
cp->wptr &= cp->ptr_mask;
cp->count_dw--;
cp->ring_free_dw--;
}
void radeon_ib_bogus_cleanup(struct radeon_device *rdev)
@ -106,14 +106,14 @@ void radeon_ib_bogus_add(struct radeon_device *rdev, struct radeon_ib *ib)
/*
* IB.
*/
int radeon_ib_get(struct radeon_device *rdev, struct radeon_ib **ib)
int radeon_ib_get(struct radeon_device *rdev, int ring, struct radeon_ib **ib)
{
struct radeon_fence *fence;
struct radeon_ib *nib;
int r = 0, i, c;
*ib = NULL;
r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
r = radeon_fence_create(rdev, &fence, ring);
if (r) {
dev_err(rdev->dev, "failed to create fence for new IB\n");
return r;
@ -178,16 +178,17 @@ void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib **ib)
int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib)
{
struct radeon_cp *cp = &rdev->cp;
int r = 0;
if (!ib->length_dw || !rdev->cp.ready) {
if (!ib->length_dw || !cp->ready) {
/* TODO: Nothings in the ib we should report. */
DRM_ERROR("radeon: couldn't schedule IB(%u).\n", ib->idx);
return -EINVAL;
}
/* 64 dwords should be enough for fence too */
r = radeon_ring_lock(rdev, 64);
r = radeon_ring_lock(rdev, cp, 64);
if (r) {
DRM_ERROR("radeon: scheduling IB failed (%d).\n", r);
return r;
@ -198,7 +199,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib)
/* once scheduled IB is considered free and protected by the fence */
ib->free = true;
mutex_unlock(&rdev->ib_pool.mutex);
radeon_ring_unlock_commit(rdev);
radeon_ring_unlock_commit(rdev, cp);
return 0;
}
@ -283,7 +284,7 @@ void radeon_ib_pool_fini(struct radeon_device *rdev)
/*
* Ring.
*/
void radeon_ring_free_size(struct radeon_device *rdev)
void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_cp *cp)
{
if (rdev->wb.enabled)
rdev->cp.rptr = le32_to_cpu(rdev->wb.wb[RADEON_WB_CP_RPTR_OFFSET/4]);
@ -294,122 +295,123 @@ void radeon_ring_free_size(struct radeon_device *rdev)
rdev->cp.rptr = RREG32(RADEON_CP_RB_RPTR);
}
/* This works because ring_size is a power of 2 */
rdev->cp.ring_free_dw = (rdev->cp.rptr + (rdev->cp.ring_size / 4));
rdev->cp.ring_free_dw -= rdev->cp.wptr;
rdev->cp.ring_free_dw &= rdev->cp.ptr_mask;
if (!rdev->cp.ring_free_dw) {
rdev->cp.ring_free_dw = rdev->cp.ring_size / 4;
cp->ring_free_dw = (cp->rptr + (cp->ring_size / 4));
cp->ring_free_dw -= cp->wptr;
cp->ring_free_dw &= cp->ptr_mask;
if (!cp->ring_free_dw) {
cp->ring_free_dw = cp->ring_size / 4;
}
}
int radeon_ring_alloc(struct radeon_device *rdev, unsigned ndw)
int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ndw)
{
int r;
/* Align requested size with padding so unlock_commit can
* pad safely */
ndw = (ndw + rdev->cp.align_mask) & ~rdev->cp.align_mask;
while (ndw > (rdev->cp.ring_free_dw - 1)) {
radeon_ring_free_size(rdev);
if (ndw < rdev->cp.ring_free_dw) {
ndw = (ndw + cp->align_mask) & ~cp->align_mask;
while (ndw > (cp->ring_free_dw - 1)) {
radeon_ring_free_size(rdev, cp);
if (ndw < cp->ring_free_dw) {
break;
}
r = radeon_fence_wait_next(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r)
return r;
}
rdev->cp.count_dw = ndw;
rdev->cp.wptr_old = rdev->cp.wptr;
cp->count_dw = ndw;
cp->wptr_old = cp->wptr;
return 0;
}
int radeon_ring_lock(struct radeon_device *rdev, unsigned ndw)
int radeon_ring_lock(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ndw)
{
int r;
mutex_lock(&rdev->cp.mutex);
r = radeon_ring_alloc(rdev, ndw);
mutex_lock(&cp->mutex);
r = radeon_ring_alloc(rdev, cp, ndw);
if (r) {
mutex_unlock(&rdev->cp.mutex);
mutex_unlock(&cp->mutex);
return r;
}
return 0;
}
void radeon_ring_commit(struct radeon_device *rdev)
void radeon_ring_commit(struct radeon_device *rdev, struct radeon_cp *cp)
{
unsigned count_dw_pad;
unsigned i;
/* We pad to match fetch size */
count_dw_pad = (rdev->cp.align_mask + 1) -
(rdev->cp.wptr & rdev->cp.align_mask);
count_dw_pad = (cp->align_mask + 1) -
(cp->wptr & cp->align_mask);
for (i = 0; i < count_dw_pad; i++) {
radeon_ring_write(rdev, 2 << 30);
radeon_ring_write(cp, 2 << 30);
}
DRM_MEMORYBARRIER();
radeon_cp_commit(rdev);
radeon_cp_commit(rdev, cp);
}
void radeon_ring_unlock_commit(struct radeon_device *rdev)
void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_cp *cp)
{
radeon_ring_commit(rdev);
mutex_unlock(&rdev->cp.mutex);
radeon_ring_commit(rdev, cp);
mutex_unlock(&cp->mutex);
}
void radeon_ring_unlock_undo(struct radeon_device *rdev)
void radeon_ring_unlock_undo(struct radeon_device *rdev, struct radeon_cp *cp)
{
rdev->cp.wptr = rdev->cp.wptr_old;
mutex_unlock(&rdev->cp.mutex);
cp->wptr = cp->wptr_old;
mutex_unlock(&cp->mutex);
}
int radeon_ring_init(struct radeon_device *rdev, unsigned ring_size)
int radeon_ring_init(struct radeon_device *rdev, struct radeon_cp *cp, unsigned ring_size)
{
int r;
rdev->cp.ring_size = ring_size;
cp->ring_size = ring_size;
/* Allocate ring buffer */
if (rdev->cp.ring_obj == NULL) {
r = radeon_bo_create(rdev, rdev->cp.ring_size, PAGE_SIZE, true,
if (cp->ring_obj == NULL) {
r = radeon_bo_create(rdev, cp->ring_size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_GTT,
&rdev->cp.ring_obj);
&cp->ring_obj);
if (r) {
dev_err(rdev->dev, "(%d) ring create failed\n", r);
return r;
}
r = radeon_bo_reserve(rdev->cp.ring_obj, false);
r = radeon_bo_reserve(cp->ring_obj, false);
if (unlikely(r != 0))
return r;
r = radeon_bo_pin(rdev->cp.ring_obj, RADEON_GEM_DOMAIN_GTT,
&rdev->cp.gpu_addr);
r = radeon_bo_pin(cp->ring_obj, RADEON_GEM_DOMAIN_GTT,
&cp->gpu_addr);
if (r) {
radeon_bo_unreserve(rdev->cp.ring_obj);
radeon_bo_unreserve(cp->ring_obj);
dev_err(rdev->dev, "(%d) ring pin failed\n", r);
return r;
}
r = radeon_bo_kmap(rdev->cp.ring_obj,
(void **)&rdev->cp.ring);
radeon_bo_unreserve(rdev->cp.ring_obj);
r = radeon_bo_kmap(cp->ring_obj,
(void **)&cp->ring);
radeon_bo_unreserve(cp->ring_obj);
if (r) {
dev_err(rdev->dev, "(%d) ring map failed\n", r);
return r;
}
}
rdev->cp.ptr_mask = (rdev->cp.ring_size / 4) - 1;
rdev->cp.ring_free_dw = rdev->cp.ring_size / 4;
cp->ptr_mask = (cp->ring_size / 4) - 1;
cp->ring_free_dw = cp->ring_size / 4;
return 0;
}
void radeon_ring_fini(struct radeon_device *rdev)
void radeon_ring_fini(struct radeon_device *rdev, struct radeon_cp *cp)
{
int r;
struct radeon_bo *ring_obj;
mutex_lock(&rdev->cp.mutex);
ring_obj = rdev->cp.ring_obj;
rdev->cp.ring = NULL;
rdev->cp.ring_obj = NULL;
mutex_unlock(&rdev->cp.mutex);
mutex_lock(&cp->mutex);
ring_obj = cp->ring_obj;
cp->ring = NULL;
cp->ring_obj = NULL;
mutex_unlock(&cp->mutex);
if (ring_obj) {
r = radeon_bo_reserve(ring_obj, false);
@ -422,7 +424,6 @@ void radeon_ring_fini(struct radeon_device *rdev)
}
}
/*
* Debugfs info
*/

4
drivers/gpu/drm/radeon/radeon_semaphore.c
View File

@ -121,13 +121,13 @@ int radeon_semaphore_create(struct radeon_device *rdev,
void radeon_semaphore_emit_signal(struct radeon_device *rdev, int ring,
struct radeon_semaphore *semaphore)
{
radeon_semaphore_ring_emit(rdev, semaphore, ring, false);
radeon_semaphore_ring_emit(rdev, &rdev->cp, semaphore, false);
}
void radeon_semaphore_emit_wait(struct radeon_device *rdev, int ring,
struct radeon_semaphore *semaphore)
{
radeon_semaphore_ring_emit(rdev, semaphore, ring, true);
radeon_semaphore_ring_emit(rdev, &rdev->cp, semaphore, true);
}
void radeon_semaphore_free(struct radeon_device *rdev,

3
drivers/gpu/drm/radeon/radeon_test.c
View File

@ -42,7 +42,8 @@ void radeon_test_moves(struct radeon_device *rdev)
/* Number of tests =
* (Total GTT - IB pool - writeback page - ring buffers) / test size
*/
n = rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024 - rdev->cp.ring_size;
n = rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024;
n -= rdev->cp.ring_size;
if (rdev->wb.wb_obj)
n -= RADEON_GPU_PAGE_SIZE;
if (rdev->ih.ring_obj)

81
drivers/gpu/drm/radeon/rv515.c
View File

@ -55,44 +55,45 @@ void rv515_debugfs(struct radeon_device *rdev)
void rv515_ring_start(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
r = radeon_ring_lock(rdev, 64);
r = radeon_ring_lock(rdev, cp, 64);
if (r) {
return;
}
radeon_ring_write(rdev, PACKET0(ISYNC_CNTL, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(ISYNC_CNTL, 0));
radeon_ring_write(cp,
ISYNC_ANY2D_IDLE3D |
ISYNC_ANY3D_IDLE2D |
ISYNC_WAIT_IDLEGUI |
ISYNC_CPSCRATCH_IDLEGUI);
radeon_ring_write(rdev, PACKET0(WAIT_UNTIL, 0));
radeon_ring_write(rdev, WAIT_2D_IDLECLEAN | WAIT_3D_IDLECLEAN);
radeon_ring_write(rdev, PACKET0(R300_DST_PIPE_CONFIG, 0));
radeon_ring_write(rdev, R300_PIPE_AUTO_CONFIG);
radeon_ring_write(rdev, PACKET0(GB_SELECT, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(GB_ENABLE, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(R500_SU_REG_DEST, 0));
radeon_ring_write(rdev, (1 << rdev->num_gb_pipes) - 1);
radeon_ring_write(rdev, PACKET0(VAP_INDEX_OFFSET, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, RB3D_DC_FLUSH | RB3D_DC_FREE);
radeon_ring_write(rdev, PACKET0(ZB_ZCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, ZC_FLUSH | ZC_FREE);
radeon_ring_write(rdev, PACKET0(WAIT_UNTIL, 0));
radeon_ring_write(rdev, WAIT_2D_IDLECLEAN | WAIT_3D_IDLECLEAN);
radeon_ring_write(rdev, PACKET0(GB_AA_CONFIG, 0));
radeon_ring_write(rdev, 0);
radeon_ring_write(rdev, PACKET0(RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, RB3D_DC_FLUSH | RB3D_DC_FREE);
radeon_ring_write(rdev, PACKET0(ZB_ZCACHE_CTLSTAT, 0));
radeon_ring_write(rdev, ZC_FLUSH | ZC_FREE);
radeon_ring_write(rdev, PACKET0(GB_MSPOS0, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(WAIT_UNTIL, 0));
radeon_ring_write(cp, WAIT_2D_IDLECLEAN | WAIT_3D_IDLECLEAN);
radeon_ring_write(cp, PACKET0(R300_DST_PIPE_CONFIG, 0));
radeon_ring_write(cp, R300_PIPE_AUTO_CONFIG);
radeon_ring_write(cp, PACKET0(GB_SELECT, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(GB_ENABLE, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(R500_SU_REG_DEST, 0));
radeon_ring_write(cp, (1 << rdev->num_gb_pipes) - 1);
radeon_ring_write(cp, PACKET0(VAP_INDEX_OFFSET, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, RB3D_DC_FLUSH | RB3D_DC_FREE);
radeon_ring_write(cp, PACKET0(ZB_ZCACHE_CTLSTAT, 0));
radeon_ring_write(cp, ZC_FLUSH | ZC_FREE);
radeon_ring_write(cp, PACKET0(WAIT_UNTIL, 0));
radeon_ring_write(cp, WAIT_2D_IDLECLEAN | WAIT_3D_IDLECLEAN);
radeon_ring_write(cp, PACKET0(GB_AA_CONFIG, 0));
radeon_ring_write(cp, 0);
radeon_ring_write(cp, PACKET0(RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(cp, RB3D_DC_FLUSH | RB3D_DC_FREE);
radeon_ring_write(cp, PACKET0(ZB_ZCACHE_CTLSTAT, 0));
radeon_ring_write(cp, ZC_FLUSH | ZC_FREE);
radeon_ring_write(cp, PACKET0(GB_MSPOS0, 0));
radeon_ring_write(cp,
((6 << MS_X0_SHIFT) |
(6 << MS_Y0_SHIFT) |
(6 << MS_X1_SHIFT) |
@ -101,8 +102,8 @@ void rv515_ring_start(struct radeon_device *rdev)
(6 << MS_Y2_SHIFT) |
(6 << MSBD0_Y_SHIFT) |
(6 << MSBD0_X_SHIFT)));
radeon_ring_write(rdev, PACKET0(GB_MSPOS1, 0));
radeon_ring_write(rdev,
radeon_ring_write(cp, PACKET0(GB_MSPOS1, 0));
radeon_ring_write(cp,
((6 << MS_X3_SHIFT) |
(6 << MS_Y3_SHIFT) |
(6 << MS_X4_SHIFT) |
@ -110,15 +111,15 @@ void rv515_ring_start(struct radeon_device *rdev)
(6 << MS_X5_SHIFT) |
(6 << MS_Y5_SHIFT) |
(6 << MSBD1_SHIFT)));
radeon_ring_write(rdev, PACKET0(GA_ENHANCE, 0));
radeon_ring_write(rdev, GA_DEADLOCK_CNTL | GA_FASTSYNC_CNTL);
radeon_ring_write(rdev, PACKET0(GA_POLY_MODE, 0));
radeon_ring_write(rdev, FRONT_PTYPE_TRIANGE | BACK_PTYPE_TRIANGE);
radeon_ring_write(rdev, PACKET0(GA_ROUND_MODE, 0));
radeon_ring_write(rdev, GEOMETRY_ROUND_NEAREST | COLOR_ROUND_NEAREST);
radeon_ring_write(rdev, PACKET0(0x20C8, 0));
radeon_ring_write(rdev, 0);
radeon_ring_unlock_commit(rdev);
radeon_ring_write(cp, PACKET0(GA_ENHANCE, 0));
radeon_ring_write(cp, GA_DEADLOCK_CNTL | GA_FASTSYNC_CNTL);
radeon_ring_write(cp, PACKET0(GA_POLY_MODE, 0));
radeon_ring_write(cp, FRONT_PTYPE_TRIANGE | BACK_PTYPE_TRIANGE);
radeon_ring_write(cp, PACKET0(GA_ROUND_MODE, 0));
radeon_ring_write(cp, GEOMETRY_ROUND_NEAREST | COLOR_ROUND_NEAREST);
radeon_ring_write(cp, PACKET0(0x20C8, 0));
radeon_ring_write(cp, 0);
radeon_ring_unlock_commit(rdev, cp);
}
int rv515_mc_wait_for_idle(struct radeon_device *rdev)

11
drivers/gpu/drm/radeon/rv770.c
View File

@ -357,7 +357,7 @@ static int rv770_cp_load_microcode(struct radeon_device *rdev)
void r700_cp_fini(struct radeon_device *rdev)
{
r700_cp_stop(rdev);
radeon_ring_fini(rdev);
radeon_ring_fini(rdev, &rdev->cp);
}
/*
@ -1043,6 +1043,7 @@ int rv770_mc_init(struct radeon_device *rdev)
static int rv770_startup(struct radeon_device *rdev)
{
struct radeon_cp *cp = &rdev->cp;
int r;
/* enable pcie gen2 link */
@ -1091,7 +1092,7 @@ static int rv770_startup(struct radeon_device *rdev)
}
r600_irq_set(rdev);
r = radeon_ring_init(rdev, rdev->cp.ring_size);
r = radeon_ring_init(rdev, cp, cp->ring_size);
if (r)
return r;
r = rv770_cp_load_microcode(rdev);
@ -1121,7 +1122,7 @@ int rv770_resume(struct radeon_device *rdev)
return r;
}
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("radeon: failed testing IB (%d).\n", r);
return r;
@ -1216,7 +1217,7 @@ int rv770_init(struct radeon_device *rdev)
return r;
rdev->cp.ring_obj = NULL;
r600_ring_init(rdev, 1024 * 1024);
r600_ring_init(rdev, &rdev->cp, 1024 * 1024);
rdev->ih.ring_obj = NULL;
r600_ih_ring_init(rdev, 64 * 1024);
@ -1242,7 +1243,7 @@ int rv770_init(struct radeon_device *rdev)
dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
rdev->accel_working = false;
} else {
r = r600_ib_test(rdev);
r = r600_ib_test(rdev, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
dev_err(rdev->dev, "IB test failed (%d).\n", r);
rdev->accel_working = false;