diff options
Diffstat (limited to 'drivers/gpu/drm/radeon/cik_sdma.c')
| -rw-r--r-- | drivers/gpu/drm/radeon/cik_sdma.c | 327 |
1 files changed, 206 insertions, 121 deletions
diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c index 8e9d0f1d858..d748963af08 100644 --- a/drivers/gpu/drm/radeon/cik_sdma.c +++ b/drivers/gpu/drm/radeon/cik_sdma.c @@ -24,6 +24,7 @@ #include <linux/firmware.h> #include <drm/drmP.h> #include "radeon.h" +#include "radeon_ucode.h" #include "radeon_asic.h" #include "radeon_trace.h" #include "cikd.h" @@ -118,6 +119,7 @@ void cik_sdma_set_wptr(struct radeon_device *rdev, reg = SDMA0_GFX_RB_WPTR + SDMA1_REGISTER_OFFSET; WREG32(reg, (ring->wptr << 2) & 0x3fffc); + (void)RREG32(reg); } /** @@ -419,7 +421,6 @@ static int cik_sdma_rlc_resume(struct radeon_device *rdev) */ static int cik_sdma_load_microcode(struct radeon_device *rdev) { - const __be32 *fw_data; int i; if (!rdev->sdma_fw) @@ -428,19 +429,48 @@ static int cik_sdma_load_microcode(struct radeon_device *rdev) /* halt the MEs */ cik_sdma_enable(rdev, false); - /* sdma0 */ - fw_data = (const __be32 *)rdev->sdma_fw->data; - WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); - for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) - WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, be32_to_cpup(fw_data++)); - WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); - - /* sdma1 */ - fw_data = (const __be32 *)rdev->sdma_fw->data; - WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); - for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) - WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, be32_to_cpup(fw_data++)); - WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + if (rdev->new_fw) { + const struct sdma_firmware_header_v1_0 *hdr = + (const struct sdma_firmware_header_v1_0 *)rdev->sdma_fw->data; + const __le32 *fw_data; + u32 fw_size; + + radeon_ucode_print_sdma_hdr(&hdr->header); + + /* sdma0 */ + fw_data = (const __le32 *) + (rdev->sdma_fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; + WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); + for (i = 0; i < fw_size; i++) + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, le32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + + /* sdma1 */ + fw_data = (const __le32 *) + (rdev->sdma_fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; + WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); + for (i = 0; i < fw_size; i++) + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, le32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + } else { + const __be32 *fw_data; + + /* sdma0 */ + fw_data = (const __be32 *)rdev->sdma_fw->data; + WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); + for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, be32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + + /* sdma1 */ + fw_data = (const __be32 *)rdev->sdma_fw->data; + WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); + for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, be32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + } WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); @@ -459,13 +489,6 @@ int cik_sdma_resume(struct radeon_device *rdev) { int r; - /* Reset dma */ - WREG32(SRBM_SOFT_RESET, SOFT_RESET_SDMA | SOFT_RESET_SDMA1); - RREG32(SRBM_SOFT_RESET); - udelay(50); - WREG32(SRBM_SOFT_RESET, 0); - RREG32(SRBM_SOFT_RESET); - r = cik_sdma_load_microcode(rdev); if (r) return r; @@ -507,18 +530,19 @@ void cik_sdma_fini(struct radeon_device *rdev) * @src_offset: src GPU address * @dst_offset: dst GPU address * @num_gpu_pages: number of GPU pages to xfer - * @fence: radeon fence object + * @resv: reservation object to sync to * * Copy GPU paging using the DMA engine (CIK). * Used by the radeon ttm implementation to move pages if * registered as the asic copy callback. */ -int cik_copy_dma(struct radeon_device *rdev, - uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, - struct radeon_fence **fence) +struct radeon_fence *cik_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct reservation_object *resv) { struct radeon_semaphore *sem = NULL; + struct radeon_fence *fence; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes; @@ -528,7 +552,7 @@ int cik_copy_dma(struct radeon_device *rdev, r = radeon_semaphore_create(rdev, &sem); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - return r; + return ERR_PTR(r); } size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); @@ -537,10 +561,10 @@ int cik_copy_dma(struct radeon_device *rdev, if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); radeon_semaphore_free(rdev, &sem, NULL); - return r; + return ERR_PTR(r); } - radeon_semaphore_sync_to(sem, *fence); + radeon_semaphore_sync_resv(rdev, sem, resv, false); radeon_semaphore_sync_rings(rdev, sem, ring->idx); for (i = 0; i < num_loops; i++) { @@ -559,17 +583,17 @@ int cik_copy_dma(struct radeon_device *rdev, dst_offset += cur_size_in_bytes; } - r = radeon_fence_emit(rdev, fence, ring->idx); + r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); radeon_semaphore_free(rdev, &sem, NULL); - return r; + return ERR_PTR(r); } - radeon_ring_unlock_commit(rdev, ring); - radeon_semaphore_free(rdev, &sem, *fence); + radeon_ring_unlock_commit(rdev, ring, false); + radeon_semaphore_free(rdev, &sem, fence); - return r; + return fence; } /** @@ -587,16 +611,19 @@ int cik_sdma_ring_test(struct radeon_device *rdev, { unsigned i; int r; - void __iomem *ptr = (void *)rdev->vram_scratch.ptr; + unsigned index; u32 tmp; + u64 gpu_addr; - if (!ptr) { - DRM_ERROR("invalid vram scratch pointer\n"); - return -EINVAL; - } + if (ring->idx == R600_RING_TYPE_DMA_INDEX) + index = R600_WB_DMA_RING_TEST_OFFSET; + else + index = CAYMAN_WB_DMA1_RING_TEST_OFFSET; + + gpu_addr = rdev->wb.gpu_addr + index; tmp = 0xCAFEDEAD; - writel(tmp, ptr); + rdev->wb.wb[index/4] = cpu_to_le32(tmp); r = radeon_ring_lock(rdev, ring, 5); if (r) { @@ -604,14 +631,14 @@ int cik_sdma_ring_test(struct radeon_device *rdev, return r; } radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); - radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr)); + radeon_ring_write(ring, lower_32_bits(gpu_addr)); + radeon_ring_write(ring, upper_32_bits(gpu_addr)); radeon_ring_write(ring, 1); /* number of DWs to follow */ radeon_ring_write(ring, 0xDEADBEEF); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_unlock_commit(rdev, ring, false); for (i = 0; i < rdev->usec_timeout; i++) { - tmp = readl(ptr); + tmp = le32_to_cpu(rdev->wb.wb[index/4]); if (tmp == 0xDEADBEEF) break; DRM_UDELAY(1); @@ -640,17 +667,20 @@ int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) { struct radeon_ib ib; unsigned i; + unsigned index; int r; - void __iomem *ptr = (void *)rdev->vram_scratch.ptr; u32 tmp = 0; + u64 gpu_addr; - if (!ptr) { - DRM_ERROR("invalid vram scratch pointer\n"); - return -EINVAL; - } + if (ring->idx == R600_RING_TYPE_DMA_INDEX) + index = R600_WB_DMA_RING_TEST_OFFSET; + else + index = CAYMAN_WB_DMA1_RING_TEST_OFFSET; + + gpu_addr = rdev->wb.gpu_addr + index; tmp = 0xCAFEDEAD; - writel(tmp, ptr); + rdev->wb.wb[index/4] = cpu_to_le32(tmp); r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256); if (r) { @@ -659,13 +689,13 @@ int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) } ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib.ptr[1] = rdev->vram_scratch.gpu_addr & 0xfffffffc; - ib.ptr[2] = upper_32_bits(rdev->vram_scratch.gpu_addr); + ib.ptr[1] = lower_32_bits(gpu_addr); + ib.ptr[2] = upper_32_bits(gpu_addr); ib.ptr[3] = 1; ib.ptr[4] = 0xDEADBEEF; ib.length_dw = 5; - r = radeon_ib_schedule(rdev, &ib, NULL); + r = radeon_ib_schedule(rdev, &ib, NULL, false); if (r) { radeon_ib_free(rdev, &ib); DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); @@ -677,7 +707,7 @@ int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) return r; } for (i = 0; i < rdev->usec_timeout; i++) { - tmp = readl(ptr); + tmp = le32_to_cpu(rdev->wb.wb[index/4]); if (tmp == 0xDEADBEEF) break; DRM_UDELAY(1); @@ -719,7 +749,43 @@ bool cik_sdma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) } /** - * cik_sdma_vm_set_page - update the page tables using sDMA + * cik_sdma_vm_copy_pages - update PTEs by copying them from the GART + * + * @rdev: radeon_device pointer + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @src: src addr to copy from + * @count: number of page entries to update + * + * Update PTEs by copying them from the GART using sDMA (CIK). + */ +void cik_sdma_vm_copy_pages(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, uint64_t src, + unsigned count) +{ + while (count) { + unsigned bytes = count * 8; + if (bytes > 0x1FFFF8) + bytes = 0x1FFFF8; + + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, + SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = bytes; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + + pe += bytes; + src += bytes; + count -= bytes / 8; + } +} + +/** + * cik_sdma_vm_write_pages - update PTEs by writing them manually * * @rdev: radeon_device pointer * @ib: indirect buffer to fill with commands @@ -729,84 +795,103 @@ bool cik_sdma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) * @incr: increase next addr by incr bytes * @flags: access flags * - * Update the page tables using sDMA (CIK). + * Update PTEs by writing them manually using sDMA (CIK). */ -void cik_sdma_vm_set_page(struct radeon_device *rdev, - struct radeon_ib *ib, - uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags) +void cik_sdma_vm_write_pages(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags) { uint64_t value; unsigned ndw; - trace_radeon_vm_set_page(pe, addr, count, incr, flags); - - if (flags == R600_PTE_GART) { - uint64_t src = rdev->gart.table_addr + (addr >> 12) * 8; - while (count) { - unsigned bytes = count * 8; - if (bytes > 0x1FFFF8) - bytes = 0x1FFFF8; - - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib->ptr[ib->length_dw++] = bytes; - ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ - ib->ptr[ib->length_dw++] = lower_32_bits(src); - ib->ptr[ib->length_dw++] = upper_32_bits(src); - ib->ptr[ib->length_dw++] = lower_32_bits(pe); - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - - pe += bytes; - src += bytes; - count -= bytes / 8; - } - } else if (flags & R600_PTE_SYSTEM) { - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { + while (count) { + ndw = count * 2; + if (ndw > 0xFFFFE) + ndw = 0xFFFFE; + + /* for non-physically contiguous pages (system) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, + SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + if (flags & R600_PTE_SYSTEM) { value = radeon_vm_map_gart(rdev, addr); value &= 0xFFFFFFFFFFFFF000ULL; - addr += incr; - value |= flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } - } - } else { - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & R600_PTE_VALID) + } else if (flags & R600_PTE_VALID) { value = addr; - else + } else { value = 0; - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ + } + addr += incr; + value |= flags; + ib->ptr[ib->length_dw++] = value; ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; } } +} + +/** + * cik_sdma_vm_set_pages - update the page tables using sDMA + * + * @rdev: radeon_device pointer + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: access flags + * + * Update the page tables using sDMA (CIK). + */ +void cik_sdma_vm_set_pages(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags) +{ + uint64_t value; + unsigned ndw; + + while (count) { + ndw = count; + if (ndw > 0x7FFFF) + ndw = 0x7FFFF; + + if (flags & R600_PTE_VALID) + value = addr; + else + value = 0; + + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); + ib->ptr[ib->length_dw++] = pe; /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = value; /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(value); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = ndw; /* number of entries */ + + pe += ndw * 8; + addr += ndw * incr; + count -= ndw; + } +} + +/** + * cik_sdma_vm_pad_ib - pad the IB to the required number of dw + * + * @ib: indirect buffer to fill with padding + * + */ +void cik_sdma_vm_pad_ib(struct radeon_ib *ib) +{ while (ib->length_dw & 0x7) ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0); } |