X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Fgpu%2Fdrm%2Fradeon%2Fr100.c;h=fa0fdc1e3457b832011e8e7fb5485460e8ee7aa4;hb=4aac047323e3082d0866b8ad3784236632105af4;hp=1b23106f95525a83d9fb5b01049f3036088a3ddf;hpb=ed10f95d60d41033d356fdcf88c240d7065bd5b4;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index 1b23106..fa0fdc1 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -29,15 +29,41 @@ #include "drmP.h" #include "drm.h" #include "radeon_drm.h" -#include "radeon_microcode.h" #include "radeon_reg.h" #include "radeon.h" +#include "r100d.h" + +#include +#include + +#include "r100_reg_safe.h" +#include "rn50_reg_safe.h" + +/* Firmware Names */ +#define FIRMWARE_R100 "radeon/R100_cp.bin" +#define FIRMWARE_R200 "radeon/R200_cp.bin" +#define FIRMWARE_R300 "radeon/R300_cp.bin" +#define FIRMWARE_R420 "radeon/R420_cp.bin" +#define FIRMWARE_RS690 "radeon/RS690_cp.bin" +#define FIRMWARE_RS600 "radeon/RS600_cp.bin" +#define FIRMWARE_R520 "radeon/R520_cp.bin" + +MODULE_FIRMWARE(FIRMWARE_R100); +MODULE_FIRMWARE(FIRMWARE_R200); +MODULE_FIRMWARE(FIRMWARE_R300); +MODULE_FIRMWARE(FIRMWARE_R420); +MODULE_FIRMWARE(FIRMWARE_RS690); +MODULE_FIRMWARE(FIRMWARE_RS600); +MODULE_FIRMWARE(FIRMWARE_R520); + +#include "r100_track.h" /* This files gather functions specifics to: * r100,rv100,rs100,rv200,rs200,r200,rv250,rs300,rv280 * * Some of these functions might be used by newer ASICs. */ +int r200_init(struct radeon_device *rdev); void r100_hdp_reset(struct radeon_device *rdev); void r100_gpu_init(struct radeon_device *rdev); int r100_gui_wait_for_idle(struct radeon_device *rdev); @@ -58,23 +84,28 @@ void r100_pci_gart_tlb_flush(struct radeon_device *rdev) * could end up in wrong address. */ } -int r100_pci_gart_enable(struct radeon_device *rdev) +int r100_pci_gart_init(struct radeon_device *rdev) { - uint32_t tmp; int r; + if (rdev->gart.table.ram.ptr) { + WARN(1, "R100 PCI GART already initialized.\n"); + return 0; + } /* Initialize common gart structure */ r = radeon_gart_init(rdev); - if (r) { + if (r) return r; - } - if (rdev->gart.table.ram.ptr == NULL) { - rdev->gart.table_size = rdev->gart.num_gpu_pages * 4; - r = radeon_gart_table_ram_alloc(rdev); - if (r) { - return r; - } - } + rdev->gart.table_size = rdev->gart.num_gpu_pages * 4; + rdev->asic->gart_tlb_flush = &r100_pci_gart_tlb_flush; + rdev->asic->gart_set_page = &r100_pci_gart_set_page; + return radeon_gart_table_ram_alloc(rdev); +} + +int r100_pci_gart_enable(struct radeon_device *rdev) +{ + uint32_t tmp; + /* discard memory request outside of configured range */ tmp = RREG32(RADEON_AIC_CNTL) | RADEON_DIS_OUT_OF_PCI_GART_ACCESS; WREG32(RADEON_AIC_CNTL, tmp); @@ -114,13 +145,11 @@ int r100_pci_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) return 0; } -int r100_gart_enable(struct radeon_device *rdev) +void r100_pci_gart_fini(struct radeon_device *rdev) { - if (rdev->flags & RADEON_IS_AGP) { - r100_pci_gart_disable(rdev); - return 0; - } - return r100_pci_gart_enable(rdev); + r100_pci_gart_disable(rdev); + radeon_gart_table_ram_free(rdev); + radeon_gart_fini(rdev); } @@ -173,8 +202,12 @@ void r100_mc_setup(struct radeon_device *rdev) DRM_ERROR("Failed to register debugfs file for R100 MC !\n"); } /* Write VRAM size in case we are limiting it */ - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); + /* Novell bug 204882 for RN50/M6/M7 with 8/16/32MB VRAM, + * if the aperture is 64MB but we have 32MB VRAM + * we report only 32MB VRAM but we have to set MC_FB_LOCATION + * to 64MB, otherwise the gpu accidentially dies */ + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(RADEON_MC_FB_TOP, tmp >> 16); tmp |= REG_SET(RADEON_MC_FB_START, rdev->mc.vram_location >> 16); WREG32(RADEON_MC_FB_LOCATION, tmp); @@ -215,7 +248,6 @@ int r100_mc_init(struct radeon_device *rdev) r100_pci_gart_disable(rdev); /* Setup GPU memory space */ - rdev->mc.vram_location = 0xFFFFFFFFUL; rdev->mc.gtt_location = 0xFFFFFFFFUL; if (rdev->flags & RADEON_IS_AGP) { r = radeon_agp_init(rdev); @@ -244,9 +276,86 @@ int r100_mc_init(struct radeon_device *rdev) void r100_mc_fini(struct radeon_device *rdev) { - r100_pci_gart_disable(rdev); - radeon_gart_table_ram_free(rdev); - radeon_gart_fini(rdev); +} + + +/* + * Interrupts + */ +int r100_irq_set(struct radeon_device *rdev) +{ + uint32_t tmp = 0; + + if (rdev->irq.sw_int) { + tmp |= RADEON_SW_INT_ENABLE; + } + if (rdev->irq.crtc_vblank_int[0]) { + tmp |= RADEON_CRTC_VBLANK_MASK; + } + if (rdev->irq.crtc_vblank_int[1]) { + tmp |= RADEON_CRTC2_VBLANK_MASK; + } + WREG32(RADEON_GEN_INT_CNTL, tmp); + return 0; +} + +void r100_irq_disable(struct radeon_device *rdev) +{ + u32 tmp; + + WREG32(R_000040_GEN_INT_CNTL, 0); + /* Wait and acknowledge irq */ + mdelay(1); + tmp = RREG32(R_000044_GEN_INT_STATUS); + WREG32(R_000044_GEN_INT_STATUS, tmp); +} + +static inline uint32_t r100_irq_ack(struct radeon_device *rdev) +{ + uint32_t irqs = RREG32(RADEON_GEN_INT_STATUS); + uint32_t irq_mask = RADEON_SW_INT_TEST | RADEON_CRTC_VBLANK_STAT | + RADEON_CRTC2_VBLANK_STAT; + + if (irqs) { + WREG32(RADEON_GEN_INT_STATUS, irqs); + } + return irqs & irq_mask; +} + +int r100_irq_process(struct radeon_device *rdev) +{ + uint32_t status; + + status = r100_irq_ack(rdev); + if (!status) { + return IRQ_NONE; + } + if (rdev->shutdown) { + return IRQ_NONE; + } + while (status) { + /* SW interrupt */ + if (status & RADEON_SW_INT_TEST) { + radeon_fence_process(rdev); + } + /* Vertical blank interrupts */ + if (status & RADEON_CRTC_VBLANK_STAT) { + drm_handle_vblank(rdev->ddev, 0); + } + if (status & RADEON_CRTC2_VBLANK_STAT) { + drm_handle_vblank(rdev->ddev, 1); + } + status = r100_irq_ack(rdev); + } + return IRQ_HANDLED; +} + +u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc) +{ + if (crtc == 0) + return RREG32(RADEON_CRTC_CRNT_FRAME); + else + return RREG32(RADEON_CRTC2_CRNT_FRAME); } @@ -298,14 +407,21 @@ int r100_wb_init(struct radeon_device *rdev) return r; } } - WREG32(0x774, rdev->wb.gpu_addr); - WREG32(0x70C, rdev->wb.gpu_addr + 1024); - WREG32(0x770, 0xff); + WREG32(R_000774_SCRATCH_ADDR, rdev->wb.gpu_addr); + WREG32(R_00070C_CP_RB_RPTR_ADDR, + S_00070C_RB_RPTR_ADDR((rdev->wb.gpu_addr + 1024) >> 2)); + WREG32(R_000770_SCRATCH_UMSK, 0xff); return 0; } +void r100_wb_disable(struct radeon_device *rdev) +{ + WREG32(R_000770_SCRATCH_UMSK, 0); +} + void r100_wb_fini(struct radeon_device *rdev) { + r100_wb_disable(rdev); if (rdev->wb.wb_obj) { radeon_object_kunmap(rdev->wb.wb_obj); radeon_object_unpin(rdev->wb.wb_obj); @@ -392,6 +508,21 @@ int r100_copy_blit(struct radeon_device *rdev, /* * CP */ +static int r100_cp_wait_for_idle(struct radeon_device *rdev) +{ + unsigned i; + u32 tmp; + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(R_000E40_RBBM_STATUS); + if (!G_000E40_CP_CMDSTRM_BUSY(tmp)) { + return 0; + } + udelay(1); + } + return -1; +} + void r100_ring_start(struct radeon_device *rdev) { int r; @@ -409,33 +540,33 @@ void r100_ring_start(struct radeon_device *rdev) radeon_ring_unlock_commit(rdev); } -static void r100_cp_load_microcode(struct radeon_device *rdev) + +/* Load the microcode for the CP */ +static int r100_cp_init_microcode(struct radeon_device *rdev) { - int i; + struct platform_device *pdev; + const char *fw_name = NULL; + int err; - if (r100_gui_wait_for_idle(rdev)) { - printk(KERN_WARNING "Failed to wait GUI idle while " - "programming pipes. Bad things might happen.\n"); - } + DRM_DEBUG("\n"); - WREG32(RADEON_CP_ME_RAM_ADDR, 0); + pdev = platform_device_register_simple("radeon_cp", 0, NULL, 0); + err = IS_ERR(pdev); + if (err) { + printk(KERN_ERR "radeon_cp: Failed to register firmware\n"); + return -EINVAL; + } if ((rdev->family == CHIP_R100) || (rdev->family == CHIP_RV100) || (rdev->family == CHIP_RV200) || (rdev->family == CHIP_RS100) || (rdev->family == CHIP_RS200)) { DRM_INFO("Loading R100 Microcode\n"); - for (i = 0; i < 256; i++) { - WREG32(RADEON_CP_ME_RAM_DATAH, R100_cp_microcode[i][1]); - WREG32(RADEON_CP_ME_RAM_DATAL, R100_cp_microcode[i][0]); - } + fw_name = FIRMWARE_R100; } else if ((rdev->family == CHIP_R200) || (rdev->family == CHIP_RV250) || (rdev->family == CHIP_RV280) || (rdev->family == CHIP_RS300)) { DRM_INFO("Loading R200 Microcode\n"); - for (i = 0; i < 256; i++) { - WREG32(RADEON_CP_ME_RAM_DATAH, R200_cp_microcode[i][1]); - WREG32(RADEON_CP_ME_RAM_DATAL, R200_cp_microcode[i][0]); - } + fw_name = FIRMWARE_R200; } else if ((rdev->family == CHIP_R300) || (rdev->family == CHIP_R350) || (rdev->family == CHIP_RV350) || @@ -443,31 +574,19 @@ static void r100_cp_load_microcode(struct radeon_device *rdev) (rdev->family == CHIP_RS400) || (rdev->family == CHIP_RS480)) { DRM_INFO("Loading R300 Microcode\n"); - for (i = 0; i < 256; i++) { - WREG32(RADEON_CP_ME_RAM_DATAH, R300_cp_microcode[i][1]); - WREG32(RADEON_CP_ME_RAM_DATAL, R300_cp_microcode[i][0]); - } + fw_name = FIRMWARE_R300; } else if ((rdev->family == CHIP_R420) || (rdev->family == CHIP_R423) || (rdev->family == CHIP_RV410)) { DRM_INFO("Loading R400 Microcode\n"); - for (i = 0; i < 256; i++) { - WREG32(RADEON_CP_ME_RAM_DATAH, R420_cp_microcode[i][1]); - WREG32(RADEON_CP_ME_RAM_DATAL, R420_cp_microcode[i][0]); - } + fw_name = FIRMWARE_R420; } else if ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740)) { DRM_INFO("Loading RS690/RS740 Microcode\n"); - for (i = 0; i < 256; i++) { - WREG32(RADEON_CP_ME_RAM_DATAH, RS690_cp_microcode[i][1]); - WREG32(RADEON_CP_ME_RAM_DATAL, RS690_cp_microcode[i][0]); - } + fw_name = FIRMWARE_RS690; } else if (rdev->family == CHIP_RS600) { DRM_INFO("Loading RS600 Microcode\n"); - for (i = 0; i < 256; i++) { - WREG32(RADEON_CP_ME_RAM_DATAH, RS600_cp_microcode[i][1]); - WREG32(RADEON_CP_ME_RAM_DATAL, RS600_cp_microcode[i][0]); - } + fw_name = FIRMWARE_RS600; } else if ((rdev->family == CHIP_RV515) || (rdev->family == CHIP_R520) || (rdev->family == CHIP_RV530) || @@ -475,9 +594,43 @@ static void r100_cp_load_microcode(struct radeon_device *rdev) (rdev->family == CHIP_RV560) || (rdev->family == CHIP_RV570)) { DRM_INFO("Loading R500 Microcode\n"); - for (i = 0; i < 256; i++) { - WREG32(RADEON_CP_ME_RAM_DATAH, R520_cp_microcode[i][1]); - WREG32(RADEON_CP_ME_RAM_DATAL, R520_cp_microcode[i][0]); + fw_name = FIRMWARE_R520; + } + + err = request_firmware(&rdev->me_fw, fw_name, &pdev->dev); + platform_device_unregister(pdev); + if (err) { + printk(KERN_ERR "radeon_cp: Failed to load firmware \"%s\"\n", + fw_name); + } else if (rdev->me_fw->size % 8) { + printk(KERN_ERR + "radeon_cp: Bogus length %zu in firmware \"%s\"\n", + rdev->me_fw->size, fw_name); + err = -EINVAL; + release_firmware(rdev->me_fw); + rdev->me_fw = NULL; + } + return err; +} +static void r100_cp_load_microcode(struct radeon_device *rdev) +{ + const __be32 *fw_data; + int i, size; + + if (r100_gui_wait_for_idle(rdev)) { + printk(KERN_WARNING "Failed to wait GUI idle while " + "programming pipes. Bad things might happen.\n"); + } + + if (rdev->me_fw) { + size = rdev->me_fw->size / 4; + fw_data = (const __be32 *)&rdev->me_fw->data[0]; + WREG32(RADEON_CP_ME_RAM_ADDR, 0); + for (i = 0; i < size; i += 2) { + WREG32(RADEON_CP_ME_RAM_DATAH, + be32_to_cpup(&fw_data[i])); + WREG32(RADEON_CP_ME_RAM_DATAL, + be32_to_cpup(&fw_data[i + 1])); } } } @@ -516,6 +669,15 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size) } else { DRM_INFO("radeon: cp idle (0x%08X)\n", tmp); } + + if (!rdev->me_fw) { + r = r100_cp_init_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load firmware!\n"); + return r; + } + } + /* Align ring size */ rb_bufsz = drm_order(ring_size / 8); ring_size = (1 << (rb_bufsz + 1)) * 4; @@ -589,9 +751,11 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size) void r100_cp_fini(struct radeon_device *rdev) { + if (r100_cp_wait_for_idle(rdev)) { + DRM_ERROR("Wait for CP idle timeout, shutting down CP.\n"); + } /* Disable ring */ - rdev->cp.ready = false; - WREG32(RADEON_CP_CSQ_CNTL, 0); + r100_cp_disable(rdev); radeon_ring_fini(rdev); DRM_INFO("radeon: cp finalized\n"); } @@ -641,6 +805,12 @@ int r100_cp_reset(struct radeon_device *rdev) return -1; } +void r100_cp_commit(struct radeon_device *rdev) +{ + WREG32(RADEON_CP_RB_WPTR, rdev->cp.wptr); + (void)RREG32(RADEON_CP_RB_WPTR); +} + /* * CS functions @@ -719,13 +889,14 @@ int r100_cs_packet_parse(struct radeon_cs_parser *p, unsigned idx) { struct radeon_cs_chunk *ib_chunk = &p->chunks[p->chunk_ib_idx]; - uint32_t header = ib_chunk->kdata[idx]; + uint32_t header; if (idx >= ib_chunk->length_dw) { DRM_ERROR("Can not parse packet at %d after CS end %d !\n", idx, ib_chunk->length_dw); return -EINVAL; } + header = ib_chunk->kdata[idx]; pkt->idx = idx; pkt->type = CP_PACKET_GET_TYPE(header); pkt->count = CP_PACKET_GET_COUNT(header); @@ -753,6 +924,102 @@ int r100_cs_packet_parse(struct radeon_cs_parser *p, } /** + * r100_cs_packet_next_vline() - parse userspace VLINE packet + * @parser: parser structure holding parsing context. + * + * Userspace sends a special sequence for VLINE waits. + * PACKET0 - VLINE_START_END + value + * PACKET0 - WAIT_UNTIL +_value + * RELOC (P3) - crtc_id in reloc. + * + * This function parses this and relocates the VLINE START END + * and WAIT UNTIL packets to the correct crtc. + * It also detects a switched off crtc and nulls out the + * wait in that case. + */ +int r100_cs_packet_parse_vline(struct radeon_cs_parser *p) +{ + struct radeon_cs_chunk *ib_chunk; + struct drm_mode_object *obj; + struct drm_crtc *crtc; + struct radeon_crtc *radeon_crtc; + struct radeon_cs_packet p3reloc, waitreloc; + int crtc_id; + int r; + uint32_t header, h_idx, reg; + + ib_chunk = &p->chunks[p->chunk_ib_idx]; + + /* parse the wait until */ + r = r100_cs_packet_parse(p, &waitreloc, p->idx); + if (r) + return r; + + /* check its a wait until and only 1 count */ + if (waitreloc.reg != RADEON_WAIT_UNTIL || + waitreloc.count != 0) { + DRM_ERROR("vline wait had illegal wait until segment\n"); + r = -EINVAL; + return r; + } + + if (ib_chunk->kdata[waitreloc.idx + 1] != RADEON_WAIT_CRTC_VLINE) { + DRM_ERROR("vline wait had illegal wait until\n"); + r = -EINVAL; + return r; + } + + /* jump over the NOP */ + r = r100_cs_packet_parse(p, &p3reloc, p->idx); + if (r) + return r; + + h_idx = p->idx - 2; + p->idx += waitreloc.count; + p->idx += p3reloc.count; + + header = ib_chunk->kdata[h_idx]; + crtc_id = ib_chunk->kdata[h_idx + 5]; + reg = ib_chunk->kdata[h_idx] >> 2; + mutex_lock(&p->rdev->ddev->mode_config.mutex); + obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC); + if (!obj) { + DRM_ERROR("cannot find crtc %d\n", crtc_id); + r = -EINVAL; + goto out; + } + crtc = obj_to_crtc(obj); + radeon_crtc = to_radeon_crtc(crtc); + crtc_id = radeon_crtc->crtc_id; + + if (!crtc->enabled) { + /* if the CRTC isn't enabled - we need to nop out the wait until */ + ib_chunk->kdata[h_idx + 2] = PACKET2(0); + ib_chunk->kdata[h_idx + 3] = PACKET2(0); + } else if (crtc_id == 1) { + switch (reg) { + case AVIVO_D1MODE_VLINE_START_END: + header &= R300_CP_PACKET0_REG_MASK; + header |= AVIVO_D2MODE_VLINE_START_END >> 2; + break; + case RADEON_CRTC_GUI_TRIG_VLINE: + header &= R300_CP_PACKET0_REG_MASK; + header |= RADEON_CRTC2_GUI_TRIG_VLINE >> 2; + break; + default: + DRM_ERROR("unknown crtc reloc\n"); + r = -EINVAL; + goto out; + } + ib_chunk->kdata[h_idx] = header; + ib_chunk->kdata[h_idx + 3] |= RADEON_ENG_DISPLAY_SELECT_CRTC1; + } +out: + mutex_unlock(&p->rdev->ddev->mode_config.mutex); + return r; +} + +/** * r100_cs_packet_next_reloc() - parse next packet which should be reloc packet3 * @parser: parser structure holding parsing context. * @data: pointer to relocation data @@ -802,193 +1069,523 @@ int r100_cs_packet_next_reloc(struct radeon_cs_parser *p, return 0; } -static int r100_packet0_check(struct radeon_cs_parser *p, - struct radeon_cs_packet *pkt) -{ - struct radeon_cs_chunk *ib_chunk; - struct radeon_cs_reloc *reloc; - volatile uint32_t *ib; - uint32_t tmp; - unsigned reg; - unsigned i; - unsigned idx; - bool onereg; - int r; - - ib = p->ib->ptr; - ib_chunk = &p->chunks[p->chunk_ib_idx]; - idx = pkt->idx + 1; - reg = pkt->reg; - onereg = false; - if (CP_PACKET0_GET_ONE_REG_WR(ib_chunk->kdata[pkt->idx])) { - onereg = true; - } - for (i = 0; i <= pkt->count; i++, idx++, reg += 4) { - switch (reg) { - /* FIXME: only allow PACKET3 blit? easier to check for out of - * range access */ - case RADEON_DST_PITCH_OFFSET: - case RADEON_SRC_PITCH_OFFSET: - r = r100_cs_packet_next_reloc(p, &reloc); - if (r) { - DRM_ERROR("No reloc for ib[%d]=0x%04X\n", - idx, reg); - r100_cs_dump_packet(p, pkt); - return r; - } - tmp = ib_chunk->kdata[idx] & 0x003fffff; - tmp += (((u32)reloc->lobj.gpu_offset) >> 10); - ib[idx] = (ib_chunk->kdata[idx] & 0xffc00000) | tmp; - break; - case RADEON_RB3D_DEPTHOFFSET: - case RADEON_RB3D_COLOROFFSET: - case R300_RB3D_COLOROFFSET0: - case R300_ZB_DEPTHOFFSET: - case R200_PP_TXOFFSET_0: - case R200_PP_TXOFFSET_1: - case R200_PP_TXOFFSET_2: - case R200_PP_TXOFFSET_3: - case R200_PP_TXOFFSET_4: - case R200_PP_TXOFFSET_5: - case RADEON_PP_TXOFFSET_0: - case RADEON_PP_TXOFFSET_1: - case RADEON_PP_TXOFFSET_2: - case R300_TX_OFFSET_0: - case R300_TX_OFFSET_0+4: - case R300_TX_OFFSET_0+8: - case R300_TX_OFFSET_0+12: - case R300_TX_OFFSET_0+16: - case R300_TX_OFFSET_0+20: - case R300_TX_OFFSET_0+24: - case R300_TX_OFFSET_0+28: - case R300_TX_OFFSET_0+32: - case R300_TX_OFFSET_0+36: - case R300_TX_OFFSET_0+40: - case R300_TX_OFFSET_0+44: - case R300_TX_OFFSET_0+48: - case R300_TX_OFFSET_0+52: - case R300_TX_OFFSET_0+56: - case R300_TX_OFFSET_0+60: - r = r100_cs_packet_next_reloc(p, &reloc); - if (r) { - DRM_ERROR("No reloc for ib[%d]=0x%04X\n", - idx, reg); - r100_cs_dump_packet(p, pkt); - return r; - } - ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); - break; - default: - /* FIXME: we don't want to allow anyothers packet */ - break; - } - if (onereg) { - /* FIXME: forbid onereg write to register on relocate */ - break; - } - } - return 0; -} - -int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p, - struct radeon_cs_packet *pkt, - struct radeon_object *robj) +static int r100_get_vtx_size(uint32_t vtx_fmt) { - struct radeon_cs_chunk *ib_chunk; - unsigned idx; - - ib_chunk = &p->chunks[p->chunk_ib_idx]; - idx = pkt->idx + 1; - if ((ib_chunk->kdata[idx+2] + 1) > radeon_object_size(robj)) { - DRM_ERROR("[drm] Buffer too small for PACKET3 INDX_BUFFER " - "(need %u have %lu) !\n", - ib_chunk->kdata[idx+2] + 1, - radeon_object_size(robj)); - return -EINVAL; - } - return 0; + int vtx_size; + vtx_size = 2; + /* ordered according to bits in spec */ + if (vtx_fmt & RADEON_SE_VTX_FMT_W0) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_FPCOLOR) + vtx_size += 3; + if (vtx_fmt & RADEON_SE_VTX_FMT_FPALPHA) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_PKCOLOR) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_FPSPEC) + vtx_size += 3; + if (vtx_fmt & RADEON_SE_VTX_FMT_FPFOG) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_PKSPEC) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_ST0) + vtx_size += 2; + if (vtx_fmt & RADEON_SE_VTX_FMT_ST1) + vtx_size += 2; + if (vtx_fmt & RADEON_SE_VTX_FMT_Q1) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_ST2) + vtx_size += 2; + if (vtx_fmt & RADEON_SE_VTX_FMT_Q2) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_ST3) + vtx_size += 2; + if (vtx_fmt & RADEON_SE_VTX_FMT_Q3) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_Q0) + vtx_size++; + /* blend weight */ + if (vtx_fmt & (0x7 << 15)) + vtx_size += (vtx_fmt >> 15) & 0x7; + if (vtx_fmt & RADEON_SE_VTX_FMT_N0) + vtx_size += 3; + if (vtx_fmt & RADEON_SE_VTX_FMT_XY1) + vtx_size += 2; + if (vtx_fmt & RADEON_SE_VTX_FMT_Z1) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_W1) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_N1) + vtx_size++; + if (vtx_fmt & RADEON_SE_VTX_FMT_Z) + vtx_size++; + return vtx_size; } -static int r100_packet3_check(struct radeon_cs_parser *p, - struct radeon_cs_packet *pkt) +static int r100_packet0_check(struct radeon_cs_parser *p, + struct radeon_cs_packet *pkt, + unsigned idx, unsigned reg) { struct radeon_cs_chunk *ib_chunk; struct radeon_cs_reloc *reloc; - unsigned idx; - unsigned i, c; + struct r100_cs_track *track; volatile uint32_t *ib; + uint32_t tmp; int r; + int i, face; + u32 tile_flags = 0; ib = p->ib->ptr; ib_chunk = &p->chunks[p->chunk_ib_idx]; - idx = pkt->idx + 1; - switch (pkt->opcode) { - case PACKET3_3D_LOAD_VBPNTR: - c = ib_chunk->kdata[idx++]; - for (i = 0; i < (c - 1); i += 2, idx += 3) { - r = r100_cs_packet_next_reloc(p, &reloc); - if (r) { - DRM_ERROR("No reloc for packet3 %d\n", - pkt->opcode); - r100_cs_dump_packet(p, pkt); - return r; - } - ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); - r = r100_cs_packet_next_reloc(p, &reloc); - if (r) { - DRM_ERROR("No reloc for packet3 %d\n", - pkt->opcode); - r100_cs_dump_packet(p, pkt); - return r; - } - ib[idx+2] = ib_chunk->kdata[idx+2] + ((u32)reloc->lobj.gpu_offset); - } - if (c & 1) { - r = r100_cs_packet_next_reloc(p, &reloc); - if (r) { - DRM_ERROR("No reloc for packet3 %d\n", - pkt->opcode); - r100_cs_dump_packet(p, pkt); - return r; - } - ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); - } - break; - case PACKET3_INDX_BUFFER: - r = r100_cs_packet_next_reloc(p, &reloc); + track = (struct r100_cs_track *)p->track; + + switch (reg) { + case RADEON_CRTC_GUI_TRIG_VLINE: + r = r100_cs_packet_parse_vline(p); if (r) { - DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode); + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); r100_cs_dump_packet(p, pkt); return r; } - ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); - r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj); - if (r) { + break; + /* FIXME: only allow PACKET3 blit? easier to check for out of + * range access */ + case RADEON_DST_PITCH_OFFSET: + case RADEON_SRC_PITCH_OFFSET: + r = r100_reloc_pitch_offset(p, pkt, idx, reg); + if (r) return r; - } break; - case 0x23: - /* FIXME: cleanup */ - /* 3D_RNDR_GEN_INDX_PRIM on r100/r200 */ + case RADEON_RB3D_DEPTHOFFSET: r = r100_cs_packet_next_reloc(p, &reloc); if (r) { - DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode); + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); r100_cs_dump_packet(p, pkt); return r; } + track->zb.robj = reloc->robj; + track->zb.offset = ib_chunk->kdata[idx]; ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); break; - case PACKET3_3D_DRAW_IMMD: + case RADEON_RB3D_COLOROFFSET: + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + track->cb[0].robj = reloc->robj; + track->cb[0].offset = ib_chunk->kdata[idx]; + ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); + break; + case RADEON_PP_TXOFFSET_0: + case RADEON_PP_TXOFFSET_1: + case RADEON_PP_TXOFFSET_2: + i = (reg - RADEON_PP_TXOFFSET_0) / 24; + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); + track->textures[i].robj = reloc->robj; + break; + case RADEON_PP_CUBIC_OFFSET_T0_0: + case RADEON_PP_CUBIC_OFFSET_T0_1: + case RADEON_PP_CUBIC_OFFSET_T0_2: + case RADEON_PP_CUBIC_OFFSET_T0_3: + case RADEON_PP_CUBIC_OFFSET_T0_4: + i = (reg - RADEON_PP_CUBIC_OFFSET_T0_0) / 4; + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + track->textures[0].cube_info[i].offset = ib_chunk->kdata[idx]; + ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); + track->textures[0].cube_info[i].robj = reloc->robj; + break; + case RADEON_PP_CUBIC_OFFSET_T1_0: + case RADEON_PP_CUBIC_OFFSET_T1_1: + case RADEON_PP_CUBIC_OFFSET_T1_2: + case RADEON_PP_CUBIC_OFFSET_T1_3: + case RADEON_PP_CUBIC_OFFSET_T1_4: + i = (reg - RADEON_PP_CUBIC_OFFSET_T1_0) / 4; + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + track->textures[1].cube_info[i].offset = ib_chunk->kdata[idx]; + ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); + track->textures[1].cube_info[i].robj = reloc->robj; + break; + case RADEON_PP_CUBIC_OFFSET_T2_0: + case RADEON_PP_CUBIC_OFFSET_T2_1: + case RADEON_PP_CUBIC_OFFSET_T2_2: + case RADEON_PP_CUBIC_OFFSET_T2_3: + case RADEON_PP_CUBIC_OFFSET_T2_4: + i = (reg - RADEON_PP_CUBIC_OFFSET_T2_0) / 4; + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + track->textures[2].cube_info[i].offset = ib_chunk->kdata[idx]; + ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); + track->textures[2].cube_info[i].robj = reloc->robj; + break; + case RADEON_RE_WIDTH_HEIGHT: + track->maxy = ((ib_chunk->kdata[idx] >> 16) & 0x7FF); + break; + case RADEON_RB3D_COLORPITCH: + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) + tile_flags |= RADEON_COLOR_TILE_ENABLE; + if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) + tile_flags |= RADEON_COLOR_MICROTILE_ENABLE; + + tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); + tmp |= tile_flags; + ib[idx] = tmp; + + track->cb[0].pitch = ib_chunk->kdata[idx] & RADEON_COLORPITCH_MASK; + break; + case RADEON_RB3D_DEPTHPITCH: + track->zb.pitch = ib_chunk->kdata[idx] & RADEON_DEPTHPITCH_MASK; + break; + case RADEON_RB3D_CNTL: + switch ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) { + case 7: + case 8: + case 9: + case 11: + case 12: + track->cb[0].cpp = 1; + break; + case 3: + case 4: + case 15: + track->cb[0].cpp = 2; + break; + case 6: + track->cb[0].cpp = 4; + break; + default: + DRM_ERROR("Invalid color buffer format (%d) !\n", + ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f)); + return -EINVAL; + } + track->z_enabled = !!(ib_chunk->kdata[idx] & RADEON_Z_ENABLE); + break; + case RADEON_RB3D_ZSTENCILCNTL: + switch (ib_chunk->kdata[idx] & 0xf) { + case 0: + track->zb.cpp = 2; + break; + case 2: + case 3: + case 4: + case 5: + case 9: + case 11: + track->zb.cpp = 4; + break; + default: + break; + } + break; + case RADEON_RB3D_ZPASS_ADDR: + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); + break; + case RADEON_PP_CNTL: + { + uint32_t temp = ib_chunk->kdata[idx] >> 4; + for (i = 0; i < track->num_texture; i++) + track->textures[i].enabled = !!(temp & (1 << i)); + } + break; + case RADEON_SE_VF_CNTL: + track->vap_vf_cntl = ib_chunk->kdata[idx]; + break; + case RADEON_SE_VTX_FMT: + track->vtx_size = r100_get_vtx_size(ib_chunk->kdata[idx]); + break; + case RADEON_PP_TEX_SIZE_0: + case RADEON_PP_TEX_SIZE_1: + case RADEON_PP_TEX_SIZE_2: + i = (reg - RADEON_PP_TEX_SIZE_0) / 8; + track->textures[i].width = (ib_chunk->kdata[idx] & RADEON_TEX_USIZE_MASK) + 1; + track->textures[i].height = ((ib_chunk->kdata[idx] & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1; + break; + case RADEON_PP_TEX_PITCH_0: + case RADEON_PP_TEX_PITCH_1: + case RADEON_PP_TEX_PITCH_2: + i = (reg - RADEON_PP_TEX_PITCH_0) / 8; + track->textures[i].pitch = ib_chunk->kdata[idx] + 32; + break; + case RADEON_PP_TXFILTER_0: + case RADEON_PP_TXFILTER_1: + case RADEON_PP_TXFILTER_2: + i = (reg - RADEON_PP_TXFILTER_0) / 24; + track->textures[i].num_levels = ((ib_chunk->kdata[idx] & RADEON_MAX_MIP_LEVEL_MASK) + >> RADEON_MAX_MIP_LEVEL_SHIFT); + tmp = (ib_chunk->kdata[idx] >> 23) & 0x7; + if (tmp == 2 || tmp == 6) + track->textures[i].roundup_w = false; + tmp = (ib_chunk->kdata[idx] >> 27) & 0x7; + if (tmp == 2 || tmp == 6) + track->textures[i].roundup_h = false; + break; + case RADEON_PP_TXFORMAT_0: + case RADEON_PP_TXFORMAT_1: + case RADEON_PP_TXFORMAT_2: + i = (reg - RADEON_PP_TXFORMAT_0) / 24; + if (ib_chunk->kdata[idx] & RADEON_TXFORMAT_NON_POWER2) { + track->textures[i].use_pitch = 1; + } else { + track->textures[i].use_pitch = 0; + track->textures[i].width = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK); + track->textures[i].height = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK); + } + if (ib_chunk->kdata[idx] & RADEON_TXFORMAT_CUBIC_MAP_ENABLE) + track->textures[i].tex_coord_type = 2; + switch ((ib_chunk->kdata[idx] & RADEON_TXFORMAT_FORMAT_MASK)) { + case RADEON_TXFORMAT_I8: + case RADEON_TXFORMAT_RGB332: + case RADEON_TXFORMAT_Y8: + track->textures[i].cpp = 1; + break; + case RADEON_TXFORMAT_AI88: + case RADEON_TXFORMAT_ARGB1555: + case RADEON_TXFORMAT_RGB565: + case RADEON_TXFORMAT_ARGB4444: + case RADEON_TXFORMAT_VYUY422: + case RADEON_TXFORMAT_YVYU422: + case RADEON_TXFORMAT_DXT1: + case RADEON_TXFORMAT_SHADOW16: + case RADEON_TXFORMAT_LDUDV655: + case RADEON_TXFORMAT_DUDV88: + track->textures[i].cpp = 2; + break; + case RADEON_TXFORMAT_ARGB8888: + case RADEON_TXFORMAT_RGBA8888: + case RADEON_TXFORMAT_DXT23: + case RADEON_TXFORMAT_DXT45: + case RADEON_TXFORMAT_SHADOW32: + case RADEON_TXFORMAT_LDUDUV8888: + track->textures[i].cpp = 4; + break; + } + track->textures[i].cube_info[4].width = 1 << ((ib_chunk->kdata[idx] >> 16) & 0xf); + track->textures[i].cube_info[4].height = 1 << ((ib_chunk->kdata[idx] >> 20) & 0xf); + break; + case RADEON_PP_CUBIC_FACES_0: + case RADEON_PP_CUBIC_FACES_1: + case RADEON_PP_CUBIC_FACES_2: + tmp = ib_chunk->kdata[idx]; + i = (reg - RADEON_PP_CUBIC_FACES_0) / 4; + for (face = 0; face < 4; face++) { + track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf); + track->textures[i].cube_info[face].height = 1 << ((tmp >> ((face * 8) + 4)) & 0xf); + } + break; + default: + printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n", + reg, idx); + return -EINVAL; + } + return 0; +} + +int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p, + struct radeon_cs_packet *pkt, + struct radeon_object *robj) +{ + struct radeon_cs_chunk *ib_chunk; + unsigned idx; + + ib_chunk = &p->chunks[p->chunk_ib_idx]; + idx = pkt->idx + 1; + if ((ib_chunk->kdata[idx+2] + 1) > radeon_object_size(robj)) { + DRM_ERROR("[drm] Buffer too small for PACKET3 INDX_BUFFER " + "(need %u have %lu) !\n", + ib_chunk->kdata[idx+2] + 1, + radeon_object_size(robj)); + return -EINVAL; + } + return 0; +} + +static int r100_packet3_check(struct radeon_cs_parser *p, + struct radeon_cs_packet *pkt) +{ + struct radeon_cs_chunk *ib_chunk; + struct radeon_cs_reloc *reloc; + struct r100_cs_track *track; + unsigned idx; + unsigned i, c; + volatile uint32_t *ib; + int r; + + ib = p->ib->ptr; + ib_chunk = &p->chunks[p->chunk_ib_idx]; + idx = pkt->idx + 1; + track = (struct r100_cs_track *)p->track; + switch (pkt->opcode) { + case PACKET3_3D_LOAD_VBPNTR: + c = ib_chunk->kdata[idx++]; + track->num_arrays = c; + for (i = 0; i < (c - 1); i += 2, idx += 3) { + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for packet3 %d\n", + pkt->opcode); + r100_cs_dump_packet(p, pkt); + return r; + } + ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); + track->arrays[i + 0].robj = reloc->robj; + track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8; + track->arrays[i + 0].esize &= 0x7F; + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for packet3 %d\n", + pkt->opcode); + r100_cs_dump_packet(p, pkt); + return r; + } + ib[idx+2] = ib_chunk->kdata[idx+2] + ((u32)reloc->lobj.gpu_offset); + track->arrays[i + 1].robj = reloc->robj; + track->arrays[i + 1].esize = ib_chunk->kdata[idx] >> 24; + track->arrays[i + 1].esize &= 0x7F; + } + if (c & 1) { + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for packet3 %d\n", + pkt->opcode); + r100_cs_dump_packet(p, pkt); + return r; + } + ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); + track->arrays[i + 0].robj = reloc->robj; + track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8; + track->arrays[i + 0].esize &= 0x7F; + } + break; + case PACKET3_INDX_BUFFER: + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode); + r100_cs_dump_packet(p, pkt); + return r; + } + ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); + r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj); + if (r) { + return r; + } + break; + case 0x23: + /* 3D_RNDR_GEN_INDX_PRIM on r100/r200 */ + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode); + r100_cs_dump_packet(p, pkt); + return r; + } + ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); + track->num_arrays = 1; + track->vtx_size = r100_get_vtx_size(ib_chunk->kdata[idx+2]); + + track->arrays[0].robj = reloc->robj; + track->arrays[0].esize = track->vtx_size; + + track->max_indx = ib_chunk->kdata[idx+1]; + + track->vap_vf_cntl = ib_chunk->kdata[idx+3]; + track->immd_dwords = pkt->count - 1; + r = r100_cs_track_check(p->rdev, track); + if (r) + return r; + break; + case PACKET3_3D_DRAW_IMMD: + if (((ib_chunk->kdata[idx+1] >> 4) & 0x3) != 3) { + DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); + return -EINVAL; + } + track->vap_vf_cntl = ib_chunk->kdata[idx+1]; + track->immd_dwords = pkt->count - 1; + r = r100_cs_track_check(p->rdev, track); + if (r) + return r; + break; /* triggers drawing using in-packet vertex data */ case PACKET3_3D_DRAW_IMMD_2: + if (((ib_chunk->kdata[idx] >> 4) & 0x3) != 3) { + DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); + return -EINVAL; + } + track->vap_vf_cntl = ib_chunk->kdata[idx]; + track->immd_dwords = pkt->count; + r = r100_cs_track_check(p->rdev, track); + if (r) + return r; + break; /* triggers drawing using in-packet vertex data */ case PACKET3_3D_DRAW_VBUF_2: + track->vap_vf_cntl = ib_chunk->kdata[idx]; + r = r100_cs_track_check(p->rdev, track); + if (r) + return r; + break; /* triggers drawing of vertex buffers setup elsewhere */ case PACKET3_3D_DRAW_INDX_2: + track->vap_vf_cntl = ib_chunk->kdata[idx]; + r = r100_cs_track_check(p->rdev, track); + if (r) + return r; + break; /* triggers drawing using indices to vertex buffer */ case PACKET3_3D_DRAW_VBUF: + track->vap_vf_cntl = ib_chunk->kdata[idx + 1]; + r = r100_cs_track_check(p->rdev, track); + if (r) + return r; + break; /* triggers drawing of vertex buffers setup elsewhere */ case PACKET3_3D_DRAW_INDX: + track->vap_vf_cntl = ib_chunk->kdata[idx + 1]; + r = r100_cs_track_check(p->rdev, track); + if (r) + return r; + break; /* triggers drawing using indices to vertex buffer */ case PACKET3_NOP: break; @@ -1002,8 +1599,12 @@ static int r100_packet3_check(struct radeon_cs_parser *p, int r100_cs_parse(struct radeon_cs_parser *p) { struct radeon_cs_packet pkt; + struct r100_cs_track *track; int r; + track = kzalloc(sizeof(*track), GFP_KERNEL); + r100_cs_track_clear(p->rdev, track); + p->track = track; do { r = r100_cs_packet_parse(p, &pkt, p->idx); if (r) { @@ -1012,7 +1613,16 @@ int r100_cs_parse(struct radeon_cs_parser *p) p->idx += pkt.count + 2; switch (pkt.type) { case PACKET_TYPE0: - r = r100_packet0_check(p, &pkt); + if (p->rdev->family >= CHIP_R200) + r = r100_cs_parse_packet0(p, &pkt, + p->rdev->config.r100.reg_safe_bm, + p->rdev->config.r100.reg_safe_bm_size, + &r200_packet0_check); + else + r = r100_cs_parse_packet0(p, &pkt, + p->rdev->config.r100.reg_safe_bm, + p->rdev->config.r100.reg_safe_bm_size, + &r100_packet0_check); break; case PACKET_TYPE2: break; @@ -1256,29 +1866,100 @@ static void r100_vram_get_type(struct radeon_device *rdev) } } -void r100_vram_info(struct radeon_device *rdev) +static u32 r100_get_accessible_vram(struct radeon_device *rdev) { - r100_vram_get_type(rdev); + u32 aper_size; + u8 byte; + + aper_size = RREG32(RADEON_CONFIG_APER_SIZE); + + /* Set HDP_APER_CNTL only on cards that are known not to be broken, + * that is has the 2nd generation multifunction PCI interface + */ + if (rdev->family == CHIP_RV280 || + rdev->family >= CHIP_RV350) { + WREG32_P(RADEON_HOST_PATH_CNTL, RADEON_HDP_APER_CNTL, + ~RADEON_HDP_APER_CNTL); + DRM_INFO("Generation 2 PCI interface, using max accessible memory\n"); + return aper_size * 2; + } + + /* Older cards have all sorts of funny issues to deal with. First + * check if it's a multifunction card by reading the PCI config + * header type... Limit those to one aperture size + */ + pci_read_config_byte(rdev->pdev, 0xe, &byte); + if (byte & 0x80) { + DRM_INFO("Generation 1 PCI interface in multifunction mode\n"); + DRM_INFO("Limiting VRAM to one aperture\n"); + return aper_size; + } + + /* Single function older card. We read HDP_APER_CNTL to see how the BIOS + * have set it up. We don't write this as it's broken on some ASICs but + * we expect the BIOS to have done the right thing (might be too optimistic...) + */ + if (RREG32(RADEON_HOST_PATH_CNTL) & RADEON_HDP_APER_CNTL) + return aper_size * 2; + return aper_size; +} + +void r100_vram_init_sizes(struct radeon_device *rdev) +{ + u64 config_aper_size; + u32 accessible; + + config_aper_size = RREG32(RADEON_CONFIG_APER_SIZE); if (rdev->flags & RADEON_IS_IGP) { uint32_t tom; /* read NB_TOM to get the amount of ram stolen for the GPU */ tom = RREG32(RADEON_NB_TOM); - rdev->mc.vram_size = (((tom >> 16) - (tom & 0xffff) + 1) << 16); - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); + rdev->mc.real_vram_size = (((tom >> 16) - (tom & 0xffff) + 1) << 16); + /* for IGPs we need to keep VRAM where it was put by the BIOS */ + rdev->mc.vram_location = (tom & 0xffff) << 16; + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); + rdev->mc.mc_vram_size = rdev->mc.real_vram_size; } else { - rdev->mc.vram_size = RREG32(RADEON_CONFIG_MEMSIZE); + rdev->mc.real_vram_size = RREG32(RADEON_CONFIG_MEMSIZE); /* Some production boards of m6 will report 0 * if it's 8 MB */ - if (rdev->mc.vram_size == 0) { - rdev->mc.vram_size = 8192 * 1024; - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); + if (rdev->mc.real_vram_size == 0) { + rdev->mc.real_vram_size = 8192 * 1024; + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); } + /* let driver place VRAM */ + rdev->mc.vram_location = 0xFFFFFFFFUL; + /* Fix for RN50, M6, M7 with 8/16/32(??) MBs of VRAM - + * Novell bug 204882 + along with lots of ubuntu ones */ + if (config_aper_size > rdev->mc.real_vram_size) + rdev->mc.mc_vram_size = config_aper_size; + else + rdev->mc.mc_vram_size = rdev->mc.real_vram_size; } + /* work out accessible VRAM */ + accessible = r100_get_accessible_vram(rdev); + rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + + if (accessible > rdev->mc.aper_size) + accessible = rdev->mc.aper_size; + + if (rdev->mc.mc_vram_size > rdev->mc.aper_size) + rdev->mc.mc_vram_size = rdev->mc.aper_size; + + if (rdev->mc.real_vram_size > rdev->mc.aper_size) + rdev->mc.real_vram_size = rdev->mc.aper_size; +} + +void r100_vram_info(struct radeon_device *rdev) +{ + r100_vram_get_type(rdev); + + r100_vram_init_sizes(rdev); } @@ -1338,28 +2019,17 @@ void r100_pll_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v) r100_pll_errata_after_data(rdev); } -uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg) -{ - if (reg < 0x10000) - return readl(((void __iomem *)rdev->rmmio) + reg); - else { - writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX); - return readl(((void __iomem *)rdev->rmmio) + RADEON_MM_DATA); - } -} - -void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v) -{ - if (reg < 0x10000) - writel(v, ((void __iomem *)rdev->rmmio) + reg); - else { - writel(reg, ((void __iomem *)rdev->rmmio) + RADEON_MM_INDEX); - writel(v, ((void __iomem *)rdev->rmmio) + RADEON_MM_DATA); - } -} - int r100_init(struct radeon_device *rdev) { + if (ASIC_IS_RN50(rdev)) { + rdev->config.r100.reg_safe_bm = rn50_reg_safe_bm; + rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(rn50_reg_safe_bm); + } else if (rdev->family < CHIP_R200) { + rdev->config.r100.reg_safe_bm = r100_reg_safe_bm; + rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(r100_reg_safe_bm); + } else { + return r200_init(rdev); + } return 0; } @@ -1533,3 +2203,987 @@ int r100_debugfs_mc_info_init(struct radeon_device *rdev) return 0; #endif } + +int r100_set_surface_reg(struct radeon_device *rdev, int reg, + uint32_t tiling_flags, uint32_t pitch, + uint32_t offset, uint32_t obj_size) +{ + int surf_index = reg * 16; + int flags = 0; + + /* r100/r200 divide by 16 */ + if (rdev->family < CHIP_R300) + flags = pitch / 16; + else + flags = pitch / 8; + + if (rdev->family <= CHIP_RS200) { + if ((tiling_flags & (RADEON_TILING_MACRO|RADEON_TILING_MICRO)) + == (RADEON_TILING_MACRO|RADEON_TILING_MICRO)) + flags |= RADEON_SURF_TILE_COLOR_BOTH; + if (tiling_flags & RADEON_TILING_MACRO) + flags |= RADEON_SURF_TILE_COLOR_MACRO; + } else if (rdev->family <= CHIP_RV280) { + if (tiling_flags & (RADEON_TILING_MACRO)) + flags |= R200_SURF_TILE_COLOR_MACRO; + if (tiling_flags & RADEON_TILING_MICRO) + flags |= R200_SURF_TILE_COLOR_MICRO; + } else { + if (tiling_flags & RADEON_TILING_MACRO) + flags |= R300_SURF_TILE_MACRO; + if (tiling_flags & RADEON_TILING_MICRO) + flags |= R300_SURF_TILE_MICRO; + } + + DRM_DEBUG("writing surface %d %d %x %x\n", reg, flags, offset, offset+obj_size-1); + WREG32(RADEON_SURFACE0_INFO + surf_index, flags); + WREG32(RADEON_SURFACE0_LOWER_BOUND + surf_index, offset); + WREG32(RADEON_SURFACE0_UPPER_BOUND + surf_index, offset + obj_size - 1); + return 0; +} + +void r100_clear_surface_reg(struct radeon_device *rdev, int reg) +{ + int surf_index = reg * 16; + WREG32(RADEON_SURFACE0_INFO + surf_index, 0); +} + +void r100_bandwidth_update(struct radeon_device *rdev) +{ + fixed20_12 trcd_ff, trp_ff, tras_ff, trbs_ff, tcas_ff; + fixed20_12 sclk_ff, mclk_ff, sclk_eff_ff, sclk_delay_ff; + fixed20_12 peak_disp_bw, mem_bw, pix_clk, pix_clk2, temp_ff, crit_point_ff; + uint32_t temp, data, mem_trcd, mem_trp, mem_tras; + fixed20_12 memtcas_ff[8] = { + fixed_init(1), + fixed_init(2), + fixed_init(3), + fixed_init(0), + fixed_init_half(1), + fixed_init_half(2), + fixed_init(0), + }; + fixed20_12 memtcas_rs480_ff[8] = { + fixed_init(0), + fixed_init(1), + fixed_init(2), + fixed_init(3), + fixed_init(0), + fixed_init_half(1), + fixed_init_half(2), + fixed_init_half(3), + }; + fixed20_12 memtcas2_ff[8] = { + fixed_init(0), + fixed_init(1), + fixed_init(2), + fixed_init(3), + fixed_init(4), + fixed_init(5), + fixed_init(6), + fixed_init(7), + }; + fixed20_12 memtrbs[8] = { + fixed_init(1), + fixed_init_half(1), + fixed_init(2), + fixed_init_half(2), + fixed_init(3), + fixed_init_half(3), + fixed_init(4), + fixed_init_half(4) + }; + fixed20_12 memtrbs_r4xx[8] = { + fixed_init(4), + fixed_init(5), + fixed_init(6), + fixed_init(7), + fixed_init(8), + fixed_init(9), + fixed_init(10), + fixed_init(11) + }; + fixed20_12 min_mem_eff; + fixed20_12 mc_latency_sclk, mc_latency_mclk, k1; + fixed20_12 cur_latency_mclk, cur_latency_sclk; + fixed20_12 disp_latency, disp_latency_overhead, disp_drain_rate, + disp_drain_rate2, read_return_rate; + fixed20_12 time_disp1_drop_priority; + int c; + int cur_size = 16; /* in octawords */ + int critical_point = 0, critical_point2; +/* uint32_t read_return_rate, time_disp1_drop_priority; */ + int stop_req, max_stop_req; + struct drm_display_mode *mode1 = NULL; + struct drm_display_mode *mode2 = NULL; + uint32_t pixel_bytes1 = 0; + uint32_t pixel_bytes2 = 0; + + if (rdev->mode_info.crtcs[0]->base.enabled) { + mode1 = &rdev->mode_info.crtcs[0]->base.mode; + pixel_bytes1 = rdev->mode_info.crtcs[0]->base.fb->bits_per_pixel / 8; + } + if (rdev->mode_info.crtcs[1]->base.enabled) { + mode2 = &rdev->mode_info.crtcs[1]->base.mode; + pixel_bytes2 = rdev->mode_info.crtcs[1]->base.fb->bits_per_pixel / 8; + } + + min_mem_eff.full = rfixed_const_8(0); + /* get modes */ + if ((rdev->disp_priority == 2) && ASIC_IS_R300(rdev)) { + uint32_t mc_init_misc_lat_timer = RREG32(R300_MC_INIT_MISC_LAT_TIMER); + mc_init_misc_lat_timer &= ~(R300_MC_DISP1R_INIT_LAT_MASK << R300_MC_DISP1R_INIT_LAT_SHIFT); + mc_init_misc_lat_timer &= ~(R300_MC_DISP0R_INIT_LAT_MASK << R300_MC_DISP0R_INIT_LAT_SHIFT); + /* check crtc enables */ + if (mode2) + mc_init_misc_lat_timer |= (1 << R300_MC_DISP1R_INIT_LAT_SHIFT); + if (mode1) + mc_init_misc_lat_timer |= (1 << R300_MC_DISP0R_INIT_LAT_SHIFT); + WREG32(R300_MC_INIT_MISC_LAT_TIMER, mc_init_misc_lat_timer); + } + + /* + * determine is there is enough bw for current mode + */ + mclk_ff.full = rfixed_const(rdev->clock.default_mclk); + temp_ff.full = rfixed_const(100); + mclk_ff.full = rfixed_div(mclk_ff, temp_ff); + sclk_ff.full = rfixed_const(rdev->clock.default_sclk); + sclk_ff.full = rfixed_div(sclk_ff, temp_ff); + + temp = (rdev->mc.vram_width / 8) * (rdev->mc.vram_is_ddr ? 2 : 1); + temp_ff.full = rfixed_const(temp); + mem_bw.full = rfixed_mul(mclk_ff, temp_ff); + + pix_clk.full = 0; + pix_clk2.full = 0; + peak_disp_bw.full = 0; + if (mode1) { + temp_ff.full = rfixed_const(1000); + pix_clk.full = rfixed_const(mode1->clock); /* convert to fixed point */ + pix_clk.full = rfixed_div(pix_clk, temp_ff); + temp_ff.full = rfixed_const(pixel_bytes1); + peak_disp_bw.full += rfixed_mul(pix_clk, temp_ff); + } + if (mode2) { + temp_ff.full = rfixed_const(1000); + pix_clk2.full = rfixed_const(mode2->clock); /* convert to fixed point */ + pix_clk2.full = rfixed_div(pix_clk2, temp_ff); + temp_ff.full = rfixed_const(pixel_bytes2); + peak_disp_bw.full += rfixed_mul(pix_clk2, temp_ff); + } + + mem_bw.full = rfixed_mul(mem_bw, min_mem_eff); + if (peak_disp_bw.full >= mem_bw.full) { + DRM_ERROR("You may not have enough display bandwidth for current mode\n" + "If you have flickering problem, try to lower resolution, refresh rate, or color depth\n"); + } + + /* Get values from the EXT_MEM_CNTL register...converting its contents. */ + temp = RREG32(RADEON_MEM_TIMING_CNTL); + if ((rdev->family == CHIP_RV100) || (rdev->flags & RADEON_IS_IGP)) { /* RV100, M6, IGPs */ + mem_trcd = ((temp >> 2) & 0x3) + 1; + mem_trp = ((temp & 0x3)) + 1; + mem_tras = ((temp & 0x70) >> 4) + 1; + } else if (rdev->family == CHIP_R300 || + rdev->family == CHIP_R350) { /* r300, r350 */ + mem_trcd = (temp & 0x7) + 1; + mem_trp = ((temp >> 8) & 0x7) + 1; + mem_tras = ((temp >> 11) & 0xf) + 4; + } else if (rdev->family == CHIP_RV350 || + rdev->family <= CHIP_RV380) { + /* rv3x0 */ + mem_trcd = (temp & 0x7) + 3; + mem_trp = ((temp >> 8) & 0x7) + 3; + mem_tras = ((temp >> 11) & 0xf) + 6; + } else if (rdev->family == CHIP_R420 || + rdev->family == CHIP_R423 || + rdev->family == CHIP_RV410) { + /* r4xx */ + mem_trcd = (temp & 0xf) + 3; + if (mem_trcd > 15) + mem_trcd = 15; + mem_trp = ((temp >> 8) & 0xf) + 3; + if (mem_trp > 15) + mem_trp = 15; + mem_tras = ((temp >> 12) & 0x1f) + 6; + if (mem_tras > 31) + mem_tras = 31; + } else { /* RV200, R200 */ + mem_trcd = (temp & 0x7) + 1; + mem_trp = ((temp >> 8) & 0x7) + 1; + mem_tras = ((temp >> 12) & 0xf) + 4; + } + /* convert to FF */ + trcd_ff.full = rfixed_const(mem_trcd); + trp_ff.full = rfixed_const(mem_trp); + tras_ff.full = rfixed_const(mem_tras); + + /* Get values from the MEM_SDRAM_MODE_REG register...converting its */ + temp = RREG32(RADEON_MEM_SDRAM_MODE_REG); + data = (temp & (7 << 20)) >> 20; + if ((rdev->family == CHIP_RV100) || rdev->flags & RADEON_IS_IGP) { + if (rdev->family == CHIP_RS480) /* don't think rs400 */ + tcas_ff = memtcas_rs480_ff[data]; + else + tcas_ff = memtcas_ff[data]; + } else + tcas_ff = memtcas2_ff[data]; + + if (rdev->family == CHIP_RS400 || + rdev->family == CHIP_RS480) { + /* extra cas latency stored in bits 23-25 0-4 clocks */ + data = (temp >> 23) & 0x7; + if (data < 5) + tcas_ff.full += rfixed_const(data); + } + + if (ASIC_IS_R300(rdev) && !(rdev->flags & RADEON_IS_IGP)) { + /* on the R300, Tcas is included in Trbs. + */ + temp = RREG32(RADEON_MEM_CNTL); + data = (R300_MEM_NUM_CHANNELS_MASK & temp); + if (data == 1) { + if (R300_MEM_USE_CD_CH_ONLY & temp) { + temp = RREG32(R300_MC_IND_INDEX); + temp &= ~R300_MC_IND_ADDR_MASK; + temp |= R300_MC_READ_CNTL_CD_mcind; + WREG32(R300_MC_IND_INDEX, temp); + temp = RREG32(R300_MC_IND_DATA); + data = (R300_MEM_RBS_POSITION_C_MASK & temp); + } else { + temp = RREG32(R300_MC_READ_CNTL_AB); + data = (R300_MEM_RBS_POSITION_A_MASK & temp); + } + } else { + temp = RREG32(R300_MC_READ_CNTL_AB); + data = (R300_MEM_RBS_POSITION_A_MASK & temp); + } + if (rdev->family == CHIP_RV410 || + rdev->family == CHIP_R420 || + rdev->family == CHIP_R423) + trbs_ff = memtrbs_r4xx[data]; + else + trbs_ff = memtrbs[data]; + tcas_ff.full += trbs_ff.full; + } + + sclk_eff_ff.full = sclk_ff.full; + + if (rdev->flags & RADEON_IS_AGP) { + fixed20_12 agpmode_ff; + agpmode_ff.full = rfixed_const(radeon_agpmode); + temp_ff.full = rfixed_const_666(16); + sclk_eff_ff.full -= rfixed_mul(agpmode_ff, temp_ff); + } + /* TODO PCIE lanes may affect this - agpmode == 16?? */ + + if (ASIC_IS_R300(rdev)) { + sclk_delay_ff.full = rfixed_const(250); + } else { + if ((rdev->family == CHIP_RV100) || + rdev->flags & RADEON_IS_IGP) { + if (rdev->mc.vram_is_ddr) + sclk_delay_ff.full = rfixed_const(41); + else + sclk_delay_ff.full = rfixed_const(33); + } else { + if (rdev->mc.vram_width == 128) + sclk_delay_ff.full = rfixed_const(57); + else + sclk_delay_ff.full = rfixed_const(41); + } + } + + mc_latency_sclk.full = rfixed_div(sclk_delay_ff, sclk_eff_ff); + + if (rdev->mc.vram_is_ddr) { + if (rdev->mc.vram_width == 32) { + k1.full = rfixed_const(40); + c = 3; + } else { + k1.full = rfixed_const(20); + c = 1; + } + } else { + k1.full = rfixed_const(40); + c = 3; + } + + temp_ff.full = rfixed_const(2); + mc_latency_mclk.full = rfixed_mul(trcd_ff, temp_ff); + temp_ff.full = rfixed_const(c); + mc_latency_mclk.full += rfixed_mul(tcas_ff, temp_ff); + temp_ff.full = rfixed_const(4); + mc_latency_mclk.full += rfixed_mul(tras_ff, temp_ff); + mc_latency_mclk.full += rfixed_mul(trp_ff, temp_ff); + mc_latency_mclk.full += k1.full; + + mc_latency_mclk.full = rfixed_div(mc_latency_mclk, mclk_ff); + mc_latency_mclk.full += rfixed_div(temp_ff, sclk_eff_ff); + + /* + HW cursor time assuming worst case of full size colour cursor. + */ + temp_ff.full = rfixed_const((2 * (cur_size - (rdev->mc.vram_is_ddr + 1)))); + temp_ff.full += trcd_ff.full; + if (temp_ff.full < tras_ff.full) + temp_ff.full = tras_ff.full; + cur_latency_mclk.full = rfixed_div(temp_ff, mclk_ff); + + temp_ff.full = rfixed_const(cur_size); + cur_latency_sclk.full = rfixed_div(temp_ff, sclk_eff_ff); + /* + Find the total latency for the display data. + */ + disp_latency_overhead.full = rfixed_const(80); + disp_latency_overhead.full = rfixed_div(disp_latency_overhead, sclk_ff); + mc_latency_mclk.full += disp_latency_overhead.full + cur_latency_mclk.full; + mc_latency_sclk.full += disp_latency_overhead.full + cur_latency_sclk.full; + + if (mc_latency_mclk.full > mc_latency_sclk.full) + disp_latency.full = mc_latency_mclk.full; + else + disp_latency.full = mc_latency_sclk.full; + + /* setup Max GRPH_STOP_REQ default value */ + if (ASIC_IS_RV100(rdev)) + max_stop_req = 0x5c; + else + max_stop_req = 0x7c; + + if (mode1) { + /* CRTC1 + Set GRPH_BUFFER_CNTL register using h/w defined optimal values. + GRPH_STOP_REQ <= MIN[ 0x7C, (CRTC_H_DISP + 1) * (bit depth) / 0x10 ] + */ + stop_req = mode1->hdisplay * pixel_bytes1 / 16; + + if (stop_req > max_stop_req) + stop_req = max_stop_req; + + /* + Find the drain rate of the display buffer. + */ + temp_ff.full = rfixed_const((16/pixel_bytes1)); + disp_drain_rate.full = rfixed_div(pix_clk, temp_ff); + + /* + Find the critical point of the display buffer. + */ + crit_point_ff.full = rfixed_mul(disp_drain_rate, disp_latency); + crit_point_ff.full += rfixed_const_half(0); + + critical_point = rfixed_trunc(crit_point_ff); + + if (rdev->disp_priority == 2) { + critical_point = 0; + } + + /* + The critical point should never be above max_stop_req-4. Setting + GRPH_CRITICAL_CNTL = 0 will thus force high priority all the time. + */ + if (max_stop_req - critical_point < 4) + critical_point = 0; + + if (critical_point == 0 && mode2 && rdev->family == CHIP_R300) { + /* some R300 cards have problem with this set to 0, when CRTC2 is enabled.*/ + critical_point = 0x10; + } + + temp = RREG32(RADEON_GRPH_BUFFER_CNTL); + temp &= ~(RADEON_GRPH_STOP_REQ_MASK); + temp |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); + temp &= ~(RADEON_GRPH_START_REQ_MASK); + if ((rdev->family == CHIP_R350) && + (stop_req > 0x15)) { + stop_req -= 0x10; + } + temp |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); + temp |= RADEON_GRPH_BUFFER_SIZE; + temp &= ~(RADEON_GRPH_CRITICAL_CNTL | + RADEON_GRPH_CRITICAL_AT_SOF | + RADEON_GRPH_STOP_CNTL); + /* + Write the result into the register. + */ + WREG32(RADEON_GRPH_BUFFER_CNTL, ((temp & ~RADEON_GRPH_CRITICAL_POINT_MASK) | + (critical_point << RADEON_GRPH_CRITICAL_POINT_SHIFT))); + +#if 0 + if ((rdev->family == CHIP_RS400) || + (rdev->family == CHIP_RS480)) { + /* attempt to program RS400 disp regs correctly ??? */ + temp = RREG32(RS400_DISP1_REG_CNTL); + temp &= ~(RS400_DISP1_START_REQ_LEVEL_MASK | + RS400_DISP1_STOP_REQ_LEVEL_MASK); + WREG32(RS400_DISP1_REQ_CNTL1, (temp | + (critical_point << RS400_DISP1_START_REQ_LEVEL_SHIFT) | + (critical_point << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); + temp = RREG32(RS400_DMIF_MEM_CNTL1); + temp &= ~(RS400_DISP1_CRITICAL_POINT_START_MASK | + RS400_DISP1_CRITICAL_POINT_STOP_MASK); + WREG32(RS400_DMIF_MEM_CNTL1, (temp | + (critical_point << RS400_DISP1_CRITICAL_POINT_START_SHIFT) | + (critical_point << RS400_DISP1_CRITICAL_POINT_STOP_SHIFT))); + } +#endif + + DRM_DEBUG("GRPH_BUFFER_CNTL from to %x\n", + /* (unsigned int)info->SavedReg->grph_buffer_cntl, */ + (unsigned int)RREG32(RADEON_GRPH_BUFFER_CNTL)); + } + + if (mode2) { + u32 grph2_cntl; + stop_req = mode2->hdisplay * pixel_bytes2 / 16; + + if (stop_req > max_stop_req) + stop_req = max_stop_req; + + /* + Find the drain rate of the display buffer. + */ + temp_ff.full = rfixed_const((16/pixel_bytes2)); + disp_drain_rate2.full = rfixed_div(pix_clk2, temp_ff); + + grph2_cntl = RREG32(RADEON_GRPH2_BUFFER_CNTL); + grph2_cntl &= ~(RADEON_GRPH_STOP_REQ_MASK); + grph2_cntl |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); + grph2_cntl &= ~(RADEON_GRPH_START_REQ_MASK); + if ((rdev->family == CHIP_R350) && + (stop_req > 0x15)) { + stop_req -= 0x10; + } + grph2_cntl |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); + grph2_cntl |= RADEON_GRPH_BUFFER_SIZE; + grph2_cntl &= ~(RADEON_GRPH_CRITICAL_CNTL | + RADEON_GRPH_CRITICAL_AT_SOF | + RADEON_GRPH_STOP_CNTL); + + if ((rdev->family == CHIP_RS100) || + (rdev->family == CHIP_RS200)) + critical_point2 = 0; + else { + temp = (rdev->mc.vram_width * rdev->mc.vram_is_ddr + 1)/128; + temp_ff.full = rfixed_const(temp); + temp_ff.full = rfixed_mul(mclk_ff, temp_ff); + if (sclk_ff.full < temp_ff.full) + temp_ff.full = sclk_ff.full; + + read_return_rate.full = temp_ff.full; + + if (mode1) { + temp_ff.full = read_return_rate.full - disp_drain_rate.full; + time_disp1_drop_priority.full = rfixed_div(crit_point_ff, temp_ff); + } else { + time_disp1_drop_priority.full = 0; + } + crit_point_ff.full = disp_latency.full + time_disp1_drop_priority.full + disp_latency.full; + crit_point_ff.full = rfixed_mul(crit_point_ff, disp_drain_rate2); + crit_point_ff.full += rfixed_const_half(0); + + critical_point2 = rfixed_trunc(crit_point_ff); + + if (rdev->disp_priority == 2) { + critical_point2 = 0; + } + + if (max_stop_req - critical_point2 < 4) + critical_point2 = 0; + + } + + if (critical_point2 == 0 && rdev->family == CHIP_R300) { + /* some R300 cards have problem with this set to 0 */ + critical_point2 = 0x10; + } + + WREG32(RADEON_GRPH2_BUFFER_CNTL, ((grph2_cntl & ~RADEON_GRPH_CRITICAL_POINT_MASK) | + (critical_point2 << RADEON_GRPH_CRITICAL_POINT_SHIFT))); + + if ((rdev->family == CHIP_RS400) || + (rdev->family == CHIP_RS480)) { +#if 0 + /* attempt to program RS400 disp2 regs correctly ??? */ + temp = RREG32(RS400_DISP2_REQ_CNTL1); + temp &= ~(RS400_DISP2_START_REQ_LEVEL_MASK | + RS400_DISP2_STOP_REQ_LEVEL_MASK); + WREG32(RS400_DISP2_REQ_CNTL1, (temp | + (critical_point2 << RS400_DISP1_START_REQ_LEVEL_SHIFT) | + (critical_point2 << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); + temp = RREG32(RS400_DISP2_REQ_CNTL2); + temp &= ~(RS400_DISP2_CRITICAL_POINT_START_MASK | + RS400_DISP2_CRITICAL_POINT_STOP_MASK); + WREG32(RS400_DISP2_REQ_CNTL2, (temp | + (critical_point2 << RS400_DISP2_CRITICAL_POINT_START_SHIFT) | + (critical_point2 << RS400_DISP2_CRITICAL_POINT_STOP_SHIFT))); +#endif + WREG32(RS400_DISP2_REQ_CNTL1, 0x105DC1CC); + WREG32(RS400_DISP2_REQ_CNTL2, 0x2749D000); + WREG32(RS400_DMIF_MEM_CNTL1, 0x29CA71DC); + WREG32(RS400_DISP1_REQ_CNTL1, 0x28FBC3AC); + } + + DRM_DEBUG("GRPH2_BUFFER_CNTL from to %x\n", + (unsigned int)RREG32(RADEON_GRPH2_BUFFER_CNTL)); + } +} + +static inline void r100_cs_track_texture_print(struct r100_cs_track_texture *t) +{ + DRM_ERROR("pitch %d\n", t->pitch); + DRM_ERROR("width %d\n", t->width); + DRM_ERROR("height %d\n", t->height); + DRM_ERROR("num levels %d\n", t->num_levels); + DRM_ERROR("depth %d\n", t->txdepth); + DRM_ERROR("bpp %d\n", t->cpp); + DRM_ERROR("coordinate type %d\n", t->tex_coord_type); + DRM_ERROR("width round to power of 2 %d\n", t->roundup_w); + DRM_ERROR("height round to power of 2 %d\n", t->roundup_h); +} + +static int r100_cs_track_cube(struct radeon_device *rdev, + struct r100_cs_track *track, unsigned idx) +{ + unsigned face, w, h; + struct radeon_object *cube_robj; + unsigned long size; + + for (face = 0; face < 5; face++) { + cube_robj = track->textures[idx].cube_info[face].robj; + w = track->textures[idx].cube_info[face].width; + h = track->textures[idx].cube_info[face].height; + + size = w * h; + size *= track->textures[idx].cpp; + + size += track->textures[idx].cube_info[face].offset; + + if (size > radeon_object_size(cube_robj)) { + DRM_ERROR("Cube texture offset greater than object size %lu %lu\n", + size, radeon_object_size(cube_robj)); + r100_cs_track_texture_print(&track->textures[idx]); + return -1; + } + } + return 0; +} + +static int r100_cs_track_texture_check(struct radeon_device *rdev, + struct r100_cs_track *track) +{ + struct radeon_object *robj; + unsigned long size; + unsigned u, i, w, h; + int ret; + + for (u = 0; u < track->num_texture; u++) { + if (!track->textures[u].enabled) + continue; + robj = track->textures[u].robj; + if (robj == NULL) { + DRM_ERROR("No texture bound to unit %u\n", u); + return -EINVAL; + } + size = 0; + for (i = 0; i <= track->textures[u].num_levels; i++) { + if (track->textures[u].use_pitch) { + if (rdev->family < CHIP_R300) + w = (track->textures[u].pitch / track->textures[u].cpp) / (1 << i); + else + w = track->textures[u].pitch / (1 << i); + } else { + w = track->textures[u].width / (1 << i); + if (rdev->family >= CHIP_RV515) + w |= track->textures[u].width_11; + if (track->textures[u].roundup_w) + w = roundup_pow_of_two(w); + } + h = track->textures[u].height / (1 << i); + if (rdev->family >= CHIP_RV515) + h |= track->textures[u].height_11; + if (track->textures[u].roundup_h) + h = roundup_pow_of_two(h); + size += w * h; + } + size *= track->textures[u].cpp; + switch (track->textures[u].tex_coord_type) { + case 0: + break; + case 1: + size *= (1 << track->textures[u].txdepth); + break; + case 2: + if (track->separate_cube) { + ret = r100_cs_track_cube(rdev, track, u); + if (ret) + return ret; + } else + size *= 6; + break; + default: + DRM_ERROR("Invalid texture coordinate type %u for unit " + "%u\n", track->textures[u].tex_coord_type, u); + return -EINVAL; + } + if (size > radeon_object_size(robj)) { + DRM_ERROR("Texture of unit %u needs %lu bytes but is " + "%lu\n", u, size, radeon_object_size(robj)); + r100_cs_track_texture_print(&track->textures[u]); + return -EINVAL; + } + } + return 0; +} + +int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track) +{ + unsigned i; + unsigned long size; + unsigned prim_walk; + unsigned nverts; + + for (i = 0; i < track->num_cb; i++) { + if (track->cb[i].robj == NULL) { + DRM_ERROR("[drm] No buffer for color buffer %d !\n", i); + return -EINVAL; + } + size = track->cb[i].pitch * track->cb[i].cpp * track->maxy; + size += track->cb[i].offset; + if (size > radeon_object_size(track->cb[i].robj)) { + DRM_ERROR("[drm] Buffer too small for color buffer %d " + "(need %lu have %lu) !\n", i, size, + radeon_object_size(track->cb[i].robj)); + DRM_ERROR("[drm] color buffer %d (%u %u %u %u)\n", + i, track->cb[i].pitch, track->cb[i].cpp, + track->cb[i].offset, track->maxy); + return -EINVAL; + } + } + if (track->z_enabled) { + if (track->zb.robj == NULL) { + DRM_ERROR("[drm] No buffer for z buffer !\n"); + return -EINVAL; + } + size = track->zb.pitch * track->zb.cpp * track->maxy; + size += track->zb.offset; + if (size > radeon_object_size(track->zb.robj)) { + DRM_ERROR("[drm] Buffer too small for z buffer " + "(need %lu have %lu) !\n", size, + radeon_object_size(track->zb.robj)); + DRM_ERROR("[drm] zbuffer (%u %u %u %u)\n", + track->zb.pitch, track->zb.cpp, + track->zb.offset, track->maxy); + return -EINVAL; + } + } + prim_walk = (track->vap_vf_cntl >> 4) & 0x3; + nverts = (track->vap_vf_cntl >> 16) & 0xFFFF; + switch (prim_walk) { + case 1: + for (i = 0; i < track->num_arrays; i++) { + size = track->arrays[i].esize * track->max_indx * 4; + if (track->arrays[i].robj == NULL) { + DRM_ERROR("(PW %u) Vertex array %u no buffer " + "bound\n", prim_walk, i); + return -EINVAL; + } + if (size > radeon_object_size(track->arrays[i].robj)) { + DRM_ERROR("(PW %u) Vertex array %u need %lu dwords " + "have %lu dwords\n", prim_walk, i, + size >> 2, + radeon_object_size(track->arrays[i].robj) >> 2); + DRM_ERROR("Max indices %u\n", track->max_indx); + return -EINVAL; + } + } + break; + case 2: + for (i = 0; i < track->num_arrays; i++) { + size = track->arrays[i].esize * (nverts - 1) * 4; + if (track->arrays[i].robj == NULL) { + DRM_ERROR("(PW %u) Vertex array %u no buffer " + "bound\n", prim_walk, i); + return -EINVAL; + } + if (size > radeon_object_size(track->arrays[i].robj)) { + DRM_ERROR("(PW %u) Vertex array %u need %lu dwords " + "have %lu dwords\n", prim_walk, i, size >> 2, + radeon_object_size(track->arrays[i].robj) >> 2); + return -EINVAL; + } + } + break; + case 3: + size = track->vtx_size * nverts; + if (size != track->immd_dwords) { + DRM_ERROR("IMMD draw %u dwors but needs %lu dwords\n", + track->immd_dwords, size); + DRM_ERROR("VAP_VF_CNTL.NUM_VERTICES %u, VTX_SIZE %u\n", + nverts, track->vtx_size); + return -EINVAL; + } + break; + default: + DRM_ERROR("[drm] Invalid primitive walk %d for VAP_VF_CNTL\n", + prim_walk); + return -EINVAL; + } + return r100_cs_track_texture_check(rdev, track); +} + +void r100_cs_track_clear(struct radeon_device *rdev, struct r100_cs_track *track) +{ + unsigned i, face; + + if (rdev->family < CHIP_R300) { + track->num_cb = 1; + if (rdev->family <= CHIP_RS200) + track->num_texture = 3; + else + track->num_texture = 6; + track->maxy = 2048; + track->separate_cube = 1; + } else { + track->num_cb = 4; + track->num_texture = 16; + track->maxy = 4096; + track->separate_cube = 0; + } + + for (i = 0; i < track->num_cb; i++) { + track->cb[i].robj = NULL; + track->cb[i].pitch = 8192; + track->cb[i].cpp = 16; + track->cb[i].offset = 0; + } + track->z_enabled = true; + track->zb.robj = NULL; + track->zb.pitch = 8192; + track->zb.cpp = 4; + track->zb.offset = 0; + track->vtx_size = 0x7F; + track->immd_dwords = 0xFFFFFFFFUL; + track->num_arrays = 11; + track->max_indx = 0x00FFFFFFUL; + for (i = 0; i < track->num_arrays; i++) { + track->arrays[i].robj = NULL; + track->arrays[i].esize = 0x7F; + } + for (i = 0; i < track->num_texture; i++) { + track->textures[i].pitch = 16536; + track->textures[i].width = 16536; + track->textures[i].height = 16536; + track->textures[i].width_11 = 1 << 11; + track->textures[i].height_11 = 1 << 11; + track->textures[i].num_levels = 12; + if (rdev->family <= CHIP_RS200) { + track->textures[i].tex_coord_type = 0; + track->textures[i].txdepth = 0; + } else { + track->textures[i].txdepth = 16; + track->textures[i].tex_coord_type = 1; + } + track->textures[i].cpp = 64; + track->textures[i].robj = NULL; + /* CS IB emission code makes sure texture unit are disabled */ + track->textures[i].enabled = false; + track->textures[i].roundup_w = true; + track->textures[i].roundup_h = true; + if (track->separate_cube) + for (face = 0; face < 5; face++) { + track->textures[i].cube_info[face].robj = NULL; + track->textures[i].cube_info[face].width = 16536; + track->textures[i].cube_info[face].height = 16536; + track->textures[i].cube_info[face].offset = 0; + } + } +} + +int r100_ring_test(struct radeon_device *rdev) +{ + uint32_t scratch; + uint32_t tmp = 0; + unsigned i; + int r; + + r = radeon_scratch_get(rdev, &scratch); + if (r) { + DRM_ERROR("radeon: cp failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + r = radeon_ring_lock(rdev, 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); + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) { + break; + } + DRM_UDELAY(1); + } + if (i < rdev->usec_timeout) { + DRM_INFO("ring test succeeded in %d usecs\n", i); + } else { + DRM_ERROR("radeon: ring test failed (sracth(0x%04X)=0x%08X)\n", + scratch, tmp); + r = -EINVAL; + } + radeon_scratch_free(rdev, scratch); + return r; +} + +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); +} + +int r100_ib_test(struct radeon_device *rdev) +{ + struct radeon_ib *ib; + uint32_t scratch; + uint32_t tmp = 0; + unsigned i; + int r; + + r = radeon_scratch_get(rdev, &scratch); + if (r) { + DRM_ERROR("radeon: failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + r = radeon_ib_get(rdev, &ib); + if (r) { + return r; + } + ib->ptr[0] = PACKET0(scratch, 0); + ib->ptr[1] = 0xDEADBEEF; + ib->ptr[2] = PACKET2(0); + ib->ptr[3] = PACKET2(0); + ib->ptr[4] = PACKET2(0); + ib->ptr[5] = PACKET2(0); + ib->ptr[6] = PACKET2(0); + ib->ptr[7] = PACKET2(0); + ib->length_dw = 8; + r = radeon_ib_schedule(rdev, ib); + if (r) { + radeon_scratch_free(rdev, scratch); + radeon_ib_free(rdev, &ib); + return r; + } + r = radeon_fence_wait(ib->fence, false); + if (r) { + return r; + } + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) { + break; + } + DRM_UDELAY(1); + } + if (i < rdev->usec_timeout) { + DRM_INFO("ib test succeeded in %u usecs\n", i); + } else { + DRM_ERROR("radeon: ib test failed (sracth(0x%04X)=0x%08X)\n", + scratch, tmp); + r = -EINVAL; + } + radeon_scratch_free(rdev, scratch); + radeon_ib_free(rdev, &ib); + return r; +} + +void r100_ib_fini(struct radeon_device *rdev) +{ + radeon_ib_pool_fini(rdev); +} + +int r100_ib_init(struct radeon_device *rdev) +{ + int r; + + r = radeon_ib_pool_init(rdev); + if (r) { + dev_err(rdev->dev, "failled initializing IB pool (%d).\n", r); + r100_ib_fini(rdev); + return r; + } + r = r100_ib_test(rdev); + if (r) { + dev_err(rdev->dev, "failled testing IB (%d).\n", r); + r100_ib_fini(rdev); + return r; + } + return 0; +} + +void r100_mc_stop(struct radeon_device *rdev, struct r100_mc_save *save) +{ + /* Shutdown CP we shouldn't need to do that but better be safe than + * sorry + */ + rdev->cp.ready = false; + WREG32(R_000740_CP_CSQ_CNTL, 0); + + /* Save few CRTC registers */ + save->GENMO_WT = RREG32(R_0003C0_GENMO_WT); + save->CRTC_EXT_CNTL = RREG32(R_000054_CRTC_EXT_CNTL); + save->CRTC_GEN_CNTL = RREG32(R_000050_CRTC_GEN_CNTL); + save->CUR_OFFSET = RREG32(R_000260_CUR_OFFSET); + if (!(rdev->flags & RADEON_SINGLE_CRTC)) { + save->CRTC2_GEN_CNTL = RREG32(R_0003F8_CRTC2_GEN_CNTL); + save->CUR2_OFFSET = RREG32(R_000360_CUR2_OFFSET); + } + + /* Disable VGA aperture access */ + WREG32(R_0003C0_GENMO_WT, C_0003C0_VGA_RAM_EN & save->GENMO_WT); + /* Disable cursor, overlay, crtc */ + WREG32(R_000260_CUR_OFFSET, save->CUR_OFFSET | S_000260_CUR_LOCK(1)); + WREG32(R_000054_CRTC_EXT_CNTL, save->CRTC_EXT_CNTL | + S_000054_CRTC_DISPLAY_DIS(1)); + WREG32(R_000050_CRTC_GEN_CNTL, + (C_000050_CRTC_CUR_EN & save->CRTC_GEN_CNTL) | + S_000050_CRTC_DISP_REQ_EN_B(1)); + WREG32(R_000420_OV0_SCALE_CNTL, + C_000420_OV0_OVERLAY_EN & RREG32(R_000420_OV0_SCALE_CNTL)); + WREG32(R_000260_CUR_OFFSET, C_000260_CUR_LOCK & save->CUR_OFFSET); + if (!(rdev->flags & RADEON_SINGLE_CRTC)) { + WREG32(R_000360_CUR2_OFFSET, save->CUR2_OFFSET | + S_000360_CUR2_LOCK(1)); + WREG32(R_0003F8_CRTC2_GEN_CNTL, + (C_0003F8_CRTC2_CUR_EN & save->CRTC2_GEN_CNTL) | + S_0003F8_CRTC2_DISPLAY_DIS(1) | + S_0003F8_CRTC2_DISP_REQ_EN_B(1)); + WREG32(R_000360_CUR2_OFFSET, + C_000360_CUR2_LOCK & save->CUR2_OFFSET); + } +} + +void r100_mc_resume(struct radeon_device *rdev, struct r100_mc_save *save) +{ + /* Update base address for crtc */ + WREG32(R_00023C_DISPLAY_BASE_ADDR, rdev->mc.vram_location); + if (!(rdev->flags & RADEON_SINGLE_CRTC)) { + WREG32(R_00033C_CRTC2_DISPLAY_BASE_ADDR, + rdev->mc.vram_location); + } + /* Restore CRTC registers */ + WREG32(R_0003C0_GENMO_WT, save->GENMO_WT); + WREG32(R_000054_CRTC_EXT_CNTL, save->CRTC_EXT_CNTL); + WREG32(R_000050_CRTC_GEN_CNTL, save->CRTC_GEN_CNTL); + if (!(rdev->flags & RADEON_SINGLE_CRTC)) { + WREG32(R_0003F8_CRTC2_GEN_CNTL, save->CRTC2_GEN_CNTL); + } +}