diff options
Diffstat (limited to 'drivers/gpu/drm/nouveau')
47 files changed, 2556 insertions, 1193 deletions
diff --git a/drivers/gpu/drm/nouveau/Kconfig b/drivers/gpu/drm/nouveau/Kconfig index d823e6319516..1175429da102 100644 --- a/drivers/gpu/drm/nouveau/Kconfig +++ b/drivers/gpu/drm/nouveau/Kconfig @@ -30,7 +30,7 @@ config DRM_NOUVEAU_DEBUG via debugfs. menu "I2C encoder or helper chips" - depends on DRM + depends on DRM && DRM_KMS_HELPER && I2C config DRM_I2C_CH7006 tristate "Chrontel ch7006 TV encoder" diff --git a/drivers/gpu/drm/nouveau/Makefile b/drivers/gpu/drm/nouveau/Makefile index 1d90d4d0144f..48c290b5da8c 100644 --- a/drivers/gpu/drm/nouveau/Makefile +++ b/drivers/gpu/drm/nouveau/Makefile @@ -8,14 +8,15 @@ nouveau-y := nouveau_drv.o nouveau_state.o nouveau_channel.o nouveau_mem.o \ nouveau_sgdma.o nouveau_dma.o \ nouveau_bo.o nouveau_fence.o nouveau_gem.o nouveau_ttm.o \ nouveau_hw.o nouveau_calc.o nouveau_bios.o nouveau_i2c.o \ - nouveau_display.o nouveau_connector.o nouveau_fbcon.o \ - nouveau_dp.o \ + nouveau_display.o nouveau_connector.o nouveau_fbcon.o \ + nouveau_dp.o nouveau_grctx.o \ nv04_timer.o \ nv04_mc.o nv40_mc.o nv50_mc.o \ nv04_fb.o nv10_fb.o nv40_fb.o \ nv04_fifo.o nv10_fifo.o nv40_fifo.o nv50_fifo.o \ nv04_graph.o nv10_graph.o nv20_graph.o \ nv40_graph.o nv50_graph.o \ + nv40_grctx.o \ nv04_instmem.o nv50_instmem.o \ nv50_crtc.o nv50_dac.o nv50_sor.o \ nv50_cursor.o nv50_display.o nv50_fbcon.o \ diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.c b/drivers/gpu/drm/nouveau/nouveau_bios.c index 5eec5ed69489..ba143972769f 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bios.c +++ b/drivers/gpu/drm/nouveau/nouveau_bios.c @@ -181,43 +181,42 @@ struct methods { const char desc[8]; void (*loadbios)(struct drm_device *, uint8_t *); const bool rw; - int score; }; static struct methods nv04_methods[] = { { "PROM", load_vbios_prom, false }, { "PRAMIN", load_vbios_pramin, true }, { "PCIROM", load_vbios_pci, true }, - { } }; static struct methods nv50_methods[] = { { "PRAMIN", load_vbios_pramin, true }, { "PROM", load_vbios_prom, false }, { "PCIROM", load_vbios_pci, true }, - { } }; +#define METHODCNT 3 + static bool NVShadowVBIOS(struct drm_device *dev, uint8_t *data) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct methods *methods, *method; + struct methods *methods; + int i; int testscore = 3; + int scores[METHODCNT]; if (nouveau_vbios) { - method = nv04_methods; - while (method->loadbios) { - if (!strcasecmp(nouveau_vbios, method->desc)) + methods = nv04_methods; + for (i = 0; i < METHODCNT; i++) + if (!strcasecmp(nouveau_vbios, methods[i].desc)) break; - method++; - } - if (method->loadbios) { + if (i < METHODCNT) { NV_INFO(dev, "Attempting to use BIOS image from %s\n", - method->desc); + methods[i].desc); - method->loadbios(dev, data); - if (score_vbios(dev, data, method->rw)) + methods[i].loadbios(dev, data); + if (score_vbios(dev, data, methods[i].rw)) return true; } @@ -229,28 +228,24 @@ static bool NVShadowVBIOS(struct drm_device *dev, uint8_t *data) else methods = nv50_methods; - method = methods; - while (method->loadbios) { + for (i = 0; i < METHODCNT; i++) { NV_TRACE(dev, "Attempting to load BIOS image from %s\n", - method->desc); + methods[i].desc); data[0] = data[1] = 0; /* avoid reuse of previous image */ - method->loadbios(dev, data); - method->score = score_vbios(dev, data, method->rw); - if (method->score == testscore) + methods[i].loadbios(dev, data); + scores[i] = score_vbios(dev, data, methods[i].rw); + if (scores[i] == testscore) return true; - method++; } while (--testscore > 0) { - method = methods; - while (method->loadbios) { - if (method->score == testscore) { + for (i = 0; i < METHODCNT; i++) { + if (scores[i] == testscore) { NV_TRACE(dev, "Using BIOS image from %s\n", - method->desc); - method->loadbios(dev, data); + methods[i].desc); + methods[i].loadbios(dev, data); return true; } - method++; } } @@ -261,10 +256,7 @@ static bool NVShadowVBIOS(struct drm_device *dev, uint8_t *data) struct init_tbl_entry { char *name; uint8_t id; - int length; - int length_offset; - int length_multiplier; - bool (*handler)(struct nvbios *, uint16_t, struct init_exec *); + int (*handler)(struct nvbios *, uint16_t, struct init_exec *); }; struct bit_entry { @@ -820,7 +812,7 @@ static uint32_t get_tmds_index_reg(struct drm_device *dev, uint8_t mlv) } } -static bool +static int init_io_restrict_prog(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -852,9 +844,10 @@ init_io_restrict_prog(struct nvbios *bios, uint16_t offset, uint32_t reg = ROM32(bios->data[offset + 7]); uint8_t config; uint32_t configval; + int len = 11 + count * 4; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, " "Shift: 0x%02X, Count: 0x%02X, Reg: 0x%08X\n", @@ -865,7 +858,7 @@ init_io_restrict_prog(struct nvbios *bios, uint16_t offset, NV_ERROR(bios->dev, "0x%04X: Config 0x%02X exceeds maximal bound 0x%02X\n", offset, config, count); - return false; + return 0; } configval = ROM32(bios->data[offset + 11 + config * 4]); @@ -874,10 +867,10 @@ init_io_restrict_prog(struct nvbios *bios, uint16_t offset, bios_wr32(bios, reg, configval); - return true; + return len; } -static bool +static int init_repeat(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -912,10 +905,10 @@ init_repeat(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) iexec->repeat = false; - return true; + return 2; } -static bool +static int init_io_restrict_pll(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -951,9 +944,10 @@ init_io_restrict_pll(struct nvbios *bios, uint16_t offset, uint32_t reg = ROM32(bios->data[offset + 8]); uint8_t config; uint16_t freq; + int len = 12 + count * 2; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, " "Shift: 0x%02X, IO Flag Condition: 0x%02X, " @@ -966,7 +960,7 @@ init_io_restrict_pll(struct nvbios *bios, uint16_t offset, NV_ERROR(bios->dev, "0x%04X: Config 0x%02X exceeds maximal bound 0x%02X\n", offset, config, count); - return false; + return 0; } freq = ROM16(bios->data[offset + 12 + config * 2]); @@ -986,10 +980,10 @@ init_io_restrict_pll(struct nvbios *bios, uint16_t offset, setPLL(bios, reg, freq * 10); - return true; + return len; } -static bool +static int init_end_repeat(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1007,12 +1001,12 @@ init_end_repeat(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) * we're not in repeat mode */ if (iexec->repeat) - return false; + return 0; - return true; + return 1; } -static bool +static int init_copy(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1041,7 +1035,7 @@ init_copy(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t crtcdata; if (!iexec->execute) - return true; + return 11; BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Shift: 0x%02X, SrcMask: 0x%02X, " "Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X\n", @@ -1060,10 +1054,10 @@ init_copy(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) crtcdata |= (uint8_t)data; bios_idxprt_wr(bios, crtcport, crtcindex, crtcdata); - return true; + return 11; } -static bool +static int init_not(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1079,10 +1073,10 @@ init_not(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) BIOSLOG(bios, "0x%04X: ------ Executing following commands ------\n", offset); iexec->execute = !iexec->execute; - return true; + return 1; } -static bool +static int init_io_flag_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1100,7 +1094,7 @@ init_io_flag_condition(struct nvbios *bios, uint16_t offset, uint8_t cond = bios->data[offset + 1]; if (!iexec->execute) - return true; + return 2; if (io_flag_condition_met(bios, offset, cond)) BIOSLOG(bios, "0x%04X: Condition fulfilled -- continuing to execute\n", offset); @@ -1109,10 +1103,10 @@ init_io_flag_condition(struct nvbios *bios, uint16_t offset, iexec->execute = false; } - return true; + return 2; } -static bool +static int init_idx_addr_latched(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1140,11 +1134,12 @@ init_idx_addr_latched(struct nvbios *bios, uint16_t offset, uint32_t mask = ROM32(bios->data[offset + 9]); uint32_t data = ROM32(bios->data[offset + 13]); uint8_t count = bios->data[offset + 17]; + int len = 18 + count * 2; uint32_t value; int i; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: ControlReg: 0x%08X, DataReg: 0x%08X, " "Mask: 0x%08X, Data: 0x%08X, Count: 0x%02X\n", @@ -1164,10 +1159,10 @@ init_idx_addr_latched(struct nvbios *bios, uint16_t offset, bios_wr32(bios, controlreg, value); } - return true; + return len; } -static bool +static int init_io_restrict_pll2(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1196,25 +1191,26 @@ init_io_restrict_pll2(struct nvbios *bios, uint16_t offset, uint8_t shift = bios->data[offset + 5]; uint8_t count = bios->data[offset + 6]; uint32_t reg = ROM32(bios->data[offset + 7]); + int len = 11 + count * 4; uint8_t config; uint32_t freq; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, " "Shift: 0x%02X, Count: 0x%02X, Reg: 0x%08X\n", offset, crtcport, crtcindex, mask, shift, count, reg); if (!reg) - return true; + return len; config = (bios_idxprt_rd(bios, crtcport, crtcindex) & mask) >> shift; if (config > count) { NV_ERROR(bios->dev, "0x%04X: Config 0x%02X exceeds maximal bound 0x%02X\n", offset, config, count); - return false; + return 0; } freq = ROM32(bios->data[offset + 11 + config * 4]); @@ -1224,10 +1220,10 @@ init_io_restrict_pll2(struct nvbios *bios, uint16_t offset, setPLL(bios, reg, freq); - return true; + return len; } -static bool +static int init_pll2(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1244,16 +1240,16 @@ init_pll2(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint32_t freq = ROM32(bios->data[offset + 5]); if (!iexec->execute) - return true; + return 9; BIOSLOG(bios, "0x%04X: Reg: 0x%04X, Freq: %dkHz\n", offset, reg, freq); setPLL(bios, reg, freq); - return true; + return 9; } -static bool +static int init_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1277,12 +1273,13 @@ init_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t i2c_index = bios->data[offset + 1]; uint8_t i2c_address = bios->data[offset + 2]; uint8_t count = bios->data[offset + 3]; + int len = 4 + count * 3; struct nouveau_i2c_chan *chan; struct i2c_msg msg; int i; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: DCBI2CIndex: 0x%02X, I2CAddress: 0x%02X, " "Count: 0x%02X\n", @@ -1290,7 +1287,7 @@ init_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) chan = init_i2c_device_find(bios->dev, i2c_index); if (!chan) - return false; + return 0; for (i = 0; i < count; i++) { uint8_t i2c_reg = bios->data[offset + 4 + i * 3]; @@ -1303,7 +1300,7 @@ init_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) msg.len = 1; msg.buf = &value; if (i2c_transfer(&chan->adapter, &msg, 1) != 1) - return false; + return 0; BIOSLOG(bios, "0x%04X: I2CReg: 0x%02X, Value: 0x%02X, " "Mask: 0x%02X, Data: 0x%02X\n", @@ -1317,14 +1314,14 @@ init_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) msg.len = 1; msg.buf = &value; if (i2c_transfer(&chan->adapter, &msg, 1) != 1) - return false; + return 0; } } - return true; + return len; } -static bool +static int init_zm_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1346,12 +1343,13 @@ init_zm_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t i2c_index = bios->data[offset + 1]; uint8_t i2c_address = bios->data[offset + 2]; uint8_t count = bios->data[offset + 3]; + int len = 4 + count * 2; struct nouveau_i2c_chan *chan; struct i2c_msg msg; int i; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: DCBI2CIndex: 0x%02X, I2CAddress: 0x%02X, " "Count: 0x%02X\n", @@ -1359,7 +1357,7 @@ init_zm_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) chan = init_i2c_device_find(bios->dev, i2c_index); if (!chan) - return false; + return 0; for (i = 0; i < count; i++) { uint8_t i2c_reg = bios->data[offset + 4 + i * 2]; @@ -1374,14 +1372,14 @@ init_zm_i2c_byte(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) msg.len = 1; msg.buf = &data; if (i2c_transfer(&chan->adapter, &msg, 1) != 1) - return false; + return 0; } } - return true; + return len; } -static bool +static int init_zm_i2c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1401,13 +1399,14 @@ init_zm_i2c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t i2c_index = bios->data[offset + 1]; uint8_t i2c_address = bios->data[offset + 2]; uint8_t count = bios->data[offset + 3]; + int len = 4 + count; struct nouveau_i2c_chan *chan; struct i2c_msg msg; uint8_t data[256]; int i; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: DCBI2CIndex: 0x%02X, I2CAddress: 0x%02X, " "Count: 0x%02X\n", @@ -1415,7 +1414,7 @@ init_zm_i2c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) chan = init_i2c_device_find(bios->dev, i2c_index); if (!chan) - return false; + return 0; for (i = 0; i < count; i++) { data[i] = bios->data[offset + 4 + i]; @@ -1429,13 +1428,13 @@ init_zm_i2c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) msg.len = count; msg.buf = data; if (i2c_transfer(&chan->adapter, &msg, 1) != 1) - return false; + return 0; } - return true; + return len; } -static bool +static int init_tmds(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1460,7 +1459,7 @@ init_tmds(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint32_t reg, value; if (!iexec->execute) - return true; + return 5; BIOSLOG(bios, "0x%04X: MagicLookupValue: 0x%02X, TMDSAddr: 0x%02X, " "Mask: 0x%02X, Data: 0x%02X\n", @@ -1468,7 +1467,7 @@ init_tmds(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) reg = get_tmds_index_reg(bios->dev, mlv); if (!reg) - return false; + return 0; bios_wr32(bios, reg, tmdsaddr | NV_PRAMDAC_FP_TMDS_CONTROL_WRITE_DISABLE); @@ -1476,10 +1475,10 @@ init_tmds(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) bios_wr32(bios, reg + 4, value); bios_wr32(bios, reg, tmdsaddr); - return true; + return 5; } -static bool +static int init_zm_tmds_group(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1500,18 +1499,19 @@ init_zm_tmds_group(struct nvbios *bios, uint16_t offset, uint8_t mlv = bios->data[offset + 1]; uint8_t count = bios->data[offset + 2]; + int len = 3 + count * 2; uint32_t reg; int i; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: MagicLookupValue: 0x%02X, Count: 0x%02X\n", offset, mlv, count); reg = get_tmds_index_reg(bios->dev, mlv); if (!reg) - return false; + return 0; for (i = 0; i < count; i++) { uint8_t tmdsaddr = bios->data[offset + 3 + i * 2]; @@ -1521,10 +1521,10 @@ init_zm_tmds_group(struct nvbios *bios, uint16_t offset, bios_wr32(bios, reg, tmdsaddr); } - return true; + return len; } -static bool +static int init_cr_idx_adr_latch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1547,11 +1547,12 @@ init_cr_idx_adr_latch(struct nvbios *bios, uint16_t offset, uint8_t crtcindex2 = bios->data[offset + 2]; uint8_t baseaddr = bios->data[offset + 3]; uint8_t count = bios->data[offset + 4]; + int len = 5 + count; uint8_t oldaddr, data; int i; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: Index1: 0x%02X, Index2: 0x%02X, " "BaseAddr: 0x%02X, Count: 0x%02X\n", @@ -1568,10 +1569,10 @@ init_cr_idx_adr_latch(struct nvbios *bios, uint16_t offset, bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex1, oldaddr); - return true; + return len; } -static bool +static int init_cr(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1592,7 +1593,7 @@ init_cr(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t value; if (!iexec->execute) - return true; + return 4; BIOSLOG(bios, "0x%04X: Index: 0x%02X, Mask: 0x%02X, Data: 0x%02X\n", offset, crtcindex, mask, data); @@ -1601,10 +1602,10 @@ init_cr(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) value |= data; bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex, value); - return true; + return 4; } -static bool +static int init_zm_cr(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1621,14 +1622,14 @@ init_zm_cr(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t data = bios->data[offset + 2]; if (!iexec->execute) - return true; + return 3; bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, crtcindex, data); - return true; + return 3; } -static bool +static int init_zm_cr_group(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1645,18 +1646,19 @@ init_zm_cr_group(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) */ uint8_t count = bios->data[offset + 1]; + int len = 2 + count * 2; int i; if (!iexec->execute) - return true; + return len; for (i = 0; i < count; i++) init_zm_cr(bios, offset + 2 + 2 * i - 1, iexec); - return true; + return len; } -static bool +static int init_condition_time(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1680,7 +1682,7 @@ init_condition_time(struct nvbios *bios, uint16_t offset, unsigned cnt; if (!iexec->execute) - return true; + return 3; if (retries > 100) retries = 100; @@ -1711,10 +1713,10 @@ init_condition_time(struct nvbios *bios, uint16_t offset, iexec->execute = false; } - return true; + return 3; } -static bool +static int init_zm_reg_sequence(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1734,10 +1736,11 @@ init_zm_reg_sequence(struct nvbios *bios, uint16_t offset, uint32_t basereg = ROM32(bios->data[offset + 1]); uint32_t count = bios->data[offset + 5]; + int len = 6 + count * 4; int i; if (!iexec->execute) - return true; + return len; BIOSLOG(bios, "0x%04X: BaseReg: 0x%08X, Count: 0x%02X\n", offset, basereg, count); @@ -1749,10 +1752,10 @@ init_zm_reg_sequence(struct nvbios *bios, uint16_t offset, bios_wr32(bios, reg, data); } - return true; + return len; } -static bool +static int init_sub_direct(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1768,7 +1771,7 @@ init_sub_direct(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint16_t sub_offset = ROM16(bios->data[offset + 1]); if (!iexec->execute) - return true; + return 3; BIOSLOG(bios, "0x%04X: Executing subroutine at 0x%04X\n", offset, sub_offset); @@ -1777,10 +1780,10 @@ init_sub_direct(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) BIOSLOG(bios, "0x%04X: End of 0x%04X subroutine\n", offset, sub_offset); - return true; + return 3; } -static bool +static int init_copy_nv_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1808,7 +1811,7 @@ init_copy_nv_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint32_t srcvalue, dstvalue; if (!iexec->execute) - return true; + return 22; BIOSLOG(bios, "0x%04X: SrcReg: 0x%08X, Shift: 0x%02X, SrcMask: 0x%08X, " "Xor: 0x%08X, DstReg: 0x%08X, DstMask: 0x%08X\n", @@ -1827,10 +1830,10 @@ init_copy_nv_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) bios_wr32(bios, dstreg, dstvalue | srcvalue); - return true; + return 22; } -static bool +static int init_zm_index_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1848,14 +1851,14 @@ init_zm_index_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t data = bios->data[offset + 4]; if (!iexec->execute) - return true; + return 5; bios_idxprt_wr(bios, crtcport, crtcindex, data); - return true; + return 5; } -static bool +static int init_compute_mem(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1904,7 +1907,7 @@ init_compute_mem(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) struct drm_nouveau_private *dev_priv = bios->dev->dev_private; if (dev_priv->card_type >= NV_50) - return true; + return 1; /* * On every card I've seen, this step gets done for us earlier in @@ -1922,10 +1925,10 @@ init_compute_mem(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) /* write back the saved configuration value */ bios_wr32(bios, NV_PFB_CFG0, bios->state.saved_nv_pfb_cfg0); - return true; + return 1; } -static bool +static int init_reset(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -1959,10 +1962,10 @@ init_reset(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) pci_nv_20 &= ~NV_PBUS_PCI_NV_20_ROM_SHADOW_ENABLED; /* 0xfffffffe */ bios_wr32(bios, NV_PBUS_PCI_NV_20, pci_nv_20); - return true; + return 13; } -static bool +static int init_configure_mem(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -1983,7 +1986,7 @@ init_configure_mem(struct nvbios *bios, uint16_t offset, uint32_t reg, data; if (bios->major_version > 2) - return false; + return 0; bios_idxprt_wr(bios, NV_VIO_SRX, NV_VIO_SR_CLOCK_INDEX, bios_idxprt_rd( bios, NV_VIO_SRX, NV_VIO_SR_CLOCK_INDEX) | 0x20); @@ -2015,10 +2018,10 @@ init_configure_mem(struct nvbios *bios, uint16_t offset, bios_wr32(bios, reg, data); } - return true; + return 1; } -static bool +static int init_configure_clk(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -2038,7 +2041,7 @@ init_configure_clk(struct nvbios *bios, uint16_t offset, int clock; if (bios->major_version > 2) - return false; + return 0; clock = ROM16(bios->data[meminitoffs + 4]) * 10; setPLL(bios, NV_PRAMDAC_NVPLL_COEFF, clock); @@ -2048,10 +2051,10 @@ init_configure_clk(struct nvbios *bios, uint16_t offset, clock *= 2; setPLL(bios, NV_PRAMDAC_MPLL_COEFF, clock); - return true; + return 1; } -static bool +static int init_configure_preinit(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -2071,15 +2074,15 @@ init_configure_preinit(struct nvbios *bios, uint16_t offset, uint8_t cr3c = ((straps << 2) & 0xf0) | (straps & (1 << 6)); if (bios->major_version > 2) - return false; + return 0; bios_idxprt_wr(bios, NV_CIO_CRX__COLOR, NV_CIO_CRE_SCRATCH4__INDEX, cr3c); - return true; + return 1; } -static bool +static int init_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2099,7 +2102,7 @@ init_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t data = bios->data[offset + 4]; if (!iexec->execute) - return true; + return 5; BIOSLOG(bios, "0x%04X: Port: 0x%04X, Mask: 0x%02X, Data: 0x%02X\n", offset, crtcport, mask, data); @@ -2158,15 +2161,15 @@ init_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) for (i = 0; i < 2; i++) bios_wr32(bios, 0x614108 + (i*0x800), bios_rd32( bios, 0x614108 + (i*0x800)) & 0x0fffffff); - return true; + return 5; } bios_port_wr(bios, crtcport, (bios_port_rd(bios, crtcport) & mask) | data); - return true; + return 5; } -static bool +static int init_sub(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2181,7 +2184,7 @@ init_sub(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t sub = bios->data[offset + 1]; if (!iexec->execute) - return true; + return 2; BIOSLOG(bios, "0x%04X: Calling script %d\n", offset, sub); @@ -2191,10 +2194,10 @@ init_sub(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) BIOSLOG(bios, "0x%04X: End of script %d\n", offset, sub); - return true; + return 2; } -static bool +static int init_ram_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -2215,7 +2218,7 @@ init_ram_condition(struct nvbios *bios, uint16_t offset, uint8_t data; if (!iexec->execute) - return true; + return 3; data = bios_rd32(bios, NV_PFB_BOOT_0) & mask; @@ -2229,10 +2232,10 @@ init_ram_condition(struct nvbios *bios, uint16_t offset, iexec->execute = false; } - return true; + return 3; } -static bool +static int init_nv_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2251,17 +2254,17 @@ init_nv_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint32_t data = ROM32(bios->data[offset + 9]); if (!iexec->execute) - return true; + return 13; BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Mask: 0x%08X, Data: 0x%08X\n", offset, reg, mask, data); bios_wr32(bios, reg, (bios_rd32(bios, reg) & mask) | data); - return true; + return 13; } -static bool +static int init_macro(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2285,7 +2288,7 @@ init_macro(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) int i; if (!iexec->execute) - return true; + return 2; BIOSLOG(bios, "0x%04X: Macro: 0x%02X, MacroTableIndex: 0x%02X, " "Count: 0x%02X\n", @@ -2300,10 +2303,10 @@ init_macro(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) bios_wr32(bios, reg, data); } - return true; + return 2; } -static bool +static int init_done(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2315,10 +2318,10 @@ init_done(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) */ /* mild retval abuse to stop parsing this table */ - return false; + return 0; } -static bool +static int init_resume(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2330,15 +2333,15 @@ init_resume(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) */ if (iexec->execute) - return true; + return 1; iexec->execute = true; BIOSLOG(bios, "0x%04X: ---- Executing following commands ----\n", offset); - return true; + return 1; } -static bool +static int init_time(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2353,7 +2356,7 @@ init_time(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) unsigned time = ROM16(bios->data[offset + 1]); if (!iexec->execute) - return true; + return 3; BIOSLOG(bios, "0x%04X: Sleeping for 0x%04X microseconds\n", offset, time); @@ -2363,10 +2366,10 @@ init_time(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) else msleep((time + 900) / 1000); - return true; + return 3; } -static bool +static int init_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2383,7 +2386,7 @@ init_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t cond = bios->data[offset + 1]; if (!iexec->execute) - return true; + return 2; BIOSLOG(bios, "0x%04X: Condition: 0x%02X\n", offset, cond); @@ -2394,10 +2397,10 @@ init_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) iexec->execute = false; } - return true; + return 2; } -static bool +static int init_io_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2414,7 +2417,7 @@ init_io_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t cond = bios->data[offset + 1]; if (!iexec->execute) - return true; + return 2; BIOSLOG(bios, "0x%04X: IO condition: 0x%02X\n", offset, cond); @@ -2425,10 +2428,10 @@ init_io_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) iexec->execute = false; } - return true; + return 2; } -static bool +static int init_index_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2451,7 +2454,7 @@ init_index_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint8_t value; if (!iexec->execute) - return true; + return 6; BIOSLOG(bios, "0x%04X: Port: 0x%04X, Index: 0x%02X, Mask: 0x%02X, " "Data: 0x%02X\n", @@ -2460,10 +2463,10 @@ init_index_io(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) value = (bios_idxprt_rd(bios, crtcport, crtcindex) & mask) | data; bios_idxprt_wr(bios, crtcport, crtcindex, value); - return true; + return 6; } -static bool +static int init_pll(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2481,16 +2484,16 @@ init_pll(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint16_t freq = ROM16(bios->data[offset + 5]); if (!iexec->execute) - return true; + return 7; BIOSLOG(bios, "0x%04X: Reg: 0x%08X, Freq: %d0kHz\n", offset, reg, freq); setPLL(bios, reg, freq * 10); - return true; + return 7; } -static bool +static int init_zm_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2507,17 +2510,17 @@ init_zm_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint32_t value = ROM32(bios->data[offset + 5]); if (!iexec->execute) - return true; + return 9; if (reg == 0x000200) value |= 1; bios_wr32(bios, reg, value); - return true; + return 9; } -static bool +static int init_ram_restrict_pll(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -2543,14 +2546,15 @@ init_ram_restrict_pll(struct nvbios *bios, uint16_t offset, uint8_t type = bios->data[offset + 1]; uint32_t freq = ROM32(bios->data[offset + 2 + (index * 4)]); uint8_t *pll_limits = &bios->data[bios->pll_limit_tbl_ptr], *entry; + int len = 2 + bios->ram_restrict_group_count * 4; int i; if (!iexec->execute) - return true; + return len; if (!bios->pll_limit_tbl_ptr || (pll_limits[0] & 0xf0) != 0x30) { NV_ERROR(dev, "PLL limits table not version 3.x\n"); - return true; /* deliberate, allow default clocks to remain */ + return len; /* deliberate, allow default clocks to remain */ } entry = pll_limits + pll_limits[1]; @@ -2563,15 +2567,15 @@ init_ram_restrict_pll(struct nvbios *bios, uint16_t offset, offset, type, reg, freq); setPLL(bios, reg, freq); - return true; + return len; } } NV_ERROR(dev, "PLL type 0x%02x not found in PLL limits table", type); - return true; + return len; } -static bool +static int init_8c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2581,10 +2585,10 @@ init_8c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) * */ - return true; + return 1; } -static bool +static int init_8d(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2594,10 +2598,10 @@ init_8d(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) * */ - return true; + return 1; } -static bool +static int init_gpio(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2615,14 +2619,17 @@ init_gpio(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) const uint8_t *gpio_entry; int i; + if (!iexec->execute) + return 1; + if (bios->bdcb.version != 0x40) { NV_ERROR(bios->dev, "DCB table not version 4.0\n"); - return false; + return 0; } if (!bios->bdcb.gpio_table_ptr) { NV_WARN(bios->dev, "Invalid pointer to INIT_8E table\n"); - return false; + return 0; } gpio_entry = gpio_table + gpio_table[1]; @@ -2660,13 +2667,10 @@ init_gpio(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) bios_wr32(bios, r, v); } - return true; + return 1; } -/* hack to avoid moving the itbl_entry array before this function */ -int init_ram_restrict_zm_reg_group_blocklen; - -static bool +static int init_ram_restrict_zm_reg_group(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -2692,21 +2696,21 @@ init_ram_restrict_zm_reg_group(struct nvbios *bios, uint16_t offset, uint8_t regincrement = bios->data[offset + 5]; uint8_t count = bios->data[offset + 6]; uint32_t strap_ramcfg, data; - uint16_t blocklen; + /* previously set by 'M' BIT table */ + uint16_t blocklen = bios->ram_restrict_group_count * 4; + int len = 7 + count * blocklen; uint8_t index; int i; - /* previously set by 'M' BIT table */ - blocklen = init_ram_restrict_zm_reg_group_blocklen; if (!iexec->execute) - return true; + return len; if (!blocklen) { NV_ERROR(bios->dev, "0x%04X: Zero block length - has the M table " "been parsed?\n", offset); - return false; + return 0; } strap_ramcfg = (bios_rd32(bios, NV_PEXTDEV_BOOT_0) >> 2) & 0xf; @@ -2724,10 +2728,10 @@ init_ram_restrict_zm_reg_group(struct nvbios *bios, uint16_t offset, reg += regincrement; } - return true; + return len; } -static bool +static int init_copy_zm_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2744,14 +2748,14 @@ init_copy_zm_reg(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) uint32_t dstreg = ROM32(bios->data[offset + 5]); if (!iexec->execute) - return true; + return 9; bios_wr32(bios, dstreg, bios_rd32(bios, srcreg)); - return true; + return 9; } -static bool +static int init_zm_reg_group_addr_latched(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -2769,20 +2773,21 @@ init_zm_reg_group_addr_latched(struct nvbios *bios, uint16_t offset, uint32_t reg = ROM32(bios->data[offset + 1]); uint8_t count = bios->data[offset + 5]; + int len = 6 + count * 4; int i; if (!iexec->execute) - return true; + return len; for (i = 0; i < count; i++) { uint32_t data = ROM32(bios->data[offset + 6 + 4 * i]); bios_wr32(bios, reg, data); } - return true; + return len; } -static bool +static int init_reserved(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2793,10 +2798,10 @@ init_reserved(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) * Seemingly does nothing */ - return true; + return 1; } -static bool +static int init_96(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2829,13 +2834,13 @@ init_96(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) val <<= bios->data[offset + 16]; if (!iexec->execute) - return true; + return 17; bios_wr32(bios, reg, (bios_rd32(bios, reg) & mask) | val); - return true; + return 17; } -static bool +static int init_97(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2859,13 +2864,13 @@ init_97(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) val = (val & mask) | ((val + add) & ~mask); if (!iexec->execute) - return true; + return 13; bios_wr32(bios, reg, val); - return true; + return 13; } -static bool +static int init_auxch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2883,32 +2888,33 @@ init_auxch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) struct drm_device *dev = bios->dev; struct nouveau_i2c_chan *auxch; uint32_t addr = ROM32(bios->data[offset + 1]); - uint8_t len = bios->data[offset + 5]; + uint8_t count = bios->data[offset + 5]; + int len = 6 + count * 2; int ret, i; if (!bios->display.output) { NV_ERROR(dev, "INIT_AUXCH: no active output\n"); - return false; + return 0; } auxch = init_i2c_device_find(dev, bios->display.output->i2c_index); if (!auxch) { NV_ERROR(dev, "INIT_AUXCH: couldn't get auxch %d\n", bios->display.output->i2c_index); - return false; + return 0; } if (!iexec->execute) - return true; + return len; offset += 6; - for (i = 0; i < len; i++, offset += 2) { + for (i = 0; i < count; i++, offset += 2) { uint8_t data; ret = nouveau_dp_auxch(auxch, 9, addr, &data, 1); if (ret) { NV_ERROR(dev, "INIT_AUXCH: rd auxch fail %d\n", ret); - return false; + return 0; } data &= bios->data[offset + 0]; @@ -2917,14 +2923,14 @@ init_auxch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) ret = nouveau_dp_auxch(auxch, 8, addr, &data, 1); if (ret) { NV_ERROR(dev, "INIT_AUXCH: wr auxch fail %d\n", ret); - return false; + return 0; } } - return true; + return len; } -static bool +static int init_zm_auxch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { /* @@ -2941,106 +2947,99 @@ init_zm_auxch(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) struct drm_device *dev = bios->dev; struct nouveau_i2c_chan *auxch; uint32_t addr = ROM32(bios->data[offset + 1]); - uint8_t len = bios->data[offset + 5]; + uint8_t count = bios->data[offset + 5]; + int len = 6 + count; int ret, i; if (!bios->display.output) { NV_ERROR(dev, "INIT_ZM_AUXCH: no active output\n"); - return false; + return 0; } auxch = init_i2c_device_find(dev, bios->display.output->i2c_index); if (!auxch) { NV_ERROR(dev, "INIT_ZM_AUXCH: couldn't get auxch %d\n", bios->display.output->i2c_index); - return false; + return 0; } if (!iexec->execute) - return true; + return len; offset += 6; - for (i = 0; i < len; i++, offset++) { + for (i = 0; i < count; i++, offset++) { ret = nouveau_dp_auxch(auxch, 8, addr, &bios->data[offset], 1); if (ret) { NV_ERROR(dev, "INIT_ZM_AUXCH: wr auxch fail %d\n", ret); - return false; + return 0; } } - return true; + return len; } static struct init_tbl_entry itbl_entry[] = { /* command name , id , length , offset , mult , command handler */ /* INIT_PROG (0x31, 15, 10, 4) removed due to no example of use */ - { "INIT_IO_RESTRICT_PROG" , 0x32, 11 , 6 , 4 , init_io_restrict_prog }, - { "INIT_REPEAT" , 0x33, 2 , 0 , 0 , init_repeat }, - { "INIT_IO_RESTRICT_PLL" , 0x34, 12 , 7 , 2 , init_io_restrict_pll }, - { "INIT_END_REPEAT" , 0x36, 1 , 0 , 0 , init_end_repeat }, - { "INIT_COPY" , 0x37, 11 , 0 , 0 , init_copy }, - { "INIT_NOT" , 0x38, 1 , 0 , 0 , init_not }, - { "INIT_IO_FLAG_CONDITION" , 0x39, 2 , 0 , 0 , init_io_flag_condition }, - { "INIT_INDEX_ADDRESS_LATCHED" , 0x49, 18 , 17 , 2 , init_idx_addr_latched }, - { "INIT_IO_RESTRICT_PLL2" , 0x4A, 11 , 6 , 4 , init_io_restrict_pll2 }, - { "INIT_PLL2" , 0x4B, 9 , 0 , 0 , init_pll2 }, - { "INIT_I2C_BYTE" , 0x4C, 4 , 3 , 3 , init_i2c_byte }, - { "INIT_ZM_I2C_BYTE" , 0x4D, 4 , 3 , 2 , init_zm_i2c_byte }, - { "INIT_ZM_I2C" , 0x4E, 4 , 3 , 1 , init_zm_i2c }, - { "INIT_TMDS" , 0x4F, 5 , 0 , 0 , init_tmds }, - { "INIT_ZM_TMDS_GROUP" , 0x50, 3 , 2 , 2 , init_zm_tmds_group }, - { "INIT_CR_INDEX_ADDRESS_LATCHED" , 0x51, 5 , 4 , 1 , init_cr_idx_adr_latch }, - { "INIT_CR" , 0x52, 4 , 0 , 0 , init_cr }, - { "INIT_ZM_CR" , 0x53, 3 , 0 , 0 , init_zm_cr }, - { "INIT_ZM_CR_GROUP" , 0x54, 2 , 1 , 2 , init_zm_cr_group }, - { "INIT_CONDITION_TIME" , 0x56, 3 , 0 , 0 , init_condition_time }, - { "INIT_ZM_REG_SEQUENCE" , 0x58, 6 , 5 , 4 , init_zm_reg_sequence }, + { "INIT_IO_RESTRICT_PROG" , 0x32, init_io_restrict_prog }, + { "INIT_REPEAT" , 0x33, init_repeat }, + { "INIT_IO_RESTRICT_PLL" , 0x34, init_io_restrict_pll }, + { "INIT_END_REPEAT" , 0x36, init_end_repeat }, + { "INIT_COPY" , 0x37, init_copy }, + { "INIT_NOT" , 0x38, init_not }, + { "INIT_IO_FLAG_CONDITION" , 0x39, init_io_flag_condition }, + { "INIT_INDEX_ADDRESS_LATCHED" , 0x49, init_idx_addr_latched }, + { "INIT_IO_RESTRICT_PLL2" , 0x4A, init_io_restrict_pll2 }, + { "INIT_PLL2" , 0x4B, init_pll2 }, + { "INIT_I2C_BYTE" , 0x4C, init_i2c_byte }, + { "INIT_ZM_I2C_BYTE" , 0x4D, init_zm_i2c_byte }, + { "INIT_ZM_I2C" , 0x4E, init_zm_i2c }, + { "INIT_TMDS" , 0x4F, init_tmds }, + { "INIT_ZM_TMDS_GROUP" , 0x50, init_zm_tmds_group }, + { "INIT_CR_INDEX_ADDRESS_LATCHED" , 0x51, init_cr_idx_adr_latch }, + { "INIT_CR" , 0x52, init_cr }, + { "INIT_ZM_CR" , 0x53, init_zm_cr }, + { "INIT_ZM_CR_GROUP" , 0x54, init_zm_cr_group }, + { "INIT_CONDITION_TIME" , 0x56, init_condition_time }, + { "INIT_ZM_REG_SEQUENCE" , 0x58, init_zm_reg_sequence }, /* INIT_INDIRECT_REG (0x5A, 7, 0, 0) removed due to no example of use */ - { "INIT_SUB_DIRECT" , 0x5B, 3 , 0 , 0 , init_sub_direct }, - { "INIT_COPY_NV_REG" , 0x5F, 22 , 0 , 0 , init_copy_nv_reg }, - { "INIT_ZM_INDEX_IO" , 0x62, 5 , 0 , 0 , init_zm_index_io }, - { "INIT_COMPUTE_MEM" , 0x63, 1 , 0 , 0 , init_compute_mem }, - { "INIT_RESET" , 0x65, 13 , 0 , 0 , init_reset }, - { "INIT_CONFIGURE_MEM" , 0x66, 1 , 0 , 0 , init_configure_mem }, - { "INIT_CONFIGURE_CLK" , 0x67, 1 , 0 , 0 , init_configure_clk }, - { "INIT_CONFIGURE_PREINIT" , 0x68, 1 , 0 , 0 , init_configure_preinit }, - { "INIT_IO" , 0x69, 5 , 0 , 0 , init_io }, - { "INIT_SUB" , 0x6B, 2 , 0 , 0 , init_sub }, - { "INIT_RAM_CONDITION" , 0x6D, 3 , 0 , 0 , init_ram_condition }, - { "INIT_NV_REG" , 0x6E, 13 , 0 , 0 , init_nv_reg }, - { "INIT_MACRO" , 0x6F, 2 , 0 , 0 , init_macro }, - { "INIT_DONE" , 0x71, 1 , 0 , 0 , init_done }, - { "INIT_RESUME" , 0x72, 1 , 0 , 0 , init_resume }, + { "INIT_SUB_DIRECT" , 0x5B, init_sub_direct }, + { "INIT_COPY_NV_REG" , 0x5F, init_copy_nv_reg }, + { "INIT_ZM_INDEX_IO" , 0x62, init_zm_index_io }, + { "INIT_COMPUTE_MEM" , 0x63, init_compute_mem }, + { "INIT_RESET" , 0x65, init_reset }, + { "INIT_CONFIGURE_MEM" , 0x66, init_configure_mem }, + { "INIT_CONFIGURE_CLK" , 0x67, init_configure_clk }, + { "INIT_CONFIGURE_PREINIT" , 0x68, init_configure_preinit }, + { "INIT_IO" , 0x69, init_io }, + { "INIT_SUB" , 0x6B, init_sub }, + { "INIT_RAM_CONDITION" , 0x6D, init_ram_condition }, + { "INIT_NV_REG" , 0x6E, init_nv_reg }, + { "INIT_MACRO" , 0x6F, init_macro }, + { "INIT_DONE" , 0x71, init_done }, + { "INIT_RESUME" , 0x72, init_resume }, /* INIT_RAM_CONDITION2 (0x73, 9, 0, 0) removed due to no example of use */ - { "INIT_TIME" , 0x74, 3 , 0 , 0 , init_time }, - { "INIT_CONDITION" , 0x75, 2 , 0 , 0 , init_condition }, - { "INIT_IO_CONDITION" , 0x76, 2 , 0 , 0 , init_io_condition }, - { "INIT_INDEX_IO" , 0x78, 6 , 0 , 0 , init_index_io }, - { "INIT_PLL" , 0x79, 7 , 0 , 0 , init_pll }, - { "INIT_ZM_REG" , 0x7A, 9 , 0 , 0 , init_zm_reg }, - /* INIT_RAM_RESTRICT_PLL's length is adjusted by the BIT M table */ - { "INIT_RAM_RESTRICT_PLL" , 0x87, 2 , 0 , 0 , init_ram_restrict_pll }, - { "INIT_8C" , 0x8C, 1 , 0 , 0 , init_8c }, - { "INIT_8D" , 0x8D, 1 , 0 , 0 , init_8d }, - { "INIT_GPIO" , 0x8E, 1 , 0 , 0 , init_gpio }, - /* INIT_RAM_RESTRICT_ZM_REG_GROUP's mult is loaded by M table in BIT */ - { "INIT_RAM_RESTRICT_ZM_REG_GROUP" , 0x8F, 7 , 6 , 0 , init_ram_restrict_zm_reg_group }, - { "INIT_COPY_ZM_REG" , 0x90, 9 , 0 , 0 , init_copy_zm_reg }, - { "INIT_ZM_REG_GROUP_ADDRESS_LATCHED" , 0x91, 6 , 5 , 4 , init_zm_reg_group_addr_latched }, - { "INIT_RESERVED" , 0x92, 1 , 0 , 0 , init_reserved }, - { "INIT_96" , 0x96, 17 , 0 , 0 , init_96 }, - { "INIT_97" , 0x97, 13 , 0 , 0 , init_97 }, - { "INIT_AUXCH" , 0x98, 6 , 5 , 2 , init_auxch }, - { "INIT_ZM_AUXCH" , 0x99, 6 , 5 , 1 , init_zm_auxch }, - { NULL , 0 , 0 , 0 , 0 , NULL } + { "INIT_TIME" , 0x74, init_time }, + { "INIT_CONDITION" , 0x75, init_condition }, + { "INIT_IO_CONDITION" , 0x76, init_io_condition }, + { "INIT_INDEX_IO" , 0x78, init_index_io }, + { "INIT_PLL" , 0x79, init_pll }, + { "INIT_ZM_REG" , 0x7A, init_zm_reg }, + { "INIT_RAM_RESTRICT_PLL" , 0x87, init_ram_restrict_pll }, + { "INIT_8C" , 0x8C, init_8c }, + { "INIT_8D" , 0x8D, init_8d }, + { "INIT_GPIO" , 0x8E, init_gpio }, + { "INIT_RAM_RESTRICT_ZM_REG_GROUP" , 0x8F, init_ram_restrict_zm_reg_group }, + { "INIT_COPY_ZM_REG" , 0x90, init_copy_zm_reg }, + { "INIT_ZM_REG_GROUP_ADDRESS_LATCHED" , 0x91, init_zm_reg_group_addr_latched }, + { "INIT_RESERVED" , 0x92, init_reserved }, + { "INIT_96" , 0x96, init_96 }, + { "INIT_97" , 0x97, init_97 }, + { "INIT_AUXCH" , 0x98, init_auxch }, + { "INIT_ZM_AUXCH" , 0x99, init_zm_auxch }, + { NULL , 0 , NULL } }; -static unsigned int get_init_table_entry_length(struct nvbios *bios, unsigned int offset, int i) -{ - /* Calculates the length of a given init table entry. */ - return itbl_entry[i].length + bios->data[offset + itbl_entry[i].length_offset]*itbl_entry[i].length_multiplier; -} - #define MAX_TABLE_OPS 1000 static int @@ -3056,7 +3055,7 @@ parse_init_table(struct nvbios *bios, unsigned int offset, * is changed back to EXECUTE. */ - int count = 0, i; + int count = 0, i, res; uint8_t id; /* @@ -3076,22 +3075,21 @@ parse_init_table(struct nvbios *bios, unsigned int offset, offset, itbl_entry[i].id, itbl_entry[i].name); /* execute eventual command handler */ - if (itbl_entry[i].handler) - if (!(*itbl_entry[i].handler)(bios, offset, iexec)) - break; + res = (*itbl_entry[i].handler)(bios, offset, iexec); + if (!res) + break; + /* + * Add the offset of the current command including all data + * of that command. The offset will then be pointing on the + * next op code. + */ + offset += res; } else { NV_ERROR(bios->dev, "0x%04X: Init table command not found: " "0x%02X\n", offset, id); return -ENOENT; } - - /* - * Add the offset of the current command including all data - * of that command. The offset will then be pointing on the - * next op code. - */ - offset += get_init_table_entry_length(bios, offset, i); } if (offset >= bios->length) @@ -3854,7 +3852,7 @@ nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent, * script tables is a pointer to the script to execute. */ - NV_DEBUG(dev, "Searching for output entry for %d %d %d\n", + NV_DEBUG_KMS(dev, "Searching for output entry for %d %d %d\n", dcbent->type, dcbent->location, dcbent->or); otable = bios_output_config_match(dev, dcbent, table[1] + bios->display.script_table_ptr, @@ -3884,7 +3882,7 @@ nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent, if (pxclk == 0) { script = ROM16(otable[6]); if (!script) { - NV_DEBUG(dev, "output script 0 not found\n"); + NV_DEBUG_KMS(dev, "output script 0 not found\n"); return 1; } @@ -3894,7 +3892,7 @@ nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent, if (pxclk == -1) { script = ROM16(otable[8]); if (!script) { - NV_DEBUG(dev, "output script 1 not found\n"); + NV_DEBUG_KMS(dev, "output script 1 not found\n"); return 1; } @@ -3907,7 +3905,7 @@ nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent, else script = 0; if (!script) { - NV_DEBUG(dev, "output script 2 not found\n"); + NV_DEBUG_KMS(dev, "output script 2 not found\n"); return 1; } @@ -3931,7 +3929,7 @@ nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent, if (script) script = clkcmptable(bios, script, -pxclk); if (!script) { - NV_DEBUG(dev, "clock script 1 not found\n"); + NV_DEBUG_KMS(dev, "clock script 1 not found\n"); return 1; } @@ -4606,10 +4604,6 @@ parse_bit_M_tbl_entry(struct drm_device *dev, struct nvbios *bios, * stuff that we don't use - their use currently unknown */ - uint16_t rr_strap_xlat; - uint8_t rr_group_count; - int i; - /* * Older bios versions don't have a sufficiently long table for * what we want @@ -4618,24 +4612,13 @@ parse_bit_M_tbl_entry(struct drm_device *dev, struct nvbios *bios, return 0; if (bitentry->id[1] < 2) { - rr_group_count = bios->data[bitentry->offset + 2]; - rr_strap_xlat = ROM16(bios->data[bitentry->offset + 3]); + bios->ram_restrict_group_count = bios->data[bitentry->offset + 2]; + bios->ram_restrict_tbl_ptr = ROM16(bios->data[bitentry->offset + 3]); } else { - rr_group_count = bios->data[bitentry->offset + 0]; - rr_strap_xlat = ROM16(bios->data[bitentry->offset + 1]); + bios->ram_restrict_group_count = bios->data[bitentry->offset + 0]; + bios->ram_restrict_tbl_ptr = ROM16(bios->data[bitentry->offset + 1]); } - /* adjust length of INIT_87 */ - for (i = 0; itbl_entry[i].name && (itbl_entry[i].id != 0x87); i++); - itbl_entry[i].length += rr_group_count * 4; - - /* set up multiplier for INIT_RAM_RESTRICT_ZM_REG_GROUP */ - for (; itbl_entry[i].name && (itbl_entry[i].id != 0x8f); i++); - itbl_entry[i].length_multiplier = rr_group_count * 4; - - init_ram_restrict_zm_reg_group_blocklen = itbl_entry[i].length_multiplier; - bios->ram_restrict_tbl_ptr = rr_strap_xlat; - return 0; } @@ -5234,7 +5217,7 @@ parse_dcb_connector_table(struct nvbios *bios) int i; if (!bios->bdcb.connector_table_ptr) { - NV_DEBUG(dev, "No DCB connector table present\n"); + NV_DEBUG_KMS(dev, "No DCB connector table present\n"); return; } diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.h b/drivers/gpu/drm/nouveau/nouveau_bios.h index 1d5f10bd78ed..058e98c76d89 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bios.h +++ b/drivers/gpu/drm/nouveau/nouveau_bios.h @@ -227,6 +227,7 @@ struct nvbios { uint16_t pll_limit_tbl_ptr; uint16_t ram_restrict_tbl_ptr; + uint8_t ram_restrict_group_count; uint16_t some_script_ptr; /* BIT I + 14 */ uint16_t init96_tbl_ptr; /* BIT I + 16 */ diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 320a14bceb99..e342a418d434 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -33,10 +33,13 @@ #include "nouveau_drv.h" #include "nouveau_dma.h" +#include <linux/log2.h> + static void nouveau_bo_del_ttm(struct ttm_buffer_object *bo) { struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev); + struct drm_device *dev = dev_priv->dev; struct nouveau_bo *nvbo = nouveau_bo(bo); ttm_bo_kunmap(&nvbo->kmap); @@ -44,12 +47,87 @@ nouveau_bo_del_ttm(struct ttm_buffer_object *bo) if (unlikely(nvbo->gem)) DRM_ERROR("bo %p still attached to GEM object\n", bo); + if (nvbo->tile) + nv10_mem_expire_tiling(dev, nvbo->tile, NULL); + spin_lock(&dev_priv->ttm.bo_list_lock); list_del(&nvbo->head); spin_unlock(&dev_priv->ttm.bo_list_lock); kfree(nvbo); } +static void +nouveau_bo_fixup_align(struct drm_device *dev, + uint32_t tile_mode, uint32_t tile_flags, + int *align, int *size) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + + /* + * Some of the tile_flags have a periodic structure of N*4096 bytes, + * align to to that as well as the page size. Overallocate memory to + * avoid corruption of other buffer objects. + */ + if (dev_priv->card_type == NV_50) { + uint32_t block_size = nouveau_mem_fb_amount(dev) >> 15; + int i; + + switch (tile_flags) { + case 0x1800: + case 0x2800: + case 0x4800: + case 0x7a00: + *size = roundup(*size, block_size); + if (is_power_of_2(block_size)) { + *size += 3 * block_size; + for (i = 1; i < 10; i++) { + *align = 12 * i * block_size; + if (!(*align % 65536)) + break; + } + } else { + *size += 6 * block_size; + for (i = 1; i < 10; i++) { + *align = 8 * i * block_size; + if (!(*align % 65536)) + break; + } + } + break; + default: + break; + } + + } else { + if (tile_mode) { + if (dev_priv->chipset >= 0x40) { + *align = 65536; + *size = roundup(*size, 64 * tile_mode); + + } else if (dev_priv->chipset >= 0x30) { + *align = 32768; + *size = roundup(*size, 64 * tile_mode); + + } else if (dev_priv->chipset >= 0x20) { + *align = 16384; + *size = roundup(*size, 64 * tile_mode); + + } else if (dev_priv->chipset >= 0x10) { + *align = 16384; + *size = roundup(*size, 32 * tile_mode); + } + } + } + + /* ALIGN works only on powers of two. */ + *size = roundup(*size, PAGE_SIZE); + + if (dev_priv->card_type == NV_50) { + *size = roundup(*size, 65536); + *align = max(65536, *align); + } +} + int nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan, int size, int align, uint32_t flags, uint32_t tile_mode, @@ -58,7 +136,7 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan, { struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_bo *nvbo; - int ret, n = 0; + int ret = 0; nvbo = kzalloc(sizeof(struct nouveau_bo), GFP_KERNEL); if (!nvbo) @@ -70,59 +148,14 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan, nvbo->tile_mode = tile_mode; nvbo->tile_flags = tile_flags; - /* - * Some of the tile_flags have a periodic structure of N*4096 bytes, - * align to to that as well as the page size. Overallocate memory to - * avoid corruption of other buffer objects. - */ - switch (tile_flags) { - case 0x1800: - case 0x2800: - case 0x4800: - case 0x7a00: - if (dev_priv->chipset >= 0xA0) { - /* This is based on high end cards with 448 bits - * memory bus, could be different elsewhere.*/ - size += 6 * 28672; - /* 8 * 28672 is the actual alignment requirement, - * but we must also align to page size. */ - align = 2 * 8 * 28672; - } else if (dev_priv->chipset >= 0x90) { - size += 3 * 16384; - align = 12 * 16834; - } else { - size += 3 * 8192; - /* 12 * 8192 is the actual alignment requirement, - * but we must also align to page size. */ - align = 2 * 12 * 8192; - } - break; - default: - break; - } - + nouveau_bo_fixup_align(dev, tile_mode, tile_flags, &align, &size); align >>= PAGE_SHIFT; - size = (size + (PAGE_SIZE - 1)) & ~(PAGE_SIZE - 1); - if (dev_priv->card_type == NV_50) { - size = (size + 65535) & ~65535; - if (align < (65536 / PAGE_SIZE)) - align = (65536 / PAGE_SIZE); - } - - if (flags & TTM_PL_FLAG_VRAM) - nvbo->placements[n++] = TTM_PL_FLAG_VRAM | TTM_PL_MASK_CACHING; - if (flags & TTM_PL_FLAG_TT) - nvbo->placements[n++] = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING; nvbo->placement.fpfn = 0; nvbo->placement.lpfn = mappable ? dev_priv->fb_mappable_pages : 0; - nvbo->placement.placement = nvbo->placements; - nvbo->placement.busy_placement = nvbo->placements; - nvbo->placement.num_placement = n; - nvbo->placement.num_busy_placement = n; + nouveau_bo_placement_set(nvbo, flags); nvbo->channel = chan; - nouveau_bo_placement_set(nvbo, flags); ret = ttm_bo_init(&dev_priv->ttm.bdev, &nvbo->bo, size, ttm_bo_type_device, &nvbo->placement, align, 0, false, NULL, size, nouveau_bo_del_ttm); @@ -154,6 +187,11 @@ nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t memtype) nvbo->placement.busy_placement = nvbo->placements; nvbo->placement.num_placement = n; nvbo->placement.num_busy_placement = n; + + if (nvbo->pin_refcnt) { + while (n--) + nvbo->placements[n] |= TTM_PL_FLAG_NO_EVICT; + } } int @@ -311,8 +349,10 @@ nouveau_bo_create_ttm_backend_entry(struct ttm_bo_device *bdev) struct drm_device *dev = dev_priv->dev; switch (dev_priv->gart_info.type) { +#if __OS_HAS_AGP case NOUVEAU_GART_AGP: return ttm_agp_backend_init(bdev, dev->agp->bridge); +#endif case NOUVEAU_GART_SGDMA: return nouveau_sgdma_init_ttm(dev); default: @@ -398,16 +438,23 @@ nouveau_bo_evict_flags(struct ttm_buffer_object *bo, struct ttm_placement *pl) struct nouveau_bo *nvbo = nouveau_bo(bo); switch (bo->mem.mem_type) { + case TTM_PL_VRAM: + nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_TT | + TTM_PL_FLAG_SYSTEM); + break; default: nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_SYSTEM); break; } + + *pl = nvbo->placement; } /* GPU-assisted copy using NV_MEMORY_TO_MEMORY_FORMAT, can access * TTM_PL_{VRAM,TT} directly. */ + static int nouveau_bo_move_accel_cleanup(struct nouveau_channel *chan, struct nouveau_bo *nvbo, bool evict, bool no_wait, @@ -442,22 +489,20 @@ nouveau_bo_mem_ctxdma(struct nouveau_bo *nvbo, struct nouveau_channel *chan, } static int -nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, int no_wait, - struct ttm_mem_reg *old_mem, struct ttm_mem_reg *new_mem) +nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr, + int no_wait, struct ttm_mem_reg *new_mem) { struct nouveau_bo *nvbo = nouveau_bo(bo); struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev); + struct ttm_mem_reg *old_mem = &bo->mem; struct nouveau_channel *chan; uint64_t src_offset, dst_offset; uint32_t page_count; int ret; chan = nvbo->channel; - if (!chan || nvbo->tile_flags || nvbo->no_vm) { + if (!chan || nvbo->tile_flags || nvbo->no_vm) chan = dev_priv->channel; - if (!chan) - return -EINVAL; - } src_offset = old_mem->mm_node->start << PAGE_SHIFT; dst_offset = new_mem->mm_node->start << PAGE_SHIFT; @@ -537,7 +582,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr, placement.fpfn = placement.lpfn = 0; placement.num_placement = placement.num_busy_placement = 1; - placement.placement = &placement_memtype; + placement.placement = placement.busy_placement = &placement_memtype; tmp_mem = *new_mem; tmp_mem.mm_node = NULL; @@ -549,7 +594,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr, if (ret) goto out; - ret = nouveau_bo_move_m2mf(bo, true, no_wait, &bo->mem, &tmp_mem); + ret = nouveau_bo_move_m2mf(bo, true, intr, no_wait, &tmp_mem); if (ret) goto out; @@ -575,7 +620,7 @@ nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr, placement.fpfn = placement.lpfn = 0; placement.num_placement = placement.num_busy_placement = 1; - placement.placement = &placement_memtype; + placement.placement = placement.busy_placement = &placement_memtype; tmp_mem = *new_mem; tmp_mem.mm_node = NULL; @@ -587,7 +632,7 @@ nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr, if (ret) goto out; - ret = nouveau_bo_move_m2mf(bo, true, no_wait, &bo->mem, new_mem); + ret = nouveau_bo_move_m2mf(bo, evict, intr, no_wait, new_mem); if (ret) goto out; @@ -602,51 +647,106 @@ out: } static int -nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, bool intr, - bool no_wait, struct ttm_mem_reg *new_mem) +nouveau_bo_vm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem, + struct nouveau_tile_reg **new_tile) { struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev); - struct nouveau_bo *nvbo = nouveau_bo(bo); struct drm_device *dev = dev_priv->dev; - struct ttm_mem_reg *old_mem = &bo->mem; + struct nouveau_bo *nvbo = nouveau_bo(bo); + uint64_t offset; int ret; - if (dev_priv->card_type == NV_50 && new_mem->mem_type == TTM_PL_VRAM && - !nvbo->no_vm) { - uint64_t offset = new_mem->mm_node->start << PAGE_SHIFT; + if (nvbo->no_vm || new_mem->mem_type != TTM_PL_VRAM) { + /* Nothing to do. */ + *new_tile = NULL; + return 0; + } + offset = new_mem->mm_node->start << PAGE_SHIFT; + + if (dev_priv->card_type == NV_50) { ret = nv50_mem_vm_bind_linear(dev, offset + dev_priv->vm_vram_base, new_mem->size, nvbo->tile_flags, offset); if (ret) return ret; + + } else if (dev_priv->card_type >= NV_10) { + *new_tile = nv10_mem_set_tiling(dev, offset, new_mem->size, + nvbo->tile_mode); } - if (dev_priv->init_state != NOUVEAU_CARD_INIT_DONE) - return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); + return 0; +} + +static void +nouveau_bo_vm_cleanup(struct ttm_buffer_object *bo, + struct nouveau_tile_reg *new_tile, + struct nouveau_tile_reg **old_tile) +{ + struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev); + struct drm_device *dev = dev_priv->dev; + if (dev_priv->card_type >= NV_10 && + dev_priv->card_type < NV_50) { + if (*old_tile) + nv10_mem_expire_tiling(dev, *old_tile, bo->sync_obj); + + *old_tile = new_tile; + } +} + +static int +nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, bool intr, + bool no_wait, struct ttm_mem_reg *new_mem) +{ + struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev); + struct nouveau_bo *nvbo = nouveau_bo(bo); + struct ttm_mem_reg *old_mem = &bo->mem; + struct nouveau_tile_reg *new_tile = NULL; + int ret = 0; + + ret = nouveau_bo_vm_bind(bo, new_mem, &new_tile); + if (ret) + return ret; + + /* Software copy if the card isn't up and running yet. */ + if (dev_priv->init_state != NOUVEAU_CARD_INIT_DONE || + !dev_priv->channel) { + ret = ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); + goto out; + } + + /* Fake bo copy. */ if (old_mem->mem_type == TTM_PL_SYSTEM && !bo->ttm) { BUG_ON(bo->mem.mm_node != NULL); bo->mem = *new_mem; new_mem->mm_node = NULL; - return 0; + goto out; } - if (new_mem->mem_type == TTM_PL_SYSTEM) { - if (old_mem->mem_type == TTM_PL_SYSTEM) - return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); - if (nouveau_bo_move_flipd(bo, evict, intr, no_wait, new_mem)) - return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); - } else if (old_mem->mem_type == TTM_PL_SYSTEM) { - if (nouveau_bo_move_flips(bo, evict, intr, no_wait, new_mem)) - return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); - } else { - if (nouveau_bo_move_m2mf(bo, evict, no_wait, old_mem, new_mem)) - return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); - } + /* Hardware assisted copy. */ + if (new_mem->mem_type == TTM_PL_SYSTEM) + ret = nouveau_bo_move_flipd(bo, evict, intr, no_wait, new_mem); + else if (old_mem->mem_type == TTM_PL_SYSTEM) + ret = nouveau_bo_move_flips(bo, evict, intr, no_wait, new_mem); + else + ret = nouveau_bo_move_m2mf(bo, evict, intr, no_wait, new_mem); - return 0; + if (!ret) + goto out; + + /* Fallback to software copy. */ + ret = ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); + +out: + if (ret) + nouveau_bo_vm_cleanup(bo, NULL, &new_tile); + else + nouveau_bo_vm_cleanup(bo, new_tile, &nvbo->tile); + + return ret; } static int diff --git a/drivers/gpu/drm/nouveau/nouveau_channel.c b/drivers/gpu/drm/nouveau/nouveau_channel.c index 9aaa972f8822..343d718a9667 100644 --- a/drivers/gpu/drm/nouveau/nouveau_channel.c +++ b/drivers/gpu/drm/nouveau/nouveau_channel.c @@ -158,6 +158,8 @@ nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret, return ret; } + nouveau_dma_pre_init(chan); + /* Locate channel's user control regs */ if (dev_priv->card_type < NV_40) user = NV03_USER(channel); @@ -235,47 +237,6 @@ nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret, return 0; } -int -nouveau_channel_idle(struct nouveau_channel *chan) -{ - struct drm_device *dev = chan->dev; - struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nouveau_engine *engine = &dev_priv->engine; - uint32_t caches; - int idle; - - if (!chan) { - NV_ERROR(dev, "no channel...\n"); - return 1; - } - - caches = nv_rd32(dev, NV03_PFIFO_CACHES); - nv_wr32(dev, NV03_PFIFO_CACHES, caches & ~1); - - if (engine->fifo.channel_id(dev) != chan->id) { - struct nouveau_gpuobj *ramfc = - chan->ramfc ? chan->ramfc->gpuobj : NULL; - - if (!ramfc) { - NV_ERROR(dev, "No RAMFC for channel %d\n", chan->id); - return 1; - } - - engine->instmem.prepare_access(dev, false); - if (nv_ro32(dev, ramfc, 0) != nv_ro32(dev, ramfc, 1)) - idle = 0; - else - idle = 1; - engine->instmem.finish_access(dev); - } else { - idle = (nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_GET) == - nv_rd32(dev, NV04_PFIFO_CACHE1_DMA_PUT)); - } - - nv_wr32(dev, NV03_PFIFO_CACHES, caches); - return idle; -} - /* stops a fifo */ void nouveau_channel_free(struct nouveau_channel *chan) @@ -414,7 +375,9 @@ nouveau_ioctl_fifo_alloc(struct drm_device *dev, void *data, init->subchan[0].grclass = 0x0039; else init->subchan[0].grclass = 0x5039; - init->nr_subchan = 1; + init->subchan[1].handle = NvSw; + init->subchan[1].grclass = NV_SW; + init->nr_subchan = 2; /* Named memory object area */ ret = drm_gem_handle_create(file_priv, chan->notifier_bo->gem, diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c b/drivers/gpu/drm/nouveau/nouveau_connector.c index 032cf098fa1c..5a10deb8bdbd 100644 --- a/drivers/gpu/drm/nouveau/nouveau_connector.c +++ b/drivers/gpu/drm/nouveau/nouveau_connector.c @@ -86,7 +86,7 @@ nouveau_connector_destroy(struct drm_connector *drm_connector) struct nouveau_connector *connector = nouveau_connector(drm_connector); struct drm_device *dev = connector->base.dev; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); if (!connector) return; @@ -420,7 +420,7 @@ nouveau_connector_native_mode(struct nouveau_connector *connector) /* Use preferred mode if there is one.. */ list_for_each_entry(mode, &connector->base.probed_modes, head) { if (mode->type & DRM_MODE_TYPE_PREFERRED) { - NV_DEBUG(dev, "native mode from preferred\n"); + NV_DEBUG_KMS(dev, "native mode from preferred\n"); return drm_mode_duplicate(dev, mode); } } @@ -445,7 +445,7 @@ nouveau_connector_native_mode(struct nouveau_connector *connector) largest = mode; } - NV_DEBUG(dev, "native mode from largest: %dx%d@%d\n", + NV_DEBUG_KMS(dev, "native mode from largest: %dx%d@%d\n", high_w, high_h, high_v); return largest ? drm_mode_duplicate(dev, largest) : NULL; } @@ -725,7 +725,7 @@ nouveau_connector_create(struct drm_device *dev, int index, int type) struct drm_encoder *encoder; int ret; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); nv_connector = kzalloc(sizeof(*nv_connector), GFP_KERNEL); if (!nv_connector) diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.c b/drivers/gpu/drm/nouveau/nouveau_dma.c index 703553687b20..7afbe8b40d51 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dma.c +++ b/drivers/gpu/drm/nouveau/nouveau_dma.c @@ -29,12 +29,22 @@ #include "nouveau_drv.h" #include "nouveau_dma.h" +void +nouveau_dma_pre_init(struct nouveau_channel *chan) +{ + chan->dma.max = (chan->pushbuf_bo->bo.mem.size >> 2) - 2; + chan->dma.put = 0; + chan->dma.cur = chan->dma.put; + chan->dma.free = chan->dma.max - chan->dma.cur; +} + int nouveau_dma_init(struct nouveau_channel *chan) { struct drm_device *dev = chan->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_gpuobj *m2mf = NULL; + struct nouveau_gpuobj *nvsw = NULL; int ret, i; /* Create NV_MEMORY_TO_MEMORY_FORMAT for buffer moves */ @@ -47,6 +57,15 @@ nouveau_dma_init(struct nouveau_channel *chan) if (ret) return ret; + /* Create an NV_SW object for various sync purposes */ + ret = nouveau_gpuobj_sw_new(chan, NV_SW, &nvsw); + if (ret) + return ret; + + ret = nouveau_gpuobj_ref_add(dev, chan, NvSw, nvsw, NULL); + if (ret) + return ret; + /* NV_MEMORY_TO_MEMORY_FORMAT requires a notifier object */ ret = nouveau_notifier_alloc(chan, NvNotify0, 32, &chan->m2mf_ntfy); if (ret) @@ -64,12 +83,6 @@ nouveau_dma_init(struct nouveau_channel *chan) return ret; } - /* Initialise DMA vars */ - chan->dma.max = (chan->pushbuf_bo->bo.mem.size >> 2) - 2; - chan->dma.put = 0; - chan->dma.cur = chan->dma.put; - chan->dma.free = chan->dma.max - chan->dma.cur; - /* Insert NOPS for NOUVEAU_DMA_SKIPS */ ret = RING_SPACE(chan, NOUVEAU_DMA_SKIPS); if (ret) @@ -87,6 +100,13 @@ nouveau_dma_init(struct nouveau_channel *chan) BEGIN_RING(chan, NvSubM2MF, NV_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 1); OUT_RING(chan, NvNotify0); + /* Initialise NV_SW */ + ret = RING_SPACE(chan, 2); + if (ret) + return ret; + BEGIN_RING(chan, NvSubSw, 0, 1); + OUT_RING(chan, NvSw); + /* Sit back and pray the channel works.. */ FIRE_RING(chan); @@ -113,7 +133,7 @@ READ_GET(struct nouveau_channel *chan, uint32_t *get) val = nvchan_rd32(chan, chan->user_get); if (val < chan->pushbuf_base || - val >= chan->pushbuf_base + chan->pushbuf_bo->bo.mem.size) { + val > chan->pushbuf_base + (chan->dma.max << 2)) { /* meaningless to dma_wait() except to know whether the * GPU has stalled or not */ diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.h b/drivers/gpu/drm/nouveau/nouveau_dma.h index 04e85d8f757e..dabfd655f93e 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dma.h +++ b/drivers/gpu/drm/nouveau/nouveau_dma.h @@ -46,10 +46,11 @@ /* Hardcoded object assignments to subchannels (subchannel id). */ enum { NvSubM2MF = 0, - NvSub2D = 1, - NvSubCtxSurf2D = 1, - NvSubGdiRect = 2, - NvSubImageBlit = 3 + NvSubSw = 1, + NvSub2D = 2, + NvSubCtxSurf2D = 2, + NvSubGdiRect = 3, + NvSubImageBlit = 4 }; /* Object handles. */ @@ -67,6 +68,7 @@ enum { NvClipRect = 0x8000000b, NvGdiRect = 0x8000000c, NvImageBlit = 0x8000000d, + NvSw = 0x8000000e, /* G80+ display objects */ NvEvoVRAM = 0x01000000, diff --git a/drivers/gpu/drm/nouveau/nouveau_dp.c b/drivers/gpu/drm/nouveau/nouveau_dp.c index de61f4640e12..9e2926c48579 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dp.c +++ b/drivers/gpu/drm/nouveau/nouveau_dp.c @@ -187,7 +187,7 @@ nouveau_dp_link_train_adjust(struct drm_encoder *encoder, uint8_t *config) if (ret) return false; - NV_DEBUG(dev, "\t\tadjust 0x%02x 0x%02x\n", request[0], request[1]); + NV_DEBUG_KMS(dev, "\t\tadjust 0x%02x 0x%02x\n", request[0], request[1]); /* Keep all lanes at the same level.. */ for (i = 0; i < nv_encoder->dp.link_nr; i++) { @@ -228,7 +228,7 @@ nouveau_dp_link_train_commit(struct drm_encoder *encoder, uint8_t *config) int or = nv_encoder->or, link = !(nv_encoder->dcb->sorconf.link & 1); int dpe_headerlen, ret, i; - NV_DEBUG(dev, "\t\tconfig 0x%02x 0x%02x 0x%02x 0x%02x\n", + NV_DEBUG_KMS(dev, "\t\tconfig 0x%02x 0x%02x 0x%02x 0x%02x\n", config[0], config[1], config[2], config[3]); dpe = nouveau_bios_dp_table(dev, nv_encoder->dcb, &dpe_headerlen); @@ -276,12 +276,12 @@ nouveau_dp_link_train(struct drm_encoder *encoder) bool cr_done, cr_max_vs, eq_done; int ret = 0, i, tries, voltage; - NV_DEBUG(dev, "link training!!\n"); + NV_DEBUG_KMS(dev, "link training!!\n"); train: cr_done = eq_done = false; /* set link configuration */ - NV_DEBUG(dev, "\tbegin train: bw %d, lanes %d\n", + NV_DEBUG_KMS(dev, "\tbegin train: bw %d, lanes %d\n", nv_encoder->dp.link_bw, nv_encoder->dp.link_nr); ret = nouveau_dp_link_bw_set(encoder, nv_encoder->dp.link_bw); @@ -297,7 +297,7 @@ train: return false; /* clock recovery */ - NV_DEBUG(dev, "\tbegin cr\n"); + NV_DEBUG_KMS(dev, "\tbegin cr\n"); ret = nouveau_dp_link_train_set(encoder, DP_TRAINING_PATTERN_1); if (ret) goto stop; @@ -314,7 +314,7 @@ train: ret = auxch_rd(encoder, DP_LANE0_1_STATUS, status, 2); if (ret) break; - NV_DEBUG(dev, "\t\tstatus: 0x%02x 0x%02x\n", + NV_DEBUG_KMS(dev, "\t\tstatus: 0x%02x 0x%02x\n", status[0], status[1]); cr_done = true; @@ -346,7 +346,7 @@ train: goto stop; /* channel equalisation */ - NV_DEBUG(dev, "\tbegin eq\n"); + NV_DEBUG_KMS(dev, "\tbegin eq\n"); ret = nouveau_dp_link_train_set(encoder, DP_TRAINING_PATTERN_2); if (ret) goto stop; @@ -357,7 +357,7 @@ train: ret = auxch_rd(encoder, DP_LANE0_1_STATUS, status, 3); if (ret) break; - NV_DEBUG(dev, "\t\tstatus: 0x%02x 0x%02x\n", + NV_DEBUG_KMS(dev, "\t\tstatus: 0x%02x 0x%02x\n", status[0], status[1]); eq_done = true; @@ -395,9 +395,9 @@ stop: /* retry at a lower setting, if possible */ if (!ret && !(eq_done && cr_done)) { - NV_DEBUG(dev, "\twe failed\n"); + NV_DEBUG_KMS(dev, "\twe failed\n"); if (nv_encoder->dp.link_bw != DP_LINK_BW_1_62) { - NV_DEBUG(dev, "retry link training at low rate\n"); + NV_DEBUG_KMS(dev, "retry link training at low rate\n"); nv_encoder->dp.link_bw = DP_LINK_BW_1_62; goto train; } @@ -418,7 +418,7 @@ nouveau_dp_detect(struct drm_encoder *encoder) if (ret) return false; - NV_DEBUG(dev, "encoder: link_bw %d, link_nr %d\n" + NV_DEBUG_KMS(dev, "encoder: link_bw %d, link_nr %d\n" "display: link_bw %d, link_nr %d version 0x%02x\n", nv_encoder->dcb->dpconf.link_bw, nv_encoder->dcb->dpconf.link_nr, @@ -446,7 +446,7 @@ nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr, uint32_t tmp, ctrl, stat = 0, data32[4] = {}; int ret = 0, i, index = auxch->rd; - NV_DEBUG(dev, "ch %d cmd %d addr 0x%x len %d\n", index, cmd, addr, data_nr); + NV_DEBUG_KMS(dev, "ch %d cmd %d addr 0x%x len %d\n", index, cmd, addr, data_nr); tmp = nv_rd32(dev, NV50_AUXCH_CTRL(auxch->rd)); nv_wr32(dev, NV50_AUXCH_CTRL(auxch->rd), tmp | 0x00100000); @@ -472,7 +472,7 @@ nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr, if (!(cmd & 1)) { memcpy(data32, data, data_nr); for (i = 0; i < 4; i++) { - NV_DEBUG(dev, "wr %d: 0x%08x\n", i, data32[i]); + NV_DEBUG_KMS(dev, "wr %d: 0x%08x\n", i, data32[i]); nv_wr32(dev, NV50_AUXCH_DATA_OUT(index, i), data32[i]); } } @@ -504,7 +504,7 @@ nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr, if (cmd & 1) { for (i = 0; i < 4; i++) { data32[i] = nv_rd32(dev, NV50_AUXCH_DATA_IN(index, i)); - NV_DEBUG(dev, "rd %d: 0x%08x\n", i, data32[i]); + NV_DEBUG_KMS(dev, "rd %d: 0x%08x\n", i, data32[i]); } memcpy(data, data32, data_nr); } diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.c b/drivers/gpu/drm/nouveau/nouveau_drv.c index 35249c35118f..06eb993e0883 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drv.c +++ b/drivers/gpu/drm/nouveau/nouveau_drv.c @@ -35,6 +35,10 @@ #include "drm_pciids.h" +MODULE_PARM_DESC(ctxfw, "Use external firmware blob for grctx init (NV40)"); +int nouveau_ctxfw = 0; +module_param_named(ctxfw, nouveau_ctxfw, int, 0400); + MODULE_PARM_DESC(noagp, "Disable AGP"); int nouveau_noagp; module_param_named(noagp, nouveau_noagp, int, 0400); @@ -273,7 +277,7 @@ nouveau_pci_resume(struct pci_dev *pdev) for (i = 0; i < dev_priv->engine.fifo.channels; i++) { chan = dev_priv->fifos[i]; - if (!chan) + if (!chan || !chan->pushbuf_bo) continue; for (j = 0; j < NOUVEAU_DMA_SKIPS; j++) @@ -341,7 +345,7 @@ static struct drm_driver driver = { .owner = THIS_MODULE, .open = drm_open, .release = drm_release, - .ioctl = drm_ioctl, + .unlocked_ioctl = drm_ioctl, .mmap = nouveau_ttm_mmap, .poll = drm_poll, .fasync = drm_fasync, diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h index 88b4c7b77e7f..026419fe8791 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drv.h +++ b/drivers/gpu/drm/nouveau/nouveau_drv.h @@ -54,15 +54,24 @@ struct nouveau_fpriv { #include "nouveau_drm.h" #include "nouveau_reg.h" #include "nouveau_bios.h" +struct nouveau_grctx; #define MAX_NUM_DCB_ENTRIES 16 #define NOUVEAU_MAX_CHANNEL_NR 128 +#define NOUVEAU_MAX_TILE_NR 15 #define NV50_VM_MAX_VRAM (2*1024*1024*1024ULL) #define NV50_VM_BLOCK (512*1024*1024ULL) #define NV50_VM_VRAM_NR (NV50_VM_MAX_VRAM / NV50_VM_BLOCK) +struct nouveau_tile_reg { + struct nouveau_fence *fence; + uint32_t addr; + uint32_t size; + bool used; +}; + struct nouveau_bo { struct ttm_buffer_object bo; struct ttm_placement placement; @@ -82,6 +91,7 @@ struct nouveau_bo { uint32_t tile_mode; uint32_t tile_flags; + struct nouveau_tile_reg *tile; struct drm_gem_object *gem; struct drm_file *cpu_filp; @@ -276,8 +286,13 @@ struct nouveau_timer_engine { }; struct nouveau_fb_engine { + int num_tiles; + int (*init)(struct drm_device *dev); void (*takedown)(struct drm_device *dev); + + void (*set_region_tiling)(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch); }; struct nouveau_fifo_engine { @@ -291,6 +306,8 @@ struct nouveau_fifo_engine { void (*disable)(struct drm_device *); void (*enable)(struct drm_device *); bool (*reassign)(struct drm_device *, bool enable); + bool (*cache_flush)(struct drm_device *dev); + bool (*cache_pull)(struct drm_device *dev, bool enable); int (*channel_id)(struct drm_device *); @@ -317,6 +334,7 @@ struct nouveau_pgraph_engine { bool accel_blocked; void *ctxprog; void *ctxvals; + int grctx_size; int (*init)(struct drm_device *); void (*takedown)(struct drm_device *); @@ -328,6 +346,9 @@ struct nouveau_pgraph_engine { void (*destroy_context)(struct nouveau_channel *); int (*load_context)(struct nouveau_channel *); int (*unload_context)(struct drm_device *); + + void (*set_region_tiling)(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch); }; struct nouveau_engine { @@ -546,6 +567,12 @@ struct drm_nouveau_private { unsigned long sg_handle; } gart_info; + /* nv10-nv40 tiling regions */ + struct { + struct nouveau_tile_reg reg[NOUVEAU_MAX_TILE_NR]; + spinlock_t lock; + } tile; + /* G8x/G9x virtual address space */ uint64_t vm_gart_base; uint64_t vm_gart_size; @@ -647,6 +674,7 @@ extern int nouveau_fbpercrtc; extern char *nouveau_tv_norm; extern int nouveau_reg_debug; extern char *nouveau_vbios; +extern int nouveau_ctxfw; /* nouveau_state.c */ extern void nouveau_preclose(struct drm_device *dev, struct drm_file *); @@ -682,6 +710,13 @@ extern void nouveau_mem_release(struct drm_file *, struct mem_block *heap); extern int nouveau_mem_init(struct drm_device *); extern int nouveau_mem_init_agp(struct drm_device *); extern void nouveau_mem_close(struct drm_device *); +extern struct nouveau_tile_reg *nv10_mem_set_tiling(struct drm_device *dev, + uint32_t addr, + uint32_t size, + uint32_t pitch); +extern void nv10_mem_expire_tiling(struct drm_device *dev, + struct nouveau_tile_reg *tile, + struct nouveau_fence *fence); extern int nv50_mem_vm_bind_linear(struct drm_device *, uint64_t virt, uint32_t size, uint32_t flags, uint64_t phys); @@ -710,7 +745,6 @@ extern int nouveau_channel_alloc(struct drm_device *dev, struct drm_file *file_priv, uint32_t fb_ctxdma, uint32_t tt_ctxdma); extern void nouveau_channel_free(struct nouveau_channel *); -extern int nouveau_channel_idle(struct nouveau_channel *chan); /* nouveau_object.c */ extern int nouveau_gpuobj_early_init(struct drm_device *); @@ -753,6 +787,8 @@ extern int nouveau_gpuobj_gart_dma_new(struct nouveau_channel *, uint32_t *o_ret); extern int nouveau_gpuobj_gr_new(struct nouveau_channel *, int class, struct nouveau_gpuobj **); +extern int nouveau_gpuobj_sw_new(struct nouveau_channel *, int class, + struct nouveau_gpuobj **); extern int nouveau_ioctl_grobj_alloc(struct drm_device *, void *data, struct drm_file *); extern int nouveau_ioctl_gpuobj_free(struct drm_device *, void *data, @@ -801,6 +837,7 @@ nouveau_debugfs_channel_fini(struct nouveau_channel *chan) #endif /* nouveau_dma.c */ +extern void nouveau_dma_pre_init(struct nouveau_channel *); extern int nouveau_dma_init(struct nouveau_channel *); extern int nouveau_dma_wait(struct nouveau_channel *, int size); @@ -876,16 +913,22 @@ extern void nv04_fb_takedown(struct drm_device *); /* nv10_fb.c */ extern int nv10_fb_init(struct drm_device *); extern void nv10_fb_takedown(struct drm_device *); +extern void nv10_fb_set_region_tiling(struct drm_device *, int, uint32_t, + uint32_t, uint32_t); /* nv40_fb.c */ extern int nv40_fb_init(struct drm_device *); extern void nv40_fb_takedown(struct drm_device *); +extern void nv40_fb_set_region_tiling(struct drm_device *, int, uint32_t, + uint32_t, uint32_t); /* nv04_fifo.c */ extern int nv04_fifo_init(struct drm_device *); extern void nv04_fifo_disable(struct drm_device *); extern void nv04_fifo_enable(struct drm_device *); extern bool nv04_fifo_reassign(struct drm_device *, bool); +extern bool nv04_fifo_cache_flush(struct drm_device *); +extern bool nv04_fifo_cache_pull(struct drm_device *, bool); extern int nv04_fifo_channel_id(struct drm_device *); extern int nv04_fifo_create_context(struct nouveau_channel *); extern void nv04_fifo_destroy_context(struct nouveau_channel *); @@ -938,6 +981,8 @@ extern void nv10_graph_destroy_context(struct nouveau_channel *); extern int nv10_graph_load_context(struct nouveau_channel *); extern int nv10_graph_unload_context(struct drm_device *); extern void nv10_graph_context_switch(struct drm_device *); +extern void nv10_graph_set_region_tiling(struct drm_device *, int, uint32_t, + uint32_t, uint32_t); /* nv20_graph.c */ extern struct nouveau_pgraph_object_class nv20_graph_grclass[]; @@ -949,6 +994,8 @@ extern int nv20_graph_unload_context(struct drm_device *); extern int nv20_graph_init(struct drm_device *); extern void nv20_graph_takedown(struct drm_device *); extern int nv30_graph_init(struct drm_device *); +extern void nv20_graph_set_region_tiling(struct drm_device *, int, uint32_t, + uint32_t, uint32_t); /* nv40_graph.c */ extern struct nouveau_pgraph_object_class nv40_graph_grclass[]; @@ -959,9 +1006,9 @@ extern int nv40_graph_create_context(struct nouveau_channel *); extern void nv40_graph_destroy_context(struct nouveau_channel *); extern int nv40_graph_load_context(struct nouveau_channel *); extern int nv40_graph_unload_context(struct drm_device *); -extern int nv40_grctx_init(struct drm_device *); -extern void nv40_grctx_fini(struct drm_device *); -extern void nv40_grctx_vals_load(struct drm_device *, struct nouveau_gpuobj *); +extern void nv40_grctx_init(struct nouveau_grctx *); +extern void nv40_graph_set_region_tiling(struct drm_device *, int, uint32_t, + uint32_t, uint32_t); /* nv50_graph.c */ extern struct nouveau_pgraph_object_class nv50_graph_grclass[]; @@ -975,6 +1022,12 @@ extern int nv50_graph_load_context(struct nouveau_channel *); extern int nv50_graph_unload_context(struct drm_device *); extern void nv50_graph_context_switch(struct drm_device *); +/* nouveau_grctx.c */ +extern int nouveau_grctx_prog_load(struct drm_device *); +extern void nouveau_grctx_vals_load(struct drm_device *, + struct nouveau_gpuobj *); +extern void nouveau_grctx_fini(struct drm_device *); + /* nv04_instmem.c */ extern int nv04_instmem_init(struct drm_device *); extern void nv04_instmem_takedown(struct drm_device *); @@ -1023,8 +1076,7 @@ extern long nouveau_compat_ioctl(struct file *file, unsigned int cmd, /* nv04_dac.c */ extern int nv04_dac_create(struct drm_device *dev, struct dcb_entry *entry); -extern enum drm_connector_status nv17_dac_detect(struct drm_encoder *encoder, - struct drm_connector *connector); +extern uint32_t nv17_dac_sample_load(struct drm_encoder *encoder); extern int nv04_dac_output_offset(struct drm_encoder *encoder); extern void nv04_dac_update_dacclk(struct drm_encoder *encoder, bool enable); @@ -1042,9 +1094,6 @@ extern int nv04_tv_create(struct drm_device *dev, struct dcb_entry *entry); /* nv17_tv.c */ extern int nv17_tv_create(struct drm_device *dev, struct dcb_entry *entry); -extern enum drm_connector_status nv17_tv_detect(struct drm_encoder *encoder, - struct drm_connector *connector, - uint32_t pin_mask); /* nv04_display.c */ extern int nv04_display_create(struct drm_device *); @@ -1207,14 +1256,24 @@ static inline void nv_wo32(struct drm_device *dev, struct nouveau_gpuobj *obj, pci_name(d->pdev), ##arg) #ifndef NV_DEBUG_NOTRACE #define NV_DEBUG(d, fmt, arg...) do { \ - if (drm_debug) { \ + if (drm_debug & DRM_UT_DRIVER) { \ + NV_PRINTK(KERN_DEBUG, d, "%s:%d - " fmt, __func__, \ + __LINE__, ##arg); \ + } \ +} while (0) +#define NV_DEBUG_KMS(d, fmt, arg...) do { \ + if (drm_debug & DRM_UT_KMS) { \ NV_PRINTK(KERN_DEBUG, d, "%s:%d - " fmt, __func__, \ __LINE__, ##arg); \ } \ } while (0) #else #define NV_DEBUG(d, fmt, arg...) do { \ - if (drm_debug) \ + if (drm_debug & DRM_UT_DRIVER) \ + NV_PRINTK(KERN_DEBUG, d, fmt, ##arg); \ +} while (0) +#define NV_DEBUG_KMS(d, fmt, arg...) do { \ + if (drm_debug & DRM_UT_KMS) \ NV_PRINTK(KERN_DEBUG, d, fmt, ##arg); \ } while (0) #endif @@ -1273,14 +1332,14 @@ nv_two_reg_pll(struct drm_device *dev) return false; } -#define NV50_NVSW 0x0000506e -#define NV50_NVSW_DMA_SEMAPHORE 0x00000060 -#define NV50_NVSW_SEMAPHORE_OFFSET 0x00000064 -#define NV50_NVSW_SEMAPHORE_ACQUIRE 0x00000068 -#define NV50_NVSW_SEMAPHORE_RELEASE 0x0000006c -#define NV50_NVSW_DMA_VBLSEM 0x0000018c -#define NV50_NVSW_VBLSEM_OFFSET 0x00000400 -#define NV50_NVSW_VBLSEM_RELEASE_VALUE 0x00000404 -#define NV50_NVSW_VBLSEM_RELEASE 0x00000408 +#define NV_SW 0x0000506e +#define NV_SW_DMA_SEMAPHORE 0x00000060 +#define NV_SW_SEMAPHORE_OFFSET 0x00000064 +#define NV_SW_SEMAPHORE_ACQUIRE 0x00000068 +#define NV_SW_SEMAPHORE_RELEASE 0x0000006c +#define NV_SW_DMA_VBLSEM 0x0000018c +#define NV_SW_VBLSEM_OFFSET 0x00000400 +#define NV_SW_VBLSEM_RELEASE_VALUE 0x00000404 +#define NV_SW_VBLSEM_RELEASE 0x00000408 #endif /* __NOUVEAU_DRV_H__ */ diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c index 36e8c5e4503a..0b05c869e0e7 100644 --- a/drivers/gpu/drm/nouveau/nouveau_fbcon.c +++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.c @@ -58,14 +58,13 @@ nouveau_fbcon_sync(struct fb_info *info) struct nouveau_channel *chan = dev_priv->channel; int ret, i; - if (!chan->accel_done || + if (!chan || !chan->accel_done || info->state != FBINFO_STATE_RUNNING || info->flags & FBINFO_HWACCEL_DISABLED) return 0; if (RING_SPACE(chan, 4)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); return 0; } @@ -86,8 +85,7 @@ nouveau_fbcon_sync(struct fb_info *info) } if (ret) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); return 0; } @@ -212,11 +210,11 @@ nouveau_fbcon_create(struct drm_device *dev, uint32_t fb_width, mode_cmd.bpp = surface_bpp; mode_cmd.pitch = mode_cmd.width * (mode_cmd.bpp >> 3); - mode_cmd.pitch = ALIGN(mode_cmd.pitch, 256); + mode_cmd.pitch = roundup(mode_cmd.pitch, 256); mode_cmd.depth = surface_depth; size = mode_cmd.pitch * mode_cmd.height; - size = ALIGN(size, PAGE_SIZE); + size = roundup(size, PAGE_SIZE); ret = nouveau_gem_new(dev, dev_priv->channel, size, 0, TTM_PL_FLAG_VRAM, 0, 0x0000, false, true, &nvbo); @@ -318,14 +316,16 @@ nouveau_fbcon_create(struct drm_device *dev, uint32_t fb_width, par->nouveau_fb = nouveau_fb; par->dev = dev; - switch (dev_priv->card_type) { - case NV_50: - nv50_fbcon_accel_init(info); - break; - default: - nv04_fbcon_accel_init(info); - break; - }; + if (dev_priv->channel) { + switch (dev_priv->card_type) { + case NV_50: + nv50_fbcon_accel_init(info); + break; + default: + nv04_fbcon_accel_init(info); + break; + }; + } nouveau_fbcon_zfill(dev); @@ -347,7 +347,7 @@ out: int nouveau_fbcon_probe(struct drm_device *dev) { - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); return drm_fb_helper_single_fb_probe(dev, 32, nouveau_fbcon_create); } @@ -378,3 +378,12 @@ nouveau_fbcon_remove(struct drm_device *dev, struct drm_framebuffer *fb) return 0; } + +void nouveau_fbcon_gpu_lockup(struct fb_info *info) +{ + struct nouveau_fbcon_par *par = info->par; + struct drm_device *dev = par->dev; + + NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); + info->flags |= FBINFO_HWACCEL_DISABLED; +} diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.h b/drivers/gpu/drm/nouveau/nouveau_fbcon.h index 8531140fedbc..462e0b87b4bd 100644 --- a/drivers/gpu/drm/nouveau/nouveau_fbcon.h +++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.h @@ -43,5 +43,6 @@ void nouveau_fbcon_zfill(struct drm_device *dev); int nv04_fbcon_accel_init(struct fb_info *info); int nv50_fbcon_accel_init(struct fb_info *info); +void nouveau_fbcon_gpu_lockup(struct fb_info *info); #endif /* __NV50_FBCON_H__ */ diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c b/drivers/gpu/drm/nouveau/nouveau_fence.c index 0cff7eb3690a..faddf53ff9ed 100644 --- a/drivers/gpu/drm/nouveau/nouveau_fence.c +++ b/drivers/gpu/drm/nouveau/nouveau_fence.c @@ -142,7 +142,7 @@ nouveau_fence_emit(struct nouveau_fence *fence) list_add_tail(&fence->entry, &chan->fence.pending); spin_unlock_irqrestore(&chan->fence.lock, flags); - BEGIN_RING(chan, NvSubM2MF, USE_REFCNT ? 0x0050 : 0x0150, 1); + BEGIN_RING(chan, NvSubSw, USE_REFCNT ? 0x0050 : 0x0150, 1); OUT_RING(chan, fence->sequence); FIRE_RING(chan); @@ -205,7 +205,7 @@ nouveau_fence_wait(void *sync_obj, void *sync_arg, bool lazy, bool intr) schedule_timeout(1); if (intr && signal_pending(current)) { - ret = -ERESTART; + ret = -ERESTARTSYS; break; } } diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c index 11f831f0ddc5..2009db2426c3 100644 --- a/drivers/gpu/drm/nouveau/nouveau_gem.c +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c @@ -220,7 +220,6 @@ nouveau_gem_set_domain(struct drm_gem_object *gem, uint32_t read_domains, } struct validate_op { - struct nouveau_fence *fence; struct list_head vram_list; struct list_head gart_list; struct list_head both_list; @@ -252,17 +251,11 @@ validate_fini_list(struct list_head *list, struct nouveau_fence *fence) } static void -validate_fini(struct validate_op *op, bool success) +validate_fini(struct validate_op *op, struct nouveau_fence* fence) { - struct nouveau_fence *fence = op->fence; - - if (unlikely(!success)) - op->fence = NULL; - - validate_fini_list(&op->vram_list, op->fence); - validate_fini_list(&op->gart_list, op->fence); - validate_fini_list(&op->both_list, op->fence); - nouveau_fence_unref((void *)&fence); + validate_fini_list(&op->vram_list, fence); + validate_fini_list(&op->gart_list, fence); + validate_fini_list(&op->both_list, fence); } static int @@ -342,8 +335,6 @@ retry: } ret = ttm_bo_wait_cpu(&nvbo->bo, false); - if (ret == -ERESTART) - ret = -EAGAIN; if (ret) return ret; goto retry; @@ -422,10 +413,6 @@ nouveau_gem_pushbuf_validate(struct nouveau_channel *chan, INIT_LIST_HEAD(&op->gart_list); INIT_LIST_HEAD(&op->both_list); - ret = nouveau_fence_new(chan, &op->fence, false); - if (ret) - return ret; - if (nr_buffers == 0) return 0; @@ -543,6 +530,7 @@ nouveau_gem_ioctl_pushbuf(struct drm_device *dev, void *data, struct drm_nouveau_gem_pushbuf_bo *bo = NULL; struct nouveau_channel *chan; struct validate_op op; + struct nouveau_fence* fence = 0; uint32_t *pushbuf = NULL; int ret = 0, do_reloc = 0, i; @@ -599,7 +587,7 @@ nouveau_gem_ioctl_pushbuf(struct drm_device *dev, void *data, OUT_RINGp(chan, pushbuf, req->nr_dwords); - ret = nouveau_fence_emit(op.fence); + ret = nouveau_fence_new(chan, &fence, true); if (ret) { NV_ERROR(dev, "error fencing pushbuf: %d\n", ret); WIND_RING(chan); @@ -607,7 +595,7 @@ nouveau_gem_ioctl_pushbuf(struct drm_device *dev, void *data, } if (nouveau_gem_pushbuf_sync(chan)) { - ret = nouveau_fence_wait(op.fence, NULL, false, false); + ret = nouveau_fence_wait(fence, NULL, false, false); if (ret) { for (i = 0; i < req->nr_dwords; i++) NV_ERROR(dev, "0x%08x\n", pushbuf[i]); @@ -616,7 +604,8 @@ nouveau_gem_ioctl_pushbuf(struct drm_device *dev, void *data, } out: - validate_fini(&op, ret == 0); + validate_fini(&op, fence); + nouveau_fence_unref((void**)&fence); mutex_unlock(&dev->struct_mutex); kfree(pushbuf); kfree(bo); @@ -636,6 +625,7 @@ nouveau_gem_ioctl_pushbuf_call(struct drm_device *dev, void *data, struct drm_gem_object *gem; struct nouveau_bo *pbbo; struct validate_op op; + struct nouveau_fence* fence = 0; int i, ret = 0, do_reloc = 0; NOUVEAU_CHECK_INITIALISED_WITH_RETURN; @@ -774,7 +764,7 @@ nouveau_gem_ioctl_pushbuf_call(struct drm_device *dev, void *data, OUT_RING(chan, 0); } - ret = nouveau_fence_emit(op.fence); + ret = nouveau_fence_new(chan, &fence, true); if (ret) { NV_ERROR(dev, "error fencing pushbuf: %d\n", ret); WIND_RING(chan); @@ -782,7 +772,8 @@ nouveau_gem_ioctl_pushbuf_call(struct drm_device *dev, void *data, } out: - validate_fini(&op, ret == 0); + validate_fini(&op, fence); + nouveau_fence_unref((void**)&fence); mutex_unlock(&dev->struct_mutex); kfree(bo); @@ -915,8 +906,6 @@ nouveau_gem_ioctl_cpu_prep(struct drm_device *dev, void *data, goto out; ret = ttm_bo_wait_cpu(&nvbo->bo, no_wait); - if (ret == -ERESTART) - ret = -EAGAIN; if (ret) goto out; } @@ -925,9 +914,6 @@ nouveau_gem_ioctl_cpu_prep(struct drm_device *dev, void *data, ret = ttm_bo_wait(&nvbo->bo, false, false, no_wait); } else { ret = ttm_bo_synccpu_write_grab(&nvbo->bo, no_wait); - if (ret == -ERESTART) - ret = -EAGAIN; - else if (ret == 0) nvbo->cpu_filp = file_priv; } diff --git a/drivers/gpu/drm/nouveau/nouveau_grctx.c b/drivers/gpu/drm/nouveau/nouveau_grctx.c new file mode 100644 index 000000000000..419f4c2b3b89 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nouveau_grctx.c @@ -0,0 +1,161 @@ +/* + * Copyright 2009 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#include <linux/firmware.h> + +#include "drmP.h" +#include "nouveau_drv.h" + +struct nouveau_ctxprog { + uint32_t signature; + uint8_t version; + uint16_t length; + uint32_t data[]; +} __attribute__ ((packed)); + +struct nouveau_ctxvals { + uint32_t signature; + uint8_t version; + uint32_t length; + struct { + uint32_t offset; + uint32_t value; + } data[]; +} __attribute__ ((packed)); + +int +nouveau_grctx_prog_load(struct drm_device *dev) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; + const int chipset = dev_priv->chipset; + const struct firmware *fw; + const struct nouveau_ctxprog *cp; + const struct nouveau_ctxvals *cv; + char name[32]; + int ret, i; + + if (pgraph->accel_blocked) + return -ENODEV; + + if (!pgraph->ctxprog) { + sprintf(name, "nouveau/nv%02x.ctxprog", chipset); + ret = request_firmware(&fw, name, &dev->pdev->dev); + if (ret) { + NV_ERROR(dev, "No ctxprog for NV%02x\n", chipset); + return ret; + } + + pgraph->ctxprog = kmalloc(fw->size, GFP_KERNEL); + if (!pgraph->ctxprog) { + NV_ERROR(dev, "OOM copying ctxprog\n"); + release_firmware(fw); + return -ENOMEM; + } + memcpy(pgraph->ctxprog, fw->data, fw->size); + + cp = pgraph->ctxprog; + if (le32_to_cpu(cp->signature) != 0x5043564e || + cp->version != 0 || + le16_to_cpu(cp->length) != ((fw->size - 7) / 4)) { + NV_ERROR(dev, "ctxprog invalid\n"); + release_firmware(fw); + nouveau_grctx_fini(dev); + return -EINVAL; + } + release_firmware(fw); + } + + if (!pgraph->ctxvals) { + sprintf(name, "nouveau/nv%02x.ctxvals", chipset); + ret = request_firmware(&fw, name, &dev->pdev->dev); + if (ret) { + NV_ERROR(dev, "No ctxvals for NV%02x\n", chipset); + nouveau_grctx_fini(dev); + return ret; + } + + pgraph->ctxvals = kmalloc(fw->size, GFP_KERNEL); + if (!pgraph->ctxprog) { + NV_ERROR(dev, "OOM copying ctxprog\n"); + release_firmware(fw); + nouveau_grctx_fini(dev); + return -ENOMEM; + } + memcpy(pgraph->ctxvals, fw->data, fw->size); + + cv = (void *)pgraph->ctxvals; + if (le32_to_cpu(cv->signature) != 0x5643564e || + cv->version != 0 || + le32_to_cpu(cv->length) != ((fw->size - 9) / 8)) { + NV_ERROR(dev, "ctxvals invalid\n"); + release_firmware(fw); + nouveau_grctx_fini(dev); + return -EINVAL; + } + release_firmware(fw); + } + + cp = pgraph->ctxprog; + + nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0); + for (i = 0; i < le16_to_cpu(cp->length); i++) + nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, + le32_to_cpu(cp->data[i])); + + return 0; +} + +void +nouveau_grctx_fini(struct drm_device *dev) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; + + if (pgraph->ctxprog) { + kfree(pgraph->ctxprog); + pgraph->ctxprog = NULL; + } + + if (pgraph->ctxvals) { + kfree(pgraph->ctxprog); + pgraph->ctxvals = NULL; + } +} + +void +nouveau_grctx_vals_load(struct drm_device *dev, struct nouveau_gpuobj *ctx) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; + struct nouveau_ctxvals *cv = pgraph->ctxvals; + int i; + + if (!cv) + return; + + for (i = 0; i < le32_to_cpu(cv->length); i++) + nv_wo32(dev, ctx, le32_to_cpu(cv->data[i].offset), + le32_to_cpu(cv->data[i].value)); +} diff --git a/drivers/gpu/drm/nouveau/nouveau_grctx.h b/drivers/gpu/drm/nouveau/nouveau_grctx.h new file mode 100644 index 000000000000..5d39c4ce8006 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nouveau_grctx.h @@ -0,0 +1,133 @@ +#ifndef __NOUVEAU_GRCTX_H__ +#define __NOUVEAU_GRCTX_H__ + +struct nouveau_grctx { + struct drm_device *dev; + + enum { + NOUVEAU_GRCTX_PROG, + NOUVEAU_GRCTX_VALS + } mode; + void *data; + + uint32_t ctxprog_max; + uint32_t ctxprog_len; + uint32_t ctxprog_reg; + int ctxprog_label[32]; + uint32_t ctxvals_pos; + uint32_t ctxvals_base; +}; + +#ifdef CP_CTX +static inline void +cp_out(struct nouveau_grctx *ctx, uint32_t inst) +{ + uint32_t *ctxprog = ctx->data; + + if (ctx->mode != NOUVEAU_GRCTX_PROG) + return; + + BUG_ON(ctx->ctxprog_len == ctx->ctxprog_max); + ctxprog[ctx->ctxprog_len++] = inst; +} + +static inline void +cp_lsr(struct nouveau_grctx *ctx, uint32_t val) +{ + cp_out(ctx, CP_LOAD_SR | val); +} + +static inline void +cp_ctx(struct nouveau_grctx *ctx, uint32_t reg, uint32_t length) +{ + ctx->ctxprog_reg = (reg - 0x00400000) >> 2; + + ctx->ctxvals_base = ctx->ctxvals_pos; + ctx->ctxvals_pos = ctx->ctxvals_base + length; + + if (length > (CP_CTX_COUNT >> CP_CTX_COUNT_SHIFT)) { + cp_lsr(ctx, length); + length = 0; + } + + cp_out(ctx, CP_CTX | (length << CP_CTX_COUNT_SHIFT) | ctx->ctxprog_reg); +} + +static inline void +cp_name(struct nouveau_grctx *ctx, int name) +{ + uint32_t *ctxprog = ctx->data; + int i; + + if (ctx->mode != NOUVEAU_GRCTX_PROG) + return; + + ctx->ctxprog_label[name] = ctx->ctxprog_len; + for (i = 0; i < ctx->ctxprog_len; i++) { + if ((ctxprog[i] & 0xfff00000) != 0xff400000) + continue; + if ((ctxprog[i] & CP_BRA_IP) != ((name) << CP_BRA_IP_SHIFT)) + continue; + ctxprog[i] = (ctxprog[i] & 0x00ff00ff) | + (ctx->ctxprog_len << CP_BRA_IP_SHIFT); + } +} + +static inline void +_cp_bra(struct nouveau_grctx *ctx, u32 mod, int flag, int state, int name) +{ + int ip = 0; + + if (mod != 2) { + ip = ctx->ctxprog_label[name] << CP_BRA_IP_SHIFT; + if (ip == 0) + ip = 0xff000000 | (name << CP_BRA_IP_SHIFT); + } + + cp_out(ctx, CP_BRA | (mod << 18) | ip | flag | + (state ? 0 : CP_BRA_IF_CLEAR)); +} +#define cp_bra(c,f,s,n) _cp_bra((c), 0, CP_FLAG_##f, CP_FLAG_##f##_##s, n) +#ifdef CP_BRA_MOD +#define cp_cal(c,f,s,n) _cp_bra((c), 1, CP_FLAG_##f, CP_FLAG_##f##_##s, n) +#define cp_ret(c,f,s) _cp_bra((c), 2, CP_FLAG_##f, CP_FLAG_##f##_##s, 0) +#endif + +static inline void +_cp_wait(struct nouveau_grctx *ctx, int flag, int state) +{ + cp_out(ctx, CP_WAIT | flag | (state ? CP_WAIT_SET : 0)); +} +#define cp_wait(c,f,s) _cp_wait((c), CP_FLAG_##f, CP_FLAG_##f##_##s) + +static inline void +_cp_set(struct nouveau_grctx *ctx, int flag, int state) +{ + cp_out(ctx, CP_SET | flag | (state ? CP_SET_1 : 0)); +} +#define cp_set(c,f,s) _cp_set((c), CP_FLAG_##f, CP_FLAG_##f##_##s) + +static inline void +cp_pos(struct nouveau_grctx *ctx, int offset) +{ + ctx->ctxvals_pos = offset; + ctx->ctxvals_base = ctx->ctxvals_pos; + + cp_lsr(ctx, ctx->ctxvals_pos); + cp_out(ctx, CP_SET_CONTEXT_POINTER); +} + +static inline void +gr_def(struct nouveau_grctx *ctx, uint32_t reg, uint32_t val) +{ + if (ctx->mode != NOUVEAU_GRCTX_VALS) + return; + + reg = (reg - 0x00400000) / 4; + reg = (reg - ctx->ctxprog_reg) + ctx->ctxvals_base; + + nv_wo32(ctx->dev, ctx->data, reg, val); +} +#endif + +#endif diff --git a/drivers/gpu/drm/nouveau/nouveau_ioc32.c b/drivers/gpu/drm/nouveau/nouveau_ioc32.c index a2c30f4611ba..475ba810bba3 100644 --- a/drivers/gpu/drm/nouveau/nouveau_ioc32.c +++ b/drivers/gpu/drm/nouveau/nouveau_ioc32.c @@ -61,12 +61,10 @@ long nouveau_compat_ioctl(struct file *filp, unsigned int cmd, if (nr < DRM_COMMAND_BASE + DRM_ARRAY_SIZE(mga_compat_ioctls)) fn = nouveau_compat_ioctls[nr - DRM_COMMAND_BASE]; #endif - lock_kernel(); /* XXX for now */ if (fn != NULL) ret = (*fn)(filp, cmd, arg); else - ret = drm_ioctl(filp->f_dentry->d_inode, filp, cmd, arg); - unlock_kernel(); + ret = drm_ioctl(filp, cmd, arg); return ret; } diff --git a/drivers/gpu/drm/nouveau/nouveau_irq.c b/drivers/gpu/drm/nouveau/nouveau_irq.c index 370c72c968d1..919a619ca7fa 100644 --- a/drivers/gpu/drm/nouveau/nouveau_irq.c +++ b/drivers/gpu/drm/nouveau/nouveau_irq.c @@ -635,6 +635,7 @@ nv50_pgraph_irq_handler(struct drm_device *dev) if ((nv_rd32(dev, 0x400500) & isb) != isb) nv_wr32(dev, 0x400500, nv_rd32(dev, 0x400500) | isb); + nv_wr32(dev, 0x400824, nv_rd32(dev, 0x400824) & ~(1 << 31)); } nv_wr32(dev, NV03_PMC_INTR_0, NV_PMC_INTR_0_PGRAPH_PENDING); diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.c b/drivers/gpu/drm/nouveau/nouveau_mem.c index 02755712ed3d..fb9bdd6edf1f 100644 --- a/drivers/gpu/drm/nouveau/nouveau_mem.c +++ b/drivers/gpu/drm/nouveau/nouveau_mem.c @@ -192,6 +192,92 @@ void nouveau_mem_release(struct drm_file *file_priv, struct mem_block *heap) } /* + * NV10-NV40 tiling helpers + */ + +static void +nv10_mem_set_region_tiling(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo; + struct nouveau_fb_engine *pfb = &dev_priv->engine.fb; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; + struct nouveau_tile_reg *tile = &dev_priv->tile.reg[i]; + + tile->addr = addr; + tile->size = size; + tile->used = !!pitch; + nouveau_fence_unref((void **)&tile->fence); + + if (!pfifo->cache_flush(dev)) + return; + + pfifo->reassign(dev, false); + pfifo->cache_flush(dev); + pfifo->cache_pull(dev, false); + + nouveau_wait_for_idle(dev); + + pgraph->set_region_tiling(dev, i, addr, size, pitch); + pfb->set_region_tiling(dev, i, addr, size, pitch); + + pfifo->cache_pull(dev, true); + pfifo->reassign(dev, true); +} + +struct nouveau_tile_reg * +nv10_mem_set_tiling(struct drm_device *dev, uint32_t addr, uint32_t size, + uint32_t pitch) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_fb_engine *pfb = &dev_priv->engine.fb; + struct nouveau_tile_reg *tile = dev_priv->tile.reg, *found = NULL; + int i; + + spin_lock(&dev_priv->tile.lock); + + for (i = 0; i < pfb->num_tiles; i++) { + if (tile[i].used) + /* Tile region in use. */ + continue; + + if (tile[i].fence && + !nouveau_fence_signalled(tile[i].fence, NULL)) + /* Pending tile region. */ + continue; + + if (max(tile[i].addr, addr) < + min(tile[i].addr + tile[i].size, addr + size)) + /* Kill an intersecting tile region. */ + nv10_mem_set_region_tiling(dev, i, 0, 0, 0); + + if (pitch && !found) { + /* Free tile region. */ + nv10_mem_set_region_tiling(dev, i, addr, size, pitch); + found = &tile[i]; + } + } + + spin_unlock(&dev_priv->tile.lock); + + return found; +} + +void +nv10_mem_expire_tiling(struct drm_device *dev, struct nouveau_tile_reg *tile, + struct nouveau_fence *fence) +{ + if (fence) { + /* Mark it as pending. */ + tile->fence = fence; + nouveau_fence_ref(fence); + } + + tile->used = false; +} + +/* * NV50 VM helpers */ int @@ -407,6 +493,7 @@ uint64_t nouveau_mem_fb_amount(struct drm_device *dev) return 0; } +#if __OS_HAS_AGP static void nouveau_mem_reset_agp(struct drm_device *dev) { uint32_t saved_pci_nv_1, saved_pci_nv_19, pmc_enable; @@ -432,10 +519,12 @@ static void nouveau_mem_reset_agp(struct drm_device *dev) nv_wr32(dev, NV04_PBUS_PCI_NV_19, saved_pci_nv_19); nv_wr32(dev, NV04_PBUS_PCI_NV_1, saved_pci_nv_1); } +#endif int nouveau_mem_init_agp(struct drm_device *dev) { +#if __OS_HAS_AGP struct drm_nouveau_private *dev_priv = dev->dev_private; struct drm_agp_info info; struct drm_agp_mode mode; @@ -471,6 +560,7 @@ nouveau_mem_init_agp(struct drm_device *dev) dev_priv->gart_info.type = NOUVEAU_GART_AGP; dev_priv->gart_info.aper_base = info.aperture_base; dev_priv->gart_info.aper_size = info.aperture_size; +#endif return 0; } @@ -509,6 +599,7 @@ nouveau_mem_init(struct drm_device *dev) INIT_LIST_HEAD(&dev_priv->ttm.bo_list); spin_lock_init(&dev_priv->ttm.bo_list_lock); + spin_lock_init(&dev_priv->tile.lock); dev_priv->fb_available_size = nouveau_mem_fb_amount(dev); diff --git a/drivers/gpu/drm/nouveau/nouveau_object.c b/drivers/gpu/drm/nouveau/nouveau_object.c index 93379bb81bea..6c2cf81716df 100644 --- a/drivers/gpu/drm/nouveau/nouveau_object.c +++ b/drivers/gpu/drm/nouveau/nouveau_object.c @@ -881,7 +881,7 @@ nouveau_gpuobj_gr_new(struct nouveau_channel *chan, int class, return 0; } -static int +int nouveau_gpuobj_sw_new(struct nouveau_channel *chan, int class, struct nouveau_gpuobj **gpuobj_ret) { diff --git a/drivers/gpu/drm/nouveau/nouveau_reg.h b/drivers/gpu/drm/nouveau/nouveau_reg.h index fa1b0e7165b9..251f1b3b38b9 100644 --- a/drivers/gpu/drm/nouveau/nouveau_reg.h +++ b/drivers/gpu/drm/nouveau/nouveau_reg.h @@ -349,19 +349,19 @@ #define NV04_PGRAPH_BLEND 0x00400824 #define NV04_PGRAPH_STORED_FMT 0x00400830 #define NV04_PGRAPH_PATT_COLORRAM 0x00400900 -#define NV40_PGRAPH_TILE0(i) (0x00400900 + (i*16)) -#define NV40_PGRAPH_TLIMIT0(i) (0x00400904 + (i*16)) -#define NV40_PGRAPH_TSIZE0(i) (0x00400908 + (i*16)) -#define NV40_PGRAPH_TSTATUS0(i) (0x0040090C + (i*16)) +#define NV20_PGRAPH_TILE(i) (0x00400900 + (i*16)) +#define NV20_PGRAPH_TLIMIT(i) (0x00400904 + (i*16)) +#define NV20_PGRAPH_TSIZE(i) (0x00400908 + (i*16)) +#define NV20_PGRAPH_TSTATUS(i) (0x0040090C + (i*16)) #define NV10_PGRAPH_TILE(i) (0x00400B00 + (i*16)) #define NV10_PGRAPH_TLIMIT(i) (0x00400B04 + (i*16)) #define NV10_PGRAPH_TSIZE(i) (0x00400B08 + (i*16)) #define NV10_PGRAPH_TSTATUS(i) (0x00400B0C + (i*16)) #define NV04_PGRAPH_U_RAM 0x00400D00 -#define NV47_PGRAPH_TILE0(i) (0x00400D00 + (i*16)) -#define NV47_PGRAPH_TLIMIT0(i) (0x00400D04 + (i*16)) -#define NV47_PGRAPH_TSIZE0(i) (0x00400D08 + (i*16)) -#define NV47_PGRAPH_TSTATUS0(i) (0x00400D0C + (i*16)) +#define NV47_PGRAPH_TILE(i) (0x00400D00 + (i*16)) +#define NV47_PGRAPH_TLIMIT(i) (0x00400D04 + (i*16)) +#define NV47_PGRAPH_TSIZE(i) (0x00400D08 + (i*16)) +#define NV47_PGRAPH_TSTATUS(i) (0x00400D0C + (i*16)) #define NV04_PGRAPH_V_RAM 0x00400D40 #define NV04_PGRAPH_W_RAM 0x00400D80 #define NV10_PGRAPH_COMBINER0_IN_ALPHA 0x00400E40 diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c index 2ed41d339f6a..09b9a46dfc0e 100644 --- a/drivers/gpu/drm/nouveau/nouveau_state.c +++ b/drivers/gpu/drm/nouveau/nouveau_state.c @@ -76,6 +76,8 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->fifo.disable = nv04_fifo_disable; engine->fifo.enable = nv04_fifo_enable; engine->fifo.reassign = nv04_fifo_reassign; + engine->fifo.cache_flush = nv04_fifo_cache_flush; + engine->fifo.cache_pull = nv04_fifo_cache_pull; engine->fifo.channel_id = nv04_fifo_channel_id; engine->fifo.create_context = nv04_fifo_create_context; engine->fifo.destroy_context = nv04_fifo_destroy_context; @@ -100,6 +102,7 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->timer.takedown = nv04_timer_takedown; engine->fb.init = nv10_fb_init; engine->fb.takedown = nv10_fb_takedown; + engine->fb.set_region_tiling = nv10_fb_set_region_tiling; engine->graph.grclass = nv10_graph_grclass; engine->graph.init = nv10_graph_init; engine->graph.takedown = nv10_graph_takedown; @@ -109,12 +112,15 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->graph.fifo_access = nv04_graph_fifo_access; engine->graph.load_context = nv10_graph_load_context; engine->graph.unload_context = nv10_graph_unload_context; + engine->graph.set_region_tiling = nv10_graph_set_region_tiling; engine->fifo.channels = 32; engine->fifo.init = nv10_fifo_init; engine->fifo.takedown = nouveau_stub_takedown; engine->fifo.disable = nv04_fifo_disable; engine->fifo.enable = nv04_fifo_enable; engine->fifo.reassign = nv04_fifo_reassign; + engine->fifo.cache_flush = nv04_fifo_cache_flush; + engine->fifo.cache_pull = nv04_fifo_cache_pull; engine->fifo.channel_id = nv10_fifo_channel_id; engine->fifo.create_context = nv10_fifo_create_context; engine->fifo.destroy_context = nv10_fifo_destroy_context; @@ -139,6 +145,7 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->timer.takedown = nv04_timer_takedown; engine->fb.init = nv10_fb_init; engine->fb.takedown = nv10_fb_takedown; + engine->fb.set_region_tiling = nv10_fb_set_region_tiling; engine->graph.grclass = nv20_graph_grclass; engine->graph.init = nv20_graph_init; engine->graph.takedown = nv20_graph_takedown; @@ -148,12 +155,15 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->graph.fifo_access = nv04_graph_fifo_access; engine->graph.load_context = nv20_graph_load_context; engine->graph.unload_context = nv20_graph_unload_context; + engine->graph.set_region_tiling = nv20_graph_set_region_tiling; engine->fifo.channels = 32; engine->fifo.init = nv10_fifo_init; engine->fifo.takedown = nouveau_stub_takedown; engine->fifo.disable = nv04_fifo_disable; engine->fifo.enable = nv04_fifo_enable; engine->fifo.reassign = nv04_fifo_reassign; + engine->fifo.cache_flush = nv04_fifo_cache_flush; + engine->fifo.cache_pull = nv04_fifo_cache_pull; engine->fifo.channel_id = nv10_fifo_channel_id; engine->fifo.create_context = nv10_fifo_create_context; engine->fifo.destroy_context = nv10_fifo_destroy_context; @@ -178,6 +188,7 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->timer.takedown = nv04_timer_takedown; engine->fb.init = nv10_fb_init; engine->fb.takedown = nv10_fb_takedown; + engine->fb.set_region_tiling = nv10_fb_set_region_tiling; engine->graph.grclass = nv30_graph_grclass; engine->graph.init = nv30_graph_init; engine->graph.takedown = nv20_graph_takedown; @@ -187,12 +198,15 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->graph.destroy_context = nv20_graph_destroy_context; engine->graph.load_context = nv20_graph_load_context; engine->graph.unload_context = nv20_graph_unload_context; + engine->graph.set_region_tiling = nv20_graph_set_region_tiling; engine->fifo.channels = 32; engine->fifo.init = nv10_fifo_init; engine->fifo.takedown = nouveau_stub_takedown; engine->fifo.disable = nv04_fifo_disable; engine->fifo.enable = nv04_fifo_enable; engine->fifo.reassign = nv04_fifo_reassign; + engine->fifo.cache_flush = nv04_fifo_cache_flush; + engine->fifo.cache_pull = nv04_fifo_cache_pull; engine->fifo.channel_id = nv10_fifo_channel_id; engine->fifo.create_context = nv10_fifo_create_context; engine->fifo.destroy_context = nv10_fifo_destroy_context; @@ -218,6 +232,7 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->timer.takedown = nv04_timer_takedown; engine->fb.init = nv40_fb_init; engine->fb.takedown = nv40_fb_takedown; + engine->fb.set_region_tiling = nv40_fb_set_region_tiling; engine->graph.grclass = nv40_graph_grclass; engine->graph.init = nv40_graph_init; engine->graph.takedown = nv40_graph_takedown; @@ -227,12 +242,15 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->graph.destroy_context = nv40_graph_destroy_context; engine->graph.load_context = nv40_graph_load_context; engine->graph.unload_context = nv40_graph_unload_context; + engine->graph.set_region_tiling = nv40_graph_set_region_tiling; engine->fifo.channels = 32; engine->fifo.init = nv40_fifo_init; engine->fifo.takedown = nouveau_stub_takedown; engine->fifo.disable = nv04_fifo_disable; engine->fifo.enable = nv04_fifo_enable; engine->fifo.reassign = nv04_fifo_reassign; + engine->fifo.cache_flush = nv04_fifo_cache_flush; + engine->fifo.cache_pull = nv04_fifo_cache_pull; engine->fifo.channel_id = nv10_fifo_channel_id; engine->fifo.create_context = nv40_fifo_create_context; engine->fifo.destroy_context = nv40_fifo_destroy_context; @@ -299,12 +317,57 @@ nouveau_vga_set_decode(void *priv, bool state) return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM; } +static int +nouveau_card_init_channel(struct drm_device *dev) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_gpuobj *gpuobj; + int ret; + + ret = nouveau_channel_alloc(dev, &dev_priv->channel, + (struct drm_file *)-2, + NvDmaFB, NvDmaTT); + if (ret) + return ret; + + gpuobj = NULL; + ret = nouveau_gpuobj_dma_new(dev_priv->channel, NV_CLASS_DMA_IN_MEMORY, + 0, nouveau_mem_fb_amount(dev), + NV_DMA_ACCESS_RW, NV_DMA_TARGET_VIDMEM, + &gpuobj); + if (ret) + goto out_err; + + ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, NvDmaVRAM, + gpuobj, NULL); + if (ret) + goto out_err; + + gpuobj = NULL; + ret = nouveau_gpuobj_gart_dma_new(dev_priv->channel, 0, + dev_priv->gart_info.aper_size, + NV_DMA_ACCESS_RW, &gpuobj, NULL); + if (ret) + goto out_err; + + ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, NvDmaGART, + gpuobj, NULL); + if (ret) + goto out_err; + + return 0; +out_err: + nouveau_gpuobj_del(dev, &gpuobj); + nouveau_channel_free(dev_priv->channel); + dev_priv->channel = NULL; + return ret; +} + int nouveau_card_init(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_engine *engine; - struct nouveau_gpuobj *gpuobj; int ret; NV_DEBUG(dev, "prev state = %d\n", dev_priv->init_state); @@ -317,7 +380,7 @@ nouveau_card_init(struct drm_device *dev) /* Initialise internal driver API hooks */ ret = nouveau_init_engine_ptrs(dev); if (ret) - return ret; + goto out; engine = &dev_priv->engine; dev_priv->init_state = NOUVEAU_CARD_INIT_FAILED; @@ -325,12 +388,12 @@ nouveau_card_init(struct drm_device *dev) if (drm_core_check_feature(dev, DRIVER_MODESET)) { ret = nouveau_bios_init(dev); if (ret) - return ret; + goto out; } ret = nouveau_gpuobj_early_init(dev); if (ret) - return ret; + goto out_bios; /* Initialise instance memory, must happen before mem_init so we * know exactly how much VRAM we're able to use for "normal" @@ -338,100 +401,68 @@ nouveau_card_init(struct drm_device *dev) */ ret = engine->instmem.init(dev); if (ret) - return ret; + goto out_gpuobj_early; /* Setup the memory manager */ ret = nouveau_mem_init(dev); if (ret) - return ret; + goto out_instmem; ret = nouveau_gpuobj_init(dev); if (ret) - return ret; + goto out_mem; /* PMC */ ret = engine->mc.init(dev); if (ret) - return ret; + goto out_gpuobj; /* PTIMER */ ret = engine->timer.init(dev); if (ret) - return ret; + goto out_mc; /* PFB */ ret = engine->fb.init(dev); if (ret) - return ret; + goto out_timer; /* PGRAPH */ ret = engine->graph.init(dev); if (ret) - return ret; + goto out_fb; /* PFIFO */ ret = engine->fifo.init(dev); if (ret) - return ret; + goto out_graph; /* this call irq_preinstall, register irq handler and * call irq_postinstall */ ret = drm_irq_install(dev); if (ret) - return ret; + goto out_fifo; ret = drm_vblank_init(dev, 0); if (ret) - return ret; + goto out_irq; /* what about PVIDEO/PCRTC/PRAMDAC etc? */ - ret = nouveau_channel_alloc(dev, &dev_priv->channel, - (struct drm_file *)-2, - NvDmaFB, NvDmaTT); - if (ret) - return ret; - - gpuobj = NULL; - ret = nouveau_gpuobj_dma_new(dev_priv->channel, NV_CLASS_DMA_IN_MEMORY, - 0, nouveau_mem_fb_amount(dev), - NV_DMA_ACCESS_RW, NV_DMA_TARGET_VIDMEM, - &gpuobj); - if (ret) - return ret; - - ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, NvDmaVRAM, - gpuobj, NULL); - if (ret) { - nouveau_gpuobj_del(dev, &gpuobj); - return ret; - } - - gpuobj = NULL; - ret = nouveau_gpuobj_gart_dma_new(dev_priv->channel, 0, - dev_priv->gart_info.aper_size, - NV_DMA_ACCESS_RW, &gpuobj, NULL); - if (ret) - return ret; - - ret = nouveau_gpuobj_ref_add(dev, dev_priv->channel, NvDmaGART, - gpuobj, NULL); - if (ret) { - nouveau_gpuobj_del(dev, &gpuobj); - return ret; + if (!engine->graph.accel_blocked) { + ret = nouveau_card_init_channel(dev); + if (ret) + goto out_irq; } if (drm_core_check_feature(dev, DRIVER_MODESET)) { - if (dev_priv->card_type >= NV_50) { + if (dev_priv->card_type >= NV_50) ret = nv50_display_create(dev); - if (ret) - return ret; - } else { + else ret = nv04_display_create(dev); - if (ret) - return ret; - } + if (ret) + goto out_irq; } ret = nouveau_backlight_init(dev); @@ -444,6 +475,32 @@ nouveau_card_init(struct drm_device *dev) drm_helper_initial_config(dev); return 0; + +out_irq: + drm_irq_uninstall(dev); +out_fifo: + engine->fifo.takedown(dev); +out_graph: + engine->graph.takedown(dev); +out_fb: + engine->fb.takedown(dev); +out_timer: + engine->timer.takedown(dev); +out_mc: + engine->mc.takedown(dev); +out_gpuobj: + nouveau_gpuobj_takedown(dev); +out_mem: + nouveau_mem_close(dev); +out_instmem: + engine->instmem.takedown(dev); +out_gpuobj_early: + nouveau_gpuobj_late_takedown(dev); +out_bios: + nouveau_bios_takedown(dev); +out: + vga_client_register(dev->pdev, NULL, NULL, NULL); + return ret; } static void nouveau_card_takedown(struct drm_device *dev) @@ -585,7 +642,10 @@ int nouveau_load(struct drm_device *dev, unsigned long flags) dev_priv->chipset = (reg0 & 0xff00000) >> 20; /* NV04 or NV05 */ } else if ((reg0 & 0xff00fff0) == 0x20004000) { - dev_priv->chipset = 0x04; + if (reg0 & 0x00f00000) + dev_priv->chipset = 0x05; + else + dev_priv->chipset = 0x04; } else dev_priv->chipset = 0xff; @@ -665,8 +725,8 @@ static void nouveau_close(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - /* In the case of an error dev_priv may not be be allocated yet */ - if (dev_priv && dev_priv->card_type) + /* In the case of an error dev_priv may not be allocated yet */ + if (dev_priv) nouveau_card_takedown(dev); } diff --git a/drivers/gpu/drm/nouveau/nouveau_ttm.c b/drivers/gpu/drm/nouveau/nouveau_ttm.c index 187eb84e4da5..c385d50f041b 100644 --- a/drivers/gpu/drm/nouveau/nouveau_ttm.c +++ b/drivers/gpu/drm/nouveau/nouveau_ttm.c @@ -28,45 +28,17 @@ #include "nouveau_drv.h" -static struct vm_operations_struct nouveau_ttm_vm_ops; -static const struct vm_operations_struct *ttm_vm_ops; - -static int -nouveau_ttm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) -{ - struct ttm_buffer_object *bo = vma->vm_private_data; - int ret; - - if (unlikely(bo == NULL)) - return VM_FAULT_NOPAGE; - - ret = ttm_vm_ops->fault(vma, vmf); - return ret; -} - int nouveau_ttm_mmap(struct file *filp, struct vm_area_struct *vma) { struct drm_file *file_priv = filp->private_data; struct drm_nouveau_private *dev_priv = file_priv->minor->dev->dev_private; - int ret; if (unlikely(vma->vm_pgoff < DRM_FILE_PAGE_OFFSET)) return drm_mmap(filp, vma); - ret = ttm_bo_mmap(filp, vma, &dev_priv->ttm.bdev); - if (unlikely(ret != 0)) - return ret; - - if (unlikely(ttm_vm_ops == NULL)) { - ttm_vm_ops = vma->vm_ops; - nouveau_ttm_vm_ops = *ttm_vm_ops; - nouveau_ttm_vm_ops.fault = &nouveau_ttm_fault; - } - - vma->vm_ops = &nouveau_ttm_vm_ops; - return 0; + return ttm_bo_mmap(filp, vma, &dev_priv->ttm.bdev); } static int diff --git a/drivers/gpu/drm/nouveau/nv04_crtc.c b/drivers/gpu/drm/nouveau/nv04_crtc.c index b91363606055..d2f143ed97c1 100644 --- a/drivers/gpu/drm/nouveau/nv04_crtc.c +++ b/drivers/gpu/drm/nouveau/nv04_crtc.c @@ -143,10 +143,10 @@ static void nv_crtc_calc_state_ext(struct drm_crtc *crtc, struct drm_display_mod state->pllsel |= nv_crtc->index ? PLLSEL_VPLL2_MASK : PLLSEL_VPLL1_MASK; if (pv->NM2) - NV_TRACE(dev, "vpll: n1 %d n2 %d m1 %d m2 %d log2p %d\n", + NV_DEBUG_KMS(dev, "vpll: n1 %d n2 %d m1 %d m2 %d log2p %d\n", pv->N1, pv->N2, pv->M1, pv->M2, pv->log2P); else - NV_TRACE(dev, "vpll: n %d m %d log2p %d\n", + NV_DEBUG_KMS(dev, "vpll: n %d m %d log2p %d\n", pv->N1, pv->M1, pv->log2P); nv_crtc->cursor.set_offset(nv_crtc, nv_crtc->cursor.offset); @@ -160,7 +160,7 @@ nv_crtc_dpms(struct drm_crtc *crtc, int mode) unsigned char seq1 = 0, crtc17 = 0; unsigned char crtc1A; - NV_TRACE(dev, "Setting dpms mode %d on CRTC %d\n", mode, + NV_DEBUG_KMS(dev, "Setting dpms mode %d on CRTC %d\n", mode, nv_crtc->index); if (nv_crtc->last_dpms == mode) /* Don't do unnecesary mode changes. */ @@ -603,7 +603,7 @@ nv_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode, struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc); struct drm_nouveau_private *dev_priv = dev->dev_private; - NV_DEBUG(dev, "CTRC mode on CRTC %d:\n", nv_crtc->index); + NV_DEBUG_KMS(dev, "CTRC mode on CRTC %d:\n", nv_crtc->index); drm_mode_debug_printmodeline(adjusted_mode); /* unlock must come after turning off FP_TG_CONTROL in output_prepare */ @@ -703,7 +703,7 @@ static void nv_crtc_destroy(struct drm_crtc *crtc) { struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc); - NV_DEBUG(crtc->dev, "\n"); + NV_DEBUG_KMS(crtc->dev, "\n"); if (!nv_crtc) return; diff --git a/drivers/gpu/drm/nouveau/nv04_dac.c b/drivers/gpu/drm/nouveau/nv04_dac.c index a5fa51714e87..d0e038d28948 100644 --- a/drivers/gpu/drm/nouveau/nv04_dac.c +++ b/drivers/gpu/drm/nouveau/nv04_dac.c @@ -205,23 +205,22 @@ out: NVWriteVgaSeq(dev, 0, NV_VIO_SR_CLOCK_INDEX, saved_seq1); if (blue == 0x18) { - NV_TRACE(dev, "Load detected on head A\n"); + NV_INFO(dev, "Load detected on head A\n"); return connector_status_connected; } return connector_status_disconnected; } -enum drm_connector_status nv17_dac_detect(struct drm_encoder *encoder, - struct drm_connector *connector) +uint32_t nv17_dac_sample_load(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; struct dcb_entry *dcb = nouveau_encoder(encoder)->dcb; - uint32_t testval, regoffset = nv04_dac_output_offset(encoder); + uint32_t sample, testval, regoffset = nv04_dac_output_offset(encoder); uint32_t saved_powerctrl_2 = 0, saved_powerctrl_4 = 0, saved_routput, saved_rtest_ctrl, saved_gpio0, saved_gpio1, temp, routput; - int head, present = 0; + int head; #define RGB_TEST_DATA(r, g, b) (r << 0 | g << 10 | b << 20) if (dcb->type == OUTPUT_TV) { @@ -287,13 +286,7 @@ enum drm_connector_status nv17_dac_detect(struct drm_encoder *encoder, temp | NV_PRAMDAC_TEST_CONTROL_TP_INS_EN_ASSERTED); msleep(5); - temp = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset); - - if (dcb->type == OUTPUT_TV) - present = (nv17_tv_detect(encoder, connector, temp) - == connector_status_connected); - else - present = temp & NV_PRAMDAC_TEST_CONTROL_SENSEB_ALLHI; + sample = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset); temp = NVReadRAMDAC(dev, head, NV_PRAMDAC_TEST_CONTROL); NVWriteRAMDAC(dev, head, NV_PRAMDAC_TEST_CONTROL, @@ -310,15 +303,25 @@ enum drm_connector_status nv17_dac_detect(struct drm_encoder *encoder, nv17_gpio_set(dev, DCB_GPIO_TVDAC1, saved_gpio1); nv17_gpio_set(dev, DCB_GPIO_TVDAC0, saved_gpio0); - if (present) { - NV_INFO(dev, "Load detected on output %c\n", '@' + ffs(dcb->or)); + return sample; +} + +static enum drm_connector_status +nv17_dac_detect(struct drm_encoder *encoder, struct drm_connector *connector) +{ + struct drm_device *dev = encoder->dev; + struct dcb_entry *dcb = nouveau_encoder(encoder)->dcb; + uint32_t sample = nv17_dac_sample_load(encoder); + + if (sample & NV_PRAMDAC_TEST_CONTROL_SENSEB_ALLHI) { + NV_INFO(dev, "Load detected on output %c\n", + '@' + ffs(dcb->or)); return connector_status_connected; + } else { + return connector_status_disconnected; } - - return connector_status_disconnected; } - static bool nv04_dac_mode_fixup(struct drm_encoder *encoder, struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode) @@ -350,14 +353,10 @@ static void nv04_dac_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode) { - struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); struct drm_device *dev = encoder->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; int head = nouveau_crtc(encoder->crtc)->index; - NV_TRACE(dev, "%s called for encoder %d\n", __func__, - nv_encoder->dcb->index); - if (nv_gf4_disp_arch(dev)) { struct drm_encoder *rebind; uint32_t dac_offset = nv04_dac_output_offset(encoder); @@ -466,7 +465,7 @@ static void nv04_dac_destroy(struct drm_encoder *encoder) { struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); - NV_DEBUG(encoder->dev, "\n"); + NV_DEBUG_KMS(encoder->dev, "\n"); drm_encoder_cleanup(encoder); kfree(nv_encoder); diff --git a/drivers/gpu/drm/nouveau/nv04_dfp.c b/drivers/gpu/drm/nouveau/nv04_dfp.c index e5b33339d595..483f875bdb6a 100644 --- a/drivers/gpu/drm/nouveau/nv04_dfp.c +++ b/drivers/gpu/drm/nouveau/nv04_dfp.c @@ -261,7 +261,7 @@ static void nv04_dfp_mode_set(struct drm_encoder *encoder, struct drm_display_mode *output_mode = &nv_encoder->mode; uint32_t mode_ratio, panel_ratio; - NV_DEBUG(dev, "Output mode on CRTC %d:\n", nv_crtc->index); + NV_DEBUG_KMS(dev, "Output mode on CRTC %d:\n", nv_crtc->index); drm_mode_debug_printmodeline(output_mode); /* Initialize the FP registers in this CRTC. */ @@ -413,7 +413,9 @@ static void nv04_dfp_commit(struct drm_encoder *encoder) struct dcb_entry *dcbe = nv_encoder->dcb; int head = nouveau_crtc(encoder->crtc)->index; - NV_TRACE(dev, "%s called for encoder %d\n", __func__, nv_encoder->dcb->index); + NV_INFO(dev, "Output %s is running on CRTC %d using output %c\n", + drm_get_connector_name(&nouveau_encoder_connector_get(nv_encoder)->base), + nv_crtc->index, '@' + ffs(nv_encoder->dcb->or)); if (dcbe->type == OUTPUT_TMDS) run_tmds_table(dev, dcbe, head, nv_encoder->mode.clock); @@ -550,7 +552,7 @@ static void nv04_dfp_destroy(struct drm_encoder *encoder) { struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); - NV_DEBUG(encoder->dev, "\n"); + NV_DEBUG_KMS(encoder->dev, "\n"); drm_encoder_cleanup(encoder); kfree(nv_encoder); diff --git a/drivers/gpu/drm/nouveau/nv04_display.c b/drivers/gpu/drm/nouveau/nv04_display.c index b47c757ff48b..ef77215fa5b9 100644 --- a/drivers/gpu/drm/nouveau/nv04_display.c +++ b/drivers/gpu/drm/nouveau/nv04_display.c @@ -99,10 +99,11 @@ nv04_display_create(struct drm_device *dev) uint16_t connector[16] = { 0 }; int i, ret; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); if (nv_two_heads(dev)) nv04_display_store_initial_head_owner(dev); + nouveau_hw_save_vga_fonts(dev, 1); drm_mode_config_init(dev); drm_mode_create_scaling_mode_property(dev); @@ -203,8 +204,6 @@ nv04_display_create(struct drm_device *dev) /* Save previous state */ NVLockVgaCrtcs(dev, false); - nouveau_hw_save_vga_fonts(dev, 1); - list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) crtc->funcs->save(crtc); @@ -223,7 +222,7 @@ nv04_display_destroy(struct drm_device *dev) struct drm_encoder *encoder; struct drm_crtc *crtc; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); /* Turn every CRTC off. */ list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { @@ -246,9 +245,9 @@ nv04_display_destroy(struct drm_device *dev) list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) crtc->funcs->restore(crtc); - nouveau_hw_save_vga_fonts(dev, 0); - drm_mode_config_cleanup(dev); + + nouveau_hw_save_vga_fonts(dev, 0); } void diff --git a/drivers/gpu/drm/nouveau/nv04_fbcon.c b/drivers/gpu/drm/nouveau/nv04_fbcon.c index 09a31071ee58..d910873c1368 100644 --- a/drivers/gpu/drm/nouveau/nv04_fbcon.c +++ b/drivers/gpu/drm/nouveau/nv04_fbcon.c @@ -39,8 +39,7 @@ nv04_fbcon_copyarea(struct fb_info *info, const struct fb_copyarea *region) return; if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 4)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); } if (info->flags & FBINFO_HWACCEL_DISABLED) { @@ -62,14 +61,12 @@ nv04_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect) struct drm_device *dev = par->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_channel *chan = dev_priv->channel; - uint32_t color = ((uint32_t *) info->pseudo_palette)[rect->color]; if (info->state != FBINFO_STATE_RUNNING) return; if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 7)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); } if (info->flags & FBINFO_HWACCEL_DISABLED) { @@ -80,7 +77,11 @@ nv04_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect) BEGIN_RING(chan, NvSubGdiRect, 0x02fc, 1); OUT_RING(chan, (rect->rop != ROP_COPY) ? 1 : 3); BEGIN_RING(chan, NvSubGdiRect, 0x03fc, 1); - OUT_RING(chan, color); + if (info->fix.visual == FB_VISUAL_TRUECOLOR || + info->fix.visual == FB_VISUAL_DIRECTCOLOR) + OUT_RING(chan, ((uint32_t *)info->pseudo_palette)[rect->color]); + else + OUT_RING(chan, rect->color); BEGIN_RING(chan, NvSubGdiRect, 0x0400, 2); OUT_RING(chan, (rect->dx << 16) | rect->dy); OUT_RING(chan, (rect->width << 16) | rect->height); @@ -109,8 +110,7 @@ nv04_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) } if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 8)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); } if (info->flags & FBINFO_HWACCEL_DISABLED) { @@ -144,8 +144,7 @@ nv04_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) int iter_len = dsize > 128 ? 128 : dsize; if (RING_SPACE(chan, iter_len + 1)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); cfb_imageblit(info, image); return; } @@ -184,6 +183,7 @@ nv04_fbcon_accel_init(struct fb_info *info) struct drm_device *dev = par->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_channel *chan = dev_priv->channel; + const int sub = NvSubCtxSurf2D; int surface_fmt, pattern_fmt, rect_fmt; int ret; @@ -242,30 +242,29 @@ nv04_fbcon_accel_init(struct fb_info *info) return ret; if (RING_SPACE(chan, 49)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); return 0; } - BEGIN_RING(chan, 1, 0x0000, 1); + BEGIN_RING(chan, sub, 0x0000, 1); OUT_RING(chan, NvCtxSurf2D); - BEGIN_RING(chan, 1, 0x0184, 2); + BEGIN_RING(chan, sub, 0x0184, 2); OUT_RING(chan, NvDmaFB); OUT_RING(chan, NvDmaFB); - BEGIN_RING(chan, 1, 0x0300, 4); + BEGIN_RING(chan, sub, 0x0300, 4); OUT_RING(chan, surface_fmt); OUT_RING(chan, info->fix.line_length | (info->fix.line_length << 16)); OUT_RING(chan, info->fix.smem_start - dev->mode_config.fb_base); OUT_RING(chan, info->fix.smem_start - dev->mode_config.fb_base); - BEGIN_RING(chan, 1, 0x0000, 1); + BEGIN_RING(chan, sub, 0x0000, 1); OUT_RING(chan, NvRop); - BEGIN_RING(chan, 1, 0x0300, 1); + BEGIN_RING(chan, sub, 0x0300, 1); OUT_RING(chan, 0x55); - BEGIN_RING(chan, 1, 0x0000, 1); + BEGIN_RING(chan, sub, 0x0000, 1); OUT_RING(chan, NvImagePatt); - BEGIN_RING(chan, 1, 0x0300, 8); + BEGIN_RING(chan, sub, 0x0300, 8); OUT_RING(chan, pattern_fmt); #ifdef __BIG_ENDIAN OUT_RING(chan, 2); @@ -279,9 +278,9 @@ nv04_fbcon_accel_init(struct fb_info *info) OUT_RING(chan, ~0); OUT_RING(chan, ~0); - BEGIN_RING(chan, 1, 0x0000, 1); + BEGIN_RING(chan, sub, 0x0000, 1); OUT_RING(chan, NvClipRect); - BEGIN_RING(chan, 1, 0x0300, 2); + BEGIN_RING(chan, sub, 0x0300, 2); OUT_RING(chan, 0); OUT_RING(chan, (info->var.yres_virtual << 16) | info->var.xres_virtual); diff --git a/drivers/gpu/drm/nouveau/nv04_fifo.c b/drivers/gpu/drm/nouveau/nv04_fifo.c index 0c3cd53c7313..f31347b8c9b0 100644 --- a/drivers/gpu/drm/nouveau/nv04_fifo.c +++ b/drivers/gpu/drm/nouveau/nv04_fifo.c @@ -71,6 +71,40 @@ nv04_fifo_reassign(struct drm_device *dev, bool enable) return (reassign == 1); } +bool +nv04_fifo_cache_flush(struct drm_device *dev) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_timer_engine *ptimer = &dev_priv->engine.timer; + uint64_t start = ptimer->read(dev); + + do { + if (nv_rd32(dev, NV03_PFIFO_CACHE1_GET) == + nv_rd32(dev, NV03_PFIFO_CACHE1_PUT)) + return true; + + } while (ptimer->read(dev) - start < 100000000); + + NV_ERROR(dev, "Timeout flushing the PFIFO cache.\n"); + + return false; +} + +bool +nv04_fifo_cache_pull(struct drm_device *dev, bool enable) +{ + uint32_t pull = nv_rd32(dev, NV04_PFIFO_CACHE1_PULL0); + + if (enable) { + nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, pull | 1); + } else { + nv_wr32(dev, NV04_PFIFO_CACHE1_PULL0, pull & ~1); + nv_wr32(dev, NV04_PFIFO_CACHE1_HASH, 0); + } + + return !!(pull & 1); +} + int nv04_fifo_channel_id(struct drm_device *dev) { diff --git a/drivers/gpu/drm/nouveau/nv04_graph.c b/drivers/gpu/drm/nouveau/nv04_graph.c index 396ee92118f6..e260986ea65a 100644 --- a/drivers/gpu/drm/nouveau/nv04_graph.c +++ b/drivers/gpu/drm/nouveau/nv04_graph.c @@ -28,6 +28,10 @@ #include "nouveau_drv.h" static uint32_t nv04_graph_ctx_regs[] = { + 0x0040053c, + 0x00400544, + 0x00400540, + 0x00400548, NV04_PGRAPH_CTX_SWITCH1, NV04_PGRAPH_CTX_SWITCH2, NV04_PGRAPH_CTX_SWITCH3, @@ -102,69 +106,69 @@ static uint32_t nv04_graph_ctx_regs[] = { NV04_PGRAPH_PATT_COLOR0, NV04_PGRAPH_PATT_COLOR1, NV04_PGRAPH_PATT_COLORRAM+0x00, - NV04_PGRAPH_PATT_COLORRAM+0x01, - NV04_PGRAPH_PATT_COLORRAM+0x02, - NV04_PGRAPH_PATT_COLORRAM+0x03, NV04_PGRAPH_PATT_COLORRAM+0x04, - NV04_PGRAPH_PATT_COLORRAM+0x05, - NV04_PGRAPH_PATT_COLORRAM+0x06, - NV04_PGRAPH_PATT_COLORRAM+0x07, NV04_PGRAPH_PATT_COLORRAM+0x08, - NV04_PGRAPH_PATT_COLORRAM+0x09, - NV04_PGRAPH_PATT_COLORRAM+0x0A, - NV04_PGRAPH_PATT_COLORRAM+0x0B, - NV04_PGRAPH_PATT_COLORRAM+0x0C, - NV04_PGRAPH_PATT_COLORRAM+0x0D, - NV04_PGRAPH_PATT_COLORRAM+0x0E, - NV04_PGRAPH_PATT_COLORRAM+0x0F, + NV04_PGRAPH_PATT_COLORRAM+0x0c, NV04_PGRAPH_PATT_COLORRAM+0x10, - NV04_PGRAPH_PATT_COLORRAM+0x11, - NV04_PGRAPH_PATT_COLORRAM+0x12, - NV04_PGRAPH_PATT_COLORRAM+0x13, NV04_PGRAPH_PATT_COLORRAM+0x14, - NV04_PGRAPH_PATT_COLORRAM+0x15, - NV04_PGRAPH_PATT_COLORRAM+0x16, - NV04_PGRAPH_PATT_COLORRAM+0x17, NV04_PGRAPH_PATT_COLORRAM+0x18, - NV04_PGRAPH_PATT_COLORRAM+0x19, - NV04_PGRAPH_PATT_COLORRAM+0x1A, - NV04_PGRAPH_PATT_COLORRAM+0x1B, - NV04_PGRAPH_PATT_COLORRAM+0x1C, - NV04_PGRAPH_PATT_COLORRAM+0x1D, - NV04_PGRAPH_PATT_COLORRAM+0x1E, - NV04_PGRAPH_PATT_COLORRAM+0x1F, + NV04_PGRAPH_PATT_COLORRAM+0x1c, NV04_PGRAPH_PATT_COLORRAM+0x20, - NV04_PGRAPH_PATT_COLORRAM+0x21, - NV04_PGRAPH_PATT_COLORRAM+0x22, - NV04_PGRAPH_PATT_COLORRAM+0x23, NV04_PGRAPH_PATT_COLORRAM+0x24, - NV04_PGRAPH_PATT_COLORRAM+0x25, - NV04_PGRAPH_PATT_COLORRAM+0x26, - NV04_PGRAPH_PATT_COLORRAM+0x27, NV04_PGRAPH_PATT_COLORRAM+0x28, - NV04_PGRAPH_PATT_COLORRAM+0x29, - NV04_PGRAPH_PATT_COLORRAM+0x2A, - NV04_PGRAPH_PATT_COLORRAM+0x2B, - NV04_PGRAPH_PATT_COLORRAM+0x2C, - NV04_PGRAPH_PATT_COLORRAM+0x2D, - NV04_PGRAPH_PATT_COLORRAM+0x2E, - NV04_PGRAPH_PATT_COLORRAM+0x2F, + NV04_PGRAPH_PATT_COLORRAM+0x2c, NV04_PGRAPH_PATT_COLORRAM+0x30, - NV04_PGRAPH_PATT_COLORRAM+0x31, - NV04_PGRAPH_PATT_COLORRAM+0x32, - NV04_PGRAPH_PATT_COLORRAM+0x33, NV04_PGRAPH_PATT_COLORRAM+0x34, - NV04_PGRAPH_PATT_COLORRAM+0x35, - NV04_PGRAPH_PATT_COLORRAM+0x36, - NV04_PGRAPH_PATT_COLORRAM+0x37, NV04_PGRAPH_PATT_COLORRAM+0x38, - NV04_PGRAPH_PATT_COLORRAM+0x39, - NV04_PGRAPH_PATT_COLORRAM+0x3A, - NV04_PGRAPH_PATT_COLORRAM+0x3B, - NV04_PGRAPH_PATT_COLORRAM+0x3C, - NV04_PGRAPH_PATT_COLORRAM+0x3D, - NV04_PGRAPH_PATT_COLORRAM+0x3E, - NV04_PGRAPH_PATT_COLORRAM+0x3F, + NV04_PGRAPH_PATT_COLORRAM+0x3c, + NV04_PGRAPH_PATT_COLORRAM+0x40, + NV04_PGRAPH_PATT_COLORRAM+0x44, + NV04_PGRAPH_PATT_COLORRAM+0x48, + NV04_PGRAPH_PATT_COLORRAM+0x4c, + NV04_PGRAPH_PATT_COLORRAM+0x50, + NV04_PGRAPH_PATT_COLORRAM+0x54, + NV04_PGRAPH_PATT_COLORRAM+0x58, + NV04_PGRAPH_PATT_COLORRAM+0x5c, + NV04_PGRAPH_PATT_COLORRAM+0x60, + NV04_PGRAPH_PATT_COLORRAM+0x64, + NV04_PGRAPH_PATT_COLORRAM+0x68, + NV04_PGRAPH_PATT_COLORRAM+0x6c, + NV04_PGRAPH_PATT_COLORRAM+0x70, + NV04_PGRAPH_PATT_COLORRAM+0x74, + NV04_PGRAPH_PATT_COLORRAM+0x78, + NV04_PGRAPH_PATT_COLORRAM+0x7c, + NV04_PGRAPH_PATT_COLORRAM+0x80, + NV04_PGRAPH_PATT_COLORRAM+0x84, + NV04_PGRAPH_PATT_COLORRAM+0x88, + NV04_PGRAPH_PATT_COLORRAM+0x8c, + NV04_PGRAPH_PATT_COLORRAM+0x90, + NV04_PGRAPH_PATT_COLORRAM+0x94, + NV04_PGRAPH_PATT_COLORRAM+0x98, + NV04_PGRAPH_PATT_COLORRAM+0x9c, + NV04_PGRAPH_PATT_COLORRAM+0xa0, + NV04_PGRAPH_PATT_COLORRAM+0xa4, + NV04_PGRAPH_PATT_COLORRAM+0xa8, + NV04_PGRAPH_PATT_COLORRAM+0xac, + NV04_PGRAPH_PATT_COLORRAM+0xb0, + NV04_PGRAPH_PATT_COLORRAM+0xb4, + NV04_PGRAPH_PATT_COLORRAM+0xb8, + NV04_PGRAPH_PATT_COLORRAM+0xbc, + NV04_PGRAPH_PATT_COLORRAM+0xc0, + NV04_PGRAPH_PATT_COLORRAM+0xc4, + NV04_PGRAPH_PATT_COLORRAM+0xc8, + NV04_PGRAPH_PATT_COLORRAM+0xcc, + NV04_PGRAPH_PATT_COLORRAM+0xd0, + NV04_PGRAPH_PATT_COLORRAM+0xd4, + NV04_PGRAPH_PATT_COLORRAM+0xd8, + NV04_PGRAPH_PATT_COLORRAM+0xdc, + NV04_PGRAPH_PATT_COLORRAM+0xe0, + NV04_PGRAPH_PATT_COLORRAM+0xe4, + NV04_PGRAPH_PATT_COLORRAM+0xe8, + NV04_PGRAPH_PATT_COLORRAM+0xec, + NV04_PGRAPH_PATT_COLORRAM+0xf0, + NV04_PGRAPH_PATT_COLORRAM+0xf4, + NV04_PGRAPH_PATT_COLORRAM+0xf8, + NV04_PGRAPH_PATT_COLORRAM+0xfc, NV04_PGRAPH_PATTERN, 0x0040080c, NV04_PGRAPH_PATTERN_SHAPE, @@ -247,14 +251,6 @@ static uint32_t nv04_graph_ctx_regs[] = { 0x004004f8, 0x0040047c, 0x004004fc, - 0x0040053c, - 0x00400544, - 0x00400540, - 0x00400548, - 0x00400560, - 0x00400568, - 0x00400564, - 0x0040056c, 0x00400534, 0x00400538, 0x00400514, @@ -341,9 +337,8 @@ static uint32_t nv04_graph_ctx_regs[] = { 0x00400500, 0x00400504, NV04_PGRAPH_VALID1, - NV04_PGRAPH_VALID2 - - + NV04_PGRAPH_VALID2, + NV04_PGRAPH_DEBUG_3 }; struct graph_state { @@ -388,6 +383,18 @@ nv04_graph_context_switch(struct drm_device *dev) pgraph->fifo_access(dev, true); } +static uint32_t *ctx_reg(struct graph_state *ctx, uint32_t reg) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(nv04_graph_ctx_regs); i++) { + if (nv04_graph_ctx_regs[i] == reg) + return &ctx->nv04[i]; + } + + return NULL; +} + int nv04_graph_create_context(struct nouveau_channel *chan) { struct graph_state *pgraph_ctx; @@ -398,15 +405,8 @@ int nv04_graph_create_context(struct nouveau_channel *chan) if (pgraph_ctx == NULL) return -ENOMEM; - /* dev_priv->fifos[channel].pgraph_ctx_user = channel << 24; */ - pgraph_ctx->nv04[0] = 0x0001ffff; - /* is it really needed ??? */ -#if 0 - dev_priv->fifos[channel].pgraph_ctx[1] = - nv_rd32(dev, NV_PGRAPH_DEBUG_4); - dev_priv->fifos[channel].pgraph_ctx[2] = - nv_rd32(dev, 0x004006b0); -#endif + *ctx_reg(pgraph_ctx, NV04_PGRAPH_DEBUG_3) = 0xfad4ff31; + return 0; } @@ -429,9 +429,13 @@ int nv04_graph_load_context(struct nouveau_channel *chan) nv_wr32(dev, nv04_graph_ctx_regs[i], pgraph_ctx->nv04[i]); nv_wr32(dev, NV04_PGRAPH_CTX_CONTROL, 0x10010100); - nv_wr32(dev, NV04_PGRAPH_CTX_USER, chan->id << 24); + + tmp = nv_rd32(dev, NV04_PGRAPH_CTX_USER) & 0x00ffffff; + nv_wr32(dev, NV04_PGRAPH_CTX_USER, tmp | chan->id << 24); + tmp = nv_rd32(dev, NV04_PGRAPH_FFINTFC_ST2); nv_wr32(dev, NV04_PGRAPH_FFINTFC_ST2, tmp & 0x000fffff); + return 0; } @@ -494,7 +498,7 @@ int nv04_graph_init(struct drm_device *dev) nv_wr32(dev, NV04_PGRAPH_STATE , 0xFFFFFFFF); nv_wr32(dev, NV04_PGRAPH_CTX_CONTROL , 0x10000100); tmp = nv_rd32(dev, NV04_PGRAPH_CTX_USER) & 0x00ffffff; - tmp |= dev_priv->engine.fifo.channels << 24; + tmp |= (dev_priv->engine.fifo.channels - 1) << 24; nv_wr32(dev, NV04_PGRAPH_CTX_USER, tmp); /* These don't belong here, they're part of a per-channel context */ @@ -533,7 +537,7 @@ nv04_graph_mthd_set_operation(struct nouveau_channel *chan, int grclass, int mthd, uint32_t data) { struct drm_device *dev = chan->dev; - uint32_t instance = nv_rd32(dev, NV04_PGRAPH_CTX_SWITCH4) & 0xffff; + uint32_t instance = (nv_rd32(dev, NV04_PGRAPH_CTX_SWITCH4) & 0xffff) << 4; int subc = (nv_rd32(dev, NV04_PGRAPH_TRAPPED_ADDR) >> 13) & 0x7; uint32_t tmp; @@ -543,11 +547,11 @@ nv04_graph_mthd_set_operation(struct nouveau_channel *chan, int grclass, nv_wi32(dev, instance, tmp); nv_wr32(dev, NV04_PGRAPH_CTX_SWITCH1, tmp); - nv_wr32(dev, NV04_PGRAPH_CTX_CACHE1 + subc, tmp); + nv_wr32(dev, NV04_PGRAPH_CTX_CACHE1 + (subc<<2), tmp); return 0; } -static struct nouveau_pgraph_object_method nv04_graph_mthds_m2mf[] = { +static struct nouveau_pgraph_object_method nv04_graph_mthds_sw[] = { { 0x0150, nv04_graph_mthd_set_ref }, {} }; @@ -558,7 +562,7 @@ static struct nouveau_pgraph_object_method nv04_graph_mthds_set_operation[] = { }; struct nouveau_pgraph_object_class nv04_graph_grclass[] = { - { 0x0039, false, nv04_graph_mthds_m2mf }, + { 0x0039, false, NULL }, { 0x004a, false, nv04_graph_mthds_set_operation }, /* gdirect */ { 0x005f, false, nv04_graph_mthds_set_operation }, /* imageblit */ { 0x0061, false, nv04_graph_mthds_set_operation }, /* ifc */ @@ -574,6 +578,7 @@ struct nouveau_pgraph_object_class nv04_graph_grclass[] = { { 0x0053, false, NULL }, /* surf3d */ { 0x0054, false, NULL }, /* tex_tri */ { 0x0055, false, NULL }, /* multitex_tri */ + { 0x506e, true, nv04_graph_mthds_sw }, {} }; diff --git a/drivers/gpu/drm/nouveau/nv10_fb.c b/drivers/gpu/drm/nouveau/nv10_fb.c index 79e2d104d70a..cc5cda44e501 100644 --- a/drivers/gpu/drm/nouveau/nv10_fb.c +++ b/drivers/gpu/drm/nouveau/nv10_fb.c @@ -3,17 +3,37 @@ #include "nouveau_drv.h" #include "nouveau_drm.h" +void +nv10_fb_set_region_tiling(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + uint32_t limit = max(1u, addr + size) - 1; + + if (pitch) { + if (dev_priv->card_type >= NV_20) + addr |= 1; + else + addr |= 1 << 31; + } + + nv_wr32(dev, NV10_PFB_TLIMIT(i), limit); + nv_wr32(dev, NV10_PFB_TSIZE(i), pitch); + nv_wr32(dev, NV10_PFB_TILE(i), addr); +} + int nv10_fb_init(struct drm_device *dev) { - uint32_t fb_bar_size; + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_fb_engine *pfb = &dev_priv->engine.fb; int i; - fb_bar_size = drm_get_resource_len(dev, 0) - 1; - for (i = 0; i < NV10_PFB_TILE__SIZE; i++) { - nv_wr32(dev, NV10_PFB_TILE(i), 0); - nv_wr32(dev, NV10_PFB_TLIMIT(i), fb_bar_size); - } + pfb->num_tiles = NV10_PFB_TILE__SIZE; + + /* Turn all the tiling regions off. */ + for (i = 0; i < pfb->num_tiles; i++) + pfb->set_region_tiling(dev, i, 0, 0, 0); return 0; } diff --git a/drivers/gpu/drm/nouveau/nv10_graph.c b/drivers/gpu/drm/nouveau/nv10_graph.c index 6bf6804bb0ef..fcf2cdd19493 100644 --- a/drivers/gpu/drm/nouveau/nv10_graph.c +++ b/drivers/gpu/drm/nouveau/nv10_graph.c @@ -389,49 +389,50 @@ struct graph_state { int nv10[ARRAY_SIZE(nv10_graph_ctx_regs)]; int nv17[ARRAY_SIZE(nv17_graph_ctx_regs)]; struct pipe_state pipe_state; + uint32_t lma_window[4]; }; +#define PIPE_SAVE(dev, state, addr) \ + do { \ + int __i; \ + nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, addr); \ + for (__i = 0; __i < ARRAY_SIZE(state); __i++) \ + state[__i] = nv_rd32(dev, NV10_PGRAPH_PIPE_DATA); \ + } while (0) + +#define PIPE_RESTORE(dev, state, addr) \ + do { \ + int __i; \ + nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, addr); \ + for (__i = 0; __i < ARRAY_SIZE(state); __i++) \ + nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, state[__i]); \ + } while (0) + static void nv10_graph_save_pipe(struct nouveau_channel *chan) { struct drm_device *dev = chan->dev; struct graph_state *pgraph_ctx = chan->pgraph_ctx; - struct pipe_state *fifo_pipe_state = &pgraph_ctx->pipe_state; - int i; -#define PIPE_SAVE(addr) \ - do { \ - nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, addr); \ - for (i = 0; i < ARRAY_SIZE(fifo_pipe_state->pipe_##addr); i++) \ - fifo_pipe_state->pipe_##addr[i] = nv_rd32(dev, NV10_PGRAPH_PIPE_DATA); \ - } while (0) - - PIPE_SAVE(0x4400); - PIPE_SAVE(0x0200); - PIPE_SAVE(0x6400); - PIPE_SAVE(0x6800); - PIPE_SAVE(0x6c00); - PIPE_SAVE(0x7000); - PIPE_SAVE(0x7400); - PIPE_SAVE(0x7800); - PIPE_SAVE(0x0040); - PIPE_SAVE(0x0000); - -#undef PIPE_SAVE + struct pipe_state *pipe = &pgraph_ctx->pipe_state; + + PIPE_SAVE(dev, pipe->pipe_0x4400, 0x4400); + PIPE_SAVE(dev, pipe->pipe_0x0200, 0x0200); + PIPE_SAVE(dev, pipe->pipe_0x6400, 0x6400); + PIPE_SAVE(dev, pipe->pipe_0x6800, 0x6800); + PIPE_SAVE(dev, pipe->pipe_0x6c00, 0x6c00); + PIPE_SAVE(dev, pipe->pipe_0x7000, 0x7000); + PIPE_SAVE(dev, pipe->pipe_0x7400, 0x7400); + PIPE_SAVE(dev, pipe->pipe_0x7800, 0x7800); + PIPE_SAVE(dev, pipe->pipe_0x0040, 0x0040); + PIPE_SAVE(dev, pipe->pipe_0x0000, 0x0000); } static void nv10_graph_load_pipe(struct nouveau_channel *chan) { struct drm_device *dev = chan->dev; struct graph_state *pgraph_ctx = chan->pgraph_ctx; - struct pipe_state *fifo_pipe_state = &pgraph_ctx->pipe_state; - int i; + struct pipe_state *pipe = &pgraph_ctx->pipe_state; uint32_t xfmode0, xfmode1; -#define PIPE_RESTORE(addr) \ - do { \ - nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, addr); \ - for (i = 0; i < ARRAY_SIZE(fifo_pipe_state->pipe_##addr); i++) \ - nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, fifo_pipe_state->pipe_##addr[i]); \ - } while (0) - + int i; nouveau_wait_for_idle(dev); /* XXX check haiku comments */ @@ -457,24 +458,22 @@ static void nv10_graph_load_pipe(struct nouveau_channel *chan) nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000008); - PIPE_RESTORE(0x0200); + PIPE_RESTORE(dev, pipe->pipe_0x0200, 0x0200); nouveau_wait_for_idle(dev); /* restore XFMODE */ nv_wr32(dev, NV10_PGRAPH_XFMODE0, xfmode0); nv_wr32(dev, NV10_PGRAPH_XFMODE1, xfmode1); - PIPE_RESTORE(0x6400); - PIPE_RESTORE(0x6800); - PIPE_RESTORE(0x6c00); - PIPE_RESTORE(0x7000); - PIPE_RESTORE(0x7400); - PIPE_RESTORE(0x7800); - PIPE_RESTORE(0x4400); - PIPE_RESTORE(0x0000); - PIPE_RESTORE(0x0040); + PIPE_RESTORE(dev, pipe->pipe_0x6400, 0x6400); + PIPE_RESTORE(dev, pipe->pipe_0x6800, 0x6800); + PIPE_RESTORE(dev, pipe->pipe_0x6c00, 0x6c00); + PIPE_RESTORE(dev, pipe->pipe_0x7000, 0x7000); + PIPE_RESTORE(dev, pipe->pipe_0x7400, 0x7400); + PIPE_RESTORE(dev, pipe->pipe_0x7800, 0x7800); + PIPE_RESTORE(dev, pipe->pipe_0x4400, 0x4400); + PIPE_RESTORE(dev, pipe->pipe_0x0000, 0x0000); + PIPE_RESTORE(dev, pipe->pipe_0x0040, 0x0040); nouveau_wait_for_idle(dev); - -#undef PIPE_RESTORE } static void nv10_graph_create_pipe(struct nouveau_channel *chan) @@ -808,6 +807,20 @@ void nv10_graph_destroy_context(struct nouveau_channel *chan) chan->pgraph_ctx = NULL; } +void +nv10_graph_set_region_tiling(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch) +{ + uint32_t limit = max(1u, addr + size) - 1; + + if (pitch) + addr |= 1 << 31; + + nv_wr32(dev, NV10_PGRAPH_TLIMIT(i), limit); + nv_wr32(dev, NV10_PGRAPH_TSIZE(i), pitch); + nv_wr32(dev, NV10_PGRAPH_TILE(i), addr); +} + int nv10_graph_init(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; @@ -832,21 +845,16 @@ int nv10_graph_init(struct drm_device *dev) (1<<31)); if (dev_priv->chipset >= 0x17) { nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x1f000000); + nv_wr32(dev, 0x400a10, 0x3ff3fb6); + nv_wr32(dev, 0x400838, 0x2f8684); + nv_wr32(dev, 0x40083c, 0x115f3f); nv_wr32(dev, 0x004006b0, 0x40000020); } else nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x00000000); - /* copy tile info from PFB */ - for (i = 0; i < NV10_PFB_TILE__SIZE; i++) { - nv_wr32(dev, NV10_PGRAPH_TILE(i), - nv_rd32(dev, NV10_PFB_TILE(i))); - nv_wr32(dev, NV10_PGRAPH_TLIMIT(i), - nv_rd32(dev, NV10_PFB_TLIMIT(i))); - nv_wr32(dev, NV10_PGRAPH_TSIZE(i), - nv_rd32(dev, NV10_PFB_TSIZE(i))); - nv_wr32(dev, NV10_PGRAPH_TSTATUS(i), - nv_rd32(dev, NV10_PFB_TSTATUS(i))); - } + /* Turn all the tiling regions off. */ + for (i = 0; i < NV10_PFB_TILE__SIZE; i++) + nv10_graph_set_region_tiling(dev, i, 0, 0, 0); nv_wr32(dev, NV10_PGRAPH_CTX_SWITCH1, 0x00000000); nv_wr32(dev, NV10_PGRAPH_CTX_SWITCH2, 0x00000000); @@ -867,6 +875,115 @@ void nv10_graph_takedown(struct drm_device *dev) { } +static int +nv17_graph_mthd_lma_window(struct nouveau_channel *chan, int grclass, + int mthd, uint32_t data) +{ + struct drm_device *dev = chan->dev; + struct graph_state *ctx = chan->pgraph_ctx; + struct pipe_state *pipe = &ctx->pipe_state; + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; + uint32_t pipe_0x0040[1], pipe_0x64c0[8], pipe_0x6a80[3], pipe_0x6ab0[3]; + uint32_t xfmode0, xfmode1; + int i; + + ctx->lma_window[(mthd - 0x1638) / 4] = data; + + if (mthd != 0x1644) + return 0; + + nouveau_wait_for_idle(dev); + + PIPE_SAVE(dev, pipe_0x0040, 0x0040); + PIPE_SAVE(dev, pipe->pipe_0x0200, 0x0200); + + PIPE_RESTORE(dev, ctx->lma_window, 0x6790); + + nouveau_wait_for_idle(dev); + + xfmode0 = nv_rd32(dev, NV10_PGRAPH_XFMODE0); + xfmode1 = nv_rd32(dev, NV10_PGRAPH_XFMODE1); + + PIPE_SAVE(dev, pipe->pipe_0x4400, 0x4400); + PIPE_SAVE(dev, pipe_0x64c0, 0x64c0); + PIPE_SAVE(dev, pipe_0x6ab0, 0x6ab0); + PIPE_SAVE(dev, pipe_0x6a80, 0x6a80); + + nouveau_wait_for_idle(dev); + + nv_wr32(dev, NV10_PGRAPH_XFMODE0, 0x10000000); + nv_wr32(dev, NV10_PGRAPH_XFMODE1, 0x00000000); + nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0); + for (i = 0; i < 4; i++) + nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + for (i = 0; i < 4; i++) + nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0); + for (i = 0; i < 3; i++) + nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + + nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80); + for (i = 0; i < 3; i++) + nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040); + nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000008); + + PIPE_RESTORE(dev, pipe->pipe_0x0200, 0x0200); + + nouveau_wait_for_idle(dev); + + PIPE_RESTORE(dev, pipe_0x0040, 0x0040); + + nv_wr32(dev, NV10_PGRAPH_XFMODE0, xfmode0); + nv_wr32(dev, NV10_PGRAPH_XFMODE1, xfmode1); + + PIPE_RESTORE(dev, pipe_0x64c0, 0x64c0); + PIPE_RESTORE(dev, pipe_0x6ab0, 0x6ab0); + PIPE_RESTORE(dev, pipe_0x6a80, 0x6a80); + PIPE_RESTORE(dev, pipe->pipe_0x4400, 0x4400); + + nv_wr32(dev, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0); + nv_wr32(dev, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nouveau_wait_for_idle(dev); + + pgraph->fifo_access(dev, true); + + return 0; +} + +static int +nv17_graph_mthd_lma_enable(struct nouveau_channel *chan, int grclass, + int mthd, uint32_t data) +{ + struct drm_device *dev = chan->dev; + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; + + nouveau_wait_for_idle(dev); + + nv_wr32(dev, NV10_PGRAPH_DEBUG_4, + nv_rd32(dev, NV10_PGRAPH_DEBUG_4) | 0x1 << 8); + nv_wr32(dev, 0x004006b0, + nv_rd32(dev, 0x004006b0) | 0x8 << 24); + + pgraph->fifo_access(dev, true); + + return 0; +} + +static struct nouveau_pgraph_object_method nv17_graph_celsius_mthds[] = { + { 0x1638, nv17_graph_mthd_lma_window }, + { 0x163c, nv17_graph_mthd_lma_window }, + { 0x1640, nv17_graph_mthd_lma_window }, + { 0x1644, nv17_graph_mthd_lma_window }, + { 0x1658, nv17_graph_mthd_lma_enable }, + {} +}; + struct nouveau_pgraph_object_class nv10_graph_grclass[] = { { 0x0030, false, NULL }, /* null */ { 0x0039, false, NULL }, /* m2mf */ @@ -887,6 +1004,6 @@ struct nouveau_pgraph_object_class nv10_graph_grclass[] = { { 0x0095, false, NULL }, /* multitex_tri */ { 0x0056, false, NULL }, /* celcius (nv10) */ { 0x0096, false, NULL }, /* celcius (nv11) */ - { 0x0099, false, NULL }, /* celcius (nv17) */ + { 0x0099, false, nv17_graph_celsius_mthds }, /* celcius (nv17) */ {} }; diff --git a/drivers/gpu/drm/nouveau/nv17_tv.c b/drivers/gpu/drm/nouveau/nv17_tv.c index 46cfd9c60478..58b917c3341b 100644 --- a/drivers/gpu/drm/nouveau/nv17_tv.c +++ b/drivers/gpu/drm/nouveau/nv17_tv.c @@ -33,13 +33,103 @@ #include "nouveau_hw.h" #include "nv17_tv.h" -enum drm_connector_status nv17_tv_detect(struct drm_encoder *encoder, - struct drm_connector *connector, - uint32_t pin_mask) +static uint32_t nv42_tv_sample_load(struct drm_encoder *encoder) { + struct drm_device *dev = encoder->dev; + struct drm_nouveau_private *dev_priv = dev->dev_private; + uint32_t testval, regoffset = nv04_dac_output_offset(encoder); + uint32_t gpio0, gpio1, fp_htotal, fp_hsync_start, fp_hsync_end, + fp_control, test_ctrl, dacclk, ctv_14, ctv_1c, ctv_6c; + uint32_t sample = 0; + int head; + +#define RGB_TEST_DATA(r, g, b) (r << 0 | g << 10 | b << 20) + testval = RGB_TEST_DATA(0x82, 0xeb, 0x82); + if (dev_priv->vbios->tvdactestval) + testval = dev_priv->vbios->tvdactestval; + + dacclk = NVReadRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset); + head = (dacclk & 0x100) >> 8; + + /* Save the previous state. */ + gpio1 = nv17_gpio_get(dev, DCB_GPIO_TVDAC1); + gpio0 = nv17_gpio_get(dev, DCB_GPIO_TVDAC0); + fp_htotal = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_HTOTAL); + fp_hsync_start = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_START); + fp_hsync_end = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_END); + fp_control = NVReadRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL); + test_ctrl = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset); + ctv_1c = NVReadRAMDAC(dev, head, 0x680c1c); + ctv_14 = NVReadRAMDAC(dev, head, 0x680c14); + ctv_6c = NVReadRAMDAC(dev, head, 0x680c6c); + + /* Prepare the DAC for load detection. */ + nv17_gpio_set(dev, DCB_GPIO_TVDAC1, true); + nv17_gpio_set(dev, DCB_GPIO_TVDAC0, true); + + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HTOTAL, 1343); + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_START, 1047); + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_END, 1183); + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL, + NV_PRAMDAC_FP_TG_CONTROL_DISPEN_POS | + NV_PRAMDAC_FP_TG_CONTROL_WIDTH_12 | + NV_PRAMDAC_FP_TG_CONTROL_READ_PROG | + NV_PRAMDAC_FP_TG_CONTROL_HSYNC_POS | + NV_PRAMDAC_FP_TG_CONTROL_VSYNC_POS); + + NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset, 0); + + NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset, + (dacclk & ~0xff) | 0x22); + msleep(1); + NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset, + (dacclk & ~0xff) | 0x21); + + NVWriteRAMDAC(dev, head, 0x680c1c, 1 << 20); + NVWriteRAMDAC(dev, head, 0x680c14, 4 << 16); + + /* Sample pin 0x4 (usually S-video luma). */ + NVWriteRAMDAC(dev, head, 0x680c6c, testval >> 10 & 0x3ff); + msleep(20); + sample |= NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset) + & 0x4 << 28; + + /* Sample the remaining pins. */ + NVWriteRAMDAC(dev, head, 0x680c6c, testval & 0x3ff); + msleep(20); + sample |= NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset) + & 0xa << 28; + + /* Restore the previous state. */ + NVWriteRAMDAC(dev, head, 0x680c1c, ctv_1c); + NVWriteRAMDAC(dev, head, 0x680c14, ctv_14); + NVWriteRAMDAC(dev, head, 0x680c6c, ctv_6c); + NVWriteRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset, dacclk); + NVWriteRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset, test_ctrl); + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_TG_CONTROL, fp_control); + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_END, fp_hsync_end); + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HSYNC_START, fp_hsync_start); + NVWriteRAMDAC(dev, head, NV_PRAMDAC_FP_HTOTAL, fp_htotal); + nv17_gpio_set(dev, DCB_GPIO_TVDAC1, gpio1); + nv17_gpio_set(dev, DCB_GPIO_TVDAC0, gpio0); + + return sample; +} + +static enum drm_connector_status +nv17_tv_detect(struct drm_encoder *encoder, struct drm_connector *connector) +{ + struct drm_device *dev = encoder->dev; + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct drm_mode_config *conf = &dev->mode_config; struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder); + struct dcb_entry *dcb = tv_enc->base.dcb; - tv_enc->pin_mask = pin_mask >> 28 & 0xe; + if (dev_priv->chipset == 0x42 || + dev_priv->chipset == 0x43) + tv_enc->pin_mask = nv42_tv_sample_load(encoder) >> 28 & 0xe; + else + tv_enc->pin_mask = nv17_dac_sample_load(encoder) >> 28 & 0xe; switch (tv_enc->pin_mask) { case 0x2: @@ -50,7 +140,7 @@ enum drm_connector_status nv17_tv_detect(struct drm_encoder *encoder, tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_SVIDEO; break; case 0xe: - if (nouveau_encoder(encoder)->dcb->tvconf.has_component_output) + if (dcb->tvconf.has_component_output) tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_Component; else tv_enc->subconnector = DRM_MODE_SUBCONNECTOR_SCART; @@ -61,11 +151,16 @@ enum drm_connector_status nv17_tv_detect(struct drm_encoder *encoder, } drm_connector_property_set_value(connector, - encoder->dev->mode_config.tv_subconnector_property, - tv_enc->subconnector); + conf->tv_subconnector_property, + tv_enc->subconnector); - return tv_enc->subconnector ? connector_status_connected : - connector_status_disconnected; + if (tv_enc->subconnector) { + NV_INFO(dev, "Load detected on output %c\n", + '@' + ffs(dcb->or)); + return connector_status_connected; + } else { + return connector_status_disconnected; + } } static const struct { @@ -219,7 +314,7 @@ static void nv17_tv_dpms(struct drm_encoder *encoder, int mode) return; nouveau_encoder(encoder)->last_dpms = mode; - NV_TRACE(dev, "Setting dpms mode %d on TV encoder (output %d)\n", + NV_INFO(dev, "Setting dpms mode %d on TV encoder (output %d)\n", mode, nouveau_encoder(encoder)->dcb->index); regs->ptv_200 &= ~1; @@ -619,7 +714,7 @@ static void nv17_tv_destroy(struct drm_encoder *encoder) { struct nv17_tv_encoder *tv_enc = to_tv_enc(encoder); - NV_DEBUG(encoder->dev, "\n"); + NV_DEBUG_KMS(encoder->dev, "\n"); drm_encoder_cleanup(encoder); kfree(tv_enc); @@ -633,7 +728,7 @@ static struct drm_encoder_helper_funcs nv17_tv_helper_funcs = { .prepare = nv17_tv_prepare, .commit = nv17_tv_commit, .mode_set = nv17_tv_mode_set, - .detect = nv17_dac_detect, + .detect = nv17_tv_detect, }; static struct drm_encoder_slave_funcs nv17_tv_slave_funcs = { diff --git a/drivers/gpu/drm/nouveau/nv20_graph.c b/drivers/gpu/drm/nouveau/nv20_graph.c index 18ba74f19703..d6fc0a82f03d 100644 --- a/drivers/gpu/drm/nouveau/nv20_graph.c +++ b/drivers/gpu/drm/nouveau/nv20_graph.c @@ -514,6 +514,27 @@ nv20_graph_rdi(struct drm_device *dev) nouveau_wait_for_idle(dev); } +void +nv20_graph_set_region_tiling(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch) +{ + uint32_t limit = max(1u, addr + size) - 1; + + if (pitch) + addr |= 1; + + nv_wr32(dev, NV20_PGRAPH_TLIMIT(i), limit); + nv_wr32(dev, NV20_PGRAPH_TSIZE(i), pitch); + nv_wr32(dev, NV20_PGRAPH_TILE(i), addr); + + nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + 4 * i); + nv_wr32(dev, NV10_PGRAPH_RDI_DATA, limit); + nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + 4 * i); + nv_wr32(dev, NV10_PGRAPH_RDI_DATA, pitch); + nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + 4 * i); + nv_wr32(dev, NV10_PGRAPH_RDI_DATA, addr); +} + int nv20_graph_init(struct drm_device *dev) { @@ -572,27 +593,10 @@ nv20_graph_init(struct drm_device *dev) nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000030); } - /* copy tile info from PFB */ - for (i = 0; i < NV10_PFB_TILE__SIZE; i++) { - nv_wr32(dev, 0x00400904 + i * 0x10, - nv_rd32(dev, NV10_PFB_TLIMIT(i))); - /* which is NV40_PGRAPH_TLIMIT0(i) ?? */ - nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + i * 4); - nv_wr32(dev, NV10_PGRAPH_RDI_DATA, - nv_rd32(dev, NV10_PFB_TLIMIT(i))); - nv_wr32(dev, 0x00400908 + i * 0x10, - nv_rd32(dev, NV10_PFB_TSIZE(i))); - /* which is NV40_PGRAPH_TSIZE0(i) ?? */ - nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + i * 4); - nv_wr32(dev, NV10_PGRAPH_RDI_DATA, - nv_rd32(dev, NV10_PFB_TSIZE(i))); - nv_wr32(dev, 0x00400900 + i * 0x10, - nv_rd32(dev, NV10_PFB_TILE(i))); - /* which is NV40_PGRAPH_TILE0(i) ?? */ - nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + i * 4); - nv_wr32(dev, NV10_PGRAPH_RDI_DATA, - nv_rd32(dev, NV10_PFB_TILE(i))); - } + /* Turn all the tiling regions off. */ + for (i = 0; i < NV10_PFB_TILE__SIZE; i++) + nv20_graph_set_region_tiling(dev, i, 0, 0, 0); + for (i = 0; i < 8; i++) { nv_wr32(dev, 0x400980 + i * 4, nv_rd32(dev, 0x100300 + i * 4)); nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0090 + i * 4); @@ -704,18 +708,9 @@ nv30_graph_init(struct drm_device *dev) nv_wr32(dev, 0x4000c0, 0x00000016); - /* copy tile info from PFB */ - for (i = 0; i < NV10_PFB_TILE__SIZE; i++) { - nv_wr32(dev, 0x00400904 + i * 0x10, - nv_rd32(dev, NV10_PFB_TLIMIT(i))); - /* which is NV40_PGRAPH_TLIMIT0(i) ?? */ - nv_wr32(dev, 0x00400908 + i * 0x10, - nv_rd32(dev, NV10_PFB_TSIZE(i))); - /* which is NV40_PGRAPH_TSIZE0(i) ?? */ - nv_wr32(dev, 0x00400900 + i * 0x10, - nv_rd32(dev, NV10_PFB_TILE(i))); - /* which is NV40_PGRAPH_TILE0(i) ?? */ - } + /* Turn all the tiling regions off. */ + for (i = 0; i < NV10_PFB_TILE__SIZE; i++) + nv20_graph_set_region_tiling(dev, i, 0, 0, 0); nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000100); nv_wr32(dev, NV10_PGRAPH_STATE , 0xFFFFFFFF); diff --git a/drivers/gpu/drm/nouveau/nv40_fb.c b/drivers/gpu/drm/nouveau/nv40_fb.c index ca1d27107a8e..3cd07d8d5bd7 100644 --- a/drivers/gpu/drm/nouveau/nv40_fb.c +++ b/drivers/gpu/drm/nouveau/nv40_fb.c @@ -3,12 +3,37 @@ #include "nouveau_drv.h" #include "nouveau_drm.h" +void +nv40_fb_set_region_tiling(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + uint32_t limit = max(1u, addr + size) - 1; + + if (pitch) + addr |= 1; + + switch (dev_priv->chipset) { + case 0x40: + nv_wr32(dev, NV10_PFB_TLIMIT(i), limit); + nv_wr32(dev, NV10_PFB_TSIZE(i), pitch); + nv_wr32(dev, NV10_PFB_TILE(i), addr); + break; + + default: + nv_wr32(dev, NV40_PFB_TLIMIT(i), limit); + nv_wr32(dev, NV40_PFB_TSIZE(i), pitch); + nv_wr32(dev, NV40_PFB_TILE(i), addr); + break; + } +} + int nv40_fb_init(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - uint32_t fb_bar_size, tmp; - int num_tiles; + struct nouveau_fb_engine *pfb = &dev_priv->engine.fb; + uint32_t tmp; int i; /* This is strictly a NV4x register (don't know about NV5x). */ @@ -23,35 +48,23 @@ nv40_fb_init(struct drm_device *dev) case 0x45: tmp = nv_rd32(dev, NV10_PFB_CLOSE_PAGE2); nv_wr32(dev, NV10_PFB_CLOSE_PAGE2, tmp & ~(1 << 15)); - num_tiles = NV10_PFB_TILE__SIZE; + pfb->num_tiles = NV10_PFB_TILE__SIZE; break; case 0x46: /* G72 */ case 0x47: /* G70 */ case 0x49: /* G71 */ case 0x4b: /* G73 */ case 0x4c: /* C51 (G7X version) */ - num_tiles = NV40_PFB_TILE__SIZE_1; + pfb->num_tiles = NV40_PFB_TILE__SIZE_1; break; default: - num_tiles = NV40_PFB_TILE__SIZE_0; + pfb->num_tiles = NV40_PFB_TILE__SIZE_0; break; } - fb_bar_size = drm_get_resource_len(dev, 0) - 1; - switch (dev_priv->chipset) { - case 0x40: - for (i = 0; i < num_tiles; i++) { - nv_wr32(dev, NV10_PFB_TILE(i), 0); - nv_wr32(dev, NV10_PFB_TLIMIT(i), fb_bar_size); - } - break; - default: - for (i = 0; i < num_tiles; i++) { - nv_wr32(dev, NV40_PFB_TILE(i), 0); - nv_wr32(dev, NV40_PFB_TLIMIT(i), fb_bar_size); - } - break; - } + /* Turn all the tiling regions off. */ + for (i = 0; i < pfb->num_tiles; i++) + pfb->set_region_tiling(dev, i, 0, 0, 0); return 0; } diff --git a/drivers/gpu/drm/nouveau/nv40_graph.c b/drivers/gpu/drm/nouveau/nv40_graph.c index d3e0a2a6acf8..53e8afe1dcd1 100644 --- a/drivers/gpu/drm/nouveau/nv40_graph.c +++ b/drivers/gpu/drm/nouveau/nv40_graph.c @@ -24,36 +24,10 @@ * */ -#include <linux/firmware.h> - #include "drmP.h" #include "drm.h" #include "nouveau_drv.h" - -MODULE_FIRMWARE("nouveau/nv40.ctxprog"); -MODULE_FIRMWARE("nouveau/nv40.ctxvals"); -MODULE_FIRMWARE("nouveau/nv41.ctxprog"); -MODULE_FIRMWARE("nouveau/nv41.ctxvals"); -MODULE_FIRMWARE("nouveau/nv42.ctxprog"); -MODULE_FIRMWARE("nouveau/nv42.ctxvals"); -MODULE_FIRMWARE("nouveau/nv43.ctxprog"); -MODULE_FIRMWARE("nouveau/nv43.ctxvals"); -MODULE_FIRMWARE("nouveau/nv44.ctxprog"); -MODULE_FIRMWARE("nouveau/nv44.ctxvals"); -MODULE_FIRMWARE("nouveau/nv46.ctxprog"); -MODULE_FIRMWARE("nouveau/nv46.ctxvals"); -MODULE_FIRMWARE("nouveau/nv47.ctxprog"); -MODULE_FIRMWARE("nouveau/nv47.ctxvals"); -MODULE_FIRMWARE("nouveau/nv49.ctxprog"); -MODULE_FIRMWARE("nouveau/nv49.ctxvals"); -MODULE_FIRMWARE("nouveau/nv4a.ctxprog"); -MODULE_FIRMWARE("nouveau/nv4a.ctxvals"); -MODULE_FIRMWARE("nouveau/nv4b.ctxprog"); -MODULE_FIRMWARE("nouveau/nv4b.ctxvals"); -MODULE_FIRMWARE("nouveau/nv4c.ctxprog"); -MODULE_FIRMWARE("nouveau/nv4c.ctxvals"); -MODULE_FIRMWARE("nouveau/nv4e.ctxprog"); -MODULE_FIRMWARE("nouveau/nv4e.ctxvals"); +#include "nouveau_grctx.h" struct nouveau_channel * nv40_graph_channel(struct drm_device *dev) @@ -83,27 +57,30 @@ nv40_graph_create_context(struct nouveau_channel *chan) { struct drm_device *dev = chan->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nouveau_gpuobj *ctx; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; int ret; - /* Allocate a 175KiB block of PRAMIN to store the context. This - * is massive overkill for a lot of chipsets, but it should be safe - * until we're able to implement this properly (will happen at more - * or less the same time we're able to write our own context programs. - */ - ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, 175*1024, 16, - NVOBJ_FLAG_ZERO_ALLOC, - &chan->ramin_grctx); + ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, pgraph->grctx_size, + 16, NVOBJ_FLAG_ZERO_ALLOC, + &chan->ramin_grctx); if (ret) return ret; - ctx = chan->ramin_grctx->gpuobj; /* Initialise default context values */ dev_priv->engine.instmem.prepare_access(dev, true); - nv40_grctx_vals_load(dev, ctx); - nv_wo32(dev, ctx, 0, ctx->im_pramin->start); - dev_priv->engine.instmem.finish_access(dev); + if (!pgraph->ctxprog) { + struct nouveau_grctx ctx = {}; + ctx.dev = chan->dev; + ctx.mode = NOUVEAU_GRCTX_VALS; + ctx.data = chan->ramin_grctx->gpuobj; + nv40_grctx_init(&ctx); + } else { + nouveau_grctx_vals_load(dev, chan->ramin_grctx->gpuobj); + } + nv_wo32(dev, chan->ramin_grctx->gpuobj, 0, + chan->ramin_grctx->gpuobj->im_pramin->start); + dev_priv->engine.instmem.finish_access(dev); return 0; } @@ -204,133 +181,46 @@ nv40_graph_unload_context(struct drm_device *dev) return ret; } -struct nouveau_ctxprog { - uint32_t signature; - uint8_t version; - uint16_t length; - uint32_t data[]; -} __attribute__ ((packed)); - -struct nouveau_ctxvals { - uint32_t signature; - uint8_t version; - uint32_t length; - struct { - uint32_t offset; - uint32_t value; - } data[]; -} __attribute__ ((packed)); - -int -nv40_grctx_init(struct drm_device *dev) -{ - struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; - const int chipset = dev_priv->chipset; - const struct firmware *fw; - const struct nouveau_ctxprog *cp; - const struct nouveau_ctxvals *cv; - char name[32]; - int ret, i; - - pgraph->accel_blocked = true; - - if (!pgraph->ctxprog) { - sprintf(name, "nouveau/nv%02x.ctxprog", chipset); - ret = request_firmware(&fw, name, &dev->pdev->dev); - if (ret) { - NV_ERROR(dev, "No ctxprog for NV%02x\n", chipset); - return ret; - } - - pgraph->ctxprog = kmalloc(fw->size, GFP_KERNEL); - if (!pgraph->ctxprog) { - NV_ERROR(dev, "OOM copying ctxprog\n"); - release_firmware(fw); - return -ENOMEM; - } - memcpy(pgraph->ctxprog, fw->data, fw->size); - - cp = pgraph->ctxprog; - if (cp->signature != 0x5043564e || cp->version != 0 || - cp->length != ((fw->size - 7) / 4)) { - NV_ERROR(dev, "ctxprog invalid\n"); - release_firmware(fw); - nv40_grctx_fini(dev); - return -EINVAL; - } - release_firmware(fw); - } - - if (!pgraph->ctxvals) { - sprintf(name, "nouveau/nv%02x.ctxvals", chipset); - ret = request_firmware(&fw, name, &dev->pdev->dev); - if (ret) { - NV_ERROR(dev, "No ctxvals for NV%02x\n", chipset); - nv40_grctx_fini(dev); - return ret; - } - - pgraph->ctxvals = kmalloc(fw->size, GFP_KERNEL); - if (!pgraph->ctxprog) { - NV_ERROR(dev, "OOM copying ctxprog\n"); - release_firmware(fw); - nv40_grctx_fini(dev); - return -ENOMEM; - } - memcpy(pgraph->ctxvals, fw->data, fw->size); - - cv = (void *)pgraph->ctxvals; - if (cv->signature != 0x5643564e || cv->version != 0 || - cv->length != ((fw->size - 9) / 8)) { - NV_ERROR(dev, "ctxvals invalid\n"); - release_firmware(fw); - nv40_grctx_fini(dev); - return -EINVAL; - } - release_firmware(fw); - } - - cp = pgraph->ctxprog; - - nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0); - for (i = 0; i < cp->length; i++) - nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, cp->data[i]); - - pgraph->accel_blocked = false; - return 0; -} - void -nv40_grctx_fini(struct drm_device *dev) +nv40_graph_set_region_tiling(struct drm_device *dev, int i, uint32_t addr, + uint32_t size, uint32_t pitch) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; - - if (pgraph->ctxprog) { - kfree(pgraph->ctxprog); - pgraph->ctxprog = NULL; - } + uint32_t limit = max(1u, addr + size) - 1; - if (pgraph->ctxvals) { - kfree(pgraph->ctxprog); - pgraph->ctxvals = NULL; - } -} + if (pitch) + addr |= 1; -void -nv40_grctx_vals_load(struct drm_device *dev, struct nouveau_gpuobj *ctx) -{ - struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; - struct nouveau_ctxvals *cv = pgraph->ctxvals; - int i; + switch (dev_priv->chipset) { + case 0x44: + case 0x4a: + case 0x4e: + nv_wr32(dev, NV20_PGRAPH_TSIZE(i), pitch); + nv_wr32(dev, NV20_PGRAPH_TLIMIT(i), limit); + nv_wr32(dev, NV20_PGRAPH_TILE(i), addr); + break; - if (!cv) - return; + case 0x46: + case 0x47: + case 0x49: + case 0x4b: + nv_wr32(dev, NV47_PGRAPH_TSIZE(i), pitch); + nv_wr32(dev, NV47_PGRAPH_TLIMIT(i), limit); + nv_wr32(dev, NV47_PGRAPH_TILE(i), addr); + nv_wr32(dev, NV40_PGRAPH_TSIZE1(i), pitch); + nv_wr32(dev, NV40_PGRAPH_TLIMIT1(i), limit); + nv_wr32(dev, NV40_PGRAPH_TILE1(i), addr); + break; - for (i = 0; i < cv->length; i++) - nv_wo32(dev, ctx, cv->data[i].offset, cv->data[i].value); + default: + nv_wr32(dev, NV20_PGRAPH_TSIZE(i), pitch); + nv_wr32(dev, NV20_PGRAPH_TLIMIT(i), limit); + nv_wr32(dev, NV20_PGRAPH_TILE(i), addr); + nv_wr32(dev, NV40_PGRAPH_TSIZE1(i), pitch); + nv_wr32(dev, NV40_PGRAPH_TLIMIT1(i), limit); + nv_wr32(dev, NV40_PGRAPH_TILE1(i), addr); + break; + } } /* @@ -347,7 +237,8 @@ nv40_graph_init(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = (struct drm_nouveau_private *)dev->dev_private; - uint32_t vramsz, tmp; + struct nouveau_fb_engine *pfb = &dev_priv->engine.fb; + uint32_t vramsz; int i, j; nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) & @@ -355,7 +246,26 @@ nv40_graph_init(struct drm_device *dev) nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) | NV_PMC_ENABLE_PGRAPH); - nv40_grctx_init(dev); + if (nouveau_ctxfw) { + nouveau_grctx_prog_load(dev); + dev_priv->engine.graph.grctx_size = 175 * 1024; + } + + if (!dev_priv->engine.graph.ctxprog) { + struct nouveau_grctx ctx = {}; + uint32_t cp[256]; + + ctx.dev = dev; + ctx.mode = NOUVEAU_GRCTX_PROG; + ctx.data = cp; + ctx.ctxprog_max = 256; + nv40_grctx_init(&ctx); + dev_priv->engine.graph.grctx_size = ctx.ctxvals_pos * 4; + + nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0); + for (i = 0; i < ctx.ctxprog_len; i++) + nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, cp[i]); + } /* No context present currently */ nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0x00000000); @@ -425,74 +335,9 @@ nv40_graph_init(struct drm_device *dev) nv_wr32(dev, 0x400b38, 0x2ffff800); nv_wr32(dev, 0x400b3c, 0x00006000); - /* copy tile info from PFB */ - switch (dev_priv->chipset) { - case 0x40: /* vanilla NV40 */ - for (i = 0; i < NV10_PFB_TILE__SIZE; i++) { - tmp = nv_rd32(dev, NV10_PFB_TILE(i)); - nv_wr32(dev, NV40_PGRAPH_TILE0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TILE1(i), tmp); - tmp = nv_rd32(dev, NV10_PFB_TLIMIT(i)); - nv_wr32(dev, NV40_PGRAPH_TLIMIT0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TLIMIT1(i), tmp); - tmp = nv_rd32(dev, NV10_PFB_TSIZE(i)); - nv_wr32(dev, NV40_PGRAPH_TSIZE0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TSIZE1(i), tmp); - tmp = nv_rd32(dev, NV10_PFB_TSTATUS(i)); - nv_wr32(dev, NV40_PGRAPH_TSTATUS0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TSTATUS1(i), tmp); - } - break; - case 0x44: - case 0x4a: - case 0x4e: /* NV44-based cores don't have 0x406900? */ - for (i = 0; i < NV40_PFB_TILE__SIZE_0; i++) { - tmp = nv_rd32(dev, NV40_PFB_TILE(i)); - nv_wr32(dev, NV40_PGRAPH_TILE0(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TLIMIT(i)); - nv_wr32(dev, NV40_PGRAPH_TLIMIT0(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TSIZE(i)); - nv_wr32(dev, NV40_PGRAPH_TSIZE0(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TSTATUS(i)); - nv_wr32(dev, NV40_PGRAPH_TSTATUS0(i), tmp); - } - break; - case 0x46: - case 0x47: - case 0x49: - case 0x4b: /* G7X-based cores */ - for (i = 0; i < NV40_PFB_TILE__SIZE_1; i++) { - tmp = nv_rd32(dev, NV40_PFB_TILE(i)); - nv_wr32(dev, NV47_PGRAPH_TILE0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TILE1(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TLIMIT(i)); - nv_wr32(dev, NV47_PGRAPH_TLIMIT0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TLIMIT1(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TSIZE(i)); - nv_wr32(dev, NV47_PGRAPH_TSIZE0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TSIZE1(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TSTATUS(i)); - nv_wr32(dev, NV47_PGRAPH_TSTATUS0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TSTATUS1(i), tmp); - } - break; - default: /* everything else */ - for (i = 0; i < NV40_PFB_TILE__SIZE_0; i++) { - tmp = nv_rd32(dev, NV40_PFB_TILE(i)); - nv_wr32(dev, NV40_PGRAPH_TILE0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TILE1(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TLIMIT(i)); - nv_wr32(dev, NV40_PGRAPH_TLIMIT0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TLIMIT1(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TSIZE(i)); - nv_wr32(dev, NV40_PGRAPH_TSIZE0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TSIZE1(i), tmp); - tmp = nv_rd32(dev, NV40_PFB_TSTATUS(i)); - nv_wr32(dev, NV40_PGRAPH_TSTATUS0(i), tmp); - nv_wr32(dev, NV40_PGRAPH_TSTATUS1(i), tmp); - } - break; - } + /* Turn all the tiling regions off. */ + for (i = 0; i < pfb->num_tiles; i++) + nv40_graph_set_region_tiling(dev, i, 0, 0, 0); /* begin RAM config */ vramsz = drm_get_resource_len(dev, 0) - 1; @@ -535,6 +380,7 @@ nv40_graph_init(struct drm_device *dev) void nv40_graph_takedown(struct drm_device *dev) { + nouveau_grctx_fini(dev); } struct nouveau_pgraph_object_class nv40_graph_grclass[] = { diff --git a/drivers/gpu/drm/nouveau/nv40_grctx.c b/drivers/gpu/drm/nouveau/nv40_grctx.c new file mode 100644 index 000000000000..11b11c31f543 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nv40_grctx.c @@ -0,0 +1,678 @@ +/* + * Copyright 2009 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +/* NVIDIA context programs handle a number of other conditions which are + * not implemented in our versions. It's not clear why NVIDIA context + * programs have this code, nor whether it's strictly necessary for + * correct operation. We'll implement additional handling if/when we + * discover it's necessary. + * + * - On context save, NVIDIA set 0x400314 bit 0 to 1 if the "3D state" + * flag is set, this gets saved into the context. + * - On context save, the context program for all cards load nsource + * into a flag register and check for ILLEGAL_MTHD. If it's set, + * opcode 0x60000d is called before resuming normal operation. + * - Some context programs check more conditions than the above. NV44 + * checks: ((nsource & 0x0857) || (0x400718 & 0x0100) || (intr & 0x0001)) + * and calls 0x60000d before resuming normal operation. + * - At the very beginning of NVIDIA's context programs, flag 9 is checked + * and if true 0x800001 is called with count=0, pos=0, the flag is cleared + * and then the ctxprog is aborted. It looks like a complicated NOP, + * its purpose is unknown. + * - In the section of code that loads the per-vs state, NVIDIA check + * flag 10. If it's set, they only transfer the small 0x300 byte block + * of state + the state for a single vs as opposed to the state for + * all vs units. It doesn't seem likely that it'll occur in normal + * operation, especially seeing as it appears NVIDIA may have screwed + * up the ctxprogs for some cards and have an invalid instruction + * rather than a cp_lsr(ctx, dwords_for_1_vs_unit) instruction. + * - There's a number of places where context offset 0 (where we place + * the PRAMIN offset of the context) is loaded into either 0x408000, + * 0x408004 or 0x408008. Not sure what's up there either. + * - The ctxprogs for some cards save 0x400a00 again during the cleanup + * path for auto-loadctx. + */ + +#define CP_FLAG_CLEAR 0 +#define CP_FLAG_SET 1 +#define CP_FLAG_SWAP_DIRECTION ((0 * 32) + 0) +#define CP_FLAG_SWAP_DIRECTION_LOAD 0 +#define CP_FLAG_SWAP_DIRECTION_SAVE 1 +#define CP_FLAG_USER_SAVE ((0 * 32) + 5) +#define CP_FLAG_USER_SAVE_NOT_PENDING 0 +#define CP_FLAG_USER_SAVE_PENDING 1 +#define CP_FLAG_USER_LOAD ((0 * 32) + 6) +#define CP_FLAG_USER_LOAD_NOT_PENDING 0 +#define CP_FLAG_USER_LOAD_PENDING 1 +#define CP_FLAG_STATUS ((3 * 32) + 0) +#define CP_FLAG_STATUS_IDLE 0 +#define CP_FLAG_STATUS_BUSY 1 +#define CP_FLAG_AUTO_SAVE ((3 * 32) + 4) +#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0 +#define CP_FLAG_AUTO_SAVE_PENDING 1 +#define CP_FLAG_AUTO_LOAD ((3 * 32) + 5) +#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0 +#define CP_FLAG_AUTO_LOAD_PENDING 1 +#define CP_FLAG_UNK54 ((3 * 32) + 6) +#define CP_FLAG_UNK54_CLEAR 0 +#define CP_FLAG_UNK54_SET 1 +#define CP_FLAG_ALWAYS ((3 * 32) + 8) +#define CP_FLAG_ALWAYS_FALSE 0 +#define CP_FLAG_ALWAYS_TRUE 1 +#define CP_FLAG_UNK57 ((3 * 32) + 9) +#define CP_FLAG_UNK57_CLEAR 0 +#define CP_FLAG_UNK57_SET 1 + +#define CP_CTX 0x00100000 +#define CP_CTX_COUNT 0x000fc000 +#define CP_CTX_COUNT_SHIFT 14 +#define CP_CTX_REG 0x00003fff +#define CP_LOAD_SR 0x00200000 +#define CP_LOAD_SR_VALUE 0x000fffff +#define CP_BRA 0x00400000 +#define CP_BRA_IP 0x0000ff00 +#define CP_BRA_IP_SHIFT 8 +#define CP_BRA_IF_CLEAR 0x00000080 +#define CP_BRA_FLAG 0x0000007f +#define CP_WAIT 0x00500000 +#define CP_WAIT_SET 0x00000080 +#define CP_WAIT_FLAG 0x0000007f +#define CP_SET 0x00700000 +#define CP_SET_1 0x00000080 +#define CP_SET_FLAG 0x0000007f +#define CP_NEXT_TO_SWAP 0x00600007 +#define CP_NEXT_TO_CURRENT 0x00600009 +#define CP_SET_CONTEXT_POINTER 0x0060000a +#define CP_END 0x0060000e +#define CP_LOAD_MAGIC_UNK01 0x00800001 /* unknown */ +#define CP_LOAD_MAGIC_NV44TCL 0x00800029 /* per-vs state (0x4497) */ +#define CP_LOAD_MAGIC_NV40TCL 0x00800041 /* per-vs state (0x4097) */ + +#include "drmP.h" +#include "nouveau_drv.h" +#include "nouveau_grctx.h" + +/* TODO: + * - get vs count from 0x1540 + * - document unimplemented bits compared to nvidia + * - nsource handling + * - R0 & 0x0200 handling + * - single-vs handling + * - 400314 bit 0 + */ + +static int +nv40_graph_4097(struct drm_device *dev) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + + if ((dev_priv->chipset & 0xf0) == 0x60) + return 0; + + return !!(0x0baf & (1 << dev_priv->chipset)); +} + +static int +nv40_graph_vs_count(struct drm_device *dev) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + + switch (dev_priv->chipset) { + case 0x47: + case 0x49: + case 0x4b: + return 8; + case 0x40: + return 6; + case 0x41: + case 0x42: + return 5; + case 0x43: + case 0x44: + case 0x46: + case 0x4a: + return 3; + case 0x4c: + case 0x4e: + case 0x67: + default: + return 1; + } +} + + +enum cp_label { + cp_check_load = 1, + cp_setup_auto_load, + cp_setup_load, + cp_setup_save, + cp_swap_state, + cp_swap_state3d_3_is_save, + cp_prepare_exit, + cp_exit, +}; + +static void +nv40_graph_construct_general(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int i; + + cp_ctx(ctx, 0x4000a4, 1); + gr_def(ctx, 0x4000a4, 0x00000008); + cp_ctx(ctx, 0x400144, 58); + gr_def(ctx, 0x400144, 0x00000001); + cp_ctx(ctx, 0x400314, 1); + gr_def(ctx, 0x400314, 0x00000000); + cp_ctx(ctx, 0x400400, 10); + cp_ctx(ctx, 0x400480, 10); + cp_ctx(ctx, 0x400500, 19); + gr_def(ctx, 0x400514, 0x00040000); + gr_def(ctx, 0x400524, 0x55555555); + gr_def(ctx, 0x400528, 0x55555555); + gr_def(ctx, 0x40052c, 0x55555555); + gr_def(ctx, 0x400530, 0x55555555); + cp_ctx(ctx, 0x400560, 6); + gr_def(ctx, 0x400568, 0x0000ffff); + gr_def(ctx, 0x40056c, 0x0000ffff); + cp_ctx(ctx, 0x40057c, 5); + cp_ctx(ctx, 0x400710, 3); + gr_def(ctx, 0x400710, 0x20010001); + gr_def(ctx, 0x400714, 0x0f73ef00); + cp_ctx(ctx, 0x400724, 1); + gr_def(ctx, 0x400724, 0x02008821); + cp_ctx(ctx, 0x400770, 3); + if (dev_priv->chipset == 0x40) { + cp_ctx(ctx, 0x400814, 4); + cp_ctx(ctx, 0x400828, 5); + cp_ctx(ctx, 0x400840, 5); + gr_def(ctx, 0x400850, 0x00000040); + cp_ctx(ctx, 0x400858, 4); + gr_def(ctx, 0x400858, 0x00000040); + gr_def(ctx, 0x40085c, 0x00000040); + gr_def(ctx, 0x400864, 0x80000000); + cp_ctx(ctx, 0x40086c, 9); + gr_def(ctx, 0x40086c, 0x80000000); + gr_def(ctx, 0x400870, 0x80000000); + gr_def(ctx, 0x400874, 0x80000000); + gr_def(ctx, 0x400878, 0x80000000); + gr_def(ctx, 0x400888, 0x00000040); + gr_def(ctx, 0x40088c, 0x80000000); + cp_ctx(ctx, 0x4009c0, 8); + gr_def(ctx, 0x4009cc, 0x80000000); + gr_def(ctx, 0x4009dc, 0x80000000); + } else { + cp_ctx(ctx, 0x400840, 20); + if (!nv40_graph_4097(ctx->dev)) { + for (i = 0; i < 8; i++) + gr_def(ctx, 0x400860 + (i * 4), 0x00000001); + } + gr_def(ctx, 0x400880, 0x00000040); + gr_def(ctx, 0x400884, 0x00000040); + gr_def(ctx, 0x400888, 0x00000040); + cp_ctx(ctx, 0x400894, 11); + gr_def(ctx, 0x400894, 0x00000040); + if (nv40_graph_4097(ctx->dev)) { + for (i = 0; i < 8; i++) + gr_def(ctx, 0x4008a0 + (i * 4), 0x80000000); + } + cp_ctx(ctx, 0x4008e0, 2); + cp_ctx(ctx, 0x4008f8, 2); + if (dev_priv->chipset == 0x4c || + (dev_priv->chipset & 0xf0) == 0x60) + cp_ctx(ctx, 0x4009f8, 1); + } + cp_ctx(ctx, 0x400a00, 73); + gr_def(ctx, 0x400b0c, 0x0b0b0b0c); + cp_ctx(ctx, 0x401000, 4); + cp_ctx(ctx, 0x405004, 1); + switch (dev_priv->chipset) { + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x403448, 1); + gr_def(ctx, 0x403448, 0x00001010); + break; + default: + cp_ctx(ctx, 0x403440, 1); + switch (dev_priv->chipset) { + case 0x40: + gr_def(ctx, 0x403440, 0x00000010); + break; + case 0x44: + case 0x46: + case 0x4a: + gr_def(ctx, 0x403440, 0x00003010); + break; + case 0x41: + case 0x42: + case 0x43: + case 0x4c: + case 0x4e: + case 0x67: + default: + gr_def(ctx, 0x403440, 0x00001010); + break; + } + break; + } +} + +static void +nv40_graph_construct_state3d(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int i; + + if (dev_priv->chipset == 0x40) { + cp_ctx(ctx, 0x401880, 51); + gr_def(ctx, 0x401940, 0x00000100); + } else + if (dev_priv->chipset == 0x46 || dev_priv->chipset == 0x47 || + dev_priv->chipset == 0x49 || dev_priv->chipset == 0x4b) { + cp_ctx(ctx, 0x401880, 32); + for (i = 0; i < 16; i++) + gr_def(ctx, 0x401880 + (i * 4), 0x00000111); + if (dev_priv->chipset == 0x46) + cp_ctx(ctx, 0x401900, 16); + cp_ctx(ctx, 0x401940, 3); + } + cp_ctx(ctx, 0x40194c, 18); + gr_def(ctx, 0x401954, 0x00000111); + gr_def(ctx, 0x401958, 0x00080060); + gr_def(ctx, 0x401974, 0x00000080); + gr_def(ctx, 0x401978, 0xffff0000); + gr_def(ctx, 0x40197c, 0x00000001); + gr_def(ctx, 0x401990, 0x46400000); + if (dev_priv->chipset == 0x40) { + cp_ctx(ctx, 0x4019a0, 2); + cp_ctx(ctx, 0x4019ac, 5); + } else { + cp_ctx(ctx, 0x4019a0, 1); + cp_ctx(ctx, 0x4019b4, 3); + } + gr_def(ctx, 0x4019bc, 0xffff0000); + switch (dev_priv->chipset) { + case 0x46: + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x4019c0, 18); + for (i = 0; i < 16; i++) + gr_def(ctx, 0x4019c0 + (i * 4), 0x88888888); + break; + } + cp_ctx(ctx, 0x401a08, 8); + gr_def(ctx, 0x401a10, 0x0fff0000); + gr_def(ctx, 0x401a14, 0x0fff0000); + gr_def(ctx, 0x401a1c, 0x00011100); + cp_ctx(ctx, 0x401a2c, 4); + cp_ctx(ctx, 0x401a44, 26); + for (i = 0; i < 16; i++) + gr_def(ctx, 0x401a44 + (i * 4), 0x07ff0000); + gr_def(ctx, 0x401a8c, 0x4b7fffff); + if (dev_priv->chipset == 0x40) { + cp_ctx(ctx, 0x401ab8, 3); + } else { + cp_ctx(ctx, 0x401ab8, 1); + cp_ctx(ctx, 0x401ac0, 1); + } + cp_ctx(ctx, 0x401ad0, 8); + gr_def(ctx, 0x401ad0, 0x30201000); + gr_def(ctx, 0x401ad4, 0x70605040); + gr_def(ctx, 0x401ad8, 0xb8a89888); + gr_def(ctx, 0x401adc, 0xf8e8d8c8); + cp_ctx(ctx, 0x401b10, dev_priv->chipset == 0x40 ? 2 : 1); + gr_def(ctx, 0x401b10, 0x40100000); + cp_ctx(ctx, 0x401b18, dev_priv->chipset == 0x40 ? 6 : 5); + gr_def(ctx, 0x401b28, dev_priv->chipset == 0x40 ? + 0x00000004 : 0x00000000); + cp_ctx(ctx, 0x401b30, 25); + gr_def(ctx, 0x401b34, 0x0000ffff); + gr_def(ctx, 0x401b68, 0x435185d6); + gr_def(ctx, 0x401b6c, 0x2155b699); + gr_def(ctx, 0x401b70, 0xfedcba98); + gr_def(ctx, 0x401b74, 0x00000098); + gr_def(ctx, 0x401b84, 0xffffffff); + gr_def(ctx, 0x401b88, 0x00ff7000); + gr_def(ctx, 0x401b8c, 0x0000ffff); + if (dev_priv->chipset != 0x44 && dev_priv->chipset != 0x4a && + dev_priv->chipset != 0x4e) + cp_ctx(ctx, 0x401b94, 1); + cp_ctx(ctx, 0x401b98, 8); + gr_def(ctx, 0x401b9c, 0x00ff0000); + cp_ctx(ctx, 0x401bc0, 9); + gr_def(ctx, 0x401be0, 0x00ffff00); + cp_ctx(ctx, 0x401c00, 192); + for (i = 0; i < 16; i++) { /* fragment texture units */ + gr_def(ctx, 0x401c40 + (i * 4), 0x00018488); + gr_def(ctx, 0x401c80 + (i * 4), 0x00028202); + gr_def(ctx, 0x401d00 + (i * 4), 0x0000aae4); + gr_def(ctx, 0x401d40 + (i * 4), 0x01012000); + gr_def(ctx, 0x401d80 + (i * 4), 0x00080008); + gr_def(ctx, 0x401e00 + (i * 4), 0x00100008); + } + for (i = 0; i < 4; i++) { /* vertex texture units */ + gr_def(ctx, 0x401e90 + (i * 4), 0x0001bc80); + gr_def(ctx, 0x401ea0 + (i * 4), 0x00000202); + gr_def(ctx, 0x401ec0 + (i * 4), 0x00000008); + gr_def(ctx, 0x401ee0 + (i * 4), 0x00080008); + } + cp_ctx(ctx, 0x400f5c, 3); + gr_def(ctx, 0x400f5c, 0x00000002); + cp_ctx(ctx, 0x400f84, 1); +} + +static void +nv40_graph_construct_state3d_2(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int i; + + cp_ctx(ctx, 0x402000, 1); + cp_ctx(ctx, 0x402404, dev_priv->chipset == 0x40 ? 1 : 2); + switch (dev_priv->chipset) { + case 0x40: + gr_def(ctx, 0x402404, 0x00000001); + break; + case 0x4c: + case 0x4e: + case 0x67: + gr_def(ctx, 0x402404, 0x00000020); + break; + case 0x46: + case 0x49: + case 0x4b: + gr_def(ctx, 0x402404, 0x00000421); + break; + default: + gr_def(ctx, 0x402404, 0x00000021); + } + if (dev_priv->chipset != 0x40) + gr_def(ctx, 0x402408, 0x030c30c3); + switch (dev_priv->chipset) { + case 0x44: + case 0x46: + case 0x4a: + case 0x4c: + case 0x4e: + case 0x67: + cp_ctx(ctx, 0x402440, 1); + gr_def(ctx, 0x402440, 0x00011001); + break; + default: + break; + } + cp_ctx(ctx, 0x402480, dev_priv->chipset == 0x40 ? 8 : 9); + gr_def(ctx, 0x402488, 0x3e020200); + gr_def(ctx, 0x40248c, 0x00ffffff); + switch (dev_priv->chipset) { + case 0x40: + gr_def(ctx, 0x402490, 0x60103f00); + break; + case 0x47: + gr_def(ctx, 0x402490, 0x40103f00); + break; + case 0x41: + case 0x42: + case 0x49: + case 0x4b: + gr_def(ctx, 0x402490, 0x20103f00); + break; + default: + gr_def(ctx, 0x402490, 0x0c103f00); + break; + } + gr_def(ctx, 0x40249c, dev_priv->chipset <= 0x43 ? + 0x00020000 : 0x00040000); + cp_ctx(ctx, 0x402500, 31); + gr_def(ctx, 0x402530, 0x00008100); + if (dev_priv->chipset == 0x40) + cp_ctx(ctx, 0x40257c, 6); + cp_ctx(ctx, 0x402594, 16); + cp_ctx(ctx, 0x402800, 17); + gr_def(ctx, 0x402800, 0x00000001); + switch (dev_priv->chipset) { + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x402864, 1); + gr_def(ctx, 0x402864, 0x00001001); + cp_ctx(ctx, 0x402870, 3); + gr_def(ctx, 0x402878, 0x00000003); + if (dev_priv->chipset != 0x47) { /* belong at end!! */ + cp_ctx(ctx, 0x402900, 1); + cp_ctx(ctx, 0x402940, 1); + cp_ctx(ctx, 0x402980, 1); + cp_ctx(ctx, 0x4029c0, 1); + cp_ctx(ctx, 0x402a00, 1); + cp_ctx(ctx, 0x402a40, 1); + cp_ctx(ctx, 0x402a80, 1); + cp_ctx(ctx, 0x402ac0, 1); + } + break; + case 0x40: + cp_ctx(ctx, 0x402844, 1); + gr_def(ctx, 0x402844, 0x00000001); + cp_ctx(ctx, 0x402850, 1); + break; + default: + cp_ctx(ctx, 0x402844, 1); + gr_def(ctx, 0x402844, 0x00001001); + cp_ctx(ctx, 0x402850, 2); + gr_def(ctx, 0x402854, 0x00000003); + break; + } + + cp_ctx(ctx, 0x402c00, 4); + gr_def(ctx, 0x402c00, dev_priv->chipset == 0x40 ? + 0x80800001 : 0x00888001); + switch (dev_priv->chipset) { + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x402c20, 40); + for (i = 0; i < 32; i++) + gr_def(ctx, 0x402c40 + (i * 4), 0xffffffff); + cp_ctx(ctx, 0x4030b8, 13); + gr_def(ctx, 0x4030dc, 0x00000005); + gr_def(ctx, 0x4030e8, 0x0000ffff); + break; + default: + cp_ctx(ctx, 0x402c10, 4); + if (dev_priv->chipset == 0x40) + cp_ctx(ctx, 0x402c20, 36); + else + if (dev_priv->chipset <= 0x42) + cp_ctx(ctx, 0x402c20, 24); + else + if (dev_priv->chipset <= 0x4a) + cp_ctx(ctx, 0x402c20, 16); + else + cp_ctx(ctx, 0x402c20, 8); + cp_ctx(ctx, 0x402cb0, dev_priv->chipset == 0x40 ? 12 : 13); + gr_def(ctx, 0x402cd4, 0x00000005); + if (dev_priv->chipset != 0x40) + gr_def(ctx, 0x402ce0, 0x0000ffff); + break; + } + + cp_ctx(ctx, 0x403400, dev_priv->chipset == 0x40 ? 4 : 3); + cp_ctx(ctx, 0x403410, dev_priv->chipset == 0x40 ? 4 : 3); + cp_ctx(ctx, 0x403420, nv40_graph_vs_count(ctx->dev)); + for (i = 0; i < nv40_graph_vs_count(ctx->dev); i++) + gr_def(ctx, 0x403420 + (i * 4), 0x00005555); + + if (dev_priv->chipset != 0x40) { + cp_ctx(ctx, 0x403600, 1); + gr_def(ctx, 0x403600, 0x00000001); + } + cp_ctx(ctx, 0x403800, 1); + + cp_ctx(ctx, 0x403c18, 1); + gr_def(ctx, 0x403c18, 0x00000001); + switch (dev_priv->chipset) { + case 0x46: + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x405018, 1); + gr_def(ctx, 0x405018, 0x08e00001); + cp_ctx(ctx, 0x405c24, 1); + gr_def(ctx, 0x405c24, 0x000e3000); + break; + } + if (dev_priv->chipset != 0x4e) + cp_ctx(ctx, 0x405800, 11); + cp_ctx(ctx, 0x407000, 1); +} + +static void +nv40_graph_construct_state3d_3(struct nouveau_grctx *ctx) +{ + int len = nv40_graph_4097(ctx->dev) ? 0x0684 : 0x0084; + + cp_out (ctx, 0x300000); + cp_lsr (ctx, len - 4); + cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_swap_state3d_3_is_save); + cp_lsr (ctx, len); + cp_name(ctx, cp_swap_state3d_3_is_save); + cp_out (ctx, 0x800001); + + ctx->ctxvals_pos += len; +} + +static void +nv40_graph_construct_shader(struct nouveau_grctx *ctx) +{ + struct drm_device *dev = ctx->dev; + struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_gpuobj *obj = ctx->data; + int vs, vs_nr, vs_len, vs_nr_b0, vs_nr_b1, b0_offset, b1_offset; + int offset, i; + + vs_nr = nv40_graph_vs_count(ctx->dev); + vs_nr_b0 = 363; + vs_nr_b1 = dev_priv->chipset == 0x40 ? 128 : 64; + if (dev_priv->chipset == 0x40) { + b0_offset = 0x2200/4; /* 33a0 */ + b1_offset = 0x55a0/4; /* 1500 */ + vs_len = 0x6aa0/4; + } else + if (dev_priv->chipset == 0x41 || dev_priv->chipset == 0x42) { + b0_offset = 0x2200/4; /* 2200 */ + b1_offset = 0x4400/4; /* 0b00 */ + vs_len = 0x4f00/4; + } else { + b0_offset = 0x1d40/4; /* 2200 */ + b1_offset = 0x3f40/4; /* 0b00 : 0a40 */ + vs_len = nv40_graph_4097(dev) ? 0x4a40/4 : 0x4980/4; + } + + cp_lsr(ctx, vs_len * vs_nr + 0x300/4); + cp_out(ctx, nv40_graph_4097(dev) ? 0x800041 : 0x800029); + + offset = ctx->ctxvals_pos; + ctx->ctxvals_pos += (0x0300/4 + (vs_nr * vs_len)); + + if (ctx->mode != NOUVEAU_GRCTX_VALS) + return; + + offset += 0x0280/4; + for (i = 0; i < 16; i++, offset += 2) + nv_wo32(dev, obj, offset, 0x3f800000); + + for (vs = 0; vs < vs_nr; vs++, offset += vs_len) { + for (i = 0; i < vs_nr_b0 * 6; i += 6) + nv_wo32(dev, obj, offset + b0_offset + i, 0x00000001); + for (i = 0; i < vs_nr_b1 * 4; i += 4) + nv_wo32(dev, obj, offset + b1_offset + i, 0x3f800000); + } +} + +void +nv40_grctx_init(struct nouveau_grctx *ctx) +{ + /* decide whether we're loading/unloading the context */ + cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save); + cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save); + + cp_name(ctx, cp_check_load); + cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load); + cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load); + cp_bra (ctx, ALWAYS, TRUE, cp_exit); + + /* setup for context load */ + cp_name(ctx, cp_setup_auto_load); + cp_wait(ctx, STATUS, IDLE); + cp_out (ctx, CP_NEXT_TO_SWAP); + cp_name(ctx, cp_setup_load); + cp_wait(ctx, STATUS, IDLE); + cp_set (ctx, SWAP_DIRECTION, LOAD); + cp_out (ctx, 0x00910880); /* ?? */ + cp_out (ctx, 0x00901ffe); /* ?? */ + cp_out (ctx, 0x01940000); /* ?? */ + cp_lsr (ctx, 0x20); + cp_out (ctx, 0x0060000b); /* ?? */ + cp_wait(ctx, UNK57, CLEAR); + cp_out (ctx, 0x0060000c); /* ?? */ + cp_bra (ctx, ALWAYS, TRUE, cp_swap_state); + + /* setup for context save */ + cp_name(ctx, cp_setup_save); + cp_set (ctx, SWAP_DIRECTION, SAVE); + + /* general PGRAPH state */ + cp_name(ctx, cp_swap_state); + cp_pos (ctx, 0x00020/4); + nv40_graph_construct_general(ctx); + cp_wait(ctx, STATUS, IDLE); + + /* 3D state, block 1 */ + cp_bra (ctx, UNK54, CLEAR, cp_prepare_exit); + nv40_graph_construct_state3d(ctx); + cp_wait(ctx, STATUS, IDLE); + + /* 3D state, block 2 */ + nv40_graph_construct_state3d_2(ctx); + + /* Some other block of "random" state */ + nv40_graph_construct_state3d_3(ctx); + + /* Per-vertex shader state */ + cp_pos (ctx, ctx->ctxvals_pos); + nv40_graph_construct_shader(ctx); + + /* pre-exit state updates */ + cp_name(ctx, cp_prepare_exit); + cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load); + cp_bra (ctx, USER_SAVE, PENDING, cp_exit); + cp_out (ctx, CP_NEXT_TO_CURRENT); + + cp_name(ctx, cp_exit); + cp_set (ctx, USER_SAVE, NOT_PENDING); + cp_set (ctx, USER_LOAD, NOT_PENDING); + cp_out (ctx, CP_END); +} + diff --git a/drivers/gpu/drm/nouveau/nv50_crtc.c b/drivers/gpu/drm/nouveau/nv50_crtc.c index f8e28a1e44e7..118d3285fd8c 100644 --- a/drivers/gpu/drm/nouveau/nv50_crtc.c +++ b/drivers/gpu/drm/nouveau/nv50_crtc.c @@ -45,7 +45,7 @@ nv50_crtc_lut_load(struct drm_crtc *crtc) void __iomem *lut = nvbo_kmap_obj_iovirtual(nv_crtc->lut.nvbo); int i; - NV_DEBUG(crtc->dev, "\n"); + NV_DEBUG_KMS(crtc->dev, "\n"); for (i = 0; i < 256; i++) { writew(nv_crtc->lut.r[i] >> 2, lut + 8*i + 0); @@ -68,8 +68,8 @@ nv50_crtc_blank(struct nouveau_crtc *nv_crtc, bool blanked) struct nouveau_channel *evo = dev_priv->evo; int index = nv_crtc->index, ret; - NV_DEBUG(dev, "index %d\n", nv_crtc->index); - NV_DEBUG(dev, "%s\n", blanked ? "blanked" : "unblanked"); + NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index); + NV_DEBUG_KMS(dev, "%s\n", blanked ? "blanked" : "unblanked"); if (blanked) { nv_crtc->cursor.hide(nv_crtc, false); @@ -139,7 +139,7 @@ nv50_crtc_set_dither(struct nouveau_crtc *nv_crtc, bool on, bool update) struct nouveau_channel *evo = dev_priv->evo; int ret; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); ret = RING_SPACE(evo, 2 + (update ? 2 : 0)); if (ret) { @@ -193,7 +193,7 @@ nv50_crtc_set_scale(struct nouveau_crtc *nv_crtc, int scaling_mode, bool update) uint32_t outX, outY, horiz, vert; int ret; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); switch (scaling_mode) { case DRM_MODE_SCALE_NONE: @@ -301,7 +301,7 @@ nv50_crtc_destroy(struct drm_crtc *crtc) struct drm_device *dev = crtc->dev; struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc); - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); if (!crtc) return; @@ -433,7 +433,7 @@ nv50_crtc_prepare(struct drm_crtc *crtc) struct drm_device *dev = crtc->dev; struct drm_encoder *encoder; - NV_DEBUG(dev, "index %d\n", nv_crtc->index); + NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index); /* Disconnect all unused encoders. */ list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { @@ -458,7 +458,7 @@ nv50_crtc_commit(struct drm_crtc *crtc) struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc); int ret; - NV_DEBUG(dev, "index %d\n", nv_crtc->index); + NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index); nv50_crtc_blank(nv_crtc, false); @@ -497,7 +497,7 @@ nv50_crtc_do_mode_set_base(struct drm_crtc *crtc, int x, int y, struct nouveau_framebuffer *fb = nouveau_framebuffer(drm_fb); int ret, format; - NV_DEBUG(dev, "index %d\n", nv_crtc->index); + NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index); switch (drm_fb->depth) { case 8: @@ -612,7 +612,7 @@ nv50_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode, *nv_crtc->mode = *adjusted_mode; - NV_DEBUG(dev, "index %d\n", nv_crtc->index); + NV_DEBUG_KMS(dev, "index %d\n", nv_crtc->index); hsync_dur = adjusted_mode->hsync_end - adjusted_mode->hsync_start; vsync_dur = adjusted_mode->vsync_end - adjusted_mode->vsync_start; @@ -706,7 +706,7 @@ nv50_crtc_create(struct drm_device *dev, int index) struct nouveau_crtc *nv_crtc = NULL; int ret, i; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); nv_crtc = kzalloc(sizeof(*nv_crtc), GFP_KERNEL); if (!nv_crtc) diff --git a/drivers/gpu/drm/nouveau/nv50_cursor.c b/drivers/gpu/drm/nouveau/nv50_cursor.c index e2e79a8f220d..753e723adb3a 100644 --- a/drivers/gpu/drm/nouveau/nv50_cursor.c +++ b/drivers/gpu/drm/nouveau/nv50_cursor.c @@ -41,7 +41,7 @@ nv50_cursor_show(struct nouveau_crtc *nv_crtc, bool update) struct drm_device *dev = nv_crtc->base.dev; int ret; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); if (update && nv_crtc->cursor.visible) return; @@ -76,7 +76,7 @@ nv50_cursor_hide(struct nouveau_crtc *nv_crtc, bool update) struct drm_device *dev = nv_crtc->base.dev; int ret; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); if (update && !nv_crtc->cursor.visible) return; @@ -116,7 +116,7 @@ nv50_cursor_set_pos(struct nouveau_crtc *nv_crtc, int x, int y) static void nv50_cursor_set_offset(struct nouveau_crtc *nv_crtc, uint32_t offset) { - NV_DEBUG(nv_crtc->base.dev, "\n"); + NV_DEBUG_KMS(nv_crtc->base.dev, "\n"); if (offset == nv_crtc->cursor.offset) return; @@ -143,7 +143,7 @@ nv50_cursor_fini(struct nouveau_crtc *nv_crtc) struct drm_device *dev = nv_crtc->base.dev; int idx = nv_crtc->index; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); nv_wr32(dev, NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(idx), 0); if (!nv_wait(NV50_PDISPLAY_CURSOR_CURSOR_CTRL2(idx), diff --git a/drivers/gpu/drm/nouveau/nv50_dac.c b/drivers/gpu/drm/nouveau/nv50_dac.c index fb5838e3be24..f08f042a8e10 100644 --- a/drivers/gpu/drm/nouveau/nv50_dac.c +++ b/drivers/gpu/drm/nouveau/nv50_dac.c @@ -44,7 +44,7 @@ nv50_dac_disconnect(struct nouveau_encoder *nv_encoder) struct nouveau_channel *evo = dev_priv->evo; int ret; - NV_DEBUG(dev, "Disconnecting DAC %d\n", nv_encoder->or); + NV_DEBUG_KMS(dev, "Disconnecting DAC %d\n", nv_encoder->or); ret = RING_SPACE(evo, 2); if (ret) { @@ -81,11 +81,11 @@ nv50_dac_detect(struct drm_encoder *encoder, struct drm_connector *connector) /* Use bios provided value if possible. */ if (dev_priv->vbios->dactestval) { load_pattern = dev_priv->vbios->dactestval; - NV_DEBUG(dev, "Using bios provided load_pattern of %d\n", + NV_DEBUG_KMS(dev, "Using bios provided load_pattern of %d\n", load_pattern); } else { load_pattern = 340; - NV_DEBUG(dev, "Using default load_pattern of %d\n", + NV_DEBUG_KMS(dev, "Using default load_pattern of %d\n", load_pattern); } @@ -103,9 +103,9 @@ nv50_dac_detect(struct drm_encoder *encoder, struct drm_connector *connector) status = connector_status_connected; if (status == connector_status_connected) - NV_DEBUG(dev, "Load was detected on output with or %d\n", or); + NV_DEBUG_KMS(dev, "Load was detected on output with or %d\n", or); else - NV_DEBUG(dev, "Load was not detected on output with or %d\n", or); + NV_DEBUG_KMS(dev, "Load was not detected on output with or %d\n", or); return status; } @@ -118,7 +118,7 @@ nv50_dac_dpms(struct drm_encoder *encoder, int mode) uint32_t val; int or = nv_encoder->or; - NV_DEBUG(dev, "or %d mode %d\n", or, mode); + NV_DEBUG_KMS(dev, "or %d mode %d\n", or, mode); /* wait for it to be done */ if (!nv_wait(NV50_PDISPLAY_DAC_DPMS_CTRL(or), @@ -173,7 +173,7 @@ nv50_dac_mode_fixup(struct drm_encoder *encoder, struct drm_display_mode *mode, struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); struct nouveau_connector *connector; - NV_DEBUG(encoder->dev, "or %d\n", nv_encoder->or); + NV_DEBUG_KMS(encoder->dev, "or %d\n", nv_encoder->or); connector = nouveau_encoder_connector_get(nv_encoder); if (!connector) { @@ -213,7 +213,7 @@ nv50_dac_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode, uint32_t mode_ctl = 0, mode_ctl2 = 0; int ret; - NV_DEBUG(dev, "or %d\n", nv_encoder->or); + NV_DEBUG_KMS(dev, "or %d\n", nv_encoder->or); nv50_dac_dpms(encoder, DRM_MODE_DPMS_ON); @@ -264,7 +264,7 @@ nv50_dac_destroy(struct drm_encoder *encoder) if (!encoder) return; - NV_DEBUG(encoder->dev, "\n"); + NV_DEBUG_KMS(encoder->dev, "\n"); drm_encoder_cleanup(encoder); kfree(nv_encoder); @@ -280,7 +280,7 @@ nv50_dac_create(struct drm_device *dev, struct dcb_entry *entry) struct nouveau_encoder *nv_encoder; struct drm_encoder *encoder; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); NV_INFO(dev, "Detected a DAC output\n"); nv_encoder = kzalloc(sizeof(*nv_encoder), GFP_KERNEL); diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c index 12c5ee63495b..90f0bf59fbcd 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.c +++ b/drivers/gpu/drm/nouveau/nv50_display.c @@ -188,7 +188,7 @@ nv50_display_init(struct drm_device *dev) uint64_t start; int ret, i; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); nv_wr32(dev, 0x00610184, nv_rd32(dev, 0x00614004)); /* @@ -232,7 +232,7 @@ nv50_display_init(struct drm_device *dev) nv_wr32(dev, NV50_PDISPLAY_UNK_380, 0); /* RAM is clamped to 256 MiB. */ ram_amount = nouveau_mem_fb_amount(dev); - NV_DEBUG(dev, "ram_amount %d\n", ram_amount); + NV_DEBUG_KMS(dev, "ram_amount %d\n", ram_amount); if (ram_amount > 256*1024*1024) ram_amount = 256*1024*1024; nv_wr32(dev, NV50_PDISPLAY_RAM_AMOUNT, ram_amount - 1); @@ -398,7 +398,7 @@ static int nv50_display_disable(struct drm_device *dev) struct drm_crtc *drm_crtc; int ret, i; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); list_for_each_entry(drm_crtc, &dev->mode_config.crtc_list, head) { struct nouveau_crtc *crtc = nouveau_crtc(drm_crtc); @@ -469,7 +469,7 @@ int nv50_display_create(struct drm_device *dev) uint32_t connector[16] = {}; int ret, i; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); /* init basic kernel modesetting */ drm_mode_config_init(dev); @@ -573,7 +573,7 @@ int nv50_display_destroy(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); drm_mode_config_cleanup(dev); @@ -617,7 +617,7 @@ nv50_display_irq_head(struct drm_device *dev, int *phead, * CRTC separately, and submission will be blocked by the GPU * until we handle each in turn. */ - NV_DEBUG(dev, "0x610030: 0x%08x\n", unk30); + NV_DEBUG_KMS(dev, "0x610030: 0x%08x\n", unk30); head = ffs((unk30 >> 9) & 3) - 1; if (head < 0) return -EINVAL; @@ -661,7 +661,7 @@ nv50_display_irq_head(struct drm_device *dev, int *phead, or = i; } - NV_DEBUG(dev, "type %d, or %d\n", type, or); + NV_DEBUG_KMS(dev, "type %d, or %d\n", type, or); if (type == OUTPUT_ANY) { NV_ERROR(dev, "unknown encoder!!\n"); return -1; @@ -690,9 +690,21 @@ nv50_display_script_select(struct drm_device *dev, struct dcb_entry *dcbent, int pxclk) { struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nouveau_connector *nv_connector = NULL; + struct drm_encoder *encoder; struct nvbios *bios = &dev_priv->VBIOS; uint32_t mc, script = 0, or; + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); + + if (nv_encoder->dcb != dcbent) + continue; + + nv_connector = nouveau_encoder_connector_get(nv_encoder); + break; + } + or = ffs(dcbent->or) - 1; mc = nv50_display_mode_ctrl(dev, dcbent->type != OUTPUT_ANALOG, or); switch (dcbent->type) { @@ -711,6 +723,11 @@ nv50_display_script_select(struct drm_device *dev, struct dcb_entry *dcbent, } else if (bios->fp.strapless_is_24bit & 1) script |= 0x0200; + + if (nv_connector && nv_connector->edid && + (nv_connector->edid->revision >= 4) && + (nv_connector->edid->input & 0x70) >= 0x20) + script |= 0x0200; } if (nouveau_uscript_lvds >= 0) { @@ -811,7 +828,7 @@ nv50_display_unk20_handler(struct drm_device *dev) pclk = nv_rd32(dev, NV50_PDISPLAY_CRTC_P(head, CLOCK)) & 0x3fffff; script = nv50_display_script_select(dev, dcbent, pclk); - NV_DEBUG(dev, "head %d pxclk: %dKHz\n", head, pclk); + NV_DEBUG_KMS(dev, "head %d pxclk: %dKHz\n", head, pclk); if (dcbent->type != OUTPUT_DP) nouveau_bios_run_display_table(dev, dcbent, 0, -2); @@ -870,7 +887,7 @@ nv50_display_irq_handler_bh(struct work_struct *work) uint32_t intr0 = nv_rd32(dev, NV50_PDISPLAY_INTR_0); uint32_t intr1 = nv_rd32(dev, NV50_PDISPLAY_INTR_1); - NV_DEBUG(dev, "PDISPLAY_INTR_BH 0x%08x 0x%08x\n", intr0, intr1); + NV_DEBUG_KMS(dev, "PDISPLAY_INTR_BH 0x%08x 0x%08x\n", intr0, intr1); if (intr1 & NV50_PDISPLAY_INTR_1_CLK_UNK10) nv50_display_unk10_handler(dev); @@ -974,7 +991,7 @@ nv50_display_irq_handler(struct drm_device *dev) uint32_t intr1 = nv_rd32(dev, NV50_PDISPLAY_INTR_1); uint32_t clock; - NV_DEBUG(dev, "PDISPLAY_INTR 0x%08x 0x%08x\n", intr0, intr1); + NV_DEBUG_KMS(dev, "PDISPLAY_INTR 0x%08x 0x%08x\n", intr0, intr1); if (!intr0 && !(intr1 & ~delayed)) break; diff --git a/drivers/gpu/drm/nouveau/nv50_fbcon.c b/drivers/gpu/drm/nouveau/nv50_fbcon.c index 6bcc6d39e9b0..e4f279ee61cf 100644 --- a/drivers/gpu/drm/nouveau/nv50_fbcon.c +++ b/drivers/gpu/drm/nouveau/nv50_fbcon.c @@ -16,9 +16,7 @@ nv50_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect) if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, rect->rop == ROP_COPY ? 7 : 11)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); } if (info->flags & FBINFO_HWACCEL_DISABLED) { @@ -31,7 +29,11 @@ nv50_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect) OUT_RING(chan, 1); } BEGIN_RING(chan, NvSub2D, 0x0588, 1); - OUT_RING(chan, rect->color); + if (info->fix.visual == FB_VISUAL_TRUECOLOR || + info->fix.visual == FB_VISUAL_DIRECTCOLOR) + OUT_RING(chan, ((uint32_t *)info->pseudo_palette)[rect->color]); + else + OUT_RING(chan, rect->color); BEGIN_RING(chan, NvSub2D, 0x0600, 4); OUT_RING(chan, rect->dx); OUT_RING(chan, rect->dy); @@ -56,9 +58,7 @@ nv50_fbcon_copyarea(struct fb_info *info, const struct fb_copyarea *region) return; if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 12)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); } if (info->flags & FBINFO_HWACCEL_DISABLED) { @@ -101,8 +101,7 @@ nv50_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) } if (!(info->flags & FBINFO_HWACCEL_DISABLED) && RING_SPACE(chan, 11)) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); } if (info->flags & FBINFO_HWACCEL_DISABLED) { @@ -135,9 +134,7 @@ nv50_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) int push = dwords > 2047 ? 2047 : dwords; if (RING_SPACE(chan, push + 1)) { - NV_ERROR(dev, - "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; + nouveau_fbcon_gpu_lockup(info); cfb_imageblit(info, image); return; } @@ -199,7 +196,7 @@ nv50_fbcon_accel_init(struct fb_info *info) ret = RING_SPACE(chan, 59); if (ret) { - NV_ERROR(dev, "GPU lockup - switching to software fbcon\n"); + nouveau_fbcon_gpu_lockup(info); return ret; } diff --git a/drivers/gpu/drm/nouveau/nv50_fifo.c b/drivers/gpu/drm/nouveau/nv50_fifo.c index 77ae1aaa0bce..39caf167587d 100644 --- a/drivers/gpu/drm/nouveau/nv50_fifo.c +++ b/drivers/gpu/drm/nouveau/nv50_fifo.c @@ -384,8 +384,8 @@ nv50_fifo_load_context(struct nouveau_channel *chan) nv_wr32(dev, NV40_PFIFO_CACHE1_DATA(ptr), nv_ro32(dev, cache, (ptr * 2) + 1)); } - nv_wr32(dev, 0x3210, cnt << 2); - nv_wr32(dev, 0x3270, 0); + nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, cnt << 2); + nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0); /* guessing that all the 0x34xx regs aren't on NV50 */ if (!IS_G80) { @@ -398,8 +398,6 @@ nv50_fifo_load_context(struct nouveau_channel *chan) dev_priv->engine.instmem.finish_access(dev); - nv_wr32(dev, NV03_PFIFO_CACHE1_GET, 0); - nv_wr32(dev, NV03_PFIFO_CACHE1_PUT, 0); nv_wr32(dev, NV03_PFIFO_CACHE1_PUSH1, chan->id | (1<<16)); return 0; } @@ -416,7 +414,7 @@ nv50_fifo_unload_context(struct drm_device *dev) NV_DEBUG(dev, "\n"); chid = pfifo->channel_id(dev); - if (chid < 0 || chid >= dev_priv->engine.fifo.channels) + if (chid < 1 || chid >= dev_priv->engine.fifo.channels - 1) return 0; chan = dev_priv->fifos[chid]; diff --git a/drivers/gpu/drm/nouveau/nv50_graph.c b/drivers/gpu/drm/nouveau/nv50_graph.c index 177d8229336f..ca79f32be44c 100644 --- a/drivers/gpu/drm/nouveau/nv50_graph.c +++ b/drivers/gpu/drm/nouveau/nv50_graph.c @@ -107,9 +107,13 @@ nv50_graph_init_regs(struct drm_device *dev) static int nv50_graph_init_ctxctl(struct drm_device *dev) { + struct drm_nouveau_private *dev_priv = dev->dev_private; + NV_DEBUG(dev, "\n"); - nv40_grctx_init(dev); + nouveau_grctx_prog_load(dev); + if (!dev_priv->engine.graph.ctxprog) + dev_priv->engine.graph.accel_blocked = true; nv_wr32(dev, 0x400320, 4); nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0); @@ -140,7 +144,7 @@ void nv50_graph_takedown(struct drm_device *dev) { NV_DEBUG(dev, "\n"); - nv40_grctx_fini(dev); + nouveau_grctx_fini(dev); } void @@ -207,7 +211,7 @@ nv50_graph_create_context(struct nouveau_channel *chan) dev_priv->engine.instmem.finish_access(dev); dev_priv->engine.instmem.prepare_access(dev, true); - nv40_grctx_vals_load(dev, ctx); + nouveau_grctx_vals_load(dev, ctx); nv_wo32(dev, ctx, 0x00000/4, chan->ramin->instance >> 12); if ((dev_priv->chipset & 0xf0) == 0xa0) nv_wo32(dev, ctx, 0x00004/4, 0x00000000); diff --git a/drivers/gpu/drm/nouveau/nv50_sor.c b/drivers/gpu/drm/nouveau/nv50_sor.c index 8c280463a664..e395c16d30f5 100644 --- a/drivers/gpu/drm/nouveau/nv50_sor.c +++ b/drivers/gpu/drm/nouveau/nv50_sor.c @@ -44,7 +44,7 @@ nv50_sor_disconnect(struct nouveau_encoder *nv_encoder) struct nouveau_channel *evo = dev_priv->evo; int ret; - NV_DEBUG(dev, "Disconnecting SOR %d\n", nv_encoder->or); + NV_DEBUG_KMS(dev, "Disconnecting SOR %d\n", nv_encoder->or); ret = RING_SPACE(evo, 2); if (ret) { @@ -70,7 +70,7 @@ nv50_sor_dp_link_train(struct drm_encoder *encoder) } if (dpe->script0) { - NV_DEBUG(dev, "SOR-%d: running DP script 0\n", nv_encoder->or); + NV_DEBUG_KMS(dev, "SOR-%d: running DP script 0\n", nv_encoder->or); nouveau_bios_run_init_table(dev, le16_to_cpu(dpe->script0), nv_encoder->dcb); } @@ -79,7 +79,7 @@ nv50_sor_dp_link_train(struct drm_encoder *encoder) NV_ERROR(dev, "SOR-%d: link training failed\n", nv_encoder->or); if (dpe->script1) { - NV_DEBUG(dev, "SOR-%d: running DP script 1\n", nv_encoder->or); + NV_DEBUG_KMS(dev, "SOR-%d: running DP script 1\n", nv_encoder->or); nouveau_bios_run_init_table(dev, le16_to_cpu(dpe->script1), nv_encoder->dcb); } @@ -93,7 +93,7 @@ nv50_sor_dpms(struct drm_encoder *encoder, int mode) uint32_t val; int or = nv_encoder->or; - NV_DEBUG(dev, "or %d mode %d\n", or, mode); + NV_DEBUG_KMS(dev, "or %d mode %d\n", or, mode); /* wait for it to be done */ if (!nv_wait(NV50_PDISPLAY_SOR_DPMS_CTRL(or), @@ -142,7 +142,7 @@ nv50_sor_mode_fixup(struct drm_encoder *encoder, struct drm_display_mode *mode, struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); struct nouveau_connector *connector; - NV_DEBUG(encoder->dev, "or %d\n", nv_encoder->or); + NV_DEBUG_KMS(encoder->dev, "or %d\n", nv_encoder->or); connector = nouveau_encoder_connector_get(nv_encoder); if (!connector) { @@ -182,7 +182,7 @@ nv50_sor_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode, uint32_t mode_ctl = 0; int ret; - NV_DEBUG(dev, "or %d\n", nv_encoder->or); + NV_DEBUG_KMS(dev, "or %d\n", nv_encoder->or); nv50_sor_dpms(encoder, DRM_MODE_DPMS_ON); @@ -246,7 +246,7 @@ nv50_sor_destroy(struct drm_encoder *encoder) if (!encoder) return; - NV_DEBUG(encoder->dev, "\n"); + NV_DEBUG_KMS(encoder->dev, "\n"); drm_encoder_cleanup(encoder); @@ -265,7 +265,7 @@ nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry) bool dum; int type; - NV_DEBUG(dev, "\n"); + NV_DEBUG_KMS(dev, "\n"); switch (entry->type) { case OUTPUT_TMDS: |