* drivers/gpu/drm/nouveau/nv50_graph.c: Adjusted, was already deblobbed. From a4cbb7f1379aa9817c85840c6d079dd222641dbd Mon Sep 17 00:00:00 2001 From: Marcin Slusarz Date: Wed, 17 Feb 2010 19:04:00 +0100 Subject: [PATCH] drm-nouveau-updates MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm/nouveau: fix pramdac_table range checking get_tmds_index_reg reads some value from stack when mlv happens to be equal to size of pramdac_table array. Fix it. Reported-by: Dan Carpenter Signed-off-by: Marcin Slusarz Signed-off-by: Francisco Jerez drm/nouveau: fix nouveau_i2c_find bounds checking Reported-by: Dan Carpenter Signed-off-by: Marcin Slusarz Signed-off-by: Francisco Jerez drm/nouveau: fix i2ctable bounds checking i2c_entries seems to be the number of i2c entries, so with index equal to this number, we could read invalid data from i2ctable. Fix it. Signed-off-by: Marcin Slusarz Signed-off-by: Francisco Jerez drm/nouveau: allow retrieval of vbios image from debugfs It's very useful to be able to access this without additional tools for debugging purposes. Signed-off-by: Ben Skeggs drm/nouveau: rename parsed_dcb_gpio to dcb_gpio_table Signed-off-by: Ben Skeggs drm/nouveau: merge parsed_dcb and bios_parsed_dcb into dcb_table Signed-off-by: Ben Skeggs drm/nouveau: merge nvbios and nouveau_bios_info Signed-off-by: Ben Skeggs drm/nouveau: reorganise bios header, add dcb connector type enums Signed-off-by: Ben Skeggs drm/nouveau: parse dcb gpio/connector tables after encoders Signed-off-by: Ben Skeggs drm/nouveau: check for known dcb connector types Signed-off-by: Ben Skeggs drm/nouveau: construct a connector table for cards that lack a real one Signed-off-by: Ben Skeggs drm/nouveau: use dcb connector table for creating drm connectors This makes this code common to both the nv04 and nv50 paths. For the moment, we keep the previous behaviour with HDMI/eDP connectors and report them as DVI-D/DP instead. This will be fixed once the rest of the code has been fixed to deal with those types. Signed-off-by: Ben Skeggs drm/nv50: enable hpd on any connector we know the gpio line for Signed-off-by: Ben Skeggs drm/nouveau: use dcb connector types throughout the driver Signed-off-by: Ben Skeggs drm/nv50: Implement ctxprog/state generation. This removes dependence on external firmware for NV50 generation cards. If the generated ctxprogs don't work for you for some reason, please report it. Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nouveau: Fix noaccel/nofbaccel option descriptions. Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nouveau: support version 0x20 displayport tables Not entirely identical to 0x21, the per-encoder table header lacks the third init table pointer. However, our current parsing of the table should work just fine. Signed-off-by: Ben Skeggs drm/nouveau: report unknown connector state if lid closed This is in preference to disconnected. If there's no other outputs connected this will cause LVDS to be programmed even with the lid closed rather than having X fail to start because of no available outputs. Signed-off-by: Ben Skeggs drm/nouveau: use ALIGN instead of open coding it CC: Ben Skeggs Signed-off-by: Matt Turner Signed-off-by: Ben Skeggs drm/nouveau: protect channel create/destroy and irq handler with a spinlock The nv50 pgraph handler (for example) could reenable pgraph fifo access and that would be bad when pgraph context is being unloaded (we need the guarantee a ctxprog isn't running). Signed-off-by: Maarten Maathuis Signed-off-by: Ben Skeggs drm/nv50: Remove redundant/incorrect ctxvals initialisation. 11c/004 offset corresponds to PGRAPH reg 0x400828, and is initialised earlier anyway by both our ctxprog generator and blob ctxvals. It's actually incorrect with the generator, since we use different layout on pre-NVA0. Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nouveau: Fix fbcon corruption with font width not divisible by 8 NV50 is nice and has a switch that autoaligns stuff for us. Pre-NV50, we need to align input bitmap width manually. Signed-off-by: Marcin Kościelnicki Signed-off-by: Francisco Jerez drm/nv50: Make ctxprog wait until interrupt handler is done. This will fix races between generated ctxprogs and interrupt handler. Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nv50: Improve PGRAPH interrupt handling. This makes nouveau recognise and report more kinds of PGRAPH errors, as well as prevent GPU lockups resulting from some of them. Lots of guesswork was involved and some part of this is probably incorrect. Some potential-lockuop situations are handled by just resetting a whole PGRAPH subunit, which doesn't sound like a "proper" solution, but seems to work just fine... for now. Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nouveau: add option to allow override of dcb connector table types Signed-off-by: Ben Skeggs drm/nouveau: Gigabyte NX85T connector table lies, it has DVI-I not HDMI Signed-off-by: Ben Skeggs drm/nv04-nv40: Fix up the programmed horizontal sync pulse delay. The calculated values were a little bit off (~16 clocks), the only effect it could have had is a slightly offset image with respect to the blob on analog outputs (bug 26790). Signed-off-by: Francisco Jerez drm/nouveau: print a message very early during suspend - In case of suspend lockups it's nice to know it happened in nouveau. Signed-off-by: Maarten Maathuis drm/nv50: add a memory barrier to pushbuf submission - This is useful for vram pushbuffers that are write combined. - pre-nv50 has one too (in WRITE_PUT). Signed-off-by: Maarten Maathuis drm/nv50: fix connector table parsing for some cards The connector table index in the DCB entry for each output type is an index into the connector table, and does *not* necessarily match up with what was previously called "index" in the connector table entries themselves. Not real sure what that index is exactly, renamed to "index2" as we still use it to prevent creating multiple TV connectors. Signed-off-by: Ben Skeggs drm/nouveau: Never evict VRAM buffers to system. VRAM->system is a synchronous operation: it involves scheduling a VRAM->TT DMA transfer and stalling the CPU until it's finished so that we can unbind the new memory from the translation tables. VRAM->TT can always be performed asynchronously, even if TT is already full and we have to move something out of it. Additionally, allowing VRAM->system behaves badly under heavy memory pressure because once we run out of TT, stuff starts to be moved back and forth between VRAM and system, and the TT contents are hardly renewed. Signed-off-by: Francisco Jerez drm/nouveau: add module option to disable TV detection Intended to be used as a workaround in cases where we falsely detect that a TV is connected when it's not. Signed-off-by: Ben Skeggs drm/nv50: add more 0x100c80 flushy magic Fixes the !vbo_fifo path in the 3D driver on certain chipsets. Still not really any good idea of what exactly the magic achieves, but it makes things work. While we're at it, in the PCIEGART path, flush on unbinding also. Signed-off-by: Ben Skeggs drm/nouveau: bail out of auxch transaction if we repeatedly recieve defers There's one known case where we never stop recieving DEFER, and loop here forever. Lets not do that.. Signed-off-by: Ben Skeggs drm/nv50: fix fbcon when framebuffer above 4GiB mark This can't actually happen right now, but lets fix it anyway. Signed-off-by: Ben Skeggs drm/nv50: Fix NEWCTX_DONE flag number Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nouveau: remove some unused members from drm_nouveau_private Signed-off-by: Ben Skeggs drm/nouveau: detect vram amount once, and save the value As opposed to repeatedly reading the amount back from the GPU every time we need to know the VRAM size. We should now fail to load gracefully on detecting no VRAM, rather than something potentially messy happening. Signed-off-by: Ben Skeggs drm/nv40: rework lvds table parsing All indications seem to be that the version 0x30 table should be handled the same way as 0x40 (as used on G80), at least for the parts that we currently try use. This commit cleans up the parsing to make it clearer about what we're actually trying to achieve, and unifies the 0x30/0x40 parsing. Signed-off-by: Ben Skeggs drm/nv40: add LVDS table quirk for Dell Latitude D620 Should fix: https://bugzilla.redhat.com/show_bug.cgi?id=505132 https://bugzilla.redhat.com/show_bug.cgi?id=543091 https://bugzilla.redhat.com/show_bug.cgi?id=530425 https://bugs.edge.launchpad.net/ubuntu/+source/xserver-xorg-video-nouveau/ +bug/539730 Signed-off-by: Ben Skeggs drm/nv50: fix instmem init on IGPs if stolen mem crosses 4GiB mark Signed-off-by: Ben Skeggs drm/nouveau: fixup the init failure paths some more Signed-off-by: Ben Skeggs drm/nv50: move pdisp init earlier, and cleanup if it fails Moving it earlier is to avoid some extra cleanup if it fails. Signed-off-by: Ben Skeggs drm/nv50: Allow using the NVA3 new compute class. Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nouveau: Make use of TTM busy_placements. Previously we were filling it the same as "placements", but in some cases there're valid alternatives that we were ignoring completely. Keeping a back-up memory type helps on several low-mem situations. Signed-off-by: Francisco Jerez drm/nv50: preserve an unknown SOR_MODECTRL value for DP encoders This value interacts with some registers we don't currently know how to program properly ourselves. The default of 5 that we were using matches what the VBIOS on early DP cards do, but later ones use 6, which would cause nouveau to program an incorrect mode on these chips. Signed-off-by: Ben Skeggs drm/nv50: punt hotplug irq handling out to workqueue On DP outputs we'll likely end up running vbios init tables here, which may sleep. Signed-off-by: Ben Skeggs drm/nv50: partially revert ec99dbe438787d62ecde3a22f8ce3f880a4f4e14 The commit mentioned above breaks the DP SOR_MODE_CTRL detection as once nv50_display_init() has been called all the MODE_CTRL registers are reset. This wasn't noticed when initially writing the DP SOR_MODE_CTRL patch as it was done on another machine, without ec99dbe..4e14 applied. This commit moves the nv50_display_init() call to back where it was, after the KMS setup, and adds the additional cleanup needed. Signed-off-by: Ben Skeggs drm/nv50: another dodgy DP hack Allows *some* DP cards to keep working in some corner cases that most people shouldn't hit. I hit it all the time with development, so this can stay for now. Signed-off-by: Ben Skeggs drm/nv50: Add NVA3 support in ctxprog/ctxvals generator. Signed-off-by: Marcin Kościelnicki Signed-off-by: Ben Skeggs drm/nv40: Init some tiling-related PGRAPH state. Fixes garbled 3D on an nv46 card. Reported-by: Francesco Marella Signed-off-by: Francisco Jerez drm/nouveau: store raw gpio table entry in bios gpio structs And use our own version of the GPIO table for the INIT_GPIO opcode. Signed-off-by: Ben Skeggs drm/nv50: parse/use some more de-magiced parts of gpio table entries Signed-off-by: Ben Skeggs drm/nv50: implement gpio set/get routines Signed-off-by: Ben Skeggs Revert "drm/nouveau: report unknown connector state if lid closed" Included in upstream stable point-release. This reverts commit b30083bdb990bcc2829fce83d871a86059ff4fc1. drm/nouveau: fix a nouveau_bo dereference after it's been destroyed Signed-off-by: Ben Skeggs drm/nouveau: bios parser fixes for eDP boards Signed-off-by: Ben Skeggs drm/nouveau: dump pll limits entries when debugging is on Signed-off-by: Ben Skeggs drm/nv50: output calculated crtc pll when debugging on Signed-off-by: Ben Skeggs drm/nv50: fix suspend/resume with DP outputs Signed-off-by: Ben Skeggs drm/nv50: store full dcb i2c entry from vbios Signed-off-by: Ben Skeggs drm/nv50: fix monitor detection on certain chipsets There appears to be some kind of switch on certain chips to control whether the DP auxch or traditional i2c bus will be operational on a connector, this commit hopefully fixes nouveau to do the right thing. Likely only relevent on chips with DP outputs. Signed-off-by: Ben Skeggs drm/nv50: send hotplug event to userspace Signed-off-by: Ben Skeggs drm/nv50: support fractional feedback divider on newer chips Signed-off-by: Ben Skeggs drm/nouveau: don't execute INIT_GPIO unless we're really running the table This resulted in accidently switching off the eDP panel on certain laptops since the default state in the GPIO table was off. Fixes rh#582621 Signed-off-by: Ben Skeggs drm/nv50: fix iommu errors caused by device reading from address 0 Signed-off-by: Ben Skeggs --- drivers/gpu/drm/nouveau/Makefile | 7 +- drivers/gpu/drm/nouveau/nouveau_bios.c | 672 +++++--- drivers/gpu/drm/nouveau/nouveau_bios.h | 132 +- drivers/gpu/drm/nouveau/nouveau_bo.c | 68 +- drivers/gpu/drm/nouveau/nouveau_calc.c | 4 +- drivers/gpu/drm/nouveau/nouveau_channel.c | 15 +- drivers/gpu/drm/nouveau/nouveau_connector.c | 163 ++- drivers/gpu/drm/nouveau/nouveau_connector.h | 3 +- drivers/gpu/drm/nouveau/nouveau_debugfs.c | 18 +- drivers/gpu/drm/nouveau/nouveau_dma.c | 5 + drivers/gpu/drm/nouveau/nouveau_dp.c | 8 +- drivers/gpu/drm/nouveau/nouveau_drv.c | 14 +- drivers/gpu/drm/nouveau/nouveau_drv.h | 59 +- drivers/gpu/drm/nouveau/nouveau_encoder.h | 3 + drivers/gpu/drm/nouveau/nouveau_gem.c | 55 +- drivers/gpu/drm/nouveau/nouveau_hw.c | 6 +- drivers/gpu/drm/nouveau/nouveau_i2c.c | 23 +- drivers/gpu/drm/nouveau/nouveau_irq.c | 615 +++++++- drivers/gpu/drm/nouveau/nouveau_mem.c | 124 +- drivers/gpu/drm/nouveau/nouveau_reg.h | 1 + drivers/gpu/drm/nouveau/nouveau_sgdma.c | 18 + drivers/gpu/drm/nouveau/nouveau_state.c | 20 +- drivers/gpu/drm/nouveau/nv04_crtc.c | 6 +- drivers/gpu/drm/nouveau/nv04_dac.c | 8 +- drivers/gpu/drm/nouveau/nv04_dfp.c | 4 +- drivers/gpu/drm/nouveau/nv04_display.c | 49 +- drivers/gpu/drm/nouveau/nv04_fbcon.c | 6 +- drivers/gpu/drm/nouveau/nv04_fifo.c | 5 + drivers/gpu/drm/nouveau/nv04_tv.c | 2 +- drivers/gpu/drm/nouveau/nv17_tv.c | 6 +- drivers/gpu/drm/nouveau/nv40_fifo.c | 7 +- drivers/gpu/drm/nouveau/nv40_graph.c | 21 + drivers/gpu/drm/nouveau/nv50_calc.c | 88 + drivers/gpu/drm/nouveau/nv50_crtc.c | 46 +- drivers/gpu/drm/nouveau/nv50_dac.c | 4 +- drivers/gpu/drm/nouveau/nv50_display.c | 111 +- drivers/gpu/drm/nouveau/nv50_display.h | 1 + drivers/gpu/drm/nouveau/nv50_fb.c | 38 + drivers/gpu/drm/nouveau/nv50_fbcon.c | 17 +- drivers/gpu/drm/nouveau/nv50_fifo.c | 5 + drivers/gpu/drm/nouveau/nv50_gpio.c | 76 + drivers/gpu/drm/nouveau/nv50_graph.c | 103 +- drivers/gpu/drm/nouveau/nv50_grctx.c | 2383 +++++++++++++++++++++++++++ drivers/gpu/drm/nouveau/nv50_instmem.c | 18 +- drivers/gpu/drm/nouveau/nv50_sor.c | 30 +- 45 files changed, 4304 insertions(+), 763 deletions(-) create mode 100644 drivers/gpu/drm/nouveau/nv50_calc.c create mode 100644 drivers/gpu/drm/nouveau/nv50_fb.c create mode 100644 drivers/gpu/drm/nouveau/nv50_gpio.c create mode 100644 drivers/gpu/drm/nouveau/nv50_grctx.c diff --git a/drivers/gpu/drm/nouveau/Makefile b/drivers/gpu/drm/nouveau/Makefile index 48c290b..acd31ed 100644 --- a/drivers/gpu/drm/nouveau/Makefile +++ b/drivers/gpu/drm/nouveau/Makefile @@ -12,17 +12,18 @@ nouveau-y := nouveau_drv.o nouveau_state.o nouveau_channel.o nouveau_mem.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_fb.o nv10_fb.o nv40_fb.o nv50_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 \ + nv40_grctx.o nv50_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 \ nv04_dac.o nv04_dfp.o nv04_tv.o nv17_tv.o nv17_tv_modes.o \ nv04_crtc.o nv04_display.o nv04_cursor.o nv04_fbcon.o \ - nv17_gpio.o + nv17_gpio.o nv50_gpio.o \ + nv50_calc.o nouveau-$(CONFIG_DRM_NOUVEAU_DEBUG) += nouveau_debugfs.o nouveau-$(CONFIG_COMPAT) += nouveau_ioc32.o diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.c b/drivers/gpu/drm/nouveau/nouveau_bios.c index 0e9cd1d..5472c76 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bios.c +++ b/drivers/gpu/drm/nouveau/nouveau_bios.c @@ -26,6 +26,7 @@ #define NV_DEBUG_NOTRACE #include "nouveau_drv.h" #include "nouveau_hw.h" +#include "nouveau_encoder.h" /* these defines are made up */ #define NV_CIO_CRE_44_HEADA 0x0 @@ -311,11 +312,11 @@ valid_reg(struct nvbios *bios, uint32_t reg) /* C51 has misaligned regs on purpose. Marvellous */ if (reg & 0x2 || - (reg & 0x1 && dev_priv->VBIOS.pub.chip_version != 0x51)) + (reg & 0x1 && dev_priv->vbios.chip_version != 0x51)) NV_ERROR(dev, "======= misaligned reg 0x%08X =======\n", reg); /* warn on C51 regs that haven't been verified accessible in tracing */ - if (reg & 0x1 && dev_priv->VBIOS.pub.chip_version == 0x51 && + if (reg & 0x1 && dev_priv->vbios.chip_version == 0x51 && reg != 0x130d && reg != 0x1311 && reg != 0x60081d) NV_WARN(dev, "=== C51 misaligned reg 0x%08X not verified ===\n", reg); @@ -420,7 +421,7 @@ bios_wr32(struct nvbios *bios, uint32_t reg, uint32_t data) LOG_OLD_VALUE(bios_rd32(bios, reg)); BIOSLOG(bios, " Write: Reg: 0x%08X, Data: 0x%08X\n", reg, data); - if (dev_priv->VBIOS.execute) { + if (dev_priv->vbios.execute) { still_alive(); nv_wr32(bios->dev, reg, data); } @@ -647,7 +648,7 @@ nv50_pll_set(struct drm_device *dev, uint32_t reg, uint32_t clk) reg0 = (reg0 & 0xfff8ffff) | (pll.log2P << 16); reg1 = (reg1 & 0xffff0000) | (pll.N1 << 8) | pll.M1; - if (dev_priv->VBIOS.execute) { + if (dev_priv->vbios.execute) { still_alive(); nv_wr32(dev, reg + 4, reg1); nv_wr32(dev, reg + 0, reg0); @@ -689,7 +690,7 @@ setPLL(struct nvbios *bios, uint32_t reg, uint32_t clk) static int dcb_entry_idx_from_crtchead(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; /* * For the results of this function to be correct, CR44 must have been @@ -700,7 +701,7 @@ static int dcb_entry_idx_from_crtchead(struct drm_device *dev) uint8_t dcb_entry = NVReadVgaCrtc5758(dev, bios->state.crtchead, 0); - if (dcb_entry > bios->bdcb.dcb.entries) { + if (dcb_entry > bios->dcb.entries) { NV_ERROR(dev, "CR58 doesn't have a valid DCB entry currently " "(%02X)\n", dcb_entry); dcb_entry = 0x7f; /* unused / invalid marker */ @@ -713,25 +714,26 @@ static struct nouveau_i2c_chan * init_i2c_device_find(struct drm_device *dev, int i2c_index) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct bios_parsed_dcb *bdcb = &dev_priv->VBIOS.bdcb; + struct dcb_table *dcb = &dev_priv->vbios.dcb; if (i2c_index == 0xff) { /* note: dcb_entry_idx_from_crtchead needs pre-script set-up */ int idx = dcb_entry_idx_from_crtchead(dev), shift = 0; - int default_indices = bdcb->i2c_default_indices; + int default_indices = dcb->i2c_default_indices; - if (idx != 0x7f && bdcb->dcb.entry[idx].i2c_upper_default) + if (idx != 0x7f && dcb->entry[idx].i2c_upper_default) shift = 4; i2c_index = (default_indices >> shift) & 0xf; } if (i2c_index == 0x80) /* g80+ */ - i2c_index = bdcb->i2c_default_indices & 0xf; + i2c_index = dcb->i2c_default_indices & 0xf; return nouveau_i2c_find(dev, i2c_index); } -static uint32_t get_tmds_index_reg(struct drm_device *dev, uint8_t mlv) +static uint32_t +get_tmds_index_reg(struct drm_device *dev, uint8_t mlv) { /* * For mlv < 0x80, it is an index into a table of TMDS base addresses. @@ -744,6 +746,7 @@ static uint32_t get_tmds_index_reg(struct drm_device *dev, uint8_t mlv) */ struct drm_nouveau_private *dev_priv = dev->dev_private; + struct nvbios *bios = &dev_priv->vbios; const int pramdac_offset[13] = { 0, 0, 0x8, 0, 0x2000, 0, 0, 0, 0x2008, 0, 0, 0, 0x2000 }; const uint32_t pramdac_table[4] = { @@ -756,13 +759,12 @@ static uint32_t get_tmds_index_reg(struct drm_device *dev, uint8_t mlv) dcb_entry = dcb_entry_idx_from_crtchead(dev); if (dcb_entry == 0x7f) return 0; - dacoffset = pramdac_offset[ - dev_priv->VBIOS.bdcb.dcb.entry[dcb_entry].or]; + dacoffset = pramdac_offset[bios->dcb.entry[dcb_entry].or]; if (mlv == 0x81) dacoffset ^= 8; return 0x6808b0 + dacoffset; } else { - if (mlv > ARRAY_SIZE(pramdac_table)) { + if (mlv >= ARRAY_SIZE(pramdac_table)) { NV_ERROR(dev, "Magic Lookup Value too big (%02X)\n", mlv); return 0; @@ -1066,6 +1068,126 @@ init_io_flag_condition(struct nvbios *bios, uint16_t offset, } static int +init_dp_condition(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) +{ + /* + * INIT_DP_CONDITION opcode: 0x3A ('') + * + * offset (8 bit): opcode + * offset + 1 (8 bit): "sub" opcode + * offset + 2 (8 bit): unknown + * + */ + + struct bit_displayport_encoder_table *dpe = NULL; + struct dcb_entry *dcb = bios->display.output; + struct drm_device *dev = bios->dev; + uint8_t cond = bios->data[offset + 1]; + int dummy; + + BIOSLOG(bios, "0x%04X: subop 0x%02X\n", offset, cond); + + if (!iexec->execute) + return 3; + + dpe = nouveau_bios_dp_table(dev, dcb, &dummy); + if (!dpe) { + NV_ERROR(dev, "0x%04X: INIT_3A: no encoder table!!\n", offset); + return -EINVAL; + } + + switch (cond) { + case 0: + { + struct dcb_connector_table_entry *ent = + &bios->dcb.connector.entry[dcb->connector]; + + if (ent->type != DCB_CONNECTOR_eDP) + iexec->execute = false; + } + break; + case 1: + case 2: + if (!(dpe->unknown & cond)) + iexec->execute = false; + break; + case 5: + { + struct nouveau_i2c_chan *auxch; + int ret; + + auxch = nouveau_i2c_find(dev, bios->display.output->i2c_index); + if (!auxch) + return -ENODEV; + + ret = nouveau_dp_auxch(auxch, 9, 0xd, &cond, 1); + if (ret) + return ret; + + if (cond & 1) + iexec->execute = false; + } + break; + default: + NV_WARN(dev, "0x%04X: unknown INIT_3A op: %d\n", offset, cond); + break; + } + + if (iexec->execute) + BIOSLOG(bios, "0x%04X: continuing to execute\n", offset); + else + BIOSLOG(bios, "0x%04X: skipping following commands\n", offset); + + return 3; +} + +static int +init_op_3b(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) +{ + /* + * INIT_3B opcode: 0x3B ('') + * + * offset (8 bit): opcode + * offset + 1 (8 bit): crtc index + * + */ + + uint8_t or = ffs(bios->display.output->or) - 1; + uint8_t index = bios->data[offset + 1]; + uint8_t data; + + if (!iexec->execute) + return 2; + + data = bios_idxprt_rd(bios, 0x3d4, index); + bios_idxprt_wr(bios, 0x3d4, index, data & ~(1 << or)); + return 2; +} + +static int +init_op_3c(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) +{ + /* + * INIT_3C opcode: 0x3C ('') + * + * offset (8 bit): opcode + * offset + 1 (8 bit): crtc index + * + */ + + uint8_t or = ffs(bios->display.output->or) - 1; + uint8_t index = bios->data[offset + 1]; + uint8_t data; + + if (!iexec->execute) + return 2; + + data = bios_idxprt_rd(bios, 0x3d4, index); + bios_idxprt_wr(bios, 0x3d4, index, data | (1 << or)); + return 2; +} + +static int init_idx_addr_latched(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) { @@ -2572,48 +2694,37 @@ init_gpio(struct nvbios *bios, uint16_t offset, struct init_exec *iexec) * each GPIO according to various values listed in each entry */ - const uint32_t nv50_gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 }; + struct drm_nouveau_private *dev_priv = bios->dev->dev_private; const uint32_t nv50_gpio_ctl[2] = { 0xe100, 0xe28c }; - const uint8_t *gpio_table = &bios->data[bios->bdcb.gpio_table_ptr]; - 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 0; - } - - if (!bios->bdcb.gpio_table_ptr) { - NV_WARN(bios->dev, "Invalid pointer to INIT_8E table\n"); - return 0; + if (dev_priv->card_type != NV_50) { + NV_ERROR(bios->dev, "INIT_GPIO on unsupported chipset\n"); + return -ENODEV; } - gpio_entry = gpio_table + gpio_table[1]; - for (i = 0; i < gpio_table[2]; i++, gpio_entry += gpio_table[3]) { - uint32_t entry = ROM32(gpio_entry[0]), r, s, v; - int line = (entry & 0x0000001f); + if (!iexec->execute) + return 1; - BIOSLOG(bios, "0x%04X: Entry: 0x%08X\n", offset, entry); + for (i = 0; i < bios->dcb.gpio.entries; i++) { + struct dcb_gpio_entry *gpio = &bios->dcb.gpio.entry[i]; + uint32_t r, s, v; - if ((entry & 0x0000ff00) == 0x0000ff00) - continue; + BIOSLOG(bios, "0x%04X: Entry: 0x%08X\n", offset, gpio->entry); - r = nv50_gpio_reg[line >> 3]; - s = (line & 0x07) << 2; - v = bios_rd32(bios, r) & ~(0x00000003 << s); - if (entry & 0x01000000) - v |= (((entry & 0x60000000) >> 29) ^ 2) << s; - else - v |= (((entry & 0x18000000) >> 27) ^ 2) << s; - bios_wr32(bios, r, v); + BIOSLOG(bios, "0x%04X: set gpio 0x%02x, state %d\n", + offset, gpio->tag, gpio->state_default); + if (bios->execute) + nv50_gpio_set(bios->dev, gpio->tag, gpio->state_default); - r = nv50_gpio_ctl[line >> 4]; - s = (line & 0x0f); + /* The NVIDIA binary driver doesn't appear to actually do + * any of this, my VBIOS does however. + */ + /* Not a clue, needs de-magicing */ + r = nv50_gpio_ctl[gpio->line >> 4]; + s = (gpio->line & 0x0f); v = bios_rd32(bios, r) & ~(0x00010001 << s); - switch ((entry & 0x06000000) >> 25) { + switch ((gpio->entry & 0x06000000) >> 25) { case 1: v |= (0x00000001 << s); break; @@ -2947,6 +3058,9 @@ static struct init_tbl_entry itbl_entry[] = { { "INIT_COPY" , 0x37, init_copy }, { "INIT_NOT" , 0x38, init_not }, { "INIT_IO_FLAG_CONDITION" , 0x39, init_io_flag_condition }, + { "INIT_DP_CONDITION" , 0x3A, init_dp_condition }, + { "INIT_OP_3B" , 0x3B, init_op_3b }, + { "INIT_OP_3C" , 0x3C, init_op_3c }, { "INIT_INDEX_ADDRESS_LATCHED" , 0x49, init_idx_addr_latched }, { "INIT_IO_RESTRICT_PLL2" , 0x4A, init_io_restrict_pll2 }, { "INIT_PLL2" , 0x4B, init_pll2 }, @@ -3123,7 +3237,7 @@ run_digital_op_script(struct drm_device *dev, uint16_t scriptptr, struct dcb_entry *dcbent, int head, bool dl) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; struct init_exec iexec = {true, false}; NV_TRACE(dev, "0x%04X: Parsing digital output script table\n", @@ -3140,7 +3254,7 @@ run_digital_op_script(struct drm_device *dev, uint16_t scriptptr, static int call_lvds_manufacturer_script(struct drm_device *dev, struct dcb_entry *dcbent, int head, enum LVDS_script script) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; uint8_t sub = bios->data[bios->fp.xlated_entry + script] + (bios->fp.link_c_increment && dcbent->or & OUTPUT_C ? 1 : 0); uint16_t scriptofs = ROM16(bios->data[bios->init_script_tbls_ptr + sub * 2]); @@ -3194,10 +3308,9 @@ static int run_lvds_table(struct drm_device *dev, struct dcb_entry *dcbent, int * of a list of pxclks and script pointers. */ struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; unsigned int outputset = (dcbent->or == 4) ? 1 : 0; uint16_t scriptptr = 0, clktable; - uint8_t clktableptr = 0; /* * For now we assume version 3.0 table - g80 support will need some @@ -3216,26 +3329,29 @@ static int run_lvds_table(struct drm_device *dev, struct dcb_entry *dcbent, int scriptptr = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 11 + outputset * 2]); break; case LVDS_RESET: + clktable = bios->fp.lvdsmanufacturerpointer + 15; + if (dcbent->or == 4) + clktable += 8; + if (dcbent->lvdsconf.use_straps_for_mode) { if (bios->fp.dual_link) - clktableptr += 2; - if (bios->fp.BITbit1) - clktableptr++; + clktable += 4; + if (bios->fp.if_is_24bit) + clktable += 2; } else { /* using EDID */ - uint8_t fallback = bios->data[bios->fp.lvdsmanufacturerpointer + 4]; - int fallbackcmpval = (dcbent->or == 4) ? 4 : 1; + int cmpval_24bit = (dcbent->or == 4) ? 4 : 1; if (bios->fp.dual_link) { - clktableptr += 2; - fallbackcmpval *= 2; + clktable += 4; + cmpval_24bit <<= 1; } - if (fallbackcmpval & fallback) - clktableptr++; + + if (bios->fp.strapless_is_24bit & cmpval_24bit) + clktable += 2; } - /* adding outputset * 8 may not be correct */ - clktable = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 15 + clktableptr * 2 + outputset * 8]); + clktable = ROM16(bios->data[clktable]); if (!clktable) { NV_ERROR(dev, "Pixel clock comparison table not found\n"); return -ENOENT; @@ -3261,7 +3377,7 @@ int call_lvds_script(struct drm_device *dev, struct dcb_entry *dcbent, int head, */ struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; uint8_t lvds_ver = bios->data[bios->fp.lvdsmanufacturerpointer]; uint32_t sel_clk_binding, sel_clk; int ret; @@ -3395,7 +3511,7 @@ static int parse_fp_mode_table(struct drm_device *dev, struct nvbios *bios) #ifndef __powerpc__ NV_ERROR(dev, "Pointer to flat panel table invalid\n"); #endif - bios->pub.digital_min_front_porch = 0x4b; + bios->digital_min_front_porch = 0x4b; return 0; } @@ -3428,7 +3544,7 @@ static int parse_fp_mode_table(struct drm_device *dev, struct nvbios *bios) * fptable[4] is the minimum * RAMDAC_FP_HCRTC -> RAMDAC_FP_HSYNC_START gap */ - bios->pub.digital_min_front_porch = fptable[4]; + bios->digital_min_front_porch = fptable[4]; ofs = -7; break; default: @@ -3467,7 +3583,7 @@ static int parse_fp_mode_table(struct drm_device *dev, struct nvbios *bios) /* nv4x cards need both a strap value and fpindex of 0xf to use DDC */ if (lth.lvds_ver > 0x10) - bios->pub.fp_no_ddc = fpstrapping != 0xf || fpindex != 0xf; + bios->fp_no_ddc = fpstrapping != 0xf || fpindex != 0xf; /* * If either the strap or xlated fpindex value are 0xf there is no @@ -3491,7 +3607,7 @@ static int parse_fp_mode_table(struct drm_device *dev, struct nvbios *bios) bool nouveau_bios_fp_mode(struct drm_device *dev, struct drm_display_mode *mode) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; uint8_t *mode_entry = &bios->data[bios->fp.mode_ptr]; if (!mode) /* just checking whether we can produce a mode */ @@ -3562,11 +3678,11 @@ int nouveau_bios_parse_lvds_table(struct drm_device *dev, int pxclk, bool *dl, b * until later, when this function should be called with non-zero pxclk */ struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; int fpstrapping = get_fp_strap(dev, bios), lvdsmanufacturerindex = 0; struct lvdstableheader lth; uint16_t lvdsofs; - int ret, chip_version = bios->pub.chip_version; + int ret, chip_version = bios->chip_version; ret = parse_lvds_manufacturer_table_header(dev, bios, <h); if (ret) @@ -3637,37 +3753,40 @@ int nouveau_bios_parse_lvds_table(struct drm_device *dev, int pxclk, bool *dl, b *if_is_24bit = bios->data[lvdsofs] & 16; break; case 0x30: - /* - * My money would be on there being a 24 bit interface bit in - * this table, but I have no example of a laptop bios with a - * 24 bit panel to confirm that. Hence we shout loudly if any - * bit other than bit 0 is set (I've not even seen bit 1) - */ - if (bios->data[lvdsofs] > 1) - NV_ERROR(dev, - "You have a very unusual laptop display; please report it\n"); + case 0x40: /* * No sign of the "power off for reset" or "reset for panel * on" bits, but it's safer to assume we should */ bios->fp.power_off_for_reset = true; bios->fp.reset_after_pclk_change = true; + /* * It's ok lvdsofs is wrong for nv4x edid case; dual_link is - * over-written, and BITbit1 isn't used + * over-written, and if_is_24bit isn't used */ bios->fp.dual_link = bios->data[lvdsofs] & 1; - bios->fp.BITbit1 = bios->data[lvdsofs] & 2; - bios->fp.duallink_transition_clk = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 5]) * 10; - break; - case 0x40: - bios->fp.dual_link = bios->data[lvdsofs] & 1; bios->fp.if_is_24bit = bios->data[lvdsofs] & 2; bios->fp.strapless_is_24bit = bios->data[bios->fp.lvdsmanufacturerpointer + 4]; bios->fp.duallink_transition_clk = ROM16(bios->data[bios->fp.lvdsmanufacturerpointer + 5]) * 10; break; } + /* Dell Latitude D620 reports a too-high value for the dual-link + * transition freq, causing us to program the panel incorrectly. + * + * It doesn't appear the VBIOS actually uses its transition freq + * (90000kHz), instead it uses the "Number of LVDS channels" field + * out of the panel ID structure (http://www.spwg.org/). + * + * For the moment, a quirk will do :) + */ + if ((dev->pdev->device == 0x01d7) && + (dev->pdev->subsystem_vendor == 0x1028) && + (dev->pdev->subsystem_device == 0x01c2)) { + bios->fp.duallink_transition_clk = 80000; + } + /* set dual_link flag for EDID case */ if (pxclk && (chip_version < 0x25 || chip_version > 0x28)) bios->fp.dual_link = (pxclk >= bios->fp.duallink_transition_clk); @@ -3682,7 +3801,7 @@ bios_output_config_match(struct drm_device *dev, struct dcb_entry *dcbent, uint16_t record, int record_len, int record_nr) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; uint32_t entry; uint16_t table; int i, v; @@ -3716,7 +3835,7 @@ nouveau_bios_dp_table(struct drm_device *dev, struct dcb_entry *dcbent, int *length) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; uint8_t *table; if (!bios->display.dp_table_ptr) { @@ -3725,7 +3844,7 @@ nouveau_bios_dp_table(struct drm_device *dev, struct dcb_entry *dcbent, } table = &bios->data[bios->display.dp_table_ptr]; - if (table[0] != 0x21) { + if (table[0] != 0x20 && table[0] != 0x21) { NV_ERROR(dev, "DisplayPort table version 0x%02x unknown\n", table[0]); return NULL; @@ -3765,7 +3884,7 @@ nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent, */ struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; uint8_t *table = &bios->data[bios->display.script_table_ptr]; uint8_t *otable = NULL; uint16_t script; @@ -3918,8 +4037,8 @@ int run_tmds_table(struct drm_device *dev, struct dcb_entry *dcbent, int head, i */ struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; - int cv = bios->pub.chip_version; + struct nvbios *bios = &dev_priv->vbios; + int cv = bios->chip_version; uint16_t clktable = 0, scriptptr; uint32_t sel_clk_binding, sel_clk; @@ -3978,8 +4097,8 @@ int get_pll_limits(struct drm_device *dev, uint32_t limit_match, struct pll_lims */ struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; - int cv = bios->pub.chip_version, pllindex = 0; + struct nvbios *bios = &dev_priv->vbios; + int cv = bios->chip_version, pllindex = 0; uint8_t pll_lim_ver = 0, headerlen = 0, recordlen = 0, entries = 0; uint32_t crystal_strap_mask, crystal_straps; @@ -4293,31 +4412,32 @@ int get_pll_limits(struct drm_device *dev, uint32_t limit_match, struct pll_lims break; } -#if 0 /* for easy debugging */ - ErrorF("pll.vco1.minfreq: %d\n", pll_lim->vco1.minfreq); - ErrorF("pll.vco1.maxfreq: %d\n", pll_lim->vco1.maxfreq); - ErrorF("pll.vco2.minfreq: %d\n", pll_lim->vco2.minfreq); - ErrorF("pll.vco2.maxfreq: %d\n", pll_lim->vco2.maxfreq); - - ErrorF("pll.vco1.min_inputfreq: %d\n", pll_lim->vco1.min_inputfreq); - ErrorF("pll.vco1.max_inputfreq: %d\n", pll_lim->vco1.max_inputfreq); - ErrorF("pll.vco2.min_inputfreq: %d\n", pll_lim->vco2.min_inputfreq); - ErrorF("pll.vco2.max_inputfreq: %d\n", pll_lim->vco2.max_inputfreq); - - ErrorF("pll.vco1.min_n: %d\n", pll_lim->vco1.min_n); - ErrorF("pll.vco1.max_n: %d\n", pll_lim->vco1.max_n); - ErrorF("pll.vco1.min_m: %d\n", pll_lim->vco1.min_m); - ErrorF("pll.vco1.max_m: %d\n", pll_lim->vco1.max_m); - ErrorF("pll.vco2.min_n: %d\n", pll_lim->vco2.min_n); - ErrorF("pll.vco2.max_n: %d\n", pll_lim->vco2.max_n); - ErrorF("pll.vco2.min_m: %d\n", pll_lim->vco2.min_m); - ErrorF("pll.vco2.max_m: %d\n", pll_lim->vco2.max_m); - - ErrorF("pll.max_log2p: %d\n", pll_lim->max_log2p); - ErrorF("pll.log2p_bias: %d\n", pll_lim->log2p_bias); - - ErrorF("pll.refclk: %d\n", pll_lim->refclk); -#endif + NV_DEBUG(dev, "pll.vco1.minfreq: %d\n", pll_lim->vco1.minfreq); + NV_DEBUG(dev, "pll.vco1.maxfreq: %d\n", pll_lim->vco1.maxfreq); + NV_DEBUG(dev, "pll.vco1.min_inputfreq: %d\n", pll_lim->vco1.min_inputfreq); + NV_DEBUG(dev, "pll.vco1.max_inputfreq: %d\n", pll_lim->vco1.max_inputfreq); + NV_DEBUG(dev, "pll.vco1.min_n: %d\n", pll_lim->vco1.min_n); + NV_DEBUG(dev, "pll.vco1.max_n: %d\n", pll_lim->vco1.max_n); + NV_DEBUG(dev, "pll.vco1.min_m: %d\n", pll_lim->vco1.min_m); + NV_DEBUG(dev, "pll.vco1.max_m: %d\n", pll_lim->vco1.max_m); + if (pll_lim->vco2.maxfreq) { + NV_DEBUG(dev, "pll.vco2.minfreq: %d\n", pll_lim->vco2.minfreq); + NV_DEBUG(dev, "pll.vco2.maxfreq: %d\n", pll_lim->vco2.maxfreq); + NV_DEBUG(dev, "pll.vco2.min_inputfreq: %d\n", pll_lim->vco2.min_inputfreq); + NV_DEBUG(dev, "pll.vco2.max_inputfreq: %d\n", pll_lim->vco2.max_inputfreq); + NV_DEBUG(dev, "pll.vco2.min_n: %d\n", pll_lim->vco2.min_n); + NV_DEBUG(dev, "pll.vco2.max_n: %d\n", pll_lim->vco2.max_n); + NV_DEBUG(dev, "pll.vco2.min_m: %d\n", pll_lim->vco2.min_m); + NV_DEBUG(dev, "pll.vco2.max_m: %d\n", pll_lim->vco2.max_m); + } + if (!pll_lim->max_p) { + NV_DEBUG(dev, "pll.max_log2p: %d\n", pll_lim->max_log2p); + NV_DEBUG(dev, "pll.log2p_bias: %d\n", pll_lim->log2p_bias); + } else { + NV_DEBUG(dev, "pll.min_p: %d\n", pll_lim->min_p); + NV_DEBUG(dev, "pll.max_p: %d\n", pll_lim->max_p); + } + NV_DEBUG(dev, "pll.refclk: %d\n", pll_lim->refclk); return 0; } @@ -4332,7 +4452,7 @@ static void parse_bios_version(struct drm_device *dev, struct nvbios *bios, uint */ bios->major_version = bios->data[offset + 3]; - bios->pub.chip_version = bios->data[offset + 2]; + bios->chip_version = bios->data[offset + 2]; NV_TRACE(dev, "Bios version %02x.%02x.%02x.%02x\n", bios->data[offset + 3], bios->data[offset + 2], bios->data[offset + 1], bios->data[offset]); @@ -4402,7 +4522,7 @@ static int parse_bit_A_tbl_entry(struct drm_device *dev, struct nvbios *bios, st } /* First entry is normal dac, 2nd tv-out perhaps? */ - bios->pub.dactestval = ROM32(bios->data[load_table_ptr + headerlen]) & 0x3ff; + bios->dactestval = ROM32(bios->data[load_table_ptr + headerlen]) & 0x3ff; return 0; } @@ -4526,8 +4646,8 @@ static int parse_bit_i_tbl_entry(struct drm_device *dev, struct nvbios *bios, st return -ENOSYS; } - bios->pub.dactestval = ROM32(bios->data[daccmpoffset + dacheaderlen]); - bios->pub.tvdactestval = ROM32(bios->data[daccmpoffset + dacheaderlen + 4]); + bios->dactestval = ROM32(bios->data[daccmpoffset + dacheaderlen]); + bios->tvdactestval = ROM32(bios->data[daccmpoffset + dacheaderlen + 4]); return 0; } @@ -4796,11 +4916,11 @@ static int parse_bmp_structure(struct drm_device *dev, struct nvbios *bios, unsi uint16_t legacy_scripts_offset, legacy_i2c_offset; /* load needed defaults in case we can't parse this info */ - bios->bdcb.dcb.i2c[0].write = NV_CIO_CRE_DDC_WR__INDEX; - bios->bdcb.dcb.i2c[0].read = NV_CIO_CRE_DDC_STATUS__INDEX; - bios->bdcb.dcb.i2c[1].write = NV_CIO_CRE_DDC0_WR__INDEX; - bios->bdcb.dcb.i2c[1].read = NV_CIO_CRE_DDC0_STATUS__INDEX; - bios->pub.digital_min_front_porch = 0x4b; + bios->dcb.i2c[0].write = NV_CIO_CRE_DDC_WR__INDEX; + bios->dcb.i2c[0].read = NV_CIO_CRE_DDC_STATUS__INDEX; + bios->dcb.i2c[1].write = NV_CIO_CRE_DDC0_WR__INDEX; + bios->dcb.i2c[1].read = NV_CIO_CRE_DDC0_STATUS__INDEX; + bios->digital_min_front_porch = 0x4b; bios->fmaxvco = 256000; bios->fminvco = 128000; bios->fp.duallink_transition_clk = 90000; @@ -4907,10 +5027,10 @@ static int parse_bmp_structure(struct drm_device *dev, struct nvbios *bios, unsi bios->legacy.i2c_indices.crt = bios->data[legacy_i2c_offset]; bios->legacy.i2c_indices.tv = bios->data[legacy_i2c_offset + 1]; bios->legacy.i2c_indices.panel = bios->data[legacy_i2c_offset + 2]; - bios->bdcb.dcb.i2c[0].write = bios->data[legacy_i2c_offset + 4]; - bios->bdcb.dcb.i2c[0].read = bios->data[legacy_i2c_offset + 5]; - bios->bdcb.dcb.i2c[1].write = bios->data[legacy_i2c_offset + 6]; - bios->bdcb.dcb.i2c[1].read = bios->data[legacy_i2c_offset + 7]; + bios->dcb.i2c[0].write = bios->data[legacy_i2c_offset + 4]; + bios->dcb.i2c[0].read = bios->data[legacy_i2c_offset + 5]; + bios->dcb.i2c[1].write = bios->data[legacy_i2c_offset + 6]; + bios->dcb.i2c[1].read = bios->data[legacy_i2c_offset + 7]; if (bmplength > 74) { bios->fmaxvco = ROM32(bmp[67]); @@ -4984,7 +5104,8 @@ read_dcb_i2c_entry(struct drm_device *dev, int dcb_version, uint8_t *i2ctable, i else NV_WARN(dev, "DCB I2C table has more entries than indexable " - "(%d entries, max index 15)\n", i2ctable[2]); + "(%d entries, max %d)\n", i2ctable[2], + DCB_MAX_NUM_I2C_ENTRIES); entry_len = i2ctable[3]; /* [4] is i2c_default_indices, read in parse_dcb_table() */ } @@ -5000,8 +5121,8 @@ read_dcb_i2c_entry(struct drm_device *dev, int dcb_version, uint8_t *i2ctable, i if (index == 0xf) return 0; - if (index > i2c_entries) { - NV_ERROR(dev, "DCB I2C index too big (%d > %d)\n", + if (index >= i2c_entries) { + NV_ERROR(dev, "DCB I2C index too big (%d >= %d)\n", index, i2ctable[2]); return -ENOENT; } @@ -5023,8 +5144,12 @@ read_dcb_i2c_entry(struct drm_device *dev, int dcb_version, uint8_t *i2ctable, i rdofs = wrofs = 0; } - if (dcb_i2c_ver >= 0x40 && port_type != 5 && port_type != 6) - NV_WARN(dev, "DCB I2C table has port type %d\n", port_type); + if (dcb_i2c_ver >= 0x40) { + if (port_type != 5 && port_type != 6) + NV_WARN(dev, "DCB I2C table has port type %d\n", port_type); + + i2c->entry = ROM32(i2ctable[headerlen + recordoffset + entry_len * index]); + } i2c->port_type = port_type; i2c->read = i2ctable[headerlen + recordoffset + rdofs + entry_len * index]; @@ -5036,7 +5161,7 @@ read_dcb_i2c_entry(struct drm_device *dev, int dcb_version, uint8_t *i2ctable, i static struct dcb_gpio_entry * new_gpio_entry(struct nvbios *bios) { - struct parsed_dcb_gpio *gpio = &bios->bdcb.gpio; + struct dcb_gpio_table *gpio = &bios->dcb.gpio; return &gpio->entry[gpio->entries++]; } @@ -5045,14 +5170,14 @@ struct dcb_gpio_entry * nouveau_bios_gpio_entry(struct drm_device *dev, enum dcb_gpio_tag tag) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; int i; - for (i = 0; i < bios->bdcb.gpio.entries; i++) { - if (bios->bdcb.gpio.entry[i].tag != tag) + for (i = 0; i < bios->dcb.gpio.entries; i++) { + if (bios->dcb.gpio.entry[i].tag != tag) continue; - return &bios->bdcb.gpio.entry[i]; + return &bios->dcb.gpio.entry[i]; } return NULL; @@ -5075,32 +5200,32 @@ parse_dcb30_gpio_entry(struct nvbios *bios, uint16_t offset) gpio->tag = tag; gpio->line = line; gpio->invert = flags != 4; + gpio->entry = ent; } static void parse_dcb40_gpio_entry(struct nvbios *bios, uint16_t offset) { + uint32_t entry = ROM32(bios->data[offset]); struct dcb_gpio_entry *gpio; - uint32_t ent = ROM32(bios->data[offset]); - uint8_t line = ent & 0x1f, - tag = ent >> 8 & 0xff; - if (tag == 0xff) + if ((entry & 0x0000ff00) == 0x0000ff00) return; gpio = new_gpio_entry(bios); - - /* Currently unused, we may need more fields parsed at some - * point. */ - gpio->tag = tag; - gpio->line = line; + gpio->tag = (entry & 0x0000ff00) >> 8; + gpio->line = (entry & 0x0000001f) >> 0; + gpio->state_default = (entry & 0x01000000) >> 24; + gpio->state[0] = (entry & 0x18000000) >> 27; + gpio->state[1] = (entry & 0x60000000) >> 29; + gpio->entry = entry; } static void parse_dcb_gpio_table(struct nvbios *bios) { struct drm_device *dev = bios->dev; - uint16_t gpio_table_ptr = bios->bdcb.gpio_table_ptr; + uint16_t gpio_table_ptr = bios->dcb.gpio_table_ptr; uint8_t *gpio_table = &bios->data[gpio_table_ptr]; int header_len = gpio_table[1], entries = gpio_table[2], @@ -5108,7 +5233,7 @@ parse_dcb_gpio_table(struct nvbios *bios) void (*parse_entry)(struct nvbios *, uint16_t) = NULL; int i; - if (bios->bdcb.version >= 0x40) { + if (bios->dcb.version >= 0x40) { if (gpio_table_ptr && entry_len != 4) { NV_WARN(dev, "Invalid DCB GPIO table entry length.\n"); return; @@ -5116,7 +5241,7 @@ parse_dcb_gpio_table(struct nvbios *bios) parse_entry = parse_dcb40_gpio_entry; - } else if (bios->bdcb.version >= 0x30) { + } else if (bios->dcb.version >= 0x30) { if (gpio_table_ptr && entry_len != 2) { NV_WARN(dev, "Invalid DCB GPIO table entry length.\n"); return; @@ -5124,7 +5249,7 @@ parse_dcb_gpio_table(struct nvbios *bios) parse_entry = parse_dcb30_gpio_entry; - } else if (bios->bdcb.version >= 0x22) { + } else if (bios->dcb.version >= 0x22) { /* * DCBs older than v3.0 don't really have a GPIO * table, instead they keep some GPIO info at fixed @@ -5158,30 +5283,82 @@ struct dcb_connector_table_entry * nouveau_bios_connector_entry(struct drm_device *dev, int index) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; struct dcb_connector_table_entry *cte; - if (index >= bios->bdcb.connector.entries) + if (index >= bios->dcb.connector.entries) return NULL; - cte = &bios->bdcb.connector.entry[index]; + cte = &bios->dcb.connector.entry[index]; if (cte->type == 0xff) return NULL; return cte; } +static enum dcb_connector_type +divine_connector_type(struct nvbios *bios, int index) +{ + struct dcb_table *dcb = &bios->dcb; + unsigned encoders = 0, type = DCB_CONNECTOR_NONE; + int i; + + for (i = 0; i < dcb->entries; i++) { + if (dcb->entry[i].connector == index) + encoders |= (1 << dcb->entry[i].type); + } + + if (encoders & (1 << OUTPUT_DP)) { + if (encoders & (1 << OUTPUT_TMDS)) + type = DCB_CONNECTOR_DP; + else + type = DCB_CONNECTOR_eDP; + } else + if (encoders & (1 << OUTPUT_TMDS)) { + if (encoders & (1 << OUTPUT_ANALOG)) + type = DCB_CONNECTOR_DVI_I; + else + type = DCB_CONNECTOR_DVI_D; + } else + if (encoders & (1 << OUTPUT_ANALOG)) { + type = DCB_CONNECTOR_VGA; + } else + if (encoders & (1 << OUTPUT_LVDS)) { + type = DCB_CONNECTOR_LVDS; + } else + if (encoders & (1 << OUTPUT_TV)) { + type = DCB_CONNECTOR_TV_0; + } + + return type; +} + +static void +apply_dcb_connector_quirks(struct nvbios *bios, int idx) +{ + struct dcb_connector_table_entry *cte = &bios->dcb.connector.entry[idx]; + struct drm_device *dev = bios->dev; + + /* Gigabyte NX85T */ + if ((dev->pdev->device == 0x0421) && + (dev->pdev->subsystem_vendor == 0x1458) && + (dev->pdev->subsystem_device == 0x344c)) { + if (cte->type == DCB_CONNECTOR_HDMI_1) + cte->type = DCB_CONNECTOR_DVI_I; + } +} + static void parse_dcb_connector_table(struct nvbios *bios) { struct drm_device *dev = bios->dev; - struct dcb_connector_table *ct = &bios->bdcb.connector; + struct dcb_connector_table *ct = &bios->dcb.connector; struct dcb_connector_table_entry *cte; - uint8_t *conntab = &bios->data[bios->bdcb.connector_table_ptr]; + uint8_t *conntab = &bios->data[bios->dcb.connector_table_ptr]; uint8_t *entry; int i; - if (!bios->bdcb.connector_table_ptr) { + if (!bios->dcb.connector_table_ptr) { NV_DEBUG_KMS(dev, "No DCB connector table present\n"); return; } @@ -5199,12 +5376,14 @@ parse_dcb_connector_table(struct nvbios *bios) entry = conntab + conntab[1]; cte = &ct->entry[0]; for (i = 0; i < conntab[2]; i++, entry += conntab[3], cte++) { + cte->index = i; if (conntab[3] == 2) cte->entry = ROM16(entry[0]); else cte->entry = ROM32(entry[0]); + cte->type = (cte->entry & 0x000000ff) >> 0; - cte->index = (cte->entry & 0x00000f00) >> 8; + cte->index2 = (cte->entry & 0x00000f00) >> 8; switch (cte->entry & 0x00033000) { case 0x00001000: cte->gpio_tag = 0x07; @@ -5226,12 +5405,43 @@ parse_dcb_connector_table(struct nvbios *bios) if (cte->type == 0xff) continue; + apply_dcb_connector_quirks(bios, i); + NV_INFO(dev, " %d: 0x%08x: type 0x%02x idx %d tag 0x%02x\n", i, cte->entry, cte->type, cte->index, cte->gpio_tag); + + /* check for known types, fallback to guessing the type + * from attached encoders if we hit an unknown. + */ + switch (cte->type) { + case DCB_CONNECTOR_VGA: + case DCB_CONNECTOR_TV_0: + case DCB_CONNECTOR_TV_1: + case DCB_CONNECTOR_TV_3: + case DCB_CONNECTOR_DVI_I: + case DCB_CONNECTOR_DVI_D: + case DCB_CONNECTOR_LVDS: + case DCB_CONNECTOR_DP: + case DCB_CONNECTOR_eDP: + case DCB_CONNECTOR_HDMI_0: + case DCB_CONNECTOR_HDMI_1: + break; + default: + cte->type = divine_connector_type(bios, cte->index); + NV_WARN(dev, "unknown type, using 0x%02x\n", cte->type); + break; + } + + if (nouveau_override_conntype) { + int type = divine_connector_type(bios, cte->index); + if (type != cte->type) + NV_WARN(dev, " -> type 0x%02x\n", cte->type); + } + } } -static struct dcb_entry *new_dcb_entry(struct parsed_dcb *dcb) +static struct dcb_entry *new_dcb_entry(struct dcb_table *dcb) { struct dcb_entry *entry = &dcb->entry[dcb->entries]; @@ -5241,7 +5451,7 @@ static struct dcb_entry *new_dcb_entry(struct parsed_dcb *dcb) return entry; } -static void fabricate_vga_output(struct parsed_dcb *dcb, int i2c, int heads) +static void fabricate_vga_output(struct dcb_table *dcb, int i2c, int heads) { struct dcb_entry *entry = new_dcb_entry(dcb); @@ -5252,7 +5462,7 @@ static void fabricate_vga_output(struct parsed_dcb *dcb, int i2c, int heads) /* "or" mostly unused in early gen crt modesetting, 0 is fine */ } -static void fabricate_dvi_i_output(struct parsed_dcb *dcb, bool twoHeads) +static void fabricate_dvi_i_output(struct dcb_table *dcb, bool twoHeads) { struct dcb_entry *entry = new_dcb_entry(dcb); @@ -5279,7 +5489,7 @@ static void fabricate_dvi_i_output(struct parsed_dcb *dcb, bool twoHeads) #endif } -static void fabricate_tv_output(struct parsed_dcb *dcb, bool twoHeads) +static void fabricate_tv_output(struct dcb_table *dcb, bool twoHeads) { struct dcb_entry *entry = new_dcb_entry(dcb); @@ -5290,13 +5500,13 @@ static void fabricate_tv_output(struct parsed_dcb *dcb, bool twoHeads) } static bool -parse_dcb20_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, +parse_dcb20_entry(struct drm_device *dev, struct dcb_table *dcb, uint32_t conn, uint32_t conf, struct dcb_entry *entry) { entry->type = conn & 0xf; entry->i2c_index = (conn >> 4) & 0xf; entry->heads = (conn >> 8) & 0xf; - if (bdcb->version >= 0x40) + if (dcb->version >= 0x40) entry->connector = (conn >> 12) & 0xf; entry->bus = (conn >> 16) & 0xf; entry->location = (conn >> 20) & 0x3; @@ -5314,7 +5524,7 @@ parse_dcb20_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, * Although the rest of a CRT conf dword is usually * zeros, mac biosen have stuff there so we must mask */ - entry->crtconf.maxfreq = (bdcb->version < 0x30) ? + entry->crtconf.maxfreq = (dcb->version < 0x30) ? (conf & 0xffff) * 10 : (conf & 0xff) * 10000; break; @@ -5323,7 +5533,7 @@ parse_dcb20_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, uint32_t mask; if (conf & 0x1) entry->lvdsconf.use_straps_for_mode = true; - if (bdcb->version < 0x22) { + if (dcb->version < 0x22) { mask = ~0xd; /* * The laptop in bug 14567 lies and claims to not use @@ -5347,7 +5557,7 @@ parse_dcb20_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, * Until we even try to use these on G8x, it's * useless reporting unknown bits. They all are. */ - if (bdcb->version >= 0x40) + if (dcb->version >= 0x40) break; NV_ERROR(dev, "Unknown LVDS configuration bits, " @@ -5357,7 +5567,7 @@ parse_dcb20_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, } case OUTPUT_TV: { - if (bdcb->version >= 0x30) + if (dcb->version >= 0x30) entry->tvconf.has_component_output = conf & (0x8 << 4); else entry->tvconf.has_component_output = false; @@ -5384,8 +5594,10 @@ parse_dcb20_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, break; case 0xe: /* weird g80 mobile type that "nv" treats as a terminator */ - bdcb->dcb.entries--; + dcb->entries--; return false; + default: + break; } /* unsure what DCB version introduces this, 3.0? */ @@ -5396,7 +5608,7 @@ parse_dcb20_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, } static bool -parse_dcb15_entry(struct drm_device *dev, struct parsed_dcb *dcb, +parse_dcb15_entry(struct drm_device *dev, struct dcb_table *dcb, uint32_t conn, uint32_t conf, struct dcb_entry *entry) { switch (conn & 0x0000000f) { @@ -5462,27 +5674,27 @@ parse_dcb15_entry(struct drm_device *dev, struct parsed_dcb *dcb, return true; } -static bool parse_dcb_entry(struct drm_device *dev, struct bios_parsed_dcb *bdcb, +static bool parse_dcb_entry(struct drm_device *dev, struct dcb_table *dcb, uint32_t conn, uint32_t conf) { - struct dcb_entry *entry = new_dcb_entry(&bdcb->dcb); + struct dcb_entry *entry = new_dcb_entry(dcb); bool ret; - if (bdcb->version >= 0x20) - ret = parse_dcb20_entry(dev, bdcb, conn, conf, entry); + if (dcb->version >= 0x20) + ret = parse_dcb20_entry(dev, dcb, conn, conf, entry); else - ret = parse_dcb15_entry(dev, &bdcb->dcb, conn, conf, entry); + ret = parse_dcb15_entry(dev, dcb, conn, conf, entry); if (!ret) return ret; - read_dcb_i2c_entry(dev, bdcb->version, bdcb->i2c_table, - entry->i2c_index, &bdcb->dcb.i2c[entry->i2c_index]); + read_dcb_i2c_entry(dev, dcb->version, dcb->i2c_table, + entry->i2c_index, &dcb->i2c[entry->i2c_index]); return true; } static -void merge_like_dcb_entries(struct drm_device *dev, struct parsed_dcb *dcb) +void merge_like_dcb_entries(struct drm_device *dev, struct dcb_table *dcb) { /* * DCB v2.0 lists each output combination separately. @@ -5534,8 +5746,7 @@ static int parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct bios_parsed_dcb *bdcb = &bios->bdcb; - struct parsed_dcb *dcb; + struct dcb_table *dcb = &bios->dcb; uint16_t dcbptr = 0, i2ctabptr = 0; uint8_t *dcbtable; uint8_t headerlen = 0x4, entries = DCB_MAX_NUM_ENTRIES; @@ -5543,9 +5754,6 @@ parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads) int recordlength = 8, confofs = 4; int i; - dcb = bios->pub.dcb = &bdcb->dcb; - dcb->entries = 0; - /* get the offset from 0x36 */ if (dev_priv->card_type > NV_04) { dcbptr = ROM16(bios->data[0x36]); @@ -5567,21 +5775,21 @@ parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads) dcbtable = &bios->data[dcbptr]; /* get DCB version */ - bdcb->version = dcbtable[0]; + dcb->version = dcbtable[0]; NV_TRACE(dev, "Found Display Configuration Block version %d.%d\n", - bdcb->version >> 4, bdcb->version & 0xf); + dcb->version >> 4, dcb->version & 0xf); - if (bdcb->version >= 0x20) { /* NV17+ */ + if (dcb->version >= 0x20) { /* NV17+ */ uint32_t sig; - if (bdcb->version >= 0x30) { /* NV40+ */ + if (dcb->version >= 0x30) { /* NV40+ */ headerlen = dcbtable[1]; entries = dcbtable[2]; recordlength = dcbtable[3]; i2ctabptr = ROM16(dcbtable[4]); sig = ROM32(dcbtable[6]); - bdcb->gpio_table_ptr = ROM16(dcbtable[10]); - bdcb->connector_table_ptr = ROM16(dcbtable[20]); + dcb->gpio_table_ptr = ROM16(dcbtable[10]); + dcb->connector_table_ptr = ROM16(dcbtable[20]); } else { i2ctabptr = ROM16(dcbtable[2]); sig = ROM32(dcbtable[4]); @@ -5593,7 +5801,7 @@ parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads) "signature (%08X)\n", sig); return -EINVAL; } - } else if (bdcb->version >= 0x15) { /* some NV11 and NV20 */ + } else if (dcb->version >= 0x15) { /* some NV11 and NV20 */ char sig[8] = { 0 }; strncpy(sig, (char *)&dcbtable[-7], 7); @@ -5641,14 +5849,11 @@ parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads) if (!i2ctabptr) NV_WARN(dev, "No pointer to DCB I2C port table\n"); else { - bdcb->i2c_table = &bios->data[i2ctabptr]; - if (bdcb->version >= 0x30) - bdcb->i2c_default_indices = bdcb->i2c_table[4]; + dcb->i2c_table = &bios->data[i2ctabptr]; + if (dcb->version >= 0x30) + dcb->i2c_default_indices = dcb->i2c_table[4]; } - parse_dcb_gpio_table(bios); - parse_dcb_connector_table(bios); - if (entries > DCB_MAX_NUM_ENTRIES) entries = DCB_MAX_NUM_ENTRIES; @@ -5673,7 +5878,7 @@ parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads) NV_TRACEWARN(dev, "Raw DCB entry %d: %08x %08x\n", dcb->entries, connection, config); - if (!parse_dcb_entry(dev, bdcb, connection, config)) + if (!parse_dcb_entry(dev, dcb, connection, config)) break; } @@ -5681,18 +5886,22 @@ parse_dcb_table(struct drm_device *dev, struct nvbios *bios, bool twoHeads) * apart for v2.1+ not being known for requiring merging, this * guarantees dcbent->index is the index of the entry in the rom image */ - if (bdcb->version < 0x21) + if (dcb->version < 0x21) merge_like_dcb_entries(dev, dcb); - return dcb->entries ? 0 : -ENXIO; + if (!dcb->entries) + return -ENXIO; + + parse_dcb_gpio_table(bios); + parse_dcb_connector_table(bios); + return 0; } static void fixup_legacy_connector(struct nvbios *bios) { - struct bios_parsed_dcb *bdcb = &bios->bdcb; - struct parsed_dcb *dcb = &bdcb->dcb; - int high = 0, i; + struct dcb_table *dcb = &bios->dcb; + int i, i2c, i2c_conn[DCB_MAX_NUM_I2C_ENTRIES] = { }; /* * DCB 3.0 also has the table in most cases, but there are some cards @@ -5700,9 +5909,11 @@ fixup_legacy_connector(struct nvbios *bios) * indices are all 0. We don't need the connector indices on pre-G80 * chips (yet?) so limit the use to DCB 4.0 and above. */ - if (bdcb->version >= 0x40) + if (dcb->version >= 0x40) return; + dcb->connector.entries = 0; + /* * No known connector info before v3.0, so make it up. the rule here * is: anything on the same i2c bus is considered to be on the same @@ -5710,37 +5921,38 @@ fixup_legacy_connector(struct nvbios *bios) * its own unique connector index. */ for (i = 0; i < dcb->entries; i++) { - if (dcb->entry[i].i2c_index == 0xf) - continue; - /* * Ignore the I2C index for on-chip TV-out, as there * are cards with bogus values (nv31m in bug 23212), * and it's otherwise useless. */ if (dcb->entry[i].type == OUTPUT_TV && - dcb->entry[i].location == DCB_LOC_ON_CHIP) { + dcb->entry[i].location == DCB_LOC_ON_CHIP) dcb->entry[i].i2c_index = 0xf; + i2c = dcb->entry[i].i2c_index; + + if (i2c_conn[i2c]) { + dcb->entry[i].connector = i2c_conn[i2c] - 1; continue; } - dcb->entry[i].connector = dcb->entry[i].i2c_index; - if (dcb->entry[i].connector > high) - high = dcb->entry[i].connector; + dcb->entry[i].connector = dcb->connector.entries++; + if (i2c != 0xf) + i2c_conn[i2c] = dcb->connector.entries; } - for (i = 0; i < dcb->entries; i++) { - if (dcb->entry[i].i2c_index != 0xf) - continue; - - dcb->entry[i].connector = ++high; + /* Fake the connector table as well as just connector indices */ + for (i = 0; i < dcb->connector.entries; i++) { + dcb->connector.entry[i].index = i; + dcb->connector.entry[i].type = divine_connector_type(bios, i); + dcb->connector.entry[i].gpio_tag = 0xff; } } static void fixup_legacy_i2c(struct nvbios *bios) { - struct parsed_dcb *dcb = &bios->bdcb.dcb; + struct dcb_table *dcb = &bios->dcb; int i; for (i = 0; i < dcb->entries; i++) { @@ -5826,7 +6038,7 @@ static int load_nv17_hw_sequencer_ucode(struct drm_device *dev, uint8_t *nouveau_bios_embedded_edid(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; const uint8_t edid_sig[] = { 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00 }; uint16_t offset = 0; @@ -5859,7 +6071,7 @@ nouveau_bios_run_init_table(struct drm_device *dev, uint16_t table, struct dcb_entry *dcbent) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; struct init_exec iexec = { true, false }; mutex_lock(&bios->lock); @@ -5872,7 +6084,7 @@ nouveau_bios_run_init_table(struct drm_device *dev, uint16_t table, static bool NVInitVBIOS(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; memset(bios, 0, sizeof(struct nvbios)); mutex_init(&bios->lock); @@ -5888,7 +6100,7 @@ static bool NVInitVBIOS(struct drm_device *dev) static int nouveau_parse_vbios_struct(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; const uint8_t bit_signature[] = { 0xff, 0xb8, 'B', 'I', 'T' }; const uint8_t bmp_signature[] = { 0xff, 0x7f, 'N', 'V', 0x0 }; int offset; @@ -5915,7 +6127,7 @@ int nouveau_run_vbios_init(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; int i, ret = 0; NVLockVgaCrtcs(dev, false); @@ -5946,9 +6158,9 @@ nouveau_run_vbios_init(struct drm_device *dev) } if (dev_priv->card_type >= NV_50) { - for (i = 0; i < bios->bdcb.dcb.entries; i++) { + for (i = 0; i < bios->dcb.entries; i++) { nouveau_bios_run_display_table(dev, - &bios->bdcb.dcb.entry[i], + &bios->dcb.entry[i], 0, 0); } } @@ -5962,11 +6174,11 @@ static void nouveau_bios_i2c_devices_takedown(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; struct dcb_i2c_entry *entry; int i; - entry = &bios->bdcb.dcb.i2c[0]; + entry = &bios->dcb.i2c[0]; for (i = 0; i < DCB_MAX_NUM_I2C_ENTRIES; i++, entry++) nouveau_i2c_fini(dev, entry); } @@ -5975,13 +6187,11 @@ int nouveau_bios_init(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct nvbios *bios = &dev_priv->vbios; uint32_t saved_nv_pextdev_boot_0; bool was_locked; int ret; - dev_priv->vbios = &bios->pub; - if (!NVInitVBIOS(dev)) return -ENODEV; @@ -6023,10 +6233,8 @@ nouveau_bios_init(struct drm_device *dev) bios_wr32(bios, NV_PEXTDEV_BOOT_0, saved_nv_pextdev_boot_0); ret = nouveau_run_vbios_init(dev); - if (ret) { - dev_priv->vbios = NULL; + if (ret) return ret; - } /* feature_byte on BMP is poor, but init always sets CR4B */ was_locked = NVLockVgaCrtcs(dev, false); diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.h b/drivers/gpu/drm/nouveau/nouveau_bios.h index fd94bd6..adf4ec2 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bios.h +++ b/drivers/gpu/drm/nouveau/nouveau_bios.h @@ -34,9 +34,72 @@ #define DCB_LOC_ON_CHIP 0 +struct dcb_i2c_entry { + uint32_t entry; + uint8_t port_type; + uint8_t read, write; + struct nouveau_i2c_chan *chan; +}; + +enum dcb_gpio_tag { + DCB_GPIO_TVDAC0 = 0xc, + DCB_GPIO_TVDAC1 = 0x2d, +}; + +struct dcb_gpio_entry { + enum dcb_gpio_tag tag; + int line; + bool invert; + uint32_t entry; + uint8_t state_default; + uint8_t state[2]; +}; + +struct dcb_gpio_table { + int entries; + struct dcb_gpio_entry entry[DCB_MAX_NUM_GPIO_ENTRIES]; +}; + +enum dcb_connector_type { + DCB_CONNECTOR_VGA = 0x00, + DCB_CONNECTOR_TV_0 = 0x10, + DCB_CONNECTOR_TV_1 = 0x11, + DCB_CONNECTOR_TV_3 = 0x13, + DCB_CONNECTOR_DVI_I = 0x30, + DCB_CONNECTOR_DVI_D = 0x31, + DCB_CONNECTOR_LVDS = 0x40, + DCB_CONNECTOR_DP = 0x46, + DCB_CONNECTOR_eDP = 0x47, + DCB_CONNECTOR_HDMI_0 = 0x60, + DCB_CONNECTOR_HDMI_1 = 0x61, + DCB_CONNECTOR_NONE = 0xff +}; + +struct dcb_connector_table_entry { + uint8_t index; + uint32_t entry; + enum dcb_connector_type type; + uint8_t index2; + uint8_t gpio_tag; +}; + +struct dcb_connector_table { + int entries; + struct dcb_connector_table_entry entry[DCB_MAX_NUM_CONNECTOR_ENTRIES]; +}; + +enum dcb_type { + OUTPUT_ANALOG = 0, + OUTPUT_TV = 1, + OUTPUT_TMDS = 2, + OUTPUT_LVDS = 3, + OUTPUT_DP = 6, + OUTPUT_ANY = -1 +}; + struct dcb_entry { int index; /* may not be raw dcb index if merging has happened */ - uint8_t type; + enum dcb_type type; uint8_t i2c_index; uint8_t heads; uint8_t connector; @@ -71,69 +134,22 @@ struct dcb_entry { bool i2c_upper_default; }; -struct dcb_i2c_entry { - uint8_t port_type; - uint8_t read, write; - struct nouveau_i2c_chan *chan; -}; +struct dcb_table { + uint8_t version; -struct parsed_dcb { int entries; struct dcb_entry entry[DCB_MAX_NUM_ENTRIES]; - struct dcb_i2c_entry i2c[DCB_MAX_NUM_I2C_ENTRIES]; -}; - -enum dcb_gpio_tag { - DCB_GPIO_TVDAC0 = 0xc, - DCB_GPIO_TVDAC1 = 0x2d, -}; - -struct dcb_gpio_entry { - enum dcb_gpio_tag tag; - int line; - bool invert; -}; - -struct parsed_dcb_gpio { - int entries; - struct dcb_gpio_entry entry[DCB_MAX_NUM_GPIO_ENTRIES]; -}; - -struct dcb_connector_table_entry { - uint32_t entry; - uint8_t type; - uint8_t index; - uint8_t gpio_tag; -}; - -struct dcb_connector_table { - int entries; - struct dcb_connector_table_entry entry[DCB_MAX_NUM_CONNECTOR_ENTRIES]; -}; - -struct bios_parsed_dcb { - uint8_t version; - - struct parsed_dcb dcb; uint8_t *i2c_table; uint8_t i2c_default_indices; + struct dcb_i2c_entry i2c[DCB_MAX_NUM_I2C_ENTRIES]; uint16_t gpio_table_ptr; - struct parsed_dcb_gpio gpio; + struct dcb_gpio_table gpio; uint16_t connector_table_ptr; struct dcb_connector_table connector; }; -enum nouveau_encoder_type { - OUTPUT_ANALOG = 0, - OUTPUT_TV = 1, - OUTPUT_TMDS = 2, - OUTPUT_LVDS = 3, - OUTPUT_DP = 6, - OUTPUT_ANY = -1 -}; - enum nouveau_or { OUTPUT_A = (1 << 0), OUTPUT_B = (1 << 1), @@ -190,8 +206,8 @@ struct pll_lims { int refclk; }; -struct nouveau_bios_info { - struct parsed_dcb *dcb; +struct nvbios { + struct drm_device *dev; uint8_t chip_version; @@ -199,11 +215,6 @@ struct nouveau_bios_info { uint32_t tvdactestval; uint8_t digital_min_front_porch; bool fp_no_ddc; -}; - -struct nvbios { - struct drm_device *dev; - struct nouveau_bios_info pub; struct mutex lock; @@ -234,7 +245,7 @@ struct nvbios { uint16_t some_script_ptr; /* BIT I + 14 */ uint16_t init96_tbl_ptr; /* BIT I + 16 */ - struct bios_parsed_dcb bdcb; + struct dcb_table dcb; struct { int crtchead; @@ -260,7 +271,6 @@ struct nvbios { bool reset_after_pclk_change; bool dual_link; bool link_c_increment; - bool BITbit1; bool if_is_24bit; int duallink_transition_clk; uint8_t strapless_is_24bit; diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 028719f..8fac10d 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -71,7 +71,7 @@ nouveau_bo_fixup_align(struct drm_device *dev, * many small buffers. */ if (dev_priv->card_type == NV_50) { - uint32_t block_size = nouveau_mem_fb_amount(dev) >> 15; + uint32_t block_size = dev_priv->vram_size >> 15; int i; switch (tile_flags) { @@ -153,17 +153,17 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan, nvbo->placement.fpfn = 0; nvbo->placement.lpfn = mappable ? dev_priv->fb_mappable_pages : 0; - nouveau_bo_placement_set(nvbo, flags); + nouveau_bo_placement_set(nvbo, flags, 0); nvbo->channel = chan; 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); - nvbo->channel = NULL; if (ret) { /* ttm will call nouveau_bo_del_ttm if it fails.. */ return ret; } + nvbo->channel = NULL; spin_lock(&dev_priv->ttm.bo_list_lock); list_add_tail(&nvbo->head, &dev_priv->ttm.bo_list); @@ -172,26 +172,33 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan, return 0; } +static void +set_placement_list(uint32_t *pl, unsigned *n, uint32_t type, uint32_t flags) +{ + *n = 0; + + if (type & TTM_PL_FLAG_VRAM) + pl[(*n)++] = TTM_PL_FLAG_VRAM | flags; + if (type & TTM_PL_FLAG_TT) + pl[(*n)++] = TTM_PL_FLAG_TT | flags; + if (type & TTM_PL_FLAG_SYSTEM) + pl[(*n)++] = TTM_PL_FLAG_SYSTEM | flags; +} + void -nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t memtype) +nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t type, uint32_t busy) { - int n = 0; - - if (memtype & TTM_PL_FLAG_VRAM) - nvbo->placements[n++] = TTM_PL_FLAG_VRAM | TTM_PL_MASK_CACHING; - if (memtype & TTM_PL_FLAG_TT) - nvbo->placements[n++] = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING; - if (memtype & TTM_PL_FLAG_SYSTEM) - nvbo->placements[n++] = TTM_PL_FLAG_SYSTEM | TTM_PL_MASK_CACHING; - nvbo->placement.placement = nvbo->placements; - 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; - } + struct ttm_placement *pl = &nvbo->placement; + uint32_t flags = TTM_PL_MASK_CACHING | + (nvbo->pin_refcnt ? TTM_PL_FLAG_NO_EVICT : 0); + + pl->placement = nvbo->placements; + set_placement_list(nvbo->placements, &pl->num_placement, + type, flags); + + pl->busy_placement = nvbo->busy_placements; + set_placement_list(nvbo->busy_placements, &pl->num_busy_placement, + type | busy, flags); } int @@ -199,7 +206,7 @@ nouveau_bo_pin(struct nouveau_bo *nvbo, uint32_t memtype) { struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev); struct ttm_buffer_object *bo = &nvbo->bo; - int ret, i; + int ret; if (nvbo->pin_refcnt && !(memtype & (1 << bo->mem.mem_type))) { NV_ERROR(nouveau_bdev(bo->bdev)->dev, @@ -215,9 +222,7 @@ nouveau_bo_pin(struct nouveau_bo *nvbo, uint32_t memtype) if (ret) goto out; - nouveau_bo_placement_set(nvbo, memtype); - for (i = 0; i < nvbo->placement.num_placement; i++) - nvbo->placements[i] |= TTM_PL_FLAG_NO_EVICT; + nouveau_bo_placement_set(nvbo, memtype, 0); ret = ttm_bo_validate(bo, &nvbo->placement, false, false); if (ret == 0) { @@ -244,7 +249,7 @@ nouveau_bo_unpin(struct nouveau_bo *nvbo) { struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev); struct ttm_buffer_object *bo = &nvbo->bo; - int ret, i; + int ret; if (--nvbo->pin_refcnt) return 0; @@ -253,8 +258,7 @@ nouveau_bo_unpin(struct nouveau_bo *nvbo) if (ret) return ret; - for (i = 0; i < nvbo->placement.num_placement; i++) - nvbo->placements[i] &= ~TTM_PL_FLAG_NO_EVICT; + nouveau_bo_placement_set(nvbo, bo->mem.placement, 0); ret = ttm_bo_validate(bo, &nvbo->placement, false, false); if (ret == 0) { @@ -395,8 +399,8 @@ nouveau_bo_init_mem_type(struct ttm_bo_device *bdev, uint32_t type, man->io_addr = NULL; man->io_offset = drm_get_resource_start(dev, 1); man->io_size = drm_get_resource_len(dev, 1); - if (man->io_size > nouveau_mem_fb_amount(dev)) - man->io_size = nouveau_mem_fb_amount(dev); + if (man->io_size > dev_priv->vram_size) + man->io_size = dev_priv->vram_size; man->gpu_offset = dev_priv->vm_vram_base; break; @@ -439,11 +443,11 @@ nouveau_bo_evict_flags(struct ttm_buffer_object *bo, struct ttm_placement *pl) switch (bo->mem.mem_type) { case TTM_PL_VRAM: - nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_TT | + nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_TT, TTM_PL_FLAG_SYSTEM); break; default: - nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_SYSTEM); + nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_SYSTEM, 0); break; } diff --git a/drivers/gpu/drm/nouveau/nouveau_calc.c b/drivers/gpu/drm/nouveau/nouveau_calc.c index ee2b845..88f9bc0 100644 --- a/drivers/gpu/drm/nouveau/nouveau_calc.c +++ b/drivers/gpu/drm/nouveau/nouveau_calc.c @@ -274,7 +274,7 @@ getMNP_single(struct drm_device *dev, struct pll_lims *pll_lim, int clk, * returns calculated clock */ struct drm_nouveau_private *dev_priv = dev->dev_private; - int cv = dev_priv->vbios->chip_version; + int cv = dev_priv->vbios.chip_version; int minvco = pll_lim->vco1.minfreq, maxvco = pll_lim->vco1.maxfreq; int minM = pll_lim->vco1.min_m, maxM = pll_lim->vco1.max_m; int minN = pll_lim->vco1.min_n, maxN = pll_lim->vco1.max_n; @@ -373,7 +373,7 @@ getMNP_double(struct drm_device *dev, struct pll_lims *pll_lim, int clk, * returns calculated clock */ struct drm_nouveau_private *dev_priv = dev->dev_private; - int chip_version = dev_priv->vbios->chip_version; + int chip_version = dev_priv->vbios.chip_version; int minvco1 = pll_lim->vco1.minfreq, maxvco1 = pll_lim->vco1.maxfreq; int minvco2 = pll_lim->vco2.minfreq, maxvco2 = pll_lim->vco2.maxfreq; int minU1 = pll_lim->vco1.min_inputfreq, minU2 = pll_lim->vco2.min_inputfreq; diff --git a/drivers/gpu/drm/nouveau/nouveau_channel.c b/drivers/gpu/drm/nouveau/nouveau_channel.c index adac0f8..1fc57ef 100644 --- a/drivers/gpu/drm/nouveau/nouveau_channel.c +++ b/drivers/gpu/drm/nouveau/nouveau_channel.c @@ -142,7 +142,6 @@ nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret, GFP_KERNEL); if (!dev_priv->fifos[channel]) return -ENOMEM; - dev_priv->fifo_alloc_count++; chan = dev_priv->fifos[channel]; INIT_LIST_HEAD(&chan->nvsw.vbl_wait); INIT_LIST_HEAD(&chan->fence.pending); @@ -280,9 +279,18 @@ nouveau_channel_free(struct nouveau_channel *chan) */ nouveau_fence_fini(chan); - /* Ensure the channel is no longer active on the GPU */ + /* This will prevent pfifo from switching channels. */ pfifo->reassign(dev, false); + /* We want to give pgraph a chance to idle and get rid of all potential + * errors. We need to do this before the lock, otherwise the irq handler + * is unable to process them. + */ + if (pgraph->channel(dev) == chan) + nouveau_wait_for_idle(dev); + + spin_lock_irqsave(&dev_priv->context_switch_lock, flags); + pgraph->fifo_access(dev, false); if (pgraph->channel(dev) == chan) pgraph->unload_context(dev); @@ -298,6 +306,8 @@ nouveau_channel_free(struct nouveau_channel *chan) pfifo->reassign(dev, true); + spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags); + /* Release the channel's resources */ nouveau_gpuobj_ref_del(dev, &chan->pushbuf); if (chan->pushbuf_bo) { @@ -310,7 +320,6 @@ nouveau_channel_free(struct nouveau_channel *chan) iounmap(chan->user); dev_priv->fifos[chan->id] = NULL; - dev_priv->fifo_alloc_count--; kfree(chan); } diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c b/drivers/gpu/drm/nouveau/nouveau_connector.c index d2f6335..8fc0145 100644 --- a/drivers/gpu/drm/nouveau/nouveau_connector.c +++ b/drivers/gpu/drm/nouveau/nouveau_connector.c @@ -218,7 +218,7 @@ nouveau_connector_set_encoder(struct drm_connector *connector, connector->interlace_allowed = true; } - if (connector->connector_type == DRM_MODE_CONNECTOR_DVII) { + if (nv_connector->dcb->type == DCB_CONNECTOR_DVI_I) { drm_connector_property_set_value(connector, dev->mode_config.dvi_i_subconnector_property, nv_encoder->dcb->type == OUTPUT_TMDS ? @@ -236,7 +236,7 @@ nouveau_connector_detect(struct drm_connector *connector) struct nouveau_i2c_chan *i2c; int type, flags; - if (connector->connector_type == DRM_MODE_CONNECTOR_LVDS) + if (nv_connector->dcb->type == DCB_CONNECTOR_LVDS) nv_encoder = find_encoder_by_type(connector, OUTPUT_LVDS); if (nv_encoder && nv_connector->native_mode) { #ifdef CONFIG_ACPI @@ -279,7 +279,7 @@ nouveau_connector_detect(struct drm_connector *connector) * same i2c channel so the value returned from ddc_detect * isn't necessarily correct. */ - if (connector->connector_type == DRM_MODE_CONNECTOR_DVII) { + if (nv_connector->dcb->type == DCB_CONNECTOR_DVI_I) { if (nv_connector->edid->input & DRM_EDID_INPUT_DIGITAL) type = OUTPUT_TMDS; else @@ -300,7 +300,7 @@ nouveau_connector_detect(struct drm_connector *connector) detect_analog: nv_encoder = find_encoder_by_type(connector, OUTPUT_ANALOG); - if (!nv_encoder) + if (!nv_encoder && !nouveau_tv_disable) nv_encoder = find_encoder_by_type(connector, OUTPUT_TV); if (nv_encoder) { struct drm_encoder *encoder = to_drm_encoder(nv_encoder); @@ -321,11 +321,11 @@ detect_analog: static void nouveau_connector_force(struct drm_connector *connector) { - struct drm_device *dev = connector->dev; + struct nouveau_connector *nv_connector = nouveau_connector(connector); struct nouveau_encoder *nv_encoder; int type; - if (connector->connector_type == DRM_MODE_CONNECTOR_DVII) { + if (nv_connector->dcb->type == DCB_CONNECTOR_DVI_I) { if (connector->force == DRM_FORCE_ON_DIGITAL) type = OUTPUT_TMDS; else @@ -335,7 +335,7 @@ nouveau_connector_force(struct drm_connector *connector) nv_encoder = find_encoder_by_type(connector, type); if (!nv_encoder) { - NV_ERROR(dev, "can't find encoder to force %s on!\n", + NV_ERROR(connector->dev, "can't find encoder to force %s on!\n", drm_get_connector_name(connector)); connector->status = connector_status_disconnected; return; @@ -369,7 +369,7 @@ nouveau_connector_set_property(struct drm_connector *connector, } /* LVDS always needs gpu scaling */ - if (connector->connector_type == DRM_MODE_CONNECTOR_LVDS && + if (nv_connector->dcb->type == DCB_CONNECTOR_LVDS && value == DRM_MODE_SCALE_NONE) return -EINVAL; @@ -535,7 +535,7 @@ nouveau_connector_get_modes(struct drm_connector *connector) /* If we're not LVDS, destroy the previous native mode, the attached * monitor could have changed. */ - if (connector->connector_type != DRM_MODE_CONNECTOR_LVDS && + if (nv_connector->dcb->type != DCB_CONNECTOR_LVDS && nv_connector->native_mode) { drm_mode_destroy(dev, nv_connector->native_mode); nv_connector->native_mode = NULL; @@ -563,7 +563,7 @@ nouveau_connector_get_modes(struct drm_connector *connector) ret = get_slave_funcs(nv_encoder)-> get_modes(to_drm_encoder(nv_encoder), connector); - if (connector->connector_type == DRM_MODE_CONNECTOR_LVDS) + if (nv_encoder->dcb->type == OUTPUT_LVDS) ret += nouveau_connector_scaler_modes_add(connector); return ret; @@ -613,6 +613,9 @@ nouveau_connector_mode_valid(struct drm_connector *connector, clock *= 3; break; + default: + BUG_ON(1); + return MODE_BAD; } if (clock < min_clock) @@ -680,7 +683,7 @@ nouveau_connector_create_lvds(struct drm_device *dev, /* Firstly try getting EDID over DDC, if allowed and I2C channel * is available. */ - if (!dev_priv->VBIOS.pub.fp_no_ddc && nv_encoder->dcb->i2c_index < 0xf) + if (!dev_priv->vbios.fp_no_ddc && nv_encoder->dcb->i2c_index < 0xf) i2c = nouveau_i2c_find(dev, nv_encoder->dcb->i2c_index); if (i2c) { @@ -695,7 +698,7 @@ nouveau_connector_create_lvds(struct drm_device *dev, */ if (!nv_connector->edid && nouveau_bios_fp_mode(dev, &native) && (nv_encoder->dcb->lvdsconf.use_straps_for_mode || - dev_priv->VBIOS.pub.fp_no_ddc)) { + dev_priv->vbios.fp_no_ddc)) { nv_connector->native_mode = drm_mode_duplicate(dev, &native); goto out; } @@ -704,7 +707,7 @@ nouveau_connector_create_lvds(struct drm_device *dev, * stored for the panel stored in them. */ if (!nv_connector->edid && !nv_connector->native_mode && - !dev_priv->VBIOS.pub.fp_no_ddc) { + !dev_priv->vbios.fp_no_ddc) { struct edid *edid = (struct edid *)nouveau_bios_embedded_edid(dev); if (edid) { @@ -739,46 +742,66 @@ out: } int -nouveau_connector_create(struct drm_device *dev, int index, int type) +nouveau_connector_create(struct drm_device *dev, + struct dcb_connector_table_entry *dcb) { struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_connector *nv_connector = NULL; struct drm_connector *connector; struct drm_encoder *encoder; - int ret; + int ret, type; NV_DEBUG_KMS(dev, "\n"); - nv_connector = kzalloc(sizeof(*nv_connector), GFP_KERNEL); - if (!nv_connector) - return -ENOMEM; - nv_connector->dcb = nouveau_bios_connector_entry(dev, index); - connector = &nv_connector->base; - - switch (type) { - case DRM_MODE_CONNECTOR_VGA: + switch (dcb->type) { + case DCB_CONNECTOR_NONE: + return 0; + case DCB_CONNECTOR_VGA: NV_INFO(dev, "Detected a VGA connector\n"); + type = DRM_MODE_CONNECTOR_VGA; break; - case DRM_MODE_CONNECTOR_DVID: - NV_INFO(dev, "Detected a DVI-D connector\n"); + case DCB_CONNECTOR_TV_0: + case DCB_CONNECTOR_TV_1: + case DCB_CONNECTOR_TV_3: + NV_INFO(dev, "Detected a TV connector\n"); + type = DRM_MODE_CONNECTOR_TV; break; - case DRM_MODE_CONNECTOR_DVII: + case DCB_CONNECTOR_DVI_I: NV_INFO(dev, "Detected a DVI-I connector\n"); + type = DRM_MODE_CONNECTOR_DVII; break; - case DRM_MODE_CONNECTOR_LVDS: - NV_INFO(dev, "Detected a LVDS connector\n"); + case DCB_CONNECTOR_DVI_D: + NV_INFO(dev, "Detected a DVI-D connector\n"); + type = DRM_MODE_CONNECTOR_DVID; break; - case DRM_MODE_CONNECTOR_TV: - NV_INFO(dev, "Detected a TV connector\n"); + case DCB_CONNECTOR_HDMI_0: + case DCB_CONNECTOR_HDMI_1: + NV_INFO(dev, "Detected a HDMI connector\n"); + type = DRM_MODE_CONNECTOR_HDMIA; break; - case DRM_MODE_CONNECTOR_DisplayPort: + case DCB_CONNECTOR_LVDS: + NV_INFO(dev, "Detected a LVDS connector\n"); + type = DRM_MODE_CONNECTOR_LVDS; + break; + case DCB_CONNECTOR_DP: NV_INFO(dev, "Detected a DisplayPort connector\n"); + type = DRM_MODE_CONNECTOR_DisplayPort; break; - default: - NV_ERROR(dev, "Unknown connector, this is not good.\n"); + case DCB_CONNECTOR_eDP: + NV_INFO(dev, "Detected an eDP connector\n"); + type = DRM_MODE_CONNECTOR_eDP; break; + default: + NV_ERROR(dev, "unknown connector type: 0x%02x!!\n", dcb->type); + return -EINVAL; } + nv_connector = kzalloc(sizeof(*nv_connector), GFP_KERNEL); + if (!nv_connector) + return -ENOMEM; + nv_connector->dcb = dcb; + connector = &nv_connector->base; + /* defaults, will get overridden in detect() */ connector->interlace_allowed = false; connector->doublescan_allowed = false; @@ -786,55 +809,65 @@ nouveau_connector_create(struct drm_device *dev, int index, int type) drm_connector_init(dev, connector, &nouveau_connector_funcs, type); drm_connector_helper_add(connector, &nouveau_connector_helper_funcs); + /* attach encoders */ + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); + + if (nv_encoder->dcb->connector != dcb->index) + continue; + + if (get_slave_funcs(nv_encoder)) + get_slave_funcs(nv_encoder)->create_resources(encoder, connector); + + drm_mode_connector_attach_encoder(connector, encoder); + } + + if (!connector->encoder_ids[0]) { + NV_WARN(dev, " no encoders, ignoring\n"); + drm_connector_cleanup(connector); + kfree(connector); + return 0; + } + /* Init DVI-I specific properties */ - if (type == DRM_MODE_CONNECTOR_DVII) { + if (dcb->type == DCB_CONNECTOR_DVI_I) { drm_mode_create_dvi_i_properties(dev); drm_connector_attach_property(connector, dev->mode_config.dvi_i_subconnector_property, 0); drm_connector_attach_property(connector, dev->mode_config.dvi_i_select_subconnector_property, 0); } - if (type != DRM_MODE_CONNECTOR_LVDS) + if (dcb->type != DCB_CONNECTOR_LVDS) nv_connector->use_dithering = false; - if (type == DRM_MODE_CONNECTOR_DVID || - type == DRM_MODE_CONNECTOR_DVII || - type == DRM_MODE_CONNECTOR_LVDS || - type == DRM_MODE_CONNECTOR_DisplayPort) { - nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; - - drm_connector_attach_property(connector, dev->mode_config.scaling_mode_property, - nv_connector->scaling_mode); - drm_connector_attach_property(connector, dev->mode_config.dithering_mode_property, - nv_connector->use_dithering ? DRM_MODE_DITHERING_ON - : DRM_MODE_DITHERING_OFF); - - } else { - nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; - - if (type == DRM_MODE_CONNECTOR_VGA && - dev_priv->card_type >= NV_50) { + switch (dcb->type) { + case DCB_CONNECTOR_VGA: + if (dev_priv->card_type >= NV_50) { drm_connector_attach_property(connector, dev->mode_config.scaling_mode_property, nv_connector->scaling_mode); } - } - - /* attach encoders */ - list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { - struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); - - if (nv_encoder->dcb->connector != index) - continue; - - if (get_slave_funcs(nv_encoder)) - get_slave_funcs(nv_encoder)->create_resources(encoder, connector); + /* fall-through */ + case DCB_CONNECTOR_TV_0: + case DCB_CONNECTOR_TV_1: + case DCB_CONNECTOR_TV_3: + nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; + break; + default: + nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; - drm_mode_connector_attach_encoder(connector, encoder); + drm_connector_attach_property(connector, + dev->mode_config.scaling_mode_property, + nv_connector->scaling_mode); + drm_connector_attach_property(connector, + dev->mode_config.dithering_mode_property, + nv_connector->use_dithering ? + DRM_MODE_DITHERING_ON : DRM_MODE_DITHERING_OFF); + break; } drm_sysfs_connector_add(connector); - if (connector->connector_type == DRM_MODE_CONNECTOR_LVDS) { + if (dcb->type == DCB_CONNECTOR_LVDS) { ret = nouveau_connector_create_lvds(dev, connector); if (ret) { connector->funcs->destroy(connector); diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.h b/drivers/gpu/drm/nouveau/nouveau_connector.h index 728b809..4ef38ab 100644 --- a/drivers/gpu/drm/nouveau/nouveau_connector.h +++ b/drivers/gpu/drm/nouveau/nouveau_connector.h @@ -49,6 +49,7 @@ static inline struct nouveau_connector *nouveau_connector( return container_of(con, struct nouveau_connector, base); } -int nouveau_connector_create(struct drm_device *dev, int i2c_index, int type); +int nouveau_connector_create(struct drm_device *, + struct dcb_connector_table_entry *); #endif /* __NOUVEAU_CONNECTOR_H__ */ diff --git a/drivers/gpu/drm/nouveau/nouveau_debugfs.c b/drivers/gpu/drm/nouveau/nouveau_debugfs.c index 89e36ee..a251886 100644 --- a/drivers/gpu/drm/nouveau/nouveau_debugfs.c +++ b/drivers/gpu/drm/nouveau/nouveau_debugfs.c @@ -137,16 +137,28 @@ nouveau_debugfs_memory_info(struct seq_file *m, void *data) { struct drm_info_node *node = (struct drm_info_node *) m->private; struct drm_minor *minor = node->minor; - struct drm_device *dev = minor->dev; + struct drm_nouveau_private *dev_priv = minor->dev->dev_private; + + seq_printf(m, "VRAM total: %dKiB\n", (int)(dev_priv->vram_size >> 10)); + return 0; +} + +static int +nouveau_debugfs_vbios_image(struct seq_file *m, void *data) +{ + struct drm_info_node *node = (struct drm_info_node *) m->private; + struct drm_nouveau_private *dev_priv = node->minor->dev->dev_private; + int i; - seq_printf(m, "VRAM total: %dKiB\n", - (int)(nouveau_mem_fb_amount(dev) >> 10)); + for (i = 0; i < dev_priv->vbios.length; i++) + seq_printf(m, "%c", dev_priv->vbios.data[i]); return 0; } static struct drm_info_list nouveau_debugfs_list[] = { { "chipset", nouveau_debugfs_chipset_info, 0, NULL }, { "memory", nouveau_debugfs_memory_info, 0, NULL }, + { "vbios.rom", nouveau_debugfs_vbios_image, 0, NULL }, }; #define NOUVEAU_DEBUGFS_ENTRIES ARRAY_SIZE(nouveau_debugfs_list) diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.c b/drivers/gpu/drm/nouveau/nouveau_dma.c index c8482a1..65c441a 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dma.c +++ b/drivers/gpu/drm/nouveau/nouveau_dma.c @@ -190,6 +190,11 @@ nv50_dma_push(struct nouveau_channel *chan, struct nouveau_bo *bo, nouveau_bo_wr32(pb, ip++, upper_32_bits(offset) | length << 8); chan->dma.ib_put = (chan->dma.ib_put + 1) & chan->dma.ib_max; + + DRM_MEMORYBARRIER(); + /* Flush writes. */ + nouveau_bo_rd32(pb, 0); + nvchan_wr32(chan, 0x8c, chan->dma.ib_put); chan->dma.ib_free--; } diff --git a/drivers/gpu/drm/nouveau/nouveau_dp.c b/drivers/gpu/drm/nouveau/nouveau_dp.c index f954ad9..deeb21c 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dp.c +++ b/drivers/gpu/drm/nouveau/nouveau_dp.c @@ -483,7 +483,7 @@ nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr, ctrl |= (cmd << NV50_AUXCH_CTRL_CMD_SHIFT); ctrl |= ((data_nr - 1) << NV50_AUXCH_CTRL_LEN_SHIFT); - for (;;) { + for (i = 0; i < 16; i++) { nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl | 0x80000000); nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl); nv_wr32(dev, NV50_AUXCH_CTRL(index), ctrl | 0x00010000); @@ -502,6 +502,12 @@ nouveau_dp_auxch(struct nouveau_i2c_chan *auxch, int cmd, int addr, break; } + if (i == 16) { + NV_ERROR(dev, "auxch DEFER too many times, bailing\n"); + ret = -EREMOTEIO; + goto out; + } + if (cmd & 1) { if ((stat & NV50_AUXCH_STAT_COUNT) != data_nr) { ret = -EREMOTEIO; diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.c b/drivers/gpu/drm/nouveau/nouveau_drv.c index da3b93b..60a709c 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drv.c +++ b/drivers/gpu/drm/nouveau/nouveau_drv.c @@ -75,14 +75,22 @@ MODULE_PARM_DESC(ignorelid, "Ignore ACPI lid status"); int nouveau_ignorelid = 0; module_param_named(ignorelid, nouveau_ignorelid, int, 0400); -MODULE_PARM_DESC(noagp, "Disable all acceleration"); +MODULE_PARM_DESC(noaccel, "Disable all acceleration"); int nouveau_noaccel = 0; module_param_named(noaccel, nouveau_noaccel, int, 0400); -MODULE_PARM_DESC(noagp, "Disable fbcon acceleration"); +MODULE_PARM_DESC(nofbaccel, "Disable fbcon acceleration"); int nouveau_nofbaccel = 0; module_param_named(nofbaccel, nouveau_nofbaccel, int, 0400); +MODULE_PARM_DESC(override_conntype, "Ignore DCB connector type"); +int nouveau_override_conntype = 0; +module_param_named(override_conntype, nouveau_override_conntype, int, 0400); + +MODULE_PARM_DESC(tv_disable, "Disable TV-out detection\n"); +int nouveau_tv_disable = 0; +module_param_named(tv_disable, nouveau_tv_disable, int, 0400); + MODULE_PARM_DESC(tv_norm, "Default TV norm.\n" "\t\tSupported: PAL, PAL-M, PAL-N, PAL-Nc, NTSC-M, NTSC-J,\n" "\t\t\thd480i, hd480p, hd576i, hd576p, hd720p, hd1080i.\n" @@ -154,9 +162,11 @@ nouveau_pci_suspend(struct pci_dev *pdev, pm_message_t pm_state) if (pm_state.event == PM_EVENT_PRETHAW) return 0; + NV_INFO(dev, "Disabling fbcon acceleration...\n"); fbdev_flags = dev_priv->fbdev_info->flags; dev_priv->fbdev_info->flags |= FBINFO_HWACCEL_DISABLED; + NV_INFO(dev, "Unpinning framebuffer(s)...\n"); list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { struct nouveau_framebuffer *nouveau_fb; diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h index 5be0cca..b908f77 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drv.h +++ b/drivers/gpu/drm/nouveau/nouveau_drv.h @@ -76,6 +76,7 @@ struct nouveau_bo { struct ttm_buffer_object bo; struct ttm_placement placement; u32 placements[3]; + u32 busy_placements[3]; struct ttm_bo_kmap_obj kmap; struct list_head head; @@ -519,6 +520,7 @@ struct drm_nouveau_private { struct workqueue_struct *wq; struct work_struct irq_work; + struct work_struct hpd_work; struct list_head vbl_waiting; @@ -533,12 +535,14 @@ struct drm_nouveau_private { struct fb_info *fbdev_info; - int fifo_alloc_count; struct nouveau_channel *fifos[NOUVEAU_MAX_CHANNEL_NR]; struct nouveau_engine engine; struct nouveau_channel *channel; + /* For PFIFO and PGRAPH. */ + spinlock_t context_switch_lock; + /* RAMIN configuration, RAMFC, RAMHT and RAMRO offsets */ struct nouveau_gpuobj *ramht; uint32_t ramin_rsvd_vram; @@ -550,12 +554,6 @@ struct drm_nouveau_private { uint32_t ramro_offset; uint32_t ramro_size; - /* base physical adresses */ - uint64_t fb_phys; - uint64_t fb_available_size; - uint64_t fb_mappable_pages; - uint64_t fb_aper_free; - struct { enum { NOUVEAU_GART_NONE = 0, @@ -569,10 +567,6 @@ struct drm_nouveau_private { struct nouveau_gpuobj *sg_ctxdma; struct page *sg_dummy_page; dma_addr_t sg_dummy_bus; - - /* nottm hack */ - struct drm_ttm_backend *sg_be; - unsigned long sg_handle; } gart_info; /* nv10-nv40 tiling regions */ @@ -581,6 +575,16 @@ struct drm_nouveau_private { spinlock_t lock; } tile; + /* VRAM/fb configuration */ + uint64_t vram_size; + uint64_t vram_sys_base; + + uint64_t fb_phys; + uint64_t fb_available_size; + uint64_t fb_mappable_pages; + uint64_t fb_aper_free; + int fb_mtrr; + /* G8x/G9x virtual address space */ uint64_t vm_gart_base; uint64_t vm_gart_size; @@ -589,10 +593,6 @@ struct drm_nouveau_private { uint64_t vm_end; struct nouveau_gpuobj *vm_vram_pt[NV50_VM_VRAM_NR]; int vm_vram_pt_nr; - uint64_t vram_sys_base; - - /* the mtrr covering the FB */ - int fb_mtrr; struct mem_block *ramin_heap; @@ -602,8 +602,7 @@ struct drm_nouveau_private { struct list_head gpuobj_list; - struct nvbios VBIOS; - struct nouveau_bios_info *vbios; + struct nvbios vbios; struct nv04_mode_state mode_reg; struct nv04_mode_state saved_reg; @@ -612,11 +611,7 @@ struct drm_nouveau_private { uint32_t dac_users[4]; struct nouveau_suspend_resume { - uint32_t fifo_mode; - uint32_t graph_ctx_control; - uint32_t graph_state; uint32_t *ramin_copy; - uint64_t ramin_size; } susres; struct backlight_device *backlight; @@ -680,6 +675,7 @@ extern int nouveau_uscript_tmds; extern int nouveau_vram_pushbuf; extern int nouveau_vram_notify; extern int nouveau_fbpercrtc; +extern int nouveau_tv_disable; extern char *nouveau_tv_norm; extern int nouveau_reg_debug; extern char *nouveau_vbios; @@ -687,6 +683,7 @@ extern int nouveau_ctxfw; extern int nouveau_ignorelid; extern int nouveau_nofbaccel; extern int nouveau_noaccel; +extern int nouveau_override_conntype; /* nouveau_state.c */ extern void nouveau_preclose(struct drm_device *dev, struct drm_file *); @@ -711,7 +708,7 @@ extern struct mem_block *nouveau_mem_alloc_block(struct mem_block *, struct drm_file *, int tail); extern void nouveau_mem_takedown(struct mem_block **heap); extern void nouveau_mem_free_block(struct mem_block *); -extern uint64_t nouveau_mem_fb_amount(struct drm_device *); +extern int nouveau_mem_detect(struct drm_device *dev); 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 *); @@ -928,6 +925,10 @@ 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); +/* nv50_fb.c */ +extern int nv50_fb_init(struct drm_device *); +extern void nv50_fb_takedown(struct drm_device *); + /* nv04_fifo.c */ extern int nv04_fifo_init(struct drm_device *); extern void nv04_fifo_disable(struct drm_device *); @@ -1027,6 +1028,7 @@ extern void nv50_graph_destroy_context(struct nouveau_channel *); 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 *); +extern int nv50_grctx_init(struct nouveau_grctx *); /* nouveau_grctx.c */ extern int nouveau_grctx_prog_load(struct drm_device *); @@ -1119,7 +1121,8 @@ extern int nouveau_bo_pin(struct nouveau_bo *, uint32_t flags); extern int nouveau_bo_unpin(struct nouveau_bo *); extern int nouveau_bo_map(struct nouveau_bo *); extern void nouveau_bo_unmap(struct nouveau_bo *); -extern void nouveau_bo_placement_set(struct nouveau_bo *, uint32_t memtype); +extern void nouveau_bo_placement_set(struct nouveau_bo *, uint32_t type, + uint32_t busy); extern u16 nouveau_bo_rd16(struct nouveau_bo *nvbo, unsigned index); extern void nouveau_bo_wr16(struct nouveau_bo *nvbo, unsigned index, u16 val); extern u32 nouveau_bo_rd32(struct nouveau_bo *nvbo, unsigned index); @@ -1163,6 +1166,16 @@ extern int nouveau_gem_ioctl_info(struct drm_device *, void *, int nv17_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag); int nv17_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state); +/* nv50_gpio.c */ +int nv50_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag); +int nv50_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state); + +/* nv50_calc. */ +int nv50_calc_pll(struct drm_device *, struct pll_lims *, int clk, + int *N1, int *M1, int *N2, int *M2, int *P); +int nv50_calc_pll2(struct drm_device *, struct pll_lims *, + int clk, int *N, int *fN, int *M, int *P); + #ifndef ioread32_native #ifdef __BIG_ENDIAN #define ioread16_native ioread16be diff --git a/drivers/gpu/drm/nouveau/nouveau_encoder.h b/drivers/gpu/drm/nouveau/nouveau_encoder.h index bc4a240..e1df820 100644 --- a/drivers/gpu/drm/nouveau/nouveau_encoder.h +++ b/drivers/gpu/drm/nouveau/nouveau_encoder.h @@ -47,6 +47,9 @@ struct nouveau_encoder { union { struct { + int mc_unknown; + uint32_t unk0; + uint32_t unk1; int dpcd_version; int link_nr; int link_bw; diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c index 8265fed..0846a1e 100644 --- a/drivers/gpu/drm/nouveau/nouveau_gem.c +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c @@ -182,40 +182,35 @@ nouveau_gem_set_domain(struct drm_gem_object *gem, uint32_t read_domains, { struct nouveau_bo *nvbo = gem->driver_private; struct ttm_buffer_object *bo = &nvbo->bo; - uint64_t flags; + uint32_t domains = valid_domains & + (write_domains ? write_domains : read_domains); + uint32_t pref_flags = 0, valid_flags = 0; - if (!valid_domains || (!read_domains && !write_domains)) + if (!domains) return -EINVAL; - if (write_domains) { - if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) && - (write_domains & NOUVEAU_GEM_DOMAIN_VRAM)) - flags = TTM_PL_FLAG_VRAM; - else - if ((valid_domains & NOUVEAU_GEM_DOMAIN_GART) && - (write_domains & NOUVEAU_GEM_DOMAIN_GART)) - flags = TTM_PL_FLAG_TT; - else - return -EINVAL; - } else { - if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) && - (read_domains & NOUVEAU_GEM_DOMAIN_VRAM) && - bo->mem.mem_type == TTM_PL_VRAM) - flags = TTM_PL_FLAG_VRAM; - else - if ((valid_domains & NOUVEAU_GEM_DOMAIN_GART) && - (read_domains & NOUVEAU_GEM_DOMAIN_GART) && - bo->mem.mem_type == TTM_PL_TT) - flags = TTM_PL_FLAG_TT; - else - if ((valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) && - (read_domains & NOUVEAU_GEM_DOMAIN_VRAM)) - flags = TTM_PL_FLAG_VRAM; - else - flags = TTM_PL_FLAG_TT; - } + if (valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) + valid_flags |= TTM_PL_FLAG_VRAM; + + if (valid_domains & NOUVEAU_GEM_DOMAIN_GART) + valid_flags |= TTM_PL_FLAG_TT; + + if ((domains & NOUVEAU_GEM_DOMAIN_VRAM) && + bo->mem.mem_type == TTM_PL_VRAM) + pref_flags |= TTM_PL_FLAG_VRAM; + + else if ((domains & NOUVEAU_GEM_DOMAIN_GART) && + bo->mem.mem_type == TTM_PL_TT) + pref_flags |= TTM_PL_FLAG_TT; + + else if (domains & NOUVEAU_GEM_DOMAIN_VRAM) + pref_flags |= TTM_PL_FLAG_VRAM; + + else + pref_flags |= TTM_PL_FLAG_TT; + + nouveau_bo_placement_set(nvbo, pref_flags, valid_flags); - nouveau_bo_placement_set(nvbo, flags); return 0; } diff --git a/drivers/gpu/drm/nouveau/nouveau_hw.c b/drivers/gpu/drm/nouveau/nouveau_hw.c index dc46792..7855b35 100644 --- a/drivers/gpu/drm/nouveau/nouveau_hw.c +++ b/drivers/gpu/drm/nouveau/nouveau_hw.c @@ -160,7 +160,7 @@ static void setPLL_single(struct drm_device *dev, uint32_t reg, struct nouveau_pll_vals *pv) { struct drm_nouveau_private *dev_priv = dev->dev_private; - int chip_version = dev_priv->vbios->chip_version; + int chip_version = dev_priv->vbios.chip_version; uint32_t oldpll = NVReadRAMDAC(dev, 0, reg); int oldN = (oldpll >> 8) & 0xff, oldM = oldpll & 0xff; uint32_t pll = (oldpll & 0xfff80000) | pv->log2P << 16 | pv->NM1; @@ -216,7 +216,7 @@ setPLL_double_highregs(struct drm_device *dev, uint32_t reg1, struct nouveau_pll_vals *pv) { struct drm_nouveau_private *dev_priv = dev->dev_private; - int chip_version = dev_priv->vbios->chip_version; + int chip_version = dev_priv->vbios.chip_version; bool nv3035 = chip_version == 0x30 || chip_version == 0x35; uint32_t reg2 = reg1 + ((reg1 == NV_RAMDAC_VPLL2) ? 0x5c : 0x70); uint32_t oldpll1 = NVReadRAMDAC(dev, 0, reg1); @@ -374,7 +374,7 @@ nouveau_hw_setpll(struct drm_device *dev, uint32_t reg1, struct nouveau_pll_vals *pv) { struct drm_nouveau_private *dev_priv = dev->dev_private; - int cv = dev_priv->vbios->chip_version; + int cv = dev_priv->vbios.chip_version; if (cv == 0x30 || cv == 0x31 || cv == 0x35 || cv == 0x36 || cv >= 0x40) { diff --git a/drivers/gpu/drm/nouveau/nouveau_i2c.c b/drivers/gpu/drm/nouveau/nouveau_i2c.c index 70e994d..316a3c7 100644 --- a/drivers/gpu/drm/nouveau/nouveau_i2c.c +++ b/drivers/gpu/drm/nouveau/nouveau_i2c.c @@ -254,16 +254,27 @@ struct nouveau_i2c_chan * nouveau_i2c_find(struct drm_device *dev, int index) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct nvbios *bios = &dev_priv->VBIOS; + struct dcb_i2c_entry *i2c = &dev_priv->vbios.dcb.i2c[index]; - if (index > DCB_MAX_NUM_I2C_ENTRIES) + if (index >= DCB_MAX_NUM_I2C_ENTRIES) return NULL; - if (!bios->bdcb.dcb.i2c[index].chan) { - if (nouveau_i2c_init(dev, &bios->bdcb.dcb.i2c[index], index)) - return NULL; + if (dev_priv->chipset >= NV_50 && (i2c->entry & 0x00000100)) { + uint32_t reg = 0xe500, val; + + if (i2c->port_type == 6) { + reg += i2c->read * 0x50; + val = 0x2002; + } else { + reg += ((i2c->entry & 0x1e00) >> 9) * 0x50; + val = 0xe001; + } + + nv_wr32(dev, reg, (nv_rd32(dev, reg) & ~0xf003) | val); } - return bios->bdcb.dcb.i2c[index].chan; + if (!i2c->chan && nouveau_i2c_init(dev, i2c, index)) + return NULL; + return i2c->chan; } diff --git a/drivers/gpu/drm/nouveau/nouveau_irq.c b/drivers/gpu/drm/nouveau/nouveau_irq.c index 447f9f6..13e73ce 100644 --- a/drivers/gpu/drm/nouveau/nouveau_irq.c +++ b/drivers/gpu/drm/nouveau/nouveau_irq.c @@ -51,6 +51,7 @@ nouveau_irq_preinstall(struct drm_device *dev) if (dev_priv->card_type == NV_50) { INIT_WORK(&dev_priv->irq_work, nv50_display_irq_handler_bh); + INIT_WORK(&dev_priv->hpd_work, nv50_display_irq_hotplug_bh); INIT_LIST_HEAD(&dev_priv->vbl_waiting); } } @@ -311,6 +312,31 @@ nouveau_print_bitfield_names_(uint32_t value, #define nouveau_print_bitfield_names(val, namelist) \ nouveau_print_bitfield_names_((val), (namelist), ARRAY_SIZE(namelist)) +struct nouveau_enum_names { + uint32_t value; + const char *name; +}; + +static void +nouveau_print_enum_names_(uint32_t value, + const struct nouveau_enum_names *namelist, + const int namelist_len) +{ + /* + * Caller must have already printed the KERN_* log level for us. + * Also the caller is responsible for adding the newline. + */ + int i; + for (i = 0; i < namelist_len; ++i) { + if (value == namelist[i].value) { + printk("%s", namelist[i].name); + return; + } + } + printk("unknown value 0x%08x", value); +} +#define nouveau_print_enum_names(val, namelist) \ + nouveau_print_enum_names_((val), (namelist), ARRAY_SIZE(namelist)) static int nouveau_graph_chid_from_grctx(struct drm_device *dev) @@ -427,14 +453,16 @@ nouveau_graph_dump_trap_info(struct drm_device *dev, const char *id, struct drm_nouveau_private *dev_priv = dev->dev_private; uint32_t nsource = trap->nsource, nstatus = trap->nstatus; - NV_INFO(dev, "%s - nSource:", id); - nouveau_print_bitfield_names(nsource, nsource_names); - printk(", nStatus:"); - if (dev_priv->card_type < NV_10) - nouveau_print_bitfield_names(nstatus, nstatus_names); - else - nouveau_print_bitfield_names(nstatus, nstatus_names_nv10); - printk("\n"); + if (dev_priv->card_type < NV_50) { + NV_INFO(dev, "%s - nSource:", id); + nouveau_print_bitfield_names(nsource, nsource_names); + printk(", nStatus:"); + if (dev_priv->card_type < NV_10) + nouveau_print_bitfield_names(nstatus, nstatus_names); + else + nouveau_print_bitfield_names(nstatus, nstatus_names_nv10); + printk("\n"); + } NV_INFO(dev, "%s - Ch %d/%d Class 0x%04x Mthd 0x%04x " "Data 0x%08x:0x%08x\n", @@ -578,27 +606,502 @@ nouveau_pgraph_irq_handler(struct drm_device *dev) } static void +nv50_pfb_vm_trap(struct drm_device *dev, int display, const char *name) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + uint32_t trap[6]; + int i, ch; + uint32_t idx = nv_rd32(dev, 0x100c90); + if (idx & 0x80000000) { + idx &= 0xffffff; + if (display) { + for (i = 0; i < 6; i++) { + nv_wr32(dev, 0x100c90, idx | i << 24); + trap[i] = nv_rd32(dev, 0x100c94); + } + for (ch = 0; ch < dev_priv->engine.fifo.channels; ch++) { + struct nouveau_channel *chan = dev_priv->fifos[ch]; + + if (!chan || !chan->ramin) + continue; + + if (trap[1] == chan->ramin->instance >> 12) + break; + } + NV_INFO(dev, "%s - VM: Trapped %s at %02x%04x%04x status %08x %08x channel %d\n", + name, (trap[5]&0x100?"read":"write"), + trap[5]&0xff, trap[4]&0xffff, + trap[3]&0xffff, trap[0], trap[2], ch); + } + nv_wr32(dev, 0x100c90, idx | 0x80000000); + } else if (display) { + NV_INFO(dev, "%s - no VM fault?\n", name); + } +} + +static struct nouveau_enum_names nv50_mp_exec_error_names[] = +{ + { 3, "STACK_UNDERFLOW" }, + { 4, "QUADON_ACTIVE" }, + { 8, "TIMEOUT" }, + { 0x10, "INVALID_OPCODE" }, + { 0x40, "BREAKPOINT" }, +}; + +static void +nv50_pgraph_mp_trap(struct drm_device *dev, int tpid, int display) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + uint32_t units = nv_rd32(dev, 0x1540); + uint32_t addr, mp10, status, pc, oplow, ophigh; + int i; + int mps = 0; + for (i = 0; i < 4; i++) { + if (!(units & 1 << (i+24))) + continue; + if (dev_priv->chipset < 0xa0) + addr = 0x408200 + (tpid << 12) + (i << 7); + else + addr = 0x408100 + (tpid << 11) + (i << 7); + mp10 = nv_rd32(dev, addr + 0x10); + status = nv_rd32(dev, addr + 0x14); + if (!status) + continue; + if (display) { + nv_rd32(dev, addr + 0x20); + pc = nv_rd32(dev, addr + 0x24); + oplow = nv_rd32(dev, addr + 0x70); + ophigh= nv_rd32(dev, addr + 0x74); + NV_INFO(dev, "PGRAPH_TRAP_MP_EXEC - " + "TP %d MP %d: ", tpid, i); + nouveau_print_enum_names(status, + nv50_mp_exec_error_names); + printk(" at %06x warp %d, opcode %08x %08x\n", + pc&0xffffff, pc >> 24, + oplow, ophigh); + } + nv_wr32(dev, addr + 0x10, mp10); + nv_wr32(dev, addr + 0x14, 0); + mps++; + } + if (!mps && display) + NV_INFO(dev, "PGRAPH_TRAP_MP_EXEC - TP %d: " + "No MPs claiming errors?\n", tpid); +} + +static void +nv50_pgraph_tp_trap(struct drm_device *dev, int type, uint32_t ustatus_old, + uint32_t ustatus_new, int display, const char *name) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + int tps = 0; + uint32_t units = nv_rd32(dev, 0x1540); + int i, r; + uint32_t ustatus_addr, ustatus; + for (i = 0; i < 16; i++) { + if (!(units & (1 << i))) + continue; + if (dev_priv->chipset < 0xa0) + ustatus_addr = ustatus_old + (i << 12); + else + ustatus_addr = ustatus_new + (i << 11); + ustatus = nv_rd32(dev, ustatus_addr) & 0x7fffffff; + if (!ustatus) + continue; + tps++; + switch (type) { + case 6: /* texture error... unknown for now */ + nv50_pfb_vm_trap(dev, display, name); + if (display) { + NV_ERROR(dev, "magic set %d:\n", i); + for (r = ustatus_addr + 4; r <= ustatus_addr + 0x10; r += 4) + NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r, + nv_rd32(dev, r)); + } + break; + case 7: /* MP error */ + if (ustatus & 0x00010000) { + nv50_pgraph_mp_trap(dev, i, display); + ustatus &= ~0x00010000; + } + break; + case 8: /* TPDMA error */ + { + uint32_t e0c = nv_rd32(dev, ustatus_addr + 4); + uint32_t e10 = nv_rd32(dev, ustatus_addr + 8); + uint32_t e14 = nv_rd32(dev, ustatus_addr + 0xc); + uint32_t e18 = nv_rd32(dev, ustatus_addr + 0x10); + uint32_t e1c = nv_rd32(dev, ustatus_addr + 0x14); + uint32_t e20 = nv_rd32(dev, ustatus_addr + 0x18); + uint32_t e24 = nv_rd32(dev, ustatus_addr + 0x1c); + nv50_pfb_vm_trap(dev, display, name); + /* 2d engine destination */ + if (ustatus & 0x00000010) { + if (display) { + NV_INFO(dev, "PGRAPH_TRAP_TPDMA_2D - TP %d - Unknown fault at address %02x%08x\n", + i, e14, e10); + NV_INFO(dev, "PGRAPH_TRAP_TPDMA_2D - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n", + i, e0c, e18, e1c, e20, e24); + } + ustatus &= ~0x00000010; + } + /* Render target */ + if (ustatus & 0x00000040) { + if (display) { + NV_INFO(dev, "PGRAPH_TRAP_TPDMA_RT - TP %d - Unknown fault at address %02x%08x\n", + i, e14, e10); + NV_INFO(dev, "PGRAPH_TRAP_TPDMA_RT - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n", + i, e0c, e18, e1c, e20, e24); + } + ustatus &= ~0x00000040; + } + /* CUDA memory: l[], g[] or stack. */ + if (ustatus & 0x00000080) { + if (display) { + if (e18 & 0x80000000) { + /* g[] read fault? */ + NV_INFO(dev, "PGRAPH_TRAP_TPDMA - TP %d - Global read fault at address %02x%08x\n", + i, e14, e10 | ((e18 >> 24) & 0x1f)); + e18 &= ~0x1f000000; + } else if (e18 & 0xc) { + /* g[] write fault? */ + NV_INFO(dev, "PGRAPH_TRAP_TPDMA - TP %d - Global write fault at address %02x%08x\n", + i, e14, e10 | ((e18 >> 7) & 0x1f)); + e18 &= ~0x00000f80; + } else { + NV_INFO(dev, "PGRAPH_TRAP_TPDMA - TP %d - Unknown CUDA fault at address %02x%08x\n", + i, e14, e10); + } + NV_INFO(dev, "PGRAPH_TRAP_TPDMA - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n", + i, e0c, e18, e1c, e20, e24); + } + ustatus &= ~0x00000080; + } + } + break; + } + if (ustatus) { + if (display) + NV_INFO(dev, "%s - TP%d: Unhandled ustatus 0x%08x\n", name, i, ustatus); + } + nv_wr32(dev, ustatus_addr, 0xc0000000); + } + + if (!tps && display) + NV_INFO(dev, "%s - No TPs claiming errors?\n", name); +} + +static void +nv50_pgraph_trap_handler(struct drm_device *dev) +{ + struct nouveau_pgraph_trap trap; + uint32_t status = nv_rd32(dev, 0x400108); + uint32_t ustatus; + int display = nouveau_ratelimit(); + + + if (!status && display) { + nouveau_graph_trap_info(dev, &trap); + nouveau_graph_dump_trap_info(dev, "PGRAPH_TRAP", &trap); + NV_INFO(dev, "PGRAPH_TRAP - no units reporting traps?\n"); + } + + /* DISPATCH: Relays commands to other units and handles NOTIFY, + * COND, QUERY. If you get a trap from it, the command is still stuck + * in DISPATCH and you need to do something about it. */ + if (status & 0x001) { + ustatus = nv_rd32(dev, 0x400804) & 0x7fffffff; + if (!ustatus && display) { + NV_INFO(dev, "PGRAPH_TRAP_DISPATCH - no ustatus?\n"); + } + + /* Known to be triggered by screwed up NOTIFY and COND... */ + if (ustatus & 0x00000001) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_DISPATCH_FAULT"); + nv_wr32(dev, 0x400500, 0); + if (nv_rd32(dev, 0x400808) & 0x80000000) { + if (display) { + if (nouveau_graph_trapped_channel(dev, &trap.channel)) + trap.channel = -1; + trap.class = nv_rd32(dev, 0x400814); + trap.mthd = nv_rd32(dev, 0x400808) & 0x1ffc; + trap.subc = (nv_rd32(dev, 0x400808) >> 16) & 0x7; + trap.data = nv_rd32(dev, 0x40080c); + trap.data2 = nv_rd32(dev, 0x400810); + nouveau_graph_dump_trap_info(dev, + "PGRAPH_TRAP_DISPATCH_FAULT", &trap); + NV_INFO(dev, "PGRAPH_TRAP_DISPATCH_FAULT - 400808: %08x\n", nv_rd32(dev, 0x400808)); + NV_INFO(dev, "PGRAPH_TRAP_DISPATCH_FAULT - 400848: %08x\n", nv_rd32(dev, 0x400848)); + } + nv_wr32(dev, 0x400808, 0); + } else if (display) { + NV_INFO(dev, "PGRAPH_TRAP_DISPATCH_FAULT - No stuck command?\n"); + } + nv_wr32(dev, 0x4008e8, nv_rd32(dev, 0x4008e8) & 3); + nv_wr32(dev, 0x400848, 0); + ustatus &= ~0x00000001; + } + if (ustatus & 0x00000002) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_DISPATCH_QUERY"); + nv_wr32(dev, 0x400500, 0); + if (nv_rd32(dev, 0x40084c) & 0x80000000) { + if (display) { + if (nouveau_graph_trapped_channel(dev, &trap.channel)) + trap.channel = -1; + trap.class = nv_rd32(dev, 0x400814); + trap.mthd = nv_rd32(dev, 0x40084c) & 0x1ffc; + trap.subc = (nv_rd32(dev, 0x40084c) >> 16) & 0x7; + trap.data = nv_rd32(dev, 0x40085c); + trap.data2 = 0; + nouveau_graph_dump_trap_info(dev, + "PGRAPH_TRAP_DISPATCH_QUERY", &trap); + NV_INFO(dev, "PGRAPH_TRAP_DISPATCH_QUERY - 40084c: %08x\n", nv_rd32(dev, 0x40084c)); + } + nv_wr32(dev, 0x40084c, 0); + } else if (display) { + NV_INFO(dev, "PGRAPH_TRAP_DISPATCH_QUERY - No stuck command?\n"); + } + ustatus &= ~0x00000002; + } + if (ustatus && display) + NV_INFO(dev, "PGRAPH_TRAP_DISPATCH - Unhandled ustatus 0x%08x\n", ustatus); + nv_wr32(dev, 0x400804, 0xc0000000); + nv_wr32(dev, 0x400108, 0x001); + status &= ~0x001; + } + + /* TRAPs other than dispatch use the "normal" trap regs. */ + if (status && display) { + nouveau_graph_trap_info(dev, &trap); + nouveau_graph_dump_trap_info(dev, + "PGRAPH_TRAP", &trap); + } + + /* M2MF: Memory to memory copy engine. */ + if (status & 0x002) { + ustatus = nv_rd32(dev, 0x406800) & 0x7fffffff; + if (!ustatus && display) { + NV_INFO(dev, "PGRAPH_TRAP_M2MF - no ustatus?\n"); + } + if (ustatus & 0x00000001) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_M2MF_NOTIFY"); + ustatus &= ~0x00000001; + } + if (ustatus & 0x00000002) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_M2MF_IN"); + ustatus &= ~0x00000002; + } + if (ustatus & 0x00000004) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_M2MF_OUT"); + ustatus &= ~0x00000004; + } + NV_INFO (dev, "PGRAPH_TRAP_M2MF - %08x %08x %08x %08x\n", + nv_rd32(dev, 0x406804), + nv_rd32(dev, 0x406808), + nv_rd32(dev, 0x40680c), + nv_rd32(dev, 0x406810)); + if (ustatus && display) + NV_INFO(dev, "PGRAPH_TRAP_M2MF - Unhandled ustatus 0x%08x\n", ustatus); + /* No sane way found yet -- just reset the bugger. */ + nv_wr32(dev, 0x400040, 2); + nv_wr32(dev, 0x400040, 0); + nv_wr32(dev, 0x406800, 0xc0000000); + nv_wr32(dev, 0x400108, 0x002); + status &= ~0x002; + } + + /* VFETCH: Fetches data from vertex buffers. */ + if (status & 0x004) { + ustatus = nv_rd32(dev, 0x400c04) & 0x7fffffff; + if (!ustatus && display) { + NV_INFO(dev, "PGRAPH_TRAP_VFETCH - no ustatus?\n"); + } + if (ustatus & 0x00000001) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_VFETCH_FAULT"); + NV_INFO (dev, "PGRAPH_TRAP_VFETCH_FAULT - %08x %08x %08x %08x\n", + nv_rd32(dev, 0x400c00), + nv_rd32(dev, 0x400c08), + nv_rd32(dev, 0x400c0c), + nv_rd32(dev, 0x400c10)); + ustatus &= ~0x00000001; + } + if (ustatus && display) + NV_INFO(dev, "PGRAPH_TRAP_VFETCH - Unhandled ustatus 0x%08x\n", ustatus); + nv_wr32(dev, 0x400c04, 0xc0000000); + nv_wr32(dev, 0x400108, 0x004); + status &= ~0x004; + } + + /* STRMOUT: DirectX streamout / OpenGL transform feedback. */ + if (status & 0x008) { + ustatus = nv_rd32(dev, 0x401800) & 0x7fffffff; + if (!ustatus && display) { + NV_INFO(dev, "PGRAPH_TRAP_STRMOUT - no ustatus?\n"); + } + if (ustatus & 0x00000001) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_STRMOUT_FAULT"); + NV_INFO (dev, "PGRAPH_TRAP_STRMOUT_FAULT - %08x %08x %08x %08x\n", + nv_rd32(dev, 0x401804), + nv_rd32(dev, 0x401808), + nv_rd32(dev, 0x40180c), + nv_rd32(dev, 0x401810)); + ustatus &= ~0x00000001; + } + if (ustatus && display) + NV_INFO(dev, "PGRAPH_TRAP_STRMOUT - Unhandled ustatus 0x%08x\n", ustatus); + /* No sane way found yet -- just reset the bugger. */ + nv_wr32(dev, 0x400040, 0x80); + nv_wr32(dev, 0x400040, 0); + nv_wr32(dev, 0x401800, 0xc0000000); + nv_wr32(dev, 0x400108, 0x008); + status &= ~0x008; + } + + /* CCACHE: Handles code and c[] caches and fills them. */ + if (status & 0x010) { + ustatus = nv_rd32(dev, 0x405018) & 0x7fffffff; + if (!ustatus && display) { + NV_INFO(dev, "PGRAPH_TRAP_CCACHE - no ustatus?\n"); + } + if (ustatus & 0x00000001) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_CCACHE_FAULT"); + NV_INFO (dev, "PGRAPH_TRAP_CCACHE_FAULT - %08x %08x %08x %08x %08x %08x %08x\n", + nv_rd32(dev, 0x405800), + nv_rd32(dev, 0x405804), + nv_rd32(dev, 0x405808), + nv_rd32(dev, 0x40580c), + nv_rd32(dev, 0x405810), + nv_rd32(dev, 0x405814), + nv_rd32(dev, 0x40581c)); + ustatus &= ~0x00000001; + } + if (ustatus && display) + NV_INFO(dev, "PGRAPH_TRAP_CCACHE - Unhandled ustatus 0x%08x\n", ustatus); + nv_wr32(dev, 0x405018, 0xc0000000); + nv_wr32(dev, 0x400108, 0x010); + status &= ~0x010; + } + + /* Unknown, not seen yet... 0x402000 is the only trap status reg + * remaining, so try to handle it anyway. Perhaps related to that + * unknown DMA slot on tesla? */ + if (status & 0x20) { + nv50_pfb_vm_trap(dev, display, "PGRAPH_TRAP_UNKC04"); + ustatus = nv_rd32(dev, 0x402000) & 0x7fffffff; + if (display) + NV_INFO(dev, "PGRAPH_TRAP_UNKC04 - Unhandled ustatus 0x%08x\n", ustatus); + nv_wr32(dev, 0x402000, 0xc0000000); + /* no status modifiction on purpose */ + } + + /* TEXTURE: CUDA texturing units */ + if (status & 0x040) { + nv50_pgraph_tp_trap (dev, 6, 0x408900, 0x408600, display, + "PGRAPH_TRAP_TEXTURE"); + nv_wr32(dev, 0x400108, 0x040); + status &= ~0x040; + } + + /* MP: CUDA execution engines. */ + if (status & 0x080) { + nv50_pgraph_tp_trap (dev, 7, 0x408314, 0x40831c, display, + "PGRAPH_TRAP_MP"); + nv_wr32(dev, 0x400108, 0x080); + status &= ~0x080; + } + + /* TPDMA: Handles TP-initiated uncached memory accesses: + * l[], g[], stack, 2d surfaces, render targets. */ + if (status & 0x100) { + nv50_pgraph_tp_trap (dev, 8, 0x408e08, 0x408708, display, + "PGRAPH_TRAP_TPDMA"); + nv_wr32(dev, 0x400108, 0x100); + status &= ~0x100; + } + + if (status) { + if (display) + NV_INFO(dev, "PGRAPH_TRAP - Unknown trap 0x%08x\n", + status); + nv_wr32(dev, 0x400108, status); + } +} + +/* There must be a *lot* of these. Will take some time to gather them up. */ +static struct nouveau_enum_names nv50_data_error_names[] = +{ + { 4, "INVALID_VALUE" }, + { 5, "INVALID_ENUM" }, + { 8, "INVALID_OBJECT" }, + { 0xc, "INVALID_BITFIELD" }, + { 0x28, "MP_NO_REG_SPACE" }, + { 0x2b, "MP_BLOCK_SIZE_MISMATCH" }, +}; + +static void nv50_pgraph_irq_handler(struct drm_device *dev) { + struct nouveau_pgraph_trap trap; + int unhandled = 0; uint32_t status; while ((status = nv_rd32(dev, NV03_PGRAPH_INTR))) { - uint32_t nsource = nv_rd32(dev, NV03_PGRAPH_NSOURCE); - + /* NOTIFY: You've set a NOTIFY an a command and it's done. */ if (status & 0x00000001) { - nouveau_pgraph_intr_notify(dev, nsource); + nouveau_graph_trap_info(dev, &trap); + if (nouveau_ratelimit()) + nouveau_graph_dump_trap_info(dev, + "PGRAPH_NOTIFY", &trap); status &= ~0x00000001; nv_wr32(dev, NV03_PGRAPH_INTR, 0x00000001); } - if (status & 0x00000010) { - nouveau_pgraph_intr_error(dev, nsource | - NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD); + /* COMPUTE_QUERY: Purpose and exact cause unknown, happens + * when you write 0x200 to 0x50c0 method 0x31c. */ + if (status & 0x00000002) { + nouveau_graph_trap_info(dev, &trap); + if (nouveau_ratelimit()) + nouveau_graph_dump_trap_info(dev, + "PGRAPH_COMPUTE_QUERY", &trap); + status &= ~0x00000002; + nv_wr32(dev, NV03_PGRAPH_INTR, 0x00000002); + } + + /* Unknown, never seen: 0x4 */ + /* ILLEGAL_MTHD: You used a wrong method for this class. */ + if (status & 0x00000010) { + nouveau_graph_trap_info(dev, &trap); + if (nouveau_pgraph_intr_swmthd(dev, &trap)) + unhandled = 1; + if (unhandled && nouveau_ratelimit()) + nouveau_graph_dump_trap_info(dev, + "PGRAPH_ILLEGAL_MTHD", &trap); status &= ~0x00000010; nv_wr32(dev, NV03_PGRAPH_INTR, 0x00000010); } + /* ILLEGAL_CLASS: You used a wrong class. */ + if (status & 0x00000020) { + nouveau_graph_trap_info(dev, &trap); + if (nouveau_ratelimit()) + nouveau_graph_dump_trap_info(dev, + "PGRAPH_ILLEGAL_CLASS", &trap); + status &= ~0x00000020; + nv_wr32(dev, NV03_PGRAPH_INTR, 0x00000020); + } + + /* DOUBLE_NOTIFY: You tried to set a NOTIFY on another NOTIFY. */ + if (status & 0x00000040) { + nouveau_graph_trap_info(dev, &trap); + if (nouveau_ratelimit()) + nouveau_graph_dump_trap_info(dev, + "PGRAPH_DOUBLE_NOTIFY", &trap); + status &= ~0x00000040; + nv_wr32(dev, NV03_PGRAPH_INTR, 0x00000040); + } + + /* CONTEXT_SWITCH: PGRAPH needs us to load a new context */ if (status & 0x00001000) { nv_wr32(dev, 0x400500, 0x00000000); nv_wr32(dev, NV03_PGRAPH_INTR, @@ -613,49 +1116,59 @@ nv50_pgraph_irq_handler(struct drm_device *dev) status &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; } - if (status & 0x00100000) { - nouveau_pgraph_intr_error(dev, nsource | - NV03_PGRAPH_NSOURCE_DATA_ERROR); + /* BUFFER_NOTIFY: Your m2mf transfer finished */ + if (status & 0x00010000) { + nouveau_graph_trap_info(dev, &trap); + if (nouveau_ratelimit()) + nouveau_graph_dump_trap_info(dev, + "PGRAPH_BUFFER_NOTIFY", &trap); + status &= ~0x00010000; + nv_wr32(dev, NV03_PGRAPH_INTR, 0x00010000); + } + /* DATA_ERROR: Invalid value for this method, or invalid + * state in current PGRAPH context for this operation */ + if (status & 0x00100000) { + nouveau_graph_trap_info(dev, &trap); + if (nouveau_ratelimit()) { + nouveau_graph_dump_trap_info(dev, + "PGRAPH_DATA_ERROR", &trap); + NV_INFO (dev, "PGRAPH_DATA_ERROR - "); + nouveau_print_enum_names(nv_rd32(dev, 0x400110), + nv50_data_error_names); + printk("\n"); + } status &= ~0x00100000; nv_wr32(dev, NV03_PGRAPH_INTR, 0x00100000); } + /* TRAP: Something bad happened in the middle of command + * execution. Has a billion types, subtypes, and even + * subsubtypes. */ if (status & 0x00200000) { - int r; - - nouveau_pgraph_intr_error(dev, nsource | - NV03_PGRAPH_NSOURCE_PROTECTION_ERROR); - - NV_ERROR(dev, "magic set 1:\n"); - for (r = 0x408900; r <= 0x408910; r += 4) - NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r, - nv_rd32(dev, r)); - nv_wr32(dev, 0x408900, - nv_rd32(dev, 0x408904) | 0xc0000000); - for (r = 0x408e08; r <= 0x408e24; r += 4) - NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r, - nv_rd32(dev, r)); - nv_wr32(dev, 0x408e08, - nv_rd32(dev, 0x408e08) | 0xc0000000); - - NV_ERROR(dev, "magic set 2:\n"); - for (r = 0x409900; r <= 0x409910; r += 4) - NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r, - nv_rd32(dev, r)); - nv_wr32(dev, 0x409900, - nv_rd32(dev, 0x409904) | 0xc0000000); - for (r = 0x409e08; r <= 0x409e24; r += 4) - NV_ERROR(dev, "\t0x%08x: 0x%08x\n", r, - nv_rd32(dev, r)); - nv_wr32(dev, 0x409e08, - nv_rd32(dev, 0x409e08) | 0xc0000000); - + nv50_pgraph_trap_handler(dev); status &= ~0x00200000; - nv_wr32(dev, NV03_PGRAPH_NSOURCE, nsource); nv_wr32(dev, NV03_PGRAPH_INTR, 0x00200000); } + /* Unknown, never seen: 0x00400000 */ + + /* SINGLE_STEP: Happens on every method if you turned on + * single stepping in 40008c */ + if (status & 0x01000000) { + nouveau_graph_trap_info(dev, &trap); + if (nouveau_ratelimit()) + nouveau_graph_dump_trap_info(dev, + "PGRAPH_SINGLE_STEP", &trap); + status &= ~0x01000000; + nv_wr32(dev, NV03_PGRAPH_INTR, 0x01000000); + } + + /* 0x02000000 happens when you pause a ctxprog... + * but the only way this can happen that I know is by + * poking the relevant MMIO register, and we don't + * do that. */ + if (status) { NV_INFO(dev, "Unhandled PGRAPH_INTR - 0x%08x\n", status); @@ -672,7 +1185,8 @@ nv50_pgraph_irq_handler(struct drm_device *dev) } nv_wr32(dev, NV03_PMC_INTR_0, NV_PMC_INTR_0_PGRAPH_PENDING); - nv_wr32(dev, 0x400824, nv_rd32(dev, 0x400824) & ~(1 << 31)); + if (nv_rd32(dev, 0x400824) & (1 << 31)) + nv_wr32(dev, 0x400824, nv_rd32(dev, 0x400824) & ~(1 << 31)); } static void @@ -691,11 +1205,14 @@ nouveau_irq_handler(DRM_IRQ_ARGS) struct drm_device *dev = (struct drm_device *)arg; struct drm_nouveau_private *dev_priv = dev->dev_private; uint32_t status, fbdev_flags = 0; + unsigned long flags; status = nv_rd32(dev, NV03_PMC_INTR_0); if (!status) return IRQ_NONE; + spin_lock_irqsave(&dev_priv->context_switch_lock, flags); + if (dev_priv->fbdev_info) { fbdev_flags = dev_priv->fbdev_info->flags; dev_priv->fbdev_info->flags |= FBINFO_HWACCEL_DISABLED; @@ -733,5 +1250,7 @@ nouveau_irq_handler(DRM_IRQ_ARGS) if (dev_priv->fbdev_info) dev_priv->fbdev_info->flags = fbdev_flags; + spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags); + return IRQ_HANDLED; } diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.c b/drivers/gpu/drm/nouveau/nouveau_mem.c index 2dc09db..775a701 100644 --- a/drivers/gpu/drm/nouveau/nouveau_mem.c +++ b/drivers/gpu/drm/nouveau/nouveau_mem.c @@ -347,6 +347,20 @@ nv50_mem_vm_bind_linear(struct drm_device *dev, uint64_t virt, uint32_t size, return -EBUSY; } + nv_wr32(dev, 0x100c80, 0x00040001); + if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) { + NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n"); + NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80)); + return -EBUSY; + } + + nv_wr32(dev, 0x100c80, 0x00060001); + if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) { + NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n"); + NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80)); + return -EBUSY; + } + return 0; } @@ -387,6 +401,20 @@ nv50_mem_vm_unbind(struct drm_device *dev, uint64_t virt, uint32_t size) if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) { NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n"); NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80)); + return; + } + + nv_wr32(dev, 0x100c80, 0x00040001); + if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) { + NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n"); + NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80)); + return; + } + + nv_wr32(dev, 0x100c80, 0x00060001); + if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) { + NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n"); + NV_ERROR(dev, "0x100c80 = 0x%08x\n", nv_rd32(dev, 0x100c80)); } } @@ -449,9 +477,30 @@ void nouveau_mem_close(struct drm_device *dev) } } -/*XXX won't work on BSD because of pci_read_config_dword */ static uint32_t -nouveau_mem_fb_amount_igp(struct drm_device *dev) +nouveau_mem_detect_nv04(struct drm_device *dev) +{ + uint32_t boot0 = nv_rd32(dev, NV03_BOOT_0); + + if (boot0 & 0x00000100) + return (((boot0 >> 12) & 0xf) * 2 + 2) * 1024 * 1024; + + switch (boot0 & NV03_BOOT_0_RAM_AMOUNT) { + case NV04_BOOT_0_RAM_AMOUNT_32MB: + return 32 * 1024 * 1024; + case NV04_BOOT_0_RAM_AMOUNT_16MB: + return 16 * 1024 * 1024; + case NV04_BOOT_0_RAM_AMOUNT_8MB: + return 8 * 1024 * 1024; + case NV04_BOOT_0_RAM_AMOUNT_4MB: + return 4 * 1024 * 1024; + } + + return 0; +} + +static uint32_t +nouveau_mem_detect_nforce(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; struct pci_dev *bridge; @@ -463,11 +512,11 @@ nouveau_mem_fb_amount_igp(struct drm_device *dev) return 0; } - if (dev_priv->flags&NV_NFORCE) { + if (dev_priv->flags & NV_NFORCE) { pci_read_config_dword(bridge, 0x7C, &mem); return (uint64_t)(((mem >> 6) & 31) + 1)*1024*1024; } else - if (dev_priv->flags&NV_NFORCE2) { + if (dev_priv->flags & NV_NFORCE2) { pci_read_config_dword(bridge, 0x84, &mem); return (uint64_t)(((mem >> 4) & 127) + 1)*1024*1024; } @@ -477,50 +526,32 @@ nouveau_mem_fb_amount_igp(struct drm_device *dev) } /* returns the amount of FB ram in bytes */ -uint64_t nouveau_mem_fb_amount(struct drm_device *dev) +int +nouveau_mem_detect(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - uint32_t boot0; - - switch (dev_priv->card_type) { - case NV_04: - boot0 = nv_rd32(dev, NV03_BOOT_0); - if (boot0 & 0x00000100) - return (((boot0 >> 12) & 0xf) * 2 + 2) * 1024 * 1024; - - switch (boot0 & NV03_BOOT_0_RAM_AMOUNT) { - case NV04_BOOT_0_RAM_AMOUNT_32MB: - return 32 * 1024 * 1024; - case NV04_BOOT_0_RAM_AMOUNT_16MB: - return 16 * 1024 * 1024; - case NV04_BOOT_0_RAM_AMOUNT_8MB: - return 8 * 1024 * 1024; - case NV04_BOOT_0_RAM_AMOUNT_4MB: - return 4 * 1024 * 1024; - } - break; - case NV_10: - case NV_20: - case NV_30: - case NV_40: - case NV_50: - default: - if (dev_priv->flags & (NV_NFORCE | NV_NFORCE2)) { - return nouveau_mem_fb_amount_igp(dev); - } else { - uint64_t mem; - mem = (nv_rd32(dev, NV04_FIFO_DATA) & - NV10_FIFO_DATA_RAM_AMOUNT_MB_MASK) >> - NV10_FIFO_DATA_RAM_AMOUNT_MB_SHIFT; - return mem * 1024 * 1024; - } - break; + + if (dev_priv->card_type == NV_04) { + dev_priv->vram_size = nouveau_mem_detect_nv04(dev); + } else + if (dev_priv->flags & (NV_NFORCE | NV_NFORCE2)) { + dev_priv->vram_size = nouveau_mem_detect_nforce(dev); + } else { + dev_priv->vram_size = nv_rd32(dev, NV04_FIFO_DATA); + dev_priv->vram_size &= NV10_FIFO_DATA_RAM_AMOUNT_MB_MASK; + if (dev_priv->chipset == 0xaa || dev_priv->chipset == 0xac) + dev_priv->vram_sys_base = nv_rd32(dev, 0x100e10) << 12; } - NV_ERROR(dev, - "Unable to detect video ram size. Please report your setup to " - DRIVER_EMAIL "\n"); - return 0; + NV_INFO(dev, "Detected %dMiB VRAM\n", (int)(dev_priv->vram_size >> 20)); + if (dev_priv->vram_sys_base) { + NV_INFO(dev, "Stolen system memory at: 0x%010llx\n", + dev_priv->vram_sys_base); + } + + if (dev_priv->vram_size) + return 0; + return -ENOMEM; } #if __OS_HAS_AGP @@ -631,15 +662,12 @@ nouveau_mem_init(struct drm_device *dev) 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); - + dev_priv->fb_available_size = dev_priv->vram_size; dev_priv->fb_mappable_pages = dev_priv->fb_available_size; if (dev_priv->fb_mappable_pages > drm_get_resource_len(dev, 1)) dev_priv->fb_mappable_pages = drm_get_resource_len(dev, 1); dev_priv->fb_mappable_pages >>= PAGE_SHIFT; - NV_INFO(dev, "%d MiB VRAM\n", (int)(dev_priv->fb_available_size >> 20)); - /* remove reserved space at end of vram from available amount */ dev_priv->fb_available_size -= dev_priv->ramin_rsvd_vram; dev_priv->fb_aper_free = dev_priv->fb_available_size; diff --git a/drivers/gpu/drm/nouveau/nouveau_reg.h b/drivers/gpu/drm/nouveau/nouveau_reg.h index aa9b310..6ca80a3 100644 --- a/drivers/gpu/drm/nouveau/nouveau_reg.h +++ b/drivers/gpu/drm/nouveau/nouveau_reg.h @@ -826,6 +826,7 @@ #define NV50_SOR_DP_CTRL_TRAINING_PATTERN_2 0x02000000 #define NV50_SOR_DP_UNK118(i,l) (0x0061c118 + (i) * 0x800 + (l) * 0x80) #define NV50_SOR_DP_UNK120(i,l) (0x0061c120 + (i) * 0x800 + (l) * 0x80) +#define NV50_SOR_DP_UNK128(i,l) (0x0061c128 + (i) * 0x800 + (l) * 0x80) #define NV50_SOR_DP_UNK130(i,l) (0x0061c130 + (i) * 0x800 + (l) * 0x80) #define NV50_PDISPLAY_USER(i) ((i) * 0x1000 + 0x00640000) diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c index ed15905..554fb45 100644 --- a/drivers/gpu/drm/nouveau/nouveau_sgdma.c +++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c @@ -171,6 +171,24 @@ nouveau_sgdma_unbind(struct ttm_backend *be) } dev_priv->engine.instmem.finish_access(nvbe->dev); + if (dev_priv->card_type == NV_50) { + nv_wr32(dev, 0x100c80, 0x00050001); + if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) { + NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n"); + NV_ERROR(dev, "0x100c80 = 0x%08x\n", + nv_rd32(dev, 0x100c80)); + return -EBUSY; + } + + nv_wr32(dev, 0x100c80, 0x00000001); + if (!nv_wait(0x100c80, 0x00000001, 0x00000000)) { + NV_ERROR(dev, "timeout: (0x100c80 & 1) == 0 (2)\n"); + NV_ERROR(dev, "0x100c80 = 0x%08x\n", + nv_rd32(dev, 0x100c80)); + return -EBUSY; + } + } + nvbe->bound = false; return 0; } diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c index a8d77c8..a11cd7f 100644 --- a/drivers/gpu/drm/nouveau/nouveau_state.c +++ b/drivers/gpu/drm/nouveau/nouveau_state.c @@ -34,7 +34,6 @@ #include "nouveau_drm.h" #include "nv50_display.h" -static int nouveau_stub_init(struct drm_device *dev) { return 0; } static void nouveau_stub_takedown(struct drm_device *dev) {} static int nouveau_init_engine_ptrs(struct drm_device *dev) @@ -276,8 +275,8 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev) engine->timer.init = nv04_timer_init; engine->timer.read = nv04_timer_read; engine->timer.takedown = nv04_timer_takedown; - engine->fb.init = nouveau_stub_init; - engine->fb.takedown = nouveau_stub_takedown; + engine->fb.init = nv50_fb_init; + engine->fb.takedown = nv50_fb_takedown; engine->graph.grclass = nv50_graph_grclass; engine->graph.init = nv50_graph_init; engine->graph.takedown = nv50_graph_takedown; @@ -340,7 +339,7 @@ nouveau_card_init_channel(struct drm_device *dev) gpuobj = NULL; ret = nouveau_gpuobj_dma_new(dev_priv->channel, NV_CLASS_DMA_IN_MEMORY, - 0, nouveau_mem_fb_amount(dev), + 0, dev_priv->vram_size, NV_DMA_ACCESS_RW, NV_DMA_TARGET_VIDMEM, &gpuobj); if (ret) @@ -391,6 +390,7 @@ nouveau_card_init(struct drm_device *dev) goto out; engine = &dev_priv->engine; dev_priv->init_state = NOUVEAU_CARD_INIT_FAILED; + spin_lock_init(&dev_priv->context_switch_lock); /* Parse BIOS tables / Run init tables if card not POSTed */ if (drm_core_check_feature(dev, DRIVER_MODESET)) { @@ -399,6 +399,10 @@ nouveau_card_init(struct drm_device *dev) goto out; } + ret = nouveau_mem_detect(dev); + if (ret) + goto out_bios; + ret = nouveau_gpuobj_early_init(dev); if (ret) goto out_bios; @@ -474,7 +478,7 @@ nouveau_card_init(struct drm_device *dev) else ret = nv04_display_create(dev); if (ret) - goto out_irq; + goto out_channel; } ret = nouveau_backlight_init(dev); @@ -488,6 +492,11 @@ nouveau_card_init(struct drm_device *dev) return 0; +out_channel: + if (dev_priv->channel) { + nouveau_channel_free(dev_priv->channel); + dev_priv->channel = NULL; + } out_irq: drm_irq_uninstall(dev); out_fifo: @@ -505,6 +514,7 @@ out_mc: out_gpuobj: nouveau_gpuobj_takedown(dev); out_mem: + nouveau_sgdma_takedown(dev); nouveau_mem_close(dev); out_instmem: engine->instmem.takedown(dev); diff --git a/drivers/gpu/drm/nouveau/nv04_crtc.c b/drivers/gpu/drm/nouveau/nv04_crtc.c index d2f143e..9986aba 100644 --- a/drivers/gpu/drm/nouveau/nv04_crtc.c +++ b/drivers/gpu/drm/nouveau/nv04_crtc.c @@ -230,9 +230,9 @@ nv_crtc_mode_set_vga(struct drm_crtc *crtc, struct drm_display_mode *mode) struct drm_framebuffer *fb = crtc->fb; /* Calculate our timings */ - int horizDisplay = (mode->crtc_hdisplay >> 3) - 1; - int horizStart = (mode->crtc_hsync_start >> 3) - 1; - int horizEnd = (mode->crtc_hsync_end >> 3) - 1; + int horizDisplay = (mode->crtc_hdisplay >> 3) - 1; + int horizStart = (mode->crtc_hsync_start >> 3) + 1; + int horizEnd = (mode->crtc_hsync_end >> 3) + 1; int horizTotal = (mode->crtc_htotal >> 3) - 5; int horizBlankStart = (mode->crtc_hdisplay >> 3) - 1; int horizBlankEnd = (mode->crtc_htotal >> 3) - 1; diff --git a/drivers/gpu/drm/nouveau/nv04_dac.c b/drivers/gpu/drm/nouveau/nv04_dac.c index 1d73b15..1cb19e3 100644 --- a/drivers/gpu/drm/nouveau/nv04_dac.c +++ b/drivers/gpu/drm/nouveau/nv04_dac.c @@ -230,13 +230,13 @@ uint32_t nv17_dac_sample_load(struct drm_encoder *encoder) if (dcb->type == OUTPUT_TV) { testval = RGB_TEST_DATA(0xa0, 0xa0, 0xa0); - if (dev_priv->vbios->tvdactestval) - testval = dev_priv->vbios->tvdactestval; + if (dev_priv->vbios.tvdactestval) + testval = dev_priv->vbios.tvdactestval; } else { testval = RGB_TEST_DATA(0x140, 0x140, 0x140); /* 0x94050140 */ - if (dev_priv->vbios->dactestval) - testval = dev_priv->vbios->dactestval; + if (dev_priv->vbios.dactestval) + testval = dev_priv->vbios.dactestval; } saved_rtest_ctrl = NVReadRAMDAC(dev, 0, NV_PRAMDAC_TEST_CONTROL + regoffset); diff --git a/drivers/gpu/drm/nouveau/nv04_dfp.c b/drivers/gpu/drm/nouveau/nv04_dfp.c index 483f875..41634d4 100644 --- a/drivers/gpu/drm/nouveau/nv04_dfp.c +++ b/drivers/gpu/drm/nouveau/nv04_dfp.c @@ -269,10 +269,10 @@ static void nv04_dfp_mode_set(struct drm_encoder *encoder, regp->fp_horiz_regs[FP_TOTAL] = output_mode->htotal - 1; if (!nv_gf4_disp_arch(dev) || (output_mode->hsync_start - output_mode->hdisplay) >= - dev_priv->vbios->digital_min_front_porch) + dev_priv->vbios.digital_min_front_porch) regp->fp_horiz_regs[FP_CRTC] = output_mode->hdisplay; else - regp->fp_horiz_regs[FP_CRTC] = output_mode->hsync_start - dev_priv->vbios->digital_min_front_porch - 1; + regp->fp_horiz_regs[FP_CRTC] = output_mode->hsync_start - dev_priv->vbios.digital_min_front_porch - 1; regp->fp_horiz_regs[FP_SYNC_START] = output_mode->hsync_start - 1; regp->fp_horiz_regs[FP_SYNC_END] = output_mode->hsync_end - 1; regp->fp_horiz_regs[FP_VALID_START] = output_mode->hskew; diff --git a/drivers/gpu/drm/nouveau/nv04_display.c b/drivers/gpu/drm/nouveau/nv04_display.c index ef77215..c7898b4 100644 --- a/drivers/gpu/drm/nouveau/nv04_display.c +++ b/drivers/gpu/drm/nouveau/nv04_display.c @@ -93,10 +93,9 @@ int nv04_display_create(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct parsed_dcb *dcb = dev_priv->vbios->dcb; + struct dcb_table *dcb = &dev_priv->vbios.dcb; struct drm_encoder *encoder; struct drm_crtc *crtc; - uint16_t connector[16] = { 0 }; int i, ret; NV_DEBUG_KMS(dev, "\n"); @@ -154,52 +153,10 @@ nv04_display_create(struct drm_device *dev) if (ret) continue; - - connector[dcbent->connector] |= (1 << dcbent->type); } - for (i = 0; i < dcb->entries; i++) { - struct dcb_entry *dcbent = &dcb->entry[i]; - uint16_t encoders; - int type; - - encoders = connector[dcbent->connector]; - if (!(encoders & (1 << dcbent->type))) - continue; - connector[dcbent->connector] = 0; - - switch (dcbent->type) { - case OUTPUT_ANALOG: - if (!MULTIPLE_ENCODERS(encoders)) - type = DRM_MODE_CONNECTOR_VGA; - else - type = DRM_MODE_CONNECTOR_DVII; - break; - case OUTPUT_TMDS: - if (!MULTIPLE_ENCODERS(encoders)) - type = DRM_MODE_CONNECTOR_DVID; - else - type = DRM_MODE_CONNECTOR_DVII; - break; - case OUTPUT_LVDS: - type = DRM_MODE_CONNECTOR_LVDS; -#if 0 - /* don't create i2c adapter when lvds ddc not allowed */ - if (dcbent->lvdsconf.use_straps_for_mode || - dev_priv->vbios->fp_no_ddc) - i2c_index = 0xf; -#endif - break; - case OUTPUT_TV: - type = DRM_MODE_CONNECTOR_TV; - break; - default: - type = DRM_MODE_CONNECTOR_Unknown; - continue; - } - - nouveau_connector_create(dev, dcbent->connector, type); - } + for (i = 0; i < dcb->connector.entries; i++) + nouveau_connector_create(dev, &dcb->connector.entry[i]); /* Save previous state */ NVLockVgaCrtcs(dev, false); diff --git a/drivers/gpu/drm/nouveau/nv04_fbcon.c b/drivers/gpu/drm/nouveau/nv04_fbcon.c index fd01caa..813b25c 100644 --- a/drivers/gpu/drm/nouveau/nv04_fbcon.c +++ b/drivers/gpu/drm/nouveau/nv04_fbcon.c @@ -118,8 +118,8 @@ nv04_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) return; } - width = (image->width + 31) & ~31; - dsize = (width * image->height) >> 5; + width = ALIGN(image->width, 8); + dsize = ALIGN(width * image->height, 32) >> 5; if (info->fix.visual == FB_VISUAL_TRUECOLOR || info->fix.visual == FB_VISUAL_DIRECTCOLOR) { @@ -136,8 +136,8 @@ nv04_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) ((image->dx + image->width) & 0xffff)); OUT_RING(chan, bg); OUT_RING(chan, fg); - OUT_RING(chan, (image->height << 16) | image->width); OUT_RING(chan, (image->height << 16) | width); + OUT_RING(chan, (image->height << 16) | image->width); OUT_RING(chan, (image->dy << 16) | (image->dx & 0xffff)); while (dsize) { diff --git a/drivers/gpu/drm/nouveau/nv04_fifo.c b/drivers/gpu/drm/nouveau/nv04_fifo.c index f31347b..66fe559 100644 --- a/drivers/gpu/drm/nouveau/nv04_fifo.c +++ b/drivers/gpu/drm/nouveau/nv04_fifo.c @@ -117,6 +117,7 @@ nv04_fifo_create_context(struct nouveau_channel *chan) { struct drm_device *dev = chan->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; + unsigned long flags; int ret; ret = nouveau_gpuobj_new_fake(dev, NV04_RAMFC(chan->id), ~0, @@ -127,6 +128,8 @@ nv04_fifo_create_context(struct nouveau_channel *chan) if (ret) return ret; + spin_lock_irqsave(&dev_priv->context_switch_lock, flags); + /* Setup initial state */ dev_priv->engine.instmem.prepare_access(dev, true); RAMFC_WR(DMA_PUT, chan->pushbuf_base); @@ -144,6 +147,8 @@ nv04_fifo_create_context(struct nouveau_channel *chan) /* enable the fifo dma operation */ nv_wr32(dev, NV04_PFIFO_MODE, nv_rd32(dev, NV04_PFIFO_MODE) | (1 << chan->id)); + + spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags); return 0; } diff --git a/drivers/gpu/drm/nouveau/nv04_tv.c b/drivers/gpu/drm/nouveau/nv04_tv.c index 9c63099..c4e3404 100644 --- a/drivers/gpu/drm/nouveau/nv04_tv.c +++ b/drivers/gpu/drm/nouveau/nv04_tv.c @@ -262,7 +262,7 @@ int nv04_tv_create(struct drm_device *dev, struct dcb_entry *entry) nv_encoder->or = ffs(entry->or) - 1; /* Run the slave-specific initialization */ - adap = &dev_priv->vbios->dcb->i2c[i2c_index].chan->adapter; + adap = &dev_priv->vbios.dcb.i2c[i2c_index].chan->adapter; was_locked = NVLockVgaCrtcs(dev, false); diff --git a/drivers/gpu/drm/nouveau/nv17_tv.c b/drivers/gpu/drm/nouveau/nv17_tv.c index 21ac6e4..74c8803 100644 --- a/drivers/gpu/drm/nouveau/nv17_tv.c +++ b/drivers/gpu/drm/nouveau/nv17_tv.c @@ -45,8 +45,8 @@ static uint32_t nv42_tv_sample_load(struct drm_encoder *encoder) #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; + if (dev_priv->vbios.tvdactestval) + testval = dev_priv->vbios.tvdactestval; dacclk = NVReadRAMDAC(dev, 0, NV_PRAMDAC_DACCLK + regoffset); head = (dacclk & 0x100) >> 8; @@ -367,7 +367,7 @@ static void nv17_tv_prepare(struct drm_encoder *encoder) !enc->crtc && nv04_dfp_get_bound_head(dev, dcb) == head) { nv04_dfp_bind_head(dev, dcb, head ^ 1, - dev_priv->VBIOS.fp.dual_link); + dev_priv->vbios.fp.dual_link); } } diff --git a/drivers/gpu/drm/nouveau/nv40_fifo.c b/drivers/gpu/drm/nouveau/nv40_fifo.c index b4f19cc..500ccfd 100644 --- a/drivers/gpu/drm/nouveau/nv40_fifo.c +++ b/drivers/gpu/drm/nouveau/nv40_fifo.c @@ -37,6 +37,7 @@ nv40_fifo_create_context(struct nouveau_channel *chan) struct drm_device *dev = chan->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; uint32_t fc = NV40_RAMFC(chan->id); + unsigned long flags; int ret; ret = nouveau_gpuobj_new_fake(dev, NV40_RAMFC(chan->id), ~0, @@ -45,6 +46,8 @@ nv40_fifo_create_context(struct nouveau_channel *chan) if (ret) return ret; + spin_lock_irqsave(&dev_priv->context_switch_lock, flags); + dev_priv->engine.instmem.prepare_access(dev, true); nv_wi32(dev, fc + 0, chan->pushbuf_base); nv_wi32(dev, fc + 4, chan->pushbuf_base); @@ -63,6 +66,8 @@ nv40_fifo_create_context(struct nouveau_channel *chan) /* enable the fifo dma operation */ nv_wr32(dev, NV04_PFIFO_MODE, nv_rd32(dev, NV04_PFIFO_MODE) | (1 << chan->id)); + + spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags); return 0; } @@ -273,7 +278,7 @@ nv40_fifo_init_ramxx(struct drm_device *dev) default: nv_wr32(dev, 0x2230, 0); nv_wr32(dev, NV40_PFIFO_RAMFC, - ((nouveau_mem_fb_amount(dev) - 512 * 1024 + + ((dev_priv->vram_size - 512 * 1024 + dev_priv->ramfc_offset) >> 16) | (3 << 16)); break; } diff --git a/drivers/gpu/drm/nouveau/nv40_graph.c b/drivers/gpu/drm/nouveau/nv40_graph.c index 53e8afe..0616c96 100644 --- a/drivers/gpu/drm/nouveau/nv40_graph.c +++ b/drivers/gpu/drm/nouveau/nv40_graph.c @@ -335,6 +335,27 @@ nv40_graph_init(struct drm_device *dev) nv_wr32(dev, 0x400b38, 0x2ffff800); nv_wr32(dev, 0x400b3c, 0x00006000); + /* Tiling related stuff. */ + switch (dev_priv->chipset) { + case 0x44: + case 0x4a: + nv_wr32(dev, 0x400bc4, 0x1003d888); + nv_wr32(dev, 0x400bbc, 0xb7a7b500); + break; + case 0x46: + nv_wr32(dev, 0x400bc4, 0x0000e024); + nv_wr32(dev, 0x400bbc, 0xb7a7b520); + break; + case 0x4c: + case 0x4e: + case 0x67: + nv_wr32(dev, 0x400bc4, 0x1003d888); + nv_wr32(dev, 0x400bbc, 0xb7a7b540); + break; + default: + 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); diff --git a/drivers/gpu/drm/nouveau/nv50_calc.c b/drivers/gpu/drm/nouveau/nv50_calc.c new file mode 100644 index 0000000..90f3f6b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nv50_calc.c @@ -0,0 +1,88 @@ +/* + * Copyright 2010 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 "drmP.h" +#include "drm_fixed.h" +#include "nouveau_drv.h" +#include "nouveau_hw.h" + +int +nv50_calc_pll(struct drm_device *dev, struct pll_lims *pll, int clk, + int *N1, int *M1, int *N2, int *M2, int *P) +{ + struct nouveau_pll_vals pll_vals; + int ret; + + ret = nouveau_calc_pll_mnp(dev, pll, clk, &pll_vals); + if (ret <= 0) + return ret; + + *N1 = pll_vals.N1; + *M1 = pll_vals.M1; + *N2 = pll_vals.N2; + *M2 = pll_vals.M2; + *P = pll_vals.log2P; + return ret; +} + +int +nv50_calc_pll2(struct drm_device *dev, struct pll_lims *pll, int clk, + int *N, int *fN, int *M, int *P) +{ + fixed20_12 fb_div, a, b; + + *P = pll->vco1.maxfreq / clk; + if (*P > pll->max_p) + *P = pll->max_p; + if (*P < pll->min_p) + *P = pll->min_p; + + /* *M = ceil(refclk / pll->vco.max_inputfreq); */ + a.full = dfixed_const(pll->refclk); + b.full = dfixed_const(pll->vco1.max_inputfreq); + a.full = dfixed_div(a, b); + a.full = dfixed_ceil(a); + *M = dfixed_trunc(a); + + /* fb_div = (vco * *M) / refclk; */ + fb_div.full = dfixed_const(clk * *P); + fb_div.full = dfixed_mul(fb_div, a); + a.full = dfixed_const(pll->refclk); + fb_div.full = dfixed_div(fb_div, a); + + /* *N = floor(fb_div); */ + a.full = dfixed_floor(fb_div); + *N = dfixed_trunc(fb_div); + + /* *fN = (fmod(fb_div, 1.0) * 8192) - 4096; */ + b.full = dfixed_const(8192); + a.full = dfixed_mul(a, b); + fb_div.full = dfixed_mul(fb_div, b); + fb_div.full = fb_div.full - a.full; + *fN = dfixed_trunc(fb_div) - 4096; + *fN &= 0xffff; + + return clk; +} + diff --git a/drivers/gpu/drm/nouveau/nv50_crtc.c b/drivers/gpu/drm/nouveau/nv50_crtc.c index d1a651e..03d0e41 100644 --- a/drivers/gpu/drm/nouveau/nv50_crtc.c +++ b/drivers/gpu/drm/nouveau/nv50_crtc.c @@ -264,32 +264,40 @@ nv50_crtc_set_scale(struct nouveau_crtc *nv_crtc, int scaling_mode, bool update) int nv50_crtc_set_clock(struct drm_device *dev, int head, int pclk) { - uint32_t pll_reg = NV50_PDISPLAY_CRTC_CLK_CTRL1(head); - struct nouveau_pll_vals pll; - struct pll_lims limits; + uint32_t reg = NV50_PDISPLAY_CRTC_CLK_CTRL1(head); + struct pll_lims pll; uint32_t reg1, reg2; - int ret; + int ret, N1, M1, N2, M2, P; - ret = get_pll_limits(dev, pll_reg, &limits); + ret = get_pll_limits(dev, reg, &pll); if (ret) return ret; - ret = nouveau_calc_pll_mnp(dev, &limits, pclk, &pll); - if (ret <= 0) - return ret; + if (pll.vco2.maxfreq) { + ret = nv50_calc_pll(dev, &pll, pclk, &N1, &M1, &N2, &M2, &P); + if (ret <= 0) + return 0; + + NV_DEBUG(dev, "pclk %d out %d NM1 %d %d NM2 %d %d P %d\n", + pclk, ret, N1, M1, N2, M2, P); - if (limits.vco2.maxfreq) { - reg1 = nv_rd32(dev, pll_reg + 4) & 0xff00ff00; - reg2 = nv_rd32(dev, pll_reg + 8) & 0x8000ff00; - nv_wr32(dev, pll_reg, 0x10000611); - nv_wr32(dev, pll_reg + 4, reg1 | (pll.M1 << 16) | pll.N1); - nv_wr32(dev, pll_reg + 8, - reg2 | (pll.log2P << 28) | (pll.M2 << 16) | pll.N2); + reg1 = nv_rd32(dev, reg + 4) & 0xff00ff00; + reg2 = nv_rd32(dev, reg + 8) & 0x8000ff00; + nv_wr32(dev, reg, 0x10000611); + nv_wr32(dev, reg + 4, reg1 | (M1 << 16) | N1); + nv_wr32(dev, reg + 8, reg2 | (P << 28) | (M2 << 16) | N2); } else { - reg1 = nv_rd32(dev, pll_reg + 4) & 0xffc00000; - nv_wr32(dev, pll_reg, 0x50000610); - nv_wr32(dev, pll_reg + 4, reg1 | - (pll.log2P << 16) | (pll.M1 << 8) | pll.N1); + ret = nv50_calc_pll2(dev, &pll, pclk, &N1, &N2, &M1, &P); + if (ret <= 0) + return 0; + + NV_DEBUG(dev, "pclk %d out %d N %d fN 0x%04x M %d P %d\n", + pclk, ret, N1, N2, M1, P); + + reg1 = nv_rd32(dev, reg + 4) & 0xffc00000; + nv_wr32(dev, reg, 0x50000610); + nv_wr32(dev, reg + 4, reg1 | (P << 16) | (M1 << 8) | N1); + nv_wr32(dev, reg + 8, N2); } return 0; diff --git a/drivers/gpu/drm/nouveau/nv50_dac.c b/drivers/gpu/drm/nouveau/nv50_dac.c index f08f042..1fd9537 100644 --- a/drivers/gpu/drm/nouveau/nv50_dac.c +++ b/drivers/gpu/drm/nouveau/nv50_dac.c @@ -79,8 +79,8 @@ 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; + if (dev_priv->vbios.dactestval) { + load_pattern = dev_priv->vbios.dactestval; NV_DEBUG_KMS(dev, "Using bios provided load_pattern of %d\n", load_pattern); } else { diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c index 90f0bf5..2c2ec5f 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.c +++ b/drivers/gpu/drm/nouveau/nv50_display.c @@ -143,7 +143,7 @@ nv50_evo_channel_new(struct drm_device *dev, struct nouveau_channel **pchan) } ret = nv50_evo_dmaobj_new(chan, 0x3d, NvEvoVRAM, 0, 0x19, - 0, nouveau_mem_fb_amount(dev)); + 0, dev_priv->vram_size); if (ret) { nv50_evo_channel_del(pchan); return ret; @@ -231,7 +231,7 @@ nv50_display_init(struct drm_device *dev) /* This used to be in crtc unblank, but seems out of place there. */ nv_wr32(dev, NV50_PDISPLAY_UNK_380, 0); /* RAM is clamped to 256 MiB. */ - ram_amount = nouveau_mem_fb_amount(dev); + ram_amount = dev_priv->vram_size; NV_DEBUG_KMS(dev, "ram_amount %d\n", ram_amount); if (ram_amount > 256*1024*1024) ram_amount = 256*1024*1024; @@ -370,9 +370,7 @@ nv50_display_init(struct drm_device *dev) struct nouveau_connector *conn = nouveau_connector(connector); struct dcb_gpio_entry *gpio; - if (connector->connector_type != DRM_MODE_CONNECTOR_DVII && - connector->connector_type != DRM_MODE_CONNECTOR_DVID && - connector->connector_type != DRM_MODE_CONNECTOR_DisplayPort) + if (conn->dcb->gpio_tag == 0xff) continue; gpio = nouveau_bios_gpio_entry(dev, conn->dcb->gpio_tag); @@ -465,8 +463,7 @@ static int nv50_display_disable(struct drm_device *dev) int nv50_display_create(struct drm_device *dev) { struct drm_nouveau_private *dev_priv = dev->dev_private; - struct parsed_dcb *dcb = dev_priv->vbios->dcb; - uint32_t connector[16] = {}; + struct dcb_table *dcb = &dev_priv->vbios.dcb; int ret, i; NV_DEBUG_KMS(dev, "\n"); @@ -522,49 +519,20 @@ int nv50_display_create(struct drm_device *dev) NV_WARN(dev, "DCB encoder %d unknown\n", entry->type); continue; } - - connector[entry->connector] |= (1 << entry->type); } - /* It appears that DCB 3.0+ VBIOS has a connector table, however, - * I'm not 100% certain how to decode it correctly yet so just - * look at what encoders are present on each connector index and - * attempt to derive the connector type from that. - */ - for (i = 0 ; i < dcb->entries; i++) { - struct dcb_entry *entry = &dcb->entry[i]; - uint16_t encoders; - int type; - - encoders = connector[entry->connector]; - if (!(encoders & (1 << entry->type))) + for (i = 0 ; i < dcb->connector.entries; i++) { + if (i != 0 && dcb->connector.entry[i].index2 == + dcb->connector.entry[i - 1].index2) continue; - connector[entry->connector] = 0; - - if (encoders & (1 << OUTPUT_DP)) { - type = DRM_MODE_CONNECTOR_DisplayPort; - } else if (encoders & (1 << OUTPUT_TMDS)) { - if (encoders & (1 << OUTPUT_ANALOG)) - type = DRM_MODE_CONNECTOR_DVII; - else - type = DRM_MODE_CONNECTOR_DVID; - } else if (encoders & (1 << OUTPUT_ANALOG)) { - type = DRM_MODE_CONNECTOR_VGA; - } else if (encoders & (1 << OUTPUT_LVDS)) { - type = DRM_MODE_CONNECTOR_LVDS; - } else { - type = DRM_MODE_CONNECTOR_Unknown; - } - - if (type == DRM_MODE_CONNECTOR_Unknown) - continue; - - nouveau_connector_create(dev, entry->connector, type); + nouveau_connector_create(dev, &dcb->connector.entry[i]); } ret = nv50_display_init(dev); - if (ret) + if (ret) { + nv50_display_destroy(dev); return ret; + } return 0; } @@ -667,8 +635,8 @@ nv50_display_irq_head(struct drm_device *dev, int *phead, return -1; } - for (i = 0; i < dev_priv->vbios->dcb->entries; i++) { - struct dcb_entry *dcbent = &dev_priv->vbios->dcb->entry[i]; + for (i = 0; i < dev_priv->vbios.dcb.entries; i++) { + struct dcb_entry *dcbent = &dev_priv->vbios.dcb.entry[i]; if (dcbent->type != type) continue; @@ -692,7 +660,7 @@ nv50_display_script_select(struct drm_device *dev, struct dcb_entry *dcbent, 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; + struct nvbios *bios = &dev_priv->vbios; uint32_t mc, script = 0, or; list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { @@ -710,7 +678,7 @@ nv50_display_script_select(struct drm_device *dev, struct dcb_entry *dcbent, switch (dcbent->type) { case OUTPUT_LVDS: script = (mc >> 8) & 0xf; - if (bios->pub.fp_no_ddc) { + if (bios->fp_no_ddc) { if (bios->fp.dual_link) script |= 0x0100; if (bios->fp.if_is_24bit) @@ -815,6 +783,37 @@ ack: } static void +nv50_display_unk20_dp_hack(struct drm_device *dev, struct dcb_entry *dcb) +{ + int or = ffs(dcb->or) - 1, link = !(dcb->dpconf.sor.link & 1); + struct drm_encoder *encoder; + uint32_t tmp, unk0 = 0, unk1 = 0; + + if (dcb->type != OUTPUT_DP) + return; + + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder); + + if (nv_encoder->dcb == dcb) { + unk0 = nv_encoder->dp.unk0; + unk1 = nv_encoder->dp.unk1; + break; + } + } + + if (unk0 || unk1) { + tmp = nv_rd32(dev, NV50_SOR_DP_CTRL(or, link)); + tmp &= 0xfffffe03; + nv_wr32(dev, NV50_SOR_DP_CTRL(or, link), tmp | unk0); + + tmp = nv_rd32(dev, NV50_SOR_DP_UNK128(or, link)); + tmp &= 0xfef080c0; + nv_wr32(dev, NV50_SOR_DP_UNK128(or, link), tmp | unk1); + } +} + +static void nv50_display_unk20_handler(struct drm_device *dev) { struct dcb_entry *dcbent; @@ -837,6 +836,8 @@ nv50_display_unk20_handler(struct drm_device *dev) nouveau_bios_run_display_table(dev, dcbent, script, pclk); + nv50_display_unk20_dp_hack(dev, dcbent); + tmp = nv_rd32(dev, NV50_PDISPLAY_CRTC_CLK_CTRL2(head)); tmp &= ~0x000000f; nv_wr32(dev, NV50_PDISPLAY_CRTC_CLK_CTRL2(head), tmp); @@ -919,10 +920,12 @@ nv50_display_error_handler(struct drm_device *dev) nv_wr32(dev, NV50_PDISPLAY_TRAPPED_ADDR, 0x90000000); } -static void -nv50_display_irq_hotplug(struct drm_device *dev) +void +nv50_display_irq_hotplug_bh(struct work_struct *work) { - struct drm_nouveau_private *dev_priv = dev->dev_private; + struct drm_nouveau_private *dev_priv = + container_of(work, struct drm_nouveau_private, hpd_work); + struct drm_device *dev = dev_priv->dev; struct drm_connector *connector; const uint32_t gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 }; uint32_t unplug_mask, plug_mask, change_mask; @@ -975,6 +978,8 @@ nv50_display_irq_hotplug(struct drm_device *dev) nv_wr32(dev, 0xe054, nv_rd32(dev, 0xe054)); if (dev_priv->chipset >= 0x90) nv_wr32(dev, 0xe074, nv_rd32(dev, 0xe074)); + + drm_sysfs_hotplug_event(dev); } void @@ -983,8 +988,10 @@ nv50_display_irq_handler(struct drm_device *dev) struct drm_nouveau_private *dev_priv = dev->dev_private; uint32_t delayed = 0; - while (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_HOTPLUG) - nv50_display_irq_hotplug(dev); + if (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_HOTPLUG) { + if (!work_pending(&dev_priv->hpd_work)) + queue_work(dev_priv->wq, &dev_priv->hpd_work); + } while (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_DISPLAY) { uint32_t intr0 = nv_rd32(dev, NV50_PDISPLAY_INTR_0); diff --git a/drivers/gpu/drm/nouveau/nv50_display.h b/drivers/gpu/drm/nouveau/nv50_display.h index 3ae8d07..581d405 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.h +++ b/drivers/gpu/drm/nouveau/nv50_display.h @@ -37,6 +37,7 @@ void nv50_display_irq_handler(struct drm_device *dev); void nv50_display_irq_handler_bh(struct work_struct *work); +void nv50_display_irq_hotplug_bh(struct work_struct *work); int nv50_display_init(struct drm_device *dev); int nv50_display_create(struct drm_device *dev); int nv50_display_destroy(struct drm_device *dev); diff --git a/drivers/gpu/drm/nouveau/nv50_fb.c b/drivers/gpu/drm/nouveau/nv50_fb.c new file mode 100644 index 0000000..32611bd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nv50_fb.c @@ -0,0 +1,38 @@ +#include "drmP.h" +#include "drm.h" +#include "nouveau_drv.h" +#include "nouveau_drm.h" + +int +nv50_fb_init(struct drm_device *dev) +{ + struct drm_nouveau_private *dev_priv = dev->dev_private; + + /* Not a clue what this is exactly. Without pointing it at a + * scratch page, VRAM->GART blits with M2MF (as in DDX DFS) + * cause IOMMU "read from address 0" errors (rh#561267) + */ + nv_wr32(dev, 0x100c08, dev_priv->gart_info.sg_dummy_bus >> 8); + + /* This is needed to get meaningful information from 100c90 + * on traps. No idea what these values mean exactly. */ + switch (dev_priv->chipset) { + case 0x50: + nv_wr32(dev, 0x100c90, 0x0707ff); + break; + case 0xa5: + case 0xa8: + nv_wr32(dev, 0x100c90, 0x0d0fff); + break; + default: + nv_wr32(dev, 0x100c90, 0x1d07ff); + break; + } + + return 0; +} + +void +nv50_fb_takedown(struct drm_device *dev) +{ +} diff --git a/drivers/gpu/drm/nouveau/nv50_fbcon.c b/drivers/gpu/drm/nouveau/nv50_fbcon.c index 0f57cdf..a8c70e7 100644 --- a/drivers/gpu/drm/nouveau/nv50_fbcon.c +++ b/drivers/gpu/drm/nouveau/nv50_fbcon.c @@ -109,7 +109,7 @@ nv50_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) return; } - width = (image->width + 31) & ~31; + width = ALIGN(image->width, 32); dwords = (width * image->height) >> 5; BEGIN_RING(chan, NvSub2D, 0x0814, 2); @@ -157,8 +157,11 @@ nv50_fbcon_accel_init(struct fb_info *info) struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_channel *chan = dev_priv->channel; struct nouveau_gpuobj *eng2d = NULL; + uint64_t fb; int ret, format; + fb = info->fix.smem_start - dev_priv->fb_phys + dev_priv->vm_vram_base; + switch (info->var.bits_per_pixel) { case 8: format = 0xf3; @@ -233,7 +236,7 @@ nv50_fbcon_accel_init(struct fb_info *info) BEGIN_RING(chan, NvSub2D, 0x0808, 3); OUT_RING(chan, 0); OUT_RING(chan, 0); - OUT_RING(chan, 0); + OUT_RING(chan, 1); BEGIN_RING(chan, NvSub2D, 0x081c, 1); OUT_RING(chan, 1); BEGIN_RING(chan, NvSub2D, 0x0840, 4); @@ -248,9 +251,8 @@ nv50_fbcon_accel_init(struct fb_info *info) OUT_RING(chan, info->fix.line_length); OUT_RING(chan, info->var.xres_virtual); OUT_RING(chan, info->var.yres_virtual); - OUT_RING(chan, 0); - OUT_RING(chan, info->fix.smem_start - dev_priv->fb_phys + - dev_priv->vm_vram_base); + OUT_RING(chan, upper_32_bits(fb)); + OUT_RING(chan, lower_32_bits(fb)); BEGIN_RING(chan, NvSub2D, 0x0230, 2); OUT_RING(chan, format); OUT_RING(chan, 1); @@ -258,9 +260,8 @@ nv50_fbcon_accel_init(struct fb_info *info) OUT_RING(chan, info->fix.line_length); OUT_RING(chan, info->var.xres_virtual); OUT_RING(chan, info->var.yres_virtual); - OUT_RING(chan, 0); - OUT_RING(chan, info->fix.smem_start - dev_priv->fb_phys + - dev_priv->vm_vram_base); + OUT_RING(chan, upper_32_bits(fb)); + OUT_RING(chan, lower_32_bits(fb)); return 0; } diff --git a/drivers/gpu/drm/nouveau/nv50_fifo.c b/drivers/gpu/drm/nouveau/nv50_fifo.c index df5335a..e20c0e2 100644 --- a/drivers/gpu/drm/nouveau/nv50_fifo.c +++ b/drivers/gpu/drm/nouveau/nv50_fifo.c @@ -243,6 +243,7 @@ nv50_fifo_create_context(struct nouveau_channel *chan) struct drm_device *dev = chan->dev; struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_gpuobj *ramfc = NULL; + unsigned long flags; int ret; NV_DEBUG(dev, "ch%d\n", chan->id); @@ -278,6 +279,8 @@ nv50_fifo_create_context(struct nouveau_channel *chan) return ret; } + spin_lock_irqsave(&dev_priv->context_switch_lock, flags); + dev_priv->engine.instmem.prepare_access(dev, true); nv_wo32(dev, ramfc, 0x48/4, chan->pushbuf->instance >> 4); @@ -306,10 +309,12 @@ nv50_fifo_create_context(struct nouveau_channel *chan) ret = nv50_fifo_channel_enable(dev, chan->id, false); if (ret) { NV_ERROR(dev, "error enabling ch%d: %d\n", chan->id, ret); + spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags); nouveau_gpuobj_ref_del(dev, &chan->ramfc); return ret; } + spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags); return 0; } diff --git a/drivers/gpu/drm/nouveau/nv50_gpio.c b/drivers/gpu/drm/nouveau/nv50_gpio.c new file mode 100644 index 0000000..c61782b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nv50_gpio.c @@ -0,0 +1,76 @@ +/* + * Copyright 2010 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 "drmP.h" +#include "nouveau_drv.h" +#include "nouveau_hw.h" + +static int +nv50_gpio_location(struct dcb_gpio_entry *gpio, uint32_t *reg, uint32_t *shift) +{ + const uint32_t nv50_gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 }; + + if (gpio->line > 32) + return -EINVAL; + + *reg = nv50_gpio_reg[gpio->line >> 3]; + *shift = (gpio->line & 7) << 2; + return 0; +} + +int +nv50_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag) +{ + struct dcb_gpio_entry *gpio; + uint32_t r, s, v; + + gpio = nouveau_bios_gpio_entry(dev, tag); + if (!gpio) + return -ENOENT; + + if (nv50_gpio_location(gpio, &r, &s)) + return -EINVAL; + + v = nv_rd32(dev, r) >> (s + 2); + return ((v & 1) == (gpio->state[1] & 1)); +} + +int +nv50_gpio_set(struct drm_device *dev, enum dcb_gpio_tag tag, int state) +{ + struct dcb_gpio_entry *gpio; + uint32_t r, s, v; + + gpio = nouveau_bios_gpio_entry(dev, tag); + if (!gpio) + return -ENOENT; + + if (nv50_gpio_location(gpio, &r, &s)) + return -EINVAL; + + v = nv_rd32(dev, r) & ~(0x3 << s); + v |= (gpio->state[state] ^ 2) << s; + nv_wr32(dev, r, v); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nv50_graph.c b/drivers/gpu/drm/nouveau/nv50_graph.c index 6d50480..b203d06 100644 --- a/drivers/gpu/drm/nouveau/nv50_graph.c +++ b/drivers/gpu/drm/nouveau/nv50_graph.c @@ -28,7 +28,7 @@ #include "drm.h" #include "nouveau_drv.h" -/*(DEBLOBBED)*/ +#include "nouveau_grctx.h" #define IS_G80 ((dev_priv->chipset & 0xf0) == 0x50) @@ -79,6 +56,10 @@ nv50_graph_init_intr(struct drm_device *dev) static void nv50_graph_init_regs__nv(struct drm_device *dev) { + struct drm_nouveau_private *dev_priv = dev->dev_private; + uint32_t units = nv_rd32(dev, 0x1540); + int i; + NV_DEBUG(dev, "\n"); nv_wr32(dev, 0x400804, 0xc0000000); @@ -88,6 +69,20 @@ nv50_graph_init_regs__nv(struct drm_device *dev) nv_wr32(dev, 0x405018, 0xc0000000); nv_wr32(dev, 0x402000, 0xc0000000); + for (i = 0; i < 16; i++) { + if (units & 1 << i) { + if (dev_priv->chipset < 0xa0) { + nv_wr32(dev, 0x408900 + (i << 12), 0xc0000000); + nv_wr32(dev, 0x408e08 + (i << 12), 0xc0000000); + nv_wr32(dev, 0x408314 + (i << 12), 0xc0000000); + } else { + nv_wr32(dev, 0x408600 + (i << 11), 0xc0000000); + nv_wr32(dev, 0x408708 + (i << 11), 0xc0000000); + nv_wr32(dev, 0x40831c + (i << 11), 0xc0000000); + } + } + } + nv_wr32(dev, 0x400108, 0xffffffff); nv_wr32(dev, 0x400824, 0x00004000); @@ -111,9 +106,34 @@ nv50_graph_init_ctxctl(struct drm_device *dev) NV_DEBUG(dev, "\n"); - nouveau_grctx_prog_load(dev); - if (!dev_priv->engine.graph.ctxprog) - dev_priv->engine.graph.accel_blocked = true; + if (nouveau_ctxfw) { + nouveau_grctx_prog_load(dev); + dev_priv->engine.graph.grctx_size = 0x70000; + } + if (!dev_priv->engine.graph.ctxprog) { + struct nouveau_grctx ctx = {}; + uint32_t *cp = kmalloc(512 * 4, GFP_KERNEL); + int i; + if (!cp) { + NV_ERROR(dev, "Couldn't alloc ctxprog! Disabling acceleration.\n"); + dev_priv->engine.graph.accel_blocked = true; + return 0; + } + ctx.dev = dev; + ctx.mode = NOUVEAU_GRCTX_PROG; + ctx.data = cp; + ctx.ctxprog_max = 512; + if (!nv50_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]); + } else { + dev_priv->engine.graph.accel_blocked = true; + } + kfree(cp); + } nv_wr32(dev, 0x400320, 4); nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0); @@ -193,13 +213,13 @@ nv50_graph_create_context(struct nouveau_channel *chan) struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_gpuobj *ramin = chan->ramin->gpuobj; struct nouveau_gpuobj *ctx; - uint32_t grctx_size = 0x70000; + struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph; int hdr, ret; NV_DEBUG(dev, "ch%d\n", chan->id); - ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, grctx_size, 0x1000, - NVOBJ_FLAG_ZERO_ALLOC | + ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, pgraph->grctx_size, + 0x1000, NVOBJ_FLAG_ZERO_ALLOC | NVOBJ_FLAG_ZERO_FREE, &chan->ramin_grctx); if (ret) return ret; @@ -209,7 +229,7 @@ nv50_graph_create_context(struct nouveau_channel *chan) dev_priv->engine.instmem.prepare_access(dev, true); nv_wo32(dev, ramin, (hdr + 0x00)/4, 0x00190002); nv_wo32(dev, ramin, (hdr + 0x04)/4, chan->ramin_grctx->instance + - grctx_size - 1); + pgraph->grctx_size - 1); nv_wo32(dev, ramin, (hdr + 0x08)/4, chan->ramin_grctx->instance); nv_wo32(dev, ramin, (hdr + 0x0c)/4, 0); nv_wo32(dev, ramin, (hdr + 0x10)/4, 0); @@ -217,12 +237,16 @@ nv50_graph_create_context(struct nouveau_channel *chan) dev_priv->engine.instmem.finish_access(dev); dev_priv->engine.instmem.prepare_access(dev, true); - nouveau_grctx_vals_load(dev, ctx); + if (!pgraph->ctxprog) { + struct nouveau_grctx ctx = {}; + ctx.dev = chan->dev; + ctx.mode = NOUVEAU_GRCTX_VALS; + ctx.data = chan->ramin_grctx->gpuobj; + nv50_grctx_init(&ctx); + } else { + 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); - else - nv_wo32(dev, ctx, 0x0011c/4, 0x00000000); dev_priv->engine.instmem.finish_access(dev); return 0; @@ -386,9 +410,10 @@ struct nouveau_pgraph_object_class nv50_graph_grclass[] = { { 0x5039, false, NULL }, /* m2mf */ { 0x502d, false, NULL }, /* 2d */ { 0x50c0, false, NULL }, /* compute */ + { 0x85c0, false, NULL }, /* compute (nva3, nva5, nva8) */ { 0x5097, false, NULL }, /* tesla (nv50) */ - { 0x8297, false, NULL }, /* tesla (nv80/nv90) */ - { 0x8397, false, NULL }, /* tesla (nva0) */ - { 0x8597, false, NULL }, /* tesla (nva8) */ + { 0x8297, false, NULL }, /* tesla (nv8x/nv9x) */ + { 0x8397, false, NULL }, /* tesla (nva0, nvaa, nvac) */ + { 0x8597, false, NULL }, /* tesla (nva3, nva5, nva8) */ {} }; diff --git a/drivers/gpu/drm/nouveau/nv50_grctx.c b/drivers/gpu/drm/nouveau/nv50_grctx.c new file mode 100644 index 0000000..42a8fb2 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nv50_grctx.c @@ -0,0 +1,2383 @@ +/* + * Copyright 2009 Marcin Kościelnicki + * + * 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. + */ + +#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_UNK01 ((0 * 32) + 1) +#define CP_FLAG_UNK01_CLEAR 0 +#define CP_FLAG_UNK01_SET 1 +#define CP_FLAG_UNK03 ((0 * 32) + 3) +#define CP_FLAG_UNK03_CLEAR 0 +#define CP_FLAG_UNK03_SET 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_UNK0B ((0 * 32) + 0xb) +#define CP_FLAG_UNK0B_CLEAR 0 +#define CP_FLAG_UNK0B_SET 1 +#define CP_FLAG_UNK1D ((0 * 32) + 0x1d) +#define CP_FLAG_UNK1D_CLEAR 0 +#define CP_FLAG_UNK1D_SET 1 +#define CP_FLAG_UNK20 ((1 * 32) + 0) +#define CP_FLAG_UNK20_CLEAR 0 +#define CP_FLAG_UNK20_SET 1 +#define CP_FLAG_STATUS ((2 * 32) + 0) +#define CP_FLAG_STATUS_BUSY 0 +#define CP_FLAG_STATUS_IDLE 1 +#define CP_FLAG_AUTO_SAVE ((2 * 32) + 4) +#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0 +#define CP_FLAG_AUTO_SAVE_PENDING 1 +#define CP_FLAG_AUTO_LOAD ((2 * 32) + 5) +#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0 +#define CP_FLAG_AUTO_LOAD_PENDING 1 +#define CP_FLAG_NEWCTX ((2 * 32) + 10) +#define CP_FLAG_NEWCTX_BUSY 0 +#define CP_FLAG_NEWCTX_DONE 1 +#define CP_FLAG_XFER ((2 * 32) + 11) +#define CP_FLAG_XFER_IDLE 0 +#define CP_FLAG_XFER_BUSY 1 +#define CP_FLAG_ALWAYS ((2 * 32) + 13) +#define CP_FLAG_ALWAYS_FALSE 0 +#define CP_FLAG_ALWAYS_TRUE 1 +#define CP_FLAG_INTR ((2 * 32) + 15) +#define CP_FLAG_INTR_NOT_PENDING 0 +#define CP_FLAG_INTR_PENDING 1 + +#define CP_CTX 0x00100000 +#define CP_CTX_COUNT 0x000f0000 +#define CP_CTX_COUNT_SHIFT 16 +#define CP_CTX_REG 0x00003fff +#define CP_LOAD_SR 0x00200000 +#define CP_LOAD_SR_VALUE 0x000fffff +#define CP_BRA 0x00400000 +#define CP_BRA_IP 0x0001ff00 +#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_NEWCTX 0x00600004 +#define CP_NEXT_TO_SWAP 0x00600005 +#define CP_SET_CONTEXT_POINTER 0x00600006 +#define CP_SET_XFER_POINTER 0x00600007 +#define CP_ENABLE 0x00600009 +#define CP_END 0x0060000c +#define CP_NEXT_TO_CURRENT 0x0060000d +#define CP_DISABLE1 0x0090ffff +#define CP_DISABLE2 0x0091ffff +#define CP_XFER_1 0x008000ff +#define CP_XFER_2 0x008800ff +#define CP_SEEK_1 0x00c000ff +#define CP_SEEK_2 0x00c800ff + +#include "drmP.h" +#include "nouveau_drv.h" +#include "nouveau_grctx.h" + +/* + * This code deals with PGRAPH contexts on NV50 family cards. Like NV40, it's + * the GPU itself that does context-switching, but it needs a special + * microcode to do it. And it's the driver's task to supply this microcode, + * further known as ctxprog, as well as the initial context values, known + * as ctxvals. + * + * Without ctxprog, you cannot switch contexts. Not even in software, since + * the majority of context [xfer strands] isn't accessible directly. You're + * stuck with a single channel, and you also suffer all the problems resulting + * from missing ctxvals, since you cannot load them. + * + * Without ctxvals, you're stuck with PGRAPH's default context. It's enough to + * run 2d operations, but trying to utilise 3d or CUDA will just lock you up, + * since you don't have... some sort of needed setup. + * + * Nouveau will just disable acceleration if not given ctxprog + ctxvals, since + * it's too much hassle to handle no-ctxprog as a special case. + */ + +/* + * How ctxprogs work. + * + * The ctxprog is written in its own kind of microcode, with very small and + * crappy set of available commands. You upload it to a small [512 insns] + * area of memory on PGRAPH, and it'll be run when PFIFO wants PGRAPH to + * switch channel. or when the driver explicitely requests it. Stuff visible + * to ctxprog consists of: PGRAPH MMIO registers, PGRAPH context strands, + * the per-channel context save area in VRAM [known as ctxvals or grctx], + * 4 flags registers, a scratch register, two grctx pointers, plus many + * random poorly-understood details. + * + * When ctxprog runs, it's supposed to check what operations are asked of it, + * save old context if requested, optionally reset PGRAPH and switch to the + * new channel, and load the new context. Context consists of three major + * parts: subset of MMIO registers and two "xfer areas". + */ + +/* TODO: + * - document unimplemented bits compared to nvidia + * - NVAx: make a TP subroutine, use it. + * - use 0x4008fc instead of 0x1540? + */ + +enum cp_label { + cp_check_load = 1, + cp_setup_auto_load, + cp_setup_load, + cp_setup_save, + cp_swap_state, + cp_prepare_exit, + cp_exit, +}; + +static void nv50_graph_construct_mmio(struct nouveau_grctx *ctx); +static void nv50_graph_construct_xfer1(struct nouveau_grctx *ctx); +static void nv50_graph_construct_xfer2(struct nouveau_grctx *ctx); + +/* Main function: construct the ctxprog skeleton, call the other functions. */ + +int +nv50_grctx_init(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + + switch (dev_priv->chipset) { + case 0x50: + case 0x84: + case 0x86: + case 0x92: + case 0x94: + case 0x96: + case 0x98: + case 0xa0: + case 0xa3: + case 0xa5: + case 0xa8: + case 0xaa: + case 0xac: + break; + default: + NV_ERROR(ctx->dev, "I don't know how to make a ctxprog for " + "your NV%x card.\n", dev_priv->chipset); + NV_ERROR(ctx->dev, "Disabling acceleration. Please contact " + "the devs.\n"); + return -ENOSYS; + } + /* 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_out (ctx, CP_DISABLE1); + cp_out (ctx, CP_DISABLE2); + cp_out (ctx, CP_ENABLE); + cp_out (ctx, CP_NEXT_TO_SWAP); + cp_set (ctx, UNK01, SET); + cp_name(ctx, cp_setup_load); + cp_out (ctx, CP_NEWCTX); + cp_wait(ctx, NEWCTX, BUSY); + cp_set (ctx, UNK1D, CLEAR); + cp_set (ctx, SWAP_DIRECTION, LOAD); + cp_bra (ctx, UNK0B, SET, cp_prepare_exit); + cp_bra (ctx, ALWAYS, TRUE, cp_swap_state); + + /* setup for context save */ + cp_name(ctx, cp_setup_save); + cp_set (ctx, UNK1D, SET); + cp_wait(ctx, STATUS, BUSY); + cp_wait(ctx, INTR, PENDING); + cp_bra (ctx, STATUS, BUSY, cp_setup_save); + cp_set (ctx, UNK01, SET); + cp_set (ctx, SWAP_DIRECTION, SAVE); + + /* general PGRAPH state */ + cp_name(ctx, cp_swap_state); + cp_set (ctx, UNK03, SET); + cp_pos (ctx, 0x00004/4); + cp_ctx (ctx, 0x400828, 1); /* needed. otherwise, flickering happens. */ + cp_pos (ctx, 0x00100/4); + nv50_graph_construct_mmio(ctx); + nv50_graph_construct_xfer1(ctx); + nv50_graph_construct_xfer2(ctx); + + cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load); + + cp_set (ctx, UNK20, SET); + cp_set (ctx, SWAP_DIRECTION, SAVE); /* no idea why this is needed, but fixes at least one lockup. */ + cp_lsr (ctx, ctx->ctxvals_base); + cp_out (ctx, CP_SET_XFER_POINTER); + cp_lsr (ctx, 4); + cp_out (ctx, CP_SEEK_1); + cp_out (ctx, CP_XFER_1); + cp_wait(ctx, XFER, BUSY); + + /* pre-exit state updates */ + cp_name(ctx, cp_prepare_exit); + cp_set (ctx, UNK01, CLEAR); + cp_set (ctx, UNK03, CLEAR); + cp_set (ctx, UNK1D, CLEAR); + + 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); + ctx->ctxvals_pos += 0x400; /* padding... no idea why you need it */ + + return 0; +} + +/* + * Constructs MMIO part of ctxprog and ctxvals. Just a matter of knowing which + * registers to save/restore and the default values for them. + */ + +static void +nv50_graph_construct_mmio(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int i, j; + int offset, base; + uint32_t units = nv_rd32 (ctx->dev, 0x1540); + + /* 0800: DISPATCH */ + cp_ctx(ctx, 0x400808, 7); + gr_def(ctx, 0x400814, 0x00000030); + cp_ctx(ctx, 0x400834, 0x32); + if (dev_priv->chipset == 0x50) { + gr_def(ctx, 0x400834, 0xff400040); + gr_def(ctx, 0x400838, 0xfff00080); + gr_def(ctx, 0x40083c, 0xfff70090); + gr_def(ctx, 0x400840, 0xffe806a8); + } + gr_def(ctx, 0x400844, 0x00000002); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + gr_def(ctx, 0x400894, 0x00001000); + gr_def(ctx, 0x4008e8, 0x00000003); + gr_def(ctx, 0x4008ec, 0x00001000); + if (dev_priv->chipset == 0x50) + cp_ctx(ctx, 0x400908, 0xb); + else if (dev_priv->chipset < 0xa0) + cp_ctx(ctx, 0x400908, 0xc); + else + cp_ctx(ctx, 0x400908, 0xe); + + if (dev_priv->chipset >= 0xa0) + cp_ctx(ctx, 0x400b00, 0x1); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + cp_ctx(ctx, 0x400b10, 0x1); + gr_def(ctx, 0x400b10, 0x0001629d); + cp_ctx(ctx, 0x400b20, 0x1); + gr_def(ctx, 0x400b20, 0x0001629d); + } + + /* 0C00: VFETCH */ + cp_ctx(ctx, 0x400c08, 0x2); + gr_def(ctx, 0x400c08, 0x0000fe0c); + + /* 1000 */ + if (dev_priv->chipset < 0xa0) { + cp_ctx(ctx, 0x401008, 0x4); + gr_def(ctx, 0x401014, 0x00001000); + } else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa) { + cp_ctx(ctx, 0x401008, 0x5); + gr_def(ctx, 0x401018, 0x00001000); + } else { + cp_ctx(ctx, 0x401008, 0x5); + gr_def(ctx, 0x401018, 0x00004000); + } + + /* 1400 */ + cp_ctx(ctx, 0x401400, 0x8); + cp_ctx(ctx, 0x401424, 0x3); + if (dev_priv->chipset == 0x50) + gr_def(ctx, 0x40142c, 0x0001fd87); + else + gr_def(ctx, 0x40142c, 0x00000187); + cp_ctx(ctx, 0x401540, 0x5); + gr_def(ctx, 0x401550, 0x00001018); + + /* 1800: STREAMOUT */ + cp_ctx(ctx, 0x401814, 0x1); + gr_def(ctx, 0x401814, 0x000000ff); + if (dev_priv->chipset == 0x50) { + cp_ctx(ctx, 0x40181c, 0xe); + gr_def(ctx, 0x401850, 0x00000004); + } else if (dev_priv->chipset < 0xa0) { + cp_ctx(ctx, 0x40181c, 0xf); + gr_def(ctx, 0x401854, 0x00000004); + } else { + cp_ctx(ctx, 0x40181c, 0x13); + gr_def(ctx, 0x401864, 0x00000004); + } + + /* 1C00 */ + cp_ctx(ctx, 0x401c00, 0x1); + switch (dev_priv->chipset) { + case 0x50: + gr_def(ctx, 0x401c00, 0x0001005f); + break; + case 0x84: + case 0x86: + case 0x94: + gr_def(ctx, 0x401c00, 0x044d00df); + break; + case 0x92: + case 0x96: + case 0x98: + case 0xa0: + case 0xaa: + case 0xac: + gr_def(ctx, 0x401c00, 0x042500df); + break; + case 0xa3: + case 0xa5: + case 0xa8: + gr_def(ctx, 0x401c00, 0x142500df); + break; + } + + /* 2400 */ + cp_ctx(ctx, 0x402400, 0x1); + if (dev_priv->chipset == 0x50) + cp_ctx(ctx, 0x402408, 0x1); + else + cp_ctx(ctx, 0x402408, 0x2); + gr_def(ctx, 0x402408, 0x00000600); + + /* 2800 */ + cp_ctx(ctx, 0x402800, 0x1); + if (dev_priv->chipset == 0x50) + gr_def(ctx, 0x402800, 0x00000006); + + /* 2C00 */ + cp_ctx(ctx, 0x402c08, 0x6); + if (dev_priv->chipset != 0x50) + gr_def(ctx, 0x402c14, 0x01000000); + gr_def(ctx, 0x402c18, 0x000000ff); + if (dev_priv->chipset == 0x50) + cp_ctx(ctx, 0x402ca0, 0x1); + else + cp_ctx(ctx, 0x402ca0, 0x2); + if (dev_priv->chipset < 0xa0) + gr_def(ctx, 0x402ca0, 0x00000400); + else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa) + gr_def(ctx, 0x402ca0, 0x00000800); + else + gr_def(ctx, 0x402ca0, 0x00000400); + cp_ctx(ctx, 0x402cac, 0x4); + + /* 3000 */ + cp_ctx(ctx, 0x403004, 0x1); + gr_def(ctx, 0x403004, 0x00000001); + + /* 3404 */ + if (dev_priv->chipset >= 0xa0) { + cp_ctx(ctx, 0x403404, 0x1); + gr_def(ctx, 0x403404, 0x00000001); + } + + /* 5000 */ + cp_ctx(ctx, 0x405000, 0x1); + switch (dev_priv->chipset) { + case 0x50: + gr_def(ctx, 0x405000, 0x00300080); + break; + case 0x84: + case 0xa0: + case 0xa3: + case 0xa5: + case 0xa8: + case 0xaa: + case 0xac: + gr_def(ctx, 0x405000, 0x000e0080); + break; + case 0x86: + case 0x92: + case 0x94: + case 0x96: + case 0x98: + gr_def(ctx, 0x405000, 0x00000080); + break; + } + cp_ctx(ctx, 0x405014, 0x1); + gr_def(ctx, 0x405014, 0x00000004); + cp_ctx(ctx, 0x40501c, 0x1); + cp_ctx(ctx, 0x405024, 0x1); + cp_ctx(ctx, 0x40502c, 0x1); + + /* 5400 or maybe 4800 */ + if (dev_priv->chipset == 0x50) { + offset = 0x405400; + cp_ctx(ctx, 0x405400, 0xea); + } else if (dev_priv->chipset < 0x94) { + offset = 0x405400; + cp_ctx(ctx, 0x405400, 0xcb); + } else if (dev_priv->chipset < 0xa0) { + offset = 0x405400; + cp_ctx(ctx, 0x405400, 0xcc); + } else if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + offset = 0x404800; + cp_ctx(ctx, 0x404800, 0xda); + } else { + offset = 0x405400; + cp_ctx(ctx, 0x405400, 0xd4); + } + gr_def(ctx, offset + 0x0c, 0x00000002); + gr_def(ctx, offset + 0x10, 0x00000001); + if (dev_priv->chipset >= 0x94) + offset += 4; + gr_def(ctx, offset + 0x1c, 0x00000001); + gr_def(ctx, offset + 0x20, 0x00000100); + gr_def(ctx, offset + 0x38, 0x00000002); + gr_def(ctx, offset + 0x3c, 0x00000001); + gr_def(ctx, offset + 0x40, 0x00000001); + gr_def(ctx, offset + 0x50, 0x00000001); + gr_def(ctx, offset + 0x54, 0x003fffff); + gr_def(ctx, offset + 0x58, 0x00001fff); + gr_def(ctx, offset + 0x60, 0x00000001); + gr_def(ctx, offset + 0x64, 0x00000001); + gr_def(ctx, offset + 0x6c, 0x00000001); + gr_def(ctx, offset + 0x70, 0x00000001); + gr_def(ctx, offset + 0x74, 0x00000001); + gr_def(ctx, offset + 0x78, 0x00000004); + gr_def(ctx, offset + 0x7c, 0x00000001); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + offset += 4; + gr_def(ctx, offset + 0x80, 0x00000001); + gr_def(ctx, offset + 0x84, 0x00000001); + gr_def(ctx, offset + 0x88, 0x00000007); + gr_def(ctx, offset + 0x8c, 0x00000001); + gr_def(ctx, offset + 0x90, 0x00000007); + gr_def(ctx, offset + 0x94, 0x00000001); + gr_def(ctx, offset + 0x98, 0x00000001); + gr_def(ctx, offset + 0x9c, 0x00000001); + if (dev_priv->chipset == 0x50) { + gr_def(ctx, offset + 0xb0, 0x00000001); + gr_def(ctx, offset + 0xb4, 0x00000001); + gr_def(ctx, offset + 0xbc, 0x00000001); + gr_def(ctx, offset + 0xc0, 0x0000000a); + gr_def(ctx, offset + 0xd0, 0x00000040); + gr_def(ctx, offset + 0xd8, 0x00000002); + gr_def(ctx, offset + 0xdc, 0x00000100); + gr_def(ctx, offset + 0xe0, 0x00000001); + gr_def(ctx, offset + 0xe4, 0x00000100); + gr_def(ctx, offset + 0x100, 0x00000001); + gr_def(ctx, offset + 0x124, 0x00000004); + gr_def(ctx, offset + 0x13c, 0x00000001); + gr_def(ctx, offset + 0x140, 0x00000100); + gr_def(ctx, offset + 0x148, 0x00000001); + gr_def(ctx, offset + 0x154, 0x00000100); + gr_def(ctx, offset + 0x158, 0x00000001); + gr_def(ctx, offset + 0x15c, 0x00000100); + gr_def(ctx, offset + 0x164, 0x00000001); + gr_def(ctx, offset + 0x170, 0x00000100); + gr_def(ctx, offset + 0x174, 0x00000001); + gr_def(ctx, offset + 0x17c, 0x00000001); + gr_def(ctx, offset + 0x188, 0x00000002); + gr_def(ctx, offset + 0x190, 0x00000001); + gr_def(ctx, offset + 0x198, 0x00000001); + gr_def(ctx, offset + 0x1ac, 0x00000003); + offset += 0xd0; + } else { + gr_def(ctx, offset + 0xb0, 0x00000001); + gr_def(ctx, offset + 0xb4, 0x00000100); + gr_def(ctx, offset + 0xbc, 0x00000001); + gr_def(ctx, offset + 0xc8, 0x00000100); + gr_def(ctx, offset + 0xcc, 0x00000001); + gr_def(ctx, offset + 0xd0, 0x00000100); + gr_def(ctx, offset + 0xd8, 0x00000001); + gr_def(ctx, offset + 0xe4, 0x00000100); + } + gr_def(ctx, offset + 0xf8, 0x00000004); + gr_def(ctx, offset + 0xfc, 0x00000070); + gr_def(ctx, offset + 0x100, 0x00000080); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + offset += 4; + gr_def(ctx, offset + 0x114, 0x0000000c); + if (dev_priv->chipset == 0x50) + offset -= 4; + gr_def(ctx, offset + 0x11c, 0x00000008); + gr_def(ctx, offset + 0x120, 0x00000014); + if (dev_priv->chipset == 0x50) { + gr_def(ctx, offset + 0x124, 0x00000026); + offset -= 0x18; + } else { + gr_def(ctx, offset + 0x128, 0x00000029); + gr_def(ctx, offset + 0x12c, 0x00000027); + gr_def(ctx, offset + 0x130, 0x00000026); + gr_def(ctx, offset + 0x134, 0x00000008); + gr_def(ctx, offset + 0x138, 0x00000004); + gr_def(ctx, offset + 0x13c, 0x00000027); + } + gr_def(ctx, offset + 0x148, 0x00000001); + gr_def(ctx, offset + 0x14c, 0x00000002); + gr_def(ctx, offset + 0x150, 0x00000003); + gr_def(ctx, offset + 0x154, 0x00000004); + gr_def(ctx, offset + 0x158, 0x00000005); + gr_def(ctx, offset + 0x15c, 0x00000006); + gr_def(ctx, offset + 0x160, 0x00000007); + gr_def(ctx, offset + 0x164, 0x00000001); + gr_def(ctx, offset + 0x1a8, 0x000000cf); + if (dev_priv->chipset == 0x50) + offset -= 4; + gr_def(ctx, offset + 0x1d8, 0x00000080); + gr_def(ctx, offset + 0x1dc, 0x00000004); + gr_def(ctx, offset + 0x1e0, 0x00000004); + if (dev_priv->chipset == 0x50) + offset -= 4; + else + gr_def(ctx, offset + 0x1e4, 0x00000003); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + gr_def(ctx, offset + 0x1ec, 0x00000003); + offset += 8; + } + gr_def(ctx, offset + 0x1e8, 0x00000001); + if (dev_priv->chipset == 0x50) + offset -= 4; + gr_def(ctx, offset + 0x1f4, 0x00000012); + gr_def(ctx, offset + 0x1f8, 0x00000010); + gr_def(ctx, offset + 0x1fc, 0x0000000c); + gr_def(ctx, offset + 0x200, 0x00000001); + gr_def(ctx, offset + 0x210, 0x00000004); + gr_def(ctx, offset + 0x214, 0x00000002); + gr_def(ctx, offset + 0x218, 0x00000004); + if (dev_priv->chipset >= 0xa0) + offset += 4; + gr_def(ctx, offset + 0x224, 0x003fffff); + gr_def(ctx, offset + 0x228, 0x00001fff); + if (dev_priv->chipset == 0x50) + offset -= 0x20; + else if (dev_priv->chipset >= 0xa0) { + gr_def(ctx, offset + 0x250, 0x00000001); + gr_def(ctx, offset + 0x254, 0x00000001); + gr_def(ctx, offset + 0x258, 0x00000002); + offset += 0x10; + } + gr_def(ctx, offset + 0x250, 0x00000004); + gr_def(ctx, offset + 0x254, 0x00000014); + gr_def(ctx, offset + 0x258, 0x00000001); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + offset += 4; + gr_def(ctx, offset + 0x264, 0x00000002); + if (dev_priv->chipset >= 0xa0) + offset += 8; + gr_def(ctx, offset + 0x270, 0x00000001); + gr_def(ctx, offset + 0x278, 0x00000002); + gr_def(ctx, offset + 0x27c, 0x00001000); + if (dev_priv->chipset == 0x50) + offset -= 0xc; + else { + gr_def(ctx, offset + 0x280, 0x00000e00); + gr_def(ctx, offset + 0x284, 0x00001000); + gr_def(ctx, offset + 0x288, 0x00001e00); + } + gr_def(ctx, offset + 0x290, 0x00000001); + gr_def(ctx, offset + 0x294, 0x00000001); + gr_def(ctx, offset + 0x298, 0x00000001); + gr_def(ctx, offset + 0x29c, 0x00000001); + gr_def(ctx, offset + 0x2a0, 0x00000001); + gr_def(ctx, offset + 0x2b0, 0x00000200); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + gr_def(ctx, offset + 0x2b4, 0x00000200); + offset += 4; + } + if (dev_priv->chipset < 0xa0) { + gr_def(ctx, offset + 0x2b8, 0x00000001); + gr_def(ctx, offset + 0x2bc, 0x00000070); + gr_def(ctx, offset + 0x2c0, 0x00000080); + gr_def(ctx, offset + 0x2cc, 0x00000001); + gr_def(ctx, offset + 0x2d0, 0x00000070); + gr_def(ctx, offset + 0x2d4, 0x00000080); + } else { + gr_def(ctx, offset + 0x2b8, 0x00000001); + gr_def(ctx, offset + 0x2bc, 0x000000f0); + gr_def(ctx, offset + 0x2c0, 0x000000ff); + gr_def(ctx, offset + 0x2cc, 0x00000001); + gr_def(ctx, offset + 0x2d0, 0x000000f0); + gr_def(ctx, offset + 0x2d4, 0x000000ff); + gr_def(ctx, offset + 0x2dc, 0x00000009); + offset += 4; + } + gr_def(ctx, offset + 0x2e4, 0x00000001); + gr_def(ctx, offset + 0x2e8, 0x000000cf); + gr_def(ctx, offset + 0x2f0, 0x00000001); + gr_def(ctx, offset + 0x300, 0x000000cf); + gr_def(ctx, offset + 0x308, 0x00000002); + gr_def(ctx, offset + 0x310, 0x00000001); + gr_def(ctx, offset + 0x318, 0x00000001); + gr_def(ctx, offset + 0x320, 0x000000cf); + gr_def(ctx, offset + 0x324, 0x000000cf); + gr_def(ctx, offset + 0x328, 0x00000001); + + /* 6000? */ + if (dev_priv->chipset == 0x50) + cp_ctx(ctx, 0x4063e0, 0x1); + + /* 6800: M2MF */ + if (dev_priv->chipset < 0x90) { + cp_ctx(ctx, 0x406814, 0x2b); + gr_def(ctx, 0x406818, 0x00000f80); + gr_def(ctx, 0x406860, 0x007f0080); + gr_def(ctx, 0x40689c, 0x007f0080); + } else { + cp_ctx(ctx, 0x406814, 0x4); + if (dev_priv->chipset == 0x98) + gr_def(ctx, 0x406818, 0x00000f80); + else + gr_def(ctx, 0x406818, 0x00001f80); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + gr_def(ctx, 0x40681c, 0x00000030); + cp_ctx(ctx, 0x406830, 0x3); + } + + /* 7000: per-ROP group state */ + for (i = 0; i < 8; i++) { + if (units & (1<<(i+16))) { + cp_ctx(ctx, 0x407000 + (i<<8), 3); + if (dev_priv->chipset == 0x50) + gr_def(ctx, 0x407000 + (i<<8), 0x1b74f820); + else if (dev_priv->chipset != 0xa5) + gr_def(ctx, 0x407000 + (i<<8), 0x3b74f821); + else + gr_def(ctx, 0x407000 + (i<<8), 0x7b74f821); + gr_def(ctx, 0x407004 + (i<<8), 0x89058001); + + if (dev_priv->chipset == 0x50) { + cp_ctx(ctx, 0x407010 + (i<<8), 1); + } else if (dev_priv->chipset < 0xa0) { + cp_ctx(ctx, 0x407010 + (i<<8), 2); + gr_def(ctx, 0x407010 + (i<<8), 0x00001000); + gr_def(ctx, 0x407014 + (i<<8), 0x0000001f); + } else { + cp_ctx(ctx, 0x407010 + (i<<8), 3); + gr_def(ctx, 0x407010 + (i<<8), 0x00001000); + if (dev_priv->chipset != 0xa5) + gr_def(ctx, 0x407014 + (i<<8), 0x000000ff); + else + gr_def(ctx, 0x407014 + (i<<8), 0x000001ff); + } + + cp_ctx(ctx, 0x407080 + (i<<8), 4); + if (dev_priv->chipset != 0xa5) + gr_def(ctx, 0x407080 + (i<<8), 0x027c10fa); + else + gr_def(ctx, 0x407080 + (i<<8), 0x827c10fa); + if (dev_priv->chipset == 0x50) + gr_def(ctx, 0x407084 + (i<<8), 0x000000c0); + else + gr_def(ctx, 0x407084 + (i<<8), 0x400000c0); + gr_def(ctx, 0x407088 + (i<<8), 0xb7892080); + + if (dev_priv->chipset < 0xa0) + cp_ctx(ctx, 0x407094 + (i<<8), 1); + else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) + cp_ctx(ctx, 0x407094 + (i<<8), 3); + else { + cp_ctx(ctx, 0x407094 + (i<<8), 4); + gr_def(ctx, 0x4070a0 + (i<<8), 1); + } + } + } + + cp_ctx(ctx, 0x407c00, 0x3); + if (dev_priv->chipset < 0x90) + gr_def(ctx, 0x407c00, 0x00010040); + else if (dev_priv->chipset < 0xa0) + gr_def(ctx, 0x407c00, 0x00390040); + else + gr_def(ctx, 0x407c00, 0x003d0040); + gr_def(ctx, 0x407c08, 0x00000022); + if (dev_priv->chipset >= 0xa0) { + cp_ctx(ctx, 0x407c10, 0x3); + cp_ctx(ctx, 0x407c20, 0x1); + cp_ctx(ctx, 0x407c2c, 0x1); + } + + if (dev_priv->chipset < 0xa0) { + cp_ctx(ctx, 0x407d00, 0x9); + } else { + cp_ctx(ctx, 0x407d00, 0x15); + } + if (dev_priv->chipset == 0x98) + gr_def(ctx, 0x407d08, 0x00380040); + else { + if (dev_priv->chipset < 0x90) + gr_def(ctx, 0x407d08, 0x00010040); + else if (dev_priv->chipset < 0xa0) + gr_def(ctx, 0x407d08, 0x00390040); + else + gr_def(ctx, 0x407d08, 0x003d0040); + gr_def(ctx, 0x407d0c, 0x00000022); + } + + /* 8000+: per-TP state */ + for (i = 0; i < 10; i++) { + if (units & (1<chipset < 0xa0) + base = 0x408000 + (i<<12); + else + base = 0x408000 + (i<<11); + if (dev_priv->chipset < 0xa0) + offset = base + 0xc00; + else + offset = base + 0x80; + cp_ctx(ctx, offset + 0x00, 1); + gr_def(ctx, offset + 0x00, 0x0000ff0a); + cp_ctx(ctx, offset + 0x08, 1); + + /* per-MP state */ + for (j = 0; j < (dev_priv->chipset < 0xa0 ? 2 : 4); j++) { + if (!(units & (1 << (j+24)))) continue; + if (dev_priv->chipset < 0xa0) + offset = base + 0x200 + (j<<7); + else + offset = base + 0x100 + (j<<7); + cp_ctx(ctx, offset, 0x20); + gr_def(ctx, offset + 0x00, 0x01800000); + gr_def(ctx, offset + 0x04, 0x00160000); + gr_def(ctx, offset + 0x08, 0x01800000); + gr_def(ctx, offset + 0x18, 0x0003ffff); + switch (dev_priv->chipset) { + case 0x50: + gr_def(ctx, offset + 0x1c, 0x00080000); + break; + case 0x84: + gr_def(ctx, offset + 0x1c, 0x00880000); + break; + case 0x86: + gr_def(ctx, offset + 0x1c, 0x008c0000); + break; + case 0x92: + case 0x96: + case 0x98: + gr_def(ctx, offset + 0x1c, 0x118c0000); + break; + case 0x94: + gr_def(ctx, offset + 0x1c, 0x10880000); + break; + case 0xa0: + case 0xa5: + gr_def(ctx, offset + 0x1c, 0x310c0000); + break; + case 0xa3: + case 0xa8: + case 0xaa: + case 0xac: + gr_def(ctx, offset + 0x1c, 0x300c0000); + break; + } + gr_def(ctx, offset + 0x40, 0x00010401); + if (dev_priv->chipset == 0x50) + gr_def(ctx, offset + 0x48, 0x00000040); + else + gr_def(ctx, offset + 0x48, 0x00000078); + gr_def(ctx, offset + 0x50, 0x000000bf); + gr_def(ctx, offset + 0x58, 0x00001210); + if (dev_priv->chipset == 0x50) + gr_def(ctx, offset + 0x5c, 0x00000080); + else + gr_def(ctx, offset + 0x5c, 0x08000080); + if (dev_priv->chipset >= 0xa0) + gr_def(ctx, offset + 0x68, 0x0000003e); + } + + if (dev_priv->chipset < 0xa0) + cp_ctx(ctx, base + 0x300, 0x4); + else + cp_ctx(ctx, base + 0x300, 0x5); + if (dev_priv->chipset == 0x50) + gr_def(ctx, base + 0x304, 0x00007070); + else if (dev_priv->chipset < 0xa0) + gr_def(ctx, base + 0x304, 0x00027070); + else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) + gr_def(ctx, base + 0x304, 0x01127070); + else + gr_def(ctx, base + 0x304, 0x05127070); + + if (dev_priv->chipset < 0xa0) + cp_ctx(ctx, base + 0x318, 1); + else + cp_ctx(ctx, base + 0x320, 1); + if (dev_priv->chipset == 0x50) + gr_def(ctx, base + 0x318, 0x0003ffff); + else if (dev_priv->chipset < 0xa0) + gr_def(ctx, base + 0x318, 0x03ffffff); + else + gr_def(ctx, base + 0x320, 0x07ffffff); + + if (dev_priv->chipset < 0xa0) + cp_ctx(ctx, base + 0x324, 5); + else + cp_ctx(ctx, base + 0x328, 4); + + if (dev_priv->chipset < 0xa0) { + cp_ctx(ctx, base + 0x340, 9); + offset = base + 0x340; + } else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) { + cp_ctx(ctx, base + 0x33c, 0xb); + offset = base + 0x344; + } else { + cp_ctx(ctx, base + 0x33c, 0xd); + offset = base + 0x344; + } + gr_def(ctx, offset + 0x0, 0x00120407); + gr_def(ctx, offset + 0x4, 0x05091507); + if (dev_priv->chipset == 0x84) + gr_def(ctx, offset + 0x8, 0x05100202); + else + gr_def(ctx, offset + 0x8, 0x05010202); + gr_def(ctx, offset + 0xc, 0x00030201); + if (dev_priv->chipset == 0xa3) + cp_ctx(ctx, base + 0x36c, 1); + + cp_ctx(ctx, base + 0x400, 2); + gr_def(ctx, base + 0x404, 0x00000040); + cp_ctx(ctx, base + 0x40c, 2); + gr_def(ctx, base + 0x40c, 0x0d0c0b0a); + gr_def(ctx, base + 0x410, 0x00141210); + + if (dev_priv->chipset < 0xa0) + offset = base + 0x800; + else + offset = base + 0x500; + cp_ctx(ctx, offset, 6); + gr_def(ctx, offset + 0x0, 0x000001f0); + gr_def(ctx, offset + 0x4, 0x00000001); + gr_def(ctx, offset + 0x8, 0x00000003); + if (dev_priv->chipset == 0x50 || dev_priv->chipset >= 0xaa) + gr_def(ctx, offset + 0xc, 0x00008000); + gr_def(ctx, offset + 0x14, 0x00039e00); + cp_ctx(ctx, offset + 0x1c, 2); + if (dev_priv->chipset == 0x50) + gr_def(ctx, offset + 0x1c, 0x00000040); + else + gr_def(ctx, offset + 0x1c, 0x00000100); + gr_def(ctx, offset + 0x20, 0x00003800); + + if (dev_priv->chipset >= 0xa0) { + cp_ctx(ctx, base + 0x54c, 2); + if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) + gr_def(ctx, base + 0x54c, 0x003fe006); + else + gr_def(ctx, base + 0x54c, 0x003fe007); + gr_def(ctx, base + 0x550, 0x003fe000); + } + + if (dev_priv->chipset < 0xa0) + offset = base + 0xa00; + else + offset = base + 0x680; + cp_ctx(ctx, offset, 1); + gr_def(ctx, offset, 0x00404040); + + if (dev_priv->chipset < 0xa0) + offset = base + 0xe00; + else + offset = base + 0x700; + cp_ctx(ctx, offset, 2); + if (dev_priv->chipset < 0xa0) + gr_def(ctx, offset, 0x0077f005); + else if (dev_priv->chipset == 0xa5) + gr_def(ctx, offset, 0x6cf7f007); + else if (dev_priv->chipset == 0xa8) + gr_def(ctx, offset, 0x6cfff007); + else if (dev_priv->chipset == 0xac) + gr_def(ctx, offset, 0x0cfff007); + else + gr_def(ctx, offset, 0x0cf7f007); + if (dev_priv->chipset == 0x50) + gr_def(ctx, offset + 0x4, 0x00007fff); + else if (dev_priv->chipset < 0xa0) + gr_def(ctx, offset + 0x4, 0x003f7fff); + else + gr_def(ctx, offset + 0x4, 0x02bf7fff); + cp_ctx(ctx, offset + 0x2c, 1); + if (dev_priv->chipset == 0x50) { + cp_ctx(ctx, offset + 0x50, 9); + gr_def(ctx, offset + 0x54, 0x000003ff); + gr_def(ctx, offset + 0x58, 0x00000003); + gr_def(ctx, offset + 0x5c, 0x00000003); + gr_def(ctx, offset + 0x60, 0x000001ff); + gr_def(ctx, offset + 0x64, 0x0000001f); + gr_def(ctx, offset + 0x68, 0x0000000f); + gr_def(ctx, offset + 0x6c, 0x0000000f); + } else if(dev_priv->chipset < 0xa0) { + cp_ctx(ctx, offset + 0x50, 1); + cp_ctx(ctx, offset + 0x70, 1); + } else { + cp_ctx(ctx, offset + 0x50, 1); + cp_ctx(ctx, offset + 0x60, 5); + } + } + } +} + +/* + * xfer areas. These are a pain. + * + * There are 2 xfer areas: the first one is big and contains all sorts of + * stuff, the second is small and contains some per-TP context. + * + * Each area is split into 8 "strands". The areas, when saved to grctx, + * are made of 8-word blocks. Each block contains a single word from + * each strand. The strands are independent of each other, their + * addresses are unrelated to each other, and data in them is closely + * packed together. The strand layout varies a bit between cards: here + * and there, a single word is thrown out in the middle and the whole + * strand is offset by a bit from corresponding one on another chipset. + * For this reason, addresses of stuff in strands are almost useless. + * Knowing sequence of stuff and size of gaps between them is much more + * useful, and that's how we build the strands in our generator. + * + * NVA0 takes this mess to a whole new level by cutting the old strands + * into a few dozen pieces [known as genes], rearranging them randomly, + * and putting them back together to make new strands. Hopefully these + * genes correspond more or less directly to the same PGRAPH subunits + * as in 400040 register. + * + * The most common value in default context is 0, and when the genes + * are separated by 0's, gene bounduaries are quite speculative... + * some of them can be clearly deduced, others can be guessed, and yet + * others won't be resolved without figuring out the real meaning of + * given ctxval. For the same reason, ending point of each strand + * is unknown. Except for strand 0, which is the longest strand and + * its end corresponds to end of the whole xfer. + * + * An unsolved mystery is the seek instruction: it takes an argument + * in bits 8-18, and that argument is clearly the place in strands to + * seek to... but the offsets don't seem to correspond to offsets as + * seen in grctx. Perhaps there's another, real, not randomly-changing + * addressing in strands, and the xfer insn just happens to skip over + * the unused bits? NV10-NV30 PIPE comes to mind... + * + * As far as I know, there's no way to access the xfer areas directly + * without the help of ctxprog. + */ + +static inline void +xf_emit(struct nouveau_grctx *ctx, int num, uint32_t val) { + int i; + if (val && ctx->mode == NOUVEAU_GRCTX_VALS) + for (i = 0; i < num; i++) + nv_wo32(ctx->dev, ctx->data, ctx->ctxvals_pos + (i << 3), val); + ctx->ctxvals_pos += num << 3; +} + +/* Gene declarations... */ + +static void nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk1(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk2(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk3(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk4(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk5(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk6(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk7(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk8(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk9(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_unk10(struct nouveau_grctx *ctx); +static void nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx); +static void nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx); + +static void +nv50_graph_construct_xfer1(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int i; + int offset; + int size = 0; + uint32_t units = nv_rd32 (ctx->dev, 0x1540); + + offset = (ctx->ctxvals_pos+0x3f)&~0x3f; + ctx->ctxvals_base = offset; + + if (dev_priv->chipset < 0xa0) { + /* Strand 0 */ + ctx->ctxvals_pos = offset; + switch (dev_priv->chipset) { + case 0x50: + xf_emit(ctx, 0x99, 0); + break; + case 0x84: + case 0x86: + xf_emit(ctx, 0x384, 0); + break; + case 0x92: + case 0x94: + case 0x96: + case 0x98: + xf_emit(ctx, 0x380, 0); + break; + } + nv50_graph_construct_gene_m2mf (ctx); + switch (dev_priv->chipset) { + case 0x50: + case 0x84: + case 0x86: + case 0x98: + xf_emit(ctx, 0x4c4, 0); + break; + case 0x92: + case 0x94: + case 0x96: + xf_emit(ctx, 0x984, 0); + break; + } + nv50_graph_construct_gene_unk5(ctx); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 0xa, 0); + else + xf_emit(ctx, 0xb, 0); + nv50_graph_construct_gene_unk4(ctx); + nv50_graph_construct_gene_unk3(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 1 */ + ctx->ctxvals_pos = offset + 0x1; + nv50_graph_construct_gene_unk6(ctx); + nv50_graph_construct_gene_unk7(ctx); + nv50_graph_construct_gene_unk8(ctx); + switch (dev_priv->chipset) { + case 0x50: + case 0x92: + xf_emit(ctx, 0xfb, 0); + break; + case 0x84: + xf_emit(ctx, 0xd3, 0); + break; + case 0x94: + case 0x96: + xf_emit(ctx, 0xab, 0); + break; + case 0x86: + case 0x98: + xf_emit(ctx, 0x6b, 0); + break; + } + xf_emit(ctx, 2, 0x4e3bfdf); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 0xb, 0); + xf_emit(ctx, 2, 0x4e3bfdf); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 2 */ + ctx->ctxvals_pos = offset + 0x2; + switch (dev_priv->chipset) { + case 0x50: + case 0x92: + xf_emit(ctx, 0xa80, 0); + break; + case 0x84: + xf_emit(ctx, 0xa7e, 0); + break; + case 0x94: + case 0x96: + xf_emit(ctx, 0xa7c, 0); + break; + case 0x86: + case 0x98: + xf_emit(ctx, 0xa7a, 0); + break; + } + xf_emit(ctx, 1, 0x3fffff); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x1fff); + xf_emit(ctx, 0xe, 0); + nv50_graph_construct_gene_unk9(ctx); + nv50_graph_construct_gene_unk2(ctx); + nv50_graph_construct_gene_unk1(ctx); + nv50_graph_construct_gene_unk10(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 3: per-ROP group state */ + ctx->ctxvals_pos = offset + 3; + for (i = 0; i < 6; i++) + if (units & (1 << (i + 16))) + nv50_graph_construct_gene_ropc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strands 4-7: per-TP state */ + for (i = 0; i < 4; i++) { + ctx->ctxvals_pos = offset + 4 + i; + if (units & (1 << (2 * i))) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << (2 * i + 1))) + nv50_graph_construct_xfer_tp(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + } else { + /* Strand 0 */ + ctx->ctxvals_pos = offset; + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0x385, 0); + else + xf_emit(ctx, 0x384, 0); + nv50_graph_construct_gene_m2mf(ctx); + xf_emit(ctx, 0x950, 0); + nv50_graph_construct_gene_unk10(ctx); + xf_emit(ctx, 1, 0x0fac6881); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 1, 1); + xf_emit(ctx, 3, 0); + } + nv50_graph_construct_gene_unk8(ctx); + if (dev_priv->chipset == 0xa0) + xf_emit(ctx, 0x189, 0); + else if (dev_priv->chipset == 0xa3) + xf_emit(ctx, 0xd5, 0); + else if (dev_priv->chipset == 0xa5) + xf_emit(ctx, 0x99, 0); + else if (dev_priv->chipset == 0xaa) + xf_emit(ctx, 0x65, 0); + else + xf_emit(ctx, 0x6d, 0); + nv50_graph_construct_gene_unk9(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 1 */ + ctx->ctxvals_pos = offset + 1; + nv50_graph_construct_gene_unk1(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 2 */ + ctx->ctxvals_pos = offset + 2; + if (dev_priv->chipset == 0xa0) { + nv50_graph_construct_gene_unk2(ctx); + } + xf_emit(ctx, 0x36, 0); + nv50_graph_construct_gene_unk5(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 3 */ + ctx->ctxvals_pos = offset + 3; + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + nv50_graph_construct_gene_unk6(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 4 */ + ctx->ctxvals_pos = offset + 4; + if (dev_priv->chipset == 0xa0) + xf_emit(ctx, 0xa80, 0); + else if (dev_priv->chipset == 0xa3) + xf_emit(ctx, 0xa7c, 0); + else + xf_emit(ctx, 0xa7a, 0); + xf_emit(ctx, 1, 0x3fffff); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x1fff); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 5 */ + ctx->ctxvals_pos = offset + 5; + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 0xb, 0); + xf_emit(ctx, 2, 0x4e3bfdf); + xf_emit(ctx, 3, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 2, 0x4e3bfdf); + xf_emit(ctx, 2, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 0); + for (i = 0; i < 8; i++) + if (units & (1<<(i+16))) + nv50_graph_construct_gene_ropc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 6 */ + ctx->ctxvals_pos = offset + 6; + nv50_graph_construct_gene_unk3(ctx); + xf_emit(ctx, 0xb, 0); + nv50_graph_construct_gene_unk4(ctx); + nv50_graph_construct_gene_unk7(ctx); + if (units & (1 << 0)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 1)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 2)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 3)) + nv50_graph_construct_xfer_tp(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 7 */ + ctx->ctxvals_pos = offset + 7; + if (dev_priv->chipset == 0xa0) { + if (units & (1 << 4)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 5)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 6)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 7)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 8)) + nv50_graph_construct_xfer_tp(ctx); + if (units & (1 << 9)) + nv50_graph_construct_xfer_tp(ctx); + } else { + nv50_graph_construct_gene_unk2(ctx); + } + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + + ctx->ctxvals_pos = offset + size * 8; + ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f; + cp_lsr (ctx, offset); + cp_out (ctx, CP_SET_XFER_POINTER); + cp_lsr (ctx, size); + cp_out (ctx, CP_SEEK_1); + cp_out (ctx, CP_XFER_1); + cp_wait(ctx, XFER, BUSY); +} + +/* + * non-trivial demagiced parts of ctx init go here + */ + +static void +nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx) +{ + /* m2mf state */ + xf_emit (ctx, 1, 0); /* DMA_NOTIFY instance >> 4 */ + xf_emit (ctx, 1, 0); /* DMA_BUFFER_IN instance >> 4 */ + xf_emit (ctx, 1, 0); /* DMA_BUFFER_OUT instance >> 4 */ + xf_emit (ctx, 1, 0); /* OFFSET_IN */ + xf_emit (ctx, 1, 0); /* OFFSET_OUT */ + xf_emit (ctx, 1, 0); /* PITCH_IN */ + xf_emit (ctx, 1, 0); /* PITCH_OUT */ + xf_emit (ctx, 1, 0); /* LINE_LENGTH */ + xf_emit (ctx, 1, 0); /* LINE_COUNT */ + xf_emit (ctx, 1, 0x21); /* FORMAT: bits 0-4 INPUT_INC, bits 5-9 OUTPUT_INC */ + xf_emit (ctx, 1, 1); /* LINEAR_IN */ + xf_emit (ctx, 1, 0x2); /* TILING_MODE_IN: bits 0-2 y tiling, bits 3-5 z tiling */ + xf_emit (ctx, 1, 0x100); /* TILING_PITCH_IN */ + xf_emit (ctx, 1, 0x100); /* TILING_HEIGHT_IN */ + xf_emit (ctx, 1, 1); /* TILING_DEPTH_IN */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_IN_Z */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_IN */ + xf_emit (ctx, 1, 1); /* LINEAR_OUT */ + xf_emit (ctx, 1, 0x2); /* TILING_MODE_OUT: bits 0-2 y tiling, bits 3-5 z tiling */ + xf_emit (ctx, 1, 0x100); /* TILING_PITCH_OUT */ + xf_emit (ctx, 1, 0x100); /* TILING_HEIGHT_OUT */ + xf_emit (ctx, 1, 1); /* TILING_DEPTH_OUT */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_OUT_Z */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_OUT */ + xf_emit (ctx, 1, 0); /* OFFSET_IN_HIGH */ + xf_emit (ctx, 1, 0); /* OFFSET_OUT_HIGH */ +} + +static void +nv50_graph_construct_gene_unk1(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + /* end of area 2 on pre-NVA0, area 1 on NVAx */ + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x80); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0x80c14); + xf_emit(ctx, 1, 0); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); + else + xf_emit(ctx, 1, 0x7ff); + switch (dev_priv->chipset) { + case 0x50: + case 0x86: + case 0x98: + case 0xaa: + case 0xac: + xf_emit(ctx, 0x542, 0); + break; + case 0x84: + case 0x92: + case 0x94: + case 0x96: + xf_emit(ctx, 0x942, 0); + break; + case 0xa0: + case 0xa3: + xf_emit(ctx, 0x2042, 0); + break; + case 0xa5: + case 0xa8: + xf_emit(ctx, 0x842, 0); + break; + } + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x80); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x27); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x26); + xf_emit(ctx, 3, 0); +} + +static void +nv50_graph_construct_gene_unk10(struct nouveau_grctx *ctx) +{ + /* end of area 2 on pre-NVA0, area 1 on NVAx */ + xf_emit(ctx, 0x10, 0x04000000); + xf_emit(ctx, 0x24, 0); + xf_emit(ctx, 2, 0x04e3bfdf); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x1fe21); +} + +static void +nv50_graph_construct_gene_unk2(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + /* middle of area 2 on pre-NVA0, beginning of area 2 on NVA0, area 7 on >NVA0 */ + if (dev_priv->chipset != 0x50) { + xf_emit(ctx, 5, 0); + xf_emit(ctx, 1, 0x80c14); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x804); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0x8100c12); + } + xf_emit(ctx, 1, 0); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x10); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 3, 0); + else + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0x804); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x1a); + if (dev_priv->chipset != 0x50) + xf_emit(ctx, 1, 0x7f); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x80c14); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x8100c12); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x10); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x8100c12); + xf_emit(ctx, 6, 0); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); + else + xf_emit(ctx, 1, 0x7ff); + xf_emit(ctx, 1, 0x80c14); + xf_emit(ctx, 0x38, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x10); + xf_emit(ctx, 0x38, 0); + xf_emit(ctx, 2, 0x88); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 0x16, 0); + xf_emit(ctx, 1, 0x26); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x3f800000); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 4, 0); + else + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x1a); + xf_emit(ctx, 1, 0x10); + if (dev_priv->chipset != 0x50) + xf_emit(ctx, 0x28, 0); + else + xf_emit(ctx, 0x25, 0); + xf_emit(ctx, 1, 0x52); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x26); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x1a); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x00ffff00); + xf_emit(ctx, 1, 0); +} + +static void +nv50_graph_construct_gene_unk3(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + /* end of area 0 on pre-NVA0, beginning of area 6 on NVAx */ + xf_emit(ctx, 1, 0x3f); + xf_emit(ctx, 0xa, 0); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 2, 0x04000000); + xf_emit(ctx, 8, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 4); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 0x10, 0); + else + xf_emit(ctx, 0x11, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x1001); + xf_emit(ctx, 4, 0xffff); + xf_emit(ctx, 0x20, 0); + xf_emit(ctx, 0x10, 0x3f800000); + xf_emit(ctx, 1, 0x10); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 1, 0); + else + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 3); + xf_emit(ctx, 2, 0); +} + +static void +nv50_graph_construct_gene_unk4(struct nouveau_grctx *ctx) +{ + /* middle of area 0 on pre-NVA0, middle of area 6 on NVAx */ + xf_emit(ctx, 2, 0x04000000); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x80); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x80); + xf_emit(ctx, 1, 0); +} + +static void +nv50_graph_construct_gene_unk5(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + /* middle of area 0 on pre-NVA0 [after m2mf], end of area 2 on NVAx */ + xf_emit(ctx, 2, 4); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0x1c4d, 0); + else + xf_emit(ctx, 0x1c4b, 0); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0x8100c12); + if (dev_priv->chipset != 0x50) + xf_emit(ctx, 1, 3); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x8100c12); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x80c14); + xf_emit(ctx, 1, 1); + if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0x80c14); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x8100c12); + xf_emit(ctx, 1, 0x27); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x3c1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x16, 0); + xf_emit(ctx, 1, 0x8100c12); + xf_emit(ctx, 1, 0); +} + +static void +nv50_graph_construct_gene_unk6(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + /* beginning of area 1 on pre-NVA0 [after m2mf], area 3 on NVAx */ + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0xf); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 8, 0); + else + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0x20); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0x11, 0); + else if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 0xf, 0); + else + xf_emit(ctx, 0xe, 0); + xf_emit(ctx, 1, 0x1a); + xf_emit(ctx, 0xd, 0); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 8); + xf_emit(ctx, 1, 0); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); + else + xf_emit(ctx, 1, 0x7ff); + if (dev_priv->chipset == 0xa8) + xf_emit(ctx, 1, 0x1e00); + xf_emit(ctx, 0xc, 0); + xf_emit(ctx, 1, 0xf); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 0x125, 0); + else if (dev_priv->chipset < 0xa0) + xf_emit(ctx, 0x126, 0); + else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa) + xf_emit(ctx, 0x124, 0); + else + xf_emit(ctx, 0x1f7, 0); + xf_emit(ctx, 1, 0xf); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 3, 0); + else + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0xa1, 0); + else + xf_emit(ctx, 0x5a, 0); + xf_emit(ctx, 1, 0xf); + if (dev_priv->chipset < 0xa0) + xf_emit(ctx, 0x834, 0); + else if (dev_priv->chipset == 0xa0) + xf_emit(ctx, 0x1873, 0); + else if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0x8ba, 0); + else + xf_emit(ctx, 0x833, 0); + xf_emit(ctx, 1, 0xf); + xf_emit(ctx, 0xf, 0); +} + +static void +nv50_graph_construct_gene_unk7(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + /* middle of area 1 on pre-NVA0 [after m2mf], middle of area 6 on NVAx */ + xf_emit(ctx, 2, 0); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 2, 1); + else + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0x100); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 8); + xf_emit(ctx, 5, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 3, 1); + xf_emit(ctx, 1, 0xcf); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 6, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 3, 1); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x15); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x4444480); + xf_emit(ctx, 0x37, 0); +} + +static void +nv50_graph_construct_gene_unk8(struct nouveau_grctx *ctx) +{ + /* middle of area 1 on pre-NVA0 [after m2mf], middle of area 0 on NVAx */ + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0x8100c12); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0x100); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x10001); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x10001); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x10001); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 2); +} + +static void +nv50_graph_construct_gene_unk9(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + /* middle of area 2 on pre-NVA0 [after m2mf], end of area 0 on NVAx */ + xf_emit(ctx, 1, 0x3f800000); + xf_emit(ctx, 6, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0x1a); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x12, 0); + xf_emit(ctx, 1, 0x00ffff00); + xf_emit(ctx, 6, 0); + xf_emit(ctx, 1, 0xf); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 0xf, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 2, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 3); + else if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 2, 0x04000000); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 5); + xf_emit(ctx, 1, 0x52); + if (dev_priv->chipset == 0x50) { + xf_emit(ctx, 0x13, 0); + } else { + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 1); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0x11, 0); + else + xf_emit(ctx, 0x10, 0); + } + xf_emit(ctx, 0x10, 0x3f800000); + xf_emit(ctx, 1, 0x10); + xf_emit(ctx, 0x26, 0); + xf_emit(ctx, 1, 0x8100c12); + xf_emit(ctx, 1, 5); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 4, 0xffff); + if (dev_priv->chipset != 0x50) + xf_emit(ctx, 1, 3); + if (dev_priv->chipset < 0xa0) + xf_emit(ctx, 0x1f, 0); + else if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0xc, 0); + else + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x00ffff00); + xf_emit(ctx, 1, 0x1a); + if (dev_priv->chipset != 0x50) { + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 3); + } + if (dev_priv->chipset < 0xa0) + xf_emit(ctx, 0x26, 0); + else + xf_emit(ctx, 0x3c, 0); + xf_emit(ctx, 1, 0x102); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 4, 4); + if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 8, 0); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); + else + xf_emit(ctx, 1, 0x7ff); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x102); + xf_emit(ctx, 9, 0); + xf_emit(ctx, 4, 4); + xf_emit(ctx, 0x2c, 0); +} + +static void +nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int magic2; + if (dev_priv->chipset == 0x50) { + magic2 = 0x00003e60; + } else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) { + magic2 = 0x001ffe67; + } else { + magic2 = 0x00087e67; + } + xf_emit(ctx, 8, 0); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, magic2); + xf_emit(ctx, 4, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 7, 0); + if (dev_priv->chipset >= 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 0x15); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x10); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 4, 0); + if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x92 || dev_priv->chipset == 0x98 || dev_priv->chipset >= 0xa0) { + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0x400); + xf_emit(ctx, 1, 0x300); + xf_emit(ctx, 1, 0x1001); + if (dev_priv->chipset != 0xa0) { + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 0); + else + xf_emit(ctx, 1, 0x15); + } + xf_emit(ctx, 3, 0); + } + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 8, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x10); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x13, 0); + xf_emit(ctx, 1, 0x10); + xf_emit(ctx, 0x10, 0); + xf_emit(ctx, 0x10, 0x3f800000); + xf_emit(ctx, 0x19, 0); + xf_emit(ctx, 1, 0x10); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x3f); + xf_emit(ctx, 6, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + if (dev_priv->chipset >= 0xa0) { + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x1001); + xf_emit(ctx, 0xb, 0); + } else { + xf_emit(ctx, 0xc, 0); + } + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0xf); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x11); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 4, 0); + else + xf_emit(ctx, 6, 0); + xf_emit(ctx, 3, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, magic2); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x0fac6881); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 1, 0); + xf_emit(ctx, 0x18, 1); + xf_emit(ctx, 8, 2); + xf_emit(ctx, 8, 1); + xf_emit(ctx, 8, 2); + xf_emit(ctx, 8, 1); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 5, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x16, 0); + } else { + if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 0x1b, 0); + else + xf_emit(ctx, 0x15, 0); + } + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 2, 1); + if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 4, 0); + else + xf_emit(ctx, 3, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 0x10, 1); + xf_emit(ctx, 8, 2); + xf_emit(ctx, 0x10, 1); + xf_emit(ctx, 8, 2); + xf_emit(ctx, 8, 1); + xf_emit(ctx, 3, 0); + } + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x5b, 0); +} + +static void +nv50_graph_construct_xfer_tp_x1(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int magic3; + if (dev_priv->chipset == 0x50) + magic3 = 0x1000; + else if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x98 || dev_priv->chipset >= 0xa8) + magic3 = 0x1e00; + else + magic3 = 0; + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 4); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0x24, 0); + else if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 0x14, 0); + else + xf_emit(ctx, 0x15, 0); + xf_emit(ctx, 2, 4); + if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 1, 0x03020100); + else + xf_emit(ctx, 1, 0x00608080); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 2, 4); + xf_emit(ctx, 1, 0x80); + if (magic3) + xf_emit(ctx, 1, magic3); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 0x24, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0x80); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0x03020100); + xf_emit(ctx, 1, 3); + if (magic3) + xf_emit(ctx, 1, magic3); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 3); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 4); + if (dev_priv->chipset == 0x94 || dev_priv->chipset == 0x96) + xf_emit(ctx, 0x1024, 0); + else if (dev_priv->chipset < 0xa0) + xf_emit(ctx, 0xa24, 0); + else if (dev_priv->chipset == 0xa0 || dev_priv->chipset >= 0xaa) + xf_emit(ctx, 0x214, 0); + else + xf_emit(ctx, 0x414, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 3); + xf_emit(ctx, 2, 0); +} + +static void +nv50_graph_construct_xfer_tp_x2(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int magic1, magic2; + if (dev_priv->chipset == 0x50) { + magic1 = 0x3ff; + magic2 = 0x00003e60; + } else if (dev_priv->chipset <= 0xa0 || dev_priv->chipset >= 0xaa) { + magic1 = 0x7ff; + magic2 = 0x001ffe67; + } else { + magic1 = 0x7ff; + magic2 = 0x00087e67; + } + xf_emit(ctx, 3, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0xc, 0); + xf_emit(ctx, 1, 0xf); + xf_emit(ctx, 0xb, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 4, 0xffff); + xf_emit(ctx, 8, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 5, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 1, 3); + xf_emit(ctx, 1, 0); + } else if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0xa, 0); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 1, 2); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 1, 0); + xf_emit(ctx, 0x18, 1); + xf_emit(ctx, 8, 2); + xf_emit(ctx, 8, 1); + xf_emit(ctx, 8, 2); + xf_emit(ctx, 8, 1); + xf_emit(ctx, 1, 0); + } + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 3, 0xcf); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0xa, 0); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 8, 1); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 1, 0xf); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, magic2); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x11); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 2, 1); + else + xf_emit(ctx, 1, 1); + if(dev_priv->chipset == 0x50) + xf_emit(ctx, 1, 0); + else + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 5, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, magic1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x28, 0); + xf_emit(ctx, 8, 8); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 8, 0x400); + xf_emit(ctx, 8, 0x300); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0xf); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x20); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 0x100); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x40); + xf_emit(ctx, 1, 0x100); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 3); + xf_emit(ctx, 4, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, magic2); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 9, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x400); + xf_emit(ctx, 1, 0x300); + xf_emit(ctx, 1, 0x1001); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 4, 0); + else + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 1, 0xf); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 0x15, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 3, 0); + } else + xf_emit(ctx, 0x17, 0); + if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 1, 0x0fac6881); + xf_emit(ctx, 1, magic2); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 3, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 2, 1); + else + xf_emit(ctx, 1, 1); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 2, 0); + else if (dev_priv->chipset != 0x50) + xf_emit(ctx, 1, 0); +} + +static void +nv50_graph_construct_xfer_tp_x3(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 2, 0); + else + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0x2a712488); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x4085c000); + xf_emit(ctx, 1, 0x40); + xf_emit(ctx, 1, 0x100); + xf_emit(ctx, 1, 0x10100); + xf_emit(ctx, 1, 0x02800000); +} + +static void +nv50_graph_construct_xfer_tp_x4(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + xf_emit(ctx, 2, 0x04e3bfdf); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x00ffff00); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 2, 1); + else + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x00ffff00); + xf_emit(ctx, 8, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0x30201000); + xf_emit(ctx, 1, 0x70605040); + xf_emit(ctx, 1, 0xb8a89888); + xf_emit(ctx, 1, 0xf8e8d8c8); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x1a); +} + +static void +nv50_graph_construct_xfer_tp_x5(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 0xfac6881); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 1); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 0xb, 0); + else + xf_emit(ctx, 0xa, 0); + xf_emit(ctx, 8, 1); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0xfac6881); + xf_emit(ctx, 1, 0xf); + xf_emit(ctx, 7, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 1); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 6, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 6, 0); + } else { + xf_emit(ctx, 0xb, 0); + } +} + +static void +nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + if (dev_priv->chipset < 0xa0) { + nv50_graph_construct_xfer_tp_x1(ctx); + nv50_graph_construct_xfer_tp_x2(ctx); + nv50_graph_construct_xfer_tp_x3(ctx); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 0xf, 0); + else + xf_emit(ctx, 0x12, 0); + nv50_graph_construct_xfer_tp_x4(ctx); + } else { + nv50_graph_construct_xfer_tp_x3(ctx); + if (dev_priv->chipset < 0xaa) + xf_emit(ctx, 0xc, 0); + else + xf_emit(ctx, 0xa, 0); + nv50_graph_construct_xfer_tp_x2(ctx); + nv50_graph_construct_xfer_tp_x5(ctx); + nv50_graph_construct_xfer_tp_x4(ctx); + nv50_graph_construct_xfer_tp_x1(ctx); + } +} + +static void +nv50_graph_construct_xfer_tp2(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int i, mpcnt; + if (dev_priv->chipset == 0x98 || dev_priv->chipset == 0xaa) + mpcnt = 1; + else if (dev_priv->chipset < 0xa0 || dev_priv->chipset >= 0xa8) + mpcnt = 2; + else + mpcnt = 3; + for (i = 0; i < mpcnt; i++) { + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x80); + xf_emit(ctx, 1, 0x80007004); + xf_emit(ctx, 1, 0x04000400); + if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 1, 0xc0); + xf_emit(ctx, 1, 0x1000); + xf_emit(ctx, 2, 0); + if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x98 || dev_priv->chipset >= 0xa8) { + xf_emit(ctx, 1, 0xe00); + xf_emit(ctx, 1, 0x1e00); + } + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 2, 0x1000); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 2); + if (dev_priv->chipset >= 0xaa) + xf_emit(ctx, 0xb, 0); + else if (dev_priv->chipset >= 0xa0) + xf_emit(ctx, 0xc, 0); + else + xf_emit(ctx, 0xa, 0); + } + xf_emit(ctx, 1, 0x08100c12); + xf_emit(ctx, 1, 0); + if (dev_priv->chipset >= 0xa0) { + xf_emit(ctx, 1, 0x1fe21); + } + xf_emit(ctx, 5, 0); + xf_emit(ctx, 4, 0xffff); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 2, 0x10001); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 0x1fe21); + xf_emit(ctx, 1, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 1); + xf_emit(ctx, 4, 0); + xf_emit(ctx, 1, 0x08100c12); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 8, 0); + xf_emit(ctx, 1, 0xfac6881); + xf_emit(ctx, 1, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) + xf_emit(ctx, 1, 3); + xf_emit(ctx, 3, 0); + xf_emit(ctx, 1, 4); + xf_emit(ctx, 9, 0); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 2, 1); + xf_emit(ctx, 1, 2); + xf_emit(ctx, 3, 1); + xf_emit(ctx, 1, 0); + if (dev_priv->chipset > 0xa0 && dev_priv->chipset < 0xaa) { + xf_emit(ctx, 8, 2); + xf_emit(ctx, 0x10, 1); + xf_emit(ctx, 8, 2); + xf_emit(ctx, 0x18, 1); + xf_emit(ctx, 3, 0); + } + xf_emit(ctx, 1, 4); + if (dev_priv->chipset == 0x50) + xf_emit(ctx, 0x3a0, 0); + else if (dev_priv->chipset < 0x94) + xf_emit(ctx, 0x3a2, 0); + else if (dev_priv->chipset == 0x98 || dev_priv->chipset == 0xaa) + xf_emit(ctx, 0x39f, 0); + else + xf_emit(ctx, 0x3a3, 0); + xf_emit(ctx, 1, 0x11); + xf_emit(ctx, 1, 0); + xf_emit(ctx, 1, 1); + xf_emit(ctx, 0x2d, 0); +} + +static void +nv50_graph_construct_xfer2(struct nouveau_grctx *ctx) +{ + struct drm_nouveau_private *dev_priv = ctx->dev->dev_private; + int i; + uint32_t offset; + uint32_t units = nv_rd32 (ctx->dev, 0x1540); + int size = 0; + + offset = (ctx->ctxvals_pos+0x3f)&~0x3f; + + if (dev_priv->chipset < 0xa0) { + for (i = 0; i < 8; i++) { + ctx->ctxvals_pos = offset + i; + if (i == 0) + xf_emit(ctx, 1, 0x08100c12); + if (units & (1 << i)) + nv50_graph_construct_xfer_tp2(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + } else { + /* Strand 0: TPs 0, 1 */ + ctx->ctxvals_pos = offset; + xf_emit(ctx, 1, 0x08100c12); + if (units & (1 << 0)) + nv50_graph_construct_xfer_tp2(ctx); + if (units & (1 << 1)) + nv50_graph_construct_xfer_tp2(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 0: TPs 2, 3 */ + ctx->ctxvals_pos = offset + 1; + if (units & (1 << 2)) + nv50_graph_construct_xfer_tp2(ctx); + if (units & (1 << 3)) + nv50_graph_construct_xfer_tp2(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 0: TPs 4, 5, 6 */ + ctx->ctxvals_pos = offset + 2; + if (units & (1 << 4)) + nv50_graph_construct_xfer_tp2(ctx); + if (units & (1 << 5)) + nv50_graph_construct_xfer_tp2(ctx); + if (units & (1 << 6)) + nv50_graph_construct_xfer_tp2(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 0: TPs 7, 8, 9 */ + ctx->ctxvals_pos = offset + 3; + if (units & (1 << 7)) + nv50_graph_construct_xfer_tp2(ctx); + if (units & (1 << 8)) + nv50_graph_construct_xfer_tp2(ctx); + if (units & (1 << 9)) + nv50_graph_construct_xfer_tp2(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + ctx->ctxvals_pos = offset + size * 8; + ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f; + cp_lsr (ctx, offset); + cp_out (ctx, CP_SET_XFER_POINTER); + cp_lsr (ctx, size); + cp_out (ctx, CP_SEEK_2); + cp_out (ctx, CP_XFER_2); + cp_wait(ctx, XFER, BUSY); +} diff --git a/drivers/gpu/drm/nouveau/nv50_instmem.c b/drivers/gpu/drm/nouveau/nv50_instmem.c index f0dc4e3..5f21df3 100644 --- a/drivers/gpu/drm/nouveau/nv50_instmem.c +++ b/drivers/gpu/drm/nouveau/nv50_instmem.c @@ -63,9 +63,10 @@ nv50_instmem_init(struct drm_device *dev) struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_channel *chan; uint32_t c_offset, c_size, c_ramfc, c_vmpd, c_base, pt_size; + uint32_t save_nv001700; + uint64_t v; struct nv50_instmem_priv *priv; int ret, i; - uint32_t v, save_nv001700; priv = kzalloc(sizeof(*priv), GFP_KERNEL); if (!priv) @@ -76,17 +77,12 @@ nv50_instmem_init(struct drm_device *dev) for (i = 0x1700; i <= 0x1710; i += 4) priv->save1700[(i-0x1700)/4] = nv_rd32(dev, i); - if (dev_priv->chipset == 0xaa || dev_priv->chipset == 0xac) - dev_priv->vram_sys_base = nv_rd32(dev, 0x100e10) << 12; - else - dev_priv->vram_sys_base = 0; - /* Reserve the last MiB of VRAM, we should probably try to avoid * setting up the below tables over the top of the VBIOS image at * some point. */ dev_priv->ramin_rsvd_vram = 1 << 20; - c_offset = nouveau_mem_fb_amount(dev) - dev_priv->ramin_rsvd_vram; + c_offset = dev_priv->vram_size - dev_priv->ramin_rsvd_vram; c_size = 128 << 10; c_vmpd = ((dev_priv->chipset & 0xf0) == 0x50) ? 0x1400 : 0x200; c_ramfc = ((dev_priv->chipset & 0xf0) == 0x50) ? 0x0 : 0x20; @@ -106,7 +102,7 @@ nv50_instmem_init(struct drm_device *dev) dev_priv->vm_gart_size = NV50_VM_BLOCK; dev_priv->vm_vram_base = dev_priv->vm_gart_base + dev_priv->vm_gart_size; - dev_priv->vm_vram_size = nouveau_mem_fb_amount(dev); + dev_priv->vm_vram_size = dev_priv->vram_size; if (dev_priv->vm_vram_size > NV50_VM_MAX_VRAM) dev_priv->vm_vram_size = NV50_VM_MAX_VRAM; dev_priv->vm_vram_size = roundup(dev_priv->vm_vram_size, NV50_VM_BLOCK); @@ -189,8 +185,8 @@ nv50_instmem_init(struct drm_device *dev) i = 0; while (v < dev_priv->vram_sys_base + c_offset + c_size) { - BAR0_WI32(priv->pramin_pt->gpuobj, i + 0, v); - BAR0_WI32(priv->pramin_pt->gpuobj, i + 4, 0x00000000); + BAR0_WI32(priv->pramin_pt->gpuobj, i + 0, lower_32_bits(v)); + BAR0_WI32(priv->pramin_pt->gpuobj, i + 4, upper_32_bits(v)); v += 0x1000; i += 8; } @@ -390,7 +386,7 @@ nv50_instmem_populate(struct drm_device *dev, struct nouveau_gpuobj *gpuobj, if (gpuobj->im_backing) return -EINVAL; - *sz = (*sz + (NV50_INSTMEM_PAGE_SIZE-1)) & ~(NV50_INSTMEM_PAGE_SIZE-1); + *sz = ALIGN(*sz, NV50_INSTMEM_PAGE_SIZE); if (*sz == 0) return -EINVAL; diff --git a/drivers/gpu/drm/nouveau/nv50_sor.c b/drivers/gpu/drm/nouveau/nv50_sor.c index c2fff54..b11eaf9 100644 --- a/drivers/gpu/drm/nouveau/nv50_sor.c +++ b/drivers/gpu/drm/nouveau/nv50_sor.c @@ -211,7 +211,7 @@ nv50_sor_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode, mode_ctl = 0x0200; break; case OUTPUT_DP: - mode_ctl |= 0x00050000; + mode_ctl |= (nv_encoder->dp.mc_unknown << 16); if (nv_encoder->dcb->sorconf.link & 1) mode_ctl |= 0x00000800; else @@ -274,6 +274,7 @@ static const struct drm_encoder_funcs nv50_sor_encoder_funcs = { int nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry) { + struct drm_nouveau_private *dev_priv = dev->dev_private; struct nouveau_encoder *nv_encoder = NULL; struct drm_encoder *encoder; bool dum; @@ -319,5 +320,32 @@ nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry) encoder->possible_crtcs = entry->heads; encoder->possible_clones = 0; + if (nv_encoder->dcb->type == OUTPUT_DP) { + int or = nv_encoder->or, link = !(entry->dpconf.sor.link & 1); + uint32_t tmp; + + if (dev_priv->chipset < 0x90 || + dev_priv->chipset == 0x92 || dev_priv->chipset == 0xa0) + tmp = nv_rd32(dev, NV50_PDISPLAY_SOR_MODE_CTRL_C(or)); + else + tmp = nv_rd32(dev, NV90_PDISPLAY_SOR_MODE_CTRL_C(or)); + + switch ((tmp & 0x00000f00) >> 8) { + case 8: + case 9: + nv_encoder->dp.mc_unknown = (tmp & 0x000f0000) >> 16; + tmp = nv_rd32(dev, NV50_SOR_DP_CTRL(or, link)); + nv_encoder->dp.unk0 = tmp & 0x000001fc; + tmp = nv_rd32(dev, NV50_SOR_DP_UNK128(or, link)); + nv_encoder->dp.unk1 = tmp & 0x010f7f3f; + break; + default: + break; + } + + if (!nv_encoder->dp.mc_unknown) + nv_encoder->dp.mc_unknown = 5; + } + return 0; } -- 1.7.0.1