Commit 6796b129 authored by Dave Airlie's avatar Dave Airlie

Merge branch 'linux-4.12' of git://github.com/skeggsb/linux into drm-next

- Re-architecture of the code to handle proprietary fw, more abstracted
to support the multitude of differences that NVIDIA introduce
- Support in the said code for GP10x ACR and GR fw, giving acceleration
support \o/
- Fix for GTX 970 GPUs that are in an odd MMU configuration

* 'linux-4.12' of git://github.com/skeggsb/linux: (60 commits)
  drm/nouveau/fb/gf100-: rework ram detection
  drm/nouveau/fb/gm200: split ram implementation from gm107
  drm/nouveau/fb/gf108: split implementation from gf100
  drm/nouveau/fb/gf100-: modify constructors to allow more customisation
  drm/nouveau/kms/nv50: use drm core i2c-over-aux algorithm
  drm/nouveau/i2c/g94-: return REPLY_M value on reads
  drm/nouveau/i2c: modify aux interface to return length actually transferred
  drm/nouveau/gp10x: enable secboot and GR
  drm/nouveau/gr/gp102: initial support
  drm/nouveau/falcon: support for gp10x msgqueue
  drm/nouveau/secboot: add gp102/gp104/gp106/gp107 support
  drm/nouveau/secboot: put HS code loading code into own file
  drm/nouveau/secboot: support for r375 ACR
  drm/nouveau/secboot: support for r367 ACR
  drm/nouveau/secboot: support for r364 ACR
  drm/nouveau/secboot: workaround bug when starting SEC2 firmware
  drm/nouveau/secboot: support standard NVIDIA HS binaries
  drm/nouveau/secboot: support for unload blob bootloader
  drm/nouveau/secboot: let callers interpret return value of blobs
  drm/nouveau/secboot: support for different load and unload falcons
  ...
parents 2e161017 97e5268d
......@@ -125,6 +125,7 @@
#define MAXWELL_B /* cl9097.h */ 0x0000b197
#define PASCAL_A /* cl9097.h */ 0x0000c097
#define PASCAL_B /* cl9097.h */ 0x0000c197
#define NV74_BSP 0x000074b0
......@@ -163,6 +164,7 @@
#define MAXWELL_COMPUTE_A 0x0000b0c0
#define MAXWELL_COMPUTE_B 0x0000b1c0
#define PASCAL_COMPUTE_A 0x0000c0c0
#define PASCAL_COMPUTE_B 0x0000c1c0
#define NV74_CIPHER 0x000074c1
#endif
......@@ -59,6 +59,7 @@ enum nvkm_devidx {
NVKM_ENGINE_NVDEC,
NVKM_ENGINE_PM,
NVKM_ENGINE_SEC,
NVKM_ENGINE_SEC2,
NVKM_ENGINE_SW,
NVKM_ENGINE_VIC,
NVKM_ENGINE_VP,
......@@ -155,9 +156,10 @@ struct nvkm_device {
struct nvkm_engine *msppp;
struct nvkm_engine *msvld;
struct nvkm_engine *nvenc[3];
struct nvkm_engine *nvdec;
struct nvkm_nvdec *nvdec;
struct nvkm_pm *pm;
struct nvkm_engine *sec;
struct nvkm_sec2 *sec2;
struct nvkm_sw *sw;
struct nvkm_engine *vic;
struct nvkm_engine *vp;
......@@ -225,9 +227,10 @@ struct nvkm_device_chip {
int (*msppp )(struct nvkm_device *, int idx, struct nvkm_engine **);
int (*msvld )(struct nvkm_device *, int idx, struct nvkm_engine **);
int (*nvenc[3])(struct nvkm_device *, int idx, struct nvkm_engine **);
int (*nvdec )(struct nvkm_device *, int idx, struct nvkm_engine **);
int (*nvdec )(struct nvkm_device *, int idx, struct nvkm_nvdec **);
int (*pm )(struct nvkm_device *, int idx, struct nvkm_pm **);
int (*sec )(struct nvkm_device *, int idx, struct nvkm_engine **);
int (*sec2 )(struct nvkm_device *, int idx, struct nvkm_sec2 **);
int (*sw )(struct nvkm_device *, int idx, struct nvkm_sw **);
int (*vic )(struct nvkm_device *, int idx, struct nvkm_engine **);
int (*vp )(struct nvkm_device *, int idx, struct nvkm_engine **);
......
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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 AUTHORS OR COPYRIGHT HOLDERS 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.
*/
#ifndef __NVKM_CORE_MSGQUEUE_H
#define __NVKM_CORE_MSGQUEUE_H
#include <core/os.h>
struct nvkm_falcon;
struct nvkm_msgqueue;
enum nvkm_secboot_falcon;
/* Hopefully we will never have firmware arguments larger than that... */
#define NVKM_MSGQUEUE_CMDLINE_SIZE 0x100
int nvkm_msgqueue_new(u32, struct nvkm_falcon *, struct nvkm_msgqueue **);
void nvkm_msgqueue_del(struct nvkm_msgqueue **);
void nvkm_msgqueue_recv(struct nvkm_msgqueue *);
int nvkm_msgqueue_reinit(struct nvkm_msgqueue *);
/* useful if we run a NVIDIA-signed firmware */
void nvkm_msgqueue_write_cmdline(struct nvkm_msgqueue *, void *);
/* interface to ACR unit running on falcon (NVIDIA signed firmware) */
int nvkm_msgqueue_acr_boot_falcon(struct nvkm_msgqueue *,
enum nvkm_secboot_falcon);
#endif
......@@ -10,6 +10,7 @@ enum nvkm_falcon_dmaidx {
FALCON_DMAIDX_PHYS_VID = 2,
FALCON_DMAIDX_PHYS_SYS_COH = 3,
FALCON_DMAIDX_PHYS_SYS_NCOH = 4,
FALCON_SEC2_DMAIDX_UCODE = 6,
};
struct nvkm_falcon {
......@@ -19,11 +20,13 @@ struct nvkm_falcon {
u32 addr;
struct mutex mutex;
struct mutex dmem_mutex;
const struct nvkm_subdev *user;
u8 version;
u8 secret;
bool debug;
bool has_emem;
struct nvkm_memory *core;
bool external;
......@@ -45,8 +48,14 @@ struct nvkm_falcon {
struct nvkm_engine engine;
};
/* This constructor must be called from the owner's oneinit() hook and
* *not* its constructor. This is to ensure that DEVINIT has been
* completed, and that the device is correctly enabled before we touch
* falcon registers.
*/
int nvkm_falcon_v1_new(struct nvkm_subdev *owner, const char *name, u32 addr,
struct nvkm_falcon **);
void nvkm_falcon_del(struct nvkm_falcon **);
int nvkm_falcon_get(struct nvkm_falcon *, const struct nvkm_subdev *);
void nvkm_falcon_put(struct nvkm_falcon *, const struct nvkm_subdev *);
......
......@@ -43,4 +43,5 @@ int gm107_gr_new(struct nvkm_device *, int, struct nvkm_gr **);
int gm200_gr_new(struct nvkm_device *, int, struct nvkm_gr **);
int gm20b_gr_new(struct nvkm_device *, int, struct nvkm_gr **);
int gp100_gr_new(struct nvkm_device *, int, struct nvkm_gr **);
int gp102_gr_new(struct nvkm_device *, int, struct nvkm_gr **);
#endif
#ifndef __NVKM_NVDEC_H__
#define __NVKM_NVDEC_H__
#define nvkm_nvdec(p) container_of((p), struct nvkm_nvdec, engine)
#include <core/engine.h>
struct nvkm_nvdec {
struct nvkm_engine engine;
struct nvkm_falcon *falcon;
};
int gp102_nvdec_new(struct nvkm_device *, int, struct nvkm_nvdec **);
#endif
#ifndef __NVKM_SEC2_H__
#define __NVKM_SEC2_H__
#include <core/engine.h>
struct nvkm_sec2 {
struct nvkm_engine engine;
struct nvkm_falcon *falcon;
struct nvkm_msgqueue *queue;
struct work_struct work;
};
int gp102_sec2_new(struct nvkm_device *, int, struct nvkm_sec2 **);
#endif
......@@ -89,6 +89,7 @@ int gt215_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
int mcp77_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
int mcp89_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
int gf100_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
int gf108_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
int gk104_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
int gk20a_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
int gm107_fb_new(struct nvkm_device *, int, struct nvkm_fb **);
......@@ -146,6 +147,12 @@ struct nvkm_ram {
};
struct nvkm_ram_func {
u64 upper;
u32 (*probe_fbp)(const struct nvkm_ram_func *, struct nvkm_device *,
int fbp, int *pltcs);
u32 (*probe_fbp_amount)(const struct nvkm_ram_func *, u32 fbpao,
struct nvkm_device *, int fbp, int *pltcs);
u32 (*probe_fbpa_amount)(struct nvkm_device *, int fbpa);
void *(*dtor)(struct nvkm_ram *);
int (*init)(struct nvkm_ram *);
......
......@@ -64,7 +64,7 @@ void nvkm_i2c_aux_monitor(struct nvkm_i2c_aux *, bool monitor);
int nvkm_i2c_aux_acquire(struct nvkm_i2c_aux *);
void nvkm_i2c_aux_release(struct nvkm_i2c_aux *);
int nvkm_i2c_aux_xfer(struct nvkm_i2c_aux *, bool retry, u8 type,
u32 addr, u8 *data, u8 size);
u32 addr, u8 *data, u8 *size);
int nvkm_i2c_aux_lnk_ctl(struct nvkm_i2c_aux *, int link_nr, int link_bw,
bool enhanced_framing);
......@@ -162,9 +162,11 @@ nvkm_probe_i2c(struct i2c_adapter *adap, u8 addr)
static inline int
nvkm_rdaux(struct nvkm_i2c_aux *aux, u32 addr, u8 *data, u8 size)
{
const u8 xfer = size;
int ret = nvkm_i2c_aux_acquire(aux);
if (ret == 0) {
ret = nvkm_i2c_aux_xfer(aux, true, 9, addr, data, size);
ret = nvkm_i2c_aux_xfer(aux, true, 9, addr, data, &size);
WARN_ON(!ret && size != xfer);
nvkm_i2c_aux_release(aux);
}
return ret;
......@@ -175,7 +177,7 @@ nvkm_wraux(struct nvkm_i2c_aux *aux, u32 addr, u8 *data, u8 size)
{
int ret = nvkm_i2c_aux_acquire(aux);
if (ret == 0) {
ret = nvkm_i2c_aux_xfer(aux, true, 8, addr, data, size);
ret = nvkm_i2c_aux_xfer(aux, true, 8, addr, data, &size);
nvkm_i2c_aux_release(aux);
}
return ret;
......
......@@ -7,6 +7,7 @@ struct nvkm_pmu {
const struct nvkm_pmu_func *func;
struct nvkm_subdev subdev;
struct nvkm_falcon *falcon;
struct nvkm_msgqueue *queue;
struct {
u32 base;
......
......@@ -30,10 +30,13 @@ enum nvkm_secboot_falcon {
NVKM_SECBOOT_FALCON_RESERVED = 1,
NVKM_SECBOOT_FALCON_FECS = 2,
NVKM_SECBOOT_FALCON_GPCCS = 3,
NVKM_SECBOOT_FALCON_END = 4,
NVKM_SECBOOT_FALCON_SEC2 = 7,
NVKM_SECBOOT_FALCON_END = 8,
NVKM_SECBOOT_FALCON_INVALID = 0xffffffff,
};
extern const char *nvkm_secboot_falcon_name[];
/**
* @wpr_set: whether the WPR region is currently set
*/
......@@ -42,6 +45,7 @@ struct nvkm_secboot {
struct nvkm_acr *acr;
struct nvkm_subdev subdev;
struct nvkm_falcon *boot_falcon;
struct nvkm_falcon *halt_falcon;
u64 wpr_addr;
u32 wpr_size;
......@@ -55,5 +59,6 @@ int nvkm_secboot_reset(struct nvkm_secboot *, enum nvkm_secboot_falcon);
int gm200_secboot_new(struct nvkm_device *, int, struct nvkm_secboot **);
int gm20b_secboot_new(struct nvkm_device *, int, struct nvkm_secboot **);
int gp102_secboot_new(struct nvkm_device *, int, struct nvkm_secboot **);
#endif
......@@ -1147,6 +1147,7 @@ nouveau_connector_aux_xfer(struct drm_dp_aux *obj, struct drm_dp_aux_msg *msg)
container_of(obj, typeof(*nv_connector), aux);
struct nouveau_encoder *nv_encoder;
struct nvkm_i2c_aux *aux;
u8 size = msg->size;
int ret;
nv_encoder = find_encoder(&nv_connector->base, DCB_OUTPUT_DP);
......@@ -1162,11 +1163,11 @@ nouveau_connector_aux_xfer(struct drm_dp_aux *obj, struct drm_dp_aux_msg *msg)
return ret;
ret = nvkm_i2c_aux_xfer(aux, false, msg->request, msg->address,
msg->buffer, msg->size);
msg->buffer, &size);
nvkm_i2c_aux_release(aux);
if (ret >= 0) {
msg->reply = ret;
return msg->size;
return size;
}
return ret;
......
......@@ -3627,7 +3627,7 @@ nv50_sor_create(struct drm_connector *connector, struct dcb_output *dcbe)
struct nvkm_i2c_aux *aux =
nvkm_i2c_aux_find(i2c, dcbe->i2c_index);
if (aux) {
nv_encoder->i2c = &aux->i2c;
nv_encoder->i2c = &nv_connector->aux.ddc;
nv_encoder->aux = aux;
}
......@@ -3777,6 +3777,7 @@ nv50_pior_func = {
static int
nv50_pior_create(struct drm_connector *connector, struct dcb_output *dcbe)
{
struct nouveau_connector *nv_connector = nouveau_connector(connector);
struct nouveau_drm *drm = nouveau_drm(connector->dev);
struct nvkm_i2c *i2c = nvxx_i2c(&drm->client.device);
struct nvkm_i2c_bus *bus = NULL;
......@@ -3794,7 +3795,7 @@ nv50_pior_create(struct drm_connector *connector, struct dcb_output *dcbe)
break;
case DCB_OUTPUT_DP:
aux = nvkm_i2c_aux_find(i2c, NVKM_I2C_AUX_EXT(dcbe->extdev));
ddc = aux ? &aux->i2c : NULL;
ddc = aux ? &nv_connector->aux.ddc : NULL;
type = DRM_MODE_ENCODER_TMDS;
break;
default:
......
......@@ -78,6 +78,7 @@ nvkm_subdev_name[NVKM_SUBDEV_NR] = {
[NVKM_ENGINE_NVDEC ] = "nvdec",
[NVKM_ENGINE_PM ] = "pm",
[NVKM_ENGINE_SEC ] = "sec",
[NVKM_ENGINE_SEC2 ] = "sec2",
[NVKM_ENGINE_SW ] = "sw",
[NVKM_ENGINE_VIC ] = "vic",
[NVKM_ENGINE_VP ] = "vp",
......
......@@ -18,6 +18,7 @@ include $(src)/nvkm/engine/nvenc/Kbuild
include $(src)/nvkm/engine/nvdec/Kbuild
include $(src)/nvkm/engine/pm/Kbuild
include $(src)/nvkm/engine/sec/Kbuild
include $(src)/nvkm/engine/sec2/Kbuild
include $(src)/nvkm/engine/sw/Kbuild
include $(src)/nvkm/engine/vic/Kbuild
include $(src)/nvkm/engine/vp/Kbuild
......@@ -1379,7 +1379,7 @@ nvc1_chipset = {
.bus = gf100_bus_new,
.clk = gf100_clk_new,
.devinit = gf100_devinit_new,
.fb = gf100_fb_new,
.fb = gf108_fb_new,
.fuse = gf100_fuse_new,
.gpio = g94_gpio_new,
.i2c = g94_i2c_new,
......@@ -2200,6 +2200,9 @@ nv132_chipset = {
.ltc = gp100_ltc_new,
.mc = gp100_mc_new,
.mmu = gf100_mmu_new,
.secboot = gp102_secboot_new,
.sec2 = gp102_sec2_new,
.nvdec = gp102_nvdec_new,
.pci = gp100_pci_new,
.pmu = gp102_pmu_new,
.timer = gk20a_timer_new,
......@@ -2211,6 +2214,8 @@ nv132_chipset = {
.disp = gp102_disp_new,
.dma = gf119_dma_new,
.fifo = gp100_fifo_new,
.gr = gp102_gr_new,
.sw = gf100_sw_new,
};
static const struct nvkm_device_chip
......@@ -2229,6 +2234,9 @@ nv134_chipset = {
.ltc = gp100_ltc_new,
.mc = gp100_mc_new,
.mmu = gf100_mmu_new,
.secboot = gp102_secboot_new,
.sec2 = gp102_sec2_new,
.nvdec = gp102_nvdec_new,
.pci = gp100_pci_new,
.pmu = gp102_pmu_new,
.timer = gk20a_timer_new,
......@@ -2240,6 +2248,8 @@ nv134_chipset = {
.disp = gp102_disp_new,
.dma = gf119_dma_new,
.fifo = gp100_fifo_new,
.gr = gp102_gr_new,
.sw = gf100_sw_new,
};
static const struct nvkm_device_chip
......@@ -2258,6 +2268,9 @@ nv136_chipset = {
.ltc = gp100_ltc_new,
.mc = gp100_mc_new,
.mmu = gf100_mmu_new,
.secboot = gp102_secboot_new,
.sec2 = gp102_sec2_new,
.nvdec = gp102_nvdec_new,
.pci = gp100_pci_new,
.pmu = gp102_pmu_new,
.timer = gk20a_timer_new,
......@@ -2269,6 +2282,8 @@ nv136_chipset = {
.disp = gp102_disp_new,
.dma = gf119_dma_new,
.fifo = gp100_fifo_new,
.gr = gp102_gr_new,
.sw = gf100_sw_new,
};
static int
......@@ -2362,9 +2377,10 @@ nvkm_device_engine(struct nvkm_device *device, int index)
_(NVENC0 , device->nvenc[0], device->nvenc[0]);
_(NVENC1 , device->nvenc[1], device->nvenc[1]);
_(NVENC2 , device->nvenc[2], device->nvenc[2]);
_(NVDEC , device->nvdec , device->nvdec);
_(NVDEC , device->nvdec , &device->nvdec->engine);
_(PM , device->pm , &device->pm->engine);
_(SEC , device->sec , device->sec);
_(SEC2 , device->sec2 , &device->sec2->engine);
_(SW , device->sw , &device->sw->engine);
_(VIC , device->vic , device->vic);
_(VP , device->vp , device->vp);
......@@ -2812,6 +2828,7 @@ nvkm_device_ctor(const struct nvkm_device_func *func,
_(NVKM_ENGINE_NVDEC , nvdec);
_(NVKM_ENGINE_PM , pm);
_(NVKM_ENGINE_SEC , sec);
_(NVKM_ENGINE_SEC2 , sec2);
_(NVKM_ENGINE_SW , sw);
_(NVKM_ENGINE_VIC , vic);
_(NVKM_ENGINE_VP , vp);
......
......@@ -41,6 +41,7 @@
#include <engine/nvdec.h>
#include <engine/pm.h>
#include <engine/sec.h>
#include <engine/sec2.h>
#include <engine/sw.h>
#include <engine/vic.h>
#include <engine/vp.h>
......
......@@ -32,6 +32,7 @@ nvkm-y += nvkm/engine/gr/gm107.o
nvkm-y += nvkm/engine/gr/gm200.o
nvkm-y += nvkm/engine/gr/gm20b.o
nvkm-y += nvkm/engine/gr/gp100.o
nvkm-y += nvkm/engine/gr/gp102.o
nvkm-y += nvkm/engine/gr/ctxnv40.o
nvkm-y += nvkm/engine/gr/ctxnv50.o
......@@ -50,3 +51,4 @@ nvkm-y += nvkm/engine/gr/ctxgm107.o
nvkm-y += nvkm/engine/gr/ctxgm200.o
nvkm-y += nvkm/engine/gr/ctxgm20b.o
nvkm-y += nvkm/engine/gr/ctxgp100.o
nvkm-y += nvkm/engine/gr/ctxgp102.o
......@@ -102,6 +102,10 @@ void gm200_grctx_generate_405b60(struct gf100_gr *);
extern const struct gf100_grctx_func gm20b_grctx;
extern const struct gf100_grctx_func gp100_grctx;
void gp100_grctx_generate_main(struct gf100_gr *, struct gf100_grctx *);
void gp100_grctx_generate_pagepool(struct gf100_grctx *);
extern const struct gf100_grctx_func gp102_grctx;
/* context init value lists */
......
......@@ -29,7 +29,7 @@
* PGRAPH context implementation
******************************************************************************/
static void
void
gp100_grctx_generate_pagepool(struct gf100_grctx *info)
{
const struct gf100_grctx_func *grctx = info->gr->func->grctx;
......@@ -123,7 +123,7 @@ gp100_grctx_generate_405b60(struct gf100_gr *gr)
nvkm_wr32(device, 0x405ba0 + (i * 4), gpcs[i]);
}
static void
void
gp100_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info)
{
struct nvkm_device *device = gr->base.engine.subdev.device;
......
/*
* Copyright 2016 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 <bskeggs@redhat.com>
*/
#include "ctxgf100.h"
#include <subdev/fb.h>
/*******************************************************************************
* PGRAPH context implementation
******************************************************************************/
static void
gp102_grctx_generate_attrib(struct gf100_grctx *info)
{
struct gf100_gr *gr = info->gr;
const struct gf100_grctx_func *grctx = gr->func->grctx;
const u32 alpha = grctx->alpha_nr;
const u32 attrib = grctx->attrib_nr;
const u32 pertpc = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max);
const u32 size = roundup(gr->tpc_total * pertpc, 0x80);
const u32 access = NV_MEM_ACCESS_RW;
const int s = 12;
const int b = mmio_vram(info, size, (1 << s), access);
const int max_batches = 0xffff;
u32 ao = 0;
u32 bo = ao + grctx->alpha_nr_max * gr->tpc_total;
int gpc, ppc, n = 0;
mmio_refn(info, 0x418810, 0x80000000, s, b);
mmio_refn(info, 0x419848, 0x10000000, s, b);
mmio_refn(info, 0x419c2c, 0x10000000, s, b);
mmio_refn(info, 0x419b00, 0x00000000, s, b);
mmio_wr32(info, 0x419b04, 0x80000000 | size >> 7);
mmio_wr32(info, 0x405830, attrib);
mmio_wr32(info, 0x40585c, alpha);
mmio_wr32(info, 0x4064c4, ((alpha / 4) << 16) | max_batches);
for (gpc = 0; gpc < gr->gpc_nr; gpc++) {
for (ppc = 0; ppc < gr->ppc_nr[gpc]; ppc++, n++) {
const u32 as = alpha * gr->ppc_tpc_nr[gpc][ppc];
const u32 bs = attrib * gr->ppc_tpc_nr[gpc][ppc];
const u32 u = 0x418ea0 + (n * 0x04);
const u32 o = PPC_UNIT(gpc, ppc, 0);
const u32 p = GPC_UNIT(gpc, 0xc44 + (ppc * 4));
if (!(gr->ppc_mask[gpc] & (1 << ppc)))
continue;
mmio_wr32(info, o + 0xc0, bs);
mmio_wr32(info, p, bs);
mmio_wr32(info, o + 0xf4, bo);
mmio_wr32(info, o + 0xf0, bs);
bo += grctx->attrib_nr_max * gr->ppc_tpc_nr[gpc][ppc];
mmio_wr32(info, o + 0xe4, as);
mmio_wr32(info, o + 0xf8, ao);
ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc];
mmio_wr32(info, u, bs);
}
}
mmio_wr32(info, 0x4181e4, 0x00000100);
mmio_wr32(info, 0x41befc, 0x00000100);
}
const struct gf100_grctx_func
gp102_grctx = {
.main = gp100_grctx_generate_main,
.unkn = gk104_grctx_generate_unkn,
.bundle = gm107_grctx_generate_bundle,
.bundle_size = 0x3000,
.bundle_min_gpm_fifo_depth = 0x180,
.bundle_token_limit = 0x900,
.pagepool = gp100_grctx_generate_pagepool,
.pagepool_size = 0x20000,
.attrib = gp102_grctx_generate_attrib,
.attrib_nr_max = 0x5d4,
.attrib_nr = 0x320,
.alpha_nr_max = 0xc00,
.alpha_nr = 0x800,
};
......@@ -1647,8 +1647,18 @@ static int
gf100_gr_oneinit(struct nvkm_gr *base)
{
struct gf100_gr *gr = gf100_gr(base);
struct nvkm_device *device = gr->base.engine.subdev.device;
struct nvkm_subdev *subdev = &gr->base.engine.subdev;
struct nvkm_device *device = subdev->device;
int i, j;
int ret;
ret = nvkm_falcon_v1_new(subdev, "FECS", 0x409000, &gr->fecs);
if (ret)
return ret;
ret = nvkm_falcon_v1_new(subdev, "GPCCS", 0x41a000, &gr->gpccs);
if (ret)
return ret;
nvkm_pmu_pgob(device->pmu, false);
......@@ -1856,24 +1866,13 @@ int
gf100_gr_ctor(const struct gf100_gr_func *func, struct nvkm_device *device,
int index, struct gf100_gr *gr)
{
struct nvkm_subdev *subdev = &gr->base.engine.subdev;
int ret;
gr->func = func;
gr->firmware = nvkm_boolopt(device->cfgopt, "NvGrUseFW",
func->fecs.ucode == NULL);
ret = nvkm_gr_ctor(&gf100_gr_, device, index,
gr->firmware || func->fecs.ucode != NULL,
&gr->base);
if (ret)
return ret;
ret = nvkm_falcon_v1_new(subdev, "FECS", 0x409000, &gr->fecs);
if (ret)
return ret;
return nvkm_falcon_v1_new(subdev, "GPCCS", 0x41a000, &gr->gpccs);
return nvkm_gr_ctor(&gf100_gr_, device, index,
gr->firmware || func->fecs.ucode != NULL,
&gr->base);
}
int
......
......@@ -124,6 +124,7 @@ struct gf100_gr_func {
void (*init_gpc_mmu)(struct gf100_gr *);
void (*init_rop_active_fbps)(struct gf100_gr *);
void (*init_ppc_exceptions)(struct gf100_gr *);
void (*init_swdx_pes_mask)(struct gf100_gr *);
void (*set_hww_esr_report_mask)(struct gf100_gr *);
const struct gf100_gr_pack *mmio;
struct {
......@@ -150,6 +151,9 @@ int gk20a_gr_init(struct gf100_gr *);
int gm200_gr_init(struct gf100_gr *);
int gm200_gr_rops(struct gf100_gr *);
int gp100_gr_init(struct gf100_gr *);
void gp100_gr_init_rop_active_fbps(struct gf100_gr *);
#define gf100_gr_chan(p) container_of((p), struct gf100_gr_chan, object)
struct gf100_gr_chan {
......
......@@ -30,7 +30,7 @@
* PGRAPH engine/subdev functions
******************************************************************************/
static void
void
gp100_gr_init_rop_active_fbps(struct gf100_gr *gr)
{
struct nvkm_device *device = gr->base.engine.subdev.device;
......@@ -40,7 +40,7 @@ gp100_gr_init_rop_active_fbps(struct gf100_gr *gr)
nvkm_mask(device, 0x408958, 0x0000000f, fbp_count); /* crop */
}
static int
int
gp100_gr_init(struct gf100_gr *gr)
{
struct nvkm_device *device = gr->base.engine.subdev.device;
......@@ -85,6 +85,8 @@ gp100_gr_init(struct gf100_gr *gr)
nvkm_wr32(device, GPC_BCAST(0x033c), nvkm_rd32(device, 0x100804));
gr->func->init_rop_active_fbps(gr);
if (gr->func->init_swdx_pes_mask)
gr->func->init_swdx_pes_mask(gr);
nvkm_wr32(device, 0x400500, 0x00010001);
nvkm_wr32(device, 0x400100, 0xffffffff);
......
/*
* Copyright 2016 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 <bskeggs@redhat.com>
*/
#include "gf100.h"
#include "ctxgf100.h"
#include <nvif/class.h>
static void
gp102_gr_init_swdx_pes_mask(struct gf100_gr *gr)
{
struct nvkm_device *device = gr->base.engine.subdev.device;
u32 mask = 0, data, gpc;
for (gpc = 0; gpc < gr->gpc_nr; gpc++) {
data = nvkm_rd32(device, GPC_UNIT(gpc, 0x0c50)) & 0x0000000f;
mask |= data << (gpc * 4);
}
nvkm_wr32(device, 0x4181d0, mask);
}
static const struct gf100_gr_func
gp102_gr = {
.init = gp100_gr_init,
.init_gpc_mmu = gm200_gr_init_gpc_mmu,
.init_rop_active_fbps = gp100_gr_init_rop_active_fbps,
.init_ppc_exceptions = gk104_gr_init_ppc_exceptions,
.init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask,
.rops = gm200_gr_rops,
.ppc_nr = 3,
.grctx = &gp102_grctx,
.sclass = {
{ -1, -1, FERMI_TWOD_A },
{ -1, -1, KEPLER_INLINE_TO_MEMORY_B },
{ -1, -1, PASCAL_B, &gf100_fermi },
{ -1, -1, PASCAL_COMPUTE_B },
{}
}
};
int
gp102_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr)
{
return gm200_gr_new_(&gp102_gr, device, index, pgr);
}
#nvkm-y += nvkm/engine/nvdec/base.o
nvkm-y += nvkm/engine/nvdec/base.o
nvkm-y += nvkm/engine/nvdec/gp102.o
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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 AUTHORS OR COPYRIGHT HOLDERS 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.
*/
#include "priv.h"
#include <engine/falcon.h>
static int
nvkm_nvdec_oneinit(struct nvkm_engine *engine)
{
struct nvkm_nvdec *nvdec = nvkm_nvdec(engine);
return nvkm_falcon_v1_new(&nvdec->engine.subdev, "NVDEC", 0x84000,
&nvdec->falcon);
}
static void *
nvkm_nvdec_dtor(struct nvkm_engine *engine)
{
struct nvkm_nvdec *nvdec = nvkm_nvdec(engine);
nvkm_falcon_del(&nvdec->falcon);
return nvdec;
}
static const struct nvkm_engine_func
nvkm_nvdec = {
.dtor = nvkm_nvdec_dtor,
.oneinit = nvkm_nvdec_oneinit,
};
int
nvkm_nvdec_new_(struct nvkm_device *device, int index,
struct nvkm_nvdec **pnvdec)
{
struct nvkm_nvdec *nvdec;
if (!(nvdec = *pnvdec = kzalloc(sizeof(*nvdec), GFP_KERNEL)))
return -ENOMEM;
return nvkm_engine_ctor(&nvkm_nvdec, device, index, true,
&nvdec->engine);
};
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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 AUTHORS OR COPYRIGHT HOLDERS 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.
*/
#include "priv.h"
int
gp102_nvdec_new(struct nvkm_device *device, int index,
struct nvkm_nvdec **pnvdec)
{
return nvkm_nvdec_new_(device, index, pnvdec);
}
#ifndef __NVKM_NVDEC_PRIV_H__
#define __NVKM_NVDEC_PRIV_H__
#include <engine/nvdec.h>
int nvkm_nvdec_new_(struct nvkm_device *, int, struct nvkm_nvdec **);
#endif
nvkm-y += nvkm/engine/sec2/base.o
nvkm-y += nvkm/engine/sec2/gp102.o
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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 AUTHORS OR COPYRIGHT HOLDERS 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.
*/
#include "priv.h"
#include <core/msgqueue.h>
#include <engine/falcon.h>
static void *
nvkm_sec2_dtor(struct nvkm_engine *engine)
{
struct nvkm_sec2 *sec2 = nvkm_sec2(engine);
nvkm_msgqueue_del(&sec2->queue);
nvkm_falcon_del(&sec2->falcon);
return sec2;
}
static void
nvkm_sec2_intr(struct nvkm_engine *engine)
{
struct nvkm_sec2 *sec2 = nvkm_sec2(engine);
struct nvkm_subdev *subdev = &engine->subdev;
struct nvkm_device *device = subdev->device;
u32 disp = nvkm_rd32(device, 0x8701c);
u32 intr = nvkm_rd32(device, 0x87008) & disp & ~(disp >> 16);
if (intr & 0x00000040) {
schedule_work(&sec2->work);
nvkm_wr32(device, 0x87004, 0x00000040);
intr &= ~0x00000040;
}
if (intr) {
nvkm_error(subdev, "unhandled intr %08x\n", intr);
nvkm_wr32(device, 0x87004, intr);
}
}
static void
nvkm_sec2_recv(struct work_struct *work)
{
struct nvkm_sec2 *sec2 = container_of(work, typeof(*sec2), work);
nvkm_msgqueue_recv(sec2->queue);
}
static int
nvkm_sec2_oneinit(struct nvkm_engine *engine)
{
struct nvkm_sec2 *sec2 = nvkm_sec2(engine);
return nvkm_falcon_v1_new(&sec2->engine.subdev, "SEC2", 0x87000,
&sec2->falcon);
}
static int
nvkm_sec2_fini(struct nvkm_engine *engine, bool suspend)
{
struct nvkm_sec2 *sec2 = nvkm_sec2(engine);
flush_work(&sec2->work);
return 0;
}
static const struct nvkm_engine_func
nvkm_sec2 = {
.dtor = nvkm_sec2_dtor,
.oneinit = nvkm_sec2_oneinit,
.fini = nvkm_sec2_fini,
.intr = nvkm_sec2_intr,
};
int
nvkm_sec2_new_(struct nvkm_device *device, int index,
struct nvkm_sec2 **psec2)
{
struct nvkm_sec2 *sec2;
if (!(sec2 = *psec2 = kzalloc(sizeof(*sec2), GFP_KERNEL)))
return -ENOMEM;
INIT_WORK(&sec2->work, nvkm_sec2_recv);
return nvkm_engine_ctor(&nvkm_sec2, device, index, true, &sec2->engine);
};
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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 AUTHORS OR COPYRIGHT HOLDERS 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.
*/
#include "priv.h"
int
gp102_sec2_new(struct nvkm_device *device, int index,
struct nvkm_sec2 **psec2)
{
return nvkm_sec2_new_(device, index, psec2);
}
#ifndef __NVKM_SEC2_PRIV_H__
#define __NVKM_SEC2_PRIV_H__
#include <engine/sec2.h>
#define nvkm_sec2(p) container_of((p), struct nvkm_sec2, engine)
int nvkm_sec2_new_(struct nvkm_device *, int, struct nvkm_sec2 **);
#endif
nvkm-y += nvkm/falcon/base.o
nvkm-y += nvkm/falcon/v1.o
nvkm-y += nvkm/falcon/msgqueue.o
nvkm-y += nvkm/falcon/msgqueue_0137c63d.o
nvkm-y += nvkm/falcon/msgqueue_0148cdec.o
......@@ -41,14 +41,22 @@ void
nvkm_falcon_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start,
u32 size, u8 port)
{
mutex_lock(&falcon->dmem_mutex);
falcon->func->load_dmem(falcon, data, start, size, port);
mutex_unlock(&falcon->dmem_mutex);
}
void
nvkm_falcon_read_dmem(struct nvkm_falcon *falcon, u32 start, u32 size, u8 port,
void *data)
{
mutex_lock(&falcon->dmem_mutex);
falcon->func->read_dmem(falcon, start, size, port, data);
mutex_unlock(&falcon->dmem_mutex);
}
void
......@@ -129,6 +137,9 @@ nvkm_falcon_clear_interrupt(struct nvkm_falcon *falcon, u32 mask)
void
nvkm_falcon_put(struct nvkm_falcon *falcon, const struct nvkm_subdev *user)
{
if (unlikely(!falcon))
return;
mutex_lock(&falcon->mutex);
if (falcon->user == user) {
nvkm_debug(falcon->user, "released %s falcon\n", falcon->name);
......@@ -159,6 +170,7 @@ nvkm_falcon_ctor(const struct nvkm_falcon_func *func,
struct nvkm_subdev *subdev, const char *name, u32 addr,
struct nvkm_falcon *falcon)
{
u32 debug_reg;
u32 reg;
falcon->func = func;
......@@ -166,6 +178,7 @@ nvkm_falcon_ctor(const struct nvkm_falcon_func *func,
falcon->name = name;
falcon->addr = addr;
mutex_init(&falcon->mutex);
mutex_init(&falcon->dmem_mutex);
reg = nvkm_falcon_rd32(falcon, 0x12c);
falcon->version = reg & 0xf;
......@@ -177,8 +190,31 @@ nvkm_falcon_ctor(const struct nvkm_falcon_func *func,
falcon->code.limit = (reg & 0x1ff) << 8;
falcon->data.limit = (reg & 0x3fe00) >> 1;
reg = nvkm_falcon_rd32(falcon, 0xc08);
falcon->debug = (reg >> 20) & 0x1;
switch (subdev->index) {
case NVKM_ENGINE_GR:
debug_reg = 0x0;
break;
case NVKM_SUBDEV_PMU:
debug_reg = 0xc08;
break;
case NVKM_ENGINE_NVDEC:
debug_reg = 0xd00;
break;
case NVKM_ENGINE_SEC2:
debug_reg = 0x408;
falcon->has_emem = true;
break;
default:
nvkm_warn(subdev, "unsupported falcon %s!\n",
nvkm_subdev_name[subdev->index]);
debug_reg = 0;
break;
}
if (debug_reg) {
u32 val = nvkm_falcon_rd32(falcon, debug_reg);
falcon->debug = (val >> 20) & 0x1;
}
}
void
......
This diff is collapsed.
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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.
*
*/
#ifndef __NVKM_CORE_FALCON_MSGQUEUE_H
#define __NVKM_CORE_FALCON_MSGQUEUE_H
#include <core/msgqueue.h>
/*
* The struct nvkm_msgqueue (named so for lack of better candidate) manages
* a firmware (typically, NVIDIA signed firmware) running under a given falcon.
*
* Such firmwares expect to receive commands (through one or several command
* queues) and will reply to such command by sending messages (using one
* message queue).
*
* Each firmware can support one or several units - ACR for managing secure
* falcons, PMU for power management, etc. A unit can be seen as a class to
* which command can be sent.
*
* One usage example would be to send a command to the SEC falcon to ask it to
* reset a secure falcon. The SEC falcon will receive the command, process it,
* and send a message to signal success or failure. Only when the corresponding
* message is received can the requester assume the request has been processed.
*
* Since we expect many variations between the firmwares NVIDIA will release
* across GPU generations, this library is built in a very modular way. Message
* formats and queues details (such as number of usage) are left to
* specializations of struct nvkm_msgqueue, while the functions in msgqueue.c
* take care of posting commands and processing messages in a fashion that is
* universal.
*
*/
enum msgqueue_msg_priority {
MSGQUEUE_MSG_PRIORITY_HIGH,
MSGQUEUE_MSG_PRIORITY_LOW,
};
/**
* struct nvkm_msgqueue_hdr - header for all commands/messages
* @unit_id: id of firmware using receiving the command/sending the message
* @size: total size of command/message
* @ctrl_flags: type of command/message
* @seq_id: used to match a message from its corresponding command
*/
struct nvkm_msgqueue_hdr {
u8 unit_id;
u8 size;
u8 ctrl_flags;
u8 seq_id;
};
/**
* struct nvkm_msgqueue_msg - base message.
*
* This is just a header and a message (or command) type. Useful when
* building command-specific structures.
*/
struct nvkm_msgqueue_msg {
struct nvkm_msgqueue_hdr hdr;
u8 msg_type;
};
struct nvkm_msgqueue;
typedef void
(*nvkm_msgqueue_callback)(struct nvkm_msgqueue *, struct nvkm_msgqueue_hdr *);
/**
* struct nvkm_msgqueue_init_func - msgqueue functions related to initialization
*
* @gen_cmdline: build the commandline into a pre-allocated buffer
* @init_callback: called to process the init message
*/
struct nvkm_msgqueue_init_func {
void (*gen_cmdline)(struct nvkm_msgqueue *, void *);
int (*init_callback)(struct nvkm_msgqueue *, struct nvkm_msgqueue_hdr *);
};
/**
* struct nvkm_msgqueue_acr_func - msgqueue functions related to ACR
*
* @boot_falcon: build and send the command to reset a given falcon
*/
struct nvkm_msgqueue_acr_func {
int (*boot_falcon)(struct nvkm_msgqueue *, enum nvkm_secboot_falcon);
};
struct nvkm_msgqueue_func {
const struct nvkm_msgqueue_init_func *init_func;
const struct nvkm_msgqueue_acr_func *acr_func;
void (*dtor)(struct nvkm_msgqueue *);
struct nvkm_msgqueue_queue *(*cmd_queue)(struct nvkm_msgqueue *,
enum msgqueue_msg_priority);
void (*recv)(struct nvkm_msgqueue *queue);
};
/**
* struct nvkm_msgqueue_queue - information about a command or message queue
*
* The number of queues is firmware-dependent. All queues must have their
* information filled by the init message handler.
*
* @mutex_lock: to be acquired when the queue is being used
* @index: physical queue index
* @offset: DMEM offset where this queue begins
* @size: size allocated to this queue in DMEM (in bytes)
* @position: current write position
* @head_reg: address of the HEAD register for this queue
* @tail_reg: address of the TAIL register for this queue
*/
struct nvkm_msgqueue_queue {
struct mutex mutex;
u32 index;
u32 offset;
u32 size;
u32 position;
u32 head_reg;
u32 tail_reg;
};
/**
* struct nvkm_msgqueue_seq - keep track of ongoing commands
*
* Every time a command is sent, a sequence is assigned to it so the
* corresponding message can be matched. Upon receiving the message, a callback
* can be called and/or a completion signaled.
*
* @id: sequence ID
* @state: current state
* @callback: callback to call upon receiving matching message
* @completion: completion to signal after callback is called
*/
struct nvkm_msgqueue_seq {
u16 id;
enum {
SEQ_STATE_FREE = 0,
SEQ_STATE_PENDING,
SEQ_STATE_USED,
SEQ_STATE_CANCELLED
} state;
nvkm_msgqueue_callback callback;
struct completion *completion;
};
/*
* We can have an arbitrary number of sequences, but realistically we will
* probably not use that much simultaneously.
*/
#define NVKM_MSGQUEUE_NUM_SEQUENCES 16
/**
* struct nvkm_msgqueue - manage a command/message based FW on a falcon
*
* @falcon: falcon to be managed
* @func: implementation of the firmware to use
* @init_msg_received: whether the init message has already been received
* @init_done: whether all init is complete and commands can be processed
* @seq_lock: protects seq and seq_tbl
* @seq: sequences to match commands and messages
* @seq_tbl: bitmap of sequences currently in use
*/
struct nvkm_msgqueue {
struct nvkm_falcon *falcon;
const struct nvkm_msgqueue_func *func;
u32 fw_version;
bool init_msg_received;
struct completion init_done;
struct mutex seq_lock;
struct nvkm_msgqueue_seq seq[NVKM_MSGQUEUE_NUM_SEQUENCES];
unsigned long seq_tbl[BITS_TO_LONGS(NVKM_MSGQUEUE_NUM_SEQUENCES)];
};
void nvkm_msgqueue_ctor(const struct nvkm_msgqueue_func *, struct nvkm_falcon *,
struct nvkm_msgqueue *);
int nvkm_msgqueue_post(struct nvkm_msgqueue *, enum msgqueue_msg_priority,
struct nvkm_msgqueue_hdr *, nvkm_msgqueue_callback,
struct completion *, bool);
void nvkm_msgqueue_process_msgs(struct nvkm_msgqueue *,
struct nvkm_msgqueue_queue *);
int msgqueue_0137c63d_new(struct nvkm_falcon *, struct nvkm_msgqueue **);
int msgqueue_0148cdec_new(struct nvkm_falcon *, struct nvkm_msgqueue **);
#endif
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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.
*
*/
#include "msgqueue.h"
#include <engine/falcon.h>
#include <subdev/secboot.h>
/* Queues identifiers */
enum {
/* High Priority Command Queue for Host -> PMU communication */
MSGQUEUE_0137C63D_COMMAND_QUEUE_HPQ = 0,
/* Low Priority Command Queue for Host -> PMU communication */
MSGQUEUE_0137C63D_COMMAND_QUEUE_LPQ = 1,
/* Message queue for PMU -> Host communication */
MSGQUEUE_0137C63D_MESSAGE_QUEUE = 4,
MSGQUEUE_0137C63D_NUM_QUEUES = 5,
};
struct msgqueue_0137c63d {
struct nvkm_msgqueue base;
struct nvkm_msgqueue_queue queue[MSGQUEUE_0137C63D_NUM_QUEUES];
};
#define msgqueue_0137c63d(q) \
container_of(q, struct msgqueue_0137c63d, base)
static struct nvkm_msgqueue_queue *
msgqueue_0137c63d_cmd_queue(struct nvkm_msgqueue *queue,
enum msgqueue_msg_priority priority)
{
struct msgqueue_0137c63d *priv = msgqueue_0137c63d(queue);
const struct nvkm_subdev *subdev = priv->base.falcon->owner;
switch (priority) {
case MSGQUEUE_MSG_PRIORITY_HIGH:
return &priv->queue[MSGQUEUE_0137C63D_COMMAND_QUEUE_HPQ];
case MSGQUEUE_MSG_PRIORITY_LOW:
return &priv->queue[MSGQUEUE_0137C63D_COMMAND_QUEUE_LPQ];
default:
nvkm_error(subdev, "invalid command queue!\n");
return ERR_PTR(-EINVAL);
}
}
static void
msgqueue_0137c63d_process_msgs(struct nvkm_msgqueue *queue)
{
struct msgqueue_0137c63d *priv = msgqueue_0137c63d(queue);
struct nvkm_msgqueue_queue *q_queue =
&priv->queue[MSGQUEUE_0137C63D_MESSAGE_QUEUE];
nvkm_msgqueue_process_msgs(&priv->base, q_queue);
}
/* Init unit */
#define MSGQUEUE_0137C63D_UNIT_INIT 0x07
enum {
INIT_MSG_INIT = 0x0,
};
static void
init_gen_cmdline(struct nvkm_msgqueue *queue, void *buf)
{
struct {
u32 reserved;
u32 freq_hz;
u32 trace_size;
u32 trace_dma_base;
u16 trace_dma_base1;
u8 trace_dma_offset;
u32 trace_dma_idx;
bool secure_mode;
bool raise_priv_sec;
struct {
u32 dma_base;
u16 dma_base1;
u8 dma_offset;
u16 fb_size;
u8 dma_idx;
} gc6_ctx;
u8 pad;
} *args = buf;
args->secure_mode = 1;
}
/* forward declaration */
static int acr_init_wpr(struct nvkm_msgqueue *queue);
static int
init_callback(struct nvkm_msgqueue *_queue, struct nvkm_msgqueue_hdr *hdr)
{
struct msgqueue_0137c63d *priv = msgqueue_0137c63d(_queue);
struct {
struct nvkm_msgqueue_msg base;
u8 pad;
u16 os_debug_entry_point;
struct {
u16 size;
u16 offset;
u8 index;
u8 pad;
} queue_info[MSGQUEUE_0137C63D_NUM_QUEUES];
u16 sw_managed_area_offset;
u16 sw_managed_area_size;
} *init = (void *)hdr;
const struct nvkm_subdev *subdev = _queue->falcon->owner;
int i;
if (init->base.hdr.unit_id != MSGQUEUE_0137C63D_UNIT_INIT) {
nvkm_error(subdev, "expected message from init unit\n");
return -EINVAL;
}
if (init->base.msg_type != INIT_MSG_INIT) {
nvkm_error(subdev, "expected PMU init msg\n");
return -EINVAL;
}
for (i = 0; i < MSGQUEUE_0137C63D_NUM_QUEUES; i++) {
struct nvkm_msgqueue_queue *queue = &priv->queue[i];
mutex_init(&queue->mutex);
queue->index = init->queue_info[i].index;
queue->offset = init->queue_info[i].offset;
queue->size = init->queue_info[i].size;
if (i != MSGQUEUE_0137C63D_MESSAGE_QUEUE) {
queue->head_reg = 0x4a0 + (queue->index * 4);
queue->tail_reg = 0x4b0 + (queue->index * 4);
} else {
queue->head_reg = 0x4c8;
queue->tail_reg = 0x4cc;
}
nvkm_debug(subdev,
"queue %d: index %d, offset 0x%08x, size 0x%08x\n",
i, queue->index, queue->offset, queue->size);
}
/* Complete initialization by initializing WPR region */
return acr_init_wpr(&priv->base);
}
static const struct nvkm_msgqueue_init_func
msgqueue_0137c63d_init_func = {
.gen_cmdline = init_gen_cmdline,
.init_callback = init_callback,
};
/* ACR unit */
#define MSGQUEUE_0137C63D_UNIT_ACR 0x0a
enum {
ACR_CMD_INIT_WPR_REGION = 0x00,
ACR_CMD_BOOTSTRAP_FALCON = 0x01,
};
static void
acr_init_wpr_callback(struct nvkm_msgqueue *queue,
struct nvkm_msgqueue_hdr *hdr)
{
struct {
struct nvkm_msgqueue_msg base;
u32 error_code;
} *msg = (void *)hdr;
const struct nvkm_subdev *subdev = queue->falcon->owner;
if (msg->error_code) {
nvkm_error(subdev, "ACR WPR init failure: %d\n",
msg->error_code);
return;
}
nvkm_debug(subdev, "ACR WPR init complete\n");
complete_all(&queue->init_done);
}
static int
acr_init_wpr(struct nvkm_msgqueue *queue)
{
/*
* region_id: region ID in WPR region
* wpr_offset: offset in WPR region
*/
struct {
struct nvkm_msgqueue_hdr hdr;
u8 cmd_type;
u32 region_id;
u32 wpr_offset;
} cmd;
memset(&cmd, 0, sizeof(cmd));
cmd.hdr.unit_id = MSGQUEUE_0137C63D_UNIT_ACR;
cmd.hdr.size = sizeof(cmd);
cmd.cmd_type = ACR_CMD_INIT_WPR_REGION;
cmd.region_id = 0x01;
cmd.wpr_offset = 0x00;
nvkm_msgqueue_post(queue, MSGQUEUE_MSG_PRIORITY_HIGH, &cmd.hdr,
acr_init_wpr_callback, NULL, false);
return 0;
}
static void
acr_boot_falcon_callback(struct nvkm_msgqueue *priv,
struct nvkm_msgqueue_hdr *hdr)
{
struct acr_bootstrap_falcon_msg {
struct nvkm_msgqueue_msg base;
u32 falcon_id;
} *msg = (void *)hdr;
const struct nvkm_subdev *subdev = priv->falcon->owner;
u32 falcon_id = msg->falcon_id;
if (falcon_id >= NVKM_SECBOOT_FALCON_END) {
nvkm_error(subdev, "in bootstrap falcon callback:\n");
nvkm_error(subdev, "invalid falcon ID 0x%x\n", falcon_id);
return;
}
nvkm_debug(subdev, "%s booted\n", nvkm_secboot_falcon_name[falcon_id]);
}
enum {
ACR_CMD_BOOTSTRAP_FALCON_FLAGS_RESET_YES = 0,
ACR_CMD_BOOTSTRAP_FALCON_FLAGS_RESET_NO = 1,
};
static int
acr_boot_falcon(struct nvkm_msgqueue *priv, enum nvkm_secboot_falcon falcon)
{
DECLARE_COMPLETION_ONSTACK(completed);
/*
* flags - Flag specifying RESET or no RESET.
* falcon id - Falcon id specifying falcon to bootstrap.
*/
struct {
struct nvkm_msgqueue_hdr hdr;
u8 cmd_type;
u32 flags;
u32 falcon_id;
} cmd;
memset(&cmd, 0, sizeof(cmd));
cmd.hdr.unit_id = MSGQUEUE_0137C63D_UNIT_ACR;
cmd.hdr.size = sizeof(cmd);
cmd.cmd_type = ACR_CMD_BOOTSTRAP_FALCON;
cmd.flags = ACR_CMD_BOOTSTRAP_FALCON_FLAGS_RESET_YES;
cmd.falcon_id = falcon;
nvkm_msgqueue_post(priv, MSGQUEUE_MSG_PRIORITY_HIGH, &cmd.hdr,
acr_boot_falcon_callback, &completed, true);
if (!wait_for_completion_timeout(&completed, msecs_to_jiffies(1000)))
return -ETIMEDOUT;
return 0;
}
static const struct nvkm_msgqueue_acr_func
msgqueue_0137c63d_acr_func = {
.boot_falcon = acr_boot_falcon,
};
static void
msgqueue_0137c63d_dtor(struct nvkm_msgqueue *queue)
{
kfree(msgqueue_0137c63d(queue));
}
static const struct nvkm_msgqueue_func
msgqueue_0137c63d_func = {
.init_func = &msgqueue_0137c63d_init_func,
.acr_func = &msgqueue_0137c63d_acr_func,
.cmd_queue = msgqueue_0137c63d_cmd_queue,
.recv = msgqueue_0137c63d_process_msgs,
.dtor = msgqueue_0137c63d_dtor,
};
int
msgqueue_0137c63d_new(struct nvkm_falcon *falcon, struct nvkm_msgqueue **queue)
{
struct msgqueue_0137c63d *ret;
ret = kzalloc(sizeof(*ret), GFP_KERNEL);
if (!ret)
return -ENOMEM;
*queue = &ret->base;
nvkm_msgqueue_ctor(&msgqueue_0137c63d_func, falcon, &ret->base);
return 0;
}
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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.
*
*/
#include "msgqueue.h"
#include <engine/falcon.h>
#include <subdev/secboot.h>
/*
* This firmware runs on the SEC falcon. It only has one command and one
* message queue, and uses a different command line and init message.
*/
enum {
MSGQUEUE_0148CDEC_COMMAND_QUEUE = 0,
MSGQUEUE_0148CDEC_MESSAGE_QUEUE = 1,
MSGQUEUE_0148CDEC_NUM_QUEUES,
};
struct msgqueue_0148cdec {
struct nvkm_msgqueue base;
struct nvkm_msgqueue_queue queue[MSGQUEUE_0148CDEC_NUM_QUEUES];
};
#define msgqueue_0148cdec(q) \
container_of(q, struct msgqueue_0148cdec, base)
static struct nvkm_msgqueue_queue *
msgqueue_0148cdec_cmd_queue(struct nvkm_msgqueue *queue,
enum msgqueue_msg_priority priority)
{
struct msgqueue_0148cdec *priv = msgqueue_0148cdec(queue);
return &priv->queue[MSGQUEUE_0148CDEC_COMMAND_QUEUE];
}
static void
msgqueue_0148cdec_process_msgs(struct nvkm_msgqueue *queue)
{
struct msgqueue_0148cdec *priv = msgqueue_0148cdec(queue);
struct nvkm_msgqueue_queue *q_queue =
&priv->queue[MSGQUEUE_0148CDEC_MESSAGE_QUEUE];
nvkm_msgqueue_process_msgs(&priv->base, q_queue);
}
/* Init unit */
#define MSGQUEUE_0148CDEC_UNIT_INIT 0x01
enum {
INIT_MSG_INIT = 0x0,
};
static void
init_gen_cmdline(struct nvkm_msgqueue *queue, void *buf)
{
struct {
u32 freq_hz;
u32 falc_trace_size;
u32 falc_trace_dma_base;
u32 falc_trace_dma_idx;
bool secure_mode;
} *args = buf;
args->secure_mode = false;
}
static int
init_callback(struct nvkm_msgqueue *_queue, struct nvkm_msgqueue_hdr *hdr)
{
struct msgqueue_0148cdec *priv = msgqueue_0148cdec(_queue);
struct {
struct nvkm_msgqueue_msg base;
u8 num_queues;
u16 os_debug_entry_point;
struct {
u32 offset;
u16 size;
u8 index;
u8 id;
} queue_info[MSGQUEUE_0148CDEC_NUM_QUEUES];
u16 sw_managed_area_offset;
u16 sw_managed_area_size;
} *init = (void *)hdr;
const struct nvkm_subdev *subdev = _queue->falcon->owner;
int i;
if (init->base.hdr.unit_id != MSGQUEUE_0148CDEC_UNIT_INIT) {
nvkm_error(subdev, "expected message from init unit\n");
return -EINVAL;
}
if (init->base.msg_type != INIT_MSG_INIT) {
nvkm_error(subdev, "expected SEC init msg\n");
return -EINVAL;
}
for (i = 0; i < MSGQUEUE_0148CDEC_NUM_QUEUES; i++) {
u8 id = init->queue_info[i].id;
struct nvkm_msgqueue_queue *queue = &priv->queue[id];
mutex_init(&queue->mutex);
queue->index = init->queue_info[i].index;
queue->offset = init->queue_info[i].offset;
queue->size = init->queue_info[i].size;
if (id == MSGQUEUE_0148CDEC_MESSAGE_QUEUE) {
queue->head_reg = 0xa30 + (queue->index * 8);
queue->tail_reg = 0xa34 + (queue->index * 8);
} else {
queue->head_reg = 0xa00 + (queue->index * 8);
queue->tail_reg = 0xa04 + (queue->index * 8);
}
nvkm_debug(subdev,
"queue %d: index %d, offset 0x%08x, size 0x%08x\n",
id, queue->index, queue->offset, queue->size);
}
complete_all(&_queue->init_done);
return 0;
}
static const struct nvkm_msgqueue_init_func
msgqueue_0148cdec_init_func = {
.gen_cmdline = init_gen_cmdline,
.init_callback = init_callback,
};
/* ACR unit */
#define MSGQUEUE_0148CDEC_UNIT_ACR 0x08
enum {
ACR_CMD_BOOTSTRAP_FALCON = 0x00,
};
static void
acr_boot_falcon_callback(struct nvkm_msgqueue *priv,
struct nvkm_msgqueue_hdr *hdr)
{
struct acr_bootstrap_falcon_msg {
struct nvkm_msgqueue_msg base;
u32 error_code;
u32 falcon_id;
} *msg = (void *)hdr;
const struct nvkm_subdev *subdev = priv->falcon->owner;
u32 falcon_id = msg->falcon_id;
if (msg->error_code) {
nvkm_error(subdev, "in bootstrap falcon callback:\n");
nvkm_error(subdev, "expected error code 0x%x\n",
msg->error_code);
return;
}
if (falcon_id >= NVKM_SECBOOT_FALCON_END) {
nvkm_error(subdev, "in bootstrap falcon callback:\n");
nvkm_error(subdev, "invalid falcon ID 0x%x\n", falcon_id);
return;
}
nvkm_debug(subdev, "%s booted\n", nvkm_secboot_falcon_name[falcon_id]);
}
enum {
ACR_CMD_BOOTSTRAP_FALCON_FLAGS_RESET_YES = 0,
ACR_CMD_BOOTSTRAP_FALCON_FLAGS_RESET_NO = 1,
};
static int
acr_boot_falcon(struct nvkm_msgqueue *priv, enum nvkm_secboot_falcon falcon)
{
DECLARE_COMPLETION_ONSTACK(completed);
/*
* flags - Flag specifying RESET or no RESET.
* falcon id - Falcon id specifying falcon to bootstrap.
*/
struct {
struct nvkm_msgqueue_hdr hdr;
u8 cmd_type;
u32 flags;
u32 falcon_id;
} cmd;
memset(&cmd, 0, sizeof(cmd));
cmd.hdr.unit_id = MSGQUEUE_0148CDEC_UNIT_ACR;
cmd.hdr.size = sizeof(cmd);
cmd.cmd_type = ACR_CMD_BOOTSTRAP_FALCON;
cmd.flags = ACR_CMD_BOOTSTRAP_FALCON_FLAGS_RESET_YES;
cmd.falcon_id = falcon;
nvkm_msgqueue_post(priv, MSGQUEUE_MSG_PRIORITY_HIGH, &cmd.hdr,
acr_boot_falcon_callback, &completed, true);
if (!wait_for_completion_timeout(&completed, msecs_to_jiffies(1000)))
return -ETIMEDOUT;
return 0;
}
const struct nvkm_msgqueue_acr_func
msgqueue_0148cdec_acr_func = {
.boot_falcon = acr_boot_falcon,
};
static void
msgqueue_0148cdec_dtor(struct nvkm_msgqueue *queue)
{
kfree(msgqueue_0148cdec(queue));
}
const struct nvkm_msgqueue_func
msgqueue_0148cdec_func = {
.init_func = &msgqueue_0148cdec_init_func,
.acr_func = &msgqueue_0148cdec_acr_func,
.cmd_queue = msgqueue_0148cdec_cmd_queue,
.recv = msgqueue_0148cdec_process_msgs,
.dtor = msgqueue_0148cdec_dtor,
};
int
msgqueue_0148cdec_new(struct nvkm_falcon *falcon, struct nvkm_msgqueue **queue)
{
struct msgqueue_0148cdec *ret;
ret = kzalloc(sizeof(*ret), GFP_KERNEL);
if (!ret)
return -ENOMEM;
*queue = &ret->base;
nvkm_msgqueue_ctor(&msgqueue_0148cdec_func, falcon, &ret->base);
return 0;
}
......@@ -40,8 +40,8 @@ nvkm_falcon_v1_load_imem(struct nvkm_falcon *falcon, void *data, u32 start,
for (i = 0; i < size / 4; i++) {
/* write new tag every 256B */
if ((i & 0x3f) == 0)
nvkm_falcon_wr32(falcon, 0x188, tag++);
nvkm_falcon_wr32(falcon, 0x184, ((u32 *)data)[i]);
nvkm_falcon_wr32(falcon, 0x188 + (port * 16), tag++);
nvkm_falcon_wr32(falcon, 0x184 + (port * 16), ((u32 *)data)[i]);
}
/*
......@@ -53,16 +53,44 @@ nvkm_falcon_v1_load_imem(struct nvkm_falcon *falcon, void *data, u32 start,
/* write new tag every 256B */
if ((i & 0x3f) == 0)
nvkm_falcon_wr32(falcon, 0x188, tag++);
nvkm_falcon_wr32(falcon, 0x184, extra & (BIT(rem * 8) - 1));
nvkm_falcon_wr32(falcon, 0x188 + (port * 16), tag++);
nvkm_falcon_wr32(falcon, 0x184 + (port * 16),
extra & (BIT(rem * 8) - 1));
++i;
}
/* code must be padded to 0x40 words */
for (; i & 0x3f; i++)
nvkm_falcon_wr32(falcon, 0x184, 0);
nvkm_falcon_wr32(falcon, 0x184 + (port * 16), 0);
}
static void
nvkm_falcon_v1_load_emem(struct nvkm_falcon *falcon, void *data, u32 start,
u32 size, u8 port)
{
u8 rem = size % 4;
int i;
size -= rem;
nvkm_falcon_wr32(falcon, 0xac0 + (port * 8), start | (0x1 << 24));
for (i = 0; i < size / 4; i++)
nvkm_falcon_wr32(falcon, 0xac4 + (port * 8), ((u32 *)data)[i]);
/*
* If size is not a multiple of 4, mask the last word to ensure garbage
* does not get written
*/
if (rem) {
u32 extra = ((u32 *)data)[i];
nvkm_falcon_wr32(falcon, 0xac4 + (port * 8),
extra & (BIT(rem * 8) - 1));
}
}
static const u32 EMEM_START_ADDR = 0x1000000;
static void
nvkm_falcon_v1_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start,
u32 size, u8 port)
......@@ -70,20 +98,53 @@ nvkm_falcon_v1_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start,
u8 rem = size % 4;
int i;
if (start >= EMEM_START_ADDR && falcon->has_emem)
return nvkm_falcon_v1_load_emem(falcon, data,
start - EMEM_START_ADDR, size,
port);
size -= rem;
nvkm_falcon_wr32(falcon, 0x1c0 + (port * 16), start | (0x1 << 24));
nvkm_falcon_wr32(falcon, 0x1c0 + (port * 8), start | (0x1 << 24));
for (i = 0; i < size / 4; i++)
nvkm_falcon_wr32(falcon, 0x1c4, ((u32 *)data)[i]);
nvkm_falcon_wr32(falcon, 0x1c4 + (port * 8), ((u32 *)data)[i]);
/*
* If size is not a multiple of 4, mask the last work to ensure garbage
* does not get read
* If size is not a multiple of 4, mask the last word to ensure garbage
* does not get written
*/
if (rem) {
u32 extra = ((u32 *)data)[i];
nvkm_falcon_wr32(falcon, 0x1c4, extra & (BIT(rem * 8) - 1));
nvkm_falcon_wr32(falcon, 0x1c4 + (port * 8),
extra & (BIT(rem * 8) - 1));
}
}
static void
nvkm_falcon_v1_read_emem(struct nvkm_falcon *falcon, u32 start, u32 size,
u8 port, void *data)
{
u8 rem = size % 4;
int i;
size -= rem;
nvkm_falcon_wr32(falcon, 0xac0 + (port * 8), start | (0x1 << 25));
for (i = 0; i < size / 4; i++)
((u32 *)data)[i] = nvkm_falcon_rd32(falcon, 0xac4 + (port * 8));
/*
* If size is not a multiple of 4, mask the last word to ensure garbage
* does not get read
*/
if (rem) {
u32 extra = nvkm_falcon_rd32(falcon, 0xac4 + (port * 8));
for (i = size; i < size + rem; i++) {
((u8 *)data)[i] = (u8)(extra & 0xff);
extra >>= 8;
}
}
}
......@@ -94,18 +155,22 @@ nvkm_falcon_v1_read_dmem(struct nvkm_falcon *falcon, u32 start, u32 size,
u8 rem = size % 4;
int i;
if (start >= EMEM_START_ADDR && falcon->has_emem)
return nvkm_falcon_v1_read_emem(falcon, start - EMEM_START_ADDR,
size, port, data);
size -= rem;
nvkm_falcon_wr32(falcon, 0x1c0 + (port * 16), start | (0x1 << 25));
nvkm_falcon_wr32(falcon, 0x1c0 + (port * 8), start | (0x1 << 25));
for (i = 0; i < size / 4; i++)
((u32 *)data)[i] = nvkm_falcon_rd32(falcon, 0x1c4);
((u32 *)data)[i] = nvkm_falcon_rd32(falcon, 0x1c4 + (port * 8));
/*
* If size is not a multiple of 4, mask the last work to ensure garbage
* If size is not a multiple of 4, mask the last word to ensure garbage
* does not get read
*/
if (rem) {
u32 extra = nvkm_falcon_rd32(falcon, 0x1c4);
u32 extra = nvkm_falcon_rd32(falcon, 0x1c4 + (port * 8));
for (i = size; i < size + rem; i++) {
((u8 *)data)[i] = (u8)(extra & 0xff);
......@@ -118,6 +183,7 @@ static void
nvkm_falcon_v1_bind_context(struct nvkm_falcon *falcon, struct nvkm_gpuobj *ctx)
{
u32 inst_loc;
u32 fbif;
/* disable instance block binding */
if (ctx == NULL) {
......@@ -125,19 +191,34 @@ nvkm_falcon_v1_bind_context(struct nvkm_falcon *falcon, struct nvkm_gpuobj *ctx)
return;
}
switch (falcon->owner->index) {
case NVKM_ENGINE_NVENC0:
case NVKM_ENGINE_NVENC1:
case NVKM_ENGINE_NVENC2:
fbif = 0x800;
break;
case NVKM_SUBDEV_PMU:
fbif = 0xe00;
break;
default:
fbif = 0x600;
break;
}
nvkm_falcon_wr32(falcon, 0x10c, 0x1);
/* setup apertures - virtual */
nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_UCODE, 0x4);
nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_VIRT, 0x0);
nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_UCODE, 0x4);
nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_VIRT, 0x0);
/* setup apertures - physical */
nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_PHYS_VID, 0x4);
nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_PHYS_SYS_COH, 0x5);
nvkm_falcon_wr32(falcon, 0xe00 + 4 * FALCON_DMAIDX_PHYS_SYS_NCOH, 0x6);
nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_PHYS_VID, 0x4);
nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_PHYS_SYS_COH, 0x5);
nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_PHYS_SYS_NCOH, 0x6);
/* Set context */
switch (nvkm_memory_target(ctx->memory)) {
case NVKM_MEM_TARGET_VRAM: inst_loc = 0; break;
case NVKM_MEM_TARGET_HOST: inst_loc = 2; break;
case NVKM_MEM_TARGET_NCOH: inst_loc = 3; break;
default:
WARN_ON(1);
......@@ -146,9 +227,12 @@ nvkm_falcon_v1_bind_context(struct nvkm_falcon *falcon, struct nvkm_gpuobj *ctx)
/* Enable context */
nvkm_falcon_mask(falcon, 0x048, 0x1, 0x1);
nvkm_falcon_wr32(falcon, 0x480,
nvkm_falcon_wr32(falcon, 0x054,
((ctx->addr >> 12) & 0xfffffff) |
(inst_loc << 28) | (1 << 30));
nvkm_falcon_mask(falcon, 0x090, 0x10000, 0x10000);
nvkm_falcon_mask(falcon, 0x0a4, 0x8, 0x8);
}
static void
......
......@@ -20,6 +20,7 @@ nvkm-y += nvkm/subdev/fb/gt215.o
nvkm-y += nvkm/subdev/fb/mcp77.o
nvkm-y += nvkm/subdev/fb/mcp89.o
nvkm-y += nvkm/subdev/fb/gf100.o
nvkm-y += nvkm/subdev/fb/gf108.o
nvkm-y += nvkm/subdev/fb/gk104.o
nvkm-y += nvkm/subdev/fb/gk20a.o
nvkm-y += nvkm/subdev/fb/gm107.o
......@@ -42,8 +43,10 @@ nvkm-y += nvkm/subdev/fb/ramnv50.o
nvkm-y += nvkm/subdev/fb/ramgt215.o
nvkm-y += nvkm/subdev/fb/rammcp77.o
nvkm-y += nvkm/subdev/fb/ramgf100.o
nvkm-y += nvkm/subdev/fb/ramgf108.o
nvkm-y += nvkm/subdev/fb/ramgk104.o
nvkm-y += nvkm/subdev/fb/ramgm107.o
nvkm-y += nvkm/subdev/fb/ramgm200.o
nvkm-y += nvkm/subdev/fb/ramgp100.o
nvkm-y += nvkm/subdev/fb/sddr2.o
nvkm-y += nvkm/subdev/fb/sddr3.o
......
/*
* Copyright 2017 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 <bskeggs@redhat.com>
*/
#include "gf100.h"
#include "ram.h"
static const struct nvkm_fb_func
gf108_fb = {
.dtor = gf100_fb_dtor,
.oneinit = gf100_fb_oneinit,
.init = gf100_fb_init,
.init_page = gf100_fb_init_page,
.intr = gf100_fb_intr,
.ram_new = gf108_ram_new,
.memtype_valid = gf100_fb_memtype_valid,
};
int
gf108_fb_new(struct nvkm_device *device, int index, struct nvkm_fb **pfb)
{
return gf100_fb_new_(&gf108_fb, device, index, pfb);
}
......@@ -68,7 +68,7 @@ gm200_fb = {
.init = gm200_fb_init,
.init_page = gm200_fb_init_page,
.intr = gf100_fb_intr,
.ram_new = gm107_ram_new,
.ram_new = gm200_ram_new,
.memtype_valid = gf100_fb_memtype_valid,
};
......
......@@ -19,13 +19,38 @@ int nv50_ram_get(struct nvkm_ram *, u64, u32, u32, u32, struct nvkm_mem **);
void nv50_ram_put(struct nvkm_ram *, struct nvkm_mem **);
void __nv50_ram_put(struct nvkm_ram *, struct nvkm_mem *);
int gf100_ram_new_(const struct nvkm_ram_func *, struct nvkm_fb *,
struct nvkm_ram **);
int gf100_ram_ctor(const struct nvkm_ram_func *, struct nvkm_fb *,
u32, struct nvkm_ram *);
struct nvkm_ram *);
u32 gf100_ram_probe_fbp(const struct nvkm_ram_func *,
struct nvkm_device *, int, int *);
u32 gf100_ram_probe_fbp_amount(const struct nvkm_ram_func *, u32,
struct nvkm_device *, int, int *);
u32 gf100_ram_probe_fbpa_amount(struct nvkm_device *, int);
int gf100_ram_get(struct nvkm_ram *, u64, u32, u32, u32, struct nvkm_mem **);
void gf100_ram_put(struct nvkm_ram *, struct nvkm_mem **);
int gf100_ram_init(struct nvkm_ram *);
int gf100_ram_calc(struct nvkm_ram *, u32);
int gf100_ram_prog(struct nvkm_ram *);
void gf100_ram_tidy(struct nvkm_ram *);
u32 gf108_ram_probe_fbp_amount(const struct nvkm_ram_func *, u32,
struct nvkm_device *, int, int *);
int gk104_ram_new_(const struct nvkm_ram_func *, struct nvkm_fb *,
struct nvkm_ram **);
void *gk104_ram_dtor(struct nvkm_ram *);
int gk104_ram_init(struct nvkm_ram *);
int gk104_ram_calc(struct nvkm_ram *, u32);
int gk104_ram_prog(struct nvkm_ram *);
void gk104_ram_tidy(struct nvkm_ram *);
u32 gm107_ram_probe_fbp(const struct nvkm_ram_func *,
struct nvkm_device *, int, int *);
int gk104_ram_ctor(struct nvkm_fb *, struct nvkm_ram **, u32);
int gk104_ram_init(struct nvkm_ram *ram);
u32 gm200_ram_probe_fbp_amount(const struct nvkm_ram_func *, u32,
struct nvkm_device *, int, int *);
/* RAM type-specific MR calculation routines */
int nvkm_sddr2_calc(struct nvkm_ram *);
......@@ -46,7 +71,9 @@ int nv50_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int gt215_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int mcp77_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int gf100_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int gf108_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int gk104_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int gm107_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int gm200_ram_new(struct nvkm_fb *, struct nvkm_ram **);
int gp100_ram_new(struct nvkm_fb *, struct nvkm_ram **);
#endif
......@@ -124,7 +124,7 @@ gf100_ram_train(struct gf100_ramfuc *fuc, u32 magic)
}
}
static int
int
gf100_ram_calc(struct nvkm_ram *base, u32 freq)
{
struct gf100_ram *ram = gf100_ram(base);
......@@ -404,7 +404,7 @@ gf100_ram_calc(struct nvkm_ram *base, u32 freq)
return 0;
}
static int
int
gf100_ram_prog(struct nvkm_ram *base)
{
struct gf100_ram *ram = gf100_ram(base);
......@@ -413,7 +413,7 @@ gf100_ram_prog(struct nvkm_ram *base)
return 0;
}
static void
void
gf100_ram_tidy(struct nvkm_ram *base)
{
struct gf100_ram *ram = gf100_ram(base);
......@@ -500,7 +500,7 @@ gf100_ram_get(struct nvkm_ram *ram, u64 size, u32 align, u32 ncmin,
return 0;
}
static int
int
gf100_ram_init(struct nvkm_ram *base)
{
static const u8 train0[] = {
......@@ -543,77 +543,96 @@ gf100_ram_init(struct nvkm_ram *base)
return 0;
}
static const struct nvkm_ram_func
gf100_ram_func = {
.init = gf100_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
.calc = gf100_ram_calc,
.prog = gf100_ram_prog,
.tidy = gf100_ram_tidy,
};
u32
gf100_ram_probe_fbpa_amount(struct nvkm_device *device, int fbpa)
{
return nvkm_rd32(device, 0x11020c + (fbpa * 0x1000));
}
u32
gf100_ram_probe_fbp_amount(const struct nvkm_ram_func *func, u32 fbpao,
struct nvkm_device *device, int fbp, int *pltcs)
{
if (!(fbpao & BIT(fbp))) {
*pltcs = 1;
return func->probe_fbpa_amount(device, fbp);
}
return 0;
}
u32
gf100_ram_probe_fbp(const struct nvkm_ram_func *func,
struct nvkm_device *device, int fbp, int *pltcs)
{
u32 fbpao = nvkm_rd32(device, 0x022554);
return func->probe_fbp_amount(func, fbpao, device, fbp, pltcs);
}
int
gf100_ram_ctor(const struct nvkm_ram_func *func, struct nvkm_fb *fb,
u32 maskaddr, struct nvkm_ram *ram)
struct nvkm_ram *ram)
{
struct nvkm_subdev *subdev = &fb->subdev;
struct nvkm_device *device = subdev->device;
struct nvkm_bios *bios = device->bios;
const u32 rsvd_head = ( 256 * 1024); /* vga memory */
const u32 rsvd_tail = (1024 * 1024); /* vbios etc */
u32 parts = nvkm_rd32(device, 0x022438);
u32 pmask = nvkm_rd32(device, maskaddr);
u64 bsize = (u64)nvkm_rd32(device, 0x10f20c) << 20;
u64 psize, size = 0;
enum nvkm_ram_type type = nvkm_fb_bios_memtype(bios);
bool uniform = true;
int ret, i;
nvkm_debug(subdev, "100800: %08x\n", nvkm_rd32(device, 0x100800));
nvkm_debug(subdev, "parts %08x mask %08x\n", parts, pmask);
/* read amount of vram attached to each memory controller */
for (i = 0; i < parts; i++) {
if (pmask & (1 << i))
continue;
psize = (u64)nvkm_rd32(device, 0x11020c + (i * 0x1000)) << 20;
if (psize != bsize) {
if (psize < bsize)
bsize = psize;
uniform = false;
u32 fbps = nvkm_rd32(device, 0x022438);
u64 total = 0, lcomm = ~0, lower, ubase, usize;
int ret, fbp, ltcs, ltcn = 0;
nvkm_debug(subdev, "%d FBP(s)\n", fbps);
for (fbp = 0; fbp < fbps; fbp++) {
u32 size = func->probe_fbp(func, device, fbp, &ltcs);
if (size) {
nvkm_debug(subdev, "FBP %d: %4d MiB, %d LTC(s)\n",
fbp, size, ltcs);
lcomm = min(lcomm, (u64)(size / ltcs) << 20);
total += size << 20;
ltcn += ltcs;
} else {
nvkm_debug(subdev, "FBP %d: disabled\n", fbp);
}
nvkm_debug(subdev, "%d: %d MiB\n", i, (u32)(psize >> 20));
size += psize;
}
ret = nvkm_ram_ctor(func, fb, type, size, 0, ram);
lower = lcomm * ltcn;
ubase = lcomm + func->upper;
usize = total - lower;
nvkm_debug(subdev, "Lower: %4lld MiB @ %010llx\n", lower >> 20, 0ULL);
nvkm_debug(subdev, "Upper: %4lld MiB @ %010llx\n", usize >> 20, ubase);
nvkm_debug(subdev, "Total: %4lld MiB\n", total >> 20);
ret = nvkm_ram_ctor(func, fb, type, total, 0, ram);
if (ret)
return ret;
nvkm_mm_fini(&ram->vram);
/* if all controllers have the same amount attached, there's no holes */
if (uniform) {
/* Some GPUs are in what's known as a "mixed memory" configuration.
*
* This is either where some FBPs have more memory than the others,
* or where LTCs have been disabled on a FBP.
*/
if (lower != total) {
/* The common memory amount is addressed normally. */
ret = nvkm_mm_init(&ram->vram, rsvd_head >> NVKM_RAM_MM_SHIFT,
(size - rsvd_head - rsvd_tail) >>
NVKM_RAM_MM_SHIFT, 1);
(lower - rsvd_head) >> NVKM_RAM_MM_SHIFT, 1);
if (ret)
return ret;
} else {
/* otherwise, address lowest common amount from 0GiB */
ret = nvkm_mm_init(&ram->vram, rsvd_head >> NVKM_RAM_MM_SHIFT,
((bsize * parts) - rsvd_head) >>
NVKM_RAM_MM_SHIFT, 1);
/* And the rest is much higher in the physical address
* space, and may not be usable for certain operations.
*/
ret = nvkm_mm_init(&ram->vram, ubase >> NVKM_RAM_MM_SHIFT,
(usize - rsvd_tail) >> NVKM_RAM_MM_SHIFT, 1);
if (ret)
return ret;
/* and the rest starting from (8GiB + common_size) */
ret = nvkm_mm_init(&ram->vram, (0x0200000000ULL + bsize) >>
NVKM_RAM_MM_SHIFT,
(size - (bsize * parts) - rsvd_tail) >>
} else {
/* GPUs without mixed-memory are a lot nicer... */
ret = nvkm_mm_init(&ram->vram, rsvd_head >> NVKM_RAM_MM_SHIFT,
(total - rsvd_head - rsvd_tail) >>
NVKM_RAM_MM_SHIFT, 1);
if (ret)
return ret;
......@@ -624,7 +643,8 @@ gf100_ram_ctor(const struct nvkm_ram_func *func, struct nvkm_fb *fb,
}
int
gf100_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
gf100_ram_new_(const struct nvkm_ram_func *func,
struct nvkm_fb *fb, struct nvkm_ram **pram)
{
struct nvkm_subdev *subdev = &fb->subdev;
struct nvkm_bios *bios = subdev->device->bios;
......@@ -635,7 +655,7 @@ gf100_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
return -ENOMEM;
*pram = &ram->base;
ret = gf100_ram_ctor(&gf100_ram_func, fb, 0x022554, &ram->base);
ret = gf100_ram_ctor(func, fb, &ram->base);
if (ret)
return ret;
......@@ -711,3 +731,23 @@ gf100_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
ram->fuc.r_0x13d8f4 = ramfuc_reg(0x13d8f4);
return 0;
}
static const struct nvkm_ram_func
gf100_ram = {
.upper = 0x0200000000,
.probe_fbp = gf100_ram_probe_fbp,
.probe_fbp_amount = gf100_ram_probe_fbp_amount,
.probe_fbpa_amount = gf100_ram_probe_fbpa_amount,
.init = gf100_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
.calc = gf100_ram_calc,
.prog = gf100_ram_prog,
.tidy = gf100_ram_tidy,
};
int
gf100_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
{
return gf100_ram_new_(&gf100_ram, fb, pram);
}
/*
* Copyright 2017 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 <bskeggs@redhat.com>
*/
#include "ram.h"
u32
gf108_ram_probe_fbp_amount(const struct nvkm_ram_func *func, u32 fbpao,
struct nvkm_device *device, int fbp, int *pltcs)
{
u32 fbpt = nvkm_rd32(device, 0x022438);
u32 fbpat = nvkm_rd32(device, 0x02243c);
u32 fbpas = fbpat / fbpt;
u32 fbpa = fbp * fbpas;
u32 size = 0;
while (fbpas--) {
if (!(fbpao & BIT(fbpa)))
size += func->probe_fbpa_amount(device, fbpa);
fbpa++;
}
*pltcs = 1;
return size;
}
static const struct nvkm_ram_func
gf108_ram = {
.upper = 0x0200000000,
.probe_fbp = gf100_ram_probe_fbp,
.probe_fbp_amount = gf108_ram_probe_fbp_amount,
.probe_fbpa_amount = gf100_ram_probe_fbpa_amount,
.init = gf100_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
.calc = gf100_ram_calc,
.prog = gf100_ram_prog,
.tidy = gf100_ram_tidy,
};
int
gf108_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
{
return gf100_ram_new_(&gf108_ram, fb, pram);
}
......@@ -1108,7 +1108,7 @@ gk104_ram_calc_xits(struct gk104_ram *ram, struct nvkm_ram_data *next)
return ret;
}
static int
int
gk104_ram_calc(struct nvkm_ram *base, u32 freq)
{
struct gk104_ram *ram = gk104_ram(base);
......@@ -1227,7 +1227,7 @@ gk104_ram_prog_0(struct gk104_ram *ram, u32 freq)
nvkm_mask(device, 0x10f444, mask, data);
}
static int
int
gk104_ram_prog(struct nvkm_ram *base)
{
struct gk104_ram *ram = gk104_ram(base);
......@@ -1247,7 +1247,7 @@ gk104_ram_prog(struct nvkm_ram *base)
return (ram->base.next == &ram->base.xition);
}
static void
void
gk104_ram_tidy(struct nvkm_ram *base)
{
struct gk104_ram *ram = gk104_ram(base);
......@@ -1509,7 +1509,7 @@ gk104_ram_ctor_data(struct gk104_ram *ram, u8 ramcfg, int i)
return ret;
}
static void *
void *
gk104_ram_dtor(struct nvkm_ram *base)
{
struct gk104_ram *ram = gk104_ram(base);
......@@ -1522,31 +1522,14 @@ gk104_ram_dtor(struct nvkm_ram *base)
return ram;
}
static const struct nvkm_ram_func
gk104_ram_func = {
.dtor = gk104_ram_dtor,
.init = gk104_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
.calc = gk104_ram_calc,
.prog = gk104_ram_prog,
.tidy = gk104_ram_tidy,
};
int
gk104_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
{
return gk104_ram_ctor(fb, pram, 0x022554);
}
int
gk104_ram_ctor(struct nvkm_fb *fb, struct nvkm_ram **pram, u32 maskaddr)
gk104_ram_new_(const struct nvkm_ram_func *func, struct nvkm_fb *fb,
struct nvkm_ram **pram)
{
struct nvkm_subdev *subdev = &fb->subdev;
struct nvkm_device *device = subdev->device;
struct nvkm_bios *bios = device->bios;
struct nvkm_gpio *gpio = device->gpio;
struct dcb_gpio_func func;
struct dcb_gpio_func gpio;
struct gk104_ram *ram;
int ret, i;
u8 ramcfg = nvbios_ramcfg_index(subdev);
......@@ -1556,7 +1539,7 @@ gk104_ram_ctor(struct nvkm_fb *fb, struct nvkm_ram **pram, u32 maskaddr)
return -ENOMEM;
*pram = &ram->base;
ret = gf100_ram_ctor(&gk104_ram_func, fb, maskaddr, &ram->base);
ret = gf100_ram_ctor(func, fb, &ram->base);
if (ret)
return ret;
......@@ -1614,18 +1597,18 @@ gk104_ram_ctor(struct nvkm_fb *fb, struct nvkm_ram **pram, u32 maskaddr)
}
/* lookup memory voltage gpios */
ret = nvkm_gpio_find(gpio, 0, 0x18, DCB_GPIO_UNUSED, &func);
ret = nvkm_gpio_find(device->gpio, 0, 0x18, DCB_GPIO_UNUSED, &gpio);
if (ret == 0) {
ram->fuc.r_gpioMV = ramfuc_reg(0x00d610 + (func.line * 0x04));
ram->fuc.r_funcMV[0] = (func.log[0] ^ 2) << 12;
ram->fuc.r_funcMV[1] = (func.log[1] ^ 2) << 12;
ram->fuc.r_gpioMV = ramfuc_reg(0x00d610 + (gpio.line * 0x04));
ram->fuc.r_funcMV[0] = (gpio.log[0] ^ 2) << 12;
ram->fuc.r_funcMV[1] = (gpio.log[1] ^ 2) << 12;
}
ret = nvkm_gpio_find(gpio, 0, 0x2e, DCB_GPIO_UNUSED, &func);
ret = nvkm_gpio_find(device->gpio, 0, 0x2e, DCB_GPIO_UNUSED, &gpio);
if (ret == 0) {
ram->fuc.r_gpio2E = ramfuc_reg(0x00d610 + (func.line * 0x04));
ram->fuc.r_func2E[0] = (func.log[0] ^ 2) << 12;
ram->fuc.r_func2E[1] = (func.log[1] ^ 2) << 12;
ram->fuc.r_gpio2E = ramfuc_reg(0x00d610 + (gpio.line * 0x04));
ram->fuc.r_func2E[0] = (gpio.log[0] ^ 2) << 12;
ram->fuc.r_func2E[1] = (gpio.log[1] ^ 2) << 12;
}
ram->fuc.r_gpiotrig = ramfuc_reg(0x00d604);
......@@ -1717,3 +1700,24 @@ gk104_ram_ctor(struct nvkm_fb *fb, struct nvkm_ram **pram, u32 maskaddr)
ram->fuc.r_0x100750 = ramfuc_reg(0x100750);
return 0;
}
static const struct nvkm_ram_func
gk104_ram = {
.upper = 0x0200000000,
.probe_fbp = gf100_ram_probe_fbp,
.probe_fbp_amount = gf108_ram_probe_fbp_amount,
.probe_fbpa_amount = gf100_ram_probe_fbpa_amount,
.dtor = gk104_ram_dtor,
.init = gk104_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
.calc = gk104_ram_calc,
.prog = gk104_ram_prog,
.tidy = gk104_ram_tidy,
};
int
gk104_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
{
return gk104_ram_new_(&gk104_ram, fb, pram);
}
......@@ -23,8 +23,31 @@
*/
#include "ram.h"
u32
gm107_ram_probe_fbp(const struct nvkm_ram_func *func,
struct nvkm_device *device, int fbp, int *pltcs)
{
u32 fbpao = nvkm_rd32(device, 0x021c14);
return func->probe_fbp_amount(func, fbpao, device, fbp, pltcs);
}
static const struct nvkm_ram_func
gm107_ram = {
.upper = 0x1000000000,
.probe_fbp = gm107_ram_probe_fbp,
.probe_fbp_amount = gf108_ram_probe_fbp_amount,
.probe_fbpa_amount = gf100_ram_probe_fbpa_amount,
.dtor = gk104_ram_dtor,
.init = gk104_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
.calc = gk104_ram_calc,
.prog = gk104_ram_prog,
.tidy = gk104_ram_tidy,
};
int
gm107_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
{
return gk104_ram_ctor(fb, pram, 0x021c14);
return gk104_ram_new_(&gm107_ram, fb, pram);
}
/*
* Copyright 2017 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 <bskeggs@redhat.com>
*/
#include "ram.h"
u32
gm200_ram_probe_fbp_amount(const struct nvkm_ram_func *func, u32 fbpao,
struct nvkm_device *device, int fbp, int *pltcs)
{
u32 ltcs = nvkm_rd32(device, 0x022450);
u32 fbpas = nvkm_rd32(device, 0x022458);
u32 fbpa = fbp * fbpas;
u32 size = 0;
if (!(nvkm_rd32(device, 0x021d38) & BIT(fbp))) {
u32 ltco = nvkm_rd32(device, 0x021d70 + (fbp * 4));
u32 ltcm = ~ltco & ((1 << ltcs) - 1);
while (fbpas--) {
if (!(fbpao & (1 << fbpa)))
size += func->probe_fbpa_amount(device, fbpa);
fbpa++;
}
*pltcs = hweight32(ltcm);
}
return size;
}
static const struct nvkm_ram_func
gm200_ram = {
.upper = 0x1000000000,
.probe_fbp = gm107_ram_probe_fbp,
.probe_fbp_amount = gm200_ram_probe_fbp_amount,
.probe_fbpa_amount = gf100_ram_probe_fbpa_amount,
.dtor = gk104_ram_dtor,
.init = gk104_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
.calc = gk104_ram_calc,
.prog = gk104_ram_prog,
.tidy = gk104_ram_tidy,
};
int
gm200_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
{
return gk104_ram_new_(&gm200_ram, fb, pram);
}
......@@ -76,8 +76,18 @@ gp100_ram_init(struct nvkm_ram *ram)
return 0;
}
static u32
gp100_ram_probe_fbpa(struct nvkm_device *device, int fbpa)
{
return nvkm_rd32(device, 0x90020c + (fbpa * 0x4000));
}
static const struct nvkm_ram_func
gp100_ram_func = {
gp100_ram = {
.upper = 0x1000000000,
.probe_fbp = gm107_ram_probe_fbp,
.probe_fbp_amount = gm200_ram_probe_fbp_amount,
.probe_fbpa_amount = gp100_ram_probe_fbpa,
.init = gp100_ram_init,
.get = gf100_ram_get,
.put = gf100_ram_put,
......@@ -87,60 +97,10 @@ int
gp100_ram_new(struct nvkm_fb *fb, struct nvkm_ram **pram)
{
struct nvkm_ram *ram;
struct nvkm_subdev *subdev = &fb->subdev;
struct nvkm_device *device = subdev->device;
enum nvkm_ram_type type = nvkm_fb_bios_memtype(device->bios);
const u32 rsvd_head = ( 256 * 1024); /* vga memory */
const u32 rsvd_tail = (1024 * 1024); /* vbios etc */
u32 fbpa_num = nvkm_rd32(device, 0x02243c), fbpa;
u32 fbio_opt = nvkm_rd32(device, 0x021c14);
u64 part, size = 0, comm = ~0ULL;
bool mixed = false;
int ret;
nvkm_debug(subdev, "02243c: %08x\n", fbpa_num);
nvkm_debug(subdev, "021c14: %08x\n", fbio_opt);
for (fbpa = 0; fbpa < fbpa_num; fbpa++) {
if (!(fbio_opt & (1 << fbpa))) {
part = nvkm_rd32(device, 0x90020c + (fbpa * 0x4000));
nvkm_debug(subdev, "fbpa %02x: %lld MiB\n", fbpa, part);
part = part << 20;
if (part != comm) {
if (comm != ~0ULL)
mixed = true;
comm = min(comm, part);
}
size = size + part;
}
}
ret = nvkm_ram_new_(&gp100_ram_func, fb, type, size, 0, &ram);
*pram = ram;
if (ret)
return ret;
nvkm_mm_fini(&ram->vram);
if (!(ram = *pram = kzalloc(sizeof(*ram), GFP_KERNEL)))
return -ENOMEM;
if (mixed) {
ret = nvkm_mm_init(&ram->vram, rsvd_head >> NVKM_RAM_MM_SHIFT,
((comm * fbpa_num) - rsvd_head) >>
NVKM_RAM_MM_SHIFT, 1);
if (ret)
return ret;
return gf100_ram_ctor(&gp100_ram, fb, ram);
ret = nvkm_mm_init(&ram->vram, (0x1000000000ULL + comm) >>
NVKM_RAM_MM_SHIFT,
(size - (comm * fbpa_num) - rsvd_tail) >>
NVKM_RAM_MM_SHIFT, 1);
if (ret)
return ret;
} else {
ret = nvkm_mm_init(&ram->vram, rsvd_head >> NVKM_RAM_MM_SHIFT,
(size - rsvd_head - rsvd_tail) >>
NVKM_RAM_MM_SHIFT, 1);
if (ret)
return ret;
}
return 0;
}
......@@ -134,7 +134,7 @@ struct anx9805_aux {
static int
anx9805_aux_xfer(struct nvkm_i2c_aux *base, bool retry,
u8 type, u32 addr, u8 *data, u8 size)
u8 type, u32 addr, u8 *data, u8 *size)
{
struct anx9805_aux *aux = anx9805_aux(base);
struct anx9805_pad *pad = aux->pad;
......@@ -143,7 +143,7 @@ anx9805_aux_xfer(struct nvkm_i2c_aux *base, bool retry,
u8 buf[16] = {};
u8 tmp;
AUX_DBG(&aux->base, "%02x %05x %d", type, addr, size);
AUX_DBG(&aux->base, "%02x %05x %d", type, addr, *size);
tmp = nvkm_rdi2cr(adap, pad->addr, 0x07) & ~0x04;
nvkm_wri2cr(adap, pad->addr, 0x07, tmp | 0x04);
......@@ -152,12 +152,12 @@ anx9805_aux_xfer(struct nvkm_i2c_aux *base, bool retry,
nvkm_wri2cr(adap, aux->addr, 0xe4, 0x80);
if (!(type & 1)) {
memcpy(buf, data, size);
memcpy(buf, data, *size);
AUX_DBG(&aux->base, "%16ph", buf);
for (i = 0; i < size; i++)
for (i = 0; i < *size; i++)
nvkm_wri2cr(adap, aux->addr, 0xf0 + i, buf[i]);
}
nvkm_wri2cr(adap, aux->addr, 0xe5, ((size - 1) << 4) | type);
nvkm_wri2cr(adap, aux->addr, 0xe5, ((*size - 1) << 4) | type);
nvkm_wri2cr(adap, aux->addr, 0xe6, (addr & 0x000ff) >> 0);
nvkm_wri2cr(adap, aux->addr, 0xe7, (addr & 0x0ff00) >> 8);
nvkm_wri2cr(adap, aux->addr, 0xe8, (addr & 0xf0000) >> 16);
......@@ -176,10 +176,10 @@ anx9805_aux_xfer(struct nvkm_i2c_aux *base, bool retry,
}
if (type & 1) {
for (i = 0; i < size; i++)
for (i = 0; i < *size; i++)
buf[i] = nvkm_rdi2cr(adap, aux->addr, 0xf0 + i);
AUX_DBG(&aux->base, "%16ph", buf);
memcpy(data, buf, size);
memcpy(data, buf, *size);
}
ret = 0;
......
......@@ -51,7 +51,7 @@ nvkm_i2c_aux_i2c_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
if (mcnt || remaining > 16)
cmd |= 4; /* MOT */
ret = aux->func->xfer(aux, true, cmd, msg->addr, ptr, cnt);
ret = aux->func->xfer(aux, true, cmd, msg->addr, ptr, &cnt);
if (ret < 0) {
nvkm_i2c_aux_release(aux);
return ret;
......@@ -115,7 +115,7 @@ nvkm_i2c_aux_acquire(struct nvkm_i2c_aux *aux)
int
nvkm_i2c_aux_xfer(struct nvkm_i2c_aux *aux, bool retry, u8 type,
u32 addr, u8 *data, u8 size)
u32 addr, u8 *data, u8 *size)
{
return aux->func->xfer(aux, retry, type, addr, data, size);
}
......
......@@ -4,7 +4,7 @@
struct nvkm_i2c_aux_func {
int (*xfer)(struct nvkm_i2c_aux *, bool retry, u8 type,
u32 addr, u8 *data, u8 size);
u32 addr, u8 *data, u8 *size);
int (*lnk_ctl)(struct nvkm_i2c_aux *, int link_nr, int link_bw,
bool enhanced_framing);
};
......@@ -15,7 +15,7 @@ int nvkm_i2c_aux_new_(const struct nvkm_i2c_aux_func *, struct nvkm_i2c_pad *,
int id, struct nvkm_i2c_aux **);
void nvkm_i2c_aux_del(struct nvkm_i2c_aux **);
int nvkm_i2c_aux_xfer(struct nvkm_i2c_aux *, bool retry, u8 type,
u32 addr, u8 *data, u8 size);
u32 addr, u8 *data, u8 *size);
int g94_i2c_aux_new(struct nvkm_i2c_pad *, int, u8, struct nvkm_i2c_aux **);
int gm200_i2c_aux_new(struct nvkm_i2c_pad *, int, u8, struct nvkm_i2c_aux **);
......
......@@ -74,7 +74,7 @@ g94_i2c_aux_init(struct g94_i2c_aux *aux)
static int
g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
u8 type, u32 addr, u8 *data, u8 size)
u8 type, u32 addr, u8 *data, u8 *size)
{
struct g94_i2c_aux *aux = g94_i2c_aux(obj);
struct nvkm_device *device = aux->base.pad->i2c->subdev.device;
......@@ -83,7 +83,7 @@ g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
u32 xbuf[4] = {};
int ret, i;
AUX_TRACE(&aux->base, "%d: %08x %d", type, addr, size);
AUX_TRACE(&aux->base, "%d: %08x %d", type, addr, *size);
ret = g94_i2c_aux_init(aux);
if (ret < 0)
......@@ -97,7 +97,7 @@ g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
}
if (!(type & 1)) {
memcpy(xbuf, data, size);
memcpy(xbuf, data, *size);
for (i = 0; i < 16; i += 4) {
AUX_TRACE(&aux->base, "wr %08x", xbuf[i / 4]);
nvkm_wr32(device, 0x00e4c0 + base + i, xbuf[i / 4]);
......@@ -107,7 +107,7 @@ g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
ctrl = nvkm_rd32(device, 0x00e4e4 + base);
ctrl &= ~0x0001f0ff;
ctrl |= type << 12;
ctrl |= size - 1;
ctrl |= *size - 1;
nvkm_wr32(device, 0x00e4e0 + base, addr);
/* (maybe) retry transaction a number of times on failure... */
......@@ -151,7 +151,8 @@ g94_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
xbuf[i / 4] = nvkm_rd32(device, 0x00e4d0 + base + i);
AUX_TRACE(&aux->base, "rd %08x", xbuf[i / 4]);
}
memcpy(data, xbuf, size);
memcpy(data, xbuf, *size);
*size = stat & 0x0000001f;
}
out:
......
......@@ -74,7 +74,7 @@ gm200_i2c_aux_init(struct gm200_i2c_aux *aux)
static int
gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
u8 type, u32 addr, u8 *data, u8 size)
u8 type, u32 addr, u8 *data, u8 *size)
{
struct gm200_i2c_aux *aux = gm200_i2c_aux(obj);
struct nvkm_device *device = aux->base.pad->i2c->subdev.device;
......@@ -83,7 +83,7 @@ gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
u32 xbuf[4] = {};
int ret, i;
AUX_TRACE(&aux->base, "%d: %08x %d", type, addr, size);
AUX_TRACE(&aux->base, "%d: %08x %d", type, addr, *size);
ret = gm200_i2c_aux_init(aux);
if (ret < 0)
......@@ -97,7 +97,7 @@ gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
}
if (!(type & 1)) {
memcpy(xbuf, data, size);
memcpy(xbuf, data, *size);
for (i = 0; i < 16; i += 4) {
AUX_TRACE(&aux->base, "wr %08x", xbuf[i / 4]);
nvkm_wr32(device, 0x00d930 + base + i, xbuf[i / 4]);
......@@ -107,7 +107,7 @@ gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
ctrl = nvkm_rd32(device, 0x00d954 + base);
ctrl &= ~0x0001f0ff;
ctrl |= type << 12;
ctrl |= size - 1;
ctrl |= *size - 1;
nvkm_wr32(device, 0x00d950 + base, addr);
/* (maybe) retry transaction a number of times on failure... */
......@@ -151,7 +151,8 @@ gm200_i2c_aux_xfer(struct nvkm_i2c_aux *obj, bool retry,
xbuf[i / 4] = nvkm_rd32(device, 0x00d940 + base + i);
AUX_TRACE(&aux->base, "rd %08x", xbuf[i / 4]);
}
memcpy(data, xbuf, size);
memcpy(data, xbuf, *size);
*size = stat & 0x0000001f;
}
out:
......
......@@ -30,7 +30,7 @@ gf100_ibus_intr_hub(struct nvkm_subdev *ibus, int i)
u32 addr = nvkm_rd32(device, 0x122120 + (i * 0x0400));
u32 data = nvkm_rd32(device, 0x122124 + (i * 0x0400));
u32 stat = nvkm_rd32(device, 0x122128 + (i * 0x0400));
nvkm_error(ibus, "HUB%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_debug(ibus, "HUB%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_mask(device, 0x122128 + (i * 0x0400), 0x00000200, 0x00000000);
}
......@@ -41,7 +41,7 @@ gf100_ibus_intr_rop(struct nvkm_subdev *ibus, int i)
u32 addr = nvkm_rd32(device, 0x124120 + (i * 0x0400));
u32 data = nvkm_rd32(device, 0x124124 + (i * 0x0400));
u32 stat = nvkm_rd32(device, 0x124128 + (i * 0x0400));
nvkm_error(ibus, "ROP%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_debug(ibus, "ROP%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_mask(device, 0x124128 + (i * 0x0400), 0x00000200, 0x00000000);
}
......@@ -52,7 +52,7 @@ gf100_ibus_intr_gpc(struct nvkm_subdev *ibus, int i)
u32 addr = nvkm_rd32(device, 0x128120 + (i * 0x0400));
u32 data = nvkm_rd32(device, 0x128124 + (i * 0x0400));
u32 stat = nvkm_rd32(device, 0x128128 + (i * 0x0400));
nvkm_error(ibus, "GPC%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_debug(ibus, "GPC%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_mask(device, 0x128128 + (i * 0x0400), 0x00000200, 0x00000000);
}
......
......@@ -30,7 +30,7 @@ gk104_ibus_intr_hub(struct nvkm_subdev *ibus, int i)
u32 addr = nvkm_rd32(device, 0x122120 + (i * 0x0800));
u32 data = nvkm_rd32(device, 0x122124 + (i * 0x0800));
u32 stat = nvkm_rd32(device, 0x122128 + (i * 0x0800));
nvkm_error(ibus, "HUB%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_debug(ibus, "HUB%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_mask(device, 0x122128 + (i * 0x0800), 0x00000200, 0x00000000);
}
......@@ -41,7 +41,7 @@ gk104_ibus_intr_rop(struct nvkm_subdev *ibus, int i)
u32 addr = nvkm_rd32(device, 0x124120 + (i * 0x0800));
u32 data = nvkm_rd32(device, 0x124124 + (i * 0x0800));
u32 stat = nvkm_rd32(device, 0x124128 + (i * 0x0800));
nvkm_error(ibus, "ROP%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_debug(ibus, "ROP%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_mask(device, 0x124128 + (i * 0x0800), 0x00000200, 0x00000000);
}
......@@ -52,7 +52,7 @@ gk104_ibus_intr_gpc(struct nvkm_subdev *ibus, int i)
u32 addr = nvkm_rd32(device, 0x128120 + (i * 0x0800));
u32 data = nvkm_rd32(device, 0x128124 + (i * 0x0800));
u32 stat = nvkm_rd32(device, 0x128128 + (i * 0x0800));
nvkm_error(ibus, "GPC%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_debug(ibus, "GPC%d: %06x %08x (%08x)\n", i, addr, data, stat);
nvkm_mask(device, 0x128128 + (i * 0x0800), 0x00000200, 0x00000000);
}
......
......@@ -23,6 +23,7 @@
*/
#include "priv.h"
#include <core/msgqueue.h>
#include <subdev/timer.h>
void
......@@ -85,7 +86,8 @@ nvkm_pmu_reset(struct nvkm_pmu *pmu)
);
/* Reset. */
pmu->func->reset(pmu);
if (pmu->func->reset)
pmu->func->reset(pmu);
/* Wait for IMEM/DMEM scrubbing to be complete. */
nvkm_msec(device, 2000,
......@@ -113,10 +115,18 @@ nvkm_pmu_init(struct nvkm_subdev *subdev)
return ret;
}
static int
nvkm_pmu_oneinit(struct nvkm_subdev *subdev)
{
struct nvkm_pmu *pmu = nvkm_pmu(subdev);
return nvkm_falcon_v1_new(&pmu->subdev, "PMU", 0x10a000, &pmu->falcon);
}
static void *
nvkm_pmu_dtor(struct nvkm_subdev *subdev)
{
struct nvkm_pmu *pmu = nvkm_pmu(subdev);
nvkm_msgqueue_del(&pmu->queue);
nvkm_falcon_del(&pmu->falcon);
return nvkm_pmu(subdev);
}
......@@ -125,6 +135,7 @@ static const struct nvkm_subdev_func
nvkm_pmu = {
.dtor = nvkm_pmu_dtor,
.preinit = nvkm_pmu_preinit,
.oneinit = nvkm_pmu_oneinit,
.init = nvkm_pmu_init,
.fini = nvkm_pmu_fini,
.intr = nvkm_pmu_intr,
......@@ -138,7 +149,7 @@ nvkm_pmu_ctor(const struct nvkm_pmu_func *func, struct nvkm_device *device,
pmu->func = func;
INIT_WORK(&pmu->recv.work, nvkm_pmu_recv);
init_waitqueue_head(&pmu->recv.wait);
return nvkm_falcon_v1_new(&pmu->subdev, "PMU", 0x10a000, &pmu->falcon);
return 0;
}
int
......
......@@ -20,15 +20,30 @@
* DEALINGS IN THE SOFTWARE.
*/
#include <engine/falcon.h>
#include <core/msgqueue.h>
#include "priv.h"
static void
gm20b_pmu_recv(struct nvkm_pmu *pmu)
{
nvkm_msgqueue_recv(pmu->queue);
}
static const struct nvkm_pmu_func
gm20b_pmu = {
.reset = gt215_pmu_reset,
.intr = gt215_pmu_intr,
.recv = gm20b_pmu_recv,
};
int
gm20b_pmu_new(struct nvkm_device *device, int index, struct nvkm_pmu **ppmu)
{
return nvkm_pmu_new_(&gm20b_pmu, device, index, ppmu);
int ret;
ret = nvkm_pmu_new_(&gm20b_pmu, device, index, ppmu);
if (ret)
return ret;
return 0;
}
nvkm-y += nvkm/subdev/secboot/base.o
nvkm-y += nvkm/subdev/secboot/hs_ucode.o
nvkm-y += nvkm/subdev/secboot/ls_ucode_gr.o
nvkm-y += nvkm/subdev/secboot/ls_ucode_msgqueue.o
nvkm-y += nvkm/subdev/secboot/acr.o
nvkm-y += nvkm/subdev/secboot/acr_r352.o
nvkm-y += nvkm/subdev/secboot/acr_r361.o
nvkm-y += nvkm/subdev/secboot/acr_r364.o
nvkm-y += nvkm/subdev/secboot/acr_r367.o
nvkm-y += nvkm/subdev/secboot/acr_r375.o
nvkm-y += nvkm/subdev/secboot/gm200.o
nvkm-y += nvkm/subdev/secboot/gm20b.o
nvkm-y += nvkm/subdev/secboot/gp102.o
......@@ -37,12 +37,10 @@ struct nvkm_acr_func {
void (*dtor)(struct nvkm_acr *);
int (*oneinit)(struct nvkm_acr *, struct nvkm_secboot *);
int (*fini)(struct nvkm_acr *, struct nvkm_secboot *, bool);
int (*load)(struct nvkm_acr *, struct nvkm_secboot *,
int (*load)(struct nvkm_acr *, struct nvkm_falcon *,
struct nvkm_gpuobj *, u64);
int (*reset)(struct nvkm_acr *, struct nvkm_secboot *,
enum nvkm_secboot_falcon);
int (*start)(struct nvkm_acr *, struct nvkm_secboot *,
enum nvkm_secboot_falcon);
};
/**
......@@ -50,7 +48,7 @@ struct nvkm_acr_func {
*
* @boot_falcon: ID of the falcon that will perform secure boot
* @managed_falcons: bitfield of falcons managed by this ACR
* @start_address: virtual start address of the HS bootloader
* @optional_falcons: bitfield of falcons we can live without
*/
struct nvkm_acr {
const struct nvkm_acr_func *func;
......@@ -58,12 +56,15 @@ struct nvkm_acr {
enum nvkm_secboot_falcon boot_falcon;
unsigned long managed_falcons;
u32 start_address;
unsigned long optional_falcons;
};
void *nvkm_acr_load_firmware(const struct nvkm_subdev *, const char *, size_t);
struct nvkm_acr *acr_r352_new(unsigned long);
struct nvkm_acr *acr_r361_new(unsigned long);
struct nvkm_acr *acr_r364_new(unsigned long);
struct nvkm_acr *acr_r367_new(enum nvkm_secboot_falcon, unsigned long);
struct nvkm_acr *acr_r375_new(enum nvkm_secboot_falcon, unsigned long);
#endif
......@@ -24,131 +24,27 @@
#include "acr.h"
#include "ls_ucode.h"
#include "hs_ucode.h"
struct ls_ucode_img;
#define ACR_R352_MAX_APPS 8
/*
*
* LS blob structures
*
*/
/**
* struct acr_r352_lsf_lsb_header - LS firmware header
* @signature: signature to verify the firmware against
* @ucode_off: offset of the ucode blob in the WPR region. The ucode
* blob contains the bootloader, code and data of the
* LS falcon
* @ucode_size: size of the ucode blob, including bootloader
* @data_size: size of the ucode blob data
* @bl_code_size: size of the bootloader code
* @bl_imem_off: offset in imem of the bootloader
* @bl_data_off: offset of the bootloader data in WPR region
* @bl_data_size: size of the bootloader data
* @app_code_off: offset of the app code relative to ucode_off
* @app_code_size: size of the app code
* @app_data_off: offset of the app data relative to ucode_off
* @app_data_size: size of the app data
* @flags: flags for the secure bootloader
*
* This structure is written into the WPR region for each managed falcon. Each
* instance is referenced by the lsb_offset member of the corresponding
* lsf_wpr_header.
*/
struct acr_r352_lsf_lsb_header {
/**
* LS falcon signatures
* @prd_keys: signature to use in production mode
* @dgb_keys: signature to use in debug mode
* @b_prd_present: whether the production key is present
* @b_dgb_present: whether the debug key is present
* @falcon_id: ID of the falcon the ucode applies to
*/
struct {
u8 prd_keys[2][16];
u8 dbg_keys[2][16];
u32 b_prd_present;
u32 b_dbg_present;
u32 falcon_id;
} signature;
u32 ucode_off;
u32 ucode_size;
u32 data_size;
u32 bl_code_size;
u32 bl_imem_off;
u32 bl_data_off;
u32 bl_data_size;
u32 app_code_off;
u32 app_code_size;
u32 app_data_off;
u32 app_data_size;
u32 flags;
#define LSF_FLAG_LOAD_CODE_AT_0 1
#define LSF_FLAG_DMACTL_REQ_CTX 4
#define LSF_FLAG_FORCE_PRIV_LOAD 8
};
/**
* struct acr_r352_lsf_wpr_header - LS blob WPR Header
* @falcon_id: LS falcon ID
* @lsb_offset: offset of the lsb_lsf_header in the WPR region
* @bootstrap_owner: secure falcon reponsible for bootstrapping the LS falcon
* @lazy_bootstrap: skip bootstrapping by ACR
* @status: bootstrapping status
*
* An array of these is written at the beginning of the WPR region, one for
* each managed falcon. The array is terminated by an instance which falcon_id
* is LSF_FALCON_ID_INVALID.
*/
struct acr_r352_lsf_wpr_header {
u32 falcon_id;
u32 lsb_offset;
u32 bootstrap_owner;
u32 lazy_bootstrap;
u32 status;
#define LSF_IMAGE_STATUS_NONE 0
#define LSF_IMAGE_STATUS_COPY 1
#define LSF_IMAGE_STATUS_VALIDATION_CODE_FAILED 2
#define LSF_IMAGE_STATUS_VALIDATION_DATA_FAILED 3
#define LSF_IMAGE_STATUS_VALIDATION_DONE 4
#define LSF_IMAGE_STATUS_VALIDATION_SKIPPED 5
#define LSF_IMAGE_STATUS_BOOTSTRAP_READY 6
};
/**
* struct ls_ucode_img_r352 - ucode image augmented with r352 headers
*/
struct ls_ucode_img_r352 {
struct ls_ucode_img base;
struct acr_r352_lsf_wpr_header wpr_header;
struct acr_r352_lsf_lsb_header lsb_header;
};
#define ls_ucode_img_r352(i) container_of(i, struct ls_ucode_img_r352, base)
/*
* HS blob structures
*/
struct hsf_load_header_app {
u32 sec_code_off;
u32 sec_code_size;
};
static inline u32
hsf_load_header_app_off(const struct hsf_load_header *hdr, u32 app)
{
return hdr->apps[app];
}
/**
* struct hsf_load_header - HS firmware load header
*/
struct hsf_load_header {
u32 non_sec_code_off;
u32 non_sec_code_size;
u32 data_dma_base;
u32 data_size;
u32 num_apps;
struct hsf_load_header_app app[0];
};
static inline u32
hsf_load_header_app_size(const struct hsf_load_header *hdr, u32 app)
{
return hdr->apps[hdr->num_apps + app];
}
/**
* struct acr_r352_ls_func - manages a single LS firmware
......@@ -157,6 +53,7 @@ struct hsf_load_header {
* @generate_bl_desc: function called on a block of bl_desc_size to generate the
* proper bootloader descriptor for this LS firmware
* @bl_desc_size: size of the bootloader descriptor
* @post_run: hook called right after the ACR is executed
* @lhdr_flags: LS flags
*/
struct acr_r352_ls_func {
......@@ -164,6 +61,7 @@ struct acr_r352_ls_func {
void (*generate_bl_desc)(const struct nvkm_acr *,
const struct ls_ucode_img *, u64, void *);
u32 bl_desc_size;
void (*post_run)(const struct nvkm_acr *, const struct nvkm_secboot *);
u32 lhdr_flags;
};
......@@ -179,13 +77,15 @@ struct acr_r352;
struct acr_r352_func {
void (*generate_hs_bl_desc)(const struct hsf_load_header *, void *,
u64);
void (*fixup_hs_desc)(struct acr_r352 *, struct nvkm_secboot *, void *);
u32 hs_bl_desc_size;
bool shadow_blob;
struct ls_ucode_img *(*ls_ucode_img_load)(const struct acr_r352 *,
enum nvkm_secboot_falcon);
int (*ls_fill_headers)(struct acr_r352 *, struct list_head *);
int (*ls_write_wpr)(struct acr_r352 *, struct list_head *,
struct nvkm_gpuobj *, u32);
struct nvkm_gpuobj *, u64);
const struct acr_r352_ls_func *ls_func[NVKM_SECBOOT_FALCON_END];
};
......@@ -204,19 +104,22 @@ struct acr_r352 {
struct nvkm_gpuobj *load_blob;
struct {
struct hsf_load_header load_bl_header;
struct hsf_load_header_app __load_apps[ACR_R352_MAX_APPS];
u32 __load_apps[ACR_R352_MAX_APPS * 2];
};
/* HS FW - unlock WPR region (dGPU only) */
struct nvkm_gpuobj *unload_blob;
struct {
struct hsf_load_header unload_bl_header;
struct hsf_load_header_app __unload_apps[ACR_R352_MAX_APPS];
u32 __unload_apps[ACR_R352_MAX_APPS * 2];
};
/* HS bootloader */
void *hsbl_blob;
/* HS bootloader for unload blob, if using a different falcon */
void *hsbl_unload_blob;
/* LS FWs, to be loaded by the HS ACR */
struct nvkm_gpuobj *ls_blob;
......@@ -245,6 +148,8 @@ struct ls_ucode_img *acr_r352_ls_ucode_img_load(const struct acr_r352 *,
enum nvkm_secboot_falcon);
int acr_r352_ls_fill_headers(struct acr_r352 *, struct list_head *);
int acr_r352_ls_write_wpr(struct acr_r352 *, struct list_head *,
struct nvkm_gpuobj *, u32);
struct nvkm_gpuobj *, u64);
void acr_r352_fixup_hs_desc(struct acr_r352 *, struct nvkm_secboot *, void *);
#endif
......@@ -20,58 +20,23 @@
* DEALINGS IN THE SOFTWARE.
*/
#include "acr_r352.h"
#include "acr_r361.h"
#include <engine/falcon.h>
/**
* struct acr_r361_flcn_bl_desc - DMEM bootloader descriptor
* @signature: 16B signature for secure code. 0s if no secure code
* @ctx_dma: DMA context to be used by BL while loading code/data
* @code_dma_base: 256B-aligned Physical FB Address where code is located
* (falcon's $xcbase register)
* @non_sec_code_off: offset from code_dma_base where the non-secure code is
* located. The offset must be multiple of 256 to help perf
* @non_sec_code_size: the size of the nonSecure code part.
* @sec_code_off: offset from code_dma_base where the secure code is
* located. The offset must be multiple of 256 to help perf
* @sec_code_size: offset from code_dma_base where the secure code is
* located. The offset must be multiple of 256 to help perf
* @code_entry_point: code entry point which will be invoked by BL after
* code is loaded.
* @data_dma_base: 256B aligned Physical FB Address where data is located.
* (falcon's $xdbase register)
* @data_size: size of data block. Should be multiple of 256B
*
* Structure used by the bootloader to load the rest of the code. This has
* to be filled by host and copied into DMEM at offset provided in the
* hsflcn_bl_desc.bl_desc_dmem_load_off.
*/
struct acr_r361_flcn_bl_desc {
u32 reserved[4];
u32 signature[4];
u32 ctx_dma;
struct flcn_u64 code_dma_base;
u32 non_sec_code_off;
u32 non_sec_code_size;
u32 sec_code_off;
u32 sec_code_size;
u32 code_entry_point;
struct flcn_u64 data_dma_base;
u32 data_size;
};
#include <core/msgqueue.h>
#include <subdev/pmu.h>
#include <engine/sec2.h>
static void
acr_r361_generate_flcn_bl_desc(const struct nvkm_acr *acr,
const struct ls_ucode_img *_img, u64 wpr_addr,
const struct ls_ucode_img *img, u64 wpr_addr,
void *_desc)
{
struct ls_ucode_img_r352 *img = ls_ucode_img_r352(_img);
struct acr_r361_flcn_bl_desc *desc = _desc;
const struct ls_ucode_img_desc *pdesc = &img->base.ucode_desc;
const struct ls_ucode_img_desc *pdesc = &img->ucode_desc;
u64 base, addr_code, addr_data;
base = wpr_addr + img->lsb_header.ucode_off + pdesc->app_start_offset;
base = wpr_addr + img->ucode_off + pdesc->app_start_offset;
addr_code = base + pdesc->app_resident_code_offset;
addr_data = base + pdesc->app_resident_data_offset;
......@@ -84,7 +49,7 @@ acr_r361_generate_flcn_bl_desc(const struct nvkm_acr *acr,
desc->data_size = pdesc->app_resident_data_size;
}
static void
void
acr_r361_generate_hs_bl_desc(const struct hsf_load_header *hdr, void *_bl_desc,
u64 offset)
{
......@@ -94,8 +59,8 @@ acr_r361_generate_hs_bl_desc(const struct hsf_load_header *hdr, void *_bl_desc,
bl_desc->code_dma_base = u64_to_flcn64(offset);
bl_desc->non_sec_code_off = hdr->non_sec_code_off;
bl_desc->non_sec_code_size = hdr->non_sec_code_size;
bl_desc->sec_code_off = hdr->app[0].sec_code_off;
bl_desc->sec_code_size = hdr->app[0].sec_code_size;
bl_desc->sec_code_off = hsf_load_header_app_off(hdr, 0);
bl_desc->sec_code_size = hsf_load_header_app_size(hdr, 0);
bl_desc->code_entry_point = 0;
bl_desc->data_dma_base = u64_to_flcn64(offset + hdr->data_dma_base);
bl_desc->data_size = hdr->data_size;
......@@ -117,8 +82,100 @@ acr_r361_ls_gpccs_func = {
.lhdr_flags = LSF_FLAG_FORCE_PRIV_LOAD,
};
struct acr_r361_pmu_bl_desc {
u32 reserved;
u32 dma_idx;
struct flcn_u64 code_dma_base;
u32 total_code_size;
u32 code_size_to_load;
u32 code_entry_point;
struct flcn_u64 data_dma_base;
u32 data_size;
struct flcn_u64 overlay_dma_base;
u32 argc;
u32 argv;
};
static void
acr_r361_generate_pmu_bl_desc(const struct nvkm_acr *acr,
const struct ls_ucode_img *img, u64 wpr_addr,
void *_desc)
{
const struct ls_ucode_img_desc *pdesc = &img->ucode_desc;
const struct nvkm_pmu *pmu = acr->subdev->device->pmu;
struct acr_r361_pmu_bl_desc *desc = _desc;
u64 base, addr_code, addr_data;
u32 addr_args;
base = wpr_addr + img->ucode_off + pdesc->app_start_offset;
addr_code = base + pdesc->app_resident_code_offset;
addr_data = base + pdesc->app_resident_data_offset;
addr_args = pmu->falcon->data.limit;
addr_args -= NVKM_MSGQUEUE_CMDLINE_SIZE;
desc->dma_idx = FALCON_DMAIDX_UCODE;
desc->code_dma_base = u64_to_flcn64(addr_code);
desc->total_code_size = pdesc->app_size;
desc->code_size_to_load = pdesc->app_resident_code_size;
desc->code_entry_point = pdesc->app_imem_entry;
desc->data_dma_base = u64_to_flcn64(addr_data);
desc->data_size = pdesc->app_resident_data_size;
desc->overlay_dma_base = u64_to_flcn64(addr_code);
desc->argc = 1;
desc->argv = addr_args;
}
const struct acr_r352_ls_func
acr_r361_ls_pmu_func = {
.load = acr_ls_ucode_load_pmu,
.generate_bl_desc = acr_r361_generate_pmu_bl_desc,
.bl_desc_size = sizeof(struct acr_r361_pmu_bl_desc),
.post_run = acr_ls_pmu_post_run,
};
static void
acr_r361_generate_sec2_bl_desc(const struct nvkm_acr *acr,
const struct ls_ucode_img *img, u64 wpr_addr,
void *_desc)
{
const struct ls_ucode_img_desc *pdesc = &img->ucode_desc;
const struct nvkm_sec2 *sec = acr->subdev->device->sec2;
struct acr_r361_pmu_bl_desc *desc = _desc;
u64 base, addr_code, addr_data;
u32 addr_args;
base = wpr_addr + img->ucode_off + pdesc->app_start_offset;
/* For some reason we should not add app_resident_code_offset here */
addr_code = base;
addr_data = base + pdesc->app_resident_data_offset;
addr_args = sec->falcon->data.limit;
addr_args -= NVKM_MSGQUEUE_CMDLINE_SIZE;
desc->dma_idx = FALCON_SEC2_DMAIDX_UCODE;
desc->code_dma_base = u64_to_flcn64(addr_code);
desc->total_code_size = pdesc->app_size;
desc->code_size_to_load = pdesc->app_resident_code_size;
desc->code_entry_point = pdesc->app_imem_entry;
desc->data_dma_base = u64_to_flcn64(addr_data);
desc->data_size = pdesc->app_resident_data_size;
desc->overlay_dma_base = u64_to_flcn64(addr_code);
desc->argc = 1;
/* args are stored at the beginning of EMEM */
desc->argv = 0x01000000;
}
const struct acr_r352_ls_func
acr_r361_ls_sec2_func = {
.load = acr_ls_ucode_load_sec2,
.generate_bl_desc = acr_r361_generate_sec2_bl_desc,
.bl_desc_size = sizeof(struct acr_r361_pmu_bl_desc),
.post_run = acr_ls_sec2_post_run,
};
const struct acr_r352_func
acr_r361_func = {
.fixup_hs_desc = acr_r352_fixup_hs_desc,
.generate_hs_bl_desc = acr_r361_generate_hs_bl_desc,
.hs_bl_desc_size = sizeof(struct acr_r361_flcn_bl_desc),
.ls_ucode_img_load = acr_r352_ls_ucode_img_load,
......@@ -127,6 +184,8 @@ acr_r361_func = {
.ls_func = {
[NVKM_SECBOOT_FALCON_FECS] = &acr_r361_ls_fecs_func,
[NVKM_SECBOOT_FALCON_GPCCS] = &acr_r361_ls_gpccs_func,
[NVKM_SECBOOT_FALCON_PMU] = &acr_r361_ls_pmu_func,
[NVKM_SECBOOT_FALCON_SEC2] = &acr_r361_ls_sec2_func,
},
};
......
/*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
*
* 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 AUTHORS OR COPYRIGHT HOLDERS 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.
*/
#ifndef __NVKM_SECBOOT_ACR_R361_H__
#define __NVKM_SECBOOT_ACR_R361_H__
#include "acr_r352.h"
/**
* struct acr_r361_flcn_bl_desc - DMEM bootloader descriptor
* @signature: 16B signature for secure code. 0s if no secure code
* @ctx_dma: DMA context to be used by BL while loading code/data
* @code_dma_base: 256B-aligned Physical FB Address where code is located
* (falcon's $xcbase register)
* @non_sec_code_off: offset from code_dma_base where the non-secure code is
* located. The offset must be multiple of 256 to help perf
* @non_sec_code_size: the size of the nonSecure code part.
* @sec_code_off: offset from code_dma_base where the secure code is
* located. The offset must be multiple of 256 to help perf
* @sec_code_size: offset from code_dma_base where the secure code is
* located. The offset must be multiple of 256 to help perf
* @code_entry_point: code entry point which will be invoked by BL after
* code is loaded.
* @data_dma_base: 256B aligned Physical FB Address where data is located.
* (falcon's $xdbase register)
* @data_size: size of data block. Should be multiple of 256B
*
* Structure used by the bootloader to load the rest of the code. This has
* to be filled by host and copied into DMEM at offset provided in the
* hsflcn_bl_desc.bl_desc_dmem_load_off.
*/
struct acr_r361_flcn_bl_desc {
u32 reserved[4];
u32 signature[4];
u32 ctx_dma;
struct flcn_u64 code_dma_base;
u32 non_sec_code_off;
u32 non_sec_code_size;
u32 sec_code_off;
u32 sec_code_size;
u32 code_entry_point;
struct flcn_u64 data_dma_base;
u32 data_size;
};
void acr_r361_generate_hs_bl_desc(const struct hsf_load_header *, void *, u64);
extern const struct acr_r352_ls_func acr_r361_ls_fecs_func;
extern const struct acr_r352_ls_func acr_r361_ls_gpccs_func;
extern const struct acr_r352_ls_func acr_r361_ls_pmu_func;
extern const struct acr_r352_ls_func acr_r361_ls_sec2_func;
#endif
This diff is collapsed.
This diff is collapsed.
/*
* Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved.
*
* 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 AUTHORS OR COPYRIGHT HOLDERS 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.
*/
#ifndef __NVKM_SECBOOT_ACR_R367_H__
#define __NVKM_SECBOOT_ACR_R367_H__
#include "acr_r352.h"
void acr_r367_fixup_hs_desc(struct acr_r352 *, struct nvkm_secboot *, void *);
struct ls_ucode_img *acr_r367_ls_ucode_img_load(const struct acr_r352 *,
enum nvkm_secboot_falcon);
int acr_r367_ls_fill_headers(struct acr_r352 *, struct list_head *);
int acr_r367_ls_write_wpr(struct acr_r352 *, struct list_head *,
struct nvkm_gpuobj *, u64);
#endif
This diff is collapsed.
......@@ -87,6 +87,7 @@
#include <subdev/mc.h>
#include <subdev/timer.h>
#include <subdev/pmu.h>
#include <engine/sec2.h>
const char *
nvkm_secboot_falcon_name[] = {
......@@ -94,6 +95,7 @@ nvkm_secboot_falcon_name[] = {
[NVKM_SECBOOT_FALCON_RESERVED] = "<reserved>",
[NVKM_SECBOOT_FALCON_FECS] = "FECS",
[NVKM_SECBOOT_FALCON_GPCCS] = "GPCCS",
[NVKM_SECBOOT_FALCON_SEC2] = "SEC2",
[NVKM_SECBOOT_FALCON_END] = "<invalid>",
};
/**
......@@ -131,13 +133,20 @@ nvkm_secboot_oneinit(struct nvkm_subdev *subdev)
switch (sb->acr->boot_falcon) {
case NVKM_SECBOOT_FALCON_PMU:
sb->boot_falcon = subdev->device->pmu->falcon;
sb->halt_falcon = sb->boot_falcon = subdev->device->pmu->falcon;
break;
case NVKM_SECBOOT_FALCON_SEC2:
/* we must keep SEC2 alive forever since ACR will run on it */
nvkm_engine_ref(&subdev->device->sec2->engine);
sb->boot_falcon = subdev->device->sec2->falcon;
sb->halt_falcon = subdev->device->pmu->falcon;
break;
default:
nvkm_error(subdev, "Unmanaged boot falcon %s!\n",
nvkm_secboot_falcon_name[sb->acr->boot_falcon]);
return -EINVAL;
}
nvkm_debug(subdev, "using %s falcon for ACR\n", sb->boot_falcon->name);
/* Call chip-specific init function */
if (sb->func->oneinit)
......
......@@ -38,6 +38,7 @@ struct gm200_secboot {
int gm200_secboot_oneinit(struct nvkm_secboot *);
int gm200_secboot_fini(struct nvkm_secboot *, bool);
void *gm200_secboot_dtor(struct nvkm_secboot *);
int gm200_secboot_run_blob(struct nvkm_secboot *, struct nvkm_gpuobj *);
int gm200_secboot_run_blob(struct nvkm_secboot *, struct nvkm_gpuobj *,
struct nvkm_falcon *);
#endif
......@@ -107,9 +107,12 @@ gm20b_secboot_new(struct nvkm_device *device, int index,
struct gm200_secboot *gsb;
struct nvkm_acr *acr;
acr = acr_r352_new(BIT(NVKM_SECBOOT_FALCON_FECS));
acr = acr_r352_new(BIT(NVKM_SECBOOT_FALCON_FECS) |
BIT(NVKM_SECBOOT_FALCON_PMU));
if (IS_ERR(acr))
return PTR_ERR(acr);
/* Support the initial GM20B firmware release without PMU */
acr->optional_falcons = BIT(NVKM_SECBOOT_FALCON_PMU);
gsb = kzalloc(sizeof(*gsb), GFP_KERNEL);
if (!gsb) {
......@@ -137,3 +140,6 @@ MODULE_FIRMWARE("nvidia/gm20b/gr/sw_ctx.bin");
MODULE_FIRMWARE("nvidia/gm20b/gr/sw_nonctx.bin");
MODULE_FIRMWARE("nvidia/gm20b/gr/sw_bundle_init.bin");
MODULE_FIRMWARE("nvidia/gm20b/gr/sw_method_init.bin");
MODULE_FIRMWARE("nvidia/gm20b/pmu/desc.bin");
MODULE_FIRMWARE("nvidia/gm20b/pmu/image.bin");
MODULE_FIRMWARE("nvidia/gm20b/pmu/sig.bin");
This diff is collapsed.
This diff is collapsed.
This diff is collapsed.
......@@ -30,11 +30,10 @@ struct nvkm_secboot_func {
int (*oneinit)(struct nvkm_secboot *);
int (*fini)(struct nvkm_secboot *, bool suspend);
void *(*dtor)(struct nvkm_secboot *);
int (*run_blob)(struct nvkm_secboot *, struct nvkm_gpuobj *);
int (*run_blob)(struct nvkm_secboot *, struct nvkm_gpuobj *,
struct nvkm_falcon *);
};
extern const char *nvkm_secboot_falcon_name[];
int nvkm_secboot_ctor(const struct nvkm_secboot_func *, struct nvkm_acr *,
struct nvkm_device *, int, struct nvkm_secboot *);
int nvkm_secboot_falcon_reset(struct nvkm_secboot *);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment