nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-24 10:34:43 +03:00
Code Issues Packages Projects Releases Wiki Activity
Files
a299b8857f6dca1ef47f197fd09a897ea0b1964d
linux-nvgpu/drivers/gpu/nvgpu/common/nvlink/nvlink_gv100.c
Tejal Kudav 6a315a5a6d gpu: nvgpu: Add nvlink link_mode_transition unit 
Move code used for transitioning between link and sublink modes
into a separate unit called 'link_mode_transition' under subelement
'nvlink'.
Also aggregate all the link/sublink mode related ops under
g->ops.nvlink.link_mode_transitions.

JIRA NVGPU-2862

Change-Id: I289faa10fd1c83b6b8d028b44fe244531ca42b15
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2035794
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
2019-03-18 15:45:31 -07:00

919 lines
26 KiB
C
Raw Blame History

/*
* Copyright (c) 2018-2019, 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.
*/
#ifdef CONFIG_TEGRA_NVLINK
#include <nvgpu/nvgpu_common.h>
#include <nvgpu/bios.h>
#include <nvgpu/bitops.h>
#include <nvgpu/nvlink.h>
#include <nvgpu/enabled.h>
#include <nvgpu/io.h>
#include <nvgpu/utils.h>
#include <nvgpu/timers.h>
#include <nvgpu/gk20a.h>
#include <nvgpu/top.h>
#include <nvgpu/nvlink_minion.h>
#include <nvgpu/nvlink_link_mode_transitions.h>
#include "nvlink_gv100.h"
#include <nvgpu/hw/gv100/hw_nvlinkip_discovery_gv100.h>
#include <nvgpu/hw/gv100/hw_ioctrl_gv100.h>
#include <nvgpu/hw/gv100/hw_nvl_gv100.h>
#include <nvgpu/hw/gv100/hw_trim_gv100.h>
#define NVL_DEVICE(str) nvlinkip_discovery_common_device_##str##_v()
u32 gv100_nvlink_get_link_reset_mask(struct gk20a *g)
{
u32 reg_data;
reg_data = IOCTRL_REG_RD32(g, ioctrl_reset_r());
return ioctrl_reset_linkreset_v(reg_data);
}
static int gv100_nvlink_state_load_hal(struct gk20a *g)
{
unsigned long discovered = g->nvlink.discovered_links;
g->ops.nvlink.intr.common_intr_enable(g, discovered);
return nvgpu_nvlink_minion_load(g);
}
static const char *gv100_device_type_to_str(u32 type)
{
if (type == NVL_DEVICE(ioctrl)) {
return "IOCTRL";
}
if (type == NVL_DEVICE(dlpl)) {
return "DL/PL";
}
if (type == NVL_DEVICE(nvltlc)) {
return "NVLTLC";
}
if (type == NVL_DEVICE(ioctrlmif)) {
return "IOCTRLMIF";
}
if (type == NVL_DEVICE(nvlipt)) {
return "NVLIPT";
}
if (type == NVL_DEVICE(minion)) {
return "MINION";
}
if (type == NVL_DEVICE(dlpl_multicast)) {
return "DL/PL MULTICAST";
}
if (type == NVL_DEVICE(nvltlc_multicast)) {
return "NVLTLC MULTICAST";
}
if (type == NVL_DEVICE(ioctrlmif_multicast)) {
return "IOCTRLMIF MULTICAST";
}
if (type == NVL_DEVICE(nvltlc_multicast)) {
return "NVLTLC MULTICAST";
}
return "UNKNOWN";
}
/*
* Configure AC coupling
*/
static int gv100_nvlink_minion_configure_ac_coupling(struct gk20a *g,
unsigned long mask, bool sync)
{
int err = 0;
u32 link_id;
u32 temp;
unsigned long bit;
for_each_set_bit(bit, &mask, NVLINK_MAX_LINKS_SW) {
link_id = (u32)bit;
temp = DLPL_REG_RD32(g, link_id, nvl_link_config_r());
temp &= ~nvl_link_config_ac_safe_en_m();
temp |= nvl_link_config_ac_safe_en_on_f();
DLPL_REG_WR32(g, link_id, nvl_link_config_r(), temp);
err = g->ops.nvlink.minion.send_dlcmd(g, link_id,
NVGPU_NVLINK_MINION_DLCMD_SETACMODE, sync);
if (err != 0) {
return err;
}
}
return err;
}
static void gv100_nvlink_prog_alt_clk(struct gk20a *g)
{
u32 tmp;
/* RMW registers need to be separate */
tmp = gk20a_readl(g, trim_sys_nvl_common_clk_alt_switch_r());
tmp &= ~trim_sys_nvl_common_clk_alt_switch_slowclk_m();
tmp |= trim_sys_nvl_common_clk_alt_switch_slowclk_xtal4x_f();
gk20a_writel(g, trim_sys_nvl_common_clk_alt_switch_r(), tmp);
}
static int gv100_nvlink_enable_links_pre_top(struct gk20a *g,
unsigned long links)
{
u32 link_id;
u32 tmp;
u32 reg;
u32 delay = ioctrl_reset_sw_post_reset_delay_microseconds_v();
int err;
unsigned long bit;
nvgpu_log(g, gpu_dbg_nvlink, " enabling 0x%lx links", links);
/* Take links out of reset */
for_each_set_bit(bit, &links, NVLINK_MAX_LINKS_SW) {
link_id = (u32)bit;
reg = IOCTRL_REG_RD32(g, ioctrl_reset_r());
tmp = (BIT32(link_id) |
BIT32(g->nvlink.links[link_id].pll_master_link_id));
reg = set_field(reg, ioctrl_reset_linkreset_m(),
ioctrl_reset_linkreset_f(ioctrl_reset_linkreset_v(reg) |
tmp));
IOCTRL_REG_WR32(g, ioctrl_reset_r(), reg);
nvgpu_udelay(delay);
reg = IOCTRL_REG_RD32(g, ioctrl_debug_reset_r());
reg &= ~ioctrl_debug_reset_link_f(BIT32(link_id));
IOCTRL_REG_WR32(g, ioctrl_debug_reset_r(), reg);
nvgpu_udelay(delay);
reg |= ioctrl_debug_reset_link_f(BIT32(link_id));
IOCTRL_REG_WR32(g, ioctrl_debug_reset_r(), reg);
nvgpu_udelay(delay);
/* Before doing any link initialization, run RXDET to check
* if link is connected on other end.
*/
if (g->ops.nvlink.rxdet != NULL) {
err = g->ops.nvlink.rxdet(g, link_id);
if (err != 0) {
return err;
}
}
/* Enable Link DLPL for AN0 */
reg = DLPL_REG_RD32(g, link_id, nvl_link_config_r());
reg = set_field(reg, nvl_link_config_link_en_m(),
nvl_link_config_link_en_f(1));
DLPL_REG_WR32(g, link_id, nvl_link_config_r(), reg);
/* This should be done by the NVLINK API */
err = g->ops.nvlink.link_mode_transitions.set_sublink_mode(g,
link_id, false, nvgpu_nvlink_sublink_tx_common);
if (err != 0) {
nvgpu_err(g, "Failed to init phy of link: %u", link_id);
return err;
}
err = g->ops.nvlink.link_mode_transitions.set_sublink_mode(g,
link_id, true, nvgpu_nvlink_sublink_rx_rxcal);
if (err != 0) {
nvgpu_err(g, "Failed to RXcal on link: %u", link_id);
return err;
}
err = g->ops.nvlink.link_mode_transitions.set_sublink_mode(g,
link_id, false, nvgpu_nvlink_sublink_tx_data_ready);
if (err != 0) {
nvgpu_err(g, "Failed to set data ready link:%u",
link_id);
return err;
}
g->nvlink.enabled_links |= BIT32(link_id);
}
nvgpu_log(g, gpu_dbg_nvlink, "enabled_links=0x%08x",
g->nvlink.enabled_links);
if (g->nvlink.enabled_links != 0U) {
return 0;
}
nvgpu_err(g, " No links were enabled");
return -EINVAL;
}
void gv100_nvlink_set_sw_war(struct gk20a *g, u32 link_id)
{
u32 reg;
/* WAR for HW bug 1888034 */
reg = DLPL_REG_RD32(g, link_id, nvl_sl0_safe_ctrl2_tx_r());
reg = set_field(reg, nvl_sl0_safe_ctrl2_tx_ctr_init_m(),
nvl_sl0_safe_ctrl2_tx_ctr_init_init_f());
reg = set_field(reg, nvl_sl0_safe_ctrl2_tx_ctr_initscl_m(),
nvl_sl0_safe_ctrl2_tx_ctr_initscl_init_f());
DLPL_REG_WR32(g, link_id, nvl_sl0_safe_ctrl2_tx_r(), reg);
}
static int gv100_nvlink_enable_links_post_top(struct gk20a *g,
unsigned long links)
{
u32 link_id;
unsigned long bit;
unsigned long enabled_links = (links & g->nvlink.enabled_links) &
~g->nvlink.initialized_links;
for_each_set_bit(bit, &enabled_links, NVLINK_MAX_LINKS_SW) {
link_id = (u32)bit;
if (g->ops.nvlink.set_sw_war != NULL) {
g->ops.nvlink.set_sw_war(g, link_id);
}
g->ops.nvlink.intr.init_nvlipt_intr(g, link_id);
g->ops.nvlink.intr.enable_link_intr(g, link_id, true);
g->nvlink.initialized_links |= BIT32(link_id);
};
return 0;
}
/*
*******************************************************************************
* Internal "ops" functions *
*******************************************************************************
*/
/*
* Main Nvlink init function. Calls into the Nvlink core API
*/
int gv100_nvlink_init(struct gk20a *g)
{
int err = 0;
if (!nvgpu_is_enabled(g, NVGPU_SUPPORT_NVLINK)) {
return -ENODEV;
}
err = nvgpu_nvlink_enumerate(g);
if (err != 0) {
nvgpu_err(g, "failed to enumerate nvlink");
goto fail;
}
/* Set HSHUB and SG_PHY */
nvgpu_set_enabled(g, NVGPU_MM_USE_PHYSICAL_SG, true);
err = g->ops.fb.enable_nvlink(g);
if (err != 0) {
nvgpu_err(g, "failed switch to nvlink sysmem");
goto fail;
}
return err;
fail:
nvgpu_set_enabled(g, NVGPU_MM_USE_PHYSICAL_SG, false);
nvgpu_set_enabled(g, NVGPU_SUPPORT_NVLINK, false);
return err;
}
/*
* Query internal device topology and discover devices in nvlink local
* infrastructure. Initialize register base and offsets
*/
int gv100_nvlink_discover_link(struct gk20a *g)
{
u32 i;
u32 link_id;
u32 ioctrl_entry_addr;
u32 ioctrl_device_type;
u32 table_entry;
u32 ioctrl_info_entry_type;
u32 ioctrl_discovery_size;
bool is_chain = false;
u8 nvlink_num_devices = 0U;
unsigned long available_links = 0UL;
struct nvgpu_nvlink_device_list *device_table;
int err = 0;
unsigned long bit;
/*
* Process Entry 0 & 1 of IOCTRL table to find table size
*/
if ((g->nvlink.ioctrl_table != NULL) &&
(g->nvlink.ioctrl_table[0].pri_base_addr != 0U)) {
ioctrl_entry_addr = g->nvlink.ioctrl_table[0].pri_base_addr;
table_entry = gk20a_readl(g, ioctrl_entry_addr);
ioctrl_info_entry_type = nvlinkip_discovery_common_device_v(table_entry);
} else {
nvgpu_err(g, " Bad IOCTRL PRI Base addr");
return -EINVAL;
}
if (ioctrl_info_entry_type == NVL_DEVICE(ioctrl)) {
ioctrl_entry_addr = g->nvlink.ioctrl_table[0].pri_base_addr + 4U;
table_entry = gk20a_readl(g, ioctrl_entry_addr);
ioctrl_discovery_size = nvlinkip_discovery_common_ioctrl_length_v(table_entry);
nvgpu_log(g, gpu_dbg_nvlink, "IOCTRL size: %d", ioctrl_discovery_size);
} else {
nvgpu_err(g, " First entry of IOCTRL_DISCOVERY invalid");
return -EINVAL;
}
device_table = nvgpu_kzalloc(g, ioctrl_discovery_size *
sizeof(struct nvgpu_nvlink_device_list));
if (device_table == NULL) {
nvgpu_err(g, " Unable to allocate nvlink device table");
return -ENOMEM;
}
for (i = 0U; i < ioctrl_discovery_size; i++) {
ioctrl_entry_addr =
g->nvlink.ioctrl_table[0].pri_base_addr + 4U*i;
table_entry = gk20a_readl(g, ioctrl_entry_addr);
nvgpu_log(g, gpu_dbg_nvlink, "parsing ioctrl %d: 0x%08x", i, table_entry);
ioctrl_info_entry_type = nvlinkip_discovery_common_entry_v(table_entry);
if (ioctrl_info_entry_type ==
nvlinkip_discovery_common_entry_invalid_v()) {
continue;
}
if (ioctrl_info_entry_type ==
nvlinkip_discovery_common_entry_enum_v()) {
nvgpu_log(g, gpu_dbg_nvlink, "IOCTRL entry %d is ENUM", i);
ioctrl_device_type =
nvlinkip_discovery_common_device_v(table_entry);
if (nvlinkip_discovery_common_chain_v(table_entry) !=
nvlinkip_discovery_common_chain_enable_v()) {
nvgpu_log(g, gpu_dbg_nvlink,
"IOCTRL entry %d is ENUM but no chain",
i);
err = -EINVAL;
break;
}
is_chain = true;
device_table[nvlink_num_devices].valid = true;
device_table[nvlink_num_devices].device_type =
(u8)ioctrl_device_type;
device_table[nvlink_num_devices].device_id =
(u8)nvlinkip_discovery_common_id_v(table_entry);
device_table[nvlink_num_devices].device_version =
nvlinkip_discovery_common_version_v(
table_entry);
continue;
}
if (ioctrl_info_entry_type ==
nvlinkip_discovery_common_entry_data1_v()) {
nvgpu_log(g, gpu_dbg_nvlink, "IOCTRL entry %d is DATA1", i);
if (is_chain) {
device_table[nvlink_num_devices].pri_base_addr =
nvlinkip_discovery_common_pri_base_v(
table_entry) << 12;
device_table[nvlink_num_devices].intr_enum =
(u8)nvlinkip_discovery_common_intr_v(
table_entry);
device_table[nvlink_num_devices].reset_enum =
(u8)nvlinkip_discovery_common_reset_v(
table_entry);
nvgpu_log(g, gpu_dbg_nvlink, "IOCTRL entry %d type = %d base: 0x%08x intr: %d reset: %d",
i,
device_table[nvlink_num_devices].device_type,
device_table[nvlink_num_devices].pri_base_addr,
device_table[nvlink_num_devices].intr_enum,
device_table[nvlink_num_devices].reset_enum);
if (device_table[nvlink_num_devices].device_type ==
NVL_DEVICE(dlpl)) {
device_table[nvlink_num_devices].num_tx =
(u8)nvlinkip_discovery_common_dlpl_num_tx_v(table_entry);
device_table[nvlink_num_devices].num_rx =
(u8)nvlinkip_discovery_common_dlpl_num_rx_v(table_entry);
nvgpu_log(g, gpu_dbg_nvlink, "DLPL tx: %d rx: %d",
device_table[nvlink_num_devices].num_tx,
device_table[nvlink_num_devices].num_rx);
}
if (nvlinkip_discovery_common_chain_v(table_entry) !=
nvlinkip_discovery_common_chain_enable_v()) {
is_chain = false;
nvlink_num_devices++;
}
}
continue;
}
if (ioctrl_info_entry_type ==
nvlinkip_discovery_common_entry_data2_v()) {
nvgpu_log(g, gpu_dbg_nvlink, "IOCTRL entry %d is DATA2", i);
if (is_chain) {
if (nvlinkip_discovery_common_dlpl_data2_type_v(table_entry) != 0U) {
device_table[nvlink_num_devices].pll_master =
(u8)nvlinkip_discovery_common_dlpl_data2_master_v(table_entry);
device_table[nvlink_num_devices].pll_master_id =
(u8)nvlinkip_discovery_common_dlpl_data2_masterid_v(table_entry);
nvgpu_log(g, gpu_dbg_nvlink, "PLL info: Master: %d, Master ID: %d",
device_table[nvlink_num_devices].pll_master,
device_table[nvlink_num_devices].pll_master_id);
}
if (nvlinkip_discovery_common_chain_v(table_entry) !=
nvlinkip_discovery_common_chain_enable_v()) {
is_chain = false;
nvlink_num_devices++;
}
}
continue;
}
}
g->nvlink.device_table = device_table;
g->nvlink.num_devices = nvlink_num_devices;
/*
* Print table
*/
for (i = 0; i < nvlink_num_devices; i++) {
if (device_table[i].valid) {
nvgpu_log(g, gpu_dbg_nvlink, "Device %d - %s", i,
gv100_device_type_to_str(
device_table[i].device_type));
nvgpu_log(g, gpu_dbg_nvlink, "+Link/Device Id: %d", device_table[i].device_id);
nvgpu_log(g, gpu_dbg_nvlink, "+Version: %d", device_table[i].device_version);
nvgpu_log(g, gpu_dbg_nvlink, "+Base Addr: 0x%08x", device_table[i].pri_base_addr);
nvgpu_log(g, gpu_dbg_nvlink, "+Intr Enum: %d", device_table[i].intr_enum);
nvgpu_log(g, gpu_dbg_nvlink, "+Reset Enum: %d", device_table[i].reset_enum);
if ((device_table[i].device_type == NVL_DEVICE(dlpl)) ||
(device_table[i].device_type == NVL_DEVICE(nvlink))) {
nvgpu_log(g, gpu_dbg_nvlink, "+TX: %d", device_table[i].num_tx);
nvgpu_log(g, gpu_dbg_nvlink, "+RX: %d", device_table[i].num_rx);
nvgpu_log(g, gpu_dbg_nvlink, "+PLL Master: %d", device_table[i].pll_master);
nvgpu_log(g, gpu_dbg_nvlink, "+PLL Master ID: %d", device_table[i].pll_master_id);
}
}
}
for (i = 0; i < nvlink_num_devices; i++) {
if (device_table[i].valid) {
if (device_table[i].device_type == NVL_DEVICE(ioctrl)) {
g->nvlink.ioctrl_type =
device_table[i].device_type;
g->nvlink.ioctrl_base =
device_table[i].pri_base_addr;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(dlpl)) {
g->nvlink.dlpl_type =
device_table[i].device_type;
g->nvlink.dlpl_base[device_table[i].device_id] =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].valid = true;
g->nvlink.links[device_table[i].device_id].g = g;
g->nvlink.links[device_table[i].device_id].dlpl_version =
device_table[i].device_version;
g->nvlink.links[device_table[i].device_id].dlpl_base =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].intr_enum =
device_table[i].intr_enum;
g->nvlink.links[device_table[i].device_id].reset_enum =
device_table[i].reset_enum;
g->nvlink.links[device_table[i].device_id].link_id =
device_table[i].device_id;
/* initiate the PLL master and slave link id to max */
g->nvlink.links[device_table[i].device_id].pll_master_link_id =
NVLINK_MAX_LINKS_SW;
g->nvlink.links[device_table[i].device_id].pll_slave_link_id =
NVLINK_MAX_LINKS_SW;
/* Update Pll master */
if (device_table[i].pll_master != 0U) {
g->nvlink.links[device_table[i].device_id].pll_master_link_id =
g->nvlink.links[device_table[i].device_id].link_id;
} else {
g->nvlink.links[device_table[i].device_id].pll_master_link_id =
device_table[i].pll_master_id;
g->nvlink.links[device_table[i].device_id].pll_slave_link_id =
g->nvlink.links[device_table[i].device_id].link_id;
g->nvlink.links[device_table[i].pll_master_id].pll_slave_link_id =
g->nvlink.links[device_table[i].device_id].link_id;
}
available_links |= BIT64(
device_table[i].device_id);
continue;
}
if (device_table[i].device_type == NVL_DEVICE(nvltlc)) {
g->nvlink.tl_type = device_table[i].device_type;
g->nvlink.tl_base[device_table[i].device_id] =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].tl_base =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].tl_version =
device_table[i].device_version;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(nvltlc)) {
g->nvlink.tl_type = device_table[i].device_type;
g->nvlink.tl_base[device_table[i].device_id] =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].tl_base =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].tl_version =
device_table[i].device_version;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(ioctrlmif)) {
g->nvlink.mif_type = device_table[i].device_type;
g->nvlink.mif_base[device_table[i].device_id] =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].mif_base =
device_table[i].pri_base_addr;
g->nvlink.links[device_table[i].device_id].mif_version =
device_table[i].device_version;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(nvlipt)) {
g->nvlink.ipt_type =
device_table[i].device_type;
g->nvlink.ipt_base =
device_table[i].pri_base_addr;
g->nvlink.ipt_version =
device_table[i].device_version;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(minion)) {
g->nvlink.minion_type =
device_table[i].device_type;
g->nvlink.minion_base =
device_table[i].pri_base_addr;
g->nvlink.minion_version =
device_table[i].device_version;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(dlpl_multicast)) {
g->nvlink.dlpl_multicast_type =
device_table[i].device_type;
g->nvlink.dlpl_multicast_base =
device_table[i].pri_base_addr;
g->nvlink.dlpl_multicast_version =
device_table[i].device_version;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(nvltlc_multicast)) {
g->nvlink.tl_multicast_type =
device_table[i].device_type;
g->nvlink.tl_multicast_base =
device_table[i].pri_base_addr;
g->nvlink.tl_multicast_version =
device_table[i].device_version;
continue;
}
if (device_table[i].device_type == NVL_DEVICE(ioctrlmif_multicast)) {
g->nvlink.mif_multicast_type =
device_table[i].device_type;
g->nvlink.mif_multicast_base =
device_table[i].pri_base_addr;
g->nvlink.mif_multicast_version =
device_table[i].device_version;
continue;
}
}
}
g->nvlink.discovered_links = (u32) available_links;
nvgpu_log(g, gpu_dbg_nvlink, "Nvlink Tree:");
nvgpu_log(g, gpu_dbg_nvlink, "+ Available Links: 0x%08lx", available_links);
nvgpu_log(g, gpu_dbg_nvlink, "+ Per-Link Devices:");
for_each_set_bit(bit, &available_links, NVLINK_MAX_LINKS_SW) {
link_id = (u32)bit;
nvgpu_log(g, gpu_dbg_nvlink, "-- Link %d Dl/Pl Base: 0x%08x TLC Base: 0x%08x MIF Base: 0x%08x",
link_id, g->nvlink.dlpl_base[link_id], g->nvlink.tl_base[link_id], g->nvlink.mif_base[link_id]);
}
nvgpu_log(g, gpu_dbg_nvlink, "+ IOCTRL Base: 0x%08x", g->nvlink.ioctrl_base);
nvgpu_log(g, gpu_dbg_nvlink, "+ NVLIPT Base: 0x%08x", g->nvlink.ipt_base);
nvgpu_log(g, gpu_dbg_nvlink, "+ MINION Base: 0x%08x", g->nvlink.minion_base);
nvgpu_log(g, gpu_dbg_nvlink, "+ DLPL MCAST Base: 0x%08x", g->nvlink.dlpl_multicast_base);
nvgpu_log(g, gpu_dbg_nvlink, "+ TLC MCAST Base: 0x%08x", g->nvlink.tl_multicast_base);
nvgpu_log(g, gpu_dbg_nvlink, "+ MIF MCAST Base: 0x%08x", g->nvlink.mif_multicast_base);
if (g->nvlink.minion_version == 0U) {
nvgpu_err(g, "Unsupported MINION version");
nvgpu_kfree(g, device_table);
g->nvlink.device_table = NULL;
g->nvlink.num_devices = 0;
return -EINVAL;
}
return err;
}
/*
* Query IOCTRL for device discovery
*/
int gv100_nvlink_discover_ioctrl(struct gk20a *g)
{
int ret = 0;
u32 i;
struct nvgpu_nvlink_ioctrl_list *ioctrl_table;
u32 ioctrl_num_entries = 0U;
if (g->ops.top.get_num_engine_type_entries != NULL) {
ioctrl_num_entries = g->ops.top.get_num_engine_type_entries(g,
NVGPU_ENGINE_IOCTRL);
nvgpu_log_info(g, "ioctrl_num_entries: %d", ioctrl_num_entries);
}
if (ioctrl_num_entries == 0U) {
nvgpu_err(g, "No NVLINK IOCTRL entry found in dev_info table");
return -EINVAL;
}
ioctrl_table = nvgpu_kzalloc(g, ioctrl_num_entries *
sizeof(struct nvgpu_nvlink_ioctrl_list));
if (ioctrl_table == NULL) {
nvgpu_err(g, "Failed to allocate memory for nvlink io table");
return -ENOMEM;
}
for (i = 0U; i < ioctrl_num_entries; i++) {
struct nvgpu_device_info dev_info;
ret = g->ops.top.get_device_info(g, &dev_info,
NVGPU_ENGINE_IOCTRL, i);
if (ret != 0) {
nvgpu_err(g, "Failed to parse dev_info table"
"for engine %d",
NVGPU_ENGINE_IOCTRL);
nvgpu_kfree(g, ioctrl_table);
return -EINVAL;
}
ioctrl_table[i].valid = true;
ioctrl_table[i].intr_enum = (u8)dev_info.intr_id;
ioctrl_table[i].reset_enum = (u8)dev_info.reset_id;
ioctrl_table[i].pri_base_addr = dev_info.pri_base;
nvgpu_log(g, gpu_dbg_nvlink,
"Dev %d: Pri_Base = 0x%0x Intr = %d Reset = %d",
i, ioctrl_table[i].pri_base_addr,
ioctrl_table[i].intr_enum,
ioctrl_table[i].reset_enum);
}
g->nvlink.ioctrl_table = ioctrl_table;
g->nvlink.io_num_entries = ioctrl_num_entries;
return 0;
}
/*
*******************************************************************************
* NVLINK API FUNCTIONS *
*******************************************************************************
*/
/*
* Performs link level initialization like phy inits, AN0 and interrupts
*/
int gv100_nvlink_link_early_init(struct gk20a *g, unsigned long mask)
{
int err;
err = gv100_nvlink_enable_links_pre_top(g, mask);
if (err != 0) {
nvgpu_err(g, "Pre topology failed for links 0x%lx", mask);
return err;
}
nvgpu_log(g, gpu_dbg_nvlink, "pretopology enabled: 0x%lx",
mask & g->nvlink.enabled_links);
err = gv100_nvlink_enable_links_post_top(g, mask);
return err;
}
/*
* Performs memory interface initialization
*/
int gv100_nvlink_interface_init(struct gk20a *g)
{
unsigned long mask = g->nvlink.enabled_links;
u32 link_id;
int err;
unsigned long bit;
for_each_set_bit(bit, &mask, NVLINK_MAX_LINKS_SW) {
link_id = (u32)bit;
g->ops.nvlink.intr.init_mif_intr(g, link_id);
g->ops.nvlink.intr.mif_intr_enable(g, link_id, true);
}
err = g->ops.fb.init_nvlink(g);
if (err != 0) {
nvgpu_err(g, "failed to setup nvlinks for sysmem");
return err;
}
return 0;
}
int gv100_nvlink_interface_disable(struct gk20a *g)
{
return 0;
}
/*
* Shutdown device. This should tear down Nvlink connection.
* For now return.
*/
int gv100_nvlink_shutdown(struct gk20a *g)
{
nvgpu_falcon_sw_free(g, FALCON_ID_MINION);
return 0;
}
/* Hardcode the link_mask while we wait for VBIOS link_disable_mask field
* to be updated.
*/
void gv100_nvlink_get_connected_link_mask(u32 *link_mask)
{
*link_mask = GV100_CONNECTED_LINK_MASK;
}
/*
* Performs nvlink device level initialization by discovering the topology
* taking device out of reset, boot minion, set clocks up and common interrupts
*/
int gv100_nvlink_early_init(struct gk20a *g)
{
int err = 0;
u32 mc_reset_nvlink_mask;
if (!nvgpu_is_enabled(g, NVGPU_SUPPORT_NVLINK)) {
return -EINVAL;
}
err = nvgpu_bios_get_lpwr_nvlink_table_hdr(g);
if (err != 0) {
nvgpu_err(g, "Failed to read LWPR_NVLINK_TABLE header\n");
goto exit;
}
err = nvgpu_bios_get_nvlink_config_data(g);
if (err != 0) {
nvgpu_err(g, "failed to read nvlink vbios data");
goto exit;
}
err = g->ops.nvlink.discover_ioctrl(g);
if (err != 0) {
goto exit;
}
/* Enable NVLINK in MC */
mc_reset_nvlink_mask = BIT32(g->nvlink.ioctrl_table[0].reset_enum);
nvgpu_log(g, gpu_dbg_nvlink, "mc_reset_nvlink_mask: 0x%x",
mc_reset_nvlink_mask);
g->ops.mc.reset(g, mc_reset_nvlink_mask);
err = g->ops.nvlink.discover_link(g);
if ((err != 0) || (g->nvlink.discovered_links == 0U)) {
nvgpu_err(g, "No links available");
goto exit;
}
err = nvgpu_falcon_sw_init(g, FALCON_ID_MINION);
if (err != 0) {
nvgpu_err(g, "failed to sw init FALCON_ID_MINION");
goto exit;
}
g->nvlink.discovered_links &= ~g->nvlink.link_disable_mask;
nvgpu_log(g, gpu_dbg_nvlink, "link_disable_mask = 0x%08x (from VBIOS)",
g->nvlink.link_disable_mask);
/* Links in reset should be removed from initialized link sw state */
g->nvlink.initialized_links &= g->ops.nvlink.get_link_reset_mask(g);
/* VBIOS link_disable_mask should be sufficient to find the connected
* links. As VBIOS is not updated with correct mask, we parse the DT
* node where we hardcode the link_id. DT method is not scalable as same
* DT node is used for different dGPUs connected over PCIE.
* Remove the DT parsing of link id and use HAL to get link_mask based
* on the GPU. This is temporary WAR while we get the VBIOS updated with
* correct mask.
*/
g->ops.nvlink.get_connected_link_mask(&(g->nvlink.connected_links));
nvgpu_log(g, gpu_dbg_nvlink, "connected_links = 0x%08x",
g->nvlink.connected_links);
/* Track only connected links */
g->nvlink.discovered_links &= g->nvlink.connected_links;
nvgpu_log(g, gpu_dbg_nvlink, "discovered_links = 0x%08x (combination)",
g->nvlink.discovered_links);
if (hweight32(g->nvlink.discovered_links) > 1U) {
nvgpu_err(g, "more than one link enabled");
err = -EINVAL;
goto nvlink_init_exit;
}
err = gv100_nvlink_state_load_hal(g);
if (err != 0) {
nvgpu_err(g, " failed Nvlink state load");
goto nvlink_init_exit;
}
err = gv100_nvlink_minion_configure_ac_coupling(g,
g->nvlink.ac_coupling_mask, true);
if (err != 0) {
nvgpu_err(g, " failed Nvlink state load");
goto nvlink_init_exit;
}
/* Program clocks */
gv100_nvlink_prog_alt_clk(g);
nvlink_init_exit:
nvgpu_falcon_sw_free(g, FALCON_ID_MINION);
exit:
return err;
}
int gv100_nvlink_speed_config(struct gk20a *g)
{
g->nvlink.speed = nvgpu_nvlink_speed_20G;
g->nvlink.initpll_ordinal = INITPLL_1;
g->nvlink.initpll_cmd = NVGPU_NVLINK_MINION_DLCMD_INITPLL_1;
return 0;
}
#endif /* CONFIG_TEGRA_NVLINK */
Reference in New Issue View Git Blame Copy Permalink