nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-23 01:50:07 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: refectored sim_readl/sim_writel

Browse Source 
refactored sim_readl and sim_writel to use
os-agnostic structures.

converted all sim buffers to the type nvgpu_mem and
replaced all alloc_page and free_page calls with
corresponding nvgpu_dma_alloc/nvgpu_dma_free calls.

JIRA VQRM-2368

Change-Id: Ia9d29119d31f239ed16be932cfd16c334002c727
Signed-off-by: Antony Clince Alex <aalex@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1702050
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: Sourab Gupta <sourabg@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
This commit is contained in:
Antony Clince Alex
2018-04-25 10:19:01 +05:30
committed by mobile promotions
parent 703d00d730
commit 2a23cd2494
3 changed files with 168 additions and 187 deletions
Download Patch File Download Diff File Expand all files Collapse all files

172
drivers/gpu/nvgpu/common/linux/sim.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
* *
* This program is free software; you can redistribute it and/or modify it * This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License, * under the terms and conditions of the GNU General Public License,
@@ -20,47 +20,67 @@
#include <nvgpu/log.h> #include <nvgpu/log.h>
#include <nvgpu/linux/vm.h> #include <nvgpu/linux/vm.h>
#include <nvgpu/bitops.h>
#include <nvgpu/nvgpu_mem.h>
#include <nvgpu/dma.h>
#include "gk20a/gk20a.h" #include "gk20a/gk20a.h"
#include "sim.h" #include "sim.h"
#include <nvgpu/hw/gk20a/hw_sim_gk20a.h> #include <nvgpu/hw/gk20a/hw_sim_gk20a.h>
static inline void sim_writel(struct sim_gk20a_linux *sim_linux, u32 r, u32 v) static inline void sim_writel(struct sim_gk20a *sim, u32 r, u32 v)
{ {
struct sim_gk20a_linux *sim_linux =
container_of(sim, struct sim_gk20a_linux, sim);
writel(v, sim_linux->regs + r); writel(v, sim_linux->regs + r);
} }
static inline u32 sim_readl(struct sim_gk20a_linux *sim_linux, u32 r) static inline u32 sim_readl(struct sim_gk20a *sim, u32 r)
{ {
struct sim_gk20a_linux *sim_linux =
container_of(sim, struct sim_gk20a_linux, sim);
return readl(sim_linux->regs + r); return readl(sim_linux->regs + r);
} }
static void kunmap_and_free_iopage(void **kvaddr, struct page **page) static int gk20a_alloc_sim_buffer(struct gk20a *g, struct nvgpu_mem *mem)
{ {
if (*kvaddr) { int err;
kunmap(*kvaddr);
*kvaddr = NULL; err = nvgpu_dma_alloc(g, PAGE_SIZE, mem);
if (err)
return err;
/*
* create a valid cpu_va mapping
*/
nvgpu_mem_begin(g, mem);
return 0;
} }
if (*page) {
__free_page(*page); static void gk20a_free_sim_buffer(struct gk20a *g, struct nvgpu_mem *mem)
*page = NULL; {
if (nvgpu_mem_is_valid(mem)) {
/*
* invalidate the cpu_va mapping
*/
nvgpu_mem_end(g, mem);
nvgpu_dma_free(g, mem);
} }
memset(mem, 0, sizeof(*mem));
} }
static void gk20a_free_sim_support(struct gk20a *g) static void gk20a_free_sim_support(struct gk20a *g)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
/* free sim mappings, bfrs */
kunmap_and_free_iopage(&sim_linux->send_bfr.kvaddr,
&sim_linux->send_bfr.page);
kunmap_and_free_iopage(&sim_linux->recv_bfr.kvaddr, gk20a_free_sim_buffer(g, &sim_linux->send_bfr);
&sim_linux->recv_bfr.page); gk20a_free_sim_buffer(g, &sim_linux->recv_bfr);
gk20a_free_sim_buffer(g, &sim_linux->msg_bfr);
kunmap_and_free_iopage(&sim_linux->msg_bfr.kvaddr,
&sim_linux->msg_bfr.page);
} }
static void gk20a_remove_sim_support(struct sim_gk20a *s) static void gk20a_remove_sim_support(struct sim_gk20a *s)
@@ -70,7 +90,7 @@ static void gk20a_remove_sim_support(struct sim_gk20a *s)
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
if (sim_linux->regs) if (sim_linux->regs)
sim_writel(sim_linux, sim_config_r(), sim_config_mode_disabled_v()); sim_writel(s, sim_config_r(), sim_config_mode_disabled_v());
gk20a_free_sim_support(g); gk20a_free_sim_support(g);
if (sim_linux->regs) { if (sim_linux->regs) {
@@ -82,35 +102,6 @@ static void gk20a_remove_sim_support(struct sim_gk20a *s)
g->sim = NULL; g->sim = NULL;
} }
static int alloc_and_kmap_iopage(struct gk20a *g,
void **kvaddr,
u64 *phys,
struct page **page)
{
int err = 0;
*page = alloc_page(GFP_KERNEL);
if (!*page) {
err = -ENOMEM;
nvgpu_err(g, "couldn't allocate io page");
goto fail;
}
*kvaddr = kmap(*page);
if (!*kvaddr) {
err = -ENOMEM;
nvgpu_err(g, "couldn't kmap io page");
goto fail;
}
*phys = page_to_phys(*page);
return 0;
fail:
kunmap_and_free_iopage(kvaddr, page);
return err;
}
static inline u32 sim_msg_header_size(void) static inline u32 sim_msg_header_size(void)
{ {
return 24;/*TBD: fix the header to gt this from NV_VGPU_MSG_HEADER*/ return 24;/*TBD: fix the header to gt this from NV_VGPU_MSG_HEADER*/
@@ -120,7 +111,11 @@ static inline u32 *sim_msg_bfr(struct gk20a *g, u32 byte_offset)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
return (u32 *)(sim_linux->msg_bfr.kvaddr + byte_offset); u8 *cpu_va;
cpu_va = (u8 *)sim_linux->msg_bfr.cpu_va;
return (u32 *)(cpu_va + byte_offset);
} }
static inline u32 *sim_msg_hdr(struct gk20a *g, u32 byte_offset) static inline u32 *sim_msg_hdr(struct gk20a *g, u32 byte_offset)
@@ -153,7 +148,11 @@ static u32 *sim_send_ring_bfr(struct gk20a *g, u32 byte_offset)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
return (u32 *)(sim_linux->send_bfr.kvaddr + byte_offset); u8 *cpu_va;
cpu_va = (u8 *)sim_linux->send_bfr.cpu_va;
return (u32 *)(cpu_va + byte_offset);
} }
static int rpc_send_message(struct gk20a *g) static int rpc_send_message(struct gk20a *g)
@@ -169,18 +168,17 @@ static int rpc_send_message(struct gk20a *g)
sim_dma_target_phys_pci_coherent_f() | sim_dma_target_phys_pci_coherent_f() |
sim_dma_status_valid_f() | sim_dma_status_valid_f() |
sim_dma_size_4kb_f() | sim_dma_size_4kb_f() |
sim_dma_addr_lo_f(sim_linux->msg_bfr.phys >> PAGE_SHIFT); sim_dma_addr_lo_f(nvgpu_mem_get_addr(g, &sim_linux->msg_bfr) >> PAGE_SHIFT);
*sim_send_ring_bfr(g, dma_hi_offset*sizeof(u32)) = *sim_send_ring_bfr(g, dma_hi_offset*sizeof(u32)) =
u64_hi32(sim_linux->msg_bfr.phys); u64_hi32(nvgpu_mem_get_addr(g, &sim_linux->msg_bfr));
*sim_msg_hdr(g, sim_msg_sequence_r()) = g->sim->sequence_base++; *sim_msg_hdr(g, sim_msg_sequence_r()) = g->sim->sequence_base++;
g->sim->send_ring_put = (g->sim->send_ring_put + 2 * sizeof(u32)) % g->sim->send_ring_put = (g->sim->send_ring_put + 2 * sizeof(u32)) % PAGE_SIZE;
PAGE_SIZE;
/* Update the put pointer. This will trap into the host. */ /* Update the put pointer. This will trap into the host. */
sim_writel(sim_linux, sim_send_put_r(), g->sim->send_ring_put); sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put);
return 0; return 0;
} }
@@ -189,7 +187,11 @@ static inline u32 *sim_recv_ring_bfr(struct gk20a *g, u32 byte_offset)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
return (u32 *)(sim_linux->recv_bfr.kvaddr + byte_offset); u8 *cpu_va;
cpu_va = (u8 *)sim_linux->recv_bfr.cpu_va;
return (u32 *)(cpu_va + byte_offset);
} }
static int rpc_recv_poll(struct gk20a *g) static int rpc_recv_poll(struct gk20a *g)
@@ -203,7 +205,7 @@ static int rpc_recv_poll(struct gk20a *g)
/* Poll the recv ring get pointer in an infinite loop*/ /* Poll the recv ring get pointer in an infinite loop*/
do { do {
g->sim->recv_ring_put = sim_readl(sim_linux, sim_recv_put_r()); g->sim->recv_ring_put = sim_readl(g->sim, sim_recv_put_r());
} while (g->sim->recv_ring_put == g->sim->recv_ring_get); } while (g->sim->recv_ring_put == g->sim->recv_ring_get);
/* process all replies */ /* process all replies */
@@ -220,19 +222,19 @@ static int rpc_recv_poll(struct gk20a *g)
recv_phys_addr = (u64)recv_phys_addr_hi << 32 | recv_phys_addr = (u64)recv_phys_addr_hi << 32 |
(u64)recv_phys_addr_lo << PAGE_SHIFT; (u64)recv_phys_addr_lo << PAGE_SHIFT;
if (recv_phys_addr != sim_linux->msg_bfr.phys) { if (recv_phys_addr !=
nvgpu_mem_get_addr(g, &sim_linux->msg_bfr)) {
nvgpu_err(g, "%s Error in RPC reply", nvgpu_err(g, "%s Error in RPC reply",
__func__); __func__);
return -1; return -1;
} }
/* Update GET pointer */ /* Update GET pointer */
g->sim->recv_ring_get = (g->sim->recv_ring_get + 2*sizeof(u32)) % g->sim->recv_ring_get = (g->sim->recv_ring_get + 2*sizeof(u32)) % PAGE_SIZE;
PAGE_SIZE;
sim_writel(sim_linux, sim_recv_get_r(), g->sim->recv_ring_get); sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get);
g->sim->recv_ring_put = sim_readl(sim_linux, sim_recv_put_r()); g->sim->recv_ring_put = sim_readl(g->sim, sim_recv_put_r());
} }
return 0; return 0;
@@ -295,53 +297,41 @@ int gk20a_init_sim_support(struct gk20a *g)
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
/* allocate sim event/msg buffers */ /* allocate sim event/msg buffers */
err = alloc_and_kmap_iopage(g, &sim_linux->send_bfr.kvaddr, err = gk20a_alloc_sim_buffer(g, &sim_linux->send_bfr);
&sim_linux->send_bfr.phys, err = err || gk20a_alloc_sim_buffer(g, &sim_linux->recv_bfr);
&sim_linux->send_bfr.page); err = err || gk20a_alloc_sim_buffer(g, &sim_linux->msg_bfr);
err = err || alloc_and_kmap_iopage(g, &sim_linux->recv_bfr.kvaddr, if (err)
&sim_linux->recv_bfr.phys,
&sim_linux->recv_bfr.page);
err = err || alloc_and_kmap_iopage(g, &sim_linux->msg_bfr.kvaddr,
&sim_linux->msg_bfr.phys,
&sim_linux->msg_bfr.page);
if (!(sim_linux->send_bfr.kvaddr && sim_linux->recv_bfr.kvaddr &&
sim_linux->msg_bfr.kvaddr)) {
nvgpu_err(g, "couldn't allocate all sim buffers");
goto fail; goto fail;
}
/*mark send ring invalid*/ /*mark send ring invalid*/
sim_writel(sim_linux, sim_send_ring_r(), sim_send_ring_status_invalid_f()); sim_writel(g->sim, sim_send_ring_r(), sim_send_ring_status_invalid_f());
/*read get pointer and make equal to put*/ /*read get pointer and make equal to put*/
g->sim->send_ring_put = sim_readl(sim_linux, sim_send_get_r()); g->sim->send_ring_put = sim_readl(g->sim, sim_send_get_r());
sim_writel(sim_linux, sim_send_put_r(), g->sim->send_ring_put); sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put);
/*write send ring address and make it valid*/ /*write send ring address and make it valid*/
phys = sim_linux->send_bfr.phys; phys = nvgpu_mem_get_addr(g, &sim_linux->send_bfr);
sim_writel(sim_linux, sim_send_ring_hi_r(), sim_writel(g->sim, sim_send_ring_hi_r(),
sim_send_ring_hi_addr_f(u64_hi32(phys))); sim_send_ring_hi_addr_f(u64_hi32(phys)));
sim_writel(sim_linux, sim_send_ring_r(), sim_writel(g->sim, sim_send_ring_r(),
sim_send_ring_status_valid_f() | sim_send_ring_status_valid_f() |
sim_send_ring_target_phys_pci_coherent_f() | sim_send_ring_target_phys_pci_coherent_f() |
sim_send_ring_size_4kb_f() | sim_send_ring_size_4kb_f() |
sim_send_ring_addr_lo_f(phys >> PAGE_SHIFT)); sim_send_ring_addr_lo_f(phys >> PAGE_SHIFT));
/*repeat for recv ring (but swap put,get as roles are opposite) */ /*repeat for recv ring (but swap put,get as roles are opposite) */
sim_writel(sim_linux, sim_recv_ring_r(), sim_recv_ring_status_invalid_f()); sim_writel(g->sim, sim_recv_ring_r(), sim_recv_ring_status_invalid_f());
/*read put pointer and make equal to get*/ /*read put pointer and make equal to get*/
g->sim->recv_ring_get = sim_readl(sim_linux, sim_recv_put_r()); g->sim->recv_ring_get = sim_readl(g->sim, sim_recv_put_r());
sim_writel(sim_linux, sim_recv_get_r(), g->sim->recv_ring_get); sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get);
/*write send ring address and make it valid*/ /*write send ring address and make it valid*/
phys = sim_linux->recv_bfr.phys; phys = nvgpu_mem_get_addr(g, &sim_linux->recv_bfr);
sim_writel(sim_linux, sim_recv_ring_hi_r(), sim_writel(g->sim, sim_recv_ring_hi_r(),
sim_recv_ring_hi_addr_f(u64_hi32(phys))); sim_recv_ring_hi_addr_f(u64_hi32(phys)));
sim_writel(sim_linux, sim_recv_ring_r(), sim_writel(g->sim, sim_recv_ring_r(),
sim_recv_ring_status_valid_f() | sim_recv_ring_status_valid_f() |
sim_recv_ring_target_phys_pci_coherent_f() | sim_recv_ring_target_phys_pci_coherent_f() |
sim_recv_ring_size_4kb_f() | sim_recv_ring_size_4kb_f() |

11
drivers/gpu/nvgpu/common/linux/sim.h
View File

@@ -2,7 +2,7 @@
* *
* GK20A sim support * GK20A sim support
* *
* Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
* *
* This program is free software; you can redistribute it and/or modify it * This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License, * under the terms and conditions of the GNU General Public License,
@@ -20,17 +20,16 @@
#ifndef __SIM_LINUX_H__ #ifndef __SIM_LINUX_H__
#define __SIM_LINUX_H__ #define __SIM_LINUX_H__
#include <nvgpu/nvgpu_mem.h>
#include "gk20a/sim_gk20a.h" #include "gk20a/sim_gk20a.h"
struct sim_gk20a_linux { struct sim_gk20a_linux {
struct sim_gk20a sim; struct sim_gk20a sim;
struct resource *reg_mem; struct resource *reg_mem;
void __iomem *regs; void __iomem *regs;
struct { struct nvgpu_mem send_bfr;
struct page *page; struct nvgpu_mem recv_bfr;
void *kvaddr; struct nvgpu_mem msg_bfr;
u64 phys;
} send_bfr, recv_bfr, msg_bfr;
}; };
int gk20a_init_sim_support(struct gk20a *g); int gk20a_init_sim_support(struct gk20a *g);

168
drivers/gpu/nvgpu/common/linux/sim_pci.c
View File

@@ -20,47 +20,67 @@
#include <nvgpu/log.h> #include <nvgpu/log.h>
#include <nvgpu/linux/vm.h> #include <nvgpu/linux/vm.h>
#include <nvgpu/bitops.h>
#include <nvgpu/nvgpu_mem.h>
#include <nvgpu/dma.h>
#include "gk20a/gk20a.h" #include "gk20a/gk20a.h"
#include "os_linux.h" #include "os_linux.h"
#include "sim.h" #include "sim.h"
#include "hw_sim_pci.h" #include "hw_sim_pci.h"
static inline void sim_writel(struct sim_gk20a_linux *sim_linux, u32 r, u32 v) static inline void sim_writel(struct sim_gk20a *sim, u32 r, u32 v)
{ {
struct sim_gk20a_linux *sim_linux =
container_of(sim, struct sim_gk20a_linux, sim);
writel(v, sim_linux->regs + r); writel(v, sim_linux->regs + r);
} }
static inline u32 sim_readl(struct sim_gk20a_linux *sim_linux, u32 r) static inline u32 sim_readl(struct sim_gk20a *sim, u32 r)
{ {
struct sim_gk20a_linux *sim_linux =
container_of(sim, struct sim_gk20a_linux, sim);
return readl(sim_linux->regs + r); return readl(sim_linux->regs + r);
} }
static void kunmap_and_free_iopage(void **kvaddr, struct page **page) static int gk20a_alloc_sim_buffer(struct gk20a *g, struct nvgpu_mem *mem)
{ {
if (*kvaddr) { int err;
kunmap(*kvaddr);
*kvaddr = NULL; err = nvgpu_dma_alloc(g, PAGE_SIZE, mem);
if (err)
return err;
/*
* create a valid cpu_va mapping
*/
nvgpu_mem_begin(g, mem);
return 0;
} }
if (*page) {
__free_page(*page); static void gk20a_free_sim_buffer(struct gk20a *g, struct nvgpu_mem *mem)
*page = NULL; {
if (nvgpu_mem_is_valid(mem)) {
/*
* invalidate the cpu_va mapping
*/
nvgpu_mem_end(g, mem);
nvgpu_dma_free(g, mem);
} }
memset(mem, 0, sizeof(*mem));
} }
static void gk20a_free_sim_support(struct gk20a *g) static void gk20a_free_sim_support(struct gk20a *g)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
/* free sim mappings, bfrs */
kunmap_and_free_iopage(&sim_linux->send_bfr.kvaddr,
&sim_linux->send_bfr.page);
kunmap_and_free_iopage(&sim_linux->recv_bfr.kvaddr, gk20a_free_sim_buffer(g, &sim_linux->send_bfr);
&sim_linux->recv_bfr.page); gk20a_free_sim_buffer(g, &sim_linux->recv_bfr);
gk20a_free_sim_buffer(g, &sim_linux->msg_bfr);
kunmap_and_free_iopage(&sim_linux->msg_bfr.kvaddr,
&sim_linux->msg_bfr.page);
} }
static void gk20a_remove_sim_support(struct sim_gk20a *s) static void gk20a_remove_sim_support(struct sim_gk20a *s)
@@ -70,7 +90,7 @@ static void gk20a_remove_sim_support(struct sim_gk20a *s)
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
if (sim_linux->regs) if (sim_linux->regs)
sim_writel(sim_linux, sim_config_r(), sim_config_mode_disabled_v()); sim_writel(s, sim_config_r(), sim_config_mode_disabled_v());
gk20a_free_sim_support(g); gk20a_free_sim_support(g);
if (sim_linux->regs) { if (sim_linux->regs) {
@@ -82,35 +102,6 @@ static void gk20a_remove_sim_support(struct sim_gk20a *s)
g->sim = NULL; g->sim = NULL;
} }
static int alloc_and_kmap_iopage(struct gk20a *g,
void **kvaddr,
u64 *phys,
struct page **page)
{
int err = 0;
*page = alloc_page(GFP_KERNEL);
if (!*page) {
err = -ENOMEM;
nvgpu_err(g, "couldn't allocate io page");
goto fail;
}
*kvaddr = kmap(*page);
if (!*kvaddr) {
err = -ENOMEM;
nvgpu_err(g, "couldn't kmap io page");
goto fail;
}
*phys = page_to_phys(*page);
return 0;
fail:
kunmap_and_free_iopage(kvaddr, page);
return err;
}
static inline u32 sim_msg_header_size(void) static inline u32 sim_msg_header_size(void)
{ {
return 32U; return 32U;
@@ -120,7 +111,11 @@ static inline u32 *sim_msg_bfr(struct gk20a *g, u32 byte_offset)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
return (u32 *)(sim_linux->msg_bfr.kvaddr + byte_offset); u8 *cpu_va;
cpu_va = (u8 *)sim_linux->msg_bfr.cpu_va;
return (u32 *)(cpu_va + byte_offset);
} }
static inline u32 *sim_msg_hdr(struct gk20a *g, u32 byte_offset) static inline u32 *sim_msg_hdr(struct gk20a *g, u32 byte_offset)
@@ -155,7 +150,11 @@ static u32 *sim_send_ring_bfr(struct gk20a *g, u32 byte_offset)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
return (u32 *)(sim_linux->send_bfr.kvaddr + byte_offset); u8 *cpu_va;
cpu_va = (u8 *)sim_linux->send_bfr.cpu_va;
return (u32 *)(cpu_va + byte_offset);
} }
static int rpc_send_message(struct gk20a *g) static int rpc_send_message(struct gk20a *g)
@@ -171,10 +170,10 @@ static int rpc_send_message(struct gk20a *g)
sim_dma_target_phys_pci_coherent_f() | sim_dma_target_phys_pci_coherent_f() |
sim_dma_status_valid_f() | sim_dma_status_valid_f() |
sim_dma_size_4kb_f() | sim_dma_size_4kb_f() |
sim_dma_addr_lo_f(sim_linux->msg_bfr.phys >> PAGE_SHIFT); sim_dma_addr_lo_f(nvgpu_mem_get_addr(g, &sim_linux->msg_bfr) >> PAGE_SHIFT);
*sim_send_ring_bfr(g, dma_hi_offset*sizeof(u32)) = *sim_send_ring_bfr(g, dma_hi_offset*sizeof(u32)) =
u64_hi32(sim_linux->msg_bfr.phys); u64_hi32(nvgpu_mem_get_addr(g, &sim_linux->msg_bfr));
*sim_msg_hdr(g, sim_msg_sequence_r()) = g->sim->sequence_base++; *sim_msg_hdr(g, sim_msg_sequence_r()) = g->sim->sequence_base++;
@@ -182,7 +181,7 @@ static int rpc_send_message(struct gk20a *g)
PAGE_SIZE; PAGE_SIZE;
/* Update the put pointer. This will trap into the host. */ /* Update the put pointer. This will trap into the host. */
sim_writel(sim_linux, sim_send_put_r(), g->sim->send_ring_put); sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put);
return 0; return 0;
} }
@@ -191,7 +190,11 @@ static inline u32 *sim_recv_ring_bfr(struct gk20a *g, u32 byte_offset)
{ {
struct sim_gk20a_linux *sim_linux = struct sim_gk20a_linux *sim_linux =
container_of(g->sim, struct sim_gk20a_linux, sim); container_of(g->sim, struct sim_gk20a_linux, sim);
return (u32 *)(sim_linux->recv_bfr.kvaddr + byte_offset); u8 *cpu_va;
cpu_va = (u8 *)sim_linux->recv_bfr.cpu_va;
return (u32 *)(cpu_va + byte_offset);
} }
static int rpc_recv_poll(struct gk20a *g) static int rpc_recv_poll(struct gk20a *g)
@@ -202,7 +205,7 @@ static int rpc_recv_poll(struct gk20a *g)
/* Poll the recv ring get pointer in an infinite loop */ /* Poll the recv ring get pointer in an infinite loop */
do { do {
g->sim->recv_ring_put = sim_readl(sim_linux, sim_recv_put_r()); g->sim->recv_ring_put = sim_readl(g->sim, sim_recv_put_r());
} while (g->sim->recv_ring_put == g->sim->recv_ring_get); } while (g->sim->recv_ring_put == g->sim->recv_ring_get);
/* process all replies */ /* process all replies */
@@ -219,18 +222,19 @@ static int rpc_recv_poll(struct gk20a *g)
recv_phys_addr = (u64)recv_phys_addr_hi << 32 | recv_phys_addr = (u64)recv_phys_addr_hi << 32 |
(u64)recv_phys_addr_lo << PAGE_SHIFT; (u64)recv_phys_addr_lo << PAGE_SHIFT;
if (recv_phys_addr != sim_linux->msg_bfr.phys) { if (recv_phys_addr !=
nvgpu_mem_get_addr(g, &sim_linux->msg_bfr)) {
nvgpu_err(g, "Error in RPC reply"); nvgpu_err(g, "Error in RPC reply");
return -EINVAL; return -EINVAL;
} }
/* Update GET pointer */ /* Update GET pointer */
g->sim->recv_ring_get = (g->sim->recv_ring_get + 2*sizeof(u32)) % g->sim->recv_ring_get = (g->sim->recv_ring_get + 2*sizeof(u32))
PAGE_SIZE; % PAGE_SIZE;
sim_writel(sim_linux, sim_recv_get_r(), g->sim->recv_ring_get); sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get);
g->sim->recv_ring_put = sim_readl(sim_linux, sim_recv_put_r()); g->sim->recv_ring_put = sim_readl(g->sim, sim_recv_put_r());
} }
return 0; return 0;
@@ -316,53 +320,41 @@ int nvgpu_pci_init_sim_support(struct gk20a *g)
sim_linux->regs = l->regs + sim_r(); sim_linux->regs = l->regs + sim_r();
/* allocate sim event/msg buffers */ /* allocate sim event/msg buffers */
err = alloc_and_kmap_iopage(g, &sim_linux->send_bfr.kvaddr, err = gk20a_alloc_sim_buffer(g, &sim_linux->send_bfr);
&sim_linux->send_bfr.phys, err = err || gk20a_alloc_sim_buffer(g, &sim_linux->recv_bfr);
&sim_linux->send_bfr.page); err = err || gk20a_alloc_sim_buffer(g, &sim_linux->msg_bfr);
err = err || alloc_and_kmap_iopage(g, &sim_linux->recv_bfr.kvaddr, if (err)
&sim_linux->recv_bfr.phys,
&sim_linux->recv_bfr.page);
err = err || alloc_and_kmap_iopage(g, &sim_linux->msg_bfr.kvaddr,
&sim_linux->msg_bfr.phys,
&sim_linux->msg_bfr.page);
if (!(sim_linux->send_bfr.kvaddr && sim_linux->recv_bfr.kvaddr &&
sim_linux->msg_bfr.kvaddr)) {
nvgpu_err(g, "couldn't allocate all sim buffers");
goto fail; goto fail;
}
/* mark send ring invalid */ /* mark send ring invalid */
sim_writel(sim_linux, sim_send_ring_r(), sim_send_ring_status_invalid_f()); sim_writel(g->sim, sim_send_ring_r(), sim_send_ring_status_invalid_f());
/* read get pointer and make equal to put */ /* read get pointer and make equal to put */
g->sim->send_ring_put = sim_readl(sim_linux, sim_send_get_r()); g->sim->send_ring_put = sim_readl(g->sim, sim_send_get_r());
sim_writel(sim_linux, sim_send_put_r(), g->sim->send_ring_put); sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put);
/* write send ring address and make it valid */ /* write send ring address and make it valid */
phys = sim_linux->send_bfr.phys; phys = nvgpu_mem_get_addr(g, &sim_linux->send_bfr);
sim_writel(sim_linux, sim_send_ring_hi_r(), sim_writel(g->sim, sim_send_ring_hi_r(),
sim_send_ring_hi_addr_f(u64_hi32(phys))); sim_send_ring_hi_addr_f(u64_hi32(phys)));
sim_writel(sim_linux, sim_send_ring_r(), sim_writel(g->sim, sim_send_ring_r(),
sim_send_ring_status_valid_f() | sim_send_ring_status_valid_f() |
sim_send_ring_target_phys_pci_coherent_f() | sim_send_ring_target_phys_pci_coherent_f() |
sim_send_ring_size_4kb_f() | sim_send_ring_size_4kb_f() |
sim_send_ring_addr_lo_f(phys >> PAGE_SHIFT)); sim_send_ring_addr_lo_f(phys >> PAGE_SHIFT));
/* repeat for recv ring (but swap put,get as roles are opposite) */ /* repeat for recv ring (but swap put,get as roles are opposite) */
sim_writel(sim_linux, sim_recv_ring_r(), sim_recv_ring_status_invalid_f()); sim_writel(g->sim, sim_recv_ring_r(), sim_recv_ring_status_invalid_f());
/* read put pointer and make equal to get */ /* read put pointer and make equal to get */
g->sim->recv_ring_get = sim_readl(sim_linux, sim_recv_put_r()); g->sim->recv_ring_get = sim_readl(g->sim, sim_recv_put_r());
sim_writel(sim_linux, sim_recv_get_r(), g->sim->recv_ring_get); sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get);
/* write send ring address and make it valid */ /* write send ring address and make it valid */
phys = sim_linux->recv_bfr.phys; phys = nvgpu_mem_get_addr(g, &sim_linux->recv_bfr);
sim_writel(sim_linux, sim_recv_ring_hi_r(), sim_writel(g->sim, sim_recv_ring_hi_r(),
sim_recv_ring_hi_addr_f(u64_hi32(phys))); sim_recv_ring_hi_addr_f(u64_hi32(phys)));
sim_writel(sim_linux, sim_recv_ring_r(), sim_writel(g->sim, sim_recv_ring_r(),
sim_recv_ring_status_valid_f() | sim_recv_ring_status_valid_f() |
sim_recv_ring_target_phys_pci_coherent_f() | sim_recv_ring_target_phys_pci_coherent_f() |
sim_recv_ring_size_4kb_f() | sim_recv_ring_size_4kb_f() |