nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-22 17:36:20 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: sim: make ring buffer independent of PAGE_SIZE

Browse Source 
The simulator ring buffer DMA interface supports buffers of the following sizes:
4, 8, 12 and 16K. At present, it is configured to 4K and it  happens to match
with the kernel PAGE_SIZE, which is used to wrap back the GET/PUT pointers once
4K is reached. However, this is not always true; for instance, take 64K pages.
Hence, replace PAGE_SIZE with SIM_BFR_SIZE.

Introduce macro NVGPU_CPU_PAGE_SIZE which aliases to PAGE_SIZE and replace
latter with former.

Bug 200658101
Jira NVGPU-6018

Change-Id: I83cc62b87291734015c51f3e5a98173549e065de
Signed-off-by: Antony Clince Alex <aalex@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2420728
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
This commit is contained in:
Antony Clince Alex
2020-09-28 16:44:40 +05:30
committed by Alex Waterman
parent 09857ecd91
commit c36752fe3d
37 changed files with 120 additions and 103 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
drivers/gpu/nvgpu/common/fifo/userd.c
View File

@@ -37,7 +37,7 @@ int nvgpu_userd_init_slabs(struct gk20a *g)
nvgpu_mutex_init(&f->userd_mutex); nvgpu_mutex_init(&f->userd_mutex);
f->num_channels_per_slab = PAGE_SIZE / g->ops.userd.entry_size(g); f->num_channels_per_slab = NVGPU_CPU_PAGE_SIZE / g->ops.userd.entry_size(g);
f->num_userd_slabs = f->num_userd_slabs =
DIV_ROUND_UP(f->num_channels, f->num_channels_per_slab); DIV_ROUND_UP(f->num_channels, f->num_channels_per_slab);
@@ -88,7 +88,7 @@ int nvgpu_userd_init_channel(struct gk20a *g, struct nvgpu_channel *c)
nvgpu_mutex_acquire(&f->userd_mutex); nvgpu_mutex_acquire(&f->userd_mutex);
if (!nvgpu_mem_is_valid(mem)) { if (!nvgpu_mem_is_valid(mem)) {
err = nvgpu_dma_alloc_sys(g, PAGE_SIZE, mem); err = nvgpu_dma_alloc_sys(g, NVGPU_CPU_PAGE_SIZE, mem);
if (err != 0) { if (err != 0) {
nvgpu_err(g, "userd allocation failed, err=%d", err); nvgpu_err(g, "userd allocation failed, err=%d", err);
goto done; goto done;
@@ -96,7 +96,7 @@ int nvgpu_userd_init_channel(struct gk20a *g, struct nvgpu_channel *c)
if (g->ops.mm.is_bar1_supported(g)) { if (g->ops.mm.is_bar1_supported(g)) {
mem->gpu_va = g->ops.mm.bar1_map_userd(g, mem, mem->gpu_va = g->ops.mm.bar1_map_userd(g, mem,
slab * PAGE_SIZE); slab * NVGPU_CPU_PAGE_SIZE);
} }
} }
c->userd_mem = mem; c->userd_mem = mem;
@@ -128,9 +128,9 @@ int nvgpu_userd_setup_sw(struct gk20a *g)
} }
size = f->num_channels * g->ops.userd.entry_size(g); size = f->num_channels * g->ops.userd.entry_size(g);
num_pages = DIV_ROUND_UP(size, PAGE_SIZE); num_pages = DIV_ROUND_UP(size, NVGPU_CPU_PAGE_SIZE);
err = nvgpu_vm_area_alloc(g->mm.bar1.vm, err = nvgpu_vm_area_alloc(g->mm.bar1.vm,
num_pages, PAGE_SIZE, &f->userd_gpu_va, 0); num_pages, NVGPU_CPU_PAGE_SIZE, &f->userd_gpu_va, 0);
if (err != 0) { if (err != 0) {
nvgpu_err(g, "userd gpu va allocation failed, err=%d", err); nvgpu_err(g, "userd gpu va allocation failed, err=%d", err);
goto clean_up; goto clean_up;

4
drivers/gpu/nvgpu/common/gr/gr.c
View File

@@ -328,8 +328,8 @@ static int gr_init_access_map(struct gk20a *g, struct nvgpu_gr *gr)
struct nvgpu_mem *mem; struct nvgpu_mem *mem;
u32 nr_pages = u32 nr_pages =
DIV_ROUND_UP(NVGPU_GR_GLOBAL_CTX_PRIV_ACCESS_MAP_SIZE, DIV_ROUND_UP(NVGPU_GR_GLOBAL_CTX_PRIV_ACCESS_MAP_SIZE,
PAGE_SIZE); NVGPU_CPU_PAGE_SIZE);
u32 nr_pages_size = nvgpu_safe_mult_u32(PAGE_SIZE, nr_pages); u32 nr_pages_size = nvgpu_safe_mult_u32(NVGPU_CPU_PAGE_SIZE, nr_pages);
#ifdef CONFIG_NVGPU_SET_FALCON_ACCESS_MAP #ifdef CONFIG_NVGPU_SET_FALCON_ACCESS_MAP
u32 *whitelist = NULL; u32 *whitelist = NULL;
u32 w, num_entries = 0U; u32 w, num_entries = 0U;

2
drivers/gpu/nvgpu/common/init/nvgpu_init.c
View File

@@ -529,7 +529,7 @@ static int nvgpu_init_syncpt_mem(struct gk20a *g)
if (nvgpu_has_syncpoints(g) && (g->syncpt_unit_size != 0UL)) { if (nvgpu_has_syncpoints(g) && (g->syncpt_unit_size != 0UL)) {
if (!nvgpu_mem_is_valid(&g->syncpt_mem)) { if (!nvgpu_mem_is_valid(&g->syncpt_mem)) {
nr_pages = U64(DIV_ROUND_UP(g->syncpt_unit_size, nr_pages = U64(DIV_ROUND_UP(g->syncpt_unit_size,
PAGE_SIZE)); NVGPU_CPU_PAGE_SIZE));
err = nvgpu_mem_create_from_phys(g, &g->syncpt_mem, err = nvgpu_mem_create_from_phys(g, &g->syncpt_mem,
g->syncpt_unit_base, nr_pages); g->syncpt_unit_base, nr_pages);
if (err != 0) { if (err != 0) {

4
drivers/gpu/nvgpu/common/mm/gmmu/page_table.c
View File

@@ -195,14 +195,14 @@ int nvgpu_gmmu_init_page_table(struct vm_gk20a *vm)
* aligned. Although lower PDE tables can be aligned at 256B boundaries * aligned. Although lower PDE tables can be aligned at 256B boundaries
* the PDB must be 4K aligned. * the PDB must be 4K aligned.
* *
* Currently PAGE_SIZE is used, even when 64K, to work around an issue * Currently NVGPU_CPU_PAGE_SIZE is used, even when 64K, to work around an issue
* with the PDB TLB invalidate code not being pd_cache aware yet. * with the PDB TLB invalidate code not being pd_cache aware yet.
* *
* Similarly, we can't use nvgpu_pd_alloc() here, because the top-level * Similarly, we can't use nvgpu_pd_alloc() here, because the top-level
* PD must have mem_offs be 0 for the invalidate code to work, so we * PD must have mem_offs be 0 for the invalidate code to work, so we
* can't use the PD cache. * can't use the PD cache.
*/ */
pdb_size = ALIGN(pd_get_size(&vm->mmu_levels[0], &attrs), PAGE_SIZE); pdb_size = ALIGN(pd_get_size(&vm->mmu_levels[0], &attrs), NVGPU_CPU_PAGE_SIZE);
err = nvgpu_pd_cache_alloc_direct(vm->mm->g, &vm->pdb, pdb_size); err = nvgpu_pd_cache_alloc_direct(vm->mm->g, &vm->pdb, pdb_size);
if (err != 0) { if (err != 0) {

6
drivers/gpu/nvgpu/common/mm/gmmu/pd_cache.c
View File

@@ -171,7 +171,7 @@ int nvgpu_pd_cache_alloc_direct(struct gk20a *g,
} }
/* /*
* If bytes == PAGE_SIZE then it's impossible to get a discontiguous DMA * If bytes == NVGPU_CPU_PAGE_SIZE then it's impossible to get a discontiguous DMA
* allocation. Some DMA implementations may, despite this fact, still * allocation. Some DMA implementations may, despite this fact, still
* use the contiguous pool for page sized allocations. As such only * use the contiguous pool for page sized allocations. As such only
* request explicitly contiguous allocs if the page directory is larger * request explicitly contiguous allocs if the page directory is larger
@@ -180,7 +180,7 @@ int nvgpu_pd_cache_alloc_direct(struct gk20a *g,
* going to be virtually contiguous and we don't have to force the * going to be virtually contiguous and we don't have to force the
* underlying allocations to be physically contiguous as well. * underlying allocations to be physically contiguous as well.
*/ */
if (!nvgpu_iommuable(g) && (bytes > PAGE_SIZE)) { if (!nvgpu_iommuable(g) && (bytes > NVGPU_CPU_PAGE_SIZE)) {
flags = NVGPU_DMA_PHYSICALLY_ADDRESSED; flags = NVGPU_DMA_PHYSICALLY_ADDRESSED;
} }
@@ -218,7 +218,7 @@ static int nvgpu_pd_cache_alloc_new(struct gk20a *g,
return -ENOMEM; return -ENOMEM;
} }
if (!nvgpu_iommuable(g) && (NVGPU_PD_CACHE_SIZE > PAGE_SIZE)) { if (!nvgpu_iommuable(g) && (NVGPU_PD_CACHE_SIZE > NVGPU_CPU_PAGE_SIZE)) {
flags = NVGPU_DMA_PHYSICALLY_ADDRESSED; flags = NVGPU_DMA_PHYSICALLY_ADDRESSED;
} }

2
drivers/gpu/nvgpu/common/mm/gmmu/pd_cache_priv.h
View File

@@ -99,7 +99,7 @@
* PD cache size to be 64K if PAGE_SIZE > 4K (i.e PAGE_SIZE == 64K). * PD cache size to be 64K if PAGE_SIZE > 4K (i.e PAGE_SIZE == 64K).
*/ */
#ifdef __KERNEL__ #ifdef __KERNEL__
# if PAGE_SIZE > 4096 # if NVGPU_CPU_PAGE_SIZE > 4096
# define NVGPU_PD_CACHE_COUNT 8UL # define NVGPU_PD_CACHE_COUNT 8UL
# else # else
# define NVGPU_PD_CACHE_COUNT 4UL # define NVGPU_PD_CACHE_COUNT 4UL

3
drivers/gpu/nvgpu/common/mm/nvgpu_mem.c
View File

@@ -402,7 +402,8 @@ int nvgpu_mem_create_from_phys(struct gk20a *g, struct nvgpu_mem *dest,
(void) memset(dest, 0, sizeof(*dest)); (void) memset(dest, 0, sizeof(*dest));
dest->aperture = APERTURE_SYSMEM; dest->aperture = APERTURE_SYSMEM;
dest->size = nvgpu_safe_mult_u64(nr_pages, (u64)SZ_4K); dest->size = nvgpu_safe_mult_u64(nr_pages,
(u64)NVGPU_CPU_PAGE_SIZE);
dest->aligned_size = dest->size; dest->aligned_size = dest->size;
dest->mem_flags = NVGPU_MEM_FLAG_NO_DMA; dest->mem_flags = NVGPU_MEM_FLAG_NO_DMA;
dest->phys_sgt = sgt; dest->phys_sgt = sgt;

4
drivers/gpu/nvgpu/common/mm/vm.c
View File

@@ -373,7 +373,7 @@ static int nvgpu_init_sema_pool(struct vm_gk20a *vm)
nvgpu_semaphore_sea_allocate_gpu_va(sema_sea, &vm->kernel, nvgpu_semaphore_sea_allocate_gpu_va(sema_sea, &vm->kernel,
nvgpu_safe_sub_u64(vm->va_limit, nvgpu_safe_sub_u64(vm->va_limit,
mm->channel.kernel_size), mm->channel.kernel_size),
512U * PAGE_SIZE, 512U * NVGPU_CPU_PAGE_SIZE,
nvgpu_safe_cast_u64_to_u32(SZ_4K)); nvgpu_safe_cast_u64_to_u32(SZ_4K));
if (nvgpu_semaphore_sea_get_gpu_va(sema_sea) == 0ULL) { if (nvgpu_semaphore_sea_get_gpu_va(sema_sea) == 0ULL) {
nvgpu_free(&vm->kernel, nvgpu_free(&vm->kernel,
@@ -741,7 +741,7 @@ static int nvgpu_vm_init_attributes(struct mm_gk20a *mm,
nvgpu_safe_cast_u64_to_u32(SZ_4K); nvgpu_safe_cast_u64_to_u32(SZ_4K);
vm->gmmu_page_sizes[GMMU_PAGE_SIZE_BIG] = big_page_size; vm->gmmu_page_sizes[GMMU_PAGE_SIZE_BIG] = big_page_size;
vm->gmmu_page_sizes[GMMU_PAGE_SIZE_KERNEL] = vm->gmmu_page_sizes[GMMU_PAGE_SIZE_KERNEL] =
nvgpu_safe_cast_u64_to_u32(PAGE_SIZE); nvgpu_safe_cast_u64_to_u32(NVGPU_CPU_PAGE_SIZE);
/* Set up vma pointers. */ /* Set up vma pointers. */
vm->vma[GMMU_PAGE_SIZE_SMALL] = &vm->user; vm->vma[GMMU_PAGE_SIZE_SMALL] = &vm->user;

2
drivers/gpu/nvgpu/common/semaphore/semaphore_hw.c
View File

@@ -46,7 +46,7 @@ int nvgpu_hw_semaphore_init(struct vm_gk20a *vm, u32 chid,
/* Find an available HW semaphore. */ /* Find an available HW semaphore. */
hw_sema_idx = semaphore_bitmap_alloc(p->semas_alloced, hw_sema_idx = semaphore_bitmap_alloc(p->semas_alloced,
PAGE_SIZE / SEMAPHORE_SIZE); NVGPU_CPU_PAGE_SIZE / SEMAPHORE_SIZE);
if (hw_sema_idx < 0) { if (hw_sema_idx < 0) {
ret = hw_sema_idx; ret = hw_sema_idx;
goto fail; goto fail;

4
drivers/gpu/nvgpu/common/semaphore/semaphore_pool.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2014-2019, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2014-2020, NVIDIA CORPORATION. All rights reserved.
* *
* Permission is hereby granted, free of charge, to any person obtaining a * Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"), * copy of this software and associated documentation files (the "Software"),
@@ -244,7 +244,7 @@ u64 nvgpu_semaphore_pool_gpu_va(struct nvgpu_semaphore_pool *p, bool global)
return p->gpu_va; return p->gpu_va;
} }
return p->gpu_va_ro + (PAGE_SIZE * p->page_idx); return p->gpu_va_ro + (NVGPU_CPU_PAGE_SIZE * p->page_idx);
} }
/* /*

2
drivers/gpu/nvgpu/common/semaphore/semaphore_priv.h
View File

@@ -93,7 +93,7 @@ struct nvgpu_semaphore_pool {
u64 gpu_va_ro; /* GPU access to the pool. */ u64 gpu_va_ro; /* GPU access to the pool. */
u64 page_idx; /* Index into sea bitmap. */ u64 page_idx; /* Index into sea bitmap. */
DECLARE_BITMAP(semas_alloced, PAGE_SIZE / SEMAPHORE_SIZE); DECLARE_BITMAP(semas_alloced, NVGPU_CPU_PAGE_SIZE / SEMAPHORE_SIZE);
struct nvgpu_semaphore_sea *sema_sea; /* Sea that owns this pool. */ struct nvgpu_semaphore_sea *sema_sea; /* Sea that owns this pool. */

8
drivers/gpu/nvgpu/common/semaphore/semaphore_sea.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2014-2019, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2014-2020, NVIDIA CORPORATION. All rights reserved.
* *
* Permission is hereby granted, free of charge, to any person obtaining a * Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"), * copy of this software and associated documentation files (the "Software"),
@@ -50,21 +50,21 @@ static int semaphore_sea_grow(struct nvgpu_semaphore_sea *sea)
nvgpu_semaphore_sea_lock(sea); nvgpu_semaphore_sea_lock(sea);
ret = nvgpu_dma_alloc_sys(g, ret = nvgpu_dma_alloc_sys(g,
PAGE_SIZE * SEMAPHORE_POOL_COUNT, NVGPU_CPU_PAGE_SIZE * SEMAPHORE_POOL_COUNT,
&sea->sea_mem); &sea->sea_mem);
if (ret != 0) { if (ret != 0) {
goto out; goto out;
} }
sea->size = SEMAPHORE_POOL_COUNT; sea->size = SEMAPHORE_POOL_COUNT;
sea->map_size = SEMAPHORE_POOL_COUNT * PAGE_SIZE; sea->map_size = SEMAPHORE_POOL_COUNT * NVGPU_CPU_PAGE_SIZE;
/* /*
* Start the semaphores at values that will soon overflow the 32-bit * Start the semaphores at values that will soon overflow the 32-bit
* integer range. This way any buggy comparisons would start to fail * integer range. This way any buggy comparisons would start to fail
* sooner rather than later. * sooner rather than later.
*/ */
for (i = 0U; i < PAGE_SIZE * SEMAPHORE_POOL_COUNT; i += 4U) { for (i = 0U; i < NVGPU_CPU_PAGE_SIZE * SEMAPHORE_POOL_COUNT; i += 4U) {
nvgpu_mem_wr(g, &sea->sea_mem, i, 0xfffffff0U); nvgpu_mem_wr(g, &sea->sea_mem, i, 0xfffffff0U);
} }

8
drivers/gpu/nvgpu/common/sim/sim.c
View File

@@ -36,7 +36,7 @@ int nvgpu_alloc_sim_buffer(struct gk20a *g, struct nvgpu_mem *mem)
int err = 0; int err = 0;
if (!nvgpu_mem_is_valid(mem)) { if (!nvgpu_mem_is_valid(mem)) {
err = nvgpu_dma_alloc_sys(g, PAGE_SIZE, mem); err = nvgpu_dma_alloc_sys(g, NVGPU_CPU_PAGE_SIZE, mem);
} }
return err; return err;
@@ -67,7 +67,7 @@ void nvgpu_remove_sim_support(struct gk20a *g)
void sim_write_hdr(struct gk20a *g, u32 func, u32 size) void sim_write_hdr(struct gk20a *g, u32 func, u32 size)
{ {
/*memset(g->sim->msg_bfr.kvaddr,0,min(PAGE_SIZE,size));*/ /*memset(g->sim->msg_bfr.kvaddr,0,min(NVGPU_CPU_PAGE_SIZE,size));*/
*sim_msg_hdr(g, sim_msg_signature_r()) = sim_msg_signature_valid_v(); *sim_msg_hdr(g, sim_msg_signature_r()) = sim_msg_signature_valid_v();
*sim_msg_hdr(g, sim_msg_result_r()) = sim_msg_result_rpc_pending_v(); *sim_msg_hdr(g, sim_msg_result_r()) = sim_msg_result_rpc_pending_v();
*sim_msg_hdr(g, sim_msg_spare_r()) = sim_msg_spare__init_v(); *sim_msg_hdr(g, sim_msg_spare_r()) = sim_msg_spare__init_v();
@@ -104,7 +104,7 @@ static int rpc_send_message(struct gk20a *g)
*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; % SIM_BFR_SIZE;
/* Update the put pointer. This will trap into the host. */ /* Update the put pointer. This will trap into the host. */
sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put); sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put);
@@ -156,7 +156,7 @@ static int rpc_recv_poll(struct gk20a *g)
/* 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; % SIM_BFR_SIZE;
sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get); sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get);

4
drivers/gpu/nvgpu/common/sim/sim_pci.c
View File

@@ -83,7 +83,7 @@ static int rpc_send_message(struct gk20a *g)
*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; SIM_BFR_SIZE;
/* Update the put pointer. This will trap into the host. */ /* Update the put pointer. This will trap into the host. */
sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put); sim_writel(g->sim, sim_send_put_r(), g->sim->send_ring_put);
@@ -131,7 +131,7 @@ static int rpc_recv_poll(struct gk20a *g)
/* 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; % SIM_BFR_SIZE;
sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get); sim_writel(g->sim, sim_recv_get_r(), g->sim->recv_ring_get);

2
drivers/gpu/nvgpu/common/vgpu/fifo/fifo_vgpu.c
View File

@@ -118,7 +118,7 @@ int vgpu_init_fifo_setup_hw(struct gk20a *g)
nvgpu_log_fn(g, " "); nvgpu_log_fn(g, " ");
/* allocate and map first userd slab for bar1 test. */ /* allocate and map first userd slab for bar1 test. */
err = nvgpu_dma_alloc_sys(g, PAGE_SIZE, mem); err = nvgpu_dma_alloc_sys(g, NVGPU_CPU_PAGE_SIZE, mem);
if (err != 0) { if (err != 0) {
nvgpu_err(g, "userd allocation failed, err=%d", err); nvgpu_err(g, "userd allocation failed, err=%d", err);
return err; return err;

4
drivers/gpu/nvgpu/hal/fb/fb_tu104.c
View File

@@ -191,7 +191,7 @@ int tu104_fb_apply_pdb_cache_war(struct gk20a *g)
/* Bind 256 instance blocks to unused engine ID 0x0 */ /* Bind 256 instance blocks to unused engine ID 0x0 */
for (i = 0U; i < 256U; i++) { for (i = 0U; i < 256U; i++) {
inst_blk_addr = u64_lo32((inst_blk_base_addr + inst_blk_addr = u64_lo32((inst_blk_base_addr +
(U64(i) * U64(PAGE_SIZE))) (U64(i) * U64(NVGPU_CPU_PAGE_SIZE)))
>> fb_mmu_bind_imb_addr_alignment_v()); >> fb_mmu_bind_imb_addr_alignment_v());
nvgpu_writel(g, fb_mmu_bind_imb_r(), nvgpu_writel(g, fb_mmu_bind_imb_r(),
@@ -236,7 +236,7 @@ int tu104_fb_apply_pdb_cache_war(struct gk20a *g)
} }
/* Bind 257th (last) instance block that reserves PDB cache entry 255 */ /* Bind 257th (last) instance block that reserves PDB cache entry 255 */
inst_blk_addr = u64_lo32((inst_blk_base_addr + (256ULL * U64(PAGE_SIZE))) inst_blk_addr = u64_lo32((inst_blk_base_addr + (256ULL * U64(NVGPU_CPU_PAGE_SIZE)))
>> U64(fb_mmu_bind_imb_addr_alignment_v())); >> U64(fb_mmu_bind_imb_addr_alignment_v()));
nvgpu_writel(g, fb_mmu_bind_imb_r(), nvgpu_writel(g, fb_mmu_bind_imb_r(),

18
drivers/gpu/nvgpu/hal/fifo/ramin_tu104.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2018-2019, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2018-2020, NVIDIA CORPORATION. All rights reserved.
* *
* Permission is hereby granted, free of charge, to any person obtaining a * Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"), * copy of this software and associated documentation files (the "Software"),
@@ -32,7 +32,7 @@
int tu104_ramin_init_pdb_cache_war(struct gk20a *g) int tu104_ramin_init_pdb_cache_war(struct gk20a *g)
{ {
u32 size = PAGE_SIZE * 258U; u32 size = NVGPU_CPU_PAGE_SIZE * 258U;
u64 last_bind_pdb_addr; u64 last_bind_pdb_addr;
u64 pdb_addr; u64 pdb_addr;
u32 pdb_addr_lo, pdb_addr_hi; u32 pdb_addr_lo, pdb_addr_hi;
@@ -57,9 +57,9 @@ int tu104_ramin_init_pdb_cache_war(struct gk20a *g)
* valid memory * valid memory
* First 256 binds can happen to dummy addresses * First 256 binds can happen to dummy addresses
*/ */
pdb_addr = PAGE_SIZE; pdb_addr = NVGPU_CPU_PAGE_SIZE;
last_bind_pdb_addr = nvgpu_mem_get_addr(g, &g->pdb_cache_war_mem) + last_bind_pdb_addr = nvgpu_mem_get_addr(g, &g->pdb_cache_war_mem) +
(257U * PAGE_SIZE); (257U * NVGPU_CPU_PAGE_SIZE);
/* Setup first 256 instance blocks */ /* Setup first 256 instance blocks */
for (i = 0U; i < 256U; i++) { for (i = 0U; i < 256U; i++) {
@@ -67,7 +67,7 @@ int tu104_ramin_init_pdb_cache_war(struct gk20a *g)
pdb_addr_hi = u64_hi32(pdb_addr); pdb_addr_hi = u64_hi32(pdb_addr);
nvgpu_mem_wr32(g, &g->pdb_cache_war_mem, nvgpu_mem_wr32(g, &g->pdb_cache_war_mem,
ram_in_page_dir_base_lo_w() + (i * PAGE_SIZE / 4U), ram_in_page_dir_base_lo_w() + (i * NVGPU_CPU_PAGE_SIZE / 4U),
nvgpu_aperture_mask(g, &g->pdb_cache_war_mem, nvgpu_aperture_mask(g, &g->pdb_cache_war_mem,
ram_in_page_dir_base_target_sys_mem_ncoh_f(), ram_in_page_dir_base_target_sys_mem_ncoh_f(),
ram_in_page_dir_base_target_sys_mem_coh_f(), ram_in_page_dir_base_target_sys_mem_coh_f(),
@@ -78,10 +78,10 @@ int tu104_ramin_init_pdb_cache_war(struct gk20a *g)
ram_in_use_ver2_pt_format_true_f()); ram_in_use_ver2_pt_format_true_f());
nvgpu_mem_wr32(g, &g->pdb_cache_war_mem, nvgpu_mem_wr32(g, &g->pdb_cache_war_mem,
ram_in_page_dir_base_hi_w() + (i * PAGE_SIZE / 4U), ram_in_page_dir_base_hi_w() + (i * NVGPU_CPU_PAGE_SIZE / 4U),
ram_in_page_dir_base_hi_f(pdb_addr_hi)); ram_in_page_dir_base_hi_f(pdb_addr_hi));
pdb_addr += PAGE_SIZE; pdb_addr += NVGPU_CPU_PAGE_SIZE;
} }
/* Setup 257th instance block */ /* Setup 257th instance block */
@@ -89,7 +89,7 @@ int tu104_ramin_init_pdb_cache_war(struct gk20a *g)
pdb_addr_hi = u64_hi32(last_bind_pdb_addr); pdb_addr_hi = u64_hi32(last_bind_pdb_addr);
nvgpu_mem_wr32(g, &g->pdb_cache_war_mem, nvgpu_mem_wr32(g, &g->pdb_cache_war_mem,
ram_in_page_dir_base_lo_w() + (256U * PAGE_SIZE / 4U), ram_in_page_dir_base_lo_w() + (256U * NVGPU_CPU_PAGE_SIZE / 4U),
nvgpu_aperture_mask(g, &g->pdb_cache_war_mem, nvgpu_aperture_mask(g, &g->pdb_cache_war_mem,
ram_in_page_dir_base_target_sys_mem_ncoh_f(), ram_in_page_dir_base_target_sys_mem_ncoh_f(),
ram_in_page_dir_base_target_sys_mem_coh_f(), ram_in_page_dir_base_target_sys_mem_coh_f(),
@@ -100,7 +100,7 @@ int tu104_ramin_init_pdb_cache_war(struct gk20a *g)
ram_in_use_ver2_pt_format_true_f()); ram_in_use_ver2_pt_format_true_f());
nvgpu_mem_wr32(g, &g->pdb_cache_war_mem, nvgpu_mem_wr32(g, &g->pdb_cache_war_mem,
ram_in_page_dir_base_hi_w() + (256U * PAGE_SIZE / 4U), ram_in_page_dir_base_hi_w() + (256U * NVGPU_CPU_PAGE_SIZE / 4U),
ram_in_page_dir_base_hi_f(pdb_addr_hi)); ram_in_page_dir_base_hi_f(pdb_addr_hi));
return 0; return 0;

2
drivers/gpu/nvgpu/hal/fifo/tsg_gv11b_fusa.c
View File

@@ -108,7 +108,7 @@ int gv11b_tsg_init_eng_method_buffers(struct gk20a *g, struct nvgpu_tsg *tsg)
int err = 0; int err = 0;
int i; int i;
unsigned int runque, buffer_size; unsigned int runque, buffer_size;
u32 page_size = U32(PAGE_SIZE); u32 page_size = U32(NVGPU_CPU_PAGE_SIZE);
unsigned int num_pbdma = nvgpu_get_litter_value(g, GPU_LIT_HOST_NUM_PBDMA); unsigned int num_pbdma = nvgpu_get_litter_value(g, GPU_LIT_HOST_NUM_PBDMA);
if (tsg->eng_method_buffers != NULL) { if (tsg->eng_method_buffers != NULL) {

4
drivers/gpu/nvgpu/hal/mm/mm_gk20a.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2019-2020, NVIDIA CORPORATION. All rights reserved.
* *
* Permission is hereby granted, free of charge, to any person obtaining a * Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"), * copy of this software and associated documentation files (the "Software"),
@@ -51,7 +51,7 @@ u64 gk20a_mm_bar1_map_userd(struct gk20a *g, struct nvgpu_mem *mem, u32 offset)
u64 gpu_va = f->userd_gpu_va + offset; u64 gpu_va = f->userd_gpu_va + offset;
return nvgpu_gmmu_map_fixed(g->mm.bar1.vm, mem, gpu_va, return nvgpu_gmmu_map_fixed(g->mm.bar1.vm, mem, gpu_va,
PAGE_SIZE, 0, NVGPU_CPU_PAGE_SIZE, 0,
gk20a_mem_flag_none, false, gk20a_mem_flag_none, false,
mem->aperture); mem->aperture);
} }

2
drivers/gpu/nvgpu/hal/sync/syncpt_cmdbuf_gv11b_fusa.c
View File

@@ -74,7 +74,7 @@ int gv11b_syncpt_alloc_buf(struct nvgpu_channel *c,
return err; return err;
} }
nr_pages = DIV_ROUND_UP(g->syncpt_size, PAGE_SIZE); nr_pages = DIV_ROUND_UP(g->syncpt_size, NVGPU_CPU_PAGE_SIZE);
err = nvgpu_mem_create_from_phys(g, syncpt_buf, err = nvgpu_mem_create_from_phys(g, syncpt_buf,
nvgpu_safe_add_u64(g->syncpt_unit_base, nvgpu_safe_add_u64(g->syncpt_unit_base,
nvgpu_nvhost_syncpt_unit_interface_get_byte_offset(g, nvgpu_nvhost_syncpt_unit_interface_get_byte_offset(g,

2
drivers/gpu/nvgpu/include/nvgpu/gmmu.h
View File

@@ -239,7 +239,7 @@ static inline const char *nvgpu_gmmu_perm_str(enum gk20a_mem_rw_flag p)
* aligned. Although lower PDE tables can be aligned at 256B boundaries * aligned. Although lower PDE tables can be aligned at 256B boundaries
* the PDB must be 4K aligned. * the PDB must be 4K aligned.
* *
* Currently PAGE_SIZE is used, even when 64K, to work around an issue * Currently NVGPU_CPU_PAGE_SIZE is used, even when 64K, to work around an issue
* with the PDB TLB invalidate code not being pd_cache aware yet. * with the PDB TLB invalidate code not being pd_cache aware yet.
* *
* @return 0 in case of success. * @return 0 in case of success.

4
drivers/gpu/nvgpu/include/nvgpu/gr/ctx.h
View File

@@ -55,9 +55,9 @@ struct zcull_ctx_desc;
*/ */
/** Number of slots required in patch buffer per entry. */ /** Number of slots required in patch buffer per entry. */
#define PATCH_CTX_SLOTS_REQUIRED_PER_ENTRY 2U #define PATCH_CTX_SLOTS_REQUIRED_PER_ENTRY 2U
/** Number of slots per PAGE_SIZE. */ /** Number of slots per NVGPU_CPU_PAGE_SIZE. */
#define PATCH_CTX_SLOTS_PER_PAGE \ #define PATCH_CTX_SLOTS_PER_PAGE \
(PAGE_SIZE/(PATCH_CTX_SLOTS_REQUIRED_PER_ENTRY * (u32)sizeof(u32))) (NVGPU_CPU_PAGE_SIZE/(PATCH_CTX_SLOTS_REQUIRED_PER_ENTRY * (u32)sizeof(u32)))
/** Get number of entries in patch buffer given the size of buffer. */ /** Get number of entries in patch buffer given the size of buffer. */
#define PATCH_CTX_ENTRIES_FROM_SIZE(size) ((size)/sizeof(u32)) #define PATCH_CTX_ENTRIES_FROM_SIZE(size) ((size)/sizeof(u32))

10
drivers/gpu/nvgpu/include/nvgpu/sim.h
View File

@@ -32,6 +32,16 @@
#endif #endif
/** @endcond DOXYGEN_SHOULD_SKIP_THIS */ /** @endcond DOXYGEN_SHOULD_SKIP_THIS */
/*
* Size of SIM ring buffers.
*
* Although, each buffer is allocated with NVGPU_CPU_PAGE_SIZE bytes.
* The send and receive interface can only be configured to work with buffers of
* sizes: 4K, 8K, 12K and 16K. Furthermore, this size should match with size
* configured in the fmodel chiplib. At present, the agreed buffer size is 4K.
*/
#define SIM_BFR_SIZE (SZ_4K)
struct sim_nvgpu { struct sim_nvgpu {
struct gk20a *g; struct gk20a *g;
u32 send_ring_put; u32 send_ring_put;

6
drivers/gpu/nvgpu/include/nvgpu/utils.h
View File

@@ -32,6 +32,12 @@
#include <nvgpu/posix/utils.h> #include <nvgpu/posix/utils.h>
#endif #endif
/*
* PAGE_SIZE is OS specific and can vary across OSes. Depending on the OS it maybe
* defined to 4K or 64K.
*/
#define NVGPU_CPU_PAGE_SIZE PAGE_SIZE
/** /**
* @file * @file
* *

4
drivers/gpu/nvgpu/os/linux/fecs_trace_linux.c
View File

@@ -200,7 +200,7 @@ int nvgpu_gr_fecs_trace_ring_alloc(struct gk20a *g,
{ {
struct nvgpu_ctxsw_ring_header *hdr; struct nvgpu_ctxsw_ring_header *hdr;
*size = round_up(*size, PAGE_SIZE); *size = round_up(*size, NVGPU_CPU_PAGE_SIZE);
hdr = vmalloc_user(*size); hdr = vmalloc_user(*size);
if (!hdr) if (!hdr)
return -ENOMEM; return -ENOMEM;
@@ -531,7 +531,7 @@ int gk20a_ctxsw_dev_mmap(struct file *filp, struct vm_area_struct *vma)
unsigned long vsize = vma->vm_end - vma->vm_start; unsigned long vsize = vma->vm_end - vma->vm_start;
size = min(mmapsize, vsize); size = min(mmapsize, vsize);
size = round_up(size, PAGE_SIZE); size = round_up(size, NVGPU_CPU_PAGE_SIZE);
ret = remap_pfn_range(vma, vma->vm_start, ret = remap_pfn_range(vma, vma->vm_start,
(unsigned long) mmapaddr, (unsigned long) mmapaddr,

4
drivers/gpu/nvgpu/os/linux/fecs_trace_linux.h
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2017-2019, NVIDIA Corporation. All rights reserved. * Copyright (c) 2017-2020, 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,
@@ -20,7 +20,7 @@
#include <nvgpu/types.h> #include <nvgpu/types.h>
#define GK20A_CTXSW_TRACE_NUM_DEVS 1 #define GK20A_CTXSW_TRACE_NUM_DEVS 1
#define GK20A_CTXSW_TRACE_MAX_VM_RING_SIZE (128*PAGE_SIZE) #define GK20A_CTXSW_TRACE_MAX_VM_RING_SIZE (128*NVGPU_CPU_PAGE_SIZE)
struct file; struct file;
struct inode; struct inode;

10
drivers/gpu/nvgpu/os/linux/kmem.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2017-2020, 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,
@@ -40,7 +40,7 @@ void *nvgpu_big_alloc_impl(struct gk20a *g, size_t size, bool clear)
{ {
void *p; void *p;
if (size > PAGE_SIZE) { if (size > NVGPU_CPU_PAGE_SIZE) {
if (clear) if (clear)
p = nvgpu_vzalloc(g, size); p = nvgpu_vzalloc(g, size);
else else
@@ -314,12 +314,12 @@ static int __nvgpu_free_kmem_alloc(struct nvgpu_mem_alloc_tracker *tracker,
static void __nvgpu_check_valloc_size(unsigned long size) static void __nvgpu_check_valloc_size(unsigned long size)
{ {
WARN(size < PAGE_SIZE, "Alloc smaller than page size! (%lu)!\n", size); WARN(size < NVGPU_CPU_PAGE_SIZE, "Alloc smaller than page size! (%lu)!\n", size);
} }
static void __nvgpu_check_kalloc_size(size_t size) static void __nvgpu_check_kalloc_size(size_t size)
{ {
WARN(size > PAGE_SIZE, "Alloc larger than page size! (%zu)!\n", size); WARN(size > NVGPU_CPU_PAGE_SIZE, "Alloc larger than page size! (%zu)!\n", size);
} }
void *__nvgpu_track_vmalloc(struct gk20a *g, unsigned long size, void *__nvgpu_track_vmalloc(struct gk20a *g, unsigned long size,
@@ -570,7 +570,7 @@ int nvgpu_kmem_init(struct gk20a *g)
nvgpu_mutex_init(&g->vmallocs->lock); nvgpu_mutex_init(&g->vmallocs->lock);
nvgpu_mutex_init(&g->kmallocs->lock); nvgpu_mutex_init(&g->kmallocs->lock);
g->vmallocs->min_alloc = PAGE_SIZE; g->vmallocs->min_alloc = NVGPU_CPU_PAGE_SIZE;
g->kmallocs->min_alloc = KMALLOC_MIN_SIZE; g->kmallocs->min_alloc = KMALLOC_MIN_SIZE;
/* /*

10
drivers/gpu/nvgpu/os/linux/linux-dma.c
View File

@@ -180,7 +180,7 @@ static void *nvgpu_dma_alloc_no_iommu(struct device *dev, size_t size,
struct page **pages; struct page **pages;
int i = 0; int i = 0;
if (array_size <= PAGE_SIZE) if (array_size <= NVGPU_CPU_PAGE_SIZE)
pages = kzalloc(array_size, GFP_KERNEL); pages = kzalloc(array_size, GFP_KERNEL);
else else
pages = vzalloc(array_size); pages = vzalloc(array_size);
@@ -205,7 +205,7 @@ static void *nvgpu_dma_alloc_no_iommu(struct device *dev, size_t size,
pages[i + j] = pages[i] + j; pages[i + j] = pages[i] + j;
} }
memset(page_address(pages[i]), 0, PAGE_SIZE << order); memset(page_address(pages[i]), 0, NVGPU_CPU_PAGE_SIZE << order);
i += 1 << order; i += 1 << order;
count -= 1 << order; count -= 1 << order;
@@ -216,7 +216,7 @@ static void *nvgpu_dma_alloc_no_iommu(struct device *dev, size_t size,
return (void *)pages; return (void *)pages;
error: error:
__nvgpu_dma_free_no_iommu(pages, i, array_size > PAGE_SIZE); __nvgpu_dma_free_no_iommu(pages, i, array_size > NVGPU_CPU_PAGE_SIZE);
return NULL; return NULL;
} }
@@ -228,7 +228,7 @@ static void nvgpu_dma_free_no_iommu(size_t size, void *vaddr)
WARN_ON(!pages); WARN_ON(!pages);
__nvgpu_dma_free_no_iommu(pages, count, array_size > PAGE_SIZE); __nvgpu_dma_free_no_iommu(pages, count, array_size > NVGPU_CPU_PAGE_SIZE);
} }
/* Check if IOMMU is available and if GPU uses it */ /* Check if IOMMU is available and if GPU uses it */
@@ -570,7 +570,7 @@ int nvgpu_get_sgtable_from_pages(struct gk20a *g, struct sg_table **sgt,
} }
err = sg_alloc_table_from_pages(tbl, pages, err = sg_alloc_table_from_pages(tbl, pages,
DIV_ROUND_UP(size, PAGE_SIZE), DIV_ROUND_UP(size, NVGPU_CPU_PAGE_SIZE),
0, size, GFP_KERNEL); 0, size, GFP_KERNEL);
if (err) if (err)
goto fail; goto fail;

6
drivers/gpu/nvgpu/os/linux/nvgpu_mem.c
View File

@@ -158,8 +158,8 @@ int nvgpu_mem_create_from_mem(struct gk20a *g,
u64 start_page, size_t nr_pages) u64 start_page, size_t nr_pages)
{ {
int ret; int ret;
u64 start = start_page * PAGE_SIZE; u64 start = start_page * NVGPU_CPU_PAGE_SIZE;
u64 size = nr_pages * PAGE_SIZE; u64 size = nr_pages * NVGPU_CPU_PAGE_SIZE;
dma_addr_t new_iova; dma_addr_t new_iova;
if (src->aperture != APERTURE_SYSMEM) if (src->aperture != APERTURE_SYSMEM)
@@ -178,7 +178,7 @@ int nvgpu_mem_create_from_mem(struct gk20a *g,
/* Re-use the CPU mapping only if the mapping was made by the DMA API */ /* Re-use the CPU mapping only if the mapping was made by the DMA API */
if (!(src->priv.flags & NVGPU_DMA_NO_KERNEL_MAPPING)) if (!(src->priv.flags & NVGPU_DMA_NO_KERNEL_MAPPING))
dest->cpu_va = src->cpu_va + (PAGE_SIZE * start_page); dest->cpu_va = src->cpu_va + (NVGPU_CPU_PAGE_SIZE * start_page);
dest->priv.pages = src->priv.pages + start_page; dest->priv.pages = src->priv.pages + start_page;
dest->priv.flags = src->priv.flags; dest->priv.flags = src->priv.flags;

2
drivers/gpu/nvgpu/os/linux/pci_power.c
View File

@@ -117,7 +117,7 @@ static ssize_t probed_gpus_show(struct device_driver *drv, char *buf)
ssize_t count = 0; ssize_t count = 0;
list_for_each_entry_safe(pp, tmp_pp, &nvgpu_pci_power_devs, list) { list_for_each_entry_safe(pp, tmp_pp, &nvgpu_pci_power_devs, list) {
count += snprintf(buf, PAGE_SIZE - count, "pci-%s\t%s\n", count += snprintf(buf, NVGPU_CPU_PAGE_SIZE - count, "pci-%s\t%s\n",
pp->pci_dev_name, pp->pci_dev_name,
pp->pci_dev ? "PoweredOn" : "PoweredOff"); pp->pci_dev ? "PoweredOn" : "PoweredOff");
} }

2
drivers/gpu/nvgpu/os/linux/platform_gk20a_tegra.c
View File

@@ -666,7 +666,7 @@ int gk20a_tegra_init_secure_alloc(struct gk20a_platform *platform)
return 0; return 0;
} }
#if PAGE_SIZE > 4096 #if NVGPU_CPU_PAGE_SIZE > 4096
platform->secure_buffer_size += SZ_64K; platform->secure_buffer_size += SZ_64K;
#endif #endif
(void)dma_alloc_attrs(&tegra_vpr_dev, platform->secure_buffer_size, &iova, (void)dma_alloc_attrs(&tegra_vpr_dev, platform->secure_buffer_size, &iova,

2
drivers/gpu/nvgpu/os/linux/platform_gp10b_tegra.c
View File

@@ -206,7 +206,7 @@ static int gp10b_tegra_probe(struct device *dev)
return ret; return ret;
platform->disable_bigpage = !iommu_get_domain_for_dev(dev) && platform->disable_bigpage = !iommu_get_domain_for_dev(dev) &&
(PAGE_SIZE < SZ_64K); (NVGPU_CPU_PAGE_SIZE < SZ_64K);
#ifdef CONFIG_OF #ifdef CONFIG_OF
of_chosen = of_find_node_by_path("/chosen"); of_chosen = of_find_node_by_path("/chosen");

2
drivers/gpu/nvgpu/os/linux/platform_gv11b_tegra.c
View File

@@ -96,7 +96,7 @@ static int gv11b_tegra_probe(struct device *dev)
return err; return err;
platform->disable_bigpage = !(iommu_get_domain_for_dev(dev)) && platform->disable_bigpage = !(iommu_get_domain_for_dev(dev)) &&
(PAGE_SIZE < SZ_64K); (NVGPU_CPU_PAGE_SIZE < SZ_64K);
#ifdef CONFIG_OF #ifdef CONFIG_OF
of_chosen = of_find_node_by_path("/chosen"); of_chosen = of_find_node_by_path("/chosen");

50
drivers/gpu/nvgpu/os/linux/sysfs.c
View File

@@ -79,7 +79,7 @@ static ssize_t elcg_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->elcg_enabled ? 1 : 0); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->elcg_enabled ? 1 : 0);
} }
static DEVICE_ATTR(elcg_enable, ROOTRW, elcg_enable_read, elcg_enable_store); static DEVICE_ATTR(elcg_enable, ROOTRW, elcg_enable_read, elcg_enable_store);
@@ -119,7 +119,7 @@ static ssize_t blcg_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->blcg_enabled ? 1 : 0); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->blcg_enabled ? 1 : 0);
} }
@@ -165,7 +165,7 @@ static ssize_t slcg_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->slcg_enabled ? 1 : 0); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->slcg_enabled ? 1 : 0);
} }
static DEVICE_ATTR(slcg_enable, ROOTRW, slcg_enable_read, slcg_enable_store); static DEVICE_ATTR(slcg_enable, ROOTRW, slcg_enable_read, slcg_enable_store);
@@ -189,7 +189,7 @@ static ssize_t ptimer_scale_factor_show(struct device *dev,
((u32)(src_freq_hz) / ((u32)(src_freq_hz) /
(u32)(PTIMER_FP_FACTOR)); (u32)(PTIMER_FP_FACTOR));
res = snprintf(buf, res = snprintf(buf,
PAGE_SIZE, NVGPU_CPU_PAGE_SIZE,
"%u.%u\n", "%u.%u\n",
scaling_factor_fp / PTIMER_FP_FACTOR, scaling_factor_fp / PTIMER_FP_FACTOR,
scaling_factor_fp % PTIMER_FP_FACTOR); scaling_factor_fp % PTIMER_FP_FACTOR);
@@ -217,7 +217,7 @@ static ssize_t ptimer_ref_freq_show(struct device *dev,
return -EINVAL; return -EINVAL;
} }
res = snprintf(buf, PAGE_SIZE, "%u\n", PTIMER_REF_FREQ_HZ); res = snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u\n", PTIMER_REF_FREQ_HZ);
return res; return res;
@@ -242,7 +242,7 @@ static ssize_t ptimer_src_freq_show(struct device *dev,
return -EINVAL; return -EINVAL;
} }
res = snprintf(buf, PAGE_SIZE, "%u\n", src_freq_hz); res = snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u\n", src_freq_hz);
return res; return res;
@@ -260,7 +260,7 @@ static ssize_t gpu_powered_on_show(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%s\n", nvgpu_get_power_state(g)); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%s\n", nvgpu_get_power_state(g));
} }
static DEVICE_ATTR(gpu_powered_on, S_IRUGO, gpu_powered_on_show, NULL); static DEVICE_ATTR(gpu_powered_on, S_IRUGO, gpu_powered_on_show, NULL);
@@ -318,7 +318,7 @@ static ssize_t railgate_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n",
nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE) ? 1 : 0); nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE) ? 1 : 0);
} }
@@ -360,7 +360,7 @@ static ssize_t railgate_delay_show(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->railgate_delay); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->railgate_delay);
} }
static DEVICE_ATTR(railgate_delay, ROOTRW, railgate_delay_show, static DEVICE_ATTR(railgate_delay, ROOTRW, railgate_delay_show,
railgate_delay_store); railgate_delay_store);
@@ -374,7 +374,7 @@ static ssize_t is_railgated_show(struct device *dev,
if (platform->is_railgated) if (platform->is_railgated)
is_railgated = platform->is_railgated(dev); is_railgated = platform->is_railgated(dev);
return snprintf(buf, PAGE_SIZE, "%s\n", is_railgated ? "yes" : "no"); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%s\n", is_railgated ? "yes" : "no");
} }
static DEVICE_ATTR(is_railgated, S_IRUGO, is_railgated_show, NULL); static DEVICE_ATTR(is_railgated, S_IRUGO, is_railgated_show, NULL);
@@ -387,7 +387,7 @@ static ssize_t counters_show(struct device *dev,
nvgpu_pmu_get_load_counters(g, &busy_cycles, &total_cycles); nvgpu_pmu_get_load_counters(g, &busy_cycles, &total_cycles);
res = snprintf(buf, PAGE_SIZE, "%u %u\n", busy_cycles, total_cycles); res = snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u %u\n", busy_cycles, total_cycles);
return res; return res;
} }
@@ -427,7 +427,7 @@ static ssize_t gk20a_load_show(struct device *dev,
gk20a_idle(g); gk20a_idle(g);
} }
res = snprintf(buf, PAGE_SIZE, "%u\n", busy_time); res = snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u\n", busy_time);
return res; return res;
} }
@@ -468,7 +468,7 @@ static ssize_t elpg_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n",
nvgpu_pg_elpg_is_enabled(g) ? 1 : 0); nvgpu_pg_elpg_is_enabled(g) ? 1 : 0);
} }
@@ -521,7 +521,7 @@ static ssize_t ldiv_slowdown_factor_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->ldiv_slowdown_factor); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->ldiv_slowdown_factor);
} }
static DEVICE_ATTR(ldiv_slowdown_factor, ROOTRW, static DEVICE_ATTR(ldiv_slowdown_factor, ROOTRW,
@@ -588,7 +588,7 @@ static ssize_t mscg_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->mscg_enabled ? 1 : 0); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->mscg_enabled ? 1 : 0);
} }
static DEVICE_ATTR(mscg_enable, ROOTRW, mscg_enable_read, mscg_enable_store); static DEVICE_ATTR(mscg_enable, ROOTRW, mscg_enable_read, mscg_enable_store);
@@ -641,7 +641,7 @@ static ssize_t aelpg_param_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, return snprintf(buf, NVGPU_CPU_PAGE_SIZE,
"%d %d %d %d %d\n", g->pmu->pg->aelpg_param[0], "%d %d %d %d %d\n", g->pmu->pg->aelpg_param[0],
g->pmu->pg->aelpg_param[1], g->pmu->pg->aelpg_param[2], g->pmu->pg->aelpg_param[1], g->pmu->pg->aelpg_param[2],
g->pmu->pg->aelpg_param[3], g->pmu->pg->aelpg_param[4]); g->pmu->pg->aelpg_param[3], g->pmu->pg->aelpg_param[4]);
@@ -697,7 +697,7 @@ static ssize_t aelpg_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->aelpg_enabled ? 1 : 0); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->aelpg_enabled ? 1 : 0);
} }
static DEVICE_ATTR(aelpg_enable, ROOTRW, static DEVICE_ATTR(aelpg_enable, ROOTRW,
@@ -709,7 +709,7 @@ static ssize_t allow_all_enable_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->allow_all ? 1 : 0); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->allow_all ? 1 : 0);
} }
static ssize_t allow_all_enable_store(struct device *dev, static ssize_t allow_all_enable_store(struct device *dev,
@@ -751,7 +751,7 @@ static ssize_t emc3d_ratio_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->emc3d_ratio); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->emc3d_ratio);
} }
static DEVICE_ATTR(emc3d_ratio, ROOTRW, emc3d_ratio_read, emc3d_ratio_store); static DEVICE_ATTR(emc3d_ratio, ROOTRW, emc3d_ratio_read, emc3d_ratio_store);
@@ -765,7 +765,7 @@ static ssize_t fmax_at_vmin_safe_read(struct device *dev,
if (g->ops.clk.get_fmax_at_vmin_safe) if (g->ops.clk.get_fmax_at_vmin_safe)
gpu_fmax_at_vmin_hz = g->ops.clk.get_fmax_at_vmin_safe(g); gpu_fmax_at_vmin_hz = g->ops.clk.get_fmax_at_vmin_safe(g);
return snprintf(buf, PAGE_SIZE, "%d\n", (int)(gpu_fmax_at_vmin_hz)); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", (int)(gpu_fmax_at_vmin_hz));
} }
static DEVICE_ATTR(fmax_at_vmin_safe, S_IRUGO, fmax_at_vmin_safe_read, NULL); static DEVICE_ATTR(fmax_at_vmin_safe, S_IRUGO, fmax_at_vmin_safe_read, NULL);
@@ -813,7 +813,7 @@ static ssize_t force_idle_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->forced_idle ? 1 : 0); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->forced_idle ? 1 : 0);
} }
static DEVICE_ATTR(force_idle, ROOTRW, force_idle_read, force_idle_store); static DEVICE_ATTR(force_idle, ROOTRW, force_idle_read, force_idle_store);
@@ -824,7 +824,7 @@ static ssize_t tpc_pg_mask_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%d\n", g->tpc_pg_mask); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%d\n", g->tpc_pg_mask);
} }
static bool is_tpc_mask_valid(struct gk20a *g, u32 tpc_mask) static bool is_tpc_mask_valid(struct gk20a *g, u32 tpc_mask)
@@ -963,7 +963,7 @@ static ssize_t tpc_fs_mask_read(struct device *dev,
gk20a_idle(g); gk20a_idle(g);
return snprintf(buf, PAGE_SIZE, "0x%x\n", tpc_fs_mask); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "0x%x\n", tpc_fs_mask);
} }
static DEVICE_ATTR(tpc_fs_mask, ROOTRW, tpc_fs_mask_read, tpc_fs_mask_store); static DEVICE_ATTR(tpc_fs_mask, ROOTRW, tpc_fs_mask_read, tpc_fs_mask_store);
@@ -973,7 +973,7 @@ static ssize_t tsg_timeslice_min_us_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%u\n", g->tsg_timeslice_min_us); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u\n", g->tsg_timeslice_min_us);
} }
static ssize_t tsg_timeslice_min_us_store(struct device *dev, static ssize_t tsg_timeslice_min_us_store(struct device *dev,
@@ -1001,7 +1001,7 @@ static ssize_t tsg_timeslice_max_us_read(struct device *dev,
{ {
struct gk20a *g = get_gk20a(dev); struct gk20a *g = get_gk20a(dev);
return snprintf(buf, PAGE_SIZE, "%u\n", g->tsg_timeslice_max_us); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u\n", g->tsg_timeslice_max_us);
} }
static ssize_t tsg_timeslice_max_us_store(struct device *dev, static ssize_t tsg_timeslice_max_us_store(struct device *dev,

6
drivers/gpu/nvgpu/os/linux/vgpu/sysfs_vgpu.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2017-2020, 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,
@@ -37,7 +37,7 @@ static ssize_t vgpu_load_show(struct device *dev,
if (err) if (err)
return err; return err;
return snprintf(buf, PAGE_SIZE, "%u\n", p->load); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u\n", p->load);
} }
static DEVICE_ATTR(load, S_IRUGO, vgpu_load_show, NULL); static DEVICE_ATTR(load, S_IRUGO, vgpu_load_show, NULL);
@@ -64,7 +64,7 @@ static ssize_t vgpu_ecc_stat_show(struct device *dev,
return err; return err;
} }
return snprintf(buf, PAGE_SIZE, "%u\n", p->value); return snprintf(buf, NVGPU_CPU_PAGE_SIZE, "%u\n", p->value);
} }
static int vgpu_create_ecc_sysfs(struct device *dev) static int vgpu_create_ecc_sysfs(struct device *dev)

2
drivers/gpu/nvgpu/os/linux/vgpu/vgpu_linux.c
View File

@@ -111,7 +111,7 @@ static void vgpu_init_vars(struct gk20a *g, struct gk20a_platform *platform)
nvgpu_set_enabled(g, NVGPU_CAN_RAILGATE, platform->can_railgate_init); nvgpu_set_enabled(g, NVGPU_CAN_RAILGATE, platform->can_railgate_init);
g->railgate_delay = platform->railgate_delay_init; g->railgate_delay = platform->railgate_delay_init;
g->mm.disable_bigpage = PAGE_SIZE < SZ_64K; g->mm.disable_bigpage = NVGPU_CPU_PAGE_SIZE < SZ_64K;
nvgpu_set_enabled(g, NVGPU_MM_UNIFIED_MEMORY, nvgpu_set_enabled(g, NVGPU_MM_UNIFIED_MEMORY,
platform->unified_memory); platform->unified_memory);
nvgpu_set_enabled(g, NVGPU_MM_UNIFY_ADDRESS_SPACES, nvgpu_set_enabled(g, NVGPU_MM_UNIFY_ADDRESS_SPACES,

6
drivers/gpu/nvgpu/os/posix/posix-nvgpu_mem.c
View File

@@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2018-2019, NVIDIA CORPORATION. All rights reserved. * Copyright (c) 2018-2020, NVIDIA CORPORATION. All rights reserved.
* *
* Permission is hereby granted, free of charge, to any person obtaining a * Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"), * copy of this software and associated documentation files (the "Software"),
@@ -298,8 +298,8 @@ int nvgpu_mem_create_from_mem(struct gk20a *g,
struct nvgpu_mem *dest, struct nvgpu_mem *src, struct nvgpu_mem *dest, struct nvgpu_mem *src,
u64 start_page, size_t nr_pages) u64 start_page, size_t nr_pages)
{ {
u64 start = start_page * U64(PAGE_SIZE); u64 start = start_page * U64(NVGPU_CPU_PAGE_SIZE);
u64 size = U64(nr_pages) * U64(PAGE_SIZE); u64 size = U64(nr_pages) * U64(NVGPU_CPU_PAGE_SIZE);
if (src->aperture != APERTURE_SYSMEM) { if (src->aperture != APERTURE_SYSMEM) {
return -EINVAL; return -EINVAL;