From 5d30a5cda37ca349b4d9cb7e1985c7a0849001b6 Mon Sep 17 00:00:00 2001 From: Mahantesh Kumbar Date: Thu, 6 Sep 2018 20:44:27 +0530 Subject: gpu: nvgpu: ACR code refactor -Created struct nvgpu_acr to hold acr module related member within single struct which are currently spread across multiple structs like nvgpu_pmu, pmu_ops & gk20a. -Created struct hs_flcn_bl struct to hold ACR HS bootloader specific members -Created struct hs_acr to hold ACR ucode specific members like bootloader data using struct hs_flcn_bl, acr type & falcon info on which ACR ucode need to run. -Created acr ops under struct nvgpu_acr to perform ACR specific operation, currently ACR ops were part PMU which caused to have always dependence on PMU even though ACR was not executing on PMU. -Added acr_remove_support ops which will be called as part of gk20a_remove_support() method, earlier acr cleanup was part of pmu remove_support method. -Created define for ACR types, -Ops acr_sw_init() function helps to set ACR properties statically for chip currently in execution & assign ops to point to needed functions as per chip. -Ops acr_sw_init execute at early as nvgpu_init_mm_support calls acr function to alloc blob space. -Created ops to fill bootloader descriptor & to patch WPR info to ACR uocde based on interfaces used to bootstrap ACR ucode. -Created function gm20b_bootstrap_hs_acr() function which is now common HAL for all chips to bootstrap ACR, earlier had 3 different function for gm20b/gp10b, gv11b & for all dgpu based on interface needed. -Removed duplicate code for falcon engine wherever common falcon code can be used. -Removed ACR code dependent on PMU & made changes to use from nvgpu_acr. JIRA NVGPU-1148 Change-Id: I39951d2fc9a0bb7ee6057e0fa06da78045d47590 Signed-off-by: Mahantesh Kumbar Reviewed-on: https://git-master.nvidia.com/r/1813231 GVS: Gerrit_Virtual_Submit Reviewed-by: svc-misra-checker Reviewed-by: Vijayakumar Subbu Reviewed-by: mobile promotions Tested-by: mobile promotions --- drivers/gpu/nvgpu/gm20b/acr_gm20b.c | 374 +++++++++++++++++++++++++++++++++++- drivers/gpu/nvgpu/gm20b/acr_gm20b.h | 17 +- drivers/gpu/nvgpu/gm20b/hal_gm20b.c | 6 +- drivers/gpu/nvgpu/gm20b/pmu_gm20b.c | 60 ++++++ drivers/gpu/nvgpu/gm20b/pmu_gm20b.h | 4 +- 5 files changed, 450 insertions(+), 11 deletions(-) (limited to 'drivers/gpu/nvgpu/gm20b') diff --git a/drivers/gpu/nvgpu/gm20b/acr_gm20b.c b/drivers/gpu/nvgpu/gm20b/acr_gm20b.c index 87f3ef54..a4657ff3 100644 --- a/drivers/gpu/nvgpu/gm20b/acr_gm20b.c +++ b/drivers/gpu/nvgpu/gm20b/acr_gm20b.c @@ -38,8 +38,10 @@ #include #include "mm_gm20b.h" +#include "pmu_gm20b.h" #include "acr_gm20b.h" +#include #include typedef int (*get_ucode_details)(struct gk20a *g, struct flcn_ucode_img *udata); @@ -358,7 +360,7 @@ int prepare_ucode_blob(struct gk20a *g) g->ops.fb.vpr_info_fetch(g); gr_gk20a_init_ctxsw_ucode(g); - g->ops.pmu.get_wpr(g, &wpr_inf); + g->acr.get_wpr_info(g, &wpr_inf); nvgpu_pmu_dbg(g, "wpr carveout base:%llx\n", wpr_inf.wpr_base); nvgpu_pmu_dbg(g, "wpr carveout size :%llx\n", wpr_inf.size); @@ -377,7 +379,7 @@ int prepare_ucode_blob(struct gk20a *g) } /*Alloc memory to hold ucode blob contents*/ - err = g->ops.pmu.alloc_blob_space(g, plsfm->wpr_size + err = g->acr.alloc_blob_space(g, plsfm->wpr_size , &g->acr.ucode_blob); if (err) { goto free_sgt; @@ -450,7 +452,7 @@ static int lsfm_discover_ucode_images(struct gk20a *g, /*0th index is always PMU which is already handled in earlier if condition*/ - for (i = 1; i < (MAX_SUPPORTED_LSFM); i++) { + for (i = 1; i < g->acr.max_supported_lsfm; i++) { memset(&ucode_img, 0, sizeof(ucode_img)); if (pmu_acr_supp_ucode_list[i](g, &ucode_img) == 0) { if (ucode_img.lsf_desc != NULL) { @@ -520,7 +522,7 @@ int gm20b_pmu_populate_loader_cfg(struct gk20a *g, * physical addresses of each respective segment. */ addr_base = p_lsfm->lsb_header.ucode_off; - g->ops.pmu.get_wpr(g, &wpr_inf); + g->acr.get_wpr_info(g, &wpr_inf); addr_base += wpr_inf.wpr_base; nvgpu_pmu_dbg(g, "pmu loader cfg u32 addrbase %x\n", (u32)addr_base); /*From linux*/ @@ -596,7 +598,7 @@ int gm20b_flcn_populate_bl_dmem_desc(struct gk20a *g, * physical addresses of each respective segment. */ addr_base = p_lsfm->lsb_header.ucode_off; - g->ops.pmu.get_wpr(g, &wpr_inf); + g->acr.get_wpr_info(g, &wpr_inf); addr_base += wpr_inf.wpr_base; nvgpu_pmu_dbg(g, "gen loader cfg %x u32 addrbase %x ID\n", (u32)addr_base, @@ -874,7 +876,7 @@ static int lsfm_add_ucode_img(struct gk20a *g, struct ls_flcn_mgr *plsfm, /* Fill in static WPR header info*/ pnode->wpr_header.falcon_id = falcon_id; - pnode->wpr_header.bootstrap_owner = LSF_BOOTSTRAP_OWNER_DEFAULT; + pnode->wpr_header.bootstrap_owner = g->acr.bootstrap_owner; pnode->wpr_header.status = LSF_IMAGE_STATUS_COPY; pnode->wpr_header.lazy_bootstrap = @@ -1112,3 +1114,363 @@ void gm20b_update_lspmu_cmdline_args(struct gk20a *g) (u8 *)(g->ops.pmu_ver.get_pmu_cmdline_args_ptr(pmu)), g->ops.pmu_ver.get_pmu_cmdline_args_size(pmu), 0); } + +static int nvgpu_gm20b_acr_wait_for_completion(struct gk20a *g, + struct nvgpu_falcon *flcn, unsigned int timeout) +{ + u32 base_addr = flcn->flcn_base; + int completion = 0; + u32 data = 0; + + nvgpu_log_fn(g, " "); + + completion = nvgpu_flcn_wait_for_halt(flcn, timeout); + if (completion != 0U) { + nvgpu_err(g, "flcn-%d: ACR boot timed out", flcn->flcn_id); + goto exit; + } + + nvgpu_pmu_dbg(g, "flcn-%d: ACR capabilities %x\n", flcn->flcn_id, + nvgpu_flcn_mailbox_read(flcn, FALCON_MAILBOX_1)); + + data = nvgpu_flcn_mailbox_read(flcn, FALCON_MAILBOX_0); + if (data != 0U) { + nvgpu_err(g, "flcn-%d: ACR boot failed, err %x", flcn->flcn_id, + data); + completion = -EAGAIN; + goto exit; + } + + nvgpu_pmu_dbg(g, "flcn-%d: sctl reg %x", flcn->flcn_id, + gk20a_readl(g, base_addr + falcon_falcon_sctl_r())); + nvgpu_pmu_dbg(g, "flcn-%d: cpuctl reg %x", flcn->flcn_id, + gk20a_readl(g, base_addr + falcon_falcon_cpuctl_r())); + +exit: + return completion; +} + +static int gm20b_acr_hs_bl_exec(struct gk20a *g, struct nvgpu_acr *acr, + struct hs_acr *acr_desc, bool b_wait_for_halt) +{ + struct nvgpu_firmware *hs_bl_fw = acr_desc->acr_hs_bl.hs_bl_fw; + struct hsflcn_bl_desc *hs_bl_desc; + struct nvgpu_falcon_bl_info bl_info; + struct hs_flcn_bl *hs_bl = &acr_desc->acr_hs_bl; + struct mm_gk20a *mm = &g->mm; + struct vm_gk20a *vm = mm->pmu.vm; + u32 *hs_bl_code = NULL; + int err = 0; + u32 bl_sz; + + nvgpu_pmu_dbg(g, "Executing ACR HS Bootloader %s on Falcon-ID - %d", + hs_bl->bl_fw_name, acr_desc->acr_flcn->flcn_id); + + if (hs_bl_fw == NULL) { + hs_bl_fw = nvgpu_request_firmware(g, hs_bl->bl_fw_name, 0U); + if (hs_bl_fw == NULL) { + nvgpu_err(g, "ACR HS BL ucode load fail"); + return -ENOENT; + } + + hs_bl->hs_bl_fw = hs_bl_fw; + hs_bl->hs_bl_bin_hdr = (struct bin_hdr *)hs_bl_fw->data; + hs_bl->hs_bl_desc = (struct hsflcn_bl_desc *)(hs_bl_fw->data + + hs_bl->hs_bl_bin_hdr->header_offset); + + hs_bl_desc = hs_bl->hs_bl_desc; + hs_bl_code = (u32 *)(hs_bl_fw->data + + hs_bl->hs_bl_bin_hdr->data_offset); + + bl_sz = ALIGN(hs_bl_desc->bl_img_hdr.bl_code_size, 256U); + + hs_bl->hs_bl_ucode.size = bl_sz; + + err = nvgpu_dma_alloc_sys(g, bl_sz, &hs_bl->hs_bl_ucode); + if (err) { + nvgpu_err(g, "ACR HS BL failed to allocate memory"); + goto err_done; + } + + hs_bl->hs_bl_ucode.gpu_va = nvgpu_gmmu_map(vm, + &hs_bl->hs_bl_ucode, + bl_sz, + 0U, /* flags */ + gk20a_mem_flag_read_only, false, + hs_bl->hs_bl_ucode.aperture); + if (hs_bl->hs_bl_ucode.gpu_va == 0U) { + nvgpu_err(g, "ACR HS BL failed to map ucode memory!!"); + goto err_free_ucode; + } + + nvgpu_mem_wr_n(g, &hs_bl->hs_bl_ucode, 0U, hs_bl_code, bl_sz); + + nvgpu_pmu_dbg(g, "Copied BL ucode to bl_cpuva"); + } + + /* Fill HS BL info */ + bl_info.bl_src = hs_bl->hs_bl_ucode.cpu_va; + bl_info.bl_desc = acr_desc->ptr_bl_dmem_desc; + bl_info.bl_desc_size = acr_desc->bl_dmem_desc_size; + bl_info.bl_size = hs_bl->hs_bl_ucode.size; + bl_info.bl_start_tag = hs_bl->hs_bl_desc->bl_start_tag; + + /* + * 1. Dose falcon reset + * 2. setup falcon apertures + * 3. bootstrap falcon + */ + acr_desc->acr_flcn_setup_hw_and_bl_bootstrap(g, acr_desc, &bl_info); + + if (b_wait_for_halt) { + /* wait for ACR halt*/ + err = nvgpu_gm20b_acr_wait_for_completion(g, acr_desc->acr_flcn, + ACR_COMPLETION_TIMEOUT_MS); + if (err != 0U) { + goto err_unmap_bl; + } + } + + return 0U; +err_unmap_bl: + nvgpu_gmmu_unmap(vm, &hs_bl->hs_bl_ucode, hs_bl->hs_bl_ucode.gpu_va); +err_free_ucode: + nvgpu_dma_free(g, &hs_bl->hs_bl_ucode); +err_done: + nvgpu_release_firmware(g, hs_bl_fw); + + return err; +} + +int gm20b_acr_patch_wpr_info_to_ucode(struct gk20a *g, + struct nvgpu_acr *acr, struct hs_acr *acr_desc, bool is_recovery) +{ + struct nvgpu_firmware *acr_fw = acr_desc->acr_fw; + struct acr_fw_header *acr_fw_hdr = NULL; + struct bin_hdr *acr_fw_bin_hdr = NULL; + struct flcn_acr_desc *acr_dmem_desc; + u32 *acr_ucode_header = NULL; + u32 *acr_ucode_data = NULL; + + nvgpu_log_fn(g, " "); + + if (is_recovery) { + acr_desc->acr_dmem_desc->nonwpr_ucode_blob_size = 0U; + } else { + acr_fw_bin_hdr = (struct bin_hdr *)acr_fw->data; + acr_fw_hdr = (struct acr_fw_header *) + (acr_fw->data + acr_fw_bin_hdr->header_offset); + + acr_ucode_data = (u32 *)(acr_fw->data + + acr_fw_bin_hdr->data_offset); + + acr_ucode_header = (u32 *)(acr_fw->data + + acr_fw_hdr->hdr_offset); + + /* During recovery need to update blob size as 0x0*/ + acr_desc->acr_dmem_desc = (struct flcn_acr_desc *)((u8 *)( + acr_desc->acr_ucode.cpu_va) + acr_ucode_header[2U]); + + /* Patch WPR info to ucode */ + acr_dmem_desc = (struct flcn_acr_desc *) + &(((u8 *)acr_ucode_data)[acr_ucode_header[2U]]); + + acr_dmem_desc->nonwpr_ucode_blob_start = + nvgpu_mem_get_addr(g, &g->acr.ucode_blob); + acr_dmem_desc->nonwpr_ucode_blob_size = + g->acr.ucode_blob.size; + acr_dmem_desc->regions.no_regions = 1U; + acr_dmem_desc->wpr_offset = 0U; + } + + return 0; +} + +int gm20b_acr_fill_bl_dmem_desc(struct gk20a *g, + struct nvgpu_acr *acr, struct hs_acr *acr_desc, + u32 *acr_ucode_header) +{ + struct flcn_bl_dmem_desc *bl_dmem_desc = &acr_desc->bl_dmem_desc; + + nvgpu_log_fn(g, " "); + + memset(bl_dmem_desc, 0U, sizeof(struct flcn_bl_dmem_desc)); + + bl_dmem_desc->signature[0] = 0U; + bl_dmem_desc->signature[1] = 0U; + bl_dmem_desc->signature[2] = 0U; + bl_dmem_desc->signature[3] = 0U; + bl_dmem_desc->ctx_dma = GK20A_PMU_DMAIDX_VIRT; + bl_dmem_desc->code_dma_base = + (unsigned int)(((u64)acr_desc->acr_ucode.gpu_va >> 8U)); + bl_dmem_desc->code_dma_base1 = 0x0U; + bl_dmem_desc->non_sec_code_off = acr_ucode_header[0U]; + bl_dmem_desc->non_sec_code_size = acr_ucode_header[1U]; + bl_dmem_desc->sec_code_off = acr_ucode_header[5U]; + bl_dmem_desc->sec_code_size = acr_ucode_header[6U]; + bl_dmem_desc->code_entry_point = 0U; /* Start at 0th offset */ + bl_dmem_desc->data_dma_base = + bl_dmem_desc->code_dma_base + + ((acr_ucode_header[2U]) >> 8U); + bl_dmem_desc->data_dma_base1 = 0x0U; + bl_dmem_desc->data_size = acr_ucode_header[3U]; + + return 0; +} + +/* + * Loads ACR bin to SYSMEM/FB and bootstraps ACR with bootloader code + * start and end are addresses of ucode blob in non-WPR region + */ +int gm20b_bootstrap_hs_acr(struct gk20a *g, struct nvgpu_acr *acr, + struct hs_acr *acr_desc) +{ + struct mm_gk20a *mm = &g->mm; + struct vm_gk20a *vm = mm->pmu.vm; + struct nvgpu_firmware *acr_fw = acr_desc->acr_fw; + struct bin_hdr *acr_fw_bin_hdr = NULL; + struct acr_fw_header *acr_fw_hdr = NULL; + struct nvgpu_mem *acr_ucode_mem = &acr_desc->acr_ucode; + u32 img_size_in_bytes = 0; + u32 *acr_ucode_data; + u32 *acr_ucode_header; + u32 status = 0U; + + nvgpu_pmu_dbg(g, "ACR TYPE %x ", acr_desc->acr_type); + + if (acr_fw != NULL) { + acr->patch_wpr_info_to_ucode(g, acr, acr_desc, true); + } else { + acr_fw = nvgpu_request_firmware(g, acr_desc->acr_fw_name, + NVGPU_REQUEST_FIRMWARE_NO_SOC); + if (acr_fw == NULL) { + nvgpu_err(g, "%s ucode get fail for %s", + acr_desc->acr_fw_name, g->name); + return -ENOENT; + } + + acr_desc->acr_fw = acr_fw; + + acr_fw_bin_hdr = (struct bin_hdr *)acr_fw->data; + + acr_fw_hdr = (struct acr_fw_header *) + (acr_fw->data + acr_fw_bin_hdr->header_offset); + + acr_ucode_header = (u32 *)(acr_fw->data + + acr_fw_hdr->hdr_offset); + + acr_ucode_data = (u32 *)(acr_fw->data + + acr_fw_bin_hdr->data_offset); + + img_size_in_bytes = ALIGN((acr_fw_bin_hdr->data_size), 256U); + + /* Lets patch the signatures first.. */ + if (acr_ucode_patch_sig(g, acr_ucode_data, + (u32 *)(acr_fw->data + acr_fw_hdr->sig_prod_offset), + (u32 *)(acr_fw->data + acr_fw_hdr->sig_dbg_offset), + (u32 *)(acr_fw->data + acr_fw_hdr->patch_loc), + (u32 *)(acr_fw->data + acr_fw_hdr->patch_sig)) < 0) { + nvgpu_err(g, "patch signatures fail"); + status = -1; + goto err_release_acr_fw; + } + + status = nvgpu_dma_alloc_map_sys(vm, img_size_in_bytes, + acr_ucode_mem); + if (status != 0U) { + status = -ENOMEM; + goto err_release_acr_fw; + } + + acr->patch_wpr_info_to_ucode(g, acr, acr_desc, false); + + nvgpu_mem_wr_n(g, acr_ucode_mem, 0U, acr_ucode_data, + img_size_in_bytes); + + /* + * In order to execute this binary, we will be using + * a bootloader which will load this image into + * FALCON IMEM/DMEM. + * Fill up the bootloader descriptor to use.. + * TODO: Use standard descriptor which the generic bootloader is + * checked in. + */ + acr->acr_fill_bl_dmem_desc(g, acr, acr_desc, acr_ucode_header); + } + + status = gm20b_acr_hs_bl_exec(g, acr, acr_desc, true); + if (status != 0U) { + goto err_free_ucode_map; + } + + return 0; +err_free_ucode_map: + nvgpu_dma_unmap_free(vm, acr_ucode_mem); +err_release_acr_fw: + nvgpu_release_firmware(g, acr_fw); + acr_fw = NULL; + return status; +} + +void gm20b_remove_acr_support(struct nvgpu_acr *acr) +{ + struct gk20a *g = acr->g; + struct mm_gk20a *mm = &g->mm; + struct vm_gk20a *vm = mm->pmu.vm; + + if (acr->acr.acr_fw != NULL) { + nvgpu_release_firmware(g, acr->acr.acr_fw); + } + + if (acr->acr.acr_hs_bl.hs_bl_fw != NULL) { + nvgpu_release_firmware(g, acr->acr.acr_hs_bl.hs_bl_fw); + } + + nvgpu_dma_unmap_free(vm, &acr->acr.acr_ucode); + nvgpu_dma_unmap_free(vm, &acr->acr.acr_hs_bl.hs_bl_ucode); +} + +static void gm20b_acr_default_sw_init(struct gk20a *g, struct hs_acr *hs_acr) +{ + struct hs_flcn_bl *hs_bl = &hs_acr->acr_hs_bl; + + nvgpu_log_fn(g, " "); + + /* ACR HS bootloader ucode name */ + hs_bl->bl_fw_name = HSBIN_ACR_BL_UCODE_IMAGE; + + /* ACR HS ucode type & f/w name*/ + hs_acr->acr_type = ACR_DEFAULT; + hs_acr->acr_fw_name = HSBIN_ACR_UCODE_IMAGE; + + /* bootlader interface used by ACR HS bootloader*/ + hs_acr->ptr_bl_dmem_desc = &hs_acr->bl_dmem_desc; + hs_acr->bl_dmem_desc_size = sizeof(struct flcn_bl_dmem_desc); + + /* set on which falcon ACR need to execute*/ + hs_acr->acr_flcn = &g->pmu_flcn; + hs_acr->acr_flcn_setup_hw_and_bl_bootstrap = + gm20b_pmu_setup_hw_and_bl_bootstrap; +} + +void nvgpu_gm20b_acr_sw_init(struct gk20a *g, struct nvgpu_acr *acr) +{ + nvgpu_log_fn(g, " "); + + acr->g = g; + + acr->bootstrap_owner = LSF_FALCON_ID_PMU; + acr->max_supported_lsfm = MAX_SUPPORTED_LSFM; + + gm20b_acr_default_sw_init(g, &acr->acr); + + acr->get_wpr_info = gm20b_wpr_info; + acr->alloc_blob_space = gm20b_alloc_blob_space; + acr->bootstrap_hs_acr = gm20b_bootstrap_hs_acr; + acr->patch_wpr_info_to_ucode = + gm20b_acr_patch_wpr_info_to_ucode; + acr->acr_fill_bl_dmem_desc = + gm20b_acr_fill_bl_dmem_desc; + + acr->remove_support = gm20b_remove_acr_support; +} diff --git a/drivers/gpu/nvgpu/gm20b/acr_gm20b.h b/drivers/gpu/nvgpu/gm20b/acr_gm20b.h index 329d53b8..cae6ab6a 100644 --- a/drivers/gpu/nvgpu/gm20b/acr_gm20b.h +++ b/drivers/gpu/nvgpu/gm20b/acr_gm20b.h @@ -37,21 +37,32 @@ bool gm20b_is_pmu_supported(struct gk20a *g); int prepare_ucode_blob(struct gk20a *g); bool gm20b_is_lazy_bootstrap(u32 falcon_id); bool gm20b_is_priv_load(u32 falcon_id); -void gm20b_wpr_info(struct gk20a *g, struct wpr_carveout_info *inf); -int gm20b_alloc_blob_space(struct gk20a *g, size_t size, struct nvgpu_mem *mem); int gm20b_pmu_populate_loader_cfg(struct gk20a *g, void *lsfm, u32 *p_bl_gen_desc_size); int gm20b_flcn_populate_bl_dmem_desc(struct gk20a *g, void *lsfm, u32 *p_bl_gen_desc_size, u32 falconid); void gm20b_update_lspmu_cmdline_args(struct gk20a *g); void gm20b_setup_apertures(struct gk20a *g); - int gm20b_pmu_setup_sw(struct gk20a *g); int gm20b_init_nspmu_setup_hw1(struct gk20a *g); + int acr_ucode_patch_sig(struct gk20a *g, unsigned int *p_img, unsigned int *p_prod_sig, unsigned int *p_dbg_sig, unsigned int *p_patch_loc, unsigned int *p_patch_ind); +int gm20b_alloc_blob_space(struct gk20a *g, + size_t size, struct nvgpu_mem *mem); +void gm20b_wpr_info(struct gk20a *g, struct wpr_carveout_info *inf); +int gm20b_acr_patch_wpr_info_to_ucode(struct gk20a *g, + struct nvgpu_acr *acr, struct hs_acr *acr_desc, bool is_recovery); +int gm20b_acr_fill_bl_dmem_desc(struct gk20a *g, + struct nvgpu_acr *acr, struct hs_acr *acr_desc, + u32 *acr_ucode_header); +int gm20b_bootstrap_hs_acr(struct gk20a *g, struct nvgpu_acr *acr, + struct hs_acr *acr_desc); +void gm20b_remove_acr_support(struct nvgpu_acr *acr); +void nvgpu_gm20b_acr_sw_init(struct gk20a *g, struct nvgpu_acr *acr); + #endif /*NVGPU_GM20B_ACR_GM20B_H*/ diff --git a/drivers/gpu/nvgpu/gm20b/hal_gm20b.c b/drivers/gpu/nvgpu/gm20b/hal_gm20b.c index 39c902d7..52f86dab 100644 --- a/drivers/gpu/nvgpu/gm20b/hal_gm20b.c +++ b/drivers/gpu/nvgpu/gm20b/hal_gm20b.c @@ -656,6 +656,9 @@ static const struct gpu_ops gm20b_ops = { .read_vin_cal_slope_intercept_fuse = NULL, .read_vin_cal_gain_offset_fuse = NULL, }, + .acr = { + .acr_sw_init = nvgpu_gm20b_acr_sw_init, + }, .chip_init_gpu_characteristics = gk20a_init_gpu_characteristics, .get_litter_value = gm20b_get_litter_value, }; @@ -703,6 +706,8 @@ int gm20b_init_hal(struct gk20a *g) gops->fuse = gm20b_ops.fuse; + gops->acr = gm20b_ops.acr; + /* Lone functions */ gops->chip_init_gpu_characteristics = gm20b_ops.chip_init_gpu_characteristics; @@ -750,7 +755,6 @@ int gm20b_init_hal(struct gk20a *g) __nvgpu_set_enabled(g, NVGPU_PMU_FECS_BOOTSTRAP_DONE, false); g->pmu_lsf_pmu_wpr_init_done = 0; - g->bootstrap_owner = LSF_BOOTSTRAP_OWNER_DEFAULT; g->name = "gm20b"; diff --git a/drivers/gpu/nvgpu/gm20b/pmu_gm20b.c b/drivers/gpu/nvgpu/gm20b/pmu_gm20b.c index c1ec4d8e..6e764ac5 100644 --- a/drivers/gpu/nvgpu/gm20b/pmu_gm20b.c +++ b/drivers/gpu/nvgpu/gm20b/pmu_gm20b.c @@ -277,3 +277,63 @@ bool gm20b_pmu_is_debug_mode_en(struct gk20a *g) u32 ctl_stat = gk20a_readl(g, pwr_pmu_scpctl_stat_r()); return pwr_pmu_scpctl_stat_debug_mode_v(ctl_stat) != 0U; } + + +static int gm20b_bl_bootstrap(struct gk20a *g, + struct nvgpu_falcon_bl_info *bl_info) +{ + struct mm_gk20a *mm = &g->mm; + + nvgpu_log_fn(g, " "); + + gk20a_writel(g, pwr_falcon_itfen_r(), + gk20a_readl(g, pwr_falcon_itfen_r()) | + pwr_falcon_itfen_ctxen_enable_f()); + gk20a_writel(g, pwr_pmu_new_instblk_r(), + pwr_pmu_new_instblk_ptr_f( + nvgpu_inst_block_addr(g, &mm->pmu.inst_block) >> 12U) | + pwr_pmu_new_instblk_valid_f(1U) | + (nvgpu_is_enabled(g, NVGPU_USE_COHERENT_SYSMEM) ? + pwr_pmu_new_instblk_target_sys_coh_f() : + pwr_pmu_new_instblk_target_sys_ncoh_f())) ; + + nvgpu_flcn_bl_bootstrap(&g->pmu_flcn, bl_info); + + return 0; +} + +int gm20b_pmu_setup_hw_and_bl_bootstrap(struct gk20a *g, + struct hs_acr *acr_desc, + struct nvgpu_falcon_bl_info *bl_info) +{ + struct nvgpu_pmu *pmu = &g->pmu; + int err; + + nvgpu_log_fn(g, " "); + + nvgpu_mutex_acquire(&pmu->isr_mutex); + /* + * disable irqs for hs falcon booting + * as we will poll for halt + */ + g->ops.pmu.pmu_enable_irq(pmu, false); + pmu->isr_enabled = false; + err = nvgpu_flcn_reset(acr_desc->acr_flcn); + if (err != 0) { + nvgpu_mutex_release(&pmu->isr_mutex); + goto exit; + } + nvgpu_mutex_release(&pmu->isr_mutex); + + if (g->ops.pmu.setup_apertures) { + g->ops.pmu.setup_apertures(g); + } + + /*Clearing mailbox register used to reflect capabilities*/ + gk20a_writel(g, pwr_falcon_mailbox1_r(), 0); + + err = gm20b_bl_bootstrap(g, bl_info); + +exit: + return err; +} diff --git a/drivers/gpu/nvgpu/gm20b/pmu_gm20b.h b/drivers/gpu/nvgpu/gm20b/pmu_gm20b.h index f1b6cd93..37634783 100644 --- a/drivers/gpu/nvgpu/gm20b/pmu_gm20b.h +++ b/drivers/gpu/nvgpu/gm20b/pmu_gm20b.h @@ -34,5 +34,7 @@ void gm20b_pmu_load_lsf(struct gk20a *g, u32 falcon_id, u32 flags); int gm20b_pmu_init_acr(struct gk20a *g); void gm20b_write_dmatrfbase(struct gk20a *g, u32 addr); bool gm20b_pmu_is_debug_mode_en(struct gk20a *g); - +int gm20b_pmu_setup_hw_and_bl_bootstrap(struct gk20a *g, + struct hs_acr *acr_desc, + struct nvgpu_falcon_bl_info *bl_info); #endif /*NVGPU_GM20B_PMU_GM20B_H*/ -- cgit v1.2.2