nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-23 09:57:08 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: report MMU page fault errors to 3LSS

Browse Source 
This patch adds support to report MMU page fault errors to 3LSS.

JIRA NVGPU-3459

Change-Id: I3f06e594a75ae79bf4deef9acdc1829a002ea869
Signed-off-by: Rajesh Devaraj <rdevaraj@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2142742
GVS: Gerrit_Virtual_Submit
Reviewed-by: Antony Clince Alex <aalex@nvidia.com>
Reviewed-by: Ankur Kishore <ankkishore@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
This commit is contained in:
Rajesh Devaraj
2019-06-25 09:52:43 +05:30
committed by mobile promotions
parent b43fbafa93
commit cd4fa084c1
5 changed files with 60 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
drivers/gpu/nvgpu/hal/fb/fb_mmu_fault_gv11b_fusa.c
View File

@@ -581,6 +581,12 @@ void gv11b_fb_handle_mmu_fault(struct gk20a *g, u32 niso_intr)
if ((niso_intr &
fb_niso_intr_mmu_other_fault_notify_m()) != 0U) {
(void) nvgpu_report_mmu_err(g, NVGPU_ERR_MODULE_HUBMMU,
GPU_HUBMMU_PAGE_FAULT_ERROR,
NULL,
fault_status,
GPU_HUBMMU_OTHER_FAULT_NOTIFY);
gv11b_fb_handle_dropped_mmu_fault(g, fault_status);
gv11b_mm_mmu_fault_handle_other_fault_notify(g, fault_status);
@@ -604,6 +610,12 @@ void gv11b_fb_handle_mmu_fault(struct gk20a *g, u32 niso_intr)
if ((niso_intr &
fb_niso_intr_mmu_nonreplayable_fault_overflow_m()) != 0U) {
(void) nvgpu_report_mmu_err(g, NVGPU_ERR_MODULE_HUBMMU,
GPU_HUBMMU_PAGE_FAULT_ERROR,
NULL,
fault_status,
GPU_HUBMMU_NONREPLAYABLE_FAULT_OVERFLOW);
gv11b_fb_handle_nonreplay_fault_overflow(g,
fault_status);
}
@@ -622,6 +634,12 @@ void gv11b_fb_handle_mmu_fault(struct gk20a *g, u32 niso_intr)
if ((niso_intr &
fb_niso_intr_mmu_replayable_fault_overflow_m()) != 0U) {
(void) nvgpu_report_mmu_err(g, NVGPU_ERR_MODULE_HUBMMU,
GPU_HUBMMU_PAGE_FAULT_ERROR,
NULL,
fault_status,
GPU_HUBMMU_REPLAYABLE_FAULT_OVERFLOW);
gv11b_fb_handle_replay_fault_overflow(g,
fault_status);
}

13
drivers/gpu/nvgpu/hal/mm/mmu_fault/mmu_fault_gv11b_fusa.c
View File

@@ -449,6 +449,7 @@ void gv11b_mm_mmu_fault_handle_nonreplay_replay_fault(struct gk20a *g,
u32 invalidate_replay_val = 0U;
u64 prev_fault_addr = 0ULL;
u64 next_fault_addr = 0ULL;
u32 sub_err_type = 0U;
if (gv11b_fb_is_fault_buffer_empty(g, index, &get_indx)) {
nvgpu_log(g, gpu_dbg_intr,
@@ -481,6 +482,18 @@ void gv11b_mm_mmu_fault_handle_nonreplay_replay_fault(struct gk20a *g,
gv11b_fb_copy_from_hw_fault_buf(g, mem, offset, mmufault);
if (index == NVGPU_MMU_FAULT_REPLAY_REG_INDX) {
sub_err_type = GPU_HUBMMU_REPLAYABLE_FAULT_NOTIFY;
} else {
sub_err_type = GPU_HUBMMU_NONREPLAYABLE_FAULT_NOTIFY;
}
(void) nvgpu_report_mmu_err(g, NVGPU_ERR_MODULE_HUBMMU,
GPU_HUBMMU_PAGE_FAULT_ERROR,
mmufault,
fault_status,
sub_err_type);
nvgpu_assert(get_indx < U32_MAX);
get_indx = (get_indx + 1U) % entries;
nvgpu_log(g, gpu_dbg_intr, "new get index = %d", get_indx);

13
drivers/gpu/nvgpu/include/nvgpu/nvgpu_err.h
View File

@@ -26,6 +26,7 @@
#include <nvgpu/types.h>
struct gk20a;
struct mmu_fault_info;
#define NVGPU_ERR_MODULE_HOST (0U)
#define NVGPU_ERR_MODULE_SM (1U)
@@ -169,6 +170,14 @@ struct gr_exception_info {
#define GPU_HUBMMU_PTE_DATA_ECC_UNCORRECTED (5U)
#define GPU_HUBMMU_PDE0_DATA_ECC_CORRECTED (6U)
#define GPU_HUBMMU_PDE0_DATA_ECC_UNCORRECTED (7U)
#define GPU_HUBMMU_PAGE_FAULT_ERROR (8U)
/* Sub-errors in GPU_HUBMMU_PAGE_FAULT_ERROR */
#define GPU_HUBMMU_REPLAYABLE_FAULT_OVERFLOW (0U)
#define GPU_HUBMMU_REPLAYABLE_FAULT_NOTIFY (1U)
#define GPU_HUBMMU_NONREPLAYABLE_FAULT_OVERFLOW (2U)
#define GPU_HUBMMU_NONREPLAYABLE_FAULT_NOTIFY (3U)
#define GPU_HUBMMU_OTHER_FAULT_NOTIFY (4U)
#define GPU_PRI_TIMEOUT_ERROR (0U)
#define GPU_PRI_ACCESS_VIOLATION (1U)
@@ -227,4 +236,8 @@ int nvgpu_report_pmu_err(struct gk20a *g, u32 hw_unit, u32 err_id,
int nvgpu_report_pri_err(struct gk20a *g, u32 hw_unit, u32 inst,
u32 err_type, u32 err_addr, u32 err_code);
int nvgpu_report_mmu_err(struct gk20a *g, u32 hw_unit,
u32 err_type, struct mmu_fault_info *fault_info,
u32 status, u32 sub_err_type);
#endif

8
drivers/gpu/nvgpu/os/linux/sdl/sdl_stub.c
View File

@@ -23,6 +23,7 @@
#include <nvgpu/nvgpu_err.h>
struct gk20a;
struct mmu_fault_info;
int nvgpu_report_host_err(struct gk20a *g, u32 hw_unit,
u32 inst, u32 err_id, u32 intr_info)
@@ -65,3 +66,10 @@ int nvgpu_report_ctxsw_err(struct gk20a *g, u32 hw_unit, u32 err_id,
{
return 0;
}
int nvgpu_report_mmu_err(struct gk20a *g, u32 hw_unit,
u32 err_type, struct mmu_fault_info *fault_info,
u32 status, u32 sub_err_type)
{
return 0;
}

8
drivers/gpu/nvgpu/os/posix/stubs.c
View File

@@ -30,6 +30,7 @@
#include <nvgpu/nvgpu_err.h>
struct gk20a;
struct mmu_fault_info;
#ifdef CONFIG_NVGPU_DEBUGGER
void nvgpu_dbg_session_post_event(struct dbg_session_gk20a *dbg_s)
@@ -87,3 +88,10 @@ int nvgpu_report_ctxsw_err(struct gk20a *g, u32 hw_unit, u32 err_id,
{
return 0;
}
int nvgpu_report_mmu_err(struct gk20a *g, u32 hw_unit,
u32 err_type, struct mmu_fault_info *fault_info,
u32 status, u32 sub_err_type)
{
return 0;
}