nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-24 10:34:43 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: fix ecc counter free

Browse Source 
ECC counter structures are freed without removing the node from the
stats_list. This can lead to invalid access due to dangling pointers.

Update the ecc counter free logic to set them to NULL upon free, to
remove them from stats_list and free them by validation.

Also updated some of the ecc init paths where error was not propa-
gated to callers and full ecc counters deallocation was not done.

Now, calling unit ecc_free from any context (with counters alloc-
ated or not) is harmless as requisite checks are in place.

bug 3326612
bug 3345977

Change-Id: I05eb6ed226cff9197ad37776912da9dcb7e0716d
Signed-off-by: Sagar Kamble <skamble@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2565264
Tested-by: Ashish Mhetre <amhetre@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Reviewed-by: svc_kernel_abi <svc_kernel_abi@nvidia.com>
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
This commit is contained in:
Sagar Kamble
2021-06-17 11:34:36 +05:30
committed by mobile promotions
parent 2887d06e3b
commit 40064ef1ec
33 changed files with 546 additions and 218 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
drivers/gpu/nvgpu/include/nvgpu/ecc.h
View File

@@ -77,6 +77,7 @@
#include <nvgpu/types.h>
#include <nvgpu/list.h>
#include <nvgpu/lock.h>
#define NVGPU_ECC_STAT_NAME_MAX_SIZE 100UL
@@ -268,6 +269,8 @@ struct nvgpu_ecc {
/** Contains the head to the list of error statistics. */
struct nvgpu_list_node stats_list;
/** Lock to protect the stats_list updates. */
struct nvgpu_mutex stats_lock;
/** Contains the number of error statistics. */
int stats_count;
/**
@@ -281,29 +284,50 @@ struct nvgpu_ecc {
* @brief Allocates, initializes an error counter with specified name.
*
* @param g [in] The GPU driver struct.
* @param stat [out] Pointer to array of tpc error counters.
* @param statp [out] Pointer to error counter pointer.
* @param name [in] Unique name for error counter.
*
* Allocate memory for one error counter, initializes the counter with 0 and the
* specified string identifier. Finally the counter is added to the status_list
* specified string identifier. Finally the counter is added to the stats_list
* of struct nvgpu_ecc.
*
* @return 0 in case of success, less than 0 for failure.
* @return -ENOMEM if there is not enough memory to allocate ecc stats.
*/
int nvgpu_ecc_counter_init(struct gk20a *g,
struct nvgpu_ecc_stat **stat, const char *name);
struct nvgpu_ecc_stat **statp, const char *name);
/**
* @brief Concatenates the error counter to status list.
* @brief Deallocates an error counter.
*
* @param g [in] The GPU driver struct.
* @param statp [in] Pointer to error counter pointer.
*
* Delete the counter from the nvgpu_ecc stats_list. Deallocate memory for the
* error counter.
*/
void nvgpu_ecc_counter_deinit(struct gk20a *g, struct nvgpu_ecc_stat **statp);
/**
* @brief Concatenates the error counter to stats list.
*
* @param g [in] The GPU driver struct.
* @param stat [in] Pointer to error counter.
*
* The counter is added to the status_list of struct nvgpu_ecc.
* The counter is added to the stats_list of struct nvgpu_ecc.
*/
void nvgpu_ecc_stat_add(struct gk20a *g, struct nvgpu_ecc_stat *stat);
/**
* @brief Deletes the error counter from the stats list.
*
* @param g [in] The GPU driver struct.
* @param stat [in] Pointer to error counter.
*
* The counter is removed from the stats_list of struct nvgpu_ecc.
*/
void nvgpu_ecc_stat_del(struct gk20a *g, struct nvgpu_ecc_stat *stat);
/**
* @brief Release memory associated with all error counters.
*

21
drivers/gpu/nvgpu/include/nvgpu/gops/gr.h
View File

@@ -105,6 +105,27 @@ struct gops_gr_ecc {
*/
int (*fecs_ecc_init)(struct gk20a *g);
/**
* @brief Deinitialize GR unit ECC support.
*
* @param g [in] Pointer to GPU driver struct.
*
* This function deallocates memory to track the ecc error counts
* for GR unit and subunits of GR (like GPCs, TPCs etc) and removes
* it from global list.
*/
void (*gpc_tpc_ecc_deinit)(struct gk20a *g);
/**
* @brief Deinitialize GR unit ECC support.
*
* @param g [in] Pointer to GPU driver struct.
*
* This function deallocates memory to track the ecc error counts
* for FECS in GR and removes it from global list.
*/
void (*fecs_ecc_deinit)(struct gk20a *g);
/**
* @brief Detect ECC enabled units in GR engine.
*

69
drivers/gpu/nvgpu/include/nvgpu/gr/gr_ecc.h
View File

@@ -49,7 +49,7 @@ int nvgpu_ecc_counter_init_per_tpc(struct gk20a *g,
* @param stat [in] Address of pointer to struct nvgpu_ecc_stat.
*
*/
#define NVGPU_ECC_COUNTER_INIT_PER_TPC(stat) \
#define NVGPU_ECC_COUNTER_INIT_PER_TPC_OR_RETURN(stat) \
do { \
int err = 0; \
err = nvgpu_ecc_counter_init_per_tpc(g, \
@@ -59,6 +59,27 @@ int nvgpu_ecc_counter_init_per_tpc(struct gk20a *g,
} \
} while (false)
/**
* @brief Free error counter of all tpc instances in all gpc instances.
*
* @param g [in] The GPU driver struct.
* @param stats_p [out] Pointer to 2D array of error counters in tpcs in gpcs.
*
* Removes the error counter of all gpc instances from stats_list in struct
* nvgpu_ecc and frees the memory allocated for it.
*/
void nvgpu_ecc_counter_deinit_per_tpc(struct gk20a *g,
struct nvgpu_ecc_stat ***stats_p);
/*
* @brief Frees counters for memories shared across a TPCs in GPCs.
*
* @param stat [in] error counter member from g->ecc.gr.
*
*/
#define NVGPU_ECC_COUNTER_DEINIT_PER_TPC(stat) \
nvgpu_ecc_counter_deinit_per_tpc(g, &g->ecc.gr.stat)
/**
* @brief Allocate and initialize error counter specified by name for all gpc
* instances.
@@ -79,12 +100,33 @@ int nvgpu_ecc_counter_init_per_gpc(struct gk20a *g,
/*
* @brief Allocate and initialize counters for memories shared across a GPC.
*
* @param stat [in] Address of pointer to struct nvgpu_ecc_stat.
* @param stat [in] error counter member from g->ecc.gr.
*
*/
#define NVGPU_ECC_COUNTER_INIT_PER_GPC(stat) \
nvgpu_ecc_counter_init_per_gpc(g, &g->ecc.gr.stat, #stat)
/**
* @brief Free error counter of all gpc instances.
*
* @param g [in] The GPU driver struct.
* @param stats_p [out] Pointer to array of gpc error counters.
*
* Removes the error counter of all gpc instances from stats_list in struct
* nvgpu_ecc and frees the memory allocated for it.
*/
void nvgpu_ecc_counter_deinit_per_gpc(struct gk20a *g,
struct nvgpu_ecc_stat **stats_p);
/*
* @brief Frees counters for memories shared across a GPC.
*
* @param stat [in] error counter member from g->ecc.gr.
*
*/
#define NVGPU_ECC_COUNTER_DEINIT_PER_GPC(stat) \
nvgpu_ecc_counter_deinit_per_gpc(g, &g->ecc.gr.stat)
/**
* @brief Allocate and initialize error counter specified by name for all gr
* instances.
@@ -107,9 +149,30 @@ int nvgpu_ecc_counter_init_per_gr(struct gk20a *g,
* @param stat [in] Address of pointer to struct nvgpu_ecc_stat.
*
*/
#define NVGPU_ECC_COUNTER_INIT_GR(stat) \
#define NVGPU_ECC_COUNTER_INIT_PER_GR(stat) \
nvgpu_ecc_counter_init_per_gr(g, &g->ecc.gr.stat, #stat)
/**
* @brief Free error counter of all gr instances.
*
* @param g [in] The GPU driver struct.
* @param stats_p [out] Pointer to array of gr error counters.
*
* Removes the error counter of all gr instances from stats_list in struct
* nvgpu_ecc and frees the memory allocated for it.
*/
void nvgpu_ecc_counter_deinit_per_gr(struct gk20a *g,
struct nvgpu_ecc_stat **stats_p);
/*
* @brief Frees counters for memories shared across a GR instances.
*
* @param stat [in] error counter member from g->ecc.gr.
*
*/
#define NVGPU_ECC_COUNTER_DEINIT_PER_GR(stat) \
nvgpu_ecc_counter_deinit_per_gr(g, &g->ecc.gr.stat)
/**
* @brief Release all GR ECC stats counters.
*

11
drivers/gpu/nvgpu/include/nvgpu/pmu.h
View File

@@ -528,7 +528,7 @@ int nvgpu_pmu_early_init(struct gk20a *g);
void nvgpu_pmu_remove_support(struct gk20a *g, struct nvgpu_pmu *pmu);
/*
* @brief Allocate and initialize counter for memories within PMU.
* @brief Allocate and initialize ECC counter for memories within PMU.
*
* @param stat [in] Address of pointer to struct nvgpu_ecc_stat.
*
@@ -536,5 +536,14 @@ void nvgpu_pmu_remove_support(struct gk20a *g, struct nvgpu_pmu *pmu);
#define NVGPU_ECC_COUNTER_INIT_PMU(stat) \
nvgpu_ecc_counter_init(g, &g->ecc.pmu.stat, #stat)
/*
* @brief Remove ECC counter from the list and free the counter.
*
* @param stat [in] Address of pointer to struct nvgpu_ecc_stat.
*
*/
#define NVGPU_ECC_COUNTER_FREE_PMU(stat) \
nvgpu_ecc_counter_deinit(g, &g->ecc.pmu.stat)
#endif /* NVGPU_PMU_H */