Commit 3d72d2e5 authored by YiPeng Chai's avatar YiPeng Chai Committed by Alex Deucher
Browse files

drm/amdgpu: Intercept ras interrupts to ras module 



Intercept ras interrupts to ras module.

V2:
  Change function names in ras module.

Signed-off-by: default avatarYiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: default avatarTao Zhou <tao.zhou1@amd.com>
Reviewed-by: default avatarHawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent 4b6ec94f

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c

+15 −0
Original line number Diff line number Diff line
@@ -36,6 +36,7 @@ 
#include "amdgpu_ras.h"

#include "amdgpu_umc.h"

#include "amdgpu_reset.h"

#include "amdgpu_ras_mgr.h"



/* Total memory size in system memory and all GPU VRAM. Used to

 * estimate worst case amount of memory to reserve for page tables
@@ -746,6 +747,20 @@ void amdgpu_amdkfd_ras_pasid_poison_consumption_handler(struct amdgpu_device *ad 
				enum amdgpu_ras_block block, uint16_t pasid,

				pasid_notify pasid_fn, void *data, uint32_t reset)

{



	if (amdgpu_uniras_enabled(adev)) {

		struct ras_ih_info ih_info;



		memset(&ih_info, 0, sizeof(ih_info));

		ih_info.block = block;

		ih_info.pasid = pasid;

		ih_info.reset = reset;

		ih_info.pasid_fn = pasid_fn;

		ih_info.data = data;

		amdgpu_ras_mgr_handle_consumer_interrupt(adev, &ih_info);

		return;

	}



	amdgpu_umc_pasid_poison_handler(adev, block, pasid, pasid_fn, data, reset);

}

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c

+16 −0
Original line number Diff line number Diff line
@@ -41,6 +41,7 @@ 
#include "atom.h"

#include "amdgpu_reset.h"

#include "amdgpu_psp.h"

#include "amdgpu_ras_mgr.h"



#ifdef CONFIG_X86_MCE_AMD

#include <asm/mce.h>
@@ -2241,6 +2242,11 @@ void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev) 
	    amdgpu_ras_is_err_state(adev, AMDGPU_RAS_BLOCK__ANY))

		return;



	if (amdgpu_uniras_enabled(adev)) {

		amdgpu_ras_mgr_handle_fatal_interrupt(adev, NULL);

		return;

	}



	if (adev->nbio.ras &&

	    adev->nbio.ras->handle_ras_controller_intr_no_bifring)

		adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev);
@@ -2411,6 +2417,16 @@ int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev, 
	struct ras_manager *obj;

	struct ras_ih_data *data;



	if (amdgpu_uniras_enabled(adev)) {

		struct ras_ih_info ih_info;



		memset(&ih_info, 0, sizeof(ih_info));

		ih_info.block = info->head.block;

		memcpy(&ih_info.iv_entry, info->entry, sizeof(struct amdgpu_iv_entry));



		return amdgpu_ras_mgr_handle_controller_interrupt(adev, &ih_info);

	}



	obj = amdgpu_ras_find_obj(adev, &info->head);

	if (!obj)

		return -EINVAL;

drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c

+2 −2
Original line number Diff line number Diff line
@@ -381,7 +381,7 @@ static const struct amd_ip_funcs __maybe_unused ras_v1_0_ip_funcs = { 
	.hw_fini = amdgpu_ras_mgr_hw_fini,

};



int amdgpu_enable_unified_ras(struct amdgpu_device *adev, bool enable)

int amdgpu_enable_uniras(struct amdgpu_device *adev, bool enable)

{

	struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev);
@@ -395,7 +395,7 @@ int amdgpu_enable_unified_ras(struct amdgpu_device *adev, bool enable) 
	return ras_core_set_status(ras_mgr->ras_core, enable);

}



bool amdgpu_unified_ras_enabled(struct amdgpu_device *adev)

bool amdgpu_uniras_enabled(struct amdgpu_device *adev)

{

	struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev);

drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h

+2 −2
Original line number Diff line number Diff line
@@ -56,8 +56,8 @@ struct amdgpu_ras_mgr { 


struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(

			struct amdgpu_device *adev);

int amdgpu_enable_unified_ras(struct amdgpu_device *adev, bool enable);

bool amdgpu_unified_ras_enabled(struct amdgpu_device *adev);

int amdgpu_enable_uniras(struct amdgpu_device *adev, bool enable);

bool amdgpu_uniras_enabled(struct amdgpu_device *adev);

int amdgpu_ras_mgr_handle_fatal_interrupt(struct amdgpu_device *adev, void *data);

int amdgpu_ras_mgr_handle_controller_interrupt(struct amdgpu_device *adev, void *data);

int amdgpu_ras_mgr_handle_consumer_interrupt(struct amdgpu_device *adev, void *data);