drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switch
Subsystems (2)
PATCH 1/1 drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switch
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14 +++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22 +++++++++++++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +13 -1
@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct amdgpu_device *adev)if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control;
@@ -5859,3 +5866,8 @@ void amdgpu_ras_post_reset(struct amdgpu_device *adev,amdgpu_ras_mgr_post_reset (tmp_adev); } } + +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) +{ +amdgpu_ras_mgr_resume_after_reset (adev); +}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct amdgpu_device *adev, struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endif
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10 -0
@@ -1669,6 +1669,16 @@ static void amdgpu_xgmi_reset_on_init_work(struct work_struct *work)if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization" ); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +22 -0
@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)return 0 ; } +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev) +{ +struct amdgpu_ras *con =amdgpu_ras_get_context (adev); +struct amdgpu_ras_mgr *ras_mgr =amdgpu_ras_mgr_get_context (adev); +struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block =amdgpu_device_ip_get_ip_block (adev,AMD_IP_BLOCK_TYPE_RAS ); +if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); +} +struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context (struct amdgpu_device *adev) {if (!adev || !adev->psp.ras_context.ras)
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1 -0
@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct amdgpu_device *adev, void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_lookup_bad_pages_in_a_row (struct amdgpu_device *adev,uint64_t addr,uint64_t *nps_page_addr,uint32_t max_page_count);#endif
Signature · 1 line
2.43.0
Quoted 5 lines · Stanley.Yang
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
After reset, we set it to default level.
https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c#L5211
The default level includes all IP blocks, hence it's supposed to resume all blocks.
With RAS as a separate ip block, ideally it should have resumed. Could you check why/how RAS IP block is missed?
Thanks,
Lijo
Quoted 114 lines
Signed-off-by Stanley.Yang <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14 +++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22 +++++++++++++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +13
-1@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct amdgpu_device *adev)if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control;@@ -5859,3 +5866,8 @@ void amdgpu_ras_post_reset(struct amdgpu_device *adev,amdgpu_ras_mgr_post_reset (tmp_adev); } } + +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) +{ +amdgpu_ras_mgr_resume_after_reset (adev); +}drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1
-0@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct amdgpu_device *adev, struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endif drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10
-0@@ -1669,6 +1669,16 @@ static void amdgpu_xgmi_reset_on_init_work(struct work_struct *work)if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization" ); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +22
-0@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)return 0 ; } +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev) +{ +struct amdgpu_ras *con =amdgpu_ras_get_context (adev); +struct amdgpu_ras_mgr *ras_mgr =amdgpu_ras_mgr_get_context (adev); +struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block =amdgpu_device_ip_get_ip_block (adev,AMD_IP_BLOCK_TYPE_RAS ); +if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); +} +struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context (struct amdgpu_device *adev) {if (!adev || !adev->psp.ras_context.ras)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1
-0@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct amdgpu_device *adev, void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_lookup_bad_pages_in_a_row (struct amdgpu_device *adev,uint64_t addr,uint64_t *nps_page_addr,uint32_t max_page_count);#endif
AMD General
Quoted 18 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 4:16 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 5 lines · Stanley.Yang
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
After reset, we set it to default level.
Quoted 8 lines
https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amdg
pu/amdgpu_device.c#L5211The default level includes all IP blocks, hence it's supposed to resume all blocks.
With RAS as a separate ip block, ideally it should have resumed. Could you check why/how RAS IP block is missed?
static const struct amd_ip_funcs __maybe_unused ras_v1_0_ip_funcs = {
.name = "ras_v1_0",
.sw_init = amdgpu_ras_mgr_sw_init,
.sw_fini = amdgpu_ras_mgr_sw_fini,
.hw_init = amdgpu_ras_mgr_hw_init,
.hw_fini = amdgpu_ras_mgr_hw_fini,
};
The RAS IP block does not register suspend and resume callback function, so RAS IP block hw init function missed under this case.
Regards,
Stanley
Quoted 131 lines
Thanks,
LijoQuoted 6 lines
Signed-off-by Stanley.Yang <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14 +++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22+++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c int amdgpu_ras_init_badpage_info(struct
Quoted 9 lines
.../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +0
-0@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct
amdgpu_device*adev)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c void amdgpu_ras_post_reset(struct
Quoted 16 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +8
-1if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control;@@ -5859,3 +5866,8 @@ void amdgpu_ras_post_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h void amdgpu_ras_pre_reset(struct
Quoted 13 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +4
-0amdgpu_ras_mgr_post_reset (tmp_adev); } } + +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) { +amdgpu_ras_mgr_resume_after_reset (adev); +}drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +0
-0@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c static void
Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1
-0struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endif drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +0
-0@@ -1669,6 +1669,16 @@ static void
amdgpu_xgmi_reset_on_init_work(struct work_struct *work)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct
Quoted 22 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10
-0if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization" ); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct
amdgpu_ip_block*ip_block)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +3
-0return 0 ; } +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev) { +struct amdgpu_ras *con =amdgpu_ras_get_context (adev); + struct amdgpu_ras_mgr *ras_mgr =
amdgpu_ras_mgr_get_context(adev);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 12 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +12
-0+struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block = amdgpu_device_ip_get_ip_block(adev,
AMD_IP_BLOCK_TYPE_RAS);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +5
-0+if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); } + struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(struct
amdgpu_device*adev)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct
Quoted 8 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0{ if (!adev || !adev->psp.ras_context.ras) diff --git
a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hb/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hindex4f44a917d48b..3f80b9f1f0ac100644drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct
amdgpu_device*adev,drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct amdgpu_device *adev,
Quoted 5 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1
-0void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev); int amdgpu_ras_mgr_lookup_bad_pages_in_a_row(struct amdgpu_device*adev,
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct · line 87
Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0uint64_t addr, uint64_t *nps_page_addr, uint32_t
max_page_count);Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0#endif
Quoted 37 lines · Stanley
AMD General
Quoted 18 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 4:16 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 5 lines · Stanley.Yang
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
After reset, we set it to default level.
Quoted 8 lines
https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amdg
pu/amdgpu_device.c#L5211The default level includes all IP blocks, hence it's supposed to resume all blocks.
With RAS as a separate ip block, ideally it should have resumed. Could you check why/how RAS IP block is missed?
static const struct
amd_ip_funcs__maybe_unusedras_v1_0_ip_funcs= {
.name = "ras_v1_0",
.sw_init =amdgpu_ras_mgr_sw_init,
.sw_fini =amdgpu_ras_mgr_sw_fini,
.hw_init =amdgpu_ras_mgr_hw_init,
.hw_fini =amdgpu_ras_mgr_hw_fini,
};
Quoted 2 lines
The RAS IP block does not register suspend and resume callback function, so RAS IP block hw init function missed under this case.
How does it work for regular reset? For regular reset also, driver calls the resume sequence for the IP block.
I think the proper fix is to add a resume sequence which calls hw_init (if those paths are identical).
Thanks,
Lijo
Quoted 134 lines
Regards,
StanleyQuoted 131 lines
Thanks,
LijoQuoted 6 lines
Signed-off-by Stanley.Yang <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14 +++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22+++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c int amdgpu_ras_init_badpage_info(struct
Quoted 9 lines
.../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +0
-0@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct
amdgpu_device*adev)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c void amdgpu_ras_post_reset(struct
Quoted 16 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +8
-1if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control;@@ -5859,3 +5866,8 @@ void amdgpu_ras_post_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h void amdgpu_ras_pre_reset(struct
Quoted 13 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +4
-0amdgpu_ras_mgr_post_reset (tmp_adev); } } + +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) { +amdgpu_ras_mgr_resume_after_reset (adev); +}drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +0
-0@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c static void
Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1
-0struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endif drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +0
-0@@ -1669,6 +1669,16 @@ static void
amdgpu_xgmi_reset_on_init_work(struct work_struct *work)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct
Quoted 22 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10
-0if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization" ); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct
amdgpu_ip_block*ip_block)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +3
-0return 0 ; } +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev) { +struct amdgpu_ras *con =amdgpu_ras_get_context (adev); + struct amdgpu_ras_mgr *ras_mgr =
amdgpu_ras_mgr_get_context(adev);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 12 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +12
-0+struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block = amdgpu_device_ip_get_ip_block(adev,
AMD_IP_BLOCK_TYPE_RAS);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +5
-0+if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); } + struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(struct
amdgpu_device*adev)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct
Quoted 8 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0{ if (!adev || !adev->psp.ras_context.ras) diff --git
a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hb/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hindex4f44a917d48b..3f80b9f1f0ac100644drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct
amdgpu_device*adev,drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct amdgpu_device *adev,
Quoted 5 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1
-0void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev); int amdgpu_ras_mgr_lookup_bad_pages_in_a_row(struct amdgpu_device*adev,
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct · line 87
Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0uint64_t addr, uint64_t *nps_page_addr, uint32_t
max_page_count);Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0#endif
AMD General
Quoted 50 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 5:44 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 22 lines · Stanley
AMD General
Quoted 20 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 4:16 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 5 lines · Stanley.Yang
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
After reset, we set it to default level.
https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amd
Quoted 9 lines
Quoted 8 lines
g
pu/amdgpu_device.c#L5211The default level includes all IP blocks, hence it's supposed to resume all blocks.
With RAS as a separate ip block, ideally it should have resumed. Could you check why/how RAS IP block is missed?
Quoted 8 lines
static const struct
amd_ip_funcs__maybe_unusedras_v1_0_ip_funcs= {
.name = "ras_v1_0",
.sw_init =amdgpu_ras_mgr_sw_init,
.sw_fini =amdgpu_ras_mgr_sw_fini,
.hw_init =amdgpu_ras_mgr_hw_init,
.hw_fini =amdgpu_ras_mgr_hw_fini, };The RAS IP block does not register suspend and resume callback function, so
Quoted 8 lines
RAS IP block hw init function missed under this case.
Quoted 1 line
How does it work for regular reset? For regular reset also, driver calls the resume sequence for the IP block.
I think the proper fix is to add a resume sequence which calls
hw_init(if those paths are identical).
[Stanley]: Normal cold start is at the DEFAULT init level, and RAS hw_init is executing normally, so uniras is enabled; Normal GPU reset, because there is no. suspend, ras_is_read remains true and the state is preserved - so only NPS switching to this XGMI reset on int path will trigger this bug.
Not suggest add .resume sequence because during RAS IP block hw_init process need hold reset domain semaphore but the semaphore has already hold before, the other reason is adding .resume will break regular reset, it doesn't need to be reinitialized due to all RAS error info is cached.
Regards,
Stanley
Quoted 141 lines
Thanks,
LijoQuoted 131 lines
Regards,
StanleyQuoted 129 lines
Thanks,
LijoQuoted 6 lines
Signed-off-by Stanley.Yang <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14 +++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22+++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c int amdgpu_ras_init_badpage_info(struct
Quoted 9 lines
.../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +0
-0@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct
amdgpu_device*adev)Quoted 15 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +8
-1if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control; @@ -5859,3 +5866,8 @@ voidQuoted 1 line
amdgpu_ras_post_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h void amdgpu_ras_pre_reset(struct
Quoted 13 lines
amdgpu_ras_mgr_post_reset(tmp_adev); }}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +4
-0+ +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) { +amdgpu_ras_mgr_resume_after_reset (adev); +}drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +0
-0@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c static void
Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1
-0struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endifdrivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +0
-0@@ -1669,6 +1669,16 @@ static void
amdgpu_xgmi_reset_on_init_work(struct work_struct *work)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct
Quoted 23 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10
-0if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization"); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct
amdgpu_ip_block*ip_block)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +3
-0return 0 ; } +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev) { +struct amdgpu_ras *con =amdgpu_ras_get_context (adev); + struct amdgpu_ras_mgr *ras_mgr =
amdgpu_ras_mgr_get_context(adev);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 12 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +12
-0+struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block = amdgpu_device_ip_get_ip_block(adev,
AMD_IP_BLOCK_TYPE_RAS);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +5
-0+if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); } + struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(struct
amdgpu_device*adev)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct
Quoted 8 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0{ if (!adev || !adev->psp.ras_context.ras) diff --git
a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hb/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hindex4f44a917d48b..3f80b9f1f0ac100644drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct
amdgpu_device*adev,Quoted 5 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1
-0void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev); int amdgpu_ras_mgr_lookup_bad_pages_in_a_row(struct
amdgpu_deviceQuoted 5 lines
Quoted 4 lines
*adev,
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct · line 87
Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0uint64_t addr, uint64_t *nps_page_addr, uint32_t
max_page_count);Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0#endif
Quoted 58 lines · Stanley
AMD General
Quoted 56 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 5:44 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 22 lines · Stanley
AMD General
Quoted 20 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 4:16 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 5 lines · Stanley.Yang
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
After reset, we set it to default level.
https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amd
Quoted 9 lines
Quoted 8 lines
g
pu/amdgpu_device.c#L5211The default level includes all IP blocks, hence it's supposed to resume all blocks.
With RAS as a separate ip block, ideally it should have resumed. Could you check why/how RAS IP block is missed?
Quoted 8 lines
static const struct
amd_ip_funcs__maybe_unusedras_v1_0_ip_funcs= {
.name = "ras_v1_0",
.sw_init =amdgpu_ras_mgr_sw_init,
.sw_fini =amdgpu_ras_mgr_sw_fini,
.hw_init =amdgpu_ras_mgr_hw_init,
.hw_fini =amdgpu_ras_mgr_hw_fini, };The RAS IP block does not register suspend and resume callback function, so
RAS IP block hw init function missed under this case.
Quoted 1 line
How does it work for regular reset? For regular reset also, driver calls the resume sequence for the IP block.
Quoted 7 lines
Quoted 2 lines
I think the proper fix is to add a resume sequence which calls
hw_init(if those paths are identical).[Stanley]: Normal cold start is at the DEFAULT init level, and RAS
hw_initis executing normally, so uniras is enabled; Normal GPU reset, because there is no. suspend,ras_is_readremains true and the state is preserved - so only NPS switching to this XGMI reset on int path will trigger this bug.Not suggest add .resume sequence because during RAS IP block
hw_initprocess need hold reset domain semaphore but the semaphore has already hold before, the other reason is adding .resume will break regular reset, it doesn't need to be reinitialized due to all RAS error info is cached.
Then a dedicated resume sequence needs to be added which checks if device is in reset (being resumed from a reset) before taking hold of the semaphore.
As an IP block, it will need suspend/resume for other scenarios like device runtime pm or system suspend/resume. This can be used only as a temporary workaround till that point.
Thanks,
Lijo
Quoted 144 lines
Regards,
StanleyQuoted 141 lines
Thanks,
LijoQuoted 131 lines
Regards,
StanleyQuoted 129 lines
Thanks,
LijoQuoted 6 lines
Signed-off-by Stanley.Yang <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14 +++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22+++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c int amdgpu_ras_init_badpage_info(struct
Quoted 9 lines
.../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +0
-0@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct
amdgpu_device*adev)Quoted 15 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +8
-1if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control; @@ -5859,3 +5866,8 @@ voidQuoted 1 line
amdgpu_ras_post_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h void amdgpu_ras_pre_reset(struct
Quoted 13 lines
amdgpu_ras_mgr_post_reset(tmp_adev); } }drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +4
-0+ +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) { +amdgpu_ras_mgr_resume_after_reset (adev); +}drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +0
-0@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c static void
Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1
-0struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endifdrivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +0
-0@@ -1669,6 +1669,16 @@ static void
amdgpu_xgmi_reset_on_init_work(struct work_struct *work)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct
Quoted 23 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10
-0if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization"); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct
amdgpu_ip_block*ip_block)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +3
-0return 0 ; } +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev) { +struct amdgpu_ras *con =amdgpu_ras_get_context (adev); + struct amdgpu_ras_mgr *ras_mgr =
amdgpu_ras_mgr_get_context(adev);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 12 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +12
-0+struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block = amdgpu_device_ip_get_ip_block(adev,
AMD_IP_BLOCK_TYPE_RAS);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +5
-0+if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); } + struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(struct
amdgpu_device*adev)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct
Quoted 8 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0{ if (!adev || !adev->psp.ras_context.ras) diff --git
a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hb/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hindex4f44a917d48b..3f80b9f1f0ac100644drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct
amdgpu_device*adev,Quoted 5 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1
-0void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev); int amdgpu_ras_mgr_lookup_bad_pages_in_a_row(struct
amdgpu_deviceQuoted 5 lines
Quoted 4 lines
*adev,
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct · line 87
Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0uint64_t addr, uint64_t *nps_page_addr, uint32_t
max_page_count);Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0#endif
AMD General
Quoted 79 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 7:37 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 36 lines · Stanley
AMD General
Quoted 34 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 5:44 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 22 lines · Stanley
AMD General
Quoted 20 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 4:16 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 5 lines · Stanley.Yang
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
After reset, we set it to default level.
https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amd
Quoted 21 lines
Quoted 21 lines
Quoted 16 lines
Quoted 8 lines
g
pu/amdgpu_device.c#L5211The default level includes all IP blocks, hence it's supposed to resume all blocks.
With RAS as a separate ip block, ideally it should have resumed. Could you check why/how RAS IP block is missed?
static const struct
amd_ip_funcs__maybe_unusedras_v1_0_ip_funcs= {
.name = "ras_v1_0",
.sw_init =amdgpu_ras_mgr_sw_init,
.sw_fini =amdgpu_ras_mgr_sw_fini,
.hw_init =amdgpu_ras_mgr_hw_init,
.hw_fini =amdgpu_ras_mgr_hw_fini, };Quoted 2 lines
The RAS IP block does not register suspend and resume callback function, so
RAS IP block hw init function missed under this case.
Quoted 1 line
Quoted 7 lines
Quoted 5 lines
How does it work for regular reset? For regular reset also, driver calls the resume sequence for the IP block.
I think the proper fix is to add a resume sequence which calls
hw_init(if those paths are identical).[Stanley]: Normal cold start is at the DEFAULT init level, and RAS
hw_initisexecuting normally, so uniras is enabled; Normal GPU reset, because there is no. suspend,
ras_is_readremains true and the state is preserved - so only NPS switching to this XGMI reset on int path will trigger this bug.Quoted 1 line
Quoted 8 lines
Quoted 1 line
Not suggest add .resume sequence because during RAS IP block
hw_initprocess need hold reset domain semaphore but the semaphore has already hold before, the other reason is adding .resume will break regular reset, it doesn't need to be reinitialized due to all RAS error info is cached.
Quoted 1 line
Then a dedicated resume sequence needs to be added which checks if device is in reset (being resumed from a reset) before taking hold of the semaphore.
Yes, controlling the semaphore by checking if it is in the reset state can indeed solve the semaphore problem, but as I said there is no need to suspend and then resume the RAS IP block in regular reset.
Regards,
Stanley
Quoted 155 lines
As an IP block, it will need suspend/resume for other scenarios like device runtime pm or system suspend/resume. This can be used only as a temporary workaround till that point.
Thanks,
LijoQuoted 14 lines
Regards,
StanleyQuoted 12 lines
Thanks,
LijoQuoted 8 lines
Regards,
StanleyQuoted 6 lines
Thanks,
LijoQuoted 3 lines
Signed-off-by Stanley.Yang <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14+++++++++++-
Quoted 82 lines
Quoted 82 lines
Quoted 82 lines
Quoted 82 lines
Quoted 3 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22+++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c int amdgpu_ras_init_badpage_info(struct
Quoted 9 lines
.../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +0
-0@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct
amdgpu_device*adev)Quoted 15 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +8
-1if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control; @@ -5859,3 +5866,8 @@ voidQuoted 1 line
amdgpu_ras_post_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h void amdgpu_ras_pre_reset(struct
Quoted 13 lines
amdgpu_ras_mgr_post_reset(tmp_adev); } }drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +4
-0+ +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) { +amdgpu_ras_mgr_resume_after_reset (adev); +}drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +0
-0@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c static void
Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1
-0struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endifdrivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +0
-0@@ -1669,6 +1669,16 @@ static void
amdgpu_xgmi_reset_on_init_work(struct work_struct *work)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct
Quoted 23 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10
-0if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization"); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct
amdgpu_ip_block*ip_block)Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 4 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +1
-0return 0 ; } +int amdgpu_ras_mgr_resume_after_reset(struct amdgpu_device*adev) {
Quoted 45 lines
Quoted 44 lines
Quoted 32 lines
Quoted 32 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct · line 470
Quoted 2 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +2
-0+struct amdgpu_ras *con =amdgpu_ras_get_context (adev); + struct amdgpu_ras_mgr *ras_mgr =
amdgpu_ras_mgr_get_context(adev);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 12 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +12
-0+struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block = amdgpu_device_ip_get_ip_block(adev,
AMD_IP_BLOCK_TYPE_RAS);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +5
-0+if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); } + struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(struct
amdgpu_device*adev)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct
Quoted 8 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0{ if (!adev || !adev->psp.ras_context.ras) diff --git
a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hb/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hindex4f44a917d48b..3f80b9f1f0ac100644drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct
amdgpu_device*adev,Quoted 6 lines
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct amdgpu_device *adev,
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +2
-0void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device +*adev); int amdgpu_ras_mgr_lookup_bad_pages_in_a_row(struct
amdgpu_deviceQuoted 5 lines
Quoted 4 lines
*adev,
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct · line 88
Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0uint64_t addr, uint64_t *nps_page_addr, uint32_t
max_page_count);Quoted 1 line
#endif
Lazar, Lijo
Reviewed-byQuoted 87 lines · Stanley
AMD General
Quoted 85 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 7:37 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 36 lines · Stanley
AMD General
Quoted 34 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 5:44 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 22 lines · Stanley
AMD General
Quoted 20 lines
-----Original Message-----
From: Lazar, Lijo <[email protected]>
Sent: Friday, June 26, 2026 4:16 PM
To: Yang, Stanley <[email protected]>; [email protected]
Subject: Re: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switchQuoted 5 lines · Stanley.Yang
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
After reset, we set it to default level.
https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amd
Quoted 21 lines
Quoted 21 lines
Quoted 16 lines
Quoted 8 lines
g
pu/amdgpu_device.c#L5211The default level includes all IP blocks, hence it's supposed to resume all blocks.
With RAS as a separate ip block, ideally it should have resumed. Could you check why/how RAS IP block is missed?
static const struct
amd_ip_funcs__maybe_unusedras_v1_0_ip_funcs= {
.name = "ras_v1_0",
.sw_init =amdgpu_ras_mgr_sw_init,
.sw_fini =amdgpu_ras_mgr_sw_fini,
.hw_init =amdgpu_ras_mgr_hw_init,
.hw_fini =amdgpu_ras_mgr_hw_fini, };Quoted 2 lines
The RAS IP block does not register suspend and resume callback function, so
RAS IP block hw init function missed under this case.
Quoted 1 line
Quoted 7 lines
Quoted 5 lines
How does it work for regular reset? For regular reset also, driver calls the resume sequence for the IP block.
I think the proper fix is to add a resume sequence which calls
hw_init(if those paths are identical).[Stanley]: Normal cold start is at the DEFAULT init level, and RAS
hw_initisexecuting normally, so uniras is enabled; Normal GPU reset, because there is no. suspend,
ras_is_readremains true and the state is preserved - so only NPS switching to this XGMI reset on int path will trigger this bug.Quoted 2 lines
Not suggest add .resume sequence because during RAS IP block
hw_initprocess need hold reset domain semaphore but the semaphore has already hold before, the other reason is adding .resume will break regular reset, it doesn't need to be reinitialized due to all RAS error info is cached.
Quoted 1 line
Quoted 5 lines
Quoted 2 lines
Then a dedicated resume sequence needs to be added which checks if device is in reset (being resumed from a reset) before taking hold of the semaphore.
Yes, controlling the semaphore by checking if it is in the reset state can indeed solve the semaphore problem, but as I said there is no need to suspend and then resume the RAS IP block in regular reset.
Right. Eventually ras block also will need suspend/resume sequence like others for other cases like S3 suspend/resume (or during VF migration sequence). It's better to add a TODO comment (remove this once suspend/resume sequence is added or similar) in the existing code.
With that change
Thanks,
Lijo
Quoted 158 lines
Regards,
StanleyQuoted 155 lines
As an IP block, it will need suspend/resume for other scenarios like device runtime pm or system suspend/resume. This can be used only as a temporary workaround till that point.
Thanks,
LijoQuoted 14 lines
Regards,
StanleyQuoted 12 lines
Thanks,
LijoQuoted 8 lines
Regards,
StanleyQuoted 6 lines
Thanks,
LijoQuoted 3 lines
Signed-off-by Stanley.Yang <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14+++++++++++-
Quoted 82 lines
Quoted 82 lines
Quoted 82 lines
Quoted 82 lines
Quoted 3 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22+++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c int amdgpu_ras_init_badpage_info(struct
Quoted 9 lines
.../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +0
-0@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct
amdgpu_device*adev)Quoted 15 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +8
-1if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control; @@ -5859,3 +5866,8 @@ voidQuoted 1 line
amdgpu_ras_post_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h void amdgpu_ras_pre_reset(struct
Quoted 13 lines
amdgpu_ras_mgr_post_reset(tmp_adev); } }drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +4
-0+ +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) { +amdgpu_ras_mgr_resume_after_reset (adev); +}drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +0
-0@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct
amdgpu_device*adev,drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c static void
Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1
-0struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endifdrivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +0
-0@@ -1669,6 +1669,16 @@ static void
amdgpu_xgmi_reset_on_init_work(struct work_struct *work)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct
Quoted 23 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10
-0if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization"); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct
amdgpu_ip_block*ip_block)Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct · line 468
Quoted 4 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +1
-0return 0 ; } +int amdgpu_ras_mgr_resume_after_reset(struct amdgpu_device*adev) {
Quoted 45 lines
Quoted 44 lines
Quoted 32 lines
Quoted 32 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct · line 470
Quoted 2 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +2
-0+struct amdgpu_ras *con =amdgpu_ras_get_context (adev); + struct amdgpu_ras_mgr *ras_mgr =
amdgpu_ras_mgr_get_context(adev);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct · line 482
Quoted 12 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +12
-0+struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block = amdgpu_device_ip_get_ip_block(adev,
AMD_IP_BLOCK_TYPE_RAS);drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c static int amdgpu_ras_mgr_hw_fini(struct · line 488
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +5
-0+if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); } + struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(struct
amdgpu_device*adev)drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct
Quoted 8 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +0
-0{ if (!adev || !adev->psp.ras_context.ras) diff --git
a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hb/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.hindex4f44a917d48b..3f80b9f1f0ac100644drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct
amdgpu_device*adev,Quoted 6 lines
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct · line 87
Quoted 6 lines
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +2
-0void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device +*adev); int amdgpu_ras_mgr_lookup_bad_pages_in_a_row(struct
amdgpu_deviceQuoted 5 lines
Quoted 4 lines
*adev,
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h int amdgpu_ras_mgr_handle_ras_cmd(struct · line 88
Quoted 1 line
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0
-0uint64_t addr, uint64_t *nps_page_addr, uint32_t
max_page_count);Quoted 1 line
#endif
AMD General
Regards,
Hawking
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Stanley.Yang
Sent: Friday, June 26, 2026 3:18 PM
To: [email protected]
Cc: Yang, Stanley <[email protected]>
Subject: [PATCH 1/1] drm/amdgpu/ras: Resum RAS IP hw init during nps dynamic switch
On an XGMI reset-on-init (NPS memory patition mode swith), RAS IP hw fini, sw fini is called but hw init is skipped due to RAS IP block is not included in hwinit mask, so need call RAS IP hw init during XGMI reset-on-init.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 14 +++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 10 +++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 22 +++++++++++++++++++ .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h | 1 + 5 files changed, 47 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +12 -1
@@ -3857,7 +3857,14 @@ int amdgpu_ras_init_badpage_info(struct amdgpu_device *adev)if (!con ||amdgpu_sriov_vf (adev))return 0 ; -if (amdgpu_uniras_enabled (adev)) +/* +* For the reset-on-init path (e.g. an NPS memory partition, +* switch) the RAS IP block hw_init has not been enabled and +* the amdgpu_uniras_enabled return false, check amdgpu ras +* context uniras_enabled flag, eepron init will be called +* during RAS IP block hw_init. +*/ +if (amdgpu_uniras_enabled (adev) || con->uniras_enabled)return 0 ; control = &con->eeprom_control;
@@ -5859,3 +5866,8 @@ void amdgpu_ras_post_reset(struct amdgpu_device *adev,amdgpu_ras_mgr_post_reset (tmp_adev); } } + +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev) { +amdgpu_ras_mgr_resume_after_reset (adev); +}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
@@ -1045,4 +1045,5 @@ void amdgpu_ras_pre_reset(struct amdgpu_device *adev, struct list_head *device_list);void amdgpu_ras_post_reset (struct amdgpu_device *adev,struct list_head *device_list); +void amdgpu_ras_resume_after_reset (struct amdgpu_device *adev);#endif
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +10 -0
@@ -1669,6 +1669,16 @@ static void amdgpu_xgmi_reset_on_init_work(struct work_struct *work)if (r && r != -EHWPOISON )dev_err (tmp_adev->dev,"error during bad page data initialization" ); + +/* +* For the reset-on-init path (e.g. an NPS memory partition +* switch) the RAS IP block hw_init was skipped under the +* minimal init level, so uniras was never enabled. Bring it +* up now that the reset domain has been unlocked. This is a +* no-op for any other reset path where RAS is already +* initialized, and for non-uniras devices. +*/ +amdgpu_ras_resume_after_reset (tmp_adev); } }
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +20 -0
@@ -465,6 +465,28 @@ static int amdgpu_ras_mgr_hw_fini(struct amdgpu_ip_block *ip_block)return 0 ; } +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev) { +struct amdgpu_ras *con =amdgpu_ras_get_context (adev); +struct amdgpu_ras_mgr *ras_mgr =amdgpu_ras_mgr_get_context (adev); +struct amdgpu_ip_block *ip_block; + +if (!con || !con->uniras_enabled) +return 0 ; + +if (!ras_mgr || !ras_mgr->ras_core) +return -EINVAL ; + +if (ras_mgr->ras_is_ready) +return 0 ; + + ip_block =amdgpu_device_ip_get_ip_block (adev,AMD_IP_BLOCK_TYPE_RAS ); +if (!ip_block) +return -EINVAL ; + +return amdgpu_ras_mgr_hw_init (ip_block); } +struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context (struct amdgpu_device *adev) { if (!adev || !adev->psp.ras_context.ras) diff --git a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h b/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h
index 4f44a917d48b..3f80b9f1f0ac 100644
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1 -1
@@ -82,6 +82,7 @@ int amdgpu_ras_mgr_handle_ras_cmd(struct amdgpu_device *adev, void *output, uint32_t out_size);int amdgpu_ras_mgr_pre_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_post_reset (struct amdgpu_device *adev); +int amdgpu_ras_mgr_resume_after_reset (struct amdgpu_device *adev);int amdgpu_ras_mgr_lookup_bad_pages_in_a_row (struct amdgpu_device *adev,uint64_t addr,uint64_t *nps_page_addr,uint32_t max_page_count);#endif --
2.43.0