RE: [PATCH 1/2] drm/amdgpu: Add fatal error detected flag

["Kamal, Asad" <Asad.Kamal@xxxxxxx>]

[AMD Official Use Only - General]

Series is Reviewed-by: Asad Kamal <asad.kamal@xxxxxxx>

Thanks & Regards
Asad

-----Original Message-----
From: Lazar, Lijo <Lijo.Lazar@xxxxxxx>
Sent: Thursday, February 22, 2024 3:47 PM
To: amd-gfx@xxxxxxxxxxxxxxxxxxxxx
Cc: Zhang, Hawking <Hawking.Zhang@xxxxxxx>; Deucher, Alexander <Alexander.Deucher@xxxxxxx>; Kuehling, Felix <Felix.Kuehling@xxxxxxx>; Joshi, Mukul <Mukul.Joshi@xxxxxxx>; Kamal, Asad <Asad.Kamal@xxxxxxx>
Subject: [PATCH 1/2] drm/amdgpu: Add fatal error detected flag

For a RAS error that needs a full reset to recover, set the fatal error status. Clear the status once the device is reset.

Signed-off-by: Lijo Lazar <lijo.lazar@xxxxxxx>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 32 ++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  6 ++++
 3 files changed, 39 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 1ef892bea488..d475c54c0a08 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -5308,6 +5308,7 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
        list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
                if (need_full_reset) {
                        /* post card */
+                       amdgpu_ras_set_fed(tmp_adev, false);
                        r = amdgpu_device_asic_init(tmp_adev);
                        if (r) {
                                dev_warn(tmp_adev->dev, "asic atom init failed!"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 46f3d1013e8c..2c94de305c69 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -2439,6 +2439,18 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
                                ras->gpu_reset_flags &= ~AMDGPU_RAS_GPU_RESET_MODE1_RESET;
                                set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);

+                               /* For any RAS error that needs a full reset to
+                                * recover, set the fatal error status
+                                */
+                               if (hive) {
+                                       list_for_each_entry(remote_adev,
+                                                           &hive->device_list,
+                                                           gmc.xgmi.head)
+                                               amdgpu_ras_set_fed(remote_adev,
+                                                                  true);
+                               } else {
+                                       amdgpu_ras_set_fed(adev, true);
+                               }
                                psp_fatal_error_recovery_quirk(&adev->psp);
                        }
                }
@@ -3440,6 +3452,26 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
        return 0;
 }

+bool amdgpu_ras_get_fed_status(struct amdgpu_device *adev) {
+       struct amdgpu_ras *ras;
+
+       ras = amdgpu_ras_get_context(adev);
+       if (!ras)
+               return false;
+
+       return atomic_read(&ras->fed);
+}
+
+void amdgpu_ras_set_fed(struct amdgpu_device *adev, bool status) {
+       struct amdgpu_ras *ras;
+
+       ras = amdgpu_ras_get_context(adev);
+       if (ras)
+               atomic_set(&ras->fed, !!status);
+}
+
 void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)  {
        if (atomic_cmpxchg(&amdgpu_ras_in_intr, 0, 1) == 0) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
index d10e5bb0e52f..e0f8ce9d8440 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
@@ -477,6 +477,8 @@ struct amdgpu_ras {
        wait_queue_head_t page_retirement_wq;
        struct mutex page_retirement_lock;
        atomic_t page_retirement_req_cnt;
+       /* Fatal error detected flag */
+       atomic_t fed;
 };

 struct ras_fs_data {
@@ -873,4 +875,8 @@ void amdgpu_ras_add_mca_err_addr(struct ras_err_info *err_info,

 void amdgpu_ras_del_mca_err_addr(struct ras_err_info *err_info,
                struct ras_err_addr *mca_err_addr);
+
+void amdgpu_ras_set_fed(struct amdgpu_device *adev, bool status); bool
+amdgpu_ras_get_fed_status(struct amdgpu_device *adev);
+
 #endif
--
2.25.1