[PATCH 19/21] drm/amdgpu: parse legacy RAS bad page mixed with new data in various NPS modes

[Tao Zhou]

All legacy RAS bad pages are generated in NPS1 mode, but new bad page can be
generated in any NPS mode, so we can't use retired_page stored on eeprom
directly in non-nps1 mode even for legacy data. We need to take different
actions for different data, new data can be identified from old data by
UMC_CHANNEL_IDX_V2 flag.

Signed-off-by: Tao Zhou <tao.zhou1 at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 45 +++++++++++++++++++------
 1 file changed, 35 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index be56938932d1..76cd095f96b6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -2757,12 +2757,7 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
 			is_mca_add = false;
 	}
 
-	mutex_lock(&con->recovery_lock);
-	data = con->eh_data;
-	if (!data)
-		goto out;
-
-	if (is_mca_add) {
+	if (from_rom) {
 		err_data.err_addr =
 			kcalloc(adev->umc.retire_unit,
 				sizeof(struct eeprom_table_record), GFP_KERNEL);
@@ -2773,11 +2768,17 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
 			goto out;
 		}
 
+		err_rec = err_data.err_addr;
 		loop_cnt = adev->umc.retire_unit;
 		if (adev->gmc.gmc_funcs->query_mem_partition_mode)
 			nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
 	}
 
+	mutex_lock(&con->recovery_lock);
+	data = con->eh_data;
+	if (!data)
+		goto free;
+
 	for (i = 0; i < pages; i++) {
 		if (is_mca_add) {
 			if (!find_pages_per_pa) {
@@ -2799,10 +2800,34 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
 						bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT))
 					goto free;
 			}
-
-			err_rec = err_data.err_addr;
 		} else {
-			err_rec = &bps[i];
+			if (from_rom && !find_pages_per_pa) {
+				if (bps[i].retired_page & UMC_CHANNEL_IDX_V2) {
+					/* bad page in any NPS mode in eeprom */
+					if (amdgpu_ras_mca2pa_by_idx(adev, &bps[i], &err_data))
+						goto free;
+				} else {
+					/* legacy bad page in eeprom, generated only in NPS1 mode */
+					if (amdgpu_ras_mca2pa(adev, &bps[i], &err_data)) {
+						/* old RAS TA or ASICs which don't support to convert addrss
+						 * via mca address
+						 */
+						if (!i && nps == AMDGPU_NPS1_PARTITION_MODE) {
+							find_pages_per_pa = true;
+							err_rec = &bps[i];
+							loop_cnt = 1;
+						} else {
+							/* non-nps1 mode, old RAS TA can't support it */
+							goto free;
+						}
+					}
+				}
+
+				if (!find_pages_per_pa)
+					i += (adev->umc.retire_unit - 1);
+			} else {
+				err_rec = &bps[i];
+			}
 		}
 
 		for (j = 0; j < loop_cnt; j++) {
@@ -2826,7 +2851,7 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
 	}
 
 free:
-	if (is_mca_add)
+	if (from_rom)
 		kfree(err_data.err_addr);
 out:
 	mutex_unlock(&con->recovery_lock);
-- 
2.34.1