drm/ttm: fix error handling if no BO can be swapped out v4

In case that all pre-allocated BOs are busy, just continue to populate
BOs since likely half of system memory in total is still free.

v4 (chk): fix code moved to VMWGFX as well

Signed-off-by: Shiwu Zhang <shiwu.zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210422115757.3946-1-christian.koenig@amd.com

Author: amd-morriszhang

drivers/gpu/drm/ttm/ttm_device.c

@@ -112,7 +112,7 @@ int ttm_global_swapout(struct ttm_operation_ctx *ctx, gfp_t gfp_flags)
 {
 	struct ttm_global *glob = &ttm_glob;
 	struct ttm_device *bdev;
-	int ret = -EBUSY;
+	int ret = 0;
 
 	mutex_lock(&ttm_global_mutex);
 	list_for_each_entry(bdev, &glob->device_list, device_list) {

drivers/gpu/drm/ttm/ttm_tt.c

@@ -329,6 +329,8 @@ int ttm_tt_populate(struct ttm_device *bdev,
 	       ttm_dma32_pages_limit) {
 
 		ret = ttm_global_swapout(ctx, GFP_KERNEL);
+		if (ret == 0)
+			break;
 		if (ret < 0)
 			goto error;
 	}

drivers/gpu/drm/vmwgfx/ttm_memory.c

@@ -280,7 +280,7 @@ static void ttm_shrink(struct ttm_mem_global *glob, bool from_wq,
 		spin_unlock(&glob->lock);
 		ret = ttm_global_swapout(ctx, GFP_KERNEL);
 		spin_lock(&glob->lock);
-		if (unlikely(ret < 0))
+		if (unlikely(ret <= 0))
 			break;
 	}