Re: [PATCH 1/4] btrfs: replace BUG() with error handling in compression.c

[Qu Wenruo]

在 2026/2/28 05:01, Adarsh Das 写道:
> Replace BUG() calls with proper error handling. Where fs_info is
> available, use btrfs_err() and return -EUCLEAN. Where fs_info is
> not available,

btrfs_err() can accept a NULL pointer as @fs_info.

[...]
>   
> @@ -874,11 +862,8 @@ static struct list_head *get_workspace(struct btrfs_fs_info *fs_info, int type,
>   	case BTRFS_COMPRESS_LZO:  return btrfs_get_workspace(fs_info, type, level);
>   	case BTRFS_COMPRESS_ZSTD: return zstd_get_workspace(fs_info, level);
>   	default:
> -		/*
> -		 * This can't happen, the type is validated several times
> -		 * before we get here.
> -		 */
> -		BUG();
> +		btrfs_err(fs_info, "invalid compression type %d", type);
> +		return ERR_PTR(-EUCLEAN);

Although for such unknown compression level case, the workspace user and 
put_workspace() are all erroring out correctly, it's not following the 
common pattern of checking the error first.

I'd prefer to just remove those default: branch, and add an ASSERT() 
checking the type is inside the BTRFS_NR_COMPRESS_TYPE before calling 
switch().

Just as the comment said, the @type has been verify too many times, we 
don't need to add a new error pattern where no one is checking.

Thanks,
Qu

>   	}
>   }
>   
> @@ -925,11 +910,8 @@ static void put_workspace(struct btrfs_fs_info *fs_info, int type, struct list_h
>   	case BTRFS_COMPRESS_LZO:  return btrfs_put_workspace(fs_info, type, ws);
>   	case BTRFS_COMPRESS_ZSTD: return zstd_put_workspace(fs_info, ws);
>   	default:
> -		/*
> -		 * This can't happen, the type is validated several times
> -		 * before we get here.
> -		 */
> -		BUG();
> +		btrfs_err(fs_info, "invalid compression type %d", type);
> +		return;
>   	}
>   }
>