How does Linux calculate the total block count when running /bin/ls -l?

Question

I am trying to figure out how the program /bin/ls -l calculates the total size (block count) for a directory. By this I mean the output total number that it prints right before the directory contents.

There is a similar question here: https://stackoverflow.com/questions/7401704/what-is-that-total-in-the-very-first-line-after-ls-l but it doesn't fully answer the question nor explain exactly how it is calculated.

I've tried adding the numbers of 512B blocks allocated for all the (non-hidden) files in a directory.  Here is how I am trying to go about it (in C):

 int getBlockSize(char* directory) {
   int size = 0;

   DIR *d;
   struct dirent *dir;
   struct stat fileStat;
   d = opendir(directory);
   if (d) {
       while ((dir = readdir(d)) != NULL) {
           if (dir->d_name[0] != '.') { // Ignore hidden files
               // Create the path to stat
               char info_path[PATH_MAX + 1];
               strcpy(info_path, directory);
               if (directory[strlen(directory) - 1] != '/')
                   strcat(info_path, "/");
               strcat(info_path, dir->d_name);

               stat(info_path, &fileStat);

               size += fileStat.st_blocks;
           }
       }
   }

   return size;
}

However this is giving me a much different number compared to the ls command.

What is 'wrong' with my approach?  How does ls compute the total?

Edit:

To test I made a folder which contains files test_file1.txt and test_file2.txt each containing the text Hello World!. When I run ls -l I get the following output

total 1
-rw-------. 1 aaa111 ugrad 13 Oct 27 13:17 test_file1.txt
-rw-------. 1 aaa111 ugrad 13 Oct 27 13:17 test_file2.txt

However when I run my code using the method above I get

total 2
-rw-------. 1 aaa111 ugrad 13 Oct 27 13:17 test_file1.txt
-rw-------. 1 aaa111 ugrad 13 Oct 27 13:17 test_file2.txt 

Answer 1

On Ubuntu, the default ls is GNU ls, which defaults to 1024-byte block size for its “total” line. This explains the difference in output between ls and your approach: your approach shows double the number of blocks because it’s counting 512-byte blocks.

There are various ways to force GNU ls to count in 512-byte blocks (see the link above); the most reliable is to set LS_BLOCK_SIZE:

LS_BLOCK_SIZE=512 ls -l

The other implementation of ls which you’re liable to run into on Linux is BusyBox ls; it also uses a 1024-byte block size for the “total” line, and it can’t be configured to use any other size.