Programming

Lets say my program reads one block (4096 bytes) at a time from a file. It needs to report whether that block is a hole.

The program currently uses read() to read 1 block into a 4KB buffer.

Now I know that even if that block is a hole, it will be replaced by a sequence of NULL bytes '\0' when read into a buffer. And there is no way to detect if a particular block is a hole.

So for every block thats read in I will need to check if it contains consecutive stretches of zeros, and if so then report it as a hole.

Can anyone think of a way to do this ???

(yes I know that incase that block was not a hole but actually had sequence of zeros stored, my program will incorrectly report it as a hole)

Thanks

static char zeroes_4k[4096];

int its_a_hole=0;
char buffer[sizeof(zeroes_4k);

i=read(fd,buffer,sizeof(buffer));

if (i==sizeof(buffer))
{
if (!memcmp(buffer,zeroes_4k,sizeof(zeroes_4k))
{
its_a_hole=1;
}
}

sparse files can have less than a block of 0x0 characters. especially uncopied (or unrestored) files.

The easiest way: copy the file and see a difference in stat results for st_size between file_uncopied vs file_copied.

When you copy a sparse file (if it not alrerady a copy or a restore) it becomes larger.
Note that sysadmins detest these files because of what they do to disk space when restored from tape. Are you playing with zum?

st_size will not change when a sparse file is copied. st_blocks is what will change. You can stat a file and see if st_blocks is large enough to cover st_size. If not, then the file is sparse.

MY bad - it is just st_blocks that is affected.

Thanks for your reply guys !

Yes if st_blocks is not as large as st_size that means the file is sparse. And I can even calculate the number of holes it has. But my problem was to know exactly which blocks are holes.

Thanks porter !!! is there a particular reason why you have declared the char array as static ?

Thanks again everyone.. !

I think that you can tell the difference if you are willing to destroy the source file. Copy the file block-by-block starting from the end and truncate the source file as you go. stat the file after truncation to see if you freed a real block (that is, did st_blocks decrease?). This will be implementation dependant... posix tells us So some experimentation may be needed to how this works on your system.