hi
i'm trying to write a "pure" block allocator i.e. it only allocates
blocks, not writing to them( imagine a file which i know that will grow
to same limit, like this i can pre-allocate all its block to sit as
close as possible)
as far as i understood the way block allocating is done in ext2fs is
like this:
user writes to file -> ext2fs try to find the block -> if block doesn't
exists -> block is allocated -> buffer head is filled with zeros ->
buffer head is returned to caller -> later to be written to disk.
since i'm gonna do only the allocation and write nothing to the disk,
and since files are gonna be very big( GBs ), it is a big slow down if
those buffer caches ever be written to disk with all those boring zeros,
it is also gonna slow down other processes since their buffer cache will
be trashed.
so i'm thinking of adding a little HACK as follows:
let ext2fs allocate the block, and relay on the fact that all writes are
performs only to buffer cache, and simply discard the cache after each
allocation so those buffers never get written to the disk, and other
processes won't suffer since the same buffer_head will be reused again
and again.
allocating the block will be done with ext2_getblk()
but then i need some way to mark the buffer_head as if it was written to
disk already AND put back the buffer_head in the end of the free list so
it will be reused again in the next call and we won't be trashing the
buffer cache
it looks like that put_last_free(bh) will do the last part, but how do
i mark the buffer_head as if it was written to the disk ?
will mark_buffer_clean() do the work
the following code suppose to allocate block_num_to_allocate to the
given indoe
buffer_head * bh;
for( block =0; block < block_num_to_alloacte; block++){
bh= ext2_getblk( &inode, block, 1, &err); //allocate the
block
mark_buffer_clean( bh ); // ???
put_last_free( bh ); // reuse buffer head
}
does it looks like this hack work or that i'm gonna create
inconsistency in the file system ?
THX
/gabriel
please CC me for any answer
10x again
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun May 07 2000 - 21:00:12 EST