8.10 Code: Inode content
The on-disk inode structure,
struct dinode,
contains a size and an array of block numbers (see
Figure 8.3).
The inode data is found in the blocks listed
in the
dinode ’s
addrs
array.
The first
NDIRECT
blocks of data are listed in the first
NDIRECT
entries in the array; these blocks are called
direct blocks.
The next
NINDIRECT
blocks of data are listed not in the inode
but in a data block called the
indirect block.
The last entry in the
addrs
array gives the address of the indirect block.
Thus the first 12 kB (
NDIRECT
x
BSIZE)
bytes of a file can be loaded from blocks listed in the inode,
while the next
256 kB (
NINDIRECT
x
BSIZE)
bytes can only be loaded after consulting the indirect block.
This is a good on-disk representation but a
complex one for clients.
The function
bmap
manages the representation so that higher-level routines, such as
readi
and
writei,
which we will see shortly, do not need to manage this complexity.
bmap
returns the disk block number of the
bn’th
data block for the inode
ip.
If
ip
does not have such a block yet,
bmap
allocates one.
The function
bmap
(kernel/fs.c:383)
begins by picking off the easy case: the first
NDIRECT
blocks are listed in the inode itself
(kernel/fs.c:388-396).
The next
NINDIRECT
blocks are listed in the indirect block at
ip->addrs[NDIRECT].
bmap
reads the indirect block
(kernel/fs.c:407)
and then reads a block number from the right
position within the block
(kernel/fs.c:408).
If the block number exceeds
NDIRECT+NINDIRECT,
bmap
panics;
writei
contains the check that prevents this from happening
(kernel/fs.c:513).
bmap
allocates blocks as needed.
An
ip->addrs[]
or indirect
entry of zero indicates that no block is allocated.
As
bmap
encounters zeros, it replaces them with the numbers of fresh blocks,
allocated on demand
(kernel/fs.c:389-390)
(kernel/fs.c:401-402).
itrunc
frees a file’s blocks, resetting the inode’s size to zero.
itrunc
(kernel/fs.c:426)
starts by freeing the direct blocks
(kernel/fs.c:432-437),
then the ones listed in the indirect block
(kernel/fs.c:442-445),
and finally the indirect block itself
(kernel/fs.c:447-448).
bmap
makes it easy for
readi
and
writei
to get at an inode’s data.
readi
(kernel/fs.c:472)
starts by
making sure that the offset and count are not
beyond the end of the file.
Reads that start beyond the end of the file return an error
(kernel/fs.c:477-478)
while reads that start at or cross the end of the file
return fewer bytes than requested
(kernel/fs.c:479-480).
The main loop processes each block of the file,
copying data from the buffer into
dst
(kernel/fs.c:482-494).
writei
(kernel/fs.c:506)
is identical to
readi,
with three exceptions:
writes that start at or cross the end of the file
grow the file, up to the maximum file size
(kernel/fs.c:513-514);
the loop copies data into the buffers instead of out
(kernel/fs.c:522);
and if the write has extended the file,
writei
must update its size
(kernel/fs.c:530-531).
The function
stati
(kernel/fs.c:458)
copies inode metadata into the
stat
structure, which is exposed to user programs
via the
stat
system call.