summaryrefslogtreecommitdiff
path: root/src/inodedb/README
blob: 82d9ecc5913df395c2d28834cfe2fbb755c1859c (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
Not all UNIXes provide an easy way to get the filename if you have
the device it is mounted on and its inode.

Well, under Linux we can simply read /proc/<pid>/maps to get the filenames
but I don't see any way to get this under FreeBSD.

In this case you can give configure the optional `--with-libgtop-inodedb'
parameter to build this stuff here.

It uses the GNU database library `gdbm' to find the inode in a database
which is created when libgtop is installed. This should be reasonable
fast since the entire filesystem only needs to be traversed once when
libgtop is installed (and each time you install new software, of cause).

We have two databases:

* First the system administrator may place a system-wide database
  in `$(prefix)/var/libgtop/inodedb.db' when installing libgtop.

* Every user can have his/her own one in `~/var/libgtop/inodedb.db'.
  This one has precedence over the system wide one.

The `mkinodedb' program which is build in this directory takes two
command line arguments: the full pathname of the database to be created
and the name of a configuration file consisting of directory and file names
each on a line by itself - see `/etc/ld.so.conf' for an example.

Putting a directory name in this file means all regular files found in this
directory are included in the database, but it will not recursively descend
into subdirectories (for instance, we want everythink in `/usr/lib' but not
every single file in `/usr/lib/sgml'). You can also use filenames to include
a single file.

To use this interface in your program, you first have to call
`glibtop_inodedb_open ()':

glibtop_inodedb *
glibtop_inodedb_open (unsigned databases, unsigned long cachesize)

glibtop_inodedb *
glibtop_inodedb_open_s (glibtop *server, unsigned databases,
			unsigned long cachesize)

There are some constants defined in <glibtop/inodedb.h> for the `databases'
parameter - you can use zero as default:

#define GLIBTOP_INODEDB_SYSTEM		1
#define GLIBTOP_INODEDB_USER		2
#define GLIBTOP_INODEDB_CACHE		4

The `cachesize' gives the size of a not yet implemented in-memory cache for
looked up entries.

This function will return a pointer of type `glibtop_inodedb *' which you need
to lookup an entry in the database.

To look up an entry, use the `glibtop_inodedb_lookup ()' function:

const char *
glibtop_inodedb_lookup (glibtop_inodedb *inodedb,
			u_int64_t device, u_int64_t inode)
const char *
glibtop_inodedb_lookup_s (glibtop *server, glibtop_inodedb *inodedb,
			  u_int64_t device, u_int64_t inode)

It will either return a pointer to the filename - which you have to
`glibtop_free ()' once done with it - or NULL on error.

You can use the `file_by_inode' program in this directory to test the
interface - just call it with two command line arguments: the device number
and the inode number.

Martin <martin@home-of-linux.org>