5 guestfs - Library for accessing and modifying virtual machine images
11 guestfs_h *handle = guestfs_create ();
12 guestfs_add_drive (handle, "guest.img");
13 guestfs_launch (handle);
14 guestfs_mount (handle, "/dev/sda1", "/");
15 guestfs_touch (handle, "/hello");
16 guestfs_sync (handle);
17 guestfs_close (handle);
21 Libguestfs is a library for accessing and modifying guest disk images.
22 Amongst the things this is good for: making batch configuration
23 changes to guests, getting disk used/free statistics (see also:
24 virt-df), migrating between virtualization systems (see also:
25 virt-p2v), performing partial backups, performing partial guest
26 clones, cloning guests and changing registry/UUID/hostname info, and
29 Libguestfs uses Linux kernel and qemu code, and can access any type of
30 guest filesystem that Linux and qemu can, including but not limited
31 to: ext2/3/4, btrfs, FAT and NTFS, LVM, many different disk partition
32 schemes, qcow, qcow2, vmdk.
34 Libguestfs provides ways to enumerate guest storage (eg. partitions,
35 LVs, what filesystem is in each LV, etc.). It can also run commands
36 in the context of the guest. Also you can access filesystems over
39 Libguestfs is a library that can be linked with C and C++ management
40 programs (or management programs written in OCaml, Perl, Python, Ruby,
41 Java, Haskell or C#). You can also use it from shell scripts or the
44 You don't need to be root to use libguestfs, although obviously you do
45 need enough permissions to access the disk images.
47 Libguestfs is a large API because it can do many things. For a gentle
48 introduction, please read the L</API OVERVIEW> section next.
52 This section provides a gentler overview of the libguestfs API. We
53 also try to group API calls together, where that may not be obvious
54 from reading about the individual calls below.
58 Before you can use libguestfs calls, you have to create a handle.
59 Then you must add at least one disk image to the handle, followed by
60 launching the handle, then performing whatever operations you want,
61 and finally closing the handle. So the general structure of all
62 libguestfs-using programs looks like this:
64 guestfs_h *handle = guestfs_create ();
66 /* Call guestfs_add_drive additional times if there are
67 * multiple disk images.
69 guestfs_add_drive (handle, "guest.img");
71 /* Most manipulation calls won't work until you've launched
72 * the handle. You have to do this _after_ adding drives
73 * and _before_ other commands.
75 guestfs_launch (handle);
77 /* Now you can examine what partitions, LVs etc are available.
79 char **partitions = guestfs_list_partitions (handle);
80 char **logvols = guestfs_lvs (handle);
82 /* To access a filesystem in the image, you must mount it.
84 guestfs_mount (handle, "/dev/sda1", "/");
86 /* Now you can perform filesystem actions on the guest
89 guestfs_touch (handle, "/hello");
91 /* You only need to call guestfs_sync if you have made
92 * changes to the guest image.
94 guestfs_sync (handle);
96 /* Close the handle. */
97 guestfs_close (handle);
99 The code above doesn't include any error checking. In real code you
100 should check return values carefully for errors. In general all
101 functions that return integers return C<-1> on error, and all
102 functions that return pointers return C<NULL> on error. See section
103 L</ERROR HANDLING> below for how to handle errors, and consult the
104 documentation for each function call below to see precisely how they
105 return error indications.
109 The image filename (C<"guest.img"> in the example above) could be a
110 disk image from a virtual machine, a L<dd(1)> copy of a physical hard
111 disk, an actual block device, or simply an empty file of zeroes that
112 you have created through L<posix_fallocate(3)>. Libguestfs lets you
113 do useful things to all of these.
115 You can add a disk read-only using C<guestfs_add_drive_ro>, in which
116 case libguestfs won't modify the file.
118 Be extremely cautious if the disk image is in use, eg. if it is being
119 used by a virtual machine. Adding it read-write will almost certainly
120 cause disk corruption, but adding it read-only is safe.
122 You must add at least one disk image, and you may add multiple disk
123 images. In the API, the disk images are usually referred to as
124 C</dev/sda> (for the first one you added), C</dev/sdb> (for the second
127 Once C<guestfs_launch> has been called you cannot add any more images.
128 You can call C<guestfs_list_devices> to get a list of the device
129 names, in the order that you added them. See also L</BLOCK DEVICE
134 Before you can read or write files, create directories and so on in a
135 disk image that contains filesystems, you have to mount those
136 filesystems using C<guestfs_mount>. If you already know that a disk
137 image contains (for example) one partition with a filesystem on that
138 partition, then you can mount it directly:
140 guestfs_mount (handle, "/dev/sda1", "/");
142 where C</dev/sda1> means literally the first partition (C<1>) of the
143 first disk image that we added (C</dev/sda>). If the disk contains
144 Linux LVM2 logical volumes you could refer to those instead (eg. C</dev/VG/LV>).
146 If you are given a disk image and you don't know what it contains then
147 you have to find out. Libguestfs can do that too: use
148 C<guestfs_list_partitions> and C<guestfs_lvs> to list possible
149 partitions and LVs, and either try mounting each to see what is
150 mountable, or else examine them with C<guestfs_file>. But you might
151 find it easier to look at higher level programs built on top of
152 libguestfs, in particular L<virt-inspector(1)>.
154 To mount a disk image read-only, use C<guestfs_mount_ro>. There are
155 several other variations of the C<guestfs_mount_*> call.
157 =head2 FILESYSTEM ACCESS AND MODIFICATION
159 The majority of the libguestfs API consists of fairly low-level calls
160 for accessing and modifying the files, directories, symlinks etc on
161 mounted filesystems. There are over a hundred such calls which you
162 can find listed in detail below in this man page, and we don't even
163 pretend to cover them all in this overview.
165 Specify filenames as full paths including the mount point.
167 For example, if you mounted a filesystem at C<"/"> and you want to
168 read the file called C<"etc/passwd"> then you could do:
170 char *data = guestfs_cat (handle, "/etc/passwd");
172 This would return C<data> as a newly allocated buffer containing the
173 full content of that file (with some conditions: see also
174 L</DOWNLOADING> below), or C<NULL> if there was an error.
176 As another example, to create a top-level directory on that filesystem
177 called C<"var"> you would do:
179 guestfs_mkdir (handle, "/var");
181 To create a symlink you could do:
183 guestfs_ln_s (handle, "/etc/init.d/portmap",
184 "/etc/rc3.d/S30portmap");
186 Libguestfs will reject attempts to use relative paths. There is no
187 concept of a current working directory. Libguestfs can return errors
188 in many situations: for example if the filesystem isn't writable, or
189 if a file or directory that you requested doesn't exist. If you are
190 using the C API (documented here) you have to check for those error
191 conditions after each call. (Other language bindings turn these
192 errors into exceptions).
194 File writes are affected by the per-handle umask, set by calling
195 C<guestfs_umask> and defaulting to 022.
199 Libguestfs contains API calls to read, create and modify partition
200 tables on disk images.
202 In the common case where you want to create a single partition
203 covering the whole disk, you should use the C<guestfs_part_disk>
206 const char *parttype = "mbr";
207 if (disk_is_larger_than_2TB)
209 guestfs_part_disk (g, "/dev/sda", parttype);
211 Obviously this effectively wipes anything that was on that disk image
214 In general MBR partitions are both unnecessarily complicated and
215 depend on archaic details, namely the Cylinder-Head-Sector (CHS)
216 geometry of the disk. C<guestfs_sfdiskM> can be used to
217 create more complex arrangements where the relative sizes are
218 expressed in megabytes instead of cylinders, which is a small win.
219 C<guestfs_sfdiskM> will choose the nearest cylinder to approximate the
220 requested size. There's a lot of crazy stuff to do with IDE and
221 virtio disks having different, incompatible CHS geometries, that you
222 probably don't want to know about.
224 My advice: make a single partition to cover the whole disk, then use
229 Libguestfs provides access to a large part of the LVM2 API, such as
230 C<guestfs_lvcreate> and C<guestfs_vgremove>. It won't make much sense
231 unless you familiarize yourself with the concepts of physical volumes,
232 volume groups and logical volumes.
234 This author strongly recommends reading the LVM HOWTO, online at
235 L<http://tldp.org/HOWTO/LVM-HOWTO/>.
239 Use C<guestfs_cat> to download small, text only files. This call
240 is limited to files which are less than 2 MB and which cannot contain
241 any ASCII NUL (C<\0>) characters. However it has a very simple
244 C<guestfs_read_file> can be used to read files which contain
245 arbitrary 8 bit data, since it returns a (pointer, size) pair.
246 However it is still limited to "small" files, less than 2 MB.
248 C<guestfs_download> can be used to download any file, with no
249 limits on content or size (even files larger than 4 GB).
251 To download multiple files, see C<guestfs_tar_out> and
256 It's often the case that you want to write a file or files to the disk
259 For small, single files, use C<guestfs_write_file>. This call
260 currently contains a bug which limits the call to plain text files
261 (not containing ASCII NUL characters).
263 To upload a single file, use C<guestfs_upload>. This call has no
264 limits on file content or size (even files larger than 4 GB).
266 To upload multiple files, see C<guestfs_tar_in> and C<guestfs_tgz_in>.
268 However the fastest way to upload I<large numbers of arbitrary files>
269 is to turn them into a squashfs or CD ISO (see L<mksquashfs(8)> and
270 L<mkisofs(8)>), then attach this using C<guestfs_add_drive_ro>. If
271 you add the drive in a predictable way (eg. adding it last after all
272 other drives) then you can get the device name from
273 C<guestfs_list_devices> and mount it directly using
274 C<guestfs_mount_ro>. Note that squashfs images are sometimes
275 non-portable between kernel versions, and they don't support labels or
276 UUIDs. If you want to pre-build an image or you need to mount it
277 using a label or UUID, use an ISO image instead.
281 There are various different commands for copying between files and
282 devices and in and out of the guest filesystem. These are summarised
287 =item B<file> to B<file>
289 Use L</guestfs_cp> to copy a single file, or
290 L</guestfs_cp_a> to copy directories recursively.
292 =item B<file or device> to B<file or device>
294 Use L</guestfs_dd> which efficiently uses L<dd(1)>
295 to copy between files and devices in the guest.
297 Example: duplicate the contents of an LV:
299 guestfs_dd (g, "/dev/VG/Original", "/dev/VG/Copy");
301 The destination (C</dev/VG/Copy>) must be at least as large as the
302 source (C</dev/VG/Original>).
304 =item B<file on the host> to B<file or device>
306 Use L</guestfs_upload>. See L</UPLOADING> above.
308 =item B<file or device> to B<file on the host>
310 Use L</guestfs_download>. See L</DOWNLOADING> above.
316 C<guestfs_ll> is just designed for humans to read (mainly when using
317 the L<guestfish(1)>-equivalent command C<ll>).
319 C<guestfs_ls> is a quick way to get a list of files in a directory
320 from programs, as a flat list of strings.
322 C<guestfs_readdir> is a programmatic way to get a list of files in a
323 directory, plus additional information about each one. It is more
324 equivalent to using the L<readdir(3)> call on a local filesystem.
326 C<guestfs_find> can be used to recursively list files.
328 =head2 RUNNING COMMANDS
330 Although libguestfs is a primarily an API for manipulating files
331 inside guest images, we also provide some limited facilities for
332 running commands inside guests.
334 There are many limitations to this:
340 The kernel version that the command runs under will be different
341 from what it expects.
345 If the command needs to communicate with daemons, then most likely
346 they won't be running.
350 The command will be running in limited memory.
354 Only supports Linux guests (not Windows, BSD, etc).
358 Architecture limitations (eg. won't work for a PPC guest on
363 For SELinux guests, you may need to enable SELinux and load policy
364 first. See L</SELINUX> in this manpage.
368 The two main API calls to run commands are C<guestfs_command> and
369 C<guestfs_sh> (there are also variations).
371 The difference is that C<guestfs_sh> runs commands using the shell, so
372 any shell globs, redirections, etc will work.
374 =head2 CONFIGURATION FILES
376 To read and write configuration files in Linux guest filesystems, we
377 strongly recommend using Augeas. For example, Augeas understands how
378 to read and write, say, a Linux shadow password file or X.org
379 configuration file, and so avoids you having to write that code.
381 The main Augeas calls are bound through the C<guestfs_aug_*> APIs. We
382 don't document Augeas itself here because there is excellent
383 documentation on the L<http://augeas.net/> website.
385 If you don't want to use Augeas (you fool!) then try calling
386 C<guestfs_read_lines> to get the file as a list of lines which
387 you can iterate over.
391 We support SELinux guests. To ensure that labeling happens correctly
392 in SELinux guests, you need to enable SELinux and load the guest's
399 Before launching, do:
401 guestfs_set_selinux (g, 1);
405 After mounting the guest's filesystem(s), load the policy. This
406 is best done by running the L<load_policy(8)> command in the
409 guestfs_sh (g, "/usr/sbin/load_policy");
411 (Older versions of C<load_policy> require you to specify the
412 name of the policy file).
416 Optionally, set the security context for the API. The correct
417 security context to use can only be known by inspecting the
418 guest. As an example:
420 guestfs_setcon (g, "unconfined_u:unconfined_r:unconfined_t:s0");
424 This will work for running commands and editing existing files.
426 When new files are created, you may need to label them explicitly,
427 for example by running the external command
428 C<restorecon pathname>.
430 =head2 SPECIAL CONSIDERATIONS FOR WINDOWS GUESTS
432 Libguestfs can mount NTFS partitions. It does this using the
433 L<http://www.ntfs-3g.org/> driver.
435 DOS and Windows still use drive letters, and the filesystems are
436 always treated as case insensitive by Windows itself, and therefore
437 you might find a Windows configuration file referring to a path like
438 C<c:\windows\system32>. When the filesystem is mounted in libguestfs,
439 that directory might be referred to as C</WINDOWS/System32>.
441 Drive letter mappings are outside the scope of libguestfs. You have
442 to use libguestfs to read the appropriate Windows Registry and
443 configuration files, to determine yourself how drives are mapped (see
444 also L<virt-inspector(1)>).
446 Replacing backslash characters with forward slash characters is also
447 outside the scope of libguestfs, but something that you can easily do.
449 Where we can help is in resolving the case insensitivity of paths.
450 For this, call C<guestfs_case_sensitive_path>.
452 Libguestfs also provides some help for decoding Windows Registry
453 "hive" files, through the library C<libhivex> which is part of
454 libguestfs. You have to locate and download the hive file(s)
455 yourself, and then pass them to C<libhivex> functions. See also the
456 programs L<hivexml(1)>, L<hivexget(1)> and L<virt-win-reg(1)> for more
459 =head2 USING LIBGUESTFS WITH OTHER PROGRAMMING LANGUAGES
461 Although we don't want to discourage you from using the C API, we will
462 mention here that the same API is also available in other languages.
464 The API is broadly identical in all supported languages. This means
465 that the C call C<guestfs_mount(handle,path)> is
466 C<$handle-E<gt>mount($path)> in Perl, C<handle.mount(path)> in Python,
467 and C<Guestfs.mount handle path> in OCaml. In other words, a
468 straightforward, predictable isomorphism between each language.
470 Error messages are automatically transformed
471 into exceptions if the language supports it.
473 We don't try to "object orientify" parts of the API in OO languages,
474 although contributors are welcome to write higher level APIs above
475 what we provide in their favourite languages if they wish.
481 You can use the I<guestfs.h> header file from C++ programs. The C++
482 API is identical to the C API. C++ classes and exceptions are
487 The C# bindings are highly experimental. Please read the warnings
488 at the top of C<csharp/Libguestfs.cs>.
492 This is the only language binding that working but incomplete. Only
493 calls which return simple integers have been bound in Haskell, and we
494 are looking for help to complete this binding.
498 Full documentation is contained in the Javadoc which is distributed
503 For documentation see the file C<guestfs.mli>.
507 For documentation see L<Sys::Guestfs(3)>.
511 For documentation do:
519 Use the Guestfs module. There is no Ruby-specific documentation, but
520 you can find examples written in Ruby in the libguestfs source.
522 =item B<shell scripts>
524 For documentation see L<guestfish(1)>.
528 =head2 LIBGUESTFS GOTCHAS
530 L<http://en.wikipedia.org/wiki/Gotcha_(programming)>: "A feature of a
531 system [...] that works in the way it is documented but is
532 counterintuitive and almost invites mistakes."
534 Since we developed libguestfs and the associated tools, there are
535 several things we would have designed differently, but are now stuck
536 with for backwards compatibility or other reasons. If there is ever a
537 libguestfs 2.0 release, you can expect these to change. Beware of
542 =item Autosync / forgetting to sync.
544 When modifying a filesystem from C or another language, you B<must>
545 unmount all filesystems and call L</guestfs_sync> explicitly before
546 you close the libguestfs handle. You can also call:
548 guestfs_set_autosync (handle, 1);
550 to have the unmount/sync done automatically for you when the handle is
551 closed. (This feature is called "autosync", L</guestfs_set_autosync>
554 If you forget to do this, then it is entirely possible that your
555 changes won't be written out, or will be partially written, or (very
556 rarely) that you'll get disk corruption.
558 Note that in L<guestfish(3)> I<autosync is the default>. So quick and
559 dirty guestfish scripts that forget to sync will work just fine, which
560 can make this extra-puzzling if you are trying to debug a problem.
562 =item Mount option C<-o sync> should not be the default.
564 If you use C<guestfs_mount>, then C<-o sync,noatime> are added
565 implicitly. However C<-o sync> does not add any reliability benefit,
566 but does have a very large performance impact.
568 The work around is to use C<guestfs_mount_options> and set the mount
569 options that you actually want to use.
571 =item Read-only should be the default.
573 In L<guestfish(3)>, I<--ro> should be the default, and you should
574 have to specify I<--rw> if you want to make changes to the image.
576 This would reduce the potential to corrupt live VM images.
578 Note that many filesystems change the disk when you just mount and
579 unmount, even if you didn't perform any writes. You need to use
580 C<guestfs_add_drive_ro> to guarantee that the disk is not changed.
582 =item guestfish command line is hard to use.
584 C<guestfish disk.img> doesn't do what people expect (open C<disk.img>
585 for examination). It tries to run a guestfish command C<disk.img>
586 which doesn't exist, so it fails, and it fails with a strange and
587 unintuitive error message. Like the Bourne shell, we should have used
588 C<guestfish -c command> to run commands.
592 =head1 CONNECTION MANAGEMENT
596 C<guestfs_h> is the opaque type representing a connection handle.
597 Create a handle by calling C<guestfs_create>. Call C<guestfs_close>
598 to free the handle and release all resources used.
600 For information on using multiple handles and threads, see the section
601 L</MULTIPLE HANDLES AND MULTIPLE THREADS> below.
603 =head2 guestfs_create
605 guestfs_h *guestfs_create (void);
607 Create a connection handle.
609 You have to call C<guestfs_add_drive> on the handle at least once.
611 This function returns a non-NULL pointer to a handle on success or
614 After configuring the handle, you have to call C<guestfs_launch>.
616 You may also want to configure error handling for the handle. See
617 L</ERROR HANDLING> section below.
621 void guestfs_close (guestfs_h *handle);
623 This closes the connection handle and frees up all resources used.
625 =head1 ERROR HANDLING
627 The convention in all functions that return C<int> is that they return
628 C<-1> to indicate an error. You can get additional information on
629 errors by calling C<guestfs_last_error> and/or by setting up an error
630 handler with C<guestfs_set_error_handler>.
632 The default error handler prints the information string to C<stderr>.
634 Out of memory errors are handled differently. The default action is
635 to call L<abort(3)>. If this is undesirable, then you can set a
636 handler using C<guestfs_set_out_of_memory_handler>.
638 =head2 guestfs_last_error
640 const char *guestfs_last_error (guestfs_h *handle);
642 This returns the last error message that happened on C<handle>. If
643 there has not been an error since the handle was created, then this
646 The lifetime of the returned string is until the next error occurs, or
647 C<guestfs_close> is called.
649 The error string is not localized (ie. is always in English), because
650 this makes searching for error messages in search engines give the
651 largest number of results.
653 =head2 guestfs_set_error_handler
655 typedef void (*guestfs_error_handler_cb) (guestfs_h *handle,
658 void guestfs_set_error_handler (guestfs_h *handle,
659 guestfs_error_handler_cb cb,
662 The callback C<cb> will be called if there is an error. The
663 parameters passed to the callback are an opaque data pointer and the
664 error message string.
666 Note that the message string C<msg> is freed as soon as the callback
667 function returns, so if you want to stash it somewhere you must make
670 The default handler prints messages on C<stderr>.
672 If you set C<cb> to C<NULL> then I<no> handler is called.
674 =head2 guestfs_get_error_handler
676 guestfs_error_handler_cb guestfs_get_error_handler (guestfs_h *handle,
679 Returns the current error handler callback.
681 =head2 guestfs_set_out_of_memory_handler
683 typedef void (*guestfs_abort_cb) (void);
684 int guestfs_set_out_of_memory_handler (guestfs_h *handle,
687 The callback C<cb> will be called if there is an out of memory
688 situation. I<Note this callback must not return>.
690 The default is to call L<abort(3)>.
692 You cannot set C<cb> to C<NULL>. You can't ignore out of memory
695 =head2 guestfs_get_out_of_memory_handler
697 guestfs_abort_fn guestfs_get_out_of_memory_handler (guestfs_h *handle);
699 This returns the current out of memory handler.
703 Libguestfs needs a kernel and initrd.img, which it finds by looking
704 along an internal path.
706 By default it looks for these in the directory C<$libdir/guestfs>
707 (eg. C</usr/local/lib/guestfs> or C</usr/lib64/guestfs>).
709 Use C<guestfs_set_path> or set the environment variable
710 C<LIBGUESTFS_PATH> to change the directories that libguestfs will
711 search in. The value is a colon-separated list of paths. The current
712 directory is I<not> searched unless the path contains an empty element
713 or C<.>. For example C<LIBGUESTFS_PATH=:/usr/lib/guestfs> would
714 search the current directory and then C</usr/lib/guestfs>.
716 =head1 HIGH-LEVEL API ACTIONS
720 We guarantee the libguestfs ABI (binary interface), for public,
721 high-level actions as outlined in this section. Although we will
722 deprecate some actions, for example if they get replaced by newer
723 calls, we will keep the old actions forever. This allows you the
724 developer to program in confidence against libguestfs.
734 =head2 GROUPS OF FUNCTIONALITY IN THE APPLIANCE
736 Using L</guestfs_available> you can test availability of
737 the following groups of functions. This test queries the
738 appliance to see if the appliance you are currently using
739 supports the functionality.
743 =head2 SINGLE CALLS AT COMPILE TIME
745 If you need to test whether a single libguestfs function is
746 available at compile time, we recommend using build tools
747 such as autoconf or cmake. For example in autotools you could
750 AC_CHECK_LIB([guestfs],[guestfs_create])
751 AC_CHECK_FUNCS([guestfs_dd])
753 which would result in C<HAVE_GUESTFS_DD> being either defined
754 or not defined in your program.
756 =head2 SINGLE CALLS AT RUN TIME
758 Testing at compile time doesn't guarantee that a function really
759 exists in the library. The reason is that you might be dynamically
760 linked against a previous I<libguestfs.so> (dynamic library)
761 which doesn't have the call. This situation unfortunately results
762 in a segmentation fault, which is a shortcoming of the C dynamic
763 linking system itself.
765 You can use L<dlopen(3)> to test if a function is available
766 at run time, as in this example program (note that you still
767 need the compile time check as well):
779 #ifdef HAVE_GUESTFS_DD
783 /* Test if the function guestfs_dd is really available. */
784 dl = dlopen (NULL, RTLD_LAZY);
786 fprintf (stderr, "dlopen: %s\n", dlerror ());
789 has_function = dlsym (dl, "guestfs_dd") != NULL;
793 printf ("this libguestfs.so does NOT have guestfs_dd function\n");
795 printf ("this libguestfs.so has guestfs_dd function\n");
796 /* Now it's safe to call
797 guestfs_dd (g, "foo", "bar");
801 printf ("guestfs_dd function was not found at compile time\n");
805 You may think the above is an awful lot of hassle, and it is.
806 There are other ways outside of the C linking system to ensure
807 that this kind of incompatibility never arises, such as using
810 Requires: libguestfs >= 1.0.80
814 <!-- old anchor for the next section -->
815 <a name="state_machine_and_low_level_event_api"/>
821 Internally, libguestfs is implemented by running an appliance (a
822 special type of small virtual machine) using L<qemu(1)>. Qemu runs as
823 a child process of the main program.
829 | | child process / appliance
830 | | __________________________
832 +-------------------+ RPC | +-----------------+ |
833 | libguestfs <--------------------> guestfsd | |
834 | | | +-----------------+ |
835 \___________________/ | | Linux kernel | |
836 | +--^--------------+ |
837 \_________|________________/
845 The library, linked to the main program, creates the child process and
846 hence the appliance in the L</guestfs_launch> function.
848 Inside the appliance is a Linux kernel and a complete stack of
849 userspace tools (such as LVM and ext2 programs) and a small
850 controlling daemon called C<guestfsd>. The library talks to
851 C<guestfsd> using remote procedure calls (RPC). There is a mostly
852 one-to-one correspondence between libguestfs API calls and RPC calls
853 to the daemon. Lastly the disk image(s) are attached to the qemu
854 process which translates device access by the appliance's Linux kernel
855 into accesses to the image.
857 A common misunderstanding is that the appliance "is" the virtual
858 machine. Although the disk image you are attached to might also be
859 used by some virtual machine, libguestfs doesn't know or care about
860 this. (But you will care if both libguestfs's qemu process and your
861 virtual machine are trying to update the disk image at the same time,
862 since these usually results in massive disk corruption).
866 libguestfs uses a state machine to model the child process:
877 / | \ \ guestfs_launch
888 \______/ <------ \________/
890 The normal transitions are (1) CONFIG (when the handle is created, but
891 there is no child process), (2) LAUNCHING (when the child process is
892 booting up), (3) alternating between READY and BUSY as commands are
893 issued to, and carried out by, the child process.
895 The guest may be killed by C<guestfs_kill_subprocess>, or may die
896 asynchronously at any time (eg. due to some internal error), and that
897 causes the state to transition back to CONFIG.
899 Configuration commands for qemu such as C<guestfs_add_drive> can only
900 be issued when in the CONFIG state.
902 The high-level API offers two calls that go from CONFIG through
903 LAUNCHING to READY. C<guestfs_launch> blocks until the child process
904 is READY to accept commands (or until some failure or timeout).
905 C<guestfs_launch> internally moves the state from CONFIG to LAUNCHING
908 High-level API actions such as C<guestfs_mount> can only be issued
909 when in the READY state. These high-level API calls block waiting for
910 the command to be carried out (ie. the state to transition to BUSY and
911 then back to READY). But using the low-level event API, you get
912 non-blocking versions. (But you can still only carry out one
913 operation per handle at a time - that is a limitation of the
914 communications protocol we use).
916 Finally, the child process sends asynchronous messages back to the
917 main program, such as kernel log messages. Mostly these are ignored
918 by the high-level API, but using the low-level event API you can
919 register to receive these messages.
921 =head2 SETTING CALLBACKS TO HANDLE EVENTS
923 The child process generates events in some situations. Current events
924 include: receiving a log message, the child process exits.
926 Use the C<guestfs_set_*_callback> functions to set a callback for
927 different types of events.
929 Only I<one callback of each type> can be registered for each handle.
930 Calling C<guestfs_set_*_callback> again overwrites the previous
931 callback of that type. Cancel all callbacks of this type by calling
932 this function with C<cb> set to C<NULL>.
934 =head2 guestfs_set_log_message_callback
936 typedef void (*guestfs_log_message_cb) (guestfs_h *g, void *opaque,
938 void guestfs_set_log_message_callback (guestfs_h *handle,
939 guestfs_log_message_cb cb,
942 The callback function C<cb> will be called whenever qemu or the guest
943 writes anything to the console.
945 Use this function to capture kernel messages and similar.
947 Normally there is no log message handler, and log messages are just
950 =head2 guestfs_set_subprocess_quit_callback
952 typedef void (*guestfs_subprocess_quit_cb) (guestfs_h *g, void *opaque);
953 void guestfs_set_subprocess_quit_callback (guestfs_h *handle,
954 guestfs_subprocess_quit_cb cb,
957 The callback function C<cb> will be called when the child process
958 quits, either asynchronously or if killed by
959 C<guestfs_kill_subprocess>. (This corresponds to a transition from
960 any state to the CONFIG state).
962 =head2 guestfs_set_launch_done_callback
964 typedef void (*guestfs_launch_done_cb) (guestfs_h *g, void *opaque);
965 void guestfs_set_launch_done_callback (guestfs_h *handle,
969 The callback function C<cb> will be called when the child process
970 becomes ready first time after it has been launched. (This
971 corresponds to a transition from LAUNCHING to the READY state).
973 =head1 BLOCK DEVICE NAMING
975 In the kernel there is now quite a profusion of schemata for naming
976 block devices (in this context, by I<block device> I mean a physical
977 or virtual hard drive). The original Linux IDE driver used names
978 starting with C</dev/hd*>. SCSI devices have historically used a
979 different naming scheme, C</dev/sd*>. When the Linux kernel I<libata>
980 driver became a popular replacement for the old IDE driver
981 (particularly for SATA devices) those devices also used the
982 C</dev/sd*> scheme. Additionally we now have virtual machines with
983 paravirtualized drivers. This has created several different naming
984 systems, such as C</dev/vd*> for virtio disks and C</dev/xvd*> for Xen
987 As discussed above, libguestfs uses a qemu appliance running an
988 embedded Linux kernel to access block devices. We can run a variety
989 of appliances based on a variety of Linux kernels.
991 This causes a problem for libguestfs because many API calls use device
992 or partition names. Working scripts and the recipe (example) scripts
993 that we make available over the internet could fail if the naming
996 Therefore libguestfs defines C</dev/sd*> as the I<standard naming
997 scheme>. Internally C</dev/sd*> names are translated, if necessary,
998 to other names as required. For example, under RHEL 5 which uses the
999 C</dev/hd*> scheme, any device parameter C</dev/sda2> is translated to
1000 C</dev/hda2> transparently.
1002 Note that this I<only> applies to parameters. The
1003 C<guestfs_list_devices>, C<guestfs_list_partitions> and similar calls
1004 return the true names of the devices and partitions as known to the
1007 =head2 ALGORITHM FOR BLOCK DEVICE NAME TRANSLATION
1009 Usually this translation is transparent. However in some (very rare)
1010 cases you may need to know the exact algorithm. Such cases include
1011 where you use C<guestfs_config> to add a mixture of virtio and IDE
1012 devices to the qemu-based appliance, so have a mixture of C</dev/sd*>
1013 and C</dev/vd*> devices.
1015 The algorithm is applied only to I<parameters> which are known to be
1016 either device or partition names. Return values from functions such
1017 as C<guestfs_list_devices> are never changed.
1023 Is the string a parameter which is a device or partition name?
1027 Does the string begin with C</dev/sd>?
1031 Does the named device exist? If so, we use that device.
1032 However if I<not> then we continue with this algorithm.
1036 Replace initial C</dev/sd> string with C</dev/hd>.
1038 For example, change C</dev/sda2> to C</dev/hda2>.
1040 If that named device exists, use it. If not, continue.
1044 Replace initial C</dev/sd> string with C</dev/vd>.
1046 If that named device exists, use it. If not, return an error.
1050 =head2 PORTABILITY CONCERNS
1052 Although the standard naming scheme and automatic translation is
1053 useful for simple programs and guestfish scripts, for larger programs
1054 it is best not to rely on this mechanism.
1056 Where possible for maximum future portability programs using
1057 libguestfs should use these future-proof techniques:
1063 Use C<guestfs_list_devices> or C<guestfs_list_partitions> to list
1064 actual device names, and then use those names directly.
1066 Since those device names exist by definition, they will never be
1071 Use higher level ways to identify filesystems, such as LVM names,
1072 UUIDs and filesystem labels.
1078 =head2 COMMUNICATION PROTOCOL
1080 Don't rely on using this protocol directly. This section documents
1081 how it currently works, but it may change at any time.
1083 The protocol used to talk between the library and the daemon running
1084 inside the qemu virtual machine is a simple RPC mechanism built on top
1085 of XDR (RFC 1014, RFC 1832, RFC 4506).
1087 The detailed format of structures is in C<src/guestfs_protocol.x>
1088 (note: this file is automatically generated).
1090 There are two broad cases, ordinary functions that don't have any
1091 C<FileIn> and C<FileOut> parameters, which are handled with very
1092 simple request/reply messages. Then there are functions that have any
1093 C<FileIn> or C<FileOut> parameters, which use the same request and
1094 reply messages, but they may also be followed by files sent using a
1097 =head3 ORDINARY FUNCTIONS (NO FILEIN/FILEOUT PARAMS)
1099 For ordinary functions, the request message is:
1101 total length (header + arguments,
1102 but not including the length word itself)
1103 struct guestfs_message_header (encoded as XDR)
1104 struct guestfs_<foo>_args (encoded as XDR)
1106 The total length field allows the daemon to allocate a fixed size
1107 buffer into which it slurps the rest of the message. As a result, the
1108 total length is limited to C<GUESTFS_MESSAGE_MAX> bytes (currently
1109 4MB), which means the effective size of any request is limited to
1110 somewhere under this size.
1112 Note also that many functions don't take any arguments, in which case
1113 the C<guestfs_I<foo>_args> is completely omitted.
1115 The header contains the procedure number (C<guestfs_proc>) which is
1116 how the receiver knows what type of args structure to expect, or none
1119 The reply message for ordinary functions is:
1121 total length (header + ret,
1122 but not including the length word itself)
1123 struct guestfs_message_header (encoded as XDR)
1124 struct guestfs_<foo>_ret (encoded as XDR)
1126 As above the C<guestfs_I<foo>_ret> structure may be completely omitted
1127 for functions that return no formal return values.
1129 As above the total length of the reply is limited to
1130 C<GUESTFS_MESSAGE_MAX>.
1132 In the case of an error, a flag is set in the header, and the reply
1133 message is slightly changed:
1135 total length (header + error,
1136 but not including the length word itself)
1137 struct guestfs_message_header (encoded as XDR)
1138 struct guestfs_message_error (encoded as XDR)
1140 The C<guestfs_message_error> structure contains the error message as a
1143 =head3 FUNCTIONS THAT HAVE FILEIN PARAMETERS
1145 A C<FileIn> parameter indicates that we transfer a file I<into> the
1146 guest. The normal request message is sent (see above). However this
1147 is followed by a sequence of file chunks.
1149 total length (header + arguments,
1150 but not including the length word itself,
1151 and not including the chunks)
1152 struct guestfs_message_header (encoded as XDR)
1153 struct guestfs_<foo>_args (encoded as XDR)
1154 sequence of chunks for FileIn param #0
1155 sequence of chunks for FileIn param #1 etc.
1157 The "sequence of chunks" is:
1159 length of chunk (not including length word itself)
1160 struct guestfs_chunk (encoded as XDR)
1162 struct guestfs_chunk (encoded as XDR)
1165 struct guestfs_chunk (with data.data_len == 0)
1167 The final chunk has the C<data_len> field set to zero. Additionally a
1168 flag is set in the final chunk to indicate either successful
1169 completion or early cancellation.
1171 At time of writing there are no functions that have more than one
1172 FileIn parameter. However this is (theoretically) supported, by
1173 sending the sequence of chunks for each FileIn parameter one after
1174 another (from left to right).
1176 Both the library (sender) I<and> the daemon (receiver) may cancel the
1177 transfer. The library does this by sending a chunk with a special
1178 flag set to indicate cancellation. When the daemon sees this, it
1179 cancels the whole RPC, does I<not> send any reply, and goes back to
1180 reading the next request.
1182 The daemon may also cancel. It does this by writing a special word
1183 C<GUESTFS_CANCEL_FLAG> to the socket. The library listens for this
1184 during the transfer, and if it gets it, it will cancel the transfer
1185 (it sends a cancel chunk). The special word is chosen so that even if
1186 cancellation happens right at the end of the transfer (after the
1187 library has finished writing and has started listening for the reply),
1188 the "spurious" cancel flag will not be confused with the reply
1191 This protocol allows the transfer of arbitrary sized files (no 32 bit
1192 limit), and also files where the size is not known in advance
1193 (eg. from pipes or sockets). However the chunks are rather small
1194 (C<GUESTFS_MAX_CHUNK_SIZE>), so that neither the library nor the
1195 daemon need to keep much in memory.
1197 =head3 FUNCTIONS THAT HAVE FILEOUT PARAMETERS
1199 The protocol for FileOut parameters is exactly the same as for FileIn
1200 parameters, but with the roles of daemon and library reversed.
1202 total length (header + ret,
1203 but not including the length word itself,
1204 and not including the chunks)
1205 struct guestfs_message_header (encoded as XDR)
1206 struct guestfs_<foo>_ret (encoded as XDR)
1207 sequence of chunks for FileOut param #0
1208 sequence of chunks for FileOut param #1 etc.
1210 =head3 INITIAL MESSAGE
1212 Because the underlying channel (QEmu -net channel) doesn't have any
1213 sort of connection control, when the daemon launches it sends an
1214 initial word (C<GUESTFS_LAUNCH_FLAG>) which indicates that the guest
1215 and daemon is alive. This is what C<guestfs_launch> waits for.
1217 =head1 MULTIPLE HANDLES AND MULTIPLE THREADS
1219 All high-level libguestfs actions are synchronous. If you want
1220 to use libguestfs asynchronously then you must create a thread.
1222 Only use the handle from a single thread. Either use the handle
1223 exclusively from one thread, or provide your own mutex so that two
1224 threads cannot issue calls on the same handle at the same time.
1226 =head1 QEMU WRAPPERS
1228 If you want to compile your own qemu, run qemu from a non-standard
1229 location, or pass extra arguments to qemu, then you can write a
1230 shell-script wrapper around qemu.
1232 There is one important rule to remember: you I<must C<exec qemu>> as
1233 the last command in the shell script (so that qemu replaces the shell
1234 and becomes the direct child of the libguestfs-using program). If you
1235 don't do this, then the qemu process won't be cleaned up correctly.
1237 Here is an example of a wrapper, where I have built my own copy of
1241 qemudir=/home/rjones/d/qemu
1242 exec $qemudir/x86_64-softmmu/qemu-system-x86_64 -L $qemudir/pc-bios "$@"
1244 Save this script as C</tmp/qemu.wrapper> (or wherever), C<chmod +x>,
1245 and then use it by setting the LIBGUESTFS_QEMU environment variable.
1248 LIBGUESTFS_QEMU=/tmp/qemu.wrapper guestfish
1250 Note that libguestfs also calls qemu with the -help and -version
1251 options in order to determine features.
1253 =head1 ENVIRONMENT VARIABLES
1257 =item LIBGUESTFS_APPEND
1259 Pass additional options to the guest kernel.
1261 =item LIBGUESTFS_DEBUG
1263 Set C<LIBGUESTFS_DEBUG=1> to enable verbose messages. This
1264 has the same effect as calling C<guestfs_set_verbose (handle, 1)>.
1266 =item LIBGUESTFS_MEMSIZE
1268 Set the memory allocated to the qemu process, in megabytes. For
1271 LIBGUESTFS_MEMSIZE=700
1273 =item LIBGUESTFS_PATH
1275 Set the path that libguestfs uses to search for kernel and initrd.img.
1276 See the discussion of paths in section PATH above.
1278 =item LIBGUESTFS_QEMU
1280 Set the default qemu binary that libguestfs uses. If not set, then
1281 the qemu which was found at compile time by the configure script is
1284 See also L</QEMU WRAPPERS> above.
1286 =item LIBGUESTFS_TRACE
1288 Set C<LIBGUESTFS_TRACE=1> to enable command traces. This
1289 has the same effect as calling C<guestfs_set_trace (handle, 1)>.
1293 Location of temporary directory, defaults to C</tmp>.
1295 If libguestfs was compiled to use the supermin appliance then each
1296 handle will require rather a large amount of space in this directory
1297 for short periods of time (~ 80 MB). You can use C<$TMPDIR> to
1298 configure another directory to use in case C</tmp> is not large
1308 L<http://libguestfs.org/>.
1310 Tools with a similar purpose:
1319 To get a list of bugs against libguestfs use this link:
1321 L<https://bugzilla.redhat.com/buglist.cgi?component=libguestfs&product=Virtualization+Tools>
1323 To report a new bug against libguestfs use this link:
1325 L<https://bugzilla.redhat.com/enter_bug.cgi?component=libguestfs&product=Virtualization+Tools>
1327 When reporting a bug, please check:
1333 That the bug hasn't been reported already.
1337 That you are testing a recent version.
1341 Describe the bug accurately, and give a way to reproduce it.
1345 Run libguestfs-test-tool and paste the B<complete, unedited>
1346 output into the bug report.
1352 Richard W.M. Jones (C<rjones at redhat dot com>)
1356 Copyright (C) 2009 Red Hat Inc.
1357 L<http://libguestfs.org/>
1359 This library is free software; you can redistribute it and/or
1360 modify it under the terms of the GNU Lesser General Public
1361 License as published by the Free Software Foundation; either
1362 version 2 of the License, or (at your option) any later version.
1364 This library is distributed in the hope that it will be useful,
1365 but WITHOUT ANY WARRANTY; without even the implied warranty of
1366 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
1367 Lesser General Public License for more details.
1369 You should have received a copy of the GNU Lesser General Public
1370 License along with this library; if not, write to the Free Software
1371 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA