mirror of
https://github.com/libguestfs/libguestfs.git
synced 2026-03-21 22:53:37 +00:00
2158 lines
66 KiB
C
2158 lines
66 KiB
C
/* libguestfs
|
|
* Copyright (C) 2010-2012 Red Hat Inc.
|
|
*
|
|
* This library is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2 of the License, or (at your option) any later version.
|
|
*
|
|
* This library is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* Lesser General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
* License along with this library; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
*/
|
|
|
|
#include <config.h>
|
|
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <inttypes.h>
|
|
#include <unistd.h>
|
|
#include <string.h>
|
|
#include <libintl.h>
|
|
#include <errno.h>
|
|
|
|
#ifdef HAVE_ENDIAN_H
|
|
#include <endian.h>
|
|
#endif
|
|
|
|
#include "c-ctype.h"
|
|
#include "ignore-value.h"
|
|
#include "hash-pjw.h"
|
|
|
|
#include "guestfs.h"
|
|
#include "guestfs-internal.h"
|
|
|
|
COMPILE_REGEXP (re_fedora, "Fedora release (\\d+)", 0)
|
|
COMPILE_REGEXP (re_rhel_old, "Red Hat.*release (\\d+).*Update (\\d+)", 0)
|
|
COMPILE_REGEXP (re_rhel, "Red Hat.*release (\\d+)\\.(\\d+)", 0)
|
|
COMPILE_REGEXP (re_rhel_no_minor, "Red Hat.*release (\\d+)", 0)
|
|
COMPILE_REGEXP (re_centos_old, "CentOS.*release (\\d+).*Update (\\d+)", 0)
|
|
COMPILE_REGEXP (re_centos, "CentOS.*release (\\d+)\\.(\\d+)", 0)
|
|
COMPILE_REGEXP (re_centos_no_minor, "CentOS.*release (\\d+)", 0)
|
|
COMPILE_REGEXP (re_scientific_linux_old,
|
|
"Scientific Linux.*release (\\d+).*Update (\\d+)", 0)
|
|
COMPILE_REGEXP (re_scientific_linux,
|
|
"Scientific Linux.*release (\\d+)\\.(\\d+)", 0)
|
|
COMPILE_REGEXP (re_scientific_linux_no_minor,
|
|
"Scientific Linux.*release (\\d+)", 0)
|
|
COMPILE_REGEXP (re_oracle_linux_old,
|
|
"Oracle Linux.*release (\\d+).*Update (\\d+)", 0)
|
|
COMPILE_REGEXP (re_oracle_linux,
|
|
"Oracle Linux.*release (\\d+)\\.(\\d+)", 0)
|
|
COMPILE_REGEXP (re_oracle_linux_no_minor, "Oracle Linux.*release (\\d+)", 0)
|
|
COMPILE_REGEXP (re_xdev, "^/dev/(h|s|v|xv)d([a-z]+)(\\d*)$", 0)
|
|
COMPILE_REGEXP (re_cciss, "^/dev/(cciss/c\\d+d\\d+)(?:p(\\d+))?$", 0)
|
|
COMPILE_REGEXP (re_mdN, "^(/dev/md\\d+)$", 0)
|
|
COMPILE_REGEXP (re_freebsd_mbr,
|
|
"^/dev/(ada{0,1}|vtbd)(\\d+)s(\\d+)([a-z])$", 0)
|
|
COMPILE_REGEXP (re_freebsd_gpt, "^/dev/(ada{0,1}|vtbd)(\\d+)p(\\d+)$", 0)
|
|
COMPILE_REGEXP (re_diskbyid, "^/dev/disk/by-id/.*-part(\\d+)$", 0)
|
|
COMPILE_REGEXP (re_netbsd, "^NetBSD (\\d+)\\.(\\d+)", 0)
|
|
COMPILE_REGEXP (re_opensuse, "^(openSUSE|SuSE Linux|SUSE LINUX) ", 0)
|
|
COMPILE_REGEXP (re_sles, "^SUSE (Linux|LINUX) Enterprise ", 0)
|
|
COMPILE_REGEXP (re_nld, "^Novell Linux Desktop ", 0)
|
|
COMPILE_REGEXP (re_opensuse_version, "^VERSION = (\\d+)\\.(\\d+)", 0)
|
|
COMPILE_REGEXP (re_sles_version, "^VERSION = (\\d+)", 0)
|
|
COMPILE_REGEXP (re_sles_patchlevel, "^PATCHLEVEL = (\\d+)", 0)
|
|
COMPILE_REGEXP (re_minix, "^(\\d+)\\.(\\d+)(\\.(\\d+))?", 0)
|
|
COMPILE_REGEXP (re_hurd_dev, "^/dev/(h)d(\\d+)s(\\d+)$", 0)
|
|
COMPILE_REGEXP (re_openbsd, "^OpenBSD (\\d+|\\?)\\.(\\d+|\\?)", 0)
|
|
COMPILE_REGEXP (re_openbsd_duid, "^[0-9a-f]{16}\\.[a-z]", 0)
|
|
COMPILE_REGEXP (re_openbsd_dev, "^/dev/(s|w)d([0-9])([a-z])$", 0)
|
|
COMPILE_REGEXP (re_netbsd_dev, "^/dev/(l|s)d([0-9])([a-z])$", 0)
|
|
COMPILE_REGEXP (re_altlinux, " (?:(\\d+)(?:\\.(\\d+)(?:[\\.\\d]+)?)?)\\s+\\((?:[^)]+)\\)$", 0)
|
|
COMPILE_REGEXP (re_frugalware, "Frugalware (\\d+)\\.(\\d+)", 0)
|
|
COMPILE_REGEXP (re_pldlinux, "(\\d+)\\.(\\d+) PLD Linux", 0)
|
|
|
|
static void check_architecture (guestfs_h *g, struct inspect_fs *fs);
|
|
static int check_hostname_unix (guestfs_h *g, struct inspect_fs *fs);
|
|
static int check_hostname_redhat (guestfs_h *g, struct inspect_fs *fs);
|
|
static int check_hostname_freebsd (guestfs_h *g, struct inspect_fs *fs);
|
|
static int check_fstab (guestfs_h *g, struct inspect_fs *fs);
|
|
static void add_fstab_entry (guestfs_h *g, struct inspect_fs *fs,
|
|
const char *mountable, const char *mp);
|
|
static char *resolve_fstab_device (guestfs_h *g, const char *spec,
|
|
Hash_table *md_map,
|
|
enum inspect_os_type os_type);
|
|
static int inspect_with_augeas (guestfs_h *g, struct inspect_fs *fs, const char **configfiles, int (*f) (guestfs_h *, struct inspect_fs *));
|
|
static void canonical_mountpoint (char *mp);
|
|
|
|
/* Hash structure for uuid->path lookups */
|
|
typedef struct md_uuid {
|
|
uint32_t uuid[4];
|
|
char *path;
|
|
} md_uuid;
|
|
|
|
static size_t uuid_hash(const void *x, size_t table_size);
|
|
static bool uuid_cmp(const void *x, const void *y);
|
|
static void md_uuid_free(void *x);
|
|
|
|
static int parse_uuid(const char *str, uint32_t *uuid);
|
|
|
|
/* Hash structure for path(mdadm)->path(appliance) lookup */
|
|
typedef struct {
|
|
char *mdadm;
|
|
char *app;
|
|
} mdadm_app;
|
|
|
|
static size_t mdadm_app_hash(const void *x, size_t table_size);
|
|
static bool mdadm_app_cmp(const void *x, const void *y);
|
|
static void mdadm_app_free(void *x);
|
|
|
|
static ssize_t map_app_md_devices (guestfs_h *g, Hash_table **map);
|
|
static int map_md_devices(guestfs_h *g, Hash_table **map);
|
|
|
|
/* Set fs->product_name to the first line of the release file. */
|
|
static int
|
|
parse_release_file (guestfs_h *g, struct inspect_fs *fs,
|
|
const char *release_filename)
|
|
{
|
|
fs->product_name = guestfs_int_first_line_of_file (g, release_filename);
|
|
if (fs->product_name == NULL)
|
|
return -1;
|
|
if (STREQ (fs->product_name, "")) {
|
|
free (fs->product_name);
|
|
fs->product_name = NULL;
|
|
error (g, _("release file %s is empty or malformed"), release_filename);
|
|
return -1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/* Parse a os-release file.
|
|
*
|
|
* Only few fields are parsed, falling back to the usual detection if we
|
|
* cannot read all of them.
|
|
*
|
|
* For the format of os-release, see also:
|
|
* http://www.freedesktop.org/software/systemd/man/os-release.html
|
|
*/
|
|
static int
|
|
parse_os_release (guestfs_h *g, struct inspect_fs *fs, const char *filename)
|
|
{
|
|
int64_t size;
|
|
CLEANUP_FREE_STRING_LIST char **lines = NULL;
|
|
size_t i;
|
|
enum inspect_os_distro distro = OS_DISTRO_UNKNOWN;
|
|
CLEANUP_FREE char *product_name = NULL;
|
|
struct version version;
|
|
guestfs_int_version_from_values (&version, -1, -1, 0);
|
|
|
|
/* Don't trust guestfs_read_lines not to break with very large files.
|
|
* Check the file size is something reasonable first.
|
|
*/
|
|
size = guestfs_filesize (g, filename);
|
|
if (size == -1)
|
|
/* guestfs_filesize failed and has already set error in handle */
|
|
return -1;
|
|
if (size > MAX_SMALL_FILE_SIZE) {
|
|
error (g, _("size of %s is unreasonably large (%" PRIi64 " bytes)"),
|
|
filename, size);
|
|
return -1;
|
|
}
|
|
|
|
lines = guestfs_read_lines (g, filename);
|
|
if (lines == NULL)
|
|
return -1;
|
|
|
|
for (i = 0; lines[i] != NULL; ++i) {
|
|
const char *line = lines[i];
|
|
const char *value;
|
|
size_t value_len;
|
|
|
|
if (line[0] == '#')
|
|
continue;
|
|
|
|
value = strchr (line, '=');
|
|
if (value == NULL)
|
|
continue;
|
|
|
|
++value;
|
|
value_len = strlen (line) - (value - line);
|
|
if (value_len > 1 && value[0] == '"' && value[value_len-1] == '"') {
|
|
++value;
|
|
value_len -= 2;
|
|
}
|
|
|
|
#define VALUE_IS(a) STREQLEN(value, a, value_len)
|
|
if (STRPREFIX (line, "ID=")) {
|
|
if (VALUE_IS ("alpine"))
|
|
distro = OS_DISTRO_ALPINE_LINUX;
|
|
else if (VALUE_IS ("altlinux"))
|
|
distro = OS_DISTRO_ALTLINUX;
|
|
else if (VALUE_IS ("arch"))
|
|
distro = OS_DISTRO_ARCHLINUX;
|
|
else if (VALUE_IS ("centos"))
|
|
distro = OS_DISTRO_CENTOS;
|
|
else if (VALUE_IS ("coreos"))
|
|
distro = OS_DISTRO_COREOS;
|
|
else if (VALUE_IS ("debian"))
|
|
distro = OS_DISTRO_DEBIAN;
|
|
else if (VALUE_IS ("fedora"))
|
|
distro = OS_DISTRO_FEDORA;
|
|
else if (VALUE_IS ("frugalware"))
|
|
distro = OS_DISTRO_FRUGALWARE;
|
|
else if (VALUE_IS ("mageia"))
|
|
distro = OS_DISTRO_MAGEIA;
|
|
else if (VALUE_IS ("opensuse"))
|
|
distro = OS_DISTRO_OPENSUSE;
|
|
else if (VALUE_IS ("pld"))
|
|
distro = OS_DISTRO_PLD_LINUX;
|
|
else if (VALUE_IS ("rhel"))
|
|
distro = OS_DISTRO_RHEL;
|
|
else if (VALUE_IS ("sles") || VALUE_IS ("sled"))
|
|
distro = OS_DISTRO_SLES;
|
|
else if (VALUE_IS ("ubuntu"))
|
|
distro = OS_DISTRO_UBUNTU;
|
|
else if (VALUE_IS ("void"))
|
|
distro = OS_DISTRO_VOID_LINUX;
|
|
} else if (STRPREFIX (line, "PRETTY_NAME=")) {
|
|
free (product_name);
|
|
product_name = safe_strndup (g, value, value_len);
|
|
} else if (STRPREFIX (line, "VERSION_ID=")) {
|
|
CLEANUP_FREE char *buf =
|
|
safe_asprintf (g, "%.*s", (int) value_len, value);
|
|
if (guestfs_int_version_from_x_y_or_x (g, &version, buf) == -1)
|
|
return -1;
|
|
}
|
|
#undef VALUE_IS
|
|
}
|
|
|
|
/* If we haven't got all the fields, exit right away. */
|
|
if (distro == OS_DISTRO_UNKNOWN || product_name == NULL)
|
|
return 0;
|
|
|
|
if (version.v_major == -1 || version.v_minor == -1) {
|
|
/* Void Linux has no VERSION_ID (yet), but since it's a rolling
|
|
* distro and has no other version/release-like file. */
|
|
if (distro == OS_DISTRO_VOID_LINUX)
|
|
version_init_null (&version);
|
|
else
|
|
return 0;
|
|
}
|
|
|
|
/* Apparently, os-release in Debian and CentOS does not provide the full
|
|
* version number in VERSION_ID, but just the "major" part of it.
|
|
* Hence, if version.v_minor is 0, act as there was no information in
|
|
* os-release, which will continue the inspection using the release files
|
|
* as done previously.
|
|
*/
|
|
if ((distro == OS_DISTRO_DEBIAN || distro == OS_DISTRO_CENTOS) &&
|
|
version.v_minor == 0)
|
|
return 0;
|
|
|
|
/* We got everything, so set the fields and report the inspection
|
|
* was successful.
|
|
*/
|
|
fs->distro = distro;
|
|
fs->product_name = product_name;
|
|
product_name = NULL;
|
|
fs->version = version;
|
|
|
|
return 1;
|
|
}
|
|
|
|
/* Ubuntu has /etc/lsb-release containing:
|
|
* DISTRIB_ID=Ubuntu # Distro
|
|
* DISTRIB_RELEASE=10.04 # Version
|
|
* DISTRIB_CODENAME=lucid
|
|
* DISTRIB_DESCRIPTION="Ubuntu 10.04.1 LTS" # Product name
|
|
*
|
|
* [Ubuntu-derived ...] Linux Mint was found to have this:
|
|
* DISTRIB_ID=LinuxMint
|
|
* DISTRIB_RELEASE=10
|
|
* DISTRIB_CODENAME=julia
|
|
* DISTRIB_DESCRIPTION="Linux Mint 10 Julia"
|
|
* Linux Mint also has /etc/linuxmint/info with more information,
|
|
* but we can use the LSB file.
|
|
*
|
|
* Mandriva has:
|
|
* LSB_VERSION=lsb-4.0-amd64:lsb-4.0-noarch
|
|
* DISTRIB_ID=MandrivaLinux
|
|
* DISTRIB_RELEASE=2010.1
|
|
* DISTRIB_CODENAME=Henry_Farman
|
|
* DISTRIB_DESCRIPTION="Mandriva Linux 2010.1"
|
|
* Mandriva also has a normal release file called /etc/mandriva-release.
|
|
*
|
|
* CoreOS has a /etc/lsb-release link to /usr/share/coreos/lsb-release containing:
|
|
* DISTRIB_ID=CoreOS
|
|
* DISTRIB_RELEASE=647.0.0
|
|
* DISTRIB_CODENAME="Red Dog"
|
|
* DISTRIB_DESCRIPTION="CoreOS 647.0.0"
|
|
*/
|
|
static int
|
|
parse_lsb_release (guestfs_h *g, struct inspect_fs *fs, const char *filename)
|
|
{
|
|
int64_t size;
|
|
CLEANUP_FREE_STRING_LIST char **lines = NULL;
|
|
size_t i;
|
|
int r = 0;
|
|
|
|
/* Don't trust guestfs_head_n not to break with very large files.
|
|
* Check the file size is something reasonable first.
|
|
*/
|
|
size = guestfs_filesize (g, filename);
|
|
if (size == -1)
|
|
/* guestfs_filesize failed and has already set error in handle */
|
|
return -1;
|
|
if (size > MAX_SMALL_FILE_SIZE) {
|
|
error (g, _("size of %s is unreasonably large (%" PRIi64 " bytes)"),
|
|
filename, size);
|
|
return -1;
|
|
}
|
|
|
|
lines = guestfs_head_n (g, 10, filename);
|
|
if (lines == NULL)
|
|
return -1;
|
|
|
|
for (i = 0; lines[i] != NULL; ++i) {
|
|
if (fs->distro == 0 &&
|
|
STREQ (lines[i], "DISTRIB_ID=Ubuntu")) {
|
|
fs->distro = OS_DISTRO_UBUNTU;
|
|
r = 1;
|
|
}
|
|
else if (fs->distro == 0 &&
|
|
STREQ (lines[i], "DISTRIB_ID=LinuxMint")) {
|
|
fs->distro = OS_DISTRO_LINUX_MINT;
|
|
r = 1;
|
|
}
|
|
else if (fs->distro == 0 &&
|
|
STREQ (lines[i], "DISTRIB_ID=MandrivaLinux")) {
|
|
fs->distro = OS_DISTRO_MANDRIVA;
|
|
r = 1;
|
|
}
|
|
else if (fs->distro == 0 &&
|
|
STREQ (lines[i], "DISTRIB_ID=\"Mageia\"")) {
|
|
fs->distro = OS_DISTRO_MAGEIA;
|
|
r = 1;
|
|
}
|
|
else if (fs->distro == 0 &&
|
|
STREQ (lines[i], "DISTRIB_ID=CoreOS")) {
|
|
fs->distro = OS_DISTRO_COREOS;
|
|
r = 1;
|
|
}
|
|
else if (STRPREFIX (lines[i], "DISTRIB_RELEASE=")) {
|
|
if (guestfs_int_version_from_x_y_or_x (g, &fs->version, &lines[i][16]) == -1)
|
|
return -1;
|
|
}
|
|
else if (fs->product_name == NULL &&
|
|
(STRPREFIX (lines[i], "DISTRIB_DESCRIPTION=\"") ||
|
|
STRPREFIX (lines[i], "DISTRIB_DESCRIPTION='"))) {
|
|
const size_t len = strlen (lines[i]) - 21 - 1;
|
|
fs->product_name = safe_strndup (g, &lines[i][21], len);
|
|
r = 1;
|
|
}
|
|
else if (fs->product_name == NULL &&
|
|
STRPREFIX (lines[i], "DISTRIB_DESCRIPTION=")) {
|
|
const size_t len = strlen (lines[i]) - 20;
|
|
fs->product_name = safe_strndup (g, &lines[i][20], len);
|
|
r = 1;
|
|
}
|
|
}
|
|
|
|
/* The unnecessary construct in the next line is required to
|
|
* workaround -Wstrict-overflow warning in gcc 4.5.1.
|
|
*/
|
|
return r ? 1 : 0;
|
|
}
|
|
|
|
static int
|
|
parse_suse_release (guestfs_h *g, struct inspect_fs *fs, const char *filename)
|
|
{
|
|
int64_t size;
|
|
CLEANUP_FREE_STRING_LIST char **lines = NULL;
|
|
int r = -1;
|
|
|
|
/* Don't trust guestfs_head_n not to break with very large files.
|
|
* Check the file size is something reasonable first.
|
|
*/
|
|
size = guestfs_filesize (g, filename);
|
|
if (size == -1)
|
|
/* guestfs_filesize failed and has already set error in handle */
|
|
return -1;
|
|
if (size > MAX_SMALL_FILE_SIZE) {
|
|
error (g, _("size of %s is unreasonably large (%" PRIi64 " bytes)"),
|
|
filename, size);
|
|
return -1;
|
|
}
|
|
|
|
lines = guestfs_head_n (g, 10, filename);
|
|
if (lines == NULL)
|
|
return -1;
|
|
|
|
if (lines[0] == NULL)
|
|
goto out;
|
|
|
|
/* First line is dist release name */
|
|
fs->product_name = safe_strdup (g, lines[0]);
|
|
|
|
/* Match SLES first because openSuSE regex overlaps some SLES release strings */
|
|
if (match (g, fs->product_name, re_sles) || match (g, fs->product_name, re_nld)) {
|
|
char *major, *minor;
|
|
|
|
fs->distro = OS_DISTRO_SLES;
|
|
|
|
/* Second line contains version string */
|
|
if (lines[1] == NULL)
|
|
goto out;
|
|
major = match1 (g, lines[1], re_sles_version);
|
|
if (major == NULL)
|
|
goto out;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1)
|
|
goto out;
|
|
|
|
/* Third line contains service pack string */
|
|
if (lines[2] == NULL)
|
|
goto out;
|
|
minor = match1 (g, lines[2], re_sles_patchlevel);
|
|
if (minor == NULL)
|
|
goto out;
|
|
fs->version.v_minor = guestfs_int_parse_unsigned_int (g, minor);
|
|
free (minor);
|
|
if (fs->version.v_minor == -1)
|
|
goto out;
|
|
}
|
|
else if (match (g, fs->product_name, re_opensuse)) {
|
|
fs->distro = OS_DISTRO_OPENSUSE;
|
|
|
|
/* Second line contains version string */
|
|
if (lines[1] == NULL)
|
|
goto out;
|
|
if (guestfs_int_version_from_x_y_re (g, &fs->version, lines[1],
|
|
re_opensuse_version) == -1)
|
|
goto out;
|
|
}
|
|
|
|
r = 0;
|
|
|
|
out:
|
|
return r;
|
|
}
|
|
|
|
/* The currently mounted device is known to be a Linux root. Try to
|
|
* determine from this the distro, version, etc. Also parse
|
|
* /etc/fstab to determine the arrangement of mountpoints and
|
|
* associated devices.
|
|
*/
|
|
int
|
|
guestfs_int_check_linux_root (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
int r;
|
|
char *major, *minor;
|
|
|
|
fs->type = OS_TYPE_LINUX;
|
|
|
|
if (guestfs_is_file_opts (g, "/etc/os-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
r = parse_os_release (g, fs, "/etc/os-release");
|
|
if (r == -1) /* error */
|
|
return -1;
|
|
if (r == 1) /* ok - detected the release from this file */
|
|
goto skip_release_checks;
|
|
}
|
|
|
|
if (guestfs_is_file_opts (g, "/etc/lsb-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
r = parse_lsb_release (g, fs, "/etc/lsb-release");
|
|
if (r == -1) /* error */
|
|
return -1;
|
|
if (r == 1) /* ok - detected the release from this file */
|
|
goto skip_release_checks;
|
|
}
|
|
|
|
/* RHEL-based distros include a "/etc/redhat-release" file, hence their
|
|
* checks need to be performed before the Red-Hat one.
|
|
*/
|
|
if (guestfs_is_file_opts (g, "/etc/oracle-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
|
|
fs->distro = OS_DISTRO_ORACLE_LINUX;
|
|
|
|
if (parse_release_file (g, fs, "/etc/oracle-release") == -1)
|
|
return -1;
|
|
|
|
if (match2 (g, fs->product_name, re_oracle_linux_old, &major, &minor) ||
|
|
match2 (g, fs->product_name, re_oracle_linux, &major, &minor)) {
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1) {
|
|
free (minor);
|
|
return -1;
|
|
}
|
|
fs->version.v_minor = guestfs_int_parse_unsigned_int (g, minor);
|
|
free (minor);
|
|
if (fs->version.v_minor == -1)
|
|
return -1;
|
|
} else if ((major = match1 (g, fs->product_name, re_oracle_linux_no_minor)) != NULL) {
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1)
|
|
return -1;
|
|
fs->version.v_minor = 0;
|
|
}
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/centos-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_CENTOS;
|
|
|
|
if (parse_release_file (g, fs, "/etc/centos-release") == -1)
|
|
return -1;
|
|
|
|
if (match2 (g, fs->product_name, re_centos_old, &major, &minor) ||
|
|
match2 (g, fs->product_name, re_centos, &major, &minor)) {
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1) {
|
|
free (minor);
|
|
return -1;
|
|
}
|
|
fs->version.v_minor = guestfs_int_parse_unsigned_int (g, minor);
|
|
free (minor);
|
|
if (fs->version.v_minor == -1)
|
|
return -1;
|
|
}
|
|
else if ((major = match1 (g, fs->product_name, re_centos_no_minor)) != NULL) {
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1)
|
|
return -1;
|
|
fs->version.v_minor = 0;
|
|
}
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/altlinux-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_ALTLINUX;
|
|
|
|
if (parse_release_file (g, fs, "/etc/altlinux-release") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_version_from_x_y_re (g, &fs->version, fs->product_name,
|
|
re_altlinux) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/redhat-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_REDHAT_BASED; /* Something generic Red Hat-like. */
|
|
|
|
if (parse_release_file (g, fs, "/etc/redhat-release") == -1)
|
|
return -1;
|
|
|
|
if ((major = match1 (g, fs->product_name, re_fedora)) != NULL) {
|
|
fs->distro = OS_DISTRO_FEDORA;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1)
|
|
return -1;
|
|
}
|
|
else if (match2 (g, fs->product_name, re_rhel_old, &major, &minor) ||
|
|
match2 (g, fs->product_name, re_rhel, &major, &minor)) {
|
|
fs->distro = OS_DISTRO_RHEL;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1) {
|
|
free (minor);
|
|
return -1;
|
|
}
|
|
fs->version.v_minor = guestfs_int_parse_unsigned_int (g, minor);
|
|
free (minor);
|
|
if (fs->version.v_minor == -1)
|
|
return -1;
|
|
}
|
|
else if ((major = match1 (g, fs->product_name, re_rhel_no_minor)) != NULL) {
|
|
fs->distro = OS_DISTRO_RHEL;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1)
|
|
return -1;
|
|
fs->version.v_minor = 0;
|
|
}
|
|
else if (match2 (g, fs->product_name, re_centos_old, &major, &minor) ||
|
|
match2 (g, fs->product_name, re_centos, &major, &minor)) {
|
|
fs->distro = OS_DISTRO_CENTOS;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1) {
|
|
free (minor);
|
|
return -1;
|
|
}
|
|
fs->version.v_minor = guestfs_int_parse_unsigned_int (g, minor);
|
|
free (minor);
|
|
if (fs->version.v_minor == -1)
|
|
return -1;
|
|
}
|
|
else if ((major = match1 (g, fs->product_name, re_centos_no_minor)) != NULL) {
|
|
fs->distro = OS_DISTRO_CENTOS;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1)
|
|
return -1;
|
|
fs->version.v_minor = 0;
|
|
}
|
|
else if (match2 (g, fs->product_name, re_scientific_linux_old, &major, &minor) ||
|
|
match2 (g, fs->product_name, re_scientific_linux, &major, &minor)) {
|
|
fs->distro = OS_DISTRO_SCIENTIFIC_LINUX;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1) {
|
|
free (minor);
|
|
return -1;
|
|
}
|
|
fs->version.v_minor = guestfs_int_parse_unsigned_int (g, minor);
|
|
free (minor);
|
|
if (fs->version.v_minor == -1)
|
|
return -1;
|
|
}
|
|
else if ((major = match1 (g, fs->product_name, re_scientific_linux_no_minor)) != NULL) {
|
|
fs->distro = OS_DISTRO_SCIENTIFIC_LINUX;
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
free (major);
|
|
if (fs->version.v_major == -1)
|
|
return -1;
|
|
fs->version.v_minor = 0;
|
|
}
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/debian_version",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_DEBIAN;
|
|
|
|
if (parse_release_file (g, fs, "/etc/debian_version") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/pardus-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_PARDUS;
|
|
|
|
if (parse_release_file (g, fs, "/etc/pardus-release") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/arch-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_ARCHLINUX;
|
|
|
|
/* /etc/arch-release file is empty and I can't see a way to
|
|
* determine the actual release or product string.
|
|
*/
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/gentoo-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_GENTOO;
|
|
|
|
if (parse_release_file (g, fs, "/etc/gentoo-release") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/meego-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_MEEGO;
|
|
|
|
if (parse_release_file (g, fs, "/etc/meego-release") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/slackware-version",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_SLACKWARE;
|
|
|
|
if (parse_release_file (g, fs, "/etc/slackware-version") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/ttylinux-target",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_TTYLINUX;
|
|
|
|
if (parse_release_file (g, fs, "/etc/ttylinux-target") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/SuSE-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_SUSE_BASED;
|
|
|
|
if (parse_suse_release (g, fs, "/etc/SuSE-release") == -1)
|
|
return -1;
|
|
|
|
}
|
|
/* CirrOS versions providing a own version file.
|
|
*/
|
|
else if (guestfs_is_file_opts (g, "/etc/cirros/version",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_CIRROS;
|
|
|
|
if (parse_release_file (g, fs, "/etc/cirros/version") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
/* Buildroot (http://buildroot.net) is an embedded Linux distro
|
|
* toolkit. It is used by specific distros such as Cirros.
|
|
*/
|
|
else if (guestfs_is_file_opts (g, "/etc/br-version",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
if (guestfs_is_file_opts (g, "/usr/share/cirros/logo",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0)
|
|
fs->distro = OS_DISTRO_CIRROS;
|
|
else
|
|
fs->distro = OS_DISTRO_BUILDROOT;
|
|
|
|
/* /etc/br-version has the format YYYY.MM[-git/hg/svn release] */
|
|
if (parse_release_file (g, fs, "/etc/br-version") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/alpine-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_ALPINE_LINUX;
|
|
|
|
if (parse_release_file (g, fs, "/etc/alpine-release") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/frugalware-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_FRUGALWARE;
|
|
|
|
if (parse_release_file (g, fs, "/etc/frugalware-release") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_version_from_x_y_re (g, &fs->version, fs->product_name,
|
|
re_frugalware) == -1)
|
|
return -1;
|
|
}
|
|
else if (guestfs_is_file_opts (g, "/etc/pld-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_PLD_LINUX;
|
|
|
|
if (parse_release_file (g, fs, "/etc/pld-release") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_version_from_x_y_re (g, &fs->version, fs->product_name,
|
|
re_pldlinux) == -1)
|
|
return -1;
|
|
}
|
|
|
|
skip_release_checks:;
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
/* We already know /etc/fstab exists because it's part of the test
|
|
* for Linux root above. We must now parse this file to determine
|
|
* which filesystems are used by the operating system and how they
|
|
* are mounted.
|
|
*/
|
|
const char *configfiles[] = { "/etc/fstab", "/etc/mdadm.conf", NULL };
|
|
if (inspect_with_augeas (g, fs, configfiles, check_fstab) == -1)
|
|
return -1;
|
|
|
|
/* Determine hostname. */
|
|
if (check_hostname_unix (g, fs) == -1)
|
|
return -1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device looks like a Linux /usr. */
|
|
int
|
|
guestfs_int_check_linux_usr (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
int r;
|
|
|
|
fs->type = OS_TYPE_LINUX;
|
|
fs->role = OS_ROLE_USR;
|
|
|
|
if (guestfs_is_file_opts (g, "/lib/os-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
r = parse_os_release (g, fs, "/lib/os-release");
|
|
if (r == -1) /* error */
|
|
return -1;
|
|
if (r == 1) /* ok - detected the release from this file */
|
|
goto skip_release_checks;
|
|
}
|
|
|
|
skip_release_checks:;
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device is known to be a FreeBSD root. */
|
|
int
|
|
guestfs_int_check_freebsd_root (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
fs->type = OS_TYPE_FREEBSD;
|
|
fs->distro = OS_DISTRO_FREEBSD;
|
|
|
|
/* FreeBSD has no authoritative version file. The version number is
|
|
* in /etc/motd, which the system administrator might edit, but
|
|
* we'll use that anyway.
|
|
*/
|
|
|
|
if (guestfs_is_file_opts (g, "/etc/motd",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
if (parse_release_file (g, fs, "/etc/motd") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
/* We already know /etc/fstab exists because it's part of the test above. */
|
|
const char *configfiles[] = { "/etc/fstab", NULL };
|
|
if (inspect_with_augeas (g, fs, configfiles, check_fstab) == -1)
|
|
return -1;
|
|
|
|
/* Determine hostname. */
|
|
if (check_hostname_unix (g, fs) == -1)
|
|
return -1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device is maybe to be a *BSD root. */
|
|
int
|
|
guestfs_int_check_netbsd_root (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
|
|
if (guestfs_is_file_opts (g, "/etc/release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
int result;
|
|
if (parse_release_file (g, fs, "/etc/release") == -1)
|
|
return -1;
|
|
|
|
result = guestfs_int_version_from_x_y_re (g, &fs->version,
|
|
fs->product_name, re_netbsd);
|
|
if (result == -1) {
|
|
return -1;
|
|
} else if (result == 1) {
|
|
fs->type = OS_TYPE_NETBSD;
|
|
fs->distro = OS_DISTRO_NETBSD;
|
|
}
|
|
} else {
|
|
return -1;
|
|
}
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
/* We already know /etc/fstab exists because it's part of the test above. */
|
|
const char *configfiles[] = { "/etc/fstab", NULL };
|
|
if (inspect_with_augeas (g, fs, configfiles, check_fstab) == -1)
|
|
return -1;
|
|
|
|
/* Determine hostname. */
|
|
if (check_hostname_unix (g, fs) == -1)
|
|
return -1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device may be an OpenBSD root. */
|
|
int
|
|
guestfs_int_check_openbsd_root (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
if (guestfs_is_file_opts (g, "/etc/motd",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
CLEANUP_FREE char *major = NULL, *minor = NULL;
|
|
|
|
/* The first line of this file gets automatically updated at boot. */
|
|
if (parse_release_file (g, fs, "/etc/motd") == -1)
|
|
return -1;
|
|
|
|
if (match2 (g, fs->product_name, re_openbsd, &major, &minor)) {
|
|
fs->type = OS_TYPE_OPENBSD;
|
|
fs->distro = OS_DISTRO_OPENBSD;
|
|
|
|
/* Before the first boot, the first line will look like this:
|
|
*
|
|
* OpenBSD ?.? (UNKNOWN)
|
|
*/
|
|
if ((fs->product_name[8] != '?') && (fs->product_name[10] != '?')) {
|
|
fs->version.v_major = guestfs_int_parse_unsigned_int (g, major);
|
|
if (fs->version.v_major == -1)
|
|
return -1;
|
|
|
|
fs->version.v_minor = guestfs_int_parse_unsigned_int (g, minor);
|
|
if (fs->version.v_minor == -1)
|
|
return -1;
|
|
}
|
|
}
|
|
} else {
|
|
return -1;
|
|
}
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
/* We already know /etc/fstab exists because it's part of the test above. */
|
|
const char *configfiles[] = { "/etc/fstab", NULL };
|
|
if (inspect_with_augeas (g, fs, configfiles, check_fstab) == -1)
|
|
return -1;
|
|
|
|
/* Determine hostname. */
|
|
if (check_hostname_unix (g, fs) == -1)
|
|
return -1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device may be a Hurd root. Hurd has distros
|
|
* just like Linux.
|
|
*/
|
|
int
|
|
guestfs_int_check_hurd_root (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
fs->type = OS_TYPE_HURD;
|
|
|
|
if (guestfs_is_file_opts (g, "/etc/debian_version",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
fs->distro = OS_DISTRO_DEBIAN;
|
|
|
|
if (parse_release_file (g, fs, "/etc/debian_version") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_parse_major_minor (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
|
|
/* Arch Hurd also exists, but inconveniently it doesn't have
|
|
* the normal /etc/arch-release file. XXX
|
|
*/
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
if (guestfs_is_file (g, "/etc/fstab") > 0) {
|
|
const char *configfiles[] = { "/etc/fstab", NULL };
|
|
if (inspect_with_augeas (g, fs, configfiles, check_fstab) == -1)
|
|
return -1;
|
|
}
|
|
|
|
/* Determine hostname. */
|
|
if (check_hostname_unix (g, fs) == -1)
|
|
return -1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device is maybe to be a Minix root. */
|
|
int
|
|
guestfs_int_check_minix_root (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
fs->type = OS_TYPE_MINIX;
|
|
|
|
if (guestfs_is_file_opts (g, "/etc/version",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
if (parse_release_file (g, fs, "/etc/version") == -1)
|
|
return -1;
|
|
|
|
if (guestfs_int_version_from_x_y_re (g, &fs->version, fs->product_name,
|
|
re_minix) == -1)
|
|
return -1;
|
|
} else {
|
|
return -1;
|
|
}
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
/* TODO: enable fstab inspection once resolve_fstab_device implements
|
|
* the proper mapping from the Minix device names to the appliance names
|
|
*/
|
|
|
|
/* Determine hostname. */
|
|
if (check_hostname_unix (g, fs) == -1)
|
|
return -1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device is a CoreOS root. From this partition we can
|
|
* only determine the hostname. All immutable OS files are under a separate
|
|
* read-only /usr partition.
|
|
*/
|
|
int
|
|
guestfs_int_check_coreos_root (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
fs->type = OS_TYPE_LINUX;
|
|
fs->distro = OS_DISTRO_COREOS;
|
|
|
|
/* Determine hostname. */
|
|
if (check_hostname_unix (g, fs) == -1)
|
|
return -1;
|
|
|
|
/* CoreOS does not contain /etc/fstab to determine the mount points.
|
|
* Associate this filesystem with the "/" mount point.
|
|
*/
|
|
add_fstab_entry (g, fs, fs->mountable, "/");
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The currently mounted device looks like a CoreOS /usr. In CoreOS
|
|
* the read-only /usr contains the OS version. The /etc/os-release is a
|
|
* link to /usr/share/coreos/os-release.
|
|
*/
|
|
int
|
|
guestfs_int_check_coreos_usr (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
int r;
|
|
|
|
fs->type = OS_TYPE_LINUX;
|
|
fs->distro = OS_DISTRO_COREOS;
|
|
fs->role = OS_ROLE_USR;
|
|
|
|
if (guestfs_is_file_opts (g, "/lib/os-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
r = parse_os_release (g, fs, "/lib/os-release");
|
|
if (r == -1) /* error */
|
|
return -1;
|
|
if (r == 1) /* ok - detected the release from this file */
|
|
goto skip_release_checks;
|
|
}
|
|
|
|
if (guestfs_is_file_opts (g, "/share/coreos/lsb-release",
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
r = parse_lsb_release (g, fs, "/share/coreos/lsb-release");
|
|
if (r == -1) /* error */
|
|
return -1;
|
|
}
|
|
|
|
skip_release_checks:;
|
|
|
|
/* Determine the architecture. */
|
|
check_architecture (g, fs);
|
|
|
|
/* CoreOS does not contain /etc/fstab to determine the mount points.
|
|
* Associate this filesystem with the "/usr" mount point.
|
|
*/
|
|
add_fstab_entry (g, fs, fs->mountable, "/usr");
|
|
|
|
return 0;
|
|
}
|
|
|
|
static void
|
|
check_architecture (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
const char *binaries[] =
|
|
{ "/bin/bash", "/bin/ls", "/bin/echo", "/bin/rm", "/bin/sh" };
|
|
size_t i;
|
|
char *arch = NULL;
|
|
|
|
for (i = 0; i < sizeof binaries / sizeof binaries[0]; ++i) {
|
|
/* Allow symlinks when checking the binaries:,so in case they are
|
|
* relative ones (which can be resolved within the same partition),
|
|
* then we can check the architecture of their target.
|
|
*/
|
|
if (guestfs_is_file_opts (g, binaries[i],
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) > 0) {
|
|
CLEANUP_FREE char *resolved = NULL;
|
|
|
|
/* Ignore errors from realpath and file_architecture calls. */
|
|
guestfs_push_error_handler (g, NULL, NULL);
|
|
resolved = guestfs_realpath (g, binaries[i]);
|
|
/* If is_file above succeeded realpath should too, but better
|
|
* be safe than sorry.
|
|
*/
|
|
if (resolved)
|
|
arch = guestfs_file_architecture (g, resolved);
|
|
guestfs_pop_error_handler (g);
|
|
|
|
if (arch) {
|
|
/* String will be owned by handle, freed by
|
|
* guestfs_int_free_inspect_info.
|
|
*/
|
|
fs->arch = arch;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Try several methods to determine the hostname from a Linux or
|
|
* FreeBSD guest. Note that type and distro have been set, so we can
|
|
* use that information to direct the search.
|
|
*/
|
|
static int
|
|
check_hostname_unix (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
switch (fs->type) {
|
|
case OS_TYPE_LINUX:
|
|
case OS_TYPE_HURD:
|
|
/* Red Hat-derived would be in /etc/sysconfig/network or
|
|
* /etc/hostname (RHEL 7+, F18+). Debian-derived in the file
|
|
* /etc/hostname. Very old Debian and SUSE use /etc/HOSTNAME.
|
|
* It's best to just look for each of these files in turn, rather
|
|
* than try anything clever based on distro.
|
|
*/
|
|
if (guestfs_is_file (g, "/etc/HOSTNAME")) {
|
|
fs->hostname = guestfs_int_first_line_of_file (g, "/etc/HOSTNAME");
|
|
if (fs->hostname == NULL)
|
|
return -1;
|
|
if (STREQ (fs->hostname, "")) {
|
|
free (fs->hostname);
|
|
fs->hostname = NULL;
|
|
}
|
|
}
|
|
|
|
if (!fs->hostname && guestfs_is_file (g, "/etc/hostname")) {
|
|
fs->hostname = guestfs_int_first_line_of_file (g, "/etc/hostname");
|
|
if (fs->hostname == NULL)
|
|
return -1;
|
|
if (STREQ (fs->hostname, "")) {
|
|
free (fs->hostname);
|
|
fs->hostname = NULL;
|
|
}
|
|
}
|
|
|
|
if (!fs->hostname && guestfs_is_file (g, "/etc/sysconfig/network")) {
|
|
const char *configfiles[] = { "/etc/sysconfig/network", NULL };
|
|
if (inspect_with_augeas (g, fs, configfiles,
|
|
check_hostname_redhat) == -1)
|
|
return -1;
|
|
}
|
|
break;
|
|
|
|
case OS_TYPE_FREEBSD:
|
|
case OS_TYPE_NETBSD:
|
|
/* /etc/rc.conf contains the hostname, but there is no Augeas lens
|
|
* for this file.
|
|
*/
|
|
if (guestfs_is_file (g, "/etc/rc.conf")) {
|
|
if (check_hostname_freebsd (g, fs) == -1)
|
|
return -1;
|
|
}
|
|
break;
|
|
|
|
case OS_TYPE_OPENBSD:
|
|
if (guestfs_is_file (g, "/etc/myname")) {
|
|
fs->hostname = guestfs_int_first_line_of_file (g, "/etc/myname");
|
|
if (fs->hostname == NULL)
|
|
return -1;
|
|
if (STREQ (fs->hostname, "")) {
|
|
free (fs->hostname);
|
|
fs->hostname = NULL;
|
|
}
|
|
}
|
|
break;
|
|
|
|
case OS_TYPE_MINIX:
|
|
if (guestfs_is_file (g, "/etc/hostname.file")) {
|
|
fs->hostname = guestfs_int_first_line_of_file (g, "/etc/hostname.file");
|
|
if (fs->hostname == NULL)
|
|
return -1;
|
|
if (STREQ (fs->hostname, "")) {
|
|
free (fs->hostname);
|
|
fs->hostname = NULL;
|
|
}
|
|
}
|
|
break;
|
|
|
|
case OS_TYPE_WINDOWS: /* not here, see check_windows_system_registry */
|
|
case OS_TYPE_DOS:
|
|
case OS_TYPE_UNKNOWN:
|
|
/* nothing */;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* Parse the hostname from /etc/sysconfig/network. This must be
|
|
* called from the inspect_with_augeas wrapper. Note that F18+ and
|
|
* RHEL7+ use /etc/hostname just like Debian.
|
|
*/
|
|
static int
|
|
check_hostname_redhat (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
char *hostname;
|
|
|
|
/* Errors here are not fatal (RHBZ#726739), since it could be
|
|
* just missing HOSTNAME field in the file.
|
|
*/
|
|
guestfs_push_error_handler (g, NULL, NULL);
|
|
hostname = guestfs_aug_get (g, "/files/etc/sysconfig/network/HOSTNAME");
|
|
guestfs_pop_error_handler (g);
|
|
|
|
/* This is freed by guestfs_int_free_inspect_info. Note that hostname
|
|
* could be NULL because we ignored errors above.
|
|
*/
|
|
fs->hostname = hostname;
|
|
return 0;
|
|
}
|
|
|
|
/* Parse the hostname from /etc/rc.conf. On FreeBSD this file
|
|
* contains comments, blank lines and:
|
|
* hostname="freebsd8.example.com"
|
|
* ifconfig_re0="DHCP"
|
|
* keymap="uk.iso"
|
|
* sshd_enable="YES"
|
|
*/
|
|
static int
|
|
check_hostname_freebsd (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
const char *filename = "/etc/rc.conf";
|
|
int64_t size;
|
|
CLEANUP_FREE_STRING_LIST char **lines = NULL;
|
|
size_t i;
|
|
|
|
/* Don't trust guestfs_read_lines not to break with very large files.
|
|
* Check the file size is something reasonable first.
|
|
*/
|
|
size = guestfs_filesize (g, filename);
|
|
if (size == -1)
|
|
/* guestfs_filesize failed and has already set error in handle */
|
|
return -1;
|
|
if (size > MAX_SMALL_FILE_SIZE) {
|
|
error (g, _("size of %s is unreasonably large (%" PRIi64 " bytes)"),
|
|
filename, size);
|
|
return -1;
|
|
}
|
|
|
|
lines = guestfs_read_lines (g, filename);
|
|
if (lines == NULL)
|
|
return -1;
|
|
|
|
for (i = 0; lines[i] != NULL; ++i) {
|
|
if (STRPREFIX (lines[i], "hostname=\"") ||
|
|
STRPREFIX (lines[i], "hostname='")) {
|
|
const size_t len = strlen (lines[i]) - 10 - 1;
|
|
fs->hostname = safe_strndup (g, &lines[i][10], len);
|
|
break;
|
|
} else if (STRPREFIX (lines[i], "hostname=")) {
|
|
const size_t len = strlen (lines[i]) - 9;
|
|
fs->hostname = safe_strndup (g, &lines[i][9], len);
|
|
break;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
check_fstab (guestfs_h *g, struct inspect_fs *fs)
|
|
{
|
|
CLEANUP_FREE_STRING_LIST char **entries = NULL;
|
|
char **entry;
|
|
char augpath[256];
|
|
CLEANUP_HASH_FREE Hash_table *md_map = NULL;
|
|
bool is_bsd = (fs->type == OS_TYPE_FREEBSD ||
|
|
fs->type == OS_TYPE_NETBSD ||
|
|
fs->type == OS_TYPE_OPENBSD);
|
|
|
|
/* Generate a map of MD device paths listed in /etc/mdadm.conf to MD device
|
|
* paths in the guestfs appliance */
|
|
if (map_md_devices (g, &md_map) == -1) return -1;
|
|
|
|
entries = guestfs_aug_match (g, "/files/etc/fstab/*[label() != '#comment']");
|
|
if (entries == NULL)
|
|
return -1;
|
|
|
|
for (entry = entries; *entry != NULL; entry++) {
|
|
CLEANUP_FREE char *spec = NULL;
|
|
CLEANUP_FREE char *mp = NULL;
|
|
CLEANUP_FREE char *mountable = NULL;
|
|
CLEANUP_FREE char *vfstype = NULL;
|
|
|
|
snprintf (augpath, sizeof augpath, "%s/spec", *entry);
|
|
spec = guestfs_aug_get (g, augpath);
|
|
if (spec == NULL)
|
|
return -1;
|
|
|
|
/* Ignore /dev/fd (floppy disks) (RHBZ#642929) and CD-ROM drives.
|
|
*
|
|
* /dev/iso9660/FREEBSD_INSTALL can be found in FreeBSDs installation
|
|
* discs.
|
|
*/
|
|
if ((STRPREFIX (spec, "/dev/fd") && c_isdigit (spec[7])) ||
|
|
STREQ (spec, "/dev/floppy") ||
|
|
STREQ (spec, "/dev/cdrom") ||
|
|
STRPREFIX (spec, "/dev/iso9660/"))
|
|
continue;
|
|
|
|
snprintf (augpath, sizeof augpath, "%s/file", *entry);
|
|
mp = guestfs_aug_get (g, augpath);
|
|
if (mp == NULL)
|
|
return -1;
|
|
|
|
/* Canonicalize the path, so "///usr//local//" -> "/usr/local" */
|
|
canonical_mountpoint (mp);
|
|
|
|
/* Ignore certain mountpoints. */
|
|
if (STRPREFIX (mp, "/dev/") ||
|
|
STREQ (mp, "/dev") ||
|
|
STRPREFIX (mp, "/media/") ||
|
|
STRPREFIX (mp, "/proc/") ||
|
|
STREQ (mp, "/proc") ||
|
|
STRPREFIX (mp, "/selinux/") ||
|
|
STREQ (mp, "/selinux") ||
|
|
STRPREFIX (mp, "/sys/") ||
|
|
STREQ (mp, "/sys"))
|
|
continue;
|
|
|
|
/* Resolve UUID= and LABEL= to the actual device. */
|
|
if (STRPREFIX (spec, "UUID=")) {
|
|
CLEANUP_FREE char *s = guestfs_int_shell_unquote (&spec[5]);
|
|
if (s == NULL) { perrorf (g, "guestfs_int_shell_unquote"); return -1; }
|
|
mountable = guestfs_findfs_uuid (g, s);
|
|
}
|
|
else if (STRPREFIX (spec, "LABEL=")) {
|
|
CLEANUP_FREE char *s = guestfs_int_shell_unquote (&spec[6]);
|
|
if (s == NULL) { perrorf (g, "guestfs_int_shell_unquote"); return -1; }
|
|
mountable = guestfs_findfs_label (g, s);
|
|
}
|
|
/* Ignore "/.swap" (Pardus) and pseudo-devices like "tmpfs". */
|
|
else if (STREQ (spec, "/dev/root") || (is_bsd && STREQ (mp, "/")))
|
|
/* Resolve /dev/root to the current device.
|
|
* Do the same for the / partition of the *BSD systems, since the
|
|
* BSD -> Linux device translation is not straight forward.
|
|
*/
|
|
mountable = safe_strdup (g, fs->mountable);
|
|
else if (STRPREFIX (spec, "/dev/"))
|
|
/* Resolve guest block device names. */
|
|
mountable = resolve_fstab_device (g, spec, md_map, fs->type);
|
|
else if (match (g, spec, re_openbsd_duid)) {
|
|
/* In OpenBSD's fstab you can specify partitions on a disk by appending a
|
|
* period and a partition letter to a Disklable Unique Identifier. The
|
|
* DUID is a 16 hex digit field found in the OpenBSD's altered BSD
|
|
* disklabel. For more info see here:
|
|
* http://www.openbsd.org/faq/faq14.html#intro
|
|
*/
|
|
char device[10]; /* /dev/sd[0-9][a-z] */
|
|
char part = spec[17];
|
|
|
|
/* We cannot peep into disklables, we can only assume that this is the
|
|
* first disk.
|
|
*/
|
|
snprintf(device, 10, "%s%c", "/dev/sd0", part);
|
|
mountable = resolve_fstab_device (g, device, md_map, fs->type);
|
|
}
|
|
|
|
/* If we haven't resolved the device successfully by this point,
|
|
* we don't care, just ignore it.
|
|
*/
|
|
if (mountable == NULL)
|
|
continue;
|
|
|
|
snprintf (augpath, sizeof augpath, "%s/vfstype", *entry);
|
|
vfstype = guestfs_aug_get (g, augpath);
|
|
if (vfstype == NULL) return -1;
|
|
|
|
if (STREQ (vfstype, "btrfs")) {
|
|
size_t i;
|
|
|
|
snprintf (augpath, sizeof augpath, "%s/opt", *entry);
|
|
CLEANUP_FREE_STRING_LIST char **opts = guestfs_aug_match (g, augpath);
|
|
if (opts == NULL) return -1;
|
|
|
|
for (i = 0; opts[i] != NULL; ++i) {
|
|
CLEANUP_FREE char *optname = NULL, *optvalue = NULL, *subvol = NULL;
|
|
char *old_mountable;
|
|
|
|
optname = guestfs_aug_get (g, opts[i]);
|
|
if (optname == NULL) return -1;
|
|
|
|
if (STREQ (optname, "subvol")) {
|
|
optvalue = safe_asprintf (g, "%s/value", opts[i]);
|
|
|
|
subvol = guestfs_aug_get (g, optvalue);
|
|
if (subvol == NULL) return -1;
|
|
|
|
old_mountable = mountable;
|
|
mountable = safe_asprintf (g, "btrfsvol:%s/%s", mountable, subvol);
|
|
free (old_mountable);
|
|
}
|
|
}
|
|
}
|
|
|
|
add_fstab_entry (g, fs, mountable, mp);
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* Add a filesystem and possibly a mountpoint entry for
|
|
* the root filesystem 'fs'.
|
|
*
|
|
* 'spec' is the fstab spec field, which might be a device name or a
|
|
* pseudodevice or 'UUID=...' or 'LABEL=...'.
|
|
*
|
|
* 'mp' is the mount point, which could also be 'swap' or 'none'.
|
|
*/
|
|
static void
|
|
add_fstab_entry (guestfs_h *g, struct inspect_fs *fs,
|
|
const char *mountable, const char *mountpoint)
|
|
{
|
|
/* Add this to the fstab entry in 'fs'.
|
|
* Note these are further filtered by guestfs_inspect_get_mountpoints
|
|
* and guestfs_inspect_get_filesystems.
|
|
*/
|
|
const size_t n = fs->nr_fstab + 1;
|
|
struct inspect_fstab_entry *p;
|
|
|
|
p = safe_realloc (g, fs->fstab, n * sizeof (struct inspect_fstab_entry));
|
|
|
|
fs->fstab = p;
|
|
fs->nr_fstab = n;
|
|
|
|
/* These are owned by the handle and freed by guestfs_int_free_inspect_info. */
|
|
fs->fstab[n-1].mountable = safe_strdup (g, mountable);
|
|
fs->fstab[n-1].mountpoint = safe_strdup (g, mountpoint);
|
|
|
|
debug (g, "fstab: mountable=%s mountpoint=%s", mountable, mountpoint);
|
|
}
|
|
|
|
/* Compute a uuid hash as a simple xor of of its 4 32bit components */
|
|
static size_t
|
|
uuid_hash(const void *x, size_t table_size)
|
|
{
|
|
const md_uuid *a = x;
|
|
size_t h, i;
|
|
|
|
h = a->uuid[0];
|
|
for (i = 1; i < 4; i++) {
|
|
h ^= a->uuid[i];
|
|
}
|
|
|
|
return h % table_size;
|
|
}
|
|
|
|
static bool
|
|
uuid_cmp(const void *x, const void *y)
|
|
{
|
|
const md_uuid *a = x;
|
|
const md_uuid *b = y;
|
|
size_t i;
|
|
|
|
for (i = 0; i < 1; i++) {
|
|
if (a->uuid[i] != b->uuid[i]) return 0;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static void
|
|
md_uuid_free(void *x)
|
|
{
|
|
md_uuid *a = x;
|
|
free(a->path);
|
|
free(a);
|
|
}
|
|
|
|
/* Taken from parse_uuid in mdadm */
|
|
static int
|
|
parse_uuid (const char *str, uint32_t *uuid)
|
|
{
|
|
size_t hit = 0; /* number of Hex digIT */
|
|
char c;
|
|
size_t i;
|
|
int n;
|
|
|
|
for (i = 0; i < 4; i++)
|
|
uuid[i] = 0;
|
|
|
|
while ((c = *str++)) {
|
|
if (c >= '0' && c <= '9')
|
|
n = c - '0';
|
|
else if (c >= 'a' && c <= 'f')
|
|
n = 10 + c - 'a';
|
|
else if (c >= 'A' && c <= 'F')
|
|
n = 10 + c - 'A';
|
|
else if (strchr (":. -", c))
|
|
continue;
|
|
else
|
|
return -1;
|
|
|
|
if (hit < 32) {
|
|
uuid[hit / 8] <<= 4;
|
|
uuid[hit / 8] += n;
|
|
}
|
|
hit++;
|
|
}
|
|
if (hit == 32) return 0;
|
|
|
|
return -1;
|
|
}
|
|
|
|
/* Create a mapping of uuids to appliance md device names */
|
|
static ssize_t
|
|
map_app_md_devices (guestfs_h *g, Hash_table **map)
|
|
{
|
|
CLEANUP_FREE_STRING_LIST char **mds = NULL;
|
|
size_t n = 0;
|
|
char **md;
|
|
|
|
/* A hash mapping uuids to md device names */
|
|
*map = hash_initialize(16, NULL, uuid_hash, uuid_cmp, md_uuid_free);
|
|
if (*map == NULL) g->abort_cb();
|
|
|
|
mds = guestfs_list_md_devices(g);
|
|
if (mds == NULL) goto error;
|
|
|
|
for (md = mds; *md != NULL; md++) {
|
|
char **i;
|
|
CLEANUP_FREE_STRING_LIST char **detail = guestfs_md_detail (g, *md);
|
|
if (detail == NULL) goto error;
|
|
|
|
/* Iterate over keys until we find uuid */
|
|
for (i = detail; *i != NULL; i += 2) {
|
|
if (STREQ(*i, "uuid")) break;
|
|
}
|
|
|
|
/* We found it */
|
|
if (*i) {
|
|
md_uuid *entry;
|
|
|
|
/* Next item is the uuid value */
|
|
i++;
|
|
|
|
entry = safe_malloc(g, sizeof(md_uuid));
|
|
entry->path = safe_strdup(g, *md);
|
|
|
|
if (parse_uuid(*i, entry->uuid) == -1) {
|
|
/* Invalid UUID is weird, but not fatal. */
|
|
debug(g, "inspect-os: guestfs_md_detail returned invalid "
|
|
"uuid for %s: %s", *md, *i);
|
|
md_uuid_free(entry);
|
|
continue;
|
|
}
|
|
|
|
const void *matched = NULL;
|
|
switch (hash_insert_if_absent(*map, entry, &matched)) {
|
|
case -1:
|
|
g->abort_cb();
|
|
|
|
case 0:
|
|
/* Duplicate uuid in for md device is weird, but not fatal. */
|
|
debug(g, "inspect-os: md devices %s and %s have the same uuid",
|
|
((md_uuid *)matched)->path, entry->path);
|
|
md_uuid_free(entry);
|
|
break;
|
|
|
|
default:
|
|
n++;
|
|
}
|
|
}
|
|
}
|
|
|
|
return n;
|
|
|
|
error:
|
|
hash_free (*map); *map = NULL;
|
|
|
|
return -1;
|
|
}
|
|
|
|
static size_t
|
|
mdadm_app_hash(const void *x, size_t table_size)
|
|
{
|
|
const mdadm_app *a = x;
|
|
return hash_pjw(a->mdadm, table_size);
|
|
}
|
|
|
|
static bool
|
|
mdadm_app_cmp(const void *x, const void *y)
|
|
{
|
|
const mdadm_app *a = x;
|
|
const mdadm_app *b = y;
|
|
|
|
return STREQ (a->mdadm, b->mdadm);
|
|
}
|
|
|
|
static void
|
|
mdadm_app_free(void *x)
|
|
{
|
|
mdadm_app *a = x;
|
|
free(a->mdadm);
|
|
free(a->app);
|
|
free(a);
|
|
}
|
|
|
|
/* Get a map of md device names in mdadm.conf to their device names in the
|
|
* appliance */
|
|
static int
|
|
map_md_devices(guestfs_h *g, Hash_table **map)
|
|
{
|
|
CLEANUP_HASH_FREE Hash_table *app_map = NULL;
|
|
CLEANUP_FREE_STRING_LIST char **matches = NULL;
|
|
ssize_t n_app_md_devices;
|
|
|
|
*map = NULL;
|
|
|
|
/* Get a map of md device uuids to their device names in the appliance */
|
|
n_app_md_devices = map_app_md_devices (g, &app_map);
|
|
if (n_app_md_devices == -1) goto error;
|
|
|
|
/* Nothing to do if there are no md devices */
|
|
if (n_app_md_devices == 0)
|
|
return 0;
|
|
|
|
/* Get all arrays listed in mdadm.conf */
|
|
matches = guestfs_aug_match(g, "/files/etc/mdadm.conf/array");
|
|
if (!matches) goto error;
|
|
|
|
/* Log a debug message if we've got md devices, but nothing in mdadm.conf */
|
|
if (matches[0] == NULL) {
|
|
debug(g, "Appliance has MD devices, but augeas returned no array matches "
|
|
"in mdadm.conf");
|
|
return 0;
|
|
}
|
|
|
|
*map = hash_initialize(16, NULL, mdadm_app_hash, mdadm_app_cmp,
|
|
mdadm_app_free);
|
|
if (!*map) g->abort_cb();
|
|
|
|
for (char **m = matches; *m != NULL; m++) {
|
|
/* Get device name and uuid for each array */
|
|
CLEANUP_FREE char *dev_path = safe_asprintf (g, "%s/devicename", *m);
|
|
char *dev = guestfs_aug_get (g, dev_path);
|
|
if (!dev) goto error;
|
|
|
|
CLEANUP_FREE char *uuid_path = safe_asprintf (g, "%s/uuid", *m);
|
|
CLEANUP_FREE char *uuid = guestfs_aug_get (g, uuid_path);
|
|
if (!uuid) {
|
|
free (dev);
|
|
continue;
|
|
}
|
|
|
|
/* Parse the uuid into an md_uuid structure so we can look it up in the
|
|
* uuid->appliance device map */
|
|
md_uuid mdadm;
|
|
mdadm.path = dev;
|
|
if (parse_uuid(uuid, mdadm.uuid) == -1) {
|
|
/* Invalid uuid. Weird, but not fatal. */
|
|
debug(g, "inspect-os: mdadm.conf contains invalid uuid for %s: %s",
|
|
dev, uuid);
|
|
free (dev);
|
|
continue;
|
|
}
|
|
|
|
/* If there's a corresponding uuid in the appliance, create a new
|
|
* entry in the transitive map */
|
|
md_uuid *app = hash_lookup(app_map, &mdadm);
|
|
if (app) {
|
|
mdadm_app *entry = safe_malloc(g, sizeof(mdadm_app));
|
|
entry->mdadm = dev;
|
|
entry->app = safe_strdup(g, app->path);
|
|
|
|
switch (hash_insert_if_absent(*map, entry, NULL)) {
|
|
case -1:
|
|
g->abort_cb();
|
|
|
|
case 0:
|
|
/* Duplicate uuid in for md device is weird, but not fatal. */
|
|
debug(g, "inspect-os: mdadm.conf contains multiple entries for %s",
|
|
app->path);
|
|
mdadm_app_free(entry);
|
|
continue;
|
|
}
|
|
} else
|
|
free (dev);
|
|
}
|
|
|
|
return 0;
|
|
|
|
error:
|
|
if (*map) hash_free (*map);
|
|
|
|
return -1;
|
|
}
|
|
|
|
static int
|
|
resolve_fstab_device_xdev (guestfs_h *g, const char *type, const char *disk,
|
|
const char *part, char **device_ret)
|
|
{
|
|
CLEANUP_FREE char *name = NULL;
|
|
char *device;
|
|
CLEANUP_FREE_STRING_LIST char **devices = NULL;
|
|
size_t i, count;
|
|
struct drive *drv;
|
|
const char *p;
|
|
|
|
/* type: (h|s|v|xv)
|
|
* disk: ([a-z]+)
|
|
* part: (\d*)
|
|
*/
|
|
|
|
devices = guestfs_list_devices (g);
|
|
if (devices == NULL)
|
|
return -1;
|
|
|
|
/* Check any hints we were passed for a non-heuristic mapping */
|
|
name = safe_asprintf (g, "%sd%s", type, disk);
|
|
ITER_DRIVES (g, i, drv) {
|
|
if (drv->name && STREQ (drv->name, name)) {
|
|
device = safe_asprintf (g, "%s%s", devices[i], part);
|
|
if (!guestfs_int_is_partition (g, device)) {
|
|
free (device);
|
|
return 0;
|
|
}
|
|
*device_ret = device;
|
|
break;
|
|
}
|
|
}
|
|
|
|
/* Guess the appliance device name if we didn't find a matching hint */
|
|
if (!*device_ret) {
|
|
/* Count how many disks the libguestfs appliance has */
|
|
for (count = 0; devices[count] != NULL; count++)
|
|
;
|
|
|
|
/* Calculate the numerical index of the disk */
|
|
i = disk[0] - 'a';
|
|
for (p = disk + 1; *p != '\0'; p++) {
|
|
i += 1; i *= 26;
|
|
i += *p - 'a';
|
|
}
|
|
|
|
/* Check the index makes sense wrt the number of disks the appliance has.
|
|
* If it does, map it to an appliance disk.
|
|
*/
|
|
if (i < count) {
|
|
device = safe_asprintf (g, "%s%s", devices[i], part);
|
|
if (!guestfs_int_is_partition (g, device)) {
|
|
free (device);
|
|
return 0;
|
|
}
|
|
*device_ret = device;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
resolve_fstab_device_cciss (guestfs_h *g, const char *disk, const char *part,
|
|
char **device_ret)
|
|
{
|
|
char *device;
|
|
CLEANUP_FREE_STRING_LIST char **devices = NULL;
|
|
size_t i;
|
|
struct drive *drv;
|
|
|
|
/* disk: (cciss/c\d+d\d+)
|
|
* part: (\d+)?
|
|
*/
|
|
|
|
devices = guestfs_list_devices (g);
|
|
if (devices == NULL)
|
|
return -1;
|
|
|
|
/* Check any hints we were passed for a non-heuristic mapping */
|
|
ITER_DRIVES (g, i, drv) {
|
|
if (drv->name && STREQ (drv->name, disk)) {
|
|
if (part) {
|
|
device = safe_asprintf (g, "%s%s", devices[i], part);
|
|
if (!guestfs_int_is_partition (g, device)) {
|
|
free (device);
|
|
return 0;
|
|
}
|
|
*device_ret = device;
|
|
}
|
|
else
|
|
*device_ret = safe_strdup (g, devices[i]);
|
|
break;
|
|
}
|
|
}
|
|
|
|
/* We don't try to guess mappings for cciss devices */
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
resolve_fstab_device_diskbyid (guestfs_h *g, const char *part,
|
|
char **device_ret)
|
|
{
|
|
int nr_devices;
|
|
char *device;
|
|
|
|
/* For /dev/disk/by-id there is a limit to what we can do because
|
|
* original SCSI ID information has likely been lost. This
|
|
* heuristic will only work for guests that have a single block
|
|
* device.
|
|
*
|
|
* So the main task here is to make sure the assumptions above are
|
|
* true.
|
|
*
|
|
* XXX Use hints from virt-p2v if available.
|
|
* See also: https://bugzilla.redhat.com/show_bug.cgi?id=836573#c3
|
|
*/
|
|
|
|
nr_devices = guestfs_nr_devices (g);
|
|
if (nr_devices == -1)
|
|
return -1;
|
|
|
|
/* If #devices isn't 1, give up trying to translate this fstab entry. */
|
|
if (nr_devices != 1)
|
|
return 0;
|
|
|
|
/* Make the partition name and check it exists. */
|
|
device = safe_asprintf (g, "/dev/sda%s", part);
|
|
if (!guestfs_int_is_partition (g, device)) {
|
|
free (device);
|
|
return 0;
|
|
}
|
|
|
|
*device_ret = device;
|
|
return 0;
|
|
}
|
|
|
|
/* Resolve block device name to the libguestfs device name, eg.
|
|
* /dev/xvdb1 => /dev/vdb1; and /dev/mapper/VG-LV => /dev/VG/LV. This
|
|
* assumes that disks were added in the same order as they appear to
|
|
* the real VM, which is a reasonable assumption to make. Return
|
|
* anything we don't recognize unchanged.
|
|
*/
|
|
static char *
|
|
resolve_fstab_device (guestfs_h *g, const char *spec, Hash_table *md_map,
|
|
enum inspect_os_type os_type)
|
|
{
|
|
char *device = NULL;
|
|
char *type, *slice, *disk, *part;
|
|
int r;
|
|
|
|
if (STRPREFIX (spec, "/dev/mapper/")) {
|
|
/* LVM2 does some strange munging on /dev/mapper paths for VGs and
|
|
* LVs which contain '-' character:
|
|
*
|
|
* ><fs> lvcreate LV--test VG--test 32
|
|
* ><fs> debug ls /dev/mapper
|
|
* VG----test-LV----test
|
|
*
|
|
* This makes it impossible to reverse those paths directly, so
|
|
* we have implemented lvm_canonical_lv_name in the daemon.
|
|
*/
|
|
guestfs_push_error_handler (g, NULL, NULL);
|
|
device = guestfs_lvm_canonical_lv_name (g, spec);
|
|
guestfs_pop_error_handler (g);
|
|
if (device == NULL) {
|
|
if (guestfs_last_errno (g) == ENOENT) {
|
|
/* Ignore devices that don't exist. (RHBZ#811872) */
|
|
} else {
|
|
guestfs_int_error_errno (g, guestfs_last_errno (g), "%s", guestfs_last_error (g));
|
|
return NULL;
|
|
}
|
|
}
|
|
}
|
|
else if (match3 (g, spec, re_xdev, &type, &disk, &part)) {
|
|
r = resolve_fstab_device_xdev (g, type, disk, part, &device);
|
|
free (type);
|
|
free (disk);
|
|
free (part);
|
|
if (r == -1)
|
|
return NULL;
|
|
}
|
|
else if (match2 (g, spec, re_cciss, &disk, &part)) {
|
|
r = resolve_fstab_device_cciss (g, disk, part, &device);
|
|
free (disk);
|
|
free (part);
|
|
if (r == -1)
|
|
return NULL;
|
|
}
|
|
else if (md_map && (disk = match1 (g, spec, re_mdN)) != NULL) {
|
|
mdadm_app entry;
|
|
entry.mdadm = disk;
|
|
|
|
mdadm_app *app = hash_lookup (md_map, &entry);
|
|
if (app) device = safe_strdup (g, app->app);
|
|
|
|
free (disk);
|
|
}
|
|
else if (match3 (g, spec, re_freebsd_gpt, &type, &disk, &part)) {
|
|
/* If the FreeBSD disk contains GPT partitions, the translation to Linux
|
|
* device names is straight forward. Partitions on a virtio disk are
|
|
* prefixed with vtbd. IDE hard drives used to be prefixed with ad and now
|
|
* are with ada.
|
|
*/
|
|
const int disk_i = guestfs_int_parse_unsigned_int (g, disk);
|
|
const int part_i = guestfs_int_parse_unsigned_int (g, part);
|
|
free (type);
|
|
free (disk);
|
|
free (part);
|
|
|
|
if (disk_i != -1 && disk_i <= 26 && part_i > 0 && part_i <= 128)
|
|
device = safe_asprintf (g, "/dev/sd%c%d", disk_i + 'a', part_i);
|
|
}
|
|
else if (match4 (g, spec, re_freebsd_mbr, &type, &disk, &slice, &part)) {
|
|
/* FreeBSD disks are organized quite differently. See:
|
|
* http://www.freebsd.org/doc/handbook/disk-organization.html
|
|
* FreeBSD "partitions" are exposed as quasi-extended partitions
|
|
* numbered from 5 in Linux. I have no idea what happens when you
|
|
* have multiple "slices" (the FreeBSD term for MBR partitions).
|
|
*/
|
|
const int disk_i = guestfs_int_parse_unsigned_int (g, disk);
|
|
const int slice_i = guestfs_int_parse_unsigned_int (g, slice);
|
|
int part_i = part[0] - 'a' /* counting from 0 */;
|
|
free (type);
|
|
free (disk);
|
|
free (slice);
|
|
free (part);
|
|
|
|
if (part_i > 2)
|
|
/* Partition 'c' has the size of the enclosing slice. Not mapped under Linux. */
|
|
part_i -= 1;
|
|
|
|
if (disk_i != -1 && disk_i <= 26 &&
|
|
slice_i > 0 && slice_i <= 1 /* > 4 .. see comment above */ &&
|
|
part_i >= 0 && part_i < 25) {
|
|
device = safe_asprintf (g, "/dev/sd%c%d", disk_i + 'a', part_i + 5);
|
|
}
|
|
}
|
|
else if ((os_type == OS_TYPE_NETBSD) &&
|
|
match3 (g, spec, re_netbsd_dev, &type, &disk, &part)) {
|
|
const int disk_i = guestfs_int_parse_unsigned_int (g, disk);
|
|
int part_i = part[0] - 'a'; /* counting from 0 */
|
|
free (type);
|
|
free (disk);
|
|
free (part);
|
|
|
|
if (part_i > 3)
|
|
/* Partition 'c' is the disklabel partition and 'd' the hard disk itself.
|
|
* Not mapped under Linux.
|
|
*/
|
|
part_i -= 2;
|
|
|
|
if (disk_i != -1 && part_i >= 0 && part_i < 24)
|
|
device = safe_asprintf (g, "/dev/sd%c%d", disk_i + 'a', part_i + 5);
|
|
}
|
|
else if ((os_type == OS_TYPE_OPENBSD) &&
|
|
match3 (g, spec, re_openbsd_dev, &type, &disk, &part)) {
|
|
const int disk_i = guestfs_int_parse_unsigned_int (g, disk);
|
|
int part_i = part[0] - 'a'; /* counting from 0 */
|
|
free (type);
|
|
free (disk);
|
|
free (part);
|
|
|
|
if (part_i > 2)
|
|
/* Partition 'c' is the hard disk itself. Not mapped under Linux */
|
|
part_i -= 1;
|
|
|
|
/* In OpenBSD MAXPARTITIONS is defined to 16 for all architectures */
|
|
if (disk_i != -1 && part_i >= 0 && part_i < 15)
|
|
device = safe_asprintf (g, "/dev/sd%c%d", disk_i + 'a', part_i + 5);
|
|
}
|
|
else if ((part = match1 (g, spec, re_diskbyid)) != NULL) {
|
|
r = resolve_fstab_device_diskbyid (g, part, &device);
|
|
free (part);
|
|
if (r == -1)
|
|
return NULL;
|
|
}
|
|
else if (match3 (g, spec, re_hurd_dev, &type, &disk, &part)) {
|
|
/* Hurd disk devices are like /dev/hdNsM, where hdN is the
|
|
* N-th disk and M is the M-th partition on that disk.
|
|
* Turn the disk number into a letter-based identifier, so
|
|
* we can resolve it easily.
|
|
*/
|
|
const int disk_i = guestfs_int_parse_unsigned_int (g, disk);
|
|
const char disk_as_letter[2] = { disk_i + 'a', 0 };
|
|
r = resolve_fstab_device_xdev (g, type, disk_as_letter, part, &device);
|
|
free (type);
|
|
free (disk);
|
|
free (part);
|
|
if (r == -1)
|
|
return NULL;
|
|
}
|
|
|
|
/* Didn't match device pattern, return original spec unchanged. */
|
|
if (device == NULL)
|
|
device = safe_strdup (g, spec);
|
|
|
|
return device;
|
|
}
|
|
|
|
static char *make_augeas_path_expression (guestfs_h *g, const char **configfiles);
|
|
|
|
/* Call 'f' with Augeas opened and having parsed 'configfiles' (these
|
|
* files must exist). As a security measure, this bails if any file
|
|
* is too large for a reasonable configuration file. After the call
|
|
* to 'f' the Augeas handle is closed.
|
|
*/
|
|
static int
|
|
inspect_with_augeas (guestfs_h *g, struct inspect_fs *fs,
|
|
const char **configfiles,
|
|
int (*f) (guestfs_h *, struct inspect_fs *))
|
|
{
|
|
size_t i;
|
|
int64_t size;
|
|
int r;
|
|
CLEANUP_FREE char *pathexpr = NULL;
|
|
CLEANUP_FREE_STRING_LIST char **matches = NULL;
|
|
char **match;
|
|
|
|
/* Security: Refuse to do this if a config file is too large. */
|
|
for (i = 0; configfiles[i] != NULL; ++i) {
|
|
if (guestfs_is_file_opts (g, configfiles[i],
|
|
GUESTFS_IS_FILE_OPTS_FOLLOWSYMLINKS, 1, -1) == 0)
|
|
continue;
|
|
|
|
size = guestfs_filesize (g, configfiles[i]);
|
|
if (size == -1)
|
|
/* guestfs_filesize failed and has already set error in handle */
|
|
return -1;
|
|
if (size > MAX_AUGEAS_FILE_SIZE) {
|
|
error (g, _("size of %s is unreasonably large (%" PRIi64 " bytes)"),
|
|
configfiles[i], size);
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
if (guestfs_aug_init (g, "/", 16|32 /* AUG_SAVE_NOOP|AUG_NO_LOAD */) == -1)
|
|
return -1;
|
|
|
|
r = -1;
|
|
|
|
/* Tell Augeas to only load configfiles and no other files. This
|
|
* prevents a rogue guest from performing a denial of service attack
|
|
* by having large, over-complicated configuration files which are
|
|
* unrelated to the task at hand. (Thanks Dominic Cleal).
|
|
* Note this requires Augeas >= 1.0.0 because of RHBZ#975412.
|
|
*/
|
|
pathexpr = make_augeas_path_expression (g, configfiles);
|
|
if (guestfs_aug_rm (g, pathexpr) == -1)
|
|
goto out;
|
|
|
|
if (guestfs_aug_load (g) == -1)
|
|
goto out;
|
|
|
|
/* Check that augeas did not get a parse error for any of the configfiles,
|
|
* otherwise we are silently missing information.
|
|
*/
|
|
matches = guestfs_aug_match (g, "/augeas/files//error");
|
|
for (match = matches; *match != NULL; ++match) {
|
|
for (i = 0; configfiles[i] != NULL; ++i) {
|
|
CLEANUP_FREE char *errorpath =
|
|
safe_asprintf (g, "/augeas/files%s/error", configfiles[i]);
|
|
|
|
if (STREQ (*match, errorpath)) {
|
|
/* Get the various error details. */
|
|
guestfs_push_error_handler (g, NULL, NULL);
|
|
CLEANUP_FREE char *messagepath =
|
|
safe_asprintf (g, "%s/message", errorpath);
|
|
CLEANUP_FREE char *message = guestfs_aug_get (g, messagepath);
|
|
CLEANUP_FREE char *linepath =
|
|
safe_asprintf (g, "%s/line", errorpath);
|
|
CLEANUP_FREE char *line = guestfs_aug_get (g, linepath);
|
|
CLEANUP_FREE char *charpath =
|
|
safe_asprintf (g, "%s/char", errorpath);
|
|
CLEANUP_FREE char *charp = guestfs_aug_get (g, charpath);
|
|
guestfs_pop_error_handler (g);
|
|
|
|
error (g, _("%s:%s:%s: augeas parse failure: %s"),
|
|
configfiles[i],
|
|
line ? : "<none>",
|
|
charp ? : "<none>",
|
|
message ? : "<none>");
|
|
goto out;
|
|
}
|
|
}
|
|
}
|
|
|
|
r = f (g, fs);
|
|
|
|
out:
|
|
guestfs_aug_close (g);
|
|
|
|
return r;
|
|
}
|
|
|
|
/* Explained here: https://bugzilla.redhat.com/show_bug.cgi?id=975412#c0 */
|
|
static char *
|
|
make_augeas_path_expression (guestfs_h *g, const char **configfiles)
|
|
{
|
|
size_t i;
|
|
size_t nr_files;
|
|
CLEANUP_FREE_STRING_LIST char **subexprs = NULL;
|
|
CLEANUP_FREE char *subexpr = NULL;
|
|
char *ret;
|
|
|
|
nr_files = guestfs_int_count_strings ((char **) configfiles);
|
|
subexprs = safe_malloc (g, sizeof (char *) * (nr_files + 1));
|
|
|
|
for (i = 0; i < nr_files; ++i) {
|
|
subexprs[i] = /* v NB trailing '/' after filename */
|
|
safe_asprintf (g, "\"%s/\" !~ regexp('^') + glob(incl) + regexp('/.*')",
|
|
configfiles[i]);
|
|
}
|
|
subexprs[nr_files] = NULL;
|
|
|
|
subexpr = guestfs_int_join_strings (" and ", subexprs);
|
|
if (subexpr == NULL)
|
|
g->abort_cb ();
|
|
|
|
ret = safe_asprintf (g, "/augeas/load/*[ %s ]", subexpr);
|
|
debug (g, "augeas pathexpr = %s", ret);
|
|
return ret;
|
|
}
|
|
|
|
/* Canonicalize the path, so "///usr//local//" -> "/usr/local"
|
|
*
|
|
* The path is modified in place because the result is always
|
|
* the same length or shorter than the argument passed.
|
|
*/
|
|
static void
|
|
canonical_mountpoint (char *s)
|
|
{
|
|
size_t len = strlen (s);
|
|
char *orig = s;
|
|
|
|
s = strchr (s, '/');
|
|
while (s != NULL && *s != 0) {
|
|
char *pos = s + 1;
|
|
char *p = pos;
|
|
/* Find how many consecutive slashes are there after the one found,
|
|
* and shift the characters after them accordingly. */
|
|
while (*p == '/')
|
|
++p;
|
|
if (p - pos > 0) {
|
|
memmove (pos, p, len - (p - orig) + 1);
|
|
len -= p - pos;
|
|
}
|
|
|
|
s = strchr (pos, '/');
|
|
}
|
|
/* Ignore the trailing slash, but avoid removing it for "/". */
|
|
if (len > 1 && orig[len-1] == '/')
|
|
--len;
|
|
orig[len] = 0;
|
|
}
|