Compare commits

...

No commits in common. 'c9' and 'i8c' have entirely different histories.
c9 ... i8c

2
.gitignore vendored

@ -1 +1 @@
SOURCES/libosinfo-1.10.0.tar.xz
SOURCES/libosinfo-1.9.0.tar.xz

@ -1 +1 @@
b57c594b7f45a33575070df86862df78be2c4efb SOURCES/libosinfo-1.10.0.tar.xz
1f928e1412ac29e044e6b02254c9d2b2bf969c31 SOURCES/libosinfo-1.9.0.tar.xz

@ -0,0 +1,45 @@
From d3b1587f7b77b630bae8ab3f4760eda69bd7fe66 Mon Sep 17 00:00:00 2001
From: Victor Toso <victortoso@redhat.com>
Date: Fri, 26 Nov 2021 17:36:09 +0100
Subject: [PATCH] loader: add check for unknown file type
So we can provide a meaningful error message in case the provided path
is not accessible to running process.
e.g: running HOME=/root osinfo-query os we would get
`Error loading OS data: Unexpected file type`
and now we get
`Error loading OS data: Can't read path /root/.config/osinfo`
This error was first hit with v2v [0] that was leaking $USER and $HOME
of root user when osinfo-query as vsdm user with `sudo -c vdsm`. The
example above is a simple way to show lack of permision of
osinfo-query to read the root's $HOME.
[0] https://bugzilla.redhat.com/show_bug.cgi?id=1901423
Related: https://bugzilla.redhat.com/show_bug.cgi?id=1902720
Signed-off-by: Victor Toso <victortoso@redhat.com>
---
osinfo/osinfo_loader.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/osinfo/osinfo_loader.c b/osinfo/osinfo_loader.c
index 3f04fa5..96ca6ee 100644
--- a/osinfo/osinfo_loader.c
+++ b/osinfo/osinfo_loader.c
@@ -2374,6 +2374,11 @@ static void osinfo_loader_find_files(OsinfoLoader *loader,
}
g_object_unref(ents);
g_list_free(children);
+ } else if (type == G_FILE_TYPE_UNKNOWN) {
+ g_autofree gchar *path = g_file_get_path(file);
+ g_autofree gchar *msg = g_strdup_printf("Can't read path %s", path);
+ OSINFO_LOADER_SET_ERROR(&error, msg);
+ g_propagate_error(err, error);
} else {
OSINFO_LOADER_SET_ERROR(&error, "Unexpected file type");
g_propagate_error(err, error);
--
2.33.1

@ -0,0 +1,220 @@
From e5bdc6759195dbcfc4e7dcb02bf59190a3debe06 Mon Sep 17 00:00:00 2001
From: Laszlo Ersek <lersek@redhat.com>
Date: Wed, 9 Feb 2022 13:14:54 +0100
Subject: [PATCH] loader: work around lstat()/EACCES regression in
_g_local_file_info_get()
In glib commit 71e7b5800a31 ("Handle MLS selinux policy better",
2010-07-08), which was made for
<https://bugzilla.gnome.org/show_bug.cgi?id=623692>, an lstat() failure
with error code EACCES was *masked* in function _g_local_file_info_get().
Consequently, if osinfo_loader_find_files() calls g_file_query_info() on a
file that is inaccessible due to (e.g.) a missing "x" (search) permission
on a leading directory, then g_file_query_info() succeeds, our
"skipMissing" branch is dead, g_file_info_get_attribute_uint32() is
reached, and it returns G_FILE_TYPE_UNKNOWN.
As a consequence, the outer osinfo_loader_process_default_path() function
can fail, even though it passes skipMissing=TRUE to
osinfo_loader_process_list(). Example:
> $ HOME=/root \
> OSINFO_SYSTEM_DIR=/usr/share/osinfo \
> build/tools/osinfo-query os
> Error loading OS data: Can't read path /root/.config/osinfo
Arguably, this situation should be handled by simply skipping the
inaccessible path, as if all leading directories could be searched, and
only the last pathname compontent (the filename entry) didn't exist in its
direct parent directory.
The glib regression was reported in 2017:
https://bugzilla.gnome.org/show_bug.cgi?id=777187
and then migrated to gitlab:
https://gitlab.gnome.org/GNOME/glib/-/issues/1237
but it's still not solved today.
Work around the issue by honoring "skipMissing" on the G_FILE_TYPE_UNKNOWN
branch. Demonstration:
> $ HOME=/root \
> OSINFO_SYSTEM_DIR=/usr/share/osinfo \
> build/tools/osinfo-query os
>
> ** (osinfo-query:9924): WARNING **: 13:23:12.776: Can't read path /root/.config/osinfo
> Short ID | Name | Version | ID
> ----------------+------------------+---------+----------------------------------------
> alpinelinux3.5 | Alpine Linux 3.5 | 3.5 | http://alpinelinux.org/alpinelinux/3.5
> ...
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2051559
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
---
osinfo/osinfo_loader.c | 15 ++++++
tests/test-loader.c | 105 +++++++++++++++++++++++++++++++++++++++++
2 files changed, 120 insertions(+)
diff --git a/osinfo/osinfo_loader.c b/osinfo/osinfo_loader.c
index 96ca6ee..e244b3f 100644
--- a/osinfo/osinfo_loader.c
+++ b/osinfo/osinfo_loader.c
@@ -2377,6 +2377,21 @@ static void osinfo_loader_find_files(OsinfoLoader *loader,
} else if (type == G_FILE_TYPE_UNKNOWN) {
g_autofree gchar *path = g_file_get_path(file);
g_autofree gchar *msg = g_strdup_printf("Can't read path %s", path);
+ if (skipMissing) {
+ /* This is a work-around for
+ * <https://gitlab.gnome.org/GNOME/glib/-/issues/1237>. If the
+ * lstat() call underlying our g_file_query_info() call at the top
+ * of this function fails for "path" with EACCES, then
+ * g_file_query_info() should fail, and the "skipMissing" branch up
+ * there should suppress the error and return cleanly.
+ * Unfortunately, _g_local_file_info_get() masks the lstat()
+ * failure, g_file_info_get_attribute_uint32() is reached above,
+ * and returns G_FILE_TYPE_UNKNOWN for the file that could never be
+ * accessed. So we need to consider "skipMissing" here too.
+ */
+ g_warning("%s", msg);
+ return;
+ }
OSINFO_LOADER_SET_ERROR(&error, msg);
g_propagate_error(err, error);
} else {
diff --git a/tests/test-loader.c b/tests/test-loader.c
index 6644943..bb86585 100644
--- a/tests/test-loader.c
+++ b/tests/test-loader.c
@@ -16,6 +16,8 @@
*/
#include <osinfo/osinfo.h>
+#include <glib/gstdio.h>
+#include <unistd.h>
static void
test_basic(void)
@@ -31,6 +33,101 @@ test_basic(void)
g_object_unref(loader);
}
+typedef struct {
+ gchar *tmp_parent;
+ gchar *tmp_child;
+ gchar *orig_userdir;
+ gchar *expected_warning;
+} TestEaccesFixture;
+
+static void
+eacces_fixture_setup(TestEaccesFixture *fixture, gconstpointer user_data)
+{
+ gpointer rp;
+ gint ri;
+ gboolean rb;
+
+ /* create a temporary directory with permissions 0700 */
+ fixture->tmp_parent = g_strdup_printf("%s/%s", g_get_tmp_dir(),
+ "test_eacces.XXXXXX");
+ rp = g_mkdtemp_full(fixture->tmp_parent, 0700);
+ g_assert_nonnull(rp);
+
+ /* create a child directory called "osinfo" in it, with permissions 0700 */
+ fixture->tmp_child = g_strdup_printf("%s/osinfo", fixture->tmp_parent);
+ ri = g_mkdir(fixture->tmp_child, 0700);
+ g_assert_cmpint(ri, ==, 0);
+
+ /* revoke the search permission (0100) from the parent */
+ ri = g_chmod(fixture->tmp_parent, 0600);
+ g_assert_cmpint(ri, ==, 0);
+
+ /* stash the current value of OSINFO_USER_DIR */
+ fixture->orig_userdir = g_strdup(g_getenv("OSINFO_USER_DIR"));
+
+ /* point osinfo_loader_get_user_path() inside
+ * osinfo_loader_process_default_path() to the child directory
+ */
+ rb = g_setenv("OSINFO_USER_DIR", fixture->tmp_child, TRUE);
+ g_assert_true(rb);
+
+ /* format the pattern for the warning expected later on */
+ fixture->expected_warning = g_strdup_printf("Can't read path %s",
+ fixture->tmp_child);
+}
+
+static void
+eacces_fixture_teardown(TestEaccesFixture *fixture, gconstpointer user_data)
+{
+ gboolean rb;
+ gint ri;
+
+ /* free the expected warning pattern */
+ g_free(fixture->expected_warning);
+
+ /* restore the OSINFO_USER_DIR variable */
+ if (fixture->orig_userdir) {
+ rb = g_setenv("OSINFO_USER_DIR", fixture->orig_userdir, TRUE);
+ g_assert_true(rb);
+ g_free(fixture->orig_userdir);
+ } else {
+ g_unsetenv("OSINFO_USER_DIR");
+ }
+
+ /* restore search permission on the parent */
+ ri = g_chmod(fixture->tmp_parent, 0700);
+ g_assert_cmpint(ri, ==, 0);
+
+ /* remove both directories */
+ ri = g_rmdir(fixture->tmp_child);
+ g_assert_cmpint(ri, ==, 0);
+ g_free(fixture->tmp_child);
+
+ ri = g_rmdir(fixture->tmp_parent);
+ g_assert_cmpint(ri, ==, 0);
+ g_free(fixture->tmp_parent);
+}
+
+static void
+test_eacces(TestEaccesFixture *fixture, gconstpointer user_data)
+{
+ OsinfoLoader *loader = osinfo_loader_new();
+ GError *error = NULL;
+
+ g_assert_true(OSINFO_IS_LOADER(loader));
+
+ /* this should trigger an EACCES in glib's lstat(), but not break db
+ * loading; also we expect the warning here
+ */
+ g_test_expect_message(G_LOG_DOMAIN, G_LOG_LEVEL_WARNING,
+ fixture->expected_warning);
+ osinfo_loader_process_default_path(loader, &error);
+ g_assert_no_error(error);
+ g_test_assert_expected_messages();
+
+ g_object_unref(loader);
+}
+
int
main(int argc, char *argv[])
{
@@ -38,6 +135,14 @@ main(int argc, char *argv[])
g_test_add_func("/loader/basic", test_basic);
+ /* the following test depends on a directory with file mode bits 0600 being
+ * unsearchable for the owner, so skip it if the test is running as root
+ */
+ if (geteuid() != 0) {
+ g_test_add("/loader/eacces", TestEaccesFixture, NULL,
+ eacces_fixture_setup, test_eacces, eacces_fixture_teardown);
+ }
+
/* Upfront so we don't confuse valgrind */
osinfo_entity_get_type();
osinfo_db_get_type();
--
2.34.1

@ -2,13 +2,15 @@
Summary: A library for managing OS information for virtualization
Name: libosinfo
Version: 1.10.0
Release: 1%{?dist}
Version: 1.9.0
Release: 3%{?dist}
License: LGPLv2+
Source: https://releases.pagure.io/%{name}/%{name}-%{version}.tar.xz
URL: https://libosinfo.org/
### Patches ###
Patch0001: 0001-loader-add-check-for-unknown-file-type.patch
Patch0002: 0002-loader-work-around-lstat-EACCES-regression-in-_g_loc.patch
BuildRequires: meson
BuildRequires: gcc
@ -96,102 +98,46 @@ Libraries, includes, etc. to compile with the libosinfo library
%{_datadir}/vala/vapi/libosinfo-1.0.vapi
%changelog
* Wed Nov 30 2022 Victor Toso <victortoso@redhat.com> - 1.10.0-1
- Update to 1.10.0 release
* Wed Jul 26 2023 MSVSphere Packaging Team <packager@msvsphere.ru> - 1.9.0-3
- Rebuilt for MSVSphere 8.8
* Fri Feb 11 2022 Victor Toso <victortoso@redhat.com> - 1.9.0-5
- Warn on EACCES error when skipMissing is set
Resolves: rhbz#2051559
* Fri Feb 11 2022 Victor Toso <victortoso@redhat.com> - 1.9.0-3
- Resolves: rhbz#2053272
* Wed Dec 01 2021 Victor Toso <victortoso@redhat.com> - 1.9.0-4
- Improve error message on failure to access database due permission
Related: rhbz#1942431
* Thu Dec 16 2021 Victor Toso <victortoso@redhat.com> - 1.9.0-2
- Resolves: rhbz#1902720
* Mon Aug 09 2021 Mohan Boddu <mboddu@redhat.com> - 1.9.0-3
- Rebuilt for IMA sigs, glibc 2.34, aarch64 flags
Related: rhbz#1991688
* Thu Feb 04 2021 Danilo C. L. de Paula <ddepaula@redhat.com> - 1.9.0-1
- Resolves: rhbz#1903299 - Rebase to latest Fedora
* Fri Apr 16 2021 Mohan Boddu <mboddu@redhat.com> - 1.9.0-2
- Rebuilt for RHEL 9 BETA on Apr 15th 2021. Related: rhbz#1947937
* Tue Feb 02 2021 Fabiano Fidêncio <fidencio@redhat.com> - 1.9.0-1
- Update to 1.9.0 release
* Tue Jan 26 2021 Fedora Release Engineering <releng@fedoraproject.org> - 1.8.0-6
- Rebuilt for https://fedoraproject.org/wiki/Fedora_34_Mass_Rebuild
* Sat Aug 01 2020 Fedora Release Engineering <releng@fedoraproject.org> - 1.8.0-5
- Second attempt - Rebuilt for
https://fedoraproject.org/wiki/Fedora_33_Mass_Rebuild
* Tue Jul 28 2020 Fedora Release Engineering <releng@fedoraproject.org> - 1.8.0-4
- Rebuilt for https://fedoraproject.org/wiki/Fedora_33_Mass_Rebuild
* Sat May 30 2020 Fabiano Fidêncio <fidencio@redhat.com> - 1.8.0-1
- Update to 1.8.0 release
* Wed Jan 29 2020 Fedora Release Engineering <releng@fedoraproject.org> - 1.7.1-3
- Rebuilt for https://fedoraproject.org/wiki/Fedora_32_Mass_Rebuild
* Mon Jan 13 2020 Fabiano Fidêncio <fidencio@redhat.com> - 1.7.1-2
- Fix OsinfoList ABI breakage
* Wed Dec 04 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.7.1-1
- Update to 1.7.1 release
* Fri Nov 29 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.7.0-1
- Update to 1.7.0 release
* Fri Nov 08 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.6.0-2
- Improve ISO detection mechanism
* Fri Jul 26 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.6.0-1
- Update to 1.6.0 release
* Thu Jul 25 2019 Fedora Release Engineering <releng@fedoraproject.org> - 1.5.0-4
- Rebuilt for https://fedoraproject.org/wiki/Fedora_31_Mass_Rebuild
* Sun May 31 2020 Fabiano Fidêncio <fidencio@redhat.com> - 1.8.0-1
- Resolves: rhbz#1815158 - Rebase to libosinfo the latest upstream release
- Resolves: rhbz#1754394 - Provide information about UEFI support for guests (libosinfo)
- Resolves: rhbz#1032520 - WHQL'ed drivers should be made available for Boxes/libosinfo users
* Wed Jul 10 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.5.0-3
- rhbz#1727767 - CVE-2019-13313 libosinfo: osinfo-install-script
option leaks password via command line argument
- Resolves: rhbz#1727843 - CVE-2019-13313 libosinfo: osinfo-install-script
option leaks password via command line argument
* Mon Jun 03 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.5.0-2
- Fix coverity issues
* Wed May 22 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.5.0-2
- Resolves: rhbz#1712425 - New defects found in libosinfo-1.5.0-1.el8
* Thu May 09 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.5.0-1
* Fri May 10 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.5.0-1
- Update to 1.5.0 release
- Resolves: rhbz#1699988 - Rebase to the latest upstream release
* Thu Apr 11 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.4.0-3
- rhbz#1698845: Require GVFS
* Wed Apr 10 2019 Fabiano Fidêncio <fidencio@redhat.com> - 1.4.0-2
- Fix usage of application ID
- Fix images' load
- Remove tests depending on osinfo-db
* Fri Mar 01 2019 Fabiano Fidêncio <fidencio@redhat.com> 1.4.0-1
- Update to 1.4.0 release
* Mon Feb 04 2019 Kalev Lember <klember@redhat.com> - 1.3.0-3
- Use standard vala packaging pattern where vapi files are in -devel
* Fri Feb 01 2019 Fedora Release Engineering <releng@fedoraproject.org> - 1.3.0-2
- Rebuilt for https://fedoraproject.org/wiki/Fedora_30_Mass_Rebuild
* Wed Jan 30 2019 Daniel P. Berrangé <berrange@redhat.com> - 1.3.0-1
- Update to 1.3.0 release
* Thu Oct 11 2018 Fabiano Fidêncio <fabiano@fidencio.org> - 1.2.0-5
- Do not force anchored patterns on libosinfo, leave it for osinfo-db
* Fri Nov 30 2018 Fabiano Fidêncio <fidencio@redhat.com> - 1.2.0-5
- Related: rhbz#1650197 - Fix volume-ids for rhel8.0 entry
* Thu Sep 20 2018 Fabiano Fidêncio <fabiano@fidencio.org> - 1.2.0-4
- Require osinfo-db >= 20180920-1
* Wed Nov 14 2018 Fabiano Fidêncio <fidencio@redhat.com> - 1.2.0-4
- Resolves: rhbz#1649632 - libosinfo test suite should collect+report all
failures, not exit on first error
* Thu Sep 20 2018 Fabiano Fidêncio <fabiano@fidencio.org> - 1.2.0-3
- Force anchored patterns when matching regex
* Mon Oct 08 2018 Fabiano Fidêncio <fidencio@redhat.com> - 1.2.0-3
- Related: rhbz#1628027 - Revert ca945cdf04f
* Fri Jul 13 2018 Fedora Release Engineering <releng@fedoraproject.org> - 1.2.0-2
- Rebuilt for https://fedoraproject.org/wiki/Fedora_29_Mass_Rebuild
* Fri Sep 21 2018 Fabiano Fidêncio <fidencio@redhat.com> - 1.2.0-2
- Resolves: rhbz#1628027 - Force anchored patterns when matching regex
* Wed Jun 20 2018 Daniel P. Berrangé <berrange@redhat.com> - 1.2.0-1
- Update to 1.2.0 release

Loading…
Cancel
Save