On Sun, Feb 10, 2013 at 4:36 AM, Loic Dachary <loic@xxxxxxxxxxx> wrote: > Hi, > > On 02/06/2013 05:47 PM, Sage Weil wrote: >> On Wed, 6 Feb 2013, Loic Dachary wrote: >>> Hi, >>> >>> The patch below adds unit tests for FileStore::_detect_fs, but it needs to run >>> as root in order to mount the ext3, ext4 and btrfs file systems created for test purposes. >>> >>> Is there a way to mount filesystems without root privileges ? >> >> Not that I know of. >> >> There are several other tests that are meant ot be run separate from the >> build environment that use gtest (test_filejournal, the api functional >> tests, etc.). As long as this is built to a separate binary, I don't >> think it's a problem. We should converge on unittest_* for the stuff >> 'make check' runs and test_* for stuff that is meant to run elsewhere. >> Having our normal qa harness be root is not a problem. > > If I understand correctly, these test_* binaries can then be included in scripts found in ceph/qa/workunits and used by teuthology. The simplest example being https://github.com/ceph/ceph/blob/master/qa/workunits/rbd/test_librbd.sh Yep that's right. Note that the test_* binaries were recently renamed to ceph_test_* -sam > > Cheers > >>> configure.ac does not detect the presence of commands such as >>> mkfs.ext4 or mkfs.btrfs. The patch below does not add >>> AC_CHECK_TOOL([BTRFS], [mkfs.btrfs], ...) because it has no use >>> except for unit testing. Instead, the TEST_F(StoreTest, _detect_fs) >>> function checks if /sbin/mkfs.btrfs is an executable. If it is not, >>> "SKIP btrfs because /sbin/mkfs.btrfs is not executable" is displayed >>> and nothing is done. >>> >>> The drawback is that it is quite difficult to figure out what tools >>> must be installed to maximize test coverage. >>> >>> How would you implement unit test tools detection ? >>> >>> The error returned when ext4 is mounted without user_xattr is the same >>> as the error returned when ext4 is is mounted with user_xattr but >>> without the filestore_xattr_use_omap option : -ENOTSUP. From the point >>> of view of the unit tests, they cannot be distinguished, except by >>> parsing the messages sent to derr which show >>> >>> Extended attributes don't appear to work. >>> >>> or >>> >>> limited size xattrs -- enable filestore_xattr_use_omap >>> >>> Is parsing the output a good practice to assert success or failure during unit testing ? >> >> I'm not sure it is worth the complexity. IMO it would be better to change >> the return value (suggestions?) or not both distinguishing between the two >> cases in the test. >> >> sage >> >>> >>> Cheers >>> >>> diff --git a/src/test/filestore/store_test.cc b/src/test/filestore/store_test.cc >>> index c98ffb0..7ba38e2 100644 >>> --- a/src/test/filestore/store_test.cc >>> +++ b/src/test/filestore/store_test.cc >>> @@ -16,6 +16,7 @@ >>> #include <string.h> >>> #include <iostream> >>> #include <time.h> >>> +#include <sys/mount.h> >>> #include "os/FileStore.h" >>> #include "include/Context.h" >>> #include "common/ceph_argparse.h" >>> @@ -38,6 +39,7 @@ public: >>> >>> StoreTest() : store(0) {} >>> virtual void SetUp() { >>> + ::system("rm -fr store_test_temp_dir store_test_temp_journal"); >>> ::mkdir("store_test_temp_dir", 0777); >>> ObjectStore *store_ = new FileStore(string("store_test_temp_dir"), string("store_test_temp_journal")); >>> store.reset(store_); >>> @@ -823,6 +825,95 @@ TEST_F(StoreTest, ColSplitTest3) { >>> } >>> #endif >>> >>> +TEST_F(StoreTest, _detect_fs) { >>> + if (getuid() != 0) { >>> + cerr << "SKIP because it does not run as root" << std::endl; >>> + return; >>> + } >>> + >>> + if (access("/dev/zero", R_OK) != 0) { >>> + cerr << "SKIP because /dev/zero is not a readable file" << std::endl; >>> + return; >>> + } >>> + >>> + const string mnt("/tmp/test_filestore"); >>> + ::mkdir(mnt.c_str(), 0755); >>> + ::umount(mnt.c_str()); >>> + const string disk(mnt + ".disk"); >>> + ::unlink(disk.c_str()); >>> + ASSERT_EQ(::system(string("dd if=/dev/zero of=" + disk + " bs=1 count=0 seek=1G").c_str()), 0); >>> + >>> + const string dir("store_test_temp_dir"); >>> + const string journal("store_test_temp_journal"); >>> + >>> + const string mkfs_ext4("/sbin/mkfs.ext4"); >>> + if (::access(mkfs_ext4.c_str(), X_OK) == 0) { >>> + >>> + ASSERT_EQ(::system((mkfs_ext4 + " -q -F " + disk).c_str()), 0); >>> + // >>> + // without user_xattr, ext4 fails >>> + // >>> + { >>> + g_ceph_context->_conf->set_val("filestore_xattr_use_omap", "true"); >>> + ASSERT_EQ(::system((string("mount -o loop,nouser_xattr ") + disk + " " + mnt).c_str()), 0); >>> + ASSERT_EQ(::chdir(mnt.c_str()), 0); >>> + ASSERT_EQ(::mkdir(dir.c_str(), 0755), 0); >>> + FileStore store(dir, journal); >>> + ASSERT_EQ(store._detect_fs(), -ENOTSUP); >>> + ASSERT_EQ(::chdir("/tmp"), 0); >>> + ASSERT_EQ(::umount(mnt.c_str()), 0); >>> + } >>> + // >>> + // mounted with user_xattr, ext4 fails if filestore_xattr_use_omap is false >>> + // >>> + { >>> + g_ceph_context->_conf->set_val("filestore_xattr_use_omap", "false"); >>> + ASSERT_EQ(::system((string("mount -o loop,user_xattr ") + disk + " " + mnt).c_str()), 0); >>> + ASSERT_EQ(::chdir(mnt.c_str()), 0); >>> + FileStore store(dir, journal); >>> + ASSERT_EQ(store._detect_fs(), -ENOTSUP); >>> + ASSERT_EQ(::chdir("/tmp"), 0); >>> + ASSERT_EQ(::umount(mnt.c_str()), 0); >>> + } >>> + // >>> + // mounted with user_xattr, ext4 succeeds if filestore_xattr_use_omap is true >>> + // >>> + { >>> + g_ceph_context->_conf->set_val("filestore_xattr_use_omap", "true"); >>> + ASSERT_EQ(::system((string("mount -o loop,user_xattr ") + disk + " " + mnt).c_str()), 0); >>> + ASSERT_EQ(::chdir(mnt.c_str()), 0); >>> + FileStore store(dir, journal); >>> + ASSERT_EQ(store._detect_fs(), 0); >>> + ASSERT_EQ(::chdir("/tmp"), 0); >>> + ASSERT_EQ(::umount(mnt.c_str()), 0); >>> + } >>> + } else { >>> + cerr << "SKIP ext4 because " << mkfs_ext4 << " is not executable" << std::endl; >>> + } >>> + >>> + g_ceph_context->_conf->set_val("filestore_xattr_use_omap", "false"); >>> + ASSERT_EQ(::unlink(disk.c_str()), 0); >>> + ASSERT_EQ(::system(string("dd if=/dev/zero of=" + disk + " bs=1 count=0 seek=1G").c_str()), 0); >>> + const string mkfs_btrfs("/sbin/mkfs.btrfs"); >>> + if (::access(mkfs_btrfs.c_str(), X_OK) == 0) { >>> + ASSERT_EQ(::system((mkfs_btrfs + " " + disk).c_str()), 0); >>> + // >>> + // btrfs succeeds >>> + // >>> + { >>> + ASSERT_EQ(::system((string("mount -o loop ") + disk + " " + mnt).c_str()), 0); >>> + ASSERT_EQ(::chdir(mnt.c_str()), 0); >>> + ASSERT_EQ(::mkdir(dir.c_str(), 0755), 0); >>> + FileStore store(dir, journal); >>> + ASSERT_EQ(store._detect_fs(), 0); >>> + ASSERT_EQ(::chdir("/tmp"), 0); >>> + ASSERT_EQ(::umount(mnt.c_str()), 0); >>> + } >>> + } else { >>> + cerr << "SKIP btrfs because " << mkfs_btrfs << " is not executable" << std::endl; >>> + } >>> +} >>> + >>> >>> -- >>> Lo?c Dachary, Artisan Logiciel Libre >>> >>> > > -- > Loïc Dachary, Artisan Logiciel Libre > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
