e2fsck/e2fsck.shared
e2fsck/e2fsck.static
e2fsck/gen_crc32table
+e2fsck/lfsck
+e2fsck/lfsck.8
e2fsck/prof_err.c
e2fsck/prof_err.h
e2fsck/tst_crc32
DEPLIBQUOTA = @DEPSTATIC_LIBQUOTA@
DEPLIBBLKID = @DEPLIBBLKID@ @PRIVATE_LIBS_CMT@ $(DEPLIBUUID)
TESTENV = LD_LIBRARY_PATH="$(LIB):$${LD_LIBRARY_PATH}" DYLD_LIBRARY_PATH="$(LIB):$${DYLD_LIBRARY_PATH}"
+#Version of libdb lib found by configure
+LIBDB = @libdb@
STATIC_LIBSS = $(LIB)/libss@STATIC_LIB_EXT@ @DLOPEN_LIB@
STATIC_LIBCOM_ERR = $(LIB)/libcom_err@STATIC_LIB_EXT@ @SEM_INIT_LIB@
ET_DIR
SS_DIR
LDFLAG_STATIC
+WITH_LUSTRE
+LUSTRE
+ENABLE_LFSCK
+LFSCK_MAN
+LFSCK_CMT
root_sysconfdir
root_libdir
root_sbindir
LINUX_CMT
UNI_DIFF_OPTS
SEM_INIT_LIB
+DB4VERSION
SOCKET_LIB
SIZEOF_OFF_T
SIZEOF_LONG_LONG
with_libiconv_prefix
with_included_gettext
with_libintl_prefix
+enable_lfsck
+with_lustre
with_multiarch
'
ac_precious_vars='build_alias
--disable-uuidd disable building the uuid daemon
--disable-nls do not use Native Language Support
--disable-rpath do not hardcode runtime library paths
+ --disable-lfsck disable EXPERIMENTAL Lustre lfsck support
Optional Packages:
--with-PACKAGE[=ARG] use PACKAGE [ARG=yes]
--with-included-gettext use the GNU gettext library included here
--with-libintl-prefix[=DIR] search for libintl in DIR/include and DIR/lib
--without-libintl-prefix don't search for libintl in includedir and libdir
+ --with-lustre=LUSTRE path to Lustre sources for lfsck (default=/usr for RPM)
--with-multiarch=ARCH specify the multiarch triplet
Some influential environment variables:
fi
+if test "x$enable_lfsck" != "xno"; then :
+
+for ac_header in db.h
+do :
+ ac_fn_c_check_header_mongrel "$LINENO" "db.h" "ac_cv_header_db_h" "$ac_includes_default"
+if test "x$ac_cv_header_db_h" = xyes; then :
+ cat >>confdefs.h <<_ACEOF
+#define HAVE_DB_H 1
+_ACEOF
+
+fi
+
+done
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for db_env_create in -ldb-4.1" >&5
+$as_echo_n "checking for db_env_create in -ldb-4.1... " >&6; }
+if ${ac_cv_lib_db_4_1_db_env_create+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ ac_check_lib_save_LIBS=$LIBS
+LIBS="-ldb-4.1 $LIBS"
+cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+
+/* Override any GCC internal prototype to avoid an error.
+ Use char because int might match the return type of a GCC
+ builtin and then its argument prototype would still apply. */
+#ifdef __cplusplus
+extern "C"
+#endif
+char db_env_create ();
+int
+main ()
+{
+return db_env_create ();
+ ;
+ return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+ ac_cv_lib_db_4_1_db_env_create=yes
+else
+ ac_cv_lib_db_4_1_db_env_create=no
+fi
+rm -f core conftest.err conftest.$ac_objext \
+ conftest$ac_exeext conftest.$ac_ext
+LIBS=$ac_check_lib_save_LIBS
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_db_4_1_db_env_create" >&5
+$as_echo "$ac_cv_lib_db_4_1_db_env_create" >&6; }
+if test "x$ac_cv_lib_db_4_1_db_env_create" = xyes; then :
+ DB4VERSION=4.1
+
+$as_echo "#define HAVE_DB4 1" >>confdefs.h
+
+
+fi
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for db_env_create in -ldb-4.2" >&5
+$as_echo_n "checking for db_env_create in -ldb-4.2... " >&6; }
+if ${ac_cv_lib_db_4_2_db_env_create+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ ac_check_lib_save_LIBS=$LIBS
+LIBS="-ldb-4.2 $LIBS"
+cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+
+/* Override any GCC internal prototype to avoid an error.
+ Use char because int might match the return type of a GCC
+ builtin and then its argument prototype would still apply. */
+#ifdef __cplusplus
+extern "C"
+#endif
+char db_env_create ();
+int
+main ()
+{
+return db_env_create ();
+ ;
+ return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+ ac_cv_lib_db_4_2_db_env_create=yes
+else
+ ac_cv_lib_db_4_2_db_env_create=no
+fi
+rm -f core conftest.err conftest.$ac_objext \
+ conftest$ac_exeext conftest.$ac_ext
+LIBS=$ac_check_lib_save_LIBS
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_db_4_2_db_env_create" >&5
+$as_echo "$ac_cv_lib_db_4_2_db_env_create" >&6; }
+if test "x$ac_cv_lib_db_4_2_db_env_create" = xyes; then :
+ DB4VERSION=4.2
+
+$as_echo "#define HAVE_DB4 1" >>confdefs.h
+
+
+fi
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for db_env_create in -ldb-4.3" >&5
+$as_echo_n "checking for db_env_create in -ldb-4.3... " >&6; }
+if ${ac_cv_lib_db_4_3_db_env_create+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ ac_check_lib_save_LIBS=$LIBS
+LIBS="-ldb-4.3 $LIBS"
+cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+
+/* Override any GCC internal prototype to avoid an error.
+ Use char because int might match the return type of a GCC
+ builtin and then its argument prototype would still apply. */
+#ifdef __cplusplus
+extern "C"
+#endif
+char db_env_create ();
+int
+main ()
+{
+return db_env_create ();
+ ;
+ return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+ ac_cv_lib_db_4_3_db_env_create=yes
+else
+ ac_cv_lib_db_4_3_db_env_create=no
+fi
+rm -f core conftest.err conftest.$ac_objext \
+ conftest$ac_exeext conftest.$ac_ext
+LIBS=$ac_check_lib_save_LIBS
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_db_4_3_db_env_create" >&5
+$as_echo "$ac_cv_lib_db_4_3_db_env_create" >&6; }
+if test "x$ac_cv_lib_db_4_3_db_env_create" = xyes; then :
+ DB4VERSION=4.3
+
+$as_echo "#define HAVE_DB4 1" >>confdefs.h
+
+
+fi
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for db_env_create in -ldb-4.5" >&5
+$as_echo_n "checking for db_env_create in -ldb-4.5... " >&6; }
+if ${ac_cv_lib_db_4_5_db_env_create+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ ac_check_lib_save_LIBS=$LIBS
+LIBS="-ldb-4.5 $LIBS"
+cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+
+/* Override any GCC internal prototype to avoid an error.
+ Use char because int might match the return type of a GCC
+ builtin and then its argument prototype would still apply. */
+#ifdef __cplusplus
+extern "C"
+#endif
+char db_env_create ();
+int
+main ()
+{
+return db_env_create ();
+ ;
+ return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+ ac_cv_lib_db_4_5_db_env_create=yes
+else
+ ac_cv_lib_db_4_5_db_env_create=no
+fi
+rm -f core conftest.err conftest.$ac_objext \
+ conftest$ac_exeext conftest.$ac_ext
+LIBS=$ac_check_lib_save_LIBS
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_db_4_5_db_env_create" >&5
+$as_echo "$ac_cv_lib_db_4_5_db_env_create" >&6; }
+if test "x$ac_cv_lib_db_4_5_db_env_create" = xyes; then :
+ DB4VERSION=4.5
+
+$as_echo "#define HAVE_DB4 1" >>confdefs.h
+
+
+fi
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for db_env_create in -ldb-4.6" >&5
+$as_echo_n "checking for db_env_create in -ldb-4.6... " >&6; }
+if ${ac_cv_lib_db_4_6_db_env_create+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ ac_check_lib_save_LIBS=$LIBS
+LIBS="-ldb-4.6 $LIBS"
+cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+
+/* Override any GCC internal prototype to avoid an error.
+ Use char because int might match the return type of a GCC
+ builtin and then its argument prototype would still apply. */
+#ifdef __cplusplus
+extern "C"
+#endif
+char db_env_create ();
+int
+main ()
+{
+return db_env_create ();
+ ;
+ return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+ ac_cv_lib_db_4_6_db_env_create=yes
+else
+ ac_cv_lib_db_4_6_db_env_create=no
+fi
+rm -f core conftest.err conftest.$ac_objext \
+ conftest$ac_exeext conftest.$ac_ext
+LIBS=$ac_check_lib_save_LIBS
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_db_4_6_db_env_create" >&5
+$as_echo "$ac_cv_lib_db_4_6_db_env_create" >&6; }
+if test "x$ac_cv_lib_db_4_6_db_env_create" = xyes; then :
+ DB4VERSION=4.6
+
+$as_echo "#define HAVE_DB4 1" >>confdefs.h
+
+
+fi
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for db_env_create in -ldb-4.7" >&5
+$as_echo_n "checking for db_env_create in -ldb-4.7... " >&6; }
+if ${ac_cv_lib_db_4_7_db_env_create+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ ac_check_lib_save_LIBS=$LIBS
+LIBS="-ldb-4.7 $LIBS"
+cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+
+/* Override any GCC internal prototype to avoid an error.
+ Use char because int might match the return type of a GCC
+ builtin and then its argument prototype would still apply. */
+#ifdef __cplusplus
+extern "C"
+#endif
+char db_env_create ();
+int
+main ()
+{
+return db_env_create ();
+ ;
+ return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+ ac_cv_lib_db_4_7_db_env_create=yes
+else
+ ac_cv_lib_db_4_7_db_env_create=no
+fi
+rm -f core conftest.err conftest.$ac_objext \
+ conftest$ac_exeext conftest.$ac_ext
+LIBS=$ac_check_lib_save_LIBS
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_db_4_7_db_env_create" >&5
+$as_echo "$ac_cv_lib_db_4_7_db_env_create" >&6; }
+if test "x$ac_cv_lib_db_4_7_db_env_create" = xyes; then :
+ DB4VERSION=4.7
+
+$as_echo "#define HAVE_DB4 1" >>confdefs.h
+
+
+fi
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for db_env_create in -ldb-4.8" >&5
+$as_echo_n "checking for db_env_create in -ldb-4.8... " >&6; }
+if ${ac_cv_lib_db_4_8_db_env_create+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ ac_check_lib_save_LIBS=$LIBS
+LIBS="-ldb-4.8 $LIBS"
+cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h. */
+
+/* Override any GCC internal prototype to avoid an error.
+ Use char because int might match the return type of a GCC
+ builtin and then its argument prototype would still apply. */
+#ifdef __cplusplus
+extern "C"
+#endif
+char db_env_create ();
+int
+main ()
+{
+return db_env_create ();
+ ;
+ return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+ ac_cv_lib_db_4_8_db_env_create=yes
+else
+ ac_cv_lib_db_4_8_db_env_create=no
+fi
+rm -f core conftest.err conftest.$ac_objext \
+ conftest$ac_exeext conftest.$ac_ext
+LIBS=$ac_check_lib_save_LIBS
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_db_4_8_db_env_create" >&5
+$as_echo "$ac_cv_lib_db_4_8_db_env_create" >&6; }
+if test "x$ac_cv_lib_db_4_8_db_env_create" = xyes; then :
+ DB4VERSION=4.8
+
+$as_echo "#define HAVE_DB4 1" >>confdefs.h
+
+
+fi
+
+
+
+fi
{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for optreset" >&5
$as_echo_n "checking for optreset... " >&6; }
if test "${ac_cv_have_optreset+set}" = set; then :
+# Check whether --enable-lfsck was given.
+if test "${enable_lfsck+set}" = set; then :
+ enableval=$enable_lfsck; if test "$enableval" = "no"
+then
+ LFSCK_CMT="#"
+ LFSCK_MAN='.\"'
+ ENABLE_LFSCK="--disable-lfsck"
+ echo "Disabling Lustre lfsck support"
+else
+ LFSCK_CMT=
+ LFSCK_MAN=
+ ENABLE_LFSCK="--enable-lfsck"
+
+$as_echo "#define HAVE_LFSCK 1" >>confdefs.h
+
+ echo "Enabling Lustre lfsck support"
+fi
+
+else
+ LFSCK_CMT=
+LFSCK_MAN=
+ENABLE_LFSCK=
+
+$as_echo "#define HAVE_LFSCK 1" >>confdefs.h
+
+echo "Enabling Lustre lfsck support by default"
+
+fi
+
+
+
+
+
+# Check whether --with-lustre was given.
+if test "${with_lustre+set}" = set; then :
+ withval=$with_lustre; { $as_echo "$as_me:${as_lineno-$LINENO}: result: LUSTRE is in $withval" >&5
+$as_echo "LUSTRE is in $withval" >&6; }
+ LUSTRE="$withval"
+ WITH_LUSTRE="--with-lustre=$withval"
+else
+ LUSTRE="/usr"
+ WITH_LUSTRE=
+
+fi
+
+
+if test "x$enable_lfsck" != "xno"; then :
+
+ as_ac_File=`$as_echo "ac_cv_file_$LUSTRE/include/lustre/lustreapi.h" | $as_tr_sh`
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $LUSTRE/include/lustre/lustreapi.h" >&5
+$as_echo_n "checking for $LUSTRE/include/lustre/lustreapi.h... " >&6; }
+if eval \${$as_ac_File+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ test "$cross_compiling" = yes &&
+ as_fn_error $? "cannot check for file existence when cross compiling" "$LINENO" 5
+if test -r "$LUSTRE/include/lustre/lustreapi.h"; then
+ eval "$as_ac_File=yes"
+else
+ eval "$as_ac_File=no"
+fi
+fi
+eval ac_res=\$$as_ac_File
+ { $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_res" >&5
+$as_echo "$ac_res" >&6; }
+if eval test \"x\$"$as_ac_File"\" = x"yes"; then :
+
+$as_echo "#define HAVE_LUSTRE_LUSTREAPI_H 1" >>confdefs.h
+
+else
+ as_ac_File=`$as_echo "ac_cv_file_$LUSTRE/lustre/include/lustre/lustreapi.h" | $as_tr_sh`
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $LUSTRE/lustre/include/lustre/lustreapi.h" >&5
+$as_echo_n "checking for $LUSTRE/lustre/include/lustre/lustreapi.h... " >&6; }
+if eval \${$as_ac_File+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ test "$cross_compiling" = yes &&
+ as_fn_error $? "cannot check for file existence when cross compiling" "$LINENO" 5
+if test -r "$LUSTRE/lustre/include/lustre/lustreapi.h"; then
+ eval "$as_ac_File=yes"
+else
+ eval "$as_ac_File=no"
+fi
+fi
+eval ac_res=\$$as_ac_File
+ { $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_res" >&5
+$as_echo "$ac_res" >&6; }
+if eval test \"x\$"$as_ac_File"\" = x"yes"; then :
+
+$as_echo "#define HAVE_LUSTRE_LUSTREAPI_H 1" >>confdefs.h
+
+else
+ as_ac_File=`$as_echo "ac_cv_file_$LUSTRE/include/lustre/liblustreapi.h" | $as_tr_sh`
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $LUSTRE/include/lustre/liblustreapi.h" >&5
+$as_echo_n "checking for $LUSTRE/include/lustre/liblustreapi.h... " >&6; }
+if eval \${$as_ac_File+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ test "$cross_compiling" = yes &&
+ as_fn_error $? "cannot check for file existence when cross compiling" "$LINENO" 5
+if test -r "$LUSTRE/include/lustre/liblustreapi.h"; then
+ eval "$as_ac_File=yes"
+else
+ eval "$as_ac_File=no"
+fi
+fi
+eval ac_res=\$$as_ac_File
+ { $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_res" >&5
+$as_echo "$ac_res" >&6; }
+if eval test \"x\$"$as_ac_File"\" = x"yes"; then :
+
+$as_echo "#define HAVE_LUSTRE_LIBLUSTREAPI_H 1" >>confdefs.h
+
+else
+ as_ac_File=`$as_echo "ac_cv_file_$LUSTRE/lustre/include/lustre/liblustreapi.h" | $as_tr_sh`
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $LUSTRE/lustre/include/lustre/liblustreapi.h" >&5
+$as_echo_n "checking for $LUSTRE/lustre/include/lustre/liblustreapi.h... " >&6; }
+if eval \${$as_ac_File+:} false; then :
+ $as_echo_n "(cached) " >&6
+else
+ test "$cross_compiling" = yes &&
+ as_fn_error $? "cannot check for file existence when cross compiling" "$LINENO" 5
+if test -r "$LUSTRE/lustre/include/lustre/liblustreapi.h"; then
+ eval "$as_ac_File=yes"
+else
+ eval "$as_ac_File=no"
+fi
+fi
+eval ac_res=\$$as_ac_File
+ { $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_res" >&5
+$as_echo "$ac_res" >&6; }
+if eval test \"x\$"$as_ac_File"\" = x"yes"; then :
+
+$as_echo "#define HAVE_LUSTRE_LIBLUSTREAPI_H 1" >>confdefs.h
+
+else
+ as_fn_error $? "No lustre includes found." "$LINENO" 5
+fi
+
+fi
+
+fi
+
+
+fi
+
+
+fi
# Check whether --with-multiarch was given.
if test "${with_multiarch+set}" = set; then :
AC_CHECK_LIB(socket, socket, [SOCKET_LIB=-lsocket])
AC_SUBST(SOCKET_LIB)
dnl
+dnl Check to see if libdb exists for the sake of lfsck
+dnl
+AS_IF([test "x$enable_lfsck" != "xno"],
+[
+AC_CHECK_HEADERS(db.h)
+AC_CHECK_LIB(db-4.1, db_env_create,
+ [DB4VERSION=4.1
+ AC_DEFINE(HAVE_DB4, 1, [Define to 1 if DB4 library is present])
+])
+AC_CHECK_LIB(db-4.2, db_env_create,
+ [DB4VERSION=4.2
+ AC_DEFINE(HAVE_DB4, 1, [Define to 1 if DB4 library is present])
+])
+AC_CHECK_LIB(db-4.3, db_env_create,
+ [DB4VERSION=4.3
+ AC_DEFINE(HAVE_DB4, 1, [Define to 1 if DB4 library is present])
+])
+AC_CHECK_LIB(db-4.5, db_env_create,
+ [DB4VERSION=4.5
+ AC_DEFINE(HAVE_DB4, 1, [Define to 1 if DB4 library is present])
+])
+AC_CHECK_LIB(db-4.6, db_env_create,
+ [DB4VERSION=4.6
+ AC_DEFINE(HAVE_DB4, 1, [Define to 1 if DB4 library is present])
+])
+AC_CHECK_LIB(db-4.7, db_env_create,
+ [DB4VERSION=4.7
+ AC_DEFINE(HAVE_DB4, 1, [Define to 1 if DB4 library is present])
+])
+AC_CHECK_LIB(db-4.8, db_env_create,
+ [DB4VERSION=4.8
+ AC_DEFINE(HAVE_DB4, 1, [Define to 1 if DB4 library is present])
+])
+AC_SUBST(DB4VERSION)
+])
+dnl
dnl See if optreset exists
dnl
AC_MSG_CHECKING(for optreset)
AC_SUBST(root_libdir)
AC_SUBST(root_sysconfdir)
dnl
+dnl handle --enable-lfsck
+dnl
+AC_ARG_ENABLE([lfsck],
+[ --disable-lfsck disable EXPERIMENTAL Lustre lfsck support],
+if test "$enableval" = "no"
+then
+ LFSCK_CMT="#"
+ LFSCK_MAN='.\"'
+ ENABLE_LFSCK="--disable-lfsck"
+ echo "Disabling Lustre lfsck support"
+else
+ LFSCK_CMT=
+ LFSCK_MAN=
+ ENABLE_LFSCK="--enable-lfsck"
+ AC_DEFINE(HAVE_LFSCK, 1, [Define to 1 if Lustre lfsck is enabled])
+ echo "Enabling Lustre lfsck support"
+fi
+,
+LFSCK_CMT=
+LFSCK_MAN=
+ENABLE_LFSCK=
+AC_DEFINE(HAVE_LFSCK, 1, [Define to 1 if Lustre lfsck is enabled])
+echo "Enabling Lustre lfsck support by default"
+)
+AC_SUBST(LFSCK_CMT)
+AC_SUBST(LFSCK_MAN)
+AC_SUBST(ENABLE_LFSCK)
+dnl
+dnl set lustre include path and build lfsck
+dnl
+AC_ARG_WITH([lustre],
+[ --with-lustre=LUSTRE path to Lustre sources for lfsck (default=/usr for RPM)],
+AC_MSG_RESULT(LUSTRE is in $withval)
+ LUSTRE="$withval"
+ WITH_LUSTRE="--with-lustre=$withval",
+ LUSTRE="/usr"
+ WITH_LUSTRE=
+)dnl
+AC_SUBST(LUSTRE)
+AC_SUBST(WITH_LUSTRE)
+dnl
+AS_IF([test "x$enable_lfsck" != "xno"],
+[
+dnl # Lustre 2.4 and later have lustreapi.h instead of liblustreapi.h.
+dnl # While the latter includes the former for compatibility reasons, the
+dnl # new liblustreapi.h wrapper emits a deprecation compile warning on
+dnl # each use that I'd rather avoid, and it will eventually be removed.
+ AC_CHECK_FILE($LUSTRE/include/lustre/lustreapi.h,
+ AC_DEFINE(HAVE_LUSTRE_LUSTREAPI_H, 1,
+ [Define to 1 if lustre/lustreapi.h found]),
+ AC_CHECK_FILE($LUSTRE/lustre/include/lustre/lustreapi.h,
+ AC_DEFINE(HAVE_LUSTRE_LUSTREAPI_H, 1,
+ [Define to 1 if lustre/lustreapi.h found]),
+ AC_CHECK_FILE($LUSTRE/include/lustre/liblustreapi.h,
+ AC_DEFINE(HAVE_LUSTRE_LIBLUSTREAPI_H, 1,
+ [Define to 1 if lustre/liblustreapi.h found]),
+ AC_CHECK_FILE($LUSTRE/lustre/include/lustre/liblustreapi.h,
+ AC_DEFINE(HAVE_LUSTRE_LIBLUSTREAPI_H, 1,
+ [Define to 1 if lustre/liblustreapi.h found]),
+ AC_MSG_ERROR([No lustre includes found.]))))
+ )
+dnl # Disable for old autoconf that doesn't check header usability.
+dnl # Supposed to check if lustre_idl.h is buildable from userspace.
+dnl AC_CHECK_HEADER($LUSTRE/lustre/include/lustre/lustre_idl.h,
+dnl AC_DEFINE(HAVE_LUSTRE_LUSTRE_IDL_H, 1),
+dnl AC_CHECK_HEADER($LUSTRE/include/lustre/lustre_idl.h,
+dnl AC_DEFINE(HAVE_LUSTRE_LUSTRE_IDL_H, 1),)
+dnl )
+])
+dnl
dnl Allow specification of the multiarch arch
dnl
AC_ARG_WITH([multiarch],
.c.o:
$(E) " CC $<"
- $(Q) $(CC) -c $(ALL_CFLAGS) $< -o $@
+ $(Q) $(CC) -c $(ALL_CFLAGS) $(LUSTRE_INC) $< -o $@
$(Q) $(CHECK_CMD) $(ALL_CFLAGS) $<
all:: $(PROGS) $(MANPAGES)
@MCONFIG@
PROGS= e2fsck
+@LFSCK_CMT@USPROGS= lfsck
MANPAGES= e2fsck.8
FMANPAGES= e2fsck.conf.5
+@LFSCK_CMT@MANPAGES += lfsck.8
+XTRA_CFLAGS= -DRESOURCE_TRACK -I.
+
+@LFSCK_CMT@LFSCK_LIBS=-ldb-@DB4VERSION@
+@LFSCK_CMT@LUSTRE_INC=-I @LUSTRE@/lustre/include -I @LUSTRE@/include -I @LUSTRE@/libcfs/include -Wall
+@LFSCK_CMT@LUSTRE_LIBS=-L@LUSTRE@/lustre/utils -L@LUSTRE@/lib64 -L@LUSTRE@/lib \
+@LFSCK_CMT@ -Wl,-Bstatic -llustreapi -Wl,-Bdynamic
LIBS= $(LIBQUOTA) $(LIBEXT2FS) $(LIBCOM_ERR) $(LIBBLKID) $(LIBUUID) \
- $(LIBINTL) $(LIBE2P) $(SYSLIBS)
+ $(LIBINTL) $(LIBE2P) $(SYSLIBS) $(LFSCK_LIBS)
DEPLIBS= $(DEPLIBQUOTA) $(LIBEXT2FS) $(DEPLIBCOM_ERR) $(DEPLIBBLKID) \
$(DEPLIBUUID) $(DEPLIBE2P)
STATIC_LIBS= $(STATIC_LIBQUOTA) $(STATIC_LIBEXT2FS) $(STATIC_LIBCOM_ERR) \
$(STATIC_LIBBLKID) $(STATIC_LIBUUID) $(LIBINTL) $(STATIC_LIBE2P) \
- $(SYSLIBS)
+ $(SYSLIBS) $(LFSCK_LIBS)
STATIC_DEPLIBS= $(DEPSTATIC_LIBQUOTA) $(STATIC_LIBEXT2FS) \
$(DEPSTATIC_LIBCOM_ERR) $(DEPSTATIC_LIBBLKID) \
$(DEPSTATIC_LIBUUID) $(DEPSTATIC_LIBE2P)
PROFILED_LIBS= $(PROFILED_LIBQUOTA) $(PROFILED_LIBEXT2FS) \
$(PROFILED_LIBCOM_ERR) $(PROFILED_LIBBLKID) $(PROFILED_LIBUUID) \
- $(PROFILED_LIBE2P) $(LIBINTL) $(SYSLIBS)
+ $(PROFILED_LIBE2P) $(LIBINTL) $(SYSLIBS) $(LFSCK_LIBS)
PROFILED_DEPLIBS= $(DEPPROFILED_LIBQUOTA) $(PROFILED_LIBEXT2FS) \
$(DEPPROFILED_LIBCOM_ERR) $(DEPPROFILED_LIBBLKID) \
$(DEPPROFILED_LIBUUID) $(DEPPROFILED_LIBE2P)
.c.o:
$(E) " CC $<"
- $(Q) $(CC) -c $(ALL_CFLAGS) $< -o $@
+ $(Q) $(CC) -c $(ALL_CFLAGS) $(LUSTRE_INC) $< -o $@
$(Q) $(CHECK_CMD) $(ALL_CFLAGS) $<
-@PROFILE_CMT@ $(Q) $(CC) $(ALL_CFLAGS) -g -pg -o profiled/$*.o -c $<
+@PROFILE_CMT@ $(Q) $(CC) $(ALL_CFLAGS) $(LUSTRE_INC) -g -pg -o profiled/$*.o -c $<
#
# Flags for doing mtrace --- uncomment to produce mtracing e2fsck
#MCHECK= -DMCHECK
OBJS= crc32.o dict.o unix.o e2fsck.o super.o pass1.o pass1b.o pass2.o \
- pass3.o pass4.o pass5.o journal.o badblocks.o util.o dirinfo.o \
+ pass3.o pass4.o pass5.o pass6.o journal.o badblocks.o util.o dirinfo.o \
dx_dirinfo.o ehandler.o problem.o message.o quota.o recovery.o \
region.o revoke.o ea_refcount.o rehash.o profile.o prof_err.o \
logfile.o sigcatcher.o $(MTRACE_OBJ)
+@LFSCK_CMT@OBJS += lfsck_common.o
+
+@LFSCK_CMT@LFSCK_OBJS = lfsck_common.o lfsck.o
PROFILED_OBJS= profiled/dict.o profiled/unix.o profiled/e2fsck.o \
profiled/super.o profiled/pass1.o profiled/pass1b.o \
profiled/pass2.o profiled/pass3.o profiled/pass4.o profiled/pass5.o \
- profiled/journal.o profiled/badblocks.o profiled/util.o \
+ profiled/pass6.o profiled/journal.o profiled/badblocks.o profiled/util.o \
profiled/dirinfo.o profiled/dx_dirinfo.o profiled/ehandler.o \
profiled/message.o profiled/problem.o profiled/quota.o \
profiled/recovery.o profiled/region.o profiled/revoke.o \
profiled/ea_refcount.o profiled/rehash.o profiled/profile.o \
profiled/crc32.o profiled/prof_err.o profiled/logfile.o \
profiled/sigcatcher.o
+@LFSCK_CMT@PROFILED_OBJS += profiled/lfsck_common.o
SRCS= $(srcdir)/e2fsck.c \
$(srcdir)/crc32.c \
$(srcdir)/pass3.c \
$(srcdir)/pass4.c \
$(srcdir)/pass5.c \
+ $(srcdir)/pass6.c \
$(srcdir)/journal.c \
$(srcdir)/recovery.c \
$(srcdir)/revoke.c \
$(srcdir)/quota.c \
$(MTRACE_SRC)
-all:: profiled $(PROGS) e2fsck $(MANPAGES) $(FMANPAGES)
+@LFSCK_CMT@SRCS += $(srcdir)/lfsck_common.c
+@LFSCK_CMT@LFSCK_SRCS = $(srcdir)/lfsck_common.c $(srcdir)/lfsck.c
+
+all:: profiled $(PROGS) $(USPROGS) e2fsck $(MANPAGES) $(FMANPAGES)
@PROFILE_CMT@all:: e2fsck.profiled
$(TESTENV) ./tst_crc32
$(TESTENV) ./tst_problem
+@LFSCK_CMT@lfsck: $(LFSCK_OBJS) $(DEPLIBS)
+@LFSCK_CMT@ @echo " LD $@"
+@LFSCK_CMT@ $(LD) -lpthread $(ALL_LDFLAGS) $(LFSCK_OBJS) -o lfsck $(LIBS) $(LUSTRE_LIBS)
+
extend: extend.o
$(E) " LD $@"
$(Q) $(LD) $(ALL_LDFLAGS) -o extend extend.o $(CHECKLIB)
@PROFILE_CMT@ $(E) " MKDIR $@"
@PROFILE_CMT@ $(Q) mkdir profiled
+lfsck.8: $(DEP_SUBSTITUTE) $(srcdir)/lfsck.8.in
+ $(SUBSTITUTE_UPTIME) $(srcdir)/lfsck.8.in lfsck.8
+
e2fsck.8: $(DEP_SUBSTITUTE) $(srcdir)/e2fsck.8.in
$(E) " SUBST $@"
$(Q) $(SUBSTITUTE_UPTIME) $(srcdir)/e2fsck.8.in e2fsck.8
$(Q) $(SUBSTITUTE_UPTIME) $(srcdir)/e2fsck.conf.5.in e2fsck.conf.5
installdirs:
- $(E) " MKINSTALLDIRS $(root_sbindir) $(man8dir)"
+ $(E) " MKINSTALLDIRS $(root_sbindir) $(sbindir) $(man8dir)"
$(Q) $(MKINSTALLDIRS) $(DESTDIR)$(root_sbindir) \
- $(DESTDIR)$(man8dir) $(DESTDIR)$(man5dir)
+ $(DESTDIR)$(man8dir) $(DESTDIR)$(man5dir) $(DESTDIR)$(sbindir)
-install: $(PROGS) $(MANPAGES) $(FMANPAGES) installdirs
+install: $(PROGS) $(USPROGS) $(MANPAGES) $(FMANPAGES) installdirs
$(Q) for i in $(PROGS); do \
$(ES) " INSTALL $(root_sbindir)/$$i"; \
$(INSTALL_PROGRAM) $$i $(DESTDIR)$(root_sbindir)/$$i; \
done
+ @for i in $(USPROGS); do \
+ echo " INSTALL $(sbindir)/$$i"; \
+ $(INSTALL_PROGRAM) $$i $(DESTDIR)$(sbindir)/$$i; \
+ done
$(Q) for i in ext2 ext3 ext4 ext4dev; do \
$(ES) " LINK $(root_sbindir)/fsck.$$i"; \
(cd $(DESTDIR)$(root_sbindir); \
for i in $(PROGS); do \
$(RM) -f $(DESTDIR)$(root_sbindir)/$$i; \
done
+ for i in $(USPROGS); do \
+ $(RM) -f $(DESTDIR)$(sbindir)/$$i; \
+ done
$(RM) -f $(DESTDIR)$(root_sbindir)/fsck.ext2 \
$(DESTDIR)$(root_sbindir)/fsck.ext3 \
$(DESTDIR)$(root_sbindir)/fsck.ext4 \
$(srcdir)/profile.h prof_err.h $(top_srcdir)/lib/quota/quotaio.h \
$(top_srcdir)/lib/quota/dqblk_v2.h $(top_srcdir)/lib/quota/quotaio_tree.h \
$(top_srcdir)/lib/../e2fsck/dict.h $(srcdir)/problem.h
+pass6.o: $(srcdir)/pass6.c $(top_builddir)/lib/config.h $(srcdir)/e2fsck.h \
+ $(top_srcdir)/lib/ext2fs/ext2_fs.h $(top_builddir)/lib/ext2fs/ext2_types.h \
+ $(top_srcdir)/lib/ext2fs/ext2fs.h $(srcdir)/profile.h prof_err.h \
+ $(top_srcdir)/lib/et/com_err.h $(top_srcdir)/lib/ext2fs/ext2_io.h \
+ $(top_builddir)/lib/ext2fs/ext2_err.h $(top_srcdir)/lib/ext2fs/bitops.h \
+ $(srcdir)/problem.h $(top_srcdir)/lib/ext2fs/lfsck.h
journal.o: $(srcdir)/journal.c $(top_builddir)/lib/config.h \
$(top_builddir)/lib/dirpaths.h $(srcdir)/jfs_user.h $(srcdir)/e2fsck.h \
$(top_srcdir)/lib/ext2fs/ext2_fs.h $(top_builddir)/lib/ext2fs/ext2_types.h \
profile.o: $(srcdir)/profile.c $(top_builddir)/lib/config.h \
$(top_builddir)/lib/dirpaths.h $(top_srcdir)/lib/et/com_err.h \
$(srcdir)/profile.h prof_err.h
+lfsck.o: $(srcdir)/lfsck.c $(top_builddir)/lib/config.h \
+ $(top_builddir)/lib/dirpaths.h $(srcdir)/e2fsck.h \
+ $(top_srcdir)/lib/ext2fs/lfsck.h $(top_srcdir)/lib/ext2fs/ext2_fs.h \
+ $(top_builddir)/lib/ext2fs/ext2_types.h $(top_srcdir)/lib/ext2fs/ext2fs.h
+lfsck_common.o: $(srcdir)/lfsck_common.c $(top_builddir)/lib/config.h \
+ $(top_builddir)/lib/dirpaths.h $(srcdir)/e2fsck.h \
+ $(top_srcdir)/lib/ext2fs/lfsck.h $(top_srcdir)/lib/ext2fs/ext2_fs.h \
+ $(top_builddir)/lib/ext2fs/ext2_types.h $(top_srcdir)/lib/ext2fs/ext2fs.h
sigcatcher.o: $(srcdir)/sigcatcher.c $(top_builddir)/lib/config.h \
$(top_builddir)/lib/dirpaths.h $(srcdir)/e2fsck.h \
$(top_srcdir)/lib/ext2fs/ext2_fs.h $(top_builddir)/lib/ext2fs/ext2_types.h \
or
.B \-p
options.
+@LFSCK_MAN@.TP
+@LFSCK_MAN@.BI \-\-mdsdb " mds_database_filename"
+@LFSCK_MAN@Specify which file to use for an MDS database on a Lustre filesystem.
+@LFSCK_MAN@The mdsdb also needs to be specified when creating an OST database with
+@LFSCK_MAN@.B \-\-ostdb
+@LFSCK_MAN@but the mdsdb file used in a read-only manner.
+@LFSCK_MAN@.TP
+@LFSCK_MAN@.BI \-\-ostdb " ostN_database_filename"
+@LFSCK_MAN@Specify which file to use for an OST database on a Lustre filesystem.
+@LFSCK_MAN@A separate database file should be specified for each OST device in
+@LFSCK_MAN@the filesystem. For convenience a stub
+@LFSCK_MAN@.BR mdsdb .hdr
+@LFSCK_MAN@file is generated for use by the OST e2fsck to avoid the need to
+@LFSCK_MAN@wait for the MDS e2fsck to finish or copy the full mdsdb to the OSTs.
.SH EXIT CODE
The exit code returned by
.B e2fsck
typedef void (*pass_t)(e2fsck_t ctx);
static pass_t e2fsck_passes[] = {
- e2fsck_pass1, e2fsck_pass2, e2fsck_pass3, e2fsck_pass4,
- e2fsck_pass5, 0 };
+ e2fsck_pass1, e2fsck_pass2, e2fsck_pass3, e2fsck_pass4, e2fsck_pass5,
+#ifdef HAVE_LFSCK
+ e2fsck_pass6,
+#endif
+ 0 };
#define E2F_FLAG_RUN_RETURN (E2F_FLAG_SIGNAL_MASK|E2F_FLAG_RESTART)
#define E2F_OPT_FRAGCHECK 0x0800
#define E2F_OPT_JOURNAL_ONLY 0x1000 /* only replay the journal */
#define E2F_OPT_DISCARD 0x2000
+#define E2F_OPT_VERBOSE 0x4000
/*
* E2fsck flags
* Ext4 quota support
*/
quota_ctx_t qctx;
+
+ /* lustre support */
+ int lustre_devtype;
+ char *lustre_mdsdb;
+ char *lustre_ostdb;
+ struct lfsck_outdb_info *lfsck_oinfo;
+
#ifdef RESOURCE_TRACK
/*
* For timing purposes
extern void e2fsck_pass3(e2fsck_t ctx);
extern void e2fsck_pass4(e2fsck_t ctx);
extern void e2fsck_pass5(e2fsck_t ctx);
+extern void e2fsck_pass6(e2fsck_t ctx);
/* e2fsck.c */
extern errcode_t e2fsck_allocate_context(e2fsck_t *ret);
--- /dev/null
+.\" -*- nroff -*-
+.\" Copyright 2004 by Andreas Dilger. All Rights Reserved.
+.\" This file may be copied under the terms of the GNU Public License.
+.\"
+.TH LFSCK 8 "@E2FSPROGS_MONTH@ @E2FSPROGS_YEAR@" "E2fsprogs version @E2FSPROGS_VERSION@"
+.SH NAME
+lfsck \- check a Lustre distributed filesystem
+.SH SYNOPSIS
+.B lfsck
+[
+.B \-cdfhlnv
+]
+.B \--mdsdb
+.I mds_database_file
+.B \--ostdb
+.IR "ost1_database_file " [ "ost2_database_file " ...]
+.I filesystem
+.SH DESCRIPTION
+.B lfsck
+is used to check and repair the distributed coherency of a Lustre filesystem.
+.PP
+.I filesystem
+is the mountpoint of the Lustre filesystem to be checked. e.g.
+.IR /mnt/lustre ).
+.SH DESCRIPTION
+Lfsck is used to do the distributed coherency checking of a Lustre filesystem
+after
+.B e2fsck
+has been run on each of the local MDS and OST device filesystems.
+.PP
+.B Note that the lfsck command is obsolete for Lustre 2.6 and later.
+.B Starting in Lustre 2.4 the
+.BR lctl (8)
+.B start_lfsck command can do online verification of the Lustre-specific
+.B state on the MDT without the need for external database files, and in
+.B Lustre 2.6 this includes distributed checking of the MDT-OST consistency,
+.B including cleanup of orphan OST objects.
+.PP
+.B For older versions of Lustre, lfsck
+will ensure that objects are not referenced by multiple MDS files, that there
+are no orphan objects on the OSTs (objects that do not have any file on the
+MDS which references them), and that all of the objects referenced by the
+MDS exist. Under normal circumstances Lustre will maintain such coherency
+by distributed logging mechanisms, but under exceptional circumstances that
+may fail (e.g. disk failure, filesystem corruption leading to e2fsck repair).
+.SH OPTIONS
+.TP
+.B \-c
+Create (empty) missing OST objects referenced by MDS inodes.
+.TP
+.B \-d
+Delete orphaned objects from the filesystem. Since objects on the OST are
+often only one of several stripes of a file it can be difficult to put
+multiple objects back together into a single usable file.
+.TP
+.B \-h
+Print a brief help message.
+.TP
+.B \-l
+Put orphaned objects into a lost+found directory in the root of the filesystem.
+.TP
+.B \-n
+Do not repair the filesystem, just perform a read-only check (default).
+.TP
+.B \-v
+Verbose operation - more verbosity by specifing option multiple times.
+.TP
+.BI \-\-mdsdb " mds_database_file"
+The MDS database file created by running
+.B e2fsck --mdsdb mds_database_file device
+on the MDS backing device. This is required.
+.TP
+.BI \-\-ostdb " ost1_database_file " [ "ost2_database_file " ...]
+The OST database files created by running
+.B e2fsck --ostdb ost_database_file device
+on each of the OST backing devices. These are required, unless an OST is
+unavailable, in which case all objects thereon will be considered missing.
+.SH REPORTING BUGS
+Bugs should be reported to Sun Microsystems, Inc. via Bugzilla:
+http://bugzilla.lustre.org/
+.SH AUTHOR
+This version of
+.B lfsck
+was originally written by Liam Kelleher <liam.kelleher@hp.com>
+and maintained by Andreas Dilger <andreas.dilger@intel.com>
+.SH SEE ALSO
+.BR e2fsck (8)
--- /dev/null
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim:shiftwidth=8:tabstop=8:
+ *
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
+ *
+ * This program 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
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2004 Hewlett-Packard Co.
+ * Copyright 2004, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ * Copyright (c) 2011 Whamcloud, Inc.
+ */
+/***********************************************************************
+ * This program takes as input files generated by running e2fsck on
+ * the mds and ost filesystems. The file generated for each ost
+ * contains a table including the object id and size for each object
+ * extant on the ost in each entry.
+ * The file generated from scanning the mds filesystem with e2fsck
+ * contains multiple tables one for each ost. In each table an entry
+ * contains the mds fid as well as the object id on the appropriate
+ * ost. In addition there is an additional table that holds the mds_fid
+ * and the containing directory fid for each entry. This is used for
+ * name lookup.
+ * There are three basic checks
+ * 1) Make sure that multiple mds entries do not reference the same object
+ * 2) Cross reference each object on each ost to make sure a "containing"
+ * file for this exists on the mds
+ * 3) For each file on the mds make sure that the associated objects exist
+ * on the osts
+ * These checks and potential correction for errors found are run from
+ * run_pass*
+ * Each of these checks is just iterate through the appropriate table and
+ * cross check against another table and if errors are found repair.
+ ***************************************************************************/
+#include "config.h"
+#ifndef _GNU_SOURCE
+# define _GNU_SOURCE 1
+#endif
+#include <pthread.h>
+#include <stdio.h>
+#include <stdarg.h>
+#include <string.h>
+#include <time.h>
+#include <errno.h>
+#include <limits.h>
+#include <stdlib.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <utime.h>
+#include <unistd.h>
+#include <sys/mman.h>
+#include <getopt.h>
+#include <mntent.h>
+#include <dirent.h>
+#ifdef HAVE_SYS_IOCTL_H
+#include <sys/ioctl.h>
+#endif
+#include "../version.h"
+#include "e2fsck.h"
+#include "ext2fs/lfsck.h"
+#include "ext2fs/ext2fs.h"
+
+struct lfsck_fids {
+ int depth;
+ struct lu_fid *fids;
+};
+
+struct lfsck_thread_info {
+ struct lfsck_mds_hdr *mds_hdr;
+ DB *mds_direntdb;
+ DB *mds_sizeinfodb;
+ __u32 start_ost_idx;
+ __u32 end_ost_idx;
+ int status;
+};
+
+struct lfsck_saved_duplicates {
+ struct lu_fid ld_mds_fid;
+ struct ostid ld_oi;
+ __u32 ld_ost_idx;
+ char *ld_link;
+};
+
+#ifndef O_LARGEFILE
+#define O_LARGEFILE 0
+#endif
+
+#define LOG_PATH "/var/log/lfsck.log"
+#define RLIMIT 1024
+
+/* Procedure declarations */
+
+char *progname = "lfsck";
+FILE *logfile;
+
+int lfsck_help;
+int lfsck_save;
+int lfsck_delete;
+int lfsck_create;
+int lfsck_force;
+int lfsck_verbose;
+int lfsck_yes;
+
+int num_threads = 1;
+
+char mnt_path[PATH_MAX];
+char *mds_file;
+char lostandfounddir[PATH_MAX];
+char dupedir[PATH_MAX];
+char *ost_files[LOV_MAX_OSTS];
+int num_ost_files;
+
+struct obd_uuid lfsck_uuid[LOV_MAX_OSTS];
+int lov_tgt_count = LOV_MAX_OSTS;
+
+struct lfsck_saved_duplicates *lfsck_duplicates;
+int lfsck_dup_saved;
+int fixed;
+int fix_failed;
+
+pthread_cond_t init_cond = PTHREAD_COND_INITIALIZER;
+pthread_mutex_t init_mutex = PTHREAD_MUTEX_INITIALIZER;
+pthread_mutex_t phase_lock = PTHREAD_MUTEX_INITIALIZER;
+pthread_mutex_t size_lock = PTHREAD_MUTEX_INITIALIZER;
+int all_started;
+
+int lfsck_path2fid(const char *path, struct lu_fid *fid)
+{
+ struct stat st;
+ long generation;
+ int fd;
+ int rc = 0;
+
+#ifdef LL_IOC_PATH2FID
+ rc = llapi_path2fid(path, fid);
+ if (rc == 0 || rc != -ENOTTY)
+ return rc;
+#endif
+ /* LL_IOC_PATH2FID was landed in 1.8.2. If it doesn't exist at compile
+ * time, or it fails at runtime with a return ENOTTY indicating that
+ * the ioctl is unimplemented, emulate it here for the older clients.
+ * Assume the server is running Lustre 1.x and create an IGIF FID,
+ * since < 1.8.4 will not work properly with 2.x servers anyway. */
+ fd = open(path, O_RDONLY);
+ if (fd < 0)
+ return -errno;
+
+ rc = fstat(fd, &st);
+ if (rc < 0) {
+ rc = -errno;
+ goto out;
+ }
+ fid->f_seq = st.st_ino;
+
+ rc = ioctl(fd, FSFILT_IOC_GETVERSION, &generation);
+ if (rc < 0) {
+ rc = -errno;
+ goto out;
+ }
+ fid->f_oid = generation;
+ fid->f_ver = 0;
+
+out:
+ close(fd);
+ return rc;
+}
+
+#define VERBOSE(lvl, fmt, args...) \
+do { if (lfsck_verbose >= lvl) printf(fmt, ## args); } while (0)
+
+int log_open()
+{
+ time_t tm;
+
+ logfile = fopen(LOG_PATH, "a");
+ if (logfile == NULL) {
+ fprintf(stderr, "%s: Failed to open log file %s\n",
+ progname, LOG_PATH);
+ return -EPERM;
+ }
+
+ time(&tm);
+ fprintf(logfile, "===============================================\n\n");
+ fprintf(logfile, "Starting lfsck %s\n", ctime(&tm));
+ return 0;
+}
+
+int log_close(int status)
+{
+ time_t tm;
+
+ if (logfile == NULL)
+ return 0;
+
+ time(&tm);
+ if (status < 0)
+ fprintf(logfile, "ERROR: lfsck aborted\n");
+ else
+ fprintf(logfile, "lfsck run completed: %s\n", ctime(&tm));
+
+ fprintf(logfile, "===============================================\n\n");
+
+ fclose(logfile);
+ logfile = NULL;
+ return 0;
+}
+
+static void log_write(char *fmt, ...) __attribute__ ((format (printf, 1, 2)));
+
+static void log_write(char *fmt, ...)
+{
+ va_list args;
+
+ if (logfile) {
+ va_start(args, fmt);
+ vfprintf(logfile, fmt, args);
+ va_end(args);
+ }
+ va_start(args, fmt);
+ vfprintf(stderr, fmt, args);
+ va_end(args);
+}
+
+void usage()
+{
+ printf("\n");
+ printf("usage: lfsck [-cdfhlnv] --mdsdb mdsdb "
+ "--ostdb ostdb1 [ostdb2 ...] filesystem\n\n");
+ printf("\t-m|--mdsdb mdsdb MDS database from e2fsck --mdsdb\n");
+ printf("\t-o|--ostdb ostdb OST databases from e2fsck --ostdb\n");
+ printf("\tfilesytem full path of lustre mountpoint\n");
+ printf("\t[-c|--create] create missing objects\n");
+ printf("\t[-d|--delete] delete orphan objects\n");
+ printf("\t[-f|--force] force running if fs appears unmounted\n");
+ printf("\t[-h|--help] print this message\n");
+ printf("\t[-l|--lostfound] save orphans objects to lost+found\n");
+ printf("\t[-n|--nofix] do not fix filesystem errors (default)\n");
+ printf("\t[-v|--verbose] print verbose runtime messages\n");
+ /* printf("\t[-y|--yes] do all cleanup automatically\n"); */
+ printf("\n");
+}
+
+/*
+ * Get the lov config for the filesystem - this is primarily used to correlate
+ * each ost db file with its index in the lov configuration. Obviously this is
+ * critical.
+ */
+int get_lov_config()
+{
+ int fd, rc;
+
+ fd = open(mnt_path, O_RDONLY);
+ if (fd < 0) {
+ log_write("Error: opening %s\n", mnt_path);
+ return -errno;
+ }
+
+ rc = llapi_lov_get_uuids(fd, lfsck_uuid, &lov_tgt_count);
+
+ close(fd);
+ return rc;
+}
+
+int parse_args(int argc, char *argv[])
+{
+ int option_index = 0;
+ char *path_name = NULL;
+ struct option long_options[] = {
+ { "create", 1, NULL, 'c' },
+ { "delete", 0, NULL, 'd' },
+ { "force", 0, NULL, 'f' },
+ { "help", 0, NULL, 'h' },
+ { "lostfound", 0, NULL, 'l' },
+ { "mdsdb", 1, NULL, 'm' },
+ { "mdtdb", 1, NULL, 'm' },
+ { "nofix", 0, NULL, 'n' },
+ { "ostdb", 1, NULL, 'o' },
+ { "threads", 1, NULL, 't' },
+ { "verbose", 0, NULL, 'v' },
+ /* { "yes", 0, NULL, 'y' }, */
+ { 0, 0, 0, 0 }
+ };
+ struct mntent *mnt_info;
+ char tmp[PATH_MAX];
+ char *dbpath;
+ int c, found;
+ char *p1;
+ FILE *fp;
+
+ if (argc < 6 )
+ return -EINVAL;
+
+ while ((c = getopt_long(argc, argv, "-cdfhlm:no:t:vy",
+ long_options, &option_index)) != -1) {
+ switch (c) {
+ case 'c':
+ lfsck_create++;
+ break;
+ case 'd':
+ lfsck_delete++;
+ break;
+ case 'f':
+ lfsck_force++;
+ break;
+ case 'h':
+ lfsck_help++;
+ break;
+ case 'l':
+ lfsck_save++;
+ break;
+ case 'm':
+ VERBOSE(1, "MDSDB: %s\n", optarg);
+ dbpath = malloc(PATH_MAX);
+ if (dbpath == NULL) {
+ fprintf(stderr, "error allocating dbpath\n");
+ return -ENOMEM;
+ }
+ strcpy(tmp, optarg);
+ if (realpath(my_dirname(tmp), dbpath) == NULL) {
+ fprintf(stderr, "Failure to resolve path %s\n",
+ optarg);
+ free(dbpath);
+ exit(1);
+ }
+
+ strcpy(tmp, optarg);
+ sprintf(dbpath+strlen(dbpath), "/%s", my_basename(tmp));
+ mds_file = dbpath;
+ break;
+ case 'n':
+ lfsck_create = 0;
+ lfsck_delete = 0;
+ lfsck_save = 0;
+ break;
+ case 1:
+ if (optind == argc) { /* last one is mountpoint */
+ VERBOSE(1, "MOUNTPOINT: %s\n", optarg);
+ path_name = optarg;
+ break;
+ }
+ /* Otherwise it is another ostdb file */
+ case 'o':
+ {
+ char *ost_path;
+
+ VERBOSE(1, "OSTDB[%u]: %s\n", num_ost_files, optarg);
+ p1 = optarg;
+ do {
+ dbpath = malloc(PATH_MAX);
+ if (dbpath == NULL) {
+ fprintf(stderr,
+ "error allocate ost_files[%d]\n",
+ num_ost_files);
+ return -ENOMEM;
+ }
+
+ /* Old-style arguments are comma separated */
+ ost_path = strsep(&p1, ",");
+ strcpy(tmp, ost_path);
+ if (realpath(my_dirname(tmp), dbpath) == NULL) {
+ fprintf(stderr, "Failure to resolve "
+ "path %s\n", optarg);
+ for (c = 0; c < num_ost_files; c++)
+ free(ost_files[c]);
+ free(dbpath);
+ exit(1);
+ }
+
+ strcpy(tmp, ost_path);
+ sprintf(dbpath+strlen(dbpath), "/%s",
+ my_basename(tmp));
+ ost_files[num_ost_files] = dbpath;
+ num_ost_files++;
+ } while (p1 != NULL);
+ break;
+ }
+ case 't':
+ num_threads = strtol(optarg, NULL, 0);
+ if (num_threads == ULONG_MAX)
+ return -EINVAL;
+
+ if (num_threads > 128)
+ num_threads = 128;
+ if (num_threads == 0)
+ num_threads = 1;
+ break;
+ case 'v':
+ lfsck_verbose++;
+ break;
+ case 'y':
+ lfsck_yes++;
+ break;
+ default:
+ fprintf(stderr, "unknown option %c\n", c);
+ return -EINVAL;
+ }
+ }
+
+ if (path_name == NULL)
+ path_name = argv[optind];
+
+ if (lfsck_yes && !lfsck_save && !lfsck_delete && !lfsck_create) {
+ fprintf(stderr, "--yes requires either --save or --delete, or"
+ "--create\n");
+ return -EINVAL;
+ }
+
+ if (realpath(path_name, mnt_path) == NULL) {
+ fprintf(stderr, "error getting real mount path_name\n");
+ return -EINVAL;
+ }
+ fp = setmntent(MOUNTED, "r");
+ if (fp == NULL) {
+ fprintf(stderr, "error opening /etc/mtab\n");
+ return -EINVAL;
+ }
+
+ found = 0;
+ while ((mnt_info = getmntent(fp)) != NULL) {
+ if (strcmp(mnt_info->mnt_dir, mnt_path) == 0 &&
+ llapi_is_lustre_mnttype(mnt_info->mnt_type)) {
+ found = 1;
+ break;
+ }
+ }
+ endmntent(fp);
+
+ if (!found) {
+ if (lfsck_force) {
+ fprintf(stderr, "lfsck: %s unmounted? forcing\n",
+ mnt_path);
+ } else {
+ fprintf(stderr, "lfsck: %s not mounted\n", mnt_path);
+ return -EINVAL;
+ }
+ }
+
+ if (mds_file == NULL || ost_files[0] == NULL) {
+ fprintf(stderr, "--mdsdb or --ostdb unspecified\n");
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+/*
+ * This is called from lfsck_get_path and also recursively.
+ * This function is used on error paths when the name of an mds fid has
+ * to be determined. It relies on the order of directory search that was
+ * run in e2fsck when generating the mds_fid + containing dir table.
+ * It searches through the table for the correct mds_fid. When this
+ * is found a list of fids which are the fids of the directory tree up
+ * to the containing directory of the sought for fid is returned.
+ * When called recursively it continues search from the current point and
+ * when the recursive call returns the search is continued from the
+ * current search point as well. Basically is just traverses the list once.
+ * For a file like <mntpt>/aaa/ccc/ddd the fids of aaa ccc and the fid
+ * for ddd would also be returned.
+ */
+static int lfsck_get_fids(struct lu_fid *mds_fid, DB *mds_direntdb,
+ int depth, struct lfsck_fids *lfidp)
+{
+ struct lfsck_mds_dirent mds_dirent;
+ int rc = 0;
+ DBT key, data;
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = mds_fid;
+ key.size = sizeof(*mds_fid);
+ data.data = &mds_dirent;
+ data.size = data.ulen = sizeof(mds_dirent);
+ data.flags = DB_DBT_USERMEM;
+
+ rc = mds_direntdb->get(mds_direntdb, NULL, &key, &data, 0);
+ if (rc && !lfsck_is_dirfid_root(mds_fid)) {
+ log_write("Failed to find fid "DFID": %s\n", PFID(mds_fid),
+ db_strerror(rc));
+ return -ENOENT;
+ }
+ letocpu_mds_dirent(&mds_dirent);
+ if (lfsck_is_dirfid_root(&mds_dirent.mds_dirfid)) {
+ lfidp->fids = malloc(sizeof(*lfidp->fids) * (depth + 1));
+ if (lfidp->fids == NULL)
+ return -ENOMEM;
+
+ lfidp->depth = depth;
+ lfidp->fids[depth] = mds_dirent.mds_fid;
+ return 0;
+ }
+ rc = lfsck_get_fids(&mds_dirent.mds_dirfid, mds_direntdb,
+ depth + 1, lfidp);
+ if (rc)
+ return rc;
+
+ lfidp->fids[depth] = mds_dirent.mds_fid;
+ return 0;
+}
+
+/* This function determines a path to a file given an mds fid.
+ * The workhorse function is lfsck_get_fids which once given a
+ * fid return a list of directory fids from the "root" directory to
+ * the fid in question. Using these fids we can construct the path to
+ * the file by using readir()
+ */
+static int lfsck_get_path(struct lu_fid *mds_fid, DB *mds_direntdb,
+ char *path, int path_len)
+{
+ struct lfsck_fids lfids;
+ DIR *dir;
+ struct dirent *dent;
+ int rc, i;
+ int cur_len = 0;
+
+ VERBOSE(2, "lookup path for FID "DFID"\n", PFID(mds_fid));
+
+ lfids.fids = NULL;
+ lfids.depth = 0;
+
+ rc = lfsck_get_fids(mds_fid, mds_direntdb, 0, &lfids);
+ if (rc != 0) {
+ rc = -ENOENT;
+ goto out;
+ }
+
+ if (strlen(mnt_path) + 1 > path_len) {
+ rc = -ENOMEM;
+ goto out;
+ }
+ cur_len = strlen(mnt_path);
+ path[strlen(mnt_path)] = 0;
+ memcpy(path, mnt_path, strlen(mnt_path));
+ /* Skip the first dir since this would be "ROOT" */
+ rc = 0;
+ for (i = lfids.depth - 1; i >= 0; i--) {
+ dir = opendir(path);
+ if (dir == NULL) {
+ rc = -errno;
+ goto out;
+ }
+ while (1) {
+ char path_tmp[PATH_MAX];
+ struct lu_fid fid;
+
+ dent = readdir(dir);
+ if (dent == NULL) {
+ closedir(dir);
+ rc = -ENOENT;
+ goto out;
+ }
+
+ if (strlen(path) + strlen(dent->d_name)+2 >= PATH_MAX) {
+ closedir(dir);
+ rc = -ENAMETOOLONG;
+ goto out;
+ }
+
+ sprintf(path_tmp, "%s/%s", path, dent->d_name);
+ rc = lfsck_path2fid(path_tmp, &fid);
+ if (rc)
+ continue;
+
+ rc = 0;
+ if (lfsck_fidcmp(&fid, &lfids.fids[i]) == 0) {
+ if (cur_len + 1 + strlen(dent->d_name) >
+ path_len) {
+ rc = -ENOMEM;
+ closedir(dir);
+ goto out;
+ }
+ path[cur_len] = '/';
+ cur_len++;
+ memcpy(&path[cur_len], dent->d_name,
+ strlen(dent->d_name));
+ cur_len += strlen(dent->d_name);
+ path[cur_len] = 0;
+ closedir(dir);
+ break;
+ }
+ }
+ }
+out:
+ if (lfids.fids)
+ free(lfids.fids);
+ return rc;
+}
+
+/*
+ * Used by pass1 to save the ids of files which reference the same
+ * objects. This is then used by pass4 to repair these files
+ */
+static int lfsck_save_duplicate(const struct lfsck_mds_objent *mds_obj)
+{
+ VERBOSE(2, "save duplicate object %u:"DOSTID" FID "DFID"\n",
+ mds_obj->mds_ostidx, POSTID(&mds_obj->mds_oi),
+ PFID(&mds_obj->mds_fid));
+
+ pthread_mutex_lock(&phase_lock);
+ if (lfsck_duplicates == NULL) {
+ lfsck_duplicates = malloc(sizeof(*lfsck_duplicates) *
+ RLIMIT);
+ if (lfsck_duplicates == NULL)
+ return -EINVAL;
+
+ } else if (!((lfsck_dup_saved + 1) % RLIMIT)) {
+ size_t size = (((lfsck_dup_saved + 1) / RLIMIT) + 1) *
+ sizeof(*lfsck_duplicates) * RLIMIT;
+ void *tmp = realloc(lfsck_duplicates, size);
+
+ if (tmp == NULL)
+ return -ENOMEM;
+
+ lfsck_duplicates = tmp;
+ }
+ lfsck_duplicates[lfsck_dup_saved].ld_mds_fid = mds_obj->mds_fid;
+ lfsck_duplicates[lfsck_dup_saved].ld_oi = mds_obj->mds_oi;
+ lfsck_duplicates[lfsck_dup_saved].ld_ost_idx = mds_obj->mds_ostidx;
+ lfsck_duplicates[lfsck_dup_saved].ld_link = NULL;
+ lfsck_dup_saved++;
+ pthread_mutex_unlock(&phase_lock);
+ return 0;
+}
+
+/*
+ * Check for duplicate ost objects on mds. Run through the table of
+ * mds_fid/ost object to make sure that each ost object is only
+ * refrenced by one mds entry. If a duplicate is found save the information
+ * for repair in pass4
+ */
+int lfsck_run_pass1(__u32 ost_idx, DB *mds_ostdb, DB *ost_db, DB *mds_direntdb)
+{
+ int i = 0;
+ int error = 0;
+ int rc = 0;
+ struct lfsck_mds_objent mds_obj, mds_obj2;
+ unsigned long count = 0;
+ DBT key, data;
+ DBC *dbcp = NULL;
+
+ log_write("%s: ost_idx %d: pass1: check for duplicate objects\n",
+ progname, ost_idx);
+
+ rc = mds_ostdb->cursor(mds_ostdb, NULL, &dbcp, 0);
+ if (rc != 0) {
+ log_write("%s: error acquiring cursor for database: %s\n",
+ progname, db_strerror(rc));
+ goto out;
+ }
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ data.data = &mds_obj;
+ data.size = data.ulen = sizeof(mds_obj);
+ data.flags = DB_DBT_USERMEM;
+
+ while ((rc = dbcp->c_get(dbcp, &key, &data, DB_NEXT)) == 0) {
+ DBT data_dup;
+ db_recno_t num_dup;
+
+ count++;
+ rc = dbcp->c_count(dbcp, &num_dup, 0);
+ if (rc != 0) {
+ log_write("%s: [%u] getting object refcount: %s\n",
+ progname, ost_idx, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+ if (num_dup <= 1)
+ continue;
+
+ letocpu_mds_objent(&mds_obj);
+
+ if (lfsck_save_duplicate(&mds_obj))
+ fix_failed++;
+
+ for (i = 1; i < num_dup; i++) {
+ memset(&data_dup, 0, sizeof(data_dup));
+ data_dup.data = &mds_obj2;
+ data_dup.size = data_dup.ulen = sizeof(mds_obj2);
+ data_dup.flags = DB_DBT_USERMEM;
+ rc = dbcp->c_get(dbcp, &key, &data_dup, DB_NEXT);
+ if (rc != 0) {
+ log_write("%s: acquiring duplicate info: %s\n",
+ progname, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+ letocpu_mds_objent(&mds_obj2);
+
+ if (!lfsck_fidcmp(&mds_obj.mds_fid, &mds_obj2.mds_fid)){
+ log_write("%s: [%u] hard link on FID "DFID" is"
+ " not a duplicate object "DOSTID"\n",
+ progname, ost_idx,
+ PFID(&mds_obj.mds_fid),
+ POSTID(&mds_obj.mds_oi));
+ continue;
+ }
+
+ if (lfsck_save_duplicate(&mds_obj2))
+ fix_failed++;
+ }
+ }
+
+ if (rc != DB_NOTFOUND) {
+ log_write("%s: error reading from inode database: %s\n",
+ progname, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+
+ rc = 0;
+ if (error == 0) {
+ log_write("%s: ost_idx %d: pass1 OK (%lu files total)\n",
+ progname, ost_idx, count);
+ } else {
+ log_write("%s: ost_idx %d: pass1 ERROR: %d duplicate "
+ "entries found (fixed in pass4) (%lu files total)\n",
+ progname, ost_idx, error, count);
+ }
+out:
+ if (dbcp)
+ dbcp->c_close(dbcp);
+
+ return rc;
+}
+
+#ifndef LL_IOC_RECREATE_OBJ
+#define LL_IOC_RECREATE_OBJ _IOW('f', 157, long) /* 1.x object IDs */
+#endif
+#ifndef LL_IOC_RECREATE_FID
+#define LL_IOC_RECREATE_FID _IOW('f', 157, struct lu_fid) /* 2.x FIDs */
+#endif
+
+/* If an MDS file is missing an object recreate object using an ioctl call */
+static int lfsck_recreate_obj(int cmd, void *creat, struct ostid *oi,
+ __u32 ost_idx, char *path)
+{
+
+ int fd;
+ int rc;
+
+ if (!lfsck_create) {
+ log_write("[%u]: %s object %s "DOSTID" not created\n", ost_idx,
+ path, cmd == LL_IOC_RECREATE_FID ? "FID":"ID",
+ POSTID(oi));
+ return 0;
+ }
+
+ fd = open(path, O_LARGEFILE | O_RDONLY, 0);
+ if (fd < 0) {
+ rc = -errno;
+ log_write("[%u]: FAILED to open %s missing obj "DOSTID"\n",
+ ost_idx, path, POSTID(oi));
+ fix_failed++;
+ return rc;
+ }
+
+ rc = ioctl(fd, cmd, creat);
+ if (rc) {
+ rc = -errno;
+ log_write("[%u]: failed to recreate %s missing obj "DOSTID"\n",
+ ost_idx, path, POSTID(oi));
+ fix_failed++;
+ } else {
+ log_write("[%u]: recreated %s missing obj "DOSTID"\n",
+ ost_idx, path, POSTID(oi));
+ fixed++;
+ }
+ close(fd);
+
+ return rc;
+}
+
+/*
+ * If size checking is enabled see if this ost is "adding" to the file size
+ * if it is then just calculate the new size and save.
+ */
+static int lfsck_calc_size(struct lfsck_mds_objent *mds_obj,
+ struct lfsck_ost_objent *ost_obj,
+ DB *mds_sizeinfodb)
+{
+ int rc = 0;
+#ifdef LFSCK_CHECK_SIZE
+ struct lfsck_mds_szinfo mds_szinfo1;
+ __u64 calc_size;
+ DBT key, data;
+ __u64 chunks, rem;
+
+ if (ost_obj->ost_size == 0)
+ return 0;
+
+ pthread_mutex_lock(&size_lock);
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &mds_obj->mds_fid;
+ key.size = sizeof(mds_obj->mds_fid);
+ data.data = &mds_szinfo1;
+ data.size = data.ulen = sizeof(mds_szinfo1);
+ data.flags = DB_DBT_USERMEM;
+ rc = mds_sizeinfodb->get(mds_sizeinfodb, NULL, &key, &data, 0);
+ if (rc != 0) {
+ log_write("Failure to get sizeinfo "LPU64"\n",
+ mds_obj->mds_fid);
+ pthread_mutex_unlock(&size_lock);
+ return -ENOENT;
+ }
+ letocpu_mds_szinfo(&mds_szinfo1);
+ assert (mds_szinfo1.mds_stripe_pattern == LOV_PATTERN_RAID0);
+ chunks = ost_obj->ost_size / mds_szinfo1.mds_stripe_size;
+ rem = ost_obj->ost_size % mds_szinfo1.mds_stripe_size;
+ if (rem == 0) {
+ calc_size = (((chunks - 1 ) * mds_szinfo1.mds_stripe_size) *
+ mds_szinfo1.mds_stripe_count);
+ calc_size += mds_szinfo1.mds_stripe_size *
+ (mds_obj->mds_ostoffset + 1);
+ } else {
+ calc_size = ((chunks * mds_szinfo1.mds_stripe_size)
+ * mds_szinfo1.mds_stripe_count);
+ if (mds_obj->mds_ostoffset == 0) {
+ calc_size += rem;
+ } else {
+ calc_size += mds_szinfo1.mds_stripe_size *
+ mds_obj->mds_ostoffset;
+ calc_size += rem;
+ }
+ }
+ if (calc_size > mds_szinfo1.mds_calc_size) {
+ mds_szinfo1.mds_calc_size = calc_size;
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &mds_obj->mds_fid;
+ key.size = sizeof(mds_obj->mds_fid);
+ data.data = &mds_szinfo1;
+ data.size = sizeof(mds_szinfo1);
+ cputole_mds_szinfo(&mds_szinfo1);
+ /* Make sure we overwrite */
+ rc = mds_sizeinfodb->put(mds_sizeinfodb, NULL, &key, &data, 0);
+ if (rc != 0) {
+ log_write("Failure to update sizeinfo data\n");
+ pthread_mutex_unlock(&size_lock);
+ return -EIO;
+ }
+ }
+ pthread_mutex_unlock(&size_lock);
+#endif /* LFSCK_CHECK_SIZE */
+ return rc;
+}
+
+/*
+ * Check for dangling inode.
+ * pass runs through the mds table for an ost and checks again the ost table
+ * that the object refrenced on the mds exists on the ost
+ */
+int lfsck_run_pass2(__u32 ost_idx, struct lfsck_mds_hdr *mds_hdr,
+ DB *mds_ostdb, DB *ostdb,
+ DB *mds_direntdb, DB *mds_sizeinfodb)
+{
+ struct lfsck_mds_objent mds_obj;
+ struct lfsck_ost_objent ost_obj;
+ int error = 0, rc = 0;
+ unsigned long count = 0;
+ char *path;
+ DBC *dbcp = NULL;
+ DBT key, data;
+ __u64 max_objid = mds_hdr->mds_max_ost_id[ost_idx];
+ __u64 mds_connect_flags = 0;
+
+ rc = llapi_get_connect_flags(mnt_path, &mds_connect_flags);
+ /* Ignore the error here, and assume it is an older 1.8.x without
+ * LL_IOC_GET_CONNECT_FLAGS. We only use this for 2.x detection. */
+
+ log_write("lfsck: ost_idx %d: pass2: check for missing inode objects\n",
+ ost_idx);
+
+ path = malloc(PATH_MAX);
+ if (path == NULL) {
+ log_write("lfsck: [%u]: pass2 ERROR: out of memory\n",
+ ost_idx);
+ return -ENOMEM;
+ }
+
+ rc = mds_ostdb->cursor(mds_ostdb, NULL, &dbcp, 0);
+ if (rc != 0) {
+ log_write("[%u]: error acquiring cursor for mds table: %s\n",
+ ost_idx, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ data.data = &mds_obj;
+ data.size = data.ulen = sizeof(mds_obj);
+ data.flags = DB_DBT_USERMEM;
+ while ((rc = dbcp->c_get(dbcp, &key, &data, DB_NEXT)) == 0) {
+ DBT key_ost, data_ost;
+
+ count++;
+ letocpu_mds_objent(&mds_obj);
+
+ if (mds_hdr->mds_flags & E2F_OPT_READONLY &&
+ mds_obj.mds_oi.oi_id > max_objid) {
+ VERBOSE(2, "[%u] skipping MDS FID "DFID": object "DOSTID
+ " > max "LPU64"\n", ost_idx,
+ PFID(&mds_obj.mds_fid), POSTID(&mds_obj.mds_oi),
+ max_objid);
+ continue;
+ }
+
+ memset(&key_ost, 0, sizeof(key_ost));
+ memset(&data_ost, 0, sizeof(data_ost));
+ key_ost.data = &mds_obj.mds_oi;
+ key_ost.size = sizeof(mds_obj.mds_oi);
+ data_ost.data = &ost_obj;
+ data_ost.size = data_ost.ulen = sizeof(ost_obj);
+ data_ost.flags = DB_DBT_USERMEM;
+ rc = ostdb->get(ostdb, NULL, &key_ost, &data_ost, 0);
+ if (rc != 0) {
+ letocpu_ost_objent(&ost_obj);
+ if (rc == DB_NOTFOUND) {
+ struct lu_fid fid;
+ struct create18 {
+ __u64 lrc_id;
+ __u32 lrc_ost_idx;
+ } create18;
+ void *create;
+ int cmd;
+
+ if (lfsck_get_path(&mds_obj.mds_fid,
+ mds_direntdb, path,
+ PATH_MAX)) {
+ VERBOSE(1,"[%u]: MDS FID "DFID" object "
+ DOSTID" deleted?\n", ost_idx,
+ PFID(&mds_obj.mds_fid),
+ POSTID(&mds_obj.mds_oi));
+ continue;
+ }
+ error++;
+ if (mds_connect_flags & OBD_CONNECT_FID) {
+ ostid_idif_unpack(&mds_obj.mds_oi,
+ &fid, ost_idx);
+ create = &fid;
+ cmd = LL_IOC_RECREATE_FID;
+ } else {
+ create18.lrc_id = mds_obj.mds_oi.oi_id;
+ create18.lrc_ost_idx = ost_idx;
+ create = &create18;
+ cmd = LL_IOC_RECREATE_OBJ;
+ }
+
+ lfsck_recreate_obj(cmd, create,&mds_obj.mds_oi,
+ ost_idx, path);
+ } else {
+ log_write("[%u]: error looking up object "DOSTID
+ ": %s\n", ost_idx,
+ POSTID(&mds_obj.mds_oi),
+ db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+ }
+ if (lfsck_calc_size(&mds_obj, &ost_obj, mds_sizeinfodb)) {
+ log_write("[%u]: error updating file size for object "
+ DOSTID": %s\n", ost_idx,
+ POSTID(&mds_obj.mds_oi), strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+ }
+ if (rc != DB_NOTFOUND) {
+ log_write("[%u]: error getting next inode: %s\n",
+ ost_idx, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+ rc = 0;
+ if (error == 0) {
+ log_write("lfsck: ost_idx %d: pass2 OK (%lu objects)\n",
+ ost_idx, count);
+ } else {
+ log_write("lfsck: ost_idx %d: pass2 ERROR: %d dangling inodes "
+ "found (%lu files total)\n", ost_idx, error, count);
+ }
+
+out:
+ dbcp->c_close(dbcp);
+ free(path);
+ return 0;
+}
+
+/*
+ * If an object exists on an ost but is not referenced by an entry on the mds
+ * then create a lost+found entry and set the EA on the file so that the
+ * orphaned object is picked up. If the object is requested to be deleted
+ * an unlink on this lost+found file will now delete same
+ */
+int lfsck_fix_orphan(__u32 ost_idx, struct ostid *oi,
+ struct obd_uuid *uuid, int delete)
+{
+ struct lov_user_md *lum;
+ char file[PATH_MAX];
+ int fd, lum_size, rc = 0;
+ struct utimbuf utimbuf = { 0, 0 };
+
+ lum_size = LOV_EA_SIZE(lum, 1);
+
+ lum = malloc(lum_size);
+ if (lum == NULL) {
+ log_write("%s: out of memory on EA (%u) orphan %u:"DOSTID"\n",
+ progname, lum_size, ost_idx, POSTID(oi));
+ return -ENOMEM;
+ }
+
+ memset(file, 0, PATH_MAX);
+ sprintf(file, "%s/%s-"DOSTID, lostandfounddir, uuid->uuid, POSTID(oi));
+
+ fd = open(file, O_CREAT|O_EXCL|O_LOV_DELAY_CREATE, 0600);
+ if (fd < 0) {
+ rc = -errno;
+ log_write("%s: unable to open %s for orphan %u:"DOSTID": %s\n",
+ progname, file, ost_idx, POSTID(oi), strerror(-rc));
+ goto out_free;
+ }
+ lum->lmm_magic = LOV_USER_MAGIC;
+ lum->lmm_pattern = LOV_PATTERN_RAID0;
+ lum->lmm_stripe_size = 1048576;
+ lum->lmm_stripe_offset = 0;
+ lum->lmm_stripe_count = 1;
+ memcpy(&lum->lmm_objects[0].l_ost_oi, oi, sizeof(*oi));
+ lum->lmm_objects[0].l_ost_gen = 0;
+ lum->lmm_objects[0].l_ost_idx = ost_idx;
+
+ /* reset the MDS timestamps so we can see the OST timestamps */
+ utime(file, &utimbuf);
+
+ if (ioctl(fd, LL_IOC_LOV_SETEA, lum) < 0) {
+ rc = -errno;
+ log_write("%s: unable to open %s for orphan %u:"DOSTID": %s\n",
+ progname, file, ost_idx, POSTID(oi), strerror(-rc));
+ }
+
+ close(fd);
+ if (rc != 0 || delete) {
+ int err = unlink(file);
+ if (err != 0 && errno != ENOENT) {
+ rc = rc ? rc : -errno;
+ log_write("%s: failed to unlink %s for orphan %u:"DOSTID
+ ": %s\n", progname, file, ost_idx,
+ POSTID(oi), strerror(-rc));
+ }
+ }
+out_free:
+ free(lum);
+ return rc;
+}
+
+/*
+ * Check for orphans
+ * Run through each entry in ost table and check the mds ost table for
+ * a corresponding entry. If not found report and repair.
+ */
+int lfsck_run_pass3(__u32 ost_idx, DB *mds_ostdb, DB *ostdb,
+ struct obd_uuid *uuid, __u64 last_id)
+{
+ int rc = 0;
+ struct lfsck_mds_objent mds_obj;
+ struct lfsck_ost_objent ost_obj;
+ unsigned long error = 0, count = 0;
+ DBT key, data;
+ DBC *dbcp = NULL;
+ __u64 bytes = 0;
+
+ log_write("lfsck: ost_idx %d: pass3: check for orphan objects\n",
+ ost_idx);
+
+ VERBOSE(1, "[%u] uuid %s\n", ost_idx, uuid->uuid);
+ VERBOSE(1, "[%u] last_id "LPU64"\n", ost_idx, last_id);
+
+ rc = ostdb->cursor(ostdb, NULL, &dbcp, 0);
+ if (rc != 0) {
+ log_write("[%u]: error acquiring cursor for mds table: %s\n",
+ ost_idx, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ data.data = &ost_obj;
+ data.size = data.ulen = sizeof(ost_obj);
+ data.flags = DB_DBT_USERMEM;
+
+ while ((rc = dbcp->c_get(dbcp, &key, &data, DB_NEXT)) == 0) {
+ DBT key_mdt, data_mdt;
+ struct ostid *oi;
+
+ count++;
+ letocpu_ost_objent(&ost_obj);
+ oi = &ost_obj.ost_oi;
+
+ if (oi->oi_id > last_id) {
+ VERBOSE(2, "[%u] skipping objid "DOSTID" > "LPU64"\n",
+ ost_idx, POSTID(oi), last_id);
+ continue;
+ }
+ VERBOSE(2, "[%u] processing objid "DOSTID"\n", ost_idx,
+ POSTID(oi));
+
+ memset(&key_mdt, 0, sizeof(key_mdt));
+ memset(&data_mdt, 0, sizeof(data_mdt));
+ key_mdt.data = oi;
+ key_mdt.size = sizeof(*oi);
+ data_mdt.data = &mds_obj;
+ data_mdt.size = data_mdt.ulen = sizeof(mds_obj);
+ data_mdt.flags = DB_DBT_USERMEM;
+ rc = mds_ostdb->get(mds_ostdb, NULL, &key_mdt, &data_mdt, 0);
+ if (rc == 0) {
+ VERBOSE(2, "[%u] found object "DOSTID" reference\n",
+ ost_idx, POSTID(oi));
+ continue;
+ }
+
+ letocpu_mds_objent(&mds_obj);
+ if (rc != DB_NOTFOUND) {
+ log_write("Failed to check mds db for entry\n");
+ rc = -EINVAL;
+ goto out;
+ }
+ if (ost_obj.ost_size == 0) {
+ /* don't report errors for normal orphan recovery */
+ VERBOSE(1, "[%u] zero-length orphan objid "DOSTID"\n",
+ ost_idx, POSTID(oi));
+ if (lfsck_save || lfsck_delete) {
+ /* No reason to save just delete*/
+ rc = lfsck_fix_orphan(ost_idx, oi, uuid, 1);
+ if (rc) {
+ log_write("lfsck: [%u]: pass3 "
+ "error fixing zero-length "
+ "orphan objid "DOSTID"\n",
+ ost_idx, POSTID(oi));
+ fix_failed++;
+ } else {
+ fixed++;
+ }
+ }
+ continue;
+ }
+
+ error++;
+ bytes += ost_obj.ost_bytes;
+ if (lfsck_save || lfsck_delete) {
+ rc = lfsck_fix_orphan(ost_idx, oi, uuid, lfsck_delete);
+ if (rc) {
+ log_write("lfsck: [%u]: failed to fix orphan "
+ "object "DOSTID", "LPU64" bytes\n",
+ ost_idx, POSTID(oi), ost_obj.ost_bytes);
+ fix_failed++;
+ } else {
+ log_write("lfsck: [%u]: pass3 %s orphan object "
+ DOSTID", "LPU64" bytes\n", ost_idx,
+ lfsck_save ? "saved" : "unlinked",
+ POSTID(oi), ost_obj.ost_bytes);
+ fixed++;
+ }
+ } else {
+ error++;
+ log_write("lfsck: [%u]: pass3 orphan found objid "
+ DOSTID", "LPU64" bytes\n", ost_idx,
+ POSTID(oi), ost_obj.ost_bytes);
+ }
+ }
+ if (rc != DB_NOTFOUND) {
+ log_write("[%u]: error getting next object in db %s\n",
+ ost_idx, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+
+ if (error == 0) {
+ log_write("lfsck: ost_idx %d: pass3 OK (%lu files total)\n",
+ ost_idx, count);
+ } else {
+ log_write("lfsck: ost_idx %d: pass3 %s: %4gMB of orphan "
+ "data (%lu of %lu files total)\n", ost_idx,
+ (lfsck_save | lfsck_delete) ? "FIXED" : "ERROR",
+ (double)bytes / (1024 * 1024), error, count);
+ }
+out:
+ if (dbcp)
+ dbcp->c_close(dbcp);
+ return 0;
+}
+
+/* Missing ost information report affected file names */
+int lfsck_list_affected_files(char *mds_file, struct lfsck_mds_hdr *mds_hdr,
+ DB *mds_direntdb, __u32 ost_idx)
+{
+ struct lfsck_mds_objent mds_obj;
+ char dbname[256];
+ char *path;
+ DB *mds_db = NULL;
+ DBT key, data;
+ DBC *dbcp = NULL;
+ int rc = 0;
+
+ path = malloc(PATH_MAX);
+ if (path == NULL)
+ return -ENOMEM;
+
+ sprintf(dbname, "%s.%d", MDS_OSTDB, ost_idx);
+ rc = lfsck_opendb(mds_file, dbname, &mds_db, 1, 0, 0);
+ if (rc != 0) {
+ log_write("%s: failed to open mds db file %s: rc %d\n",
+ progname, mds_file, rc);
+ rc = -EINVAL;
+ goto out;
+ }
+
+ if ((rc = mds_db->cursor(mds_db, NULL, &dbcp, 0)) != 0) {
+ log_write("Failed to acquire cursor for mds table\n");
+ rc = -EINVAL;
+ goto out;
+ }
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ data.data = &mds_obj;
+ data.size = data.ulen = sizeof(mds_obj);
+ data.flags = DB_DBT_USERMEM;
+
+ log_write("Files affected by missing ost info are : -\n");
+ while ((rc = dbcp->c_get(dbcp, &key, &data, DB_NEXT)) == 0) {
+ letocpu_mds_objent(&mds_obj);
+
+ if (lfsck_get_path(&mds_obj.mds_fid, mds_direntdb,
+ path, PATH_MAX)) {
+ log_write("Failed to get path for fid "DFID"\n",
+ PFID(&mds_obj.mds_fid));
+ fix_failed++;
+ } else {
+ log_write("%s\n", path);
+ }
+ }
+ if (rc != DB_NOTFOUND) {
+ log_write("Error getting next element in db %d\n", rc);
+ rc = -EINVAL;
+ goto out;
+ }
+ rc = 0;
+
+out:
+ if (dbcp)
+ dbcp->c_close(dbcp);
+ if (mds_db)
+ mds_db->close(mds_db, 0);
+ return rc;
+}
+
+/*
+ * For each ost index run checks 1 2 and 3.
+ * 1) Check for object referenced by more than one file
+ * 2) Check that objects exist on ost
+ * 3) Check that containg mds entry exists for an object
+ */
+int run_test(__u32 ost_idx, struct lfsck_mds_hdr *mds_hdr,
+ DB *mds_direntdb, DB *mds_sizeinfodb)
+{
+ struct lfsck_ost_hdr *ost_hdr = NULL;
+ char dbname[256];
+ DB *mds_ostdb = NULL;
+ DB *ost_db = NULL;
+ DBT key, data;
+ __u64 last_id;
+ int i, rc;
+
+ sprintf(dbname, "%s.%d", MDS_OSTDB, ost_idx);
+
+ VERBOSE(2, "testing ost_idx %d\n", ost_idx);
+
+ rc = lfsck_opendb(mds_file, dbname, &mds_ostdb, 1, 0, 0);
+ if (rc != 0) {
+ log_write("%s: failed to open mds db file %s: rc %d\n",
+ progname, mds_file, rc);
+ goto out;
+ }
+
+ ost_hdr = malloc(sizeof(*ost_hdr));
+ if (ost_hdr == NULL) {
+ log_write("Failure to alloc memory\n");
+ rc = -ENOMEM;
+ goto out;
+ }
+
+
+ VERBOSE(2, "looking for index %u UUID %s\n", ost_idx,
+ lfsck_uuid[ost_idx].uuid);
+
+ for (i = 0; i < num_ost_files; i++) {
+ VERBOSE(2, "checking file %s\n", ost_files[i]);
+ rc = lfsck_opendb(ost_files[i], OST_HDR, &ost_db, 0, 0, 0);
+ if (rc != 0) {
+ log_write("%s: error opening ost_data_file %s: rc %d\n",
+ progname, ost_files[i], rc);
+ goto out;
+ }
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ ost_hdr->ost_magic = OST_MAGIC;
+ key.data = &ost_hdr->ost_magic;
+ key.size = sizeof(ost_hdr->ost_magic);
+ data.size = data.ulen = sizeof(*ost_hdr);
+ data.data = ost_hdr;
+ data.flags = DB_DBT_USERMEM;
+
+ rc = ost_db->get(ost_db, NULL, &key, &data, 0);
+ ost_db->close(ost_db, 0);
+ ost_db = NULL;
+ if (rc != 0) {
+ log_write("Invalid ost magic on file %s: rc %s\n",
+ ost_files[i], db_strerror(rc));
+ continue;
+ }
+
+ letocpu_ost_hdr(ost_hdr);
+ VERBOSE(2, "%s has ost UUID %s\n", ost_files[i],
+ ost_hdr->ost_uuid.uuid);
+
+ if (obd_uuid_equals(&lfsck_uuid[ost_idx], &ost_hdr->ost_uuid)) {
+ if (ost_hdr->ost_index != ost_idx) {
+ log_write("Requested ost_idx %u doesn't match "
+ "index %u found in %s\n", ost_idx,
+ ost_hdr->ost_index, ost_files[i]);
+ continue;
+ }
+
+ break;
+ }
+ }
+
+ if (i == num_ost_files) {
+ log_write("lfsck: can't find file for ost_idx %d\n", ost_idx);
+ rc = lfsck_list_affected_files(mds_file, mds_hdr,
+ mds_direntdb, ost_idx);
+ goto out;
+ }
+ rc = lfsck_opendb(ost_files[i], OST_OSTDB, &ost_db, 0, 0, 0);
+ if (rc != 0) {
+ log_write("%s: error opening ost_data_file %s: rc %d\n",
+ progname, ost_files[i], rc);
+ goto out;
+ }
+
+ VERBOSE(1, "MDS: max_id "LPU64" OST: max_id "LPU64"\n",
+ mds_hdr->mds_max_ost_id[ost_idx], ost_hdr->ost_last_id);
+
+ rc = lfsck_run_pass1(ost_idx, mds_ostdb, ost_db, mds_direntdb);
+ if (rc != 0) {
+ log_write("error in running pass1\n");
+ goto out;
+ }
+
+ rc = lfsck_run_pass2(ost_idx, mds_hdr, mds_ostdb, ost_db, mds_direntdb,
+ mds_sizeinfodb);
+ if (rc != 0) {
+ log_write("error in running pass2\n");
+ goto out;
+ }
+
+ last_id = (ost_hdr->ost_flags & E2F_OPT_READONLY ||
+ mds_hdr->mds_flags & E2F_OPT_READONLY) ?
+ mds_hdr->mds_max_ost_id[ost_idx] : ost_hdr->ost_last_id;
+
+ rc = lfsck_run_pass3(ost_idx, mds_ostdb, ost_db, &ost_hdr->ost_uuid,
+ last_id);
+ if (rc != 0) {
+ log_write("error in running pass3\n");
+ goto out;
+ }
+ rc = 0;
+
+out:
+ if (ost_hdr)
+ free(ost_hdr);
+ if (mds_ostdb)
+ mds_ostdb->close(mds_ostdb, 0);
+ if (ost_db)
+ ost_db->close(ost_db, 0);
+
+ return rc;
+}
+
+static int lfsck_validate_duplicate(struct lfsck_saved_duplicates *dup,
+ const char *path)
+{
+ struct lov_user_md *lum;
+ struct lov_user_ost_data_v1 *loi;
+ struct stat64 st;
+ struct lu_fid fid;
+ int rc, i;
+
+ VERBOSE(2,"[%u] check duplicate FID "DFID" object "DOSTID"\n for\t%s\n",
+ dup->ld_ost_idx, PFID(&dup->ld_mds_fid), POSTID(&dup->ld_oi),
+ path);
+
+ /* first, validate that the paths are still valid */
+ if (stat64(path, &st) < 0) {
+ rc = -errno;
+ log_write("%s: duplicate file %s error: %s\n",
+ progname, path, strerror(-rc));
+ return rc;
+ }
+
+ rc = lfsck_path2fid(path, &fid);
+ if (rc < 0) {
+ log_write("%s: unable to get LMA EA on %s: %s\n",
+ progname, path, strerror(-rc));
+ return rc;
+ }
+ if (lfsck_fidcmp(&dup->ld_mds_fid, &fid)) {
+ log_write("%s: duplicate file %s is no longer FID "DFID"\n",
+ progname, path, PFID(&dup->ld_mds_fid));
+ return -EBADF;
+ }
+
+ lum = malloc(LOV_EA_MAX(lum));
+ if (lum == NULL) {
+ log_write("%s: out of memory allocating LOV EA (%d)\n",
+ progname, (int)LOV_EA_MAX(lum));
+ return -ENOMEM;
+ }
+
+ rc = llapi_file_get_stripe(path, lum);
+ if (rc < 0) {
+ log_write("%s: unable to get LOV EA on %s: %s\n",
+ progname, path, strerror(-rc));
+ goto out;
+ }
+
+ if (lum->lmm_pattern != LOV_PATTERN_RAID0) {
+ log_write("%s: unknown LOV stripe pattern %#08x\n",
+ progname, lum->lmm_pattern);
+ rc = -EINVAL;
+ goto out;
+ }
+
+ if (lum->lmm_magic == LOV_USER_MAGIC_V1) {
+ loi = lum->lmm_objects;
+ } else if (lum->lmm_magic == LOV_USER_MAGIC_V3) {
+ loi = ((struct lov_user_md_v3 *)lum)->lmm_objects;
+ } else {
+ log_write("%s: unknown LOV magic %#08x\n",
+ progname, lum->lmm_magic);
+ rc = -EINVAL;
+ goto out;
+ }
+
+ /* Verify that the object in question is still in the file */
+ for (i = 0; i < lum->lmm_stripe_count; i++, loi++) {
+ if (loi->l_ost_idx == dup->ld_ost_idx &&
+ ostid_id(&loi->l_ost_oi) == ostid_id(&dup->ld_oi) &&
+ ostid_seq(&loi->l_ost_oi) == ostid_seq(&dup->ld_oi))
+ break;
+ }
+
+ if (i == lum->lmm_stripe_count) {
+ log_write("%s: cannot find object %u:"DOSTID" in\n\t%s\n",
+ progname, dup->ld_ost_idx, POSTID(&dup->ld_oi), path);
+ rc = -EBADF;
+ goto out;
+ }
+
+out:
+ free(lum);
+ return rc;
+}
+
+#ifndef HAVE_LLAPI_CANCEL_OSC_LOCKS
+#define NAMESPACES "/proc/fs/lustre/ldlm/namespaces"
+void llapi_cancel_osc_locks(const char *mnt_path)
+{
+ DIR *namespaces;
+ char path[PATH_MAX];
+ struct dirent *dent;
+ int rc, fd;
+
+ namespaces = opendir(NAMESPACES);
+ if (namespaces == NULL) {
+ rc = -errno;
+ log_write("%s: error opening %s: %s\n",
+ progname, NAMESPACES, strerror(-rc));
+ return;
+ }
+
+ while ((dent = readdir(namespaces)) != NULL) {
+ if (strcmp(dent->d_name, ".") == 0 ||
+ strcmp(dent->d_name, "..") == 0)
+ continue;
+
+ if (strstr(dent->d_name, "osc") == NULL &&
+ strstr(dent->d_name, "OSC") == NULL)
+ continue;
+
+ snprintf(path, sizeof(path) - 1, "%s/%s/lru_size",
+ NAMESPACES, dent->d_name);
+
+ fd = open(path, O_WRONLY);
+ if (fd < 0) {
+ log_write("%s: error opening %s to cancel locks: %s\n",
+ progname, path, strerror(errno));
+ continue;
+ }
+ VERBOSE(3, "clearing locks in %s\n", path);
+ rc = write(fd, "clear", 6);
+ close(fd);
+ }
+
+ closedir(namespaces);
+}
+#endif
+
+/* Remove inodes from the client cache to avoid hitting an LASSERTF() on
+ * the client if it tries to attach two inodes to the same object */
+static void lfsck_drop_caches(void)
+{
+ int fd, rc;
+
+ sync();
+ llapi_cancel_osc_locks(mnt_path);
+ fd = open("/proc/sys/vm/drop_caches", O_WRONLY);
+ if (fd < 0)
+ return;
+ VERBOSE(3, "flushing vm cache\n");
+ rc = write(fd, "3", 2);
+ close(fd);
+}
+
+/* Duplicate an object that is referenced by multiple files and point one
+ * of the files to use the duplicated object */
+static int lfsck_fix_duplicate(struct lfsck_saved_duplicates *dup,
+ DB *mds_direntdb, const char *path)
+{
+ char path_tmp[PATH_MAX] = { 0 };
+ char tmp[PATH_MAX * 2 + 10] = { 0 };
+ const char *base;
+ int rc;
+
+ lfsck_drop_caches();
+
+ if (!lfsck_create) {
+ VERBOSE(1, "%s: [%u]: not duplicating FID "DFID
+ " object "DOSTID" by request\n on\t%s\n",
+ progname, dup->ld_ost_idx, PFID(&dup->ld_mds_fid),
+ POSTID(&dup->ld_oi), path);
+ return 0;
+ }
+
+ rc = lfsck_validate_duplicate(dup, path);
+ if (rc < 0)
+ goto out;
+
+ snprintf(path_tmp, sizeof(path_tmp) - 1, "%s.lfsck_tmp", path);
+ snprintf(tmp, sizeof(tmp) - 1, "cp -p '%s' '%s'", path, path_tmp);
+ VERBOSE(2, "%s\n", tmp);
+ rc = system(tmp);
+ if (rc) {
+ rc = -errno;
+ log_write("%s: duplicating object for %u:"DOSTID" %s: %s\n",
+ progname, dup->ld_ost_idx, POSTID(&dup->ld_oi),
+ path, strerror(-rc));
+ goto out;
+ }
+
+ base = strrchr(path, '/');
+ if (base == NULL)
+ base = path;
+ else
+ base++;
+
+ rc = asprintf(&dup->ld_link, "%s/%u-"DOSTID"-"DFID":%s", dupedir,
+ dup->ld_ost_idx, POSTID(&dup->ld_oi),
+ PFID(&dup->ld_mds_fid), base);
+ if (rc < 0) {
+ rc = -errno;
+ goto out;
+ }
+
+ VERBOSE(2, "ln %s %s\n", path, dup->ld_link);
+ rc = link(path, dup->ld_link);
+ if (rc) {
+ rc = -errno;
+ log_write("%s: error linking %s to %s: %s\n",
+ progname, path, dup->ld_link, strerror(-rc));
+ free(dup->ld_link);
+ dup->ld_link = NULL;
+ goto out;
+ }
+
+ VERBOSE(2, "rename %s %s\n", path_tmp, path);
+ rc = rename(path_tmp, path);
+ if (rc) {
+ rc = -errno;
+ log_write("%s: error renaming %s to %s: %s\n",
+ progname, path_tmp, path, strerror(-rc));
+ free(dup->ld_link);
+ dup->ld_link = NULL;
+ } else {
+ log_write("%s: [%u]: fixed duplicate FID "DFID" object "
+ DOSTID":\n\t%s\n", progname, dup->ld_ost_idx,
+ PFID(&dup->ld_mds_fid), POSTID(&dup->ld_oi), path);
+ }
+out:
+ if (rc) {
+ VERBOSE(2, "unlink %s\n", path_tmp);
+ if (unlink(path_tmp))
+ log_write("%s: unlink %s failed: %s\n", progname,
+ path_tmp, strerror(errno));
+ }
+
+ return rc;
+}
+
+/*
+ * Check for files found that reference the same ost objects
+ * (found in pass1) and repair now if necessary
+ */
+int lfsck_run_pass4(DB *mds_direntdb)
+{
+ char tmp[PATH_MAX + 512];
+ char path[PATH_MAX];
+ int failed_get_path, fixed_dup;
+ int i, j;
+
+ log_write("lfsck: pass4: check for %u duplicate object references\n",
+ lfsck_dup_saved);
+ if (lfsck_dup_saved == 0) {
+ log_write("lfsck: pass4 OK (no duplicates)\n");
+ return 0;
+ }
+
+ do {
+ struct lu_fid *mds_fid;
+
+ failed_get_path = 0;
+ fixed_dup = 0;
+
+ lfsck_drop_caches();
+
+ for (i = 0; i < lfsck_dup_saved; i++) {
+ mds_fid = &lfsck_duplicates[i].ld_mds_fid;
+ if (mds_fid->f_oid == 0)
+ continue;
+
+ if (lfsck_get_path(mds_fid, mds_direntdb, path,
+ sizeof(path))) {
+ failed_get_path++;
+ continue;
+ }
+ if (lfsck_fix_duplicate(&lfsck_duplicates[i],
+ mds_direntdb, path)) {
+ fix_failed++;
+ continue;
+ }
+
+ fixed++;
+ fixed_dup++;
+
+ /* don't duplicate a file multiple times even if it has
+ * multiple shared objects */
+ for (j = i + 1; j < lfsck_dup_saved; j++) {
+ if (lfsck_fidcmp(&lfsck_duplicates[j].ld_mds_fid,
+ mds_fid) == 0)
+ lfsck_duplicates[j].ld_mds_fid.f_oid =0;
+ }
+ lfsck_duplicates[i].ld_mds_fid.f_oid = 0;
+ }
+ } while (failed_get_path && fixed_dup);
+
+ for (i = 0; i < lfsck_dup_saved; i++) {
+ lfsck_drop_caches();
+
+ if (!lfsck_duplicates[i].ld_link)
+ continue;
+
+ if (unlink(lfsck_duplicates[i].ld_link))
+ log_write("%s: failed to unlink %s: %s\n", progname,
+ lfsck_duplicates[i].ld_link, strerror(errno));
+ else
+ log_write("%s: %s unlinked\n", progname,
+ lfsck_duplicates[i].ld_link);
+ free(lfsck_duplicates[i].ld_link);
+ lfsck_duplicates[i].ld_link = NULL;
+ }
+
+ snprintf(tmp, sizeof(tmp) - 1, "rm -rvf '%s'", dupedir);
+ VERBOSE(1, "%s\n", tmp);
+ if (system(tmp) == -1)
+ VERBOSE(1, "%s failed", tmp);
+
+ log_write("lfsck: pass4 finished\n");
+
+ return 0;
+}
+
+/*
+ * This is a placeholder to check for filesize correctness no fixup is in
+ * place right now since file size is still obtained from osts
+ */
+int lfsck_run_pass5(DB *mds_direntdb, DB *mds_sizeinfodb)
+{
+ int rc = 0;
+#ifdef LFSCK_CHECK_SIZE
+ struct lfsck_mds_szinfo mds_szinfo1;
+ char path[PATH_MAX];
+ struct stat64 statbuf;
+ DBT key,data;
+ DBC *dbcp;
+
+ log_write("lfsck: pass5: file size correctness\n");
+
+ rc = mds_sizeinfodb->cursor(mds_sizeinfodb, NULL, &dbcp, 0);
+ if (rc != 0) {
+ log_write("%s: error acquiring cursor for database: %s\n",
+ progname, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ data.data = &mds_szinfo1;
+ data.size = data.ulen = sizeof(mds_szinfo1);
+ data.flags = DB_DBT_USERMEM;
+ while ((rc = dbcp->c_get(dbcp, &key, &data, DB_NEXT)) == 0) {
+ letocpu_mds_szinfo(&mds_szinfo1);
+
+ if (mds_szinfo1.mds_size != mds_szinfo1.mds_calc_size) {
+ if (lfsck_get_path(mds_szinfo1.mds_fid, mds_direntdb,
+ path, sizeof(path))) {
+ log_write("%s: failed to get path and update "
+ "size for fid "LPU64"\n",
+ progname, mds_szinfo1.mds_fid);
+ fix_failed++;
+ continue;
+ }
+
+ if (stat64(path, &statbuf)) {
+ log_write("%s: pass5: failed to stat %s\n",
+ progname, path);
+ fix_failed++;
+ continue;
+ }
+ if (statbuf.st_size == mds_szinfo1.mds_calc_size) {
+ VERBOSE(2, "%s: %s really has right size\n",
+ progname, path);
+ } else {
+ log_write("%s: %s size "LPU64" != "LPU64"\n",
+ progname, path, statbuf.st_size,
+ mds_szinfo1.mds_calc_size);
+ fixed++;
+ }
+ }
+ }
+ if (rc != DB_NOTFOUND) {
+ log_write("%s: error getting next element in db: %s\n",
+ progname, db_strerror(rc));
+ rc = -EINVAL;
+ goto out;
+ }
+ rc = 0;
+ log_write("%s: pass5 finished\n", progname);
+out:
+ dbcp->c_close(dbcp);
+#endif /* LFSCK_CHECK_SIZE */
+ return rc;
+}
+
+int get_response()
+{
+ char yes[] = "Yy";
+ char no[] = "Nn";
+ char c;
+ int rc = -1;
+
+ while (1) {
+ c = getchar();
+ if (c == EOF)
+ break;
+
+ if (strchr(yes, c)) {
+ rc = 1;
+ break;
+ }
+
+ if (strchr(no, c)) {
+ rc = 0;
+ break;
+ }
+ }
+ return rc;
+}
+
+/* Starting point for each thread */
+void *lfsck_start_thread(void *arg)
+{
+ struct lfsck_thread_info *tinfo = (struct lfsck_thread_info *)arg;
+ int i, rc;
+
+ tinfo->status = 0;
+ pthread_mutex_lock(&init_mutex);
+ if (all_started)
+ pthread_mutex_unlock(&init_mutex);
+ else
+ pthread_cond_wait(&init_cond, &init_mutex);
+
+ if (!all_started)
+ pthread_exit(NULL);
+ for (i = tinfo->start_ost_idx; i < tinfo->end_ost_idx; i++) {
+ rc = run_test(i, tinfo->mds_hdr, tinfo->mds_direntdb,
+ tinfo->mds_sizeinfodb);
+ if (rc) {
+ log_write("lfsck: ost_idx %d: error running check\n",i);
+ tinfo->status = rc;
+ }
+ }
+ pthread_exit(NULL);
+}
+
+/* Start threads and run filesystem checks and repair */
+int lfsck_run_checks()
+{
+ struct lfsck_mds_hdr *mds_hdr = NULL;
+ struct lfsck_thread_info *tinfo = NULL;
+ pthread_t *threads = NULL;
+ int rc, i;
+ DB *mds_direntdb = NULL;
+ DB *mds_hdrdb = NULL;
+ DB *mds_sizeinfodb = NULL;
+ DBT key, data;
+ int num_osts;
+
+ rc = lfsck_opendb(mds_file, MDS_HDR, &mds_hdrdb, 0, 0, 0);
+ if (rc != 0) {
+ log_write("%s: error opening mds_hdr in %s: rc %d\n",
+ progname, mds_file, rc);
+ return -EINVAL;
+ }
+ mds_hdr = malloc(sizeof(*mds_hdr));
+ if (mds_hdr == NULL) {
+ log_write("%s: out of memory allocating DB header (%d)\n",
+ progname, (int)sizeof(*mds_hdr));
+ rc = -ENOMEM;
+ goto out;
+ }
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ mds_hdr->mds_magic = MDS_MAGIC;
+ key.data = &mds_hdr->mds_magic;
+ key.size = sizeof(mds_hdr->mds_magic);
+ data.data = mds_hdr;
+ data.size = sizeof(*mds_hdr);
+ data.ulen = sizeof(*mds_hdr);
+ data.flags = DB_DBT_USERMEM;
+ rc = mds_hdrdb->get(mds_hdrdb, NULL, &key, &data, 0);
+ if (rc != 0) {
+ log_write("%s: error getting mds_hdr info %s: %s\n",
+ progname, mds_file, db_strerror(rc));
+ goto out;
+ }
+ letocpu_mds_hdr(mds_hdr);
+
+ rc = lfsck_opendb(mds_file, MDS_DIRINFO, &mds_direntdb, 0, 0, 0);
+ if (rc != 0) {
+ log_write("%s: error opening dirinfo db %s: rc %d\n",
+ progname, mds_file, rc);
+ goto out;
+ }
+
+ rc = lfsck_opendb(mds_file, MDS_SIZEINFO, &mds_sizeinfodb, 0, 0, 0);
+ if (rc != 0) {
+ log_write("%s: error opening sizeinfo db %s: rc %d\n",
+ progname, mds_file, rc);
+ goto out;
+ }
+
+ if (lov_tgt_count > mds_hdr->mds_num_osts) {
+ fprintf(stderr, "%s: number of osts in lov (%u) > "
+ "num referenced in mds (%u) (new ost or "
+ "empty filesystem?)\n", progname,
+ lov_tgt_count, mds_hdr->mds_num_osts);
+ fprintf(stderr, "Do you wish to continue? (y/n)\n");
+ rc = get_response();
+ if (rc != 1) {
+ log_write("%s: exiting \n", progname);
+ goto out;
+ }
+ fprintf(stderr, "\n");
+
+ num_osts = lov_tgt_count;
+ } else {
+ num_osts = mds_hdr->mds_num_osts;
+ }
+ if (num_threads > num_osts)
+ num_threads = num_osts;
+
+ tinfo = calloc(num_threads, sizeof(*tinfo));
+ if (tinfo == NULL) {
+ log_write("%s: out of memory for thread info\n", progname);
+ rc = -ENOMEM;
+ goto out;
+ }
+ threads = calloc(num_threads, sizeof(pthread_t));
+ if (threads == NULL) {
+ log_write("%s: out of memory for threads\n", progname);
+ rc = -ENOMEM;
+ goto out;
+ }
+
+ all_started = 0;
+ for (i = 0; i < num_threads; i++) {
+ __u32 end_ost_idx;
+ __u32 chunk;
+
+ chunk = num_osts / num_threads;
+ if (num_osts % num_threads)
+ chunk++;
+ tinfo[i].mds_hdr = mds_hdr;
+ tinfo[i].mds_direntdb = mds_direntdb;
+ tinfo[i].mds_sizeinfodb = mds_sizeinfodb;
+ tinfo[i].status = 0;
+ tinfo[i].start_ost_idx = (chunk) * i;
+ end_ost_idx = (chunk) * (i + 1);
+ end_ost_idx = end_ost_idx > num_osts ?
+ num_osts : end_ost_idx;
+ tinfo[i].end_ost_idx = end_ost_idx;
+ rc = pthread_create(&threads[i], NULL, lfsck_start_thread,
+ &tinfo[i]);
+ if (rc) {
+ log_write("%s: error starting thread waiting for other"
+ " threads to exit\n", progname);
+ pthread_mutex_lock(&init_mutex);
+ pthread_cond_broadcast(&init_cond);
+ pthread_mutex_unlock(&init_mutex);
+ for (--i; i >= 0; i--)
+ pthread_cancel(threads[i]);
+
+ rc = -ENOMEM;
+ goto out;
+ }
+ }
+ pthread_mutex_lock(&init_mutex);
+ all_started = 1;
+ pthread_cond_broadcast(&init_cond);
+ pthread_mutex_unlock(&init_mutex);
+ for (i = 0; i < num_threads; i++) {
+ rc = pthread_join(threads[i], NULL);
+ if (tinfo[i].status) {
+ log_write("%s: error running thread %u\n", progname, i);
+ rc = -EINVAL;
+ }
+ }
+
+ rc = lfsck_run_pass4(mds_direntdb);
+ if (rc != 0)
+ goto out;
+
+ rc = lfsck_run_pass5(mds_direntdb, mds_sizeinfodb);
+
+out:
+ if (threads)
+ free(threads);
+ if (tinfo)
+ free(tinfo);
+ if (mds_hdr)
+ free(mds_hdr);
+ if (mds_direntdb)
+ mds_direntdb->close(mds_direntdb, 0);
+ if (mds_hdrdb)
+ mds_hdrdb->close(mds_hdrdb, 0);
+ if (mds_sizeinfodb)
+ mds_sizeinfodb->close(mds_sizeinfodb, 0);
+
+ return rc;
+}
+
+int create_lostandfound()
+{
+ struct stat statbuf;
+
+ snprintf(lostandfounddir, PATH_MAX - 1, "%s/lost+found", mnt_path);
+ lostandfounddir[PATH_MAX - 1] = '\0';
+
+ VERBOSE(2, "%s: creating %s\n", progname, lostandfounddir);
+ if (mkdir(lostandfounddir, 0700)) {
+ if (errno != EEXIST) {
+ fprintf(stderr, "%s: error creating %s: %s\n",
+ progname, lostandfounddir, strerror(errno));
+ return -errno;
+ }
+
+ if (stat(lostandfounddir, &statbuf)) {
+ fprintf(stderr, "%s: error stat %s: %s\n",
+ progname, lostandfounddir, strerror(errno));
+ return -errno;
+ }
+
+ if (!S_ISDIR(statbuf.st_mode)) {
+ fprintf(stderr, "%s: error %s is not a directory\n",
+ progname, lostandfounddir);
+ return -EINVAL;
+ }
+ }
+
+ snprintf(dupedir, sizeof(dupedir), "%s/duplicates", lostandfounddir);
+ dupedir[PATH_MAX - 1] = '\0';
+
+ VERBOSE(2, "%s: creating %s\n", progname, dupedir);
+ if (mkdir(dupedir, 0700)) {
+ if (errno != EEXIST) {
+ fprintf(stderr, "%s: error creating %s: %s\n",
+ progname, dupedir, strerror(errno));
+ return -errno;
+ }
+
+ if (stat(lostandfounddir, &statbuf)) {
+ fprintf(stderr, "%s: error stat %s: %s\n",
+ progname, dupedir, strerror(errno));
+ return -errno;
+ }
+
+ if (!S_ISDIR(statbuf.st_mode)) {
+ fprintf(stderr, "%s: error %s is not a directory\n",
+ progname, dupedir);
+ return -EINVAL;
+ }
+ }
+
+ return 0;
+}
+
+int main(int argc, char *argv[])
+{
+ uid_t myuid;
+ int i;
+
+ fprintf(stderr, "lfsck %s (%s)\n", E2FSPROGS_VERSION, E2FSPROGS_DATE);
+
+ if (parse_args(argc, argv)) {
+ usage();
+ exit(16);
+ }
+
+ myuid = getuid();
+ if (myuid != 0 && !lfsck_force) {
+ fprintf(stderr, "%s: can only be run by root user\n", progname);
+ exit(16);
+ }
+
+ log_open();
+
+ if ((lfsck_save || lfsck_delete) && create_lostandfound() != 0) {
+ log_write("%s: failed to create lost+found directory\n",
+ progname);
+ log_close(-1);
+ exit(8);
+ }
+
+ if (get_lov_config()) {
+ log_close(-1);
+ exit(8);
+ }
+
+ if (lfsck_run_checks())
+ log_close(-1);
+
+ if (mds_file)
+ free(mds_file);
+ for (i = 0; i < LOV_MAX_OSTS; i++) {
+ if (ost_files[i])
+ free(ost_files[i]);
+ }
+ if (lfsck_duplicates)
+ free(lfsck_duplicates);
+
+ log_close(0);
+ if (fix_failed) {
+ fprintf(stderr, "%s: exit with %u unfixed errors\n",
+ progname, fix_failed);
+ return 2;
+ } else {
+ printf("%s: fixed %u errors\n", progname, fixed);
+ return !!fixed;
+ }
+}
--- /dev/null
+/*
+ * Copyright (c) 2004 Hewlett-Packard Co.
+ * Copyright (c) 2011 Whamcloud, Inc
+ */
+/*****************************************************************************
+ * e2fsck extentions: code for gathering data from the OST & MDT filesystems
+ * when e2fsck is run against them. The best description and knowledge of
+ * the layout and information gathered is in lfsck.h where the structures
+ * defining each entry in the tables are declared. Basically the ost file
+ * contains one table with each entry holding the object id and size.
+ * In addition there is header information at the start of the file.
+ * The MDT file contains multiple tables, one per OST. Each MDT/OST table
+ * contains an entry describing the MDT FID and the OST object associated
+ * with this FID on an OST. In addition the MDT also contains a table
+ * with the mds_fid and the FID of the containg directory. Header information
+ * for each table is also included.
+ * lfsck is run afterwards where the data gathered and stored here is cross
+ * checked to ensure consistency and correctness
+ *
+ *****************************************************************************/
+#include "config.h"
+#include <string.h>
+#include <time.h>
+#include <errno.h>
+#include <limits.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include "ext2fs/lfsck.h"
+
+#ifdef HAVE_LFSCK
+char *my_dirname(char *path)
+{
+ if (path != NULL) {
+ char *tmp = strrchr(path, '/');
+ if (tmp != NULL) {
+ *tmp = '\0';
+ return path;
+ }
+ }
+
+ return ".";
+}
+
+const char *my_basename(const char *path)
+{
+ if (path != NULL) {
+ char *tmp = strrchr(path, '/');
+ if (tmp != NULL)
+ return tmp + 1;
+ }
+
+ return path;
+}
+
+DB_ENV *dbenv;
+
+#define HASH_SIZE 131072
+
+u_int32_t lfsck_hash_raw_fn(const void *p)
+{
+ u_int32_t *c = (u_int32_t *)p;
+ u_int32_t rc = 0;
+
+ rc = (*c >> 7) & (HASH_SIZE - 1) ;
+
+ return rc;
+}
+
+
+u_int32_t lfsck_hash_fn(DB *dbp, const void *p, u_int32_t len)
+{
+ u_int32_t rc = 0 ;
+
+ if (len < sizeof(u_int32_t)) {
+ printf("Hash size error");
+ exit(128);
+ }
+ rc = lfsck_hash_raw_fn(p);
+
+ return rc;
+}
+
+int lfsck_create_dbenv(const char *progname)
+{
+ int rc;
+ size_t pagesize;
+ long pages;
+ unsigned long cachesize;
+
+ pagesize = getpagesize();
+ pages = sysconf(_SC_AVPHYS_PAGES);
+
+ cachesize = ((pagesize * 3) / 4) * pages;
+ if (cachesize > 500UL * 1024 * 1024)
+ cachesize = 500UL * 1024 * 1024;
+ else if (cachesize < 10 * 1024 * 1024)
+ cachesize = 10 * 1024 * 1024;
+
+ rc = db_env_create(&dbenv, 0);
+ if (rc != 0) {
+ fprintf(stderr, "%s: error creating dbenv: %s\n",
+ progname, db_strerror(rc));
+ return -EINVAL;
+ }
+
+ rc = dbenv->set_cachesize(dbenv, 0, cachesize, 0);
+ if (rc != 0) {
+ dbenv->err(dbenv, rc, "set_cachesize");
+ dbenv->close(dbenv, 0);
+ return -EINVAL;
+ }
+ rc = dbenv->set_data_dir(dbenv, "/");
+ if (rc != 0) {
+ dbenv->err(dbenv, rc, "set_data_dir");
+ dbenv->close(dbenv, 0);
+ return -EINVAL;
+ }
+
+ /* Open the environment with full transactional support. */
+ rc = dbenv->open(dbenv, "/tmp", DB_CREATE | DB_PRIVATE |
+ DB_INIT_MPOOL | DB_INIT_LOCK | DB_THREAD, 0);
+ if (rc != 0) {
+ dbenv->err(dbenv, rc, "environment open: ");
+ dbenv->close(dbenv, 0);
+ return -EINVAL;
+ }
+ return 0;
+}
+
+int lfsck_opendb(const char *fname, const char *dbname, DB **dbpp,
+ int allow_dup, int keydata_size, int num_files)
+{
+ static int dbenv_set;
+ DB *dbp;
+ int rc;
+ int pagesize = 512;
+ int h_ffactor = 0;
+
+ if (!dbenv_set) {
+ if (lfsck_create_dbenv(dbname))
+ return -EIO;
+ dbenv_set = 1;
+ }
+
+ rc = db_create(&dbp, dbenv, 0);
+ if (rc) {
+ fprintf(stderr, "%s: error db_create: %s\n",
+ dbname, db_strerror(rc));
+ return EIO;
+ }
+
+ rc = dbp->set_pagesize(dbp, pagesize);
+ if (rc != 0) {
+ dbp->err(dbp, rc, "set_pagesize");
+ dbp->close(dbp, 0);
+ return EIO;
+ }
+
+ rc = dbp->set_lorder(dbp, 1234);
+ if (rc != 0) {
+ dbp->err(dbp, rc, "set_lorder");
+ dbp->close(dbp, 0);
+ return EIO;
+ }
+
+ if (keydata_size && num_files) {
+ h_ffactor = (pagesize - 32) / (keydata_size + 8);
+ rc = dbp->set_h_ffactor(dbp, h_ffactor);
+ if (rc != 0)
+ dbp->err(dbp, rc, "set_h_ffactor");
+
+ rc = dbp->set_h_nelem(dbp, num_files);
+ if (rc != 0)
+ dbp->err(dbp, rc, "set_h_nelem");
+ }
+
+ rc = dbp->set_h_hash(dbp, lfsck_hash_fn);
+ if (rc != 0) {
+ dbp->err(dbp, rc, "set_h_hash");
+ dbp->close(dbp, 0);
+ return EIO;
+ }
+
+ if (allow_dup) {
+ rc = dbp->set_flags(dbp, DB_DUPSORT);
+ if (rc != 0) {
+ fprintf(stderr, "Failure to allow duplicates\n");
+ dbp->close(dbp, 0);
+ return EIO;
+ }
+ }
+
+ rc = dbp->open(dbp, NULL, fname, dbname, DB_HASH,
+ DB_CREATE | DB_THREAD, 0664);
+ if (rc != 0) {
+ dbp->err(dbp, rc, "%s:%s\n", fname, dbname);
+ dbp->close(dbp, 0);
+ return EIO;
+ }
+ *dbpp = dbp;
+ return 0;
+}
+
+void cputole_mds_hdr(struct lfsck_mds_hdr *mds_hdr)
+{
+ int i, num_osts = mds_hdr->mds_num_osts;
+
+ mds_hdr->mds_magic = ext2fs_cpu_to_le64(mds_hdr->mds_magic);
+ mds_hdr->mds_flags = ext2fs_cpu_to_le64(mds_hdr->mds_flags);
+ mds_hdr->mds_max_files = ext2fs_cpu_to_le64(mds_hdr->mds_max_files);
+ mds_hdr->mds_num_osts = ext2fs_cpu_to_le64(mds_hdr->mds_num_osts);
+
+ for (i = 0; i < num_osts; i++)
+ mds_hdr->mds_max_ost_id[i] =
+ ext2fs_cpu_to_le64(mds_hdr->mds_max_ost_id[i]);
+}
+
+void letocpu_mds_hdr(struct lfsck_mds_hdr *mds_hdr)
+{
+ int i;
+
+ mds_hdr->mds_magic = ext2fs_le64_to_cpu(mds_hdr->mds_magic);
+ mds_hdr->mds_flags = ext2fs_le64_to_cpu(mds_hdr->mds_flags);
+ mds_hdr->mds_max_files = ext2fs_le64_to_cpu(mds_hdr->mds_max_files);
+ mds_hdr->mds_num_osts = ext2fs_le64_to_cpu(mds_hdr->mds_num_osts);
+
+ for (i = 0; i < mds_hdr->mds_num_osts; i++)
+ mds_hdr->mds_max_ost_id[i] =
+ ext2fs_le64_to_cpu(mds_hdr->mds_max_ost_id[i]);
+}
+
+void cputole_ost_hdr(struct lfsck_ost_hdr *ost_hdr)
+{
+ ost_hdr->ost_magic = ext2fs_cpu_to_le64(ost_hdr->ost_magic);
+ ost_hdr->ost_flags = ext2fs_cpu_to_le64(ost_hdr->ost_flags);
+ ost_hdr->ost_num_files = ext2fs_cpu_to_le64(ost_hdr->ost_num_files);
+ ost_hdr->ost_last_id = ext2fs_cpu_to_le64(ost_hdr->ost_last_id);
+}
+
+void letocpu_ost_hdr(struct lfsck_ost_hdr *ost_hdr)
+{
+ ost_hdr->ost_magic = ext2fs_le64_to_cpu(ost_hdr->ost_magic);
+ ost_hdr->ost_flags = ext2fs_le64_to_cpu(ost_hdr->ost_flags);
+ ost_hdr->ost_num_files = ext2fs_le64_to_cpu(ost_hdr->ost_num_files);
+ ost_hdr->ost_last_id = ext2fs_le64_to_cpu(ost_hdr->ost_last_id);
+}
+
+void cputole_fid(struct lu_fid *fid)
+{
+ fid->f_seq = ext2fs_cpu_to_le64(fid->f_seq);
+ fid->f_oid = ext2fs_cpu_to_le32(fid->f_oid);
+ fid->f_ver = ext2fs_cpu_to_le32(fid->f_ver);
+}
+
+void letocpu_fid(struct lu_fid *fid)
+{
+ fid->f_seq = ext2fs_le64_to_cpu(fid->f_seq);
+ fid->f_oid = ext2fs_le32_to_cpu(fid->f_oid);
+ fid->f_ver = ext2fs_le32_to_cpu(fid->f_ver);
+}
+
+void cputole_oi(void *oip)
+{
+ struct ostid *oi = oip;
+
+ oi->oi_id = ext2fs_cpu_to_le64(oi->oi_id);
+ oi->oi_seq = ext2fs_cpu_to_le64(oi->oi_seq);
+}
+
+void letocpu_oi(void *oip)
+{
+ struct ostid *oi = oip;
+
+ oi->oi_id = ext2fs_le64_to_cpu(oi->oi_id);
+ oi->oi_seq = ext2fs_le64_to_cpu(oi->oi_seq);
+}
+
+void cputole_mds_dirent(struct lfsck_mds_dirent *mds_dirent)
+{
+ cputole_fid(&mds_dirent->mds_dirfid);
+ cputole_fid(&mds_dirent->mds_fid);
+}
+
+void letocpu_mds_dirent(struct lfsck_mds_dirent *mds_dirent)
+{
+ letocpu_fid(&mds_dirent->mds_dirfid);
+ letocpu_fid(&mds_dirent->mds_fid);
+}
+
+void cputole_mds_szinfo(struct lfsck_mds_szinfo *mds_szinfo)
+{
+ mds_szinfo->mds_fid = ext2fs_cpu_to_le64(mds_szinfo->mds_fid);
+ mds_szinfo->mds_seq = ext2fs_cpu_to_le64(mds_szinfo->mds_seq);
+ mds_szinfo->mds_size = ext2fs_cpu_to_le64(mds_szinfo->mds_size);
+ mds_szinfo->mds_calc_size =
+ ext2fs_cpu_to_le64(mds_szinfo->mds_calc_size);
+ mds_szinfo->mds_stripe_size =
+ ext2fs_cpu_to_le32(mds_szinfo->mds_stripe_size);
+ mds_szinfo->mds_stripe_pattern =
+ ext2fs_cpu_to_le32(mds_szinfo->mds_stripe_pattern);
+ mds_szinfo->mds_stripe_count =
+ ext2fs_cpu_to_le16(mds_szinfo->mds_stripe_count);
+ mds_szinfo->mds_stripe_start =
+ ext2fs_cpu_to_le16(mds_szinfo->mds_stripe_start);
+}
+
+void letocpu_mds_szinfo(struct lfsck_mds_szinfo *mds_szinfo)
+{
+ mds_szinfo->mds_fid = ext2fs_le64_to_cpu(mds_szinfo->mds_fid);
+ mds_szinfo->mds_seq = ext2fs_le64_to_cpu(mds_szinfo->mds_seq);
+ mds_szinfo->mds_size = ext2fs_le64_to_cpu(mds_szinfo->mds_size);
+ mds_szinfo->mds_calc_size =
+ ext2fs_le64_to_cpu(mds_szinfo->mds_calc_size);
+ mds_szinfo->mds_stripe_size =
+ ext2fs_le32_to_cpu(mds_szinfo->mds_stripe_size);
+ mds_szinfo->mds_stripe_pattern =
+ ext2fs_le32_to_cpu(mds_szinfo->mds_stripe_pattern);
+ mds_szinfo->mds_stripe_count =
+ ext2fs_le16_to_cpu(mds_szinfo->mds_stripe_count);
+ mds_szinfo->mds_stripe_start =
+ ext2fs_le16_to_cpu(mds_szinfo->mds_stripe_start);
+}
+
+void cputole_mds_objent(struct lfsck_mds_objent *mds_objent)
+{
+ cputole_fid(&mds_objent->mds_fid);
+ cputole_oi(&mds_objent->mds_oi);
+ mds_objent->mds_ostidx = ext2fs_cpu_to_le32(mds_objent->mds_ostidx);
+ mds_objent->mds_ostoffset =
+ ext2fs_cpu_to_le32(mds_objent->mds_ostoffset);
+}
+
+void letocpu_mds_objent(struct lfsck_mds_objent *mds_objent)
+{
+ letocpu_fid(&mds_objent->mds_fid);
+ letocpu_oi(&mds_objent->mds_oi);
+ mds_objent->mds_ostidx = ext2fs_le32_to_cpu(mds_objent->mds_ostidx);
+ mds_objent->mds_ostoffset =
+ ext2fs_le32_to_cpu(mds_objent->mds_ostoffset);
+}
+
+void cputole_ost_objent(struct lfsck_ost_objent *ost_objent)
+{
+ cputole_oi(&ost_objent->ost_oi);
+ ost_objent->ost_size = ext2fs_cpu_to_le64(ost_objent->ost_size);
+ ost_objent->ost_bytes = ext2fs_cpu_to_le64(ost_objent->ost_bytes);
+}
+
+void letocpu_ost_objent(struct lfsck_ost_objent *ost_objent)
+{
+ letocpu_oi(&ost_objent->ost_oi);
+ ost_objent->ost_oi.oi_id = ext2fs_le64_to_cpu(ost_objent->ost_oi.oi_id);
+ ost_objent->ost_oi.oi_seq=ext2fs_le64_to_cpu(ost_objent->ost_oi.oi_seq);
+ ost_objent->ost_size = ext2fs_le64_to_cpu(ost_objent->ost_size);
+ ost_objent->ost_bytes = ext2fs_le64_to_cpu(ost_objent->ost_bytes);
+}
+
+void letocpu_lov_user_md(struct lov_user_md *lmm)
+{
+ struct lov_user_ost_data_v1 *loi;
+ int i;
+
+ lmm->lmm_magic = ext2fs_le32_to_cpu(lmm->lmm_magic);
+ lmm->lmm_pattern = ext2fs_le32_to_cpu(lmm->lmm_pattern);
+ letocpu_oi(&lmm->lmm_oi);
+ lmm->lmm_stripe_size = ext2fs_le32_to_cpu(lmm->lmm_stripe_size);
+ lmm->lmm_stripe_count = ext2fs_le16_to_cpu(lmm->lmm_stripe_count);
+ /* No swabbing needed for the lov_user_md_v3 lmm_pool_name */
+
+ if (lmm->lmm_magic == LOV_USER_MAGIC_V3)
+ loi = ((struct lov_user_md_v3 *)lmm)->lmm_objects;
+ else /* if (lmm->lmm_magic == LOV_USER_MAGIC_V1) */
+ loi = lmm->lmm_objects;
+ /* If there is a bad magic, this will be found immediately in the
+ * call to lfsck_check_lov_ea() following this function. */
+
+ for (i = 0; i < lmm->lmm_stripe_count; i++, loi++) {
+ letocpu_oi(&loi->l_ost_oi);
+ loi->l_ost_gen = ext2fs_le32_to_cpu(loi->l_ost_gen);
+ loi->l_ost_idx = ext2fs_le32_to_cpu(loi->l_ost_idx);
+ }
+}
+
+int lfsck_get_fid(ext2_filsys fs, ino_t ino, struct lu_fid *fid)
+{
+ struct ext2_inode *inode;
+ errcode_t rc;
+ char buf[LMA_OLD_SIZE];
+ struct lustre_mdt_attrs *lma = (struct lustre_mdt_attrs *)buf;
+ int size;
+
+ rc = ext2fs_get_mem(EXT2_INODE_SIZE(fs->super), &inode);
+ if (rc) {
+ com_err("ext2fs_get_mem", rc, "allocating %d bytes\n",
+ EXT2_INODE_SIZE(fs->super));
+ return rc;
+ }
+ rc = ext2fs_read_inode_full(fs, ino, inode, EXT2_INODE_SIZE(fs->super));
+ if (rc) {
+ com_err("ext2fs_read_inode_full", rc,
+ "reading inode %lu\n", ino);
+ ext2fs_free_mem(&inode);
+ return rc;
+ }
+ rc = ext2fs_attr_get(fs, inode, EXT2_ATTR_INDEX_TRUSTED,
+ LUSTRE_XATTR_MDT_LMA, buf, sizeof(buf), &size);
+ if (rc) {
+ if (rc != EXT2_ET_EA_NAME_NOT_FOUND &&
+ rc != EXT2_ET_EA_BAD_MAGIC) {
+ ext2fs_free_mem(&inode);
+ return rc;
+ }
+ /* compose igif */
+ fid->f_seq = ino;
+ fid->f_oid = inode->i_generation;
+ fid->f_ver = 0;
+ } else {
+ *fid = lma->lma_self_fid;
+ }
+ ext2fs_free_mem(&inode);
+ return 0;
+}
+
+int lfsck_is_dirfid_root(const struct lu_fid *dirfid)
+{
+ if (dirfid->f_seq == EXT2_ROOT_INO &&
+ dirfid->f_oid == 0 && dirfid->f_ver == 0)
+ return 1;
+ return 0;
+}
+
+int lfsck_fidcmp(const struct lu_fid *fid1, const struct lu_fid *fid2)
+{
+ if (fid_is_igif(fid1) && fid_is_igif(fid2)) {
+ /* do not compare f_ver for comparing igif-s */
+ if (fid1->f_seq == fid2->f_seq && fid1->f_oid == fid2->f_oid)
+ return 0;
+ return 1;
+ }
+ if (!fid_is_igif(fid1) && !fid_is_igif(fid2)) {
+ if (fid1->f_seq == fid2->f_seq && fid1->f_oid == fid2->f_oid &&
+ fid1->f_ver == fid2->f_ver)
+ return 0;
+ return 1;
+ }
+ return 1;
+}
+#endif
#include "e2fsck.h"
#include <ext2fs/ext2_ext_attr.h>
+#include "ext2fs/lfsck.h"
#include "problem.h"
struct ext2_ext_attr_entry *entry;
char *start;
unsigned int storage_size, remain;
+ struct lov_user_md_v1 *lmm = NULL;
+ struct lustre_mdt_attrs *lma = NULL;
problem_t problem = 0;
inode = (struct ext2_inode_large *) pctx->inode;
goto fix;
}
+ e2fsck_lfsck_find_ea(ctx, inode, entry,
+ start + entry->e_value_offs, &lmm, &lma);
+
/* If EA value is stored in external inode then it does not
* consume space here */
if (entry->e_value_inum == 0)
entry = EXT2_EXT_ATTR_NEXT(entry);
}
+
+ if (lmm)
+ e2fsck_lfsck_save_ea(ctx, pctx->ino, inode->i_generation,
+ lmm, lma);
fix:
+ if (lmm)
+ ext2fs_free_mem(&lmm);
+ if (lma)
+ ext2fs_free_mem(&lma);
/*
* it seems like a corruption. it's very unlikely we could repair
* EA(s) in automatic fashion -bzzz
ext2fs_mark_block_bitmap2(ctx->block_found_map,
fs->super->s_mmp_block);
+ if ((ctx->lustre_devtype & LUSTRE_TYPE) == LUSTRE_MDS) {
+ if (!(ctx->options & E2F_OPT_READONLY) &&
+ e2fsck_lfsck_remove_pending(ctx, block_buf))
+ return;
+ if (e2fsck_lfsck_check_layout(ctx, block_buf) != 0) {
+ com_err(ctx->program_name, 0,
+ _("this lfsck not allowed on this filesystem"));
+ com_err(ctx->program_name, 0,
+ _("see 'lctl lfsck_start' to check Lustre"));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto endit;
+ }
+ }
+
while (1) {
if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
if (e2fsck_mmp_update(fs))
}
e2fsck_pass1_dupblocks(ctx, block_buf);
}
+
+ e2fsck_lfsck_flush_ea(ctx);
+
ext2fs_free_mem(&inodes_to_process);
endit:
e2fsck_use_inode_shortcuts(ctx, 0);
struct ext2_ext_attr_entry *entry;
int count;
region_t region = 0;
+ struct lov_user_md_v1 *lmm = NULL;
+ struct lustre_mdt_attrs *lma = NULL;
int ret;
blk = ext2fs_file_acl_block(fs, inode);
entry->e_hash = hash;
}
+ if (e2fsck_lfsck_find_ea(ctx, (struct ext2_inode_large *)inode,
+ entry, block_buf + entry->e_value_offs,
+ &lmm, &lma) != 0) {
+ if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
+ return 0;
+ }
+
entry = EXT2_EXT_ATTR_NEXT(entry);
}
+
+ if (lmm)
+ e2fsck_lfsck_save_ea(ctx, ino, inode->i_generation, lmm, lma);
+
if (region_allocate(region, (char *)entry - (char *)header, 4)) {
if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
goto clear_extattr;
}
region_free(region);
+ if (lmm)
+ ext2fs_free_mem(&lmm);
+ if (lma)
+ ext2fs_free_mem(&lma);
count = header->h_refcount - 1;
if (count)
return 1;
clear_extattr:
+ if (lmm)
+ ext2fs_free_mem(&lmm);
+ if (lma)
+ ext2fs_free_mem(&lma);
+
if (region)
region_free(region);
ext2fs_file_acl_block_set(fs, inode, 0);
--- /dev/null
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim:shiftwidth=8:tabstop=8:
+ *
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
+ *
+ * This program 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
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
+ *
+ *
+ * Copyright (c) 2004 Hewlett-Packard Co.
+ */
+/*****************************************************************************
+ * e2fsck extentions: code for gathering data from the ost & mds filesystems
+ * when e2fsck is run against them. The best description and knowledge of the
+ * layout and information gathered is in lfsck.h where the structures
+ * defining each entry in the tables are declared. Basically the ost file
+ * contains one table with each entry holding the object id and size.
+ * In addition there is header information at the start of the file.
+ * The mds file contains multiple tables one per ost. Each mds/ost table
+ * contains an entry describing the mds fid and the ost object associated
+ * with this fid on an ost. In addition the mds also contains a table
+ * with the mds_fid and the fid of the containg directory. Header information
+ * for each table is also included.
+ * lfsck is run afterwards where the data gathered and stored here is cross
+ * checked to ensure consistency and correctness
+ *
+ *****************************************************************************/
+#include "config.h"
+#include <string.h>
+#include <time.h>
+#include <errno.h>
+#include <limits.h>
+#include <stdlib.h>
+#include "ext2fs/ext2_fs.h"
+#include "ext2fs/ext2fs.h"
+
+#ifdef HAVE_LFSCK
+#include "e2fsck.h"
+#include "ext2fs/lfsck.h"
+#include "problem.h"
+//#define LOG_REMOVAL
+
+#define VERBOSE(ctx, fmt, args...) \
+do { if (ctx->options & E2F_OPT_VERBOSE) printf(fmt, ##args); } while (0)
+
+#define DEBUG(ctx, fmt, args...) \
+do { if (ctx->options & E2F_OPT_DEBUG) printf(fmt, ##args); } while (0)
+
+struct lfsck_mds_ctx {
+ e2fsck_t ctx;
+ DB *outdb;
+ ext2_ino_t dot;
+ ext2_ino_t dotdot;
+ struct lu_fid dotfid;
+ int numfiles;
+};
+
+struct lfsck_ost_ctx {
+ e2fsck_t ctx;
+ DB *outdb;
+ ext2_ino_t dirinode;
+ int numfiles;
+ int status;
+ __u64 max_objid;
+};
+
+int e2fsck_lfsck_check_lustre_version(char *buf, int buflen)
+{
+ FILE *f;
+ int maj, min, rel;
+ int rc = 0;
+
+ f = fopen("/proc/fs/lustre/version", "r");
+ if (f == NULL)
+ return 0;
+
+ /* /proc/fs/lustre/version should have the first line like:
+ *
+ * lustre: 2.5.55
+ *
+ * and optionally some other lines after it, which we ignore */
+ if (fgets(buf, buflen, f) == NULL) {
+ rc = 0;
+ } else if (sscanf(buf, "lustre: %u.%u.%u", &maj, &min, &rel) != 3) {
+ rc = 1;
+ } else if (maj >= 3 || (maj == 2 && min >= 6) ||
+ (maj == 2 && min == 5 && rel >= 55)) {
+ rc = 1;
+ }
+
+ fclose(f);
+
+ return rc;
+}
+
+/* If the new in-kernel LFSCK has been run on this filesystem to
+ * repair OST orphans, we shouldn't be running this lfsck anymore. */
+int e2fsck_lfsck_check_layout(e2fsck_t ctx, char *block_buf)
+{
+ ext2_filsys fs = ctx->fs;
+ ext2_ino_t tinode;
+ int rc = 0;
+
+ rc = ext2fs_lookup(fs, EXT2_ROOT_INO, LFSCK_LAYOUT,
+ strlen(LFSCK_LAYOUT), block_buf, &tinode);
+ if (rc == 0) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return 1;
+ }
+
+ return 0;
+}
+
+int e2fsck_lfsck_cleanupdb(e2fsck_t ctx)
+{
+ int i;
+ int rc = 0;
+ DB *dbp;
+
+ if (ctx->lfsck_oinfo == NULL)
+ return 0;
+
+ for (i = 0; i < ctx->lfsck_oinfo->ost_count; i++) {
+ if (ctx->lfsck_oinfo->ofile_ctx[i].dbp != NULL) {
+ dbp = ctx->lfsck_oinfo->ofile_ctx[i].dbp;
+ rc += dbp->close(dbp, 0);
+ ctx->lfsck_oinfo->ofile_ctx[i].dbp = NULL;
+ }
+ }
+ if (ctx->lfsck_oinfo->mds_sizeinfo_dbp != NULL) {
+ dbp = ctx->lfsck_oinfo->mds_sizeinfo_dbp;
+ rc += dbp->close(dbp, 0);
+ ctx->lfsck_oinfo->mds_sizeinfo_dbp = NULL;
+ }
+ if (ctx->lfsck_oinfo->ofile_ctx)
+ ext2fs_free_mem(ctx->lfsck_oinfo->ofile_ctx);
+ ext2fs_free_mem(&ctx->lfsck_oinfo);
+
+ return rc;
+}
+
+/* What is the last object id for the OST on the MDS */
+int e2fsck_get_lov_objids(e2fsck_t ctx, struct lfsck_outdb_info *outdb)
+{
+ ext2_filsys fs = ctx->fs;
+ ext2_ino_t inode;
+ ext2_file_t e2_file;
+ __u64 *lov_objids = NULL;
+ unsigned int got;
+ char *block_buf;
+ int i, rc = 0;
+
+ block_buf = e2fsck_allocate_memory(ctx, fs->blocksize * 3,
+ "block iterate buffer");
+
+ rc = ext2fs_lookup(fs, EXT2_ROOT_INO, LOV_OBJID,
+ strlen(LOV_OBJID), block_buf, &inode);
+ if (rc)
+ goto out;
+
+ lov_objids = e2fsck_allocate_memory(ctx,
+ sizeof(*lov_objids) * LOV_MAX_OSTS,
+ "lov_objids array");
+ if (lov_objids == NULL) {
+ rc = ENOMEM;
+ goto out;
+ }
+
+ rc = ext2fs_file_open(fs, inode, 0, &e2_file);
+ if (rc)
+ goto out;
+
+ rc = ext2fs_file_read(e2_file, lov_objids,
+ sizeof(*lov_objids) * LOV_MAX_OSTS, &got);
+ rc = ext2fs_file_close(e2_file);
+
+ outdb->ost_count = got / sizeof(*lov_objids);
+ for (i = 0; i < outdb->ost_count; i++) {
+ VERBOSE(ctx,"MDS: ost_idx %d max_id "LPU64"\n",i,lov_objids[i]);
+ outdb->ofile_ctx[i].max_id = lov_objids[i];
+ outdb->ofile_ctx[i].have_max_id = 1;
+ outdb->have_ost_count = 1;
+ }
+
+out:
+ ext2fs_free_mem(&block_buf);
+ if (lov_objids)
+ ext2fs_free_mem(&lov_objids);
+ if (rc)
+ VERBOSE(ctx, "MDS: unable to read lov_objids: rc %d\n", rc);
+ else
+ VERBOSE(ctx, "MDS: got %d bytes = %d entries in lov_objids\n",
+ got, outdb->ost_count);
+ return rc;
+}
+
+static int lfsck_write_mds_hdrinfo(e2fsck_t ctx, struct lfsck_outdb_info *outdb)
+{
+ struct lfsck_mds_hdr mds_hdr;
+ ext2_filsys fs = ctx->fs;
+ char *mds_hdrname;
+ DB *mds_hdrdb = NULL;
+ DBT key, data;
+ int rc = 0;
+ int i;
+
+ mds_hdrname = e2fsck_allocate_memory(ctx, PATH_MAX,
+ "mds_hdr filename");
+ sprintf(mds_hdrname, "%s.mdshdr", ctx->lustre_mdsdb);
+
+ if (unlink(mds_hdrname)) {
+ if (errno != ENOENT) {
+ fprintf(stderr, "Failure to remove old db file %s\n",
+ mds_hdrname);
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -EINVAL;
+ }
+ }
+
+ rc = lfsck_opendb(mds_hdrname, MDS_HDR, &mds_hdrdb, 0, 0, 0);
+ if (rc != 0) {
+ fprintf(stderr, "failure to open database for mdsdhr "
+ "info%s: %s\n", MDS_HDR, db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ ext2fs_free_mem(&mds_hdrname);
+ return rc;
+ }
+
+ /* read in e2fsck_lfsck_save_ea() already if we opened read/write */
+ if (ctx->lfsck_oinfo->ost_count == 0)
+ e2fsck_get_lov_objids(ctx, ctx->lfsck_oinfo);
+
+ memset(&mds_hdr, 0, sizeof(mds_hdr));
+ mds_hdr.mds_magic = MDS_MAGIC;
+ mds_hdr.mds_flags = ctx->options & E2F_OPT_READONLY;
+ mds_hdr.mds_max_files = fs->super->s_inodes_count -
+ fs->super->s_free_inodes_count;
+ VERBOSE(ctx, "MDS: max_files = "LPU64"\n", mds_hdr.mds_max_files);
+ mds_hdr.mds_num_osts = ctx->lfsck_oinfo->ost_count;
+ VERBOSE(ctx, "MDS: num_osts = %u\n", mds_hdr.mds_num_osts);
+ for (i = 0; i < mds_hdr.mds_num_osts; i++) {
+ mds_hdr.mds_max_ost_id[i] =
+ ctx->lfsck_oinfo->ofile_ctx[i].max_id;
+ }
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &mds_hdr.mds_magic;
+ key.size = sizeof(mds_hdr.mds_magic);
+ cputole_mds_hdr(&mds_hdr);
+ data.data = &mds_hdr;
+ data.size = sizeof(mds_hdr);
+ rc = mds_hdrdb->put(mds_hdrdb, NULL, &key, &data, 0);
+ if (rc != 0) {
+ fprintf(stderr, "error: db put %s: %s\n", MDS_HDR,
+ db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+out:
+ mds_hdrdb->close(mds_hdrdb, 0);
+ ext2fs_free_mem(&mds_hdrname);
+ if (rc == 0) {
+ printf("mds info db file written \n");
+ fflush(stdout);
+
+ }
+ return rc;
+}
+
+int e2fsck_lfsck_save_ea(e2fsck_t ctx, ext2_ino_t ino, __u32 generation,
+ struct lov_user_md *lmm, struct lustre_mdt_attrs *lma)
+{
+ ext2_filsys fs = ctx->fs;
+#ifdef LFSCK_CHECK_SIZE
+ struct lfsck_mds_szinfo szinfo;
+#endif /* LFSCK_CHECK_SIZE */
+ struct lov_user_ost_data_v1 *loi;
+ int rc, i;
+ DBT key, data;
+ DB *dbp;
+ __u32 numfiles = fs->super->s_inodes_count -
+ fs->super->s_free_inodes_count;
+
+ if (!ctx->lfsck_oinfo) {
+ /* remove old db file */
+ if (unlink(ctx->lustre_mdsdb)) {
+ rc = errno;
+ if (rc != ENOENT) {
+ fprintf(stderr,"Error removing old db %s: %s\n",
+ ctx->lustre_mdsdb, strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ return rc;
+ }
+ }
+
+ rc = ext2fs_get_mem(sizeof(struct lfsck_outdb_info),
+ &ctx->lfsck_oinfo);
+ if (rc) {
+ ctx->lfsck_oinfo = NULL;
+ ctx->flags |= E2F_FLAG_ABORT;
+ return rc;
+ }
+ memset(ctx->lfsck_oinfo, 0, sizeof(struct lfsck_outdb_info));
+ rc = ext2fs_get_mem(sizeof(struct lfsck_ofile_ctx)*LOV_MAX_OSTS,
+ &ctx->lfsck_oinfo->ofile_ctx);
+ if (rc) {
+ ext2fs_free_mem(&ctx->lfsck_oinfo);
+ ctx->flags |= E2F_FLAG_ABORT;
+ return rc;
+ }
+ memset(ctx->lfsck_oinfo->ofile_ctx, 0,
+ sizeof(struct lfsck_ofile_ctx) * LOV_MAX_OSTS);
+#ifdef LFSCK_CHECK_SIZE
+ if (lfsck_opendb(ctx->lustre_mdsdb, MDS_SIZEINFO,
+ &ctx->lfsck_oinfo->mds_sizeinfo_dbp, 0,
+ sizeof(szinfo.mds_fid) + sizeof(szinfo),
+ numfiles)) {
+ fprintf(stderr, "Failed to open db file %s\n",
+ MDS_SIZEINFO);
+ ctx->flags |= E2F_FLAG_ABORT;
+ return EIO;
+ }
+#endif /* LFSCK_CHECK_SIZE */
+
+ if (ctx->options & E2F_OPT_READONLY) {
+ e2fsck_get_lov_objids(ctx, ctx->lfsck_oinfo);
+ lfsck_write_mds_hdrinfo(ctx, ctx->lfsck_oinfo);
+ }
+ }
+ if (lmm->lmm_magic == LOV_USER_MAGIC_V3)
+ loi = ((struct lov_user_md_v3 *)lmm)->lmm_objects;
+ else /* if (lmm->lmm_magic == LOV_USER_MAGIC_V1) */
+ loi = lmm->lmm_objects;
+
+#ifdef LFSCK_CHECK_SIZE
+ /* XXX: We don't save the layout type here. This doesn't matter for
+ * now, we don't really need the pool information for lfsck, but
+ * in the future we may need it for RAID-1 and other layouts. */
+ memset(&szinfo, 0, sizeof(szinfo));
+ szinfo.mds_fid = ino;
+ szinfo.mds_seq = 0;
+ szinfo.mds_stripe_size = lmm->lmm_stripe_size;
+ szinfo.mds_stripe_start = loi->l_ost_idx;
+ szinfo.mds_calc_size = 0;
+ szinfo.mds_stripe_pattern = lmm->lmm_pattern;
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &szinfo.mds_fid;
+ key.size = sizeof(szinfo.mds_fid);
+ cputole_mds_szinfo(&szinfo);
+ data.data = &szinfo;
+ data.size = sizeof(szinfo);
+ dbp = ctx->lfsck_oinfo->mds_sizeinfo_dbp;
+
+ rc = dbp->put(dbp, NULL, &key, &data, 0);
+ if (rc != 0) {
+ dbp->err(ctx->lfsck_oinfo->mds_sizeinfo_dbp, rc,
+ "db->put failed\n");
+ e2fsck_lfsck_cleanupdb(ctx);
+ ctx->flags |= E2F_FLAG_ABORT;
+
+ return EIO;
+ }
+#endif /* LFSCK_CHECK_SIZE */
+ for (i = 0; i < lmm->lmm_stripe_count; i++, loi++) {
+ int ost_idx = loi->l_ost_idx;
+ struct lfsck_mds_objent mds_ent;
+ struct lfsck_ofile_ctx *ofile_ctx =
+ &ctx->lfsck_oinfo->ofile_ctx[ost_idx];
+
+ if (ost_idx >= LOV_MAX_OSTS) {
+ fprintf(stderr, "invalid OST index %u ino %u[%d]\n",
+ ost_idx, ino, i);
+ continue;
+ }
+
+ if (ost_idx + 1 > ctx->lfsck_oinfo->ost_count) {
+ if (ctx->lfsck_oinfo->have_ost_count) {
+ fprintf(stderr, "bad OST index %u ino %u[%d]\n",
+ ost_idx, ino, i);
+ continue;
+ }
+ ctx->lfsck_oinfo->ost_count = ost_idx + 1;
+ }
+
+ if (ofile_ctx->dbp == NULL) {
+ char dbname[256];
+
+ sprintf(dbname, "%s.%d", MDS_OSTDB, ost_idx);
+ rc = lfsck_opendb(ctx->lustre_mdsdb, dbname,
+ &ofile_ctx->dbp, 1,
+ sizeof(mds_ent), numfiles);
+ if (rc) {
+ e2fsck_lfsck_cleanupdb(ctx);
+ ctx->flags |= E2F_FLAG_ABORT;
+ return EIO;
+ }
+ }
+
+ memset(&mds_ent, 0, sizeof(mds_ent));
+ if (lma) {
+ mds_ent.mds_fid = lma->lma_self_fid;
+ } else {
+ mds_ent.mds_fid.f_seq = ino;
+ mds_ent.mds_fid.f_oid = generation;
+ mds_ent.mds_fid.f_ver = 0;
+ }
+ mds_ent.mds_oi = *(struct ostid *)&loi->l_ost_oi;
+ mds_ent.mds_ostidx = ost_idx;
+ mds_ent.mds_ostoffset = i;
+
+ if (ostid_id(&mds_ent.mds_oi) > ofile_ctx->max_id) {
+ if (ofile_ctx->have_max_id) {
+ DEBUG(ctx,
+ "[%d] skip obj "LPU64" > max "LPU64"\n",
+ ost_idx, ostid_id(&mds_ent.mds_oi),
+ ofile_ctx->max_id);
+ continue;
+ }
+ ofile_ctx->max_id = ostid_id(&mds_ent.mds_oi);
+ }
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &mds_ent.mds_oi;
+ key.size = sizeof(mds_ent.mds_oi);
+ cputole_mds_objent(&mds_ent);
+ data.data = &mds_ent;
+ data.size = sizeof(mds_ent);
+ dbp = ofile_ctx->dbp;
+#if 0
+ DEBUG(ctx, "OST[%u]: inode FID "DFID" oi "DOSTID"\n", ost_idx,
+ PFID(&mds_ent.mds_fid), POSTID(&mds_ent.mds_oi));
+#endif
+ rc = dbp->put(dbp, NULL, &key, &data, 0);
+ if (rc != 0) {
+ dbp->err(dbp, rc, "db->put failed\n");
+ e2fsck_lfsck_cleanupdb(ctx);
+ ctx->flags |= E2F_FLAG_ABORT;
+ /* XXX - Free lctx memory */
+ return EIO;
+ }
+ }
+ return 0;
+}
+
+static int lfsck_check_lov_ea(e2fsck_t ctx, struct lov_user_md *lmm)
+{
+ if (lmm->lmm_magic != LOV_USER_MAGIC_V1 &&
+ lmm->lmm_magic != LOV_USER_MAGIC_V3) {
+ VERBOSE(ctx, "error: only handle v1/v3 LOV EAs, not %08x\n",
+ lmm->lmm_magic);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+static int lfsck_check_lma_ea(e2fsck_t ctx, struct lustre_mdt_attrs *lma)
+{
+ /* if we find a DNE LMA xattr, then we shouldn't be running on
+ * this filesystem at all. */
+ if ((lma->lma_incompat & ~LMA_INCOMPAT_SUPP) != 0) {
+ VERBOSE(ctx, "error: do not handle lma_incompat %08x\n",
+ lma->lma_incompat & ~LMA_INCOMPAT_SUPP);
+ return -EINVAL;
+ }
+ return 0;
+}
+
+/*
+ * e2fsck pass1 has found a file with an EA let's save the information in
+ * the correct table(s). This is only called for an MDS search.
+ */
+int e2fsck_lfsck_find_ea(e2fsck_t ctx, struct ext2_inode_large *inode,
+ struct ext2_ext_attr_entry *entry, void *value,
+ struct lov_user_md **lmm,
+ struct lustre_mdt_attrs **lma)
+{
+ void *ea = NULL;
+ int retval;
+ unsigned int got;
+ ext2_file_t file;
+
+ /* This ensures that we don't open the file here if traversing an OST */
+ if ((ctx->lustre_devtype & LUSTRE_TYPE) != LUSTRE_MDS)
+ return 0;
+
+ if (!LINUX_S_ISREG(inode->i_mode))
+ return 0;
+
+ if (entry->e_name_len == 0)
+ return 0;
+
+ if (entry->e_name_index != EXT2_ATTR_INDEX_TRUSTED &&
+ entry->e_name_index != EXT2_ATTR_INDEX_LUSTRE)
+ return 0;
+
+ ea = e2fsck_allocate_memory(ctx, entry->e_value_size, "EA");
+
+ if (entry->e_value_inum != 0) {
+ /* EA in external inode */
+ retval = ext2fs_file_open(ctx->fs, entry->e_value_inum,
+ 0, &file);
+ if (!retval) {
+ retval = ext2fs_file_read(file, ea,
+ entry->e_value_size, &got);
+ ext2fs_file_close(file);
+ if (retval != 0) {
+ ext2fs_free_mem(&ea);
+ return 0;
+ }
+ } else {
+ ext2fs_free_mem(&ea);
+ return 0;
+ }
+ } else {
+ memcpy(ea, value, entry->e_value_size);
+ }
+
+ if (entry->e_name_len == strlen(LUSTRE_XATTR_MDT_LOV) &&
+ strncmp(entry->e_name, LUSTRE_XATTR_MDT_LOV,
+ entry->e_name_len) == 0) {
+ if (*lmm) {
+ ext2fs_free_mem(&ea);
+ return -EINVAL;
+ }
+ *lmm = ea;
+ letocpu_lov_user_md(*lmm);
+
+ if (lfsck_check_lov_ea(ctx, *lmm)) {
+ *lmm = NULL;
+ ctx->flags |= E2F_FLAG_ABORT;
+ ext2fs_free_mem(&ea);
+ return -EINVAL;
+ }
+ } else if (entry->e_name_len == strlen(LUSTRE_XATTR_MDT_LMA) &&
+ strncmp(entry->e_name, LUSTRE_XATTR_MDT_LMA,
+ entry->e_name_len) == 0) {
+ if (*lma) {
+ ext2fs_free_mem(&ea);
+ return -EINVAL;
+ }
+ *lma = ea;
+ if (lfsck_check_lma_ea(ctx, *lma)) {
+ *lma = NULL;
+ ctx->flags |= E2F_FLAG_ABORT;
+ ext2fs_free_mem(&ea);
+ return -EINVAL;
+ }
+ } else {
+ ext2fs_free_mem(&ea);
+ }
+
+ return 0;
+}
+
+/* make sure that the mds data is on file */
+int e2fsck_lfsck_flush_ea(e2fsck_t ctx)
+{
+ int i, rc = 0;
+ DB *dbp;
+
+ if ((ctx->lustre_devtype & LUSTRE_TYPE) != LUSTRE_MDS)
+ return 0;
+
+ if (ctx->lfsck_oinfo == 0)
+ return 0;
+
+ for (i = 0; i < ctx->lfsck_oinfo->ost_count; i++) {
+ if (ctx->lfsck_oinfo->ofile_ctx == NULL)
+ break;
+
+ if (ctx->lfsck_oinfo->ofile_ctx[i].dbp != NULL) {
+ dbp = ctx->lfsck_oinfo->ofile_ctx[i].dbp;
+ rc += dbp->close(dbp, 0);
+ ctx->lfsck_oinfo->ofile_ctx[i].dbp = NULL;
+ }
+ }
+ if (ctx->lfsck_oinfo->mds_sizeinfo_dbp != NULL) {
+ dbp = ctx->lfsck_oinfo->mds_sizeinfo_dbp;
+ rc += dbp->close(dbp, 0);
+ ctx->lfsck_oinfo->mds_sizeinfo_dbp = NULL;
+ }
+
+ if (rc)
+ ctx->flags |= E2F_FLAG_ABORT;
+
+ return rc;
+}
+
+/* From debugfs.c for file removal */
+static int lfsck_release_blocks_proc(ext2_filsys fs, blk_t *blocknr,
+ int blockcnt, void *private)
+{
+ blk_t block;
+
+ block = *blocknr;
+ ext2fs_block_alloc_stats(fs, block, -1);
+ return 0;
+}
+
+static void lfsck_kill_file_by_inode(ext2_filsys fs, ext2_ino_t inode)
+{
+ struct ext2_inode inode_buf;
+
+ if (ext2fs_read_inode(fs, inode, &inode_buf))
+ return;
+
+ inode_buf.i_dtime = time(NULL);
+ if (ext2fs_write_inode(fs, inode, &inode_buf))
+ return;
+
+ ext2fs_block_iterate(fs, inode, 0, NULL,
+ lfsck_release_blocks_proc, NULL);
+ ext2fs_inode_alloc_stats2(fs, inode, -1,
+ LINUX_S_ISDIR(inode_buf.i_mode));
+}
+
+/*
+ * remove a file. Currently this removes the lov_objids file
+ * since otherwise the automatic deletion of precreated objects on
+ * mds/ost connection could potentially remove objects with
+ * data - this would be especially the case if the mds has being
+ * restored from backup
+ */
+static int lfsck_rm_file(e2fsck_t ctx, ext2_ino_t dir, char *name)
+{
+ ext2_filsys fs = ctx->fs;
+ ext2_ino_t ino;
+ struct ext2_inode inode;
+ int rc;
+
+ rc = ext2fs_lookup(fs, dir, name, strlen(name),
+ NULL, &ino);
+ if (rc)
+ return 0;
+
+ if (ext2fs_read_inode(fs, ino, &inode))
+ return -EINVAL;
+
+ --inode.i_links_count;
+
+ if (ext2fs_write_inode(fs, ino, &inode))
+ return -EINVAL;
+
+ if (ext2fs_unlink(fs, dir, name, ino, 0))
+ return -EIO;
+
+ if (inode.i_links_count == 0)
+ lfsck_kill_file_by_inode(fs, ino);
+
+ return 0;
+}
+
+/* called for each ost object - save the object id and size */
+static int lfsck_list_objs(ext2_ino_t dir, int entry,
+ struct ext2_dir_entry *dirent, int offset,
+ int blocksize, char *buf, void *priv_data)
+{
+ struct lfsck_ost_ctx *lctx = priv_data;
+ struct lfsck_ost_objent objent;
+ struct ext2_inode inode;
+ DBT key, data;
+ DB *dbp;
+ char name[32]; /* same as filter_fid2dentry() */
+
+ if (!ext2fs_check_directory(lctx->ctx->fs, dirent->inode))
+ return 0;
+
+ memset(name, 0, sizeof(name));
+ strncpy(name, dirent->name, dirent->name_len & 0xFF);
+ memset(&objent, 0, sizeof(objent));
+ objent.ost_oi.oi_id = strtoul(name, NULL, 0);
+ if (objent.ost_oi.oi_id == ULONG_MAX) {
+ lctx->status = 1;
+ lctx->ctx->flags |= E2F_FLAG_ABORT;
+ return DIRENT_ABORT;
+ }
+
+ if (ext2fs_read_inode(lctx->ctx->fs, dirent->inode, &inode)) {
+ lctx->status = 1;
+ lctx->ctx->flags |= E2F_FLAG_ABORT;
+ return DIRENT_ABORT;
+ }
+
+ if (LINUX_S_ISREG(inode.i_mode))
+ objent.ost_size = EXT2_I_SIZE(&inode);
+ else
+ objent.ost_size = inode.i_size;
+ objent.ost_bytes = (__u64)inode.i_blocks * 512;
+
+ if (objent.ost_oi.oi_id > lctx->max_objid)
+ lctx->max_objid = objent.ost_oi.oi_id;
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &objent.ost_oi;
+ key.size = sizeof(objent.ost_oi);
+ cputole_ost_objent(&objent);
+ data.data = &objent;
+ data.size = sizeof(objent);
+ dbp = lctx->outdb;
+ if (dbp->put(dbp, NULL, &key, &data, 0) != 0) {
+ fprintf(stderr, "Failure to put data into db\n");
+ lctx->ctx->flags |= E2F_FLAG_ABORT;
+ return DIRENT_ABORT;
+ }
+
+ lctx->numfiles++;
+ return 0;
+}
+
+/* For each file on the mds save the fid and the containing directory */
+static int lfsck_mds_dirs(ext2_ino_t dir, int entry,
+ struct ext2_dir_entry *dirent, int offset,
+ int blocksize, char *buf, void *priv_data)
+{
+ struct lfsck_mds_ctx *lctx = priv_data;
+ struct lfsck_mds_ctx lctx2;
+ struct lfsck_mds_dirent mds_dirent;
+ DBT key, data;
+ DB *dbp = lctx->outdb;
+ int file_type = (dirent->name_len >> 8) & EXT2_FT_MASK;
+ int rc = 0;
+
+ DEBUG(lctx->ctx, "MDT: inode %u, file %.*s, type %u\n", dirent->inode,
+ dirent->name_len & EXT2_NAME_LEN, dirent->name, file_type);
+ if (dirent->inode == lctx->dot || dirent->inode == lctx->dotdot)
+ return 0;
+
+ /* only the low 4 bits are used to specify file type */
+ if (file_type != EXT2_FT_DIR && file_type != EXT2_FT_REG_FILE)
+ return 0;
+
+ lctx->numfiles++;
+
+ rc = lfsck_get_fid(lctx->ctx->fs, dirent->inode, &mds_dirent.mds_fid);
+ if (rc != 0)
+ goto follow_on;
+
+ DEBUG(lctx->ctx, "MDT: dirfid "DFID" child "DFID" file %.*s\n",
+ PFID(&mds_dirent.mds_dirfid), PFID(&mds_dirent.mds_fid),
+ dirent->name_len & EXT2_NAME_LEN, dirent->name);
+
+ mds_dirent.mds_dirfid = lctx->dotfid;
+ cputole_mds_dirent(&mds_dirent);
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &mds_dirent.mds_fid;
+ key.size = sizeof(mds_dirent.mds_fid);
+ data.data = &mds_dirent;
+ data.size = sizeof(mds_dirent);
+
+ rc = dbp->put(dbp, NULL, &key, &data, 0);
+ if (rc != 0) {
+ if (rc != DB_KEYEXIST) {
+ fprintf(stderr,
+ "error adding MDS inode %.*s (inum %u): %s\n",
+ dirent->name_len & 0xFF, dirent->name,
+ dirent->inode, db_strerror(rc));
+ lctx->ctx->flags |= E2F_FLAG_ABORT;
+ return DIRENT_ABORT;
+ }
+ }
+follow_on:
+ if (file_type == EXT2_FT_DIR) {
+ lctx2 = *lctx;
+ lctx2.dot = dirent->inode;
+ lctx2.dotdot = lctx->dot;
+ lctx2.dotfid = mds_dirent.mds_fid;
+ if (ext2fs_dir_iterate2(lctx->ctx->fs, dirent->inode, 0, NULL,
+ lfsck_mds_dirs, &lctx2)) {
+ return DIRENT_ABORT;
+ }
+ lctx->numfiles = lctx2.numfiles;
+ }
+ return 0;
+}
+
+/* For each directory get the objects and save the data */
+static int lfsck_iterate_obj_dirs(ext2_ino_t dir, int entry,
+ struct ext2_dir_entry *dirent, int offset,
+ int blocksize, char *buf, void *priv_data)
+{
+ struct lfsck_ost_ctx *lctx = priv_data;
+
+ if (ext2fs_check_directory(lctx->ctx->fs, dirent->inode))
+ return 0;
+
+ /* Traverse the d* directories */
+ if (*dirent->name != 'd')
+ return 0;
+
+ ext2fs_dir_iterate2(lctx->ctx->fs, dirent->inode, 0, NULL,
+ lfsck_list_objs, priv_data);
+ if (lctx->status != 0)
+ return DIRENT_ABORT;
+
+ return 0;
+}
+
+static int lfsck_check_entry(ext2_ino_t dir, int entry,
+ struct ext2_dir_entry *dirent, int offset,
+ int blocksize, char *buf, void *priv_data)
+{
+ int *dne = (int *)priv_data;
+ __u64 seq;
+ char *endptr;
+
+ seq = strtoll(dirent->name, &endptr, 16);
+ if (*endptr == '\0') {
+ if (seq >= FID_SEQ_NORMAL) {
+ *dne = 1;
+ return DIRENT_ABORT;
+ }
+ }
+ return 0;
+}
+
+/* Get the starting point of where the objects reside */
+static int lfsck_get_object_dir(e2fsck_t ctx, char *block_buf,ext2_ino_t *inode)
+{
+ ext2_filsys fs = ctx->fs;
+ ext2_ino_t tinode;
+ int rc, dne = 0;
+
+ rc = ext2fs_lookup(fs, EXT2_ROOT_INO, OBJECT_DIR, strlen(OBJECT_DIR),
+ block_buf, &tinode);
+ if (rc) {
+ fprintf(stderr, "error looking up OST object parent dir\n");
+ return ENOENT;
+ }
+ rc = ext2fs_check_directory(fs, tinode);
+ if (rc)
+ return ENOENT;
+
+ /* We don't support DNE so far */
+ rc = ext2fs_dir_iterate2(fs, tinode, 0, NULL, lfsck_check_entry, &dne);
+ if (dne != 0) {
+ fprintf(stderr, "DNE mode isn't supported!\n");
+ return ENOENT;
+ }
+
+ rc = ext2fs_lookup(fs, tinode, OBJECT_DIR_V1, strlen(OBJECT_DIR_V1),
+ block_buf, inode);
+ if (rc) {
+ rc = ext2fs_lookup(fs, tinode, OBJECT_DIR_V2,
+ strlen(OBJECT_DIR_V2), block_buf, inode);
+ if (rc) {
+ fprintf(stderr, "error looking up OST object subdir\n");
+ return ENOENT;
+ }
+ }
+ rc = ext2fs_check_directory(fs, *inode);
+ if (rc)
+ return ENOENT;
+
+ return 0;
+}
+
+/* create a new LAST_ID file */
+static int create_last_id_file(ext2_filsys fs, ext2_ino_t dir_ino,
+ ext2_ino_t *new_ino)
+{
+ struct ext2_inode inode;
+ ext2_ino_t inum;
+ int rc;
+
+ rc = ext2fs_new_inode(fs, dir_ino, S_IFREG | S_IRWXU, 0, &inum);
+ if (rc) {
+ fprintf(stderr, "Creating LAST_ID file failed!\n");
+ return rc;
+ }
+
+ /* Allocate and write new inode before adding the direntry.
+ * If that fails or the program is interrupted, at worst some inode
+ * space will be leaked. But without visible direntry issues */
+ ext2fs_inode_alloc_stats2(fs, inum, +1, 0);
+
+ memset(&inode, 0, sizeof(inode));
+ inode.i_mode = LINUX_S_IFREG | S_IRUSR | S_IWUSR;
+ inode.i_atime = inode.i_ctime = inode.i_mtime =
+ fs->now ? fs->now : time(NULL);
+ inode.i_links_count = 1;
+ inode.i_size = 0;
+ rc = ext2fs_write_new_inode(fs, inum, &inode);
+ if (rc) {
+ fprintf(stderr, "Failed to write new inode\n");
+ return rc;
+ }
+
+ rc = ext2fs_link(fs, dir_ino, "LAST_ID", inum, EXT2_FT_REG_FILE);
+ if (rc == EXT2_ET_DIR_NO_SPACE) {
+ rc = ext2fs_expand_dir(fs, dir_ino);
+ if (rc) {
+ fprintf(stderr, "Failed to expand directory");
+ return rc;
+ }
+ rc = ext2fs_link(fs, dir_ino, "LAST_ID", inum, EXT2_FT_REG_FILE);
+ if (rc) {
+ fprintf(stderr, "Failed to link directory");
+ return rc;
+ }
+ }
+
+ *new_ino = inum;
+ return 0;
+}
+
+int lfsck_set_last_id(e2fsck_t ctx, __u64 last_id)
+{
+ ext2_filsys fs = ctx->fs;
+ ext2_ino_t inode, tinode;
+ ext2_file_t e2_file;
+ char *block_buf;
+ unsigned int written;
+ int rc;
+
+ block_buf = e2fsck_allocate_memory(ctx, fs->blocksize * 3,
+ "lookup buffer");
+
+ rc = lfsck_get_object_dir(ctx, block_buf, &inode);
+ if (rc)
+ goto out;
+
+ rc = ext2fs_lookup(fs, inode, LAST_ID,
+ strlen(LAST_ID), block_buf, &tinode);
+ if (rc == EXT2_ET_FILE_NOT_FOUND) {
+ /* Create a new file */
+ VERBOSE(ctx, "Recreating missing LAST_ID\n");
+ rc = create_last_id_file(fs, inode, &tinode);
+ if (rc)
+ goto out;
+ } else if (rc) {
+ fprintf(stderr, "LAST_ID lookup failed: %d\n",
+ rc);
+ goto out;
+ }
+
+ rc = ext2fs_file_open(fs, tinode, EXT2_FILE_WRITE, &e2_file);
+ if (rc)
+ goto out;
+
+ if (ext2fs_file_get_size(e2_file) < sizeof(__u64)) {
+ rc = ext2fs_file_set_size(e2_file, sizeof(__u64));
+ if (rc) {
+ fprintf(stderr, "Failed to resize LAST_ID\n");
+ goto out;
+ }
+ }
+
+ last_id = ext2fs_cpu_to_le64(last_id);
+
+ rc = ext2fs_file_write(e2_file, &last_id, sizeof(__u64), &written);
+ if (rc) {
+ fprintf(stderr, "Failure to update last id on file\n");
+ ext2fs_file_close(e2_file);
+ goto out;
+ }
+
+ if (written != sizeof(__u64)) {
+ rc = EIO;
+ fprintf(stderr, "Failure to update last id on file\n");
+ ext2fs_file_close(e2_file);
+ goto out;
+ }
+
+ rc = ext2fs_file_close(e2_file);
+
+out:
+ ext2fs_free_mem(&block_buf);
+ return rc;
+}
+
+int e2fsck_get_last_rcvd_info(e2fsck_t ctx, struct obd_uuid *local_uuid,
+ struct obd_uuid *peer_uuid, __u32 *subdircount,
+ __u32 *index, __u32 *compat, __u32 *rocompat,
+ __u32 *incompat)
+{
+ ext2_filsys fs = ctx->fs;
+ ext2_ino_t inode;
+ ext2_file_t e2_file;
+ struct lustre_server_data *lsd = NULL;
+ unsigned int got;
+ char *block_buf;
+ __u32 cmp, inc;
+ int rc = 0;
+
+ block_buf = e2fsck_allocate_memory(ctx, fs->blocksize * 3,
+ "block iterate buffer");
+
+ rc = ext2fs_lookup(fs, EXT2_ROOT_INO, LAST_RCVD, strlen(LAST_RCVD),
+ block_buf, &inode);
+ if (rc)
+ goto out;
+
+ rc = ext2fs_file_open(fs, inode, 0, &e2_file);
+ if (rc)
+ goto out;
+
+ lsd = e2fsck_allocate_memory(ctx, sizeof(*lsd), "lustre server data");
+ if (lsd == NULL) {
+ rc = ENOMEM;
+ goto out;
+ }
+
+ rc = ext2fs_file_read(e2_file, lsd, sizeof(*lsd), &got);
+ if (rc)
+ goto out;
+ if (got != sizeof(*lsd)) {
+ rc = EIO;
+ goto out;
+ }
+
+ if (local_uuid)
+ memcpy(local_uuid, &lsd->lsd_uuid, sizeof(lsd->lsd_uuid));
+
+ if (peer_uuid)
+ memcpy(peer_uuid, &lsd->lsd_peeruuid,sizeof(lsd->lsd_peeruuid));
+
+ if (subdircount)
+ *subdircount = ext2fs_le16_to_cpu(lsd->lsd_subdir_count);
+
+ if (compat == NULL)
+ compat = &cmp;
+ *compat = ext2fs_le32_to_cpu(lsd->lsd_feature_compat);
+ if (rocompat)
+ *rocompat = ext2fs_le32_to_cpu(lsd->lsd_feature_rocompat);
+ if (incompat == NULL)
+ incompat = &inc;
+ *incompat = ext2fs_le32_to_cpu(lsd->lsd_feature_incompat);
+ if (index) {
+ if (*compat & OBD_COMPAT_OST || *incompat & OBD_INCOMPAT_OST)
+ *index = ext2fs_le32_to_cpu(lsd->lsd_ost_index);
+ else if (*compat & OBD_COMPAT_MDT||*incompat & OBD_INCOMPAT_MDT)
+ *index = ext2fs_le32_to_cpu(lsd->lsd_mdt_index);
+ else
+ *index = -1;
+ }
+
+ rc = ext2fs_file_close(e2_file);
+
+out:
+ ext2fs_free_mem(&block_buf);
+ if (lsd)
+ ext2fs_free_mem(&lsd);
+ return rc;
+}
+
+int lfsck_rm_log(ext2_ino_t dir, int entry, struct ext2_dir_entry *dirent,
+ int offset, int blocksize, char *buf, void *priv_data)
+{
+ struct lfsck_ost_ctx *lctx = priv_data;
+ char name[EXT2_NAME_LEN + 1];
+
+ if (!ext2fs_check_directory(lctx->ctx->fs, dirent->inode))
+ return 0;
+
+ strncpy(name, dirent->name, dirent->name_len & 0xFF);
+ name[EXT2_NAME_LEN] = '\0';
+ if (memcmp(name, LAST_ID, strlen(LAST_ID)) == 0)
+ return 0;
+
+
+ if (lfsck_rm_file(lctx->ctx, lctx->dirinode, name))
+ return DIRENT_ABORT;
+
+ return 0;
+}
+
+/* Not 100% sure that this is correct so not activated yet */
+int lfsck_remove_ost_logs(e2fsck_t ctx, char *block_buf)
+{
+ ext2_filsys fs = ctx->fs;
+ struct lfsck_ost_ctx lctx;
+ ext2_ino_t inode;
+ ext2_ino_t tinode;
+ int rc;
+
+ if (lfsck_rm_file(ctx, EXT2_ROOT_INO, CATLIST)) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -EINVAL;
+ }
+
+ rc = ext2fs_lookup(fs, EXT2_ROOT_INO, OBJECT_DIR, strlen(OBJECT_DIR),
+ block_buf, &tinode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+ rc = ext2fs_check_directory(fs, tinode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+
+ rc = ext2fs_lookup(fs, tinode, LOG_DIR, strlen(LOG_DIR),
+ block_buf, &inode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+ rc = ext2fs_check_directory(fs, inode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+ lctx.ctx = ctx;
+ lctx.dirinode = inode;
+
+ if (ext2fs_dir_iterate2(fs, inode, 0, block_buf, lfsck_rm_log, &lctx)) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -EIO;
+ }
+ return 0;
+}
+
+/* Remove files from PENDING dir - this needs to be done before getting ea from
+ * blocks but we need the inode_map bitmap loaded beforehand so load write any
+ * changes then remove references
+ */
+int e2fsck_lfsck_remove_pending(e2fsck_t ctx, char *block_buf)
+{
+ ext2_filsys fs = ctx->fs;
+ struct lfsck_ost_ctx lctx;
+ ext2_ino_t tinode;
+ int rc = 0;
+
+ rc = ext2fs_lookup(fs, EXT2_ROOT_INO, PENDING_DIR, strlen(PENDING_DIR),
+ block_buf, &tinode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+ rc = ext2fs_check_directory(fs, tinode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+
+ lctx.ctx = ctx;
+ lctx.dirinode = tinode;
+
+ e2fsck_read_bitmaps(ctx);
+
+ if (ext2fs_dir_iterate2(fs, tinode, 0, block_buf, lfsck_rm_log, &lctx)){
+ ctx->flags |= E2F_FLAG_ABORT;
+ rc = -EIO;
+ }
+ e2fsck_write_bitmaps(ctx);
+ ext2fs_free_inode_bitmap(fs->inode_map);
+ ext2fs_free_block_bitmap(fs->block_map);
+ fs->inode_map = NULL;
+ fs->block_map = NULL;
+ return rc;
+}
+
+/* partially using code from debugfs do_write() */
+int lfsck_create_objid(e2fsck_t ctx, __u64 objid)
+{
+ int rc = 0;
+ char dirname[32];
+ char name[32];
+ int len, dirlen;
+ __u32 compat, incompat, subdircount;
+ ext2_ino_t inode, tinode, cinode;
+ struct ext2_inode ext2inode;
+ char *block_buf;
+
+ block_buf = e2fsck_allocate_memory(ctx, ctx->fs->blocksize * 3,
+ "lookup buffer");
+
+ memset(name, 0, 32);
+ memset(dirname, 0, 32);
+
+ len = sprintf(name, LPU64, objid);
+
+ fprintf(stderr, "creating %s\n", name);
+
+ rc = e2fsck_get_last_rcvd_info(ctx, NULL, NULL, &subdircount, NULL,
+ &compat, NULL, &incompat);
+ if (rc) {
+ fprintf(stderr, "Error: reading OST last_rcvd file\n");
+ rc = EINVAL;
+ goto out;
+ }
+
+ if (compat & OBD_COMPAT_MDT || incompat & OBD_INCOMPAT_MDT) {
+ fprintf(stderr, "Error: MDS last_rcvd file doing OST check\n");
+ rc = EINVAL;
+ goto out;
+ }
+
+ if (lfsck_get_object_dir(ctx, block_buf, &inode)) {
+ rc = EINVAL;
+ goto out;
+ }
+
+ dirlen = sprintf(dirname, "d%u", (int)objid & (subdircount - 1));
+
+ rc = ext2fs_lookup(ctx->fs, inode, dirname,
+ dirlen, block_buf, &tinode);
+ if (rc) {
+ rc = EINVAL;
+ goto out;
+ }
+
+ if (ext2fs_namei(ctx->fs, EXT2_ROOT_INO, tinode, name, &cinode) == 0) {
+ fprintf(stderr, "Failure to create obj\n");
+ rc = EINVAL;
+ goto out;
+ }
+
+ rc = ext2fs_new_inode(ctx->fs, tinode, 010755, 0, &cinode);
+ if (rc) {
+ fprintf(stderr, "Failure to create obj\n");
+ rc = EINVAL;
+ goto out;
+ }
+
+ rc = ext2fs_link(ctx->fs, tinode, name, cinode, EXT2_FT_REG_FILE);
+ if (rc) {
+ fprintf(stderr, "Failure to create obj\n");
+ rc = EINVAL;
+ goto out;
+ }
+
+ if (ext2fs_test_inode_bitmap2(ctx->fs->inode_map, cinode))
+ fprintf(stderr, "Warning: inode already set");
+
+ ext2fs_inode_alloc_stats2(ctx->fs, cinode, +1, 0);
+ memset(&ext2inode, 0, sizeof(ext2inode));
+ ext2inode.i_mode = LINUX_S_IFREG;
+ ext2inode.i_atime = ext2inode.i_ctime = ext2inode.i_mtime = time(NULL);
+ ext2inode.i_links_count = 1;
+ ext2inode.i_size = 0;
+ if (ext2fs_write_inode(ctx->fs, cinode, &ext2inode)) {
+ fprintf(stderr, "Failure to create obj\n");
+ rc = EINVAL;
+ goto out;
+ }
+
+out:
+ ext2fs_free_mem((void *)&(block_buf));
+ return rc;
+}
+
+/*
+ * For on ost iterate for the direcories and save the object information.
+ */
+void e2fsck_pass6_ost(e2fsck_t ctx)
+{
+ ext2_filsys fs = ctx->fs;
+ struct lfsck_ost_ctx lctx;
+ struct lfsck_ost_hdr ost_hdr;
+ struct lfsck_mds_hdr mds_hdr;
+ struct lfsck_ost_objent objent;
+ DB *outdb = NULL;
+ DB *mds_hdrdb = NULL;
+ DB *osthdr = NULL;
+ DBT key, data;
+ ext2_ino_t dir;
+ __u32 compat, rocompat, incompat;
+ int i, rc;
+ char *block_buf = NULL;
+
+ if (unlink(ctx->lustre_ostdb)) {
+ if (errno != ENOENT) {
+ fprintf(stderr, "Failure to remove old db file %s\n",
+ ctx->lustre_ostdb);
+ ctx->flags |= E2F_FLAG_ABORT;
+ return;
+ }
+ }
+
+ block_buf = e2fsck_allocate_memory(ctx, fs->blocksize * 3,
+ "block iterate buffer");
+
+ rc = lfsck_opendb(ctx->lustre_mdsdb, MDS_HDR, &mds_hdrdb, 0, 0, 0);
+ if (rc != 0) {
+ fprintf(stderr, "failure to open database %s: %s\n",
+ MDS_HDR, db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ memset(&mds_hdr, 0, sizeof(mds_hdr));
+ mds_hdr.mds_magic = MDS_MAGIC;
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &mds_hdr.mds_magic;
+ key.size = sizeof(mds_hdr.mds_magic);
+ data.data = &mds_hdr;
+ data.size = sizeof(mds_hdr);
+ data.ulen = sizeof(mds_hdr);
+ data.flags = DB_DBT_USERMEM;
+ rc = mds_hdrdb->get(mds_hdrdb, NULL, &key, &data, 0);
+ if (rc) {
+ fprintf(stderr,"error getting mds_hdr ("LPU64":%u) in %s: %s\n",
+ mds_hdr.mds_magic, (int)sizeof(mds_hdr.mds_magic),
+ ctx->lustre_mdsdb, db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ memcpy(&mds_hdr, data.data, sizeof(mds_hdr));
+ letocpu_mds_hdr(&mds_hdr);
+
+ rc = lfsck_opendb(ctx->lustre_ostdb, OST_HDR, &osthdr, 0, 0, 0);
+ if (rc != 0) {
+ fprintf(stderr, "failure to open database %s: %s\n",
+ OST_HDR, db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ rc = lfsck_opendb(ctx->lustre_ostdb, OST_OSTDB, &outdb, 0,
+ sizeof(objent.ost_oi) + sizeof(objent),
+ fs->super->s_inodes_count -
+ fs->super->s_free_inodes_count);
+ if (rc != 0) {
+ fprintf(stderr, "error getting ost_hdr in %s: %s\n",
+ ctx->lustre_ostdb, db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ memset(&ost_hdr, 0, sizeof(ost_hdr));
+ if (e2fsck_get_last_rcvd_info(ctx, &ost_hdr.ost_uuid,
+ &ost_hdr.ost_mds_uuid, NULL,
+ &ost_hdr.ost_index,
+ &compat, &rocompat, &incompat)) {
+ fprintf(stderr, "Failure to read OST last_rcvd file\n");
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ VERBOSE(ctx, "OST: '%s' ost idx %u: compat %#x rocomp %#x incomp %#x\n",
+ (char *)&ost_hdr.ost_uuid.uuid, ost_hdr.ost_index,
+ compat, rocompat, incompat);
+
+ if (compat & OBD_COMPAT_MDT) {
+ fprintf(stderr, "Found MDS last_rcvd file doing OST check\n");
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ /*
+ * Get /O/R or /O/0 directory
+ * for each entry scan all the dirents and get the object id
+ */
+ if (lfsck_get_object_dir(ctx, block_buf, &dir)) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ /*
+ * Okay so we have the containing directory so let's iterate over the
+ * containing d* dirs and then iterate again inside
+ */
+ lctx.ctx = ctx;
+ lctx.outdb = outdb;
+ lctx.status = 0;
+ lctx.numfiles = 0;
+ lctx.max_objid = 0;
+ lctx.status = ext2fs_dir_iterate2(fs, dir, 0, block_buf,
+ lfsck_iterate_obj_dirs, &lctx);
+ if (lctx.status) {
+ fprintf(stderr, "Failure in iterating object dirs\n");
+ ctx->flags |= E2F_FLAG_ABORT;
+ return;
+ }
+
+ ost_hdr.ost_magic = OST_MAGIC;
+ ost_hdr.ost_flags = ctx->options & E2F_OPT_READONLY;
+ ost_hdr.ost_num_files = lctx.numfiles;
+ VERBOSE(ctx, "OST: num files = %u\n", lctx.numfiles);
+
+ /* Update the last_id value on the OST if necessary/possible to the
+ * MDS value if larger. Otherwise we risk creating duplicate objects.
+ * If running read-only, we skip this so new objects are ignored. */
+ ost_hdr.ost_last_id = lctx.max_objid;
+ if (!(ctx->options & E2F_OPT_READONLY) &&
+ !(mds_hdr.mds_flags & E2F_OPT_READONLY)) {
+ for (i = 0; i < mds_hdr.mds_num_osts; i++) {
+ if (strcmp((char *)mds_hdr.mds_ost_info[i].uuid,
+ (char *)ost_hdr.ost_uuid.uuid) == 0 &&
+ mds_hdr.mds_max_ost_id[i] >= ost_hdr.ost_last_id)
+ ost_hdr.ost_last_id=mds_hdr.mds_max_ost_id[i]+1;
+ }
+
+ if (lfsck_set_last_id(ctx, ost_hdr.ost_last_id)) {
+ fprintf(stderr, "Failure to set last id\n");
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+#ifdef LOG_REMOVAL
+ if (lfsck_remove_ost_logs(ctx, block_buf))
+ ctx->flags |= E2F_FLAG_ABORT;
+#endif
+ }
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &ost_hdr.ost_magic;
+ key.size = sizeof(ost_hdr.ost_magic);
+ cputole_ost_hdr(&ost_hdr);
+ data.data = &ost_hdr;
+ data.size = sizeof(ost_hdr);
+ if (osthdr->put(osthdr, NULL, &key, &data, 0)) {
+ fprintf(stderr, "Failed to db_put data\n");
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+out:
+ if (mds_hdrdb)
+ mds_hdrdb->close(mds_hdrdb, 0);
+ if (outdb)
+ outdb->close(outdb, 0);
+ if (osthdr)
+ osthdr->close(osthdr, 0);
+ if (block_buf)
+ ext2fs_free_mem((void *)&(block_buf));
+ return;
+}
+
+int lfsck_remove_mds_logs(e2fsck_t ctx)
+{
+ ext2_filsys fs = ctx->fs;
+ struct lfsck_ost_ctx lctx;
+ ext2_ino_t tinode;
+ int rc = 0;
+
+ if (lfsck_rm_file(ctx, EXT2_ROOT_INO, CATLIST)) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -EINVAL;
+ }
+
+ rc = ext2fs_lookup(fs, EXT2_ROOT_INO, OBJECTS, strlen(OBJECTS),
+ NULL, &tinode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+ rc = ext2fs_check_directory(fs, tinode);
+ if (rc) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return -ENOENT;
+ }
+
+ lctx.ctx = ctx;
+ lctx.dirinode = tinode;
+
+ if (ext2fs_dir_iterate2(fs, tinode, 0, NULL, lfsck_rm_log, &lctx)) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ rc = -EIO;
+ }
+ return rc;
+}
+
+
+/*
+ * On the mds save the fid and directory information for each file.
+ * The mds ost tables have already been populated by pass1
+ */
+void e2fsck_pass6_mdt(e2fsck_t ctx)
+{
+ ext2_filsys fs = ctx->fs;
+ struct problem_context pctx;
+ struct lfsck_mds_ctx lctx;
+ struct lfsck_mds_dirent mds_dirent;
+ struct lfsck_mds_hdr mds_hdr;
+ DBT key, data;
+ DB *outdb = NULL, *dbhdr = NULL;
+ __u32 compat, rocompat, incompat, index;
+ int rc, i;
+
+ clear_problem_context(&pctx);
+
+ memset(&lctx, 0, sizeof(lctx));
+ lctx.ctx = ctx;
+
+ /* Found no files with EA on filesystem - empty */
+ if (ctx->lfsck_oinfo == NULL) {
+ if (unlink(ctx->lustre_mdsdb)) {
+ if (errno != ENOENT) {
+ fprintf(stderr, "Failure to remove old "
+ "db file %s\n", ctx->lustre_mdsdb);
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+ }
+ rc = ext2fs_get_mem(sizeof(struct lfsck_outdb_info),
+ &ctx->lfsck_oinfo);
+ if (rc) {
+ ctx->lfsck_oinfo = NULL;
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+ memset(ctx->lfsck_oinfo, 0, sizeof(struct lfsck_outdb_info));
+ rc = ext2fs_get_mem(sizeof(struct lfsck_ofile_ctx)*LOV_MAX_OSTS,
+ &ctx->lfsck_oinfo->ofile_ctx);
+ if (rc) {
+ ext2fs_free_mem(&ctx->lfsck_oinfo);
+ ctx->lfsck_oinfo = NULL;
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+ memset(ctx->lfsck_oinfo->ofile_ctx, 0,
+ sizeof(struct lfsck_ofile_ctx) * LOV_MAX_OSTS);
+ }
+
+ if (!(ctx->options & E2F_OPT_READONLY))
+ lfsck_write_mds_hdrinfo(ctx, ctx->lfsck_oinfo);
+
+ if (lfsck_opendb(ctx->lustre_mdsdb, MDS_DIRINFO, &outdb, 1,
+ sizeof(mds_dirent.mds_fid) + sizeof(mds_dirent),
+ fs->super->s_inodes_count -
+ fs->super->s_free_inodes_count)) {
+ fprintf(stderr, "failure to open database %s\n", MDS_DIRINFO);
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ lctx.outdb = outdb;
+ lctx.numfiles = 0;
+ lctx.dot = EXT2_ROOT_INO;
+ lctx.dotdot = EXT2_ROOT_INO;
+ lctx.dotfid.f_seq = EXT2_ROOT_INO;
+
+ rc = ext2fs_dir_iterate2(fs, EXT2_ROOT_INO,0,NULL,lfsck_mds_dirs,&lctx);
+ if (rc != 0) {
+ fprintf(stderr, "Error iterating directories: %d\n", rc);
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ /* read in e2fsck_lfsck_save_ea() already if we opened read/write */
+ if (ctx->lfsck_oinfo->ost_count == 0)
+ e2fsck_get_lov_objids(ctx, ctx->lfsck_oinfo);
+
+ memset(&mds_hdr, 0, sizeof(mds_hdr));
+ mds_hdr.mds_magic = MDS_MAGIC;
+ mds_hdr.mds_flags = ctx->options & E2F_OPT_READONLY;
+ mds_hdr.mds_max_files = fs->super->s_inodes_count -
+ fs->super->s_free_inodes_count;
+ VERBOSE(ctx, "MDS: max_files = "LPU64"\n", mds_hdr.mds_max_files);
+ mds_hdr.mds_num_osts = ctx->lfsck_oinfo->ost_count;
+ VERBOSE(ctx, "MDS: num_osts = %u\n", mds_hdr.mds_num_osts);
+ for (i = 0; i < mds_hdr.mds_num_osts; i++) {
+ mds_hdr.mds_max_ost_id[i] =
+ ctx->lfsck_oinfo->ofile_ctx[i].max_id;
+ }
+
+ if (e2fsck_get_last_rcvd_info(ctx, &mds_hdr.mds_uuid, NULL, NULL,
+ &index, &compat, &rocompat, &incompat)) {
+ fprintf(stderr, "Failure to read MDS last_rcvd file\n");
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ VERBOSE(ctx, "MDS: '%s' mdt idx %u: compat %#x rocomp %#x incomp %#x\n",
+ (char *)&mds_hdr.mds_uuid.uuid, index,compat,rocompat,incompat);
+
+ if (compat & OBD_COMPAT_OST || incompat & OBD_INCOMPAT_OST) {
+ fprintf(stderr, "Found OST last_rcvd file doing MDS check\n");
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ if (!(ctx->options & E2F_OPT_READONLY)) {
+ if (lfsck_rm_file(ctx, EXT2_ROOT_INO, LOV_OBJID)) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+#ifdef LOG_REMOVAL
+ if (lfsck_remove_mds_logs(ctx)) {
+ ctx->flags |= E2F_FLAG_ABORT;
+ return;
+ }
+#endif
+ }
+
+ rc = lfsck_opendb(ctx->lustre_mdsdb, MDS_HDR, &dbhdr, 0, 0, 0);
+ if (rc != 0) {
+ fprintf(stderr, "failure to open database %s: %s\n", MDS_HDR,
+ db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+
+ memset(&key, 0, sizeof(key));
+ memset(&data, 0, sizeof(data));
+ key.data = &mds_hdr.mds_magic;
+ key.size = sizeof(mds_hdr.mds_magic);
+ cputole_mds_hdr(&mds_hdr);
+ data.data = &mds_hdr;
+ data.size = sizeof(mds_hdr);
+ rc = dbhdr->put(dbhdr, NULL, &key, &data, 0);
+ if (rc != 0) {
+ fprintf(stderr, "error: db put %s: %s\n", MDS_HDR,
+ db_strerror(rc));
+ ctx->flags |= E2F_FLAG_ABORT;
+ goto out;
+ }
+out:
+ if (dbhdr)
+ dbhdr->close(dbhdr, 0);
+ if (outdb)
+ outdb->close(outdb, 0);
+}
+
+/* If lfsck checking requested then gather the data */
+void e2fsck_pass6(e2fsck_t ctx)
+{
+ if (ctx->lustre_devtype == LUSTRE_NULL)
+ return;
+
+ printf("Pass 6: Acquiring %s information for lfsck\n",
+ ctx->lustre_devtype & LUSTRE_OST ? "OST" :
+ ctx->lustre_devtype & LUSTRE_MDS ? "MDT" : "device");
+
+ fflush(stdout);
+
+ if (ctx->lustre_devtype & LUSTRE_OST)
+ e2fsck_pass6_ost(ctx);
+ else if (ctx->lustre_devtype & LUSTRE_MDS)
+ e2fsck_pass6_mdt(ctx);
+ else
+ fprintf(stderr, "Invalid lustre dev %x\n", ctx->lustre_devtype);
+
+ return;
+}
+#endif /* HAVE_LFSCK */
*/
#define _XOPEN_SOURCE 600 /* for inclusion of sa_handler in Solaris */
+#define _GNU_SOURCE
#include "config.h"
#include <stdio.h>
#ifdef HAVE_DIRENT_H
#include <dirent.h>
#endif
+#ifdef HAVE_LIMITS_H
+#include <limits.h>
+#endif
#include "e2p/e2p.h"
#include "et/com_err.h"
#include "problem.h"
#include "../version.h"
+#include "ext2fs/lfsck.h"
+
+static struct option long_options[] = {
+#ifdef HAVE_LFSCK
+ { "mdsdb", 1, NULL, 1 },
+ { "mdtdb", 1, NULL, 1 },
+ { "ostdb", 1, NULL, 2 },
+#endif
+ { 0, 0, 0, 0 }
+};
+
/* Command line options */
static int cflag; /* check disk */
static int show_version_only;
fs->super->s_checkinterval*2))
reason = 0;
}
+#ifdef HAVE_LFSCK
+ if (ctx->lustre_devtype & LUSTRE_TYPE) {
+ if (!reason || ctx->options & E2F_OPT_READONLY)
+ ctx->lustre_devtype |= LUSTRE_ONLY;
+ if (!reason)
+ reason = _(" lustre database creation");
+ }
+#endif
+
if (reason) {
log_out(ctx, "%s", ctx->device_name);
log_out(ctx, reason, reason_arg);
log_out(ctx, "\n");
skip:
ext2fs_close_free(&ctx->fs);
+#ifdef HAVE_LFSCK
+ if (ctx->lustre_mdsdb)
+ free(ctx->lustre_mdsdb);
+ if (ctx->lustre_ostdb)
+ free(ctx->lustre_ostdb);
+ if (ctx->lfsck_oinfo)
+ e2fsck_lfsck_cleanupdb(ctx);
+#endif /* HAVE_LFSCK */
+
e2fsck_free_context(ctx);
exit(FSCK_OK);
}
{
int flush = 0;
int c, fd;
+ int option_index;
#ifdef MTRACE
extern void *mallwatch;
#endif
ctx->inode_badness_threshold = BADNESS_THRESHOLD;
- while ((c = getopt (argc, argv, "panyrcC:B:dE:fvtFVM:b:I:j:P:l:L:N:SsDk")) != EOF)
+ ctx->lustre_devtype = LUSTRE_NULL;
+
+ while ((c = getopt_long(argc, argv,
+ "panyrcC:B:dE:fvtFVM:b:I:j:P:l:L:N:SsDk",
+ long_options, &option_index)) != EOF)
switch (c) {
+#ifdef HAVE_LFSCK
+ case 1: {
+ char *dbpath, *tmp;
+
+ if (!optarg)
+ usage(ctx);
+
+ dbpath = malloc(PATH_MAX);
+ if (dbpath == NULL) {
+ fprintf(stderr, "Out of memory\n");
+ exit(1);
+ }
+ tmp = malloc(PATH_MAX);
+ if (tmp == NULL) {
+ fprintf(stderr, "Out of memory\n");
+ exit(1);
+ }
+
+ strcpy(tmp, optarg);
+ if (realpath(my_dirname(tmp), dbpath) == NULL) {
+ fprintf(stderr, "Failure to resolve path %s\n",
+ optarg);
+ exit(1);
+ }
+
+ strcpy(tmp, optarg);
+ sprintf(dbpath+strlen(dbpath), "/%s", my_basename(tmp));
+ ctx->lustre_mdsdb = dbpath;
+ ctx->lustre_devtype |= LUSTRE_MDS;
+
+ free(tmp);
+ break;
+ }
+ case 2: {
+ char *dbpath, *tmp;
+
+ dbpath = malloc(PATH_MAX);
+ if (dbpath == NULL) {
+ fprintf(stderr, "Out of memory\n");
+ exit(1);
+ }
+ tmp = malloc(PATH_MAX);
+ if (tmp == NULL) {
+ fprintf(stderr, "Out of memory\n");
+ exit(1);
+ }
+
+ strcpy(tmp, optarg);
+ if (realpath(my_dirname(tmp), dbpath) == NULL) {
+ fprintf(stderr, "Failure to resolve path %s\n",
+ optarg);
+ exit(1);
+ }
+
+ strcpy(tmp, optarg);
+ sprintf(dbpath+strlen(dbpath), "/%s", my_basename(tmp));
+ ctx->lustre_ostdb = dbpath;
+ ctx->lustre_devtype |= LUSTRE_OST;
+
+ free(tmp);
+ break;
+ }
+#endif /* HAVE_LFSCK */
case 'C':
ctx->progress = e2fsck_update_progress;
res = sscanf(optarg, "%d", &ctx->progress_fd);
break;
case 'v':
verbose = 1;
+ ctx->options |= E2F_OPT_VERBOSE;
break;
case 'V':
show_version_only = 1;
default:
usage(ctx);
}
+#ifdef HAVE_LFSCK
+ if (ctx->lustre_devtype) {
+ char buf[128];
+
+ if ((ctx->lustre_devtype != LUSTRE_MDS) &&
+ ctx->lustre_devtype != (LUSTRE_MDS | LUSTRE_OST)) {
+ com_err(ctx->program_name, 0,
+ _("must specify --mdsdb with --ostdb"));
+ usage(ctx);
+ }
+ if (e2fsck_lfsck_check_lustre_version(buf, sizeof(buf)) != 0) {
+ com_err(ctx->program_name, 0,
+ _("this lfsck not allowed on version %s"), buf);
+ com_err(ctx->program_name, 0,
+ _("see 'lctl lfsck_start' to check Lustre"));
+ fatal_error(ctx, 0);
+ }
+ }
+#endif /* HAVE_LFSCK */
if (show_version_only)
return 0;
if (optind != argc - 1)
}
no_journal:
- if (ctx->qctx) {
+ if (run_result & E2F_FLAG_ABORT) {
+ fatal_error(ctx, _("aborted"));
+ } else if (run_result & E2F_FLAG_CANCEL) {
+ log_out(ctx, _("%s: e2fsck canceled.\n"), ctx->device_name ?
+ ctx->device_name : ctx->filesystem_name);
+ exit_value |= FSCK_CANCELED;
+ } else if (ctx->qctx) {
int i, needs_writeout;
for (i = 0; i < MAXQUOTAS; i++) {
if (qtype != -1 && qtype != i)
ext2fs_close_free(&ctx->fs);
goto restart;
}
- if (run_result & E2F_FLAG_ABORT)
- fatal_error(ctx, _("aborted"));
#ifdef MTRACE
mtrace_print("Cleanup");
#endif
was_changed = ext2fs_test_changed(fs);
- if (run_result & E2F_FLAG_CANCEL) {
- log_out(ctx, _("%s: e2fsck canceled.\n"), ctx->device_name ?
- ctx->device_name : ctx->filesystem_name);
- exit_value |= FSCK_CANCELED;
- } else if (!(ctx->options & E2F_OPT_READONLY)) {
+ if (!(ctx->options & E2F_OPT_READONLY)) {
if (ext2fs_test_valid(fs)) {
if (!(sb->s_state & EXT2_VALID_FS))
exit_value |= FSCK_NONDESTRUCT;
ext2fs_close_free(&ctx->fs);
free(ctx->journal_name);
+#ifdef HAVE_LFSCK
+ if (ctx->lfsck_oinfo)
+ e2fsck_lfsck_cleanupdb(ctx);
+ if (ctx->lustre_mdsdb)
+ free(ctx->lustre_mdsdb);
+ if (ctx->lustre_ostdb)
+ free(ctx->lustre_ostdb);
+#endif /* HAVE_LFSCK */
e2fsck_free_context(ctx);
remove_error_table(&et_ext2_error_table);
Url: https://downloads.hpdd.intel.com/public/e2fsprogs/
Provides: ldiskfsprogs = %{version}
+@LFSCK_CMT@Requires: db4 >= @DB4VERSION@
+@LFSCK_CMT@BuildRequires: db4-devel >= 4.1 texinfo
BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}-root-%(%{__id_u} -n)
Requires: e2fsprogs-libs = %{version}-%{release}
Requires: libcom_err >= 1.42.2
--disable-e2initrd-helper --disable-libblkid --disable-libuuid \
--disable-defrag --enable-symlink-install \
--enable-quota \
+ @WITH_LUSTRE@ @ENABLE_LFSCK@ \
%{?extra_config_flags:%extra_config_flags}
make %{?_smp_mflags}
%{_root_sbindir}/fsck.ext3
%{_root_sbindir}/fsck.ext4
%{_root_sbindir}/fsck.ext4dev
+@LFSCK_CMT@%{_sbindir}/lfsck
%{_root_sbindir}/logsave
%{_root_sbindir}/mke2fs
%{_root_sbindir}/mkfs.ext2
%{_mandir}/man8/e2image.8*
%{_mandir}/man8/e2label.8*
%{_mandir}/man8/e2undo.8*
+@LFSCK_CMT@%{_mandir}/man8/lfsck.8*
%{_mandir}/man8/logsave.8*
%{_mandir}/man8/mke2fs.8*
%{_mandir}/man8/mkfs.ext2.8*
--disable-uuidd \
--disable-fsck \
--enable-quota \
+ @WITH_LUSTRE@ @ENABLE_LFSCK@ \
%{?extra_config_flags:%extra_config_flags} CFLAGS="$RPM_OPT_FLAGS"
make V=1
/sbin/fsck.ext2
/sbin/fsck.ext3
/sbin/fsck.ext4
+@LFSCK_CMT@/usr/sbin/lfsck
/sbin/mke2fs
/sbin/mkfs.ext2
/sbin/mkfs.ext3
BuildRoot: %{_tmppath}/%{name}-root
%if %{_vendor} == "suse"
Group: System/Filesystems
+@LFSCK_CMT@Requires: db >= @DB4VERSION@
+@LFSCK_CMT@BuildRequires: db-devel >= 4.1 texinfo
Provides: e2fsbn ext2fs libcom_err = %{version} ldiskfsprogs = %{version}
Obsoletes: ext2fs libcom_err < %{version}
%else
Group: System Environment/Base
+@LFSCK_CMT@Requires: db4 >= @DB4VERSION@
+@LFSCK_CMT@BuildRequires: db4-devel >= 4.1 texinfo
Provides: e2fsprogs-libs = %{version} ldiskfsprogs = %{version}
Obsoletes: e4fsprogs e2fsprogs-libs < %{version}
%endif
%build
%configure --enable-elf-shlibs --enable-nls --disable-defrag \
--enable-quota \
+ @WITH_LUSTRE@ @ENABLE_LFSCK@ \
%{?extra_config_flags:%extra_config_flags}
make
make check
%{_root_sbindir}/fsck.ext3
%{_root_sbindir}/fsck.ext4
%{_root_sbindir}/fsck.ext4dev
+@LFSCK_CMT@%{_sbindir}/lfsck
%{_root_sbindir}/logsave
%{_root_sbindir}/mke2fs
%{_root_etcdir}/mke2fs.conf
%{_mandir}/man8/debugfs.8*
%{_mandir}/man8/dumpe2fs.8*
%{_mandir}/man8/e2fsck.8*
+@LFSCK_CMT@%{_mandir}/man8/lfsck.8*
%{_mandir}/man8/findfs.8*
%{_mandir}/man8/fsck.ext2.8*
%{_mandir}/man8/fsck.ext3.8*
/* Define to 1 if you have the `chflags' function. */
#undef HAVE_CHFLAGS
+/* Define to 1 if DB4 library is present */
+#undef HAVE_DB4
+
+/* Define to 1 if you have the <db.h> header file. */
+#undef HAVE_DB_H
+
/* Define if the GNU dcgettext() function is already present or preinstalled.
*/
#undef HAVE_DCGETTEXT
/* Define if your <locale.h> file defines LC_MESSAGES. */
#undef HAVE_LC_MESSAGES
+/* Define to 1 if Lustre lfsck is enabled */
+#undef HAVE_LFSCK
+
/* Define to 1 if you have the <limits.h> header file. */
#undef HAVE_LIMITS_H
/* Define to 1 if lseek64 declared in unistd.h */
#undef HAVE_LSEEK64_PROTOTYPE
+/* Define to 1 if lustre/liblustreapi.h found */
+#undef HAVE_LUSTRE_LIBLUSTREAPI_H
+
+/* Define to 1 if lustre/lustreapi.h found */
+#undef HAVE_LUSTRE_LUSTREAPI_H
+
/* Define to 1 if you have the `mallinfo' function. */
#undef HAVE_MALLINFO
@MAKEFILE_BSDLIB@
@MAKEFILE_PROFILE@
+@LFSCK_CMT@LUSTRE_INC=-I @LUSTRE@/lustre/include -I @LUSTRE@/include -I @LUSTRE@/libcfs/include -Wall
+
all:: ext2fs.pc
.c.o:
$(E) " CC $<"
- $(Q) $(CC) $(ALL_CFLAGS) -c $< -o $@
+ $(Q) $(CC) $(ALL_CFLAGS) $(LUSTRE_INC) -c $< -o $@
$(Q) $(CHECK_CMD) $(ALL_CFLAGS) $<
@PROFILE_CMT@ $(Q) $(CC) $(ALL_CFLAGS) -g -pg -o profiled/$*.o -c $<
@ELF_CMT@ $(Q) $(CC) $(ALL_CFLAGS) -fPIC -o elfshared/$*.o -c $<
debugfs.o: $(top_srcdir)/debugfs/debugfs.c
$(E) " CC $<"
- $(Q) $(CC) $(ALL_CFLAGS) -c $< -o $@
+ $(Q) $(CC) $(ALL_CFLAGS) $(LUSTRE_INC) -c $< -o $@
extent_inode.o: $(top_srcdir)/debugfs/extent_inode.c
$(E) " CC $<"
ls.o: $(top_srcdir)/debugfs/ls.c
$(E) " CC $<"
- $(Q) $(CC) $(ALL_CFLAGS) -c $< -o $@
+ $(Q) $(CC) $(ALL_CFLAGS) $(LUSTRE_INC) -c $< -o $@
lsdel.o: $(top_srcdir)/debugfs/lsdel.c
$(E) " CC $<"
}
memcpy(buffer, block_buf + entry->e_value_offs,
entry->e_value_size);
- error = 0;
}
cleanup:
/* This is unfortunately needed for older lustre_user.h to be usable */
#define LASSERT(cond) do { } while (0)
+#include "../lib/ext2fs/ext2fsP.h"
+#include <ext2fs/ext2_ext_attr.h>
#ifdef HAVE_LUSTRE_LUSTREAPI_H
#include <lustre/lustreapi.h>
#elif HAVE_LUSTRE_LIBLUSTREAPI_H
dst->f_oid = ext2fs_le32_to_cpu(src->f_oid);
dst->f_ver = ext2fs_le32_to_cpu(src->f_ver);
}
+#ifndef LMA_INCOMPAT_SUPP
+#define LMAI_RELEASED 0x00000001
+#define LMA_INCOMPAT_SUPP (LMAI_RELEASED)
+#endif
+
#endif /* HAVE_LUSTRE_LUSTRE_IDL_H */
#define LUSTRE_XATTR_MDT_LOV "lov"
#define LUSTRE_XATTR_OST_FID "fid"
#ifndef LMA_OLD_SIZE
-#ifndef LMA_INCOMPAT_SUPP
struct lustre_mdt_attrs {
__u32 lma_compat;
__u32 lma_incompat;
struct lu_fid lma_self_fid;
};
-#endif
struct filter_fid_old {
struct lu_fid ff_parent;
#define LMA_OLD_SIZE 64
#endif /* !LMA_OLD_SIZE */
+/* Defines for Lustre */
+#define LUSTRE_NULL 0x0000
+#define LUSTRE_MDS 0x0001
+#define LUSTRE_OST 0x0002
+#define LUSTRE_TYPE 0x000f
+#define LUSTRE_ONLY 0x1000
+
+#ifdef HAVE_LFSCK /* compiling with Lustre lfsck support */
+
+#ifdef DOSTID
+#define lmm_oi lmm_oi.oi
+#define l_ost_oi l_ost_oi.oi
+#else
+struct ostid {
+ __u64 oi_id;
+ __u64 oi_seq;
+};
+#define DOSTID LPX64":"LPX64
+#define POSTID(oi) (oi)->oi_seq, (oi)->oi_id
+
+#define l_ost_oi l_object_id /* for lov_ost_data_v1 */
+#define lmm_oi lmm_object_id /* for lov_mds_md_v1/3 */
+#endif /* DOSTID */
+
+#ifndef HAVE_LUSTRE_LUSTRE_IDL_H
+#define OBD_CONNECT_FID 0x40000000ULL
+
+enum fid_seq {
+ FID_SEQ_OST_MDT0 = 0ULL,
+ FID_SEQ_IGIF = 12ULL,
+ FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
+ FID_SEQ_IDIF = 0x100000000ULL,
+ FID_SEQ_NORMAL = 0x200000400ULL,
+};
+
+static inline int fid_seq_is_igif(const __u64 seq)
+{
+ return seq >= FID_SEQ_IGIF && seq <= FID_SEQ_IGIF_MAX;
+}
+
+static inline int fid_is_igif(const struct lu_fid *fid)
+{
+ return fid_seq_is_igif(fid_seq(fid));
+}
+
+static inline __u64 ostid_id(const void *oip)
+{
+ const struct ostid *oi = oip;
+
+ if (oi->oi_seq == FID_SEQ_OST_MDT0)
+ return oi->oi_id;
+
+ return oi->oi_seq;
+}
+
+static inline __u64 ostid_seq(const void *oip)
+{
+ const struct ostid *oi = oip;
+
+ if (oi->oi_seq == FID_SEQ_OST_MDT0)
+ return FID_SEQ_OST_MDT0;
+
+ return oi->oi_id;
+}
+
+/* convert an OST objid + index into an IDIF FID SEQ number */
+static inline __u64 fid_idif_seq(__u64 id, __u32 ost_idx)
+{
+ return FID_SEQ_IDIF | (ost_idx << 16) | ((id >> 32) & 0xffff);
+}
+
+/* convert ost_id from 1.x compatible OST protocol into FID for future usage */
+static inline void ostid_idif_unpack(struct ostid *oi, struct lu_fid *fid,
+ __u32 idx)
+{
+ fid->f_seq = fid_idif_seq(ostid_id(oi), idx);
+ fid->f_oid = ostid_id(oi); /* truncate to 32 bits by assignment */
+ fid->f_ver = ostid_id(oi) >> 48;/* in theory, not currently used */
+}
+#endif /* HAVE_LUSTRE_LUSTRE_IDL_H */
+
+#ifndef LPU64
+#if (__WORDSIZE == 32) || defined(__x86_64__)
+# define LPU64 "%llu"
+# define LPD64 "%lld"
+# define LPX64 "%#llx"
+# define LPSZ "%u"
+# define LPSSZ "%d"
+#elif (__WORDSIZE == 64)
+# define LPU64 "%lu"
+# define LPD64 "%ld"
+# define LPX64 "%#lx"
+# define LPSZ "%lu"
+# define LPSSZ "%ld"
+#endif
+#endif /* !LPU64 */
+
+/* Get O/R or O/0 dir */
+#define OBJECT_DIR "O"
+#define OBJECT_DIR_V1 "R"
+#define OBJECT_DIR_V2 "0"
+#define LOG_DIR "1"
+#define PENDING_DIR "PENDING"
+#define LFSCK_LAYOUT "lfsck_layout"
+#define OBJECTS "OBJECTS"
+#define CATLIST "CATALOGS"
+#define LAST_ID "LAST_ID"
+#define LAST_RCVD "last_rcvd"
+#define LOV_OBJID "lov_objid"
+
+/* Database names */
+#define MDS_HDR "mdshdr"
+#define MDS_DIRINFO "mds_dirinfo"
+#define MDS_SIZEINFO "mds_sizeinfo"
+#define MDS_OSTDB "mds_ostdb"
+#define OST_HDR "osthdr"
+#define OST_OSTDB "ost_db"
+
+#define MDS_MAGIC 0xDBABCD01
+#define OST_MAGIC 0xDB123402
+
+#define OBD_COMPAT_OST 0x00000002 /* this is an OST (1.6+) */
+#define OBD_COMPAT_MDT 0x00000004 /* this is an MDT (1.6+) */
+
+#define OBD_INCOMPAT_OST 0x00000002 /* this is an OST (1.8+) */
+#define OBD_INCOMPAT_MDT 0x00000004 /* this is an MDS (1.8+) */
+
+#define LOV_MAX_OSTS 2048 /* Arbitrary limit, can be increased */
+#define LOV_EA_SIZE(lum, num) (sizeof(*lum) + num * sizeof(*lum->lmm_objects))
+#define LOV_EA_MAX(lum) LOV_EA_SIZE(lum, LOV_MAX_OSTS)
+
+struct lustre_server_data {
+ __u8 lsd_uuid[40]; /* server UUID */
+ __u64 lsd_last_transno; /* last completed transaction ID */
+ __u64 lsd_compat14; /* reserved - compat with old last_rcvd */
+ __u64 lsd_mount_count; /* incarnation number */
+ __u32 lsd_feature_compat; /* compatible feature flags */
+ __u32 lsd_feature_rocompat;/* read-only compatible feature flags */
+ __u32 lsd_feature_incompat;/* incompatible feature flags */
+ __u32 lsd_server_size; /* size of server data area */
+ __u32 lsd_client_start; /* start of per-client data area */
+ __u16 lsd_client_size; /* size of per-client data area */
+ __u16 lsd_subdir_count; /* number of subdirectories for objects */
+ __u64 lsd_catalog_oid; /* recovery catalog object id */
+ __u32 lsd_catalog_ogen; /* recovery catalog inode generation */
+ __u8 lsd_peeruuid[40]; /* UUID of LOV/OSC associated with MDS */
+ __u32 lsd_ost_index; /* index number of OST in LOV */
+ __u32 lsd_mdt_index; /* index number of MDT in LMV */
+};
+
+struct lfsck_mds_hdr {
+ __u64 mds_magic;
+ __u64 mds_flags;
+ __u64 mds_max_files;
+ __u32 mds_num_osts;
+ __u32 mds_unused;
+ __u64 mds_max_ost_id[LOV_MAX_OSTS];
+ struct obd_uuid mds_uuid;
+ struct obd_uuid mds_ost_info[LOV_MAX_OSTS];
+};
+
+struct lfsck_ost_hdr {
+ __u64 ost_magic;
+ __u64 ost_flags;
+ __u64 ost_num_files;
+ __u64 ost_last_id;
+ __u32 ost_index;
+ __u32 ost_unused;
+ struct obd_uuid ost_mds_uuid;
+ struct obd_uuid ost_uuid;
+};
+
+struct lfsck_mds_dirent {
+ struct lu_fid mds_dirfid;
+ struct lu_fid mds_fid;
+};
+
+struct lfsck_mds_szinfo {
+ __u64 mds_fid;
+ __u64 mds_seq;
+ __u64 mds_size;
+ __u64 mds_calc_size;
+ __u32 mds_stripe_size;
+ __u32 mds_stripe_pattern;
+ __u16 mds_stripe_count;
+ __u16 mds_stripe_start;
+};
+
+struct lfsck_mds_objent {
+ struct lu_fid mds_fid;
+ struct ostid mds_oi;
+ __u32 mds_ostidx;
+ __u32 mds_ostoffset;
+};
+
+struct lfsck_ost_objent {
+ struct ostid ost_oi;
+ __u64 ost_size;
+ __u64 ost_bytes;
+};
+
+#ifdef HAVE_DB_H
+#include <db.h>
+#endif
+
+struct lfsck_ofile_ctx {
+ DB *dbp;
+ __u64 max_id;
+ int have_max_id;
+};
+
+struct lfsck_outdb_info {
+ __u32 ost_count;
+ int have_ost_count;
+ DB *mds_sizeinfo_dbp;
+ struct lfsck_ofile_ctx *ofile_ctx;
+};
+
+/* pass6.c */
+#ifdef FSCK_OK /* compiling for e2fsck or lfsck */
+extern int e2fsck_lfsck_check_lustre_version(char *buf, int buflen);
+extern int e2fsck_lfsck_check_layout(e2fsck_t ctx, char *block_buf);
+extern int e2fsck_lfsck_find_ea(e2fsck_t ctx, struct ext2_inode_large *inode,
+ struct ext2_ext_attr_entry *entry, void *value,
+ struct lov_user_md_v1 **lmm,
+ struct lustre_mdt_attrs **lma);
+extern int e2fsck_lfsck_save_ea(e2fsck_t ctx, ext2_ino_t ino, __u32 generation,
+ struct lov_user_md_v1 *lmm,
+ struct lustre_mdt_attrs *lma);
+extern int e2fsck_lfsck_flush_ea(e2fsck_t ctx);
+extern int e2fsck_lfsck_cleanupdb(e2fsck_t ctx);
+extern int e2fsck_lfsck_remove_pending(e2fsck_t ctx, char *block_buf);
+
+/* lfsck_common.c */
+extern char *my_dirname(char *path);
+extern const char *my_basename(const char *path);
+extern int lfsck_create_dbenv(const char *progname);
+extern int lfsck_opendb(const char *fname, const char *dbname, DB **dbpp,
+ int allow_dup, int keydata_size, int num_files);
+extern void cputole_mds_hdr(struct lfsck_mds_hdr *mds_hdr);
+extern void letocpu_mds_hdr(struct lfsck_mds_hdr *mds_hdr);
+extern void cputole_ost_hdr(struct lfsck_ost_hdr *ost_hdr);
+extern void letocpu_ost_hdr(struct lfsck_ost_hdr *ost_hdr);
+extern void cputole_fid(struct lu_fid *fid);
+extern void letocpu_fid(struct lu_fid *fid);
+extern void cputole_mds_dirent(struct lfsck_mds_dirent *mds_dirent);
+extern void letocpu_mds_dirent(struct lfsck_mds_dirent *mds_dirent);
+extern void cputole_mds_szinfo(struct lfsck_mds_szinfo *mds_szinfo);
+extern void letocpu_mds_szinfo(struct lfsck_mds_szinfo *mds_szinfo);
+extern void cputole_mds_objent(struct lfsck_mds_objent *mds_objent);
+extern void letocpu_mds_objent(struct lfsck_mds_objent *mds_objent);
+extern void cputole_ost_objent(struct lfsck_ost_objent *ost_objent);
+extern void letocpu_ost_objent(struct lfsck_ost_objent *ost_objent);
+extern void letocpu_lov_user_md(struct lov_user_md *lmm);
+
+int lfsck_get_fid(ext2_filsys fs, ino_t ino, struct lu_fid *fid);
+int lfsck_is_dirfid_root(const struct lu_fid *dirfid);
+int lfsck_fidcmp(const struct lu_fid *fid1, const struct lu_fid *fid2);
+#endif /* FSCK_OK */
+
+#define MDS_START_DIRENT_TABLE sizeof(struct lfsck_mds_hdr)
+
+#define MDS_START_SZINFO_TABLE(numfiles) \
+ (sizeof(struct lfsck_mds_hdr) + (sizeof(struct lfsck_mds_dirent) * numfiles))
+
+#define MDS_START_OST_TABLE_OFFSET(idx, numfiles) \
+ (sizeof(struct lfsck_mds_hdr) + (sizeof(struct lfsck_mds_dirent) * numfiles)+\
+ (sizeof(struct lfsck_mds_szinfo) * numfiles) + \
+ (sizeof(struct lfsck_mds_objent_hdr) + \
+ ((sizeof(struct lfsck_mds_objent) * numfiles)) * (idx)) + \
+ sizeof(struct lfsck_mds_objent_hdr))
+
+#define MDS_START_OST_HDR_OFFSET(idx, numfiles) \
+ (sizeof(struct lfsck_mds_hdr) + (sizeof(struct lfsck_mds_dirent) * numfiles)+\
+ (sizeof(struct lfsck_mds_szinfo) * numfiles) + \
+ (sizeof(struct lfsck_mds_objent_hdr) + \
+ ((sizeof(struct lfsck_mds_objent) * numfiles)) * (idx)))
+
+#define OST_START_OFFSET sizeof(struct lfsck_ost_hdr)
+
+#else /* !HAVE_LFSCK */
+#define e2fsck_lfsck_check_layout(ctx, block_buf) (0)
+#define e2fsck_lfsck_find_ea(ctx, inode, entry, value, lmm, lma) (0)
+#define e2fsck_lfsck_save_ea(ctx, ino, generation, lmm, lma) do {} while(0)
+#define e2fsck_lfsck_flush_ea(ctx) (0)
+#define e2fsck_lfsck_cleanupdb(ctx) (0)
+#define e2fsck_lfsck_remove_pending(ctx, block_buf) (0)
+#endif /* HAVE_LFSCK */
+
#endif /* LFSCK_H */
E2FSPROGS_MONTH @E2FSPROGS_MONTH@
E2FSPROGS_YEAR @E2FSPROGS_YEAR@
E2FSPROGS_VERSION @E2FSPROGS_VERSION@
+LFSCK_CMT @LFSCK_CMT@
+LFSCK_MAN @LFSCK_MAN@
SIZEOF_LONG_LONG @SIZEOF_LONG_LONG@
SIZEOF_LONG @SIZEOF_LONG@
SIZEOF_INT @SIZEOF_INT@
git://git.whamcloud.com / tools / e2fsprogs.git / commitdiff