On Mon, 23 May 2011 08:24:33 -0500
Shirish Pargaonkar <shirishpargaonkar@xxxxxxxxx> wrote:
> On Mon, May 23, 2011 at 8:08 AM, Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> > On Mon, 23 May 2011 08:04:09 -0500
> > Shirish Pargaonkar <shirishpargaonkar@xxxxxxxxx> wrote:
> >
> >> On Mon, May 23, 2011 at 7:42 AM, Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> >> > On Mon, 23 May 2011 07:26:49 -0500
> >> > shirishpargaonkar@xxxxxxxxx wrote:
> >> >
> >> >> From: Shirish Pargaonkar <shirishpargaonkar@xxxxxxxxx>
> >> >>
> >> >> Handle cifs.idmap type of key. Extract a SID string from the description
> >> >> and map it to either an uid or gid using winbind APIs.
> >> >> If that fails (e.g. because winbind is not installed/running or winbind returns
> >> >> an error), try to obtain uid of 'nobody' and gid of 'nogroup'.
> >> >> And if that fails, kernel assigns uid and gid (from mount superblock).
> >> >>
> >> >> Enable including winbind header files and idmapping code conditional
> >> >> to winbind devel rpms (header and library).
> >> >>
> >> >> An entry such as this
> >> >>
> >> >> create Âcifs.idmap  *    *        /usr/sbin/cifs.idmap %k
> >> >>
> >> >> is needed in the file /etc/request-key.conf.
> >> >>
> >> >>
> >> >> Signed-off-by: Shirish Pargaonkar <shirishpargaonkar@xxxxxxxxx>
> >> >> ---
> >> >> ÂMakefile.am   Â|  11 +++-
> >> >> Âaclocal/idmap.m4 | Â 45 +++++++++++
> >> >> Âcifs.idmap.c   | Â221 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
> >> >> Âcifs.upcall.c  Â|  Â1 -
> >> >> Âconfigure.ac   |  20 +++++-
> >> >> Â5 files changed, 294 insertions(+), 4 deletions(-)
> >> >> Âcreate mode 100644 aclocal/idmap.m4
> >> >> Âcreate mode 100644 cifs.idmap.c
> >> >>
> >> >> diff --git a/Makefile.am b/Makefile.am
> >> >> index 67a0190..c6eeb22 100644
> >> >> --- a/Makefile.am
> >> >> +++ b/Makefile.am
> >> >> @@ -8,8 +8,10 @@ mount_cifs_LDADD = $(LIBCAP) $(CAPNG_LDADD)
> >> >>
> >> >> Âman_MANS = mount.cifs.8
> >> >>
> >> >> +sbin_PROGRAMS =
> >> >> +
> >> >> Âif CONFIG_CIFSUPCALL
> >> >> -sbin_PROGRAMS = cifs.upcall
> >> >> +sbin_PROGRAMS += cifs.upcall
> >> >> Âcifs_upcall_SOURCES = cifs.upcall.c data_blob.c asn1.c spnego.c util.c
> >> >> Âcifs_upcall_LDADD = -ltalloc -lkeyutils $(KRB5_LDADD)
> >> >> Âman_MANS += cifs.upcall.8
> >> >> @@ -30,3 +32,10 @@ bin_PROGRAMS = cifscreds
> >> >> Âcifscreds_SOURCES = cifscreds.c resolve_host.c util.c
> >> >> Âcifscreds_LDADD = -lkeyutils
> >> >> Âendif
> >> >> +
> >> >> +if CONFIG_CIFSIDMAP
> >> >> +sbin_PROGRAMS += cifs.idmap
> >> >> +cifs_idmap_SOURCES = cifs.idmap.c
> >> >> +cifs_idmap_LDADD = -lkeyutils $(WINB_LDADD)
> >> >> +endif
> >> >> +
> >> >> diff --git a/aclocal/idmap.m4 b/aclocal/idmap.m4
> >> >> new file mode 100644
> >> >> index 0000000..211d372
> >> >> --- /dev/null
> >> >> +++ b/aclocal/idmap.m4
> >> >> @@ -0,0 +1,45 @@
> >> >> +dnl Headers needed by wbclient.h
> >> >> +dnl
> >> >> +AC_DEFUN([AC_WBCH_COMPL],[
> >> >> +[
> >> >> +#ifdef HAVE_STDINT_H
> >> >> +#include <stdint.h>
> >> >> +#endif
> >> >> +]
> >> >> +[#ifdef HAVE_STDBOOL_H
> >> >> +#include <stdbool.h>
> >> >> +#endif
> >> >> +]
> >> >> +[#ifdef HAVE_STDIO_H
> >> >> +#include <stdio.h>
> >> >> +#endif
> >> >> +]
> >> >> +[#ifdef HAVE_STDLIB_H
> >> >> +#include <stdlib.h>
> >> >> +#endif
> >> >> +]
> >> >> +[#ifdef HAVE_ERRNO_H
> >> >> +#include <errno.h>
> >> >> +#endif
> >> >> +]])
> >> >> +
> >> >> +dnl Check for wbclient.h header and libwbclietn.so
> >> >> +dnl
> >> >> +AC_DEFUN([AC_TEST_WBCHL],[
> >> >> +if test $enable_cifsidmap != "no"; then
> >> >> + Â Â AC_CHECK_HEADERS([wbclient.h], , [
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â if test "$enable_cifsidmap" = "yes"; then
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AC_MSG_ERROR([wbclient.h not found, consider installing libwbclient-devel.])
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â else
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â AC_MSG_WARN([wbclient.h not found, consider installing libwbclient-devel. Disabling cifs.idmap.])
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â enable_cifsidmap="no"
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â fi
> >> >> + Â Â Â Â Â Â Â Â Â Â ], [ AC_WBCH_COMPL ])
> >> >> +fi
> >> >> +
> >> >> +if test $enable_cifsidmap != "no"; then
> >> >> + Â Â AC_CHECK_LIB([wbclient], [wbcStringToSid],
> >> >> + Â Â Â Â Â Â [ WINB_LDADD='-lwbclient' ] [ AC_DEFINE(HAVE_LIBWBCLIENT, 1, ["Define var have_libwbclient"]) ], [AC_MSG_ERROR([No functioning wbclient library found!])])
> >> >> + Â Â AC_SUBST(WINB_LDADD)
> >> >> +fi
> >> >> +])
> >> >> diff --git a/cifs.idmap.c b/cifs.idmap.c
> >> >> new file mode 100644
> >> >> index 0000000..7788bf2
> >> >> --- /dev/null
> >> >> +++ b/cifs.idmap.c
> >> >> @@ -0,0 +1,221 @@
> >> >> +/*
> >> >> +* CIFS idmap helper.
> >> >> +* Copyright (C) Shirish Pargaonkar (shirishp@xxxxxxxxxx) 2011
> >> >> +*
> >> >> +* Used by /sbin/request-key.conf for handling
> >> >> +* cifs upcall for SID to uig/gid and uid/gid to SID mapping.
> >> >> +* You should have keyutils installed and add
> >> >> +* this lines to /etc/request-key.conf file:
> >> >> +
> >> >> + Â Âcreate cifs.idmap * * /usr/local/sbin/cifs.idmap %k
> >> >> +
> >> >> +* This program is free software; you can redistribute it and/or modify
> >> >> +* it under the terms of the GNU General Public License as published by
> >> >> +* the Free Software Foundation; either version 2 of the License, or
> >> >> +* (at your option) any later version.
> >> >> +* 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 for more details.
> >> >> +* You should have received a copy of the GNU General Public License
> >> >> +* along with this program; if not, write to the Free Software
> >> >> +* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
> >> >> +*/
> >> >> +
> >> >> +#ifdef HAVE_CONFIG_H
> >> >> +#include "config.h"
> >> >> +#endif /* HAVE_CONFIG_H */
> >> >> +
> >> >> +#include <string.h>
> >> >> +#include <getopt.h>
> >> >> +#include <syslog.h>
> >> >> +#include <dirent.h>
> >> >> +#include <sys/types.h>
> >> >> +#include <sys/stat.h>
> >> >> +#include <unistd.h>
> >> >> +#include <keyutils.h>
> >> >> +#include <stdint.h>
> >> >> +#include <stdbool.h>
> >> >> +#include <stdio.h>
> >> >> +#include <stdlib.h>
> >> >> +#include <errno.h>
> >> >> +#include <limits.h>
> >> >> +#include <wbclient.h>
> >> >> +
> >> >> +static const char *prog = "cifs.idmap";
> >> >> +
> >> >> +static void usage(void)
> >> >> +{
> >> >> + Â Â fprintf(stderr, "Usage: %s key_serial\n", prog);
> >> >> +}
> >> >> +
> >> >> +char *strget(const char *str, char *substr)
> >> >> +{
> >> >> + Â Â int len, sublen, retlen;
> >> >> + Â Â char *retstr, *substrptr;
> >> >> +
> >> >> + Â Â sublen = strlen(substr);
> >> >> + Â Â substrptr = strstr(str, substr);
> >> >> + Â Â if (substrptr) {
> >> >> + Â Â Â Â Â Â len = strlen(substrptr);
> >> >> + Â Â Â Â Â Â substrptr += sublen;
> >> >> +
> >> >> + Â Â Â Â Â Â retlen = len - sublen;
> >> >> + Â Â Â Â Â Â if (retlen > 0) {
> >> >> + Â Â Â Â Â Â Â Â Â Â retstr = malloc(retlen + 1);
> >> >> + Â Â Â Â Â Â Â Â Â Â if (retstr) {
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â strncpy(retstr, substrptr, retlen);
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â return retstr;
> >> >> + Â Â Â Â Â Â Â Â Â Â }
> >> >> + Â Â Â Â Â Â }
> >> >> + Â Â }
> >> >> +
> >> >> + Â Â return NULL;
> >> >> +}
> >> >> +
> >> >
> >> > I still worry a bit about bounds checking here, but I think we can trust
> >> > keyctl_describe_alloc to give us a NULL terminated string, so let's not
> >> > sweat it for now.
> >> >
> >> >> +static int
> >> >> +cifs_idmap(const key_serial_t key, const char *key_descr)
> >> >> +{
> >> >> + Â Â uid_t uid = 0;
> >> >> + Â Â gid_t gid = 0;;
> >> >> + Â Â wbcErr rc = 1;
> >> >> + Â Â char *sidstr = NULL;
> >> >> + Â Â struct wbcDomainSid sid;
> >> >> + Â Â struct passwd *pw;
> >> >> + Â Â struct group *gr;
> >> >> +
> >> >> + Â Â /*
> >> >> + Â Â Â* Use winbind to convert received string to a SID and lookup
> >> >> + Â Â Â* name and map that SID to an uid. ÂIf either of these
> >> >> + Â Â Â* function calls return with an error, Âuse system calls to obtain
> >> >> + Â Â Â* uid of user "nobody". If winbind fails to map a SID to an UID
> >> >> + Â Â Â* and there is no user named "nobody", return error to the
> >> >> + Â Â Â* upcall caller. Otherwise instanticate a key using that uid.
> >> >> + Â Â Â*
> >> >> + Â Â Â* The same applies to SID and gid mapping. ÂInstead of a
> >> >> + Â Â Â* user "nobody", user "nogroup" is looked up if winbind
> >> >> + Â Â Â* fails to map a SID to a gid.
> >> >> + Â Â Â*/
> >> >> + Â Â sidstr = strget(key_descr, "os:");
> >> >> + Â Â if (sidstr) {
> >> >> + Â Â Â Â Â Â rc = wbcStringToSid(sidstr, &sid);
> >> >> + Â Â Â Â Â Â if (rc)
> >> >> + Â Â Â Â Â Â Â Â Â Â syslog(LOG_DEBUG, "Invalid owner string: %s, rc: %d",
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â key_descr, rc);
> >> >> + Â Â Â Â Â Â else {
> >> >> + Â Â Â Â Â Â Â Â Â Â rc = wbcSidToUid(&sid, &uid);
> >> >> + Â Â Â Â Â Â Â Â Â Â if (rc)
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â syslog(LOG_DEBUG, "SID %s to uid wbc error: %d",
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â key_descr, rc);
> >> >> + Â Â Â Â Â Â }
> >> >> + Â Â Â Â Â Â if (rc) { /* either of the two wbcSid functions failed */
> >> >> + Â Â Â Â Â Â Â Â Â Â pw = getpwnam("nobody");
> >> >> + Â Â Â Â Â Â Â Â Â Â if (!pw)
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â syslog(LOG_DEBUG, "SID %s to uid pw error: %d",
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â key_descr, rc);
> >> >> + Â Â Â Â Â Â Â Â Â Â else {
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â uid = pw->pw_uid;
> >> >> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â rc = 0;
> >> >> + Â Â Â Â Â Â Â Â Â Â }
> >> > Â Â Â Â Â Â Â Â Â Â Â Â^^^^^^^^^^^
> >> > Â Â Â Â Â Â Â Â Â Â Â ÂIs it ever possible that the error from the wb
> >> > Â Â Â Â Â Â Â Â Â Â Â Âlibs will be transient? If so, then mapping
> >> > Â Â Â Â Â Â Â Â Â Â Â Âthis to nobody could be problematic, right?
> >> >
> >>
> >> We can't say. It is possible that the error from winbind is
> >> transiet and it is also possible that winbind is not running
> >> at all.
> >>
> >> > Â Â Â Â Â Â Â Â Â Â Â ÂMight it be better to not map it to anything
> >> > Â Â Â Â Â Â Â Â Â Â Â Âand let the kernel turn this into the mount
> >> > Â Â Â Â Â Â Â Â Â Â Â Âuid/gid? If "nobody" really is better then you
> >> > Â Â Â Â Â Â Â Â Â Â Â Âprobably ought to set a short timeout on this
> >> > Â Â Â Â Â Â Â Â Â Â Â Âkey.
> >>
> >> Even if the key is changed to a transiet, the mapping in cache in kernel
> >> is "permanent" i.e. at least till that cache entry is evicted by shrinker,
> >> if at all.
> >>
> >> What we can do is perhaps return different codes to kernel to only
> >> mark an entry as mapped when key handling code returns 0, say
> >> if it returns 1, then it is nobody/nogroup and do not mark an entry
> >> as mapped. If it is error (negative ret code), do not mark an entry
> >> as mapped but assign sb uid/gid.
> >>
> >
> > That sounds a bit complicated. Why not just exit without instantiating
> > the key if a successful mapping can't be made? With this scheme, you'll
> > get different behavior depending on whether cifs.idmap couldn't be
> > called, or if it got called and the mapping didn't get made.
> >
> > I think it might be best to have the user get the same behavior in the
> > event of any failure. Thoughts?
> >
> > --
> > Jeff Layton <jlayton@xxxxxxxxxx>
> >
>
> I think I am OK with that. So basically we are ruling out assigning
> nobody/nogroup as uid/gid.
I think that's probably best. It gives us simpler to document behavior
which is generally a good thing for people deploying this. When you
write the manpage for this (and you are going to write it, right?), you
should probably make note that people should take great care to set
their uid=/gid= options.
--
Jeff Layton <jlayton@xxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
