Tomas Mraz [Wed, 17 Feb 2016 13:57:15 +0000 (14:57 +0100)]
pam_env: Document the /etc/environment file.
* modules/pam_env/Makefile.am: Add the environment.5 soelim stub.
* modules/pam_env/pam_env.8.xml: Add environ(7) reference.
* modules/pam_env/pam_env.conf.5.xml: Add environment alias name.
Add a paragraph about /etc/environment. Add environ(7) reference.
Tomas Mraz [Wed, 17 Feb 2016 13:21:41 +0000 (14:21 +0100)]
pam_unix: Add no_pass_expiry option to ignore password expiration.
* modules/pam_unix/pam_unix.8.xml: Document the no_pass_expiry option.
* modules/pam_unix/pam_unix_acct.c (pam_sm_acct_mgmt): If no_pass_expiry
is on and return value data is not set to PAM_SUCCESS then ignore
PAM_NEW_AUTHTOK_REQD and PAM_AUTHTOK_EXPIRED returns.
* modules/pam_unix/pam_unix_auth.c (pam_sm_authenticate): Always set the
return value data.
(pam_sm_setcred): Test for likeauth option and use the return value data
only if set.
* modules/pam_unix/support.h: Add the no_pass_expiry option.
Thorsten Kukuk [Mon, 22 Jun 2015 12:53:01 +0000 (14:53 +0200)]
Release version 1.2.1
Security fix: CVE-2015-3238
If the process executing pam_sm_authenticate or pam_sm_chauthtok method
of pam_unix is not privileged enough to check the password, e.g.
if selinux is enabled, the _unix_run_helper_binary function is called.
When a long enough password is supplied (16 pages or more, i.e. 65536+
bytes on a system with 4K pages), this helper function hangs
indefinitely, blocked in the write(2) call while writing to a blocking
pipe that has a limited capacity.
With this fix, the verifiable password length will be limited to
PAM_MAX_RESP_SIZE bytes (i.e. 512 bytes) for pam_exec and pam_unix.
* NEWS: Update
* configure.ac: Bump version
* libpam/Makefile.am: Bump version of libpam
* libpam_misc/Makefile.am: Bump version of libpam_misc
* po/*: Regenerate po files
Tomas Mraz [Tue, 7 Apr 2015 08:52:16 +0000 (10:52 +0200)]
Use crypt_r if available in pam_userdb and in pam_unix.
* modules/pam_unix/passverify.c (create_password_hash): Call crypt_r()
instead of crypt() if available.
* modules/pam_userdb/pam_userdb.c (user_lookup): Call crypt_r()
instead of crypt() if available.
Dmitry V. Levin [Tue, 6 Jan 2015 22:29:11 +0000 (22:29 +0000)]
build: raise gettext version requirement
Raise gettext requirement to the latest oldstable version 0.18.3.
This fixes the following automake warning:
configure.ac:581: warning: The 'AM_PROG_MKDIR_P' macro is deprecated, and its use is discouraged.
configure.ac:581: You should use the Autoconf-provided 'AC_PROG_MKDIR_P' macro instead,
configure.ac:581: and use '$(MKDIR_P)' instead of '$(mkdir_p)'in your Makefile.am files.
* configure.ac (AM_GNU_GETTEXT_VERSION): Raise from 0.15 to 0.18.3.
* po/Makevars: Update from gettext-0.18.3.
Dmitry V. Levin [Tue, 6 Jan 2015 22:58:27 +0000 (22:58 +0000)]
Remove unmodified GNU gettext files installed by autopoint
These files are part of GNU gettext; we have not modified them, they are
installed by autopoint which is called by autoreconf, so they had to be
removed from this repository along with ABOUT-NLS, config.rpath, and
mkinstalldirs files that were removed by commit
Linux-PAM-1_1_5-7-g542ec8b.
Luke Shumaker [Mon, 22 Dec 2014 20:46:43 +0000 (15:46 -0500)]
libpam: Only print "Password change aborted" when it's true.
pam_get_authtok() may be used any time that a password needs to be entered,
unlike pam_get_authtok_{no,}verify(), which may only be used when
changing a password; yet when the user aborts, it prints "Password change
aborted." whether or not that was the operation being performed.
This bug was non-obvious because none of the modules distributed with
Linux-PAM use it for anything but changing passwords; pam_unix has its
own utility function that it uses instead. As an example, the
nss-pam-ldapd package uses it in pam_sm_authenticate().
libpam/pam_get_authtok.c (pam_get_authtok_internal): check that the
password is trying to be changed before printing a message about the
password change being aborted.
Dmitry V. Levin [Tue, 9 Dec 2014 21:47:40 +0000 (21:47 +0000)]
build: extend cross compiling check to cover CPPFLAGS (ticket #21)
Use BUILD_CPPFLAGS variable to override CPPFLAGS where necessary in
case of cross compiling, in addition to CC_FOR_BUILD, BUILD_CFLAGS,
and BUILD_LDFLAGS variables introduced earlier to override CC,
CFLAGS, and LDFLAGS, respectively.
* configure.in (BUILD_CPPFLAGS): Define.
* doc/specs/Makefile.am (CPPFLAGS): Define to @BUILD_CPPFLAGS@.
Tomas Mraz [Fri, 17 Oct 2014 06:34:24 +0000 (08:34 +0200)]
pam_succeed_if: Use long long type for numeric values
The currently used long with additional conversion to int is
too small for uids and gids.
modules/pam_succeed_if/pam_succeed_if.c (evaluate_num): Replace
strtol() with strtoll() and int with long long in the parameters
of comparison functions.
Tomas Mraz [Fri, 5 Sep 2014 07:09:37 +0000 (09:09 +0200)]
Add grantor field to audit records of libpam.
The grantor field gives audit trail of PAM modules which granted access
for successful return from libpam calls. In case of failed return
the grantor field is set to '?'.
libpam/pam_account.c (pam_acct_mgmt): Remove _pam_auditlog() call.
libpam/pam_auth.c (pam_authenticate, pam_setcred): Likewise.
libpam/pam_password.c (pam_chauthtok): Likewise.
libpam/pam_session.c (pam_open_session, pam_close_session): Likewise.
libpam/pam_audit.c (_pam_audit_writelog): Add grantors parameter,
add grantor= field to the message if grantors is set.
(_pam_list_grantors): New function creating the string with grantors list.
(_pam_auditlog): Add struct handler pointer parameter, call _pam_list_grantors()
to list the grantors from the handler list.
(_pam_audit_end): Add NULL handler parameter to _pam_auditlog() call.
(pam_modutil_audit_write): Add NULL grantors parameter to _pam_audit_writelog().
libpam/pam_dispatch.c (_pam_dispatch_aux): Set h->grantor where appropriate.
(_pam_clear_grantors): New function to clear grantor field of handler.
(_pam_dispatch): Call _pam_clear_grantors() before executing the stack.
Call _pam_auditlog() when appropriate.
libpam/pam_handlers.c (extract_modulename): Do not allow empty module name
or just "?" to avoid confusing audit trail.
(_pam_add_handler): Test for NULL return from extract_modulename().
Clear grantor field of handler.
libpam/pam_private.h: Add grantor field to struct handler, add handler pointer
parameter to _pam_auditlog().
Tomas Mraz [Mon, 21 Jul 2014 14:31:38 +0000 (16:31 +0200)]
Make pam_pwhistory and pam_unix tolerant of corrupted opasswd file.
* modules/pam_pwhistory/opasswd.c (parse_entry): Test for missing fields
in opasswd entry and return error.
* modules/pam_unix/passverify.c (save_old_password): Test for missing fields
in opasswd entry and skip it.
Dmitry V. Levin [Thu, 8 May 2014 23:36:58 +0000 (23:36 +0000)]
doc: fix install in case of out of tree build (ticket #31)
* doc/adg/Makefile.am (install-data-local, releasedocs): Fall back
to srcdir if documentation files haven't been found in builddir.
(releasedocs): Treat missing documentation files as an error.
* doc/mwg/Makefile.am: Likewise.
* doc/sag/Makefile.am: Likewise.
pam_timestamp uses values of PAM_RUSER and PAM_TTY as components of
the timestamp pathname it creates, so extra care should be taken to
avoid potential directory traversal issues.
* modules/pam_timestamp/pam_timestamp.c (check_tty): Treat
"." and ".." tty values as invalid.
(get_ruser): Treat "." and ".." ruser values, as well as any ruser
value containing '/', as invalid.
Dmitry V. Levin [Tue, 28 Jan 2014 15:01:24 +0000 (15:01 +0000)]
Change tarball name back to "Linux-PAM"
As a side effect of commit Linux-PAM-1_1_8-11-g3fa23ce, tarball name
changed accidentally from "Linux-PAM" to "linux-pam".
This change brings it back to "Linux-PAM".
Dmitry V. Levin [Fri, 24 Jan 2014 15:32:08 +0000 (15:32 +0000)]
Introduce pam_modutil_sanitize_helper_fds
This change introduces pam_modutil_sanitize_helper_fds - a new function
that redirects standard descriptors and closes all other descriptors.
pam_modutil_sanitize_helper_fds supports three types of input and output
redirection:
- PAM_MODUTIL_IGNORE_FD: do not redirect at all.
- PAM_MODUTIL_PIPE_FD: redirect to a pipe. For stdin, it is implemented
by creating a pipe, closing its write end, and redirecting stdin to
its read end. Likewise, for stdout/stderr it is implemented by
creating a pipe, closing its read end, and redirecting to its write
end. Unlike stdin redirection, stdout/stderr redirection to a pipe
has a side effect that a process writing to such descriptor should be
prepared to handle SIGPIPE appropriately.
- PAM_MODUTIL_NULL_FD: redirect to /dev/null. For stdin, it is
implemented via PAM_MODUTIL_PIPE_FD because there is no functional
difference. For stdout/stderr, it is classic redirection to
/dev/null.
PAM_MODUTIL_PIPE_FD is usually more suitable due to linux kernel
security restrictions, but when the helper process might be writing to
the corresponding descriptor and termination of the helper process by
SIGPIPE is not desirable, one should choose PAM_MODUTIL_NULL_FD.
* libpam/pam_modutil_sanitize.c: New file.
* libpam/Makefile.am (libpam_la_SOURCES): Add it.
* libpam/include/security/pam_modutil.h (pam_modutil_redirect_fd,
pam_modutil_sanitize_helper_fds): New declarations.
* libpam/libpam.map (LIBPAM_MODUTIL_1.1.9): New interface.
* modules/pam_exec/pam_exec.c (call_exec): Use
pam_modutil_sanitize_helper_fds.
* modules/pam_mkhomedir/pam_mkhomedir.c (create_homedir): Likewise.
* modules/pam_unix/pam_unix_acct.c (_unix_run_verify_binary): Likewise.
* modules/pam_unix/pam_unix_passwd.c (_unix_run_update_binary):
Likewise.
* modules/pam_unix/support.c (_unix_run_helper_binary): Likewise.
* modules/pam_xauth/pam_xauth.c (run_coprocess): Likewise.
* modules/pam_unix/support.h (MAX_FD_NO): Remove.
Dmitry V. Levin [Fri, 24 Jan 2014 23:53:09 +0000 (23:53 +0000)]
Fix gratuitous use of strdup and x_strdup
There is no need to copy strings passed as arguments to execve,
the only potentially noticeable effect of using strdup/x_strdup
would be a malformed argument list in case of memory allocation error.
Also, x_strdup, being a thin wrapper around strdup, is of no benefit
when its argument is known to be non-NULL, and should not be used in
such cases.
* modules/pam_cracklib/pam_cracklib.c (password_check): Use strdup
instead of x_strdup, the latter is of no benefit in this case.
* modules/pam_ftp/pam_ftp.c (lookup): Likewise.
* modules/pam_userdb/pam_userdb.c (user_lookup): Likewise.
* modules/pam_userdb/pam_userdb.h (x_strdup): Remove.
* modules/pam_mkhomedir/pam_mkhomedir.c (create_homedir): Do not use
x_strdup for strings passed as arguments to execve.
* modules/pam_unix/pam_unix_acct.c (_unix_run_verify_binary): Likewise.
* modules/pam_unix/pam_unix_passwd.c (_unix_run_update_binary): Likewise.
* modules/pam_unix/support.c (_unix_run_helper_binary): Likewise.
(_unix_verify_password): Use strdup instead of x_strdup, the latter
is of no benefit in this case.
* modules/pam_xauth/pam_xauth.c (run_coprocess): Do not use strdup for
strings passed as arguments to execv.
Dmitry V. Levin [Fri, 24 Jan 2014 22:18:32 +0000 (22:18 +0000)]
pam_userdb: fix password hash comparison
Starting with commit Linux-PAM-0-77-28-g0b3e583 that introduced hashed
passwords support in pam_userdb, hashes are compared case-insensitively.
This bug leads to accepting hashes for completely different passwords in
addition to those that should be accepted.
Additionally, commit Linux-PAM-1_1_6-13-ge2a8187 that added support for
modern password hashes with different lengths and settings, did not
update the hash comparison accordingly, which leads to accepting
computed hashes longer than stored hashes when the latter is a prefix
of the former.
* modules/pam_userdb/pam_userdb.c (user_lookup): Reject the computed
hash whose length differs from the stored hash length.
Compare computed and stored hashes case-sensitively.
Fixes CVE-2013-7041.
Dmitry V. Levin [Wed, 22 Jan 2014 02:34:03 +0000 (02:34 +0000)]
libpam_misc: fix an inconsistency in handling memory allocation errors
When misc_conv fails to allocate memory for pam_response array, it
returns PAM_CONV_ERR. However, when read_string fails to allocate
memory for a response string, it loses the response string and silently
ignores the error, with net result as if EOF has been read.
* libpam_misc/misc_conv.c (read_string): Use strdup instead of x_strdup,
the latter is of no benefit in this case.
Do not ignore potential memory allocation errors returned by strdup,
forward them to misc_conv.
Dmitry V. Levin [Mon, 20 Jan 2014 16:24:18 +0000 (16:24 +0000)]
pam_limits: fix utmp->ut_user handling
ut_user member of struct utmp is a string that is not necessarily
null-terminated, so extra care should be taken when using it.
* modules/pam_limits/pam_limits.c (check_logins): Convert ut->UT_USER to
a null-terminated string and consistently use it where a null-terminated
string is expected.
Dmitry V. Levin [Mon, 20 Jan 2014 02:29:41 +0000 (02:29 +0000)]
pam_mkhomedir: check and create home directory for the same user (ticket #22)
Before pam_mkhomedir helper was introduced in commit 7b14630ef39e71f603aeca0c47edf2f384717176, pam_mkhomedir was checking for
existance and creating the same directory - the home directory of the
user NAME returned by pam_get_item(PAM_USER).
The change in behaviour accidentally introduced along with
mkhomedir_helper is not consistent: while the module still checks for
getpwnam(NAME)->pw_dir, the directory created by mkhomedir_helper is
getpwnam(getpwnam(NAME)->pw_name)->pw_dir, which is not necessarily
the same as the directory being checked.
This change brings check and creation back in sync, both handling
getpwnam(NAME)->pw_dir.
* modules/pam_mkhomedir/pam_mkhomedir.c (create_homedir): Replace
"struct passwd *" argument with user's name and home directory.
Pass user's name to MKHOMEDIR_HELPER.
(pam_sm_open_session): Update create_homedir call.
Stéphane Graber [Fri, 17 Jan 2014 23:24:16 +0000 (18:24 -0500)]
pam_loginuid: Always return PAM_IGNORE in userns
The previous patch to support user namespaces works fine with containers
that are started from a desktop/terminal session but fails when dealing
with containers that were started from a remote session such as ssh.
I haven't looked at the exact reason for that in the kernel but on the
userspace side of things, the difference is that containers started from
an ssh session will happily let pam open /proc/self/loginuid read-write,
will let it read its content but will then fail with EPERM when trying
to write to it.
So to make the userns support bullet proof, this commit moves the userns
check earlier in the function (which means a small performance impact as
it'll now happen everytime on kernels that have userns support) and will
set rc = PAM_IGNORE instead of rc = PAM_ERROR.
The rest of the code is still executed in the event that PAM is run on a
future kernel where we have some kind of audit namespace that includes a
working loginuid.
Signed-off-by: Stéphane Graber <stgraber@ubuntu.com> Signed-off-by: Steve Langasek <vorlon@debian.org> Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Steve Langasek [Wed, 15 Jan 2014 03:48:51 +0000 (19:48 -0800)]
pam_namespace: don't use bashisms in default namespace.init script
* modules/pam_namespace/pam_namespace.c: call setuid() before execing the
namespace init script, so that scripts run with maximum privilege regardless
of the shell implementation.
* modules/pam_namespace/namespace.init: drop the '-p' bashism from the
shebang line
This is not a POSIX standard option, it's a bashism. The bash manpage says
that it's used to prevent the effective user id from being reset to the real
user id on startup, and to ignore certain unsafe variables from the
environment.
In the case of pam_namespace, the -p is not necessary for environment
sanitizing because the PAM module (properly) sanitizes the environment
before execing the script.
The stated reason given in CVS history for passing -p is to "preserve euid
when called from setuid apps (su, newrole)." This should be done more
portably, by calling setuid() before spawning the shell.
Signed-off-by: Steve Langasek <vorlon@debian.org>
Bug-Debian: http://bugs.debian.org/624842
Bug-Ubuntu: https://bugs.launchpad.net/bugs/1081323
Stéphane Graber [Wed, 8 Jan 2014 00:12:03 +0000 (16:12 -0800)]
pam_loginuid: Ignore failure in user namespaces
When running pam_loginuid in a container using the user namespaces, even
uid 0 isn't allowed to set the loginuid property.
This change catches the EACCES from opening loginuid, checks if the user
is in the host namespace (by comparing the uid_map with the host's one)
and only if that's the case, sets rc to 1.
Should uid_map not exist or be unreadable for some reason, it'll be
assumed that the process is running on the host's namespace.
The initial reason behind this change was failure to ssh into an
unprivileged container (using a 3.13 kernel and current LXC) when using
a standard pam profile for sshd (which requires success from
pam_loginuid).
I believe this solution doesn't have any drawback and will allow people
to use unprivileged containers normally. An alternative would be to have
all distros set pam_loginuid as optional but that'd be bad for any of
the other potential failure case which people may care about.
There has also been some discussions to get some of the audit features
tied with the user namespaces but currently none of that has been merged
upstream and the currently proposed implementation doesn't cover
loginuid (nor is it clear how this should even work when loginuid is set
as immutable after initial write).
Signed-off-by: Steve Langasek <vorlon@debian.org> Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Dmitry V. Levin [Wed, 8 Jan 2014 23:53:30 +0000 (15:53 -0800)]
pam_loginuid: return PAM_IGNORE when /proc/self/loginuid does not exist
When /proc/self/loginuid does not exist, return PAM_IGNORE instead of
PAM_SUCCESS, so that we can distinguish between "loginuid set
successfully" and "loginuid not set, but this is expected".
Suggested by Steve Langasek.
* modules/pam_loginuid/pam_loginuid.c (set_loginuid): Change return
code semantics: return PAM_SUCCESS on success, PAM_IGNORE when loginuid
does not exist, PAM_SESSION_ERR in case of any other error.
(_pam_loginuid): Forward the PAM error code returned by set_loginuid.
Dmitry V. Levin [Tue, 19 Nov 2013 21:55:40 +0000 (21:55 +0000)]
pam_warn: log flags passed to the module (ticket #25)
* modules/pam_warn/pam_warn.c (log_items): Take "flags" argument and
log it using pam_syslog.
(pam_sm_authenticate, pam_sm_setcred, pam_sm_chauthtok,
pam_sm_acct_mgmt, pam_sm_open_session, pam_sm_close_session): Pass
"flags" argument to log_items.
Thorsten Kukuk [Thu, 29 Aug 2013 12:09:39 +0000 (14:09 +0200)]
Restart waitpid if it returns with EINTR (ticket #17)
* modules/pam_unix/pam_unix_acct.c: run waitpid in a while loop.
* modules/pam_unix/pam_unix_passwd.c: Likewise.
* modules/pam_unix/support.c: Likewise.
projects / linux-pam / log