* src/chage.c: New function: fail_exit(). Change most of the exit()
to a fail_exit, which makes sure the files are unlocked (new global
variables: pw_locked, spw_locked), the PAM transaction is ended, and
the failure is logged to libaudit (use a global user_name and user_uid
for logging).
* src/chage.c: Compilation fix for PAM support (pamh needs to be
global since the function split).
* src/chage.c: Document process_flags(), check_flags(), check_perms(),
open_files(), and close_files().
* src/chage.c: Split update_age() and get_defaults() out of main()
* src/chage.c: Drop the privileges just after opening the files.
* src/chage.c: Do not log to audit only if the user has an entry in
the shadow file.
* NEWS, src/chage.c (open_files): Also open the password file for
writing. This fix chage when the user only has a password entry (and
no shadow entries).
* src/chage.c (get_defaults): Use default values that don't change the
behavior of the account for the fields that are not specified when the
user has no shadow entry.
* src/chage.c: Before pam_end(), the return value of the previous
pam API was already checked. No need to validate it again.
* src/chage.c: main() split in new functions: process_flags(),
check_flags(), check_perms(), open_files(), and close_files().
* libmisc/copydir.c: Split copy_tree() in more maintainable functions:
copy_entry(), copy_dir(), copy_symlink(), copy_hardlink(),
copy_special(), and copy_file().
* libmisc/copydir.c: -1 is used to indicate an error, directly set err
to -1, instead of incrementing it, and checking if not nul at the
end.
* NEWS, libmisc/setupenv.c: Export PATH according to ENV_PATH and
ENV_SUPATH, as for su. This impacts login.
* man/login.1.xml: PATH and SUPATH are now used both when PAM support
is disabled and enabled.
* libmisc/Makefile.am, lib/prototypes.h, libmisc/yesno.c, src/grpck.c,
src/pwck.c: move yes_or_no() from grpck/pwck to a separate
libmisc/yesno.c (with a read_only argument).
* libmisc/fields.c, libmisc/yesno.c: Make sure stdout is flushed before
reading the user's answer.
* NEWS: Document that usermod will now preserve user's file modification
and access time.
* libmisc/copydir.c: Preserve the access and modification time of copied
files. This is important for usermod. This will also impact useradd, for
the skeleton files, but this is not important.
* libmisc/copydir.c: Stop and return an error if a file could not be
closed after during a copy.
Merge RedHat's patch shadow-4.0.18.1-findNewUidOnce.patch:
* src/useradd.c (usr_update): Do not call find_new_uid(). The UID was
already either specified or found by another call to find_new_uid().
* src/useradd.c (find_new_uid): Always start with uid_min (find_new_uid()
is never called when user_id was already specified).
* src/useradd.c (find_new_uid): Fix the comments (find_new_uid() is not
called when the UID is specified (uflg)).
* src/useradd.c (main): Only call find_new_uid() if (!oflg) and (!uflg).
If uflg is set (but not oflg), check the UID uniqueness.
* src/useradd.c (find_new_uid): Don't check the uid and user name
uniqueness in find_new_uid(). The user name uniqueness is already checked
during the parameter validation. UID uniqueness is also checked (see
above).
* src/useradd.c (find_new_uid): Don't check uflg in find_new_uid().
* src/useradd.c (find_new_uid): Make sure that find_new_uid() is not
called when uflg is set (assert).
Cleanups in find_new_gid:
* src/useradd.c (find_new_gid): Check that gflg is not set (assert).
* src/useradd.c (find_new_gid): Do not check the group name uniqueness
(already checked in main).
* src/useradd.c (find_new_gid): Avoid a "continue" in the loop.
* src/useradd.c (find_new_gid): Remove irrelevant comments.
* src/useradd.c (find_new_gid): Fix the function definition's comment.
* man/chpasswd.8.xml, man/chgpasswd.8.xml: Document the NONE crypt
method.
* man/login.defs.d/MAIL_DIR.xml: Add comment regarding useradd not
using MAIL_FILE.
* man/login.defs.d/ERASECHAR.xml, man/login.defs.d/KILLCHAR.xml,
man/login.defs.d/CONSOLE_GROUPS.xml, man/login.defs.d/ENV_HZ.xml,
man/login.defs.d/ENV_PATH.xml, man/login.defs.d/ENV_SUPATH.xml:
These variables are also used by some tools when compiled with PAM
support.
* man/login.defs.d/ENV_HZ.xml: Add note that it is only used by
sulogin when compiled with PAM support.
* man/login.defs.d/ENV_SUPATH.xml: Typos: ENV_PATH -> ENV_SUPATH,
and mention sbin in the path.
* man/login.defs.d/LOGIN_STRING.xml: Fix typo: confition ->
condition.
* man/sg.1.xml: Add CONFIGURATION section (SYSLOG_SG_ENAB).
* man/su.1.xml: ENV_HZ, LOGIN_STRING, MAIL_DIR, USERGROUPS_ENAB
are only used when su is compiled without PAM support.
* man/login.defs.5.xml: Added variables: OBSCURE_CHECKS_ENAB
PASS_ALWAYS_WARN PASS_CHANGE_TRIES SULOG_FILE SU_NAME
SU_WHEEL_ONLY SYSLOG_SG_ENAB SYSLOG_SU_ENAB.
* man/login.defs.5.xml: ENVIRON_FILE is only used when compiled
without PAM support.
* man/login.defs.5.xml: sulogin uses variables even when compiled
with PAM support.
* man/login.1.xml: ENV_HZ ENV_PATH ENV_SUPATH MAIL_DIR UMASK are
only used when login is not compiled with PAM support.
man/po/POTFILES.in, man/Makefile.am: Generate the PO files for the
manpages in the man/po directory (instead of man/<lang>). Use a
Makefile.in.in based on gettext's one. This ensure that the PO are
generated before being used in the <lang> directories.
* man/generate_mans.mak, man/generate_translations.mak,
man/Makefile.am: New makefile for the generation of manpages from
XML (generate_mans.mak). This avoid duplicate chunks in
generate_translations.mak and Makefile.am
* man/de/de.po, man/fr/fr.po, man/it/it.po, man/pl/pl.po,
man/ru/ru.po, man/sv/sv.po: Moved to...
* man/po/de.po, man/po/fr.po, man/po/it.po, man/po/pl.po,
man/po/ru.po, man/po/sv.po: ... here.
default. Keeping the feature enabled is safe. Disabling it permits
to disable the references to the SHA256 and SHA512 password
encryption algorithms from the usage help and manuals (in addition
to the support for these algorithms in the code).
* libmisc/obscure.c, libmisc/salt.c, src/newusers.c,
src/chpasswd.c, src/chgpasswd.c, src/passwd.c: ENCRYPT_METHOD is
always supported in login.defs. Remove the ENCRYPTMETHOD_SELECT
preprocessor condition.
* libmisc/obscure.c, libmisc/salt.c, src/newusers.c,
src/chpasswd.c, src/chgpasswd.c, src/passwd.c: Disable SHA256 and
SHA512 if USE_SHA_CRYPT is not defined (this corresponds to a
subset of the ENCRYPTMETHOD_SELECT sections).
password longer than 13 chars. This protects against the GNU crypt() which
does not return NULL if the algorithm is not supported, and return a DES
encrypted password.
Thanks to Dan Kopecek <dkopecek@redhat.com>.
* src/chpasswd.c, src/chgpasswd.c: Do not use DES by default, but
the system default define in /Etc/login.defs. Thanks to Dan
Kopecek <dkopecek@redhat.com>.
* NEWS, man/chpasswd.8.xml, man/chgpasswd.8.xml: Do not mention
DES as the default algorithm.
* src/chpasswd.c, src/chgpasswd.c: Tag the ENCRYPTMETHOD_SELECT
dependent code accordingly.
* src/chpasswd.c: Added --sha-rounds to the usage().
* libmisc/Makefile.am, libmisc/getlong.c, src/chgpasswd.c,
src/chpasswd.c: New getlong function. Replace chpasswd's and
chgpasswd's getnumber.
users. (was sometimes <emphasis remap='I'>)
* Use <option> vor the variable names. This
makes the manpage much more readable.
* (ENCRYPT_METHOD, MD5_CRYPT_ENAB,
SHA_CRYPT_MIN_ROUNDS, SHA_CRYPT_MAX_ROUNDS): Mention that command
line option may supersede the system setting.
* Document the variables used by chpasswd
and chgpasswd.
variable: MAX_MEMBERS_PER_GROUP. Used for the split groups support.
* lib/commonio.c, lib/commonio.h: Add an open_hook and close_hook
operation. They are called after the database is actually opened
and parse, or before it is closed.
* lib/groupio.c: Add an open_hook to merge split groups, and an
close group to split groups if MAX_MEMBERS_PER_GROUP is set.
This fixes gpasswd and chgpasswd when split groups are used.
* lib/sgroupio.c, lib/shadowio.c, lib/pwio.c: No open or close
hooks for these databases. (unsure about what should be the gshadow
behavior for split groups)
gr_locate and sgr_locate. gpasswd write in the file database. Thus
it should read information from the file database, not using
getgrnam. The change to sgr_locate is just for consistency. This
requires opening the group databases (read only) using
gr_open/sgr_open.
* NEWS: Indicate that manpages should be re-generated if configure
option are changed, due to conditions.
generation of manpages.
* man/generate_translations.mak: Added pam/no_pam condition (like
in man/Makefile.am).
* man/Makefile.am, man/generate_translations.mak: Added
gshadow/no_gshadow condition.
* man/gpasswd.1.xml: Use the gshadow/no_gshadow condition to
change the manpage depending on the shadow group support.
and /etc/passwd to section FILES.
* man/newusers.8.xml: Mentions that PAM is not used to set the
passwords.
* man/chpasswd.8.xml: Added section FILES (/etc/passwd,
/etc/shadow, /etc/login.defs).
* man/chpasswd.8.xml: Use the same paragraph as in newusers.8.xml
to indicate that PAM is not used.
* man/chgpasswd.8.xml: Added section FILES (/etc/group,
/etc/gshadow, /etc/login.defs).
This was not the case when the GID is not specified, and a GID
exist with an ID higher than the all the UIDs.
* Typo in comment: contrained -> constrained.
SHA256, and SHA512 exactly (not only the first 3/6 chars).
* libmisc/salt.c (SHA_salt_rounds): Set rounds to the specified
prefered_rounds value, if specified.
* src/gpasswd.c, libmisc/salt.c: Fix compilation warnings (use
size_t for lengths).
* src/chpasswd.c, src/chgpasswd.c: Add missing parenthesis.
encryption algorithm is chosen for the passwords. Document the new
-c and -s options. Add a reference to login.defs(5).
* man/login.defs.5.xml: Document the ENCRYPT_METHOD,
MD5_CRYPT_ENAB, SHA_CRYPT_MIN_ROUNDS, and SHA_CRYPT_MAX_ROUNDS
variables.
* etc/login.defs: Indicate that MD5_CRYPT_ENAB is deprecated.
Document the relationship with PAM for MD5_CRYPT_ENAB and
ENCRYPT_METHOD.
crypt_make_salt to force the crypt method and number of rounds.
* libmisc/salt.c: Add parameter to SHA_salt_rounds to force the
number of rounds.
* libmisc/salt.c, lib/getdef.c: ENCRYPT_METHOD and MD5_CRYPT_ENAB
are needed also when USE_PAM (e.g. for chpasswd).
* src/newusers.c, src/gpasswd.c: Use the new crypt_make_salt prototype.
* src/chpasswd.c, src/chgpasswd.c: Add option -c, --crypt-method
and -s, --sha-rounds to specify the crypt method and number of
rounds in case of one of the SHA methods. The new prototype of
crypt_make_salt simplifies the handling of -m, --md5.
bytes).
* lib/getdef.c, etc/login.defs: Add definitions for
SHA_CRYPT_MIN_ROUNDS and SHA_CRYPT_MAX_ROUNDS.
* libmisc/salt.c: Use SHA_CRYPT_MIN_ROUNDS and SHA_CRYPT_MAX_ROUNDS
to add a random number of rounds if needed.
is then used with strcat).
* libmisc/salt.c (crypt_make_salt): Initialize result[0] to nul at
the beginning (was not initialized when USE_PAM).
* libmisc/salt.c (crypt_make_salt): Check that ENCRYPT_METHOD is a
valid crypt method.
shadow-4.0.18.1-sha256.patch. Thanks to Peter Vrabec. Hardly no changes
except re-indent and changes related to recent modifications (max_salt_len
in crypt_make_salt). Changes in lib/defines.h not applied (definition of
ENCRYPTMETHOD_SELECT). I will add a configure check or flag.
* src/login.c: "dereferencing type-punned pointer will break
strict-aliasing rules", add a variable indirection: ptr_pam_user.
* lib/commonio.c: do not initialize the sb stat structure.
* lib/pwio.c, lib/shadowio.c, lib/sgroupio.c, lib/groupio.c:
initialize the security context if WITH_SELINUX.
* lib/nscd.c: The service argument is not const (used in the exec*
parameters). This matches with the prototype definition.
* src/groupmems.c: Avoid ++i when i is also used in the same line.
* src/newusers.c: i is positive every time it is compared. Add
cast to unsigned int.
* src/nologin.c: Use a main() prototype with no arguments.
* libmisc/getdate.y: Initialize the type and value fields of the
terminating entry for each TABLE.
* libmisc/tz.c: Use "TZ=CST6CDT" as the default timezone.
libmisc/xgetXXbyYY.c, libmisc/xgetpwnam.c, libmisc/xgetpwuid.c,
libmisc/xgetgrnam.c, libmisc/xgetgrgid.c, libmisc/xgetspnam.c:
Added functions xgetpwnam(), xgetpwuid(), xgetgrnam(),
xgetgrgid(), and xgetspnam(). They allocate memory for the
returned structure and are more robust to successive calls. They
are implemented with the libc's getxxyyy_r() functions if
available.
* libmisc/limits.c, libmisc/entry.c, libmisc/chowntty.c,
libmisc/addgrps.c, libmisc/myname.c, libmisc/rlogin.c,
libmisc/pwdcheck.c, src/newgrp.c, src/login_nopam.c,
src/userdel.c, src/lastlog.c, src/grpck.c, src/gpasswd.c,
src/newusers.c, src/chpasswd.c, src/chfn.c, src/groupmems.c,
src/usermod.c, src/expiry.c, src/groupdel.c, src/chgpasswd.c,
src/su.c, src/useradd.c, src/groupmod.c, src/passwd.c, src/pwck.c,
src/groupadd.c, src/chage.c, src/login.c, src/suauth.c,
src/faillog.c, src/groups.c, src/chsh.c, src/id.c: Review all the
usage of one of the getpwnam(), getpwuid(), getgrnam(),
getgrgid(), and getspnam() functions. It was noticed on
http://bugs.debian.org/341230 that chfn and chsh use a passwd
structure after calling a pam function, which result in using
information from the passwd structure requested by pam, not the
original one. It is much easier to use the new xget... functions
to avoid these issues. I've checked which call to the original
get... functions could be left (reducing the scope of the
structure if possible), and I've left comments to ease future
reviews (e.g. /* local, no need for xgetpwnam */).
Note: the getpwent/getgrent calls should probably be checked also.
* src/groupdel.c, src/expiry.c: Fix typos in comments.
* src/groupmod.c: Re-indent.
* libmisc/Makefile.am, lib/groupmem.c, lib/groupio.c, lib/pwmem.c,
lib/pwio.c, lib/shadowmem.c, lib/shadowio.c: Move the __<xx>_dup
functions (used by the xget... functions) from the <xx>io.c files
to the new <xx>mem.c files. This avoid linking some utils against
the SELinux library.
cleanup(2) by spw_unlock and remove the calls to cleanup(1).
* src/chage.c: Remove variable pwrw. It is always set to 0. The
password database is always read only.
generated translated manpages (if ENABLE_REGENERATE_MAN).
* man/Makefile.am: Removed rules for all the generated translated
manpages.
* man/sv/Makefile.am, man/de/Makefile.am, man/fr/Makefile.am,
man/pl/Makefile.am, man/ru/Makefile.am, man/it/Makefile.am:
Include generate_translations.mak to handle the generated
translations (XML and roff files).
* man/Makefile.am: Translated XML files moved from the CLEANFILES
variable of man/Makefile.am to the various languages Makefiles.
* man/useradd.8.xml: Indicate that the NIS caveats is also valid
for any external database as LDAP.
* man/groupadd.8.xml: Likewise.
* man/groupadd.8.xml: Reorder and reformat the caveats bullets.
glibc socket to flush the nscd tables. This comes from the RedHat
patch shadow-4.0.16-nscd.c.
* lib/commonio.c: Forbid inheritance of the passwd and group files
to the spawed processes (like nscd). This comes from the RedHat
patch shadow-4.0.17-notInheritFd.patch.
* lib/nscd.h: Update header.
spw_locked, gr_locked, and sgr_locked to indicate which files must
be unlocked.
* src/usermod.c (open_files, close_files): Open and close the
group files as well as the passwd files. This permit to check if
the group files modification are allowed before writing the passwd
files.
* src/usermod.c (grp_update, update_gshadow, update_group): Do not
return a status code, but call fail_exit() in case of error. The
group files are no more opened and closed in update_gshadow() and
update_group().
* src/usermod.c (main): move the call to grp_update between
open_files and close_files.
* src/usermod.c: Differentiate failure to add a group entry and
failure to add a shadow group entry.
* NEWS, lib/commonio.c (commonio_update): When an entry is updated, make
sure that there are no other entry with the same name. This fixes
an infinite loop in userdel and usermod when an (erroneous) group
file contains two entries with the same name.
(https://bugzilla.redhat.com/show_bug.cgi?id=240915)
right place (either 8th, or 11th position).
* NEWS, src/chgpasswd.c, src/chpasswd.c: The protocol + salt does
not need 15 chars. No need for a temporary buffer.
This change the fix committed on 2007-11-10. The salt provided to
pw_encrypt could have been too long.
