872 lines
23 KiB
C
872 lines
23 KiB
C
/*
|
|
* Copyright (c) 1997-2003 Erez Zadok
|
|
* Copyright (c) 1989 Jan-Simon Pendry
|
|
* Copyright (c) 1989 Imperial College of Science, Technology & Medicine
|
|
* Copyright (c) 1989 The Regents of the University of California.
|
|
* All rights reserved.
|
|
*
|
|
* This code is derived from software contributed to Berkeley by
|
|
* Jan-Simon Pendry at Imperial College, London.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
* 3. All advertising materials mentioning features or use of this software
|
|
* must display the following acknowledgment:
|
|
* This product includes software developed by the University of
|
|
* California, Berkeley and its contributors.
|
|
* 4. Neither the name of the University nor the names of its contributors
|
|
* may be used to endorse or promote products derived from this software
|
|
* without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
* SUCH DAMAGE.
|
|
*
|
|
* %W% (Berkeley) %G%
|
|
*
|
|
* $Id: homedir.c,v 1.5.2.10 2002/12/27 22:45:08 ezk Exp $
|
|
* $FreeBSD$
|
|
*
|
|
* HLFSD was written at Columbia University Computer Science Department, by
|
|
* Erez Zadok <ezk@cs.columbia.edu> and Alexander Dupuy <dupuy@cs.columbia.edu>
|
|
* It is being distributed under the same terms and conditions as amd does.
|
|
*/
|
|
|
|
#ifdef HAVE_CONFIG_H
|
|
# include <config.h>
|
|
#endif /* HAVE_CONFIG_H */
|
|
#include <am_defs.h>
|
|
#include <hlfsd.h>
|
|
|
|
|
|
/*
|
|
* STATIC VARIABLES AND FUNCTIONS:
|
|
*/
|
|
static FILE *passwd_fp = NULL;
|
|
static char pw_name[16], pw_dir[128];
|
|
static int cur_pwtab_num = 0, max_pwtab_num = 0;
|
|
static int hlfsd_diskspace(char *);
|
|
static int hlfsd_stat(char *, struct stat *);
|
|
static int passwd_line = 0;
|
|
static int plt_reset(void);
|
|
static struct passwd passwd_ent;
|
|
static uid2home_t *lastchild;
|
|
static uid2home_t *pwtab;
|
|
static void delay(uid2home_t *, int);
|
|
static void table_add(int, const char *, const char *);
|
|
static char mboxfile[MAXPATHLEN];
|
|
static char *root_home; /* root's home directory */
|
|
|
|
/* GLOBAL FUNCTIONS */
|
|
char *homeof(char *username);
|
|
int uidof(char *username);
|
|
|
|
/* GLOBALS VARIABLES */
|
|
username2uid_t *untab; /* user name table */
|
|
|
|
/*
|
|
* Return the home directory pathname for the user with uid "userid".
|
|
*/
|
|
char *
|
|
homedir(int userid, int groupid)
|
|
{
|
|
static char linkval[MAXPATHLEN + 1];
|
|
static struct timeval tp;
|
|
uid2home_t *found;
|
|
char *homename;
|
|
struct stat homestat;
|
|
int old_groupid, old_userid;
|
|
|
|
clock_valid = 0; /* invalidate logging clock */
|
|
|
|
if ((found = plt_search(userid)) == (uid2home_t *) NULL) {
|
|
return alt_spooldir; /* use alt spool for unknown uid */
|
|
}
|
|
homename = found->home;
|
|
|
|
if (homename[0] != '/' || homename[1] == '\0') {
|
|
found->last_status = 1;
|
|
return alt_spooldir; /* use alt spool for / or rel. home */
|
|
}
|
|
if ((int) userid == 0) /* force all uid 0 to use root's home */
|
|
sprintf(linkval, "%s/%s", root_home, home_subdir);
|
|
else
|
|
sprintf(linkval, "%s/%s", homename, home_subdir);
|
|
|
|
if (noverify) {
|
|
found->last_status = 0;
|
|
return linkval;
|
|
}
|
|
|
|
/*
|
|
* To optimize hlfsd, we don't actually check the validity of the
|
|
* symlink if it has been checked in the last N seconds. It is
|
|
* very likely that the link, machine, and filesystem are still
|
|
* valid, as long as N is small. But if N is large, that may not be
|
|
* true. That's why the default N is 5 minutes, but we allow the
|
|
* user to override this value via a command line option. Note that
|
|
* we do not update the last_access_time each time it is accessed,
|
|
* but only once every N seconds.
|
|
*/
|
|
if (gettimeofday(&tp, (struct timezone *) NULL) < 0) {
|
|
tp.tv_sec = 0;
|
|
} else {
|
|
if ((tp.tv_sec - found->last_access_time) < cache_interval) {
|
|
if (found->last_status == 0) {
|
|
return linkval;
|
|
} else {
|
|
return alt_spooldir;
|
|
}
|
|
} else {
|
|
found->last_access_time = tp.tv_sec;
|
|
}
|
|
}
|
|
|
|
#ifdef DEBUG
|
|
/*
|
|
* only run this forking code if asked for -D fork
|
|
* or if did not ask for -D nofork
|
|
*/
|
|
amuDebug(D_FORK) {
|
|
#endif /* DEBUG */
|
|
/* fork child to process request if none in progress */
|
|
if (found->child && kill(found->child, 0) < 0)
|
|
found->child = 0;
|
|
|
|
if (found->child)
|
|
delay(found, 5); /* wait a bit if in progress */
|
|
|
|
#if defined(DEBUG) && defined(HAVE_WAITPID)
|
|
if (found->child) {
|
|
/* perhaps it's a child we lost count of? let's wait on it */
|
|
int status, child;
|
|
if ((child = waitpid((pid_t) found->child, &status, WNOHANG)) > 0) {
|
|
plog(XLOG_ERROR, "found lost child %d", child);
|
|
found->child = 0;
|
|
if (WIFEXITED(status))
|
|
found->last_status = WEXITSTATUS(status);
|
|
else if (WIFSIGNALED(status))
|
|
found->last_status = -WTERMSIG(status);
|
|
else {
|
|
plog(XLOG_ERROR, "unknown child exit status (%d) ???", status);
|
|
found->last_status = 255;
|
|
}
|
|
}
|
|
}
|
|
#endif /* DEBUG && HAVE_WAITPID */
|
|
|
|
if (found->child) {
|
|
found->last_status = 1; /* better safe than sorry - maybe */
|
|
return alt_spooldir;
|
|
}
|
|
if ((found->child = fork()) < 0) {
|
|
found->last_status = 1;
|
|
return alt_spooldir;
|
|
}
|
|
if (found->child) { /* PARENT */
|
|
#ifdef DEBUG
|
|
if (lastchild)
|
|
plog(XLOG_INFO, "cache spill uid = %ld, pid = %ld, home = %s",
|
|
(long) lastchild->uid, (long) lastchild->child,
|
|
lastchild->home);
|
|
#endif /* DEBUG */
|
|
lastchild = found;
|
|
return (char *) NULL; /* return NULL to parent, so it can continue */
|
|
}
|
|
#ifdef DEBUG
|
|
} /* end of Debug(D_FORK) */
|
|
#endif /* DEBUG */
|
|
|
|
/*
|
|
* CHILD: (or parent if -D nofork)
|
|
*
|
|
* Check and create dir if needed.
|
|
* Check disk space and/or quotas too.
|
|
*
|
|
* We don't need to set the _last_status field of found after the fork
|
|
* in the child, b/c that information would be later determined in
|
|
* nfsproc_readlink_2() and the correct exit status would be returned
|
|
* to the parent upon SIGCHLD in interlock().
|
|
*
|
|
*/
|
|
am_set_mypid(); /* for logging routines */
|
|
if ((old_groupid = setgid(groupid)) < 0) {
|
|
plog(XLOG_WARNING, "could not setgid to %d: %m", groupid);
|
|
return linkval;
|
|
}
|
|
if ((old_userid = seteuid(userid)) < 0) {
|
|
plog(XLOG_WARNING, "could not seteuid to %d: %m", userid);
|
|
setgid(old_groupid);
|
|
return linkval;
|
|
}
|
|
if (hlfsd_stat(linkval, &homestat) < 0) {
|
|
if (errno == ENOENT) { /* make the spool dir if possible */
|
|
/* don't use recursive mkdirs here */
|
|
if (mkdir(linkval, PERS_SPOOLMODE) < 0) {
|
|
seteuid(old_userid);
|
|
setgid(old_groupid);
|
|
plog(XLOG_WARNING, "can't make directory %s: %m", linkval);
|
|
return alt_spooldir;
|
|
}
|
|
/* fall through to testing the disk space / quota */
|
|
} else { /* the home dir itself must not exist then */
|
|
seteuid(old_userid);
|
|
setgid(old_groupid);
|
|
plog(XLOG_WARNING, "bad link to %s: %m", linkval);
|
|
return alt_spooldir;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* If gets here, then either the spool dir in the home dir exists,
|
|
* or it was just created. In either case, we now need to
|
|
* test if we can create a small file and write at least one
|
|
* byte into it. This will test that we have both enough inodes
|
|
* and disk blocks to spare, or they fall within the user's quotas too.
|
|
* We are still seteuid to the user at this point.
|
|
*/
|
|
if (hlfsd_diskspace(linkval) < 0) {
|
|
seteuid(old_userid);
|
|
setgid(old_groupid);
|
|
plog(XLOG_WARNING, "no more space in %s: %m", linkval);
|
|
return alt_spooldir;
|
|
} else {
|
|
seteuid(old_userid);
|
|
setgid(old_groupid);
|
|
return linkval;
|
|
}
|
|
}
|
|
|
|
|
|
static int
|
|
hlfsd_diskspace(char *path)
|
|
{
|
|
char buf[MAXPATHLEN];
|
|
int fd, len;
|
|
|
|
clock_valid = 0; /* invalidate logging clock */
|
|
|
|
sprintf(buf, "%s/._hlfstmp_%lu", path, (long) getpid());
|
|
if ((fd = open(buf, O_RDWR | O_CREAT, 0600)) < 0) {
|
|
plog(XLOG_ERROR, "cannot open %s: %m", buf);
|
|
return -1;
|
|
}
|
|
len = strlen(buf);
|
|
if (write(fd, buf, len) < len) {
|
|
plog(XLOG_ERROR, "cannot write \"%s\" (%d bytes) to %s : %m", buf, len, buf);
|
|
close(fd);
|
|
unlink(buf); /* cleanup just in case */
|
|
return -1;
|
|
}
|
|
if (unlink(buf) < 0) {
|
|
plog(XLOG_ERROR, "cannot unlink %s : %m", buf);
|
|
}
|
|
close(fd);
|
|
return 0;
|
|
}
|
|
|
|
|
|
static int
|
|
hlfsd_stat(char *path, struct stat *statp)
|
|
{
|
|
if (stat(path, statp) < 0)
|
|
return -1;
|
|
else if (!S_ISDIR(statp->st_mode)) {
|
|
errno = ENOTDIR;
|
|
return -1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
|
|
static void
|
|
delay(uid2home_t *found, int secs)
|
|
{
|
|
struct timeval tv;
|
|
|
|
#ifdef DEBUG
|
|
if (found)
|
|
dlog("delaying on child %ld for %d seconds", (long) found->child, secs);
|
|
#endif /* DEBUG */
|
|
|
|
tv.tv_usec = 0;
|
|
|
|
do {
|
|
tv.tv_sec = secs;
|
|
if (select(0, 0, 0, 0, &tv) == 0)
|
|
break;
|
|
} while (--secs && found->child);
|
|
}
|
|
|
|
|
|
/*
|
|
* This function is called when a child has terminated after
|
|
* servicing an nfs request. We need to check the exit status and
|
|
* update the last_status field of the requesting user.
|
|
*/
|
|
RETSIGTYPE
|
|
interlock(int signum)
|
|
{
|
|
int child;
|
|
uid2home_t *lostchild;
|
|
int status;
|
|
int max_errors = 10; /* avoid infinite loops */
|
|
|
|
#ifdef HAVE_WAITPID
|
|
while ((child = waitpid((pid_t) -1, &status, WNOHANG)) != 0) {
|
|
#else /* not HAVE_WAITPID */
|
|
while ((child = wait3(&status, WNOHANG, (struct rusage *) 0)) != 0) {
|
|
#endif /* not HAVE_WAITPID */
|
|
|
|
if (child < 0) {
|
|
plog(XLOG_WARNING, "waitpid/wait3: %m");
|
|
if (--max_errors > 0)
|
|
continue;
|
|
else
|
|
break;
|
|
}
|
|
|
|
/* high chances this was the last child forked */
|
|
if (lastchild && lastchild->child == child) {
|
|
lastchild->child = 0;
|
|
|
|
if (WIFEXITED(status))
|
|
lastchild->last_status = WEXITSTATUS(status);
|
|
else if (WIFSIGNALED(status))
|
|
lastchild->last_status = -WTERMSIG(status);
|
|
else {
|
|
plog(XLOG_ERROR, "unknown child exit status (%d) ???", status);
|
|
lastchild->last_status = 255;
|
|
}
|
|
lastchild = (uid2home_t *) NULL;
|
|
} else {
|
|
/* and if not, we have to search for it... */
|
|
int found = 0;
|
|
for (lostchild = pwtab; lostchild < &pwtab[cur_pwtab_num]; lostchild++) {
|
|
if (lostchild->child == child) {
|
|
lostchild->child = 0;
|
|
if (WIFEXITED(status))
|
|
lostchild->last_status = WEXITSTATUS(status);
|
|
else if (WIFSIGNALED(status))
|
|
lostchild->last_status = -WTERMSIG(status);
|
|
else {
|
|
plog(XLOG_ERROR, "unknown child exit status (%d) ???", status);
|
|
lostchild->last_status = 255;
|
|
}
|
|
found = 1;
|
|
break;
|
|
}
|
|
}
|
|
if (!found)
|
|
plog(XLOG_ERROR, "no record of child %d found???", child);
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* PASSWORD AND USERNAME LOOKUP TABLES FUNCTIONS
|
|
*/
|
|
|
|
/*
|
|
* get index of UserName table entry which matches username.
|
|
* must not return uid_t because we want to return a negative number.
|
|
*/
|
|
int
|
|
untab_index(char *username)
|
|
{
|
|
int max, min, mid, cmp;
|
|
|
|
max = cur_pwtab_num - 1;
|
|
min = 0;
|
|
|
|
do {
|
|
mid = (max + min) / 2;
|
|
cmp = strcmp(untab[mid].username, username);
|
|
if (cmp == 0) /* record found! */
|
|
return mid;
|
|
if (cmp > 0)
|
|
max = mid;
|
|
else
|
|
min = mid;
|
|
} while (max > min + 1);
|
|
|
|
if (STREQ(untab[max].username, username))
|
|
return max;
|
|
if (STREQ(untab[min].username, username))
|
|
return min;
|
|
|
|
/* if gets here then record was not found */
|
|
return -1;
|
|
}
|
|
|
|
|
|
/*
|
|
* Don't make this return a uid_t, because we need to return negative
|
|
* numbers as well (error codes.)
|
|
*/
|
|
int
|
|
uidof(char *username)
|
|
{
|
|
int idx;
|
|
|
|
if ((idx = untab_index(username)) < 0) /* not found */
|
|
return INVALIDID; /* an invalid user id */
|
|
return untab[idx].uid;
|
|
}
|
|
|
|
|
|
/*
|
|
* Don't make this return a uid_t, because we need to return negative
|
|
* numbers as well (error codes.)
|
|
*/
|
|
char *
|
|
homeof(char *username)
|
|
{
|
|
int idx;
|
|
|
|
if ((idx = untab_index(username)) < 0) /* not found */
|
|
return (char *) NULL; /* an invalid user id */
|
|
return untab[idx].home;
|
|
}
|
|
|
|
|
|
char *
|
|
mailbox(int uid, char *username)
|
|
{
|
|
char *home;
|
|
|
|
if (uid < 0)
|
|
return (char *) NULL; /* not found */
|
|
|
|
if ((home = homeof(username)) == (char *) NULL)
|
|
return (char *) NULL;
|
|
if (STREQ(home, "/"))
|
|
sprintf(mboxfile, "/%s/%s", home_subdir, username);
|
|
else
|
|
sprintf(mboxfile, "%s/%s/%s", home, home_subdir, username);
|
|
return mboxfile;
|
|
}
|
|
|
|
|
|
static int
|
|
plt_compare_fxn(const voidp x, const voidp y)
|
|
{
|
|
uid2home_t *i = (uid2home_t *) x;
|
|
uid2home_t *j = (uid2home_t *) y;
|
|
|
|
return i->uid - j->uid;
|
|
}
|
|
|
|
|
|
static int
|
|
unt_compare_fxn(const voidp x, const voidp y)
|
|
{
|
|
username2uid_t *i = (username2uid_t *) x;
|
|
username2uid_t *j = (username2uid_t *) y;
|
|
|
|
return strcmp(i->username, j->username);
|
|
}
|
|
|
|
|
|
/* perform initialization of user passwd database */
|
|
static void
|
|
hlfsd_setpwent(void)
|
|
{
|
|
if (!passwdfile) {
|
|
setpwent();
|
|
return;
|
|
}
|
|
|
|
passwd_fp = fopen(passwdfile, "r");
|
|
if (!passwd_fp) {
|
|
plog(XLOG_ERROR, "unable to read passwd file %s: %m", passwdfile);
|
|
return;
|
|
}
|
|
plog(XLOG_INFO, "reading password entries from file %s", passwdfile);
|
|
|
|
passwd_line = 0;
|
|
memset((char *) &passwd_ent, 0, sizeof(struct passwd));
|
|
passwd_ent.pw_name = (char *) &pw_name;
|
|
passwd_ent.pw_dir = (char *) &pw_dir;
|
|
}
|
|
|
|
|
|
/* perform de-initialization of user passwd database */
|
|
static void
|
|
hlfsd_endpwent(void)
|
|
{
|
|
if (!passwdfile) {
|
|
/*
|
|
* Don't actually run this because we will be making more passwd calls
|
|
* afterwards. On Solaris 2.5.1, making getpwent() calls after calling
|
|
* endpwent() results in a memory leak! (and no, even Purify didn't
|
|
* detect it...)
|
|
*
|
|
endpwent();
|
|
*/
|
|
return;
|
|
}
|
|
|
|
if (passwd_fp) {
|
|
fclose(passwd_fp);
|
|
}
|
|
}
|
|
|
|
|
|
/* perform record reading/parsing of individual passwd database records */
|
|
static struct passwd *
|
|
hlfsd_getpwent(void)
|
|
{
|
|
char buf[256], *cp;
|
|
|
|
/* check if to perform standard unix function */
|
|
if (!passwdfile) {
|
|
return getpwent();
|
|
}
|
|
|
|
clock_valid = 0; /* invalidate logging clock */
|
|
|
|
/* return here to read another entry */
|
|
readent:
|
|
|
|
/* return NULL if reached end of file */
|
|
if (feof(passwd_fp))
|
|
return NULL;
|
|
|
|
pw_name[0] = pw_dir[0] = '\0';
|
|
|
|
/* read records */
|
|
buf[0] = '\0';
|
|
fgets(buf, 256, passwd_fp);
|
|
passwd_line++;
|
|
if (!buf || buf[0] == '\0')
|
|
goto readent;
|
|
|
|
/* read user name */
|
|
cp = strtok(buf, ":");
|
|
if (!cp || cp[0] == '\0') {
|
|
plog(XLOG_ERROR, "no user name on line %d of %s", passwd_line, passwdfile);
|
|
goto readent;
|
|
}
|
|
strcpy(pw_name, cp); /* will show up in passwd_ent.pw_name */
|
|
|
|
/* skip passwd */
|
|
strtok(NULL, ":");
|
|
|
|
/* read uid */
|
|
cp = strtok(NULL, ":");
|
|
if (!cp || cp[0] == '\0') {
|
|
plog(XLOG_ERROR, "no uid on line %d of %s", passwd_line, passwdfile);
|
|
goto readent;
|
|
}
|
|
passwd_ent.pw_uid = atoi(cp);
|
|
|
|
/* skip gid and gcos */
|
|
strtok(NULL, ":");
|
|
strtok(NULL, ":");
|
|
|
|
/* read home dir */
|
|
cp = strtok(NULL, ":");
|
|
if (!cp || cp[0] == '\0') {
|
|
plog(XLOG_ERROR, "no home dir on line %d of %s", passwd_line, passwdfile);
|
|
goto readent;
|
|
}
|
|
strcpy(pw_dir, cp); /* will show up in passwd_ent.pw_dir */
|
|
|
|
/* the rest of the fields are unimportant and not being considered */
|
|
|
|
plog(XLOG_USER, "hlfsd_getpwent: name=%s, uid=%ld, dir=%s",
|
|
passwd_ent.pw_name, (long) passwd_ent.pw_uid, passwd_ent.pw_dir);
|
|
|
|
return &passwd_ent;
|
|
}
|
|
|
|
|
|
/*
|
|
* read and hash the passwd file or NIS map
|
|
*/
|
|
void
|
|
plt_init(void)
|
|
{
|
|
struct passwd *pent_p;
|
|
|
|
if (plt_reset() < 0) /* could not reset table. skip. */
|
|
return;
|
|
|
|
plog(XLOG_INFO, "reading password map");
|
|
|
|
hlfsd_setpwent(); /* prepare to read passwd entries */
|
|
while ((pent_p = hlfsd_getpwent()) != (struct passwd *) NULL) {
|
|
table_add(pent_p->pw_uid, pent_p->pw_dir, pent_p->pw_name);
|
|
if (STREQ("root", pent_p->pw_name)) {
|
|
int len;
|
|
if (root_home)
|
|
XFREE(root_home);
|
|
root_home = strdup(pent_p->pw_dir);
|
|
len = strlen(root_home);
|
|
/* remove any trailing '/' chars from root's home (even if just one) */
|
|
while (len > 0 && root_home[len - 1] == '/') {
|
|
len--;
|
|
root_home[len] = '\0';
|
|
}
|
|
}
|
|
}
|
|
hlfsd_endpwent();
|
|
|
|
qsort((char *) pwtab, cur_pwtab_num, sizeof(uid2home_t),
|
|
plt_compare_fxn);
|
|
qsort((char *) untab, cur_pwtab_num, sizeof(username2uid_t),
|
|
unt_compare_fxn);
|
|
|
|
if (!root_home)
|
|
root_home = strdup("");
|
|
|
|
plog(XLOG_INFO, "password map read and sorted");
|
|
}
|
|
|
|
|
|
/*
|
|
* This is essentially so that we don't reset known good lookup tables when a
|
|
* YP server goes down.
|
|
*/
|
|
static int
|
|
plt_reset(void)
|
|
{
|
|
int i;
|
|
|
|
clock_valid = 0; /* invalidate logging clock */
|
|
|
|
hlfsd_setpwent();
|
|
if (hlfsd_getpwent() == (struct passwd *) NULL) {
|
|
hlfsd_endpwent();
|
|
return -1; /* did not reset table */
|
|
}
|
|
hlfsd_endpwent();
|
|
|
|
lastchild = (uid2home_t *) NULL;
|
|
|
|
if (max_pwtab_num > 0) /* was used already. cleanup old table */
|
|
for (i = 0; i < cur_pwtab_num; ++i) {
|
|
if (pwtab[i].home) {
|
|
XFREE(pwtab[i].home);
|
|
pwtab[i].home = (char *) NULL;
|
|
}
|
|
pwtab[i].uid = INVALIDID; /* not a valid uid (yet...) */
|
|
pwtab[i].child = (pid_t) 0;
|
|
pwtab[i].uname = (char *) NULL; /* only a ptr to untab[i].username */
|
|
if (untab[i].username) {
|
|
XFREE(untab[i].username);
|
|
untab[i].username = (char *) NULL;
|
|
}
|
|
untab[i].uid = INVALIDID; /* invalid uid */
|
|
untab[i].home = (char *) NULL; /* only a ptr to pwtab[i].home */
|
|
}
|
|
cur_pwtab_num = 0; /* zero current size */
|
|
|
|
if (root_home)
|
|
XFREE(root_home);
|
|
|
|
return 0; /* resetting ok */
|
|
}
|
|
|
|
|
|
/*
|
|
* u: uid number
|
|
* h: home directory
|
|
* n: user ID name
|
|
*/
|
|
static void
|
|
table_add(int u, const char *h, const char *n)
|
|
{
|
|
int i;
|
|
|
|
clock_valid = 0; /* invalidate logging clock */
|
|
|
|
if (max_pwtab_num <= 0) { /* was never initialized */
|
|
max_pwtab_num = 1;
|
|
pwtab = (uid2home_t *) xmalloc(max_pwtab_num *
|
|
sizeof(uid2home_t));
|
|
memset((char *) &pwtab[0], 0, max_pwtab_num * sizeof(uid2home_t));
|
|
untab = (username2uid_t *) xmalloc(max_pwtab_num *
|
|
sizeof(username2uid_t));
|
|
memset((char *) &untab[0], 0, max_pwtab_num * sizeof(username2uid_t));
|
|
}
|
|
|
|
/* check if need more space. */
|
|
if (cur_pwtab_num + 1 > max_pwtab_num) {
|
|
/* need more space in table */
|
|
max_pwtab_num *= 2;
|
|
plog(XLOG_INFO, "reallocating table spaces to %d entries", max_pwtab_num);
|
|
pwtab = (uid2home_t *) xrealloc(pwtab,
|
|
sizeof(uid2home_t) * max_pwtab_num);
|
|
untab = (username2uid_t *) xrealloc(untab,
|
|
sizeof(username2uid_t) *
|
|
max_pwtab_num);
|
|
/* zero out newly added entries */
|
|
for (i=cur_pwtab_num; i<max_pwtab_num; ++i) {
|
|
memset((char *) &pwtab[i], 0, sizeof(uid2home_t));
|
|
memset((char *) &untab[i], 0, sizeof(username2uid_t));
|
|
}
|
|
}
|
|
|
|
/* do NOT add duplicate entries (this is an O(N^2) algorithm... */
|
|
for (i=0; i<cur_pwtab_num; ++i)
|
|
if (u == pwtab[i].uid && u != 0 ) {
|
|
#ifdef DEBUG
|
|
dlog("ignoring duplicate home %s for uid %d (already %s)",
|
|
h, u, pwtab[i].home);
|
|
#endif /* DEBUG */
|
|
return;
|
|
}
|
|
|
|
/* add new password entry */
|
|
pwtab[cur_pwtab_num].home = strdup(h);
|
|
pwtab[cur_pwtab_num].child = 0;
|
|
pwtab[cur_pwtab_num].last_access_time = 0;
|
|
pwtab[cur_pwtab_num].last_status = 0; /* assume best: used homedir */
|
|
pwtab[cur_pwtab_num].uid = u;
|
|
|
|
/* add new userhome entry */
|
|
untab[cur_pwtab_num].username = strdup(n);
|
|
|
|
/* just a second pointer */
|
|
pwtab[cur_pwtab_num].uname = untab[cur_pwtab_num].username;
|
|
untab[cur_pwtab_num].uid = u;
|
|
untab[cur_pwtab_num].home = pwtab[cur_pwtab_num].home; /* a ptr */
|
|
|
|
/* increment counter */
|
|
++cur_pwtab_num;
|
|
}
|
|
|
|
|
|
/*
|
|
* return entry in lookup table
|
|
*/
|
|
uid2home_t *
|
|
plt_search(int u)
|
|
{
|
|
int max, min, mid;
|
|
|
|
/*
|
|
* empty table should not happen,
|
|
* but I have a bug with signals to trace...
|
|
*/
|
|
if (pwtab == (uid2home_t *) NULL)
|
|
return (uid2home_t *) NULL;
|
|
|
|
max = cur_pwtab_num - 1;
|
|
min = 0;
|
|
|
|
do {
|
|
mid = (max + min) / 2;
|
|
if (pwtab[mid].uid == u) /* record found! */
|
|
return &pwtab[mid];
|
|
if (pwtab[mid].uid > u)
|
|
max = mid;
|
|
else
|
|
min = mid;
|
|
} while (max > min + 1);
|
|
|
|
if (pwtab[max].uid == u)
|
|
return &pwtab[max];
|
|
if (pwtab[min].uid == u)
|
|
return &pwtab[min];
|
|
|
|
/* if gets here then record was not found */
|
|
return (uid2home_t *) NULL;
|
|
}
|
|
|
|
|
|
#if defined(DEBUG) || defined(DEBUG_PRINT)
|
|
void
|
|
plt_print(int signum)
|
|
{
|
|
FILE *dumpfile;
|
|
int dumpfd;
|
|
char dumptmp[] = "/usr/tmp/hlfsd.dump.XXXXXX";
|
|
int i;
|
|
|
|
#ifdef HAVE_MKSTEMP
|
|
dumpfd = mkstemp(dumptmp);
|
|
#else /* not HAVE_MKSTEMP */
|
|
mktemp(dumptmp);
|
|
if (!dumptmp) {
|
|
plog(XLOG_ERROR, "cannot create temporary dump file");
|
|
return;
|
|
}
|
|
dumpfd = open(dumptmp, O_RDONLY);
|
|
#endif /* not HAVE_MKSTEMP */
|
|
if (dumpfd < 0) {
|
|
plog(XLOG_ERROR, "cannot open temporary dump file");
|
|
return;
|
|
}
|
|
if ((dumpfile = fdopen(dumpfd, "a")) != NULL) {
|
|
plog(XLOG_INFO, "dumping internal state to file %s", dumptmp);
|
|
fprintf(dumpfile, "\n\nNew plt_dump():\n");
|
|
for (i = 0; i < cur_pwtab_num; ++i)
|
|
fprintf(dumpfile,
|
|
"%4d %5lu %10lu %1d %4lu \"%s\" uname=\"%s\"\n",
|
|
i,
|
|
(long) pwtab[i].child,
|
|
pwtab[i].last_access_time,
|
|
pwtab[i].last_status,
|
|
(long) pwtab[i].uid,
|
|
pwtab[i].home,
|
|
pwtab[i].uname);
|
|
fprintf(dumpfile, "\nUserName table by plt_print():\n");
|
|
for (i = 0; i < cur_pwtab_num; ++i)
|
|
fprintf(dumpfile, "%4d : \"%s\" %4lu \"%s\"\n", i,
|
|
untab[i].username, (long) untab[i].uid, untab[i].home);
|
|
close(dumpfd);
|
|
fclose(dumpfile);
|
|
}
|
|
}
|
|
|
|
|
|
void
|
|
plt_dump(uid2home_t *lastc, pid_t this)
|
|
{
|
|
FILE *dumpfile;
|
|
int i;
|
|
|
|
if ((dumpfile = fopen("/var/tmp/hlfsdump", "a")) != NULL) {
|
|
fprintf(dumpfile, "\n\nNEW PLT_DUMP -- ");
|
|
fprintf(dumpfile, "lastchild->child=%d ",
|
|
(int) (lastc ? lastc->child : -999));
|
|
fprintf(dumpfile, ", child from wait3=%lu:\n", (long) this);
|
|
for (i = 0; i < cur_pwtab_num; ++i)
|
|
fprintf(dumpfile, "%4d %5lu: %4lu \"%s\" uname=\"%s\"\n", i,
|
|
(long) pwtab[i].child, (long) pwtab[i].uid,
|
|
pwtab[i].home, pwtab[i].uname);
|
|
fprintf(dumpfile, "\nUserName table by plt_dump():\n");
|
|
for (i = 0; i < cur_pwtab_num; ++i)
|
|
fprintf(dumpfile, "%4d : \"%s\" %4lu \"%s\"\n", i,
|
|
untab[i].username, (long) untab[i].uid, untab[i].home);
|
|
fprintf(dumpfile, "ezk: ent=%d, uid=%lu, home=\"%s\"\n",
|
|
untab_index("ezk"),
|
|
(long) untab[untab_index("ezk")].uid,
|
|
pwtab[untab[untab_index("ezk")].uid].home);
|
|
fprintf(dumpfile, "rezk: ent=%d, uid=%lu, home=\"%s\"\n",
|
|
untab_index("rezk"),
|
|
(long) untab[untab_index("rezk")].uid,
|
|
pwtab[untab[untab_index("rezk")].uid].home);
|
|
fclose(dumpfile);
|
|
}
|
|
}
|
|
#endif /* defined(DEBUG) || defined(DEBUG_PRINT) */
|