[PATCH] Updated audit failure query functionality
Lisa Smith
lisa.m.smith at hp.com
Tue Jun 27 13:32:09 UTC 2006
Tim,
Thanks for your comments.
Other than the missing brackets in get_auditfail_action (good catch,
thanks), all the other functions were pulled directly from
auditd-config.c. It was requested that the new audit failure query code
use the functions from auditd for ease of code maintainability. As you
pointed out, the parsing only works with a very specific format, which
is why I rewrote it in the original patch. This is the parsing
mechanism used for auditd.conf.
Steve and others, I assume you'd like to leave the functions as they
are, so they match what is in auditd-config.c. If not, please comment
on how to handle each of the concerns Tim points out below.
Thanks,
Lisa
Timothy R. Chavez wrote:
> On Mon, 2006-06-26 at 15:23 -0400, Lisa Smith wrote:
>
>> This is an updated patch for the audit failure query
>> functionality. The patch has been completely rewritten
>> to match the programming style and algorithms used in
>> auditd-config.c.
>>
> <snip>
>
>> Lisa
>>
>
> Hi Lisa,
>
> I had some comments below. Thanks!
>
> -tim
>
>
>> -------------------
>>
>> libaudit.c | 246 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>> libaudit.h | 11 ++
>> 2 files changed, 257 insertions(+)
>>
>> diff -burN orig/libaudit.c src/libaudit.c
>> --- orig/libaudit.c 2006-05-25 17:39:52.000000000 -0400
>> +++ src/libaudit.c 2006-06-26 15:00:56.000000000 -0400
>> @@ -42,13 +42,56 @@
>> #include "libaudit.h"
>> #include "private.h"
>>
>> +/* Local prototypes */
>> +struct nv_pair
>> +{
>> + const char *name;
>> + const char *value;
>> +};
>> +
>> +struct kw_pair
>> +{
>> + const char *name;
>> + int (*parser)(const char *, int);
>> +};
>> +
>> +struct nv_list
>> +{
>> + const char *name;
>> + int option;
>> +};
>> +
>> +struct libaudit_conf
>> +{
>> + auditfail_t failure_action;
>> +};
>>
>> int audit_archadded = 0;
>> int audit_syscalladded = 0;
>> unsigned int audit_elf = 0U;
>> +static struct libaudit_conf config;
>> +static int load_libaudit_config();
>> +static char *get_line(FILE *f, char *buf);
>> +static int nv_split(char *buf, struct nv_pair *nv);
>> +static const struct kw_pair *kw_lookup(const char *val);
>> +static int audit_failure_parser(const char *val, int line);
>> static int name_to_uid(const char *name, uid_t *uid);
>> static int name_to_gid(const char *name, gid_t *gid);
>>
>> +static const struct kw_pair keywords[] =
>> +{
>> + {"failure_action", audit_failure_parser },
>> + { NULL, NULL }
>> +};
>> +
>> +static const struct nv_list failure_actions[] =
>> +{
>> + {"ignore", FA_IGNORE },
>> + {"log", FA_LOG },
>> + {"terminate", FA_TERMINATE },
>> + { NULL, 0 }
>> +};
>> +
>>
>> int audit_request_status(int fd)
>> {
>> @@ -68,6 +111,209 @@
>> return rc;
>> }
>>
>> +/*
>> + * Set everything to its default value
>> +*/
>> +static void clear_config()
>> +{
>> + config.failure_action = FA_IGNORE;
>> +}
>> +
>> +static int load_libaudit_config()
>> +{
>> + int fd, rc, lineno = 1;
>> + struct stat st;
>> + FILE *f;
>> + char buf[128];
>> +
>> + clear_config();
>> +
>> + /* open the file */
>> + rc = open(CONFIG_FILE, O_NOFOLLOW|O_RDONLY);
>> + if (rc < 0) {
>> + if (errno != ENOENT) {
>> + audit_msg(LOG_ERR, "Error opening config file (%s)",
>> + strerror(errno));
>> + return 1;
>> + }
>> + audit_msg(LOG_WARNING,
>> + "Config file %s doesn't exist, skipping", CONFIG_FILE);
>> + return 0;
>> + }
>> + fd = rc;
>> +
>> + /* check the file's permissions: owned by root, not world writable,
>> + * not symlink.
>> + */
>> + audit_msg(LOG_DEBUG, "Config file %s opened for parsing",
>> + CONFIG_FILE);
>> + if (fstat(fd, &st) < 0) {
>> + audit_msg(LOG_ERR, "Error fstat'ing config file (%s)",
>> + strerror(errno));
>> + return 1;
>> + }
>> + if (st.st_uid != 0) {
>> + audit_msg(LOG_ERR, "Error - %s isn't owned by root",
>> + CONFIG_FILE);
>> + return 1;
>> + }
>>
>
> Hm, you really want to hard-code in a root check? What if you're acting
> as another type of administrator with access to audit? Done know, I
> could be out of line here.
>
> [..]
>
>> + if ((st.st_mode & S_IWOTH) == S_IWOTH) {
>> + audit_msg(LOG_ERR, "Error - %s is world writable",
>> + CONFIG_FILE);
>> + return 1;
>> + }
>> + if (!S_ISREG(st.st_mode)) {
>> + audit_msg(LOG_ERR, "Error - %s is not a regular file",
>> + CONFIG_FILE);
>> + return 1;
>> + }
>> +
>> + /* it's ok, read line by line */
>> + f = fdopen(fd, "r");
>> + if (f == NULL) {
>> + audit_msg(LOG_ERR, "Error - fdopen failed (%s)",
>> + strerror(errno));
>> + return 1;
>> + }
>> +
>> + while (get_line(f, buf)) {
>> + // convert line into name-value pair
>> + const struct kw_pair *kw;
>> + struct nv_pair nv;
>> + rc = nv_split(buf, &nv);
>> + switch (rc) {
>> + case 0: // fine
>> + break;
>> + case 1: // not the right number of tokens.
>> + audit_msg(LOG_ERR,
>> + "Wrong number of arguments for line %d in %s",
>> + lineno, CONFIG_FILE);
>> + break;
>> + case 2: // no '=' sign
>> + audit_msg(LOG_ERR,
>> + "Missing equal sign for line %d in %s",
>> + lineno, CONFIG_FILE);
>> + break;
>> + default: // something else went wrong...
>> + audit_msg(LOG_ERR,
>> + "Unknown error for line %d in %s",
>> + lineno, CONFIG_FILE);
>> + break;
>> + }
>> + if (nv.name == NULL) {
>> + lineno++;
>> + continue;
>> + }
>> +
>> + /* identify keyword or error */
>> + kw = kw_lookup(nv.name);
>> + if (kw->name == NULL) {
>> + audit_msg(LOG_ERR,
>> + "Unknown keyword \"%s\" in line %d of %s",
>> + nv.name, lineno, CONFIG_FILE);
>> + fclose(f);
>> + return 1;
>> + }
>> +
>> + /* dispatch to keyword's local parser */
>> + rc = kw->parser(nv.value, lineno);
>> + if (rc != 0) {
>> + fclose(f);
>> + return 1; // local parser puts message out
>> + }
>> +
>> + lineno++;
>> + }
>> +
>> + fclose(f);
>> + return 0;
>> +}
>> +
>> +int get_auditfail_action(auditfail_t *failmode)
>> +{
>> + if (load_libaudit_config())
>> + *failmode = config.failure_action;
>> + return 1;
>>
>
> Missing brackets?
>
> [..]
>
>> +
>> + *failmode = config.failure_action;
>> + return 0;
>> +}
>> +
>> +
>> +static char *get_line(FILE *f, char *buf)
>> +{
>> + if (fgets_unlocked(buf, 128, f)) {
>> + /* remove newline */
>> + char *ptr = strchr(buf, 0x0a);
>> + if (ptr)
>> + *ptr = 0;
>> + return buf;
>> + }
>> + return NULL;
>> +}
>> +
>> +static int nv_split(char *buf, struct nv_pair *nv)
>> +{
>> + /* Get the name part */
>> + char *ptr;
>> +
>> + nv->name = NULL;
>> + nv->value = NULL;
>> + ptr = strtok(buf, " ");
>> + if (ptr == NULL)
>> + return 0; /* If there's nothing, go to next line */
>> + if (ptr[0] == '#')
>> + return 0; /* If there's a comment, go to next line */
>>
>
> I still think this will break if ptr[0] happens to be a space followed
> by a '#'. I suppose it could break by design if the first character of
> the line is not '#', but I wouldn't like that much as a user myself.
>
> [..]
>
>> + nv->name = ptr;
>> +
>> + /* Check for a '=' */
>> + ptr = strtok(NULL, " ");
>> + if (ptr == NULL)
>> + return 1;
>> + if (strcmp(ptr, "=") != 0)
>>
>
> No real need to do a strcmp here.
>
> if (ptr && *ptr == '=')
>
> [..]
>
>> + return 2;
>> +
>> + /* get the value */
>> + ptr = strtok(NULL, " ");
>> + if (ptr == NULL)
>> + return 1;
>> + nv->value = ptr;
>>
>
> I'll be the first to admit that I relatively little experience with
> strtok, but is this pointer assignment even safe? You're pointing into
> a buffer that could potentially be overwritten by another app, right?
>
> Here's my off-the-top-of-the-head, uncompiled, and untested example
> using strtok_r...
>
> char *buf;
> char *ptr;
>
> buf = (char *)malloc(128);
> if (!buf) {
> /* handle memory error */
> }
> ptr = strtok_r(line, "=", buf);
> if (!ptr || *buf == '\0') {
> /* not "token=value", free memory */
> }
> /* <do comment check here> */
> nv->name = (char*)malloc(strlen(ptr)+1);
> if (!nv->name) {
> /* handle memory error */
> }
> strcpy(nv->name, ptr);
> nv->value = (char*)malloc(strlen(buf)+1);
> if (!nv->value) {
> /* handle memory error */
> }
> strcpy(nv->value, buf);
>
> /* clean up memory */
> return 0;
>
> You may not care about whether or not you're thread-safe or being able
> to have multiple word token values... but at least by providing your own
> buffer, you can keep track of your position in the parsed string in a
> lovely way.
>
> [..]
>
>> +
>> + /* Make sure there's nothing else */
>> + ptr = strtok(NULL, " ");
>> + if (ptr)
>> + return 1;
>> +
>> + /* Everything is OK */
>> + return 0;
>> +}
>> +
>> +static const struct kw_pair *kw_lookup(const char *val)
>> +{
>> + int i = 0;
>> + while (keywords[i].name != NULL) {
>> + if (strcasecmp(keywords[i].name, val) == 0)
>> + break;
>> + i++;
>> + }
>> + return &keywords[i];
>>
>
> So I think this would look better a for-loop as you're in fact iterating
> anyway :) *oh my, looks down at next function*
>
> [..]
>
>> +}
>> +
>> +static int audit_failure_parser(const char *val, int line)
>> +{
>> + int i;
>> +
>> + audit_msg(LOG_DEBUG, "audit_failure_parser called with: %s", val);
>> + for (i=0; failure_actions[i].name != NULL; i++) {
>> + if (strcasecmp(val, failure_actions[i].name) == 0) {
>> + config.failure_action = failure_actions[i].option;
>> + return 0;
>> + }
>> + }
>>
>
> :)
>
> [..]
>
>> + audit_msg(LOG_ERR, "Option %s not found - line %d", val, line);
>> + return 1;
>> +}
>> +
>> int audit_set_enabled(int fd, uint32_t enabled)
>> {
>> int rc;
>> diff -burN orig/libaudit.h src/libaudit.h
>> --- orig/libaudit.h 2006-05-25 17:38:21.000000000 -0400
>> +++ src/libaudit.h 2006-06-26 14:58:53.000000000 -0400
>> @@ -248,6 +248,16 @@
>> MACH_ALPHA
>> } machine_t;
>>
>> +/* These are the valid audit failure tunable enum values */
>> +typedef enum {
>> + FA_IGNORE=0,
>> + FA_LOG,
>> + FA_TERMINATE
>> +} auditfail_t;
>> +
>> +/* #defines for the audit failure query */
>> +#define CONFIG_FILE "/etc/libaudit.conf"
>> +
>> /*
>> * audit_rule_data supports filter rules with both integer and string
>> * fields. It corresponds with AUDIT_ADD_RULE, AUDIT_DEL_RULE and
>> @@ -362,6 +372,7 @@
>> /* AUDIT_GET */
>> extern int audit_request_status(int fd);
>> extern int audit_is_enabled(int fd);
>> +extern int get_auditfail_action(auditfail_t *failmode);
>>
>> /* AUDIT_SET */
>> typedef enum { WAIT_NO, WAIT_YES } rep_wait_t;
>>
>
>
>
More information about the Linux-audit
mailing list