Skip to contentSkip to conversation

secilc/docs: selinuxuser actually takes a string not identifier

PATCH In review bauen1 2 participants 4 messages Aug 28, 2022 – Aug 31, 2022 SELinux security module

lore
Share & tools
lore

bauen1

author
#0 lore raw

PATCH 1/3 secilc/docs: selinuxuser actually takes a string not identifier

Update parameter table and add note to example.
This can very easily lead to confusion, because secilc does not reject e.g. (selinuxuser "admin_1" admin low_low).

---
 secilc/docs/cil_user_statements.md | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
secilc/docs/cil_user_statements.md +3 -2
index d5674f12..45a52140 100644
@@ -429,7 +429,7 @@ Associates a GNU/Linux user to a previously declared [`user`](cil_user_statement
 </tr>
 <tr class="odd">
 <td align="left"><p><code>user_id</code></p></td>
-<td align="left"><p>A previously declared SELinux <code>user</code> identifier.</p></td>
+<td align="left"><p>A <b>string</b> referencing a previously declared SELinux <code>user</code> identifier.</p></td>
 </tr>
 <tr class="even">
 <td align="left"><p><code>userrange_id</code></p></td>
@@ -445,7 +445,8 @@ This example will associate `unconfined.admin` user with a GNU / Linux user "`ad
 ```secil
     (block unconfined
         (user admin)
-        (selinuxuser admin_1 admin low_low)
+        ; XXX: Because the user_id is a string, the fully qualified identifier needs to be used here:
+        (selinuxuser "admin_1" "unconfined.admin" low_low)
     )
 ```
 
Signature · 1 line
2.36.1
#1 lore raw
Quoted 24 lines · On Sun, Aug 28, 2022 at 8:18 AM bauen1 <[email protected]> wrote:

Update parameter table and add note to example.
This can very easily lead to confusion, because secilc does not reject e.g. (selinuxuser "admin_1" admin low_low).

---
  secilc/docs/cil_user_statements.md | 5 +++--
  1 file changed, 3 insertions(+), 2 deletions(-)
secilc/docs/cil_user_statements.md +1 -1
index d5674f12..45a52140 100644
@@ -429,7 +429,7 @@ Associates a GNU/Linux user to a previously declared [`user`](cil_user_statement
  </tr>
  <tr class="odd">
  <td align="left"><p><code>user_id</code></p></td>
-<td align="left"><p>A previously declared SELinux <code>user</code> identifier.</p></td>
+<td align="left"><p>A <b>string</b> referencing a previously declared SELinux <code>user</code> identifier.</p></td>
  </tr>
  <tr class="even">
  <td align="left"><p><code>userrange_id</code></p></td>

secilc/docs/cil_user_statements.md This example will associate `unconfined.admin` user with a GNU / Linux user "`ad · line 449

Quoted 6 lines
secilc/docs/cil_user_statements.md +2 -1
@@ -445,7 +445,8 @@ This example will associate `unconfined.admin` user with a GNU / Linux user "`ad
  ```secil
      (block unconfined
          (user admin)
-        (selinuxuser admin_1 admin low_low)
+        ; XXX: Because the user_id is a string, the fully qualified identifier needs to be used here:
+        (selinuxuser "admin_1" "unconfined.admin" low_low)

I don't see this behavior.

The following:
(block b2
(user user2_u)
(userrole user2_u ROLE)
(userlevel user2_u (SENS))
(userrange user2_u ((SENS) (SENS (CAT)))) (userprefix user2_u user2)
(selinuxuser name2 user2_u ((SENS) (SENS (CAT))))
)
Gives me:
name2:b2.user2_u:SENS-SENS:CAT

It works with quotes as well, but it doesn't require them.

The following:
(block b3
(user user3_u)
(userrole user3_u ROLE)
(userlevel user3_u (SENS))
(userrange user3_u ((SENS) (SENS (CAT)))) (userprefix user3_u user3)
(selinuxuser "name3" "user3_u" ((SENS) (SENS (CAT))))
)
Gives me:
name3:b3.user3_u:SENS-SENS:CAT

Thanks,
Jim

#2 lore raw
Quoted 61 lines · James Carter
Quoted 24 lines · On Sun, Aug 28, 2022 at 8:18 AM bauen1 <[email protected]> wrote:

Update parameter table and add note to example.
This can very easily lead to confusion, because secilc does not reject e.g. (selinuxuser "admin_1" admin low_low).

---
  secilc/docs/cil_user_statements.md | 5 +++--
  1 file changed, 3 insertions(+), 2 deletions(-)
secilc/docs/cil_user_statements.md +1 -1
index d5674f12..45a52140 100644
@@ -429,7 +429,7 @@ Associates a GNU/Linux user to a previously declared [`user`](cil_user_statement
  </tr>
  <tr class="odd">
  <td align="left"><p><code>user_id</code></p></td>
-<td align="left"><p>A previously declared SELinux <code>user</code> identifier.</p></td>
+<td align="left"><p>A <b>string</b> referencing a previously declared SELinux <code>user</code> identifier.</p></td>
  </tr>
  <tr class="even">
  <td align="left"><p><code>userrange_id</code></p></td>

secilc/docs/cil_user_statements.md This example will associate `unconfined.admin` user with a GNU / Linux user "`ad · line 449

Quoted 6 lines
secilc/docs/cil_user_statements.md +2 -1
@@ -445,7 +445,8 @@ This example will associate `unconfined.admin` user with a GNU / Linux user "`ad
  ```secil
      (block unconfined
          (user admin)
-        (selinuxuser admin_1 admin low_low)
+        ; XXX: Because the user_id is a string, the fully qualified identifier needs to be used here:
+        (selinuxuser "admin_1" "unconfined.admin" low_low)

I don't see this behavior.

The following:
(block b2
(user user2_u)
(userrole user2_u ROLE)
(userlevel user2_u (SENS))
(userrange user2_u ((SENS) (SENS (CAT)))) (userprefix user2_u user2)
(selinuxuser name2 user2_u ((SENS) (SENS (CAT))))
)
Gives me:
name2:b2.user2_u:SENS-SENS:CAT

It works with quotes as well, but it doesn't require them.

The following:
(block b3
(user user3_u)
(userrole user3_u ROLE)
(userlevel user3_u (SENS))
(userrange user3_u ((SENS) (SENS (CAT)))) (userprefix user3_u user3)
(selinuxuser "name3" "user3_u" ((SENS) (SENS (CAT))))
)
Gives me:
name3:b3.user3_u:SENS-SENS:CAT

Quoted 2 lines

Thanks,
Jim

Attached is the policy I used to test and a program that calls cil_selinuxusers_to_string().
Jim

#include <stdlib.h>
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#include <sys/stat.h>

#include <sepol/cil/cil.h>
#include <sepol/policydb.h>

static int write_out_selinuxusers(struct cil_db *db)

{
	int rc;
	char *data;
	size_t size, i;

	rc = cil_selinuxusers_to_string(db, &data, &size);
	if (rc != SEPOL_OK) {
		return rc;
	}

	if (size > 0) {
		printf("####################\n");
		for (i=0; i<size; i++) {
			printf("%c", data[i]);
		}
		printf("####################\n");
	}

	return 0;
}

/*
gcc -I/home/jim/local/usr/include -L/home/jim/local/usr/lib -lsepol -o get_selinuxusers get_selinuxusers.c */
int main(int argc, char *argv[])

{
	struct cil_db *db = NULL;
	FILE *file = NULL;
	char *buffer = NULL;
	struct stat filedata;
	uint32_t file_size;
	int i;
	int rc = SEPOL_ERR;

	if (argc < 2) {
		fprintf(stderr, "No cil files specified\n");
		rc = SEPOL_ERR;
		goto exit;
	}

	cil_set_log_level(CIL_ERR);

	cil_db_init(&db);
	cil_set_mls(db, 1);
	cil_set_preserve_tunables(db, 0);
	cil_set_qualified_names(db, 0);
	cil_set_attrs_expand_generated(db, 0);
	cil_set_attrs_expand_size(db, 0);

	for (i = 1; i < argc; i++) {
		file = fopen(argv[i], "r");
		if (!file) {
			fprintf(stderr, "Could not open file: %s\n", argv[i]);
			rc = SEPOL_ERR;
			goto exit;
		}
		rc = stat(argv[i], &filedata);
		if (rc == -1) {
			fprintf(stderr, "Could not stat file: %s\n", argv[i]);
			goto exit;
		}
		file_size = filedata.st_size;

		buffer = malloc(file_size);
		rc = fread(buffer, file_size, 1, file);
		if (rc != 1) {
			fprintf(stderr, "Failure reading file: %s\n", argv[i]);
			goto exit;
		}
		fclose(file);
		file = NULL;

		rc = cil_add_file(db, argv[i], buffer, file_size);
		if (rc != SEPOL_OK) {
			fprintf(stderr, "Failure adding %s\n", argv[i]);
			goto exit;
		}

		free(buffer);
		buffer = NULL;
	}

	rc = cil_compile(db);
	if (rc != SEPOL_OK) {
		fprintf(stderr, "Failed to compile CIL policy\n");
		goto exit;
	}

	rc = write_out_selinuxusers(db);
	if (rc != SEPOL_OK) {
		fprintf(stderr, "Failed to write selinux users\n");
		goto exit;
	}

exit:

	free(buffer);
	cil_db_destroy(&db);
	return rc;
}
author
#3 lore raw
Quoted 30 lines · James Carter
Quoted 29 lines · James Carter

I don't see this behavior.

The following:
(block b2
(user user2_u)
(userrole user2_u ROLE)
(userlevel user2_u (SENS))
(userrange user2_u ((SENS) (SENS (CAT)))) (userprefix user2_u user2)
(selinuxuser name2 user2_u ((SENS) (SENS (CAT))))
)
Gives me:
name2:b2.user2_u:SENS-SENS:CAT

It works with quotes as well, but it doesn't require them.

The following:
(block b3
(user user3_u)
(userrole user3_u ROLE)
(userlevel user3_u (SENS))
(userrange user3_u ((SENS) (SENS (CAT)))) (userprefix user3_u user3)
(selinuxuser "name3" "user3_u" ((SENS) (SENS (CAT))))
)
Gives me:
name3:b3.user3_u:SENS-SENS:CAT

Quoted 6 lines
Quoted 2 lines

Thanks,
Jim

Attached is the policy I used to test and a program that calls cil_selinuxusers_to_string().
Jim

Thanks, I'll take another look at this some time. Obviously the other patches can be merged independently.

Signature · 1 line
bauen1