Skip to contentSkip to conversation

selinux: replace strlcat() with seq_buf in selinux_ima_collect_state()

PATCH In review Ian Bridges 5 participants 7 messages Jun 26 – Jul 13 SELinux security module

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 selinux: replace strlcat() with seq_buf in selinux_ima_collect_state()

In preparation for removing the deprecated strlcat() API[1], replace the strscpy()/strlcat() chain in selinux_ima_collect_state() with a struct seq_buf, which tracks the write position and remaining space internally.

Each field is written with seq_buf_printf() using a "=%d;" format, which removes the open-coded "=1;"/"=0;" constants. The seven per-append WARN_ON(rc >= buf_len) truncation checks are replaced by a single seq_buf_has_overflowed() check after the string is built.

---
Changed in v2: replace the v1 seq_buf_puts() pairs with seq_buf_printf() using a "=%d;" format, which drops the open-coded "=1;"/"=0;" constants.

v1: https://lore.kernel.org/all/ajlN94VO7BYNUTAy@dev/

I didn't change the precomputation of the string size. An alternative, which is used by other seq_buf callers (e.g. kernel/rcu/refscale.c, mm/memcontrol.c), is to drop the precomputation and allocate an oversized fixed buffer, relying on the seq_buf overflow check as a backstop. I'm happy to rework the patch to adopt that alternative.

 security/selinux/ima.c | 40 +++++++++++++---------------------------
 1 file changed, 13 insertions(+), 27 deletions(-)
security/selinux/ima.c +13 -27
index aa34da9b0aeb..cb0efa2fc1ad 100644
@@ -9,6 +9,7 @@
  */
 #include <linux/vmalloc.h>
 #include <linux/ima.h>
+#include <linux/seq_buf.h>
 #include "security.h"
 #include "ima.h"
 
@@ -20,46 +21,31 @@
  */
 static char *selinux_ima_collect_state(void)
 {
-	const char *on = "=1;", *off = "=0;";
+	struct seq_buf s;
 	char *buf;
-	int buf_len, len, i, rc;
+	int buf_len, suffix_len, i;
 
 	buf_len = strlen("initialized=0;enforcing=0;checkreqprot=0;") + 1;
+	suffix_len = strlen("=0;");
 
-	len = strlen(on);
 	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
-		buf_len += strlen(selinux_policycap_names[i]) + len;
+		buf_len += strlen(selinux_policycap_names[i]) + suffix_len;
 
 	buf = kzalloc(buf_len, GFP_KERNEL);
 	if (!buf)
 		return NULL;
 
-	rc = strscpy(buf, "initialized", buf_len);
-	WARN_ON(rc < 0);
+	seq_buf_init(&s, buf, buf_len);
 
-	rc = strlcat(buf, selinux_initialized() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
+	seq_buf_printf(&s, "initialized=%d;enforcing=%d;checkreqprot=%d;",
+		       selinux_initialized(), enforcing_enabled(),
+		       checkreqprot_get());
 
-	rc = strlcat(buf, "enforcing", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, enforcing_enabled() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, "checkreqprot", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, checkreqprot_get() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	for (i = 0; i < __POLICYDB_CAP_MAX; i++) {
-		rc = strlcat(buf, selinux_policycap_names[i], buf_len);
-		WARN_ON(rc >= buf_len);
+	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
+		seq_buf_printf(&s, "%s=%d;", selinux_policycap_names[i],
+			       selinux_state.policycap[i]);
 
-		rc = strlcat(buf, selinux_state.policycap[i] ? on : off,
-			buf_len);
-		WARN_ON(rc >= buf_len);
-	}
+	WARN_ON(seq_buf_has_overflowed(&s));
 
 	return buf;
 }
Signature · 1 line
2.47.3
#1 lore raw
Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 11 lines

In preparation for removing the deprecated strlcat() API[1], replace the strscpy()/strlcat() chain in selinux_ima_collect_state() with a struct seq_buf, which tracks the write position and remaining space internally.

Each field is written with seq_buf_printf() using a "=%d;" format, which removes the open-coded "=1;"/"=0;" constants. The seven per-append WARN_ON(rc >= buf_len) truncation checks are replaced by a single seq_buf_has_overflowed() check after the string is built.

Sashiko has reviewed this patch and found no issues. It looks great!

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/aj6hVjECikvYtnED@dev?part=1
maintainer Reviewed-by on v2 1/1
#2 lore raw
In reply to Ian Bridges
Quoted 5 lines · Ian Bridges

In preparation for removing the deprecated strlcat() API[1], replace the strscpy()/strlcat() chain in selinux_ima_collect_state() with a struct seq_buf, which tracks the write position and remaining space internally.

Quoted 7 lines

Each field is written with seq_buf_printf() using a "=%d;" format, which removes the open-coded "=1;"/"=0;" constants. The seven per-append WARN_ON(rc >= buf_len) truncation checks are replaced by a single seq_buf_has_overflowed() check after the string is built.

Quoted 89 lines

---
Changed in v2: replace the v1 seq_buf_puts() pairs with seq_buf_printf() using a "=%d;" format, which drops the open-coded "=1;"/"=0;" constants.

v1: https://lore.kernel.org/all/ajlN94VO7BYNUTAy@dev/

I didn't change the precomputation of the string size. An alternative, which is used by other seq_buf callers (e.g. kernel/rcu/refscale.c, mm/memcontrol.c), is to drop the precomputation and allocate an oversized fixed buffer, relying on the seq_buf overflow check as a backstop. I'm happy to rework the patch to adopt that alternative.

 security/selinux/ima.c | 40 +++++++++++++---------------------------
 1 file changed, 13 insertions(+), 27 deletions(-)
security/selinux/ima.c +13 -28
index aa34da9b0aeb..cb0efa2fc1ad 100644
@@ -9,6 +9,7 @@
  */
 #include <linux/vmalloc.h>
 #include <linux/ima.h>
+#include <linux/seq_buf.h>
 #include "security.h"
 #include "ima.h"
 
@@ -20,46 +21,31 @@
  */
 static char *selinux_ima_collect_state(void)
 {
-       const char *on = "=1;", *off = "=0;";
+       struct seq_buf s;
        char *buf;
-       int buf_len, len, i, rc;
+       int buf_len, suffix_len, i;
 
        buf_len = strlen("initialized=0;enforcing=0;checkreqprot=0;") + 1;
+       suffix_len = strlen("=0;");
 
-       len = strlen(on);
        for (i = 0; i < __POLICYDB_CAP_MAX; i++)
-               buf_len += strlen(selinux_policycap_names[i]) + len;
+               buf_len += strlen(selinux_policycap_names[i]) + suffix_len;
 
        buf = kzalloc(buf_len, GFP_KERNEL);
        if (!buf)
                return NULL;
 
-       rc = strscpy(buf, "initialized", buf_len);
-       WARN_ON(rc < 0);
+       seq_buf_init(&s, buf, buf_len);
 
-       rc = strlcat(buf, selinux_initialized() ? on : off, buf_len);
-       WARN_ON(rc >= buf_len);
+       seq_buf_printf(&s, "initialized=%d;enforcing=%d;checkreqprot=%d;",
+                      selinux_initialized(), enforcing_enabled(),
+                      checkreqprot_get());
 
-       rc = strlcat(buf, "enforcing", buf_len);
-       WARN_ON(rc >= buf_len);
-
-       rc = strlcat(buf, enforcing_enabled() ? on : off, buf_len);
-       WARN_ON(rc >= buf_len);
-
-       rc = strlcat(buf, "checkreqprot", buf_len);
-       WARN_ON(rc >= buf_len);
-
-       rc = strlcat(buf, checkreqprot_get() ? on : off, buf_len);
-       WARN_ON(rc >= buf_len);
-
-       for (i = 0; i < __POLICYDB_CAP_MAX; i++) {
-               rc = strlcat(buf, selinux_policycap_names[i], buf_len);
-               WARN_ON(rc >= buf_len);
+       for (i = 0; i < __POLICYDB_CAP_MAX; i++)
+               seq_buf_printf(&s, "%s=%d;", selinux_policycap_names[i],
+                              selinux_state.policycap[i]);
 
-               rc = strlcat(buf, selinux_state.policycap[i] ? on : off,
-                       buf_len);
-               WARN_ON(rc >= buf_len);
-       }
+       WARN_ON(seq_buf_has_overflowed(&s));
 
        return buf;
 }
--

2.47.3

#3 lore raw
In reply to Ian Bridges
Quoted 9 lines · On Fri, 26 Jun 2026 10:57:10 -0500 Ian Bridges <[email protected]> wrote:

In preparation for removing the deprecated strlcat() API[1], replace the strscpy()/strlcat() chain in selinux_ima_collect_state() with a struct seq_buf, which tracks the write position and remaining space internally.

Each field is written with seq_buf_printf() using a "=%d;" format, which removes the open-coded "=1;"/"=0;" constants. The seven per-append WARN_ON(rc >= buf_len) truncation checks are replaced by a single seq_buf_has_overflowed() check after the string is built.

Quoted 5 lines

---
Changed in v2: replace the v1 seq_buf_puts() pairs with seq_buf_printf() using a "=%d;" format, which drops the open-coded "=1;"/"=0;" constants.

Did you verify that all the values are 0/1 otherwise a !! prefix might be needed.

Quoted 8 lines

v1: https://lore.kernel.org/all/ajlN94VO7BYNUTAy@dev/

I didn't change the precomputation of the string size. An alternative, which is used by other seq_buf callers (e.g. kernel/rcu/refscale.c, mm/memcontrol.c), is to drop the precomputation and allocate an oversized fixed buffer, relying on the seq_buf overflow check as a backstop. I'm happy to rework the patch to adopt that alternative.

That would be reasonable, the output is always exactly the same length and the buffer is freed just after being allocated.
A comment that there are 3 + 15 strings and all are under 32 bytes so 1k is plenty would suffice.

David

Quoted 75 lines
 security/selinux/ima.c | 40 +++++++++++++---------------------------
 1 file changed, 13 insertions(+), 27 deletions(-)
security/selinux/ima.c +13 -27
index aa34da9b0aeb..cb0efa2fc1ad 100644
@@ -9,6 +9,7 @@
  */
 #include <linux/vmalloc.h>
 #include <linux/ima.h>
+#include <linux/seq_buf.h>
 #include "security.h"
 #include "ima.h"
 
@@ -20,46 +21,31 @@
  */
 static char *selinux_ima_collect_state(void)
 {
-	const char *on = "=1;", *off = "=0;";
+	struct seq_buf s;
 	char *buf;
-	int buf_len, len, i, rc;
+	int buf_len, suffix_len, i;
 
 	buf_len = strlen("initialized=0;enforcing=0;checkreqprot=0;") + 1;
+	suffix_len = strlen("=0;");
 
-	len = strlen(on);
 	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
-		buf_len += strlen(selinux_policycap_names[i]) + len;
+		buf_len += strlen(selinux_policycap_names[i]) + suffix_len;
 
 	buf = kzalloc(buf_len, GFP_KERNEL);
 	if (!buf)
 		return NULL;
 
-	rc = strscpy(buf, "initialized", buf_len);
-	WARN_ON(rc < 0);
+	seq_buf_init(&s, buf, buf_len);
 
-	rc = strlcat(buf, selinux_initialized() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
+	seq_buf_printf(&s, "initialized=%d;enforcing=%d;checkreqprot=%d;",
+		       selinux_initialized(), enforcing_enabled(),
+		       checkreqprot_get());
 
-	rc = strlcat(buf, "enforcing", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, enforcing_enabled() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, "checkreqprot", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, checkreqprot_get() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	for (i = 0; i < __POLICYDB_CAP_MAX; i++) {
-		rc = strlcat(buf, selinux_policycap_names[i], buf_len);
-		WARN_ON(rc >= buf_len);
+	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
+		seq_buf_printf(&s, "%s=%d;", selinux_policycap_names[i],
+			       selinux_state.policycap[i]);
 
-		rc = strlcat(buf, selinux_state.policycap[i] ? on : off,
-			buf_len);
-		WARN_ON(rc >= buf_len);
-	}
+	WARN_ON(seq_buf_has_overflowed(&s));
 
 	return buf;
 }
#4 lore raw
Quoted 12 lines · David Laight
Quoted 9 lines · On Fri, 26 Jun 2026 10:57:10 -0500 Ian Bridges <[email protected]> wrote:

In preparation for removing the deprecated strlcat() API[1], replace the strscpy()/strlcat() chain in selinux_ima_collect_state() with a struct seq_buf, which tracks the write position and remaining space internally.

Each field is written with seq_buf_printf() using a "=%d;" format, which removes the open-coded "=1;"/"=0;" constants. The seven per-append WARN_ON(rc >= buf_len) truncation checks are replaced by a single seq_buf_has_overflowed() check after the string is built.

Quoted 8 lines
Quoted 5 lines

---
Changed in v2: replace the v1 seq_buf_puts() pairs with seq_buf_printf() using a "=%d;" format, which drops the open-coded "=1;"/"=0;" constants.

Did you verify that all the values are 0/1 otherwise a !! prefix might be needed.

Yes. selinux_initialized(), enforcing_enabled() and checkreqprot_get() all return bool, and selinux_state.policycap[] is a bool array. checkreqprot_get() always returns 0. The others read bool fields that are written only by normal assignment (WRITE_ONCE() or smp_store_release()), which normalizes the stored value to 0 or 1. So "%d" prints "0" or "1", the same as the old ? "=1;" : "=0;" mapping, and no !! is needed.

Quoted 10 lines
Quoted 9 lines

v1: https://lore.kernel.org/all/ajlN94VO7BYNUTAy@dev/

I didn't change the precomputation of the string size. An alternative, which is used by other seq_buf callers (e.g. kernel/rcu/refscale.c, mm/memcontrol.c), is to drop the precomputation and allocate an oversized fixed buffer, relying on the seq_buf overflow check as a backstop. I'm happy to rework the patch to adopt that alternative.

Quoted 7 lines

That would be reasonable, the output is always exactly the same length and the buffer is freed just after being allocated.
A comment that there are 3 + 15 strings and all are under 32 bytes so 1k is plenty would suffice.

David

I didn't originally take this route because a fixed buffer bakes in an invariant (the whole string must stay under 1K) with nothing to enforce it at compile time. If more capabilities are added later, or a name grows longer, the string could overflow. That shows up only at runtime, as the seq_buf_has_overflowed() WARN, and the measured selinux-state string is silently truncated.

Thanks for the review.

Ian

Quoted 76 lines
Quoted 75 lines
 security/selinux/ima.c | 40 +++++++++++++---------------------------
 1 file changed, 13 insertions(+), 27 deletions(-)
security/selinux/ima.c +13 -27
index aa34da9b0aeb..cb0efa2fc1ad 100644
@@ -9,6 +9,7 @@
  */
 #include <linux/vmalloc.h>
 #include <linux/ima.h>
+#include <linux/seq_buf.h>
 #include "security.h"
 #include "ima.h"
 
@@ -20,46 +21,31 @@
  */
 static char *selinux_ima_collect_state(void)
 {
-	const char *on = "=1;", *off = "=0;";
+	struct seq_buf s;
 	char *buf;
-	int buf_len, len, i, rc;
+	int buf_len, suffix_len, i;
 
 	buf_len = strlen("initialized=0;enforcing=0;checkreqprot=0;") + 1;
+	suffix_len = strlen("=0;");
 
-	len = strlen(on);
 	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
-		buf_len += strlen(selinux_policycap_names[i]) + len;
+		buf_len += strlen(selinux_policycap_names[i]) + suffix_len;
 
 	buf = kzalloc(buf_len, GFP_KERNEL);
 	if (!buf)
 		return NULL;
 
-	rc = strscpy(buf, "initialized", buf_len);
-	WARN_ON(rc < 0);
+	seq_buf_init(&s, buf, buf_len);
 
-	rc = strlcat(buf, selinux_initialized() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
+	seq_buf_printf(&s, "initialized=%d;enforcing=%d;checkreqprot=%d;",
+		       selinux_initialized(), enforcing_enabled(),
+		       checkreqprot_get());
 
-	rc = strlcat(buf, "enforcing", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, enforcing_enabled() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, "checkreqprot", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, checkreqprot_get() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	for (i = 0; i < __POLICYDB_CAP_MAX; i++) {
-		rc = strlcat(buf, selinux_policycap_names[i], buf_len);
-		WARN_ON(rc >= buf_len);
+	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
+		seq_buf_printf(&s, "%s=%d;", selinux_policycap_names[i],
+			       selinux_state.policycap[i]);
 
-		rc = strlcat(buf, selinux_state.policycap[i] ? on : off,
-			buf_len);
-		WARN_ON(rc >= buf_len);
-	}
+	WARN_ON(seq_buf_has_overflowed(&s));
 
 	return buf;
 }
#5 lore raw
Quoted 3 lines · On Wed, 1 Jul 2026 00:54:11 -0500 Ian Bridges <[email protected]> wrote:
Quoted 2 lines · David Laight

On Fri, 26 Jun 2026 10:57:10 -0500
Ian Bridges <[email protected]> wrote:

...

Quoted 17 lines
Quoted 9 lines
Quoted 8 lines

v1: https://lore.kernel.org/all/ajlN94VO7BYNUTAy@dev/

I didn't change the precomputation of the string size. An alternative, which is used by other seq_buf callers (e.g. kernel/rcu/refscale.c, mm/memcontrol.c), is to drop the precomputation and allocate an oversized fixed buffer, relying on the seq_buf overflow check as a backstop. I'm happy to rework the patch to adopt that alternative.

Quoted 7 lines

That would be reasonable, the output is always exactly the same length and the buffer is freed just after being allocated.
A comment that there are 3 + 15 strings and all are under 32 bytes so 1k is plenty would suffice.

David

Quoted 6 lines

I didn't originally take this route because a fixed buffer bakes in an invariant (the whole string must stay under 1K) with nothing to enforce it at compile time. If more capabilities are added later, or a name grows longer, the string could overflow. That shows up only at runtime, as the seq_buf_has_overflowed() WARN, and the measured selinux-state string is silently truncated.

You could use '32 * (3 + NUM_CAPAPILITIES)'.
IIRC the longest is 24 characters (plus 2 for the =[01]) so that will be plenty.

David

Quoted 81 lines

Thanks for the review.

Ian

Quoted 76 lines
Quoted 75 lines
 security/selinux/ima.c | 40 +++++++++++++---------------------------
 1 file changed, 13 insertions(+), 27 deletions(-)
security/selinux/ima.c +13 -27
index aa34da9b0aeb..cb0efa2fc1ad 100644
@@ -9,6 +9,7 @@
  */
 #include <linux/vmalloc.h>
 #include <linux/ima.h>
+#include <linux/seq_buf.h>
 #include "security.h"
 #include "ima.h"
 
@@ -20,46 +21,31 @@
  */
 static char *selinux_ima_collect_state(void)
 {
-	const char *on = "=1;", *off = "=0;";
+	struct seq_buf s;
 	char *buf;
-	int buf_len, len, i, rc;
+	int buf_len, suffix_len, i;
 
 	buf_len = strlen("initialized=0;enforcing=0;checkreqprot=0;") + 1;
+	suffix_len = strlen("=0;");
 
-	len = strlen(on);
 	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
-		buf_len += strlen(selinux_policycap_names[i]) + len;
+		buf_len += strlen(selinux_policycap_names[i]) + suffix_len;
 
 	buf = kzalloc(buf_len, GFP_KERNEL);
 	if (!buf)
 		return NULL;
 
-	rc = strscpy(buf, "initialized", buf_len);
-	WARN_ON(rc < 0);
+	seq_buf_init(&s, buf, buf_len);
 
-	rc = strlcat(buf, selinux_initialized() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
+	seq_buf_printf(&s, "initialized=%d;enforcing=%d;checkreqprot=%d;",
+		       selinux_initialized(), enforcing_enabled(),
+		       checkreqprot_get());
 
-	rc = strlcat(buf, "enforcing", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, enforcing_enabled() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, "checkreqprot", buf_len);
-	WARN_ON(rc >= buf_len);
-
-	rc = strlcat(buf, checkreqprot_get() ? on : off, buf_len);
-	WARN_ON(rc >= buf_len);
-
-	for (i = 0; i < __POLICYDB_CAP_MAX; i++) {
-		rc = strlcat(buf, selinux_policycap_names[i], buf_len);
-		WARN_ON(rc >= buf_len);
+	for (i = 0; i < __POLICYDB_CAP_MAX; i++)
+		seq_buf_printf(&s, "%s=%d;", selinux_policycap_names[i],
+			       selinux_state.policycap[i]);
 
-		rc = strlcat(buf, selinux_state.policycap[i] ? on : off,
-			buf_len);
-		WARN_ON(rc >= buf_len);
-	}
+	WARN_ON(seq_buf_has_overflowed(&s));
 
 	return buf;
 }  
maintainer
#6 lore raw
In reply to Ian Bridges
Quoted 17 lines · On Jun 26, 2026 Ian Bridges <[email protected]> wrote:

In preparation for removing the deprecated strlcat() API[1], replace the strscpy()/strlcat() chain in selinux_ima_collect_state() with a struct seq_buf, which tracks the write position and remaining space internally.

Each field is written with seq_buf_printf() using a "=%d;" format, which removes the open-coded "=1;"/"=0;" constants. The seven per-append WARN_ON(rc >= buf_len) truncation checks are replaced by a single seq_buf_has_overflowed() check after the string is built.

---
Changed in v2: replace the v1 seq_buf_puts() pairs with seq_buf_printf() using a "=%d;" format, which drops the open-coded "=1;"/"=0;" constants.

Quoted 7 lines

v1: https://lore.kernel.org/all/ajlN94VO7BYNUTAy@dev/

I didn't change the precomputation of the string size. An alternative, which is used by other seq_buf callers (e.g. kernel/rcu/refscale.c, mm/memcontrol.c), is to drop the precomputation and allocate an oversized fixed buffer, relying on the seq_buf overflow check as a backstop. I'm happy to rework the patch to adopt that alternative.

Another option would be to split out the length calculation and only do it once at boot, e.g. call a selinux_ima_calculate_len() function from selinux_init(), and then reuse that length in selinux_ima_collect_state().

What you're doing here is no worse than what we currently have so I'm going to go ahead and merge it into selinux/dev, but if you could split out the length calculation I think that would be worthwhile. Thanks!

Quoted 2 lines
 security/selinux/ima.c | 40 +++++++++++++---------------------------
 1 file changed, 13 insertions(+), 27 deletions(-)

--
paul-moore.com