Skip to contentSkip to conversation

Potential NULL pointer deference in drbg_ctr_df

PATCH In review Stephan Mueller 4 participants 6 messages Jun 21, 2014 – Jul 5, 2014 2 messages not in the archive Crypto API

lore
Share & tools
lore
#1 lore raw

PATCH 1/1 Potential NULL pointer deference in drbg_ctr_df

In reply to a message not in the archive

The handling of additional input data / personalization string data may be subject to a NULL pointer deference for the CTR DRBG. The caller-provided data may be NULL which must be caught by the DRBG.

---
 crypto/drbg.c | 23 +++++++++++++----------
 1 file changed, 13 insertions(+), 10 deletions(-)
crypto/drbg.c +13 -10
index faaa2ce..8e7c302 100644
@@ -513,17 +513,20 @@ static int drbg_ctr_df(struct drbg_state *drbg,
 	drbg_string_fill(&S2, L_N, sizeof(L_N));
 	drbg_string_fill(&S4, pad, padlen);
 	S1.next = &S2;
-	S2.next = addtl;
 
-	/*
-	 * splice in addtl between S2 and S4 -- we place S4 at the end of the
-	 * input data chain
-	 */
-	tempstr = addtl;
-	for (; NULL != tempstr; tempstr = tempstr->next)
-		if (NULL == tempstr->next)
-			break;
-	tempstr->next = &S4;
+	if (NULL == addtl) {
+		S2.next = &S4;
+	} else {
+		/*
+		 * splice in addtl between S2 and S4 -- we place S4 at the end
+		 * of the input data chain
+		 */
+		S2.next = addtl;
+		tempstr = addtl;
+		while (tempstr->next)
+			tempstr = tempstr->next;
+		tempstr->next = &S4;
+	}
 
 	/* 10.4.2 step 9 */
 	while (templen < (drbg_keylen(drbg) + (drbg_blocklen(drbg)))) {
Signature · 1 line
1.9.3
maintainer
#2 lore raw
Quoted 31 lines · Stephan Mueller

The handling of additional input data / personalization string data may be subject to a NULL pointer deference for the CTR DRBG. The caller-provided data may be NULL which must be caught by the DRBG.

---
 crypto/drbg.c | 23 +++++++++++++----------
 1 file changed, 13 insertions(+), 10 deletions(-)
crypto/drbg.c +2 -10
index faaa2ce..8e7c302 100644
@@ -513,17 +513,20 @@ static int drbg_ctr_df(struct drbg_state *drbg,
 	drbg_string_fill(&S2, L_N, sizeof(L_N));
 	drbg_string_fill(&S4, pad, padlen);
 	S1.next = &S2;
-	S2.next = addtl;
 
-	/*
-	 * splice in addtl between S2 and S4 -- we place S4 at the end of the
-	 * input data chain
-	 */
-	tempstr = addtl;
-	for (; NULL != tempstr; tempstr = tempstr->next)
-		if (NULL == tempstr->next)
-			break;
-	tempstr->next = &S4;
+	if (NULL == addtl) {
+		S2.next = &S4;

crypto/drbg.c static int drbg_ctr_df(struct drbg_state *drbg, · line 528

Quoted 10 lines
crypto/drbg.c +10 -0
+	} else {
+		/*
+		 * splice in addtl between S2 and S4 -- we place S4 at the end
+		 * of the input data chain
+		 */
+		S2.next = addtl;
+		tempstr = addtl;
+		while (tempstr->next)
+			tempstr = tempstr->next;
+		tempstr->next = &S4;

You've still got exactly the same NULL dereference.

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
maintainer
#7 lore raw

Re: [cryptodev:master 9/28] crypto/drbg.c:526 drbg_ctr_df() error: we previously assumed 'tempstr' could be null (see line 523)

In reply to a message not in the archive
Quoted 10 lines · Stephan Mueller

Am Freitag, 20. Juni 2014, 23:24:18 schrieb kbuild test robot:

Hi,

first of all, thanks for the information. As you have seen, I created a patch covering your comments on the NULL pointer handling.

Quoted 3 lines

The kernel also provides standard linked list features and you should consider using those instead of rolling your own.

Quoted 7 lines

There is a reason for this approach: the very same code (with exception of the interface code part) is also available for libgcrypt -- see [1] and gcrypt- devel. I did not want to have different code, because any bug fix relevant to the one implementation is also important to the other. Thus, I feel that it is beneficial to use the linked list feature as implemented in the current DRBG.

[1] http://www.chronox.de/drbg.html

This is unacceptable. We are not going to add crap to the kernel just because the same crap exists elsewhere.

Please fix this or I will be reverting your patches.

Thanks,

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#3 lore raw
In reply to Herbert Xu
Quoted 6 lines · Herbert Xu
Quoted 5 lines · Stephan Mueller

The handling of additional input data / personalization string data may be subject to a NULL pointer deference for the CTR DRBG. The caller-provided data may be NULL which must be caught by the DRBG.

Oh, huh. This bug was actually reported by me but I forgot to change the from header and apparently my smtp server allows me to send emails as if I work for Intel. :P

Fengguang is much nicer than I am. :P

Quoted 30 lines
Quoted 30 lines
---
 crypto/drbg.c | 23 +++++++++++++----------
 1 file changed, 13 insertions(+), 10 deletions(-)
crypto/drbg.c +6 -10
index faaa2ce..8e7c302 100644
@@ -513,17 +513,20 @@ static int drbg_ctr_df(struct drbg_state *drbg,
 	drbg_string_fill(&S2, L_N, sizeof(L_N));
 	drbg_string_fill(&S4, pad, padlen);
 	S1.next = &S2;
-	S2.next = addtl;
 
-	/*
-	 * splice in addtl between S2 and S4 -- we place S4 at the end of the
-	 * input data chain
-	 */
-	tempstr = addtl;
-	for (; NULL != tempstr; tempstr = tempstr->next)
-		if (NULL == tempstr->next)
-			break;
-	tempstr->next = &S4;
+	if (NULL == addtl) {
+		S2.next = &S4;
+	} else {
+		/*
+		 * splice in addtl between S2 and S4 -- we place S4 at the end
+		 * of the input data chain
Quoted 8 lines

crypto/drbg.c static int drbg_ctr_df(struct drbg_state *drbg, · line 528

Quoted 6 lines
crypto/drbg.c +6 -0
+		 */
+		S2.next = addtl;
+		tempstr = addtl;
+		while (tempstr->next)
+			tempstr = tempstr->next;
+		tempstr->next = &S4;

You've still got exactly the same NULL dereference.

I was offline for a bit so I'm coming into this late. It's weird that Stephan isn't defending his patch but it looks ok to me...

regards,
dan carpenter

#4 lore raw

Am Freitag, 4. Juli 2014, 13:50:03 schrieb Dan Carpenter:

Hi Dan,

Quoted 6 lines · Herbert Xu
Quoted 5 lines · Stephan Mueller
Quoted 4 lines

The handling of additional input data / personalization string data may be subject to a NULL pointer deference for the CTR DRBG. The caller-provided data may be NULL which must be caught by the DRBG.

Quoted 7 lines
Quoted 1 line
Quoted 1 line

Oh, huh. This bug was actually reported by me but I forgot to change the from header and apparently my smtp server allows me to send emails as if I work for Intel. :P

Fengguang is much nicer than I am. :P

Too late, your colleague's name is now in the patches ;-)

Quoted 43 lines
Quoted 34 lines
Quoted 34 lines

---

 crypto/drbg.c | 23 +++++++++++++----------
 1 file changed, 13 insertions(+), 10 deletions(-)
crypto/drbg.c +6 -10
index faaa2ce..8e7c302 100644
@@ -513,17 +513,20 @@ static int drbg_ctr_df(struct drbg_state *drbg,
 
 	drbg_string_fill(&S2, L_N, sizeof(L_N));
 	drbg_string_fill(&S4, pad, padlen);
 	S1.next = &S2;
 
-	S2.next = addtl;
 
-	/*
-	 * splice in addtl between S2 and S4 -- we place S4 at the end of the
-	 * input data chain
-	 */
-	tempstr = addtl;
-	for (; NULL != tempstr; tempstr = tempstr->next)
-		if (NULL == tempstr->next)
-			break;
-	tempstr->next = &S4;
+	if (NULL == addtl) {
+		S2.next = &S4;
+	} else {
+		/*
+		 * splice in addtl between S2 and S4 -- we place S4 at the end
+		 * of the input data chain
Quoted 8 lines

crypto/drbg.c static int drbg_ctr_df(struct drbg_state *drbg, · line 528

Quoted 6 lines
crypto/drbg.c +6 -0
+		 */
+		S2.next = addtl;
+		tempstr = addtl;
+		while (tempstr->next)
+			tempstr = tempstr->next;
+		tempstr->next = &S4;

You've still got exactly the same NULL dereference.

Quoted 2 lines

I was offline for a bit so I'm coming into this late. It's weird that Stephan isn't defending his patch but it looks ok to me...

It has been clarified in a later email that there was no NULL pointer after all. I erroneously sent this message and initially missed the check that guards against the NULL pointer.

Quoted 3 lines

regards,
dan carpenter

Signature · 2 lines
Ciao
Stephan
#5 lore raw
Quoted 12 lines · Stephan Mueller

Am Freitag, 4. Juli 2014, 13:50:03 schrieb Dan Carpenter:

Hi Dan,

Quoted 8 lines · Herbert Xu
Quoted 6 lines · Stephan Mueller
Quoted 5 lines

The handling of additional input data / personalization string data may be subject to a NULL pointer deference for the CTR DRBG. The caller-provided data may be NULL which must be caught by the DRBG.

Quoted 7 lines
Quoted 5 lines

Oh, huh. This bug was actually reported by me but I forgot to change the from header and apparently my smtp server allows me to send emails as if I work for Intel. :P

Fengguang is much nicer than I am. :P

Too late, your colleague's name is now in the patches ;-)

Thank you nice guys! :P

Thanks,
Fengguang