Skip to contentSkip to conversation

bpf: btf: Fix vsnprintf return value check

PATCH In review Fedor Tokarev 4 participants 6 messages Jul 11, 2022 – Jul 25, 2022 BPF

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bpf: btf: Fix vsnprintf return value check

vsnprintf returns the number of characters which would have been written if enough space had been available, excluding the terminating null byte. Thus, the return value of 'len_left' means that the last character has been dropped.

---
 kernel/bpf/btf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/bpf/btf.c +1 -1
index eb12d4f705cc..a9c1c98017d4 100644
@@ -6519,7 +6519,7 @@ static void btf_snprintf_show(struct btf_show *show, const char *fmt,
 	if (len < 0) {
 		ssnprintf->len_left = 0;
 		ssnprintf->len = len;
-	} else if (len > ssnprintf->len_left) {
+	} else if (len >= ssnprintf->len_left) {
 		/* no space, drive on to get length we would have written */
 		ssnprintf->len_left = 0;
 		ssnprintf->len += len;
Signature · 1 line
2.25.1
#1 lore raw
Quoted 4 lines · Fedor Tokarev

vsnprintf returns the number of characters which would have been written if enough space had been available, excluding the terminating null byte. Thus, the return value of 'len_left' means that the last character has been dropped.

should we have test for this in progs/test_snprintf.c ?

jirka

Quoted 22 lines
---
 kernel/bpf/btf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/bpf/btf.c +1 -1
index eb12d4f705cc..a9c1c98017d4 100644
@@ -6519,7 +6519,7 @@ static void btf_snprintf_show(struct btf_show *show, const char *fmt,
 	if (len < 0) {
 		ssnprintf->len_left = 0;
 		ssnprintf->len = len;
-	} else if (len > ssnprintf->len_left) {
+	} else if (len >= ssnprintf->len_left) {
 		/* no space, drive on to get length we would have written */
 		ssnprintf->len_left = 0;
 		ssnprintf->len += len;
Signature · 1 line
2.25.1
#2 lore raw
Quoted 8 lines · Jiri Olsa
Quoted 4 lines · Fedor Tokarev

vsnprintf returns the number of characters which would have been written if enough space had been available, excluding the terminating null byte. Thus, the return value of 'len_left' means that the last character has been dropped.

should we have test for this in progs/test_snprintf.c ?

It might be too annoying to set up such test, and given the fix is pretty trivial IMO it's ok without extra test. But cc Alan for ack. Alan, please take a look as well.

Quoted 25 lines

jirka

Quoted 22 lines
---
 kernel/bpf/btf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/bpf/btf.c +1 -2
index eb12d4f705cc..a9c1c98017d4 100644
@@ -6519,7 +6519,7 @@ static void btf_snprintf_show(struct btf_show *show, const char *fmt,
      if (len < 0) {
              ssnprintf->len_left = 0;
              ssnprintf->len = len;
-     } else if (len > ssnprintf->len_left) {
+     } else if (len >= ssnprintf->len_left) {
              /* no space, drive on to get length we would have written */
              ssnprintf->len_left = 0;
              ssnprintf->len += len;
--

2.25.1

Acked-by
#3 lore raw
Quoted 8 lines · Andrii Nakryiko
Quoted 7 lines · Jiri Olsa
Quoted 4 lines · Fedor Tokarev

vsnprintf returns the number of characters which would have been written if enough space had been available, excluding the terminating null byte. Thus, the return value of 'len_left' means that the last character has been dropped.

Quoted 6 lines
Quoted 1 line

should we have test for this in progs/test_snprintf.c ?

It might be too annoying to set up such test, and given the fix is pretty trivial IMO it's ok without extra test. But cc Alan for ack. Alan, please take a look as well.

I can follow up with a test, it should be okay I think (we can use the "don't show types" flag and tryp to print "10" with a 2-byte len or similar).

In terms of the fix, it looks good, but given that the code is tricky, it might be good to expand a bit on the explanation. Something like the below?

"When using btf_type_snprintf_show(), the user passes in a "len" value, and we use it to initialize ssnprintf.len_left, indicating how much space remains for the string representation, including the null byte, so "len - 1" bytes are actually available for the actual string data, leaving one for the terminating null byte.

In btf_snprintf_show() - which is passed the ssnprintf data as an argument - vsnprintf() returns the len that would have been written, and this _excludes_ the null terminator. But we want to handle cases where the length of the string to be written (excluding the null terminator) exactly matches the original len value we passed in (len == len_left) in the same way was we do other overflow cases (len > len_left)."

Quoted 25 lines
Quoted 25 lines

jirka

Quoted 22 lines
---
 kernel/bpf/btf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/bpf/btf.c +1 -2
index eb12d4f705cc..a9c1c98017d4 100644
@@ -6519,7 +6519,7 @@ static void btf_snprintf_show(struct btf_show *show, const char *fmt,
      if (len < 0) {
              ssnprintf->len_left = 0;
              ssnprintf->len = len;
-     } else if (len > ssnprintf->len_left) {
+     } else if (len >= ssnprintf->len_left) {
              /* no space, drive on to get length we would have written */
              ssnprintf->len_left = 0;
              ssnprintf->len += len;
--

2.25.1

#4 lore raw
Quoted 11 lines · Alan Maguire
Quoted 10 lines · Andrii Nakryiko
Quoted 8 lines · Jiri Olsa
Quoted 4 lines · Fedor Tokarev

vsnprintf returns the number of characters which would have been written if enough space had been available, excluding the terminating null byte. Thus, the return value of 'len_left' means that the last character has been dropped.

should we have test for this in progs/test_snprintf.c ?

Quoted 8 lines
Quoted 4 lines

It might be too annoying to set up such test, and given the fix is pretty trivial IMO it's ok without extra test. But cc Alan for ack. Alan, please take a look as well.

I can follow up with a test, it should be okay I think (we can use the "don't show types" flag and tryp to print "10" with a 2-byte len or similar).

I'll gladly give it a try.

Quoted 3 lines

In terms of the fix, it looks good, but given that the code is tricky, it might be good to expand a bit on the explanation. Something like the below?

Agreed.

Quoted 40 lines

"When using btf_type_snprintf_show(), the user passes in a "len" value, and we use it to initialize ssnprintf.len_left, indicating how much space remains for the string representation, including the null byte, so "len - 1" bytes are actually available for the actual string data, leaving one for the terminating null byte.

In btf_snprintf_show() - which is passed the ssnprintf data as an argument - vsnprintf() returns the len that would have been written, and this _excludes_ the null terminator. But we want to handle cases where the length of the string to be written (excluding the null terminator) exactly matches the original len value we passed in (len == len_left) in the same way was we do other overflow cases (len > len_left)."

Quoted 25 lines
Quoted 25 lines

jirka

Quoted 22 lines
---
 kernel/bpf/btf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/bpf/btf.c +1 -2
index eb12d4f705cc..a9c1c98017d4 100644
@@ -6519,7 +6519,7 @@ static void btf_snprintf_show(struct btf_show *show, const char *fmt,
      if (len < 0) {
              ssnprintf->len_left = 0;
              ssnprintf->len = len;
-     } else if (len > ssnprintf->len_left) {
+     } else if (len >= ssnprintf->len_left) {
              /* no space, drive on to get length we would have written */
              ssnprintf->len_left = 0;
              ssnprintf->len += len;
--

2.25.1

#5 lore raw
Quoted 17 lines · Fedor Tokarev
Quoted 16 lines · Alan Maguire
Quoted 14 lines · Andrii Nakryiko
Quoted 8 lines · Jiri Olsa
Quoted 4 lines · Fedor Tokarev

vsnprintf returns the number of characters which would have been written if enough space had been available, excluding the terminating null byte. Thus, the return value of 'len_left' means that the last character has been dropped.

should we have test for this in progs/test_snprintf.c ?

It might be too annoying to set up such test, and given the fix is pretty trivial IMO it's ok without extra test. But cc Alan for ack. Alan, please take a look as well.

Quoted 5 lines
Quoted 3 lines

I can follow up with a test, it should be okay I think (we can use the "don't show types" flag and tryp to print "10" with a 2-byte len or similar).

I'll gladly give it a try.

Thanks! I've sent

https://lore.kernel.org/bpf/[email protected]/

If you could give it a try that would be great; tested at my end with your fix and all works well. I'd suggest pulling it into a 2-patch series comprised of your fix + the selftest, but since the fix targets bpf and the tests are new (so would be more like a bpf-next candidate), not sure if that's the right way to handle this..

If not I can follow up with the test once the fix lands.

Anyway thanks again for finding and fixing this!

Alan

Quoted 46 lines
Quoted 4 lines

In terms of the fix, it looks good, but given that the code is tricky, it might be good to expand a bit on the explanation. Something like the below?

Agreed.

Quoted 40 lines

"When using btf_type_snprintf_show(), the user passes in a "len" value, and we use it to initialize ssnprintf.len_left, indicating how much space remains for the string representation, including the null byte, so "len - 1" bytes are actually available for the actual string data, leaving one for the terminating null byte.

In btf_snprintf_show() - which is passed the ssnprintf data as an argument - vsnprintf() returns the len that would have been written, and this _excludes_ the null terminator. But we want to handle cases where the length of the string to be written (excluding the null terminator) exactly matches the original len value we passed in (len == len_left) in the same way was we do other overflow cases (len > len_left)."

Quoted 25 lines
Quoted 25 lines

jirka

Quoted 22 lines
---
 kernel/bpf/btf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/bpf/btf.c +1 -2
index eb12d4f705cc..a9c1c98017d4 100644
@@ -6519,7 +6519,7 @@ static void btf_snprintf_show(struct btf_show *show, const char *fmt,
      if (len < 0) {
              ssnprintf->len_left = 0;
              ssnprintf->len = len;
-     } else if (len > ssnprintf->len_left) {
+     } else if (len >= ssnprintf->len_left) {
              /* no space, drive on to get length we would have written */
              ssnprintf->len_left = 0;
              ssnprintf->len += len;
--

2.25.1