RE: [PATCH] replace strncopy by memcpy

RE: [PATCH] replace strncopy by memcpy

[VAUTRIN Emmanuel (Canal Plus Prestataire)]

Hello Lukáš

Thank you for your patch, but, even if I have not checked the 
related sources, I really doubt about your fix. Maybe it will not
be a problem, due to the following check:
if ((uptr + ulen) > uncomp_end)
 			goto out;

> memcpy is equivalent to strncpy here as pointed by Santtu.
Please be careful, no more warning does not always imply that
the problem has been fixed. The warning has only disappear here
because you have replaced the call to strncpy by another one.
memcpy and  strncpy are not equivalent.
strncpy is dedicated to string processing, and enables to avoid not
null terminated strings, when the string length is equal to the destination
buffer size.
The good fix is just to replace, in the strncpy call,  ulen by  ulen -1 or
strlen(name).
If you really want to propose a patch, please refer to the commits
(sources code + commit message) I mentioned
(90f0e412569b327d8b25b834623b814f1fddbd19
& 3dbcc963a66d43e4ac51c97de4fe116ec72eb9f6),
it will be a better approach.


Best Regards,

Emmanuel

Re: [PATCH] replace strncopy by memcpy

[Santtu Lakkala]

On 18.6.2021 15.38, VAUTRIN Emmanuel (Canal Plus Prestataire) wrote:
> Thank you for your patch, but, even if I have not checked the
> related sources, I really doubt about your fix. Maybe it will not
> be a problem, due to the following check:
> if ((uptr + ulen) > uncomp_end)
>   			goto out;
> 
>> memcpy is equivalent to strncpy here as pointed by Santtu.
> Please be careful, no more warning does not always imply that
> the problem has been fixed. The warning has only disappear here
> because you have replaced the call to strncpy by another one.
> memcpy and  strncpy are not equivalent.
> strncpy is dedicated to string processing, and enables to avoid not
> null terminated strings, when the string length is equal to the destination
> buffer size.
> The good fix is just to replace, in the strncpy call,  ulen by  ulen -1 or
> strlen(name).
> If you really want to propose a patch, please refer to the commits
> (sources code + commit message) I mentioned
> (90f0e412569b327d8b25b834623b814f1fddbd19
> & 3dbcc963a66d43e4ac51c97de4fe116ec72eb9f6),
> it will be a better approach.

This is a different warning, this warning is about using strlen(src) as 
strncpy() length argument. I see a few possible fixes to acknowledge the 
warning:
  1) replace strncpy with memcpy (as done in this patch, no change in 
behavior, only theoretically slightly more performant)
  2) replace strncpy with strcpy (available space is already checked)
  3) replace strncpy length argument with available buffer space (would 
unnecessarily zero the buffer to the end)
  4) change the whole paradigm, and do something like:
     uptr += snprintf(uptr, uncomp_end - uptr, "%s", name);
     if (uptr >= uncomp_end)
         goto out;
  5) do 4 with strlcpy, but we don't have that

-- 
Santtu

Re: [PATCH] replace strncopy by memcpy

[Daniel Wagner]

Hi,

On Fri, Jun 18, 2021 at 08:53:25PM +0300, Santtu Lakkala wrote:
> On 18.6.2021 15.38, VAUTRIN Emmanuel (Canal Plus Prestataire) wrote:
> > Thank you for your patch, but, even if I have not checked the
> > related sources, I really doubt about your fix. Maybe it will not
> > be a problem, due to the following check:
> > if ((uptr + ulen) > uncomp_end)
> >   			goto out;
> > 
> > > memcpy is equivalent to strncpy here as pointed by Santtu.
> > Please be careful, no more warning does not always imply that
> > the problem has been fixed. The warning has only disappear here
> > because you have replaced the call to strncpy by another one.
> > memcpy and  strncpy are not equivalent.
> > strncpy is dedicated to string processing, and enables to avoid not
> > null terminated strings, when the string length is equal to the destination
> > buffer size.
> > The good fix is just to replace, in the strncpy call,  ulen by  ulen -1 or
> > strlen(name).
> > If you really want to propose a patch, please refer to the commits
> > (sources code + commit message) I mentioned
> > (90f0e412569b327d8b25b834623b814f1fddbd19
> > & 3dbcc963a66d43e4ac51c97de4fe116ec72eb9f6),
> > it will be a better approach.
> 
> This is a different warning, this warning is about using strlen(src) as
> strncpy() length argument. I see a few possible fixes to acknowledge the
> warning:
>  1) replace strncpy with memcpy (as done in this patch, no change in
> behavior, only theoretically slightly more performant)
>  2) replace strncpy with strcpy (available space is already checked)
>  3) replace strncpy length argument with available buffer space (would
> unnecessarily zero the buffer to the end)
>  4) change the whole paradigm, and do something like:
>     uptr += snprintf(uptr, uncomp_end - uptr, "%s", name);
>     if (uptr >= uncomp_end)
>         goto out;
>  5) do 4 with strlcpy, but we don't have that

I've applied the patch. While option 4) is probably a bit nicer to read,
the rest of the function uses memcpy. So let's use memcpy for the first
operation as well at least for the time being.

If someone wants to rewrite the dnsproxy code to be more robust please
give it a go. We had at least 3 CVE in dnsproxy and I woulnd't be
surprised if there are more sever bugs hidden.

Thanks a lot!
Daniel

[PATCH] replace strncopy by memcpy

[Lukáš Karas]

[-- Attachment #1.1: Type: text/plain, Size: 390 bytes --]

This patch fixes build with GCC 10.3.0, that fails with warning/error:

  '__builtin_strncpy' specified bound depends on the length
  of the source argument [-Werror=stringop-overflow=]

Destination buffer size is is checked properly,
memcpy is equivalent to strncpy here as pointed by Santtu.

This patch replaces my previous one:
"dnsproxy: Suppress GCC warning stringop-overflow"

Lukas

[-- Attachment #1.2: 0001-replace-strncopy-by-memcpy.patch --]
[-- Type: text/x-patch, Size: 956 bytes --]

From 3dca8ec863d60a3bfd81baeb56255742c242ad93 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Luk=C3=A1=C5=A1=20Karas?= <lukas.karas@centrum.cz>
Date: Fri, 18 Jun 2021 11:09:04 +0200
Subject: [PATCH] replace strncopy by memcpy

Destination buffer size is is checked properly,
but GCC 10.3.0 analysis fails to detect it and prints stringop-overflow warning.
Memcpy is equivalent to strncpy here, because ulen = strlen(name) + 1.
---
 src/dnsproxy.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/dnsproxy.c b/src/dnsproxy.c
index 38dbdd71..4b65b379 100644
--- a/src/dnsproxy.c
+++ b/src/dnsproxy.c
@@ -1791,7 +1791,7 @@ static char *uncompress(int16_t field_count, char *start, char *end,
 		ulen = strlen(name) + 1;
 		if ((uptr + ulen) > uncomp_end)
 			goto out;
-		strncpy(uptr, name, ulen);
+		memcpy(uptr, name, ulen);
 
 		debug("pos %d ulen %d left %d name %s", pos, ulen,
 			(int)(uncomp_end - (uptr + ulen)), uptr);
-- 
2.27.0


[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 488 bytes --]