Re: [PATCH] Fix off by one in tools/perf strncpy size argument

[Arnaldo Carvalho de Melo]

Em Mon, Mar 09, 2020 at 09:39:40AM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Mon, Mar 09, 2020 at 11:48:53AM +0100, Dominik 'disconnect3d' Czarnota escreveu:
> > From: disconnect3d <dominik.b.czarnota@xxxxxxxxx>
> > 
> > This patch fixes an off-by-one error in strncpy size argument in
> > tools/perf/util/map.c. The issue is that in:
> > 
> >         strncmp(filename, "/system/lib/", 11)
> > 
> > the passed string literal: "/system/lib/" has 12 bytes (without the NULL
> > byte) and the passed size argument is 11. As a result, the logic won't
> > match the ending "/" byte and will pass filepaths that are stored in
> > other directories e.g. "/system/libmalicious/bin" or just
> > "/system/libmalicious".
> > 
> > This functionality seems to be present only on Android. I assume the
> > /system/ directory is only writable by the root user, so I don't
> > think this bug has much (or any) security impact.
> > 
> > Signed-off-by: disconnect3d <dominik.b.czarnota@xxxxxxxxx>
> > ---
> > 
> > Notes:
> >     I can't test this patch, so if someone can, please, do so.
> >     
> >     The bug could also be fixed by changing the size argument to `sizeof("string literal")-1` but I am not proposing this change as that would have to be changed in other places.
> 
> So, there are parts of this tools/perf/util/map.c that uses the idiom
> you mention, for instance:
> 
> static inline int is_anon_memory(const char *filename, u32 flags)
> {
>         return flags & MAP_HUGETLB ||
>                !strcmp(filename, "//anon") ||
>                !strncmp(filename, "/dev/zero", sizeof("/dev/zero") - 1) ||
>                !strncmp(filename, "/anon_hugepage", sizeof("/anon_hugepage") - 1);
> }
> 
> So I think we should make all cases use this idim to avoid these
> problems.
> 
> So I'll add your patch, then another, on top, that fixes the other
> off-by-one errors introduced by the android specific code in this patch:

This is the only such off-by-one in that file, and I remembered we have
strstarts(), that we adopted from the kernel sources, so I think its a
better fit, no?

[acme@five linux]$ find . -name "*.c" | grep -v tools/ | xargs grep -w strstarts | wc -l
55
[acme@five linux]$ find tools/ -name "*.c" | xargs grep -w strstarts | wc -l
40
[acme@five linux]$
 
> Fixes: eca818369996 ("perf tools: Add automatic remapping of Android libraries")
> 
> Put this in perf/urgent and then in perf/core move to the more robust
> idiom,
> 
> Thanks,
> 
> - Arnaldo
>      
> >     There are also more cases like this in kernel sources which I am going to report soon.
> >     
> >     Also please note that other path comparisons in this file lack the "/" at the end and it seems they may imply similar issue. I haven't analysed the code deeply to see if that is a real issue.
> >     
> >     This bug has been found by running a massive grep-like search using Google's BigQuery on GitHub repositories data. I am also going to work on a CodeQL/Semmle query to be able to find more sophisticated cases like this that can't be found via grepping.
> > 
> >  tools/perf/util/map.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c
> > index a08ca276098e..addd7edb0486 100644
> > --- a/tools/perf/util/map.c
> > +++ b/tools/perf/util/map.c
> > @@ -89,7 +89,7 @@ static inline bool replace_android_lib(const char *filename, char *newfilename)
> >  		return true;
> >  	}
> >  
> > -	if (!strncmp(filename, "/system/lib/", 11)) {
> > +	if (!strncmp(filename, "/system/lib/", 12)) {
> >  		char *ndk, *app;
> >  		const char *arch;
> >  		size_t ndk_length;
> > -- 
> > 2.25.1
> > 
> 
> -- 
> 
> - Arnaldo

-- 

- Arnaldo