[PATCH] link_alt_odb_entry: fix read over array bounds reported by valgrind

[PATCH] link_alt_odb_entry: fix read over array bounds reported by valgrind

[Heiko Voigt]

pfxlen can be longer than the path in objdir when relative_base contains
the path to gits object directory.

Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
---
 sha1_file.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/sha1_file.c b/sha1_file.c
index 4ccaf7a..631d0dd 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -251,7 +251,7 @@ static int link_alt_odb_entry(const char * entry, int len, const char * relative
 	const char *objdir = get_object_directory();
 	struct alternate_object_database *ent;
 	struct alternate_object_database *alt;
-	int pfxlen, entlen;
+	int pfxlen, entlen, objdirlen;
 	struct strbuf pathbuf = STRBUF_INIT;
 
 	if (!is_absolute_path(entry) && relative_base) {
@@ -298,7 +298,8 @@ static int link_alt_odb_entry(const char * entry, int len, const char * relative
 			return -1;
 		}
 	}
-	if (!memcmp(ent->base, objdir, pfxlen)) {
+	objdirlen = strlen(objdir);
+	if (!memcmp(ent->base, objdir, pfxlen > objdirlen ? objdirlen : pfxlen)) {
 		free(ent);
 		return -1;
 	}
-- 
1.7.12.rc0.23.gc9a5ac4

Re: [PATCH] link_alt_odb_entry: fix read over array bounds reported by valgrind

[Junio C Hamano]

Heiko Voigt <hvoigt@hvoigt.net> writes:

> pfxlen can be longer than the path in objdir when relative_base contains
> the path to gits object directory.

s/gits/????/ perhaps "Git's", but I am not sure.

> Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
> ---
>  sha1_file.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/sha1_file.c b/sha1_file.c
> index 4ccaf7a..631d0dd 100644
> --- a/sha1_file.c
> +++ b/sha1_file.c
> @@ -251,7 +251,7 @@ static int link_alt_odb_entry(const char * entry, int len, const char * relative
>  	const char *objdir = get_object_directory();
>  	struct alternate_object_database *ent;
>  	struct alternate_object_database *alt;
> -	int pfxlen, entlen;
> +	int pfxlen, entlen, objdirlen;
>  	struct strbuf pathbuf = STRBUF_INIT;
>  
>  	if (!is_absolute_path(entry) && relative_base) {
> @@ -298,7 +298,8 @@ static int link_alt_odb_entry(const char * entry, int len, const char * relative
>  			return -1;
>  		}
>  	}
> -	if (!memcmp(ent->base, objdir, pfxlen)) {
> +	objdirlen = strlen(objdir);
> +	if (!memcmp(ent->base, objdir, pfxlen > objdirlen ? objdirlen : pfxlen)) {

The new code tells us to compare up to the shorter length between
objdir (i.e. path/to/.git/objects) and the given alternate object
directory (i.e. alt/path/to/.git/objects), but is that really what
we want?  What happens if the given alternate object directory were
"path/to/.git/objects-not-quite", with objdir "path/to/.git/objects"?

They are not the same directory, and this check is about avoiding
"the common mistake of listing ... object directory itself", no?

>  		free(ent);
>  		return -1;
>  	}

Re: [PATCH] link_alt_odb_entry: fix read over array bounds reported by valgrind

[Junio C Hamano]

Junio C Hamano <gitster@pobox.com> writes:

> Heiko Voigt <hvoigt@hvoigt.net> writes:
>
>> pfxlen can be longer than the path in objdir when relative_base contains
>> the path to gits object directory.
>
> s/gits/????/ perhaps "Git's", but I am not sure.
>
>> Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
>> ---
>>  sha1_file.c | 5 +++--
>>  1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/sha1_file.c b/sha1_file.c
>> index 4ccaf7a..631d0dd 100644
>> --- a/sha1_file.c
>> +++ b/sha1_file.c
>>  			return -1;
>>  		}
>>  	}
>> -	if (!memcmp(ent->base, objdir, pfxlen)) {
>> +	objdirlen = strlen(objdir);
>> +	if (!memcmp(ent->base, objdir, pfxlen > objdirlen ? objdirlen : pfxlen)) {
>
> The new code tells us to compare up to the shorter length between
> objdir (i.e. path/to/.git/objects) and the given alternate object
> directory (i.e. alt/path/to/.git/objects), but is that really what
> we want?  What happens if the given alternate object directory were
> "path/to/.git/objects-not-quite", with objdir "path/to/.git/objects"?
>
> They are not the same directory, and this check is about avoiding
> "the common mistake of listing ... object directory itself", no?
>
>>  		free(ent);
>>  		return -1;
>>  	}

In other words, wouldn't this be sufficient?  We NUL terminate
ent->base[pfxlen] when we prepare that buffer with

	LEADING PATH\0XX/XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX\0

in preparation for these "duplicate check" step, and then we turn
the NUL at ent->base[pfxlen] to '/' before leaving the function to
make it

	LEADING PATH/XX/XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX\0

so that we can fill XX when probing for loose objects.

 sha1_file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sha1_file.c b/sha1_file.c
index 4f06a0e..a1f3bee 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -298,7 +298,7 @@ static int link_alt_odb_entry(const char * entry, int len, const char * relative
 			return -1;
 		}
 	}
-	if (!memcmp(ent->base, objdir, pfxlen)) {
+	if (!strcmp(ent->base, objdir)) {
 		free(ent);
 		return -1;
 	}

[PATCH v2] link_alt_odb_entry: fix read over array bounds reported by valgrind

[Heiko Voigt]

pfxlen can be longer than the path in objdir when relative_base contains
the path to Git's object directory.

Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
---

On Sun, Jul 29, 2012 at 05:54:02PM -0700, Junio C Hamano wrote:
> Heiko Voigt <hvoigt@hvoigt.net> writes:
> 
> > pfxlen can be longer than the path in objdir when relative_base contains
> > the path to gits object directory.
> 
> s/gits/????/ perhaps "Git's", but I am not sure.

Git's is correct.

> > @@ -298,7 +298,8 @@ static int link_alt_odb_entry(const char * entry, int len, const char * relative
> >  			return -1;
> >  		}
> >  	}
> > -	if (!memcmp(ent->base, objdir, pfxlen)) {
> > +	objdirlen = strlen(objdir);
> > +	if (!memcmp(ent->base, objdir, pfxlen > objdirlen ? objdirlen : pfxlen)) {
> 
> The new code tells us to compare up to the shorter length between
> objdir (i.e. path/to/.git/objects) and the given alternate object
> directory (i.e. alt/path/to/.git/objects), but is that really what
> we want?  What happens if the given alternate object directory were
> "path/to/.git/objects-not-quite", with objdir "path/to/.git/objects"?
> 
> They are not the same directory, and this check is about avoiding
> "the common mistake of listing ... object directory itself", no?

You are right. strcmp is the correct solution for this. I tried your
change and it also fixes the error valgrind reported.

 sha1_file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sha1_file.c b/sha1_file.c
index 4ccaf7a..af5cfbd 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -298,7 +298,7 @@ static int link_alt_odb_entry(const char * entry, int len, const char * relative
 			return -1;
 		}
 	}
-	if (!memcmp(ent->base, objdir, pfxlen)) {
+	if (!strcmp(ent->base, objdir)) {
 		free(ent);
 		return -1;
 	}
-- 
1.7.12.rc0.24.g3fa2a49