Re: [GNUnet-developers] [PATCH] namecache: fix NULL-pointer dereference in namecache-flat

[Daniel Golle]

Hi,

I had to set
[namecache]
DATABASE = flat

[namecache-flat]
FILENAME = '/path/to/somewhere'

to make sure it actually starts the namecache service with the
flat plugin loaded...

I can file a bug, however, discussing code (as the fix is trivial)
is more conveniant on the mailing list imho as I can use
git send-email and other conveniance tools instead of copy-pasting
stuff into a web-form...

Looking at the test-case, it's quite clear it won't work, see the
patch making it work:
diff --git a/src/namecache/test_plugin_namecache_flat.conf 
b/src/namecache/test_plugin_namecache_flat.conf
index 2909d76..8b18c3b 100644
--- a/src/namecache/test_plugin_namecache_flat.conf
+++ b/src/namecache/test_plugin_namecache_flat.conf
@@ -1,2 +1,2 @@
-[namestore-flat]
-FILENAME = /tmp/gnunet-test-plugin-namestore-sqlite/flatdb
+[namecache-flat]
+FILENAME = /tmp/gnunet-test-plugin-namecache-flat/flatdb


Cheers


Daniel
On Fri, Dec 04, 2015 at 10:42:04PM +0100, Martin Schanzenbach wrote:
> Hi,
> 
> This is strange since the test should cover this and does not fail (or
> crash) for me (TM).
> I don't think we need a parsergenerator to parse csv files, though. As
> long as it works, of course.
> Can you please file a bug instead of posting it here? Thanks!
> 
> - Martin
> 
> On Fri, 2015-12-04 at 22:11 +0100, Daniel Golle wrote:
> > namecache-flat crashes if started on an empty or non-existent file.
> > gnunet-service-[2568]: segfault at 0 ip 00007fe32dc8bb3c sp
> > 00007ffde11826f0 error 4 in
> > libgnunet_plugin_namecache_flat.so[7fe32dc8a000+202000]
> > gnunet-service-[2576]: segfault at 0 ip 00007f2a6be75b3c sp
> > 00007ffdf9b6d880 error 4 in
> > libgnunet_plugin_namecache_flat.so[7f2a6be74000+202000]
> > gnunet-service-[2581]: segfault at 0 ip 00007ff1dcfb1b3c sp
> > 00007fff834070f0 error 4 in
> > libgnunet_plugin_namecache_flat.so[7ff1dcfb0000+202000]
> > 
> > This fix at least prevents parsing a 0-byte file, however, looking at
> > the code it is clear that more needs to be done to prevent crashes in
> > case of corrupted or incompletely written cache data.
> > Are you sure it's at all a good idea to implement parsers using
> > standard C string functions...? (what about bison and flex?)
> > 
> > Signed-off-by: Daniel Golle <address@hidden>
> > ---
> >  src/namecache/plugin_namecache_flat.c | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/src/namecache/plugin_namecache_flat.c
> > b/src/namecache/plugin_namecache_flat.c
> > index 409b058..85bb692 100644
> > --- a/src/namecache/plugin_namecache_flat.c
> > +++ b/src/namecache/plugin_namecache_flat.c
> > @@ -132,6 +132,9 @@ database_setup (struct Plugin *plugin)
> >      return GNUNET_SYSERR;
> >    }
> >  
> > +  if (0 == size)
> > +    return GNUNET_OK;
> > +
> >    buffer = GNUNET_malloc (size);
> >  
> >    if (GNUNET_SYSERR == GNUNET_DISK_file_read (fh,