BUG/MINOR: mux-h2: harden h2_dump_h2s_info() against potentially null h2s->sd

Not really sure if this can happen but since the code already takes care
of testing the sd, better make sure everywhere it's dereferenced it's
tested. And h2s_sc() does dereference h2s->sd which is later tested in
that function.

Under normal circumstances it's fine (when called from the stream via
h2_show_fd()) but it's less certain when dumping FDs via h2_show_fd()
which is called by the CLI's "show fd" debugging command. Let's just
guard all h2s_sc() under an sd test then.

Thanks to Ilya for reporting this Coverity finding in GH issue #3475.

This can be backported wherever it applies.

diff --git a/src/mux_h2.c b/src/mux_h2.c
index 937d5f4db8e4f2377ec54094a06469a7b7afc9ee..7704de822e564abdaaa7a294339f3df9b1c524ef 100644 (file)
--- a/src/mux_h2.c
+++ b/src/mux_h2.c
@@ -8534,10 +8534,13 @@ static int h2_dump_h2s_info(struct buffer *msg, const struct h2s *h2s, const cha
        if (pfx)
                chunk_appendf(msg, "\n%s", pfx);
 
-       chunk_appendf(msg, " .sc=%p", h2s_sc(h2s));
-       if (h2s_sc(h2s))
-               chunk_appendf(msg, "(.flg=0x%08x .app=%p)",
-                             h2s_sc(h2s)->flags, h2s_sc(h2s)->app);
+       if (h2s->sd) {
+               /* h2s_sc requires an sd! */
+               chunk_appendf(msg, " .sc=%p", h2s_sc(h2s));
+               if (h2s_sc(h2s))
+                       chunk_appendf(msg, "(.flg=0x%08x .app=%p)",
+                                     h2s_sc(h2s)->flags, h2s_sc(h2s)->app);
+       }
 
        chunk_appendf(msg, " .sd=%p", h2s->sd);
        if (h2s->sd)