[nginx] SSL: support ALPN (IETF's successor to NPN).

Valentin V. Bartenev vbart at nginx.com
Thu Jan 30 13:06:02 UTC 2014


On Thursday 30 January 2014 16:45:18 Maxim Dounin wrote:
> Hello!
> 
> On Thu, Jan 30, 2014 at 01:55:23PM +0400, Ruslan Ermilov wrote:
> 
> > On Wed, Jan 29, 2014 at 03:51:50PM +0000, Valentin Bartenev wrote:
> > > details:   http://hg.nginx.org/nginx/rev/01e2a5bcdd8f
> > > branches:  
> > > changeset: 5545:01e2a5bcdd8f
> > > user:      Piotr Sikora <piotr at cloudflare.com>
> > > date:      Tue Jan 28 15:33:49 2014 -0800
> > > description:
> > > SSL: support ALPN (IETF's successor to NPN).
> > > 
> > > Signed-off-by: Piotr Sikora <piotr at cloudflare.com>
> > 
> > This change breaks compilation with clang on systems
> > where OpenSSL doesn't have ALPN support:
> > 
> > : cc -c -pipe  -O -Wall -Wextra -Wpointer-arith -Wconditional-uninitialized -Wno-unused-parameter -Werror -g  -I src/core -I src/event -I src/event/modules -I src/os/unix -I /opt/local/include -I objs -I src/http -I src/http/modules \
> > :                 -o objs/src/http/ngx_http_request.o \
> > :                 src/http/ngx_http_request.c
> > : src/http/ngx_http_request.c:728:44: error: variable 'data' may be uninitialized when used here [-Werror,-Wconditional-uninitialized]
> > :         if (len == spdy.len && ngx_strncmp(data, spdy.data, spdy.len) == 0) {
> > :                                            ^~~~
> > : src/core/ngx_string.h:53:56: note: expanded from macro 'ngx_strncmp'
> > : #define ngx_strncmp(s1, s2, n)  strncmp((const char *) s1, (const char *) s2, n)
> > :                                                        ^
> > : src/http/ngx_http_request.c:713:38: note: initialize the variable 'data' to silence this warning
> > :         const unsigned char     *data;
> > :                                      ^
> > :                                       = NULL
> > : 1 error generated.
> > : make[1]: *** [objs/src/http/ngx_http_request.o] Error 1
> > : make: *** [build] Error 2
> > 
> > clang(1) mistakenly thinks that "len" can be non-zero, and under
> > this condition "data" may be uninitialized.
> > 
> > It's possible to rewrite the code such as that it doesn't trigger
> > the warning:
> > 
> > diff --git a/src/http/ngx_http_request.c b/src/http/ngx_http_request.c
> > --- a/src/http/ngx_http_request.c
> > +++ b/src/http/ngx_http_request.c
> > @@ -713,11 +713,8 @@ ngx_http_ssl_handshake_handler(ngx_conne
> >          const unsigned char     *data;
> >          static const ngx_str_t   spdy = ngx_string(NGX_SPDY_NPN_NEGOTIATED);
> >  
> > -        len = 0;
> > -
> >  #ifdef TLSEXT_TYPE_application_layer_protocol_negotiation
> >          SSL_get0_alpn_selected(c->ssl->connection, &data, &len);
> > -#endif
> >  
> >  #ifdef TLSEXT_TYPE_next_proto_neg
> >          if (len == 0) {
> > @@ -725,6 +722,10 @@ ngx_http_ssl_handshake_handler(ngx_conne
> >          }
> >  #endif
> >  
> > +#else /* TLSEXT_TYPE_next_proto_neg */
> > +        SSL_get0_next_proto_negotiated(c->ssl->connection, &data, &len);
> > +#endif
> > +
> >          if (len == spdy.len && ngx_strncmp(data, spdy.data, spdy.len) == 0) {
> >              ngx_http_spdy_init(c->read);
> >              return;
> 
> That's clearly a false positive warning, and I would rather add "data = NULL"
> under NGX_SUPPRESS_WARN as we usually do, something like this:
[..]

Of course you have the final word, but I like Ruslan's solution.

  wbr, Valentin V. Bartenev



More information about the nginx-devel mailing list