Segfault in 1.0.5.

Maxim Dounin mdounin at
Sat Jul 23 07:00:44 UTC 2011


On Sat, Jul 23, 2011 at 03:37:11PM +1000, Nigel Cunningham wrote:

> Hi everyone.
> I'm seeing a segfault in 1.0.5 in worker processes, most reliably caused
> when a visitor attempts to access what is actually an invalid URL
> (index.php/index.php?...). I can however also reproduce it on another
> vhost with valid URLs that fail occasionally and work when you retry.
> The segfault is always in the call to uscf->peer.init(r, uscf) (line 622
> in src/http/ngx_http_upstream.c) and occurs because uscf is not a valid
> pointer. Most of the time it is NULL, but I have also seen it be in the
> low positive integer range. The function itself is working fine, but
> r->upstream->conf->upstream is not getting set right, presumably at
> initialisation time.

You are hitting "/crash" example from

Maxim Dounin

More information about the nginx-devel mailing list