However, the API is not going to make this trivial to fix. ap_get_remote_host is connection-based. And that is what mod_authz_host is currently relying upon. It seems that there needs to be a way for mod_remoteip to override the existing behavior, perhaps ap_set_remote_host(), that will cache the request-based on for the lifetime of the request pool. In the request pool cleanup, ap_set_remote_host(c, NULL) would clear that overridden request-based host, popping the value back to the cached c-> fields. There is also the issue of the timing of setting the scoreboard record. All three issues are intertwined. --------- Original Message --------- Subject: Re: mod_status: Apache 2.4 incorrect IP (proxy, not useragent_ip) on server-status page From: "Jim Jagielski" <[email protected]> Date: 9/11/14 9:46 am To: [email protected]
Yeah, the more I think about it, ap_get_remote_host() is currently broken wrt how it handles useragent_ip and client_ip. Will likely try to patch this on trunk sometime today... On Sep 11, 2014, at 9:35 AM, Martynas Bendorius <[email protected]> wrote: > Yes, we may re-phrase it like that, if we'd like to fix it in apache source > (and not documentation) :) Currently ap_get_remote_host in server/core.c > doesn't return useragent_ip, and instead of it we get conn->client_ip. > > Best regards, > Martynas Bendorius > > On 9/11/14 4:21 PM, Jim Jagielski wrote: >> isn't the question rather "What should ap_get_remote_host() >> return?"? >> >> On Sep 11, 2014, at 8:17 AM, Martynas Bendorius <[email protected]> >> wrote: >> >>> Hello, >>> >>> Would it be possible to change the documentation of mod_remoteip for 2.4 >>> (http://httpd.apache.org/docs/2.4/mod/mod_remoteip.html), and get "is >>> reported by mod_status" removed from the page? As it leds Apache customers >>> to believe that it will report a real (useragent) IP instead of a proxy >>> one in server-status page. useragent_ip is not even available in >>> scoreboard, which is used by mod_status, so it's not available for >>> mod_status. >>> >>> This has been already discussed here: >>> https://issues.apache.org/bugzilla/show_bug.cgi?id=55886 >>> >>> Thank you! >>> >>> Best regards, >>> Martynas Bendorius >>> >> >
