248545 - PAC: DNS resolution failures not noticed

Status: VERIFIED DUPLICATE of bug 208287

(Core :: Networking, defect)

Description

[Jason Haar]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7b) Gecko/20040316
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7b) Gecko/20040316

We run Mozilla via "automatic proxy"/WPAD configuration. The WPAD file says to
resolve the hostname, and connect direct if it's an internal IP address, and
forward to the proxy if it's not.

Right now if I go to www.m23.org Mozilla will hang for over 2 minutes before
handing off the request to the proxy - which then hangs for only 5-10 secs
before giving up an error page.

Running tcpdump on my Fedora workstation shows

> 50496+ A? www.m23.org.
< 50496 ServFail 0/0/0

within miliseconds of each other - so I can't understand why Mozilla appears to
be ignoring the ServFail, and instead keeps asking the same A query - and
getting the same answer.

I also run iptables, so specifically disabled it - no difference.

Any ideas?

BTW: "time dig www.m23.org." returns "no servers could be reached" in 12 secs.


Reproducible: Always
Steps to Reproduce:
1. go to www.m23.org.
2.
3.

Actual Results:  
browser hangs. Totally. Looks to me like Mozilla blocks on DNS lookups?

Expected Results:  
I'd have thought reacting to the DNS error would have been the thing to do.

Comment 1

[R.K.Aa.]

dup of bug 239358?

Comment 2

[Christian :Biesinger (don't email me, ping me on IRC)]

jason, does your PAC file contain dnsResolve calls?

Comment 3

[Jason Haar]

(In reply to comment #2)
> jason, does your PAC file contain dnsResolve calls?

Yup. It's got bunches of isinnet, and dnsDomainIs calls.

It looks to see if the URL is internal (vis dnsdomainis), and resolves the name
to an IP address to see if the IP address is internal - then it goes DIRECT,
else PROXY.

Comment 4

[Christian :Biesinger (don't email me, ping me on IRC)]

caching negative dns results is bug 208312 (not sure whether that'd cover
ServFail, I suppose it would)

strangely I can't find a bug about dnsResolve blocking the UI thread... there's
bug 235853 and bug 208287... hm...

Comment 5

[Christian :Biesinger (don't email me, ping me on IRC)]

...but you really said you are using isInNet, so marking dup of that bug.

*** This bug has been marked as a duplicate of 208287 ***

Comment 6

[benc]

V/dupe.

althought I'm not sure it should be hanging as long as Jason describes... My DNS
is rusty, is 

> 50496+ A? www.m23.org.
< 50496 ServFail 0/0/0

that bug we have w/ IPv6 DNS queries?

Comment 7

[Christian :Biesinger (don't email me, ping me on IRC)]

(In reply to comment #6)
> > 50496+ A? www.m23.org.

I'd think that IPv6 would show up as AAAA? or something like this