I have received already multiple reports [1] of crashes in 2.86 version.
I thought commit de372d6914ae20a1f9997815f258efbf3b14c39b might fix that
issue, but even more reports arrived with version including this commit
[2] [3].

I do not have any coredump available for analysis (yet), but bugs
contain automated backtraces including some variables. It seems
crop_query operation can overflow to some wrong value. Then qlen gets
outside of original domain name char buffer and process crashes.

I did not find the failing place yet. I suspect order() function might
not always enforce qlen > daemon->serverarray[try]->domain_len. But
without coredump file I am unsure.

Thread 1 (LWP 3842):

#0  0x0000558a903382be in lookup_domain (domain=0x558a91fd7cd0 
"dns.msftncsi.com", flags=128, lowout=0x7fff41395fe8, highout=0x7fff41395fe4) 
at /usr/src/debug/dnsmasq-2.86-2.fc35.x86_64/src/domain-match.c:234
        rc = <optimized out>
        crop_query = <optimized out>
        nodots = 0
        qlen = 51543
        try = <optimized out>
        high = <optimized out>
        low = <optimized out>
        nlow = 0
        nhigh = 0
        cp = <optimized out>
        qdomain = 0x558a91fcb389 <error: Cannot access memory at address 
0x558a91fcb389>
#1  0x0000558a90308a2a in forward_query (udpfd=5, udpaddr=0x7fff41396170, 
dst_addr=0x7fff41396140, dst_iface=5, header=0x558a91fd9e30, plen=34, 
limit=0x558a91fda030 "", now=1633667814, forward=0x0, ad_reqd=0, do_bit=0) at 
/usr/src/debug/dnsmasq-2.86-2.fc35.x86_64/src/forward.c:258
        flags = 0
        fwd_flags = 0
        is_dnssec = 0
        master = <optimized out>
        gotname = 128
        hash = 0x558a91fd9b70
        oph = 0x0
        old_src = 0
        first = 12
        last = 0
        start = 0
        subnet = 0
        cacheable = 0
        forwarded = 0
        edns0_len = 94053643165266
        pheader = 0x558a91fd9e52 ""
        ede = -1

Is there any commit after v2.86, which should make this fixed? Does
anyone else met a crash like this and can provide more debugging?
Especially useful would be a way to reproduce. It seems like nothing
very special is required to trigger this bug.

Any thoughts? Fixes maybe?

Cheers,
Petr


1. https://bugzilla.redhat.com/show_bug.cgi?id=2009975
2. https://bugzilla.redhat.com/show_bug.cgi?id=2012151
3. https://bugzilla.redhat.com/show_bug.cgi?id=2012547

-- 
Petr Menšík
Software Engineer
Red Hat, http://www.redhat.com/
email: pemen...@redhat.com
PGP: DFCF908DB7C87E8E529925BC4931CA5B6C9FC5CB
_______________________________________________
Dnsmasq-discuss mailing list
Dnsmasq-discuss@lists.thekelleys.org.uk
https://lists.thekelleys.org.uk/cgi-bin/mailman/listinfo/dnsmasq-discuss

Reply via email to