DNS hangs on invalid domain

[daurnimator]

This hangs:

print(require"cqueues.socket".connect("wrong.invalid", 80):connect())

Whereas dig returns NXDOMAIN almost instantly.

Relevant strace output of lua sample (exited during hang):

socket(AF_INET, SOCK_DGRAM, IPPROTO_IP) = 4
fcntl(4, F_SETFD, FD_CLOEXEC)           = 0
fcntl(4, F_GETFL)                       = 0x2 (flags O_RDWR)
fcntl(4, F_SETFL, O_RDWR|O_NONBLOCK)    = 0
open("/dev/urandom", O_RDONLY|O_NOCTTY|O_NONBLOCK) = 5
fstat(5, {st_mode=S_IFCHR|0666, st_rdev=makedev(1, 9), ...}) = 0
poll([{fd=5, events=POLLIN}], 1, 10)    = 1 ([{fd=5, revents=POLLIN}])
read(5, "\241\253\337\243A\241\fs\320\363(\255\366\2505\346\364\215\25\233\262\325YN\312\201\324\21\352\255\177\304", 32) = 32
close(5)                                = 0
getuid()                                = 1000
bind(4, {sa_family=AF_INET, sin_port=htons(33483), sin_addr=inet_addr("0.0.0.0")}, 16) = 0
brk(0x967000)                           = 0x967000
connect(4, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.1.1.1")}, 16) = 0
sendto(4, "\263+\1\0\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\1\0\1", 31, 0, NULL, 0) = 31
recvfrom(4, 0x947064, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
recvfrom(4, 0x947064, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
epoll_create1(EPOLL_CLOEXEC)            = 5
eventfd2(0, EFD_CLOEXEC|EFD_NONBLOCK)   = 6
epoll_ctl(5, EPOLL_CTL_ADD, 6, {EPOLLIN, {u32=9729320, u64=9729320}}) = 0
write(6, "\1\0\0\0\0\0\0\0", 8)         = 8
epoll_wait(5, [{EPOLLIN, {u32=9729320, u64=9729320}}], 32, 0) = 1
epoll_ctl(5, EPOLL_CTL_ADD, 4, {EPOLLIN, {u32=9730880, u64=9730880}}) = 0
read(6, "\1\0\0\0\0\0\0\0", 8)          = 8
epoll_wait(5, [{EPOLLIN, {u32=9730880, u64=9730880}}], 32, -1) = 1
write(6, "\1\0\0\0\0\0\0\0", 8)         = 8
epoll_ctl(5, EPOLL_CTL_DEL, 4, 0x7ffff26147f0) = 0
recvfrom(4, "\263+\201\203\0\1\0\0\0\1\0\0\5wrong\7invalid\0\0\1\0\1\0"..., 768, 0, NULL, NULL) = 106
connect(4, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.1.1.1")}, 16) = 0
sendto(4, "\372\372\1\0\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\0\1\0\1", 32, 0, NULL, 0) = 32
recvfrom(4, 0x947da4, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
epoll_wait(5, [{EPOLLIN, {u32=9729320, u64=9729320}}], 32, 0) = 1
epoll_ctl(5, EPOLL_CTL_ADD, 4, {EPOLLIN, {u32=9730880, u64=9730880}}) = 0
read(6, "\1\0\0\0\0\0\0\0", 8)          = 8
epoll_wait(5, ^Cstrace: Process 8661 detached

[wahern]

Can you capture the entire reply? Also, can you paste your /etc/resolv.conf? I can't reproduce, but I think the problem is related to search generation.

In trying to reproduce an environment that can reproduce the issue I discovered that your patch (4d66661) to cqs_newmetatable broke dns.config. The :set method added to the config userdata object is lost, and so dns.config.new(t) doesn't work. In fact, I'm surprised everything else appears to work, and can't figure out how I missed the problem on review.

[daurnimator]

It doesn't seem to replicate since I restarted my computer. I had previously moved between various networks.... so I'm not completely sure what the state was.

In general if you send a query to a non-existant server it hangs. But I don't think that's the same issue

sudo ip addr add 10.1.1.5/24 dev wlp6s0 # add a new subnet that doesn't have any devices on it
echo nameserver 10.1.1.1 | sudo tee /etc/resolv.conf # replace resolv.conf with non-listening dns server
strace lua -e 'require"cqueues.auxlib" print(require"cqueues.socket".connect("wrong.invalid", 80):connect())' # Replicated?

• in the strace above I can see data was received.

[daurnimator]

Closing as I can't replicate

[daurnimator]

This might have just cropped up again.

$ strace lua -e 'require"cqueues.auxlib"; print(require"cqueues.socket".connect({host="wrong.invalid", port=80}):connect())'
<snip>
open("/etc/resolv.conf", O_RDONLY|O_CLOEXEC) = 3
fstat(3, {st_mode=S_IFREG|0644, st_size=69, ...}) = 0
lseek(3, 0, SEEK_SET)                   = 0
read(3, "# Generated by resolvconf\nnamese"..., 4096) = 69
read(3, "", 4096)                       = 0
close(3)                                = 0
open("/etc/nsswitch.conf", O_RDONLY|O_CLOEXEC) = 3
fstat(3, {st_mode=S_IFREG|0644, st_size=234, ...}) = 0
read(3, "# Begin /etc/nsswitch.conf\n\npass"..., 4096) = 234
read(3, "", 4096)                       = 0
read(3, "", 4096)                       = 0
read(3, "", 4096)                       = 0
close(3)                                = 0
open("/etc/hosts", O_RDONLY|O_CLOEXEC)  = 3
fstat(3, {st_mode=S_IFREG|0644, st_size=195, ...}) = 0
lseek(3, 0, SEEK_SET)                   = 0
read(3, "#\n# /etc/hosts: static lookup ta"..., 4096) = 195
brk(0x20d3000)                          = 0x20d3000
read(3, "", 4096)                       = 0
close(3)                                = 0
socket(AF_INET, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, IPPROTO_IP) = 3
open("/dev/urandom", O_RDONLY|O_NOCTTY|O_NONBLOCK) = 4
fstat(4, {st_mode=S_IFCHR|0666, st_rdev=makedev(1, 9), ...}) = 0
poll([{fd=4, events=POLLIN}], 1, 10)    = 1 ([{fd=4, revents=POLLIN}])
read(4, "A\6\215`r\342X\254\321z\rp\316|\267foGi\2}\376\t\217m\210\302z\223\4\200\354", 32) = 32
close(4)                                = 0
getuid()                                = 1000
bind(3, {sa_family=AF_INET, sin_port=htons(55374), sin_addr=inet_addr("0.0.0.0")}, 16) = 0
connect(3, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.1.2.1")}, 16) = 0
sendto(3, "_\252\1\0\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\1\0\1", 31, 0, NULL, 0) = 31
recvfrom(3, 0x20b501c, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
recvfrom(3, 0x20b501c, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
epoll_create1(EPOLL_CLOEXEC)            = 4
eventfd2(0, EFD_CLOEXEC|EFD_NONBLOCK)   = 5
epoll_ctl(4, EPOLL_CTL_ADD, 5, {EPOLLIN, {u32=34297176, u64=34297176}}) = 0
write(5, "\1\0\0\0\0\0\0\0", 8)         = 8
epoll_wait(4, [{EPOLLIN, {u32=34297176, u64=34297176}}], 32, 0) = 1
epoll_ctl(4, EPOLL_CTL_ADD, 3, {EPOLLIN, {u32=34299040, u64=34299040}}) = 0
read(5, "\1\0\0\0\0\0\0\0", 8)          = 8
epoll_wait(4, [{EPOLLIN, {u32=34299040, u64=34299040}}], 32, -1) = 1
write(5, "\1\0\0\0\0\0\0\0", 8)         = 8
epoll_ctl(4, EPOLL_CTL_DEL, 3, 0x7ffedda261d0) = 0
recvfrom(3, "_\252\201\203\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\1\0\1", 768, 0, NULL, NULL) = 31
connect(3, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.1.2.1")}, 16) = 0
sendto(3, "\254\304\1\0\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\0\1\0\1", 32, 0, NULL, 0) = 32
recvfrom(3, 0x20b5e0c, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
epoll_wait(4, [{EPOLLIN, {u32=34297176, u64=34297176}}], 32, 0) = 1
epoll_ctl(4, EPOLL_CTL_ADD, 3, {EPOLLIN, {u32=34299040, u64=34299040}}) = 0
read(5, "\1\0\0\0\0\0\0\0", 8)          = 8
epoll_wait(4, ^Cstrace: Process 15454 detached
 <detached ...>
lua: /usr/share/lua/5.3/cqueues.lua:77: interrupted!

$ cat /etc/resolv.conf
# Generated by resolvconf
nameserver 10.1.2.1
nameserver 192.168.1.1

It never got to the 2nd nameserver. Removing that from resolv.conf doesn't seem to fix anything.

Passing a family argument (of either 2 or 10) to connect doesn't seem to help.

[daurnimator]

I think this might be easier to replicate with just dns.c.
Though I'm not 100% confident this is exhibiting the same behaviour.

$ gcc -D DNS_DEBUG -D DNS_MAIN dns.c
$ ./a.out  -q wrong.invalid show-resconf
; SOURCES
;   /etc/resolv.conf
;   /etc/nsswitch.conf
;
nameserver 10.1.2.1
search .
; hosts: files dns
lookup file bind
options ndots:1 timeout:5 attempts:2
interface 0.0.0.0 0
$ ./a.out -vvvv -q wrong.invalid addrinfo-stub
@@ BEGIN * * * * * * * * * * * *
@@ ASKING: hints.local./10.1.2.1 @ DEPTH: 0)
;; [HEADER]
;;    qid : 53039
;;     qr : QUERY(0)
;; opcode : QUERY(0)
;;     aa : NON-AUTHORITATIVE(0)
;;     tc : NOT-TRUNCATED(0)
;;     rd : RECURSION-DESIRED(1)
;;     ra : RECURSION-NOT-ALLOWED(0)
;;  rcode : NOERROR(0)

;; [QUESTION:1]
;wrong.invalid. IN A
@@ END * * * * * * * * * * * * *

@@ BEGIN * * * * * * * * * * * *
@@ ANSWER @ DEPTH: 0)
;; [HEADER]
;;    qid : 53039
;;     qr : RESPONSE(1)
;; opcode : QUERY(0)
;;     aa : NON-AUTHORITATIVE(0)
;;     tc : NOT-TRUNCATED(0)
;;     rd : RECURSION-DESIRED(1)
;;     ra : RECURSION-ALLOWED(1)
;;  rcode : NXDOMAIN(3)

;; [QUESTION:1]
;wrong.invalid. IN A
@@ END * * * * * * * * * * * * *

@@ BEGIN * * * * * * * * * * * *
@@ ASKING: hints.local./10.1.2.1 @ DEPTH: 0)
;; [HEADER]
;;    qid : 51133
;;     qr : QUERY(0)
;; opcode : QUERY(0)
;;     aa : NON-AUTHORITATIVE(0)
;;     tc : NOT-TRUNCATED(0)
;;     rd : RECURSION-DESIRED(1)
;;     ra : RECURSION-NOT-ALLOWED(0)
;;  rcode : NOERROR(0)

;; [QUESTION:1]
;wrong.invalid. 256 0
@@ END * * * * * * * * * * * * *

^C

[daurnimator]

I still have this issue semi-regularly.

 $ strace -e network lua -e 'print(require"cqueues.socket".connect("wrong.invalid", 80):connect())'
socket(AF_INET, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, IPPROTO_IP) = 3
bind(3, {sa_family=AF_INET, sin_port=htons(53399), sin_addr=inet_addr("0.0.0.0")}, 16) = 0
connect(3, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.1.1.2")}, 16) = 0
sendto(3, "\276\234\1\0\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\1\0\1", 31, 0, NULL, 0) = 31
recvfrom(3, 0x2494a2c, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
recvfrom(3, 0x2494a2c, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
recvfrom(3, "\276\234\201\203\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\1\0\1", 768, 0, NULL, NULL) = 31
connect(3, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.1.1.2")}, 16) = 0
sendto(3, "se\1\0\0\1\0\0\0\0\0\0\5wrong\7invalid\0\0\0\1\0\1", 32, 0, NULL, 0) = 32
recvfrom(3, 0x249d26c, 768, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
^Cstrace: Process 24159 detached

Looking at this in wireshark was more educational. The second dns query is invalid:
as hex:

7365010000010000000000000577726f6e6707696e76616c6964000000010001

Which is an invalid type and an invalid class.

Which we should have noticed in the dump in my earlier comment...

> ;; [QUESTION:1]
> ;wrong.invalid. 256 0

[daurnimator]

@wahern are you able to look into this issue with the dump above? A user aside from myself seems to be running into it. daurnimator/lua-http#87