tinc - Jul 2014 - Error while waiting for input: Bad file descriptor

Hello,

Thanks to recent fix of 'Failed to decrypt and verify packet' issue by
Etienne, I decided to upgrade tinc on a few nodes.
I had to go back to 1.1pre9 immediately because tinc on my laptop didn't
survive the restart of other nodes. It aborted with the following message:
  Error while waiting for input: Bad file descriptor

I could reproduce the issue. Here is strace output:

select(14, [4 5 6 7 8 9 10 12 13], [], NULL, {0, 377947}) = 0 (Timeout)
sendto(3, "<30>Jul  2 01:39:52 tinc.jmuchem"..., 107,
MSG_NOSIGNAL, NULL, 0) = 107
sendto(3, "<29>Jul  2 01:39:52 tinc.jmuchem"..., 95,
MSG_NOSIGNAL, NULL, 0) = 95
sendto(3, "<31>Jul  2 01:39:52 tinc.jmuchem"..., 106,
MSG_NOSIGNAL, NULL, 0) = 106
sendto(3, "<31>Jul  2 01:39:52 tinc.jmuchem"..., 100,
MSG_NOSIGNAL, NULL, 0) = 100
close(12)                         = 0
sendto(3, "<30>Jul  2 01:39:52 tinc.jmuchem"..., 90,
MSG_NOSIGNAL, NULL, 0) = 90
socket(PF_INET, SOCK_STREAM, IPPROTO_TCP) = 12
fcntl(12, F_GETFL)                = 0x2 (flags O_RDWR)
fcntl(12, F_SETFL, O_RDWR|O_NONBLOCK) = 0
setsockopt(12, SOL_TCP, TCP_NODELAY, [1], 4) = 0
setsockopt(12, SOL_IP, IP_TOS, [16], 4) = 0
fcntl(12, F_SETFD, FD_CLOEXEC)    = 0
connect(12, {sa_family=AF_INET, sin_port=htons(655),
sin_addr=inet_addr("192.168.1.1")}, 16) = -1 ENETUNREACH (Network is
unreachable)
sendto(3, "<27>Jul  2 01:39:52 tinc.jmuchem"..., 114,
MSG_NOSIGNAL, NULL, 0) = 114
close(12)                         = 0
sendto(3, "<27>Jul  2 01:39:52 tinc.jmuchem"..., 84,
MSG_NOSIGNAL, NULL, 0) = 84
sendto(3, "<29>Jul  2 01:39:52 tinc.jmuchem"..., 97,
MSG_NOSIGNAL, NULL, 0) = 97
wait4(-1, NULL, WNOHANG, NULL)    = -1 ECHILD (No child processes)
sendto(3, "<31>Jul  2 01:39:52 tinc.jmuchem"..., 87,
MSG_NOSIGNAL, NULL, 0) = 87
sendto(3, "<31>Jul  2 01:39:52 tinc.jmuchem"..., 99,
MSG_NOSIGNAL, NULL, 0) = 99
select(14, [4 5 6 7 8 9 10 12 13], [], NULL, {1, 683412}) = -1 EBADF (Bad file
descriptor)
sendto(3, "<27>Jul  2 01:39:52 tinc.jmuchem"..., 92,
MSG_NOSIGNAL, NULL, 0) = 92

On 07/02/2014 01:05 PM, Julien Muchembled wrote:
> Thanks to recent fix of 'Failed to decrypt and verify packet' issue
by Etienne, I decided to upgrade tinc on a few nodes.
> I had to go back to 1.1pre9 immediately because tinc on my laptop
didn't survive the restart of other nodes. It aborted with the following
message:
>    Error while waiting for input: Bad file descriptor
I probably won't have time to investigate this before the week-end, but 
you might want to indicate the commit you're building from. It would 
also be helpful to have debug level 4 (-d4) output mixed with the strace 
output so that we can see what tinc thinks it's doing while this is 
happening. In addition, more log history would probably be useful.

-- 
Etienne Dechamps

On 02/07/2014 13:05, Julien Muchembled wrote:
> Hello,
>
> Thanks to recent fix of 'Failed to decrypt and verify packet' issue
by Etienne, I decided to upgrade tinc on a few nodes.
> I had to go back to 1.1pre9 immediately because tinc on my laptop
didn't survive the restart of other nodes. It aborted with the following
message:
>    Error while waiting for input: Bad file descriptor
Never mind, I managed to reproduce. It seems like this happens every 
time there is a timeout.

For anyone interested: http://pastebin.com/nvbxSzWJ

-- 
Etienne Dechamps