1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
|
Version 1.3
- fix AUTOSSH_DEBUG for Solaris and AIX
- fix attempt to free() static storage (affected platforms without
getaddrinfo() -- mostly cygwin) (Andrew Schulman)
- change test and placement of typedef for socklen_t under OS X; new
OS X defines it
- add ability to signal autossh to kill and restart ssh child, using
SIGUSR1
- add hostname to monitor message (Ron Yorston)
- check on looping on poll() where connection has been lost (spinning
and high CPU)
- fix bug where length argument to accept() was not initialised
- fix arg parsing bug where stripping -f from arguments would strip
from a parameter to the argument: e.g -L8808:foo:80 would become
-L8808:oo:80 (pointed out by Eric Larson)
- pull out r/w loops in conn_test() into separate functions, makes
logic more apparent
- add echo port support: have the remote server use the inetd
echo service to echo our test string back to us. Or use some other
echo service. Idea and patch from Ron Yorston. This makes it 1.3.
- remove bad strcpy() (left over from some testing?) thanks to Ron
Yorston, change to memset read buffer to all zeros
- fix ssh args when AUTOSSH_PORT=0 is used to turn monitor loop off
(Karl Berry)
- add more descriptive usage output, by popular request
Version 1.2g
- add AUTOSSH_NTSERVICE (Andrew Schulman)
- fix bad calculation for seconds left in poll time (again from
Andrew Schulman)
- from Andrew Schulman: add support for older networking and cygwin
- add AUTOSSH_MAXSTART (from Hugo Haas)
- loop around waitpid() in ssh_kill() in case interrupted (thanks to
Jens Krabbenhoeft)
- update ssh argument string
- move openlog above port error checking
- handle environment variables and -M arg being set to the
empty string (reported by Dan Christensen via Filippo Giunche)
- add some rudimetary auto-adjust to the network timeouts for
low poll times. So by default 15secs on each of accept() and
poll()to deal with high-latency connections, but as poll time
falls below 30secs, then adjust timeouts down as well.
- adjust division in grace_time() to allow for low poll time
- don't call shutdown() and close() on invalid socket ()
(found by Dmitry V. Levin)
Version 1.2f
- by popular request, support the -f flag by doing the fork/background
ourselves, and not passing the flag to ssh (thanks to Dan Christensen
for prodding me into it)
- change timeout to from 5000 to 15000 ms to deal with reported issues
with high-latency links; unused var removal; and man page typo (thanks
to Michael Shields)
Version 1.2e
- check for attempt to use -f flag; exit as error
- if AUTOSSH_GATETIME is 0, then disable any of the startup
failure bailouts, and just retry
- set SO_REUSEADDR on listening socket (Jedi One's suggestion)
- show port number in some error messages
- note in man page and README about -M 0 turning port monitoring off
- remove duplicate include of sys/socket.h
Version 1.2d
- AIX support (thanks to Stefan Rodenstein)
- fix argument rewrite bug when using AUTOSSH_PORT (thanks
to Peter Williams)
- log pid of ssh child just after fork
Version 1.2c
- use Marcus Friedl's suggestion to simply connect to "127.0.0.1"
rather than "localhost", as not really using IPv6 anyway. And this
gets rid of annoying ipv6 localhost message on Solaris at least.
- support for MacOS X using Greg Parker's fakepoll.
Version 1.2b
(Thanks to Simon Easter for bug reports and trials)
- fix file descriptor leak
- setsockopt()'s don't work for Solaris either. Give up on them.
- set close-on-exec for read socket so will be closed when
ssh executed
- do shutdown() before close() (paranoia)
- close read socket before exit
- pull read socket open and close up out of ssh_run()
- cosmetic changes to some loops, get rid of newlines in some
errlog strings.
Version 1.2a
- setsockopt() will not set timeouts on socket read/write ops
under Linux 2.2 kernels (and earlier, I presume). So unless
someone tells me we really need them, I've #ifdef'd the
setsockopt()s out for Linux.
- check value of fd returned by accept().
- Oh, there's a man page now. Unfortunately, it doesn't render
for Solaris. Sorry, Solaris users are still stuck with the
README.
Version 1.2
- Major restructuring of code, mostly to support the change to
a loop of forwarded ports to test the network connection.
- Incremental back off on rate of connection attempts when
there are rapid failures to establish/maintain a connection.
Version 1.1b
- change handling of ssh exit status
- introduce "starting gate" time: ssh process that dies too
early is deemed not to have made it out of the starting
gate, and autossh exits.
Version 1.1a
- fix race after failure to exec ssh where parent will
loop attempting to restart
- add -V version flag
- use strtoul() for environment option parsing
- drop useless intermediate function ssh_setwatch()
Version 1.1
Initial release
|