[Varnish] #865: Opensolaris/snv_134 and others: CNT_Session - TCP_blocking failing with ECONNREFUSED (was: CNT_Session - TCP_blocking failing with ECONNREFUSED)

[Varnish]

#865: Opensolaris/snv_134 and others: CNT_Session - TCP_blocking failing with
ECONNREFUSED
----------------------+-----------------------------------------------------
 Reporter:  slink     |       Owner:  slink   
     Type:  defect    |      Status:  assigned
 Priority:  normal    |   Milestone:          
Component:  varnishd  |     Version:  2.1.4   
 Severity:  normal    |    Keywords:          
----------------------+-----------------------------------------------------

Comment(by slink):

 Status update:
 The issue also happens with tcp_random_anon_port set to 0 (see
 http://opensolaris.org/jive/thread.jspa?messageID=476738).

 If we're seeing a delayed error, then it does not seem to make sense to
 me: For the (reproducible) case in varnishtest, it would be an error from
 connect(), but the remote socket is on the local machine is definitely
 open.

 For the core dumps with the stack trace given in the initial description,
 ECONNREFUSED makes even less sense to me, because we have successfully
 accept()ed a connection.

 For completeness, here's the trace of the last system calls before the
 assertion error, but with the original ioctl code in place:

 {{{
 13584/4:        so_socket(PF_INET, SOCK_STREAM, IPPROTO_IP, 0x00000000,
 SOV_DEFAULT) = 9
 13584/4:        ioctl(9, FIONBIO, 0xFDA6DEBC)                   = 0
 13584/4:        connect(9, 0x080B5218, 16, SOV_DEFAULT)         Err#150
 EINPROGRESS
 13584/4:        pollsys(0xFDA6DF28, 1, 0xFDA6DED8, 0x00000000)  = 1
 13584/4:        ioctl(9, FIONBIO, 0xFDA6DF3C)                   Err#146
 ECONNREFUSED
 13584/4:        lwp_sigmask(SIG_SETMASK, 0xFFBFFEFF, 0x0000FFF7) =
 0xFFBFFEFF [0x0000FFFF]
 13584/4:
 open("/usr/lib/locale/en_US.UTF-8/LC_MESSAGES/SUNW_OST_SGS.mo", O_RDONLY)
 Err#2 ENOENT
 13584/4:        lwp_exit()
 }}}

 I must say I really don't understand this and I am tempted to close this
 bug as INVALID, because this does so much look like an OpenSolaris issue.

 Nils

-- 
Ticket URL: <http://www.varnish-cache.org/trac/ticket/865#comment:4>
Varnish <http://varnish-cache.org/>
The Varnish HTTP Accelerator