339466 - SSL server cache watching thread is a problem for fork

Status: NEW

(NSS :: Libraries, defect, P2)

Description

[Nelson Bolyard (seldom reads bugmail)]

On unix and BEOS OSes, when an application calls SSL_ConfigMPServerSIDCache
to initialize a shared memory server session cache, to be shared among
server processes, SSL_ConfigMPServerSIDCache launches a "lock poller" 
thread.  This thread exists to monitor the locks in the shared memory, to 
detect locks that are held by sharing processes that have died, and free 
them, so the rest of the cache sharing processes can go on about their 
work.  This is only needed on OSes whose shared memory locks do not get
automatically released by the OS when the locker dies.  

The problem is that the existence of this lock poller thread is a problem
if the application forks after the lock poller thread is created.  Some
OSes really want any process that calls fork to have only one thread 
when it does so.  The lock poller thread is a problem for them.

This is an issue for NSS-based servers on Solaris.

NSS (libSSL) doesn't provide any way for an application to stop that rhread,
or explicitly restart it, or even ask if it is running.  

It has been suggested that the pthread_atfork method can be used to enable
libSSL to stop the thread just before forking, and restart it in the parent
just after forking.  That might help.  But we still want an explicit way to 
shut it down.  

Also, don't know if BEOS has anything like pthread_atfork.

Comment 1

[Julien Pierre]

Retargetting all P2s to 3.11.3 .

Comment 2

[Nelson Bolyard (seldom reads bugmail)]

remove target milestone, since the target was missed.

Comment 3

[BugBot [:suhaib / :marco/ :calixte]]

The bug assignee is inactive on Bugzilla, and this bug has priority 'P2'.
:beurdouche, could you have a look please?

For more information, please visit auto_nag documentation.

Comment 4

[Marco Castelluccio [:marco]]

We have modified the bot to only consider P1 as high priority, so I'm cancelling the needinfo here.