FW: ibuf_empty delayed efd

Scott Neugroschl scott_n at xypro.com
Sat Jun 14 07:07:38 EST 2008

 On Wednesday, June 11, 2008 Damien Miller responded
> On Wed, 11 Jun 2008, Scott Neugroschl wrote:
> > > I'm not sure what is going wrong here, but could you please try a
> > > snapshot? We fixed a long-existing race in the efd handling @
> > > case and the fix may serendibitously fix your case :)
> > >
> > Is this race condition check the changes in
> channel_register_cleanup(),
> > where c->istate and c->ostate are being interrogated?
> >
> > I'd really like to avoid having to add all the changes in channel.c
> (the
> > queueing, changed external interfaces, etc...) if I could avoid it.
> I think this it is, though there may be other changes of relevance. It
> would probably be best if you tested a whole snapshot and then looked
> to
> extract specific changes if you find they work.

Found the issue.  There were two, actually, one of which is fixed in the
snapshots, and the
other was wholly mine, and doesn't affect the mainline code.

First issue was in do_exec_no_pty(), if USE_PIPES is not defined, the
socketpair isn't closed after
The dup2() calls.  This is corrected in the snapshots.

The other issue was a typo in some platform specific code I'd put into
which caused channel_close_efd(&c->efd) not to be called.

Corrected both these issues and Bob's your uncle!

Thanks for the help, Damien!


More information about the openssh-unix-dev mailing list