IRC, freenode, #hurd, 2011-03-28
<pinotree> basically, i'm trying to implement socket credentials for local
sockets, and i guessed doing it in pflocal would be the appropriate place
<pinotree> what i thought was filling the cmsg data for MSG_CRED at
<pinotree> in case i missed it, would there be a way to "identify" the
other side of the port associated to the sock_user of that call?
<pochu> pinotree: that's needed by dbus right? cool! (and I don't know)
<pinotree> (yes, and gamin)
<youpi> pinotree: you have them already, they're just not stored
<youpi> see S_io_reauthenticate
<youpi> Throw away the ids we went through all that trouble to get...
* pinotree looks
<pinotree> hm, and who calls that rpc?
<youpi> since that's how ext2fs knows the permission to apply, for instance
<pinotree> ah, i was referring to the reauthenticate of pflocal, not
<youpi> that's what I'm saying
<youpi> see __hurd_file_name_lookup_retry, which is the very internal part
<youpi> it calls io_reauthenticate()
<youpi> to authenticate itself to the underlying translator of the opened
<pinotree> youpi: so, hm, could be an option make the result of pflocal's
S_io_reauthenticate cached in the sock_user struct?
<pinotree> nice thanks, i will try that change first
IRC, OFTC, #debian-hurd, 2013-02-20
<pinotree> youpi: while debugging #700530, it seems that xorg does not have
working socket credentials on kfreebsd (and hurd too)
<pinotree> julien provided sune with
http://people.debian.org/~jcristau/kbsd-peercred.diff to test, but of
course that won't work for us (even if we would have working socket
credentials with cmsg)
<pinotree> (that patch is not tested yet)
<pinotree> at least, we're aware there's another place in need for working
socket credentials now
<pinotree> youpi: (the patch above has been confirmed to work, with
<youpi> 0 ?!
See also pflocal reauth and sendmsg scm creds.