Re: file sync on close(2)
Re: file sync on close(2)
- Subject: Re: file sync on close(2)
- From: Jim Magee <email@hidden>
- Date: Thu, 24 Jun 2004 19:30:55 -0400
Sorry for the long delay. I've been a bit busy... ;-)
On Jun 14, 2004, at 2:31 PM, Justin Walker wrote:
>
On Jun 14, 2004, at 11:01, Jay A. Kreibich wrote:
>
> The "single user mode" you get on boot, and the "single user mode"
>
> you get via a SIGTERM to init(1) are two different things.
>
>
Not really.
Actually, and unfortunately, Jay is correct. Going back down to
single-user-mode is something that Darwin does not do very gracefully
[or correctly] at the moment. And, as Jay stated, that "second"
single-user-mode is quite a bit less functional than the first.
Mainly, this is because mach_init disappears. And with it goes lots of
mechanism that is expected to be available by the processes one expects
to work with in single-user mode. Init also thinks "the old mach_init
is still the current one." So, as it launches processes on the way
back up - they all will try to talk to the dead mach_init - quite
unsuccessfully as you would imagine.
But there are also lots of other things that weren't as graceful as
they should be. One example is that the dynamic_pager process, which
manages the swap file creation/deletion for the kernel, goes away. Any
new one that comes up after a trip into single-user mode and back won't
know anything about the swap files that were created by the first one.
The current protocol between the kernel and this daemon doesn't help in
that regard - so confusion reigns there as well. There are a handful of
other similar issues. The result is that going back to
single-user-mode without the reboot is not a very safe or pleasant
thing to do at the moment.
Luckily, the reboot really doesn't add much time to the process - so
most people don't care about all this. But if you are trying to "add
cleanup" at that stage before the reboot, it becomes all too apparent.
Mac OS X doesn't support the notion of single-user-mode on the way
down, so this isn't a problem there. But people using Darwin in ways
Mac OS X doesn't have noticed these and filed the appropriate bug
reports already.
And before the questions come, yes we are addressing these issues as
best we can. But until we nail them all satisfactorily, we don't want
to "fix" the mach_init thing. It will make too many things "appear" to
work - when they system is none the happier overall.
--Jim
[demime 0.98b removed an attachment of type application/pkcs7-signature which had a name of smime.p7s]
_______________________________________________
darwin-kernel mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/darwin-kernel
Do not post admin requests to the list. They will be ignored.