Hi All,
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Attached is a patch to prevent this from happening.
Thanks to Mads Kiilerich <killerix a gmail.com> for pointing this out in https://hosted.fedoraproject.org/projects/revisor/ticket/257
Kind regards,
Jeroen van Meeuwen -kanarip
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote:
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
Jeremy
Jeremy Katz wrote:
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote:
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
+1
Kind regards,
Jeroen van Meeuwen -kanarip
On Mon, 2007-08-06 at 23:09 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote:
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
+1
Want to send a revised patch? :-)
Jeremy
Jeremy Katz wrote:
On Mon, 2007-08-06 at 23:09 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote:
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
+1
Want to send a revised patch? :-)
Well, as it is now it does apply at the very last moment, exactly the way the installer works, right?
Kind regards,
Jeroen van Meeuwen -kanarip
On Tue, 2007-08-07 at 00:20 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Mon, 2007-08-06 at 23:09 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote:
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
+1
Want to send a revised patch? :-)
Well, as it is now it does apply at the very last moment, exactly the way the installer works, right?
There are a few steps (relabeling, prelinking, making the initrd) that are happening after runPost() of your patch. If the %post is going to be the absolute last thing, it should probably be after all of these
Jeremy
Jeremy Katz wrote:
On Tue, 2007-08-07 at 00:20 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Mon, 2007-08-06 at 23:09 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote:
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
+1
Want to send a revised patch? :-)
Well, as it is now it does apply at the very last moment, exactly the way the installer works, right?
There are a few steps (relabeling, prelinking, making the initrd) that are happening after runPost() of your patch. If the %post is going to be the absolute last thing, it should probably be after all of these
This might be the wrong list to be asking this, but out of curiosity-
Why can't relabeling be done if the host is running with selinux disabled? (selinux=0)
It's just writing some metadata to the fs right? Seems like it should be possible.
And on a seperate note, I agree that prelinking really ought to be done in %post. That's something that a kickstart user might be interested in doing in the non-livecd case as well, for the same reasons, right?
-dmc
On Tue, 2007-08-07 at 16:42 -0500, Douglas McClendon wrote:
This might be the wrong list to be asking this, but out of curiosity-
Why can't relabeling be done if the host is running with selinux disabled? (selinux=0) It's just writing some metadata to the fs right? Seems like it should be possible.
Because the kernel developers have deemed it unsafe to write out any security xattrs which aren't understood by the kernel. The fact that they then get mapped to unlabeled_t doesn't seem to make much difference. Frankly, I think they're wrong, but lengthy attempts to convince them in the past have been unsuccessful. Maybe it's worth tilting at that windmill again. Dunno.
And on a seperate note, I agree that prelinking really ought to be done in %post. That's something that a kickstart user might be interested in doing in the non-livecd case as well, for the same reasons, right?
Yep.
Jeremy
Jeremy Katz wrote:
On Tue, 2007-08-07 at 16:42 -0500, Douglas McClendon wrote:
This might be the wrong list to be asking this, but out of curiosity-
Why can't relabeling be done if the host is running with selinux disabled? (selinux=0) It's just writing some metadata to the fs right? Seems like it should be possible.
Because the kernel developers have deemed it unsafe to write out any security xattrs which aren't understood by the kernel. The fact that they then get mapped to unlabeled_t doesn't seem to make much difference. Frankly, I think they're wrong, but lengthy attempts to convince them in the past have been unsuccessful. Maybe it's worth tilting at that windmill again. Dunno.
Thats the sort of info I was looking for. I'm probably part of a very small group of people who would notice, and even I don't care that much. I'll post a little script one of these days that gets around it :)
-dmc
Jeremy Katz wrote:
On Tue, 2007-08-07 at 00:20 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Mon, 2007-08-06 at 23:09 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote:
livecd-creator's InstallationTarget.install() seems to run self.configureSystem(), then run self.configureNetwork(), possible overriding some configuration files generated with %post.
Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
+1
Want to send a revised patch? :-)
Well, as it is now it does apply at the very last moment, exactly the way the installer works, right?
There are a few steps (relabeling, prelinking, making the initrd) that are happening after runPost() of your patch. If the %post is going to be the absolute last thing, it should probably be after all of these
It'd be plausible to have relabeling, prelinking and making the initrd occur after %post, but %post should apply to whatever else we do with kickstart information, right? I think where it is now, it fits perfectly.
Kind regards,
Jeroen van Meeuwen -kanarip
On Wed, 2007-08-08 at 12:53 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Tue, 2007-08-07 at 00:20 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Mon, 2007-08-06 at 23:09 +0200, Jeroen van Meeuwen wrote:
Jeremy Katz wrote:
On Wed, 2007-08-01 at 14:16 +0200, Jeroen van Meeuwen wrote: > livecd-creator's InstallationTarget.install() seems to run > self.configureSystem(), then run self.configureNetwork(), possible > overriding some configuration files generated with %post. Hmm, the question is when is the best point at which to do this. After configuring the network or really as last as possible? The latter seems to fit better with when its run on an install
+1
Want to send a revised patch? :-)
Well, as it is now it does apply at the very last moment, exactly the way the installer works, right?
There are a few steps (relabeling, prelinking, making the initrd) that are happening after runPost() of your patch. If the %post is going to be the absolute last thing, it should probably be after all of these
It'd be plausible to have relabeling, prelinking and making the initrd occur after %post, but %post should apply to whatever else we do with kickstart information, right? I think where it is now, it fits perfectly.
Yeah, I can see it being plausible basically anywhere. That's why the position of %post has moved around from "right after packages installed" to "at the very end" to "at the very end, minus the CD eject" in anaconda over time. Hence, I lean towards at the very end just for consistency with current anaconda behavior
Jeremy
livecd@lists.fedoraproject.org
-
Douglas McClendon -
Jeremy Katz -
Jeroen van Meeuwen