dieter wrote:
On Fri, 28 Jun 2013 01:35:10 -0700 Linda Walsh wrote:
Cristian RodrC-guez wrote:
No, it is not just kernel modules.. in current incarnations with systemd/dracut the initrd is also used for clean shutdown, not just for system startup.
dracut wouldn't be needed for an initrd that has only kernel modules would it? What's on initrd that systemd would need that isn't on the hard disk?
Er... wait a sec, you mean initrd isn't deallocated after boot? ?
Do you mean to say that the large initrd's that people are complaining about stay around in memory AFTER boot just so systemd can use it to shut down? I fully agree that small initrds are desirable and I would also prefer if it does not contain parts which are not really needed. But on the other hand for "modern PCs" I would think it is not really an issue. Even if the initrd is kept during the uptime, in case the memory would be needed it would be paged out, I assume without noticable impact on the performance.
If the systemd project recommends getting rid of initd for efficiency & speed, I'm sure it wouldn't be too difficult for us to do so. In my understanding there are significantly different use cases, especially small/mobile/embedded devices vs. "PCs". For the first it is known quite precisely in advance which HW can be included and therefore it is "easier" to build a monolythic kernel - and my interpretation of the systemd configuration without initrd aims at these.
For the latter there is a huge range of possible configurations, and therefore it is "reasonable" to use the initrd.
How is a PC not small, mobile or embedded compared to a mainframe, or, at the other end, compared to a cellphone? It's all relative. However, to any individual PC owner, it can be known (for those who wish to know) precisely, in advance what HW is our computers. A distro manufacturer may not know every device out there, but that's why they created scripts to create initrd disks that are tailored to users' individual computers, right? How would configuring a kernel via a similar script be any different than the current situation where it takes 10-20 minutes to generate all of the initrd scripts on my machine when SW is updated, vs. about 3-4 minutes to generate a kernel? Why not go for the lighter-weight solution and only generate the parts that need to be different for users -- their boot kernel. The rest can be read in as soon as the kernel is up. The less we put on initrd's the better. Surprise initrd-generations are a real pain -- they take 10-20 minutes vs. about 3.6 minutes to build a new kernel. So how do users benefit by having to wait so long when they upgrade SW on their system? Supposedly the benefit of booting from a ram disk was users didn't have to wait for a kernel to be recompiled... but 3.6 minutes vs. 10-20? Um... -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org