[PATCH] Mask the tmp.mount service to avoid tmpfs

Will Woods wwoods at redhat.com
Thu Aug 23 20:28:05 UTC 2012


On Thu, 2012-08-16 at 09:33 -0700, Jesse Keating wrote:
> On 08/16/2012 08:30 AM, Will Woods wrote:
> > On Wed, 2012-08-15 at 13:59 -0700, Jesse Keating wrote:
> >> Anaconda runtime is already in memory, no need to use tmpfs here.  In
> >> fact use of tmpfs here will overwrite any updates content that was put
> >> in place by dracut.
> >
> > Not a great idea - we want to avoid using the live overlay unless
> > necessary:
> >
> > 1) filling the overlay leads to Bad Explodey Problems,
> > 2) tmpfs is can be swapped to disk, but the overlay can't,
> 
> That's assuming you have disk available right?  That requires a system 
> where swap is going to be used, but that doesn't come into play until 
> after the partitioning screen right, pretty late in the game?

Yes. But I'm not sure what else goes into /tmp (yum metadata? downloaded
packages?), and I think our peak memory use was historically sometime
during package install - which is when we'd have swap available.. 

> > 3) we shouldn't depend on root+usr being writable [NA here though]
> >
> > Also we should avoid diverging from Standard System Setup when we can.
> >
> > When switch_root happens, /run gets moved to the new system.
> > Maybe we should be writing our updates etc. under /run instead?
> >
> 
> Moving our updates infrastructure to /run/ could work, but would require 
> more changes in anaconda than I think we're comfortable with at this 
> moment given F18.  

We don't need to necessarily move everything to /run - there are ways to
get the contents of initramfs /tmp into anaconda's /tmp without
modifying all of ananconda's code to point to '/run/updates' etc.

> I'll look into what it would take to do that, but in 
> the mean time we have no working updates infrastructure.  Do we want to 
> wait to see how much code it'll take, and hope it works, or do we want 
> to use this fix for now in the short term?

I'm fine with this fix for the short term (i.e. F18 Alpha) but surely we
can work out a better solution post-Alpha.

-w



More information about the anaconda-patches mailing list