Re: 2.4.5-rc1 shm paging blocker

[Andrea Arcangeli <andrea@suse.de>]

it seems the testing I did on 2.6.5-rc1 was incorrect (after testing
another 2 combinations of kernels were the problem exists I may have not
waited long enough while testing 2.6.5-rc1 for the first time and I
mistaken a sluggishness for a live lock and I clicked reboot too early,
that's the only reasonable explanation), it's working correctly now
(like it works fine with 2.6.5-rc1 - backout + objrmap + anon_vma so
it's impossible that it was my changes triggering it). If it happens
again I'll let you know. I still have to find what cause it in another
combinations of vm patches. Andrew, now that you've the testcase could
you test the -mm tree, and see if happens there, some -mm patch is one
of the diffs between the working tree and the non-working one.

However now I'll repeat all tests, if they works all flawlessy today I
will be very annoyed, since the thing was definitely not swapping at all
yesterday, no matter how long I awaited.

Now I start to wonder if kbuild may have screwed my kernel, effectively
it was a development tree, I don't run make distclean anymore in between
the kernel compiles (I almost don't feel the need of ccache anymoe ;),
but maybe I'm wrong trusting the buildsystem can be smart enough.