Re: [Qemu-devel] [PULL] Memory core space reduction

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Feb 28, 2012 at 01:20:47PM -0600, Anthony Liguori wrote:
> On 02/28/2012 01:17 PM, Avi Kivity wrote:
> >On 02/28/2012 09:14 PM, Anthony Liguori wrote:
> >>On 02/28/2012 12:15 PM, Avi Kivity wrote:
> >>>On 02/28/2012 08:13 PM, Anthony Liguori wrote:
> >>>>
> >>>>FWIW, I'm processing this PULL request right now and I'm seeing a SEGV
> >>>>too.  The backtrace is a malloc failure in QOM.
> >>>>
> >>>
> >>>How do we reproduce this?
> >>
> >>I don't trust this bisect completely, but here are the results:
> >>
> >>
> >>  5f0e841a5c8c0bc0663e5582432eb788a3e0f9e3 is the first bad commit
> >>commit 5f0e841a5c8c0bc0663e5582432eb788a3e0f9e3
> >>Author: Avi Kivity<avi@redhat.com>
> >>Date:   Sun Jan 8 19:46:17 2012 +0200
> >>
> >>     ioport: change portio_list not to use memory_region_set_offset()
> >>
> >>     memory_region_set_offset() will be going away soon, so don't use it.
> >>     Use an alias instead.
> >>
> >>     Signed-off-by: Avi Kivity<avi@redhat.com>
> >>     Reviewed-by: Richard Henderson<rth@twiddle.net>
> >>
> >>:100644 100644 36fa3a477ebde72de4745bf4e13ad5146f4686fd
> >>505b252491d1d4e618a5059d75f3cb560a24c61f M    ioport.c
> >>:100644 100644 ae3e9da0b5487e68a16f28c459889496160e8e16
> >>ab29c89fb3ac6bbe72b2b622172cb9ef7c462e62 M    ioport.h
> >>bisect run success
> >
> >That's the very first commit.  You'd get this result if either this was
> >the bad commit, of if the input to 'git bisect good' was also bad.  Can
> >you double-check this?
> 
> Looks like it was a bad bisect :-(  I thought I had a 100%
> reproducible test case but it turned out not to be.
> 
> Regards,
> 
> Anthony Liguroi
> 
> 


OK, here it reproduces without issues, so I did a bisect.
This is the 1st bad commit:

memory: allow phys_map tree paths to terminate early

When storing large contiguous ranges in phys_map, all values tend to
be the same pointers to a single MemoryRegionSection.  Collapse them
by marking nodes with level > 0 as leaves.  This reduces tree memory
usage dramatically.

Signed-off-by: Avi Kivity <avi@redhat.com>


What I did, to allow bisect, is rebase Avi's patches on top
of my bridge implementation, then run qemu with a bridge.
bridge without Avi's patches at least starts booting, with
Avi's patches crashes before guest start.

If you want to play with that, take it from branch bisectme
on my qemu tree on github.

-- 
MST