Re: [Qemu-ppc] [PULL 00/67] ppc-for-2.9 queue 20170112

[David Gibson]

On Fri, Jan 13, 2017 at 10:54:33AM +0000, Peter Maydell wrote:
> On 12 January 2017 at 02:02, David Gibson <address@hidden> wrote:
> > The following changes since commit b44486dfb9447c88e4b216e730adcc780190852c:
> >
> >   Merge remote-tracking branch 'remotes/kraxel/tags/pull-ui-20170110-1' 
> > into staging (2017-01-10 14:52:34 +0000)
> >
> > are available in the git repository at:
> >
> >   git://github.com/dgibson/qemu.git tags/ppc-for-2.9-20170112
> >
> > for you to fetch changes up to 229b5b6b56db9c3e6b8f44b4b55a62fa32979f47:
> >
> >   ppc: Fix a warning in bcdcfz code and improve BCD_DIG_BYTE macro 
> > (2017-01-12 10:21:49 +1100)
> >
> > ----------------------------------------------------------------
> > ppc patch queue 2017-01-12
> >
> > This is the first ppc pull request for qemu-2.9.  It's been a while
> > coming, partly due to dealing with some problems running my usual set
> > of tests.  So, there's quite a lot in here.
> >
> >     * More POWER9 instruction implementations for TCG
> >     * The simpler parts of my CPU compatibility mode cleanup
> >         * This changes behaviour to prefer compatibility modes over
> >           "raW" mode for new machine type versions
> >     * New "40p" machine type which is essentially a modernized and
> >       cleaned up "prep".  The intention is that it will replace "prep"
> >       once it has some more testing and polish.
> >     * Add pseries-2.9 machine type
> >     * Implement H_SIGNAL_SYS_RESET hypercall
> >     * Consolidate the two alternate CPU init paths in pseries by
> >       making it always go through CPU core objects to initialize CPU
> >     * A handful of bugfixes and cleanups
> >
> > There are also some changes not strictly related to ppc code, but for
> > its benefit:
> >
> >     * Limit the pxi-expander-bridge (PXB) device to x86 guests only
> >       (it's essentially a hack to work around historical x86
> >       limitations)
> >     * Revise a number of qtests and enable them for ppc
> 
> Hi -- this fails "make check" on OSX host:

Bother.  I saw the failure on Travis, but I was misled by all the SASL
warnings, and assumed it was an upstream / library problem, not
something with my patches.

>   GTESTER check-qtest-ppc64
> qemu-system-ppc64: -object
> memory-backend-file,id=mb1,size=1M,share,mem-path=/dev/shm/qtest-68237-4247260626:
> invalid object type: memory-backend-file
> Broken pipe
> GTester: last random seed: R02Sc13bf53b8b3055a3c30552f67d7c263b
> **
> ERROR:/Users/pm215/src/qemu-for-merges/tests/libqos/pci.c:412:void
> qpci_plug_device_test(const char *, const char *, uint8_t, const char
> *): assertion failed: (!qdict_haskey(response, "error"))
> GTester: last random seed: R02Sf3594a21c711ec941d5f4c89724ce12d
> qemu-system-ppc64: -device ivshmem-plain,memdev=mb1: 'ivshmem-plain'
> is not a valid device model name
> Broken pipe
> 
> It looks like you've enabled the ivshmem test unconditionally,
> but it should only be turned on if CONFIG_EVENTFD is true, see
> how we do this for x86:
> 
> check-qtest-pci-$(CONFIG_EVENTFD) += tests/ivshmem-test$(EXESUF)
> gcov-files-pci-y += hw/misc/ivshmem.c

Ah, right, thanks.

> Q: can we just enable these tests for PCI with
> check-qtest-ppc64-y += $(check-qtest-pci-y)
> gcov-files-ppc64-y += $(gcov-files-pci-y)
> 
> the way we do for i386/x86_64 ? Or is that doing too much
> unwanted extra testing?

As Laurent says, that doesn't quite work, but we can still put the
test on CONFIG_EVENTFD in.  I've put a draft fix into my tree, but I
probably won't get a chance to properly test and resend until after
I'm back from LCA and holidays, around the end of Jan.

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature