Re: [Qemu-devel] [PULL v1 0/2] Merge qcrypto-next 2016/07/20

[Daniel P. Berrange]

On Wed, Jul 20, 2016 at 08:31:09PM +0100, Peter Maydell wrote:
> On 20 July 2016 at 12:32, Daniel P. Berrange <address@hidden> wrote:
> > The following changes since commit a3b343772114c8c98986508f3352a631164f913c:
> >
> >   Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into 
> > staging (2016-07-19 15:08:05 +0100)
> >
> > are available in the git repository at:
> >
> >   git://github.com/berrange/qemu tags/pull-qcrypto-2016-07-20-1
> >
> > for you to fetch changes up to 718083de8051736bba02f10dc4a88bd14ff6b79b:
> >
> >   crypto: don't open-code qcrypto_hash_supports (2016-07-19 16:46:23 +0100)
> >
> > ----------------------------------------------------------------
> > Merge qcrypto-next 2016/07/20 v1
> >
> > ----------------------------------------------------------------
> > Daniel P. Berrange (2):
> >   crypto: use glib as fallback for hash algorithm
> >   crypto: don't open-code qcrypto_hash_supports
> >
> 
> Fails to build with minimum-glib-version:
> 
> /Users/pm215/src/qemu-for-merges/crypto/hash-glib.c:32:33: error: use
> of undeclared identifier 'G_CHECKSUM_SHA512'; did you mean
> 'G_CHECKSUM_SHA1'?
>     [QCRYPTO_HASH_ALG_SHA512] = G_CHECKSUM_SHA512,
>                                 ^~~~~~~~~~~~~~~~~
>                                 G_CHECKSUM_SHA1
> /sw/include/glib-2.0/glib/gchecksum.h:48:3: note: 'G_CHECKSUM_SHA1'
> declared here
>   G_CHECKSUM_SHA1,
> 
> SHA512 doesn't exist in our minimum-supported glib version,
> I think (which has just MD5, SHA1 and SHA256).

Yay, glib docs simply say since 2.16, never mentioning that
SHA512 was not added until 2.36. Filed a bug against glib
to fix this

  https://bugzilla.gnome.org/show_bug.cgi?id=769027

It is easy enough to workaround in QEMU, so will do that.

> I also get test failures on the ppc64 build:
> 
> TEST: tests/test-crypto-afsplit... (pid=36787)
>   /crypto/afsplit/sha256/5:
> Unexpected error in qcrypto_random_bytes() at
> /home/pm215/qemu/crypto/random-stub.c:29:
> No random byte source provided in this build
> FAIL
> GTester: last random seed: R02S49a560b2cb6ee4cfc024d35cc33004f6
> (pid=36789)
>   /crypto/afsplit/sha256/5000:
> Unexpected error in qcrypto_random_bytes() at
> /home/pm215/qemu/crypto/random-stub.c:29:
> No random byte source provided in this build
> FAIL
> GTester: last random seed: R02S7e066aa40979ffe0de63a18e2e6a4bfa
> (pid=36790)
>   /crypto/afsplit/sha256/big:
> Unexpected error in qcrypto_random_bytes() at
> /home/pm215/qemu/crypto/random-stub.c:29:
> No random byte source provided in this build
> FAIL
> GTester: last random seed: R02S91b927f7808472d3ae4672d177078877
> (pid=36791)
>   /crypto/afsplit/sha1/1000:
> Unexpected error in qcrypto_random_bytes() at
> /home/pm215/qemu/crypto/random-stub.c:29:
> No random byte source provided in this build
> FAIL
> GTester: last random seed: R02S0a50082cedeb926cabdf8df1028d4296
> (pid=36792)
> FAIL: tests/test-crypto-afsplit

Ah fun, so afsplit used to automatically skip it tests if
no checksum impl was provided. It never bothered to check
if random byte impl was provided, since the check condition
for that matched that for checksums. Now we're always providing
a checksum impl, the lack of checking for random byte impl rears
its head.  Rather than fixing the test, I'm jst going to add a
general fallback random byte source using /dev/[u]random.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|