On 30/01/2018 07:29, Juan Quintela wrote:
> Peter Maydell <peter.mayd...@linaro.org> wrote:
>> On 29 January 2018 at 12:09, Juan Quintela <quint...@redhat.com> wrote:
>>> Hi
>>>
>>> - migrate_fd_connect cleanup (dave)
>>> - use MAX_VM_PACKAGED_SIZE
>>> have to change (1<<32) for UINT32_MAX to get no warnings on win32
>>> - Use right s->threashord_size(wei)
>>> - patches reviewed from info/migrate, migration tests and multifd (me)
>>>
>>> Later, Juan.
>>>
>>> The following changes since commit 6233b4a8c2a32ef6955a921246fa08705bbb3676:
>>>
>>> Merge remote-tracking branch
>>> 'remotes/ericb/tags/pull-nbd-2018-01-26' into staging (2018-01-26
>>> 17:29:14 +0000)
>>>
>>> are available in the Git repository at:
>>>
>>> git://github.com/juanquintela/qemu.git tags/migration/20180129
>>>
>>> for you to fetch changes up to 4b6bb0a77761b8a906a540e927c7a9dde99bc10f:
>>>
>>> tests: Adjust sleeps for migration test (2018-01-29 12:10:22 +0100)
>>>
>>> ----------------------------------------------------------------
>>> migration/next for 20180129
>>
>> Hi. On aarch32 host I get this assert in the tests:
>>
>> TEST: tests/migration-test... (pid=605)
>> /ppc64/migration/deprecated: OK
>> /ppc64/migration/precopy/unix: OK
>> /ppc64/migration/postcopy/unix: OK
>> /ppc64/migration/xbzrle/unix:
>> Unexpected 32 on dest_serial serial
>> **
>> ERROR:/home/peter.maydell/qemu/tests/migration-test.c:218:wait_for_serial:
>> code should not be reached
>
> 1st time I see a problem on the xbzrle test. I can resent the pull
> without that patch, but that is a *test*, the problem is that we have a
> xbzrle problem on aarch32 that we didn't know about yet.
>
> So I will just try to do the PULL request anyways. I will try to get
> access to an arm32 system tomorrow and see what is going on.
>
> Thanks for checking it.
I have tested this on ppc32 (G4@900MHz) and I found it never converges.
If I increase the downtime limit, it works:
--- a/tests/migration-test.c
+++ b/tests/migration-test.c
@@ -674,7 +674,7 @@ static void test_xbzrle(const char *uri)
wait_for_migration_pass(from);
/* 300ms should converge */
- migrate_set_parameter(from, "downtime-limit", "300");
+ migrate_set_parameter(from, "downtime-limit", "2000");
The test is also disabled if USERFAULTFD is disabled in the kernel,
whereas it doesn't need it. You should update the "if
(!ufd_version_check())" part.
Thanks,
Laurent
