[Qemu-devel] [PATCH v2 0/4] rtc port to VMState (and fix rtc-td hack)

[Juan Quintela <quintela@redhat.com>]

v2:
- fix compilation for non i386 targets.
- remove rtc_mm_init() (not used since 2008).

v1:
- minor indentation cleanup
- integrate rtc-td-hack properly in normal rtc state
- port to VMState

>From the fix rtc-td-hack that didn't change from last sent.
v2:

Only call rtc_coalesced_timer_update() if rtc_td_hack is true.
Patch was failing to load its own images.

v1:

There hasn't still released a version with the rtc_td hack.
Do we want to release 0.11 with this hack or change to a more maintenable mode,
i.e. just upgrade the version to 3 and save always the new values.
We should also propably add rtd_hack to the state, and make sure that value is the
same in origing and destination of th emigration.
Another problem is what we do with kvm, they have already done releases
with the "mc146818rtc-td" savevm section.

Notice that rtc-td-hack should really, really be enabled by defaulet.  It is
needed to have windows guest running correctly under load.

What do you think?  What should we do?

Later, Juan.

PD. Once that I was looking here, this driver has several features that are not
    used anywhere else.
- rtc_mm_init() is defined and exported but not used
- rtc_init_sqw() is defined, but only called from rtc_init() with sqw_irq = NULL
- sqw_irq is only assigned in rtc_init_sqw() and always to NULL value.

Is there a reason why we have the functionality that is not used anywhere?

Juan Quintela (4):
  mc145818rtc: fix saving of rtc-td hack properly upgrading the version
    number
  mc146818rtc: port rtc to vmstate
  mc146818rtc: fix indentation
  mc146818rtc: remove rtc_mm_init()

 hw/mc146818rtc.c |  220 +++++++++---------------------------------------------
 hw/pc.h          |    2 -
 2 files changed, 37 insertions(+), 185 deletions(-)