On 1/10/22 15:33, Daniel Henrique Barboza wrote:
> Hi,
>
> This v3 contains new versions of pnv-phb4 exclusive patches from
> version 2. Patches 1-10 are already accepted.
>
> I changed how patch 9 (v2 patch 17) works by doing everything possible
> in extra patches/cleanups beforehand, and then using patch 9 to flip the
> switch in a single step. This means that handling the default initialization
> of pnv-phb4s is done at the same time we enable user creatable pnv-phb4s.
>
> There's also a change in how XSCOM initializion is being handled. We're not
> using a flag to do a partial XSCOM initialization during phb4_realize() anymore.
> Intead, we moved XSCOM initialization code, as less intrusive as we could, to
> phb4_realize().
>
> This time I also took the precaution of testing the default case
> (i.e. running without -nodefaults) in every patch. v2 was breaking
> this default run between some patches.
>
> changes from v2:
> - former patch 16: removed
> - patch 10 (v2 18): unchanged
> - patches 4,5,7,8: new
> - patch 9 (former 17):
> * added g_assert() if stack == NULL
> * added a comment explaining why we shouldn't assert on user error
> with wrong chip-id/index values
> - minor changes across the patches due to the design changes
> - v2 link: https://lists.gnu.org/archive/html/qemu-devel/2022-01/msg00671.html
>
> Daniel Henrique Barboza (10):
> pnv_phb4.c: introduce pnv_phb4_set_stack_phb_props()
> pnv_phb4_pec.c: move pnv_pec_phb_offset() to pnv_phb4.c
> pnv_phb4_pec: use pnv_phb4_pec_get_phb_id() in pnv_pec_dt_xscom()
> pnv_phb4_pec.c: remove stack 'phb-id' alias
> pnv_phb4_pec.c: move phb4 properties setup to pec_realize()
> ppc/pnv: turn 'phb' into a pointer in struct PnvPhb4PecStack
> ppc/pnv: move PHB4 related XSCOM init to phb4_realize()
> pnv_phb4.c: check stack->phb not NULL in phb4_update_regions()
> ppc/pnv: Introduce user creatable pnv-phb4 devices
> pnv_phb4.c: change TYPE_PNV_PHB4_ROOT_BUS name
I have taken Patches 10, 2, 3 in this order. There is still some PHB4
code outside the PHB4 realize routine and this is making the code too
complex. We are getting closer !
Could you use the 'ppc/pnv:' prefix ?
Thanks,
C.