RE: [PATCH V3 1/5] powerpc/fsl-pci: Unify pci/pcie initialization code

[Jia Hongtao-B38951 <B38951@freescale.com>]

> -----Original Message-----
> From: Kumar Gala [mailto:galak@kernel.crashing.org]
> Sent: Friday, July 27, 2012 8:47 PM
> To: Jia Hongtao-B38951
> Cc: linuxppc-dev@lists.ozlabs.org; Wood Scott-B07421; Li Yang-R58472
> Subject: Re: [PATCH V3 1/5] powerpc/fsl-pci: Unify pci/pcie
> initialization code
>=20
>=20
> On Jul 27, 2012, at 3:35 AM, Jia Hongtao-B38951 wrote:
>=20
> >
> >
> >> -----Original Message-----
> >> From: Kumar Gala [mailto:galak@kernel.crashing.org]
> >> Sent: Friday, July 27, 2012 2:15 AM
> >> To: Jia Hongtao-B38951
> >> Cc: linuxppc-dev@lists.ozlabs.org; Wood Scott-B07421; Li Yang-R58472
> >> Subject: Re: [PATCH V3 1/5] powerpc/fsl-pci: Unify pci/pcie
> >> initialization code
> >>
> >>
> >> On Jul 26, 2012, at 7:30 AM, Jia Hongtao wrote:
> >>
> >>> We unified the Freescale pci/pcie initialization by changing the
> >> fsl_pci
> >>> to a platform driver. In previous PCI code architecture the
> >> initialization
> >>> routine is called at board_setup_arch stage. Now the initialization
> is
> >> done
> >>> in probe function which is architectural better. Also It's convenient
> >> for
> >>> adding PM support for PCI controller in later patch.
> >>>
> >>> One issue introduced by this architecture is the timing of
> swiotlb_init.
> >>> During PCI initialization the need of swiotlb is determined and this
> >> should
> >>> be done before swiotlb_init. So a new function to determine swiotlb
> by
> >>> parsing pci ranges is made. This function is called at
> board_setup_arch
> >>> stage which is earlier than swiotlb_init.
> >>>
> >>> Signed-off-by: Jia Hongtao <B38951@freescale.com>
> >>> Signed-off-by: Li Yang <leoli@freescale.com>
> >>> ---
> >>> Changed for V3:
> >>> - Rebase the patch set on the latest tree
> >>> - merge PCI unify and swiotlb patch into one
> >>>
> >>> arch/powerpc/sysdev/fsl_pci.c |  155
> ++++++++++++++++++++++++++++++++--
> >> -------
> >>> arch/powerpc/sysdev/fsl_pci.h |    9 +--
> >>> 2 files changed, 125 insertions(+), 39 deletions(-)
> >>
> >> I'd like the SWIOTLB refactoring as a separate patch.  Additionally,
> the
> >> order of patches should be as follows:
> >>
> >> 1. refactor PCI node parsing code
> >> 2. add pci_determine_swiotlb (should rename to
> fsl_pci_determine_swiotlb)
> >> 3. Determine primary bus by looking for ISA node
> >> 4. convert all boards over to fsl_pci_init
> >> 5. convert fsl pci to platform driver (edac and other fixes should be
> >> merged in here)
> >> 6. PM support
> >>
> >> - k
> >
> > Should I convert all boards over to fsl_pci_init first and then convert
> them
> > over to platform driver again or just convert them direct to platform
> driver?
>=20
> Yes do the fsl_pci_init conversion first.  The reason is we should NOT
> break functionality from one patch to another.
>=20
> - k


Actually, the functionality is not broken, other boards just use the old
Way to init pci controller and it still works.

-Hongtao.