On Thu, Dec 10, 2015 at 01:07:12PM -0600, Scott Wood wrote:
> Originally the mpc85xx-pci-edac driver bound directly to the PCI
> controller node.
>
> Commit 905e75c46dba5f30 ("powerpc/fsl-pci: Unify pci/pcie
> initialization code") turned the PCI controller code into a platform
> device. Since we can't have two drivers binding to the same device,
> the edac code was changed to be called into as a library-style
> submodule. However, this doesn't work if the edac driver is built as a
> module.
>
> Commit 8d8fcba6d1eab ("EDAC: Rip out the edac_subsys reference
> counting") exposed another problem with this approach --
> mpc85xx_pci_err_probe() was being called in the same early boot phase
> that the PCI controller is initialized, rather than in the
> device_initcall phase that the EDAC layer expects. This caused a crash
> on boot.
>
> To fix this, the PCI controller code now creates a child platform
> device specifically for EDAC, which the mpc85xx-pci-edac driver binds
> to.
>
> Signed-off-by: Scott Wood <scottw...@freescale.com>
> Cc: Jia Hongtao <b38...@freescale.com>
> Cc: Borislav Petkov <b...@suse.de>
> Cc: Johannes Thumshirn <jthumsh...@suse.de>
> Cc: Michael Ellerman <m...@ellerman.id.au>
> ---
> v3: Fix build with CONFIG_PCI disabled
>
> arch/powerpc/sysdev/fsl_pci.c | 28 +++++++++++++++++++++++++++-
> arch/powerpc/sysdev/fsl_pci.h | 9 ---------
> drivers/edac/mpc85xx_edac.c | 38 +++++++++++++++++++++++++++++++++-----
> include/linux/fsl/edac.h | 8 ++++++++
> 4 files changed, 68 insertions(+), 15 deletions(-)
> create mode 100644 include/linux/fsl/edac.h
Applied and rebased the tree so that this fix comes before the
rip-out-reference counting patch so that there's no hole during
bisection.
Thanks guys.
--
Regards/Gruss,
Boris.
SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284
(AG Nürnberg)
--
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
