Re: [PATCH 0/6] constify pci_error_handlers structures

Re: [PATCH 0/6] constify pci_error_handlers structures

[Christoph Hellwig]

On Sat, Aug 12, 2017 at 07:44:28AM +0200, Julia Lawall wrote:
> These pci_error_handlers structures are only stored in the err_handler
> field of a pci_driver structure, and this field is declared as const.  Thus
> the pci_error_handlers structures can be const too.
> 
> Done with the help of Coccinelle.

If you're doing a scripted conversion of the pci_error_handlers
structured I'd much rather see that structure killed off and folded
into the pci_driver one.

Re: [PATCH 0/6] constify pci_error_handlers structures

[Julia Lawall]

On Sat, 12 Aug 2017, Christoph Hellwig wrote:

> On Sat, Aug 12, 2017 at 07:44:28AM +0200, Julia Lawall wrote:
> > These pci_error_handlers structures are only stored in the err_handler
> > field of a pci_driver structure, and this field is declared as const.  Thus
> > the pci_error_handlers structures can be const too.
> >
> > Done with the help of Coccinelle.
>
> If you're doing a scripted conversion of the pci_error_handlers
> structured I'd much rather see that structure killed off and folded
> into the pci_driver one.

OK, sure.  So to be precise, you want the fields error_detected,
mmio_enabled, etc to be added as new fields to the pci_driver structure?

They both have a resume field, though.  What should the pci_error_handlers
resume function be renamed to?  Would resume_after_error be too much?

julia

Re: [PATCH 0/6] constify pci_error_handlers structures

[Christoph Hellwig]

On Sat, Aug 12, 2017 at 09:52:28AM +0200, Julia Lawall wrote:
> OK, sure.  So to be precise, you want the fields error_detected,
> mmio_enabled, etc to be added as new fields to the pci_driver structure?

Yes.

> They both have a resume field, though.  What should the pci_error_handlers
> resume function be renamed to?  Would resume_after_error be too much?

error_resume maybe?

FYI, I already killed it for the PCIe port drivers a while ago:

https://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci.git/commit/?h=pci/aer&id=c5dc3c69f17a7e77359f10c342d1816390bc8846

Re: [PATCH 0/6] constify pci_error_handlers structures

[Julia Lawall]

On Sat, 12 Aug 2017, Christoph Hellwig wrote:

> On Sat, Aug 12, 2017 at 09:52:28AM +0200, Julia Lawall wrote:
> > OK, sure.  So to be precise, you want the fields error_detected,
> > mmio_enabled, etc to be added as new fields to the pci_driver structure?
>
> Yes.
>
> > They both have a resume field, though.  What should the pci_error_handlers
> > resume function be renamed to?  Would resume_after_error be too much?
>
> error_resume maybe?

OK

>
> FYI, I already killed it for the PCIe port drivers a while ago:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci.git/commit/?h=pci/aer&id=c5dc3c69f17a7e77359f10c342d1816390bc8846

Thanks for the pointer.

julia

Re: [PATCH 0/6] constify pci_error_handlers structures

[Julia Lawall]

Another issue arises in the files drivers/infiniband/hw/hfi1/pcie.c and
drivers/infiniband/hw/qib/qib_pcie.c, where the pci_error_handlers
structure is defined in one file and used in another file.  The structure
definition references various functions that are static in the same file.
Should I try to move those functions to the file containing the pci_driver
structure?  Or leave the functions where they are and remove the static
annotation?

thanks,
julia

Re: [PATCH 0/6] constify pci_error_handlers structures

[Christoph Hellwig]

On Sat, Aug 12, 2017 at 01:51:21PM +0200, Julia Lawall wrote:
> Another issue arises in the files drivers/infiniband/hw/hfi1/pcie.c and
> drivers/infiniband/hw/qib/qib_pcie.c, where the pci_error_handlers
> structure is defined in one file and used in another file.  The structure
> definition references various functions that are static in the same file.
> Should I try to move those functions to the file containing the pci_driver
> structure?  Or leave the functions where they are and remove the static
> annotation?

I'd just leave them as-is and remove the static for now.