Re: [PATCH 3/4] pci: move devres code from pci.c to devres.c

[Philipp Stanner <pstanner@redhat.com>]

On Tue, 2023-11-21 at 11:17 +0100, Arnd Bergmann wrote:
> On Mon, Nov 20, 2023, at 22:59, Philipp Stanner wrote:
> > The file pci.c is very large and contains a number of devres-
> > functions.
> > These functions should now reside in devres.c
> > 
> > There are a few callers left in pci.c that do devres operations.
> > These
> > should be ported in the future. Corresponding TODOs are added by
> > this
> > commit.
> > 
> > The reason they are not moved right now in this commit is that
> > pci's
> > devres currently implements a sort of "hybrid-mode":
> > pci_request_region(), for instance, does not have a corresponding
> > pcim_
> > equivalent, yet. Instead, the function can be made managed by
> > previously
> > calling pcim_enable_device() (instead of pci_enable_device()). This
> > makes it unreasonable to move pci_request_region() to devres.c
> > Moving the functions would require changes to pci's API and is,
> > therefore, left for future work.
> > 
> > Move as much devres-specific code from pci.c to devres.c as
> > possible.
> > 
> > Suggested-by: Danilo Krummrich <dakr@redhat.com>
> > Signed-off-by: Philipp Stanner <pstanner@redhat.com>
> > ---
> >  drivers/pci/devres.c | 243
> > +++++++++++++++++++++++++++++++++++++++++
> >  drivers/pci/pci.c    | 249 ---------------------------------------
> > ----
> >  drivers/pci/pci.h    |  24 +++++
> >  3 files changed, 267 insertions(+), 249 deletions(-)
> 
> I had just commented in the other mail that you'd have to move
> these functions to devres.c for the file to make sense, but that
> I think the existing state is better.
> 
> Just to clarify again here: this patch does not seem to improve
> anything to me,

Have you read the cover letter? It elaborates on that.

The idea behind centralizing devres-pci-code in a separate file is that
the current implementation is strangely torn.
My mid-term goal would be to fix that, but that's beyond the scope of
this series.

PCI has some separate devres functions, prefixed with pcim_ – and then
there are some other functions that use a crazy hybrid mode:

drivers/pci/pci.c:

void pci_release_region(struct pci_dev *pdev, int bar)
{
	/* ..... SNIP ......... */

 dr = find_pci_dr(pdev);
 if (dr)
 dr->region_mask &= ~(1 << bar);
}

Some functions without pcim_ prefix switch to managed mode if
pcim_enable_device() instead of pci_enable_device() was called.

So some functions are sometimes managed and others never are, which is
totally inconsistent.

This is bug-provoking because programmers won't know without looking
very closely which functions become managed and which don't.

That should be fixed. And a first step towards that goal is to cleanly
split them. Everything managed should reside, on the long term, in
drivers/pci/devres.c

>  I'd much prefer leaving it the way it is, and
> moving the pcim_iomap family to corresponding drivers/pci/iomap.c. 

We could branch that change out of the patch series and handle the
topics independently

P.

> 
>      Arnd
>