On Fri, 2024-11-15 at 01:46 +0100, Thomas Gleixner wrote:

On Thu, Nov 14 2024 at 10:05, Philipp Stanner wrote:

quoted

On Wed, 2024-11-13 at 17:22 +0100, Thomas Gleixner wrote:

quoted

On Wed, Nov 13 2024 at 13:41, Philipp Stanner wrote:

quoted

pci_intx() is a hybrid function which can sometimes be managed
through
devres. This hybrid nature is undesirable.

Since all users of pci_intx() have by now been ported either to
always-managed pcim_intx() or never-managed
pci_intx_unmanaged(),
the
devres functionality can be removed from pci_intx().

Consequently, pci_intx_unmanaged() is now redundant, because
pci_intx()
itself is now unmanaged.

Remove the devres functionality from pci_intx(). Have all users
of
pci_intx_unmanaged() call pci_intx(). Remove
pci_intx_unmanaged().

Signed-off-by: Philipp Stanner <redacted>
---
 drivers/misc/cardreader/rtsx_pcr.c            |  2 +-
 drivers/misc/tifm_7xx1.c                      |  6 +--
 .../net/ethernet/broadcom/bnx2x/bnx2x_main.c  |  2 +-
 drivers/net/ethernet/brocade/bna/bnad.c       |  2 +-
 drivers/ntb/hw/amd/ntb_hw_amd.c               |  4 +-
 drivers/ntb/hw/intel/ntb_hw_gen1.c            |  2 +-
 drivers/pci/devres.c                          |  4 +-
 drivers/pci/msi/api.c                         |  2 +-
 drivers/pci/msi/msi.c                         |  2 +-
 drivers/pci/pci.c                             | 43 +----------
----
----
 drivers/vfio/pci/vfio_pci_core.c              |  2 +-
 drivers/vfio/pci/vfio_pci_intrs.c             | 10 ++---
 drivers/xen/xen-pciback/conf_space_header.c   |  2 +-
 include/linux/pci.h                           |  1 -
 14 files changed, 22 insertions(+), 62 deletions(-)

Now I'm utterly confused. This undoes the pci_intx_unmanaged()
churn
which you carefully split into several patches first.

Have you read the email I have linked?

There is also the cover-letter (does anyone in the community ever
read
those?) which explicitly states:

"Patch "Remove devres from pci_intx()" obviously reverts the
previous
patches that made drivers use pci_intx_unmanaged(). But this way
it's
easier to review and approve. It also makes sure that each checked
out
commit should provide correct behavior, not just the entire series
as a
whole."

I read it and I assume your intention was to force an eye on every
use
case of pci_intx() and not just on those which need to be converted
to
pcim_intx().

I'm not convinced that this is needed, but fair enough.

Whether pcim_enable_device() is really not used could have been
overlooked, or the driver could move to "managed mode" in parallel for
v6.13 for example. Then a bug would be silently introduced into those
drivers.

Besides, me touching pci_intx() unfortunately caused a few explosions
in the past already, in

fc8c818e756991f5f50b8dfab07f970a18da2556 and
00f89ae4e759a7eef07e4188e1534af7dd2c7e9c

So this time I prefer to be rather safe than sorry.


BTW, if you can review the MSI patch and check whether removing devres
from there really is fine, that would be helpful.


Regards,
P.