From: Marc Zyngier Date: Tue, 19 Oct 2021 08:39:51 +0100 Subject: [Intel-wired-lan] [PATCH v3 0/3] PCI/MSI: Clarify the IRQ sysfs ABI for PCI devices In-Reply-To: <20211018215047.GA2265015@bhelgaas> References: <20210825102636.52757-1-21cnbao@gmail.com> <20211018215047.GA2265015@bhelgaas> Message-ID: <87wnm9o43c.wl-maz@kernel.org> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: intel-wired-lan@osuosl.org List-ID: On Mon, 18 Oct 2021 22:50:47 +0100, Bjorn Helgaas wrote: > > On Wed, Aug 25, 2021 at 06:26:33PM +0800, Barry Song wrote: > > From: Barry Song > > > > > > /sys/bus/pci/devices/.../irq has been there for many years but it has never > > been documented. This patchset is trying to clarify it. > > > > -v3: > > - Don't attempt to modify the current behaviour of IRQ ABI for MSI-X > > - Make MSI IRQ ABI more explicit(return 1st IRQ of the IRQ vector) > > - Add Marc's patch of removing default_irq from the previous comment to > > the series. > > Note patch 3/3 indirectly changed the code of pci_restore_msi_state(), > > drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c is the only driver > > calling this API. I would appreciate testing done from this driver's > > maintainers. > > > > -v2: > > - split into two patches according to Bjorn's comments; > > - Add Greg's Acked-by, thanks for reviewing! > > https://lore.kernel.org/lkml/20210820223744.8439-1-21cnbao at gmail.com/ > > > > -v1: > > https://lore.kernel.org/lkml/20210813122650.25764-1-21cnbao at gmail.com/#t > > > > Barry Song (2): > > Documentation: ABI: sysfs-bus-pci: Add description for IRQ entry > > PCI/sysfs: Don't depend on pci_dev.irq for IRQ entry > > I applied the first two (above) to pci/msi for v5.16, thanks! > > As far as I can tell from the discussion so far, they should be safe > and should preserve all existing behavior. The second patch should > remove the sysfs dependency on the PCI core to swap the INTx and first > MSI IRQ values in dev->irq. > > Marc's patch below is certainly desirable but my understanding is that > it requires some driver updates first. Indeed. It means that there will be an uphill battle to remove the INTx/MSI confusion in a lot of drivers (Barry says about 700 of them, and I can't say I'm surprised). We can probably tackle that as a background task for the next few releases (I need to relearn Coccinelle for the 27th time...). Thanks, M. -- Without deviation from the norm, progress is not possible.