| Message ID | 20231122060406.14695-7-dlemoal@kernel.org |
|---|---|
| State | New |
| Headers | show |
| Series | Cleanup IRQ type definitions | expand |
On Wed, Nov 22, 2023 at 03:03:56PM +0900, Damien Le Moal wrote: > In the PCI Express Port Bus Driver, use the macro PCI_IRQ_INTX instead > of the now deprecated PCI_IRQ_LEGACY macro. I'd prefer to just script these cleanups in one big run to be honest..
On 11/22/23 15:17, Christoph Hellwig wrote: > On Wed, Nov 22, 2023 at 03:03:56PM +0900, Damien Le Moal wrote: >> In the PCI Express Port Bus Driver, use the macro PCI_IRQ_INTX instead >> of the now deprecated PCI_IRQ_LEGACY macro. > > I'd prefer to just script these cleanups in one big run to be honest.. I did not want to go as far as changing all drivers everywhere and limited the series to drivers/pci. We could do a coccinel script for everything else.
On Wed, Nov 22, 2023 at 03:22:54PM +0900, Damien Le Moal wrote: > I did not want to go as far as changing all drivers everywhere and limited the > series to drivers/pci. We could do a coccinel script for everything else. Yes. This is actually even trivial enough for sed :)
On 11/22/23 15:23, Christoph Hellwig wrote: > On Wed, Nov 22, 2023 at 03:22:54PM +0900, Damien Le Moal wrote: >> I did not want to go as far as changing all drivers everywhere and limited the >> series to drivers/pci. We could do a coccinel script for everything else. > > Yes. This is actually even trivial enough for sed :) Surprisingly, only 44 files use PCI_IRQ_LEGACY. Let me see how a patch look with the change.
On Wed, Nov 22, 2023 at 03:33:04PM +0900, Damien Le Moal wrote: > On 11/22/23 15:23, Christoph Hellwig wrote: > > On Wed, Nov 22, 2023 at 03:22:54PM +0900, Damien Le Moal wrote: > >> I did not want to go as far as changing all drivers everywhere and limited the > >> series to drivers/pci. We could do a coccinel script for everything else. > > > > Yes. This is actually even trivial enough for sed :) > > Surprisingly, only 44 files use PCI_IRQ_LEGACY. Let me see how a patch look > with the change. As mentioned in reply 1 I think this is perfect for a scripted run after -rc1. I'm actually surprised PCI_IRQ_LEGACY is used even 44 times. There is really no point in using the APIs based on PCI_IRQ_ for legacy irqs, and the case where it is just supposed as a fallback are covered by PCI_IRQ_ALL_TYPES.
On 11/22/23 15:36, Christoph Hellwig wrote: > On Wed, Nov 22, 2023 at 03:33:04PM +0900, Damien Le Moal wrote: >> On 11/22/23 15:23, Christoph Hellwig wrote: >>> On Wed, Nov 22, 2023 at 03:22:54PM +0900, Damien Le Moal wrote: >>>> I did not want to go as far as changing all drivers everywhere and limited the >>>> series to drivers/pci. We could do a coccinel script for everything else. >>> >>> Yes. This is actually even trivial enough for sed :) >> >> Surprisingly, only 44 files use PCI_IRQ_LEGACY. Let me see how a patch look >> with the change. > > As mentioned in reply 1 I think this is perfect for a scripted run > after -rc1. You mean 6.8-rc1 next cycle ? > > I'm actually surprised PCI_IRQ_LEGACY is used even 44 times. There is > really no point in using the APIs based on PCI_IRQ_ for legacy irqs, > and the case where it is just supposed as a fallback are covered by > PCI_IRQ_ALL_TYPES. I had a closer look and if we want to do this correctly, there is more to do than just the rename of PCI_IRQ_LEGACY to PCI_IRQ_INTX. There is also NR_IRQS_LEGACY, some Kconfig options, functions that use "legacy_irq" in there name, etc. A big chunk is just the rename, but some drivers will need finer handling as otherwise we'll endup mixing up intx and legacy in the same code, which is not pretty/confusing.
On Wed, Nov 22, 2023 at 03:49:28PM +0900, Damien Le Moal wrote: > > As mentioned in reply 1 I think this is perfect for a scripted run > > after -rc1. > > You mean 6.8-rc1 next cycle ? Yes. 6.7-rc1 is over :)
On 11/22/23 15:54, Christoph Hellwig wrote: > On Wed, Nov 22, 2023 at 03:49:28PM +0900, Damien Le Moal wrote: >>> As mentioned in reply 1 I think this is perfect for a scripted run >>> after -rc1. >> >> You mean 6.8-rc1 next cycle ? > > Yes. 6.7-rc1 is over :) OK. Bjorn, I can resend without this patch, or maybe you can drop it when applying. Let me know what you prefer.
On Wed, Nov 22, 2023 at 03:59:47PM +0900, Damien Le Moal wrote: > On 11/22/23 15:54, Christoph Hellwig wrote: > > On Wed, Nov 22, 2023 at 03:49:28PM +0900, Damien Le Moal wrote: > >>> As mentioned in reply 1 I think this is perfect for a scripted run > >>> after -rc1. > >> > >> You mean 6.8-rc1 next cycle ? > > > > Yes. 6.7-rc1 is over :) > > OK. > > Bjorn, > > I can resend without this patch, or maybe you can drop it when applying. Let me > know what you prefer. Krzysztof diligently made me notice, thanks. I have now dropped it and repushed out the resulting irq-clean-up branch. Lorenzo
On 12/19/23 19:07, Lorenzo Pieralisi wrote: > On Wed, Nov 22, 2023 at 03:59:47PM +0900, Damien Le Moal wrote: >> On 11/22/23 15:54, Christoph Hellwig wrote: >>> On Wed, Nov 22, 2023 at 03:49:28PM +0900, Damien Le Moal wrote: >>>>> As mentioned in reply 1 I think this is perfect for a scripted run >>>>> after -rc1. >>>> >>>> You mean 6.8-rc1 next cycle ? >>> >>> Yes. 6.7-rc1 is over :) >> >> OK. >> >> Bjorn, >> >> I can resend without this patch, or maybe you can drop it when applying. Let me >> know what you prefer. > > Krzysztof diligently made me notice, thanks. > > I have now dropped it and repushed out the resulting irq-clean-up > branch. OK. Thanks. I will work on removing what remains of "legacy" naming once everything is in linux next.
diff --git a/drivers/pci/pcie/portdrv.c b/drivers/pci/pcie/portdrv.c index 14a4b89a3b83..bb65dfe43409 100644 --- a/drivers/pci/pcie/portdrv.c +++ b/drivers/pci/pcie/portdrv.c @@ -187,15 +187,15 @@ static int pcie_init_service_irqs(struct pci_dev *dev, int *irqs, int mask) * interrupt. */ if ((mask & PCIE_PORT_SERVICE_PME) && pcie_pme_no_msi()) - goto legacy_irq; + goto intx_irq; /* Try to use MSI-X or MSI if supported */ if (pcie_port_enable_irq_vec(dev, irqs, mask) == 0) return 0; -legacy_irq: - /* fall back to legacy IRQ */ - ret = pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_LEGACY); +intx_irq: + /* fall back to INTX IRQ */ + ret = pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_INTX); if (ret < 0) return -ENODEV;
In the PCI Express Port Bus Driver, use the macro PCI_IRQ_INTX instead of the now deprecated PCI_IRQ_LEGACY macro. Signed-off-by: Damien Le Moal <dlemoal@kernel.org> --- drivers/pci/pcie/portdrv.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)