Re: MSI interrupts and disable_irq

[Jeff Garzik]

Yinghai Lu wrote:
> On 10/16/07, Jeff Garzik <jgarzik@xxxxxxxxx> wrote:
> > Yinghai Lu wrote:
> > > On 10/15/07, Jeff Garzik <jgarzik@xxxxxxxxx> wrote:
> > > > Manfred Spraul wrote:
> > > > > Jeff Garzik wrote:
> > > > > > I think the scenario you outline is an illustration of the approach's
> > > > > > fragility:  disable_irq() is a heavy hammer that originated with INTx,
> > > > > > and it relies on a chip-specific disable method (kernel/irq/manage.c)
> > > > > > that practically guarantees behavior will vary across MSI/INTx/etc.
> > > > > I checked the code: IRQ_DISABLE is implemented in software, i.e.
> > > > > handle_level_irq() only calls handle_IRQ_event() [and then the nic irq
> > > > > handler] if IRQ_DISABLE is not set.
> > > > > OTHO: The last trace looks as if nv_do_nic_poll() is interrupted by an irq.
> > > > >
> > > > > Perhaps something corrupts dev->irq? The irq is requested with
> > > > >    request_irq(np->pci_dev->irq, handler, IRQF_SHARED, dev->name, dev)
> > > > > and disabled with
> > > > >    disable_irq_lockdep(dev->irq);
> > > > >
> > > > > Someone around with a MSI capable board? The forcedeth driver does
> > > > >    dev->irq = pci_dev->irq
> > > > > in nv_probe(), especially before pci_enable_msi().
> > > > > Does pci_enable_msi() change pci_dev->irq? Then we would disable the
> > > > > wrong interrupt....
> > > > Remember, fundamentally MSI-X is a one-to-many relationship, when you
> > > > consider a single PCI device might have multiple vectors.
> > > msi-x is using other entry
> > >
> > >                if (np->msi_flags & NV_MSI_X_ENABLED)
> > >
> > > enable_irq_lockdep(np->msi_x_entry[NV_MSI_X_VECTOR_ALL].vector);
> > Correct, but the overall point was that MSI-X conceptually conflicts
> > with the existing "lockless" disable_irq() schedule, which was written
> > when there was a one-one relationship between irq, PCI device, and work
> > to be done.
>
> Can I use your new driver with RHEL 5 or RHEL 5.1?

Not without modification, since it depends on the napi_struct work 
currently in torvalds/linux-2.6.git.

But I am currently rewriting the fe-lock yet again, and most of those 
changes can be applied to pre-napi_struct forcedeth.

	Jeff



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/