| Message ID | 564EECA9.9070400@arm.com |
|---|---|
| State | Superseded |
| Headers | show |
Hi Marc, On 20 November 2015 09:49, Marc Zyngier wrote: > On 18/11/15 18:01, Phil Edworthy wrote: > > Hi Marc, > > > > On 16 November 2015 18:31, Marc Zyngier wrote: > >> On 13/11/15 09:36, Phil Edworthy wrote: > > <snip> > >>> Since the stack trace doesn't help that much I added some tracing: > >>> pci_msi_setup_msi_irqs() > >>> calls pci_msi_get_domain() > >>> calls dev_get_msi_domain(), gets a non-NULL domain. > >>> pci_msi_setup_msi_irqs() > >>> calls pci_msi_domain_alloc_irqs() > >>> calls msi_domain_alloc_irqs() > >>> msi_domain_alloc_irqs:273: ops=ffffffc03193a810 > >>> msi_domain_alloc_irqs:274: ops->msi_check=ffffffc031161418 > >>> systemd-udevd[1311]: undefined instruction: pc=ffffffc03116141c > >>> That looks to me as though msi_check is off pointing to the weeds. > >> > >> So the next step is to find out who initializes msi_check. Assuming > >> someone does... > > Nothing initializes msi_check... > > > > > >>> By passing a NULL domain into irq_domain_add_linear() you get: > >>> pci_msi_setup_msi_irqs() > >>> calls pci_msi_get_domain() > >>> calls dev_get_msi_domain(), gets a NULL domain. > >>> calls arch_setup_msi_irq() > >>> All ok then. > >> > >> Yes, because you're sidestepping the issue. Any chance you could dig a > >> bit deeper? I'd really like to nail this one down (before we convert > >> your PCI driver to the right API... ;-). > > The problem appears to be that when the pci host driver enables msi > > it calls the following: > > msi->domain = irq_domain_add_linear(pcie->dev->of_node, > INT_PCI_MSI_NR, > > &msi_domain_ops, &msi->chip); > > The last arg is documented as: > > * @host_data: Controller private data pointer > > In _irq_domain_add() this ptr is stored in struct irq_domain's host_data. > > > > However, msi_domain_alloc_irqs() expects host_data to be a ptr to a > > struct msi_domain_info. > > > > It seems that a number of other pci host drivers do the same, so I am > > surprised that no one else has seen this. > > Can you please give this hack a go and let me know if that helps? Works for me! Many thanks Phil > diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c > index 53e4632..7eaa4c8 100644 > --- a/drivers/pci/msi.c > +++ b/drivers/pci/msi.c > @@ -54,7 +54,7 @@ static int pci_msi_setup_msi_irqs(struct pci_dev *dev, int > nvec, int type) > struct irq_domain *domain; > > domain = pci_msi_get_domain(dev); > - if (domain) > + if (domain && irq_domain_is_hierarchy(domain)) > return pci_msi_domain_alloc_irqs(domain, dev, nvec, type); > > return arch_setup_msi_irqs(dev, nvec, type); > @@ -65,7 +65,7 @@ static void pci_msi_teardown_msi_irqs(struct pci_dev *dev) > struct irq_domain *domain; > > domain = pci_msi_get_domain(dev); > - if (domain) > + if (domain && irq_domain_is_hierarchy(domain)) > pci_msi_domain_free_irqs(domain, dev); > else > arch_teardown_msi_irqs(dev); > > -- > Jazz is not dead. It just smells funny... -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, 23 Nov 2015 09:44:10 +0000 Phil Edworthy <phil.edworthy@renesas.com> wrote: > Hi Marc, > > On 20 November 2015 09:49, Marc Zyngier wrote: > > On 18/11/15 18:01, Phil Edworthy wrote: > > > Hi Marc, > > > > > > On 16 November 2015 18:31, Marc Zyngier wrote: > > >> On 13/11/15 09:36, Phil Edworthy wrote: > > > <snip> > > >>> Since the stack trace doesn't help that much I added some tracing: > > >>> pci_msi_setup_msi_irqs() > > >>> calls pci_msi_get_domain() > > >>> calls dev_get_msi_domain(), gets a non-NULL domain. > > >>> pci_msi_setup_msi_irqs() > > >>> calls pci_msi_domain_alloc_irqs() > > >>> calls msi_domain_alloc_irqs() > > >>> msi_domain_alloc_irqs:273: ops=ffffffc03193a810 > > >>> msi_domain_alloc_irqs:274: ops->msi_check=ffffffc031161418 > > >>> systemd-udevd[1311]: undefined instruction: pc=ffffffc03116141c > > >>> That looks to me as though msi_check is off pointing to the weeds. > > >> > > >> So the next step is to find out who initializes msi_check. Assuming > > >> someone does... > > > Nothing initializes msi_check... > > > > > > > > >>> By passing a NULL domain into irq_domain_add_linear() you get: > > >>> pci_msi_setup_msi_irqs() > > >>> calls pci_msi_get_domain() > > >>> calls dev_get_msi_domain(), gets a NULL domain. > > >>> calls arch_setup_msi_irq() > > >>> All ok then. > > >> > > >> Yes, because you're sidestepping the issue. Any chance you could dig a > > >> bit deeper? I'd really like to nail this one down (before we convert > > >> your PCI driver to the right API... ;-). > > > The problem appears to be that when the pci host driver enables msi > > > it calls the following: > > > msi->domain = irq_domain_add_linear(pcie->dev->of_node, > > INT_PCI_MSI_NR, > > > &msi_domain_ops, &msi->chip); > > > The last arg is documented as: > > > * @host_data: Controller private data pointer > > > In _irq_domain_add() this ptr is stored in struct irq_domain's host_data. > > > > > > However, msi_domain_alloc_irqs() expects host_data to be a ptr to a > > > struct msi_domain_info. > > > > > > It seems that a number of other pci host drivers do the same, so I am > > > surprised that no one else has seen this. > > > > Can you please give this hack a go and let me know if that helps? > Works for me! Cool. I'll post a patch with your Reported-by/Tested-by later today. Thanks a lot. M.
Hi Marc, > > > Can you please give this hack a go and let me know if that helps? > > Works for me! > > Cool. I'll post a patch with your Reported-by/Tested-by later today. Please put me on CC as well. Thanks, Wolfram
diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c index 53e4632..7eaa4c8 100644 --- a/drivers/pci/msi.c +++ b/drivers/pci/msi.c @@ -54,7 +54,7 @@ static int pci_msi_setup_msi_irqs(struct pci_dev *dev, int nvec, int type) struct irq_domain *domain; domain = pci_msi_get_domain(dev); - if (domain) + if (domain && irq_domain_is_hierarchy(domain)) return pci_msi_domain_alloc_irqs(domain, dev, nvec, type); return arch_setup_msi_irqs(dev, nvec, type); @@ -65,7 +65,7 @@ static void pci_msi_teardown_msi_irqs(struct pci_dev *dev) struct irq_domain *domain; domain = pci_msi_get_domain(dev); - if (domain) + if (domain && irq_domain_is_hierarchy(domain)) pci_msi_domain_free_irqs(domain, dev); else arch_teardown_msi_irqs(dev);