diff mbox

PCI: dra7xx: Use PCI_NUM_INTX

Message ID 20170817165354.GI28977@bhelgaas-glaptop.roam.corp.google.com (mailing list archive)
State New, archived
Delegated to: Bjorn Helgaas
Headers show

Commit Message

Bjorn Helgaas Aug. 17, 2017, 4:53 p.m. UTC 
On Thu, Aug 17, 2017 at 09:51:29PM +0530, Kishon Vijay Abraham I wrote:
> Hi Bjorn,
> 
> On Wednesday 16 August 2017 03:22 AM, Bjorn Helgaas wrote:
> > On Tue, Aug 15, 2017 at 04:38:38PM -0500, Bjorn Helgaas wrote:
> >> Use the PCI_NUM_INTX macro to indicate the number of PCI INTx interrupts
> >> rather than the magic number 4. This makes it clearer where the number
> >> comes from & what it relates to.
> >>
> >> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
> >> Cc: Kishon Vijay Abraham I <kishon@ti.com>
> >> ---
> >>  drivers/pci/dwc/pci-dra7xx.c |    2 +-
> >>  1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/pci/dwc/pci-dra7xx.c b/drivers/pci/dwc/pci-dra7xx.c
> >> index f2fc5f47064e..30131ecaadea 100644
> >> --- a/drivers/pci/dwc/pci-dra7xx.c
> >> +++ b/drivers/pci/dwc/pci-dra7xx.c
> >> @@ -238,7 +238,7 @@ static int dra7xx_pcie_init_irq_domain(struct pcie_port *pp)
> >>  		return -ENODEV;
> >>  	}
> >>  
> >> -	dra7xx->irq_domain = irq_domain_add_linear(pcie_intc_node, 4,
> >> +	dra7xx->irq_domain = irq_domain_add_linear(pcie_intc_node, PCI_NUM_INTX,
> >>  						   &intx_domain_ops, pp);
> >>  	if (!dra7xx->irq_domain) {
> >>  		dev_err(dev, "Failed to get a INTx IRQ domain\n");
> >>
> > 
> > Oops.  I think this patch is OK as far as it goes, but Kishon
> > confirmed [1] that INTD was broken in dra7xx, and AFAIK it's still
> > broken, so maybe we need to also use pci_irqd_intx_xlate() as in the
> > following?
> > 
> > [1] https://lkml.org/lkml/2016/9/14/241
> > 
> > 
> > 
> > commit 569f29aa4ff0ab4bd4d1782b3a806a7561fe9db5
> > Author: Bjorn Helgaas <bhelgaas@google.com>
> > Date:   Tue Aug 15 16:28:27 2017 -0500
> > 
> >     PCI: dra7xx: Translate INTx range to hwirqs 0-3
> >     
> >     The pci-dra7xx driver creates an IRQ domain of size 4 for legacy PCI INTx
> >     interrupts, which at first glance seems reasonable since there are 4 possible
> >     such interrupts. Unfortunately the driver then proceeds to use the range 1-4
> >     as the hwirq numbers for INTA-INTD, causing warnings & broken interrupts when
> >     attempting to use INTD/hwirq=4 due to it being beyond the range of the IRQ
> >     domain:
> >     
> >       WARNING: CPU: 0 PID: 1 at kernel/irq/irqdomain.c:342 irq_domain_associate+0x12c/0x1c4
> >       error: hwirq 0x4 is too large for dummy
> >       Modules linked in:
> >       CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.8.0-rc6-01351-gd4fcf3d-dirty #15
> >       Hardware name: Generic DRA72X (Flattened Device Tree)
> >         (unwind_backtrace) from [<c010c264>] (show_stack+0x10/0x14)
> >         (show_stack) from [<c048f3ec>] (dump_stack+0xac/0xe0)
> >         (dump_stack) from [<c0138a74>] (__warn+0xd8/0x104)
> >         (__warn) from [<c0138ad4>] (warn_slowpath_fmt+0x34/0x44)
> >         (warn_slowpath_fmt) from [<c01a9944>] (irq_domain_associate+0x12c/0x1c4)
> >         (irq_domain_associate) from [<c01aa158>] (irq_create_mapping+0x64/0xcc)
> >         (irq_create_mapping) from [<c01aa848>] (irq_create_fwspec_mapping+0xac/0x2fc)
> >         (irq_create_fwspec_mapping) from [<c01aaaec>] (irq_create_of_mapping+0x54/0x5c)
> >         (irq_create_of_mapping) from [<c0668b04>] (of_irq_parse_and_map_pci+0x24/0x2c)
> >         (of_irq_parse_and_map_pci) from [<c04ea578>] (pci_fixup_irqs+0x44/0xb8)
> >         (pci_fixup_irqs) from [<c04eb8dc>] (dw_pcie_host_init+0x234/0x3e4)
> >         (dw_pcie_host_init) from [<c0b38fa4>] (dra7xx_pcie_probe+0x418/0x5c4)
> >         (dra7xx_pcie_probe) from [<c0557260>] (platform_drv_probe+0x4c/0xb0)
> >     
> >     Fix this by making use of the new pci_irqd_intx_xlate() helper to translate
> >     the INTx 1-4 range into the 0-3 range suitable for the IRQ domain of size
> >     4, and stop adding 1 to the hwirq number decoded from the interrupt FIFO
> >     which is already in the range 0-3.
> >     
> >     Whilst we're here we switch to using PCI_NUM_INTX rather than the magic
> >     number 4, making it clearer what the 4 means.
> >     
> >     Based-on-similar-patches-by: Paul Burton <paul.burton@imgtec.com>
> >     Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
> >     Cc: Kishon Vijay Abraham I <kishon@ti.com>
> > 
> > diff --git a/drivers/pci/dwc/pci-dra7xx.c b/drivers/pci/dwc/pci-dra7xx.c
> > index f2fc5f47064e..55f16fcf701d 100644
> > --- a/drivers/pci/dwc/pci-dra7xx.c
> > +++ b/drivers/pci/dwc/pci-dra7xx.c
> > @@ -223,6 +223,7 @@ static int dra7xx_pcie_intx_map(struct irq_domain *domain, unsigned int irq,
> >  
> >  static const struct irq_domain_ops intx_domain_ops = {
> >  	.map = dra7xx_pcie_intx_map,
> > +	.xlate = pci_irqd_intx_xlate,
> >  };
> >  
> >  static int dra7xx_pcie_init_irq_domain(struct pcie_port *pp)
> > @@ -238,7 +239,7 @@ static int dra7xx_pcie_init_irq_domain(struct pcie_port *pp)
> >  		return -ENODEV;
> >  	}
> >  
> > -	dra7xx->irq_domain = irq_domain_add_linear(pcie_intc_node, 4,
> > +	dra7xx->irq_domain = irq_domain_add_linear(pcie_intc_node, PCI_NUM_INTX,
> >  						   &intx_domain_ops, pp);
> >  	if (!dra7xx->irq_domain) {
> >  		dev_err(dev, "Failed to get a INTx IRQ domain\n");
> 
> With this patch, I don't see [1] mentioned anymore.
> However I think there are other issues in pci-dra7xx which doesn't allow me to
> use INTD. When I tried to set 0x4 in interrupt_pin of pcie endpoint and raise
> legacy interrupt, I get the following dump in RC.
> [  102.348957] irq 0, desc: ee80d800, depth: 1, count: 0, unhandled: 0
> [  102.355536] ->handle_irq():  c01a6994,
> [  102.355553] handle_bad_irq+0x0/0x268
> [  102.363311] ->irq_data.chip(): c0d46f50,
> [  102.363325] no_irq_chip+0x0/0x88
> [  102.370903] ->action():   (null)
> [  102.374290]    IRQ_NOPROBE set
> [  102.377490]  IRQ_NOREQUEST set
> [  102.380692] unexpected IRQ trap at vector 00
> 
> I'll look into this.

Thanks for trying this out.  In the meantime, I dropped the
translation part of the patch, resulting in the patch below.
I don't want to exchange the "error: hwirq 0x4 is too large for dummy"
problem for another unknown problem.

If/when you fix the INTD issue, pci_irqd_intx_xlate() might be part of
the solution, but sounds like there's more to it.

[1] https://lkml.org/lkml/2016/9/14/241


commit 61534d1a4c79501261bb8c534f992e8c8e1353da
Author: Bjorn Helgaas <bhelgaas@google.com>
Date:   Tue Aug 15 16:28:27 2017 -0500

    PCI: dra7xx: Use PCI_NUM_INTX
    
    Use the PCI_NUM_INTX macro to indicate the number of PCI INTx interrupts
    rather than the magic number 4. This makes it clearer where the number
    comes from & what it relates to.
    
    Based-on-similar-patches-by: Paul Burton <paul.burton@imgtec.com>
    Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
    Cc: Kishon Vijay Abraham I <kishon@ti.com>
diff mbox

Patch

diff --git a/drivers/pci/dwc/pci-dra7xx.c b/drivers/pci/dwc/pci-dra7xx.c
index f2fc5f47064e..30131ecaadea 100644
--- a/drivers/pci/dwc/pci-dra7xx.c
+++ b/drivers/pci/dwc/pci-dra7xx.c
@@ -238,7 +238,7 @@  static int dra7xx_pcie_init_irq_domain(struct pcie_port *pp)
 		return -ENODEV;
 	}
 
-	dra7xx->irq_domain = irq_domain_add_linear(pcie_intc_node, 4,
+	dra7xx->irq_domain = irq_domain_add_linear(pcie_intc_node, PCI_NUM_INTX,
 						   &intx_domain_ops, pp);
 	if (!dra7xx->irq_domain) {
 		dev_err(dev, "Failed to get a INTx IRQ domain\n");