Re: [PATCH v2 3/6] powerpc/pci: use pci_is_pcie() to simplify code

On 2013/10/11 14:53, Gavin Shan wrote:

On Fri, Oct 11, 2013 at 02:33:58PM +0800, Yijing Wang wrote:

quoted

On 2013/10/11 14:16, Gavin Shan wrote:

quoted

On Fri, Oct 11, 2013 at 04:49:56PM +1100, Benjamin Herrenschmidt wrote:

quoted

On Fri, 2013-09-06 at 14:30 -0600, Bjorn Helgaas wrote:

quoted

On Thu, Sep 05, 2013 at 03:55:27PM +0800, Yijing Wang wrote:

.../...

quoted

quoted

quoted

quoted

quoted

Use pci_is_pcie() to simplify code.

diff --git a/arch/powerpc/kernel/eeh.c b/arch/powerpc/kernel/eeh.c
index 55593ee..6ebbe54 100644
--- a/arch/powerpc/kernel/eeh.c
+++ b/arch/powerpc/kernel/eeh.c

@@ -189,8 +189,7 @@ static size_t eeh_gather_pci_data(struct eeh_dev *edev, char * buf, size_t len)
 	}
 
 	/* If PCI-E capable, dump PCI-E cap 10, and the AER */
-	cap = pci_find_capability(dev, PCI_CAP_ID_EXP);
-	if (cap) {
+	if (pci_is_pcie(dev)) {
 		n += scnprintf(buf+n, len-n, "pci-e cap10:\n");
 		printk(KERN_WARNING
 		       "EEH: PCI-E capabilities and status follow:\n");

So we remove reading of "cap", but slightly further down the code does:

		for (i=0; i<=8; i++) {
			eeh_ops->read_config(dn, cap+4*i, 4, &cfg);
			n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg);
			printk(KERN_WARNING "EEH: PCI-E %02x: %08x\n", i, cfg);
		}

Which actually *uses* the value of "cap" ... oops :-)

It's my fault and I should have looked into the changes more closely.
How about changing it like this:

	cap = pci_is_pcie(dev) ? pci_pcie_cap(dev) :
	      pci_find_capability(dev, PCI_CAP_ID_EXP);
	if (cap) {
		...
	}

It would save some PCI-CFG access cycles for most cases :-)

Hi Gavin,  it's not your fault, it's my fault. :)

Because pci_pcie_cap(dev) == dev->pcie_cap == pci_find_capability(dev, PCI_CAP_ID_EXP);

so I think it's ok to use dev->pcie_cap instead of stale "cap".

Yijing, There has one exception: dev->pcie_cap isn't updated yet.

In my idea, dev->pcie_cap(here is pci_dev->pcie_cap) will update in set_pcie_port_type() function,
and this function always be called after allocate pci device. We get pci_dev by eeh_dev_to_pci_dev(),
I think pci_dev has been initialized completely.

This function has possibility to be invoked before that. However,
we don't have the binding (eeh device <-> PCI device) for the case.
So the piece of code shouldn't be running

In PCI core, I knew

pci_scan_device()
   pci_setup_device()
       set_pcie_port_type()
            pci_dev->pcie_cap = pci_find_capability(pdev, PCI_CAP_ID_EXP);

In powerpc, I also found

of_scan_pci_dev()
   of_create_pci_dev()
       set_pcie_port_type()
	    pci_dev->pcie_cap = pci_find_capability(pdev, PCI_CAP_ID_EXP);

However, it's a bit safer to have pci_find_capability(dev, PCI_CAP_ID_EXP)
as well even though we needn't it for 99.9% cases if you agree :-)

I agree, this function is not the performance bottleneck,
safety is more important. :)
So if Bjorn and Benjamin think it's not safe, it's ok to drop it. :)

Thanks!
Yijing.

Thanks,
Gavin


.

-- 
Thanks!
Yijing

On Fri, Oct 11, 2013 at 03:28:02PM +0800, Yijing Wang wrote:

On 2013/10/11 14:53, Gavin Shan wrote:

quoted

On Fri, Oct 11, 2013 at 02:33:58PM +0800, Yijing Wang wrote:

quoted

On 2013/10/11 14:16, Gavin Shan wrote:

quoted

On Fri, Oct 11, 2013 at 04:49:56PM +1100, Benjamin Herrenschmidt wrote:

quoted

On Fri, 2013-09-06 at 14:30 -0600, Bjorn Helgaas wrote:

quoted

On Thu, Sep 05, 2013 at 03:55:27PM +0800, Yijing Wang wrote:

.../...

quoted

quoted

quoted

quoted

quoted

Use pci_is_pcie() to simplify code.

diff --git a/arch/powerpc/kernel/eeh.c b/arch/powerpc/kernel/eeh.c
index 55593ee..6ebbe54 100644
--- a/arch/powerpc/kernel/eeh.c
+++ b/arch/powerpc/kernel/eeh.c

@@ -189,8 +189,7 @@ static size_t eeh_gather_pci_data(struct eeh_dev *edev, char * buf, size_t len)
 	}
 
 	/* If PCI-E capable, dump PCI-E cap 10, and the AER */
-	cap = pci_find_capability(dev, PCI_CAP_ID_EXP);
-	if (cap) {
+	if (pci_is_pcie(dev)) {
 		n += scnprintf(buf+n, len-n, "pci-e cap10:\n");
 		printk(KERN_WARNING
 		       "EEH: PCI-E capabilities and status follow:\n");

So we remove reading of "cap", but slightly further down the code does:

		for (i=0; i<=8; i++) {
			eeh_ops->read_config(dn, cap+4*i, 4, &cfg);
			n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg);
			printk(KERN_WARNING "EEH: PCI-E %02x: %08x\n", i, cfg);
		}

Which actually *uses* the value of "cap" ... oops :-)

It's my fault and I should have looked into the changes more closely.
How about changing it like this:

	cap = pci_is_pcie(dev) ? pci_pcie_cap(dev) :
	      pci_find_capability(dev, PCI_CAP_ID_EXP);
	if (cap) {
		...
	}

It would save some PCI-CFG access cycles for most cases :-)

Hi Gavin,  it's not your fault, it's my fault. :)

Because pci_pcie_cap(dev) == dev->pcie_cap == pci_find_capability(dev, PCI_CAP_ID_EXP);

so I think it's ok to use dev->pcie_cap instead of stale "cap".

Yijing, There has one exception: dev->pcie_cap isn't updated yet.

In my idea, dev->pcie_cap(here is pci_dev->pcie_cap) will update in set_pcie_port_type() function,
and this function always be called after allocate pci device. We get pci_dev by eeh_dev_to_pci_dev(),
I think pci_dev has been initialized completely.

quoted

This function has possibility to be invoked before that. However,
we don't have the binding (eeh device <-> PCI device) for the case.
So the piece of code shouldn't be running

In PCI core, I knew

pci_scan_device()
  pci_setup_device()
      set_pcie_port_type()
           pci_dev->pcie_cap = pci_find_capability(pdev, PCI_CAP_ID_EXP);

In powerpc, I also found

of_scan_pci_dev()
  of_create_pci_dev()
      set_pcie_port_type()
    pci_dev->pcie_cap = pci_find_capability(pdev, PCI_CAP_ID_EXP);

quoted

However, it's a bit safer to have pci_find_capability(dev, PCI_CAP_ID_EXP)
as well even though we needn't it for 99.9% cases if you agree :-)

I agree, this function is not the performance bottleneck,
safety is more important. :)
So if Bjorn and Benjamin think it's not safe, it's ok to drop it. :)

No, it's not what I mean. Anyway, "v3" looks good to me.
At least, it can save PCI-CFG access cycles find locate
the PCIe capability position :-)

Thanks,
Gavin