Skip to contentSkip to conversation

PCI/portdrv: Set master earlier

PATCH In review Jean Delvare 4 participants 6 messages Oct 9, 2013 – Dec 17, 2013 PCI

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 PCI/portdrv: Set master earlier

Since kernel 3.12-rc3, I get the following warning messages at boot:
pcieport 0000:00:07.0: driver skip pci_set_master, fix it! pcieport 0000:00:01.0: driver skip pci_set_master, fix it!

These are:
00:01.0 PCI bridge: Intel Corporation 5520/5500/X58 I/O Hub PCI Express Root Port 1 (rev 13) 00:07.0 PCI bridge: Intel Corporation 5520/5500/X58 I/O Hub PCI Express Root Port 7 (rev 13)

Calling pci_set_master() immediately after pci_enable_device() makes the warning messages go away.

---
Please note that I am far from certain that this is the correct fix. Another approach would be to disable the device if get_port_device_capability() returns no capabilities. I suppose the device would then be re-enabled later if really needed. What's the best approach?

 drivers/pci/pcie/portdrv_core.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/pcie/portdrv_core.c +1 -1
@@ -366,13 +366,13 @@ int pcie_port_device_register(struct pci
 	status = pci_enable_device(dev);
 	if (status)
 		return status;
+	pci_set_master(dev);
 
 	/* Get and check PCI Express port services */
 	capabilities = get_port_device_capability(dev);
 	if (!capabilities)
 		return 0;
 
-	pci_set_master(dev);
 	/*
 	 * Initialize service irqs. Don't use service devices that
 	 * require interrupts if there is no way to generate them.
Signature · 2 lines
Jean Delvare
Suse L3 Support
maintainer
#1 lore raw

[+cc Yinghai]

Quoted 3 lines · Jean Delvare

Since kernel 3.12-rc3, I get the following warning messages at boot:
pcieport 0000:00:07.0: driver skip pci_set_master, fix it! pcieport 0000:00:01.0: driver skip pci_set_master, fix it!

cf3e1feba7f9 ("PCI: Workaround missing pci_set_master in pci drivers") added this warning and, at the same time, turned on bus mastering.

There didn't seem to be any reason for the warning and no reason why bus mastering should be enabled in the driver rather than the core, so fbeeb822f6f4 ("PCI: Drop warning about drivers that don't use pci_set_master()") just dropped the warning.

This should resolve the problem, so I don't think there's a need for this patch. Let me know if otherwise.

Bjorn

Quoted 41 lines

These are:
00:01.0 PCI bridge: Intel Corporation 5520/5500/X58 I/O Hub PCI Express Root Port 1 (rev 13) 00:07.0 PCI bridge: Intel Corporation 5520/5500/X58 I/O Hub PCI Express Root Port 7 (rev 13)

Calling pci_set_master() immediately after pci_enable_device() makes the warning messages go away.

---
Please note that I am far from certain that this is the correct fix. Another approach would be to disable the device if get_port_device_capability() returns no capabilities. I suppose the device would then be re-enabled later if really needed. What's the best approach?

 drivers/pci/pcie/portdrv_core.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/pcie/portdrv_core.c 2013-10-09 15:43:54.205943783 +0200 renamed from drivers/pci/pcie/portdrv_core.c 2013-09-24 00:41:09.000000000 +0200 +1 -2
@@ -366,13 +366,13 @@ int pcie_port_device_register(struct pci
        status = pci_enable_device(dev);
        if (status)
                return status;
+       pci_set_master(dev);
 
        /* Get and check PCI Express port services */
        capabilities = get_port_device_capability(dev);
        if (!capabilities)
                return 0;
 
-       pci_set_master(dev);
        /*
         * Initialize service irqs. Don't use service devices that
         * require interrupts if there is no way to generate them.
 
 
--

Jean Delvare
Suse L3 Support

#2 lore raw

[+to BenH]

Quoted 10 lines · Bjorn Helgaas

[+cc Yinghai]

Quoted 3 lines · Jean Delvare

Since kernel 3.12-rc3, I get the following warning messages at boot:
pcieport 0000:00:07.0: driver skip pci_set_master, fix it! pcieport 0000:00:01.0: driver skip pci_set_master, fix it!

cf3e1feba7f9 ("PCI: Workaround missing pci_set_master in pci drivers") added this warning and, at the same time, turned on bus mastering.

Quoted 7 lines

There didn't seem to be any reason for the warning and no reason why bus mastering should be enabled in the driver rather than the core, so fbeeb822f6f4 ("PCI: Drop warning about drivers that don't use pci_set_master()") just dropped the warning.

This should resolve the problem, so I don't think there's a need for this patch. Let me know if otherwise.

BenH mentioned that get_port_device_capability have some problem on powerpc and will return null.

Quoted 23 lines
Quoted 23 lines
 drivers/pci/pcie/portdrv_core.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/pcie/portdrv_core.c 2013-10-09 15:43:54.205943783 +0200 renamed from drivers/pci/pcie/portdrv_core.c 2013-09-24 00:41:09.000000000 +0200 +1 -1
@@ -366,13 +366,13 @@ int pcie_port_device_register(struct pci
        status = pci_enable_device(dev);
        if (status)
                return status;
+       pci_set_master(dev);
 
        /* Get and check PCI Express port services */
        capabilities = get_port_device_capability(dev);
        if (!capabilities)
                return 0;
 
-       pci_set_master(dev);
        /*
         * Initialize service irqs. Don't use service devices that
         * require interrupts if there is no way to generate them.
#3 lore raw
Quoted 2 lines · Yinghai Lu

BenH mentioned that get_port_device_capability have some problem on powerpc and will return null.

Well, it has problems on anything that doesn't have ACPI :-)

I don't care what the "right" fix is but we need to makes sure nobody does a pci_enable_device() on a bridge and doesn't also enable bus master, otherwise the bridge bus master will never be set. This is what was happening an breaking us.

Ben.

maintainer
#4 lore raw
Quoted 10 lines · Benjamin Herrenschmidt
Quoted 2 lines · Yinghai Lu

BenH mentioned that get_port_device_capability have some problem on powerpc and will return null.

Well, it has problems on anything that doesn't have ACPI :-)

I don't care what the "right" fix is but we need to makes sure nobody does a pci_enable_device() on a bridge and doesn't also enable bus master, otherwise the bridge bus master will never be set. This is what was happening an breaking us.

I assume there's nothing more I need to do here. If there is, please let me know.

Bjorn

#5 lore raw
In reply to Bjorn Helgaas

Hi Bjorn,

Quoted 10 lines · Bjorn Helgaas

[+cc Yinghai]

Quoted 3 lines · Jean Delvare

Since kernel 3.12-rc3, I get the following warning messages at boot:
pcieport 0000:00:07.0: driver skip pci_set_master, fix it! pcieport 0000:00:01.0: driver skip pci_set_master, fix it!

cf3e1feba7f9 ("PCI: Workaround missing pci_set_master in pci drivers") added this warning and, at the same time, turned on bus mastering.

Quoted 7 lines

There didn't seem to be any reason for the warning and no reason why bus mastering should be enabled in the driver rather than the core, so fbeeb822f6f4 ("PCI: Drop warning about drivers that don't use pci_set_master()") just dropped the warning.

This should resolve the problem, so I don't think there's a need for this patch. Let me know if otherwise.

This is fine with me, all I really wanted was to get rid of the warning. I am not aware of any problem either (but I am no PCI expert by any means.)

Thanks,

Signature · 2 lines
Jean Delvare
Suse L3 Support