On 10/10/2011 12:00 PM, Bjorn Helgaas wrote:
> On Mon, Oct 10, 2011 at 9:17 AM, Prarit Bhargava <prarit@xxxxxxxxxx> wrote:
>>
>>>
>>> what is the point sriov=1? are you going to enable sriov forcefully?
>>>
>>> pci=nosriov
>>
>> Hi Yinghai,
>>
>> I'm looking at this from a distribution's point-of-view.
>>
>> SRIOV is built-in but purposely disabled by the distro for a particular system via an SMBIOS check so that pci_sriov_enabled = 0.
>>
>> In that case, when the system was functional, through a FW update, I will want to turn it on without having to rebuild the kernel. ie) specifying pci=sriov=1 makes sense to test or debug in that case.
>>
>> If that isn't acceptable to upstream, I would have no objection to pci=nosriov.
>
> What about
>
> pci=sriov
> pci=nosriov
>
> The existing PCI options that use "=" are mostly things that require
> an entire value (bitmask, size, etc.) rather than just a boolean, and
> there's some precedent for using "OPTION and noOPTION" for boolean
> things (bios/nobios, rom/norom, bfsort/nobfsort).
Sure ... I can do that. But Yinghai is requesting just "pci=nosriov"
which I'll implement.
>
> Would you mind mentioning "PCI" in the printk message along with
> "SRIOV" just to help dmesg readers put it in context?
Done.
>
> Would it be overkill to mention either in
> Documentation/kernel-parameters.txt or in the dmesg that if you needed
> to use this option, you should report a bug? I've been disappointed
> by how many times I've found reports on the web saying "pci=use_crs"
> is the final fix, when it's really only a band-aid that we should fix
> properly in the kernel. I don't think end-users should ever have to
> know about options like this, but sometimes they become part of the
> folklore of random things that people try, which reflects poorly on
> Linux.
I completely agree with this Bjorn but ...
I'm clueless about asking an end-user to report a bug
because the question becomes where/which company or group
do they report the bug to?
To lkml? linux-pci? bugzilla.kernel.org? The distribution's bug
reporting tool? The hardware vendor?
Either way, I have put
"Please report a bug if you use this kernel parameter."
On 10/10/2011 02:58 PM, Yinghai Lu wrote:
> pci=nosriov should be good enough.
Okay.
P.
[PATCH] Add pci=nosriov to disable SRIOV.
To help debugging, add a kernel parameter, pci=nosriov which disables SRIOV.
Signed-off-by: Prarit Bhargava <prarit@xxxxxxxxxx>
Cc: ddutile@xxxxxxxxxx
Cc: bhelgaas@xxxxxxxxxx
---
Documentation/kernel-parameters.txt | 3 +++
drivers/pci/iov.c | 19 ++++++++++++++++++-
drivers/pci/pci.c | 2 ++
drivers/pci/pci.h | 3 +++
4 files changed, 26 insertions(+), 1 deletions(-)
diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 854ed5c..429c7e6 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -2032,6 +2032,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
on: Turn ECRC on.
realloc reallocate PCI resources if allocations done by BIOS
are erroneous.
+ nosriov= [HW]
+ Disable SRIOV support in kernel.
+ Please report a bug if you use this parameter.
pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
Management.
diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
index 42fae47..260c94d 100644
--- a/drivers/pci/iov.c
+++ b/drivers/pci/iov.c
@@ -408,6 +408,8 @@ static void sriov_disable(struct pci_dev *dev)
iov->nr_virtfn = 0;
}
+unsigned int pci_sriov_enabled = 1;
+
static int sriov_init(struct pci_dev *dev, int pos)
{
int i;
@@ -419,6 +421,12 @@ static int sriov_init(struct pci_dev *dev, int pos)
struct resource *res;
struct pci_dev *pdev;
+ if (!pci_sriov_enabled) {
+ rc = -EPERM;
+ /* in theory no flags should have been set ... */
+ goto failed;
+ }
+
if (dev->pcie_type != PCI_EXP_TYPE_RC_END &&
dev->pcie_type != PCI_EXP_TYPE_ENDPOINT)
return -ENODEV;
@@ -614,7 +622,10 @@ resource_size_t pci_sriov_resource_alignment(struct pci_dev *dev, int resno)
struct resource tmp;
enum pci_bar_type type;
int reg = pci_iov_resource_bar(dev, resno, &type);
-
+
+ if (!pci_sriov_enabled)
+ return 0;
+
if (!reg)
return 0;
@@ -667,6 +678,9 @@ int pci_enable_sriov(struct pci_dev *dev, int nr_virtfn)
{
might_sleep();
+ if (!pci_sriov_enabled)
+ return -EPERM;
+
if (!dev->is_physfn)
return -ENODEV;
@@ -682,6 +696,9 @@ void pci_disable_sriov(struct pci_dev *dev)
{
might_sleep();
+ if (!pci_sriov_enabled)
+ return;
+
if (!dev->is_physfn)
return;
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index e9651f0..11175cb 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -3576,6 +3576,8 @@ static int __init pci_setup(char *str)
pcie_bus_config = PCIE_BUS_PERFORMANCE;
} else if (!strncmp(str, "pcie_bus_peer2peer", 18)) {
pcie_bus_config = PCIE_BUS_PEER2PEER;
+ } else if (!strncmp(str, "nosriov", 7)) {
+ pci_sriov_enabled = 0;
} else {
printk(KERN_ERR "PCI: Unknown option `%s'\n",
str);
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index b74084e..9eeb56d 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -229,6 +229,9 @@ resource_size_t pci_specified_resource_alignment(struct pci_dev *dev);
extern void pci_disable_bridge_window(struct pci_dev *dev);
#endif
+/* Use "pci=nosriov" to disable SRIOV */
+extern unsigned int pci_sriov_enabled;
+
/* Single Root I/O Virtualization */
struct pci_sriov {
int pos; /* capability position */
--
1.7.1
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
