Re: [PATCH v2] iov: restore NumVFs register to 0 before return from virtfn_max_buses()

From: Alexander Duyck
Date: Tue Oct 27 2015 - 01:48:27 EST


On 10/15/2015 10:16 AM, Bjorn Helgaas wrote:
Hi Ethan,

On Wed, Sep 16, 2015 at 12:19:53PM +0900, Ethan Zhao wrote:
After commit 4449f079722c ("PCI: Calculate maximum number of buses
required for VFs"),the initial value of NumVFs register was left to
non-zero after sriov_init() and no VFs was enabled in device driver.
this changed the behaviour of kernel exported by lspci and sysfs etc.
so this patch restore the NumVFs register to zero after the
calculation of max_VF_buses was done and before return from
virtfn_max_buses().

Tested on stable 4.1 and passed building on stable 4.3-rc1

Signed-off-by: Ethan Zhao <ethan.zhao@xxxxxxxxxx>
Tested-by: Sriharsha Yadagudde <sriharsha.devdas@xxxxxxxxxx>
Can you test the patch below? I'm trying to avoid touching
PCI_SRIOV_NUM_VF in more than one place, and I think it's OK to set it
and test offset/stride at the end, instead of setting NUM_VF to zero,
testing offset/stride, computing max_bus, then setting NUM_VF to zero
again.

Bjorn


commit 8e20e89658f23b8d16b1e21810e9f63c8625129c
Author: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
Date: Thu Oct 15 11:31:21 2015 -0500

PCI: Set SR-IOV NumVFs to zero after enumeration
The enumeration path should leave NumVFs set to zero. But after
4449f079722c ("PCI: Calculate maximum number of buses required for VFs"),
we call virtfn_max_buses() in the enumeration path, which changes NumVFs.
This NumVFs change is visible via lspci and sysfs until a driver enables
SR-IOV.
Set NumVFs to zero after virtfn_max_buses() computes the maximum number of
buses.
Fixes: 4449f079722c ("PCI: Calculate maximum number of buses required for VFs")
Based-on-patch-by: Ethan Zhao <ethan.zhao@xxxxxxxxxx>
Signed-off-by: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>

diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
index ee0ebff..0202ab0 100644
--- a/drivers/pci/iov.c
+++ b/drivers/pci/iov.c
@@ -384,7 +384,7 @@ static int sriov_init(struct pci_dev *dev, int pos)
int rc;
int nres;
u32 pgsz;
- u16 ctrl, total, offset, stride;
+ u16 ctrl, total;
struct pci_sriov *iov;
struct resource *res;
struct pci_dev *pdev;
@@ -414,11 +414,6 @@ static int sriov_init(struct pci_dev *dev, int pos)
found:
pci_write_config_word(dev, pos + PCI_SRIOV_CTRL, ctrl);
- pci_write_config_word(dev, pos + PCI_SRIOV_NUM_VF, 0);
- pci_read_config_word(dev, pos + PCI_SRIOV_VF_OFFSET, &offset);
- pci_read_config_word(dev, pos + PCI_SRIOV_VF_STRIDE, &stride);
- if (!offset || (total > 1 && !stride))
- return -EIO;
pci_read_config_dword(dev, pos + PCI_SRIOV_SUP_PGSIZE, &pgsz);
i = PAGE_SHIFT > 12 ? PAGE_SHIFT - 12 : 0;
@@ -456,8 +451,6 @@ found:
iov->nres = nres;
iov->ctrl = ctrl;
iov->total_VFs = total;
- iov->offset = offset;
- iov->stride = stride;
iov->pgsz = pgsz;
iov->self = dev;
pci_read_config_dword(dev, pos + PCI_SRIOV_CAP, &iov->cap);
@@ -475,6 +468,11 @@ found:
dev->sriov = iov;
dev->is_physfn = 1;
iov->max_VF_buses = virtfn_max_buses(dev);
+ pci_iov_set_numvfs(dev, 0);
+ if (!iov->offset || (total > 1 && !iov->stride)) {
+ rc = -EIO;
+ goto failed;
+ }
return 0;

You might want to reorder this a bit. The problem is offset and stride can be 0 if numvfs is 0. So you should probably test offset and stride first, and then reset numvfs to 0.

- Alex

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/