Re: [PATCH 1/2] PCI: Add support for VF Resizable Bar extended cap

From: Michał Winiarski
Date: Fri Dec 17 2021 - 07:38:22 EST


On 16.12.2021 01:21, Krzysztof Wilczyński wrote:
Hi Michał,

[...]
+static int pci_memory_decoding(struct pci_dev *dev)
+{
+ u16 cmd;
+
+ pci_read_config_word(dev, PCI_COMMAND, &cmd);
+ if (cmd & PCI_COMMAND_MEMORY)
+ return -EBUSY;
+
+ return 0;
+}
+
+#ifdef CONFIG_PCI_IOV
+static int pci_vf_memory_decoding(struct pci_dev *dev)
+{
+ u16 cmd;
+
+ pci_read_config_word(dev, dev->sriov->pos + PCI_SRIOV_CTRL, &cmd);
+ if (cmd & PCI_SRIOV_CTRL_MSE)
+ return -EBUSY;
+
+ return 0;
+}
+#endif
+
int pci_resize_resource(struct pci_dev *dev, int resno, int size)
{
struct resource *res = dev->resource + resno;
struct pci_host_bridge *host;
int old, ret;
u32 sizes;
- u16 cmd;
/* Check if we must preserve the firmware's resource assignment */
host = pci_find_host_bridge(dev->bus);
@@ -424,9 +447,14 @@ int pci_resize_resource(struct pci_dev *dev, int resno, int size)
if (!(res->flags & IORESOURCE_UNSET))
return -EBUSY;
- pci_read_config_word(dev, PCI_COMMAND, &cmd);
- if (cmd & PCI_COMMAND_MEMORY)
- return -EBUSY;
+#ifdef CONFIG_PCI_IOV
+ if (resno >= PCI_IOV_RESOURCES)
+ ret = pci_vf_memory_decoding(dev);
+ else
+#endif
+ ret = pci_memory_decoding(dev);
+ if (ret)
+ return ret;

The above else works, however, it does trip our static analysis tools, and
lack of the indentation makes it slightly confusing to read, at least at
the first glance. Thus, I wonder whether it would be possible to combine
the pci_vf_memory_decoding() and pci_memory_decoding() somehow neatly into
a private helper that takes a boolean to indicate whether it's a VF or not.
Then, we could drop the if-statement, since some of the logic could live
within the helper.

What do you think?

Sure - implementing a helper in a way that Christian suggested (the helper still takes resno though) should also take care of that.

Thanks
-Michał


Krzysztof