PCI/MSI: Add "msi_bus" sysfs MSI/MSI-X control for endpoints

The "msi_bus" sysfs file for bridges sets a bus flag to allow or disallow
future driver requests for MSI or MSI-X.  Previously, the sysfs file
existed for endpoints but did nothing.

Add "msi_bus" support for endpoints, so an administrator can prevent the
use of MSI and MSI-X for individual devices.

Note that as for bridges, these changes only affect future driver requests
for MSI or MSI-X, so drivers may need to be reloaded.

Add documentation for the "msi_bus" sysfs file.

[bhelgaas: changelog, comments, add "subordinate", add endpoint printk,
rework bus_flags setting, make bus_flags printk unconditional]
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
This commit is contained in:
Yijing Wang 2014-09-23 13:27:24 +08:00 committed by Bjorn Helgaas
parent 48c3c38f00
commit 468ff15a3a
2 changed files with 29 additions and 20 deletions

View file

@ -65,6 +65,16 @@ Description:
force a rescan of all PCI buses in the system, and force a rescan of all PCI buses in the system, and
re-discover previously removed devices. re-discover previously removed devices.
What: /sys/bus/pci/devices/.../msi_bus
Date: September 2014
Contact: Linux PCI developers <linux-pci@vger.kernel.org>
Description:
Writing a zero value to this attribute disallows MSI and
MSI-X for any future drivers of the device. If the device
is a bridge, MSI and MSI-X will be disallowed for future
drivers of all child devices under the bridge. Drivers
must be reloaded for the new setting to take effect.
What: /sys/bus/pci/devices/.../msi_irqs/ What: /sys/bus/pci/devices/.../msi_irqs/
Date: September, 2011 Date: September, 2011
Contact: Neil Horman <nhorman@tuxdriver.com> Contact: Neil Horman <nhorman@tuxdriver.com>

View file

@ -250,46 +250,45 @@ static ssize_t msi_bus_show(struct device *dev, struct device_attribute *attr,
char *buf) char *buf)
{ {
struct pci_dev *pdev = to_pci_dev(dev); struct pci_dev *pdev = to_pci_dev(dev);
struct pci_bus *subordinate = pdev->subordinate;
if (!pdev->subordinate) return sprintf(buf, "%u\n", subordinate ?
return 0; !(subordinate->bus_flags & PCI_BUS_FLAGS_NO_MSI)
: !pdev->no_msi);
return sprintf(buf, "%u\n",
!(pdev->subordinate->bus_flags & PCI_BUS_FLAGS_NO_MSI));
} }
static ssize_t msi_bus_store(struct device *dev, struct device_attribute *attr, static ssize_t msi_bus_store(struct device *dev, struct device_attribute *attr,
const char *buf, size_t count) const char *buf, size_t count)
{ {
struct pci_dev *pdev = to_pci_dev(dev); struct pci_dev *pdev = to_pci_dev(dev);
struct pci_bus *subordinate = pdev->subordinate;
unsigned long val; unsigned long val;
if (kstrtoul(buf, 0, &val) < 0) if (kstrtoul(buf, 0, &val) < 0)
return -EINVAL; return -EINVAL;
/*
* Bad things may happen if the no_msi flag is changed
* while drivers are loaded.
*/
if (!capable(CAP_SYS_ADMIN)) if (!capable(CAP_SYS_ADMIN))
return -EPERM; return -EPERM;
/* /*
* Maybe devices without subordinate buses shouldn't have this * "no_msi" and "bus_flags" only affect what happens when a driver
* attribute in the first place? * requests MSI or MSI-X. They don't affect any drivers that have
* already requested MSI or MSI-X.
*/ */
if (!pdev->subordinate) if (!subordinate) {
pdev->no_msi = !val;
dev_info(&pdev->dev, "MSI/MSI-X %s for future drivers\n",
val ? "allowed" : "disallowed");
return count; return count;
/* Is the flag going to change, or keep the value it already had? */
if (!(pdev->subordinate->bus_flags & PCI_BUS_FLAGS_NO_MSI) ^
!!val) {
pdev->subordinate->bus_flags ^= PCI_BUS_FLAGS_NO_MSI;
dev_warn(&pdev->dev, "forced subordinate bus to%s support MSI, bad things could happen\n",
val ? "" : " not");
} }
if (val)
subordinate->bus_flags &= ~PCI_BUS_FLAGS_NO_MSI;
else
subordinate->bus_flags |= PCI_BUS_FLAGS_NO_MSI;
dev_info(&subordinate->dev, "MSI/MSI-X %s for future drivers of devices on this bus\n",
val ? "allowed" : "disallowed");
return count; return count;
} }
static DEVICE_ATTR_RW(msi_bus); static DEVICE_ATTR_RW(msi_bus);