[libvirt PATCH v2 6/7] util: make virPCIDeviceIsPCIExpress() more intelligent

[Laine Stump]

Until now there has been an extra bit of code in
qemuDomainDeviceCalculatePCIConnectFlag() (one of the two callers of
virPCIDeviceIsPCIExpress()) that tries to determine if a device is
PCIe by looking at the *length* of its sysfs config file; it only does
this when libvirt is running as a non-root process.

This patch takes advantage of our newfound ability to tell the
difference between "I read a 0 from the device PCI config file" and "I
couldn't read the PCI Express Capabilities because I don't have
sufficient permission" to put the file length check down in
virPCIDeviceIsPCIExpress(), and do that check any time we fail while
reading the config file (not only when the process is non-root).

Fixes: https://bugzilla.redhat.com/1901685
Signed-off-by: Laine Stump <laine at redhat.com>
---
 src/util/virpci.c | 66 ++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 62 insertions(+), 4 deletions(-)

diff --git a/src/util/virpci.c b/src/util/virpci.c
index 5ec559dec0..9bfc743fbd 100644
--- a/src/util/virpci.c
+++ b/src/util/virpci.c
@@ -73,10 +73,18 @@ struct _virPCIDevice {
     char          *used_by_drvname;
     char          *used_by_domname;
 
+    /* The following 5 items are only valid after virPCIDeviceInit()
+     * has been called for the virPCIDevice object. This is *not* done
+     * in most cases (because it creates extra overhead, and parts of
+     * it can fail if libvirtd is running unprivileged)
+     */
     unsigned int  pcie_cap_pos;
     unsigned int  pci_pm_cap_pos;
     bool          has_flr;
     bool          has_pm_reset;
+    bool          is_pcie;
+    /**/
+
     bool          managed;
 
     virPCIStubDriver stubDriver;
@@ -629,7 +637,8 @@ virPCIDeviceDetectFunctionLevelReset(virPCIDevicePtr dev, int cfgfd)
      * the same thing, except for conventional PCI
      * devices. This is not common yet.
      */
-    virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_AF, &pos);
+    if (virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_AF, &pos) < 0)
+        goto error;
 
     if (pos) {
         caps = virPCIDeviceRead16(dev, cfgfd, pos + PCI_AF_CAP);
@@ -654,8 +663,8 @@ virPCIDeviceDetectFunctionLevelReset(virPCIDevicePtr dev, int cfgfd)
         return true;
     }
 
+ error:
     VIR_DEBUG("%s %s: no FLR capability found", dev->id, dev->name);
-
     return false;
 }
 
@@ -937,10 +946,59 @@ virPCIDeviceTryPowerManagementReset(virPCIDevicePtr dev, int cfgfd)
     return 0;
 }
 
+/**
+ * virPCIDeviceInit:
+ * @dev: virPCIDevice object needing its PCI capabilities info initialized
+ * @cfgfd: open file descriptor for device config file in sysfs
+ *
+ * Initialize the PCI capabilities attributes of a virPCIDevice object
+ * (i.e. pcie_cap_pos, pci_pm_cap_pos, has_flr, has_pm_reset, and
+ * is_pcie). This is done by walking the info in the (already-opened)
+ * device PCI config file in sysfs. This function can be called
+ * regardless of whether a process has sufficient privilege to read
+ * the entire file (unprivileged processes can only read the 1st 64
+ * bytes, while the Express Capabilities are all located beyond that
+ * boundary).
+ *
+ * In the case that we are unable to read a capability
+ * directly, we will attempt to infer its value by other means. In
+ * particular, we can determine that a device is (almost surely) PCIe
+ * by checking that the length of the config file is != 256 (since all
+ * conventional PCI config files are 256 bytes), and we know that any
+ * device that is an SR-IOV VF will have FLR available (since that is
+ * required by the SR-IOV spec.)
+ *
+ * Always returns success (0) (for now)
+ */
 static int
 virPCIDeviceInit(virPCIDevicePtr dev, int cfgfd)
 {
-    virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_EXP, &dev->pcie_cap_pos);
+    dev->is_pcie = false;
+    if (virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_EXP, &dev->pcie_cap_pos) < 0) {
+        /* an unprivileged process is unable to read *all* of a
+         * device's PCI config (it can only read the first 64
+         * bytes, which isn't enough for see the Express
+         * Capabilities data). If virPCIDeviceFindCapabilityOffset
+         * returns failure (and not just a pcie_cap_pos == 0,
+         * which is *success* at determining the device is *not*
+         * PCIe) we make an educated guess based on the length of
+         * the device's config file - if it is 256 bytes, then it
+         * is definitely a legacy PCI device. If it's larger than
+         * that, then it is *probably PCIe (although it could be
+         * PCI-x, but those are extremely rare). If the config
+         * file can't be found (in which case the "length" will be
+         * -1), then we blindly assume the most likely outcome -
+         * PCIe.
+         */
+        off_t configLen = virFileLength(virPCIDeviceGetConfigPath(dev), -1);
+
+        if (configLen != 256)
+            dev->is_pcie = true;
+
+    } else {
+        dev->is_pcie = (dev->pcie_cap_pos != 0);
+    }
+
     virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_PM, &dev->pci_pm_cap_pos);
     dev->has_flr = virPCIDeviceDetectFunctionLevelReset(dev, cfgfd);
     dev->has_pm_reset = virPCIDeviceDetectPowerManagementReset(dev, cfgfd);
@@ -2644,7 +2702,7 @@ virPCIDeviceIsPCIExpress(virPCIDevicePtr dev)
     if (virPCIDeviceInit(dev, fd) < 0)
         goto cleanup;
 
-    ret = dev->pcie_cap_pos != 0;
+    ret = dev->is_pcie;
 
  cleanup:
     virPCIDeviceConfigClose(dev, fd);
-- 
2.28.0