When VFIO passes through a PCI device to a guest, it does not allow
the guest to mmap BARs that are smaller than PAGE_SIZE unless it
can reserve the rest of the page (see vfio_pci_probe_mmaps()). This
is because a page might contain several small BARs for unrelated
devices and a guest should not be able to access all of them.
VFIO emulates guest accesses to non-mappable BARs, which is functional
but slow. On systems with large page sizes, e.g., PowerNV with 64K pages,
BARs are more likely to share a page and performance is more likely to
be a problem.
Add a weak function to set default alignment for all PCI devices.
An arch can override it to force the PCI core to place memory BARs on
their own pages.
Signed-off-by: Yongji Xie <elohimes@xxxxxxxxx>
---
drivers/pci/pci.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 7904d02..02f1255 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -4947,6 +4947,11 @@ void pci_ignore_hotplug(struct pci_dev *dev)
}
EXPORT_SYMBOL_GPL(pci_ignore_hotplug);
+resource_size_t __weak pcibios_default_alignment(struct pci_dev *dev)
+{
+ return 0;
+}
+
#define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
static DEFINE_SPINLOCK(resource_alignment_lock);
@@ -4962,14 +4967,15 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev)
{
int seg, bus, slot, func, align_order, count;
unsigned short vendor, device, subsystem_vendor, subsystem_device;
- resource_size_t align = 0;
+ resource_size_t align = pcibios_default_alignment(dev);
char *p;
spin_lock(&resource_alignment_lock);
p = resource_alignment_param;
- if (!*p)
+ if (!*p && !align)
goto out;
if (pci_has_flag(PCI_PROBE_ONLY)) {
+ align = 0;
pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
goto out;
}
--
1.7.9.5
