qemu-kvm/SOURCES/0013-vfio-cap-number-of-devices-that-can-be-assigned.patch

115 lines
4.0 KiB
Diff
Raw Normal View History

From de433da59448eaad4ac1b902d07d57b57f922aff Mon Sep 17 00:00:00 2001
From: Bandan Das <bsd@redhat.com>
Date: Tue, 3 Dec 2013 20:05:13 +0100
Subject: vfio: cap number of devices that can be assigned
RH-Author: Bandan Das <bsd@redhat.com>
Message-id: <1386101113-31560-3-git-send-email-bsd@redhat.com>
Patchwork-id: 55984
O-Subject: [PATCH RHEL7 qemu-kvm v2 2/2] vfio: cap number of devices that can be assigned
Bugzilla: 678368
RH-Acked-by: Alex Williamson <alex.williamson@redhat.com>
RH-Acked-by: Marcelo Tosatti <mtosatti@redhat.com>
RH-Acked-by: Michael S. Tsirkin <mst@redhat.com>
Go through all groups to get count of total number of devices
active to enforce limit
Reasoning from Alex for the limit(32) - Assuming 3 slots per
device, with 125 slots (number of memory slots for RHEL 7),
we can support almost 40 devices and still have few slots left
for other uses. Stepping down a bit, the number 32 arbitrarily
matches the number of slots on a PCI bus and is also a nice power
of two.
Signed-off-by: Bandan Das <bsd@redhat.com>
Rebase notes (2.8.0):
- removed return value for vfio_realize (commit 1a22aca)
Merged patches (2.9.0):
- 17eb774 vfio: Use error_setg when reporting max assigned device overshoot
Merged patches (4.1.0-rc3):
- 2b89558 vfio: increase the cap on number of assigned devices to 64
(cherry picked from commit 9fa3c9fc6dfcde76d80db1aa601b2d577f72ceec)
(cherry picked from commit 3cb35556dc7d994f203d732fe952f95fcdb03c0a)
Signed-off-by: Danilo C. L. de Paula <ddepaula@redhat.com>
---
hw/vfio/pci.c | 29 ++++++++++++++++++++++++++++-
hw/vfio/pci.h | 1 +
2 files changed, 29 insertions(+), 1 deletion(-)
diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
index c8534d3035..309535f306 100644
--- a/hw/vfio/pci.c
+++ b/hw/vfio/pci.c
@@ -47,6 +47,9 @@
#define TYPE_VFIO_PCI_NOHOTPLUG "vfio-pci-nohotplug"
+/* RHEL only: Set once for the first assigned dev */
+static uint16_t device_limit;
+
static void vfio_disable_interrupts(VFIOPCIDevice *vdev);
static void vfio_mmap_set_enabled(VFIOPCIDevice *vdev, bool enabled);
@@ -2722,9 +2725,30 @@ static void vfio_realize(PCIDevice *pdev, Error **errp)
ssize_t len;
struct stat st;
int groupid;
- int i, ret;
+ int ret, i = 0;
bool is_mdev;
+ if (device_limit && device_limit != vdev->assigned_device_limit) {
+ error_setg(errp, "Assigned device limit has been redefined. "
+ "Old:%d, New:%d",
+ device_limit, vdev->assigned_device_limit);
+ return;
+ } else {
+ device_limit = vdev->assigned_device_limit;
+ }
+
+ QLIST_FOREACH(group, &vfio_group_list, next) {
+ QLIST_FOREACH(vbasedev_iter, &group->device_list, next) {
+ i++;
+ }
+ }
+
+ if (i >= vdev->assigned_device_limit) {
+ error_setg(errp, "Maximum supported vfio devices (%d) "
+ "already attached", vdev->assigned_device_limit);
+ return;
+ }
+
if (!vdev->vbasedev.sysfsdev) {
if (!(~vdev->host.domain || ~vdev->host.bus ||
~vdev->host.slot || ~vdev->host.function)) {
@@ -3167,6 +3191,9 @@ static Property vfio_pci_dev_properties[] = {
DEFINE_PROP_BOOL("x-no-kvm-msix", VFIOPCIDevice, no_kvm_msix, false),
DEFINE_PROP_BOOL("x-no-geforce-quirks", VFIOPCIDevice,
no_geforce_quirks, false),
+ /* RHEL only */
+ DEFINE_PROP_UINT16("x-assigned-device-limit", VFIOPCIDevice,
+ assigned_device_limit, 64),
DEFINE_PROP_BOOL("x-no-kvm-ioeventfd", VFIOPCIDevice, no_kvm_ioeventfd,
false),
DEFINE_PROP_BOOL("x-no-vfio-ioeventfd", VFIOPCIDevice, no_vfio_ioeventfd,
diff --git a/hw/vfio/pci.h b/hw/vfio/pci.h
index 35626cd63e..0cd4803aee 100644
--- a/hw/vfio/pci.h
+++ b/hw/vfio/pci.h
@@ -135,6 +135,7 @@ typedef struct VFIOPCIDevice {
EventNotifier err_notifier;
EventNotifier req_notifier;
int (*resetfn)(struct VFIOPCIDevice *);
+ uint16_t assigned_device_limit;
uint32_t vendor_id;
uint32_t device_id;
uint32_t sub_vendor_id;
--
2.21.0