qemu_domain: Increase memlock limit for NVMe disks

When starting QEMU, or when hotplugging a PCI device QEMU might
lock some memory. How much? Well, that's an undecidable problem.

But despite that, we try to guess. And it more or less works,
until there's a counter example. This time, it's a guest with
both <hostdev/> and an NVMe <disk/>. I've started a simple guest
with 4GiB of memory:

  # virsh dominfo fedora
  Max memory:     4194304 KiB
  Used memory:    4194304 KiB

And here are the amounts of memory that QEMU tried to lock,
obtained via:

  grep VmLck /proc/$(pgrep qemu-kvm)/status

  1) with just one <hostdev/>
     VmLck:   4194308 kB

  2) with just one NVMe <disk/>
     VmLck:   4328544 kB

  3) with one <hostdev/> and one NVMe <disk/>
     VmLck:   8522852 kB

Now, what's surprising is case 2) where the locked memory exceeds
the VM memory. It almost resembles VDPA. Therefore, treat is as
such.

Unfortunately, I don't have a box with two or more spare NVMe-s
so I can't tell for sure. But setting limit too tight means QEMU
refuses to start.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2014030
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>

diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
index 63b13b6875c9f1530e264654e2cbb09c9fbb3d52..41db98880cd9ba096b010ed664c2e11ddb0475c3 100644 (file)
--- a/src/qemu/qemu_domain.c
+++ b/src/qemu/qemu_domain.c
@@ -9532,7 +9532,7 @@ getPPC64MemLockLimitBytes(virDomainDef *def,
 
 
 static int
-qemuDomainGetNumVFIODevices(const virDomainDef *def)
+qemuDomainGetNumVFIOHostdevs(const virDomainDef *def)
 {
     size_t i;
     int n = 0;
@@ -9542,10 +9542,22 @@ qemuDomainGetNumVFIODevices(const virDomainDef *def)
             virHostdevIsMdevDevice(def->hostdevs[i]))
             n++;
     }
+
+    return n;
+}
+
+
+static int
+qemuDomainGetNumNVMeDisks(const virDomainDef *def)
+{
+    size_t i;
+    int n = 0;
+
     for (i = 0; i < def->ndisks; i++) {
         if (virStorageSourceChainHasNVMe(def->disks[i]->src))
             n++;
     }
+
     return n;
 }
 
@@ -9585,6 +9597,7 @@ qemuDomainGetMemLockLimitBytes(virDomainDef *def,
 {
     unsigned long long memKB = 0;
     int nvfio;
+    int nnvme;
     int nvdpa;
 
     /* prefer the hard limit */
@@ -9604,7 +9617,8 @@ qemuDomainGetMemLockLimitBytes(virDomainDef *def,
     if (ARCH_IS_PPC64(def->os.arch) && def->virtType == VIR_DOMAIN_VIRT_KVM)
         return getPPC64MemLockLimitBytes(def, forceVFIO);
 
-    nvfio = qemuDomainGetNumVFIODevices(def);
+    nvfio = qemuDomainGetNumVFIOHostdevs(def);
+    nnvme = qemuDomainGetNumNVMeDisks(def);
     nvdpa = qemuDomainGetNumVDPANetDevices(def);
     /* For device passthrough using VFIO the guest memory and MMIO memory
      * regions need to be locked persistent in order to allow DMA.
@@ -9624,16 +9638,17 @@ qemuDomainGetMemLockLimitBytes(virDomainDef *def,
      *
      * Note that this may not be valid for all platforms.
      */
-    if (forceVFIO || nvfio || nvdpa) {
+    if (forceVFIO || nvfio || nnvme || nvdpa) {
         /* At present, the full memory needs to be locked for each VFIO / VDPA
-         * device. For VFIO devices, this only applies when there is a vIOMMU
-         * present. Yes, this may result in a memory limit that is greater than
-         * the host physical memory, which is not ideal. The long-term solution
-         * is a new userspace iommu interface (iommufd) which should eliminate
-         * this duplicate memory accounting. But for now this is the only way
-         * to enable configurations with e.g. multiple vdpa devices.
+         * NVMe device. For VFIO devices, this only applies when there is a
+         * vIOMMU present. Yes, this may result in a memory limit that is
+         * greater than the host physical memory, which is not ideal. The
+         * long-term solution is a new userspace iommu interface (iommufd)
+         * which should eliminate this duplicate memory accounting. But for now
+         * this is the only way to enable configurations with e.g. multiple
+         * VDPA/NVMe devices.
          */
-        int factor = nvdpa;
+        int factor = nvdpa + nnvme;
 
         if (nvfio || forceVFIO) {
             if (nvfio && def->iommu)

diff --git a/tests/qemumemlockdata/qemumemlock-pc-hostdev-nvme.xml b/tests/qemumemlockdata/qemumemlock-pc-hostdev-nvme.xml
new file mode 100644 (file)
index 0000000..06f1496
--- /dev/null
+++ b/tests/qemumemlockdata/qemumemlock-pc-hostdev-nvme.xml
@@ -0,0 +1,24 @@
+<domain type='kvm'>
+  <name>guest</name>
+  <memory unit='KiB'>1048576</memory>
+  <vcpu placement='static'>1</vcpu>
+  <os>
+    <type arch='x86_64' machine='pc'>hvm</type>
+  </os>
+  <devices>
+    <emulator>/usr/bin/qemu-system-x86_64</emulator>
+    <disk type='nvme' device='disk'>
+      <driver name='qemu' type='raw'/>
+      <source type='pci' managed='yes' namespace='1'>
+        <address domain='0x0003' bus='0x02' slot='0x00' function='0x0'/>
+      </source>
+      <target dev='vda' bus='virtio'/>
+    </disk>
+    <hostdev mode='subsystem' type='pci' managed='yes'>
+      <driver name='vfio'/>
+      <source>
+        <address domain='0x0001' bus='0x01' slot='0x00' function='0x0'/>
+      </source>
+    </hostdev>
+  </devices>
+</domain>

diff --git a/tests/qemumemlocktest.c b/tests/qemumemlocktest.c
index 61b73e1d793364e3755766d0d5253be86902eaac..c53905a7dd666117ac46a09a9a8126688a93ab49 100644 (file)
--- a/tests/qemumemlocktest.c
+++ b/tests/qemumemlocktest.c
@@ -97,6 +97,7 @@ mymain(void)
     DO_TEST("pc-hardlimit", 2147483648);
     DO_TEST("pc-locked", VIR_DOMAIN_MEMORY_PARAM_UNLIMITED);
     DO_TEST("pc-hostdev", 2147483648);
+    DO_TEST("pc-hostdev-nvme", 3221225472);
 
     DO_TEST("pc-hardlimit+locked", 2147483648);
     DO_TEST("pc-hardlimit+hostdev", 2147483648);