Userfaultfd is by default allowed only for privileged processes.
Since
libvirt runs QEMU unprivileged, we need to enable unprivileged access to
userfaultfd before starting post-copy migration.
Rather than providing a static sysctl configuration file, we set the
sysctl knob in runtime once post-copy migration is requested. This way
unprivileged_userfaultfd is only enabled once actually used.
https://bugzilla.redhat.com/show_bug.cgi?id=1945420
Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com>
---
src/qemu/qemu_migration_params.c | 25 +++++++++++++++++++++++++
1 file changed, 25 insertions(+)
diff --git a/src/qemu/qemu_migration_params.c b/src/qemu/qemu_migration_params.c
index dbc3219826..a9449ed1ff 100644
--- a/src/qemu/qemu_migration_params.c
+++ b/src/qemu/qemu_migration_params.c
@@ -804,6 +804,24 @@ qemuMigrationCapsToJSON(virBitmap *caps,
}
+static void
+qemuMigrationParamsEnableUserfaultfd(virDomainObj *vm)
+{
+ const char *sysctl = "/proc/sys/vm/unprivileged_userfaultfd";
+
+ if (!virFileExists(sysctl))
+ return;
+
+ VIR_DEBUG("Enabling unprivileged userfaultfd for post-copy migration of "
+ "domain %s", vm->def->name);
+
+ if (virFileWriteStr(sysctl, "1", 0) < 0) {
+ virReportSystemError(errno, "%s",
+ _("failed to enable unprivileged userfaultfd"));
+ }
I can imagine a wild scenario where libvirtd is run inside a restricted
container but with the proper value set. How about we check whether the
expected value isn't set already?
No need to resend, the change is trivial enough to be done before push.
Reviewed-by: Michal Privoznik <mprivozn(a)redhat.com>
Michal