bhyve guests: what clock source to use?
Hi all, As part of debugging a problem I'm having, I'd like to understand: What clock source should bhyve guests use? tsc? hpet? acpi_pm? What are pros/cons of the choices? Does it depend on the guest OS? (for me FreeBSD and Ubuntu) Should it match the host? Thanks, Sean
[Bug 273557] Regression preventing bhyve from running inside a jail without IP after f74147e26999838e03a522bf59ea33bef470d356) breaks support for jailing bhyve with IPv4 and IPv6 disabled. Patch inclu
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=273557 Mark Johnston changed: What|Removed |Added CC||ma...@freebsd.org --- Comment #12 from Mark Johnston --- I think I agree with the proposed patch to use VMIO_SIOCSIFFLAGS. Even if IP is disabled on the jail, tap is an L2 device. The fact that we have to use an IP socket to configure the interface is bizarre; this overloading of sockets is also problematic for capsicum. OTOH, I think we can configure IFF_UP with any kind of socket, no? We could use a PF_ROUTE or PF_NETLINK socket instead. I don't think netlink itself buys us anything here, ifconfig still uses socket ioctls to set interface flags. One other reason to avoid VMIO_SIOCSIFFLAGS is that it's really just there for vmnet(4) compatibility, I believe. -- You are receiving this mail because: You are the assignee for the bug.
[Bug 273557] Regression preventing bhyve from running inside a jail without IP after f74147e26999838e03a522bf59ea33bef470d356) breaks support for jailing bhyve with IPv4 and IPv6 disabled. Patch inclu
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=273557 --- Comment #13 from cr...@rlwinm.de --- The patch changes the bhyve tap network backend code invoked to open devices with names starting with "tap" or "vmnet". Since the code is already exclusive to the if_tuntap driver in tap or vmnet mode I can't see any problem with relying on one *more* driver specific ioctl(). As far as I can tell there is no possible configuration where the existing code works as intended, but the patch wouldn't also work. Netlink sockets shouldn't offer any way for a bhyve process attached to a vnet enabled jail to even address network interface in its parent jail (normally the unjailed host) to bring it up (or down). Anything else would be a security problem. The device file descriptor is the natural API endpoint to bring an interface up. It is the capability to manipulate a tap device through the single opening allowed per tap device (at a time). Having to query the tap device for its interface name or index to reference the interface in a request to change interface state by name or index is also a race condition since the interface could've been destroyed and its name or index reused. Using the file descriptor precludes any race condition similar to funlinkat() vs unlinkat(). -- You are receiving this mail because: You are the assignee for the bug.
