There are two problems with qvm-run -a: 1) even with -q flag, it spits to stdout (actually, "xl create" does it), and this garbage is received by rpc client 2) even with -q flag, it steals input (actually, "qrexec ... wait for session") These two can be manually fixed (by passing /dev/null appropriately); hovewer, this is prone to disaster if qvm-run is enhanced/broken later. We could do if is_domain_running() ; then run qrexec client else qvm-run -a domain true </dev/null >/dev/null run qrexec client fi which looks safer; but is_domain_running() is a bit expensive even in "running" case - we need to xl_context.list_domains anyway. Gotta decide on one of these. |
||
|---|---|---|
| .. | ||
| .gitignore | ||
| buffer.c | ||
| buffer.h | ||
| exec.c | ||
| glue.h | ||
| Makefile | ||
| qrexec_agent.c | ||
| qrexec_client_vm.c | ||
| qrexec_client.c | ||
| qrexec_daemon.c | ||
| qrexec_policy | ||
| qrexec.h | ||
| qubes_rpc_multiplexer | ||
| txrx-vchan.c | ||
| unix_server.c | ||
| write_stdin.c | ||