Re: [libvirt PATCH] qemu: allow passt to self-daemonize

[Stefano Brivio <sbrivio@redhat.com>]

On Thu, 9 Feb 2023 11:10:21 +0100
Michal Prívozník <mprivozn@redhat.com> wrote:

> On 2/9/23 10:56, Daniel P. Berrangé wrote:
> > On Thu, Feb 09, 2023 at 09:52:00AM +0100, Michal Prívozník wrote:  
> >> On 2/9/23 00:13, Laine Stump wrote:  
> >>> I initially had the passt process being started in an identical
> >>> fashion to the slirp-helper - libvirt was daemonizing the new process
> >>> and recording its pid in a pidfile. The problem with this is that,
> >>> since it is daemonized immediately, any startup error in passt happens
> >>> after the daemonization, and thus isn't seen by libvirt - libvirt
> >>> believes that the process has started successfully and continues on
> >>> its merry way. The result was that sometimes a guest would be started,
> >>> but there would be no passt process for qemu to use for network
> >>> traffic.
> >>>
> >>> Instead, we should be starting passt in the same manner we start
> >>> dnsmasq - we just exec it as normal (along with a request that passt
> >>> create the pidfile, which is just another option on the passt
> >>> commandline) and wait for the child process to exit; passt then has a
> >>> chance to parse its commandline and complete all the setup prior to
> >>> daemonizing itself; if it encounters an error and exits with a non-0
> >>> code, libvirt will see the code and know about the failure. We can
> >>> then grab the output from stderr, log that so the "user" has some idea
> >>> of what went wrong, and then fail the guest startup.
> >>>
> >>> Signed-off-by: Laine Stump <laine@redhat.com>
> >>> ---
> >>>  src/qemu/qemu_passt.c | 9 ++++-----
> >>>  1 file changed, 4 insertions(+), 5 deletions(-)
> >>>
> >>> diff --git a/src/qemu/qemu_passt.c b/src/qemu/qemu_passt.c
> >>> index 0f09bf3db8..f640a69c00 100644
> >>> --- a/src/qemu/qemu_passt.c
> >>> +++ b/src/qemu/qemu_passt.c
> >>> @@ -141,24 +141,23 @@ qemuPasstStart(virDomainObj *vm,
> >>>      g_autofree char *passtSocketName = qemuPasstCreateSocketPath(vm, net);
> >>>      g_autoptr(virCommand) cmd = NULL;
> >>>      g_autofree char *pidfile = qemuPasstCreatePidFilename(vm, net);
> >>> +    g_autofree char *errbuf = NULL;
> >>>      char macaddr[VIR_MAC_STRING_BUFLEN];
> >>>      size_t i;
> >>>      pid_t pid = (pid_t) -1;
> >>>      int exitstatus = 0;
> >>>      int cmdret = 0;
> >>> -    VIR_AUTOCLOSE errfd = -1;
> >>>  
> >>>      cmd = virCommandNew(PASST);
> >>>  
> >>>      virCommandClearCaps(cmd);
> >>> -    virCommandSetPidFile(cmd, pidfile);
> >>> -    virCommandSetErrorFD(cmd, &errfd);
> >>> -    virCommandDaemonize(cmd);
> >>> +    virCommandSetErrorBuffer(cmd, &errbuf);
> >>>  
> >>>      virCommandAddArgList(cmd,
> >>>                           "--one-off",
> >>>                           "--socket", passtSocketName,
> >>>                           "--mac-addr", virMacAddrFormat(&net->mac, macaddr),
> >>> +                         "--pid", pidfile,  
> >>
> >> The only problem with this approach is that our virPidFile*() functions
> >> rely on locking the very first byte. And when reading the pidfile, we
> >> try to lock the file and if we succeeded it means the file wasn't locked
> >> which means the process holding the lock died and thus the pid in the
> >> pidfile is stale.
> >>
> >> Now, I don't see passt locking the pidfile at all. So effectively, after
> >> this patch qemuPasstStop() would do nothing (well, okay, it'll remove
> >> the pidfile), qemuPasstSetupCgroup() does nothing, etc.
> >>
> >> What we usually do in this case, is: we let our code write the pidfile
> >> (just like the current code does), but then have a loop that waits a bit
> >> for socket to show up. If it doesn't in say 5 seconds we kill the child
> >> process (which we know the PID of). You can take inspiration from:
> >> qemuDBusStart() or qemuProcessStartManagedPRDaemon().  
> > 
> > Busy waiting for sockets is nasty though. Depending on how passt is
> > written it might not be needed. If passt creates the listen()
> > socket and does all the important initialization steps that are liable
> > to fail, *before* it daemonizes, then we can synchronize without busy
> > waiting.

It does. In my opinion it could simply be handled like it's done for
dnsmasq -- from networkStartDhcpDaemon():

    if (virCommandRun(cmd, NULL) < 0)
        return -1;

    /*
     * There really is no race here - when dnsmasq daemonizes, its
     * leader process stays around until its child has actually
     * written its pidfile. So by time virCommandRun exits it has
     * waitpid'd and guaranteed the proess has started and written a
     * pid
     */

> > ie waitpid() for passt leader process to exit. Then check if
> > the socket exists. If it does, then passt has daemonized and is listening
> > and running, if it does not, then passt failed.  
> 
> That still requires passt to hold the pidfile open and locked, neither
> of which is happening with the current code.

...is this still a requirement even if qemuPasstStop() just needs to
remove the PID file?

-- 
Stefano