Re: [PATCH 4/4] softmmu: move parsing of -runas, -chroot and -daemonize code

[Daniel P . Berrangé]

On Fri, Mar 04, 2022 at 11:56:57AM +0000, Daniel P. Berrangé wrote:
> With the future intent to try to move to a fully QAPI driven
> configuration system, we want to have any current command
> parsing well isolated from logic that applies the resulting
> configuration.
> 
> We also don't want os-posix.c to contain code that is specific
> to the system emulators, as this file is linked to other binaries
> too.
> 
> To satisfy these goals, we move parsing of the -runas, -chroot and
> -daemonize code out of the os-posix.c helper code, and pass the
> parsed data into APIs in os-posix.c.
> 
> As a side benefit the 'os_daemonize()' function now lives upto to
> its name and will always daemonize, instead of using global state
> to decide to be a no-op sometimes.
> 
> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
> ---
>  include/sysemu/os-posix.h |   4 +-
>  include/sysemu/os-win32.h |   1 -
>  os-posix.c                | 148 +++++++++++---------------------------
>  os-win32.c                |   9 ---
>  softmmu/vl.c              |  76 ++++++++++++++++++--
>  5 files changed, 113 insertions(+), 125 deletions(-)


> diff --git a/softmmu/vl.c b/softmmu/vl.c
> index 1fe028800f..cdf27b6027 100644
> --- a/softmmu/vl.c
> +++ b/softmmu/vl.c
> @@ -2602,11 +2602,13 @@ static void qemu_process_help_options(void)
>      }
>  }
>  
> -static void qemu_maybe_daemonize(const char *pid_file)
> +static void qemu_maybe_daemonize(bool daemonize, const char *pid_file)
>  {
>      Error *err = NULL;
>  
> -    os_daemonize();
> +    if (daemonize) {
> +        os_daemonize();
> +    }
>      rcu_disable_atfork();
>  
>      if (pid_file && !qemu_write_pidfile(pid_file, &err)) {


> @@ -3683,7 +3743,7 @@ void qemu_init(int argc, char **argv, char **envp)
>      qemu_process_early_options();
>  
>      qemu_process_help_options();
> -    qemu_maybe_daemonize(pid_file);
> +    qemu_maybe_daemonize(daemonize, pid_file);

This commit is a bit flawed, because we're until we call the
os_daemonize() method, the is_daemonized() method won't return
true. Unfortunately some callers rely is_daemonized() returning
true merely for the request, even though we've not yet put it
into action. ie the method would have been better called
is_daemonize_requested()

The upshot is that we fail to properly close stderr.

I'll send a v2 that handles this by fully removing the
is_daemonize() method.


Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|