Bug 119039

Created attachment 98817 [details]
adds initreq.h to the RPM

Of course, header files belong in -devel packages, but it seems hardly worth it
for one very small header file. If the maintainer feels it's worth it, I guess
technically it should be split up.

Why do the volume/brightness buttons need to signal *init*?

int
The daemon has two functions:
- those buttons
- power management

init is used in the latter function. More specifically, here's where
it's used to send SIGPWR to init:

send_sigpwr (char mode)
{
  struct init_request initreq;
  FILE *fd;
  int err = -1;
                                                                     
          
  initreq.magic     = INIT_MAGIC;
  initreq.cmd       = mode == 'F' ? INIT_CMD_POWERFAIL : INIT_CMD_POWEROK;
  initreq.sleeptime = 5;
                                                                     
          
  if ((fd = fopen(INIT_FIFO, "r+")) != NULL) {
    err = 0;
    if ((fwrite(&initreq, sizeof(initreq), 1, fd)) != 1)
      err = -1;
    fclose(fd);
  }
  if (err == -1) {
    if ((fd = fopen("/etc/powerstatus", "w")) != NULL) {
      err = 0;
      if ((fwrite(&mode, 1, 1, fd)) != 1)
        err = -1;
      fclose(fd);
    }
    err = kill(1, SIGPWR);  /* send SIGPWR to the INIT process */
  }
  return err;
}

Hm, it would be simpler if those were gated through the same mechanism
as the ACPI buttons.

I forwarded on this bug report to the author of pbbuttonsd. Here is
his response:

---------------------------------------------------------------------
pbbuttonsd is a power management daemon for apple powerpc machines 
which don't support ACPI. 
 
PBButtonsd can be configured to tell init about the low battery 
condition. Init has appropriate mechanisms to handle such emergency 
conditions. Usually an application writes the file /etc/powerstatus 
and send the signal SIGPWR to the init process but the man page of 
init (8) depreciated this method: 
--- quote --- 
Usage of SIGPWR and /etc/powerstatus is discouraged. Someone 
wanting to interact with init should use the /dev/initctl 
control channel - see the source code of the sysvinit package 
for more documentation about this. 
--- quote off --- 
You see installing the include file initreq.h is a step forward. 
 
 Best Regards 
   Matthias Grimm 
---------------------------------------------------------------------

Also from the init man page:

[quote]
INTERFACE
Init  listens  on  a fifo in /dev, /dev/initctl, for messages. 
Telinit uses this to communicate with init. The interface is not very
well documented or finished. Those interested should study the
initreq.h file in the src/ subdirectory of the init source code tar
archive.
[/quote]

So, to use the /dev/initctl interface, you need the header that patch
provides. From initreq.h:

#define INIT_FIFO "/dev/initctl"

and from the code listing above:

if ((fd = fopen(INIT_FIFO, "r+")) != NULL)

So, this looks like the recommended way, from reading the man page.

Yes, it's just that there are also acpi-using buttons for doing
soft-power off. In the Grand Scheme, it would be nice to have them all
going through the same interface. :)

That may well be so (or not, my knowledge is limited in this area).

However, either way, the initreq.h header should be included in any
case, because it's inconsistent with the init(8) man page not to
include it. The man page says "we recommend you do this, see this
header file" (ie. use what's in there, implying #include it). Without
the header file included, there is no way *any* application could
follow the recommended practice from the man page.

Whether or not pbbuttonsd should be using this header file is another
discussion, I guess.

Will be fixed in 2.86-1.

Package name is now sysvinit in Fedora.