Add a new call to process spank options from a plugin.

The spank_option_getopt() function will search the current
spank environment for use of the option passed as an argument.
The current option cache, and the local environment are checked
for the use of the given spank option. This call is an alternative
to use of a global variable in combination with the option callback,
and is also needed for processing options in the isolated contexts
of slurm_spank_job_prolog() and slurm_spank_job_epilog().

Add spank_clear_remote_options_env() to clear any spank options
passed through the environment after they are no longer needed.
This is done in slurmd after running the spank job prolog || epilog,
as well as in the spank_post_opt function, after the env has been
searched for spank variables.

Always set spank options in the environemnt and spank job environment
to ensure that used options are propagated to the job prolog and
epilog.  (Previously, spank options were set in the environment
only in allocator context)

In slurmd and job prolog/epilog contexts, avoid loading plugins that
have no callbacks in the context in which they are loaded. That is
for slurmd, if there are no slurm_spank_slurmd_init or
slurm_spank_slurmd_exit callbacks, there is no reason to keep the
current plugin loaded.

We now want to return error on failure of either spank prolog/epilog
or regular prolog/epilog scripts, so add a common function _run_job_script
to handle return of shared error code.

For now, we continue to run the normal prolog or epilog even if the spank
prolog/epilog fail. In the future, a failure the spank prolog/epilog may
short-circuit the run of the normal scripts.

Call spank_job_prolog() and spank_job_epilog() at prolog/epilog
time by invoking "slurmstepd spank [prolog|epilog]"

The prolog and epilog spank plugin hooks are not called within the
virtual address space of slurmd for at least a couple of reasons,
including

 1. Plugins dlopened in the address space of slurmd cannot be dlopened
   a second time. Therefore, static and global state in the DSO may
   be "dirty" in that some state may be preserved from the last epilog
   or prolog call, or even from the slurmd_init callback.

 2. The prolog and epilog need to be guaranteed reentrant. The safest
   way to guarantee this is to ensure prolog/epilog hooks are called
   from a separate address space.

 3. To satisfy "principle of least surprise" we want to have new plugins
   installed run their prolog/epilog hooks on the next job, just as
   if an update to the prolog/epilog script was made. The only way to
   guarantee this is to reload the spank plugin stack from plugstack.conf
   on each run. Because of #1 above, this needs to be done in a separate
   process.

Greatly simplify ability of code to get at current slurmstepd path
by setting slurmd's conf->stepd_loc to the default slurmstped path
if that path was not overridden on the command line.

This allows slurmd code to directly use conf->stepd_loc, instead of
requiring the duplicated code that created the default slurmstepd
path if conf->stepd_loc was not set at each call site.

Make waitpid_timeout() return more quickly when the child exits before
1s but after the initial call to waitpid(2).

Abstract the code for a waitpid(2) with timeout into a waitpid_timeout()
function for future use from other callers. For now, the function goes
into src/slurmd/common/run_script.c, since that is the original use
of the functionality.

Add new handle_spank_mode() function in slurmstepd to handle
when slurmstepd is called with "spank prolog" or "spank epilog".
In this function, the slurmd_conf_lite is read to handle reinitializing
the log facility as defined by slurmd config.

Factor out the read and write of the packed slurmd_conf_lite
data between slurmd and slurmstepd. This simplifies the code
in which that data is handled, and will allow for other callers
in the future.

The spank_job_prolog() and spank_job_epilog() spank calls need
to be run in a different address space from slurmd. This not allows
reinitializing the spank plugin stack on each run of the prolog or
epilog, but also ensures that any static data in plugins does not
propagate to each invocation of the job prolog and epilog (e.g. global
variables). Additionally, it is much safer to run these plugins
in a new process because we may be calling prolog/epilog for multiple
jobs at the same time.

This patch runs spank_job_prolog() or spank_job_epilog() from slurmstepd
when slurmstepd is invoked as

 slurmstepd spank [prolog|epilog]

The environment variables SLURM_JOBID and SLURM_UID are used to set
the jobid and uid for the prolog/epilog. Spank plugin options may
also be passed through the current environment.

Move special handling of slurmstepd cmdline to a function for
future expansion.

Add slurm_spank_job_prolog and slurm_spank_job_epilog callbacks
to the spank API, to be called just before the job prolog and epilog
scripts are executed.

These callbacks are not active until the hooks spank_job_prolog and
spank_job_epilog are added to slurmd.

Add new spank context "job script" for use during job prolog/epilog.

Add support for slurm_spank_slurmd_init and slurm_spank_slurmd_exit
symbols in spank plugins, to be called at slurmd startup and shutdown.

These are not functional yet until slurmd calls spank_slurmd_init()
and spank_slurmd_exit().

Currently spank_get_item and spank_job_control* are not valid in
slurmd context. Handle this case in relevant fucntions.

Prepare for spank plugins run in the context of slurmd daemon by
adding a new S_CTX_SLURMD context type.

The spank_set_remote_options_env() function is not used anywhere except
internal to plugstack.c, so remove it from plugstack.h. Then redefine
it to take a spank_stack argument so that it doesn't refer to the
global_spank_stack. Finally rename to spank_stack_set_remote_options_env()
to clarify the intent.

Refactor the post_opt handling code embedded in _spank_init() into
a spank_stack_post_opt() function, then call this in remote context
from a new spank_init_remote() function.

Instead of trying to handle missing plugstack.conf early in the code,
just treat missing plugstack.conf the same as an empty config.

Move struct spank_stack initialization code into a spank_stack_init()
function so that it can be called from multiple call sites.

Simplify code in _do_call_stack() by extracting case statement
to assign current callback symbol to its own function. Since all
spank functions have the same prototype we can then use the same
code to call _all_ callbacks, reducing greatly the number of lines
of code required.

Consolidate common code in spank_getenv, spank_setenv, spank_unsetenv
which checks for validity of the current context, spank handle, etc.

Consilidate checks for correct spank context in spank_job_control*
functions to avoid code duplication.

The use of globals in plugstack.c is cumbersome and prevents the
future expansion of spank plugins, e.g. calling spank plugins from
multiple contexts within the same process or reinitializing the
spank plugin state.

This patch consolidates the current globals (spank_stack, spank_ctx,
spank_optval, and option_cache) into a global "spank stack" structure
and expands many of the functions internal to plugstack.c to operate
on a struct spank_stack instead of globally.

There was likely a typo/thinko/patcho in the handling of the
return code from _do_call_stack(SPANK_INIT_POST_OPT) in _spank_init
in "remote" context. This error caused spank_init() to always
succeed, since the test less than zero would always return 0 or 1.

Avoid loading the same plugin more than once in plugstack.c.
Most likely this will be a configuration error, so we should
catch it early. If the same .so appears in the plugin stack
more than once, it is likely to cause very strange errors,
since dlopen() will only map the library a single time.

Change the owner of slurmctld and slurmdbd log files to the appropriate
user. Without this change the files will be created by and owned by the
user starting the daemons (likely user root).

CRAY: Fix support for configuration with SlurmdTimeout=0 (never mark
node that is DOWN in ALPS as DOWN in SLURM).

in the tightly coupled functions slurmd:stepd_completion and
slurmstepd:_handle_completion, a jobacct structure is
send from the main daemon to the step daemon to provide
the statistics of the children slurmstepd and do the aggregation.

The methodology used to send the structure is the use of
jobacct_gather_g_{setinfo,getinfo} over a pipe (JOBACCT_DATA_PIPE).
As {setinfo,getinfo} use a common internal lock and reading
or writing to a pipe is equivalent to holding a lock, slurmd and
slurmstepd have to avoid using both setinfo and getinfo over a
pipe or deadlock situations can occured. For example :
slurmd(lockforread,write)/slurmstepd(write,lockforread).

This patch remove the call to jobacct_gather_g_setinfo in slurmd
and the call to jobacct_gather_g_getinfo in slurmstepd ensuring
that slurmd only do getinfo operations over a pipe and slurmstepd
only do setinfo over a pipe. Instead jobacct_gather_g_{pack,unpack}
are used to marshall/unmarshall the data for transmission over the
pipe.
Patch by Matthieu Hautreux, CEA.

The patch committed here is a variation on the work by Matthieu.
Specifically, the logic is added to slurmstepd to read a new format
of RPC including an RPC version number and buffer with the data
structure. The slurmd however will not send the RPC in the new format
until SLURM version 2.5.