Dynamically load .so modules to expand functionality

[imp]

Summary

o Dynamically load all the .so files found in /libexec/nvmecontrol and

/usr/local/libexec/nvmecontrol.

o Link nvmecontrol -rdynamic so that its symbols are visible to the

libraries we load.

o Create mirrored linker sets that we dynamically expand.
o Add the linked-in top and logpage linker sets to the mirrors for them

and add those sets to the mirrors when we load a new .so.

o Add some macros to help hide the names of the linker sets.
o Update the man page.

Test Plan

run through all the commands before / after the change and make sure they work.
Use my experimental builds of intel.so and wdc.so to proof the concept.
A future commit will create the modules.

[cem]

Mostly LGTM, thanks.

sbin/nvmecontrol/nvmecontrol.c

89

style nit: separate line for static void

145–153

Seems like this could be simplified somewhat with realloc:

m->begin = reallocarray(m->begin, cur_n + add_n, sizeof(void *));
if (m->begin == NULL)
    err(1, "expanding mirror set");
memcpy(m->begin + cur_n, bp, add_n * sizeof(void *));

(err can be used, malloc/realloc/etc set errno on failure.)

305

extra space between NULL and ;

306

Could use dent->d_namlen in place of strlen(dent->d_name).

308–310

I'd err() this failure — if this tiny allocation failed, we're not going to be able to make much progress anyway.

332

closedir(dir); ?

342–343

I'm not sure if libexec is the right place for these. They might just be 'lib'. libexec is generally for helper programs, not modules. There is some precedence here in, e.g., /lib/geom. (Either way, it should be documented in hier.7.)

sbin/nvmecontrol/nvmecontrol.h

92

How did you land on the name "mirror?" It's confusing to me. Something like "contiguous_set" would make more sense to me, since the goal is to contain multiple .so's set tables in a single contiguous VA for easy searching.

[imp]

sbin/nvmecontrol/nvmecontrol.c

89

used in one place, and less gross than before, so expanded inline.

145–153

Good idea.

305

thanks

306

true. Don't really need to optimize here, but it's easy.

308–310

OK.

332

nice catch.

342–343

I Was all set to argue, but we have /lib/geom, /lib/casper (annd libexec/casper which just has executables) and /usr/lib/clang. So I'm sold.

[imp]

updates per cem@

[cem]

LGTM. I still find the 'mirror' name confusing, but maybe that's a personal problem ;-).

[imp]

In D18455#393317, @cem wrote:

LGTM. I still find the 'mirror' name confusing, but maybe that's a personal problem ;-).

I'll change it to set_concat which is less confusing.