Fix RTLD_NEXT handling to not return the symbol from the current library.

[bdrewery]

Summary

The root of the problem here is that TAILQ_FOREACH_FROM will default to
the head of the list if passed NULL.

Before r294373, this code was:

if (handle == RTLD_NEXT)
    obj = obj->next;
for (; obj != NULL; obj = obj->next) {

After r294373 it was:

if (handle == RTLD_NEXT)
      obj = globallist_next(obj);
TAILQ_FOREACH_FROM(obj, &obj_list, next) {

TAILQ_FOREACH_FROM expands to:

for ((var) = ((var) ? (var) : TAILQ_FIRST((head)));
    (var);
    (var) = TAILQ_NEXT((var), field))

Thus if passed NULL it defaults to the head of the list. The code here would
end up iterating over all shared libraries and eventually find the current
shared library and return the symbol from itself.

Sponsored by: EMC / Isilon Storage Division
MFC after: 1 week

Test Plan

Running https://github.com/bdrewery/RTLD_NEXT

[bdrewery]

By the way, the whitespace is a disaster in this file :(

[bdrewery]

What may be unclear is that in my test case, obj is becomes NULL from globallist_next(obj) since there are no further libraries loaded.

[kib]

I agree with the analysis and the direction of the fix, but think that the actual code changes add unneccessary layers there. Why not replace TAILQ_FOREACH_FROM() line with

for (; obj != NULL; obj = TAILQ_NEXT(obj, next)) {
...

? We never need to start iteration from the obj_list head.

[bdrewery]

In D7216#150086, @kib wrote:

I agree with the analysis and the direction of the fix, but think that the actual code changes add unneccessary layers there. Why not replace TAILQ_FOREACH_FROM() line with

for (; obj != NULL; obj = TAILQ_NEXT(obj, next)) {
...

? We never need to start iteration from the obj_list head.

Sure

[bdrewery]

Simplify the change