vxworks osloader.c functions not using mutex

[skliper]

The osloader.c functions do not properly protect the OS_module_table, OS_sym_table_file_id, and OS_symbol_table_size with the OS_module_table_mut.

These items are being read from and changed outside of the mutex.

(Discovered as part of #45 coverage testing.)

[skliper]

Imported from trac issue 76. Created by abrown4 on 2015-07-15T11:47:13, last modified: 2015-12-01T14:13:24

[skliper]

Trac comment by abrown4 on 2015-07-20 15:43:12:

Discovered OS_ModuleLoad() also wasn't always cleaning up when encountering an error and performing an early return with
{{{
OS_module_table[possible_moduleid].free = TRUE ;
}}}
Put in cleanup code and added test checks.

[skliper]

Trac comment by abrown4 on 2015-07-20 16:14:32:

commit [changeset:ecfe66e] Trac #99, osloader mutex unit test update.
commit [changeset:be8d33e] Trac #99, osloader mutex fix.

BEFORE THE osloader.c FIX:
Test failures:
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 365
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 459
FAIL: retval != OS_SUCCESS, File: osloader_testcase.c, Line: 603
FAIL: Terminating filename null in table.name, File: osloader_testcase.c, Line: 605
FAIL: getNSemTake() == 0, File: osloader_testcase.c, Line: 609
FAIL: Table is empty, File: osloader_testcase.c, Line: 612
FAIL: Table is empty, File: osloader_testcase.c, Line: 739
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 831
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 876
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 900
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 932
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 1006
FAIL: getNSemTake() > 0, File: osloader_testcase.c, Line: 1032

Tests Executed: 34
Assert Pass Count: 273
Assert Fail Count: 13

AFTER THE FIX:
Tests failing due to #98 (module_name too long)
FAIL: retval != OS_SUCCESS, File: osloader_testcase.c, Line: 603
FAIL: Terminating filename null in table.name, File: osloader_testcase.c, Line: 605
FAIL: getNSemTake() == 0, File: osloader_testcase.c, Line: 609
FAIL: Table is empty, File: osloader_testcase.c, Line: 612

Tests Executed: 34
Assert Pass Count: 282
Assert Fail Count: 4

[skliper]

Trac comment by abrown4 on 2015-07-28 15:09:15:

Comment for review: The changes for this issue come down to the premise that it is safer to globally lock the OS_module_table and serialize access during OS_SymbolTableDump(), OS_ModuleLoad(), OS_ModuleUnload(), and OS_ModuleInfo(). If that is not the case then recommend removing these changes and addressing this issue at a later time.

[skliper]

Trac comment by abrown4 on 2015-08-20 18:29:17:

From the above commits, as of #45 [changeset:6a9db70], the white-box unit test still shows that OS_SymbolTableDump(), and OS_ModuleInfo() don't safely access the OS_module_table (5 test failures from osloader_wb test).

[skliper]

Trac comment by abrown4 on 2015-11-25 12:59:12:

Note, after an 8/19 code inspection (#45), [changeset:ab00bb5] removed proposed locking for OS_SymbolTableDump and [changeset:abdf238] removed proposed locking for OS_ModuleInfo. Note, locking is still not present in those functions.

Traceability: the remaining osloader.c locking changes are reflected in #138, [changeset:ab2d6d7]. Recommend closure of this issue in favor of #138 with the caveat that several osloader.c functions are (still) not thread-safe.