HarryR / nt365 / 26492732040

44%

Build Type

push

github

Committed by HarryR

Commit Message

lpc: NtReplyWaitReceivePortEx + decouple _KiKernelDispatch from gensrv

Three threads bundled because they hit the same files:

  1. New syscall.  NtReplyWaitReceivePortEx takes a Timeout argument
     that NtReplyWaitReceivePort had hard-coded NULL.  LPCRECV.C
     renames the body to the Ex variant (probes Timeout under the
     existing try/except, threads it into the pure-receive
     KeWaitForSingleObject), and NtReplyWaitReceivePort becomes a
     5-line wrapper passing NULL.  Plumbed end-to-end: SERVICES.TAB
     entry, ntlpcapi.h prototype, ntdll.lib + kernel32.lib refreshed,
     pkg/nt/dll/lpc.lua wrapper + a live LPC round-trip test in
     pkg/test/lpc.lua that uses the timeout to make a deadlock surface
     as a clean FAIL instead of hanging the whole selftest.

  2. _KiKernelDispatch out of gensrv-generated sysstubs.asm.  The
     handwritten kernel-mode Zw* direct-dispatch helper was wedged
     into sysstubs.asm between the STUBS_BEGIN macros and the
     SYSSTUBS_ENTRY1 invocations, so every services.tab change risked
     gensrv silently wiping it.  Move it to its own KIKDISP.ASM,
     declare the extrn _inside_ the SYSSTUBS_ENTRY1 macro in
     SERVICES.STB (so the symbol only enters the COFF symbol table
     when the macro actually expands, keeping the user-side
     usrstubs.asm from breaking with an unresolved external on the
     kernel-only helper).  Adding a new syscall is now a one-line
     edit to SERVICES.TAB plus a rebuild.

  3. Friendlier bugcheck text.  The boilerplate
       'Restart and set the recovery options in the system control
        panel or the /CRASHDEBUG system start option.'
     ships to MicroNT testers as bad advice (we don't have a control
     panel and our STOPs are interesting).  Point them at the GitHub
     issue tracker instead.

The companion KE/WAIT.C fix in the previous commit (KeReleaseWaitFor-
Semaphore fast path leaks per-thread timer) is what makes the new
timeout-using LPC service safe ... (continued)

Coverage Stats

73 of 185 new or added lines in 3 files covered. (39.46%)

237 existing lines in 28 files now uncovered.

46772 of 114848 relevant lines covered (40.73%)

1861.57 hits per line