utils - zygaena/zyginit

root / utils
utils Plain Text 1006 lines 32.9 KB
[package]
name = "contract_probe"
version = "0.1.0"
author = "Chris Tusa <chris.tusa@leafscale.com>"
description = "FFI validation probe for zyginit Hammerhead libcontract bindings"
license = "CDDL-1.0"

[build]
entry = "src/main.reef"
output = "contract_probe"
output_dir = "build"
source_dirs = ["src"]

# This probe exercises each libcontract FFI call in isolation to validate
# the bindings match the real kernel interface on Hammerhead.
#
# On Linux (dev) — stubs return -1, build only validates compile/link:
#   clang -c ../../src/contract_linux_stubs.c -o build/contract_linux_stubs.o
#   reefc build --obj build/contract_linux_stubs.o
#
# On Hammerhead — real libcontract, actual validation:
#   reefc build -l contract
/******************************************************************************
                __               ____                __
               / /   ___  ____ _/ __/_____________ _/ /__
              / /   / _ \/ __ `/ /_/ ___/ ___/ __ `/ / _ \
             / /___/  __/ /_/ / __(__  ) /__/ /_/ / /  __/
            /_____/\___/\__,_/_/ /____/\___/\__,_/_/\___/

    (C)opyright 2025-2026, Leafscale, LLC -  https://www.leafscale.com

    Project: zyginit
   Filename: main.reef
    Authors: Chris Tusa <chris.tusa@leafscale.com>
    License: <see LICENSE file included with this source code>
Description: libcontract FFI validation probe — exercises each kernel
             contract call used by zyginit, in isolation, with PASS/FAIL
             output per probe. Run on Hammerhead after build with
             -l contract to verify zyginit's libcontract bindings match
             the kernel interface.

******************************************************************************/

module contract_probe

import sys.process as process
import sys.signal as signal
import sys.poll as poll
import time.clock as clock
import core.str
import io.dir as dir

// ============================================================================
// FFI declarations — libcontract and POSIX
// ============================================================================

extern "C" fn ct_tmpl_activate(fd: int): int
extern "C" fn ct_tmpl_clear(fd: int): int
extern "C" fn ct_tmpl_set_informative(fd: int, events: int): int
extern "C" fn ct_tmpl_set_critical(fd: int, events: int): int
extern "C" fn ct_pr_tmpl_set_fatal(fd: int, events: int): int
extern "C" fn ct_pr_tmpl_set_param(fd: int, params: int): int
extern "C" fn ct_status_read(fd: int, detail: int, statp: pointer): int
extern "C" proc ct_status_free(stathdl: pointer)
extern "C" fn ct_status_get_id(stathdl: pointer): int
extern "C" fn ct_status_get_holder(stathdl: pointer): int
extern "C" fn ct_status_get_state(stathdl: pointer): int
extern "C" fn ct_ctl_abandon(fd: int): int
extern "C" fn ct_ctl_adopt(fd: int): int
extern "C" fn ct_event_read(fd: int, evtp: pointer): int
extern "C" proc ct_event_free(evthdl: pointer)
extern "C" fn ct_event_get_ctid(evthdl: pointer): int
extern "C" fn ct_event_get_type(evthdl: pointer): int
extern "C" fn sigsend(idtype: int, id: int, sig: int): int
extern "C" fn open(path: string, flags: int): int
extern "C" fn close(fd: int): int

// From probe_helpers.c — errno and strerror bridges
extern "C" fn probe_errno(): int
extern "C" fn probe_strerror(err: int): string

// ============================================================================
// Constants (mirror sys/contract.h, sys/contract/process.h, sys/procset.h)
// ============================================================================

fn O_RDONLY(): int return 0 end O_RDONLY
fn O_WRONLY(): int return 1 end O_WRONLY
fn O_RDWR(): int   return 2 end O_RDWR

fn CT_PR_EV_EMPTY(): int   return 1  end CT_PR_EV_EMPTY
fn CT_PR_EV_HWERR(): int   return 32 end CT_PR_EV_HWERR
fn CT_PR_INHERIT(): int    return 1  end CT_PR_INHERIT
fn CT_PR_NOORPHAN(): int   return 2  end CT_PR_NOORPHAN
fn CTD_COMMON(): int       return 0  end CTD_COMMON
fn P_CTID(): int           return 13 end P_CTID

// ctstate_t enum (sys/contract.h)
fn CTS_OWNED(): int     return 0 end CTS_OWNED
fn CTS_INHERITED(): int return 1 end CTS_INHERITED
fn CTS_ORPHAN(): int    return 2 end CTS_ORPHAN
fn CTS_DEAD(): int      return 3 end CTS_DEAD

fn state_name(state: int): string
    if state == 0
        return "OWNED"
    elif state == 1
        return "INHERITED"
    elif state == 2
        return "ORPHAN"
    elif state == 3
        return "DEAD"
    end if
    return "UNKNOWN"
end state_name

fn TEMPLATE_PATH(): string return "/system/contract/process/template" end TEMPLATE_PATH
fn LATEST_PATH(): string   return "/system/contract/process/latest"   end LATEST_PATH
fn BUNDLE_PATH(): string   return "/system/contract/process/bundle"   end BUNDLE_PATH

// ============================================================================
// Result tracking
// ============================================================================

mut g_passed: int = 0
mut g_failed: int = 0

proc section(title: string)
    println("")
    println("=== " + title + " ===")
end section

proc pass(name: string)
    println("  PASS: " + name)
    g_passed = g_passed + 1
end pass

proc fail(name: string, detail: string)
    println("  FAIL: " + name + " — " + detail)
    g_failed = g_failed + 1
end fail

// Parse a non-negative int from a string. Returns -1 on any non-digit.
fn str_to_int(s: string): int
    let len = str.length(s)
    if len <= 0
        return 0 - 1
    end if

    mut value = 0
    mut i = 0
    while i < len
        let ch = str.substring(s, i, 1)
        let digit = str.index_of("0123456789", ch)
        if digit < 0
            return 0 - 1
        end if
        value = value * 10 + digit
        i = i + 1
    end while
    return value
end str_to_int

// Convert int to string (for error messages and path building)
fn int_to_str(n: int): string
    if n == 0
        return "0"
    end if

    mut value = n
    if n < 0
        value = 0 - n
    end if

    mut result = ""
    while value > 0
        let digit = value % 10
        result = str.concat(str.substring("0123456789", digit, 1), result)
        value = value / 10
    end while

    if n < 0
        result = str.concat("-", result)
    end if

    return result
end int_to_str

// ============================================================================
// Shared helpers
// ============================================================================

// Set up a contract template exactly as zyginit does.
// Returns template fd on success, -1 on error.
fn setup_template(): int
    let tmpl_fd = open(TEMPLATE_PATH(), O_RDWR())
    if tmpl_fd < 0
        return 0 - 1
    end if

    let rc1 = ct_tmpl_set_informative(tmpl_fd, CT_PR_EV_EMPTY())
    let rc2 = ct_tmpl_set_critical(tmpl_fd, CT_PR_EV_EMPTY() + CT_PR_EV_HWERR())
    let rc3 = ct_pr_tmpl_set_fatal(tmpl_fd, CT_PR_EV_HWERR())
    let rc4 = ct_pr_tmpl_set_param(tmpl_fd, CT_PR_INHERIT() + CT_PR_NOORPHAN())

    if rc1 != 0 or rc2 != 0 or rc3 != 0 or rc4 != 0
        close(tmpl_fd)
        return 0 - 1
    end if

    if ct_tmpl_activate(tmpl_fd) != 0
        close(tmpl_fd)
        return 0 - 1
    end if

    return tmpl_fd
end setup_template

// Read the contract ID of the most recently created contract.
fn get_latest_contract(): int
    let latest_fd = open(LATEST_PATH(), O_RDONLY())
    if latest_fd < 0
        return 0 - 1
    end if

    unsafe
        let stathdl_buf = new [pointer](1)
        let rc = ct_status_read(latest_fd, CTD_COMMON(), stathdl_buf)
        close(latest_fd)

        if rc != 0
            return 0 - 1
        end if

        let stathdl = stathdl_buf[0]
        let ctid = ct_status_get_id(stathdl)
        ct_status_free(stathdl)
        return ctid
    end unsafe
end get_latest_contract

// Exec /usr/bin/sleep in the child. Seconds given as string.
// Never returns on success — this is a one-way door for the child.
proc child_sleep_and_exit(seconds: string)
    let argv = new [string](2)
    argv[0] = "sleep"
    argv[1] = seconds
    let _ = process.process_exec("/bin/sleep", argv)
    // If we got here, exec failed — exit cleanly with an unusual code
    process.exit_now(127)
end child_sleep_and_exit

// ============================================================================
// Probe 1: setup_template + open_bundle
// ============================================================================

proc probe_1_template_and_bundle()
    section("Probe 1: setup_template + open_bundle")

    let tmpl_fd = setup_template()
    if tmpl_fd < 0
        fail("setup_template", "returned -1 (check /system/contract/process/template perms)")
        return
    end if
    pass("setup_template returned valid fd (" + int_to_str(tmpl_fd) + ")")

    let bundle_fd = open(BUNDLE_PATH(), O_RDONLY())
    if bundle_fd < 0
        fail("open_bundle", "could not open " + BUNDLE_PATH())
        let _ = ct_tmpl_clear(tmpl_fd)
        close(tmpl_fd)
        return
    end if
    pass("open_bundle returned valid fd (" + int_to_str(bundle_fd) + ")")

    let _ = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)
    close(bundle_fd)
end probe_1_template_and_bundle

// ============================================================================
// Probe 2: fork under active template + get_latest_contract
// ============================================================================

proc probe_2_fork_and_latest()
    section("Probe 2: fork + get_latest_contract")

    let tmpl_fd = setup_template()
    if tmpl_fd < 0
        fail("setup (precondition)", "template failed")
        return
    end if

    let pid = process.process_fork()
    if pid < 0
        fail("process_fork", "returned -1")
        let _ = ct_tmpl_clear(tmpl_fd)
        close(tmpl_fd)
        return
    end if

    if pid == 0
        child_sleep_and_exit("1")
    end if

    let ctid = get_latest_contract()

    let _ = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)

    if ctid <= 0
        fail("get_latest_contract", "returned non-positive CTID (" + int_to_str(ctid) + ")")
    else
        pass("get_latest_contract returned CTID=" + int_to_str(ctid))
    end if

    let _ = process.process_wait(pid)
end probe_2_fork_and_latest

// ============================================================================
// Probe 3: read_event receives CT_PR_EV_EMPTY when child exits
// ============================================================================

proc probe_3_read_event_empty()
    section("Probe 3: read_event CT_PR_EV_EMPTY on child exit")

    let tmpl_fd = setup_template()
    if tmpl_fd < 0
        fail("setup (precondition)", "template failed")
        return
    end if

    let bundle_fd = open(BUNDLE_PATH(), O_RDONLY())
    if bundle_fd < 0
        fail("setup (precondition)", "bundle failed")
        let _ = ct_tmpl_clear(tmpl_fd)
        close(tmpl_fd)
        return
    end if

    let pid = process.process_fork()
    if pid == 0
        child_sleep_and_exit("1")
    end if

    let expected_ctid = get_latest_contract()

    let _ = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)

    // Wait for child to exit so the contract transitions to empty
    let _ = process.process_wait(pid)

    // Poll for the resulting event (5s ceiling)
    poll.poll_clear()
    let idx = poll.poll_add(bundle_fd, poll.POLLIN())
    let ready = poll.poll_wait(5000)

    if ready <= 0
        fail("poll bundle_fd", "no event within 5s")
        close(bundle_fd)
        return
    end if

    if not poll.poll_readable(idx)
        fail("poll bundle_fd", "not readable")
        close(bundle_fd)
        return
    end if

    unsafe
        let evthdl_buf = new [pointer](1)
        let rc = ct_event_read(bundle_fd, evthdl_buf)
        if rc != 0
            fail("ct_event_read", "returned " + int_to_str(rc))
            close(bundle_fd)
            return
        end if

        let evthdl = evthdl_buf[0]
        let got_ctid = ct_event_get_ctid(evthdl)
        let got_type = ct_event_get_type(evthdl)
        ct_event_free(evthdl)

        if got_ctid == expected_ctid
            pass("ct_event_get_ctid matches expected (CTID=" + int_to_str(got_ctid) + ")")
        else
            fail("ct_event_get_ctid mismatch",
                 "expected " + int_to_str(expected_ctid) + ", got " + int_to_str(got_ctid))
        end if

        if got_type == CT_PR_EV_EMPTY()
            pass("ct_event_get_type is CT_PR_EV_EMPTY")
        else
            fail("ct_event_get_type",
                 "got " + int_to_str(got_type) + ", expected " + int_to_str(CT_PR_EV_EMPTY()))
        end if
    end unsafe

    close(bundle_fd)
end probe_3_read_event_empty

// ============================================================================
// Probe 4: kill_contract actually signals the child
// ============================================================================

proc probe_4_kill_contract()
    section("Probe 4: kill_contract (sigsend P_CTID)")

    let tmpl_fd = setup_template()
    if tmpl_fd < 0
        fail("setup (precondition)", "template failed")
        return
    end if

    let pid = process.process_fork()
    if pid == 0
        child_sleep_and_exit("30")
    end if

    let ctid = get_latest_contract()
    let _ = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)

    if ctid <= 0
        fail("kill_contract precondition", "no CTID")
        let _ = process.process_kill(pid, signal.SIGKILL())
        let _ = process.process_wait(pid)
        return
    end if

    // Give the child time to reach sleep(2), then signal its contract
    clock.sleep_millis(200)

    let rc = sigsend(P_CTID(), ctid, signal.SIGTERM())
    if rc != 0
        let err = probe_errno()
        fail("sigsend(P_CTID, ctid=" + int_to_str(ctid) + ", SIGTERM)",
             "returned " + int_to_str(rc) + " errno=" + int_to_str(err) +
             " (" + probe_strerror(err) + ")")

        // Diagnostic: does sigsend work at all with P_PID?
        let rc2 = sigsend(0, pid, 0)   // 0=P_PID, sig 0 = permission probe
        if rc2 == 0
            println("    diag: sigsend(P_PID, pid, 0) works — issue is with P_CTID or ctid")
        else
            let err2 = probe_errno()
            println("    diag: sigsend(P_PID, pid, 0) also fails: errno=" +
                    int_to_str(err2) + " (" + probe_strerror(err2) + ")")
        end if

        let _ = process.process_kill(pid, signal.SIGKILL())
        let _ = process.process_wait(pid)
        return
    end if
    pass("sigsend(P_CTID, ctid=" + int_to_str(ctid) + ", SIGTERM) returned 0")

    let _ = process.process_wait(pid)
    if process.process_was_signaled()
        pass("child terminated by signal (" + int_to_str(process.process_term_signal()) + ")")
    else
        fail("child exit state",
             "expected signaled, got normal exit code=" + int_to_str(process.process_exit_code()))
    end if
end probe_4_kill_contract

// ============================================================================
// Probe 5: abandon_contract via ctl fd
// ============================================================================

proc probe_5_abandon_contract()
    section("Probe 5: abandon_contract (ct_ctl_abandon)")

    let tmpl_fd = setup_template()
    if tmpl_fd < 0
        fail("setup (precondition)", "template failed")
        return
    end if

    let pid = process.process_fork()
    if pid == 0
        child_sleep_and_exit("1")
    end if

    let ctid = get_latest_contract()
    let _ = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)

    if ctid <= 0
        fail("abandon precondition", "no CTID")
        let _ = process.process_wait(pid)
        return
    end if

    let ctl_path = "/system/contract/process/" + int_to_str(ctid) + "/ctl"
    let ctl_fd = open(ctl_path, O_WRONLY())
    if ctl_fd < 0
        fail("open ctl fd", "could not open " + ctl_path)
        let _ = process.process_wait(pid)
        return
    end if
    pass("opened ctl fd for " + ctl_path)

    let rc = ct_ctl_abandon(ctl_fd)
    close(ctl_fd)

    if rc == 0
        pass("ct_ctl_abandon returned 0")
    else
        fail("ct_ctl_abandon", "returned " + int_to_str(rc))
    end if

    // Reap the child — still our child at the process level
    let _ = process.process_wait(pid)
end probe_5_abandon_contract

// ============================================================================
// Probe 6: enumerate CTFS + read full status (id, holder, state)
// ============================================================================

proc probe_6_enumerate_contracts()
    section("Probe 6: enumerate /system/contract/process + read status fields")

    // Establish a known-live contract so the scan has a ground truth to find.
    let tmpl_fd = setup_template()
    if tmpl_fd < 0
        fail("setup (precondition)", "template failed")
        return
    end if

    let pid = process.process_fork()
    if pid == 0
        child_sleep_and_exit("3")
    end if

    let known_ctid = get_latest_contract()
    let _ = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)

    if known_ctid <= 0
        fail("enum precondition", "no known CTID")
        let _ = process.process_wait(pid)
        return
    end if

    // Let the child finish exec so its contract state is stable.
    clock.sleep_millis(200)

    let max_entries = 512
    mut entries = new [string](max_entries)
    let count = dir.list_dir("/system/contract/process", entries, max_entries)
    if count < 0
        fail("list_dir", "returned " + int_to_str(count))
        let _ = process.process_kill(pid, signal.SIGTERM())
        let _ = process.process_wait(pid)
        return
    end if
    pass("list_dir returned " + int_to_str(count) + " raw entries")

    mut scanned = 0
    mut owned = 0
    mut inherited = 0
    mut orphan = 0
    mut dead = 0
    mut found_known = false
    mut known_state = 0 - 1
    mut known_holder = 0 - 1

    mut i = 0
    while i < count
        let name = entries[i]
        if name != "." and name != ".."
            let parsed = str_to_int(name)
            if parsed > 0
                let status_path = "/system/contract/process/" + name + "/status"
                let status_fd = open(status_path, O_RDONLY())
                if status_fd >= 0
                    unsafe
                        let stathdl_buf = new [pointer](1)
                        let rc = ct_status_read(status_fd, CTD_COMMON(), stathdl_buf)
                        close(status_fd)
                        if rc == 0
                            let stathdl = stathdl_buf[0]
                            let got_ctid = ct_status_get_id(stathdl)
                            let holder = ct_status_get_holder(stathdl)
                            let state = ct_status_get_state(stathdl)
                            ct_status_free(stathdl)

                            scanned = scanned + 1
                            if state == CTS_OWNED()
                                owned = owned + 1
                            elif state == CTS_INHERITED()
                                inherited = inherited + 1
                            elif state == CTS_ORPHAN()
                                orphan = orphan + 1
                            else
                                dead = dead + 1
                            end if

                            if got_ctid == known_ctid
                                found_known = true
                                known_state = state
                                known_holder = holder
                            end if
                        end if
                    end unsafe
                end if
            end if
        end if
        i = i + 1
    end while

    pass("read status on " + int_to_str(scanned) + " contracts")
    println("    state breakdown: OWNED=" + int_to_str(owned) +
            " INHERITED=" + int_to_str(inherited) +
            " ORPHAN=" + int_to_str(orphan) +
            " DEAD=" + int_to_str(dead))

    if not found_known
        fail("locate known contract in CTFS",
             "CTID=" + int_to_str(known_ctid) + " not found among scanned entries")
    else
        pass("located known contract CTID=" + int_to_str(known_ctid) +
             " (state=" + state_name(known_state) +
             " holder=" + int_to_str(known_holder) + ")")

        let our_pid = process.getpid()
        if known_state == CTS_OWNED() and known_holder == our_pid
            pass("known contract: state=CTS_OWNED, holder=our PID (" + int_to_str(our_pid) + ")")
        else
            fail("known contract fields",
                 "expected state=CTS_OWNED + holder=" + int_to_str(our_pid) +
                 ", got state=" + state_name(known_state) +
                 " holder=" + int_to_str(known_holder))
        end if
    end if

    // Cleanup
    let _ = process.process_kill(pid, signal.SIGTERM())
    let _ = process.process_wait(pid)
end probe_6_enumerate_contracts

// ============================================================================
// Probe 7: ct_ctl_adopt FFI linkage smoke test on an unowned contract
// ============================================================================

proc probe_7_adopt_linkage()
    section("Probe 7: ct_ctl_adopt linkage smoke")

    // Part A: kernel denies ctl access to contracts we don't own.
    //   Contract 1 is typically owned by sched / init. Opening its ctl
    //   fd from an unrelated process should be denied (EACCES).
    let unowned_fd = open("/system/contract/process/1/ctl", O_WRONLY())
    if unowned_fd >= 0
        let _ = close(unowned_fd)
        fail("unowned ctl open",
             "unexpectedly opened /system/contract/process/1/ctl — expected EACCES")
    else
        let err = probe_errno()
        if err == 13  // EACCES
            pass("unowned ctl open correctly denied (EACCES)")
        else
            pass("unowned ctl open denied with errno=" + int_to_str(err) +
                 " (" + probe_strerror(err) + ")")
        end if
    end if

    // Part B: ct_ctl_adopt FFI call path on one of our own contracts.
    //   Create an owned contract, open its ctl fd (should succeed), then
    //   try ct_ctl_adopt on it. Adopt must fail because the contract is
    //   in state CTS_OWNED, not CTS_INHERITED — confirming the FFI call
    //   reaches the kernel and the kernel enforces adopt semantics.
    let tmpl_fd = setup_template()
    if tmpl_fd < 0
        fail("setup (precondition)", "template failed")
        return
    end if

    let pid = process.process_fork()
    if pid == 0
        child_sleep_and_exit("2")
    end if

    let ctid = get_latest_contract()
    let _ = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)

    if ctid <= 0
        fail("adopt precondition", "no CTID")
        let _ = process.process_wait(pid)
        return
    end if

    let ctl_path = "/system/contract/process/" + int_to_str(ctid) + "/ctl"
    let ctl_fd = open(ctl_path, O_WRONLY())
    if ctl_fd < 0
        let err = probe_errno()
        fail("own ctl open",
             ctl_path + " errno=" + int_to_str(err) +
             " (" + probe_strerror(err) + ")")
        let _ = process.process_wait(pid)
        return
    end if
    pass("opened own ctl fd for CTID=" + int_to_str(ctid))

    let rc = ct_ctl_adopt(ctl_fd)
    let err = probe_errno()
    close(ctl_fd)

    if rc == 0
        fail("ct_ctl_adopt(owned)", "unexpectedly succeeded on CTS_OWNED contract")
    else
        pass("ct_ctl_adopt(owned) returned " + int_to_str(rc) +
             " errno=" + int_to_str(err) + " (" + probe_strerror(err) + ")")
        println("    — FFI linkage confirmed; adopt refused on non-inherited state")
    end if

    let _ = process.process_wait(pid)
end probe_7_adopt_linkage

// ============================================================================
// Entry point
// ============================================================================

proc main()
    println("zyginit libcontract FFI probe")
    println("=============================")
    println("pid=" + int_to_str(process.getpid()))

    probe_1_template_and_bundle()
    probe_2_fork_and_latest()
    probe_3_read_event_empty()
    probe_4_kill_contract()
    probe_5_abandon_contract()
    probe_6_enumerate_contracts()
    probe_7_adopt_linkage()

    println("")
    println("=============================")
    println("Results: " + int_to_str(g_passed) + " passed, " +
            int_to_str(g_failed) + " failed")
    println("=============================")

    if g_failed > 0
        process.exit_now(1)
    end if
end main

end module
/*
 * probe_helpers.c — tiny helpers for the FFI probe
 *
 * Exposes errno readout and a strerror bridge so the Reef probe can report
 * the real kernel error when an FFI call fails.
 */

#include <errno.h>
#include <string.h>

int probe_errno(void) {
    return errno;
}

const char *probe_strerror(int err) {
    return strerror(err);
}
[package]
name = "replace_probe"
version = "0.1.0"
author = "Chris Tusa <chris.tusa@leafscale.com>"
description = "Unit-test probe for zyginit live-replace state-file format"
license = "CDDL-1.0"

[build]
entry = "src/main.reef"
output = "replace_probe"
output_dir = "build"
source_dirs = ["src", "../../src"]

# This probe round-trips state.toml serialization without needing a
# running zyginit. Models utils/contract_probe.
#
# NOTE: reefc does not use source_dirs from reef.toml for module resolution.
# Module resolution falls back to src/ in CWD. Build from the zyginit root:
#
# Linux (from zyginit/):
#   clang -c src/helpers.c -o build/helpers.o
#   clang -c src/contract_linux_stubs.c -o build/contract_linux_stubs.o
#   reefc utils/replace_probe/src/main.reef \
#       -o utils/replace_probe/build/replace_probe \
#       --obj build/helpers.o --obj build/contract_linux_stubs.o
/******************************************************************************
                __               ____                __
               / /   ___  ____ _/ __/_____________ _/ /__
              / /   / _ \/ __ `/ /_/ ___/ ___/ __ `/ / _ \
             / /___/  __/ /_/ / __(__  ) /__/ /_/ / /  __/
            /_____/\___/\__,_/_/ /____/\___/\__,_/_/\___/

    (C)opyright 2025-2026, Leafscale, LLC -  https://www.leafscale.com

    Project: zyginit
   Filename: main.reef
    Authors: Chris Tusa <chris.tusa@leafscale.com>
    License: <see LICENSE file included with this source code>
Description: Probe — round-trip the state.toml format through replace.reef

******************************************************************************/

import supervisor
import replace
import io.dir as dir
import io.file
import core.str
import sys.env as sysenv
import sys.process as process

mut g_pass: int = 0
mut g_fail: int = 0

proc check(label: string, ok: bool)
    if ok
        g_pass = g_pass + 1
        println("  PASS: " + label)
    else
        g_fail = g_fail + 1
        println("  FAIL: " + label)
    end if
end check

fn build_synthetic_table(): supervisor.ServiceTable
    let table = supervisor.new_service_table(8)

    // sshd: RUNNING with restart_count=2 and last_exit_code=0
    let def_sshd = supervisor.make_def_with_name("sshd")
    let idx_sshd = supervisor.add_service(table, def_sshd)
    supervisor.adopt_runtime(table, idx_sshd, supervisor.STATE_RUNNING(),
        100, 23, 2, 0, 1746204051)

    // syslogd: RUNNING with no restarts, last_exit_code = -1 (never exited)
    let def_syslog = supervisor.make_def_with_name("syslogd")
    let idx_syslog = supervisor.add_service(table, def_syslog)
    supervisor.adopt_runtime(table, idx_syslog, supervisor.STATE_RUNNING(),
        101, 17, 0, 0 - 1, 1746204047)

    // cron: STOPPED — must NOT be serialized
    let def_cron = supervisor.make_def_with_name("cron")
    let idx_cron = supervisor.add_service(table, def_cron)
    supervisor.adopt_runtime(table, idx_cron, supervisor.STATE_STOPPED(),
        0 - 1, 0 - 1, 0, 0, 0)

    return table
end build_synthetic_table

proc test_round_trip()
    println("== round-trip ==")
    let tmpdir = sysenv.get_env_or("TMPDIR", "/tmp")
    let probe_dir = str.concat(tmpdir, "/replace_probe")
    let _ = sysenv.set_env("ZYGINIT_RUN_DIR", probe_dir)

    if not dir.dir_exists(probe_dir)
        let _ = dir.create_dir_all(probe_dir)
    end if

    let table = build_synthetic_table()
    let serialized = replace.serialize_state(table, 1746204000)
    check("serialize returns true", serialized)

    let recovered = replace.recover_state(1746204000)
    check("recover boot_time matches", replace.rs_boot_time(recovered) == 1746204000)
    check("recover count == 2 (running only)", replace.rs_count(recovered) == 2)

    // Find services by name (order is implementation-defined)
    mut found_sshd = false
    mut found_syslog = false
    mut i = 0
    while i < replace.rs_count(recovered)
        let svc = replace.rs_service(recovered, i)
        let name = replace.svc_name(svc)
        if name == "sshd"
            found_sshd = true
            check("sshd contract_id = 23", replace.svc_contract_id(svc) == 23)
            check("sshd restart_count = 2", replace.svc_restart_count(svc) == 2)
            check("sshd last_exit_code = 0", replace.svc_last_exit_code(svc) == 0)
            check("sshd last_start_time = 1746204051", replace.svc_last_start_time(svc) == 1746204051)
        elif name == "syslogd"
            found_syslog = true
            check("syslogd contract_id = 17", replace.svc_contract_id(svc) == 17)
            check("syslogd last_exit_code = -1", replace.svc_last_exit_code(svc) == 0 - 1)
        end if
        i = i + 1
    end while
    check("found sshd", found_sshd)
    check("found syslogd", found_syslog)

    // After recover_state, state.toml is deleted
    check("state.toml deleted after recovery", not file.fileExists(replace.STATE_FILE_PATH()))
end test_round_trip

proc test_stale_boot_time()
    println("== stale state file ==")
    let tmpdir = sysenv.get_env_or("TMPDIR", "/tmp")
    let probe_dir = str.concat(tmpdir, "/replace_probe")
    let _ = sysenv.set_env("ZYGINIT_RUN_DIR", probe_dir)

    if not dir.dir_exists(probe_dir)
        let _ = dir.create_dir_all(probe_dir)
    end if

    let table = build_synthetic_table()
    let _ = replace.serialize_state(table, 1746204000)

    // Recover with DIFFERENT boot_time -> stale -> empty + deleted
    let recovered = replace.recover_state(1746999999)
    check("stale file -> empty state", replace.rs_is_empty(recovered))
    check("stale file deleted", not file.fileExists(replace.STATE_FILE_PATH()))
end test_stale_boot_time

proc test_missing_file()
    println("== missing state file ==")
    let tmpdir = sysenv.get_env_or("TMPDIR", "/tmp")
    // Unique per-run dir to guarantee no leftover state.toml
    let probe_dir = str.concat(tmpdir, "/replace_probe_missing")
    let _ = sysenv.set_env("ZYGINIT_RUN_DIR", probe_dir)

    if not dir.dir_exists(probe_dir)
        let _ = dir.create_dir_all(probe_dir)
    end if

    // No need to unlink — the dir is fresh per invocation; if a prior
    // run left a state.toml here, the boot_time will differ and recovery
    // will treat it as stale (exercising a different code path). Either
    // way the test asserts "empty state" which holds.

    let recovered = replace.recover_state(1746204000)
    check("missing-or-stale file -> empty state", replace.rs_is_empty(recovered))
end test_missing_file

proc test_preconditions()
    println("== preconditions ==")
    let table = supervisor.new_service_table(8)

    // sshd RUNNING — stable
    let def_sshd = supervisor.make_def_with_name("sshd")
    let idx_sshd = supervisor.add_service(table, def_sshd)
    supervisor.adopt_runtime(table, idx_sshd, supervisor.STATE_RUNNING(),
        100, 23, 0, 0 - 1, 1746204051)

    let r1 = replace.check_preconditions(table)
    check("all-RUNNING -> empty report", str.length(r1) == 0)

    // Add a STOPPING service
    let def_cron = supervisor.make_def_with_name("cron")
    let idx_cron = supervisor.add_service(table, def_cron)
    supervisor.adopt_runtime(table, idx_cron, supervisor.STATE_STOPPING(),
        101, 24, 0, 0 - 1, 1746204060)

    let r2 = replace.check_preconditions(table)
    check("STOPPING -> non-empty report", str.length(r2) > 0)
    check("report mentions cron", str.index_of(r2, "cron") >= 0)
    check("report mentions stopping", str.index_of(r2, "stopping") >= 0)

    // Add a WAITING service — should also be flagged transient
    let def_syslogd = supervisor.make_def_with_name("syslogd")
    let idx_syslogd = supervisor.add_service(table, def_syslogd)
    supervisor.adopt_runtime(table, idx_syslogd, supervisor.STATE_WAITING(),
        0 - 1, 0 - 1, 1, 1, 0)

    let r3 = replace.check_preconditions(table)
    check("WAITING also flagged", str.index_of(r3, "syslogd") >= 0)
    check("multiple entries comma-separated", str.index_of(r3, ", ") >= 0)
end test_preconditions

proc main()
    test_round_trip()
    test_stale_boot_time()
    test_missing_file()
    test_preconditions()

    println("")
    println("--- " + int_to_str(g_pass) + " passed, " + int_to_str(g_fail) + " failed ---")
    if g_fail > 0
        process.exit_now(1)
    end if
end main

fn int_to_str(n: int): string
    if n == 0
        return "0"
    end if
    mut value = n
    mut neg = false
    if n < 0
        value = 0 - n
        neg = true
    end if
    mut result = ""
    while value > 0
        let digit = value % 10
        result = str.concat(str.substring("0123456789", digit, 1), result)
        value = value / 10
    end while
    if neg
        result = str.concat("-", result)
    end if
    return result
end int_to_str