contract.reef - zygaena/zyginit

root / src / contract.reef
contract.reef Reef 502 lines 15.0 KB
/******************************************************************************
                __               ____                __
               / /   ___  ____ _/ __/_____________ _/ /__
              / /   / _ \/ __ `/ /_/ ___/ ___/ __ `/ / _ \
             / /___/  __/ /_/ / __(__  ) /__/ /_/ / /  __/
            /_____/\___/\__,_/_/ /____/\___/\__,_/_/\___/

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

    Project: zyginit
   Filename: contract.reef
    Authors: Chris Tusa <chris.tusa@leafscale.com>
    License: <see LICENSE file included with this source code>
Description: Hammerhead process contract FFI bindings (libcontract)

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

// Hammerhead process contracts are the kernel mechanism for tracking service
// processes. This module provides Reef wrappers around libcontract.
//
// This module ONLY works on Hammerhead. On Linux, the extern "C"
// functions will fail at link time unless stub implementations are provided.
//
// Reference: Hammerhead libcontract(3LIB), contract(5), process(5)
// Source: hammerhead/usr/src/lib/libcontract/common/

module contract

import sys.fd as fd
import core.str

export
    type Contract

    // Contract event flags (from sys/contract/process.h)
    fn CT_PR_EV_EMPTY(): int
    fn CT_PR_EV_FORK(): int
    fn CT_PR_EV_EXIT(): int
    fn CT_PR_EV_CORE(): int
    fn CT_PR_EV_SIGNAL(): int
    fn CT_PR_EV_HWERR(): int
    fn CT_PR_ALLEVENT(): int
    fn CT_PR_ALLFATAL(): int

    // Contract parameter flags
    fn CT_PR_INHERIT(): int
    fn CT_PR_NOORPHAN(): int
    fn CT_PR_PGRPONLY(): int
    fn CT_PR_REGENT(): int

    // Status detail levels (from sys/contract.h)
    fn CTD_COMMON(): int
    fn CTD_FIXED(): int
    fn CTD_ALL(): int

    // sigsend id type
    fn P_CTID(): int

    // CTFS paths
    fn CTFS_ROOT(): string
    fn CTFS_PROCESS_TEMPLATE(): string
    fn CTFS_PROCESS_LATEST(): string
    fn CTFS_PROCESS_BUNDLE(): string

    // Contract handle constructor
    fn new_contract(id: int, service_name: string): Contract

    // Contract accessors
    fn contract_id(ct: Contract): int
    fn contract_service(ct: Contract): string
    fn contract_bundle_fd(ct: Contract): int

    // Contract lifecycle
    fn setup_template(): int
    fn get_latest_contract(): int
    fn clear_template(tmpl_fd: int): int
    fn open_bundle(): int
    fn kill_contract(contract_id: int, sig: int): int
    fn abandon_contract(contract_id: int): int
    fn adopt_contract(contract_id: int): int
    fn is_contract_empty(contract_id: int): bool

    // Event reading
    fn read_event(bundle_fd: int, out_ctid: [int], out_type: [int]): bool
    fn ack_event(contract_id: int, event_id: int): int
end export

// ============================================================================
// FFI declarations — libcontract
// ============================================================================

// Template management
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

// Process contract template
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

// Contract status — ct_stathdl_t is an opaque void*
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

// Process-contract member list — fills *members and *n. Memory is
// owned by the stathdl, freed by ct_status_free. Requires the status
// to have been read at CTD_FIXED detail or higher (CTD_COMMON does
// not include the member list).
extern "C" fn ct_pr_status_get_members(stathdl: pointer, members: [pointer], n: [int]): int

// Contract control
extern "C" fn ct_ctl_abandon(fd: int): int
extern "C" fn ct_ctl_adopt(fd: int): int

// Contract events — ct_evthdl_t is an opaque void*
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 ct_event_get_flags(evthdl: pointer): int

// Process contract event details
extern "C" fn ct_pr_event_get_exitstatus(evthdl: pointer, status: pointer): int
extern "C" fn ct_pr_event_get_pid(evthdl: pointer, pid: pointer): int

// POSIX — sigsend for contract-wide signaling
extern "C" fn sigsend(idtype: int, id: int, sig: int): int

// POSIX — open/close for ctfs paths
extern "C" fn open(path: string, flags: int): int
extern "C" fn close(fd: int): int

// ============================================================================
// Constants — values from Hammerhead kernel headers
// ============================================================================

// Contract event flags (sys/contract/process.h)
fn CT_PR_EV_EMPTY(): int
    return 1
end CT_PR_EV_EMPTY

fn CT_PR_EV_FORK(): int
    return 2
end CT_PR_EV_FORK

fn CT_PR_EV_EXIT(): int
    return 4
end CT_PR_EV_EXIT

fn CT_PR_EV_CORE(): int
    return 8
end CT_PR_EV_CORE

fn CT_PR_EV_SIGNAL(): int
    return 16
end CT_PR_EV_SIGNAL

fn CT_PR_EV_HWERR(): int
    return 32
end CT_PR_EV_HWERR

fn CT_PR_ALLEVENT(): int
    return 63
end CT_PR_ALLEVENT

fn CT_PR_ALLFATAL(): int
    return 56
end CT_PR_ALLFATAL

// Contract parameter flags (sys/contract/process.h)
fn CT_PR_INHERIT(): int
    return 1
end CT_PR_INHERIT

fn CT_PR_NOORPHAN(): int
    return 2
end CT_PR_NOORPHAN

fn CT_PR_PGRPONLY(): int
    return 4
end CT_PR_PGRPONLY

fn CT_PR_REGENT(): int
    return 8
end CT_PR_REGENT

// Status detail levels (sys/contract.h)
fn CTD_COMMON(): int
    return 0
end CTD_COMMON

fn CTD_FIXED(): int
    return 1
end CTD_FIXED

fn CTD_ALL(): int
    return 2
end CTD_ALL

// sigsend id type for contract (sys/procset.h enum position 13,
// between P_ZONEID=12 and P_CPUID=14). Verified on Hammerhead via
// utils/contract_probe.
fn P_CTID(): int
    return 13
end P_CTID

// CTFS paths
fn CTFS_ROOT(): string
    return "/system/contract"
end CTFS_ROOT

fn CTFS_PROCESS_TEMPLATE(): string
    return "/system/contract/process/template"
end CTFS_PROCESS_TEMPLATE

fn CTFS_PROCESS_LATEST(): string
    return "/system/contract/process/latest"
end CTFS_PROCESS_LATEST

fn CTFS_PROCESS_BUNDLE(): string
    return "/system/contract/process/bundle"
end CTFS_PROCESS_BUNDLE

// Open flags
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

// ============================================================================
// Contract type
// ============================================================================

type Contract = struct
    id: int
    service_name: string
    bundle_fd: int
end Contract

fn new_contract(id: int, service_name: string): Contract
    return Contract{
        id: id,
        service_name: service_name,
        bundle_fd: 0 - 1
    }
end new_contract

fn contract_id(ct: Contract): int
    return ct.id
end contract_id

fn contract_service(ct: Contract): string
    return ct.service_name
end contract_service

fn contract_bundle_fd(ct: Contract): int
    return ct.bundle_fd
end contract_bundle_fd

// ============================================================================
// Contract lifecycle functions
// ============================================================================

// Set up a contract template for forking a service.
// Configures: informative=EMPTY, critical=EMPTY|HWERR, fatal=HWERR,
// params=INHERIT|NOORPHAN. Returns template fd on success, -1 on error.
fn setup_template(): int
    let tmpl_fd = open(CTFS_PROCESS_TEMPLATE(), O_RDWR())
    if tmpl_fd < 0
        return 0 - 1
    end if

    // Want informative notification when contract becomes empty
    let rc1 = ct_tmpl_set_informative(tmpl_fd, CT_PR_EV_EMPTY())
    // Critical events: empty (all exited) and hardware error
    let rc2 = ct_tmpl_set_critical(tmpl_fd, CT_PR_EV_EMPTY() + CT_PR_EV_HWERR())
    // Fatal events: hardware error causes contract death
    let rc3 = ct_pr_tmpl_set_fatal(tmpl_fd, CT_PR_EV_HWERR())
    // Parameters: inherit contracts on fork, kill orphans on abandon
    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

    let rc5 = ct_tmpl_activate(tmpl_fd)
    if rc5 != 0
        close(tmpl_fd)
        return 0 - 1
    end if

    return tmpl_fd
end setup_template

// After fork(), retrieve the contract ID for the new child process.
// The kernel writes the latest contract to /system/contract/process/latest.
// Returns contract ID on success, -1 on error.
fn get_latest_contract(): int
    let latest_fd = open(CTFS_PROCESS_LATEST(), O_RDONLY())
    if latest_fd < 0
        return 0 - 1
    end if

    // ct_status_read writes a handle (void*) to an output parameter.
    // We pass a [pointer] array — in C, this is void**, exactly what ct_status_read expects.
    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

// Clear (deactivate) a contract template after fork.
// Returns 0 on success.
fn clear_template(tmpl_fd: int): int
    let rc = ct_tmpl_clear(tmpl_fd)
    close(tmpl_fd)
    return rc
end clear_template

// Open the contract bundle fd for poll()-based event monitoring.
// Returns bundle fd on success, -1 on error.
fn open_bundle(): int
    // O_NONBLOCK (0x80 on Hammerhead): ct_event_read on a blocking fd
    // BLOCKS waiting for events. Our main loop only enters
    // handle_contract_events when poll says the fd is readable, but
    // after we drain the events present at that moment, the next
    // ct_event_read call sees no event and would block indefinitely
    // — wedging the entire main loop, including reap_children for
    // SIGCHLD. With O_NONBLOCK, ct_event_read returns EAGAIN once
    // we've drained, read_event returns false, the while loop in
    // handle_contract_events exits cleanly.
    let bundle_fd = open(CTFS_PROCESS_BUNDLE(), O_RDONLY() + 0x80)
    return bundle_fd
end open_bundle

// Kill all processes in a contract with the given signal.
// Uses sigsend(P_CTID, contract_id, signal).
// Returns 0 on success, -1 on error.
fn kill_contract(contract_id: int, sig: int): int
    return sigsend(P_CTID(), contract_id, sig)
end kill_contract

// Abandon a contract (release ownership, orphan handling applies).
// Opens the contract's ctl fd and issues abandon.
// Returns 0 on success, -1 on error.
fn abandon_contract(contract_id: int): int
    // Build path: /system/contract/process/<id>/ctl
    // ctl file mode is --w--w--w-, so open with O_WRONLY (O_RDWR fails).
    let ctl_path = str.concat(str.concat("/system/contract/process/", int_to_str(contract_id)), "/ctl")
    let ctl_fd = open(ctl_path, O_WRONLY())
    if ctl_fd < 0
        return 0 - 1
    end if
    let rc = ct_ctl_abandon(ctl_fd)
    close(ctl_fd)
    return rc
end abandon_contract

// Adopt an existing contract (for zyginit restart recovery).
// Opens the contract's ctl fd and issues adopt.
// Returns 0 on success, -1 on error.
fn adopt_contract(contract_id: int): int
    let ctl_path = str.concat(str.concat("/system/contract/process/", int_to_str(contract_id)), "/ctl")
    let ctl_fd = open(ctl_path, O_WRONLY())
    if ctl_fd < 0
        return 0 - 1
    end if
    let rc = ct_ctl_adopt(ctl_fd)
    close(ctl_fd)
    return rc
end adopt_contract

// Check whether a contract has zero member processes. Used by the
// post-recovery idempotency step in main.apply_recovered_state — if a
// contract was reported in state.toml but the kernel says it's empty,
// the service exited during the exec gap and we should apply restart
// policy as if a contract-empty event arrived.
//
// Returns true if the contract has zero members OR if the status
// read fails (treat unknown as empty so the recovery path applies
// restart policy rather than leaving a phantom service in RUNNING).
fn is_contract_empty(contract_id: int): bool
    let status_path = str.concat(str.concat("/system/contract/process/", int_to_str(contract_id)), "/status")
    let st_fd = open(status_path, O_RDONLY())
    if st_fd < 0
        return true
    end if

    unsafe
        let stathdl_buf = new [pointer](1)
        // CTD_FIXED includes process-contract-specific fields like the
        // member-PID list, which CTD_COMMON omits.
        let rc = ct_status_read(st_fd, CTD_FIXED(), stathdl_buf)
        close(st_fd)

        if rc != 0
            return true
        end if

        let stathdl = stathdl_buf[0]
        let members_buf = new [pointer](1)
        let n_buf = new [int](1)
        let mrc = ct_pr_status_get_members(stathdl, members_buf, n_buf)
        let n = n_buf[0]
        ct_status_free(stathdl)

        if mrc != 0
            return true
        end if
        return n == 0
    end unsafe
end is_contract_empty

// ============================================================================
// Event reading
// ============================================================================

// Read an event from the contract bundle fd.
// Writes the contract ID to out_ctid[0] and event type to out_type[0].
// Returns true on success, false on error or no event.
fn read_event(bundle_fd: int, out_ctid: [int], out_type: [int]): bool
    unsafe
        let evthdl_buf = new [pointer](1)
        let rc = ct_event_read(bundle_fd, evthdl_buf)

        if rc != 0
            return false
        end if

        let evthdl = evthdl_buf[0]
        out_ctid[0] = ct_event_get_ctid(evthdl)
        out_type[0] = ct_event_get_type(evthdl)

        ct_event_free(evthdl)
        return true
    end unsafe
end read_event

// Acknowledge a contract event.
// Opens the contract's event fd and issues ack.
fn ack_event(contract_id: int, event_id: int): int
    let ev_path = str.concat(str.concat("/system/contract/process/", int_to_str(contract_id)), "/events")
    let ev_fd = open(ev_path, O_RDONLY())
    if ev_fd < 0
        return 0 - 1
    end if
    // For now, we just close — full ack requires ct_ctl_ack on the ctl fd
    close(ev_fd)
    return 0
end ack_event

// ============================================================================
// Helpers
// ============================================================================

// Convert int to string (for building ctfs paths)
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

end module