Files
meeting-assistant/tools/ci/macos-native-readiness.sh
T

363 lines
15 KiB
Bash

#!/bin/bash
# Existing macOS Recovery/launchd runtime hook; never installs or erases anything.
set -u
PATH="/usr/bin:/bin:/usr/sbin:/sbin"
export PATH
PROOF_TOKEN="@@PROOF_TOKEN@@"
STATE_DIR="/Volumes/installstate"
PROOF_LOG="$STATE_DIR/proof.log"
RESULT="$STATE_DIR/result.json"
EXPECTED_BYTES=68719476736
MAX_LOG_BYTES=4194304
MAX_OUTPUT_BYTES=524288
TIMER_FIFO="/tmp/native-diagnostic-$PROOF_TOKEN-$$.fifo"
PENDING_OUTPUTS=()
ACTIVE_COMMAND=""
ACTIVE_TIMER=""
# BEGIN disk IPC diagnostic
ACTIVE_OBSERVER=""
# END disk IPC diagnostic
os_version=""
architecture=""
uid=-1
system_exit=-1
arbitration_exit=-1
recovery_exit=-1
disk_list_exit=-1
selected_disk=""
disk_bytes=0
count=0
while [ ! -d "$STATE_DIR" ] && (( count < 120 )); do
/sbin/mount_9p installstate >/dev/null 2>&1 || :
count=$((count + 1))
sleep 1
done
[ -d "$STATE_DIR" ] || exit 1
: > "$PROOF_LOG" || exit 1
exec 3>> "$PROOF_LOG" || exit 1
rm -f "$RESULT" "$RESULT.tmp"
printf '[proof-token] %s\n' "$PROOF_TOKEN" >&3
finish() {
local success="$1" reason="$2"
flush_outputs || { success=false; reason=diagnostic_log_budget_exceeded; }
printf '[proof-result] %s: %s\n' "$success" "$reason" >&3
printf '{"token":"%s","success":%s,"reason":"%s","osVersion":"%s","architecture":"%s","uid":%s,"systemExit":%s,"diskArbitrationExit":%s,"recoveryExit":%s,"diskListExit":%s,"disk":"%s","diskBytes":%s,"readOnly":false}\n' \
"$PROOF_TOKEN" "$success" "$reason" "$os_version" "$architecture" "$uid" \
"$system_exit" "$arbitration_exit" "$recovery_exit" "$disk_list_exit" \
"$selected_disk" "$disk_bytes" > "$RESULT.tmp"
/bin/mv -f "$RESULT.tmp" "$RESULT" || exit 1
exec 9>&-
[ ! -p "$TIMER_FIFO" ] || /bin/rm -f "$TIMER_FIFO"
# Keep the service alive for the bounded host diagnostic to capture evidence.
while :; do sleep 60; done
}
init_timer_fifo() {
# Recovery has Bash 3.2 before any SDK is installed. Its read timeout uses
# alarm(), avoiding a separate sleep process for every command and grace period.
[ ! -e "$TIMER_FIFO" ] || exit 1
/usr/bin/mkfifo -m 600 "$TIMER_FIFO" || exit 1
exec 9<> "$TIMER_FIFO" || exit 1
}
flush_outputs() {
(( ${#PENDING_OUTPUTS[@]} > 0 )) || return 0
local started=$SECONDS sizes="/tmp/native-diagnostic-$$.sizes" proof_size output_size raw_size
local raw_count=0 raw_valid=1 pending_count=${#PENDING_OUTPUTS[@]}
local bounded="/tmp/native-diagnostic-$$.flush"
# One bounded native copy per group, rather than tail/stat startup per command.
# Keep native byte-oriented copying: Bash 3.2 read -n would read large outputs
# one byte per system call. Small scalar reads below have a separate tight bound.
/usr/bin/tail -c "$MAX_OUTPUT_BYTES" "${PENDING_OUTPUTS[@]}" > "$bounded" || return 1
/usr/bin/stat -f '%z' "$PROOF_LOG" "$bounded" "${PENDING_OUTPUTS[@]}" > "$sizes" || return 1
{
IFS= read -r proof_size; IFS= read -r output_size
while IFS= read -r raw_size; do
raw_count=$((raw_count + 1))
[[ "$raw_size" =~ ^[0-9]+$ ]] && (( raw_size <= MAX_OUTPUT_BYTES )) || raw_valid=0
done
} < "$sizes"
PENDING_OUTPUTS=()
[[ "$proof_size" =~ ^[0-9]+$ && "$output_size" =~ ^[0-9]+$ ]] || return 1
(( raw_valid == 1 && raw_count == pending_count )) || return 1
(( proof_size + output_size + 1024 <= MAX_LOG_BYTES )) || return 1
/bin/cat "$bounded" >&3 || return 1
printf '\n[proof-flush] outputs-bytes=%s elapsed=%ss\n' "$output_size" "$((SECONDS - started))" >&3
}
read_scalar() {
local value status
# All three values are short native machine/uid/version scalars. Reject excess
# content instead of accepting a truncated first line as a successful gate.
IFS= read -r -n 65 -d '' value < "$LAST_OUTPUT"; status=$?
# EOF is mandatory: the byte bound or a NUL delimiter must never hide a suffix.
(( status == 1 && ${#value} < 65 )) || return 1
value=${value%$'\n'}
[[ "$value" != *$'\n'* ]] || return 1
SCALAR="$value"
}
# BEGIN successful sw_vers version parser
read_product_version() {
local value status line version="" fields=0
# Read the entire successful native output. EOF is mandatory; a NUL delimiter
# or reaching the 1025-byte sentinel must never hide a suffix.
LC_ALL=C IFS= read -r -n 1025 -d '' value < "$LAST_OUTPUT"; status=$?
(( status == 1 && ${#value} <= 1024 )) || return 1
while IFS= read -r line || [ -n "$line" ]; do
if [[ "$line" =~ ^[[:blank:]]*ProductVersion: ]]; then
fields=$((fields + 1))
(( fields == 1 )) || return 1
[[ "$line" =~ ^[[:blank:]]*ProductVersion:[[:blank:]]*([0-9]+\.[0-9]+(\.[0-9]+)?)[[:blank:]]*$ ]] || return 1
version="${BASH_REMATCH[1]}"
fi
done <<< "$value"
(( fields == 1 )) || return 1
SCALAR="$version"
}
# END successful sw_vers version parser
# BEGIN disk IPC diagnostic
# Optional observations have their own child/timer ownership. No service is
# loaded, restarted or changed, and samples target only this probe's diskutil.
observe_disk_query() {
local disk_process="$1" output="$2" sample_output="$3" observation_child="" observation_timer=""
cancel_observation() {
trap '' TERM INT
if [ -n "$observation_child" ]; then
kill -TERM "$observation_child" 2>/dev/null || :
IFS= read -r -t 2 -u 9 unused || :
kill -KILL "$observation_child" 2>/dev/null || :
wait "$observation_child" 2>/dev/null || :
fi
[ -z "$observation_timer" ] || { kill -TERM "$observation_timer" 2>/dev/null || :; wait "$observation_timer" 2>/dev/null || :; }
printf '[disk-observation] stopped after the owned disk query\n' >> "$output"
exit 143
}
observe_command() {
local name="$1" status started=$SECONDS
shift
printf '\n[disk-observation-command] %s:' "$name" >> "$output"
printf ' %s' "$@" >> "$output"
printf '\n' >> "$output"
"$@" >> "$output" 2>&1 &
observation_child=$!
(
trap 'exit 0' TERM INT
IFS= read -r -t 60 -u 9 unused || :
printf '[disk-observation-timeout] %s child=%s limit=60s\n' "$name" "$observation_child" >> "$output"
kill -TERM "$observation_child" 2>/dev/null || :
IFS= read -r -t 2 -u 9 unused || :
kill -KILL "$observation_child" 2>/dev/null || :
) &
observation_timer=$!
wait "$observation_child"; status=$?
kill -TERM "$observation_timer" 2>/dev/null || :
wait "$observation_timer" 2>/dev/null || :
printf '[disk-observation-exit] %s status=%s elapsed=%ss\n' "$name" "$status" "$((SECONDS - started))" >> "$output"
observation_child=""; observation_timer=""
}
trap cancel_observation TERM INT
printf '[disk-observation] owned-diskutil-child=%s parent-shell=%s\n' "$disk_process" "$$" >> "$output"
if [ -x /usr/bin/sample ]; then
if kill -0 "$disk_process" 2>/dev/null; then
observe_command diskutil-sample /usr/bin/sample "$disk_process" 3 100 -file "$sample_output"
else
printf '[disk-observation-unavailable] diskutil already exited before sample\n' >> "$output"
fi
else
printf '[disk-observation-unavailable] /usr/bin/sample is unavailable\n' >> "$output"
fi
}
stop_disk_observation() {
[ -n "$ACTIVE_OBSERVER" ] || return 0
kill -TERM "$ACTIVE_OBSERVER" 2>/dev/null || :
wait "$ACTIVE_OBSERVER" 2>/dev/null || :
ACTIVE_OBSERVER=""
}
# END disk IPC diagnostic
cancel_probe() {
trap '' TERM INT
# BEGIN disk IPC diagnostic
stop_disk_observation
# END disk IPC diagnostic
if [ -n "$ACTIVE_COMMAND" ]; then
kill -TERM "$ACTIVE_COMMAND" 2>/dev/null || :
IFS= read -r -t 2 -u 9 unused || :
kill -KILL "$ACTIVE_COMMAND" 2>/dev/null || :
wait "$ACTIVE_COMMAND" 2>/dev/null || :
fi
[ -z "$ACTIVE_TIMER" ] || { kill -TERM "$ACTIVE_TIMER" 2>/dev/null || :; wait "$ACTIVE_TIMER" 2>/dev/null || :; }
ACTIVE_COMMAND=""; ACTIVE_TIMER=""
finish false probe_cancelled
}
run_command() {
local name="$1"
shift
local process timer exit_code started waited command_limit=45
# Run 4161: even native uname/ps startup took 34-42s under TCG.
# Isolate only the failed UID gate; every other watchdog remains unchanged.
[[ "$name" != uid ]] || command_limit=180
# BEGIN disk IPC diagnostic
if [[ "$name" == disks && "${attempt:-0}" == 1 ]]; then command_limit=120; fi
# END disk IPC diagnostic
LAST_OUTPUT="/tmp/native-diagnostic-$name.out"
printf '\n[proof-command] %s:' "$name" >&3
printf ' %s' "$@" >&3
printf '\n' >&3
started=$SECONDS
"$@" > "$LAST_OUTPUT" 2>&1 &
process=$!
ACTIVE_COMMAND="$process"
printf '[proof-start] %s child=%s shell=%s parent=%s seconds=%s\n' "$name" "$process" "$$" "$PPID" "$started" >&3
printf '[proof-limit] %s %ss\n' "$name" "$command_limit" >&3
(
trap 'exit 0' TERM INT
IFS= read -r -t "$command_limit" -u 9 unused || :
printf '[proof-timeout] %s child=%s elapsed=%ss signal=TERM\n' "$name" "$process" "$((SECONDS - started))" >&3
kill -TERM "$process" 2>/dev/null || :
IFS= read -r -t 2 -u 9 unused || :
kill -KILL "$process" 2>/dev/null || :
) &
timer=$!
ACTIVE_TIMER="$timer"
# BEGIN disk IPC diagnostic
if [[ "$name" == disks && "$attempt" == 1 ]]; then
local observation_output="/tmp/native-diagnostic-disk-observation.out"
local sample_output="/tmp/native-diagnostic-disk-sample.out"
: > "$observation_output"
: > "$sample_output"
observe_disk_query "$process" "$observation_output" "$sample_output" &
ACTIVE_OBSERVER=$!
printf '[disk-observation-start] observer=%s owned-diskutil-child=%s\n' "$ACTIVE_OBSERVER" "$process" >&3
PENDING_OUTPUTS+=("$observation_output" "$sample_output")
fi
# END disk IPC diagnostic
wait "$process"
exit_code=$?
waited=$SECONDS
# Includes fork/exec/wait, but excludes timer cleanup and evidence copying.
printf '[proof-native-wait] %s child=%s elapsed=%ss exit=%s\n' "$name" "$process" "$((waited - started))" "$exit_code" >&3
kill -TERM "$timer" 2>/dev/null || :
wait "$timer" 2>/dev/null || :
# BEGIN disk IPC diagnostic
stop_disk_observation
# END disk IPC diagnostic
ACTIVE_COMMAND=""; ACTIVE_TIMER=""
printf '[proof-cleanup] %s child=%s elapsed=%ss total=%ss\n' "$name" "$process" "$((SECONDS - waited))" "$((SECONDS - started))" >&3
printf '[proof-exit] %s\n' "$exit_code" >&3
LAST_EXIT="$exit_code"
PENDING_OUTPUTS+=("$LAST_OUTPUT")
return 0
}
diagnose_failure() {
run_command kernel /usr/bin/uname -a
run_command account /usr/bin/id
run_command context /usr/sbin/sysctl kern.bootargs machdep.cpu.brand_string machdep.cpu.features machdep.cpu.leaf7_features
run_command parent /bin/ps -p "$$" -p "$PPID" -o pid=,ppid=,comm=
run_command processes /bin/ps -axo pid,ppid,comm
}
fail_probe() {
local reason="$1"
flush_outputs || finish false diagnostic_log_budget_exceeded
diagnose_failure
finish false "$reason"
}
init_timer_fifo
trap cancel_probe TERM INT
# Test the required native gates before optional process/CPU diagnostics.
run_command architecture /usr/bin/uname -m
(( LAST_EXIT == 0 )) || fail_probe architecture_probe_failed
read_scalar || fail_probe architecture_output_invalid
architecture="$SCALAR"
[ "$architecture" = x86_64 ] || fail_probe unexpected_guest_architecture
run_command uid /usr/bin/id -u
(( LAST_EXIT == 0 )) || fail_probe uid_probe_failed
read_scalar || fail_probe uid_output_invalid
uid="$SCALAR"
[ "$uid" = 0 ] || fail_probe recovery_account_not_root
run_command platform /usr/bin/sw_vers
platform_exit="$LAST_EXIT"
flush_outputs || finish false diagnostic_log_budget_exceeded
if (( platform_exit != 0 )); then
run_command system /bin/launchctl print system
system_exit="$LAST_EXIT"
run_command arbitration /bin/launchctl print system/com.apple.diskarbitrationd
arbitration_exit="$LAST_EXIT"
run_command recovery /bin/launchctl print system/com.apple.recoveryosd
recovery_exit="$LAST_EXIT"
printf '[proof-retry] sw_vers once after native service context; same 45-second deadline\n' >&3
run_command platform-warm /usr/bin/sw_vers
platform_exit="$LAST_EXIT"
fi
(( platform_exit == 0 )) || fail_probe sw_vers_failed
# BEGIN successful sw_vers version extraction
read_product_version || fail_probe product_version_invalid
# END successful sw_vers version extraction
os_version="$SCALAR"
[[ "$os_version" =~ ^[0-9]+\.[0-9]+(\.[0-9]+)?$ ]] || fail_probe product_version_invalid
(( ${os_version%%.*} >= 13 )) || fail_probe unsupported_macos_version
flush_outputs || finish false diagnostic_log_budget_exceeded
# BEGIN disk IPC diagnostic
run_command arbitration_before /bin/launchctl print system/com.apple.diskarbitrationd
run_command management_before /bin/launchctl print system/com.apple.diskmanagementd
run_command media_before /usr/sbin/ioreg -r -c IOMedia -l -w 0
flush_outputs || finish false diagnostic_log_budget_exceeded
# END disk IPC diagnostic
# Bound readiness independently of the host's 40-minute overall deadline.
readiness_start=$SECONDS
attempt=0
while (( attempt < 1 && SECONDS - readiness_start < 600 )); do
attempt=$((attempt + 1))
printf '\n[readiness-attempt] %s\n' "$attempt" >&3
run_command disks /usr/sbin/diskutil list physical
disk_list_exit="$LAST_EXIT"
if (( disk_list_exit == 0 )); then
disk_list=$(cat "$LAST_OUTPUT")
candidates=0
while IFS= read -r disk; do
[ -n "$disk" ] || continue
run_command "info-$disk" /usr/sbin/diskutil info "/dev/$disk"
(( LAST_EXIT == 0 )) || continue
info=$(cat "$LAST_OUTPUT")
if printf '%s\n' "$info" | grep -Eq '^[[:space:]]*(Read-Only (Media|Device)|(Media|Device) Read-Only):[[:space:]]*Yes'; then
continue
fi
printf '%s\n' "$info" | grep -Eq '^[[:space:]]*(Read-Only (Media|Device)|(Media|Device) Read-Only):[[:space:]]*No' || continue
size=$(printf '%s\n' "$info" | sed -nE 's/^[[:space:]]*Disk Size:.*\(([0-9]+) Bytes\).*/\1/p' | head -n 1)
[[ "$size" =~ ^[0-9]+$ ]] || continue
(( size == EXPECTED_BYTES )) || continue
candidates=$((candidates + 1))
selected_disk="/dev/$disk"
disk_bytes="$size"
printf '[writable-target] %s %s bytes\n' "$selected_disk" "$disk_bytes" >&3
done < <(printf '%s\n' "$disk_list" | sed -nE 's#^/dev/(disk[0-9]+).*#\1#p')
(( candidates <= 1 )) || fail_probe ambiguous_writable_64g_disks
if (( candidates == 1 )); then
# Re-probe live launchd domains after disk readiness, preserving native exits.
run_command system_ready /bin/launchctl print system
system_exit="$LAST_EXIT"
run_command arbitration_ready /bin/launchctl print system/com.apple.diskarbitrationd
arbitration_exit="$LAST_EXIT"
run_command recovery_ready /bin/launchctl print system/com.apple.recoveryosd
recovery_exit="$LAST_EXIT"
(( system_exit == 0 && arbitration_exit == 0 && recovery_exit == 0 )) || fail_probe service_domain_not_ready
finish true native_recovery_and_writable_64g_disk_ready
fi
fi
flush_outputs || finish false diagnostic_log_budget_exceeded
IFS= read -r -t 5 -u 9 unused || :
done
fail_probe disk_management_or_writable_target_not_ready