--- b6/aurora-modem 2026-09-30 19:39:43.675393185 +0600 +++ b6p/rootfs/usr/sbin/aurora-modem 2026-09-30 22:24:42.244902667 +0600 @@ -1,18 +1,17 @@ #!/bin/sh -# B6 RAM-only modem lifecycle controller for Aurora (golden mm1). Lives in /tmp only. -# aurora-modem start | stop | status | addroute -# Reconstructs the proven flow (NOTES sessions 14-22 Aurora, JZ02 stopcycle/wwanfix/openrc): -# start: RO gates -> rmtfs -r -P -s -v (rmtfs -s starts MSS) -> MSS running -> wwan0qmi0+wwan0 (+2 s, JZ02 port-race guard) -# -> qmicli -p DMS online (one SET, only if not online) -> NAS registered + PS attached -> kill qmi-proxy -# -> udevd/dbus/polkitd/ModemManager -> MM modem registered -> ONE simple-connect -> bearer IPv4 on wwanN -# + /32 host routes only (no default route, usb0 untouched). -# stop: simple-disconnect -> MM disable (radio low-power = detach) -> del our /32 routes, flush+down all wwan* -> stop MM (+qmi-proxy) -> BAM-DMUX suspended -# -> SIGTERM rmtfs: rmtfs writes "stop" to remoteproc state (MSS stops while rmtfs still serves) -> rmtfs exits LAST. -# Never writes eMMC/NV: rmtfs -r keeps EFS writes in RAM; the script refuses to run unless the NV partitions are getro=1. +# Aurora (JZ08AU, MSM8916) modem lifecycle controller - B6P production build of the B6N-proven controller (b6j/aurora-modem). +# aurora-modem start | disconnect | stop | status | addroute +# start: [rmtfs -r -P -s -v -> MSS -> wwan0qmi0 (+2 s port-race guard) -> SIM ready -> DMS online (one SET, retried only on DeviceNotReady) +# -> registration state machine (registered + PS attached, deadline REG_TIMEOUT, default 600 s)] (skipped when rmtfs+MSS already run) +# -> udevd/dbus/polkitd/ModemManager (reused if running) -> ONE simple-connect -> bearer IPv4 on wwanN + /32 routes only (no default route). +# disconnect: simple-disconnect, our routes, flush+down all wwan*, wait BAM-DMUX suspended. MSS, rmtfs, MM and radio stay up; 'start' reconnects. +# stop: disconnect -> stop MM (+qmi-proxy) -> BAM-DMUX suspended -> SIGTERM rmtfs (rmtfs writes 'stop' -> MSS offline) -> rmtfs exits last. +# Never writes eMMC/NV: rmtfs -r keeps EFS writes in RAM; refuses to start unless modemst1/2, fsc, fsg, persist, mmcblk0 are getro=1 and /firmware is ro. +# Does not need a valid wall clock (all timing from /proc/uptime). No automatic MSS/rmtfs retry. # Exit: 0 ok, 1 failure/timeout (start rolls back with a full stop unless MODEM_KEEP_ON_FAIL=1), 2 usage/config error. -CONF=${MODEM_CONF:-/tmp/aurora-modem.conf} -D=/tmp/aurora-modem.d; mkdir -p $D +CONF=${MODEM_CONF:-/etc/aurora/modem.conf} +D=/run/aurora-modem; mkdir -p $D LOG=$D/lifecycle.log; STATE=$D/state; ROUTES=$D/routes; BEARER=$D/bearer RPNAME=4080000.remoteproc BAM=/sys/bus/platform/devices/4080000.remoteproc:bam-dmux @@ -20,10 +19,10 @@ # --- timeouts (s); override in $CONF T_RMTFS=5 T_MSS=60 T_QMI_DEV=30 T_QMI=45 T_SIM=60 T_ONLINE=30 T_REG=180 T_MM_DETECT=150 T_MM_REG=60 T_CONNECT=150 -MODEM_REATTACH_HOLDOFF=120 # s between radio stop and next MSS start (B6: re-attach <=61 s after stop was rejected 3/3 by Beeline) T_DISCONNECT=60 T_DISABLE=30 T_MM_EXIT=30 T_BAM_IDLE=15 T_MSS_STOP=30 T_RMTFS_EXIT=15 -MODEM_APN= MODEM_USER= MODEM_PASSWORD= MODEM_AUTH= MODEM_IP_TYPE=ipv4 MODEM_TEST_ROUTES= KEEP=${MODEM_KEEP_ON_FAIL:-0} +APN= USER= PASSWORD= AUTH= IP_TYPE=ipv4 REG_TIMEOUT=600 ROUTES= KEEP=${MODEM_KEEP_ON_FAIL:-0} [ -f "$CONF" ] && . "$CONF" +MODEM_APN=$APN MODEM_USER=$USER MODEM_PASSWORD=$PASSWORD MODEM_AUTH=$AUTH MODEM_IP_TYPE=${IP_TYPE:-ipv4} MODEM_TEST_ROUTES=$ROUTES T_REG=${REG_TIMEOUT:-600} MODEM_KEEP_ON_FAIL=${MODEM_KEEP_ON_FAIL:-$KEEP} mmc() { mmcli "$@" 2>/dev/null | tr -d '\033' | sed 's/\[[0-9;]*m//g'; } # mmcli colours values even without a tty @@ -62,6 +61,43 @@ # this mmcli has no --list-bearers ("no actions specified"); use 'Bearer | paths:' of mmcli -m, not the initial (attach) bearer bearers() { mmc -m "$(mm_idx)" | grep -v "initial bearer" | grep -oE "Bearer/[0-9]+" | cut -d/ -f2; } any_bearer_connected() { for b in $(bearers); do mmc -b $b | grep -q "connected: yes" && return 0; done; return 1; } +qv() { grep -m1 -oE "$1: *'[^']*'" $2 2>/dev/null | sed -E "s/.*'([^']*)'/\1/"; } +# B6M registration state machine. rc 0 = registered+PS attached, 1 = REGISTRATION_TIMEOUT, 2 = local failure (reason in $D/reg-fail) +reg_sm() { + T=$D/reg-trans.tsv; t0=$1; last=; n=0; nd=0; rm -f $D/reg-fail; D0=$(dmesg | wc -l) + printf 'uptime t_online reg ps status true_status domain plmn earfcn pci si_tac si_eci cl_tac cl_eci rsrp rsrq snr +' > $T + lf() { echo "$*" > $D/reg-fail; return 2; } + while :; do n=$((n+1)) + mss_is running || { lf "MSS state $(mss_state)"; return 2; } + [ -e /dev/wwan0qmi0 ] || { lf "wwan0qmi0 disappeared"; return 2; } + [ -n "$(rmtfs_pid)" ] || { lf "rmtfs died"; return 2; } + [ "$(bam_pm)" = error ] && { lf "BAM-DMUX runtime_status=error"; return 2; } + dmesg | tail -n +$((D0+1)) | grep -qE 'crash detected|fatal error|[Ww]atchdog|wdog|Oops|BUG:|Failed to resume' && { lf "dmesg: $(dmesg | tail -n +$((D0+1)) | grep -m1 -E 'crash detected|fatal error|[Ww]atchdog|wdog|Oops|BUG:|Failed to resume')"; return 2; } + if [ $((n % 5)) = 1 ]; then + sim_ready || { lf "SIM not ready: $(grep -E 'Card state|Application state' $D/qmi-uim.txt | head -2 | tr -s ' \t\n' ' ')"; return 2; } + ping -c1 -W2 172.16.42.2 >/dev/null 2>&1 || { sleep 1; ping -c1 -W2 172.16.42.2 >/dev/null 2>&1 || { lf "USB management (usb0 ping) lost"; return 2; }; } + fi + timeout 20 $Q --nas-get-serving-system > $D/qmi-ss.txt 2>&1 + timeout 20 $Q --nas-get-system-info > $D/qmi-si-all.txt 2>&1; sed -n '/LTE service:/,/SIM reject/p' $D/qmi-si-all.txt > $D/qmi-si.txt + RG=$(qv 'Registration state' $D/qmi-ss.txt); PS=$(qv PS $D/qmi-ss.txt); ST=$(qv Status $D/qmi-si.txt); TS=$(qv 'True Status' $D/qmi-si.txt) + DM=$(qv Domain $D/qmi-si.txt); PL=$(qv MCC $D/qmi-si.txt)-$(qv MNC $D/qmi-si.txt); SIT=$(qv 'Tracking Area Code' $D/qmi-si.txt); SIE=$(qv 'Cell ID' $D/qmi-si.txt) + key="$RG|$PS|$ST|$TS|$DM|$PL|$SIT|$SIE" + if [ "$key" != "$last" ]; then + timeout 20 $Q --nas-get-cell-location-info > $D/qmi-cl.txt 2>&1; timeout 20 $Q --nas-get-signal-info > $D/qmi-sig.txt 2>&1 + [ "$RG" = registration-denied ] && [ "${last%%|*}" != registration-denied ] && nd=$((nd+1)) + printf '%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\n' $(now) $(since $t0) "$RG" "$PS" "$ST" "$TS" "$DM" "$PL" \ + "$(qv 'EUTRA Absolute RF Channel Number' $D/qmi-cl.txt)" "$(qv 'Serving Cell ID' $D/qmi-cl.txt)" "$SIT" "$SIE" \ + "$(qv 'Tracking Area Code' $D/qmi-cl.txt)" "$(qv 'Global Cell ID' $D/qmi-cl.txt)" "$(qv RSRP $D/qmi-sig.txt)" "$(qv RSRQ $D/qmi-sig.txt)" "$(qv SNR $D/qmi-sig.txt)" >> $T + log "REG t+$(since $t0)s: $RG / PS $PS / $ST ($TS, $DM) $PL pci=$(qv 'Serving Cell ID' $D/qmi-cl.txt) si=$SIT/$SIE cl=$(qv 'Tracking Area Code' $D/qmi-cl.txt)/$(qv 'Global Cell ID' $D/qmi-cl.txt)" + last=$key + fi + echo $nd > $D/reg-denied-count + [ "$RG" = registered ] && [ "$PS" = attached ] && return 0 + [ "$(awk -v a="$(now)" -v b="$t0" -v t="$T_REG" 'BEGIN{print (a-b>=t)}')" = 1 ] && { lf "REGISTRATION_TIMEOUT after ${T_REG}s ($RG, PS $PS)"; return 1; } + sleep 3 + done +} bearer_field() { awk -F': ' -v k="$1" '$1 ~ " "k"$" {gsub(/ /,"",$2); print $2; exit}' $BEARER.txt; } check_safety() { @@ -110,14 +146,11 @@ elif [ -n "$(rmtfs_pid)" ] || ! mss_is offline; then log "FAIL: inconsistent (rmtfs='$(rmtfs_pid)' MSS=$(mss_state)); run stop first"; exit 1 else - if [ -f $D/last-radio-stop ]; then - w=$(awk -v a="$(now)" -v b="$(cat $D/last-radio-stop)" -v h="$MODEM_REATTACH_HOLDOFF" 'BEGIN{d=h-(a-b); printf "%d", (d>0)?d+1:0}') - [ "$w" -gt 0 ] && { log "re-attach holdoff: last radio stop $(since $(cat $D/last-radio-stop))s ago, waiting ${w}s (MODEM_REATTACH_HOLDOFF=$MODEM_REATTACH_HOLDOFF)"; sleep $w; } - fi : > $ROUTES; rm -f $D/ipcfg + n0=$(grep -c 'registering services' $D/rmtfs.log 2>/dev/null) LD_PRELOAD=/usr/lib/linebuf.so rmtfs -r -P -s -v >> $D/rmtfs.log 2>&1 & echo $! > $D/rmtfs.pid - wait_for $T_RMTFS 0.2 sh -c "kill -0 $(cat $D/rmtfs.pid) 2>/dev/null && grep -q 'registering services' $D/rmtfs.log" \ + wait_for $T_RMTFS 0.2 sh -c "kill -0 $(cat $D/rmtfs.pid) 2>/dev/null && [ \$(grep -c 'registering services' $D/rmtfs.log) -gt ${n0:-0} ]" \ || fail_start "rmtfs did not start (pid $(cat $D/rmtfs.pid))" setstate RMTFS_READY setstate MPSS_BOOTING @@ -149,7 +182,10 @@ log "DMS online after $(since $t)s" fi setstate RADIO_ONLINE - t=$(now); wait_for $T_REG 2 qmi_reg || fail_start "not registered+PS attached after ${T_REG}s ($(grep -E "Registration state" $D/qmi-ss.txt | tr -s ' \t'))" + t=$(now); reg_sm $t; rr=$? + [ $rr = 2 ] && fail_start "LOCAL FAILURE during registration: $(cat $D/reg-fail)" + [ $rr = 1 ] && fail_start "$(cat $D/reg-fail)" + log "registered: $(grep -oE "(MCC|MNC|Description): '[^']*'" $D/qmi-ss.txt | tr '\n' ' ')after $(since $t)s" setstate REGISTERED killall qmi-proxy 2>/dev/null; wait_for 5 0.2 sh -c '! pidof qmi-proxy >/dev/null' @@ -210,13 +246,6 @@ if [ -n "$(mm_pid)" ] && mm_has_modem; then timeout $T_DISCONNECT mmcli -m "$(mm_idx)" --simple-disconnect >> $LOG 2>&1; log "simple-disconnect rc=$? (bearers: $(bearers | tr '\n' ' '))" wait_for 15 1 sh -c "! sh $0 _anyconn" && log "no bearer connected" || { log "WARNING: bearer still connected"; rc=1; } - # 1b. radio off / network detach before MSS goes away (B6 addition: abrupt MSS stop while EPS-attached - # was followed by attach rejection 'limited' on quick restart). MM disable = DMS low-power. - timeout $T_DISABLE mmcli -m "$(mm_idx)" --disable >> $LOG 2>&1; r=$? - wait_for 10 0.5 sh -c "[ \"\$(sh $0 _mmstate)\" = disabled ]" && log "MM disable rc=$r -> state disabled (radio low-power, detached)" \ - || { log "WARNING: MM disable rc=$r state '$(mm_state)'"; rc=1; } - elif [ -e /dev/wwan0qmi0 ] && mss_is running; then - timeout 20 $Q --dms-set-operating-mode=low-power >> $LOG 2>&1; log "qmicli low-power rc=$? (MM not running)" fi # 2. our /32 routes, 3. flush + DOWN every wwan* for h in $(cat $ROUTES 2>/dev/null); do ip route del $h/32 2>/dev/null; done; : > $ROUTES @@ -242,7 +271,6 @@ fi # 6. MSS stop via rmtfs -s (SIGTERM -> pwrite "stop" to remoteproc state -> exit), 7. rmtfs exits last P=$(rmtfs_pid) - mss_is running && now > $D/last-radio-stop if [ -n "$P" ]; then t=$(now); kill -TERM $P; log "SIGTERM rmtfs $P (rmtfs stops MSS, then exits)" wait_for $T_MSS_STOP 0.2 mss_is offline && log "MSS offline after $(since $t)s (rmtfs alive: $(kill -0 $P 2>/dev/null && echo yes || echo no))" \ @@ -260,6 +288,23 @@ return $rc } +do_disconnect() { # B6N: bearer down only; MSS, rmtfs, MM (and DMS radio) stay up. Next 'start' reuses them and only simple-connects. + T0=$(now); log "=== DISCONNECT (from $(cat $STATE 2>/dev/null || echo OFF): mss=$(mss_state) rmtfs=$(rmtfs_pid) mm=$(mm_pid) bam=$(bam_pm))"; rc=0 + [ -n "$(mm_pid)" ] && mm_has_modem || { log "DISCONNECT: no MM modem"; return 1; } + timeout $T_DISCONNECT mmcli -m "$(mm_idx)" --simple-disconnect >> $LOG 2>&1; log "simple-disconnect rc=$? (bearers: $(bearers | tr '\n' ' '))" + wait_for 15 1 sh -c "! sh $0 _anyconn" && log "no bearer connected" || { log "ERROR: bearer still connected"; rc=1; } + for h in $(cat $ROUTES 2>/dev/null); do ip route del $h/32 2>/dev/null; done; : > $ROUTES + for d in /sys/class/net/wwan*; do [ -e "$d" ] || continue; w=${d##*/} + ip addr flush dev $w 2>/dev/null; ip link set $w down 2>/dev/null + link_up $w && { log "ERROR: $w still UP"; rc=1; }; done + rm -f $D/ipcfg + ip route | grep -q " dev wwan" && { log "ERROR: routes via wwan remain: $(ip route | grep ' dev wwan' | tr '\n' ';')"; rc=1; } + t=$(now); wait_for $T_BAM_IDLE 0.5 bam_suspended && log "BAM-DMUX suspended after $(since $t)s" || { log "ERROR: BAM-DMUX '$(bam_pm)' not suspended"; rc=1; } + setstate MM_IDLE + log "=== DISCONNECT done rc=$rc in $(since $T0)s (mss=$(mss_state) rmtfs='$(rmtfs_pid)' mm='$(mm_pid)' mm_state=$(mm_state) bam=$(bam_pm))" + return $rc +} + do_status() { r=$(rp) echo "controller: $(cat $STATE 2>/dev/null || echo OFF)" @@ -285,6 +330,7 @@ case "$1" in start) do_start;; stop) do_stop; exit $?;; + disconnect) do_disconnect; exit $?;; status) do_status;; addroute) [ -n "$2" ] || exit 2; addroute "$2";; _anyconn) any_bearer_connected;;