j613 (MacBook Air 13" M3): lid-closed s2idle on battery drains 30% → 0% overnight (≥1.35 W), no hibernation or RTC wake fallback #403

Mac and installed build

• Mac: MacBook Air (13-inch, M3, 2024), apple,j613 / apple,t8122
• Installed: Omarchy 4.5.0-3, omarchy-mac 0.2.3-2, linux-aurora 7.2.11.aurora14-1 (uname -r: 7.2.11-14-1-ARCH)
• Boot chain: m1n1-aurora 1.6.1.aurora22-1, uboot-aurora 2026.07.aurora6-1
• Distribution: Arch Linux ARM, btrfs on LUKS2, zram swap only. No local kernel patches.
• systemd 262.1-1, Hyprland 0.56.2-7, mesa-aurora 26.3.0_devel.20261031

Both firmware versions and provisioning

asahi,os-fw-version: 14.7
asahi,system-fw-version: 27.0

Reproduction, expected result, and actual result

To reproduce, close the lid on battery with nothing attached. logind suspends to s2idle, which is the only mode offered (/sys/power/mem_sleep: [s2idle]), and the machine is left closed overnight.

Expected: a closed Air loses a few percent overnight, as it does under macOS.

Actual: the battery runs completely flat. The machine never resumes, and the next start is a cold boot at 0%.

Timeline from the journal and UPower history (GetHistory charge):

So about 30% of a 41.0 Wh pack (~12.3 Wh) was drained in at most 9.1 h. That puts the s2idle draw at ≥ 1.35 W (~3.3 %/h). It is a lower bound, because the time the battery hit empty is unknown.

That is lower than the T600x/T6030 figures in #290 (~1.9–2.0 W) and #291 (~2.65 W), but still more than 10× the ~0.1 W macOS figure in #290. As far as I can find, this is the first sleep-drain data point for J613.

Other details:

• This is not a failed suspend. Two short s2idle cycles a few minutes earlier in the same boot resumed correctly on lid open. The final cycle entered cleanly, and nothing is logged after PM: suspend entry (s2idle).
• The RTC survived. On the next boot macsmc-rtc set the correct wall-clock time (setting system clock to 2026-11-07T12:01:45 UTC). That differs from #779, where the RTC reset to 1970/1976.
• The gauge read 0% right after boot (energy_now 0, voltage_now ~11.50 V while charging at ~21 W). It recovered to normal readings after a short charge.
• Battery: bq40z651, energy_full 41.0 Wh of 52.0 Wh design, 308 cycles. The charge thresholds (75/80) are configured.
• No fallback is available while asleep on this machine either. /sys/power/state is freeze mem with no disk, so there is no hibernation. /sys/class/rtc/rtc0 (macsmc-rtc) has no wakealarm, so there is no timed wake to power off. iw phy phy0 wowlan show returns -95. As a result, the stock lid-close behaviour on battery is "suspend and drain to zero".

Graphics or session details (if relevant)

The display is on DCP (apple-t8122-display: enabled ... /soc/dcp@28ec00000). Hyprland 0.56.2 runs with Omarchy's lid-close lock. The two short resumes before the final sleep came back to the locked session without a compositor restart.

Relevant error lines

2026-11-07T03:53:56+01:00 systemd-logind[835]: Lid closed.
2026-11-07T03:53:56+01:00 systemd-logind[835]: Suspending...
2026-11-07T03:53:57+01:00 fprintd[8701]: Failed to install a sleep delay inhibitor: GDBus.Error:org.freedesktop.login1.OperationInProgress: The operation inhibition has been requested for is already running
2026-11-07T03:53:57+01:00 systemd-sleep[8968]: Performing sleep operation 'suspend'...
2026-11-07T03:53:57+01:00 kernel: PM: suspend entry (s2idle)
-- (end of boot; next boot starts 13:01:45) --
2026-11-07T13:01:45+01:00 kernel: macsmc-rtc macsmc-rtc: setting system clock to 2026-11-07T12:01:45 UTC (1794052905)

The previous boot logs no Battery critical line from macsmc-power.

Related issues, pull requests, or code examples

• j314s (MacBook Pro M1 Pro): s2idle draws ~1.9–2.0 W; measurements and what's ruled out #290: J314s (M1 Pro) s2idle ~1.9–2.0 W, with a breakdown of the PMGR domains that stay powered (display coprocessors, always-on SoC blocks).
• j514s (MacBook Pro 14" M3 Pro): s2idle draws ~2.65 W overnight, with no hibernation or RTC wake to fall back on #291: J514s (M3 Pro) s2idle ~2.65 W, with no hibernation or RTC wake.
• Apple Silicon: closed lid on battery drains to 0% in s2idle (~3 W), no hibernation or timed wake to stop it omarchy-mac#779: closed lid on battery drains to 0% (~3 W, M1 Max). That issue also proposes a battery-aware lid policy.
• t600x: PMU RTC wake alarm (SPMI notify at 192, Maverick event clear, DT) #351: PMU RTC wake alarm, T600x only. Something equivalent for T8122 would allow a suspend → timed wake → poweroff fallback here.

Workaround: power off instead of suspending when leaving the machine closed on battery.

What was not tested

• Only one overnight measurement, taken from UPower history (1% granularity). There are no charge_counter short runs yet.
• The exact time the battery reached empty is unknown, so 1.35 W is a lower bound.
• No per-domain (pmgr-pwrstate), cpuidle s2idle counter, or interrupt breakdown during sleep. The counters were lost with the dead boot.
• No comparison with radios soft-blocked or USB-C wakeup disabled.
• No macOS sleep measurement on this unit.
• No external display, dock or USB devices were attached. The Bluetooth mouse (MX Master 3S) was paired.

I can run short charge_counter lid-closed measurements, or test a CI kernel package, on this J613 if that helps.