Caveat from my previous comment is closed. That mainline 7.1.12 test ran
without the
nvidia DKMS module (mainline is built with GCC 14 and passes
-fmin-function-alignment=16,
which gcc-13 on 24.04 rejects). Building the module with gcc-14 fixes that, and
7.1.12 has
now been re-tested twice with nvidia 580.173.02 loaded — dwells of 2 h 45 min
and 7 h 16 min,
both with zero flip_done timed out, Failed to bring PHY, 0c70 failed, port_clock
mismatch or powerdown state, and sub-second resumes. The earlier result was not
an
artifact of the missing module.
Correction — please disregard my "fails to hold the low-power state" paragraph.
I
compared /sys/power/suspend_stats/last_hw_sleep against wall-clock dwell and
concluded
that 7.0 does not stay in hardware sleep while 7.1.12 does. That was wrong:
last_hw_sleep
does not report S0ix residency correctly on this machine. Measuring one 7 h 16
min s2idle
dwell with every counter at once:
slp_s0_residency_usec delta 26050 s (99.55% of dwell)
S0i2.1 substate delta 26048 s (99.54%)
battery drain 1420 mWh = 0.195 W average
total_hw_sleep delta 280 s (1.07%)
The PMC counters and the battery agree the SoC slept deeply at a normal draw;
only
last_hw_sleep disagrees, by ~90x. Any residency figure I derived from it is
meaningless.
(For anyone measuring s2idle on these platforms, /sys/kernel/debug/pmc_core/
before/after
deltas are the reliable source.)
Nothing else in that comment depended on this metric, and the rest stands: the
Bspec 65101
backport onto 7.0.0-30 does not fix the resume stall (46 s, unchanged
signature, patch
verified present in the built binary), mainline 7.1.12 does, and the difference
is not in
intel_cx0_phy.c, whose message-bus path differs from 7.0 only by a rename.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2150605
Title:
`i915 Arrow Lake-S: PHY A / C10 DPLL state mismatch on resume from
long s2idle dwell — slow wake (5-10s) with retry storm`
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2150605/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
