Package: src:linux Version: 6.12.113-1 Severity: normal Tags: upstream Dear Maintainer,
Sierra Wireless EM7430 (USB 1199:9071) on a Panasonic Let's Note CF-SV8R11VS
loses its MBIM channel repeatedly during the same boot. ModemManager 1.24.0
reports 10 consecutive transaction timeouts on /dev/cdc-wdm0 (or ttyUSB2),
declares the modem invalid and removes it, which drops mobile data.
Environment:
Debian 13 (trixie), linux-image 6.12.113-1, ModemManager 1.24.0
Firmware SWI9X30C_02.33.03.00 (Panasonic OEM, PRI 002.072_001)
SIM: China Unicom, APN 3gnet, AT+CPIN? -> READY
Observed ModemManager messages:
<wrn> [modem0] ATDS signal query failed: Transaction timed out
<wrn> [modem0] port cdc-wdm0 timed out 2 consecutive times
<wrn> [modem0/bearer0] reloading stats failed: Transaction timed out
<err> [modem0] port cdc-wdm0 timed out 10 consecutive times, marking
modem as invalid
Three occurrences in one day: 08:27:59 (cdc-wdm0), 09:09:17 (ttyUSB2),
10:03:53 (cdc-wdm0). One followed suspend/resume, the others happened while
the machine stayed awake.
Ruled out:
- USB runtime autosuspend: power/control=on, runtime_status=active,
runtime_suspended_time=0 during the failures
- USB link loss: device stayed on the bus (lsusb visible, authorized=1),
/dev/cdc-wdm0 and /dev/ttyUSB* nodes present
- SIM/registration: AT+CPIN? returns READY, CREG/COPS fine
- GPS polling: A/B test (15 min GPS off vs 15 min GPS on at 2.5 Hz)
showed no difference in timeout counts
- Firmware: 02.33.03.00 is the newest Panasonic OEM build; the Sierra
generic 02.38 installer fails at the PRI step and rolls back
- Hardware: AT port always answers, USB re-enumeration restores the link
Suspicious kernel-side detail, printed at every boot:
usb 1-5: config 1 has an invalid interface number: 12 but max is 4
usb 1-5: config 1 has an invalid interface number: 13 but max is 4
usb 1-5: config 1 has no interface number 1
usb 1-5: config 1 has no interface number 4
bNumInterfaces is 5 but the module uses interface numbers 0/2/3/12/13
(0/2/3 via qcserial, 12/13 via cdc_mbim). Older kernels tolerated this.
With 6.12.113, after a USB re-enumeration cdc_mbim frequently fails to
bind 1.12/1.13 automatically; the interfaces stay driverless until
1-5:1.12 is written into /sys/bus/usb/drivers/cdc_mbim/bind by hand.
dmesg contains no -71/-110/over-current/reset entries for this device; the
only "device disconnected" lines come from the authorized 0/1 toggling we do
ourselves as a recovery step.
Reproducible by keeping the modem connected for a few hours while
ModemManager polls bearer statistics and ATDS every 30 seconds.
I can provide the full ModemManager log (with DEBUG), per-boot kernel USB
logs, lsusb -t output and a sysfs snapshot of the failure state if useful.
I am also able to test a different kernel (6.12.111 is installed) if that
helps narrow it down.
Regards
lte-evidence-small.tar.gz
Description: application/gzip
