Kernel net maintainer flags LLM patches in PTP time-card series
Jakub Kicinski says AI-generated submissions are hurting review capacity as a seven-patch ptp_ocp series draws process and technical pushback.
Linux networking maintainer Jakub Kicinski has called out a seven-patch series for the Open Compute Project time-card PTP driver as likely LLM-generated, writing that the influx of such patches has a hugely negative impact on the community.
The series, posted by Ahmad Byagowi for net-next, aims to describe board peripherals behind the FPGA I2C controller so the driver can instantiate R4006 and Time Card V9 topologies from software nodes. That includes a new binding and LED-class driver for the Lumissil IS32FL3207 controller, I2C-mux changes so software-node channel descriptions reach child adapters, independent tracking of board ID versus serial EEPROM data, and profile-driven mux topology with cleanup before the root adapter disappears.
Kicinski asked what an end user would actually do with the new LEDs, then followed with detailed AI-assisted reviews he judged plausible. Those notes questioned whether the IS32FL3207 binding can tell apart its unit-address and dash-suffix multicolor node forms, whether the driver rejects device trees the binding would accept, a possible unsafe firmware-node dereference in I2C adapter lookup during client teardown, and whether an EEPROM handling change that unblocks profile selection when only the board ID is present should carry a Fixes tag.
Separately, device-tree binding maintainer Krzysztof Kozlowski objected that a prior Reviewed-by tag had been dropped without changelog explanation, warning that reviewers asked to redo the same work may simply skip the series. Byagowi replied that v5 materially changed the binding (dash-suffixed multicolor groups, optional reg for outputs 16 and 17, and RISET/current-limit documentation), apologized for the omission, and said he would not carry the tag unless it is renewed.
The exchange lands in a subsystem already under heavy patch load. Core I2C and LED binding changes need acks beyond net-next, and maintainer patience for unexplained process slips and opaque machine-written churn is visibly thin.