Repository navigation
Conversation
During EPR entry/exit the EC intentionally puts the system into a transitional clamp state: charger input current is collapsed to 0mA (learn mode on) and a hard AP throttle is asserted via update_pmf_events(BIT(PD_PROGRESS_ENTER/EXIT_EPR_MODE), 1). On a healthy transition this lasts ~1.1s and both are released after the new contract lands. When the transition fails, two pieces of that transitional state currently persist beyond the transition: 1. Neither failure handler releases the PMF transition event. The normal release path, clear_prochot(), only clears the ENTER event when cypd_get_ac_power() > 100W, which is never true after a failed EPR entry (the standing SPR contract is at most 100W), so the hard throttle outlives the transition it was meant to cover and is left to the 3s force_clear_pmf_prochot backstop at best. Release the ENTER/EXIT transition events at both points where the flow is abandoned: the EPR_MODE_ENTER_FAILED/default branch of cypd_update_epr_state(), and the retry-exhausted branch of epr_flow_pending_deferred() (the path taken when the EPR event from the PD controller never arrives). The failure handler now leaves the system in the plain-SPR state (input limit is already restored by the existing cypd_update_port_state() call on both paths) instead of the mid-transition state. 2. The "EPR fail, do not retry" latch (epr_active = 0xff) is only cleared by clear_port_state() on a PORT_DISCONNECT/hard-reset response. cypd_update_port_state() deliberately preserves it, and a CCG_RESPONSE_PORT_CONNECT does not reset it, so if the disconnect response is missed (the same event-loss class the deferred retry exists for) or EC RAM outlives a PD controller reset, the latch outlives the connection that caused it and the port can never enter EPR again. Clear the latch on PORT_CONNECT: a fresh attach is a new connection, so the do-not-retry intent (no EPR retry on the contract that failed) is preserved while the latch can no longer outlive that connection. Field evidence (FW16 Ryzen AI 300 "tulip", RTX 5070, 240W EPR, EC tulip-4.0.1): instrumented captures show the system latching into the transitional state - charger input stuck at ~0/500-800mA, hard throttle asserted (BIOS runs slow, survives OS reboot), and a port stuck attach-without-contract where EC_CMD_GET_PD_PORT_STATE returns zeros and the EC reports "AC not connected". Resets of either or both PD controllers do not clear it (EC RAM state re-imposes it after every reset); only a physical AC removal recovers, because only the PD_PROGRESS_DISCONNECTED path clears the progress/event state. A healthy reference trace of the same transition (throttle on, input 500mA, ~1.1s, then release) was captured for comparison. Issue: tyeth/framework-system#1 Telemetry/data: https://github.com/tyeth/fw16-gpu-watchdog Related: FrameworkComputer/SoftwareFirmwareIssueTracker #160 #193 #241 #243 Compile-untested: no Zephyr build environment was available; changes were limited to call sites and symbols already used in these files and verified by read-through. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…oned On boards with the RAA489300 3-level buck (tulip), enter_epr_mode() / exit_epr_mode() first collapse the charger input current to 0mA (learn mode on) and assert the PMF transition throttle, then hand off to buck_ready_for_epr() to sequence the buck mode change before the CCG EPR command is sent. If the buck readiness check exhausts its retries, the handler reverts the buck mode and goes idle - but the CCG command is never sent, so no EPR event ever arrives, and because epr_flow_pending_deferred() is only scheduled from handle_pd_epr_mode() (after buck ready), the event-loss backstop never runs either. The pd_epr_in_progress port bits, the collapsed input current and the transition throttle are all left standing: a second independent route into the latched slow state that otherwise only a physical AC removal clears. Give the abandonment branch the same cleanup the retry-exhausted branch of epr_flow_pending_deferred() already performs: clear the in-progress port bits (keeping the process flags so the existing restore block in cypd_update_port_state() runs), release the ENTER/EXIT transition throttle events, and refresh the active charge port so the input current limit is restored from the standing contract. The buck mode revert, retry count and timing are unchanged; on an abandoned exit the standing contract is the still-active EPR contract and epr_active stays 1, so a later exit attempt remains possible. This stacks on "fwk: epr: make a failed EPR transition recoverable without AC removal" (the two compose: that change covers the failure paths after the CCG command is sent, this one covers the flow being abandoned before it is sent). Issue: tyeth/framework-system#1 Telemetry/data: https://github.com/tyeth/fw16-gpu-watchdog Compile-untested: no Zephyr build environment was available; the change uses only symbols already used in this file and was verified by read-through. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Add fork-build.yml: builds the tulip EC image on ubuntu-latest using zmake with the Zephyr SDK toolchain, Framework zephyr fwk-main, and the cros cmsis chromeos-main checkout, per docs/zephyr/out_of_chroot.md. Remove zephyr-build.yml on this branch only: it targets the Framework self-hosted build server and would queue forever on a fork. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
CI-only branch: the two stacked EPR patches from #1 and #2 plus a hosted-runner build workflow. Opened to get a checkpatch lint run against fwk-tulip-29169. Not for merge.
🤖 Generated with Claude Code