U-Boot flare_ab: drop the unproven p54 attribution from the reset note
ci / test (push) Successful in 10s
ci / bench (push) Successful in 15s
ci / mcdc (push) Successful in 42s
ci / prune-artifacts (push) Skipped
ci / kernel-build (push) Skipped
ci / patches-apply (push) Successful in 1m25s
ci / kernel-symbols (push) Successful in 2m13s
ci / quality (push) Successful in 1m30s
ci / badges (push) Skipped
ci / qemu-tools (push) Successful in 6m33s

The comment above flare_ab_global_reset() said that reset_cpu() ending in
hang() at the boot splash was the stall the p54 record shows
(flare-edge#3118). flare-edge#4569 records that this attribution has no
evidence: the p54 record shows attempt 3 on the recovery branch, where no
reset is issued, and the screen was never captured on serial.

Keep only the source-level finding (the WARM request writes CRU
glb_srst_snd, which does not reset this SoC; flare-edge#4554) and point to
flare-edge#4569 for the open p54 cause.

Comment-only change to 0003-flare-ab-slots.patch. The new-file index id of
common/flare_ab.c moves from 2b41ab8ca to ec685444d; the flare-edge pin and
the wiki pins that cite this commit move with it. check-patch-index.py
passes on the patch.

Refs flare-edge#5555, flare-edge#4554, flare-edge#4569.
Refs flare-edge#4565: the no-committed escalation message (printf in
flare_ab.c) was fixed in the previous commit, which did not cite it.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B1W6qGNpGNtFT7nhShpH56
This commit is contained in:
BFE Engineering
2026-11-14 19:05:37 -06:00
co-authored by Claude Haiku 5.5
parent 81d609fed3
commit 9705273f92
@@ -142,7 +142,7 @@ index 52cd612d1..a81071566 100644
obj-$(CONFIG_AUTOBOOT) += autoboot.o
diff --git a/sysdrv/source/uboot/u-boot/common/flare_ab.c b/sysdrv/source/uboot/u-boot/common/flare_ab.c
new file mode 100644
index 000000000..2b41ab8ca
index 000000000..ec685444d
--- /dev/null
+++ b/sysdrv/source/uboot/u-boot/common/flare_ab.c
@@ -0,0 +1,709 @@
@@ -721,11 +721,11 @@ index 000000000..2b41ab8ca
+ * Escalating reset after a failed storage boot.
+ *
+ * Do not call do_reset() here. On this SoC the U-Boot WARM request
+ * (sysreset_rockchip, ARM32 path) writes CRU glb_srst_snd (0xeca8), which does
+ * not reset the chip, and reset_cpu() then ends in hang() at the boot splash:
+ * the stall the p54 record shows (flare-edge#3118). The reset that works is the
+ * CRU glb_srst_fst poke that the Linux restart path and the bench scenarios
+ * use (docs/boot-flow.md, "Watchdog / reset facts"). It is a warm reset, so
+ * (sysreset_rockchip, ARM32 path) writes CRU glb_srst_snd (0xeca8), which
+ * does not reset the chip (flare-edge#4554). The p54 splash stall cause is
+ * not established (flare-edge#4569). The reset that works is the CRU
+ * glb_srst_fst poke that the Linux restart path and the bench scenarios use
+ * (docs/boot-flow.md, "Watchdog / reset facts"). It is a warm reset, so
+ * glb_rst_st stays nonzero and OS_REG4 carries the attempt count.
+ */
+static void flare_ab_global_reset(void)