[DjDeveloperr]

Stacked on #1285.

Cause. The sleep's debug-link cut selects a page of the debug port's controller (sn201202x 1-0c) with a single write, then reads the selection back 2000 times. That write is sometimes dropped while the controller wakes. Every read then returns the page tipd used last (0x20), the cut gives up with debug link: the stop failed: -110, and the link stays up through the sleep. Seen on the J700 with #1208 merged (kit dlg, 1 of 18 cuts on a rerun: select 0x8 ... read 0x20 after 2000).

Fix. While the readback is neither the page nor a selection in progress (page | 0x80), send WAKEUP and the select again every 20 reads.

Hardware (J700).

• With a forced stale select (TEST kit; the first write goes to page 0x20):
  • without the retry: -110;
  • with it: read 0x20 at 19, sent again, then the cut completed and the contract came back.
• Deep sleep (suspend-to-RAM) for the MacBook Neo (T8140) at EL2 #1208-merged tree with the fix: 20 of 20 sleeps, no -110.
• This PR's code on t8140: rearm the debug link reliably after a sleep #1285's head (d44fd35 + this commit): 40 of 40 sleeps over 4 boots, each 1 deep + 9 DRAM-kept.
  • The PD contract, the host's port state and the Debug USB were up after every sleep.
  • The host's reboot VDM worked after each 10-sleep block.
  • NVMe, USB hub and keyboard, panel and Wi-Fi were OK after every wake.
  • One -110 remains, from a different step: the S3 notice the cut sends after its reset didn't complete in time. The link was already cut and that sleep was fine.
• The merged branch head: 6 of 6 sleeps plus the mic row (HPAI and LPAI recording before and after 2 sleeps).
• This branch also confirms t8140: rearm the debug link reliably after a sleep #1285's contract poll and second reset: every PD loss after the first reset came back on reset 2.
  • Real sleeps: 3 of 25 on t8140: rearm the debug link reliably after a sleep #1285's head and 12 of 38 on the Deep sleep (suspend-to-RAM) for the MacBook Neo (T8140) at EL2 #1208 tree.
  • Awake loop: 11 of 200 resets, the same no-plug state, each recovered by a second reset.

🤖 Generated with Claude Code

[DjDeveloperr]

Superseded by #1445 (merged as 2b05915); its changes are in integration/aurora. The page-select retry is in sleep_hpm_select() in drivers/soc/apple/sleep.c.