Bluetooth: controller: Fix encryption setup to be queueable

[cvinayak]

Fix the controller implementation to make start encryption
queueable if there is any control procedure in progress.

The context related to encryption procedure is not shared so
that it will be used as the ongoing procedure completes.

The fix here maintains the old functionality of serializing
the queued data and LL Encryption Request PDU, so that data
queued before start encryption is acknowledged.

Fixes #15012. (which was marked as fixed but still had an outstanding issue)

Signed-off-by: Vinayak Kariappa Chettimada vich@nordicsemi.no

[cvinayak]

@joerchan @carlescufi Ran encryption related 37 conformance tests, no regression.

[cvinayak]

@thoh-ot is unable to review this until after Easter holidays. Hence, could we find another reviewer?

[mped-oticon]

This pertains to the pre-split controller, which only permits Oticon to do a shallow review.
Nevertheless, I have read through it and found no issues.

[Qbicz]

#15012 which was fixed in #15186 still works with this PR. Security can still be established after any number of reconnections. Tested with nRF52 Desktop firmware on pca10059 + pca20041.

CC @carlescufi

[cvinayak]

Thanks @mped-oticon and @Qbicz.

I have fixed commit message for misspelled 'not' -> 'now' and 'at' -> 'after'.

[carlescufi]

@Qbicz thank you for testing.

[Qbicz]

@cvinayak Actually now I noticed after reconnecting the peripheral 4 times, I get an assertion on central. I can reproduce it every time (after rebooting central 4th reconnection from peripheral causes this error again).
The assertion occurs in bt_recv_prio for hdr->evt = 26.

[02021262] <err> bt_hci_core: F: bt_recv_prio, evt 14
[02022348] <inf> ble_state: Connected to c4:ac:5a:2a:d5:21 (random)
[02022353] <err> bt_hci_core: F: bt_recv_prio, evt 15
[02022419] <err> bt_hci_core: F: bt_recv_prio, evt 15
[02024077] <err> bt_hci_core: F: bt_recv_prio, evt 15
[02040435] <inf> ble_state: Security with c4:ac:5a:2a:d5:21 (random) level 2
[02040447] <err> bt_ctlr_hci: Tx Buffer Overflow
[02040448] <err> bt_hci_core: F: bt_recv_prio, evt 26
[02040449] <err> bt_hci_core: assert: 'bt_hci_evt_is_prio(hdr->evt)' failed
***** Kernel OOPS! *****
Current thread ID = 0x200022dc
Faulting instruction address = 0x1a30e
Fatal fault in thread 0x200022dc! Aborting.
[0ASSERTION FAIL [out_ctx->control_block->offset <= out_ctx->size] @ /home/fi/ncs/zephyr/subsys/logging/log_output.c:101
***** Kernel Panic! *****
Current thread ID = 0x20003d14
Faulting instruction address = 0x8b3c

[cvinayak]

With @Qbicz and @joerchan help found the bug, in the form of missing code to update data queue head and tail after the enc_req pdu was moved to end of control queue. PR commit updated.

[Qbicz]

@cvinayak tested with latest pushed changes. I did connect+encrypt+disconnect 15 times and cannot see any issue anymore. Thanks for help @carlescufi and @joerchan