sv6160: retract the tx_batch ceiling; the stalls tracked the write regime
The previous commit claimed the firmware refuses num_frames_per_desc above 9, on the strength of tx_batch=16 stalling TX 4 times out of 4 and 12 and 10 stalling on their first transfer. That does not replicate. Six transfers at each of 8, 10, 12 and 16 stalled ZERO times, at 1.81-1.91 MB/s throughout. The claim was under-powered where it mattered: 10 and 12 had exactly ONE trial each, which is the single-trial A/B this driver is known to punish. Only 16 had repeats, and those all fell inside one bad window. What actually separates the two passes is the SDIO write regime, not the batch size. Every stall observed happened while throughput sat in the degraded ~1.0 MB/s regime; every clean run, including all 24 trials of the replication, has been in the healthy ~1.85 MB/s one. debugfs tx_tokens `body_avg_us` plus its byte count tells them apart at a glance: ~7-9 B/us healthy, ~2 B/us degraded. So batch size above 9 may be a probability dial on the link-death stall while the chip is already degraded, and harmless while it is not. That is a statement about that open bug, not about a descriptor limit -- and it is a lead worth having, because it is the first thing found that correlates with a failure that otherwise arrives hours apart with no warning. The two negative results from that work do hold and are kept: we are not truncating the descriptor (it sums to exactly SV6160_TX_CMD_SIZE with 16 table slots, all transmitted), and it is not a global firmware frame pool (descs_per_ac=1 + tx_batch=16 is 16 frames in flight, fewer than the 18 that descs_per_ac=2 + tx_batch=9 runs happily on, and it stalled anyway). The 8-vs-9 throughput figure also stands: it came from 45 paired trials with the arms alternating every cycle, so both arms saw the same regime. Default stays 8, for the same reason as before but on honest grounds: 9 is the highest value never yet seen to stall, and +3% does not pay for spending that margin while the stall itself is not understood. Comment only; the binary is unchanged (md5 7280ff0261eb). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JSNniou44nN9EFGqTmZ6my
This commit is contained in:
1 parent
1a30deb20c
commit
714280f648
1 file changed
+35
-25
@@ -599,38 +599,48 @@ struct rpu_cmd_txq_params {
|
||||
* tx_batch=8 1.37 / 1.50 MB/s
|
||||
* -- with no sign of flattening at 8.
|
||||
*
|
||||
* MEASURED 2026-09-16, and the firmware's nominal 16 is NOT usable. Probed on
|
||||
* hardware, 3 MB uplink transfers, module reloaded before each arm:
|
||||
* tx_batch=16 stalls TX on the first transfer, 3 tries out of 3
|
||||
* tx_batch=12 stalls on the first transfer
|
||||
* tx_batch=10 stalls on the first transfer
|
||||
* tx_batch=9 clean, and fills to 7.86 frames/descriptor
|
||||
* tx_batch=8 clean, 9 transfers out of 9
|
||||
* A stall here is not graceful: the descriptors in flight are never completed,
|
||||
* and the SDIO bus can end up at -16 needing an unbind/bind of ff380000.mmc.
|
||||
* No reboot occurred in any of these trials (uptime held throughout), so the
|
||||
* historical reboot at 16 really was the separate sustained-uplink bug, as the
|
||||
* note this replaced suspected.
|
||||
* NO FIRMWARE CEILING HAS BEEN DEMONSTRATED. A first pass on 2026-09-16 read
|
||||
* like a hard cliff at 9 -- 16 stalled TX on 4 of 4 attempts, 12 and 10 on
|
||||
* their first -- but that was under-powered (one trial each at 10 and 12) and
|
||||
* it did not survive replication: a later run of SIX transfers at each of 8,
|
||||
* 10, 12 and 16 stalled ZERO times, at 1.81-1.91 MB/s throughout.
|
||||
*
|
||||
* The ceiling is PER-DESCRIPTOR, not total frames in flight: descs_per_ac=1
|
||||
* with tx_batch=16 puts only 16 frames in flight -- fewer than the 18 that
|
||||
* descs_per_ac=2 with tx_batch=9 runs happily -- and it still stalls on the
|
||||
* first transfer. So a firmware frame pool does not explain it; the firmware
|
||||
* simply does not honour num_frames_per_desc above 9, whatever its tables say.
|
||||
* (Those tables really are 16 wide and really are all transmitted: the struct
|
||||
* sums to exactly SV6160_TX_CMD_SIZE with SV6160_MAX_TX_CMDS=16, so this is
|
||||
* not us truncating the descriptor.)
|
||||
* What separates the two passes is not the batch size, it is the SDIO write
|
||||
* regime (see the excursion note above sv6160_tx_xfer_len). Every stall
|
||||
* happened while throughput sat in the degraded ~1.0 MB/s regime; every clean
|
||||
* run has been in the healthy ~1.85 MB/s one. Read `body_avg_us` and its
|
||||
* byte count out of debugfs tx_tokens to tell them apart: ~7-9 B/us is
|
||||
* healthy, ~2 B/us is degraded. So the honest statement is that batch size
|
||||
* above 9 MAY be a trigger for the link-death stall while the chip is already
|
||||
* degraded, and is harmless while it is not -- which is a statement about
|
||||
* that open bug, not about a descriptor limit.
|
||||
*
|
||||
* Do not conclude "16 is fine" from this either. Both passes are consistent
|
||||
* with batch size being a probability dial on that stall rather than a cliff,
|
||||
* and a stall is not graceful: the descriptors in flight are never completed,
|
||||
* ieee80211_restart_hw does NOT clear it, and it has left the SDIO bus at -16
|
||||
* needing an unbind/bind of ff380000.mmc from the serial console.
|
||||
*
|
||||
* Ruled out along the way, and these do hold: it is not us truncating the
|
||||
* descriptor (the struct sums to exactly SV6160_TX_CMD_SIZE with
|
||||
* SV6160_MAX_TX_CMDS=16, so all 16 table slots are transmitted), and it is
|
||||
* not a global firmware frame pool (descs_per_ac=1 with tx_batch=16 puts only
|
||||
* 16 frames in flight, fewer than the 18 that descs_per_ac=2 with tx_batch=9
|
||||
* runs happily on, and it stalled anyway). No reboot occurred in any trial,
|
||||
* so the historical reboot at 16 was the separate sustained-uplink bug, as
|
||||
* the note this replaced suspected.
|
||||
*
|
||||
* 9 does work and is worth about 3%: 45 paired trials alternating 8 and 9
|
||||
* every cycle with NO reload between the arms (so both see the same RF and
|
||||
* the same throughput regime) gave median 1.873 -> 1.937 MB/s, 9 ahead in
|
||||
* 34 of 45 pairs, zero stalls on either arm over 135 MB.
|
||||
*
|
||||
* Left at 8 anyway. 9 is the LAST working value -- 10 fails immediately and
|
||||
* hard -- and 3% is not worth operating with no margin against a failure that
|
||||
* wedges the link until a module reload. Every array here scales off this
|
||||
* constant (bodies[], batch[], tx_batch[][]; frm_status[] in the event is
|
||||
* already 16), so changing it stays a one-line edit. */
|
||||
* Left at 8 anyway. The +3% is small, 9 was the highest value never yet seen
|
||||
* to stall, and raising it trades that margin for a failure mode that needs a
|
||||
* module reload to clear -- a bad trade until the link-death stall itself is
|
||||
* understood. Every array here scales off this constant (bodies[], batch[],
|
||||
* tx_batch[][]; frm_status[] in the event is already 16), so changing it
|
||||
* stays a one-line edit. */
|
||||
#define SV6160_TX_BATCH_MAX 8
|
||||
/* Most pieces mac80211 may hand us in one (A-)MSDU; we flatten them anyway,
|
||||
* this just bounds the copy. */
|
||||
|
||||
Reference in new issue
Block a user