Hi,
v5.4.y commit 3cf391e4174a ("wifi: ath10k: Don't touch the CE
interrupt registers after power up"), which is commit 170c75d43a77
upstream, unleashed multiple DB845c(sdm845) regressions ranging
from random RCU stalls to UFS crashes as also reported here
https://lore.kernel.org/lkml/20230630151842.1.If764ede23c4e09a43a842771c2ddf99608f25f8e@changeid/
Taking a cue from the commit message of 170c75d43a77, I tried
backporting upstream commit d66d24ac300c ("ath10k: Keep track of
which interrupts fired, don't poll them") and other relevant fixes
and that seem to have done the trick.
We no longer see any of the above reported regressions with the
following patchset. This upstream patchset is just an educated
guess and there may be one or more fixes in this series which are
not needed at all but I have not tested them individually and
marked all of them as Stable-dep-of: 170c75d43a77 ("ath10k: Don't
touch the CE interrupt registers after power up") instead.
Douglas Anderson (3):
ath10k: Wait until copy complete is actually done before completing
ath10k: Keep track of which interrupts fired, don't poll them
ath10k: Get rid of "per_ce_irq" hw param
Rakesh Pillai (1):
ath10k: Add interrupt summary based CE processing
drivers/net/wireless/ath/ath10k/ce.c | 79 ++++++++++++++------------
drivers/net/wireless/ath/ath10k/ce.h | 15 +++--
drivers/net/wireless/ath/ath10k/core.c | 13 -----
drivers/net/wireless/ath/ath10k/hw.h | 3 -
drivers/net/wireless/ath/ath10k/snoc.c | 19 +++++--
drivers/net/wireless/ath/ath10k/snoc.h | 1 +
6 files changed, 64 insertions(+), 66 deletions(-)
--
2.25.1
