The early version of XuanTie C910 core has a store merge buffer
delay problem. The store merge buffer could improve the store queue
performance by merging multi-store requests, but when there are not
continued store requests, the prior single store request would be
waiting in the store queue for a long time. That would cause
significant problems for communication between multi-cores. This
problem was found on sg2042 & th1520 platforms with the qspinlock
lock torture test.
So appending a fence w.o could immediately flush the store merge
buffer and let other cores see the write result.
This will apply the WRITE_ONCE errata to handle the non-standard
behavior via appending a fence w.o instruction for WRITE_ONCE().
This problem is only observed on the sg2042 hardware platform by
running the lock_torture test program for half an hour. The problem
was not found in the user space application, because interrupt can
break the livelock.
Reviewed-by: Leonardo Bras <leobras@redhat.com>
Signed-off-by: Guo Ren (Alibaba DAMO Academy) <guoren@kernel.org>
Link: https://lore.kernel.org/r/20250713155321.2064856-3-guoren@kernel.org
Signed-off-by: Han Gao <rabenda.cn@gmail.com>
0d94e08ccd