Introduction of lockless subqueues broke the class statistics.
Before the change stats were accumulated in `bstats' and `qstats'
on the stack which was then copied to struct gnet_dump.

After the change the `bstats' and `qstats' are initialized to 0
and never updated, yet still fed to gnet_dump. The code updates
the global qdisc->cpu_bstats and qdisc->cpu_qstats instead,
clobbering them. Most likely a copy-paste error from the code in
mqprio_dump().

__gnet_stats_copy_basic() and __gnet_stats_copy_queue() accumulate
the values for per-CPU case but for global stats they overwrite
the value, so only stats from the last loop iteration / tc end up
in sch->[bq]stats.

Use the on-stack [bq]stats variables again and add the stats manually
in the global case.

Fixes: ce679e8d ("net: sched: add support for TCQ_F_NOLOCK subqueues to sch_mqprio")
Cc: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
https://lore.kernel.org/all/20211007175000.2334713-2-bigeasy@linutronix.de/Signed-off-by: Jakub Kicinski <kuba@kernel.org>