On 7/17/24 14:24, tj@xxxxxxxxxx wrote:
Hello,
On Wed, Jul 17, 2024 at 02:18:39PM -0400, Waiman Long wrote:
Well, it can be confusing whether we are dealing with blkg->iostat or
blkg->iostat_cpu. In many cases, we are dealing with iostat_cpu instead of
iostat like __blkcg_rstat_flush() and blkg_clear_stat(). So we can't
eliminate the use of u64_stats_update_begin_irqsave() in those cases.
I mean, we need to distinguish them. For 32bits, blkg->iostat has multiple
updaters, so we can't use u64_sync; however, blkg->iostat_cpu has only one
updater (except blkg_clear_stat() which I don't think we need to worry too
much about), so u64_sync is fine.
I was wrong about __blkcg_rstat_flush(). Right, the main updater of
iostat_cpu is blk_cgroup_bio_start(). We do need to drop down some
comment on what is protected by u64_sync and what is by blkg_stat_lock
though. It can be confusing.
Cheers,
Longman
