On (23/03/03 15:29), Andrew Morton wrote: > From: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> > Subject: zsmalloc-fine-grained-inuse-ratio-based-fullness-grouping-fix-fix > Date: Fri Mar 3 03:25:23 PM PST 2023 > > fix zs_stats_size_show() accumulation, per Yosry [..] > --- a/mm/zsmalloc.c~zsmalloc-fine-grained-inuse-ratio-based-fullness-grouping-fix-fix > +++ a/mm/zsmalloc.c > @@ -651,9 +651,9 @@ static int zs_stats_size_show(struct seq > * stats. > */ > for (fg = ZS_INUSE_RATIO_70; fg <= ZS_INUSE_RATIO_99; fg++) > - class_almost_full = zs_stat_get(class, fg); > + class_almost_full += zs_stat_get(class, fg); > for (fg = ZS_INUSE_RATIO_10; fg < ZS_INUSE_RATIO_70; fg++) > - class_almost_empty = zs_stat_get(class, fg); > + class_almost_empty += zs_stat_get(class, fg); > > obj_allocated = zs_stat_get(class, ZS_OBJS_ALLOCATED); > obj_used = zs_stat_get(class, ZS_OBJS_INUSE); This also needs to set class_almost_full and class_almost_empty to 0 for each class. Andrew, I'll send v4 in a bit.
