From 25cd1bee0c8025150275252e26c88a7cdf51adb9 Mon Sep 17 00:00:00 2001 From: Zhao Mengmeng Date: Tue, 11 Apr 2023 05:52:24 -0400 Subject: slabinfo: add comment for cache_size calculation One of our physical machine shows that the "CACHE SIZE" column of slabtop output is extremely high, three times of the products of objs nums and objs size. After some analysis, we found that the order of slab, which decides "pages per slab", will shrink when memory pressure is high and normal order allocation failed. So we think it might help to add these comments to the man help. Minor fix: add the "memory." back, which is lost after "aa461df0: docs: Minor manpage fixes" Signed-off-by: Zhao Mengmeng --- library/include/slabinfo.h | 2 +- library/slabinfo.c | 4 ++++ man/slabtop.1 | 8 +++++++- 3 files changed, 12 insertions(+), 2 deletions(-) diff --git a/library/include/slabinfo.h b/library/include/slabinfo.h index 128bad9..f63ca79 100644 --- a/library/include/slabinfo.h +++ b/library/include/slabinfo.h @@ -40,7 +40,7 @@ enum slabinfo_item { SLAB_ACTIVE_SLABS, // u_int " SLAB_PAGES_PER_SLAB, // u_int " SLAB_PERCENT_USED, // u_int derived from ACTIVE_OBJS / NUM_OBJS - SLAB_SIZE_TOTAL, // ul_int derived from page size * NUM_OBJS * PAGES_PER_SLAB + SLAB_SIZE_TOTAL, // ul_int derived from page size * NUMS_SLABS * PAGES_PER_SLAB SLABS_CACHES_TOTAL, // u_int derived from all caches SLABS_CACHES_ACTIVE, // u_int " diff --git a/library/slabinfo.c b/library/slabinfo.c index a4386d6..1d4453f 100644 --- a/library/slabinfo.c +++ b/library/slabinfo.c @@ -434,6 +434,10 @@ static int parse_slabinfo20 ( if (node->obj_size > slabs->max_obj_size) slabs->max_obj_size = node->obj_size; + /* cache_size is not accurate, it's the upper limit of memory used by this slab. + * When system using slub(most common case) is under high memory pressure, there + * are slab order fallbacks, which means pages_per_slab is not constant and may decrease. + */ node->cache_size = (unsigned long)node->nr_slabs * node->pages_per_slab * page_size; if (node->nr_objs) { diff --git a/man/slabtop.1 b/man/slabtop.1 index edf4784..089fb0e 100644 --- a/man/slabtop.1 +++ b/man/slabtop.1 @@ -104,7 +104,13 @@ The .B slabtop statistic header is tracking how many bytes of slabs are being used and is not a measure of physical memory. The 'Slab' field in the -\fI/proc/meminfo\fR file is tracking information about used slab physical +\fI/proc/meminfo\fR file is tracking information about used slab physical memory. +.PP +The +.B CACHE SIZE +column is not accurate, it's the upper limit of memory used by specific slab. When system +using slub (most common case) is under high memory pressure, there are slab order +fallbacks, which means "pages per slab" is not constant and may decrease. .SH AUTHORS Written by Chris Rivera and Robert Love. .PP -- cgit v1.2.1