[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230926073333epcms1p14c9798232b395007eb20becb5dbc4b4e@epcms1p1>
Date: Tue, 26 Sep 2023 16:33:33 +0900
From: 김재원 <jaewon31.kim@...sung.com>
To: 김재원 <jaewon31.kim@...sung.com>,
Yu Zhao <yuzhao@...gle.com>
CC: "rostedt@...dmis.org" <rostedt@...dmis.org>,
"tjmercier@...gle.com" <tjmercier@...gle.com>,
"kaleshsingh@...gle.com" <kaleshsingh@...gle.com>,
"akpm@...ux-foundation.org" <akpm@...ux-foundation.org>,
"vbabka@...e.cz" <vbabka@...e.cz>,
"hannes@...xchg.org" <hannes@...xchg.org>,
"sj@...nel.org" <sj@...nel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"linux-trace-kernel@...r.kernel.org"
<linux-trace-kernel@...r.kernel.org>,
"linux-mm@...ck.org" <linux-mm@...ck.org>,
"jaewon31.kim@...il.com" <jaewon31.kim@...il.com>
Subject: RE: [PATCH v4] vmscan: add trace events for lru_gen
>>On Mon, Sep 25, 2023 at 10:20?PM Jaewon Kim <jaewon31.kim@...sung.com> wrote:
>>>
>>> As the legacy lru provides, the lru_gen needs some trace events for
>>> debugging.
>>>
>>> This commit introduces 2 trace events.
>>> trace_mm_vmscan_lru_gen_scan
>>> trace_mm_vmscan_lru_gen_evict
>>>
>>> Each event is similar to the following legacy events.
>>> trace_mm_vmscan_lru_isolate,
>>> trace_mm_vmscan_lru_shrink_[in]active
>>
>>We should just reuse trace_mm_vmscan_lru_isolate and
>>trace_mm_vmscan_lru_shrink_inactive instead of adding new tracepoints.
>>
>>To reuse trace_mm_vmscan_lru_isolate, we'd just need to append two new
>>names to LRU_NAMES.
>>
>>The naming of trace_mm_vmscan_lru_shrink_inactive might seem confusing
>>but it's how MGLRU maintains the compatibility, e.g., the existing
>>active/inactive counters in /proc/vmstat.
>
>
>Hello
>
>Actually I had tried to reuse them. But some value was not that compatible.
>Let me try that way again.
>
>>
Hello Yu Zhao
Could you look into what I tried below? I reused the legacy trace events as you recommened.
For the nr_scanned for trace_mm_vmscan_lru_shrink_inactive, I just used the scanned returned from isolate_folios.
I thought this is right as scan_folios also uses its isolated.
__count_vm_events(PGSCAN_ANON + type, isolated);
But I guess the scanned in scan_folios is actually the one used in shrink_inactive_list
I tested this on both 0 and 7 of /sys/kernel/mm/lru_gen/enabled
diff --git a/mm/vmscan.c b/mm/vmscan.c
index a4e44f1c97c1..b61a0156559c 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -4328,6 +4328,7 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
int sorted = 0;
int scanned = 0;
int isolated = 0;
+ int skipped = 0;
int remaining = MAX_LRU_BATCH;
struct lru_gen_folio *lrugen = &lruvec->lrugen;
struct mem_cgroup *memcg = lruvec_memcg(lruvec);
@@ -4341,7 +4342,7 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
for (i = MAX_NR_ZONES; i > 0; i--) {
LIST_HEAD(moved);
- int skipped = 0;
+ int skipped_zone = 0;
int zone = (sc->reclaim_idx + i) % MAX_NR_ZONES;
struct list_head *head = &lrugen->folios[gen][type][zone];
@@ -4363,16 +4364,17 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
isolated += delta;
} else {
list_move(&folio->lru, &moved);
- skipped += delta;
+ skipped_zone += delta;
}
- if (!--remaining || max(isolated, skipped) >= MIN_LRU_BATCH)
+ if (!--remaining || max(isolated, skipped_zone) >= MIN_LRU_BATCH)
break;
}
- if (skipped) {
+ if (skipped_zone) {
list_splice(&moved, head);
- __count_zid_vm_events(PGSCAN_SKIP, zone, skipped);
+ __count_zid_vm_events(PGSCAN_SKIP, zone, skipped_zone);
+ skipped += skipped_zone;
}
if (!remaining || isolated >= MIN_LRU_BATCH)
@@ -4387,6 +4389,9 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
__count_memcg_events(memcg, item, isolated);
__count_memcg_events(memcg, PGREFILL, sorted);
__count_vm_events(PGSCAN_ANON + type, isolated);
+ trace_mm_vmscan_lru_isolate(sc->reclaim_idx, sc->order, MAX_LRU_BATCH,
+ scanned, skipped, isolated,
+ type ? LRU_INACTIVE_FILE : LRU_INACTIVE_ANON);
/*
* There might not be eligible folios due to reclaim_idx. Check the
@@ -4517,6 +4522,9 @@ static int evict_folios(struct lruvec *lruvec, struct scan_control *sc, int swap
retry:
reclaimed = shrink_folio_list(&list, pgdat, sc, &stat, false);
sc->nr_reclaimed += reclaimed;
+ trace_mm_vmscan_lru_shrink_inactive(pgdat->node_id,
+ scanned, reclaimed, &stat, sc->priority,
+ type ? LRU_INACTIVE_FILE : LRU_INACTIVE_ANON);
list_for_each_entry_safe_reverse(folio, next, &list, lru) {
if (!folio_evictable(folio)) {
>>> Here's an example
>>> mm_vmscan_lru_gen_scan: classzone=2 order=0 nr_requested=4096 nr_scanned=64 nr_skipped=0 nr_taken=64 lru=anon
>>> mm_vmscan_lru_gen_evict: nid=0 nr_reclaimed=64 nr_dirty=0 nr_writeback=0 nr_congested=0 nr_immediate=0 nr_activate_anon=0 nr_activate_file=0 nr_ref_keep=0 nr_unmap_fail=0 priority=2 flags=RECLAIM_WB_ANON|RECLAIM_WB_ASYNC
>>> mm_vmscan_lru_gen_scan: classzone=1 order=0 nr_requested=4096 nr_scanned=64 nr_skipped=0 nr_taken=64 lru=file
>>> mm_vmscan_lru_gen_evict: nid=0 nr_reclaimed=64 nr_dirty=0 nr_writeback=0 nr_congested=0 nr_immediate=0 nr_activate_anon=0 nr_activate_file=0 nr_ref_keep=0 nr_unmap_fail=0 priority=12 flags=RECLAIM_WB_FILE|RECLAIM_WB_ASYNC
>>>
>>> Signed-off-by: Jaewon Kim <jaewon31.kim@...sung.com>
>>> ---
>>> v4: wrap with #ifdef CONFIG_LRU_GEN
>>> v3: change printk format
>>> v2: use condition and make it aligned
>>> v1: introduce trace events
>>> ---
>>> include/trace/events/mmflags.h | 9 ++++
>>> include/trace/events/vmscan.h | 96 ++++++++++++++++++++++++++++++++++
>>> mm/vmscan.c | 20 +++++--
>>> 3 files changed, 120 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/include/trace/events/mmflags.h b/include/trace/events/mmflags.h
>>> index 1478b9dd05fa..6dfe85bd4e81 100644
>>> --- a/include/trace/events/mmflags.h
>>> +++ b/include/trace/events/mmflags.h
>>> @@ -274,6 +274,12 @@ IF_HAVE_VM_SOFTDIRTY(VM_SOFTDIRTY, "softdirty" ) \
>>> EM (LRU_ACTIVE_FILE, "active_file") \
>>> EMe(LRU_UNEVICTABLE, "unevictable")
>>>
>>> +#ifdef CONFIG_LRU_GEN
>>> +#define LRU_GEN_NAMES \
>>> + EM (LRU_GEN_ANON, "anon") \
>>> + EMe(LRU_GEN_FILE, "file")
>>> +#endif
>>> +
>>> /*
>>> * First define the enums in the above macros to be exported to userspace
>>> * via TRACE_DEFINE_ENUM().
>>> @@ -288,6 +294,9 @@ COMPACTION_PRIORITY
>>> /* COMPACTION_FEEDBACK are defines not enums. Not needed here. */
>>> ZONE_TYPE
>>> LRU_NAMES
>>> +#ifdef CONFIG_LRU_GEN
>>> +LRU_GEN_NAMES
>>> +#endif
>>>
>>> /*
>>> * Now redefine the EM() and EMe() macros to map the enums to the strings
>>> diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
>>> index d2123dd960d5..2080ef742f89 100644
>>> --- a/include/trace/events/vmscan.h
>>> +++ b/include/trace/events/vmscan.h
>>> @@ -327,6 +327,102 @@ TRACE_EVENT(mm_vmscan_lru_isolate,
>>> __print_symbolic(__entry->lru, LRU_NAMES))
>>> );
>>>
>>> +#ifdef CONFIG_LRU_GEN
>>> +TRACE_EVENT_CONDITION(mm_vmscan_lru_gen_scan,
>>> + TP_PROTO(int highest_zoneidx,
>>> + int order,
>>> + unsigned long nr_requested,
>>> + unsigned long nr_scanned,
>>> + unsigned long nr_skipped,
>>> + unsigned long nr_taken,
>>> + int lru),
>>> +
>>> + TP_ARGS(highest_zoneidx, order, nr_requested, nr_scanned, nr_skipped, nr_taken, lru),
>>> +
>>> + TP_CONDITION(nr_scanned),
>>> +
>>> + TP_STRUCT__entry(
>>> + __field(int, highest_zoneidx)
>>> + __field(int, order)
>>> + __field(unsigned long, nr_requested)
>>> + __field(unsigned long, nr_scanned)
>>> + __field(unsigned long, nr_skipped)
>>> + __field(unsigned long, nr_taken)
>>> + __field(int, lru)
>>> + ),
>>> +
>>> + TP_fast_assign(
>>> + __entry->highest_zoneidx = highest_zoneidx;
>>> + __entry->order = order;
>>> + __entry->nr_requested = nr_requested;
>>> + __entry->nr_scanned = nr_scanned;
>>> + __entry->nr_skipped = nr_skipped;
>>> + __entry->nr_taken = nr_taken;
>>> + __entry->lru = lru;
>>> + ),
>>> +
>>> + /*
>>> + * classzone is previous name of the highest_zoneidx.
>>> + * Reason not to change it is the ABI requirement of the tracepoint.
>>> + */
>>> + TP_printk("classzone=%d order=%d nr_requested=%lu nr_scanned=%lu nr_skipped=%lu nr_taken=%lu lru=%s",
>>> + __entry->highest_zoneidx,
>>> + __entry->order,
>>> + __entry->nr_requested,
>>> + __entry->nr_scanned,
>>> + __entry->nr_skipped,
>>> + __entry->nr_taken,
>>> + __print_symbolic(__entry->lru, LRU_GEN_NAMES))
>>> +);
>>> +
>>> +TRACE_EVENT(mm_vmscan_lru_gen_evict,
>>> +
>>> + TP_PROTO(int nid, unsigned long nr_reclaimed,
>>> + struct reclaim_stat *stat, int priority, int file),
>>> +
>>> + TP_ARGS(nid, nr_reclaimed, stat, priority, file),
>>> +
>>> + TP_STRUCT__entry(
>>> + __field(unsigned long, nr_reclaimed)
>>> + __field(unsigned long, nr_dirty)
>>> + __field(unsigned long, nr_writeback)
>>> + __field(unsigned long, nr_congested)
>>> + __field(unsigned long, nr_immediate)
>>> + __field(unsigned int, nr_activate0)
>>> + __field(unsigned int, nr_activate1)
>>> + __field(unsigned long, nr_ref_keep)
>>> + __field(unsigned long, nr_unmap_fail)
>>> + __field(int, nid)
>>> + __field(int, priority)
>>> + __field(int, reclaim_flags)
>>> + ),
>>> +
>>> + TP_fast_assign(
>>> + __entry->nid = nid;
>>> + __entry->nr_reclaimed = nr_reclaimed;
>>> + __entry->nr_dirty = stat->nr_dirty;
>>> + __entry->nr_writeback = stat->nr_writeback;
>>> + __entry->nr_congested = stat->nr_congested;
>>> + __entry->nr_immediate = stat->nr_immediate;
>>> + __entry->nr_activate0 = stat->nr_activate[0];
>>> + __entry->nr_activate1 = stat->nr_activate[1];
>>> + __entry->nr_ref_keep = stat->nr_ref_keep;
>>> + __entry->nr_unmap_fail = stat->nr_unmap_fail;
>>> + __entry->priority = priority;
>>> + __entry->reclaim_flags = trace_reclaim_flags(file);
>>> + ),
>>> +
>>> + TP_printk("nid=%d nr_reclaimed=%lu nr_dirty=%lu nr_writeback=%lu nr_congested=%lu nr_immediate=%lu nr_activate_anon=%u nr_activate_file=%u nr_ref_keep=%lu nr_unmap_fail=%lu priority=%d flags=%s",
>>> + __entry->nid, __entry->nr_reclaimed,
>>> + __entry->nr_dirty, __entry->nr_writeback,
>>> + __entry->nr_congested, __entry->nr_immediate,
>>> + __entry->nr_activate0, __entry->nr_activate1,
>>> + __entry->nr_ref_keep, __entry->nr_unmap_fail,
>>> + __entry->priority,
>>> + show_reclaim_flags(__entry->reclaim_flags))
>>> +);
>>> +#endif
>>> +
>>> TRACE_EVENT(mm_vmscan_write_folio,
>>>
>>> TP_PROTO(struct folio *folio),
>>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>>> index 6f13394b112e..0c8b48bcb461 100644
>>> --- a/mm/vmscan.c
>>> +++ b/mm/vmscan.c
>>> @@ -5005,6 +5005,7 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
>>> int sorted = 0;
>>> int scanned = 0;
>>> int isolated = 0;
>>> + int skipped = 0;
>>> int remaining = MAX_LRU_BATCH;
>>> struct lru_gen_folio *lrugen = &lruvec->lrugen;
>>> struct mem_cgroup *memcg = lruvec_memcg(lruvec);
>>> @@ -5018,7 +5019,7 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
>>>
>>> for (i = MAX_NR_ZONES; i > 0; i--) {
>>> LIST_HEAD(moved);
>>> - int skipped = 0;
>>> + int skipped_zone = 0;
>>> int zone = (sc->reclaim_idx + i) % MAX_NR_ZONES;
>>> struct list_head *head = &lrugen->folios[gen][type][zone];
>>>
>>> @@ -5040,16 +5041,17 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
>>> isolated += delta;
>>> } else {
>>> list_move(&folio->lru, &moved);
>>> - skipped += delta;
>>> + skipped_zone += delta;
>>> }
>>>
>>> - if (!--remaining || max(isolated, skipped) >= MIN_LRU_BATCH)
>>> + if (!--remaining || max(isolated, skipped_zone) >= MIN_LRU_BATCH)
>>> break;
>>> }
>>>
>>> - if (skipped) {
>>> + if (skipped_zone) {
>>> list_splice(&moved, head);
>>> - __count_zid_vm_events(PGSCAN_SKIP, zone, skipped);
>>> + __count_zid_vm_events(PGSCAN_SKIP, zone, skipped_zone);
>>> + skipped += skipped_zone;
>>> }
>>>
>>> if (!remaining || isolated >= MIN_LRU_BATCH)
>>> @@ -5065,6 +5067,10 @@ static int scan_folios(struct lruvec *lruvec, struct scan_control *sc,
>>> __count_memcg_events(memcg, PGREFILL, sorted);
>>> __count_vm_events(PGSCAN_ANON + type, isolated);
>>>
>>> +#ifdef CONFIG_LRU_GEN
>>> + trace_mm_vmscan_lru_gen_scan(sc->reclaim_idx, sc->order, MAX_LRU_BATCH,
>>> + scanned, skipped, isolated, type);
>>> +#endif
>>
>>These functions are already within CONFIG_LRU_GEN.
>>
>>> /*
>>> * There might not be eligible folios due to reclaim_idx. Check the
>>> * remaining to prevent livelock if it's not making progress.
>>> @@ -5194,6 +5200,10 @@ static int evict_folios(struct lruvec *lruvec, struct scan_control *sc, int swap
>>> retry:
>>> reclaimed = shrink_folio_list(&list, pgdat, sc, &stat, false);
>>> sc->nr_reclaimed += reclaimed;
>>> +#ifdef CONFIG_LRU_GEN
>>> + trace_mm_vmscan_lru_gen_evict(pgdat->node_id, reclaimed, &stat,
>>> + sc->priority, type);
>>> +#endif
>>
>>Ditto.
Powered by blists - more mailing lists