crash/SOURCES/0008-Improve-the-ps-perform...

150 lines
4.3 KiB
Diff
Raw Permalink Normal View History

2022-03-01 11:34:58 +00:00
From e389667cf62ef5db82f9796cdbc0134ec38612dc Mon Sep 17 00:00:00 2001
From: Tao Liu <ltao@redhat.com>
Date: Fri, 21 Jan 2022 13:43:09 +0800
Subject: [PATCH 08/10] Improve the ps performance for vmcores with large
number of threads
Previously, the ps command will iterate over all threads which
have the same tgid, to accumulate their rss value, in order to
get a thread/process's final rss value as part of the final output.
For non-live systems, the rss accumulation values are identical for
threads which have the same tgid, so there is no need to do the
iteration and accumulation repeatly, thus a lot of readmem calls are
skipped. Otherwise it will be the performance bottleneck if the
vmcores have a large number of threads.
In this patch, the rss accumulation value will be stored in a cache,
next time a thread with the same tgid will take it directly without
the iteration.
For example, we can monitor the performance issue when a vmcore has
~65k processes, most of which are threads for several specific
processes. Without the patch, it will take ~7h for ps command
to finish. With the patch, ps command will finish in 1min.
Signed-off-by: Tao Liu <ltao@redhat.com>
---
defs.h | 1 +
memory.c | 70 +++++++++++++++++++++++++++++++-------------------------
task.c | 1 +
3 files changed, 41 insertions(+), 31 deletions(-)
diff --git a/defs.h b/defs.h
index b63741c7d78b..55600d56ef1c 100644
--- a/defs.h
+++ b/defs.h
@@ -829,6 +829,7 @@ struct task_context { /* context stored for each task */
struct tgid_context { /* tgid and task stored for each task */
ulong tgid;
ulong task;
+ long rss_cache;
};
struct task_table { /* kernel/local task table data */
diff --git a/memory.c b/memory.c
index 5af45fd7d834..e80c59ea4534 100644
--- a/memory.c
+++ b/memory.c
@@ -4665,7 +4665,7 @@ void
get_task_mem_usage(ulong task, struct task_mem_usage *tm)
{
struct task_context *tc;
- long rss = 0;
+ long rss = 0, rss_cache = 0;
BZERO(tm, sizeof(struct task_mem_usage));
@@ -4730,38 +4730,46 @@ get_task_mem_usage(ulong task, struct task_mem_usage *tm)
(last->tgid == (last + 1)->tgid))
last++;
- while (first <= last)
- {
- /* count 0 -> filepages */
- if (!readmem(first->task +
- OFFSET(task_struct_rss_stat) +
- OFFSET(task_rss_stat_count), KVADDR,
- &sync_rss,
- sizeof(int),
- "task_struct rss_stat MM_FILEPAGES",
- RETURN_ON_ERROR))
- continue;
-
- rss += sync_rss;
-
- /* count 1 -> anonpages */
- if (!readmem(first->task +
- OFFSET(task_struct_rss_stat) +
- OFFSET(task_rss_stat_count) +
- sizeof(int),
- KVADDR, &sync_rss,
- sizeof(int),
- "task_struct rss_stat MM_ANONPAGES",
- RETURN_ON_ERROR))
- continue;
-
- rss += sync_rss;
-
- if (first == last)
- break;
- first++;
+ /*
+ * Using rss cache for dumpfile is more beneficial than live debug
+ * because its value never changes in dumpfile.
+ */
+ if (ACTIVE() || last->rss_cache == UNINITIALIZED) {
+ while (first <= last)
+ {
+ /* count 0 -> filepages */
+ if (!readmem(first->task +
+ OFFSET(task_struct_rss_stat) +
+ OFFSET(task_rss_stat_count), KVADDR,
+ &sync_rss,
+ sizeof(int),
+ "task_struct rss_stat MM_FILEPAGES",
+ RETURN_ON_ERROR))
+ continue;
+
+ rss_cache += sync_rss;
+
+ /* count 1 -> anonpages */
+ if (!readmem(first->task +
+ OFFSET(task_struct_rss_stat) +
+ OFFSET(task_rss_stat_count) +
+ sizeof(int),
+ KVADDR, &sync_rss,
+ sizeof(int),
+ "task_struct rss_stat MM_ANONPAGES",
+ RETURN_ON_ERROR))
+ continue;
+
+ rss_cache += sync_rss;
+
+ if (first == last)
+ break;
+ first++;
+ }
+ last->rss_cache = rss_cache;
}
+ rss += last->rss_cache;
tt->last_tgid = last;
}
}
diff --git a/task.c b/task.c
index a79ed0d96fb5..864c838637ee 100644
--- a/task.c
+++ b/task.c
@@ -2947,6 +2947,7 @@ add_context(ulong task, char *tp)
tg = tt->tgid_array + tt->running_tasks;
tg->tgid = *tgid_addr;
tg->task = task;
+ tg->rss_cache = UNINITIALIZED;
if (do_verify && !verify_task(tc, do_verify)) {
error(INFO, "invalid task address: %lx\n", tc->task);
--
2.20.1