packages/sysSentry
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

!96 fix ebpf collect iodump dead loop

Browse Source 
From: @znzjugod 
Reviewed-by: @gaoruoshu 
Signed-off-by: @gaoruoshu
This commit is contained in:
openeuler-ci-bot 2024-10-14 06:10:45 +00:00 committed by Gitee
commit a7a7340e1e
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
3 changed files with 122 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
add-ebpf-collector.patch
View File

@ -3478,4 +3478,5 @@ index ac35be2..a83bd9b 100644
if not lat_sec and not iodump_sec:
return common_param
--
2.33.0
2.33.0

111
ebpf-fix-dead-loop.patch Normal file
View File

@ -0,0 +1,111 @@
From b2881b6f3e441061afba3be5f9802b639fb4dca2 Mon Sep 17 00:00:00 2001
From: zhangnan <zhangnan134@huawei.com>
Date: Mon, 14 Oct 2024 12:01:27 +0800
Subject: [PATCH] ebpf fix dead loop
---
src/c/ebpf_collector/ebpf_collector.c | 53 +++++++++++++++++----------
1 file changed, 33 insertions(+), 20 deletions(-)
diff --git a/src/c/ebpf_collector/ebpf_collector.c b/src/c/ebpf_collector/ebpf_collector.c
index 6e981da..0885a5f 100644
--- a/src/c/ebpf_collector/ebpf_collector.c
+++ b/src/c/ebpf_collector/ebpf_collector.c
@@ -117,41 +117,46 @@ char* find_device_name(dev_t dev) {
return device_name;
}
-static int print_map_res(struct bpf_map *map_res, struct bpf_map *map_res_2, char *stage, int *map_size)
-{
- int err;
- struct stage_data counter;
+static void update_io_dump(struct bpf_map *map_res, int *io_dump, int *map_size, char *stage) {
struct time_range_io_count time_count;
- int key = 0;
- int io_dump[MAP_SIZE] = {0};
u32 io_dump_key = 0, io_dump_next_key = 0;
-
struct sysinfo info;
sysinfo(&info);
- while (bpf_map_get_next_key(map_res_2, &io_dump_key, &io_dump_next_key) == 0) {
- err = bpf_map_lookup_elem(map_res_2, &io_dump_next_key, &time_count);
+ while (bpf_map_get_next_key(map_res, &io_dump_key, &io_dump_next_key) == 0) {
+ int err = bpf_map_lookup_elem(map_res, &io_dump_next_key, &time_count);
if (err < 0) {
fprintf(stderr, "failed to lookup %s io dump: %d\n", stage, err);
- continue;
+ break;
+ }
+ if (io_dump_key == io_dump_next_key) {
+ break;
}
+
io_dump_key = io_dump_next_key;
- if ((info.uptime - io_dump_key) > 2) {
+
+ if ((info.uptime - io_dump_key) >= 2) {
int isempty = 1;
- for (key = 0; key < map_size; key++){
+ for (int key = 0; key < map_size; key++) {
if (time_count.count[key] > 0) {
io_dump[key] += time_count.count[key];
isempty = 0;
}
}
if (isempty || (info.uptime - io_dump_key) > IO_DUMP_THRESHOLD) {
- bpf_map_delete_elem(map_res_2, &io_dump_key);
+ bpf_map_delete_elem(map_res, &io_dump_key);
}
}
}
+}
+
+static int print_map_res(struct bpf_map *map_res, char *stage, int *map_size, int *io_dump)
+{
+ struct stage_data counter;
+ int key = 0;
for (key = 0; key < map_size; key++) {
- err = bpf_map_lookup_elem(map_res, &key, &counter);
+ int err = bpf_map_lookup_elem(map_res, &key, &counter);
if (err < 0) {
fprintf(stderr, "failed to lookup %s map_res: %d\n", stage, err);
return -1;
@@ -274,19 +279,27 @@ int main(int argc, char **argv) {
sleep(1);
- err = print_map_res(BLK_RES, BLK_RES_2, "rq_driver", device_count);
+ int io_dump_blk[MAP_SIZE] = {0};
+ update_io_dump(BLK_RES_2, io_dump_blk, device_count,"rq_driver");
+ err = print_map_res(BLK_RES, "rq_driver", device_count, io_dump_blk);
if (err)
break;
- err = print_map_res(BIO_RES, BIO_RES_2, "bio", device_count);
+ int io_dump_bio[MAP_SIZE] = {0};
+ update_io_dump(BIO_RES_2, io_dump_bio, device_count,"bio");
+ err = print_map_res(BIO_RES, "bio", device_count, io_dump_bio);
if (err)
break;
-
- err = print_map_res(TAG_RES, TAG_RES_2, "gettag", device_count);
+
+ int io_dump_tag[MAP_SIZE] = {0};
+ update_io_dump(TAG_RES_2, io_dump_tag, device_count,"gettag");
+ err = print_map_res(TAG_RES, "gettag", device_count, io_dump_tag);
if (err)
break;
-
- err = print_map_res(WBT_RES, WBT_RES_2, "wbt", device_count);
+
+ int io_dump_wbt[MAP_SIZE] = {0};
+ update_io_dump(WBT_RES_2, io_dump_wbt, device_count,"wbt");
+ err = print_map_res(WBT_RES, "wbt", device_count, io_dump_wbt);
if (err)
break;
--
2.33.0

11
sysSentry.spec
View File

@ -4,7 +4,7 @@
Summary: System Inspection Framework
Name: sysSentry
Version: 1.0.2
Release: 41
Release: 42
License: Mulan PSL v2
Group: System Environment/Daemons
Source0: https://gitee.com/openeuler/sysSentry/releases/download/v%{version}/%{name}-%{version}.tar.gz
@ -57,6 +57,7 @@ Patch44: add-root-cause-analysis.patch
Patch45: update-collect-log.patch
Patch46: modify-abnormal-stack-when-the-disk-field-is-not-con.patch
Patch47: precise-alarm-query-time.patch
Patch48: ebpf-fix-dead-loop.patch
BuildRequires: cmake gcc-c++
BuildRequires: python3 python3-setuptools
@ -329,6 +330,12 @@ rm -rf %{buildroot}
%attr(0550,root,root) %{python3_sitelib}/sentryCollector/__pycache__/collect_plugin*
%changelog
* Mon Oct 14 2024 zhangnan <zhangnan134@huawei.com> - 1.0.2-42
- Type:bugfix
- CVE:NA
- SUG:NA
- DESC:fix dead loop when find next bpf map key
* Mon Oct 14 2024 zhuofeng <zhuofeng2@huawei.com> - 1.0.2-41
- Type:bugfix
- CVE:NA
@ -378,7 +385,7 @@ rm -rf %{buildroot}
- SUG:NA
- DESC:add parameter valication for time_range and alarm_id and alarm_clear_time
* Thu Oct 11 2024 gaoruoshu <gaoruoshu@huawei.com> - 1.0.2-33
* Fri Oct 11 2024 gaoruoshu <gaoruoshu@huawei.com> - 1.0.2-33
- Type:requirement
- CVE:NA
- SUG:NA