Skip to contentSkip to conversation

Broken kernel/mem/ksm01 test

LTP Lukáš Doktor 3 participants 4 messages Jul 20, 2012 – Jul 27, 2012

lore
Share & tools
lore

Lukáš Doktor

author
#0 lore raw

Hi guys,

I'm testing LTP-20120614 on Fedora 17 and it always stuck on ksm01 test.

I went through sources and the problem is I have KSM already enabled so the full_scans is a huge number. The function _group_check waits till KSM does 3 * full_scans which in my case never ends (well it might finish in a couple of years).

To me it doesn't even make sense to wait 3 * full_scans as in the first check it waits 1-2 seconds, in second check it waits few seconds and it keeps growing further you goes.

Would you please take look on this issue? I'd really appreciate it.

Kind regards,
Lukáš Doktor

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

author
#1 lore raw
Quoted 9 lines · Lukáš Doktor

Hi guys,

I'm testing LTP-20120614 on Fedora 17 and it always stuck on ksm01 test.

I went through sources and the problem is I have KSM already enabled so the full_scans is a huge number. The function _group_check waits till KSM does 3 * full_scans which in my case never ends (well it might finish in a couple of years).

Quoted 8 lines

To me it doesn't even make sense to wait 3 * full_scans as in the first check it waits 1-2 seconds, in second check it waits few seconds and it keeps growing further you goes.

Would you please take look on this issue? I'd really appreciate it.

Kind regards,
Lukáš Doktor

Anyway if you think it should stay 3 * full_scans would you please at least reset the counter before the test start?

mem.c 2012-07-20 14:58:52.192157167 +0200 renamed from _mem.c 2012-07-20 13:39:20.351738873 +0200 +1 -0
@@ -415,6 +415,7 @@ void create_same_memory(int size, int nu
                 }
         }
         tst_resm(TINFO, "KSM merging...");
+       write_file(PATH_KSM "run", "2");
         write_file(PATH_KSM "run", "1");
         snprintf(buf, BUFSIZ, "%ld", size * pages * num);
         write_file(PATH_KSM "pages_to_scan", buf);

I tested it and it works better (it fails, but at least it finishes). Anyway IMO it would be better to calculate the number of necessarily scans differently.

regards,
Lukáš

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

#2 lore raw

Zhouping, would you like to have a look at the issue below?

Caspar

Quoted 40 lines · Lukáš Doktor
Quoted 9 lines · Lukáš Doktor

Hi guys,

I'm testing LTP-20120614 on Fedora 17 and it always stuck on ksm01 test.

I went through sources and the problem is I have KSM already enabled so the full_scans is a huge number. The function _group_check waits till KSM does 3 * full_scans which in my case never ends (well it might finish in a couple of years).

Quoted 8 lines

To me it doesn't even make sense to wait 3 * full_scans as in the first check it waits 1-2 seconds, in second check it waits few seconds and it keeps growing further you goes.

Would you please take look on this issue? I'd really appreciate it.

Kind regards,
Lukáš Doktor

Anyway if you think it should stay 3 * full_scans would you please at least reset the counter before the test start?

mem.c 2012-07-20 14:58:52.192157167 +0200 renamed from _mem.c 2012-07-20 13:39:20.351738873 +0200 +1 -0
@@ -415,6 +415,7 @@ void create_same_memory(int size, int nu
                  }
          }
          tst_resm(TINFO, "KSM merging...");
+       write_file(PATH_KSM "run", "2");
          write_file(PATH_KSM "run", "1");
          snprintf(buf, BUFSIZ, "%ld", size * pages * num);
          write_file(PATH_KSM "pages_to_scan", buf);

I tested it and it works better (it fails, but at least it finishes). Anyway IMO it would be better to calculate the number of necessarily scans differently.

regards,
Lukáš

Quoted 11 lines

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

#3 lore raw
In reply to Lukáš Doktor

----- Original Message -----

Quoted 9 lines

From: "Lukáš Doktor" <[email protected]>
To: [email protected], "Cleber Rosa" <[email protected]>, [email protected]
Sent: Friday, July 20, 2012 8:45:32 PM
Subject: [LTP] Broken kernel/mem/ksm01 test

Hi guys,

I'm testing LTP-20120614 on Fedora 17 and it always stuck on ksm01 test.

I guess ksm service(user space) is on, which will set /sys/kernel/mm/ksm/run to 0, then full_scans will stop to count, this lead to the process(ksm01) stay in the following 'while()' all the time.

        while (new_num < old_num * 3) {
                sleep(1);
                read_file(PATH_KSM "full_scans", buf);
                new_num = SAFE_STRTOL(cleanup, buf, 0, LONG_MAX);
        }

that's why the testing stuck on ksm01.

so you need to stop ksm service:

 # systemctl stop ksm.service

after the ksm.service is stopped, I'm sure ksm01 can finish its job at least. also the ksm0* cases have some error[1], the latest version has fixed them. I suggest you use git version, which includes the fix(commit 60c41a5465cc5)

[1] http://sourceforge.net/mailarchive/message.php?msg_id=29088640

Quoted 31 lines

I went through sources and the problem is I have KSM already enabled so
the full_scans is a huge number. The function _group_check waits till KSM does 3 * full_scans which in my case never ends (well it might finish in a couple of years).

To me it doesn't even make sense to wait 3 * full_scans as in the first
check it waits 1-2 seconds, in second check it waits few seconds and it
keeps growing further you goes.

Would you please take look on this issue? I'd really appreciate it.

Kind regards,
Lukáš Doktor

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Signature · 13 lines
Thanks,
Zhouping

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list