From: Cyril Hrubis <chrubis@suse.cz>
To: Li Wang <liwang@redhat.com>
Cc: ltp-list@lists.sourceforge.net
Subject: Re: [LTP] [PATCH] futex: regression test for take hugepages into account when generating futex_key
Date: Tue, 16 Jun 2015 15:21:09 +0200 [thread overview]
Message-ID: <20150616132109.GA17408@rei.suse.de> (raw)
In-Reply-To: <1434444799-29951-1-git-send-email-liwang@redhat.com>
Hi!
> +#if __i386__ || __x86_64__
> +#define HUGE_SIZE (2 * 1024 * 1024)
> +
> +#elif __powerpc__ || __powerpc64__
> +#define HUGE_SIZE (16 * 1024 * 1024)
> +
> +#elif __s390__ || __s390x__
> +#define HUGE_SIZE (1 * 1024 * 1024)
> +
> +#else
> +#define HUGE_SIZE (2 * 1024 * 1024)
> +#endif
Can we use Hugepagesize from /proc/meminfo instead?
> +#define PATH_NR_HUGEPAGES "/proc/sys/vm/nr_hugepages"
> +
> +const char *TCID = "futex_wake04";
> +const int TST_TOTAL = 1;
> +
> +static futex_t *futex1, *futex2;
> +
> +static long th2_wait_time;
> +static int th2_wait_done;
> +
> +static long orig_hugepages;
> +
> +static void setup(void)
> +{
> + tst_require_root(NULL);
> + tst_tmpdir();
> +
> + SAFE_FILE_SCANF(NULL, PATH_NR_HUGEPAGES, "%ld", &orig_hugepages);
> + SAFE_FILE_PRINTF(NULL, PATH_NR_HUGEPAGES, "%d", 1);
> +
> + TEST_PAUSE;
> +}
> +
> +static void cleanup(void)
> +{
> + SAFE_FILE_PRINTF(NULL, PATH_NR_HUGEPAGES, "%ld", orig_hugepages);
> +
> + tst_rmdir();
> +}
> +
> +static void *wait_thread1(void *arg LTP_ATTRIBUTE_UNUSED)
> +{
> + futex_wait(futex1, *futex1, NULL, 0);
> +
> + return NULL;
> +}
> +
> +static void *wait_thread2(void *arg LTP_ATTRIBUTE_UNUSED)
> +{
> + struct timeval tv;
> +
> + gettimeofday(&tv, NULL);
> + th2_wait_time = tv.tv_sec;
> + futex_wait(futex2, *futex2, NULL, 0);
Eh, futex_wait can have timeout parameter. Why don't you set the timeout
here and fail the test if the futex timeouted? That would be the
simplest solution.
> + th2_wait_done = 1;
> + tst_resm(TPASS, "Hi hydra, thread2 awake!");
> +
> + return NULL;
> +}
> +
> +static void wakeup_thread2(void)
> +{
> + void *addr;
> + int pgsz, wait_max_time = 30;
> + pthread_t th1, th2;
> + struct timeval tv;
> +
> + /*allocate some shared memory*/
> + addr = mmap(NULL,
> + HUGE_SIZE,
> + PROT_READ | PROT_WRITE,
> + MAP_SHARED | MAP_ANONYMOUS | MAP_HUGETLB,
> + -1,
> + 0);
> + if (addr == MAP_FAILED) {
> + printf("errno=%d\n", errno);
> + perror("mmap");
> + if (errno == ENOMEM) {
> + tst_brkm(TBROK | TERRNO, NULL,
> + "Probably system didn't actually create any huge pages.");
> + }
> + }
> +
> + pgsz = getpagesize();
> +
> + /*apply the first subpage to futex1*/
> + futex1 = addr;
> + *futex1 = 0;
> + /*apply the second subpage to futex2*/
> + futex2 = (futex_t *)((char *)addr + pgsz);
> + *futex2 = 0;
> +
> + /*thread1 block on futex1 first,then thread2 block on futex2*/
> + pthread_create(&th1, NULL, wait_thread1, NULL);
> + sleep(2);
> + pthread_create(&th2, NULL, wait_thread2, NULL);
> + sleep(2);
No sleeps in testcases, you have to use proper synchronization
primitives.
If you need to run the test in several threads have a look at
futex_wake03.c how to assert that threads are sleeping on futex.
If the bug can be reproduced using child processes as well, you can use
TST_PROCESS_STATE_WAIT() to ensure children are sleeping in kernel and
tst_record_childstatus() to propagate the test result from child
process.
> + /*try to wake up thread2*/
> + futex_wake(futex2, 1, 0);
> +
> + /*see if thread2 can be woke up*/
> + while (!th2_wait_done) {
> + gettimeofday(&tv, NULL);
> + /*thread2 block over 30 secs, test fail*/
> + if (tv.tv_sec > (th2_wait_time + wait_max_time)) {
> + tst_resm(TFAIL,
> + "Bug: wait_thread2 did not wake after %ld secs.",
> + tv.tv_sec - th2_wait_time);
> + break;
> + }
> + }
> +
> + munmap(addr, HUGE_SIZE);
SAFE_MUNMAP()
You should also wake the second thread here and join both of them,
otherwise with test looping the program will create more and more
threads until it fails.
> +}
> +
> +int main(int argc, char *argv[])
> +{
> + int lc;
> +
> + tst_parse_opts(argc, argv, NULL, NULL);
> +
> + setup();
> +
> + for (lc = 0; TEST_LOOPING(lc); lc++)
> + wakeup_thread2();
> +
> + cleanup();
> + tst_exit();
> +}
> --
> 1.8.3.1
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Ltp-list mailing list
> Ltp-list@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/ltp-list
--
Cyril Hrubis
chrubis@suse.cz
------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list
next prev parent reply other threads:[~2015-06-16 13:22 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-06-16 8:53 [LTP] [PATCH] futex: regression test for take hugepages into account when generating futex_key Li Wang
2015-06-16 13:21 ` Cyril Hrubis [this message]
[not found] ` <266085602.15945003.1434547937242.JavaMail.zimbra@redhat.com>
2015-06-17 13:41 ` Cyril Hrubis
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20150616132109.GA17408@rei.suse.de \
--to=chrubis@suse.cz \
--cc=liwang@redhat.com \
--cc=ltp-list@lists.sourceforge.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.