Hello chrubis,
On 04/15/2013 08:43 PM, [email protected] wrote:
> Hi!
>> +static void khugepaged_scan_done(void)
>> +{
>> + int changing = 1, count = 0;
>> + long old_pages_collapsed, old_max_ptes_none, old_pages_to_scan;
>> + long pages_collapsed = 0, max_ptes_none = 0, pages_to_scan = 0;
>> +
>> + while (changing) {
>> + /*
>> + * as 'khugepaged' run 100% during testing, so 10s
>> + * is an enough for us to recognize if 'khugepaged'
>> + * finish scanning proceses' anonymouse hugepages
> ^
> anonymous
Thanks for reviewing, I'll updated it on new version
>
>> + * or not.
>> + */
>> + sleep(10);
>> + count++;
>> +
>> + SAFE_FILE_SCANF(cleanup, PATH_KHPD "pages_collapsed",
>> + "%ld", &pages_collapsed);
>> + SAFE_FILE_SCANF(cleanup, PATH_KHPD "max_ptes_none",
>> + "%ld", &max_ptes_none);
>> + SAFE_FILE_SCANF(cleanup, PATH_KHPD "pages_to_scan",
>> + "%ld", &pages_to_scan);
>> +
>> + if (pages_collapsed != old_pages_collapsed ||
>> + max_ptes_none != old_max_ptes_none ||
>> + pages_to_scan != old_pages_to_scan) {
>> + old_pages_collapsed = pages_collapsed;
>> + old_max_ptes_none = max_ptes_none;
>> + old_pages_to_scan = pages_to_scan;
>> + } else {
>> + changing = 0;
>> + }
>> + }
>> +
>> + tst_resm(TINFO, "khugepaged daemon takes %ds to scan all thp pages",
>> + count * 10);
>> +}
> Ah, so it acutally does polling, this is fine.
>
> How long scanning takes. If it is about 10 seconds, we should poll
> faster so that the testcase has chance to finish reasonably fast.
> (if it takes 10 seconds and we miss the first windows it will sleep
> doing noting for another ten seconds)
so how about 5 seconds? I will test the value on different systems.
>
>> +static void verify_thp_size(int *child, int nr_child, int nr_thps)
>> +{
>> + FILE *fp;
>> + char path[BUFSIZ], buf[BUFSIZ], line[BUFSIZ];
>> + int i, ret;
>> + long expect_thps; /* the amount of per child's transparent hugepages */
>> + long val, actual_thps;
>> + long hugepagesize;
>> +
>> + hugepagesize = read_meminfo("Hugepagesize:");
>> + expect_thps = nr_thps * hugepagesize;
>> +
>> + for (i = 0; i < nr_child; i++) {
>> + actual_thps = 0;
>> +
>> + snprintf(path, BUFSIZ, "/proc/%d/smaps", child[i]);
>> + fp = fopen(path, "r");
>> + while (fgets(line, BUFSIZ, fp) != NULL) {
>> + ret = sscanf(line, "%64s %ld", buf, &val);
>> + if (ret == 2 && val != 0) {
>> + if (strcmp(buf, "AnonHugePages:") == 0)
>> + actual_thps += val;
>> + }
>> + }
>> +
>> + if (actual_thps != expect_thps)
>> + tst_resm(TFAIL, "child[%d] got %ldKB thps - expect %ld"
>> + "KB thps", getpid(), actual_thps, expect_thps);
>> + fclose(fp);
>> + }
>> +}
>> +
>> +void test_transparent_hugepage(int nr_child, int nr_thps, int hg_aligned)
>> +{
>> + unsigned long hugepagesize, memfree;
>> + int i, *pid, ret, status;
>> + char path[BUFSIZ];
>> +
>> + memfree = read_meminfo("MemFree:");
>> + tst_resm(TINFO, "The current MemFree is %luMB", memfree / KB);
>> + if (memfree < MB)
>> + tst_resm(TWARN, "The system has low memory, which maybe "
>> + "cause the case failed");
> Hmm, TWARN is most likely wrong here, as it gets propagated into the
> test return value. If it is known to fail on low memory, I would just
> exit with TCONF and "Not enough memory for testing" message. Comments
> anybody?
Agreed.
>
>> + hugepagesize = read_meminfo("Hugepagesize:");
>> +
>> + pid = malloc(nr_child * sizeof(int));
>> + if (pid == NULL)
>> + tst_brkm(TBROK | TERRNO, cleanup, "malloc");
> This is really small nit, but I would have named this array as 'pids'.
Agreed.
>
>> + for (i = 0; i < nr_child; i++) {
>> + switch (pid[i] = fork()) {
>> + case -1:
>> + tst_brkm(TBROK | TERRNO, cleanup, "fork");
>> +
>> + case 0:
>> + ret = alloc_transparent_hugepages(nr_thps, hg_aligned);
>> + exit(ret);
>> + }
>> + }
>> +
>> + tst_resm(TINFO, "Stop all children...");
>> + for (i = 0; i < nr_child; i++) {
>> + if (waitpid(pid[i], &status, WUNTRACED) == -1)
>> + tst_brkm(TBROK|TERRNO, cleanup, "waitpid");
>> + if (!WIFSTOPPED(status))
>> + tst_brkm(TBROK, cleanup,
>> + "child[%d] was not stoppted", pid[i]);
>> + }
>> +
>> + tst_resm(TINFO, "Start to scan all transparent hugepages...");
>> + khugepaged_scan_done();
>> +
>> + tst_resm(TINFO, "Start to verify transparent hugepage size...");
>> + verify_thp_size(pid, nr_child, nr_thps);
>> +
>> + tst_resm(TINFO, "Wake up all children...");
>> + for (i = 0; i < nr_child; i++) {
>> + if (kill(pid[i], SIGCONT) == -1)
>> + tst_brkm(TBROK | TERRNO, cleanup,
>> + "signal continue child[%d]", pid[i]);
>> + }
>> +
>> + /* wait all children finish himself task */
> ^
> their tasks
OK.
>> + for (i = 0; i < nr_child; i++) {
>> + if (waitpid(pid[i], &status, 0) == -1)
>> + tst_brkm(TBROK|TERRNO, cleanup, "waitpid %d", pid[i]);
>> +
>> + if (!WIFEXITED(status) || WEXITSTATUS(status) != 0)
>> + tst_resm(TFAIL, "the child[%d] unexpectedly failed:"
>> + " %d", pid[i], status);
>> + }
>> +}
>> +
>> +void check_thp_options(int *nr_child, int *nr_thps)
>> +{
>> + if (opt_nr_child)
>> + *nr_child = SAFE_STRTOL(NULL, opt_nr_child_str, 0, LONG_MAX);
>> + if (opt_nr_thps)
>> + *nr_thps = SAFE_STRTOL(NULL, opt_nr_thps_str, 0, LONG_MAX);
>> +}
>> +
>> +void thp_usage(void)
>> +{
>> + printf(" -n Number of processes\n");
>> + printf(" -N Number of transparent hugepages\n");
>> +}
>> +
>> /* cpuset/memcg */
>>
>> static void gather_node_cpus(char *cpus, long nd)
>> diff --git a/testcases/kernel/mem/thp/thp04.c
>> b/testcases/kernel/mem/thp/thp04.c
>> new file mode 100644
>> index 0000000..0f6c553
>> --- /dev/null
>> +++ b/testcases/kernel/mem/thp/thp04.c
>> @@ -0,0 +1,136 @@
>> +/*
>> + * Copyright (C) 2013 Linux Test Project
>> + *
>> + * This program is free software; you can redistribute it and/or
>> + * modify it under the terms of version 2 of the GNU General Public
>> + * License as published by the Free Software Foundation.
>> + *
>> + * This program is distributed in the hope that it would be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
>> + *
>> + * Further, this software is distributed without any warranty that it
>> + * is free of the rightful claim of any third person regarding
>> + * infringement or the like. Any license provided herein, whether
>> + * implied or otherwise, applies only to this software file. Patent
>> + * licenses, if any, provided herein do not apply to combinations of
>> + * this program with other software, or any other product whatsoever.
>> + *
>> + * You should have received a copy of the GNU General Public License
>> + * along with this program; if not, write the Free Software
>> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
>> + * 02110-1301, USA.
>> + */
>> +
>> +/*
>> + * The case is designed to test the functionality of transparent
>> + * hugepage - THP
>> + *
>> + * when one process allocate hugepage aligned anonymouse pages,
>> + * kernel thread 'khugepaged' controlled by sysfs knobs
>> + * /sys/kernel/mm/transparent_hugepage/ will scan them, and make
>> + * them as transparent hugepage if they are suited, you can find out
>> + * how many transparent hugepages are there in one process from
>> + * /proc/<pid>/smaps, among the file contents, 'AnonHugePages' entry
>> + * stand for transparent hugepage.
>> + */
>> +
>> +#include <sys/types.h>
>> +#include <sys/mman.h>
>> +#include <sys/stat.h>
>> +#include <sys/wait.h>
>> +#include <errno.h>
>> +#include <fcntl.h>
>> +#include <signal.h>
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <string.h>
>> +#include <unistd.h>
>> +#include "test.h"
>> +#include "usctest.h"
>> +#include "mem.h"
>> +
>> +char *TCID = "thp04";
>> +int TST_TOTAL = 1;
>> +
>> +option_t thp_options[] = {
>> + {"n:", &opt_nr_child, &opt_nr_child_str},
>> + {"N:", &opt_nr_thps, &opt_nr_thps_str},
>> + {NULL, NULL, NULL}
>> +};
>> +
>> +static int pre_thp_scan_sleep_millisecs;
>> +static int pre_thp_alloc_sleep_millisecs;
>> +static char pre_thp_enabled[BUFSIZ];
>> +
>> +int main(int argc, char *argv[])
>> +{
>> + int lc;
>> + char *msg;
>> + int nr_child = 2, nr_thps = 64;
>> +
>> + msg = parse_opts(argc, argv, thp_options, thp_usage);
>> + if (msg != NULL)
>> + tst_brkm(TBROK, tst_exit, "OPTION PARSING ERROR - %s", msg);
>> + check_thp_options(&nr_child, &nr_thps);
>> +
>> + setup();
>> +
>> + tst_resm(TINFO, "Start to test transparent hugepage...");
>> + tst_resm(TINFO, "There are %d children allocating %d "
>> + "transparent hugepages", nr_child, nr_thps);
>> +
>> + for (lc = 0; TEST_LOOPING(lc); lc++) {
>> + tst_count = 0;
>> +
>> + test_transparent_hugepage(nr_child, nr_thps, 1);
>> + }
>> +
>> + cleanup();
>> + tst_exit();
>> +}
>> +
>> +void setup(void)
>> +{
>> + char path[BUFSIZ];
>> +
>> + tst_require_root(NULL);
>> +
>> + if (access(PATH_THP, F_OK) == -1)
>> + tst_brkm(TCONF, NULL, "THP is not enabled");
>> +
>> + snprintf(path, BUFSIZ, PATH_KHPD "scan_sleep_millisecs");
>> + SAFE_FILE_SCANF(NULL, path, "%d", &pre_thp_scan_sleep_millisecs);
> The PATH_KPHD and "scan_sleep_millisecs" are both static strings and are
> merged before they are passed to the snprintf() at compile time, which
> results in printf format string __without__ any characters to
> substitute. This is really weird.
OK, I will remove snprintf()
>
>> + /* set 0 to khugepaged/scan_sleep_millisecs to run khugepaged 100% */
>> + SAFE_FILE_PRINTF(cleanup, path, "%d", 0);
> I'm confused here, this is the only function in the setup that takes a
> pointer to the cleanup defined bellow.
yeah, you are right, I will remove cleanup on new version.
>
>> + snprintf(path, BUFSIZ, PATH_KHPD "alloc_sleep_millisecs");
>> + SAFE_FILE_SCANF(NULL, path, "%d", &pre_thp_alloc_sleep_millisecs);
>> + /*
>> + * set 0 to khugepaged/alloc_sleep_millisecs to make sure khugepaged
>> + * don't stop if there's a hugepage allcation failure.
>> + */
>> + SAFE_FILE_PRINTF(NULL, path, "%d", 0);
>> +
>> + snprintf(path, BUFSIZ, PATH_THP "enabled");
> Here as well.
OK.
>
>> + write_file(path, "always");
> Where this came from? Does the function check for failures? Why not to
> use SAFE_FILE_PRINTF() as the rest of the file?
yes, there's some confused issue here.
but I have tried to use SAFE_FILE_PRINTF, but it didn't work fine for
transparent_hugepage/enabled,
if # cat /sys/kernel/mm/transparent_hugepage/enabled
always [madvise] never
SAFE_FILE_PRINTF only get 'always', expected 'madvise'
if # cat /sys/kernel/mm/transparent_hugepage/enabled
[always] madvise never
SAFE_FILE_PRINTF get '[always]', expected 'always'
I didn't find the reason why SAFE_FILE_PRINTF couldn't get the expected
value?
>
>> + tst_sig(FORK, DEF_HANDLER, NULL);
>> + TEST_PAUSE;
>> +}
>> +
>> +void cleanup(void)
>> +{
>> + char path[BUFSIZ];
>> +
>> + snprintf(path, BUFSIZ, PATH_KHPD "scan_sleep_millisecs");
>> + SAFE_FILE_PRINTF(NULL, path, "%d", pre_thp_scan_sleep_millisecs);
>> +
>> + snprintf(path, BUFSIZ, PATH_KHPD "alloc_sleep_millisecs");
>> + SAFE_FILE_PRINTF(NULL, path, "%d", pre_thp_alloc_sleep_millisecs);
>> +
>> + snprintf(path, BUFSIZ, PATH_THP "enabled");
>> + write_file(path, pre_thp_enabled);
>> +
> Same here.
Same with the previous reason.
>
>> + TEST_CLEANUP;
>> +}
>> --
>> 1.7.11.7
>>
>>
>> ------------------------------------------------------------------------------
>> Precog is a next-generation analytics platform capable of advanced
>> analytics on semi-structured data. The platform includes APIs for building
>> apps and a phenomenal toolset for data science. Developers can use
>> our toolset for easy data analysis & visualization. Get a free account!
>> http://www2.precog.com/precogplatform/slashdotnewsletter
>> _______________________________________________
>> Ltp-list mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/ltp-list
------------------------------------------------------------------------------
Precog is a next-generation analytics platform capable of advanced
analytics on semi-structured data. The platform includes APIs for building
apps and a phenomenal toolset for data science. Developers can use
our toolset for easy data analysis & visualization. Get a free account!
http://www2.precog.com/precogplatform/slashdotnewsletter
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
