From: Cyril Hrubis <chrubis@suse.cz>
To: ltp@lists.linux.it
Subject: [LTP] [PATCH 2/2] nfs05: rewrite the test, make use of new library
Date: Wed, 22 Jun 2016 14:30:17 +0200 [thread overview]
Message-ID: <20160622123017.GC13962@rei.lan> (raw)
In-Reply-To: <1466076425-12790-2-git-send-email-alexey.kodanev@oracle.com>
Hi!
> --- a/testcases/network/nfs/nfs_stress/Makefile
> +++ b/testcases/network/nfs/nfs_stress/Makefile
> @@ -19,7 +19,7 @@ top_srcdir ?= ../../../..
> include $(top_srcdir)/include/mk/env_pre.mk
>
> nfs04_create_file: CPPFLAGS += -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> -nfs05_make_tree: LDLIBS += -lpthread
> +nfs05_make_tree: LDLIBS += -I../../../kernel/include -lltp -lpthread
Hmm, shouldn't we rather include testcase.mk to get the -lltp and
the include path?
And even if we don't the -I directive should go to CPPFLAGS instead.
> INSTALL_TARGETS := nfs_lib.sh \
> nfs01 \
> diff --git a/testcases/network/nfs/nfs_stress/nfs05 b/testcases/network/nfs/nfs_stress/nfs05
> index 0e15c8c..26977ab 100755
> --- a/testcases/network/nfs/nfs_stress/nfs05
> +++ b/testcases/network/nfs/nfs_stress/nfs05
> @@ -34,8 +34,12 @@ THREAD_NUM=${THREAD_NUM:-"8"}
> nfs_setup
Shouldn't we check that make and gcc is installed at this point?
> tst_resm TINFO "start nfs05_make_tree -d $DIR_NUM -f $FILE_NUM -t $THREAD_NUM"
> -ROD nfs05_make_tree -d $DIR_NUM -f $FILE_NUM -t $THREAD_NUM
> +TMPDIR=$(pwd) nfs05_make_tree -d $DIR_NUM -f $FILE_NUM -t $THREAD_NUM
>
> -tst_resm TPASS "test finished"
> +if [ $? -ne 0 ]; then
> + tst_resm TFAIL "'make' test failed"
> +else
> + tst_resm TPASS "'make' test finished"
> +fi
>
> tst_exit
> diff --git a/testcases/network/nfs/nfs_stress/nfs05_make_tree.c b/testcases/network/nfs/nfs_stress/nfs05_make_tree.c
> index 4163988..5d128dd 100644
> --- a/testcases/network/nfs/nfs_stress/nfs05_make_tree.c
> +++ b/testcases/network/nfs/nfs_stress/nfs05_make_tree.c
> @@ -1,179 +1,50 @@
> -/******************************************************************************/
> -/* */
> -/* Copyright (c) International Business Machines Corp., 2001 */
> -/* */
> -/* This program is free software; you can redistribute it and/or modify */
> -/* it under the terms of the GNU General Public License as published by */
> -/* the Free Software Foundation; either version 2 of the License, or */
> -/* (at your option) any later version. */
> -/* */
> -/* This program is distributed in the hope that it will be useful, */
> -/* but WITHOUT ANY WARRANTY; without even the implied warranty of */
> -/* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See */
> -/* the GNU General Public License for more details. */
> -/* */
> -/* You should have received a copy of the GNU General Public License */
> -/* along with this program; if not, write to the Free Software */
> -/* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */
> -/* */
> -/******************************************************************************/
> -
> -/******************************************************************************/
> -/* */
> -/* History: Oct - 10 - 2001 Created - Manoj Iyer, IBM Austin TX. */
> -/* email:manjo@austin.ibm.com */
> -/* - create a directory tree that is */
> -/* unique to each process. The base directory */
> -/* looks like hostname.<pid of the process> */
> -/* the subdirectories will be <pid>.0 <pid.1> etc*/
> -/* eg: */
> -/* hostname.1234 */
> -/* |_ 1234.0 */
> -/* |_ 1234.1 */
> -/* |_1234.2 */
> -/* |.... */
> -/* hostname - hostname of the machine */
> -/* 1234 - pid of the current process. */
> -/* Each of these directories are populated with */
> -/* N number of ".c" files and a makefile that can*/
> -/* compile the ".c" files and also initiate */
> -/* compile of ".c" files in the subdirectories */
> -/* under it. */
> -/* */
> -/* Oct - 11 - 2001 Modified */
> -/* - fixed a bug in the makefiles, the last make-*/
> -/* file was expecting subdirectories. Added */
> -/* code to generate a different makefile for */
> -/* the last subdirectory. */
> -/* - Added logic to first compile all the c files*/
> -/* and upon completion remove them. */
> -/* - Added multithreading, arguments handling. */
> -/* By default the program will generate 8 */
> -/* threads, each creating by default 100 deep */
> -/* directory tree each containing default 100 */
> -/* ".c" files and one makefile. */
> -/* - Added usage message. */
> -/* */
> -/* Oct - 12 - 2001 Modified */
> -/* - Added logic to print missing arguments to */
> -/* options. */
> -/* */
> -/* Oct - 15 - 2001 Modified */
> -/* - Added logic to remove the files, makefiles */
> -/* and subdirectories that were created. */
> -/* - Added logic to print debug messages. */
> -/* */
> -/* Oct - 16 - 2001 Modified */
> -/* - Added sync() calls to commit changes. */
> -/* - Fixed bug. pthread_join() returns 0 when */
> -/* pthread_join fails, if the thread function */
> -/* fails pthread_join() will put the exit value*/
> -/* of the thread function in the thread_return */
> -/* output argument. */
> -/* - Debugging function crte_mk_rm fails to */
> -/* create fies, problem appears only in multi- */
> -/* threaded case. */
> -/* */
> -/* Oct - 17 - 2001 Checked in */
> -/* - GPL statement was added and the initial ver */
> -/* - checked into CVS. */
> -/* - note: this version works only if it is run */
> -/* single threaded, when its run multithreaded */
> -/* random thread will fail on open() sys call */
> -/* problem currently under investigation. */
> -/* */
> -/* Oct - 20 - 2001 Modified */
> -/* - fixed a whole bunch of problems. */
> -/* - created function init_compile. Apparently */
> -/* this code works!!. */
> -/* - removed system() system call that was doing */
> -/* make and make clean. init_compile() replaces*/
> -/* this piece of code. */
> -/* - on supplying the full pathname to unlink() */
> -/* solved most of the problems with rm_file_mk */
> -/* function. */
> -/* - reset the default vaulues for MAXT = 8 */
> -/* MAXD = 100 and MAXF = 100. */
> -/* ie. maximum number of threads = 8 */
> -/* directory depth (num of sub dirs) = 100 */
> -/* numeber of .c fils in each sub dir = 100*/
> -/* - finally program is now in working state. */
> -/* */
> -/* Nov - 01 - 2001 Modified. */
> -/* - fixed usage message default MAXT is 8 not 1 */
> -/* - fixed make to compile the files silently */
> -/* */
> -/* Nov - 19 - 2001 Modified. */
> -/* - changed th_status in function main() from */
> -/* dynamic variable to static array. */
> -/* */
> -/* File: make_tree.c */
> -/* */
> -/* Description: This program is designed stress the NFS implimentation. */
> -/* Many bugs were uncovered in the AIX operating system */
> -/* implimentation of NFS when AIX kernel was built over NFS. */
> -/* Source directory on a remote machine (one server many clients)*/
> -/* NFS-mounted on to a directory on a local machine from which */
> -/* the kernel build was initiated. Apparently many defects/bugs */
> -/* were uncovered when multiple users tried to build the kernel */
> -/* by NFS mounting the kernel source from a remote machine and */
> -/* tried to build the kernel on a local machine. AIX build envi- */
> -/* ronment is set up to create the object files and executable */
> -/* on the local machine. */
> -/* This testcase will try to recreate such a senario. */
> -/* Spawn N number of threads. Each thread does the following. */
> -/* * Create a directory tree. */
> -/* * Populate it with ".c" files and makefiles. */
> -/* * initate a build. Executable will print hello world when exed*/
> -/* * clean up all the executables that were created. */
> -/* * recurssively remove each subdir and its contents. */
> -/* The test is aimed at stressing the NFS client and server. */
> -/* hostname.1234 */
> -/* | */
> -/* | - 1234.0.0.c */
> -/* | - 1234.0.1.c */
> -/* | - .......... */
> -/* | - makefile */
> -/* | */
> -/* |_ 1234.0 */
> -/* | */
> -/* | - 1234.1.0.c */
> -/* | - 1234.1.1.c */
> -/* | - .......... */
> -/* | - makefile */
> -/* | */
> -/* |_ 1234.1 */
> -/* | */
> -/* | - 1234.2.0.c */
> -/* | - 1234.2.1.c */
> -/* | - .......... */
> -/* | - makefile */
> -/* | */
> -/* |_1234.2 */
> -/* |.... */
> -/* */
> -/* Setup: - on the server side: */
> -/* * create a directory /nfs_test */
> -/* * make an entry in /etc/exports file like this... */
> -/* "/nfs_test *(rw,no_root_squash)" */
> -/* * run command "exportfs -a" */
> -/* - on client side: */
> -/* * create a directory say for eg: /nfs_cli */
> -/* * mount -t nfs servername:/nfs_test /nfs_cli */
> -/* * set up the tescase in /nfs_cli directory */
> -/* - I reccomend that you have atleast 8 client machines running */
> -/* this test, linux has 8 NFSD's running by default, you might*/
> -/* have to increase it as per your requirement. */
> -/* */
> -/* Note: - assumed that NFS services are installed and configured */
> -/* - you have atleast 2 machines to act as client and server */
> -/* (you can have muiltiple client machines and one server) */
> -/* - large amount of disk space, this depends on the number of */
> -/* of clients you will have, if you have only one client, I */
> -/* reccomend that the server have atleast 4 Giga bytes of */
> -/* disk space (paranoid!). */
> -/* */
> -/******************************************************************************/
> +/*
> + * Copyright (c) 2016 Oracle and/or its affiliates. All Rights Reserved.
> + * Copyright (c) International Business Machines Corp., 2001
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License as
> + * published by the Free Software Foundation; either version 2 of
> + * the License, or (at your option) any later version.
> + *
> + * 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. See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program. If not, see <http://www.gnu.org/licenses/>.
> +
> + * Description:
> + * This program is designed to stress the NFS implimentation. Many bugs were
> + * uncovered in the AIX operating system implimentation of NFS when AIX kernel
> + * was built over NFS. Source directory on a remote machine (one server many
> + * clients) NFS-mounted on to a directory on a local machine from which the
> + * kernel build was initiated. Apparently many defects/bugs were uncovered when
> + * multiple users tried to build the kernel by NFS mounting the kernel source
> + * from a remote machine and tried to build the kernel on a local machine.
> + *
> + * The test's aimed to stress NFS client/server and recreates such a senario.
> + * Spawn N number of threads. Each thread does the following:
> + * * create a directory tree;
> + * * populate it with ".c" files and makefiles;
> + * hostname.1234
> + * | - 1234.0.0.c
> + * | - ..........
> + * | - makefile
> + * |_ 1234.0
> + * |
> + * | - 1234.1.0.c
> + * | - ..........
> + * | - makefile
> + * |_ 1234.1
> + * |....
> + *
> + * * initate a build, executable will print hello world;
> + * * clean up all the executables that were created;
> + * * recurssively remove each subdir and its contents.
> + *
> + */
>
> #include <stdio.h>
> #include <sys/stat.h>
> @@ -188,609 +59,133 @@
> #include <errno.h>
> #include <linux/unistd.h>
>
> -#define gettid() syscall(__NR_gettid)
> -
> -#ifdef DEBUG
> -#define dprt(fmt, args...) printf(fmt, ## args)
> -#else
> -#define dprt(fmt, args...)
> -#endif
> -
> -#define MAKE_EXE 1 /* initate a make */
> -#define MAKE_CLEAN 0 /* initate a make clean */
> -
> -#define PTHREAD_EXIT(val) do {\
> - exit_val = val; \
> - dprt("pid[%d]: exiting with %d\n", gettid(),exit_val); \
> - pthread_exit((void *)exit_val); \
> - } while (0)
> -
> -#define OPT_MISSING(prog, opt) do{\
> - fprintf(stderr, "%s: option -%c ", prog, opt); \
> - fprintf(stderr, "requires an argument\n"); \
> - usage(prog); \
> - } while (0)
> -
> -#define MAXD 100 /* default number of directories to create. */
> -#define MAXF 100 /* default number of files to create. */
> -#define MAXT 8 /* default number of threads to create. */
> -
> -/******************************************************************************/
> -/* */
> -/* Function: usage */
> -/* */
> -/* Description: Print the usage message. */
> -/* */
> -/* Return: exits with -1 */
> -/* */
> -/******************************************************************************/
> -static void usage(char *progname)
> -{ /* name of this program */
> - fprintf(stderr,
> - "Usage: %s -d NUMDIR -f NUMFILES -h -t NUMTHRD\n"
> - "\t -d Number of subdirectories to generate: Default: 100\n"
> - "\t -f Number of c files in each subdirectory: Default: 100\n"
> - "\t -h Help!\n"
> - "\t -t Number of threads to generate: Default: 8\n",
> - progname);
> - exit(-1);
> -}
> -
> -/******************************************************************************/
> -/* */
> -/* Function: init_compile */
> -/* */
> -/* Description: This function compiles the .c files and removes the exeutables*/
> -/* This function does the same function as the system() system */
> -/* call, the code is available in the system() man page. When */
> -/* called with the parameter MAKE_EXE it will initiate make in */
> -/* the first directory created, the makefile is designed to build*/
> -/* recursively all the files in the subdirectories below. */
> -/* When called with the MAKE_CLEAN parameter it will remove the */
> -/* executables that were created design is similar to the case */
> -/* were it initiates a make. */
> -/* */
> -/* Return: exits with 1 on error, 0 on success */
> -/* */
> -/******************************************************************************/
> -static int init_compile(int what_todo, /* do a compile or clean */
> - char *base_dir, /* base directory of the test */
> - char *hname)
> -{ /* hostname of the machine */
> - int status; /* return status of execve process */
> - pid_t pid; /* pid of the process that does compile */
> - char *dirname; /* location where compile is initated */
> - char *command; /* make or make clean command. */
> -
> - if ((dirname = malloc(sizeof(char) * 2048)) == NULL) { /* just paranoid */
> - perror("init_compile(): dirname malloc()");
> - return 1;
> - }
> -
> - if ((command = malloc(sizeof(char) * 1024)) == NULL) { /* just paranoid */
> - perror("init_compile(): dirname malloc()");
> - return 1;
> - }
> -
> - what_todo ? sprintf(command, "make -s") : sprintf(command,
> - "make -s clean");
> +#include "lapi/mkdirat.h"
> +#include "tst_safe_stdio.h"
> +#include "tst_test.h"
>
> - sprintf(dirname, "%s/%s.%ld", base_dir, hname, gettid());
> +#define gettid() syscall(__NR_gettid)
>
> - if (chdir(dirname) == -1) {
> - dprt("pid[%d]: init_compile(): dir name = %s\n", gettid(),
> - dirname);
> - perror("init_compile() chdir()");
> - free(dirname);
> - return 1;
> - }
> +static int thrd_num = 8;
> +static int dirs_num = 100;
> +static int file_num = 100;
>
> - dprt("pid[%d]: init_compile(): command = %s\n", gettid(), command);
> +static char *t_arg, *d_arg, *f_arg;
>
> - if ((pid = fork()) == -1) {
> - perror("init_compile(): fork()");
> - return 1;
> - }
> - if (!pid) {
> - char *argv[4];
> +static struct tst_option opts[] = {
> + {"t:", &t_arg, "-t x Number of threads to generate, default: 8\n"},
> + {"d:", &d_arg, "-d x Number of subdirs to generate, default: 100\n"},
> + {"f:", &f_arg, "-f x Number of c files in each dir, default: 100\n"},
> + {NULL, NULL, NULL}
> +};
>
> - argv[0] = "/bin/sh";
> - argv[1] = "-c";
> - argv[2] = command;
> - argv[3] = 0;
> -
> - if (execv("/bin/sh", argv) == -1) {
> - perror("init_compile(): execv()");
> - return 1;
> - }
> - }
> - do {
> - if (waitpid(pid, &status, 0) == -1) {
> - if (errno != EINTR) {
> - fprintf(stderr,
> - "init_compile(): waitpid() failed\n");
> - return 1;
> - }
> +static void *thread_fn(LTP_ATTRIBUTE_UNUSED void *args)
> +{
> + const char prog_buf[] = "main()\n{\n\t printf(\"hello world\");\n}\n";
> + const size_t prog_buf_size = strlen(prog_buf);
^
sizeof(prog_buf) - 1
> + int i, k, fd, dirfd, len, ret;
> + char *dirname, *tmpdir;
> + char cfile[PATH_MAX];
> + char make_buf[1024];
> + char hostname[256];
> + pid_t tid = gettid();
> +
> + SAFE_GETHOSTNAME(hostname, 256);
> + SAFE_ASPRINTF(&dirname, "%s.%ld", hostname, tid);
> + SAFE_ASPRINTF(&tmpdir, "%ld", tid);
> +
> + SAFE_MKDIR(dirname, 0755);
> + dirfd = SAFE_OPEN(dirname, O_DIRECTORY);
> +
> + for (i = 0; i < dirs_num; ++i) {
> + if (i == dirs_num - 1) {
> + len = snprintf(make_buf, 1024,
> + "CFLAGS := -O -w -g\n"
> + "SUBDIRS = %s\n"
> + "SRCS=$(wildcard *.c)\n"
> + "TARGETS=$(patsubst %%.c,\%%,$(SRCS))\n"
> + "all:\t $(TARGETS)\n"
> + "clean:\n"
> + "\trm -f $(TARGETS)\n",
> + tmpdir);
The SUBDIRS variable seems to be unused here.
> } else {
> - if (chdir(base_dir) == -1) {
> - dprt("pid[%d]: init_compile(): dir = %s\n",
> - gettid(), dirname);
> - perror("init_compile(): chdir()");
> - return 1;
> + len = snprintf(make_buf, 1024,
> + "CFLAGS := -O -w -g\n"
> + "SUBDIRS = %s\n"
> + "SRCS=$(wildcard *.c)\n"
> + "TARGETS=$(patsubst %%.c,\%%,$(SRCS))\n"
> + "all:\t $(TARGETS)\n\t@for i in $(SUBDIRS);"
> + "do $(MAKE) -C $$i ; done\nclean:\n"
> + "\trm -f $(TARGETS)\n\t@for i in $(SUBDIRS);"
> + "do $(MAKE) -C $$i clean ; done\n", tmpdir);
> + }
> +
> + fd = openat(dirfd, "makefile", O_CREAT | O_RDWR,
> + S_IRWXU | S_IRWXG | S_IRWXO);
> + if (fd < 0)
> + tst_brk(TFAIL | TERRNO, "openat(makefile) failed");
> +
> + SAFE_WRITE(1, fd, make_buf, len);
> + SAFE_CLOSE(fd);
> +
> + for (k = 0; k < file_num; ++k) {
> + snprintf(cfile, PATH_MAX, "%d.%d.%d.c", tid, i, k);
> + fd = openat(dirfd, cfile, O_CREAT | O_RDWR,
> + S_IRWXU | S_IRWXG | S_IRWXO);
> + if (fd < 0) {
> + tst_brk(TFAIL | TERRNO,
> + "openat(%s) failed", cfile);
> }
>
> - dprt("pid[%d]: init_compile(): status = %d\n",
> - gettid(), status);
> - dprt("we are here %d\n", __LINE__);
> - return status;
> + SAFE_WRITE(1, fd, prog_buf, prog_buf_size);
> + fsync(fd);
Why do we fsync(fd) here? Isn't the whole point of this test to find a
race conditions?
--
Cyril Hrubis
chrubis@suse.cz
next prev parent reply other threads:[~2016-06-22 12:30 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-06-16 11:27 [LTP] [PATCH 1/2] lib: add safe_gethostname() Alexey Kodanev
2016-06-16 11:27 ` [LTP] [PATCH 2/2] nfs05: rewrite the test, make use of new library Alexey Kodanev
2016-06-22 12:30 ` Cyril Hrubis [this message]
2016-06-23 11:56 ` Alexey Kodanev
2016-06-23 11:54 ` Cyril Hrubis
2016-06-23 12:38 ` Alexey Kodanev
2016-06-23 12:41 ` Cyril Hrubis
2016-06-22 12:05 ` [LTP] [PATCH 1/2] lib: add safe_gethostname() Cyril Hrubis
2016-06-23 11:58 ` Alexey Kodanev
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=20160622123017.GC13962@rei.lan \
--to=chrubis@suse.cz \
--cc=ltp@lists.linux.it \
/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.