Skip to contentSkip to conversation

splice02: Generate input in C

PATCH Superseded Petr Vorel 2 participants 7 messages Apr 13, 2021 – Apr 16, 2021

v3 of 6 Superseded Latest is v6 →

lore
Share & tools
lore

Petr Vorel

author
#0 lore raw

PATCH v3 1/2 splice02: Generate input in C

instead relying on shell piping data into it.

Found on SLES JeOS (https://progress.opensuse.org/issues/77260).

Also add metadata docs.

---
changes v2->v3:
* fix bugs found by Cyril (use !num_writes, !SAFE_FORK(), break if !TST_RET)

BTW I'm still not sure if break on !TST_RET is needed, but it does not harm. If you really like it, I can use while (!TST_RET).

Kind regards,
Petr

 runtest/smoketest                           |   2 +-
 runtest/syscalls                            |   2 +-
 testcases/kernel/syscalls/splice/splice02.c | 104 +++++++++++++++++---
 3 files changed, 94 insertions(+), 14 deletions(-)
runtest/smoketest +1 -1
index 0c24fc1fa..ec0c088cb 100644
@@ -11,5 +11,5 @@ symlink01 symlink01
 stat04 symlink01 -T stat04
 utime01A symlink01 -T utime01
 rename01A symlink01 -T rename01
-splice02 seq 1 20 | splice02
+splice02 splice02 -n 20
 route4-change-dst route-change-dst.sh
runtest/syscalls +1 -1
index 2d1e7b648..b89c545f0 100644
@@ -1451,7 +1451,7 @@ socketpair02 socketpair02
 sockioctl01 sockioctl01
 
 splice01 splice01
-splice02 seq 1 20000 | splice02
+splice02 splice02
 splice03 splice03
 splice04 splice04
 splice05 splice05
testcases/kernel/syscalls/splice/splice02.c +92 -12

132 lines not on this page · raw message

Signature · 1 line
2.31.1
author
#1 lore raw

PATCH v3 2/2 commands: Add shell pipe test

This adds basic shell pipe testing, which was removed from splice02 in 85cebe238 ("splice02: Generate input in C").

---
The same as v2.

 runtest/commands                         |  1 +
 runtest/smoketest                        |  1 +
 testcases/commands/shell/shell_pipe01.sh | 17 +++++++++++++++++
 3 files changed, 19 insertions(+)
 create mode 100755 testcases/commands/shell/shell_pipe01.sh
runtest/commands +1 -0
index 058266b54..8cfad0449 100644
@@ -41,3 +41,4 @@ gdb01_sh gdb01.sh
 unshare01_sh unshare01.sh
 sysctl01_sh sysctl01.sh
 sysctl02_sh sysctl02.sh
+shell_test01 echo "SUCCESS" | shell_pipe01.sh
runtest/smoketest +1 -0
index ec0c088cb..85f377192 100644
@@ -13,3 +13,4 @@ utime01A symlink01 -T utime01
 rename01A symlink01 -T rename01
 splice02 splice02 -n 20
 route4-change-dst route-change-dst.sh
+shell_test01 echo "SUCCESS" | shell_pipe01.sh
testcases/commands/shell/shell_pipe01.sh new file +17 -0
new file mode 100755
index 000000000..3c8ef49fb
@@ -0,0 +1,17 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0-or-later
+# Copyright (c) 2021 Petr Vorel <[email protected]>
+
+TST_TESTFUNC=do_test
+
+. tst_test.sh
+
+do_test()
+{
+	tst_res TINFO "expecting SUCCESS string passed from stdin"
+
+	read line
+	EXPECT_PASS [ "$line" = "SUCCESS" ]
+}
+
+tst_run
Signature · 1 line
2.31.1
#2 lore raw

PATCH v3 1/2 splice02: Generate input in C

In reply to Petr Vorel

Hi!

Quoted 6 lines

---
changes v2->v3:
* fix bugs found by Cyril (use !num_writes, !SAFE_FORK(), break if !TST_RET)

BTW I'm still not sure if break on !TST_RET is needed, but it does not harm. If you really like it, I can use while (!TST_RET).

I meant that we could simplify the test with:

testcases/kernel/syscalls/splice/splice02.c +4 -9
index fd71f2995..1b566a2f4 100644
@@ -54,18 +54,13 @@ static void do_child(void)
 
 	fd = SAFE_OPEN(TEST_FILENAME, O_WRONLY | O_CREAT | O_TRUNC, 0644);
 
-	while (to_write > 0) {
+	do {
 		TEST(splice(STDIN_FILENO, NULL, fd, NULL, WRITE_SIZE, 0));
-		tst_res(TINFO, "splice() wrote %ld, remaining %d", TST_RET, to_write);
 		if (TST_RET < 0) {
 			tst_res(TFAIL | TTERRNO, "splice failed");
 			goto cleanup;
-		} else if (!TST_RET) {
-			break;
-		} else {
-			to_write -= TST_RET;
 		}
-	}
+	} while (TST_RET > 0);
 
 	stat(TEST_FILENAME, &st);
 	if (st.st_size != num_writes) {
@@ -100,10 +95,10 @@ static void run(void)
 	for (i = 0; i < num_writes; i++)
 		SAFE_WRITE(1, pipe_fd[1], "x", 1);
 
-	tst_reap_children();
-
 	SAFE_CLOSE(pipe_fd[0]);
 	SAFE_CLOSE(pipe_fd[1]);
+
+	tst_reap_children();
 }
 
 static struct tst_test test = {

If you close the pipe in the parent before reaping children the last splice() will return 0 and exit the loop. No need to count anything.

Signature · 2 lines
Cyril Hrubis
[email protected]
author
#3 lore raw

PATCH v3 1/2 splice02: Generate input in C

Hi Cyril,

Quoted 42 lines

I meant that we could simplify the test with:

testcases/kernel/syscalls/splice/splice02.c +4 -9
index fd71f2995..1b566a2f4 100644
@@ -54,18 +54,13 @@ static void do_child(void)
 
 	fd = SAFE_OPEN(TEST_FILENAME, O_WRONLY | O_CREAT | O_TRUNC, 0644);
 
-	while (to_write > 0) {
+	do {
 		TEST(splice(STDIN_FILENO, NULL, fd, NULL, WRITE_SIZE, 0));
-		tst_res(TINFO, "splice() wrote %ld, remaining %d", TST_RET, to_write);
 		if (TST_RET < 0) {
 			tst_res(TFAIL | TTERRNO, "splice failed");
 			goto cleanup;
-		} else if (!TST_RET) {
-			break;
-		} else {
-			to_write -= TST_RET;
 		}
-	}
+	} while (TST_RET > 0);
 
 	stat(TEST_FILENAME, &st);
 	if (st.st_size != num_writes) {
@@ -100,10 +95,10 @@ static void run(void)
 	for (i = 0; i < num_writes; i++)
 		SAFE_WRITE(1, pipe_fd[1], "x", 1);
 
-	tst_reap_children();
-
 	SAFE_CLOSE(pipe_fd[0]);
 	SAFE_CLOSE(pipe_fd[1]);
+
+	tst_reap_children();
 }
 
 static struct tst_test test = {
Quoted 2 lines

If you close the pipe in the parent before reaping children the last splice() will return 0 and exit the loop. No need to count anything.

LGTM, going to merge it with these changes (+ cleanup to_write) and your Reviewed-by: tags.
Thanks a lot!

Kind regards,
Petr

author
#4 lore raw

PATCH v3 1/2 splice02: Generate input in C

Hi Cyril,

Quoted 3 lines

LGTM, going to merge it with these changes (+ cleanup to_write) and your Reviewed-by: tags.
Thanks a lot!

Maybe one more change: INT_MAX is a bit too much, it timeouts. I'll use 262144 as max allowed -n (the default on intel - 2x max pipe size).

Kind regards,
Petr

#5 lore raw

PATCH v3 1/2 splice02: Generate input in C

Hi!

Quoted 6 lines
Quoted 3 lines

LGTM, going to merge it with these changes (+ cleanup to_write) and your Reviewed-by: tags.
Thanks a lot!

Maybe one more change: INT_MAX is a bit too much, it timeouts. I'll use 262144 as max allowed -n (the default on intel - 2x max pipe size).

I guess that the main problem here is that we write one character at a time, which means that we spend most of the time in syscall handler code rather than copying the data. It may work much better if we change the code to write 512 bytes at a time. I.e. something as:

	#define BUFSIZE 512

	char buf[BUFSIZE];

	memset(buf, 'a', BUFSIZE);

	while (to_write > 0) {
		size_t size = to_write > BUFSIZE ? BUFSIZE : to_write;

		...

		to_write -= written;
	}
Signature · 2 lines
Cyril Hrubis
[email protected]
author
#6 lore raw

PATCH v3 1/2 splice02: Generate input in C

Hi Cyril,

...

Quoted 7 lines
Quoted 2 lines

Maybe one more change: INT_MAX is a bit too much, it timeouts. I'll use 262144 as max allowed -n (the default on intel - 2x max pipe size).

I guess that the main problem here is that we write one character at a time, which means that we spend most of the time in syscall handler code rather than copying the data. It may work much better if we change the code to write 512 bytes at a time. I.e. something as:

Sounds very reasonable, thx! I'll send v4.

Kind regards,
Petr

Quoted 13 lines
	#define BUFSIZE 512

	char buf[BUFSIZE];

	memset(buf, 'a', BUFSIZE);

	while (to_write > 0) {
		size_t size = to_write > BUFSIZE ? BUFSIZE : to_write;

		...

		to_write -= written;
	}