Skip to contentSkip to conversation

kunit: fix failure to build without printk

PATCH Superseded Brendan Higgins 7 participants 21 messages Aug 28, 2019 – Sep 2, 2019

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 kunit: fix failure to build without printk

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

---
 include/kunit/test.h | 11 ++++-----
 kunit/test.c         | 57 +++++---------------------------------------
 2 files changed, 11 insertions(+), 57 deletions(-)
include/kunit/test.h +5 -6
index 8b7eb03d4971..efad2eacd6ba 100644
@@ -339,9 +339,8 @@ static inline void *kunit_kzalloc(struct kunit *test, size_t size, gfp_t gfp)
 
 void kunit_cleanup(struct kunit *test);
 
-void __printf(3, 4) kunit_printk(const char *level,
-				 const struct kunit *test,
-				 const char *fmt, ...);
+#define kunit_print_level(KERN_LEVEL, test, fmt, ...) \
+	printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)
 
 /**
  * kunit_info() - Prints an INFO level message associated with @test.
@@ -353,7 +352,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Takes a variable number of format parameters just like printk().
  */
 #define kunit_info(test, fmt, ...) \
-	kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
+	kunit_print_level(KERN_INFO, test, fmt, ##__VA_ARGS__)
 
 /**
  * kunit_warn() - Prints a WARN level message associated with @test.
@@ -364,7 +363,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints a warning level message.
  */
 #define kunit_warn(test, fmt, ...) \
-	kunit_printk(KERN_WARNING, test, fmt, ##__VA_ARGS__)
+	kunit_print_level(KERN_WARNING, test, fmt, ##__VA_ARGS__)
 
 /**
  * kunit_err() - Prints an ERROR level message associated with @test.
@@ -375,7 +374,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints an error level message.
  */
 #define kunit_err(test, fmt, ...) \
-	kunit_printk(KERN_ERR, test, fmt, ##__VA_ARGS__)
+	kunit_print_level(KERN_ERR, test, fmt, ##__VA_ARGS__)
 
 /**
  * KUNIT_SUCCEED() - A no-op expectation. Only exists for code clarity.
kunit/test.c +6 -51

92 lines not on this page · raw message

Signature · 1 line
2.23.0.187.g17f5b7556c-goog
#1 lore raw

On (08/28/19 02:31), Brendan Higgins wrote:
[..]

Quoted 9 lines

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

[..]

kunit/test.c static void kunit_set_failure(struct kunit *test) · line 41

Quoted 6 lines
kunit/test.c +0 -6
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}

This patch looks good to me. I like the removal of recursive vsprintf() (%pV).

	-ss
#2 lore raw
Quoted 14 lines · Sergey Senozhatsky

On (08/28/19 02:31), Brendan Higgins wrote:
[..]

Quoted 9 lines

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

[..]

Quoted 9 lines

kunit/test.c static void kunit_set_failure(struct kunit *test) · line 41

Quoted 6 lines
kunit/test.c +0 -6
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}

This patch looks good to me. I like the removal of recursive vsprintf() (%pV).

Same here. And I am happy that we did not add more external vprintk_emit() callers. It would be great to rework dev_printk() as well.

Best Regards,
Petr

maintainer
#3 lore raw
In reply to Sergey Senozhatsky
Quoted 24 lines · Sergey Senozhatsky

On (08/28/19 02:31), Brendan Higgins wrote:
[..]

Quoted 9 lines

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

[..]

kunit/test.c static void kunit_set_failure(struct kunit *test) · line 41

Quoted 6 lines
kunit/test.c +0 -6
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}

This patch looks good to me. I like the removal of recursive vsprintf() (%pV).

Quoted 2 lines
	-ss

Hi Sergey,

What are the guidelines for using printk(). I recall some discussion about not using printk(). I am seeing the following from checkpatch script:

WARNING: Prefer [subsystem eg: netdev]_level([subsystem]dev, ... then dev_level(dev, ... then pr_level(... to printk(KERN_LEVEL ... #105: FILE: include/kunit/test.h:343:
+ printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)

Is there supposed to be pr_level() - I can find dev_level()

cc'ing Joe Perches for his feedback on this message recommending pr_level() which isn't in 5.3.

thanks,
-- Shuah

thanks,
-- Shuah

#4 lore raw
Quoted 41 lines · shuah
Quoted 24 lines · Sergey Senozhatsky

On (08/28/19 02:31), Brendan Higgins wrote:
[..]

Quoted 9 lines

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

[..]

Review

Quoted 6 lines
(quoted diff) +0 -6
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}

This patch looks good to me. I like the removal of recursive vsprintf() (%pV).

Quoted 2 lines
	-ss

Hi Sergey,

What are the guidelines for using printk(). I recall some discussion about not using printk(). I am seeing the following from checkpatch script:

WARNING: Prefer [subsystem eg: netdev]_level([subsystem]dev, ... then dev_level(dev, ... then pr_level(... to printk(KERN_LEVEL ... #105: FILE: include/kunit/test.h:343:

(quoted diff) +1 -0
+	printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)
Quoted 4 lines

Is there supposed to be pr_level() - I can find dev_level()

cc'ing Joe Perches for his feedback on this message recommending pr_level() which isn't in 5.3.

I don't care for pr_level or KERN_LEVEL in a printk.

I think this is somewhat overly complicated.

I think I'd write it like:

---
 include/kunit/test.h | 11 ++++-----
 kunit/test.c         | 69 ++++++++++++++++------------------------------------
 2 files changed, 26 insertions(+), 54 deletions(-)
include/kunit/test.h +5 -6
index 8b7eb03d4971..aa4abf0a22a5 100644
@@ -339,9 +339,8 @@ static inline void *kunit_kzalloc(struct kunit *test, size_t size, gfp_t gfp)
 
 void kunit_cleanup(struct kunit *test);
 
-void __printf(3, 4) kunit_printk(const char *level,
-				 const struct kunit *test,
-				 const char *fmt, ...);
+__printf(2, 3)
+void kunit_printk(const struct kunit *test, const char *fmt, ...);
 
 /**
  * kunit_info() - Prints an INFO level message associated with @test.
@@ -353,7 +352,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Takes a variable number of format parameters just like printk().
  */
 #define kunit_info(test, fmt, ...) \
-	kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_INFO fmt, ##__VA_ARGS__)
 
 /**
  * kunit_warn() - Prints a WARN level message associated with @test.
@@ -364,7 +363,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints a warning level message.
  */
 #define kunit_warn(test, fmt, ...) \
-	kunit_printk(KERN_WARNING, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_WARNING fmt, ##__VA_ARGS__)
 
 /**
  * kunit_err() - Prints an ERROR level message associated with @test.
@@ -375,7 +374,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints an error level message.
  */
 #define kunit_err(test, fmt, ...) \
-	kunit_printk(KERN_ERR, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_ERR fmt, ##__VA_ARGS__)
 
 /**
  * KUNIT_SUCCEED() - A no-op expectation. Only exists for code clarity.
kunit/test.c +21 -48

111 lines not on this page · raw message

#5 lore raw
Quoted 1 line · Joe Perches

On Thu, 2019-08-29 at 11:01 -0600, shuah wrote:

[]

Quoted 7 lines
Quoted 7 lines

WARNING: Prefer [subsystem eg: netdev]_level([subsystem]dev, ... then dev_level(dev, ... then pr_level(... to printk(KERN_LEVEL ... #105: FILE: include/kunit/test.h:343:

(quoted diff) +1 -0
+	printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)

Is there supposed to be pr_level() - I can find dev_level()

btw: the checkpatch message is meant to be interpreted as

Prefer [subsystem eg: netdev]_<level>([subsystem]dev, ...) then dev_<level>(dev, ...) then pr_<level>(...), to printk(KERN_<LEVEL> ...)

btw2:

dev_level is actually not a function, but a convenience macro argument which indirects to an actual specific logging function.

So no, there is not supposed to be a pr_level.

#6 lore raw
In reply to Joe Perches
Quoted 42 lines · Joe Perches
Quoted 41 lines · shuah
Quoted 24 lines · Sergey Senozhatsky

On (08/28/19 02:31), Brendan Higgins wrote:
[..]

Quoted 9 lines

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

[..]

kunit/test.c static void kunit_set_failure(struct kunit *test) · line 41

Quoted 6 lines
kunit/test.c +0 -6
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}

This patch looks good to me. I like the removal of recursive vsprintf() (%pV).

Quoted 2 lines
	-ss

Hi Sergey,

What are the guidelines for using printk(). I recall some discussion about not using printk(). I am seeing the following from checkpatch script:

WARNING: Prefer [subsystem eg: netdev]_level([subsystem]dev, ... then dev_level(dev, ... then pr_level(... to printk(KERN_LEVEL ... #105: FILE: include/kunit/test.h:343:

(quoted diff) +1 -0
+	printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)
Quoted 6 lines
Quoted 4 lines

Is there supposed to be pr_level() - I can find dev_level()

cc'ing Joe Perches for his feedback on this message recommending pr_level() which isn't in 5.3.

I don't care for pr_level or KERN_LEVEL in a printk.

I don't think I follow, how does your version fix this?

Quoted 161 lines

I think this is somewhat overly complicated.

I think I'd write it like:

---
 include/kunit/test.h | 11 ++++-----
 kunit/test.c         | 69 ++++++++++++++++------------------------------------
 2 files changed, 26 insertions(+), 54 deletions(-)
include/kunit/test.h +5 -6
index 8b7eb03d4971..aa4abf0a22a5 100644
@@ -339,9 +339,8 @@ static inline void *kunit_kzalloc(struct kunit *test, size_t size, gfp_t gfp)
 
 void kunit_cleanup(struct kunit *test);
 
-void __printf(3, 4) kunit_printk(const char *level,
-				 const struct kunit *test,
-				 const char *fmt, ...);
+__printf(2, 3)
+void kunit_printk(const struct kunit *test, const char *fmt, ...);
 
 /**
  * kunit_info() - Prints an INFO level message associated with @test.
@@ -353,7 +352,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Takes a variable number of format parameters just like printk().
  */
 #define kunit_info(test, fmt, ...) \
-	kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_INFO fmt, ##__VA_ARGS__)
 
 /**
  * kunit_warn() - Prints a WARN level message associated with @test.
@@ -364,7 +363,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints a warning level message.
  */
 #define kunit_warn(test, fmt, ...) \
-	kunit_printk(KERN_WARNING, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_WARNING fmt, ##__VA_ARGS__)
 
 /**
  * kunit_err() - Prints an ERROR level message associated with @test.
@@ -375,7 +374,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints an error level message.
  */
 #define kunit_err(test, fmt, ...) \
-	kunit_printk(KERN_ERR, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_ERR fmt, ##__VA_ARGS__)
 
 /**
  * KUNIT_SUCCEED() - A no-op expectation. Only exists for code clarity.
kunit/test.c +18 -47
index b2ca9b94c353..ddb9bffb5a5d 100644
@@ -16,40 +16,6 @@ static void kunit_set_failure(struct kunit *test)
 	WRITE_ONCE(test->success, false);
 }
 
-static int kunit_vprintk_emit(int level, const char *fmt, va_list args)
-{
-	return vprintk_emit(0, level, NULL, 0, fmt, args);
-}
-
-static int kunit_printk_emit(int level, const char *fmt, ...)
-{
-	va_list args;
-	int ret;
-
-	va_start(args, fmt);
-	ret = kunit_vprintk_emit(level, fmt, args);
-	va_end(args);
-
-	return ret;
-}
-
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}
-
-static void kunit_print_tap_version(void)
-{
-	static bool kunit_has_printed_tap_version;
-
-	if (!kunit_has_printed_tap_version) {
-		kunit_printk_emit(LOGLEVEL_INFO, "TAP version 14\n");
-		kunit_has_printed_tap_version = true;
-	}
-}
-
 static size_t kunit_test_cases_len(struct kunit_case *test_cases)
 {
 	struct kunit_case *test_case;
@@ -63,11 +29,9 @@ static size_t kunit_test_cases_len(struct kunit_case *test_cases)
 
 static void kunit_print_subtest_start(struct kunit_suite *suite)
 {
-	kunit_print_tap_version();
-	kunit_printk_emit(LOGLEVEL_INFO, "\t# Subtest: %s\n", suite->name);
-	kunit_printk_emit(LOGLEVEL_INFO,
-			  "\t1..%zd\n",
-			  kunit_test_cases_len(suite->test_cases));
+	pr_info_once("TAP version 14\n");
+	pr_info("\t# Subtest: %s\n", suite->name);
+	pr_info("\t1..%zd\n", kunit_test_cases_len(suite->test_cases));
 }
 
 static void kunit_print_ok_not_ok(bool should_indent,
@@ -87,9 +51,8 @@ static void kunit_print_ok_not_ok(bool should_indent,
 	else
 		ok_not_ok = "not ok";
 
-	kunit_printk_emit(LOGLEVEL_INFO,
-			  "%s%s %zd - %s\n",
-			  indent, ok_not_ok, test_number, description);
+	pr_info("%s%s %zd - %s\n",
+		indent, ok_not_ok, test_number, description);
 }
 
 static bool kunit_suite_has_succeeded(struct kunit_suite *suite)
@@ -133,11 +96,11 @@ static void kunit_print_string_stream(struct kunit *test,
 		kunit_err(test,
 			  "Could not allocate buffer, dumping stream:\n");
 		list_for_each_entry(fragment, &stream->fragments, node) {
-			kunit_err(test, fragment->fragment);
+			kunit_err(test, "%s", fragment->fragment);
 		}
 		kunit_err(test, "\n");
 	} else {
-		kunit_err(test, buf);
+		kunit_err(test, "%s", buf);
 		kunit_kfree(test, buf);
 	}
 }
@@ -505,19 +468,29 @@ void kunit_cleanup(struct kunit *test)
 	}
 }
 
-void kunit_printk(const char *level,
-		  const struct kunit *test,
-		  const char *fmt, ...)
+void kunit_printk(const struct kunit *test, const char *fmt, ...)
 {
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 	struct va_format vaf;
 	va_list args;
+	int kern_level;
 
 	va_start(args, fmt);
 
+	while ((kern_level = printk_get_level(fmt)) != 0) {
+		size_t size = printk_skip_level(fmt) - fmt;
+
+		if (kern_level >= '0' && kern_level <= '7') {
+			memcpy(lvl, fmt,  size);
+			lvl[size] = '\0';
+		}
+		fmt += size;

kunit/test.c void kunit_cleanup(struct kunit *test) · line 496

Quoted 10 lines
kunit/test.c +3 -1
+	}
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
 
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", lvl, test->name, &vaf);
 
 	va_end(args);
 }

How is this simpler?

If we are okay with dynamically adding the KERN_<LEVEL> and %pV (and I don't think that Sergey is), then wouldn't it be easier to pass in the kernel level as a separate parameter and then strip off all printk headers like this:

void kunit_printk(const char *level,
const struct kunit *test,
const char *fmt, ...)
{
struct va_format vaf;
va_list args;

va_start(args, fmt);

+ fmt = printk_skip_headers(fmt);
+
vaf.fmt = fmt;
vaf.va = &args;

- kunit_vprintk(test, level, &vaf);
+ printk("%s\t# %s %pV\n", level, test->name, &vaf);

va_end(args);
}

Then the kunit_printk function is much simpler, and I don't think my header file has to change at all.

I don't know. I am clearly not an expert on this topic, but I don't see the merit of the while loop you added above or dropping the level param.

#7 lore raw
Quoted 221 lines · Brendan Higgins
Quoted 42 lines · Joe Perches
Quoted 41 lines · shuah
Quoted 24 lines · Sergey Senozhatsky

On (08/28/19 02:31), Brendan Higgins wrote:
[..]

Quoted 9 lines

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

[..]

kunit/test.c static void kunit_set_failure(struct kunit *test) · line 41

Quoted 6 lines
kunit/test.c +0 -6
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}

This patch looks good to me. I like the removal of recursive vsprintf() (%pV).

Quoted 2 lines
	-ss

Hi Sergey,

What are the guidelines for using printk(). I recall some discussion about not using printk(). I am seeing the following from checkpatch script:

WARNING: Prefer [subsystem eg: netdev]_level([subsystem]dev, ... then dev_level(dev, ... then pr_level(... to printk(KERN_LEVEL ... #105: FILE: include/kunit/test.h:343:

(quoted diff) +1 -0
+	printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)
Quoted 6 lines
Quoted 4 lines

Is there supposed to be pr_level() - I can find dev_level()

cc'ing Joe Perches for his feedback on this message recommending pr_level() which isn't in 5.3.

I don't care for pr_level or KERN_LEVEL in a printk.

I don't think I follow, how does your version fix this?

Quoted 169 lines

I think this is somewhat overly complicated.

I think I'd write it like:

---
 include/kunit/test.h | 11 ++++-----
 kunit/test.c         | 69 ++++++++++++++++------------------------------------
 2 files changed, 26 insertions(+), 54 deletions(-)
include/kunit/test.h +5 -6
index 8b7eb03d4971..aa4abf0a22a5 100644
@@ -339,9 +339,8 @@ static inline void *kunit_kzalloc(struct kunit *test, size_t size, gfp_t gfp)
 
 void kunit_cleanup(struct kunit *test);
 
-void __printf(3, 4) kunit_printk(const char *level,
-				 const struct kunit *test,
-				 const char *fmt, ...);
+__printf(2, 3)
+void kunit_printk(const struct kunit *test, const char *fmt, ...);
 
 /**
  * kunit_info() - Prints an INFO level message associated with @test.
@@ -353,7 +352,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Takes a variable number of format parameters just like printk().
  */
 #define kunit_info(test, fmt, ...) \
-	kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_INFO fmt, ##__VA_ARGS__)
 
 /**
  * kunit_warn() - Prints a WARN level message associated with @test.
@@ -364,7 +363,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints a warning level message.
  */
 #define kunit_warn(test, fmt, ...) \
-	kunit_printk(KERN_WARNING, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_WARNING fmt, ##__VA_ARGS__)
 
 /**
  * kunit_err() - Prints an ERROR level message associated with @test.
@@ -375,7 +374,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints an error level message.
  */
 #define kunit_err(test, fmt, ...) \
-	kunit_printk(KERN_ERR, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_ERR fmt, ##__VA_ARGS__)
 
 /**
  * KUNIT_SUCCEED() - A no-op expectation. Only exists for code clarity.
kunit/test.c +21 -48
index b2ca9b94c353..ddb9bffb5a5d 100644
@@ -16,40 +16,6 @@ static void kunit_set_failure(struct kunit *test)
 	WRITE_ONCE(test->success, false);
 }
 
-static int kunit_vprintk_emit(int level, const char *fmt, va_list args)
-{
-	return vprintk_emit(0, level, NULL, 0, fmt, args);
-}
-
-static int kunit_printk_emit(int level, const char *fmt, ...)
-{
-	va_list args;
-	int ret;
-
-	va_start(args, fmt);
-	ret = kunit_vprintk_emit(level, fmt, args);
-	va_end(args);
-
-	return ret;
-}
-
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}
-
-static void kunit_print_tap_version(void)
-{
-	static bool kunit_has_printed_tap_version;
-
-	if (!kunit_has_printed_tap_version) {
-		kunit_printk_emit(LOGLEVEL_INFO, "TAP version 14\n");
-		kunit_has_printed_tap_version = true;
-	}
-}
-
 static size_t kunit_test_cases_len(struct kunit_case *test_cases)
 {
 	struct kunit_case *test_case;
@@ -63,11 +29,9 @@ static size_t kunit_test_cases_len(struct kunit_case *test_cases)
 
 static void kunit_print_subtest_start(struct kunit_suite *suite)
 {
-	kunit_print_tap_version();
-	kunit_printk_emit(LOGLEVEL_INFO, "\t# Subtest: %s\n", suite->name);
-	kunit_printk_emit(LOGLEVEL_INFO,
-			  "\t1..%zd\n",
-			  kunit_test_cases_len(suite->test_cases));
+	pr_info_once("TAP version 14\n");
+	pr_info("\t# Subtest: %s\n", suite->name);
+	pr_info("\t1..%zd\n", kunit_test_cases_len(suite->test_cases));
 }
 
 static void kunit_print_ok_not_ok(bool should_indent,
@@ -87,9 +51,8 @@ static void kunit_print_ok_not_ok(bool should_indent,
 	else
 		ok_not_ok = "not ok";
 
-	kunit_printk_emit(LOGLEVEL_INFO,
-			  "%s%s %zd - %s\n",
-			  indent, ok_not_ok, test_number, description);
+	pr_info("%s%s %zd - %s\n",
+		indent, ok_not_ok, test_number, description);
 }
 
 static bool kunit_suite_has_succeeded(struct kunit_suite *suite)
@@ -133,11 +96,11 @@ static void kunit_print_string_stream(struct kunit *test,
 		kunit_err(test,
 			  "Could not allocate buffer, dumping stream:\n");
 		list_for_each_entry(fragment, &stream->fragments, node) {
-			kunit_err(test, fragment->fragment);
+			kunit_err(test, "%s", fragment->fragment);
 		}
 		kunit_err(test, "\n");
 	} else {
-		kunit_err(test, buf);
+		kunit_err(test, "%s", buf);
 		kunit_kfree(test, buf);
 	}
 }
@@ -505,19 +468,29 @@ void kunit_cleanup(struct kunit *test)
 	}
 }
 
-void kunit_printk(const char *level,
-		  const struct kunit *test,
-		  const char *fmt, ...)
+void kunit_printk(const struct kunit *test, const char *fmt, ...)
 {
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 	struct va_format vaf;
 	va_list args;
+	int kern_level;
 
 	va_start(args, fmt);
 
+	while ((kern_level = printk_get_level(fmt)) != 0) {
+		size_t size = printk_skip_level(fmt) - fmt;
+
+		if (kern_level >= '0' && kern_level <= '7') {
+			memcpy(lvl, fmt,  size);
+			lvl[size] = '\0';
+		}
+		fmt += size;
+	}
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
 
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", lvl, test->name, &vaf);
 
Quoted 7 lines

kunit/test.c void kunit_cleanup(struct kunit *test) · line 496

Quoted 2 lines
kunit/test.c +0 -0
 	va_end(args);
 }

How is this simpler?

If we are okay with dynamically adding the KERN_<LEVEL> and %pV (and I don't think that Sergey is),

Sergey may well be in the minority overall as %pV is now very frequently
used throughout the kernel.

$ git grep "%pV" | wc -l
241

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall object size. Consolidated formats with the embedded KERN_<LEVEL> like suggested are smaller overall object size.

This style is also already used in the kernel.

Quoted 2 lines

I don't know. I am clearly not an expert on this topic, but I don't see the merit of the while loop you added above or dropping the level param.

The while use is only to avoid misuses with consecutive KERN_<LEVEL> formats, which had happened in the past.

#8 lore raw
Quoted 244 lines

-----Original Message-----
From: Joe Perches

Quoted 6 lines · Brendan Higgins
Quoted 5 lines · Joe Perches
Quoted 4 lines · shuah
Quoted 3 lines · Sergey Senozhatsky

On (08/28/19 02:31), Brendan Higgins wrote:
[..]

Quoted 1 line

Previously KUnit assumed that printk would always be present,

which is

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

[email protected]/T/#t

Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 3 lines
Cc Stephen Rothwell <[email protected]>
Cc Sergey Senozhatsky <[email protected]>

[..]

kunit/test.c static void kunit_set_failure(struct kunit *test)

Quoted 5 lines
kunit/test.c +0 -5
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name,

vaf);

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
Quoted 1 line

-}

This patch looks good to me. I like the removal of recursive vsprintf() (%pV).

	-ss
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Hi Sergey,

What are the guidelines for using printk(). I recall some discussion about not using printk(). I am seeing the following from checkpatch script:

WARNING: Prefer [subsystem eg: netdev]_level([subsystem]dev, ...

then

Quoted 22 lines
Quoted 11 lines
Quoted 9 lines

dev_level(dev, ... then pr_level(... to printk(KERN_LEVEL ... #105: FILE: include/kunit/test.h:343:

(quoted diff) +1 -0
+	printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)

Is there supposed to be pr_level() - I can find dev_level()

cc'ing Joe Perches for his feedback on this message recommending pr_level() which isn't in 5.3.

I don't care for pr_level or KERN_LEVEL in a printk.

I don't think I follow, how does your version fix this?

Quoted 8 lines

I think this is somewhat overly complicated.

I think I'd write it like:

---
 include/kunit/test.h | 11 ++++-----
 kunit/test.c         | 69 ++++++++++++++++------------------------------------
 2 files changed, 26 insertions(+), 54 deletions(-)
Quoted 5 lines

include/kunit/test.h static inline void *kunit_kzalloc(struct kunit *test,

Quoted 5 lines
include/kunit/test.h +0 -0
index 8b7eb03d4971..aa4abf0a22a5 100644
@@ -339,9 +339,8 @@ static inline void *kunit_kzalloc(struct kunit *test,

size_t size, gfp_t gfp)

Quoted 76 lines
Quoted 76 lines
include/kunit/test.h +5 -6
 void kunit_cleanup(struct kunit *test);
 
-void __printf(3, 4) kunit_printk(const char *level,
-				 const struct kunit *test,
-				 const char *fmt, ...);
+__printf(2, 3)
+void kunit_printk(const struct kunit *test, const char *fmt, ...);
 
 /**
  * kunit_info() - Prints an INFO level message associated with @test.
@@ -353,7 +352,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Takes a variable number of format parameters just like printk().
  */
 #define kunit_info(test, fmt, ...) \
-	kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_INFO fmt, ##__VA_ARGS__)
 
 /**
  * kunit_warn() - Prints a WARN level message associated with @test.
@@ -364,7 +363,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints a warning level message.
  */
 #define kunit_warn(test, fmt, ...) \
-	kunit_printk(KERN_WARNING, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_WARNING fmt, ##__VA_ARGS__)
 
 /**
  * kunit_err() - Prints an ERROR level message associated with @test.
@@ -375,7 +374,7 @@ void __printf(3, 4) kunit_printk(const char *level,
  * Prints an error level message.
  */
 #define kunit_err(test, fmt, ...) \
-	kunit_printk(KERN_ERR, test, fmt, ##__VA_ARGS__)
+	kunit_printk(test, KERN_ERR fmt, ##__VA_ARGS__)
 
 /**
  * KUNIT_SUCCEED() - A no-op expectation. Only exists for code clarity.
kunit/test.c +0 -30
index b2ca9b94c353..ddb9bffb5a5d 100644
@@ -16,40 +16,6 @@ static void kunit_set_failure(struct kunit *test)
 	WRITE_ONCE(test->success, false);
 }
 
-static int kunit_vprintk_emit(int level, const char *fmt, va_list args)
-{
-	return vprintk_emit(0, level, NULL, 0, fmt, args);
-}
-
-static int kunit_printk_emit(int level, const char *fmt, ...)
-{
-	va_list args;
-	int ret;
-
-	va_start(args, fmt);
-	ret = kunit_vprintk_emit(level, fmt, args);
-	va_end(args);
-
-	return ret;
-}
-
-static void kunit_vprintk(const struct kunit *test,
-			  const char *level,
-			  struct va_format *vaf)
-{
-	kunit_printk_emit(level[1] - '0', "\t# %s: %pV", test->name, vaf);
-}
-
-static void kunit_print_tap_version(void)
-{
-	static bool kunit_has_printed_tap_version;
-
-	if (!kunit_has_printed_tap_version) {
-		kunit_printk_emit(LOGLEVEL_INFO, "TAP version 14\n");
Quoted 8 lines

kunit/test.c static size_t kunit_test_cases_len(struct kunit_case

Quoted 8 lines
kunit/test.c +0 -4
-		kunit_has_printed_tap_version = true;
-	}
-}
-
 static size_t kunit_test_cases_len(struct kunit_case *test_cases)
 {
 	struct kunit_case *test_case;
@@ -63,11 +29,9 @@ static size_t kunit_test_cases_len(struct kunit_case

*test_cases)

Quoted 14 lines

kunit/test.c static size_t kunit_test_cases_len(struct kunit_case *test_cases)

Quoted 5 lines
kunit/test.c +0 -2
 static void kunit_print_subtest_start(struct kunit_suite *suite)
 {
-	kunit_print_tap_version();
-	kunit_printk_emit(LOGLEVEL_INFO, "\t# Subtest: %s\n", suite-

name);

Quoted 8 lines
kunit/test.c +3 -3
-	kunit_printk_emit(LOGLEVEL_INFO,
-			  "\t1..%zd\n",
-			  kunit_test_cases_len(suite->test_cases));
+	pr_info_once("TAP version 14\n");
+	pr_info("\t# Subtest: %s\n", suite->name);
+	pr_info("\t1..%zd\n", kunit_test_cases_len(suite->test_cases));
 }
 
Quoted 2 lines

kunit/test.c static void kunit_print_ok_not_ok(bool

Quoted 2 lines
kunit/test.c +0 -0
 static void kunit_print_ok_not_ok(bool should_indent,
@@ -87,9 +51,8 @@ static void kunit_print_ok_not_ok(bool

should_indent,

Quoted 10 lines
Quoted 10 lines
kunit/test.c +2 -3
 	else
 		ok_not_ok = "not ok";
 
-	kunit_printk_emit(LOGLEVEL_INFO,
-			  "%s%s %zd - %s\n",
-			  indent, ok_not_ok, test_number, description);
+	pr_info("%s%s %zd - %s\n",
+		indent, ok_not_ok, test_number, description);
 }
 
Quoted 2 lines

kunit/test.c static void kunit_print_string_stream(struct kunit

Quoted 2 lines
kunit/test.c +0 -0
 static bool kunit_suite_has_succeeded(struct kunit_suite *suite)
@@ -133,11 +96,11 @@ static void kunit_print_string_stream(struct kunit

*test,

Quoted 45 lines
Quoted 45 lines
kunit/test.c +16 -6
 		kunit_err(test,
 			  "Could not allocate buffer, dumping stream:\n");
 		list_for_each_entry(fragment, &stream->fragments, node) {
-			kunit_err(test, fragment->fragment);
+			kunit_err(test, "%s", fragment->fragment);
 		}
 		kunit_err(test, "\n");
 	} else {
-		kunit_err(test, buf);
+		kunit_err(test, "%s", buf);
 		kunit_kfree(test, buf);
 	}
 }
@@ -505,19 +468,29 @@ void kunit_cleanup(struct kunit *test)
 	}
 }
 
-void kunit_printk(const char *level,
-		  const struct kunit *test,
-		  const char *fmt, ...)
+void kunit_printk(const struct kunit *test, const char *fmt, ...)
 {
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 	struct va_format vaf;
 	va_list args;
+	int kern_level;
 
 	va_start(args, fmt);
 
+	while ((kern_level = printk_get_level(fmt)) != 0) {
+		size_t size = printk_skip_level(fmt) - fmt;
+
+		if (kern_level >= '0' && kern_level <= '7') {
+			memcpy(lvl, fmt,  size);
+			lvl[size] = '\0';
+		}
+		fmt += size;
+	}
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
 
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", lvl, test->name, &vaf);
 
Quoted 7 lines

kunit/test.c void kunit_cleanup(struct kunit *test) · line 496

Quoted 2 lines
kunit/test.c +0 -0
 	va_end(args);
 }

How is this simpler?

If we are okay with dynamically adding the KERN_<LEVEL> and %pV (and I don't think that Sergey is),

Quoted 6 lines

Sergey may well be in the minority overall as %pV is now very frequently
used throughout the kernel.

$ git grep "%pV" | wc -l
241

Hmm. IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Quoted 9 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall object size. Consolidated formats with the embedded KERN_<LEVEL> like suggested are smaller overall object size.

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

-- Tim

Quoted 8 lines

This style is also already used in the kernel.

Quoted 2 lines

I don't know. I am clearly not an expert on this topic, but I don't see the merit of the while loop you added above or dropping the level param.

The while use is only to avoid misuses with consecutive KERN_<LEVEL> formats, which had happened in the past.

#9 lore raw
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 11 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

As am I.

#10 lore raw
Quoted 19 lines · Joe Perches
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 9 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

Quoted 4 lines
Quoted 2 lines

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

As am I.

Sorry, to be clear, we are talking about the object size penalty due to adding a single parameter to a function. Is that right?

#11 lore raw
Quoted 23 lines

-----Original Message-----
From: Brendan Higgins

Quoted 19 lines · Joe Perches
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 9 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

As am I.

Sorry, to be clear, we are talking about the object size penalty due to adding a single parameter to a function. Is that right?

Not exactly. The argument is that pre-pending the different KERN_LEVEL strings onto format strings can result in several versions of nearly identical strings being compiled into the object file. By parameterizing this (that is, adding '%s' into the format string, and putting the level into the string as an argument), it prevents this duplication of format strings.

I haven't seen the data on duplication of format strings, and how much this affects it, but little things can add up. Whether it matters in this case depends on whether the format strings that kunit uses are also used elsewhere in the kernel, and whether these same format strings are used with multiple kernel message levels.
-- Tim

#12 lore raw
Quoted 36 lines · On Fri, 2019-08-30 at 23:22 +0000, [email protected] wrote:
Quoted 22 lines

-----Original Message-----
From: Brendan Higgins

Quoted 18 lines · Joe Perches
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 9 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

As am I.

Sorry, to be clear, we are talking about the object size penalty due to adding a single parameter to a function. Is that right?

Not exactly. The argument is that pre-pending the different KERN_LEVEL strings onto format strings can result in several versions of nearly identical strings being compiled into the object file. By parameterizing this (that is, adding '%s' into the format string, and putting the level into the string as an argument), it prevents this duplication of format strings.

Quoted 4 lines

I haven't seen the data on duplication of format strings, and how much this affects it, but little things can add up. Whether it matters in this case depends on whether the format strings that kunit uses are also used elsewhere in the kernel, and whether these same format strings are used with multiple kernel message levels.

deduplication can matter as well, but so far there is little content with kunit_(err|warn|info(=)

kunit/example-test.c:   kunit_info(test, "initializing\n");
kunit/test.c:           kunit_err(test,
kunit/test.c:                   kunit_err(test, "%s", fragment->fragment);
kunit/test.c:           kunit_err(test, "\n");
kunit/test.c:           kunit_err(test, "%s", buf);
kunit/test.c:                   kunit_err(test, "failed to initialize: %d\n", ret);
kunit/test.c:                   kunit_err(test, "test case timed out\n");
kunit/test.c:                   kunit_err(test, "internal error occurred preventing test case from running: %d\n",
kunit/try-catch.c:              kunit_err(test, "try timed out\n");
kunit/try-catch.c:              kunit_err(test, "wake_up_process() was never called\n");
kunit/try-catch.c:              kunit_err(test, "Unknown error: %d\n", exit_code);

Of these, only two do match other kernel uses.

"initializing\n", "failed to initialize: %d\n"

#13 lore raw
In reply to [email protected]
Quoted 37 lines · On Fri, Aug 30, 2019 at 11:22:43PM +0000, [email protected] wrote:
Quoted 23 lines

-----Original Message-----
From: Brendan Higgins

Quoted 19 lines · Joe Perches
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 9 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

As am I.

Sorry, to be clear, we are talking about the object size penalty due to adding a single parameter to a function. Is that right?

Not exactly. The argument is that pre-pending the different KERN_LEVEL strings onto format strings can result in several versions of nearly identical strings being compiled into the object file. By parameterizing this (that is, adding '%s' into the format string, and putting the level into the string as an argument), it prevents this duplication of format strings.

Quoted 5 lines

I haven't seen the data on duplication of format strings, and how much this affects it, but little things can add up. Whether it matters in this case depends on whether the format strings that kunit uses are also used elsewhere in the kernel, and whether these same format strings are used with multiple kernel message levels.
-- Tim

I thought this portion of the discussion was about whether Joe's version of kunit_printk was better or my critique of his version of kunit_printk:

Joe's:

Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
(quoted diff) +13 -3
-void kunit_printk(const char *level,
-		  const struct kunit *test,
-		  const char *fmt, ...)
+void kunit_printk(const struct kunit *test, const char *fmt, ...)
 {
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 	struct va_format vaf;
 	va_list args;
+	int kern_level;
 
 	va_start(args, fmt);
 
+	while ((kern_level = printk_get_level(fmt)) != 0) {
+		size_t size = printk_skip_level(fmt) - fmt;
+
+		if (kern_level >= '0' && kern_level <= '7') {
+			memcpy(lvl, fmt,  size);
+			lvl[size] = '\0';
+		}
+		fmt += size;
+	}
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

Review

Quoted 5 lines
(quoted diff) +1 -1
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", lvl, test->name, &vaf);
 
 	va_end(args);
 }

Mine:

Quoted 9 lines

void kunit_printk(const char *level,

(quoted diff) +0 -0
 		  const struct kunit *test,
 		  const char *fmt, ...)
 {
 	struct va_format vaf;
 	va_list args;
 
 	va_start(args, fmt);
 

Review

Quoted 10 lines
(quoted diff) +3 -1
+	fmt = printk_skip_headers(fmt);
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
 
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", level, test->name, &vaf);
 
 	va_end(args);
 }

I thought you and Joe were arguing that "Joe's" resulted in a smaller object size than "Mine" (not to be confused with the actual patch I presented here, which is what Sergey suggested I do on a different thread).

I really don't feel strongly about what Sergey suggested I do (which is what this patch originally introduced), versus, what Joe suggested, versus what I suggested in response to Joe (or any of the things suggested on other threads). I just want to pick one, fix the breakage in linux-next, and move on with my life.

Cheers

#14 lore raw
Quoted 10 lines · Brendan Higgins

I thought you and Joe were arguing that "Joe's" resulted in a smaller object size than "Mine" (not to be confused with the actual patch I presented here, which is what Sergey suggested I do on a different thread).

I really don't feel strongly about what Sergey suggested I do (which is what this patch originally introduced), versus, what Joe suggested, versus what I suggested in response to Joe (or any of the things suggested on other threads). I just want to pick one, fix the breakage in linux-next, and move on with my life.

Well, if we are voting, I vote for mine! ;)

cheers, Joe

#15 lore raw
In reply to Brendan Higgins
Quoted 113 lines

-----Original Message-----
From: Brendan Higgins

Quoted 26 lines · On Fri, Aug 30, 2019 at 11:22:43PM +0000, [email protected] wrote:
Quoted 26 lines

-----Original Message-----
From: Brendan Higgins

Quoted 22 lines · Joe Perches
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 11 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

As am I.

Sorry, to be clear, we are talking about the object size penalty due to adding a single parameter to a function. Is that right?

Not exactly. The argument is that pre-pending the different KERN_LEVEL strings onto format strings can result in several versions of nearly identical

strings

Quoted 2 lines

being compiled into the object file. By parameterizing this (that is, adding '%s' into the format string, and putting the level into the string as an

argument),

Quoted 4 lines

it prevents this duplication of format strings.

I haven't seen the data on duplication of format strings, and how much this affects it, but little things can add up. Whether it matters in this case

depends

Quoted 1 line

on whether the format strings that kunit uses are also used elsewhere in

the kernel,

Quoted 1 line

and whether these same format strings are used with multiple kernel

message levels.

Quoted 1 line

-- Tim

I thought this portion of the discussion was about whether Joe's version of kunit_printk was better or my critique of his version of kunit_printk:

Joe's:

Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
(quoted diff) +13 -3
-void kunit_printk(const char *level,
-		  const struct kunit *test,
-		  const char *fmt, ...)
+void kunit_printk(const struct kunit *test, const char *fmt, ...)
 {
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 	struct va_format vaf;
 	va_list args;
+	int kern_level;
 
 	va_start(args, fmt);
 
+	while ((kern_level = printk_get_level(fmt)) != 0) {
+		size_t size = printk_skip_level(fmt) - fmt;
+
+		if (kern_level >= '0' && kern_level <= '7') {
+			memcpy(lvl, fmt,  size);
+			lvl[size] = '\0';
+		}
+		fmt += size;
+	}
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

Review

Quoted 5 lines
(quoted diff) +1 -1
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", lvl, test->name, &vaf);
 
 	va_end(args);
 }

Mine:

Quoted 9 lines

void kunit_printk(const char *level,

(quoted diff) +0 -0
 		  const struct kunit *test,
 		  const char *fmt, ...)
 {
 	struct va_format vaf;
 	va_list args;
 
 	va_start(args, fmt);
 

Review

Quoted 10 lines
(quoted diff) +3 -1
+	fmt = printk_skip_headers(fmt);
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
 
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", level, test->name, &vaf);
 
 	va_end(args);
 }

I thought you and Joe were arguing that "Joe's" resulted in a smaller object size than "Mine" (not to be confused with the actual patch I presented here, which is what Sergey suggested I do on a different thread).

Quoted 5 lines

I really don't feel strongly about what Sergey suggested I do (which is what this patch originally introduced), versus, what Joe suggested, versus what I suggested in response to Joe (or any of the things suggested on other threads). I just want to pick one, fix the breakage in linux-next, and move on with my life.

When in doubt, do what the sub-system maintainer says. I'd go with Sergey's suggestion. Maintainers often are juggling a host of issues, and weighing new features and usages of their system against their long-term plans for their sub-system. Sometimes they have time to communicate all the intricacies of their counter-proposals, and sometimes not.

But they know their system best, and much more often than not provide sound advice.

If you don't have a strong feeling about it, just do what they say.
-- Tim

Reviewed-by
#16 lore raw
In reply to Brendan Higgins
Quoted 105 lines · Brendan Higgins
Quoted 37 lines · On Fri, Aug 30, 2019 at 11:22:43PM +0000, [email protected] wrote:
Quoted 23 lines

-----Original Message-----
From: Brendan Higgins

Quoted 19 lines · Joe Perches
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 9 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

As am I.

Sorry, to be clear, we are talking about the object size penalty due to adding a single parameter to a function. Is that right?

Not exactly. The argument is that pre-pending the different KERN_LEVEL strings onto format strings can result in several versions of nearly identical strings being compiled into the object file. By parameterizing this (that is, adding '%s' into the format string, and putting the level into the string as an argument), it prevents this duplication of format strings.

Quoted 5 lines

I haven't seen the data on duplication of format strings, and how much this affects it, but little things can add up. Whether it matters in this case depends on whether the format strings that kunit uses are also used elsewhere in the kernel, and whether these same format strings are used with multiple kernel message levels.
-- Tim

I thought this portion of the discussion was about whether Joe's version of kunit_printk was better or my critique of his version of kunit_printk:

Joe's:

Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
(quoted diff) +13 -3
-void kunit_printk(const char *level,
-		  const struct kunit *test,
-		  const char *fmt, ...)
+void kunit_printk(const struct kunit *test, const char *fmt, ...)
 {
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 	struct va_format vaf;
 	va_list args;
+	int kern_level;
 
 	va_start(args, fmt);
 
+	while ((kern_level = printk_get_level(fmt)) != 0) {
+		size_t size = printk_skip_level(fmt) - fmt;
+
+		if (kern_level >= '0' && kern_level <= '7') {
+			memcpy(lvl, fmt,  size);
+			lvl[size] = '\0';
+		}
+		fmt += size;
+	}
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

Review

Quoted 5 lines
(quoted diff) +1 -1
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", lvl, test->name, &vaf);
 
 	va_end(args);
 }

Mine:

Quoted 9 lines

void kunit_printk(const char *level,

(quoted diff) +0 -0
 		  const struct kunit *test,
 		  const char *fmt, ...)
 {
 	struct va_format vaf;
 	va_list args;
 
 	va_start(args, fmt);
 

Review

Quoted 10 lines
(quoted diff) +3 -1
+	fmt = printk_skip_headers(fmt);
+
 	vaf.fmt = fmt;
 	vaf.va = &args;
 
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", level, test->name, &vaf);
 
 	va_end(args);
 }

I thought you and Joe were arguing that "Joe's" resulted in a smaller object size than "Mine" (not to be confused with the actual patch I presented here, which is what Sergey suggested I do on a different thread).

Quoted 5 lines

I really don't feel strongly about what Sergey suggested I do (which is what this patch originally introduced), versus, what Joe suggested, versus what I suggested in response to Joe (or any of the things suggested on other threads). I just want to pick one, fix the breakage in linux-next, and move on with my life.

I am a bit lost in all the versions ;-) Though, I like most this patch. I think that it is based on Sergey's suggestion.

I think that object size is not a huge concern for unit testing. Also if I get it correctly, the object is bigger only when the same string is used with different log levels. I am not sure how often this happen.

Feel free to use for this patch:

Best Regards,
Petr

Quoted 1 line

Cheers

maintainer
#17 lore raw
Quoted 106 lines · Petr Mladek
Quoted 105 lines · Brendan Higgins
Quoted 37 lines · On Fri, Aug 30, 2019 at 11:22:43PM +0000, [email protected] wrote:
Quoted 23 lines

-----Original Message-----
From: Brendan Higgins

Quoted 19 lines · Joe Perches
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

Quoted 9 lines
Quoted 8 lines

then wouldn't it be easier to pass in the

Quoted 2 lines

kernel level as a separate parameter and then strip off all printk headers like this:

Depends on whether or not you care for overall
object size. Consolidated formats with the
embedded KERN_<LEVEL> like suggested are smaller
overall object size.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

This is an argument I can agree with. I'm generally in favor of things that lessen kernel size creep. :-)

As am I.

Sorry, to be clear, we are talking about the object size penalty due to adding a single parameter to a function. Is that right?

Not exactly. The argument is that pre-pending the different KERN_LEVEL strings onto format strings can result in several versions of nearly identical strings being compiled into the object file. By parameterizing this (that is, adding '%s' into the format string, and putting the level into the string as an argument), it prevents this duplication of format strings.

Quoted 5 lines

I haven't seen the data on duplication of format strings, and how much this affects it, but little things can add up. Whether it matters in this case depends on whether the format strings that kunit uses are also used elsewhere in the kernel, and whether these same format strings are used with multiple kernel message levels.
-- Tim

I thought this portion of the discussion was about whether Joe's version of kunit_printk was better or my critique of his version of kunit_printk:

Joe's:

Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
(quoted diff) +13 -3
-void kunit_printk(const char *level,
-		  const struct kunit *test,
-		  const char *fmt, ...)
+void kunit_printk(const struct kunit *test, const char *fmt, ...)
  {
+	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
  	struct va_format vaf;
  	va_list args;
+	int kern_level;
 
  	va_start(args, fmt);
 
+	while ((kern_level = printk_get_level(fmt)) != 0) {
+		size_t size = printk_skip_level(fmt) - fmt;
+
+		if (kern_level >= '0' && kern_level <= '7') {
+			memcpy(lvl, fmt,  size);
+			lvl[size] = '\0';
+		}
+		fmt += size;
+	}
+
  	vaf.fmt = fmt;
  	vaf.va = &args;
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

Review

Quoted 5 lines
(quoted diff) +1 -1
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", lvl, test->name, &vaf);
 
  	va_end(args);
  }

Mine:

Quoted 9 lines

void kunit_printk(const char *level,
const struct kunit *test,
const char *fmt, ...)
{
struct va_format vaf;
va_list args;

  	va_start(args, fmt);

Review

Quoted 10 lines
(quoted diff) +3 -1
+	fmt = printk_skip_headers(fmt);
+
  	vaf.fmt = fmt;
  	vaf.va = &args;
 
-	kunit_vprintk(test, level, &vaf);
+	printk("%s\t# %s %pV\n", level, test->name, &vaf);
 
  	va_end(args);
  }

I thought you and Joe were arguing that "Joe's" resulted in a smaller object size than "Mine" (not to be confused with the actual patch I presented here, which is what Sergey suggested I do on a different thread).

Quoted 9 lines
Quoted 5 lines

I really don't feel strongly about what Sergey suggested I do (which is what this patch originally introduced), versus, what Joe suggested, versus what I suggested in response to Joe (or any of the things suggested on other threads). I just want to pick one, fix the breakage in linux-next, and move on with my life.

I am a bit lost in all the versions ;-) Though, I like most this patch. I think that it is based on Sergey's suggestion.

I am too.

Quoted 9 lines

I think that object size is not a huge concern for unit testing. Also if I get it correctly, the object is bigger only when the same string is used with different log levels. I am not sure how often this happen.

Feel free to use for this patch:

Brendan,

Send me the version Sergey suggested with a short summary of the discussion in the commit log. Tag it v3 so I don't pull the wrong patch in.

I am going to just ignore the checkpatch warn on this and get it in. Thanks for the discussion. It helped me clarify my understanding of the printk.

thanks,
-- Shuah

#18 lore raw
In reply to Joe Perches
Quoted 5 lines

-----Original Message-----
From: Joe Perches

Quoted 8 lines
Quoted 1 line · On Fri, 2019-08-30 at 21:58 +0000, [email protected] wrote:
Quoted 1 line

From: Joe Perches

[]

Quoted 2 lines

IMHO %pV should be avoided if possible. Just because people are doing it doesn't mean it should be used when it is not necessary.

Well, as the guy that created %pV, I of course
have a different opinion.

LOL. Well I stepped in that one.

I don't have any data to support my position on this particular printk feature, but having worked for a while on stack size reduction for a few Sony products, I'm always a bit leery of recursive routines in the kernel. I vaguely recall some recursive printk routines giving me problems on a product that used a sub-4K stack configuration I did many years ago. I don't recall if it was specifically %pV or not. Anyway YMMV.
-- Tim

#19 lore raw
In reply to shuah

On (08/29/19 11:01), shuah wrote:
[..]

Quoted 5 lines

Hi Sergey,

What are the guidelines for using printk(). I recall some discussion about not using printk(). I am seeing the following from checkpatch script:

Hello,

include/kunit/test.h static inline void *kunit_kzalloc(struct kunit *test, size_t size, gfp_t gfp) · line 343

Quoted 5 lines

WARNING: Prefer [subsystem eg: netdev]_level([subsystem]dev, ... then dev_level(dev, ... then pr_level(... to printk(KERN_LEVEL ... #105: FILE: include/kunit/test.h:343:

include/kunit/test.h +1 -0
+	printk(KERN_LEVEL "\t# %s: " fmt, (test)->name, ##__VA_ARGS__)

Oh, right.
So we sort of want people to use pr_err()/pr_info()/pr_"level"() because, otherwise, when people use plain printk(), they tend to forget to add KERN_LEVEL.

In kunit case everything looks fine. KERN_LEVEL is there so I'm fine with the patch.

You still can switch to pr_info()/pr_err()/pr_etc, just to make checkpatch happier, but that's up to you.

Quoted 1 line

Is there supposed to be pr_level() - I can find dev_level()

No, not really. pr_level() stands for pr_"debug"()/pr_"info"()/etc.

E.g.

#define pr_emerg(fmt, ...) \
        printk(KERN_EMERG pr_fmt(fmt), ##__VA_ARGS__)
#define pr_alert(fmt, ...) \
        printk(KERN_ALERT pr_fmt(fmt), ##__VA_ARGS__)
#define pr_crit(fmt, ...) \
        printk(KERN_CRIT pr_fmt(fmt), ##__VA_ARGS__)
#define pr_err(fmt, ...) \
        printk(KERN_ERR pr_fmt(fmt), ##__VA_ARGS__)
#define pr_warning(fmt, ...) \
        printk(KERN_WARNING pr_fmt(fmt), ##__VA_ARGS__)
#define pr_warn pr_warning
#define pr_notice(fmt, ...) \
        printk(KERN_NOTICE pr_fmt(fmt), ##__VA_ARGS__)
#define pr_info(fmt, ...) \
        printk(KERN_INFO pr_fmt(fmt), ##__VA_ARGS__)

	-ss
Acked-by on v2 1/1
#20 lore raw
In reply to Brendan Higgins
Quoted 9 lines · Brendan Higgins

Previously KUnit assumed that printk would always be present, which is not a valid assumption to make. Fix that by removing call to vprintk_emit, and calling printk directly.

Quoted 4 lines
---
 include/kunit/test.h | 11 ++++-----
 kunit/test.c         | 57 +++++---------------------------------------
 2 files changed, 11 insertions(+), 57 deletions(-)

thanks.

Signature · 1 line
~Randy