Skip to contentSkip to conversation

printf stuff for 4.5

PATCH Superseded Rasmus Villemoes 3 participants 28 messages Nov 23, 2015 – Nov 26, 2015

Subsystems (2)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Rasmus Villemoes

author reviewer
#0 lore raw

cover printf stuff for 4.5

I was too late for 4.4, so here's hoping to get this into 4.5.

1-3 fix a theoretical race in printing strings via %s - since we're reusing existing code from the dentry printer, we actually also win on code size. Ingo, can I perhaps get you to turn your "looks good to" into an ack?

4 fixes a problem introduced by converting all bitmap formatting to go via %pb[l] - it turns out that bitmaps with >= 1<<15 bits are actually printed sometimes. This isn't necessarily the best fix, but the discussion died out, so I'm proposing this for now.

5 is just a minor optimization (maybe not so much on register-challenged arches).

6 I'm not too sure about, but maybe the bitmap problem had been discovered sooner (it took a little over half a year to be reported) if these had been in place.

7 is another microoptimization.

Paranoid-me wanted me to include patch 8, but I don't have strong feelings for it.

9-14 are minor additions to the test module (some with acks from Kees).

Rasmus Villemoes (14):
lib/vsprintf.c: pull out padding code from dentry_name() lib/vsprintf.c: move string() below widen_string() lib/vsprintf.c: eliminate potential race in string() lib/vsprintf.c: expand field_width to 24 bits lib/vsprintf.c: help gcc make number() smaller lib/vsprintf.c: warn about too large precisions and field widths lib/vsprintf.c: slightly refactor vscnprintf() lib/kasprintf.c: add sanity check to kvasprintf lib/test_printf.c: don't BUG
lib/test_printf.c: check for out-of-bound writes lib/test_printf.c: test precision quirks lib/test_printf.c: account for kvasprintf tests lib/test_printf.c: add test for large bitmaps lib/test_printf.c: test dentry printing

 lib/kasprintf.c   |  10 +--
 lib/test_printf.c |  96 ++++++++++++++++++++++----
 lib/vsprintf.c    | 196 +++++++++++++++++++++++++++++++-----------------------
 3 files changed, 203 insertions(+), 99 deletions(-)
Signature · 1 line
2.6.1
author reviewer
#1 lore raw

PATCH 1/14 lib/vsprintf.c: pull out padding code from dentry_name()

Pull out the logic in dentry_name() which handles field width space padding, in preparation for reusing it from string(). Rename the widen() helper to move_right(), since it is used for handling the !(flags & LEFT) case.

---
 lib/vsprintf.c | 46 +++++++++++++++++++++++++++++++---------------
 1 file changed, 31 insertions(+), 15 deletions(-)
lib/vsprintf.c +31 -15
index f9cee8e1233c..d7452563a6a6 100644
@@ -538,7 +538,7 @@ char *string(char *buf, char *end, const char *s, struct printf_spec spec)
 	return buf;
 }
 
-static void widen(char *buf, char *end, unsigned len, unsigned spaces)
+static void move_right(char *buf, char *end, unsigned len, unsigned spaces)
 {
 	size_t size;
 	if (buf >= end)	/* nowhere to put anything */
@@ -556,6 +556,35 @@ static void widen(char *buf, char *end, unsigned len, unsigned spaces)
 	memset(buf, ' ', spaces);
 }
 
+/*
+ * Handle field width padding for a string.
+ * @buf: current buffer position
+ * @n: length of string
+ * @end: end of output buffer
+ * @spec: for field width and flags
+ * Returns: new buffer position after padding.
+ */
+static noinline_for_stack
+char *widen_string(char *buf, int n, char *end, struct printf_spec spec)
+{
+	unsigned spaces;
+
+	if (likely(n >= spec.field_width))
+		return buf;
+	/* we want to pad the sucker */
+	spaces = spec.field_width - n;
+	if (!(spec.flags & LEFT)) {
+		move_right(buf - n, end, n, spaces);
+		return buf + spaces;
+	}
+	while (spaces--) {
+		if (buf < end)
+			*buf = ' ';
+		++buf;
+	}
+	return buf;
+}
+
 static noinline_for_stack
 char *dentry_name(char *buf, char *end, const struct dentry *d, struct printf_spec spec,
 		  const char *fmt)
@@ -597,20 +626,7 @@ char *dentry_name(char *buf, char *end, const struct dentry *d, struct printf_sp
 			*buf = c;
 	}
 	rcu_read_unlock();
-	if (n < spec.field_width) {
-		/* we want to pad the sucker */
-		unsigned spaces = spec.field_width - n;
-		if (!(spec.flags & LEFT)) {
-			widen(buf - n, end, n, spaces);
-			return buf + spaces;
-		}
-		while (spaces--) {
-			if (buf < end)
-				*buf = ' ';
-			++buf;
-		}
-	}
-	return buf;
+	return widen_string(buf, n, end, spec);
 }
 
 static noinline_for_stack
Signature · 1 line
2.6.1
#2 lore raw

Re: [PATCH 01/14] lib/vsprintf.c: pull out padding code from dentry_name()

Quoted 26 lines · Rasmus Villemoes

Pull out the logic in dentry_name() which handles field width space padding, in preparation for reusing it from string(). Rename the widen() helper to move_right(), since it is used for handling the !(flags & LEFT) case.

---
 lib/vsprintf.c | 46 +++++++++++++++++++++++++++++++---------------
 1 file changed, 31 insertions(+), 15 deletions(-)
lib/vsprintf.c +1 -1
index f9cee8e1233c..d7452563a6a6 100644
@@ -538,7 +538,7 @@ char *string(char *buf, char *end, const char *s, struct printf_spec spec)
        return buf;
 }
 
-static void widen(char *buf, char *end, unsigned len, unsigned spaces)
+static void move_right(char *buf, char *end, unsigned len, unsigned spaces)
 {
        size_t size;
        if (buf >= end) /* nowhere to put anything */

lib/vsprintf.c static void widen(char *buf, char *end, unsigned len, unsigned spaces) · line 559

Quoted 4 lines
lib/vsprintf.c +1 -0
@@ -556,6 +556,35 @@ static void widen(char *buf, char *end, unsigned len, unsigned spaces)
        memset(buf, ' ', spaces);
 }
 
+/*

Perhaps /**

Quoted 62 lines
lib/vsprintf.c +28 -0
+ * Handle field width padding for a string.
+ * @buf: current buffer position
+ * @n: length of string
+ * @end: end of output buffer
+ * @spec: for field width and flags
+ * Returns: new buffer position after padding.
+ */
+static noinline_for_stack
+char *widen_string(char *buf, int n, char *end, struct printf_spec spec)
+{
+       unsigned spaces;
+
+       if (likely(n >= spec.field_width))
 
+               return buf;
+       /* we want to pad the sucker */
+       spaces = spec.field_width - n;
+       if (!(spec.flags & LEFT)) {
+               move_right(buf - n, end, n, spaces);
+               return buf + spaces;
+       }
+       while (spaces--) {
+               if (buf < end)
+                       *buf = ' ';
+               ++buf;
+       }
+       return buf;
+}
+
 static noinline_for_stack
 char *dentry_name(char *buf, char *end, const struct dentry *d, struct printf_spec spec,

const char *fmt)

lib/vsprintf.c +1 -15
@@ -597,20 +626,7 @@ char *dentry_name(char *buf, char *end, const struct dentry *d, struct printf_sp
                        *buf = c;
        }
        rcu_read_unlock();
-       if (n < spec.field_width) {
-               /* we want to pad the sucker */
-               unsigned spaces = spec.field_width - n;
-               if (!(spec.flags & LEFT)) {
-                       widen(buf - n, end, n, spaces);
-                       return buf + spaces;
-               }
-               while (spaces--) {
-                       if (buf < end)
-                               *buf = ' ';
-                       ++buf;
-               }
-       }
-       return buf;
+       return widen_string(buf, n, end, spec);
 }
 
 static noinline_for_stack
--

2.6.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author reviewer
#3 lore raw

Re: [PATCH 01/14] lib/vsprintf.c: pull out padding code from dentry_name()

Quoted 20 lines · Andy Shevchenko
Quoted 18 lines · Rasmus Villemoes

Pull out the logic in dentry_name() which handles field width space padding, in preparation for reusing it from string(). Rename the widen() helper to move_right(), since it is used for handling the !(flags & LEFT) case.

lib/vsprintf.c +1 -1
-static void widen(char *buf, char *end, unsigned len, unsigned spaces)
+static void move_right(char *buf, char *end, unsigned len, unsigned spaces)
 {
        size_t size;
        if (buf >= end) /* nowhere to put anything */
@@ -556,6 +556,35 @@ static void widen(char *buf, char *end, unsigned len, unsigned spaces)
        memset(buf, ' ', spaces);
 }
 
Quoted 3 lines

lib/vsprintf.c static void widen(char *buf, char *end, unsigned len, unsigned spaces) · line 559

Quoted 1 line
lib/vsprintf.c +1 -0
+/*

Perhaps /**

Nah, it's not an exported function, so I didn't want to do a formal kernel-doc thing. I can make it proper kernel-doc if you insist, but I don't think the whole "current buffer position", "end of output buffer", "new buffer position" makes sense outside the context of vsprintf.c with its own slightly peculiar way of doing things.

Quoted 10 lines

lib/vsprintf.c static void widen(char *buf, char *end, unsigned len, unsigned spaces) · line 569

Quoted 10 lines
lib/vsprintf.c +10 -0
+ * Handle field width padding for a string.
+ * @buf: current buffer position
+ * @n: length of string
+ * @end: end of output buffer
+ * @spec: for field width and flags
+ * Returns: new buffer position after padding.
+ */
+static noinline_for_stack
+char *widen_string(char *buf, int n, char *end, struct printf_spec spec)
+{

Rasmus

author reviewer
#4 lore raw

PATCH 2/14 lib/vsprintf.c: move string() below widen_string()

In reply to Rasmus Villemoes

This is pure code movement, making sure the widen_string() helper is defined before the string() function.

---
 lib/vsprintf.c | 62 +++++++++++++++++++++++++++++-----------------------------
 1 file changed, 31 insertions(+), 31 deletions(-)
lib/vsprintf.c +31 -31
index d7452563a6a6..a021e6380404 100644
@@ -507,37 +507,6 @@ char *number(char *buf, char *end, unsigned long long num,
 	return buf;
 }
 
-static noinline_for_stack
-char *string(char *buf, char *end, const char *s, struct printf_spec spec)
-{
-	int len, i;
-
-	if ((unsigned long)s < PAGE_SIZE)
-		s = "(null)";
-
-	len = strnlen(s, spec.precision);
-
-	if (!(spec.flags & LEFT)) {
-		while (len < spec.field_width--) {
-			if (buf < end)
-				*buf = ' ';
-			++buf;
-		}
-	}
-	for (i = 0; i < len; ++i) {
-		if (buf < end)
-			*buf = *s;
-		++buf; ++s;
-	}
-	while (len < spec.field_width--) {
-		if (buf < end)
-			*buf = ' ';
-		++buf;
-	}
-
-	return buf;
-}
-
 static void move_right(char *buf, char *end, unsigned len, unsigned spaces)
 {
 	size_t size;
@@ -586,6 +555,37 @@ char *widen_string(char *buf, int n, char *end, struct printf_spec spec)
 }
 
 static noinline_for_stack
+char *string(char *buf, char *end, const char *s, struct printf_spec spec)
+{
+	int len, i;
+
+	if ((unsigned long)s < PAGE_SIZE)
+		s = "(null)";
+
+	len = strnlen(s, spec.precision);
+
+	if (!(spec.flags & LEFT)) {
+		while (len < spec.field_width--) {
+			if (buf < end)
+				*buf = ' ';
+			++buf;
+		}
+	}
+	for (i = 0; i < len; ++i) {
+		if (buf < end)
+			*buf = *s;
+		++buf; ++s;
+	}
+	while (len < spec.field_width--) {
+		if (buf < end)
+			*buf = ' ';
+		++buf;
+	}
+
+	return buf;
+}
+
+static noinline_for_stack
 char *dentry_name(char *buf, char *end, const struct dentry *d, struct printf_spec spec,
 		  const char *fmt)
 {
Signature · 1 line
2.6.1
author reviewer
#5 lore raw

PATCH 3/14 lib/vsprintf.c: eliminate potential race in string()

In reply to Rasmus Villemoes

If the string corresponding to a %s specifier can change under us, we might end up copying a \0 byte to the output buffer. There might be callers who expect the output buffer to contain a genuine C string whose length is exactly the snprintf return value (assuming truncation hasn't happened or has been checked for).

We can avoid this by only passing over the source string once, stopping the first time we meet a nul byte (or when we reach the given precision), and then letting widen_string() handle left/right space padding. As a small bonus, this code reuse also makes the generated code slightly smaller.

---
 lib/vsprintf.c | 28 +++++++++-------------------
 1 file changed, 9 insertions(+), 19 deletions(-)
lib/vsprintf.c +9 -19
index a021e6380404..63ca52366049 100644
@@ -557,32 +557,22 @@ char *widen_string(char *buf, int n, char *end, struct printf_spec spec)
 static noinline_for_stack
 char *string(char *buf, char *end, const char *s, struct printf_spec spec)
 {
-	int len, i;
+	int len = 0;
+	size_t lim = spec.precision;
 
 	if ((unsigned long)s < PAGE_SIZE)
 		s = "(null)";
 
-	len = strnlen(s, spec.precision);
-
-	if (!(spec.flags & LEFT)) {
-		while (len < spec.field_width--) {
-			if (buf < end)
-				*buf = ' ';
-			++buf;
-		}
-	}
-	for (i = 0; i < len; ++i) {
-		if (buf < end)
-			*buf = *s;
-		++buf; ++s;
-	}
-	while (len < spec.field_width--) {
+	while (lim--) {
+		char c = *s++;
+		if (!c)
+			break;
 		if (buf < end)
-			*buf = ' ';
+			*buf = c;
 		++buf;
+		++len;
 	}
-
-	return buf;
+	return widen_string(buf, len, end, spec);
 }
 
 static noinline_for_stack
Signature · 1 line
2.6.1
#6 lore raw

Re: [PATCH 03/14] lib/vsprintf.c: eliminate potential race in string()

Quoted 6 lines · Rasmus Villemoes

If the string corresponding to a %s specifier can change under us, we might end up copying a \0 byte to the output buffer. There might be callers who expect the output buffer to contain a genuine C string whose length is exactly the snprintf return value (assuming truncation hasn't happened or has been checked for).

Quoted 6 lines

We can avoid this by only passing over the source string once, stopping the first time we meet a nul byte (or when we reach the given precision), and then letting widen_string() handle left/right space padding. As a small bonus, this code reuse also makes the generated code slightly smaller.

Could it be pair of patches: a) re-use, b) optimize for fuzzy strings?

Quoted 11 lines
---
 lib/vsprintf.c | 28 +++++++++-------------------
 1 file changed, 9 insertions(+), 19 deletions(-)
lib/vsprintf.c +0 -0
index a021e6380404..63ca52366049 100644

lib/vsprintf.c char *widen_string(char *buf, int n, char *end, struct printf_spec spec) · line 561

Quoted 6 lines
lib/vsprintf.c +2 -1
@@ -557,32 +557,22 @@ char *widen_string(char *buf, int n, char *end, struct printf_spec spec)
 static noinline_for_stack
 char *string(char *buf, char *end, const char *s, struct printf_spec spec)
 {
-       int len, i;
+       int len = 0;
+       size_t lim = spec.precision;

Just a nitpick: maybe longer first?

Quoted 43 lines
lib/vsprintf.c +7 -19
        if ((unsigned long)s < PAGE_SIZE)
                s = "(null)";
 
-       len = strnlen(s, spec.precision);
-
-       if (!(spec.flags & LEFT)) {
-               while (len < spec.field_width--) {
-                       if (buf < end)
-                               *buf = ' ';
-                       ++buf;
-               }
-       }
-       for (i = 0; i < len; ++i) {
-               if (buf < end)
-                       *buf = *s;
-               ++buf; ++s;
-       }
-       while (len < spec.field_width--) {
+       while (lim--) {
+               char c = *s++;
+               if (!c)
+                       break;
                if (buf < end)
-                       *buf = ' ';
+                       *buf = c;
                ++buf;
+               ++len;
        }
-
-       return buf;
+       return widen_string(buf, len, end, spec);
 }
 
 static noinline_for_stack
--

2.6.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author reviewer
#7 lore raw

Re: [PATCH 03/14] lib/vsprintf.c: eliminate potential race in string()

Quoted 8 lines · Andy Shevchenko
Quoted 6 lines · Rasmus Villemoes

If the string corresponding to a %s specifier can change under us, we might end up copying a \0 byte to the output buffer. There might be callers who expect the output buffer to contain a genuine C string whose length is exactly the snprintf return value (assuming truncation hasn't happened or has been checked for).

Quoted 8 lines
Quoted 6 lines

We can avoid this by only passing over the source string once, stopping the first time we meet a nul byte (or when we reach the given precision), and then letting widen_string() handle left/right space padding. As a small bonus, this code reuse also makes the generated code slightly smaller.

Could it be pair of patches: a) re-use, b) optimize for fuzzy strings?

I'm afraid I have no idea what you mean. The patch is already broken into three (pull out from dentry(), move helper, do the actual thing to string()). What's a 'fuzzy string', and what optimization do you think of?

This patch already gives us the bonus of only passing over the source once instead of twice. (Well, at the expense of a little complicated logic in case we have a larger field width and not the LEFT flag set, but these are so extremely rare compared to plain %s that it's not worth caring about. And in any case, the logic already existed.)

Quoted 6 lines
Quoted 6 lines
---
 lib/vsprintf.c | 28 +++++++++-------------------
 1 file changed, 9 insertions(+), 19 deletions(-)
Quoted 13 lines

lib/vsprintf.c char *widen_string(char *buf, int n, char *end, struct printf_spec spec) · line 561

Quoted 11 lines
lib/vsprintf.c +2 -1
index a021e6380404..63ca52366049 100644
@@ -557,32 +557,22 @@ char *widen_string(char *buf, int n, char *end, struct printf_spec spec)
 static noinline_for_stack
 char *string(char *buf, char *end, const char *s, struct printf_spec spec)
 {
-       int len, i;
+       int len = 0;
+       size_t lim = spec.precision;

Just a nitpick: maybe longer first?

Why?

Rasmus

author reviewer
#8 lore raw

PATCH 4/14 lib/vsprintf.c: expand field_width to 24 bits

In reply to Rasmus Villemoes

Maurizio Lombardi reported a problem [1] with the %pb extension: It doesn't work for sufficiently large bitmaps, since the size is stashed in the field_width field of the struct printf_spec, which is currently an s16. Concretely, this manifested itself in /sys/bus/pseudo/drivers/scsi_debug/map being empty, since the bitmap printer got a size of 0, which is the 16 bit truncation of the actual bitmap size.

We do want to keep struct printf_spec at 8 bytes so that it can cheaply be passed by value. The qualifier field is only used for internal bookkeeping in format_decode, so we might as well use a local variable for that. This gives us an additional 8 bits, which we can then use for the field width.

To stay in 8 bytes, we need to do a little rearranging and make the type member a bitfield as well. For consistency, change all the members to bit fields. gcc doesn't generate much worse code with these changes (in fact, bloat-o-meter says we save 300 bytes - which I think is a little surprising).

I didn't find a BUILD_BUG/compiletime_assertion/... which would work outside function context, so for now I just open-coded it.

[1] http://thread.gmane.org/gmane.linux.kernel/2034835

---
 lib/vsprintf.c | 41 +++++++++++++++++++++--------------------
 1 file changed, 21 insertions(+), 20 deletions(-)
lib/vsprintf.c +21 -20
index 63ca52366049..01c3aa638582 100644
@@ -380,13 +380,13 @@ enum format_type {
 };
 
 struct printf_spec {
-	u8	type;		/* format_type enum */
-	u8	flags;		/* flags to number() */
-	u8	base;		/* number base, 8, 10 or 16 only */
-	u8	qualifier;	/* number qualifier, one of 'hHlLtzZ' */
-	s16	field_width;	/* width of output field */
-	s16	precision;	/* # of digits/chars */
-};
+	unsigned int	type:8;		/* format_type enum */
+	signed int	field_width:24;	/* width of output field */
+	unsigned int	flags:8;	/* flags to number() */
+	unsigned int	base:8;		/* number base, 8, 10 or 16 only */
+	signed int	precision:16;	/* # of digits/chars */
+} __packed;
+extern char __check_printf_spec[1-2*(sizeof(struct printf_spec) != 8)];
 
 static noinline_for_stack
 char *number(char *buf, char *end, unsigned long long num,
@@ -1641,6 +1641,7 @@ static noinline_for_stack
 int format_decode(const char *fmt, struct printf_spec *spec)
 {
 	const char *start = fmt;
+	char qualifier;
 
 	/* we finished early by reading the field width */
 	if (spec->type == FORMAT_TYPE_WIDTH) {
@@ -1723,16 +1724,16 @@ precision:
 
 qualifier:
 	/* get the conversion qualifier */
-	spec->qualifier = -1;
+	qualifier = 0;
 	if (*fmt == 'h' || _tolower(*fmt) == 'l' ||
 	    _tolower(*fmt) == 'z' || *fmt == 't') {
-		spec->qualifier = *fmt++;
-		if (unlikely(spec->qualifier == *fmt)) {
-			if (spec->qualifier == 'l') {
-				spec->qualifier = 'L';
+		qualifier = *fmt++;
+		if (unlikely(qualifier == *fmt)) {
+			if (qualifier == 'l') {
+				qualifier = 'L';
 				++fmt;
-			} else if (spec->qualifier == 'h') {
-				spec->qualifier = 'H';
+			} else if (qualifier == 'h') {
+				qualifier = 'H';
 				++fmt;
 			}
 		}
@@ -1789,19 +1790,19 @@ qualifier:
 		return fmt - start;
 	}
 
-	if (spec->qualifier == 'L')
+	if (qualifier == 'L')
 		spec->type = FORMAT_TYPE_LONG_LONG;
-	else if (spec->qualifier == 'l') {
+	else if (qualifier == 'l') {
 		BUILD_BUG_ON(FORMAT_TYPE_ULONG + SIGN != FORMAT_TYPE_LONG);
 		spec->type = FORMAT_TYPE_ULONG + (spec->flags & SIGN);
-	} else if (_tolower(spec->qualifier) == 'z') {
+	} else if (_tolower(qualifier) == 'z') {
 		spec->type = FORMAT_TYPE_SIZE_T;
-	} else if (spec->qualifier == 't') {
+	} else if (qualifier == 't') {
 		spec->type = FORMAT_TYPE_PTRDIFF;
-	} else if (spec->qualifier == 'H') {
+	} else if (qualifier == 'H') {
 		BUILD_BUG_ON(FORMAT_TYPE_UBYTE + SIGN != FORMAT_TYPE_BYTE);
 		spec->type = FORMAT_TYPE_UBYTE + (spec->flags & SIGN);
-	} else if (spec->qualifier == 'h') {
+	} else if (qualifier == 'h') {
 		BUILD_BUG_ON(FORMAT_TYPE_USHORT + SIGN != FORMAT_TYPE_SHORT);
 		spec->type = FORMAT_TYPE_USHORT + (spec->flags & SIGN);
 	} else {
Signature · 1 line
2.6.1
#9 lore raw

Re: [PATCH 04/14] lib/vsprintf.c: expand field_width to 24 bits

Quoted 14 lines · Rasmus Villemoes

Maurizio Lombardi reported a problem [1] with the %pb extension: It doesn't work for sufficiently large bitmaps, since the size is stashed in the field_width field of the struct printf_spec, which is currently an s16. Concretely, this manifested itself in /sys/bus/pseudo/drivers/scsi_debug/map being empty, since the bitmap printer got a size of 0, which is the 16 bit truncation of the actual bitmap size.

We do want to keep struct printf_spec at 8 bytes so that it can cheaply be passed by value. The qualifier field is only used for internal bookkeeping in format_decode, so we might as well use a local variable for that. This gives us an additional 8 bits, which we can then use for the field width.

Quoted 8 lines

To stay in 8 bytes, we need to do a little rearranging and make the type member a bitfield as well. For consistency, change all the members to bit fields. gcc doesn't generate much worse code with these changes (in fact, bloat-o-meter says we save 300 bytes - which I think is a little surprising).

I didn't find a BUILD_BUG/compiletime_assertion/... which would work outside function context, so for now I just open-coded it.

And any objections to put it into vsnprintf() ?

Quoted 37 lines

[1] http://thread.gmane.org/gmane.linux.kernel/2034835

---
 lib/vsprintf.c | 41 +++++++++++++++++++++--------------------
 1 file changed, 21 insertions(+), 20 deletions(-)
lib/vsprintf.c +7 -7
index 63ca52366049..01c3aa638582 100644
@@ -380,13 +380,13 @@ enum format_type {
 };
 
 struct printf_spec {
-       u8      type;           /* format_type enum */
-       u8      flags;          /* flags to number() */
-       u8      base;           /* number base, 8, 10 or 16 only */
-       u8      qualifier;      /* number qualifier, one of 'hHlLtzZ' */
-       s16     field_width;    /* width of output field */
-       s16     precision;      /* # of digits/chars */
-};
+       unsigned int    type:8;         /* format_type enum */
+       signed int      field_width:24; /* width of output field */
+       unsigned int    flags:8;        /* flags to number() */
+       unsigned int    base:8;         /* number base, 8, 10 or 16 only */
+       signed int      precision:16;   /* # of digits/chars */
+} __packed;
+extern char __check_printf_spec[1-2*(sizeof(struct printf_spec) != 8)];
 
 static noinline_for_stack
 char *number(char *buf, char *end, unsigned long long num,

lib/vsprintf.c static noinline_for_stack · line 1644

Quoted 4 lines
lib/vsprintf.c +1 -0
@@ -1641,6 +1641,7 @@ static noinline_for_stack
 int format_decode(const char *fmt, struct printf_spec *spec)
 {
        const char *start = fmt;
+       char qualifier;

u8 ?

Quoted 61 lines
lib/vsprintf.c +13 -14
        /* we finished early by reading the field width */
        if (spec->type == FORMAT_TYPE_WIDTH) {
@@ -1723,16 +1724,16 @@ precision:
 
 qualifier:
        /* get the conversion qualifier */
-       spec->qualifier = -1;
+       qualifier = 0;
        if (*fmt == 'h' || _tolower(*fmt) == 'l' ||
            _tolower(*fmt) == 'z' || *fmt == 't') {
-               spec->qualifier = *fmt++;
-               if (unlikely(spec->qualifier == *fmt)) {
-                       if (spec->qualifier == 'l') {
-                               spec->qualifier = 'L';
+               qualifier = *fmt++;
+               if (unlikely(qualifier == *fmt)) {
+                       if (qualifier == 'l') {
+                               qualifier = 'L';
                                ++fmt;
-                       } else if (spec->qualifier == 'h') {
-                               spec->qualifier = 'H';
+                       } else if (qualifier == 'h') {
+                               qualifier = 'H';
                                ++fmt;
                        }
                }
@@ -1789,19 +1790,19 @@ qualifier:
                return fmt - start;
        }
 
-       if (spec->qualifier == 'L')
+       if (qualifier == 'L')
                spec->type = FORMAT_TYPE_LONG_LONG;
-       else if (spec->qualifier == 'l') {
+       else if (qualifier == 'l') {
                BUILD_BUG_ON(FORMAT_TYPE_ULONG + SIGN != FORMAT_TYPE_LONG);
                spec->type = FORMAT_TYPE_ULONG + (spec->flags & SIGN);
-       } else if (_tolower(spec->qualifier) == 'z') {
+       } else if (_tolower(qualifier) == 'z') {
                spec->type = FORMAT_TYPE_SIZE_T;
-       } else if (spec->qualifier == 't') {
+       } else if (qualifier == 't') {
                spec->type = FORMAT_TYPE_PTRDIFF;
-       } else if (spec->qualifier == 'H') {
+       } else if (qualifier == 'H') {
                BUILD_BUG_ON(FORMAT_TYPE_UBYTE + SIGN != FORMAT_TYPE_BYTE);
                spec->type = FORMAT_TYPE_UBYTE + (spec->flags & SIGN);
-       } else if (spec->qualifier == 'h') {
+       } else if (qualifier == 'h') {
                BUILD_BUG_ON(FORMAT_TYPE_USHORT + SIGN != FORMAT_TYPE_SHORT);
                spec->type = FORMAT_TYPE_USHORT + (spec->flags & SIGN);
        } else {
--

2.6.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author reviewer
#10 lore raw

Re: [PATCH 04/14] lib/vsprintf.c: expand field_width to 24 bits

Quoted 22 lines · Andy Shevchenko
Quoted 20 lines · Rasmus Villemoes

Maurizio Lombardi reported a problem [1] with the %pb extension: It doesn't work for sufficiently large bitmaps, since the size is stashed in the field_width field of the struct printf_spec, which is currently an s16. Concretely, this manifested itself in /sys/bus/pseudo/drivers/scsi_debug/map being empty, since the bitmap printer got a size of 0, which is the 16 bit truncation of the actual bitmap size.

We do want to keep struct printf_spec at 8 bytes so that it can cheaply be passed by value. The qualifier field is only used for internal bookkeeping in format_decode, so we might as well use a local variable for that. This gives us an additional 8 bits, which we can then use for the field width.

To stay in 8 bytes, we need to do a little rearranging and make the type member a bitfield as well. For consistency, change all the members to bit fields. gcc doesn't generate much worse code with these changes (in fact, bloat-o-meter says we save 300 bytes - which I think is a little surprising).

Quoted 4 lines
Quoted 2 lines

I didn't find a BUILD_BUG/compiletime_assertion/... which would work outside function context, so for now I just open-coded it.

And any objections to put it into vsnprintf() ?

I'd like to keep it close to the type definition. And I was hoping someone would come forward and say "yeah, that's been bugging me too, here's a patch I've been sitting on to fix that". Almost every compiler released this decade has _Static_assert, it's about time we start using that instead of the current mess of homegrown workarounds...

Rasmus

Acked-by on v1 4/14
#11 lore raw

Re: [PATCH 04/14] lib/vsprintf.c: expand field_width to 24 bits

In reply to Rasmus Villemoes
Quoted 23 lines · Rasmus Villemoes

Maurizio Lombardi reported a problem [1] with the %pb extension: It doesn't work for sufficiently large bitmaps, since the size is stashed in the field_width field of the struct printf_spec, which is currently an s16. Concretely, this manifested itself in /sys/bus/pseudo/drivers/scsi_debug/map being empty, since the bitmap printer got a size of 0, which is the 16 bit truncation of the actual bitmap size.

We do want to keep struct printf_spec at 8 bytes so that it can cheaply be passed by value. The qualifier field is only used for internal bookkeeping in format_decode, so we might as well use a local variable for that. This gives us an additional 8 bits, which we can then use for the field width.

To stay in 8 bytes, we need to do a little rearranging and make the type member a bitfield as well. For consistency, change all the members to bit fields. gcc doesn't generate much worse code with these changes (in fact, bloat-o-meter says we save 300 bytes - which I think is a little surprising).

I didn't find a BUILD_BUG/compiletime_assertion/... which would work outside function context, so for now I just open-coded it.

Quoted 6 lines

[1] http://thread.gmane.org/gmane.linux.kernel/2034835

FWIW,

Thanks.

Signature · 1 line
tejun
author reviewer
#12 lore raw

PATCH 5/14 lib/vsprintf.c: help gcc make number() smaller

In reply to Rasmus Villemoes

One consequence of the reorganization of struct printf_spec to make field_width 24 bits was that number() gained about 180 bytes. Since spec is never passed to other functions, we can help gcc make number() lose most of that extra weight by using local variables for the field width and precision.

---
 lib/vsprintf.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)
lib/vsprintf.c +14 -12
index 01c3aa638582..d7e27c54fa00 100644
@@ -399,6 +399,8 @@ char *number(char *buf, char *end, unsigned long long num,
 	int need_pfx = ((spec.flags & SPECIAL) && spec.base != 10);
 	int i;
 	bool is_zero = num == 0LL;
+	int field_width = spec.field_width;
+	int precision = spec.precision;
 
 	/* locase = 0 or 0x20. ORing digits or letters with 'locase'
 	 * produces same digits or (maybe lowercased) letters */
@@ -410,20 +412,20 @@ char *number(char *buf, char *end, unsigned long long num,
 		if ((signed long long)num < 0) {
 			sign = '-';
 			num = -(signed long long)num;
-			spec.field_width--;
+			field_width--;
 		} else if (spec.flags & PLUS) {
 			sign = '+';
-			spec.field_width--;
+			field_width--;
 		} else if (spec.flags & SPACE) {
 			sign = ' ';
-			spec.field_width--;
+			field_width--;
 		}
 	}
 	if (need_pfx) {
 		if (spec.base == 16)
-			spec.field_width -= 2;
+			field_width -= 2;
 		else if (!is_zero)
-			spec.field_width--;
+			field_width--;
 	}
 
 	/* generate full string in tmp[], in reverse order */
@@ -445,12 +447,12 @@ char *number(char *buf, char *end, unsigned long long num,
 	}
 
 	/* printing 100 using %2d gives "100", not "00" */
-	if (i > spec.precision)
-		spec.precision = i;
+	if (i > precision)
+		precision = i;
 	/* leading space padding */
-	spec.field_width -= spec.precision;
+	field_width -= precision;
 	if (!(spec.flags & (ZEROPAD | LEFT))) {
-		while (--spec.field_width >= 0) {
+		while (--field_width >= 0) {
 			if (buf < end)
 				*buf = ' ';
 			++buf;
@@ -479,14 +481,14 @@ char *number(char *buf, char *end, unsigned long long num,
 	if (!(spec.flags & LEFT)) {
 		char c = ' ' + (spec.flags & ZEROPAD);
 		BUILD_BUG_ON(' ' + ZEROPAD != '0');
-		while (--spec.field_width >= 0) {
+		while (--field_width >= 0) {
 			if (buf < end)
 				*buf = c;
 			++buf;
 		}
 	}
 	/* hmm even more zero padding? */
-	while (i <= --spec.precision) {
+	while (i <= --precision) {
 		if (buf < end)
 			*buf = '0';
 		++buf;
@@ -498,7 +500,7 @@ char *number(char *buf, char *end, unsigned long long num,
 		++buf;
 	}
 	/* trailing space padding */
-	while (--spec.field_width >= 0) {
+	while (--field_width >= 0) {
 		if (buf < end)
 			*buf = ' ';
 		++buf;
Signature · 1 line
2.6.1
Reviewed-by on v1 5/14
#13 lore raw

Re: [PATCH 05/14] lib/vsprintf.c: help gcc make number() smaller

Quoted 6 lines · Rasmus Villemoes

One consequence of the reorganization of struct printf_spec to make field_width 24 bits was that number() gained about 180 bytes. Since spec is never passed to other functions, we can help gcc make number() lose most of that extra weight by using local variables for the field width and precision.

One minor nitpick below

Quoted 10 lines
---
 lib/vsprintf.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)
lib/vsprintf.c +0 -0
index 01c3aa638582..d7e27c54fa00 100644

lib/vsprintf.c char *number(char *buf, char *end, unsigned long long num, · line 403

Quoted 5 lines
lib/vsprintf.c +2 -0
@@ -399,6 +399,8 @@ char *number(char *buf, char *end, unsigned long long num,
        int need_pfx = ((spec.flags & SPECIAL) && spec.base != 10);
        int i;
        bool is_zero = num == 0LL;
+       int field_width = spec.field_width;
+       int precision = spec.precision;

I would put these before int i; But your choice.

Quoted 80 lines
lib/vsprintf.c +12 -13
        /* locase = 0 or 0x20. ORing digits or letters with 'locase'
         * produces same digits or (maybe lowercased) letters */
@@ -410,20 +412,20 @@ char *number(char *buf, char *end, unsigned long long num,
                if ((signed long long)num < 0) {
                        sign = '-';
                        num = -(signed long long)num;
-                       spec.field_width--;
+                       field_width--;
                } else if (spec.flags & PLUS) {
                        sign = '+';
-                       spec.field_width--;
+                       field_width--;
                } else if (spec.flags & SPACE) {
                        sign = ' ';
-                       spec.field_width--;
+                       field_width--;
                }
        }
        if (need_pfx) {
                if (spec.base == 16)
-                       spec.field_width -= 2;
+                       field_width -= 2;
                else if (!is_zero)
-                       spec.field_width--;
+                       field_width--;
        }
 
        /* generate full string in tmp[], in reverse order */
@@ -445,12 +447,12 @@ char *number(char *buf, char *end, unsigned long long num,
        }
 
        /* printing 100 using %2d gives "100", not "00" */
-       if (i > spec.precision)
-               spec.precision = i;
+       if (i > precision)
+               precision = i;
        /* leading space padding */
-       spec.field_width -= spec.precision;
+       field_width -= precision;
        if (!(spec.flags & (ZEROPAD | LEFT))) {
-               while (--spec.field_width >= 0) {
+               while (--field_width >= 0) {
                        if (buf < end)
                                *buf = ' ';
                        ++buf;
@@ -479,14 +481,14 @@ char *number(char *buf, char *end, unsigned long long num,
        if (!(spec.flags & LEFT)) {
                char c = ' ' + (spec.flags & ZEROPAD);
                BUILD_BUG_ON(' ' + ZEROPAD != '0');
-               while (--spec.field_width >= 0) {
+               while (--field_width >= 0) {
                        if (buf < end)
                                *buf = c;
                        ++buf;
                }
        }
        /* hmm even more zero padding? */
-       while (i <= --spec.precision) {
+       while (i <= --precision) {
                if (buf < end)
                        *buf = '0';
                ++buf;
@@ -498,7 +500,7 @@ char *number(char *buf, char *end, unsigned long long num,
                ++buf;
        }
        /* trailing space padding */
-       while (--spec.field_width >= 0) {
+       while (--field_width >= 0) {
                if (buf < end)
                        *buf = ' ';
                ++buf;
--

2.6.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author reviewer
#14 lore raw

PATCH 6/14 lib/vsprintf.c: warn about too large precisions and field widths

In reply to Rasmus Villemoes

The field width is overloaded to pass some extra information for some %p extensions (e.g. #bits for %pb). But we might silently truncate the passed value when we stash it in struct printf_spec (see e.g. "lib/vsprintf.c: expand field_width to 24 bits"). Hopefully 23 value bits should now be enough for everybody, but if not, let's make some noise.

Do the same for the precision. In both cases, clamping seems more sensible than truncating. While, according to POSIX, "A negative precision is taken as if the precision were omitted.", the kernel's printf has always treated that case as if the precision was 0, so we use that as lower bound. For the field width, the smallest representable value is actually -(1<<23), but a negative field width means 'set the LEFT flag and use the absolute value', so we want the absolute value to fit.

---
 lib/vsprintf.c | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)
lib/vsprintf.c +22 -2
index d7e27c54fa00..8af5535fd738 100644
@@ -386,6 +386,8 @@ struct printf_spec {
 	unsigned int	base:8;		/* number base, 8, 10 or 16 only */
 	signed int	precision:16;	/* # of digits/chars */
 } __packed;
+#define FIELD_WIDTH_MAX ((1 << 23) - 1)
+#define PRECISION_MAX ((1 << 15) - 1)
 extern char __check_printf_spec[1-2*(sizeof(struct printf_spec) != 8)];
 
 static noinline_for_stack
@@ -1815,6 +1817,24 @@ qualifier:
 	return ++fmt - start;
 }
 
+static inline void
+set_field_width(struct printf_spec *spec, int width)
+{
+	spec->field_width = width;
+	if (WARN_ONCE(spec->field_width != width, "field width %d too large", width)) {
+		spec->field_width = clamp(width, -FIELD_WIDTH_MAX, FIELD_WIDTH_MAX);
+	}
+}
+
+static inline void
+set_precision(struct printf_spec *spec, int prec)
+{
+	spec->precision = prec;
+	if (WARN_ONCE(spec->precision != prec, "precision %d too large", prec)) {
+		spec->precision = clamp(prec, 0, PRECISION_MAX);
+	}
+}
+
 /**
  * vsnprintf - Format a string and place it in a buffer
  * @buf: The buffer to place the result into
@@ -1882,11 +1902,11 @@ int vsnprintf(char *buf, size_t size, const char *fmt, va_list args)
 		}
 
 		case FORMAT_TYPE_WIDTH:
-			spec.field_width = va_arg(args, int);
+			set_field_width(&spec, va_arg(args, int));
 			break;
 
 		case FORMAT_TYPE_PRECISION:
-			spec.precision = va_arg(args, int);
+			set_precision(&spec, va_arg(args, int));
 			break;
 
 		case FORMAT_TYPE_CHAR: {
Signature · 1 line
2.6.1
#15 lore raw

Re: [PATCH 06/14] lib/vsprintf.c: warn about too large precisions and field widths

Quoted 7 lines · Rasmus Villemoes

The field width is overloaded to pass some extra information for some %p extensions (e.g. #bits for %pb). But we might silently truncate the passed value when we stash it in struct printf_spec (see e.g. "lib/vsprintf.c: expand field_width to 24 bits"). Hopefully 23 value bits should now be enough for everybody, but if not, let's make some noise.

Quoted 9 lines

Do the same for the precision. In both cases, clamping seems more sensible than truncating. While, according to POSIX, "A negative precision is taken as if the precision were omitted.", the kernel's printf has always treated that case as if the precision was 0, so we use that as lower bound. For the field width, the smallest representable value is actually -(1<<23), but a negative field width means 'set the LEFT flag and use the absolute value', so we want the absolute value to fit.

Do we need to do the same for bstr_printf() ?

Quoted 65 lines
---
 lib/vsprintf.c | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)
lib/vsprintf.c +22 -3
index d7e27c54fa00..8af5535fd738 100644
@@ -386,6 +386,8 @@ struct printf_spec {
        unsigned int    base:8;         /* number base, 8, 10 or 16 only */
        signed int      precision:16;   /* # of digits/chars */
 } __packed;
+#define FIELD_WIDTH_MAX ((1 << 23) - 1)
+#define PRECISION_MAX ((1 << 15) - 1)
 extern char __check_printf_spec[1-2*(sizeof(struct printf_spec) != 8)];
 
 static noinline_for_stack
@@ -1815,6 +1817,24 @@ qualifier:
        return ++fmt - start;
 }
 
+static inline void
+set_field_width(struct printf_spec *spec, int width)
+{
+       spec->field_width = width;
+       if (WARN_ONCE(spec->field_width != width, "field width %d too large", width)) {
+               spec->field_width = clamp(width, -FIELD_WIDTH_MAX, FIELD_WIDTH_MAX);
+       }
+}
+
+static inline void
+set_precision(struct printf_spec *spec, int prec)
+{
+       spec->precision = prec;
+       if (WARN_ONCE(spec->precision != prec, "precision %d too large", prec)) {
+               spec->precision = clamp(prec, 0, PRECISION_MAX);
+       }
+}
+
 /**
  * vsnprintf - Format a string and place it in a buffer
  * @buf: The buffer to place the result into
@@ -1882,11 +1902,11 @@ int vsnprintf(char *buf, size_t size, const char *fmt, va_list args)
                }
 
                case FORMAT_TYPE_WIDTH:
-                       spec.field_width = va_arg(args, int);
+                       set_field_width(&spec, va_arg(args, int));
                        break;
 
                case FORMAT_TYPE_PRECISION:
-                       spec.precision = va_arg(args, int);
+                       set_precision(&spec, va_arg(args, int));
                        break;
 
                case FORMAT_TYPE_CHAR: {
--

2.6.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author reviewer
#16 lore raw

Re: [PATCH 06/14] lib/vsprintf.c: warn about too large precisions and field widths

Quoted 19 lines · Andy Shevchenko
Quoted 7 lines · Rasmus Villemoes

The field width is overloaded to pass some extra information for some %p extensions (e.g. #bits for %pb). But we might silently truncate the passed value when we stash it in struct printf_spec (see e.g. "lib/vsprintf.c: expand field_width to 24 bits"). Hopefully 23 value bits should now be enough for everybody, but if not, let's make some noise.

Quoted 9 lines

Do the same for the precision. In both cases, clamping seems more sensible than truncating. While, according to POSIX, "A negative precision is taken as if the precision were omitted.", the kernel's printf has always treated that case as if the precision was 0, so we use that as lower bound. For the field width, the smallest representable value is actually -(1<<23), but a negative field width means 'set the LEFT flag and use the absolute value', so we want the absolute value to fit.

Quoted 2 lines

Do we need to do the same for bstr_printf() ?

Heh, apparently I didn't learn anything from 762abb51. Thanks, will fix in next spin.

Rasmus

author reviewer
#17 lore raw

PATCH 7/14 lib/vsprintf.c: slightly refactor vscnprintf()

In reply to Rasmus Villemoes

If we're given a size of 0, the vsnprintf() won't have any side effects, and neither "i < size" or "size != 0" will trigger. So we might as well return 0 immediately.

---
 lib/vsprintf.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
lib/vsprintf.c +4 -3
index 8af5535fd738..e22a6189548f 100644
@@ -2036,13 +2036,14 @@ int vscnprintf(char *buf, size_t size, const char *fmt, va_list args)
 {
 	int i;
 
+	if (unlikely(!size))
+		return 0;
+
 	i = vsnprintf(buf, size, fmt, args);
 
 	if (likely(i < size))
 		return i;
-	if (size != 0)
-		return size - 1;
-	return 0;
+	return size - 1;
 }
 EXPORT_SYMBOL(vscnprintf);
 
Signature · 1 line
2.6.1
#18 lore raw

Re: [PATCH 07/14] lib/vsprintf.c: slightly refactor vscnprintf()

Quoted 14 lines · Rasmus Villemoes

If we're given a size of 0, the vsnprintf() won't have any side effects, and neither "i < size" or "size != 0" will trigger. So we might as well return 0 immediately.

---
 lib/vsprintf.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
lib/vsprintf.c +0 -0
index 8af5535fd738..e22a6189548f 100644

lib/vsprintf.c int vscnprintf(char *buf, size_t size, const char *fmt, va_list args) · line 2041

Quoted 6 lines
lib/vsprintf.c +3 -0
@@ -2036,13 +2036,14 @@ int vscnprintf(char *buf, size_t size, const char *fmt, va_list args)
 {
        int i;
 
+       if (unlikely(!size))
+               return 0;
+

Might it potentially shadow any issue when run vsnprintf(buf, 0, fmt, args); with certain arguments?
I can imagine something like %pV with unstable pointer.

Quoted 19 lines
lib/vsprintf.c +1 -4
        i = vsnprintf(buf, size, fmt, args);
 
        if (likely(i < size))
                return i;
-       if (size != 0)
-               return size - 1;
-       return 0;
+       return size - 1;
 }
 EXPORT_SYMBOL(vscnprintf);
 
--

2.6.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author reviewer
#19 lore raw

Re: [PATCH 07/14] lib/vsprintf.c: slightly refactor vscnprintf()

Quoted 19 lines · Andy Shevchenko
Quoted 17 lines · Rasmus Villemoes

If we're given a size of 0, the vsnprintf() won't have any side effects, and neither "i < size" or "size != 0" will trigger. So we might as well return 0 immediately.

---
 lib/vsprintf.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
lib/vsprintf.c +0 -0
index 8af5535fd738..e22a6189548f 100644
@@ -2036,13 +2036,14 @@ int vscnprintf(char *buf, size_t size, const char *fmt, va_list args)
 {
        int i;
 
Quoted 6 lines

lib/vsprintf.c int vscnprintf(char *buf, size_t size, const char *fmt, va_list args) · line 2041

Quoted 3 lines
lib/vsprintf.c +3 -0
+       if (unlikely(!size))
+               return 0;
+

Might it potentially shadow any issue when run vsnprintf(buf, 0, fmt, args); with certain arguments?

Only if we ever come up with a %p extension with side effects, but then people couldn't rely on them happening exactly once anyway (kasprintf would make them happen twice). printf-like calls are also often compiled out or disabled (dyndebug, ratelimit, ...) without it being obvious at the call site whether they'll run or not, so I think such a hypothetical %p extension would meet some resistance.

Quoted 1 line

I can imagine something like %pV with unstable pointer.

I don't see how %pV is different than any other current %p extensions.

Rasmus

author reviewer
#20 lore raw

PATCH 8/14 lib/kasprintf.c: add sanity check to kvasprintf

In reply to Rasmus Villemoes

kasprintf relies on being able to replay the formatting and getting the same result (in particular, the same length). This will almost always work, but it is possible that the object pointed to by a %s or %p argument changed under us (so we might get truncated output). Add a somewhat paranoid sanity check and let's see if it ever triggers.

---
 lib/kasprintf.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
lib/kasprintf.c +6 -4
index f194e6e593e1..7f6c506a4942 100644
@@ -13,19 +13,21 @@
 /* Simplified asprintf. */
 char *kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
 {
-	unsigned int len;
+	unsigned int first, second;
 	char *p;
 	va_list aq;
 
 	va_copy(aq, ap);
-	len = vsnprintf(NULL, 0, fmt, aq);
+	first = vsnprintf(NULL, 0, fmt, aq);
 	va_end(aq);
 
-	p = kmalloc_track_caller(len+1, gfp);
+	p = kmalloc_track_caller(first+1, gfp);
 	if (!p)
 		return NULL;
 
-	vsnprintf(p, len+1, fmt, ap);
+	second = vsnprintf(p, first+1, fmt, ap);
+	WARN(first != second, "different return values (%u and %u) from vsnprintf(\"%s\", ...)",
+	     first, second, fmt);
 
 	return p;
 }
Signature · 1 line
2.6.1
#21 lore raw

Re: [PATCH 08/14] lib/kasprintf.c: add sanity check to kvasprintf

Quoted 16 lines · Rasmus Villemoes

kasprintf relies on being able to replay the formatting and getting the same result (in particular, the same length). This will almost always work, but it is possible that the object pointed to by a %s or %p argument changed under us (so we might get truncated output). Add a somewhat paranoid sanity check and let's see if it ever triggers.

---
 lib/kasprintf.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
lib/kasprintf.c +0 -0
index f194e6e593e1..7f6c506a4942 100644

lib/kasprintf.c line 16

Quoted 5 lines
lib/kasprintf.c +1 -1
@@ -13,19 +13,21 @@
 /* Simplified asprintf. */
 char *kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
 {
-       unsigned int len;
+       unsigned int first, second;

vsnprintf() returns plain int.
Perhaps change the type to follow the function.

Quoted 28 lines
lib/kasprintf.c +5 -4
        char *p;
        va_list aq;
 
        va_copy(aq, ap);
-       len = vsnprintf(NULL, 0, fmt, aq);
+       first = vsnprintf(NULL, 0, fmt, aq);
        va_end(aq);
 
-       p = kmalloc_track_caller(len+1, gfp);
+       p = kmalloc_track_caller(first+1, gfp);
        if (!p)
                return NULL;
 
-       vsnprintf(p, len+1, fmt, ap);
+       second = vsnprintf(p, first+1, fmt, ap);
+       WARN(first != second, "different return values (%u and %u) from vsnprintf(\"%s\", ...)",
+            first, second, fmt);
 
        return p;
 }
--

2.6.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author reviewer
#22 lore raw

PATCH 9/14 lib/test_printf.c: don't BUG

In reply to Rasmus Villemoes

BUG is a completely unnecessarily big hammer, and we're more likely to get the internal bug reported if we just pr_err() and ensure the test suite fails.

---
 lib/test_printf.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
lib/test_printf.c +6 -1
index c5a666af9ba5..9232a2add28c 100644
@@ -91,7 +91,12 @@ __test(const char *expect, int elen, const char *fmt, ...)
 	int rand;
 	char *p;
 
-	BUG_ON(elen >= BUF_SIZE);
+	if (elen >= BUF_SIZE) {
+		pr_err("error in test suite: expected output length %d too long. Format was '%s'.\n",
+		       elen, fmt);
+		failed_tests++;
+		return;
+	}
 
 	va_start(ap, fmt);
 
Signature · 1 line
2.6.1
author reviewer
#23 lore raw

PATCH 10/14 lib/test_printf.c: check for out-of-bound writes

In reply to Rasmus Villemoes

Add a few padding bytes on either side of the test buffer, and check that these (and the part of the buffer not used) are untouched by vsnprintf.

---
 lib/test_printf.c | 24 +++++++++++++++++++-----
 1 file changed, 19 insertions(+), 5 deletions(-)
lib/test_printf.c +19 -5
index 9232a2add28c..1ce1a1dd8faf 100644
@@ -16,6 +16,7 @@
 #include <linux/in.h>
 
 #define BUF_SIZE 256
+#define PAD_SIZE 16
 #define FILL_CHAR '$'
 
 #define PTR1 ((void*)0x01234567)
@@ -39,6 +40,7 @@
 static unsigned total_tests __initdata;
 static unsigned failed_tests __initdata;
 static char *test_buffer __initdata;
+static char *alloced_buffer __initdata;
 
 static int __printf(4, 0) __init
 do_test(int bufsize, const char *expect, int elen,
@@ -49,7 +51,7 @@ do_test(int bufsize, const char *expect, int elen,
 
 	total_tests++;
 
-	memset(test_buffer, FILL_CHAR, BUF_SIZE);
+	memset(alloced_buffer, FILL_CHAR, BUF_SIZE + 2*PAD_SIZE);
 	va_copy(aq, ap);
 	ret = vsnprintf(test_buffer, bufsize, fmt, aq);
 	va_end(aq);
@@ -60,8 +62,13 @@ do_test(int bufsize, const char *expect, int elen,
 		return 1;
 	}
 
+	if (memchr_inv(alloced_buffer, FILL_CHAR, PAD_SIZE)) {
+		pr_warn("vsnprintf(buf, %d, \"%s\", ...) wrote before buffer\n", bufsize, fmt);
+		return 1;
+	}
+
 	if (!bufsize) {
-		if (memchr_inv(test_buffer, FILL_CHAR, BUF_SIZE)) {
+		if (memchr_inv(test_buffer, FILL_CHAR, BUF_SIZE + PAD_SIZE)) {
 			pr_warn("vsnprintf(buf, 0, \"%s\", ...) wrote to buffer\n",
 				fmt);
 			return 1;
@@ -76,6 +83,12 @@ do_test(int bufsize, const char *expect, int elen,
 		return 1;
 	}
 
+	if (memchr_inv(test_buffer + written + 1, FILL_CHAR, BUF_SIZE + PAD_SIZE - (written + 1))) {
+		pr_warn("vsnprintf(buf, %d, \"%s\", ...) wrote beyond the nul-terminator\n",
+			bufsize, fmt);
+		return 1;
+	}
+
 	if (memcmp(test_buffer, expect, written)) {
 		pr_warn("vsnprintf(buf, %d, \"%s\", ...) wrote '%s', expected '%.*s'\n",
 			bufsize, fmt, test_buffer, written, expect);
@@ -342,16 +355,17 @@ test_pointer(void)
 static int __init
 test_printf_init(void)
 {
-	test_buffer = kmalloc(BUF_SIZE, GFP_KERNEL);
-	if (!test_buffer)
+	alloced_buffer = kmalloc(BUF_SIZE + 2*PAD_SIZE, GFP_KERNEL);
+	if (!alloced_buffer)
 		return -ENOMEM;
+	test_buffer = alloced_buffer + PAD_SIZE;
 
 	test_basic();
 	test_number();
 	test_string();
 	test_pointer();
 
-	kfree(test_buffer);
+	kfree(alloced_buffer);
 
 	if (failed_tests == 0)
 		pr_info("all %u tests passed\n", total_tests);
Signature · 1 line
2.6.1
author reviewer
#24 lore raw

PATCH 11/14 lib/test_printf.c: test precision quirks

In reply to Rasmus Villemoes

The kernel's printf doesn't follow the standards in a few corner cases (which are probably mostly irrelevant). Add tests that document the current behaviour.

---
 lib/test_printf.c | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)
lib/test_printf.c +14 -6
index 1ce1a1dd8faf..5f742b99cfdc 100644
@@ -166,14 +166,22 @@ test_string(void)
 	test("", "%s%.0s", "", "123");
 	test("ABCD|abc|123", "%s|%.3s|%.*s", "ABCD", "abcdef", 3, "123456");
 	test("1  |  2|3  |  4|5  ", "%-3s|%3s|%-*s|%*s|%*s", "1", "2", 3, "3", 3, "4", -3, "5");
+	test("1234      ", "%-10.4s", "123456");
+	test("      1234", "%10.4s", "123456");
 	/*
-	 * POSIX and C99 say that a missing precision should be
-	 * treated as a precision of 0. However, the kernel's printf
-	 * implementation treats this case as if the . wasn't
-	 * present. Let's add a test case documenting the current
-	 * behaviour; should anyone ever feel the need to follow the
-	 * standards more closely, this can be revisited.
+	 * POSIX and C99 say that a negative precision (which is only
+	 * possible to pass via a * argument) should be treated as if
+	 * the precision wasn't present, and that if the precision is
+	 * omitted (as in %.s), the precision should be taken to be
+	 * 0. However, the kernel's printf behave exactly opposite,
+	 * treating a negative precision as 0 and treating an omitted
+	 * precision specifier as if no precision was given.
+	 *
+	 * These test cases document the current behaviour; should
+	 * anyone ever feel the need to follow the standards more
+	 * closely, this can be revisited.
 	 */
+	test("    ", "%4.*s", -5, "123456");
 	test("a||", "%.s|%.0s|%.*s", "a", "b", 0, "c");
 	test("a  |   |   ", "%-3.s|%-3.0s|%-3.*s", "a", "b", 0, "c");
 }
Signature · 1 line
2.6.1
author reviewer
#25 lore raw

PATCH 12/14 lib/test_printf.c: account for kvasprintf tests

In reply to Rasmus Villemoes

These should also count as performed tests.

---
 lib/test_printf.c | 1 +
 1 file changed, 1 insertion(+)
lib/test_printf.c +1 -0
index 5f742b99cfdc..fc9169d360ba 100644
@@ -127,6 +127,7 @@ __test(const char *expect, int elen, const char *fmt, ...)
 
 	p = kvasprintf(GFP_KERNEL, fmt, ap);
 	if (p) {
+		total_tests++;
 		if (memcmp(p, expect, elen+1)) {
 			pr_warn("kvasprintf(..., \"%s\", ...) returned '%s', expected '%s'\n",
 				fmt, p, expect);
Signature · 1 line
2.6.1
author reviewer
#26 lore raw

PATCH 13/14 lib/test_printf.c: add test for large bitmaps

In reply to Rasmus Villemoes

Following "lib/vsprintf.c: expand field_width to 24 bits", let's add a test to see that we now actually support bitmaps with 65536 bits.

---
 lib/test_printf.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
lib/test_printf.c +17 -0
index fc9169d360ba..705907aec901 100644
@@ -12,6 +12,7 @@
 #include <linux/slab.h>
 #include <linux/string.h>
 
+#include <linux/bitmap.h>
 #include <linux/socket.h>
 #include <linux/in.h>
 
@@ -316,6 +317,20 @@ struct_clk(void)
 }
 
 static void __init
+large_bitmap(void)
+{
+	const int nbits = 1 << 16;
+	unsigned long *bits = kcalloc(BITS_TO_LONGS(nbits), sizeof(long), GFP_KERNEL);
+	if (!bits)
+		return;
+
+	bitmap_set(bits, 1, 20);
+	bitmap_set(bits, 60000, 15);
+	test("1-20,60000-60014", "%*pbl", nbits, bits);
+	kfree(bits);
+}
+
+static void __init
 bitmap(void)
 {
 	DECLARE_BITMAP(bits, 20);
@@ -334,6 +349,8 @@ bitmap(void)
 	bitmap_fill(bits, 20);
 	test("fffff|fffff", "%20pb|%*pb", bits, 20, bits);
 	test("0-19|0-19", "%20pbl|%*pbl", bits, 20, bits);
+
+	large_bitmap();
 }
 
 static void __init
Signature · 1 line
2.6.1
author reviewer
#27 lore raw

PATCH 14/14 lib/test_printf.c: test dentry printing

In reply to Rasmus Villemoes
---
 lib/test_printf.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)
lib/test_printf.c +27 -0
index 705907aec901..6b7c4a1ec141 100644
@@ -13,6 +13,7 @@
 #include <linux/string.h>
 
 #include <linux/bitmap.h>
+#include <linux/dcache.h>
 #include <linux/socket.h>
 #include <linux/in.h>
 
@@ -301,9 +302,35 @@ uuid(void)
 	test("03020100-0504-0706-0809-0A0B0C0D0E0F", "%pUL", uuid);
 }
 
+static struct dentry test_dentry[4] __initdata = {
+	{ .d_parent = &test_dentry[0],
+	  .d_name = { .len = 3, .name = test_dentry[0].d_iname },
+	  .d_iname = "foo" },
+	{ .d_parent = &test_dentry[0],
+	  .d_name = { .len = 5, .name = test_dentry[1].d_iname },
+	  .d_iname = "bravo" },
+	{ .d_parent = &test_dentry[1],
+	  .d_name = { .len = 4, .name = test_dentry[2].d_iname },
+	  .d_iname = "alfa" },
+	{ .d_parent = &test_dentry[2],
+	  .d_name = { .len = 5, .name = test_dentry[3].d_iname },
+	  .d_iname = "romeo" },
+};
+
 static void __init
 dentry(void)
 {
+	test("foo", "%pd", &test_dentry[0]);
+	test("foo", "%pd2", &test_dentry[0]);
+
+	test("romeo", "%pd", &test_dentry[3]);
+	test("alfa/romeo", "%pd2", &test_dentry[3]);
+	test("bravo/alfa/romeo", "%pd3", &test_dentry[3]);
+	test("/bravo/alfa/romeo", "%pd4", &test_dentry[3]);
+	test("/bravo/alfa", "%pd4", &test_dentry[2]);
+
+	test("bravo/alfa  |bravo/alfa  ", "%-12pd2|%*pd2", &test_dentry[2], -12, &test_dentry[2]);
+	test("  bravo/alfa|  bravo/alfa", "%12pd2|%*pd2", &test_dentry[2], 12, &test_dentry[2]);
 }
 
 static void __init
Signature · 1 line
2.6.1