out of range struct tm fields in strftime

out of range struct tm fields in strftime

[Szabolcs Nagy]

[-- Attachment #1: Type: text/plain, Size: 186 bytes --]

out of range tm fields should not be treated as ub
as noted in the thread
http://sourceware.org/ml/libc-alpha/2015-09/msg00546.html

i have a patch but there might be simpler approaches

[-- Attachment #2: 0001-fix-strftime-to-handle-out-of-range-tm-fields-withou.patch --]
[-- Type: text/x-diff, Size: 4543 bytes --]

From 042305a745991d2de8e7fd0111f158086d283f35 Mon Sep 17 00:00:00 2001
From: Szabolcs Nagy <nsz@port70.net>
Date: Sun, 20 Sep 2015 12:02:21 +0000
Subject: [PATCH] fix strftime to handle out of range tm fields without UB

Calling strftime with out of range tm fields is not undefined
behaviour, it should return a result (though in this case the stored
string is unspecified).

tm_wday, tm_yday, tm_mon and tm_year fields are used in signed int
arithmetics that may overflow. For tm_year long long arithmetics is
used and the rest is fixed by limiting the input.
---
 src/time/strftime.c | 46 +++++++++++++++++++++++++---------------------
 1 file changed, 25 insertions(+), 21 deletions(-)

diff --git a/src/time/strftime.c b/src/time/strftime.c
index e945bb7..464b546 100644
--- a/src/time/strftime.c
+++ b/src/time/strftime.c
@@ -19,27 +19,27 @@ static int is_leap(int y)
 	return !(y%4) && ((y%100) || !(y%400));
 }
 
-static int week_num(const struct tm *tm)
+static int week_num(int wday, int yday, int year)
 {
-	int val = (tm->tm_yday + 7 - (tm->tm_wday+6)%7) / 7;
+	int val = (yday + 7 - (wday+6)%7) / 7;
 	/* If 1 Jan is just 1-3 days past Monday,
 	 * the previous week is also in this year. */
-	if ((tm->tm_wday - tm->tm_yday - 2 + 371) % 7 <= 2)
+	if ((wday - yday - 2 + 371) % 7 <= 2)
 		val++;
 	if (!val) {
 		val = 52;
 		/* If 31 December of prev year a Thursday,
 		 * or Friday of a leap year, then the
 		 * prev year has 53 weeks. */
-		int dec31 = (tm->tm_wday - tm->tm_yday - 1 + 7) % 7;
-		if (dec31 == 4 || (dec31 == 5 && is_leap(tm->tm_year%400-1)))
+		int dec31 = (wday - yday - 1 + 7) % 7;
+		if (dec31 == 4 || (dec31 == 5 && is_leap(year%400-1)))
 			val++;
 	} else if (val == 53) {
 		/* If 1 January is not a Thursday, and not
 		 * a Wednesday of a leap year, then this
 		 * year has only 52 weeks. */
-		int jan1 = (tm->tm_wday - tm->tm_yday + 371) % 7;
-		if (jan1 != 4 && (jan1 != 3 || !is_leap(tm->tm_year)))
+		int jan1 = (wday - yday + 371) % 7;
+		if (jan1 != 4 && (jan1 != 3 || !is_leap(year)))
 			val = 1;
 	}
 	return val;
@@ -55,19 +55,23 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 	const char *fmt;
 	int width = 2;
 
+	int wday = tm->tm_wday < 7U ? tm->tm_wday : 0;
+	int yday = tm->tm_yday < 366U ? tm->tm_yday : 0;
+	int mon = tm->tm_mon < 12U ? tm->tm_mon : 0;
+
 	switch (f) {
 	case 'a':
-		item = ABDAY_1 + tm->tm_wday;
+		item = ABDAY_1 + wday;
 		goto nl_strcat;
 	case 'A':
-		item = DAY_1 + tm->tm_wday;
+		item = DAY_1 + wday;
 		goto nl_strcat;
 	case 'h':
 	case 'b':
-		item = ABMON_1 + tm->tm_mon;
+		item = ABMON_1 + mon;
 		goto nl_strcat;
 	case 'B':
-		item = MON_1 + tm->tm_mon;
+		item = MON_1 + mon;
 		goto nl_strcat;
 	case 'c':
 		item = D_T_FMT;
@@ -90,8 +94,8 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 	case 'g':
 	case 'G':
 		val = tm->tm_year + 1900LL;
-		if (tm->tm_yday < 3 && week_num(tm) != 1) val--;
-		else if (tm->tm_yday > 360 && week_num(tm) == 1) val++;
+		if (yday < 3 && week_num(wday, yday, tm->tm_year) != 1) val--;
+		else if (yday > 360 && week_num(wday, yday, tm->tm_year) == 1) val++;
 		if (f=='g') val %= 100;
 		else width = 4;
 		goto number;
@@ -104,11 +108,11 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 		else if (val > 12) val -= 12;
 		goto number;
 	case 'j':
-		val = tm->tm_yday+1;
+		val = yday+1;
 		width = 3;
 		goto number;
 	case 'm':
-		val = tm->tm_mon+1;
+		val = mon+1;
 		goto number;
 	case 'M':
 		val = tm->tm_min;
@@ -139,20 +143,20 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 		fmt = "%H:%M:%S";
 		goto recu_strftime;
 	case 'u':
-		val = tm->tm_wday ? tm->tm_wday : 7;
+		val = wday ? wday : 7;
 		width = 1;
 		goto number;
 	case 'U':
-		val = (tm->tm_yday + 7 - tm->tm_wday) / 7;
+		val = (yday + 7 - wday) / 7;
 		goto number;
 	case 'W':
-		val = (tm->tm_yday + 7 - (tm->tm_wday+6)%7) / 7;
+		val = (yday + 7 - (wday+6)%7) / 7;
 		goto number;
 	case 'V':
-		val = week_num(tm);
+		val = week_num(wday, yday, tm->tm_year);
 		goto number;
 	case 'w':
-		val = tm->tm_wday;
+		val = wday;
 		width = 1;
 		goto number;
 	case 'x':
@@ -165,7 +169,7 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 		val = tm->tm_year % 100;
 		goto number;
 	case 'Y':
-		val = tm->tm_year + 1900;
+		val = tm->tm_year + 1900LL;
 		if (val >= 10000) {
 			*l = snprintf(*s, sizeof *s, "+%lld", val);
 			return *s;
-- 
2.4.1

Re: out of range struct tm fields in strftime

[Rich Felker]

On Sun, Sep 20, 2015 at 02:44:50PM +0200, Szabolcs Nagy wrote:
> out of range tm fields should not be treated as ub
> as noted in the thread
> http://sourceware.org/ml/libc-alpha/2015-09/msg00546.html
> 
> i have a patch but there might be simpler approaches

Wouldn't it be less invasive to just make some small changes like
putting a U on some of the constants so that the arithmetic happens as
unsigned?

Rich

Re: out of range struct tm fields in strftime

[Szabolcs Nagy]

* Rich Felker <dalias@libc.org> [2015-09-20 12:36:29 -0400]:
> On Sun, Sep 20, 2015 at 02:44:50PM +0200, Szabolcs Nagy wrote:
> > out of range tm fields should not be treated as ub
> > as noted in the thread
> > http://sourceware.org/ml/libc-alpha/2015-09/msg00546.html
> > 
> > i have a patch but there might be simpler approaches
> 
> Wouldn't it be less invasive to just make some small changes like
> putting a U on some of the constants so that the arithmetic happens as
> unsigned?
> 

that might be simpler (and probably generates better
code for div,mod by const)

but for the nl_langinfo item computation the range
has to be limited properly (tm_wday and tm_mon are
affected).

Re: out of range struct tm fields in strftime

[Szabolcs Nagy]

[-- Attachment #1: Type: text/plain, Size: 845 bytes --]

* Szabolcs Nagy <nsz@port70.net> [2015-09-20 18:44:35 +0200]:
> * Rich Felker <dalias@libc.org> [2015-09-20 12:36:29 -0400]:
> > On Sun, Sep 20, 2015 at 02:44:50PM +0200, Szabolcs Nagy wrote:
> > > out of range tm fields should not be treated as ub
> > > as noted in the thread
> > > http://sourceware.org/ml/libc-alpha/2015-09/msg00546.html
> > > 
> > > i have a patch but there might be simpler approaches
> > 
> > Wouldn't it be less invasive to just make some small changes like
> > putting a U on some of the constants so that the arithmetic happens as
> > unsigned?
> > 
> 
> that might be simpler (and probably generates better
> code for div,mod by const)
> 
> but for the nl_langinfo item computation the range
> has to be limited properly (tm_wday and tm_mon are
> affected).

implemented this approach with wday%7U and using some 0U+

[-- Attachment #2: 0001-fix-strftime-to-handle-out-of-range-tm-fields-withou.patch --]
[-- Type: text/x-diff, Size: 3058 bytes --]

From e8891b06e0c698c0334f08e996f3b0d733f8ede7 Mon Sep 17 00:00:00 2001
From: Szabolcs Nagy <nsz@port70.net>
Date: Sun, 20 Sep 2015 19:41:23 +0000
Subject: [PATCH] fix strftime to handle out of range tm fields without UB

strftime returns unspecifed result with out of range tm fields, but
it should not invoke undefined behaviour.

tm_wday, tm_yday, tm_mon and tm_year fields were used in signed int
arithmetics that could overflow.
---
 src/time/strftime.c | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

diff --git a/src/time/strftime.c b/src/time/strftime.c
index e945bb7..a1db37c 100644
--- a/src/time/strftime.c
+++ b/src/time/strftime.c
@@ -21,24 +21,24 @@ static int is_leap(int y)
 
 static int week_num(const struct tm *tm)
 {
-	int val = (tm->tm_yday + 7 - (tm->tm_wday+6)%7) / 7;
+	int val = (tm->tm_yday + 7U - (tm->tm_wday+6U)%7) / 7;
 	/* If 1 Jan is just 1-3 days past Monday,
 	 * the previous week is also in this year. */
-	if ((tm->tm_wday - tm->tm_yday - 2 + 371) % 7 <= 2)
+	if ((0U + tm->tm_wday - tm->tm_yday - 2 + 371) % 7 <= 2)
 		val++;
 	if (!val) {
 		val = 52;
 		/* If 31 December of prev year a Thursday,
 		 * or Friday of a leap year, then the
 		 * prev year has 53 weeks. */
-		int dec31 = (tm->tm_wday - tm->tm_yday - 1 + 7) % 7;
+		int dec31 = (0U + tm->tm_wday - tm->tm_yday - 1 + 7) % 7;
 		if (dec31 == 4 || (dec31 == 5 && is_leap(tm->tm_year%400-1)))
 			val++;
 	} else if (val == 53) {
 		/* If 1 January is not a Thursday, and not
 		 * a Wednesday of a leap year, then this
 		 * year has only 52 weeks. */
-		int jan1 = (tm->tm_wday - tm->tm_yday + 371) % 7;
+		int jan1 = (0U + tm->tm_wday - tm->tm_yday + 371) % 7;
 		if (jan1 != 4 && (jan1 != 3 || !is_leap(tm->tm_year)))
 			val = 1;
 	}
@@ -57,17 +57,17 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 
 	switch (f) {
 	case 'a':
-		item = ABDAY_1 + tm->tm_wday;
+		item = ABDAY_1 + tm->tm_wday%7U;
 		goto nl_strcat;
 	case 'A':
-		item = DAY_1 + tm->tm_wday;
+		item = DAY_1 + tm->tm_wday%7U;
 		goto nl_strcat;
 	case 'h':
 	case 'b':
-		item = ABMON_1 + tm->tm_mon;
+		item = ABMON_1 + tm->tm_mon%12U;
 		goto nl_strcat;
 	case 'B':
-		item = MON_1 + tm->tm_mon;
+		item = MON_1 + tm->tm_mon%12U;
 		goto nl_strcat;
 	case 'c':
 		item = D_T_FMT;
@@ -143,10 +143,10 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 		width = 1;
 		goto number;
 	case 'U':
-		val = (tm->tm_yday + 7 - tm->tm_wday) / 7;
+		val = (tm->tm_yday + 7U - tm->tm_wday) / 7;
 		goto number;
 	case 'W':
-		val = (tm->tm_yday + 7 - (tm->tm_wday+6)%7) / 7;
+		val = (tm->tm_yday + 7U - (tm->tm_wday+6U)%7) / 7;
 		goto number;
 	case 'V':
 		val = week_num(tm);
@@ -165,7 +165,7 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
 		val = tm->tm_year % 100;
 		goto number;
 	case 'Y':
-		val = tm->tm_year + 1900;
+		val = tm->tm_year + 1900LL;
 		if (val >= 10000) {
 			*l = snprintf(*s, sizeof *s, "+%lld", val);
 			return *s;
-- 
2.4.1

Re: out of range struct tm fields in strftime

[Rich Felker]

On Sun, Sep 20, 2015 at 09:54:50PM +0200, Szabolcs Nagy wrote:
> * Szabolcs Nagy <nsz@port70.net> [2015-09-20 18:44:35 +0200]:
> > * Rich Felker <dalias@libc.org> [2015-09-20 12:36:29 -0400]:
> > > On Sun, Sep 20, 2015 at 02:44:50PM +0200, Szabolcs Nagy wrote:
> > > > out of range tm fields should not be treated as ub
> > > > as noted in the thread
> > > > http://sourceware.org/ml/libc-alpha/2015-09/msg00546.html
> > > > 
> > > > i have a patch but there might be simpler approaches
> > > 
> > > Wouldn't it be less invasive to just make some small changes like
> > > putting a U on some of the constants so that the arithmetic happens as
> > > unsigned?
> > > 
> > 
> > that might be simpler (and probably generates better
> > code for div,mod by const)
> > 
> > but for the nl_langinfo item computation the range
> > has to be limited properly (tm_wday and tm_mon are
> > affected).
> 
> implemented this approach with wday%7U and using some 0U+

>  static int week_num(const struct tm *tm)
>  {
> -	int val = (tm->tm_yday + 7 - (tm->tm_wday+6)%7) / 7;
> +	int val = (tm->tm_yday + 7U - (tm->tm_wday+6U)%7) / 7;
>  	/* If 1 Jan is just 1-3 days past Monday,
>  	 * the previous week is also in this year. */
> -	if ((tm->tm_wday - tm->tm_yday - 2 + 371) % 7 <= 2)
> +	if ((0U + tm->tm_wday - tm->tm_yday - 2 + 371) % 7 <= 2)
>  		val++;
>  	if (!val) {
>  		val = 52;
>  		/* If 31 December of prev year a Thursday,
>  		 * or Friday of a leap year, then the
>  		 * prev year has 53 weeks. */
> -		int dec31 = (tm->tm_wday - tm->tm_yday - 1 + 7) % 7;
> +		int dec31 = (0U + tm->tm_wday - tm->tm_yday - 1 + 7) % 7;

This is okay but it might (or might not) be less ugly to just reorder
the constants to avoid the 0U. Thoughts?

> @@ -57,17 +57,17 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
>  
>  	switch (f) {
>  	case 'a':
> -		item = ABDAY_1 + tm->tm_wday;
> +		item = ABDAY_1 + tm->tm_wday%7U;

This is going to be a significant code size increase on many archs,
and possibly significant performance cost too. I wonder if there's a
better way.

> @@ -143,10 +143,10 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
>  		width = 1;
>  		goto number;
>  	case 'U':
> -		val = (tm->tm_yday + 7 - tm->tm_wday) / 7;
> +		val = (tm->tm_yday + 7U - tm->tm_wday) / 7;

These look fine.
> @@ -165,7 +165,7 @@ const char *__strftime_fmt_1(char (*s)[100], size_t *l, int f, const struct tm *
>  		val = tm->tm_year % 100;
>  		goto number;
>  	case 'Y':
> -		val = tm->tm_year + 1900;
> +		val = tm->tm_year + 1900LL;

Also looks ok. Hopefully the compiler does not actually evaluate in
64-bit.

Rich