Skip to contentSkip to conversation

string.h: Add strncmp_prefix() helper macro

RFC PATCH In review Bryana Rostedt 3 participants 6 messages Dec 19, 2018 – Dec 20, 2018 Generic string library

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 string.h: Add strncmp_prefix() helper macro

A discussion came up in the trace triggers thread about converting a bunch of:

strncmp(str, "const", sizeof("const") - 1)

use cases into a helper macro. It started with:

#define strncmp_const(str, const) \
strncmp(str, const, sizeof(const) - 1)

But then Joe Perches mentioned that if a const is not used, the sizeof() will be the size of a pointer, which can be bad. And that gcc will optimize strlen("const") into "sizeof("const") - 1".

Thinking about this more, a quick grep in the kernel tree found several (thousands!) of cases that use this construct. A quick grep also revealed that there's probably several bugs in that use case. Some are that people forgot the "- 1" (which I found) and others could be that the constant for the sizeof is different than the constant (although, I haven't found any of those, but I also didn't look hard).

I figured the best thing to do is to create a helper macro and place it into include/linux/string.h. And go around and fix all the open coded versions of it later.

I plan on only applying this patch and updating the tracing hooks for this merge window. And perhaps use it to fix some of the bugs that were found.

I was going to just use:

#define strncmp_prefix(str, prefix) \
strncmp(str, prefix, strlen(prefix))

but then realized that "prefix" is used twice, and will break if someone does something like:

	strncmp_prefix(str, ptr++);

So instead I check with __builtin_constant_p() to see if the second parameter is truly a constant, which I use sizeof() anyway (why bother gcc to optimize it, if we already know it's a constant), and copy the parameter into a local variable and use that local variable for the non constant part (with strlen).

---

include/linux/string.h +21 -0
index 27d0482e5e05..3dc743e3a0ba 100644
@@ -14,6 +14,27 @@ extern void *memdup_user(const void __user *, size_t);
 extern void *vmemdup_user(const void __user *, size_t);
 extern void *memdup_user_nul(const void __user *, size_t);
 
+/*
+ * A common way to test a prefix of a string is to do:
+ *  strncmp(str, prefix, sizeof(prefix) - 1)
+ *
+ * But this can lead to bugs due to typos, or if prefix is a pointer
+ * and not a constant. Instead use strncmp_prefix().
+ */
+#define strncmp_prefix(str, prefix)					\
+	({								\
+		int ____strcmp_prefix_ret____;				\
+		char *____strcmp_prefix____ = prefix;			\
+		if (__builtin_constant_p(&prefix))			\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, prefix, sizeof(prefix) - 1); \
+		else							\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, ____strcmp_prefix____,	\
+					strlen(____strcmp_prefix____));	\
+		____strcmp_prefix_ret____;				\
+	})
+
 /*
  * Include machine specific inline routines
  */
#1 lore raw
Quoted 58 lines · Bryana Rostedt

A discussion came up in the trace triggers thread about converting a bunch of:

strncmp(str, "const", sizeof("const") - 1)

use cases into a helper macro. It started with:

#define strncmp_const(str, const) \
strncmp(str, const, sizeof(const) - 1)

But then Joe Perches mentioned that if a const is not used, the sizeof() will be the size of a pointer, which can be bad. And that gcc will optimize strlen("const") into "sizeof("const") - 1".

Thinking about this more, a quick grep in the kernel tree found several (thousands!) of cases that use this construct. A quick grep also revealed that there's probably several bugs in that use case. Some are that people forgot the "- 1" (which I found) and others could be that the constant for the sizeof is different than the constant (although, I haven't found any of those, but I also didn't look hard).

I figured the best thing to do is to create a helper macro and place it into include/linux/string.h. And go around and fix all the open coded versions of it later.

I plan on only applying this patch and updating the tracing hooks for this merge window. And perhaps use it to fix some of the bugs that were found.

I was going to just use:

#define strncmp_prefix(str, prefix) \
strncmp(str, prefix, strlen(prefix))

but then realized that "prefix" is used twice, and will break if someone does something like:

	strncmp_prefix(str, ptr++);

So instead I check with __builtin_constant_p() to see if the second parameter is truly a constant, which I use sizeof() anyway (why bother gcc to optimize it, if we already know it's a constant), and copy the parameter into a local variable and use that local variable for the non constant part (with strlen).

---

include/linux/string.h +1 -0
index 27d0482e5e05..3dc743e3a0ba 100644
@@ -14,6 +14,27 @@ extern void *memdup_user(const void __user *, size_t);
 extern void *vmemdup_user(const void __user *, size_t);
 extern void *memdup_user_nul(const void __user *, size_t);
 
+/*

include/linux/string.h extern void *memdup_user(const void __user *, size_t); · line 27

Quoted 10 lines
include/linux/string.h +10 -0
+ * A common way to test a prefix of a string is to do:
+ *  strncmp(str, prefix, sizeof(prefix) - 1)
+ *
+ * But this can lead to bugs due to typos, or if prefix is a pointer
+ * and not a constant. Instead use strncmp_prefix().
+ */
+#define strncmp_prefix(str, prefix)					\e
+	({								\
+		int ____strcmp_prefix_ret____;				\
+		char *____strcmp_prefix____ = prefix;			\

This creates a warning and discards any const from prefix when used like

	static const char foo[] = "bar";

	strncmp_prefix(str, foo);

include/linux/string.h extern void *memdup_user(const void __user *, size_t); · line 37

Quoted 10 lines
include/linux/string.h +10 -0
+		if (__builtin_constant_p(&prefix))			\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, prefix, sizeof(prefix) - 1); \
+		else							\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, ____strcmp_prefix____,	\
+					strlen(____strcmp_prefix____));	\
+		____strcmp_prefix_ret____;				\
+	})
+

Perhaps

#define strncmp_prefix(str, prefix) \

({									\
	typeof(prefix) p = (prefix);					\
	strncmp(str, p, strlen(p));					\
})
#2 lore raw

Seems that I sent this as my daughter.

I'm not home now but will resend again as myself when I am.

-- Steve

Quoted 104 lines · Joe Perches
Quoted 7 lines · Bryana Rostedt

A discussion came up in the trace triggers thread about converting a bunch of:

strncmp(str, "const", sizeof("const") - 1)

use cases into a helper macro. It started with:

Quoted 8 lines

#define strncmp_const(str, const) \
strncmp(str, const, sizeof(const) - 1)

But then Joe Perches mentioned that if a const is not used, the sizeof() will be the size of a pointer, which can be bad. And that gcc will optimize strlen("const") into "sizeof("const") - 1".

Thinking about this more, a quick grep in the kernel tree found

several

Quoted 2 lines

(thousands!) of cases that use this construct. A quick grep also revealed that there's probably several bugs in that use case. Some

are

Quoted 2 lines

that people forgot the "- 1" (which I found) and others could be that the constant for the sizeof is different than the constant (although,

I

Quoted 3 lines

haven't found any of those, but I also didn't look hard).

I figured the best thing to do is to create a helper macro and place

it

Quoted 5 lines

into include/linux/string.h. And go around and fix all the open coded versions of it later.

I plan on only applying this patch and updating the tracing hooks for this merge window. And perhaps use it to fix some of the bugs that

were

Quoted 7 lines

found.

I was going to just use:

#define strncmp_prefix(str, prefix) \
strncmp(str, prefix, strlen(prefix))

Quoted 7 lines

but then realized that "prefix" is used twice, and will break if someone does something like:

	strncmp_prefix(str, ptr++);

So instead I check with __builtin_constant_p() to see if the second parameter is truly a constant, which I use sizeof() anyway (why

bother

Quoted 2 lines

gcc to optimize it, if we already know it's a constant), and copy the parameter into a local variable and use that local variable for the

non

Quoted 3 lines

constant part (with strlen).

Link:

http://lkml.kernel.org/r/e3e754f2bd18e56eaa8baf79bee619316ebf4cfc.1545161087.git.tom.zanussi@linux.intel.com

include/linux/string.h extern void *memdup_user(const void __user *,

Quoted 8 lines

---

include/linux/string.h +0 -0
index 27d0482e5e05..3dc743e3a0ba 100644
@@ -14,6 +14,27 @@ extern void *memdup_user(const void __user *,

size_t);

Quoted 4 lines
include/linux/string.h +1 -0
 extern void *vmemdup_user(const void __user *, size_t);
 extern void *memdup_user_nul(const void __user *, size_t);
 
+/*

include/linux/string.h extern void *memdup_user(const void __user *, size_t); · line 27

Quoted 10 lines
include/linux/string.h +10 -0
+ * A common way to test a prefix of a string is to do:
+ *  strncmp(str, prefix, sizeof(prefix) - 1)
+ *
+ * But this can lead to bugs due to typos, or if prefix is a pointer
+ * and not a constant. Instead use strncmp_prefix().
+ */
+#define strncmp_prefix(str, prefix)					\e
+	({								\
+		int ____strcmp_prefix_ret____;				\
+		char *____strcmp_prefix____ = prefix;			\

This creates a warning and discards any const from prefix when used like

	static const char foo[] = "bar";

	strncmp_prefix(str, foo);

include/linux/string.h extern void *memdup_user(const void __user *, size_t); · line 37

Quoted 10 lines
include/linux/string.h +10 -0
+		if (__builtin_constant_p(&prefix))			\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, prefix, sizeof(prefix) - 1); \
+		else							\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, ____strcmp_prefix____,	\
+					strlen(____strcmp_prefix____));	\
+		____strcmp_prefix_ret____;				\
+	})
+

Perhaps

#define strncmp_prefix(str, prefix) \

({									\
	typeof(prefix) p = (prefix);					\
	strncmp(str, p, strlen(p));					\
})
Signature · 1 line
Sent from my Android device with K-9 Mail. Please excuse my brevity and top posting.
#3 lore raw
In reply to Joe Perches

[ Sending now as myself and not my daughter :-p ]

Quoted 33 lines · Joe Perches

include/linux/string.h extern void *memdup_user(const void __user *, size_t); · line 27

Quoted 12 lines · Bryana Rostedt
include/linux/string.h +11 -0
+/*
+ * A common way to test a prefix of a string is to do:
+ *  strncmp(str, prefix, sizeof(prefix) - 1)
+ *
+ * But this can lead to bugs due to typos, or if prefix is a pointer
+ * and not a constant. Instead use strncmp_prefix().
+ */
+#define strncmp_prefix(str, prefix)					\e
+	({								\
+		int ____strcmp_prefix_ret____;				\
+		char *____strcmp_prefix____ = prefix;			\

This creates a warning and discards any const from prefix when used like

	static const char foo[] = "bar";

	strncmp_prefix(str, foo);

include/linux/string.h extern void *memdup_user(const void __user *, size_t); · line 37

Quoted 10 lines
include/linux/string.h +10 -0
+		if (__builtin_constant_p(&prefix))			\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, prefix, sizeof(prefix) - 1); \
+		else							\
+			____strcmp_prefix_ret____ =			\
+				strncmp(str, ____strcmp_prefix____,	\
+					strlen(____strcmp_prefix____));	\
+		____strcmp_prefix_ret____;				\
+	})
+

Perhaps

Quoted 4 lines

#define strncmp_prefix(str, prefix) \

({									\
	typeof(prefix) p = (prefix);					\
	strncmp(str, p, strlen(p));					\

I'll have to see how all supported gcc optimizes that.

I could just move the prefix assignment in the else part, which gets compiled out on constants anyway, just to be safe, and also shows exactly what is happening and not worrying about gcc doing the "right" thing.

-- Steve

Quoted 2 lines

})

#4 lore raw
In reply to Bryana Rostedt
Quoted 6 lines · Bryana Rostedt

a quick grep in the kernel tree found several
(thousands!) of cases that use this construct. A quick grep also revealed that there's probably several bugs in that use case. Some are that people forgot the "- 1" (which I found) and others could be that the constant for the sizeof is different than the constant (although, I haven't found any of those, but I also didn't look hard).

Several dozen cases exist like strncmp(foo, "bar", 4) where it might as well use strcmp instead.

And a few mismatches exist like:

drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c:2285: if (!strncmp(opt, "eee_timer:", 6)) {
tools/power/x86/x86_energy_perf_policy/x86_energy_perf_policy.c:318: if (!strncmp(s, "power", 9))
tools/power/x86/x86_energy_perf_policy/x86_energy_perf_policy.c:321: if (!strncmp(s, "balance-power", 17))
j

#5 lore raw

[ Another email not as my daughter. ]

Quoted 11 lines · Joe Perches
Quoted 6 lines · Bryana Rostedt

a quick grep in the kernel tree found several
(thousands!) of cases that use this construct. A quick grep also revealed that there's probably several bugs in that use case. Some are that people forgot the "- 1" (which I found) and others could be that the constant for the sizeof is different than the constant (although, I haven't found any of those, but I also didn't look hard).

Several dozen cases exist like strncmp(foo, "bar", 4) where it might as well use strcmp instead.

Quoted 6 lines

And a few mismatches exist like:

drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c:2285: if (!strncmp(opt, "eee_timer:", 6)) {
tools/power/x86/x86_energy_perf_policy/x86_energy_perf_policy.c:318: if (!strncmp(s, "power", 9))
tools/power/x86/x86_energy_perf_policy/x86_energy_perf_policy.c:321: if (!strncmp(s, "balance-power", 17))
j

Yeah I saw these too. But that's just more justification to get a helper function applied.

I'll look at the options, and then just add it to my queue for the next merge window. I'll have to add the commit that is added as a requirement of other patches that will sure need to go to stable.

-- Steve