Skip to contentSkip to conversation

kernel/module.c: for looping, need use 'goto' instead of 'break' to jump out in time

PATCH In review Chen Gang 2 participants 11 messages May 7, 2013 – May 23, 2013

lore
Share & tools
lore

Chen Gang

author
#0 lore raw

PATCH 1/1 kernel/module.c: for looping, need use 'goto' instead of 'break' to jump out in time

In the 'for' looping, when error occurs, the 'break' only jump out of 'switch', and still in 'for' looping. If error occurs multiple times, the original error value will be overwrite.

Currently, that will not cause issue, but still better to improve it, so that let it return the first real error code in time.

The related commit: "1da177e Linux-2.6.12-rc"

---
 kernel/module.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)
kernel/module.c +3 -2
index b049939..7e012ff 100644
@@ -1977,7 +1977,7 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			printk("%s: please compile with -fno-common\n",
 			       mod->name);
 			ret = -ENOEXEC;
-			break;
+			goto tail;
 
 		case SHN_ABS:
 			/* Don't need to do anything */
@@ -2000,7 +2000,7 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
 			       mod->name, name, PTR_ERR(ksym));
 			ret = PTR_ERR(ksym) ?: -ENOENT;
-			break;
+			goto tail;
 
 		default:
 			/* Divert to percpu allocation if a percpu var. */
@@ -2013,6 +2013,7 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 		}
 	}
 
+tail:
 	return ret;
 }
 
Signature · 1 line
1.7.7.6
#1 lore raw
Quoted 6 lines · Chen Gang

In the 'for' looping, when error occurs, the 'break' only jump out of 'switch', and still in 'for' looping. If error occurs multiple times, the original error value will be overwrite.

Currently, that will not cause issue, but still better to improve it, so that let it return the first real error code in time.

We choose to print all the problems, rather than just one. I don't really mind though.

It we want this patch, it would be neater to just 'return -ENOEXEC' and 'return PTR_ERR(ksym) ?: -ENOENT'.

Cheers,
Rusty.

Quoted 45 lines

The related commit: "1da177e Linux-2.6.12-rc"

---
 kernel/module.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)
kernel/module.c +3 -2
index b049939..7e012ff 100644
@@ -1977,7 +1977,7 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			printk("%s: please compile with -fno-common\n",
 			       mod->name);
 			ret = -ENOEXEC;
-			break;
+			goto tail;
 
 		case SHN_ABS:
 			/* Don't need to do anything */
@@ -2000,7 +2000,7 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
 			       mod->name, name, PTR_ERR(ksym));
 			ret = PTR_ERR(ksym) ?: -ENOENT;
-			break;
+			goto tail;
 
 		default:
 			/* Divert to percpu allocation if a percpu var. */
@@ -2013,6 +2013,7 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 		}
 	}
 
+tail:
 	return ret;
 }
 
Signature · 6 lines
1.7.7.6
--
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/
author
#2 lore raw
Quoted 10 lines · Rusty Russell
Quoted 6 lines · Chen Gang
Quoted 6 lines

In the 'for' looping, when error occurs, the 'break' only jump out of 'switch', and still in 'for' looping. If error occurs multiple times, the original error value will be overwrite.

Currently, that will not cause issue, but still better to improve it, so that let it return the first real error code in time.

We choose to print all the problems, rather than just one. I don't really mind though.

It sounds good: "choose to print all the problems, rather than just one"

If so, it seems enough to only return a bool value to known whether success or fail, do not need the error details any more (since they are already been printed)

Quoted 2 lines

It we want this patch, it would be neater to just 'return -ENOEXEC' and 'return PTR_ERR(ksym) ?: -ENOENT'.

If we really want this patch (still only print the first error, and return the real error value), I should send patch v2 (also 'ret' is obsoleted)

Thanks.

Signature · 3 lines
Chen Gang

Asianux Corporation
#3 lore raw
Quoted 18 lines · Chen Gang
Quoted 10 lines · Rusty Russell
Quoted 6 lines · Chen Gang
Quoted 6 lines

In the 'for' looping, when error occurs, the 'break' only jump out of 'switch', and still in 'for' looping. If error occurs multiple times, the original error value will be overwrite.

Currently, that will not cause issue, but still better to improve it, so that let it return the first real error code in time.

We choose to print all the problems, rather than just one. I don't really mind though.

It sounds good: "choose to print all the problems, rather than just one"

If so, it seems enough to only return a bool value to known whether success or fail, do not need the error details any more (since they are already been printed)

Quoted 6 lines
Quoted 2 lines

It we want this patch, it would be neater to just 'return -ENOEXEC' and 'return PTR_ERR(ksym) ?: -ENOENT'.

If we really want this patch (still only print the first error, and return the real error value), I should send patch v2 (also 'ret' is obsoleted)

I would take such a patch, since it makes things a little neater. But if you don't write it, I wouldn't write it myself, since it's borderline.

Thanks,
Rusty.

author
#4 lore raw
Quoted 22 lines · Rusty Russell
Quoted 21 lines · Chen Gang
Quoted 21 lines · Rusty Russell
Quoted 10 lines
Quoted 10 lines
Quoted 6 lines · Chen Gang
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

In the 'for' looping, when error occurs, the 'break' only jump out of 'switch', and still in 'for' looping. If error occurs multiple times, the original error value will be overwrite.

Currently, that will not cause issue, but still better to improve it, so that let it return the first real error code in time.

We choose to print all the problems, rather than just one. I don't really mind though.

It sounds good: "choose to print all the problems, rather than just one"

If so, it seems enough to only return a bool value to known whether success or fail, do not need the error details any more (since they are already been printed)

Quoted 2 lines
Quoted 2 lines

It we want this patch, it would be neater to just 'return -ENOEXEC' and 'return PTR_ERR(ksym) ?: -ENOENT'.

Quoted 6 lines
Quoted 3 lines
Quoted 3 lines

If we really want this patch (still only print the first error, and return the real error value), I should send patch v2 (also 'ret' is obsoleted)

I would take such a patch, since it makes things a little neater. But if you don't write it, I wouldn't write it myself, since it's borderline.

OK, thank, I should write it, and now I will "choose to print all the problems, and return a bool value", and it will be as a cleanup patch.

I will finish it within 2 days (2013-05-15).

If have additional suggestions or completions, please reply.

Thanks.

Signature · 3 lines
Chen Gang

Asianux Corporation
author
#5 lore raw

PATCH v2 1/1 kernel/module.c: cleanup patch for looping, let return 'bool' value instead of real error number.

For looping in simplify_symbols(), when multiple errors occur, the original error number will be override (the original related commit "1da177e Linux-2.6.12-rc").

We focus on printing all errors as much as possible, and not focus on the return error number (only know whether succeed or not is enough),

So when error occurs during looping, it is correct to only mark "need return failure" and continue looping.

Since simplify_symbols() is a static function (not an API to outside), it is better to change the function return type to 'bool', so that can skip processing error number details incorrectly.

---
 kernel/module.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
kernel/module.c +12 -7
index b049939..445a960 100644
@@ -1956,14 +1956,18 @@ static int verify_export_symbols(struct module *mod)
 	return 0;
 }
 
-/* Change all symbols so that st_value encodes the pointer directly. */
-static int simplify_symbols(struct module *mod, const struct load_info *info)
+/*
+ * Change all symbols so that st_value encodes the pointer directly.
+ *
+ * Return true if succeed, or false if at least 1 failure occurs.
+ */
+static bool simplify_symbols(struct module *mod, const struct load_info *info)
 {
 	Elf_Shdr *symsec = &info->sechdrs[info->index.sym];
 	Elf_Sym *sym = (void *)symsec->sh_addr;
 	unsigned long secbase;
 	unsigned int i;
-	int ret = 0;
+	bool ret = true;
 	const struct kernel_symbol *ksym;
 
 	for (i = 1; i < symsec->sh_size / sizeof(Elf_Sym); i++) {
@@ -1976,7 +1980,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			pr_debug("Common symbol: %s\n", name);
 			printk("%s: please compile with -fno-common\n",
 			       mod->name);
-			ret = -ENOEXEC;
+			/* Mark return result and continue looping */
+			ret = false;
 			break;
 
 		case SHN_ABS:
@@ -1999,7 +2004,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 
 			printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
 			       mod->name, name, PTR_ERR(ksym));
-			ret = PTR_ERR(ksym) ?: -ENOENT;
+			/* Mark return result and continue looping */
+			ret = false;
 			break;
 
 		default:
@@ -3267,8 +3273,7 @@ static int load_module(struct load_info *info, const char __user *uargs,
 	setup_modinfo(mod, info);
 
 	/* Fix up syms, so that st_value is a pointer to location. */
-	err = simplify_symbols(mod, info);
-	if (err < 0)
+	if (!simplify_symbols(mod, info))
 		goto free_modinfo;
 
 	err = apply_relocations(mod, info);
Signature · 1 line
1.7.11.7
author
#6 lore raw

Re: [PATCH v2] kernel/module.c: cleanup patch for looping, let return 'bool' value instead of real error number.

Oh, sorry, my fix is incorrect !!

The return value is used by caller of caller ... it seems the user mode can sense it !! So we still should return the correct error code to outside when error occurs.

I think we can reference 'using compiler': it prints all errors and warnings as far as it can, but the user usually mainly focus on the first error or warning.

So I think the suitable fix is to still print all problems, but should return the first error code (not return the last error code).

Please help check, thanks.

Quoted 78 lines · Chen Gang

For looping in simplify_symbols(), when multiple errors occur, the original error number will be override (the original related commit "1da177e Linux-2.6.12-rc").

We focus on printing all errors as much as possible, and not focus on the return error number (only know whether succeed or not is enough),

So when error occurs during looping, it is correct to only mark "need return failure" and continue looping.

Since simplify_symbols() is a static function (not an API to outside), it is better to change the function return type to 'bool', so that can skip processing error number details incorrectly.

---
 kernel/module.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
kernel/module.c +12 -7
index b049939..445a960 100644
@@ -1956,14 +1956,18 @@ static int verify_export_symbols(struct module *mod)
 	return 0;
 }
 
-/* Change all symbols so that st_value encodes the pointer directly. */
-static int simplify_symbols(struct module *mod, const struct load_info *info)
+/*
+ * Change all symbols so that st_value encodes the pointer directly.
+ *
+ * Return true if succeed, or false if at least 1 failure occurs.
+ */
+static bool simplify_symbols(struct module *mod, const struct load_info *info)
 {
 	Elf_Shdr *symsec = &info->sechdrs[info->index.sym];
 	Elf_Sym *sym = (void *)symsec->sh_addr;
 	unsigned long secbase;
 	unsigned int i;
-	int ret = 0;
+	bool ret = true;
 	const struct kernel_symbol *ksym;
 
 	for (i = 1; i < symsec->sh_size / sizeof(Elf_Sym); i++) {
@@ -1976,7 +1980,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			pr_debug("Common symbol: %s\n", name);
 			printk("%s: please compile with -fno-common\n",
 			       mod->name);
-			ret = -ENOEXEC;
+			/* Mark return result and continue looping */
+			ret = false;
 			break;
 
 		case SHN_ABS:
@@ -1999,7 +2004,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 
 			printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
 			       mod->name, name, PTR_ERR(ksym));
-			ret = PTR_ERR(ksym) ?: -ENOENT;
+			/* Mark return result and continue looping */
+			ret = false;
 			break;
 
 		default:
@@ -3267,8 +3273,7 @@ static int load_module(struct load_info *info, const char __user *uargs,
 	setup_modinfo(mod, info);
 
 	/* Fix up syms, so that st_value is a pointer to location. */
-	err = simplify_symbols(mod, info);
-	if (err < 0)
+	if (!simplify_symbols(mod, info))
 		goto free_modinfo;
 
 	err = apply_relocations(mod, info);
Signature · 3 lines
Chen Gang

Asianux Corporation
author
#7 lore raw

PATCH v3 1/1 kernel/module.c: need return the first error code to upper caller when error occurs

When multiple errors occur, simplify_symbols() will return the last error code to the upper caller.

In this case, better to return the first error code to the upper caller.

Just like "using compiler": it will print all errors and warnings as much as it can, but the user usually mainly focus on the first error or warning. Since 'user' only can get one error return code, 'he/she' usually assume the error code will match the first error print line.

---
 kernel/module.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)
kernel/module.c +4 -2
index e4ee1bf..9e6c96d 100644
@@ -1976,7 +1976,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			pr_debug("Common symbol: %s\n", name);
 			printk("%s: please compile with -fno-common\n",
 			       mod->name);
-			ret = -ENOEXEC;
+			if (!ret)
+				ret = -ENOEXEC;
 			break;
 
 		case SHN_ABS:
@@ -1999,7 +2000,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 
 			printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
 			       mod->name, name, PTR_ERR(ksym));
-			ret = PTR_ERR(ksym) ?: -ENOENT;
+			if (!ret)
+				ret = PTR_ERR(ksym) ?: -ENOENT;
 			break;
 
 		default:
Signature · 1 line
1.7.7.6
author
#8 lore raw

Re: [PATCH v3] kernel/module.c: need return the first error code to upper caller when error occurs

Hello Maintainers:

Please help check the patch whether OK or not, when you have time.

Thanks.

Quoted 42 lines · Chen Gang

When multiple errors occur, simplify_symbols() will return the last error code to the upper caller.

In this case, better to return the first error code to the upper caller.

Just like "using compiler": it will print all errors and warnings as much as it can, but the user usually mainly focus on the first error or warning. Since 'user' only can get one error return code, 'he/she' usually assume the error code will match the first error print line.

---
 kernel/module.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)
kernel/module.c +4 -2
index e4ee1bf..9e6c96d 100644
@@ -1976,7 +1976,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			pr_debug("Common symbol: %s\n", name);
 			printk("%s: please compile with -fno-common\n",
 			       mod->name);
-			ret = -ENOEXEC;
+			if (!ret)
+				ret = -ENOEXEC;
 			break;
 
 		case SHN_ABS:
@@ -1999,7 +2000,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 
 			printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
 			       mod->name, name, PTR_ERR(ksym));
-			ret = PTR_ERR(ksym) ?: -ENOENT;
+			if (!ret)
+				ret = PTR_ERR(ksym) ?: -ENOENT;
 			break;
 
 		default:
Signature · 3 lines
Chen Gang

Asianux Corporation
#9 lore raw

Re: [PATCH v3] kernel/module.c: need return the first error code to upper caller when error occurs

Quoted 3 lines · Chen Gang

Hello Maintainers:

Please help check the patch whether OK or not, when you have time.

Hi Chen,

There's nothing *wrong* with the patch, but I don't think it's
worthwhile. If this were your very first kernel patch, I'd probably apply it just to encourage you, but you're not a newbie any more!

Cheers,
Rusty.

Quoted 49 lines · Chen Gang
Quoted 42 lines

When multiple errors occur, simplify_symbols() will return the last error code to the upper caller.

In this case, better to return the first error code to the upper caller.

Just like "using compiler": it will print all errors and warnings as much as it can, but the user usually mainly focus on the first error or warning. Since 'user' only can get one error return code, 'he/she' usually assume the error code will match the first error print line.

---
 kernel/module.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)
kernel/module.c +4 -2
index e4ee1bf..9e6c96d 100644
@@ -1976,7 +1976,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 			pr_debug("Common symbol: %s\n", name);
 			printk("%s: please compile with -fno-common\n",
 			       mod->name);
-			ret = -ENOEXEC;
+			if (!ret)
+				ret = -ENOEXEC;
 			break;
 
 		case SHN_ABS:
@@ -1999,7 +2000,8 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
 
 			printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
 			       mod->name, name, PTR_ERR(ksym));
-			ret = PTR_ERR(ksym) ?: -ENOENT;
+			if (!ret)
+				ret = PTR_ERR(ksym) ?: -ENOENT;
 			break;
 
 		default:
Signature · 3 lines
Chen Gang

Asianux Corporation
author
#10 lore raw

Re: [PATCH v3] kernel/module.c: need return the first error code to upper caller when error occurs

Quoted 9 lines · Rusty Russell
Quoted 3 lines · Chen Gang
Quoted 3 lines

Hello Maintainers:

Please help check the patch whether OK or not, when you have time.

Hi Chen,

There's nothing *wrong* with the patch, but I don't think it's
worthwhile. If this were your very first kernel patch, I'd probably apply it just to encourage you, but you're not a newbie any more!

OK, I can understand.

Thanks.

Signature · 3 lines
Chen Gang

Asianux Corporation