Skip to contentSkip to conversation

ERR_PTR: if errno value is known at compile time, make sure it's valid

PATCH New Marcin Slusarz 2 participants 4 messages May 22, 2008 – May 27, 2008

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ERR_PTR: if errno value is known at compile time, make sure it's valid

ERR_PTR is easy to call with wrong argument (positive errno), and this error lead to catastrophic event - oops or kernel panic (dereference of invalid pointer).

As most of error handling code paths are rarely tested, this kind of bug can be hidden for years. Currently there are > 1400 calls of ERR_PTR with constant argument.

Cc Andrew Morton <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Al Viro <[email protected]>
Cc Alexey Dobriyan <[email protected]>
Cc Johannes Weiner <[email protected]>
---
 include/linux/err.h |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)
include/linux/err.h +10 -1
index ec87f31..4773ed3 100644
@@ -18,12 +18,21 @@
 #ifndef __ASSEMBLY__
 
 #define IS_ERR_VALUE(x) unlikely((x) >= (unsigned long)-MAX_ERRNO)
+#define VALID_ERR_PTR_ARG(error) (error == 0 || IS_ERR_VALUE(error))
 
-static inline void *ERR_PTR(long error)
+static inline void *__ERR_PTR(long error)
 {
 	return (void *) error;
 }
 
+/*
+ * implementation note: we have to make it a macro, otherwise
+ * gcc won't break the build on wrong argument
+ */
+#define ERR_PTR(error) (BUILD_BUG_ON(__builtin_constant_p(error) && \
+					!VALID_ERR_PTR_ARG(error)), \
+			__ERR_PTR(error))
+
 static inline long PTR_ERR(const void *ptr)
 {
 	return (long) ptr;
Signature · 1 line
1.5.4.5
#1 lore raw

PATCH 1/1 ERR_PTR: warn when ERR_PTR parameter is valid argument

Check at runtime whether error argument of ERR_PTR is valid. It can catch bugs which possibly lead to oops or panic earlier.

Currently there are > 600 calls of ERR_PTR with non-constant argument.

Cc Andrew Morton <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Al Viro <[email protected]>
Cc Alexey Dobriyan <[email protected]>
Cc Johannes Weiner <[email protected]>
---
 include/linux/err.h |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/err.h +2 -0
index 4773ed3..f7e098e 100644
@@ -3,6 +3,7 @@
 
 #include <linux/compiler.h>
 
+#include <asm/bug.h>
 #include <asm/errno.h>
 
 /*
@@ -22,6 +23,7 @@
 
 static inline void *__ERR_PTR(long error)
 {
+	WARN_ON(!VALID_ERR_PTR_ARG(error));
 	return (void *) error;
 }
 
Signature · 1 line
1.5.4.5
#2 lore raw

Re: [PATCH] ERR_PTR: warn when ERR_PTR parameter is invalid

Sorry, wrong patch title. Should be:
"ERR_PTR: warn when ERR_PTR parameter is invalid"

I forgot to add that this patch sould be applied after:
"vfs: fix ERR_PTR abuse in generic_readlink"
http://lkml.org/lkml/2008/5/18/316
which is alredy in -mm

Cheers,
Marcin

#3 lore raw

Re: [PATCH] ERR_PTR: warn when ERR_PTR parameter is valid argument

In reply to Marcin Slusarz
Quoted 28 lines · On Thu, 22 May 2008 18:50:19 +0200 Marcin Slusarz <[email protected]>…

Check at runtime whether error argument of ERR_PTR is valid. It can catch bugs which possibly lead to oops or panic earlier.

Currently there are > 600 calls of ERR_PTR with non-constant argument.

Cc Andrew Morton <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Al Viro <[email protected]>
Cc Alexey Dobriyan <[email protected]>
Cc Johannes Weiner <[email protected]>
---
 include/linux/err.h |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/err.h +1 -0
index 4773ed3..f7e098e 100644
@@ -3,6 +3,7 @@
 
 #include <linux/compiler.h>
 
+#include <asm/bug.h>
 #include <asm/errno.h>
 
 /*

include/linux/err.h line 28

Quoted 6 lines
include/linux/err.h +1 -0
@@ -22,6 +23,7 @@
 
 static inline void *__ERR_PTR(long error)
 {
+	WARN_ON(!VALID_ERR_PTR_ARG(error));
 	return (void *) error;
 }

It would be regrettable to add source-level complexity and runtime cost to detect this particular bug. I think it would be better to do this via static source-code checking if at all possible.

Is there _any_ legitimate use of non-negative EFOO? There might be some baroque bits of code which are using non-negative constants in a non-buggy fashion, but I bet they could be reworked to use negative constants.

In which case I'd have thought that a script which

a) extracts all the EFOO identifiers from include/*/errno.h and

b) greps the tree for non-negative uses of those

would have 100% coverage?

We might need to touch up some code sites to avoid triggering false positives and make that script's life a bit easier, but that's fine.