On Wed, 12 Nov 2014, Mike Frysinger wrote:
On 12 Nov 2014 15:02, Sami Kerola wrote:
On 12 November 2014 14:25, Karel Zak <kzak@xxxxxxxxxx> wrote:
On Tue, Nov 11, 2014 at 10:49:02PM +0000, Sami Kerola wrote:
printf(_("Setting up swapspace version %d, size = %s (%ju bytes)\n"),
version, strsz, sz);
+ free(strsz);
free-before-exit, you know what I think about it, right? :)
Normally I would agree, but this case is bizarre. The mkswap will
return 1 when compiled with clang -fsanitize=address without any
message what is wrong. That's probably clang bug, which is a different
topic all together. Adding the free() calls seems to make the
immediate tests failure issue to go away, that is needed to automate
sanitize. In case travis && address sanitation is not needed at all
then these changes does not make sense.
i don't think adding a free to avoid a bug in clang is justified
at least gcc defines __SANITIZE_ADDRESS__ ... so you might be able to gate on
that with a comment pointing to an open bug report on the clang side
Hi Mike,
I like the idea, let me give a try to it.
$ git diff
diff --git a/disk-utils/mkswap.c b/disk-utils/mkswap.c
index d34e2fd..895d1ec 100644
--- a/disk-utils/mkswap.c
+++ b/disk-utils/mkswap.c
@@ -347,7 +347,7 @@ static void write_header_to_device(struct mkswap_control *ctl)
ctl->devname);
}
-int main(int argc, char **argv)
+int UL_ASAN_BLACKLIST main(int argc, char **argv)
{
struct mkswap_control ctl = { .fd = -1 };
int c;
@@ -489,7 +489,6 @@ int main(int argc, char **argv)
printf(_("Setting up swapspace version %d, size = %s (%ju bytes)\n"),
version, strsz, sz);
- free(strsz);
set_signature(&ctl);
set_uuid_and_label(&ctl);
diff --git a/include/c.h b/include/c.h
index 0f6e5b2..c581d73 100644
--- a/include/c.h
+++ b/include/c.h
@@ -313,4 +313,21 @@ static inline int xusleep(useconds_t usec)
#define stringify_value(s) stringify(s)
#define stringify(s) #s
+/*
+ * UL_ASAN_BLACKLIST is a macro to tell AddressSanitizer (a compile-time
+ * instrumentation shipped with Clang and GCC) to not instrument the
+ * annotated function. Furthermore, it will prevent the compiler from
+ * inlining the function because inlining currently breaks the
+ * blacklisting mechanism of AddressSanitizer.
+ */
+#if defined(__has_feature)
+# if __has_feature(address_sanitizer)
+# define UL_ASAN_BLACKLIST __attribute__((no_sanitize_address))
+# else
+# define UL_ASAN_BLACKLIST /* nothing */
+# endif
+#else
+# define UL_ASAN_BLACKLIST /* nothing */
+#endif
+
#endif /* UTIL_LINUX_C_H */
I borrowed the above from mozilla source.
http://dxr.mozilla.org/mozilla-central/source/mfbt/Attributes.h
There is just one catch. After compiling with no_sanitize_address
instrumentation the return value issue is back.
$ ./mkswap test ; echo $?
lt-mkswap: test: warning: wiping old swap signature.
Setting up swapspace version 1, size = 96 KiB (98304 bytes)
no label, UUID=1aeb88e8-6442-4e0f-93a5-118271b026c0
1 <--- is the issue
--
Sami Kerola
http://www.iki.fi/kerolasa/
--
To unsubscribe from this list: send the line "unsubscribe util-linux" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
