|author||Denys Vlasenko <email@example.com>||2013-06-28 19:59:54 (GMT)|
|committer||Denys Vlasenko <firstname.lastname@example.org>||2013-06-28 19:59:54 (GMT)|
MIPS: set _NSIG to 128, not 129. This matches glibc.
Prompted by lkml discussion of a MIPS bug where sending signal 128 was found to be able to crash the machine :/ Signed-off-by: Denys Vlasenko <email@example.com>
1 files changed, 15 insertions, 2 deletions
diff --git a/libc/sysdeps/linux/mips/bits/signum.h b/libc/sysdeps/linux/mips/bits/signum.h
index cf9b834..e83250e 100644
@@ -53,7 +53,20 @@
#define SIGXCPU 30 /* CPU limit exceeded (4.2 BSD). */
#define SIGXFSZ 31 /* File size limit exceeded (4.2 BSD). */
-/* Biggest signal number + 1 (including real-time signals). */
-#define _NSIG 129
+/* MIPS is special by having 128 signals.
+ * All (?) other architectures have at most 64 signals.
+ * Having 128 signals is problematic because signal nos are 1-based
+ * and last signal number is then 128.
+ * This plays havoc with WIFSIGNALED and WCOREDUMP in waitpid status word,
+ * when process dies from signal 128.
+ * Linux kernel 3.9 accepts signal 128, with awful results :/
+ * It is being fixed.
+ * glibc (accidentally?) papers over this issue by declaring _NSIG to be 128,
+ * not 129 (despite claiming that _NSIG is "biggest signal number + 1"
+ * in the comment above that definition). We follow suit.
+ * Note that this results in __SIGRTMAX == 127. It is intended.
+#define _NSIG 128
#endif /* <signal.h> included. */