user/sven/linux.git/include/linux/signal.h, branch v3.9

Fix breakage in MIPS siginfo handling

2013-03-19T18:15:52Z

MIPS's siginfo handling has been broken since this commit: commit 574c4866e33d648520a8bd5bf6f573ea6e554e88 Author: Al Viro Date: Sun Nov 25 22:24:19 2012 -0500 consolidate kernel-side struct sigaction declarations for 64-bit BE MIPS CPUs. The UAPI variant looks like this: struct sigaction { unsigned int sa_flags; __sighandler_t sa_handler; sigset_t sa_mask; }; but the core kernel's variant looks like this: struct sigaction { #ifndef __ARCH_HAS_ODD_SIGACTION __sighandler_t sa_handler; unsigned long sa_flags; #else unsigned long sa_flags; __sighandler_t sa_handler; #endif #ifdef __ARCH_HAS_SA_RESTORER __sigrestore_t sa_restorer; #endif sigset_t sa_mask; }; The problem is that sa_flags has been changed from an unsigned int to an unsigned long. Fix this by making sa_flags unsigned int if __ARCH_HAS_ODD_SIGACTION is defined. Whilst we're at it, rename __ARCH_HAS_ODD_SIGACTION to __ARCH_HAS_IRIX_SIGACTION. Signed-off-by: David Howells Cc: linux-mips@linux-mips.org Cc: linux-kernel@vger.kernel.org Cc: stable@vger.kernel.org Acked-by: Al Viro Signed-off-by: Ralf Baechle

new helper: signal_setup_done()

2013-02-03T20:09:26Z

usual "call force_sigsegv or signal_delivered" logics. Takes ksignal instead of separate siginfo/k_sigaction. Signed-off-by: Al Viro

new helper: get_signal()

2013-02-03T20:09:25Z

On success get_signal_to_deliver() fills k_sigaction and siginfo. _All_ users pass it addresses of the local variables sitting in the same function. Then they proceed to pass those addresses pretty much in tandem to a bunch of helper functions; again, all callers of those helpers are passing them such a pair, and one that had been through get_signal_to_deliver() at that. The obvious cleanup: introduce a new type that would contain a pair (struct ksignal) and begin switching to using it. Turns out that it's convenient to store the signal number in the same object. New helper, taking that sucker is a wrapper for get_signal_to_deliver(); takes struct ksignal * and returns bool. On success fills ksignal with the information for signal handler to be invoked. For now it's a macro (to avoid header ordering headache), but eventually it'll be a function in kernel/signal.c, with get_signal_to_deliver() folded into it. Signed-off-by: Al Viro

generic sys_sigaction() and compat_sys_sigaction()

2013-02-03T20:09:23Z

conditional on OLD_SIGACTION/COMPAT_OLD_SIGACTION Signed-off-by: Al Viro

consolidate kernel-side struct sigaction declarations

2013-02-03T20:09:22Z

Signed-off-by: Al Viro

consolidate declarations of k_sigaction

2013-02-03T20:09:22Z

Only alpha and sparc are unusual - they have ka_restorer in it. And nobody needs that exposed to userland. Signed-off-by: Al Viro

switch rt_tgsigqueueinfo to COMPAT_SYSCALL_DEFINE

2013-02-03T20:09:21Z

C ABI violations on sparc, ppc and mips Signed-off-by: Al Viro

generic compat_sys_rt_sigpending()

2013-02-03T20:09:19Z

conditional on GENERIC_COMPAT_RT_SIGPENDING; by the end of that series it will become the same thing as COMPAT and conditional will die out. Signed-off-by: Al Viro

new helpers: __save_altstack/__compat_save_altstack, switch x86 and um to those

2012-12-19T23:07:41Z

note that they are relying on access_ok() already checked by caller. Signed-off-by: Al Viro

new helper: restore_altstack()

2012-12-19T23:07:40Z

to be used by rt_sigreturn instances Signed-off-by: Al Viro