[klibc] signal handling broken on i386 (Re: fix i386 build)

[H. Peter Anvin]

On 05/10/2010 01:28 PM, Sergey Vlasov wrote:
> 
> This compiles, but fails at runtime on i386, because
> arch/x86/include/asm/signal.h in the kernel does not provide the
> correct definition of sigset_t needed for klibc:
> 
> | #ifdef __KERNEL__
> | #include <linux/linkage.h>
> | 
> | /* Most things should be clean enough to redefine this at will, if care
> |    is taken to make libc match.  */
> | 
> | #define _NSIG           64
> | 
> | #ifdef __i386__
> | # define _NSIG_BPW      32
> | #else
> | # define _NSIG_BPW      64
> | #endif
> | 
> | #define _NSIG_WORDS     (_NSIG / _NSIG_BPW)
> | 
> | typedef unsigned long old_sigset_t;             /* at least 32 bits */
> | 
> | typedef struct {
> |         unsigned long sig[_NSIG_WORDS];
> | } sigset_t;
> | 
> | #else
> | /* Here we must cater to libcs that poke about in kernel headers.  */
> | 
> | #define NSIG            32
> | typedef unsigned long sigset_t;
> | 
> | #endif /* __KERNEL__ */
> 
> This gives sizeof(sigset_t) == 4, which is then rejected with EINVAL
> by rt_sigprocmask() and other syscalls.  The subsequent definition of
> struct sigaction is also wrong for klibc.
> 
> Looks like the libc5 crap is not really dead yet...
> 

The above is, in fact, said libc5 crap.

	-hpa