On Sat, May 13, 2017 at 07:04:13PM +0100, Al Viro wrote:
> My point is, this stuff needs looking at. Even this quick look in arch/x86
> has shown several fairly different classes of that stuff, probably needing
> different approaches. And that - on an architecture that had tons of TLC
> around signal delivery; I'm not saying that result is optimal (asm-goto sounds
> potentially useful there), but it had a lot of attention given to it...
BTW, even in arch/* they tend to nest. E.g. arch/alpha has 133 callers
total. Distribution by files:
35 arch/alpha/kernel/osf_sys.c
92 arch/alpha/kernel/signal.c
1 arch/alpha/kernel/traps.c
4 arch/alpha/lib/csum_partial_copy.c
1 arch/alpha/mm/fault.c
Distribution by functions:
1 osf_getdomainname() [1]
2 osf_sigstack()
2 get_tv32()
2 put_tv32()
4 get_it32()
4 put_it32()
2 osf_select()
18 osf_wait4() [2]
6 osf_sigaction()
34 restore_sigcontext()
1 do_sigreturn()
42 setup_sigcontext()
3 setup_frame()
6 setup_rt_frame()
1 dik_show_code() [3]
2 csum_partial_cfu_aligned()
2 csum_partial_cfu_src_aligned()
1 do_page_fault() [4]
[1] insane, BTW - should be strnlen() + copy_to_user(); should report -EFAULT
on failure, while we are at it.
[2] with fairly disgusting use of set_fs() in the mix.
[3] would break with get_user() - it's oopser fetching code to printk.
[4] this:
/* As of EV6, a load into $31/$f31 is a prefetch, and never faults
(or is suppressed by the PALcode). Support that for older CPUs
by ignoring such an instruction. */
if (cause == 0) {
unsigned int insn;
__get_user(insn, (unsigned int __user *)regs->pc);
if ((insn >> 21 & 0x1f) == 0x1f &&
/* ldq ldl ldt lds ldg ldf ldwu ldbu */
(1ul << (insn >> 26) & 0x30f00001400ul)) {
regs->pc += 4;
return;
}
}
