linux/arch/parisc/kernel/signal.c
Helge Deller 3847dab774 parisc: Add alternative coding infrastructure
This patch adds the necessary code to patch a running kernel at runtime
to improve performance.

The current implementation offers a few optimizations variants:

- When running a SMP kernel on a single UP processor, unwanted assembler
  statements like locking functions are overwritten with NOPs. When
  multiple instructions shall be skipped, one branch instruction is used
  instead of multiple nop instructions.

- In the UP case, some pdtlb and pitlb instructions are patched to
  become pdtlb,l and pitlb,l which only flushes the CPU-local tlb
  entries instead of broadcasting the flush to other CPUs in the system
  and thus may improve performance.

- fic and fdc instructions are skipped if no I- or D-caches are
  installed.  This should speed up qemu emulation and cacheless systems.

- If no cache coherence is needed for IO operations, the relevant fdc
  and sync instructions in the sba and ccio drivers are replaced by
  nops.

- On systems which share I- and D-TLBs and thus don't have a seperate
  instruction TLB, the pitlb instruction is replaced by a nop.

Live-patching is done early in the boot process, just after having run
the system inventory. No drivers are running and thus no external
interrupts should arrive. So the hope is that no TLB exceptions will
occur during the patching. If this turns out to be wrong we will
probably need to do the patching in real-mode.

Signed-off-by: Helge Deller <deller@gmx.de>
2018-10-17 17:22:26 +02:00

615 lines
18 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* linux/arch/parisc/kernel/signal.c: Architecture-specific signal
* handling support.
*
* Copyright (C) 2000 David Huggins-Daines <dhd@debian.org>
* Copyright (C) 2000 Linuxcare, Inc.
*
* Based on the ia64, i386, and alpha versions.
*
* Like the IA-64, we are a recent enough port (we are *starting*
* with glibc2.2) that we do not need to support the old non-realtime
* Linux signals. Therefore we don't.
*/
#include <linux/sched.h>
#include <linux/sched/debug.h>
#include <linux/mm.h>
#include <linux/smp.h>
#include <linux/kernel.h>
#include <linux/signal.h>
#include <linux/errno.h>
#include <linux/wait.h>
#include <linux/ptrace.h>
#include <linux/tracehook.h>
#include <linux/unistd.h>
#include <linux/stddef.h>
#include <linux/compat.h>
#include <linux/elf.h>
#include <asm/ucontext.h>
#include <asm/rt_sigframe.h>
#include <linux/uaccess.h>
#include <asm/pgalloc.h>
#include <asm/cacheflush.h>
#include <asm/asm-offsets.h>
#ifdef CONFIG_COMPAT
#include "signal32.h"
#endif
#define DEBUG_SIG 0
#define DEBUG_SIG_LEVEL 2
#if DEBUG_SIG
#define DBG(LEVEL, ...) \
((DEBUG_SIG_LEVEL >= LEVEL) \
? printk(__VA_ARGS__) : (void) 0)
#else
#define DBG(LEVEL, ...)
#endif
/* gcc will complain if a pointer is cast to an integer of different
* size. If you really need to do this (and we do for an ELF32 user
* application in an ELF64 kernel) then you have to do a cast to an
* integer of the same size first. The A() macro accomplishes
* this. */
#define A(__x) ((unsigned long)(__x))
/*
* Do a signal return - restore sigcontext.
*/
/* Trampoline for calling rt_sigreturn() */
#define INSN_LDI_R25_0 0x34190000 /* ldi 0,%r25 (in_syscall=0) */
#define INSN_LDI_R25_1 0x34190002 /* ldi 1,%r25 (in_syscall=1) */
#define INSN_LDI_R20 0x3414015a /* ldi __NR_rt_sigreturn,%r20 */
#define INSN_BLE_SR2_R0 0xe4008200 /* be,l 0x100(%sr2,%r0),%sr0,%r31 */
/* For debugging */
#define INSN_DIE_HORRIBLY 0x68000ccc /* stw %r0,0x666(%sr0,%r0) */
static long
restore_sigcontext(struct sigcontext __user *sc, struct pt_regs *regs)
{
long err = 0;
err |= __copy_from_user(regs->gr, sc->sc_gr, sizeof(regs->gr));
err |= __copy_from_user(regs->fr, sc->sc_fr, sizeof(regs->fr));
err |= __copy_from_user(regs->iaoq, sc->sc_iaoq, sizeof(regs->iaoq));
err |= __copy_from_user(regs->iasq, sc->sc_iasq, sizeof(regs->iasq));
err |= __get_user(regs->sar, &sc->sc_sar);
DBG(2,"restore_sigcontext: iaoq is %#lx / %#lx\n",
regs->iaoq[0],regs->iaoq[1]);
DBG(2,"restore_sigcontext: r28 is %ld\n", regs->gr[28]);
return err;
}
void
sys_rt_sigreturn(struct pt_regs *regs, int in_syscall)
{
struct rt_sigframe __user *frame;
sigset_t set;
unsigned long usp = (regs->gr[30] & ~(0x01UL));
unsigned long sigframe_size = PARISC_RT_SIGFRAME_SIZE;
#ifdef CONFIG_64BIT
struct compat_rt_sigframe __user * compat_frame;
if (is_compat_task())
sigframe_size = PARISC_RT_SIGFRAME_SIZE32;
#endif
current->restart_block.fn = do_no_restart_syscall;
/* Unwind the user stack to get the rt_sigframe structure. */
frame = (struct rt_sigframe __user *)
(usp - sigframe_size);
DBG(2,"sys_rt_sigreturn: frame is %p\n", frame);
regs->orig_r28 = 1; /* no restarts for sigreturn */
#ifdef CONFIG_64BIT
compat_frame = (struct compat_rt_sigframe __user *)frame;
if (is_compat_task()) {
DBG(2,"sys_rt_sigreturn: ELF32 process.\n");
if (get_compat_sigset(&set, &compat_frame->uc.uc_sigmask))
goto give_sigsegv;
} else
#endif
{
if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
goto give_sigsegv;
}
set_current_blocked(&set);
/* Good thing we saved the old gr[30], eh? */
#ifdef CONFIG_64BIT
if (is_compat_task()) {
DBG(1,"sys_rt_sigreturn: compat_frame->uc.uc_mcontext 0x%p\n",
&compat_frame->uc.uc_mcontext);
// FIXME: Load upper half from register file
if (restore_sigcontext32(&compat_frame->uc.uc_mcontext,
&compat_frame->regs, regs))
goto give_sigsegv;
DBG(1,"sys_rt_sigreturn: usp %#08lx stack 0x%p\n",
usp, &compat_frame->uc.uc_stack);
if (compat_restore_altstack(&compat_frame->uc.uc_stack))
goto give_sigsegv;
} else
#endif
{
DBG(1,"sys_rt_sigreturn: frame->uc.uc_mcontext 0x%p\n",
&frame->uc.uc_mcontext);
if (restore_sigcontext(&frame->uc.uc_mcontext, regs))
goto give_sigsegv;
DBG(1,"sys_rt_sigreturn: usp %#08lx stack 0x%p\n",
usp, &frame->uc.uc_stack);
if (restore_altstack(&frame->uc.uc_stack))
goto give_sigsegv;
}
/* If we are on the syscall path IAOQ will not be restored, and
* if we are on the interrupt path we must not corrupt gr31.
*/
if (in_syscall)
regs->gr[31] = regs->iaoq[0];
#if DEBUG_SIG
DBG(1,"sys_rt_sigreturn: returning to %#lx, DUMPING REGS:\n", regs->iaoq[0]);
show_regs(regs);
#endif
return;
give_sigsegv:
DBG(1,"sys_rt_sigreturn: Sending SIGSEGV\n");
force_sig(SIGSEGV, current);
return;
}
/*
* Set up a signal frame.
*/
static inline void __user *
get_sigframe(struct k_sigaction *ka, unsigned long sp, size_t frame_size)
{
/*FIXME: ELF32 vs. ELF64 has different frame_size, but since we
don't use the parameter it doesn't matter */
DBG(1,"get_sigframe: ka = %#lx, sp = %#lx, frame_size = %#lx\n",
(unsigned long)ka, sp, frame_size);
/* Align alternate stack and reserve 64 bytes for the signal
handler's frame marker. */
if ((ka->sa.sa_flags & SA_ONSTACK) != 0 && ! sas_ss_flags(sp))
sp = (current->sas_ss_sp + 0x7f) & ~0x3f; /* Stacks grow up! */
DBG(1,"get_sigframe: Returning sp = %#lx\n", (unsigned long)sp);
return (void __user *) sp; /* Stacks grow up. Fun. */
}
static long
setup_sigcontext(struct sigcontext __user *sc, struct pt_regs *regs, int in_syscall)
{
unsigned long flags = 0;
long err = 0;
if (on_sig_stack((unsigned long) sc))
flags |= PARISC_SC_FLAG_ONSTACK;
if (in_syscall) {
flags |= PARISC_SC_FLAG_IN_SYSCALL;
/* regs->iaoq is undefined in the syscall return path */
err |= __put_user(regs->gr[31], &sc->sc_iaoq[0]);
err |= __put_user(regs->gr[31]+4, &sc->sc_iaoq[1]);
err |= __put_user(regs->sr[3], &sc->sc_iasq[0]);
err |= __put_user(regs->sr[3], &sc->sc_iasq[1]);
DBG(1,"setup_sigcontext: iaoq %#lx / %#lx (in syscall)\n",
regs->gr[31], regs->gr[31]+4);
} else {
err |= __copy_to_user(sc->sc_iaoq, regs->iaoq, sizeof(regs->iaoq));
err |= __copy_to_user(sc->sc_iasq, regs->iasq, sizeof(regs->iasq));
DBG(1,"setup_sigcontext: iaoq %#lx / %#lx (not in syscall)\n",
regs->iaoq[0], regs->iaoq[1]);
}
err |= __put_user(flags, &sc->sc_flags);
err |= __copy_to_user(sc->sc_gr, regs->gr, sizeof(regs->gr));
err |= __copy_to_user(sc->sc_fr, regs->fr, sizeof(regs->fr));
err |= __put_user(regs->sar, &sc->sc_sar);
DBG(1,"setup_sigcontext: r28 is %ld\n", regs->gr[28]);
return err;
}
static long
setup_rt_frame(struct ksignal *ksig, sigset_t *set, struct pt_regs *regs,
int in_syscall)
{
struct rt_sigframe __user *frame;
unsigned long rp, usp;
unsigned long haddr, sigframe_size;
unsigned long start, end;
int err = 0;
#ifdef CONFIG_64BIT
struct compat_rt_sigframe __user * compat_frame;
#endif
usp = (regs->gr[30] & ~(0x01UL));
/*FIXME: frame_size parameter is unused, remove it. */
frame = get_sigframe(&ksig->ka, usp, sizeof(*frame));
DBG(1,"SETUP_RT_FRAME: START\n");
DBG(1,"setup_rt_frame: frame %p info %p\n", frame, ksig->info);
#ifdef CONFIG_64BIT
compat_frame = (struct compat_rt_sigframe __user *)frame;
if (is_compat_task()) {
DBG(1,"setup_rt_frame: frame->info = 0x%p\n", &compat_frame->info);
err |= copy_siginfo_to_user32(&compat_frame->info, &ksig->info);
err |= __compat_save_altstack( &compat_frame->uc.uc_stack, regs->gr[30]);
DBG(1,"setup_rt_frame: frame->uc = 0x%p\n", &compat_frame->uc);
DBG(1,"setup_rt_frame: frame->uc.uc_mcontext = 0x%p\n", &compat_frame->uc.uc_mcontext);
err |= setup_sigcontext32(&compat_frame->uc.uc_mcontext,
&compat_frame->regs, regs, in_syscall);
err |= put_compat_sigset(&compat_frame->uc.uc_sigmask, set,
sizeof(compat_sigset_t));
} else
#endif
{
DBG(1,"setup_rt_frame: frame->info = 0x%p\n", &frame->info);
err |= copy_siginfo_to_user(&frame->info, &ksig->info);
err |= __save_altstack(&frame->uc.uc_stack, regs->gr[30]);
DBG(1,"setup_rt_frame: frame->uc = 0x%p\n", &frame->uc);
DBG(1,"setup_rt_frame: frame->uc.uc_mcontext = 0x%p\n", &frame->uc.uc_mcontext);
err |= setup_sigcontext(&frame->uc.uc_mcontext, regs, in_syscall);
/* FIXME: Should probably be converted as well for the compat case */
err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set));
}
if (err)
return -EFAULT;
/* Set up to return from userspace. If provided, use a stub
already in userspace. The first words of tramp are used to
save the previous sigrestartblock trampoline that might be
on the stack. We start the sigreturn trampoline at
SIGRESTARTBLOCK_TRAMP+X. */
err |= __put_user(in_syscall ? INSN_LDI_R25_1 : INSN_LDI_R25_0,
&frame->tramp[SIGRESTARTBLOCK_TRAMP+0]);
err |= __put_user(INSN_LDI_R20,
&frame->tramp[SIGRESTARTBLOCK_TRAMP+1]);
err |= __put_user(INSN_BLE_SR2_R0,
&frame->tramp[SIGRESTARTBLOCK_TRAMP+2]);
err |= __put_user(INSN_NOP, &frame->tramp[SIGRESTARTBLOCK_TRAMP+3]);
#if DEBUG_SIG
/* Assert that we're flushing in the correct space... */
{
unsigned long sid;
asm ("mfsp %%sr3,%0" : "=r" (sid));
DBG(1,"setup_rt_frame: Flushing 64 bytes at space %#x offset %p\n",
sid, frame->tramp);
}
#endif
start = (unsigned long) &frame->tramp[0];
end = (unsigned long) &frame->tramp[TRAMP_SIZE];
flush_user_dcache_range_asm(start, end);
flush_user_icache_range_asm(start, end);
/* TRAMP Words 0-4, Length 5 = SIGRESTARTBLOCK_TRAMP
* TRAMP Words 5-9, Length 4 = SIGRETURN_TRAMP
* So the SIGRETURN_TRAMP is at the end of SIGRESTARTBLOCK_TRAMP
*/
rp = (unsigned long) &frame->tramp[SIGRESTARTBLOCK_TRAMP];
if (err)
return -EFAULT;
haddr = A(ksig->ka.sa.sa_handler);
/* The sa_handler may be a pointer to a function descriptor */
#ifdef CONFIG_64BIT
if (is_compat_task()) {
#endif
if (haddr & PA_PLABEL_FDESC) {
Elf32_Fdesc fdesc;
Elf32_Fdesc __user *ufdesc = (Elf32_Fdesc __user *)A(haddr & ~3);
err = __copy_from_user(&fdesc, ufdesc, sizeof(fdesc));
if (err)
return -EFAULT;
haddr = fdesc.addr;
regs->gr[19] = fdesc.gp;
}
#ifdef CONFIG_64BIT
} else {
Elf64_Fdesc fdesc;
Elf64_Fdesc __user *ufdesc = (Elf64_Fdesc __user *)A(haddr & ~3);
err = __copy_from_user(&fdesc, ufdesc, sizeof(fdesc));
if (err)
return -EFAULT;
haddr = fdesc.addr;
regs->gr[19] = fdesc.gp;
DBG(1,"setup_rt_frame: 64 bit signal, exe=%#lx, r19=%#lx, in_syscall=%d\n",
haddr, regs->gr[19], in_syscall);
}
#endif
/* The syscall return path will create IAOQ values from r31.
*/
sigframe_size = PARISC_RT_SIGFRAME_SIZE;
#ifdef CONFIG_64BIT
if (is_compat_task())
sigframe_size = PARISC_RT_SIGFRAME_SIZE32;
#endif
if (in_syscall) {
regs->gr[31] = haddr;
#ifdef CONFIG_64BIT
if (!test_thread_flag(TIF_32BIT))
sigframe_size |= 1;
#endif
} else {
unsigned long psw = USER_PSW;
#ifdef CONFIG_64BIT
if (!test_thread_flag(TIF_32BIT))
psw |= PSW_W;
#endif
/* If we are singlestepping, arrange a trap to be delivered
when we return to userspace. Note the semantics -- we
should trap before the first insn in the handler is
executed. Ref:
http://sources.redhat.com/ml/gdb/2004-11/msg00245.html
*/
if (pa_psw(current)->r) {
pa_psw(current)->r = 0;
psw |= PSW_R;
mtctl(-1, 0);
}
regs->gr[0] = psw;
regs->iaoq[0] = haddr | 3;
regs->iaoq[1] = regs->iaoq[0] + 4;
}
regs->gr[2] = rp; /* userland return pointer */
regs->gr[26] = ksig->sig; /* signal number */
#ifdef CONFIG_64BIT
if (is_compat_task()) {
regs->gr[25] = A(&compat_frame->info); /* siginfo pointer */
regs->gr[24] = A(&compat_frame->uc); /* ucontext pointer */
} else
#endif
{
regs->gr[25] = A(&frame->info); /* siginfo pointer */
regs->gr[24] = A(&frame->uc); /* ucontext pointer */
}
DBG(1,"setup_rt_frame: making sigreturn frame: %#lx + %#lx = %#lx\n",
regs->gr[30], sigframe_size,
regs->gr[30] + sigframe_size);
/* Raise the user stack pointer to make a proper call frame. */
regs->gr[30] = (A(frame) + sigframe_size);
DBG(1,"setup_rt_frame: sig deliver (%s,%d) frame=0x%p sp=%#lx iaoq=%#lx/%#lx rp=%#lx\n",
current->comm, current->pid, frame, regs->gr[30],
regs->iaoq[0], regs->iaoq[1], rp);
return 0;
}
/*
* OK, we're invoking a handler.
*/
static void
handle_signal(struct ksignal *ksig, struct pt_regs *regs, int in_syscall)
{
int ret;
sigset_t *oldset = sigmask_to_save();
DBG(1,"handle_signal: sig=%ld, ka=%p, info=%p, oldset=%p, regs=%p\n",
ksig->sig, ksig->ka, ksig->info, oldset, regs);
/* Set up the stack frame */
ret = setup_rt_frame(ksig, oldset, regs, in_syscall);
signal_setup_done(ret, ksig, test_thread_flag(TIF_SINGLESTEP) ||
test_thread_flag(TIF_BLOCKSTEP));
DBG(1,KERN_DEBUG "do_signal: Exit (success), regs->gr[28] = %ld\n",
regs->gr[28]);
}
/*
* Check how the syscall number gets loaded into %r20 within
* the delay branch in userspace and adjust as needed.
*/
static void check_syscallno_in_delay_branch(struct pt_regs *regs)
{
u32 opcode, source_reg;
u32 __user *uaddr;
int err;
/* Usually we don't have to restore %r20 (the system call number)
* because it gets loaded in the delay slot of the branch external
* instruction via the ldi instruction.
* In some cases a register-to-register copy instruction might have
* been used instead, in which case we need to copy the syscall
* number into the source register before returning to userspace.
*/
/* A syscall is just a branch, so all we have to do is fiddle the
* return pointer so that the ble instruction gets executed again.
*/
regs->gr[31] -= 8; /* delayed branching */
/* Get assembler opcode of code in delay branch */
uaddr = (unsigned int *) ((regs->gr[31] & ~3) + 4);
err = get_user(opcode, uaddr);
if (err)
return;
/* Check if delay branch uses "ldi int,%r20" */
if ((opcode & 0xffff0000) == 0x34140000)
return; /* everything ok, just return */
/* Check if delay branch uses "nop" */
if (opcode == INSN_NOP)
return;
/* Check if delay branch uses "copy %rX,%r20" */
if ((opcode & 0xffe0ffff) == 0x08000254) {
source_reg = (opcode >> 16) & 31;
regs->gr[source_reg] = regs->gr[20];
return;
}
pr_warn("syscall restart: %s (pid %d): unexpected opcode 0x%08x\n",
current->comm, task_pid_nr(current), opcode);
}
static inline void
syscall_restart(struct pt_regs *regs, struct k_sigaction *ka)
{
if (regs->orig_r28)
return;
regs->orig_r28 = 1; /* no more restarts */
/* Check the return code */
switch (regs->gr[28]) {
case -ERESTART_RESTARTBLOCK:
case -ERESTARTNOHAND:
DBG(1,"ERESTARTNOHAND: returning -EINTR\n");
regs->gr[28] = -EINTR;
break;
case -ERESTARTSYS:
if (!(ka->sa.sa_flags & SA_RESTART)) {
DBG(1,"ERESTARTSYS: putting -EINTR\n");
regs->gr[28] = -EINTR;
break;
}
/* fallthrough */
case -ERESTARTNOINTR:
check_syscallno_in_delay_branch(regs);
break;
}
}
static inline void
insert_restart_trampoline(struct pt_regs *regs)
{
if (regs->orig_r28)
return;
regs->orig_r28 = 1; /* no more restarts */
switch(regs->gr[28]) {
case -ERESTART_RESTARTBLOCK: {
/* Restart the system call - no handlers present */
unsigned int *usp = (unsigned int *)regs->gr[30];
unsigned long start = (unsigned long) &usp[2];
unsigned long end = (unsigned long) &usp[5];
long err = 0;
/* Setup a trampoline to restart the syscall
* with __NR_restart_syscall
*
* 0: <return address (orig r31)>
* 4: <2nd half for 64-bit>
* 8: ldw 0(%sp), %r31
* 12: be 0x100(%sr2, %r0)
* 16: ldi __NR_restart_syscall, %r20
*/
#ifdef CONFIG_64BIT
err |= put_user(regs->gr[31] >> 32, &usp[0]);
err |= put_user(regs->gr[31] & 0xffffffff, &usp[1]);
err |= put_user(0x0fc010df, &usp[2]);
#else
err |= put_user(regs->gr[31], &usp[0]);
err |= put_user(0x0fc0109f, &usp[2]);
#endif
err |= put_user(0xe0008200, &usp[3]);
err |= put_user(0x34140000, &usp[4]);
WARN_ON(err);
/* flush data/instruction cache for new insns */
flush_user_dcache_range_asm(start, end);
flush_user_icache_range_asm(start, end);
regs->gr[31] = regs->gr[30] + 8;
return;
}
case -ERESTARTNOHAND:
case -ERESTARTSYS:
case -ERESTARTNOINTR:
check_syscallno_in_delay_branch(regs);
return;
default:
break;
}
}
/*
* Note that 'init' is a special process: it doesn't get signals it doesn't
* want to handle. Thus you cannot kill init even with a SIGKILL even by
* mistake.
*
* We need to be able to restore the syscall arguments (r21-r26) to
* restart syscalls. Thus, the syscall path should save them in the
* pt_regs structure (it's okay to do so since they are caller-save
* registers). As noted below, the syscall number gets restored for
* us due to the magic of delayed branching.
*/
asmlinkage void
do_signal(struct pt_regs *regs, long in_syscall)
{
struct ksignal ksig;
DBG(1,"\ndo_signal: regs=0x%p, sr7 %#lx, in_syscall=%d\n",
regs, regs->sr[7], in_syscall);
if (get_signal(&ksig)) {
DBG(3,"do_signal: signr = %d, regs->gr[28] = %ld\n", signr, regs->gr[28]);
/* Restart a system call if necessary. */
if (in_syscall)
syscall_restart(regs, &ksig.ka);
handle_signal(&ksig, regs, in_syscall);
return;
}
/* Did we come from a system call? */
if (in_syscall)
insert_restart_trampoline(regs);
DBG(1,"do_signal: Exit (not delivered), regs->gr[28] = %ld\n",
regs->gr[28]);
restore_saved_sigmask();
}
void do_notify_resume(struct pt_regs *regs, long in_syscall)
{
if (test_thread_flag(TIF_SIGPENDING))
do_signal(regs, in_syscall);
if (test_thread_flag(TIF_NOTIFY_RESUME)) {
clear_thread_flag(TIF_NOTIFY_RESUME);
tracehook_notify_resume(regs);
}
}