[S390] add kprobes support.
authorMichael Grundy <grundym@us.ibm.com>
Wed, 20 Sep 2006 13:58:39 +0000 (15:58 +0200)
committerMartin Schwidefsky <schwidefsky@de.ibm.com>
Wed, 20 Sep 2006 13:58:39 +0000 (15:58 +0200)
Signed-off-by: Michael Grundy <grundym@us.ibm.com>
Signed-off-by: David Wilder <dwilder@us.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
arch/s390/Kconfig
arch/s390/kernel/Makefile
arch/s390/kernel/entry.S
arch/s390/kernel/entry64.S
arch/s390/kernel/kprobes.c [new file with mode: 0644]
arch/s390/kernel/traps.c
arch/s390/kernel/vmlinux.lds.S
arch/s390/mm/fault.c
include/asm-s390/kdebug.h [new file with mode: 0644]
include/asm-s390/kprobes.h [new file with mode: 0644]

index 2f4f70c4dbb293caf433de6ee3f3f15202278f73..76122ce1e6cbc1977384a139d6452860cfadf21e 100644 (file)
@@ -487,8 +487,22 @@ source "drivers/net/Kconfig"
 
 source "fs/Kconfig"
 
+menu "Instrumentation Support"
+
 source "arch/s390/oprofile/Kconfig"
 
+config KPROBES
+       bool "Kprobes (EXPERIMENTAL)"
+       depends on EXPERIMENTAL && MODULES
+       help
+         Kprobes allows you to trap at almost any kernel address and
+         execute a callback function.  register_kprobe() establishes
+         a probepoint and specifies the callback.  Kprobes is useful
+         for kernel debugging, non-intrusive instrumentation and testing.
+         If in doubt, say "N".
+
+endmenu
+
 source "arch/s390/Kconfig.debug"
 
 source "security/Kconfig"
index 9a33ed6ca69605915adbfc68e23ec9ea645beddc..33a5069c0e1611893b367a1133c6f16818205320 100644 (file)
@@ -24,6 +24,7 @@ obj-$(CONFIG_COMPAT)          += compat_linux.o compat_signal.o \
 
 obj-$(CONFIG_VIRT_TIMER)       += vtime.o
 obj-$(CONFIG_STACKTRACE)       += stacktrace.o
+obj-$(CONFIG_KPROBES)          += kprobes.o
 
 # Kexec part
 S390_KEXEC_OBJS := machine_kexec.o crash.o
index 5b5799ac8f8397012b81512821419213129e57b5..0c712b78a7e82480f5392381c89112e8550d5da0 100644 (file)
@@ -505,6 +505,8 @@ pgm_no_vtime2:
        mvc     __THREAD_per+__PER_address(4,%r1),__LC_PER_ADDRESS
        mvc     __THREAD_per+__PER_access_id(1,%r1),__LC_PER_ACCESS_ID
        oi      __TI_flags+3(%r9),_TIF_SINGLE_STEP # set TIF_SINGLE_STEP
+       tm      SP_PSW+1(%r15),0x01     # kernel per event ?
+       bz      BASED(kernel_per)
        l       %r3,__LC_PGM_ILC         # load program interruption code
        la      %r8,0x7f
        nr      %r8,%r3                  # clear per-event-bit and ilc
@@ -536,6 +538,16 @@ pgm_no_vtime3:
        stosm   __SF_EMPTY(%r15),0x03   # reenable interrupts
        b       BASED(sysc_do_svc)
 
+#
+# per was called from kernel, must be kprobes
+#
+kernel_per:
+       mvi     SP_TRAP+1(%r15),0x28    # set trap indication to pgm check
+       la      %r2,SP_PTREGS(%r15)     # address of register-save area
+       l       %r1,BASED(.Lhandle_per) # load adr. of per handler
+       la      %r14,BASED(sysc_leave)  # load adr. of system return
+       br      %r1                     # branch to do_single_step
+
 /*
  * IO interrupt handler routine
  */
index 56f5f613b868f6da5d07bb294e2435c2c6c6f3fc..8b956d1538f552acd464ec8b5b374fe43e7f65b7 100644 (file)
@@ -518,6 +518,8 @@ pgm_no_vtime2:
 #endif
        lg      %r9,__LC_THREAD_INFO    # load pointer to thread_info struct
        lg      %r1,__TI_task(%r9)
+       tm      SP_PSW+1(%r15),0x01     # kernel per event ?
+       jz      kernel_per
        mvc     __THREAD_per+__PER_atmid(2,%r1),__LC_PER_ATMID
        mvc     __THREAD_per+__PER_address(8,%r1),__LC_PER_ADDRESS
        mvc     __THREAD_per+__PER_access_id(1,%r1),__LC_PER_ACCESS_ID
@@ -553,6 +555,16 @@ pgm_no_vtime3:
        stosm   __SF_EMPTY(%r15),0x03   # reenable interrupts
        j       sysc_do_svc
 
+#
+# per was called from kernel, must be kprobes
+#
+kernel_per:
+       lhi     %r0,__LC_PGM_OLD_PSW
+       sth     %r0,SP_TRAP(%r15)       # set trap indication to pgm check
+       la      %r2,SP_PTREGS(%r15)     # address of register-save area
+       larl    %r14,sysc_leave         # load adr. of system ret, no work
+       jg      do_single_step          # branch to do_single_step
+
 /*
  * IO interrupt handler routine
  */
diff --git a/arch/s390/kernel/kprobes.c b/arch/s390/kernel/kprobes.c
new file mode 100644 (file)
index 0000000..ca28fb0
--- /dev/null
@@ -0,0 +1,657 @@
+/*
+ *  Kernel Probes (KProbes)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright (C) IBM Corporation, 2002, 2006
+ *
+ * s390 port, used ppc64 as template. Mike Grundy <grundym@us.ibm.com>
+ */
+
+#include <linux/config.h>
+#include <linux/kprobes.h>
+#include <linux/ptrace.h>
+#include <linux/preempt.h>
+#include <linux/stop_machine.h>
+#include <asm/cacheflush.h>
+#include <asm/kdebug.h>
+#include <asm/sections.h>
+#include <asm/uaccess.h>
+#include <linux/module.h>
+
+DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
+DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);
+
+int __kprobes arch_prepare_kprobe(struct kprobe *p)
+{
+       /* Make sure the probe isn't going on a difficult instruction */
+       if (is_prohibited_opcode((kprobe_opcode_t *) p->addr))
+               return -EINVAL;
+
+       if ((unsigned long)p->addr & 0x01) {
+               printk("Attempt to register kprobe at an unaligned address\n");
+               return -EINVAL;
+               }
+
+       /* Use the get_insn_slot() facility for correctness */
+       if (!(p->ainsn.insn = get_insn_slot()))
+               return -ENOMEM;
+
+       memcpy(p->ainsn.insn, p->addr, MAX_INSN_SIZE * sizeof(kprobe_opcode_t));
+
+       get_instruction_type(&p->ainsn);
+       p->opcode = *p->addr;
+       return 0;
+}
+
+int __kprobes is_prohibited_opcode(kprobe_opcode_t *instruction)
+{
+       switch (*(__u8 *) instruction) {
+       case 0x0c:      /* bassm */
+       case 0x0b:      /* bsm   */
+       case 0x83:      /* diag  */
+       case 0x44:      /* ex    */
+               return -EINVAL;
+       }
+       switch (*(__u16 *) instruction) {
+       case 0x0101:    /* pr    */
+       case 0xb25a:    /* bsa   */
+       case 0xb240:    /* bakr  */
+       case 0xb258:    /* bsg   */
+       case 0xb218:    /* pc    */
+       case 0xb228:    /* pt    */
+               return -EINVAL;
+       }
+       return 0;
+}
+
+void __kprobes get_instruction_type(struct arch_specific_insn *ainsn)
+{
+       /* default fixup method */
+       ainsn->fixup = FIXUP_PSW_NORMAL;
+
+       /* save r1 operand */
+       ainsn->reg = (*ainsn->insn & 0xf0) >> 4;
+
+       /* save the instruction length (pop 5-5) in bytes */
+       switch (*(__u8 *) (ainsn->insn) >> 4) {
+       case 0:
+               ainsn->ilen = 2;
+               break;
+       case 1:
+       case 2:
+               ainsn->ilen = 4;
+               break;
+       case 3:
+               ainsn->ilen = 6;
+               break;
+       }
+
+       switch (*(__u8 *) ainsn->insn) {
+       case 0x05:      /* balr */
+       case 0x0d:      /* basr */
+               ainsn->fixup = FIXUP_RETURN_REGISTER;
+               /* if r2 = 0, no branch will be taken */
+               if ((*ainsn->insn & 0x0f) == 0)
+                       ainsn->fixup |= FIXUP_BRANCH_NOT_TAKEN;
+               break;
+       case 0x06:      /* bctr */
+       case 0x07:      /* bcr  */
+               ainsn->fixup = FIXUP_BRANCH_NOT_TAKEN;
+               break;
+       case 0x45:      /* bal  */
+       case 0x4d:      /* bas  */
+               ainsn->fixup = FIXUP_RETURN_REGISTER;
+               break;
+       case 0x47:      /* bc   */
+       case 0x46:      /* bct  */
+       case 0x86:      /* bxh  */
+       case 0x87:      /* bxle */
+               ainsn->fixup = FIXUP_BRANCH_NOT_TAKEN;
+               break;
+       case 0x82:      /* lpsw */
+               ainsn->fixup = FIXUP_NOT_REQUIRED;
+               break;
+       case 0xb2:      /* lpswe */
+               if (*(((__u8 *) ainsn->insn) + 1) == 0xb2) {
+                       ainsn->fixup = FIXUP_NOT_REQUIRED;
+               }
+               break;
+       case 0xa7:      /* bras */
+               if ((*ainsn->insn & 0x0f) == 0x05) {
+                       ainsn->fixup |= FIXUP_RETURN_REGISTER;
+               }
+               break;
+       case 0xc0:
+               if ((*ainsn->insn & 0x0f) == 0x00  /* larl  */
+                       || (*ainsn->insn & 0x0f) == 0x05) /* brasl */
+               ainsn->fixup |= FIXUP_RETURN_REGISTER;
+               break;
+       case 0xeb:
+               if (*(((__u8 *) ainsn->insn) + 5 ) == 0x44 ||   /* bxhg  */
+                       *(((__u8 *) ainsn->insn) + 5) == 0x45) {/* bxleg */
+                       ainsn->fixup = FIXUP_BRANCH_NOT_TAKEN;
+               }
+               break;
+       case 0xe3:      /* bctg */
+               if (*(((__u8 *) ainsn->insn) + 5) == 0x46) {
+                       ainsn->fixup = FIXUP_BRANCH_NOT_TAKEN;
+               }
+               break;
+       }
+}
+
+static int __kprobes swap_instruction(void *aref)
+{
+       struct ins_replace_args *args = aref;
+       int err = -EFAULT;
+
+       asm volatile(
+               "0: mvc  0(2,%2),0(%3)\n"
+               "1: la   %0,0\n"
+               "2:\n"
+               EX_TABLE(0b,2b)
+               : "+d" (err), "=m" (*args->ptr)
+               : "a" (args->ptr), "a" (&args->new), "m" (args->new));
+       return err;
+}
+
+void __kprobes arch_arm_kprobe(struct kprobe *p)
+{
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+       unsigned long status = kcb->kprobe_status;
+       struct ins_replace_args args;
+
+       args.ptr = p->addr;
+       args.old = p->opcode;
+       args.new = BREAKPOINT_INSTRUCTION;
+
+       kcb->kprobe_status = KPROBE_SWAP_INST;
+       stop_machine_run(swap_instruction, &args, NR_CPUS);
+       kcb->kprobe_status = status;
+}
+
+void __kprobes arch_disarm_kprobe(struct kprobe *p)
+{
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+       unsigned long status = kcb->kprobe_status;
+       struct ins_replace_args args;
+
+       args.ptr = p->addr;
+       args.old = BREAKPOINT_INSTRUCTION;
+       args.new = p->opcode;
+
+       kcb->kprobe_status = KPROBE_SWAP_INST;
+       stop_machine_run(swap_instruction, &args, NR_CPUS);
+       kcb->kprobe_status = status;
+}
+
+void __kprobes arch_remove_kprobe(struct kprobe *p)
+{
+       mutex_lock(&kprobe_mutex);
+       free_insn_slot(p->ainsn.insn);
+       mutex_unlock(&kprobe_mutex);
+}
+
+static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
+{
+       per_cr_bits kprobe_per_regs[1];
+
+       memset(kprobe_per_regs, 0, sizeof(per_cr_bits));
+       regs->psw.addr = (unsigned long)p->ainsn.insn | PSW_ADDR_AMODE;
+
+       /* Set up the per control reg info, will pass to lctl */
+       kprobe_per_regs[0].em_instruction_fetch = 1;
+       kprobe_per_regs[0].starting_addr = (unsigned long)p->ainsn.insn;
+       kprobe_per_regs[0].ending_addr = (unsigned long)p->ainsn.insn + 1;
+
+       /* Set the PER control regs, turns on single step for this address */
+       __ctl_load(kprobe_per_regs, 9, 11);
+       regs->psw.mask |= PSW_MASK_PER;
+       regs->psw.mask &= ~(PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK);
+}
+
+static void __kprobes save_previous_kprobe(struct kprobe_ctlblk *kcb)
+{
+       kcb->prev_kprobe.kp = kprobe_running();
+       kcb->prev_kprobe.status = kcb->kprobe_status;
+       kcb->prev_kprobe.kprobe_saved_imask = kcb->kprobe_saved_imask;
+       memcpy(kcb->prev_kprobe.kprobe_saved_ctl, kcb->kprobe_saved_ctl,
+                                       sizeof(kcb->kprobe_saved_ctl));
+}
+
+static void __kprobes restore_previous_kprobe(struct kprobe_ctlblk *kcb)
+{
+       __get_cpu_var(current_kprobe) = kcb->prev_kprobe.kp;
+       kcb->kprobe_status = kcb->prev_kprobe.status;
+       kcb->kprobe_saved_imask = kcb->prev_kprobe.kprobe_saved_imask;
+       memcpy(kcb->kprobe_saved_ctl, kcb->prev_kprobe.kprobe_saved_ctl,
+                                       sizeof(kcb->kprobe_saved_ctl));
+}
+
+static void __kprobes set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
+                                               struct kprobe_ctlblk *kcb)
+{
+       __get_cpu_var(current_kprobe) = p;
+       /* Save the interrupt and per flags */
+       kcb->kprobe_saved_imask = regs->psw.mask &
+           (PSW_MASK_PER | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK);
+       /* Save the control regs that govern PER */
+       __ctl_store(kcb->kprobe_saved_ctl, 9, 11);
+}
+
+/* Called with kretprobe_lock held */
+void __kprobes arch_prepare_kretprobe(struct kretprobe *rp,
+                                       struct pt_regs *regs)
+{
+       struct kretprobe_instance *ri;
+
+       if ((ri = get_free_rp_inst(rp)) != NULL) {
+               ri->rp = rp;
+               ri->task = current;
+               ri->ret_addr = (kprobe_opcode_t *) regs->gprs[14];
+
+               /* Replace the return addr with trampoline addr */
+               regs->gprs[14] = (unsigned long)&kretprobe_trampoline;
+
+               add_rp_inst(ri);
+       } else {
+               rp->nmissed++;
+       }
+}
+
+static int __kprobes kprobe_handler(struct pt_regs *regs)
+{
+       struct kprobe *p;
+       int ret = 0;
+       unsigned long *addr = (unsigned long *)
+               ((regs->psw.addr & PSW_ADDR_INSN) - 2);
+       struct kprobe_ctlblk *kcb;
+
+       /*
+        * We don't want to be preempted for the entire
+        * duration of kprobe processing
+        */
+       preempt_disable();
+       kcb = get_kprobe_ctlblk();
+
+       /* Check we're not actually recursing */
+       if (kprobe_running()) {
+               p = get_kprobe(addr);
+               if (p) {
+                       if (kcb->kprobe_status == KPROBE_HIT_SS &&
+                           *p->ainsn.insn == BREAKPOINT_INSTRUCTION) {
+                               regs->psw.mask &= ~PSW_MASK_PER;
+                               regs->psw.mask |= kcb->kprobe_saved_imask;
+                               goto no_kprobe;
+                       }
+                       /* We have reentered the kprobe_handler(), since
+                        * another probe was hit while within the handler.
+                        * We here save the original kprobes variables and
+                        * just single step on the instruction of the new probe
+                        * without calling any user handlers.
+                        */
+                       save_previous_kprobe(kcb);
+                       set_current_kprobe(p, regs, kcb);
+                       kprobes_inc_nmissed_count(p);
+                       prepare_singlestep(p, regs);
+                       kcb->kprobe_status = KPROBE_REENTER;
+                       return 1;
+               } else {
+                       p = __get_cpu_var(current_kprobe);
+                       if (p->break_handler && p->break_handler(p, regs)) {
+                               goto ss_probe;
+                       }
+               }
+               goto no_kprobe;
+       }
+
+       p = get_kprobe(addr);
+       if (!p) {
+               if (*addr != BREAKPOINT_INSTRUCTION) {
+                       /*
+                        * The breakpoint instruction was removed right
+                        * after we hit it.  Another cpu has removed
+                        * either a probepoint or a debugger breakpoint
+                        * at this address.  In either case, no further
+                        * handling of this interrupt is appropriate.
+                        *
+                        */
+                       ret = 1;
+               }
+               /* Not one of ours: let kernel handle it */
+               goto no_kprobe;
+       }
+
+       kcb->kprobe_status = KPROBE_HIT_ACTIVE;
+       set_current_kprobe(p, regs, kcb);
+       if (p->pre_handler && p->pre_handler(p, regs))
+               /* handler has already set things up, so skip ss setup */
+               return 1;
+
+ss_probe:
+       prepare_singlestep(p, regs);
+       kcb->kprobe_status = KPROBE_HIT_SS;
+       return 1;
+
+no_kprobe:
+       preempt_enable_no_resched();
+       return ret;
+}
+
+/*
+ * Function return probe trampoline:
+ *     - init_kprobes() establishes a probepoint here
+ *     - When the probed function returns, this probe
+ *             causes the handlers to fire
+ */
+void __kprobes kretprobe_trampoline_holder(void)
+{
+       asm volatile(".global kretprobe_trampoline\n"
+                    "kretprobe_trampoline: bcr 0,0\n");
+}
+
+/*
+ * Called when the probe at kretprobe trampoline is hit
+ */
+int __kprobes trampoline_probe_handler(struct kprobe *p, struct pt_regs *regs)
+{
+       struct kretprobe_instance *ri = NULL;
+       struct hlist_head *head;
+       struct hlist_node *node, *tmp;
+       unsigned long flags, orig_ret_address = 0;
+       unsigned long trampoline_address = (unsigned long)&kretprobe_trampoline;
+
+       spin_lock_irqsave(&kretprobe_lock, flags);
+       head = kretprobe_inst_table_head(current);
+
+       /*
+        * It is possible to have multiple instances associated with a given
+        * task either because an multiple functions in the call path
+        * have a return probe installed on them, and/or more then one return
+        * return probe was registered for a target function.
+        *
+        * We can handle this because:
+        *     - instances are always inserted at the head of the list
+        *     - when multiple return probes are registered for the same
+        *       function, the first instance's ret_addr will point to the
+        *       real return address, and all the rest will point to
+        *       kretprobe_trampoline
+        */
+       hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
+               if (ri->task != current)
+                       /* another task is sharing our hash bucket */
+                       continue;
+
+               if (ri->rp && ri->rp->handler)
+                       ri->rp->handler(ri, regs);
+
+               orig_ret_address = (unsigned long)ri->ret_addr;
+               recycle_rp_inst(ri);
+
+               if (orig_ret_address != trampoline_address) {
+                       /*
+                        * This is the real return address. Any other
+                        * instances associated with this task are for
+                        * other calls deeper on the call stack
+                        */
+                       break;
+               }
+       }
+       BUG_ON(!orig_ret_address || (orig_ret_address == trampoline_address));
+       regs->psw.addr = orig_ret_address | PSW_ADDR_AMODE;
+
+       reset_current_kprobe();
+       spin_unlock_irqrestore(&kretprobe_lock, flags);
+       preempt_enable_no_resched();
+
+       /*
+        * By returning a non-zero value, we are telling
+        * kprobe_handler() that we don't want the post_handler
+        * to run (and have re-enabled preemption)
+        */
+       return 1;
+}
+
+/*
+ * Called after single-stepping.  p->addr is the address of the
+ * instruction whose first byte has been replaced by the "breakpoint"
+ * instruction.  To avoid the SMP problems that can occur when we
+ * temporarily put back the original opcode to single-step, we
+ * single-stepped a copy of the instruction.  The address of this
+ * copy is p->ainsn.insn.
+ */
+static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
+{
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+
+       regs->psw.addr &= PSW_ADDR_INSN;
+
+       if (p->ainsn.fixup & FIXUP_PSW_NORMAL)
+               regs->psw.addr = (unsigned long)p->addr +
+                               ((unsigned long)regs->psw.addr -
+                                (unsigned long)p->ainsn.insn);
+
+       if (p->ainsn.fixup & FIXUP_BRANCH_NOT_TAKEN)
+               if ((unsigned long)regs->psw.addr -
+                   (unsigned long)p->ainsn.insn == p->ainsn.ilen)
+                       regs->psw.addr = (unsigned long)p->addr + p->ainsn.ilen;
+
+       if (p->ainsn.fixup & FIXUP_RETURN_REGISTER)
+               regs->gprs[p->ainsn.reg] = ((unsigned long)p->addr +
+                                               (regs->gprs[p->ainsn.reg] -
+                                               (unsigned long)p->ainsn.insn))
+                                               | PSW_ADDR_AMODE;
+
+       regs->psw.addr |= PSW_ADDR_AMODE;
+       /* turn off PER mode */
+       regs->psw.mask &= ~PSW_MASK_PER;
+       /* Restore the original per control regs */
+       __ctl_load(kcb->kprobe_saved_ctl, 9, 11);
+       regs->psw.mask |= kcb->kprobe_saved_imask;
+}
+
+static int __kprobes post_kprobe_handler(struct pt_regs *regs)
+{
+       struct kprobe *cur = kprobe_running();
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+
+       if (!cur)
+               return 0;
+
+       if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) {
+               kcb->kprobe_status = KPROBE_HIT_SSDONE;
+               cur->post_handler(cur, regs, 0);
+       }
+
+       resume_execution(cur, regs);
+
+       /*Restore back the original saved kprobes variables and continue. */
+       if (kcb->kprobe_status == KPROBE_REENTER) {
+               restore_previous_kprobe(kcb);
+               goto out;
+       }
+       reset_current_kprobe();
+out:
+       preempt_enable_no_resched();
+
+       /*
+        * if somebody else is singlestepping across a probe point, psw mask
+        * will have PER set, in which case, continue the remaining processing
+        * of do_single_step, as if this is not a probe hit.
+        */
+       if (regs->psw.mask & PSW_MASK_PER) {
+               return 0;
+       }
+
+       return 1;
+}
+
+static int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
+{
+       struct kprobe *cur = kprobe_running();
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+       const struct exception_table_entry *entry;
+
+       switch(kcb->kprobe_status) {
+       case KPROBE_SWAP_INST:
+               /* We are here because the instruction replacement failed */
+               return 0;
+       case KPROBE_HIT_SS:
+       case KPROBE_REENTER:
+               /*
+                * We are here because the instruction being single
+                * stepped caused a page fault. We reset the current
+                * kprobe and the nip points back to the probe address
+                * and allow the page fault handler to continue as a
+                * normal page fault.
+                */
+               regs->psw.addr = (unsigned long)cur->addr | PSW_ADDR_AMODE;
+               regs->psw.mask &= ~PSW_MASK_PER;
+               regs->psw.mask |= kcb->kprobe_saved_imask;
+               if (kcb->kprobe_status == KPROBE_REENTER)
+                       restore_previous_kprobe(kcb);
+               else
+                       reset_current_kprobe();
+               preempt_enable_no_resched();
+               break;
+       case KPROBE_HIT_ACTIVE:
+       case KPROBE_HIT_SSDONE:
+               /*
+                * We increment the nmissed count for accounting,
+                * we can also use npre/npostfault count for accouting
+                * these specific fault cases.
+                */
+               kprobes_inc_nmissed_count(cur);
+
+               /*
+                * We come here because instructions in the pre/post
+                * handler caused the page_fault, this could happen
+                * if handler tries to access user space by
+                * copy_from_user(), get_user() etc. Let the
+                * user-specified handler try to fix it first.
+                */
+               if (cur->fault_handler && cur->fault_handler(cur, regs, trapnr))
+                       return 1;
+
+               /*
+                * In case the user-specified fault handler returned
+                * zero, try to fix up.
+                */
+               entry = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN);
+               if (entry) {
+                       regs->psw.addr = entry->fixup | PSW_ADDR_AMODE;
+                       return 1;
+               }
+
+               /*
+                * fixup_exception() could not handle it,
+                * Let do_page_fault() fix it.
+                */
+               break;
+       default:
+               break;
+       }
+       return 0;
+}
+
+/*
+ * Wrapper routine to for handling exceptions.
+ */
+int __kprobes kprobe_exceptions_notify(struct notifier_block *self,
+                                      unsigned long val, void *data)
+{
+       struct die_args *args = (struct die_args *)data;
+       int ret = NOTIFY_DONE;
+
+       switch (val) {
+       case DIE_BPT:
+               if (kprobe_handler(args->regs))
+                       ret = NOTIFY_STOP;
+               break;
+       case DIE_SSTEP:
+               if (post_kprobe_handler(args->regs))
+                       ret = NOTIFY_STOP;
+               break;
+       case DIE_TRAP:
+       case DIE_PAGE_FAULT:
+               /* kprobe_running() needs smp_processor_id() */
+               preempt_disable();
+               if (kprobe_running() &&
+                   kprobe_fault_handler(args->regs, args->trapnr))
+                       ret = NOTIFY_STOP;
+               preempt_enable();
+               break;
+       default:
+               break;
+       }
+       return ret;
+}
+
+int __kprobes setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
+{
+       struct jprobe *jp = container_of(p, struct jprobe, kp);
+       unsigned long addr;
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+
+       memcpy(&kcb->jprobe_saved_regs, regs, sizeof(struct pt_regs));
+
+       /* setup return addr to the jprobe handler routine */
+       regs->psw.addr = (unsigned long)(jp->entry) | PSW_ADDR_AMODE;
+
+       /* r14 is the function return address */
+       kcb->jprobe_saved_r14 = (unsigned long)regs->gprs[14];
+       /* r15 is the stack pointer */
+       kcb->jprobe_saved_r15 = (unsigned long)regs->gprs[15];
+       addr = (unsigned long)kcb->jprobe_saved_r15;
+
+       memcpy(kcb->jprobes_stack, (kprobe_opcode_t *) addr,
+              MIN_STACK_SIZE(addr));
+       return 1;
+}
+
+void __kprobes jprobe_return(void)
+{
+       asm volatile(".word 0x0002");
+}
+
+void __kprobes jprobe_return_end(void)
+{
+       asm volatile("bcr 0,0");
+}
+
+int __kprobes longjmp_break_handler(struct kprobe *p, struct pt_regs *regs)
+{
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+       unsigned long stack_addr = (unsigned long)(kcb->jprobe_saved_r15);
+
+       /* Put the regs back */
+       memcpy(regs, &kcb->jprobe_saved_regs, sizeof(struct pt_regs));
+       /* put the stack back */
+       memcpy((kprobe_opcode_t *) stack_addr, kcb->jprobes_stack,
+              MIN_STACK_SIZE(stack_addr));
+       preempt_enable_no_resched();
+       return 1;
+}
+
+static struct kprobe trampoline_p = {
+       .addr = (kprobe_opcode_t *) & kretprobe_trampoline,
+       .pre_handler = trampoline_probe_handler
+};
+
+int __init arch_init_kprobes(void)
+{
+       return register_kprobe(&trampoline_p);
+}
index bde1d1d598586cc483bcb84ca0211008d3a70711..c4982c963424921aebaaad4fa15560b05ca22bd5 100644 (file)
@@ -29,6 +29,7 @@
 #include <linux/module.h>
 #include <linux/kallsyms.h>
 #include <linux/reboot.h>
+#include <linux/kprobes.h>
 
 #include <asm/system.h>
 #include <asm/uaccess.h>
@@ -39,6 +40,7 @@
 #include <asm/s390_ext.h>
 #include <asm/lowcore.h>
 #include <asm/debug.h>
+#include <asm/kdebug.h>
 
 /* Called from entry.S only */
 extern void handle_per_exception(struct pt_regs *regs);
@@ -74,6 +76,20 @@ static int kstack_depth_to_print = 12;
 static int kstack_depth_to_print = 20;
 #endif /* CONFIG_64BIT */
 
+ATOMIC_NOTIFIER_HEAD(s390die_chain);
+
+int register_die_notifier(struct notifier_block *nb)
+{
+       return atomic_notifier_chain_register(&s390die_chain, nb);
+}
+EXPORT_SYMBOL(register_die_notifier);
+
+int unregister_die_notifier(struct notifier_block *nb)
+{
+       return atomic_notifier_chain_unregister(&s390die_chain, nb);
+}
+EXPORT_SYMBOL(unregister_die_notifier);
+
 /*
  * For show_trace we have tree different stack to consider:
  *   - the panic stack which is used if the kernel stack has overflown
@@ -305,8 +321,9 @@ report_user_fault(long interruption_code, struct pt_regs *regs)
 #endif
 }
 
-static void inline do_trap(long interruption_code, int signr, char *str,
-                           struct pt_regs *regs, siginfo_t *info)
+static void __kprobes inline do_trap(long interruption_code, int signr,
+                                       char *str, struct pt_regs *regs,
+                                       siginfo_t *info)
 {
        /*
         * We got all needed information from the lowcore and can
@@ -315,6 +332,10 @@ static void inline do_trap(long interruption_code, int signr, char *str,
         if (regs->psw.mask & PSW_MASK_PSTATE)
                local_irq_enable();
 
+       if (notify_die(DIE_TRAP, str, regs, interruption_code,
+                               interruption_code, signr) == NOTIFY_STOP)
+               return;
+
         if (regs->psw.mask & PSW_MASK_PSTATE) {
                 struct task_struct *tsk = current;
 
@@ -336,8 +357,12 @@ static inline void __user *get_check_address(struct pt_regs *regs)
        return (void __user *)((regs->psw.addr-S390_lowcore.pgm_ilc) & PSW_ADDR_INSN);
 }
 
-void do_single_step(struct pt_regs *regs)
+void __kprobes do_single_step(struct pt_regs *regs)
 {
+       if (notify_die(DIE_SSTEP, "sstep", regs, 0, 0,
+                                       SIGTRAP) == NOTIFY_STOP){
+               return;
+       }
        if ((current->ptrace & PT_PTRACED) != 0)
                force_sig(SIGTRAP, current);
 }
index ff5f7bb34f75bc68f26a7f54c60167e9bcf3cf9a..df0c16ab8e9200f23c09c97fd9531bd7f01bbebc 100644 (file)
@@ -24,6 +24,7 @@ SECTIONS
        *(.text)
        SCHED_TEXT
        LOCK_TEXT
+       KPROBES_TEXT
        *(.fixup)
        *(.gnu.warning)
        } = 0x0700
index 7cd82575813d1c7be34a85b109f75e1ff787b482..44f0cda7e72e44bb9ae47e6fc42fd284ceb430ab 100644 (file)
 #include <linux/console.h>
 #include <linux/module.h>
 #include <linux/hardirq.h>
+#include <linux/kprobes.h>
 
 #include <asm/system.h>
 #include <asm/uaccess.h>
 #include <asm/pgtable.h>
+#include <asm/kdebug.h>
 
 #ifndef CONFIG_64BIT
 #define __FAIL_ADDR_MASK 0x7ffff000
@@ -48,6 +50,38 @@ extern int sysctl_userprocess_debug;
 
 extern void die(const char *,struct pt_regs *,long);
 
+#ifdef CONFIG_KPROBES
+ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
+int register_page_fault_notifier(struct notifier_block *nb)
+{
+       return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
+}
+
+int unregister_page_fault_notifier(struct notifier_block *nb)
+{
+       return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
+}
+
+static inline int notify_page_fault(enum die_val val, const char *str,
+                       struct pt_regs *regs, long err, int trap, int sig)
+{
+       struct die_args args = {
+               .regs = regs,
+               .str = str,
+               .err = err,
+               .trapnr = trap,
+               .signr = sig
+       };
+       return atomic_notifier_call_chain(&notify_page_fault_chain, val, &args);
+}
+#else
+static inline int notify_page_fault(enum die_val val, const char *str,
+                       struct pt_regs *regs, long err, int trap, int sig)
+{
+       return NOTIFY_DONE;
+}
+#endif
+
 extern spinlock_t timerlist_lock;
 
 /*
@@ -159,7 +193,7 @@ static void do_sigsegv(struct pt_regs *regs, unsigned long error_code,
  *   11       Page translation     ->  Not present       (nullification)
  *   3b       Region third trans.  ->  Not present       (nullification)
  */
-static inline void
+static inline void __kprobes
 do_exception(struct pt_regs *regs, unsigned long error_code, int is_protection)
 {
         struct task_struct *tsk;
@@ -173,6 +207,10 @@ do_exception(struct pt_regs *regs, unsigned long error_code, int is_protection)
         tsk = current;
         mm = tsk->mm;
        
+       if (notify_page_fault(DIE_PAGE_FAULT, "page fault", regs, error_code, 14,
+                                       SIGSEGV) == NOTIFY_STOP)
+               return;
+
        /* 
          * Check for low-address protection.  This needs to be treated
         * as a special case because the translation exception code 
diff --git a/include/asm-s390/kdebug.h b/include/asm-s390/kdebug.h
new file mode 100644 (file)
index 0000000..40cc680
--- /dev/null
@@ -0,0 +1,59 @@
+#ifndef _S390_KDEBUG_H
+#define _S390_KDEBUG_H
+
+/*
+ * Feb 2006 Ported to s390 <grundym@us.ibm.com>
+ */
+#include <linux/notifier.h>
+
+struct pt_regs;
+
+struct die_args {
+       struct pt_regs *regs;
+       const char *str;
+       long err;
+       int trapnr;
+       int signr;
+};
+
+/* Note - you should never unregister because that can race with NMIs.
+ * If you really want to do it first unregister - then synchronize_sched
+ *  - then free.
+ */
+extern int register_die_notifier(struct notifier_block *);
+extern int unregister_die_notifier(struct notifier_block *);
+extern int register_page_fault_notifier(struct notifier_block *);
+extern int unregister_page_fault_notifier(struct notifier_block *);
+extern struct atomic_notifier_head s390die_chain;
+
+
+enum die_val {
+       DIE_OOPS = 1,
+       DIE_BPT,
+       DIE_SSTEP,
+       DIE_PANIC,
+       DIE_NMI,
+       DIE_DIE,
+       DIE_NMIWATCHDOG,
+       DIE_KERNELDEBUG,
+       DIE_TRAP,
+       DIE_GPF,
+       DIE_CALL,
+       DIE_NMI_IPI,
+       DIE_PAGE_FAULT,
+};
+
+static inline int notify_die(enum die_val val, const char *str,
+                       struct pt_regs *regs, long err, int trap, int sig)
+{
+       struct die_args args = {
+               .regs = regs,
+               .str = str,
+               .err = err,
+               .trapnr = trap,
+               .signr = sig
+       };
+       return atomic_notifier_call_chain(&s390die_chain, val, &args);
+}
+
+#endif
diff --git a/include/asm-s390/kprobes.h b/include/asm-s390/kprobes.h
new file mode 100644 (file)
index 0000000..b847ff0
--- /dev/null
@@ -0,0 +1,114 @@
+#ifndef _ASM_S390_KPROBES_H
+#define _ASM_S390_KPROBES_H
+/*
+ *  Kernel Probes (KProbes)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright (C) IBM Corporation, 2002, 2006
+ *
+ * 2002-Oct    Created by Vamsi Krishna S <vamsi_krishna@in.ibm.com> Kernel
+ *             Probes initial implementation ( includes suggestions from
+ *             Rusty Russell).
+ * 2004-Nov    Modified for PPC64 by Ananth N Mavinakayanahalli
+ *             <ananth@in.ibm.com>
+ * 2005-Dec    Used as a template for s390 by Mike Grundy
+ *             <grundym@us.ibm.com>
+ */
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/percpu.h>
+
+#define  __ARCH_WANT_KPROBES_INSN_SLOT
+struct pt_regs;
+struct kprobe;
+
+typedef u16 kprobe_opcode_t;
+#define BREAKPOINT_INSTRUCTION 0x0002
+
+/* Maximum instruction size is 3 (16bit) halfwords: */
+#define MAX_INSN_SIZE          0x0003
+#define MAX_STACK_SIZE         64
+#define MIN_STACK_SIZE(ADDR) (((MAX_STACK_SIZE) < \
+       (((unsigned long)current_thread_info()) + THREAD_SIZE - (ADDR))) \
+       ? (MAX_STACK_SIZE) \
+       : (((unsigned long)current_thread_info()) + THREAD_SIZE - (ADDR)))
+
+#define JPROBE_ENTRY(pentry) (kprobe_opcode_t *)(pentry)
+
+#define ARCH_SUPPORTS_KRETPROBES
+#define ARCH_INACTIVE_KPROBE_COUNT 0
+
+#define KPROBE_SWAP_INST       0x10
+
+#define FIXUP_PSW_NORMAL       0x08
+#define FIXUP_BRANCH_NOT_TAKEN 0x04
+#define FIXUP_RETURN_REGISTER  0x02
+#define FIXUP_NOT_REQUIRED     0x01
+
+/* Architecture specific copy of original instruction */
+struct arch_specific_insn {
+       /* copy of original instruction */
+       kprobe_opcode_t *insn;
+       int fixup;
+       int ilen;
+       int reg;
+};
+
+struct ins_replace_args {
+       kprobe_opcode_t *ptr;
+       kprobe_opcode_t old;
+       kprobe_opcode_t new;
+};
+struct prev_kprobe {
+       struct kprobe *kp;
+       unsigned long status;
+       unsigned long saved_psw;
+       unsigned long kprobe_saved_imask;
+       unsigned long kprobe_saved_ctl[3];
+};
+
+/* per-cpu kprobe control block */
+struct kprobe_ctlblk {
+       unsigned long kprobe_status;
+       unsigned long kprobe_saved_imask;
+       unsigned long kprobe_saved_ctl[3];
+       struct pt_regs jprobe_saved_regs;
+       unsigned long jprobe_saved_r14;
+       unsigned long jprobe_saved_r15;
+       struct prev_kprobe prev_kprobe;
+       kprobe_opcode_t jprobes_stack[MAX_STACK_SIZE];
+};
+
+void arch_remove_kprobe(struct kprobe *p);
+void kretprobe_trampoline(void);
+int  is_prohibited_opcode(kprobe_opcode_t *instruction);
+void get_instruction_type(struct arch_specific_insn *ainsn);
+
+#define flush_insn_slot(p)     do { } while (0)
+
+#endif /* _ASM_S390_KPROBES_H */
+
+#ifdef CONFIG_KPROBES
+
+extern int kprobe_exceptions_notify(struct notifier_block *self,
+                                       unsigned long val, void *data);
+#else  /* !CONFIG_KPROBES */
+static inline int kprobe_exceptions_notify(struct notifier_block *self,
+                                               unsigned long val, void *data)
+{
+       return 0;
+}
+#endif