[PATCH] Kprobes: Track kprobe on a per_cpu basis - x86_64 changes
authorAnanth N Mavinakayanahalli <ananth@in.ibm.com>
Mon, 7 Nov 2005 09:00:12 +0000 (01:00 -0800)
committerLinus Torvalds <torvalds@g5.osdl.org>
Mon, 7 Nov 2005 15:53:46 +0000 (07:53 -0800)
x86_64 changes to track kprobe execution on a per-cpu basis.  We now track the
kprobe state machine independently on each cpu using a arch specific kprobe
control block.

Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
arch/x86_64/kernel/kprobes.c
include/asm-x86_64/kprobes.h

index ebfa2c9..6cb40d1 100644 (file)
 #include <asm/kdebug.h>
 
 static DECLARE_MUTEX(kprobe_mutex);
-
-static struct kprobe *current_kprobe;
-static unsigned long kprobe_status, kprobe_old_rflags, kprobe_saved_rflags;
-static struct kprobe *kprobe_prev;
-static unsigned long kprobe_status_prev, kprobe_old_rflags_prev, kprobe_saved_rflags_prev;
-static struct pt_regs jprobe_saved_regs;
-static long *jprobe_saved_rsp;
 void jprobe_return_end(void);
 
-/* copy of the kernel stack at the probe fire time */
-static kprobe_opcode_t jprobes_stack[MAX_STACK_SIZE];
+DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
+DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);
 
 /*
  * returns non-zero if opcode modifies the interrupt flag.
@@ -236,29 +229,30 @@ void __kprobes arch_remove_kprobe(struct kprobe *p)
        up(&kprobe_mutex);
 }
 
-static inline void save_previous_kprobe(void)
+static inline void save_previous_kprobe(struct kprobe_ctlblk *kcb)
 {
-       kprobe_prev = current_kprobe;
-       kprobe_status_prev = kprobe_status;
-       kprobe_old_rflags_prev = kprobe_old_rflags;
-       kprobe_saved_rflags_prev = kprobe_saved_rflags;
+       kcb->prev_kprobe.kp = kprobe_running();
+       kcb->prev_kprobe.status = kcb->kprobe_status;
+       kcb->prev_kprobe.old_rflags = kcb->kprobe_old_rflags;
+       kcb->prev_kprobe.saved_rflags = kcb->kprobe_saved_rflags;
 }
 
-static inline void restore_previous_kprobe(void)
+static inline void restore_previous_kprobe(struct kprobe_ctlblk *kcb)
 {
-       current_kprobe = kprobe_prev;
-       kprobe_status = kprobe_status_prev;
-       kprobe_old_rflags = kprobe_old_rflags_prev;
-       kprobe_saved_rflags = kprobe_saved_rflags_prev;
+       __get_cpu_var(current_kprobe) = kcb->prev_kprobe.kp;
+       kcb->kprobe_status = kcb->prev_kprobe.status;
+       kcb->kprobe_old_rflags = kcb->prev_kprobe.old_rflags;
+       kcb->kprobe_saved_rflags = kcb->prev_kprobe.saved_rflags;
 }
 
-static inline void set_current_kprobe(struct kprobe *p, struct pt_regs *regs)
+static inline void set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
+                               struct kprobe_ctlblk *kcb)
 {
-       current_kprobe = p;
-       kprobe_saved_rflags = kprobe_old_rflags
+       __get_cpu_var(current_kprobe) = p;
+       kcb->kprobe_saved_rflags = kcb->kprobe_old_rflags
                = (regs->eflags & (TF_MASK | IF_MASK));
        if (is_IF_modifier(p->ainsn.insn))
-               kprobe_saved_rflags &= ~IF_MASK;
+               kcb->kprobe_saved_rflags &= ~IF_MASK;
 }
 
 static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
@@ -301,6 +295,7 @@ int __kprobes kprobe_handler(struct pt_regs *regs)
        struct kprobe *p;
        int ret = 0;
        kprobe_opcode_t *addr = (kprobe_opcode_t *)(regs->rip - sizeof(kprobe_opcode_t));
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
 
        /* Check we're not actually recursing */
        if (kprobe_running()) {
@@ -308,13 +303,13 @@ int __kprobes kprobe_handler(struct pt_regs *regs)
                   Disarm the probe we just hit, and ignore it. */
                p = get_kprobe(addr);
                if (p) {
-                       if (kprobe_status == KPROBE_HIT_SS &&
+                       if (kcb->kprobe_status == KPROBE_HIT_SS &&
                                *p->ainsn.insn == BREAKPOINT_INSTRUCTION) {
                                regs->eflags &= ~TF_MASK;
-                               regs->eflags |= kprobe_saved_rflags;
+                               regs->eflags |= kcb->kprobe_saved_rflags;
                                unlock_kprobes();
                                goto no_kprobe;
-                       } else if (kprobe_status == KPROBE_HIT_SSDONE) {
+                       } else if (kcb->kprobe_status == KPROBE_HIT_SSDONE) {
                                /* TODO: Provide re-entrancy from
                                 * post_kprobes_handler() and avoid exception
                                 * stack corruption while single-stepping on
@@ -322,6 +317,7 @@ int __kprobes kprobe_handler(struct pt_regs *regs)
                                 */
                                arch_disarm_kprobe(p);
                                regs->rip = (unsigned long)p->addr;
+                               reset_current_kprobe();
                                ret = 1;
                        } else {
                                /* We have reentered the kprobe_handler(), since
@@ -331,15 +327,15 @@ int __kprobes kprobe_handler(struct pt_regs *regs)
                                 * of the new probe without calling any user
                                 * handlers.
                                 */
-                               save_previous_kprobe();
-                               set_current_kprobe(p, regs);
+                               save_previous_kprobe(kcb);
+                               set_current_kprobe(p, regs, kcb);
                                p->nmissed++;
                                prepare_singlestep(p, regs);
-                               kprobe_status = KPROBE_REENTER;
+                               kcb->kprobe_status = KPROBE_REENTER;
                                return 1;
                        }
                } else {
-                       p = current_kprobe;
+                       p = __get_cpu_var(current_kprobe);
                        if (p->break_handler && p->break_handler(p, regs)) {
                                goto ss_probe;
                        }
@@ -374,8 +370,8 @@ int __kprobes kprobe_handler(struct pt_regs *regs)
         * in post_kprobe_handler()
         */
        preempt_disable();
-       kprobe_status = KPROBE_HIT_ACTIVE;
-       set_current_kprobe(p, regs);
+       set_current_kprobe(p, regs, kcb);
+       kcb->kprobe_status = KPROBE_HIT_ACTIVE;
 
        if (p->pre_handler && p->pre_handler(p, regs))
                /* handler has already set things up, so skip ss setup */
@@ -383,7 +379,7 @@ int __kprobes kprobe_handler(struct pt_regs *regs)
 
 ss_probe:
        prepare_singlestep(p, regs);
-       kprobe_status = KPROBE_HIT_SS;
+       kcb->kprobe_status = KPROBE_HIT_SS;
        return 1;
 
 no_kprobe:
@@ -451,6 +447,7 @@ int __kprobes trampoline_probe_handler(struct kprobe *p, struct pt_regs *regs)
        BUG_ON(!orig_ret_address || (orig_ret_address == trampoline_address));
        regs->rip = orig_ret_address;
 
+       reset_current_kprobe();
        unlock_kprobes();
        preempt_enable_no_resched();
 
@@ -484,7 +481,8 @@ int __kprobes trampoline_probe_handler(struct kprobe *p, struct pt_regs *regs)
  * that is atop the stack is the address following the copied instruction.
  * We need to make it the address following the original instruction.
  */
-static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
+static void __kprobes resume_execution(struct kprobe *p,
+               struct pt_regs *regs, struct kprobe_ctlblk *kcb)
 {
        unsigned long *tos = (unsigned long *)regs->rsp;
        unsigned long next_rip = 0;
@@ -499,7 +497,7 @@ static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
        switch (*insn) {
        case 0x9c:              /* pushfl */
                *tos &= ~(TF_MASK | IF_MASK);
-               *tos |= kprobe_old_rflags;
+               *tos |= kcb->kprobe_old_rflags;
                break;
        case 0xc3:              /* ret/lret */
        case 0xcb:
@@ -544,24 +542,28 @@ static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
  */
 int __kprobes post_kprobe_handler(struct pt_regs *regs)
 {
-       if (!kprobe_running())
+       struct kprobe *cur = kprobe_running();
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+
+       if (!cur)
                return 0;
 
-       if ((kprobe_status != KPROBE_REENTER) && current_kprobe->post_handler) {
-               kprobe_status = KPROBE_HIT_SSDONE;
-               current_kprobe->post_handler(current_kprobe, regs, 0);
+       if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) {
+               kcb->kprobe_status = KPROBE_HIT_SSDONE;
+               cur->post_handler(cur, regs, 0);
        }
 
-       resume_execution(current_kprobe, regs);
-       regs->eflags |= kprobe_saved_rflags;
+       resume_execution(cur, regs, kcb);
+       regs->eflags |= kcb->kprobe_saved_rflags;
 
        /* Restore the original saved kprobes variables and continue. */
-       if (kprobe_status == KPROBE_REENTER) {
-               restore_previous_kprobe();
+       if (kcb->kprobe_status == KPROBE_REENTER) {
+               restore_previous_kprobe(kcb);
                goto out;
        } else {
                unlock_kprobes();
        }
+       reset_current_kprobe();
 out:
        preempt_enable_no_resched();
 
@@ -579,14 +581,17 @@ out:
 /* Interrupts disabled, kprobe_lock held. */
 int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr)
 {
-       if (current_kprobe->fault_handler
-           && current_kprobe->fault_handler(current_kprobe, regs, trapnr))
+       struct kprobe *cur = kprobe_running();
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+
+       if (cur->fault_handler && cur->fault_handler(cur, regs, trapnr))
                return 1;
 
-       if (kprobe_status & KPROBE_HIT_SS) {
-               resume_execution(current_kprobe, regs);
-               regs->eflags |= kprobe_old_rflags;
+       if (kcb->kprobe_status & KPROBE_HIT_SS) {
+               resume_execution(cur, regs, kcb);
+               regs->eflags |= kcb->kprobe_old_rflags;
 
+               reset_current_kprobe();
                unlock_kprobes();
                preempt_enable_no_resched();
        }
@@ -629,10 +634,11 @@ 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();
 
-       jprobe_saved_regs = *regs;
-       jprobe_saved_rsp = (long *) regs->rsp;
-       addr = (unsigned long)jprobe_saved_rsp;
+       kcb->jprobe_saved_regs = *regs;
+       kcb->jprobe_saved_rsp = (long *) regs->rsp;
+       addr = (unsigned long)(kcb->jprobe_saved_rsp);
        /*
         * As Linus pointed out, gcc assumes that the callee
         * owns the argument space and could overwrite it, e.g.
@@ -640,7 +646,8 @@ int __kprobes setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
         * we also save and restore enough stack bytes to cover
         * the argument area.
         */
-       memcpy(jprobes_stack, (kprobe_opcode_t *) addr, MIN_STACK_SIZE(addr));
+       memcpy(kcb->jprobes_stack, (kprobe_opcode_t *)addr,
+                       MIN_STACK_SIZE(addr));
        regs->eflags &= ~IF_MASK;
        regs->rip = (unsigned long)(jp->entry);
        return 1;
@@ -648,34 +655,38 @@ int __kprobes setjmp_pre_handler(struct kprobe *p, struct pt_regs *regs)
 
 void __kprobes jprobe_return(void)
 {
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
+
        asm volatile ("       xchg   %%rbx,%%rsp     \n"
                      "       int3                      \n"
                      "       .globl jprobe_return_end  \n"
                      "       jprobe_return_end:        \n"
                      "       nop                       \n"::"b"
-                     (jprobe_saved_rsp):"memory");
+                     (kcb->jprobe_saved_rsp):"memory");
 }
 
 int __kprobes longjmp_break_handler(struct kprobe *p, struct pt_regs *regs)
 {
+       struct kprobe_ctlblk *kcb = get_kprobe_ctlblk();
        u8 *addr = (u8 *) (regs->rip - 1);
-       unsigned long stack_addr = (unsigned long)jprobe_saved_rsp;
+       unsigned long stack_addr = (unsigned long)(kcb->jprobe_saved_rsp);
        struct jprobe *jp = container_of(p, struct jprobe, kp);
 
        if ((addr > (u8 *) jprobe_return) && (addr < (u8 *) jprobe_return_end)) {
-               if ((long *)regs->rsp != jprobe_saved_rsp) {
+               if ((long *)regs->rsp != kcb->jprobe_saved_rsp) {
                        struct pt_regs *saved_regs =
-                           container_of(jprobe_saved_rsp, struct pt_regs, rsp);
+                           container_of(kcb->jprobe_saved_rsp,
+                                           struct pt_regs, rsp);
                        printk("current rsp %p does not match saved rsp %p\n",
-                              (long *)regs->rsp, jprobe_saved_rsp);
+                              (long *)regs->rsp, kcb->jprobe_saved_rsp);
                        printk("Saved registers for jprobe %p\n", jp);
                        show_registers(saved_regs);
                        printk("Current registers\n");
                        show_registers(regs);
                        BUG();
                }
-               *regs = jprobe_saved_regs;
-               memcpy((kprobe_opcode_t *) stack_addr, jprobes_stack,
+               *regs = kcb->jprobe_saved_regs;
+               memcpy((kprobe_opcode_t *) stack_addr, kcb->jprobes_stack,
                       MIN_STACK_SIZE(stack_addr));
                return 1;
        }
index 6d6d883..4dd7a7e 100644 (file)
@@ -25,6 +25,7 @@
  */
 #include <linux/types.h>
 #include <linux/ptrace.h>
+#include <linux/percpu.h>
 
 struct pt_regs;
 
@@ -48,6 +49,24 @@ struct arch_specific_insn {
        kprobe_opcode_t *insn;
 };
 
+struct prev_kprobe {
+       struct kprobe *kp;
+       unsigned long status;
+       unsigned long old_rflags;
+       unsigned long saved_rflags;
+};
+
+/* per-cpu kprobe control block */
+struct kprobe_ctlblk {
+       unsigned long kprobe_status;
+       unsigned long kprobe_old_rflags;
+       unsigned long kprobe_saved_rflags;
+       long *jprobe_saved_rsp;
+       struct pt_regs jprobe_saved_regs;
+       kprobe_opcode_t jprobes_stack[MAX_STACK_SIZE];
+       struct prev_kprobe prev_kprobe;
+};
+
 /* trap3/1 are intr gates for kprobes.  So, restore the status of IF,
  * if necessary, before executing the original int3/1 (trap) handler.
  */