[IA64] kprobe opcode 16 bytes alignment on IA64

On IA64 instruction opcode must be 16 bytes alignment, in kprobe structure
there is one element to save original instruction, currently saved opcode
is not statically allocated in kprobe structure, that can not assure
16 bytes alignment. This patch dynamically allocated kprobe instruction
opcode to assure 16 bytes alignment.

Signed-off-by: bibo mao <[e-mail]>
Acked-by: Anil S Keshavamurthy <[e-mail]>
Signed-off-by: Tony Luck <[e-mail]>

arch/ia64/kernel/kprobes.c

···
 static int __kprobes unsupported_inst(uint template, uint  slot,
 				      uint major_opcode,
 				      unsigned long kprobe_inst,
-				      struct kprobe *p)
+				      unsigned long addr)
 {
-	unsigned long addr = (unsigned long)p->addr;
-
 	if (bundle_encoding[template][slot] == I) {
 		switch (major_opcode) {
 			case 0x0: //I_UNIT_MISC_OPCODE:
···
 					 struct kprobe *p)
 {
 	unsigned long break_inst = BREAK_INST;
-	bundle_t *bundle = &p->ainsn.insn.bundle;
+	bundle_t *bundle = &p->opcode.bundle;
 
 	/*
 	 * Copy the original kprobe_inst qualifying predicate(qp)
···
 	unsigned long *kprobe_addr = (unsigned long *)(addr & ~0xFULL);
 	unsigned long kprobe_inst=0;
 	unsigned int slot = addr & 0xf, template, major_opcode = 0;
-	bundle_t *bundle = &p->ainsn.insn.bundle;
+	bundle_t *bundle;
 
-	memcpy(&p->opcode.bundle, kprobe_addr, sizeof(bundle_t));
-	memcpy(&p->ainsn.insn.bundle, kprobe_addr, sizeof(bundle_t));
-
+	bundle = &((kprobe_opcode_t *)kprobe_addr)->bundle;
  	template = bundle->quad0.template;
 
 	if(valid_kprobe_addr(template, slot, addr))
···
 	/* Get kprobe_inst and major_opcode from the bundle */
 	get_kprobe_inst(bundle, slot, &kprobe_inst, &major_opcode);
 
-	if (unsupported_inst(template, slot, major_opcode, kprobe_inst, p))
+	if (unsupported_inst(template, slot, major_opcode, kprobe_inst, addr))
 			return -EINVAL;
+
+
+	p->ainsn.insn = get_insn_slot();
+	if (!p->ainsn.insn)
+		return -ENOMEM;
+	memcpy(&p->opcode, kprobe_addr, sizeof(kprobe_opcode_t));
+	memcpy(p->ainsn.insn, kprobe_addr, sizeof(kprobe_opcode_t));
 
 	prepare_break_inst(template, slot, major_opcode, kprobe_inst, p);
 
 	return 0;
-}
-
-void __kprobes flush_insn_slot(struct kprobe *p)
-{
-	unsigned long arm_addr;
-
-	arm_addr = ((unsigned long)&p->opcode.bundle) & ~0xFULL;
-	flush_icache_range(arm_addr, arm_addr + sizeof(bundle_t));
 }
 
 void __kprobes arch_arm_kprobe(struct kprobe *p)
···
 	unsigned long addr = (unsigned long)p->addr;
 	unsigned long arm_addr = addr & ~0xFULL;
 
-	flush_insn_slot(p);
-	memcpy((char *)arm_addr, &p->ainsn.insn.bundle, sizeof(bundle_t));
-	flush_icache_range(arm_addr, arm_addr + sizeof(bundle_t));
+	flush_icache_range((unsigned long)p->ainsn.insn,
+			(unsigned long)p->ainsn.insn + sizeof(kprobe_opcode_t));
+	memcpy((char *)arm_addr, &p->opcode, sizeof(kprobe_opcode_t));
+	flush_icache_range(arm_addr, arm_addr + sizeof(kprobe_opcode_t));
 }
 
 void __kprobes arch_disarm_kprobe(struct kprobe *p)
···
 	unsigned long addr = (unsigned long)p->addr;
 	unsigned long arm_addr = addr & ~0xFULL;
 
-	/* p->opcode contains the original unaltered bundle */
-	memcpy((char *) arm_addr, (char *) &p->opcode.bundle, sizeof(bundle_t));
-	flush_icache_range(arm_addr, arm_addr + sizeof(bundle_t));
+	/* p->ainsn.insn contains the original unaltered kprobe_opcode_t */
+	memcpy((char *) arm_addr, (char *) p->ainsn.insn,
+					 sizeof(kprobe_opcode_t));
+	flush_icache_range(arm_addr, arm_addr + sizeof(kprobe_opcode_t));
 }
 
+void __kprobes arch_remove_kprobe(struct kprobe *p)
+{
+	mutex_lock(&kprobe_mutex);
+	free_insn_slot(p->ainsn.insn);
+	mutex_unlock(&kprobe_mutex);
+}
 /*
  * We are resuming execution after a single step fault, so the pt_regs
  * structure reflects the register state after we executed the instruction
···
  */
 static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
 {
-  	unsigned long bundle_addr = ((unsigned long) (&p->opcode.bundle)) & ~0xFULL;
+  	unsigned long bundle_addr = (unsigned long) (&p->ainsn.insn->bundle);
   	unsigned long resume_addr = (unsigned long)p->addr & ~0xFULL;
  	unsigned long template;
  	int slot = ((unsigned long)p->addr & 0xf);
 
-	template = p->opcode.bundle.quad0.template;
+	template = p->ainsn.insn->bundle.quad0.template;
 
  	if (slot == 1 && bundle_encoding[template][1] == L)
  		slot = 2;
···
 
 static void __kprobes prepare_ss(struct kprobe *p, struct pt_regs *regs)
 {
-	unsigned long bundle_addr = (unsigned long) &p->opcode.bundle;
+	unsigned long bundle_addr = (unsigned long) &p->ainsn.insn->bundle;
 	unsigned long slot = (unsigned long)p->addr & 0xf;
 
 	/* single step inline if break instruction */

include/asm-ia64/kprobes.h

···
 #include <linux/percpu.h>
 #include <asm/break.h>
 
-#define MAX_INSN_SIZE   16
+#define __ARCH_WANT_KPROBES_INSN_SLOT
+#define MAX_INSN_SIZE   1
 #define BREAK_INST	(long)(__IA64_BREAK_KPROBE << 6)
 
 typedef union cmp_inst {
···
 #define IP_RELATIVE_PREDICT_OPCODE	(7)
 #define LONG_BRANCH_OPCODE		(0xC)
 #define LONG_CALL_OPCODE		(0xD)
-#define arch_remove_kprobe(p)		do {} while (0)
+#define flush_insn_slot(p)		do { } while (0)
 
 typedef struct kprobe_opcode {
 	bundle_t bundle;
···
 /* Architecture specific copy of original instruction*/
 struct arch_specific_insn {
 	/* copy of the instruction to be emulated */
-	kprobe_opcode_t insn;
+	kprobe_opcode_t *insn;
  #define INST_FLAG_FIX_RELATIVE_IP_ADDR		1
  #define INST_FLAG_FIX_BRANCH_REG		2
  #define INST_FLAG_BREAK_INST			4
···
 }
 extern void invalidate_stacked_regs(void);
 extern void flush_register_stack(void);
-extern void flush_insn_slot(struct kprobe *p);
+extern void arch_remove_kprobe(struct kprobe *p);
 
 #endif				/* _ASM_KPROBES_H */