@@ -96,6 +96,7 @@ config KPROBES_ON_FTRACE
config UPROBES
def_bool n
+ depends on ARCH_SUPPORTS_UPROBES
help
Uprobes is the user-space counterpart to kprobes: they
enable instrumentation applications (such as 'perf probe')
@@ -147,6 +147,9 @@ config PGTABLE_LEVELS
default 4 if 64BIT
default 3
+config ARCH_SUPPORTS_UPROBES
+ def_bool y if SPARC64
+
source "init/Kconfig"
source "kernel/Kconfig.freezer"
@@ -213,6 +213,7 @@ CONFIG_SCHEDSTATS=y
# CONFIG_RCU_CPU_STALL_DETECTOR is not set
CONFIG_SYSCTL_SYSCALL_CHECK=y
CONFIG_BLK_DEV_IO_TRACE=y
+CONFIG_UPROBE_EVENTS=y
CONFIG_KEYS=y
CONFIG_CRYPTO_NULL=m
CONFIG_CRYPTO_TEST=m
@@ -10,6 +10,8 @@ enum die_val {
DIE_OOPS = 1,
DIE_DEBUG, /* ta 0x70 */
DIE_DEBUG_2, /* ta 0x71 */
+ DIE_BPT, /* ta 0x73 */
+ DIE_SSTEP, /* ta 0x74 */
DIE_DIE,
DIE_TRAP,
DIE_TRAP_TL1,
@@ -61,7 +61,10 @@ extern union global_cpu_snapshot global_cpu_snapshot[NR_CPUS];
#define force_successful_syscall_return() set_thread_noerror(1)
#define user_mode(regs) (!((regs)->tstate & TSTATE_PRIV))
#define instruction_pointer(regs) ((regs)->tpc)
-#define instruction_pointer_set(regs, val) ((regs)->tpc = (val))
+#define instruction_pointer_set(regs, val) do { \
+ (regs)->tpc = (val); \
+ (regs)->tnpc = (val)+4; \
+ } while (0)
#define user_stack_pointer(regs) ((regs)->u_regs[UREG_FP])
static inline int is_syscall_success(struct pt_regs *regs)
{
@@ -77,6 +80,34 @@ unsigned long profile_pc(struct pt_regs *);
#else
#define profile_pc(regs) instruction_pointer(regs)
#endif
+
+#define MAX_REG_OFFSET (offsetof(struct pt_regs, magic))
+
+extern int regs_query_register_offset(const char *name);
+
+/**
+ * regs_get_register() - get register value from its offset
+ * @regs: pt_regs from which register value is gotten
+ * @offset: offset number of the register.
+ *
+ * regs_get_register returns the value of a register whose
+ * offset from @regs. The @offset is the offset of the register
+ * in struct pt_regs. If @offset is bigger than MAX_REG_OFFSET,
+ * this returns 0.
+ */
+static inline unsigned long regs_get_register(struct pt_regs *regs,
+ unsigned long offset)
+{
+ if (unlikely(offset >= MAX_REG_OFFSET))
+ return 0;
+ return *(unsigned long *)((unsigned long)regs + offset);
+}
+
+/* Valid only for Kernel mode traps. */
+static inline unsigned long kernel_stack_pointer(struct pt_regs *regs)
+{
+ return regs->u_regs[UREG_I6];
+}
#else /* __ASSEMBLY__ */
#endif /* __ASSEMBLY__ */
#else /* (defined(__sparc__) && defined(__arch64__)) */
@@ -180,7 +180,7 @@ register struct thread_info *current_thread_info_reg asm("g6");
#define TIF_NEED_RESCHED 3 /* rescheduling necessary */
/* flag bit 4 is available */
#define TIF_UNALIGNED 5 /* allowed to do unaligned accesses */
-/* flag bit 6 is available */
+#define TIF_UPROBE 6 /* breakpointed or singlestepped */
#define TIF_32BIT 7 /* 32-bit binary */
#define TIF_NOHZ 8 /* in adaptive nohz mode */
#define TIF_SECCOMP 9 /* secure computing */
@@ -199,6 +199,7 @@ register struct thread_info *current_thread_info_reg asm("g6");
#define _TIF_SIGPENDING (1<<TIF_SIGPENDING)
#define _TIF_NEED_RESCHED (1<<TIF_NEED_RESCHED)
#define _TIF_UNALIGNED (1<<TIF_UNALIGNED)
+#define _TIF_UPROBE (1<<TIF_UPROBE)
#define _TIF_32BIT (1<<TIF_32BIT)
#define _TIF_NOHZ (1<<TIF_NOHZ)
#define _TIF_SECCOMP (1<<TIF_SECCOMP)
@@ -209,7 +210,8 @@ register struct thread_info *current_thread_info_reg asm("g6");
#define _TIF_USER_WORK_MASK ((0xff << TI_FLAG_WSAVED_SHIFT) | \
_TIF_DO_NOTIFY_RESUME_MASK | \
_TIF_NEED_RESCHED)
-#define _TIF_DO_NOTIFY_RESUME_MASK (_TIF_NOTIFY_RESUME | _TIF_SIGPENDING)
+#define _TIF_DO_NOTIFY_RESUME_MASK (_TIF_NOTIFY_RESUME | \
+ _TIF_SIGPENDING | _TIF_UPROBE)
#define is_32bit_task() (test_thread_flag(TIF_32BIT))
@@ -186,6 +186,12 @@
#define KPROBES_TRAP(lvl) TRAP_ARG(bad_trap, lvl)
#endif
+#ifdef CONFIG_UPROBES
+#define UPROBES_TRAP(lvl) TRAP_ARG(uprobe_trap, lvl)
+#else
+#define UPROBES_TRAP(lvl) TRAP_ARG(bad_trap, lvl)
+#endif
+
#ifdef CONFIG_KGDB
#define KGDB_TRAP(lvl) TRAP_IRQ(kgdb_trap, lvl)
#else
new file mode 100644
@@ -0,0 +1,59 @@
+#ifndef _ASM_UPROBES_H
+#define _ASM_UPROBES_H
+/*
+ * User-space Probes (UProbes) for sparc
+ *
+ * Copyright (C) 2013 Oracle, Inc.
+ *
+ * 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, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ * Jose E. Marchesi <jose.marchesi@oracle.com>
+ * Eric Saint Etienne <eric.saint.etienne@oracle.com>
+ */
+
+typedef u32 uprobe_opcode_t;
+
+#define MAX_UINSN_BYTES 4
+#define UPROBE_XOL_SLOT_BYTES (MAX_UINSN_BYTES * 2)
+
+#define UPROBE_SWBP_INSN_SIZE 4
+#define UPROBE_SWBP_INSN 0x91d02073 /* ta 0x73 */
+#define UPROBE_STP_INSN 0x91d02074 /* ta 0x74 */
+
+#define ANNUL_BIT (1 << 29)
+
+struct arch_uprobe {
+ union {
+ u8 insn[MAX_UINSN_BYTES];
+ u32 ixol;
+ };
+};
+
+struct arch_uprobe_task {
+ u32 saved_tpc;
+ u32 saved_tnpc;
+};
+
+struct task_struct;
+struct notifier_block;
+
+extern int arch_uprobe_analyze_insn(struct arch_uprobe *aup, struct mm_struct *mm, unsigned long addr);
+extern int arch_uprobe_pre_xol(struct arch_uprobe *aup, struct pt_regs *regs);
+extern int arch_uprobe_post_xol(struct arch_uprobe *aup, struct pt_regs *regs);
+extern bool arch_uprobe_xol_was_trapped(struct task_struct *tsk);
+extern int arch_uprobe_exception_notify(struct notifier_block *self, unsigned long val, void *data);
+extern void arch_uprobe_abort_xol(struct arch_uprobe *aup, struct pt_regs *regs);
+
+#endif /* _ASM_UPROBES_H */
@@ -116,4 +116,5 @@ obj-$(CONFIG_COMPAT) += $(audit--y)
pc--$(CONFIG_PERF_EVENTS) := perf_event.o
obj-$(CONFIG_SPARC64) += $(pc--y)
+obj-$(CONFIG_UPROBES) += uprobes.o
obj-$(CONFIG_SPARC64) += jump_label.o
@@ -46,6 +46,43 @@
/* #define ALLOW_INIT_TRACING */
+struct pt_regs_offset {
+ const char *name;
+ int offset;
+};
+
+#define REG_OFFSET_NAME(n, r) \
+ {.name = n, .offset = (PT_V9_##r)}
+#define REG_OFFSET_END {.name = NULL, .offset = 0}
+
+static const struct pt_regs_offset regoffset_table[] = {
+ REG_OFFSET_NAME("g0", G0),
+ REG_OFFSET_NAME("g1", G1),
+ REG_OFFSET_NAME("g2", G2),
+ REG_OFFSET_NAME("g3", G3),
+ REG_OFFSET_NAME("g4", G4),
+ REG_OFFSET_NAME("g5", G5),
+ REG_OFFSET_NAME("g6", G6),
+ REG_OFFSET_NAME("g7", G7),
+
+ REG_OFFSET_NAME("i0", I0),
+ REG_OFFSET_NAME("i1", I1),
+ REG_OFFSET_NAME("i2", I2),
+ REG_OFFSET_NAME("i3", I3),
+ REG_OFFSET_NAME("i4", I4),
+ REG_OFFSET_NAME("i5", I5),
+ REG_OFFSET_NAME("i6", I6),
+ REG_OFFSET_NAME("i7", I7),
+
+ REG_OFFSET_NAME("tstate", TSTATE),
+ REG_OFFSET_NAME("pc", TPC),
+ REG_OFFSET_NAME("npc", TNPC),
+ REG_OFFSET_NAME("y", Y),
+ REG_OFFSET_NAME("lr", I7),
+
+ REG_OFFSET_END,
+};
+
/*
* Called by kernel/ptrace.c when detaching..
*
@@ -1099,3 +1136,20 @@ asmlinkage void syscall_trace_leave(struct pt_regs *regs)
if (test_thread_flag(TIF_NOHZ))
user_enter();
}
+
+/**
+ * regs_query_register_offset() - query register offset from its name
+ * @name: the name of a register
+ *
+ * regs_query_register_offset() returns the offset of a register in struct
+ * pt_regs from its name. If the name is invalid, this returns -EINVAL;
+ */
+int regs_query_register_offset(const char *name)
+{
+ const struct pt_regs_offset *roff;
+
+ for (roff = regoffset_table; roff->name != NULL; roff++)
+ if (!strcmp(roff->name, name))
+ return roff->offset;
+ return -EINVAL;
+}
@@ -545,6 +545,8 @@ static void do_signal(struct pt_regs *regs, unsigned long orig_i0)
void do_notify_resume(struct pt_regs *regs, unsigned long orig_i0, unsigned long thread_info_flags)
{
user_exit();
+ if (thread_info_flags & _TIF_UPROBE)
+ uprobe_notify_resume(regs);
if (thread_info_flags & _TIF_SIGPENDING)
do_signal(regs, orig_i0);
if (thread_info_flags & _TIF_NOTIFY_RESUME) {
@@ -165,7 +165,7 @@ tl0_resv169: BTRAP(0x169) BTRAP(0x16a) BTRAP(0x16b) BTRAP(0x16c)
tl0_linux64: LINUX_64BIT_SYSCALL_TRAP
tl0_gsctx: TRAP(sparc64_get_context) TRAP(sparc64_set_context)
tl0_resv170: KPROBES_TRAP(0x170) KPROBES_TRAP(0x171) KGDB_TRAP(0x172)
-tl0_resv173: BTRAP(0x173) BTRAP(0x174) BTRAP(0x175) BTRAP(0x176) BTRAP(0x177)
+tl0_resv173: UPROBES_TRAP(0x173) UPROBES_TRAP(0x174) BTRAP(0x175) BTRAP(0x176) BTRAP(0x177)
tl0_resv178: BTRAP(0x178) BTRAP(0x179) BTRAP(0x17a) BTRAP(0x17b) BTRAP(0x17c)
tl0_resv17d: BTRAP(0x17d) BTRAP(0x17e) BTRAP(0x17f)
#define BTRAPS(x) BTRAP(x) BTRAP(x+1) BTRAP(x+2) BTRAP(x+3) BTRAP(x+4) BTRAP(x+5) BTRAP(x+6) BTRAP(x+7)
new file mode 100644
@@ -0,0 +1,331 @@
+/*
+ * User-space Probes (UProbes) for sparc
+ *
+ * Copyright (C) 2013 Oracle Inc.
+ *
+ * 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, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ * Jose E. Marchesi <jose.marchesi@oracle.com>
+ * Eric Saint Etienne <eric.saint.etienne@oracle.com>
+ */
+
+#include <linux/kernel.h>
+#include <linux/highmem.h>
+#include <linux/uprobes.h>
+#include <linux/uaccess.h>
+#include <linux/sched.h> /* For struct task_struct */
+#include <linux/kdebug.h>
+
+#include <asm/cacheflush.h>
+#include <asm/uaccess.h>
+
+/* Compute the address of the breakpoint instruction and return it.
+ *
+ * Note that uprobe_get_swbp_addr is defined as a weak symbol in
+ * kernel/events/uprobe.c.
+ */
+unsigned long uprobe_get_swbp_addr(struct pt_regs *regs)
+{
+ return instruction_pointer(regs);
+}
+
+static void copy_to_page(struct page *page, unsigned long vaddr,
+ const void *src, int len)
+{
+ void *kaddr = kmap_atomic(page);
+
+ memcpy(kaddr + (vaddr & ~PAGE_MASK), src, len);
+ kunmap_atomic(kaddr);
+}
+
+/* Fill in the xol area with the probed instruction followed by the
+ * single-step trap. Some fixups in the copied instruction are
+ * performed at this point.
+ *
+ * Note that uprobe_xol_copy is defined as a weak symbol in
+ * kernel/events/uprobe.c.
+ */
+void arch_uprobe_copy_ixol(struct page *page, unsigned long vaddr,
+ void *src, unsigned long len)
+{
+ const u32 stp_insn = UPROBE_STP_INSN;
+ u32 insn = *(u32 *) src;
+
+ /* Branches annulling their delay slot must be fixed to not do
+ * so. Clearing the annul bit on these instructions we can be
+ * sure the single-step breakpoint in the XOL slot will be
+ * executed.
+ */
+
+ u32 op = (insn >> 30) & 0x3;
+ u32 op2 = (insn >> 22) & 0x7;
+
+ if (op == 0 &&
+ (op2 == 1 || op2 == 2 || op2 == 3 || op2 == 5 || op2 == 6) &&
+ (insn & ANNUL_BIT) == ANNUL_BIT)
+ insn &= ~ANNUL_BIT;
+
+ copy_to_page(page, vaddr, &insn, len);
+ copy_to_page(page, vaddr+len, &stp_insn, 4);
+}
+
+
+/* Instruction analysis/validity.
+ *
+ * This function returns 0 on success or a -ve number on error.
+ */
+int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe,
+ struct mm_struct *mm, unsigned long addr)
+{
+ /* Any unsupported instruction? Then return -EINVAL */
+ return 0;
+}
+
+/* If INSN is a relative control transfer instruction, return the
+ * corrected branch destination value.
+ *
+ * Note that regs->tpc and regs->tnpc still hold the values of the
+ * program counters at the time of the single-step trap due to the
+ * execution of the UPROBE_STP_INSN at utask->xol_vaddr + 4.
+ *
+ */
+static unsigned long relbranch_fixup(u32 insn, struct uprobe_task *utask,
+ struct pt_regs *regs)
+{
+ /* Branch not taken, no mods necessary. */
+ if (regs->tnpc == regs->tpc + 0x4UL)
+ return utask->autask.saved_tnpc + 0x4UL;
+
+ /* The three cases are call, branch w/prediction,
+ * and traditional branch.
+ */
+ if ((insn & 0xc0000000) == 0x40000000 ||
+ (insn & 0xc1c00000) == 0x00400000 ||
+ (insn & 0xc1c00000) == 0x00800000) {
+ unsigned long real_pc = (unsigned long) utask->vaddr;
+ unsigned long ixol_addr = utask->xol_vaddr;
+
+ /* The instruction did all the work for us
+ * already, just apply the offset to the correct
+ * instruction location.
+ */
+ return (real_pc + (regs->tnpc - ixol_addr));
+ }
+
+ /* It is jmpl or some other absolute PC modification instruction,
+ * leave NPC as-is.
+ */
+ return regs->tnpc;
+}
+
+/* If INSN is an instruction which writes its PC location
+ * into a destination register, fix that up.
+ */
+static int retpc_fixup(struct pt_regs *regs, u32 insn,
+ unsigned long real_pc)
+{
+ unsigned long *slot = NULL;
+ int rc = 0;
+
+ /* Simplest case is 'call', which always uses %o7 */
+ if ((insn & 0xc0000000) == 0x40000000)
+ slot = ®s->u_regs[UREG_I7];
+
+ /* 'jmpl' encodes the register inside of the opcode */
+ if ((insn & 0xc1f80000) == 0x81c00000) {
+ unsigned long rd = ((insn >> 25) & 0x1f);
+
+ if (rd <= 15) {
+ slot = ®s->u_regs[rd];
+ } else {
+ unsigned long fp = regs->u_regs[UREG_FP];
+ /* Hard case, it goes onto the stack. */
+ flushw_all();
+
+ rd -= 16;
+ if (test_thread_64bit_stack(fp)) {
+ unsigned long __user *uslot =
+ (unsigned long __user *) (fp + STACK_BIAS) + rd;
+ rc = __put_user(real_pc, uslot);
+ } else {
+ unsigned int __user *uslot = (unsigned int
+ __user *) fp + rd;
+ rc = __put_user((u32) real_pc, uslot);
+ }
+ }
+ }
+ if (slot != NULL)
+ *slot = real_pc;
+ return rc;
+}
+
+/* Single-stepping can be avoided for certain instructions: NOPs and
+ * instructions that can be emulated. This function determines
+ * whether the instruction where the uprobe is installed falls in one
+ * of these cases and emulates it.
+ *
+ * This function returns true if the single-stepping can be skipped,
+ * false otherwise.
+ */
+bool arch_uprobe_skip_sstep(struct arch_uprobe *auprobe, struct pt_regs *regs)
+{
+ /* We currently only emulate NOP instructions.
+ */
+
+ if (auprobe->ixol == (1 << 24)) {
+ regs->tnpc += 4;
+ regs->tpc += 4;
+ return true;
+ }
+
+ return false;
+}
+
+/* Prepare to execute out of line. At this point
+ * current->utask->xol_vaddr points to an allocated XOL slot properly
+ * initialized with the original instruction and the single-stepping
+ * trap instruction.
+ *
+ * This function returns 0 on success, any other number on error.
+ */
+int arch_uprobe_pre_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
+{
+ struct uprobe_task *utask = current->utask;
+ struct arch_uprobe_task *autask = ¤t->utask->autask;
+
+ /* Save the current program counters so they can be restored
+ * later.
+ */
+ autask->saved_tpc = regs->tpc;
+ autask->saved_tnpc = regs->tnpc;
+
+ /* Adjust PC and NPC so the first instruction in the XOL slot
+ * will be executed by the user task.
+ */
+ instruction_pointer_set(regs, utask->xol_vaddr);
+
+ return 0;
+}
+
+/* Prepare to resume execution after the single-step. Called after
+ * single-stepping. 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.
+ *
+ * This function returns 0 on success, any other number on error.
+ */
+int arch_uprobe_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
+{
+ struct uprobe_task *utask = current->utask;
+ struct arch_uprobe_task *autask = &utask->autask;
+ u32 insn = auprobe->ixol;
+ int rc = 0;
+
+ if (utask->state == UTASK_SSTEP_ACK) {
+ regs->tnpc = relbranch_fixup(insn, utask, regs);
+ regs->tpc = autask->saved_tnpc;
+ rc = retpc_fixup(regs, insn, (unsigned long) utask->vaddr);
+ } else {
+ regs->tnpc = utask->vaddr+4;
+ regs->tpc = autask->saved_tnpc+4;
+ }
+ return rc;
+}
+
+/* Handler for uprobe traps. This is called from the traps table and
+ * triggers the proper die notification.
+ */
+asmlinkage void uprobe_trap(struct pt_regs *regs,
+ unsigned long trap_level)
+{
+ BUG_ON(trap_level != 0x173 && trap_level != 0x174);
+
+ /* We are only interested in user-mode code. Uprobe traps
+ * shall not be present in kernel code.
+ */
+ if (!user_mode(regs)) {
+ local_irq_enable();
+ bad_trap(regs, trap_level);
+ return;
+ }
+
+ /* trap_level == 0x173 --> ta 0x73
+ * trap_level == 0x174 --> ta 0x74
+ */
+ if (notify_die((trap_level == 0x173) ? DIE_BPT : DIE_SSTEP,
+ (trap_level == 0x173) ? "bpt" : "sstep",
+ regs, 0, trap_level, SIGTRAP) != NOTIFY_STOP)
+ bad_trap(regs, trap_level);
+}
+
+/* Callback routine for handling die notifications.
+*/
+int arch_uprobe_exception_notify(struct notifier_block *self,
+ unsigned long val, void *data)
+{
+ int ret = NOTIFY_DONE;
+ struct die_args *args = (struct die_args *)data;
+
+ /* We are only interested in userspace traps */
+ if (args->regs && !user_mode(args->regs))
+ return NOTIFY_DONE;
+
+ switch (val) {
+ case DIE_BPT:
+ if (uprobe_pre_sstep_notifier(args->regs))
+ ret = NOTIFY_STOP;
+ break;
+
+ case DIE_SSTEP:
+ if (uprobe_post_sstep_notifier(args->regs))
+ ret = NOTIFY_STOP;
+
+ default:
+ break;
+ }
+
+ return ret;
+}
+
+/* This function gets called when a XOL instruction either gets
+ * trapped or the thread has a fatal signal, so reset the instruction
+ * pointer to its probed address.
+ */
+void arch_uprobe_abort_xol(struct arch_uprobe *auprobe, struct pt_regs *regs)
+{
+ struct uprobe_task *utask = current->utask;
+
+ instruction_pointer_set(regs, utask->vaddr);
+}
+
+/* If xol insn itself traps and generates a signal(Say,
+ * SIGILL/SIGSEGV/etc), then detect the case where a singlestepped
+ * instruction jumps back to its own address.
+ */
+bool arch_uprobe_xol_was_trapped(struct task_struct *t)
+{
+ return false;
+}
+
+unsigned long
+arch_uretprobe_hijack_return_addr(unsigned long trampoline_vaddr,
+ struct pt_regs *regs)
+{
+ unsigned long orig_ret_vaddr = regs->u_regs[UREG_I7];
+
+ regs->u_regs[UREG_I7] = trampoline_vaddr-8;
+
+ return orig_ret_vaddr + 8;
+}
@@ -27,6 +27,7 @@
#include <linux/errno.h>
#include <linux/rbtree.h>
#include <linux/types.h>
+#include <linux/wait.h>
struct vm_area_struct;
struct mm_struct;