From 421d1069cd85f6fee9f36984a071a73b6a431f65 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Sun, 18 Nov 2018 18:41:46 -0500 Subject: arm64: function_graph: Remove use of FTRACE_NOTRACE_DEPTH Functions in the set_graph_notrace no longer subtract FTRACE_NOTRACE_DEPTH from curr_ret_stack, as that is now implemented via the trace_recursion flags. Access to curr_ret_stack no longer needs to worry about checking for this. curr_ret_stack is still initialized to -1, when there's not a shadow stack allocated. Cc: Catalin Marinas Cc: linux-arm-kernel@lists.infradead.org Acked-by: Will Deacon Reviewed-by: Joel Fernandes (Google) Signed-off-by: Steven Rostedt (VMware) --- arch/arm64/kernel/stacktrace.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'arch') diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c index 4989f7ea1e59..7723dadf25be 100644 --- a/arch/arm64/kernel/stacktrace.c +++ b/arch/arm64/kernel/stacktrace.c @@ -61,9 +61,6 @@ int notrace unwind_frame(struct task_struct *tsk, struct stackframe *frame) (frame->pc == (unsigned long)return_to_handler)) { if (WARN_ON_ONCE(frame->graph == -1)) return -EINVAL; - if (frame->graph < -1) - frame->graph += FTRACE_NOTRACE_DEPTH; - /* * This is a case where function graph tracer has * modified a return address (LR) in a stack frame -- cgit v1.2.3 From e4c07bf9867aeaec14bac042fbbd50d885f6ed3a Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Wed, 5 Dec 2018 12:48:54 -0500 Subject: arm64: ftrace: Set FTRACE_MAY_SLEEP before ftrace_modify_all_code() It has been reported that ftrace_replace_code() which is called by ftrace_modify_all_code() can cause a soft lockup warning for an allmodconfig kernel. This is because all the debug options enabled causes the loop in ftrace_replace_code() (which loops over all the functions being enabled where there can be 10s of thousands), is too slow, and never schedules out. To solve this, setting FTRACE_MAY_SLEEP to the command passed into ftrace_replace_code() will make it call cond_resched() in the loop, which prevents the soft lockup warning from triggering. Link: http://lkml.kernel.org/r/20181204192903.8193-1-anders.roxell@linaro.org Link: http://lkml.kernel.org/r/20181205183304.000714627@goodmis.org Acked-by: Will Deacon Reported-by: Anders Roxell Tested-by: Anders Roxell Signed-off-by: Steven Rostedt (VMware) --- arch/arm64/kernel/ftrace.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/arm64/kernel/ftrace.c b/arch/arm64/kernel/ftrace.c index 57e962290df3..57d4e936a176 100644 --- a/arch/arm64/kernel/ftrace.c +++ b/arch/arm64/kernel/ftrace.c @@ -193,6 +193,7 @@ int ftrace_make_nop(struct module *mod, struct dyn_ftrace *rec, void arch_ftrace_update_code(int command) { + command |= FTRACE_MAY_SLEEP; ftrace_modify_all_code(command); } -- cgit v1.2.3 From d2a68c4effd821f0871d20368f76b609349c8a3b Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Sat, 8 Dec 2018 12:58:51 -0500 Subject: x86/ftrace: Do not call function graph from dynamic trampolines Since commit 79922b8009c07 ("ftrace: Optimize function graph to be called directly"), dynamic trampolines should not be calling the function graph tracer at the end. If they do, it could cause the function graph tracer to trace functions that it filtered out. Right now it does not cause a problem because there's a test to check if the function graph tracer is attached to the same function as the function tracer, which for now is true. But the function graph tracer is undergoing changes that can make this no longer true which will cause the function graph tracer to trace other functions. For example: # cd /sys/kernel/tracing/ # echo do_IRQ > set_ftrace_filter # mkdir instances/foo # echo ip_rcv > instances/foo/set_ftrace_filter # echo function_graph > current_tracer # echo function > instances/foo/current_tracer Would cause the function graph tracer to trace both do_IRQ and ip_rcv, if the current tests change. As the current tests prevent this from being a problem, this code does not need to be backported. But it does make the code cleaner. Cc: Thomas Gleixner Cc: Borislav Petkov Cc: "H. Peter Anvin" Cc: x86@kernel.org Signed-off-by: Steven Rostedt (VMware) --- arch/x86/kernel/ftrace.c | 41 ++++++++++++++++++++++------------------- arch/x86/kernel/ftrace_64.S | 8 ++++---- 2 files changed, 26 insertions(+), 23 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c index 7ee8067cbf45..8257a59704ae 100644 --- a/arch/x86/kernel/ftrace.c +++ b/arch/x86/kernel/ftrace.c @@ -733,18 +733,20 @@ union ftrace_op_code_union { } __attribute__((packed)); }; +#define RET_SIZE 1 + static unsigned long create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) { - unsigned const char *jmp; unsigned long start_offset; unsigned long end_offset; unsigned long op_offset; unsigned long offset; unsigned long size; - unsigned long ip; + unsigned long retq; unsigned long *ptr; void *trampoline; + void *ip; /* 48 8b 15 is movq (%rip), %rdx */ unsigned const char op_ref[] = { 0x48, 0x8b, 0x15 }; union ftrace_op_code_union op_ptr; @@ -764,27 +766,27 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) /* * Allocate enough size to store the ftrace_caller code, - * the jmp to ftrace_epilogue, as well as the address of - * the ftrace_ops this trampoline is used for. + * the iret , as well as the address of the ftrace_ops this + * trampoline is used for. */ - trampoline = alloc_tramp(size + MCOUNT_INSN_SIZE + sizeof(void *)); + trampoline = alloc_tramp(size + RET_SIZE + sizeof(void *)); if (!trampoline) return 0; - *tramp_size = size + MCOUNT_INSN_SIZE + sizeof(void *); + *tramp_size = size + RET_SIZE + sizeof(void *); /* Copy ftrace_caller onto the trampoline memory */ ret = probe_kernel_read(trampoline, (void *)start_offset, size); - if (WARN_ON(ret < 0)) { - tramp_free(trampoline, *tramp_size); - return 0; - } + if (WARN_ON(ret < 0)) + goto fail; - ip = (unsigned long)trampoline + size; + ip = trampoline + size; - /* The trampoline ends with a jmp to ftrace_epilogue */ - jmp = ftrace_jmp_replace(ip, (unsigned long)ftrace_epilogue); - memcpy(trampoline + size, jmp, MCOUNT_INSN_SIZE); + /* The trampoline ends with ret(q) */ + retq = (unsigned long)ftrace_stub; + ret = probe_kernel_read(ip, (void *)retq, RET_SIZE); + if (WARN_ON(ret < 0)) + goto fail; /* * The address of the ftrace_ops that is used for this trampoline @@ -794,17 +796,15 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) * the global function_trace_op variable. */ - ptr = (unsigned long *)(trampoline + size + MCOUNT_INSN_SIZE); + ptr = (unsigned long *)(trampoline + size + RET_SIZE); *ptr = (unsigned long)ops; op_offset -= start_offset; memcpy(&op_ptr, trampoline + op_offset, OP_REF_SIZE); /* Are we pointing to the reference? */ - if (WARN_ON(memcmp(op_ptr.op, op_ref, 3) != 0)) { - tramp_free(trampoline, *tramp_size); - return 0; - } + if (WARN_ON(memcmp(op_ptr.op, op_ref, 3) != 0)) + goto fail; /* Load the contents of ptr into the callback parameter */ offset = (unsigned long)ptr; @@ -819,6 +819,9 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) ops->flags |= FTRACE_OPS_FL_ALLOC_TRAMP; return (unsigned long)trampoline; +fail: + tramp_free(trampoline, *tramp_size); + return 0; } static unsigned long calc_trampoline_call_offset(bool save_regs) diff --git a/arch/x86/kernel/ftrace_64.S b/arch/x86/kernel/ftrace_64.S index 91b2cff4b79a..75f2b36b41a6 100644 --- a/arch/x86/kernel/ftrace_64.S +++ b/arch/x86/kernel/ftrace_64.S @@ -171,9 +171,6 @@ GLOBAL(ftrace_call) restore_mcount_regs /* - * The copied trampoline must call ftrace_epilogue as it - * still may need to call the function graph tracer. - * * The code up to this label is copied into trampolines so * think twice before adding any new code or changing the * layout here. @@ -185,7 +182,10 @@ GLOBAL(ftrace_graph_call) jmp ftrace_stub #endif -/* This is weak to keep gas from relaxing the jumps */ +/* + * This is weak to keep gas from relaxing the jumps. + * It is also used to copy the retq for trampolines. + */ WEAK(ftrace_stub) retq ENDPROC(ftrace_caller) -- cgit v1.2.3 From 0fad8bfef7b08a68507178f8e278d013b60ff966 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Fri, 7 Dec 2018 12:35:47 -0500 Subject: powerpc/frace: Use ftrace_graph_get_ret_stack() instead of curr_ret_stack The structure of the ret_stack array on the task struct is going to change, and accessing it directly via the curr_ret_stack index will no longer give the ret_stack entry that holds the return address. To access that, architectures must now use ftrace_graph_get_ret_stack() to get the associated ret_stack that matches the saved return address. Cc: Benjamin Herrenschmidt Cc: Paul Mackerras Cc: linuxppc-dev@lists.ozlabs.org Acked-by: Michael Ellerman Signed-off-by: Steven Rostedt (VMware) --- arch/powerpc/kernel/process.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c index 96f34730010f..ce393df243aa 100644 --- a/arch/powerpc/kernel/process.c +++ b/arch/powerpc/kernel/process.c @@ -2061,9 +2061,10 @@ void show_stack(struct task_struct *tsk, unsigned long *stack) int count = 0; int firstframe = 1; #ifdef CONFIG_FUNCTION_GRAPH_TRACER - int curr_frame = current->curr_ret_stack; + struct ftrace_ret_stack *ret_stack; extern void return_to_handler(void); unsigned long rth = (unsigned long)return_to_handler; + int curr_frame = 0; #endif sp = (unsigned long) stack; @@ -2089,9 +2090,13 @@ void show_stack(struct task_struct *tsk, unsigned long *stack) printk("["REG"] ["REG"] %pS", sp, ip, (void *)ip); #ifdef CONFIG_FUNCTION_GRAPH_TRACER if ((ip == rth) && curr_frame >= 0) { - pr_cont(" (%pS)", - (void *)current->ret_stack[curr_frame].ret); - curr_frame--; + ret_stack = ftrace_graph_get_ret_stack(current, + curr_frame++); + if (ret_stack) + pr_cont(" (%pS)", + (void *)ret_stack->ret); + else + curr_frame = -1; } #endif if (firstframe) -- cgit v1.2.3 From 945626db0961d8388543b2c96b6f16df57947392 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Fri, 7 Dec 2018 12:51:27 -0500 Subject: sparc64: Use ftrace_graph_get_ret_stack() instead of curr_ret_stack The structure of the ret_stack array on the task struct is going to change, and accessing it directly via the curr_ret_stack index will no longer give the ret_stack entry that holds the return address. To access that, architectures must now use ftrace_graph_get_ret_stack() to get the associated ret_stack that matches the saved return address. Cc: sparclinux@vger.kernel.org Acked-by: David S. Miller Signed-off-by: Steven Rostedt (VMware) --- arch/sparc/kernel/perf_event.c | 8 +++++--- arch/sparc/kernel/stacktrace.c | 8 +++++--- arch/sparc/kernel/traps_64.c | 7 ++++--- 3 files changed, 14 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c index 47c871394ccb..6de7c684c29f 100644 --- a/arch/sparc/kernel/perf_event.c +++ b/arch/sparc/kernel/perf_event.c @@ -1767,9 +1767,11 @@ void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, perf_callchain_store(entry, pc); #ifdef CONFIG_FUNCTION_GRAPH_TRACER if ((pc + 8UL) == (unsigned long) &return_to_handler) { - int index = current->curr_ret_stack; - if (current->ret_stack && index >= graph) { - pc = current->ret_stack[index - graph].ret; + struct ftrace_ret_stack *ret_stack; + ret_stack = ftrace_graph_get_ret_stack(current, + graph); + if (ret_stack) { + pc = ret_stack->ret; perf_callchain_store(entry, pc); graph++; } diff --git a/arch/sparc/kernel/stacktrace.c b/arch/sparc/kernel/stacktrace.c index be4c14cccc05..dd654e651500 100644 --- a/arch/sparc/kernel/stacktrace.c +++ b/arch/sparc/kernel/stacktrace.c @@ -57,9 +57,11 @@ static void __save_stack_trace(struct thread_info *tp, trace->entries[trace->nr_entries++] = pc; #ifdef CONFIG_FUNCTION_GRAPH_TRACER if ((pc + 8UL) == (unsigned long) &return_to_handler) { - int index = t->curr_ret_stack; - if (t->ret_stack && index >= graph) { - pc = t->ret_stack[index - graph].ret; + struct ftrace_ret_stack *ret_stack; + ret_stack = ftrace_graph_get_ret_stack(t, + graph); + if (ret_stack) { + pc = ret_stack->ret; if (trace->nr_entries < trace->max_entries) trace->entries[trace->nr_entries++] = pc; diff --git a/arch/sparc/kernel/traps_64.c b/arch/sparc/kernel/traps_64.c index aa624ed79db1..0cd02a64a451 100644 --- a/arch/sparc/kernel/traps_64.c +++ b/arch/sparc/kernel/traps_64.c @@ -2502,9 +2502,10 @@ void show_stack(struct task_struct *tsk, unsigned long *_ksp) printk(" [%016lx] %pS\n", pc, (void *) pc); #ifdef CONFIG_FUNCTION_GRAPH_TRACER if ((pc + 8UL) == (unsigned long) &return_to_handler) { - int index = tsk->curr_ret_stack; - if (tsk->ret_stack && index >= graph) { - pc = tsk->ret_stack[index - graph].ret; + struct ftrace_ret_stack *ret_stack; + ret_stack = ftrace_graph_get_ret_stack(tsk, graph); + if (ret_stack) { + pc = ret_stack->ret; printk(" [%016lx] %pS\n", pc, (void *) pc); graph++; } -- cgit v1.2.3 From cec8d0e7f06e08b981e9d61bef267c8c36d536f5 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Fri, 7 Dec 2018 13:06:04 -0500 Subject: sh: ftrace: Use ftrace_graph_get_ret_stack() instead of curr_ret_stack The structure of the ret_stack array on the task struct is going to change, and accessing it directly via the curr_ret_stack index will no longer give the ret_stack entry that holds the return address. To access that, architectures must now use ftrace_graph_get_ret_stack() to get the associated ret_stack that matches the saved return address. Cc: linux-sh@vger.kernel.org Cc: Yoshinori Sato Cc: Rich Felker Signed-off-by: Steven Rostedt (VMware) --- arch/sh/kernel/dumpstack.c | 11 +++++++---- arch/sh/kernel/dwarf.c | 9 +++++---- 2 files changed, 12 insertions(+), 8 deletions(-) (limited to 'arch') diff --git a/arch/sh/kernel/dumpstack.c b/arch/sh/kernel/dumpstack.c index b564b1eae4ae..2c2e151bf39e 100644 --- a/arch/sh/kernel/dumpstack.c +++ b/arch/sh/kernel/dumpstack.c @@ -59,17 +59,20 @@ print_ftrace_graph_addr(unsigned long addr, void *data, struct thread_info *tinfo, int *graph) { struct task_struct *task = tinfo->task; + struct ftrace_ret_stack *ret_stack; unsigned long ret_addr; - int index = task->curr_ret_stack; if (addr != (unsigned long)return_to_handler) return; - if (!task->ret_stack || index < *graph) + if (!task->ret_stack) return; - index -= *graph; - ret_addr = task->ret_stack[index].ret; + ret_stack = ftrace_graph_get_ret_stack(task, *graph); + if (!ret_stack) + return; + + ret_addr = ret_stack->ret; ops->address(data, ret_addr, 1); diff --git a/arch/sh/kernel/dwarf.c b/arch/sh/kernel/dwarf.c index bb511e2d9d68..df0fd6efe758 100644 --- a/arch/sh/kernel/dwarf.c +++ b/arch/sh/kernel/dwarf.c @@ -608,17 +608,18 @@ struct dwarf_frame *dwarf_unwind_stack(unsigned long pc, * expected to find the real return address. */ if (pc == (unsigned long)&return_to_handler) { - int index = current->curr_ret_stack; + struct ftrace_ret_stack *ret_stack; + ret_stack = ftrace_graph_get_ret_stack(current, 0); + if (ret_stack) + pc = ret_stack->ret; /* * We currently have no way of tracking how many * return_to_handler()'s we've seen. If there is more * than one patched return address on our stack, * complain loudly. */ - WARN_ON(index > 0); - - pc = current->ret_stack[index].ret; + WARN_ON(ftrace_graph_get_ret_stack(current, 1); } #endif -- cgit v1.2.3 From a448276ce515c91cde4675be497364b91c764d95 Mon Sep 17 00:00:00 2001 From: "Steven Rostedt (VMware)" Date: Fri, 7 Dec 2018 13:13:28 -0500 Subject: arm64: Use ftrace_graph_get_ret_stack() instead of curr_ret_stack The structure of the ret_stack array on the task struct is going to change, and accessing it directly via the curr_ret_stack index will no longer give the ret_stack entry that holds the return address. To access that, architectures must now use ftrace_graph_get_ret_stack() to get the associated ret_stack that matches the saved return address. Cc: linux-arm-kernel@lists.infradead.org Cc: Will Deacon Cc: Mark Rutland Cc: Catalin Marinas Signed-off-by: Steven Rostedt (VMware) --- arch/arm64/kernel/perf_callchain.c | 2 +- arch/arm64/kernel/process.c | 2 +- arch/arm64/kernel/return_address.c | 2 +- arch/arm64/kernel/stacktrace.c | 12 +++++++----- arch/arm64/kernel/time.c | 2 +- arch/arm64/kernel/traps.c | 2 +- 6 files changed, 12 insertions(+), 10 deletions(-) (limited to 'arch') diff --git a/arch/arm64/kernel/perf_callchain.c b/arch/arm64/kernel/perf_callchain.c index bcafd7dcfe8b..1b792b46604e 100644 --- a/arch/arm64/kernel/perf_callchain.c +++ b/arch/arm64/kernel/perf_callchain.c @@ -164,7 +164,7 @@ void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, frame.fp = regs->regs[29]; frame.pc = regs->pc; #ifdef CONFIG_FUNCTION_GRAPH_TRACER - frame.graph = current->curr_ret_stack; + frame.graph = 0; #endif walk_stackframe(current, &frame, callchain_trace, entry); diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c index d9a4c2d6dd8b..37a66394b07d 100644 --- a/arch/arm64/kernel/process.c +++ b/arch/arm64/kernel/process.c @@ -459,7 +459,7 @@ unsigned long get_wchan(struct task_struct *p) frame.fp = thread_saved_fp(p); frame.pc = thread_saved_pc(p); #ifdef CONFIG_FUNCTION_GRAPH_TRACER - frame.graph = p->curr_ret_stack; + frame.graph = 0; #endif do { if (unwind_frame(p, &frame)) diff --git a/arch/arm64/kernel/return_address.c b/arch/arm64/kernel/return_address.c index 933adbc0f654..53c40196b607 100644 --- a/arch/arm64/kernel/return_address.c +++ b/arch/arm64/kernel/return_address.c @@ -44,7 +44,7 @@ void *return_address(unsigned int level) frame.fp = (unsigned long)__builtin_frame_address(0); frame.pc = (unsigned long)return_address; /* dummy */ #ifdef CONFIG_FUNCTION_GRAPH_TRACER - frame.graph = current->curr_ret_stack; + frame.graph = 0; #endif walk_stackframe(current, &frame, save_return_addr, &data); diff --git a/arch/arm64/kernel/stacktrace.c b/arch/arm64/kernel/stacktrace.c index 7723dadf25be..1a29f2695ff2 100644 --- a/arch/arm64/kernel/stacktrace.c +++ b/arch/arm64/kernel/stacktrace.c @@ -59,15 +59,17 @@ int notrace unwind_frame(struct task_struct *tsk, struct stackframe *frame) #ifdef CONFIG_FUNCTION_GRAPH_TRACER if (tsk->ret_stack && (frame->pc == (unsigned long)return_to_handler)) { - if (WARN_ON_ONCE(frame->graph == -1)) - return -EINVAL; + struct ftrace_ret_stack *ret_stack; /* * This is a case where function graph tracer has * modified a return address (LR) in a stack frame * to hook a function return. * So replace it to an original value. */ - frame->pc = tsk->ret_stack[frame->graph--].ret; + ret_stack = ftrace_graph_get_ret_stack(tsk, frame->graph++); + if (WARN_ON_ONCE(!ret_stack)) + return -EINVAL; + frame->pc = ret_stack->ret; } #endif /* CONFIG_FUNCTION_GRAPH_TRACER */ @@ -134,7 +136,7 @@ void save_stack_trace_regs(struct pt_regs *regs, struct stack_trace *trace) frame.fp = regs->regs[29]; frame.pc = regs->pc; #ifdef CONFIG_FUNCTION_GRAPH_TRACER - frame.graph = current->curr_ret_stack; + frame.graph = 0; #endif walk_stackframe(current, &frame, save_trace, &data); @@ -165,7 +167,7 @@ static noinline void __save_stack_trace(struct task_struct *tsk, frame.pc = (unsigned long)__save_stack_trace; } #ifdef CONFIG_FUNCTION_GRAPH_TRACER - frame.graph = tsk->curr_ret_stack; + frame.graph = 0; #endif walk_stackframe(tsk, &frame, save_trace, &data); diff --git a/arch/arm64/kernel/time.c b/arch/arm64/kernel/time.c index f258636273c9..a777ae90044d 100644 --- a/arch/arm64/kernel/time.c +++ b/arch/arm64/kernel/time.c @@ -52,7 +52,7 @@ unsigned long profile_pc(struct pt_regs *regs) frame.fp = regs->regs[29]; frame.pc = regs->pc; #ifdef CONFIG_FUNCTION_GRAPH_TRACER - frame.graph = current->curr_ret_stack; + frame.graph = 0; #endif do { int ret = unwind_frame(NULL, &frame); diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c index 5f4d9acb32f5..49ebf3771391 100644 --- a/arch/arm64/kernel/traps.c +++ b/arch/arm64/kernel/traps.c @@ -122,7 +122,7 @@ void dump_backtrace(struct pt_regs *regs, struct task_struct *tsk) frame.pc = thread_saved_pc(tsk); } #ifdef CONFIG_FUNCTION_GRAPH_TRACER - frame.graph = tsk->curr_ret_stack; + frame.graph = 0; #endif skip = !!regs; -- cgit v1.2.3