linux_dsm_epyc7002/arch/x86/kernel/ftrace_64.S

333 lines
7.8 KiB
ArmAsm
Raw Normal View History

/*
* Copyright (C) 2014 Steven Rostedt, Red Hat Inc
*/
#include <linux/linkage.h>
#include <asm/ptrace.h>
#include <asm/ftrace.h>
#include <asm/export.h>
.code64
.section .entry.text, "ax"
#ifdef CC_USING_FENTRY
# define function_hook __fentry__
EXPORT_SYMBOL(__fentry__)
#else
# define function_hook mcount
EXPORT_SYMBOL(mcount)
#endif
/* All cases save the original rbp (8 bytes) */
#ifdef CONFIG_FRAME_POINTER
# ifdef CC_USING_FENTRY
/* Save parent and function stack frames (rip and rbp) */
# define MCOUNT_FRAME_SIZE (8+16*2)
# else
/* Save just function stack frame (rip and rbp) */
# define MCOUNT_FRAME_SIZE (8+16)
# endif
#else
/* No need to save a stack frame */
# define MCOUNT_FRAME_SIZE 8
#endif /* CONFIG_FRAME_POINTER */
/* Size of stack used to save mcount regs in save_mcount_regs */
#define MCOUNT_REG_SIZE (SS+8 + MCOUNT_FRAME_SIZE)
/*
* gcc -pg option adds a call to 'mcount' in most functions.
* When -mfentry is used, the call is to 'fentry' and not 'mcount'
* and is done before the function's stack frame is set up.
* They both require a set of regs to be saved before calling
* any C code and restored before returning back to the function.
*
* On boot up, all these calls are converted into nops. When tracing
* is enabled, the call can jump to either ftrace_caller or
* ftrace_regs_caller. Callbacks (tracing functions) that require
* ftrace_regs_caller (like kprobes) need to have pt_regs passed to
* it. For this reason, the size of the pt_regs structure will be
* allocated on the stack and the required mcount registers will
* be saved in the locations that pt_regs has them in.
*/
/*
* @added: the amount of stack added before calling this
*
* After this is called, the following registers contain:
*
* %rdi - holds the address that called the trampoline
* %rsi - holds the parent function (traced function's return address)
* %rdx - holds the original %rbp
*/
.macro save_mcount_regs added=0
/* Always save the original rbp */
pushq %rbp
#ifdef CONFIG_FRAME_POINTER
/*
* Stack traces will stop at the ftrace trampoline if the frame pointer
* is not set up properly. If fentry is used, we need to save a frame
* pointer for the parent as well as the function traced, because the
* fentry is called before the stack frame is set up, where as mcount
* is called afterward.
*/
#ifdef CC_USING_FENTRY
/* Save the parent pointer (skip orig rbp and our return address) */
pushq \added+8*2(%rsp)
pushq %rbp
movq %rsp, %rbp
/* Save the return address (now skip orig rbp, rbp and parent) */
pushq \added+8*3(%rsp)
#else
/* Can't assume that rip is before this (unless added was zero) */
pushq \added+8(%rsp)
#endif
pushq %rbp
movq %rsp, %rbp
#endif /* CONFIG_FRAME_POINTER */
/*
* We add enough stack to save all regs.
*/
subq $(MCOUNT_REG_SIZE - MCOUNT_FRAME_SIZE), %rsp
movq %rax, RAX(%rsp)
movq %rcx, RCX(%rsp)
movq %rdx, RDX(%rsp)
movq %rsi, RSI(%rsp)
movq %rdi, RDI(%rsp)
movq %r8, R8(%rsp)
movq %r9, R9(%rsp)
/*
* Save the original RBP. Even though the mcount ABI does not
* require this, it helps out callers.
*/
movq MCOUNT_REG_SIZE-8(%rsp), %rdx
movq %rdx, RBP(%rsp)
/* Copy the parent address into %rsi (second parameter) */
#ifdef CC_USING_FENTRY
movq MCOUNT_REG_SIZE+8+\added(%rsp), %rsi
#else
/* %rdx contains original %rbp */
movq 8(%rdx), %rsi
#endif
/* Move RIP to its proper location */
movq MCOUNT_REG_SIZE+\added(%rsp), %rdi
movq %rdi, RIP(%rsp)
/*
* Now %rdi (the first parameter) has the return address of
* where ftrace_call returns. But the callbacks expect the
* address of the call itself.
*/
subq $MCOUNT_INSN_SIZE, %rdi
.endm
.macro restore_mcount_regs
movq R9(%rsp), %r9
movq R8(%rsp), %r8
movq RDI(%rsp), %rdi
movq RSI(%rsp), %rsi
movq RDX(%rsp), %rdx
movq RCX(%rsp), %rcx
movq RAX(%rsp), %rax
/* ftrace_regs_caller can modify %rbp */
movq RBP(%rsp), %rbp
addq $MCOUNT_REG_SIZE, %rsp
.endm
#ifdef CONFIG_DYNAMIC_FTRACE
ENTRY(function_hook)
retq
END(function_hook)
ENTRY(ftrace_caller)
/* save_mcount_regs fills in first two parameters */
save_mcount_regs
GLOBAL(ftrace_caller_op_ptr)
/* Load the ftrace_ops into the 3rd parameter */
movq function_trace_op(%rip), %rdx
/* regs go into 4th parameter (but make it NULL) */
movq $0, %rcx
GLOBAL(ftrace_call)
call ftrace_stub
restore_mcount_regs
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops The current method of handling multiple function callbacks is to register a list function callback that calls all the other callbacks based on their hash tables and compare it to the function that the callback was called on. But this is very inefficient. For example, if you are tracing all functions in the kernel and then add a kprobe to a function such that the kprobe uses ftrace, the mcount trampoline will switch from calling the function trace callback to calling the list callback that will iterate over all registered ftrace_ops (in this case, the function tracer and the kprobes callback). That means for every function being traced it checks the hash of the ftrace_ops for function tracing and kprobes, even though the kprobes is only set at a single function. The kprobes ftrace_ops is checked for every function being traced! Instead of calling the list function for functions that are only being traced by a single callback, we can call a dynamically allocated trampoline that calls the callback directly. The function graph tracer already uses a direct call trampoline when it is being traced by itself but it is not dynamically allocated. It's trampoline is static in the kernel core. The infrastructure that called the function graph trampoline can also be used to call a dynamically allocated one. For now, only ftrace_ops that are not dynamically allocated can have a trampoline. That is, users such as function tracer or stack tracer. kprobes and perf allocate their ftrace_ops, and until there's a safe way to free the trampoline, it can not be used. The dynamically allocated ftrace_ops may, although, use the trampoline if the kernel is not compiled with CONFIG_PREEMPT. But that will come later. Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Tested-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 10:23:31 +07:00
/*
* The copied trampoline must call ftrace_epilogue as it
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops The current method of handling multiple function callbacks is to register a list function callback that calls all the other callbacks based on their hash tables and compare it to the function that the callback was called on. But this is very inefficient. For example, if you are tracing all functions in the kernel and then add a kprobe to a function such that the kprobe uses ftrace, the mcount trampoline will switch from calling the function trace callback to calling the list callback that will iterate over all registered ftrace_ops (in this case, the function tracer and the kprobes callback). That means for every function being traced it checks the hash of the ftrace_ops for function tracing and kprobes, even though the kprobes is only set at a single function. The kprobes ftrace_ops is checked for every function being traced! Instead of calling the list function for functions that are only being traced by a single callback, we can call a dynamically allocated trampoline that calls the callback directly. The function graph tracer already uses a direct call trampoline when it is being traced by itself but it is not dynamically allocated. It's trampoline is static in the kernel core. The infrastructure that called the function graph trampoline can also be used to call a dynamically allocated one. For now, only ftrace_ops that are not dynamically allocated can have a trampoline. That is, users such as function tracer or stack tracer. kprobes and perf allocate their ftrace_ops, and until there's a safe way to free the trampoline, it can not be used. The dynamically allocated ftrace_ops may, although, use the trampoline if the kernel is not compiled with CONFIG_PREEMPT. But that will come later. Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Tested-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 10:23:31 +07:00
* 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.
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops The current method of handling multiple function callbacks is to register a list function callback that calls all the other callbacks based on their hash tables and compare it to the function that the callback was called on. But this is very inefficient. For example, if you are tracing all functions in the kernel and then add a kprobe to a function such that the kprobe uses ftrace, the mcount trampoline will switch from calling the function trace callback to calling the list callback that will iterate over all registered ftrace_ops (in this case, the function tracer and the kprobes callback). That means for every function being traced it checks the hash of the ftrace_ops for function tracing and kprobes, even though the kprobes is only set at a single function. The kprobes ftrace_ops is checked for every function being traced! Instead of calling the list function for functions that are only being traced by a single callback, we can call a dynamically allocated trampoline that calls the callback directly. The function graph tracer already uses a direct call trampoline when it is being traced by itself but it is not dynamically allocated. It's trampoline is static in the kernel core. The infrastructure that called the function graph trampoline can also be used to call a dynamically allocated one. For now, only ftrace_ops that are not dynamically allocated can have a trampoline. That is, users such as function tracer or stack tracer. kprobes and perf allocate their ftrace_ops, and until there's a safe way to free the trampoline, it can not be used. The dynamically allocated ftrace_ops may, although, use the trampoline if the kernel is not compiled with CONFIG_PREEMPT. But that will come later. Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Tested-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 10:23:31 +07:00
*/
GLOBAL(ftrace_epilogue)
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
GLOBAL(ftrace_graph_call)
jmp ftrace_stub
#endif
ftrace/x86: Set ftrace_stub to weak to prevent gcc from using short jumps to it Matt Fleming reported seeing crashes when enabling and disabling function profiling which uses function graph tracer. Later Namhyung Kim hit a similar issue and he found that the issue was due to the jmp to ftrace_stub in ftrace_graph_call was only two bytes, and when it was changed to jump to the tracing code, it overwrote the ftrace_stub that was after it. Masami Hiramatsu bisected this down to a binutils change: 8dcea93252a9ea7dff57e85220a719e2a5e8ab41 is the first bad commit commit 8dcea93252a9ea7dff57e85220a719e2a5e8ab41 Author: H.J. Lu <hjl.tools@gmail.com> Date: Fri May 15 03:17:31 2015 -0700 Add -mshared option to x86 ELF assembler This patch adds -mshared option to x86 ELF assembler. By default, assembler will optimize out non-PLT relocations against defined non-weak global branch targets with default visibility. The -mshared option tells the assembler to generate code which may go into a shared library where all non-weak global branch targets with default visibility can be preempted. The resulting code is slightly bigger. This option only affects the handling of branch instructions. Declaring ftrace_stub as a weak call prevents gas from using two byte jumps to it, which would be converted to a jump to the function graph code. Link: http://lkml.kernel.org/r/20160516230035.1dbae571@gandalf.local.home Reported-by: Matt Fleming <matt@codeblueprint.co.uk> Reported-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Matt Fleming <matt@codeblueprint.co.uk> Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2016-05-17 10:00:35 +07:00
/* This is weak to keep gas from relaxing the jumps */
WEAK(ftrace_stub)
retq
END(ftrace_caller)
ENTRY(ftrace_regs_caller)
/* Save the current flags before any operations that can change them */
pushfq
/* added 8 bytes to save flags */
save_mcount_regs 8
/* save_mcount_regs fills in first two parameters */
GLOBAL(ftrace_regs_caller_op_ptr)
/* Load the ftrace_ops into the 3rd parameter */
movq function_trace_op(%rip), %rdx
/* Save the rest of pt_regs */
movq %r15, R15(%rsp)
movq %r14, R14(%rsp)
movq %r13, R13(%rsp)
movq %r12, R12(%rsp)
movq %r11, R11(%rsp)
movq %r10, R10(%rsp)
movq %rbx, RBX(%rsp)
/* Copy saved flags */
movq MCOUNT_REG_SIZE(%rsp), %rcx
movq %rcx, EFLAGS(%rsp)
/* Kernel segments */
movq $__KERNEL_DS, %rcx
movq %rcx, SS(%rsp)
movq $__KERNEL_CS, %rcx
movq %rcx, CS(%rsp)
/* Stack - skipping return address and flags */
leaq MCOUNT_REG_SIZE+8*2(%rsp), %rcx
movq %rcx, RSP(%rsp)
/* regs go into 4th parameter */
leaq (%rsp), %rcx
GLOBAL(ftrace_regs_call)
call ftrace_stub
/* Copy flags back to SS, to restore them */
movq EFLAGS(%rsp), %rax
movq %rax, MCOUNT_REG_SIZE(%rsp)
/* Handlers can change the RIP */
movq RIP(%rsp), %rax
movq %rax, MCOUNT_REG_SIZE+8(%rsp)
/* restore the rest of pt_regs */
movq R15(%rsp), %r15
movq R14(%rsp), %r14
movq R13(%rsp), %r13
movq R12(%rsp), %r12
movq R10(%rsp), %r10
movq RBX(%rsp), %rbx
restore_mcount_regs
/* Restore flags */
popfq
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops The current method of handling multiple function callbacks is to register a list function callback that calls all the other callbacks based on their hash tables and compare it to the function that the callback was called on. But this is very inefficient. For example, if you are tracing all functions in the kernel and then add a kprobe to a function such that the kprobe uses ftrace, the mcount trampoline will switch from calling the function trace callback to calling the list callback that will iterate over all registered ftrace_ops (in this case, the function tracer and the kprobes callback). That means for every function being traced it checks the hash of the ftrace_ops for function tracing and kprobes, even though the kprobes is only set at a single function. The kprobes ftrace_ops is checked for every function being traced! Instead of calling the list function for functions that are only being traced by a single callback, we can call a dynamically allocated trampoline that calls the callback directly. The function graph tracer already uses a direct call trampoline when it is being traced by itself but it is not dynamically allocated. It's trampoline is static in the kernel core. The infrastructure that called the function graph trampoline can also be used to call a dynamically allocated one. For now, only ftrace_ops that are not dynamically allocated can have a trampoline. That is, users such as function tracer or stack tracer. kprobes and perf allocate their ftrace_ops, and until there's a safe way to free the trampoline, it can not be used. The dynamically allocated ftrace_ops may, although, use the trampoline if the kernel is not compiled with CONFIG_PREEMPT. But that will come later. Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Tested-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 10:23:31 +07:00
/*
* As this jmp to ftrace_epilogue can be a short jump
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops The current method of handling multiple function callbacks is to register a list function callback that calls all the other callbacks based on their hash tables and compare it to the function that the callback was called on. But this is very inefficient. For example, if you are tracing all functions in the kernel and then add a kprobe to a function such that the kprobe uses ftrace, the mcount trampoline will switch from calling the function trace callback to calling the list callback that will iterate over all registered ftrace_ops (in this case, the function tracer and the kprobes callback). That means for every function being traced it checks the hash of the ftrace_ops for function tracing and kprobes, even though the kprobes is only set at a single function. The kprobes ftrace_ops is checked for every function being traced! Instead of calling the list function for functions that are only being traced by a single callback, we can call a dynamically allocated trampoline that calls the callback directly. The function graph tracer already uses a direct call trampoline when it is being traced by itself but it is not dynamically allocated. It's trampoline is static in the kernel core. The infrastructure that called the function graph trampoline can also be used to call a dynamically allocated one. For now, only ftrace_ops that are not dynamically allocated can have a trampoline. That is, users such as function tracer or stack tracer. kprobes and perf allocate their ftrace_ops, and until there's a safe way to free the trampoline, it can not be used. The dynamically allocated ftrace_ops may, although, use the trampoline if the kernel is not compiled with CONFIG_PREEMPT. But that will come later. Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Tested-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 10:23:31 +07:00
* it must not be copied into the trampoline.
* The trampoline will add the code to jump
* to the return.
*/
GLOBAL(ftrace_regs_caller_end)
jmp ftrace_epilogue
END(ftrace_regs_caller)
#else /* ! CONFIG_DYNAMIC_FTRACE */
ENTRY(function_hook)
cmpq $ftrace_stub, ftrace_trace_function
jnz trace
fgraph_trace:
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
cmpq $ftrace_stub, ftrace_graph_return
jnz ftrace_graph_caller
cmpq $ftrace_graph_entry_stub, ftrace_graph_entry
jnz ftrace_graph_caller
#endif
GLOBAL(ftrace_stub)
retq
trace:
/* save_mcount_regs fills in first two parameters */
save_mcount_regs
/*
* When DYNAMIC_FTRACE is not defined, ARCH_SUPPORTS_FTRACE_OPS is not
* set (see include/asm/ftrace.h and include/linux/ftrace.h). Only the
* ip and parent ip are used and the list function is called when
* function tracing is enabled.
*/
call *ftrace_trace_function
restore_mcount_regs
jmp fgraph_trace
END(function_hook)
#endif /* CONFIG_DYNAMIC_FTRACE */
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
ENTRY(ftrace_graph_caller)
/* Saves rbp into %rdx and fills first parameter */
save_mcount_regs
#ifdef CC_USING_FENTRY
leaq MCOUNT_REG_SIZE+8(%rsp), %rsi
movq $0, %rdx /* No framepointers needed */
#else
/* Save address of the return address of traced function */
leaq 8(%rdx), %rsi
/* ftrace does sanity checks against frame pointers */
movq (%rdx), %rdx
#endif
call prepare_ftrace_return
restore_mcount_regs
retq
END(ftrace_graph_caller)
GLOBAL(return_to_handler)
subq $24, %rsp
/* Save the return values */
movq %rax, (%rsp)
movq %rdx, 8(%rsp)
movq %rbp, %rdi
call ftrace_return_to_handler
movq %rax, %rdi
movq 8(%rsp), %rdx
movq (%rsp), %rax
addq $24, %rsp
jmp *%rdi
#endif