/* GNU/Linux on ARM target support. Copyright 1999, 2000, 2001, 2002, 2003 Free Software Foundation, Inc. This file is part of GDB. 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, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ #include "defs.h" #include "target.h" #include "value.h" #include "gdbtypes.h" #include "floatformat.h" #include "gdbcore.h" #include "frame.h" #include "regcache.h" #include "doublest.h" #include "solib-svr4.h" #include "osabi.h" #include "arm-tdep.h" #include "glibc-tdep.h" /* Under ARM GNU/Linux the traditional way of performing a breakpoint is to execute a particular software interrupt, rather than use a particular undefined instruction to provoke a trap. Upon exection of the software interrupt the kernel stops the inferior with a SIGTRAP, and wakes the debugger. Since ARM GNU/Linux doesn't support Thumb at the moment we only override the ARM breakpoints. */ static const char arm_linux_arm_le_breakpoint[] = { 0x01, 0x00, 0x9f, 0xef }; static const char arm_linux_arm_be_breakpoint[] = { 0xef, 0x9f, 0x00, 0x01 }; /* DEPRECATED_CALL_DUMMY_WORDS: This sequence of words is the instructions mov lr, pc mov pc, r4 swi bkpt_swi Note this is 12 bytes. */ LONGEST arm_linux_call_dummy_words[] = { 0xe1a0e00f, 0xe1a0f004, 0xef9f001 }; /* Description of the longjmp buffer. */ #define ARM_LINUX_JB_ELEMENT_SIZE INT_REGISTER_RAW_SIZE #define ARM_LINUX_JB_PC 21 /* Extract from an array REGBUF containing the (raw) register state a function return value of type TYPE, and copy that, in virtual format, into VALBUF. */ /* FIXME rearnsha/2002-02-23: This function shouldn't be necessary. The ARM generic one should be able to handle the model used by linux and the low-level formatting of the registers should be hidden behind the regcache abstraction. */ static void arm_linux_extract_return_value (struct type *type, char regbuf[], char *valbuf) { /* ScottB: This needs to be looked at to handle the different floating point emulators on ARM GNU/Linux. Right now the code assumes that fetch inferior registers does the right thing for GDB. I suspect this won't handle NWFPE registers correctly, nor will the default ARM version (arm_extract_return_value()). */ int regnum = ((TYPE_CODE_FLT == TYPE_CODE (type)) ? ARM_F0_REGNUM : ARM_A1_REGNUM); memcpy (valbuf, ®buf[DEPRECATED_REGISTER_BYTE (regnum)], TYPE_LENGTH (type)); } /* Note: ScottB This function does not support passing parameters using the FPA variant of the APCS. It passes any floating point arguments in the general registers and/or on the stack. FIXME: This and arm_push_arguments should be merged. However this function breaks on a little endian host, big endian target using the COFF file format. ELF is ok. ScottB. */ /* Addresses for calling Thumb functions have the bit 0 set. Here are some macros to test, set, or clear bit 0 of addresses. */ #define IS_THUMB_ADDR(addr) ((addr) & 1) #define MAKE_THUMB_ADDR(addr) ((addr) | 1) #define UNMAKE_THUMB_ADDR(addr) ((addr) & ~1) static CORE_ADDR arm_linux_push_arguments (int nargs, struct value **args, CORE_ADDR sp, int struct_return, CORE_ADDR struct_addr) { char *fp; int argnum, argreg, nstack_size; /* Walk through the list of args and determine how large a temporary stack is required. Need to take care here as structs may be passed on the stack, and we have to to push them. */ nstack_size = -4 * DEPRECATED_REGISTER_SIZE; /* Some arguments go into A1-A4. */ if (struct_return) /* The struct address goes in A1. */ nstack_size += DEPRECATED_REGISTER_SIZE; /* Walk through the arguments and add their size to nstack_size. */ for (argnum = 0; argnum < nargs; argnum++) { int len; struct type *arg_type; arg_type = check_typedef (VALUE_TYPE (args[argnum])); len = TYPE_LENGTH (arg_type); /* ANSI C code passes float arguments as integers, K&R code passes float arguments as doubles. Correct for this here. */ if (TYPE_CODE_FLT == TYPE_CODE (arg_type) && DEPRECATED_REGISTER_SIZE == len) nstack_size += FP_REGISTER_VIRTUAL_SIZE; else nstack_size += len; } /* Allocate room on the stack, and initialize our stack frame pointer. */ fp = NULL; if (nstack_size > 0) { sp -= nstack_size; fp = (char *) sp; } /* Initialize the integer argument register pointer. */ argreg = ARM_A1_REGNUM; /* The struct_return pointer occupies the first parameter passing register. */ if (struct_return) write_register (argreg++, struct_addr); /* Process arguments from left to right. Store as many as allowed in the parameter passing registers (A1-A4), and save the rest on the temporary stack. */ for (argnum = 0; argnum < nargs; argnum++) { int len; char *val; CORE_ADDR regval; enum type_code typecode; struct type *arg_type, *target_type; arg_type = check_typedef (VALUE_TYPE (args[argnum])); target_type = TYPE_TARGET_TYPE (arg_type); len = TYPE_LENGTH (arg_type); typecode = TYPE_CODE (arg_type); val = (char *) VALUE_CONTENTS (args[argnum]); /* ANSI C code passes float arguments as integers, K&R code passes float arguments as doubles. The .stabs record for for ANSI prototype floating point arguments records the type as FP_INTEGER, while a K&R style (no prototype) .stabs records the type as FP_FLOAT. In this latter case the compiler converts the float arguments to double before calling the function. */ if (TYPE_CODE_FLT == typecode && DEPRECATED_REGISTER_SIZE == len) { DOUBLEST dblval; dblval = deprecated_extract_floating (val, len); len = TARGET_DOUBLE_BIT / TARGET_CHAR_BIT; val = alloca (len); deprecated_store_floating (val, len, dblval); } /* If the argument is a pointer to a function, and it is a Thumb function, set the low bit of the pointer. */ if (TYPE_CODE_PTR == typecode && NULL != target_type && TYPE_CODE_FUNC == TYPE_CODE (target_type)) { CORE_ADDR regval = extract_unsigned_integer (val, len); if (arm_pc_is_thumb (regval)) store_unsigned_integer (val, len, MAKE_THUMB_ADDR (regval)); } /* Copy the argument to general registers or the stack in register-sized pieces. Large arguments are split between registers and stack. */ while (len > 0) { int partial_len = len < DEPRECATED_REGISTER_SIZE ? len : DEPRECATED_REGISTER_SIZE; if (argreg <= ARM_LAST_ARG_REGNUM) { /* It's an argument being passed in a general register. */ regval = extract_unsigned_integer (val, partial_len); write_register (argreg++, regval); } else { /* Push the arguments onto the stack. */ write_memory ((CORE_ADDR) fp, val, DEPRECATED_REGISTER_SIZE); fp += DEPRECATED_REGISTER_SIZE; } len -= partial_len; val += partial_len; } } /* Return adjusted stack pointer. */ return sp; } /* Dynamic Linking on ARM GNU/Linux -------------------------------- Note: PLT = procedure linkage table GOT = global offset table As much as possible, ELF dynamic linking defers the resolution of jump/call addresses until the last minute. The technique used is inspired by the i386 ELF design, and is based on the following constraints. 1) The calling technique should not force a change in the assembly code produced for apps; it MAY cause changes in the way assembly code is produced for position independent code (i.e. shared libraries). 2) The technique must be such that all executable areas must not be modified; and any modified areas must not be executed. To do this, there are three steps involved in a typical jump: 1) in the code 2) through the PLT 3) using a pointer from the GOT When the executable or library is first loaded, each GOT entry is initialized to point to the code which implements dynamic name resolution and code finding. This is normally a function in the program interpreter (on ARM GNU/Linux this is usually ld-linux.so.2, but it does not have to be). On the first invocation, the function is located and the GOT entry is replaced with the real function address. Subsequent calls go through steps 1, 2 and 3 and end up calling the real code. 1) In the code: b function_call bl function_call This is typical ARM code using the 26 bit relative branch or branch and link instructions. The target of the instruction (function_call is usually the address of the function to be called. In position independent code, the target of the instruction is actually an entry in the PLT when calling functions in a shared library. Note that this call is identical to a normal function call, only the target differs. 2) In the PLT: The PLT is a synthetic area, created by the linker. It exists in both executables and libraries. It is an array of stubs, one per imported function call. It looks like this: PLT[0]: str lr, [sp, #-4]! @push the return address (lr) ldr lr, [pc, #16] @load from 6 words ahead add lr, pc, lr @form an address for GOT[0] ldr pc, [lr, #8]! @jump to the contents of that addr The return address (lr) is pushed on the stack and used for calculations. The load on the second line loads the lr with &GOT[3] - . - 20. The addition on the third leaves: lr = (&GOT[3] - . - 20) + (. + 8) lr = (&GOT[3] - 12) lr = &GOT[0] On the fourth line, the pc and lr are both updated, so that: pc = GOT[2] lr = &GOT[0] + 8 = &GOT[2] NOTE: PLT[0] borrows an offset .word from PLT[1]. This is a little "tight", but allows us to keep all the PLT entries the same size. PLT[n+1]: ldr ip, [pc, #4] @load offset from gotoff add ip, pc, ip @add the offset to the pc ldr pc, [ip] @jump to that address gotoff: .word GOT[n+3] - . The load on the first line, gets an offset from the fourth word of the PLT entry. The add on the second line makes ip = &GOT[n+3], which contains either a pointer to PLT[0] (the fixup trampoline) or a pointer to the actual code. 3) In the GOT: The GOT contains helper pointers for both code (PLT) fixups and data fixups. The first 3 entries of the GOT are special. The next M entries (where M is the number of entries in the PLT) belong to the PLT fixups. The next D (all remaining) entries belong to various data fixups. The actual size of the GOT is 3 + M + D. The GOT is also a synthetic area, created by the linker. It exists in both executables and libraries. When the GOT is first initialized , all the GOT entries relating to PLT fixups are pointing to code back at PLT[0]. The special entries in the GOT are: GOT[0] = linked list pointer used by the dynamic loader GOT[1] = pointer to the reloc table for this module GOT[2] = pointer to the fixup/resolver code The first invocation of function call comes through and uses the fixup/resolver code. On the entry to the fixup/resolver code: ip = &GOT[n+3] lr = &GOT[2] stack[0] = return address (lr) of the function call [r0, r1, r2, r3] are still the arguments to the function call This is enough information for the fixup/resolver code to work with. Before the fixup/resolver code returns, it actually calls the requested function and repairs &GOT[n+3]. */ /* Fetch, and possibly build, an appropriate link_map_offsets structure for ARM linux targets using the struct offsets defined in . Note, however, that link.h is not actually referred to in this file. Instead, the relevant structs offsets were obtained from examining link.h. (We can't refer to link.h from this file because the host system won't necessarily have it, or if it does, the structs which it defines will refer to the host system, not the target). */ static struct link_map_offsets * arm_linux_svr4_fetch_link_map_offsets (void) { static struct link_map_offsets lmo; static struct link_map_offsets *lmp = 0; if (lmp == 0) { lmp = &lmo; lmo.r_debug_size = 8; /* Actual size is 20, but this is all we need. */ lmo.r_map_offset = 4; lmo.r_map_size = 4; lmo.link_map_size = 20; /* Actual size is 552, but this is all we need. */ lmo.l_addr_offset = 0; lmo.l_addr_size = 4; lmo.l_name_offset = 4; lmo.l_name_size = 4; lmo.l_next_offset = 12; lmo.l_next_size = 4; lmo.l_prev_offset = 16; lmo.l_prev_size = 4; } return lmp; } CORE_ADDR arm_linux_skip_solib_resolver (CORE_ADDR pc) { return glibc_skip_solib_resolver (pc); } /* The constants below were determined by examining the following files in the linux kernel sources: arch/arm/kernel/signal.c - see SWI_SYS_SIGRETURN and SWI_SYS_RT_SIGRETURN include/asm-arm/unistd.h - see __NR_sigreturn, __NR_rt_sigreturn, and __NR_SYSCALL_BASE */ #define ARM_LINUX_SIGRETURN_INSTR 0xef900077 #define ARM_LINUX_RT_SIGRETURN_INSTR 0xef9000ad /* arm_linux_in_sigtramp determines if PC points at one of the instructions which cause control to return to the Linux kernel upon return from a signal handler. FUNC_NAME is unused. */ int arm_linux_in_sigtramp (CORE_ADDR pc, char *func_name) { unsigned long inst; inst = read_memory_integer (pc, 4); return (inst == ARM_LINUX_SIGRETURN_INSTR || inst == ARM_LINUX_RT_SIGRETURN_INSTR); } /* arm_linux_sigcontext_register_address returns the address in the sigcontext of register REGNO given a stack pointer value SP and program counter value PC. The value 0 is returned if PC is not pointing at one of the signal return instructions or if REGNO is not saved in the sigcontext struct. */ CORE_ADDR arm_linux_sigcontext_register_address (CORE_ADDR sp, CORE_ADDR pc, int regno) { unsigned long inst; CORE_ADDR reg_addr = 0; inst = read_memory_integer (pc, 4); if (inst == ARM_LINUX_SIGRETURN_INSTR || inst == ARM_LINUX_RT_SIGRETURN_INSTR) { CORE_ADDR sigcontext_addr; /* The sigcontext structure is at different places for the two signal return instructions. For ARM_LINUX_SIGRETURN_INSTR, it starts at the SP value. For ARM_LINUX_RT_SIGRETURN_INSTR, it is at SP+8. For the latter instruction, it may also be the case that the address of this structure may be determined by reading the 4 bytes at SP, but I'm not convinced this is reliable. In any event, these magic constants (0 and 8) may be determined by examining struct sigframe and struct rt_sigframe in arch/arm/kernel/signal.c in the Linux kernel sources. */ if (inst == ARM_LINUX_RT_SIGRETURN_INSTR) sigcontext_addr = sp + 8; else /* inst == ARM_LINUX_SIGRETURN_INSTR */ sigcontext_addr = sp + 0; /* The layout of the sigcontext structure for ARM GNU/Linux is in include/asm-arm/sigcontext.h in the Linux kernel sources. There are three 4-byte fields which precede the saved r0 field. (This accounts for the 12 in the code below.) The sixteen registers (4 bytes per field) follow in order. The PSR value follows the sixteen registers which accounts for the constant 19 below. */ if (0 <= regno && regno <= ARM_PC_REGNUM) reg_addr = sigcontext_addr + 12 + (4 * regno); else if (regno == ARM_PS_REGNUM) reg_addr = sigcontext_addr + 19 * 4; } return reg_addr; } static void arm_linux_init_abi (struct gdbarch_info info, struct gdbarch *gdbarch) { struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch); tdep->lowest_pc = 0x8000; if (info.byte_order == BFD_ENDIAN_BIG) tdep->arm_breakpoint = arm_linux_arm_be_breakpoint; else tdep->arm_breakpoint = arm_linux_arm_le_breakpoint; tdep->arm_breakpoint_size = sizeof (arm_linux_arm_le_breakpoint); tdep->fp_model = ARM_FLOAT_FPA; tdep->jb_pc = ARM_LINUX_JB_PC; tdep->jb_elt_size = ARM_LINUX_JB_ELEMENT_SIZE; set_solib_svr4_fetch_link_map_offsets (gdbarch, arm_linux_svr4_fetch_link_map_offsets); set_gdbarch_deprecated_call_dummy_words (gdbarch, arm_linux_call_dummy_words); set_gdbarch_deprecated_sizeof_call_dummy_words (gdbarch, sizeof (arm_linux_call_dummy_words)); /* The following two overrides shouldn't be needed. */ set_gdbarch_deprecated_extract_return_value (gdbarch, arm_linux_extract_return_value); set_gdbarch_deprecated_push_arguments (gdbarch, arm_linux_push_arguments); /* Shared library handling. */ set_gdbarch_in_solib_call_trampoline (gdbarch, in_plt_section); set_gdbarch_skip_trampoline_code (gdbarch, find_solib_trampoline_target); } void _initialize_arm_linux_tdep (void) { gdbarch_register_osabi (bfd_arch_arm, 0, GDB_OSABI_LINUX, arm_linux_init_abi); }