X-Git-Url: http://drtracing.org/?a=blobdiff_plain;f=bfd%2Fdwarf2.c;h=dd3568a8532c2254084ddc3ee6a12b26d45693b4;hb=d835a58baae720abe909795cb68763040d1750a8;hp=b05711435908ce5d2f1b2a79dcb35773c1191c59;hpb=0112cd268b205d8176b8b4d00988a334822956cf;p=deliverable%2Fbinutils-gdb.git diff --git a/bfd/dwarf2.c b/bfd/dwarf2.c index b057114359..dd3568a853 100644 --- a/bfd/dwarf2.c +++ b/bfd/dwarf2.c @@ -1,6 +1,5 @@ /* DWARF 2 support. - Copyright 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, - 2004, 2005, 2006 Free Software Foundation, Inc. + Copyright (C) 1994-2020 Free Software Foundation, Inc. Adapted from gdb/dwarf2read.c by Gavin Koch of Cygnus Solutions (gavin@cygnus.com). @@ -17,7 +16,7 @@ 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 + the Free Software Foundation; either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but @@ -27,14 +26,16 @@ 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., 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */ + Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, + MA 02110-1301, USA. */ -#include "bfd.h" #include "sysdep.h" +#include "bfd.h" #include "libiberty.h" #include "libbfd.h" #include "elf-bfd.h" -#include "elf/dwarf2.h" +#include "dwarf2.h" +#include "hashtab.h" /* The data in the .debug_line statement prologue looks like this. */ @@ -44,6 +45,7 @@ struct line_head unsigned short version; bfd_vma prologue_length; unsigned char minimum_instruction_length; + unsigned char maximum_ops_per_insn; unsigned char default_is_stmt; int line_base; unsigned char line_range; @@ -74,68 +76,127 @@ struct dwarf_block bfd_byte *data; }; -struct loadable_section +struct adjusted_section { asection *section; bfd_vma adj_vma; }; -struct dwarf2_debug +struct dwarf2_debug_file { - /* A list of all previously read comp_units. */ - struct comp_unit *all_comp_units; + /* The actual bfd from which debug info was loaded. Might be + different to orig_bfd because of gnu_debuglink sections. */ + bfd *bfd_ptr; - /* The next unread compilation unit within the .debug_info section. - Zero indicates that the .debug_info section has not been loaded - into a buffer yet. */ - bfd_byte *info_ptr; + /* Pointer to the symbol table. */ + asymbol **syms; - /* Pointer to the end of the .debug_info section memory buffer. */ - bfd_byte *info_ptr_end; + /* The current info pointer for the .debug_info section being parsed. */ + bfd_byte *info_ptr; - /* Pointer to the section and address of the beginning of the - section. */ - asection *sec; - bfd_byte *sec_info_ptr; + /* A pointer to the memory block allocated for .debug_info sections. */ + bfd_byte *dwarf_info_buffer; - /* Pointer to the symbol table. */ - asymbol **syms; + /* Length of the loaded .debug_info sections. */ + bfd_size_type dwarf_info_size; /* Pointer to the .debug_abbrev section loaded into memory. */ bfd_byte *dwarf_abbrev_buffer; /* Length of the loaded .debug_abbrev section. */ - unsigned long dwarf_abbrev_size; + bfd_size_type dwarf_abbrev_size; /* Buffer for decode_line_info. */ bfd_byte *dwarf_line_buffer; /* Length of the loaded .debug_line section. */ - unsigned long dwarf_line_size; + bfd_size_type dwarf_line_size; /* Pointer to the .debug_str section loaded into memory. */ bfd_byte *dwarf_str_buffer; /* Length of the loaded .debug_str section. */ - unsigned long dwarf_str_size; + bfd_size_type dwarf_str_size; + + /* Pointer to the .debug_line_str section loaded into memory. */ + bfd_byte *dwarf_line_str_buffer; - /* Pointer to the .debug_ranges section loaded into memory. */ + /* Length of the loaded .debug_line_str section. */ + bfd_size_type dwarf_line_str_size; + + /* Pointer to the .debug_ranges section loaded into memory. */ bfd_byte *dwarf_ranges_buffer; - /* Length of the loaded .debug_ranges section. */ - unsigned long dwarf_ranges_size; + /* Length of the loaded .debug_ranges section. */ + bfd_size_type dwarf_ranges_size; + + /* A list of all previously read comp_units. */ + struct comp_unit *all_comp_units; + + /* Last comp unit in list above. */ + struct comp_unit *last_comp_unit; + + /* Line table at line_offset zero. */ + struct line_info_table *line_table; + + /* Hash table to map offsets to decoded abbrevs. */ + htab_t abbrev_offsets; +}; + +struct dwarf2_debug +{ + /* Names of the debug sections. */ + const struct dwarf_debug_section *debug_sections; + + /* Per-file stuff. */ + struct dwarf2_debug_file f, alt; + + /* Pointer to the original bfd for which debug was loaded. This is what + we use to compare and so check that the cached debug data is still + valid - it saves having to possibly dereference the gnu_debuglink each + time. */ + bfd *orig_bfd; /* If the most recent call to bfd_find_nearest_line was given an address in an inlined function, preserve a pointer into the calling chain for subsequent calls to bfd_find_inliner_info to - use. */ + use. */ struct funcinfo *inliner_chain; - /* Number of loadable sections. */ - unsigned int loadable_section_count; + /* Section VMAs at the time the stash was built. */ + bfd_vma *sec_vma; + /* Number of sections in the SEC_VMA table. */ + unsigned int sec_vma_count; + + /* Number of sections whose VMA we must adjust. */ + int adjusted_section_count; + + /* Array of sections with adjusted VMA. */ + struct adjusted_section *adjusted_sections; + + /* Number of times find_line is called. This is used in + the heuristic for enabling the info hash tables. */ + int info_hash_count; + +#define STASH_INFO_HASH_TRIGGER 100 - /* Array of loadable sections. */ - struct loadable_section *loadable_sections; + /* Hash table mapping symbol names to function infos. */ + struct info_hash_table *funcinfo_hash_table; + + /* Hash table mapping symbol names to variable infos. */ + struct info_hash_table *varinfo_hash_table; + + /* Head of comp_unit list in the last hash table update. */ + struct comp_unit *hash_units_head; + + /* Status of info hash. */ + int info_hash_status; +#define STASH_INFO_HASH_OFF 0 +#define STASH_INFO_HASH_ON 1 +#define STASH_INFO_HASH_DISABLED 2 + + /* True if we opened bfd_ptr. */ + bfd_boolean close_on_cleanup; }; struct arange @@ -153,6 +214,10 @@ struct comp_unit /* Chain the previously read compilation units. */ struct comp_unit *next_unit; + /* Likewise, chain the compilation unit read after this one. + The comp units are stored in reversed reading order. */ + struct comp_unit *prev_unit; + /* Keep the bfd convenient (for memory allocation). */ bfd *abfd; @@ -166,6 +231,9 @@ struct comp_unit /* The abbrev hash table. */ struct abbrev_info **abbrevs; + /* DW_AT_language. */ + int lang; + /* Note that an error was found by comp_unit_find_nearest_line. */ int error; @@ -194,11 +262,21 @@ struct comp_unit /* A list of the functions found in this comp. unit. */ struct funcinfo *function_table; + /* A table of function information references searchable by address. */ + struct lookup_funcinfo *lookup_funcinfo_table; + + /* Number of functions in the function_table and sorted_function_table. */ + bfd_size_type number_of_functions; + /* A list of the variables found in this comp. unit. */ struct varinfo *variable_table; - /* Pointer to dwarf2_debug structure. */ + /* Pointers to dwarf2_debug structures. */ struct dwarf2_debug *stash; + struct dwarf2_debug_file *file; + + /* DWARF format version for this unit - from unit header. */ + int version; /* Address size for this unit - from unit header. */ unsigned char addr_size; @@ -209,6 +287,9 @@ struct comp_unit /* Base address for this unit - from DW_AT_low_pc attribute of DW_TAG_compile_unit DIE */ bfd_vma base_address; + + /* TRUE if symbols are cached in hash table for faster lookup by name. */ + bfd_boolean cached; }; /* This data structure holds the information of an abbrev. */ @@ -226,8 +307,78 @@ struct attr_abbrev { enum dwarf_attribute name; enum dwarf_form form; + bfd_vma implicit_const; +}; + +/* Map of uncompressed DWARF debug section name to compressed one. It + is terminated by NULL uncompressed_name. */ + +const struct dwarf_debug_section dwarf_debug_sections[] = +{ + { ".debug_abbrev", ".zdebug_abbrev" }, + { ".debug_aranges", ".zdebug_aranges" }, + { ".debug_frame", ".zdebug_frame" }, + { ".debug_info", ".zdebug_info" }, + { ".debug_info", ".zdebug_info" }, + { ".debug_line", ".zdebug_line" }, + { ".debug_loc", ".zdebug_loc" }, + { ".debug_macinfo", ".zdebug_macinfo" }, + { ".debug_macro", ".zdebug_macro" }, + { ".debug_pubnames", ".zdebug_pubnames" }, + { ".debug_pubtypes", ".zdebug_pubtypes" }, + { ".debug_ranges", ".zdebug_ranges" }, + { ".debug_static_func", ".zdebug_static_func" }, + { ".debug_static_vars", ".zdebug_static_vars" }, + { ".debug_str", ".zdebug_str", }, + { ".debug_str", ".zdebug_str", }, + { ".debug_line_str", ".zdebug_line_str", }, + { ".debug_types", ".zdebug_types" }, + /* GNU DWARF 1 extensions */ + { ".debug_sfnames", ".zdebug_sfnames" }, + { ".debug_srcinfo", ".zebug_srcinfo" }, + /* SGI/MIPS DWARF 2 extensions */ + { ".debug_funcnames", ".zdebug_funcnames" }, + { ".debug_typenames", ".zdebug_typenames" }, + { ".debug_varnames", ".zdebug_varnames" }, + { ".debug_weaknames", ".zdebug_weaknames" }, + { NULL, NULL }, +}; + +/* NB/ Numbers in this enum must match up with indices + into the dwarf_debug_sections[] array above. */ +enum dwarf_debug_section_enum +{ + debug_abbrev = 0, + debug_aranges, + debug_frame, + debug_info, + debug_info_alt, + debug_line, + debug_loc, + debug_macinfo, + debug_macro, + debug_pubnames, + debug_pubtypes, + debug_ranges, + debug_static_func, + debug_static_vars, + debug_str, + debug_str_alt, + debug_line_str, + debug_types, + debug_sfnames, + debug_srcinfo, + debug_funcnames, + debug_typenames, + debug_varnames, + debug_weaknames, + debug_max }; +/* A static assertion. */ +extern int dwarf_debug_section_assert[ARRAY_SIZE (dwarf_debug_sections) + == debug_max + 1 ? 1 : -1]; + #ifndef ABBREV_HASH_SIZE #define ABBREV_HASH_SIZE 121 #endif @@ -235,131 +386,502 @@ struct attr_abbrev #define ATTR_ALLOC_CHUNK 4 #endif -/* VERBATIM - The following function up to the END VERBATIM mark are - copied directly from dwarf2read.c. */ +/* Variable and function hash tables. This is used to speed up look-up + in lookup_symbol_in_var_table() and lookup_symbol_in_function_table(). + In order to share code between variable and function infos, we use + a list of untyped pointer for all variable/function info associated with + a symbol. We waste a bit of memory for list with one node but that + simplifies the code. */ + +struct info_list_node +{ + struct info_list_node *next; + void *info; +}; + +/* Info hash entry. */ +struct info_hash_entry +{ + struct bfd_hash_entry root; + struct info_list_node *head; +}; + +struct info_hash_table +{ + struct bfd_hash_table base; +}; + +/* Function to create a new entry in info hash table. */ + +static struct bfd_hash_entry * +info_hash_table_newfunc (struct bfd_hash_entry *entry, + struct bfd_hash_table *table, + const char *string) +{ + struct info_hash_entry *ret = (struct info_hash_entry *) entry; + + /* Allocate the structure if it has not already been allocated by a + derived class. */ + if (ret == NULL) + { + ret = (struct info_hash_entry *) bfd_hash_allocate (table, + sizeof (* ret)); + if (ret == NULL) + return NULL; + } + + /* Call the allocation method of the base class. */ + ret = ((struct info_hash_entry *) + bfd_hash_newfunc ((struct bfd_hash_entry *) ret, table, string)); + + /* Initialize the local fields here. */ + if (ret) + ret->head = NULL; + + return (struct bfd_hash_entry *) ret; +} + +/* Function to create a new info hash table. It returns a pointer to the + newly created table or NULL if there is any error. We need abfd + solely for memory allocation. */ + +static struct info_hash_table * +create_info_hash_table (bfd *abfd) +{ + struct info_hash_table *hash_table; + + hash_table = ((struct info_hash_table *) + bfd_alloc (abfd, sizeof (struct info_hash_table))); + if (!hash_table) + return hash_table; + + if (!bfd_hash_table_init (&hash_table->base, info_hash_table_newfunc, + sizeof (struct info_hash_entry))) + { + bfd_release (abfd, hash_table); + return NULL; + } + + return hash_table; +} + +/* Insert an info entry into an info hash table. We do not check of + duplicate entries. Also, the caller need to guarantee that the + right type of info in inserted as info is passed as a void* pointer. + This function returns true if there is no error. */ + +static bfd_boolean +insert_info_hash_table (struct info_hash_table *hash_table, + const char *key, + void *info, + bfd_boolean copy_p) +{ + struct info_hash_entry *entry; + struct info_list_node *node; + + entry = (struct info_hash_entry*) bfd_hash_lookup (&hash_table->base, + key, TRUE, copy_p); + if (!entry) + return FALSE; + + node = (struct info_list_node *) bfd_hash_allocate (&hash_table->base, + sizeof (*node)); + if (!node) + return FALSE; + + node->info = info; + node->next = entry->head; + entry->head = node; + + return TRUE; +} + +/* Look up an info entry list from an info hash table. Return NULL + if there is none. */ + +static struct info_list_node * +lookup_info_hash_table (struct info_hash_table *hash_table, const char *key) +{ + struct info_hash_entry *entry; + + entry = (struct info_hash_entry*) bfd_hash_lookup (&hash_table->base, key, + FALSE, FALSE); + return entry ? entry->head : NULL; +} + +/* Read a section into its appropriate place in the dwarf2_debug + struct (indicated by SECTION_BUFFER and SECTION_SIZE). If SYMS is + not NULL, use bfd_simple_get_relocated_section_contents to read the + section contents, otherwise use bfd_get_section_contents. Fail if + the located section does not contain at least OFFSET bytes. */ + +static bfd_boolean +read_section (bfd * abfd, + const struct dwarf_debug_section *sec, + asymbol ** syms, + bfd_uint64_t offset, + bfd_byte ** section_buffer, + bfd_size_type * section_size) +{ + asection *msec; + const char *section_name = sec->uncompressed_name; + bfd_byte *contents = *section_buffer; + bfd_size_type amt; + + /* The section may have already been read. */ + if (contents == NULL) + { + msec = bfd_get_section_by_name (abfd, section_name); + if (! msec) + { + section_name = sec->compressed_name; + if (section_name != NULL) + msec = bfd_get_section_by_name (abfd, section_name); + } + if (! msec) + { + _bfd_error_handler (_("DWARF error: can't find %s section."), + sec->uncompressed_name); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } + + *section_size = msec->rawsize ? msec->rawsize : msec->size; + /* Paranoia - alloc one extra so that we can make sure a string + section is NUL terminated. */ + amt = *section_size + 1; + if (amt == 0) + { + bfd_set_error (bfd_error_no_memory); + return FALSE; + } + contents = (bfd_byte *) bfd_malloc (amt); + if (contents == NULL) + return FALSE; + if (syms + ? !bfd_simple_get_relocated_section_contents (abfd, msec, contents, + syms) + : !bfd_get_section_contents (abfd, msec, contents, 0, *section_size)) + { + free (contents); + return FALSE; + } + contents[*section_size] = 0; + *section_buffer = contents; + } + + /* It is possible to get a bad value for the offset into the section + that the client wants. Validate it here to avoid trouble later. */ + if (offset != 0 && offset >= *section_size) + { + /* xgettext: c-format */ + _bfd_error_handler (_("DWARF error: offset (%" PRIu64 ")" + " greater than or equal to %s size (%" PRIu64 ")"), + (uint64_t) offset, section_name, + (uint64_t) *section_size); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } + + return TRUE; +} /* Read dwarf information from a buffer. */ static unsigned int -read_1_byte (bfd *abfd ATTRIBUTE_UNUSED, bfd_byte *buf) +read_1_byte (bfd *abfd ATTRIBUTE_UNUSED, bfd_byte *buf, bfd_byte *end) { + if (buf + 1 > end) + return 0; return bfd_get_8 (abfd, buf); } static int -read_1_signed_byte (bfd *abfd ATTRIBUTE_UNUSED, bfd_byte *buf) +read_1_signed_byte (bfd *abfd ATTRIBUTE_UNUSED, bfd_byte *buf, bfd_byte *end) { + if (buf + 1 > end) + return 0; return bfd_get_signed_8 (abfd, buf); } static unsigned int -read_2_bytes (bfd *abfd, bfd_byte *buf) +read_2_bytes (bfd *abfd, bfd_byte *buf, bfd_byte *end) { + if (buf + 2 > end) + return 0; return bfd_get_16 (abfd, buf); } static unsigned int -read_4_bytes (bfd *abfd, bfd_byte *buf) +read_4_bytes (bfd *abfd, bfd_byte *buf, bfd_byte *end) { + if (buf + 4 > end) + return 0; return bfd_get_32 (abfd, buf); } static bfd_uint64_t -read_8_bytes (bfd *abfd, bfd_byte *buf) +read_8_bytes (bfd *abfd, bfd_byte *buf, bfd_byte *end) { + if (buf + 8 > end) + return 0; return bfd_get_64 (abfd, buf); } static bfd_byte * -read_n_bytes (bfd *abfd ATTRIBUTE_UNUSED, - bfd_byte *buf, - unsigned int size ATTRIBUTE_UNUSED) +read_n_bytes (bfd_byte * buf, + bfd_byte * end, + struct dwarf_block * block) { - /* If the size of a host char is 8 bits, we can return a pointer - to the buffer, otherwise we have to copy the data to a buffer - allocated on the temporary obstack. */ - return buf; + unsigned int size = block->size; + bfd_byte * block_end = buf + size; + + if (block_end > end || block_end < buf) + { + block->data = NULL; + block->size = 0; + return end; + } + else + { + block->data = buf; + return block_end; + } } +/* Scans a NUL terminated string starting at BUF, returning a pointer to it. + Returns the number of characters in the string, *including* the NUL byte, + in BYTES_READ_PTR. This value is set even if the function fails. Bytes + at or beyond BUF_END will not be read. Returns NULL if there was a + problem, or if the string is empty. */ + static char * -read_string (bfd *abfd ATTRIBUTE_UNUSED, - bfd_byte *buf, - unsigned int *bytes_read_ptr) +read_string (bfd * abfd ATTRIBUTE_UNUSED, + bfd_byte * buf, + bfd_byte * buf_end, + unsigned int * bytes_read_ptr) { - /* Return a pointer to the embedded string. */ - char *str = (char *) buf; + bfd_byte *str = buf; + + if (buf >= buf_end) + { + * bytes_read_ptr = 0; + return NULL; + } + if (*str == '\0') { - *bytes_read_ptr = 1; + * bytes_read_ptr = 1; + return NULL; + } + + while (buf < buf_end) + if (* buf ++ == 0) + { + * bytes_read_ptr = buf - str; + return (char *) str; + } + + * bytes_read_ptr = buf - str; + return NULL; +} + +/* Reads an offset from BUF and then locates the string at this offset + inside the debug string section. Returns a pointer to the string. + Returns the number of bytes read from BUF, *not* the length of the string, + in BYTES_READ_PTR. This value is set even if the function fails. Bytes + at or beyond BUF_END will not be read from BUF. Returns NULL if there was + a problem, or if the string is empty. Does not check for NUL termination + of the string. */ + +static char * +read_indirect_string (struct comp_unit * unit, + bfd_byte * buf, + bfd_byte * buf_end, + unsigned int * bytes_read_ptr) +{ + bfd_uint64_t offset; + struct dwarf2_debug *stash = unit->stash; + struct dwarf2_debug_file *file = unit->file; + char *str; + + if (buf + unit->offset_size > buf_end) + { + * bytes_read_ptr = 0; return NULL; } - *bytes_read_ptr = strlen (str) + 1; + if (unit->offset_size == 4) + offset = read_4_bytes (unit->abfd, buf, buf_end); + else + offset = read_8_bytes (unit->abfd, buf, buf_end); + + *bytes_read_ptr = unit->offset_size; + + if (! read_section (unit->abfd, &stash->debug_sections[debug_str], + file->syms, offset, + &file->dwarf_str_buffer, &file->dwarf_str_size)) + return NULL; + + str = (char *) file->dwarf_str_buffer + offset; + if (*str == '\0') + return NULL; return str; } +/* Like read_indirect_string but from .debug_line_str section. */ + static char * -read_indirect_string (struct comp_unit* unit, - bfd_byte *buf, - unsigned int *bytes_read_ptr) +read_indirect_line_string (struct comp_unit * unit, + bfd_byte * buf, + bfd_byte * buf_end, + unsigned int * bytes_read_ptr) { bfd_uint64_t offset; struct dwarf2_debug *stash = unit->stash; + struct dwarf2_debug_file *file = unit->file; char *str; + if (buf + unit->offset_size > buf_end) + { + * bytes_read_ptr = 0; + return NULL; + } + if (unit->offset_size == 4) - offset = read_4_bytes (unit->abfd, buf); + offset = read_4_bytes (unit->abfd, buf, buf_end); else - offset = read_8_bytes (unit->abfd, buf); + offset = read_8_bytes (unit->abfd, buf, buf_end); + *bytes_read_ptr = unit->offset_size; - if (! stash->dwarf_str_buffer) + if (! read_section (unit->abfd, &stash->debug_sections[debug_line_str], + file->syms, offset, + &file->dwarf_line_str_buffer, + &file->dwarf_line_str_size)) + return NULL; + + str = (char *) file->dwarf_line_str_buffer + offset; + if (*str == '\0') + return NULL; + return str; +} + +/* Like read_indirect_string but uses a .debug_str located in + an alternate file pointed to by the .gnu_debugaltlink section. + Used to impement DW_FORM_GNU_strp_alt. */ + +static char * +read_alt_indirect_string (struct comp_unit * unit, + bfd_byte * buf, + bfd_byte * buf_end, + unsigned int * bytes_read_ptr) +{ + bfd_uint64_t offset; + struct dwarf2_debug *stash = unit->stash; + char *str; + + if (buf + unit->offset_size > buf_end) { - asection *msec; - bfd *abfd = unit->abfd; - bfd_size_type sz; + * bytes_read_ptr = 0; + return NULL; + } - msec = bfd_get_section_by_name (abfd, ".debug_str"); - if (! msec) - { - (*_bfd_error_handler) - (_("Dwarf Error: Can't find .debug_str section.")); - bfd_set_error (bfd_error_bad_value); - return NULL; - } + if (unit->offset_size == 4) + offset = read_4_bytes (unit->abfd, buf, buf_end); + else + offset = read_8_bytes (unit->abfd, buf, buf_end); + + *bytes_read_ptr = unit->offset_size; + + if (stash->alt.bfd_ptr == NULL) + { + bfd *debug_bfd; + char *debug_filename = bfd_follow_gnu_debugaltlink (unit->abfd, DEBUGDIR); - sz = msec->rawsize ? msec->rawsize : msec->size; - stash->dwarf_str_size = sz; - stash->dwarf_str_buffer = bfd_alloc (abfd, sz); - if (! stash->dwarf_str_buffer) + if (debug_filename == NULL) return NULL; - if (! bfd_get_section_contents (abfd, msec, stash->dwarf_str_buffer, - 0, sz)) + debug_bfd = bfd_openr (debug_filename, NULL); + free (debug_filename); + if (debug_bfd == NULL) + /* FIXME: Should we report our failure to follow the debuglink ? */ return NULL; - } - if (offset >= stash->dwarf_str_size) - { - (*_bfd_error_handler) (_("Dwarf Error: DW_FORM_strp offset (%lu) greater than or equal to .debug_str size (%lu)."), - (unsigned long) offset, stash->dwarf_str_size); - bfd_set_error (bfd_error_bad_value); - return NULL; + if (!bfd_check_format (debug_bfd, bfd_object)) + { + bfd_close (debug_bfd); + return NULL; + } + stash->alt.bfd_ptr = debug_bfd; } - str = (char *) stash->dwarf_str_buffer + offset; + if (! read_section (unit->stash->alt.bfd_ptr, + stash->debug_sections + debug_str_alt, + stash->alt.syms, offset, + &stash->alt.dwarf_str_buffer, + &stash->alt.dwarf_str_size)) + return NULL; + + str = (char *) stash->alt.dwarf_str_buffer + offset; if (*str == '\0') return NULL; + return str; } -/* END VERBATIM */ +/* Resolve an alternate reference from UNIT at OFFSET. + Returns a pointer into the loaded alternate CU upon success + or NULL upon failure. */ + +static bfd_byte * +read_alt_indirect_ref (struct comp_unit * unit, + bfd_uint64_t offset) +{ + struct dwarf2_debug *stash = unit->stash; + + if (stash->alt.bfd_ptr == NULL) + { + bfd *debug_bfd; + char *debug_filename = bfd_follow_gnu_debugaltlink (unit->abfd, DEBUGDIR); + + if (debug_filename == NULL) + return NULL; + + debug_bfd = bfd_openr (debug_filename, NULL); + free (debug_filename); + if (debug_bfd == NULL) + /* FIXME: Should we report our failure to follow the debuglink ? */ + return NULL; + + if (!bfd_check_format (debug_bfd, bfd_object)) + { + bfd_close (debug_bfd); + return NULL; + } + stash->alt.bfd_ptr = debug_bfd; + } + + if (! read_section (unit->stash->alt.bfd_ptr, + stash->debug_sections + debug_info_alt, + stash->alt.syms, offset, + &stash->alt.dwarf_info_buffer, + &stash->alt.dwarf_info_size)) + return NULL; + + return stash->alt.dwarf_info_buffer + offset; +} static bfd_uint64_t -read_address (struct comp_unit *unit, bfd_byte *buf) +read_address (struct comp_unit *unit, bfd_byte *buf, bfd_byte * buf_end) { - int signed_vma = get_elf_backend_data (unit->abfd)->sign_extend_vma; + int signed_vma = 0; + + if (bfd_get_flavour (unit->abfd) == bfd_target_elf_flavour) + signed_vma = get_elf_backend_data (unit->abfd)->sign_extend_vma; + + if (buf + unit->addr_size > buf_end) + return 0; if (signed_vma) { @@ -413,112 +935,155 @@ lookup_abbrev (unsigned int number, struct abbrev_info **abbrevs) return NULL; } +/* We keep a hash table to map .debug_abbrev section offsets to the + array of abbrevs, so that compilation units using the same set of + abbrevs do not waste memory. */ + +struct abbrev_offset_entry +{ + size_t offset; + struct abbrev_info **abbrevs; +}; + +static hashval_t +hash_abbrev (const void *p) +{ + const struct abbrev_offset_entry *ent = p; + return htab_hash_pointer ((void *) ent->offset); +} + +static int +eq_abbrev (const void *pa, const void *pb) +{ + const struct abbrev_offset_entry *a = pa; + const struct abbrev_offset_entry *b = pb; + return a->offset == b->offset; +} + +static void +del_abbrev (void *p) +{ + struct abbrev_offset_entry *ent = p; + struct abbrev_info **abbrevs = ent->abbrevs; + size_t i; + + for (i = 0; i < ABBREV_HASH_SIZE; i++) + { + struct abbrev_info *abbrev = abbrevs[i]; + + while (abbrev) + { + free (abbrev->attrs); + abbrev = abbrev->next; + } + } + free (ent); +} + /* In DWARF version 2, the description of the debugging information is stored in a separate .debug_abbrev section. Before we read any dies from a section we read in all abbreviations and install them in a hash table. */ static struct abbrev_info** -read_abbrevs (bfd *abfd, bfd_uint64_t offset, struct dwarf2_debug *stash) +read_abbrevs (bfd *abfd, bfd_uint64_t offset, struct dwarf2_debug *stash, + struct dwarf2_debug_file *file) { struct abbrev_info **abbrevs; bfd_byte *abbrev_ptr; + bfd_byte *abbrev_end; struct abbrev_info *cur_abbrev; unsigned int abbrev_number, bytes_read, abbrev_name; unsigned int abbrev_form, hash_number; bfd_size_type amt; + void **slot; + struct abbrev_offset_entry ent = { offset, NULL }; - if (! stash->dwarf_abbrev_buffer) - { - asection *msec; + if (ent.offset != offset) + return NULL; - msec = bfd_get_section_by_name (abfd, ".debug_abbrev"); - if (! msec) - { - (*_bfd_error_handler) (_("Dwarf Error: Can't find .debug_abbrev section.")); - bfd_set_error (bfd_error_bad_value); - return 0; - } - - stash->dwarf_abbrev_size = msec->size; - stash->dwarf_abbrev_buffer - = bfd_simple_get_relocated_section_contents (abfd, msec, NULL, - stash->syms); - if (! stash->dwarf_abbrev_buffer) - return 0; - } + slot = htab_find_slot (file->abbrev_offsets, &ent, INSERT); + if (slot == NULL) + return NULL; + if (*slot != NULL) + return ((struct abbrev_offset_entry *) (*slot))->abbrevs; - if (offset >= stash->dwarf_abbrev_size) - { - (*_bfd_error_handler) (_("Dwarf Error: Abbrev offset (%lu) greater than or equal to .debug_abbrev size (%lu)."), - (unsigned long) offset, stash->dwarf_abbrev_size); - bfd_set_error (bfd_error_bad_value); - return 0; - } + if (! read_section (abfd, &stash->debug_sections[debug_abbrev], + file->syms, offset, + &file->dwarf_abbrev_buffer, + &file->dwarf_abbrev_size)) + return NULL; amt = sizeof (struct abbrev_info*) * ABBREV_HASH_SIZE; - abbrevs = bfd_zalloc (abfd, amt); + abbrevs = (struct abbrev_info **) bfd_zalloc (abfd, amt); + if (abbrevs == NULL) + return NULL; - abbrev_ptr = stash->dwarf_abbrev_buffer + offset; - abbrev_number = read_unsigned_leb128 (abfd, abbrev_ptr, &bytes_read); + abbrev_ptr = file->dwarf_abbrev_buffer + offset; + abbrev_end = file->dwarf_abbrev_buffer + file->dwarf_abbrev_size; + abbrev_number = _bfd_safe_read_leb128 (abfd, abbrev_ptr, &bytes_read, + FALSE, abbrev_end); abbrev_ptr += bytes_read; /* Loop until we reach an abbrev number of 0. */ while (abbrev_number) { amt = sizeof (struct abbrev_info); - cur_abbrev = bfd_zalloc (abfd, amt); + cur_abbrev = (struct abbrev_info *) bfd_zalloc (abfd, amt); + if (cur_abbrev == NULL) + goto fail; /* Read in abbrev header. */ cur_abbrev->number = abbrev_number; cur_abbrev->tag = (enum dwarf_tag) - read_unsigned_leb128 (abfd, abbrev_ptr, &bytes_read); + _bfd_safe_read_leb128 (abfd, abbrev_ptr, &bytes_read, + FALSE, abbrev_end); abbrev_ptr += bytes_read; - cur_abbrev->has_children = read_1_byte (abfd, abbrev_ptr); + cur_abbrev->has_children = read_1_byte (abfd, abbrev_ptr, abbrev_end); abbrev_ptr += 1; /* Now read in declarations. */ - abbrev_name = read_unsigned_leb128 (abfd, abbrev_ptr, &bytes_read); - abbrev_ptr += bytes_read; - abbrev_form = read_unsigned_leb128 (abfd, abbrev_ptr, &bytes_read); - abbrev_ptr += bytes_read; - - while (abbrev_name) + for (;;) { + /* Initialize it just to avoid a GCC false warning. */ + bfd_vma implicit_const = -1; + + abbrev_name = _bfd_safe_read_leb128 (abfd, abbrev_ptr, &bytes_read, + FALSE, abbrev_end); + abbrev_ptr += bytes_read; + abbrev_form = _bfd_safe_read_leb128 (abfd, abbrev_ptr, &bytes_read, + FALSE, abbrev_end); + abbrev_ptr += bytes_read; + if (abbrev_form == DW_FORM_implicit_const) + { + implicit_const = _bfd_safe_read_leb128 (abfd, abbrev_ptr, + &bytes_read, TRUE, + abbrev_end); + abbrev_ptr += bytes_read; + } + + if (abbrev_name == 0) + break; + if ((cur_abbrev->num_attrs % ATTR_ALLOC_CHUNK) == 0) { struct attr_abbrev *tmp; amt = cur_abbrev->num_attrs + ATTR_ALLOC_CHUNK; amt *= sizeof (struct attr_abbrev); - tmp = bfd_realloc (cur_abbrev->attrs, amt); + tmp = (struct attr_abbrev *) bfd_realloc (cur_abbrev->attrs, amt); if (tmp == NULL) - { - size_t i; - - for (i = 0; i < ABBREV_HASH_SIZE; i++) - { - struct abbrev_info *abbrev = abbrevs[i]; - - while (abbrev) - { - free (abbrev->attrs); - abbrev = abbrev->next; - } - } - return NULL; - } + goto fail; cur_abbrev->attrs = tmp; } cur_abbrev->attrs[cur_abbrev->num_attrs].name = (enum dwarf_attribute) abbrev_name; - cur_abbrev->attrs[cur_abbrev->num_attrs++].form + cur_abbrev->attrs[cur_abbrev->num_attrs].form = (enum dwarf_form) abbrev_form; - abbrev_name = read_unsigned_leb128 (abfd, abbrev_ptr, &bytes_read); - abbrev_ptr += bytes_read; - abbrev_form = read_unsigned_leb128 (abfd, abbrev_ptr, &bytes_read); - abbrev_ptr += bytes_read; + cur_abbrev->attrs[cur_abbrev->num_attrs].implicit_const + = implicit_const; + ++cur_abbrev->num_attrs; } hash_number = abbrev_number % ABBREV_HASH_SIZE; @@ -532,142 +1097,243 @@ read_abbrevs (bfd *abfd, bfd_uint64_t offset, struct dwarf2_debug *stash) already read (which means we are about to read the abbreviations for the next compile unit) or if the end of the abbreviation table is reached. */ - if ((unsigned int) (abbrev_ptr - stash->dwarf_abbrev_buffer) - >= stash->dwarf_abbrev_size) + if ((size_t) (abbrev_ptr - file->dwarf_abbrev_buffer) + >= file->dwarf_abbrev_size) break; - abbrev_number = read_unsigned_leb128 (abfd, abbrev_ptr, &bytes_read); + abbrev_number = _bfd_safe_read_leb128 (abfd, abbrev_ptr, + &bytes_read, FALSE, abbrev_end); abbrev_ptr += bytes_read; - if (lookup_abbrev (abbrev_number,abbrevs) != NULL) + if (lookup_abbrev (abbrev_number, abbrevs) != NULL) break; } + *slot = bfd_malloc (sizeof ent); + if (!*slot) + goto fail; + ent.abbrevs = abbrevs; + memcpy (*slot, &ent, sizeof ent); return abbrevs; + + fail: + if (abbrevs != NULL) + { + size_t i; + + for (i = 0; i < ABBREV_HASH_SIZE; i++) + { + struct abbrev_info *abbrev = abbrevs[i]; + + while (abbrev) + { + free (abbrev->attrs); + abbrev = abbrev->next; + } + } + free (abbrevs); + } + return NULL; } -/* Read an attribute value described by an attribute form. */ +/* Returns true if the form is one which has a string value. */ + +static inline bfd_boolean +is_str_attr (enum dwarf_form form) +{ + return (form == DW_FORM_string || form == DW_FORM_strp + || form == DW_FORM_line_strp || form == DW_FORM_GNU_strp_alt); +} + +/* Read and fill in the value of attribute ATTR as described by FORM. + Read data starting from INFO_PTR, but never at or beyond INFO_PTR_END. + Returns an updated INFO_PTR taking into account the amount of data read. */ static bfd_byte * -read_attribute_value (struct attribute *attr, - unsigned form, - struct comp_unit *unit, - bfd_byte *info_ptr) +read_attribute_value (struct attribute * attr, + unsigned form, + bfd_vma implicit_const, + struct comp_unit * unit, + bfd_byte * info_ptr, + bfd_byte * info_ptr_end) { bfd *abfd = unit->abfd; unsigned int bytes_read; struct dwarf_block *blk; bfd_size_type amt; + if (info_ptr >= info_ptr_end && form != DW_FORM_flag_present) + { + _bfd_error_handler (_("DWARF error: info pointer extends beyond end of attributes")); + bfd_set_error (bfd_error_bad_value); + return info_ptr; + } + attr->form = (enum dwarf_form) form; switch (form) { - case DW_FORM_addr: - /* FIXME: DWARF3 draft says DW_FORM_ref_addr is offset_size. */ case DW_FORM_ref_addr: - attr->u.val = read_address (unit, info_ptr); + /* DW_FORM_ref_addr is an address in DWARF2, and an offset in + DWARF3. */ + if (unit->version == 3 || unit->version == 4) + { + if (unit->offset_size == 4) + attr->u.val = read_4_bytes (unit->abfd, info_ptr, info_ptr_end); + else + attr->u.val = read_8_bytes (unit->abfd, info_ptr, info_ptr_end); + info_ptr += unit->offset_size; + break; + } + /* FALLTHROUGH */ + case DW_FORM_addr: + attr->u.val = read_address (unit, info_ptr, info_ptr_end); info_ptr += unit->addr_size; break; + case DW_FORM_GNU_ref_alt: + case DW_FORM_sec_offset: + if (unit->offset_size == 4) + attr->u.val = read_4_bytes (unit->abfd, info_ptr, info_ptr_end); + else + attr->u.val = read_8_bytes (unit->abfd, info_ptr, info_ptr_end); + info_ptr += unit->offset_size; + break; case DW_FORM_block2: amt = sizeof (struct dwarf_block); - blk = bfd_alloc (abfd, amt); - blk->size = read_2_bytes (abfd, info_ptr); + blk = (struct dwarf_block *) bfd_alloc (abfd, amt); + if (blk == NULL) + return NULL; + blk->size = read_2_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 2; - blk->data = read_n_bytes (abfd, info_ptr, blk->size); - info_ptr += blk->size; + info_ptr = read_n_bytes (info_ptr, info_ptr_end, blk); attr->u.blk = blk; break; case DW_FORM_block4: amt = sizeof (struct dwarf_block); - blk = bfd_alloc (abfd, amt); - blk->size = read_4_bytes (abfd, info_ptr); + blk = (struct dwarf_block *) bfd_alloc (abfd, amt); + if (blk == NULL) + return NULL; + blk->size = read_4_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 4; - blk->data = read_n_bytes (abfd, info_ptr, blk->size); - info_ptr += blk->size; + info_ptr = read_n_bytes (info_ptr, info_ptr_end, blk); attr->u.blk = blk; break; case DW_FORM_data2: - attr->u.val = read_2_bytes (abfd, info_ptr); + attr->u.val = read_2_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 2; break; case DW_FORM_data4: - attr->u.val = read_4_bytes (abfd, info_ptr); + attr->u.val = read_4_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 4; break; case DW_FORM_data8: - attr->u.val = read_8_bytes (abfd, info_ptr); + attr->u.val = read_8_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 8; break; case DW_FORM_string: - attr->u.str = read_string (abfd, info_ptr, &bytes_read); + attr->u.str = read_string (abfd, info_ptr, info_ptr_end, &bytes_read); info_ptr += bytes_read; break; case DW_FORM_strp: - attr->u.str = read_indirect_string (unit, info_ptr, &bytes_read); + attr->u.str = read_indirect_string (unit, info_ptr, info_ptr_end, &bytes_read); + info_ptr += bytes_read; + break; + case DW_FORM_line_strp: + attr->u.str = read_indirect_line_string (unit, info_ptr, info_ptr_end, &bytes_read); + info_ptr += bytes_read; + break; + case DW_FORM_GNU_strp_alt: + attr->u.str = read_alt_indirect_string (unit, info_ptr, info_ptr_end, &bytes_read); info_ptr += bytes_read; break; + case DW_FORM_exprloc: case DW_FORM_block: amt = sizeof (struct dwarf_block); - blk = bfd_alloc (abfd, amt); - blk->size = read_unsigned_leb128 (abfd, info_ptr, &bytes_read); + blk = (struct dwarf_block *) bfd_alloc (abfd, amt); + if (blk == NULL) + return NULL; + blk->size = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + FALSE, info_ptr_end); info_ptr += bytes_read; - blk->data = read_n_bytes (abfd, info_ptr, blk->size); - info_ptr += blk->size; + info_ptr = read_n_bytes (info_ptr, info_ptr_end, blk); attr->u.blk = blk; break; case DW_FORM_block1: amt = sizeof (struct dwarf_block); - blk = bfd_alloc (abfd, amt); - blk->size = read_1_byte (abfd, info_ptr); + blk = (struct dwarf_block *) bfd_alloc (abfd, amt); + if (blk == NULL) + return NULL; + blk->size = read_1_byte (abfd, info_ptr, info_ptr_end); info_ptr += 1; - blk->data = read_n_bytes (abfd, info_ptr, blk->size); - info_ptr += blk->size; + info_ptr = read_n_bytes (info_ptr, info_ptr_end, blk); attr->u.blk = blk; break; case DW_FORM_data1: - attr->u.val = read_1_byte (abfd, info_ptr); + attr->u.val = read_1_byte (abfd, info_ptr, info_ptr_end); info_ptr += 1; break; case DW_FORM_flag: - attr->u.val = read_1_byte (abfd, info_ptr); + attr->u.val = read_1_byte (abfd, info_ptr, info_ptr_end); info_ptr += 1; break; + case DW_FORM_flag_present: + attr->u.val = 1; + break; case DW_FORM_sdata: - attr->u.sval = read_signed_leb128 (abfd, info_ptr, &bytes_read); + attr->u.sval = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + TRUE, info_ptr_end); info_ptr += bytes_read; break; case DW_FORM_udata: - attr->u.val = read_unsigned_leb128 (abfd, info_ptr, &bytes_read); + attr->u.val = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + FALSE, info_ptr_end); info_ptr += bytes_read; break; case DW_FORM_ref1: - attr->u.val = read_1_byte (abfd, info_ptr); + attr->u.val = read_1_byte (abfd, info_ptr, info_ptr_end); info_ptr += 1; break; case DW_FORM_ref2: - attr->u.val = read_2_bytes (abfd, info_ptr); + attr->u.val = read_2_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 2; break; case DW_FORM_ref4: - attr->u.val = read_4_bytes (abfd, info_ptr); + attr->u.val = read_4_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 4; break; case DW_FORM_ref8: - attr->u.val = read_8_bytes (abfd, info_ptr); + attr->u.val = read_8_bytes (abfd, info_ptr, info_ptr_end); + info_ptr += 8; + break; + case DW_FORM_ref_sig8: + attr->u.val = read_8_bytes (abfd, info_ptr, info_ptr_end); info_ptr += 8; break; case DW_FORM_ref_udata: - attr->u.val = read_unsigned_leb128 (abfd, info_ptr, &bytes_read); + attr->u.val = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + FALSE, info_ptr_end); info_ptr += bytes_read; break; case DW_FORM_indirect: - form = read_unsigned_leb128 (abfd, info_ptr, &bytes_read); + form = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + FALSE, info_ptr_end); info_ptr += bytes_read; - info_ptr = read_attribute_value (attr, form, unit, info_ptr); + if (form == DW_FORM_implicit_const) + { + implicit_const = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + TRUE, info_ptr_end); + info_ptr += bytes_read; + } + info_ptr = read_attribute_value (attr, form, implicit_const, unit, + info_ptr, info_ptr_end); + break; + case DW_FORM_implicit_const: + attr->form = DW_FORM_sdata; + attr->u.sval = implicit_const; break; default: - (*_bfd_error_handler) (_("Dwarf Error: Invalid or unhandled FORM value: %u."), - form); + _bfd_error_handler (_("DWARF error: invalid or unhandled FORM value: %#x"), + form); bfd_set_error (bfd_error_bad_value); + return NULL; } return info_ptr; } @@ -675,16 +1341,45 @@ read_attribute_value (struct attribute *attr, /* Read an attribute described by an abbreviated attribute. */ static bfd_byte * -read_attribute (struct attribute *attr, - struct attr_abbrev *abbrev, - struct comp_unit *unit, - bfd_byte *info_ptr) +read_attribute (struct attribute * attr, + struct attr_abbrev * abbrev, + struct comp_unit * unit, + bfd_byte * info_ptr, + bfd_byte * info_ptr_end) { attr->name = abbrev->name; - info_ptr = read_attribute_value (attr, abbrev->form, unit, info_ptr); + info_ptr = read_attribute_value (attr, abbrev->form, abbrev->implicit_const, + unit, info_ptr, info_ptr_end); return info_ptr; } +/* Return whether DW_AT_name will return the same as DW_AT_linkage_name + for a function. */ + +static bfd_boolean +non_mangled (int lang) +{ + switch (lang) + { + default: + return FALSE; + + case DW_LANG_C89: + case DW_LANG_C: + case DW_LANG_Ada83: + case DW_LANG_Cobol74: + case DW_LANG_Cobol85: + case DW_LANG_Fortran77: + case DW_LANG_Pascal83: + case DW_LANG_C99: + case DW_LANG_Ada95: + case DW_LANG_PLI: + case DW_LANG_UPC: + case DW_LANG_C11: + return TRUE; + } +} + /* Source line information table routines. */ #define FILE_ALLOC_CHUNK 5 @@ -692,51 +1387,87 @@ read_attribute (struct attribute *attr, struct line_info { - struct line_info* prev_line; - bfd_vma address; - char *filename; - unsigned int line; - unsigned int column; - int end_sequence; /* End of (sequential) code sequence. */ + struct line_info * prev_line; + bfd_vma address; + char * filename; + unsigned int line; + unsigned int column; + unsigned int discriminator; + unsigned char op_index; + unsigned char end_sequence; /* End of (sequential) code sequence. */ }; struct fileinfo { - char *name; - unsigned int dir; - unsigned int time; - unsigned int size; + char * name; + unsigned int dir; + unsigned int time; + unsigned int size; +}; + +struct line_sequence +{ + bfd_vma low_pc; + struct line_sequence* prev_sequence; + struct line_info* last_line; /* Largest VMA. */ + struct line_info** line_info_lookup; + bfd_size_type num_lines; }; struct line_info_table { - bfd* abfd; - unsigned int num_files; - unsigned int num_dirs; - char *comp_dir; - char **dirs; - struct fileinfo* files; - struct line_info* last_line; /* largest VMA */ - struct line_info* lcl_head; /* local head; used in 'add_line_info' */ + bfd * abfd; + unsigned int num_files; + unsigned int num_dirs; + unsigned int num_sequences; + char * comp_dir; + char ** dirs; + struct fileinfo* files; + struct line_sequence* sequences; + struct line_info* lcl_head; /* Local head; used in 'add_line_info'. */ }; /* Remember some information about each function. If the function is inlined (DW_TAG_inlined_subroutine) it may have two additional attributes, DW_AT_call_file and DW_AT_call_line, which specify the - source code location where this function was inlined. */ + source code location where this function was inlined. */ struct funcinfo { - struct funcinfo *prev_func; /* Pointer to previous function in list of all functions */ - struct funcinfo *caller_func; /* Pointer to function one scope higher */ - char *caller_file; /* Source location file name where caller_func inlines this func */ - int caller_line; /* Source location line number where caller_func inlines this func */ - char *file; /* Source location file name */ - int line; /* Source location line number */ - int tag; - char *name; - struct arange arange; - asection *sec; /* Where the symbol is defined */ + /* Pointer to previous function in list of all functions. */ + struct funcinfo * prev_func; + /* Pointer to function one scope higher. */ + struct funcinfo * caller_func; + /* Source location file name where caller_func inlines this func. */ + char * caller_file; + /* Source location file name. */ + char * file; + /* Source location line number where caller_func inlines this func. */ + int caller_line; + /* Source location line number. */ + int line; + int tag; + bfd_boolean is_linkage; + const char * name; + struct arange arange; + /* Where the symbol is defined. */ + asection * sec; +}; + +struct lookup_funcinfo +{ + /* Function information corresponding to this lookup table entry. */ + struct funcinfo * funcinfo; + + /* The lowest address for this specific function. */ + bfd_vma low_addr; + + /* The highest address of this function before the lookup table is sorted. + The highest address of all prior functions after the lookup table is + sorted, which is used for binary search. */ + bfd_vma high_addr; + /* Index of this function, used to ensure qsort is stable. */ + unsigned int idx; }; struct varinfo @@ -763,7 +1494,7 @@ new_line_sorts_after (struct line_info *new_line, struct line_info *line) { return (new_line->address > line->address || (new_line->address == line->address - && new_line->end_sequence < line->end_sequence)); + && new_line->op_index > line->op_index)); } @@ -772,28 +1503,38 @@ new_line_sorts_after (struct line_info *new_line, struct line_info *line) highest to lowest VMA (with possible duplicates); that is, line_info->prev_line always accesses an equal or smaller VMA. */ -static void +static bfd_boolean add_line_info (struct line_info_table *table, bfd_vma address, + unsigned char op_index, char *filename, unsigned int line, unsigned int column, + unsigned int discriminator, int end_sequence) { bfd_size_type amt = sizeof (struct line_info); - struct line_info* info = bfd_alloc (table->abfd, amt); + struct line_sequence* seq = table->sequences; + struct line_info* info = (struct line_info *) bfd_alloc (table->abfd, amt); + + if (info == NULL) + return FALSE; /* Set member data of 'info'. */ + info->prev_line = NULL; info->address = address; + info->op_index = op_index; info->line = line; info->column = column; + info->discriminator = discriminator; info->end_sequence = end_sequence; if (filename && filename[0]) { - info->filename = bfd_alloc (table->abfd, strlen (filename) + 1); - if (info->filename) - strcpy (info->filename, filename); + info->filename = (char *) bfd_alloc (table->abfd, strlen (filename) + 1); + if (info->filename == NULL) + return FALSE; + strcpy (info->filename, filename); } else info->filename = NULL; @@ -808,17 +1549,43 @@ add_line_info (struct line_info_table *table, p...z a...j (where a < j < p < z) Note: table->lcl_head is used to head an *actual* or *possible* - sequence within the list (such as a...j) that is not directly + sub-sequence within the list (such as a...j) that is not directly headed by table->last_line Note: we may receive duplicate entries from 'decode_line_info'. */ - if (!table->last_line - || new_line_sorts_after (info, table->last_line)) + if (seq + && seq->last_line->address == address + && seq->last_line->op_index == op_index + && seq->last_line->end_sequence == end_sequence) + { + /* We only keep the last entry with the same address and end + sequence. See PR ld/4986. */ + if (table->lcl_head == seq->last_line) + table->lcl_head = info; + info->prev_line = seq->last_line->prev_line; + seq->last_line = info; + } + else if (!seq || seq->last_line->end_sequence) { - /* Normal case: add 'info' to the beginning of the list */ - info->prev_line = table->last_line; - table->last_line = info; + /* Start a new line sequence. */ + amt = sizeof (struct line_sequence); + seq = (struct line_sequence *) bfd_malloc (amt); + if (seq == NULL) + return FALSE; + seq->low_pc = address; + seq->prev_sequence = table->sequences; + seq->last_line = info; + table->lcl_head = info; + table->sequences = seq; + table->num_sequences++; + } + else if (info->end_sequence + || new_line_sorts_after (info, seq->last_line)) + { + /* Normal case: add 'info' to the beginning of the current sequence. */ + info->prev_line = seq->last_line; + seq->last_line = info; /* lcl_head: initialize to head a *possible* sequence at the end. */ if (!table->lcl_head) @@ -834,9 +1601,9 @@ add_line_info (struct line_info_table *table, } else { - /* Abnormal and hard: Neither 'last_line' nor 'lcl_head' are valid - heads for 'info'. Reset 'lcl_head'. */ - struct line_info* li2 = table->last_line; /* always non-NULL */ + /* Abnormal and hard: Neither 'last_line' nor 'lcl_head' + are valid heads for 'info'. Reset 'lcl_head'. */ + struct line_info* li2 = seq->last_line; /* Always non-NULL. */ struct line_info* li1 = li2->prev_line; while (li1) @@ -851,7 +1618,10 @@ add_line_info (struct line_info_table *table, table->lcl_head = li2; info->prev_line = table->lcl_head->prev_line; table->lcl_head->prev_line = info; + if (address < seq->low_pc) + seq->low_pc = address; } + return TRUE; } /* Extract a fully qualified filename from a line info table. @@ -863,51 +1633,83 @@ concat_filename (struct line_info_table *table, unsigned int file) { char *filename; - if (file - 1 >= table->num_files) + if (table == NULL || file - 1 >= table->num_files) { /* FILE == 0 means unknown. */ if (file) - (*_bfd_error_handler) - (_("Dwarf Error: mangled line number section (bad file number).")); + _bfd_error_handler + (_("DWARF error: mangled line number section (bad file number)")); return strdup (""); } filename = table->files[file - 1].name; + if (filename == NULL) + return strdup (""); - if (! IS_ABSOLUTE_PATH (filename)) + if (!IS_ABSOLUTE_PATH (filename)) { - char *dirname = (table->files[file - 1].dir - ? table->dirs[table->files[file - 1].dir - 1] - : table->comp_dir); - - /* Not all tools set DW_AT_comp_dir, so dirname may be unknown. - The best we can do is return the filename part. */ - if (dirname != NULL) + char *dir_name = NULL; + char *subdir_name = NULL; + char *name; + size_t len; + + if (table->files[file - 1].dir + /* PR 17512: file: 0317e960. */ + && table->files[file - 1].dir <= table->num_dirs + /* PR 17512: file: 7f3d2e4b. */ + && table->dirs != NULL) + subdir_name = table->dirs[table->files[file - 1].dir - 1]; + + if (!subdir_name || !IS_ABSOLUTE_PATH (subdir_name)) + dir_name = table->comp_dir; + + if (!dir_name) { - unsigned int len = strlen (dirname) + strlen (filename) + 2; - char * name; + dir_name = subdir_name; + subdir_name = NULL; + } + + if (!dir_name) + return strdup (filename); + + len = strlen (dir_name) + strlen (filename) + 2; - name = bfd_malloc (len); + if (subdir_name) + { + len += strlen (subdir_name) + 1; + name = (char *) bfd_malloc (len); + if (name) + sprintf (name, "%s/%s/%s", dir_name, subdir_name, filename); + } + else + { + name = (char *) bfd_malloc (len); if (name) - sprintf (name, "%s/%s", dirname, filename); - return name; + sprintf (name, "%s/%s", dir_name, filename); } + + return name; } return strdup (filename); } -static void -arange_add (bfd *abfd, struct arange *first_arange, bfd_vma low_pc, bfd_vma high_pc) +static bfd_boolean +arange_add (const struct comp_unit *unit, struct arange *first_arange, + bfd_vma low_pc, bfd_vma high_pc) { struct arange *arange; - /* If the first arange is empty, use it. */ + /* Ignore empty ranges. */ + if (low_pc == high_pc) + return TRUE; + + /* If the first arange is empty, use it. */ if (first_arange->high == 0) { first_arange->low = low_pc; first_arange->high = high_pc; - return; + return TRUE; } /* Next see if we can cheaply extend an existing range. */ @@ -917,207 +1719,601 @@ arange_add (bfd *abfd, struct arange *first_arange, bfd_vma low_pc, bfd_vma high if (low_pc == arange->high) { arange->high = high_pc; - return; + return TRUE; } if (high_pc == arange->low) { arange->low = low_pc; - return; + return TRUE; } arange = arange->next; } while (arange); /* Need to allocate a new arange and insert it into the arange list. - Order isn't significant, so just insert after the first arange. */ - arange = bfd_zalloc (abfd, sizeof (*arange)); + Order isn't significant, so just insert after the first arange. */ + arange = (struct arange *) bfd_alloc (unit->abfd, sizeof (*arange)); + if (arange == NULL) + return FALSE; arange->low = low_pc; arange->high = high_pc; arange->next = first_arange->next; first_arange->next = arange; + return TRUE; } -/* Decode the line number information for UNIT. */ +/* Compare function for line sequences. */ -static struct line_info_table* -decode_line_info (struct comp_unit *unit, struct dwarf2_debug *stash) +static int +compare_sequences (const void* a, const void* b) { - bfd *abfd = unit->abfd; - struct line_info_table* table; - bfd_byte *line_ptr; - bfd_byte *line_end; - struct line_head lh; - unsigned int i, bytes_read, offset_size; - char *cur_file, *cur_dir; - unsigned char op_code, extended_op, adj_opcode; - bfd_size_type amt; + const struct line_sequence* seq1 = a; + const struct line_sequence* seq2 = b; + + /* Sort by low_pc as the primary key. */ + if (seq1->low_pc < seq2->low_pc) + return -1; + if (seq1->low_pc > seq2->low_pc) + return 1; + + /* If low_pc values are equal, sort in reverse order of + high_pc, so that the largest region comes first. */ + if (seq1->last_line->address < seq2->last_line->address) + return 1; + if (seq1->last_line->address > seq2->last_line->address) + return -1; + + if (seq1->last_line->op_index < seq2->last_line->op_index) + return 1; + if (seq1->last_line->op_index > seq2->last_line->op_index) + return -1; + + /* num_lines is initially an index, to make the sort stable. */ + if (seq1->num_lines < seq2->num_lines) + return -1; + if (seq1->num_lines > seq2->num_lines) + return 1; + return 0; +} + +/* Construct the line information table for quick lookup. */ + +static bfd_boolean +build_line_info_table (struct line_info_table * table, + struct line_sequence * seq) +{ + bfd_size_type amt; + struct line_info** line_info_lookup; + struct line_info* each_line; + unsigned int num_lines; + unsigned int line_index; + + if (seq->line_info_lookup != NULL) + return TRUE; + + /* Count the number of line information entries. We could do this while + scanning the debug information, but some entries may be added via + lcl_head without having a sequence handy to increment the number of + lines. */ + num_lines = 0; + for (each_line = seq->last_line; each_line; each_line = each_line->prev_line) + num_lines++; + + seq->num_lines = num_lines; + if (num_lines == 0) + return TRUE; + + /* Allocate space for the line information lookup table. */ + amt = sizeof (struct line_info*) * num_lines; + line_info_lookup = (struct line_info**) bfd_alloc (table->abfd, amt); + seq->line_info_lookup = line_info_lookup; + if (line_info_lookup == NULL) + return FALSE; + + /* Create the line information lookup table. */ + line_index = num_lines; + for (each_line = seq->last_line; each_line; each_line = each_line->prev_line) + line_info_lookup[--line_index] = each_line; + + BFD_ASSERT (line_index == 0); + return TRUE; +} + +/* Sort the line sequences for quick lookup. */ + +static bfd_boolean +sort_line_sequences (struct line_info_table* table) +{ + bfd_size_type amt; + struct line_sequence* sequences; + struct line_sequence* seq; + unsigned int n = 0; + unsigned int num_sequences = table->num_sequences; + bfd_vma last_high_pc; + + if (num_sequences == 0) + return TRUE; + + /* Allocate space for an array of sequences. */ + amt = sizeof (struct line_sequence) * num_sequences; + sequences = (struct line_sequence *) bfd_alloc (table->abfd, amt); + if (sequences == NULL) + return FALSE; - if (! stash->dwarf_line_buffer) + /* Copy the linked list into the array, freeing the original nodes. */ + seq = table->sequences; + for (n = 0; n < num_sequences; n++) { - asection *msec; + struct line_sequence* last_seq = seq; + + BFD_ASSERT (seq); + sequences[n].low_pc = seq->low_pc; + sequences[n].prev_sequence = NULL; + sequences[n].last_line = seq->last_line; + sequences[n].line_info_lookup = NULL; + sequences[n].num_lines = n; + seq = seq->prev_sequence; + free (last_seq); + } + BFD_ASSERT (seq == NULL); - msec = bfd_get_section_by_name (abfd, ".debug_line"); - if (! msec) + qsort (sequences, n, sizeof (struct line_sequence), compare_sequences); + + /* Make the list binary-searchable by trimming overlapping entries + and removing nested entries. */ + num_sequences = 1; + last_high_pc = sequences[0].last_line->address; + for (n = 1; n < table->num_sequences; n++) + { + if (sequences[n].low_pc < last_high_pc) { - (*_bfd_error_handler) (_("Dwarf Error: Can't find .debug_line section.")); - bfd_set_error (bfd_error_bad_value); - return 0; - } + if (sequences[n].last_line->address <= last_high_pc) + /* Skip nested entries. */ + continue; - stash->dwarf_line_size = msec->size; - stash->dwarf_line_buffer - = bfd_simple_get_relocated_section_contents (abfd, msec, NULL, - stash->syms); - if (! stash->dwarf_line_buffer) - return 0; + /* Trim overlapping entries. */ + sequences[n].low_pc = last_high_pc; + } + last_high_pc = sequences[n].last_line->address; + if (n > num_sequences) + { + /* Close up the gap. */ + sequences[num_sequences].low_pc = sequences[n].low_pc; + sequences[num_sequences].last_line = sequences[n].last_line; + } + num_sequences++; } - /* It is possible to get a bad value for the line_offset. Validate - it here so that we won't get a segfault below. */ - if (unit->line_offset >= stash->dwarf_line_size) + table->sequences = sequences; + table->num_sequences = num_sequences; + return TRUE; +} + +/* Add directory to TABLE. CUR_DIR memory ownership is taken by TABLE. */ + +static bfd_boolean +line_info_add_include_dir (struct line_info_table *table, char *cur_dir) +{ + if ((table->num_dirs % DIR_ALLOC_CHUNK) == 0) { - (*_bfd_error_handler) (_("Dwarf Error: Line offset (%lu) greater than or equal to .debug_line size (%lu)."), - unit->line_offset, stash->dwarf_line_size); - bfd_set_error (bfd_error_bad_value); - return 0; - } + char **tmp; + bfd_size_type amt; - amt = sizeof (struct line_info_table); - table = bfd_alloc (abfd, amt); - table->abfd = abfd; - table->comp_dir = unit->comp_dir; + amt = table->num_dirs + DIR_ALLOC_CHUNK; + amt *= sizeof (char *); - table->num_files = 0; - table->files = NULL; + tmp = (char **) bfd_realloc (table->dirs, amt); + if (tmp == NULL) + return FALSE; + table->dirs = tmp; + } - table->num_dirs = 0; - table->dirs = NULL; + table->dirs[table->num_dirs++] = cur_dir; + return TRUE; +} - table->files = NULL; - table->last_line = NULL; - table->lcl_head = NULL; +static bfd_boolean +line_info_add_include_dir_stub (struct line_info_table *table, char *cur_dir, + unsigned int dir ATTRIBUTE_UNUSED, + unsigned int xtime ATTRIBUTE_UNUSED, + unsigned int size ATTRIBUTE_UNUSED) +{ + return line_info_add_include_dir (table, cur_dir); +} - line_ptr = stash->dwarf_line_buffer + unit->line_offset; +/* Add file to TABLE. CUR_FILE memory ownership is taken by TABLE. */ - /* Read in the prologue. */ - lh.total_length = read_4_bytes (abfd, line_ptr); - line_ptr += 4; - offset_size = 4; - if (lh.total_length == 0xffffffff) +static bfd_boolean +line_info_add_file_name (struct line_info_table *table, char *cur_file, + unsigned int dir, unsigned int xtime, + unsigned int size) +{ + if ((table->num_files % FILE_ALLOC_CHUNK) == 0) { - lh.total_length = read_8_bytes (abfd, line_ptr); - line_ptr += 8; - offset_size = 8; + struct fileinfo *tmp; + bfd_size_type amt; + + amt = table->num_files + FILE_ALLOC_CHUNK; + amt *= sizeof (struct fileinfo); + + tmp = (struct fileinfo *) bfd_realloc (table->files, amt); + if (tmp == NULL) + return FALSE; + table->files = tmp; } - else if (lh.total_length == 0 && unit->addr_size == 8) + + table->files[table->num_files].name = cur_file; + table->files[table->num_files].dir = dir; + table->files[table->num_files].time = xtime; + table->files[table->num_files].size = size; + table->num_files++; + return TRUE; +} + +/* Read directory or file name entry format, starting with byte of + format count entries, ULEB128 pairs of entry formats, ULEB128 of + entries count and the entries themselves in the described entry + format. */ + +static bfd_boolean +read_formatted_entries (struct comp_unit *unit, bfd_byte **bufp, + bfd_byte *buf_end, struct line_info_table *table, + bfd_boolean (*callback) (struct line_info_table *table, + char *cur_file, + unsigned int dir, + unsigned int time, + unsigned int size)) +{ + bfd *abfd = unit->abfd; + bfd_byte format_count, formati; + bfd_vma data_count, datai; + bfd_byte *buf = *bufp; + bfd_byte *format_header_data; + unsigned int bytes_read; + + format_count = read_1_byte (abfd, buf, buf_end); + buf += 1; + format_header_data = buf; + for (formati = 0; formati < format_count; formati++) { - /* Handle (non-standard) 64-bit DWARF2 formats. */ - lh.total_length = read_4_bytes (abfd, line_ptr); - line_ptr += 4; - offset_size = 8; + _bfd_safe_read_leb128 (abfd, buf, &bytes_read, FALSE, buf_end); + buf += bytes_read; + _bfd_safe_read_leb128 (abfd, buf, &bytes_read, FALSE, buf_end); + buf += bytes_read; } - line_end = line_ptr + lh.total_length; - lh.version = read_2_bytes (abfd, line_ptr); - line_ptr += 2; - if (offset_size == 4) - lh.prologue_length = read_4_bytes (abfd, line_ptr); - else - lh.prologue_length = read_8_bytes (abfd, line_ptr); - line_ptr += offset_size; - lh.minimum_instruction_length = read_1_byte (abfd, line_ptr); - line_ptr += 1; - lh.default_is_stmt = read_1_byte (abfd, line_ptr); - line_ptr += 1; - lh.line_base = read_1_signed_byte (abfd, line_ptr); - line_ptr += 1; - lh.line_range = read_1_byte (abfd, line_ptr); - line_ptr += 1; - lh.opcode_base = read_1_byte (abfd, line_ptr); - line_ptr += 1; - amt = lh.opcode_base * sizeof (unsigned char); - lh.standard_opcode_lengths = bfd_alloc (abfd, amt); - lh.standard_opcode_lengths[0] = 1; + data_count = _bfd_safe_read_leb128 (abfd, buf, &bytes_read, FALSE, buf_end); + buf += bytes_read; + if (format_count == 0 && data_count != 0) + { + _bfd_error_handler (_("DWARF error: zero format count")); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } - for (i = 1; i < lh.opcode_base; ++i) + /* PR 22210. Paranoia check. Don't bother running the loop + if we know that we are going to run out of buffer. */ + if (data_count > (bfd_vma) (buf_end - buf)) { - lh.standard_opcode_lengths[i] = read_1_byte (abfd, line_ptr); - line_ptr += 1; + _bfd_error_handler + (_("DWARF error: data count (%" PRIx64 ") larger than buffer size"), + (uint64_t) data_count); + bfd_set_error (bfd_error_bad_value); + return FALSE; } - /* Read directory table. */ - while ((cur_dir = read_string (abfd, line_ptr, &bytes_read)) != NULL) + for (datai = 0; datai < data_count; datai++) { - line_ptr += bytes_read; + bfd_byte *format = format_header_data; + struct fileinfo fe; - if ((table->num_dirs % DIR_ALLOC_CHUNK) == 0) + memset (&fe, 0, sizeof fe); + for (formati = 0; formati < format_count; formati++) { - char **tmp; + bfd_vma content_type, form; + char *string_trash; + char **stringp = &string_trash; + unsigned int uint_trash, *uintp = &uint_trash; + struct attribute attr; + + content_type = _bfd_safe_read_leb128 (abfd, format, &bytes_read, + FALSE, buf_end); + format += bytes_read; + switch (content_type) + { + case DW_LNCT_path: + stringp = &fe.name; + break; + case DW_LNCT_directory_index: + uintp = &fe.dir; + break; + case DW_LNCT_timestamp: + uintp = &fe.time; + break; + case DW_LNCT_size: + uintp = &fe.size; + break; + case DW_LNCT_MD5: + break; + default: + _bfd_error_handler + (_("DWARF error: unknown format content type %" PRIu64), + (uint64_t) content_type); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } - amt = table->num_dirs + DIR_ALLOC_CHUNK; - amt *= sizeof (char *); + form = _bfd_safe_read_leb128 (abfd, format, &bytes_read, FALSE, + buf_end); + format += bytes_read; - tmp = bfd_realloc (table->dirs, amt); - if (tmp == NULL) + buf = read_attribute_value (&attr, form, 0, unit, buf, buf_end); + if (buf == NULL) + return FALSE; + switch (form) { - free (table->dirs); - return NULL; + case DW_FORM_string: + case DW_FORM_line_strp: + *stringp = attr.u.str; + break; + + case DW_FORM_data1: + case DW_FORM_data2: + case DW_FORM_data4: + case DW_FORM_data8: + case DW_FORM_udata: + *uintp = attr.u.val; + break; } - table->dirs = tmp; } - table->dirs[table->num_dirs++] = cur_dir; + if (!callback (table, fe.name, fe.dir, fe.time, fe.size)) + return FALSE; } - line_ptr += bytes_read; + *bufp = buf; + return TRUE; +} + +/* Decode the line number information for UNIT. */ + +static struct line_info_table* +decode_line_info (struct comp_unit *unit) +{ + bfd *abfd = unit->abfd; + struct dwarf2_debug *stash = unit->stash; + struct dwarf2_debug_file *file = unit->file; + struct line_info_table* table; + bfd_byte *line_ptr; + bfd_byte *line_end; + struct line_head lh; + unsigned int i, bytes_read, offset_size; + char *cur_file, *cur_dir; + unsigned char op_code, extended_op, adj_opcode; + unsigned int exop_len; + bfd_size_type amt; + + if (unit->line_offset == 0 && file->line_table) + return file->line_table; + + if (! read_section (abfd, &stash->debug_sections[debug_line], + file->syms, unit->line_offset, + &file->dwarf_line_buffer, &file->dwarf_line_size)) + return NULL; - /* Read file name table. */ - while ((cur_file = read_string (abfd, line_ptr, &bytes_read)) != NULL) + if (file->dwarf_line_size < 16) { - line_ptr += bytes_read; + _bfd_error_handler + (_("DWARF error: line info section is too small (%" PRId64 ")"), + (int64_t) file->dwarf_line_size); + bfd_set_error (bfd_error_bad_value); + return NULL; + } + line_ptr = file->dwarf_line_buffer + unit->line_offset; + line_end = file->dwarf_line_buffer + file->dwarf_line_size; + + /* Read in the prologue. */ + lh.total_length = read_4_bytes (abfd, line_ptr, line_end); + line_ptr += 4; + offset_size = 4; + if (lh.total_length == 0xffffffff) + { + lh.total_length = read_8_bytes (abfd, line_ptr, line_end); + line_ptr += 8; + offset_size = 8; + } + else if (lh.total_length == 0 && unit->addr_size == 8) + { + /* Handle (non-standard) 64-bit DWARF2 formats. */ + lh.total_length = read_4_bytes (abfd, line_ptr, line_end); + line_ptr += 4; + offset_size = 8; + } + + if (lh.total_length > (size_t) (line_end - line_ptr)) + { + _bfd_error_handler + /* xgettext: c-format */ + (_("DWARF error: line info data is bigger (%#" PRIx64 ")" + " than the space remaining in the section (%#lx)"), + (uint64_t) lh.total_length, (unsigned long) (line_end - line_ptr)); + bfd_set_error (bfd_error_bad_value); + return NULL; + } + + line_end = line_ptr + lh.total_length; + + lh.version = read_2_bytes (abfd, line_ptr, line_end); + if (lh.version < 2 || lh.version > 5) + { + _bfd_error_handler + (_("DWARF error: unhandled .debug_line version %d"), lh.version); + bfd_set_error (bfd_error_bad_value); + return NULL; + } + line_ptr += 2; + + if (line_ptr + offset_size + (lh.version >= 5 ? 8 : (lh.version >= 4 ? 6 : 5)) + >= line_end) + { + _bfd_error_handler + (_("DWARF error: ran out of room reading prologue")); + bfd_set_error (bfd_error_bad_value); + return NULL; + } + + if (lh.version >= 5) + { + unsigned int segment_selector_size; + + /* Skip address size. */ + read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; - if ((table->num_files % FILE_ALLOC_CHUNK) == 0) + segment_selector_size = read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; + if (segment_selector_size != 0) { - struct fileinfo *tmp; + _bfd_error_handler + (_("DWARF error: line info unsupported segment selector size %u"), + segment_selector_size); + bfd_set_error (bfd_error_bad_value); + return NULL; + } + } - amt = table->num_files + FILE_ALLOC_CHUNK; - amt *= sizeof (struct fileinfo); + if (offset_size == 4) + lh.prologue_length = read_4_bytes (abfd, line_ptr, line_end); + else + lh.prologue_length = read_8_bytes (abfd, line_ptr, line_end); + line_ptr += offset_size; - tmp = bfd_realloc (table->files, amt); - if (tmp == NULL) - { - free (table->files); - free (table->dirs); - return NULL; - } - table->files = tmp; + lh.minimum_instruction_length = read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; + + if (lh.version >= 4) + { + lh.maximum_ops_per_insn = read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; + } + else + lh.maximum_ops_per_insn = 1; + + if (lh.maximum_ops_per_insn == 0) + { + _bfd_error_handler + (_("DWARF error: invalid maximum operations per instruction")); + bfd_set_error (bfd_error_bad_value); + return NULL; + } + + lh.default_is_stmt = read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; + + lh.line_base = read_1_signed_byte (abfd, line_ptr, line_end); + line_ptr += 1; + + lh.line_range = read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; + + lh.opcode_base = read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; + + if (line_ptr + (lh.opcode_base - 1) >= line_end) + { + _bfd_error_handler (_("DWARF error: ran out of room reading opcodes")); + bfd_set_error (bfd_error_bad_value); + return NULL; + } + + amt = lh.opcode_base * sizeof (unsigned char); + lh.standard_opcode_lengths = (unsigned char *) bfd_alloc (abfd, amt); + + lh.standard_opcode_lengths[0] = 1; + + for (i = 1; i < lh.opcode_base; ++i) + { + lh.standard_opcode_lengths[i] = read_1_byte (abfd, line_ptr, line_end); + line_ptr += 1; + } + + amt = sizeof (struct line_info_table); + table = (struct line_info_table *) bfd_alloc (abfd, amt); + if (table == NULL) + return NULL; + table->abfd = abfd; + table->comp_dir = unit->comp_dir; + + table->num_files = 0; + table->files = NULL; + + table->num_dirs = 0; + table->dirs = NULL; + + table->num_sequences = 0; + table->sequences = NULL; + + table->lcl_head = NULL; + + if (lh.version >= 5) + { + /* Read directory table. */ + if (!read_formatted_entries (unit, &line_ptr, line_end, table, + line_info_add_include_dir_stub)) + goto fail; + + /* Read file name table. */ + if (!read_formatted_entries (unit, &line_ptr, line_end, table, + line_info_add_file_name)) + goto fail; + } + else + { + /* Read directory table. */ + while ((cur_dir = read_string (abfd, line_ptr, line_end, &bytes_read)) != NULL) + { + line_ptr += bytes_read; + + if (!line_info_add_include_dir (table, cur_dir)) + goto fail; } - table->files[table->num_files].name = cur_file; - table->files[table->num_files].dir = - read_unsigned_leb128 (abfd, line_ptr, &bytes_read); - line_ptr += bytes_read; - table->files[table->num_files].time = - read_unsigned_leb128 (abfd, line_ptr, &bytes_read); line_ptr += bytes_read; - table->files[table->num_files].size = - read_unsigned_leb128 (abfd, line_ptr, &bytes_read); + + /* Read file name table. */ + while ((cur_file = read_string (abfd, line_ptr, line_end, &bytes_read)) != NULL) + { + unsigned int dir, xtime, size; + + line_ptr += bytes_read; + + dir = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, FALSE, line_end); + line_ptr += bytes_read; + xtime = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, FALSE, line_end); + line_ptr += bytes_read; + size = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, FALSE, line_end); + line_ptr += bytes_read; + + if (!line_info_add_file_name (table, cur_file, dir, xtime, size)) + goto fail; + } + line_ptr += bytes_read; - table->num_files++; } - line_ptr += bytes_read; - /* Read the statement sequences until there's nothing left. */ while (line_ptr < line_end) { /* State machine registers. */ bfd_vma address = 0; + unsigned char op_index = 0; char * filename = table->num_files ? concat_filename (table, 1) : NULL; unsigned int line = 1; unsigned int column = 0; + unsigned int discriminator = 0; int is_stmt = lh.default_is_stmt; int end_sequence = 0; + unsigned int dir, xtime, size; /* eraxxon@alumni.rice.edu: Against the DWARF2 specs, some compilers generate address sequences that are wildly out of order using DW_LNE_set_address (e.g. Intel C++ 6.0 compiler @@ -1127,20 +2323,34 @@ decode_line_info (struct comp_unit *unit, struct dwarf2_debug *stash) bfd_vma high_pc = 0; /* Decode the table. */ - while (! end_sequence) + while (!end_sequence && line_ptr < line_end) { - op_code = read_1_byte (abfd, line_ptr); + op_code = read_1_byte (abfd, line_ptr, line_end); line_ptr += 1; if (op_code >= lh.opcode_base) { /* Special operand. */ adj_opcode = op_code - lh.opcode_base; - address += (adj_opcode / lh.line_range) - * lh.minimum_instruction_length; + if (lh.line_range == 0) + goto line_fail; + if (lh.maximum_ops_per_insn == 1) + address += (adj_opcode / lh.line_range + * lh.minimum_instruction_length); + else + { + address += ((op_index + adj_opcode / lh.line_range) + / lh.maximum_ops_per_insn + * lh.minimum_instruction_length); + op_index = ((op_index + adj_opcode / lh.line_range) + % lh.maximum_ops_per_insn); + } line += lh.line_base + (adj_opcode % lh.line_range); /* Append row to matrix using current values. */ - add_line_info (table, address, filename, line, column, 0); + if (!add_line_info (table, address, op_index, filename, + line, column, discriminator, 0)) + goto line_fail; + discriminator = 0; if (address < low_pc) low_pc = address; if (address > high_pc) @@ -1149,98 +2359,116 @@ decode_line_info (struct comp_unit *unit, struct dwarf2_debug *stash) else switch (op_code) { case DW_LNS_extended_op: - /* Ignore length. */ - line_ptr += 1; - extended_op = read_1_byte (abfd, line_ptr); + exop_len = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); + line_ptr += bytes_read; + extended_op = read_1_byte (abfd, line_ptr, line_end); line_ptr += 1; switch (extended_op) { case DW_LNE_end_sequence: end_sequence = 1; - add_line_info (table, address, filename, line, column, - end_sequence); + if (!add_line_info (table, address, op_index, filename, line, + column, discriminator, end_sequence)) + goto line_fail; + discriminator = 0; if (address < low_pc) low_pc = address; if (address > high_pc) high_pc = address; - arange_add (unit->abfd, &unit->arange, low_pc, high_pc); + if (!arange_add (unit, &unit->arange, low_pc, high_pc)) + goto line_fail; break; case DW_LNE_set_address: - address = read_address (unit, line_ptr); + address = read_address (unit, line_ptr, line_end); + op_index = 0; line_ptr += unit->addr_size; break; case DW_LNE_define_file: - cur_file = read_string (abfd, line_ptr, &bytes_read); + cur_file = read_string (abfd, line_ptr, line_end, &bytes_read); line_ptr += bytes_read; - if ((table->num_files % FILE_ALLOC_CHUNK) == 0) - { - struct fileinfo *tmp; - - amt = table->num_files + FILE_ALLOC_CHUNK; - amt *= sizeof (struct fileinfo); - tmp = bfd_realloc (table->files, amt); - if (tmp == NULL) - { - free (table->files); - free (table->dirs); - free (filename); - return NULL; - } - table->files = tmp; - } - table->files[table->num_files].name = cur_file; - table->files[table->num_files].dir = - read_unsigned_leb128 (abfd, line_ptr, &bytes_read); + dir = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); line_ptr += bytes_read; - table->files[table->num_files].time = - read_unsigned_leb128 (abfd, line_ptr, &bytes_read); + xtime = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); line_ptr += bytes_read; - table->files[table->num_files].size = - read_unsigned_leb128 (abfd, line_ptr, &bytes_read); + size = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); line_ptr += bytes_read; - table->num_files++; + if (!line_info_add_file_name (table, cur_file, dir, + xtime, size)) + goto line_fail; + break; + case DW_LNE_set_discriminator: + discriminator = + _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); + line_ptr += bytes_read; + break; + case DW_LNE_HP_source_file_correlation: + line_ptr += exop_len - 1; break; default: - (*_bfd_error_handler) (_("Dwarf Error: mangled line number section.")); + _bfd_error_handler + (_("DWARF error: mangled line number section")); bfd_set_error (bfd_error_bad_value); - free (filename); - free (table->files); - free (table->dirs); - return NULL; + line_fail: + if (filename != NULL) + free (filename); + goto fail; } break; case DW_LNS_copy: - add_line_info (table, address, filename, line, column, 0); + if (!add_line_info (table, address, op_index, + filename, line, column, discriminator, 0)) + goto line_fail; + discriminator = 0; if (address < low_pc) low_pc = address; if (address > high_pc) high_pc = address; break; case DW_LNS_advance_pc: - address += lh.minimum_instruction_length - * read_unsigned_leb128 (abfd, line_ptr, &bytes_read); + if (lh.maximum_ops_per_insn == 1) + address += (lh.minimum_instruction_length + * _bfd_safe_read_leb128 (abfd, line_ptr, + &bytes_read, + FALSE, line_end)); + else + { + bfd_vma adjust = _bfd_safe_read_leb128 (abfd, line_ptr, + &bytes_read, + FALSE, line_end); + address = ((op_index + adjust) / lh.maximum_ops_per_insn + * lh.minimum_instruction_length); + op_index = (op_index + adjust) % lh.maximum_ops_per_insn; + } line_ptr += bytes_read; break; case DW_LNS_advance_line: - line += read_signed_leb128 (abfd, line_ptr, &bytes_read); + line += _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + TRUE, line_end); line_ptr += bytes_read; break; case DW_LNS_set_file: { - unsigned int file; + unsigned int filenum; /* The file and directory tables are 0 based, the references are 1 based. */ - file = read_unsigned_leb128 (abfd, line_ptr, &bytes_read); + filenum = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); line_ptr += bytes_read; if (filename) free (filename); - filename = concat_filename (table, file); + filename = concat_filename (table, filenum); break; } case DW_LNS_set_column: - column = read_unsigned_leb128 (abfd, line_ptr, &bytes_read); + column = _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); line_ptr += bytes_read; break; case DW_LNS_negate_stmt: @@ -1249,24 +2477,34 @@ decode_line_info (struct comp_unit *unit, struct dwarf2_debug *stash) case DW_LNS_set_basic_block: break; case DW_LNS_const_add_pc: - address += lh.minimum_instruction_length - * ((255 - lh.opcode_base) / lh.line_range); + if (lh.line_range == 0) + goto line_fail; + if (lh.maximum_ops_per_insn == 1) + address += (lh.minimum_instruction_length + * ((255 - lh.opcode_base) / lh.line_range)); + else + { + bfd_vma adjust = ((255 - lh.opcode_base) / lh.line_range); + address += (lh.minimum_instruction_length + * ((op_index + adjust) + / lh.maximum_ops_per_insn)); + op_index = (op_index + adjust) % lh.maximum_ops_per_insn; + } break; case DW_LNS_fixed_advance_pc: - address += read_2_bytes (abfd, line_ptr); + address += read_2_bytes (abfd, line_ptr, line_end); + op_index = 0; line_ptr += 2; break; default: - { - int i; - - /* Unknown standard opcode, ignore it. */ - for (i = 0; i < lh.standard_opcode_lengths[op_code]; i++) - { - (void) read_unsigned_leb128 (abfd, line_ptr, &bytes_read); - line_ptr += bytes_read; - } - } + /* Unknown standard opcode, ignore it. */ + for (i = 0; i < lh.standard_opcode_lengths[op_code]; i++) + { + (void) _bfd_safe_read_leb128 (abfd, line_ptr, &bytes_read, + FALSE, line_end); + line_ptr += bytes_read; + } + break; } } @@ -1274,181 +2512,283 @@ decode_line_info (struct comp_unit *unit, struct dwarf2_debug *stash) free (filename); } - return table; + if (unit->line_offset == 0) + file->line_table = table; + if (sort_line_sequences (table)) + return table; + + fail: + while (table->sequences != NULL) + { + struct line_sequence* seq = table->sequences; + table->sequences = table->sequences->prev_sequence; + free (seq); + } + if (table->files != NULL) + free (table->files); + if (table->dirs != NULL) + free (table->dirs); + return NULL; } -/* If ADDR is within TABLE set the output parameters and return TRUE, - otherwise return FALSE. The output parameters, FILENAME_PTR and - LINENUMBER_PTR, are pointers to the objects to be filled in. */ +/* If ADDR is within TABLE set the output parameters and return the + range of addresses covered by the entry used to fill them out. + Otherwise set * FILENAME_PTR to NULL and return 0. + The parameters FILENAME_PTR, LINENUMBER_PTR and DISCRIMINATOR_PTR + are pointers to the objects to be filled in. */ -static bfd_boolean +static bfd_vma lookup_address_in_line_info_table (struct line_info_table *table, bfd_vma addr, - struct funcinfo *function, const char **filename_ptr, - unsigned int *linenumber_ptr) + unsigned int *linenumber_ptr, + unsigned int *discriminator_ptr) { - /* Note: table->last_line should be a descendingly sorted list. */ - struct line_info* next_line = table->last_line; - struct line_info* each_line = NULL; - *filename_ptr = NULL; - - if (!next_line) - return FALSE; - - each_line = next_line->prev_line; - - /* Check for large addresses */ - if (addr > next_line->address) - each_line = NULL; /* ensure we skip over the normal case */ - - /* Normal case: search the list; save */ - while (each_line && next_line) + struct line_sequence *seq = NULL; + struct line_info *info; + int low, high, mid; + + /* Binary search the array of sequences. */ + low = 0; + high = table->num_sequences; + while (low < high) { - /* If we have an address match, save this info. This allows us - to return as good as results as possible for strange debugging - info. */ - bfd_boolean addr_match = FALSE; - if (each_line->address <= addr && addr < next_line->address) - { - addr_match = TRUE; - - /* If this line appears to span functions, and addr is in the - later function, return the first line of that function instead - of the last line of the earlier one. This check is for GCC - 2.95, which emits the first line number for a function late. */ + mid = (low + high) / 2; + seq = &table->sequences[mid]; + if (addr < seq->low_pc) + high = mid; + else if (addr >= seq->last_line->address) + low = mid + 1; + else + break; + } - if (function != NULL) - { - bfd_vma lowest_pc; - struct arange *arange; - - /* Find the lowest address in the function's range list */ - lowest_pc = function->arange.low; - for (arange = &function->arange; - arange; - arange = arange->next) - { - if (function->arange.low < lowest_pc) - lowest_pc = function->arange.low; - } - /* Check for spanning function and set outgoing line info */ - if (addr >= lowest_pc - && each_line->address < lowest_pc - && next_line->address > lowest_pc) - { - *filename_ptr = next_line->filename; - *linenumber_ptr = next_line->line; - } - else - { - *filename_ptr = each_line->filename; - *linenumber_ptr = each_line->line; - } - } - else - { - *filename_ptr = each_line->filename; - *linenumber_ptr = each_line->line; - } - } + /* Check for a valid sequence. */ + if (!seq || addr < seq->low_pc || addr >= seq->last_line->address) + goto fail; - if (addr_match && !each_line->end_sequence) - return TRUE; /* we have definitely found what we want */ + if (!build_line_info_table (table, seq)) + goto fail; - next_line = each_line; - each_line = each_line->prev_line; + /* Binary search the array of line information. */ + low = 0; + high = seq->num_lines; + info = NULL; + while (low < high) + { + mid = (low + high) / 2; + info = seq->line_info_lookup[mid]; + if (addr < info->address) + high = mid; + else if (addr >= seq->line_info_lookup[mid + 1]->address) + low = mid + 1; + else + break; } - /* At this point each_line is NULL but next_line is not. If we found - a candidate end-of-sequence point in the loop above, we can return - that (compatibility with a bug in the Intel compiler); otherwise, - assuming that we found the containing function for this address in - this compilation unit, return the first line we have a number for - (compatibility with GCC 2.95). */ - if (*filename_ptr == NULL && function != NULL) + /* Check for a valid line information entry. */ + if (info + && addr >= info->address + && addr < seq->line_info_lookup[mid + 1]->address + && !(info->end_sequence || info == seq->last_line)) { - *filename_ptr = next_line->filename; - *linenumber_ptr = next_line->line; - return TRUE; + *filename_ptr = info->filename; + *linenumber_ptr = info->line; + if (discriminator_ptr) + *discriminator_ptr = info->discriminator; + return seq->last_line->address - seq->low_pc; } - return FALSE; +fail: + *filename_ptr = NULL; + return 0; } -/* Read in the .debug_ranges section for future reference */ +/* Read in the .debug_ranges section for future reference. */ static bfd_boolean -read_debug_ranges (struct comp_unit *unit) +read_debug_ranges (struct comp_unit * unit) { struct dwarf2_debug *stash = unit->stash; - if (! stash->dwarf_ranges_buffer) + struct dwarf2_debug_file *file = unit->file; + + return read_section (unit->abfd, &stash->debug_sections[debug_ranges], + file->syms, 0, + &file->dwarf_ranges_buffer, &file->dwarf_ranges_size); +} + +/* Function table functions. */ + +static int +compare_lookup_funcinfos (const void * a, const void * b) +{ + const struct lookup_funcinfo * lookup1 = a; + const struct lookup_funcinfo * lookup2 = b; + + if (lookup1->low_addr < lookup2->low_addr) + return -1; + if (lookup1->low_addr > lookup2->low_addr) + return 1; + if (lookup1->high_addr < lookup2->high_addr) + return -1; + if (lookup1->high_addr > lookup2->high_addr) + return 1; + + if (lookup1->idx < lookup2->idx) + return -1; + if (lookup1->idx > lookup2->idx) + return 1; + return 0; +} + +static bfd_boolean +build_lookup_funcinfo_table (struct comp_unit * unit) +{ + struct lookup_funcinfo *lookup_funcinfo_table = unit->lookup_funcinfo_table; + unsigned int number_of_functions = unit->number_of_functions; + struct funcinfo *each; + struct lookup_funcinfo *entry; + size_t func_index; + struct arange *range; + bfd_vma low_addr, high_addr; + + if (lookup_funcinfo_table || number_of_functions == 0) + return TRUE; + + /* Create the function info lookup table. */ + lookup_funcinfo_table = (struct lookup_funcinfo *) + bfd_malloc (number_of_functions * sizeof (struct lookup_funcinfo)); + if (lookup_funcinfo_table == NULL) + return FALSE; + + /* Populate the function info lookup table. */ + func_index = number_of_functions; + for (each = unit->function_table; each; each = each->prev_func) { - bfd *abfd = unit->abfd; - asection *msec; + entry = &lookup_funcinfo_table[--func_index]; + entry->funcinfo = each; + entry->idx = func_index; - msec = bfd_get_section_by_name (abfd, ".debug_ranges"); - if (! msec) + /* Calculate the lowest and highest address for this function entry. */ + low_addr = entry->funcinfo->arange.low; + high_addr = entry->funcinfo->arange.high; + + for (range = entry->funcinfo->arange.next; range; range = range->next) { - (*_bfd_error_handler) (_("Dwarf Error: Can't find .debug_ranges section.")); - bfd_set_error (bfd_error_bad_value); - return FALSE; + if (range->low < low_addr) + low_addr = range->low; + if (range->high > high_addr) + high_addr = range->high; } - stash->dwarf_ranges_size = msec->size; - stash->dwarf_ranges_buffer - = bfd_simple_get_relocated_section_contents (abfd, msec, NULL, - stash->syms); - if (! stash->dwarf_ranges_buffer) - return FALSE; + entry->low_addr = low_addr; + entry->high_addr = high_addr; + } + + BFD_ASSERT (func_index == 0); + + /* Sort the function by address. */ + qsort (lookup_funcinfo_table, + number_of_functions, + sizeof (struct lookup_funcinfo), + compare_lookup_funcinfos); + + /* Calculate the high watermark for each function in the lookup table. */ + high_addr = lookup_funcinfo_table[0].high_addr; + for (func_index = 1; func_index < number_of_functions; func_index++) + { + entry = &lookup_funcinfo_table[func_index]; + if (entry->high_addr > high_addr) + high_addr = entry->high_addr; + else + entry->high_addr = high_addr; } + + unit->lookup_funcinfo_table = lookup_funcinfo_table; return TRUE; } -/* Function table functions. */ - -/* If ADDR is within TABLE, set FUNCTIONNAME_PTR, and return TRUE. - Note that we need to find the function that has the smallest - range that contains ADDR, to handle inlined functions without - depending upon them being ordered in TABLE by increasing range. */ +/* If ADDR is within UNIT's function tables, set FUNCTION_PTR, and return + TRUE. Note that we need to find the function that has the smallest range + that contains ADDR, to handle inlined functions without depending upon + them being ordered in TABLE by increasing range. */ static bfd_boolean lookup_address_in_function_table (struct comp_unit *unit, bfd_vma addr, - struct funcinfo **function_ptr, - const char **functionname_ptr) + struct funcinfo **function_ptr) { - struct funcinfo* each_func; + unsigned int number_of_functions = unit->number_of_functions; + struct lookup_funcinfo* lookup_funcinfo = NULL; + struct funcinfo* funcinfo = NULL; struct funcinfo* best_fit = NULL; + bfd_vma best_fit_len = 0; + bfd_size_type low, high, mid, first; struct arange *arange; - for (each_func = unit->function_table; - each_func; - each_func = each_func->prev_func) + if (number_of_functions == 0) + return FALSE; + + if (!build_lookup_funcinfo_table (unit)) + return FALSE; + + if (unit->lookup_funcinfo_table[number_of_functions - 1].high_addr < addr) + return FALSE; + + /* Find the first function in the lookup table which may contain the + specified address. */ + low = 0; + high = number_of_functions; + first = high; + while (low < high) { - for (arange = &each_func->arange; - arange; - arange = arange->next) + mid = (low + high) / 2; + lookup_funcinfo = &unit->lookup_funcinfo_table[mid]; + if (addr < lookup_funcinfo->low_addr) + high = mid; + else if (addr >= lookup_funcinfo->high_addr) + low = mid + 1; + else + high = first = mid; + } + + /* Find the 'best' match for the address. The prior algorithm defined the + best match as the function with the smallest address range containing + the specified address. This definition should probably be changed to the + innermost inline routine containing the address, but right now we want + to get the same results we did before. */ + while (first < number_of_functions) + { + if (addr < unit->lookup_funcinfo_table[first].low_addr) + break; + funcinfo = unit->lookup_funcinfo_table[first].funcinfo; + + for (arange = &funcinfo->arange; arange; arange = arange->next) { - if (addr >= arange->low && addr < arange->high) + if (addr < arange->low || addr >= arange->high) + continue; + + if (!best_fit + || arange->high - arange->low < best_fit_len + /* The following comparison is designed to return the same + match as the previous algorithm for routines which have the + same best fit length. */ + || (arange->high - arange->low == best_fit_len + && funcinfo > best_fit)) { - if (!best_fit || - ((arange->high - arange->low) < (best_fit->arange.high - best_fit->arange.low))) - best_fit = each_func; + best_fit = funcinfo; + best_fit_len = arange->high - arange->low; } } - } - if (best_fit) - { - *functionname_ptr = best_fit->name; - *function_ptr = best_fit; - return TRUE; - } - else - { - return FALSE; + first++; } + + if (!best_fit) + return FALSE; + + *function_ptr = best_fit; + return TRUE; } /* If SYM at ADDR is within function table of UNIT, set FILENAME_PTR @@ -1463,9 +2803,10 @@ lookup_symbol_in_function_table (struct comp_unit *unit, { struct funcinfo* each_func; struct funcinfo* best_fit = NULL; + bfd_vma best_fit_len = 0; struct arange *arange; const char *name = bfd_asymbol_name (sym); - asection *sec = bfd_get_section (sym); + asection *sec = bfd_asymbol_section (sym); for (each_func = unit->function_table; each_func; @@ -1481,10 +2822,12 @@ lookup_symbol_in_function_table (struct comp_unit *unit, && each_func->name && strcmp (name, each_func->name) == 0 && (!best_fit - || ((arange->high - arange->low) - < (best_fit->arange.high - best_fit->arange.low)))) - best_fit = each_func; - } + || arange->high - arange->low < best_fit_len)) + { + best_fit = each_func; + best_fit_len = arange->high - arange->low; + } + } } if (best_fit) @@ -1511,7 +2854,7 @@ lookup_symbol_in_variable_table (struct comp_unit *unit, unsigned int *linenumber_ptr) { const char *name = bfd_asymbol_name (sym); - asection *sec = bfd_get_section (sym); + asection *sec = bfd_asymbol_section (sym); struct varinfo* each; for (each = unit->variable_table; each; each = each->prev_var) @@ -1530,22 +2873,172 @@ lookup_symbol_in_variable_table (struct comp_unit *unit, *linenumber_ptr = each->line; return TRUE; } - else - return FALSE; + + return FALSE; } -static char * -find_abstract_instance_name (struct comp_unit *unit, bfd_uint64_t die_ref) +static struct comp_unit *stash_comp_unit (struct dwarf2_debug *, + struct dwarf2_debug_file *); +static bfd_boolean comp_unit_maybe_decode_line_info (struct comp_unit *); + +static bfd_boolean +find_abstract_instance (struct comp_unit *unit, + struct attribute *attr_ptr, + unsigned int recur_count, + const char **pname, + bfd_boolean *is_linkage, + char **filename_ptr, + int *linenumber_ptr) { bfd *abfd = unit->abfd; - bfd_byte *info_ptr; + bfd_byte *info_ptr = NULL; + bfd_byte *info_ptr_end; unsigned int abbrev_number, bytes_read, i; struct abbrev_info *abbrev; + bfd_uint64_t die_ref = attr_ptr->u.val; struct attribute attr; - char *name = 0; + const char *name = NULL; + + if (recur_count == 100) + { + _bfd_error_handler + (_("DWARF error: abstract instance recursion detected")); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } + + /* DW_FORM_ref_addr can reference an entry in a different CU. It + is an offset from the .debug_info section, not the current CU. */ + if (attr_ptr->form == DW_FORM_ref_addr) + { + /* We only support DW_FORM_ref_addr within the same file, so + any relocations should be resolved already. Check this by + testing for a zero die_ref; There can't be a valid reference + to the header of a .debug_info section. + DW_FORM_ref_addr is an offset relative to .debug_info. + Normally when using the GNU linker this is accomplished by + emitting a symbolic reference to a label, because .debug_info + sections are linked at zero. When there are multiple section + groups containing .debug_info, as there might be in a + relocatable object file, it would be reasonable to assume that + a symbolic reference to a label in any .debug_info section + might be used. Since we lay out multiple .debug_info + sections at non-zero VMAs (see place_sections), and read + them contiguously into dwarf_info_buffer, that means the + reference is relative to dwarf_info_buffer. */ + size_t total; + + info_ptr = unit->file->dwarf_info_buffer; + info_ptr_end = info_ptr + unit->file->dwarf_info_size; + total = info_ptr_end - info_ptr; + if (!die_ref) + return TRUE; + else if (die_ref >= total) + { + _bfd_error_handler + (_("DWARF error: invalid abstract instance DIE ref")); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } + info_ptr += die_ref; + } + else if (attr_ptr->form == DW_FORM_GNU_ref_alt) + { + bfd_boolean first_time = unit->stash->alt.dwarf_info_buffer == NULL; + + info_ptr = read_alt_indirect_ref (unit, die_ref); + if (first_time) + unit->stash->alt.info_ptr = unit->stash->alt.dwarf_info_buffer; + if (info_ptr == NULL) + { + _bfd_error_handler + (_("DWARF error: unable to read alt ref %" PRIu64), + (uint64_t) die_ref); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } + info_ptr_end = (unit->stash->alt.dwarf_info_buffer + + unit->stash->alt.dwarf_info_size); + if (unit->stash->alt.all_comp_units) + unit = unit->stash->alt.all_comp_units; + } - info_ptr = unit->info_ptr_unit + die_ref; - abbrev_number = read_unsigned_leb128 (abfd, info_ptr, &bytes_read); + if (attr_ptr->form == DW_FORM_ref_addr + || attr_ptr->form == DW_FORM_GNU_ref_alt) + { + /* Now find the CU containing this pointer. */ + if (info_ptr >= unit->info_ptr_unit && info_ptr < unit->end_ptr) + info_ptr_end = unit->end_ptr; + else + { + /* Check other CUs to see if they contain the abbrev. */ + struct comp_unit *u; + + for (u = unit->prev_unit; u != NULL; u = u->prev_unit) + if (info_ptr >= u->info_ptr_unit && info_ptr < u->end_ptr) + break; + + if (u == NULL) + for (u = unit->next_unit; u != NULL; u = u->next_unit) + if (info_ptr >= u->info_ptr_unit && info_ptr < u->end_ptr) + break; + + if (attr_ptr->form == DW_FORM_ref_addr) + while (u == NULL) + { + u = stash_comp_unit (unit->stash, &unit->stash->f); + if (u == NULL) + break; + if (info_ptr >= u->info_ptr_unit && info_ptr < u->end_ptr) + break; + u = NULL; + } + + if (attr_ptr->form == DW_FORM_GNU_ref_alt) + while (u == NULL) + { + u = stash_comp_unit (unit->stash, &unit->stash->alt); + if (u == NULL) + break; + if (info_ptr >= u->info_ptr_unit && info_ptr < u->end_ptr) + break; + u = NULL; + } + + if (u == NULL) + { + _bfd_error_handler + (_("DWARF error: unable to locate abstract instance DIE ref %" + PRIu64), (uint64_t) die_ref); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } + unit = u; + info_ptr_end = unit->end_ptr; + } + } + else + { + /* DW_FORM_ref1, DW_FORM_ref2, DW_FORM_ref4, DW_FORM_ref8 or + DW_FORM_ref_udata. These are all references relative to the + start of the current CU. */ + size_t total; + + info_ptr = unit->info_ptr_unit; + info_ptr_end = unit->end_ptr; + total = info_ptr_end - info_ptr; + if (!die_ref || die_ref >= total) + { + _bfd_error_handler + (_("DWARF error: invalid abstract instance DIE ref")); + bfd_set_error (bfd_error_bad_value); + return FALSE; + } + info_ptr += die_ref; + } + + abbrev_number = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + FALSE, info_ptr_end); info_ptr += bytes_read; if (abbrev_number) @@ -1553,27 +3046,55 @@ find_abstract_instance_name (struct comp_unit *unit, bfd_uint64_t die_ref) abbrev = lookup_abbrev (abbrev_number, unit->abbrevs); if (! abbrev) { - (*_bfd_error_handler) (_("Dwarf Error: Could not find abbrev number %u."), - abbrev_number); + _bfd_error_handler + (_("DWARF error: could not find abbrev number %u"), abbrev_number); bfd_set_error (bfd_error_bad_value); + return FALSE; } else { for (i = 0; i < abbrev->num_attrs; ++i) { - info_ptr = read_attribute (&attr, &abbrev->attrs[i], unit, info_ptr); + info_ptr = read_attribute (&attr, &abbrev->attrs[i], unit, + info_ptr, info_ptr_end); + if (info_ptr == NULL) + break; switch (attr.name) { case DW_AT_name: - /* Prefer DW_AT_MIPS_linkage_name over DW_AT_name. */ - if (name == NULL) - name = attr.u.str; + /* Prefer DW_AT_MIPS_linkage_name or DW_AT_linkage_name + over DW_AT_name. */ + if (name == NULL && is_str_attr (attr.form)) + { + name = attr.u.str; + if (non_mangled (unit->lang)) + *is_linkage = TRUE; + } break; case DW_AT_specification: - name = find_abstract_instance_name (unit, attr.u.val); + if (!find_abstract_instance (unit, &attr, recur_count + 1, + &name, is_linkage, + filename_ptr, linenumber_ptr)) + return FALSE; break; + case DW_AT_linkage_name: case DW_AT_MIPS_linkage_name: - name = attr.u.str; + /* PR 16949: Corrupt debug info can place + non-string forms into these attributes. */ + if (is_str_attr (attr.form)) + { + name = attr.u.str; + *is_linkage = TRUE; + } + break; + case DW_AT_decl_file: + if (!comp_unit_maybe_decode_line_info (unit)) + return FALSE; + *filename_ptr = concat_filename (unit->line_table, + attr.u.val); + break; + case DW_AT_decl_line: + *linenumber_ptr = attr.u.val; break; default: break; @@ -1581,48 +3102,55 @@ find_abstract_instance_name (struct comp_unit *unit, bfd_uint64_t die_ref) } } } - return (name); + *pname = name; + return TRUE; } -static void -read_rangelist (struct comp_unit *unit, struct arange *arange, bfd_uint64_t offset) +static bfd_boolean +read_rangelist (struct comp_unit *unit, struct arange *arange, + bfd_uint64_t offset) { bfd_byte *ranges_ptr; + bfd_byte *ranges_end; bfd_vma base_address = unit->base_address; - if (! unit->stash->dwarf_ranges_buffer) + if (! unit->file->dwarf_ranges_buffer) { if (! read_debug_ranges (unit)) - return; + return FALSE; } - ranges_ptr = unit->stash->dwarf_ranges_buffer + offset; + + ranges_ptr = unit->file->dwarf_ranges_buffer + offset; + if (ranges_ptr < unit->file->dwarf_ranges_buffer) + return FALSE; + ranges_end = unit->file->dwarf_ranges_buffer + unit->file->dwarf_ranges_size; for (;;) { bfd_vma low_pc; bfd_vma high_pc; - if (unit->addr_size == 4) - { - low_pc = read_4_bytes (unit->abfd, ranges_ptr); - ranges_ptr += 4; - high_pc = read_4_bytes (unit->abfd, ranges_ptr); - ranges_ptr += 4; - } - else - { - low_pc = read_8_bytes (unit->abfd, ranges_ptr); - ranges_ptr += 8; - high_pc = read_8_bytes (unit->abfd, ranges_ptr); - ranges_ptr += 8; - } + /* PR 17512: file: 62cada7d. */ + if (ranges_ptr + 2 * unit->addr_size > ranges_end) + return FALSE; + + low_pc = read_address (unit, ranges_ptr, ranges_end); + ranges_ptr += unit->addr_size; + high_pc = read_address (unit, ranges_ptr, ranges_end); + ranges_ptr += unit->addr_size; + if (low_pc == 0 && high_pc == 0) break; if (low_pc == -1UL && high_pc != -1UL) base_address = high_pc; else - arange_add (unit->abfd, arange, base_address + low_pc, base_address + high_pc); + { + if (!arange_add (unit, arange, + base_address + low_pc, base_address + high_pc)) + return FALSE; + } } + return TRUE; } /* DWARF2 Compilation unit functions. */ @@ -1635,19 +3163,23 @@ scan_unit_for_symbols (struct comp_unit *unit) { bfd *abfd = unit->abfd; bfd_byte *info_ptr = unit->first_child_die_ptr; - int nesting_level = 1; - struct funcinfo **nested_funcs; + bfd_byte *info_ptr_end = unit->end_ptr; + int nesting_level = 0; + struct nest_funcinfo { + struct funcinfo *func; + } *nested_funcs; int nested_funcs_size; /* Maintain a stack of in-scope functions and inlined functions, which we can use to set the caller_func field. */ nested_funcs_size = 32; - nested_funcs = bfd_malloc (nested_funcs_size * sizeof (struct funcinfo *)); + nested_funcs = (struct nest_funcinfo *) + bfd_malloc (nested_funcs_size * sizeof (*nested_funcs)); if (nested_funcs == NULL) return FALSE; - nested_funcs[nesting_level] = 0; + nested_funcs[nesting_level].func = 0; - while (nesting_level) + while (nesting_level >= 0) { unsigned int abbrev_number, bytes_read, i; struct abbrev_info *abbrev; @@ -1656,8 +3188,14 @@ scan_unit_for_symbols (struct comp_unit *unit) struct varinfo *var; bfd_vma low_pc = 0; bfd_vma high_pc = 0; + bfd_boolean high_pc_relative = FALSE; + + /* PR 17512: file: 9f405d9d. */ + if (info_ptr >= info_ptr_end) + goto fail; - abbrev_number = read_unsigned_leb128 (abfd, info_ptr, &bytes_read); + abbrev_number = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + FALSE, info_ptr_end); info_ptr += bytes_read; if (! abbrev_number) @@ -1666,14 +3204,21 @@ scan_unit_for_symbols (struct comp_unit *unit) continue; } - abbrev = lookup_abbrev (abbrev_number,unit->abbrevs); + abbrev = lookup_abbrev (abbrev_number, unit->abbrevs); if (! abbrev) { - (*_bfd_error_handler) (_("Dwarf Error: Could not find abbrev number %u."), - abbrev_number); + static unsigned int previous_failed_abbrev = -1U; + + /* Avoid multiple reports of the same missing abbrev. */ + if (abbrev_number != previous_failed_abbrev) + { + _bfd_error_handler + (_("DWARF error: could not find abbrev number %u"), + abbrev_number); + previous_failed_abbrev = abbrev_number; + } bfd_set_error (bfd_error_bad_value); - free (nested_funcs); - return FALSE; + goto fail; } var = NULL; @@ -1682,19 +3227,23 @@ scan_unit_for_symbols (struct comp_unit *unit) || abbrev->tag == DW_TAG_inlined_subroutine) { bfd_size_type amt = sizeof (struct funcinfo); - func = bfd_zalloc (abfd, amt); + func = (struct funcinfo *) bfd_zalloc (abfd, amt); + if (func == NULL) + goto fail; func->tag = abbrev->tag; func->prev_func = unit->function_table; unit->function_table = func; + unit->number_of_functions++; + BFD_ASSERT (!unit->cached); if (func->tag == DW_TAG_inlined_subroutine) - for (i = nesting_level - 1; i >= 1; i--) - if (nested_funcs[i]) + for (i = nesting_level; i-- != 0; ) + if (nested_funcs[i].func) { - func->caller_func = nested_funcs[i]; + func->caller_func = nested_funcs[i].func; break; } - nested_funcs[nesting_level] = func; + nested_funcs[nesting_level].func = func; } else { @@ -1702,27 +3251,35 @@ scan_unit_for_symbols (struct comp_unit *unit) if (abbrev->tag == DW_TAG_variable) { bfd_size_type amt = sizeof (struct varinfo); - var = bfd_zalloc (abfd, amt); + var = (struct varinfo *) bfd_zalloc (abfd, amt); + if (var == NULL) + goto fail; var->tag = abbrev->tag; var->stack = 1; var->prev_var = unit->variable_table; unit->variable_table = var; + /* PR 18205: Missing debug information can cause this + var to be attached to an already cached unit. */ } /* No inline function in scope at this nesting level. */ - nested_funcs[nesting_level] = 0; + nested_funcs[nesting_level].func = 0; } for (i = 0; i < abbrev->num_attrs; ++i) { - info_ptr = read_attribute (&attr, &abbrev->attrs[i], unit, info_ptr); + info_ptr = read_attribute (&attr, &abbrev->attrs[i], + unit, info_ptr, info_ptr_end); + if (info_ptr == NULL) + goto fail; if (func) { switch (attr.name) { case DW_AT_call_file: - func->caller_file = concat_filename (unit->line_table, attr.u.val); + func->caller_file = concat_filename (unit->line_table, + attr.u.val); break; case DW_AT_call_line: @@ -1730,17 +3287,35 @@ scan_unit_for_symbols (struct comp_unit *unit) break; case DW_AT_abstract_origin: - func->name = find_abstract_instance_name (unit, attr.u.val); + case DW_AT_specification: + if (!find_abstract_instance (unit, &attr, 0, + &func->name, + &func->is_linkage, + &func->file, + &func->line)) + goto fail; break; case DW_AT_name: - /* Prefer DW_AT_MIPS_linkage_name over DW_AT_name. */ - if (func->name == NULL) - func->name = attr.u.str; + /* Prefer DW_AT_MIPS_linkage_name or DW_AT_linkage_name + over DW_AT_name. */ + if (func->name == NULL && is_str_attr (attr.form)) + { + func->name = attr.u.str; + if (non_mangled (unit->lang)) + func->is_linkage = TRUE; + } break; + case DW_AT_linkage_name: case DW_AT_MIPS_linkage_name: - func->name = attr.u.str; + /* PR 16949: Corrupt debug info can place + non-string forms into these attributes. */ + if (is_str_attr (attr.form)) + { + func->name = attr.u.str; + func->is_linkage = TRUE; + } break; case DW_AT_low_pc: @@ -1749,10 +3324,12 @@ scan_unit_for_symbols (struct comp_unit *unit) case DW_AT_high_pc: high_pc = attr.u.val; + high_pc_relative = attr.form != DW_FORM_addr; break; case DW_AT_ranges: - read_rangelist (unit, &func->arange, attr.u.val); + if (!read_rangelist (unit, &func->arange, attr.u.val)) + goto fail; break; case DW_AT_decl_file: @@ -1773,7 +3350,8 @@ scan_unit_for_symbols (struct comp_unit *unit) switch (attr.name) { case DW_AT_name: - var->name = attr.u.str; + if (is_str_attr (attr.form)) + var->name = attr.u.str; break; case DW_AT_decl_file: @@ -1797,7 +3375,9 @@ scan_unit_for_symbols (struct comp_unit *unit) case DW_FORM_block1: case DW_FORM_block2: case DW_FORM_block4: - if (*attr.u.blk->data == DW_OP_addr) + case DW_FORM_exprloc: + if (attr.u.blk->data != NULL + && *attr.u.blk->data == DW_OP_addr) { var->stack = 0; @@ -1825,9 +3405,13 @@ scan_unit_for_symbols (struct comp_unit *unit) } } + if (high_pc_relative) + high_pc += low_pc; + if (func && high_pc != 0) { - arange_add (unit->abfd, &func->arange, low_pc, high_pc); + if (!arange_add (unit, &func->arange, low_pc, high_pc)) + goto fail; } if (abbrev->has_children) @@ -1836,28 +3420,29 @@ scan_unit_for_symbols (struct comp_unit *unit) if (nesting_level >= nested_funcs_size) { - struct funcinfo **tmp; + struct nest_funcinfo *tmp; nested_funcs_size *= 2; - tmp = bfd_realloc (nested_funcs, - (nested_funcs_size - * sizeof (struct funcinfo *))); + tmp = (struct nest_funcinfo *) + bfd_realloc (nested_funcs, + nested_funcs_size * sizeof (*nested_funcs)); if (tmp == NULL) - { - free (nested_funcs); - return FALSE; - } + goto fail; nested_funcs = tmp; } - nested_funcs[nesting_level] = 0; + nested_funcs[nesting_level].func = 0; } } free (nested_funcs); return TRUE; + + fail: + free (nested_funcs); + return FALSE; } -/* Parse a DWARF2 compilation unit starting at INFO_PTR. This +/* Parse a DWARF2 compilation unit starting at INFO_PTR. UNIT_LENGTH includes the compilation unit header that proceeds the DIE's, but does not include the length field that precedes each compilation unit header. END_PTR points one past the end of this comp unit. @@ -1867,8 +3452,9 @@ scan_unit_for_symbols (struct comp_unit *unit) to get to the line number information for the compilation unit. */ static struct comp_unit * -parse_comp_unit (bfd *abfd, - struct dwarf2_debug *stash, +parse_comp_unit (struct dwarf2_debug *stash, + struct dwarf2_debug_file *file, + bfd_byte *info_ptr, bfd_vma unit_length, bfd_byte *info_ptr_unit, unsigned int offset_size) @@ -1876,88 +3462,137 @@ parse_comp_unit (bfd *abfd, struct comp_unit* unit; unsigned int version; bfd_uint64_t abbrev_offset = 0; - unsigned int addr_size; + /* Initialize it just to avoid a GCC false warning. */ + unsigned int addr_size = -1; struct abbrev_info** abbrevs; unsigned int abbrev_number, bytes_read, i; struct abbrev_info *abbrev; struct attribute attr; - bfd_byte *info_ptr = stash->info_ptr; bfd_byte *end_ptr = info_ptr + unit_length; bfd_size_type amt; bfd_vma low_pc = 0; bfd_vma high_pc = 0; + bfd *abfd = file->bfd_ptr; + bfd_boolean high_pc_relative = FALSE; + enum dwarf_unit_type unit_type; - version = read_2_bytes (abfd, info_ptr); + version = read_2_bytes (abfd, info_ptr, end_ptr); info_ptr += 2; + if (version < 2 || version > 5) + { + /* PR 19872: A version number of 0 probably means that there is padding + at the end of the .debug_info section. Gold puts it there when + performing an incremental link, for example. So do not generate + an error, just return a NULL. */ + if (version) + { + _bfd_error_handler + (_("DWARF error: found dwarf version '%u', this reader" + " only handles version 2, 3, 4 and 5 information"), version); + bfd_set_error (bfd_error_bad_value); + } + return NULL; + } + + if (version < 5) + unit_type = DW_UT_compile; + else + { + unit_type = read_1_byte (abfd, info_ptr, end_ptr); + info_ptr += 1; + + addr_size = read_1_byte (abfd, info_ptr, end_ptr); + info_ptr += 1; + } + BFD_ASSERT (offset_size == 4 || offset_size == 8); if (offset_size == 4) - abbrev_offset = read_4_bytes (abfd, info_ptr); + abbrev_offset = read_4_bytes (abfd, info_ptr, end_ptr); else - abbrev_offset = read_8_bytes (abfd, info_ptr); + abbrev_offset = read_8_bytes (abfd, info_ptr, end_ptr); info_ptr += offset_size; - addr_size = read_1_byte (abfd, info_ptr); - info_ptr += 1; - if (version != 2) + if (version < 5) { - (*_bfd_error_handler) (_("Dwarf Error: found dwarf version '%u', this reader only handles version 2 information."), version); - bfd_set_error (bfd_error_bad_value); - return 0; + addr_size = read_1_byte (abfd, info_ptr, end_ptr); + info_ptr += 1; + } + + if (unit_type == DW_UT_type) + { + /* Skip type signature. */ + info_ptr += 8; + + /* Skip type offset. */ + info_ptr += offset_size; } if (addr_size > sizeof (bfd_vma)) { - (*_bfd_error_handler) (_("Dwarf Error: found address size '%u', this reader can not handle sizes greater than '%u'."), - addr_size, - (unsigned int) sizeof (bfd_vma)); + _bfd_error_handler + /* xgettext: c-format */ + (_("DWARF error: found address size '%u', this reader" + " can not handle sizes greater than '%u'"), + addr_size, + (unsigned int) sizeof (bfd_vma)); bfd_set_error (bfd_error_bad_value); - return 0; + return NULL; } if (addr_size != 2 && addr_size != 4 && addr_size != 8) { - (*_bfd_error_handler) ("Dwarf Error: found address size '%u', this reader can only handle address sizes '2', '4' and '8'.", addr_size); + _bfd_error_handler + ("DWARF error: found address size '%u', this reader" + " can only handle address sizes '2', '4' and '8'", addr_size); bfd_set_error (bfd_error_bad_value); - return 0; + return NULL; } /* Read the abbrevs for this compilation unit into a table. */ - abbrevs = read_abbrevs (abfd, abbrev_offset, stash); + abbrevs = read_abbrevs (abfd, abbrev_offset, stash, file); if (! abbrevs) - return 0; + return NULL; - abbrev_number = read_unsigned_leb128 (abfd, info_ptr, &bytes_read); + abbrev_number = _bfd_safe_read_leb128 (abfd, info_ptr, &bytes_read, + FALSE, end_ptr); info_ptr += bytes_read; if (! abbrev_number) { - (*_bfd_error_handler) (_("Dwarf Error: Bad abbrev number: %u."), - abbrev_number); - bfd_set_error (bfd_error_bad_value); - return 0; + /* PR 19872: An abbrev number of 0 probably means that there is padding + at the end of the .debug_abbrev section. Gold puts it there when + performing an incremental link, for example. So do not generate + an error, just return a NULL. */ + return NULL; } abbrev = lookup_abbrev (abbrev_number, abbrevs); if (! abbrev) { - (*_bfd_error_handler) (_("Dwarf Error: Could not find abbrev number %u."), - abbrev_number); + _bfd_error_handler (_("DWARF error: could not find abbrev number %u"), + abbrev_number); bfd_set_error (bfd_error_bad_value); - return 0; + return NULL; } amt = sizeof (struct comp_unit); - unit = bfd_zalloc (abfd, amt); + unit = (struct comp_unit *) bfd_zalloc (abfd, amt); + if (unit == NULL) + return NULL; unit->abfd = abfd; + unit->version = version; unit->addr_size = addr_size; unit->offset_size = offset_size; unit->abbrevs = abbrevs; unit->end_ptr = end_ptr; unit->stash = stash; + unit->file = file; unit->info_ptr_unit = info_ptr_unit; for (i = 0; i < abbrev->num_attrs; ++i) { - info_ptr = read_attribute (&attr, &abbrev->attrs[i], unit, info_ptr); + info_ptr = read_attribute (&attr, &abbrev->attrs[i], unit, info_ptr, end_ptr); + if (info_ptr == NULL) + return NULL; /* Store the data if it is of an attribute we want to keep in a partial symbol table. */ @@ -1969,28 +3604,41 @@ parse_comp_unit (bfd *abfd, break; case DW_AT_name: - unit->name = attr.u.str; + if (is_str_attr (attr.form)) + unit->name = attr.u.str; break; case DW_AT_low_pc: low_pc = attr.u.val; /* If the compilation unit DIE has a DW_AT_low_pc attribute, this is the base address to use when reading location - lists or range lists. */ - unit->base_address = low_pc; + lists or range lists. */ + if (abbrev->tag == DW_TAG_compile_unit) + unit->base_address = low_pc; break; case DW_AT_high_pc: high_pc = attr.u.val; + high_pc_relative = attr.form != DW_FORM_addr; break; case DW_AT_ranges: - read_rangelist (unit, &unit->arange, attr.u.val); + if (!read_rangelist (unit, &unit->arange, attr.u.val)) + return NULL; break; case DW_AT_comp_dir: { char *comp_dir = attr.u.str; + + /* PR 17512: file: 1fe726be. */ + if (! is_str_attr (attr.form)) + { + _bfd_error_handler + (_("DWARF error: DW_AT_comp_dir attribute encountered with a non-string form")); + comp_dir = NULL; + } + if (comp_dir) { /* Irix 6.2 native cc prepends .: to the compilation @@ -2004,13 +3652,20 @@ parse_comp_unit (bfd *abfd, break; } + case DW_AT_language: + unit->lang = attr.u.val; + break; + default: break; } } + if (high_pc_relative) + high_pc += low_pc; if (high_pc != 0) { - arange_add (unit->abfd, &unit->arange, low_pc, high_pc); + if (!arange_add (unit, &unit->arange, low_pc, high_pc)) + return NULL; } unit->first_child_die_ptr = info_ptr; @@ -2045,24 +3700,43 @@ comp_unit_contains_address (struct comp_unit *unit, bfd_vma addr) /* If UNIT contains ADDR, set the output parameters to the values for the line containing ADDR. The output parameters, FILENAME_PTR, - FUNCTIONNAME_PTR, and LINENUMBER_PTR, are pointers to the objects + FUNCTION_PTR, and LINENUMBER_PTR, are pointers to the objects to be filled in. - Return TRUE if UNIT contains ADDR, and no errors were encountered; - FALSE otherwise. */ + Returns the range of addresses covered by the entry that was used + to fill in *LINENUMBER_PTR or 0 if it was not filled in. */ -static bfd_boolean +static bfd_vma comp_unit_find_nearest_line (struct comp_unit *unit, bfd_vma addr, const char **filename_ptr, - const char **functionname_ptr, + struct funcinfo **function_ptr, unsigned int *linenumber_ptr, - struct dwarf2_debug *stash) + unsigned int *discriminator_ptr) { - bfd_boolean line_p; bfd_boolean func_p; - struct funcinfo *function; + if (!comp_unit_maybe_decode_line_info (unit)) + return FALSE; + + *function_ptr = NULL; + func_p = lookup_address_in_function_table (unit, addr, function_ptr); + if (func_p && (*function_ptr)->tag == DW_TAG_inlined_subroutine) + unit->stash->inliner_chain = *function_ptr; + + return lookup_address_in_line_info_table (unit->line_table, addr, + filename_ptr, + linenumber_ptr, + discriminator_ptr); +} + +/* Check to see if line info is already decoded in a comp_unit. + If not, decode it. Returns TRUE if no errors were encountered; + FALSE otherwise. */ + +static bfd_boolean +comp_unit_maybe_decode_line_info (struct comp_unit *unit) +{ if (unit->error) return FALSE; @@ -2074,7 +3748,7 @@ comp_unit_find_nearest_line (struct comp_unit *unit, return FALSE; } - unit->line_table = decode_line_info (unit, stash); + unit->line_table = decode_line_info (unit); if (! unit->line_table) { @@ -2090,15 +3764,7 @@ comp_unit_find_nearest_line (struct comp_unit *unit, } } - function = NULL; - func_p = lookup_address_in_function_table (unit, addr, - &function, functionname_ptr); - if (func_p && (function->tag == DW_TAG_inlined_subroutine)) - stash->inliner_chain = function; - line_p = lookup_address_in_line_info_table (unit->line_table, addr, - function, filename_ptr, - linenumber_ptr); - return line_p || func_p; + return TRUE; } /* If UNIT contains SYM at ADDR, set the output parameters to the @@ -2114,407 +3780,1019 @@ comp_unit_find_line (struct comp_unit *unit, asymbol *sym, bfd_vma addr, const char **filename_ptr, - unsigned int *linenumber_ptr, - struct dwarf2_debug *stash) + unsigned int *linenumber_ptr) { - if (unit->error) + if (!comp_unit_maybe_decode_line_info (unit)) return FALSE; - if (! unit->line_table) + if (sym->flags & BSF_FUNCTION) + return lookup_symbol_in_function_table (unit, sym, addr, + filename_ptr, + linenumber_ptr); + + return lookup_symbol_in_variable_table (unit, sym, addr, + filename_ptr, + linenumber_ptr); +} + +static struct funcinfo * +reverse_funcinfo_list (struct funcinfo *head) +{ + struct funcinfo *rhead; + struct funcinfo *temp; + + for (rhead = NULL; head; head = temp) { - if (! unit->stmtlist) - { - unit->error = 1; - return FALSE; - } + temp = head->prev_func; + head->prev_func = rhead; + rhead = head; + } + return rhead; +} - unit->line_table = decode_line_info (unit, stash); +static struct varinfo * +reverse_varinfo_list (struct varinfo *head) +{ + struct varinfo *rhead; + struct varinfo *temp; - if (! unit->line_table) - { - unit->error = 1; - return FALSE; - } + for (rhead = NULL; head; head = temp) + { + temp = head->prev_var; + head->prev_var = rhead; + rhead = head; + } + return rhead; +} - if (unit->first_child_die_ptr < unit->end_ptr - && ! scan_unit_for_symbols (unit)) - { - unit->error = 1; - return FALSE; - } +/* Extract all interesting funcinfos and varinfos of a compilation + unit into hash tables for faster lookup. Returns TRUE if no + errors were enountered; FALSE otherwise. */ + +static bfd_boolean +comp_unit_hash_info (struct dwarf2_debug *stash, + struct comp_unit *unit, + struct info_hash_table *funcinfo_hash_table, + struct info_hash_table *varinfo_hash_table) +{ + struct funcinfo* each_func; + struct varinfo* each_var; + bfd_boolean okay = TRUE; + + BFD_ASSERT (stash->info_hash_status != STASH_INFO_HASH_DISABLED); + + if (!comp_unit_maybe_decode_line_info (unit)) + return FALSE; + + BFD_ASSERT (!unit->cached); + + /* To preserve the original search order, we went to visit the function + infos in the reversed order of the list. However, making the list + bi-directional use quite a bit of extra memory. So we reverse + the list first, traverse the list in the now reversed order and + finally reverse the list again to get back the original order. */ + unit->function_table = reverse_funcinfo_list (unit->function_table); + for (each_func = unit->function_table; + each_func && okay; + each_func = each_func->prev_func) + { + /* Skip nameless functions. */ + if (each_func->name) + /* There is no need to copy name string into hash table as + name string is either in the dwarf string buffer or + info in the stash. */ + okay = insert_info_hash_table (funcinfo_hash_table, each_func->name, + (void*) each_func, FALSE); } + unit->function_table = reverse_funcinfo_list (unit->function_table); + if (!okay) + return FALSE; - if (sym->flags & BSF_FUNCTION) - return lookup_symbol_in_function_table (unit, sym, addr, - filename_ptr, - linenumber_ptr); - else - return lookup_symbol_in_variable_table (unit, sym, addr, - filename_ptr, - linenumber_ptr); + /* We do the same for variable infos. */ + unit->variable_table = reverse_varinfo_list (unit->variable_table); + for (each_var = unit->variable_table; + each_var && okay; + each_var = each_var->prev_var) + { + /* Skip stack vars and vars with no files or names. */ + if (each_var->stack == 0 + && each_var->file != NULL + && each_var->name != NULL) + /* There is no need to copy name string into hash table as + name string is either in the dwarf string buffer or + info in the stash. */ + okay = insert_info_hash_table (varinfo_hash_table, each_var->name, + (void*) each_var, FALSE); + } + + unit->variable_table = reverse_varinfo_list (unit->variable_table); + unit->cached = TRUE; + return okay; } /* Locate a section in a BFD containing debugging info. The search starts from the section after AFTER_SEC, or from the first section in the BFD if AFTER_SEC is NULL. The search works by examining the names of the - sections. There are two permissiable names. The first is .debug_info. - This is the standard DWARF2 name. The second is a prefix .gnu.linkonce.wi. + sections. There are three permissiable names. The first two are given + by DEBUG_SECTIONS[debug_info] (whose standard DWARF2 names are .debug_info + and .zdebug_info). The third is a prefix .gnu.linkonce.wi. This is a variation on the .debug_info section which has a checksum describing the contents appended onto the name. This allows the linker to identify and discard duplicate debugging sections for different compilation units. */ -#define DWARF2_DEBUG_INFO ".debug_info" #define GNU_LINKONCE_INFO ".gnu.linkonce.wi." static asection * -find_debug_info (bfd *abfd, asection *after_sec) +find_debug_info (bfd *abfd, const struct dwarf_debug_section *debug_sections, + asection *after_sec) { - asection * msec; + asection *msec; + const char *look; - if (after_sec) - msec = after_sec->next; - else - msec = abfd->sections; + if (after_sec == NULL) + { + look = debug_sections[debug_info].uncompressed_name; + msec = bfd_get_section_by_name (abfd, look); + if (msec != NULL) + return msec; + + look = debug_sections[debug_info].compressed_name; + if (look != NULL) + { + msec = bfd_get_section_by_name (abfd, look); + if (msec != NULL) + return msec; + } + + for (msec = abfd->sections; msec != NULL; msec = msec->next) + if (CONST_STRNEQ (msec->name, GNU_LINKONCE_INFO)) + return msec; + + return NULL; + } - while (msec) + for (msec = after_sec->next; msec != NULL; msec = msec->next) { - if (strcmp (msec->name, DWARF2_DEBUG_INFO) == 0) + look = debug_sections[debug_info].uncompressed_name; + if (strcmp (msec->name, look) == 0) return msec; - if (CONST_STRNEQ (msec->name, GNU_LINKONCE_INFO)) + look = debug_sections[debug_info].compressed_name; + if (look != NULL && strcmp (msec->name, look) == 0) return msec; - msec = msec->next; + if (CONST_STRNEQ (msec->name, GNU_LINKONCE_INFO)) + return msec; } return NULL; } -/* Unset vmas for loadable sections in STASH. */ +/* Transfer VMAs from object file to separate debug file. */ + +static void +set_debug_vma (bfd *orig_bfd, bfd *debug_bfd) +{ + asection *s, *d; + + for (s = orig_bfd->sections, d = debug_bfd->sections; + s != NULL && d != NULL; + s = s->next, d = d->next) + { + if ((d->flags & SEC_DEBUGGING) != 0) + break; + /* ??? Assumes 1-1 correspondence between sections in the + two files. */ + if (strcmp (s->name, d->name) == 0) + { + d->output_section = s->output_section; + d->output_offset = s->output_offset; + d->vma = s->vma; + } + } +} + +/* If the dwarf2 info was found in a separate debug file, return the + debug file section corresponding to the section in the original file + and the debug file symbols. */ + +static void +_bfd_dwarf2_stash_syms (struct dwarf2_debug *stash, bfd *abfd, + asection **sec, asymbol ***syms) +{ + if (stash->f.bfd_ptr != abfd) + { + asection *s, *d; + + if (*sec == NULL) + { + *syms = stash->f.syms; + return; + } + + for (s = abfd->sections, d = stash->f.bfd_ptr->sections; + s != NULL && d != NULL; + s = s->next, d = d->next) + { + if ((d->flags & SEC_DEBUGGING) != 0) + break; + if (s == *sec + && strcmp (s->name, d->name) == 0) + { + *sec = d; + *syms = stash->f.syms; + break; + } + } + } +} + +/* Unset vmas for adjusted sections in STASH. */ static void unset_sections (struct dwarf2_debug *stash) { - unsigned int i; - struct loadable_section *p; + int i; + struct adjusted_section *p; - i = stash->loadable_section_count; - p = stash->loadable_sections; + i = stash->adjusted_section_count; + p = stash->adjusted_sections; for (; i > 0; i--, p++) p->section->vma = 0; } -/* Set unique vmas for loadable sections in ABFD and save vmas in - STASH for unset_sections. */ +/* Set VMAs for allocated and .debug_info sections in ORIG_BFD, a + relocatable object file. VMAs are normally all zero in relocatable + object files, so if we want to distinguish locations in sections by + address we need to set VMAs so the sections do not overlap. We + also set VMA on .debug_info so that when we have multiple + .debug_info sections (or the linkonce variant) they also do not + overlap. The multiple .debug_info sections make up a single + logical section. ??? We should probably do the same for other + debug sections. */ static bfd_boolean -place_sections (bfd *abfd, struct dwarf2_debug *stash) +place_sections (bfd *orig_bfd, struct dwarf2_debug *stash) { - struct loadable_section *p; - unsigned int i; + bfd *abfd; + struct adjusted_section *p; + int i; + const char *debug_info_name; - if (stash->loadable_section_count != 0) + if (stash->adjusted_section_count != 0) { - i = stash->loadable_section_count; - p = stash->loadable_sections; + i = stash->adjusted_section_count; + p = stash->adjusted_sections; for (; i > 0; i--, p++) p->section->vma = p->adj_vma; + return TRUE; } - else + + debug_info_name = stash->debug_sections[debug_info].uncompressed_name; + i = 0; + abfd = orig_bfd; + while (1) { asection *sect; - bfd_vma last_vma = 0; - bfd_size_type amt; - struct loadable_section *p; - i = 0; for (sect = abfd->sections; sect != NULL; sect = sect->next) { - bfd_size_type sz; + int is_debug_info; - if (sect->vma != 0 || (sect->flags & SEC_LOAD) == 0) + if ((sect->output_section != NULL + && sect->output_section != sect + && (sect->flags & SEC_DEBUGGING) == 0) + || sect->vma != 0) continue; - sz = sect->rawsize ? sect->rawsize : sect->size; - if (sz == 0) + is_debug_info = (strcmp (sect->name, debug_info_name) == 0 + || CONST_STRNEQ (sect->name, GNU_LINKONCE_INFO)); + + if (!((sect->flags & SEC_ALLOC) != 0 && abfd == orig_bfd) + && !is_debug_info) continue; i++; } + if (abfd == stash->f.bfd_ptr) + break; + abfd = stash->f.bfd_ptr; + } - amt = i * sizeof (struct loadable_section); - p = (struct loadable_section *) bfd_zalloc (abfd, amt); - if (! p) + if (i <= 1) + stash->adjusted_section_count = -1; + else + { + bfd_vma last_vma = 0, last_dwarf = 0; + bfd_size_type amt = i * sizeof (struct adjusted_section); + + p = (struct adjusted_section *) bfd_malloc (amt); + if (p == NULL) return FALSE; - stash->loadable_sections = p; - stash->loadable_section_count = i; + stash->adjusted_sections = p; + stash->adjusted_section_count = i; - for (sect = abfd->sections; sect != NULL; sect = sect->next) + abfd = orig_bfd; + while (1) { - bfd_size_type sz; + asection *sect; - if (sect->vma != 0 || (sect->flags & SEC_LOAD) == 0) - continue; + for (sect = abfd->sections; sect != NULL; sect = sect->next) + { + bfd_size_type sz; + int is_debug_info; - sz = sect->rawsize ? sect->rawsize : sect->size; - if (sz == 0) - continue; + if ((sect->output_section != NULL + && sect->output_section != sect + && (sect->flags & SEC_DEBUGGING) == 0) + || sect->vma != 0) + continue; + + is_debug_info = (strcmp (sect->name, debug_info_name) == 0 + || CONST_STRNEQ (sect->name, GNU_LINKONCE_INFO)); + + if (!((sect->flags & SEC_ALLOC) != 0 && abfd == orig_bfd) + && !is_debug_info) + continue; + + sz = sect->rawsize ? sect->rawsize : sect->size; + + if (is_debug_info) + { + BFD_ASSERT (sect->alignment_power == 0); + sect->vma = last_dwarf; + last_dwarf += sz; + } + else + { + /* Align the new address to the current section + alignment. */ + last_vma = ((last_vma + + ~(-((bfd_vma) 1 << sect->alignment_power))) + & (-((bfd_vma) 1 << sect->alignment_power))); + sect->vma = last_vma; + last_vma += sz; + } + + p->section = sect; + p->adj_vma = sect->vma; + p++; + } + if (abfd == stash->f.bfd_ptr) + break; + abfd = stash->f.bfd_ptr; + } + } - p->section = sect; - if (last_vma != 0) + if (orig_bfd != stash->f.bfd_ptr) + set_debug_vma (orig_bfd, stash->f.bfd_ptr); + + return TRUE; +} + +/* Look up a funcinfo by name using the given info hash table. If found, + also update the locations pointed to by filename_ptr and linenumber_ptr. + + This function returns TRUE if a funcinfo that matches the given symbol + and address is found with any error; otherwise it returns FALSE. */ + +static bfd_boolean +info_hash_lookup_funcinfo (struct info_hash_table *hash_table, + asymbol *sym, + bfd_vma addr, + const char **filename_ptr, + unsigned int *linenumber_ptr) +{ + struct funcinfo* each_func; + struct funcinfo* best_fit = NULL; + bfd_vma best_fit_len = 0; + struct info_list_node *node; + struct arange *arange; + const char *name = bfd_asymbol_name (sym); + asection *sec = bfd_asymbol_section (sym); + + for (node = lookup_info_hash_table (hash_table, name); + node; + node = node->next) + { + each_func = (struct funcinfo *) node->info; + for (arange = &each_func->arange; + arange; + arange = arange->next) + { + if ((!each_func->sec || each_func->sec == sec) + && addr >= arange->low + && addr < arange->high + && (!best_fit + || arange->high - arange->low < best_fit_len)) { - /* Align the new address to the current section - alignment. */ - last_vma = ((last_vma - + ~((bfd_vma) -1 << sect->alignment_power)) - & ((bfd_vma) -1 << sect->alignment_power)); - sect->vma = last_vma; + best_fit = each_func; + best_fit_len = arange->high - arange->low; } - p->adj_vma = sect->vma; - last_vma += sect->vma + sz; + } + } - p++; + if (best_fit) + { + best_fit->sec = sec; + *filename_ptr = best_fit->file; + *linenumber_ptr = best_fit->line; + return TRUE; + } + + return FALSE; +} + +/* Look up a varinfo by name using the given info hash table. If found, + also update the locations pointed to by filename_ptr and linenumber_ptr. + + This function returns TRUE if a varinfo that matches the given symbol + and address is found with any error; otherwise it returns FALSE. */ + +static bfd_boolean +info_hash_lookup_varinfo (struct info_hash_table *hash_table, + asymbol *sym, + bfd_vma addr, + const char **filename_ptr, + unsigned int *linenumber_ptr) +{ + const char *name = bfd_asymbol_name (sym); + asection *sec = bfd_asymbol_section (sym); + struct varinfo* each; + struct info_list_node *node; + + for (node = lookup_info_hash_table (hash_table, name); + node; + node = node->next) + { + each = (struct varinfo *) node->info; + if (each->addr == addr + && (!each->sec || each->sec == sec)) + { + each->sec = sec; + *filename_ptr = each->file; + *linenumber_ptr = each->line; + return TRUE; + } + } + + return FALSE; +} + +/* Update the funcinfo and varinfo info hash tables if they are + not up to date. Returns TRUE if there is no error; otherwise + returns FALSE and disable the info hash tables. */ + +static bfd_boolean +stash_maybe_update_info_hash_tables (struct dwarf2_debug *stash) +{ + struct comp_unit *each; + + /* Exit if hash tables are up-to-date. */ + if (stash->f.all_comp_units == stash->hash_units_head) + return TRUE; + + if (stash->hash_units_head) + each = stash->hash_units_head->prev_unit; + else + each = stash->f.last_comp_unit; + + while (each) + { + if (!comp_unit_hash_info (stash, each, stash->funcinfo_hash_table, + stash->varinfo_hash_table)) + { + stash->info_hash_status = STASH_INFO_HASH_DISABLED; + return FALSE; } + each = each->prev_unit; } + stash->hash_units_head = stash->f.all_comp_units; return TRUE; } -/* The DWARF2 version of find_nearest_line. Return TRUE if the line - is found without error. ADDR_SIZE is the number of bytes in the - initial .debug_info length field and in the abbreviation offset. - You may use zero to indicate that the default value should be - used. */ +/* Check consistency of info hash tables. This is for debugging only. */ -bfd_boolean -_bfd_dwarf2_find_nearest_line (bfd *abfd, - asection *section, - asymbol **symbols, - bfd_vma offset, - const char **filename_ptr, - const char **functionname_ptr, - unsigned int *linenumber_ptr, - unsigned int addr_size, - void **pinfo) +static void ATTRIBUTE_UNUSED +stash_verify_info_hash_table (struct dwarf2_debug *stash) { - /* Read each compilation unit from the section .debug_info, and check - to see if it contains the address we are searching for. If yes, - lookup the address, and return the line number info. If no, go - on to the next compilation unit. + struct comp_unit *each_unit; + struct funcinfo *each_func; + struct varinfo *each_var; + struct info_list_node *node; + bfd_boolean found; + + for (each_unit = stash->f.all_comp_units; + each_unit; + each_unit = each_unit->next_unit) + { + for (each_func = each_unit->function_table; + each_func; + each_func = each_func->prev_func) + { + if (!each_func->name) + continue; + node = lookup_info_hash_table (stash->funcinfo_hash_table, + each_func->name); + BFD_ASSERT (node); + found = FALSE; + while (node && !found) + { + found = node->info == each_func; + node = node->next; + } + BFD_ASSERT (found); + } - We keep a list of all the previously read compilation units, and - a pointer to the next un-read compilation unit. Check the - previously read units before reading more. */ - struct dwarf2_debug *stash; + for (each_var = each_unit->variable_table; + each_var; + each_var = each_var->prev_var) + { + if (!each_var->name || !each_var->file || each_var->stack) + continue; + node = lookup_info_hash_table (stash->varinfo_hash_table, + each_var->name); + BFD_ASSERT (node); + found = FALSE; + while (node && !found) + { + found = node->info == each_var; + node = node->next; + } + BFD_ASSERT (found); + } + } +} - /* What address are we looking for? */ - bfd_vma addr; +/* Check to see if we want to enable the info hash tables, which consume + quite a bit of memory. Currently we only check the number times + bfd_dwarf2_find_line is called. In the future, we may also want to + take the number of symbols into account. */ - struct comp_unit* each; +static void +stash_maybe_enable_info_hash_tables (bfd *abfd, struct dwarf2_debug *stash) +{ + BFD_ASSERT (stash->info_hash_status == STASH_INFO_HASH_OFF); - bfd_vma found = FALSE; + if (stash->info_hash_count++ < STASH_INFO_HASH_TRIGGER) + return; - stash = *pinfo; + /* FIXME: Maybe we should check the reduce_memory_overheads + and optimize fields in the bfd_link_info structure ? */ - if (! stash) + /* Create hash tables. */ + stash->funcinfo_hash_table = create_info_hash_table (abfd); + stash->varinfo_hash_table = create_info_hash_table (abfd); + if (!stash->funcinfo_hash_table || !stash->varinfo_hash_table) { - bfd_size_type amt = sizeof (struct dwarf2_debug); + /* Turn off info hashes if any allocation above fails. */ + stash->info_hash_status = STASH_INFO_HASH_DISABLED; + return; + } + /* We need a forced update so that the info hash tables will + be created even though there is no compilation unit. That + happens if STASH_INFO_HASH_TRIGGER is 0. */ + if (stash_maybe_update_info_hash_tables (stash)) + stash->info_hash_status = STASH_INFO_HASH_ON; +} - stash = bfd_zalloc (abfd, amt); - if (! stash) - return FALSE; +/* Find the file and line associated with a symbol and address using the + info hash tables of a stash. If there is a match, the function returns + TRUE and update the locations pointed to by filename_ptr and linenumber_ptr; + otherwise it returns FALSE. */ + +static bfd_boolean +stash_find_line_fast (struct dwarf2_debug *stash, + asymbol *sym, + bfd_vma addr, + const char **filename_ptr, + unsigned int *linenumber_ptr) +{ + BFD_ASSERT (stash->info_hash_status == STASH_INFO_HASH_ON); + + if (sym->flags & BSF_FUNCTION) + return info_hash_lookup_funcinfo (stash->funcinfo_hash_table, sym, addr, + filename_ptr, linenumber_ptr); + return info_hash_lookup_varinfo (stash->varinfo_hash_table, sym, addr, + filename_ptr, linenumber_ptr); +} + +/* Save current section VMAs. */ + +static bfd_boolean +save_section_vma (const bfd *abfd, struct dwarf2_debug *stash) +{ + asection *s; + unsigned int i; + + if (abfd->section_count == 0) + return TRUE; + stash->sec_vma = bfd_malloc (sizeof (*stash->sec_vma) * abfd->section_count); + if (stash->sec_vma == NULL) + return FALSE; + stash->sec_vma_count = abfd->section_count; + for (i = 0, s = abfd->sections; + s != NULL && i < abfd->section_count; + i++, s = s->next) + { + if (s->output_section != NULL) + stash->sec_vma[i] = s->output_section->vma + s->output_offset; + else + stash->sec_vma[i] = s->vma; } + return TRUE; +} - /* In a relocatable file, 2 functions may have the same address. - We change the section vma so that they won't overlap. */ - if ((abfd->flags & (EXEC_P | DYNAMIC)) == 0) +/* Compare current section VMAs against those at the time the stash + was created. If find_nearest_line is used in linker warnings or + errors early in the link process, the debug info stash will be + invalid for later calls. This is because we relocate debug info + sections, so the stashed section contents depend on symbol values, + which in turn depend on section VMAs. */ + +static bfd_boolean +section_vma_same (const bfd *abfd, const struct dwarf2_debug *stash) +{ + asection *s; + unsigned int i; + + /* PR 24334: If the number of sections in ABFD has changed between + when the stash was created and now, then we cannot trust the + stashed vma information. */ + if (abfd->section_count != stash->sec_vma_count) + return FALSE; + + for (i = 0, s = abfd->sections; + s != NULL && i < abfd->section_count; + i++, s = s->next) { - if (! place_sections (abfd, stash)) + bfd_vma vma; + + if (s->output_section != NULL) + vma = s->output_section->vma + s->output_offset; + else + vma = s->vma; + if (vma != stash->sec_vma[i]) return FALSE; } + return TRUE; +} - addr = offset; - if (section->output_section) - addr += section->output_section->vma + section->output_offset; +/* Read debug information from DEBUG_BFD when DEBUG_BFD is specified. + If DEBUG_BFD is not specified, we read debug information from ABFD + or its gnu_debuglink. The results will be stored in PINFO. + The function returns TRUE iff debug information is ready. */ + +bfd_boolean +_bfd_dwarf2_slurp_debug_info (bfd *abfd, bfd *debug_bfd, + const struct dwarf_debug_section *debug_sections, + asymbol **symbols, + void **pinfo, + bfd_boolean do_place) +{ + bfd_size_type amt = sizeof (struct dwarf2_debug); + bfd_size_type total_size; + asection *msec; + struct dwarf2_debug *stash = (struct dwarf2_debug *) *pinfo; + + if (stash != NULL) + { + if (stash->orig_bfd == abfd + && section_vma_same (abfd, stash)) + { + /* Check that we did previously find some debug information + before attempting to make use of it. */ + if (stash->f.bfd_ptr != NULL) + { + if (do_place && !place_sections (abfd, stash)) + return FALSE; + return TRUE; + } + + return FALSE; + } + _bfd_dwarf2_cleanup_debug_info (abfd, pinfo); + memset (stash, 0, amt); + } else - addr += section->vma; - *filename_ptr = NULL; - *functionname_ptr = NULL; - *linenumber_ptr = 0; + { + stash = (struct dwarf2_debug *) bfd_zalloc (abfd, amt); + if (! stash) + return FALSE; + } + stash->orig_bfd = abfd; + stash->debug_sections = debug_sections; + stash->f.syms = symbols; + if (!save_section_vma (abfd, stash)) + return FALSE; + + stash->f.abbrev_offsets = htab_create_alloc (10, hash_abbrev, eq_abbrev, + del_abbrev, calloc, free); + if (!stash->f.abbrev_offsets) + return FALSE; + + stash->alt.abbrev_offsets = htab_create_alloc (10, hash_abbrev, eq_abbrev, + del_abbrev, calloc, free); + if (!stash->alt.abbrev_offsets) + return FALSE; - /* The DWARF2 spec says that the initial length field, and the - offset of the abbreviation table, should both be 4-byte values. - However, some compilers do things differently. */ - if (addr_size == 0) - addr_size = 4; - BFD_ASSERT (addr_size == 4 || addr_size == 8); + *pinfo = stash; - if (! *pinfo) + if (debug_bfd == NULL) + debug_bfd = abfd; + + msec = find_debug_info (debug_bfd, debug_sections, NULL); + if (msec == NULL && abfd == debug_bfd) { - bfd_size_type total_size; - asection *msec; + char * debug_filename; - *pinfo = stash; + debug_filename = bfd_follow_build_id_debuglink (abfd, DEBUGDIR); + if (debug_filename == NULL) + debug_filename = bfd_follow_gnu_debuglink (abfd, DEBUGDIR); - msec = find_debug_info (abfd, NULL); - if (! msec) - /* No dwarf2 info. Note that at this point the stash - has been allocated, but contains zeros, this lets - future calls to this function fail quicker. */ - goto done; - - /* There can be more than one DWARF2 info section in a BFD these days. - Read them all in and produce one large stash. We do this in two - passes - in the first pass we just accumulate the section sizes. - In the second pass we read in the section's contents. The allows - us to avoid reallocing the data as we add sections to the stash. */ - for (total_size = 0; msec; msec = find_debug_info (abfd, msec)) - total_size += msec->size; - - stash->info_ptr = bfd_alloc (abfd, total_size); - if (stash->info_ptr == NULL) - goto done; - - stash->info_ptr_end = stash->info_ptr; - - for (msec = find_debug_info (abfd, NULL); + if (debug_filename == NULL) + /* No dwarf2 info, and no gnu_debuglink to follow. + Note that at this point the stash has been allocated, but + contains zeros. This lets future calls to this function + fail more quickly. */ + return FALSE; + + debug_bfd = bfd_openr (debug_filename, NULL); + free (debug_filename); + if (debug_bfd == NULL) + /* FIXME: Should we report our failure to follow the debuglink ? */ + return FALSE; + + /* Set BFD_DECOMPRESS to decompress debug sections. */ + debug_bfd->flags |= BFD_DECOMPRESS; + if (!bfd_check_format (debug_bfd, bfd_object) + || (msec = find_debug_info (debug_bfd, + debug_sections, NULL)) == NULL + || !bfd_generic_link_read_symbols (debug_bfd)) + { + bfd_close (debug_bfd); + return FALSE; + } + + symbols = bfd_get_outsymbols (debug_bfd); + stash->f.syms = symbols; + stash->close_on_cleanup = TRUE; + } + stash->f.bfd_ptr = debug_bfd; + + if (do_place + && !place_sections (abfd, stash)) + return FALSE; + + /* There can be more than one DWARF2 info section in a BFD these + days. First handle the easy case when there's only one. If + there's more than one, try case two: none of the sections is + compressed. In that case, read them all in and produce one + large stash. We do this in two passes - in the first pass we + just accumulate the section sizes, and in the second pass we + read in the section's contents. (The allows us to avoid + reallocing the data as we add sections to the stash.) If + some or all sections are compressed, then do things the slow + way, with a bunch of reallocs. */ + + if (! find_debug_info (debug_bfd, debug_sections, msec)) + { + /* Case 1: only one info section. */ + total_size = msec->size; + if (! read_section (debug_bfd, &stash->debug_sections[debug_info], + symbols, 0, + &stash->f.dwarf_info_buffer, &total_size)) + return FALSE; + } + else + { + /* Case 2: multiple sections. */ + for (total_size = 0; msec; - msec = find_debug_info (abfd, msec)) + msec = find_debug_info (debug_bfd, debug_sections, msec)) + { + /* Catch PR25070 testcase overflowing size calculation here. */ + if (total_size + msec->size < total_size + || total_size + msec->size < msec->size) + { + bfd_set_error (bfd_error_no_memory); + return FALSE; + } + total_size += msec->size; + } + + stash->f.dwarf_info_buffer = (bfd_byte *) bfd_malloc (total_size); + if (stash->f.dwarf_info_buffer == NULL) + return FALSE; + + total_size = 0; + for (msec = find_debug_info (debug_bfd, debug_sections, NULL); + msec; + msec = find_debug_info (debug_bfd, debug_sections, msec)) { bfd_size_type size; - bfd_size_type start; size = msec->size; if (size == 0) continue; - start = stash->info_ptr_end - stash->info_ptr; + if (!(bfd_simple_get_relocated_section_contents + (debug_bfd, msec, stash->f.dwarf_info_buffer + total_size, + symbols))) + return FALSE; - if ((bfd_simple_get_relocated_section_contents - (abfd, msec, stash->info_ptr + start, symbols)) == NULL) - continue; - - stash->info_ptr_end = stash->info_ptr + start + size; + total_size += size; } - - BFD_ASSERT (stash->info_ptr_end == stash->info_ptr + total_size); - - stash->sec = find_debug_info (abfd, NULL); - stash->sec_info_ptr = stash->info_ptr; - stash->syms = symbols; } - /* A null info_ptr indicates that there is no dwarf2 info - (or that an error occured while setting up the stash). */ - if (! stash->info_ptr) - goto done; + stash->f.info_ptr = stash->f.dwarf_info_buffer; + stash->f.dwarf_info_size = total_size; + return TRUE; +} - stash->inliner_chain = NULL; +/* Parse the next DWARF2 compilation unit at FILE->INFO_PTR. */ - /* Check the previously read comp. units first. */ - for (each = stash->all_comp_units; each; each = each->next_unit) - if (comp_unit_contains_address (each, addr) - && comp_unit_find_nearest_line (each, addr, filename_ptr, - functionname_ptr, - linenumber_ptr, stash)) - { - found = TRUE; - goto done; - } +static struct comp_unit * +stash_comp_unit (struct dwarf2_debug *stash, struct dwarf2_debug_file *file) +{ + bfd_size_type length; + unsigned int offset_size; + bfd_byte *info_ptr_unit = file->info_ptr; + bfd_byte *info_ptr_end = file->dwarf_info_buffer + file->dwarf_info_size; - /* Read each remaining comp. units checking each as they are read. */ - while (stash->info_ptr < stash->info_ptr_end) + if (file->info_ptr >= info_ptr_end) + return NULL; + + length = read_4_bytes (file->bfd_ptr, file->info_ptr, info_ptr_end); + /* A 0xffffff length is the DWARF3 way of indicating + we use 64-bit offsets, instead of 32-bit offsets. */ + if (length == 0xffffffff) + { + offset_size = 8; + length = read_8_bytes (file->bfd_ptr, file->info_ptr + 4, + info_ptr_end); + file->info_ptr += 12; + } + /* A zero length is the IRIX way of indicating 64-bit offsets, + mostly because the 64-bit length will generally fit in 32 + bits, and the endianness helps. */ + else if (length == 0) { - bfd_vma length; - unsigned int offset_size = addr_size; - bfd_byte *info_ptr_unit = stash->info_ptr; + offset_size = 8; + length = read_4_bytes (file->bfd_ptr, file->info_ptr + 4, + info_ptr_end); + file->info_ptr += 8; + } + /* In the absence of the hints above, we assume 32-bit DWARF2 + offsets even for targets with 64-bit addresses, because: + a) most of the time these targets will not have generated + more than 2Gb of debug info and so will not need 64-bit + offsets, + and + b) if they do use 64-bit offsets but they are not using + the size hints that are tested for above then they are + not conforming to the DWARF3 standard anyway. */ + else + { + offset_size = 4; + file->info_ptr += 4; + } - length = read_4_bytes (abfd, stash->info_ptr); - /* A 0xffffff length is the DWARF3 way of indicating we use - 64-bit offsets, instead of 32-bit offsets. */ - if (length == 0xffffffff) - { - offset_size = 8; - length = read_8_bytes (abfd, stash->info_ptr + 4); - stash->info_ptr += 12; - } - /* A zero length is the IRIX way of indicating 64-bit offsets, - mostly because the 64-bit length will generally fit in 32 - bits, and the endianness helps. */ - else if (length == 0) - { - offset_size = 8; - length = read_4_bytes (abfd, stash->info_ptr + 4); - stash->info_ptr += 8; - } - /* In the absence of the hints above, we assume addr_size-sized - offsets, for backward-compatibility with pre-DWARF3 64-bit - platforms. */ - else if (addr_size == 8) + if (length != 0 + && file->info_ptr + length <= info_ptr_end + && file->info_ptr + length > file->info_ptr) + { + struct comp_unit *each = parse_comp_unit (stash, file, + file->info_ptr, length, + info_ptr_unit, offset_size); + if (each) { - length = read_8_bytes (abfd, stash->info_ptr); - stash->info_ptr += 8; + if (file->all_comp_units) + file->all_comp_units->prev_unit = each; + else + file->last_comp_unit = each; + + each->next_unit = file->all_comp_units; + file->all_comp_units = each; + + file->info_ptr += length; + return each; } - else - stash->info_ptr += 4; + } - if (length > 0) - { - each = parse_comp_unit (abfd, stash, length, info_ptr_unit, - offset_size); - stash->info_ptr += length; + /* Don't trust any of the DWARF info after a corrupted length or + parse error. */ + file->info_ptr = info_ptr_end; + return NULL; +} - if ((bfd_vma) (stash->info_ptr - stash->sec_info_ptr) - == stash->sec->size) - { - stash->sec = find_debug_info (abfd, stash->sec); - stash->sec_info_ptr = stash->info_ptr; - } +/* Hash function for an asymbol. */ - if (each) - { - each->next_unit = stash->all_comp_units; - stash->all_comp_units = each; - - /* DW_AT_low_pc and DW_AT_high_pc are optional for - compilation units. If we don't have them (i.e., - unit->high == 0), we need to consult the line info - table to see if a compilation unit contains the given - address. */ - if ((each->arange.high == 0 - || comp_unit_contains_address (each, addr)) - && comp_unit_find_nearest_line (each, addr, - filename_ptr, - functionname_ptr, - linenumber_ptr, - stash)) - { - found = TRUE; - goto done; - } - } +static hashval_t +hash_asymbol (const void *sym) +{ + const asymbol *asym = sym; + return htab_hash_string (asym->name); +} + +/* Equality function for asymbols. */ + +static int +eq_asymbol (const void *a, const void *b) +{ + const asymbol *sa = a; + const asymbol *sb = b; + return strcmp (sa->name, sb->name) == 0; +} + +/* Scan the debug information in PINFO looking for a DW_TAG_subprogram + abbrev with a DW_AT_low_pc attached to it. Then lookup that same + symbol in SYMBOLS and return the difference between the low_pc and + the symbol's address. Returns 0 if no suitable symbol could be found. */ + +bfd_signed_vma +_bfd_dwarf2_find_symbol_bias (asymbol ** symbols, void ** pinfo) +{ + struct dwarf2_debug *stash; + struct comp_unit * unit; + htab_t sym_hash; + bfd_signed_vma result = 0; + asymbol ** psym; + + stash = (struct dwarf2_debug *) *pinfo; + + if (stash == NULL || symbols == NULL) + return 0; + + sym_hash = htab_create_alloc (10, hash_asymbol, eq_asymbol, + NULL, xcalloc, free); + for (psym = symbols; * psym != NULL; psym++) + { + asymbol * sym = * psym; + + if (sym->flags & BSF_FUNCTION && sym->section != NULL) + { + void **slot = htab_find_slot (sym_hash, sym, INSERT); + *slot = sym; } } -done: - if ((abfd->flags & (EXEC_P | DYNAMIC)) == 0) - unset_sections (stash); + for (unit = stash->f.all_comp_units; unit; unit = unit->next_unit) + { + struct funcinfo * func; - return found; -} + comp_unit_maybe_decode_line_info (unit); + + for (func = unit->function_table; func != NULL; func = func->prev_func) + if (func->name && func->arange.low) + { + asymbol search, *sym; -/* The DWARF2 version of find_line. Return TRUE if the line is found - without error. */ + /* FIXME: Do we need to scan the aranges looking for the lowest pc value ? */ -bfd_boolean -_bfd_dwarf2_find_line (bfd *abfd, - asymbol **symbols, - asymbol *symbol, - const char **filename_ptr, - unsigned int *linenumber_ptr, - unsigned int addr_size, - void **pinfo) + search.name = func->name; + sym = htab_find (sym_hash, &search); + if (sym != NULL) + { + result = ((bfd_signed_vma) func->arange.low) - + ((bfd_signed_vma) (sym->value + sym->section->vma)); + goto done; + } + } + } + + done: + htab_delete (sym_hash); + return result; +} + +/* Find the source code location of SYMBOL. If SYMBOL is NULL + then find the nearest source code location corresponding to + the address SECTION + OFFSET. + Returns 1 if the line is found without error and fills in + FILENAME_PTR and LINENUMBER_PTR. In the case where SYMBOL was + NULL the FUNCTIONNAME_PTR is also filled in. + Returns 2 if partial information from _bfd_elf_find_function is + returned (function and maybe file) by looking at symbols. DWARF2 + info is present but not regarding the requested code location. + Returns 0 otherwise. + SYMBOLS contains the symbol table for ABFD. + DEBUG_SECTIONS contains the name of the dwarf debug sections. */ + +int +_bfd_dwarf2_find_nearest_line (bfd *abfd, + asymbol **symbols, + asymbol *symbol, + asection *section, + bfd_vma offset, + const char **filename_ptr, + const char **functionname_ptr, + unsigned int *linenumber_ptr, + unsigned int *discriminator_ptr, + const struct dwarf_debug_section *debug_sections, + void **pinfo) { /* Read each compilation unit from the section .debug_info, and check to see if it contains the address we are searching for. If yes, @@ -2525,200 +4803,235 @@ _bfd_dwarf2_find_line (bfd *abfd, a pointer to the next un-read compilation unit. Check the previously read units before reading more. */ struct dwarf2_debug *stash; - /* What address are we looking for? */ bfd_vma addr; - struct comp_unit* each; + struct funcinfo *function = NULL; + int found = FALSE; + bfd_boolean do_line; - asection *section; - - bfd_boolean found = FALSE; + *filename_ptr = NULL; + if (functionname_ptr != NULL) + *functionname_ptr = NULL; + *linenumber_ptr = 0; + if (discriminator_ptr) + *discriminator_ptr = 0; - section = bfd_get_section (symbol); + if (! _bfd_dwarf2_slurp_debug_info (abfd, NULL, debug_sections, + symbols, pinfo, + (abfd->flags & (EXEC_P | DYNAMIC)) == 0)) + return FALSE; - stash = *pinfo; + stash = (struct dwarf2_debug *) *pinfo; - if (! stash) + do_line = symbol != NULL; + if (do_line) { - bfd_size_type amt = sizeof (struct dwarf2_debug); - - stash = bfd_zalloc (abfd, amt); - if (! stash) - return FALSE; + BFD_ASSERT (section == NULL && offset == 0 && functionname_ptr == NULL); + section = bfd_asymbol_section (symbol); + addr = symbol->value; } - - /* In a relocatable file, 2 functions may have the same address. - We change the section vma so that they won't overlap. */ - if (!stash && (abfd->flags & (EXEC_P | DYNAMIC)) == 0) + else { - if (! place_sections (abfd, stash)) - return FALSE; + BFD_ASSERT (section != NULL && functionname_ptr != NULL); + addr = offset; + + /* If we have no SYMBOL but the section we're looking at is not a + code section, then take a look through the list of symbols to see + if we have a symbol at the address we're looking for. If we do + then use this to look up line information. This will allow us to + give file and line results for data symbols. We exclude code + symbols here, if we look up a function symbol and then look up the + line information we'll actually return the line number for the + opening '{' rather than the function definition line. This is + because looking up by symbol uses the line table, in which the + first line for a function is usually the opening '{', while + looking up the function by section + offset uses the + DW_AT_decl_line from the function DW_TAG_subprogram for the line, + which will be the line of the function name. */ + if (symbols != NULL && (section->flags & SEC_CODE) == 0) + { + asymbol **tmp; + + for (tmp = symbols; (*tmp) != NULL; ++tmp) + if ((*tmp)->the_bfd == abfd + && (*tmp)->section == section + && (*tmp)->value == offset + && ((*tmp)->flags & BSF_SECTION_SYM) == 0) + { + symbol = *tmp; + do_line = TRUE; + /* For local symbols, keep going in the hope we find a + global. */ + if ((symbol->flags & BSF_GLOBAL) != 0) + break; + } + } } - addr = symbol->value; if (section->output_section) addr += section->output_section->vma + section->output_offset; else addr += section->vma; - *filename_ptr = NULL; - *filename_ptr = NULL; - *linenumber_ptr = 0; - - if (! *pinfo) - { - bfd_size_type total_size; - asection *msec; - - *pinfo = stash; - - msec = find_debug_info (abfd, NULL); - if (! msec) - /* No dwarf2 info. Note that at this point the stash - has been allocated, but contains zeros, this lets - future calls to this function fail quicker. */ - goto done; - - /* There can be more than one DWARF2 info section in a BFD these days. - Read them all in and produce one large stash. We do this in two - passes - in the first pass we just accumulate the section sizes. - In the second pass we read in the section's contents. The allows - us to avoid reallocing the data as we add sections to the stash. */ - for (total_size = 0; msec; msec = find_debug_info (abfd, msec)) - total_size += msec->size; - - stash->info_ptr = bfd_alloc (abfd, total_size); - if (stash->info_ptr == NULL) - goto done; - - stash->info_ptr_end = stash->info_ptr; - - for (msec = find_debug_info (abfd, NULL); - msec; - msec = find_debug_info (abfd, msec)) - { - bfd_size_type size; - bfd_size_type start; - - size = msec->size; - if (size == 0) - continue; - - start = stash->info_ptr_end - stash->info_ptr; - - if ((bfd_simple_get_relocated_section_contents - (abfd, msec, stash->info_ptr + start, symbols)) == NULL) - continue; - - stash->info_ptr_end = stash->info_ptr + start + size; - } - - BFD_ASSERT (stash->info_ptr_end == stash->info_ptr + total_size); - - stash->sec = find_debug_info (abfd, NULL); - stash->sec_info_ptr = stash->info_ptr; - stash->syms = symbols; - } - /* A null info_ptr indicates that there is no dwarf2 info (or that an error occured while setting up the stash). */ - if (! stash->info_ptr) - goto done; + if (! stash->f.info_ptr) + return FALSE; stash->inliner_chain = NULL; /* Check the previously read comp. units first. */ - for (each = stash->all_comp_units; each; each = each->next_unit) - if ((symbol->flags & BSF_FUNCTION) == 0 - || comp_unit_contains_address (each, addr)) - { - found = comp_unit_find_line (each, symbol, addr, filename_ptr, - linenumber_ptr, stash); - if (found) - goto done; - } - - /* The DWARF2 spec says that the initial length field, and the - offset of the abbreviation table, should both be 4-byte values. - However, some compilers do things differently. */ - if (addr_size == 0) - addr_size = 4; - BFD_ASSERT (addr_size == 4 || addr_size == 8); - - /* Read each remaining comp. units checking each as they are read. */ - while (stash->info_ptr < stash->info_ptr_end) + if (do_line) { - bfd_vma length; - unsigned int offset_size = addr_size; - bfd_byte *info_ptr_unit = stash->info_ptr; - - length = read_4_bytes (abfd, stash->info_ptr); - /* A 0xffffff length is the DWARF3 way of indicating we use - 64-bit offsets, instead of 32-bit offsets. */ - if (length == 0xffffffff) + /* The info hash tables use quite a bit of memory. We may not want to + always use them. We use some heuristics to decide if and when to + turn it on. */ + if (stash->info_hash_status == STASH_INFO_HASH_OFF) + stash_maybe_enable_info_hash_tables (abfd, stash); + + /* Keep info hash table up to date if they are available. Note that we + may disable the hash tables if there is any error duing update. */ + if (stash->info_hash_status == STASH_INFO_HASH_ON) + stash_maybe_update_info_hash_tables (stash); + + if (stash->info_hash_status == STASH_INFO_HASH_ON) + { + found = stash_find_line_fast (stash, symbol, addr, filename_ptr, + linenumber_ptr); + if (found) + goto done; + } + else { - offset_size = 8; - length = read_8_bytes (abfd, stash->info_ptr + 4); - stash->info_ptr += 12; + /* Check the previously read comp. units first. */ + for (each = stash->f.all_comp_units; each; each = each->next_unit) + if ((symbol->flags & BSF_FUNCTION) == 0 + || each->arange.high == 0 + || comp_unit_contains_address (each, addr)) + { + found = comp_unit_find_line (each, symbol, addr, filename_ptr, + linenumber_ptr); + if (found) + goto done; + } } - /* A zero length is the IRIX way of indicating 64-bit offsets, - mostly because the 64-bit length will generally fit in 32 - bits, and the endianness helps. */ - else if (length == 0) + } + else + { + bfd_vma min_range = (bfd_vma) -1; + const char * local_filename = NULL; + struct funcinfo *local_function = NULL; + unsigned int local_linenumber = 0; + unsigned int local_discriminator = 0; + + for (each = stash->f.all_comp_units; each; each = each->next_unit) { - offset_size = 8; - length = read_4_bytes (abfd, stash->info_ptr + 4); - stash->info_ptr += 8; + bfd_vma range = (bfd_vma) -1; + + found = ((each->arange.high == 0 + || comp_unit_contains_address (each, addr)) + && (range = (comp_unit_find_nearest_line + (each, addr, &local_filename, + &local_function, &local_linenumber, + &local_discriminator))) != 0); + if (found) + { + /* PRs 15935 15994: Bogus debug information may have provided us + with an erroneous match. We attempt to counter this by + selecting the match that has the smallest address range + associated with it. (We are assuming that corrupt debug info + will tend to result in extra large address ranges rather than + extra small ranges). + + This does mean that we scan through all of the CUs associated + with the bfd each time this function is called. But this does + have the benefit of producing consistent results every time the + function is called. */ + if (range <= min_range) + { + if (filename_ptr && local_filename) + * filename_ptr = local_filename; + if (local_function) + function = local_function; + if (discriminator_ptr && local_discriminator) + * discriminator_ptr = local_discriminator; + if (local_linenumber) + * linenumber_ptr = local_linenumber; + min_range = range; + } + } } - /* In the absence of the hints above, we assume addr_size-sized - offsets, for backward-compatibility with pre-DWARF3 64-bit - platforms. */ - else if (addr_size == 8) + + if (* linenumber_ptr) { - length = read_8_bytes (abfd, stash->info_ptr); - stash->info_ptr += 8; + found = TRUE; + goto done; } + } + + /* Read each remaining comp. units checking each as they are read. */ + while ((each = stash_comp_unit (stash, &stash->f)) != NULL) + { + /* DW_AT_low_pc and DW_AT_high_pc are optional for + compilation units. If we don't have them (i.e., + unit->high == 0), we need to consult the line info table + to see if a compilation unit contains the given + address. */ + if (do_line) + found = (((symbol->flags & BSF_FUNCTION) == 0 + || each->arange.high == 0 + || comp_unit_contains_address (each, addr)) + && comp_unit_find_line (each, symbol, addr, + filename_ptr, linenumber_ptr)); else - stash->info_ptr += 4; + found = ((each->arange.high == 0 + || comp_unit_contains_address (each, addr)) + && comp_unit_find_nearest_line (each, addr, + filename_ptr, + &function, + linenumber_ptr, + discriminator_ptr) != 0); + + if (found) + break; + } - if (length > 0) - { - each = parse_comp_unit (abfd, stash, length, info_ptr_unit, - offset_size); - stash->info_ptr += length; + done: + if (functionname_ptr && function && function->is_linkage) + *functionname_ptr = function->name; + else if (functionname_ptr + && (!*functionname_ptr + || (function && !function->is_linkage))) + { + asymbol *fun; + asymbol **syms = symbols; + asection *sec = section; - if ((bfd_vma) (stash->info_ptr - stash->sec_info_ptr) - == stash->sec->size) - { - stash->sec = find_debug_info (abfd, stash->sec); - stash->sec_info_ptr = stash->info_ptr; - } + _bfd_dwarf2_stash_syms (stash, abfd, &sec, &syms); + fun = _bfd_elf_find_function (abfd, syms, sec, offset, + *filename_ptr ? NULL : filename_ptr, + functionname_ptr); - if (each) - { - each->next_unit = stash->all_comp_units; - stash->all_comp_units = each; - - /* DW_AT_low_pc and DW_AT_high_pc are optional for - compilation units. If we don't have them (i.e., - unit->high == 0), we need to consult the line info - table to see if a compilation unit contains the given - address. */ - found = (((symbol->flags & BSF_FUNCTION) == 0 - || each->arange.high <= 0 - || comp_unit_contains_address (each, addr)) - && comp_unit_find_line (each, symbol, addr, - filename_ptr, - linenumber_ptr, - stash)); - if (found) - goto done; - } + if (!found && fun != NULL) + found = 2; + + if (function && !function->is_linkage) + { + bfd_vma sec_vma; + + sec_vma = section->vma; + if (section->output_section != NULL) + sec_vma = section->output_section->vma + section->output_offset; + if (fun != NULL + && fun->value + sec_vma == function->arange.low) + function->name = *functionname_ptr; + /* Even if we didn't find a linkage name, say that we have + to stop a repeated search of symbols. */ + function->is_linkage = TRUE; } } -done: if ((abfd->flags & (EXEC_P | DYNAMIC)) == 0) unset_sections (stash); @@ -2734,61 +5047,212 @@ _bfd_dwarf2_find_inliner_info (bfd *abfd ATTRIBUTE_UNUSED, { struct dwarf2_debug *stash; - stash = *pinfo; + stash = (struct dwarf2_debug *) *pinfo; if (stash) { struct funcinfo *func = stash->inliner_chain; + if (func && func->caller_func) { *filename_ptr = func->caller_file; *functionname_ptr = func->caller_func->name; *linenumber_ptr = func->caller_line; stash->inliner_chain = func->caller_func; - return (TRUE); + return TRUE; } } - return (FALSE); + return FALSE; } void -_bfd_dwarf2_cleanup_debug_info (bfd *abfd) +_bfd_dwarf2_cleanup_debug_info (bfd *abfd, void **pinfo) { + struct dwarf2_debug *stash = (struct dwarf2_debug *) *pinfo; struct comp_unit *each; - struct dwarf2_debug *stash; + struct dwarf2_debug_file *file; - if (abfd == NULL || elf_tdata (abfd) == NULL) + if (abfd == NULL || stash == NULL) return; - stash = elf_tdata (abfd)->dwarf2_find_line_info; - - if (stash == NULL) - return; + if (stash->varinfo_hash_table) + bfd_hash_table_free (&stash->varinfo_hash_table->base); + if (stash->funcinfo_hash_table) + bfd_hash_table_free (&stash->funcinfo_hash_table->base); - for (each = stash->all_comp_units; each; each = each->next_unit) + file = &stash->f; + while (1) { - struct abbrev_info **abbrevs = each->abbrevs; - size_t i; - - for (i = 0; i < ABBREV_HASH_SIZE; i++) + for (each = file->all_comp_units; each; each = each->next_unit) { - struct abbrev_info *abbrev = abbrevs[i]; + struct funcinfo *function_table = each->function_table; + struct varinfo *variable_table = each->variable_table; - while (abbrev) + if (each->line_table && each->line_table != file->line_table) { - free (abbrev->attrs); - abbrev = abbrev->next; + free (each->line_table->files); + free (each->line_table->dirs); + } + + if (each->lookup_funcinfo_table) + { + free (each->lookup_funcinfo_table); + each->lookup_funcinfo_table = NULL; } + + while (function_table) + { + if (function_table->file) + { + free (function_table->file); + function_table->file = NULL; + } + if (function_table->caller_file) + { + free (function_table->caller_file); + function_table->caller_file = NULL; + } + function_table = function_table->prev_func; + } + + while (variable_table) + { + if (variable_table->file) + { + free (variable_table->file); + variable_table->file = NULL; + } + variable_table = variable_table->prev_var; + } + } + + if (file->line_table) + { + free (file->line_table->files); + free (file->line_table->dirs); } + htab_delete (file->abbrev_offsets); + + free (file->dwarf_line_str_buffer); + free (file->dwarf_str_buffer); + free (file->dwarf_ranges_buffer); + free (file->dwarf_line_buffer); + free (file->dwarf_abbrev_buffer); + free (file->dwarf_info_buffer); + if (file == &stash->alt) + break; + file = &stash->alt; + } + free (stash->sec_vma); + free (stash->adjusted_sections); + if (stash->close_on_cleanup) + bfd_close (stash->f.bfd_ptr); + if (stash->alt.bfd_ptr) + bfd_close (stash->alt.bfd_ptr); +} + +/* Find the function to a particular section and offset, + for error reporting. */ + +asymbol * +_bfd_elf_find_function (bfd *abfd, + asymbol **symbols, + asection *section, + bfd_vma offset, + const char **filename_ptr, + const char **functionname_ptr) +{ + struct elf_find_function_cache + { + asection *last_section; + asymbol *func; + const char *filename; + bfd_size_type func_size; + } *cache; + + if (symbols == NULL) + return NULL; - if (each->line_table) + if (bfd_get_flavour (abfd) != bfd_target_elf_flavour) + return NULL; + + cache = elf_tdata (abfd)->elf_find_function_cache; + if (cache == NULL) + { + cache = bfd_zalloc (abfd, sizeof (*cache)); + elf_tdata (abfd)->elf_find_function_cache = cache; + if (cache == NULL) + return NULL; + } + if (cache->last_section != section + || cache->func == NULL + || offset < cache->func->value + || offset >= cache->func->value + cache->func_size) + { + asymbol *file; + bfd_vma low_func; + asymbol **p; + /* ??? Given multiple file symbols, it is impossible to reliably + choose the right file name for global symbols. File symbols are + local symbols, and thus all file symbols must sort before any + global symbols. The ELF spec may be interpreted to say that a + file symbol must sort before other local symbols, but currently + ld -r doesn't do this. So, for ld -r output, it is possible to + make a better choice of file name for local symbols by ignoring + file symbols appearing after a given local symbol. */ + enum { nothing_seen, symbol_seen, file_after_symbol_seen } state; + const struct elf_backend_data *bed = get_elf_backend_data (abfd); + + file = NULL; + low_func = 0; + state = nothing_seen; + cache->filename = NULL; + cache->func = NULL; + cache->func_size = 0; + cache->last_section = section; + + for (p = symbols; *p != NULL; p++) { - free (each->line_table->dirs); - free (each->line_table->files); + asymbol *sym = *p; + bfd_vma code_off; + bfd_size_type size; + + if ((sym->flags & BSF_FILE) != 0) + { + file = sym; + if (state == symbol_seen) + state = file_after_symbol_seen; + continue; + } + + size = bed->maybe_function_sym (sym, section, &code_off); + if (size != 0 + && code_off <= offset + && (code_off > low_func + || (code_off == low_func + && size > cache->func_size))) + { + cache->func = sym; + cache->func_size = size; + cache->filename = NULL; + low_func = code_off; + if (file != NULL + && ((sym->flags & BSF_LOCAL) != 0 + || state != file_after_symbol_seen)) + cache->filename = bfd_asymbol_name (file); + } + if (state == nothing_seen) + state = symbol_seen; } } - free (stash->dwarf_abbrev_buffer); - free (stash->dwarf_line_buffer); - free (stash->dwarf_ranges_buffer); + if (cache->func == NULL) + return NULL; + + if (filename_ptr) + *filename_ptr = cache->filename; + if (functionname_ptr) + *functionname_ptr = bfd_asymbol_name (cache->func); + + return cache->func; }