binutils-gdb/libctf/ctf-open.c

1690 lines
48 KiB
C
Raw Normal View History

libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
/* Opening CTF files.
Copyright (C) 2019 Free Software Foundation, Inc.
This file is part of libctf.
libctf 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 3, or (at your option) any later
version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; see the file COPYING. If not see
<http://www.gnu.org/licenses/>. */
#include <ctf-impl.h>
#include <stddef.h>
#include <string.h>
#include <sys/types.h>
#include <elf.h>
#include <assert.h>
#include "swap.h"
#include <bfd.h>
#include <zlib.h>
#include "elf-bfd.h"
static const ctf_dmodel_t _libctf_models[] = {
{"ILP32", CTF_MODEL_ILP32, 4, 1, 2, 4, 4},
{"LP64", CTF_MODEL_LP64, 8, 1, 2, 4, 8},
{NULL, 0, 0, 0, 0, 0, 0}
};
const char _CTF_SECTION[] = ".ctf";
const char _CTF_NULLSTR[] = "";
/* Version-sensitive accessors. */
static uint32_t
get_kind_v1 (uint32_t info)
{
return (CTF_V1_INFO_KIND (info));
}
static uint32_t
get_root_v1 (uint32_t info)
{
return (CTF_V1_INFO_ISROOT (info));
}
static uint32_t
get_vlen_v1 (uint32_t info)
{
return (CTF_V1_INFO_VLEN (info));
}
static uint32_t
get_kind_v2 (uint32_t info)
{
return (CTF_V2_INFO_KIND (info));
}
static uint32_t
get_root_v2 (uint32_t info)
{
return (CTF_V2_INFO_ISROOT (info));
}
static uint32_t
get_vlen_v2 (uint32_t info)
{
return (CTF_V2_INFO_VLEN (info));
}
static inline ssize_t
get_ctt_size_common (const ctf_file_t *fp _libctf_unused_,
const ctf_type_t *tp _libctf_unused_,
ssize_t *sizep, ssize_t *incrementp, size_t lsize,
size_t csize, size_t ctf_type_size,
size_t ctf_stype_size, size_t ctf_lsize_sent)
{
ssize_t size, increment;
if (csize == ctf_lsize_sent)
{
size = lsize;
increment = ctf_type_size;
}
else
{
size = csize;
increment = ctf_stype_size;
}
if (sizep)
*sizep = size;
if (incrementp)
*incrementp = increment;
return size;
}
static ssize_t
get_ctt_size_v1 (const ctf_file_t *fp, const ctf_type_t *tp,
ssize_t *sizep, ssize_t *incrementp)
{
ctf_type_v1_t *t1p = (ctf_type_v1_t *) tp;
return (get_ctt_size_common (fp, tp, sizep, incrementp,
CTF_TYPE_LSIZE (t1p), t1p->ctt_size,
sizeof (ctf_type_v1_t), sizeof (ctf_stype_v1_t),
CTF_LSIZE_SENT_V1));
}
/* Return the size that a v1 will be once it is converted to v2. */
static ssize_t
get_ctt_size_v2_unconverted (const ctf_file_t *fp, const ctf_type_t *tp,
ssize_t *sizep, ssize_t *incrementp)
{
ctf_type_v1_t *t1p = (ctf_type_v1_t *) tp;
return (get_ctt_size_common (fp, tp, sizep, incrementp,
CTF_TYPE_LSIZE (t1p), t1p->ctt_size,
sizeof (ctf_type_t), sizeof (ctf_stype_t),
CTF_LSIZE_SENT));
}
static ssize_t
get_ctt_size_v2 (const ctf_file_t *fp, const ctf_type_t *tp,
ssize_t *sizep, ssize_t *incrementp)
{
return (get_ctt_size_common (fp, tp, sizep, incrementp,
CTF_TYPE_LSIZE (tp), tp->ctt_size,
sizeof (ctf_type_t), sizeof (ctf_stype_t),
CTF_LSIZE_SENT));
}
static ssize_t
get_vbytes_common (unsigned short kind, ssize_t size _libctf_unused_,
size_t vlen)
{
switch (kind)
{
case CTF_K_INTEGER:
case CTF_K_FLOAT:
return (sizeof (uint32_t));
case CTF_K_SLICE:
libctf: endianness fixes Testing of the first code to generate CTF_K_SLICEs on big-endian revealed a bunch of new problems in this area. Most importantly, the trick we did earlier to avoid wasting two bytes on padding in the ctf_slice_t is best avoided: because it leads to the whole file after that point no longer being naturally aligned, all multibyte accesses from then on must use memmove() to avoid unaligned access on platforms where that is fatal. In future, this is planned, but for now we are still doing direct access in many places, so we must revert to making ctf_slice_t properly aligned for storage in an array. Rather than wasting bytes on padding, we boost the size of cts_offset and cts_bits. This is still a waste of space (we cannot have offsets or bits in bitfields > 256) but it cannot be avoided for now, and slices are not so common that this will be a serious problem. A possibly-worse endianness problem fixed at the same time involves a codepath used only for foreign-endian, uncompressed CTF files, where we were not copying the actual CTF data into the buffer, leading to libctf reading only zeroes (or, possibly, uninitialized garbage). Finally, when we read in a CTF file, we copy the header and work from the copy. We were flipping the endianness of the header copy, and of the body of the file buffer, but not of the header in the file buffer itself: so if we write the file back out again we end up with an unreadable frankenfile with header and body of different endiannesses. Fix by flipping both copies of the header. include/ * ctf.h (ctf_slice_t): Make cts_offset and cts_bits unsigned short, so following structures are properly aligned. libctf/ * ctf-open.c (get_vbytes_common): Return the new slice size. (ctf_bufopen): Flip the endianness of the CTF-section header copy. Remember to copy in the CTF data when opening an uncompressed foreign-endian CTF file. Prune useless variable manipulation.
2019-06-19 13:34:56 +02:00
return (sizeof (ctf_slice_t));
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
case CTF_K_ENUM:
return (sizeof (ctf_enum_t) * vlen);
case CTF_K_FORWARD:
case CTF_K_UNKNOWN:
case CTF_K_POINTER:
case CTF_K_TYPEDEF:
case CTF_K_VOLATILE:
case CTF_K_CONST:
case CTF_K_RESTRICT:
return 0;
default:
ctf_dprintf ("detected invalid CTF kind -- %x\n", kind);
return ECTF_CORRUPT;
}
}
static ssize_t
get_vbytes_v1 (unsigned short kind, ssize_t size, size_t vlen)
{
switch (kind)
{
case CTF_K_ARRAY:
return (sizeof (ctf_array_v1_t));
case CTF_K_FUNCTION:
return (sizeof (unsigned short) * (vlen + (vlen & 1)));
case CTF_K_STRUCT:
case CTF_K_UNION:
if (size < CTF_LSTRUCT_THRESH_V1)
return (sizeof (ctf_member_v1_t) * vlen);
else
return (sizeof (ctf_lmember_v1_t) * vlen);
}
return (get_vbytes_common (kind, size, vlen));
}
static ssize_t
get_vbytes_v2 (unsigned short kind, ssize_t size, size_t vlen)
{
switch (kind)
{
case CTF_K_ARRAY:
return (sizeof (ctf_array_t));
case CTF_K_FUNCTION:
return (sizeof (uint32_t) * (vlen + (vlen & 1)));
case CTF_K_STRUCT:
case CTF_K_UNION:
if (size < CTF_LSTRUCT_THRESH)
return (sizeof (ctf_member_t) * vlen);
else
return (sizeof (ctf_lmember_t) * vlen);
}
return (get_vbytes_common (kind, size, vlen));
}
static const ctf_fileops_t ctf_fileops[] = {
{NULL, NULL, NULL, NULL, NULL},
/* CTF_VERSION_1 */
{get_kind_v1, get_root_v1, get_vlen_v1, get_ctt_size_v1, get_vbytes_v1},
/* CTF_VERSION_1_UPGRADED_3 */
{get_kind_v2, get_root_v2, get_vlen_v2, get_ctt_size_v2, get_vbytes_v2},
/* CTF_VERSION_2 */
{get_kind_v2, get_root_v2, get_vlen_v2, get_ctt_size_v2, get_vbytes_v2},
/* CTF_VERSION_3, identical to 2: only new type kinds */
{get_kind_v2, get_root_v2, get_vlen_v2, get_ctt_size_v2, get_vbytes_v2},
};
/* Initialize the symtab translation table by filling each entry with the
offset of the CTF type or function data corresponding to each STT_FUNC or
STT_OBJECT entry in the symbol table. */
static int
init_symtab (ctf_file_t *fp, const ctf_header_t *hp,
const ctf_sect_t *sp, const ctf_sect_t *strp)
{
const unsigned char *symp = sp->cts_data;
uint32_t *xp = fp->ctf_sxlate;
uint32_t *xend = xp + fp->ctf_nsyms;
uint32_t objtoff = hp->cth_objtoff;
uint32_t funcoff = hp->cth_funcoff;
uint32_t info, vlen;
Elf64_Sym sym, *gsp;
const char *name;
/* The CTF data object and function type sections are ordered to match
the relative order of the respective symbol types in the symtab.
If no type information is available for a symbol table entry, a
pad is inserted in the CTF section. As a further optimization,
anonymous or undefined symbols are omitted from the CTF data. */
for (; xp < xend; xp++, symp += sp->cts_entsize)
{
if (sp->cts_entsize == sizeof (Elf32_Sym))
gsp = ctf_sym_to_elf64 ((Elf32_Sym *) (uintptr_t) symp, &sym);
else
gsp = (Elf64_Sym *) (uintptr_t) symp;
if (gsp->st_name < strp->cts_size)
name = (const char *) strp->cts_data + gsp->st_name;
else
name = _CTF_NULLSTR;
if (gsp->st_name == 0 || gsp->st_shndx == SHN_UNDEF
|| strcmp (name, "_START_") == 0 || strcmp (name, "_END_") == 0)
{
*xp = -1u;
continue;
}
switch (ELF64_ST_TYPE (gsp->st_info))
{
case STT_OBJECT:
if (objtoff >= hp->cth_funcoff
|| (gsp->st_shndx == SHN_EXTABS && gsp->st_value == 0))
{
*xp = -1u;
break;
}
*xp = objtoff;
objtoff += sizeof (uint32_t);
break;
case STT_FUNC:
if (funcoff >= hp->cth_typeoff)
{
*xp = -1u;
break;
}
*xp = funcoff;
info = *(uint32_t *) ((uintptr_t) fp->ctf_buf + funcoff);
vlen = LCTF_INFO_VLEN (fp, info);
/* If we encounter a zero pad at the end, just skip it. Otherwise
skip over the function and its return type (+2) and the argument
list (vlen).
*/
if (LCTF_INFO_KIND (fp, info) == CTF_K_UNKNOWN && vlen == 0)
funcoff += sizeof (uint32_t); /* Skip pad. */
else
funcoff += sizeof (uint32_t) * (vlen + 2);
break;
default:
*xp = -1u;
break;
}
}
ctf_dprintf ("loaded %lu symtab entries\n", fp->ctf_nsyms);
return 0;
}
/* Set the CTF base pointer and derive the buf pointer from it, initializing
everything in the ctf_file that depends on the base or buf pointers. */
static void
ctf_set_base (ctf_file_t *fp, const ctf_header_t *hp, void *base)
{
fp->ctf_base = base;
fp->ctf_buf = fp->ctf_base + sizeof (ctf_header_t);
fp->ctf_vars = (ctf_varent_t *) ((const char *) fp->ctf_buf +
hp->cth_varoff);
fp->ctf_nvars = (hp->cth_typeoff - hp->cth_varoff) / sizeof (ctf_varent_t);
fp->ctf_str[CTF_STRTAB_0].cts_strs = (const char *) fp->ctf_buf
+ hp->cth_stroff;
fp->ctf_str[CTF_STRTAB_0].cts_len = hp->cth_strlen;
/* If we have a parent container name and label, store the relocated
string pointers in the CTF container for easy access later. */
/* Note: before conversion, these will be set to values that will be
immediately invalidated by the conversion process, but the conversion
process will call ctf_set_base() again to fix things up. */
if (hp->cth_parlabel != 0)
fp->ctf_parlabel = ctf_strptr (fp, hp->cth_parlabel);
if (hp->cth_parname != 0)
fp->ctf_parname = ctf_strptr (fp, hp->cth_parname);
ctf_dprintf ("ctf_set_base: parent name %s (label %s)\n",
fp->ctf_parname ? fp->ctf_parname : "<NULL>",
fp->ctf_parlabel ? fp->ctf_parlabel : "<NULL>");
}
/* Free a ctf_base pointer: the pointer passed, or (if NULL) fp->ctf_base. */
static void
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
ctf_free_base (ctf_file_t *fp, unsigned char *ctf_base)
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
{
unsigned char *base;
if (ctf_base)
base = ctf_base;
else
base = (unsigned char *) fp->ctf_base;
if (base != fp->ctf_data.cts_data && base != NULL)
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
ctf_free (base);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
}
/* Set the version of the CTF file. */
/* When this is reset, LCTF_* changes behaviour, but there is no guarantee that
the variable data list associated with each type has been upgraded: the
caller must ensure this has been done in advance. */
static void
ctf_set_version (ctf_file_t * fp, ctf_header_t * cth, int ctf_version)
{
fp->ctf_version = ctf_version;
cth->cth_version = ctf_version;
fp->ctf_fileops = &ctf_fileops[ctf_version];
}
/* Upgrade the type table to CTF_VERSION_3 (really CTF_VERSION_1_UPGRADED_3).
The upgrade is not done in-place: the ctf_base is moved. ctf_strptr() must
not be called before reallocation is complete.
Type kinds not checked here due to nonexistence in older formats:
CTF_K_SLICE. */
static int
upgrade_types (ctf_file_t *fp, ctf_header_t *cth)
{
const ctf_type_v1_t *tbuf;
const ctf_type_v1_t *tend;
unsigned char *ctf_base, *old_ctf_base = (unsigned char *) fp->ctf_base;
ctf_type_t *t2buf;
ssize_t increase = 0, size, increment, v2increment, vbytes, v2bytes;
const ctf_type_v1_t *tp;
ctf_type_t *t2p;
ctf_header_t *new_cth;
tbuf = (ctf_type_v1_t *) (fp->ctf_buf + cth->cth_typeoff);
tend = (ctf_type_v1_t *) (fp->ctf_buf + cth->cth_stroff);
/* Much like init_types(), this is a two-pass process.
First, figure out the new type-section size needed. (It is possible,
in theory, for it to be less than the old size, but this is very
unlikely. It cannot be so small that cth_typeoff ends up of negative
size. We validate this with an assertion below.)
We must cater not only for changes in vlen and types sizes but also
for changes in 'increment', which happen because v2 places some types
into ctf_stype_t where v1 would be forced to use the larger non-stype. */
for (tp = tbuf; tp < tend;
tp = (ctf_type_v1_t *) ((uintptr_t) tp + increment + vbytes))
{
unsigned short kind = CTF_V1_INFO_KIND (tp->ctt_info);
unsigned long vlen = CTF_V1_INFO_VLEN (tp->ctt_info);
size = get_ctt_size_v1 (fp, (const ctf_type_t *) tp, NULL, &increment);
vbytes = get_vbytes_v1 (kind, size, vlen);
get_ctt_size_v2_unconverted (fp, (const ctf_type_t *) tp, NULL,
&v2increment);
v2bytes = get_vbytes_v2 (kind, size, vlen);
if ((vbytes < 0) || (size < 0))
return ECTF_CORRUPT;
increase += v2increment - increment; /* May be negative. */
increase += v2bytes - vbytes;
}
/* Allocate enough room for the new buffer, then copy everything but the
type section into place, and reset the base accordingly. Leave the
version number unchanged, so that LCTF_INFO_* still works on the
as-yet-untranslated type info. */
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
if ((ctf_base = ctf_alloc (fp->ctf_size + increase)) == NULL)
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
return ECTF_ZALLOC;
memcpy (ctf_base, fp->ctf_base, sizeof (ctf_header_t) + cth->cth_typeoff);
memcpy (ctf_base + sizeof (ctf_header_t) + cth->cth_stroff + increase,
fp->ctf_base + sizeof (ctf_header_t) + cth->cth_stroff,
cth->cth_strlen);
memset (ctf_base + sizeof (ctf_header_t) + cth->cth_typeoff, 0,
cth->cth_stroff - cth->cth_typeoff + increase);
/* The cth here is an automatic variable in ctf_bufopen(), and transient
(a copy maintained because at that stage the header read out of the
ctf file may be read-only). We make all modifications in the
canonical copy at ctf_base (by now, writable), then copy it back into
cth at the end. */
new_cth = (ctf_header_t *) ctf_base;
new_cth->cth_stroff += increase;
fp->ctf_size += increase;
assert (new_cth->cth_stroff >= new_cth->cth_typeoff);
ctf_set_base (fp, new_cth, ctf_base);
t2buf = (ctf_type_t *) (fp->ctf_buf + new_cth->cth_typeoff);
/* Iterate through all the types again, upgrading them.
Everything that hasn't changed can just be outright memcpy()ed.
Things that have changed need field-by-field consideration. */
for (tp = tbuf, t2p = t2buf; tp < tend;
tp = (ctf_type_v1_t *) ((uintptr_t) tp + increment + vbytes),
t2p = (ctf_type_t *) ((uintptr_t) t2p + v2increment + v2bytes))
{
unsigned short kind = CTF_V1_INFO_KIND (tp->ctt_info);
int isroot = CTF_V1_INFO_ISROOT (tp->ctt_info);
unsigned long vlen = CTF_V1_INFO_VLEN (tp->ctt_info);
ssize_t v2size;
void *vdata, *v2data;
size = get_ctt_size_v1 (fp, (const ctf_type_t *) tp, NULL, &increment);
vbytes = get_vbytes_v1 (kind, size, vlen);
t2p->ctt_name = tp->ctt_name;
t2p->ctt_info = CTF_TYPE_INFO (kind, isroot, vlen);
switch (kind)
{
case CTF_K_FUNCTION:
case CTF_K_FORWARD:
case CTF_K_TYPEDEF:
case CTF_K_POINTER:
case CTF_K_VOLATILE:
case CTF_K_CONST:
case CTF_K_RESTRICT:
t2p->ctt_type = tp->ctt_type;
break;
case CTF_K_INTEGER:
case CTF_K_FLOAT:
case CTF_K_ARRAY:
case CTF_K_STRUCT:
case CTF_K_UNION:
case CTF_K_ENUM:
case CTF_K_UNKNOWN:
libctf: fix a number of build problems found on Solaris and NetBSD - Use of nonportable <endian.h> - Use of qsort_r - Use of zlib without appropriate magic to pull in the binutils zlib - Use of off64_t without checking (fixed by dropping the unused fields that need off64_t entirely) - signedness problems due to long being too short a type on 32-bit platforms: ctf_id_t is now 'unsigned long', and CTF_ERR must be used only for functions that return ctf_id_t - One lingering use of bzero() and of <sys/errno.h> All fixed, using code from gnulib where possible. Relatedly, set cts_size in a couple of places it was missed (string table and symbol table loading upon ctf_bfdopen()). binutils/ * objdump.c (make_ctfsect): Drop cts_type, cts_flags, and cts_offset. * readelf.c (shdr_to_ctf_sect): Likewise. include/ * ctf-api.h (ctf_sect_t): Drop cts_type, cts_flags, and cts_offset. (ctf_id_t): This is now an unsigned type. (CTF_ERR): Cast it to ctf_id_t. Note that it should only be used for ctf_id_t-returning functions. libctf/ * Makefile.am (ZLIB): New. (ZLIBINC): Likewise. (AM_CFLAGS): Use them. (libctf_a_LIBADD): New, for LIBOBJS. * configure.ac: Check for zlib, endian.h, and qsort_r. * ctf-endian.h: New, providing htole64 and le64toh. * swap.h: Code style fixes. (bswap_identity_64): New. * qsort_r.c: New, from gnulib (with one added #include). * ctf-decls.h: New, providing a conditional qsort_r declaration, and unconditional definitions of MIN and MAX. * ctf-impl.h: Use it. Do not use <sys/errno.h>. (ctf_set_errno): Now returns unsigned long. * ctf-util.c (ctf_set_errno): Adjust here too. * ctf-archive.c: Use ctf-endian.h. (ctf_arc_open_by_offset): Use memset, not bzero. Drop cts_type, cts_flags and cts_offset. (ctf_arc_write): Drop debugging dependent on the size of off_t. * ctf-create.c: Provide a definition of roundup if not defined. (ctf_create): Drop cts_type, cts_flags and cts_offset. (ctf_add_reftype): Do not check if type IDs are below zero. (ctf_add_slice): Likewise. (ctf_add_typedef): Likewise. (ctf_add_member_offset): Cast error-returning ssize_t's to size_t when known error-free. Drop CTF_ERR usage for functions returning int. (ctf_add_member_encoded): Drop CTF_ERR usage for functions returning int. (ctf_add_variable): Likewise. (enumcmp): Likewise. (enumadd): Likewise. (membcmp): Likewise. (ctf_add_type): Likewise. Cast error-returning ssize_t's to size_t when known error-free. * ctf-dump.c (ctf_is_slice): Drop CTF_ERR usage for functions returning int: use CTF_ERR for functions returning ctf_type_id. (ctf_dump_label): Likewise. (ctf_dump_objts): Likewise. * ctf-labels.c (ctf_label_topmost): Likewise. (ctf_label_iter): Likewise. (ctf_label_info): Likewise. * ctf-lookup.c (ctf_func_args): Likewise. * ctf-open.c (upgrade_types): Cast to size_t where appropriate. (ctf_bufopen): Likewise. Use zlib types as needed. * ctf-types.c (ctf_member_iter): Drop CTF_ERR usage for functions returning int. (ctf_enum_iter): Likewise. (ctf_type_size): Likewise. (ctf_type_align): Likewise. Cast to size_t where appropriate. (ctf_type_kind_unsliced): Likewise. (ctf_type_kind): Likewise. (ctf_type_encoding): Likewise. (ctf_member_info): Likewise. (ctf_array_info): Likewise. (ctf_enum_value): Likewise. (ctf_type_rvisit): Likewise. * ctf-open-bfd.c (ctf_bfdopen): Drop cts_type, cts_flags and cts_offset. (ctf_simple_open): Likewise. (ctf_bfdopen_ctfsect): Likewise. Set cts_size properly. * Makefile.in: Regenerate. * aclocal.m4: Likewise. * config.h: Likewise. * configure: Likewise.
2019-05-31 11:10:51 +02:00
if ((size_t) size <= CTF_MAX_SIZE)
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
t2p->ctt_size = size;
else
{
t2p->ctt_lsizehi = CTF_SIZE_TO_LSIZE_HI (size);
t2p->ctt_lsizelo = CTF_SIZE_TO_LSIZE_LO (size);
}
break;
}
v2size = get_ctt_size_v2 (fp, t2p, NULL, &v2increment);
v2bytes = get_vbytes_v2 (kind, v2size, vlen);
/* Catch out-of-sync get_ctt_size_*(). The count goes wrong if
these are not identical (and having them different makes no
sense semantically). */
assert (size == v2size);
/* Now the varlen info. */
vdata = (void *) ((uintptr_t) tp + increment);
v2data = (void *) ((uintptr_t) t2p + v2increment);
switch (kind)
{
case CTF_K_ARRAY:
{
const ctf_array_v1_t *ap = (const ctf_array_v1_t *) vdata;
ctf_array_t *a2p = (ctf_array_t *) v2data;
a2p->cta_contents = ap->cta_contents;
a2p->cta_index = ap->cta_index;
a2p->cta_nelems = ap->cta_nelems;
break;
}
case CTF_K_STRUCT:
case CTF_K_UNION:
{
ctf_member_t tmp;
const ctf_member_v1_t *m1 = (const ctf_member_v1_t *) vdata;
const ctf_lmember_v1_t *lm1 = (const ctf_lmember_v1_t *) m1;
ctf_member_t *m2 = (ctf_member_t *) v2data;
ctf_lmember_t *lm2 = (ctf_lmember_t *) m2;
unsigned long i;
/* We walk all four pointers forward, but only reference the two
that are valid for the given size, to avoid quadruplicating all
the code. */
for (i = vlen; i != 0; i--, m1++, lm1++, m2++, lm2++)
{
size_t offset;
if (size < CTF_LSTRUCT_THRESH_V1)
{
offset = m1->ctm_offset;
tmp.ctm_name = m1->ctm_name;
tmp.ctm_type = m1->ctm_type;
}
else
{
offset = CTF_LMEM_OFFSET (lm1);
tmp.ctm_name = lm1->ctlm_name;
tmp.ctm_type = lm1->ctlm_type;
}
if (size < CTF_LSTRUCT_THRESH)
{
m2->ctm_name = tmp.ctm_name;
m2->ctm_type = tmp.ctm_type;
m2->ctm_offset = offset;
}
else
{
lm2->ctlm_name = tmp.ctm_name;
lm2->ctlm_type = tmp.ctm_type;
lm2->ctlm_offsethi = CTF_OFFSET_TO_LMEMHI (offset);
lm2->ctlm_offsetlo = CTF_OFFSET_TO_LMEMLO (offset);
}
}
break;
}
case CTF_K_FUNCTION:
{
unsigned long i;
unsigned short *a1 = (unsigned short *) vdata;
uint32_t *a2 = (uint32_t *) v2data;
for (i = vlen; i != 0; i--, a1++, a2++)
*a2 = *a1;
}
/* FALLTHRU */
default:
/* Catch out-of-sync get_vbytes_*(). */
assert (vbytes == v2bytes);
memcpy (v2data, vdata, vbytes);
}
}
/* Verify that the entire region was converted. If not, we are either
converting too much, or too little (leading to a buffer overrun either here
or at read time, in init_types().) */
assert ((size_t) t2p - (size_t) fp->ctf_buf == new_cth->cth_stroff);
ctf_set_version (fp, (ctf_header_t *) ctf_base, CTF_VERSION_1_UPGRADED_3);
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
ctf_free_base (fp, old_ctf_base);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
memcpy (cth, new_cth, sizeof (ctf_header_t));
return 0;
}
/* Initialize the type ID translation table with the byte offset of each type,
and initialize the hash tables of each named type. Upgrade the type table to
the latest supported representation in the process, if needed, and if this
recension of libctf supports upgrading. */
static int
init_types (ctf_file_t *fp, ctf_header_t *cth)
{
const ctf_type_t *tbuf;
const ctf_type_t *tend;
unsigned long pop[CTF_K_MAX + 1] = { 0 };
const ctf_type_t *tp;
ctf_hash_t *hp;
uint32_t id, dst;
uint32_t *xp;
/* We determine whether the container is a child or a parent based on
the value of cth_parname. */
int child = cth->cth_parname != 0;
int nlstructs = 0, nlunions = 0;
int err;
if (_libctf_unlikely_ (fp->ctf_version == CTF_VERSION_1))
{
int err;
if ((err = upgrade_types (fp, cth)) != 0)
return err; /* Upgrade failed. */
}
tbuf = (ctf_type_t *) (fp->ctf_buf + cth->cth_typeoff);
tend = (ctf_type_t *) (fp->ctf_buf + cth->cth_stroff);
/* We make two passes through the entire type section. In this first
pass, we count the number of each type and the total number of types. */
for (tp = tbuf; tp < tend; fp->ctf_typemax++)
{
unsigned short kind = LCTF_INFO_KIND (fp, tp->ctt_info);
unsigned long vlen = LCTF_INFO_VLEN (fp, tp->ctt_info);
ssize_t size, increment, vbytes;
(void) ctf_get_ctt_size (fp, tp, &size, &increment);
vbytes = LCTF_VBYTES (fp, kind, size, vlen);
if (vbytes < 0)
return ECTF_CORRUPT;
if (kind == CTF_K_FORWARD)
{
/* For forward declarations, ctt_type is the CTF_K_* kind for the tag,
so bump that population count too. If ctt_type is unknown, treat
the tag as a struct. */
if (tp->ctt_type == CTF_K_UNKNOWN || tp->ctt_type >= CTF_K_MAX)
pop[CTF_K_STRUCT]++;
else
pop[tp->ctt_type]++;
}
tp = (ctf_type_t *) ((uintptr_t) tp + increment + vbytes);
pop[kind]++;
}
if (child)
{
ctf_dprintf ("CTF container %p is a child\n", (void *) fp);
fp->ctf_flags |= LCTF_CHILD;
}
else
ctf_dprintf ("CTF container %p is a parent\n", (void *) fp);
/* Now that we've counted up the number of each type, we can allocate
the hash tables, type translation table, and pointer table. */
if ((fp->ctf_structs = ctf_hash_create (pop[CTF_K_STRUCT], ctf_hash_string,
ctf_hash_eq_string)) == NULL)
return ENOMEM;
if ((fp->ctf_unions = ctf_hash_create (pop[CTF_K_UNION], ctf_hash_string,
ctf_hash_eq_string)) == NULL)
return ENOMEM;
if ((fp->ctf_enums = ctf_hash_create (pop[CTF_K_ENUM], ctf_hash_string,
ctf_hash_eq_string)) == NULL)
return ENOMEM;
if ((fp->ctf_names = ctf_hash_create (pop[CTF_K_INTEGER] +
pop[CTF_K_FLOAT] +
pop[CTF_K_FUNCTION] +
pop[CTF_K_TYPEDEF] +
pop[CTF_K_POINTER] +
pop[CTF_K_VOLATILE] +
pop[CTF_K_CONST] +
pop[CTF_K_RESTRICT],
ctf_hash_string,
ctf_hash_eq_string)) == NULL)
return ENOMEM;
fp->ctf_txlate = ctf_alloc (sizeof (uint32_t) * (fp->ctf_typemax + 1));
fp->ctf_ptrtab = ctf_alloc (sizeof (uint32_t) * (fp->ctf_typemax + 1));
if (fp->ctf_txlate == NULL || fp->ctf_ptrtab == NULL)
return ENOMEM; /* Memory allocation failed. */
xp = fp->ctf_txlate;
*xp++ = 0; /* Type id 0 is used as a sentinel value. */
memset (fp->ctf_txlate, 0, sizeof (uint32_t) * (fp->ctf_typemax + 1));
memset (fp->ctf_ptrtab, 0, sizeof (uint32_t) * (fp->ctf_typemax + 1));
/* In the second pass through the types, we fill in each entry of the
type and pointer tables and add names to the appropriate hashes. */
for (id = 1, tp = tbuf; tp < tend; xp++, id++)
{
unsigned short kind = LCTF_INFO_KIND (fp, tp->ctt_info);
unsigned short flag = LCTF_INFO_ISROOT (fp, tp->ctt_info);
unsigned long vlen = LCTF_INFO_VLEN (fp, tp->ctt_info);
ssize_t size, increment, vbytes;
const char *name;
(void) ctf_get_ctt_size (fp, tp, &size, &increment);
name = ctf_strptr (fp, tp->ctt_name);
vbytes = LCTF_VBYTES (fp, kind, size, vlen);
switch (kind)
{
case CTF_K_INTEGER:
case CTF_K_FLOAT:
/* Names are reused by bit-fields, which are differentiated by their
encodings, and so typically we'd record only the first instance of
a given intrinsic. However, we replace an existing type with a
root-visible version so that we can be sure to find it when
checking for conflicting definitions in ctf_add_type(). */
if (((ctf_hash_lookup_type (fp->ctf_names, fp, name)) == 0)
|| (flag & CTF_ADD_ROOT))
{
err = ctf_hash_define_type (fp->ctf_names, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
}
break;
/* These kinds have no name, so do not need interning into any
hashtables. */
case CTF_K_ARRAY:
case CTF_K_SLICE:
break;
case CTF_K_FUNCTION:
err = ctf_hash_insert_type (fp->ctf_names, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
break;
case CTF_K_STRUCT:
err = ctf_hash_define_type (fp->ctf_structs, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
if (size >= CTF_LSTRUCT_THRESH)
nlstructs++;
break;
case CTF_K_UNION:
err = ctf_hash_define_type (fp->ctf_unions, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
if (size >= CTF_LSTRUCT_THRESH)
nlunions++;
break;
case CTF_K_ENUM:
err = ctf_hash_define_type (fp->ctf_enums, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
break;
case CTF_K_TYPEDEF:
err = ctf_hash_insert_type (fp->ctf_names, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
break;
case CTF_K_FORWARD:
/* Only insert forward tags into the given hash if the type or tag
name is not already present. */
switch (tp->ctt_type)
{
case CTF_K_STRUCT:
hp = fp->ctf_structs;
break;
case CTF_K_UNION:
hp = fp->ctf_unions;
break;
case CTF_K_ENUM:
hp = fp->ctf_enums;
break;
default:
hp = fp->ctf_structs;
}
if (ctf_hash_lookup_type (hp, fp, name) == 0)
{
err = ctf_hash_insert_type (hp, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
}
break;
case CTF_K_POINTER:
/* If the type referenced by the pointer is in this CTF container,
then store the index of the pointer type in
fp->ctf_ptrtab[ index of referenced type ]. */
if (LCTF_TYPE_ISCHILD (fp, tp->ctt_type) == child
&& LCTF_TYPE_TO_INDEX (fp, tp->ctt_type) <= fp->ctf_typemax)
fp->ctf_ptrtab[LCTF_TYPE_TO_INDEX (fp, tp->ctt_type)] = id;
/*FALLTHRU*/
case CTF_K_VOLATILE:
case CTF_K_CONST:
case CTF_K_RESTRICT:
err = ctf_hash_insert_type (fp->ctf_names, fp,
LCTF_INDEX_TO_TYPE (fp, id, child),
tp->ctt_name);
if (err != 0 && err != ECTF_STRTAB)
return err;
break;
default:
ctf_dprintf ("unhandled CTF kind in endianness conversion -- %x\n",
kind);
return ECTF_CORRUPT;
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
}
*xp = (uint32_t) ((uintptr_t) tp - (uintptr_t) fp->ctf_buf);
tp = (ctf_type_t *) ((uintptr_t) tp + increment + vbytes);
}
ctf_dprintf ("%lu total types processed\n", fp->ctf_typemax);
ctf_dprintf ("%u enum names hashed\n", ctf_hash_size (fp->ctf_enums));
ctf_dprintf ("%u struct names hashed (%d long)\n",
ctf_hash_size (fp->ctf_structs), nlstructs);
ctf_dprintf ("%u union names hashed (%d long)\n",
ctf_hash_size (fp->ctf_unions), nlunions);
ctf_dprintf ("%u base type names hashed\n", ctf_hash_size (fp->ctf_names));
/* Make an additional pass through the pointer table to find pointers that
point to anonymous typedef nodes. If we find one, modify the pointer table
so that the pointer is also known to point to the node that is referenced
by the anonymous typedef node. */
for (id = 1; id <= fp->ctf_typemax; id++)
{
if ((dst = fp->ctf_ptrtab[id]) != 0)
{
tp = LCTF_INDEX_TO_TYPEPTR (fp, id);
if (LCTF_INFO_KIND (fp, tp->ctt_info) == CTF_K_TYPEDEF &&
strcmp (ctf_strptr (fp, tp->ctt_name), "") == 0 &&
LCTF_TYPE_ISCHILD (fp, tp->ctt_type) == child &&
LCTF_TYPE_TO_INDEX (fp, tp->ctt_type) <= fp->ctf_typemax)
fp->ctf_ptrtab[LCTF_TYPE_TO_INDEX (fp, tp->ctt_type)] = dst;
}
}
return 0;
}
/* Endianness-flipping routines.
We flip everything, mindlessly, even 1-byte entities, so that future
expansions do not require changes to this code. */
/* < C11? define away static assertions. */
#if !defined (__STDC_VERSION__) || __STDC_VERSION__ < 201112L
#define _Static_assert(cond, err)
#endif
/* Swap the endianness of something. */
#define swap_thing(x) \
do { \
_Static_assert (sizeof (x) == 1 || (sizeof (x) % 2 == 0 \
&& sizeof (x) <= 8), \
"Invalid size, update endianness code"); \
switch (sizeof (x)) { \
case 2: x = bswap_16 (x); break; \
case 4: x = bswap_32 (x); break; \
case 8: x = bswap_64 (x); break; \
case 1: /* Nothing needs doing */ \
break; \
} \
} while (0);
/* Flip the endianness of the CTF header. */
static void
flip_header (ctf_header_t *cth)
{
swap_thing (cth->cth_preamble.ctp_magic);
swap_thing (cth->cth_preamble.ctp_version);
swap_thing (cth->cth_preamble.ctp_flags);
swap_thing (cth->cth_parlabel);
swap_thing (cth->cth_parname);
swap_thing (cth->cth_objtoff);
swap_thing (cth->cth_funcoff);
swap_thing (cth->cth_varoff);
swap_thing (cth->cth_typeoff);
swap_thing (cth->cth_stroff);
swap_thing (cth->cth_strlen);
}
/* Flip the endianness of the label section, an array of ctf_lblent_t. */
static void
flip_lbls (void *start, size_t len)
{
ctf_lblent_t *lbl = start;
for (ssize_t i = len / sizeof (struct ctf_lblent); i > 0; lbl++, i--)
{
swap_thing (lbl->ctl_label);
swap_thing (lbl->ctl_type);
}
}
/* Flip the endianness of the data-object or function sections, an array of
uint32_t. (The function section has more internal structure, but that
structure is an array of uint32_t, so can be treated as one big array for
byte-swapping.) */
static void
flip_objts (void *start, size_t len)
{
uint32_t *obj = start;
for (ssize_t i = len / sizeof (uint32_t); i > 0; obj++, i--)
swap_thing (*obj);
}
/* Flip the endianness of the variable section, an array of ctf_varent_t. */
static void
flip_vars (void *start, size_t len)
{
ctf_varent_t *var = start;
for (ssize_t i = len / sizeof (struct ctf_varent); i > 0; var++, i--)
{
swap_thing (var->ctv_name);
swap_thing (var->ctv_type);
}
}
/* Flip the endianness of the type section, a tagged array of ctf_type or
ctf_stype followed by variable data. */
static int
flip_types (void *start, size_t len)
{
ctf_type_t *t = start;
while ((uintptr_t) t < ((uintptr_t) start) + len)
{
swap_thing (t->ctt_name);
swap_thing (t->ctt_info);
swap_thing (t->ctt_size);
uint32_t kind = CTF_V2_INFO_KIND (t->ctt_info);
size_t size = t->ctt_size;
uint32_t vlen = CTF_V2_INFO_VLEN (t->ctt_info);
size_t vbytes = get_vbytes_v2 (kind, size, vlen);
if (_libctf_unlikely_ (size == CTF_LSIZE_SENT))
{
swap_thing (t->ctt_lsizehi);
swap_thing (t->ctt_lsizelo);
size = CTF_TYPE_LSIZE (t);
t = (ctf_type_t *) ((uintptr_t) t + sizeof (ctf_type_t));
}
else
t = (ctf_type_t *) ((uintptr_t) t + sizeof (ctf_stype_t));
switch (kind)
{
case CTF_K_FORWARD:
case CTF_K_UNKNOWN:
case CTF_K_POINTER:
case CTF_K_TYPEDEF:
case CTF_K_VOLATILE:
case CTF_K_CONST:
case CTF_K_RESTRICT:
/* These types have no vlen data to swap. */
assert (vbytes == 0);
break;
case CTF_K_INTEGER:
case CTF_K_FLOAT:
{
/* These types have a single uint32_t. */
uint32_t *item = (uint32_t *) t;
swap_thing (*item);
break;
}
case CTF_K_FUNCTION:
{
/* This type has a bunch of uint32_ts. */
uint32_t *item = (uint32_t *) t;
for (ssize_t i = vlen; i > 0; item++, i--)
swap_thing (*item);
break;
}
case CTF_K_ARRAY:
{
/* This has a single ctf_array_t. */
ctf_array_t *a = (ctf_array_t *) t;
assert (vbytes == sizeof (ctf_array_t));
swap_thing (a->cta_contents);
swap_thing (a->cta_index);
swap_thing (a->cta_nelems);
break;
}
case CTF_K_SLICE:
{
/* This has a single ctf_slice_t. */
ctf_slice_t *s = (ctf_slice_t *) t;
assert (vbytes == sizeof (ctf_slice_t));
swap_thing (s->cts_type);
swap_thing (s->cts_offset);
swap_thing (s->cts_bits);
break;
}
case CTF_K_STRUCT:
case CTF_K_UNION:
{
/* This has an array of ctf_member or ctf_lmember, depending on
size. We could consider it to be a simple array of uint32_t,
but for safety's sake in case these structures ever acquire
non-uint32_t members, do it member by member. */
if (_libctf_unlikely_ (size >= CTF_LSTRUCT_THRESH))
{
ctf_lmember_t *lm = (ctf_lmember_t *) t;
for (ssize_t i = vlen; i > 0; i--, lm++)
{
swap_thing (lm->ctlm_name);
swap_thing (lm->ctlm_offsethi);
swap_thing (lm->ctlm_type);
swap_thing (lm->ctlm_offsetlo);
}
}
else
{
ctf_member_t *m = (ctf_member_t *) t;
for (ssize_t i = vlen; i > 0; i--, m++)
{
swap_thing (m->ctm_name);
swap_thing (m->ctm_offset);
swap_thing (m->ctm_type);
}
}
break;
}
case CTF_K_ENUM:
{
/* This has an array of ctf_enum_t. */
ctf_enum_t *item = (ctf_enum_t *) t;
for (ssize_t i = vlen; i > 0; item++, i--)
{
swap_thing (item->cte_name);
swap_thing (item->cte_value);
}
break;
}
default:
ctf_dprintf ("unhandled CTF kind in endianness conversion -- %x\n",
kind);
return ECTF_CORRUPT;
}
t = (ctf_type_t *) ((uintptr_t) t + vbytes);
}
return 0;
}
/* Flip the endianness of BASE, given the offsets in the (already endian-
converted) CTH.
All of this stuff happens before the header is fully initialized, so the
LCTF_*() macros cannot be used yet. Since we do not try to endian-convert v1
data, this is no real loss. */
static int
flip_ctf (ctf_header_t *cth, unsigned char *base)
{
base += sizeof (ctf_header_t);
flip_lbls (base + cth->cth_lbloff, cth->cth_objtoff - cth->cth_lbloff);
flip_objts (base + cth->cth_objtoff, cth->cth_funcoff - cth->cth_objtoff);
flip_objts (base + cth->cth_funcoff, cth->cth_varoff - cth->cth_funcoff);
flip_vars (base + cth->cth_varoff, cth->cth_typeoff - cth->cth_varoff);
return flip_types (base + cth->cth_typeoff, cth->cth_stroff - cth->cth_typeoff);
}
/* Open a CTF file, mocking up a suitable ctf_sect. */
ctf_file_t *ctf_simple_open (const char *ctfsect, size_t ctfsect_size,
const char *symsect, size_t symsect_size,
size_t symsect_entsize,
const char *strsect, size_t strsect_size,
int *errp)
{
ctf_sect_t skeleton;
ctf_sect_t ctf_sect, sym_sect, str_sect;
ctf_sect_t *ctfsectp = NULL;
ctf_sect_t *symsectp = NULL;
ctf_sect_t *strsectp = NULL;
skeleton.cts_name = _CTF_SECTION;
skeleton.cts_entsize = 1;
if (ctfsect)
{
memcpy (&ctf_sect, &skeleton, sizeof (struct ctf_sect));
ctf_sect.cts_data = ctfsect;
ctf_sect.cts_size = ctfsect_size;
ctfsectp = &ctf_sect;
}
if (symsect)
{
memcpy (&sym_sect, &skeleton, sizeof (struct ctf_sect));
sym_sect.cts_data = symsect;
sym_sect.cts_size = symsect_size;
sym_sect.cts_entsize = symsect_entsize;
symsectp = &sym_sect;
}
if (strsect)
{
memcpy (&str_sect, &skeleton, sizeof (struct ctf_sect));
str_sect.cts_data = strsect;
str_sect.cts_size = strsect_size;
strsectp = &str_sect;
}
return ctf_bufopen (ctfsectp, symsectp, strsectp, errp);
}
/* Decode the specified CTF buffer and optional symbol table, and create a new
CTF container representing the symbolic debugging information. This code can
be used directly by the debugger, or it can be used as the engine for
ctf_fdopen() or ctf_open(), below. */
ctf_file_t *
ctf_bufopen (const ctf_sect_t *ctfsect, const ctf_sect_t *symsect,
const ctf_sect_t *strsect, int *errp)
{
const ctf_preamble_t *pp;
ctf_header_t hp;
ctf_file_t *fp;
libctf: endianness fixes Testing of the first code to generate CTF_K_SLICEs on big-endian revealed a bunch of new problems in this area. Most importantly, the trick we did earlier to avoid wasting two bytes on padding in the ctf_slice_t is best avoided: because it leads to the whole file after that point no longer being naturally aligned, all multibyte accesses from then on must use memmove() to avoid unaligned access on platforms where that is fatal. In future, this is planned, but for now we are still doing direct access in many places, so we must revert to making ctf_slice_t properly aligned for storage in an array. Rather than wasting bytes on padding, we boost the size of cts_offset and cts_bits. This is still a waste of space (we cannot have offsets or bits in bitfields > 256) but it cannot be avoided for now, and slices are not so common that this will be a serious problem. A possibly-worse endianness problem fixed at the same time involves a codepath used only for foreign-endian, uncompressed CTF files, where we were not copying the actual CTF data into the buffer, leading to libctf reading only zeroes (or, possibly, uninitialized garbage). Finally, when we read in a CTF file, we copy the header and work from the copy. We were flipping the endianness of the header copy, and of the body of the file buffer, but not of the header in the file buffer itself: so if we write the file back out again we end up with an unreadable frankenfile with header and body of different endiannesses. Fix by flipping both copies of the header. include/ * ctf.h (ctf_slice_t): Make cts_offset and cts_bits unsigned short, so following structures are properly aligned. libctf/ * ctf-open.c (get_vbytes_common): Return the new slice size. (ctf_bufopen): Flip the endianness of the CTF-section header copy. Remember to copy in the CTF data when opening an uncompressed foreign-endian CTF file. Prune useless variable manipulation.
2019-06-19 13:34:56 +02:00
void *base;
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
size_t size, hdrsz;
int foreign_endian = 0;
int err;
libctf_init_debug();
if (ctfsect == NULL || ((symsect == NULL) != (strsect == NULL)))
return (ctf_set_open_errno (errp, EINVAL));
if (symsect != NULL && symsect->cts_entsize != sizeof (Elf32_Sym) &&
symsect->cts_entsize != sizeof (Elf64_Sym))
return (ctf_set_open_errno (errp, ECTF_SYMTAB));
if (symsect != NULL && symsect->cts_data == NULL)
return (ctf_set_open_errno (errp, ECTF_SYMBAD));
if (strsect != NULL && strsect->cts_data == NULL)
return (ctf_set_open_errno (errp, ECTF_STRBAD));
if (ctfsect->cts_size < sizeof (ctf_preamble_t))
return (ctf_set_open_errno (errp, ECTF_NOCTFBUF));
pp = (const ctf_preamble_t *) ctfsect->cts_data;
ctf_dprintf ("ctf_bufopen: magic=0x%x version=%u\n",
pp->ctp_magic, pp->ctp_version);
/* Validate each part of the CTF header.
First, we validate the preamble (common to all versions). At that point,
we know the endianness and specific header version, and can validate the
version-specific parts including section offsets and alignments.
We specifically do not support foreign-endian old versions. */
if (_libctf_unlikely_ (pp->ctp_magic != CTF_MAGIC))
{
if (pp->ctp_magic == bswap_16 (CTF_MAGIC))
{
if (pp->ctp_version != CTF_VERSION_3)
return (ctf_set_open_errno (errp, ECTF_CTFVERS));
foreign_endian = 1;
}
else
return (ctf_set_open_errno (errp, ECTF_NOCTFBUF));
}
if (_libctf_unlikely_ ((pp->ctp_version < CTF_VERSION_1)
|| (pp->ctp_version > CTF_VERSION_3)))
return (ctf_set_open_errno (errp, ECTF_CTFVERS));
if ((symsect != NULL) && (pp->ctp_version < CTF_VERSION_2))
{
/* The symtab can contain function entries which contain embedded ctf
info. We do not support dynamically upgrading such entries (none
should exist in any case, since dwarf2ctf does not create them). */
ctf_dprintf ("ctf_bufopen: CTF version %d symsect not "
"supported\n", pp->ctp_version);
return (ctf_set_open_errno (errp, ECTF_NOTSUP));
}
if (ctfsect->cts_size < sizeof (ctf_header_t))
return (ctf_set_open_errno (errp, ECTF_NOCTFBUF));
memcpy (&hp, ctfsect->cts_data, sizeof (hp));
if (foreign_endian)
flip_header (&hp);
ctf_dprintf ("header offsets: %x/%x/%x/%x/%x/%x/%x\n",
hp.cth_lbloff, hp.cth_objtoff, hp.cth_funcoff, hp.cth_varoff,
hp.cth_typeoff, hp.cth_stroff, hp.cth_strlen);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
hdrsz = sizeof (ctf_header_t);
size = hp.cth_stroff + hp.cth_strlen;
ctf_dprintf ("ctf_bufopen: uncompressed size=%lu\n", (unsigned long) size);
if (hp.cth_lbloff > size || hp.cth_objtoff > size
|| hp.cth_funcoff > size || hp.cth_typeoff > size || hp.cth_stroff > size)
return (ctf_set_open_errno (errp, ECTF_CORRUPT));
if (hp.cth_lbloff > hp.cth_objtoff
|| hp.cth_objtoff > hp.cth_funcoff
|| hp.cth_funcoff > hp.cth_typeoff
|| hp.cth_funcoff > hp.cth_varoff
|| hp.cth_varoff > hp.cth_typeoff || hp.cth_typeoff > hp.cth_stroff)
return (ctf_set_open_errno (errp, ECTF_CORRUPT));
if ((hp.cth_lbloff & 3) || (hp.cth_objtoff & 1)
|| (hp.cth_funcoff & 1) || (hp.cth_varoff & 3) || (hp.cth_typeoff & 3))
return (ctf_set_open_errno (errp, ECTF_CORRUPT));
/* Once everything is determined to be valid, attempt to decompress the CTF
data buffer if it is compressed, or copy it into new storage if it is not
compressed but needs endian-flipping. Otherwise we just put the data
section's buffer pointer into ctf_buf, below. */
/* Note: if this is a v1 buffer, it will be reallocated and expanded by
init_types(). */
if (hp.cth_flags & CTF_F_COMPRESS)
{
libctf: fix a number of build problems found on Solaris and NetBSD - Use of nonportable <endian.h> - Use of qsort_r - Use of zlib without appropriate magic to pull in the binutils zlib - Use of off64_t without checking (fixed by dropping the unused fields that need off64_t entirely) - signedness problems due to long being too short a type on 32-bit platforms: ctf_id_t is now 'unsigned long', and CTF_ERR must be used only for functions that return ctf_id_t - One lingering use of bzero() and of <sys/errno.h> All fixed, using code from gnulib where possible. Relatedly, set cts_size in a couple of places it was missed (string table and symbol table loading upon ctf_bfdopen()). binutils/ * objdump.c (make_ctfsect): Drop cts_type, cts_flags, and cts_offset. * readelf.c (shdr_to_ctf_sect): Likewise. include/ * ctf-api.h (ctf_sect_t): Drop cts_type, cts_flags, and cts_offset. (ctf_id_t): This is now an unsigned type. (CTF_ERR): Cast it to ctf_id_t. Note that it should only be used for ctf_id_t-returning functions. libctf/ * Makefile.am (ZLIB): New. (ZLIBINC): Likewise. (AM_CFLAGS): Use them. (libctf_a_LIBADD): New, for LIBOBJS. * configure.ac: Check for zlib, endian.h, and qsort_r. * ctf-endian.h: New, providing htole64 and le64toh. * swap.h: Code style fixes. (bswap_identity_64): New. * qsort_r.c: New, from gnulib (with one added #include). * ctf-decls.h: New, providing a conditional qsort_r declaration, and unconditional definitions of MIN and MAX. * ctf-impl.h: Use it. Do not use <sys/errno.h>. (ctf_set_errno): Now returns unsigned long. * ctf-util.c (ctf_set_errno): Adjust here too. * ctf-archive.c: Use ctf-endian.h. (ctf_arc_open_by_offset): Use memset, not bzero. Drop cts_type, cts_flags and cts_offset. (ctf_arc_write): Drop debugging dependent on the size of off_t. * ctf-create.c: Provide a definition of roundup if not defined. (ctf_create): Drop cts_type, cts_flags and cts_offset. (ctf_add_reftype): Do not check if type IDs are below zero. (ctf_add_slice): Likewise. (ctf_add_typedef): Likewise. (ctf_add_member_offset): Cast error-returning ssize_t's to size_t when known error-free. Drop CTF_ERR usage for functions returning int. (ctf_add_member_encoded): Drop CTF_ERR usage for functions returning int. (ctf_add_variable): Likewise. (enumcmp): Likewise. (enumadd): Likewise. (membcmp): Likewise. (ctf_add_type): Likewise. Cast error-returning ssize_t's to size_t when known error-free. * ctf-dump.c (ctf_is_slice): Drop CTF_ERR usage for functions returning int: use CTF_ERR for functions returning ctf_type_id. (ctf_dump_label): Likewise. (ctf_dump_objts): Likewise. * ctf-labels.c (ctf_label_topmost): Likewise. (ctf_label_iter): Likewise. (ctf_label_info): Likewise. * ctf-lookup.c (ctf_func_args): Likewise. * ctf-open.c (upgrade_types): Cast to size_t where appropriate. (ctf_bufopen): Likewise. Use zlib types as needed. * ctf-types.c (ctf_member_iter): Drop CTF_ERR usage for functions returning int. (ctf_enum_iter): Likewise. (ctf_type_size): Likewise. (ctf_type_align): Likewise. Cast to size_t where appropriate. (ctf_type_kind_unsliced): Likewise. (ctf_type_kind): Likewise. (ctf_type_encoding): Likewise. (ctf_member_info): Likewise. (ctf_array_info): Likewise. (ctf_enum_value): Likewise. (ctf_type_rvisit): Likewise. * ctf-open-bfd.c (ctf_bfdopen): Drop cts_type, cts_flags and cts_offset. (ctf_simple_open): Likewise. (ctf_bfdopen_ctfsect): Likewise. Set cts_size properly. * Makefile.in: Regenerate. * aclocal.m4: Likewise. * config.h: Likewise. * configure: Likewise.
2019-05-31 11:10:51 +02:00
size_t srclen;
uLongf dstlen;
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
const void *src;
int rc = Z_OK;
libctf: endianness fixes Testing of the first code to generate CTF_K_SLICEs on big-endian revealed a bunch of new problems in this area. Most importantly, the trick we did earlier to avoid wasting two bytes on padding in the ctf_slice_t is best avoided: because it leads to the whole file after that point no longer being naturally aligned, all multibyte accesses from then on must use memmove() to avoid unaligned access on platforms where that is fatal. In future, this is planned, but for now we are still doing direct access in many places, so we must revert to making ctf_slice_t properly aligned for storage in an array. Rather than wasting bytes on padding, we boost the size of cts_offset and cts_bits. This is still a waste of space (we cannot have offsets or bits in bitfields > 256) but it cannot be avoided for now, and slices are not so common that this will be a serious problem. A possibly-worse endianness problem fixed at the same time involves a codepath used only for foreign-endian, uncompressed CTF files, where we were not copying the actual CTF data into the buffer, leading to libctf reading only zeroes (or, possibly, uninitialized garbage). Finally, when we read in a CTF file, we copy the header and work from the copy. We were flipping the endianness of the header copy, and of the body of the file buffer, but not of the header in the file buffer itself: so if we write the file back out again we end up with an unreadable frankenfile with header and body of different endiannesses. Fix by flipping both copies of the header. include/ * ctf.h (ctf_slice_t): Make cts_offset and cts_bits unsigned short, so following structures are properly aligned. libctf/ * ctf-open.c (get_vbytes_common): Return the new slice size. (ctf_bufopen): Flip the endianness of the CTF-section header copy. Remember to copy in the CTF data when opening an uncompressed foreign-endian CTF file. Prune useless variable manipulation.
2019-06-19 13:34:56 +02:00
void *buf;
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
if ((base = ctf_alloc (size + hdrsz)) == NULL)
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
return (ctf_set_open_errno (errp, ECTF_ZALLOC));
memcpy (base, ctfsect->cts_data, hdrsz);
((ctf_preamble_t *) base)->ctp_flags &= ~CTF_F_COMPRESS;
buf = (unsigned char *) base + hdrsz;
src = (unsigned char *) ctfsect->cts_data + hdrsz;
srclen = ctfsect->cts_size - hdrsz;
dstlen = size;
if ((rc = uncompress (buf, &dstlen, src, srclen)) != Z_OK)
{
ctf_dprintf ("zlib inflate err: %s\n", zError (rc));
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
free (base);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
return (ctf_set_open_errno (errp, ECTF_DECOMPRESS));
}
libctf: fix a number of build problems found on Solaris and NetBSD - Use of nonportable <endian.h> - Use of qsort_r - Use of zlib without appropriate magic to pull in the binutils zlib - Use of off64_t without checking (fixed by dropping the unused fields that need off64_t entirely) - signedness problems due to long being too short a type on 32-bit platforms: ctf_id_t is now 'unsigned long', and CTF_ERR must be used only for functions that return ctf_id_t - One lingering use of bzero() and of <sys/errno.h> All fixed, using code from gnulib where possible. Relatedly, set cts_size in a couple of places it was missed (string table and symbol table loading upon ctf_bfdopen()). binutils/ * objdump.c (make_ctfsect): Drop cts_type, cts_flags, and cts_offset. * readelf.c (shdr_to_ctf_sect): Likewise. include/ * ctf-api.h (ctf_sect_t): Drop cts_type, cts_flags, and cts_offset. (ctf_id_t): This is now an unsigned type. (CTF_ERR): Cast it to ctf_id_t. Note that it should only be used for ctf_id_t-returning functions. libctf/ * Makefile.am (ZLIB): New. (ZLIBINC): Likewise. (AM_CFLAGS): Use them. (libctf_a_LIBADD): New, for LIBOBJS. * configure.ac: Check for zlib, endian.h, and qsort_r. * ctf-endian.h: New, providing htole64 and le64toh. * swap.h: Code style fixes. (bswap_identity_64): New. * qsort_r.c: New, from gnulib (with one added #include). * ctf-decls.h: New, providing a conditional qsort_r declaration, and unconditional definitions of MIN and MAX. * ctf-impl.h: Use it. Do not use <sys/errno.h>. (ctf_set_errno): Now returns unsigned long. * ctf-util.c (ctf_set_errno): Adjust here too. * ctf-archive.c: Use ctf-endian.h. (ctf_arc_open_by_offset): Use memset, not bzero. Drop cts_type, cts_flags and cts_offset. (ctf_arc_write): Drop debugging dependent on the size of off_t. * ctf-create.c: Provide a definition of roundup if not defined. (ctf_create): Drop cts_type, cts_flags and cts_offset. (ctf_add_reftype): Do not check if type IDs are below zero. (ctf_add_slice): Likewise. (ctf_add_typedef): Likewise. (ctf_add_member_offset): Cast error-returning ssize_t's to size_t when known error-free. Drop CTF_ERR usage for functions returning int. (ctf_add_member_encoded): Drop CTF_ERR usage for functions returning int. (ctf_add_variable): Likewise. (enumcmp): Likewise. (enumadd): Likewise. (membcmp): Likewise. (ctf_add_type): Likewise. Cast error-returning ssize_t's to size_t when known error-free. * ctf-dump.c (ctf_is_slice): Drop CTF_ERR usage for functions returning int: use CTF_ERR for functions returning ctf_type_id. (ctf_dump_label): Likewise. (ctf_dump_objts): Likewise. * ctf-labels.c (ctf_label_topmost): Likewise. (ctf_label_iter): Likewise. (ctf_label_info): Likewise. * ctf-lookup.c (ctf_func_args): Likewise. * ctf-open.c (upgrade_types): Cast to size_t where appropriate. (ctf_bufopen): Likewise. Use zlib types as needed. * ctf-types.c (ctf_member_iter): Drop CTF_ERR usage for functions returning int. (ctf_enum_iter): Likewise. (ctf_type_size): Likewise. (ctf_type_align): Likewise. Cast to size_t where appropriate. (ctf_type_kind_unsliced): Likewise. (ctf_type_kind): Likewise. (ctf_type_encoding): Likewise. (ctf_member_info): Likewise. (ctf_array_info): Likewise. (ctf_enum_value): Likewise. (ctf_type_rvisit): Likewise. * ctf-open-bfd.c (ctf_bfdopen): Drop cts_type, cts_flags and cts_offset. (ctf_simple_open): Likewise. (ctf_bfdopen_ctfsect): Likewise. Set cts_size properly. * Makefile.in: Regenerate. * aclocal.m4: Likewise. * config.h: Likewise. * configure: Likewise.
2019-05-31 11:10:51 +02:00
if ((size_t) dstlen != size)
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
{
ctf_dprintf ("zlib inflate short -- got %lu of %lu "
"bytes\n", (unsigned long) dstlen, (unsigned long) size);
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
free (base);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
return (ctf_set_open_errno (errp, ECTF_CORRUPT));
}
}
else if (foreign_endian)
{
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
if ((base = ctf_alloc (size + hdrsz)) == NULL)
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
return (ctf_set_open_errno (errp, ECTF_ZALLOC));
libctf: endianness fixes Testing of the first code to generate CTF_K_SLICEs on big-endian revealed a bunch of new problems in this area. Most importantly, the trick we did earlier to avoid wasting two bytes on padding in the ctf_slice_t is best avoided: because it leads to the whole file after that point no longer being naturally aligned, all multibyte accesses from then on must use memmove() to avoid unaligned access on platforms where that is fatal. In future, this is planned, but for now we are still doing direct access in many places, so we must revert to making ctf_slice_t properly aligned for storage in an array. Rather than wasting bytes on padding, we boost the size of cts_offset and cts_bits. This is still a waste of space (we cannot have offsets or bits in bitfields > 256) but it cannot be avoided for now, and slices are not so common that this will be a serious problem. A possibly-worse endianness problem fixed at the same time involves a codepath used only for foreign-endian, uncompressed CTF files, where we were not copying the actual CTF data into the buffer, leading to libctf reading only zeroes (or, possibly, uninitialized garbage). Finally, when we read in a CTF file, we copy the header and work from the copy. We were flipping the endianness of the header copy, and of the body of the file buffer, but not of the header in the file buffer itself: so if we write the file back out again we end up with an unreadable frankenfile with header and body of different endiannesses. Fix by flipping both copies of the header. include/ * ctf.h (ctf_slice_t): Make cts_offset and cts_bits unsigned short, so following structures are properly aligned. libctf/ * ctf-open.c (get_vbytes_common): Return the new slice size. (ctf_bufopen): Flip the endianness of the CTF-section header copy. Remember to copy in the CTF data when opening an uncompressed foreign-endian CTF file. Prune useless variable manipulation.
2019-06-19 13:34:56 +02:00
memcpy (base, ctfsect->cts_data, size + hdrsz);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
}
else
libctf: endianness fixes Testing of the first code to generate CTF_K_SLICEs on big-endian revealed a bunch of new problems in this area. Most importantly, the trick we did earlier to avoid wasting two bytes on padding in the ctf_slice_t is best avoided: because it leads to the whole file after that point no longer being naturally aligned, all multibyte accesses from then on must use memmove() to avoid unaligned access on platforms where that is fatal. In future, this is planned, but for now we are still doing direct access in many places, so we must revert to making ctf_slice_t properly aligned for storage in an array. Rather than wasting bytes on padding, we boost the size of cts_offset and cts_bits. This is still a waste of space (we cannot have offsets or bits in bitfields > 256) but it cannot be avoided for now, and slices are not so common that this will be a serious problem. A possibly-worse endianness problem fixed at the same time involves a codepath used only for foreign-endian, uncompressed CTF files, where we were not copying the actual CTF data into the buffer, leading to libctf reading only zeroes (or, possibly, uninitialized garbage). Finally, when we read in a CTF file, we copy the header and work from the copy. We were flipping the endianness of the header copy, and of the body of the file buffer, but not of the header in the file buffer itself: so if we write the file back out again we end up with an unreadable frankenfile with header and body of different endiannesses. Fix by flipping both copies of the header. include/ * ctf.h (ctf_slice_t): Make cts_offset and cts_bits unsigned short, so following structures are properly aligned. libctf/ * ctf-open.c (get_vbytes_common): Return the new slice size. (ctf_bufopen): Flip the endianness of the CTF-section header copy. Remember to copy in the CTF data when opening an uncompressed foreign-endian CTF file. Prune useless variable manipulation.
2019-06-19 13:34:56 +02:00
base = (void *) ctfsect->cts_data;
/* Flip the endianness of the copy of the header in the section, to avoid
ending up with a partially-endian-flipped file. */
if (foreign_endian)
flip_header ((ctf_header_t *) base);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
/* Once we have uncompressed and validated the CTF data buffer, we can
proceed with allocating a ctf_file_t and initializing it.
Nothing that depends on buf or base should be set directly in this function
before the init_types() call, because it may be reallocated during
transparent upgrade if this recension of libctf is so configured: see
ctf_set_base() and ctf_realloc_base(). */
if ((fp = ctf_alloc (sizeof (ctf_file_t))) == NULL)
return (ctf_set_open_errno (errp, ENOMEM));
memset (fp, 0, sizeof (ctf_file_t));
ctf_set_version (fp, &hp, hp.cth_version);
libctf: deduplicate and sort the string table ctf.h states: > [...] the CTF string table does not contain any duplicated strings. Unfortunately this is entirely untrue: libctf has before now made no attempt whatsoever to deduplicate the string table. It computes the string table's length on the fly as it adds new strings to the dynamic CTF file, and ctf_update() just writes each string to the table and notes the current write position as it traverses the dynamic CTF file's data structures and builds the final CTF buffer. There is no global view of the strings and no deduplication. Fix this by erasing the ctf_dtvstrlen dead-reckoning length, and adding a new dynhash table ctf_str_atoms that maps unique strings to a list of references to those strings: a reference is a simple uint32_t * to some value somewhere in the under-construction CTF buffer that needs updating to note the string offset when the strtab is laid out. Adding a string is now a simple matter of calling ctf_str_add_ref(), which adds a new atom to the atoms table, if one doesn't already exist, and adding the location of the reference to this atom to the refs list attached to the atom: this works reliably as long as one takes care to only call ctf_str_add_ref() once the final location of the offset is known (so you can't call it on a temporary structure and then memcpy() that structure into place in the CTF buffer, because the ref will still point to the old location: ctf_update() changes accordingly). Generating the CTF string table is a matter of calling ctf_str_write_strtab(), which counts the length and number of elements in the atoms table using the ctf_dynhash_iter() function we just added, populating an array of pointers into the atoms table and sorting it into order (to help compressors), then traversing this table and emitting it, updating the refs to each atom as we go. The only complexity here is arranging to keep the null string at offset zero, since a lot of code in libctf depends on being able to leave strtab references at 0 to indicate 'no name'. Once the table is constructed and the refs updated, we know how long it is, so we can realloc() the partial CTF buffer we allocated earlier and can copy the table on to the end of it (and purge the refs because they're not needed any more and have been invalidated by the realloc() call in any case). The net effect of all this is a reduction in uncompressed strtab sizes of about 30% (perhaps a quarter to a half of all strings across the Linux kernel are eliminated as duplicates). Of course, duplicated strings are highly redundant, so the space saving after compression is only about 20%: when the other non-strtab sections are factored in, CTF sizes shrink by about 10%. No change in externally-visible API or file format (other than the reduction in pointless redundancy). libctf/ * ctf-impl.h: (struct ctf_strs_writable): New, non-const version of struct ctf_strs. (struct ctf_dtdef): Note that dtd_data.ctt_name is unpopulated. (struct ctf_str_atom): New, disambiguated single string. (struct ctf_str_atom_ref): New, points to some other location that references this string's offset. (struct ctf_file): New members ctf_str_atoms and ctf_str_num_refs. Remove member ctf_dtvstrlen: we no longer track the total strlen as we add strings. (ctf_str_create_atoms): Declare new function in ctf-string.c. (ctf_str_free_atoms): Likewise. (ctf_str_add): Likewise. (ctf_str_add_ref): Likewise. (ctf_str_purge_refs): Likewise. (ctf_str_write_strtab): Likewise. (ctf_realloc): Declare new function in ctf-util.c. * ctf-open.c (ctf_bufopen): Create the atoms table. (ctf_file_close): Destroy it. * ctf-create.c (ctf_update): Copy-and-free it on update. No longer special-case the position of the parname string. Construct the strtab by calling ctf_str_add_ref and ctf_str_write_strtab after the rest of each buffer element is constructed, not via open-coding: realloc the CTF buffer and append the strtab to it. No longer maintain ctf_dtvstrlen. Sort the variable entry table later, after strtab construction. (ctf_copy_membnames): Remove: integrated into ctf_copy_{s,l,e}members. (ctf_copy_smembers): Drop the string offset: call ctf_str_add_ref after buffer element construction instead. (ctf_copy_lmembers): Likewise. (ctf_copy_emembers): Likewise. (ctf_create): No longer maintain the ctf_dtvstrlen. (ctf_dtd_delete): Likewise. (ctf_dvd_delete): Likewise. (ctf_add_generic): Likewise. (ctf_add_enumerator): Likewise. (ctf_add_member_offset): Likewise. (ctf_add_variable): Likewise. (membadd): Likewise. * ctf-util.c (ctf_realloc): New, wrapper around realloc that aborts if there are active ctf_str_num_refs. (ctf_strraw): Move to ctf-string.c. (ctf_strptr): Likewise. * ctf-string.c: New file, strtab manipulation. * Makefile.am (libctf_a_SOURCES): Add it. * Makefile.in: Regenerate.
2019-06-27 14:51:10 +02:00
ctf_str_create_atoms (fp);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
if (_libctf_unlikely_ (hp.cth_version < CTF_VERSION_2))
fp->ctf_parmax = CTF_MAX_PTYPE_V1;
else
fp->ctf_parmax = CTF_MAX_PTYPE;
memcpy (&fp->ctf_data, ctfsect, sizeof (ctf_sect_t));
if (symsect != NULL)
{
memcpy (&fp->ctf_symtab, symsect, sizeof (ctf_sect_t));
memcpy (&fp->ctf_strtab, strsect, sizeof (ctf_sect_t));
}
if (fp->ctf_data.cts_name != NULL)
fp->ctf_data.cts_name = ctf_strdup (fp->ctf_data.cts_name);
if (fp->ctf_symtab.cts_name != NULL)
fp->ctf_symtab.cts_name = ctf_strdup (fp->ctf_symtab.cts_name);
if (fp->ctf_strtab.cts_name != NULL)
fp->ctf_strtab.cts_name = ctf_strdup (fp->ctf_strtab.cts_name);
if (fp->ctf_data.cts_name == NULL)
fp->ctf_data.cts_name = _CTF_NULLSTR;
if (fp->ctf_symtab.cts_name == NULL)
fp->ctf_symtab.cts_name = _CTF_NULLSTR;
if (fp->ctf_strtab.cts_name == NULL)
fp->ctf_strtab.cts_name = _CTF_NULLSTR;
if (strsect != NULL)
{
fp->ctf_str[CTF_STRTAB_1].cts_strs = strsect->cts_data;
fp->ctf_str[CTF_STRTAB_1].cts_len = strsect->cts_size;
}
if (foreign_endian &&
(err = flip_ctf (&hp, base)) != 0)
{
/* We can be certain that flip_ctf() will have endian-flipped everything
other than the types table when we return. In particular the header
is fine, so set it, to allow freeing to use the usual code path. */
(void) ctf_set_open_errno (errp, err);
ctf_set_base (fp, &hp, base);
goto bad;
}
ctf_set_base (fp, &hp, base);
fp->ctf_size = size + hdrsz;
if ((err = init_types (fp, &hp)) != 0)
{
(void) ctf_set_open_errno (errp, err);
goto bad;
}
/* If we have a symbol table section, allocate and initialize
the symtab translation table, pointed to by ctf_sxlate. */
if (symsect != NULL)
{
fp->ctf_nsyms = symsect->cts_size / symsect->cts_entsize;
fp->ctf_sxlate = ctf_alloc (fp->ctf_nsyms * sizeof (uint32_t));
if (fp->ctf_sxlate == NULL)
{
(void) ctf_set_open_errno (errp, ENOMEM);
goto bad;
}
if ((err = init_symtab (fp, &hp, symsect, strsect)) != 0)
{
(void) ctf_set_open_errno (errp, err);
goto bad;
}
}
/* Initialize the ctf_lookup_by_name top-level dictionary. We keep an
array of type name prefixes and the corresponding ctf_hash to use.
NOTE: This code must be kept in sync with the code in ctf_update(). */
fp->ctf_lookups[0].ctl_prefix = "struct";
fp->ctf_lookups[0].ctl_len = strlen (fp->ctf_lookups[0].ctl_prefix);
fp->ctf_lookups[0].ctl_hash = fp->ctf_structs;
fp->ctf_lookups[1].ctl_prefix = "union";
fp->ctf_lookups[1].ctl_len = strlen (fp->ctf_lookups[1].ctl_prefix);
fp->ctf_lookups[1].ctl_hash = fp->ctf_unions;
fp->ctf_lookups[2].ctl_prefix = "enum";
fp->ctf_lookups[2].ctl_len = strlen (fp->ctf_lookups[2].ctl_prefix);
fp->ctf_lookups[2].ctl_hash = fp->ctf_enums;
fp->ctf_lookups[3].ctl_prefix = _CTF_NULLSTR;
fp->ctf_lookups[3].ctl_len = strlen (fp->ctf_lookups[3].ctl_prefix);
fp->ctf_lookups[3].ctl_hash = fp->ctf_names;
fp->ctf_lookups[4].ctl_prefix = NULL;
fp->ctf_lookups[4].ctl_len = 0;
fp->ctf_lookups[4].ctl_hash = NULL;
if (symsect != NULL)
{
if (symsect->cts_entsize == sizeof (Elf64_Sym))
(void) ctf_setmodel (fp, CTF_MODEL_LP64);
else
(void) ctf_setmodel (fp, CTF_MODEL_ILP32);
}
else
(void) ctf_setmodel (fp, CTF_MODEL_NATIVE);
fp->ctf_refcnt = 1;
return fp;
bad:
ctf_file_close (fp);
return NULL;
}
/* Close the specified CTF container and free associated data structures. Note
that ctf_file_close() is a reference counted operation: if the specified file
is the parent of other active containers, its reference count will be greater
than one and it will be freed later when no active children exist. */
void
ctf_file_close (ctf_file_t *fp)
{
ctf_dtdef_t *dtd, *ntd;
ctf_dvdef_t *dvd, *nvd;
if (fp == NULL)
return; /* Allow ctf_file_close(NULL) to simplify caller code. */
ctf_dprintf ("ctf_file_close(%p) refcnt=%u\n", (void *) fp, fp->ctf_refcnt);
if (fp->ctf_refcnt > 1)
{
fp->ctf_refcnt--;
return;
}
if (fp->ctf_dynparname != NULL)
ctf_free (fp->ctf_dynparname);
if (fp->ctf_parent != NULL)
ctf_file_close (fp->ctf_parent);
for (dtd = ctf_list_next (&fp->ctf_dtdefs); dtd != NULL; dtd = ntd)
{
ntd = ctf_list_next (dtd);
ctf_dtd_delete (fp, dtd);
}
ctf_dynhash_destroy (fp->ctf_dthash);
ctf_dynhash_destroy (fp->ctf_dtbyname);
for (dvd = ctf_list_next (&fp->ctf_dvdefs); dvd != NULL; dvd = nvd)
{
nvd = ctf_list_next (dvd);
ctf_dvd_delete (fp, dvd);
}
ctf_dynhash_destroy (fp->ctf_dvhash);
libctf: deduplicate and sort the string table ctf.h states: > [...] the CTF string table does not contain any duplicated strings. Unfortunately this is entirely untrue: libctf has before now made no attempt whatsoever to deduplicate the string table. It computes the string table's length on the fly as it adds new strings to the dynamic CTF file, and ctf_update() just writes each string to the table and notes the current write position as it traverses the dynamic CTF file's data structures and builds the final CTF buffer. There is no global view of the strings and no deduplication. Fix this by erasing the ctf_dtvstrlen dead-reckoning length, and adding a new dynhash table ctf_str_atoms that maps unique strings to a list of references to those strings: a reference is a simple uint32_t * to some value somewhere in the under-construction CTF buffer that needs updating to note the string offset when the strtab is laid out. Adding a string is now a simple matter of calling ctf_str_add_ref(), which adds a new atom to the atoms table, if one doesn't already exist, and adding the location of the reference to this atom to the refs list attached to the atom: this works reliably as long as one takes care to only call ctf_str_add_ref() once the final location of the offset is known (so you can't call it on a temporary structure and then memcpy() that structure into place in the CTF buffer, because the ref will still point to the old location: ctf_update() changes accordingly). Generating the CTF string table is a matter of calling ctf_str_write_strtab(), which counts the length and number of elements in the atoms table using the ctf_dynhash_iter() function we just added, populating an array of pointers into the atoms table and sorting it into order (to help compressors), then traversing this table and emitting it, updating the refs to each atom as we go. The only complexity here is arranging to keep the null string at offset zero, since a lot of code in libctf depends on being able to leave strtab references at 0 to indicate 'no name'. Once the table is constructed and the refs updated, we know how long it is, so we can realloc() the partial CTF buffer we allocated earlier and can copy the table on to the end of it (and purge the refs because they're not needed any more and have been invalidated by the realloc() call in any case). The net effect of all this is a reduction in uncompressed strtab sizes of about 30% (perhaps a quarter to a half of all strings across the Linux kernel are eliminated as duplicates). Of course, duplicated strings are highly redundant, so the space saving after compression is only about 20%: when the other non-strtab sections are factored in, CTF sizes shrink by about 10%. No change in externally-visible API or file format (other than the reduction in pointless redundancy). libctf/ * ctf-impl.h: (struct ctf_strs_writable): New, non-const version of struct ctf_strs. (struct ctf_dtdef): Note that dtd_data.ctt_name is unpopulated. (struct ctf_str_atom): New, disambiguated single string. (struct ctf_str_atom_ref): New, points to some other location that references this string's offset. (struct ctf_file): New members ctf_str_atoms and ctf_str_num_refs. Remove member ctf_dtvstrlen: we no longer track the total strlen as we add strings. (ctf_str_create_atoms): Declare new function in ctf-string.c. (ctf_str_free_atoms): Likewise. (ctf_str_add): Likewise. (ctf_str_add_ref): Likewise. (ctf_str_purge_refs): Likewise. (ctf_str_write_strtab): Likewise. (ctf_realloc): Declare new function in ctf-util.c. * ctf-open.c (ctf_bufopen): Create the atoms table. (ctf_file_close): Destroy it. * ctf-create.c (ctf_update): Copy-and-free it on update. No longer special-case the position of the parname string. Construct the strtab by calling ctf_str_add_ref and ctf_str_write_strtab after the rest of each buffer element is constructed, not via open-coding: realloc the CTF buffer and append the strtab to it. No longer maintain ctf_dtvstrlen. Sort the variable entry table later, after strtab construction. (ctf_copy_membnames): Remove: integrated into ctf_copy_{s,l,e}members. (ctf_copy_smembers): Drop the string offset: call ctf_str_add_ref after buffer element construction instead. (ctf_copy_lmembers): Likewise. (ctf_copy_emembers): Likewise. (ctf_create): No longer maintain the ctf_dtvstrlen. (ctf_dtd_delete): Likewise. (ctf_dvd_delete): Likewise. (ctf_add_generic): Likewise. (ctf_add_enumerator): Likewise. (ctf_add_member_offset): Likewise. (ctf_add_variable): Likewise. (membadd): Likewise. * ctf-util.c (ctf_realloc): New, wrapper around realloc that aborts if there are active ctf_str_num_refs. (ctf_strraw): Move to ctf-string.c. (ctf_strptr): Likewise. * ctf-string.c: New file, strtab manipulation. * Makefile.am (libctf_a_SOURCES): Add it. * Makefile.in: Regenerate.
2019-06-27 14:51:10 +02:00
ctf_str_free_atoms (fp);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
ctf_free (fp->ctf_tmp_typeslice);
if (fp->ctf_data.cts_name != _CTF_NULLSTR &&
fp->ctf_data.cts_name != NULL)
ctf_free ((char *) fp->ctf_data.cts_name);
if (fp->ctf_symtab.cts_name != _CTF_NULLSTR &&
fp->ctf_symtab.cts_name != NULL)
ctf_free ((char *) fp->ctf_symtab.cts_name);
if (fp->ctf_strtab.cts_name != _CTF_NULLSTR &&
fp->ctf_strtab.cts_name != NULL)
ctf_free ((char *) fp->ctf_strtab.cts_name);
else if (fp->ctf_data_mmapped)
ctf_munmap (fp->ctf_data_mmapped, fp->ctf_data_mmapped_len);
libctf: drop mmap()-based CTF data allocator This allocator has the ostensible benefit that it lets us mprotect() the memory used for CTF storage: but in exchange for this it adds considerable complexity, since we have to track allocation sizes ourselves for use at freeing time, note whether the data we are storing was ctf_data_alloc()ed or not so we know if we can safely mprotect() it... and while the mprotect()ing has found few bugs, it *has* been the cause of more than one due to errors in all this tracking leading to us mprotect()ing bits of the heap and stuff like that. We are about to start composing CTF buffers from pieces so that we can do usage-based optimizations on the strtab. This means we need realloc(), which needs nonportable mremap() and *more* tracking of the *original* allocation size, and the complexity and bureaucracy of all of this is just too high for its negligible benefits. Drop the whole thing and just use malloc() like everyone else. It knows better than we do when it is safe to use mmap() under the covers, anyway. While we're at it, don't leak the entire buffer if ctf_compress_write() fails to compress it. libctf/ * ctf-subr.c (_PAGESIZE): Remove. (ctf_data_alloc): Likewise. (ctf_data_free): Likewise. (ctf_data_protect): Likewise. * ctf-impl.h: Remove declarations. * ctf-create.c (ctf_update): No longer call ctf_data_protect: use ctf_free, not ctf_data_free. (ctf_compress_write): Use ctf_data_alloc, not ctf_alloc. Free the buffer again on compression error. * ctf-open.c (ctf_set_base): No longer track the size: call ctf_free, not ctf_data_free. (upgrade_types): Likewise. Call ctf_alloc, not ctf_data_alloc. (ctf_bufopen): Likewise. No longer call ctf_data_protect.
2019-06-19 13:20:47 +02:00
ctf_free_base (fp, NULL);
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
if (fp->ctf_sxlate != NULL)
ctf_free (fp->ctf_sxlate);
if (fp->ctf_txlate != NULL)
ctf_free (fp->ctf_txlate);
if (fp->ctf_ptrtab != NULL)
ctf_free (fp->ctf_ptrtab);
ctf_hash_destroy (fp->ctf_structs);
ctf_hash_destroy (fp->ctf_unions);
ctf_hash_destroy (fp->ctf_enums);
ctf_hash_destroy (fp->ctf_names);
ctf_free (fp);
}
/* The converse of ctf_open(). ctf_open() disguises whatever it opens as an
archive, so closing one is just like closing an archive. */
void
ctf_close (ctf_archive_t *arc)
{
ctf_arc_close (arc);
}
libctf: mmappable archives If you need to store a large number of CTF containers somewhere, this provides a dedicated facility for doing so: an mmappable archive format like a very simple tar or ar without all the system-dependent format horrors or need for heavy file copying, with built-in compression of files above a particular size threshold. libctf automatically mmap()s uncompressed elements of these archives, or uncompresses them, as needed. (If the platform does not support mmap(), copying into dynamically-allocated buffers is used.) Archive iteration operations are partitioned into raw and non-raw forms. Raw operations pass thhe raw archive contents to the callback: non-raw forms open each member with ctf_bufopen() and pass the resulting ctf_file_t to the iterator instead. This lets you manipulate the raw data in the archive, or the contents interpreted as a CTF file, as needed. It is not yet known whether we will store CTF archives in a linked ELF object in one of these (akin to debugdata) or whether they'll get one section per TU plus one parent container for types shared between them. (In the case of ELF objects with very large numbers of TUs, an archive of all of them would seem preferable, so we might just use an archive, and add lzma support so you can assume that .gnu_debugdata and .ctf are compressed using the same algorithm if both are present.) To make usage easier, the ctf_archive_t is not the on-disk representation but an abstraction over both ctf_file_t's and archives of many ctf_file_t's: users see both CTF archives and raw CTF files as ctf_archive_t's upon opening, the only difference being that a raw CTF file has only a single "archive member", named ".ctf" (the default if a null pointer is passed in as the name). The next commit will make use of this facility, in addition to providing the public interface to actually open archives. (In the future, it should be possible to have all CTF sections in an ELF file appear as an "archive" in the same fashion.) This machinery is also used to allow library-internal creators of ctf_archive_t's (such as the next commit) to stash away an ELF string and symbol table, so that all opens of members in a given archive will use them. This lets CTF archives exploit the ELF string and symbol table just like raw CTF files can. (All this leads to somewhat confusing type naming. The ctf_archive_t is a typedef for the opaque internal type, struct ctf_archive_internal: the non-internal "struct ctf_archive" is the on-disk structure meant for other libraries manipulating CTF files. It is probably clearest to use the struct name for struct ctf_archive_internal inside the program, and the typedef names outside.) libctf/ * ctf-archive.c: New. * ctf-impl.h (ctf_archive_internal): New type. (ctf_arc_open_internal): New declaration. (ctf_arc_bufopen): Likewise. (ctf_arc_close_internal): Likewise. include/ * ctf.h (CTFA_MAGIC): New. (struct ctf_archive): New. (struct ctf_archive_modent): Likewise. * ctf-api.h (ctf_archive_member_f): New. (ctf_archive_raw_member_f): Likewise. (ctf_arc_write): Likewise. (ctf_arc_close): Likewise. (ctf_arc_open_by_name): Likewise. (ctf_archive_iter): Likewise. (ctf_archive_raw_iter): Likewise. (ctf_get_arc): Likewise.
2019-04-24 12:30:17 +02:00
/* Get the CTF archive from which this ctf_file_t is derived. */
ctf_archive_t *
ctf_get_arc (const ctf_file_t *fp)
{
return fp->ctf_archive;
}
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
/* Return the ctfsect out of the core ctf_impl. Useful for freeing the
ctfsect's data * after ctf_file_close(), which is why we return the actual
structure, not a pointer to it, since that is likely to become a pointer to
freed data before the return value is used under the expected use case of
ctf_getsect()/ ctf_file_close()/free(). */
extern ctf_sect_t
ctf_getdatasect (const ctf_file_t *fp)
{
return fp->ctf_data;
}
/* Return the CTF handle for the parent CTF container, if one exists.
Otherwise return NULL to indicate this container has no imported parent. */
ctf_file_t *
ctf_parent_file (ctf_file_t *fp)
{
return fp->ctf_parent;
}
/* Return the name of the parent CTF container, if one exists. Otherwise
return NULL to indicate this container is a root container. */
const char *
ctf_parent_name (ctf_file_t *fp)
{
return fp->ctf_parname;
}
/* Set the parent name. It is an error to call this routine without calling
ctf_import() at some point. */
void
ctf_parent_name_set (ctf_file_t *fp, const char *name)
{
if (fp->ctf_dynparname != NULL)
ctf_free (fp->ctf_dynparname);
fp->ctf_dynparname = ctf_strdup (name);
fp->ctf_parname = fp->ctf_dynparname;
}
/* Import the types from the specified parent container by storing a pointer
to it in ctf_parent and incrementing its reference count. Only one parent
is allowed: if a parent already exists, it is replaced by the new parent. */
int
ctf_import (ctf_file_t *fp, ctf_file_t *pfp)
{
if (fp == NULL || fp == pfp || (pfp != NULL && pfp->ctf_refcnt == 0))
return (ctf_set_errno (fp, EINVAL));
if (pfp != NULL && pfp->ctf_dmodel != fp->ctf_dmodel)
return (ctf_set_errno (fp, ECTF_DMODEL));
if (fp->ctf_parent != NULL)
ctf_file_close (fp->ctf_parent);
if (pfp != NULL)
{
fp->ctf_flags |= LCTF_CHILD;
pfp->ctf_refcnt++;
if (fp->ctf_parname == NULL)
ctf_parent_name_set (fp, "PARENT");
}
fp->ctf_parent = pfp;
return 0;
}
/* Set the data model constant for the CTF container. */
int
ctf_setmodel (ctf_file_t *fp, int model)
{
const ctf_dmodel_t *dp;
for (dp = _libctf_models; dp->ctd_name != NULL; dp++)
{
if (dp->ctd_code == model)
{
fp->ctf_dmodel = dp;
return 0;
}
}
return (ctf_set_errno (fp, EINVAL));
}
/* Return the data model constant for the CTF container. */
int
ctf_getmodel (ctf_file_t *fp)
{
return fp->ctf_dmodel->ctd_code;
}
libctf: fix a number of build problems found on Solaris and NetBSD - Use of nonportable <endian.h> - Use of qsort_r - Use of zlib without appropriate magic to pull in the binutils zlib - Use of off64_t without checking (fixed by dropping the unused fields that need off64_t entirely) - signedness problems due to long being too short a type on 32-bit platforms: ctf_id_t is now 'unsigned long', and CTF_ERR must be used only for functions that return ctf_id_t - One lingering use of bzero() and of <sys/errno.h> All fixed, using code from gnulib where possible. Relatedly, set cts_size in a couple of places it was missed (string table and symbol table loading upon ctf_bfdopen()). binutils/ * objdump.c (make_ctfsect): Drop cts_type, cts_flags, and cts_offset. * readelf.c (shdr_to_ctf_sect): Likewise. include/ * ctf-api.h (ctf_sect_t): Drop cts_type, cts_flags, and cts_offset. (ctf_id_t): This is now an unsigned type. (CTF_ERR): Cast it to ctf_id_t. Note that it should only be used for ctf_id_t-returning functions. libctf/ * Makefile.am (ZLIB): New. (ZLIBINC): Likewise. (AM_CFLAGS): Use them. (libctf_a_LIBADD): New, for LIBOBJS. * configure.ac: Check for zlib, endian.h, and qsort_r. * ctf-endian.h: New, providing htole64 and le64toh. * swap.h: Code style fixes. (bswap_identity_64): New. * qsort_r.c: New, from gnulib (with one added #include). * ctf-decls.h: New, providing a conditional qsort_r declaration, and unconditional definitions of MIN and MAX. * ctf-impl.h: Use it. Do not use <sys/errno.h>. (ctf_set_errno): Now returns unsigned long. * ctf-util.c (ctf_set_errno): Adjust here too. * ctf-archive.c: Use ctf-endian.h. (ctf_arc_open_by_offset): Use memset, not bzero. Drop cts_type, cts_flags and cts_offset. (ctf_arc_write): Drop debugging dependent on the size of off_t. * ctf-create.c: Provide a definition of roundup if not defined. (ctf_create): Drop cts_type, cts_flags and cts_offset. (ctf_add_reftype): Do not check if type IDs are below zero. (ctf_add_slice): Likewise. (ctf_add_typedef): Likewise. (ctf_add_member_offset): Cast error-returning ssize_t's to size_t when known error-free. Drop CTF_ERR usage for functions returning int. (ctf_add_member_encoded): Drop CTF_ERR usage for functions returning int. (ctf_add_variable): Likewise. (enumcmp): Likewise. (enumadd): Likewise. (membcmp): Likewise. (ctf_add_type): Likewise. Cast error-returning ssize_t's to size_t when known error-free. * ctf-dump.c (ctf_is_slice): Drop CTF_ERR usage for functions returning int: use CTF_ERR for functions returning ctf_type_id. (ctf_dump_label): Likewise. (ctf_dump_objts): Likewise. * ctf-labels.c (ctf_label_topmost): Likewise. (ctf_label_iter): Likewise. (ctf_label_info): Likewise. * ctf-lookup.c (ctf_func_args): Likewise. * ctf-open.c (upgrade_types): Cast to size_t where appropriate. (ctf_bufopen): Likewise. Use zlib types as needed. * ctf-types.c (ctf_member_iter): Drop CTF_ERR usage for functions returning int. (ctf_enum_iter): Likewise. (ctf_type_size): Likewise. (ctf_type_align): Likewise. Cast to size_t where appropriate. (ctf_type_kind_unsliced): Likewise. (ctf_type_kind): Likewise. (ctf_type_encoding): Likewise. (ctf_member_info): Likewise. (ctf_array_info): Likewise. (ctf_enum_value): Likewise. (ctf_type_rvisit): Likewise. * ctf-open-bfd.c (ctf_bfdopen): Drop cts_type, cts_flags and cts_offset. (ctf_simple_open): Likewise. (ctf_bfdopen_ctfsect): Likewise. Set cts_size properly. * Makefile.in: Regenerate. * aclocal.m4: Likewise. * config.h: Likewise. * configure: Likewise.
2019-05-31 11:10:51 +02:00
/* The caller can hang an arbitrary pointer off each ctf_file_t using this
function. */
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
void
ctf_setspecific (ctf_file_t *fp, void *data)
{
fp->ctf_specific = data;
}
libctf: fix a number of build problems found on Solaris and NetBSD - Use of nonportable <endian.h> - Use of qsort_r - Use of zlib without appropriate magic to pull in the binutils zlib - Use of off64_t without checking (fixed by dropping the unused fields that need off64_t entirely) - signedness problems due to long being too short a type on 32-bit platforms: ctf_id_t is now 'unsigned long', and CTF_ERR must be used only for functions that return ctf_id_t - One lingering use of bzero() and of <sys/errno.h> All fixed, using code from gnulib where possible. Relatedly, set cts_size in a couple of places it was missed (string table and symbol table loading upon ctf_bfdopen()). binutils/ * objdump.c (make_ctfsect): Drop cts_type, cts_flags, and cts_offset. * readelf.c (shdr_to_ctf_sect): Likewise. include/ * ctf-api.h (ctf_sect_t): Drop cts_type, cts_flags, and cts_offset. (ctf_id_t): This is now an unsigned type. (CTF_ERR): Cast it to ctf_id_t. Note that it should only be used for ctf_id_t-returning functions. libctf/ * Makefile.am (ZLIB): New. (ZLIBINC): Likewise. (AM_CFLAGS): Use them. (libctf_a_LIBADD): New, for LIBOBJS. * configure.ac: Check for zlib, endian.h, and qsort_r. * ctf-endian.h: New, providing htole64 and le64toh. * swap.h: Code style fixes. (bswap_identity_64): New. * qsort_r.c: New, from gnulib (with one added #include). * ctf-decls.h: New, providing a conditional qsort_r declaration, and unconditional definitions of MIN and MAX. * ctf-impl.h: Use it. Do not use <sys/errno.h>. (ctf_set_errno): Now returns unsigned long. * ctf-util.c (ctf_set_errno): Adjust here too. * ctf-archive.c: Use ctf-endian.h. (ctf_arc_open_by_offset): Use memset, not bzero. Drop cts_type, cts_flags and cts_offset. (ctf_arc_write): Drop debugging dependent on the size of off_t. * ctf-create.c: Provide a definition of roundup if not defined. (ctf_create): Drop cts_type, cts_flags and cts_offset. (ctf_add_reftype): Do not check if type IDs are below zero. (ctf_add_slice): Likewise. (ctf_add_typedef): Likewise. (ctf_add_member_offset): Cast error-returning ssize_t's to size_t when known error-free. Drop CTF_ERR usage for functions returning int. (ctf_add_member_encoded): Drop CTF_ERR usage for functions returning int. (ctf_add_variable): Likewise. (enumcmp): Likewise. (enumadd): Likewise. (membcmp): Likewise. (ctf_add_type): Likewise. Cast error-returning ssize_t's to size_t when known error-free. * ctf-dump.c (ctf_is_slice): Drop CTF_ERR usage for functions returning int: use CTF_ERR for functions returning ctf_type_id. (ctf_dump_label): Likewise. (ctf_dump_objts): Likewise. * ctf-labels.c (ctf_label_topmost): Likewise. (ctf_label_iter): Likewise. (ctf_label_info): Likewise. * ctf-lookup.c (ctf_func_args): Likewise. * ctf-open.c (upgrade_types): Cast to size_t where appropriate. (ctf_bufopen): Likewise. Use zlib types as needed. * ctf-types.c (ctf_member_iter): Drop CTF_ERR usage for functions returning int. (ctf_enum_iter): Likewise. (ctf_type_size): Likewise. (ctf_type_align): Likewise. Cast to size_t where appropriate. (ctf_type_kind_unsliced): Likewise. (ctf_type_kind): Likewise. (ctf_type_encoding): Likewise. (ctf_member_info): Likewise. (ctf_array_info): Likewise. (ctf_enum_value): Likewise. (ctf_type_rvisit): Likewise. * ctf-open-bfd.c (ctf_bfdopen): Drop cts_type, cts_flags and cts_offset. (ctf_simple_open): Likewise. (ctf_bfdopen_ctfsect): Likewise. Set cts_size properly. * Makefile.in: Regenerate. * aclocal.m4: Likewise. * config.h: Likewise. * configure: Likewise.
2019-05-31 11:10:51 +02:00
/* Retrieve the arbitrary pointer again. */
libctf: opening This fills in the other half of the opening/creation puzzle: opening of already-existing CTF files. Such files are always read-only: if you want to add to a CTF file opened with one of the opening functions in this file, use ctf_add_type(), in a later commit, to copy appropriate types into a newly ctf_create()d, writable container. The lowest-level opening functions are in here: ctf_bufopen(), which takes ctf_sect_t structures akin to ELF section headers, and ctf_simple_open(), which can be used if you don't have an entire ELF section header to work from. Both will malloc() new space for the buffers only if necessary, will mmap() directly from the file if requested, and will mprotect() it afterwards to prevent accidental corruption of the types. These functions are also used by ctf_update() when converting types in a writable container into read-only types that can be looked up using the lookup functions (in later commits). The files are always of the native endianness of the system that created them: at read time, the endianness of the header magic number is used to determine whether or not the file needs byte-swapping, and the entire thing is aggressively byte-swapped. The agggressive nature of this swapping avoids complicating the rest of the code with endianness conversions, while the native endianness introduces no byte-swapping overhead in the common case. (The endianness-independence code is also much newer than everything else in this file, and deserves closer scrutiny.) The accessors at the top of the file are there to transparently support older versions of the CTF file format, allowing translation from older formats that have different sizes for the structures in ctf.h: currently, these older formats are intermingled with the newer ones in ctf.h: they will probably migrate to a compatibility header in time, to ease readability. The ctf_set_base() function is split out for the same reason: when conversion code to a newer format is written, it would need to malloc() new storage for the entire ctf_file_t if a file format change causes it to grow, and for that we need ctf_set_base() to be a separate function. One pair of linked data structures supported by this file has no creation code in libctf yet: the data and function object sections read by init_symtab(). These will probably arrive soon, when the linker comes to need them. (init_symtab() has hardly been changed since 2009, but if any code in libctf has rotted over time, this will.) A few simple accessors are also present that can even be called on read-only containers because they don't actually modify them, since the relevant things are not stored in the container but merely change its operation: ctf_setmodel(), which lets you specify whether a container is LP64 or not (used to statically determine the sizes of a few types), ctf_import(), which is the only way to associate a parent container with a child container, and ctf_setspecific(), which lets the caller associate an arbitrary pointer with the CTF container for any use. If the user doesn't call these functions correctly, libctf will misbehave: this is particularly important for ctf_import(), since a container built against a given parent container will not be able to resolve types that depend on types in the parent unless it is ctf_import()ed with a parent container with the same set of types at the same IDs, or a superset. Possible future extensions (also noted in the ctf-hash.c file) include storing a count of things so that we don't need to do one pass over the CTF file counting everything, and computing a perfect hash at CTF creation time in some compact form, storing it in the CTF file, and using it to hash things so we don't need to do a second pass over the entire CTF file to set up the hashes used to go from names to type IDs. (There are multiple such hashes, one for each C type namespace: types, enums, structs, and unions.) libctf/ * ctf-open.c: New file. * swap.h: Likewise. include/ * ctf-api.h (ctf_file_close): New declaration. (ctf_getdatasect): Likewise. (ctf_parent_file): Likewise. (ctf_parent_name): Likewise. (ctf_parent_name_set): Likewise. (ctf_import): Likewise. (ctf_setmodel): Likewise. (ctf_getmodel): Likewise. (ctf_setspecific): Likewise. (ctf_getspecific): Likewise.
2019-04-24 11:17:13 +02:00
void *
ctf_getspecific (ctf_file_t *fp)
{
return fp->ctf_specific;
}