binutils-gdb/libctf/ctf-serialize.c

1427 lines
43 KiB
C
Raw Normal View History

/* CTF dict creation.
Copyright (C) 2019-2022 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 <assert.h>
#include <string.h>
#include <unistd.h>
#include <zlib.h>
#include <elf.h>
#include "elf-bfd.h"
/* Symtypetab sections. */
/* Symtypetab emission flags. */
#define CTF_SYMTYPETAB_EMIT_FUNCTION 0x1
#define CTF_SYMTYPETAB_EMIT_PAD 0x2
#define CTF_SYMTYPETAB_FORCE_INDEXED 0x4
/* Properties of symtypetab emission, shared by symtypetab section
sizing and symtypetab emission itself. */
typedef struct emit_symtypetab_state
{
/* True if linker-reported symbols are being filtered out. symfp is set if
this is true: otherwise, indexing is forced and the symflags indicate as
much. */
int filter_syms;
/* True if symbols are being sorted. */
int sort_syms;
/* Flags for symtypetab emission. */
int symflags;
/* The dict to which the linker has reported symbols. */
ctf_dict_t *symfp;
/* The maximum number of objects seen. */
size_t maxobjt;
/* The maximum number of func info entris seen. */
size_t maxfunc;
} emit_symtypetab_state_t;
/* Determine if a symbol is "skippable" and should never appear in the
symtypetab sections. */
int
ctf_symtab_skippable (ctf_link_sym_t *sym)
{
/* Never skip symbols whose name is not yet known. */
if (sym->st_nameidx_set)
return 0;
return (sym->st_name == NULL || sym->st_name[0] == 0
|| sym->st_shndx == SHN_UNDEF
|| strcmp (sym->st_name, "_START_") == 0
|| strcmp (sym->st_name, "_END_") == 0
|| (sym->st_type == STT_OBJECT && sym->st_shndx == SHN_EXTABS
&& sym->st_value == 0));
}
/* Get the number of symbols in a symbol hash, the count of symbols, the maximum
seen, the eventual size, without any padding elements, of the func/data and
(if generated) index sections, and the size of accumulated padding elements.
The linker-reported set of symbols is found in SYMFP: it may be NULL if
symbol filtering is not desired, in which case CTF_SYMTYPETAB_FORCE_INDEXED
will always be set in the flags.
Also figure out if any symbols need to be moved to the variable section, and
add them (if not already present). */
_libctf_nonnull_ ((1,3,4,5,6,7,8))
static int
symtypetab_density (ctf_dict_t *fp, ctf_dict_t *symfp, ctf_dynhash_t *symhash,
size_t *count, size_t *max, size_t *unpadsize,
size_t *padsize, size_t *idxsize, int flags)
{
ctf_next_t *i = NULL;
const void *name;
const void *ctf_sym;
ctf_dynhash_t *linker_known = NULL;
int err;
int beyond_max = 0;
*count = 0;
*max = 0;
*unpadsize = 0;
*idxsize = 0;
*padsize = 0;
if (!(flags & CTF_SYMTYPETAB_FORCE_INDEXED))
{
/* Make a dynhash citing only symbols reported by the linker of the
appropriate type, then traverse all potential-symbols we know the types
of, removing them from linker_known as we go. Once this is done, the
only symbols remaining in linker_known are symbols we don't know the
types of: we must emit pads for those symbols that are below the
maximum symbol we will emit (any beyond that are simply skipped).
If there are none, this symtypetab will be empty: just report that. */
if (!symfp->ctf_dynsyms)
return 0;
if ((linker_known = ctf_dynhash_create (ctf_hash_string, ctf_hash_eq_string,
NULL, NULL)) == NULL)
return (ctf_set_errno (fp, ENOMEM));
while ((err = ctf_dynhash_cnext (symfp->ctf_dynsyms, &i,
&name, &ctf_sym)) == 0)
{
ctf_link_sym_t *sym = (ctf_link_sym_t *) ctf_sym;
if (((flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& sym->st_type != STT_FUNC)
|| (!(flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& sym->st_type != STT_OBJECT))
continue;
if (ctf_symtab_skippable (sym))
continue;
/* This should only be true briefly before all the names are
finalized, long before we get this far. */
if (!ctf_assert (fp, !sym->st_nameidx_set))
return -1; /* errno is set for us. */
if (ctf_dynhash_cinsert (linker_known, name, ctf_sym) < 0)
{
ctf_dynhash_destroy (linker_known);
return (ctf_set_errno (fp, ENOMEM));
}
}
if (err != ECTF_NEXT_END)
{
ctf_err_warn (fp, 0, err, _("iterating over linker-known symbols during "
"serialization"));
ctf_dynhash_destroy (linker_known);
return (ctf_set_errno (fp, err));
}
}
while ((err = ctf_dynhash_cnext (symhash, &i, &name, NULL)) == 0)
{
ctf_link_sym_t *sym;
if (!(flags & CTF_SYMTYPETAB_FORCE_INDEXED))
{
/* Linker did not report symbol in symtab. Remove it from the
set of known data symbols and continue. */
if ((sym = ctf_dynhash_lookup (symfp->ctf_dynsyms, name)) == NULL)
{
ctf_dynhash_remove (symhash, name);
continue;
}
/* We don't remove skippable symbols from the symhash because we don't
want them to be migrated into variables. */
if (ctf_symtab_skippable (sym))
continue;
if ((flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& sym->st_type != STT_FUNC)
{
ctf_err_warn (fp, 1, 0, _("symbol %s (%x) added to CTF as a "
"function but is of type %x. "
"The symbol type lookup tables "
"are probably corrupted"),
sym->st_name, sym->st_symidx, sym->st_type);
ctf_dynhash_remove (symhash, name);
continue;
}
else if (!(flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& sym->st_type != STT_OBJECT)
{
ctf_err_warn (fp, 1, 0, _("symbol %s (%x) added to CTF as a "
"data object but is of type %x. "
"The symbol type lookup tables "
"are probably corrupted"),
sym->st_name, sym->st_symidx, sym->st_type);
ctf_dynhash_remove (symhash, name);
continue;
}
ctf_dynhash_remove (linker_known, name);
}
*unpadsize += sizeof (uint32_t);
(*count)++;
if (!(flags & CTF_SYMTYPETAB_FORCE_INDEXED))
{
if (*max < sym->st_symidx)
*max = sym->st_symidx;
}
else
(*max)++;
}
if (err != ECTF_NEXT_END)
{
ctf_err_warn (fp, 0, err, _("iterating over CTF symtypetab during "
"serialization"));
ctf_dynhash_destroy (linker_known);
return (ctf_set_errno (fp, err));
}
if (!(flags & CTF_SYMTYPETAB_FORCE_INDEXED))
{
while ((err = ctf_dynhash_cnext (linker_known, &i, NULL, &ctf_sym)) == 0)
{
ctf_link_sym_t *sym = (ctf_link_sym_t *) ctf_sym;
if (sym->st_symidx > *max)
beyond_max++;
}
if (err != ECTF_NEXT_END)
{
ctf_err_warn (fp, 0, err, _("iterating over linker-known symbols "
"during CTF serialization"));
ctf_dynhash_destroy (linker_known);
return (ctf_set_errno (fp, err));
}
}
*idxsize = *count * sizeof (uint32_t);
if (!(flags & CTF_SYMTYPETAB_FORCE_INDEXED))
*padsize = (ctf_dynhash_elements (linker_known) - beyond_max) * sizeof (uint32_t);
ctf_dynhash_destroy (linker_known);
return 0;
}
/* Emit an objt or func symtypetab into DP in a particular order defined by an
array of ctf_link_sym_t or symbol names passed in. The index has NIDX
elements in it: unindexed output would terminate at symbol OUTMAX and is in
any case no larger than SIZE bytes. Some index elements are expected to be
skipped: see symtypetab_density. The linker-reported set of symbols (if any)
is found in SYMFP. */
static int
emit_symtypetab (ctf_dict_t *fp, ctf_dict_t *symfp, uint32_t *dp,
ctf_link_sym_t **idx, const char **nameidx, uint32_t nidx,
uint32_t outmax, int size, int flags)
{
uint32_t i;
uint32_t *dpp = dp;
ctf_dynhash_t *symhash;
ctf_dprintf ("Emitting table of size %i, outmax %u, %u symtypetab entries, "
"flags %i\n", size, outmax, nidx, flags);
/* Empty table? Nothing to do. */
if (size == 0)
return 0;
if (flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
symhash = fp->ctf_funchash;
else
symhash = fp->ctf_objthash;
for (i = 0; i < nidx; i++)
{
const char *sym_name;
void *type;
/* If we have a linker-reported set of symbols, we may be given that set
to work from, or a set of symbol names. In both cases we want to look
at the corresponding linker-reported symbol (if any). */
if (!(flags & CTF_SYMTYPETAB_FORCE_INDEXED))
{
ctf_link_sym_t *this_link_sym;
if (idx)
this_link_sym = idx[i];
else
this_link_sym = ctf_dynhash_lookup (symfp->ctf_dynsyms, nameidx[i]);
/* Unreported symbol number. No pad, no nothing. */
if (!this_link_sym)
continue;
/* Symbol of the wrong type, or skippable? This symbol is not in this
table. */
if (((flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& this_link_sym->st_type != STT_FUNC)
|| (!(flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& this_link_sym->st_type != STT_OBJECT))
continue;
if (ctf_symtab_skippable (this_link_sym))
continue;
sym_name = this_link_sym->st_name;
/* Linker reports symbol of a different type to the symbol we actually
added? Skip the symbol. No pad, since the symbol doesn't actually
belong in this table at all. (Warned about in
symtypetab_density.) */
if ((this_link_sym->st_type == STT_FUNC)
&& (ctf_dynhash_lookup (fp->ctf_objthash, sym_name)))
continue;
if ((this_link_sym->st_type == STT_OBJECT)
&& (ctf_dynhash_lookup (fp->ctf_funchash, sym_name)))
continue;
}
else
sym_name = nameidx[i];
/* Symbol in index but no type set? Silently skip and (optionally)
pad. (In force-indexed mode, this is also where we track symbols of
the wrong type for this round of insertion.) */
if ((type = ctf_dynhash_lookup (symhash, sym_name)) == NULL)
{
if (flags & CTF_SYMTYPETAB_EMIT_PAD)
*dpp++ = 0;
continue;
}
if (!ctf_assert (fp, (((char *) dpp) - (char *) dp) < size))
return -1; /* errno is set for us. */
*dpp++ = (ctf_id_t) (uintptr_t) type;
/* When emitting unindexed output, all later symbols are pads: stop
early. */
if ((flags & CTF_SYMTYPETAB_EMIT_PAD) && idx[i]->st_symidx == outmax)
break;
}
return 0;
}
/* Emit an objt or func symtypetab index into DP in a paticular order defined by
an array of symbol names passed in. Stop at NIDX. The linker-reported set
of symbols (if any) is found in SYMFP. */
static int
emit_symtypetab_index (ctf_dict_t *fp, ctf_dict_t *symfp, uint32_t *dp,
const char **idx, uint32_t nidx, int size, int flags)
{
uint32_t i;
uint32_t *dpp = dp;
ctf_dynhash_t *symhash;
ctf_dprintf ("Emitting index of size %i, %u entries reported by linker, "
"flags %i\n", size, nidx, flags);
/* Empty table? Nothing to do. */
if (size == 0)
return 0;
if (flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
symhash = fp->ctf_funchash;
else
symhash = fp->ctf_objthash;
/* Indexes should always be unpadded. */
if (!ctf_assert (fp, !(flags & CTF_SYMTYPETAB_EMIT_PAD)))
return -1; /* errno is set for us. */
for (i = 0; i < nidx; i++)
{
const char *sym_name;
void *type;
if (!(flags & CTF_SYMTYPETAB_FORCE_INDEXED))
{
ctf_link_sym_t *this_link_sym;
this_link_sym = ctf_dynhash_lookup (symfp->ctf_dynsyms, idx[i]);
/* This is an index: unreported symbols should never appear in it. */
if (!ctf_assert (fp, this_link_sym != NULL))
return -1; /* errno is set for us. */
/* Symbol of the wrong type, or skippable? This symbol is not in this
table. */
if (((flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& this_link_sym->st_type != STT_FUNC)
|| (!(flags & CTF_SYMTYPETAB_EMIT_FUNCTION)
&& this_link_sym->st_type != STT_OBJECT))
continue;
if (ctf_symtab_skippable (this_link_sym))
continue;
sym_name = this_link_sym->st_name;
/* Linker reports symbol of a different type to the symbol we actually
added? Skip the symbol. */
if ((this_link_sym->st_type == STT_FUNC)
&& (ctf_dynhash_lookup (fp->ctf_objthash, sym_name)))
continue;
if ((this_link_sym->st_type == STT_OBJECT)
&& (ctf_dynhash_lookup (fp->ctf_funchash, sym_name)))
continue;
}
else
sym_name = idx[i];
/* Symbol in index and reported by linker, but no type set? Silently skip
and (optionally) pad. (In force-indexed mode, this is also where we
track symbols of the wrong type for this round of insertion.) */
if ((type = ctf_dynhash_lookup (symhash, sym_name)) == NULL)
continue;
ctf_str_add_ref (fp, sym_name, dpp++);
if (!ctf_assert (fp, (((char *) dpp) - (char *) dp) <= size))
return -1; /* errno is set for us. */
}
return 0;
}
include, libctf, ld: extend variable section to contain functions too The CTF variable section is an optional (usually-not-present) section in the CTF dict which contains name -> type mappings corresponding to data symbols that are present in the linker input but not in the output symbol table: the idea is that programs that use their own symbol- resolution mechanisms can use this section to look up the types of symbols they have found using their own mechanism. Because these removed symbols (mostly static variables, functions, etc) all have names that are unlikely to appear in the ELF symtab and because very few programs have their own symbol-resolution mechanisms, a special linker flag (--ctf-variables) is needed to emit this section. Historically, we emitted only removed data symbols into the variable section. This seemed to make sense at the time, but in hindsight it really doesn't: functions are symbols too, and a C program can look them up just like any other type. So extend the variable section so that it contains all static function symbols too (if it is emitted at all), with types of kind CTF_K_FUNCTION. This is a little fiddly. We relied on compiler assistance for data symbols: the compiler simply emits all data symbols twice, once into the symtypetab as an indexed symbol and once into the variable section. Rather than wait for a suitably adjusted compiler that does the same for function symbols, we can pluck unreported function symbols out of the symtab and add them to the variable section ourselves. While we're at it, we do the same with data symbols: this is redundant right now because the compiler does it, but it costs very little time and lets the compiler drop this kludge and save a little space in .o files. include/ * ctf.h: Mention the new things we can see in the variable section. ld/ * testsuite/ld-ctf/data-func-conflicted-vars.d: New test. libctf/ * ctf-link.c (ctf_link_deduplicating_variables): Duplicate symbols into the variable section too. * ctf-serialize.c (symtypetab_delete_nonstatic_vars): Rename to... (symtypetab_delete_nonstatics): ... this. Check the funchash when pruning redundant variables. (ctf_symtypetab_sect_sizes): Adjust accordingly. * NEWS: Describe this change.
2022-03-16 23:29:25 +08:00
/* Delete symbols that have been assigned names from the variable section. Must
be called from within ctf_serialize, because that is the only place you can
safely delete variables without messing up ctf_rollback. */
static int
include, libctf, ld: extend variable section to contain functions too The CTF variable section is an optional (usually-not-present) section in the CTF dict which contains name -> type mappings corresponding to data symbols that are present in the linker input but not in the output symbol table: the idea is that programs that use their own symbol- resolution mechanisms can use this section to look up the types of symbols they have found using their own mechanism. Because these removed symbols (mostly static variables, functions, etc) all have names that are unlikely to appear in the ELF symtab and because very few programs have their own symbol-resolution mechanisms, a special linker flag (--ctf-variables) is needed to emit this section. Historically, we emitted only removed data symbols into the variable section. This seemed to make sense at the time, but in hindsight it really doesn't: functions are symbols too, and a C program can look them up just like any other type. So extend the variable section so that it contains all static function symbols too (if it is emitted at all), with types of kind CTF_K_FUNCTION. This is a little fiddly. We relied on compiler assistance for data symbols: the compiler simply emits all data symbols twice, once into the symtypetab as an indexed symbol and once into the variable section. Rather than wait for a suitably adjusted compiler that does the same for function symbols, we can pluck unreported function symbols out of the symtab and add them to the variable section ourselves. While we're at it, we do the same with data symbols: this is redundant right now because the compiler does it, but it costs very little time and lets the compiler drop this kludge and save a little space in .o files. include/ * ctf.h: Mention the new things we can see in the variable section. ld/ * testsuite/ld-ctf/data-func-conflicted-vars.d: New test. libctf/ * ctf-link.c (ctf_link_deduplicating_variables): Duplicate symbols into the variable section too. * ctf-serialize.c (symtypetab_delete_nonstatic_vars): Rename to... (symtypetab_delete_nonstatics): ... this. Check the funchash when pruning redundant variables. (ctf_symtypetab_sect_sizes): Adjust accordingly. * NEWS: Describe this change.
2022-03-16 23:29:25 +08:00
symtypetab_delete_nonstatics (ctf_dict_t *fp, ctf_dict_t *symfp)
{
ctf_dvdef_t *dvd, *nvd;
ctf_id_t type;
for (dvd = ctf_list_next (&fp->ctf_dvdefs); dvd != NULL; dvd = nvd)
{
nvd = ctf_list_next (dvd);
include, libctf, ld: extend variable section to contain functions too The CTF variable section is an optional (usually-not-present) section in the CTF dict which contains name -> type mappings corresponding to data symbols that are present in the linker input but not in the output symbol table: the idea is that programs that use their own symbol- resolution mechanisms can use this section to look up the types of symbols they have found using their own mechanism. Because these removed symbols (mostly static variables, functions, etc) all have names that are unlikely to appear in the ELF symtab and because very few programs have their own symbol-resolution mechanisms, a special linker flag (--ctf-variables) is needed to emit this section. Historically, we emitted only removed data symbols into the variable section. This seemed to make sense at the time, but in hindsight it really doesn't: functions are symbols too, and a C program can look them up just like any other type. So extend the variable section so that it contains all static function symbols too (if it is emitted at all), with types of kind CTF_K_FUNCTION. This is a little fiddly. We relied on compiler assistance for data symbols: the compiler simply emits all data symbols twice, once into the symtypetab as an indexed symbol and once into the variable section. Rather than wait for a suitably adjusted compiler that does the same for function symbols, we can pluck unreported function symbols out of the symtab and add them to the variable section ourselves. While we're at it, we do the same with data symbols: this is redundant right now because the compiler does it, but it costs very little time and lets the compiler drop this kludge and save a little space in .o files. include/ * ctf.h: Mention the new things we can see in the variable section. ld/ * testsuite/ld-ctf/data-func-conflicted-vars.d: New test. libctf/ * ctf-link.c (ctf_link_deduplicating_variables): Duplicate symbols into the variable section too. * ctf-serialize.c (symtypetab_delete_nonstatic_vars): Rename to... (symtypetab_delete_nonstatics): ... this. Check the funchash when pruning redundant variables. (ctf_symtypetab_sect_sizes): Adjust accordingly. * NEWS: Describe this change.
2022-03-16 23:29:25 +08:00
if ((((type = (ctf_id_t) (uintptr_t)
ctf_dynhash_lookup (fp->ctf_objthash, dvd->dvd_name)) > 0)
|| (type = (ctf_id_t) (uintptr_t)
ctf_dynhash_lookup (fp->ctf_funchash, dvd->dvd_name)) > 0)
&& ctf_dynhash_lookup (symfp->ctf_dynsyms, dvd->dvd_name) != NULL
&& type == dvd->dvd_type)
ctf_dvd_delete (fp, dvd);
}
return 0;
}
/* Figure out the sizes of the symtypetab sections, their indexed state,
etc. */
static int
ctf_symtypetab_sect_sizes (ctf_dict_t *fp, emit_symtypetab_state_t *s,
ctf_header_t *hdr, size_t *objt_size,
size_t *func_size, size_t *objtidx_size,
size_t *funcidx_size)
{
size_t nfuncs, nobjts;
size_t objt_unpadsize, func_unpadsize, objt_padsize, func_padsize;
/* If doing a writeout as part of linking, and the link flags request it,
filter out reported symbols from the variable section, and filter out all
other symbols from the symtypetab sections. (If we are not linking, the
symbols are sorted; if we are linking, don't bother sorting if we are not
filtering out reported symbols: this is almost certaily an ld -r and only
the linker is likely to consume these symtypetabs again. The linker
doesn't care what order the symtypetab entries is in, since it only
iterates over symbols and does not use the ctf_lookup_by_symbol* API.) */
s->sort_syms = 1;
if (fp->ctf_flags & LCTF_LINKING)
{
s->filter_syms = !(fp->ctf_link_flags & CTF_LINK_NO_FILTER_REPORTED_SYMS);
if (!s->filter_syms)
s->sort_syms = 0;
}
/* Find the dict to which the linker has reported symbols, if any. */
if (s->filter_syms)
{
if (!fp->ctf_dynsyms && fp->ctf_parent && fp->ctf_parent->ctf_dynsyms)
s->symfp = fp->ctf_parent;
else
s->symfp = fp;
}
/* If not filtering, keep all potential symbols in an unsorted, indexed
dict. */
if (!s->filter_syms)
s->symflags = CTF_SYMTYPETAB_FORCE_INDEXED;
else
hdr->cth_flags |= CTF_F_IDXSORTED;
if (!ctf_assert (fp, (s->filter_syms && s->symfp)
|| (!s->filter_syms && !s->symfp
&& ((s->symflags & CTF_SYMTYPETAB_FORCE_INDEXED) != 0))))
return -1;
/* Work out the sizes of the object and function sections, and work out the
number of pad (unassigned) symbols in each, and the overall size of the
sections. */
if (symtypetab_density (fp, s->symfp, fp->ctf_objthash, &nobjts, &s->maxobjt,
&objt_unpadsize, &objt_padsize, objtidx_size,
s->symflags) < 0)
return -1; /* errno is set for us. */
ctf_dprintf ("Object symtypetab: %i objects, max %i, unpadded size %i, "
"%i bytes of pads, index size %i\n", (int) nobjts,
(int) s->maxobjt, (int) objt_unpadsize, (int) objt_padsize,
(int) *objtidx_size);
if (symtypetab_density (fp, s->symfp, fp->ctf_funchash, &nfuncs, &s->maxfunc,
&func_unpadsize, &func_padsize, funcidx_size,
s->symflags | CTF_SYMTYPETAB_EMIT_FUNCTION) < 0)
return -1; /* errno is set for us. */
ctf_dprintf ("Function symtypetab: %i functions, max %i, unpadded size %i, "
"%i bytes of pads, index size %i\n", (int) nfuncs,
(int) s->maxfunc, (int) func_unpadsize, (int) func_padsize,
(int) *funcidx_size);
/* It is worth indexing each section if it would save space to do so, due to
reducing the number of pads sufficiently. A pad is the same size as a
single index entry: but index sections compress relatively poorly compared
to constant pads, so it takes a lot of contiguous padding to equal one
index section entry. It would be nice to be able to *verify* whether we
would save space after compression rather than guessing, but this seems
difficult, since it would require complete reserialization. Regardless, if
the linker has not reported any symbols (e.g. if this is not a final link
but just an ld -r), we must emit things in indexed fashion just as the
compiler does. */
*objt_size = objt_unpadsize;
if (!(s->symflags & CTF_SYMTYPETAB_FORCE_INDEXED)
&& ((objt_padsize + objt_unpadsize) * CTF_INDEX_PAD_THRESHOLD
> objt_padsize))
{
*objt_size += objt_padsize;
*objtidx_size = 0;
}
*func_size = func_unpadsize;
if (!(s->symflags & CTF_SYMTYPETAB_FORCE_INDEXED)
&& ((func_padsize + func_unpadsize) * CTF_INDEX_PAD_THRESHOLD
> func_padsize))
{
*func_size += func_padsize;
*funcidx_size = 0;
}
/* If we are filtering symbols out, those symbols that the linker has not
reported have now been removed from the ctf_objthash and ctf_funchash.
include, libctf, ld: extend variable section to contain functions too The CTF variable section is an optional (usually-not-present) section in the CTF dict which contains name -> type mappings corresponding to data symbols that are present in the linker input but not in the output symbol table: the idea is that programs that use their own symbol- resolution mechanisms can use this section to look up the types of symbols they have found using their own mechanism. Because these removed symbols (mostly static variables, functions, etc) all have names that are unlikely to appear in the ELF symtab and because very few programs have their own symbol-resolution mechanisms, a special linker flag (--ctf-variables) is needed to emit this section. Historically, we emitted only removed data symbols into the variable section. This seemed to make sense at the time, but in hindsight it really doesn't: functions are symbols too, and a C program can look them up just like any other type. So extend the variable section so that it contains all static function symbols too (if it is emitted at all), with types of kind CTF_K_FUNCTION. This is a little fiddly. We relied on compiler assistance for data symbols: the compiler simply emits all data symbols twice, once into the symtypetab as an indexed symbol and once into the variable section. Rather than wait for a suitably adjusted compiler that does the same for function symbols, we can pluck unreported function symbols out of the symtab and add them to the variable section ourselves. While we're at it, we do the same with data symbols: this is redundant right now because the compiler does it, but it costs very little time and lets the compiler drop this kludge and save a little space in .o files. include/ * ctf.h: Mention the new things we can see in the variable section. ld/ * testsuite/ld-ctf/data-func-conflicted-vars.d: New test. libctf/ * ctf-link.c (ctf_link_deduplicating_variables): Duplicate symbols into the variable section too. * ctf-serialize.c (symtypetab_delete_nonstatic_vars): Rename to... (symtypetab_delete_nonstatics): ... this. Check the funchash when pruning redundant variables. (ctf_symtypetab_sect_sizes): Adjust accordingly. * NEWS: Describe this change.
2022-03-16 23:29:25 +08:00
Delete entries from the variable section that duplicate newly-added
symbols. There's no need to migrate new ones in: we do that (if necessary)
in ctf_link_deduplicating_variables. */
if (s->filter_syms && s->symfp->ctf_dynsyms &&
include, libctf, ld: extend variable section to contain functions too The CTF variable section is an optional (usually-not-present) section in the CTF dict which contains name -> type mappings corresponding to data symbols that are present in the linker input but not in the output symbol table: the idea is that programs that use their own symbol- resolution mechanisms can use this section to look up the types of symbols they have found using their own mechanism. Because these removed symbols (mostly static variables, functions, etc) all have names that are unlikely to appear in the ELF symtab and because very few programs have their own symbol-resolution mechanisms, a special linker flag (--ctf-variables) is needed to emit this section. Historically, we emitted only removed data symbols into the variable section. This seemed to make sense at the time, but in hindsight it really doesn't: functions are symbols too, and a C program can look them up just like any other type. So extend the variable section so that it contains all static function symbols too (if it is emitted at all), with types of kind CTF_K_FUNCTION. This is a little fiddly. We relied on compiler assistance for data symbols: the compiler simply emits all data symbols twice, once into the symtypetab as an indexed symbol and once into the variable section. Rather than wait for a suitably adjusted compiler that does the same for function symbols, we can pluck unreported function symbols out of the symtab and add them to the variable section ourselves. While we're at it, we do the same with data symbols: this is redundant right now because the compiler does it, but it costs very little time and lets the compiler drop this kludge and save a little space in .o files. include/ * ctf.h: Mention the new things we can see in the variable section. ld/ * testsuite/ld-ctf/data-func-conflicted-vars.d: New test. libctf/ * ctf-link.c (ctf_link_deduplicating_variables): Duplicate symbols into the variable section too. * ctf-serialize.c (symtypetab_delete_nonstatic_vars): Rename to... (symtypetab_delete_nonstatics): ... this. Check the funchash when pruning redundant variables. (ctf_symtypetab_sect_sizes): Adjust accordingly. * NEWS: Describe this change.
2022-03-16 23:29:25 +08:00
symtypetab_delete_nonstatics (fp, s->symfp) < 0)
return -1;
return 0;
}
static int
ctf_emit_symtypetab_sects (ctf_dict_t *fp, emit_symtypetab_state_t *s,
unsigned char **tptr, size_t objt_size,
size_t func_size, size_t objtidx_size,
size_t funcidx_size)
{
unsigned char *t = *tptr;
size_t nsymtypes = 0;
const char **sym_name_order = NULL;
int err;
/* Sort the linker's symbols into name order if need be. */
if ((objtidx_size != 0) || (funcidx_size != 0))
{
ctf_next_t *i = NULL;
void *symname;
const char **walk;
if (s->filter_syms)
{
if (s->symfp->ctf_dynsyms)
nsymtypes = ctf_dynhash_elements (s->symfp->ctf_dynsyms);
else
nsymtypes = 0;
}
else
nsymtypes = ctf_dynhash_elements (fp->ctf_objthash)
+ ctf_dynhash_elements (fp->ctf_funchash);
if ((sym_name_order = calloc (nsymtypes, sizeof (const char *))) == NULL)
goto oom;
walk = sym_name_order;
if (s->filter_syms)
{
if (s->symfp->ctf_dynsyms)
{
while ((err = ctf_dynhash_next_sorted (s->symfp->ctf_dynsyms, &i,
&symname, NULL,
ctf_dynhash_sort_by_name,
NULL)) == 0)
*walk++ = (const char *) symname;
if (err != ECTF_NEXT_END)
goto symerr;
}
}
else
{
ctf_hash_sort_f sort_fun = NULL;
/* Since we partition the set of symbols back into objt and func,
we can sort the two independently without harm. */
if (s->sort_syms)
sort_fun = ctf_dynhash_sort_by_name;
while ((err = ctf_dynhash_next_sorted (fp->ctf_objthash, &i, &symname,
NULL, sort_fun, NULL)) == 0)
*walk++ = (const char *) symname;
if (err != ECTF_NEXT_END)
goto symerr;
while ((err = ctf_dynhash_next_sorted (fp->ctf_funchash, &i, &symname,
NULL, sort_fun, NULL)) == 0)
*walk++ = (const char *) symname;
if (err != ECTF_NEXT_END)
goto symerr;
}
}
/* Emit the object and function sections, and if necessary their indexes.
Emission is done in symtab order if there is no index, and in index
(name) order otherwise. */
if ((objtidx_size == 0) && s->symfp && s->symfp->ctf_dynsymidx)
{
ctf_dprintf ("Emitting unindexed objt symtypetab\n");
if (emit_symtypetab (fp, s->symfp, (uint32_t *) t,
s->symfp->ctf_dynsymidx, NULL,
s->symfp->ctf_dynsymmax + 1, s->maxobjt,
objt_size, s->symflags | CTF_SYMTYPETAB_EMIT_PAD) < 0)
goto err; /* errno is set for us. */
}
else
{
ctf_dprintf ("Emitting indexed objt symtypetab\n");
if (emit_symtypetab (fp, s->symfp, (uint32_t *) t, NULL,
sym_name_order, nsymtypes, s->maxobjt,
objt_size, s->symflags) < 0)
goto err; /* errno is set for us. */
}
t += objt_size;
if ((funcidx_size == 0) && s->symfp && s->symfp->ctf_dynsymidx)
{
ctf_dprintf ("Emitting unindexed func symtypetab\n");
if (emit_symtypetab (fp, s->symfp, (uint32_t *) t,
s->symfp->ctf_dynsymidx, NULL,
s->symfp->ctf_dynsymmax + 1, s->maxfunc,
func_size, s->symflags | CTF_SYMTYPETAB_EMIT_FUNCTION
| CTF_SYMTYPETAB_EMIT_PAD) < 0)
goto err; /* errno is set for us. */
}
else
{
ctf_dprintf ("Emitting indexed func symtypetab\n");
if (emit_symtypetab (fp, s->symfp, (uint32_t *) t, NULL, sym_name_order,
nsymtypes, s->maxfunc, func_size,
s->symflags | CTF_SYMTYPETAB_EMIT_FUNCTION) < 0)
goto err; /* errno is set for us. */
}
t += func_size;
if (objtidx_size > 0)
if (emit_symtypetab_index (fp, s->symfp, (uint32_t *) t, sym_name_order,
nsymtypes, objtidx_size, s->symflags) < 0)
goto err;
t += objtidx_size;
if (funcidx_size > 0)
if (emit_symtypetab_index (fp, s->symfp, (uint32_t *) t, sym_name_order,
nsymtypes, funcidx_size,
s->symflags | CTF_SYMTYPETAB_EMIT_FUNCTION) < 0)
goto err;
t += funcidx_size;
free (sym_name_order);
*tptr = t;
return 0;
oom:
ctf_set_errno (fp, EAGAIN);
goto err;
symerr:
ctf_err_warn (fp, 0, err, _("error serializing symtypetabs"));
err:
free (sym_name_order);
return -1;
}
/* Type section. */
/* Iterate through the dynamic type definition list and compute the
size of the CTF type section. */
static size_t
ctf_type_sect_size (ctf_dict_t *fp)
{
ctf_dtdef_t *dtd;
size_t type_size;
for (type_size = 0, dtd = ctf_list_next (&fp->ctf_dtdefs);
dtd != NULL; dtd = ctf_list_next (dtd))
{
uint32_t kind = LCTF_INFO_KIND (fp, dtd->dtd_data.ctt_info);
uint32_t vlen = LCTF_INFO_VLEN (fp, dtd->dtd_data.ctt_info);
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
size_t type_ctt_size = dtd->dtd_data.ctt_size;
/* Shrink ctf_type_t-using types from a ctf_type_t to a ctf_stype_t
if possible. */
if (kind == CTF_K_STRUCT || kind == CTF_K_UNION)
{
size_t lsize = CTF_TYPE_LSIZE (&dtd->dtd_data);
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
if (lsize <= CTF_MAX_SIZE)
type_ctt_size = lsize;
}
if (type_ctt_size != CTF_LSIZE_SENT)
type_size += sizeof (ctf_stype_t);
else
type_size += sizeof (ctf_type_t);
switch (kind)
{
case CTF_K_INTEGER:
case CTF_K_FLOAT:
type_size += sizeof (uint32_t);
break;
case CTF_K_ARRAY:
type_size += sizeof (ctf_array_t);
break;
case CTF_K_SLICE:
type_size += sizeof (ctf_slice_t);
break;
case CTF_K_FUNCTION:
type_size += sizeof (uint32_t) * (vlen + (vlen & 1));
break;
case CTF_K_STRUCT:
case CTF_K_UNION:
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
if (type_ctt_size < CTF_LSTRUCT_THRESH)
type_size += sizeof (ctf_member_t) * vlen;
else
type_size += sizeof (ctf_lmember_t) * vlen;
break;
case CTF_K_ENUM:
type_size += sizeof (ctf_enum_t) * vlen;
break;
}
}
return type_size;
}
/* Take a final lap through the dynamic type definition list and copy the
appropriate type records to the output buffer, noting down the strings as
we go. */
static void
ctf_emit_type_sect (ctf_dict_t *fp, unsigned char **tptr)
{
unsigned char *t = *tptr;
ctf_dtdef_t *dtd;
for (dtd = ctf_list_next (&fp->ctf_dtdefs);
dtd != NULL; dtd = ctf_list_next (dtd))
{
uint32_t kind = LCTF_INFO_KIND (fp, dtd->dtd_data.ctt_info);
uint32_t vlen = LCTF_INFO_VLEN (fp, dtd->dtd_data.ctt_info);
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
size_t type_ctt_size = dtd->dtd_data.ctt_size;
size_t len;
ctf_stype_t *copied;
const char *name;
libctf: eliminate dtd_u, part 4: enums This is the first tricky one, the first complex multi-entry vlen containing strings. To handle this in vlen form, we have to handle pending refs moving around on realloc. We grow vlen regions using a new ctf_grow_vlen function, and iterate through the existing enums every time a grow happens, telling the string machinery the distance between the old and new vlen region and letting it adjust the pending refs accordingly. (This avoids traversing all outstanding refs to find the refs that need adjusting, at the cost of having to traverse one enum: an obvious major performance win.) Addition of enums themselves (and also structs/unions later) is a bit trickier than earlier forms, because the type might be being promoted from a forward, and forwards have no vlen: so we have to spot that and create it if needed. Serialization of enums simplifies down to just telling the string machinery about the string refs; all the enum type-lookup code loses all its dynamic member lookup complexity entirely. A new test is added that iterates over (and gets values of) an enum with enough members to force a round of vlen growth. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtd_vlen_alloc>: New. (ctf_str_move_pending): Declare. * ctf-string.c (ctf_str_add_ref_internal): Fix error return. (ctf_str_move_pending): New. * ctf-create.c (ctf_grow_vlen): New. (ctf_dtd_delete): Zero out the vlen_alloc after free. Free the vlen later: iterate over it and free enum name refs first. (ctf_add_generic): Populate dtd_vlen_alloc from vlen. (ctf_add_enum): populate the vlen; do it by hand if promoting forwards. (ctf_add_enumerator): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the enumerand names as pending strings. * ctf-serialize.c (ctf_copy_emembers): Remove. (ctf_emit_type_sect): Copy the vlen into place and ref the strings. * ctf-types.c (ctf_enum_next): The dynamic portion now uses the same code as the non-dynamic. (ctf_enum_name): Likewise. (ctf_enum_value): Likewise. * testsuite/libctf-lookup/enum-many-ctf.c: New test. * testsuite/libctf-lookup/enum-many.lk: New test.
2021-03-18 20:37:52 +08:00
size_t i;
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
/* Shrink ctf_type_t-using types from a ctf_type_t to a ctf_stype_t
if possible. */
if (kind == CTF_K_STRUCT || kind == CTF_K_UNION)
{
size_t lsize = CTF_TYPE_LSIZE (&dtd->dtd_data);
if (lsize <= CTF_MAX_SIZE)
type_ctt_size = lsize;
}
if (type_ctt_size != CTF_LSIZE_SENT)
len = sizeof (ctf_stype_t);
else
len = sizeof (ctf_type_t);
memcpy (t, &dtd->dtd_data, len);
copied = (ctf_stype_t *) t; /* name is at the start: constant offset. */
if (copied->ctt_name
&& (name = ctf_strraw (fp, copied->ctt_name)) != NULL)
libctf: do not corrupt strings across ctf_serialize The preceding change revealed a new bug: the string table is sorted for better compression, so repeated serialization with type (or member) additions in the middle can move strings around. But every serialization flushes the set of refs (the memory locations that are automatically updated with a final string offset when the strtab is updated), so if we are not to have string offsets go stale, we must do all ref additions within the serialization code (which walks the complete set of types and symbols anyway). Unfortunately, we were adding one ref in another place: the type name in the dynamic type definitions, which has a ref added to it by ctf_add_generic. So adding a type, serializing (via, say, one of the ctf_write functions), adding another type with a name that sorts earlier, and serializing again will corrupt the name of the first type because it no longer had a ref pointing to its dtd entry's name when its string offset was shifted later in the strtab to mae way for the other type. To ensure that we don't miss strings, we also maintain a set of *pending refs* that will be added later (during serialization), and remove entries from that set when the ref is finally added. We always use ctf_str_add_pending outside ctf-serialize.c, ensure that ctf_serialize adds all strtab offsets as refs (even those in the dtds) on every serialization, and mandate that no refs are live on entry to ctf_serialize and that all pending refs are gone before strtab finalization. (Of necessity ctf_serialize has to traverse all strtab offsets in the dtds in order to serialize them, so adding them as refs at the same time is easy.) (Note that we still can't erase unused atoms when we roll back, though we can erase unused refs: members and enums are still not removed by rollbacks and might reference strings added after the snapshot.) libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-hash.c (ctf_dynset_elements): New. * ctf-impl.h (ctf_dynset_elements): Declare it. (ctf_str_add_pending): Likewise. (ctf_dict_t) <ctf_str_pending_ref>: New, set of refs that must be added during serialization. * ctf-string.c (ctf_str_create_atoms): Initialize it. (CTF_STR_ADD_REF): New flag. (CTF_STR_MAKE_PROVISIONAL): Likewise. (CTF_STR_PENDING_REF): Likewise. (ctf_str_add_ref_internal): Take a flags word rather than int params. Populate, and clear out, ctf_str_pending_ref. (ctf_str_add): Adjust accordingly. (ctf_str_add_external): Likewise. (ctf_str_add_pending): New. (ctf_str_remove_ref): Also remove the potential ref if it is a pending ref. * ctf-serialize.c (ctf_serialize): Prohibit addition of strings with ctf_str_add_ref before serialization. Ensure that the ctf_str_pending_ref set is empty before strtab finalization. (ctf_emit_type_sect): Add a ref to the ctt_name. * ctf-create.c (ctf_add_generic): Add the ctt_name as a pending ref. * testsuite/libctf-writable/reserialize-strtab-corruption.*: New test.
2021-03-18 20:37:52 +08:00
{
ctf_str_add_ref (fp, name, &copied->ctt_name);
ctf_str_add_ref (fp, name, &dtd->dtd_data.ctt_name);
}
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
copied->ctt_size = type_ctt_size;
t += len;
switch (kind)
{
case CTF_K_INTEGER:
case CTF_K_FLOAT:
libctf: eliminate dtd_u, part 1: int/float/slice This series eliminates a lot of special-case code to handle dynamic types (types added to writable dicts and not yet serialized). Historically, when such types have variable-length data in their final CTF representations, libctf has always worked by adding such types to a special union (ctf_dtdef_t.dtd_u) in the dynamic type definition structure, then picking the members out of this structure at serialization time and packing them into their final form. This has the advantage that the ctf_add_* code doesn't need to know anything about the final CTF representation, but the significant disadvantage that all code that looks up types in any way needs two code paths, one for dynamic types, one for all others. Historically libctf "handled" this by not supporting most type lookups on dynamic types at all until ctf_update was called to do a complete reserialization of the entire dict (it didn't emit an error, it just emitted wrong results). Since commit 676c3ecbad6e9c4, which eliminated ctf_update in favour of the internal-only ctf_serialize function, all the type-lookup paths grew an extra branch to handle dynamic types. We can eliminate this branch again by dropping the dtd_u stuff and simply writing out the vlen in (close to) its final form at ctf_add_* time: type lookup for types using this approach is then identical for types in writable dicts and types that are in read-only ones, and serialization is also simplified (we just need to write out the vlen we already created). The only complexity lies in type kinds for which multiple vlen representations are valid depending on properties of the type, e.g. structures. But we can start simple, adjusting ints, floats, and slices to work this way, and leaving everything else as is. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtd_u.dtu_enc>: Remove. <dtd_u.dtu_slice>: Likewise. <dtd_vlen>: New. * ctf-create.c (ctf_add_generic): Perhaps allocate it. All callers adjusted. (ctf_dtd_delete): Free it. (ctf_add_slice): Use the dtd_vlen, not dtu_enc. (ctf_add_encoded): Likewise. Assert that this must be an int or float. * ctf-serialize.c (ctf_emit_type_sect): Just copy the dtd_vlen. * ctf-dedup.c (ctf_dedup_rhash_type): Use the dtd_vlen, not dtu_slice. * ctf-types.c (ctf_type_reference): Likewise. (ctf_type_encoding): Remove most dynamic-type-specific code: just get the vlen from the right place. Report failure to look up the underlying type's encoding.
2021-03-18 20:37:52 +08:00
memcpy (t, dtd->dtd_vlen, sizeof (uint32_t));
t += sizeof (uint32_t);
break;
case CTF_K_SLICE:
libctf: eliminate dtd_u, part 1: int/float/slice This series eliminates a lot of special-case code to handle dynamic types (types added to writable dicts and not yet serialized). Historically, when such types have variable-length data in their final CTF representations, libctf has always worked by adding such types to a special union (ctf_dtdef_t.dtd_u) in the dynamic type definition structure, then picking the members out of this structure at serialization time and packing them into their final form. This has the advantage that the ctf_add_* code doesn't need to know anything about the final CTF representation, but the significant disadvantage that all code that looks up types in any way needs two code paths, one for dynamic types, one for all others. Historically libctf "handled" this by not supporting most type lookups on dynamic types at all until ctf_update was called to do a complete reserialization of the entire dict (it didn't emit an error, it just emitted wrong results). Since commit 676c3ecbad6e9c4, which eliminated ctf_update in favour of the internal-only ctf_serialize function, all the type-lookup paths grew an extra branch to handle dynamic types. We can eliminate this branch again by dropping the dtd_u stuff and simply writing out the vlen in (close to) its final form at ctf_add_* time: type lookup for types using this approach is then identical for types in writable dicts and types that are in read-only ones, and serialization is also simplified (we just need to write out the vlen we already created). The only complexity lies in type kinds for which multiple vlen representations are valid depending on properties of the type, e.g. structures. But we can start simple, adjusting ints, floats, and slices to work this way, and leaving everything else as is. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtd_u.dtu_enc>: Remove. <dtd_u.dtu_slice>: Likewise. <dtd_vlen>: New. * ctf-create.c (ctf_add_generic): Perhaps allocate it. All callers adjusted. (ctf_dtd_delete): Free it. (ctf_add_slice): Use the dtd_vlen, not dtu_enc. (ctf_add_encoded): Likewise. Assert that this must be an int or float. * ctf-serialize.c (ctf_emit_type_sect): Just copy the dtd_vlen. * ctf-dedup.c (ctf_dedup_rhash_type): Use the dtd_vlen, not dtu_slice. * ctf-types.c (ctf_type_reference): Likewise. (ctf_type_encoding): Remove most dynamic-type-specific code: just get the vlen from the right place. Report failure to look up the underlying type's encoding.
2021-03-18 20:37:52 +08:00
memcpy (t, dtd->dtd_vlen, sizeof (struct ctf_slice));
t += sizeof (struct ctf_slice);
break;
case CTF_K_ARRAY:
memcpy (t, dtd->dtd_vlen, sizeof (struct ctf_array));
t += sizeof (struct ctf_array);
break;
case CTF_K_FUNCTION:
/* Functions with no args also have no vlen. */
if (dtd->dtd_vlen)
memcpy (t, dtd->dtd_vlen, sizeof (uint32_t) * (vlen + (vlen & 1)));
t += sizeof (uint32_t) * (vlen + (vlen & 1));
break;
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
/* These need to be copied across element by element, depending on
their ctt_size. */
case CTF_K_STRUCT:
case CTF_K_UNION:
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
{
ctf_lmember_t *dtd_vlen = (ctf_lmember_t *) dtd->dtd_vlen;
ctf_lmember_t *t_lvlen = (ctf_lmember_t *) t;
ctf_member_t *t_vlen = (ctf_member_t *) t;
for (i = 0; i < vlen; i++)
{
const char *name = ctf_strraw (fp, dtd_vlen[i].ctlm_name);
ctf_str_add_ref (fp, name, &dtd_vlen[i].ctlm_name);
if (type_ctt_size < CTF_LSTRUCT_THRESH)
{
t_vlen[i].ctm_name = dtd_vlen[i].ctlm_name;
t_vlen[i].ctm_type = dtd_vlen[i].ctlm_type;
t_vlen[i].ctm_offset = CTF_LMEM_OFFSET (&dtd_vlen[i]);
ctf_str_add_ref (fp, name, &t_vlen[i].ctm_name);
}
else
{
t_lvlen[i] = dtd_vlen[i];
ctf_str_add_ref (fp, name, &t_lvlen[i].ctlm_name);
}
}
}
if (type_ctt_size < CTF_LSTRUCT_THRESH)
t += sizeof (ctf_member_t) * vlen;
else
libctf: eliminate dtd_u, part 5: structs / unions Eliminate the dynamic member storage for structs and unions as we have for other dynamic types. This is much like the previous enum elimination, except that structs and unions are the only types for which a full-sized ctf_type_t might be needed. Up to now, this decision has been made in the individual ctf_add_{struct,union}_sized functions and duplicated in ctf_add_member_offset. The vlen machinery lets us simplify this, always allocating a ctf_lmember_t and setting the dtd_data's ctt_size to CTF_LSIZE_SENT: we figure out whether this is really justified and (almost always) repack things down into a ctf_stype_t at ctf_serialize time. This allows us to eliminate the dynamic member paths from the iterators and query functions in ctf-types.c in favour of always using the large-structure vlen stuff for dynamic types (the diff is ugly but that's just because of the volume of reindentation this calls for). This also means the large-structure vlen stuff gets more heavily tested, which is nice because it was an almost totally unused code path before now (it only kicked in for structures of size >4GiB, and how often do you see those?) The only extra complexity here is ctf_add_type. Back in the days of the nondeduplicating linker this was called a ridiculous number of times for countless identical copies of structures: eschewing the repeated lookups of the dtd in ctf_add_member_offset and adding the members directly saved an amazing amount of time. Now the nondeduplicating linker is gone, this is extreme overoptimization: we can rip out the direct addition and use ctf_member_next and ctf_add_member_offset, just like ctf_dedup_emit does. We augment a ctf_add_type test to try adding a self-referential struct, the only thing the ctf_add_type part of this change really perturbs. This completes the elimination of dtd_u. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtu_members>: Remove. <dtd_u>: Likewise. (ctf_dmdef_t): Remove. (struct ctf_next) <u.ctn_dmd>: Remove. * ctf-create.c (INITIAL_VLEN): New, more-or-less arbitrary initial vlen size. (ctf_add_enum): Use it. (ctf_dtd_delete): Do not free the (removed) dmd; remove string refs from the vlen on struct deletion. (ctf_add_struct_sized): Populate the vlen: do it by hand if promoting forwards. Always populate the full-size lsizehi/lsizelo members. (ctf_add_union_sized): Likewise. (ctf_add_member_offset): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the member names as pending strings. Always populate the full-size lsizehi/lsizelo members. (membadd): Remove, folding back into... (ctf_add_type_internal): ... here, adding via an ordinary ctf_add_struct_sized and _next iteration rather than doing everything by hand. * ctf-serialize.c (ctf_copy_smembers): Remove this... (ctf_copy_lmembers): ... and this... (ctf_emit_type_sect): ... folding into here. Figure out if a ctf_stype_t is needed here, not in ctf_add_*_sized. (ctf_type_sect_size): Figure out the ctf_stype_t stuff the same way here. * ctf-types.c (ctf_member_next): Remove the dmd path and always use the vlen. Force large-structure usage for dynamic types. (ctf_type_align): Likewise. (ctf_member_info): Likewise. (ctf_type_rvisit): Likewise. * testsuite/libctf-regression/type-add-unnamed-struct-ctf.c: Add a self-referential type to this test. * testsuite/libctf-regression/type-add-unnamed-struct.c: Adjusted accordingly. * testsuite/libctf-regression/type-add-unnamed-struct.lk: Likewise.
2021-03-18 20:37:52 +08:00
t += sizeof (ctf_lmember_t) * vlen;
break;
case CTF_K_ENUM:
libctf: eliminate dtd_u, part 4: enums This is the first tricky one, the first complex multi-entry vlen containing strings. To handle this in vlen form, we have to handle pending refs moving around on realloc. We grow vlen regions using a new ctf_grow_vlen function, and iterate through the existing enums every time a grow happens, telling the string machinery the distance between the old and new vlen region and letting it adjust the pending refs accordingly. (This avoids traversing all outstanding refs to find the refs that need adjusting, at the cost of having to traverse one enum: an obvious major performance win.) Addition of enums themselves (and also structs/unions later) is a bit trickier than earlier forms, because the type might be being promoted from a forward, and forwards have no vlen: so we have to spot that and create it if needed. Serialization of enums simplifies down to just telling the string machinery about the string refs; all the enum type-lookup code loses all its dynamic member lookup complexity entirely. A new test is added that iterates over (and gets values of) an enum with enough members to force a round of vlen growth. libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-impl.h (ctf_dtdef_t) <dtd_vlen_alloc>: New. (ctf_str_move_pending): Declare. * ctf-string.c (ctf_str_add_ref_internal): Fix error return. (ctf_str_move_pending): New. * ctf-create.c (ctf_grow_vlen): New. (ctf_dtd_delete): Zero out the vlen_alloc after free. Free the vlen later: iterate over it and free enum name refs first. (ctf_add_generic): Populate dtd_vlen_alloc from vlen. (ctf_add_enum): populate the vlen; do it by hand if promoting forwards. (ctf_add_enumerator): Set up the vlen rather than the dmd. Expand it as needed, repointing string refs via ctf_str_move_pending. Add the enumerand names as pending strings. * ctf-serialize.c (ctf_copy_emembers): Remove. (ctf_emit_type_sect): Copy the vlen into place and ref the strings. * ctf-types.c (ctf_enum_next): The dynamic portion now uses the same code as the non-dynamic. (ctf_enum_name): Likewise. (ctf_enum_value): Likewise. * testsuite/libctf-lookup/enum-many-ctf.c: New test. * testsuite/libctf-lookup/enum-many.lk: New test.
2021-03-18 20:37:52 +08:00
{
ctf_enum_t *dtd_vlen = (struct ctf_enum *) dtd->dtd_vlen;
ctf_enum_t *t_vlen = (struct ctf_enum *) t;
memcpy (t, dtd->dtd_vlen, sizeof (struct ctf_enum) * vlen);
for (i = 0; i < vlen; i++)
{
const char *name = ctf_strraw (fp, dtd_vlen[i].cte_name);
ctf_str_add_ref (fp, name, &t_vlen[i].cte_name);
ctf_str_add_ref (fp, name, &dtd_vlen[i].cte_name);
}
t += sizeof (struct ctf_enum) * vlen;
break;
}
}
}
*tptr = t;
}
/* Variable section. */
/* Sort a newly-constructed static variable array. */
typedef struct ctf_sort_var_arg_cb
{
ctf_dict_t *fp;
ctf_strs_t *strtab;
} ctf_sort_var_arg_cb_t;
static int
ctf_sort_var (const void *one_, const void *two_, void *arg_)
{
const ctf_varent_t *one = one_;
const ctf_varent_t *two = two_;
ctf_sort_var_arg_cb_t *arg = arg_;
return (strcmp (ctf_strraw_explicit (arg->fp, one->ctv_name, arg->strtab),
ctf_strraw_explicit (arg->fp, two->ctv_name, arg->strtab)));
}
/* Overall serialization. */
/* If the specified CTF dict is writable and has been modified, reload this dict
with the updated type definitions, ready for serialization. In order to make
this code and the rest of libctf as simple as possible, we perform updates by
taking the dynamic type definitions and creating an in-memory CTF dict
containing the definitions, and then call ctf_simple_open_internal() on it.
We perform one extra trick here for the benefit of callers and to keep our
code simple: ctf_simple_open_internal() will return a new ctf_dict_t, but we
want to keep the fp constant for the caller, so after
ctf_simple_open_internal() returns, we use memcpy to swap the interior of the
old and new ctf_dict_t's, and then free the old. */
int
ctf_serialize (ctf_dict_t *fp)
{
ctf_dict_t ofp, *nfp;
ctf_header_t hdr, *hdrp;
ctf_dvdef_t *dvd;
ctf_varent_t *dvarents;
ctf_strs_writable_t strtab;
int err;
libctf: do not corrupt strings across ctf_serialize The preceding change revealed a new bug: the string table is sorted for better compression, so repeated serialization with type (or member) additions in the middle can move strings around. But every serialization flushes the set of refs (the memory locations that are automatically updated with a final string offset when the strtab is updated), so if we are not to have string offsets go stale, we must do all ref additions within the serialization code (which walks the complete set of types and symbols anyway). Unfortunately, we were adding one ref in another place: the type name in the dynamic type definitions, which has a ref added to it by ctf_add_generic. So adding a type, serializing (via, say, one of the ctf_write functions), adding another type with a name that sorts earlier, and serializing again will corrupt the name of the first type because it no longer had a ref pointing to its dtd entry's name when its string offset was shifted later in the strtab to mae way for the other type. To ensure that we don't miss strings, we also maintain a set of *pending refs* that will be added later (during serialization), and remove entries from that set when the ref is finally added. We always use ctf_str_add_pending outside ctf-serialize.c, ensure that ctf_serialize adds all strtab offsets as refs (even those in the dtds) on every serialization, and mandate that no refs are live on entry to ctf_serialize and that all pending refs are gone before strtab finalization. (Of necessity ctf_serialize has to traverse all strtab offsets in the dtds in order to serialize them, so adding them as refs at the same time is easy.) (Note that we still can't erase unused atoms when we roll back, though we can erase unused refs: members and enums are still not removed by rollbacks and might reference strings added after the snapshot.) libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-hash.c (ctf_dynset_elements): New. * ctf-impl.h (ctf_dynset_elements): Declare it. (ctf_str_add_pending): Likewise. (ctf_dict_t) <ctf_str_pending_ref>: New, set of refs that must be added during serialization. * ctf-string.c (ctf_str_create_atoms): Initialize it. (CTF_STR_ADD_REF): New flag. (CTF_STR_MAKE_PROVISIONAL): Likewise. (CTF_STR_PENDING_REF): Likewise. (ctf_str_add_ref_internal): Take a flags word rather than int params. Populate, and clear out, ctf_str_pending_ref. (ctf_str_add): Adjust accordingly. (ctf_str_add_external): Likewise. (ctf_str_add_pending): New. (ctf_str_remove_ref): Also remove the potential ref if it is a pending ref. * ctf-serialize.c (ctf_serialize): Prohibit addition of strings with ctf_str_add_ref before serialization. Ensure that the ctf_str_pending_ref set is empty before strtab finalization. (ctf_emit_type_sect): Add a ref to the ctt_name. * ctf-create.c (ctf_add_generic): Add the ctt_name as a pending ref. * testsuite/libctf-writable/reserialize-strtab-corruption.*: New test.
2021-03-18 20:37:52 +08:00
int num_missed_str_refs;
unsigned char *t;
unsigned long i;
size_t buf_size, type_size, objt_size, func_size;
size_t funcidx_size, objtidx_size;
size_t nvars;
unsigned char *buf = NULL, *newbuf;
emit_symtypetab_state_t symstate;
memset (&symstate, 0, sizeof (emit_symtypetab_state_t));
if (!(fp->ctf_flags & LCTF_RDWR))
return (ctf_set_errno (fp, ECTF_RDONLY));
/* Update required? */
if (!(fp->ctf_flags & LCTF_DIRTY))
return 0;
libctf: do not corrupt strings across ctf_serialize The preceding change revealed a new bug: the string table is sorted for better compression, so repeated serialization with type (or member) additions in the middle can move strings around. But every serialization flushes the set of refs (the memory locations that are automatically updated with a final string offset when the strtab is updated), so if we are not to have string offsets go stale, we must do all ref additions within the serialization code (which walks the complete set of types and symbols anyway). Unfortunately, we were adding one ref in another place: the type name in the dynamic type definitions, which has a ref added to it by ctf_add_generic. So adding a type, serializing (via, say, one of the ctf_write functions), adding another type with a name that sorts earlier, and serializing again will corrupt the name of the first type because it no longer had a ref pointing to its dtd entry's name when its string offset was shifted later in the strtab to mae way for the other type. To ensure that we don't miss strings, we also maintain a set of *pending refs* that will be added later (during serialization), and remove entries from that set when the ref is finally added. We always use ctf_str_add_pending outside ctf-serialize.c, ensure that ctf_serialize adds all strtab offsets as refs (even those in the dtds) on every serialization, and mandate that no refs are live on entry to ctf_serialize and that all pending refs are gone before strtab finalization. (Of necessity ctf_serialize has to traverse all strtab offsets in the dtds in order to serialize them, so adding them as refs at the same time is easy.) (Note that we still can't erase unused atoms when we roll back, though we can erase unused refs: members and enums are still not removed by rollbacks and might reference strings added after the snapshot.) libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-hash.c (ctf_dynset_elements): New. * ctf-impl.h (ctf_dynset_elements): Declare it. (ctf_str_add_pending): Likewise. (ctf_dict_t) <ctf_str_pending_ref>: New, set of refs that must be added during serialization. * ctf-string.c (ctf_str_create_atoms): Initialize it. (CTF_STR_ADD_REF): New flag. (CTF_STR_MAKE_PROVISIONAL): Likewise. (CTF_STR_PENDING_REF): Likewise. (ctf_str_add_ref_internal): Take a flags word rather than int params. Populate, and clear out, ctf_str_pending_ref. (ctf_str_add): Adjust accordingly. (ctf_str_add_external): Likewise. (ctf_str_add_pending): New. (ctf_str_remove_ref): Also remove the potential ref if it is a pending ref. * ctf-serialize.c (ctf_serialize): Prohibit addition of strings with ctf_str_add_ref before serialization. Ensure that the ctf_str_pending_ref set is empty before strtab finalization. (ctf_emit_type_sect): Add a ref to the ctt_name. * ctf-create.c (ctf_add_generic): Add the ctt_name as a pending ref. * testsuite/libctf-writable/reserialize-strtab-corruption.*: New test.
2021-03-18 20:37:52 +08:00
/* The strtab refs table must be empty at this stage. Any refs already added
will be corrupted by any modifications, including reserialization, after
strtab finalization is complete. Only this function, and functions it
calls, may add refs, and all memory locations (including in the dtds)
containing strtab offsets must be traversed as part of serialization, and
refs added. */
if (!ctf_assert (fp, fp->ctf_str_num_refs == 0))
return -1; /* errno is set for us. */
/* Fill in an initial CTF header. We will leave the label, object,
and function sections empty and only output a header, type section,
and string table. The type section begins at a 4-byte aligned
boundary past the CTF header itself (at relative offset zero). The flag
indicating a new-style function info section (an array of CTF_K_FUNCTION
type IDs in the types section) is flipped on. */
memset (&hdr, 0, sizeof (hdr));
hdr.cth_magic = CTF_MAGIC;
hdr.cth_version = CTF_VERSION;
/* This is a new-format func info section, and the symtab and strtab come out
of the dynsym and dynstr these days. */
hdr.cth_flags = (CTF_F_NEWFUNCINFO | CTF_F_DYNSTR);
if (ctf_symtypetab_sect_sizes (fp, &symstate, &hdr, &objt_size, &func_size,
&objtidx_size, &funcidx_size) < 0)
return -1; /* errno is set for us. */
for (nvars = 0, dvd = ctf_list_next (&fp->ctf_dvdefs);
dvd != NULL; dvd = ctf_list_next (dvd), nvars++);
type_size = ctf_type_sect_size (fp);
/* Compute the size of the CTF buffer we need, sans only the string table,
then allocate a new buffer and memcpy the finished header to the start of
the buffer. (We will adjust this later with strtab length info.) */
hdr.cth_lbloff = hdr.cth_objtoff = 0;
hdr.cth_funcoff = hdr.cth_objtoff + objt_size;
hdr.cth_objtidxoff = hdr.cth_funcoff + func_size;
hdr.cth_funcidxoff = hdr.cth_objtidxoff + objtidx_size;
hdr.cth_varoff = hdr.cth_funcidxoff + funcidx_size;
hdr.cth_typeoff = hdr.cth_varoff + (nvars * sizeof (ctf_varent_t));
hdr.cth_stroff = hdr.cth_typeoff + type_size;
hdr.cth_strlen = 0;
buf_size = sizeof (ctf_header_t) + hdr.cth_stroff + hdr.cth_strlen;
if ((buf = malloc (buf_size)) == NULL)
return (ctf_set_errno (fp, EAGAIN));
memcpy (buf, &hdr, sizeof (ctf_header_t));
t = (unsigned char *) buf + sizeof (ctf_header_t) + hdr.cth_objtoff;
hdrp = (ctf_header_t *) buf;
if ((fp->ctf_flags & LCTF_CHILD) && (fp->ctf_parname != NULL))
ctf_str_add_ref (fp, fp->ctf_parname, &hdrp->cth_parname);
if (fp->ctf_cuname != NULL)
ctf_str_add_ref (fp, fp->ctf_cuname, &hdrp->cth_cuname);
if (ctf_emit_symtypetab_sects (fp, &symstate, &t, objt_size, func_size,
objtidx_size, funcidx_size) < 0)
goto err;
assert (t == (unsigned char *) buf + sizeof (ctf_header_t) + hdr.cth_varoff);
/* Work over the variable list, translating everything into ctf_varent_t's and
prepping the string table. */
dvarents = (ctf_varent_t *) t;
for (i = 0, dvd = ctf_list_next (&fp->ctf_dvdefs); dvd != NULL;
dvd = ctf_list_next (dvd), i++)
{
ctf_varent_t *var = &dvarents[i];
ctf_str_add_ref (fp, dvd->dvd_name, &var->ctv_name);
var->ctv_type = (uint32_t) dvd->dvd_type;
}
assert (i == nvars);
t += sizeof (ctf_varent_t) * nvars;
assert (t == (unsigned char *) buf + sizeof (ctf_header_t) + hdr.cth_typeoff);
ctf_emit_type_sect (fp, &t);
assert (t == (unsigned char *) buf + sizeof (ctf_header_t) + hdr.cth_stroff);
libctf: do not corrupt strings across ctf_serialize The preceding change revealed a new bug: the string table is sorted for better compression, so repeated serialization with type (or member) additions in the middle can move strings around. But every serialization flushes the set of refs (the memory locations that are automatically updated with a final string offset when the strtab is updated), so if we are not to have string offsets go stale, we must do all ref additions within the serialization code (which walks the complete set of types and symbols anyway). Unfortunately, we were adding one ref in another place: the type name in the dynamic type definitions, which has a ref added to it by ctf_add_generic. So adding a type, serializing (via, say, one of the ctf_write functions), adding another type with a name that sorts earlier, and serializing again will corrupt the name of the first type because it no longer had a ref pointing to its dtd entry's name when its string offset was shifted later in the strtab to mae way for the other type. To ensure that we don't miss strings, we also maintain a set of *pending refs* that will be added later (during serialization), and remove entries from that set when the ref is finally added. We always use ctf_str_add_pending outside ctf-serialize.c, ensure that ctf_serialize adds all strtab offsets as refs (even those in the dtds) on every serialization, and mandate that no refs are live on entry to ctf_serialize and that all pending refs are gone before strtab finalization. (Of necessity ctf_serialize has to traverse all strtab offsets in the dtds in order to serialize them, so adding them as refs at the same time is easy.) (Note that we still can't erase unused atoms when we roll back, though we can erase unused refs: members and enums are still not removed by rollbacks and might reference strings added after the snapshot.) libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-hash.c (ctf_dynset_elements): New. * ctf-impl.h (ctf_dynset_elements): Declare it. (ctf_str_add_pending): Likewise. (ctf_dict_t) <ctf_str_pending_ref>: New, set of refs that must be added during serialization. * ctf-string.c (ctf_str_create_atoms): Initialize it. (CTF_STR_ADD_REF): New flag. (CTF_STR_MAKE_PROVISIONAL): Likewise. (CTF_STR_PENDING_REF): Likewise. (ctf_str_add_ref_internal): Take a flags word rather than int params. Populate, and clear out, ctf_str_pending_ref. (ctf_str_add): Adjust accordingly. (ctf_str_add_external): Likewise. (ctf_str_add_pending): New. (ctf_str_remove_ref): Also remove the potential ref if it is a pending ref. * ctf-serialize.c (ctf_serialize): Prohibit addition of strings with ctf_str_add_ref before serialization. Ensure that the ctf_str_pending_ref set is empty before strtab finalization. (ctf_emit_type_sect): Add a ref to the ctt_name. * ctf-create.c (ctf_add_generic): Add the ctt_name as a pending ref. * testsuite/libctf-writable/reserialize-strtab-corruption.*: New test.
2021-03-18 20:37:52 +08:00
/* Every string added outside serialization by ctf_str_add_pending should
now have been added by ctf_add_ref. */
num_missed_str_refs = ctf_dynset_elements (fp->ctf_str_pending_ref);
if (!ctf_assert (fp, num_missed_str_refs == 0))
goto err; /* errno is set for us. */
/* Construct the final string table and fill out all the string refs with the
final offsets. Then purge the refs list, because we're about to move this
strtab onto the end of the buf, invalidating all the offsets. */
strtab = ctf_str_write_strtab (fp);
ctf_str_purge_refs (fp);
if (strtab.cts_strs == NULL)
goto oom;
/* Now the string table is constructed, we can sort the buffer of
ctf_varent_t's. */
ctf_sort_var_arg_cb_t sort_var_arg = { fp, (ctf_strs_t *) &strtab };
ctf_qsort_r (dvarents, nvars, sizeof (ctf_varent_t), ctf_sort_var,
&sort_var_arg);
if ((newbuf = ctf_realloc (fp, buf, buf_size + strtab.cts_len)) == NULL)
{
free (strtab.cts_strs);
goto oom;
}
buf = newbuf;
memcpy (buf + buf_size, strtab.cts_strs, strtab.cts_len);
hdrp = (ctf_header_t *) buf;
hdrp->cth_strlen = strtab.cts_len;
buf_size += hdrp->cth_strlen;
free (strtab.cts_strs);
/* Finally, we are ready to ctf_simple_open() the new dict. If this is
successful, we then switch nfp and fp and free the old dict. */
if ((nfp = ctf_simple_open_internal ((char *) buf, buf_size, NULL, 0,
0, NULL, 0, fp->ctf_syn_ext_strtab,
1, &err)) == NULL)
{
free (buf);
return (ctf_set_errno (fp, err));
}
(void) ctf_setmodel (nfp, ctf_getmodel (fp));
nfp->ctf_parent = fp->ctf_parent;
nfp->ctf_parent_unreffed = fp->ctf_parent_unreffed;
nfp->ctf_refcnt = fp->ctf_refcnt;
nfp->ctf_flags |= fp->ctf_flags & ~LCTF_DIRTY;
if (nfp->ctf_dynbase == NULL)
nfp->ctf_dynbase = buf; /* Make sure buf is freed on close. */
nfp->ctf_dthash = fp->ctf_dthash;
nfp->ctf_dtdefs = fp->ctf_dtdefs;
nfp->ctf_dvhash = fp->ctf_dvhash;
nfp->ctf_dvdefs = fp->ctf_dvdefs;
nfp->ctf_dtoldid = fp->ctf_dtoldid;
nfp->ctf_add_processing = fp->ctf_add_processing;
nfp->ctf_snapshots = fp->ctf_snapshots + 1;
nfp->ctf_specific = fp->ctf_specific;
nfp->ctf_nfuncidx = fp->ctf_nfuncidx;
nfp->ctf_nobjtidx = fp->ctf_nobjtidx;
nfp->ctf_objthash = fp->ctf_objthash;
nfp->ctf_funchash = fp->ctf_funchash;
nfp->ctf_dynsyms = fp->ctf_dynsyms;
nfp->ctf_ptrtab = fp->ctf_ptrtab;
nfp->ctf_pptrtab = fp->ctf_pptrtab;
nfp->ctf_typemax = fp->ctf_typemax;
nfp->ctf_dynsymidx = fp->ctf_dynsymidx;
nfp->ctf_dynsymmax = fp->ctf_dynsymmax;
nfp->ctf_ptrtab_len = fp->ctf_ptrtab_len;
nfp->ctf_pptrtab_len = fp->ctf_pptrtab_len;
nfp->ctf_link_inputs = fp->ctf_link_inputs;
nfp->ctf_link_outputs = fp->ctf_link_outputs;
nfp->ctf_errs_warnings = fp->ctf_errs_warnings;
nfp->ctf_funcidx_names = fp->ctf_funcidx_names;
nfp->ctf_objtidx_names = fp->ctf_objtidx_names;
nfp->ctf_funcidx_sxlate = fp->ctf_funcidx_sxlate;
nfp->ctf_objtidx_sxlate = fp->ctf_objtidx_sxlate;
nfp->ctf_str_prov_offset = fp->ctf_str_prov_offset;
nfp->ctf_syn_ext_strtab = fp->ctf_syn_ext_strtab;
nfp->ctf_pptrtab_typemax = fp->ctf_pptrtab_typemax;
nfp->ctf_in_flight_dynsyms = fp->ctf_in_flight_dynsyms;
nfp->ctf_link_in_cu_mapping = fp->ctf_link_in_cu_mapping;
nfp->ctf_link_out_cu_mapping = fp->ctf_link_out_cu_mapping;
nfp->ctf_link_type_mapping = fp->ctf_link_type_mapping;
nfp->ctf_link_memb_name_changer = fp->ctf_link_memb_name_changer;
nfp->ctf_link_memb_name_changer_arg = fp->ctf_link_memb_name_changer_arg;
nfp->ctf_link_variable_filter = fp->ctf_link_variable_filter;
nfp->ctf_link_variable_filter_arg = fp->ctf_link_variable_filter_arg;
nfp->ctf_symsect_little_endian = fp->ctf_symsect_little_endian;
nfp->ctf_link_flags = fp->ctf_link_flags;
nfp->ctf_dedup_atoms = fp->ctf_dedup_atoms;
nfp->ctf_dedup_atoms_alloc = fp->ctf_dedup_atoms_alloc;
memcpy (&nfp->ctf_dedup, &fp->ctf_dedup, sizeof (fp->ctf_dedup));
nfp->ctf_snapshot_lu = fp->ctf_snapshots;
memcpy (&nfp->ctf_lookups, fp->ctf_lookups, sizeof (fp->ctf_lookups));
nfp->ctf_structs = fp->ctf_structs;
nfp->ctf_unions = fp->ctf_unions;
nfp->ctf_enums = fp->ctf_enums;
nfp->ctf_names = fp->ctf_names;
fp->ctf_dthash = NULL;
ctf_str_free_atoms (nfp);
nfp->ctf_str_atoms = fp->ctf_str_atoms;
nfp->ctf_prov_strtab = fp->ctf_prov_strtab;
libctf: do not corrupt strings across ctf_serialize The preceding change revealed a new bug: the string table is sorted for better compression, so repeated serialization with type (or member) additions in the middle can move strings around. But every serialization flushes the set of refs (the memory locations that are automatically updated with a final string offset when the strtab is updated), so if we are not to have string offsets go stale, we must do all ref additions within the serialization code (which walks the complete set of types and symbols anyway). Unfortunately, we were adding one ref in another place: the type name in the dynamic type definitions, which has a ref added to it by ctf_add_generic. So adding a type, serializing (via, say, one of the ctf_write functions), adding another type with a name that sorts earlier, and serializing again will corrupt the name of the first type because it no longer had a ref pointing to its dtd entry's name when its string offset was shifted later in the strtab to mae way for the other type. To ensure that we don't miss strings, we also maintain a set of *pending refs* that will be added later (during serialization), and remove entries from that set when the ref is finally added. We always use ctf_str_add_pending outside ctf-serialize.c, ensure that ctf_serialize adds all strtab offsets as refs (even those in the dtds) on every serialization, and mandate that no refs are live on entry to ctf_serialize and that all pending refs are gone before strtab finalization. (Of necessity ctf_serialize has to traverse all strtab offsets in the dtds in order to serialize them, so adding them as refs at the same time is easy.) (Note that we still can't erase unused atoms when we roll back, though we can erase unused refs: members and enums are still not removed by rollbacks and might reference strings added after the snapshot.) libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-hash.c (ctf_dynset_elements): New. * ctf-impl.h (ctf_dynset_elements): Declare it. (ctf_str_add_pending): Likewise. (ctf_dict_t) <ctf_str_pending_ref>: New, set of refs that must be added during serialization. * ctf-string.c (ctf_str_create_atoms): Initialize it. (CTF_STR_ADD_REF): New flag. (CTF_STR_MAKE_PROVISIONAL): Likewise. (CTF_STR_PENDING_REF): Likewise. (ctf_str_add_ref_internal): Take a flags word rather than int params. Populate, and clear out, ctf_str_pending_ref. (ctf_str_add): Adjust accordingly. (ctf_str_add_external): Likewise. (ctf_str_add_pending): New. (ctf_str_remove_ref): Also remove the potential ref if it is a pending ref. * ctf-serialize.c (ctf_serialize): Prohibit addition of strings with ctf_str_add_ref before serialization. Ensure that the ctf_str_pending_ref set is empty before strtab finalization. (ctf_emit_type_sect): Add a ref to the ctt_name. * ctf-create.c (ctf_add_generic): Add the ctt_name as a pending ref. * testsuite/libctf-writable/reserialize-strtab-corruption.*: New test.
2021-03-18 20:37:52 +08:00
nfp->ctf_str_pending_ref = fp->ctf_str_pending_ref;
fp->ctf_str_atoms = NULL;
fp->ctf_prov_strtab = NULL;
libctf: do not corrupt strings across ctf_serialize The preceding change revealed a new bug: the string table is sorted for better compression, so repeated serialization with type (or member) additions in the middle can move strings around. But every serialization flushes the set of refs (the memory locations that are automatically updated with a final string offset when the strtab is updated), so if we are not to have string offsets go stale, we must do all ref additions within the serialization code (which walks the complete set of types and symbols anyway). Unfortunately, we were adding one ref in another place: the type name in the dynamic type definitions, which has a ref added to it by ctf_add_generic. So adding a type, serializing (via, say, one of the ctf_write functions), adding another type with a name that sorts earlier, and serializing again will corrupt the name of the first type because it no longer had a ref pointing to its dtd entry's name when its string offset was shifted later in the strtab to mae way for the other type. To ensure that we don't miss strings, we also maintain a set of *pending refs* that will be added later (during serialization), and remove entries from that set when the ref is finally added. We always use ctf_str_add_pending outside ctf-serialize.c, ensure that ctf_serialize adds all strtab offsets as refs (even those in the dtds) on every serialization, and mandate that no refs are live on entry to ctf_serialize and that all pending refs are gone before strtab finalization. (Of necessity ctf_serialize has to traverse all strtab offsets in the dtds in order to serialize them, so adding them as refs at the same time is easy.) (Note that we still can't erase unused atoms when we roll back, though we can erase unused refs: members and enums are still not removed by rollbacks and might reference strings added after the snapshot.) libctf/ChangeLog 2021-03-18 Nick Alcock <nick.alcock@oracle.com> * ctf-hash.c (ctf_dynset_elements): New. * ctf-impl.h (ctf_dynset_elements): Declare it. (ctf_str_add_pending): Likewise. (ctf_dict_t) <ctf_str_pending_ref>: New, set of refs that must be added during serialization. * ctf-string.c (ctf_str_create_atoms): Initialize it. (CTF_STR_ADD_REF): New flag. (CTF_STR_MAKE_PROVISIONAL): Likewise. (CTF_STR_PENDING_REF): Likewise. (ctf_str_add_ref_internal): Take a flags word rather than int params. Populate, and clear out, ctf_str_pending_ref. (ctf_str_add): Adjust accordingly. (ctf_str_add_external): Likewise. (ctf_str_add_pending): New. (ctf_str_remove_ref): Also remove the potential ref if it is a pending ref. * ctf-serialize.c (ctf_serialize): Prohibit addition of strings with ctf_str_add_ref before serialization. Ensure that the ctf_str_pending_ref set is empty before strtab finalization. (ctf_emit_type_sect): Add a ref to the ctt_name. * ctf-create.c (ctf_add_generic): Add the ctt_name as a pending ref. * testsuite/libctf-writable/reserialize-strtab-corruption.*: New test.
2021-03-18 20:37:52 +08:00
fp->ctf_str_pending_ref = NULL;
memset (&fp->ctf_dtdefs, 0, sizeof (ctf_list_t));
memset (&fp->ctf_errs_warnings, 0, sizeof (ctf_list_t));
fp->ctf_add_processing = NULL;
fp->ctf_ptrtab = NULL;
fp->ctf_pptrtab = NULL;
fp->ctf_funcidx_names = NULL;
fp->ctf_objtidx_names = NULL;
fp->ctf_funcidx_sxlate = NULL;
fp->ctf_objtidx_sxlate = NULL;
fp->ctf_objthash = NULL;
fp->ctf_funchash = NULL;
fp->ctf_dynsyms = NULL;
fp->ctf_dynsymidx = NULL;
fp->ctf_link_inputs = NULL;
fp->ctf_link_outputs = NULL;
fp->ctf_syn_ext_strtab = NULL;
fp->ctf_link_in_cu_mapping = NULL;
fp->ctf_link_out_cu_mapping = NULL;
fp->ctf_link_type_mapping = NULL;
fp->ctf_dedup_atoms = NULL;
fp->ctf_dedup_atoms_alloc = NULL;
fp->ctf_parent_unreffed = 1;
fp->ctf_dvhash = NULL;
memset (&fp->ctf_dvdefs, 0, sizeof (ctf_list_t));
memset (fp->ctf_lookups, 0, sizeof (fp->ctf_lookups));
memset (&fp->ctf_in_flight_dynsyms, 0, sizeof (fp->ctf_in_flight_dynsyms));
memset (&fp->ctf_dedup, 0, sizeof (fp->ctf_dedup));
fp->ctf_structs.ctn_writable = NULL;
fp->ctf_unions.ctn_writable = NULL;
fp->ctf_enums.ctn_writable = NULL;
fp->ctf_names.ctn_writable = NULL;
memcpy (&ofp, fp, sizeof (ctf_dict_t));
memcpy (fp, nfp, sizeof (ctf_dict_t));
memcpy (nfp, &ofp, sizeof (ctf_dict_t));
nfp->ctf_refcnt = 1; /* Force nfp to be freed. */
ctf_dict_close (nfp);
return 0;
oom:
free (buf);
return (ctf_set_errno (fp, EAGAIN));
err:
free (buf);
return -1; /* errno is set for us. */
}
/* File writing. */
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
/* Write the compressed CTF data stream to the specified gzFile descriptor. The
whole stream is compressed, and cannot be read by CTF opening functions in
this library until it is decompressed. (The functions below this one leave
the header uncompressed, and the CTF opening functions work on them without
manual decompression.)
No support for (testing-only) endian-flipping. */
int
ctf_gzwrite (ctf_dict_t *fp, gzFile fd)
{
const unsigned char *buf;
ssize_t resid;
ssize_t len;
resid = sizeof (ctf_header_t);
buf = (unsigned char *) fp->ctf_header;
while (resid != 0)
{
if ((len = gzwrite (fd, buf, resid)) <= 0)
return (ctf_set_errno (fp, errno));
resid -= len;
buf += len;
}
resid = fp->ctf_size;
buf = fp->ctf_buf;
while (resid != 0)
{
if ((len = gzwrite (fd, buf, resid)) <= 0)
return (ctf_set_errno (fp, errno));
resid -= len;
buf += len;
}
return 0;
}
/* Optionally compress the specified CTF data stream and return it as a new
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
dynamically-allocated string. Possibly write it with reversed
endianness. */
unsigned char *
ctf_write_mem (ctf_dict_t *fp, size_t *size, size_t threshold)
{
unsigned char *buf;
unsigned char *bp;
ctf_header_t *hp;
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
unsigned char *flipped, *src;
ssize_t header_len = sizeof (ctf_header_t);
ssize_t compress_len;
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
int flip_endian;
int uncompressed;
int rc;
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
flip_endian = getenv ("LIBCTF_WRITE_FOREIGN_ENDIAN") != NULL;
uncompressed = (fp->ctf_size < threshold);
if (ctf_serialize (fp) < 0)
return NULL; /* errno is set for us. */
compress_len = compressBound (fp->ctf_size);
if (fp->ctf_size < threshold)
compress_len = fp->ctf_size;
if ((buf = malloc (compress_len
+ sizeof (struct ctf_header))) == NULL)
{
ctf_set_errno (fp, ENOMEM);
ctf_err_warn (fp, 0, 0, _("ctf_write_mem: cannot allocate %li bytes"),
(unsigned long) (compress_len + sizeof (struct ctf_header)));
return NULL;
}
hp = (ctf_header_t *) buf;
memcpy (hp, fp->ctf_header, header_len);
bp = buf + sizeof (struct ctf_header);
*size = sizeof (struct ctf_header);
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
if (uncompressed)
hp->cth_flags &= ~CTF_F_COMPRESS;
else
hp->cth_flags |= CTF_F_COMPRESS;
src = fp->ctf_buf;
flipped = NULL;
if (flip_endian)
{
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
if ((flipped = malloc (fp->ctf_size)) == NULL)
{
ctf_set_errno (fp, ENOMEM);
ctf_err_warn (fp, 0, 0, _("ctf_write_mem: cannot allocate %li bytes"),
(unsigned long) fp->ctf_size + sizeof (struct ctf_header));
return NULL;
}
ctf_flip_header (hp);
memcpy (flipped, fp->ctf_buf, fp->ctf_size);
if (ctf_flip (fp, fp->ctf_header, flipped, 1) < 0)
{
free (buf);
free (flipped);
return NULL; /* errno is set for us. */
}
src = flipped;
}
if (uncompressed)
{
memcpy (bp, src, fp->ctf_size);
*size += fp->ctf_size;
}
else
{
if ((rc = compress (bp, (uLongf *) &compress_len,
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
src, fp->ctf_size)) != Z_OK)
{
ctf_set_errno (fp, ECTF_COMPRESS);
ctf_err_warn (fp, 0, 0, _("zlib deflate err: %s"), zError (rc));
free (buf);
return NULL;
}
*size += compress_len;
}
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
free (flipped);
return buf;
}
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
/* Compress the specified CTF data stream and write it to the specified file
descriptor. */
int
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
ctf_compress_write (ctf_dict_t *fp, int fd)
{
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
unsigned char *buf;
unsigned char *bp;
size_t tmp;
ssize_t buf_len;
ssize_t len;
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
int err = 0;
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
if ((buf = ctf_write_mem (fp, &tmp, 0)) == NULL)
return -1; /* errno is set for us. */
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
buf_len = tmp;
bp = buf;
while (buf_len > 0)
{
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
if ((len = write (fd, bp, buf_len)) < 0)
{
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
err = ctf_set_errno (fp, errno);
ctf_err_warn (fp, 0, 0, _("ctf_compress_write: error writing"));
goto ret;
}
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
buf_len -= len;
bp += len;
}
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
ret:
free (buf);
return err;
}
/* Write the uncompressed CTF data stream to the specified file descriptor. */
int
ctf_write (ctf_dict_t *fp, int fd)
{
unsigned char *buf;
unsigned char *bp;
size_t tmp;
ssize_t buf_len;
ssize_t len;
int err = 0;
if ((buf = ctf_write_mem (fp, &tmp, (size_t) -1)) == NULL)
return -1; /* errno is set for us. */
buf_len = tmp;
bp = buf;
while (buf_len > 0)
{
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
if ((len = write (fd, bp, buf_len)) < 0)
{
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
err = ctf_set_errno (fp, errno);
ctf_err_warn (fp, 0, 0, _("ctf_compress_write: error writing"));
goto ret;
}
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
buf_len -= len;
bp += len;
}
libctf: add LIBCTF_WRITE_FOREIGN_ENDIAN debugging option libctf has always handled endianness differences by detecting foreign-endian CTF dicts on the input and endian-flipping them: dicts are always written in native endianness. This makes endian-awareness very low overhead, but it means that the foreign-endian code paths almost never get routinely tested, since "make check" usually reads in dicts ld has just written out: only a few corrupted-CTF tests are actually in fixed endianness, and even they only test the foreign- endian code paths when you run make check on a big-endian machine. (And the fix is surely not to add more .s-based tests like that, because they are a nightmare to maintain compared to the C-code-based ones.) To improve on this, add a new environment variable, LIBCTF_WRITE_FOREIGN_ENDIAN, which causes libctf to unconditionally endian-flip at ctf_write time, so the output is always in the wrong endianness. This then tests the foreign-endian read paths properly at open time. Make this easier by restructuring the writeout code in ctf-serialize.c, which duplicates the maybe-gzip-and-write-out code three times (once for ctf_write_mem, with thresholding, and once each for ctf_compress_write and ctf_write just so those can avoid thresholding and/or compression). Instead, have the latter two call the former with thresholds of 0 or (size_t) -1, respectively. The endian-flipping code itself gains a bit of complexity, because one single endian-flipper (flip_types) was assuming the input to be in foreign-endian form and assuming it could pull things out of the input once they had been flipped and make sense of them. At the cost of a few lines of duplicated initializations, teach it to read before flipping if we're flipping to foreign-endianness instead of away from it. libctf/ * ctf-impl.h (ctf_flip_header): No longer static. (ctf_flip): Likewise. * ctf-open.c (flip_header): Rename to... (ctf_flip_header): ... this, now it is not private to one file. (flip_ctf): Rename... (ctf_flip): ... this too. Add FOREIGN_ENDIAN arg. (flip_types): Likewise. Use it. (ctf_bufopen_internal): Adjust calls. * ctf-serialize.c (ctf_write_mem): Add flip_endian path via a newly-allocated bounce buffer. (ctf_compress_write): Move below ctf_write_mem and reimplement in terms of it. (ctf_write): Likewise. (ctf_gzwrite): Note that this obscure writeout function does not support endian-flipping.
2022-03-18 21:20:29 +08:00
ret:
free (buf);
return err;
}