1//===- UnwindInfoSection.cpp ----------------------------------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8
9#include "UnwindInfoSection.h"
10#include "InputSection.h"
11#include "Layout.h"
12#include "OutputSection.h"
13#include "OutputSegment.h"
14#include "SymbolTable.h"
15#include "Symbols.h"
16#include "SyntheticSections.h"
17#include "Target.h"
18
19#include "lld/Common/ErrorHandler.h"
20#include "lld/Common/Memory.h"
21#include "llvm/ADT/DenseMap.h"
22#include "llvm/ADT/STLExtras.h"
23#include "llvm/BinaryFormat/MachO.h"
24#include "llvm/Support/Parallel.h"
25
26#include "mach-o/compact_unwind_encoding.h"
27
28#include <numeric>
29
30using namespace llvm;
31using namespace llvm::MachO;
32using namespace llvm::support::endian;
33using namespace lld;
34using namespace lld::macho;
35
36#define COMMON_ENCODINGS_MAX 127
37#define COMPACT_ENCODINGS_MAX 256
38
39#define SECOND_LEVEL_PAGE_BYTES 4096
40#define SECOND_LEVEL_PAGE_WORDS (SECOND_LEVEL_PAGE_BYTES / sizeof(uint32_t))
41#define REGULAR_SECOND_LEVEL_ENTRIES_MAX \
42 ((SECOND_LEVEL_PAGE_BYTES - \
43 sizeof(unwind_info_regular_second_level_page_header)) / \
44 sizeof(unwind_info_regular_second_level_entry))
45#define COMPRESSED_SECOND_LEVEL_ENTRIES_MAX \
46 ((SECOND_LEVEL_PAGE_BYTES - \
47 sizeof(unwind_info_compressed_second_level_page_header)) / \
48 sizeof(uint32_t))
49
50#define COMPRESSED_ENTRY_FUNC_OFFSET_BITS 24
51#define COMPRESSED_ENTRY_FUNC_OFFSET_MASK \
52 UNWIND_INFO_COMPRESSED_ENTRY_FUNC_OFFSET(~0)
53
54static_assert(static_cast<uint32_t>(UNWIND_X86_64_DWARF_SECTION_OFFSET) ==
55 static_cast<uint32_t>(UNWIND_ARM64_DWARF_SECTION_OFFSET) &&
56 static_cast<uint32_t>(UNWIND_X86_64_DWARF_SECTION_OFFSET) ==
57 static_cast<uint32_t>(UNWIND_X86_DWARF_SECTION_OFFSET));
58
59constexpr uint64_t DWARF_SECTION_OFFSET = UNWIND_X86_64_DWARF_SECTION_OFFSET;
60
61// Compact Unwind format is a Mach-O evolution of DWARF Unwind that
62// optimizes space and exception-time lookup. Most DWARF unwind
63// entries can be replaced with Compact Unwind entries, but the ones
64// that cannot are retained in DWARF form.
65//
66// This comment will address macro-level organization of the pre-link
67// and post-link compact unwind tables. For micro-level organization
68// pertaining to the bitfield layout of the 32-bit compact unwind
69// entries, see libunwind/include/mach-o/compact_unwind_encoding.h
70//
71// Important clarifying factoids:
72//
73// * __LD,__compact_unwind is the compact unwind format for compiler
74// output and linker input. It is never a final output. It could be
75// an intermediate output with the `-r` option which retains relocs.
76//
77// * __TEXT,__unwind_info is the compact unwind format for final
78// linker output. It is never an input.
79//
80// * __TEXT,__eh_frame is the DWARF format for both linker input and output.
81//
82// * __TEXT,__unwind_info entries are divided into 4 KiB pages (2nd
83// level) by ascending address, and the pages are referenced by an
84// index (1st level) in the section header.
85//
86// * Following the headers in __TEXT,__unwind_info, the bulk of the
87// section contains a vector of compact unwind entries
88// `{functionOffset, encoding}` sorted by ascending `functionOffset`.
89// Adjacent entries with the same encoding can be folded to great
90// advantage, achieving a 3-order-of-magnitude reduction in the
91// number of entries.
92//
93// Refer to the definition of unwind_info_section_header in
94// compact_unwind_encoding.h for an overview of the format we are encoding
95// here.
96
97// TODO(gkm): how do we align the 2nd-level pages?
98
99// The various fields in the on-disk representation of each compact unwind
100// entry.
101#define FOR_EACH_CU_FIELD(DO) \
102 DO(Ptr, functionAddress) \
103 DO(uint32_t, functionLength) \
104 DO(compact_unwind_encoding_t, encoding) \
105 DO(Ptr, personality) \
106 DO(Ptr, lsda)
107
108CREATE_LAYOUT_CLASS(CompactUnwind, FOR_EACH_CU_FIELD);
109
110#undef FOR_EACH_CU_FIELD
111
112// LLD's internal representation of a compact unwind entry.
113struct CompactUnwindEntry {
114 uint64_t functionAddress;
115 uint32_t functionLength;
116 compact_unwind_encoding_t encoding;
117 Symbol *personality;
118 InputSection *lsda;
119};
120
121using EncodingMap = DenseMap<compact_unwind_encoding_t, size_t>;
122
123struct SecondLevelPage {
124 uint32_t kind;
125 size_t entryIndex;
126 size_t entryCount;
127 size_t byteCount;
128 std::vector<compact_unwind_encoding_t> localEncodings;
129 EncodingMap localEncodingIndexes;
130};
131
132// UnwindInfoSectionImpl allows us to avoid cluttering our header file with a
133// lengthy definition of UnwindInfoSection.
134class UnwindInfoSectionImpl final : public UnwindInfoSection {
135public:
136 UnwindInfoSectionImpl() : cuLayout(target->wordSize) {}
137 uint64_t getSize() const override { return unwindInfoSize; }
138 void prepare() override;
139 void finalize() override;
140 void writeTo(uint8_t *buf) const override;
141
142private:
143 void prepareRelocations(ConcatInputSection *);
144 void relocateCompactUnwind(std::vector<CompactUnwindEntry> &);
145 void encodePersonalities();
146 Symbol *canonicalizePersonality(Symbol *);
147
148 uint64_t unwindInfoSize = 0;
149 SmallVector<decltype(symbols)::value_type, 0> symbolsVec;
150 CompactUnwindLayout cuLayout;
151 std::vector<std::pair<compact_unwind_encoding_t, size_t>> commonEncodings;
152 EncodingMap commonEncodingIndexes;
153 // The entries here will be in the same order as their originating symbols
154 // in symbolsVec.
155 std::vector<CompactUnwindEntry> cuEntries;
156 // Indices into the cuEntries vector.
157 std::vector<size_t> cuIndices;
158 std::vector<Symbol *> personalities;
159 SmallDenseMap<std::pair<InputSection *, uint64_t /* addend */>, Symbol *>
160 personalityTable;
161 // Indices into cuEntries for CUEs with a non-null LSDA.
162 std::vector<size_t> entriesWithLsda;
163 // Map of cuEntries index to an index within the LSDA array.
164 DenseMap<size_t, uint32_t> lsdaIndex;
165 std::vector<SecondLevelPage> secondLevelPages;
166 uint64_t level2PagesOffset = 0;
167 // The highest-address function plus its size. The unwinder needs this to
168 // determine the address range that is covered by unwind info.
169 uint64_t cueEndBoundary = 0;
170};
171
172UnwindInfoSection::UnwindInfoSection()
173 : SyntheticSection(segment_names::text, section_names::unwindInfo) {
174 align = 4;
175}
176
177// Record function symbols that may need entries emitted in __unwind_info, which
178// stores unwind data for address ranges.
179//
180// Note that if several adjacent functions have the same unwind encoding and
181// personality function and no LSDA, they share one unwind entry. For this to
182// work, functions without unwind info need explicit "no unwind info" unwind
183// entries -- else the unwinder would think they have the unwind info of the
184// closest function with unwind info right before in the image. Thus, we add
185// function symbols for each unique address regardless of whether they have
186// associated unwind info.
187void UnwindInfoSection::addSymbol(const Defined *d) {
188 if (d->unwindEntry())
189 allEntriesAreOmitted = false;
190 // We don't yet know the final output address of this symbol, but we know that
191 // they are uniquely determined by a combination of the isec and value, so
192 // we use that as the key here.
193 auto p = symbols.insert(KV: {{d->isec(), d->value}, d});
194 // If we have multiple symbols at the same address, only one of them can have
195 // an associated unwind entry.
196 if (!p.second && d->unwindEntry()) {
197 assert(p.first->second == d || !p.first->second->unwindEntry());
198 p.first->second = d;
199 }
200}
201
202void UnwindInfoSectionImpl::prepare() {
203 // This iteration needs to be deterministic, since prepareRelocations may add
204 // entries to the GOT. Hence the use of a MapVector for
205 // UnwindInfoSection::symbols.
206 for (const Defined *d : make_second_range(c&: symbols))
207 if (d->unwindEntry()) {
208 if (d->unwindEntry()->getName() == section_names::compactUnwind) {
209 prepareRelocations(d->unwindEntry());
210 } else {
211 // We don't have to add entries to the GOT here because FDEs have
212 // explicit GOT relocations, so Writer::scanRelocations() will add those
213 // GOT entries. However, we still need to canonicalize the personality
214 // pointers (like prepareRelocations() does for CU entries) in order
215 // to avoid overflowing the 3-personality limit.
216 FDE &fde = cast<ObjFile>(Val: d->getFile())->fdes[d->unwindEntry()];
217 fde.personality = canonicalizePersonality(fde.personality);
218 }
219 }
220}
221
222// Compact unwind relocations have different semantics, so we handle them in a
223// separate code path from regular relocations. First, we do not wish to add
224// rebase opcodes for __LD,__compact_unwind, because that section doesn't
225// actually end up in the final binary. Second, personality pointers always
226// reside in the GOT and must be treated specially.
227void UnwindInfoSectionImpl::prepareRelocations(ConcatInputSection *isec) {
228 assert(!isec->shouldOmitFromOutput() &&
229 "__compact_unwind section should not be omitted");
230
231 // FIXME: Make this skip relocations for CompactUnwindEntries that
232 // point to dead-stripped functions. That might save some amount of
233 // work. But since there are usually just few personality functions
234 // that are referenced from many places, at least some of them likely
235 // live, it wouldn't reduce number of got entries.
236 for (size_t i = 0; i < isec->relocs.size(); ++i) {
237 Reloc &r = isec->relocs[i];
238 assert(target->hasAttr(r.type, RelocAttrBits::UNSIGNED));
239 // Since compact unwind sections aren't part of the inputSections vector,
240 // they don't get canonicalized by scanRelocations(), so we have to do the
241 // canonicalization here.
242 if (auto *referentIsec = r.referent.dyn_cast<InputSection *>())
243 r.referent = referentIsec->canonical();
244
245 // Functions and LSDA entries always reside in the same object file as the
246 // compact unwind entries that references them, and thus appear as section
247 // relocs. There is no need to prepare them. We only prepare relocs for
248 // personality functions.
249 if (r.offset != cuLayout.personalityOffset)
250 continue;
251
252 if (auto *s = r.referent.dyn_cast<Symbol *>()) {
253 // Personality functions are nearly always system-defined (e.g.,
254 // ___gxx_personality_v0 for C++) and relocated as dylib symbols. When an
255 // application provides its own personality function, it might be
256 // referenced by an extern Defined symbol reloc, or a local section reloc.
257 if (auto *defined = dyn_cast<Defined>(Val: s)) {
258 // XXX(vyng) This is a special case for handling duplicate personality
259 // symbols. Note that LD64's behavior is a bit different and it is
260 // inconsistent with how symbol resolution usually work
261 //
262 // So we've decided not to follow it. Instead, simply pick the symbol
263 // with the same name from the symbol table to replace the local one.
264 //
265 // (See discussions/alternatives already considered on D107533)
266 if (!defined->isExternal())
267 if (Symbol *sym = symtab->find(name: defined->getName()))
268 if (!sym->isLazy())
269 r.referent = s = sym;
270 }
271 if (auto *undefined = dyn_cast<Undefined>(Val: s)) {
272 treatUndefinedSymbol(*undefined, isec, offset: r.offset);
273 // treatUndefinedSymbol() can replace s with a DylibSymbol; re-check.
274 if (isa<Undefined>(Val: s))
275 continue;
276 }
277
278 // Similar to canonicalizePersonality(), but we also register a GOT entry.
279 if (auto *defined = dyn_cast<Defined>(Val: s)) {
280 // Check if we have created a synthetic symbol at the same address.
281 Symbol *&personality =
282 personalityTable[{defined->isec(), defined->value}];
283 if (personality == nullptr) {
284 personality = defined;
285 in.got->addEntry(sym: defined);
286 } else if (personality != defined) {
287 r.referent = personality;
288 }
289 continue;
290 }
291
292 assert(isa<DylibSymbol>(s));
293 in.got->addEntry(sym: s);
294 continue;
295 }
296
297 if (auto *referentIsec = r.referent.dyn_cast<InputSection *>()) {
298 assert(!isCoalescedWeak(referentIsec));
299 // Personality functions can be referenced via section relocations
300 // if they live in the same object file. Create placeholder synthetic
301 // symbols for them in the GOT. If the corresponding symbol is already
302 // in the GOT, use that to avoid creating a duplicate entry. All GOT
303 // entries needed by non-unwind sections will have already been added
304 // by this point.
305 Symbol *&s = personalityTable[{referentIsec, r.addend}];
306 if (s == nullptr) {
307 Defined *const *gotEntry =
308 llvm::find_if(Range&: referentIsec->symbols, P: [&](Defined const *d) {
309 return d->value == static_cast<uint64_t>(r.addend) &&
310 d->isInGot();
311 });
312 if (gotEntry != referentIsec->symbols.end()) {
313 s = *gotEntry;
314 } else {
315 // This runs after dead stripping, so the noDeadStrip argument does
316 // not matter.
317 s = make<Defined>(args: "<internal>", /*file=*/args: nullptr, args&: referentIsec,
318 args&: r.addend, /*size=*/args: 0, /*isWeakDef=*/args: false,
319 /*isExternal=*/args: false, /*isPrivateExtern=*/args: false,
320 /*includeInSymtab=*/args: true,
321 /*isReferencedDynamically=*/args: false,
322 /*noDeadStrip=*/args: false);
323 s->used = true;
324 in.got->addEntry(sym: s);
325 }
326 }
327 r.referent = s;
328 r.addend = 0;
329 }
330 }
331}
332
333Symbol *UnwindInfoSectionImpl::canonicalizePersonality(Symbol *personality) {
334 if (auto *defined = dyn_cast_or_null<Defined>(Val: personality)) {
335 // Check if we have created a synthetic symbol at the same address.
336 Symbol *&synth = personalityTable[{defined->isec(), defined->value}];
337 if (synth == nullptr)
338 synth = defined;
339 else if (synth != defined)
340 return synth;
341 }
342 return personality;
343}
344
345// We need to apply the relocations to the pre-link compact unwind section
346// before converting it to post-link form. There should only be absolute
347// relocations here: since we are not emitting the pre-link CU section, there
348// is no source address to make a relative location meaningful.
349void UnwindInfoSectionImpl::relocateCompactUnwind(
350 std::vector<CompactUnwindEntry> &cuEntries) {
351 parallelFor(Begin: 0, End: symbolsVec.size(), Fn: [&](size_t i) {
352 CompactUnwindEntry &cu = cuEntries[i];
353 const Defined *d = symbolsVec[i].second;
354 cu.functionAddress = d->getVA();
355 if (!d->unwindEntry())
356 return;
357
358 // If we have DWARF unwind info, create a slimmed-down CU entry that points
359 // to it.
360 if (d->unwindEntry()->getName() == section_names::ehFrame) {
361 // The unwinder will look for the DWARF entry starting at the hint,
362 // assuming the hint points to a valid CFI record start. If it
363 // fails to find the record, it proceeds in a linear search through the
364 // contiguous CFI records from the hint until the end of the section.
365 // Ideally, in the case where the offset is too large to be encoded, we
366 // would instead encode the largest possible offset to a valid CFI record,
367 // but since we don't keep track of that, just encode zero -- the start of
368 // the section is always the start of a CFI record.
369 uint64_t dwarfOffsetHint =
370 d->unwindEntry()->outSecOff <= DWARF_SECTION_OFFSET
371 ? d->unwindEntry()->outSecOff
372 : 0;
373 cu.encoding = target->modeDwarfEncoding | dwarfOffsetHint;
374 const FDE &fde = cast<ObjFile>(Val: d->getFile())->fdes[d->unwindEntry()];
375 cu.functionLength = fde.funcLength;
376 // Omit the DWARF personality from compact-unwind entry so that we
377 // don't need to encode it.
378 cu.personality = nullptr;
379 cu.lsda = fde.lsda;
380 return;
381 }
382
383 assert(d->unwindEntry()->getName() == section_names::compactUnwind);
384
385 auto buf =
386 reinterpret_cast<const uint8_t *>(d->unwindEntry()->data.data()) -
387 target->wordSize;
388 cu.functionLength =
389 support::endian::read32le(P: buf + cuLayout.functionLengthOffset);
390 cu.encoding = support::endian::read32le(P: buf + cuLayout.encodingOffset);
391 for (const Reloc &r : d->unwindEntry()->relocs) {
392 if (r.offset == cuLayout.personalityOffset)
393 cu.personality = r.referent.get<Symbol *>();
394 else if (r.offset == cuLayout.lsdaOffset)
395 cu.lsda = r.getReferentInputSection();
396 }
397 });
398}
399
400// There should only be a handful of unique personality pointers, so we can
401// encode them as 2-bit indices into a small array.
402void UnwindInfoSectionImpl::encodePersonalities() {
403 for (size_t idx : cuIndices) {
404 CompactUnwindEntry &cu = cuEntries[idx];
405 if (cu.personality == nullptr)
406 continue;
407 // Linear search is fast enough for a small array.
408 auto it = find(Range&: personalities, Val: cu.personality);
409 uint32_t personalityIndex; // 1-based index
410 if (it != personalities.end()) {
411 personalityIndex = std::distance(first: personalities.begin(), last: it) + 1;
412 } else {
413 personalities.push_back(x: cu.personality);
414 personalityIndex = personalities.size();
415 }
416 cu.encoding |=
417 personalityIndex << llvm::countr_zero(
418 Val: static_cast<compact_unwind_encoding_t>(UNWIND_PERSONALITY_MASK));
419 }
420 if (personalities.size() > 3)
421 error(msg: "too many personalities (" + Twine(personalities.size()) +
422 ") for compact unwind to encode");
423}
424
425static bool canFoldEncoding(compact_unwind_encoding_t encoding) {
426 // From compact_unwind_encoding.h:
427 // UNWIND_X86_64_MODE_STACK_IND:
428 // A "frameless" (RBP not used as frame pointer) function large constant
429 // stack size. This case is like the previous, except the stack size is too
430 // large to encode in the compact unwind encoding. Instead it requires that
431 // the function contains "subq $nnnnnnnn,RSP" in its prolog. The compact
432 // encoding contains the offset to the nnnnnnnn value in the function in
433 // UNWIND_X86_64_FRAMELESS_STACK_SIZE.
434 // Since this means the unwinder has to look at the `subq` in the function
435 // of the unwind info's unwind address, two functions that have identical
436 // unwind info can't be folded if it's using this encoding since both
437 // entries need unique addresses.
438 static_assert(static_cast<uint32_t>(UNWIND_X86_64_MODE_STACK_IND) ==
439 static_cast<uint32_t>(UNWIND_X86_MODE_STACK_IND));
440 if ((target->cpuType == CPU_TYPE_X86_64 || target->cpuType == CPU_TYPE_X86) &&
441 (encoding & UNWIND_MODE_MASK) == UNWIND_X86_64_MODE_STACK_IND) {
442 // FIXME: Consider passing in the two function addresses and getting
443 // their two stack sizes off the `subq` and only returning false if they're
444 // actually different.
445 return false;
446 }
447 return true;
448}
449
450// Scan the __LD,__compact_unwind entries and compute the space needs of
451// __TEXT,__unwind_info and __TEXT,__eh_frame.
452void UnwindInfoSectionImpl::finalize() {
453 if (symbols.empty())
454 return;
455
456 // At this point, the address space for __TEXT,__text has been
457 // assigned, so we can relocate the __LD,__compact_unwind entries
458 // into a temporary buffer. Relocation is necessary in order to sort
459 // the CU entries by function address. Sorting is necessary so that
460 // we can fold adjacent CU entries with identical encoding+personality
461 // and without any LSDA. Folding is necessary because it reduces the
462 // number of CU entries by as much as 3 orders of magnitude!
463 cuEntries.resize(new_size: symbols.size());
464 // The "map" part of the symbols MapVector was only needed for deduplication
465 // in addSymbol(). Now that we are done adding, move the contents to a plain
466 // std::vector for indexed access.
467 symbolsVec = symbols.takeVector();
468 relocateCompactUnwind(cuEntries);
469
470 // Rather than sort & fold the 32-byte entries directly, we create a
471 // vector of indices to entries and sort & fold that instead.
472 cuIndices.resize(new_size: cuEntries.size());
473 std::iota(first: cuIndices.begin(), last: cuIndices.end(), value: 0);
474 llvm::sort(C&: cuIndices, Comp: [&](size_t a, size_t b) {
475 return cuEntries[a].functionAddress < cuEntries[b].functionAddress;
476 });
477
478 // Record the ending boundary before we fold the entries.
479 cueEndBoundary = cuEntries[cuIndices.back()].functionAddress +
480 cuEntries[cuIndices.back()].functionLength;
481
482 // Fold adjacent entries with matching encoding+personality and without LSDA
483 // We use three iterators on the same cuIndices to fold in-situ:
484 // (1) `foldBegin` is the first of a potential sequence of matching entries
485 // (2) `foldEnd` is the first non-matching entry after `foldBegin`.
486 // The semi-open interval [ foldBegin .. foldEnd ) contains a range
487 // entries that can be folded into a single entry and written to ...
488 // (3) `foldWrite`
489 auto foldWrite = cuIndices.begin();
490 for (auto foldBegin = cuIndices.begin(); foldBegin < cuIndices.end();) {
491 auto foldEnd = foldBegin;
492 // Common LSDA encodings (e.g. for C++ and Objective-C) contain offsets from
493 // a base address. The base address is normally not contained directly in
494 // the LSDA, and in that case, the personality function treats the starting
495 // address of the function (which is computed by the unwinder) as the base
496 // address and interprets the LSDA accordingly. The unwinder computes the
497 // starting address of a function as the address associated with its CU
498 // entry. For this reason, we cannot fold adjacent entries if they have an
499 // LSDA, because folding would make the unwinder compute the wrong starting
500 // address for the functions with the folded entries, which in turn would
501 // cause the personality function to misinterpret the LSDA for those
502 // functions. In the very rare case where the base address is encoded
503 // directly in the LSDA, two functions at different addresses would
504 // necessarily have different LSDAs, so their CU entries would not have been
505 // folded anyway.
506 while (++foldEnd < cuIndices.end() &&
507 cuEntries[*foldBegin].encoding == cuEntries[*foldEnd].encoding &&
508 !cuEntries[*foldBegin].lsda && !cuEntries[*foldEnd].lsda &&
509 // If we've gotten to this point, we don't have an LSDA, which should
510 // also imply that we don't have a personality function, since in all
511 // likelihood a personality function needs the LSDA to do anything
512 // useful. It can be technically valid to have a personality function
513 // and no LSDA though (e.g. the C++ personality __gxx_personality_v0
514 // is just a no-op without LSDA), so we still check for personality
515 // function equivalence to handle that case.
516 cuEntries[*foldBegin].personality ==
517 cuEntries[*foldEnd].personality &&
518 canFoldEncoding(encoding: cuEntries[*foldEnd].encoding))
519 ;
520 *foldWrite++ = *foldBegin;
521 foldBegin = foldEnd;
522 }
523 cuIndices.erase(first: foldWrite, last: cuIndices.end());
524
525 encodePersonalities();
526
527 // Count frequencies of the folded encodings
528 EncodingMap encodingFrequencies;
529 for (size_t idx : cuIndices)
530 encodingFrequencies[cuEntries[idx].encoding]++;
531
532 // Make a vector of encodings, sorted by descending frequency
533 for (const auto &frequency : encodingFrequencies)
534 commonEncodings.emplace_back(args: frequency);
535 llvm::sort(C&: commonEncodings,
536 Comp: [](const std::pair<compact_unwind_encoding_t, size_t> &a,
537 const std::pair<compact_unwind_encoding_t, size_t> &b) {
538 if (a.second == b.second)
539 // When frequencies match, secondarily sort on encoding
540 // to maintain parity with validate-unwind-info.py
541 return a.first > b.first;
542 return a.second > b.second;
543 });
544
545 // Truncate the vector to 127 elements.
546 // Common encoding indexes are limited to 0..126, while encoding
547 // indexes 127..255 are local to each second-level page
548 if (commonEncodings.size() > COMMON_ENCODINGS_MAX)
549 commonEncodings.resize(COMMON_ENCODINGS_MAX);
550
551 // Create a map from encoding to common-encoding-table index
552 for (size_t i = 0; i < commonEncodings.size(); i++)
553 commonEncodingIndexes[commonEncodings[i].first] = i;
554
555 // Split folded encodings into pages, where each page is limited by ...
556 // (a) 4 KiB capacity
557 // (b) 24-bit difference between first & final function address
558 // (c) 8-bit compact-encoding-table index,
559 // for which 0..126 references the global common-encodings table,
560 // and 127..255 references a local per-second-level-page table.
561 // First we try the compact format and determine how many entries fit.
562 // If more entries fit in the regular format, we use that.
563 for (size_t i = 0; i < cuIndices.size();) {
564 size_t idx = cuIndices[i];
565 secondLevelPages.emplace_back();
566 SecondLevelPage &page = secondLevelPages.back();
567 page.entryIndex = i;
568 uint64_t functionAddressMax =
569 cuEntries[idx].functionAddress + COMPRESSED_ENTRY_FUNC_OFFSET_MASK;
570 size_t n = commonEncodings.size();
571 size_t wordsRemaining =
572 SECOND_LEVEL_PAGE_WORDS -
573 sizeof(unwind_info_compressed_second_level_page_header) /
574 sizeof(uint32_t);
575 while (wordsRemaining >= 1 && i < cuIndices.size()) {
576 idx = cuIndices[i];
577 const CompactUnwindEntry *cuPtr = &cuEntries[idx];
578 if (cuPtr->functionAddress >= functionAddressMax)
579 break;
580 if (commonEncodingIndexes.count(Val: cuPtr->encoding) ||
581 page.localEncodingIndexes.count(Val: cuPtr->encoding)) {
582 i++;
583 wordsRemaining--;
584 } else if (wordsRemaining >= 2 && n < COMPACT_ENCODINGS_MAX) {
585 page.localEncodings.emplace_back(args: cuPtr->encoding);
586 page.localEncodingIndexes[cuPtr->encoding] = n++;
587 i++;
588 wordsRemaining -= 2;
589 } else {
590 break;
591 }
592 }
593 page.entryCount = i - page.entryIndex;
594
595 // If this is not the final page, see if it's possible to fit more entries
596 // by using the regular format. This can happen when there are many unique
597 // encodings, and we saturated the local encoding table early.
598 if (i < cuIndices.size() &&
599 page.entryCount < REGULAR_SECOND_LEVEL_ENTRIES_MAX) {
600 page.kind = UNWIND_SECOND_LEVEL_REGULAR;
601 page.entryCount = std::min(REGULAR_SECOND_LEVEL_ENTRIES_MAX,
602 b: cuIndices.size() - page.entryIndex);
603 i = page.entryIndex + page.entryCount;
604 } else {
605 page.kind = UNWIND_SECOND_LEVEL_COMPRESSED;
606 }
607 }
608
609 for (size_t idx : cuIndices) {
610 lsdaIndex[idx] = entriesWithLsda.size();
611 if (cuEntries[idx].lsda)
612 entriesWithLsda.push_back(x: idx);
613 }
614
615 // compute size of __TEXT,__unwind_info section
616 level2PagesOffset = sizeof(unwind_info_section_header) +
617 commonEncodings.size() * sizeof(uint32_t) +
618 personalities.size() * sizeof(uint32_t) +
619 // The extra second-level-page entry is for the sentinel
620 (secondLevelPages.size() + 1) *
621 sizeof(unwind_info_section_header_index_entry) +
622 entriesWithLsda.size() *
623 sizeof(unwind_info_section_header_lsda_index_entry);
624 unwindInfoSize =
625 level2PagesOffset + secondLevelPages.size() * SECOND_LEVEL_PAGE_BYTES;
626}
627
628// All inputs are relocated and output addresses are known, so write!
629
630void UnwindInfoSectionImpl::writeTo(uint8_t *buf) const {
631 assert(!cuIndices.empty() && "call only if there is unwind info");
632
633 // section header
634 auto *uip = reinterpret_cast<unwind_info_section_header *>(buf);
635 uip->version = 1;
636 uip->commonEncodingsArraySectionOffset = sizeof(unwind_info_section_header);
637 uip->commonEncodingsArrayCount = commonEncodings.size();
638 uip->personalityArraySectionOffset =
639 uip->commonEncodingsArraySectionOffset +
640 (uip->commonEncodingsArrayCount * sizeof(uint32_t));
641 uip->personalityArrayCount = personalities.size();
642 uip->indexSectionOffset = uip->personalityArraySectionOffset +
643 (uip->personalityArrayCount * sizeof(uint32_t));
644 uip->indexCount = secondLevelPages.size() + 1;
645
646 // Common encodings
647 auto *i32p = reinterpret_cast<uint32_t *>(&uip[1]);
648 for (const auto &encoding : commonEncodings)
649 *i32p++ = encoding.first;
650
651 // Personalities
652 for (const Symbol *personality : personalities)
653 *i32p++ = personality->getGotVA() - in.header->addr;
654
655 // FIXME: LD64 checks and warns aboutgaps or overlapse in cuEntries address
656 // ranges. We should do the same too
657
658 // Level-1 index
659 uint32_t lsdaOffset =
660 uip->indexSectionOffset +
661 uip->indexCount * sizeof(unwind_info_section_header_index_entry);
662 uint64_t l2PagesOffset = level2PagesOffset;
663 auto *iep = reinterpret_cast<unwind_info_section_header_index_entry *>(i32p);
664 for (const SecondLevelPage &page : secondLevelPages) {
665 size_t idx = cuIndices[page.entryIndex];
666 iep->functionOffset = cuEntries[idx].functionAddress - in.header->addr;
667 iep->secondLevelPagesSectionOffset = l2PagesOffset;
668 iep->lsdaIndexArraySectionOffset =
669 lsdaOffset + lsdaIndex.lookup(Val: idx) *
670 sizeof(unwind_info_section_header_lsda_index_entry);
671 iep++;
672 l2PagesOffset += SECOND_LEVEL_PAGE_BYTES;
673 }
674 // Level-1 sentinel
675 // XXX(vyng): Note that LD64 adds +1 here.
676 // Unsure whether it's a bug or it's their workaround for something else.
677 // See comments from https://reviews.llvm.org/D138320.
678 iep->functionOffset = cueEndBoundary - in.header->addr;
679 iep->secondLevelPagesSectionOffset = 0;
680 iep->lsdaIndexArraySectionOffset =
681 lsdaOffset + entriesWithLsda.size() *
682 sizeof(unwind_info_section_header_lsda_index_entry);
683 iep++;
684
685 // LSDAs
686 auto *lep =
687 reinterpret_cast<unwind_info_section_header_lsda_index_entry *>(iep);
688 for (size_t idx : entriesWithLsda) {
689 const CompactUnwindEntry &cu = cuEntries[idx];
690 lep->lsdaOffset = cu.lsda->getVA(/*off=*/0) - in.header->addr;
691 lep->functionOffset = cu.functionAddress - in.header->addr;
692 lep++;
693 }
694
695 // Level-2 pages
696 auto *pp = reinterpret_cast<uint32_t *>(lep);
697 for (const SecondLevelPage &page : secondLevelPages) {
698 if (page.kind == UNWIND_SECOND_LEVEL_COMPRESSED) {
699 uintptr_t functionAddressBase =
700 cuEntries[cuIndices[page.entryIndex]].functionAddress;
701 auto *p2p =
702 reinterpret_cast<unwind_info_compressed_second_level_page_header *>(
703 pp);
704 p2p->kind = page.kind;
705 p2p->entryPageOffset =
706 sizeof(unwind_info_compressed_second_level_page_header);
707 p2p->entryCount = page.entryCount;
708 p2p->encodingsPageOffset =
709 p2p->entryPageOffset + p2p->entryCount * sizeof(uint32_t);
710 p2p->encodingsCount = page.localEncodings.size();
711 auto *ep = reinterpret_cast<uint32_t *>(&p2p[1]);
712 for (size_t i = 0; i < page.entryCount; i++) {
713 const CompactUnwindEntry &cue =
714 cuEntries[cuIndices[page.entryIndex + i]];
715 auto it = commonEncodingIndexes.find(Val: cue.encoding);
716 if (it == commonEncodingIndexes.end())
717 it = page.localEncodingIndexes.find(Val: cue.encoding);
718 *ep++ = (it->second << COMPRESSED_ENTRY_FUNC_OFFSET_BITS) |
719 (cue.functionAddress - functionAddressBase);
720 }
721 if (!page.localEncodings.empty())
722 memcpy(dest: ep, src: page.localEncodings.data(),
723 n: page.localEncodings.size() * sizeof(uint32_t));
724 } else {
725 auto *p2p =
726 reinterpret_cast<unwind_info_regular_second_level_page_header *>(pp);
727 p2p->kind = page.kind;
728 p2p->entryPageOffset =
729 sizeof(unwind_info_regular_second_level_page_header);
730 p2p->entryCount = page.entryCount;
731 auto *ep = reinterpret_cast<uint32_t *>(&p2p[1]);
732 for (size_t i = 0; i < page.entryCount; i++) {
733 const CompactUnwindEntry &cue =
734 cuEntries[cuIndices[page.entryIndex + i]];
735 *ep++ = cue.functionAddress;
736 *ep++ = cue.encoding;
737 }
738 }
739 pp += SECOND_LEVEL_PAGE_WORDS;
740 }
741}
742
743UnwindInfoSection *macho::makeUnwindInfoSection() {
744 return make<UnwindInfoSectionImpl>();
745}
746