[ELF] - Set DF_STATIC_TLS flag for i386 target.

[grimar]

Summary

This is PR32437.

DF_STATIC_TLS
If set in a shared object or executable, this flag instructs the dynamic linker to reject attempts to load this file dynamically. It indicates that the shared object or executable contains code using a static thread-local storage scheme. Implementations need not support any form of thread-local storage.

Patch checks if IE/LE relocations were used to check if code uses static model. If so it sets the DF_STATIC_TLS flag.
Patch contains testcases for i386 target for now.

[ruiu]

ELF/Relocations.cpp

207

I don't want to bury this code deep inside the relocation processing like this. Can you create a different for loop that visits each input section's Relocations vector elements to see if there's a TLS relocation?

208

No other members of Target is writable, so you want to treat a Target as an immutable object.

[silvas]

ELF/Relocations.cpp

207

Need to measure performance if it is a separate for loop though (like for any change adding O(number of relocations) work). Putting it here does a very small amount of extra work (since it guarded by e.g. Body.isTLS()), though it is less clean.

[grimar]

Review comments addressed. Perfomance difference seems varies.
I observed from no difference to about 2%. Each run has 25 iterations.

Chromium:
Before patch:

3024,298420      task-clock (msec)         #    1,000 CPUs utilized            ( +-  0,23% )
          1      context-switches          #    0,000 K/sec                    ( +- 11,47% )
          1      cpu-migrations            #    0,000 K/sec                    ( +- 16,67% )
    168 021      page-faults               #    0,056 M/sec                    ( +-  0,15% )

3,024534666 seconds time elapsed                                          ( +-  0,23% )

After patch:

3028,392936      task-clock (msec)         #    1,000 CPUs utilized            ( +-  0,17% )
          1      context-switches          #    0,000 K/sec                    ( +- 11,47% )
          1      cpu-migrations            #    0,000 K/sec                    ( +- 11,47% )
    167 843      page-faults               #    0,055 M/sec                    ( +-  0,14% )

3,028637012 seconds time elapsed                                          ( +-  0,17% )

3,028637012 / 3,024534666 == 1.001x

Clang:
Before patch:

 444,407019      task-clock (msec)         #    0,998 CPUs utilized            ( +-  0,23% )
          0      context-switches          #    0,000 K/sec                    ( +- 46,77% )
          0      cpu-migrations            #    0,000 K/sec                    ( +- 55,28% )
     54 471      page-faults               #    0,123 M/sec                    ( +-  0,08% )

0,445156259 seconds time elapsed                                          ( +-  0,26% )

After patch:

 447,810554      task-clock (msec)         #    0,999 CPUs utilized            ( +-  0,24% )
          0      context-switches          #    0,000 K/sec                    ( +- 40,82% )
          0      cpu-migrations            #    0,000 K/sec                    ( +- 69,22% )
     54 492      page-faults               #    0,122 M/sec                    ( +-  0,04% )
0,448175370 seconds time elapsed                                          ( +-  0,25% )

0,448175370 / 0,445156259 == 1.003x

ScyllaDB:
Before patch:

2301,014053      task-clock (msec)         #    0,999 CPUs utilized            ( +-  0,24% )
          1      context-switches          #    0,000 K/sec                    ( +- 21,08% )
          1      cpu-migrations            #    0,000 K/sec                    ( +- 18,09% )
    288 479      page-faults               #    0,125 M/sec                    ( +-  0,06% )
2,302234392 seconds time elapsed                                          ( +-  0,24% )

After patch:

2347,685157      task-clock (msec)         #    1,000 CPUs utilized            ( +-  0,23% )
          1      context-switches          #    0,000 K/sec                    ( +- 17,46% )
          1      cpu-migrations            #    0,000 K/sec                    ( +- 18,09% )
    288 911      page-faults               #    0,123 M/sec                    ( +-  0,04% )
2,348484871 seconds time elapsed                                          ( +-  0,23% )

2,348484871 / 2,302234392 == 1.02x

Not sure how much that is acceptable. Should we use approach from previous diff which never
affected on perfomance ?

[ruiu]

0.3% is negligible but 2% isn't. Why this patch impact only ScyllaDB?

[grimar]

In D32354#738306, @ruiu wrote:

0.3% is negligible but 2% isn't. Why this patch impact only ScyllaDB?

I dont know. I can debug it to check this, if you think it worth to do.

[ruiu]

I think it's worth to do. I'm actually curious how that can happen, if the observation is correct.

[grimar]

I also can suggest as an option not to burry logic inside tls handling method like first diff do, but also do not introduce a new loop.
And do all things somewhere at the begining of scanRelocs:

template <class ELFT, class RelTy>
static void scanRelocs(InputSectionBase &Sec, ArrayRef<RelTy> Rels) {
  OffsetGetter GetOffset(Sec);
  for (auto I = Rels.begin(), End = Rels.end(); I != End; ++I) {
    const RelTy &Rel = *I;

    <DO LOGIC HERE>

That will avoid second loop and should be effective.

[grimar]

In D32354#738320, @ruiu wrote:

I think it's worth to do. I'm actually curious how that can happen, if the observation is correct.

I checked few times, I believe them are. I'll debug it then.

[ruiu]

The other way of doing this is to add code to relocateOne, to record whether we handle TLS IE/LD relocations or not. This is probably the least expensive way as it doesn't involve any virtual function calls.

[grimar]

In D32354#738320, @ruiu wrote:

I think it's worth to do. I'm actually curious how that can happen, if the observation is correct.

I really don't know why that happened, but today I was unable to reproduce the difference when linked ScyllaDB observed earlier.
Moreover after debugging I think there should be no reasons for that.
Scylla has 661610 relocations in total for me, and after 19248 HasStaticTlsModel flag is set, what stops scanning for TLS model.
In compare, clang has 1541964 relocations and flag is never set.
So scan of relocations for clang should take much more time, though that does not happen, it does not affect on a link time for me.

Results from 100 runs:

linking clang, before patch:
Performance counter stats for './../lld -flavor gnu --hash-style=gnu --no-add-needed --build-id --eh-frame-hdr -m elf_x86_64 -dynamic-linker /lib64/ld-linux-x86-64.so.2 -o t crt1.o crti.o crtbegin.o --strip-all -allow-shlib-undefined --export-dynamic -O3 --gc-sections tools/clang/tools/driver/CMakeFiles/clang.dir/driver.cpp.o tools/clang/tools/driver/CMakeFiles/clang.dir/cc1_main.cpp.o tools/clang/tools/driver/CMakeFiles/clang.dir/cc1as_main.cpp.o lib/libLLVMAArch64CodeGen.a lib/libLLVMAArch64AsmPrinter.a lib/libLLVMAArch64AsmParser.a lib/libLLVMAArch64Desc.a lib/libLLVMAArch64Info.a lib/libLLVMAArch64Disassembler.a lib/libLLVMAMDGPUCodeGen.a lib/libLLVMAMDGPUAsmPrinter.a lib/libLLVMAMDGPUAsmParser.a lib/libLLVMAMDGPUDesc.a lib/libLLVMAMDGPUInfo.a lib/libLLVMAMDGPUDisassembler.a lib/libLLVMARMCodeGen.a lib/libLLVMARMAsmPrinter.a lib/libLLVMARMAsmParser.a lib/libLLVMARMDesc.a lib/libLLVMARMInfo.a lib/libLLVMARMDisassembler.a lib/libLLVMBPFCodeGen.a lib/libLLVMBPFAsmPrinter.a lib/libLLVMBPFDesc.a lib/libLLVMBPFInfo.a lib/libLLVMCppBackendCodeGen.a lib/libLLVMCppBackendInfo.a lib/libLLVMHexagonCodeGen.a lib/libLLVMHexagonAsmParser.a lib/libLLVMHexagonDesc.a lib/libLLVMHexagonInfo.a lib/libLLVMHexagonDisassembler.a lib/libLLVMMipsCodeGen.a lib/libLLVMMipsAsmPrinter.a lib/libLLVMMipsAsmParser.a lib/libLLVMMipsDesc.a lib/libLLVMMipsInfo.a lib/libLLVMMipsDisassembler.a lib/libLLVMMSP430CodeGen.a lib/libLLVMMSP430AsmPrinter.a lib/libLLVMMSP430Desc.a lib/libLLVMMSP430Info.a lib/libLLVMNVPTXCodeGen.a lib/libLLVMNVPTXAsmPrinter.a lib/libLLVMNVPTXDesc.a lib/libLLVMNVPTXInfo.a lib/libLLVMPowerPCCodeGen.a lib/libLLVMPowerPCAsmPrinter.a lib/libLLVMPowerPCAsmParser.a lib/libLLVMPowerPCDesc.a lib/libLLVMPowerPCInfo.a lib/libLLVMPowerPCDisassembler.a lib/libLLVMSparcCodeGen.a lib/libLLVMSparcAsmPrinter.a lib/libLLVMSparcAsmParser.a lib/libLLVMSparcDesc.a lib/libLLVMSparcInfo.a lib/libLLVMSparcDisassembler.a lib/libLLVMSystemZCodeGen.a lib/libLLVMSystemZAsmPrinter.a lib/libLLVMSystemZAsmParser.a lib/libLLVMSystemZDesc.a lib/libLLVMSystemZInfo.a lib/libLLVMSystemZDisassembler.a lib/libLLVMX86CodeGen.a lib/libLLVMX86AsmPrinter.a lib/libLLVMX86AsmParser.a lib/libLLVMX86Desc.a lib/libLLVMX86Info.a lib/libLLVMX86Disassembler.a lib/libLLVMXCoreCodeGen.a lib/libLLVMXCoreAsmPrinter.a lib/libLLVMXCoreDesc.a lib/libLLVMXCoreInfo.a lib/libLLVMXCoreDisassembler.a lib/libLLVMAnalysis.a lib/libLLVMCodeGen.a lib/libLLVMCore.a lib/libLLVMipo.a lib/libLLVMInstCombine.a lib/libLLVMInstrumentation.a lib/libLLVMMC.a lib/libLLVMMCParser.a lib/libLLVMObjCARCOpts.a lib/libLLVMOption.a lib/libLLVMScalarOpts.a lib/libLLVMSupport.a lib/libLLVMTransformUtils.a lib/libLLVMVectorize.a lib/libclangBasic.a lib/libclangCodeGen.a lib/libclangDriver.a lib/libclangFrontend.a lib/libclangFrontendTool.a lib/libLLVMGlobalISel.a lib/libLLVMAArch64Desc.a lib/libLLVMAArch64AsmPrinter.a lib/libLLVMAArch64Info.a lib/libLLVMAArch64Utils.a lib/libLLVMAMDGPUDesc.a lib/libLLVMAMDGPUAsmPrinter.a lib/libLLVMAMDGPUInfo.a lib/libLLVMAMDGPUUtils.a lib/libLLVMARMDesc.a lib/libLLVMARMAsmPrinter.a lib/libLLVMARMInfo.a lib/libLLVMBPFAsmPrinter.a lib/libLLVMHexagonDesc.a lib/libLLVMHexagonInfo.a lib/libLLVMMipsAsmPrinter.a lib/libLLVMMipsInfo.a lib/libLLVMMSP430AsmPrinter.a lib/libLLVMNVPTXAsmPrinter.a lib/libLLVMPowerPCAsmPrinter.a lib/libLLVMPowerPCInfo.a lib/libLLVMSparcAsmPrinter.a lib/libLLVMSparcInfo.a lib/libLLVMSystemZDesc.a lib/libLLVMSystemZAsmPrinter.a lib/libLLVMSystemZInfo.a lib/libLLVMX86AsmPrinter.a lib/libLLVMX86Utils.a lib/libLLVMX86Info.a lib/libLLVMXCoreAsmPrinter.a lib/libLLVMAsmPrinter.a lib/libLLVMDebugInfoCodeView.a lib/libLLVMSelectionDAG.a lib/libLLVMCodeGen.a lib/libLLVMXCoreInfo.a lib/libLLVMMCDisassembler.a lib/libclangCodeGen.a lib/libLLVMipo.a lib/libLLVMVectorize.a lib/libLLVMInstrumentation.a lib/libLLVMObjCARCOpts.a lib/libLLVMScalarOpts.a lib/libLLVMInstCombine.a lib/libLLVMTarget.a lib/libLLVMBitWriter.a lib/libLLVMIRReader.a lib/libLLVMAsmParser.a lib/libLLVMLinker.a lib/libLLVMTransformUtils.a lib/libLLVMAnalysis.a lib/libclangRewriteFrontend.a lib/libclangARCMigrate.a lib/libclangStaticAnalyzerFrontend.a lib/libclangFrontend.a lib/libclangDriver.a lib/libLLVMOption.a lib/libLLVMProfileData.a lib/libLLVMObject.a lib/libclangParse.a lib/libLLVMMCParser.a lib/libclangSerialization.a lib/libLLVMBitReader.a lib/libclangSema.a lib/libclangEdit.a lib/libclangStaticAnalyzerCheckers.a lib/libclangStaticAnalyzerCore.a lib/libclangAnalysis.a lib/libclangAST.a lib/libclangRewrite.a lib/libclangLex.a lib/libclangBasic.a lib/libLLVMCore.a lib/libLLVMMC.a lib/libLLVMSupport.a librt.so libdl.so libtinfo.so libpthread.so.0 libz.so libm.so.6 -rpath $ORIGIN/../lib libstdc++.so libgcc_s.so libc.so.6 libc_nonshared.a --as-needed ld-linux-x86-64.so.2 --no-as-needed libgcc_s.so crtend.o crtn.o' (100 runs):

     467,482313      task-clock (msec)         #    0,999 CPUs utilized            ( +-  0,14% )
            134      context-switches          #    0,286 K/sec                    ( +-  0,19% )
              0      cpu-migrations            #    0,000 K/sec                  
         55 282      page-faults               #    0,118 M/sec                    ( +-  0,05% )
<not supported>      cycles                   
<not supported>      stalled-cycles-frontend  
<not supported>      stalled-cycles-backend   
<not supported>      instructions             
<not supported>      branches                 
<not supported>      branch-misses            

    0,467917575 seconds time elapsed                                          ( +-  0,14% )

linking clang after patch:

     470,285320      task-clock (msec)         #    0,999 CPUs utilized            ( +-  0,14% )
            134      context-switches          #    0,284 K/sec                    ( +-  0,19% )
              0      cpu-migrations            #    0,000 K/sec                  
         55 225      page-faults               #    0,117 M/sec                    ( +-  0,06% )
<not supported>      cycles                   
<not supported>      stalled-cycles-frontend  
<not supported>      stalled-cycles-backend   
<not supported>      instructions             
<not supported>      branches                 
<not supported>      branch-misses            

    0,470744480 seconds time elapsed                                          ( +-  0,14% )

Difference is still minimal here: 0,470744480 / 0,467917575 == 1.00604

linking scylladb before patch:
Performance counter stats for './../lld -flavor gnu -build-id --no-add-needed --eh-frame-hdr --hash-style=gnu -m elf_x86_64 -dynamic-linker /lib64/ld-linux-x86-64.so.2 -o t crt1.o crti.o crtbegin.o --whole-archive seastar/build/release/libseastar.a --no-whole-archive --no-as-needed libaio.so libboost_program_options.so libboost_system.so libboost_filesystem.so libstdc++.so libboost_unit_test_framework.so libboost_thread.so.1.58.0 libcryptopp.so librt.so libgnutls.so libgnutlsxx.so libhwloc.so libnuma.so libpciaccess.so libxml2.so libz.so build/release/main.o build/release/database.o build/release/schema.o build/release/frozen_schema.o build/release/schema_registry.o build/release/bytes.o build/release/mutation.o build/release/row_cache.o build/release/canonical_mutation.o build/release/frozen_mutation.o build/release/memtable.o build/release/schema_mutations.o build/release/release.o build/release/utils/logalloc.o build/release/utils/large_bitset.o build/release/mutation_partition.o build/release/mutation_partition_view.o build/release/mutation_partition_serializer.o build/release/mutation_reader.o build/release/mutation_query.o build/release/key_reader.o build/release/keys.o build/release/sstables/sstables.o build/release/sstables/compress.o build/release/sstables/row.o build/release/sstables/key.o build/release/sstables/partition.o build/release/sstables/filter.o build/release/sstables/compaction.o build/release/sstables/compaction_manager.o build/release/log.o build/release/transport/event.o build/release/transport/event_notifier.o build/release/transport/server.o build/release/cql3/abstract_marker.o build/release/cql3/attributes.o build/release/cql3/cf_name.o build/release/cql3/cql3_type.o build/release/cql3/operation.o build/release/cql3/index_name.o build/release/cql3/keyspace_element_name.o build/release/cql3/lists.o build/release/cql3/sets.o build/release/cql3/maps.o build/release/cql3/functions/functions.o build/release/cql3/statements/cf_prop_defs.o build/release/cql3/statements/cf_statement.o build/release/cql3/statements/create_keyspace_statement.o build/release/cql3/statements/create_table_statement.o build/release/cql3/statements/drop_keyspace_statement.o build/release/cql3/statements/drop_table_statement.o build/release/cql3/statements/schema_altering_statement.o build/release/cql3/statements/ks_prop_defs.o build/release/cql3/statements/modification_statement.o build/release/cql3/statements/parsed_statement.o build/release/cql3/statements/property_definitions.o build/release/cql3/statements/update_statement.o build/release/cql3/statements/delete_statement.o build/release/cql3/statements/batch_statement.o build/release/cql3/statements/select_statement.o build/release/cql3/statements/use_statement.o build/release/cql3/statements/index_prop_defs.o build/release/cql3/statements/index_target.o build/release/cql3/statements/create_index_statement.o build/release/cql3/statements/truncate_statement.o build/release/cql3/statements/alter_table_statement.o build/release/cql3/update_parameters.o build/release/cql3/ut_name.o build/release/thrift/handler.o build/release/thrift/server.o build/release/thrift/thrift_validation.o build/release/utils/runtime.o build/release/utils/murmur_hash.o build/release/utils/uuid.o build/release/utils/big_decimal.o build/release/types.o build/release/validation.o build/release/service/priority_manager.o build/release/service/migration_manager.o build/release/service/storage_proxy.o build/release/cql3/operator.o build/release/cql3/relation.o build/release/cql3/column_identifier.o build/release/cql3/constants.o build/release/cql3/query_processor.o build/release/cql3/query_options.o build/release/cql3/single_column_relation.o build/release/cql3/token_relation.o build/release/cql3/column_condition.o build/release/cql3/user_types.o build/release/cql3/untyped_result_set.o build/release/cql3/selection/abstract_function_selector.o build/release/cql3/selection/simple_selector.o build/release/cql3/selection/selectable.o build/release/cql3/selection/selector_factories.o build/release/cql3/selection/selection.o build/release/cql3/selection/selector.o build/release/cql3/restrictions/statement_restrictions.o build/release/db/consistency_level.o build/release/db/system_keyspace.o build/release/db/schema_tables.o build/release/db/commitlog/commitlog.o build/release/db/commitlog/commitlog_replayer.o build/release/db/serializer.o build/release/db/config.o build/release/db/index/secondary_index.o build/release/db/marshal/type_parser.o build/release/db/batchlog_manager.o build/release/io/io.o build/release/utils/utils.o build/release/utils/UUID_gen.o build/release/utils/i_filter.o build/release/utils/bloom_filter.o build/release/utils/bloom_calculations.o build/release/utils/rate_limiter.o build/release/utils/file_lock.o build/release/utils/dynamic_bitset.o build/release/gms/version_generator.o build/release/gms/versioned_value.o build/release/gms/gossiper.o build/release/gms/failure_detector.o build/release/gms/gossip_digest_syn.o build/release/gms/gossip_digest_ack.o build/release/gms/gossip_digest_ack2.o build/release/gms/endpoint_state.o build/release/gms/application_state.o build/release/dht/i_partitioner.o build/release/dht/murmur3_partitioner.o build/release/dht/byte_ordered_partitioner.o build/release/dht/boot_strapper.o build/release/dht/range_streamer.o build/release/unimplemented.o build/release/query.o build/release/query-result-set.o build/release/locator/abstract_replication_strategy.o build/release/locator/simple_strategy.o build/release/locator/local_strategy.o build/release/locator/network_topology_strategy.o build/release/locator/token_metadata.o build/release/locator/locator.o build/release/locator/snitch_base.o build/release/locator/simple_snitch.o build/release/locator/rack_inferring_snitch.o build/release/locator/gossiping_property_file_snitch.o build/release/locator/production_snitch_base.o build/release/locator/ec2_snitch.o build/release/locator/ec2_multi_region_snitch.o build/release/message/messaging_service.o build/release/service/client_state.o build/release/service/migration_task.o build/release/service/storage_service.o build/release/service/pending_range_calculator_service.o build/release/service/load_broadcaster.o build/release/service/pager/paging_state.o build/release/service/pager/query_pagers.o build/release/streaming/stream_task.o build/release/streaming/stream_session.o build/release/streaming/stream_request.o build/release/streaming/stream_summary.o build/release/streaming/stream_transfer_task.o build/release/streaming/stream_receive_task.o build/release/streaming/stream_plan.o build/release/streaming/progress_info.o build/release/streaming/session_info.o build/release/streaming/stream_coordinator.o build/release/streaming/stream_manager.o build/release/streaming/stream_result_future.o build/release/streaming/stream_session_state.o build/release/gc_clock.o build/release/partition_slice_builder.o build/release/init.o build/release/repair/repair.o build/release/exceptions/exceptions.o build/release/dns.o build/release/auth/auth.o build/release/auth/authenticated_user.o build/release/auth/authenticator.o build/release/auth/data_resource.o build/release/auth/password_authenticator.o build/release/auth/permission.o build/release/api/api.o build/release/api/storage_service.o build/release/api/commitlog.o build/release/api/gossiper.o build/release/api/failure_detector.o build/release/api/column_family.o build/release/api/messaging_service.o build/release/api/storage_proxy.o build/release/api/cache_service.o build/release/api/collectd.o build/release/api/endpoint_snitch.o build/release/api/compaction_manager.o build/release/api/hinted_handoff.o build/release/api/lsa.o build/release/api/stream_manager.o build/release/api/system.o build/release/gen/cql3/CqlLexer.o build/release/gen/cql3/CqlParser.o build/release/gen/cassandra_types.o build/release/gen/cassandra_constants.o build/release/gen/Cassandra.o seastar/build/release/libseastar.a libthrift.so libboost_system.so libyaml-cpp.so liblz4.so libz.so libsnappy.so libjsoncpp.so libboost_filesystem.so libcrypt.so libstdc++.so libm.so.6 libgcc_s.so libpthread.so.0 libc.so.6 libc_nonshared.a libgcc_s.so crtend.o crtn.o' (100 runs):

    2364,001445      task-clock (msec)         #    1,000 CPUs utilized            ( +-  0,22% )
            378      context-switches          #    0,160 K/sec                    ( +- 10,91% )
              0      cpu-migrations            #    0,000 K/sec                  
        291 093      page-faults               #    0,123 M/sec                    ( +-  0,02% )
<not supported>      cycles                   
<not supported>      stalled-cycles-frontend  
<not supported>      stalled-cycles-backend   
<not supported>      instructions             
<not supported>      branches                 
<not supported>      branch-misses            

    2,364667169 seconds time elapsed                                            ( +-  0,22% )

linking scylladb after patch:

    2359,664993      task-clock (msec)         #    0,998 CPUs utilized            ( +-  0,20% )
            367      context-switches          #    0,156 K/sec                    ( +- 10,72% )
              0      cpu-migrations            #    0,000 K/sec                  
        291 161      page-faults               #    0,123 M/sec                    ( +-  0,02% )
<not supported>      cycles                   
<not supported>      stalled-cycles-frontend  
<not supported>      stalled-cycles-backend   
<not supported>      instructions             
<not supported>      branches                 
<not supported>      branch-misses            

    2,364471647 seconds time elapsed                                          ( +-  0,26% )

2,364471647 / 2,364667169 == 0.999917

[grimar]

In D32354#738340, @ruiu wrote:

The other way of doing this is to add code to relocateOne, to record whether we handle TLS IE/LD relocations or not. This is probably the least expensive way as it doesn't involve any virtual function calls.

relocateOne is called really late, during writeTo. That is much later than finalizeContents() where we finish updating the DT_* entries. It does not look clean to do it there, probably ?

[ruiu]

ELF/Relocations.cpp

829

This is not beautiful. Since setting DF_STATIC_TLS is not actually that important, I don't want to mess the code up for that feature. What if you add it to getRelExpr? That is also ugly, but it might be acceptable.

[grimar]

• Addressed review comment.

ELF/Relocations.cpp

829

Placing in getRelExpr probably looks better and does not conflict with D32758. Done.

[ruiu]

LGTM

Adding this piece of code to getRelExpr is tricky as getRelExpr is now not only do get but also has side effect. But this seems to be most less intrusive way of doing this.

ELF/Target.cpp

369

I don't think that comment is useful. When you describe it, you want to make it useful for those who are not familiar with details about what you are dealing with. I'd write

// There are 4 different TLS variable models with varying degrees of flexibility and performance. LocalExec and InitialExec models are fast but less-flexible models. They cannot be used for dlopen(). If they are in use, we set DF_STATIC_TLS in the ELF header so that the runtime can reject such DSOs.

[grimar]

ELF/Target.cpp

369

Ok, thanks !

ELF/Relocations.cpp

handleTlsRelocation(uint32_t Type, SymbolBody &Body, InputSectionBase &C,
   if (!Body.isTls())
     return 0;
 
   if (Config->EMachine == EM_ARM)
     return handleARMTlsRelocation<ELFT>(Type, Body, C, Offset, Addend, Expr);
   if (Config->EMachine == EM_MIPS)
     return handleMipsTlsRelocation<ELFT>(Type, Body, C, Offset, Addend, Expr);
 
+  if (Target->isTlsInitialExecRel(Type) || Target->isTlsLocalExecRel(Type))
+    Target->StaticTlsModel = true;
+
   bool IsPreemptible = isPreemptible(Body, Type);
   if (isRelExprOneOf<R_TLSDESC, R_TLSDESC_PAGE, R_TLSDESC_CALL>(Expr) &&
       Config->Shared) {
     if (In<ELFT>::Got->addDynTlsEntry(Body)) {
       uint64_t Off = In<ELFT>::Got->getGlobalDynOffset(Body);
       In<ELFT>::RelaDyn->addReloc({Target->TlsDescRel, In<ELFT>::Got, Off,
                                    !IsPreemptible, &Body, 0});
     }

 // * Write the rest of the file.
 // This would have some drawbacks. For example, we would only know if .rela.dyn
 // is needed after applying relocations. If it is, it will go after rw and rx
 // sections. Given that it is ro, we will need an extra PT_LOAD. This
 // complicates things for the dynamic linker and means we would have to reserve
 // space for the extra PT_LOAD even if we end up not using it.
 template <class ELFT, class RelTy>
 static void scanRelocs(InputSectionBase &Sec, ArrayRef<RelTy> Rels) {
   OffsetGetter GetOffset(Sec);
 
   for (auto I = Rels.begin(), End = Rels.end(); I != End; ++I) {
     const RelTy &Rel = *I;
     SymbolBody &Body = Sec.getFile<ELFT>()->getRelocTargetSym(Rel);
     uint32_t Type = Rel.getType(Config->IsMips64EL);
 
     if (Config->MipsN32Abi) {
       uint32_t Processed;

ELF/SyntheticSections.cpp

if (!Config->RPath.empty())
     add({Config->EnableNewDtags ? DT_RUNPATH : DT_RPATH,
          In<ELFT>::DynStrTab->addString(Config->RPath)});
   for (SharedFile<ELFT> *F : Symtab<ELFT>::X->getSharedFiles())
     if (F->isNeeded())
       add({DT_NEEDED, In<ELFT>::DynStrTab->addString(F->getSoName())});
   if (!Config->SoName.empty())
     add({DT_SONAME, In<ELFT>::DynStrTab->addString(Config->SoName)});
 
+  if (!Config->Shared && !Config->Relocatable)
+    add({DT_DEBUG, (uint64_t)0});
+}
+
+// Add remaining entries to complete .dynamic contents.
+template <class ELFT> void DynamicSection<ELFT>::finalizeContents() {
+  if (this->Size)
+    return; // Already finalized.
+
   // Set DT_FLAGS and DT_FLAGS_1.
   uint32_t DtFlags = 0;
   uint32_t DtFlags1 = 0;
   if (Config->Bsymbolic)
     DtFlags |= DF_SYMBOLIC;
   if (Config->ZNodelete)
     DtFlags1 |= DF_1_NODELETE;
   if (Config->ZNodlopen)
     DtFlags1 |= DF_1_NOOPEN;
   if (Config->ZNow) {
     DtFlags |= DF_BIND_NOW;
     DtFlags1 |= DF_1_NOW;
   }
   if (Config->ZOrigin) {
     DtFlags |= DF_ORIGIN;
     DtFlags1 |= DF_1_ORIGIN;
   }
+  if (Target->StaticTlsModel)
+    DtFlags |= DF_STATIC_TLS;
 
   if (DtFlags)
     add({DT_FLAGS, DtFlags});
   if (DtFlags1)
     add({DT_FLAGS_1, DtFlags1});
 
-  if (!Config->Shared && !Config->Relocatable)
-    add({DT_DEBUG, (uint64_t)0});
-}
-
-// Add remaining entries to complete .dynamic contents.
-template <class ELFT> void DynamicSection<ELFT>::finalizeContents() {
-  if (this->Size)
-    return; // Already finalized.
-
   this->Link = In<ELFT>::DynStrTab->OutSec->SectionIndex;
   if (In<ELFT>::RelaDyn->OutSec->Size > 0) {
     bool IsRela = Config->IsRela;
     add({IsRela ? DT_RELA : DT_REL, In<ELFT>::RelaDyn});
     add({IsRela ? DT_RELASZ : DT_RELSZ, In<ELFT>::RelaDyn->OutSec->Size});
     add({IsRela ? DT_RELAENT : DT_RELENT,
          uint64_t(IsRela ? sizeof(Elf_Rela) : sizeof(Elf_Rel))});
 

ELF/Target.h

 namespace elf {
 class InputFile;
 class SymbolBody;
 
 class TargetInfo {
 public:
   virtual bool isTlsInitialExecRel(uint32_t Type) const;
   virtual bool isTlsLocalDynamicRel(uint32_t Type) const;
+  virtual bool isTlsLocalExecRel(uint32_t Type) const;
   virtual bool isPicRel(uint32_t Type) const { return true; }
   virtual uint32_t getDynRel(uint32_t Type) const { return Type; }
   virtual void writeGotPltHeader(uint8_t *Buf) const {}
   virtual void writeGotPlt(uint8_t *Buf, const SymbolBody &S) const {};
   virtual void writeIgotPlt(uint8_t *Buf, const SymbolBody &S) const;
   virtual int64_t getImplicitAddend(const uint8_t *Buf, uint32_t Type) const;
 
   // If lazy binding is supported, the first entry of the PLT has code
public:
 
   // At least on x86_64 positions 1 and 2 are used by the first plt entry
   // to support lazy loading.
   unsigned GotPltHeaderEntriesNum = 3;
 
   // Set to 0 for variant 2
   unsigned TcbSize = 0;
 
+  // Indicates that output contains code that use static TLS model.
+  bool StaticTlsModel = false;
+
   bool NeedsThunks = false;
 
   // A 4-byte field corresponding to one or more trap instructions, used to pad
   // executable OutputSections.
   uint32_t TrapInstr = 0;
 
   virtual RelExpr adjustRelaxExpr(uint32_t Type, const uint8_t *Data,
                                   RelExpr Expr) const;

ELF/Target.cpp

 public:
   X86TargetInfo();
   RelExpr getRelExpr(uint32_t Type, const SymbolBody &S,
                      const uint8_t *Loc) const override;
   int64_t getImplicitAddend(const uint8_t *Buf, uint32_t Type) const override;
   void writeGotPltHeader(uint8_t *Buf) const override;
   uint32_t getDynRel(uint32_t Type) const override;
   bool isTlsLocalDynamicRel(uint32_t Type) const override;
+  bool isTlsLocalExecRel(uint32_t Type) const override;
   bool isTlsInitialExecRel(uint32_t Type) const override;
   void writeGotPlt(uint8_t *Buf, const SymbolBody &S) const override;
   void writeIgotPlt(uint8_t *Buf, const SymbolBody &S) const override;
   void writePltHeader(uint8_t *Buf) const override;
   void writePlt(uint8_t *Buf, uint64_t GotPltEntryAddr, uint64_t PltEntryAddr,
                 int32_t Index, unsigned RelOff) const override;
   void relocateOne(uint8_t *Loc, uint32_t Type, uint64_t Val) const override;
 
bool TargetInfo::needsThunk(RelExpr Expr, uint32_t RelocType,
                             const InputFile *File, const SymbolBody &S) const {
   return false;
 }
 
 bool TargetInfo::isTlsInitialExecRel(uint32_t Type) const { return false; }
 
 bool TargetInfo::isTlsLocalDynamicRel(uint32_t Type) const { return false; }
 
+bool TargetInfo::isTlsLocalExecRel(uint32_t Type) const { return false; }
+
 void TargetInfo::writeIgotPlt(uint8_t *Buf, const SymbolBody &S) const {
   writeGotPlt(Buf, S);
 }
 
 RelExpr TargetInfo::adjustRelaxExpr(uint32_t Type, const uint8_t *Data,
                                     RelExpr Expr) const {
   return Expr;
 }
X86TargetInfo::X86TargetInfo() {
   PltHeaderSize = 16;
   TlsGdRelaxSkip = 2;
   // 0xCC is the "int3" (call debug exception handler) instruction.
   TrapInstr = 0xcccccccc;
 }
 
 RelExpr X86TargetInfo::getRelExpr(uint32_t Type, const SymbolBody &S,
                                   const uint8_t *Loc) const {
   switch (Type) {
   case R_386_8:
   case R_386_16:
   case R_386_32:
   case R_386_TLS_LDO_32:
     return R_ABS;
   case R_386_TLS_GD:
     return R_TLSGD;
   case R_386_TLS_LDM:
if (Type == R_386_TLS_LE_32)
     return R_386_TLS_TPOFF32;
   return Type;
 }
 
 bool X86TargetInfo::isTlsLocalDynamicRel(uint32_t Type) const {
   return Type == R_386_TLS_LDO_32 || Type == R_386_TLS_LDM;
 }
 
+bool X86TargetInfo::isTlsLocalExecRel(uint32_t Type) const {
+  return Type == R_386_TLS_LE_32 || Type == R_386_TLS_LE;
+}
+
 bool X86TargetInfo::isTlsInitialExecRel(uint32_t Type) const {
   return Type == R_386_TLS_IE || Type == R_386_TLS_GOTIE;
 }
 
 void X86TargetInfo::writePltHeader(uint8_t *Buf) const {
   if (Config->Pic) {
     const uint8_t V[] = {
         0xff, 0xb3, 0x04, 0x00, 0x00, 0x00, // pushl GOTPLT+4(%ebx)

test/ELF/Inputs/i386-static-tls-model1.s

+.section ".tdata", "awT", @progbits
+.globl var
+var:
+
+.section .foo, "aw"
+.global _start
+_start:
+ movl $var@tpoff, %edx # R_386_TLS_LE_32
+ movl %gs:0, %ecx
+ subl %edx, %eax

test/ELF/Inputs/i386-static-tls-model2.s

+.section ".tdata", "awT", @progbits
+.globl var
+var:
+
+.section .foo, "aw"
+.global _start
+_start: 
+ movl %gs:0, %eax
+ addl var@gotntpoff(%ebx),%eax # R_386_TLS_GOTIE

test/ELF/Inputs/i386-static-tls-model3.s

+.section ".tdata", "awT", @progbits
+.globl var
+var:
+
+.section .foo, "aw"
+.global _start
+_start:
+ movl %gs:0, %eax
+ addl var@indntpoff, %eax #R_386_TLS_IE

test/ELF/Inputs/i386-static-tls-model4.s

+.section ".tdata", "awT", @progbits
+.globl var
+var:
+
+.section .foo, "aw"
+.global _start
+_start:
+ movl %gs:0, %eax
+ leal var@ntpoff(%eax), %eax #R_386_TLS_LE

test/ELF/aarch64-tls-ie.s

 #RELOC:      Section {
 #RELOC:        Index:
 #RELOC:        Name: .got
 #RELOC-NEXT:   Type: SHT_PROGBITS
 #RELOC-NEXT:   Flags [
 #RELOC-NEXT:     SHF_ALLOC
 #RELOC-NEXT:     SHF_WRITE
 #RELOC-NEXT:   ]
-#RELOC-NEXT:   Address: 0x300B0
-#RELOC-NEXT:   Offset: 0x200B0
+#RELOC-NEXT:   Address: 0x300C0
+#RELOC-NEXT:   Offset: 0x200C0
 #RELOC-NEXT:   Size: 16
 #RELOC-NEXT:   Link: 0
 #RELOC-NEXT:   Info: 0
 #RELOC-NEXT:   AddressAlignment: 8
 #RELOC-NEXT:   EntrySize: 0
 #RELOC-NEXT: }
 #RELOC:      Relocations [
 #RELOC-NEXT:  Section ({{.*}}) .rela.dyn {
-#RELOC-NEXT:    0x300B8 R_AARCH64_TLS_TPREL64 bar 0x0
-#RELOC-NEXT:    0x300B0 R_AARCH64_TLS_TPREL64 foo 0x0
+#RELOC-NEXT:    0x300C8 R_AARCH64_TLS_TPREL64 bar 0x0
+#RELOC-NEXT:    0x300C0 R_AARCH64_TLS_TPREL64 foo 0x0
 #RELOC-NEXT:  }
 #RELOC-NEXT:]
 
-# Page(0x300B0) - Page(0x20000) = 0x10000 = 65536
-# 0x300B0 & 0xff8 = 0xB0 = 176
+# Page(0x300C0) - Page(0x20000) = 0x10000 = 65536
+# 0x300C0 & 0xff8 = 0xC0 = 192
 # Page(0x300B8) - Page(0x20000) = 0x10000 = 65536
-# 0x300B8 & 0xff8 = 0xB8 = 184
+# 0x300C0 & 0xff8 = 0xC8 = 200
 #CHECK: Disassembly of section .text:
 #CHECK: _start:
 #CHECK:  20000: 80 00 00 90 adrp x0, #65536
-#CHECK:  20004: 00 58 40 f9 ldr  x0, [x0, #176]
+#CHECK:  20004: 00 60 40 f9 ldr  x0, [x0, #192]
 #CHECK:  20008: 80 00 00 90 adrp x0, #65536
-#CHECK:  2000c: 00 5c 40 f9 ldr  x0, [x0, #184]
+#CHECK:  2000c: 00 64 40 f9 ldr  x0, [x0, #200]
 
 .globl _start
 _start:
  adrp x0, :gottprel:foo
  ldr x0, [x0, #:gottprel_lo12:foo]
 
  adrp x0, :gottprel:bar
  ldr x0, [x0, #:gottprel_lo12:bar]

test/ELF/i386-static-tls-model.s

+# REQUIRES: x86
+
+# RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %S/Inputs/i386-static-tls-model1.s -o %t.o
+# RUN: ld.lld %t.o -o %t1 -shared
+# RUN: llvm-readobj  -dynamic-table %t1 | FileCheck %s
+
+# RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %S/Inputs/i386-static-tls-model2.s -o %t.o
+# RUN: ld.lld %t.o -o %t2 -shared
+# RUN: llvm-readobj  -dynamic-table %t2 | FileCheck %s
+
+# RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %S/Inputs/i386-static-tls-model3.s -o %t.o
+# RUN: ld.lld %t.o -o %t3 -shared
+# RUN: llvm-readobj  -dynamic-table %t3 | FileCheck %s
+
+# RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %S/Inputs/i386-static-tls-model4.s -o %t.o
+# RUN: ld.lld %t.o -o %t4 -shared
+# RUN: llvm-readobj  -dynamic-table %t4 | FileCheck %s
+
+# CHECK: DynamicSection [
+# CHECK: FLAGS STATIC_TLS

test/ELF/i386-tls-ie-shared.s

 // RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %s -o %t.o
 // RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %p/Inputs/tls-opt-iele-i686-nopic.s -o %tso.o
 // RUN: ld.lld -shared %tso.o -o %tso
 // RUN: ld.lld -shared %t.o %tso -o %t1
 // RUN: llvm-readobj -s -r -d %t1 | FileCheck --check-prefix=GOTRELSHARED %s
 // RUN: llvm-objdump -d %t1 | FileCheck --check-prefix=DISASMSHARED %s
 
 // GOTRELSHARED:     Section {
 // GOTRELSHARED:      Index: 8
 // GOTRELSHARED:      Name: .got
 // GOTRELSHARED-NEXT:   Type: SHT_PROGBITS
 // GOTRELSHARED-NEXT:   Flags [
 // GOTRELSHARED-NEXT:     SHF_ALLOC
 // GOTRELSHARED-NEXT:     SHF_WRITE
 // GOTRELSHARED-NEXT:   ]
-// GOTRELSHARED-NEXT:   Address: 0x1058
-// GOTRELSHARED-NEXT:   Offset: 0x1058
+// GOTRELSHARED-NEXT:   Address: 0x1060
+// GOTRELSHARED-NEXT:   Offset: 0x1060
 // GOTRELSHARED-NEXT:   Size: 16
 // GOTRELSHARED-NEXT:   Link: 0
 // GOTRELSHARED-NEXT:   Info: 0
 // GOTRELSHARED-NEXT:   AddressAlignment: 4
 // GOTRELSHARED-NEXT:   EntrySize: 0
 // GOTRELSHARED-NEXT: }
 // GOTRELSHARED:      Relocations [
 // GOTRELSHARED-NEXT:   Section ({{.*}}) .rel.dyn {
 // GOTRELSHARED-NEXT:     0x2002 R_386_RELATIVE - 0x0
 // GOTRELSHARED-NEXT:     0x200A R_386_RELATIVE - 0x0
 // GOTRELSHARED-NEXT:     0x2013 R_386_RELATIVE - 0x0
 // GOTRELSHARED-NEXT:     0x201C R_386_RELATIVE - 0x0
 // GOTRELSHARED-NEXT:     0x2024 R_386_RELATIVE - 0x0
 // GOTRELSHARED-NEXT:     0x202D R_386_RELATIVE - 0x0
 // GOTRELSHARED-NEXT:     0x2036 R_386_RELATIVE - 0x0
 // GOTRELSHARED-NEXT:     0x203F R_386_RELATIVE - 0x0
-// GOTRELSHARED-NEXT:     0x1058 R_386_TLS_TPOFF tlslocal0 0x0
-// GOTRELSHARED-NEXT:     0x105C R_386_TLS_TPOFF tlslocal1 0x0
-// GOTRELSHARED-NEXT:     0x1060 R_386_TLS_TPOFF tlsshared0 0x0
-// GOTRELSHARED-NEXT:     0x1064 R_386_TLS_TPOFF tlsshared1 0x0
+// GOTRELSHARED-NEXT:     0x1060 R_386_TLS_TPOFF tlslocal0 0x0
+// GOTRELSHARED-NEXT:     0x1064 R_386_TLS_TPOFF tlslocal1 0x0
+// GOTRELSHARED-NEXT:     0x1068 R_386_TLS_TPOFF tlsshared0 0x0
+// GOTRELSHARED-NEXT:     0x106C R_386_TLS_TPOFF tlsshared1 0x0
 // GOTRELSHARED-NEXT:   }
 // GOTRELSHARED-NEXT: ]
 // GOTRELSHARED:      0x6FFFFFFA RELCOUNT             8
 
 // DISASMSHARED:       Disassembly of section test:
 // DISASMSHARED-NEXT:  _start:
-// (.got)[0] = 0x2050 = 8272
-// (.got)[1] = 0x2054 = 8276
-// (.got)[2] = 0x2058 = 8280
-// (.got)[3] = 0x205C = 8284
-// DISASMSHARED-NEXT:  2000: 8b 0d 58 10 00 00   movl  4184, %ecx
-// DISASMSHARED-NEXT:  2006: 65 8b 01  movl  %gs:(%ecx), %eax
-// DISASMSHARED-NEXT:  2009: a1 58 10 00 00  movl  4184, %eax
-// DISASMSHARED-NEXT:  200e: 65 8b 00  movl  %gs:(%eax), %eax
-// DISASMSHARED-NEXT:  2011: 03 0d 58 10 00 00   addl  4184, %ecx
-// DISASMSHARED-NEXT:  2017: 65 8b 01  movl  %gs:(%ecx), %eax
-// DISASMSHARED-NEXT:  201a: 8b 0d 5c 10 00 00   movl  4188, %ecx
-// DISASMSHARED-NEXT:  2020: 65 8b 01  movl  %gs:(%ecx), %eax
-// DISASMSHARED-NEXT:  2023: a1 5c 10 00 00  movl  4188, %eax
-// DISASMSHARED-NEXT:  2028: 65 8b 00  movl  %gs:(%eax), %eax
-// DISASMSHARED-NEXT:  202b: 03 0d 5c 10 00 00   addl  4188, %ecx
-// DISASMSHARED-NEXT:  2031: 65 8b 01  movl  %gs:(%ecx), %eax
-// DISASMSHARED-NEXT:  2034: 8b 0d 60 10 00 00   movl  4192, %ecx
-// DISASMSHARED-NEXT:  203a: 65 8b 01  movl  %gs:(%ecx), %eax
-// DISASMSHARED-NEXT:  203d: 03 0d 64 10 00 00   addl  4196, %ecx
-// DISASMSHARED-NEXT:  2043: 65 8b 01  movl  %gs:(%ecx), %eax
+// (.got)[0] = 0x1060 = 4192
+// (.got)[1] = 0x1064 = 4196
+// (.got)[2] = 0x1068 = 4200
+// (.got)[3] = 0x106C = 4204
+// DISASMSHARED-NEXT:  2000: {{.*}}  movl  4192, %ecx
+// DISASMSHARED-NEXT:  2006: {{.*}}  movl  %gs:(%ecx), %eax
+// DISASMSHARED-NEXT:  2009: {{.*}}  movl  4192, %eax
+// DISASMSHARED-NEXT:  200e: {{.*}}  movl  %gs:(%eax), %eax
+// DISASMSHARED-NEXT:  2011: {{.*}}  addl  4192, %ecx
+// DISASMSHARED-NEXT:  2017: {{.*}}  movl  %gs:(%ecx), %eax
+// DISASMSHARED-NEXT:  201a: {{.*}}  movl  4196, %ecx
+// DISASMSHARED-NEXT:  2020: {{.*}}  movl  %gs:(%ecx), %eax
+// DISASMSHARED-NEXT:  2023: {{.*}}  movl  4196, %eax
+// DISASMSHARED-NEXT:  2028: {{.*}}  movl  %gs:(%eax), %eax
+// DISASMSHARED-NEXT:  202b: {{.*}}  addl  4196, %ecx
+// DISASMSHARED-NEXT:  2031: {{.*}}  movl  %gs:(%ecx), %eax
+// DISASMSHARED-NEXT:  2034: {{.*}}  movl  4200, %ecx
+// DISASMSHARED-NEXT:  203a: {{.*}}  movl  %gs:(%ecx), %eax
+// DISASMSHARED-NEXT:  203d: {{.*}}  addl  4204, %ecx
+// DISASMSHARED-NEXT:  2043: {{.*}}  movl  %gs:(%ecx), %eax
 
 .type tlslocal0,@object
 .section .tbss,"awT",@nobits
 .globl tlslocal0
 .align 4
 tlslocal0:
  .long 0
  .size tlslocal0, 4

test/ELF/tls-dynamic-i686.s

 addl tls1@gotntpoff(%ebx),%eax
 
 // CHECK:      Name: .got (
 // CHECK-NEXT: Type: SHT_PROGBITS
 // CHECK-NEXT: Flags [
 // CHECK-NEXT:   SHF_ALLOC
 // CHECK-NEXT:   SHF_WRITE
 // CHECK-NEXT: ]
-// CHECK-NEXT: Address: 0x3068
-// CHECK-NEXT: Offset: 0x3068
+// CHECK-NEXT: Address: 0x3070
+// CHECK-NEXT: Offset: 0x3070
 // CHECK-NEXT: Size: 32
 // CHECK-NEXT: Link: 0
 // CHECK-NEXT: Info: 0
 // CHECK-NEXT: AddressAlignment: 4
 // CHECK-NEXT: EntrySize: 0
 
 // CHECK: Relocations [
 // CHECK:      Section ({{.+}}) .rel.dyn {
-// CHECK-NEXT: 0x3078 R_386_TLS_DTPMOD32 - 0x0
-// CHECK-NEXT: 0x3068 R_386_TLS_DTPMOD32 tls0 0x0
-// CHECK-NEXT: 0x306C R_386_TLS_DTPOFF32 tls0 0x0
-// CHECK-NEXT: 0x3080 R_386_TLS_TPOFF tls0 0x0
-// CHECK-NEXT: 0x3070 R_386_TLS_DTPMOD32 tls1 0x0
-// CHECK-NEXT: 0x3074 R_386_TLS_DTPOFF32 tls1 0x0
-// CHECK-NEXT: 0x3084 R_386_TLS_TPOFF tls1 0x0
+// CHECK-NEXT: 0x3080 R_386_TLS_DTPMOD32 - 0x0
+// CHECK-NEXT: 0x3070 R_386_TLS_DTPMOD32 tls0 0x0
+// CHECK-NEXT: 0x3074 R_386_TLS_DTPOFF32 tls0 0x0
+// CHECK-NEXT: 0x3088 R_386_TLS_TPOFF tls0 0x0
+// CHECK-NEXT: 0x3078 R_386_TLS_DTPMOD32 tls1 0x0
+// CHECK-NEXT: 0x307C R_386_TLS_DTPOFF32 tls1 0x0
+// CHECK-NEXT: 0x308C R_386_TLS_TPOFF tls1 0x0
 // CHECK-NEXT: }
 
 // DIS:      Disassembly of section .text:
 // DIS-NEXT: _start:
 // General dynamic model:
 // -32 and -24 are first and second GOT entries offsets.
 // Each one is a pair of records.
-// DIS-NEXT: 1000: 8d 04 1d e0 ff ff ff  leal -32(,%ebx), %eax
-// DIS-NEXT: 1007: e8 64 00 00 00        calll 100
-// DIS-NEXT: 100c: 8d 04 1d e8 ff ff ff  leal -24(,%ebx), %eax
-// DIS-NEXT: 1013: e8 58 00 00 00        calll 88
+// DIS-NEXT: 1000: {{.*}}  leal -32(,%ebx), %eax
+// DIS-NEXT: 1007: {{.*}}  calll 100
+// DIS-NEXT: 100c: {{.*}}  leal -24(,%ebx), %eax
+// DIS-NEXT: 1013: {{.*}}  calll 88
 // Local dynamic model:
 // -16 is a local module tls index offset.
-// DIS-NEXT: 1018: 8d 83 f0 ff ff ff leal -16(%ebx), %eax
-// DIS-NEXT: 101e: e8 4d 00 00 00    calll 77
-// DIS-NEXT: 1023: 8d 90 08 00 00 00 leal 8(%eax), %edx
-// DIS-NEXT: 1029: 8d 83 f0 ff ff ff leal -16(%ebx), %eax
-// DIS-NEXT: 102f: e8 3c 00 00 00    calll 60
-// DIS-NEXT: 1034: 8d 90 08 00 00 00 leal 8(%eax), %edx
+// DIS-NEXT: 1018: {{.*}}  leal -16(%ebx), %eax
+// DIS-NEXT: 101e: {{.*}}  calll 77
+// DIS-NEXT: 1023: {{.*}}  leal 8(%eax), %edx
+// DIS-NEXT: 1029: {{.*}}  leal -16(%ebx), %eax
+// DIS-NEXT: 102f: {{.*}}  calll 60
+// DIS-NEXT: 1034: {{.*}}  leal 8(%eax), %edx
 // Initial exec model:
-// DIS-NEXT: 103a: 65 a1 00 00 00 00 movl %gs:0, %eax
-// DIS-NEXT: 1040: 03 83 f8 ff ff ff addl -8(%ebx), %eax
-// DIS-NEXT: 1046: 65 a1 00 00 00 00 movl %gs:0, %eax
-// DIS-NEXT: 104c: 03 83 fc ff ff ff addl -4(%ebx), %eax
+// DIS-NEXT: 103a: {{.*}}  movl %gs:0, %eax
+// DIS-NEXT: 1040: {{.*}}  addl -8(%ebx), %eax
+// DIS-NEXT: 1046: {{.*}}  movl %gs:0, %eax
+// DIS-NEXT: 104c: {{.*}}  addl -4(%ebx), %eax

test/ELF/tls-dynamic.s

 
 // CHECK:      Sections [
 // CHECK:          Name: .got (
 // CHECK-NEXT:     Type: SHT_PROGBITS
 // CHECK-NEXT:     Flags [
 // CHECK-NEXT:       SHF_ALLOC
 // CHECK-NEXT:       SHF_WRITE
 // CHECK-NEXT:     ]
-// CHECK-NEXT:     Address: 0x30D0
+// CHECK-NEXT:     Address: 0x30E0
 // CHECK-NEXT:     Offset:
 // CHECK-NEXT:     Size: 40
 
 // CHECK:      Relocations [
 // CHECK:        Section ({{.+}}) .rela.dyn {
-// CHECK-NEXT:     0x30D0 R_X86_64_DTPMOD64 - 0x0
-// CHECK-NEXT:     0x30E0 R_X86_64_DTPMOD64 c 0x0
-// CHECK-NEXT:     0x30E8 R_X86_64_DTPOFF64 c 0x0
-// CHECK-NEXT:     0x30F0 R_X86_64_TPOFF64 c 0x0
+// CHECK-NEXT:     0x30E0 R_X86_64_DTPMOD64 - 0x0
+// CHECK-NEXT:     0x30F0 R_X86_64_DTPMOD64 c 0x0
+// CHECK-NEXT:     0x30F8 R_X86_64_DTPOFF64 c 0x0
+// CHECK-NEXT:     0x3100 R_X86_64_TPOFF64 c 0x0
 // CHECK-NEXT:   }
 
-// 4297 = (0x20D0 + -4) - (0x1000 + 3) // PC relative offset to got entry.
-// 4285 = (0x20D0 + -4) - (0x100c + 3) // PC relative offset to got entry.
-// 4267 = (0x20E0 + -4) - (0x102e + 3) // PC relative offset to got entry.
-// 4263 = (0x20F0 + -4) - (0x1042 + 3) // PC relative offset to got entry.
+// 8409 = (0x30E0 + -4) - (0x1000 + 3) // PC relative offset to got entry.
+// 8397 = (0x30F0 + -4) - (0x100c + 3) // PC relative offset to got entry.
+// 8379 = (0x30F8 + -4) - (0x102e + 3) // PC relative offset to got entry.
+// 8375 = (0x3100 + -4) - (0x1042 + 3) // PC relative offset to got entry.
 
 // DIS:      Disassembly of section .text:
 // DIS-NEXT: .text:
-// DIS-NEXT:     1000: {{.+}} leaq    8393(%rip), %rdi
+// DIS-NEXT:     1000: {{.+}} leaq    8409(%rip), %rdi
 // DIS-NEXT:     1007: {{.+}} callq
-// DIS-NEXT:     100c: {{.+}} leaq    8381(%rip), %rdi
+// DIS-NEXT:     100c: {{.+}} leaq    8397(%rip), %rdi
 // DIS-NEXT:     1013: {{.+}} callq
 // DIS-NEXT:     1018: {{.+}} leaq    (%rax), %rcx
 // DIS-NEXT:     101f: {{.+}} leaq    4(%rax), %rcx
 // DIS-NEXT:     1026: 04 00
 // DIS-NEXT:     1028: 00 00
 // DIS-NEXT:     102a: 00 00
 // DIS-NEXT:     102c: 00 00
-// DIS-NEXT:     102e: {{.+}} leaq    8363(%rip), %rdi
+// DIS-NEXT:     102e: {{.+}} leaq    8379(%rip), %rdi
 // DIS-NEXT:     1035: {{.+}} callq
 // DIS-NEXT:     103b: {{.+}} leaq    (%rax), %rcx
-// DIS-NEXT:     1042: {{.+}} movq    8359(%rip), %rax
+// DIS-NEXT:     1042: {{.+}} movq    8375(%rip), %rax
 // DIS-NEXT:     1049: {{.+}} movq    %fs:(%rax), %rax
 // DIS-NEXT:     104d: {{.+}} movabsq $0, %rax
 // DIS-NEXT:     1057: {{.+}} movabsq $4, %rax
 // DIS-NEXT:     1061: {{.+}} movabsq $0, %rax

test/ELF/tls-got.s

 // RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t1.o
 // RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %p/Inputs/tls-got.s -o %t2.o
 // RUN: ld.lld -shared %t2.o -o %t2.so
 // RUN: ld.lld -e main %t1.o %t2.so -o %t3
 // RUN: llvm-readobj -s -r %t3 | FileCheck %s
 // RUN: llvm-objdump -d %t3 | FileCheck --check-prefix=DISASM %s
 
 // CHECK:      Section {
 // CHECK:      Index: 8
 // CHECK-NEXT: Name: .got
 // CHECK-NEXT: Type: SHT_PROGBITS
 // CHECK-NEXT: Flags [
 // CHECK-NEXT:   SHF_ALLOC
 // CHECK-NEXT:   SHF_WRITE
 // CHECK-NEXT: ]
 // CHECK-NEXT: Address: [[ADDR:.*]]
-// CHECK-NEXT: Offset: 0x20B0
+// CHECK-NEXT: Offset: 0x20C0
 // CHECK-NEXT: Size: 16
 // CHECK-NEXT: Link: 0
 // CHECK-NEXT: Info: 0
 // CHECK-NEXT: AddressAlignment: 8
 // CHECK-NEXT: EntrySize: 0
 // CHECK-NEXT: }
 
 // CHECK:      Relocations [
 // CHECK-NEXT:   Section (4) .rela.dyn {
-// CHECK-NEXT:     0x2020B8 R_X86_64_TPOFF64 tls0 0x0
+// CHECK-NEXT:     0x2020C8 R_X86_64_TPOFF64 tls0 0x0
 // CHECK-NEXT:     [[ADDR]] R_X86_64_TPOFF64 tls1 0x0
 // CHECK-NEXT:   }
 // CHECK-NEXT: ]
 
-//0x201000 + 4249 + 7 = 0x2020B0
-//0x20100A + 4247 + 7 = 0x2020B8
-//0x201014 + 4237 + 7 = 0x2020B8
+//0x201000 + 4281 + 7 = 0x2020C0
+//0x20100A + 4279 + 7 = 0x2020C8
+//0x201014 + 4269 + 7 = 0x2020C8
 //DISASM:      Disassembly of section .text:
 //DISASM-NEXT: main:
-//DISASM-NEXT: 201000: 48 8b 05 a9 10 00 00 movq 4265(%rip), %rax
-//DISASM-NEXT: 201007: 64 8b 00 movl %fs:(%rax), %eax
-//DISASM-NEXT: 20100a: 48 8b 05 a7 10 00 00 movq 4263(%rip), %rax
-//DISASM-NEXT: 201011: 64 8b 00 movl %fs:(%rax), %eax
-//DISASM-NEXT: 201014: 48 8b 05 9d 10 00 00 movq 4253(%rip), %rax
-//DISASM-NEXT: 20101b: 64 8b 00 movl %fs:(%rax), %eax
+//DISASM-NEXT: 201000: {{.*}} movq 4281(%rip), %rax
+//DISASM-NEXT: 201007: {{.*}} movl %fs:(%rax), %eax
+//DISASM-NEXT: 20100a: {{.*}} movq 4279(%rip), %rax
+//DISASM-NEXT: 201011: {{.*}} movl %fs:(%rax), %eax
+//DISASM-NEXT: 201014: {{.*}} movq 4269(%rip), %rax
+//DISASM-NEXT: 20101b: {{.*}} movl %fs:(%rax), %eax
 //DISASM-NEXT: 20101e: c3 retq
 
 .section .tdata,"awT",@progbits
 
 .text
  .globl main
  .align 16, 0x90
  .type main,@function
 main:
  movq tls1@GOTTPOFF(%rip), %rax
  movl %fs:0(%rax), %eax
  movq tls0@GOTTPOFF(%rip), %rax
  movl %fs:0(%rax), %eax
  movq tls0@GOTTPOFF(%rip), %rax
  movl %fs:0(%rax), %eax
  ret

test/ELF/tls-initial-exec-local.s

 // REQUIRES: x86
 // RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o
 // RUN: ld.lld -shared %t.o -o %t
 // RUN: llvm-readobj -r -s %t | FileCheck %s
 // RUN: llvm-objdump -d %t | FileCheck --check-prefix=DISASM %s
 
 // CHECK:      Name: .got
 // CHECK-NEXT: Type: SHT_PROGBITS
 // CHECK-NEXT: Flags [
 // CHECK-NEXT:   SHF_ALLOC (0x2)
 // CHECK-NEXT:   SHF_WRITE (0x1)
 // CHECK-NEXT: ]
-// CHECK-NEXT: Address: 0x2090
+// CHECK-NEXT: Address: 0x20A0
 
 // CHECK:      Relocations [
 // CHECK-NEXT:   Section ({{.*}}) .rela.dyn {
-// CHECK-NEXT:     0x2090 R_X86_64_TPOFF64 - 0x0
-// CHECK-NEXT:     0x2098 R_X86_64_TPOFF64 - 0x4
+// CHECK-NEXT:     0x20A0 R_X86_64_TPOFF64 - 0x0
+// CHECK-NEXT:     0x20A8 R_X86_64_TPOFF64 - 0x4
 // CHECK-NEXT:   }
 // CHECK-NEXT: ]
 
-// 0x1007 + 4233 = 0x2090
-// 0x100e + 4234 = 0x2098
+// 0x1007 + 4249 = 0x20A0
+// 0x100e + 4250 = 0x20A8
 // DISASM:      Disassembly of section .text:
 // DISASM-NEXT: .text:
-// DISASM-NEXT:  1000: {{.*}} addq      4233(%rip), %rax
-// DISASM-NEXT:  1007: {{.*}} addq      4234(%rip), %rax
+// DISASM-NEXT:  1000: {{.*}} addq      4249(%rip), %rax
+// DISASM-NEXT:  1007: {{.*}} addq      4250(%rip), %rax
 
         addq    foo@GOTTPOFF(%rip), %rax
         addq    bar@GOTTPOFF(%rip), %rax
 
         .section        .tbss,"awT",@nobits
 foo:
         .long 0
 bar:
         .long 0

test/ELF/tls-opt-iele-i686-nopic.s

 // RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %s -o %t.o
 // RUN: llvm-mc -filetype=obj -triple=i686-pc-linux %p/Inputs/tls-opt-iele-i686-nopic.s -o %tso.o
 // RUN: ld.lld -shared %tso.o -o %tso
 // RUN: ld.lld %t.o %tso -o %t1
 // RUN: llvm-readobj -s -r %t1 | FileCheck --check-prefix=GOTREL %s
 // RUN: llvm-objdump -d %t1 | FileCheck --check-prefix=DISASM %s
 
 // GOTREL:      Section {
 // GOTREL:        Index:
 // GOTREL:        Name: .got
 // GOTREL-NEXT:   Type: SHT_PROGBITS
 // GOTREL-NEXT:   Flags [
 // GOTREL-NEXT:     SHF_ALLOC
 // GOTREL-NEXT:     SHF_WRITE
 // GOTREL-NEXT:   ]
-// GOTREL-NEXT:   Address: 0x12058
-// GOTREL-NEXT:   Offset: 0x2058
+// GOTREL-NEXT:   Address: 0x12060
+// GOTREL-NEXT:   Offset: 0x2060
 // GOTREL-NEXT:   Size: 8
 // GOTREL-NEXT:   Link: 0
 // GOTREL-NEXT:   Info: 0
 // GOTREL-NEXT:   AddressAlignment: 4
 // GOTREL-NEXT:   EntrySize: 0
 // GOTREL-NEXT: }
 // GOTREL:      Relocations [
 // GOTREL-NEXT: Section ({{.*}}) .rel.dyn {
-// GOTREL-NEXT:   0x12058 R_386_TLS_TPOFF tlsshared0 0x0
-// GOTREL-NEXT:   0x1205C R_386_TLS_TPOFF tlsshared1 0x0
+// GOTREL-NEXT:   0x12060 R_386_TLS_TPOFF tlsshared0 0x0
+// GOTREL-NEXT:   0x12064 R_386_TLS_TPOFF tlsshared1 0x0
 // GOTREL-NEXT:  }
 // GOTREL-NEXT: ]
 
 // DISASM:      Disassembly of section .text:
 // DISASM-NEXT: _start:
 // 4294967288 = 0xFFFFFFF8
 // 4294967292 = 0xFFFFFFFC
-// 73808 = (.got)[0] = 0x12058
-// 73812 = (.got)[1] = 0x1205C
-// DISASM-NEXT: 11000: c7 c1 f8 ff ff ff movl $4294967288, %ecx
-// DISASM-NEXT: 11006: 65 8b 01          movl %gs:(%ecx), %eax
-// DISASM-NEXT: 11009: b8 f8 ff ff ff    movl $4294967288, %eax
-// DISASM-NEXT: 1100e: 65 8b 00          movl %gs:(%eax), %eax
-// DISASM-NEXT: 11011: 81 c1 f8 ff ff ff addl $4294967288, %ecx
-// DISASM-NEXT: 11017: 65 8b 01          movl %gs:(%ecx), %eax
-// DISASM-NEXT: 1101a: c7 c1 fc ff ff ff movl $4294967292, %ecx
-// DISASM-NEXT: 11020: 65 8b 01          movl %gs:(%ecx), %eax
-// DISASM-NEXT: 11023: b8 fc ff ff ff    movl $4294967292, %eax
-// DISASM-NEXT: 11028: 65 8b 00          movl %gs:(%eax), %eax
-// DISASM-NEXT: 1102b: 81 c1 fc ff ff ff addl $4294967292, %ecx
-// DISASM-NEXT: 11031: 65 8b 01          movl %gs:(%ecx), %eax
-// DISASM-NEXT: 11034: 8b 0d 58 20 01 00 movl 73816, %ecx
-// DISASM-NEXT: 1103a: 65 8b 01          movl %gs:(%ecx), %eax
-// DISASM-NEXT: 1103d: 03 0d 5c 20 01 00 addl 73820, %ecx
-// DISASM-NEXT: 11043: 65 8b 01          movl %gs:(%ecx), %eax
+// 73824 = (.got)[0] = 0x12060
+// 73828 = (.got)[1] = 0x12064
+// DISASM-NEXT: 11000: {{.*}} movl $4294967288, %ecx
+// DISASM-NEXT: 11006: {{.*}} movl %gs:(%ecx), %eax
+// DISASM-NEXT: 11009: {{.*}} movl $4294967288, %eax
+// DISASM-NEXT: 1100e: {{.*}} movl %gs:(%eax), %eax
+// DISASM-NEXT: 11011: {{.*}} addl $4294967288, %ecx
+// DISASM-NEXT: 11017: {{.*}} movl %gs:(%ecx), %eax
+// DISASM-NEXT: 1101a: {{.*}} movl $4294967292, %ecx
+// DISASM-NEXT: 11020: {{.*}} movl %gs:(%ecx), %eax
+// DISASM-NEXT: 11023: {{.*}} movl $4294967292, %eax
+// DISASM-NEXT: 11028: {{.*}} movl %gs:(%eax), %eax
+// DISASM-NEXT: 1102b: {{.*}} addl $4294967292, %ecx
+// DISASM-NEXT: 11031: {{.*}} movl %gs:(%ecx), %eax
+// DISASM-NEXT: 11034: {{.*}} movl 73824, %ecx
+// DISASM-NEXT: 1103a: {{.*}} movl %gs:(%ecx), %eax
+// DISASM-NEXT: 1103d: {{.*}} addl 73828, %ecx
+// DISASM-NEXT: 11043: {{.*}} movl %gs:(%ecx), %eax
 
 .type tlslocal0,@object
 .section .tbss,"awT",@nobits
 .globl tlslocal0
 .align 4
 tlslocal0:
  .long 0
  .size tlslocal0, 4