1WebAssembly lld port 2==================== 3 4The WebAssembly version of lld takes WebAssembly binaries as inputs and produces 5a WebAssembly binary as its output. For the most part it tries to mimic the 6behaviour of traditional ELF linkers and specifically the ELF lld port. Where 7possible the command line flags and the semantics should be the same. 8 9 10Object file format 11------------------ 12 13The WebAssembly object file format used by LLVM and LLD is specified as part of 14the WebAssembly tool conventions on linking_. 15 16This is the object format that the llvm will produce when run with the 17``wasm32-unknown-unknown`` target. 18 19Usage 20----- 21 22The WebAssembly version of lld is installed as **wasm-ld**. It shared many 23common linker flags with **ld.lld** but also includes several 24WebAssembly-specific options: 25 26.. option:: --no-entry 27 28 Don't search for the entry point symbol (by default ``_start``). 29 30.. option:: --export-table 31 32 Export the function table to the environment. 33 34.. option:: --import-table 35 36 Import the function table from the environment. 37 38.. option:: --export-all 39 40 Export all symbols (normally combined with --no-gc-sections) 41 42 Note that this will not export linker-generated mutable globals unless 43 the resulting binaryen already includes the 'mutable-globals' features 44 since that would otherwise create and invalid binaryen. 45 46.. option:: --export-dynamic 47 48 When building an executable, export any non-hidden symbols. By default only 49 the entry point and any symbols marked as exports (either via the command line 50 or via the `export-name` source attribute) are exported. 51 52.. option:: --global-base=<value> 53 54 Address at which to place global data. 55 56.. option:: --no-merge-data-segments 57 58 Disable merging of data segments. 59 60.. option:: --stack-first 61 62 Place stack at start of linear memory rather than after data. 63 64.. option:: --compress-relocations 65 66 Relocation targets in the code section are 5-bytes wide in order to 67 potentially accommodate the largest LEB128 value. This option will cause the 68 linker to shrink the code section to remove any padding from the final 69 output. However because it affects code offset, this option is not 70 compatible with outputting debug information. 71 72.. option:: --allow-undefined 73 74 Allow undefined symbols in linked binary. This is the legacy 75 flag which corresponds to ``--unresolve-symbols=ignore`` + 76 ``--import-undefined``. 77 78.. option:: --allow-undefined-file=<filename> 79 80 Like ``--allow-undefined``, but the filename specified a flat list of 81 symbols, one per line, which are allowed to be undefined. 82 83.. option:: --unresolved-symbols=<method> 84 85 This is a more full featured version of ``--allow-undefined``. 86 The semanatics of the different methods are as follows: 87 88 report-all: 89 90 Report all unresolved symbols. This is the default. Normally the linker 91 will generate an error message for each reported unresolved symbol but the 92 option ``--warn-unresolved-symbols`` can change this to a warning. 93 94 ignore-all: 95 96 Resolve all undefined symbols to zero. For data and function addresses 97 this is trivial. For direct function calls, the linker will generate a 98 trapping stub function in place of the undefined function. 99 100 import-dynamic: 101 102 Undefined symbols generate WebAssembly imports, including undefined data 103 symbols. This is somewhat similar to the --import-undefined option but 104 works all symbol types. This options puts limitations on the type of 105 relocations that are allowed for imported data symbols. Relocations that 106 require absolute data addresses (i.e. All R_WASM_MEMORY_ADDR_I32) will 107 generate an error if they cannot be resolved statically. For clang/llvm 108 this means inputs should be compiled with `-fPIC` (i.e. `pic` or 109 `dynamic-no-pic` relocation models). This options is useful for linking 110 binaries that are themselves static (non-relocatable) but whose undefined 111 symbols are resolved by a dynamic linker. Since the dynamic linking API is 112 experimental, this option currently requires `--experimental-pic` to also 113 be specified. 114 115.. option:: --import-memory 116 117 Import memory from the environment. 118 119.. option:: --import-undefined 120 121 Generate WebAssembly imports for undefined symbols, where possible. For 122 example, for function symbols this is always possible, but in general this 123 is not possible for undefined data symbols. Undefined data symbols will 124 still be reported as normal (in accordance with ``--unresolved-symbols``). 125 126.. option:: --initial-heap=<value> 127 128 Initial size of the heap. Default: zero. 129 130.. option:: --initial-memory=<value> 131 132 Initial size of the linear memory. Default: the sum of stack, static data and heap sizes. 133 134.. option:: --max-memory=<value> 135 136 Maximum size of the linear memory. Default: unlimited. 137 138By default the function table is neither imported nor exported, but defined 139for internal use only. 140 141Behaviour 142--------- 143 144In general, where possible, the WebAssembly linker attempts to emulate the 145behaviour of a traditional ELF linker, and in particular the ELF port of lld. 146For more specific details on how this is achieved see the tool conventions on 147linking_. 148 149Function Signatures 150~~~~~~~~~~~~~~~~~~~ 151 152One way in which the WebAssembly linker differs from traditional native linkers 153is that function signature checking is strict in WebAssembly. It is a 154validation error for a module to contain a call site that doesn't agree with 155the target signature. Even though this is undefined behaviour in C/C++, it is not 156uncommon to find this in real-world C/C++ programs. For example, a call site in 157one compilation unit which calls a function defined in another compilation 158unit but with too many arguments. 159 160In order not to generate such invalid modules, lld has two modes of handling such 161mismatches: it can simply error-out or it can create stub functions that will 162trap at runtime (functions that contain only an ``unreachable`` instruction) 163and use these stub functions at the otherwise invalid call sites. 164 165The default behaviour is to generate these stub function and to produce 166a warning. The ``--fatal-warnings`` flag can be used to disable this behaviour 167and error out if mismatched are found. 168 169Exports 170~~~~~~~ 171 172When building a shared library any symbols marked as ``visibility=default`` will 173be exported. 174 175When building an executable, only the entry point (``_start``) and symbols with 176the ``WASM_SYMBOL_EXPORTED`` flag are exported by default. In LLVM the 177``WASM_SYMBOL_EXPORTED`` flag is set by the ``wasm-export-name`` attribute which 178in turn can be set using ``__attribute__((export_name))`` clang attribute. 179 180In addition, symbols can be exported via the linker command line using 181``--export`` (which will error if the symbol is not found) or 182``--export-if-defined`` (which will not). 183 184Finally, just like with native ELF linker the ``--export-dynamic`` flag can be 185used to export symbols in the executable which are marked as 186``visibility=default``. 187 188Imports 189~~~~~~~ 190 191By default no undefined symbols are allowed in the final binary. The flag 192``--allow-undefined`` results in a WebAssembly import being defined for each 193undefined symbol. It is then up to the runtime to provide such symbols. 194``--allow-undefined-file`` is the same but allows a list of symbols to be 195specified. 196 197Alternatively symbols can be marked in the source code as with the 198``import_name`` and/or ``import_module`` clang attributes which signals that 199they are expected to be undefined at static link time. 200 201Stub Libraries 202~~~~~~~~~~~~~~ 203 204Another way to specify imports and exports is via a "stub library". This 205feature is inspired by the ELF stub objects which are supported by the Solaris 206linker. Stub libraries are text files that can be passed as normal linker 207inputs, similar to how linker scripts can be passed to the ELF linker. The stub 208library is a stand-in for a set of symbols that will be available at runtime, 209but doesn't contain any actual code or data. Instead it contains just a list of 210symbols, one per line. Each symbol can specify zero or more dependencies. 211These dependencies are symbols that must be defined, and exported, by the output 212module if the symbol is question is imported/required by the output module. 213 214For example, imagine the runtime provides an external symbol ``foo`` that 215depends on the ``malloc`` and ``free``. This can be expressed simply as:: 216 217 #STUB 218 foo: malloc,free 219 220Here we are saying that ``foo`` is allowed to be imported (undefined) but that 221if it is imported, then the output module must also export ``malloc`` and 222``free`` to the runtime. If ``foo`` is imported (undefined), but the output 223module does not define ``malloc`` and ``free`` then the link will fail. 224 225Stub libraries must begin with ``#STUB`` on a line by itself. 226 227Garbage Collection 228~~~~~~~~~~~~~~~~~~ 229 230Since WebAssembly is designed with size in mind the linker defaults to 231``--gc-sections`` which means that all unused functions and data segments will 232be stripped from the binary. 233 234The symbols which are preserved by default are: 235 236- The entry point (by default ``_start``). 237- Any symbol which is to be exported. 238- Any symbol transitively referenced by the above. 239 240Weak Undefined Functions 241~~~~~~~~~~~~~~~~~~~~~~~~ 242 243On native platforms, calls to weak undefined functions end up as calls to the 244null function pointer. With WebAssembly, direct calls must reference a defined 245function (with the correct signature). In order to handle this case the linker 246will generate function a stub containing only the ``unreachable`` instruction 247and use this for any direct references to an undefined weak function. 248 249For example a runtime call to a weak undefined function ``foo`` will up trapping 250on ``unreachable`` inside and linker-generated function called 251``undefined:foo``. 252 253Missing features 254---------------- 255 256- Merging of data section similar to ``SHF_MERGE`` in the ELF world is not 257 supported. 258- No support for creating shared libraries. The spec for shared libraries in 259 WebAssembly is still in flux: 260 https://github.com/WebAssembly/tool-conventions/blob/main/DynamicLinking.md 261 262.. _linking: https://github.com/WebAssembly/tool-conventions/blob/main/Linking.md 263