SPIRV-Tools/source/validate.cpp

// Copyright (c) 2015-2016 The Khronos Group Inc.
//
// Permission is hereby granted, free of charge, to any person obtaining a
// copy of this software and/or associated documentation files (the
// "Materials"), to deal in the Materials without restriction, including
// without limitation the rights to use, copy, modify, merge, publish,
// distribute, sublicense, and/or sell copies of the Materials, and to
// permit persons to whom the Materials are furnished to do so, subject to
// the following conditions:
//
// The above copyright notice and this permission notice shall be included
// in all copies or substantial portions of the Materials.
//
// MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS
// KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS
// SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT
//    https://www.khronos.org/registry/
//
// THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
// IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
// CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
// MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.

#include "validate.h"
#include "validate_passes.h"

#include "binary.h"
#include "diagnostic.h"
#include "instruction.h"
#include "libspirv/libspirv.h"
#include "opcode.h"
#include "operand.h"
#include "spirv_constant.h"
#include "spirv_endian.h"

#include <algorithm>
#include <cassert>
#include <cstdio>
#include <functional>
#include <iterator>
#include <sstream>
#include <string>
#include <vector>

using std::function;
using std::ostream_iterator;
using std::placeholders::_1;
using std::string;
using std::stringstream;
using std::transform;
using std::vector;

using libspirv::CfgPass;
using libspirv::InstructionPass;
using libspirv::ModuleLayoutPass;
using libspirv::SsaPass;
using libspirv::ValidationState_t;

#if 0
spv_result_t spvValidateOperandsString(const uint32_t* words,
                                       const uint16_t wordCount,
                                       spv_position position,
                                       spv_diagnostic* pDiagnostic) {
  const char* str = (const char*)words;
  uint64_t strWordCount = strlen(str) / sizeof(uint32_t) + 1;
  if (strWordCount < wordCount) {
    DIAGNOSTIC << "Instruction word count is too short, string extends past "
                  "end of instruction.";
    return SPV_WARNING;
  }
  return SPV_SUCCESS;
}

spv_result_t spvValidateOperandsLiteral(const uint32_t* words,
                                        const uint32_t length,
                                        const uint16_t maxLength,
                                        spv_position position,
                                        spv_diagnostic* pDiagnostic) {
  // NOTE: A literal could either be a number consuming up to 2 words or a
  // null terminated string.
  (void)words;
  (void)length;
  (void)maxLength;
  (void)position;
  (void)pDiagnostic;
  return SPV_UNSUPPORTED;
}

spv_result_t spvValidateOperandValue(const spv_operand_type_t type,
                                     const uint32_t word,
                                     const spv_operand_table operandTable,
                                     spv_position position,
                                     spv_diagnostic* pDiagnostic) {
  switch (type) {
    case SPV_OPERAND_TYPE_ID:
    case SPV_OPERAND_TYPE_TYPE_ID:
    case SPV_OPERAND_TYPE_RESULT_ID:
    case SPV_OPERAND_TYPE_MEMORY_SEMANTICS_ID:
    case SPV_OPERAND_TYPE_SCOPE_ID: {
      // NOTE: ID's are validated in SPV_VALIDATION_LEVEL_1, this is
      // SPV_VALIDATION_LEVEL_0
    } break;
    case SPV_OPERAND_TYPE_LITERAL_INTEGER: {
      // NOTE: Implicitly valid as they are encoded as 32 bit value
    } break;
    case SPV_OPERAND_TYPE_SOURCE_LANGUAGE:
    case SPV_OPERAND_TYPE_EXECUTION_MODEL:
    case SPV_OPERAND_TYPE_ADDRESSING_MODEL:
    case SPV_OPERAND_TYPE_MEMORY_MODEL:
    case SPV_OPERAND_TYPE_EXECUTION_MODE:
    case SPV_OPERAND_TYPE_STORAGE_CLASS:
    case SPV_OPERAND_TYPE_DIMENSIONALITY:
    case SPV_OPERAND_TYPE_SAMPLER_ADDRESSING_MODE:
    case SPV_OPERAND_TYPE_SAMPLER_FILTER_MODE:
    case SPV_OPERAND_TYPE_FP_FAST_MATH_MODE:
    case SPV_OPERAND_TYPE_FP_ROUNDING_MODE:
    case SPV_OPERAND_TYPE_LINKAGE_TYPE:
    case SPV_OPERAND_TYPE_ACCESS_QUALIFIER:
    case SPV_OPERAND_TYPE_FUNCTION_PARAMETER_ATTRIBUTE:
    case SPV_OPERAND_TYPE_DECORATION:
    case SPV_OPERAND_TYPE_BUILT_IN:
    case SPV_OPERAND_TYPE_SELECTION_CONTROL:
    case SPV_OPERAND_TYPE_LOOP_CONTROL:
    case SPV_OPERAND_TYPE_FUNCTION_CONTROL:
    case SPV_OPERAND_TYPE_MEMORY_SEMANTICS_ID:
    case SPV_OPERAND_TYPE_OPTIONAL_MEMORY_ACCESS:
    case SPV_OPERAND_TYPE_SCOPE_ID:
    case SPV_OPERAND_TYPE_GROUP_OPERATION:
    case SPV_OPERAND_TYPE_KERNEL_ENQ_FLAGS:
    case SPV_OPERAND_TYPE_KERNEL_PROFILING_INFO: {
      spv_operand_desc operandEntry = nullptr;
      spv_result_t error =
          spvOperandTableValueLookup(operandTable, type, word, &operandEntry);
      if (error) {
        DIAGNOSTIC << "Invalid '" << spvOperandTypeStr(type) << "' operand '"
                   << word << "'.";
        return error;
      }
    } break;
    default:
      assert(0 && "Invalid operand types should already have been caught!");
  }
  return SPV_SUCCESS;
}

spv_result_t spvValidateBasic(const spv_instruction_t* pInsts,
                              const uint64_t instCount,
                              const spv_opcode_table opcodeTable,
                              const spv_operand_table operandTable,
                              spv_position position,
                              spv_diagnostic* pDiagnostic) {
  for (uint64_t instIndex = 0; instIndex < instCount; ++instIndex) {
    const uint32_t* words = pInsts[instIndex].words.data();
    uint16_t wordCount;
    SpvOp opcode;
    spvOpcodeSplit(words[0], &wordCount, &opcode);

    spv_opcode_desc opcodeEntry = nullptr;
    if (spvOpcodeTableValueLookup(opcodeTable, opcode, &opcodeEntry)) {
      DIAGNOSTIC << "Invalid Opcode '" << opcode << "'.";
      return SPV_ERROR_INVALID_BINARY;
    }
    position->index++;

    if (opcodeEntry->numTypes > wordCount) {
      DIAGNOSTIC << "Instruction word count '" << wordCount
                 << "' is not small, expected at least '"
                 << opcodeEntry->numTypes << "'.";
      return SPV_ERROR_INVALID_BINARY;
    }

    spv_operand_desc operandEntry = nullptr;
    for (uint16_t index = 1; index < pInsts[instIndex].words.size();
         ++index, position->index++) {
      const uint32_t word = words[index];

      // TODO(dneto): This strategy is inadequate for dealing with operations
      // with varying kinds or numbers of logical operands.  See the definition
      // of spvBinaryOperandInfo for more.
      // We should really parse the instruction and capture and use
      // the elaborated list of logical operands generated as a side effect
      // of the parse.
      spv_operand_type_t type = spvBinaryOperandInfo(
          word, index, opcodeEntry, operandTable, &operandEntry);

      if (SPV_OPERAND_TYPE_LITERAL_STRING == type) {
        spvCheckReturn(spvValidateOperandsString(
            words + index, wordCount - index, position, pDiagnostic));
        // NOTE: String literals are always at the end of Opcodes
        break;
      } else if (SPV_OPERAND_TYPE_LITERAL_INTEGER == type) {
        // spvCheckReturn(spvValidateOperandsNumber(
        //    words + index, wordCount - index, 2, position, pDiagnostic));
      } else {
        spvCheckReturn(spvValidateOperandValue(type, word, operandTable,
                                               position, pDiagnostic));
      }
    }
  }

  return SPV_SUCCESS;
}
#endif

spv_result_t spvValidateIDs(
    const spv_instruction_t* pInsts, const uint64_t count,
    const spv_opcode_table opcodeTable, const spv_operand_table operandTable,
    const spv_ext_inst_table extInstTable, const ValidationState_t& state,
    spv_position position, spv_diagnostic* pDiagnostic) {
  auto undefd = state.usedefs().FindUsesWithoutDefs();
  for (auto id : undefd) {
    DIAGNOSTIC << "Undefined ID: " << id;
  }
  position->index = SPV_INDEX_INSTRUCTION;
  spvCheckReturn(spvValidateInstructionIDs(pInsts, count, opcodeTable,
                                           operandTable, extInstTable, state,
                                           position, pDiagnostic));
  return undefd.empty() ? SPV_SUCCESS : SPV_ERROR_INVALID_ID;
}

namespace {

// TODO(umar): Validate header
// TODO(umar): The Id bound should be validated also. But you can only do that
// after you've seen all the instructions in the module.
// TODO(umar): The binary parser validates the magic word, and the length of the
// header, but nothing else.
spv_result_t setHeader(void* user_data, spv_endianness_t endian, uint32_t magic,
                       uint32_t version, uint32_t generator, uint32_t id_bound,
                       uint32_t reserved) {
  (void)user_data;
  (void)endian;
  (void)magic;
  (void)version;
  (void)generator;
  (void)id_bound;
  (void)reserved;
  return SPV_SUCCESS;
}

// Improves diagnostic messages by collecting names of IDs
// NOTE: This function returns void and is not involved in validation
void DebugInstructionPass(ValidationState_t& _,
                          const spv_parsed_instruction_t* inst) {
  switch (inst->opcode) {
    case SpvOpName: {
      const uint32_t target = *(inst->words + inst->operands[0].offset);
      const char* str =
          reinterpret_cast<const char*>(inst->words + inst->operands[1].offset);
      _.assignNameToId(target, str);
    } break;
    case SpvOpMemberName: {
      const uint32_t target = *(inst->words + inst->operands[0].offset);
      const char* str =
          reinterpret_cast<const char*>(inst->words + inst->operands[2].offset);
      _.assignNameToId(target, str);
    } break;
    case SpvOpSourceContinued:
    case SpvOpSource:
    case SpvOpSourceExtension:
    case SpvOpString:
    case SpvOpLine:
    case SpvOpNoLine:

    default:
      break;
  }
}

// Collects use-def info about an instruction's IDs.
void ProcessIds(ValidationState_t& _, const spv_parsed_instruction_t& inst) {
  if (inst.result_id) {
    _.usedefs().AddDef(
        {inst.result_id, inst.type_id, inst.opcode,
         std::vector<uint32_t>(inst.words, inst.words + inst.num_words)});
  }
  for (auto op = inst.operands; op != inst.operands + inst.num_operands; ++op) {
    if (spvIsIdType(op->type))
      _.usedefs().AddUse(inst.words[op->offset]);
  }
}

spv_result_t ProcessInstruction(void* user_data,
                                const spv_parsed_instruction_t* inst) {
  ValidationState_t& _ = *(reinterpret_cast<ValidationState_t*>(user_data));
  _.incrementInstructionCount();
  if (inst->opcode == SpvOpEntryPoint) _.entry_points().push_back(inst->words[2]);

  DebugInstructionPass(_, inst);
  // TODO(umar): Perform data rules pass
  ProcessIds(_, *inst);
  spvCheckReturn(ModuleLayoutPass(_, inst));
  spvCheckReturn(CfgPass(_, inst));
  spvCheckReturn(SsaPass(_, inst));
  spvCheckReturn(InstructionPass(_, inst));

  return SPV_SUCCESS;
}

}  // anonymous namespace

spv_result_t spvValidate(const spv_const_context context,
                         const spv_const_binary binary, const uint32_t options,
                         spv_diagnostic* pDiagnostic) {
  if (!pDiagnostic) return SPV_ERROR_INVALID_DIAGNOSTIC;

  spv_endianness_t endian;
  spv_position_t position = {};
  if (spvBinaryEndianness(binary, &endian)) {
    DIAGNOSTIC << "Invalid SPIR-V magic number.";
    return SPV_ERROR_INVALID_BINARY;
  }

  spv_header_t header;
  if (spvBinaryHeaderGet(binary, endian, &header)) {
    DIAGNOSTIC << "Invalid SPIR-V header.";
    return SPV_ERROR_INVALID_BINARY;
  }

  // NOTE: Parse the module and perform inline validation checks. These
  // checks do not require the the knowledge of the whole module.
  ValidationState_t vstate(pDiagnostic, options, context);
  spvCheckReturn(spvBinaryParse(context, &vstate, binary->code,
                                binary->wordCount, setHeader,
                                ProcessInstruction, pDiagnostic));

  // TODO(umar): Add validation checks which require the parsing of the entire
  // module. Use the information from the ProcessInstruction pass to make the
  // checks.

  if (vstate.unresolvedForwardIdCount() > 0) {
    stringstream ss;
    vector<uint32_t> ids = vstate.unresolvedForwardIds();

    transform(begin(ids), end(ids), ostream_iterator<string>(ss, " "),
              bind(&ValidationState_t::getIdName, vstate, _1));

    auto id_str = ss.str();
    return vstate.diag(SPV_ERROR_INVALID_ID)
           << "The following forward referenced IDs have not be defined:\n"
           << id_str.substr(0, id_str.size() - 1);
  }

  // NOTE: Copy each instruction for easier processing
  std::vector<spv_instruction_t> instructions;
  uint64_t index = SPV_INDEX_INSTRUCTION;
  while (index < binary->wordCount) {
    uint16_t wordCount;
    SpvOp opcode;
    spvOpcodeSplit(spvFixWord(binary->code[index], endian), &wordCount,
                   &opcode);
    spv_instruction_t inst;
    spvInstructionCopy(&binary->code[index], opcode, wordCount, endian, &inst);
    instructions.push_back(inst);
    index += wordCount;
  }

  if (spvIsInBitfield(SPV_VALIDATE_ID_BIT, options)) {
    position.index = SPV_INDEX_INSTRUCTION;
    spvCheckReturn(spvValidateIDs(instructions.data(), instructions.size(),
                                  context->opcode_table, context->operand_table,
                                  context->ext_inst_table, vstate, &position,
                                  pDiagnostic));
  }

  return SPV_SUCCESS;
}