[LLDB] [LoongArch] Add minimal LoongArch support

[seehearfeel]

Summary

Add as little code as possible to allow compiling lldb on LoongArch.
Actual functionality will be implemented later.

[SixWeining]

lldb/source/Plugins/Process/Linux/CMakeLists.txt

11

Had better use git diff -U999999 to generate the patch so that we can get more context.
See: https://llvm.org/docs/Phabricator.html#requesting-a-review-via-the-web-interface

[seehearfeel]

Use git diff -U999999 to generate the patch so that we can get more context suggested by Weining.

[xen0n]

Hi, I've edited the summary and patch title for you. It's generally not necessary to add that much "politeness" when most of it is obvious from context (e.g. the fact that you're new face here, that there's obviously no LoongArch support in LLDB, and most of the methods being stubs). It didn't help that much of the text was in Chinglish either.

As for the changes, they look reasonable to me, but as I haven't tested it out myself yet (and unable to, due to it being stub-only), I'll not give the approval myself this time. Thanks for your contribution and welcome!

[DavidSpickett]

Always good to see another architecture in lldb.

A few points up front.

Changes like this are fine and we can review them, but can only be landed once we see that they'll be built on. So please stack changes so that we can see what is upcoming. https://llvm.org/docs/Phabricator.html#creating-a-patch-series

Do you have a plan to test this configuration? I don't see any public buildbots for Loongson (please correct me if I am wrong) but that is not a requirement as long as you have some testing plan (RISCV doesn't have a public bot, as one example).

Do you have community members who can review these changes for architectural correctness? Speaking just for myself I am happy to review bits of lldb machinery but when it comes to architecture details I am not, and will not, be an expert in Loongson.

I'm ok accepting patches without a second reviewer, but I advise for your own benefit to try to find someone who already knows the details of Loongson.

[tschuett]

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.cpp

44

Stupid question: why is the register count 0?

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.h

27

Firstly, I prefer ènum class. Second, it is unused?

[seehearfeel]

In D136578#3878476, @xen0n wrote:

Hi, I've edited the summary and patch title for you. It's generally not necessary to add that much "politeness" when most of it is obvious from context (e.g. the fact that you're new face here, that there's obviously no LoongArch support in LLDB, and most of the methods being stubs). It didn't help that much of the text was in Chinglish either.

As for the changes, they look reasonable to me, but as I haven't tested it out myself yet (and unable to, due to it being stub-only), I'll not give the approval myself this time. Thanks for your contribution and welcome!

OK, thank you.

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.cpp

44

It is just a stub function here to build successfully, actually it should return k_num_gpr_registers
after the following header file is added in the later patch.

lldb/source/Plugins/Process/Utility/lldb-loongarch-register-enums.h

[SixWeining]

In D136578#3878744, @DavidSpickett wrote:

Always good to see another architecture in lldb.

A few points up front.

Changes like this are fine and we can review them, but can only be landed once we see that they'll be built on. So please stack changes so that we can see what is upcoming. https://llvm.org/docs/Phabricator.html#creating-a-patch-series

Do you have a plan to test this configuration? I don't see any public buildbots for Loongson (please correct me if I am wrong) but that is not a requirement as long as you have some testing plan (RISCV doesn't have a public bot, as one example).

Do you have community members who can review these changes for architectural correctness? Speaking just for myself I am happy to review bits of lldb machinery but when it comes to architecture details I am not, and will not, be an expert in Loongson.

I'm ok accepting patches without a second reviewer, but I advise for your own benefit to try to find someone who already knows the details of Loongson.

Hi @DavidSpickett, thanks for your advice. I (as the LoongArch backend code owner) can review these changes for architectural correctness. And I think other LoongArch community members (non Loongson employees) also can take it. Especially @xen0n and @xry111 who are quite familar with LoongArch.

What's more, AFAIK, @seehearfeel ([e-mail]) is the LoongArch port maintainer of gdb.

Regarding the public buildbot, I'm trying to set it up and maybe we can see it in one or two weeks.

lldb/source/Plugins/Process/Linux/NativeRegisterContextLinux_loongarch64.h

10

Seems #if __loongarch_grlen == 64 is enough? But I'm fine with both.

25

Pls limit the columns count to 80 which can be done by clang-format. See: https://llvm.org/docs/Contributing.html#how-to-submit-a-patch

[seehearfeel]

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.h

27

Now, GPRegSet and FPRegSet are not used,
I will remove them and update the diff in this commit,
and use
enum class { GPRegSet = 0, FPRegSet };
in the later patch, thank you.

[DavidSpickett]

What's more, AFAIK, @seehearfeel ([e-mail]) is the LoongArch port maintainer of gdb.
Regarding the public buildbot, I'm trying to set it up and maybe we can see it in one or two weeks.

Great!

[seehearfeel]

lldb/source/Plugins/Process/Linux/NativeRegisterContextLinux_loongarch64.h

10

Let us leave it as it is like riscv does, thank you.

25

OK, will do it and then update.

[SixWeining]

Build is OK on my local LoongArch machine with 299 failed tests and 2 timed out tests. Just as the summary says, I think this is expected.

[seehearfeel]

Remove GPRegSet and FPRegSet definitions which are not used in this commit.
Use git clang-format to format the patch properly.

[DavidSpickett]

This LGTM from the lldb side.

Feel free to upload a series of patches in future if you want. You don't have to do one at a time, just update the whole stack as needed. Plus it gives the reviewers some context of what the changes are working towards.

Do you have commit access? If not it's fine to wait until you have (https://llvm.org/docs/DeveloperPolicy.html#obtaining-commit-access), or someone else can land on your behalf. Just provide a name and email address you'd like to be on the commit.

[SixWeining]

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.h

27

Why is this structure and below FPR defined like this? Do you intent to keep the layout same as some structures of Linux? If yes, is it necessary?

[xry111]

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.h

27

It seems necessary because lldb invokes ptrace system call which directly stores into the pointer returned by GetGPRBuffer.

[seehearfeel]

In D136578#3881955, @DavidSpickett wrote:

This LGTM from the lldb side.

OK, thank you.

Feel free to upload a series of patches in future if you want. You don't have to do one at a time, just update the whole stack as needed. Plus it gives the reviewers some context of what the changes are working towards.

I am a newcomer, here are some of my thoughts:
(1) Add the minimal changes to fix the build errors.
(2) Submit other more patches step by step to make
the basic command "run", "breakpoint", "next" ...
can be used to debug, single patch or patch series.
(3) Add more code to make more commands work well.

Do you have commit access? If not it's fine to wait until you have (https://llvm.org/docs/DeveloperPolicy.html#obtaining-commit-access), or someone else can land on your behalf. Just provide a name and email address you'd like to be on the commit.

I have no commit access now, @SixWeining [e-mail]
can land on my behalf, here are my name and email:
Tiezhu Yang
[e-mail]

[SixWeining]

LGTM for LoongArch related changes.

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.h

27

Make sense. Thanks.

[DavidSpickett]

I am a newcomer, here are some of my thoughts:
(1) Add the minimal changes to fix the build errors.
(2) Submit other more patches step by step to make
the basic command "run", "breakpoint", "next" ...
can be used to debug, single patch or patch series.
(3) Add more code to make more commands work well.

Sounds like a good approach to me.

I believe the riscv effort focused on parts of the test suite in turn, sounds like you'll do the same.

can land on my behalf

Will do.

[DavidSpickett]

Landed as https://github.com/llvm/llvm-project/commit/a3be778ed09b7badcda20c5c8738ba19531dad48.

[SixWeining]

Regarding the public buildbot, I'm trying to set it up and maybe we can see it in one or two weeks.

FYI. D138672 is the review for adding LoongArch buildbot.

[DavidSpickett]

Nice, glad to see that.

lldb/source/Plugins/Process/Linux/CMakeLists.txt

 add_lldb_library(lldbPluginProcessLinux
   IntelPTCollector.cpp
   IntelPTSingleBufferTrace.cpp
   IntelPTMultiCoreTrace.cpp
   IntelPTPerThreadProcessTrace.cpp
   IntelPTThreadTraceCollection.cpp
   NativeProcessLinux.cpp
   NativeRegisterContextLinux.cpp
   NativeRegisterContextLinux_arm.cpp
   NativeRegisterContextLinux_arm64.cpp
+  NativeRegisterContextLinux_loongarch64.cpp
   NativeRegisterContextLinux_ppc64le.cpp
   NativeRegisterContextLinux_riscv64.cpp
   NativeRegisterContextLinux_s390x.cpp
   NativeRegisterContextLinux_x86_64.cpp
   NativeThreadLinux.cpp
   Perf.cpp
   Procfs.cpp
   SingleStepCheck.cpp

lldb/source/Plugins/Process/Linux/NativeRegisterContextLinux_loongarch64.h

+//===-- NativeRegisterContextLinux_loongarch64.h ----------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__loongarch__) && __loongarch_grlen == 64
+
+#ifndef lldb_NativeRegisterContextLinux_loongarch64_h
+#define lldb_NativeRegisterContextLinux_loongarch64_h
+
+#include "Plugins/Process/Linux/NativeRegisterContextLinux.h"
+#include "Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.h"
+
+#include <asm/ptrace.h>
+
+namespace lldb_private {
+namespace process_linux {
+
+class NativeProcessLinux;
+
+class NativeRegisterContextLinux_loongarch64
+    : public NativeRegisterContextLinux {
+public:
+  NativeRegisterContextLinux_loongarch64(
+      const ArchSpec &target_arch, NativeThreadProtocol &native_thread,
+      std::unique_ptr<RegisterInfoPOSIX_loongarch64> register_info_up);
+
+  uint32_t GetRegisterSetCount() const override;
+
+  const RegisterSet *GetRegisterSet(uint32_t set_index) const override;
+
+  Status ReadRegister(const RegisterInfo *reg_info,
+                      RegisterValue &reg_value) override;
+
+  Status WriteRegister(const RegisterInfo *reg_info,
+                       const RegisterValue &reg_value) override;
+
+  Status ReadAllRegisterValues(lldb::WritableDataBufferSP &data_sp) override;
+
+  Status WriteAllRegisterValues(const lldb::DataBufferSP &data_sp) override;
+
+protected:
+  void *GetGPRBuffer() override { return &m_gpr; }
+
+  void *GetFPRBuffer() override { return &m_fpr; }
+
+  size_t GetGPRSize() const override { return GetRegisterInfo().GetGPRSize(); }
+
+  size_t GetFPRSize() override { return GetRegisterInfo().GetFPRSize(); }
+
+private:
+  RegisterInfoPOSIX_loongarch64::GPR m_gpr;
+
+  RegisterInfoPOSIX_loongarch64::FPR m_fpr;
+
+  const RegisterInfoPOSIX_loongarch64 &GetRegisterInfo() const;
+};
+
+} // namespace process_linux
+} // namespace lldb_private
+
+#endif // #ifndef lldb_NativeRegisterContextLinux_loongarch64_h
+
+#endif // defined(__loongarch__) && __loongarch_grlen == 64

lldb/source/Plugins/Process/Linux/NativeRegisterContextLinux_loongarch64.cpp

+//===-- NativeRegisterContextLinux_loongarch64.cpp ------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__loongarch__) && __loongarch_grlen == 64
+
+#include "NativeRegisterContextLinux_loongarch64.h"
+
+#include "lldb/Host/HostInfo.h"
+#include "lldb/Utility/DataBufferHeap.h"
+#include "lldb/Utility/Log.h"
+#include "lldb/Utility/RegisterValue.h"
+#include "lldb/Utility/Status.h"
+
+#include "Plugins/Process/Linux/NativeProcessLinux.h"
+#include "Plugins/Process/Linux/Procfs.h"
+
+using namespace lldb;
+using namespace lldb_private;
+using namespace lldb_private::process_linux;
+
+std::unique_ptr<NativeRegisterContextLinux>
+NativeRegisterContextLinux::CreateHostNativeRegisterContextLinux(
+    const ArchSpec &target_arch, NativeThreadLinux &native_thread) {
+  switch (target_arch.GetMachine()) {
+  case llvm::Triple::loongarch64: {
+    Flags opt_regsets;
+    auto register_info_up = std::make_unique<RegisterInfoPOSIX_loongarch64>(
+        target_arch, opt_regsets);
+    return std::make_unique<NativeRegisterContextLinux_loongarch64>(
+        target_arch, native_thread, std::move(register_info_up));
+  }
+  default:
+    llvm_unreachable("have no register context for architecture");
+  }
+}
+
+llvm::Expected<ArchSpec>
+NativeRegisterContextLinux::DetermineArchitecture(lldb::tid_t tid) {
+  return HostInfo::GetArchitecture();
+}
+
+NativeRegisterContextLinux_loongarch64::NativeRegisterContextLinux_loongarch64(
+    const ArchSpec &target_arch, NativeThreadProtocol &native_thread,
+    std::unique_ptr<RegisterInfoPOSIX_loongarch64> register_info_up)
+    : NativeRegisterContextRegisterInfo(native_thread,
+                                        register_info_up.release()),
+      NativeRegisterContextLinux(native_thread) {
+  ::memset(&m_fpr, 0, sizeof(m_fpr));
+  ::memset(&m_gpr, 0, sizeof(m_gpr));
+}
+
+const RegisterInfoPOSIX_loongarch64 &
+NativeRegisterContextLinux_loongarch64::GetRegisterInfo() const {
+  return static_cast<const RegisterInfoPOSIX_loongarch64 &>(
+      NativeRegisterContextRegisterInfo::GetRegisterInfoInterface());
+}
+
+uint32_t NativeRegisterContextLinux_loongarch64::GetRegisterSetCount() const {
+  return GetRegisterInfo().GetRegisterSetCount();
+}
+
+const RegisterSet *NativeRegisterContextLinux_loongarch64::GetRegisterSet(
+    uint32_t set_index) const {
+  return GetRegisterInfo().GetRegisterSet(set_index);
+}
+
+Status NativeRegisterContextLinux_loongarch64::ReadRegister(
+    const RegisterInfo *reg_info, RegisterValue &reg_value) {
+  return Status("Failed to read register value");
+}
+
+Status NativeRegisterContextLinux_loongarch64::WriteRegister(
+    const RegisterInfo *reg_info, const RegisterValue &reg_value) {
+  return Status("Failed to write register value");
+}
+
+Status NativeRegisterContextLinux_loongarch64::ReadAllRegisterValues(
+    lldb::WritableDataBufferSP &data_sp) {
+  return Status("Failed to read all register values");
+}
+
+Status NativeRegisterContextLinux_loongarch64::WriteAllRegisterValues(
+    const lldb::DataBufferSP &data_sp) {
+  return Status("Failed to write all register values");
+}
+
+#endif // defined(__loongarch__) && __loongarch_grlen == 64

lldb/source/Plugins/Process/Utility/CMakeLists.txt

add_lldb_library(lldbPluginProcessUtility
   RegisterContextPOSIX_riscv64.cpp
   RegisterContextPOSIX_s390x.cpp
   RegisterContextPOSIX_x86.cpp
   RegisterContextThreadMemory.cpp
   RegisterContextWindows_i386.cpp
   RegisterContextWindows_x86_64.cpp
   RegisterInfoPOSIX_arm.cpp
   RegisterInfoPOSIX_arm64.cpp
+  RegisterInfoPOSIX_loongarch64.cpp
   RegisterInfoPOSIX_ppc64le.cpp
   RegisterInfoPOSIX_riscv64.cpp
   StopInfoMachException.cpp
   ThreadMemory.cpp
 
   LINK_LIBS
     lldbBreakpoint
     lldbCore

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.h

+//===-- RegisterInfoPOSIX_loongarch64.h -------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLDB_SOURCE_PLUGINS_PROCESS_UTILITY_REGISTERINFOPOSIX_LOONGARCH64_H
+#define LLDB_SOURCE_PLUGINS_PROCESS_UTILITY_REGISTERINFOPOSIX_LOONGARCH64_H
+
+#include "RegisterInfoAndSetInterface.h"
+#include "lldb/Target/RegisterContext.h"
+#include "lldb/lldb-private.h"
+#include <map>
+
+class RegisterInfoPOSIX_loongarch64
+    : public lldb_private::RegisterInfoAndSetInterface {
+public:
+  static const lldb_private::RegisterInfo *
+  GetRegisterInfoPtr(const lldb_private::ArchSpec &target_arch);
+  static uint32_t
+  GetRegisterInfoCount(const lldb_private::ArchSpec &target_arch);
+
+public:
+  struct GPR {
+    uint64_t gpr[32];
+
+    uint64_t orig_a0;
+    uint64_t csr_era;
+    uint64_t csr_badv;
+    uint64_t reserved[10];
+  };
+
+  struct FPR {
+    uint64_t fpr[32];
+    uint64_t fcc;
+    uint32_t fcsr;
+  };
+
+  RegisterInfoPOSIX_loongarch64(const lldb_private::ArchSpec &target_arch,
+                                lldb_private::Flags flags);
+
+  size_t GetGPRSize() const override;
+
+  size_t GetFPRSize() const override;
+
+  const lldb_private::RegisterInfo *GetRegisterInfo() const override;
+
+  uint32_t GetRegisterCount() const override;
+
+  const lldb_private::RegisterSet *
+  GetRegisterSet(size_t reg_set) const override;
+
+  size_t GetRegisterSetCount() const override;
+
+  size_t GetRegisterSetFromRegisterIndex(uint32_t reg_index) const override;
+
+private:
+  const lldb_private::RegisterInfo *m_register_info_p;
+  uint32_t m_register_info_count;
+};
+
+#endif

lldb/source/Plugins/Process/Utility/RegisterInfoPOSIX_loongarch64.cpp

+//===-- RegisterInfoPOSIX_loongarch64.cpp --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===---------------------------------------------------------------------===//
+
+#include <cassert>
+#include <lldb/Utility/Flags.h>
+#include <stddef.h>
+
+#include "lldb/lldb-defines.h"
+#include "llvm/Support/Compiler.h"
+
+#include "RegisterInfoPOSIX_loongarch64.h"
+
+const lldb_private::RegisterInfo *
+RegisterInfoPOSIX_loongarch64::GetRegisterInfoPtr(
+    const lldb_private::ArchSpec &target_arch) {
+  switch (target_arch.GetMachine()) {
+  default:
+    assert(false && "Unhandled target architecture.");
+    return nullptr;
+  }
+}
+
+uint32_t RegisterInfoPOSIX_loongarch64::GetRegisterInfoCount(
+    const lldb_private::ArchSpec &target_arch) {
+  switch (target_arch.GetMachine()) {
+  default:
+    assert(false && "Unhandled target architecture.");
+    return 0;
+  }
+}
+
+RegisterInfoPOSIX_loongarch64::RegisterInfoPOSIX_loongarch64(
+    const lldb_private::ArchSpec &target_arch, lldb_private::Flags flags)
+    : lldb_private::RegisterInfoAndSetInterface(target_arch),
+      m_register_info_p(GetRegisterInfoPtr(target_arch)),
+      m_register_info_count(GetRegisterInfoCount(target_arch)) {}
+
+uint32_t RegisterInfoPOSIX_loongarch64::GetRegisterCount() const { return 0; }
+
+size_t RegisterInfoPOSIX_loongarch64::GetGPRSize() const {
+  return sizeof(struct RegisterInfoPOSIX_loongarch64::GPR);
+}
+
+size_t RegisterInfoPOSIX_loongarch64::GetFPRSize() const {
+  return sizeof(struct RegisterInfoPOSIX_loongarch64::FPR);
+}
+
+const lldb_private::RegisterInfo *
+RegisterInfoPOSIX_loongarch64::GetRegisterInfo() const {
+  return m_register_info_p;
+}
+
+size_t RegisterInfoPOSIX_loongarch64::GetRegisterSetCount() const { return 0; }
+
+size_t RegisterInfoPOSIX_loongarch64::GetRegisterSetFromRegisterIndex(
+    uint32_t reg_index) const {
+  return LLDB_INVALID_REGNUM;
+}
+
+const lldb_private::RegisterSet *
+RegisterInfoPOSIX_loongarch64::GetRegisterSet(size_t set_index) const {
+  return nullptr;
+}