[lldb-server] Add type info for qMemoryinfo reponse by felipepiovezan · Pull Request #209235 · llvm/llvm-project

felipepiovezan · 2026-07-13T16:27:21Z

Use the name of the memory region to report the known cases where an address points to stack/heap: "[stack]" or "[heap"].

According to 1, the "[stack]" name is the main thread's stack region:

  [stack]
                     The initial process's (also known as the main
                     thread's) stack.

For other threads, we can't know their stack region because this field got removed in newer kernels:

 [stack:tid] (from Linux 3.4 to Linux 4.4)
        A thread's stack (where the tid is a thread ID).  It
        corresponds to the /proc/pid/task/tid/ path.  This
        field was removed in Linux 4.5, since providing this
        information for a process with large numbers of
        threads is expensive.

So we can only set "isStack" to Yes in some cases, and to "No" when the name is "[heap]". If there is no name, or the name is something else, we keep the "don't know" answer.

Use the name of the memory region to report the known cases of where an address points to stack/heap: "[stack]" or "[heap"]. According to [1], the "[stack]" name is only available for the main thread. So we can only set "isStack" to Yes in some cases, and to "No" when the name is "[heap]". If there is no name, or the name is something else, we keep the "don't know" answer. > [stack:tid] (from Linux 3.4 to Linux 4.4) > A thread's stack (where the tid is a thread ID). It > corresponds to the /proc/pid/task/tid/ path. This > field was removed in Linux 4.5, since providing this > information for a process with large numbers of > threads is expensive. https://man7.org/linux/man-pages/man5/proc_pid_maps.5.html

llvmorg-github-actions · 2026-07-13T16:28:03Z

@llvm/pr-subscribers-lldb

Author: Felipe de Azevedo Piovezan (felipepiovezan)

Changes

Use the name of the memory region to report the known cases of where an address points to stack/heap: "[stack]" or "[heap"].

According to [1], the "[stack]" name is only available for the main thread. So we can only set "isStack" to Yes in some cases, and to "No" when the name is "[heap]". If there is no name, or the name is something else, we keep the "don't know" answer.

> [stack:tid] (from Linux 3.4 to Linux 4.4)
> A thread's stack (where the tid is a thread ID). It
> corresponds to the /proc/pid/task/tid/ path. This
> field was removed in Linux 4.5, since providing this
> information for a process with large numbers of
> threads is expensive.

https://man7.org/linux/man-pages/man5/proc_pid_maps.5.html

Full diff: https://github.com/llvm/llvm-project/pull/209235.diff

4 Files Affected:

(modified) lldb/include/lldb/Target/MemoryRegionInfo.h (+4-1)
(modified) lldb/source/Plugins/Process/Utility/LinuxProcMaps.cpp (+8-1)
(modified) lldb/source/Plugins/Process/gdb-remote/GDBRemoteCommunicationServerLLGS.cpp (+7)
(modified) lldb/unittests/Process/Utility/LinuxProcMapsTest.cpp (+12-1)

diff --git a/lldb/include/lldb/Target/MemoryRegionInfo.h b/lldb/include/lldb/Target/MemoryRegionInfo.h
index 16faf6ed9d64b..e8a16a3ab9757 100644
--- a/lldb/include/lldb/Target/MemoryRegionInfo.h
+++ b/lldb/include/lldb/Target/MemoryRegionInfo.h
@@ -142,7 +142,10 @@ class MemoryRegionInfo {
 
   LazyBool IsStackMemory() const { return m_is_stack_memory; }
 
-  void SetIsStackMemory(LazyBool val) { m_is_stack_memory = val; }
+  MemoryRegionInfo &SetIsStackMemory(LazyBool val) {
+    m_is_stack_memory = val;
+    return *this;
+  }
 
   void SetPageSize(int pagesize) { m_pagesize = pagesize; }
 
diff --git a/lldb/source/Plugins/Process/Utility/LinuxProcMaps.cpp b/lldb/source/Plugins/Process/Utility/LinuxProcMaps.cpp
index d952b7dbdb816..bdb340332851a 100644
--- a/lldb/source/Plugins/Process/Utility/LinuxProcMaps.cpp
+++ b/lldb/source/Plugins/Process/Utility/LinuxProcMaps.cpp
@@ -114,8 +114,15 @@ ParseMemoryRegionInfoFromProcMapsLine(llvm::StringRef maps_line,
 
   line_extractor.SkipSpaces();
   const char *name = line_extractor.Peek();
-  if (name)
+  if (name) {
     region.SetName(name);
+    // /proc/maps labels only the main thread's stack; thread stacks are
+    // anonymous and the default MemoryRegionInfo::eDontKnow is kept.
+    if (llvm::StringRef(name) == "[stack]")
+      region.SetIsStackMemory(eLazyBoolYes);
+    else if (llvm::StringRef(name) == "[heap]")
+      region.SetIsStackMemory(eLazyBoolNo);
+  }
 
   return region;
 }
diff --git a/lldb/source/Plugins/Process/gdb-remote/GDBRemoteCommunicationServerLLGS.cpp b/lldb/source/Plugins/Process/gdb-remote/GDBRemoteCommunicationServerLLGS.cpp
index 769705588fbdf..451855ef80e75 100644
--- a/lldb/source/Plugins/Process/gdb-remote/GDBRemoteCommunicationServerLLGS.cpp
+++ b/lldb/source/Plugins/Process/gdb-remote/GDBRemoteCommunicationServerLLGS.cpp
@@ -2971,6 +2971,13 @@ GDBRemoteCommunicationServerLLGS::Handle_qMemoryRegionInfo(
 
     if (std::optional<unsigned> protection_key = region_info.GetProtectionKey())
       response.Printf("protection-key:%" PRIu32 ";", *protection_key);
+
+    // Type: stack/heap classification, when known.
+    LazyBool is_stack = region_info.IsStackMemory();
+    if (is_stack == eLazyBoolYes)
+      response.PutCString("type:stack;");
+    else if (is_stack == eLazyBoolNo)
+      response.PutCString("type:heap;");
   }
 
   return SendPacketNoLock(response.GetString());
diff --git a/lldb/unittests/Process/Utility/LinuxProcMapsTest.cpp b/lldb/unittests/Process/Utility/LinuxProcMapsTest.cpp
index fcfd76cc67960..84aa206b4e50a 100644
--- a/lldb/unittests/Process/Utility/LinuxProcMapsTest.cpp
+++ b/lldb/unittests/Process/Utility/LinuxProcMapsTest.cpp
@@ -99,7 +99,18 @@ INSTANTIATE_TEST_SUITE_P(
                 MemoryRegionInfo(make_range(0x55a4512f7000, 0x55a451b68000),
                                  eLazyBoolYes, eLazyBoolYes, eLazyBoolNo,
                                  eLazyBoolNo, eLazyBoolYes,
-                                 ConstString("[heap]")),
+                                 ConstString("[heap]"))
+                    .SetIsStackMemory(eLazyBoolNo),
+            },
+            ""),
+        std::make_tuple(
+            "7ffcad8f7000-7ffcad918000 rw-p 00000000 00:00 0    [stack]",
+            MemoryRegionInfos{
+                MemoryRegionInfo(make_range(0x7ffcad8f7000, 0x7ffcad918000),
+                                 eLazyBoolYes, eLazyBoolYes, eLazyBoolNo,
+                                 eLazyBoolNo, eLazyBoolYes,
+                                 ConstString("[stack]"))
+                    .SetIsStackMemory(eLazyBoolYes),
             },
             ""),
         // Multiple entries

felipepiovezan · 2026-07-13T16:28:06Z

As a bit of motivation, I'm trying to implement Linux support for some swift stuff, where CFAs are usually on the heap, and stumbled on this lack of memory information.

jasonmolenda · 2026-07-13T16:54:33Z

+    if (is_stack == eLazyBoolYes)
+      response.PutCString("type:stack;");
+    else if (is_stack == eLazyBoolNo)
+      response.PutCString("type:heap;");


I don't know linux VM but I don't think "if not stack, is heap" is a good assumption. We could be looking at shared memory or binary data or other VM types that are not heap allocations.

But that's not the assumption we're making here, right? What we're doing is:

If stack => stack.
else If heap => not stack.
else => unknown.

Ah, sorry I didn't see where IsStackMemory was initialized,

for (llvm::StringRef entry : llvm::split(value, ',')) { if (entry == "stack") region_info.SetIsStackMemory(eLazyBoolYes); else if (entry == "heap") region_info.SetIsStackMemory(eLazyBoolNo); }

Ah, I see, using a LazyBool here isn't the clearest.
"Yes" == definitely-stack, "No" == definitely-heap, "Calculate" == could be anything incl maybe stack. Yeah this change is fine.

It used to be a different kind of value, whose options were "yes/no/don't know", but I believe @DavidSpickett unified them with LazyBool.

You can do "don't know" with lazybool, it's an alias to calculate:

enum LazyBool { eLazyBoolCalculate = -1, eLazyBoolDontKnow = eLazyBoolCalculate, eLazyBoolNo = 0, eLazyBoolYes = 1 };

And yes we unified them because calculate/don't know was the only state that was different between the 2 types.

You should put a comment as the implicit else:

// Else we do not know, so don't add any type.

Or you could refactor it like:

if (is_stack != eLazyBoolDontKnow) response.Printf("type: %s", is_stack ? "stack" : "heap");

You can do "don't know" with lazybool, it's an alias to calculate:

TIL!

DavidSpickett · 2026-07-14T08:51:27Z

What you quoted is confusing me because you've quoted a field removed in 4.4, and maybe you have a reason to support that even though it's a 10 year old kernel, but the code is actually parsing this one:

              [stack]
                     The initial process's (also known as the main
                     thread's) stack.

Which doesn't have the TID number in it.

Please make it clear which one you're looking for.

This information is in maps as well as smaps, but having read the code, we're using the same logic to parse the header of a smaps entry so it should be fine (and when you were testing it, it was likely reading from smaps anyway).

DavidSpickett · 2026-07-14T08:52:16Z

+    if (llvm::StringRef(name) == "[stack]")
+      region.SetIsStackMemory(eLazyBoolYes);
+    else if (llvm::StringRef(name) == "[heap]")
+      region.SetIsStackMemory(eLazyBoolNo);


It's a few cycles, but I'd put an explicit else here. That's a stronger signal than a comment.

felipepiovezan · 2026-07-14T09:16:40Z

What you quoted is confusing me because you've quoted a field removed in 4.4, and maybe you have a reason to support that even though it's a 10 year old kernel, but the code is actually parsing this one:
              [stack]
                     The initial process's (also known as the main
                     thread's) stack.
Which doesn't have the TID number in it.

Please make it clear which one you're looking for.

This information is in maps as well as smaps, but having read the code, we're using the same logic to parse the header of a smaps entry so it should be fine (and when you were testing it, it was likely reading from smaps anyway).

Yup, good catch. I really meant just [stack] and was trying to say that this will only work for the main thread beacuse the other field was removed. Let me rephrase things.

felipepiovezan

Addressed review comments

felipepiovezan · 2026-07-14T09:14:00Z

+    if (is_stack == eLazyBoolYes)
+      response.PutCString("type:stack;");
+    else if (is_stack == eLazyBoolNo)
+      response.PutCString("type:heap;");


You can do "don't know" with lazybool, it's an alias to calculate:

TIL!

felipepiovezan · 2026-07-14T09:19:17Z

(Updated the PR description, which will be used as the final commit message)

github-actions · 2026-07-14T09:22:04Z

✅ With the latest revision this PR passed the C/C++ code formatter.

github-actions · 2026-07-14T09:40:43Z

🪟 Windows x64 Test Results

33365 tests passed
900 tests skipped

✅ The build succeeded and all tests passed.

DavidSpickett · 2026-07-14T12:32:28Z

got removed in older kernels:

In newer kernels you mean.

felipepiovezan · 2026-07-14T12:38:01Z

got removed in older kernels:

In newer kernels you mean.

Oops, yes. Fixed

felipepiovezan requested review from DavidSpickett and jasonmolenda July 13, 2026 16:27

felipepiovezan requested a review from JDevlieghere as a code owner July 13, 2026 16:27

llvmorg-github-actions Bot added the lldb label Jul 13, 2026

jasonmolenda reviewed Jul 13, 2026

View reviewed changes

DavidSpickett reviewed Jul 14, 2026

View reviewed changes

fixup! Address review comments

c1909ed

felipepiovezan commented Jul 14, 2026

View reviewed changes

felipepiovezan requested review from DavidSpickett and jasonmolenda July 14, 2026 09:19

fixup! clang-format

ab81df4

DavidSpickett approved these changes Jul 14, 2026

View reviewed changes

Uh oh!

Conversation

felipepiovezan commented Jul 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmorg-github-actions Bot commented Jul 13, 2026

Uh oh!

felipepiovezan commented Jul 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

felipepiovezan Jul 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jasonmolenda Jul 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

DavidSpickett commented Jul 14, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

felipepiovezan commented Jul 14, 2026

Uh oh!

felipepiovezan left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

felipepiovezan commented Jul 14, 2026

Uh oh!

github-actions Bot commented Jul 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jul 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🪟 Windows x64 Test Results

Uh oh!

DavidSpickett commented Jul 14, 2026

Uh oh!

felipepiovezan commented Jul 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

felipepiovezan commented Jul 13, 2026 •

edited

Loading

felipepiovezan commented Jul 13, 2026 •

edited

Loading

felipepiovezan Jul 13, 2026 •

edited

Loading

jasonmolenda Jul 14, 2026 •

edited

Loading

github-actions Bot commented Jul 14, 2026 •

edited

Loading

github-actions Bot commented Jul 14, 2026 •

edited

Loading

felipepiovezan commented Jul 14, 2026 •

edited

Loading