[Lldb-commits] [lldb] [lldb] Support partial memory reads on windows (PR #106981)

[David Spickett via lldb-commits]

================
@@ -276,16 +276,29 @@ Status ProcessDebugger::ReadMemory(lldb::addr_t vm_addr, void *buf, size_t size,
   LLDB_LOG(log, "attempting to read {0} bytes from address {1:x}", size,
            vm_addr);
 
-  HostProcess process = m_session_data->m_debugger->GetProcess();
+  lldb::process_t handle = m_session_data->m_debugger->GetProcess()
+                               .GetNativeProcess()
+                               .GetSystemHandle();
   void *addr = reinterpret_cast<void *>(vm_addr);
   SIZE_T num_of_bytes_read = 0;
-  if (!::ReadProcessMemory(process.GetNativeProcess().GetSystemHandle(), addr,
-                           buf, size, &num_of_bytes_read)) {
-    error = Status(GetLastError(), eErrorTypeWin32);
-    LLDB_LOG(log, "reading failed with error: {0}", error);
-  } else {
+  if (::ReadProcessMemory(handle, addr, buf, size, &num_of_bytes_read)) {
+    bytes_read = num_of_bytes_read;
+    return Status();
+  }
+  error = Status(GetLastError(), eErrorTypeWin32);
+  MemoryRegionInfo info;
+  if (GetMemoryRegionInfo(vm_addr, info).Fail() ||
----------------
DavidSpickett wrote:

I'm wondering what happens if the read goes across multiple memory regions then fails. For example:
```
region 1
region 2
<unmapped> <<< want to read into this region.
```
We will read region 1, 2 and then fail. vm_addr still points to the start of region 1 so we'll return only the contents of region 1, but we could have read region 2 as well.

I had to do something like this for memory tagging in `MemoryTagManagerAArch64MTE::MakeTaggedRange` which checks that all the memory regions in a range are memory tagged.

Or perhaps you:
1. Want to avoid the complication of that at this time
2. Know that memory reads here won't span a memory region so you don't need to bother


https://github.com/llvm/llvm-project/pull/106981