Make ida integration work in shared libraries (#3172)

* Make ida integration work in shared libraries

* lint

* edge case happens too often

* make the check faster (and change the contract abit)

pwndbg/aglib/vmmap.py

@@ -41,3 +41,44 @@ def find(address: int | pwndbg.dbg_mod.Value | None) -> pwndbg.lib.memory.Page |
         return page
 
     return pwndbg.aglib.vmmap_custom.explore(address)
+
+
+def addr_region_start(address: int | pwndbg.dbg_mod.Value) -> int | None:
+    """
+    Let's define a "region" as contiguous memory compromised of memory mappings
+    which all have the same object file name. Also referred to as "File (Base)" by
+    `xinfo`.
+
+    Returns:
+        The start of the memory region this address belongs to, or None if the address
+        is not mapped.
+    """
+    address = int(address)
+    if address < 0:
+        return None
+
+    mappings = sorted(pwndbg.aglib.vmmap.get(), key=lambda p: p.vaddr)
+    idx = -1
+    for i in range(len(mappings)):
+        if mappings[i].start <= address < mappings[i].end:
+            idx = i
+            break
+
+    if idx == -1:
+        # Maybe we can find the page by exploring.
+        explored_page = pwndbg.aglib.vmmap_custom.explore(address)
+        if not explored_page:
+            return None
+
+        # We know vmmap_custom.explore() can only find one page, it does
+        # not cascade a whole region so there is no need to look backwards.
+        return explored_page.start
+
+    # Look backwards from i to find all the mappings with the same name.
+    objname = mappings[i].objfile
+    while i > 0 and objname == mappings[i - 1].objfile:
+        i -= 1
+
+    # There might be other mappings with the name "objname" in the address space
+    # but they are not contiguous with us, so we don't care.
+    return mappings[i].start

pwndbg/commands/xinfo.py

@@ -61,17 +61,19 @@ def xinfo_mmap_file(page: Page, addr: int) -> None:
 
     file_name = page.objfile
 
-    objpages = filter(lambda p: p.objfile == file_name, pwndbg.aglib.vmmap.get())
-    first = sorted(objpages, key=lambda p: p.vaddr)[0]
+    region_start = pwndbg.aglib.vmmap.addr_region_start(addr)
+    if region_start is None:
+        print("The file is not contiguous in memory.")
+        return
 
     # print offset from ELF base load address
-    rva = addr - first.vaddr
-    print_line("File (Base)", addr, first.vaddr, rva, "+")
+    rva = addr - region_start
+    print_line("File (Base)", addr, region_start, rva, "+")
 
     # find possible LOAD segments that designate memory and file backings
     containing_loads = [
         seg
-        for seg in pwndbg.aglib.elf.get_containing_segments(file_name, first.vaddr, addr)
+        for seg in pwndbg.aglib.elf.get_containing_segments(file_name, region_start, addr)
         if seg["p_type"] == "PT_LOAD"
     ]
 
@@ -89,7 +91,7 @@ def xinfo_mmap_file(page: Page, addr: int) -> None:
     else:
         print(f"{'File (Disk)'.rjust(20)} {M.get(addr)} = [not file backed]")
 
-    containing_sections = pwndbg.aglib.elf.get_containing_sections(file_name, first.vaddr, addr)
+    containing_sections = pwndbg.aglib.elf.get_containing_sections(file_name, region_start, addr)
     if len(containing_sections) > 0:
         print("\n Containing ELF sections:")
         for sec in containing_sections:

pwndbg/integration/ida.py

@@ -180,18 +180,18 @@ def can_connect() -> bool:
 
 
 def l2r(addr: int) -> int:
-    exe = pwndbg.aglib.elf.exe()
-    if not exe:
-        raise Exception("Can't find EXE base")
-    result = (addr - int(exe.address) + base()) & pwndbg.aglib.arch.ptrmask
+    region_start = pwndbg.aglib.vmmap.addr_region_start(addr)
+    if region_start is None:
+        return 0
+    result = (addr - region_start + base()) & pwndbg.aglib.arch.ptrmask
     return result
 
 
 def r2l(addr: int) -> int:
-    exe = pwndbg.aglib.elf.exe()
-    if not exe:
-        raise Exception("Can't find EXE base")
-    result = (addr - base() + int(exe.address)) & pwndbg.aglib.arch.ptrmask
+    region_start = pwndbg.aglib.vmmap.addr_region_start(addr)
+    if region_start is None:
+        return 0
+    result = (addr - base() + region_start) & pwndbg.aglib.arch.ptrmask
     return result