[PATCH] [gdb/aarch64] Use safer memory read routines

[PATCH] [gdb/aarch64] Use safer memory read routines

[Luis Machado]

  PR 28796: GDB locks up after stopping on a breakpoint in invalidated memory
	    on aarch64

  As reported, we are using some memory read routines that don't handle read
  errors gracefully. Convert those to use the safe_* versions if available.

  This allows the code to handle those read errors in a more sensible way.
---
 gdb/aarch64-tdep.c | 42 ++++++++++++++++++++++++++++++++++++------
 1 file changed, 36 insertions(+), 6 deletions(-)

diff --git a/gdb/aarch64-tdep.c b/gdb/aarch64-tdep.c
index d9ddc84a140..07330356fdc 100644
--- a/gdb/aarch64-tdep.c
+++ b/gdb/aarch64-tdep.c
@@ -2940,8 +2940,18 @@ aarch64_software_single_step (struct regcache *regcache)
   CORE_ADDR breaks[2] = { CORE_ADDR_MAX, CORE_ADDR_MAX };
   CORE_ADDR loc = pc;
   CORE_ADDR closing_insn = 0;
-  uint32_t insn = read_memory_unsigned_integer (loc, insn_size,
-						byte_order_for_code);
+
+  ULONGEST insn_from_memory;
+  if (!safe_read_memory_unsigned_integer (loc, insn_size,
+					  byte_order_for_code,
+					  &insn_from_memory))
+  {
+    /* Assume we don't have a atomic sequence, as we couldn't read the
+       instruction in this location.  */
+    return {};
+  }
+
+  uint32_t insn = insn_from_memory;
   int index;
   int insn_count;
   int bc_insn_count = 0; /* Conditional branch instruction count.  */
@@ -2958,9 +2968,17 @@ aarch64_software_single_step (struct regcache *regcache)
   for (insn_count = 0; insn_count < atomic_sequence_length; ++insn_count)
     {
       loc += insn_size;
-      insn = read_memory_unsigned_integer (loc, insn_size,
-					   byte_order_for_code);
 
+      if (!safe_read_memory_unsigned_integer (loc, insn_size,
+					      byte_order_for_code,
+					      &insn_from_memory))
+      {
+	/* Assume we don't have a atomic sequence, as we couldn't read the
+	   instruction in this location.  */
+	return {};
+      }
+
+      insn = insn_from_memory;
       if (aarch64_decode_insn (insn, &inst, 1, NULL) != 0)
 	return {};
       /* Check if the instruction is a conditional branch.  */
@@ -3259,9 +3277,15 @@ aarch64_displaced_step_copy_insn (struct gdbarch *gdbarch,
 				  struct regcache *regs)
 {
   enum bfd_endian byte_order_for_code = gdbarch_byte_order_for_code (gdbarch);
-  uint32_t insn = read_memory_unsigned_integer (from, 4, byte_order_for_code);
   struct aarch64_displaced_step_data dsd;
   aarch64_inst inst;
+  ULONGEST insn_from_memory;
+
+  if (!safe_read_memory_unsigned_integer (from, 4, byte_order_for_code,
+					  &insn_from_memory))
+    return nullptr;
+
+  uint32_t insn = insn_from_memory;
 
   if (aarch64_decode_insn (insn, &inst, 1, NULL) != 0)
     return NULL;
@@ -3472,7 +3496,13 @@ aarch64_stack_frame_destroyed_p (struct gdbarch *gdbarch, CORE_ADDR pc)
     return 0;
 
   enum bfd_endian byte_order_for_code = gdbarch_byte_order_for_code (gdbarch);
-  uint32_t insn = read_memory_unsigned_integer (pc, 4, byte_order_for_code);
+
+  ULONGEST insn_from_memory;
+  if (!safe_read_memory_unsigned_integer (pc, 4, byte_order_for_code,
+					  &insn_from_memory))
+    return 0;
+
+  uint32_t insn = insn_from_memory;
 
   aarch64_inst inst;
   if (aarch64_decode_insn (insn, &inst, 1, nullptr) != 0)
-- 
2.25.1

Re: [PATCH] [gdb/aarch64] Use safer memory read routines

[Tom Tromey]

>>>>> "Luis" == Luis Machado via Gdb-patches <gdb-patches@sourceware.org> writes:

Luis>   PR 28796: GDB locks up after stopping on a breakpoint in invalidated memory
Luis> 	    on aarch64

If you write "PR tdep/28796" instead, the commit will show up in
bugzilla as well.

Also we've adopted the convention of adding a "Bug:" trailer with the
full URL, though at present this doesn't really do anything.

I think this is ok with these minor tweaks.

Tom

Re: [PATCH] [gdb/aarch64] Use safer memory read routines

[Luis Machado]

Hi Tom,

On 11/8/22 16:59, Tom Tromey wrote:
>>>>>> "Luis" == Luis Machado via Gdb-patches <gdb-patches@sourceware.org> writes:
> 
> Luis>   PR 28796: GDB locks up after stopping on a breakpoint in invalidated memory
> Luis> 	    on aarch64
> 
> If you write "PR tdep/28796" instead, the commit will show up in
> bugzilla as well.

Indeed. I keep forgetting about that pattern.

> 
> Also we've adopted the convention of adding a "Bug:" trailer with the
> full URL, though at present this doesn't really do anything.
> 
> I think this is ok with these minor tweaks.

Fixed now and pushed.

> 
> Tom