[v2,4/4] gdb, gdbserver: introduce the 'x' RSP packet for binary memory read

Message ID b86b2aaa833c3a84116256ad937b0039442f1099.1710420898.git.tankut.baris.aktemur@intel.com
State New
Headers
Series Introduce the 'x' RSP packet |

Checks

Context Check Description
linaro-tcwg-bot/tcwg_gdb_build--master-arm success Testing passed
linaro-tcwg-bot/tcwg_gdb_check--master-arm success Testing passed
linaro-tcwg-bot/tcwg_gdb_build--master-aarch64 success Testing passed
linaro-tcwg-bot/tcwg_gdb_check--master-aarch64 success Testing passed

Commit Message

Aktemur, Tankut Baris March 14, 2024, 1:07 p.m. UTC
  Introduce an RSP packet, 'x', for reading from the remote server
memory in binary format.  The binary write packet, 'X' already exists.
The 'x' packet is essentially the same as 'm', except that the
returned data is in binary format.  For transferring relatively large
data from the memory of the remote process, the 'x' packet can reduce the
transfer costs.

For example, without this patch, fetching ~100MB of data from a remote
target takes

  (gdb) dump binary memory temp.o 0x00007f3ba4c576c0 0x00007f3bab709400
  2024-03-13 16:17:42.626 - command started
  2024-03-13 16:18:24.151 - command finished
  Command execution time: 32.136785 (cpu), 41.525515 (wall)
  (gdb)

whereas with this patch, we obtain

  (gdb) dump binary memory temp.o 0x00007fec39fce6c0 0x00007fec40a80400
  2024-03-13 16:20:48.609 - command started
  2024-03-13 16:21:16.873 - command finished
  Command execution time: 20.447970 (cpu), 28.264202 (wall)
  (gdb)

We see improvements not only when reading bulk data as above, but also
when making a large number of small memory access requests.

For example, without this patch:

  (gdb) pipe x/100000xw $pc | wc -l
  2024-03-13 16:04:57.112 - command started
  25000
  2024-03-13 16:05:10.798 - command finished
  Command execution time: 9.952364 (cpu), 13.686581 (wall)

With this patch:

  (gdb) pipe x/100000xw $pc | wc -l
  2024-03-13 16:06:48.160 - command started
  25000
  2024-03-13 16:06:57.750 - command finished
  Command execution time: 6.541425 (cpu), 9.589839 (wall)
  (gdb)

Another example, where we create a core file of a GDB process.

  (gdb) gcore /tmp/core.1
  ...
  Command execution time: 85.496967 (cpu), 133.224373 (wall)

vs.

  (gdb) gcore /tmp/core.1
  ...
  Command execution time: 48.328885 (cpu), 115.032289 (wall)

Regression-tested on X86-64 using the unix (default) and
native-extended-gdbserver board files.

Reviewed-By: Eli Zaretskii <eliz@gnu.org>
---
 gdb/NEWS                  |  6 ++++
 gdb/doc/gdb.texinfo       | 29 ++++++++++++++++
 gdb/remote.c              | 48 +++++++++++++++++++++++----
 gdbserver/remote-utils.cc | 69 +++++++++++++++++++++++++++++++++------
 gdbserver/remote-utils.h  |  2 ++
 gdbserver/server.cc       | 21 ++++++++++++
 6 files changed, 159 insertions(+), 16 deletions(-)
  

Patch

diff --git a/gdb/NEWS b/gdb/NEWS
index d8ac0bb06a7..5bde7030634 100644
--- a/gdb/NEWS
+++ b/gdb/NEWS
@@ -152,6 +152,12 @@  QThreadOptions in qSupported
   QThreadOptions packet, and the qSupported response can contain the
   set of thread options the remote stub supports.
 
+x addr,length
+  Given ADDR and LENGTH, fetch LENGTH units from the memory at address
+  ADDR and send the fetched data in binary format.  This packet is
+  equivalent to 'm', except that the data in the response are in
+  binary format.
+
 *** Changes in GDB 14
 
 * GDB now supports the AArch64 Scalable Matrix Extension 2 (SME2), which
diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index 9a7d61be65e..a387ad7bd02 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -43122,6 +43122,35 @@  for success (@pxref{Stop Reply Packets})
 @cindex @samp{vStopped} packet
 @xref{Notification Packets}.
 
+@item x @var{addr},@var{length}
+@anchor{x packet}
+@cindex @samp{x} packet
+Read @var{length} addressable memory units starting at address @var{addr}
+(@pxref{addressable memory unit}).  Note that @var{addr} does not have to
+be aligned to any particular boundary.
+
+@cindex alignment of remote memory accesses
+@cindex size of remote memory accesses
+@cindex memory, alignment and size of remote accesses
+The stub need not use any particular size or alignment when gathering
+data from memory for the response; even if @var{addr} is word-aligned
+and @var{length} is a multiple of the word size, the stub is free to
+use byte accesses, or not.  For this reason, this packet may not be
+suitable for accessing memory-mapped I/O devices.
+
+This packet is used only if the stub announces support for it using
+@samp{qSupported}.
+
+Reply:
+@table @samp
+@item @var{XX@dots{}}
+Memory contents as binary data (@pxref{Binary Data}).
+The reply may contain fewer addressable memory units than requested if the
+server was able to read only part of the region of memory.
+@item E @var{NN}
+for an error
+@end table
+
 @item X @var{addr},@var{length}:@var{XX@dots{}}
 @anchor{X packet}
 @cindex @samp{X} packet
diff --git a/gdb/remote.c b/gdb/remote.c
index 14c8b020b1e..d1a7cdb2cd5 100644
--- a/gdb/remote.c
+++ b/gdb/remote.c
@@ -194,6 +194,7 @@  struct packet_result
 enum {
   PACKET_vCont = 0,
   PACKET_X,
+  PACKET_x,
   PACKET_qSymbol,
   PACKET_P,
   PACKET_p,
@@ -5763,6 +5764,7 @@  static const struct protocol_feature remote_protocol_features[] = {
   { "no-resumed", PACKET_DISABLE, remote_supported_packet, PACKET_no_resumed },
   { "memory-tagging", PACKET_DISABLE, remote_supported_packet,
     PACKET_memory_tagging_feature },
+  { "x", PACKET_DISABLE, remote_supported_packet, PACKET_x },
 };
 
 static char *remote_support_xml;
@@ -9582,24 +9584,56 @@  remote_target::remote_read_bytes_1 (CORE_ADDR memaddr, gdb_byte *myaddr,
   todo_units = std::min (len_units,
 			 (ULONGEST) (buf_size_bytes / unit_size) / 2);
 
-  /* Construct "m"<memaddr>","<len>".  */
+  /* Determine which packet format to use.  */
+  char packet_format = 'm';
+  switch (m_features.packet_support (PACKET_x))
+    {
+    case PACKET_ENABLE:
+      packet_format = 'x';
+      break;
+
+    case PACKET_DISABLE:
+      packet_format = 'm';
+      break;
+
+    case PACKET_SUPPORT_UNKNOWN:
+      internal_error (_("remote_read_bytes_1: bad internal state"));
+
+    default:
+      /* This is unexpected but we can still continue with 'm'.  */
+      break;
+    }
+
+  /* Construct "m/x"<memaddr>","<len>".  */
   memaddr = remote_address_masked (memaddr);
   p = rs->buf.data ();
-  *p++ = 'm';
+  *p++ = packet_format;
   p += hexnumstr (p, (ULONGEST) memaddr);
   *p++ = ',';
   p += hexnumstr (p, (ULONGEST) todo_units);
   *p = '\0';
   putpkt (rs->buf);
-  getpkt (&rs->buf);
+  int packet_len = getpkt (&rs->buf);
+  if (packet_len < 0)
+    return TARGET_XFER_E_IO;
+
   if (rs->buf[0] == 'E'
       && isxdigit (rs->buf[1]) && isxdigit (rs->buf[2])
       && rs->buf[3] == '\0')
     return TARGET_XFER_E_IO;
-  /* Reply describes memory byte by byte, each byte encoded as two hex
-     characters.  */
+
   p = rs->buf.data ();
-  decoded_bytes = hex2bin (p, myaddr, todo_units * unit_size);
+  if (packet_format == 'x')
+    decoded_bytes = remote_unescape_input ((const gdb_byte *) p,
+					   packet_len, myaddr,
+					   todo_units * unit_size);
+  else
+    {
+      /* Reply describes memory byte by byte, each byte encoded as two hex
+	 characters.  */
+      decoded_bytes = hex2bin (p, myaddr, todo_units * unit_size);
+    }
+
   /* Return what we have.  Let higher layers handle partial reads.  */
   *xfered_len_units = (ULONGEST) (decoded_bytes / unit_size);
   return (*xfered_len_units != 0) ? TARGET_XFER_OK : TARGET_XFER_EOF;
@@ -15839,6 +15873,8 @@  Show the maximum size of the address (in bits) in a memory packet."), NULL,
 
   add_packet_config_cmd (PACKET_X, "X", "binary-download", 1);
 
+  add_packet_config_cmd (PACKET_x, "x", "binary-read", 0);
+
   add_packet_config_cmd (PACKET_vCont, "vCont", "verbose-resume", 0);
 
   add_packet_config_cmd (PACKET_QPassSignals, "QPassSignals", "pass-signals",
diff --git a/gdbserver/remote-utils.cc b/gdbserver/remote-utils.cc
index 5e7c38056d0..995fb621ae7 100644
--- a/gdbserver/remote-utils.cc
+++ b/gdbserver/remote-utils.cc
@@ -1334,6 +1334,13 @@  decode_M_packet (const char *from, CORE_ADDR *mem_addr_ptr,
   hex2bin (from, *to_p, *len_ptr);
 }
 
+void
+decode_x_packet (const char *from, CORE_ADDR *mem_addr_ptr,
+		 unsigned int *len_ptr)
+{
+  decode_m_packet_params (from, mem_addr_ptr, len_ptr, '\0');
+}
+
 int
 decode_X_packet (char *from, int packet_len, CORE_ADDR *mem_addr_ptr,
 		 unsigned int *len_ptr, unsigned char **to_p)
@@ -1478,25 +1485,45 @@  look_up_one_symbol (const char *name, CORE_ADDR *addrp, int may_ask_gdb)
      while it figures out the address of the symbol.  */
   while (1)
     {
-      if (cs.own_buf[0] == 'm')
+      int new_len = -1;
+      if (cs.own_buf[0] == 'm' || cs.own_buf[0] == 'x')
 	{
 	  CORE_ADDR mem_addr;
 	  unsigned char *mem_buf;
 	  unsigned int mem_len;
 
-	  decode_m_packet (&cs.own_buf[1], &mem_addr, &mem_len);
+	  if (cs.own_buf[0] == 'm')
+	    decode_m_packet (&cs.own_buf[1], &mem_addr, &mem_len);
+	  else
+	    decode_x_packet (&cs.own_buf[1], &mem_addr, &mem_len);
+
 	  mem_buf = (unsigned char *) xmalloc (mem_len);
 	  if (read_inferior_memory (mem_addr, mem_buf, mem_len) == 0)
-	    bin2hex (mem_buf, cs.own_buf, mem_len);
+	    {
+	      if (cs.own_buf[0] == 'm')
+		bin2hex (mem_buf, cs.own_buf, mem_len);
+	      else
+		{
+		  int out_len_units;
+		  new_len
+		    = remote_escape_output (mem_buf, mem_len, 1,
+					    (gdb_byte *) cs.own_buf,
+					    &out_len_units, PBUFSIZ);
+		  suppress_next_putpkt_log ();
+		}
+	    }
 	  else
 	    write_enn (cs.own_buf);
 	  free (mem_buf);
-	  if (putpkt (cs.own_buf) < 0)
+
+	  int res = ((new_len == -1)
+		     ? putpkt (cs.own_buf)
+		     : putpkt_binary (cs.own_buf, new_len));
+	  if (res < 0)
 	    return -1;
 	}
       else if (cs.own_buf[0] == 'v')
 	{
-	  int new_len = -1;
 	  handle_v_requests (cs.own_buf, len, &new_len);
 	  if (new_len != -1)
 	    putpkt_binary (cs.own_buf, new_len);
@@ -1571,18 +1598,36 @@  relocate_instruction (CORE_ADDR *to, CORE_ADDR oldloc)
      wait for the qRelocInsn "response".  That requires re-entering
      the main loop.  For now, this is an adequate approximation; allow
      GDB to access memory.  */
-  while (cs.own_buf[0] == 'm' || cs.own_buf[0] == 'M' || cs.own_buf[0] == 'X')
+  while (cs.own_buf[0] == 'm' || cs.own_buf[0] == 'M'
+	 || cs.own_buf[0] == 'X' || cs.own_buf[0] == 'x')
     {
       CORE_ADDR mem_addr;
       unsigned char *mem_buf = NULL;
       unsigned int mem_len;
+      int new_packet_len = -1;
 
-      if (cs.own_buf[0] == 'm')
+      if (cs.own_buf[0] == 'm' || cs.own_buf[0] == 'x')
 	{
-	  decode_m_packet (&cs.own_buf[1], &mem_addr, &mem_len);
+	  if (cs.own_buf[0] == 'm')
+	    decode_m_packet (&cs.own_buf[1], &mem_addr, &mem_len);
+	  else
+	    decode_x_packet (&cs.own_buf[1], &mem_addr, &mem_len);
+
 	  mem_buf = (unsigned char *) xmalloc (mem_len);
 	  if (read_inferior_memory (mem_addr, mem_buf, mem_len) == 0)
-	    bin2hex (mem_buf, cs.own_buf, mem_len);
+	    {
+	      if (cs.own_buf[0] == 'm')
+		bin2hex (mem_buf, cs.own_buf, mem_len);
+	      else
+		{
+		  int out_len_units;
+		  new_packet_len
+		    = remote_escape_output (mem_buf, mem_len, 1,
+					    (gdb_byte *) cs.own_buf,
+					    &out_len_units, PBUFSIZ);
+		  suppress_next_putpkt_log ();
+		}
+	    }
 	  else
 	    write_enn (cs.own_buf);
 	}
@@ -1604,7 +1649,11 @@  relocate_instruction (CORE_ADDR *to, CORE_ADDR oldloc)
 	    write_enn (cs.own_buf);
 	}
       free (mem_buf);
-      if (putpkt (cs.own_buf) < 0)
+
+      int res = ((new_packet_len == -1)
+		 ? putpkt (cs.own_buf)
+		 : putpkt_binary (cs.own_buf, new_packet_len));
+      if (res < 0)
 	return -1;
       len = getpkt (cs.own_buf);
       if (len < 0)
diff --git a/gdbserver/remote-utils.h b/gdbserver/remote-utils.h
index 4681ca12f55..65ce604f9dc 100644
--- a/gdbserver/remote-utils.h
+++ b/gdbserver/remote-utils.h
@@ -57,6 +57,8 @@  void decode_m_packet (const char *from, CORE_ADDR * mem_addr_ptr,
 		      unsigned int *len_ptr);
 void decode_M_packet (const char *from, CORE_ADDR * mem_addr_ptr,
 		      unsigned int *len_ptr, unsigned char **to_p);
+void decode_x_packet (const char *from, CORE_ADDR *mem_addr_ptr,
+		      unsigned int *len_ptr);
 int decode_X_packet (char *from, int packet_len, CORE_ADDR * mem_addr_ptr,
 		     unsigned int *len_ptr, unsigned char **to_p);
 int decode_xfer_write (char *buf, int packet_len,
diff --git a/gdbserver/server.cc b/gdbserver/server.cc
index 74c7763d777..96e6e551784 100644
--- a/gdbserver/server.cc
+++ b/gdbserver/server.cc
@@ -2840,6 +2840,9 @@  handle_query (char *own_buf, int packet_len, int *new_packet_len_p)
       if (target_supports_memory_tagging ())
 	strcat (own_buf, ";memory-tagging+");
 
+      /* Binary memory read support.  */
+      strcat (own_buf, ";x+");
+
       /* Reinitialize components as needed for the new connection.  */
       hostio_handle_new_gdb_connection ();
       target_handle_new_gdb_connection ();
@@ -4690,6 +4693,24 @@  process_serial_event (void)
       else
 	write_enn (cs.own_buf);
       break;
+    case 'x':
+      {
+	require_running_or_break (cs.own_buf);
+	decode_x_packet (&cs.own_buf[1], &mem_addr, &len);
+	int res = gdb_read_memory (mem_addr, mem_buf, len);
+	if (res < 0)
+	  write_enn (cs.own_buf);
+	else
+	  {
+	    int out_len_units;
+	    new_packet_len
+	      = remote_escape_output (mem_buf, res, 1,
+				      (gdb_byte *) cs.own_buf,
+				      &out_len_units, PBUFSIZ);
+	    suppress_next_putpkt_log ();
+	  }
+      }
+      break;
     case 'X':
       require_running_or_break (cs.own_buf);
       if (decode_X_packet (&cs.own_buf[1], packet_len - 1,