Parsing problem of gdb 'M' packet

Hello,

I've found a problem in the code parsing the gdb 'M' packet. During gdb remote sessions, I saw strange things when writing to memory :

============================================================
(gdb) print addr
$1 = 0
(gdb) set addr=12
(gdb) print addr
$2 = 49152
============================================================

So, I went into Qemu code, and found a problem in gdbstub.c. The format of a 'M' packet is Maddr,length:XX... as stated on [1]. So the addr is separated from the length using a comma, and the length from the data using a colon.

However, the Qemu code assumed that all fields were seperated with a comma, leading to wrong analysis of the data field.

The included one-line patch fixes the problem :

=============================================================
(gdb) print addr
$1 = 0
(gdb) set addr=12
(gdb) print addr
$2 = 12
=============================================================

Thomas

[1] http://sources.redhat.com/gdb/current/onlinedocs/gdb_33.html#SEC664

PETAZZONI Thomas - [email protected]
http://thomas.enix.org - Jabber: [email protected]
KOS: http://kos.enix.org/ - SOS: http://sos.enix.org
Fingerprint : 0BE1 4CF3 CEA4 AC9D CC6E  1624 F653 CB30 98D3 F7A7

Index: gdbstub.c
===================================================================
RCS file: /cvsroot/qemu/qemu/gdbstub.c,v
retrieving revision 1.22

diff -u -u -r1.22 gdbstub.c
@@ -420,7 +420,7 @@
         if (*p == ',')
             p++;
         len = strtoul(p, (char **)&p, 16);
-        if (*p == ',')
+        if (*p == ':')
             p++;
         hextomem(mem_buf, p, len);
         if (cpu_memory_rw_debug(env, addr, mem_buf, len, 1) != 0)

Hello,

Don't hesitate to tell me if something is wrong with this patch.

Thomas

PETAZZONI Thomas - [email protected]
http://thomas.enix.org - Jabber: [email protected]
KOS: http://kos.enix.org/ - SOS: http://sos.enix.org
Fingerprint : 0BE1 4CF3 CEA4 AC9D CC6E  1624 F653 CB30 98D3 F7A7