[ofa-general] RE: dapl bug [PATCH] dapltest: include definitions for inet_ntoa.

[Davis, Arlin R]

Steve,
 
Sorry, this was fixed in v2.0 library but apparently it didn't get
pushed back v1.2. 

dapltest: include definitions for inet_ntoa.

At load time the symbol was resolved but with the
default definition of int, instead of char*, it caused
segfault. Add correct include files in dapl_mdep_user.h
for linux.

Signed-off by: Arlin Davis <ardavis at ichips.intel.com>
---
 test/dapltest/mdep/linux/dapl_mdep_user.h |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/test/dapltest/mdep/linux/dapl_mdep_user.h
b/test/dapltest/mdep/linux/dapl_mdep_user.h
index 7fadbea..16170a7 100755
--- a/test/dapltest/mdep/linux/dapl_mdep_user.h
+++ b/test/dapltest/mdep/linux/dapl_mdep_user.h
@@ -43,6 +43,11 @@
 #include <string.h>
 #include <sys/times.h>
 
+/* inet_ntoa */
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+
 /* Default Device Name */
 #define DT_MdepDeviceName    "OpenIB-cma"
 
-- 
1.5.2.5



>-----Original Message-----
>From: Steve Wise [mailto:swise at opengridcomputing.com] 
>Sent: Thursday, April 24, 2008 12:58 PM
>To: Arlin Davis
>Cc: OpenFabrics General
>Subject: dapl bug?
>
>Hey Arlin,
>
>Have you ever seen this?  I hit this 100% of the time trying the 1.2 
>version of dapltest on an ofed-1.3 system.  The debug info below was 
>obtained by builting the src rpm with debug enabled...
>
>> (gdb) r -T T -d -s vic11-10g -D chelsio -i 10 client SR 256 
>server SR 
>> 256 client SR 256 server SR 256
>> Starting program: /usr/bin/dapltest -T T -d -s vic11-10g -D 
>chelsio -i 
>> 10 client SR 256 server SR 256 client SR 256 server SR 256
>> [Thread debugging using libthread_db enabled]
>> [New Thread 46912498371600 (LWP 6654)]
>> -------------------------------------
>> TransCmd.server_name              : vic11-10g
>> TransCmd.num_iterations           : 10
>> TransCmd.num_threads              : 1
>> TransCmd.eps_per_thread           : 1
>> TransCmd.validate                 : 0
>> TransCmd.dapl_name                : chelsio
>> TransCmd.num_ops                  : 4
>> TransCmd.op[0].transfer_type      : SEND_RECV  (client)
>> TransCmd.op[0].seg_size           : 256
>> TransCmd.op[0].num_segs           : 1
>> TransCmd.op[0].reap_send_on_recv  : 0
>> TransCmd.op[1].transfer_type      : SEND_RECV  (server)
>> TransCmd.op[1].seg_size           : 256
>> TransCmd.op[1].num_segs           : 1
>> TransCmd.op[1].reap_send_on_recv  : 0
>> TransCmd.op[2].transfer_type      : SEND_RECV  (client)
>> TransCmd.op[2].seg_size           : 256
>> TransCmd.op[2].num_segs           : 1
>> TransCmd.op[2].reap_send_on_recv  : 0
>> TransCmd.op[3].transfer_type      : SEND_RECV  (server)
>> TransCmd.op[3].seg_size           : 256
>> TransCmd.op[3].num_segs           : 1
>> TransCmd.op[3].reap_send_on_recv  : 0
>> Server Name: vic11-10g
>>
>> Program received signal SIGSEGV, Segmentation fault.
>> [Switching to Thread 46912498371600 (LWP 6654)]
>> 0x00000032f04760b0 in strlen () from /lib64/libc.so.6
>> (gdb) bt
>> #0  0x00000032f04760b0 in strlen () from /lib64/libc.so.6
>> #1  0x00000032f044602b in vfprintf () from /lib64/libc.so.6
>> #2  0x00000032f044bdea in printf () from /lib64/libc.so.6
>> #3  0x0000000000403900 in DT_NetAddrLookupHostAddress 
>> (to_netaddr=0x7e16f88, hostname=0x7e1658c "vic11-10g") at 
>> cmd/dapl_netaddr.c:136
>> #4  0x00000000004026cb in DT_Params_Parse (argc=<value 
>optimized out>, 
>> argv=<value optimized out>, params_ptr=0x7e16580) at 
>cmd/dapl_params.c:205
>> #5  0x000000000040211f in dapltest (argc=22, argv=0x7fff48e9b5f8) at 
>> cmd/dapl_main.c:88
>> #6  0x00000032f041d8a4 in __libc_start_main () from /lib64/libc.so.6
>> #7  0x0000000000401f59 in _start ()
>> (gdb) 
>
>Its hurling in DT_Mdep_printf() here:
>
>> 134         /* Pull out IP address and print it as a sanity check */
>> 135         DT_Mdep_printf ("Server Name: %s \n", hostname);
>> 136         DT_Mdep_printf ("Server Net Address: %s\n",
>> 137                         inet_ntoa(((struct sockaddr_in 
>> *)target->ai_addr)->sin_addr));
>
>The ai_addr looks ok though:
>> (gdb) p/x *((struct sockaddr_in *)target->ai_addr)
>> $3 = {sin_family = 0x2, sin_port = 0x0, sin_addr = {s_addr = 
>> 0x8846a8c0}, sin_zero = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}}
>> (gdb)
>>
>
>Ever seen this?
>
>Steve.
>