gdb, target remote, threadbased application in uml

[Michael Shuldman <michael dot shuldman at ericsson dot com>]

Hello, I'm having a problem running gdb on a threadbased application.
The application is running inside uml, with pid 765.  UML is running on
the same physical machine as I'm debugging from.

Status is this:

On the target (uml) side I do the following, to start gdbserver:
$ gdbserver 172.16.254.1:2024 --attach 765
Attached; pid = 765
Listening on port 2024
Remote debugging from host 172.16.254.1


On the host side, I then start gdb and do the following:
set solib-search-path usr/local/...
target remote 172.16.254.2:2024
Remote debugging using 172.16.254.2:2024
0xffffe410 in ?? ()

[Thread debugging using libthread_db enabled]
[New Thread 1076389568 (LWP 765)]
Can't attach LWP 765: No such process
(gdb) 
(gdb) bt
#0  0xffffe410 in ?? ()
#1  0xbf502568 in ?? ()
2  0x0000081c in ?? ()
#3  0xbf501d34 in ?? ()
#4  0x4004556b in __read_nocancel () from
/lib/tls/i686/cmov/libpthread.so.0
#5  0x40026a24 in Msg__Wait (pMsgQue=0x8502d00,
ePreemptive=NON_PREEMPTIVE)
    at /home/a8648/src/oms.svn/main/basic/ipc/msg.c:235
#6  0x0804ad51 in main (argc=1, argv=0xbf502644)
    at /home/a8648/src/oms.svn/main/mpls/rcim/rcim.c:259
(gdb) bt
#0  0xffffe410 in ?? ()
#1  0xbf502568 in ?? ()
#2  0x0000081c in ?? ()
#3  0xbf501d34 in ?? ()
#4  0x4004556b in __read_nocancel () from /lib/tls/i686/cmov/libpthread.so.0
#5 ... <properly resolved functions>
(gdb) n
Cannot find bounds of current function
(gdb) c
Continuing.
0xffffe410 in ?? ()
ptrace: No such process.


I thought the "Can't attach LWP 765: No such process" errormessage was
something read from the target host, but I strace(1)-ed gdbserver on
the target without seeing any errormessages.

I then straced gdb on the host, and see that gdb on the host connects
to target, and then seems to write a bunch of instructions to gdbserver
on target:

connect(6, {sa_family=AF_INET, sin_port=htons(2024), sin_addr=inet_addr("172.16.254.2")}, 16) = -1 EINPROGRESS (Operation now in progress)
...
send(6, "+", 1, 0)                      = 1
send(6, "$Hc-1#09", 8, 0)               = 8
...

This is as I would expect.

Then the same gdb on host starts doing ptrace(2) calls, which I did
not expect:

ptrace(PTRACE_PEEKUSER, 1076389568, offsetof(struct user, u_debugreg) +
24, [0xb
7eee320]) = -1 ESRCH (No such process)
write(1, "[Switching to thread 1076389568]"..., 33) = 33
igprocmask(SIG_BLOCK, NULL, [], 8)  = 0
rt_sigprocmask(SIG_BLOCK, NULL, [], 8)  = 0
rt_sigprocmask(SIG_BLOCK, NULL, [], 8)  = 0
send(6, "$mffffe410,8#63", 15, 0)       = 15
select(7, [6], NULL, [6], {1, 0})       = 1 (in [6], left {1, 0})
...
write(1, "[Thread debugging using libthrea"..., 46) = 46
ptrace(PTRACE_PEEKTEXT, 1076389568, 0x40047cb0, [0xb7e253b4]) = -1 ESRCH
(No such process)
send(6, "$m40047cb0,4#c1", 15, 0)       = 15
select(7, [6], NULL, [6], {1, 0})       = 1 (in [6], left {1, 0})
recv(6, "+$08000000#88", 8192, 0)       = 13
send(6, "+", 1, 0)                      = 1
ptrace(PTRACE_PEEKTEXT, 1076389568, 0x4004c52c, [0xbfaacd30]) = -1 ESRCH
(No such process)
send(6, "$m4004c52c,8#c6", 15, 0)       = 15
select(7, [6], NULL, [6], {1, 0})       = 1 (in [6], left {1, 0})
recv(6, "+$8001000000000000#09", 8192, 0) = 21
send(6, "+", 1, 0)                      = 1
open("/proc/1076389568/mem", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such
file or 
directory)
ptrace(PTRACE_PEEKTEXT, 1076389568, 0x40047ca4, [0]) = -1 ESRCH (No such process
)
ptrace(PTRACE_ATTACH, 765, 0, 0)        = -1 ESRCH (No such process)
 

I don't know why gdb on host is calling ptrace(2) at all.

Thanks for any help,

-- 

  _ // 
  \X/ -- Michael Shuldman <michael.shuldman@ericsson.com>