Re: toolchain/54590: running binaries from gdb fails

[christos%zoulas.com@localhost (Christos Zoulas)]

On Oct 1,  4:45pm, martin%duskware.de@localhost (Martin Husemann) wrote:
-- Subject: Re: toolchain/54590: running binaries from gdb fails

| The following reply was made to PR toolchain/54590; it has been noted by GNATS.
| 
| From: Martin Husemann <martin%duskware.de@localhost>
| To: gnats-bugs%netbsd.org@localhost
| Cc: kamil%NetBSD.org@localhost
| Subject: Re: toolchain/54590: running binaries from gdb fails
| Date: Tue, 1 Oct 2019 18:44:32 +0200
| 
|  Example ktrace for this issue:
|  
|   24228      1 gdb      CALL  execve(0xbfffee18,0xbb8b6070,0xbb89fd00)
|   24228      1 gdb      NAMI  "/usr/pkg/bin/tcsh"
|   24228      1 gdb      NAMI  "/usr/libexec/ld.elf_so"
|   24228      1 tcsh     EMUL  "netbsd"
|    7933      1 gdb      RET   __vfork14 24228/0x5ea4
|    7933      1 gdb      CALL  __wait450(0x5ea4,0xbfffe18c,0,0)
|    7933      1 gdb      RET   __wait450 24228/0x5ea4
|    7933      1 gdb      CALL  ptrace(PT_GET_PROCESS_STATE,0x5ea4,0xbfffe190,8)
|    7933      1 gdb      RET   ptrace -1 errno 22 Invalid argument
|  
|  24228 == 0x5ea4
|  and sizeof struct ptrace_state should be == 8.
|  
|  So I guess sys_ptrace_common.c:701 strikes and I really don't get how this
|  is supposed to work.

The trap events reported in eventswitch are:

        KASSERT((code == TRAP_CHLD) || (code == TRAP_LWP) ||
		(code == TRAP_EXEC));

But the trap events that can be retrieved in get_process_stats are:

            (t->p_sigctx.ps_info._code != TRAP_CHLD &&
	    t->p_sigctx.ps_info._code != TRAP_LWP)) {

(TRAP_EXEC is missing)

christos