[Bug backtrace/16558] New: GDB Aarch64 signal frame unwinder issue

[Bug backtrace/16558] New: GDB Aarch64 signal frame unwinder issue

[teawater at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16558

            Bug ID: 16558
           Summary: GDB Aarch64 signal frame unwinder issue
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: backtrace
          Assignee: unassigned at sourceware dot org
          Reporter: teawater at gmail dot com

The signal frame unwinder in GDB on Aarch64 seems to have troubles, e.g.
the frame where the signal was triggered is skipped.  Seen in:

(gdb) PASS: gdb.base/relativedebug.exp: continue
bt
#0  0x0000007fb7e1d5dc in __GI_raise (sig=sig@entry=6) at
../nptl/sysdeps/unix/sysv/linux/raise.c:56
#1  0x0000007fb7e20aac in __GI_abort () at abort.c:89
#2  0x0000000000400850 in handler (signo=14) at
/scratch/macro/aarch64-linux-wrs/src/gdb-7.6-yocto/gdb/testsuite/gdb.base/relativedebug.c:25
#3  <signal handler called>
#4  main () at
/scratch/macro/aarch64-linux-wrs/src/gdb-7.6-yocto/gdb/testsuite/gdb.base/relativedebug.c:33
FAIL: gdb.base/relativedebug.exp: pause found in backtrace

where pause's frame is missing between #3 and #4 and also in other test
cases where also other frame data is incorrect.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug backtrace/16558] GDB Aarch64 signal frame unwinder issue

[teawater at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16558

teawater <teawater at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at sourceware dot org   |teawater at gmail dot com

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug backtrace/16558] GDB Aarch64 signal frame unwinder issue

[teawater at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16558

teawater <teawater at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #1 from teawater <teawater at gmail dot com> ---
https://sourceware.org/ml/gdb-patches/2014-02/msg00374.html

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug backtrace/16558] GDB Aarch64 signal frame unwinder issue

[teawater at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16558

--- Comment #2 from teawater <teawater at gmail dot com> ---
https://sourceware.org/ml/gdb-patches/2014-03/msg00102.html

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug backtrace/16558] GDB Aarch64 signal frame unwinder issue

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16558

--- Comment #3 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "gdb and binutils".

The branch, master has been updated
       via  f2205de0080d999c9b67872c9db471c31b53e378 (commit)
      from  cdf2a8b762199f2fe7b2bd0230bb755296867c69 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=f2205de0080d999c9b67872c9db471c31b53e378

commit f2205de0080d999c9b67872c9db471c31b53e378
Author: Hui Zhu <teawater@gmail.com>
Date:   Tue May 20 13:19:06 2014 +0800

    Fix issue #15778: GDB Aarch64 signal frame unwinder issue

    The root cause of this issue is unwinder of "#3  <signal handler called>"
    doesn't supply right values of registers.
    When GDB want to get the previous frame of "#3  <signal handler called>",
    it will call cache init function of unwinder "aarch64_linux_sigframe_init".
    The address or the value of the registers is get from this function.
    So the bug is inside thie function.

    I check the asm code of "#3  <signal handler called>":
    (gdb) frame 3
    (gdb) p $pc
    $1 = (void (*)()) 0x7f931fa4d0
    (gdb) disassemble $pc, +10
    Dump of assembler code from 0x7f931fa4d0 to 0x7f931fa4da:
    => 0x0000007f931fa4d0:    mov    x8, #0x8b                      // #139
       0x0000007f931fa4d4:    svc    #0x0
       0x0000007f931fa4d8:    nop

    This is the syscall sys_rt_sigreturn, Linux kernel function
"restore_sigframe"
    will set the frame:
        for (i = 0; i < 31; i++)
            __get_user_error(regs->regs[i], &sf->uc.uc_mcontext.regs[i],
                     err);
        __get_user_error(regs->sp, &sf->uc.uc_mcontext.sp, err);
        __get_user_error(regs->pc, &sf->uc.uc_mcontext.pc, err);
    The struct of uc_mcontext is:
    struct sigcontext {
        __u64 fault_address;
        /* AArch64 registers */
        __u64 regs[31];
        __u64 sp;
        __u64 pc;
        __u64 pstate;
        /* 4K reserved for FP/SIMD state and future expansion */
        __u8 __reserved[4096] __attribute__((__aligned__(16)));
    };

    But in GDB function "aarch64_linux_sigframe_init", the code the get address
    of registers is:
      for (i = 0; i < 31; i++)
        {
          trad_frame_set_reg_addr (this_cache,
                       AARCH64_X0_REGNUM + i,
                       sigcontext_addr + AARCH64_SIGCONTEXT_XO_OFFSET
                       + i * AARCH64_SIGCONTEXT_REG_SIZE);
        }

      trad_frame_set_reg_addr (this_cache, AARCH64_FP_REGNUM, fp);
      trad_frame_set_reg_addr (this_cache, AARCH64_LR_REGNUM, fp + 8);
      trad_frame_set_reg_addr (this_cache, AARCH64_PC_REGNUM, fp + 8);

    The code that get pc and sp is not right, so I change the code according
    to Linux kernel code:
      trad_frame_set_reg_addr (this_cache, AARCH64_SP_REGNUM,
                   sigcontext_addr + AARCH64_SIGCONTEXT_XO_OFFSET
                     + 31 * AARCH64_SIGCONTEXT_REG_SIZE);
      trad_frame_set_reg_addr (this_cache, AARCH64_PC_REGNUM,
                   sigcontext_addr + AARCH64_SIGCONTEXT_XO_OFFSET
                     + 32 * AARCH64_SIGCONTEXT_REG_SIZE);

    The issue was fixed by this change, and I did the regression test.  It
    also fixed a lot of other XFAIL and FAIL.

    2014-05-20  Hui Zhu  <hui@codesourcery.com>
            Yao Qi  <yao@codesourcery.com>

        PR backtrace/16558
        * aarch64-linux-tdep.c (aarch64_linux_sigframe_init): Update comments
        and change address of sp and pc.

-----------------------------------------------------------------------

Summary of changes:
 gdb/ChangeLog            |    7 +++++
 gdb/aarch64-linux-tdep.c |   63 ++++++++++++++++++++++++---------------------
 2 files changed, 41 insertions(+), 29 deletions(-)

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug backtrace/16558] GDB Aarch64 signal frame unwinder issue

[teawater at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16558

teawater <teawater at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #4 from teawater <teawater at gmail dot com> ---
Fixed by patch.

-- 
You are receiving this mail because:
You are on the CC list for the bug.