diff mbox

: Fix PR19061, gdb hangs/spins-on-cpu when debugging any program on Alpha

Message ID CAFULd4aFah-xkSoCQNQOHA5i1R21D24Qmkr0iJ4Rf64R4rYd_w@mail.gmail.com
State New
Headers show

Commit Message

Uros Bizjak May 31, 2018, 9:23 a.m. UTC
On Wed, May 30, 2018 at 4:53 PM, Simon Marchi <simon.marchi@ericsson.com> wrote:
> On 2017-12-15 07:11 AM, Uros Bizjak wrote:
>> Hello!
>>
>> Attached patch fixes PR19061, where gdb hangs/spins-on-cpu when
>> debugging any program on Alpha. The patch is effectively a forward
>> port of Richard's patch from the Comment #5 of the PR [1].
>>
>>
>> 2017-12-15  Uros Bizjak  <ubizjak@gmail.com>
>>         Richard Henderson  <rth@redhat.com>
>>
>>     PR gdb/19061
>>     * alpha-tdep.c (alpha_software_single_step): Call
>>     alpha_deal_with_atomic_sequence here.
>>     (set_gdbarch_software_single_step): Set to alpha_software_single_step.
>>     * nat/linux-ptrace.h [__alpha__]: Define GDB_ARCH_IS_TRAP_BRKPT
>>     and GDB_ARCH_IS_TRAP_HWBKPT.
>>
>> Patch was tested on alphaev68-linux-gnu, also tested with gcc's
>> testsuite, where it fixed all hangs in guality.exp and
>> simulate-thread.exp testcases.
>>
>> Please note that I have no commit access, so if approved, please
>> commit the patch to the source repository for me. I also have
>> functionally equivalent patch for gdb-8 branch which I plan to submit
>> later.
>>
>> [1] https://sourceware.org/bugzilla/show_bug.cgi?id=19061#c5
>>
>> Uros.
>>
>
> Hi Uros and Richard,
>
> I would need your input.  Using this cross-compiler:
>
> alphaev67-unknown-linux-gnu-gcc (crosstool-NG crosstool-ng-1.22.0-677-ga3dd55b9) 6.3.0
> Copyright (C) 2016 Free Software Foundation, Inc.
> This is free software; see the source for copying conditions.  There is NO
> warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
>
> I get this error:
>
>   CXX    linux-nat.o
> /home/simark/src/binutils-gdb/gdb/linux-nat.c: In function 'void save_stop_reason(lwp_info*)':
> /home/simark/src/binutils-gdb/gdb/linux-nat.c:2718:9: error: duplicated 'if' condition [-Werror=duplicated-cond]
>     else if (GDB_ARCH_IS_TRAP_HWBKPT (siginfo.si_code))
>          ^~
> In file included from /home/simark/src/binutils-gdb/gdb/linux-nat.c:31:0:
> /home/simark/src/binutils-gdb/gdb/nat/linux-ptrace.h:173:41: note: previously used here
>  # define GDB_ARCH_IS_TRAP_BRKPT(X) ((X) == TRAP_BRKPT)
>                                     ~~~~~^~~~~~~~~~~~~~
> /home/simark/src/binutils-gdb/gdb/linux-nat.c:2709:13: note: in expansion of macro 'GDB_ARCH_IS_TRAP_BRKPT'
>     else if (GDB_ARCH_IS_TRAP_BRKPT (siginfo.si_code))
>              ^~~~~~~~~~~~~~~~~~~~~~
>
> Does Alpha even have hardware breakpoints?  If not, I would suggest
> defining GDB_ARCH_IS_TRAP_HWBKPT to 0 for __alpha__.  It would get
> rid of the error, and be more exact (no si_code can mean "hardware
> breakpoint" on alpha).

I didn't find any mentions of hardware breakpoint support in Alpha
Architecture Handbook v.4.

Please note that save_stop_reason from linux-nat.c has some code to
detect ambigous si_code, where si_code is both TRAP_BRKPT and
TRAP_HWBKPT, and returns TARGET_STOPPED_BY_SW_BREAKPOINT in this case.
So, if we don't define anything for __alpha__ in nat/linux-ptrace.h,
we get the same result.

I have tested the attached patch, and resulting gdb works OK for me.

gdb/ChangeLog:

2018-05-31  Uros Bizjak  <ubizjak@gmail.com>

        * nat/linux-ptrace.h  [__alpha__]: Remove definitions.

--cut here--
--cut here--

Uros.

Comments

Simon Marchi May 31, 2018, 2:36 p.m. UTC | #1
On 2018-05-31 05:23, Uros Bizjak wrote:
> On Wed, May 30, 2018 at 4:53 PM, Simon Marchi 
> <simon.marchi@ericsson.com> wrote:
>> Does Alpha even have hardware breakpoints?  If not, I would suggest
>> defining GDB_ARCH_IS_TRAP_HWBKPT to 0 for __alpha__.  It would get
>> rid of the error, and be more exact (no si_code can mean "hardware
>> breakpoint" on alpha).
> 
> I didn't find any mentions of hardware breakpoint support in Alpha
> Architecture Handbook v.4.
> 
> Please note that save_stop_reason from linux-nat.c has some code to
> detect ambigous si_code, where si_code is both TRAP_BRKPT and
> TRAP_HWBKPT, and returns TARGET_STOPPED_BY_SW_BREAKPOINT in this case.
> So, if we don't define anything for __alpha__ in nat/linux-ptrace.h,
> we get the same result.

TRAP_BKPT and TRAP_HWBKPT are different, so this will never happen with 
the default definitions of 
GDB_ARCH_IS_TRAP_BRKPT/GDB_ARCH_IS_TRAP_HWBKPT.  Since the kernel for 
alpha never sets si_code to TRAP_HWBKPT, it means we'll never get in to 
the HW breakpoint branch, which is fine.

In short, the patch LGTM.

> I have tested the attached patch, and resulting gdb works OK for me.

Thanks!

Simon
diff mbox

Patch

diff --git a/gdb/nat/linux-ptrace.h b/gdb/nat/linux-ptrace.h
index dc180fbf82..98b44a82a6 100644
--- a/gdb/nat/linux-ptrace.h
+++ b/gdb/nat/linux-ptrace.h
@@ -156,8 +156,6 @@  struct buffer;
    Beginning with Linux 4.6, the MIPS port reports proper TRAP_BRKPT and
    TRAP_HWBKPT codes, so we also match them.

-   The Alpha kernel uses TRAP_BRKPT for all traps.
-
    The generic Linux target code should use GDB_ARCH_IS_TRAP_* instead
    of TRAP_* to abstract out these peculiarities.  */
 #if defined __i386__ || defined __x86_64__
@@ -169,9 +167,6 @@  struct buffer;
 #elif defined __mips__
 # define GDB_ARCH_IS_TRAP_BRKPT(X) ((X) == SI_KERNEL || (X) == TRAP_BRKPT)
 # define GDB_ARCH_IS_TRAP_HWBKPT(X) ((X) == SI_KERNEL || (X) == TRAP_HWBKPT)
-#elif defined __alpha__
-# define GDB_ARCH_IS_TRAP_BRKPT(X) ((X) == TRAP_BRKPT)
-# define GDB_ARCH_IS_TRAP_HWBKPT(X) ((X) == TRAP_BRKPT)
 #else
 # define GDB_ARCH_IS_TRAP_BRKPT(X) ((X) == TRAP_BRKPT)
 # define GDB_ARCH_IS_TRAP_HWBKPT(X) ((X) == TRAP_HWBKPT)