[Bug backtrace/18338] New: ARM: infinite stack trace with assembly function

[Bug backtrace/18338] New: ARM: infinite stack trace with assembly function

[carl at cogsystems dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18338

            Bug ID: 18338
           Summary: ARM: infinite stack trace with assembly function
           Product: gdb
           Version: unknown
            Status: NEW
          Severity: normal
          Priority: P2
         Component: backtrace
          Assignee: unassigned at sourceware dot org
          Reporter: carl at cogsystems dot com
  Target Milestone: ---

We have encountered a bug where gdb enters an infinite loop doing a
-stack-list-frames under certain conditions.

Bug initially triggered in eclipse CDT - which proceeded to hang the UI (a
separate issue), but can be reproduced in GDB directly.

Symptoms:
breakpoint or step into isr handler from assembly irq handler causes backtrace
to become infinite. This issue only occurs when symbols are loaded. Without
symbols, gdb cannot attempt a backtrace.

Steps to reproduce:
We have an irq handler in assembly which calls a C ISR handler.

Note, setting a backtrace limit, stops the infinite behaviour, but eclipse does
not do this.

The ISR handler is annotated as a function, but does not look like a C
function. I looked through the function prologue interpretting code a bit but
didn't see anything obvious.

I show two versions of the irq handler, the first causes gdb to enter an
infinite loop stuck on the same stack frame, the second works and gdb shows the
stack frames, including pre-empted functions- which seems a bit odd since it
does not know about the push of 8 bytes in the srsdb instruction. The issue
seems to be the first three instructions somehow confusing gdb's backtrace
analysis.

irq handler exposing bug:

FUNC(__irq)
    sub     lr, lr, 4
    srsdb   sp!, 0x1f
    cps     0x1f
    push.w  {r0-r5, ip, lr}

    //r4 = get_irq()

    /* Jump to C handler */
    ldr     ip, =_irq_handlers
    add.w   ip, ip, r4, lsl 3
    ldr     ip, [ip]
    cmp     ip, 0
    it      ne
    blxne   ip

    pop.w   {r0-r5, ip, lr}
    rfeia   sp!
END_FUNC()

modified version:

__irq_start:
    sub     lr, lr, 4
    srsdb   sp!, 0x1f
    cps     0x1f
FUNC(__irq)
    push.w  {r0-r5, ip, lr}

    //r4 = get_irq()

    /* Jump to C handler */
    ldr     ip, =_irq_callbacks
    add.w   ip, ip, r4, lsl 3
    ldr     ip, r1, [ip]
    cmp     ip, 0
    it      ne
    blxne   ip

    pop.w   {r0-r5, ip, lr}
    rfeia   sp!
END_FUNC()

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

[Bug backtrace/18338] ARM: infinite stack trace with assembly function

[carl at cogsystems dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18338

Carl van Schaik <carl at cogsystems dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://sourceware.org/bugz
                   |                            |illa/show_bug.cgi?id=8872

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