[Mono-dev] mono 2.2 rc2 on ARM issues

[Cliff Brake]

I have traced it down to the kernel image.  The kernel that works is
virtually identical, but likely compiled with a older toolchain.  The
only other difference is a minor defconfig change with the MTD
partition setup.

Tracing with gdb on the bad (more recent) kernel image, I get:

Program received signal SIGILL, Illegal instruction.
mono_arch_flush_icache (code=0x0, size=<value optimized out>) at mini-arm.c:595
595     }

(gdb) bt
#0  mono_arch_flush_icache (code=0x0, size=<value optimized out>)
    at mini-arm.c:595
#1  0x000d1a08 in mono_arch_create_trampoline_code_full (
    tramp_type=MONO_TRAMPOLINE_JIT, code_size=0xbeddde34, ji=0xbedddbbc, aot=0)
    at tramp-arm.c:365
#2  0x000d1d0c in mono_arch_create_trampoline_code (
    tramp_type=MONO_TRAMPOLINE_JIT) at tramp-arm.c:160
#3  0x000bce08 in mono_trampolines_init () at mini-trampolines.c:691
#4  0x0001aa2c in mini_init (filename=0xbedddf17 "test.exe",
    runtime_version=0x0) at mini.c:14310
#5  0x0009ff14 in mono_main (argc=2, argv=0xbeddde34) at driver.c:1545
#6  0x401ed9cc in __libc_start_main () from /lib/libc.so.6
#7  0x00019214 in _start ()

Looks like this could be a fun one to track down.

Cliff

On Thu, Dec 11, 2008 at 11:32 AM, Cliff Brake <cliff.brake at gmail.com> wrote:
> It appears I have some type of system problem as the 2.2 mono binaries
> run on a older rootfs, so I'm now tracking down what all has changed
> in the system.
>
> It appears that VFP is the correct option as a simple floating point
> test app works.
>
> Cliff
>
> On Wed, Dec 10, 2008 at 5:47 PM, Cliff Brake <cliff.brake at gmail.com> wrote:
>> Hello,
>>
>> I'm trying to run mono 2.2 on ARM, and mono is crashing fairly early
>> (see attached).  Note, it has not accessed the test.exe file yet.
>>
>> Setup:
>> - PXA270 CPU (armv5te)
>> - fairly recent OE toolchain (EABI,  4.2.4)
>> - hardcoded ARM_FPU_VFP=1
>>
>> Questions:
>> 1) has anyone run mono 2.2 on ARM?
>> 2) is VFP the correct FP config?  I have tried soft-float as well with
>> the same result.
>> 3) any suggestions on how to debug this or what I should look at next.
>>
>> Thanks,
>> Cliff
>>
>> --
>> =======================
>> Cliff Brake
>> http://bec-systems.com
>>
>
>
>
> --
> =======================
> Cliff Brake
> http://bec-systems.com
>



-- 
=======================
Cliff Brake
http://bec-systems.com