Hi,
I'm trying to run a "hello world" x86_64 executable in syscall
emulation mode. This works fine with the TimingSimpleCPU, but fails
when I try to use O3.
I don't think this is a problem with my configuration script, as the
same thing happens to me when using se.py from the examples directory
as well as adapting 'o3-timing.py' from the regression tests.
Comparing the trace (see below) to the assembly, it looks as though M5
is somehow decoding garbage instructions, which is causing the fault.
This appears to be x86-specific, as a "hello world" gets well past
that point in MIPS.
I see from the archives that people have got O3 working with x86, so
it should be workable. Any help would be appreciated!
This is what I get with O3, using --trace-flags=Exec,MemoryAccess:
info: Entering event queue @ 0. Starting simulation...
3000: system.physmem: IFetch of size 64 on address 0x300
41500: system.physmem: IFetch of size 64 on address 0x340
39000: system.cpu T0 : @_start. 0 : XOR_R_R : xor ebp, ebp, ebp
: IntAlu : D=0x0000000000000044
39000: system.cpu T0 : @_start+8.0 : LOOP_I : rdip t1d,
%ctrl153, : IntAlu : D=0x0000000000400328
39000: system.cpu T0 : @_start+8.1 : LOOP_I : subi rcx, rcx,
0x1 : IntAlu : D=0x0000000000000044
39000: system.cpu T0 : @_start+8.2 : LOOP_I : wripi , t1d, 0x48
: IntAlu :
81000: system.physmem: IFetch of size 64 on address 0x380
78500: system.cpu T0 : @__do_global_dtors_aux.0 : PUSH_R : stupd
rbp, SS:[rsp + 0xfffffffffffffff8] : MemWrite : D=0x0000000000000000
A=0x7fffffffee68
85000: system.physmem: Read of size 64 on address 0xa6e40
121000: system.physmem: IFetch of size 64 on address 0x3c0
122000: system.physmem: Read of size 64 on address 0xa2cc0
panic: Tried to read unmapped address 0x2a19403d.
@ cycle 122000
This is (what I believe to be) the relevant disassembly of _start:
0000000000400320 <_start>:
400320: 31 ed xor %ebp,%ebp
400322: 49 89 d1 mov %rdx,%r9
400325: 5e pop %rsi
400326: 48 89 e2 mov %rsp,%rdx
400329: 48 83 e4 f0 and $0xfffffffffffffff0,%rsp
...so there's not even an instruction beginning at _start+8.
For reference, this is the beginning of the trace I get with the
TimingSimpleCPU model, which runs fine:
info: Entering event queue @ 0. Starting simulation...
0: system.physmem: IFetch of size 8 on address 0x320 data
0x89485ed18949ed31
30000: system.cpu T0 : @_start. 0 : XOR_R_R : xor ebp, ebp, ebp
: IntAlu : D=0x0000000000000044
30000: system.physmem: IFetch of size 8 on address 0x320 data
0x89485ed18949ed31
60000: system.cpu T0 : @_start+2.0 : MOV_R_R : mov r9, r9, rdx
: IntAlu : D=0x0000000000000000
60000: system.physmem: IFetch of size 8 on address 0x320 data
0x89485ed18949ed31
90000: system.physmem: Read of size 8 on address 0xa6e70 data 0x1
90000: system.cpu T0 : @_start+5.0 : POP_R : ld t1, SS:[rsp] :
MemRead : D=0x0000000000000001 A=0x7fffffffee70
120000: system.cpu T0 : @_start+5.1 : POP_R : addi rsp, rsp, 0x8
: IntAlu : D=0x00007fffffffee78
120000: system.cpu T0 : @_start+5.2 : POP_R : mov rsi, rsi, t1 :
IntAlu : D=0x0000000000000001
120000: system.physmem: IFetch of size 8 on address 0x320 data
0x89485ed18949ed31
150000: system.physmem: IFetch of size 8 on address 0x328 data
0x495450f0e48348e2
180000: system.cpu T0 : @_start+6.0 : MOV_R_R : mov rdx, rdx,
rsp : IntAlu : D=0x00007fffffffee78
180000: system.physmem: IFetch of size 8 on address 0x328 data
0x495450f0e48348e2
210000: system.cpu T0 : @_start+9.0 : AND_R_I : limm t1,
0xfffffffffffffff0 : IntAlu : D=0xfffffffffffffff0
210000: system.cpu T0 : @_start+9.1 : AND_R_I : and rsp, rsp, t1
: IntAlu : D=0x0000000000000000
210000: system.physmem: IFetch of size 8 on address 0x328 data
0x495450f0e48348e2
240000: system.physmem: Write of size 8 on address 0xa6e68 data 0x0
[...]
Meredydd
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
