Re: [Qemu-devel] [PATCH v2] e600 core for MPC86xx processors

[Alexander Graf]

Am 24.06.2013 um 08:13 schrieb Julio Guerra <address@hidden>:

> 2013/6/24 Alexey Kardashevskiy <address@hidden>:
>> On 06/21/2013 10:03 PM, Alexander Graf wrote:
>>> 
>>> On 21.06.2013, at 14:01, Julio Guerra wrote:
>>> 
>>>> 2013/6/21 Alexander Graf <address@hidden>:
>>>>> 
>>>>> On 26.05.2013, at 19:41, Julio Guerra wrote:
>>>>> 
>>>>>> MPC86xx processors are based on the e600 core, which is not the case
>>>>>> in qemu where it is based on the 7400 processor.
>>>>>> 
>>>>>> This patch creates the e600 core and instantiates the MPC86xx
>>>>>> processors based on it. Therefore, adding the high BATs and the SPRG
>>>>>> 4..7 registers, which are e600-specific [1].
>>>>>> 
>>>>>> This allows to define the MPC8610 processor too and my program running
>>>>>> on a real MPC8610 target is now able to run on qemu :)
>>>>>> 
>>>>>> [1] http://cache.freescale.com/files/32bit/doc/ref_manual/E600CORERM.pdf
>>>>>> 
>>>>>> Signed-off-by: Julio Guerra <address@hidden>
>>>>> 
>>>>> Thanks, applied to ppc-next.
>>>> 
>>>> I just retested this patch and I noticed the `handle_mmu_fault`
>>>> default value introduced by recent A. Farber patches on the CPU
>>>> definitions is wrong with the selected memory model. Thus qemu
>>>> complains about the MMU model.
>>>> 
>>>> The following is missing in e600 CPU definition:
>>>>> #if defined(CONFIG_SOFTMMU)
>>>>>   ppc->handle_mmu_fault = ppc_hash32_handle_mmu_fault;
>>>>> #endif
>>>> 
>>>> Can you correct it or should I resend a patch (v3 or just the previous
>>>> three lines) ?
>>> 
>>> Just send a patch on top of the current one.
>> 
>> 
>> I do not really understand what this patch actually does but it breaks my
>> pseries/power7 setup. Please, help.
>> 
>> 
>> Starting program: /home/aik/qemu-system-ppc64 -L qemu-ppc64-bios/ -trace
>> events=qemu_trace_events -net nic,model=e1000,addr=0:0:0 -net
>> user,hostfwd=tcp::5000-:22 -m 1024 -machine pseries -nographic -vga none
>> -enable-kvm -kernel guest.vmlinux.n -initrd 1.cpio
>> [Thread debugging using libthread_db enabled]
>> Using host libthread_db library "/lib64/libthread_db.so.1".
>> **
>> ERROR:/home/alexey/pcipassthru/qemu-impreza/qom/object.c:152:type_get_parent:
>> assertion failed: (type->parent_type != NULL)
>> 
>> Breakpoint 4, 0x00000080c6eefb8c in .abort () from /lib64/libc.so.6
>> Missing separate debuginfos, use: debuginfo-install SDL-1.2.15-8.fc19.ppc64
>> bluez-libs-4.101-6.fc19.ppc64 boost-system-1.53.0-6.fc19.ppc64
>> boost-thread-1.53.0-6.fc19.ppc64 brlapi-0.6.0-5.fc19.ppc64
>> ceph-devel-0.56.4-1.fc19.ppc64 ceph-libs-0.56.4-1.fc19.ppc64
>> cryptopp-5.6.2-2.fc19.ppc64 cyrus-sasl-lib-2.1.26-8.fc19.ppc64p7
>> glib2-2.36.2-1.fc19.ppc64 glibc-2.17-4.fc19.ppc64p7
>> glusterfs-3.4.0-0.5.beta2.fc19.ppc64
>> glusterfs-api-3.4.0-0.5.beta2.fc19.ppc64 gmp-5.1.1-2.fc19.ppc64
>> gnutls-3.1.11-1.fc19.ppc64 keyutils-libs-1.5.5-4.fc19.ppc64
>> krb5-libs-1.11.2-10.fc19.ppc64 leveldb-1.9.0-1.fc19.ppc64
>> libX11-1.5.99.902-1.fc19.ppc64 libXau-1.0.6-7.fc19.ppc64
>> libcom_err-1.42.7-2.fc19.ppc64 libcurl-7.29.0-6.fc19.ppc64
>> libfdt-1.3.0-6.fc19.ppc64 libgcc-4.8.0-8.fc19.ppc64
>> libgcrypt-1.5.2-1.fc19.ppc64 libgpg-error-1.11-1.fc19.ppc64
>> libidn-1.26-2.fc19.ppc64 libjpeg-turbo-1.2.90-2.fc19.ppc64
>> libselinux-2.1.13-15.fc19.ppc64 libssh2-1.4.3-4.fc19.ppc64
>> libstdc++-4.8.0-8.fc19.ppc64 libtasn1-3.3-1.fc19.ppc64
>> libuuid-2.23.1-2.fc19.ppc64 libxcb-1.9-3.fc19.ppc64
>> ncurses-libs-5.9-11.20130511.fc19.ppc64 nettle-2.6-2.fc19.ppc64
>> nspr-4.9.6-1.fc19.ppc64 nss-3.14.3-13.0.fc19.ppc64
>> nss-softokn-freebl-3.14.3-1.fc19.ppc64 nss-util-3.14.3-1.fc19.ppc64
>> openldap-2.4.35-4.fc19.ppc64 openssl-libs-1.0.1e-4.fc19.ppc64p7
>> p11-kit-0.18.3-1.fc19.ppc64 pcre-8.32-6.fc19.ppc64p7
>> pixman-0.30.0-1.fc19.ppc64 snappy-1.1.0-1.fc19.ppc64 
>> zlib-1.2.7-10.fc19.ppc64p7
>> (gdb) up
>> #1  0x00000080c742d54c in .g_assertion_message () from 
>> /lib64/libglib-2.0.so.0
>> (gdb)
>> #2  0x00000080c742d5f4 in .g_assertion_message_expr () from
>> /lib64/libglib-2.0.so.0
>> (gdb) bt
>> #0  0x00000080c6eefb8c in .abort () from /lib64/libc.so.6
>> #1  0x00000080c742d54c in .g_assertion_message () from 
>> /lib64/libglib-2.0.so.0
>> #2  0x00000080c742d5f4 in .g_assertion_message_expr () from
>> /lib64/libglib-2.0.so.0
>> #3  0x0000000010260058 in type_get_parent (type=0x10f7b260)
>>    at /home/alexey/pcipassthru/qemu-impreza/qom/object.c:152
>> #4  0x0000000010260120 in type_class_get_size (ti=0x10f7b260)
>>    at /home/alexey/pcipassthru/qemu-impreza/qom/object.c:170
>> #5  0x00000000102603d8 in type_initialize (ti=0x10f7b260)
>>    at /home/alexey/pcipassthru/qemu-impreza/qom/object.c:236
>> #6  0x000000001026175c in object_class_foreach_tramp (key=0x10f7b3e0,
>> value=0x10f7b260,
>>    opaque=0x3fffffffe5e0) at
>> /home/alexey/pcipassthru/qemu-impreza/qom/object.c:626
>> #7  0x00000080c73e0ab0 in .g_hash_table_foreach () from 
>> /lib64/libglib-2.0.so.0
>> #8  0x000000001026188c in object_class_foreach (fn=
>>    @0x10681290: 0x10261988 <object_class_get_list_tramp>,
>>    implements_type=0x105c7188 "powerpc64-cpu", include_abstract=0x0,
>> opaque=0x3fffffffe680)
>>    at /home/alexey/pcipassthru/qemu-impreza/qom/object.c:647
>> #9  0x0000000010261a3c in object_class_get_list (implements_type=0x105c7188
>> "powerpc64-cpu",
>>    include_abstract=0x0) at
>> /home/alexey/pcipassthru/qemu-impreza/qom/object.c:679
>> #10 0x0000000010459dec in ppc_cpu_class_by_pvr (pvr=0x3f0201)
>>    at /home/alexey/pcipassthru/qemu-impreza/target-ppc/translate_init.c:7908
>> #11 0x00000000103ca368 in kvm_ppc_register_host_cpu_type ()
>>    at /home/alexey/pcipassthru/qemu-impreza/target-ppc/kvm.c:1781
>> #12 0x00000000103c4fcc in kvm_arch_init (s=0x10fb8ab0)
>>    at /home/alexey/pcipassthru/qemu-impreza/target-ppc/kvm.c:110
>> #13 0x0000000010376138 in kvm_init () at
>> /home/alexey/pcipassthru/qemu-impreza/kvm-all.c:1461
>> #14 0x00000000102c1b08 in configure_accelerator ()
>>    at /home/alexey/pcipassthru/qemu-impreza/vl.c:2673
>> #15 0x00000000102c5e28 in main (argc=0x15, argv=0x3ffffffff218,
>> envp=0x3ffffffff2c8)
>>    at /home/alexey/pcipassthru/qemu-impreza/vl.c:4062
>> (gdb) up
>> #3  0x0000000010260058 in type_get_parent (type=0x10f7b260)
>>    at /home/alexey/pcipassthru/qemu-impreza/qom/object.c:152
>> 152             g_assert(type->parent_type != NULL);
>> (gdb) p *type
>> $5 = {name = 0x10f7b3e0 "MPC8641-powerpc64-cpu", class_size = 0x0,
>> instance_size = 0x0,
>>  class_init = @0x1069ac10: 0x103a287c
>> <CPU_POWERPC_e600_POWERPC_SVR_8641_e600_cpu_class_init>,
>>  class_base_init = 0x0, class_finalize = 0x0, class_data = 0x0,
>> instance_init = 0x0,
>>  instance_finalize = 0x0, abstract = 0x0, parent = 0x10f7b400
>> "e600-family-powerpc64-cpu",
>>  parent_type = 0x0, class = 0x0, num_interfaces = 0x0, interfaces = {{
>>      typename = 0x0} <repeats 32 times>}}
>> (gdb)
> 
> As Alexander said, this patch is broken and incomplete, it has not
> been mainstreamed. A v3 will be re-submitted today, I'll let you know
> so that you can test your setup again.
> 
> But to answer your question, MPC8641 board is - with this patch -
> instantiated with an e600 CPU whose TLBs are hardware or software
> reloadable. I don't know how it is handled in kvm mode, maybe it is
> due to the missing `mmu_fault_handler` that will be present in v3.

No, this is QOM. With thr broken patch we add a CPU of class type "e600", but 
no class "e600".


Alex

> 
> --
> Julio Guerra