qemu-ppc
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-ppc] [Qemu-devel] [PATCH RFC 0/4] Enable MTTCG on PowerPC


From: Nikunj A Dadhania
Subject: Re: [Qemu-ppc] [Qemu-devel] [PATCH RFC 0/4] Enable MTTCG on PowerPC
Date: Sat, 03 Sep 2016 21:59:16 +0530
User-agent: Notmuch/0.21 (https://notmuchmail.org) Emacs/25.0.94.1 (x86_64-redhat-linux-gnu)

Programmingkid <address@hidden> writes:

>> On 09/02/2016 08:32 AM, Nikunj A Dadhania wrote:
>>> The series is a first attempt at enabling Multi-Threaded TCG on PowerPC.
>>> Changes that were needed to enable PowerPC are pretty simple;
>>> 
>>> Patch 01: Take a iothread lock during hcall, as hcall can generate io 
>>> requests
>>>      02: For TCG, we were harcoding smt as 1, this gets rid of the 
>>> limitation
>>>      03: Use atomic_cmpxchg in store conditional
>>>      04: With more threads, flush the entry from each cpu. 
>>>       This can be optimized further.
>>> 
>>> The patches are based on the Alex Bennee's base enabling patches for 
>>> MTTCG[1] and Emilios's cmpxchg atomics. The consolidated tree of the 
>>> above patches is here:
>>> 
>>> https://github.com/stsquad/qemu/tree/mttcg/base-patches-v4-with-cmpxchg-atomics-v2
>>> 
>>> Apart from the above, PPC patches are based out of ppc-for-2.8 and 
>>> load/store consolidation patches [2]
>>> 
>>> Series with all dependent patches available here: 
>>> https://github.com/nikunjad/qemu/tree/ppc_mttcg_v1
>>> 
>>> Testing: 
>>> ========
>>> 
>>> -smp 4,cores=1,threads=4 -accel tcg,thread=multi
>>> 
>>> TODO
>>> ====
>>> Implement msgsndp instructions(door-bell), newer kernels enable it 
>>> depending on the PVR. I have been using following workaround to boot.
>>> https://github.com/nikunjad/qemu/commit/2c10052c5f93418a6b920e6ba3ce1813fcf50bc4
>> 
>> You could also introduce a Power8 DD1 in qemu. From the kernel cputable :
>> 
>>      {       /* Power8 DD1: Does not support doorbell IPIs */
>>              .pvr_mask               = 0xffffff00,
>>              .pvr_value              = 0x004d0100,
>>              .cpu_name               = "POWER8 (raw)",
>>              .cpu_features           = CPU_FTRS_POWER8_DD1,
>>              ...
>> 
>> Cheers,
>> C.
>
> What I did:
>
> git clone https://github.com/nikunjad/qemu/tree/ppc_mttcg_v1
>
> git checkout ppc_mttcg_v1
>
> ./configure --target-list=ppc-softmmu,i386-softmmu --cxx=gcc-4.9 --cc=gcc-4.9 
> --objcc=gcc-4.9 --disable-gtk --disable-sdl && make
>
> After these steps I tried to test Windows XP and Windows 2000 using this 
> command line:
> ./i386-softmmu/qemu-system-i386 -name 'Windows XP' -hda 'Windows XP Hard 
> Drive.img' -boot c -smp 4,cores=1,threads=4 -accel tcg,thread=multi

I haven't enabled x86. Not sure if that is supported in Alex Bennee's
tree either.

>
> This causes QEMU to crash. The cause is an abort trap. This is what I 
> captured:
>
> Thread 7 Crashed:
> 0   libSystem.B.dylib                 0x00007fff825559ce 
> __semwait_signal_nocancel + 10
> 1   libSystem.B.dylib                 0x00007fff825558d0 nanosleep$NOCANCEL + 
> 129
> 2   libSystem.B.dylib                 0x00007fff825b23ce usleep$NOCANCEL + 57
> 3   libSystem.B.dylib                 0x00007fff825d1a00 abort + 93
> 4   qemu-system-i386                  0x0000000100036c65 
> qemu_tcg_cpu_thread_fn + 325 (cpus.c:1384)
> 5   libSystem.B.dylib                 0x00007fff8251bfd6 _pthread_start + 331
> 6   libSystem.B.dylib                 0x00007fff8251be89 thread_start + 13
>
> Trying Windows 2000 also causes the same crash.
>
> I then tried Mac OS 10.4. The command line was:
> ./ppc-softmmu/qemu-system-ppc -smp 4,cores=1,threads=4 -M mac99 -m 512 -hda 
> 'Mac OS 10.4.0.qcow2' -boot c -prom-env boot-args=-v -localtime

I have tested with ppc64-softmmu. Haven't tried ppc-softmmu yet.

>
> The result was this error message:
> qemu-system-ppc: Number of SMP CPUs requested (4) exceeds max CPUs
> supported by machine 'mac99' (1)

This seems to be similar error which I got when enabling ppc64. 02/04 of
my patch has got over the limitation for ppc64. I guess some patch seems
to be missing for ppc.

I thought that something like this should work:

========================================================================

diff --git a/hw/ppc/mac.h b/hw/ppc/mac.h
index 20cbddb..e6522e5 100644
--- a/hw/ppc/mac.h
+++ b/hw/ppc/mac.h
@@ -32,7 +32,7 @@
 #include "hw/input/adb.h"
 
 /* SMP is not enabled, for now */
-#define MAX_CPUS 1
+#define MAX_CPUS 4
 
 #define BIOS_SIZE     (1024 * 1024)
 #define NVRAM_SIZE        0x2000
diff --git a/hw/ppc/mac_newworld.c b/hw/ppc/mac_newworld.c
index 7d25106..13e7d1c 100644
--- a/hw/ppc/mac_newworld.c
+++ b/hw/ppc/mac_newworld.c
@@ -515,6 +515,7 @@ static void core99_machine_class_init(ObjectClass *oc, void 
*data)
 {
     MachineClass *mc = MACHINE_CLASS(oc);
 
+    fprintf(stderr, "%s: init %d\n", __func__, MAX_CPUS);
     mc->desc = "Mac99 based PowerMAC";
     mc->init = ppc_core99_init;
     mc->max_cpus = MAX_CPUS;

========================================================================


But I do not see this print at all. core99_machine_class_init() doesnt
get called ? Am I missing something?

Regards
Nikunj




reply via email to

[Prev in Thread] Current Thread [Next in Thread]