Re: [Qemu-devel] [PULL 7/7] tcg-i386: Use new return-argument ld/st helpers

[Stefan Weil]

Am 28.05.2014 19:12, schrieb Richard Henderson:
> On 05/27/2014 03:37 PM, Stefan Weil wrote:
>> Hi Richard,
>>
>> this patch has broken the 64 bit version of QEMU for Windows: a Linux
>> guest starts booting, but hangs after "Booting the kernel.". I got a bug
>> report from a user and did a "git bisect" with a Tiny Core Linux guest /
>> cross build with default options / cross test with wine64 and default
>> options. Git reported this commit:
>>
>> 401c227b0a1134245ec61c6c5a9997cfc963c8e4 is the first bad commit
>> commit 401c227b0a1134245ec61c6c5a9997cfc963c8e4
>> Author: Richard Henderson <address@hidden>
>> Date:   Thu Jul 25 07:16:52 2013 -1000
>>
>>     tcg-i386: Use new return-argument ld/st helpers
>>    
>>     Discontinue the jump-around-jump-to-jump scheme, trading it for a single
>>     immediate move instruction.  The two extra jumps always consume 7 bytes,
>>     whereas the immediate move is either 5 or 7 bytes depending on where the
>>     code_gen_buffer gets located.
>>    
>>     Signed-off-by: Richard Henderson <address@hidden>
>>
>> :040000 040000 dfd9a66c85713cd1886a3342de1e9ac95d7ea43f
>> df8673dea69bc89cc2cc979aa24415e3fea4ed53 M    include
>> :040000 040000 1f7cd5291f2c69b4126c63bd567c6b106eb332c9
>> 87e7ece766168dda860b513dc97fe5af28ec2c4b M    tcg
>>
>> 32 bit versions of QEMU for Windows don't show this problem.
> 
> I'm having problem booting any iso with wine at the moment:
> 
> $ wine64 ./x86_64-softmmu/qemu-system-x86_64.exe -L ./pc-bios \
>     -vnc :1 -cdrom ../../../Downloads/TinyCore-current.iso
> Assertion failed!
> 
> Program: Z:\home\rth\work\qemu\bld-w64\x86_64-softmmu\qemu-system-x86_64.exe
> File: /home/rth/work/qemu/qemu/qemu-coroutine-lock.c, Line 91
> 
> Expression: qemu_in_coroutine()
> 
> abnormal program termination
> 
> Naturally, this isn't happening with a native linux boot with the same 
> arguments.
> 
> But I can boot an alpha rom:
> 
> $ wine64 ./alpha-softmmu/qemu-system-alpha.exe -L ./pc-bios/ -nographic
> PCI: 00:00:0 class 0300 id 1013:00b8
> PCI:   region 0: 10000000
> PCI:   region 1: 12000000
> PCI: 00:01:0 class 0200 id 8086:100e
> PCI:   region 0: 12020000
> PCI:   region 1: 0000c000
> PCI: 00:02:0 class 0101 id 1095:0646
> PCI:   region 0: 0000c040
> PCI:   region 1: 0000c048
> PCI:   region 3: 0000c04c
>>>>
> 
> Which says to me that it's rather unlikely that this basic load/store patch
> could be the problem.
> 
> 
> r~
> 


The current coroutine implementation for Windows does not work (compiler
problem?) and triggers that assertion. Here is one of the workarounds
which help:

diff --git a/coroutine-win32.c b/coroutine-win32.c
index edc1f72..d4c40d3 100644
--- a/coroutine-win32.c
+++ b/coroutine-win32.c
@@ -42,6 +42,7 @@ CoroutineAction qemu_coroutine_switch(Coroutine
*from_, Coroutine *to_,
     CoroutineWin32 *from = DO_UPCAST(CoroutineWin32, base, from_);
     CoroutineWin32 *to = DO_UPCAST(CoroutineWin32, base, to_);

+    g_assert(current == from_);
     current = to_;

     to->action = action;

So we have the paradox situation that adding one more assertion helps to
avoid triggering another assertion.

The alpha ROM obviously does not trigger coroutines. Block I/O usually does.

Stefan