Re: [Qemu-devel] [PATCH] fw_cfg_mem: add read memory region callback

[Eric Blake <eblake@redhat.com>]

On 9/12/18 7:33 AM, Li Qiang wrote:
>>>>
>>>> Do you have a reproducer and/or a backtrace?
>>>> memory_region_dispatch_write() checks if ops->write != NULL.
>>>>

>> FWIW, looking at the codebase, there's a good number of static
>> MemoryRegionOps structures for which the "read_with_attrs" and "read"
>> members are default-initialized to NULL. It seems unlikely they are all
>> wrong.
>>
>>
> I uses the debugexit.
> 
> QEMU command: gdb --args qemu-system-x86_64  -m 2048 -hda
> /home/liqiang02/ubuntu1801.img -enable-kvm  -vnc :100 -device isa-debug-exit
> 
> guest: inw(0x501)
> 
> We can get the following backtrack.
> 
> Starting program: /usr/local/bin/qemu-system-x86_64 -m 2048 -hda
> /home/liqiang02/ubuntu1801.img -enable-kvm -vnc :100 -device isa-debug-exit

> Thread 4 "qemu-system-x86" received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0x7fffcca3c700 (LWP 52826)]
> 0x0000000000000000 in ?? ()
> (gdb) bt
> #0  0x0000000000000000 in ?? ()
> #1  0x00005555557c1f81 in memory_region_oldmmio_read_accessor
> (mr=0x5555577b32b0, addr=0, value=0x7fffcca39568, size=2, shift=0,
> mask=65535, attrs=...) at /home/liqiang02/qemu_max_cpu/qemu-2.8/memory.c:409

Doesn't that mean we should fix memory_region_oldmmio_read_accessor() to 
deal with a NULL callback, rather than hacking up a large number of 
devices to supply a no-op callback?

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org