Am 19.07.2013 04:26, schrieb Mike Qiu:
> 于 2013/7/19 1:14, Andreas Färber 写道:
>> There's some typos in the commit message, but the change looks okay to
>> me - although there were discussions to catch this on the memory API
>> side of things instead.
> You mean this patch: see below:
>
> exec: Support 64-bit operations in address_s
No, I don't. There were other segfault avoidance patches like yours over
the past months - they're all fixing individual segfault symptoms.
Question for Paolo is whether we want to continue to discover them one
by one or whether to implement a fallback inside memory code if .read or
.write is NULL.
Andreas
>
> BTW, this bug has been opened before?
>
> Thanks
> Mike
>>
>> Regards,
>> Andreas
>>
>>> Thanks
>>> Mike
>>> 2013/7/16 11:50, Mike Qiu wrote:
>>>> For usb-ehci in qemu, its caps just has read() operation,
>>>> the write() operation does not exist.
>>>>
>>>> This cause a Segmentation fault when use usb-ehci device in ppc64
>>>> platform.
>>>>
>>>> here is gdb output:
>>>>
>>>> Program received signal SIGSEGV, Segmentation fault.
>>>> [Switching to Thread 0x3fffa7fcef20 (LWP 6793)]
>>>> 0x00000000103f5244 in memory_region_oldmmio_write_accessor
>>>> (opaque=0x113e9e78, addr=9, value=0x3fffa7fce088,
>>>> size=1, shift=0, mask=255) at /home/Mike/qemu-impreza/memory.c:384
>>>> 384 mr->ops->old_mmio.write[ctz32(size)](mr->opaque, addr, tmp);
>>>> (gdb) p *mr->ops
>>>> $1 = {read = @0x10716f68: 0x1020699c <ehci_caps_read>, write = 0,
>>>> endianness = DEVICE_LITTLE_ENDIAN, valid = {min_access_size = 1,
>>>> max_access_size = 4, unaligned = false, accepts = 0}, impl =
>>>> {min_access_size = 1, max_access_size = 1, unaligned = false},
>>>> old_mmio = {read = {0, 0, 0}, write = {0, 0, 0}}}
>>>>
>>>> Becasue function write() of mr->ops has not been implement, in
>>>> function memory_region_dispatch_write(), it call
>>>> oldmmio write accessor, but at the same time old_mmio still not
>>>> been implement by default.
>>>>
>>>> That is the root cause of the Segmentation fault.
>>>>
>>>> To solve this problem, add empty function: ehci_caps_write()
>>>>
>>>> Signed-off-by: Mike Qiu <qiud...@linux.vnet.ibm.com>
>>>> ---
>>>> hw/usb/hcd-ehci.c | 7 +++++++
>>>> 1 file changed, 7 insertions(+)
>>>>
>>>> diff --git a/hw/usb/hcd-ehci.c b/hw/usb/hcd-ehci.c
>>>> index 67e4b24..6c8a439 100644
>>>> --- a/hw/usb/hcd-ehci.c
>>>> +++ b/hw/usb/hcd-ehci.c
>>>> @@ -1072,6 +1072,12 @@ static void ehci_port_write(void *ptr, hwaddr addr,
>>>> trace_usb_ehci_portsc_change(addr + s->portscbase, addr >> 2,
>>>> *portsc, old);
>>>> }
>>>>
>>>> +static void ehci_caps_write(void *ptr, hwaddr addr, uint64_t val,
>>>> + unsigned size)
>>>> +{
>>>> + /* nothing */
>>>> +}
>>>> +
>>>> static void ehci_opreg_write(void *ptr, hwaddr addr,
>>>> uint64_t val, unsigned size)
>>>> {
>>>> @@ -2380,6 +2386,7 @@ static void ehci_frame_timer(void *opaque)
>>>>
>>>> static const MemoryRegionOps ehci_mmio_caps_ops = {
>>>> .read = ehci_caps_read,
>>>> + .write = ehci_caps_write,
>>>> .valid.min_access_size = 1,
>>>> .valid.max_access_size = 4,
>>>> .impl.min_access_size = 1,
>>>
>>
>
--
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg
