Re: [qemu-s390x] [PATCH for-2.11] pc-bios/s390-ccw: Fix problem with invalid virtio-scsi LUN when rebooting

[Christian Borntraeger]

On 11/20/2017 10:00 AM, Cornelia Huck wrote:
> On Mon, 20 Nov 2017 09:48:36 +0100
> Christian Borntraeger <address@hidden> wrote:
> 
>> Thomas,
>>
>> does this patch help as well?
>>
>> diff --git a/pc-bios/s390-ccw/Makefile b/pc-bios/s390-ccw/Makefile
>> index 6d0c2ee..2687590 100644
>> --- a/pc-bios/s390-ccw/Makefile
>> +++ b/pc-bios/s390-ccw/Makefile
>> @@ -12,7 +12,7 @@ $(call set-vpath, $(SRC_PATH)/pc-bios/s390-ccw)
>>  OBJECTS = start.o main.o bootmap.o sclp.o virtio.o virtio-scsi.o 
>> virtio-blkdev.o
>>  QEMU_CFLAGS := $(filter -W%, $(QEMU_CFLAGS))
>>  QEMU_CFLAGS += -ffreestanding -fno-delete-null-pointer-checks -msoft-float
>> -QEMU_CFLAGS += -march=z900 -fPIE -fno-strict-aliasing
>> +QEMU_CFLAGS += -march=z900 -fPIE -fno-strict-aliasing 
>> -fno-zero-initialized-in-bss
>>  QEMU_CFLAGS += $(call cc-option, $(QEMU_CFLAGS), -fno-stack-protector)
>>  LDFLAGS += -Wl,-pie -nostdlib
> 
> If this would actually enable us to kill a whole bird swarm with one
> stone, I'd prefer it over the patch below, nicely short though it is.

I will resend with proper patch description so that Thomas can comment on a 
proper
patch.


> 
> 
> I plan to pack the one or the other into s390-fixes today and rebuild.
> 
>>  
>>
>>
>>
>> On 11/17/2017 07:45 PM, Christian Borntraeger wrote:
>>>
>>>
>>> On 11/17/2017 07:10 PM, Thomas Huth wrote:  
>>>> When rebooting a guest that has a virtio-scsi disk, the s390-ccw
>>>> bios sometimes bails out with an error message like this:
>>>>
>>>> ! SCSI cannot report LUNs: STATUS=02 RSPN=70 KEY=05 CODE=25 QLFR=00, sure !
>>>>
>>>> Enabling the scsi_req* tracing in QEMU shows that the ccw bios is
>>>> trying to execute the REPORT LUNS SCSI command with a LUN != 0, and
>>>> this causes the SCSI command to fail.
>>>> Looks like we neither clear the BSS of the s390-ccw bios during reboot,
>>>> nor do we explicitly set the default_scsi_device.lun value to 0, so
>>>> this variable can contain random values from the OS after the reboot.
>>>> By setting this variable explicitly to 0, the problem is fixed and
>>>> the reboots always succeed.
>>>>
>>>> Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1514352
>>>> Signed-off-by: Thomas Huth <address@hidden>  
>>>
>>> Acked-by: Christian Borntraeger <address@hidden>
>>>
>>> We had these things multile times in the past. The QEMU elf loader does not
>>> zero the BSS (it just loads the load section).  Hmm, what about letting the 
>>> BIOS zero its bss itself. IIRC the kernel does the same thing. I will have a
>>> look into that for 2.12.
>>>
>>> PS: Please do not forget to rebuild the bios  
>>>> ---
>>>>  pc-bios/s390-ccw/virtio-scsi.c | 3 ++-
>>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/pc-bios/s390-ccw/virtio-scsi.c 
>>>> b/pc-bios/s390-ccw/virtio-scsi.c
>>>> index c92f5d3..4fe4b9d 100644
>>>> --- a/pc-bios/s390-ccw/virtio-scsi.c
>>>> +++ b/pc-bios/s390-ccw/virtio-scsi.c
>>>> @@ -223,7 +223,8 @@ static void virtio_scsi_locate_device(VDev *vdev)
>>>>
>>>>      for (target = 0; target <= vdev->config.scsi.max_target; target++) {
>>>>          sdev->channel = channel;
>>>> -        sdev->target = target; /* sdev->lun will be 0 here */
>>>> +        sdev->target = target;
>>>> +        sdev->lun = 0;          /* LUN has to be 0 for REPORT LUNS */
>>>>          if (!scsi_report_luns(vdev, data, sizeof(data))) {
>>>>              if (resp.response == VIRTIO_SCSI_S_BAD_TARGET) {
>>>>                  continue;
>>>>  
>>>
>>>   
>>
>