[PATCH] hw/ufs: avoid generating the same ID string for different LU devices

[PATCH] hw/ufs: avoid generating the same ID string for different LU devices

[Akinobu Mita]

QEMU would not start when trying to create two UFS host controllers and
a UFS logical unit for each with the following options:

-device ufs,id=bus0 \
-device ufs-lu,drive=drive1,bus=bus0,lun=0 \
-device ufs,id=bus1 \
-device ufs-lu,drive=drive2,bus=bus1,lun=0 \

This is because the same ID string ("0:0:0/scsi-disk") is generated
for both UFS logical units.

To fix this issue, prepend the parent pci device's path to make
the ID string unique.
("0000:00:03.0/0:0:0/scsi-disk" and "0000:00:04.0/0:0:0/scsi-disk")

Fixes: 096434fea13a ("hw/ufs: Modify lu.c to share codes with SCSI subsystem")
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
---
 hw/ufs/ufs.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/hw/ufs/ufs.c b/hw/ufs/ufs.c
index 68c5f1f6c9..eccdb852a0 100644
--- a/hw/ufs/ufs.c
+++ b/hw/ufs/ufs.c
@@ -1323,9 +1323,17 @@ static bool ufs_bus_check_address(BusState *qbus, DeviceState *qdev,
     return true;
 }
 
+static char *ufs_bus_get_dev_path(DeviceState *dev)
+{
+    BusState *bus = qdev_get_parent_bus(dev);
+
+    return qdev_get_dev_path(bus->parent);
+}
+
 static void ufs_bus_class_init(ObjectClass *class, void *data)
 {
     BusClass *bc = BUS_CLASS(class);
+    bc->get_dev_path = ufs_bus_get_dev_path;
     bc->check_address = ufs_bus_check_address;
 }
 
-- 
2.34.1

Re: [PATCH-for-8.2?] hw/ufs: avoid generating the same ID string for different LU devices

[Philippe Mathieu-Daudé]

On 4/12/23 16:05, Akinobu Mita wrote:
> QEMU would not start when trying to create two UFS host controllers and
> a UFS logical unit for each with the following options:
> 
> -device ufs,id=bus0 \
> -device ufs-lu,drive=drive1,bus=bus0,lun=0 \
> -device ufs,id=bus1 \
> -device ufs-lu,drive=drive2,bus=bus1,lun=0 \
> 
> This is because the same ID string ("0:0:0/scsi-disk") is generated
> for both UFS logical units.
> 
> To fix this issue, prepend the parent pci device's path to make
> the ID string unique.
> ("0000:00:03.0/0:0:0/scsi-disk" and "0000:00:04.0/0:0:0/scsi-disk")
> 
> Fixes: 096434fea13a ("hw/ufs: Modify lu.c to share codes with SCSI subsystem")
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>

> ---
>   hw/ufs/ufs.c | 8 ++++++++
>   1 file changed, 8 insertions(+)
> 
> diff --git a/hw/ufs/ufs.c b/hw/ufs/ufs.c
> index 68c5f1f6c9..eccdb852a0 100644
> --- a/hw/ufs/ufs.c
> +++ b/hw/ufs/ufs.c
> @@ -1323,9 +1323,17 @@ static bool ufs_bus_check_address(BusState *qbus, DeviceState *qdev,
>       return true;
>   }
>   
> +static char *ufs_bus_get_dev_path(DeviceState *dev)
> +{
> +    BusState *bus = qdev_get_parent_bus(dev);
> +
> +    return qdev_get_dev_path(bus->parent);
> +}
> +
>   static void ufs_bus_class_init(ObjectClass *class, void *data)
>   {
>       BusClass *bc = BUS_CLASS(class);
> +    bc->get_dev_path = ufs_bus_get_dev_path;
>       bc->check_address = ufs_bus_check_address;
>   }
>   

Re: [PATCH-for-8.2?] hw/ufs: avoid generating the same ID string for different LU devices

[Philippe Mathieu-Daudé]

Hi Jeuk,

On 4/12/23 17:50, Philippe Mathieu-Daudé wrote:
> On 4/12/23 16:05, Akinobu Mita wrote:
>> QEMU would not start when trying to create two UFS host controllers and
>> a UFS logical unit for each with the following options:
>>
>> -device ufs,id=bus0 \
>> -device ufs-lu,drive=drive1,bus=bus0,lun=0 \
>> -device ufs,id=bus1 \
>> -device ufs-lu,drive=drive2,bus=bus1,lun=0 \
>>
>> This is because the same ID string ("0:0:0/scsi-disk") is generated
>> for both UFS logical units.
>>
>> To fix this issue, prepend the parent pci device's path to make
>> the ID string unique.
>> ("0000:00:03.0/0:0:0/scsi-disk" and "0000:00:04.0/0:0:0/scsi-disk")
>>
>> Fixes: 096434fea13a ("hw/ufs: Modify lu.c to share codes with SCSI 
>> subsystem")

If you think this must be fixed for the 8.2 release, please assign
a release blocker issues to the GitLab 8.2 milestone here:
https://gitlab.com/qemu-project/qemu/-/milestones/10

>> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> 
> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> 
>> ---
>>   hw/ufs/ufs.c | 8 ++++++++
>>   1 file changed, 8 insertions(+)
>>
>> diff --git a/hw/ufs/ufs.c b/hw/ufs/ufs.c
>> index 68c5f1f6c9..eccdb852a0 100644
>> --- a/hw/ufs/ufs.c
>> +++ b/hw/ufs/ufs.c
>> @@ -1323,9 +1323,17 @@ static bool ufs_bus_check_address(BusState 
>> *qbus, DeviceState *qdev,
>>       return true;
>>   }
>> +static char *ufs_bus_get_dev_path(DeviceState *dev)
>> +{
>> +    BusState *bus = qdev_get_parent_bus(dev);
>> +
>> +    return qdev_get_dev_path(bus->parent);
>> +}
>> +
>>   static void ufs_bus_class_init(ObjectClass *class, void *data)
>>   {
>>       BusClass *bc = BUS_CLASS(class);
>> +    bc->get_dev_path = ufs_bus_get_dev_path;
>>       bc->check_address = ufs_bus_check_address;
>>   }
> 

Re: [PATCH-for-8.2?] hw/ufs: avoid generating the same ID string for different LU devices

[Jeuk Kim]

Hi Philippe,

Thank you for informing me.
I want this issue is fixed for the 8.2 release.

I created an issue like you said, but I couldn't assign the milestone (I'm guessing it's a permission problem).
https://gitlab.com/qemu-project/qemu/-/issues/2018
Could you help me with this?

Regardless, I'm going to submit a pull request to fix the issue right away.

Thank you.

On 4/12/23, Philippe Mathieu-Daudé wrote:
> Hi Jeuk,
> 
> On 4/12/23 17:50, Philippe Mathieu-Daudé wrote:
>> On 4/12/23 16:05, Akinobu Mita wrote:
>>> QEMU would not start when trying to create two UFS host controllers and
>>> a UFS logical unit for each with the following options:
>>>
>>> -device ufs,id=bus0 \
>>> -device ufs-lu,drive=drive1,bus=bus0,lun=0 \
>>> -device ufs,id=bus1 \
>>> -device ufs-lu,drive=drive2,bus=bus1,lun=0 \
>>>
>>> This is because the same ID string ("0:0:0/scsi-disk") is generated
>>> for both UFS logical units.
>>>
>>> To fix this issue, prepend the parent pci device's path to make
>>> the ID string unique.
>>> ("0000:00:03.0/0:0:0/scsi-disk" and "0000:00:04.0/0:0:0/scsi-disk")
>>>
>>> Fixes: 096434fea13a ("hw/ufs: Modify lu.c to share codes with SCSI
>>> subsystem")
>
> If you think this must be fixed for the 8.2 release, please assign
> a release blocker issues to the GitLab 8.2 milestone here:
> https://gitlab.com/qemu-project/qemu/-/milestones/10
>
>>> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
>>
>> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>>
>>> ---
>>>   hw/ufs/ufs.c 8 ++++++++
>>>   1 file changed, 8 insertions(+)
>>>
>>> diff --git a/hw/ufs/ufs.c b/hw/ufs/ufs.c
>>> index 68c5f1f6c9..eccdb852a0 100644
>>> --- a/hw/ufs/ufs.c
>>> +++ b/hw/ufs/ufs.c
>>> @@ -1323,9 +1323,17 @@ static bool ufs_bus_check_address(BusState
>>> *qbus, DeviceState *qdev,
>>>       return true;
>>>   }
>>> +static char *ufs_bus_get_dev_path(DeviceState *dev)
>>> +{
>>> +    BusState *bus = qdev_get_parent_bus(dev);
>>> +
>>> +    return qdev_get_dev_path(bus->parent);
>>> +}
>>> +
>>>   static void ufs_bus_class_init(ObjectClass *class, void *data)
>>>   {
>>>       BusClass *bc = BUS_CLASS(class);
>>> +    bc->get_dev_path = ufs_bus_get_dev_path;
>>>       bc->check_address = ufs_bus_check_address;
>>>   }

Re: [PATCH-for-8.2?] hw/ufs: avoid generating the same ID string for different LU devices

[Jeuk Kim]

On 4/12/23 16:05, Akinobu Mita wrote:
> QEMU would not start when trying to create two UFS host controllers and
> a UFS logical unit for each with the following options:
>
> -device ufs,id=bus0 \
> -device ufs-lu,drive=drive1,bus=bus0,lun=0 \
> -device ufs,id=bus1 \
> -device ufs-lu,drive=drive2,bus=bus1,lun=0 \
>
> This is because the same ID string ("0:0:0/scsi-disk") is generated
> for both UFS logical units.
>
> To fix this issue, prepend the parent pci device's path to make
> the ID string unique.
> ("0000:00:03.0/0:0:0/scsi-disk" and "0000:00:04.0/0:0:0/scsi-disk")
>
> Fixes: 096434fea13a ("hw/ufs: Modify lu.c to share codes with SCSI subsystem")
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>

Reviewed-by: Jeuk Kim <jeuk20.kim@samsung.com>

Thank you!