Re: [PULL 8/8] migration: Add qtest for migration over RDMA

[Fabiano Rosas <farosas@suse.de>]

"Zhijian Li (Fujitsu)" via <qemu-devel@nongnu.org> writes:

> Hi Philippe,
>
> Thanks for your testing.
>
>
> On 08/03/2025 14:00, Philippe Mathieu-Daudé wrote:
>> Hi,
>> 
>> On 7/3/25 19:15, Fabiano Rosas wrote:
>>> From: Li Zhijian <lizhijian@fujitsu.com>
>>>
>>> This qtest requires there is a RDMA(RoCE) link in the host.
>>> In order to make the test work smoothly, introduce a
>>> scripts/rdma-migration-helper.sh to
>>> - setup a new Soft-RoCE(aka RXE) if it's root
>>> - detect existing RoCE link
>>>
>>> Test will be skipped if there is no available RoCE link.
>> 
>> Is it? Runing as user I'm getting:
>> 
>>    RDMA ERROR: RDMA host is not set!
>
>
> It's unexpected behavior.
>
> It implies that the script terminated successfully with an exit
> code of 0(there is a RDMA link), yet failed to display its canonical
> IPv4 address.
>
> Stefan also mentioned the same error...
> https://gitlab.com/qemu-project/qemu/-/jobs/9350004599#L5590
>
> I couldn't reproduce your error.
>
> Could you share the output of this script with a normal user,
> $ scripts/rdma-migration-helper.sh detect
>
> if your had a rdma/RXE link, please share the output of its ip
> $ ip -4 -o addr show dev <NIC>
>
> Where the <NIC> is an interface associated with the RoCE(RXE), for example
>
> $ rdma link
> link enp2s0_rxe/1 state ACTIVE physical_state LINK_UP netdev enp2s0
>
> then the <NIC> is enp2s0
>
>
>> 
>> Apparently called via:
>> 
>> qemu_start_incoming_migration()
>>    -> rdma_start_incoming_migration()
>>       -> qemu_rdma_dest_init()
>> 
>>>   # Start of rdma tests
>>>   # Running /x86_64/migration/precopy/rdma/plain
>>>   Command 'rdma' is not available, please install it first.
>>>   # To enable the test:
>>>   # (1) Run 'scripts/rdma-migration-helper.sh setup' with root and rerun the test
>>>   # or
>>>   # (2) Run the test with root privilege
>> 
>> Could this might be the issue, should we skip if not root, as calling
>> the script in "detect" mode makes the new_rdma_link() method to succeed.
>
> It's expected the 'detect' should succeed and print a IPv4 address
>
>> 
>>>   #
>>>   ok 1 /x86_64/migration/precopy/rdma/plain # SKIP No rdma link available
>>>   # End of rdma tests
>>>
>>> Note: Remove the newly added RXE link by executing 'modprobe -r rdma_rxe'
>>> or by specifying 'clean' within this script.
>> 
>> qtest_add() provides both setup() / teardown() methods.> 
>
> This may require a minor refactor of the migration-test framework to
> enable support for setup() and teardown() methods.
>
> Let me see...
>
>
>
>> Test leaving system in different state seems bogus to me.
>
> At this point, I'm unable to refute. It indeed dirty the
> It indeed might dirty the system.
>
>
> A palatable compromise might be that, regardless of whether one is a
> root user or not, this test is only supported on hosts with RDMA link.
>
> Otherwise, it will provide an SKIP warning.
>
>       # Run 'scripts/rdma-migration-helper.sh setup' with root and rerun the test
>       # Optional: run 'scripts/rdma-migration-helper.sh clean' to revert the 'setup'
>
> For local users, they can independently use this script to set up and clean RDMA,
> as they are aware of the modifications they have made to the system.

Yes, let's skip it unless the user has very explicitly set things up.