[PATCH 2/2] io: Add setuid tests for faccessat

[Adhemerval Zanella Netto]

On 06/11/24 12:13, Siddhesh Poyarekar wrote:
> On 2024-11-06 10:09, Adhemerval Zanella Netto wrote:
>>
>>
>> On 16/10/24 16:21, Siddhesh Poyarekar wrote:
>>> Add a new test tst-faccessat-setuid that iterates through real and
>>> effective UID/GID combination and tests the faccessat() interface for
>>> default and AT_EACCESS flags.
>>>
>>> Signed-off-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
>>> ---
>>>   io/Makefile               |   4 +
>>>   io/tst-faccessat-setuid.c | 163 ++++++++++++++++++++++++++++++++++++++
>>>   2 files changed, 167 insertions(+)
>>>   create mode 100644 io/tst-faccessat-setuid.c
>>>
>>> diff --git a/io/Makefile b/io/Makefile
>>> index a8d575e9ce..3ae26a5aa3 100644
>>> --- a/io/Makefile
>>> +++ b/io/Makefile
>>> @@ -232,6 +232,10 @@ tests := \
>>>     tst-utimes \
>>>     # tests
>>>   +xtests := \
>>> +  tst-faccessat-setuid \
>>> +  #xtests
>>> +
>>
>> Can't we use support_become_root and make it a container test? It became
>> moot to add a new xtests, since it is not usually tested on make check
>> and not everyone knows about the make xcheck.
>>
> 
> The test needs multiple user and group combinations, which doesn't seem to be something tests-container can provide.  Does it make sense instead to add this as a `test` and bail out as UNSUPPORTED if it's not run as root?  That way users know that the test is being skipped over, similar to how we do setuid tests.

Right, I was hoping that it would be easier but it seems that we will
need to add some support on test-container for extra uid/gid mapping.

Bailing as unsupported in this case is fine.