Re: [LTP] [PATCH v2 4/4] Add error coverage for landlock network support

[Andrea Cervesato via ltp <ltp@lists.linux.it>]

On 11/6/24 11:38, Cyril Hrubis wrote:
> Hi!
>>>> -	verify_landlock_is_enabled();
>>>> +	abi_current = verify_landlock_is_enabled();
>>>>    
>>>>    	ruleset_attr->handled_access_fs = LANDLOCK_ACCESS_FS_EXECUTE;
>>>>    
>>>>    	ruleset_fd = TST_EXP_FD_SILENT(tst_syscall(__NR_landlock_create_ruleset,
>>>> -		ruleset_attr, sizeof(struct tst_landlock_ruleset_attr_abi1), 0));
>>>> +		ruleset_attr, sizeof(struct tst_landlock_ruleset_attr_abi4), 0));
>>>                                  ^
>>> 			       This should be abi_current otherwise we
>>> 			       will fail on v1 only system.
>>>
>>>>    }
>>>>    
>> In what sense? abi4 is already the last one. At least the last supported
>> by LTP.
> Because if we request abi4 it will fail on kernels that only support
> abi1. We try hard to skip the abi4 tests, but we wouldn't get there at
> all on abi1 kernel because we would fail to create the ruleset_fd in the
> test setup.
>
> And we cannot initialize the abi to anything newer than abi4 either,
> because we pass abi4 structure in the test. It's fine that we pass abi4
> structure on abi1 system here, because the test only checks for invalid
> cases and all we need here is to pass a valid attr and size so that we
> get rejected by the kernel on the rest of the parameters.
>
I can instantiate both ABI1 and ABI4 structs, then I pass it to the 
tcase accordingly.

Andrea


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp