I don't see the original mail here either. Is it me or Ted, or a forward from a
private conversation? Anyway, I am OK with this and Robert had a similar diff
two months ago when this started. Just make sure this is off by default for
both type of binaries.
Paul
În 8 iulie 2020 18:42:41 EEST, Theo de Raadt <dera...@openbsd.org> a scris:
>I think we need something like this.
>
>Documenting it will be a challenge.
>
>I really don't like the name as is too generic, when the control is
>only
>for a narrow set of "current time" system calls.
>
>Ted Unangst <t...@tedunangst.com> wrote:
>
>> Not sure how useful this will be, but I think it could be helpful to
>still
>> see section (2) functions in ktrace, even if there's magic to avoid
>that.
>>
>> As proof of concept, if env LIBC_NOSYSWRAPPERS is set, the libc
>timecounters
>> are turned off. Now I see lots of gettimeofday syscalls in ktrace
>again.
>>
>> Is this better than switching to ltrace? Combined ktrace and ltrace
>output
>> is fairly messy, but it seems to work. Setting it up to trace just a
>few
>> functions and all the system calls is a bit more involved.
>>
>>
>> Index: init.c
>> ===================================================================
>> RCS file: /home/cvs/src/lib/libc/dlfcn/init.c,v
>> retrieving revision 1.8
>> diff -u -p -r1.8 init.c
>> --- init.c 6 Jul 2020 13:33:05 -0000 1.8
>> +++ init.c 8 Jul 2020 08:13:07 -0000
>> @@ -114,6 +114,8 @@ _libc_preinit(int argc, char **argv, cha
>> _timekeep->tk_version != TK_VERSION)
>> _timekeep = NULL;
>> }
>> + if (issetugid() == 0 && getenv("LIBC_NOSYSWRAPPERS"))
>> + _timekeep = NULL;
>> break;
>> }
>> }
>>
