This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Linux: Deprecate <sys/sysctl.h> and sysctl
On 07/06/2019 13:01, Florian Weimer wrote:
> * Adhemerval Zanella:
>
>> On 07/06/2019 09:01, Florian Weimer wrote:
>>> Now that there are no internal users of __sysctl left, it is possible
>>> to add an unconditional deprecation warning to <sys/sysctl.h>.
>>>
>>> To avoid a test failure due this warning in check-install-headers,
>>> skip the test for sys/sysctl.h.
>>
>> LGTM. As a side note, if I recall correctly one usage of sysctl that might
>> be useful and not really available depending of the kernel version is to
>> get some random bits from /proc/sys/kernel/random/uuid without opening
>> /proc.
>
> Right, I saw this quite a lot in existing code (although some people use
> the FreeBSD MIB on Linux, which will always fail and never provide a UUID).
>
>>> +#warning "The <sys/sysctl.h> header is deprecated and will be removed."
>>> +#warning "Directly access the /proc file system instead."
>
> Should I add this as well?
>
> #warning "On current kernels, getentropy provides random bits without /proc."
Not sure, I think a note on NEWS might be better fort this specific usage.
>
>>> diff --git a/sysdeps/unix/sysv/linux/sysctl.c b/sysdeps/unix/sysv/linux/sysctl.c
>>> index 33afdd918b..0f18c69abe 100644
>>> --- a/sysdeps/unix/sysv/linux/sysctl.c
>>> +++ b/sysdeps/unix/sysv/linux/sysctl.c
>>> @@ -17,8 +17,7 @@
>>> <http://www.gnu.org/licenses/>. */
>>>
>>> #include <errno.h>
>>> -#include <string.h> /* For the real memset prototype. */
>>> -#include <sys/sysctl.h>
>>> +#include <linux/sysctl.h>
>>>
>>> #include <sysdep.h>
>>> #include <sys/syscall.h>
>>> @@ -39,5 +38,4 @@ __sysctl (int *name, int nlen, void *oldval, size_t *oldlenp,
>>>
>>> return INLINE_SYSCALL (_sysctl, 1, &args);
>>> }
>>> -libc_hidden_def (__sysctl)
>>> weak_alias (__sysctl, sysctl)
>>>
>>
>> What about a link warning stating this interface is deprecated and it would be
>> removed? My idea is to warn when people are static linking as well.
>
> They already get two warnings (from the header and the call site). Do
> we need a third warning? It's only helpful if you don't use the header,
> I think, but even with Go's cgo, you'll see the GCC warnings.
Fair enough.
Patch LGTM, thanks.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>