Re: [PATCH] time: Make NTP optionnal

So even if the correspondong syscall are disabled and the
corresponding clocks too, you should return an -ENOSYS from the
do_adjtimex helper, in case that another component tries to use it in
the kernel, right ?

Regards,
Romain

2017-11-20 17:08 GMT+01:00 Alan Cox [off-list ref]:

On Mon, 20 Nov 2017 16:22:06 +0100
peter enderborg [off-list ref] wrote:

quoted

On 11/20/2017 04:00 PM, Romain Perier wrote:

quoted

Hi,

2017-11-20 15:10 GMT+01:00 peter enderborg [off-list ref]:

quoted

I think it should return a error code at least.

In which case ? The idea was to don't change the behaviour of these
functions (from the "API" point of view) and to avoid regressions in
the kernel components that depend on these NTP feature.
(so be transparent the most as possible)

Thanks,
Romain

do_adjtimex should return -ENOSYS in my opinion, however it is not in the manual
that it can return ENOSYS.

The standards specifications state what must be returned for certain
error cases, they don't forbid other errors being added.

Old old Linx will of course already return -1,ENOSYS to this call.

Alan