Date: Sat, 22 Jan 2011 20:51:50 -0800 From: Garrett Cooper <yanegomi@gmail.com> To: Pawel Jakub Dawidek <pjd@freebsd.org> Cc: freebsd-fs@freebsd.org Subject: Re: Concerns over shim-layers for ZFS and kernel/userland namespace pollution Message-ID: <AANLkTi=ZGA%2BeT-dSVHZ49b4zRcGhO1DKUpzP4v2TnBMW@mail.gmail.com> In-Reply-To: <AANLkTikTcLMwDP9ec9UQgzLqEJWDb7u=6v_r=2mxoo7e@mail.gmail.com> References: <AANLkTikTcLMwDP9ec9UQgzLqEJWDb7u=6v_r=2mxoo7e@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Jan 22, 2011 at 5:06 PM, Garrett Cooper <yanegomi@gmail.com> wrote:
> Hi Pawel,
> =A0 =A0I've been trying to get rid of the time.h pollution in sys/time.h
> for POSIX conformance and I've run into a bit of a roadblock
> (referring to sys/cddl/compat/opensolaris/sys/time.h):
> =A0 =A01. The clock_gettime call in gethrtime() can fail and the failure
> itself isn't captured.
> =A0 =A02. The calls in many case assume userland behavior, being "return
> value", not "return errno, assign value to address passed in"
> =A0 =A03. AFAICT the calls shouldn't be calling the clock_gettime syscall
> interface from userland; they should be calling the clock_gettime
> interface. I could be partly wrong here, but if so then some
> additional hacking will need to be made to pull in time.h as that's
> where clock_gettime must be defined according to our manpages and
> POSIX. There might be another preexisting KPI that can be used in its
> place.
> =A0 =A0I haven't looked outside of that header for now.
> =A0 =A0Comments on my above statements are more than welcome.
Nevermind. It looks like I missed the #ifdef _KERNEL ... #else ...
#endif lines. It would be nice if the KPIs were consistent with
FreeBSD KPIs, but that's less of an issue I suppose.
Thanks,
-Garrett
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTi=ZGA%2BeT-dSVHZ49b4zRcGhO1DKUpzP4v2TnBMW>