Date: Tue, 15 Mar 2016 14:56:13 -0700 (PDT) From: Don Lewis <truckman@FreeBSD.org> To: glebius@FreeBSD.org Cc: hps@selasky.org, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r296880 - in head: share/man/man9 sys/kern sys/sys Message-ID: <201603152156.u2FLuDPJ035986@gw.catspoiler.org> In-Reply-To: <20160315190549.GT1328@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 15 Mar, Gleb Smirnoff wrote: > Hans, > > On Tue, Mar 15, 2016 at 08:50:11AM +0100, Hans Petter Selasky wrote: > H> > Modified: head/sys/sys/sysctl.h > H> > ============================================================================== > H> > --- head/sys/sys/sysctl.h Mon Mar 14 23:49:16 2016 (r296879) > H> > +++ head/sys/sys/sysctl.h Tue Mar 15 00:05:00 2016 (r296880) > H> > @@ -204,6 +204,7 @@ int sysctl_handle_long(SYSCTL_HANDLER_AR > H> > int sysctl_handle_string(SYSCTL_HANDLER_ARGS); > H> > int sysctl_handle_opaque(SYSCTL_HANDLER_ARGS); > H> > int sysctl_handle_counter_u64(SYSCTL_HANDLER_ARGS); > H> > +int sysctl_handle_counter_u64_array(SYSCTL_HANDLER_ARGS); > H> > > H> > int sysctl_handle_uma_zone_max(SYSCTL_HANDLER_ARGS); > H> > int sysctl_handle_uma_zone_cur(SYSCTL_HANDLER_ARGS); > H> > @@ -648,6 +649,26 @@ TAILQ_HEAD(sysctl_ctx_list, sysctl_ctx_e > H> > __ptr, 0, sysctl_handle_counter_u64, "QU", __DESCR(descr)); \ > H> > }) > H> > > H> > +/* Oid for an array of counter(9)s. The pointer and length must be non zero. */ > H> > +#define SYSCTL_COUNTER_U64_ARRAY(parent, nbr, name, access, ptr, len, descr) \ > H> > + SYSCTL_OID(parent, nbr, name, \ > H> > + CTLTYPE_OPAQUE | CTLFLAG_MPSAFE | (access), \ > H> > + (ptr), (len), sysctl_handle_counter_u64_array, "S", descr); \ > H> > + CTASSERT(((access) & CTLTYPE) == 0 || \ > H> > + ((access) & SYSCTL_CT_ASSERT_MASK) == CTLTYPE_OPAQUE) > H> > + > H> > H> Gleb: I suggest you follow the example of the other static SYSCTLs, and > H> extend the CTASSERT to also cover the size of the array: > > I agree. > > H> CTASSERT(((access) & CTLTYPE) == 0 || \ > H> (((access) & SYSCTL_CT_ASSERT_MASK) == CTLTYPE_OPAQUE && \ > H> (sizeof(*(ptr)) == 8 * (len)))) > > Hardcoding 8 is definitely wrong here. If you come up with correct assert, > feel free to commit it. I failed to imagine one quickly. > > H> > +#define SYSCTL_ADD_COUNTER_U64_ARRAY(ctx, parent, nbr, name, access, \ > H> > + ptr, len, descr) \ > H> > +({ \ > H> > + counter_u64_t *__ptr = (ptr); \ > H> > + CTASSERT(((access) & CTLTYPE) == 0 || \ > H> > + ((access) & SYSCTL_CT_ASSERT_MASK) == CTLTYPE_OPAQUE); \ > H> > H> Same here possibly. Maybe extend the CTASSERT() to also cover the size > H> of the array? Would limit the "len" to a constant argument though. > > Of course limiting len to constant for a dynamic oid is something we > don't want. How can you do a CTASSERT() on something that isn't known at compile time?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201603152156.u2FLuDPJ035986>