Date: Sun, 10 May 2020 13:25:37 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Andriy Gapon <avg@freebsd.org> Cc: FreeBSD Current <freebsd-current@freebsd.org> Subject: Re: CHANGE_PV_LIST_LOCK_TO_PHYS is not correct when !NUMA ? Message-ID: <20200510102537.GE68906@kib.kiev.ua> In-Reply-To: <881b2aa7-b9df-da64-25b6-783f076115c4@FreeBSD.org> References: <0d7db402-621e-cc6b-2918-2078f63e2a9b@FreeBSD.org> <20200508161500.GC44519@kib.kiev.ua> <6485ab77-a3d0-8916-9431-74e4da1e3ea7@FreeBSD.org> <20200509161325.GH44519@kib.kiev.ua> <d8876981-4775-0548-e49e-2f9171a2db10@FreeBSD.org> <20200509165010.GI44519@kib.kiev.ua> <881b2aa7-b9df-da64-25b6-783f076115c4@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, May 10, 2020 at 01:02:45PM +0300, Andriy Gapon wrote:
> On 09/05/2020 19:50, Konstantin Belousov wrote:
> > On Sat, May 09, 2020 at 07:16:27PM +0300, Andriy Gapon wrote:
> >> On 09/05/2020 19:13, Konstantin Belousov wrote:
> >>> On Sat, May 09, 2020 at 06:52:24PM +0300, Andriy Gapon wrote:
> >>>> I tried this change:
> >>>> diff --git a/sys/amd64/amd64/pmap.c b/sys/amd64/amd64/pmap.c
> >>>> index 4deed86a76d1a..b834b7f0388b7 100644
> >>>> --- a/sys/amd64/amd64/pmap.c
> >>>> +++ b/sys/amd64/amd64/pmap.c
> >>>> @@ -345,7 +345,7 @@ pmap_pku_mask_bit(pmap_t pmap)
> >>>> #define NPV_LIST_LOCKS MAXCPU
> >>>>
> >>>> #define PHYS_TO_PV_LIST_LOCK(pa) \
> >>>> - (&pv_list_locks[pa_index(pa) % NPV_LIST_LOCKS])
> >>>> + (&pv_list_locks[((pa) >> PDRSHIFT) % NPV_LIST_LOCKS])
> >>>> #endif
> >>>>
> >>>> #define CHANGE_PV_LIST_LOCK_TO_PHYS(lockp, pa) do { \
> >>>>
> >>>> It fixed the original problem, but I got a new panic.
> >>>> "DI already started" in pmap_remove() -> pmap_delayed_invl_start_u().
> >>>> I guess that !NUMA variant does not get much testing, so I'll probably just
> >>>> stick with the default.
> >>> Why didn't you just removed the KASSERT from pa_index ?
> >>
> >> Well, I thought it might be useful in the NUMA case.
> >> pa_index() definition is shared between both cases.
> > Might be define the macro two times, for NUMA/non-NUMA. non-NUMA case
> > does not need the assert, because users take it mod NPV_LIST_LOCKS.
>
> Yes, this works.
> Thank you!
>
> diff --git a/sys/amd64/amd64/pmap.c b/sys/amd64/amd64/pmap.c
> index 4deed86a76d1a..8dd236acc8205 100644
> --- a/sys/amd64/amd64/pmap.c
> +++ b/sys/amd64/amd64/pmap.c
> @@ -323,12 +323,12 @@ pmap_pku_mask_bit(pmap_t pmap)
> #endif
>
> #undef pa_index
> +#ifdef NUMA
> #define pa_index(pa) ({ \
> KASSERT((pa) <= vm_phys_segs[vm_phys_nsegs - 1].end, \
> ("address %lx beyond the last segment", (pa))); \
> (pa) >> PDRSHIFT; \
> })
> -#ifdef NUMA
> #define pa_to_pmdp(pa) (&pv_table[pa_index(pa)])
> #define pa_to_pvh(pa) (&(pa_to_pmdp(pa)->pv_page))
> #define PHYS_TO_PV_LIST_LOCK(pa) ({ \
> @@ -340,6 +340,7 @@ pmap_pku_mask_bit(pmap_t pmap)
> _lock; \
> })
> #else
> +#define pa_index(pa) ((pa) >> PDRSHIFT)
> #define pa_to_pvh(pa) (&pv_table[pa_index(pa)])
>
> #define NPV_LIST_LOCKS MAXCPU
Looks good to me.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20200510102537.GE68906>