Re: [PATCH 08/13] ppc/pnv: Set POWER9, POWER10 ibm,pa-features bits

["Nicholas Piggin" <npiggin@gmail.com>]

On Tue Mar 12, 2024 at 6:06 PM AEST, Cédric Le Goater wrote:
> On 3/11/24 19:51, Nicholas Piggin wrote:
> > Copy the pa-features arrays from spapr, adjusting slightly as
> > described in comments.
> > 
> > Cc: "Cédric Le Goater" <clg@kaod.org>
> > Cc: "Frédéric Barrat" <fbarrat@linux.ibm.com>
> > Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> > ---
> >   hw/ppc/pnv.c   | 67 ++++++++++++++++++++++++++++++++++++++++++++++++--
> >   hw/ppc/spapr.c |  1 +
> >   2 files changed, 66 insertions(+), 2 deletions(-)
> > 
> > diff --git a/hw/ppc/pnv.c b/hw/ppc/pnv.c
> > index 52d964f77a..3e30c08420 100644
> > --- a/hw/ppc/pnv.c
> > +++ b/hw/ppc/pnv.c
> > @@ -332,6 +332,35 @@ static void pnv_chip_power8_dt_populate(PnvChip *chip, void *fdt)
> >       }
> >   }
> >   
> > +/*
> > + * Same as spapr pa_features_300 except pnv always enables CI largepages bit.
> > + */
> > +static const uint8_t pa_features_300[] = { 66, 0,
> > +    /* 0: MMU|FPU|SLB|RUN|DABR|NX, 1: CILRG|fri[nzpm]|DABRX|SPRG3|SLB0|PP110 */
> > +    /* 2: VPM|DS205|PPR|DS202|DS206, 3: LSD|URG, 5: LE|CFAR|EB|LSQ */
> > +    0xf6, 0x3f, 0xc7, 0xc0, 0x00, 0xf0, /* 0 - 5 */
> > +    /* 6: DS207 */
> > +    0x80, 0x00, 0x00, 0x00, 0x00, 0x00, /* 6 - 11 */
> > +    /* 16: Vector */
> > +    0x00, 0x00, 0x00, 0x00, 0x80, 0x00, /* 12 - 17 */
> > +    /* 18: Vec. Scalar, 20: Vec. XOR, 22: HTM */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 18 - 23 */
> > +    /* 24: Ext. Dec, 26: 64 bit ftrs, 28: PM ftrs */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 24 - 29 */
> > +    /* 32: LE atomic, 34: EBB + ext EBB */
> > +    0x00, 0x00, 0x80, 0x00, 0xC0, 0x00, /* 30 - 35 */
> > +    /* 40: Radix MMU */
> > +    0x00, 0x00, 0x00, 0x00, 0x80, 0x00, /* 36 - 41 */
> > +    /* 42: PM, 44: PC RA, 46: SC vec'd */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 42 - 47 */
> > +    /* 48: SIMD, 50: QP BFP, 52: String */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 48 - 53 */
> > +    /* 54: DecFP, 56: DecI, 58: SHA */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 54 - 59 */
> > +    /* 60: NM atomic, 62: RNG */
> > +    0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 60 - 65 */
> > +};
> > +
> >   static void pnv_chip_power9_dt_populate(PnvChip *chip, void *fdt)
> >   {
> >       static const char compat[] = "ibm,power9-xscom\0ibm,xscom";
> > @@ -349,7 +378,7 @@ static void pnv_chip_power9_dt_populate(PnvChip *chip, void *fdt)
> >           offset = pnv_dt_core(chip, pnv_core, fdt);
> >   
> >           _FDT((fdt_setprop(fdt, offset, "ibm,pa-features",
> > -                           pa_features_207, sizeof(pa_features_207))));
> > +                           pa_features_300, sizeof(pa_features_300))));
> >       }
> >   
> >       if (chip->ram_size) {
> > @@ -359,6 +388,40 @@ static void pnv_chip_power9_dt_populate(PnvChip *chip, void *fdt)
> >       pnv_dt_lpc(chip, fdt, 0, PNV9_LPCM_BASE(chip), PNV9_LPCM_SIZE);
> >   }
> >   
> > +/*
> > + * Same as spapr pa_features_31 except pnv always enables CI largepages bit,
> > + * always disables copy/paste.
> > + */
> > +static const uint8_t pa_features_31[] = { 74, 0,
> > +    /* 0: MMU|FPU|SLB|RUN|DABR|NX, 1: CILRG|fri[nzpm]|DABRX|SPRG3|SLB0|PP110 */
> > +    /* 2: VPM|DS205|PPR|DS202|DS206, 3: LSD|URG, 5: LE|CFAR|EB|LSQ */
> > +    0xf6, 0x3f, 0xc7, 0xc0, 0x00, 0xf0, /* 0 - 5 */
> > +    /* 6: DS207 */
> > +    0x80, 0x00, 0x00, 0x00, 0x00, 0x00, /* 6 - 11 */
> > +    /* 16: Vector */
> > +    0x00, 0x00, 0x00, 0x00, 0x80, 0x00, /* 12 - 17 */
> > +    /* 18: Vec. Scalar, 20: Vec. XOR */
> > +    0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 18 - 23 */
> > +    /* 24: Ext. Dec, 26: 64 bit ftrs, 28: PM ftrs */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 24 - 29 */
> > +    /* 32: LE atomic, 34: EBB + ext EBB */
> > +    0x00, 0x00, 0x80, 0x00, 0xC0, 0x00, /* 30 - 35 */
> > +    /* 40: Radix MMU */
> > +    0x00, 0x00, 0x00, 0x00, 0x80, 0x00, /* 36 - 41 */
> > +    /* 42: PM, 44: PC RA, 46: SC vec'd */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 42 - 47 */
> > +    /* 48: SIMD, 50: QP BFP, 52: String */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 48 - 53 */
> > +    /* 54: DecFP, 56: DecI, 58: SHA */
> > +    0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 54 - 59 */
> > +    /* 60: NM atomic, 62: RNG */
> > +    0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 60 - 65 */
> > +    /* 68: DEXCR[SBHE|IBRTPDUS|SRAPD|NPHIE|PHIE] */
> > +    0x00, 0x00, 0xce, 0x00, 0x00, 0x00, /* 66 - 71 */
> > +    /* 72: [P]HASHCHK */
> > +    0x80, 0x00,                         /* 72 - 73 */
> > +};
> > +
> >   static void pnv_chip_power10_dt_populate(PnvChip *chip, void *fdt)
> >   {
> >       static const char compat[] = "ibm,power10-xscom\0ibm,xscom";
> > @@ -376,7 +439,7 @@ static void pnv_chip_power10_dt_populate(PnvChip *chip, void *fdt)
> >           offset = pnv_dt_core(chip, pnv_core, fdt);
> >   
> >           _FDT((fdt_setprop(fdt, offset, "ibm,pa-features",
> > -                           pa_features_207, sizeof(pa_features_207))));
> > +                           pa_features_31, sizeof(pa_features_31))));
> >       }
> >   
> >       if (chip->ram_size) {
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index 128bfe11a8..b53c13e037 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -233,6 +233,7 @@ static void spapr_dt_pa_features(SpaprMachineState *spapr,
> >                                    PowerPCCPU *cpu,
> >                                    void *fdt, int offset)
> >   {
> > +    /* These should be kept in sync with pnv */
>
> yes. In that case, the array definition should be moved under target/ppc/.
> May be under PowerPCCPUClass ?

Yeah PowerPCCPUClass might be a good idea, although I'm not quite
decided whether it's best to just store the arrays there, or make a
list of features in another format and have a builder function to
turn that into the dt array. There's also a few differences between
spapr and pnv that I haven't worked out a nice way to handle yet.
I have a pi-features property to add too which is similar.

So yes this is a bit ugly but we're already duplicating and open coding
arrays so I'd like to just get this in to fix the missing P10 bits,
and refactor it afterwards.

Thanks,
Nick