Re: [PATCH v5 01/11] hw/misc: Add NPCM7xx System Global Control Registers device model

[Havard Skinnemoen <hskinnemoen@google.com>]

On Thu, Jul 9, 2020 at 9:23 AM Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
>
> On 7/9/20 8:43 AM, Havard Skinnemoen wrote:
> > On Wed, Jul 8, 2020 at 11:04 PM Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
> >> On 7/9/20 2:35 AM, Havard Skinnemoen wrote:
> >>> Implement a device model for the System Global Control Registers in the
> >>> NPCM730 and NPCM750 BMC SoCs.
> >>>
> >>> This is primarily used to enable SMP boot (the boot ROM spins reading
> >>> the SCRPAD register) and DDR memory initialization; other registers are
> >>> best effort for now.
> >>>
> >>> The reset values of the MDLR and PWRON registers are determined by the
> >>> SoC variant (730 vs 750) and board straps respectively.
> >>>
> >>> Reviewed-by: Joel Stanley <joel@jms.id.au>
> >>> Reviewed-by: Cédric Le Goater <clg@kaod.org>
> >>> Signed-off-by: Havard Skinnemoen <hskinnemoen@google.com>
> >>> ---
> [...]
> >>> diff --git a/hw/misc/npcm7xx_gcr.c b/hw/misc/npcm7xx_gcr.c
> >>> new file mode 100644
> >>> index 0000000000..9934cd238d
> >>> --- /dev/null
> >>> +++ b/hw/misc/npcm7xx_gcr.c
> >>> @@ -0,0 +1,226 @@
> >>> +/*
> >>> + * Nuvoton NPCM7xx System Global Control Registers.
> >>> + *
> >>> + * Copyright 2020 Google LLC
> >>> + *
> >>> + * This program is free software; you can redistribute it and/or modify it
> >>> + * under the terms of the GNU General Public License as published by the
> >>> + * Free Software Foundation; either version 2 of the License, or
> >>> + * (at your option) any later version.
> >>> + *
> >>> + * This program is distributed in the hope that it will be useful, but WITHOUT
> >>> + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
> >>> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
> >>> + * for more details.
> >>> + */
> >>> +
> >>> +#include "qemu/osdep.h"
> >>> +
> >>> +#include "hw/misc/npcm7xx_gcr.h"
> >>> +#include "hw/qdev-properties.h"
> >>> +#include "migration/vmstate.h"
> >>> +#include "qapi/error.h"
> >>> +#include "qemu/log.h"
> >>> +#include "qemu/module.h"
> >>> +#include "qemu/units.h"
> >>> +
> >>> +#include "trace.h"
> >>> +
> >>> +static const uint32_t cold_reset_values[NPCM7XX_GCR_NR_REGS] = {
> >>> +    [NPCM7XX_GCR_PDID]          = 0x04a92750,   /* Poleg A1 */
> >>> +    [NPCM7XX_GCR_MISCPE]        = 0x0000ffff,
> >>> +    [NPCM7XX_GCR_SPSWC]         = 0x00000003,
> >>> +    [NPCM7XX_GCR_INTCR]         = 0x0000035e,
> >>> +    [NPCM7XX_GCR_HIFCR]         = 0x0000004e,
> >>> +    [NPCM7XX_GCR_INTCR2]        = (1U << 19),   /* DDR initialized */
> >>> +    [NPCM7XX_GCR_RESSR]         = 0x80000000,
> >>> +    [NPCM7XX_GCR_DSCNT]         = 0x000000c0,
> >>> +    [NPCM7XX_GCR_DAVCLVLR]      = 0x5a00f3cf,
> >>> +    [NPCM7XX_GCR_SCRPAD]        = 0x00000008,
> >>> +    [NPCM7XX_GCR_USB1PHYCTL]    = 0x034730e4,
> >>> +    [NPCM7XX_GCR_USB2PHYCTL]    = 0x034730e4,
> >>> +};
> >>> +
> >>> +static uint64_t npcm7xx_gcr_read(void *opaque, hwaddr offset, unsigned size)
> >>> +{
> >>> +    uint32_t reg = offset / sizeof(uint32_t);
> >>> +    NPCM7xxGCRState *s = opaque;
> >>> +
> >>> +    if (reg >= NPCM7XX_GCR_NR_REGS) {
> >>> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: offset 0x%04x out of range\n",
> >>> +                      __func__, (unsigned int)offset);
> >>
> >> Maybe use HWADDR_PRIx instead of casting to int?
> >
> > Will do, thanks!
>
> Glad you are not annoyed by my review comments.
>
> FYI your series quality is in my top-4 "adding new machine to QEMU".
> I'd like to use it as example to follow.
>
> I am slowly reviewing because this is not part of my work duties,
> so I do that in my free time before/after work, which is why I can
> barely do more that 2 per day, as I have to learn the SoC and
> cross check documentation (or in this case, existing firmware code
> since there is no public doc).

Your feedback is super valuable, thanks a lot. I really appreciate it.

>
> [...]
> >>> +static void npcm7xx_gcr_realize(DeviceState *dev, Error **errp)
> >>> +{
> >>> +    NPCM7xxGCRState *s = NPCM7XX_GCR(dev);
> >>> +    uint64_t dram_size;
> >>> +    Error *err = NULL;
> >>> +    Object *obj;
> >>> +
> >>> +    obj = object_property_get_link(OBJECT(dev), "dram-mr", &err);
> >>> +    if (!obj) {
> >>> +        error_setg(errp, "%s: required dram-mr link not found: %s",
> >>> +                   __func__, error_get_pretty(err));
> >>> +        return;
> >>> +    }
> >>> +    dram_size = memory_region_size(MEMORY_REGION(obj));
> >>> +
> >>> +    /* Power-on reset value */
> >>> +    s->reset_intcr3 = 0x00001002;
> >>> +
> >>> +    /*
> >>> +     * The GMMAP (Graphics Memory Map) field is used by u-boot to detect the
> >>> +     * DRAM size, and is normally initialized by the boot block as part of DRAM
> >>> +     * training. However, since we don't have a complete emulation of the
> >>> +     * memory controller and try to make it look like it has already been
> >>> +     * initialized, the boot block will skip this initialization, and we need
> >>> +     * to make sure this field is set correctly up front.
> >>> +     *
> >>> +     * WARNING: some versions of u-boot only looks at bits 8 and 9, so 2 GiB or
> >>> +     * more of DRAM will be interpreted as 128 MiB.
> >>
> >> I'd say u-boot is right in wrapping at 2GB, as more DRAM is
> >> un-addressable.
> >
> > Ah, maybe I shouldn't have said "or more". The bug is that if you
> > specify _exactly_ 2GiB, u-boot will see 128 MiB.
> >
> > But I agree more than 2GiB doesn't make sense, so I'll add a check for that.
> >
> > Not sure if I agree that the check should be here. > 2 GiB is an
> > addressing limitation, and the GCR module shouldn't really know what
> > the SoC's address space looks like. The lower limit is because the GCR
> > module can't encode anything less than 128 MiB.
> >
> >>> +     *
> >>> +     * https://github.com/Nuvoton-Israel/u-boot/blob/2aef993bd2aafeb5408dbaad0f3ce099ee40c4aa/board/nuvoton/poleg/poleg.c#L244
> >>> +     */
> >>> +    if (dram_size >= 2 * GiB) {
> >>
> >> See my comment in this series on the machine patch.
> >>
> >>> +        s->reset_intcr3 |= 4 << 8;
> >>> +    } else if (dram_size >= 1 * GiB) {
> >>> +        s->reset_intcr3 |= 3 << 8;
> >>> +    } else if (dram_size >= 512 * MiB) {
> >>> +        s->reset_intcr3 |= 2 << 8;
> >>> +    } else if (dram_size >= 256 * MiB) {
> >>> +        s->reset_intcr3 |= 1 << 8;
> >>> +    } else if (dram_size >= 128 * MiB) {
> >>> +        s->reset_intcr3 |= 0 << 8;
>
> I think this can be simplified as:
>
>          s->reset_intcr3 = (4 - clz32(dram_size)) << 8;
>
> Which implicitly truncate to power of 2 (which is what this
> block does, as you can have only 1 bank managed per this SGC).

Good idea. I find this a little easier to understand though:

#define NPCM7XX_GCR_MIN_DRAM_SIZE   (128 * MiB)

    s->reset_intcr3 |= ctz64(dram_size / NPCM7XX_GCR_MIN_DRAM_SIZE) << 8;

> But checking is_power_of_2(dram_size) and reporting an error
> else, is probably even better.

OK, I'll add a check for this, and also explicit checks for too large
and too small. The former is currently redundant with the DRAM size
check I'm adding to npcm7xx_realize(), but I kind of like the idea of
checking for encoding limitations and addressing limitations
separately, as they may not necessarily stay the same forever.

> >>> +    } else {
> >>> +        error_setg(errp,
> >>> +                   "npcm7xx_gcr: DRAM size %" PRIu64
> >>> +                   " is too small (need 128 MiB minimum)",
> >>> +                   dram_size);
> >>
> >> Ah, so you could add the same error for >2GB. Easier.
> >>
> >> Preferably using HWADDR_PRIx, and similar error for >2GB:
> >> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>