Nouveau - Jun 2015 - [PATCH 1/6] gr: support for NVIDIA-provided firmwares

On 19 June 2015 at 00:47, Alexandre Courbot <gnurou at gmail.com>
wrote:
> NVIDIA will officially start providing signed firmwares through
> linux-firmware. Change the GR firmware lookup function to look them up
> in addition to the extracted firmwares.
I wonder if perhaps we should just replace the mechanism entirely, and
remove the support for nouveau/fuc* as we add "official" support for
NVIDIA's ucode.  The existing code is actually partially broken
anyway, and mostly works by luck and was intended as a development aid
/ workaround anyway.  There are no chipsets (aside from GM2xx...)
which we don't currently support using our own ucode, so the impact of
removing it will be very minimal.

Thoughts?

Ben.
>
> Signed-off-by: Alexandre Courbot <acourbot at nvidia.com>
> ---
>  drm/nouveau/nvkm/engine/gr/gf100.c | 66
+++++++++++++++++++++++++++++---------
>  1 file changed, 51 insertions(+), 15 deletions(-)
>
> diff --git a/drm/nouveau/nvkm/engine/gr/gf100.c
b/drm/nouveau/nvkm/engine/gr/gf100.c
> index ca11ddb..39d482f 100644
> --- a/drm/nouveau/nvkm/engine/gr/gf100.c
> +++ b/drm/nouveau/nvkm/engine/gr/gf100.c
> @@ -1544,26 +1544,62 @@ gf100_gr_dtor_fw(struct gf100_gr_fuc *fuc)
>         fuc->data = NULL;
>  }
>
> +/**
> + * gf100_gr_ctor_fw - helper for loading external GR firmwares
> + *
> + * A firmware can either be officially provided by NVIDIA (in which case
it will
> + * use a "NVIDIA name", or be extracted from the binary blob
(and use a
> + * "Nouveau name". The fwname and nvfwname are to be given the
Nouveau and
> + * NVIDIA names of a given firmware, respectively. This function will then
> + * try to load the NVIDIA firmware, then the extracted one, in that order.
> + *
> + */
>  int
>  gf100_gr_ctor_fw(struct gf100_gr_priv *priv, const char *fwname,
> -                struct gf100_gr_fuc *fuc)
> +                const char *nvfwname, struct gf100_gr_fuc *fuc)
>  {
>         struct nvkm_device *device = nv_device(priv);
>         const struct firmware *fw;
> -       char f[32];
> -       int ret;
> +       char f[64];
> +       int ret = -EINVAL;
> +       int i;
>
> -       snprintf(f, sizeof(f), "nouveau/nv%02x_%s",
device->chipset, fwname);
> -       ret = request_firmware(&fw, f, nv_device_base(device));
> -       if (ret) {
> -               snprintf(f, sizeof(f), "nouveau/%s", fwname);
> -               ret = request_firmware(&fw, f, nv_device_base(device));
> -               if (ret) {
> -                       nv_error(priv, "failed to load %s\n",
fwname);
> -                       return ret;
> +       /*
> +        * NVIDIA firmware name provided - try to load it
> +        * We try this first since most chips that require external
firmware
> +        * are supported by NVIDIA
> +        */
> +       if (nvfwname) {
> +               snprintf(f, sizeof(f), "nvidia/%s/%s.bin",
device->cname,
> +                        nvfwname);
> +               i = strlen(f);
> +               while (i) {
> +                       --i;
> +                       f[i] = tolower(f[i]);
>                 }
> +               ret = request_firmware_direct(&fw, f,
nv_device_base(device));
> +               if (!ret)
> +                       goto found;
> +       }
> +
> +       /* Nouveau firmware name provided - try to load it */
> +       if (fwname) {
> +               snprintf(f, sizeof(f), "nouveau/nv%02x_%s",
device->chipset,
> +                        fwname);
> +               ret = request_firmware_direct(&fw, f,
nv_device_base(device));
> +               if (!ret)
> +                       goto found;
> +
> +               snprintf(f, sizeof(f), "nouveau/%s", fwname);
> +               ret = request_firmware_direct(&fw, f,
nv_device_base(device));
> +               if (!ret)
> +                       goto found;
>         }
>
> +       nv_error(priv, "failed to load %s / %s\n", fwname,
nvfwname);
> +       return ret;
> +
> +found:
>         fuc->size = fw->size;
>         fuc->data = kmemdup(fw->data, fuc->size, GFP_KERNEL);
>         release_firmware(fw);
> @@ -1615,10 +1651,10 @@ gf100_gr_ctor(struct nvkm_object *parent, struct
nvkm_object *engine,
>
>         if (use_ext_fw) {
>                 nv_info(priv, "using external firmware\n");
> -               if (gf100_gr_ctor_fw(priv, "fuc409c",
&priv->fuc409c) ||
> -                   gf100_gr_ctor_fw(priv, "fuc409d",
&priv->fuc409d) ||
> -                   gf100_gr_ctor_fw(priv, "fuc41ac",
&priv->fuc41ac) ||
> -                   gf100_gr_ctor_fw(priv, "fuc41ad",
&priv->fuc41ad))
> +               if (gf100_gr_ctor_fw(priv, "fuc409c",
"fecs_inst", &priv->fuc409c) ||
> +                   gf100_gr_ctor_fw(priv, "fuc409d",
"fecs_data", &priv->fuc409d) ||
> +                   gf100_gr_ctor_fw(priv, "fuc41ac",
"gpccs_inst", &priv->fuc41ac) ||
> +                   gf100_gr_ctor_fw(priv, "fuc41ad",
"gpccs_data", &priv->fuc41ad))
>                         return -ENODEV;
>                 priv->firmware = true;
>         }
> --
> 2.4.3
>
> _______________________________________________
> dri-devel mailing list
> dri-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel

On Fri, Jun 19, 2015 at 1:40 PM, Ben Skeggs <skeggsb at gmail.com>
wrote:
> On 19 June 2015 at 00:47, Alexandre Courbot <gnurou at gmail.com>
wrote:
>> NVIDIA will officially start providing signed firmwares through
>> linux-firmware. Change the GR firmware lookup function to look them up
>> in addition to the extracted firmwares.
> I wonder if perhaps we should just replace the mechanism entirely, and
> remove the support for nouveau/fuc* as we add "official" support
for
> NVIDIA's ucode.  The existing code is actually partially broken
> anyway, and mostly works by luck and was intended as a development aid
> / workaround anyway.  There are no chipsets (aside from GM2xx...)
> which we don't currently support using our own ucode, so the impact of
> removing it will be very minimal.
>
> Thoughts?
I'm all for making things simpler, and if someone needs to use an
external firmware for a Nouveau-supported GPU they can always put it
under the nvidia/ firmware directory. So if you agree with it I will
remove support for firmwares in nouveau/ in GR. I am not sure whether
your statement also applies to other firmwares (falcon, xtensa)?

On 19 June 2015 at 15:22, Alexandre Courbot <gnurou at gmail.com>
wrote:
> On Fri, Jun 19, 2015 at 1:40 PM, Ben Skeggs <skeggsb at gmail.com>
wrote:
>> On 19 June 2015 at 00:47, Alexandre Courbot <gnurou at gmail.com>
wrote:
>>> NVIDIA will officially start providing signed firmwares through
>>> linux-firmware. Change the GR firmware lookup function to look them
up
>>> in addition to the extracted firmwares.
>> I wonder if perhaps we should just replace the mechanism entirely, and
>> remove the support for nouveau/fuc* as we add "official"
support for
>> NVIDIA's ucode.  The existing code is actually partially broken
>> anyway, and mostly works by luck and was intended as a development aid
>> / workaround anyway.  There are no chipsets (aside from GM2xx...)
>> which we don't currently support using our own ucode, so the impact
of
>> removing it will be very minimal.
>>
>> Thoughts?
>
> I'm all for making things simpler, and if someone needs to use an
> external firmware for a Nouveau-supported GPU they can always put it
> under the nvidia/ firmware directory. So if you agree with it I will
> remove support for firmwares in nouveau/ in GR. I am not sure whether
> your statement also applies to other firmwares (falcon, xtensa)?
I'd say leave the non-gr engines for now, at least until we know what
NVIDIA plans on doing with dGPU firmwares.