drm/nouveau/clk/gk20a: parameterize PLL settings
authorAlexandre Courbot <acourbot@nvidia.com>
Wed, 1 Jun 2016 08:39:25 +0000 (17:39 +0900)
committerBen Skeggs <bskeggs@redhat.com>
Thu, 14 Jul 2016 01:53:25 +0000 (11:53 +1000)
Make functions manipulating PLL settings take them as an argument,
instead of assuming we want to work on the copy in the gk20a_clk
structure. This makes these functions more flexible, which we will need
in GM20B.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c

index 80031c0f9e3f1e0a9a5c276f51beeb7eb6e31879..a95eda7c5df4ef3328ef153c9a6343788f9bd270 100644 (file)
@@ -149,19 +149,20 @@ gk20a_pllg_write_mnp(struct gk20a_clk *clk, const struct gk20a_pll *pll)
 }
 
 static u32
-gk20a_pllg_calc_rate(struct gk20a_clk *clk)
+gk20a_pllg_calc_rate(struct gk20a_clk *clk, struct gk20a_pll *pll)
 {
        u32 rate;
        u32 divider;
 
-       rate = clk->parent_rate * clk->pll.n;
-       divider = clk->pll.m * clk->pl_to_div(clk->pll.pl);
+       rate = clk->parent_rate * pll->n;
+       divider = pll->m * clk->pl_to_div(pll->pl);
 
        return rate / divider / 2;
 }
 
 static int
-gk20a_pllg_calc_mnp(struct gk20a_clk *clk, unsigned long rate)
+gk20a_pllg_calc_mnp(struct gk20a_clk *clk, unsigned long rate,
+                   struct gk20a_pll *pll)
 {
        struct nvkm_subdev *subdev = &clk->base.subdev;
        u32 target_clk_f, ref_clk_f, target_freq;
@@ -256,16 +257,16 @@ found_match:
                           "no best match for target @ %dMHz on gpc_pll",
                           target_clk_f / KHZ);
 
-       clk->pll.m = best_m;
-       clk->pll.n = best_n;
-       clk->pll.pl = best_pl;
+       pll->m = best_m;
+       pll->n = best_n;
+       pll->pl = best_pl;
 
-       target_freq = gk20a_pllg_calc_rate(clk);
+       target_freq = gk20a_pllg_calc_rate(clk, pll);
 
        nvkm_debug(subdev,
-                  "actual target freq %d MHz, M %d, N %d, PL %d(div%d)\n",
-                  target_freq / MHZ, clk->pll.m, clk->pll.n, clk->pll.pl,
-                  clk->pl_to_div(clk->pll.pl));
+                  "actual target freq %d KHz, M %d, N %d, PL %d(div%d)\n",
+                  target_freq / KHZ, pll->m, pll->n, pll->pl,
+                  clk->pl_to_div(pll->pl));
        return 0;
 }
 
@@ -333,7 +334,8 @@ gk20a_pllg_disable(struct gk20a_clk *clk)
 }
 
 static int
-_gk20a_pllg_program_mnp(struct gk20a_clk *clk, bool allow_slide)
+_gk20a_pllg_program_mnp(struct gk20a_clk *clk, struct gk20a_pll *pll,
+                       bool allow_slide)
 {
        struct nvkm_subdev *subdev = &clk->base.subdev;
        struct nvkm_device *device = subdev->device;
@@ -346,9 +348,9 @@ _gk20a_pllg_program_mnp(struct gk20a_clk *clk, bool allow_slide)
 
        /* do NDIV slide if there is no change in M and PL */
        cfg = nvkm_rd32(device, GPCPLL_CFG);
-       if (allow_slide && clk->pll.m == old_pll.m &&
-           clk->pll.pl == old_pll.pl && (cfg & GPCPLL_CFG_ENABLE)) {
-               return gk20a_pllg_slide(clk, clk->pll.n);
+       if (allow_slide && pll->m == old_pll.m &&
+           pll->pl == old_pll.pl && (cfg & GPCPLL_CFG_ENABLE)) {
+               return gk20a_pllg_slide(clk, pll->n);
        }
 
        /* slide down to NDIV_LO */
@@ -385,11 +387,11 @@ _gk20a_pllg_program_mnp(struct gk20a_clk *clk, bool allow_slide)
        gk20a_pllg_disable(clk);
 
        nvkm_debug(subdev, "%s: m=%d n=%d pl=%d\n", __func__,
-                  clk->pll.m, clk->pll.n, clk->pll.pl);
+                  pll->m, pll->n, pll->pl);
 
-       old_pll = clk->pll;
+       old_pll = *pll;
        if (allow_slide)
-               old_pll.n = DIV_ROUND_UP(clk->pll.m * clk->params->min_vco,
+               old_pll.n = DIV_ROUND_UP(pll->m * clk->params->min_vco,
                                         clk->parent_rate / KHZ);
        gk20a_pllg_write_mnp(clk, &old_pll);
 
@@ -425,7 +427,7 @@ _gk20a_pllg_program_mnp(struct gk20a_clk *clk, bool allow_slide)
        }
 
        /* slide up to new NDIV */
-       return allow_slide ? gk20a_pllg_slide(clk, clk->pll.n) : 0;
+       return allow_slide ? gk20a_pllg_slide(clk, pll->n) : 0;
 }
 
 static int
@@ -433,9 +435,9 @@ gk20a_pllg_program_mnp(struct gk20a_clk *clk)
 {
        int err;
 
-       err = _gk20a_pllg_program_mnp(clk, true);
+       err = _gk20a_pllg_program_mnp(clk, &clk->pll, true);
        if (err)
-               err = _gk20a_pllg_program_mnp(clk, false);
+               err = _gk20a_pllg_program_mnp(clk, &clk->pll, false);
 
        return err;
 }
@@ -540,13 +542,14 @@ gk20a_clk_read(struct nvkm_clk *base, enum nv_clk_src src)
        struct gk20a_clk *clk = gk20a_clk(base);
        struct nvkm_subdev *subdev = &clk->base.subdev;
        struct nvkm_device *device = subdev->device;
+       struct gk20a_pll pll;
 
        switch (src) {
        case nv_clk_src_crystal:
                return device->crystal;
        case nv_clk_src_gpc:
-               gk20a_pllg_read_mnp(clk, &clk->pll);
-               return gk20a_pllg_calc_rate(clk) / GK20A_CLK_GPC_MDIV;
+               gk20a_pllg_read_mnp(clk, &pll);
+               return gk20a_pllg_calc_rate(clk, &pll) / GK20A_CLK_GPC_MDIV;
        default:
                nvkm_error(subdev, "invalid clock source %d\n", src);
                return -EINVAL;
@@ -559,7 +562,7 @@ gk20a_clk_calc(struct nvkm_clk *base, struct nvkm_cstate *cstate)
        struct gk20a_clk *clk = gk20a_clk(base);
 
        return gk20a_pllg_calc_mnp(clk, cstate->domain[nv_clk_src_gpc] *
-                                        GK20A_CLK_GPC_MDIV);
+                                        GK20A_CLK_GPC_MDIV, &clk->pll);
 }
 
 int