RE: [RFC PATCH 2/4] clk: renesas: rzg2l-cpg: Add support to stack the resets instead of indexing
From: Biju Das
Date: Fri May 06 2022 - 08:11:11 EST
Hi Prabhakar,
> Subject: Re: [RFC PATCH 2/4] clk: renesas: rzg2l-cpg: Add support to stack
> the resets instead of indexing
>
> Hi Biju,
>
> Thank you for the review.
>
> On Thu, May 5, 2022 at 8:48 PM Biju Das <biju.das.jz@xxxxxxxxxxxxxx> wrote:
> >
> > Hi Lad Prabhakar,
> >
> > Thanks for the patch.
> >
> > > Subject: [RFC PATCH 2/4] clk: renesas: rzg2l-cpg: Add support to
> > > stack the resets instead of indexing
> > >
> > > Instead of indexing the resets, stack them and instead create an id
> > > member in struct rzg2l_reset to store the index. With this approach
> > > for every id we will have to loop through the resets array to match the
> id.
> > >
> > > This in preparation to add support for Renesas RZ/Five CPG in
> > > r9a07g043- cpg.c file where the resets array will be split up into
> > > three i.e. common and two SoC specific.
> > >
> > > Signed-off-by: Lad Prabhakar
> > > <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> > > ---
> > > drivers/clk/renesas/rzg2l-cpg.c | 76
> > > ++++++++++++++++++++++++++------- drivers/clk/renesas/rzg2l-cpg.h |
> > > 4 +-
> > > 2 files changed, 63 insertions(+), 17 deletions(-)
> > >
> > > diff --git a/drivers/clk/renesas/rzg2l-cpg.c
> > > b/drivers/clk/renesas/rzg2l- cpg.c index 1ce35f65682b..94fe307ec4c5
> > > 100644
> > > --- a/drivers/clk/renesas/rzg2l-cpg.c
> > > +++ b/drivers/clk/renesas/rzg2l-cpg.c
> > > @@ -681,14 +681,37 @@ rzg2l_cpg_register_mod_clk(const struct
> > > rzg2l_mod_clk *mod,
> > >
> > > #define rcdev_to_priv(x) container_of(x, struct rzg2l_cpg_priv,
> > > rcdev)
> > >
> > > +static const struct rzg2l_reset
> > > +*rzg2l_get_reset_ptr(struct rzg2l_cpg_priv *priv,
> > > + unsigned long id)
> > > +
> > > +{
> > > + const struct rzg2l_cpg_info *info = priv->info;
> > > + unsigned int i;
> > > +
> > > + for (i = 0; i < priv->num_resets; i++) {
> > > + if (info->resets[i].id == id)
> > > + return &info->resets[i];
> > > + }
> >
> > Is it not possible to use shared reset like RZ/G2L and RZ/V2L?, which
> > has optimal memory and performance wise we can avoid bigger loop.
> >
> > Like adding Last index of RZ/Five as last reset index and Handle
> > RZ/G2UL specific as invalid reset index in xlate??
> >
> So we will have to maintain an array id's which are invalid to RZ/Five SoC.
> For this too we will have to loop at runtime itself. The array for invalid
> index will be big too.
As per [1], it will be 25 resets.
if you invalidate RZ/G2L specific resets in probe, there is no runtime overhead.
when a device match found, the info->reset_callback() which is mentioned in the next mail
and invalidate the resets(resets[id].off = 0)
ie,
if(info->reset_callback)
info->reset_callback();
and on r9a07g043-cpg.c, make resets[id].off = 0 to invalidate the resets.
https://git.kernel.org/pub/scm/linux/kernel/git/geert/renesas-drivers.git/tree/include/dt-bindings/clock/r9a07g043-cpg.h
>
> >
> > > +
> > > + return NULL;
> > > +}
> > > +
> > > static int rzg2l_cpg_reset(struct reset_controller_dev *rcdev,
> > > unsigned long id) {
> > > struct rzg2l_cpg_priv *priv = rcdev_to_priv(rcdev);
> > > - const struct rzg2l_cpg_info *info = priv->info;
> > > - unsigned int reg = info->resets[id].off;
> > > - u32 dis = BIT(info->resets[id].bit);
> > > - u32 we = dis << 16;
> > > + const struct rzg2l_reset *reset;
> > > + unsigned int reg;
> > > + u32 dis, we;
> > > +
> > > + reset = rzg2l_get_reset_ptr(priv, id);
> > > + if (!reset)
> > > + return -EINVAL;
> > > +
> > > + reg = reset->off;
> > > + dis = BIT(reset->bit);
> > > + we = dis << 16;
> > >
> > > dev_dbg(rcdev->dev, "reset id:%ld offset:0x%x\n", id,
> > > CLK_RST_R(reg));
> > >
> > > @@ -708,9 +731,16 @@ static int rzg2l_cpg_assert(struct
> > > reset_controller_dev *rcdev,
> > > unsigned long id) {
> > > struct rzg2l_cpg_priv *priv = rcdev_to_priv(rcdev);
> > > - const struct rzg2l_cpg_info *info = priv->info;
> > > - unsigned int reg = info->resets[id].off;
> > > - u32 value = BIT(info->resets[id].bit) << 16;
> > > + const struct rzg2l_reset *reset;
> > > + unsigned int reg;
> > > + u32 value;
> > > +
> > > + reset = rzg2l_get_reset_ptr(priv, id);
> > > + if (!reset)
> > > + return -EINVAL;
> > > +
> > > + reg = reset->off;
> > > + value = BIT(reset->bit) << 16;
> > >
> > > dev_dbg(rcdev->dev, "assert id:%ld offset:0x%x\n", id,
> > > CLK_RST_R(reg));
> > >
> > > @@ -722,11 +752,17 @@ static int rzg2l_cpg_deassert(struct
> > > reset_controller_dev *rcdev,
> > > unsigned long id) {
> > > struct rzg2l_cpg_priv *priv = rcdev_to_priv(rcdev);
> > > - const struct rzg2l_cpg_info *info = priv->info;
> > > - unsigned int reg = info->resets[id].off;
> > > - u32 dis = BIT(info->resets[id].bit);
> > > - u32 value = (dis << 16) | dis;
> > > + const struct rzg2l_reset *reset;
> > > + unsigned int reg;
> > > + u32 dis, value;
> > > +
> > > + reset = rzg2l_get_reset_ptr(priv, id);
> > > + if (!reset)
> > > + return -EINVAL;
> > >
> > > + reg = reset->off;
> > > + dis = BIT(reset->bit);
> > > + value = (dis << 16) | dis;
> > > dev_dbg(rcdev->dev, "deassert id:%ld offset:0x%x\n", id,
> > > CLK_RST_R(reg));
> > >
> > > @@ -738,9 +774,16 @@ static int rzg2l_cpg_status(struct
> > > reset_controller_dev *rcdev,
> > > unsigned long id) {
> > > struct rzg2l_cpg_priv *priv = rcdev_to_priv(rcdev);
> > > - const struct rzg2l_cpg_info *info = priv->info;
> > > - unsigned int reg = info->resets[id].off;
> > > - u32 bitmask = BIT(info->resets[id].bit);
> > > + const struct rzg2l_reset *reset;
> > > + unsigned int reg;
> > > + u32 bitmask;
> > > +
> > > + reset = rzg2l_get_reset_ptr(priv, id);
> > > + if (!reset)
> > > + return -EINVAL;
> > > +
> > > + reg = reset->off;
> > > + bitmask = BIT(reset->bit);
> > >
> > > return !(readl(priv->base + CLK_MRST_R(reg)) & bitmask); } @@
> > > -
> > > 756,10 +799,11 @@ static int rzg2l_cpg_reset_xlate(struct
> > > reset_controller_dev *rcdev,
> > > const struct of_phandle_args
> *reset_spec) {
> > > struct rzg2l_cpg_priv *priv = rcdev_to_priv(rcdev);
> > > - const struct rzg2l_cpg_info *info = priv->info;
> > > unsigned int id = reset_spec->args[0];
> > > + const struct rzg2l_reset *reset;
> > >
> > > - if (id >= rcdev->nr_resets || !info->resets[id].off) {
> > > + reset = rzg2l_get_reset_ptr(priv, id);
> > > + if (!reset) {
> > > dev_err(rcdev->dev, "Invalid reset index %u\n", id);
> > > return -EINVAL;
> > > }
> > > diff --git a/drivers/clk/renesas/rzg2l-cpg.h
> > > b/drivers/clk/renesas/rzg2l- cpg.h index 92c88f42ca7f..a99f2ba7868f
> > > 100644
> > > --- a/drivers/clk/renesas/rzg2l-cpg.h
> > > +++ b/drivers/clk/renesas/rzg2l-cpg.h
> > > @@ -152,12 +152,14 @@ struct rzg2l_mod_clk {
> > > * @bit: reset bit
> > > */
> > > struct rzg2l_reset {
> > > + unsigned int id;
> >
> > Now you are adding 4 bytes to each reset entry in the LUT.
> >
> Agreed, on the other hand we are saving space on the entries (for example
> not all the reset entries are listed in the array and the array length will
> always be equal to last index)
Not now, But eventually we will fill all the resets, that is reason it is declared in header file.
Cheers,
Biju
> >
> > > u16 off;
> > > u8 bit;
> > > };
> >
> > >
> > > #define DEF_RST(_id, _off, _bit) \
> > > - [_id] = { \
> > > + { \
> > > + .id = (_id), \
> > > .off = (_off), \
> > > .bit = (_bit) \
> > > }
> > > --
> > > 2.25.1
> >