lkml.org 
[lkml]   [2021]   [Dec]   [15]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [PATCH 4.19 29/74] clk: qcom: regmap-mux: fix parent clock lookup
    Hi!

    > > > The function mux_get_parent() uses qcom_find_src_index() to find the
    > > > parent clock index, which is incorrect: qcom_find_src_index() uses src
    > > > enum for the lookup, while mux_get_parent() should use cfg field (which
    > > > corresponds to the register value). Add qcom_find_cfg_index() function
    > > > doing this kind of lookup and use it for mux parent lookup.
    > >
    > > This appears to have problems with error handling.
    > >
    > > > +++ b/drivers/clk/qcom/clk-regmap-mux.c
    > > > @@ -36,7 +36,7 @@ static u8 mux_get_parent(struct clk_hw *
    > > > val &= mask;
    > > > if (mux->parent_map)
    > > > - return qcom_find_src_index(hw, mux->parent_map, val);
    > > > + return qcom_find_cfg_index(hw, mux->parent_map, val);
    > > > return val;
    > > > }
    > >
    > > So this returns u8.
    > >
    > > > +int qcom_find_cfg_index(struct clk_hw *hw, const struct parent_map *map, u8 cfg)
    > > > +{
    > > > + int i, num_parents = clk_hw_get_num_parents(hw);
    > > > +
    > > > + for (i = 0; i < num_parents; i++)
    > > > + if (cfg == map[i].cfg)
    > > > + return i;
    > > > +
    > > > + return -ENOENT;
    > > > +}
    > >
    > > In case of error, -ENOENT will be cast to u8 in caller. I don't
    > > believe that is correct.
    >
    > Unfortunately there is no way to return proper error code from
    > clk_ops->get_parent() callback. However returning -ENOENT would translate to
    > 254. Then clk_core_get_parent_by_index() would determine that there is no
    > such parent and return NULL. A call to clk_set_parent would reparent the
    > clock.

    Yeah, I guess it happens to work.

    > Returning some sensible default (e.g. 0) would be much worse, since then the
    > clock subsystem would assume that the clock has correct parent. A call to
    > clk_set_parent would always result in ops->set_parent() call, reparenting
    > the clock correctly.

    Well ~0 would be sensible in this case. And a comment with explanation.

    > Most probably it would be correct to make ops->get_parent() return int
    > instead of u8 (either an index or an -ERROR). However this was out of scope
    > for this patch.

    Yep, I believe that should happen, long-term.

    Best regards,
    Pavel
    --
    DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
    HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
    [unhandled content-type:application/pgp-signature]
    \
     
     \ /
      Last update: 2021-12-15 18:55    [W:4.664 / U:0.112 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site