- Various clk rate range fixes
- Drop clk rate range constraints on clk_put() (redux)
* clk-rate-range: (28 commits)
clk: mediatek: clk-mux: Add .determine_rate() callback
clk: tests: Add tests for notifiers
clk: Update req_rate on __clk_recalc_rates()
clk: tests: Add missing test case for ranges
clk: qcom: clk-rcg2: Take clock boundaries into consideration for gfx3d
clk: Introduce the clk_hw_get_rate_range function
clk: Zero the clk_rate_request structure
clk: Stop forwarding clk_rate_requests to the parent
clk: Constify clk_has_parent()
clk: Introduce clk_core_has_parent()
clk: Switch from __clk_determine_rate to clk_core_round_rate_nolock
clk: Add our request boundaries in clk_core_init_rate_req
clk: Introduce clk_hw_init_rate_request()
clk: Move clk_core_init_rate_req() from clk_core_round_rate_nolock() to its caller
clk: Change clk_core_init_rate_req prototype
clk: Set req_rate on reparenting
clk: Take into account uncached clocks in clk_set_rate_range()
clk: tests: Add some tests for orphan with multiple parents
clk: tests: Add tests for mux with multiple parents
clk: tests: Add tests for single parent mux
...
Since commit 262ca38f4b ("clk: Stop forwarding clk_rate_requests
to the parent"), the clk_rate_request is .. as the title says, not
forwarded anymore to the parent: this produces an issue with the
MediaTek clock MUX driver during GPU DVFS on MT8195, but not on
MT8192 or others.
This is because, differently from others, like MT8192 where all of
the clocks in the MFG parents tree are of mtk_mux type, but in the
parent tree of MT8195's MFG clock, we have one mtk_mux clock and
one (clk framework generic) mux clock, like so:
names: mfg_bg3d -> mfg_ck_fast_ref -> top_mfg_core_tmp (or) mfgpll
types: mtk_gate -> mux -> mtk_mux (or) mtk_pll
To solve this issue and also keep the GPU DVFS clocks code working
as expected, wire up a .determine_rate() callback for the mtk_mux
ops; for that, the standard clk_mux_determine_rate_flags() was used
as it was possible to.
This commit was successfully tested on MT6795 Xperia M5, MT8173 Elm,
MT8192 Spherion and MT8195 Tomato; no regressions were seen.
For the sake of some more documentation about this issue here's the
trace of it:
[ 12.211587] ------------[ cut here ]------------
[ 12.211589] WARNING: CPU: 6 PID: 78 at drivers/clk/clk.c:1462 clk_core_init_rate_req+0x84/0x90
[ 12.211593] Modules linked in: stp crct10dif_ce mtk_adsp_common llc rfkill snd_sof_xtensa_dsp
panfrost(+) sbs_battery cros_ec_lid_angle cros_ec_sensors snd_sof_of
cros_ec_sensors_core hid_multitouch cros_usbpd_logger snd_sof gpu_sched
snd_sof_utils fuse ipv6
[ 12.211614] CPU: 6 PID: 78 Comm: kworker/u16:2 Tainted: G W 6.0.0-next-20221011+ #58
[ 12.211616] Hardware name: Acer Tomato (rev2) board (DT)
[ 12.211617] Workqueue: devfreq_wq devfreq_monitor
[ 12.211620] pstate: 40400009 (nZcv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[ 12.211622] pc : clk_core_init_rate_req+0x84/0x90
[ 12.211625] lr : clk_core_forward_rate_req+0xa4/0xe4
[ 12.211627] sp : ffff80000893b8e0
[ 12.211628] x29: ffff80000893b8e0 x28: ffffdddf92f9b000 x27: ffff46a2c0e8bc05
[ 12.211632] x26: ffff46a2c1041200 x25: 0000000000000000 x24: 00000000173eed80
[ 12.211636] x23: ffff80000893b9c0 x22: ffff80000893b940 x21: 0000000000000000
[ 12.211641] x20: ffff46a2c1039f00 x19: ffff46a2c1039f00 x18: 0000000000000000
[ 12.211645] x17: 0000000000000038 x16: 000000000000d904 x15: 0000000000000003
[ 12.211649] x14: ffffdddf9357ce48 x13: ffffdddf935e71c8 x12: 000000000004803c
[ 12.211653] x11: 00000000a867d7ad x10: 00000000a867d7ad x9 : ffffdddf90c28df4
[ 12.211657] x8 : ffffdddf9357a980 x7 : 0000000000000000 x6 : 0000000000000004
[ 12.211661] x5 : ffffffffffffffc8 x4 : 00000000173eed80 x3 : ffff80000893b940
[ 12.211665] x2 : 00000000173eed80 x1 : ffff80000893b940 x0 : 0000000000000000
[ 12.211669] Call trace:
[ 12.211670] clk_core_init_rate_req+0x84/0x90
[ 12.211673] clk_core_round_rate_nolock+0xe8/0x10c
[ 12.211675] clk_mux_determine_rate_flags+0x174/0x1f0
[ 12.211677] clk_mux_determine_rate+0x1c/0x30
[ 12.211680] clk_core_determine_round_nolock+0x74/0x130
[ 12.211682] clk_core_round_rate_nolock+0x58/0x10c
[ 12.211684] clk_core_round_rate_nolock+0xf4/0x10c
[ 12.211686] clk_core_set_rate_nolock+0x194/0x2ac
[ 12.211688] clk_set_rate+0x40/0x94
[ 12.211691] _opp_config_clk_single+0x38/0xa0
[ 12.211693] _set_opp+0x1b0/0x500
[ 12.211695] dev_pm_opp_set_rate+0x120/0x290
[ 12.211697] panfrost_devfreq_target+0x3c/0x50 [panfrost]
[ 12.211705] devfreq_set_target+0x8c/0x2d0
[ 12.211707] devfreq_update_target+0xcc/0xf4
[ 12.211708] devfreq_monitor+0x40/0x1d0
[ 12.211710] process_one_work+0x294/0x664
[ 12.211712] worker_thread+0x7c/0x45c
[ 12.211713] kthread+0x104/0x110
[ 12.211716] ret_from_fork+0x10/0x20
[ 12.211718] irq event stamp: 7102
[ 12.211719] hardirqs last enabled at (7101): [<ffffdddf904ea5a0>] finish_task_switch.isra.0+0xec/0x2f0
[ 12.211723] hardirqs last disabled at (7102): [<ffffdddf91794b74>] el1_dbg+0x24/0x90
[ 12.211726] softirqs last enabled at (6716): [<ffffdddf90410be4>] __do_softirq+0x414/0x588
[ 12.211728] softirqs last disabled at (6507): [<ffffdddf904171d8>] ____do_softirq+0x18/0x24
[ 12.211730] ---[ end trace 0000000000000000 ]---
Fixes: 262ca38f4b ("clk: Stop forwarding clk_rate_requests to the parent")
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20221011135548.318323-1-angelogioacchino.delregno@collabora.com
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
With device frequency scaling, the mux clock that (indirectly) feeds the
device selects between a dedicated PLL, and some other stable clocks.
When a clk rate change is requested, the (normally) upstream PLL is
reconfigured. It's possible for the clock output of the PLL to become
unstable during this process.
To avoid causing the device to glitch, the mux should temporarily be
switched over to another "stable" clock during the PLL rate change.
This is done with clk notifiers.
This patch adds common functions for notifiers to temporarily and
transparently reparent mux clocks.
This was loosely based on commit 8adfb08605 ("clk: sunxi-ng: mux: Add
clk notifier functions").
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
[Angelo: Changed mtk_mux_nb to hold a pointer to clk_ops instead of mtk_mux]
Co-developed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Link: https://lore.kernel.org/r/20220927101128.44758-4-angelogioacchino.delregno@collabora.com
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
As part of the effort to improve the MediaTek clk drivers, the next step
is to switch from the old 'struct clk' clk prodivder APIs to the new
'struct clk_hw' ones.
In a previous patch, 'struct clk_onecell_data' was replaced with
'struct clk_hw_onecell_data', with (struct clk_hw *)->clk and
__clk_get_hw() bridging the new data structures and old code.
Now switch from the old 'clk_(un)?register*()' APIs to the new
'clk_hw_(un)?register*()' ones. This is done with the coccinelle script
below.
Unfortunately this also leaves clk-mt8173.c with a compile error that
would need a coccinelle script longer than the actual diff to fix. This
last part is fixed up by hand.
// Fix prototypes
@@
identifier F =~ "^mtk_clk_register_";
@@
- struct clk *
+ struct clk_hw *
F(...);
// Fix calls to mtk_clk_register_<singular>
@ reg @
identifier F =~ "^mtk_clk_register_";
identifier FS =~ "^mtk_clk_register_[a-z_]*s";
identifier I;
expression clk_data;
expression E;
@@
FS(...) {
...
- struct clk *I;
+ struct clk_hw *hw;
...
for (...;...;...) {
...
(
- I
+ hw
=
- clk_register_fixed_rate(
+ clk_hw_register_fixed_rate(
...
);
|
- I
+ hw
=
- clk_register_fixed_factor(
+ clk_hw_register_fixed_factor(
...
);
|
- I
+ hw
=
- clk_register_divider(
+ clk_hw_register_divider(
...
);
|
- I
+ hw
=
F(...);
)
...
if (
- IS_ERR(I)
+ IS_ERR(hw)
) {
pr_err(...,
- I
+ hw
,...);
...
}
- clk_data->hws[E] = __clk_get_hw(I);
+ clk_data->hws[E] = hw;
}
...
}
@ depends on reg @
identifier reg.I;
@@
return PTR_ERR(
- I
+ hw
);
// Fix mtk_clk_register_composite to return clk_hw instead of clk
@@
identifier I, R;
expression E;
@@
- struct clk *
+ struct clk_hw *
mtk_clk_register_composite(...) {
...
- struct clk *I;
+ struct clk_hw *hw;
...
- I = clk_register_composite(
+ hw = clk_hw_register_composite(
...);
if (IS_ERR(
- I
+ hw
)) {
...
R = PTR_ERR(
- I
+ hw
);
...
}
return
- I
+ hw
;
...
}
// Fix other mtk_clk_register_<singular> to return clk_hw instead of clk
@@
identifier F =~ "^mtk_clk_register_";
identifier I, D, C;
expression E;
@@
- struct clk *
+ struct clk_hw *
F(...) {
...
- struct clk *I;
+ int ret;
...
- I = clk_register(D, E);
+ ret = clk_hw_register(D, E);
...
(
- if (IS_ERR(I))
+ if (ret) {
kfree(C);
+ return ERR_PTR(ret);
+ }
|
- if (IS_ERR(I))
+ if (ret)
{
kfree(C);
- return I;
+ return ERR_PTR(ret);
}
)
- return I;
+ return E;
}
// Fix mtk_clk_unregister_<singular> to take clk_hw instead of clk
@@
identifier F =~ "^mtk_clk_unregister_";
identifier I, I2;
@@
static void F(
- struct clk *I
+ struct clk_hw *I2
)
{
...
- struct clk_hw *I2;
...
- I2 = __clk_get_hw(I);
...
(
- clk_unregister(I);
+ clk_hw_unregister(I2);
|
- clk_unregister_composite(I);
+ clk_hw_unregister_composite(I2);
)
...
}
// Fix calls to mtk_clk_unregister_*()
@@
identifier F =~ "^mtk_clk_unregister_";
expression I;
expression E;
@@
- F(I->hws[E]->clk);
+ F(I->hws[E]);
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: Miles Chen <miles.chen@mediatek.com>
Link: https://lore.kernel.org/r/20220519071610.423372-5-wenst@chromium.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
As part of the effort to improve the MediaTek clk drivers, the next step
is to switch from the old 'struct clk' clk prodivder APIs to the new
'struct clk_hw' ones.
Instead of adding new APIs to the MediaTek clk driver library mirroring
the existing ones, moving all drivers to the new APIs, and then removing
the old ones, just migrate everything at the same time. This involves
replacing 'struct clk' with 'struct clk_hw', and 'struct clk_onecell_data'
with 'struct clk_hw_onecell_data', and fixing up all usages.
For now, the clk_register() and co. usage is retained, with __clk_get_hw()
and (struct clk_hw *)->clk used to bridge the difference between the APIs.
These will be replaced in subsequent patches.
Fix up mtk_{alloc,free}_clk_data to use 'struct clk_hw' by hand. Fix up
all other affected call sites with the following coccinelle script.
// Replace type
@@
@@
- struct clk_onecell_data
+ struct clk_hw_onecell_data
// Replace of_clk_add_provider() & of_clk_src_simple_get()
@@
expression NP, DATA;
symbol of_clk_src_onecell_get;
@@
- of_clk_add_provider(
+ of_clk_add_hw_provider(
NP,
- of_clk_src_onecell_get,
+ of_clk_hw_onecell_get,
DATA
)
// Fix register/unregister
@@
identifier CD;
expression E;
identifier fn =~ "unregister";
@@
fn(...,
- CD->clks[E]
+ CD->hws[E]->clk
,...
);
// Fix calls to clk_prepare_enable()
@@
identifier CD;
expression E;
@@
clk_prepare_enable(
- CD->clks[E]
+ CD->hws[E]->clk
);
// Fix pointer assignment
@@
identifier CD;
identifier CLK;
expression E;
@@
- CD->clks[E]
+ CD->hws[E]
=
(
- CLK
+ __clk_get_hw(CLK)
|
ERR_PTR(...)
)
;
// Fix pointer usage
@@
identifier CD;
expression E;
@@
- CD->clks[E]
+ CD->hws[E]
// Fix mtk_clk_pll_get_base()
@@
symbol clk, hw, data;
@@
mtk_clk_pll_get_base(
- struct clk *clk,
+ struct clk_hw *hw,
const struct mtk_pll_data *data
) {
- struct clk_hw *hw = __clk_get_hw(clk);
...
}
// Fix mtk_clk_pll_get_base() usage
@@
identifier CD;
expression E;
@@
mtk_clk_pll_get_base(
- CD->clks[E]
+ CD->hws[E]->clk
,...
);
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Tested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Tested-by: Miles Chen <miles.chen@mediatek.com>
Link: https://lore.kernel.org/r/20220519071610.423372-4-wenst@chromium.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
The Mediatek clk driver library handles duplicate clock IDs in two
different ways: either ignoring the duplicate entry, or overwriting
the old clk. Either way may cause unexpected behavior, and the latter
also causes an orphan clk that cannot be cleaned up.
Align the behavior so that later duplicate entries are ignored, and
a warning printed. The warning will also aid in making the issue
noticeable.
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20220208124034.414635-32-wenst@chromium.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
The mux clk type registration function does not stop or return errors
if any clk failed to be registered, nor does it implement an error
handling path. This may result in a partially working device if any
step failed.
Make the register function return proper error codes, and bail out if
errors occur. Proper cleanup, i.e. unregister any clks that were
successfully registered, is done in the new error path.
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20220208124034.414635-25-wenst@chromium.org
Reviewed-by: Chun-Jie Chen <chun-jie.chen@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
The clk registration code here currently does:
if (IS_ERR_OR_NULL(clk_data->clks[mux->id])) {
... do clk registration ...
}
This extra level of nesting wastes screen real estate.
Reduce the nesting level by reversing the conditional shown above.
Other than that, functionality is not changed.
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20220208124034.414635-24-wenst@chromium.org
Reviewed-by: Chun-Jie Chen <chun-jie.chen@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Some included headers aren't actually used anywhere, while other headers
with the declaration of functions and structures aren't directly
included.
Get rid of the unused ones, and add the ones that should be included
directly.
On the header side, replace headers that are included purely for data
structure definitions with forward declarations. This decreases the
amount of preprocessing and compilation effort required for each
inclusion.
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20220208124034.414635-12-wenst@chromium.org
Reviewed-by: Chun-Jie Chen <chun-jie.chen@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
struct mtk_clk_mux is an implementation detail of the mux clk type,
and is not used outside of the implementation.
Internalize the definition to minimize leakage of details and shrink
the header file.
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20220208124034.414635-11-wenst@chromium.org
Reviewed-by: Chun-Jie Chen <chun-jie.chen@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
The mux clk type within the MediaTek clk driver library only has a
register function, and no corresponding unregister function. This
means there is no way for its users to properly implement cleanup
and removal.
Add a matching unregister function for the mux type clk.
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20220208124034.414635-10-wenst@chromium.org
Reviewed-by: Chun-Jie Chen <chun-jie.chen@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
If %pe is used to print errors, a string representation of the error
would be printed instead of a number as with %ld. Also, all the sites
printing errors are deriving the error code from a pointer. Using %pe
is more straightforward.
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
Reviewed-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20220208124034.414635-2-wenst@chromium.org
Reviewed-by: Chun-Jie Chen <chun-jie.chen@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
modpost complains once these drivers become modules.
ERROR: modpost: "mtk_mux_gate_clr_set_upd_ops" [drivers/clk/mediatek/clk-mt6779.ko] undefined!
Let's just export them.
Cc: Hanks Chen <hanks.chen@mediatek.com>
Cc: Wendell Lin <wendell.lin@mediatek.com>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: Miles Chen <miles.chen@mediatek.com>
Fixes: 32b028fb1d ("clk: mediatek: support COMMON_CLK_MEDIATEK module build")
Link: https://lore.kernel.org/r/20210915015540.1732014-1-sboyd@kernel.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
To support COMMON_CLK_MEDIATEK module build,
add MODULE_LICENSE and export necessary symbols.
Cc: Stephen Boyd <sboyd@kernel.org>
Cc: Hanks Chen <hanks.chen@mediatek.com>
Cc: Wendell Lin <wendell.lin@mediatek.com>
Cc: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Miles Chen <miles.chen@mediatek.com>
Link: https://lore.kernel.org/r/20210901222526.31065-3-miles.chen@mediatek.com
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Not all clock providers need to be marked compatible with "syscon"
for system configuration usage, so use device_node_to_regmap() to
skip "syscon" check.
Signed-off-by: Chun-Jie Chen <chun-jie.chen@mediatek.com>
Link: https://lore.kernel.org/r/20210726105719.15793-6-chun-jie.chen@mediatek.com
Reviewed-by: Ikjoon Jang <ikjn@chromium.org>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
The mux clocks don't always correctly take the new parent into account
when the parent is updated while the clock is disabled. Set the update
bit when enabling the clock to force an update of the mux.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://lore.kernel.org/r/20210125170819.26130-3-laurent.pinchart@ideasonboard.com
Reviewed-by: Weiyi Lu <weiyi.lu@mediatek.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
When some new clock supports are introduced, e.g. [1]
it might lead to an error although it should be NULL because
clk_init_data is on the stack and it might have random values
if using without initialization.
Add the missing initial value to clk_init_data.
[1] https://android-review.googlesource.com/c/kernel/common/+/1278046
Fixes: a3ae549917 ("clk: mediatek: Add new clkmux register API")
Signed-off-by: Weiyi Lu <weiyi.lu@mediatek.com>
Reviewed-by: Matthias Brugger <matthias.bgg@gmail.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/1590560749-29136-1-git-send-email-weiyi.lu@mediatek.com
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
On both MT8183 & MT6765, there add "set/clr" register for
each clkmux setting, and one update register to trigger value change.
It is designed to prevent read-modify-write racing issue.
The sw design need to add a new API to handle this hw change with
a new mtk_clk_mux/mtk_mux struct in new file "clk-mux.c", "clk-mux.h".
Signed-off-by: Owen Chen <owen.chen@mediatek.com>
Signed-off-by: Weiyi Lu <weiyi.lu@mediatek.com>
Reviewed-by: James Liao <jamesjj.liao@mediatek.com>
Reviewed-by: Nicolas Boichat <drinkcat@chromium.org>
Tested-by: Nicolas Boichat <drinkcat@chromium.org>
[sboyd@kernel.org: Squash in flags=0 to silence warning]
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
