Skip to contentSkip to conversation

CLK: microchip: Add clkcfg driver for Microchip PolarFire SoC

PATCH Superseded [email protected] 2 participants 7 messages Nov 30, 2021 – Dec 8, 2021

Subsystems (3)

v6 of 12 Superseded Latest is v12 →

lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

v6 cover CLK: microchip: Add clkcfg driver for Microchip PolarFire SoC

From: Conor Dooley <[email protected]>

This patchset adds support for the Microchip PolarFire clkcfg hardware block.

Changes since v5:
* Dropped clock-output-names property

Major changes since v4:
* Adjusted license for microchip,mpfs-clock.h to match microchip,mpfs.yaml
* Corrected the number of clocks to 33 from 32

Major changes since v3:
* Patch reformatted so microchip,mpfs-clock.h is part of device-tree patch

Major changes since v2:
* In mpfs_cfg_clk_set_rate, return immediately if divider_get_val returns <0
* rebased to v5.12-rc1

Major changes since v1:
* Dependency on SOC_MICROCHIP_POLARFIRE
* All references to PFSOC/pfsoc changed to MPFS/mpfs
* Cleaned error handling in _probe
* Re-ordered code to place structs et al at top

Daire McNamara (2):
dt-bindings: clk: microchip: Add Microchip PolarFire host binding
clk: microchip: Add driver for Microchip PolarFire SoC

 .../bindings/clock/microchip,mpfs.yaml        |  58 +++
 drivers/clk/Kconfig                           |   1 +
 drivers/clk/Makefile                          |   2 +-
 drivers/clk/microchip/Kconfig                 |   7 +
 drivers/clk/microchip/Makefile                |   6 +-
 drivers/clk/microchip/clk-mpfs.c              | 444 ++++++++++++++++++
 .../dt-bindings/clock/microchip,mpfs-clock.h  |  45 ++
 7 files changed, 560 insertions(+), 3 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/clock/microchip,mpfs.yaml
 create mode 100644 drivers/clk/microchip/Kconfig
 create mode 100644 drivers/clk/microchip/clk-mpfs.c
 create mode 100644 include/dt-bindings/clock/microchip,mpfs-clock.h
Signature · 1 line
2.33.1
author maintainer
#1 lore raw

PATCH v6 1/2 dt-bindings: clk: microchip: Add Microchip PolarFire host binding

From: Daire McNamara <[email protected]>

Add device tree bindings for the Microchip PolarFire system clock controller

---
 .../bindings/clock/microchip,mpfs.yaml        | 58 +++++++++++++++++++
 .../dt-bindings/clock/microchip,mpfs-clock.h  | 45 ++++++++++++++
 2 files changed, 103 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/clock/microchip,mpfs.yaml
 create mode 100644 include/dt-bindings/clock/microchip,mpfs-clock.h
Documentation/devicetree/bindings/clock/microchip,mpfs.yaml new file +58 -0
new file mode 100644
index 000000000000..0c15afa2214c
@@ -0,0 +1,58 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/clock/microchip,mpfs.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Microchip PolarFire Clock Control Module Binding
+
+maintainers:
+  - Daire McNamara <[email protected]>
+
+description: |
+  Microchip PolarFire clock control (CLKCFG) is an integrated clock controller,
+  which gates and enables all peripheral clocks.
+
+  This device tree binding describes 33 gate clocks.  Clocks are referenced by
+  user nodes by the CLKCFG node phandle and the clock index in the group, from
+  0 to 32.
+
+properties:
+  compatible:
+    const: microchip,mpfs-clkcfg
+
+  reg:
+    maxItems: 1
+
+  clocks:
+    maxItems: 1
+
+  '#clock-cells':
+    const: 1
+    description: |
+      The clock consumer should specify the desired clock by having the clock
+      ID in its "clocks" phandle cell. See include/dt-bindings/clock/microchip,mpfs-clock.h
+      for the full list of PolarFire clock IDs.
+
+required:
+  - compatible
+  - reg
+  - clocks
+  - '#clock-cells'
+
+additionalProperties: false
+
+examples:
+  # Clock Config node:
+  - |
+    #include <dt-bindings/clock/microchip,mpfs-clock.h>
+    soc {
+            #address-cells = <2>;
+            #size-cells = <2>;
+            clkcfg: clock-controller@20002000 {
+                compatible = "microchip,mpfs-clkcfg";
+                reg = <0x0 0x20002000 0x0 0x1000>;
+                clocks = <&ref>;
+                #clock-cells = <1>;
+        };
+    };
include/dt-bindings/clock/microchip,mpfs-clock.h new file +45 -0

45 lines not on this page · raw message

Signature · 1 line
2.33.1
Reviewed-by on v6 1/2
#2 lore raw

Re: [PATCH v6 1/2] dt-bindings: clk: microchip: Add Microchip PolarFire host binding

Quoted 8 lines · On Tue, Nov 30, 2021 at 3:06 PM <[email protected]> wrote:

From: Daire McNamara <[email protected]>

Add device tree bindings for the Microchip PolarFire system clock controller

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#3 lore raw

PATCH v6 2/2 clk: microchip: Add driver for Microchip PolarFire SoC

In reply to [email protected]

From: Daire McNamara <[email protected]>

Add support for clock configuration on Microchip PolarFire SoC

---
 drivers/clk/Kconfig              |   1 +
 drivers/clk/Makefile             |   2 +-
 drivers/clk/microchip/Kconfig    |   7 +
 drivers/clk/microchip/Makefile   |   6 +-
 drivers/clk/microchip/clk-mpfs.c | 444 +++++++++++++++++++++++++++++++
 5 files changed, 457 insertions(+), 3 deletions(-)
 create mode 100644 drivers/clk/microchip/Kconfig
 create mode 100644 drivers/clk/microchip/clk-mpfs.c
drivers/clk/Kconfig +1 -0
index c5b3dc97396a..92d112a40269 100644
@@ -401,6 +401,7 @@ source "drivers/clk/ingenic/Kconfig"
 source "drivers/clk/keystone/Kconfig"
 source "drivers/clk/mediatek/Kconfig"
 source "drivers/clk/meson/Kconfig"
+source "drivers/clk/microchip/Kconfig"
 source "drivers/clk/mstar/Kconfig"
 source "drivers/clk/mvebu/Kconfig"
 source "drivers/clk/pistachio/Kconfig"
drivers/clk/Makefile +1 -1
index e42312121e51..b60471a6061f 100644
@@ -89,7 +89,7 @@ obj-$(CONFIG_ARCH_KEYSTONE)		+= keystone/
 obj-$(CONFIG_MACH_LOONGSON32)		+= loongson1/
 obj-y					+= mediatek/
 obj-$(CONFIG_ARCH_MESON)		+= meson/
-obj-$(CONFIG_MACH_PIC32)		+= microchip/
+obj-y					+= microchip/
 ifeq ($(CONFIG_COMMON_CLK), y)
 obj-$(CONFIG_ARCH_MMP)			+= mmp/
 endif
drivers/clk/microchip/Kconfig new file +7 -0
new file mode 100644
index 000000000000..f5edc7b3c07c
@@ -0,0 +1,7 @@
+# SPDX-License-Identifier: GPL-2.0
+
+config MCHP_CLK_MPFS
+	bool "Clk driver for PolarFire SoC"
+	depends on (RISCV && SOC_MICROCHIP_POLARFIRE) || COMPILE_TEST
+	help
+	  Supports Clock Configuration for PolarFire SoC
drivers/clk/microchip/Makefile +4 -2
index f34b247e870f..0dce0b12eac4 100644
@@ -1,3 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0-only
-obj-$(CONFIG_COMMON_CLK_PIC32) += clk-core.o
-obj-$(CONFIG_PIC32MZDA) += clk-pic32mzda.o
+
+obj-$(CONFIG_COMMON_CLK_PIC32)	+= clk-core.o
+obj-$(CONFIG_PIC32MZDA)		+= clk-pic32mzda.o
+obj-$(CONFIG_MCHP_CLK_MPFS)	+= clk-mpfs.o
drivers/clk/microchip/clk-mpfs.c new file +444 -0

444 lines not on this page · raw message

Signature · 1 line
2.33.1
#4 lore raw

Re: [PATCH v6 2/2] clk: microchip: Add driver for Microchip PolarFire SoC

Hi Conor, Daire,

Quoted 6 lines · On Tue, Nov 30, 2021 at 3:06 PM <[email protected]> wrote:

From: Daire McNamara <[email protected]>

Add support for clock configuration on Microchip PolarFire SoC

Thanks for your patch!

Quoted 17 lines
drivers/clk/microchip/clk-mpfs.c new file +14 -0
@@ -0,0 +1,444 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Daire McNamara,<[email protected]>
+ * Copyright (C) 2020 Microchip Technology Inc.  All rights reserved.
+ */
+#include <linux/clk-provider.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <dt-bindings/clock/microchip,mpfs-clock.h>
+
+/* address offset of control registers */
+#define REG_CLOCK_CONFIG_CR    0x08u

drivers/clk/microchip/clk-mpfs.c line 24

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+#define REG_SUBBLK_CLOCK_CR    0x84u
+#define REG_SUBBLK_RESET_CR    0x88u
+
+struct mpfs_clock_data {
+       void __iomem *base;
+       struct clk_hw_onecell_data hw_data;
+};
+
+struct mpfs_cfg_clock {
+       unsigned int id;

implicit gap of 4 bytes

drivers/clk/microchip/clk-mpfs.c line 27

Quoted 3 lines
drivers/clk/microchip/clk-mpfs.c new file +3 -0
+       const char *name;
+       u8 shift;
+       u8 width;

implicit gap of 6 bytes

drivers/clk/microchip/clk-mpfs.c line 29

Quoted 2 lines
drivers/clk/microchip/clk-mpfs.c new file +2 -0
+       const struct clk_div_table *table;
+       unsigned long flags;

Hence you better move id, shift, and width to the end.

Quoted 5 lines
drivers/clk/microchip/clk-mpfs.c new file +5 -0
+};
+
+struct mpfs_cfg_hw_clock {
+       struct mpfs_cfg_clock cfg;
+       void __iomem *sys_base;

drivers/clk/microchip/clk-mpfs.c line 44

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+       /* lock is used to prevent multiple writes */
+       spinlock_t *lock;
+       struct clk_hw hw;
+       struct clk_init_data init;
+};
+
+#define to_mpfs_cfg_clk(_hw) container_of(_hw, struct mpfs_cfg_hw_clock, hw)
+
+struct mpfs_periph_clock {
+       unsigned int id;

implicit gap of 4 bytes

drivers/clk/microchip/clk-mpfs.c line 46

Quoted 2 lines
drivers/clk/microchip/clk-mpfs.c new file +2 -0
+       const char *name;
+       u8 shift;

implicit gap of 7 bytes

drivers/clk/microchip/clk-mpfs.c line 54

Quoted 8 lines
drivers/clk/microchip/clk-mpfs.c new file +8 -0
+       unsigned long flags;
+};
+
+struct mpfs_periph_hw_clock {
+       struct mpfs_periph_clock periph;
+       void __iomem *sys_base;
+       /* lock is used to prevent multiple writes */
+       spinlock_t *lock;

This is not used?

drivers/clk/microchip/clk-mpfs.c line 67

Quoted 13 lines
drivers/clk/microchip/clk-mpfs.c new file +13 -0
+       struct clk_hw hw;
+};
+
+#define to_mpfs_periph_clk(_hw) container_of(_hw, struct mpfs_periph_hw_clock, hw)
+
+/*
+ * mpfs_clk_lock prevents anything else from writing to the
+ * mpfs clk block while a software locked register is being written.
+ */
+static DEFINE_SPINLOCK(mpfs_clk_lock);
+
+static struct clk_parent_data mpfs_cfg_parent[] = {
+       { .fw_name = "msspllclk", .name = "msspllclk" },

So you rely on the name of the fixed-clock reference clock? That is very fragile. Please instead use devm_clk_get() to get the reference clock from the "clocks" property of the clock-controller node.

drivers/clk/microchip/clk-mpfs.c line 92

Quoted 13 lines
drivers/clk/microchip/clk-mpfs.c new file +13 -0
+static unsigned long mpfs_cfg_clk_recalc_rate(struct clk_hw *hw, unsigned long prate)
+{
+       struct mpfs_cfg_hw_clock *cfg_hw = to_mpfs_cfg_clk(hw);
+       struct mpfs_cfg_clock *cfg = &cfg_hw->cfg;
+       void __iomem *base_addr = cfg_hw->sys_base;
+       unsigned long rate;
+       u32 val;
+
+       val = readl_relaxed(base_addr + REG_CLOCK_CONFIG_CR) >> cfg->shift;
+       val &= clk_div_mask(cfg->width);
+       rate = prate / (1u << val);
+
+       return rate;

return prate / (1u << val);

Quoted 7 lines
drivers/clk/microchip/clk-mpfs.c new file +6 -0
+}
 
+static int mpfs_cfg_clk_set_rate(struct clk_hw *hw, unsigned long rate, unsigned long prate)
+{
+       struct mpfs_cfg_hw_clock *cfg_hw = to_mpfs_cfg_clk(hw);
+       struct mpfs_cfg_clock *cfg = &cfg_hw->cfg;
+       void __iomem *base_addr = cfg_hw->sys_base;

drivers/clk/microchip/clk-mpfs.c line 117

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+       unsigned long flags = 0;
+       u32 val;
+       int divider_setting;
+
+       divider_setting = divider_get_val(rate, prate, cfg->table, cfg->width, cfg_hw->cfg.flags);
+
+       if (divider_setting < 0)
+               return divider_setting;
+
+       if (cfg_hw->lock)

Isn't this branch always taken?

Quoted 13 lines
drivers/clk/microchip/clk-mpfs.c new file +13 -0
+               spin_lock_irqsave(cfg_hw->lock, flags);
+       else
+               __acquire(cfg_hw->lock);
+
+       val = readl_relaxed(base_addr + REG_CLOCK_CONFIG_CR);
+       val &= ~(clk_div_mask(cfg->width) << cfg_hw->cfg.shift);
+       val |= divider_setting << cfg->shift;
+       writel_relaxed(val, base_addr + REG_CLOCK_CONFIG_CR);
+
+       if (cfg_hw->lock)
+               spin_unlock_irqrestore(cfg_hw->lock, flags);
+       else
+               __release(cfg_hw->lock);

drivers/clk/microchip/clk-mpfs.c line 162

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +9 -0
+
+       return 0;
+}
 
+static void mpfs_clk_unregister_cfg(struct device *dev, struct clk_hw *hw)
+{
+       struct mpfs_cfg_hw_clock *cfg_hw = to_mpfs_cfg_clk(hw);
+
+       devm_clk_hw_unregister(dev, hw);
+       kfree(cfg_hw);

This is freeing a part of the big array allocated with devm_kzalloc()?

Quoted 12 lines
drivers/clk/microchip/clk-mpfs.c new file +12 -0
+}
+
+static struct clk_hw *mpfs_clk_register_cfg(struct device *dev,
+                                           struct mpfs_cfg_hw_clock *cfg_hw,
+                                           void __iomem *sys_base)
+{
+       struct clk_hw *hw;
+       int err;
+
+       cfg_hw->sys_base = sys_base;
+       cfg_hw->lock = &mpfs_clk_lock;
+

drivers/clk/microchip/clk-mpfs.c line 184

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+       hw = &cfg_hw->hw;
+       err = devm_clk_hw_register(dev, hw);
+       if (err)
+               return ERR_PTR(err);
+
+       return hw;
+}
+
+static int mpfs_clk_register_cfgs(struct device *dev, struct mpfs_cfg_hw_clock *cfg_hws,
+                                 int num_clks, struct mpfs_clock_data *data)

unsigned int num_clks

drivers/clk/microchip/clk-mpfs.c line 195

Quoted 11 lines
drivers/clk/microchip/clk-mpfs.c new file +11 -0
+{
+       struct clk_hw *hw;
+       void __iomem *sys_base = data->base;
+       unsigned int i, id;
+
+       for (i = 0; i < num_clks; i++) {
+               struct mpfs_cfg_hw_clock *cfg_hw = &cfg_hws[i];
+
+               hw = mpfs_clk_register_cfg(dev, cfg_hw, sys_base);
+               if (IS_ERR(hw)) {
+                       dev_err(dev, "%s: failed to register clock %s\n", __func__,

I guess the __func__ can be dropped, as the clock name is unique?

Quoted 13 lines
drivers/clk/microchip/clk-mpfs.c new file +13 -0
+                               cfg_hw->cfg.name);
+                       goto err_clk;
+               }
+
+               id = cfg_hws[i].cfg.id;
+               data->hw_data.hws[id] = hw;
+       }
+
+       return 0;
+
+err_clk:
+       while (i--)
+               mpfs_clk_unregister_cfg(dev, data->hw_data.hws[cfg_hws[i].cfg.id]);

drivers/clk/microchip/clk-mpfs.c

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +9 -0
+
+       return PTR_ERR(hw);
+}
 
+static void mpfs_clk_unregister_periph(struct device *dev, struct clk_hw *hw)
+{
+       struct mpfs_periph_hw_clock *periph_hw = to_mpfs_periph_clk(hw);
+
+       devm_clk_hw_unregister(dev, hw);
+       kfree(periph_hw);'

This is freeing a part of the big array allocated with devm_kzalloc()?

Quoted 5 lines
drivers/clk/microchip/clk-mpfs.c new file +4 -0
+}
 
+static int mpfs_clk_register_periphs(struct device *dev, struct mpfs_periph_hw_clock *periph_hws,
+                                    int num_clks, struct mpfs_clock_data *data)
+{

drivers/clk/microchip/clk-mpfs.c line 356

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+       struct clk_hw *hw;
+       void __iomem *sys_base = data->base;
+       unsigned int i, id;
+
+       for (i = 0; i < num_clks; i++) {
+               struct mpfs_periph_hw_clock *periph_hw = &periph_hws[i];
+
+               hw = mpfs_clk_register_periph(dev, periph_hw, sys_base);
+               if (IS_ERR(hw)) {
+                       dev_err(dev, "%s: failed to register clock %s\n", __func__,

I guess the __func__ can be dropped, as the clock name is unique?

Quoted 13 lines
drivers/clk/microchip/clk-mpfs.c new file +13 -0
+                               periph_hw->periph.name);
+                       goto err_clk;
+               }
+
+               id = periph_hws[i].periph.id;
+               data->hw_data.hws[id] = hw;
+       }
+
+       return 0;
+
+err_clk:
+       while (i--)
+               mpfs_clk_unregister_periph(dev, data->hw_data.hws[periph_hws[i].periph.id]);

drivers/clk/microchip/clk-mpfs.c line 379

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+
+       return PTR_ERR(hw);
+}
+
+static int mpfs_clk_probe(struct platform_device *pdev)
+{
+       struct device *dev = &pdev->dev;
+       struct mpfs_clock_data *clk_data;
+       struct resource *res;
+       int num_clks;

unsigned int

drivers/clk/microchip/clk-mpfs.c line 382

Quoted 3 lines
drivers/clk/microchip/clk-mpfs.c new file +3 -0
+       int ret;
+
+       num_clks = ARRAY_SIZE(mpfs_cfg_clks) + ARRAY_SIZE(mpfs_periph_clks);

As mpfs_periph_clks[] lacks an entry for CLK_RESERVED, num_clks is one too small...

Quoted 8 lines
drivers/clk/microchip/clk-mpfs.c new file +8 -0
+
+       clk_data = devm_kzalloc(dev, struct_size(clk_data, hw_data.hws, num_clks), GFP_KERNEL);
+       if (!clk_data)
+               return -ENOMEM;
+
+       res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+       clk_data->base = devm_ioremap_resource(dev, res);
+       if (IS_ERR(clk_data->base))

drivers/clk/microchip/clk-mpfs.c line 400

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+               return PTR_ERR(clk_data->base);
+
+       clk_data->hw_data.num = num_clks;
+
+       ret = mpfs_clk_register_cfgs(dev, mpfs_cfg_clks, ARRAY_SIZE(mpfs_cfg_clks), clk_data);
+       if (ret)
+               goto err_clk;
+
+       ret = mpfs_clk_register_periphs(dev, mpfs_periph_clks, ARRAY_SIZE(mpfs_periph_clks),
+                                       clk_data);

... and initializing CLK_CFM will write beyond the end of clk_data[].

Quoted 4 lines
drivers/clk/microchip/clk-mpfs.c new file +4 -0
+       if (ret)
+               goto err_clk;
+
+       ret = devm_of_clk_add_hw_provider(dev, of_clk_hw_onecell_get, &clk_data->hw_data);

drivers/clk/microchip/clk-mpfs.c line 414

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+       if (ret)
+               goto err_clk;
+
+       dev_info(dev, "registered MPFS core clocks\n");
+       return ret;
+
+err_clk:
+       dev_err(dev, "failed to register MPFS core clocks\n");
+       return ret;
+}

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#5 lore raw

Re: [PATCH v6 2/2] clk: microchip: Add driver for Microchip PolarFire SoC

Hi Geert,
Just going to ack the other items you raised for now, but curious about this one:

Quoted 9 lines · Geert Uytterhoeven

EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe

drivers/clk/microchip/clk-mpfs.c line 162

Quoted 6 lines
drivers/clk/microchip/clk-mpfs.c new file +6 -0
+static void mpfs_clk_unregister_cfg(struct device *dev, struct clk_hw *hw)
+{
+       struct mpfs_cfg_hw_clock *cfg_hw = to_mpfs_cfg_clk(hw);
+
+       devm_clk_hw_unregister(dev, hw);
+       kfree(cfg_hw);

This is freeing a part of the big array allocated with devm_kzalloc()?

I took a look at this, and I don't think it is freeing the devm allocated array.

To me, what is actually being freed is an element in the array of structs passed to mpfs_clk_register_cfgs in the probe function. However, this struct is statically defined - so its elements shouldn't be freed at all?

drivers/clk/clk-bm1880.c has the same behaviour in the unregister function, where it calls kfree on the elements of a static array of clk structs. So if my understanding is correct it would need fixing there too.

Cheers,
Conor.

Quoted 55 lines
Quoted 12 lines
drivers/clk/microchip/clk-mpfs.c new file +12 -0
+}
+
+static struct clk_hw *mpfs_clk_register_cfg(struct device *dev,
+                                           struct mpfs_cfg_hw_clock *cfg_hw,
+                                           void __iomem *sys_base)
+{
+       struct clk_hw *hw;
+       int err;
+
+       cfg_hw->sys_base = sys_base;
+       cfg_hw->lock = &mpfs_clk_lock;
+

drivers/clk/microchip/clk-mpfs.c line 184

Quoted 10 lines
drivers/clk/microchip/clk-mpfs.c new file +10 -0
+       hw = &cfg_hw->hw;
+       err = devm_clk_hw_register(dev, hw);
+       if (err)
+               return ERR_PTR(err);
+
+       return hw;
+}
+
+static int mpfs_clk_register_cfgs(struct device *dev, struct mpfs_cfg_hw_clock *cfg_hws,
+                                 int num_clks, struct mpfs_clock_data *data)

unsigned int num_clks

drivers/clk/microchip/clk-mpfs.c line 195

Quoted 11 lines
drivers/clk/microchip/clk-mpfs.c new file +11 -0
+{
+       struct clk_hw *hw;
+       void __iomem *sys_base = data->base;
+       unsigned int i, id;
+
+       for (i = 0; i < num_clks; i++) {
+               struct mpfs_cfg_hw_clock *cfg_hw = &cfg_hws[i];
+
+               hw = mpfs_clk_register_cfg(dev, cfg_hw, sys_base);
+               if (IS_ERR(hw)) {
+                       dev_err(dev, "%s: failed to register clock %s\n", __func__,

I guess the __func__ can be dropped, as the clock name is unique?

Quoted 16 lines
drivers/clk/microchip/clk-mpfs.c new file +16 -0
+                               cfg_hw->cfg.name);
+                       goto err_clk;
+               }
+
+               id = cfg_hws[i].cfg.id;
+               data->hw_data.hws[id] = hw;
+       }
+
+       return 0;
+
+err_clk:
+       while (i--)
+               mpfs_clk_unregister_cfg(dev, data->hw_data.hws[cfg_hws[i].cfg.id]);
+
+       return PTR_ERR(hw);
+}
#6 lore raw

Re: [PATCH v6 2/2] clk: microchip: Add driver for Microchip PolarFire SoC

Hi Conor,

Quoted 12 lines · On Wed, Dec 8, 2021 at 4:30 PM <[email protected]> wrote:

Just going to ack the other items you raised for now, but curious about this one:

Quoted 9 lines · Geert Uytterhoeven

EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe

drivers/clk/microchip/clk-mpfs.c line 162

Quoted 6 lines
drivers/clk/microchip/clk-mpfs.c new file +6 -0
+static void mpfs_clk_unregister_cfg(struct device *dev, struct clk_hw *hw)
+{
+       struct mpfs_cfg_hw_clock *cfg_hw = to_mpfs_cfg_clk(hw);
+
+       devm_clk_hw_unregister(dev, hw);
+       kfree(cfg_hw);

This is freeing a part of the big array allocated with devm_kzalloc()?

Quoted 7 lines

I took a look at this, and I don't think it is freeing the devm allocated array.

To me, what is actually being freed is an element in the array of structs passed to mpfs_clk_register_cfgs in the probe function. However, this struct is statically defined - so its elements shouldn't be freed at all?

You're right, it's part of the static array. I was misled by the back-and-forth conversions between the different types. Regardless, the kfree() calls should be removed.

Quoted 3 lines

drivers/clk/clk-bm1880.c has the same behaviour in the unregister function, where it calls kfree on the elements of a static array of clk structs. So if my understanding is correct it would need fixing there too.

If that is true (didn't check), it needs to be fixed, too.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds