diff mbox series

[v6,1/1] i2c: lpi2c: cache peripheral clock rate

Message ID 20231017135436.1241650-1-alexander.stein@ew.tq-group.com (mailing list archive)
State New, archived
Headers show
Series [v6,1/1] i2c: lpi2c: cache peripheral clock rate | expand

Commit Message

Alexander Stein Oct. 17, 2023, 1:54 p.m. UTC
From: Alexander Sverdlin <alexander.sverdlin@siemens.com>

One of the reasons to do it is to save some CPU cycles on cpu_freq_get()
under mutex. The second reason if the (false-positive) lockdep splat caused
by the recursive feature of the "prepare_lock" (one clock instance is I2C
peripheral clock and another is pcf85063 RTC as clock provider):

======================================================
WARNING: possible circular locking dependency detected
5.15.71+... #1 Tainted: G           O
------------------------------------------------------
.../2332 is trying to acquire lock:
ffff8000096cae08 (prepare_lock){+.+.}-{3:3}, at: clk_prepare_lock+0x50/0xb0

but task is already holding lock:
ffff000011021100 (i2c_register_adapter){+.+.}-{3:3}, at: i2c_adapter_lock_bus+0x2c/0x3c

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #2 (i2c_register_adapter){+.+.}-{3:3}:
       lock_acquire+0x68/0x8c
       rt_mutex_lock_nested+0x88/0xe0
       i2c_adapter_lock_bus+0x2c/0x3c
       i2c_transfer+0x58/0x130
       regmap_i2c_read+0x64/0xb0
       _regmap_raw_read+0x114/0x440
       _regmap_bus_read+0x4c/0x84
       _regmap_read+0x6c/0x270
       regmap_read+0x54/0x80
       pcf85063_probe+0xec/0x4cc
       i2c_device_probe+0x10c/0x350
       really_probe+0xc4/0x470
       __driver_probe_device+0x11c/0x190
       driver_probe_device+0x48/0x110
       __device_attach_driver+0xc4/0x160
       bus_for_each_drv+0x80/0xe0
       __device_attach+0xb0/0x1f0
       device_initial_probe+0x1c/0x2c
       bus_probe_device+0xa4/0xb0
       device_add+0x398/0x8ac
       device_register+0x28/0x40
       i2c_new_client_device+0x144/0x290
       of_i2c_register_devices+0x18c/0x230
       i2c_register_adapter+0x1dc/0x6b0
       __i2c_add_numbered_adapter+0x68/0xbc
       i2c_add_adapter+0xb0/0xe0
       lpi2c_imx_probe+0x354/0x5e0
       platform_probe+0x70/0xec
       really_probe+0xc4/0x470
       __driver_probe_device+0x11c/0x190
       driver_probe_device+0x48/0x110
       __device_attach_driver+0xc4/0x160
       bus_for_each_drv+0x80/0xe0
       __device_attach+0xb0/0x1f0
       device_initial_probe+0x1c/0x2c
       bus_probe_device+0xa4/0xb0
       deferred_probe_work_func+0xa0/0xfc
       process_one_work+0x2ac/0x6f4
       worker_thread+0x7c/0x47c
       kthread+0x150/0x16c
       ret_from_fork+0x10/0x20

-> #1 (rtc_pcf85063:560:(&config->regmap)->lock){+.+.}-{3:3}:
       lock_acquire+0x68/0x8c
       __mutex_lock+0x9c/0x4d0
       mutex_lock_nested+0x48/0x5c
       regmap_lock_mutex+0x1c/0x30
       regmap_read+0x44/0x80
       pcf85063_clkout_recalc_rate+0x34/0x80
       __clk_register+0x520/0x880
       devm_clk_register+0x64/0xc4
       pcf85063_probe+0x24c/0x4cc
       i2c_device_probe+0x10c/0x350
       really_probe+0xc4/0x470
       __driver_probe_device+0x11c/0x190
       driver_probe_device+0x48/0x110
       __device_attach_driver+0xc4/0x160
       bus_for_each_drv+0x80/0xe0
       __device_attach+0xb0/0x1f0
       device_initial_probe+0x1c/0x2c
       bus_probe_device+0xa4/0xb0
       device_add+0x398/0x8ac
       device_register+0x28/0x40
       i2c_new_client_device+0x144/0x290
       of_i2c_register_devices+0x18c/0x230
       i2c_register_adapter+0x1dc/0x6b0
       __i2c_add_numbered_adapter+0x68/0xbc
       i2c_add_adapter+0xb0/0xe0
       lpi2c_imx_probe+0x354/0x5e0
       platform_probe+0x70/0xec
       really_probe+0xc4/0x470
       __driver_probe_device+0x11c/0x190
       driver_probe_device+0x48/0x110
       __device_attach_driver+0xc4/0x160
       bus_for_each_drv+0x80/0xe0
       __device_attach+0xb0/0x1f0
       device_initial_probe+0x1c/0x2c
       bus_probe_device+0xa4/0xb0
       deferred_probe_work_func+0xa0/0xfc
       process_one_work+0x2ac/0x6f4
       worker_thread+0x7c/0x47c
       kthread+0x150/0x16c
       ret_from_fork+0x10/0x20

-> #0 (prepare_lock){+.+.}-{3:3}:
       __lock_acquire+0x1298/0x20d0
       lock_acquire.part.0+0xf0/0x250
       lock_acquire+0x68/0x8c
       __mutex_lock+0x9c/0x4d0
       mutex_lock_nested+0x48/0x5c
       clk_prepare_lock+0x50/0xb0
       clk_get_rate+0x28/0x80
       lpi2c_imx_xfer+0xb0/0xa9c
       __i2c_transfer+0x174/0xa80
       i2c_transfer+0x68/0x130
       regmap_i2c_read+0x64/0xb0
       _regmap_raw_read+0x114/0x440
       regmap_raw_read+0x19c/0x28c
       regmap_bulk_read+0x1b8/0x244
       at24_read+0x14c/0x2c4
       nvmem_reg_read+0x2c/0x54
       bin_attr_nvmem_read+0x8c/0xbc
       sysfs_kf_bin_read+0x74/0x94
       kernfs_fop_read_iter+0xb0/0x1d0
       new_sync_read+0xf0/0x184
       vfs_read+0x154/0x1f0
       ksys_read+0x70/0x100
       __arm64_sys_read+0x24/0x30
       invoke_syscall+0x50/0x120
       el0_svc_common.constprop.0+0x68/0x124
       do_el0_svc+0x30/0x9c
       el0_svc+0x54/0x110
       el0t_64_sync_handler+0xa4/0x130
       el0t_64_sync+0x1a0/0x1a4

other info that might help us debug this:

Chain exists of:
  prepare_lock --> rtc_pcf85063:560:(&config->regmap)->lock --> i2c_register_adapter

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(i2c_register_adapter);
                               lock(rtc_pcf85063:560:(&config->regmap)->lock);
                               lock(i2c_register_adapter);
  lock(prepare_lock);

 *** DEADLOCK ***

4 locks held by .../2332:
 #0: ffff0000146eb288 (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_read_iter+0x74/0x1d0
 #1: ffff000010fe4400 (kn->active#72){.+.+}-{0:0}, at: kernfs_fop_read_iter+0x7c/0x1d0
 #2: ffff0000110168e8 (&at24->lock){+.+.}-{3:3}, at: at24_read+0x8c/0x2c4
 #3: ffff000011021100 (i2c_register_adapter){+.+.}-{3:3}, at: i2c_adapter_lock_bus+0x2c/0x3c

stack backtrace:
CPU: 1 PID: 2332 Comm: ... Tainted: G           O      5.15.71+... #1
Hardware name: ... (DT)
Call trace:
 dump_backtrace+0x0/0x1d4
 show_stack+0x20/0x2c
 dump_stack_lvl+0x8c/0xb8
 dump_stack+0x18/0x34
 print_circular_bug+0x1f8/0x200
 check_noncircular+0x130/0x144
 __lock_acquire+0x1298/0x20d0
 lock_acquire.part.0+0xf0/0x250
 lock_acquire+0x68/0x8c
 __mutex_lock+0x9c/0x4d0
 mutex_lock_nested+0x48/0x5c
 clk_prepare_lock+0x50/0xb0
 clk_get_rate+0x28/0x80
 lpi2c_imx_xfer+0xb0/0xa9c
 __i2c_transfer+0x174/0xa80
 i2c_transfer+0x68/0x130
 regmap_i2c_read+0x64/0xb0
 _regmap_raw_read+0x114/0x440
 regmap_raw_read+0x19c/0x28c
 regmap_bulk_read+0x1b8/0x244
 at24_read+0x14c/0x2c4
 nvmem_reg_read+0x2c/0x54
 bin_attr_nvmem_read+0x8c/0xbc
 sysfs_kf_bin_read+0x74/0x94
 kernfs_fop_read_iter+0xb0/0x1d0
 new_sync_read+0xf0/0x184
 vfs_read+0x154/0x1f0
 ksys_read+0x70/0x100
 __arm64_sys_read+0x24/0x30
 invoke_syscall+0x50/0x120
 el0_svc_common.constprop.0+0x68/0x124
 do_el0_svc+0x30/0x9c
 el0_svc+0x54/0x110
 el0t_64_sync_handler+0xa4/0x130
 el0t_64_sync+0x1a0/0x1a4

Fixes: a55fa9d0e42e ("i2c: imx-lpi2c: add low power i2c bus driver")
Signed-off-by: Alexander Sverdlin <alexander.sverdlin@siemens.com>
Reviewed-by: Alexander Stein <alexander.stein@ew.tq-group.com>
Signed-off-by: Alexander Stein <alexander.stein@ew.tq-group.com>
---
This is a v6 picking up Alexander Sverdlin's patch. Up to v5 is done by
Alexander Sverdlin.
I'm not sure if in this case both my R-b and S-o-b are necessary/desired.

Changelog:
v6: - rebased to next-20231017
    - Just replaced clk_get_rate with atomic_read to shrink the diff
v5: rebased onto Wolfram's i2c/for-next
v4: switched to atomic_t
    included clk_rate_exclusive_get()/clk_rate_exclusive_put()
v3: fixed build error reported by kernel test robot <lkp@intel.com>
    https://lore.kernel.org/oe-kbuild-all/202303102010.pAv56wKs-lkp@intel.com/
v2: added clk_notifier as Alexander suggested

 drivers/i2c/busses/i2c-imx-lpi2c.c | 41 +++++++++++++++++++++++++++++-
 1 file changed, 40 insertions(+), 1 deletion(-)

Comments

Andi Shyti Oct. 18, 2023, 4:24 a.m. UTC | #1
Hi Alexander,

On Tue, Oct 17, 2023 at 03:54:36PM +0200, Alexander Stein wrote:
> From: Alexander Sverdlin <alexander.sverdlin@siemens.com>
> 
> One of the reasons to do it is to save some CPU cycles on cpu_freq_get()

do what and how did you do it?

> under mutex. The second reason if the (false-positive) lockdep splat caused
> by the recursive feature of the "prepare_lock" (one clock instance is I2C
> peripheral clock and another is pcf85063 RTC as clock provider):

...

> +	/*
> +	 * Lock the clock rate to avoid rate change between clk_get_rate() and
> +	 * atomic_set()
> +	 */
> +	ret = clk_rate_exclusive_get(lpi2c_imx->clks[0].clk);
> +	if (ret) {
> +		dev_err(&pdev->dev, "can't lock I2C peripheral clock rate\n");
> +		return ret;

return dev_err_probe() ?

> +	}
> +	atomic_set(&lpi2c_imx->rate_per, clk_get_rate(lpi2c_imx->clks[0].clk));
> +	clk_rate_exclusive_put(lpi2c_imx->clks[0].clk);
> +	if (!atomic_read(&lpi2c_imx->rate_per)) {
> +		dev_err(&pdev->dev, "can't get I2C peripheral clock rate\n");

return dev_err_probe() ?

> +		return -EINVAL;
> +	}
diff mbox series

Patch

diff --git a/drivers/i2c/busses/i2c-imx-lpi2c.c b/drivers/i2c/busses/i2c-imx-lpi2c.c
index 678b30e90492a..e04faa261e7b1 100644
--- a/drivers/i2c/busses/i2c-imx-lpi2c.c
+++ b/drivers/i2c/busses/i2c-imx-lpi2c.c
@@ -5,6 +5,7 @@ 
  * Copyright 2016 Freescale Semiconductor, Inc.
  */
 
+#include <linux/atomic.h>
 #include <linux/clk.h>
 #include <linux/completion.h>
 #include <linux/delay.h>
@@ -99,6 +100,8 @@  struct lpi2c_imx_struct {
 	__u8			*rx_buf;
 	__u8			*tx_buf;
 	struct completion	complete;
+	struct notifier_block	clk_change_nb;
+	atomic_t		rate_per;
 	unsigned int		msglen;
 	unsigned int		delivered;
 	unsigned int		block_data;
@@ -197,6 +200,20 @@  static void lpi2c_imx_stop(struct lpi2c_imx_struct *lpi2c_imx)
 	} while (1);
 }
 
+static int lpi2c_imx_clk_change_cb(struct notifier_block *nb,
+				   unsigned long action, void *data)
+{
+	struct clk_notifier_data *ndata = data;
+	struct lpi2c_imx_struct *lpi2c_imx = container_of(nb,
+							  struct lpi2c_imx_struct,
+							  clk_change_nb);
+
+	if (action & POST_RATE_CHANGE)
+		atomic_set(&lpi2c_imx->rate_per, ndata->new_rate);
+
+	return NOTIFY_OK;
+}
+
 /* CLKLO = I2C_CLK_RATIO * CLKHI, SETHOLD = CLKHI, DATAVD = CLKHI/2 */
 static int lpi2c_imx_config(struct lpi2c_imx_struct *lpi2c_imx)
 {
@@ -207,7 +224,7 @@  static int lpi2c_imx_config(struct lpi2c_imx_struct *lpi2c_imx)
 
 	lpi2c_imx_set_mode(lpi2c_imx);
 
-	clk_rate = clk_get_rate(lpi2c_imx->clks[0].clk);
+	clk_rate = atomic_read(&lpi2c_imx->rate_per);
 	if (!clk_rate)
 		return -EINVAL;
 
@@ -590,6 +607,28 @@  static int lpi2c_imx_probe(struct platform_device *pdev)
 	if (ret)
 		return ret;
 
+	lpi2c_imx->clk_change_nb.notifier_call = lpi2c_imx_clk_change_cb;
+	ret = devm_clk_notifier_register(&pdev->dev, lpi2c_imx->clks[0].clk,
+					 &lpi2c_imx->clk_change_nb);
+	if (ret)
+		return dev_err_probe(&pdev->dev, ret,
+				     "can't register peripheral clock notifier\n");
+	/*
+	 * Lock the clock rate to avoid rate change between clk_get_rate() and
+	 * atomic_set()
+	 */
+	ret = clk_rate_exclusive_get(lpi2c_imx->clks[0].clk);
+	if (ret) {
+		dev_err(&pdev->dev, "can't lock I2C peripheral clock rate\n");
+		return ret;
+	}
+	atomic_set(&lpi2c_imx->rate_per, clk_get_rate(lpi2c_imx->clks[0].clk));
+	clk_rate_exclusive_put(lpi2c_imx->clks[0].clk);
+	if (!atomic_read(&lpi2c_imx->rate_per)) {
+		dev_err(&pdev->dev, "can't get I2C peripheral clock rate\n");
+		return -EINVAL;
+	}
+
 	pm_runtime_set_autosuspend_delay(&pdev->dev, I2C_PM_TIMEOUT);
 	pm_runtime_use_autosuspend(&pdev->dev);
 	pm_runtime_get_noresume(&pdev->dev);