Re: [PATCH v4] pinctrl: Add lock to ensure the state atomization

From: Aiqun Yu (Maria)
Date: Tue Feb 20 2024 - 01:13:34 EST

Next message: Chao Yu: "Re: [f2fs-dev] [PATCH v7] f2fs: unify the error handling of f2fs_is_valid_blkaddr"
Previous message: Zhiguo Niu: "[PATCH v4 4/4] f2fs: stop checkpoint when get a out-of-bounds segment"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On 2/9/2024 4:37 PM, Alexander Stein wrote:

Am Freitag, 2. Februar 2024, 11:58:54 CET schrieb Maria Yu:

Currently pinctrl_select_state is an export symbol and don't have
effective re-entrance protect design. During async probing of devices
it's possible to end up in pinctrl_select_state() from multiple
contexts simultaneously, so make it thread safe.
More over, when the real racy happened, the system frequently have
printk message like:
"not freeing pin xx (xxx) as part of deactivating group xxx - it is
already used for some other setting".
Finally the system crashed after the flood log.
Add per pinctrl lock to ensure the old state and new state transition
atomization.
Also move dev error print message outside the region with interrupts
disabled.
Use scoped guard to simplify the lock protection needed code.

Fixes: 4198a9b57106 ("pinctrl: avoid reload of p state in list iteration")
Signed-off-by: Maria Yu <quic_aiquny@xxxxxxxxxxx>
---
drivers/pinctrl/core.c | 143 +++++++++++++++++++++--------------------
drivers/pinctrl/core.h | 2 +
2 files changed, 75 insertions(+), 70 deletions(-)

diff --git a/drivers/pinctrl/core.c b/drivers/pinctrl/core.c
index ee56856cb80c..1f7d001d4c1e 100644
--- a/drivers/pinctrl/core.c
+++ b/drivers/pinctrl/core.c
@@ -1061,6 +1061,7 @@ static struct pinctrl *create_pinctrl(struct device
*dev, p->dev = dev;
INIT_LIST_HEAD(&p->states);
INIT_LIST_HEAD(&p->dt_maps);
+ spin_lock_init(&p->lock);

ret = pinctrl_dt_to_map(p, pctldev);
if (ret < 0) {
@@ -1257,93 +1258,95 @@ static void pinctrl_link_add(struct pinctrl_dev
*pctldev, static int pinctrl_commit_state(struct pinctrl *p, struct
pinctrl_state *state) {
struct pinctrl_setting *setting, *setting2;
- struct pinctrl_state *old_state = READ_ONCE(p->state);
+ struct pinctrl_state *old_state;
int ret;

- if (old_state) {
- /*
- * For each pinmux setting in the old state, forget SW's

record

- * of mux owner for that pingroup. Any pingroups which are
- * still owned by the new state will be re-acquired by the

call

- * to pinmux_enable_setting() in the loop below.
- */
- list_for_each_entry(setting, &old_state->settings, node) {
- if (setting->type != PIN_MAP_TYPE_MUX_GROUP)
- continue;
- pinmux_disable_setting(setting);
+ scoped_guard(spinlock_irqsave, &p->lock) {
+ old_state = p->state;
+ if (old_state) {
+ /*
+ * For each pinmux setting in the old state,

forget SW's record

+ * of mux owner for that pingroup. Any pingroups

which are

+ * still owned by the new state will be re-

acquired by the call

+ * to pinmux_enable_setting() in the loop below.
+ */
+ list_for_each_entry(setting, &old_state-
settings, node) {
+ if (setting->type !=

PIN_MAP_TYPE_MUX_GROUP)

+ continue;
+ pinmux_disable_setting(setting);
+ }
}
- }
-
- p->state = NULL;

- /* Apply all the settings for the new state - pinmux first */
- list_for_each_entry(setting, &state->settings, node) {
- switch (setting->type) {
- case PIN_MAP_TYPE_MUX_GROUP:
- ret = pinmux_enable_setting(setting);
- break;
- case PIN_MAP_TYPE_CONFIGS_PIN:
- case PIN_MAP_TYPE_CONFIGS_GROUP:
- ret = 0;
- break;
- default:
- ret = -EINVAL;
- break;
- }
+ p->state = NULL;

- if (ret < 0)
- goto unapply_new_state;
+ /* Apply all the settings for the new state - pinmux first

*/

+ list_for_each_entry(setting, &state->settings, node) {
+ switch (setting->type) {
+ case PIN_MAP_TYPE_MUX_GROUP:
+ ret = pinmux_enable_setting(setting);
+ break;
+ case PIN_MAP_TYPE_CONFIGS_PIN:
+ case PIN_MAP_TYPE_CONFIGS_GROUP:
+ ret = 0;
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }

- /* Do not link hogs (circular dependency) */
- if (p != setting->pctldev->p)
- pinctrl_link_add(setting->pctldev, p->dev);
- }
+ if (ret < 0)
+ goto unapply_new_state;

- /* Apply all the settings for the new state - pinconf after */
- list_for_each_entry(setting, &state->settings, node) {
- switch (setting->type) {
- case PIN_MAP_TYPE_MUX_GROUP:
- ret = 0;
- break;
- case PIN_MAP_TYPE_CONFIGS_PIN:
- case PIN_MAP_TYPE_CONFIGS_GROUP:
- ret = pinconf_apply_setting(setting);
- break;
- default:
- ret = -EINVAL;
- break;
+ /* Do not link hogs (circular dependency) */
+ if (p != setting->pctldev->p)
+ pinctrl_link_add(setting->pctldev, p-
dev);
}

- if (ret < 0) {
- goto unapply_new_state;
- }
+ /* Apply all the settings for the new state - pinconf

after */

+ list_for_each_entry(setting, &state->settings, node) {
+ switch (setting->type) {
+ case PIN_MAP_TYPE_MUX_GROUP:
+ ret = 0;
+ break;
+ case PIN_MAP_TYPE_CONFIGS_PIN:
+ case PIN_MAP_TYPE_CONFIGS_GROUP:
+ ret = pinconf_apply_setting(setting);
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }

- /* Do not link hogs (circular dependency) */
- if (p != setting->pctldev->p)
- pinctrl_link_add(setting->pctldev, p->dev);
- }
+ if (ret < 0)
+ goto unapply_new_state;

- p->state = state;
+ /* Do not link hogs (circular dependency) */
+ if (p != setting->pctldev->p)
+ pinctrl_link_add(setting->pctldev, p-
dev);
+ }

- return 0;
+ p->state = state;
+
+ return 0;

unapply_new_state:
- dev_err(p->dev, "Error applying setting, reverse things back\n");

- list_for_each_entry(setting2, &state->settings, node) {
- if (&setting2->node == &setting->node)
- break;
- /*
- * All we can do here is pinmux_disable_setting.
- * That means that some pins are muxed differently now
- * than they were before applying the setting (We can't
- * "unmux a pin"!), but it's not a big deal since the pins
- * are free to be muxed by another apply_setting.
- */
- if (setting2->type == PIN_MAP_TYPE_MUX_GROUP)
- pinmux_disable_setting(setting2);
+ list_for_each_entry(setting2, &state->settings, node) {
+ if (&setting2->node == &setting->node)
+ break;
+ /*
+ * All we can do here is pinmux_disable_setting.
+ * That means that some pins are muxed

differently now

+ * than they were before applying the setting

(We can't

+ * "unmux a pin"!), but it's not a big deal

since the pins

+ * are free to be muxed by another

apply_setting.

+ */
+ if (setting2->type == PIN_MAP_TYPE_MUX_GROUP)
+ pinmux_disable_setting(setting2);
+ }
}

+ dev_err(p->dev, "Error applying setting, reverse things back\n");
/* There's no infinite recursive loop here because p->state is NULL

*/

if (old_state)
pinctrl_select_state(p, old_state);
diff --git a/drivers/pinctrl/core.h b/drivers/pinctrl/core.h
index 837fd5bd903d..6844edd38b4a 100644
--- a/drivers/pinctrl/core.h
+++ b/drivers/pinctrl/core.h
@@ -12,6 +12,7 @@
#include <linux/list.h>
#include <linux/mutex.h>
#include <linux/radix-tree.h>
+#include <linux/spinlock.h>
#include <linux/types.h>

#include <linux/pinctrl/machine.h>
@@ -91,6 +92,7 @@ struct pinctrl {
struct pinctrl_state *state;
struct list_head dt_maps;
struct kref users;
+ spinlock_t lock;
};

/**

base-commit: 6613476e225e090cc9aad49be7fa504e290dd33d

This breaks pinctrl-imx on imx8qxp:

[ 1.170727] imx8qxp-pinctrl system-controller:pinctrl: initialized IMX
pinctrl driver
[ 1.283968] BUG: sleeping function called from invalid context at kernel/
locking/mutex.c:283
[ 1.292089] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 70,
name: kworker/u16:4
[ 1.300341] preempt_count: 1, expected: 0
[ 1.304337] RCU nest depth: 0, expected: 0
[ 1.308423] CPU: 2 PID: 70 Comm: kworker/u16:4 Not tainted 6.8.0-rc3-
next-20240209+ #2267 0b2aeebc4d64f1aef3abdd5fede2a9b5162eb867
[ 1.320148] Hardware name: TQ-Systems i.MX8QXP TQMa8XQP on MBa8Xx (DT)
[ 1.326667] Workqueue: events_unbound deferred_probe_work_func
[ 1.332486] Call trace:
[ 1.334918] dump_backtrace+0x90/0x10c
[ 1.338653] show_stack+0x14/0x1c
[ 1.341954] dump_stack_lvl+0x6c/0x80
[ 1.345603] dump_stack+0x14/0x1c
[ 1.348904] __might_resched+0x108/0x160
[ 1.352813] __might_sleep+0x58/0xb0
[ 1.356375] mutex_lock+0x20/0x74
[ 1.359676] imx_scu_call_rpc+0x44/0x2e8
[ 1.363586] imx_pinconf_set_scu+0x84/0x150
[ 1.367756] imx_pinconf_set+0x48/0x7c
[ 1.371491] pinconf_apply_setting+0x90/0x110
[ 1.375835] pinctrl_commit_state+0xcc/0x28c
[ 1.380092] pinctrl_select_state+0x18/0x28
[ 1.384262] pinctrl_bind_pins+0x1e4/0x26c
[ 1.388345] really_probe+0x60/0x3e0
[ 1.391907] __driver_probe_device+0x84/0x198
[ 1.396251] driver_probe_device+0x38/0x150
[ 1.400421] __device_attach_driver+0xcc/0x194
[ 1.404851] bus_for_each_drv+0x80/0xdc
[ 1.408674] __device_attach+0x9c/0x1d0
[ 1.412496] device_initial_probe+0x10/0x18
[ 1.416666] bus_probe_device+0xa4/0xa8
[ 1.420489] deferred_probe_work_func+0x9c/0xe8
[ 1.425006] process_one_work+0x14c/0x40c
[ 1.429002] worker_thread+0x304/0x414
[ 1.432738] kthread+0xf4/0x100
[ 1.435866] ret_from_fork+0x10/0x20

With this commit pin_config_set callbacks need to be atomic suddenly which is
a no-go for any device attached to i2c or spi and in this case IPC RPC.
Once reverted systems start normally again.

Thanks for the info. It is a valid scenario of your report. I will also re-visit this.

Best regards,
Alexander

--
Thx and BRs,
Aiqun(Maria) Yu

Next message: Chao Yu: "Re: [f2fs-dev] [PATCH v7] f2fs: unify the error handling of f2fs_is_valid_blkaddr"
Previous message: Zhiguo Niu: "[PATCH v4 4/4] f2fs: stop checkpoint when get a out-of-bounds segment"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]