Skip to contentSkip to conversation

ASoC: rt5682: Adjust headset volume button threshold

PATCH In review Sasha Levin 3 participants 15 messages Aug 17, 2021 – Aug 30, 2021

Subsystems (8)
lore
Share & tools
lore

Sasha Levin

author maintainer
#0 lore raw

PATCH 1/12 ASoC: rt5682: Adjust headset volume button threshold

From: Derek Fang <[email protected]>

[ Upstream commit 6d20bf7c020f417fdef1810a22da17c126603472 ]

Adjust the threshold of headset button volume+ to fix the wrong button detection issue with some brand headsets.

---
 sound/soc/codecs/rt5682.c | 1 +
 1 file changed, 1 insertion(+)
sound/soc/codecs/rt5682.c +1 -0
index abcd6f483788..51ecaa2abcd1 100644
@@ -44,6 +44,7 @@ static const struct reg_sequence patch_list[] = {
 	{RT5682_I2C_CTRL, 0x000f},
 	{RT5682_PLL2_INTERNAL, 0x8266},
 	{RT5682_SAR_IL_CMD_3, 0x8365},
+	{RT5682_SAR_IL_CMD_6, 0x0180},
 };
 
 void rt5682_apply_patch_list(struct rt5682_priv *rt5682, struct device *dev)
Signature · 1 line
2.30.2
author maintainer
#1 lore raw

PATCH 2/12 ASoC: component: Remove misplaced prefix handling in pin control functions

From: Mark Brown <[email protected]>

[ Upstream commit 31428c78748cafdd9352e1f622eb89bf453d9700 ]

When the component level pin control functions were added they for some no longer obvious reason handled adding prefixing of widget names. This meant that when the lack of prefix handling in the DAPM level pin operations was fixed by ae4fc532244b3bb4d (ASoC: dapm: use component prefix when checking widget names) the one device using the component level API ended up with the prefix being applied twice, causing all lookups to fail.

Fix this by removing the redundant prefixing from the component code, which has the nice side effect of also making that code much simpler.

---
 sound/soc/soc-component.c | 63 +++++++++++++++++----------------------
 1 file changed, 27 insertions(+), 36 deletions(-)
sound/soc/soc-component.c +27 -36
index 3a5e84e16a87..c8dfd0de30e4 100644
@@ -148,86 +148,75 @@ int snd_soc_component_set_bias_level(struct snd_soc_component *component,
 	return soc_component_ret(component, ret);
 }
 
-static int soc_component_pin(struct snd_soc_component *component,
-			     const char *pin,
-			     int (*pin_func)(struct snd_soc_dapm_context *dapm,
-					     const char *pin))
-{
-	struct snd_soc_dapm_context *dapm =
-		snd_soc_component_get_dapm(component);
-	char *full_name;
-	int ret;
-
-	if (!component->name_prefix) {
-		ret = pin_func(dapm, pin);
-		goto end;
-	}
-
-	full_name = kasprintf(GFP_KERNEL, "%s %s", component->name_prefix, pin);
-	if (!full_name) {
-		ret = -ENOMEM;
-		goto end;
-	}
-
-	ret = pin_func(dapm, full_name);
-	kfree(full_name);
-end:
-	return soc_component_ret(component, ret);
-}
-
 int snd_soc_component_enable_pin(struct snd_soc_component *component,
 				 const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_enable_pin);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_enable_pin(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_enable_pin);
 
 int snd_soc_component_enable_pin_unlocked(struct snd_soc_component *component,
 					  const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_enable_pin_unlocked);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_enable_pin_unlocked(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_enable_pin_unlocked);
 
 int snd_soc_component_disable_pin(struct snd_soc_component *component,
 				  const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_disable_pin);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_disable_pin(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_disable_pin);
 
 int snd_soc_component_disable_pin_unlocked(struct snd_soc_component *component,
 					   const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_disable_pin_unlocked);
+	struct snd_soc_dapm_context *dapm = 
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_disable_pin_unlocked(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_disable_pin_unlocked);
 
 int snd_soc_component_nc_pin(struct snd_soc_component *component,
 			     const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_nc_pin);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_nc_pin(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_nc_pin);
 
 int snd_soc_component_nc_pin_unlocked(struct snd_soc_component *component,
 				      const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_nc_pin_unlocked);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_nc_pin_unlocked(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_nc_pin_unlocked);
 
 int snd_soc_component_get_pin_status(struct snd_soc_component *component,
 				     const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_get_pin_status);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_get_pin_status(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_get_pin_status);
 
 int snd_soc_component_force_enable_pin(struct snd_soc_component *component,
 				       const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_force_enable_pin);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_force_enable_pin(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_force_enable_pin);
 
@@ -235,7 +224,9 @@ int snd_soc_component_force_enable_pin_unlocked(
 	struct snd_soc_component *component,
 	const char *pin)
 {
-	return soc_component_pin(component, pin, snd_soc_dapm_force_enable_pin_unlocked);
+	struct snd_soc_dapm_context *dapm =
+		snd_soc_component_get_dapm(component);
+	return snd_soc_dapm_force_enable_pin_unlocked(dapm, pin);
 }
 EXPORT_SYMBOL_GPL(snd_soc_component_force_enable_pin_unlocked);
 
Signature · 1 line
2.30.2
#2 lore raw

Re: [PATCH AUTOSEL 5.13 02/12] ASoC: component: Remove misplaced prefix handling in pin control functions

Hi!

Quoted 7 lines

When the component level pin control functions were added they for some no longer obvious reason handled adding prefixing of widget names. This meant that when the lack of prefix handling in the DAPM level pin operations was fixed by ae4fc532244b3bb4d (ASoC: dapm: use component prefix when checking widget names) the one device using the component level API ended up with the prefix being applied twice, causing all lookups to fail.

AFAICT ae4fc532244b3bb4d (ASoC: dapm: use component...) is not in 5.10-stable, so this one should not go into 5.10-stable, either?

(Or alternatively, both can go in).

I hope I understand it right.

Best regards,
Pavel

Signature · 2 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
author maintainer
#3 lore raw

PATCH 3/12 ASoC: wm_adsp: Let soc_cleanup_component_debugfs remove debugfs

In reply to Sasha Levin

From: Lucas Tanure <[email protected]>

[ Upstream commit acbf58e530416e167c3b323111f4013d9f2b0a7d ]

soc_cleanup_component_debugfs will debugfs_remove_recursive the component->debugfs_root, so adsp doesn't need to also remove the same entry.
By doing that adsp also creates a race with core component, which causes a NULL pointer dereference

---
 sound/soc/codecs/wm_adsp.c | 1 -
 1 file changed, 1 deletion(-)
sound/soc/codecs/wm_adsp.c +0 -1
index cef05d81c39b..6698b5343974 100644
@@ -746,7 +746,6 @@ static void wm_adsp2_init_debugfs(struct wm_adsp *dsp,
 static void wm_adsp2_cleanup_debugfs(struct wm_adsp *dsp)
 {
 	wm_adsp_debugfs_clear(dsp);
-	debugfs_remove_recursive(dsp->debugfs_root);
 }
 #else
 static inline void wm_adsp2_init_debugfs(struct wm_adsp *dsp,
Signature · 1 line
2.30.2
#4 lore raw

Re: [PATCH AUTOSEL 5.13 03/12] ASoC: wm_adsp: Let soc_cleanup_component_debugfs remove debugfs

Quoted 23 lines · Sasha Levin

From: Lucas Tanure <[email protected]>

[ Upstream commit acbf58e530416e167c3b323111f4013d9f2b0a7d ]

soc_cleanup_component_debugfs will debugfs_remove_recursive the component->debugfs_root, so adsp doesn't need to also remove the same entry.
By doing that adsp also creates a race with core component, which causes a NULL pointer dereference

---
 sound/soc/codecs/wm_adsp.c | 1 -
 1 file changed, 1 deletion(-)
sound/soc/codecs/wm_adsp.c +0 -0
index cef05d81c39b..6698b5343974 100644

sound/soc/codecs/wm_adsp.c static void wm_adsp2_init_debugfs(struct wm_adsp *dsp, · line 749

Quoted 5 lines
sound/soc/codecs/wm_adsp.c +0 -1
@@ -746,7 +746,6 @@ static void wm_adsp2_init_debugfs(struct wm_adsp *dsp,
 static void wm_adsp2_cleanup_debugfs(struct wm_adsp *dsp)
 {
 	wm_adsp_debugfs_clear(dsp);
-	debugfs_remove_recursive(dsp->debugfs_root);
 }

It might be better not to backport this patch to the stable kernels. The issue has only been seen on one out of tree driver and the patch looks a little off to me. This debugfs_remove_recursive should run before the soc_cleanup_component_debugfs one, and as such it's hard to see what is actually going on. We are currently investigating internally but we might end up reverting the change, and it only seems to be causing issues on the one not upstreamed part.

Apologies for missing the review of this one when it went up Mark, I was on holiday at the time.

Thanks,
Charles

author maintainer
#5 lore raw

Re: [PATCH AUTOSEL 5.13 03/12] ASoC: wm_adsp: Let soc_cleanup_component_debugfs remove debugfs

Quoted 39 lines · Charles Keepax
Quoted 23 lines · Sasha Levin

From: Lucas Tanure <[email protected]>

[ Upstream commit acbf58e530416e167c3b323111f4013d9f2b0a7d ]

soc_cleanup_component_debugfs will debugfs_remove_recursive the component->debugfs_root, so adsp doesn't need to also remove the same entry.
By doing that adsp also creates a race with core component, which causes a NULL pointer dereference

---
 sound/soc/codecs/wm_adsp.c | 1 -
 1 file changed, 1 deletion(-)
sound/soc/codecs/wm_adsp.c +0 -0
index cef05d81c39b..6698b5343974 100644

sound/soc/codecs/wm_adsp.c static void wm_adsp2_init_debugfs(struct wm_adsp *dsp, · line 749

Quoted 5 lines
sound/soc/codecs/wm_adsp.c +0 -1
@@ -746,7 +746,6 @@ static void wm_adsp2_init_debugfs(struct wm_adsp *dsp,
 static void wm_adsp2_cleanup_debugfs(struct wm_adsp *dsp)
 {
 	wm_adsp_debugfs_clear(dsp);
-	debugfs_remove_recursive(dsp->debugfs_root);
 }

It might be better not to backport this patch to the stable kernels. The issue has only been seen on one out of tree driver and the patch looks a little off to me. This debugfs_remove_recursive should run before the soc_cleanup_component_debugfs one, and as such it's hard to see what is actually going on. We are currently investigating internally but we might end up reverting the change, and it only seems to be causing issues on the one not upstreamed part.

Quoted 2 lines

Apologies for missing the review of this one when it went up Mark, I was on holiday at the time.

I'll drop it, it didn't go anywhere yet :) Thanks!

Signature · 2 lines
Thanks,
Sasha
author maintainer
#6 lore raw

PATCH 4/12 platform/x86: Add and use a dual_accel_detect() helper

In reply to Sasha Levin

From: Hans de Goede <[email protected]>

[ Upstream commit 153cca9caa81ca8912a70528daca4b9a523c6898 ]

Various 360 degree hinges (yoga) style 2-in-1 devices use 2 accelerometers to allow the OS to determine the angle between the display and the base of the device.

On Windows these are read by a special HingeAngleService process which calls undocumented ACPI methods, to let the firmware know if the 2-in-1 is in tablet- or laptop-mode. The firmware may use this to disable the kbd and touchpad to avoid spurious input in tablet-mode as well as to report SW_TABLET_MODE info to the OS.

Since Linux does not call these undocumented methods, the SW_TABLET_MODE info reported by various pdx86 drivers is incorrect on these devices.

Before this commit the intel-hid and thinkpad_acpi code already had 2 hardcoded checks for ACPI hardware-ids of dual-accel sensors to avoid reporting broken info.

And now we also have a bug-report about the same problem in the intel-vbtn code. Since there are at least 3 different ACPI hardware-ids in play, add a new dual_accel_detect() helper which checks for all 3, rather then adding different hardware-ids to the drivers as bug-reports trickle in. Having shared code which checks all known hardware-ids is esp. important for the intel-hid and intel-vbtn drivers as these are generic drivers which are used on a lot of devices.

The BOSC0200 hardware-id requires special handling, because often it is used for a single-accelerometer setup. Only in a few cases it refers to a dual-accel setup, in which case there will be 2 I2cSerialBus resources in the device's resource-list, so the helper checks for this.

---
 drivers/platform/x86/Kconfig             |  3 +
 drivers/platform/x86/dual_accel_detect.h | 75 ++++++++++++++++++++++++
 drivers/platform/x86/intel-hid.c         | 21 ++-----
 drivers/platform/x86/intel-vbtn.c        | 18 +++++-
 drivers/platform/x86/thinkpad_acpi.c     |  3 +-
 5 files changed, 101 insertions(+), 19 deletions(-)
 create mode 100644 drivers/platform/x86/dual_accel_detect.h
drivers/platform/x86/Kconfig +3 -0
index 60592fb88e7a..e878ac192758 100644
@@ -507,6 +507,7 @@ config THINKPAD_ACPI
 	depends on RFKILL || RFKILL = n
 	depends on ACPI_VIDEO || ACPI_VIDEO = n
 	depends on BACKLIGHT_CLASS_DEVICE
+	depends on I2C
 	select ACPI_PLATFORM_PROFILE
 	select HWMON
 	select NVRAM
@@ -701,6 +702,7 @@ config INTEL_HID_EVENT
 	tristate "INTEL HID Event"
 	depends on ACPI
 	depends on INPUT
+	depends on I2C
 	select INPUT_SPARSEKMAP
 	help
 	  This driver provides support for the Intel HID Event hotkey interface.
@@ -752,6 +754,7 @@ config INTEL_VBTN
 	tristate "INTEL VIRTUAL BUTTON"
 	depends on ACPI
 	depends on INPUT
+	depends on I2C
 	select INPUT_SPARSEKMAP
 	help
 	  This driver provides support for the Intel Virtual Button interface.
drivers/platform/x86/dual_accel_detect.h new file +75 -0

75 lines not on this page · raw message

drivers/platform/x86/intel-hid.c +6 -15
index 078648a9201b..7135d720af60 100644
@@ -14,6 +14,7 @@
 #include <linux/module.h>
 #include <linux/platform_device.h>
 #include <linux/suspend.h>
+#include "dual_accel_detect.h"
 
 /* When NOT in tablet mode, VGBS returns with the flag 0x40 */
 #define TABLET_MODE_FLAG BIT(6)
@@ -121,6 +122,7 @@ struct intel_hid_priv {
 	struct input_dev *array;
 	struct input_dev *switches;
 	bool wakeup_mode;
+	bool dual_accel;
 };
 
 #define HID_EVENT_FILTER_UUID	"eeec56b3-4442-408f-a792-4edd4d758054"
@@ -450,22 +452,9 @@ static void notify_handler(acpi_handle handle, u32 event, void *context)
 	 * SW_TABLET_MODE report, in these cases we enable support when receiving
 	 * the first event instead of during driver setup.
 	 *
-	 * Some 360 degree hinges (yoga) style 2-in-1 devices use 2 accelerometers
-	 * to allow the OS to determine the angle between the display and the base
-	 * of the device. On Windows these are read by a special HingeAngleService
-	 * process which calls an ACPI DSM (Device Specific Method) on the
-	 * ACPI KIOX010A device node for the sensor in the display, to let the
-	 * firmware know if the 2-in-1 is in tablet- or laptop-mode so that it can
-	 * disable the kbd and touchpad to avoid spurious input in tablet-mode.
-	 *
-	 * The linux kxcjk1013 driver calls the DSM for this once at probe time
-	 * to ensure that the builtin kbd and touchpad work. On some devices this
-	 * causes a "spurious" 0xcd event on the intel-hid ACPI dev. In this case
-	 * there is not a functional tablet-mode switch, so we should not register
-	 * the tablet-mode switch device.
+	 * See dual_accel_detect.h for more info on the dual_accel check.
 	 */
-	if (!priv->switches && (event == 0xcc || event == 0xcd) &&
-	    !acpi_dev_present("KIOX010A", NULL, -1)) {
+	if (!priv->switches && !priv->dual_accel && (event == 0xcc || event == 0xcd)) {
 		dev_info(&device->dev, "switch event received, enable switches supports\n");
 		err = intel_hid_switches_setup(device);
 		if (err)
@@ -606,6 +595,8 @@ static int intel_hid_probe(struct platform_device *device)
 		return -ENOMEM;
 	dev_set_drvdata(&device->dev, priv);
 
+	priv->dual_accel = dual_accel_detect();
+
 	err = intel_hid_input_setup(device);
 	if (err) {
 		pr_err("Failed to setup Intel HID hotkeys\n");
drivers/platform/x86/intel-vbtn.c +15 -3
index 888a764efad1..309166431063 100644
@@ -14,6 +14,7 @@
 #include <linux/module.h>
 #include <linux/platform_device.h>
 #include <linux/suspend.h>
+#include "dual_accel_detect.h"
 
 /* Returned when NOT in tablet mode on some HP Stream x360 11 models */
 #define VGBS_TABLET_MODE_FLAG_ALT	0x10
@@ -66,6 +67,7 @@ static const struct key_entry intel_vbtn_switchmap[] = {
 struct intel_vbtn_priv {
 	struct input_dev *buttons_dev;
 	struct input_dev *switches_dev;
+	bool dual_accel;
 	bool has_buttons;
 	bool has_switches;
 	bool wakeup_mode;
@@ -160,6 +162,10 @@ static void notify_handler(acpi_handle handle, u32 event, void *context)
 		input_dev = priv->buttons_dev;
 	} else if ((ke = sparse_keymap_entry_from_scancode(priv->switches_dev, event))) {
 		if (!priv->has_switches) {
+			/* See dual_accel_detect.h for more info */
+			if (priv->dual_accel)
+				return;
+
 			dev_info(&device->dev, "Registering Intel Virtual Switches input-dev after receiving a switch event\n");
 			ret = input_register_device(priv->switches_dev);
 			if (ret)
@@ -248,11 +254,15 @@ static const struct dmi_system_id dmi_switches_allow_list[] = {
 	{} /* Array terminator */
 };
 
-static bool intel_vbtn_has_switches(acpi_handle handle)
+static bool intel_vbtn_has_switches(acpi_handle handle, bool dual_accel)
 {
 	unsigned long long vgbs;
 	acpi_status status;
 
+	/* See dual_accel_detect.h for more info */
+	if (dual_accel)
+		return false;
+
 	if (!dmi_check_system(dmi_switches_allow_list))
 		return false;
 
@@ -263,13 +273,14 @@ static bool intel_vbtn_has_switches(acpi_handle handle)
 static int intel_vbtn_probe(struct platform_device *device)
 {
 	acpi_handle handle = ACPI_HANDLE(&device->dev);
-	bool has_buttons, has_switches;
+	bool dual_accel, has_buttons, has_switches;
 	struct intel_vbtn_priv *priv;
 	acpi_status status;
 	int err;
 
+	dual_accel = dual_accel_detect();
 	has_buttons = acpi_has_method(handle, "VBDL");
-	has_switches = intel_vbtn_has_switches(handle);
+	has_switches = intel_vbtn_has_switches(handle, dual_accel);
 
 	if (!has_buttons && !has_switches) {
 		dev_warn(&device->dev, "failed to read Intel Virtual Button driver\n");
@@ -281,6 +292,7 @@ static int intel_vbtn_probe(struct platform_device *device)
 		return -ENOMEM;
 	dev_set_drvdata(&device->dev, priv);
 
+	priv->dual_accel = dual_accel;
 	priv->has_buttons = has_buttons;
 	priv->has_switches = has_switches;
 
drivers/platform/x86/thinkpad_acpi.c +2 -1
index edd71e744d27..f8cfb529d1a2 100644
@@ -73,6 +73,7 @@
 #include <linux/uaccess.h>
 #include <acpi/battery.h>
 #include <acpi/video.h>
+#include "dual_accel_detect.h"
 
 /* ThinkPad CMOS commands */
 #define TP_CMOS_VOLUME_DOWN	0
@@ -3232,7 +3233,7 @@ static int hotkey_init_tablet_mode(void)
 		 * the laptop/tent/tablet mode to the EC. The bmc150 iio driver
 		 * does not support this, so skip the hotkey on these models.
 		 */
-		if (has_tablet_mode && !acpi_dev_present("BOSC0200", "1", -1))
+		if (has_tablet_mode && !dual_accel_detect())
 			tp_features.hotkey_tablet = TP_HOTKEY_TABLET_USES_GMMS;
 		type = "GMMS";
 	} else if (acpi_evalf(hkey_handle, &res, "MHKG", "qd")) {
Signature · 1 line
2.30.2
author maintainer
#7 lore raw

PATCH 5/12 ARC: Fix CONFIG_STACKDEPOT

In reply to Sasha Levin

From: Guenter Roeck <[email protected]>

[ Upstream commit bf79167fd86f3b97390fe2e70231d383526bd9cc ]

Enabling CONFIG_STACKDEPOT results in the following build error.

arc-elf-ld: lib/stackdepot.o: in function `filter_irq_stacks':
stackdepot.c:(.text+0x456): undefined reference to `__irqentry_text_start'
arc-elf-ld: stackdepot.c:(.text+0x456): undefined reference to `__irqentry_text_start'
arc-elf-ld: stackdepot.c:(.text+0x476): undefined reference to `__irqentry_text_end'
arc-elf-ld: stackdepot.c:(.text+0x476): undefined reference to `__irqentry_text_end'
arc-elf-ld: stackdepot.c:(.text+0x484): undefined reference to `__softirqentry_text_start'
arc-elf-ld: stackdepot.c:(.text+0x484): undefined reference to `__softirqentry_text_start'
arc-elf-ld: stackdepot.c:(.text+0x48c): undefined reference to `__softirqentry_text_end'
arc-elf-ld: stackdepot.c:(.text+0x48c): undefined reference to `__softirqentry_text_end'

Other architectures address this problem by adding IRQENTRY_TEXT and SOFTIRQENTRY_TEXT to the text segment, so do the same here.

---
 arch/arc/kernel/vmlinux.lds.S | 2 ++
 1 file changed, 2 insertions(+)
arch/arc/kernel/vmlinux.lds.S +2 -0
index e2146a8da195..529ae50f9fe2 100644
@@ -88,6 +88,8 @@ SECTIONS
 		CPUIDLE_TEXT
 		LOCK_TEXT
 		KPROBES_TEXT
+		IRQENTRY_TEXT
+		SOFTIRQENTRY_TEXT
 		*(.fixup)
 		*(.gnu.warning)
 	}
Signature · 1 line
2.30.2
author maintainer
#8 lore raw

PATCH 6/12 netfilter: ipset: Limit the maximal range of consecutive elements to add/delete

In reply to Sasha Levin

From: Jozsef Kadlecsik <[email protected]>

[ Upstream commit 5f7b51bf09baca8e4f80cbe879536842bafb5f31 ]

The range size of consecutive elements were not limited. Thus one could define a huge range which may result soft lockup errors due to the long execution time. Now the range size is limited to 2^20 entries.

---
 include/linux/netfilter/ipset/ip_set.h       |  3 +++
 net/netfilter/ipset/ip_set_hash_ip.c         |  9 ++++++++-
 net/netfilter/ipset/ip_set_hash_ipmark.c     | 10 +++++++++-
 net/netfilter/ipset/ip_set_hash_ipport.c     |  3 +++
 net/netfilter/ipset/ip_set_hash_ipportip.c   |  3 +++
 net/netfilter/ipset/ip_set_hash_ipportnet.c  |  3 +++
 net/netfilter/ipset/ip_set_hash_net.c        | 11 ++++++++++-
 net/netfilter/ipset/ip_set_hash_netiface.c   | 10 +++++++++-
 net/netfilter/ipset/ip_set_hash_netnet.c     | 16 +++++++++++++++-
 net/netfilter/ipset/ip_set_hash_netport.c    | 11 ++++++++++-
 net/netfilter/ipset/ip_set_hash_netportnet.c | 16 +++++++++++++++-
 11 files changed, 88 insertions(+), 7 deletions(-)
include/linux/netfilter/ipset/ip_set.h +3 -0
index 10279c4830ac..ada1296c87d5 100644
@@ -196,6 +196,9 @@ struct ip_set_region {
 	u32 elements;		/* Number of elements vs timeout */
 };
 
+/* Max range where every element is added/deleted in one step */
+#define IPSET_MAX_RANGE		(1<<20)
+
 /* The max revision number supported by any set type + 1 */
 #define IPSET_REVISION_MAX	9
 
net/netfilter/ipset/ip_set_hash_ip.c +8 -1
index d1bef23fd4f5..dd30c03d5a23 100644
@@ -132,8 +132,11 @@ hash_ip4_uadt(struct ip_set *set, struct nlattr *tb[],
 		ret = ip_set_get_hostipaddr4(tb[IPSET_ATTR_IP_TO], &ip_to);
 		if (ret)
 			return ret;
-		if (ip > ip_to)
+		if (ip > ip_to) {
+			if (ip_to == 0)
+				return -IPSET_ERR_HASH_ELEM;
 			swap(ip, ip_to);
+		}
 	} else if (tb[IPSET_ATTR_CIDR]) {
 		u8 cidr = nla_get_u8(tb[IPSET_ATTR_CIDR]);
 
@@ -144,6 +147,10 @@ hash_ip4_uadt(struct ip_set *set, struct nlattr *tb[],
 
 	hosts = h->netmask == 32 ? 1 : 2 << (32 - h->netmask - 1);
 
+	/* 64bit division is not allowed on 32bit */
+	if (((u64)ip_to - ip + 1) >> (32 - h->netmask) > IPSET_MAX_RANGE)
+		return -ERANGE;
+
 	if (retried) {
 		ip = ntohl(h->next.ip);
 		e.ip = htonl(ip);
net/netfilter/ipset/ip_set_hash_ipmark.c +9 -1
index 18346d18aa16..153de3457423 100644
@@ -121,6 +121,8 @@ hash_ipmark4_uadt(struct ip_set *set, struct nlattr *tb[],
 
 	e.mark = ntohl(nla_get_be32(tb[IPSET_ATTR_MARK]));
 	e.mark &= h->markmask;
+	if (e.mark == 0 && e.ip == 0)
+		return -IPSET_ERR_HASH_ELEM;
 
 	if (adt == IPSET_TEST ||
 	    !(tb[IPSET_ATTR_IP_TO] || tb[IPSET_ATTR_CIDR])) {
@@ -133,8 +135,11 @@ hash_ipmark4_uadt(struct ip_set *set, struct nlattr *tb[],
 		ret = ip_set_get_hostipaddr4(tb[IPSET_ATTR_IP_TO], &ip_to);
 		if (ret)
 			return ret;
-		if (ip > ip_to)
+		if (ip > ip_to) {
+			if (e.mark == 0 && ip_to == 0)
+				return -IPSET_ERR_HASH_ELEM;
 			swap(ip, ip_to);
+		}
 	} else if (tb[IPSET_ATTR_CIDR]) {
 		u8 cidr = nla_get_u8(tb[IPSET_ATTR_CIDR]);
 
@@ -143,6 +148,9 @@ hash_ipmark4_uadt(struct ip_set *set, struct nlattr *tb[],
 		ip_set_mask_from_to(ip, ip_to, cidr);
 	}
 
+	if (((u64)ip_to - ip + 1) > IPSET_MAX_RANGE)
+		return -ERANGE;
+
 	if (retried)
 		ip = ntohl(h->next.ip);
 	for (; ip <= ip_to; ip++) {
net/netfilter/ipset/ip_set_hash_ipport.c +3 -0
index e1ca11196515..7303138e46be 100644
@@ -173,6 +173,9 @@ hash_ipport4_uadt(struct ip_set *set, struct nlattr *tb[],
 			swap(port, port_to);
 	}
 
+	if (((u64)ip_to - ip + 1)*(port_to - port + 1) > IPSET_MAX_RANGE)
+		return -ERANGE;
+
 	if (retried)
 		ip = ntohl(h->next.ip);
 	for (; ip <= ip_to; ip++) {
net/netfilter/ipset/ip_set_hash_ipportip.c +3 -0
index ab179e064597..334fb1ad0e86 100644
@@ -180,6 +180,9 @@ hash_ipportip4_uadt(struct ip_set *set, struct nlattr *tb[],
 			swap(port, port_to);
 	}
 
+	if (((u64)ip_to - ip + 1)*(port_to - port + 1) > IPSET_MAX_RANGE)
+		return -ERANGE;
+
 	if (retried)
 		ip = ntohl(h->next.ip);
 	for (; ip <= ip_to; ip++) {
net/netfilter/ipset/ip_set_hash_ipportnet.c +3 -0
index 8f075b44cf64..7df94f437f60 100644
@@ -253,6 +253,9 @@ hash_ipportnet4_uadt(struct ip_set *set, struct nlattr *tb[],
 			swap(port, port_to);
 	}
 
+	if (((u64)ip_to - ip + 1)*(port_to - port + 1) > IPSET_MAX_RANGE)
+		return -ERANGE;
+
 	ip2_to = ip2_from;
 	if (tb[IPSET_ATTR_IP2_TO]) {
 		ret = ip_set_get_hostipaddr4(tb[IPSET_ATTR_IP2_TO], &ip2_to);
net/netfilter/ipset/ip_set_hash_net.c +10 -1
index c1a11f041ac6..1422739d9aa2 100644
@@ -140,7 +140,7 @@ hash_net4_uadt(struct ip_set *set, struct nlattr *tb[],
 	ipset_adtfn adtfn = set->variant->adt[adt];
 	struct hash_net4_elem e = { .cidr = HOST_MASK };
 	struct ip_set_ext ext = IP_SET_INIT_UEXT(set);
-	u32 ip = 0, ip_to = 0;
+	u32 ip = 0, ip_to = 0, ipn, n = 0;
 	int ret;
 
 	if (tb[IPSET_ATTR_LINENO])
@@ -188,6 +188,15 @@ hash_net4_uadt(struct ip_set *set, struct nlattr *tb[],
 		if (ip + UINT_MAX == ip_to)
 			return -IPSET_ERR_HASH_RANGE;
 	}
+	ipn = ip;
+	do {
+		ipn = ip_set_range_to_cidr(ipn, ip_to, &e.cidr);
+		n++;
+	} while (ipn++ < ip_to);
+
+	if (n > IPSET_MAX_RANGE)
+		return -ERANGE;
+
 	if (retried)
 		ip = ntohl(h->next.ip);
 	do {
net/netfilter/ipset/ip_set_hash_netiface.c +9 -1
index ddd51c2e1cb3..9810f5bf63f5 100644
@@ -202,7 +202,7 @@ hash_netiface4_uadt(struct ip_set *set, struct nlattr *tb[],
 	ipset_adtfn adtfn = set->variant->adt[adt];
 	struct hash_netiface4_elem e = { .cidr = HOST_MASK, .elem = 1 };
 	struct ip_set_ext ext = IP_SET_INIT_UEXT(set);
-	u32 ip = 0, ip_to = 0;
+	u32 ip = 0, ip_to = 0, ipn, n = 0;
 	int ret;
 
 	if (tb[IPSET_ATTR_LINENO])
@@ -256,6 +256,14 @@ hash_netiface4_uadt(struct ip_set *set, struct nlattr *tb[],
 	} else {
 		ip_set_mask_from_to(ip, ip_to, e.cidr);
 	}
+	ipn = ip;
+	do {
+		ipn = ip_set_range_to_cidr(ipn, ip_to, &e.cidr);
+		n++;
+	} while (ipn++ < ip_to);
+
+	if (n > IPSET_MAX_RANGE)
+		return -ERANGE;
 
 	if (retried)
 		ip = ntohl(h->next.ip);
net/netfilter/ipset/ip_set_hash_netnet.c +15 -1
index 6532f0505e66..3d09eefe998a 100644
@@ -168,7 +168,8 @@ hash_netnet4_uadt(struct ip_set *set, struct nlattr *tb[],
 	struct hash_netnet4_elem e = { };
 	struct ip_set_ext ext = IP_SET_INIT_UEXT(set);
 	u32 ip = 0, ip_to = 0;
-	u32 ip2 = 0, ip2_from = 0, ip2_to = 0;
+	u32 ip2 = 0, ip2_from = 0, ip2_to = 0, ipn;
+	u64 n = 0, m = 0;
 	int ret;
 
 	if (tb[IPSET_ATTR_LINENO])
@@ -244,6 +245,19 @@ hash_netnet4_uadt(struct ip_set *set, struct nlattr *tb[],
 	} else {
 		ip_set_mask_from_to(ip2_from, ip2_to, e.cidr[1]);
 	}
+	ipn = ip;
+	do {
+		ipn = ip_set_range_to_cidr(ipn, ip_to, &e.cidr[0]);
+		n++;
+	} while (ipn++ < ip_to);
+	ipn = ip2_from;
+	do {
+		ipn = ip_set_range_to_cidr(ipn, ip2_to, &e.cidr[1]);
+		m++;
+	} while (ipn++ < ip2_to);
+
+	if (n*m > IPSET_MAX_RANGE)
+		return -ERANGE;
 
 	if (retried) {
 		ip = ntohl(h->next.ip[0]);
net/netfilter/ipset/ip_set_hash_netport.c +10 -1
index ec1564a1cb5a..09cf72eb37f8 100644
@@ -158,7 +158,8 @@ hash_netport4_uadt(struct ip_set *set, struct nlattr *tb[],
 	ipset_adtfn adtfn = set->variant->adt[adt];
 	struct hash_netport4_elem e = { .cidr = HOST_MASK - 1 };
 	struct ip_set_ext ext = IP_SET_INIT_UEXT(set);
-	u32 port, port_to, p = 0, ip = 0, ip_to = 0;
+	u32 port, port_to, p = 0, ip = 0, ip_to = 0, ipn;
+	u64 n = 0;
 	bool with_ports = false;
 	u8 cidr;
 	int ret;
@@ -235,6 +236,14 @@ hash_netport4_uadt(struct ip_set *set, struct nlattr *tb[],
 	} else {
 		ip_set_mask_from_to(ip, ip_to, e.cidr + 1);
 	}
+	ipn = ip;
+	do {
+		ipn = ip_set_range_to_cidr(ipn, ip_to, &cidr);
+		n++;
+	} while (ipn++ < ip_to);
+
+	if (n*(port_to - port + 1) > IPSET_MAX_RANGE)
+		return -ERANGE;
 
 	if (retried) {
 		ip = ntohl(h->next.ip);
net/netfilter/ipset/ip_set_hash_netportnet.c +15 -1
index 0e91d1e82f1c..19bcdb3141f6 100644
@@ -182,7 +182,8 @@ hash_netportnet4_uadt(struct ip_set *set, struct nlattr *tb[],
 	struct hash_netportnet4_elem e = { };
 	struct ip_set_ext ext = IP_SET_INIT_UEXT(set);
 	u32 ip = 0, ip_to = 0, p = 0, port, port_to;
-	u32 ip2_from = 0, ip2_to = 0, ip2;
+	u32 ip2_from = 0, ip2_to = 0, ip2, ipn;
+	u64 n = 0, m = 0;
 	bool with_ports = false;
 	int ret;
 
@@ -284,6 +285,19 @@ hash_netportnet4_uadt(struct ip_set *set, struct nlattr *tb[],
 	} else {
 		ip_set_mask_from_to(ip2_from, ip2_to, e.cidr[1]);
 	}
+	ipn = ip;
+	do {
+		ipn = ip_set_range_to_cidr(ipn, ip_to, &e.cidr[0]);
+		n++;
+	} while (ipn++ < ip_to);
+	ipn = ip2_from;
+	do {
+		ipn = ip_set_range_to_cidr(ipn, ip2_to, &e.cidr[1]);
+		m++;
+	} while (ipn++ < ip2_to);
+
+	if (n*m*(port_to - port + 1) > IPSET_MAX_RANGE)
+		return -ERANGE;
 
 	if (retried) {
 		ip = ntohl(h->next.ip[0]);
Signature · 1 line
2.30.2
author maintainer
#9 lore raw

PATCH 7/12 netfilter: conntrack: collect all entries in one cycle

In reply to Sasha Levin

From: Florian Westphal <[email protected]>

[ Upstream commit 4608fdfc07e116f9fc0895beb40abad7cdb5ee3d ]

Michal Kubecek reports that conntrack gc is responsible for frequent wakeups (every 125ms) on idle systems.

On busy systems, timed out entries are evicted during lookup. The gc worker is only needed to remove entries after system becomes idle after a busy period.

To resolve this, always scan the entire table.
If the scan is taking too long, reschedule so other work_structs can run and resume from next bucket.

After a completed scan, wait for 2 minutes before the next cycle. Heuristics for faster re-schedule are removed.

GC_SCAN_INTERVAL could be exposed as a sysctl in the future to allow tuning this as-needed or even turn the gc worker off.

---
 net/netfilter/nf_conntrack_core.c | 71 ++++++++++---------------------
 1 file changed, 22 insertions(+), 49 deletions(-)
net/netfilter/nf_conntrack_core.c +22 -49
index 69079a382d3a..6ba9f4b8d145 100644
@@ -68,22 +68,17 @@ EXPORT_SYMBOL_GPL(nf_conntrack_hash);
 
 struct conntrack_gc_work {
 	struct delayed_work	dwork;
-	u32			last_bucket;
+	u32			next_bucket;
 	bool			exiting;
 	bool			early_drop;
-	long			next_gc_run;
 };
 
 static __read_mostly struct kmem_cache *nf_conntrack_cachep;
 static DEFINE_SPINLOCK(nf_conntrack_locks_all_lock);
 static __read_mostly bool nf_conntrack_locks_all;
 
-/* every gc cycle scans at most 1/GC_MAX_BUCKETS_DIV part of table */
-#define GC_MAX_BUCKETS_DIV	128u
-/* upper bound of full table scan */
-#define GC_MAX_SCAN_JIFFIES	(16u * HZ)
-/* desired ratio of entries found to be expired */
-#define GC_EVICT_RATIO	50u
+#define GC_SCAN_INTERVAL	(120u * HZ)
+#define GC_SCAN_MAX_DURATION	msecs_to_jiffies(10)
 
 static struct conntrack_gc_work conntrack_gc_work;
 
@@ -1359,17 +1354,13 @@ static bool gc_worker_can_early_drop(const struct nf_conn *ct)
 
 static void gc_worker(struct work_struct *work)
 {
-	unsigned int min_interval = max(HZ / GC_MAX_BUCKETS_DIV, 1u);
-	unsigned int i, goal, buckets = 0, expired_count = 0;
-	unsigned int nf_conntrack_max95 = 0;
+	unsigned long end_time = jiffies + GC_SCAN_MAX_DURATION;
+	unsigned int i, hashsz, nf_conntrack_max95 = 0;
+	unsigned long next_run = GC_SCAN_INTERVAL;
 	struct conntrack_gc_work *gc_work;
-	unsigned int ratio, scanned = 0;
-	unsigned long next_run;
-
 	gc_work = container_of(work, struct conntrack_gc_work, dwork.work);
 
-	goal = nf_conntrack_htable_size / GC_MAX_BUCKETS_DIV;
-	i = gc_work->last_bucket;
+	i = gc_work->next_bucket;
 	if (gc_work->early_drop)
 		nf_conntrack_max95 = nf_conntrack_max / 100u * 95u;
 
@@ -1377,15 +1368,15 @@ static void gc_worker(struct work_struct *work)
 		struct nf_conntrack_tuple_hash *h;
 		struct hlist_nulls_head *ct_hash;
 		struct hlist_nulls_node *n;
-		unsigned int hashsz;
 		struct nf_conn *tmp;
 
-		i++;
 		rcu_read_lock();
 
 		nf_conntrack_get_ht(&ct_hash, &hashsz);
-		if (i >= hashsz)
-			i = 0;
+		if (i >= hashsz) {
+			rcu_read_unlock();
+			break;
+		}
 
 		hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[i], hnnode) {
 			struct nf_conntrack_net *cnet;
@@ -1393,7 +1384,6 @@ static void gc_worker(struct work_struct *work)
 
 			tmp = nf_ct_tuplehash_to_ctrack(h);
 
-			scanned++;
 			if (test_bit(IPS_OFFLOAD_BIT, &tmp->status)) {
 				nf_ct_offload_timeout(tmp);
 				continue;
@@ -1401,7 +1391,6 @@ static void gc_worker(struct work_struct *work)
 
 			if (nf_ct_is_expired(tmp)) {
 				nf_ct_gc_expired(tmp);
-				expired_count++;
 				continue;
 			}
 
@@ -1434,7 +1423,14 @@ static void gc_worker(struct work_struct *work)
 		 */
 		rcu_read_unlock();
 		cond_resched();
-	} while (++buckets < goal);
+		i++;
+
+		if (time_after(jiffies, end_time) && i < hashsz) {
+			gc_work->next_bucket = i;
+			next_run = 0;
+			break;
+		}
+	} while (i < hashsz);
 
 	if (gc_work->exiting)
 		return;
@@ -1445,40 +1441,17 @@ static void gc_worker(struct work_struct *work)
 	 *
 	 * This worker is only here to reap expired entries when system went
 	 * idle after a busy period.
-	 *
-	 * The heuristics below are supposed to balance conflicting goals:
-	 *
-	 * 1. Minimize time until we notice a stale entry
-	 * 2. Maximize scan intervals to not waste cycles
-	 *
-	 * Normally, expire ratio will be close to 0.
-	 *
-	 * As soon as a sizeable fraction of the entries have expired
-	 * increase scan frequency.
 	 */
-	ratio = scanned ? expired_count * 100 / scanned : 0;
-	if (ratio > GC_EVICT_RATIO) {
-		gc_work->next_gc_run = min_interval;
-	} else {
-		unsigned int max = GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV;
-
-		BUILD_BUG_ON((GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV) == 0);
-
-		gc_work->next_gc_run += min_interval;
-		if (gc_work->next_gc_run > max)
-			gc_work->next_gc_run = max;
+	if (next_run) {
+		gc_work->early_drop = false;
+		gc_work->next_bucket = 0;
 	}
-
-	next_run = gc_work->next_gc_run;
-	gc_work->last_bucket = i;
-	gc_work->early_drop = false;
 	queue_delayed_work(system_power_efficient_wq, &gc_work->dwork, next_run);
 }
 
 static void conntrack_gc_work_init(struct conntrack_gc_work *gc_work)
 {
 	INIT_DEFERRABLE_WORK(&gc_work->dwork, gc_worker);
-	gc_work->next_gc_run = HZ;
 	gc_work->exiting = false;
 }
 
Signature · 1 line
2.30.2
author maintainer
#10 lore raw

PATCH 8/12 once: Fix panic when module unload

In reply to Sasha Levin

From: Kefeng Wang <[email protected]>

[ Upstream commit 1027b96ec9d34f9abab69bc1a4dc5b1ad8ab1349 ]

DO_ONCE
DEFINE_STATIC_KEY_TRUE(___once_key);
__do_once_done

  once_disable_jump(once_key);
    INIT_WORK(&w->work, once_deferred);
    struct once_work *w;
    w->key = key;
    schedule_work(&w->work);                     module unload
                                                   //*the key is

destroy*
process_one_work
once_deferred
BUG_ON(!static_key_enabled(work->key));
static_key_count((struct static_key *)x) //*access key, crash*

When module uses DO_ONCE mechanism, it could crash due to the above concurrency problem, we could reproduce it with link[1].

Fix it by add/put module refcount in the once work process.

[1] https://lore.kernel.org/netdev/[email protected]/

---
 include/linux/once.h |  4 ++--
 lib/once.c           | 11 ++++++++---
 2 files changed, 10 insertions(+), 5 deletions(-)
include/linux/once.h +2 -2
index 9225ee6d96c7..ae6f4eb41cbe 100644
@@ -7,7 +7,7 @@
 
 bool __do_once_start(bool *done, unsigned long *flags);
 void __do_once_done(bool *done, struct static_key_true *once_key,
-		    unsigned long *flags);
+		    unsigned long *flags, struct module *mod);
 
 /* Call a function exactly once. The idea of DO_ONCE() is to perform
  * a function call such as initialization of random seeds, etc, only
@@ -46,7 +46,7 @@ void __do_once_done(bool *done, struct static_key_true *once_key,
 			if (unlikely(___ret)) {				     \
 				func(__VA_ARGS__);			     \
 				__do_once_done(&___done, &___once_key,	     \
-					       &___flags);		     \
+					       &___flags, THIS_MODULE);	     \
 			}						     \
 		}							     \
 		___ret;							     \
lib/once.c +8 -3
index 8b7d6235217e..59149bf3bfb4 100644
@@ -3,10 +3,12 @@
 #include <linux/spinlock.h>
 #include <linux/once.h>
 #include <linux/random.h>
+#include <linux/module.h>
 
 struct once_work {
 	struct work_struct work;
 	struct static_key_true *key;
+	struct module *module;
 };
 
 static void once_deferred(struct work_struct *w)
@@ -16,10 +18,11 @@ static void once_deferred(struct work_struct *w)
 	work = container_of(w, struct once_work, work);
 	BUG_ON(!static_key_enabled(work->key));
 	static_branch_disable(work->key);
+	module_put(work->module);
 	kfree(work);
 }
 
-static void once_disable_jump(struct static_key_true *key)
+static void once_disable_jump(struct static_key_true *key, struct module *mod)
 {
 	struct once_work *w;
 
@@ -29,6 +32,8 @@ static void once_disable_jump(struct static_key_true *key)
 
 	INIT_WORK(&w->work, once_deferred);
 	w->key = key;
+	w->module = mod;
+	__module_get(mod);
 	schedule_work(&w->work);
 }
 
@@ -53,11 +58,11 @@ bool __do_once_start(bool *done, unsigned long *flags)
 EXPORT_SYMBOL(__do_once_start);
 
 void __do_once_done(bool *done, struct static_key_true *once_key,
-		    unsigned long *flags)
+		    unsigned long *flags, struct module *mod)
 	__releases(once_lock)
 {
 	*done = true;
 	spin_unlock_irqrestore(&once_lock, *flags);
-	once_disable_jump(once_key);
+	once_disable_jump(once_key, mod);
 }
 EXPORT_SYMBOL(__do_once_done);
Signature · 1 line
2.30.2
author maintainer
#11 lore raw

PATCH 9/12 io_uring: rsrc ref lock needs to be IRQ safe

In reply to Sasha Levin

From: Jens Axboe <[email protected]>

[ Upstream commit 4956b9eaad456a88b0d56947bef036e086250beb ]

Nadav reports running into the below splat on re-enabling softirqs:

WARNING: CPU: 2 PID: 1777 at kernel/softirq.c:364 __local_bh_enable_ip+0xaa/0xe0
Modules linked in:
CPU: 2 PID: 1777 Comm: umem Not tainted 5.13.1+ #161
Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/22/2020
RIP: 0010:__local_bh_enable_ip+0xaa/0xe0
Code: a9 00 ff ff 00 74 38 65 ff 0d a2 21 8c 7a e8 ed 1a 20 00 fb 66 0f 1f 44 00 00 5b 41 5c 5d c3 65 8b 05 e6 2d 8c 7a 85 c0 75 9a <0f> 0b eb 96 e8 2d 1f 20 00 eb a5 4c 89 e7 e8 73 4f 0c 00 eb ae 65
RSP: 0018:ffff88812e58fcc8 EFLAGS: 00010046
RAX: 0000000000000000 RBX: 0000000000000201 RCX: dffffc0000000000
RDX: 0000000000000007 RSI: 0000000000000201 RDI: ffffffff8898c5ac
RBP: ffff88812e58fcd8 R08: ffffffff8575dbbf R09: ffffed1028ef14f9
R10: ffff88814778a7c3 R11: ffffed1028ef14f8 R12: ffffffff85c9e9ae
R13: ffff88814778a000 R14: ffff88814778a7b0 R15: ffff8881086db890
FS:  00007fbcfee17700(0000) GS:ffff8881e0300000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000000c0402a5008 CR3: 000000011c1ac003 CR4: 00000000003706e0
Call Trace:
 _raw_spin_unlock_bh+0x31/0x40
 io_rsrc_node_ref_zero+0x13e/0x190
 io_dismantle_req+0x215/0x220
 io_req_complete_post+0x1b8/0x720
 __io_complete_rw.isra.0+0x16b/0x1f0
 io_complete_rw+0x10/0x20

where it's clear we end up calling the percpu count release directly from the completion path, as it's in atomic mode and we drop the last ref. For file/block IO, this can be from IRQ context already, and the softirq locking for rsrc isn't enough.

Just make the lock fully IRQ safe, and ensure we correctly safe state from the release path as we don't know the full context there.

---
 fs/io_uring.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)
fs/io_uring.c +5 -14
index 32f3df13a812..29bdf6fa0f83 100644
@@ -7099,16 +7099,6 @@ static void io_free_file_tables(struct io_file_table *table, unsigned nr_files)
 	table->files = NULL;
 }
 
-static inline void io_rsrc_ref_lock(struct io_ring_ctx *ctx)
-{
-	spin_lock_bh(&ctx->rsrc_ref_lock);
-}
-
-static inline void io_rsrc_ref_unlock(struct io_ring_ctx *ctx)
-{
-	spin_unlock_bh(&ctx->rsrc_ref_lock);
-}
-
 static void io_rsrc_node_destroy(struct io_rsrc_node *ref_node)
 {
 	percpu_ref_exit(&ref_node->refs);
@@ -7125,9 +7115,9 @@ static void io_rsrc_node_switch(struct io_ring_ctx *ctx,
 		struct io_rsrc_node *rsrc_node = ctx->rsrc_node;
 
 		rsrc_node->rsrc_data = data_to_kill;
-		io_rsrc_ref_lock(ctx);
+		spin_lock_irq(&ctx->rsrc_ref_lock);
 		list_add_tail(&rsrc_node->node, &ctx->rsrc_ref_list);
-		io_rsrc_ref_unlock(ctx);
+		spin_unlock_irq(&ctx->rsrc_ref_lock);
 
 		atomic_inc(&data_to_kill->refs);
 		percpu_ref_kill(&rsrc_node->refs);
@@ -7617,9 +7607,10 @@ static void io_rsrc_node_ref_zero(struct percpu_ref *ref)
 {
 	struct io_rsrc_node *node = container_of(ref, struct io_rsrc_node, refs);
 	struct io_ring_ctx *ctx = node->rsrc_data->ctx;
+	unsigned long flags;
 	bool first_add = false;
 
-	io_rsrc_ref_lock(ctx);
+	spin_lock_irqsave(&ctx->rsrc_ref_lock, flags);
 	node->done = true;
 
 	while (!list_empty(&ctx->rsrc_ref_list)) {
@@ -7631,7 +7622,7 @@ static void io_rsrc_node_ref_zero(struct percpu_ref *ref)
 		list_del(&node->node);
 		first_add |= llist_add(&node->llist, &ctx->rsrc_put_llist);
 	}
-	io_rsrc_ref_unlock(ctx);
+	spin_unlock_irqrestore(&ctx->rsrc_ref_lock, flags);
 
 	if (first_add)
 		mod_delayed_work(system_wq, &ctx->rsrc_put_work, HZ);
Signature · 1 line
2.30.2
author maintainer
#12 lore raw

PATCH 10/12 blk-iocost: fix lockdep warning on blkcg->lock

In reply to Sasha Levin

From: Ming Lei <[email protected]>

[ Upstream commit 11431e26c9c43fa26f6b33ee1a90989f57b86024 ]

blkcg->lock depends on q->queue_lock which may depend on another driver lock required in irq context, one example is dm-thin:

	Chain exists of:
	  &pool->lock#3 --> &q->queue_lock --> &blkcg->lock
	 Possible interrupt unsafe locking scenario:

	       CPU0                    CPU1
	       ----                    ----
	  lock(&blkcg->lock);
	                               local_irq_disable();
	                               lock(&pool->lock#3);
	                               lock(&q->queue_lock);
	  <Interrupt>
	    lock(&pool->lock#3);

Fix the issue by using spin_lock_irq(&blkcg->lock) in ioc_weight_write().

---
 block/blk-iocost.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
block/blk-iocost.c +4 -4
index 5fac3757e6e0..0e56557cacf2 100644
@@ -3061,19 +3061,19 @@ static ssize_t ioc_weight_write(struct kernfs_open_file *of, char *buf,
 		if (v < CGROUP_WEIGHT_MIN || v > CGROUP_WEIGHT_MAX)
 			return -EINVAL;
 
-		spin_lock(&blkcg->lock);
+		spin_lock_irq(&blkcg->lock);
 		iocc->dfl_weight = v * WEIGHT_ONE;
 		hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
 			struct ioc_gq *iocg = blkg_to_iocg(blkg);
 
 			if (iocg) {
-				spin_lock_irq(&iocg->ioc->lock);
+				spin_lock(&iocg->ioc->lock);
 				ioc_now(iocg->ioc, &now);
 				weight_updated(iocg, &now);
-				spin_unlock_irq(&iocg->ioc->lock);
+				spin_unlock(&iocg->ioc->lock);
 			}
 		}
-		spin_unlock(&blkcg->lock);
+		spin_unlock_irq(&blkcg->lock);
 
 		return nbytes;
 	}
Signature · 1 line
2.30.2
author maintainer
#13 lore raw

PATCH 11/12 ovl: fix uninitialized pointer read in ovl_lookup_real_one()

In reply to Sasha Levin

From: Miklos Szeredi <[email protected]>

[ Upstream commit 580c610429b3994e8db24418927747cf28443cde ]

One error path can result in release_dentry_name_snapshot() being called before "name" was initialized by take_dentry_name_snapshot().

Fix by moving the release_dentry_name_snapshot() to immediately after the only use.

---
 fs/overlayfs/export.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/overlayfs/export.c +1 -1
index 41ebf52f1bbc..ebde05c9cf62 100644
@@ -392,6 +392,7 @@ static struct dentry *ovl_lookup_real_one(struct dentry *connected,
 	 */
 	take_dentry_name_snapshot(&name, real);
 	this = lookup_one_len(name.name.name, connected, name.name.len);
+	release_dentry_name_snapshot(&name);
 	err = PTR_ERR(this);
 	if (IS_ERR(this)) {
 		goto fail;
@@ -406,7 +407,6 @@ static struct dentry *ovl_lookup_real_one(struct dentry *connected,
 	}
 
 out:
-	release_dentry_name_snapshot(&name);
 	dput(parent);
 	inode_unlock(dir);
 	return this;
Signature · 1 line
2.30.2
author maintainer
#14 lore raw

PATCH 12/12 net: mscc: Fix non-GPL export of regmap APIs

In reply to Sasha Levin

From: Mark Brown <[email protected]>

[ Upstream commit 48c812e0327744b4965296f65c23fe2405692afc ]

The ocelot driver makes use of regmap, wrapping it with driver specific operations that are thin wrappers around the core regmap APIs. These are exported with EXPORT_SYMBOL, dropping the _GPL from the core regmap exports which is frowned upon. Add _GPL suffixes to at least the APIs that are doing register I/O.

---
 drivers/net/ethernet/mscc/ocelot_io.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
drivers/net/ethernet/mscc/ocelot_io.c +8 -8
index ea4e83410fe4..7390fa3980ec 100644
@@ -21,7 +21,7 @@ u32 __ocelot_read_ix(struct ocelot *ocelot, u32 reg, u32 offset)
 		    ocelot->map[target][reg & REG_MASK] + offset, &val);
 	return val;
 }
-EXPORT_SYMBOL(__ocelot_read_ix);
+EXPORT_SYMBOL_GPL(__ocelot_read_ix);
 
 void __ocelot_write_ix(struct ocelot *ocelot, u32 val, u32 reg, u32 offset)
 {
@@ -32,7 +32,7 @@ void __ocelot_write_ix(struct ocelot *ocelot, u32 val, u32 reg, u32 offset)
 	regmap_write(ocelot->targets[target],
 		     ocelot->map[target][reg & REG_MASK] + offset, val);
 }
-EXPORT_SYMBOL(__ocelot_write_ix);
+EXPORT_SYMBOL_GPL(__ocelot_write_ix);
 
 void __ocelot_rmw_ix(struct ocelot *ocelot, u32 val, u32 mask, u32 reg,
 		     u32 offset)
@@ -45,7 +45,7 @@ void __ocelot_rmw_ix(struct ocelot *ocelot, u32 val, u32 mask, u32 reg,
 			   ocelot->map[target][reg & REG_MASK] + offset,
 			   mask, val);
 }
-EXPORT_SYMBOL(__ocelot_rmw_ix);
+EXPORT_SYMBOL_GPL(__ocelot_rmw_ix);
 
 u32 ocelot_port_readl(struct ocelot_port *port, u32 reg)
 {
@@ -58,7 +58,7 @@ u32 ocelot_port_readl(struct ocelot_port *port, u32 reg)
 	regmap_read(port->target, ocelot->map[target][reg & REG_MASK], &val);
 	return val;
 }
-EXPORT_SYMBOL(ocelot_port_readl);
+EXPORT_SYMBOL_GPL(ocelot_port_readl);
 
 void ocelot_port_writel(struct ocelot_port *port, u32 val, u32 reg)
 {
@@ -69,7 +69,7 @@ void ocelot_port_writel(struct ocelot_port *port, u32 val, u32 reg)
 
 	regmap_write(port->target, ocelot->map[target][reg & REG_MASK], val);
 }
-EXPORT_SYMBOL(ocelot_port_writel);
+EXPORT_SYMBOL_GPL(ocelot_port_writel);
 
 void ocelot_port_rmwl(struct ocelot_port *port, u32 val, u32 mask, u32 reg)
 {
@@ -77,7 +77,7 @@ void ocelot_port_rmwl(struct ocelot_port *port, u32 val, u32 mask, u32 reg)
 
 	ocelot_port_writel(port, (cur & (~mask)) | val, reg);
 }
-EXPORT_SYMBOL(ocelot_port_rmwl);
+EXPORT_SYMBOL_GPL(ocelot_port_rmwl);
 
 u32 __ocelot_target_read_ix(struct ocelot *ocelot, enum ocelot_target target,
 			    u32 reg, u32 offset)
@@ -128,7 +128,7 @@ int ocelot_regfields_init(struct ocelot *ocelot,
 
 	return 0;
 }
-EXPORT_SYMBOL(ocelot_regfields_init);
+EXPORT_SYMBOL_GPL(ocelot_regfields_init);
 
 static struct regmap_config ocelot_regmap_config = {
 	.reg_bits	= 32,
@@ -148,4 +148,4 @@ struct regmap *ocelot_regmap_init(struct ocelot *ocelot, struct resource *res)
 
 	return devm_regmap_init_mmio(ocelot->dev, regs, &ocelot_regmap_config);
 }
-EXPORT_SYMBOL(ocelot_regmap_init);
+EXPORT_SYMBOL_GPL(ocelot_regmap_init);
Signature · 1 line
2.30.2