File _service:download_files:0001-fixes-miscellaneou... of Package linux-px

Overview Repositories Revisions Requests Users Attributes Meta

File _service:download_files:0001-fixes-miscellaneous.patch of Package linux-px (Revision 85bccb547dd01b546a3d5130f3df83bb)

Currently displaying revision 85bccb547dd01b546a3d5130f3df83bb , Show latest

From 6c39130856bf1238264f80fb20f5429845cca005 Mon Sep 17 00:00:00 2001
From: Tor Vic <torvic9@mailbox.org>
Date: Mon, 20 Jun 2022 12:40:39 +0200
Subject: [PATCH 01/24] mm: Change dirty writeback defaults

Use a lower ratio of 5, but a longer period of 10 seconds for dirty writeback.
---
 mm/page-writeback.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 516b1aa24..92d3d1006 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -71,7 +71,7 @@ static long ratelimit_pages = 32;
 /*
  * Start background writeback (via writeback threads) at this percentage
  */
-static int dirty_background_ratio = 10;
+static int dirty_background_ratio = 5;
 
 /*
  * dirty_background_bytes starts at 0 (disabled) so that it is a function of
@@ -99,7 +99,7 @@ static unsigned long vm_dirty_bytes;
 /*
  * The interval between `kupdate'-style writebacks
  */
-unsigned int dirty_writeback_interval = 5 * 100; /* centiseconds */
+unsigned int dirty_writeback_interval = 10 * 100; /* centiseconds */
 
 EXPORT_SYMBOL_GPL(dirty_writeback_interval);
 
-- 
2.40.1.445.gf85cd430b1

From 170856b10093368b064a76d3778b2583ddfe94f3 Mon Sep 17 00:00:00 2001
From: Sultan Alsawaf <sultan@kerneltoast.com>
Date: Wed, 20 Oct 2021 20:50:11 -0700
Subject: [PATCH 02/24] ZEN: mm: Lower the non-hugetlbpage pageblock size to
 reduce scheduling delays

The page allocator processes free pages in groups of pageblocks, where
the size of a pageblock is typically quite large (1024 pages without
hugetlbpage support). Pageblocks are processed atomically with the zone
lock held, which can cause severe scheduling delays on both the CPU
going through the pageblock and any other CPUs waiting to acquire the
zone lock. A frequent offender is move_freepages_block(), which is used
by rmqueue() for page allocation.

As it turns out, there's no requirement for pageblocks to be so large,
so the pageblock order can simply be reduced to ease the scheduling
delays and zone lock contention. PAGE_ALLOC_COSTLY_ORDER is used as a
reasonable setting to ensure non-costly page allocation requests can
still be serviced without always needing to free up more than one
pageblock's worth of pages at a time.

This has a noticeable effect on overall system latency when memory
pressure is elevated. The various mm functions which operate on
pageblocks no longer appear in the preemptoff tracer, where previously
they would spend up to 100 ms on a mobile arm64 CPU processing a
pageblock with preemption disabled and the zone lock held.

Signed-off-by: Sultan Alsawaf <sultan@kerneltoast.com>
---
 include/linux/pageblock-flags.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/pageblock-flags.h b/include/linux/pageblock-flags.h
index 5f1ae07d7..97cda629c 100644
--- a/include/linux/pageblock-flags.h
+++ b/include/linux/pageblock-flags.h
@@ -48,7 +48,7 @@ extern unsigned int pageblock_order;
 #else /* CONFIG_HUGETLB_PAGE */
 
 /* If huge pages are not used, group by MAX_ORDER_NR_PAGES */
-#define pageblock_order		(MAX_ORDER-1)
+#define pageblock_order		PAGE_ALLOC_COSTLY_ORDER
 
 #endif /* CONFIG_HUGETLB_PAGE */
 
-- 
2.40.1.445.gf85cd430b1

From 2bd948941b130d182adeeb2333c2da336623eb56 Mon Sep 17 00:00:00 2001
From: Masahiro Yamada <masahiroy@kernel.org>
Date: Tue, 15 Nov 2022 02:46:17 +0900
Subject: [PATCH 03/24] kbuild: revive parallel execution for
 .tmp_initcalls.lds rule

Prior to commit 5d45950dfbb1 ("kbuild: move vmlinux.o link to
scripts/Makefile.vmlinux_o"), jobserver-exec was invoked from the shell
script, link-vmlinux.sh. It can get access to the jobserver because
Makefile adds '+' prefix, as in:

+$(call if_changed_dep,link_vmlinux)

Since 5d45950dfbb1, jobserver-exec is invoked from Makefile, but the
'+' prefix is missing, hence jobserver-exec has no access to the
jobserver.

Fixes: 5d45950dfbb1 ("kbuild: move vmlinux.o link to scripts/Makefile.vmlinux_o")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
---
 scripts/Makefile.vmlinux_o | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/Makefile.vmlinux_o b/scripts/Makefile.vmlinux_o
index 0edfdb403..ae52d3b3f 100644
--- a/scripts/Makefile.vmlinux_o
+++ b/scripts/Makefile.vmlinux_o
@@ -19,7 +19,7 @@ quiet_cmd_gen_initcalls_lds = GEN     $@
 
 .tmp_initcalls.lds: $(srctree)/scripts/generate_initcall_order.pl \
 		vmlinux.a $(KBUILD_VMLINUX_LIBS) FORCE
-	$(call if_changed,gen_initcalls_lds)
+	+$(call if_changed,gen_initcalls_lds)
 
 targets := .tmp_initcalls.lds
 
-- 
2.40.1.445.gf85cd430b1

From e18c9f3c086cb710f722a74f7ab734dc81aea07b Mon Sep 17 00:00:00 2001
From: Nathan Chancellor <nathan@kernel.org>
Date: Tue, 29 Nov 2022 12:01:22 -0700
Subject: [PATCH 04/24] padata: Do not mark padata_mt_helper() as __init

When building arm64 allmodconfig + ThinLTO with clang and a proposed
modpost update to account for -ffuncton-sections, the following warning
appears:

WARNING: modpost: vmlinux.o: section mismatch in reference: padata_work_init (section: .text.padata_work_init) -> padata_mt_helper (section: .init.text)
  WARNING: modpost: vmlinux.o: section mismatch in reference: padata_work_init (section: .text.padata_work_init) -> padata_mt_helper (section: .init.text)

In both cases, an __init function calls padata_work_init(), which is not
marked __init, with padata_mt_helper(), another __init function, as a
work function argument.

padata_work_init() is called from non-init paths, otherwise it could be
marked __init to resolve the warning. Instead, remove __init from
padata_mt_helper() to resolve the warning.

Signed-off-by: Nathan Chancellor <nathan@kernel.org>
---
 kernel/padata.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/padata.c b/kernel/padata.c
index e007b8a4b..7c80301ab 100644
--- a/kernel/padata.c
+++ b/kernel/padata.c
@@ -45,7 +45,7 @@ struct padata_mt_job_state {
 };
 
 static void padata_free_pd(struct parallel_data *pd);
-static void __init padata_mt_helper(struct work_struct *work);
+static void padata_mt_helper(struct work_struct *work);
 
 static int padata_index_to_cpu(struct parallel_data *pd, int cpu_index)
 {
@@ -438,7 +438,7 @@ static int padata_setup_cpumasks(struct padata_instance *pinst)
 	return err;
 }
 
-static void __init padata_mt_helper(struct work_struct *w)
+static void padata_mt_helper(struct work_struct *w)
 {
 	struct padata_work *pw = container_of(w, struct padata_work, pw_work);
 	struct padata_mt_job_state *ps = pw->pw_data;
-- 
2.40.1.445.gf85cd430b1

From 809abbb90c0db94d149afd344be55e059d06f3ae Mon Sep 17 00:00:00 2001
From: Benjamin Segall <bsegall@google.com>
Date: Tue, 6 Dec 2022 13:59:55 -0800
Subject: [PATCH 05/24] epoll: ep_autoremove_wake_function should use
 list_del_init_careful

autoremove_wake_function uses list_del_init_careful, so should epoll's
more aggressive variant. It only doesn't because it was copied from an
older wait.c rather than the most recent.

Fixes: a16ceb139610 ("epoll: autoremove wakers even more aggressively")
Signed-off-by: Ben Segall <bsegall@google.com>
Cc: stable@vger.kernel.org
---
 fs/eventpoll.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index 64659b110..8b5ca9f8f 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -1760,7 +1760,7 @@ static int ep_autoremove_wake_function(struct wait_queue_entry *wq_entry,
 {
 	int ret = default_wake_function(wq_entry, mode, sync, key);
 
-	list_del_init(&wq_entry->entry);
+	list_del_init_careful(&wq_entry->entry);
 	return ret;
 }
 
-- 
2.40.1.445.gf85cd430b1

From c22f0ddae7b22d2e18a7de171bc253f01c9d885a Mon Sep 17 00:00:00 2001
From: Vishal Rao <vishalrao@gmail.com>
Date: Mon, 19 Dec 2022 15:34:03 +0530
Subject: [PATCH 06/24] Fix sound on ASUS Zenbook UM5302TA

---
 sound/pci/hda/cs35l41_hda.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sound/pci/hda/cs35l41_hda.c b/sound/pci/hda/cs35l41_hda.c
index 75020edd3..e4455220e 100644
--- a/sound/pci/hda/cs35l41_hda.c
+++ b/sound/pci/hda/cs35l41_hda.c
@@ -1239,7 +1239,7 @@ static int cs35l41_no_acpi_dsd(struct cs35l41_hda *cs35l41, struct device *physd
 
 	if (strncmp(hid, "CLSA0100", 8) == 0) {
 		hw_cfg->bst_type = CS35L41_EXT_BOOST_NO_VSPK_SWITCH;
-	} else if (strncmp(hid, "CLSA0101", 8) == 0) {
+	} else if (strncmp(hid, "CLSA0101", 8) == 0 || strncmp(hid, "CSC3551", 7) == 0) {
 		hw_cfg->bst_type = CS35L41_EXT_BOOST;
 		hw_cfg->gpio1.func = CS35l41_VSPK_SWITCH;
 		hw_cfg->gpio1.valid = true;
-- 
2.40.1.445.gf85cd430b1

From cbc0b2c34eef6ac8cf38699e17f6bf549b2067be Mon Sep 17 00:00:00 2001
From: Peter Jung <admin@ptr1337.dev>
Date: Thu, 29 Dec 2022 12:11:30 +0100
Subject: [PATCH 07/24] Initialize ata before graphics

ATA init is the long pole in the boot process, and its asynchronous.
move the graphics init after it so that ata and graphics initialize
in parallel

Signed-off-by: Peter Jung <admin@ptr1337.dev>
---
 drivers/Makefile | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/drivers/Makefile b/drivers/Makefile
index 20b118dca..c19dee206 100644
--- a/drivers/Makefile
+++ b/drivers/Makefile
@@ -64,15 +64,8 @@ obj-y				+= char/
 # iommu/ comes before gpu as gpu are using iommu controllers
 obj-y				+= iommu/
 
-# gpu/ comes after char for AGP vs DRM startup and after iommu
-obj-y				+= gpu/
-
 obj-$(CONFIG_CONNECTOR)		+= connector/
 
-# i810fb and intelfb depend on char/agp/
-obj-$(CONFIG_FB_I810)           += video/fbdev/i810/
-obj-$(CONFIG_FB_INTEL)          += video/fbdev/intelfb/
-
 obj-$(CONFIG_PARPORT)		+= parport/
 obj-y				+= base/ block/ misc/ mfd/ nfc/
 obj-$(CONFIG_LIBNVDIMM)		+= nvdimm/
@@ -84,6 +77,14 @@ obj-y				+= macintosh/
 obj-y				+= scsi/
 obj-y				+= nvme/
 obj-$(CONFIG_ATA)		+= ata/
+
+# gpu/ comes after char for AGP vs DRM startup and after iommu
+obj-y				+= gpu/
+
+# i810fb and intelfb depend on char/agp/
+obj-$(CONFIG_FB_I810)           += video/fbdev/i810/
+obj-$(CONFIG_FB_INTEL)          += video/fbdev/intelfb/
+
 obj-$(CONFIG_TARGET_CORE)	+= target/
 obj-$(CONFIG_MTD)		+= mtd/
 obj-$(CONFIG_SPI)		+= spi/
-- 
2.40.1.445.gf85cd430b1

From bf12641b543f072deb5aa92d60b023a54d424d3f Mon Sep 17 00:00:00 2001
From: Oleksandr Natalenko <oleksandr@natalenko.name>
Date: Sun, 26 Feb 2023 00:17:09 +0100
Subject: [PATCH 08/24] Bluetooth: btusb: work around command 0xfc05 tx timeout

Instead of realoding `btusb` module to tackle this:

```
Bluetooth: hci0: command 0xfc05 tx timeout
Bluetooth: hci0: Reading Intel version command failed (-110)
```

increase `btusb_qca_cmd_timeout()` sleep duration while hoping
for the best.

This looks like an ugly hack.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=215167
Signed-off-by: Oleksandr Natalenko <oleksandr@natalenko.name>
---
 drivers/bluetooth/btusb.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/bluetooth/btusb.c b/drivers/bluetooth/btusb.c
index 092358229..d40de705b 100644
--- a/drivers/bluetooth/btusb.c
+++ b/drivers/bluetooth/btusb.c
@@ -915,7 +915,7 @@ static void btusb_qca_cmd_timeout(struct hci_dev *hdev)
 		}
 
 		gpiod_set_value_cansleep(reset_gpio, 0);
-		msleep(200);
+		usleep_range(USEC_PER_SEC / 2, USEC_PER_SEC);
 		gpiod_set_value_cansleep(reset_gpio, 1);
 
 		return;
-- 
2.40.1.445.gf85cd430b1

From 045ff05b23b0361bb73ffd33bfd25cb4c19891c1 Mon Sep 17 00:00:00 2001
From: Ian Pilcher <arequipeno@gmail.com>
Date: Tue, 27 Dec 2022 16:52:26 -0600
Subject: [PATCH 09/24] leds: trigger: Add block device LED trigger

Add "blkdev" LED trigger to blink LEDs in response to block device
activity.

Add LEDS_TRIGGER_BLKDEV (tristate) config option to control building of
the trigger.

Signed-off-by: Ian Pilcher <arequipeno@gmail.com>
Tested-by: Tor Vic <torvic9@mailbox.org> (version 12)
---
 drivers/leds/trigger/Kconfig          |    9 +
 drivers/leds/trigger/Makefile         |    1 +
 drivers/leds/trigger/ledtrig-blkdev.c | 1221 +++++++++++++++++++++++++
 3 files changed, 1231 insertions(+)
 create mode 100644 drivers/leds/trigger/ledtrig-blkdev.c

diff --git a/drivers/leds/trigger/Kconfig b/drivers/leds/trigger/Kconfig
index dc6816d36..bda249068 100644
--- a/drivers/leds/trigger/Kconfig
+++ b/drivers/leds/trigger/Kconfig
@@ -154,4 +154,13 @@ config LEDS_TRIGGER_TTY
 
 	  When build as a module this driver will be called ledtrig-tty.
 
+config LEDS_TRIGGER_BLKDEV
+	tristate "LED Trigger for block devices"
+	depends on BLOCK
+	help
+	  The blkdev LED trigger allows LEDs to be controlled by block device
+	  activity (reads and writes).
+
+	  See Documentation/leds/ledtrig-blkdev.rst.
+
 endif # LEDS_TRIGGERS
diff --git a/drivers/leds/trigger/Makefile b/drivers/leds/trigger/Makefile
index 25c4db97c..d53bab5d9 100644
--- a/drivers/leds/trigger/Makefile
+++ b/drivers/leds/trigger/Makefile
@@ -16,3 +16,4 @@ obj-$(CONFIG_LEDS_TRIGGER_NETDEV)	+= ledtrig-netdev.o
 obj-$(CONFIG_LEDS_TRIGGER_PATTERN)	+= ledtrig-pattern.o
 obj-$(CONFIG_LEDS_TRIGGER_AUDIO)	+= ledtrig-audio.o
 obj-$(CONFIG_LEDS_TRIGGER_TTY)		+= ledtrig-tty.o
+obj-$(CONFIG_LEDS_TRIGGER_BLKDEV)	+= ledtrig-blkdev.o
diff --git a/drivers/leds/trigger/ledtrig-blkdev.c b/drivers/leds/trigger/ledtrig-blkdev.c
new file mode 100644
index 000000000..067eedb00
--- /dev/null
+++ b/drivers/leds/trigger/ledtrig-blkdev.c
@@ -0,0 +1,1221 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+/*
+ *	Block device LED trigger
+ *
+ *	Copyright 2021-2022 Ian Pilcher <arequipeno@gmail.com>
+ */
+
+#include <linux/blkdev.h>
+#include <linux/leds.h>
+#include <linux/module.h>
+#include <linux/part_stat.h>
+#include <linux/xarray.h>
+
+/**
+ * DOC: Overview
+ *
+ * The ``blkdev`` LED trigger works by periodically checking the activity
+ * counters of block devices that have been linked to one or more LEDs and
+ * blinking those LED(s) if the correct type of activity has occurred.  The
+ * periodic check is scheduled with the Linux kernel's deferred work facility.
+ *
+ * Trigger-specific data about block devices and LEDs is stored in two data
+ * structures --- &struct blkdev_trig_bdev (a "BTB") and &struct blkdev_trig_led
+ * (a "BTL").  Each structure contains a &struct xarray that holds links to any
+ * linked devices of the other type.  I.e. &blkdev_trig_bdev.linked_btls
+ * contains links to all BTLs whose LEDs have been linked to the BTB's block
+ * device, and &blkdev_trig_led.linked_btbs contains links to all BTBs whose
+ * block devices have been linked to the BTL's LED.  Thus, a block device can
+ * be linked to more than one LED, and an LED can be linked to more than one
+ * block device.
+ */
+
+/* Default, minimum & maximum blink duration (milliseconds) */
+#define BLKDEV_TRIG_BLINK_DEF	75
+#define BLKDEV_TRIG_BLINK_MIN	10
+#define BLKDEV_TRIG_BLINK_MAX	86400000  /* 24 hours */
+
+/* Default, minimum & maximum activity check interval (milliseconds) */
+#define BLKDEV_TRIG_CHECK_DEF	100
+#define BLKDEV_TRIG_CHECK_MIN	25
+#define BLKDEV_TRIG_CHECK_MAX	86400000  /* 24 hours */
+
+/*
+ * If blkdev_trig_check() can't lock the mutex, how long to wait before trying
+ * again (milliseconds)
+ */
+#define BLKDEV_TRIG_CHECK_RETRY	5
+
+/* Mode argument for calls to blkdev_get_by_path() and blkdev_put() */
+#define BLKDEV_TRIG_FMODE	0
+
+/**
+ * struct blkdev_trig_bdev - Trigger-specific data about a block device.
+ * @last_checked:	Time (in jiffies) at which the trigger last checked this
+ *			block device for activity.
+ * @last_activity:	Time (in jiffies) at which the trigger last detected
+ *			activity of each type.
+ * @ios:		Activity counter values for each type, corresponding to
+ *			the timestamps in &last_activity.
+ * @index:		&xarray index, so the BTB can be included in one or more
+ *			&blkdev_trig_led.linked_btbs.
+ * @bdev:		The block device.
+ * @linked_btls:	The BTLs that represent the LEDs linked to the BTB's
+ *			block device.
+ *
+ * Every block device linked to at least one LED gets a "BTB."  A BTB is created
+ * when a block device that is not currently linked to any LEDs is linked to an
+ * LED.
+ *
+ * A BTB is freed when one of the following occurs:
+ *
+ * * The number of LEDs linked to the block device becomes zero, because it has
+ *   been unlinked from its last LED using the trigger's &sysfs interface.
+ *
+ * * The number of LEDs linked to the block device becomes zero, because the
+ *   last LED to which it was linked has been disassociated from the trigger
+ *   (which happens automatically if the LED device is removed from the system).
+ *
+ * * The BTB's block device is removed from the system.  To accomodate this
+ *   scenario, BTB's are created as device resources, so that the release
+ *   function will be called by the driver core when the device is removed.
+ */
+struct blkdev_trig_bdev {
+	unsigned long		last_checked;
+	unsigned long		last_activity[NR_STAT_GROUPS];
+	unsigned long		ios[NR_STAT_GROUPS];
+	unsigned long		index;
+	struct block_device	*bdev;
+	struct xarray		linked_btls;
+};
+
+/**
+ * struct blkdev_trig_led - Trigger-specific data about an LED.
+ * @last_checked:	Time (in jiffies) at which the trigger last checked the
+ *			the block devices linked to this LED for activity.
+ * @index:		&xarray index, so the BTL can be included in one or more
+ *			&blkdev_trig_bdev.linked_btls.
+ * @mode:		Bitmask for types of block device activity that will
+ *			cause this LED to blink --- reads, writes, discards,
+ *			etc.
+ * @led:		The LED device.
+ * @blink_msec:		Duration of a blink (milliseconds).
+ * @check_jiffies:	Frequency with which block devices linked to this LED
+ *			should be checked for activity (jiffies).
+ * @linked_btbs:	The BTBs that represent the block devices linked to the
+ *			BTL's LED.
+ * @all_btls_node:	The BTL's node in the module's list of all BTLs.
+ *
+ * Every LED associated with the block device trigger gets a "BTL."  A BTL is
+ * created when the trigger is "activated" on an LED (usually by writing
+ * ``blkdev`` to the LED's &sysfs &trigger attribute).  A BTL is freed wnen its
+ * LED is disassociated from the trigger, either through the trigger's &sysfs
+ * interface or because the LED device is removed from the system.
+ */
+struct blkdev_trig_led {
+	unsigned long		last_checked;
+	unsigned long		index;
+	unsigned long		mode;  /* must be ulong for atomic bit ops */
+	struct led_classdev	*led;
+	unsigned int		blink_msec;
+	unsigned int		check_jiffies;
+	struct xarray		linked_btbs;
+	struct hlist_node	all_btls_node;
+};
+
+/* Protects everything except atomic LED attributes */
+static DEFINE_MUTEX(blkdev_trig_mutex);
+
+/* BTB device resource release function */
+static void blkdev_trig_btb_release(struct device *dev, void *res);
+
+/* Index for next BTB or BTL */
+static unsigned long blkdev_trig_next_index;
+
+/* All LEDs associated with the trigger */
+static HLIST_HEAD(blkdev_trig_all_btls);
+
+/* Delayed work to periodically check for activity & blink LEDs */
+static void blkdev_trig_check(struct work_struct *work);
+static DECLARE_DELAYED_WORK(blkdev_trig_work, blkdev_trig_check);
+
+/* When is the delayed work scheduled to run next (jiffies) */
+static unsigned long blkdev_trig_next_check;
+
+/* Total number of BTB-to-BTL links */
+static unsigned int blkdev_trig_link_count;
+
+/* Empty sysfs attribute list for next 2 declarations */
+static struct attribute *blkdev_trig_attrs_empty[] = { NULL };
+
+/* linked_leds sysfs directory for block devs linked to 1 or more LEDs */
+static const struct attribute_group blkdev_trig_linked_leds = {
+	.name	= "linked_leds",
+	.attrs	= blkdev_trig_attrs_empty,
+};
+
+/* linked_devices sysfs directory for each LED associated with the trigger */
+static const struct attribute_group blkdev_trig_linked_devs = {
+	.name	= "linked_devices",
+	.attrs	= blkdev_trig_attrs_empty,
+};
+
+
+/*
+ *
+ *	Delayed work to check for activity & blink LEDs
+ *
+ */
+
+/**
+ * blkdev_trig_blink() - Blink an LED, if the correct type of activity has
+ *	occurred on the block device.
+ * @btl:	The BTL that represents the LED
+ * @btb:	The BTB that represents the block device
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ * Return:	&true if the LED is blinked, &false if not.
+ */
+static bool blkdev_trig_blink(const struct blkdev_trig_led *btl,
+			      const struct blkdev_trig_bdev *btb)
+{
+	unsigned long mode, mask, delay_on, delay_off;
+	enum stat_group i;
+
+	mode = READ_ONCE(btl->mode);
+
+	for (i = STAT_READ, mask = 1; i <= STAT_FLUSH; ++i, mask <<= 1) {
+
+		if (!(mode & mask))
+			continue;
+
+		if (time_before_eq(btb->last_activity[i], btl->last_checked))
+			continue;
+
+		delay_on = READ_ONCE(btl->blink_msec);
+		delay_off = 1;	/* 0 leaves LED turned on */
+
+		led_blink_set_oneshot(btl->led, &delay_on, &delay_off, 0);
+		return true;
+	}
+
+	return false;
+}
+
+/**
+ * blkdev_trig_update_btb() - Update a BTB's activity counters and timestamps.
+ * @btb:	The BTB
+ * @now:	Timestamp (in jiffies)
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ */
+static void blkdev_trig_update_btb(struct blkdev_trig_bdev *btb,
+				   unsigned long now)
+{
+	unsigned long new_ios;
+	enum stat_group i;
+
+	for (i = STAT_READ; i <= STAT_FLUSH; ++i) {
+
+		new_ios = part_stat_read(btb->bdev, ios[i]);
+
+		if (new_ios != btb->ios[i]) {
+			btb->ios[i] = new_ios;
+			btb->last_activity[i] = now;
+		}
+	}
+
+	btb->last_checked = now;
+}
+
+/**
+ * blkdev_trig_check() - Check linked devices for activity and blink LEDs.
+ * @work:	Delayed work (&blkdev_trig_work)
+ *
+ * Context:	Process context.  Takes and releases &blkdev_trig_mutex.
+ */
+static void blkdev_trig_check(struct work_struct *work)
+{
+	struct blkdev_trig_led *btl;
+	struct blkdev_trig_bdev *btb;
+	unsigned long index, delay, now, led_check, led_delay;
+	bool blinked;
+
+	if (!mutex_trylock(&blkdev_trig_mutex)) {
+		delay = msecs_to_jiffies(BLKDEV_TRIG_CHECK_RETRY);
+		goto exit_reschedule;
+	}
+
+	now = jiffies;
+	delay = ULONG_MAX;
+
+	hlist_for_each_entry (btl, &blkdev_trig_all_btls, all_btls_node) {
+
+		led_check = btl->last_checked + btl->check_jiffies;
+
+		if (time_before_eq(led_check, now)) {
+
+			blinked = false;
+
+			xa_for_each (&btl->linked_btbs, index, btb) {
+
+				if (btb->last_checked != now)
+					blkdev_trig_update_btb(btb, now);
+				if (!blinked)
+					blinked = blkdev_trig_blink(btl, btb);
+			}
+
+			btl->last_checked = now;
+			led_delay = btl->check_jiffies;
+
+		} else {
+			led_delay = led_check - now;
+		}
+
+		if (led_delay < delay)
+			delay = led_delay;
+	}
+
+	mutex_unlock(&blkdev_trig_mutex);
+
+exit_reschedule:
+	WARN_ON_ONCE(delay == ULONG_MAX);
+	WARN_ON_ONCE(!schedule_delayed_work(&blkdev_trig_work, delay));
+}
+
+/**
+ * blkdev_trig_sched_led() - Set the schedule of the delayed work when a new
+ *	LED is added to the schedule.
+ * @btl:	The BTL that represents the LED
+ *
+ * Called when the number of block devices to which an LED is linked becomes
+ * non-zero.
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ */
+static void blkdev_trig_sched_led(const struct blkdev_trig_led *btl)
+{
+	unsigned long delay = READ_ONCE(btl->check_jiffies);
+	unsigned long check_by = jiffies + delay;
+
+	/*
+	 * If no other LED-to-block device links exist, simply schedule the
+	 * delayed work according to this LED's check_interval attribute
+	 * (check_jiffies).
+	 */
+	if (blkdev_trig_link_count == 0) {
+		WARN_ON(!schedule_delayed_work(&blkdev_trig_work, delay));
+		blkdev_trig_next_check = check_by;
+		return;
+	}
+
+	/*
+	 * If the next check is already scheduled to occur soon enough to
+	 * accomodate this LED's check_interval, the schedule doesn't need
+	 * to be changed.
+	 */
+	if (time_after_eq(check_by, blkdev_trig_next_check))
+		return;
+
+	/*
+	 * Modify the schedule, so that the delayed work runs soon enough for
+	 * this LED.
+	 */
+	WARN_ON(!mod_delayed_work(system_wq, &blkdev_trig_work, delay));
+	blkdev_trig_next_check = check_by;
+}
+
+
+/*
+ *
+ *	Linking and unlinking LEDs and block devices
+ *
+ */
+
+/**
+ * blkdev_trig_link() - Link a block device to an LED.
+ * @btl:	The BTL that represents the LED
+ * @btb:	The BTB that represents the block device
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ * Return:	&0 on success, negative &errno on error.
+ */
+static int blkdev_trig_link(struct blkdev_trig_led *btl,
+			    struct blkdev_trig_bdev *btb)
+{
+	bool led_first_link;
+	int err;
+
+	led_first_link = xa_empty(&btl->linked_btbs);
+
+	err = xa_insert(&btb->linked_btls, btl->index, btl, GFP_KERNEL);
+	if (err)
+		return err;
+
+	err = xa_insert(&btl->linked_btbs, btb->index, btb, GFP_KERNEL);
+	if (err)
+		goto error_erase_btl;
+
+	/* Create /sys/class/block/<bdev>/linked_leds/<led> symlink */
+	err = sysfs_add_link_to_group(bdev_kobj(btb->bdev),
+				      blkdev_trig_linked_leds.name,
+				      &btl->led->dev->kobj, btl->led->name);
+	if (err)
+		goto error_erase_btb;
+
+	/* Create /sys/class/leds/<led>/linked_devices/<bdev> symlink */
+	err = sysfs_add_link_to_group(&btl->led->dev->kobj,
+				      blkdev_trig_linked_devs.name,
+				      bdev_kobj(btb->bdev),
+				      dev_name(&btb->bdev->bd_device));
+	if (err)
+		goto error_remove_symlink;
+
+	/*
+	 * If this is the first block device linked to this LED, the delayed
+	 * work schedule may need to be changed.
+	 */
+	if (led_first_link)
+		blkdev_trig_sched_led(btl);
+
+	++blkdev_trig_link_count;
+
+	return 0;
+
+error_remove_symlink:
+	sysfs_remove_link_from_group(bdev_kobj(btb->bdev),
+				     blkdev_trig_linked_leds.name,
+				     btl->led->name);
+error_erase_btb:
+	xa_erase(&btl->linked_btbs, btb->index);
+error_erase_btl:
+	xa_erase(&btb->linked_btls, btl->index);
+	return err;
+}
+
+/**
+ * blkdev_trig_put_btb() - Remove and free a BTB, if it is no longer needed.
+ * @btb:	The BTB
+ *
+ * Does nothing if the BTB (block device) is still linked to at least one LED.
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ */
+static void blkdev_trig_put_btb(struct blkdev_trig_bdev *btb)
+{
+	struct block_device *bdev = btb->bdev;
+	int err;
+
+	if (xa_empty(&btb->linked_btls)) {
+
+		sysfs_remove_group(bdev_kobj(bdev), &blkdev_trig_linked_leds);
+		err = devres_destroy(&bdev->bd_device, blkdev_trig_btb_release,
+				     NULL, NULL);
+		WARN_ON(err);
+	}
+}
+
+/**
+ * _blkdev_trig_unlink_always() - Perform the unconditionally required steps of
+ *	unlinking a block device from an LED.
+ * @btl:	The BTL that represents the LED
+ * @btb:	The BTB that represents the block device
+ *
+ * When a block device is unlinked from an LED, certain steps must be performed
+ * only if the block device is **not** being released.  This function performs
+ * those steps that are **always** required, whether or not the block device is
+ * being released.
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ */
+static void _blkdev_trig_unlink_always(struct blkdev_trig_led *btl,
+				       struct blkdev_trig_bdev *btb)
+{
+	--blkdev_trig_link_count;
+
+	if (blkdev_trig_link_count == 0)
+		WARN_ON(!cancel_delayed_work_sync(&blkdev_trig_work));
+
+	xa_erase(&btb->linked_btls, btl->index);
+	xa_erase(&btl->linked_btbs, btb->index);
+
+	/* Remove /sys/class/leds/<led>/linked_devices/<bdev> symlink */
+	sysfs_remove_link_from_group(&btl->led->dev->kobj,
+				     blkdev_trig_linked_devs.name,
+				     dev_name(&btb->bdev->bd_device));
+}
+
+/**
+ * blkdev_trig_unlink_norelease() - Unlink an LED from a block device that is
+ *	**not** being released.
+ * @btl:	The BTL that represents the LED.
+ * @btb:	The BTB that represents the block device.
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ */
+static void blkdev_trig_unlink_norelease(struct blkdev_trig_led *btl,
+					 struct blkdev_trig_bdev *btb)
+{
+	_blkdev_trig_unlink_always(btl, btb);
+
+	/* Remove /sys/class/block/<bdev>/linked_leds/<led> symlink */
+	sysfs_remove_link_from_group(bdev_kobj(btb->bdev),
+				     blkdev_trig_linked_leds.name,
+				     btl->led->name);
+
+	blkdev_trig_put_btb(btb);
+}
+
+/**
+ * blkdev_trig_unlink_release() - Unlink an LED from a block device that is
+ *	being released.
+ * @btl:	The BTL that represents the LED
+ * @btb:	The BTB that represents the block device
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ */
+static void blkdev_trig_unlink_release(struct blkdev_trig_led *btl,
+				       struct blkdev_trig_bdev *btb)
+{
+	_blkdev_trig_unlink_always(btl, btb);
+
+	/*
+	 * If the BTB is being released, the driver core has already removed the
+	 * device's attribute groups, and the BTB will be freed automatically,
+	 * so there's nothing else to do.
+	 */
+}
+
+
+/*
+ *
+ *	BTB creation
+ *
+ */
+
+/**
+ * blkdev_trig_btb_release() - BTB device resource release function.
+ * @dev:	The block device
+ * @res:	The BTB
+ *
+ * Called by the driver core when a block device with a BTB is removed.
+ *
+ * Context:	Process context.  Takes and releases &blkdev_trig_mutex.
+ */
+static void blkdev_trig_btb_release(struct device *dev, void *res)
+{
+	struct blkdev_trig_bdev *btb = res;
+	struct blkdev_trig_led *btl;
+	unsigned long index;
+
+	mutex_lock(&blkdev_trig_mutex);
+
+	xa_for_each (&btb->linked_btls, index, btl)
+		blkdev_trig_unlink_release(btl, btb);
+
+	mutex_unlock(&blkdev_trig_mutex);
+}
+
+/**
+ * blkdev_trig_get_bdev() - Get a block device by path.
+ * @path:	The value written to an LED's &link_dev_by_path or
+ *		&unlink_dev_by_path attribute, which should be the path to a
+ *		special file that represents a block device
+ * @len:	The number of characters in &path (not including its
+ *		terminating null)
+ *
+ * The caller must call blkdev_put() when finished with the device.
+ *
+ * Context:	Process context.
+ * Return:	The block device, or an error pointer.
+ */
+static struct block_device *blkdev_trig_get_bdev(const char *path, size_t len)
+{
+	struct block_device *bdev;
+	char *buf;
+
+	buf = kmemdup(path, len + 1, GFP_KERNEL);  /* +1 to include null */
+	if (buf == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	bdev = blkdev_get_by_path(strim(buf), BLKDEV_TRIG_FMODE, THIS_MODULE);
+	kfree(buf);
+	return bdev;
+}
+
+/**
+ * blkdev_trig_get_btb() - Find or create the BTB for a block device.
+ * @path:	The value written to an LED's &link_dev_by_path attribute,
+ *		which should be the path to a special file that represents a
+ *		block device
+ * @len:	The number of characters in &path
+ *
+ * If a new BTB is created, because the block device was not previously linked
+ * to any LEDs, the block device's &linked_leds &sysfs directory is created.
+ *
+ * Context:	Process context.  Caller must hold &blkdev_trig_mutex.
+ * Return:	Pointer to the BTB, error pointer on error.
+ */
+static struct blkdev_trig_bdev *blkdev_trig_get_btb(const char *path,
+						    size_t len)
+{
+	struct block_device *bdev;
+	struct blkdev_trig_bdev *btb;
+	int err;
+
+	bdev = blkdev_trig_get_bdev(path, len);
+	if (IS_ERR(bdev))
+		return ERR_CAST(bdev);
+
+	btb = devres_find(&bdev->bd_device, blkdev_trig_btb_release,
+			  NULL, NULL);
+	if (btb != NULL) {
+		err = 0;
+		goto exit_put_bdev;
+	}
+
+	if (blkdev_trig_next_index == ULONG_MAX) {
+		err = -EOVERFLOW;
+		goto exit_put_bdev;
+	}
+
+	btb = devres_alloc(blkdev_trig_btb_release, sizeof(*btb), GFP_KERNEL);
+	if (btb == NULL) {
+		err = -ENOMEM;
+		goto exit_put_bdev;
+	}
+
+	err = sysfs_create_group(bdev_kobj(bdev), &blkdev_trig_linked_leds);
+	if (err)
+		goto exit_free_btb;
+
+	btb->index = blkdev_trig_next_index++;
+	btb->bdev = bdev;
+	xa_init(&btb->linked_btls);
+
+	/* Populate BTB activity counters */
+	blkdev_trig_update_btb(btb, jiffies);
+
+	devres_add(&bdev->bd_device, btb);
+
+exit_free_btb:
+	if (err)
+		devres_free(btb);
+exit_put_bdev:
+	blkdev_put(bdev, BLKDEV_TRIG_FMODE);
+	return err ? ERR_PTR(err) : btb;
+}
+
+
+/*
+ *
+ *	Activating and deactivating the trigger on an LED
+ *
+ */
+
+/**
+ * blkdev_trig_activate() - Called by the LEDs subsystem when an LED is
+ *	associated with the trigger.
+ * @led:	The LED
+ *
+ * Context:	Process context.  Takes and releases &blkdev_trig_mutex.
+ * Return:	&0 on success, negative &errno on error.
+ */
+static int blkdev_trig_activate(struct led_classdev *led)
+{
+	struct blkdev_trig_led *btl;
+	int err;
+
+	btl = kzalloc(sizeof(*btl), GFP_KERNEL);
+	if (btl == NULL)
+		return -ENOMEM;
+
+	err = mutex_lock_interruptible(&blkdev_trig_mutex);
+	if (err)
+		goto exit_free;
+
+	if (blkdev_trig_next_index == ULONG_MAX) {
+		err = -EOVERFLOW;
+		goto exit_unlock;
+	}
+
+	btl->index = blkdev_trig_next_index++;
+	btl->last_checked = jiffies;
+	btl->mode = -1;  /* set all bits */
+	btl->led = led;
+	btl->blink_msec = BLKDEV_TRIG_BLINK_DEF;
+	btl->check_jiffies = msecs_to_jiffies(BLKDEV_TRIG_CHECK_DEF);
+	xa_init(&btl->linked_btbs);
+
+	hlist_add_head(&btl->all_btls_node, &blkdev_trig_all_btls);
+	led_set_trigger_data(led, btl);
+
+exit_unlock:
+	mutex_unlock(&blkdev_trig_mutex);
+exit_free:
+	if (err)
+		kfree(btl);
+	return err;
+}
+
+/**
+ * blkdev_trig_deactivate() - Called by the the LEDs subsystem when an LED is
+ *	disassociated from the trigger.
+ * @led:	The LED
+ *
+ * The LEDs subsystem also calls this function when an LED associated with the
+ * trigger is removed or when the trigger is unregistered (if the module is
+ * unloaded).
+ *
+ * Context:	Process context.  Takes and releases &blkdev_trig_mutex.
+ */
+static void blkdev_trig_deactivate(struct led_classdev *led)
+{
+	struct blkdev_trig_led *btl = led_get_trigger_data(led);
+	struct blkdev_trig_bdev *btb;
+	unsigned long index;
+
+	mutex_lock(&blkdev_trig_mutex);
+
+	xa_for_each (&btl->linked_btbs, index, btb)
+		blkdev_trig_unlink_norelease(btl, btb);
+
+	hlist_del(&btl->all_btls_node);
+	kfree(btl);
+
+	mutex_unlock(&blkdev_trig_mutex);
+}
+
+
+/*
+ *
+ *	Link-related attribute store functions
+ *
+ */
+
+/**
+ * link_dev_by_path_store() - &link_dev_by_path device attribute store function.
+ * @dev:	The LED device
+ * @attr:	The &link_dev_by_path attribute (&dev_attr_link_dev_by_path)
+ * @buf:	The value written to the attribute, which should be the path to
+ *		a special file that represents a block device to be linked to
+ *		the LED (e.g. ``/dev/sda``)
+ * @count:	The number of characters in &buf
+ *
+ * Context:	Process context.  Takes and releases &blkdev_trig_mutex.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t link_dev_by_path_store(struct device *dev,
+				      struct device_attribute *attr,
+				      const char *buf, size_t count)
+{
+	struct blkdev_trig_led *btl = led_trigger_get_drvdata(dev);
+	struct blkdev_trig_bdev *btb;
+	int err;
+
+	err = mutex_lock_interruptible(&blkdev_trig_mutex);
+	if (err)
+		return err;
+
+	btb = blkdev_trig_get_btb(buf, count);
+	if (IS_ERR(btb)) {
+		err = PTR_ERR(btb);
+		goto exit_unlock;
+	}
+
+	if (xa_load(&btb->linked_btls, btl->index) != NULL) {
+		err = -EEXIST;
+		goto exit_put_btb;
+	}
+
+	err = blkdev_trig_link(btl, btb);
+
+exit_put_btb:
+	if (err)
+		blkdev_trig_put_btb(btb);
+exit_unlock:
+	mutex_unlock(&blkdev_trig_mutex);
+	return err ? : count;
+}
+
+/**
+ * unlink_dev_by_path_store() - &unlink_dev_by_path device attribute store
+ *	function.
+ * @dev:	The LED device
+ * @attr:	The &unlink_dev_by_path attribute (&dev_attr_unlink_dev_by_path)
+ * @buf:	The value written to the attribute, which should be the path to
+ *		a special file that represents a block device to be unlinked
+ *		from the LED (e.g. ``/dev/sda``)
+ * @count:	The number of characters in &buf
+ *
+ * Context:	Process context.  Takes and releases &blkdev_trig_mutex.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t unlink_dev_by_path_store(struct device *dev,
+					struct device_attribute *attr,
+					const char *buf, size_t count)
+{
+	struct blkdev_trig_led *btl = led_trigger_get_drvdata(dev);
+	struct block_device *bdev;
+	struct blkdev_trig_bdev *btb;
+	int err;
+
+	bdev = blkdev_trig_get_bdev(buf, count);
+	if (IS_ERR(bdev))
+		return PTR_ERR(bdev);
+
+	err = mutex_lock_interruptible(&blkdev_trig_mutex);
+	if (err)
+		goto exit_put_bdev;
+
+	btb = devres_find(&bdev->bd_device, blkdev_trig_btb_release,
+			  NULL, NULL);
+	if (btb == NULL) {
+		err = -EUNATCH;  /* bdev isn't linked to any LED */
+		goto exit_unlock;
+	}
+
+	if (xa_load(&btb->linked_btls, btl->index) == NULL) {
+		err = -EUNATCH;  /* bdev isn't linked to this LED */
+		goto exit_unlock;
+	}
+
+	blkdev_trig_unlink_norelease(btl, btb);
+
+exit_unlock:
+	mutex_unlock(&blkdev_trig_mutex);
+exit_put_bdev:
+	blkdev_put(bdev, BLKDEV_TRIG_FMODE);
+	return err ? : count;
+}
+
+/**
+ * unlink_dev_by_name_store() - &unlink_dev_by_name device attribute store
+ *	function.
+ * @dev:	The LED device
+ * @attr:	The &unlink_dev_by_name attribute (&dev_attr_unlink_dev_by_name)
+ * @buf:	The value written to the attribute, which should be the kernel
+ *		name of a block device to be unlinked from the LED (e.g.
+ *		``sda``)
+ * @count:	The number of characters in &buf
+ *
+ * Context:	Process context.  Takes and releases &blkdev_trig_mutex.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t unlink_dev_by_name_store(struct device *dev,
+					struct device_attribute *attr,
+					const char *buf, size_t count)
+{
+	struct blkdev_trig_led *btl = led_trigger_get_drvdata(dev);
+	struct blkdev_trig_bdev *btb;
+	unsigned long index;
+	int err;
+
+	err = mutex_lock_interruptible(&blkdev_trig_mutex);
+	if (err)
+		return err;
+
+	err = -EUNATCH;
+
+	xa_for_each (&btl->linked_btbs, index, btb) {
+
+		if (sysfs_streq(dev_name(&btb->bdev->bd_device), buf)) {
+			blkdev_trig_unlink_norelease(btl, btb);
+			err = 0;
+			break;
+		}
+	}
+
+	mutex_unlock(&blkdev_trig_mutex);
+	return err ? : count;
+}
+
+
+/*
+ *
+ *	Atomic attribute show & store functions
+ *
+ */
+
+/**
+ * blink_time_show() - &blink_time device attribute show function.
+ * @dev:	The LED device
+ * @attr:	The &blink_time attribute (&dev_attr_blink_time)
+ * @buf:	Output buffer
+ *
+ * Writes the value of &blkdev_trig_led.blink_msec to &buf.
+ *
+ * Context:	Process context.
+ * Return:	The number of characters written to &buf.
+ */
+static ssize_t blink_time_show(struct device *dev,
+			       struct device_attribute *attr, char *buf)
+{
+	const struct blkdev_trig_led *btl = led_trigger_get_drvdata(dev);
+
+	return sysfs_emit(buf, "%u\n", READ_ONCE(btl->blink_msec));
+}
+
+/**
+ * blink_time_store() - &blink_time device attribute store function.
+ * @dev:	The LED device
+ * @attr:	The &blink_time attribute (&dev_attr_blink_time)
+ * @buf:	The new value (as written to the &sysfs attribute)
+ * @count:	The number of characters in &buf
+ *
+ * Sets &blkdev_trig_led.blink_msec to the value in &buf.
+ *
+ * Context:	Process context.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t blink_time_store(struct device *dev,
+				struct device_attribute *attr,
+				const char *buf, size_t count)
+{
+	struct blkdev_trig_led *btl = led_trigger_get_drvdata(dev);
+	unsigned int value;
+	int err;
+
+	err = kstrtouint(buf, 0, &value);
+	if (err)
+		return err;
+
+	if (value < BLKDEV_TRIG_BLINK_MIN || value > BLKDEV_TRIG_BLINK_MAX)
+		return -ERANGE;
+
+	WRITE_ONCE(btl->blink_msec, value);
+	return count;
+}
+
+/**
+ * check_interval_show() - &check_interval device attribute show function.
+ * @dev:	The LED device
+ * @attr:	The &check_interval attribute (&dev_attr_check_interval)
+ * @buf:	Output buffer
+ *
+ * Writes the value of &blkdev_trig_led.check_jiffies (converted to
+ * milliseconds) to &buf.
+ *
+ * Context:	Process context.
+ * Return:	The number of characters written to &buf.
+ */
+static ssize_t check_interval_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	struct blkdev_trig_led *btl = led_trigger_get_drvdata(dev);
+
+	return sysfs_emit(buf, "%u\n",
+			  jiffies_to_msecs(READ_ONCE(btl->check_jiffies)));
+}
+
+/**
+ * check_interval_store() - &check_interval device attribute store function
+ * @dev:	The LED device
+ * @attr:	The &check_interval attribute (&dev_attr_check_interval)
+ * @buf:	The new value (as written to the &sysfs attribute)
+ * @count:	The number of characters in &buf
+ *
+ * Sets &blkdev_trig_led.check_jiffies to the value in &buf (after converting
+ * from milliseconds).
+ *
+ * Context:	Process context.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t check_interval_store(struct device *dev,
+				    struct device_attribute *attr,
+				    const char *buf, size_t count)
+{
+	struct blkdev_trig_led *led = led_trigger_get_drvdata(dev);
+	unsigned int value;
+	int err;
+
+	err = kstrtouint(buf, 0, &value);
+	if (err)
+		return err;
+
+	if (value < BLKDEV_TRIG_CHECK_MIN || value > BLKDEV_TRIG_CHECK_MAX)
+		return -ERANGE;
+
+	WRITE_ONCE(led->check_jiffies, msecs_to_jiffies(value));
+
+	return count;
+}
+
+/**
+ * blkdev_trig_mode_show() - Helper for boolean attribute show functions.
+ * @led:	The LED
+ * @buf:	Output buffer
+ * @bit:	Which bit to show
+ *
+ * Context:	Process context.
+ * Return:	The number of characters written to &buf.
+ */
+static int blkdev_trig_mode_show(const struct blkdev_trig_led *led, char *buf,
+				 enum stat_group bit)
+{
+	return sysfs_emit(buf,
+			  READ_ONCE(led->mode) & (1 << bit) ? "Y\n" : "N\n");
+}
+
+/**
+ * blkdev_trig_mode_store() - Helper for boolean attribute store functions.
+ * @led:	The LED
+ * @buf:	The new value (as written to the &sysfs attribute)
+ * @count:	The number of characters in &buf
+ * @bit:	Which bit to set
+ *
+ * Context:	Process context.
+ * Return:	&count on success, negative &errno on error.
+ */
+static int blkdev_trig_mode_store(struct blkdev_trig_led *led,
+				  const char *buf, size_t count,
+				  enum stat_group bit)
+{
+	bool set;
+	int err;
+
+	err = kstrtobool(buf, &set);
+	if (err)
+		return err;
+
+	if (set)
+		set_bit(bit, &led->mode);
+	else
+		clear_bit(bit, &led->mode);
+
+	return count;
+}
+
+/**
+ * blink_on_read_show() - &blink_on_read device attribute show function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_read attribute (&dev_attr_blink_on_read)
+ * @buf:	Output buffer
+ *
+ * Writes ``Y`` or ``N`` to &buf, depending on whether the &STAT_READ bit in
+ * &blkdev_trig_led.mode is set or cleared.
+ *
+ * Context:	Process context.
+ * Return:	The number of characters written to &buf.
+ */
+static ssize_t blink_on_read_show(struct device *dev,
+				  struct device_attribute *attr, char *buf)
+{
+	return blkdev_trig_mode_show(led_trigger_get_drvdata(dev),
+				     buf, STAT_READ);
+}
+
+/**
+ * blink_on_read_store() - &blink_on_read device attribute store function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_read attribute (&dev_attr_blink_on_read)
+ * @buf:	The new value (as written to the &sysfs attribute)
+ * @count:	The number of characters in &buf
+ *
+ * Sets the &STAT_READ bit in &blkdev_trig_led.mode to the value in &buf
+ * (interpretted as a boolean).
+ *
+ * Context:	Process context.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t blink_on_read_store(struct device *dev,
+				   struct device_attribute *attr,
+				   const char *buf, size_t count)
+{
+	return blkdev_trig_mode_store(led_trigger_get_drvdata(dev),
+				      buf, count, STAT_READ);
+}
+
+/**
+ * blink_on_write_show() - &blink_on_write device attribute show function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_write attribute (&dev_attr_blink_on_write)
+ * @buf:	Output buffer
+ *
+ * Writes ``Y`` or ``N`` to &buf, depending on whether the &STAT_WRITE bit in
+ * in &blkdev_trig_led.mode is set or cleared.
+ *
+ * Context:	Process context.
+ * Return:	The number of characters written to &buf.
+ */
+static ssize_t blink_on_write_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	return blkdev_trig_mode_show(led_trigger_get_drvdata(dev),
+				     buf, STAT_WRITE);
+}
+
+/**
+ * blink_on_write_store() - &blink_on_write device attribute store function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_write attribute (&dev_attr_blink_on_write)
+ * @buf:	The new value (as written to the &sysfs attribute)
+ * @count:	The number of characters in &buf
+ *
+ * Sets the &STAT_WRITE bit in &blkdev_trig_led.mode to the value in &buf
+ * (interpretted as a boolean).
+ *
+ * Context:	Process context.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t blink_on_write_store(struct device *dev,
+				    struct device_attribute *attr,
+				    const char *buf, size_t count)
+{
+	return blkdev_trig_mode_store(led_trigger_get_drvdata(dev),
+				      buf, count, STAT_WRITE);
+}
+
+/**
+ * blink_on_flush_show() - &blink_on_flush device attribute show function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_flush attribute (&dev_attr_blink_on_flush)
+ * @buf:	Output buffer
+ *
+ * Writes ``Y`` or ``N`` to &buf, depending whether the &STAT_FLUSH bit in
+ * &blkdev_trig_led.mode is set or cleared.
+ *
+ * Context:	Process context.
+ * Return:	The number of characters written to &buf.
+ */
+static ssize_t blink_on_flush_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	return blkdev_trig_mode_show(led_trigger_get_drvdata(dev),
+				     buf, STAT_FLUSH);
+}
+
+/**
+ * blink_on_flush_store() - &blink_on_flush device attribute store function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_flush attribute (&dev_attr_blink_on_flush)
+ * @buf:	The new value (as written to the &sysfs attribute)
+ * @count:	The number of characters in &buf
+ *
+ * Sets the &STAT_FLUSH bit in &blkdev_trig_led.mode to the value in &buf
+ * (interpretted as a boolean).
+ *
+ * Context:	Process context.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t blink_on_flush_store(struct device *dev,
+				    struct device_attribute *attr,
+				    const char *buf, size_t count)
+{
+	return blkdev_trig_mode_store(led_trigger_get_drvdata(dev),
+				      buf, count, STAT_FLUSH);
+}
+
+/**
+ * blink_on_discard_show() - &blink_on_discard device attribute show function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_discard attribute (&dev_attr_blink_on_discard)
+ * @buf:	Output buffer
+ *
+ * Writes ``Y`` or ``N`` to &buf, depending on whether the &STAT_DISCARD bit in
+ * &blkdev_trig_led.mode is set or cleared.
+ *
+ * Context:	Process context.
+ * Return:	The number of characters written to &buf.
+ */
+static ssize_t blink_on_discard_show(struct device *dev,
+				     struct device_attribute *attr, char *buf)
+{
+	return blkdev_trig_mode_show(led_trigger_get_drvdata(dev),
+				     buf, STAT_DISCARD);
+}
+
+/**
+ * blink_on_discard_store() - &blink_on_discard device attribute store function.
+ * @dev:	The LED device
+ * @attr:	The &blink_on_discard attribute (&dev_attr_blink_on_discard)
+ * @buf:	The new value (as written to the &sysfs attribute)
+ * @count:	The number of characters in &buf
+ *
+ * Sets the &STAT_DISCARD bit in &blkdev_trig_led.mode to the value in &buf
+ * (interpretted as a boolean).
+ *
+ * Context:	Process context.
+ * Return:	&count on success, negative &errno on error.
+ */
+static ssize_t blink_on_discard_store(struct device *dev,
+				      struct device_attribute *attr,
+				      const char *buf, size_t count)
+{
+	return blkdev_trig_mode_store(led_trigger_get_drvdata(dev),
+				      buf, count, STAT_DISCARD);
+}
+
+/* Device attributes */
+static DEVICE_ATTR_WO(link_dev_by_path);
+static DEVICE_ATTR_WO(unlink_dev_by_path);
+static DEVICE_ATTR_WO(unlink_dev_by_name);
+static DEVICE_ATTR_RW(blink_time);
+static DEVICE_ATTR_RW(check_interval);
+static DEVICE_ATTR_RW(blink_on_read);
+static DEVICE_ATTR_RW(blink_on_write);
+static DEVICE_ATTR_RW(blink_on_flush);
+static DEVICE_ATTR_RW(blink_on_discard);
+
+/* Device attributes in LED directory (/sys/class/leds/<led>/...) */
+static struct attribute *blkdev_trig_attrs[] = {
+	&dev_attr_link_dev_by_path.attr,
+	&dev_attr_unlink_dev_by_path.attr,
+	&dev_attr_unlink_dev_by_name.attr,
+	&dev_attr_blink_time.attr,
+	&dev_attr_check_interval.attr,
+	&dev_attr_blink_on_read.attr,
+	&dev_attr_blink_on_write.attr,
+	&dev_attr_blink_on_flush.attr,
+	&dev_attr_blink_on_discard.attr,
+	NULL
+};
+
+/* Unnamed attribute group == no subdirectory */
+static const struct attribute_group blkdev_trig_attr_group = {
+	.attrs	= blkdev_trig_attrs,
+};
+
+/* Attribute groups for the trigger */
+static const struct attribute_group *blkdev_trig_attr_groups[] = {
+	&blkdev_trig_attr_group,   /* /sys/class/leds/<led>/... */
+	&blkdev_trig_linked_devs,  /* /sys/class/leds/<led>/linked_devices/ */
+	NULL
+};
+
+/* Trigger registration data */
+static struct led_trigger blkdev_trig_trigger = {
+	.name		= "blkdev",
+	.activate	= blkdev_trig_activate,
+	.deactivate	= blkdev_trig_deactivate,
+	.groups		= blkdev_trig_attr_groups,
+};
+
+/**
+ * blkdev_trig_init() - Block device LED trigger initialization.
+ *
+ * Registers the ``blkdev`` LED trigger.
+ *
+ * Return:	&0 on success, negative &errno on failure.
+ */
+static int __init blkdev_trig_init(void)
+{
+	return led_trigger_register(&blkdev_trig_trigger);
+}
+module_init(blkdev_trig_init);
+
+/**
+ * blkdev_trig_exit() - Block device LED trigger module exit.
+ *
+ * Unregisters the ``blkdev`` LED trigger.
+ */
+static void __exit blkdev_trig_exit(void)
+{
+	led_trigger_unregister(&blkdev_trig_trigger);
+}
+module_exit(blkdev_trig_exit);
+
+MODULE_DESCRIPTION("Block device LED trigger");
+MODULE_AUTHOR("Ian Pilcher <arequipeno@gmail.com>");
+MODULE_LICENSE("GPL v2");
-- 
2.40.1.445.gf85cd430b1

From 3e940ebfd882add5ea49b8e26797a1105fe78810 Mon Sep 17 00:00:00 2001
From: Ian Pilcher <arequipeno@gmail.com>
Date: Tue, 27 Dec 2022 16:52:25 -0600
Subject: [PATCH 10/24] docs: Add block device (blkdev) LED trigger
 documentation

Add Documentation/ABI/testing/sysfs-class-led-trigger-blkdev to
document:

* /sys/class/leds/<led>/blink_time
  * /sys/class/leds/<led>/check_interval
  * /sys/class/leds/<led>/blink_on_{read,write,discard,flush}
  * /sys/class/leds/<led>/link_dev_by_path
  * /sys/class/leds/<led>/unlink_dev_by_path
  * /sys/class/leds/<led>/unlink_dev_by_name
  * /sys/class/leds/<led>/linked_devices

Add /sys/block/<disk>/linked_leds to
Documentation/ABI/stable/sysfs-block.

Add overview in Documentation/leds/ledtrig-blkdev.rst.

Signed-off-by: Ian Pilcher <arequipeno@gmail.com>
---
 Documentation/ABI/stable/sysfs-block          |  10 ++
 .../testing/sysfs-class-led-trigger-blkdev    |  78 +++++++++
 Documentation/leds/index.rst                  |   1 +
 Documentation/leds/ledtrig-blkdev.rst         | 158 ++++++++++++++++++
 4 files changed, 247 insertions(+)
 create mode 100644 Documentation/ABI/testing/sysfs-class-led-trigger-blkdev
 create mode 100644 Documentation/leds/ledtrig-blkdev.rst

diff --git a/Documentation/ABI/stable/sysfs-block b/Documentation/ABI/stable/sysfs-block
index 282de3680..ac1dd2fbd 100644
--- a/Documentation/ABI/stable/sysfs-block
+++ b/Documentation/ABI/stable/sysfs-block
@@ -101,6 +101,16 @@ Description:
 		devices that support receiving integrity metadata.
 
 
+What:		/sys/block/<disk>/linked_leds
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Directory that contains symbolic links to all LEDs that
+		are associated with (linked to) this block device by the
+		blkdev LED trigger.  Only present when at least one LED
+		is linked.  (See Documentation/leds/ledtrig-blkdev.rst.)
+
+
 What:		/sys/block/<disk>/<partition>/alignment_offset
 Date:		April 2009
 Contact:	Martin K. Petersen <martin.petersen@oracle.com>
diff --git a/Documentation/ABI/testing/sysfs-class-led-trigger-blkdev b/Documentation/ABI/testing/sysfs-class-led-trigger-blkdev
new file mode 100644
index 000000000..28ce8c814
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-class-led-trigger-blkdev
@@ -0,0 +1,78 @@
+What:		/sys/class/leds/<led>/blink_time
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Time (in milliseconds) that the LED will be on during a single
+		"blink".
+
+What:		/sys/class/leds/<led>/check_interval
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Interval (in milliseconds) between checks of the block devices
+		linked to this LED.  The LED will be blinked if the correct type
+		of activity (see blink_on_{read,write,discard,flush} attributes)
+		has occurred on any of the linked devices since the previous
+		check.
+
+What:		/sys/class/leds/<led>/blink_on_read
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Boolean that determines whether the LED will blink in response
+		to read activity on any of its linked block devices.
+
+What:		/sys/class/leds/<led>/blink_on_write
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Boolean that determines whether the LED will blink in response
+		to write activity on any of its linked block devices.
+
+What:		/sys/class/leds/<led>/blink_on_discard
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Boolean that determines whether the LED will blink in response
+		to discard activity on any of its linked block devices.
+
+What:		/sys/class/leds/<led>/blink_on_flush
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gamil.com>
+Description:
+		Boolean that determines whether the LED will blink in response
+		to cache flush activity on any of its linked block devices.
+
+What:		/sys/class/leds/<led>/link_dev_by_path
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Associate a block device with this LED by writing the path to
+		the device special file (e.g. /dev/sda) to this attribute.
+		Symbolic links are followed.
+
+What:		/sys/class/leds/<led>/unlink_dev_by_path
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Remove the association between this LED and a block device by
+		writing the path to the device special file (e.g. /dev/sda) to
+		this attribute.  Symbolic links are followed.
+
+What:		/sys/class/leds/<led>/unlink_dev_by_name
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Remove the association between this LED and a block device by
+		writing the kernel name of the device (e.g. sda) to this
+		attribute.
+
+What:		/sys/class/leds/<led>/linked_devices
+Date:		January 2023
+Contact:	Ian Pilcher <arequipeno@gmail.com>
+Description:
+		Directory containing links to all block devices that are
+		associated with this LED.  (Note that the names of the
+		symbolic links in this directory are *kernel* names, which
+		may not match the device special file paths written to
+		link_device and unlink_device.)
diff --git a/Documentation/leds/index.rst b/Documentation/leds/index.rst
index b9ca081fa..5e37d8e7b 100644
--- a/Documentation/leds/index.rst
+++ b/Documentation/leds/index.rst
@@ -10,6 +10,7 @@ LEDs
    leds-class
    leds-class-flash
    leds-class-multicolor
+   ledtrig-blkdev
    ledtrig-oneshot
    ledtrig-transient
    ledtrig-usbport
diff --git a/Documentation/leds/ledtrig-blkdev.rst b/Documentation/leds/ledtrig-blkdev.rst
new file mode 100644
index 000000000..9ff5b99de
--- /dev/null
+++ b/Documentation/leds/ledtrig-blkdev.rst
@@ -0,0 +1,158 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+=================================
+Block Device (blkdev) LED Trigger
+=================================
+
+Available when ``CONFIG_LEDS_TRIGGER_BLKDEV=y`` or
+``CONFIG_LEDS_TRIGGER_BLKDEV=m``.
+
+See also:
+
+* ``Documentation/ABI/testing/sysfs-class-led-trigger-blkdev``
+* ``Documentation/ABI/stable/sysfs-block`` (``/sys/block/<disk>/linked_leds``)
+
+Overview
+========
+
+.. note::
+	The examples below use ``<LED>`` to refer to the name of a
+	system-specific LED.  If no suitable LED is available on a test
+	system (in a virtual machine, for example), it is possible to
+	use a userspace LED.  (See ``Documentation/leds/uleds.rst``.)
+
+Verify that the ``blkdev`` LED trigger is available::
+
+	# grep blkdev /sys/class/leds/<LED>/trigger
+	... rfkill-none blkdev
+
+(If the previous command produces no output, you may need to load the trigger
+module - ``modprobe ledtrig_blkdev``.  If the module is not available, check
+the value of ``CONFIG_LEDS_TRIGGER_BLKDEV`` in your kernel configuration.)
+
+Associate the LED with the ``blkdev`` LED trigger::
+
+	# echo blkdev > /sys/class/leds/<LED>/trigger
+
+	# cat /sys/class/leds/<LED>/trigger
+	... rfkill-none [blkdev]
+
+Note that several new device attributes are available in the
+``/sys/class/leds/<LED>`` directory.
+
+* ``link_dev_by_path``, ``unlink_dev_by_path``, and ``unlink_dev_by_name`` are
+  used to manage the set of block devices associated with this LED.  The LED
+  will blink when activity occurs on any of its linked devices.
+
+* ``blink_on_read``, ``blink_on_write``, ``blink_on_discard``, and
+  ``blink_on_flush`` are boolean values that determine whether the LED will
+  blink when a particular type of activity is detected on one of its linked
+  block devices.
+
+* ``blink_time`` is the duration (in milliseconds) of each blink of this LED.
+  (The minimum value is 10 milliseconds.)
+
+* ``check_interval`` is the frequency (in milliseconds) with which block devices
+  linked to this LED will be checked for activity and the LED blinked (if the
+  correct type of activity has occurred).
+
+* The ``linked_devices`` directory will contain a symbolic link to every device
+  that is associated with this LED.
+
+Link a block device to the LED::
+
+	# echo /dev/sda > /sys/class/leds/<LED>/link_dev_by_path
+
+	# ls /sys/class/leds/<LED>/linked_devices
+	sda
+
+(The value written to ``link_dev_by_path`` must be the path of the device
+special file, such as ``/dev/sda``, that represents the block device - or the
+path of a symbolic link to such a device special file.)
+
+Activity on the device will now cause the LED to blink.  The duration of each
+blink (in milliseconds) can be adjusted by setting
+``/sys/class/leds/<LED>/blink_time``.  (But see **check_interval and
+blink_time** below.)
+
+Associate a second device with the LED::
+
+	# echo /dev/sdb > /sys/class/leds/<LED>/link_dev_by_path
+
+	# ls /sys/class/leds/<LED>/linked_devices
+	sda  sdb
+
+When a block device is linked to one or more LEDs, the LEDs are linked from
+the device's ``linked_leds`` directory::
+
+	# ls /sys/class/block/sd{a,b}/linked_leds
+	/sys/class/block/sda/linked_leds:
+	<LED>
+
+	/sys/class/block/sdb/linked_leds:
+	<LED>
+
+(The ``linked_leds`` directory only exists when the block device is linked to
+at least one LED.)
+
+``check_interval`` and ``blink_time``
+=====================================
+
+* By default, linked block devices are checked for activity every 100
+  milliseconds.  This frequency can be changed for an LED via the
+  ``/sys/class/leds/<led>/check_interval`` attribute.  (The minimum value is 25
+  milliseconds.)
+
+* All block devices associated with an LED are checked for activity every
+  ``check_interval`` milliseconds, and a blink is triggered if the correct type
+  of activity (as determined by the LED's ``blink_on_*`` attributes) is
+  detected.  The duration of an LED's blink is determined by its ``blink_time``
+  attribute.  Thus (when the correct type of activity is detected), the LED will
+  be on for ``blink_time`` milliseconds and off for
+  ``check_interval - blink_time`` milliseconds.
+
+* The LED subsystem ignores new blink requests for an LED that is already in
+  in the process of blinking, so setting a ``blink_time`` greater than or equal
+  to ``check_interval`` will cause some blinks to be missed.
+
+* Because of processing times, scheduling latencies, etc., avoiding missed
+  blinks actually requires a difference of at least a few milliseconds between
+  the ``blink_time`` and ``check_interval``.  The required difference is likely
+  to vary from system to system.  As a  reference, a Thecus N5550 NAS requires a
+  difference of 7 milliseconds (e.g. ``check_interval == 100``,
+  ``blink_time == 93``).
+
+* The default values (``check_interval == 100``, ``blink_time == 75``) cause the
+  LED associated with a continuously active device to blink rapidly.  For a more
+  "always on" effect, increase the ``blink_time`` (but not too much; see the
+  previous bullet).
+
+Other Notes
+===========
+
+* Many (possibly all) types of block devices work with this trigger, including:
+
+  * SCSI (including SATA and USB) hard disk drives and SSDs
+  * SCSI (including SATA and USB) optical drives
+  * NVMe SSDs
+  * SD cards
+  * loopback block devices (``/dev/loop*``)
+  * device mapper devices, such as LVM logical volumes
+  * MD RAID devices
+  * zRAM compressed RAM-disks
+  * partitions on block devices that support them
+
+* The names of the symbolic links in ``/sys/class/leds/<LED>/linked_devices``
+  are **kernel** names, which may not match the paths used for
+  ``link_dev_by_path`` and ``unlink_dev_by_path``.  This is most likely when a
+  symbolic link is used to refer to the device (as is common with logical
+  volumes), but it can be true for any device, because nothing prevents the
+  creation of device special files with arbitrary names (e.g.
+  ``sudo mknod /foo b 8 0``).
+
+  Kernel names can be used to unlink block devices from LEDs by writing them to
+  the LED's ``unlink_dev_by_name`` attribute.
+
+* The ``blkdev`` LED trigger supports many-to-many device/LED associations.
+  A device can be associated with multiple LEDs, and an LED can be associated
+  with multiple devices.
-- 
2.40.1.445.gf85cd430b1

From 49e8518e778d7ab40d2754f5c4201f71c71e5989 Mon Sep 17 00:00:00 2001
From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Sat, 25 Feb 2023 22:44:10 +0100
Subject: [PATCH 11/24] r8169: use spinlock to protect mac ocp register access

For disabling ASPM during NAPI poll we'll have to access mac ocp
registers in atomic context. This could result in races because
a mac ocp read consists of a write to register OCPDR, followed
by a read from the same register. Therefore add a spinlock to
protext access to mac ocp registers.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 37 ++++++++++++++++++++---
 1 file changed, 33 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 45147a101..259eac5b0 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -613,6 +613,8 @@ struct rtl8169_private {
 		struct work_struct work;
 	} wk;
 
+	spinlock_t mac_ocp_lock;
+
 	unsigned supports_gmii:1;
 	unsigned aspm_manageable:1;
 	dma_addr_t counters_phys_addr;
@@ -847,7 +849,7 @@ static int r8168_phy_ocp_read(struct rtl8169_private *tp, u32 reg)
 		(RTL_R32(tp, GPHY_OCP) & 0xffff) : -ETIMEDOUT;
 }
 
-static void r8168_mac_ocp_write(struct rtl8169_private *tp, u32 reg, u32 data)
+static void __r8168_mac_ocp_write(struct rtl8169_private *tp, u32 reg, u32 data)
 {
 	if (rtl_ocp_reg_failure(reg))
 		return;
@@ -855,7 +857,16 @@ static void r8168_mac_ocp_write(struct rtl8169_private *tp, u32 reg, u32 data)
 	RTL_W32(tp, OCPDR, OCPAR_FLAG | (reg << 15) | data);
 }
 
-static u16 r8168_mac_ocp_read(struct rtl8169_private *tp, u32 reg)
+static void r8168_mac_ocp_write(struct rtl8169_private *tp, u32 reg, u32 data)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&tp->mac_ocp_lock, flags);
+	__r8168_mac_ocp_write(tp, reg, data);
+	spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
+}
+
+static u16 __r8168_mac_ocp_read(struct rtl8169_private *tp, u32 reg)
 {
 	if (rtl_ocp_reg_failure(reg))
 		return 0;
@@ -865,12 +876,28 @@ static u16 r8168_mac_ocp_read(struct rtl8169_private *tp, u32 reg)
 	return RTL_R32(tp, OCPDR);
 }
 
+static u16 r8168_mac_ocp_read(struct rtl8169_private *tp, u32 reg)
+{
+	unsigned long flags;
+	u16 val;
+
+	spin_lock_irqsave(&tp->mac_ocp_lock, flags);
+	val = __r8168_mac_ocp_read(tp, reg);
+	spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
+
+	return val;
+}
+
 static void r8168_mac_ocp_modify(struct rtl8169_private *tp, u32 reg, u16 mask,
 				 u16 set)
 {
-	u16 data = r8168_mac_ocp_read(tp, reg);
+	unsigned long flags;
+	u16 data;
 
-	r8168_mac_ocp_write(tp, reg, (data & ~mask) | set);
+	spin_lock_irqsave(&tp->mac_ocp_lock, flags);
+	data = __r8168_mac_ocp_read(tp, reg);
+	__r8168_mac_ocp_write(tp, reg, (data & ~mask) | set);
+	spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
 }
 
 /* Work around a hw issue with RTL8168g PHY, the quirk disables
@@ -5176,6 +5203,8 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	tp->eee_adv = -1;
 	tp->ocp_base = OCP_STD_PHY_BASE;
 
+	spin_lock_init(&tp->mac_ocp_lock);
+
 	dev->tstats = devm_netdev_alloc_pcpu_stats(&pdev->dev,
 						   struct pcpu_sw_netstats);
 	if (!dev->tstats)
-- 
2.40.1.445.gf85cd430b1

From f0954f7b3ba9d7d90efb7d190a5301c1c8d203f5 Mon Sep 17 00:00:00 2001
From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Sat, 25 Feb 2023 22:44:50 +0100
Subject: [PATCH 12/24] r8169: use spinlock to protect access to registers
 Config2 and Config5

For disabling ASPM during NAPI poll we'll have to access both registers
in atomic context. Use a spinlock to protect access.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 47 ++++++++++++++++++-----
 1 file changed, 37 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 259eac5b0..e6f3f1947 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -613,6 +613,7 @@ struct rtl8169_private {
 		struct work_struct work;
 	} wk;
 
+	spinlock_t config25_lock;
 	spinlock_t mac_ocp_lock;
 
 	unsigned supports_gmii:1;
@@ -677,6 +678,28 @@ static void rtl_pci_commit(struct rtl8169_private *tp)
 	RTL_R8(tp, ChipCmd);
 }
 
+static void rtl_mod_config2(struct rtl8169_private *tp, u8 clear, u8 set)
+{
+	unsigned long flags;
+	u8 val;
+
+	spin_lock_irqsave(&tp->config25_lock, flags);
+	val = RTL_R8(tp, Config2);
+	RTL_W8(tp, Config2, (val & ~clear) | set);
+	spin_unlock_irqrestore(&tp->config25_lock, flags);
+}
+
+static void rtl_mod_config5(struct rtl8169_private *tp, u8 clear, u8 set)
+{
+	unsigned long flags;
+	u8 val;
+
+	spin_lock_irqsave(&tp->config25_lock, flags);
+	val = RTL_R8(tp, Config5);
+	RTL_W8(tp, Config5, (val & ~clear) | set);
+	spin_unlock_irqrestore(&tp->config25_lock, flags);
+}
+
 static bool rtl_is_8125(struct rtl8169_private *tp)
 {
 	return tp->mac_version >= RTL_GIGA_MAC_VER_61;
@@ -1363,6 +1386,7 @@ static void __rtl8169_set_wol(struct rtl8169_private *tp, u32 wolopts)
 		{ WAKE_MAGIC, Config3, MagicPacket }
 	};
 	unsigned int i, tmp = ARRAY_SIZE(cfg);
+	unsigned long flags;
 	u8 options;
 
 	rtl_unlock_config_regs(tp);
@@ -1381,12 +1405,14 @@ static void __rtl8169_set_wol(struct rtl8169_private *tp, u32 wolopts)
 			r8168_mac_ocp_modify(tp, 0xc0b6, BIT(0), 0);
 	}
 
+	spin_lock_irqsave(&tp->config25_lock, flags);
 	for (i = 0; i < tmp; i++) {
 		options = RTL_R8(tp, cfg[i].reg) & ~cfg[i].mask;
 		if (wolopts & cfg[i].opt)
 			options |= cfg[i].mask;
 		RTL_W8(tp, cfg[i].reg, options);
 	}
+	spin_unlock_irqrestore(&tp->config25_lock, flags);
 
 	switch (tp->mac_version) {
 	case RTL_GIGA_MAC_VER_02 ... RTL_GIGA_MAC_VER_06:
@@ -1398,10 +1424,10 @@ static void __rtl8169_set_wol(struct rtl8169_private *tp, u32 wolopts)
 	case RTL_GIGA_MAC_VER_34:
 	case RTL_GIGA_MAC_VER_37:
 	case RTL_GIGA_MAC_VER_39 ... RTL_GIGA_MAC_VER_63:
-		options = RTL_R8(tp, Config2) & ~PME_SIGNAL;
 		if (wolopts)
-			options |= PME_SIGNAL;
-		RTL_W8(tp, Config2, options);
+			rtl_mod_config2(tp, 0, PME_SIGNAL);
+		else
+			rtl_mod_config2(tp, PME_SIGNAL, 0);
 		break;
 	default:
 		break;
@@ -2704,8 +2730,8 @@ static void rtl_hw_aspm_clkreq_enable(struct rtl8169_private *tp, bool enable)
 {
 	/* Don't enable ASPM in the chip if OS can't control ASPM */
 	if (enable && tp->aspm_manageable) {
-		RTL_W8(tp, Config5, RTL_R8(tp, Config5) | ASPM_en);
-		RTL_W8(tp, Config2, RTL_R8(tp, Config2) | ClkReqEn);
+		rtl_mod_config5(tp, 0, ASPM_en);
+		rtl_mod_config2(tp, 0, ClkReqEn);
 
 		switch (tp->mac_version) {
 		case RTL_GIGA_MAC_VER_46 ... RTL_GIGA_MAC_VER_48:
@@ -2728,8 +2754,8 @@ static void rtl_hw_aspm_clkreq_enable(struct rtl8169_private *tp, bool enable)
 			break;
 		}
 
-		RTL_W8(tp, Config2, RTL_R8(tp, Config2) & ~ClkReqEn);
-		RTL_W8(tp, Config5, RTL_R8(tp, Config5) & ~ASPM_en);
+		rtl_mod_config2(tp, ClkReqEn, 0);
+		rtl_mod_config5(tp, ASPM_en, 0);
 	}
 
 	udelay(10);
@@ -2890,7 +2916,7 @@ static void rtl_hw_start_8168e_1(struct rtl8169_private *tp)
 	RTL_W32(tp, MISC, RTL_R32(tp, MISC) | TXPLA_RST);
 	RTL_W32(tp, MISC, RTL_R32(tp, MISC) & ~TXPLA_RST);
 
-	RTL_W8(tp, Config5, RTL_R8(tp, Config5) & ~Spi_en);
+	rtl_mod_config5(tp, Spi_en, 0);
 }
 
 static void rtl_hw_start_8168e_2(struct rtl8169_private *tp)
@@ -2923,7 +2949,7 @@ static void rtl_hw_start_8168e_2(struct rtl8169_private *tp)
 
 	RTL_W8(tp, DLLPR, RTL_R8(tp, DLLPR) | PFM_EN);
 	RTL_W32(tp, MISC, RTL_R32(tp, MISC) | PWM_EN);
-	RTL_W8(tp, Config5, RTL_R8(tp, Config5) & ~Spi_en);
+	rtl_mod_config5(tp, Spi_en, 0);
 
 	rtl_hw_aspm_clkreq_enable(tp, true);
 }
@@ -2946,7 +2972,7 @@ static void rtl_hw_start_8168f(struct rtl8169_private *tp)
 	RTL_W8(tp, MCU, RTL_R8(tp, MCU) & ~NOW_IS_OOB);
 	RTL_W8(tp, DLLPR, RTL_R8(tp, DLLPR) | PFM_EN);
 	RTL_W32(tp, MISC, RTL_R32(tp, MISC) | PWM_EN);
-	RTL_W8(tp, Config5, RTL_R8(tp, Config5) & ~Spi_en);
+	rtl_mod_config5(tp, Spi_en, 0);
 
 	rtl8168_config_eee_mac(tp);
 }
@@ -5203,6 +5229,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	tp->eee_adv = -1;
 	tp->ocp_base = OCP_STD_PHY_BASE;
 
+	spin_lock_init(&tp->config25_lock);
 	spin_lock_init(&tp->mac_ocp_lock);
 
 	dev->tstats = devm_netdev_alloc_pcpu_stats(&pdev->dev,
-- 
2.40.1.445.gf85cd430b1

From 06781e39bdb72cfaba7ce242518be7da96a3269b Mon Sep 17 00:00:00 2001
From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Sat, 25 Feb 2023 22:45:28 +0100
Subject: [PATCH 13/24] r8169: enable cfg9346 config register access in atomic
 context

For disabling ASPM during NAPI poll we'll have to unlock access
to the config registers in atomic context. Other code parts
running with config register access unlocked are partially
longer and can sleep. Add a usage counter to enable parallel
execution of code parts requiring unlocked config registers.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index e6f3f1947..61cbf498f 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -616,6 +616,9 @@ struct rtl8169_private {
 	spinlock_t config25_lock;
 	spinlock_t mac_ocp_lock;
 
+	spinlock_t cfg9346_usage_lock;
+	int cfg9346_usage_count;
+
 	unsigned supports_gmii:1;
 	unsigned aspm_manageable:1;
 	dma_addr_t counters_phys_addr;
@@ -664,12 +667,22 @@ static inline struct device *tp_to_dev(struct rtl8169_private *tp)
 
 static void rtl_lock_config_regs(struct rtl8169_private *tp)
 {
-	RTL_W8(tp, Cfg9346, Cfg9346_Lock);
+	unsigned long flags;
+
+	spin_lock_irqsave(&tp->cfg9346_usage_lock, flags);
+	if (!--tp->cfg9346_usage_count)
+		RTL_W8(tp, Cfg9346, Cfg9346_Lock);
+	spin_unlock_irqrestore(&tp->cfg9346_usage_lock, flags);
 }
 
 static void rtl_unlock_config_regs(struct rtl8169_private *tp)
 {
-	RTL_W8(tp, Cfg9346, Cfg9346_Unlock);
+	unsigned long flags;
+
+	spin_lock_irqsave(&tp->cfg9346_usage_lock, flags);
+	if (!tp->cfg9346_usage_count++)
+		RTL_W8(tp, Cfg9346, Cfg9346_Unlock);
+	spin_unlock_irqrestore(&tp->cfg9346_usage_lock, flags);
 }
 
 static void rtl_pci_commit(struct rtl8169_private *tp)
@@ -5229,6 +5242,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	tp->eee_adv = -1;
 	tp->ocp_base = OCP_STD_PHY_BASE;
 
+	spin_lock_init(&tp->cfg9346_usage_lock);
 	spin_lock_init(&tp->config25_lock);
 	spin_lock_init(&tp->mac_ocp_lock);
 
-- 
2.40.1.445.gf85cd430b1

From f55b397cb1dec5c6850253d01969a09c5737af00 Mon Sep 17 00:00:00 2001
From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Sat, 25 Feb 2023 22:46:05 +0100
Subject: [PATCH 14/24] r8169: prepare rtl_hw_aspm_clkreq_enable for usage in
 atomic context

Bail out if the function is used with chip versions that don't support
ASPM configuration. In addition remove the delay, it tuned out that
it's not needed, also vendor driver r8125 doesn't have it.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 61cbf498f..96af31aea 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -2741,6 +2741,9 @@ static void rtl_disable_exit_l1(struct rtl8169_private *tp)
 
 static void rtl_hw_aspm_clkreq_enable(struct rtl8169_private *tp, bool enable)
 {
+	if (tp->mac_version < RTL_GIGA_MAC_VER_32)
+		return;
+
 	/* Don't enable ASPM in the chip if OS can't control ASPM */
 	if (enable && tp->aspm_manageable) {
 		rtl_mod_config5(tp, 0, ASPM_en);
@@ -2770,8 +2773,6 @@ static void rtl_hw_aspm_clkreq_enable(struct rtl8169_private *tp, bool enable)
 		rtl_mod_config2(tp, ClkReqEn, 0);
 		rtl_mod_config5(tp, ASPM_en, 0);
 	}
-
-	udelay(10);
 }
 
 static void rtl_set_fifo_size(struct rtl8169_private *tp, u16 rx_stat,
-- 
2.40.1.445.gf85cd430b1

From da50dd39edf8f6ffc4167e33a4b4be3fbfd86de1 Mon Sep 17 00:00:00 2001
From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Sat, 25 Feb 2023 22:46:46 +0100
Subject: [PATCH 15/24] r8169: disable ASPM during NAPI poll

Several chip versions have problems with ASPM, what may result in
rx_missed errors or tx timeouts. The root cause isn't known but
experience shows that disabling ASPM during NAPI poll can avoid
these problems.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 96af31aea..2897b9bf2 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -4577,6 +4577,10 @@ static irqreturn_t rtl8169_interrupt(int irq, void *dev_instance)
 	}
 
 	if (napi_schedule_prep(&tp->napi)) {
+		rtl_unlock_config_regs(tp);
+		rtl_hw_aspm_clkreq_enable(tp, false);
+		rtl_lock_config_regs(tp);
+
 		rtl_irq_disable(tp);
 		__napi_schedule(&tp->napi);
 	}
@@ -4636,9 +4640,14 @@ static int rtl8169_poll(struct napi_struct *napi, int budget)
 
 	work_done = rtl_rx(dev, tp, budget);
 
-	if (work_done < budget && napi_complete_done(napi, work_done))
+	if (work_done < budget && napi_complete_done(napi, work_done)) {
 		rtl_irq_enable(tp);
 
+		rtl_unlock_config_regs(tp);
+		rtl_hw_aspm_clkreq_enable(tp, true);
+		rtl_lock_config_regs(tp);
+	}
+
 	return work_done;
 }
 
-- 
2.40.1.445.gf85cd430b1

From fabf76b71df1e95e3e678fbb18e767935c4576f3 Mon Sep 17 00:00:00 2001
From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Sat, 25 Feb 2023 22:47:32 +0100
Subject: [PATCH 16/24] r8169: remove ASPM restrictions now that ASPM is
 disabled during NAPI poll

Now that  ASPM is disabled during NAPI poll, we can remove all ASPM
restrictions. This allows for higher power savings if the network
isn't fully loaded.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 27 +----------------------
 1 file changed, 1 insertion(+), 26 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 2897b9bf2..6563e4c6a 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -620,7 +620,6 @@ struct rtl8169_private {
 	int cfg9346_usage_count;
 
 	unsigned supports_gmii:1;
-	unsigned aspm_manageable:1;
 	dma_addr_t counters_phys_addr;
 	struct rtl8169_counters *counters;
 	struct rtl8169_tc_offsets tc_offset;
@@ -2744,8 +2743,7 @@ static void rtl_hw_aspm_clkreq_enable(struct rtl8169_private *tp, bool enable)
 	if (tp->mac_version < RTL_GIGA_MAC_VER_32)
 		return;
 
-	/* Don't enable ASPM in the chip if OS can't control ASPM */
-	if (enable && tp->aspm_manageable) {
+	if (enable) {
 		rtl_mod_config5(tp, 0, ASPM_en);
 		rtl_mod_config2(tp, 0, ClkReqEn);
 
@@ -5221,16 +5219,6 @@ static void rtl_init_mac_address(struct rtl8169_private *tp)
 	rtl_rar_set(tp, mac_addr);
 }
 
-/* register is set if system vendor successfully tested ASPM 1.2 */
-static bool rtl_aspm_is_safe(struct rtl8169_private *tp)
-{
-	if (tp->mac_version >= RTL_GIGA_MAC_VER_61 &&
-	    r8168_mac_ocp_read(tp, 0xc0b2) & 0xf)
-		return true;
-
-	return false;
-}
-
 static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 {
 	struct rtl8169_private *tp;
@@ -5302,19 +5290,6 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	tp->mac_version = chipset;
 
-	/* Disable ASPM L1 as that cause random device stop working
-	 * problems as well as full system hangs for some PCIe devices users.
-	 * Chips from RTL8168h partially have issues with L1.2, but seem
-	 * to work fine with L1 and L1.1.
-	 */
-	if (rtl_aspm_is_safe(tp))
-		rc = 0;
-	else if (tp->mac_version >= RTL_GIGA_MAC_VER_46)
-		rc = pci_disable_link_state(pdev, PCIE_LINK_STATE_L1_2);
-	else
-		rc = pci_disable_link_state(pdev, PCIE_LINK_STATE_L1);
-	tp->aspm_manageable = !rc;
-
 	tp->dash_type = rtl_check_dash(tp);
 
 	tp->cp_cmd = RTL_R16(tp, CPlusCmd) & CPCMD_MASK;
-- 
2.40.1.445.gf85cd430b1

From 3de1275eff91135f01cace2da1bb51d182c803ba Mon Sep 17 00:00:00 2001
From: Paolo Abeni <pabeni@redhat.com>
Date: Tue, 7 Mar 2023 19:46:37 +0100
Subject: [PATCH 17/24] epoll: use refcount to reduce ep_mutex contention

We are observing huge contention on the epmutex during an http
connection/rate test:

83.17% 0.25%  nginx            [kernel.kallsyms]         [k] entry_SYSCALL_64_after_hwframe
[...]
           |--66.96%--__fput
                      |--60.04%--eventpoll_release_file
                                 |--58.41%--__mutex_lock.isra.6
                                           |--56.56%--osq_lock

The application is multi-threaded, creates a new epoll entry for each
incoming connection, and does not delete it before the connection shutdown
- that is, before the connection's fd close().

Many different threads compete frequently for the epmutex lock, affecting
the overall performance.

To reduce the contention this patch introduces explicit reference counting
for the eventpoll struct.  Each registered event acquires a reference, and
references are released at ep_remove() time.

The eventpoll struct is released by whoever - among EP file close() and
and the monitored file close() drops its last reference.

Additionally, this introduces a new 'dying' flag to prevent races between
the EP file close() and the monitored file close().
ep_eventpoll_release() marks, under f_lock spinlock, each epitem as dying
before removing it, while EP file close() does not touch dying epitems.

The above is needed as both close operations could run concurrently and
drop the EP reference acquired via the epitem entry.  Without the above
flag, the monitored file close() could reach the EP struct via the epitem
list while the epitem is still listed and then try to put it after its
disposal.

An alternative could be avoiding touching the references acquired via the
epitems at EP file close() time, but that could leave the EP struct alive
for potentially unlimited time after EP file close(), with nasty side
effects.

With all the above in place, we can drop the epmutex usage at disposal
time.

Overall this produces a significant performance improvement in the
mentioned connection/rate scenario: the mutex operations disappear from
the topmost offenders in the perf report, and the measured
connections/rate grows by ~60%.

To make the change more readable this additionally renames ep_free() to
ep_clear_and_put(), and moves the actual memory cleanup in a separate
ep_free() helper.

Link: https://lkml.kernel.org/r/e8228f0048977456466bc33b42600e929fedd319.1678213651.git.pabeni@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tested-by: Xiumei Mu <xmu@redhat.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Carlos Maiolino <cmaiolino@redhat.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Eric Biggers <ebiggers@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
 fs/eventpoll.c | 185 +++++++++++++++++++++++++++++++------------------
 1 file changed, 116 insertions(+), 69 deletions(-)

diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index 8b5ca9f8f..5e5e26311 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -57,13 +57,7 @@
  * we need a lock that will allow us to sleep. This lock is a
  * mutex (ep->mtx). It is acquired during the event transfer loop,
  * during epoll_ctl(EPOLL_CTL_DEL) and during eventpoll_release_file().
- * Then we also need a global mutex to serialize eventpoll_release_file()
- * and ep_free().
- * This mutex is acquired by ep_free() during the epoll file
- * cleanup path and it is also acquired by eventpoll_release_file()
- * if a file has been pushed inside an epoll set and it is then
- * close()d without a previous call to epoll_ctl(EPOLL_CTL_DEL).
- * It is also acquired when inserting an epoll fd onto another epoll
+ * The epmutex is acquired when inserting an epoll fd onto another epoll
  * fd. We do this so that we walk the epoll tree and ensure that this
  * insertion does not create a cycle of epoll file descriptors, which
  * could lead to deadlock. We need a global mutex to prevent two
@@ -153,6 +147,13 @@ struct epitem {
 	/* The file descriptor information this item refers to */
 	struct epoll_filefd ffd;
 
+	/*
+	 * Protected by file->f_lock, true for to-be-released epitem already
+	 * removed from the "struct file" items list; together with
+	 * eventpoll->refcount orchestrates "struct eventpoll" disposal
+	 */
+	bool dying;
+
 	/* List containing poll wait queues */
 	struct eppoll_entry *pwqlist;
 
@@ -217,6 +218,12 @@ struct eventpoll {
 	u64 gen;
 	struct hlist_head refs;
 
+	/*
+	 * usage count, used together with epitem->dying to
+	 * orchestrate the disposal of this struct
+	 */
+	refcount_t refcount;
+
 #ifdef CONFIG_NET_RX_BUSY_POLL
 	/* used to track busy poll napi_id */
 	unsigned int napi_id;
@@ -240,9 +247,7 @@ struct ep_pqueue {
 /* Maximum number of epoll watched descriptors, per user */
 static long max_user_watches __read_mostly;
 
-/*
- * This mutex is used to serialize ep_free() and eventpoll_release_file().
- */
+/* Used for cycles detection */
 static DEFINE_MUTEX(epmutex);
 
 static u64 loop_check_gen = 0;
@@ -557,8 +562,7 @@ static void ep_remove_wait_queue(struct eppoll_entry *pwq)
 
 /*
  * This function unregisters poll callbacks from the associated file
- * descriptor.  Must be called with "mtx" held (or "epmutex" if called from
- * ep_free).
+ * descriptor.  Must be called with "mtx" held.
  */
 static void ep_unregister_pollwait(struct eventpoll *ep, struct epitem *epi)
 {
@@ -681,11 +685,38 @@ static void epi_rcu_free(struct rcu_head *head)
 	kmem_cache_free(epi_cache, epi);
 }
 
+static void ep_get(struct eventpoll *ep)
+{
+	refcount_inc(&ep->refcount);
+}
+
+/*
+ * Returns true if the event poll can be disposed
+ */
+static bool ep_refcount_dec_and_test(struct eventpoll *ep)
+{
+	if (!refcount_dec_and_test(&ep->refcount))
+		return false;
+
+	WARN_ON_ONCE(!RB_EMPTY_ROOT(&ep->rbr.rb_root));
+	return true;
+}
+
+static void ep_free(struct eventpoll *ep)
+{
+	mutex_destroy(&ep->mtx);
+	free_uid(ep->user);
+	wakeup_source_unregister(ep->ws);
+	kfree(ep);
+}
+
 /*
  * Removes a "struct epitem" from the eventpoll RB tree and deallocates
  * all the associated resources. Must be called with "mtx" held.
+ * If the dying flag is set, do the removal only if force is true.
+ * Returns true if the eventpoll can be disposed.
  */
-static int ep_remove(struct eventpoll *ep, struct epitem *epi)
+static bool __ep_remove(struct eventpoll *ep, struct epitem *epi, bool force)
 {
 	struct file *file = epi->ffd.file;
 	struct epitems_head *to_free;
@@ -700,6 +731,11 @@ static int ep_remove(struct eventpoll *ep, struct epitem *epi)
 
 	/* Remove the current item from the list of epoll hooks */
 	spin_lock(&file->f_lock);
+	if (epi->dying && !force) {
+		spin_unlock(&file->f_lock);
+		return false;
+	}
+
 	to_free = NULL;
 	head = file->f_ep;
 	if (head->first == &epi->fllink && !epi->fllink.next) {
@@ -733,28 +769,28 @@ static int ep_remove(struct eventpoll *ep, struct epitem *epi)
 	call_rcu(&epi->rcu, epi_rcu_free);
 
 	percpu_counter_dec(&ep->user->epoll_watches);
+	return ep_refcount_dec_and_test(ep);
+}
 
-	return 0;
+/*
+ * ep_remove variant for callers owing an additional reference to the ep
+ */
+static void ep_remove_safe(struct eventpoll *ep, struct epitem *epi)
+{
+	WARN_ON_ONCE(__ep_remove(ep, epi, false));
 }
 
-static void ep_free(struct eventpoll *ep)
+static void ep_clear_and_put(struct eventpoll *ep)
 {
 	struct rb_node *rbp;
 	struct epitem *epi;
+	bool dispose;
 
 	/* We need to release all tasks waiting for these file */
 	if (waitqueue_active(&ep->poll_wait))
 		ep_poll_safewake(ep, NULL, 0);
 
-	/*
-	 * We need to lock this because we could be hit by
-	 * eventpoll_release_file() while we're freeing the "struct eventpoll".
-	 * We do not need to hold "ep->mtx" here because the epoll file
-	 * is on the way to be removed and no one has references to it
-	 * anymore. The only hit might come from eventpoll_release_file() but
-	 * holding "epmutex" is sufficient here.
-	 */
-	mutex_lock(&epmutex);
+	mutex_lock(&ep->mtx);
 
 	/*
 	 * Walks through the whole tree by unregistering poll callbacks.
@@ -768,25 +804,21 @@ static void ep_free(struct eventpoll *ep)
 
 	/*
 	 * Walks through the whole tree by freeing each "struct epitem". At this
-	 * point we are sure no poll callbacks will be lingering around, and also by
-	 * holding "epmutex" we can be sure that no file cleanup code will hit
-	 * us during this operation. So we can avoid the lock on "ep->lock".
-	 * We do not need to lock ep->mtx, either, we only do it to prevent
-	 * a lockdep warning.
+	 * point we are sure no poll callbacks will be lingering around.
+	 * Since we still own a reference to the eventpoll struct, the loop can't
+	 * dispose it.
 	 */
-	mutex_lock(&ep->mtx);
 	while ((rbp = rb_first_cached(&ep->rbr)) != NULL) {
 		epi = rb_entry(rbp, struct epitem, rbn);
-		ep_remove(ep, epi);
+		ep_remove_safe(ep, epi);
 		cond_resched();
 	}
+
+	dispose = ep_refcount_dec_and_test(ep);
 	mutex_unlock(&ep->mtx);
 
-	mutex_unlock(&epmutex);
-	mutex_destroy(&ep->mtx);
-	free_uid(ep->user);
-	wakeup_source_unregister(ep->ws);
-	kfree(ep);
+	if (dispose)
+		ep_free(ep);
 }
 
 static int ep_eventpoll_release(struct inode *inode, struct file *file)
@@ -794,7 +826,7 @@ static int ep_eventpoll_release(struct inode *inode, struct file *file)
 	struct eventpoll *ep = file->private_data;
 
 	if (ep)
-		ep_free(ep);
+		ep_clear_and_put(ep);
 
 	return 0;
 }
@@ -906,33 +938,35 @@ void eventpoll_release_file(struct file *file)
 {
 	struct eventpoll *ep;
 	struct epitem *epi;
-	struct hlist_node *next;
+	bool dispose;
 
 	/*
-	 * We don't want to get "file->f_lock" because it is not
-	 * necessary. It is not necessary because we're in the "struct file"
-	 * cleanup path, and this means that no one is using this file anymore.
-	 * So, for example, epoll_ctl() cannot hit here since if we reach this
-	 * point, the file counter already went to zero and fget() would fail.
-	 * The only hit might come from ep_free() but by holding the mutex
-	 * will correctly serialize the operation. We do need to acquire
-	 * "ep->mtx" after "epmutex" because ep_remove() requires it when called
-	 * from anywhere but ep_free().
-	 *
-	 * Besides, ep_remove() acquires the lock, so we can't hold it here.
+	 * Use the 'dying' flag to prevent a concurrent ep_cleat_and_put() from
+	 * touching the epitems list before eventpoll_release_file() can access
+	 * the ep->mtx.
 	 */
-	mutex_lock(&epmutex);
-	if (unlikely(!file->f_ep)) {
-		mutex_unlock(&epmutex);
-		return;
-	}
-	hlist_for_each_entry_safe(epi, next, file->f_ep, fllink) {
+again:
+	spin_lock(&file->f_lock);
+	if (file->f_ep && file->f_ep->first) {
+		/* detach from ep tree */
+		epi = hlist_entry(file->f_ep->first, struct epitem, fllink);
+		epi->dying = true;
+		spin_unlock(&file->f_lock);
+
+		/*
+		 * ep access is safe as we still own a reference to the ep
+		 * struct
+		 */
 		ep = epi->ep;
-		mutex_lock_nested(&ep->mtx, 0);
-		ep_remove(ep, epi);
+		mutex_lock(&ep->mtx);
+		dispose = __ep_remove(ep, epi, true);
 		mutex_unlock(&ep->mtx);
+
+		if (dispose)
+			ep_free(ep);
+		goto again;
 	}
-	mutex_unlock(&epmutex);
+	spin_unlock(&file->f_lock);
 }
 
 static int ep_alloc(struct eventpoll **pep)
@@ -955,6 +989,7 @@ static int ep_alloc(struct eventpoll **pep)
 	ep->rbr = RB_ROOT_CACHED;
 	ep->ovflist = EP_UNACTIVE_PTR;
 	ep->user = user;
+	refcount_set(&ep->refcount, 1);
 
 	*pep = ep;
 
@@ -1223,10 +1258,10 @@ static int ep_poll_callback(wait_queue_entry_t *wait, unsigned mode, int sync, v
 		 */
 		list_del_init(&wait->entry);
 		/*
-		 * ->whead != NULL protects us from the race with ep_free()
-		 * or ep_remove(), ep_remove_wait_queue() takes whead->lock
-		 * held by the caller. Once we nullify it, nothing protects
-		 * ep/epi or even wait.
+		 * ->whead != NULL protects us from the race with
+		 * ep_clear_and_put() or ep_remove(), ep_remove_wait_queue()
+		 * takes whead->lock held by the caller. Once we nullify it,
+		 * nothing protects ep/epi or even wait.
 		 */
 		smp_store_release(&ep_pwq_from_wait(wait)->whead, NULL);
 	}
@@ -1496,16 +1531,22 @@ static int ep_insert(struct eventpoll *ep, const struct epoll_event *event,
 	if (tep)
 		mutex_unlock(&tep->mtx);
 
+	/*
+	 * ep_remove_safe() calls in the later error paths can't lead to
+	 * ep_free() as the ep file itself still holds an ep reference.
+	 */
+	ep_get(ep);
+
 	/* now check if we've created too many backpaths */
 	if (unlikely(full_check && reverse_path_check())) {
-		ep_remove(ep, epi);
+		ep_remove_safe(ep, epi);
 		return -EINVAL;
 	}
 
 	if (epi->event.events & EPOLLWAKEUP) {
 		error = ep_create_wakeup_source(epi);
 		if (error) {
-			ep_remove(ep, epi);
+			ep_remove_safe(ep, epi);
 			return error;
 		}
 	}
@@ -1529,7 +1570,7 @@ static int ep_insert(struct eventpoll *ep, const struct epoll_event *event,
 	 * high memory pressure.
 	 */
 	if (unlikely(!epq.epi)) {
-		ep_remove(ep, epi);
+		ep_remove_safe(ep, epi);
 		return -ENOMEM;
 	}
 
@@ -2025,7 +2066,7 @@ static int do_epoll_create(int flags)
 out_free_fd:
 	put_unused_fd(fd);
 out_free_ep:
-	ep_free(ep);
+	ep_clear_and_put(ep);
 	return error;
 }
 
@@ -2167,10 +2208,16 @@ int do_epoll_ctl(int epfd, int op, int fd, struct epoll_event *epds,
 			error = -EEXIST;
 		break;
 	case EPOLL_CTL_DEL:
-		if (epi)
-			error = ep_remove(ep, epi);
-		else
+		if (epi) {
+			/*
+			 * The eventpoll itself is still alive: the refcount
+			 * can't go to zero here.
+			 */
+			ep_remove_safe(ep, epi);
+			error = 0;
+		} else {
 			error = -ENOENT;
+		}
 		break;
 	case EPOLL_CTL_MOD:
 		if (epi) {
-- 
2.40.1.445.gf85cd430b1

From cbe5fc21bb7c0ae48c25120bf9e3ff976e3d0863 Mon Sep 17 00:00:00 2001
From: Paolo Abeni <pabeni@redhat.com>
Date: Wed, 8 Mar 2023 22:51:31 +0100
Subject: [PATCH 18/24] epoll-use-refcount-to-reduce-ep_mutex-contention-v5

update some comments

Link: https://lkml.kernel.org/r/323de732635cc3513c1837c6cbb98f012174f994.1678312201.git.pabeni@redhat.com
Tested-by: Xiumei Mu <xmu@redhiat.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Carlos Maiolino <cmaiolino@redhat.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Eric Biggers <ebiggers@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
 fs/eventpoll.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index 5e5e26311..4cad49002 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -714,6 +714,8 @@ static void ep_free(struct eventpoll *ep)
  * Removes a "struct epitem" from the eventpoll RB tree and deallocates
  * all the associated resources. Must be called with "mtx" held.
  * If the dying flag is set, do the removal only if force is true.
+ * This prevents ep_clear_and_put() from dropping all the ep references
+ * while running concurrently with eventpoll_release_file().
  * Returns true if the eventpoll can be disposed.
  */
 static bool __ep_remove(struct eventpoll *ep, struct epitem *epi, bool force)
@@ -941,14 +943,13 @@ void eventpoll_release_file(struct file *file)
 	bool dispose;
 
 	/*
-	 * Use the 'dying' flag to prevent a concurrent ep_cleat_and_put() from
+	 * Use the 'dying' flag to prevent a concurrent ep_clear_and_put() from
 	 * touching the epitems list before eventpoll_release_file() can access
 	 * the ep->mtx.
 	 */
 again:
 	spin_lock(&file->f_lock);
 	if (file->f_ep && file->f_ep->first) {
-		/* detach from ep tree */
 		epi = hlist_entry(file->f_ep->first, struct epitem, fllink);
 		epi->dying = true;
 		spin_unlock(&file->f_lock);
-- 
2.40.1.445.gf85cd430b1

From b89b91dc2ff1001da97acbb01341bacbde721b95 Mon Sep 17 00:00:00 2001
From: Alexander Halbuer <halbuer@sra.uni-hannover.de>
Date: Wed, 1 Feb 2023 17:25:49 +0100
Subject: [PATCH 19/24] mm: reduce lock contention of pcp buffer refill

rmqueue_bulk() batches the allocation of multiple elements to refill the
per-CPU buffers into a single hold of the zone lock.  Each element is
allocated and checked using check_pcp_refill().  The check touches every
related struct page which is especially expensive for higher order
allocations (huge pages).

This patch reduces the time holding the lock by moving the check out of
the critical section similar to rmqueue_buddy() which allocates a single
element.

Measurements of parallel allocation-heavy workloads show a reduction of
the average huge page allocation latency of 50 percent for two cores and
nearly 90 percent for 24 cores.

Link: https://lkml.kernel.org/r/20230201162549.68384-1-halbuer@sra.uni-hannover.de
Signed-off-by: Alexander Halbuer <halbuer@sra.uni-hannover.de>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
 mm/page_alloc.c | 22 ++++++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index afcfb2a94..c92b8df24 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3138,6 +3138,8 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 {
 	unsigned long flags;
 	int i, allocated = 0;
+	struct list_head *prev_tail = list->prev;
+	struct page *pos, *n;
 
 	spin_lock_irqsave(&zone->lock, flags);
 	for (i = 0; i < count; ++i) {
@@ -3146,9 +3148,6 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 		if (unlikely(page == NULL))
 			break;
 
-		if (unlikely(check_pcp_refill(page, order)))
-			continue;
-
 		/*
 		 * Split buddy pages returned by expand() are received here in
 		 * physical page order. The page is added to the tail of
@@ -3160,7 +3159,6 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 		 * pages are ordered properly.
 		 */
 		list_add_tail(&page->pcp_list, list);
-		allocated++;
 		if (is_migrate_cma(get_pcppage_migratetype(page)))
 			__mod_zone_page_state(zone, NR_FREE_CMA_PAGES,
 					      -(1 << order));
@@ -3174,6 +3172,22 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 	 */
 	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
 	spin_unlock_irqrestore(&zone->lock, flags);
+
+	/*
+	 * Pages are appended to the pcp list without checking to reduce the
+	 * time holding the zone lock. Checking the appended pages happens right
+	 * after the critical section while still holding the pcp lock.
+	 */
+	pos = list_first_entry(prev_tail, struct page, pcp_list);
+	list_for_each_entry_safe_from(pos, n, list, pcp_list) {
+		if (unlikely(check_pcp_refill(pos, order))) {
+			list_del(&pos->pcp_list);
+			continue;
+		}
+
+		allocated++;
+	}
+
 	return allocated;
 }
 
-- 
2.40.1.445.gf85cd430b1

From 68aed5c944fc39e614c6de2299c0a6b56d52580a Mon Sep 17 00:00:00 2001
From: Uros Bizjak <ubizjak@gmail.com>
Date: Mon, 20 Mar 2023 22:20:12 +0100
Subject: [PATCH 20/24] x86/ACPI/boot: Improve __acpi_acquire_global_lock

Improve __acpi_acquire_global_lock by using a temporary variable.
This enables compiler to perform if-conversion and improves generated
code from:

...
 72a:	d1 ea                	shr    %edx
 72c:	83 e1 fc             	and    $0xfffffffc,%ecx
 72f:	83 e2 01             	and    $0x1,%edx
 732:	09 ca                	or     %ecx,%edx
 734:	83 c2 02             	add    $0x2,%edx
 737:	f0 0f b1 17          	lock cmpxchg %edx,(%rdi)
 73b:	75 e9                	jne    726 <__acpi_acquire_global_lock+0x6>
 73d:	83 e2 03             	and    $0x3,%edx
 740:	31 c0                	xor    %eax,%eax
 742:	83 fa 03             	cmp    $0x3,%edx
 745:	0f 95 c0             	setne  %al
 748:	f7 d8                	neg    %eax

to:

...
 72a:	d1 e9                	shr    %ecx
 72c:	83 e2 fc             	and    $0xfffffffc,%edx
 72f:	83 e1 01             	and    $0x1,%ecx
 732:	09 ca                	or     %ecx,%edx
 734:	83 c2 02             	add    $0x2,%edx
 737:	f0 0f b1 17          	lock cmpxchg %edx,(%rdi)
 73b:	75 e9                	jne    726 <__acpi_acquire_global_lock+0x6>
 73d:	8d 41 ff             	lea    -0x1(%rcx),%eax

BTW: the compiler could generate:

lea 0x2(%rcx,%rdx,1),%edx

instead of:

or     %ecx,%edx
	add    $0x2,%edx

but unwated conversion from add to or when bits are known to be zero
prevents this improvement. This is GCC PR108477.

No functional change intended.

Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/all/20230320212012.12704-1-ubizjak%40gmail.com
---
 arch/x86/kernel/acpi/boot.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index 0dac4ab5b..21b542a68 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -1858,13 +1858,18 @@ early_param("acpi_sci", setup_acpi_sci);
 
 int __acpi_acquire_global_lock(unsigned int *lock)
 {
-	unsigned int old, new;
+	unsigned int old, new, val;
 
 	old = READ_ONCE(*lock);
 	do {
-		new = (((old & ~0x3) + 2) + ((old >> 1) & 0x1));
+		val = (old >> 1) & 0x1;
+		new = (old & ~0x3) + 2 + val;
 	} while (!try_cmpxchg(lock, &old, new));
-	return ((new & 0x3) < 3) ? -1 : 0;
+
+	if (val)
+		return 0;
+
+	return -1;
 }
 
 int __acpi_release_global_lock(unsigned int *lock)
-- 
2.40.1.445.gf85cd430b1

From 69c4018ca131abbf66bd3cf9628c34b16b5fd355 Mon Sep 17 00:00:00 2001
From: Joan Bruguera <joanbrugueram@gmail.com>
Date: Sat, 7 Jan 2023 05:14:56 +0000
Subject: [PATCH 21/24] bpf, x86: Fix IP after emitting call depth accounting

Adjust the IP passed to `emit_patch` so it calculates the correct offset
for the CALL instruction if `x86_call_depth_emit_accounting` emits code.
Otherwise we will skip some instructions and most likely crash.

Fixes: b2e9dfe54be4 ("x86/bpf: Emit call depth accounting if required")
Link: https://lore.kernel.org/lkml/20230105214922.250473-1-joanbrugueram@gmail.com/
Signed-off-by: Joan Bruguera <joanbrugueram@gmail.com>
For: https://bugs.archlinux.org/task/77601
---
 arch/x86/net/bpf_jit_comp.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
index 1056bbf55..212bfd151 100644
--- a/arch/x86/net/bpf_jit_comp.c
+++ b/arch/x86/net/bpf_jit_comp.c
@@ -343,9 +343,10 @@ static int emit_call(u8 **pprog, void *func, void *ip)
 
 static int emit_rsb_call(u8 **pprog, void *func, void *ip)
 {
+	void *adjusted_ip;
 	OPTIMIZER_HIDE_VAR(func);
-	x86_call_depth_emit_accounting(pprog, func);
-	return emit_patch(pprog, func, ip, 0xE8);
+	adjusted_ip = (u8 *)ip + x86_call_depth_emit_accounting(pprog, func);
+	return emit_patch(pprog, func, adjusted_ip, 0xE8);
 }
 
 static int emit_jump(u8 **pprog, void *func, void *ip)
-- 
2.40.1.445.gf85cd430b1

From e47327149c5e8397efb650ad72ade0ce1c1b9633 Mon Sep 17 00:00:00 2001
From: Arnd Bergmann <arnd@arndb.de>
Date: Wed, 17 May 2023 15:19:31 +0200
Subject: [PATCH 22/24] decompressor: provide missing prototypes

The entry points for the decompressor don't always have a prototype
included in the .c file:

lib/decompress_inflate.c:42:17: error: no previous prototype for '__gunzip' [-Werror=missing-prototypes]
lib/decompress_unxz.c:251:17: error: no previous prototype for 'unxz' [-Werror=missing-prototypes]
lib/decompress_unzstd.c:331:17: error: no previous prototype for 'unzstd' [-Werror=missing-prototypes]

Include the correct headers for unxz and unzstd, and mark the inflate
function above as unconditionally 'static' to avoid these warnings.

Link: https://lkml.kernel.org/r/20230517131936.936840-1-arnd@kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Nick Terrell <terrelln@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
 lib/decompress_inflate.c | 2 +-
 lib/decompress_unxz.c    | 2 ++
 lib/decompress_unzstd.c  | 2 ++
 3 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/lib/decompress_inflate.c b/lib/decompress_inflate.c
index 6130c42b8..e19199f4a 100644
--- a/lib/decompress_inflate.c
+++ b/lib/decompress_inflate.c
@@ -39,7 +39,7 @@ static long INIT nofill(void *buffer, unsigned long len)
 }
 
 /* Included from initramfs et al code */
-STATIC int INIT __gunzip(unsigned char *buf, long len,
+static int INIT __gunzip(unsigned char *buf, long len,
 		       long (*fill)(void*, unsigned long),
 		       long (*flush)(void*, unsigned long),
 		       unsigned char *out_buf, long out_len,
diff --git a/lib/decompress_unxz.c b/lib/decompress_unxz.c
index 9f4262ee3..353268b9f 100644
--- a/lib/decompress_unxz.c
+++ b/lib/decompress_unxz.c
@@ -102,6 +102,8 @@
  */
 #ifdef STATIC
 #	define XZ_PREBOOT
+#else
+#include <linux/decompress/unxz.h>
 #endif
 #ifdef __KERNEL__
 #	include <linux/decompress/mm.h>
diff --git a/lib/decompress_unzstd.c b/lib/decompress_unzstd.c
index a512b99ae..bba2c0bb1 100644
--- a/lib/decompress_unzstd.c
+++ b/lib/decompress_unzstd.c
@@ -69,6 +69,8 @@
 # define UNZSTD_PREBOOT
 # include "xxhash.c"
 # include "zstd/decompress_sources.h"
+#else
+#include <linux/decompress/unzstd.h>
 #endif
 
 #include <linux/decompress/mm.h>
-- 
2.40.1.445.gf85cd430b1

From 4fd224b231ea5430c57fbc359dd371e107b7492e Mon Sep 17 00:00:00 2001
From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Date: Mon, 22 May 2023 15:41:21 +0200
Subject: [PATCH 23/24] r8169: Use a raw_spinlock_t for the register locks.

The driver's interrupt service routine is requested with the
IRQF_NO_THREAD if MSI is available. This means that the routine is
invoked in hardirq context even on PREEMPT_RT. The routine itself is
relatively short and schedules a worker, performs register access and
schedules NAPI. On PREEMPT_RT, scheduling NAPI from hardirq results in
waking ksoftirqd for further processing so using NAPI threads with this
driver is highly recommended since it NULL routes the threaded-IRQ
efforts.

Adding rtl_hw_aspm_clkreq_enable() to the ISR is problematic on
PREEMPT_RT because the function uses spinlock_t locks which become
sleeping locks on PREEMPT_RT. The locks are only used to protect
register access and don't nest into other functions or locks. They are
also not used for unbounded period of time. Therefore it looks okay to
convert them to raw_spinlock_t.

Convert the three locks which are used from the interrupt service
routine to raw_spinlock_t.

Fixes: e1ed3e4d91112 ("r8169: disable ASPM during NAPI poll")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 drivers/net/ethernet/realtek/r8169_main.c | 44 +++++++++++------------
 1 file changed, 22 insertions(+), 22 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 6563e4c6a..7b8f381f1 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -613,10 +613,10 @@ struct rtl8169_private {
 		struct work_struct work;
 	} wk;
 
-	spinlock_t config25_lock;
-	spinlock_t mac_ocp_lock;
+	raw_spinlock_t config25_lock;
+	raw_spinlock_t mac_ocp_lock;
 
-	spinlock_t cfg9346_usage_lock;
+	raw_spinlock_t cfg9346_usage_lock;
 	int cfg9346_usage_count;
 
 	unsigned supports_gmii:1;
@@ -668,20 +668,20 @@ static void rtl_lock_config_regs(struct rtl8169_private *tp)
 {
 	unsigned long flags;
 
-	spin_lock_irqsave(&tp->cfg9346_usage_lock, flags);
+	raw_spin_lock_irqsave(&tp->cfg9346_usage_lock, flags);
 	if (!--tp->cfg9346_usage_count)
 		RTL_W8(tp, Cfg9346, Cfg9346_Lock);
-	spin_unlock_irqrestore(&tp->cfg9346_usage_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->cfg9346_usage_lock, flags);
 }
 
 static void rtl_unlock_config_regs(struct rtl8169_private *tp)
 {
 	unsigned long flags;
 
-	spin_lock_irqsave(&tp->cfg9346_usage_lock, flags);
+	raw_spin_lock_irqsave(&tp->cfg9346_usage_lock, flags);
 	if (!tp->cfg9346_usage_count++)
 		RTL_W8(tp, Cfg9346, Cfg9346_Unlock);
-	spin_unlock_irqrestore(&tp->cfg9346_usage_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->cfg9346_usage_lock, flags);
 }
 
 static void rtl_pci_commit(struct rtl8169_private *tp)
@@ -695,10 +695,10 @@ static void rtl_mod_config2(struct rtl8169_private *tp, u8 clear, u8 set)
 	unsigned long flags;
 	u8 val;
 
-	spin_lock_irqsave(&tp->config25_lock, flags);
+	raw_spin_lock_irqsave(&tp->config25_lock, flags);
 	val = RTL_R8(tp, Config2);
 	RTL_W8(tp, Config2, (val & ~clear) | set);
-	spin_unlock_irqrestore(&tp->config25_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->config25_lock, flags);
 }
 
 static void rtl_mod_config5(struct rtl8169_private *tp, u8 clear, u8 set)
@@ -706,10 +706,10 @@ static void rtl_mod_config5(struct rtl8169_private *tp, u8 clear, u8 set)
 	unsigned long flags;
 	u8 val;
 
-	spin_lock_irqsave(&tp->config25_lock, flags);
+	raw_spin_lock_irqsave(&tp->config25_lock, flags);
 	val = RTL_R8(tp, Config5);
 	RTL_W8(tp, Config5, (val & ~clear) | set);
-	spin_unlock_irqrestore(&tp->config25_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->config25_lock, flags);
 }
 
 static bool rtl_is_8125(struct rtl8169_private *tp)
@@ -896,9 +896,9 @@ static void r8168_mac_ocp_write(struct rtl8169_private *tp, u32 reg, u32 data)
 {
 	unsigned long flags;
 
-	spin_lock_irqsave(&tp->mac_ocp_lock, flags);
+	raw_spin_lock_irqsave(&tp->mac_ocp_lock, flags);
 	__r8168_mac_ocp_write(tp, reg, data);
-	spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
 }
 
 static u16 __r8168_mac_ocp_read(struct rtl8169_private *tp, u32 reg)
@@ -916,9 +916,9 @@ static u16 r8168_mac_ocp_read(struct rtl8169_private *tp, u32 reg)
 	unsigned long flags;
 	u16 val;
 
-	spin_lock_irqsave(&tp->mac_ocp_lock, flags);
+	raw_spin_lock_irqsave(&tp->mac_ocp_lock, flags);
 	val = __r8168_mac_ocp_read(tp, reg);
-	spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
 
 	return val;
 }
@@ -929,10 +929,10 @@ static void r8168_mac_ocp_modify(struct rtl8169_private *tp, u32 reg, u16 mask,
 	unsigned long flags;
 	u16 data;
 
-	spin_lock_irqsave(&tp->mac_ocp_lock, flags);
+	raw_spin_lock_irqsave(&tp->mac_ocp_lock, flags);
 	data = __r8168_mac_ocp_read(tp, reg);
 	__r8168_mac_ocp_write(tp, reg, (data & ~mask) | set);
-	spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->mac_ocp_lock, flags);
 }
 
 /* Work around a hw issue with RTL8168g PHY, the quirk disables
@@ -1417,14 +1417,14 @@ static void __rtl8169_set_wol(struct rtl8169_private *tp, u32 wolopts)
 			r8168_mac_ocp_modify(tp, 0xc0b6, BIT(0), 0);
 	}
 
-	spin_lock_irqsave(&tp->config25_lock, flags);
+	raw_spin_lock_irqsave(&tp->config25_lock, flags);
 	for (i = 0; i < tmp; i++) {
 		options = RTL_R8(tp, cfg[i].reg) & ~cfg[i].mask;
 		if (wolopts & cfg[i].opt)
 			options |= cfg[i].mask;
 		RTL_W8(tp, cfg[i].reg, options);
 	}
-	spin_unlock_irqrestore(&tp->config25_lock, flags);
+	raw_spin_unlock_irqrestore(&tp->config25_lock, flags);
 
 	switch (tp->mac_version) {
 	case RTL_GIGA_MAC_VER_02 ... RTL_GIGA_MAC_VER_06:
@@ -5240,9 +5240,9 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	tp->eee_adv = -1;
 	tp->ocp_base = OCP_STD_PHY_BASE;
 
-	spin_lock_init(&tp->cfg9346_usage_lock);
-	spin_lock_init(&tp->config25_lock);
-	spin_lock_init(&tp->mac_ocp_lock);
+	raw_spin_lock_init(&tp->cfg9346_usage_lock);
+	raw_spin_lock_init(&tp->config25_lock);
+	raw_spin_lock_init(&tp->mac_ocp_lock);
 
 	dev->tstats = devm_netdev_alloc_pcpu_stats(&pdev->dev,
 						   struct pcpu_sw_netstats);
-- 
2.40.1.445.gf85cd430b1

From 91fe8a95afe885fd6e038fb6bd9d9c6cbd62a00e Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 10 May 2023 20:10:02 -0500
Subject: [PATCH 24/24] x86/csum: Improve performance of `csum_partial`

1) Add special case for len == 40 as that is the hottest value. The
   nets a ~8-9% latency improvement and a ~30% throughput improvement
   in the len == 40 case.

2) Use multiple accumulators in the 64-byte loop. This dramatically
   improves ILP and results in up to a 40% latency/throughput
   improvement (better for more iterations).

Results from benchmarking on Icelake. Times measured with rdtsc()
 len   lat_new   lat_old      r    tput_new  tput_old      r
   8      3.58      3.47  1.032        3.58      3.51  1.021
  16      4.14      4.02  1.028        3.96      3.78  1.046
  24      4.99      5.03  0.992        4.23      4.03  1.050
  32      5.09      5.08  1.001        4.68      4.47  1.048
  40      5.57      6.08  0.916        3.05      4.43  0.690
  48      6.65      6.63  1.003        4.97      4.69  1.059
  56      7.74      7.72  1.003        5.22      4.95  1.055
  64      6.65      7.22  0.921        6.38      6.42  0.994
  96      9.43      9.96  0.946        7.46      7.54  0.990
 128      9.39     12.15  0.773        8.90      8.79  1.012
 200     12.65     18.08  0.699       11.63     11.60  1.002
 272     15.82     23.37  0.677       14.43     14.35  1.005
 440     24.12     36.43  0.662       21.57     22.69  0.951
 952     46.20     74.01  0.624       42.98     53.12  0.809
1024     47.12     78.24  0.602       46.36     58.83  0.788
1552     72.01    117.30  0.614       71.92     96.78  0.743
2048     93.07    153.25  0.607       93.28    137.20  0.680
2600    114.73    194.30  0.590      114.28    179.32  0.637
3608    156.34    268.41  0.582      154.97    254.02  0.610
4096    175.01    304.03  0.576      175.89    292.08  0.602

There is no such thing as a free lunch, however, and the special case
for len == 40 does add overhead to the len != 40 cases. This seems to
amount to be ~5% throughput and slightly less in terms of latency.

Testing:
Part of this change is a new kunit test. The tests check all
alignment X length pairs in [0, 64) X [0, 512).
There are three cases.
    1) Precomputed random inputs/seed. The expected results where
       generated use the generic implementation (which is assumed to be
       non-buggy).
    2) An input of all 1s. The goal of this test is to catch any case
       a carry is missing.
    3) An input that never carries. The goal of this test si to catch
       any case of incorrectly carrying.

More exhaustive tests that test all alignment X length pairs in
[0, 8192) X [0, 8192] on random data are also available here:
https://github.com/goldsteinn/csum-reproduction

The reposity also has the code for reproducing the above benchmark
numbers.

Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Link: https://lore.kernel.org/all/20230511011002.935690-1-goldstein.w.n%40gmail.com
---
 arch/x86/lib/csum-partial_64.c |  97 ++++++----
 lib/Kconfig.debug              |  17 ++
 lib/Makefile                   |   1 +
 lib/checksum_kunit.c           | 334 +++++++++++++++++++++++++++++++++
 4 files changed, 417 insertions(+), 32 deletions(-)
 create mode 100644 lib/checksum_kunit.c

diff --git a/arch/x86/lib/csum-partial_64.c b/arch/x86/lib/csum-partial_64.c
index 50734a230..fe5861951 100644
--- a/arch/x86/lib/csum-partial_64.c
+++ b/arch/x86/lib/csum-partial_64.c
@@ -5,22 +5,32 @@
  * This file contains network checksum routines that are better done
  * in an architecture-specific manner due to speed.
  */
- 
+
 #include <linux/compiler.h>
 #include <linux/export.h>
 #include <asm/checksum.h>
 #include <asm/word-at-a-time.h>
 
-static inline unsigned short from32to16(unsigned a) 
+static inline unsigned short from32to16(unsigned a)
 {
-	unsigned short b = a >> 16; 
+	unsigned short b = a >> 16;
 	asm("addw %w2,%w0\n\t"
-	    "adcw $0,%w0\n" 
+	    "adcw $0,%w0\n"
 	    : "=r" (b)
 	    : "0" (b), "r" (a));
 	return b;
 }
 
+static inline __wsum csum_tail(unsigned int result, u64 temp64, int odd)
+{
+	result = add32_with_carry(temp64 >> 32, temp64 & 0xffffffff);
+	if (unlikely(odd)) {
+		result = from32to16(result);
+		result = ((result >> 8) & 0xff) | ((result & 0xff) << 8);
+	}
+	return (__force __wsum)result;
+}
+
 /*
  * Do a checksum on an arbitrary memory area.
  * Returns a 32bit checksum.
@@ -47,21 +57,52 @@ __wsum csum_partial(const void *buff, int len, __wsum sum)
 		buff++;
 	}
 
-	while (unlikely(len >= 64)) {
+	/*
+	 * len == 40 is the hot case due to IPv6 headers, but annotating it likely()
+	 * has noticeable negative affect on codegen for all other cases with
+	 * minimal performance benefit here.
+	 */
+	if (len == 40) {
 		asm("addq 0*8(%[src]),%[res]\n\t"
 		    "adcq 1*8(%[src]),%[res]\n\t"
 		    "adcq 2*8(%[src]),%[res]\n\t"
 		    "adcq 3*8(%[src]),%[res]\n\t"
 		    "adcq 4*8(%[src]),%[res]\n\t"
-		    "adcq 5*8(%[src]),%[res]\n\t"
-		    "adcq 6*8(%[src]),%[res]\n\t"
-		    "adcq 7*8(%[src]),%[res]\n\t"
 		    "adcq $0,%[res]"
-		    : [res] "+r" (temp64)
-		    : [src] "r" (buff)
-		    : "memory");
-		buff += 64;
-		len -= 64;
+		    : [res] "+r"(temp64)
+		    : [src] "r"(buff), "m"(*(const char(*)[40])buff));
+		return csum_tail(result, temp64, odd);
+	}
+	if (unlikely(len >= 64)) {
+		/*
+		 * Extra accumulators for better ILP in the loop.
+		 */
+		u64 tmp_accum, tmp_carries;
+
+		asm("xorl %k[tmp_accum],%k[tmp_accum]\n\t"
+		    "xorl %k[tmp_carries],%k[tmp_carries]\n\t"
+		    "subl $64, %[len]\n\t"
+		    "1:\n\t"
+		    "addq 0*8(%[src]),%[res]\n\t"
+		    "adcq 1*8(%[src]),%[res]\n\t"
+		    "adcq 2*8(%[src]),%[res]\n\t"
+		    "adcq 3*8(%[src]),%[res]\n\t"
+		    "adcl $0,%k[tmp_carries]\n\t"
+		    "addq 4*8(%[src]),%[tmp_accum]\n\t"
+		    "adcq 5*8(%[src]),%[tmp_accum]\n\t"
+		    "adcq 6*8(%[src]),%[tmp_accum]\n\t"
+		    "adcq 7*8(%[src]),%[tmp_accum]\n\t"
+		    "adcl $0,%k[tmp_carries]\n\t"
+		    "addq $64, %[src]\n\t"
+		    "subl $64, %[len]\n\t"
+		    "jge 1b\n\t"
+		    "addq %[tmp_accum],%[res]\n\t"
+		    "adcq %[tmp_carries],%[res]\n\t"
+		    "adcq $0,%[res]"
+		    : [tmp_accum] "=&r"(tmp_accum),
+		      [tmp_carries] "=&r"(tmp_carries), [res] "+r"(temp64),
+		      [len] "+r"(len), [src] "+r"(buff)
+		    : "m"(*(const char *)buff));
 	}
 
 	if (len & 32) {
@@ -70,45 +111,37 @@ __wsum csum_partial(const void *buff, int len, __wsum sum)
 		    "adcq 2*8(%[src]),%[res]\n\t"
 		    "adcq 3*8(%[src]),%[res]\n\t"
 		    "adcq $0,%[res]"
-			: [res] "+r" (temp64)
-			: [src] "r" (buff)
-			: "memory");
+		    : [res] "+r"(temp64)
+		    : [src] "r"(buff), "m"(*(const char(*)[32])buff));
 		buff += 32;
 	}
 	if (len & 16) {
 		asm("addq 0*8(%[src]),%[res]\n\t"
 		    "adcq 1*8(%[src]),%[res]\n\t"
 		    "adcq $0,%[res]"
-			: [res] "+r" (temp64)
-			: [src] "r" (buff)
-			: "memory");
+		    : [res] "+r"(temp64)
+		    : [src] "r"(buff), "m"(*(const char(*)[16])buff));
 		buff += 16;
 	}
 	if (len & 8) {
 		asm("addq 0*8(%[src]),%[res]\n\t"
 		    "adcq $0,%[res]"
-			: [res] "+r" (temp64)
-			: [src] "r" (buff)
-			: "memory");
+		    : [res] "+r"(temp64)
+		    : [src] "r"(buff), "m"(*(const char(*)[8])buff));
 		buff += 8;
 	}
 	if (len & 7) {
-		unsigned int shift = (8 - (len & 7)) * 8;
+		unsigned int shift = (-len << 3) & 63;
 		unsigned long trail;
 
 		trail = (load_unaligned_zeropad(buff) << shift) >> shift;
 
 		asm("addq %[trail],%[res]\n\t"
 		    "adcq $0,%[res]"
-			: [res] "+r" (temp64)
-			: [trail] "r" (trail));
+		    : [res] "+r"(temp64)
+		    : [trail] "r"(trail));
 	}
-	result = add32_with_carry(temp64 >> 32, temp64 & 0xffffffff);
-	if (unlikely(odd)) {
-		result = from32to16(result);
-		result = ((result >> 8) & 0xff) | ((result & 0xff) << 8);
-	}
-	return (__force __wsum)result;
+	return csum_tail(result, temp64, odd);
 }
 EXPORT_SYMBOL(csum_partial);
 
@@ -118,6 +151,6 @@ EXPORT_SYMBOL(csum_partial);
  */
 __sum16 ip_compute_csum(const void *buff, int len)
 {
-	return csum_fold(csum_partial(buff,len,0));
+	return csum_fold(csum_partial(buff, len, 0));
 }
 EXPORT_SYMBOL(ip_compute_csum);
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 39d1d9316..b959cdd22 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -2422,6 +2422,23 @@ config BITFIELD_KUNIT
 
 	  If unsure, say N.
 
+config CHECKSUM_KUNIT
+	tristate "KUnit test checksum functions at runtime" if !KUNIT_ALL_TESTS
+	depends on KUNIT
+	default KUNIT_ALL_TESTS
+	help
+	  Enable this option to test the checksum functions at boot.
+
+	  KUnit tests run during boot and output the results to the debug log
+	  in TAP format (http://testanything.org/). Only useful for kernel devs
+	  running the KUnit test harness, and not intended for inclusion into a
+	  production build.
+
+	  For more information on KUnit and unit tests in general please refer
+	  to the KUnit documentation in Documentation/dev-tools/kunit/.
+
+	  If unsure, say N.
+
 config HASH_KUNIT_TEST
 	tristate "KUnit Test for integer hash functions" if !KUNIT_ALL_TESTS
 	depends on KUNIT
diff --git a/lib/Makefile b/lib/Makefile
index baf2821f7..5c5bcf39c 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -374,6 +374,7 @@ obj-$(CONFIG_PLDMFW) += pldmfw/
 # KUnit tests
 CFLAGS_bitfield_kunit.o := $(DISABLE_STRUCTLEAK_PLUGIN)
 obj-$(CONFIG_BITFIELD_KUNIT) += bitfield_kunit.o
+obj-$(CONFIG_CHECKSUM_KUNIT) += checksum_kunit.o
 obj-$(CONFIG_LIST_KUNIT_TEST) += list-test.o
 obj-$(CONFIG_HASHTABLE_KUNIT_TEST) += hashtable_test.o
 obj-$(CONFIG_LINEAR_RANGES_TEST) += test_linear_ranges.o
diff --git a/lib/checksum_kunit.c b/lib/checksum_kunit.c
new file mode 100644
index 000000000..ace3c4799
--- /dev/null
+++ b/lib/checksum_kunit.c
@@ -0,0 +1,334 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Test cases csum_partial and csum_fold
+ */
+
+#include <kunit/test.h>
+#include <asm/checksum.h>
+
+#define MAX_LEN 512
+#define MAX_ALIGN 64
+#define TEST_BUFLEN (MAX_LEN + MAX_ALIGN)
+
+static const __wsum random_init_sum = 0x2847aab;
+static const u8 random_buf[] = {
+	0xac, 0xd7, 0x76, 0x69, 0x6e, 0xf2, 0x93, 0x2c, 0x1f, 0xe0, 0xde, 0x86,
+	0x8f, 0x54, 0x33, 0x90, 0x95, 0xbf, 0xff, 0xb9, 0xea, 0x62, 0x6e, 0xb5,
+	0xd3, 0x4f, 0xf5, 0x60, 0x50, 0x5c, 0xc7, 0xfa, 0x6d, 0x1a, 0xc7, 0xf0,
+	0xd2, 0x2c, 0x12, 0x3d, 0x88, 0xe3, 0x14, 0x21, 0xb1, 0x5e, 0x45, 0x31,
+	0xa2, 0x85, 0x36, 0x76, 0xba, 0xd8, 0xad, 0xbb, 0x9e, 0x49, 0x8f, 0xf7,
+	0xce, 0xea, 0xef, 0xca, 0x2c, 0x29, 0xf7, 0x15, 0x5c, 0x1d, 0x4d, 0x09,
+	0x1f, 0xe2, 0x14, 0x31, 0x8c, 0x07, 0x57, 0x23, 0x1f, 0x6f, 0x03, 0xe1,
+	0x93, 0x19, 0x53, 0x03, 0x45, 0x49, 0x9a, 0x3b, 0x8e, 0x0c, 0x12, 0x5d,
+	0x8a, 0xb8, 0x9b, 0x8c, 0x9a, 0x03, 0xe5, 0xa2, 0x43, 0xd2, 0x3b, 0x4e,
+	0x7e, 0x30, 0x3c, 0x22, 0x2d, 0xc5, 0xfc, 0x9e, 0xdb, 0xc6, 0xf9, 0x69,
+	0x12, 0x39, 0x1f, 0xa0, 0x11, 0x0c, 0x3f, 0xf5, 0x53, 0xc9, 0x30, 0xfb,
+	0xb0, 0xdd, 0x21, 0x1d, 0x34, 0xe2, 0x65, 0x30, 0xf1, 0xe8, 0x1b, 0xe7,
+	0x55, 0x0d, 0xeb, 0xbd, 0xcc, 0x9d, 0x24, 0xa4, 0xad, 0xa7, 0x93, 0x47,
+	0x19, 0x2e, 0xc4, 0x5c, 0x3b, 0xc7, 0x6d, 0x95, 0x0c, 0x47, 0x60, 0xaf,
+	0x5b, 0x47, 0xee, 0xdc, 0x31, 0x31, 0x14, 0x12, 0x7e, 0x9e, 0x45, 0xb1,
+	0xc1, 0x69, 0x4b, 0x84, 0xfc, 0x88, 0xc1, 0x9e, 0x46, 0xb4, 0xc2, 0x25,
+	0xc5, 0x6c, 0x4c, 0x22, 0x58, 0x5c, 0xbe, 0xff, 0xea, 0x88, 0x88, 0x7a,
+	0xcb, 0x1c, 0x5d, 0x63, 0xa1, 0xf2, 0x33, 0x0c, 0xa2, 0x16, 0x0b, 0x6e,
+	0x2b, 0x79, 0x58, 0xf7, 0xac, 0xd3, 0x6a, 0x3f, 0x81, 0x57, 0x48, 0x45,
+	0xe3, 0x7c, 0xdc, 0xd6, 0x34, 0x7e, 0xe6, 0x73, 0xfa, 0xcb, 0x31, 0x18,
+	0xa9, 0x0b, 0xee, 0x6b, 0x99, 0xb9, 0x2d, 0xde, 0x22, 0x0e, 0x71, 0x57,
+	0x0e, 0x9b, 0x11, 0xd1, 0x15, 0x41, 0xd0, 0x6b, 0x50, 0x8a, 0x23, 0x64,
+	0xe3, 0x9c, 0xb3, 0x55, 0x09, 0xe9, 0x32, 0x67, 0xf9, 0xe0, 0x73, 0xf1,
+	0x60, 0x66, 0x0b, 0x88, 0x79, 0x8d, 0x4b, 0x52, 0x83, 0x20, 0x26, 0x78,
+	0x49, 0x27, 0xe7, 0x3e, 0x29, 0xa8, 0x18, 0x82, 0x41, 0xdd, 0x1e, 0xcc,
+	0x3b, 0xc4, 0x65, 0xd1, 0x21, 0x40, 0x72, 0xb2, 0x87, 0x5e, 0x16, 0x10,
+	0x80, 0x3f, 0x4b, 0x58, 0x1c, 0xc2, 0x79, 0x20, 0xf0, 0xe0, 0x80, 0xd3,
+	0x52, 0xa5, 0x19, 0x6e, 0x47, 0x90, 0x08, 0xf5, 0x50, 0xe2, 0xd6, 0xae,
+	0xe9, 0x2e, 0xdc, 0xd5, 0xb4, 0x90, 0x1f, 0x79, 0x49, 0x82, 0x21, 0x84,
+	0xa0, 0xb5, 0x2f, 0xff, 0x30, 0x71, 0xed, 0x80, 0x68, 0xb1, 0x6d, 0xef,
+	0xf6, 0xcf, 0xb8, 0x41, 0x79, 0xf5, 0x01, 0xbc, 0x0c, 0x9b, 0x0e, 0x06,
+	0xf3, 0xb0, 0xbb, 0x97, 0xb8, 0xb1, 0xfd, 0x51, 0x4e, 0xef, 0x0a, 0x3d,
+	0x7a, 0x3d, 0xbd, 0x61, 0x00, 0xa2, 0xb3, 0xf0, 0x1d, 0x77, 0x7b, 0x6c,
+	0x01, 0x61, 0xa5, 0xa3, 0xdb, 0xd5, 0xd5, 0xf4, 0xb5, 0x28, 0x9f, 0x0a,
+	0xa3, 0x82, 0x5f, 0x4b, 0x40, 0x0f, 0x05, 0x0e, 0x78, 0xed, 0xbf, 0x17,
+	0xf6, 0x5a, 0x8a, 0x7d, 0xf9, 0x45, 0xc1, 0xd7, 0x1b, 0x9d, 0x6c, 0x07,
+	0x88, 0xf3, 0xbc, 0xf1, 0xea, 0x28, 0x1f, 0xb8, 0x7a, 0x60, 0x3c, 0xce,
+	0x3e, 0x50, 0xb2, 0x0b, 0xcf, 0xe5, 0x08, 0x1f, 0x48, 0x04, 0xf9, 0x35,
+	0x29, 0x15, 0xbe, 0x82, 0x96, 0xc2, 0x55, 0x04, 0x6c, 0x19, 0x45, 0x29,
+	0x0b, 0xb6, 0x49, 0x12, 0xfb, 0x8d, 0x1b, 0x75, 0x8b, 0xd9, 0x6a, 0x5c,
+	0xbe, 0x46, 0x2b, 0x41, 0xfe, 0x21, 0xad, 0x1f, 0x75, 0xe7, 0x90, 0x3d,
+	0xe1, 0xdf, 0x4b, 0xe1, 0x81, 0xe2, 0x17, 0x02, 0x7b, 0x58, 0x8b, 0x92,
+	0x1a, 0xac, 0x46, 0xdd, 0x2e, 0xce, 0x40, 0x09
+};
+static const __sum16 expected_results[] = {
+	0x82d0, 0x8224, 0xab23, 0xaaad, 0x41ad, 0x413f, 0x4f3e, 0x4eab, 0x22ab,
+	0x228c, 0x428b, 0x41ad, 0xbbac, 0xbb1d, 0x671d, 0x66ea, 0xd6e9, 0xd654,
+	0x1754, 0x1655, 0x5d54, 0x5c6a, 0xfa69, 0xf9fb, 0x44fb, 0x4428, 0xf527,
+	0xf432, 0x9432, 0x93e2, 0x37e2, 0x371b, 0x3d1a, 0x3cad, 0x22ad, 0x21e6,
+	0x31e5, 0x3113, 0x0513, 0x0501, 0xc800, 0xc778, 0xe477, 0xe463, 0xc363,
+	0xc2b2, 0x64b2, 0x646d, 0x336d, 0x32cb, 0xadca, 0xad94, 0x3794, 0x36da,
+	0x5ed9, 0x5e2c, 0xa32b, 0xa28d, 0x598d, 0x58fe, 0x61fd, 0x612f, 0x772e,
+	0x763f, 0xac3e, 0xac12, 0x8312, 0x821b, 0x6d1b, 0x6cbf, 0x4fbf, 0x4f72,
+	0x4672, 0x4653, 0x6452, 0x643e, 0x333e, 0x32b2, 0x2bb2, 0x2b5b, 0x085b,
+	0x083c, 0x993b, 0x9938, 0xb837, 0xb7a4, 0x9ea4, 0x9e51, 0x9b51, 0x9b0c,
+	0x520c, 0x5172, 0x1672, 0x15e4, 0x09e4, 0x09d2, 0xacd1, 0xac47, 0xf446,
+	0xf3ab, 0x67ab, 0x6711, 0x6411, 0x632c, 0xc12b, 0xc0e8, 0xeee7, 0xeeac,
+	0xa0ac, 0xa02e, 0x702e, 0x6ff2, 0x4df2, 0x4dc5, 0x88c4, 0x87c8, 0xe9c7,
+	0xe8ec, 0x22ec, 0x21f3, 0xb8f2, 0xb8e0, 0x7fe0, 0x7fc1, 0xdfc0, 0xdfaf,
+	0xd3af, 0xd370, 0xde6f, 0xde1c, 0x151c, 0x14ec, 0x19eb, 0x193b, 0x3c3a,
+	0x3c19, 0x1f19, 0x1ee5, 0x3ce4, 0x3c7f, 0x0c7f, 0x0b8e, 0x238d, 0x2372,
+	0x3c71, 0x3c1c, 0x2f1c, 0x2e31, 0x7130, 0x7064, 0xd363, 0xd33f, 0x2f3f,
+	0x2e92, 0x8791, 0x86fe, 0x3ffe, 0x3fe5, 0x11e5, 0x1121, 0xb520, 0xb4e5,
+	0xede4, 0xed77, 0x5877, 0x586b, 0x116b, 0x110b, 0x620a, 0x61af, 0x1aaf,
+	0x19c1, 0x3dc0, 0x3d8f, 0x0c8f, 0x0c7b, 0xfa7a, 0xf9fc, 0x5bfc, 0x5bb7,
+	0xaab6, 0xa9f5, 0x40f5, 0x40aa, 0xbca9, 0xbbad, 0x33ad, 0x32ec, 0x94eb,
+	0x94a5, 0xe0a4, 0xdfe2, 0xbae2, 0xba1d, 0x4e1d, 0x4dd1, 0x2bd1, 0x2b79,
+	0xcf78, 0xceba, 0xcfb9, 0xcecf, 0x46cf, 0x4647, 0xcc46, 0xcb7b, 0xaf7b,
+	0xaf1e, 0x4c1e, 0x4b7d, 0x597c, 0x5949, 0x4d49, 0x4ca7, 0x36a7, 0x369c,
+	0xc89b, 0xc870, 0x4f70, 0x4f18, 0x5817, 0x576b, 0x846a, 0x8400, 0x4500,
+	0x447f, 0xed7e, 0xed36, 0xa836, 0xa753, 0x2b53, 0x2a77, 0x5476, 0x5442,
+	0xd641, 0xd55b, 0x625b, 0x6161, 0x9660, 0x962f, 0x7e2f, 0x7d86, 0x7286,
+	0x7198, 0x0698, 0x05ff, 0x4cfe, 0x4cd1, 0x6ed0, 0x6eae, 0x60ae, 0x603d,
+	0x093d, 0x092f, 0x6e2e, 0x6e1d, 0x9d1c, 0x9d07, 0x5c07, 0x5b37, 0xf036,
+	0xefe6, 0x65e6, 0x65c3, 0x01c3, 0x00e0, 0x64df, 0x642c, 0x0f2c, 0x0f23,
+	0x2622, 0x25f0, 0xbeef, 0xbdf6, 0xddf5, 0xdd82, 0xec81, 0xec21, 0x8621,
+	0x8616, 0xfe15, 0xfd9c, 0x709c, 0x7051, 0x1e51, 0x1dce, 0xfdcd, 0xfda7,
+	0x85a7, 0x855e, 0x5e5e, 0x5d77, 0x1f77, 0x1f4e, 0x774d, 0x7735, 0xf534,
+	0xf4f3, 0x17f3, 0x17d5, 0x4bd4, 0x4b99, 0x8798, 0x8733, 0xb632, 0xb611,
+	0x7611, 0x759f, 0xc39e, 0xc317, 0x6517, 0x6501, 0x5501, 0x5481, 0x1581,
+	0x1536, 0xbd35, 0xbd19, 0xfb18, 0xfa9f, 0xda9f, 0xd9af, 0xf9ae, 0xf92e,
+	0x262e, 0x25dc, 0x80db, 0x80c2, 0x12c2, 0x127b, 0x827a, 0x8272, 0x8d71,
+	0x8d21, 0xab20, 0xaa4a, 0xfc49, 0xfb60, 0xcd60, 0xcc84, 0xf783, 0xf6cf,
+	0x66cf, 0x66b0, 0xedaf, 0xed66, 0x6b66, 0x6b45, 0xe744, 0xe6a4, 0x31a4,
+	0x3175, 0x3274, 0x3244, 0xc143, 0xc056, 0x4056, 0x3fee, 0x8eed, 0x8e80,
+	0x9f7f, 0x9e89, 0xcf88, 0xced0, 0x8dd0, 0x8d57, 0x9856, 0x9855, 0xdc54,
+	0xdc48, 0x4148, 0x413a, 0x3b3a, 0x3a47, 0x8a46, 0x898b, 0xf28a, 0xf1d2,
+	0x40d2, 0x3fd5, 0xeed4, 0xee86, 0xff85, 0xff7b, 0xc27b, 0xc201, 0x8501,
+	0x8444, 0x2344, 0x2344, 0x8143, 0x8090, 0x908f, 0x9072, 0x1972, 0x18f7,
+	0xacf6, 0xacf5, 0x4bf5, 0x4b50, 0xa84f, 0xa774, 0xd273, 0xd19e, 0xdd9d,
+	0xdce8, 0xb4e8, 0xb449, 0xaa49, 0xa9a6, 0x27a6, 0x2747, 0xdc46, 0xdc06,
+	0xcd06, 0xcd01, 0xbf01, 0xbe89, 0xd188, 0xd0c9, 0xb9c9, 0xb8d3, 0x5ed3,
+	0x5e49, 0xe148, 0xe04f, 0x9b4f, 0x9a8e, 0xc38d, 0xc372, 0x2672, 0x2606,
+	0x1f06, 0x1e7e, 0x2b7d, 0x2ac1, 0x39c0, 0x38d6, 0x10d6, 0x10b7, 0x58b6,
+	0x583c, 0xf83b, 0xf7ff, 0x29ff, 0x29c1, 0xd9c0, 0xd90e, 0xce0e, 0xcd3f,
+	0xe83e, 0xe836, 0xc936, 0xc8ee, 0xc4ee, 0xc3f5, 0x8ef5, 0x8ecc, 0x79cc,
+	0x790e, 0xf70d, 0xf677, 0x3477, 0x3422, 0x3022, 0x2fb6, 0x16b6, 0x1671,
+	0xed70, 0xed65, 0x3765, 0x371c, 0x251c, 0x2421, 0x9720, 0x9705, 0x2205,
+	0x217a, 0x4879, 0x480f, 0xec0e, 0xeb50, 0xa550, 0xa525, 0x6425, 0x6327,
+	0x4227, 0x417a, 0x227a, 0x2205, 0x3b04, 0x3a74, 0xfd73, 0xfc92, 0x1d92,
+	0x1d47, 0x3c46, 0x3bc5, 0x59c4, 0x59ad, 0x57ad, 0x5732, 0xff31, 0xfea6,
+	0x6ca6, 0x6c8c, 0xc08b, 0xc045, 0xe344, 0xe316, 0x1516, 0x14d6,
+};
+static const __wsum init_sums_no_overflow[] = {
+	0xffffffff, 0xfffffffb, 0xfffffbfb, 0xfffffbf7, 0xfffff7f7, 0xfffff7f3,
+	0xfffff3f3, 0xfffff3ef, 0xffffefef, 0xffffefeb, 0xffffebeb, 0xffffebe7,
+	0xffffe7e7, 0xffffe7e3, 0xffffe3e3, 0xffffe3df, 0xffffdfdf, 0xffffdfdb,
+	0xffffdbdb, 0xffffdbd7, 0xffffd7d7, 0xffffd7d3, 0xffffd3d3, 0xffffd3cf,
+	0xffffcfcf, 0xffffcfcb, 0xffffcbcb, 0xffffcbc7, 0xffffc7c7, 0xffffc7c3,
+	0xffffc3c3, 0xffffc3bf, 0xffffbfbf, 0xffffbfbb, 0xffffbbbb, 0xffffbbb7,
+	0xffffb7b7, 0xffffb7b3, 0xffffb3b3, 0xffffb3af, 0xffffafaf, 0xffffafab,
+	0xffffabab, 0xffffaba7, 0xffffa7a7, 0xffffa7a3, 0xffffa3a3, 0xffffa39f,
+	0xffff9f9f, 0xffff9f9b, 0xffff9b9b, 0xffff9b97, 0xffff9797, 0xffff9793,
+	0xffff9393, 0xffff938f, 0xffff8f8f, 0xffff8f8b, 0xffff8b8b, 0xffff8b87,
+	0xffff8787, 0xffff8783, 0xffff8383, 0xffff837f, 0xffff7f7f, 0xffff7f7b,
+	0xffff7b7b, 0xffff7b77, 0xffff7777, 0xffff7773, 0xffff7373, 0xffff736f,
+	0xffff6f6f, 0xffff6f6b, 0xffff6b6b, 0xffff6b67, 0xffff6767, 0xffff6763,
+	0xffff6363, 0xffff635f, 0xffff5f5f, 0xffff5f5b, 0xffff5b5b, 0xffff5b57,
+	0xffff5757, 0xffff5753, 0xffff5353, 0xffff534f, 0xffff4f4f, 0xffff4f4b,
+	0xffff4b4b, 0xffff4b47, 0xffff4747, 0xffff4743, 0xffff4343, 0xffff433f,
+	0xffff3f3f, 0xffff3f3b, 0xffff3b3b, 0xffff3b37, 0xffff3737, 0xffff3733,
+	0xffff3333, 0xffff332f, 0xffff2f2f, 0xffff2f2b, 0xffff2b2b, 0xffff2b27,
+	0xffff2727, 0xffff2723, 0xffff2323, 0xffff231f, 0xffff1f1f, 0xffff1f1b,
+	0xffff1b1b, 0xffff1b17, 0xffff1717, 0xffff1713, 0xffff1313, 0xffff130f,
+	0xffff0f0f, 0xffff0f0b, 0xffff0b0b, 0xffff0b07, 0xffff0707, 0xffff0703,
+	0xffff0303, 0xffff02ff, 0xfffffefe, 0xfffffefa, 0xfffffafa, 0xfffffaf6,
+	0xfffff6f6, 0xfffff6f2, 0xfffff2f2, 0xfffff2ee, 0xffffeeee, 0xffffeeea,
+	0xffffeaea, 0xffffeae6, 0xffffe6e6, 0xffffe6e2, 0xffffe2e2, 0xffffe2de,
+	0xffffdede, 0xffffdeda, 0xffffdada, 0xffffdad6, 0xffffd6d6, 0xffffd6d2,
+	0xffffd2d2, 0xffffd2ce, 0xffffcece, 0xffffceca, 0xffffcaca, 0xffffcac6,
+	0xffffc6c6, 0xffffc6c2, 0xffffc2c2, 0xffffc2be, 0xffffbebe, 0xffffbeba,
+	0xffffbaba, 0xffffbab6, 0xffffb6b6, 0xffffb6b2, 0xffffb2b2, 0xffffb2ae,
+	0xffffaeae, 0xffffaeaa, 0xffffaaaa, 0xffffaaa6, 0xffffa6a6, 0xffffa6a2,
+	0xffffa2a2, 0xffffa29e, 0xffff9e9e, 0xffff9e9a, 0xffff9a9a, 0xffff9a96,
+	0xffff9696, 0xffff9692, 0xffff9292, 0xffff928e, 0xffff8e8e, 0xffff8e8a,
+	0xffff8a8a, 0xffff8a86, 0xffff8686, 0xffff8682, 0xffff8282, 0xffff827e,
+	0xffff7e7e, 0xffff7e7a, 0xffff7a7a, 0xffff7a76, 0xffff7676, 0xffff7672,
+	0xffff7272, 0xffff726e, 0xffff6e6e, 0xffff6e6a, 0xffff6a6a, 0xffff6a66,
+	0xffff6666, 0xffff6662, 0xffff6262, 0xffff625e, 0xffff5e5e, 0xffff5e5a,
+	0xffff5a5a, 0xffff5a56, 0xffff5656, 0xffff5652, 0xffff5252, 0xffff524e,
+	0xffff4e4e, 0xffff4e4a, 0xffff4a4a, 0xffff4a46, 0xffff4646, 0xffff4642,
+	0xffff4242, 0xffff423e, 0xffff3e3e, 0xffff3e3a, 0xffff3a3a, 0xffff3a36,
+	0xffff3636, 0xffff3632, 0xffff3232, 0xffff322e, 0xffff2e2e, 0xffff2e2a,
+	0xffff2a2a, 0xffff2a26, 0xffff2626, 0xffff2622, 0xffff2222, 0xffff221e,
+	0xffff1e1e, 0xffff1e1a, 0xffff1a1a, 0xffff1a16, 0xffff1616, 0xffff1612,
+	0xffff1212, 0xffff120e, 0xffff0e0e, 0xffff0e0a, 0xffff0a0a, 0xffff0a06,
+	0xffff0606, 0xffff0602, 0xffff0202, 0xffff01fe, 0xfffffdfd, 0xfffffdf9,
+	0xfffff9f9, 0xfffff9f5, 0xfffff5f5, 0xfffff5f1, 0xfffff1f1, 0xfffff1ed,
+	0xffffeded, 0xffffede9, 0xffffe9e9, 0xffffe9e5, 0xffffe5e5, 0xffffe5e1,
+	0xffffe1e1, 0xffffe1dd, 0xffffdddd, 0xffffddd9, 0xffffd9d9, 0xffffd9d5,
+	0xffffd5d5, 0xffffd5d1, 0xffffd1d1, 0xffffd1cd, 0xffffcdcd, 0xffffcdc9,
+	0xffffc9c9, 0xffffc9c5, 0xffffc5c5, 0xffffc5c1, 0xffffc1c1, 0xffffc1bd,
+	0xffffbdbd, 0xffffbdb9, 0xffffb9b9, 0xffffb9b5, 0xffffb5b5, 0xffffb5b1,
+	0xffffb1b1, 0xffffb1ad, 0xffffadad, 0xffffada9, 0xffffa9a9, 0xffffa9a5,
+	0xffffa5a5, 0xffffa5a1, 0xffffa1a1, 0xffffa19d, 0xffff9d9d, 0xffff9d99,
+	0xffff9999, 0xffff9995, 0xffff9595, 0xffff9591, 0xffff9191, 0xffff918d,
+	0xffff8d8d, 0xffff8d89, 0xffff8989, 0xffff8985, 0xffff8585, 0xffff8581,
+	0xffff8181, 0xffff817d, 0xffff7d7d, 0xffff7d79, 0xffff7979, 0xffff7975,
+	0xffff7575, 0xffff7571, 0xffff7171, 0xffff716d, 0xffff6d6d, 0xffff6d69,
+	0xffff6969, 0xffff6965, 0xffff6565, 0xffff6561, 0xffff6161, 0xffff615d,
+	0xffff5d5d, 0xffff5d59, 0xffff5959, 0xffff5955, 0xffff5555, 0xffff5551,
+	0xffff5151, 0xffff514d, 0xffff4d4d, 0xffff4d49, 0xffff4949, 0xffff4945,
+	0xffff4545, 0xffff4541, 0xffff4141, 0xffff413d, 0xffff3d3d, 0xffff3d39,
+	0xffff3939, 0xffff3935, 0xffff3535, 0xffff3531, 0xffff3131, 0xffff312d,
+	0xffff2d2d, 0xffff2d29, 0xffff2929, 0xffff2925, 0xffff2525, 0xffff2521,
+	0xffff2121, 0xffff211d, 0xffff1d1d, 0xffff1d19, 0xffff1919, 0xffff1915,
+	0xffff1515, 0xffff1511, 0xffff1111, 0xffff110d, 0xffff0d0d, 0xffff0d09,
+	0xffff0909, 0xffff0905, 0xffff0505, 0xffff0501, 0xffff0101, 0xffff00fd,
+	0xfffffcfc, 0xfffffcf8, 0xfffff8f8, 0xfffff8f4, 0xfffff4f4, 0xfffff4f0,
+	0xfffff0f0, 0xfffff0ec, 0xffffecec, 0xffffece8, 0xffffe8e8, 0xffffe8e4,
+	0xffffe4e4, 0xffffe4e0, 0xffffe0e0, 0xffffe0dc, 0xffffdcdc, 0xffffdcd8,
+	0xffffd8d8, 0xffffd8d4, 0xffffd4d4, 0xffffd4d0, 0xffffd0d0, 0xffffd0cc,
+	0xffffcccc, 0xffffccc8, 0xffffc8c8, 0xffffc8c4, 0xffffc4c4, 0xffffc4c0,
+	0xffffc0c0, 0xffffc0bc, 0xffffbcbc, 0xffffbcb8, 0xffffb8b8, 0xffffb8b4,
+	0xffffb4b4, 0xffffb4b0, 0xffffb0b0, 0xffffb0ac, 0xffffacac, 0xffffaca8,
+	0xffffa8a8, 0xffffa8a4, 0xffffa4a4, 0xffffa4a0, 0xffffa0a0, 0xffffa09c,
+	0xffff9c9c, 0xffff9c98, 0xffff9898, 0xffff9894, 0xffff9494, 0xffff9490,
+	0xffff9090, 0xffff908c, 0xffff8c8c, 0xffff8c88, 0xffff8888, 0xffff8884,
+	0xffff8484, 0xffff8480, 0xffff8080, 0xffff807c, 0xffff7c7c, 0xffff7c78,
+	0xffff7878, 0xffff7874, 0xffff7474, 0xffff7470, 0xffff7070, 0xffff706c,
+	0xffff6c6c, 0xffff6c68, 0xffff6868, 0xffff6864, 0xffff6464, 0xffff6460,
+	0xffff6060, 0xffff605c, 0xffff5c5c, 0xffff5c58, 0xffff5858, 0xffff5854,
+	0xffff5454, 0xffff5450, 0xffff5050, 0xffff504c, 0xffff4c4c, 0xffff4c48,
+	0xffff4848, 0xffff4844, 0xffff4444, 0xffff4440, 0xffff4040, 0xffff403c,
+	0xffff3c3c, 0xffff3c38, 0xffff3838, 0xffff3834, 0xffff3434, 0xffff3430,
+	0xffff3030, 0xffff302c, 0xffff2c2c, 0xffff2c28, 0xffff2828, 0xffff2824,
+	0xffff2424, 0xffff2420, 0xffff2020, 0xffff201c, 0xffff1c1c, 0xffff1c18,
+	0xffff1818, 0xffff1814, 0xffff1414, 0xffff1410, 0xffff1010, 0xffff100c,
+	0xffff0c0c, 0xffff0c08, 0xffff0808, 0xffff0804, 0xffff0404, 0xffff0400,
+	0xffff0000, 0xfffffffb,
+};
+
+static u8 tmp_buf[TEST_BUFLEN];
+
+#define full_csum(buff, len, sum) csum_fold(csum_partial(buff, len, sum))
+
+#define CHECK_EQ(lhs, rhs) KUNIT_ASSERT_EQ(test, lhs, rhs)
+
+static void assert_setup_correct(struct kunit *test)
+{
+	CHECK_EQ(sizeof(random_buf) / sizeof(random_buf[0]), MAX_LEN);
+	CHECK_EQ(sizeof(expected_results) / sizeof(expected_results[0]),
+		 MAX_LEN);
+	CHECK_EQ(sizeof(init_sums_no_overflow) /
+			 sizeof(init_sums_no_overflow[0]),
+		 MAX_LEN);
+}
+
+/*
+ * Test with randomized input (pre determined random with known results).
+ */
+static void test_csum_fixed_random_inputs(struct kunit *test)
+{
+	int len, align;
+	__wsum result, expec, sum;
+
+	assert_setup_correct(test);
+	for (align = 0; align < TEST_BUFLEN; ++align) {
+		memcpy(&tmp_buf[align], random_buf,
+		       min(MAX_LEN, TEST_BUFLEN - align));
+		for (len = 0; len < MAX_LEN && (align + len) < TEST_BUFLEN;
+		     ++len) {
+			/*
+			 * Test the precomputed random input.
+			 */
+			sum = random_init_sum;
+			result = full_csum(&tmp_buf[align], len, sum);
+			expec = expected_results[len];
+			CHECK_EQ(result, expec);
+		}
+	}
+}
+
+/*
+ * All ones input test. If there are any missing carry operations, it fails.
+ */
+static void test_csum_all_carry_inputs(struct kunit *test)
+{
+	int len, align;
+	__wsum result, expec, sum;
+
+	assert_setup_correct(test);
+	memset(tmp_buf, 0xff, TEST_BUFLEN);
+	for (align = 0; align < TEST_BUFLEN; ++align) {
+		for (len = 0; len < MAX_LEN && (align + len) < TEST_BUFLEN;
+		     ++len) {
+			/*
+			 * All carries from input and initial sum.
+			 */
+			sum = 0xffffffff;
+			result = full_csum(&tmp_buf[align], len, sum);
+			expec = (len & 1) ? 0xff00 : 0;
+			CHECK_EQ(result, expec);
+
+			/*
+			 * All carries from input.
+			 */
+			sum = 0;
+			result = full_csum(&tmp_buf[align], len, sum);
+			if (len & 1)
+				expec = 0xff00;
+			else if (len)
+				expec = 0;
+			else
+				expec = 0xffff;
+			CHECK_EQ(result, expec);
+		}
+	}
+}
+
+/*
+ * Test with input that alone doesn't cause any carries. By selecting the
+ * maximum initial sum, this allows us to test that there are no carries
+ * where there shouldn't be.
+ */
+static void test_csum_no_carry_inputs(struct kunit *test)
+{
+	int len, align;
+	__wsum result, expec, sum;
+
+	assert_setup_correct(test);
+	memset(tmp_buf, 0x4, TEST_BUFLEN);
+	for (align = 0; align < TEST_BUFLEN; ++align) {
+		for (len = 0; len < MAX_LEN && (align + len) < TEST_BUFLEN;
+		     ++len) {
+			/*
+			 * Expect no carries.
+			 */
+			sum = init_sums_no_overflow[len];
+			result = full_csum(&tmp_buf[align], len, sum);
+			expec = 0;
+			CHECK_EQ(result, expec);
+
+			/*
+			 * Expect one carry.
+			 */
+			sum = init_sums_no_overflow[len] + 1;
+			result = full_csum(&tmp_buf[align], len, sum);
+			expec = len ? 0xfffe : 0xffff;
+			CHECK_EQ(result, expec);
+		}
+	}
+}
+
+static struct kunit_case __refdata checksum_test_cases[] = {
+	KUNIT_CASE(test_csum_fixed_random_inputs),
+	KUNIT_CASE(test_csum_all_carry_inputs),
+	KUNIT_CASE(test_csum_no_carry_inputs),
+	{}
+};
+
+static struct kunit_suite checksum_test_suite = {
+	.name = "checksum",
+	.test_cases = checksum_test_cases,
+};
+
+kunit_test_suites(&checksum_test_suite);
+
+MODULE_AUTHOR("Noah Goldstein <goldstein.w.n@gmail.com>");
+MODULE_LICENSE("GPL");
-- 
2.40.1.445.gf85cd430b1

Places

File _service:download_files:0001-fixes-miscellaneous.patch of Package linux-px (Revision 85bccb547dd01b546a3d5130f3df83bb)

Places