[PATCH printk v7 00/35] wire up write_atomic() printing

[PATCH printk v7 00/35] wire up write_atomic() printing

[John Ogness]

Hi,

This is v7 of a series to wire up the nbcon consoles so that
they actually perform printing using their write_atomic()
callback. v6 is here [0]. For information about the motivation
of the atomic consoles, please read the cover letter of v1 [1].

The main focus of this series:

- For nbcon consoles, always call write_atomic() directly from
  printk() caller context for the panic CPU.

- For nbcon consoles, call write_atomic() when unlocking the
  console lock.

- Only perform the console lock/unlock dance if legacy or boot
  consoles are registered.

- For legacy consoles, if nbcon consoles are registered, do not
  attempt to print from printk() caller context for the panic
  CPU until nbcon consoles have had a chance to print the most
  significant messages.

- Mark emergency sections. In these sections, every printk()
  call will attempt to directly flush to the consoles using the
  EMERGENCY priority.

This series does _not_ include threaded printing or nbcon
drivers. Those features will be added in separate follow-up
series.

The changes since v6:

- When a CPU is in an emergency state, printk() calls will now
  flush directly to all consoles using the EMERGENCY priority.

- Remove usage and implementation of
  nbcon_cpu_emergency_flush().

- Add printk_get_console_flush_type() to query available flush
  types and use it at all flushing call sites.

- In vprintk_emit(), always use nbcon_atomic flushing if it is
  available.

- When exiting nbcon_device_release(), flush using the legacy
  loop if nbcon_atomic flushing is not available.

- In console_flush_on_panic(), make sure nbcon_atomic flushing
  is allowed before flushing.

- In pr_flush(), always take the console_lock because there is
  no synchronization against register_console() and legacy
  consoles could register while waiting.

- In __wake_up_klogd(), remove check if PENDING_OUTPUT is
  needed. If PENDING_OUTPUT is set, it is needed.

- Rename is_printk_deferred() to is_printk_legacy_deferred().

- Remove nbcon_init(), consolidate it into nbcon_alloc().

- Improve documentation about nbcon owner/waiter matching
  rules.

- Allow nbcon_get_cpu_emergency_nesting() to run without
  migration disabled.

- Clarify that nbcon_get_default_prio() can run without
  migration disabled.

John Ogness

[0] https://lore.kernel.org/lkml/20240527063749.391035-1-john.ogness@linutronix.de
[1] https://lore.kernel.org/lkml/20230302195618.156940-1-john.ogness@linutronix.de

John Ogness (30):
  printk: Add notation to console_srcu locking
  printk: nbcon: Consolidate alloc() and init()
  printk: nbcon: Clarify rules of the owner/waiter matching
  printk: nbcon: Remove return value for write_atomic()
  printk: nbcon: Add detailed doc for write_atomic()
  printk: nbcon: Add callbacks to synchronize with driver
  printk: nbcon: Use driver synchronization while (un)registering
  serial: core: Provide low-level functions to lock port
  serial: core: Introduce wrapper to set @uart_port->cons
  console: Improve console_srcu_read_flags() comments
  nbcon: Add API to acquire context for non-printing operations
  serial: core: Acquire nbcon context in port->lock wrapper
  printk: nbcon: Do not rely on proxy headers
  printk: Make console_is_usable() available to nbcon.c
  printk: Let console_is_usable() handle nbcon
  printk: Add @flags argument for console_is_usable()
  printk: nbcon: Add helper to assign priority based on CPU state
  printk: Track registered boot consoles
  printk: nbcon: Use nbcon consoles in console_flush_all()
  printk: Add is_printk_legacy_deferred()
  printk: nbcon: Flush new records on device_release()
  printk: Flush nbcon consoles first on panic
  printk: nbcon: Add unsafe flushing on panic
  printk: Avoid console_lock dance if no legacy or boot consoles
  printk: Track nbcon consoles
  printk: Coordinate direct printing in panic
  printk: Add helper for flush type logic
  panic: Mark emergency section in oops
  rcu: Mark emergency sections in rcu stalls
  lockdep: Mark emergency sections in lockdep splats

Petr Mladek (1):
  printk: Properly deal with nbcon consoles on seq init

Sebastian Andrzej Siewior (1):
  printk: Check printk_deferred_enter()/_exit() usage

Thomas Gleixner (3):
  printk: nbcon: Provide function to flush using write_atomic()
  printk: nbcon: Implement emergency sections
  panic: Mark emergency section in warn

 drivers/tty/serial/8250/8250_core.c |   6 +-
 drivers/tty/serial/amba-pl011.c     |   2 +-
 drivers/tty/serial/serial_core.c    |  16 +-
 include/linux/console.h             | 110 +++++-
 include/linux/printk.h              |  33 +-
 include/linux/serial_core.h         | 117 +++++-
 kernel/locking/lockdep.c            |  83 +++-
 kernel/panic.c                      |   9 +
 kernel/printk/internal.h            | 130 ++++++-
 kernel/printk/nbcon.c               | 565 +++++++++++++++++++++++++---
 kernel/printk/printk.c              | 250 +++++++++---
 kernel/printk/printk_ringbuffer.h   |   2 +
 kernel/printk/printk_safe.c         |  23 +-
 kernel/rcu/tree_exp.h               |   7 +
 kernel/rcu/tree_stall.h             |   9 +
 15 files changed, 1211 insertions(+), 151 deletions(-)


base-commit: 8bf100092d60bf586bbc1a3a2cd833bb212d9d53
-- 
2.39.2

[PATCH printk v7 01/35] printk: Add notation to console_srcu locking

[John Ogness]

kernel/printk/printk.c:284:5: sparse: sparse: context imbalance in
'console_srcu_read_lock' - wrong count at exit
include/linux/srcu.h:301:9: sparse: sparse: context imbalance in
'console_srcu_read_unlock' - unexpected unlock

Fixes: 6c4afa79147e ("printk: Prepare for SRCU console list protection")
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/printk.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 054c0e7784fd..b3a4c0363a96 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -282,6 +282,7 @@ EXPORT_SYMBOL(console_list_unlock);
  * Return: A cookie to pass to console_srcu_read_unlock().
  */
 int console_srcu_read_lock(void)
+	__acquires(&console_srcu)
 {
 	return srcu_read_lock_nmisafe(&console_srcu);
 }
@@ -295,6 +296,7 @@ EXPORT_SYMBOL(console_srcu_read_lock);
  * Counterpart to console_srcu_read_lock()
  */
 void console_srcu_read_unlock(int cookie)
+	__releases(&console_srcu)
 {
 	srcu_read_unlock_nmisafe(&console_srcu, cookie);
 }
-- 
2.39.2

[PATCH printk v7 02/35] printk: nbcon: Consolidate alloc() and init()

[John Ogness]

Rather than splitting the nbcon allocation and initialization into
two pieces, perform all initialization in nbcon_alloc(). Later,
the initial sequence is calculated and can be explicitly set using
nbcon_seq_force(). This removes the need for the strong rules of
nbcon_init() that even included a BUG_ON().

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 kernel/printk/internal.h |  2 --
 kernel/printk/nbcon.c    | 37 +++++++++++--------------------------
 kernel/printk/printk.c   |  2 +-
 3 files changed, 12 insertions(+), 29 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 19dcc5832651..398ecb40d279 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -75,7 +75,6 @@ u16 printk_parse_prefix(const char *text, int *level,
 u64 nbcon_seq_read(struct console *con);
 void nbcon_seq_force(struct console *con, u64 seq);
 bool nbcon_alloc(struct console *con);
-void nbcon_init(struct console *con);
 void nbcon_free(struct console *con);
 
 #else
@@ -96,7 +95,6 @@ static inline bool printk_percpu_data_ready(void) { return false; }
 static inline u64 nbcon_seq_read(struct console *con) { return 0; }
 static inline void nbcon_seq_force(struct console *con, u64 seq) { }
 static inline bool nbcon_alloc(struct console *con) { return false; }
-static inline void nbcon_init(struct console *con) { }
 static inline void nbcon_free(struct console *con) { }
 
 #endif /* CONFIG_PRINTK */
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index c8093bcc01fe..670692dc9b10 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -929,17 +929,22 @@ static bool nbcon_emit_next_record(struct nbcon_write_context *wctxt)
 }
 
 /**
- * nbcon_alloc - Allocate buffers needed by the nbcon console
- * @con:	Console to allocate buffers for
+ * nbcon_alloc - Allocate and init the nbcon console specific data
+ * @con:	Console to initialize
  *
- * Return:	True on success. False otherwise and the console cannot
- *		be used.
+ * Return:	True if the console was fully allocated and initialized.
+ *		Otherwise @con must not be registered.
  *
- * This is not part of nbcon_init() because buffer allocation must
- * be performed earlier in the console registration process.
+ * When allocation and init was successful, the console must be properly
+ * freed using nbcon_free() once it is no longer needed.
  */
 bool nbcon_alloc(struct console *con)
 {
+	struct nbcon_state state = { };
+
+	nbcon_state_set(con, &state);
+	atomic_long_set(&ACCESS_PRIVATE(con, nbcon_seq), 0);
+
 	if (con->flags & CON_BOOT) {
 		/*
 		 * Boot console printing is synchronized with legacy console
@@ -958,26 +963,6 @@ bool nbcon_alloc(struct console *con)
 	return true;
 }
 
-/**
- * nbcon_init - Initialize the nbcon console specific data
- * @con:	Console to initialize
- *
- * nbcon_alloc() *must* be called and succeed before this function
- * is called.
- *
- * This function expects that the legacy @con->seq has been set.
- */
-void nbcon_init(struct console *con)
-{
-	struct nbcon_state state = { };
-
-	/* nbcon_alloc() must have been called and successful! */
-	BUG_ON(!con->pbufs);
-
-	nbcon_seq_force(con, con->seq);
-	nbcon_state_set(con, &state);
-}
-
 /**
  * nbcon_free - Free and cleanup the nbcon console specific data
  * @con:	Console to free/cleanup nbcon data
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index b3a4c0363a96..15c61c9ea87e 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3618,7 +3618,7 @@ void register_console(struct console *newcon)
 	console_init_seq(newcon, bootcon_registered);
 
 	if (newcon->flags & CON_NBCON)
-		nbcon_init(newcon);
+		nbcon_seq_force(newcon, newcon->seq);
 
 	/*
 	 * Put this console in the list - keep the
-- 
2.39.2

[PATCH printk v7 03/35] printk: Properly deal with nbcon consoles on seq init

[John Ogness]

From: Petr Mladek <pmladek@suse.com>

If a non-boot console is registering and boot consoles exist,
the consoles are flushed before being unregistered. This allows
the non-boot console to continue where the boot console left
off.

If for whatever reason flushing fails, the lowest seq found from
any of the enabled boot consoles is used. Until now con->seq was
checked. However, if it is an nbcon boot console, the function
nbcon_seq_read() must be used to read seq because con->seq is
not updated for nbcon consoles.

Check if it is an nbcon boot console and if so call
nbcon_seq_read() to read seq.

Also, avoid usage of con->seq as temporary storage of the
starting record. Instead, rename console_init_seq() to
get_init_console_seq() and just return the value. For nbcon
consoles set the sequence via nbcon_seq_force(), for legacy
consoles set con->seq.

The cleaned design should make sure that the value stays and is
set before the console is added to the console list. It also
unifies the sequence number initialization for legacy and nbcon
consoles.

Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: John Ogness <john.ogness@linutronix.de>
---
 kernel/printk/nbcon.c  |  3 ---
 kernel/printk/printk.c | 41 +++++++++++++++++++++++++++++------------
 2 files changed, 29 insertions(+), 15 deletions(-)

diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 670692dc9b10..776746d20fc0 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -172,9 +172,6 @@ void nbcon_seq_force(struct console *con, u64 seq)
 	u64 valid_seq = max_t(u64, seq, prb_first_valid_seq(prb));
 
 	atomic_long_set(&ACCESS_PRIVATE(con, nbcon_seq), __u64seq_to_ulseq(valid_seq));
-
-	/* Clear con->seq since nbcon consoles use con->nbcon_seq instead. */
-	con->seq = 0;
 }
 
 /**
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 15c61c9ea87e..6ac333938ffc 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3448,19 +3448,21 @@ static void try_enable_default_console(struct console *newcon)
 		newcon->flags |= CON_CONSDEV;
 }
 
-static void console_init_seq(struct console *newcon, bool bootcon_registered)
+/* Return the starting sequence number for a newly registered console. */
+static u64 get_init_console_seq(struct console *newcon, bool bootcon_registered)
 {
 	struct console *con;
 	bool handover;
+	u64 init_seq;
 
 	if (newcon->flags & (CON_PRINTBUFFER | CON_BOOT)) {
 		/* Get a consistent copy of @syslog_seq. */
 		mutex_lock(&syslog_lock);
-		newcon->seq = syslog_seq;
+		init_seq = syslog_seq;
 		mutex_unlock(&syslog_lock);
 	} else {
 		/* Begin with next message added to ringbuffer. */
-		newcon->seq = prb_next_seq(prb);
+		init_seq = prb_next_seq(prb);
 
 		/*
 		 * If any enabled boot consoles are due to be unregistered
@@ -3481,7 +3483,7 @@ static void console_init_seq(struct console *newcon, bool bootcon_registered)
 			 * Flush all consoles and set the console to start at
 			 * the next unprinted sequence number.
 			 */
-			if (!console_flush_all(true, &newcon->seq, &handover)) {
+			if (!console_flush_all(true, &init_seq, &handover)) {
 				/*
 				 * Flushing failed. Just choose the lowest
 				 * sequence of the enabled boot consoles.
@@ -3494,19 +3496,30 @@ static void console_init_seq(struct console *newcon, bool bootcon_registered)
 				if (handover)
 					console_lock();
 
-				newcon->seq = prb_next_seq(prb);
+				init_seq = prb_next_seq(prb);
 				for_each_console(con) {
-					if ((con->flags & CON_BOOT) &&
-					    (con->flags & CON_ENABLED) &&
-					    con->seq < newcon->seq) {
-						newcon->seq = con->seq;
+					u64 seq;
+
+					if (!(con->flags & CON_BOOT) ||
+					    !(con->flags & CON_ENABLED)) {
+						continue;
 					}
+
+					if (con->flags & CON_NBCON)
+						seq = nbcon_seq_read(con);
+					else
+						seq = con->seq;
+
+					if (seq < init_seq)
+						init_seq = seq;
 				}
 			}
 
 			console_unlock();
 		}
 	}
+
+	return init_seq;
 }
 
 #define console_first()				\
@@ -3538,6 +3551,7 @@ void register_console(struct console *newcon)
 	struct console *con;
 	bool bootcon_registered = false;
 	bool realcon_registered = false;
+	u64 init_seq;
 	int err;
 
 	console_list_lock();
@@ -3615,10 +3629,13 @@ void register_console(struct console *newcon)
 	}
 
 	newcon->dropped = 0;
-	console_init_seq(newcon, bootcon_registered);
+	init_seq = get_init_console_seq(newcon, bootcon_registered);
 
-	if (newcon->flags & CON_NBCON)
-		nbcon_seq_force(newcon, newcon->seq);
+	if (newcon->flags & CON_NBCON) {
+		nbcon_seq_force(newcon, init_seq);
+	} else {
+		newcon->seq = init_seq;
+	}
 
 	/*
 	 * Put this console in the list - keep the
-- 
2.39.2

[PATCH printk v7 04/35] printk: Check printk_deferred_enter()/_exit() usage

[John Ogness]

From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

Add validation that printk_deferred_enter()/_exit() are called in
non-migration contexts.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/printk.h      |  9 +++++----
 kernel/printk/internal.h    |  3 +++
 kernel/printk/printk_safe.c | 12 ++++++++++++
 3 files changed, 20 insertions(+), 4 deletions(-)

diff --git a/include/linux/printk.h b/include/linux/printk.h
index b937cefcb31c..eee8e97da681 100644
--- a/include/linux/printk.h
+++ b/include/linux/printk.h
@@ -161,15 +161,16 @@ int _printk(const char *fmt, ...);
  */
 __printf(1, 2) __cold int _printk_deferred(const char *fmt, ...);
 
-extern void __printk_safe_enter(void);
-extern void __printk_safe_exit(void);
+extern void __printk_deferred_enter(void);
+extern void __printk_deferred_exit(void);
+
 /*
  * The printk_deferred_enter/exit macros are available only as a hack for
  * some code paths that need to defer all printk console printing. Interrupts
  * must be disabled for the deferred duration.
  */
-#define printk_deferred_enter __printk_safe_enter
-#define printk_deferred_exit __printk_safe_exit
+#define printk_deferred_enter() __printk_deferred_enter()
+#define printk_deferred_exit() __printk_deferred_exit()
 
 /*
  * Please don't use printk_ratelimit(), because it shares ratelimiting state
diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 398ecb40d279..dc8bc0890fd2 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -53,6 +53,9 @@ int vprintk_store(int facility, int level,
 __printf(1, 0) int vprintk_default(const char *fmt, va_list args);
 __printf(1, 0) int vprintk_deferred(const char *fmt, va_list args);
 
+void __printk_safe_enter(void);
+void __printk_safe_exit(void);
+
 bool printk_percpu_data_ready(void);
 
 #define printk_safe_enter_irqsave(flags)	\
diff --git a/kernel/printk/printk_safe.c b/kernel/printk/printk_safe.c
index 6d10927a07d8..4421ccac3113 100644
--- a/kernel/printk/printk_safe.c
+++ b/kernel/printk/printk_safe.c
@@ -26,6 +26,18 @@ void __printk_safe_exit(void)
 	this_cpu_dec(printk_context);
 }
 
+void __printk_deferred_enter(void)
+{
+	cant_migrate();
+	__printk_safe_enter();
+}
+
+void __printk_deferred_exit(void)
+{
+	cant_migrate();
+	__printk_safe_exit();
+}
+
 asmlinkage int vprintk(const char *fmt, va_list args)
 {
 #ifdef CONFIG_KGDB_KDB
-- 
2.39.2

[PATCH printk v7 05/35] printk: nbcon: Clarify rules of the owner/waiter matching

[John Ogness]

The functions nbcon_owner_matches() and nbcon_waiter_matches()
use a minimal set of data to determine if a context matches.
The existing kerneldoc and comments were not clear enough and
caused the printk folks to re-prove that the functions are
indeed reliable in all cases.

Update and expand the explanations so that it is clear that the
implementations are sufficient for all cases.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/nbcon.c | 56 +++++++++++++++++++++++++++++++++++--------
 1 file changed, 46 insertions(+), 10 deletions(-)

diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 776746d20fc0..931b8f086902 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -228,6 +228,13 @@ static int nbcon_context_try_acquire_direct(struct nbcon_context *ctxt,
 	struct nbcon_state new;
 
 	do {
+		/*
+		 * Panic does not imply that the console is owned. However, it
+		 * is critical that non-panic CPUs during panic are unable to
+		 * acquire ownership in order to satisfy the assumptions of
+		 * nbcon_waiter_matches(). In particular, the assumption that
+		 * lower priorities are ignored during panic.
+		 */
 		if (other_cpu_in_panic())
 			return -EPERM;
 
@@ -259,18 +266,29 @@ static bool nbcon_waiter_matches(struct nbcon_state *cur, int expected_prio)
 	/*
 	 * The request context is well defined by the @req_prio because:
 	 *
-	 * - Only a context with a higher priority can take over the request.
+	 * - Only a context with a priority higher than the owner can become
+	 *   a waiter.
+	 * - Only a context with a priority higher than the waiter can
+	 *   directly take over the request.
 	 * - There are only three priorities.
 	 * - Only one CPU is allowed to request PANIC priority.
 	 * - Lower priorities are ignored during panic() until reboot.
 	 *
 	 * As a result, the following scenario is *not* possible:
 	 *
-	 * 1. Another context with a higher priority directly takes ownership.
-	 * 2. The higher priority context releases the ownership.
-	 * 3. A lower priority context takes the ownership.
-	 * 4. Another context with the same priority as this context
+	 * 1. This context is currently a waiter.
+	 * 2. Another context with a higher priority than this context
+	 *    directly takes ownership.
+	 * 3. The higher priority context releases the ownership.
+	 * 4. Another lower priority context takes the ownership.
+	 * 5. Another context with the same priority as this context
 	 *    creates a request and starts waiting.
+	 *
+	 * Event #1 implies this context is EMERGENCY.
+	 * Event #2 implies the new context is PANIC.
+	 * Event #3 occurs when panic() has flushed the console.
+	 * Events #4 and #5 are not possible due to the other_cpu_in_panic()
+	 * check in nbcon_context_try_acquire_direct().
 	 */
 
 	return (cur->req_prio == expected_prio);
@@ -578,11 +596,29 @@ static bool nbcon_owner_matches(struct nbcon_state *cur, int expected_cpu,
 				int expected_prio)
 {
 	/*
-	 * Since consoles can only be acquired by higher priorities,
-	 * owning contexts are uniquely identified by @prio. However,
-	 * since contexts can unexpectedly lose ownership, it is
-	 * possible that later another owner appears with the same
-	 * priority. For this reason @cpu is also needed.
+	 * A similar function, nbcon_waiter_matches(), only deals with
+	 * EMERGENCY and PANIC priorities. However, this function must also
+	 * deal with the NORMAL priority, which requires additional checks
+	 * and constraints.
+	 *
+	 * For the case where preemption and interrupts are disabled, it is
+	 * enough to also verify that the owning CPU has not changed.
+	 *
+	 * For the case where preemption or interrupts are enabled, an
+	 * external synchronization method *must* be used. In particular,
+	 * the driver-specific locking mechanism used in device_lock()
+	 * (including disabling migration) should be used. It prevents
+	 * scenarios such as:
+	 *
+	 * 1. [Task A] owns a context with NBCON_PRIO_NORMAL on [CPU X] and
+	 *    is scheduled out.
+	 * 2. Another context takes over the lock with NBCON_PRIO_EMERGENCY
+	 *    and releases it.
+	 * 3. [Task B] acquires a context with NBCON_PRIO_NORMAL on [CPU X]
+	 *    and is scheduled out.
+	 * 4. [Task A] gets running on [CPU X] and sees that the console is
+	 *    still owned by a task on [CPU X] with NBON_PRIO_NORMAL. Thus
+	 *    [Task A] thinks it is the owner when it is not.
 	 */
 
 	if (cur->prio != expected_prio)
-- 
2.39.2

[PATCH printk v7 06/35] printk: nbcon: Remove return value for write_atomic()

[John Ogness]

The return value of write_atomic() does not provide any useful
information. On the contrary, it makes things more complicated
for the caller to appropriately deal with the information.

Change write_atomic() to not have a return value. If the
message did not get printed due to loss of ownership, the
caller will notice this on its own. If ownership was not lost,
it will be assumed that the driver successfully printed the
message and the sequence number for that console will be
incremented.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/console.h |  2 +-
 kernel/printk/nbcon.c   | 15 +++++++--------
 2 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index 31a8f5b85f5d..577b157fe4fb 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -345,7 +345,7 @@ struct console {
 	struct hlist_node	node;
 
 	/* nbcon console specific members */
-	bool			(*write_atomic)(struct console *con,
+	void			(*write_atomic)(struct console *con,
 						struct nbcon_write_context *wctxt);
 	atomic_t		__private nbcon_state;
 	atomic_long_t		__private nbcon_seq;
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 931b8f086902..f279f839741a 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -885,7 +885,6 @@ static bool nbcon_emit_next_record(struct nbcon_write_context *wctxt)
 	unsigned long con_dropped;
 	struct nbcon_state cur;
 	unsigned long dropped;
-	bool done;
 
 	/*
 	 * The printk buffers are filled within an unsafe section. This
@@ -925,16 +924,16 @@ static bool nbcon_emit_next_record(struct nbcon_write_context *wctxt)
 	wctxt->unsafe_takeover = cur.unsafe_takeover;
 
 	if (con->write_atomic) {
-		done = con->write_atomic(con, wctxt);
+		con->write_atomic(con, wctxt);
 	} else {
-		nbcon_context_release(ctxt);
+		/*
+		 * This function should never be called for legacy consoles.
+		 * Handle it as if ownership was lost and try to continue.
+		 */
 		WARN_ON_ONCE(1);
-		done = false;
-	}
-
-	/* If not done, the emit was aborted. */
-	if (!done)
+		nbcon_context_release(ctxt);
 		return false;
+	}
 
 	/*
 	 * Since any dropped message was successfully output, reset the
-- 
2.39.2

[PATCH printk v7 07/35] printk: nbcon: Add detailed doc for write_atomic()

[John Ogness]

The write_atomic() callback has special requirements and is
allowed to use special helper functions. Provide detailed
documentation of the callback so that a developer has a
chance of implementing it correctly.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/console.h | 33 +++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index 577b157fe4fb..35c64ee3827b 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -303,7 +303,7 @@ struct nbcon_write_context {
 /**
  * struct console - The console descriptor structure
  * @name:		The name of the console driver
- * @write:		Write callback to output messages (Optional)
+ * @write:		Legacy write callback to output messages (Optional)
  * @read:		Read callback for console input (Optional)
  * @device:		The underlying TTY device driver (Optional)
  * @unblank:		Callback to unblank the console (Optional)
@@ -320,7 +320,6 @@ struct nbcon_write_context {
  * @data:		Driver private data
  * @node:		hlist node for the console list
  *
- * @write_atomic:	Write callback for atomic context
  * @nbcon_state:	State for nbcon consoles
  * @nbcon_seq:		Sequence number of the next record for nbcon to print
  * @pbufs:		Pointer to nbcon private buffer
@@ -345,8 +344,34 @@ struct console {
 	struct hlist_node	node;
 
 	/* nbcon console specific members */
-	void			(*write_atomic)(struct console *con,
-						struct nbcon_write_context *wctxt);
+
+	/**
+	 * @write_atomic:
+	 *
+	 * NBCON callback to write out text in any context. (Optional)
+	 *
+	 * This callback is called with the console already acquired. However,
+	 * a higher priority context is allowed to take it over by default.
+	 *
+	 * The callback must call nbcon_enter_unsafe() and nbcon_exit_unsafe()
+	 * around any code where the takeover is not safe, for example, when
+	 * manipulating the serial port registers.
+	 *
+	 * nbcon_enter_unsafe() will fail if the context has lost the console
+	 * ownership in the meantime. In this case, the callback is no longer
+	 * allowed to go forward. It must back out immediately and carefully.
+	 * The buffer content is also no longer trusted since it no longer
+	 * belongs to the context.
+	 *
+	 * The callback should allow the takeover whenever it is safe. It
+	 * increases the chance to see messages when the system is in trouble.
+	 *
+	 * The callback can be called from any context (including NMI).
+	 * Therefore it must avoid usage of any locking and instead rely
+	 * on the console ownership for synchronization.
+	 */
+	void (*write_atomic)(struct console *con, struct nbcon_write_context *wctxt);
+
 	atomic_t		__private nbcon_state;
 	atomic_long_t		__private nbcon_seq;
 	struct printk_buffers	*pbufs;
-- 
2.39.2

[PATCH printk v7 08/35] printk: nbcon: Add callbacks to synchronize with driver

[John Ogness]

Console drivers typically must deal with access to the hardware
via user input/output (such as an interactive login shell) and
output of kernel messages via printk() calls. To provide the
necessary synchronization, usually some driver-specific locking
mechanism is used (for example, the port spinlock for uart
serial consoles).

Until now, usage of this driver-specific locking has been hidden
from the printk-subsystem and implemented within the various
console callbacks. However, nbcon consoles would need to use it
even in the generic code.

Add device_lock() and device_unlock() callback which will need
to get implemented by nbcon consoles.

The callbacks will use whatever synchronization mechanism the
driver is using for itself. The minimum requirement is to
prevent CPU migration. It would allow a context friendly
acquiring of nbcon console ownership in non-emergency and
non-panic context.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/console.h | 43 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)

diff --git a/include/linux/console.h b/include/linux/console.h
index 35c64ee3827b..46b3c210b931 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -372,6 +372,49 @@ struct console {
 	 */
 	void (*write_atomic)(struct console *con, struct nbcon_write_context *wctxt);
 
+	/**
+	 * @device_lock:
+	 *
+	 * NBCON callback to begin synchronization with driver code.
+	 *
+	 * Console drivers typically must deal with access to the hardware
+	 * via user input/output (such as an interactive login shell) and
+	 * output of kernel messages via printk() calls. This callback is
+	 * called by the printk-subsystem whenever it needs to synchronize
+	 * with hardware access by the driver. It should be implemented to
+	 * use whatever synchronization mechanism the driver is using for
+	 * itself (for example, the port lock for uart serial consoles).
+	 *
+	 * The callback is always called from task context. It may use any
+	 * synchronization method required by the driver.
+	 *
+	 * IMPORTANT: The callback MUST disable migration. The console driver
+	 *	may be using a synchronization mechanism that already takes
+	 *	care of this (such as spinlocks). Otherwise this function must
+	 *	explicitly call migrate_disable().
+	 *
+	 * The flags argument is provided as a convenience to the driver. It
+	 * will be passed again to device_unlock(). It can be ignored if the
+	 * driver does not need it.
+	 */
+	void (*device_lock)(struct console *con, unsigned long *flags);
+
+	/**
+	 * @device_unlock:
+	 *
+	 * NBCON callback to finish synchronization with driver code.
+	 *
+	 * It is the counterpart to device_lock().
+	 *
+	 * This callback is always called from task context. It must
+	 * appropriately re-enable migration (depending on how device_lock()
+	 * disabled migration).
+	 *
+	 * The flags argument is the value of the same variable that was
+	 * passed to device_lock().
+	 */
+	void (*device_unlock)(struct console *con, unsigned long flags);
+
 	atomic_t		__private nbcon_state;
 	atomic_long_t		__private nbcon_seq;
 	struct printk_buffers	*pbufs;
-- 
2.39.2

[PATCH printk v7 09/35] printk: nbcon: Use driver synchronization while (un)registering

[John Ogness]

Console drivers typically have to deal with access to the
hardware via user input/output (such as an interactive login
shell) and output of kernel messages via printk() calls.

They use some classic driver-specific locking mechanism in most
situations. But console->write_atomic() callbacks, used by nbcon
consoles, are synchronized only by acquiring the console
context.

The synchronization via the console context ownership is possible
only when the console driver is registered. It is when a
particular device driver is connected with a particular console
driver.

The two synchronization mechanisms must be synchronized between
each other. It is tricky because the console context ownership
is quite special. It might be taken over by a higher priority
context. Also CPU migration must be disabled. The most tricky
part is to (dis)connect these two mechanisms during the console
(un)registration.

Use the driver-specific locking callbacks: device_lock(),
device_unlock(). They allow taking the device-specific lock
while the device is being (un)registered by the related console
driver.

For example, these callbacks lock/unlock the port lock for
serial port drivers.

Note that the driver-specific locking is only needed during
(un)register if it is an nbcon console with the write_atomic()
callback implemented. If write_atomic() is not implemented, the
driver should never attempt to access the hardware without
first acquiring its driver-specific lock.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/printk.c | 33 ++++++++++++++++++++++++++++++++-
 1 file changed, 32 insertions(+), 1 deletion(-)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 6ac333938ffc..7027f0868ef4 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3548,9 +3548,11 @@ static int unregister_console_locked(struct console *console);
  */
 void register_console(struct console *newcon)
 {
-	struct console *con;
+	bool use_device_lock = (newcon->flags & CON_NBCON) && newcon->write_atomic;
 	bool bootcon_registered = false;
 	bool realcon_registered = false;
+	struct console *con;
+	unsigned long flags;
 	u64 init_seq;
 	int err;
 
@@ -3637,6 +3639,19 @@ void register_console(struct console *newcon)
 		newcon->seq = init_seq;
 	}
 
+	/*
+	 * If another context is actively using the hardware of this new
+	 * console, it will not be aware of the nbcon synchronization. This
+	 * is a risk that two contexts could access the hardware
+	 * simultaneously if this new console is used for atomic printing
+	 * and the other context is still using the hardware.
+	 *
+	 * Use the driver synchronization to ensure that the hardware is not
+	 * in use while this new console transitions to being registered.
+	 */
+	if (use_device_lock)
+		newcon->device_lock(newcon, &flags);
+
 	/*
 	 * Put this console in the list - keep the
 	 * preferred driver at the head of the list.
@@ -3661,6 +3676,10 @@ void register_console(struct console *newcon)
 	 * register_console() completes.
 	 */
 
+	/* This new console is now registered. */
+	if (use_device_lock)
+		newcon->device_unlock(newcon, flags);
+
 	console_sysfs_notify();
 
 	/*
@@ -3689,6 +3708,8 @@ EXPORT_SYMBOL(register_console);
 /* Must be called under console_list_lock(). */
 static int unregister_console_locked(struct console *console)
 {
+	bool use_device_lock = (console->flags & CON_NBCON) && console->write_atomic;
+	unsigned long flags;
 	int res;
 
 	lockdep_assert_console_list_lock_held();
@@ -3707,8 +3728,18 @@ static int unregister_console_locked(struct console *console)
 	if (!console_is_registered_locked(console))
 		return -ENODEV;
 
+	/*
+	 * Use the driver synchronization to ensure that the hardware is not
+	 * in use while this console transitions to being unregistered.
+	 */
+	if (use_device_lock)
+		console->device_lock(console, &flags);
+
 	hlist_del_init_rcu(&console->node);
 
+	if (use_device_lock)
+		console->device_unlock(console, flags);
+
 	/*
 	 * <HISTORICAL>
 	 * If this isn't the last console and it has CON_CONSDEV set, we
-- 
2.39.2

[PATCH printk v7 10/35] serial: core: Provide low-level functions to lock port

[John Ogness]

It will be necessary at times for the uart nbcon console
drivers to acquire the port lock directly (without the
additional nbcon functionality of the port lock wrappers).
These are special cases such as the implementation of the
device_lock()/device_unlock() callbacks or for internal
port lock wrapper synchronization.

Provide low-level variants __uart_port_lock_irqsave() and
__uart_port_unlock_irqrestore() for this purpose.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 include/linux/serial_core.h | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/include/linux/serial_core.h b/include/linux/serial_core.h
index aea25eef9a1a..8872cd21e70a 100644
--- a/include/linux/serial_core.h
+++ b/include/linux/serial_core.h
@@ -590,6 +590,24 @@ struct uart_port {
 	void			*private_data;		/* generic platform data pointer */
 };
 
+/*
+ * Only for console->device_lock()/_unlock() callbacks and internal
+ * port lock wrapper synchronization.
+ */
+static inline void __uart_port_lock_irqsave(struct uart_port *up, unsigned long *flags)
+{
+	spin_lock_irqsave(&up->lock, *flags);
+}
+
+/*
+ * Only for console->device_lock()/_unlock() callbacks and internal
+ * port lock wrapper synchronization.
+ */
+static inline void __uart_port_unlock_irqrestore(struct uart_port *up, unsigned long flags)
+{
+	spin_unlock_irqrestore(&up->lock, flags);
+}
+
 /**
  * uart_port_lock - Lock the UART port
  * @up:		Pointer to UART port structure
-- 
2.39.2

[PATCH printk v7 11/35] serial: core: Introduce wrapper to set @uart_port->cons

[John Ogness]

Introduce uart_port_set_cons() as a wrapper to set @cons of a
uart_port. The wrapper sets @cons under the port lock in order
to prevent @cons from disappearing while another context is
holding the port lock. This is necessary for a follow-up
commit relating to the port lock wrappers, which rely on @cons
not changing between lock and unlock.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Tested-by: Théo Lebrun <theo.lebrun@bootlin.com> # EyeQ5, AMBA-PL011
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 drivers/tty/serial/8250/8250_core.c |  6 +++---
 drivers/tty/serial/amba-pl011.c     |  2 +-
 drivers/tty/serial/serial_core.c    | 16 ++++++++--------
 include/linux/serial_core.h         | 17 +++++++++++++++++
 4 files changed, 29 insertions(+), 12 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_core.c b/drivers/tty/serial/8250/8250_core.c
index 29e4b83e0376..5f9f06911795 100644
--- a/drivers/tty/serial/8250/8250_core.c
+++ b/drivers/tty/serial/8250/8250_core.c
@@ -423,11 +423,11 @@ static int univ8250_console_setup(struct console *co, char *options)
 
 	port = &serial8250_ports[co->index].port;
 	/* link port to console */
-	port->cons = co;
+	uart_port_set_cons(port, co);
 
 	retval = serial8250_console_setup(port, options, false);
 	if (retval != 0)
-		port->cons = NULL;
+		uart_port_set_cons(port, NULL);
 	return retval;
 }
 
@@ -485,7 +485,7 @@ static int univ8250_console_match(struct console *co, char *name, int idx,
 			continue;
 
 		co->index = i;
-		port->cons = co;
+		uart_port_set_cons(port, co);
 		return serial8250_console_setup(port, options, true);
 	}
 
diff --git a/drivers/tty/serial/amba-pl011.c b/drivers/tty/serial/amba-pl011.c
index 8b1644f5411e..7d0134ecd82f 100644
--- a/drivers/tty/serial/amba-pl011.c
+++ b/drivers/tty/serial/amba-pl011.c
@@ -2480,7 +2480,7 @@ static int pl011_console_match(struct console *co, char *name, int idx,
 			continue;
 
 		co->index = i;
-		port->cons = co;
+		uart_port_set_cons(port, co);
 		return pl011_console_setup(co, options);
 	}
 
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 9a18d0b95a41..61c7e1268957 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -3168,8 +3168,15 @@ static int serial_core_add_one_port(struct uart_driver *drv, struct uart_port *u
 	state->uart_port = uport;
 	uport->state = state;
 
+	/*
+	 * If this port is in use as a console then the spinlock is already
+	 * initialised.
+	 */
+	if (!uart_console_registered(uport))
+		uart_port_spin_lock_init(uport);
+
 	state->pm_state = UART_PM_STATE_UNDEFINED;
-	uport->cons = drv->cons;
+	uart_port_set_cons(uport, drv->cons);
 	uport->minor = drv->tty_driver->minor_start + uport->line;
 	uport->name = kasprintf(GFP_KERNEL, "%s%d", drv->dev_name,
 				drv->tty_driver->name_base + uport->line);
@@ -3178,13 +3185,6 @@ static int serial_core_add_one_port(struct uart_driver *drv, struct uart_port *u
 		goto out;
 	}
 
-	/*
-	 * If this port is in use as a console then the spinlock is already
-	 * initialised.
-	 */
-	if (!uart_console_registered(uport))
-		uart_port_spin_lock_init(uport);
-
 	if (uport->cons && uport->dev)
 		of_console_check(uport->dev->of_node, uport->cons->name, uport->line);
 
diff --git a/include/linux/serial_core.h b/include/linux/serial_core.h
index 8872cd21e70a..2cf03ff2056a 100644
--- a/include/linux/serial_core.h
+++ b/include/linux/serial_core.h
@@ -608,6 +608,23 @@ static inline void __uart_port_unlock_irqrestore(struct uart_port *up, unsigned
 	spin_unlock_irqrestore(&up->lock, flags);
 }
 
+/**
+ * uart_port_set_cons - Safely set the @cons field for a uart
+ * @up:		The uart port to set
+ * @con:	The new console to set to
+ *
+ * This function must be used to set @up->cons. It uses the port lock to
+ * synchronize with the port lock wrappers in order to ensure that the console
+ * cannot change or disappear while another context is holding the port lock.
+ */
+static inline void uart_port_set_cons(struct uart_port *up, struct console *con)
+{
+	unsigned long flags;
+
+	__uart_port_lock_irqsave(up, &flags);
+	up->cons = con;
+	__uart_port_unlock_irqrestore(up, flags);
+}
 /**
  * uart_port_lock - Lock the UART port
  * @up:		Pointer to UART port structure
-- 
2.39.2

[PATCH printk v7 12/35] console: Improve console_srcu_read_flags() comments

[John Ogness]

It was not clear when exactly console_srcu_read_flags() must be
used vs. directly reading @console->flags.

Refactor and clarify that console_srcu_read_flags() is only
needed if the console is registered or the caller is in a
context where the registration status of the console may change
(due to another context).

The function requires the caller holds @console_srcu, which will
ensure that the caller sees an appropriate @flags value for the
registered console and that exit/cleanup routines will not run
if the console is in the process of unregistration.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/console.h | 28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index 46b3c210b931..aafe3121b74e 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -446,28 +446,34 @@ extern void console_list_unlock(void) __releases(console_mutex);
 extern struct hlist_head console_list;
 
 /**
- * console_srcu_read_flags - Locklessly read the console flags
+ * console_srcu_read_flags - Locklessly read flags of a possibly registered
+ *				console
  * @con:	struct console pointer of console to read flags from
  *
- * This function provides the necessary READ_ONCE() and data_race()
- * notation for locklessly reading the console flags. The READ_ONCE()
- * in this function matches the WRITE_ONCE() when @flags are modified
- * for registered consoles with console_srcu_write_flags().
+ * Locklessly reading @con->flags provides a consistent read value because
+ * there is at most one CPU modifying @con->flags and that CPU is using only
+ * read-modify-write operations to do so.
  *
- * Only use this function to read console flags when locklessly
- * iterating the console list via srcu.
+ * Requires console_srcu_read_lock to be held, which implies that @con might
+ * be a registered console. The purpose of holding console_srcu_read_lock is
+ * to guarantee that the console state is valid (CON_SUSPENDED/CON_ENABLED)
+ * and that no exit/cleanup routines will run if the console is currently
+ * undergoing unregistration.
+ *
+ * If the caller is holding the console_list_lock or it is _certain_ that
+ * @con is not and will not become registered, the caller may read
+ * @con->flags directly instead.
  *
  * Context: Any context.
+ * Return: The current value of the @con->flags field.
  */
 static inline short console_srcu_read_flags(const struct console *con)
 {
 	WARN_ON_ONCE(!console_srcu_read_lock_is_held());
 
 	/*
-	 * Locklessly reading console->flags provides a consistent
-	 * read value because there is at most one CPU modifying
-	 * console->flags and that CPU is using only read-modify-write
-	 * operations to do so.
+	 * The READ_ONCE() matches the WRITE_ONCE() when @flags are modified
+	 * for registered consoles with console_srcu_write_flags().
 	 */
 	return data_race(READ_ONCE(con->flags));
 }
-- 
2.39.2

[PATCH printk v7 13/35] nbcon: Add API to acquire context for non-printing operations

[John Ogness]

Provide functions nbcon_device_try_acquire() and
nbcon_device_release() which will try to acquire the nbcon
console ownership with NBCON_PRIO_NORMAL and mark it unsafe for
handover/takeover.

These functions are to be used together with the device-specific
locking when performing non-printing activities on the console
device. They will allow synchronization against the
atomic_write() callback which will be serialized, for higher
priority contexts, only by acquiring the console context
ownership.

Pitfalls:

The API requires to be called in a context with migration
disabled because it uses per-CPU variables internally.

The context is set unsafe for a takeover all the time. It
guarantees full serialization against any atomic_write() caller
except for the final flush in panic() which might try an unsafe
takeover.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/console.h |  2 ++
 include/linux/printk.h  | 14 ++++++++++
 kernel/printk/nbcon.c   | 58 ++++++++++++++++++++++++++++++++++++++++-
 3 files changed, 73 insertions(+), 1 deletion(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index aafe3121b74e..3706f944de46 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -322,6 +322,7 @@ struct nbcon_write_context {
  *
  * @nbcon_state:	State for nbcon consoles
  * @nbcon_seq:		Sequence number of the next record for nbcon to print
+ * @nbcon_device_ctxt:	Context available for non-printing operations
  * @pbufs:		Pointer to nbcon private buffer
  */
 struct console {
@@ -417,6 +418,7 @@ struct console {
 
 	atomic_t		__private nbcon_state;
 	atomic_long_t		__private nbcon_seq;
+	struct nbcon_context	__private nbcon_device_ctxt;
 	struct printk_buffers	*pbufs;
 };
 
diff --git a/include/linux/printk.h b/include/linux/printk.h
index eee8e97da681..9687089f5ace 100644
--- a/include/linux/printk.h
+++ b/include/linux/printk.h
@@ -9,6 +9,8 @@
 #include <linux/ratelimit_types.h>
 #include <linux/once_lite.h>
 
+struct console;
+
 extern const char linux_banner[];
 extern const char linux_proc_banner[];
 
@@ -198,6 +200,8 @@ extern asmlinkage void dump_stack_lvl(const char *log_lvl) __cold;
 extern asmlinkage void dump_stack(void) __cold;
 void printk_trigger_flush(void);
 void console_try_replay_all(void);
+extern bool nbcon_device_try_acquire(struct console *con);
+extern void nbcon_device_release(struct console *con);
 #else
 static inline __printf(1, 0)
 int vprintk(const char *s, va_list args)
@@ -280,6 +284,16 @@ static inline void printk_trigger_flush(void)
 static inline void console_try_replay_all(void)
 {
 }
+
+static inline bool nbcon_device_try_acquire(struct console *con)
+{
+	return false;
+}
+
+static inline void nbcon_device_release(struct console *con)
+{
+}
+
 #endif
 
 bool this_cpu_in_panic(void);
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index f279f839741a..61f0ae6a4809 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -5,7 +5,9 @@
 #include <linux/kernel.h>
 #include <linux/console.h>
 #include <linux/delay.h>
+#include <linux/export.h>
 #include <linux/slab.h>
+#include <linux/string.h>
 #include "internal.h"
 /*
  * Printk console printing implementation for consoles which does not depend
@@ -546,6 +548,7 @@ static struct printk_buffers panic_nbcon_pbufs;
  * nbcon_context_try_acquire - Try to acquire nbcon console
  * @ctxt:	The context of the caller
  *
+ * Context:	Under @ctxt->con->device_lock() or local_irq_save().
  * Return:	True if the console was acquired. False otherwise.
  *
  * If the caller allowed an unsafe hostile takeover, on success the
@@ -553,7 +556,6 @@ static struct printk_buffers panic_nbcon_pbufs;
  * in an unsafe state. Otherwise, on success the caller may assume
  * the console is not in an unsafe state.
  */
-__maybe_unused
 static bool nbcon_context_try_acquire(struct nbcon_context *ctxt)
 {
 	unsigned int cpu = smp_processor_id();
@@ -1011,3 +1013,57 @@ void nbcon_free(struct console *con)
 
 	con->pbufs = NULL;
 }
+
+/**
+ * nbcon_device_try_acquire - Try to acquire nbcon console and enter unsafe
+ *				section
+ * @con:	The nbcon console to acquire
+ *
+ * Context:	Under the locking mechanism implemented in
+ *		@con->device_lock() including disabling migration.
+ * Return:	True if the console was acquired. False otherwise.
+ *
+ * Console drivers will usually use their own internal synchronization
+ * mechasism to synchronize between console printing and non-printing
+ * activities (such as setting baud rates). However, nbcon console drivers
+ * supporting atomic consoles may also want to mark unsafe sections when
+ * performing non-printing activities in order to synchronize against their
+ * atomic_write() callback.
+ *
+ * This function acquires the nbcon console using priority NBCON_PRIO_NORMAL
+ * and marks it unsafe for handover/takeover.
+ */
+bool nbcon_device_try_acquire(struct console *con)
+{
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(con, nbcon_device_ctxt);
+
+	cant_migrate();
+
+	memset(ctxt, 0, sizeof(*ctxt));
+	ctxt->console	= con;
+	ctxt->prio	= NBCON_PRIO_NORMAL;
+
+	if (!nbcon_context_try_acquire(ctxt))
+		return false;
+
+	if (!nbcon_context_enter_unsafe(ctxt))
+		return false;
+
+	return true;
+}
+EXPORT_SYMBOL_GPL(nbcon_device_try_acquire);
+
+/**
+ * nbcon_device_release - Exit unsafe section and release the nbcon console
+ * @con:	The nbcon console acquired in nbcon_device_try_acquire()
+ */
+void nbcon_device_release(struct console *con)
+{
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(con, nbcon_device_ctxt);
+
+	if (!nbcon_context_exit_unsafe(ctxt))
+		return;
+
+	nbcon_context_release(ctxt);
+}
+EXPORT_SYMBOL_GPL(nbcon_device_release);
-- 
2.39.2

[PATCH printk v7 14/35] serial: core: Acquire nbcon context in port->lock wrapper

[John Ogness]

Currently the port->lock wrappers uart_port_lock(),
uart_port_unlock() (and their variants) only lock/unlock
the spin_lock.

If the port is an nbcon console that has implemented the
write_atomic() callback, the wrappers must also acquire/release
the console context and mark the region as unsafe. This allows
general port->lock synchronization to be synchronized against
the nbcon write_atomic() callback.

Note that __uart_port_using_nbcon() relies on the port->lock
being held while a console is added and removed from the
console list (i.e. all uart nbcon drivers *must* take the
port->lock in their device_lock() callbacks).

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/serial_core.h | 82 ++++++++++++++++++++++++++++++++++++-
 1 file changed, 80 insertions(+), 2 deletions(-)

diff --git a/include/linux/serial_core.h b/include/linux/serial_core.h
index 2cf03ff2056a..4ab65874a850 100644
--- a/include/linux/serial_core.h
+++ b/include/linux/serial_core.h
@@ -11,6 +11,8 @@
 #include <linux/compiler.h>
 #include <linux/console.h>
 #include <linux/interrupt.h>
+#include <linux/lockdep.h>
+#include <linux/printk.h>
 #include <linux/spinlock.h>
 #include <linux/sched.h>
 #include <linux/tty.h>
@@ -625,6 +627,60 @@ static inline void uart_port_set_cons(struct uart_port *up, struct console *con)
 	up->cons = con;
 	__uart_port_unlock_irqrestore(up, flags);
 }
+
+/* Only for internal port lock wrapper usage. */
+static inline bool __uart_port_using_nbcon(struct uart_port *up)
+{
+	lockdep_assert_held_once(&up->lock);
+
+	if (likely(!uart_console(up)))
+		return false;
+
+	/*
+	 * @up->cons is only modified under the port lock. Therefore it is
+	 * certain that it cannot disappear here.
+	 *
+	 * @up->cons->node is added/removed from the console list under the
+	 * port lock. Therefore it is certain that the registration status
+	 * cannot change here, thus @up->cons->flags can be read directly.
+	 */
+	if (hlist_unhashed_lockless(&up->cons->node) ||
+	    !(up->cons->flags & CON_NBCON) ||
+	    !up->cons->write_atomic) {
+		return false;
+	}
+
+	return true;
+}
+
+/* Only for internal port lock wrapper usage. */
+static inline bool __uart_port_nbcon_try_acquire(struct uart_port *up)
+{
+	if (!__uart_port_using_nbcon(up))
+		return true;
+
+	return nbcon_device_try_acquire(up->cons);
+}
+
+/* Only for internal port lock wrapper usage. */
+static inline void __uart_port_nbcon_acquire(struct uart_port *up)
+{
+	if (!__uart_port_using_nbcon(up))
+		return;
+
+	while (!nbcon_device_try_acquire(up->cons))
+		cpu_relax();
+}
+
+/* Only for internal port lock wrapper usage. */
+static inline void __uart_port_nbcon_release(struct uart_port *up)
+{
+	if (!__uart_port_using_nbcon(up))
+		return;
+
+	nbcon_device_release(up->cons);
+}
+
 /**
  * uart_port_lock - Lock the UART port
  * @up:		Pointer to UART port structure
@@ -632,6 +688,7 @@ static inline void uart_port_set_cons(struct uart_port *up, struct console *con)
 static inline void uart_port_lock(struct uart_port *up)
 {
 	spin_lock(&up->lock);
+	__uart_port_nbcon_acquire(up);
 }
 
 /**
@@ -641,6 +698,7 @@ static inline void uart_port_lock(struct uart_port *up)
 static inline void uart_port_lock_irq(struct uart_port *up)
 {
 	spin_lock_irq(&up->lock);
+	__uart_port_nbcon_acquire(up);
 }
 
 /**
@@ -651,6 +709,7 @@ static inline void uart_port_lock_irq(struct uart_port *up)
 static inline void uart_port_lock_irqsave(struct uart_port *up, unsigned long *flags)
 {
 	spin_lock_irqsave(&up->lock, *flags);
+	__uart_port_nbcon_acquire(up);
 }
 
 /**
@@ -661,7 +720,15 @@ static inline void uart_port_lock_irqsave(struct uart_port *up, unsigned long *f
  */
 static inline bool uart_port_trylock(struct uart_port *up)
 {
-	return spin_trylock(&up->lock);
+	if (!spin_trylock(&up->lock))
+		return false;
+
+	if (!__uart_port_nbcon_try_acquire(up)) {
+		spin_unlock(&up->lock);
+		return false;
+	}
+
+	return true;
 }
 
 /**
@@ -673,7 +740,15 @@ static inline bool uart_port_trylock(struct uart_port *up)
  */
 static inline bool uart_port_trylock_irqsave(struct uart_port *up, unsigned long *flags)
 {
-	return spin_trylock_irqsave(&up->lock, *flags);
+	if (!spin_trylock_irqsave(&up->lock, *flags))
+		return false;
+
+	if (!__uart_port_nbcon_try_acquire(up)) {
+		spin_unlock_irqrestore(&up->lock, *flags);
+		return false;
+	}
+
+	return true;
 }
 
 /**
@@ -682,6 +757,7 @@ static inline bool uart_port_trylock_irqsave(struct uart_port *up, unsigned long
  */
 static inline void uart_port_unlock(struct uart_port *up)
 {
+	__uart_port_nbcon_release(up);
 	spin_unlock(&up->lock);
 }
 
@@ -691,6 +767,7 @@ static inline void uart_port_unlock(struct uart_port *up)
  */
 static inline void uart_port_unlock_irq(struct uart_port *up)
 {
+	__uart_port_nbcon_release(up);
 	spin_unlock_irq(&up->lock);
 }
 
@@ -701,6 +778,7 @@ static inline void uart_port_unlock_irq(struct uart_port *up)
  */
 static inline void uart_port_unlock_irqrestore(struct uart_port *up, unsigned long flags)
 {
+	__uart_port_nbcon_release(up);
 	spin_unlock_irqrestore(&up->lock, flags);
 }
 
-- 
2.39.2

[PATCH printk v7 15/35] printk: nbcon: Do not rely on proxy headers

[John Ogness]

The headers kernel.h, serial_core.h, and console.h allow for the
definitions of many types and functions from other headers.
Rather than relying on these as proxy headers, explicitly
include all headers providing needed definitions. Also sort the
list alphabetically to be able to easily detect duplicates.

Suggested-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/internal.h          |  8 ++++++--
 kernel/printk/nbcon.c             | 13 ++++++++++++-
 kernel/printk/printk_ringbuffer.h |  2 ++
 3 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index dc8bc0890fd2..ccb916688178 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -2,11 +2,12 @@
 /*
  * internal.h - printk internal definitions
  */
-#include <linux/percpu.h>
 #include <linux/console.h>
-#include "printk_ringbuffer.h"
+#include <linux/percpu.h>
+#include <linux/types.h>
 
 #if defined(CONFIG_PRINTK) && defined(CONFIG_SYSCTL)
+struct ctl_table;
 void __init printk_sysctl_init(void);
 int devkmsg_sysctl_set_loglvl(const struct ctl_table *table, int write,
 			      void *buffer, size_t *lenp, loff_t *ppos);
@@ -43,6 +44,9 @@ enum printk_info_flags {
 	LOG_CONT	= 8,	/* text is a fragment of a continuation line */
 };
 
+struct printk_ringbuffer;
+struct dev_printk_info;
+
 extern struct printk_ringbuffer *prb;
 
 __printf(4, 0)
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 61f0ae6a4809..e8ddcb6f7053 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -2,13 +2,24 @@
 // Copyright (C) 2022 Linutronix GmbH, John Ogness
 // Copyright (C) 2022 Intel, Thomas Gleixner
 
-#include <linux/kernel.h>
+#include <linux/atomic.h>
+#include <linux/bug.h>
 #include <linux/console.h>
 #include <linux/delay.h>
+#include <linux/errno.h>
 #include <linux/export.h>
+#include <linux/init.h>
+#include <linux/irqflags.h>
+#include <linux/minmax.h>
+#include <linux/percpu.h>
+#include <linux/preempt.h>
 #include <linux/slab.h>
+#include <linux/smp.h>
+#include <linux/stddef.h>
 #include <linux/string.h>
+#include <linux/types.h>
 #include "internal.h"
+#include "printk_ringbuffer.h"
 /*
  * Printk console printing implementation for consoles which does not depend
  * on the legacy style console_lock mechanism.
diff --git a/kernel/printk/printk_ringbuffer.h b/kernel/printk/printk_ringbuffer.h
index 52626d0f1fa3..bd2a892deac1 100644
--- a/kernel/printk/printk_ringbuffer.h
+++ b/kernel/printk/printk_ringbuffer.h
@@ -5,6 +5,8 @@
 
 #include <linux/atomic.h>
 #include <linux/dev_printk.h>
+#include <linux/stddef.h>
+#include <linux/types.h>
 
 /*
  * Meta information about each stored message.
-- 
2.39.2

[PATCH printk v7 16/35] printk: Make console_is_usable() available to nbcon.c

[John Ogness]

Move console_is_usable() as-is into internal.h so that it can
be used by nbcon printing functions as well.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/internal.h | 32 ++++++++++++++++++++++++++++++++
 kernel/printk/printk.c   | 30 ------------------------------
 2 files changed, 32 insertions(+), 30 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index ccb916688178..5d9deb56b582 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -84,6 +84,36 @@ void nbcon_seq_force(struct console *con, u64 seq);
 bool nbcon_alloc(struct console *con);
 void nbcon_free(struct console *con);
 
+/*
+ * Check if the given console is currently capable and allowed to print
+ * records.
+ *
+ * Requires the console_srcu_read_lock.
+ */
+static inline bool console_is_usable(struct console *con)
+{
+	short flags = console_srcu_read_flags(con);
+
+	if (!(flags & CON_ENABLED))
+		return false;
+
+	if ((flags & CON_SUSPENDED))
+		return false;
+
+	if (!con->write)
+		return false;
+
+	/*
+	 * Console drivers may assume that per-cpu resources have been
+	 * allocated. So unless they're explicitly marked as being able to
+	 * cope (CON_ANYTIME) don't call them until this CPU is officially up.
+	 */
+	if (!cpu_online(raw_smp_processor_id()) && !(flags & CON_ANYTIME))
+		return false;
+
+	return true;
+}
+
 #else
 
 #define PRINTK_PREFIX_MAX	0
@@ -104,6 +134,8 @@ static inline void nbcon_seq_force(struct console *con, u64 seq) { }
 static inline bool nbcon_alloc(struct console *con) { return false; }
 static inline void nbcon_free(struct console *con) { }
 
+static inline bool console_is_usable(struct console *con) { return false; }
+
 #endif /* CONFIG_PRINTK */
 
 extern struct printk_buffers printk_shared_pbufs;
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 7027f0868ef4..5a1fa8eebe5a 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -2767,36 +2767,6 @@ int is_console_locked(void)
 }
 EXPORT_SYMBOL(is_console_locked);
 
-/*
- * Check if the given console is currently capable and allowed to print
- * records.
- *
- * Requires the console_srcu_read_lock.
- */
-static inline bool console_is_usable(struct console *con)
-{
-	short flags = console_srcu_read_flags(con);
-
-	if (!(flags & CON_ENABLED))
-		return false;
-
-	if ((flags & CON_SUSPENDED))
-		return false;
-
-	if (!con->write)
-		return false;
-
-	/*
-	 * Console drivers may assume that per-cpu resources have been
-	 * allocated. So unless they're explicitly marked as being able to
-	 * cope (CON_ANYTIME) don't call them until this CPU is officially up.
-	 */
-	if (!cpu_online(raw_smp_processor_id()) && !(flags & CON_ANYTIME))
-		return false;
-
-	return true;
-}
-
 static void __console_unlock(void)
 {
 	console_locked = 0;
-- 
2.39.2

[PATCH printk v7 17/35] printk: Let console_is_usable() handle nbcon

[John Ogness]

The nbcon consoles use a different printing callback. For nbcon
consoles, check for the write_atomic() callback instead of
write().

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/internal.h | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 5d9deb56b582..448a5fcd5228 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -86,6 +86,8 @@ void nbcon_free(struct console *con);
 
 /*
  * Check if the given console is currently capable and allowed to print
+ * records. Note that this function does not consider the current context,
+ * which can also play a role in deciding if @con can be used to print
  * records.
  *
  * Requires the console_srcu_read_lock.
@@ -100,8 +102,13 @@ static inline bool console_is_usable(struct console *con)
 	if ((flags & CON_SUSPENDED))
 		return false;
 
-	if (!con->write)
-		return false;
+	if (flags & CON_NBCON) {
+		if (!con->write_atomic)
+			return false;
+	} else {
+		if (!con->write)
+			return false;
+	}
 
 	/*
 	 * Console drivers may assume that per-cpu resources have been
-- 
2.39.2

[PATCH printk v7 18/35] printk: Add @flags argument for console_is_usable()

[John Ogness]

The caller of console_is_usable() usually needs @console->flags
for its own checks. Rather than having console_is_usable() read
its own copy, make the caller pass in the @flags. This also
ensures that the caller saw the same @flags value.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/internal.h | 8 ++------
 kernel/printk/printk.c   | 5 +++--
 2 files changed, 5 insertions(+), 8 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 448a5fcd5228..fe8d84d78f1c 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -89,13 +89,9 @@ void nbcon_free(struct console *con);
  * records. Note that this function does not consider the current context,
  * which can also play a role in deciding if @con can be used to print
  * records.
- *
- * Requires the console_srcu_read_lock.
  */
-static inline bool console_is_usable(struct console *con)
+static inline bool console_is_usable(struct console *con, short flags)
 {
-	short flags = console_srcu_read_flags(con);
-
 	if (!(flags & CON_ENABLED))
 		return false;
 
@@ -141,7 +137,7 @@ static inline void nbcon_seq_force(struct console *con, u64 seq) { }
 static inline bool nbcon_alloc(struct console *con) { return false; }
 static inline void nbcon_free(struct console *con) { }
 
-static inline bool console_is_usable(struct console *con) { return false; }
+static inline bool console_is_usable(struct console *con, short flags) { return false; }
 
 #endif /* CONFIG_PRINTK */
 
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 5a1fa8eebe5a..92703249221e 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3012,9 +3012,10 @@ static bool console_flush_all(bool do_cond_resched, u64 *next_seq, bool *handove
 
 		cookie = console_srcu_read_lock();
 		for_each_console_srcu(con) {
+			short flags = console_srcu_read_flags(con);
 			bool progress;
 
-			if (!console_is_usable(con))
+			if (!console_is_usable(con, flags))
 				continue;
 			any_usable = true;
 
@@ -3925,7 +3926,7 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 			 * that they make forward progress, so only increment
 			 * @diff for usable consoles.
 			 */
-			if (!console_is_usable(c))
+			if (!console_is_usable(c, flags))
 				continue;
 
 			if (flags & CON_NBCON) {
-- 
2.39.2

[PATCH printk v7 19/35] printk: nbcon: Add helper to assign priority based on CPU state

[John Ogness]

Add a helper function to use the current state of the CPU to
determine which priority to assign to the printing context.

The EMERGENCY priority handling is added in a follow-up commit.
It will use a per-CPU variable.

Note: nbcon_device_try_acquire(), which is used by console
      drivers to acquire the nbcon console for non-printing
      activities, is hard-coded to always use NORMAL priority.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/internal.h |  2 ++
 kernel/printk/nbcon.c    | 19 +++++++++++++++++++
 2 files changed, 21 insertions(+)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index fe8d84d78f1c..72f229382cfa 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -83,6 +83,7 @@ u64 nbcon_seq_read(struct console *con);
 void nbcon_seq_force(struct console *con, u64 seq);
 bool nbcon_alloc(struct console *con);
 void nbcon_free(struct console *con);
+enum nbcon_prio nbcon_get_default_prio(void);
 
 /*
  * Check if the given console is currently capable and allowed to print
@@ -136,6 +137,7 @@ static inline u64 nbcon_seq_read(struct console *con) { return 0; }
 static inline void nbcon_seq_force(struct console *con, u64 seq) { }
 static inline bool nbcon_alloc(struct console *con) { return false; }
 static inline void nbcon_free(struct console *con) { }
+static inline enum nbcon_prio nbcon_get_default_prio(void) { return NBCON_PRIO_NONE; }
 
 static inline bool console_is_usable(struct console *con, short flags) { return false; }
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index e8ddcb6f7053..c6a9aa9f62f6 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -973,6 +973,25 @@ static bool nbcon_emit_next_record(struct nbcon_write_context *wctxt)
 	return nbcon_context_exit_unsafe(ctxt);
 }
 
+/**
+ * nbcon_get_default_prio - The appropriate nbcon priority to use for nbcon
+ *				printing on the current CPU
+ *
+ * Context:	Any context.
+ * Return:	The nbcon_prio to use for acquiring an nbcon console in this
+ *		context for printing.
+ *
+ * The function is safe for reading per-CPU data in any context because
+ * preemption is disabled if the current CPU is in the panic state.
+ */
+enum nbcon_prio nbcon_get_default_prio(void)
+{
+	if (this_cpu_in_panic())
+		return NBCON_PRIO_PANIC;
+
+	return NBCON_PRIO_NORMAL;
+}
+
 /**
  * nbcon_alloc - Allocate and init the nbcon console specific data
  * @con:	Console to initialize
-- 
2.39.2

[PATCH printk v7 20/35] printk: nbcon: Provide function to flush using write_atomic()

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

Provide nbcon_atomic_flush_pending() to perform flushing of all
registered nbcon consoles using their write_atomic() callback.

Unlike console_flush_all(), nbcon_atomic_flush_pending() will
only flush up through the newest record at the time of the
call. This prevents a CPU from printing unbounded when other
CPUs are adding records. If new records are added while
flushing, it is expected that the dedicated printer threads
will print those records. If the printer thread is not
available (which is always the case at this point in the
rework), nbcon_atomic_flush_pending() _will_ flush all records
in the ringbuffer.

Unlike console_flush_all(), nbcon_atomic_flush_pending() will
fully flush one console before flushing the next. This helps to
guarantee that a block of pending records (such as a stack
trace in an emergency situation) can be printed atomically at
once before releasing console ownership.

nbcon_atomic_flush_pending() is safe in any context because it
uses write_atomic() and acquires with unsafe_takeover disabled.

Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
---
 kernel/printk/internal.h |   2 +
 kernel/printk/nbcon.c    | 151 ++++++++++++++++++++++++++++++++++++++-
 2 files changed, 152 insertions(+), 1 deletion(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 72f229382cfa..0dc9b92b5dd0 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -84,6 +84,7 @@ void nbcon_seq_force(struct console *con, u64 seq);
 bool nbcon_alloc(struct console *con);
 void nbcon_free(struct console *con);
 enum nbcon_prio nbcon_get_default_prio(void);
+void nbcon_atomic_flush_pending(void);
 
 /*
  * Check if the given console is currently capable and allowed to print
@@ -138,6 +139,7 @@ static inline void nbcon_seq_force(struct console *con, u64 seq) { }
 static inline bool nbcon_alloc(struct console *con) { return false; }
 static inline void nbcon_free(struct console *con) { }
 static inline enum nbcon_prio nbcon_get_default_prio(void) { return NBCON_PRIO_NONE; }
+static inline void nbcon_atomic_flush_pending(void) { }
 
 static inline bool console_is_usable(struct console *con, short flags) { return false; }
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index c6a9aa9f62f6..3982d68979d6 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -886,7 +886,6 @@ EXPORT_SYMBOL_GPL(nbcon_exit_unsafe);
  * When true is returned, @wctxt->ctxt.backlog indicates whether there are
  * still records pending in the ringbuffer,
  */
-__maybe_unused
 static bool nbcon_emit_next_record(struct nbcon_write_context *wctxt)
 {
 	struct nbcon_context *ctxt = &ACCESS_PRIVATE(wctxt, ctxt);
@@ -992,6 +991,156 @@ enum nbcon_prio nbcon_get_default_prio(void)
 	return NBCON_PRIO_NORMAL;
 }
 
+/*
+ * __nbcon_atomic_flush_pending_con - Flush specified nbcon console using its
+ *					write_atomic() callback
+ * @con:			The nbcon console to flush
+ * @stop_seq:			Flush up until this record
+ *
+ * Return:	0 if @con was flushed up to @stop_seq Otherwise, error code on
+ *		failure.
+ *
+ * Errors:
+ *
+ *	-EPERM:		Unable to acquire console ownership.
+ *
+ *	-EAGAIN:	Another context took over ownership while printing.
+ *
+ *	-ENOENT:	A record before @stop_seq is not available.
+ *
+ * If flushing up to @stop_seq was not successful, it only makes sense for the
+ * caller to try again when -EAGAIN was returned. When -EPERM is returned,
+ * this context is not allowed to acquire the console. When -ENOENT is
+ * returned, it cannot be expected that the unfinalized record will become
+ * available.
+ */
+static int __nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
+{
+	struct nbcon_write_context wctxt = { };
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(&wctxt, ctxt);
+	int err = 0;
+
+	ctxt->console			= con;
+	ctxt->spinwait_max_us		= 2000;
+	ctxt->prio			= nbcon_get_default_prio();
+
+	if (!nbcon_context_try_acquire(ctxt))
+		return -EPERM;
+
+	while (nbcon_seq_read(con) < stop_seq) {
+		/*
+		 * nbcon_emit_next_record() returns false when the console was
+		 * handed over or taken over. In both cases the context is no
+		 * longer valid.
+		 */
+		if (!nbcon_emit_next_record(&wctxt))
+			return -EAGAIN;
+
+		if (!ctxt->backlog) {
+			/* Are there reserved but not yet finalized records? */
+			if (nbcon_seq_read(con) < stop_seq)
+				err = -ENOENT;
+			break;
+		}
+	}
+
+	nbcon_context_release(ctxt);
+	return err;
+}
+
+/**
+ * nbcon_atomic_flush_pending_con - Flush specified nbcon console using its
+ *					write_atomic() callback
+ * @con:			The nbcon console to flush
+ * @stop_seq:			Flush up until this record
+ *
+ * This will stop flushing before @stop_seq if another context has ownership.
+ * That context is then responsible for the flushing. Likewise, if new records
+ * are added while this context was flushing and there is no other context
+ * to handle the printing, this context must also flush those records.
+ */
+static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
+{
+	unsigned long flags;
+	int err;
+
+again:
+	/*
+	 * Atomic flushing does not use console driver synchronization (i.e.
+	 * it does not hold the port lock for uart consoles). Therefore IRQs
+	 * must be disabled to avoid being interrupted and then calling into
+	 * a driver that will deadlock trying to acquire console ownership.
+	 */
+	local_irq_save(flags);
+
+	err = __nbcon_atomic_flush_pending_con(con, stop_seq);
+
+	local_irq_restore(flags);
+
+	/*
+	 * If there was a new owner (-EPERM, -EAGAIN), that context is
+	 * responsible for completing.
+	 *
+	 * Do not wait for records not yet finalized (-ENOENT) to avoid a
+	 * possible deadlock. They will either get flushed by the writer or
+	 * eventually skipped on panic CPU.
+	 */
+	if (err)
+		return;
+
+	/*
+	 * If flushing was successful but more records are available, this
+	 * context must flush those remaining records because there is no
+	 * other context that will do it.
+	 */
+	if (prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
+		stop_seq = prb_next_reserve_seq(prb);
+		goto again;
+	}
+}
+
+/**
+ * __nbcon_atomic_flush_pending - Flush all nbcon consoles using their
+ *					write_atomic() callback
+ * @stop_seq:			Flush up until this record
+ */
+static void __nbcon_atomic_flush_pending(u64 stop_seq)
+{
+	struct console *con;
+	int cookie;
+
+	cookie = console_srcu_read_lock();
+	for_each_console_srcu(con) {
+		short flags = console_srcu_read_flags(con);
+
+		if (!(flags & CON_NBCON))
+			continue;
+
+		if (!console_is_usable(con, flags))
+			continue;
+
+		if (nbcon_seq_read(con) >= stop_seq)
+			continue;
+
+		nbcon_atomic_flush_pending_con(con, stop_seq);
+	}
+	console_srcu_read_unlock(cookie);
+}
+
+/**
+ * nbcon_atomic_flush_pending - Flush all nbcon consoles using their
+ *				write_atomic() callback
+ *
+ * Flush the backlog up through the currently newest record. Any new
+ * records added while flushing will not be flushed if there is another
+ * context available to handle the flushing. This is to avoid one CPU
+ * printing unbounded because other CPUs continue to add records.
+ */
+void nbcon_atomic_flush_pending(void)
+{
+	__nbcon_atomic_flush_pending(prb_next_reserve_seq(prb));
+}
+
 /**
  * nbcon_alloc - Allocate and init the nbcon console specific data
  * @con:	Console to initialize
-- 
2.39.2

[PATCH printk v7 21/35] printk: Track registered boot consoles

[John Ogness]

Unfortunately it is not known if a boot console and a regular
(legacy or nbcon) console use the same hardware. For this reason
they must not be allowed to print simultaneously.

For legacy consoles this is not an issue because they are
already synchronized with the boot consoles using the console
lock. However nbcon consoles can be triggered separately.

Add a global flag @have_boot_console to identify if any boot
consoles are registered. This will be used in follow-up commits
to ensure that boot consoles and nbcon consoles cannot print
simultaneously.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/printk.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 92703249221e..be4b9c491d8d 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -463,6 +463,14 @@ static int console_msg_format = MSG_FORMAT_DEFAULT;
 /* syslog_lock protects syslog_* variables and write access to clear_seq. */
 static DEFINE_MUTEX(syslog_lock);
 
+/*
+ * Specifies if a boot console is registered. If boot consoles are present,
+ * nbcon consoles cannot print simultaneously and must be synchronized by
+ * the console lock. This is because boot consoles and nbcon consoles may
+ * have mapped the same hardware.
+ */
+static bool have_boot_console;
+
 #ifdef CONFIG_PRINTK
 DECLARE_WAIT_QUEUE_HEAD(log_wait);
 /* All 3 protected by @syslog_lock. */
@@ -3610,6 +3618,9 @@ void register_console(struct console *newcon)
 		newcon->seq = init_seq;
 	}
 
+	if (newcon->flags & CON_BOOT)
+		have_boot_console = true;
+
 	/*
 	 * If another context is actively using the hardware of this new
 	 * console, it will not be aware of the nbcon synchronization. This
@@ -3680,7 +3691,9 @@ EXPORT_SYMBOL(register_console);
 static int unregister_console_locked(struct console *console)
 {
 	bool use_device_lock = (console->flags & CON_NBCON) && console->write_atomic;
+	bool found_boot_con = false;
 	unsigned long flags;
+	struct console *c;
 	int res;
 
 	lockdep_assert_console_list_lock_held();
@@ -3738,6 +3751,17 @@ static int unregister_console_locked(struct console *console)
 	if (console->exit)
 		res = console->exit(console);
 
+	/*
+	 * With this console gone, the global flags tracking registered
+	 * console types may have changed. Update them.
+	 */
+	for_each_console(c) {
+		if (c->flags & CON_BOOT)
+			found_boot_con = true;
+	}
+	if (!found_boot_con)
+		have_boot_console = found_boot_con;
+
 	return res;
 }
 
-- 
2.39.2

[PATCH printk v7 22/35] printk: nbcon: Use nbcon consoles in console_flush_all()

[John Ogness]

Allow nbcon consoles to print messages in the legacy printk()
caller context (printing via unlock) by integrating them into
console_flush_all(). The write_atomic() callback is used for
printing.

Provide nbcon_legacy_emit_next_record(), which acts as the
nbcon variant of console_emit_next_record(). Call this variant
within console_flush_all() for nbcon consoles. Since nbcon
consoles use their own @nbcon_seq variable to track the next
record to print, this also must be appropriately handled in
console_flush_all().

Note that the legacy printing logic uses @handover to detect
handovers for printing all consoles. For nbcon consoles,
handovers/takeovers occur on a per-console basis and thus do
not cause the console_flush_all() loop to abort.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/internal.h |  6 +++
 kernel/printk/nbcon.c    | 87 ++++++++++++++++++++++++++++++++++++++++
 kernel/printk/printk.c   | 17 +++++---
 3 files changed, 105 insertions(+), 5 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 0dc9b92b5dd0..44468f3828f3 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -78,6 +78,8 @@ void defer_console_output(void);
 
 u16 printk_parse_prefix(const char *text, int *level,
 			enum printk_info_flags *flags);
+void console_lock_spinning_enable(void);
+int console_lock_spinning_disable_and_check(int cookie);
 
 u64 nbcon_seq_read(struct console *con);
 void nbcon_seq_force(struct console *con, u64 seq);
@@ -85,6 +87,8 @@ bool nbcon_alloc(struct console *con);
 void nbcon_free(struct console *con);
 enum nbcon_prio nbcon_get_default_prio(void);
 void nbcon_atomic_flush_pending(void);
+bool nbcon_legacy_emit_next_record(struct console *con, bool *handover,
+				   int cookie);
 
 /*
  * Check if the given console is currently capable and allowed to print
@@ -140,6 +144,8 @@ static inline bool nbcon_alloc(struct console *con) { return false; }
 static inline void nbcon_free(struct console *con) { }
 static inline enum nbcon_prio nbcon_get_default_prio(void) { return NBCON_PRIO_NONE; }
 static inline void nbcon_atomic_flush_pending(void) { }
+static inline bool nbcon_legacy_emit_next_record(struct console *con, bool *handover,
+						 int cookie) { return false; }
 
 static inline bool console_is_usable(struct console *con, short flags) { return false; }
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 3982d68979d6..d09c084c9af4 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -992,6 +992,93 @@ enum nbcon_prio nbcon_get_default_prio(void)
 }
 
 /*
+ * nbcon_atomic_emit_one - Print one record for an nbcon console using the
+ *				write_atomic() callback
+ * @wctxt:	An initialized write context struct to use for this context
+ *
+ * Return:	True, when a record has been printed and there are still
+ *		pending records. The caller might want to continue flushing.
+ *
+ *		False, when there is no pending record, or when the console
+ *		context cannot be acquired, or the ownership has been lost.
+ *		The caller should give up. Either the job is done, cannot be
+ *		done, or will be handled by the owning context.
+ *
+ * This is an internal helper to handle the locking of the console before
+ * calling nbcon_emit_next_record().
+ */
+static bool nbcon_atomic_emit_one(struct nbcon_write_context *wctxt)
+{
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(wctxt, ctxt);
+
+	if (!nbcon_context_try_acquire(ctxt))
+		return false;
+
+	/*
+	 * nbcon_emit_next_record() returns false when the console was
+	 * handed over or taken over. In both cases the context is no
+	 * longer valid.
+	 *
+	 * The higher priority printing context takes over responsibility
+	 * to print the pending records.
+	 */
+	if (!nbcon_emit_next_record(wctxt))
+		return false;
+
+	nbcon_context_release(ctxt);
+
+	return ctxt->backlog;
+}
+
+/**
+ * nbcon_legacy_emit_next_record - Print one record for an nbcon console
+ *					in legacy contexts
+ * @con:	The console to print on
+ * @handover:	Will be set to true if a printk waiter has taken over the
+ *		console_lock, in which case the caller is no longer holding
+ *		both the console_lock and the SRCU read lock. Otherwise it
+ *		is set to false.
+ * @cookie:	The cookie from the SRCU read lock.
+ *
+ * Context:	Any context except NMI.
+ * Return:	True, when a record has been printed and there are still
+ *		pending records. The caller might want to continue flushing.
+ *
+ *		False, when there is no pending record, or when the console
+ *		context cannot be acquired, or the ownership has been lost.
+ *		The caller should give up. Either the job is done, cannot be
+ *		done, or will be handled by the owning context.
+ *
+ * This function is meant to be called by console_flush_all() to print records
+ * on nbcon consoles from legacy context (printing via console unlocking).
+ * Essentially it is the nbcon version of console_emit_next_record().
+ */
+bool nbcon_legacy_emit_next_record(struct console *con, bool *handover,
+				   int cookie)
+{
+	struct nbcon_write_context wctxt = { };
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(&wctxt, ctxt);
+	unsigned long flags;
+	bool progress;
+
+	/* Use the same procedure as console_emit_next_record(). */
+	printk_safe_enter_irqsave(flags);
+	console_lock_spinning_enable();
+	stop_critical_timings();
+
+	ctxt->console	= con;
+	ctxt->prio	= nbcon_get_default_prio();
+
+	progress = nbcon_atomic_emit_one(&wctxt);
+
+	start_critical_timings();
+	*handover = console_lock_spinning_disable_and_check(cookie);
+	printk_safe_exit_irqrestore(flags);
+
+	return progress;
+}
+
+/**
  * __nbcon_atomic_flush_pending_con - Flush specified nbcon console using its
  *					write_atomic() callback
  * @con:			The nbcon console to flush
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index be4b9c491d8d..5f5fc5df2f09 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -1860,7 +1860,7 @@ static bool console_waiter;
  * there may be a waiter spinning (like a spinlock). Also it must be
  * ready to hand over the lock at the end of the section.
  */
-static void console_lock_spinning_enable(void)
+void console_lock_spinning_enable(void)
 {
 	/*
 	 * Do not use spinning in panic(). The panic CPU wants to keep the lock.
@@ -1899,7 +1899,7 @@ static void console_lock_spinning_enable(void)
  *
  * Return: 1 if the lock rights were passed, 0 otherwise.
  */
-static int console_lock_spinning_disable_and_check(int cookie)
+int console_lock_spinning_disable_and_check(int cookie)
 {
 	int waiter;
 
@@ -3021,13 +3021,20 @@ static bool console_flush_all(bool do_cond_resched, u64 *next_seq, bool *handove
 		cookie = console_srcu_read_lock();
 		for_each_console_srcu(con) {
 			short flags = console_srcu_read_flags(con);
+			u64 printk_seq;
 			bool progress;
 
 			if (!console_is_usable(con, flags))
 				continue;
 			any_usable = true;
 
-			progress = console_emit_next_record(con, handover, cookie);
+			if (flags & CON_NBCON) {
+				progress = nbcon_legacy_emit_next_record(con, handover, cookie);
+				printk_seq = nbcon_seq_read(con);
+			} else {
+				progress = console_emit_next_record(con, handover, cookie);
+				printk_seq = con->seq;
+			}
 
 			/*
 			 * If a handover has occurred, the SRCU read lock
@@ -3037,8 +3044,8 @@ static bool console_flush_all(bool do_cond_resched, u64 *next_seq, bool *handove
 				return false;
 
 			/* Track the next of the highest seq flushed. */
-			if (con->seq > *next_seq)
-				*next_seq = con->seq;
+			if (printk_seq > *next_seq)
+				*next_seq = printk_seq;
 
 			if (!progress)
 				continue;
-- 
2.39.2

[PATCH printk v7 23/35] printk: Add is_printk_legacy_deferred()

[John Ogness]

If printk has been explicitly deferred or is called from NMI
context, legacy console printing must be deferred to an irq_work
context. Introduce a helper function is_printk_legacy_deferred()
for a CPU to query if it must defer legacy console printing.

In follow-up commits this helper will be needed at other call
sites as well.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 kernel/printk/internal.h    |  2 ++
 kernel/printk/printk_safe.c | 11 ++++++++++-
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 44468f3828f3..84706c1c934b 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -75,6 +75,7 @@ bool printk_percpu_data_ready(void);
 	} while (0)
 
 void defer_console_output(void);
+bool is_printk_legacy_deferred(void);
 
 u16 printk_parse_prefix(const char *text, int *level,
 			enum printk_info_flags *flags);
@@ -138,6 +139,7 @@ static inline bool console_is_usable(struct console *con, short flags)
 #define printk_safe_exit_irqrestore(flags) local_irq_restore(flags)
 
 static inline bool printk_percpu_data_ready(void) { return false; }
+static inline bool is_printk_legacy_deferred(void) { return false; }
 static inline u64 nbcon_seq_read(struct console *con) { return 0; }
 static inline void nbcon_seq_force(struct console *con, u64 seq) { }
 static inline bool nbcon_alloc(struct console *con) { return false; }
diff --git a/kernel/printk/printk_safe.c b/kernel/printk/printk_safe.c
index 4421ccac3113..86439fd20aab 100644
--- a/kernel/printk/printk_safe.c
+++ b/kernel/printk/printk_safe.c
@@ -38,6 +38,15 @@ void __printk_deferred_exit(void)
 	__printk_safe_exit();
 }
 
+bool is_printk_legacy_deferred(void)
+{
+	/*
+	 * The per-CPU variable @printk_context can be read safely in any
+	 * context. CPU migration is always disabled when set.
+	 */
+	return (this_cpu_read(printk_context) || in_nmi());
+}
+
 asmlinkage int vprintk(const char *fmt, va_list args)
 {
 #ifdef CONFIG_KGDB_KDB
@@ -50,7 +59,7 @@ asmlinkage int vprintk(const char *fmt, va_list args)
 	 * Use the main logbuf even in NMI. But avoid calling console
 	 * drivers that might have their own locks.
 	 */
-	if (this_cpu_read(printk_context) || in_nmi())
+	if (is_printk_legacy_deferred())
 		return vprintk_deferred(fmt, args);
 
 	/* No obstacles. */
-- 
2.39.2

[PATCH printk v7 24/35] printk: nbcon: Flush new records on device_release()

[John Ogness]

There may be new records that were added while a driver was
holding the nbcon context for non-printing purposes. These
new records must be flushed by the nbcon_device_release()
context because no other context will do it.

If boot consoles are registered, the legacy loop is used
(either direct or per irq_work) to handle the flushing.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 kernel/printk/internal.h |  2 ++
 kernel/printk/nbcon.c    | 20 ++++++++++++++++++++
 kernel/printk/printk.c   |  2 +-
 3 files changed, 23 insertions(+), 1 deletion(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 84706c1c934b..7679e18f24b3 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -153,6 +153,8 @@ static inline bool console_is_usable(struct console *con, short flags) { return
 
 #endif /* CONFIG_PRINTK */
 
+extern bool have_boot_console;
+
 extern struct printk_buffers printk_shared_pbufs;
 
 /**
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index d09c084c9af4..269aeed18064 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -1326,10 +1326,30 @@ EXPORT_SYMBOL_GPL(nbcon_device_try_acquire);
 void nbcon_device_release(struct console *con)
 {
 	struct nbcon_context *ctxt = &ACCESS_PRIVATE(con, nbcon_device_ctxt);
+	int cookie;
 
 	if (!nbcon_context_exit_unsafe(ctxt))
 		return;
 
 	nbcon_context_release(ctxt);
+
+	/*
+	 * This context must flush any new records added while the console
+	 * was locked. The console_srcu_read_lock must be taken to ensure
+	 * the console is usable throughout flushing.
+	 */
+	cookie = console_srcu_read_lock();
+	if (console_is_usable(con, console_srcu_read_flags(con)) &&
+	    prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
+		if (!have_boot_console) {
+			__nbcon_atomic_flush_pending_con(con, prb_next_reserve_seq(prb));
+		} else if (!is_printk_legacy_deferred()) {
+			if (console_trylock())
+				console_unlock();
+		} else {
+			printk_trigger_flush();
+		}
+	}
+	console_srcu_read_unlock(cookie);
 }
 EXPORT_SYMBOL_GPL(nbcon_device_release);
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 5f5fc5df2f09..d54eae0d06e9 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -469,7 +469,7 @@ static DEFINE_MUTEX(syslog_lock);
  * the console lock. This is because boot consoles and nbcon consoles may
  * have mapped the same hardware.
  */
-static bool have_boot_console;
+bool have_boot_console;
 
 #ifdef CONFIG_PRINTK
 DECLARE_WAIT_QUEUE_HEAD(log_wait);
-- 
2.39.2

[PATCH printk v7 25/35] printk: Flush nbcon consoles first on panic

[John Ogness]

In console_flush_on_panic(), flush the nbcon consoles before
flushing legacy consoles. The legacy write() callbacks are not
fully safe when oops_in_progress is set.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 kernel/printk/printk.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index d54eae0d06e9..f61ecb84c3f6 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3269,6 +3269,9 @@ void console_flush_on_panic(enum con_flush_mode mode)
 	if (mode == CONSOLE_REPLAY_ALL)
 		__console_rewind_all();
 
+	if (!have_boot_console)
+		nbcon_atomic_flush_pending();
+
 	console_flush_all(false, &next_seq, &handover);
 }
 
-- 
2.39.2

[PATCH printk v7 26/35] printk: nbcon: Add unsafe flushing on panic

[John Ogness]

Add nbcon_atomic_flush_unsafe() to flush all nbcon consoles
using the write_atomic() callback and allowing unsafe hostile
takeovers. Call this at the end of panic() as a final attempt
to flush any pending messages.

Note that legacy consoles use unsafe methods for flushing
from the beginning of panic (see bust_spinlocks()). Therefore,
systems using both legacy and nbcon consoles may still fail to
see panic messages due to unsafe legacy console usage.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 include/linux/printk.h |  5 +++++
 kernel/panic.c         |  1 +
 kernel/printk/nbcon.c  | 32 +++++++++++++++++++++++++-------
 3 files changed, 31 insertions(+), 7 deletions(-)

diff --git a/include/linux/printk.h b/include/linux/printk.h
index 9687089f5ace..2e083f01f8a3 100644
--- a/include/linux/printk.h
+++ b/include/linux/printk.h
@@ -202,6 +202,7 @@ void printk_trigger_flush(void);
 void console_try_replay_all(void);
 extern bool nbcon_device_try_acquire(struct console *con);
 extern void nbcon_device_release(struct console *con);
+void nbcon_atomic_flush_unsafe(void);
 #else
 static inline __printf(1, 0)
 int vprintk(const char *s, va_list args)
@@ -294,6 +295,10 @@ static inline void nbcon_device_release(struct console *con)
 {
 }
 
+static inline void nbcon_atomic_flush_unsafe(void)
+{
+}
+
 #endif
 
 bool this_cpu_in_panic(void);
diff --git a/kernel/panic.c b/kernel/panic.c
index f861bedc1925..a58631f42e79 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -457,6 +457,7 @@ void panic(const char *fmt, ...)
 	 * Explicitly flush the kernel log buffer one last time.
 	 */
 	console_flush_on_panic(CONSOLE_FLUSH_PENDING);
+	nbcon_atomic_flush_unsafe();
 
 	local_irq_enable();
 	for (i = 0; ; i += PANIC_TIMER_STEP) {
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 269aeed18064..afdb16c1c733 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -1083,6 +1083,7 @@ bool nbcon_legacy_emit_next_record(struct console *con, bool *handover,
  *					write_atomic() callback
  * @con:			The nbcon console to flush
  * @stop_seq:			Flush up until this record
+ * @allow_unsafe_takeover:	True, to allow unsafe hostile takeovers
  *
  * Return:	0 if @con was flushed up to @stop_seq Otherwise, error code on
  *		failure.
@@ -1101,7 +1102,8 @@ bool nbcon_legacy_emit_next_record(struct console *con, bool *handover,
  * returned, it cannot be expected that the unfinalized record will become
  * available.
  */
-static int __nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
+static int __nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq,
+					    bool allow_unsafe_takeover)
 {
 	struct nbcon_write_context wctxt = { };
 	struct nbcon_context *ctxt = &ACCESS_PRIVATE(&wctxt, ctxt);
@@ -1110,6 +1112,7 @@ static int __nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
 	ctxt->console			= con;
 	ctxt->spinwait_max_us		= 2000;
 	ctxt->prio			= nbcon_get_default_prio();
+	ctxt->allow_unsafe_takeover	= allow_unsafe_takeover;
 
 	if (!nbcon_context_try_acquire(ctxt))
 		return -EPERM;
@@ -1140,13 +1143,15 @@ static int __nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
  *					write_atomic() callback
  * @con:			The nbcon console to flush
  * @stop_seq:			Flush up until this record
+ * @allow_unsafe_takeover:	True, to allow unsafe hostile takeovers
  *
  * This will stop flushing before @stop_seq if another context has ownership.
  * That context is then responsible for the flushing. Likewise, if new records
  * are added while this context was flushing and there is no other context
  * to handle the printing, this context must also flush those records.
  */
-static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
+static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq,
+					   bool allow_unsafe_takeover)
 {
 	unsigned long flags;
 	int err;
@@ -1160,7 +1165,7 @@ static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
 	 */
 	local_irq_save(flags);
 
-	err = __nbcon_atomic_flush_pending_con(con, stop_seq);
+	err = __nbcon_atomic_flush_pending_con(con, stop_seq, allow_unsafe_takeover);
 
 	local_irq_restore(flags);
 
@@ -1190,8 +1195,9 @@ static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq)
  * __nbcon_atomic_flush_pending - Flush all nbcon consoles using their
  *					write_atomic() callback
  * @stop_seq:			Flush up until this record
+ * @allow_unsafe_takeover:	True, to allow unsafe hostile takeovers
  */
-static void __nbcon_atomic_flush_pending(u64 stop_seq)
+static void __nbcon_atomic_flush_pending(u64 stop_seq, bool allow_unsafe_takeover)
 {
 	struct console *con;
 	int cookie;
@@ -1209,7 +1215,7 @@ static void __nbcon_atomic_flush_pending(u64 stop_seq)
 		if (nbcon_seq_read(con) >= stop_seq)
 			continue;
 
-		nbcon_atomic_flush_pending_con(con, stop_seq);
+		nbcon_atomic_flush_pending_con(con, stop_seq, allow_unsafe_takeover);
 	}
 	console_srcu_read_unlock(cookie);
 }
@@ -1225,7 +1231,19 @@ static void __nbcon_atomic_flush_pending(u64 stop_seq)
  */
 void nbcon_atomic_flush_pending(void)
 {
-	__nbcon_atomic_flush_pending(prb_next_reserve_seq(prb));
+	__nbcon_atomic_flush_pending(prb_next_reserve_seq(prb), false);
+}
+
+/**
+ * nbcon_atomic_flush_unsafe - Flush all nbcon consoles using their
+ *	write_atomic() callback and allowing unsafe hostile takeovers
+ *
+ * Flush the backlog up through the currently newest record. Unsafe hostile
+ * takeovers will be performed, if necessary.
+ */
+void nbcon_atomic_flush_unsafe(void)
+{
+	__nbcon_atomic_flush_pending(prb_next_reserve_seq(prb), true);
 }
 
 /**
@@ -1342,7 +1360,7 @@ void nbcon_device_release(struct console *con)
 	if (console_is_usable(con, console_srcu_read_flags(con)) &&
 	    prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
 		if (!have_boot_console) {
-			__nbcon_atomic_flush_pending_con(con, prb_next_reserve_seq(prb));
+			__nbcon_atomic_flush_pending_con(con, prb_next_reserve_seq(prb), false);
 		} else if (!is_printk_legacy_deferred()) {
 			if (console_trylock())
 				console_unlock();
-- 
2.39.2

[PATCH printk v7 27/35] printk: Avoid console_lock dance if no legacy or boot consoles

[John Ogness]

Currently the console lock is used to attempt legacy-type
printing even if there are no legacy or boot consoles registered.
If no such consoles are registered, the console lock does not
need to be taken.

Add tracking of legacy console registration and use it with
boot console tracking to avoid unnecessary code paths, i.e.
do not use the console lock if there are no boot consoles
and no legacy consoles.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 kernel/printk/printk.c | 39 ++++++++++++++++++++++++++++++++++-----
 1 file changed, 34 insertions(+), 5 deletions(-)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index f61ecb84c3f6..a172c545e8dd 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -463,6 +463,13 @@ static int console_msg_format = MSG_FORMAT_DEFAULT;
 /* syslog_lock protects syslog_* variables and write access to clear_seq. */
 static DEFINE_MUTEX(syslog_lock);
 
+/*
+ * Specifies if a legacy console is registered. If legacy consoles are
+ * present, it is necessary to perform the console lock/unlock dance
+ * whenever console flushing should occur.
+ */
+static bool have_legacy_console;
+
 /*
  * Specifies if a boot console is registered. If boot consoles are present,
  * nbcon consoles cannot print simultaneously and must be synchronized by
@@ -471,6 +478,14 @@ static DEFINE_MUTEX(syslog_lock);
  */
 bool have_boot_console;
 
+/*
+ * Specifies if the console lock/unlock dance is needed for console
+ * printing. If @have_boot_console is true, the nbcon consoles will
+ * be printed serially along with the legacy consoles because nbcon
+ * consoles cannot print simultaneously with boot consoles.
+ */
+#define printing_via_unlock (have_legacy_console || have_boot_console)
+
 #ifdef CONFIG_PRINTK
 DECLARE_WAIT_QUEUE_HEAD(log_wait);
 /* All 3 protected by @syslog_lock. */
@@ -2339,7 +2354,7 @@ asmlinkage int vprintk_emit(int facility, int level,
 	printed_len = vprintk_store(facility, level, dev_info, fmt, args);
 
 	/* If called from the scheduler, we can not call up(). */
-	if (!in_sched) {
+	if (!in_sched && printing_via_unlock) {
 		/*
 		 * The caller may be holding system-critical or
 		 * timing-sensitive locks. Disable preemption during
@@ -2359,7 +2374,7 @@ asmlinkage int vprintk_emit(int facility, int level,
 		preempt_enable();
 	}
 
-	if (in_sched)
+	if (in_sched && printing_via_unlock)
 		defer_console_output();
 	else
 		wake_up_klogd();
@@ -2718,7 +2733,7 @@ void resume_console(void)
  */
 static int console_cpu_notify(unsigned int cpu)
 {
-	if (!cpuhp_tasks_frozen) {
+	if (!cpuhp_tasks_frozen && printing_via_unlock) {
 		/* If trylock fails, someone else is doing the printing */
 		if (console_trylock())
 			console_unlock();
@@ -3625,6 +3640,7 @@ void register_console(struct console *newcon)
 	if (newcon->flags & CON_NBCON) {
 		nbcon_seq_force(newcon, init_seq);
 	} else {
+		have_legacy_console = true;
 		newcon->seq = init_seq;
 	}
 
@@ -3701,6 +3717,7 @@ EXPORT_SYMBOL(register_console);
 static int unregister_console_locked(struct console *console)
 {
 	bool use_device_lock = (console->flags & CON_NBCON) && console->write_atomic;
+	bool found_legacy_con = false;
 	bool found_boot_con = false;
 	unsigned long flags;
 	struct console *c;
@@ -3768,9 +3785,13 @@ static int unregister_console_locked(struct console *console)
 	for_each_console(c) {
 		if (c->flags & CON_BOOT)
 			found_boot_con = true;
+		if (!(c->flags & CON_NBCON))
+			found_legacy_con = true;
 	}
 	if (!found_boot_con)
 		have_boot_console = found_boot_con;
+	if (!found_legacy_con)
+		have_legacy_console = found_legacy_con;
 
 	return res;
 }
@@ -3931,8 +3952,10 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 	seq = prb_next_reserve_seq(prb);
 
 	/* Flush the consoles so that records up to @seq are printed. */
-	console_lock();
-	console_unlock();
+	if (printing_via_unlock) {
+		console_lock();
+		console_unlock();
+	}
 
 	for (;;) {
 		unsigned long begin_jiffies;
@@ -3945,6 +3968,12 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 		 * console->seq. Releasing console_lock flushes more
 		 * records in case @seq is still not printed on all
 		 * usable consoles.
+		 *
+		 * Holding the console_lock is not necessary if there
+		 * are no legacy or boot consoles. However, such a
+		 * console could register at any time. Always hold the
+		 * console_lock as a precaution rather than
+		 * synchronizing against register_console().
 		 */
 		console_lock();
 
-- 
2.39.2

[PATCH printk v7 28/35] printk: Track nbcon consoles

[John Ogness]

Add a global flag @have_nbcon_console to identify if any nbcon
consoles are registered. This will be used in follow-up commits
to preserve legacy behavior when no nbcon consoles are registered.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/printk/printk.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index a172c545e8dd..2277871e1f12 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -470,6 +470,11 @@ static DEFINE_MUTEX(syslog_lock);
  */
 static bool have_legacy_console;
 
+/*
+ * Specifies if an nbcon console is registered.
+ */
+static bool have_nbcon_console;
+
 /*
  * Specifies if a boot console is registered. If boot consoles are present,
  * nbcon consoles cannot print simultaneously and must be synchronized by
@@ -3638,6 +3643,7 @@ void register_console(struct console *newcon)
 	init_seq = get_init_console_seq(newcon, bootcon_registered);
 
 	if (newcon->flags & CON_NBCON) {
+		have_nbcon_console = true;
 		nbcon_seq_force(newcon, init_seq);
 	} else {
 		have_legacy_console = true;
@@ -3718,6 +3724,7 @@ static int unregister_console_locked(struct console *console)
 {
 	bool use_device_lock = (console->flags & CON_NBCON) && console->write_atomic;
 	bool found_legacy_con = false;
+	bool found_nbcon_con = false;
 	bool found_boot_con = false;
 	unsigned long flags;
 	struct console *c;
@@ -3785,13 +3792,18 @@ static int unregister_console_locked(struct console *console)
 	for_each_console(c) {
 		if (c->flags & CON_BOOT)
 			found_boot_con = true;
-		if (!(c->flags & CON_NBCON))
+
+		if (c->flags & CON_NBCON)
+			found_nbcon_con = true;
+		else
 			found_legacy_con = true;
 	}
 	if (!found_boot_con)
 		have_boot_console = found_boot_con;
 	if (!found_legacy_con)
 		have_legacy_console = found_legacy_con;
+	if (!found_nbcon_con)
+		have_nbcon_console = found_nbcon_con;
 
 	return res;
 }
-- 
2.39.2

[PATCH printk v7 29/35] printk: Coordinate direct printing in panic

[John Ogness]

If legacy and nbcon consoles are registered and the nbcon
consoles are allowed to flush (i.e. no boot consoles
registered), the legacy consoles will no longer perform
direct printing on the panic CPU until after the backtrace
has been stored. This will give the safe nbcon consoles a
chance to print the panic messages before allowing the
unsafe legacy consoles to print.

If no nbcon consoles are registered or they are not allowed
to flush, there is no change in behavior (i.e. legacy
consoles will always attempt to print from the printk()
caller context).

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 include/linux/printk.h   |  5 ++++
 kernel/panic.c           |  2 ++
 kernel/printk/internal.h |  2 ++
 kernel/printk/printk.c   | 54 ++++++++++++++++++++++++++++++++++------
 4 files changed, 56 insertions(+), 7 deletions(-)

diff --git a/include/linux/printk.h b/include/linux/printk.h
index 2e083f01f8a3..eca9bb2ee637 100644
--- a/include/linux/printk.h
+++ b/include/linux/printk.h
@@ -200,6 +200,7 @@ extern asmlinkage void dump_stack_lvl(const char *log_lvl) __cold;
 extern asmlinkage void dump_stack(void) __cold;
 void printk_trigger_flush(void);
 void console_try_replay_all(void);
+void printk_legacy_allow_panic_sync(void);
 extern bool nbcon_device_try_acquire(struct console *con);
 extern void nbcon_device_release(struct console *con);
 void nbcon_atomic_flush_unsafe(void);
@@ -286,6 +287,10 @@ static inline void console_try_replay_all(void)
 {
 }
 
+static inline void printk_legacy_allow_panic_sync(void)
+{
+}
+
 static inline bool nbcon_device_try_acquire(struct console *con)
 {
 	return false;
diff --git a/kernel/panic.c b/kernel/panic.c
index a58631f42e79..84488daa14ce 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -368,6 +368,8 @@ void panic(const char *fmt, ...)
 
 	panic_other_cpus_shutdown(_crash_kexec_post_notifiers);
 
+	printk_legacy_allow_panic_sync();
+
 	/*
 	 * Run any panic handlers, including those that might need to
 	 * add information to the kmsg dump output.
diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 7679e18f24b3..dbda90f0dc08 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -130,6 +130,8 @@ static inline bool console_is_usable(struct console *con, short flags)
 #define PRINTK_MESSAGE_MAX	0
 #define PRINTKRB_RECORD_MAX	0
 
+#define legacy_allow_panic_sync	false
+
 /*
  * In !PRINTK builds we still export console_sem
  * semaphore and some of console functions (console_unlock()/etc.), so
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 2277871e1f12..480c0993abd5 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -471,7 +471,9 @@ static DEFINE_MUTEX(syslog_lock);
 static bool have_legacy_console;
 
 /*
- * Specifies if an nbcon console is registered.
+ * Specifies if an nbcon console is registered. If nbcon consoles are present,
+ * synchronous printing of legacy consoles will not occur during panic until
+ * the backtrace has been stored to the ringbuffer.
  */
 static bool have_nbcon_console;
 
@@ -2330,12 +2332,30 @@ int vprintk_store(int facility, int level,
 	return ret;
 }
 
+static bool legacy_allow_panic_sync;
+
+/*
+ * This acts as a one-way switch to allow legacy consoles to print from
+ * the printk() caller context on a panic CPU. It also attempts to flush
+ * the legacy consoles in this context.
+ */
+void printk_legacy_allow_panic_sync(void)
+{
+	legacy_allow_panic_sync = true;
+
+	if (printing_via_unlock && !is_printk_legacy_deferred()) {
+		if (console_trylock())
+			console_unlock();
+	}
+}
+
 asmlinkage int vprintk_emit(int facility, int level,
 			    const struct dev_printk_info *dev_info,
 			    const char *fmt, va_list args)
 {
+	bool do_trylock_unlock = printing_via_unlock;
+	bool defer_legacy = !do_trylock_unlock;
 	int printed_len;
-	bool in_sched = false;
 
 	/* Suppress unimportant messages after panic happens */
 	if (unlikely(suppress_printk))
@@ -2349,17 +2369,35 @@ asmlinkage int vprintk_emit(int facility, int level,
 	if (other_cpu_in_panic())
 		return 0;
 
+	/* If called from the scheduler, we can not call up(). */
 	if (level == LOGLEVEL_SCHED) {
 		level = LOGLEVEL_DEFAULT;
-		in_sched = true;
+		do_trylock_unlock = false;
+		defer_legacy = true;
 	}
 
 	printk_delay(level);
 
 	printed_len = vprintk_store(facility, level, dev_info, fmt, args);
 
-	/* If called from the scheduler, we can not call up(). */
-	if (!in_sched && printing_via_unlock) {
+	if (have_nbcon_console && !have_boot_console) {
+		nbcon_atomic_flush_pending();
+
+		/*
+		 * In panic, the legacy consoles are not allowed to print from
+		 * the printk calling context unless explicitly allowed. This
+		 * gives the safe nbcon consoles a chance to print out all the
+		 * panic messages first. This restriction only applies if
+		 * there are nbcon consoles registered and they are allowed to
+		 * flush.
+		 */
+		if (this_cpu_in_panic() && !legacy_allow_panic_sync) {
+			do_trylock_unlock = false;
+			defer_legacy = false;
+		}
+	}
+
+	if (do_trylock_unlock) {
 		/*
 		 * The caller may be holding system-critical or
 		 * timing-sensitive locks. Disable preemption during
@@ -2379,7 +2417,7 @@ asmlinkage int vprintk_emit(int facility, int level,
 		preempt_enable();
 	}
 
-	if (in_sched && printing_via_unlock)
+	if (defer_legacy)
 		defer_console_output();
 	else
 		wake_up_klogd();
@@ -3292,7 +3330,9 @@ void console_flush_on_panic(enum con_flush_mode mode)
 	if (!have_boot_console)
 		nbcon_atomic_flush_pending();
 
-	console_flush_all(false, &next_seq, &handover);
+	/* Flush legacy consoles once allowed, even when dangerous. */
+	if (legacy_allow_panic_sync)
+		console_flush_all(false, &next_seq, &handover);
 }
 
 /*
-- 
2.39.2

[PATCH printk v7 30/35] printk: Add helper for flush type logic

[John Ogness]

There are many call sites where console flushing occur.
Depending on the system state and types of consoles, the flush
methods to use are different. A flush call site generally must
consider:

	@have_boot_console
	@have_nbcon_console
	@have_legacy_console
	@legacy_allow_panic_sync
	is_printk_preferred()

and take into account the current CPU state:

	NBCON_PRIO_NORMAL
	NBCON_PRIO_EMERGENCY
	NBCON_PRIO_PANIC

in order to decide if it should:

	flush nbcon directly via atomic_write() callback
	flush legacy directly via console_unlock
	flush legacy via offload to irq_work

All of these call sites use their own logic to make this
decision, which is complicated and error prone. Especially
later when two more flush methods will be introduced:

	flush nbcon via offload to kthread
	flush legacy via offload to kthread

Introduce a new internal struct console_flush_type that
specifies the flush method(s) that are available for a
particular call site to use.

Introduce a helper function to fill out console_flush_type to
be used for flushing call sites.

In many system states it is acceptable to flush legacy directly
via console_unlock or via offload to irq_work. The caller must
decide which of these methods it prefers.

Replace the logic of all flushing call sites to use the new
helper.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 kernel/printk/internal.h | 65 +++++++++++++++++++++++++++++++++++++++
 kernel/printk/nbcon.c    | 25 ++++++++++++---
 kernel/printk/printk.c   | 66 ++++++++++++++++++----------------------
 3 files changed, 114 insertions(+), 42 deletions(-)

diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index dbda90f0dc08..58eb2a58b07b 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -48,6 +48,7 @@ struct printk_ringbuffer;
 struct dev_printk_info;
 
 extern struct printk_ringbuffer *prb;
+extern bool legacy_allow_panic_sync;
 
 __printf(4, 0)
 int vprintk_store(int facility, int level,
@@ -156,6 +157,70 @@ static inline bool console_is_usable(struct console *con, short flags) { return
 #endif /* CONFIG_PRINTK */
 
 extern bool have_boot_console;
+extern bool have_nbcon_console;
+extern bool have_legacy_console;
+
+/**
+ * struct console_flush_type - Define available console flush methods
+ * @nbcon_atomic:	Flush directly using nbcon_atomic() callback
+ * @legacy_direct:	Call the legacy loop in this context
+ * @legacy_offload:	Offload the legacy loop into IRQ
+ *
+ * Note that the legacy loop also flushes the nbcon consoles.
+ */
+struct console_flush_type {
+	bool	nbcon_atomic;
+	bool	legacy_direct;
+	bool	legacy_offload;
+};
+
+/*
+ * Identify which console flushing methods are available to the context of
+ * the caller. The caller can then decide which of the available flushing
+ * methods it will use.
+ */
+static inline void printk_get_console_flush_type(struct console_flush_type *ft)
+{
+	memset(ft, 0, sizeof(*ft));
+
+	switch (nbcon_get_default_prio()) {
+	case NBCON_PRIO_NORMAL:
+		if (have_nbcon_console && !have_boot_console)
+			ft->nbcon_atomic = true;
+
+		if (have_legacy_console || have_boot_console) {
+			ft->legacy_offload = true;
+			if (!is_printk_legacy_deferred())
+				ft->legacy_direct = true;
+		}
+		break;
+
+	case NBCON_PRIO_PANIC:
+		/*
+		 * In panic, the nbcon consoles will directly print. But
+		 * only allowed if there are no boot consoles.
+		 */
+		if (have_nbcon_console && !have_boot_console)
+			ft->nbcon_atomic = true;
+
+		/*
+		 * In panic, if nbcon atomic printing occurs, the legacy
+		 * consoles must remain silent until explicitly allowed.
+		 * Also, legacy consoles cannot print when deferred. However,
+		 * console_flush_on_panic() will flush them anyway, even if
+		 * unsafe.
+		 */
+		if ((legacy_allow_panic_sync || !ft->nbcon_atomic) &&
+		    !is_printk_legacy_deferred()) {
+			ft->legacy_direct = true;
+		}
+		break;
+
+	default:
+		WARN_ON_ONCE(1);
+		break;
+	}
+}
 
 extern struct printk_buffers printk_shared_pbufs;
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index afdb16c1c733..9e13327b4fe3 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -1153,6 +1153,7 @@ static int __nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq,
 static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq,
 					   bool allow_unsafe_takeover)
 {
+	struct console_flush_type ft;
 	unsigned long flags;
 	int err;
 
@@ -1186,8 +1187,16 @@ static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq,
 	 * other context that will do it.
 	 */
 	if (prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
-		stop_seq = prb_next_reserve_seq(prb);
-		goto again;
+		printk_get_console_flush_type(&ft);
+		/*
+		 * If nbcon_atomic flushing is not available, this printing
+		 * must be occurring via the legacy loop. Let that loop be
+		 * responsible for flushing the new records.
+		 */
+		if (ft.nbcon_atomic) {
+			stop_seq = prb_next_reserve_seq(prb);
+			goto again;
+		}
 	}
 }
 
@@ -1344,6 +1353,7 @@ EXPORT_SYMBOL_GPL(nbcon_device_try_acquire);
 void nbcon_device_release(struct console *con)
 {
 	struct nbcon_context *ctxt = &ACCESS_PRIVATE(con, nbcon_device_ctxt);
+	struct console_flush_type ft;
 	int cookie;
 
 	if (!nbcon_context_exit_unsafe(ctxt))
@@ -1359,12 +1369,17 @@ void nbcon_device_release(struct console *con)
 	cookie = console_srcu_read_lock();
 	if (console_is_usable(con, console_srcu_read_flags(con)) &&
 	    prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
-		if (!have_boot_console) {
+		/*
+		 * If nbcon_atomic flushing is not available, fallback to
+		 * using the legacy loop.
+		 */
+		printk_get_console_flush_type(&ft);
+		if (ft.nbcon_atomic) {
 			__nbcon_atomic_flush_pending_con(con, prb_next_reserve_seq(prb), false);
-		} else if (!is_printk_legacy_deferred()) {
+		} else if (ft.legacy_direct) {
 			if (console_trylock())
 				console_unlock();
-		} else {
+		} else if (ft.legacy_offload) {
 			printk_trigger_flush();
 		}
 	}
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 480c0993abd5..6bed91b5a55b 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -468,14 +468,14 @@ static DEFINE_MUTEX(syslog_lock);
  * present, it is necessary to perform the console lock/unlock dance
  * whenever console flushing should occur.
  */
-static bool have_legacy_console;
+bool have_legacy_console;
 
 /*
  * Specifies if an nbcon console is registered. If nbcon consoles are present,
  * synchronous printing of legacy consoles will not occur during panic until
  * the backtrace has been stored to the ringbuffer.
  */
-static bool have_nbcon_console;
+bool have_nbcon_console;
 
 /*
  * Specifies if a boot console is registered. If boot consoles are present,
@@ -485,14 +485,6 @@ static bool have_nbcon_console;
  */
 bool have_boot_console;
 
-/*
- * Specifies if the console lock/unlock dance is needed for console
- * printing. If @have_boot_console is true, the nbcon consoles will
- * be printed serially along with the legacy consoles because nbcon
- * consoles cannot print simultaneously with boot consoles.
- */
-#define printing_via_unlock (have_legacy_console || have_boot_console)
-
 #ifdef CONFIG_PRINTK
 DECLARE_WAIT_QUEUE_HEAD(log_wait);
 /* All 3 protected by @syslog_lock. */
@@ -2332,7 +2324,7 @@ int vprintk_store(int facility, int level,
 	return ret;
 }
 
-static bool legacy_allow_panic_sync;
+bool legacy_allow_panic_sync;
 
 /*
  * This acts as a one-way switch to allow legacy consoles to print from
@@ -2341,9 +2333,12 @@ static bool legacy_allow_panic_sync;
  */
 void printk_legacy_allow_panic_sync(void)
 {
+	struct console_flush_type ft;
+
 	legacy_allow_panic_sync = true;
 
-	if (printing_via_unlock && !is_printk_legacy_deferred()) {
+	printk_get_console_flush_type(&ft);
+	if (ft.legacy_direct) {
 		if (console_trylock())
 			console_unlock();
 	}
@@ -2353,8 +2348,8 @@ asmlinkage int vprintk_emit(int facility, int level,
 			    const struct dev_printk_info *dev_info,
 			    const char *fmt, va_list args)
 {
-	bool do_trylock_unlock = printing_via_unlock;
-	bool defer_legacy = !do_trylock_unlock;
+	struct console_flush_type ft;
+	bool do_trylock_unlock;
 	int printed_len;
 
 	/* Suppress unimportant messages after panic happens */
@@ -2369,34 +2364,23 @@ asmlinkage int vprintk_emit(int facility, int level,
 	if (other_cpu_in_panic())
 		return 0;
 
+	printk_get_console_flush_type(&ft);
+
 	/* If called from the scheduler, we can not call up(). */
 	if (level == LOGLEVEL_SCHED) {
 		level = LOGLEVEL_DEFAULT;
 		do_trylock_unlock = false;
-		defer_legacy = true;
+	} else {
+		do_trylock_unlock = ft.legacy_direct;
 	}
 
 	printk_delay(level);
 
 	printed_len = vprintk_store(facility, level, dev_info, fmt, args);
 
-	if (have_nbcon_console && !have_boot_console) {
+	if (ft.nbcon_atomic)
 		nbcon_atomic_flush_pending();
 
-		/*
-		 * In panic, the legacy consoles are not allowed to print from
-		 * the printk calling context unless explicitly allowed. This
-		 * gives the safe nbcon consoles a chance to print out all the
-		 * panic messages first. This restriction only applies if
-		 * there are nbcon consoles registered and they are allowed to
-		 * flush.
-		 */
-		if (this_cpu_in_panic() && !legacy_allow_panic_sync) {
-			do_trylock_unlock = false;
-			defer_legacy = false;
-		}
-	}
-
 	if (do_trylock_unlock) {
 		/*
 		 * The caller may be holding system-critical or
@@ -2417,7 +2401,7 @@ asmlinkage int vprintk_emit(int facility, int level,
 		preempt_enable();
 	}
 
-	if (defer_legacy)
+	if (!do_trylock_unlock && ft.legacy_offload)
 		defer_console_output();
 	else
 		wake_up_klogd();
@@ -2776,10 +2760,14 @@ void resume_console(void)
  */
 static int console_cpu_notify(unsigned int cpu)
 {
-	if (!cpuhp_tasks_frozen && printing_via_unlock) {
-		/* If trylock fails, someone else is doing the printing */
-		if (console_trylock())
-			console_unlock();
+	struct console_flush_type ft;
+
+	if (!cpuhp_tasks_frozen) {
+		printk_get_console_flush_type(&ft);
+		if (ft.legacy_direct) {
+			if (console_trylock())
+				console_unlock();
+		}
 	}
 	return 0;
 }
@@ -3304,6 +3292,7 @@ static void __console_rewind_all(void)
  */
 void console_flush_on_panic(enum con_flush_mode mode)
 {
+	struct console_flush_type ft;
 	bool handover;
 	u64 next_seq;
 
@@ -3327,7 +3316,8 @@ void console_flush_on_panic(enum con_flush_mode mode)
 	if (mode == CONSOLE_REPLAY_ALL)
 		__console_rewind_all();
 
-	if (!have_boot_console)
+	printk_get_console_flush_type(&ft);
+	if (ft.nbcon_atomic)
 		nbcon_atomic_flush_pending();
 
 	/* Flush legacy consoles once allowed, even when dangerous. */
@@ -3991,6 +3981,7 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 {
 	unsigned long timeout_jiffies = msecs_to_jiffies(timeout_ms);
 	unsigned long remaining_jiffies = timeout_jiffies;
+	struct console_flush_type ft;
 	struct console *c;
 	u64 last_diff = 0;
 	u64 printk_seq;
@@ -4004,7 +3995,8 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 	seq = prb_next_reserve_seq(prb);
 
 	/* Flush the consoles so that records up to @seq are printed. */
-	if (printing_via_unlock) {
+	printk_get_console_flush_type(&ft);
+	if (ft.legacy_direct) {
 		console_lock();
 		console_unlock();
 	}
-- 
2.39.2

[PATCH printk v7 31/35] printk: nbcon: Implement emergency sections

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

In emergency situations (something has gone wrong but the
system continues to operate), usually important information
(such as a backtrace) is generated via printk(). This
information should be pushed out to the consoles ASAP.

Add per-CPU emergency nesting tracking because an emergency
can arise while in an emergency situation.

Add functions to mark the beginning and end of emergency
sections where the urgent messages are generated.

Perform direct console flushing at the emergency priority if
the current CPU is in an emergency state and it is safe to do
so.

Note that the emergency state is not system-wide. While one CPU
is in an emergency state, another CPU may attempt to print
console messages at normal priority.

Also note that printk() already attempts to flush consoles in
the caller context for normal priority. However, follow-up
changes will introduce printing kthreads, in which case the
normal priority printk() calls will offload to the kthreads.

Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
---
 include/linux/console.h  |  4 +++
 kernel/printk/internal.h |  1 +
 kernel/printk/nbcon.c    | 75 +++++++++++++++++++++++++++++++++++++++-
 3 files changed, 79 insertions(+), 1 deletion(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index 3706f944de46..9a13f91b0c43 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -553,10 +553,14 @@ static inline bool console_is_registered(const struct console *con)
 	hlist_for_each_entry(con, &console_list, node)
 
 #ifdef CONFIG_PRINTK
+extern void nbcon_cpu_emergency_enter(void);
+extern void nbcon_cpu_emergency_exit(void);
 extern bool nbcon_can_proceed(struct nbcon_write_context *wctxt);
 extern bool nbcon_enter_unsafe(struct nbcon_write_context *wctxt);
 extern bool nbcon_exit_unsafe(struct nbcon_write_context *wctxt);
 #else
+static inline void nbcon_cpu_emergency_enter(void) { }
+static inline void nbcon_cpu_emergency_exit(void) { }
 static inline bool nbcon_can_proceed(struct nbcon_write_context *wctxt) { return false; }
 static inline bool nbcon_enter_unsafe(struct nbcon_write_context *wctxt) { return false; }
 static inline bool nbcon_exit_unsafe(struct nbcon_write_context *wctxt) { return false; }
diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 58eb2a58b07b..f51225a56f5f 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -185,6 +185,7 @@ static inline void printk_get_console_flush_type(struct console_flush_type *ft)
 
 	switch (nbcon_get_default_prio()) {
 	case NBCON_PRIO_NORMAL:
+	case NBCON_PRIO_EMERGENCY:
 		if (have_nbcon_console && !have_boot_console)
 			ft->nbcon_atomic = true;
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 9e13327b4fe3..a5551d8cbd50 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -972,6 +972,36 @@ static bool nbcon_emit_next_record(struct nbcon_write_context *wctxt)
 	return nbcon_context_exit_unsafe(ctxt);
 }
 
+/* Track the nbcon emergency nesting per CPU. */
+static DEFINE_PER_CPU(unsigned int, nbcon_pcpu_emergency_nesting);
+static unsigned int early_nbcon_pcpu_emergency_nesting __initdata;
+
+/**
+ * nbcon_get_cpu_emergency_nesting - Get the per CPU emergency nesting pointer
+ *
+ * Context:	For reading, any context. For writing, any context which could
+ *		not be migrated to another CPU.
+ * Return:	Either a pointer to the per CPU emergency nesting counter of
+ *		the current CPU or to the init data during early boot.
+ *
+ * The function is safe for reading per-CPU variables in any context because
+ * preemption is disabled if the current CPU is in the emergency state. See
+ * also nbcon_cpu_emergency_enter().
+ */
+static __ref unsigned int *nbcon_get_cpu_emergency_nesting(void)
+{
+	/*
+	 * The value of __printk_percpu_data_ready gets set in normal
+	 * context and before SMP initialization. As a result it could
+	 * never change while inside an nbcon emergency section.
+	 */
+	if (!printk_percpu_data_ready())
+		return &early_nbcon_pcpu_emergency_nesting;
+
+	/* Open code this_cpu_ptr() without checking migration. */
+	return per_cpu_ptr(&nbcon_pcpu_emergency_nesting, raw_smp_processor_id());
+}
+
 /**
  * nbcon_get_default_prio - The appropriate nbcon priority to use for nbcon
  *				printing on the current CPU
@@ -981,13 +1011,20 @@ static bool nbcon_emit_next_record(struct nbcon_write_context *wctxt)
  *		context for printing.
  *
  * The function is safe for reading per-CPU data in any context because
- * preemption is disabled if the current CPU is in the panic state.
+ * preemption is disabled if the current CPU is in the emergency or panic
+ * state.
  */
 enum nbcon_prio nbcon_get_default_prio(void)
 {
+	unsigned int *cpu_emergency_nesting;
+
 	if (this_cpu_in_panic())
 		return NBCON_PRIO_PANIC;
 
+	cpu_emergency_nesting = nbcon_get_cpu_emergency_nesting();
+	if (*cpu_emergency_nesting)
+		return NBCON_PRIO_EMERGENCY;
+
 	return NBCON_PRIO_NORMAL;
 }
 
@@ -1255,6 +1292,42 @@ void nbcon_atomic_flush_unsafe(void)
 	__nbcon_atomic_flush_pending(prb_next_reserve_seq(prb), true);
 }
 
+/**
+ * nbcon_cpu_emergency_enter - Enter an emergency section where printk()
+ *				messages for that CPU are flushed directly
+ *
+ * Context:	Any context. Disables preemption.
+ *
+ * When within an emergency section, printk() calls will attempt to flush any
+ * pending messages in the ringbuffer.
+ */
+void nbcon_cpu_emergency_enter(void)
+{
+	unsigned int *cpu_emergency_nesting;
+
+	preempt_disable();
+
+	cpu_emergency_nesting = nbcon_get_cpu_emergency_nesting();
+	(*cpu_emergency_nesting)++;
+}
+
+/**
+ * nbcon_cpu_emergency_exit - Exit an emergency section
+ *
+ * Context:	Within an emergency section. Enables preemption.
+ */
+void nbcon_cpu_emergency_exit(void)
+{
+	unsigned int *cpu_emergency_nesting;
+
+	cpu_emergency_nesting = nbcon_get_cpu_emergency_nesting();
+
+	if (!WARN_ON_ONCE(*cpu_emergency_nesting == 0))
+		(*cpu_emergency_nesting)--;
+
+	preempt_enable();
+}
+
 /**
  * nbcon_alloc - Allocate and init the nbcon console specific data
  * @con:	Console to initialize
-- 
2.39.2

[PATCH printk v7 32/35] panic: Mark emergency section in warn

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

Mark the full contents of __warn() as an emergency section. In
this section, every printk() call will attempt to directly
flush to the consoles using the EMERGENCY priority.

Co-developed-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/panic.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/panic.c b/kernel/panic.c
index 84488daa14ce..5fc44fd3b3dc 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -712,6 +712,8 @@ struct warn_args {
 void __warn(const char *file, int line, void *caller, unsigned taint,
 	    struct pt_regs *regs, struct warn_args *args)
 {
+	nbcon_cpu_emergency_enter();
+
 	disable_trace_on_warning();
 
 	if (file)
@@ -747,6 +749,8 @@ void __warn(const char *file, int line, void *caller, unsigned taint,
 
 	/* Just a warning, don't kill lockdep. */
 	add_taint(taint, LOCKDEP_STILL_OK);
+
+	nbcon_cpu_emergency_exit();
 }
 
 #ifdef CONFIG_BUG
-- 
2.39.2

[PATCH printk v7 33/35] panic: Mark emergency section in oops

[John Ogness]

Mark an emergency section beginning with oops_enter() until the
end of oops_exit(). In this section, every printk() call will
attempt to directly flush to the consoles using the EMERGENCY
priority.

The very end of oops_exit() performs a kmsg_dump(). This is not
included in the emergency section because it is another
flushing mechanism that should occur after the consoles have
flushed the oops messages.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/panic.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/panic.c b/kernel/panic.c
index 5fc44fd3b3dc..0843a275531a 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -679,6 +679,7 @@ bool oops_may_print(void)
  */
 void oops_enter(void)
 {
+	nbcon_cpu_emergency_enter();
 	tracing_off();
 	/* can't trust the integrity of the kernel anymore: */
 	debug_locks_off();
@@ -701,6 +702,7 @@ void oops_exit(void)
 {
 	do_oops_enter_exit();
 	print_oops_end_marker();
+	nbcon_cpu_emergency_exit();
 	kmsg_dump(KMSG_DUMP_OOPS);
 }
 
-- 
2.39.2

[PATCH printk v7 34/35] rcu: Mark emergency sections in rcu stalls

[John Ogness]

Mark emergency sections wherever multiple lines of
rcu stall information are generated. In an emergency
section, every printk() call will attempt to directly
flush to the consoles using the EMERGENCY priority.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/rcu/tree_exp.h   | 7 +++++++
 kernel/rcu/tree_stall.h | 9 +++++++++
 2 files changed, 16 insertions(+)

diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h
index 4acd29d16fdb..f6b35a0585a8 100644
--- a/kernel/rcu/tree_exp.h
+++ b/kernel/rcu/tree_exp.h
@@ -7,6 +7,7 @@
  * Authors: Paul E. McKenney <paulmck@linux.ibm.com>
  */
 
+#include <linux/console.h>
 #include <linux/lockdep.h>
 
 static void rcu_exp_handler(void *unused);
@@ -590,6 +591,9 @@ static void synchronize_rcu_expedited_wait(void)
 			return;
 		if (rcu_stall_is_suppressed())
 			continue;
+
+		nbcon_cpu_emergency_enter();
+
 		j = jiffies;
 		rcu_stall_notifier_call_chain(RCU_STALL_NOTIFY_EXP, (void *)(j - jiffies_start));
 		trace_rcu_stall_warning(rcu_state.name, TPS("ExpeditedStall"));
@@ -643,6 +647,9 @@ static void synchronize_rcu_expedited_wait(void)
 			rcu_exp_print_detail_task_stall_rnp(rnp);
 		}
 		jiffies_stall = 3 * rcu_exp_jiffies_till_stall_check() + 3;
+
+		nbcon_cpu_emergency_exit();
+
 		panic_on_rcu_stall();
 	}
 }
diff --git a/kernel/rcu/tree_stall.h b/kernel/rcu/tree_stall.h
index 4b0e9d7c4c68..b3a6943127bc 100644
--- a/kernel/rcu/tree_stall.h
+++ b/kernel/rcu/tree_stall.h
@@ -7,6 +7,7 @@
  * Author: Paul E. McKenney <paulmck@linux.ibm.com>
  */
 
+#include <linux/console.h>
 #include <linux/kvm_para.h>
 #include <linux/rcu_notifier.h>
 
@@ -605,6 +606,8 @@ static void print_other_cpu_stall(unsigned long gp_seq, unsigned long gps)
 	if (rcu_stall_is_suppressed())
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	/*
 	 * OK, time to rat on our buddy...
 	 * See Documentation/RCU/stallwarn.rst for info on how to debug
@@ -657,6 +660,8 @@ static void print_other_cpu_stall(unsigned long gp_seq, unsigned long gps)
 	rcu_check_gp_kthread_expired_fqs_timer();
 	rcu_check_gp_kthread_starvation();
 
+	nbcon_cpu_emergency_exit();
+
 	panic_on_rcu_stall();
 
 	rcu_force_quiescent_state();  /* Kick them all. */
@@ -677,6 +682,8 @@ static void print_cpu_stall(unsigned long gps)
 	if (rcu_stall_is_suppressed())
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	/*
 	 * OK, time to rat on ourselves...
 	 * See Documentation/RCU/stallwarn.rst for info on how to debug
@@ -706,6 +713,8 @@ static void print_cpu_stall(unsigned long gps)
 			   jiffies + 3 * rcu_jiffies_till_stall_check() + 3);
 	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
 
+	nbcon_cpu_emergency_exit();
+
 	panic_on_rcu_stall();
 
 	/*
-- 
2.39.2

[PATCH printk v7 35/35] lockdep: Mark emergency sections in lockdep splats

[John Ogness]

Mark emergency sections wherever multiple lines of
lock debugging output are generated. In an emergency
section, every printk() call will attempt to directly
flush to the consoles using the EMERGENCY priority.

Note that debug_show_all_locks() and
lockdep_print_held_locks() rely on their callers to
enter the emergency section. This is because these
functions can also be called in non-emergency
situations (such as sysrq).

Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
---
 kernel/locking/lockdep.c | 83 +++++++++++++++++++++++++++++++++++++++-
 1 file changed, 81 insertions(+), 2 deletions(-)

diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 58c88220a478..fea538f7d131 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -56,6 +56,7 @@
 #include <linux/kprobes.h>
 #include <linux/lockdep.h>
 #include <linux/context_tracking.h>
+#include <linux/console.h>
 
 #include <asm/sections.h>
 
@@ -573,8 +574,10 @@ static struct lock_trace *save_trace(void)
 		if (!debug_locks_off_graph_unlock())
 			return NULL;
 
+		nbcon_cpu_emergency_enter();
 		print_lockdep_off("BUG: MAX_STACK_TRACE_ENTRIES too low!");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 
 		return NULL;
 	}
@@ -887,11 +890,13 @@ look_up_lock_class(const struct lockdep_map *lock, unsigned int subclass)
 	if (unlikely(subclass >= MAX_LOCKDEP_SUBCLASSES)) {
 		instrumentation_begin();
 		debug_locks_off();
+		nbcon_cpu_emergency_enter();
 		printk(KERN_ERR
 			"BUG: looking up invalid subclass: %u\n", subclass);
 		printk(KERN_ERR
 			"turning off the locking correctness validator.\n");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 		instrumentation_end();
 		return NULL;
 	}
@@ -968,11 +973,13 @@ static bool assign_lock_key(struct lockdep_map *lock)
 	else {
 		/* Debug-check: all keys must be persistent! */
 		debug_locks_off();
+		nbcon_cpu_emergency_enter();
 		pr_err("INFO: trying to register non-static key.\n");
 		pr_err("The code is fine but needs lockdep annotation, or maybe\n");
 		pr_err("you didn't initialize this object before use?\n");
 		pr_err("turning off the locking correctness validator.\n");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 		return false;
 	}
 
@@ -1316,8 +1323,10 @@ register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
 			return NULL;
 		}
 
+		nbcon_cpu_emergency_enter();
 		print_lockdep_off("BUG: MAX_LOCKDEP_KEYS too low!");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 		return NULL;
 	}
 	nr_lock_classes++;
@@ -1349,11 +1358,13 @@ register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
 	if (verbose(class)) {
 		graph_unlock();
 
+		nbcon_cpu_emergency_enter();
 		printk("\nnew class %px: %s", class->key, class->name);
 		if (class->name_version > 1)
 			printk(KERN_CONT "#%d", class->name_version);
 		printk(KERN_CONT "\n");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 
 		if (!graph_lock()) {
 			return NULL;
@@ -1392,8 +1403,10 @@ static struct lock_list *alloc_list_entry(void)
 		if (!debug_locks_off_graph_unlock())
 			return NULL;
 
+		nbcon_cpu_emergency_enter();
 		print_lockdep_off("BUG: MAX_LOCKDEP_ENTRIES too low!");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 		return NULL;
 	}
 	nr_list_entries++;
@@ -2039,6 +2052,8 @@ static noinline void print_circular_bug(struct lock_list *this,
 
 	depth = get_lock_depth(target);
 
+	nbcon_cpu_emergency_enter();
+
 	print_circular_bug_header(target, depth, check_src, check_tgt);
 
 	parent = get_lock_parent(target);
@@ -2057,6 +2072,8 @@ static noinline void print_circular_bug(struct lock_list *this,
 
 	printk("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 static noinline void print_bfs_bug(int ret)
@@ -2569,6 +2586,8 @@ print_bad_irq_dependency(struct task_struct *curr,
 	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("=====================================================\n");
 	pr_warn("WARNING: %s-safe -> %s-unsafe lock order detected\n",
@@ -2618,11 +2637,13 @@ print_bad_irq_dependency(struct task_struct *curr,
 	pr_warn(" and %s-irq-unsafe lock:\n", irqclass);
 	next_root->trace = save_trace();
 	if (!next_root->trace)
-		return;
+		goto out;
 	print_shortest_lock_dependencies(forwards_entry, next_root);
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+out:
+	nbcon_cpu_emergency_exit();
 }
 
 static const char *state_names[] = {
@@ -2987,6 +3008,8 @@ print_deadlock_bug(struct task_struct *curr, struct held_lock *prev,
 	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("============================================\n");
 	pr_warn("WARNING: possible recursive locking detected\n");
@@ -3009,6 +3032,8 @@ print_deadlock_bug(struct task_struct *curr, struct held_lock *prev,
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 /*
@@ -3606,6 +3631,8 @@ static void print_collision(struct task_struct *curr,
 			struct held_lock *hlock_next,
 			struct lock_chain *chain)
 {
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("============================\n");
 	pr_warn("WARNING: chain_key collision\n");
@@ -3622,6 +3649,8 @@ static void print_collision(struct task_struct *curr,
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 #endif
 
@@ -3712,8 +3741,10 @@ static inline int add_chain_cache(struct task_struct *curr,
 		if (!debug_locks_off_graph_unlock())
 			return 0;
 
+		nbcon_cpu_emergency_enter();
 		print_lockdep_off("BUG: MAX_LOCKDEP_CHAINS too low!");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 		return 0;
 	}
 	chain->chain_key = chain_key;
@@ -3730,8 +3761,10 @@ static inline int add_chain_cache(struct task_struct *curr,
 		if (!debug_locks_off_graph_unlock())
 			return 0;
 
+		nbcon_cpu_emergency_enter();
 		print_lockdep_off("BUG: MAX_LOCKDEP_CHAIN_HLOCKS too low!");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 		return 0;
 	}
 
@@ -3970,6 +4003,8 @@ print_usage_bug(struct task_struct *curr, struct held_lock *this,
 	if (!debug_locks_off() || debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("================================\n");
 	pr_warn("WARNING: inconsistent lock state\n");
@@ -3998,6 +4033,8 @@ print_usage_bug(struct task_struct *curr, struct held_lock *this,
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 /*
@@ -4032,6 +4069,8 @@ print_irq_inversion_bug(struct task_struct *curr,
 	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("========================================================\n");
 	pr_warn("WARNING: possible irq lock inversion dependency detected\n");
@@ -4072,11 +4111,13 @@ print_irq_inversion_bug(struct task_struct *curr,
 	pr_warn("\nthe shortest dependencies between 2nd lock and 1st lock:\n");
 	root->trace = save_trace();
 	if (!root->trace)
-		return;
+		goto out;
 	print_shortest_lock_dependencies(other, root);
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+out:
+	nbcon_cpu_emergency_exit();
 }
 
 /*
@@ -4153,6 +4194,8 @@ void print_irqtrace_events(struct task_struct *curr)
 {
 	const struct irqtrace_events *trace = &curr->irqtrace;
 
+	nbcon_cpu_emergency_enter();
+
 	printk("irq event stamp: %u\n", trace->irq_events);
 	printk("hardirqs last  enabled at (%u): [<%px>] %pS\n",
 		trace->hardirq_enable_event, (void *)trace->hardirq_enable_ip,
@@ -4166,6 +4209,8 @@ void print_irqtrace_events(struct task_struct *curr)
 	printk("softirqs last disabled at (%u): [<%px>] %pS\n",
 		trace->softirq_disable_event, (void *)trace->softirq_disable_ip,
 		(void *)trace->softirq_disable_ip);
+
+	nbcon_cpu_emergency_exit();
 }
 
 static int HARDIRQ_verbose(struct lock_class *class)
@@ -4686,10 +4731,12 @@ static int mark_lock(struct task_struct *curr, struct held_lock *this,
 	 * We must printk outside of the graph_lock:
 	 */
 	if (ret == 2) {
+		nbcon_cpu_emergency_enter();
 		printk("\nmarked lock as {%s}:\n", usage_str[new_bit]);
 		print_lock(this);
 		print_irqtrace_events(curr);
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 	}
 
 	return ret;
@@ -4730,6 +4777,8 @@ print_lock_invalid_wait_context(struct task_struct *curr,
 	if (debug_locks_silent)
 		return 0;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("=============================\n");
 	pr_warn("[ BUG: Invalid wait context ]\n");
@@ -4749,6 +4798,8 @@ print_lock_invalid_wait_context(struct task_struct *curr,
 	pr_warn("stack backtrace:\n");
 	dump_stack();
 
+	nbcon_cpu_emergency_exit();
+
 	return 0;
 }
 
@@ -4956,6 +5007,8 @@ print_lock_nested_lock_not_held(struct task_struct *curr,
 	if (debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("==================================\n");
 	pr_warn("WARNING: Nested lock was not taken\n");
@@ -4976,6 +5029,8 @@ print_lock_nested_lock_not_held(struct task_struct *curr,
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 static int __lock_is_held(const struct lockdep_map *lock, int read);
@@ -5024,11 +5079,13 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
 	debug_class_ops_inc(class);
 
 	if (very_verbose(class)) {
+		nbcon_cpu_emergency_enter();
 		printk("\nacquire class [%px] %s", class->key, class->name);
 		if (class->name_version > 1)
 			printk(KERN_CONT "#%d", class->name_version);
 		printk(KERN_CONT "\n");
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 	}
 
 	/*
@@ -5155,6 +5212,7 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
 #endif
 	if (unlikely(curr->lockdep_depth >= MAX_LOCK_DEPTH)) {
 		debug_locks_off();
+		nbcon_cpu_emergency_enter();
 		print_lockdep_off("BUG: MAX_LOCK_DEPTH too low!");
 		printk(KERN_DEBUG "depth: %i  max: %lu!\n",
 		       curr->lockdep_depth, MAX_LOCK_DEPTH);
@@ -5162,6 +5220,7 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
 		lockdep_print_held_locks(current);
 		debug_show_all_locks();
 		dump_stack();
+		nbcon_cpu_emergency_exit();
 
 		return 0;
 	}
@@ -5181,6 +5240,8 @@ static void print_unlock_imbalance_bug(struct task_struct *curr,
 	if (debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("=====================================\n");
 	pr_warn("WARNING: bad unlock balance detected!\n");
@@ -5197,6 +5258,8 @@ static void print_unlock_imbalance_bug(struct task_struct *curr,
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 static noinstr int match_held_lock(const struct held_lock *hlock,
@@ -5901,6 +5964,8 @@ static void print_lock_contention_bug(struct task_struct *curr,
 	if (debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("=================================\n");
 	pr_warn("WARNING: bad contention detected!\n");
@@ -5917,6 +5982,8 @@ static void print_lock_contention_bug(struct task_struct *curr,
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 static void
@@ -6530,6 +6597,8 @@ print_freed_lock_bug(struct task_struct *curr, const void *mem_from,
 	if (debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("=========================\n");
 	pr_warn("WARNING: held lock freed!\n");
@@ -6542,6 +6611,8 @@ print_freed_lock_bug(struct task_struct *curr, const void *mem_from,
 
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 static inline int not_in_range(const void* mem_from, unsigned long mem_len,
@@ -6588,6 +6659,8 @@ static void print_held_locks_bug(void)
 	if (debug_locks_silent)
 		return;
 
+	nbcon_cpu_emergency_enter();
+
 	pr_warn("\n");
 	pr_warn("====================================\n");
 	pr_warn("WARNING: %s/%d still has locks held!\n",
@@ -6597,6 +6670,8 @@ static void print_held_locks_bug(void)
 	lockdep_print_held_locks(current);
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+
+	nbcon_cpu_emergency_exit();
 }
 
 void debug_check_no_locks_held(void)
@@ -6654,6 +6729,7 @@ asmlinkage __visible void lockdep_sys_exit(void)
 	if (unlikely(curr->lockdep_depth)) {
 		if (!debug_locks_off())
 			return;
+		nbcon_cpu_emergency_enter();
 		pr_warn("\n");
 		pr_warn("================================================\n");
 		pr_warn("WARNING: lock held when returning to user space!\n");
@@ -6662,6 +6738,7 @@ asmlinkage __visible void lockdep_sys_exit(void)
 		pr_warn("%s/%d is leaving the kernel with locks still held!\n",
 				curr->comm, curr->pid);
 		lockdep_print_held_locks(curr);
+		nbcon_cpu_emergency_exit();
 	}
 
 	/*
@@ -6678,6 +6755,7 @@ void lockdep_rcu_suspicious(const char *file, const int line, const char *s)
 	bool rcu = warn_rcu_enter();
 
 	/* Note: the following can be executed concurrently, so be careful. */
+	nbcon_cpu_emergency_enter();
 	pr_warn("\n");
 	pr_warn("=============================\n");
 	pr_warn("WARNING: suspicious RCU usage\n");
@@ -6716,6 +6794,7 @@ void lockdep_rcu_suspicious(const char *file, const int line, const char *s)
 	lockdep_print_held_locks(curr);
 	pr_warn("\nstack backtrace:\n");
 	dump_stack();
+	nbcon_cpu_emergency_exit();
 	warn_rcu_exit(rcu);
 }
 EXPORT_SYMBOL_GPL(lockdep_rcu_suspicious);
-- 
2.39.2

Re: [PATCH printk v7 02/35] printk: nbcon: Consolidate alloc() and init()

[Petr Mladek]

On Sun 2024-08-04 02:57:05, John Ogness wrote:
> Rather than splitting the nbcon allocation and initialization into
> two pieces, perform all initialization in nbcon_alloc(). Later,
> the initial sequence is calculated and can be explicitly set using
> nbcon_seq_force(). This removes the need for the strong rules of
> nbcon_init() that even included a BUG_ON().
> 
> Signed-off-by: John Ogness <john.ogness@linutronix.de>

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr

Re: [PATCH printk v7 20/35] printk: nbcon: Provide function to flush using write_atomic()

[Petr Mladek]

On Sun 2024-08-04 02:57:23, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> Provide nbcon_atomic_flush_pending() to perform flushing of all
> registered nbcon consoles using their write_atomic() callback.
> 
> Unlike console_flush_all(), nbcon_atomic_flush_pending() will
> only flush up through the newest record at the time of the
> call. This prevents a CPU from printing unbounded when other
> CPUs are adding records. If new records are added while
> flushing, it is expected that the dedicated printer threads
> will print those records. If the printer thread is not
> available (which is always the case at this point in the
> rework), nbcon_atomic_flush_pending() _will_ flush all records
> in the ringbuffer.
> 
> Unlike console_flush_all(), nbcon_atomic_flush_pending() will
> fully flush one console before flushing the next. This helps to
> guarantee that a block of pending records (such as a stack
> trace in an emergency situation) can be printed atomically at
> once before releasing console ownership.
> 
> nbcon_atomic_flush_pending() is safe in any context because it
> uses write_atomic() and acquires with unsafe_takeover disabled.
> 
> Co-developed-by: John Ogness <john.ogness@linutronix.de>
> Signed-off-by: John Ogness <john.ogness@linutronix.de>
> Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr

Re: [PATCH printk v7 23/35] printk: Add is_printk_legacy_deferred()

[Petr Mladek]

On Sun 2024-08-04 02:57:26, John Ogness wrote:
> If printk has been explicitly deferred or is called from NMI
> context, legacy console printing must be deferred to an irq_work
> context. Introduce a helper function is_printk_legacy_deferred()
> for a CPU to query if it must defer legacy console printing.
> 
> In follow-up commits this helper will be needed at other call
> sites as well.
> 
> Signed-off-by: John Ogness <john.ogness@linutronix.de>

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr

Re: [PATCH printk v7 24/35] printk: nbcon: Flush new records on device_release()

[Petr Mladek]

On Sun 2024-08-04 02:57:27, John Ogness wrote:
> There may be new records that were added while a driver was
> holding the nbcon context for non-printing purposes. These
> new records must be flushed by the nbcon_device_release()
> context because no other context will do it.
> 
> If boot consoles are registered, the legacy loop is used
> (either direct or per irq_work) to handle the flushing.
> 
> Signed-off-by: John Ogness <john.ogness@linutronix.de>

It makes some sense and seems to work. I do not know how to
make it better.

Reviewed-by: Petr Mladek <pmladek@suse.com>

But it makes me nervous a bit, see below:

> --- a/kernel/printk/nbcon.c
> +++ b/kernel/printk/nbcon.c
> @@ -1326,10 +1326,30 @@ EXPORT_SYMBOL_GPL(nbcon_device_try_acquire);
>  void nbcon_device_release(struct console *con)
>  {
>  	struct nbcon_context *ctxt = &ACCESS_PRIVATE(con, nbcon_device_ctxt);
> +	int cookie;
>  
>  	if (!nbcon_context_exit_unsafe(ctxt))
>  		return;
>  
>  	nbcon_context_release(ctxt);
> +
> +	/*
> +	 * This context must flush any new records added while the console
> +	 * was locked. The console_srcu_read_lock must be taken to ensure
> +	 * the console is usable throughout flushing.
> +	 */
> +	cookie = console_srcu_read_lock();
> +	if (console_is_usable(con, console_srcu_read_flags(con)) &&
> +	    prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
> +		if (!have_boot_console) {
> +			__nbcon_atomic_flush_pending_con(con, prb_next_reserve_seq(prb));
> +		} else if (!is_printk_legacy_deferred()) {
> +			if (console_trylock())
> +				console_unlock();

nbcon_device_release() is going to be called in uart_port_unlock*()
still under the port->lock.

=> It smells with a potential deadlock. The console_flush_all() in
   console_unlock() might want to flush this console under the
   port->lock as well.

   And it almost happens because nbcon_legacy_emit_next_record()
   might eventually take con->device_lock() when called in
   a task context.

   It will not happen here because this code uses console_trylock()
   which would set @console_may_schedule to false.

Anyway, it would look more safe when the flush was done after releasing
the port->lock.

I still have to think about this.

> +		} else {
> +			printk_trigger_flush();
> +		}
> +	}
> +	console_srcu_read_unlock(cookie);
>  }
>  EXPORT_SYMBOL_GPL(nbcon_device_release);

Best Regards,
Petr

Re: [PATCH printk v7 25/35] printk: Flush nbcon consoles first on panic

[Petr Mladek]

On Sun 2024-08-04 02:57:28, John Ogness wrote:
> In console_flush_on_panic(), flush the nbcon consoles before
> flushing legacy consoles. The legacy write() callbacks are not
> fully safe when oops_in_progress is set.
> 
> Signed-off-by: John Ogness <john.ogness@linutronix.de>

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr

Re: [PATCH printk v7 27/35] printk: Avoid console_lock dance if no legacy or boot consoles

[Petr Mladek]

On Sun 2024-08-04 02:57:30, John Ogness wrote:
> Currently the console lock is used to attempt legacy-type
> printing even if there are no legacy or boot consoles registered.
> If no such consoles are registered, the console lock does not
> need to be taken.
> 
> Add tracking of legacy console registration and use it with
> boot console tracking to avoid unnecessary code paths, i.e.
> do not use the console lock if there are no boot consoles
> and no legacy consoles.
> 
> Signed-off-by: John Ogness <john.ogness@linutronix.de>

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr

Re: [PATCH printk v7 29/35] printk: Coordinate direct printing in panic

[Petr Mladek]

Hi Linus,

On Sun 2024-08-04 02:57:32, John Ogness wrote:
> If legacy and nbcon consoles are registered and the nbcon
> consoles are allowed to flush (i.e. no boot consoles
> registered), the legacy consoles will no longer perform
> direct printing on the panic CPU until after the backtrace
> has been stored. This will give the safe nbcon consoles a
> chance to print the panic messages before allowing the
> unsafe legacy consoles to print.
> 
> If no nbcon consoles are registered or they are not allowed
> to flush, there is no change in behavior (i.e. legacy
> consoles will always attempt to print from the printk()
> caller context).

I want to be sure that this is acceptable to you.

This behavior has already been in the rejected pull request
for 6.11, see https://lore.kernel.org/r/Zp-_7R49fIHgIhaq@pathway.suse.cz

You did not complain about this particular change. But it is
yet another buffering in critical situation which you do not
like in general.

It is a bit different from the buffering during Oops. In this case,
the new nbcon consoles will still be flushed immediately. And
the legacy consoles will be blocked only when there is a nbcon
console.

The buffering here should increase the chance to see the messages
on the more safe (nbcon) consoles. The legacy consoles are less
safe primary because of the bust_spinlocks(1) called earlier.

I personally do not have strong opinion. The change makes sense.
It looks like it should make more good than harm. But it is not
a clear win.

Best Regards,
Petr

Re: [PATCH printk v7 29/35] printk: Coordinate direct printing in panic

[Petr Mladek]

On Sun 2024-08-04 02:57:32, John Ogness wrote:
> If legacy and nbcon consoles are registered and the nbcon
> consoles are allowed to flush (i.e. no boot consoles
> registered), the legacy consoles will no longer perform
> direct printing on the panic CPU until after the backtrace
> has been stored. This will give the safe nbcon consoles a
> chance to print the panic messages before allowing the
> unsafe legacy consoles to print.
> 
> If no nbcon consoles are registered or they are not allowed
> to flush, there is no change in behavior (i.e. legacy
> consoles will always attempt to print from the printk()
> caller context).
> 
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -2330,12 +2332,30 @@ int vprintk_store(int facility, int level,
>  	return ret;
>  }
>  
> +static bool legacy_allow_panic_sync;
> +
> +/*
> + * This acts as a one-way switch to allow legacy consoles to print from
> + * the printk() caller context on a panic CPU. It also attempts to flush
> + * the legacy consoles in this context.
> + */
> +void printk_legacy_allow_panic_sync(void)
> +{
> +	legacy_allow_panic_sync = true;
> +
> +	if (printing_via_unlock && !is_printk_legacy_deferred()) {
> +		if (console_trylock())
> +			console_unlock();
> +	}
> +}
> +
>  asmlinkage int vprintk_emit(int facility, int level,
>  			    const struct dev_printk_info *dev_info,
>  			    const char *fmt, va_list args)
>  {
> +	bool do_trylock_unlock = printing_via_unlock;
> +	bool defer_legacy = !do_trylock_unlock;

This should be:

	bool defer_legacy = false;

If we do not need to call the legacy loop then we do not need
to do defer it. The nbcon consoles are flushed directly, see below.


>  	int printed_len;
> -	bool in_sched = false;
>  
>  	/* Suppress unimportant messages after panic happens */
>  	if (unlikely(suppress_printk))
> @@ -2349,17 +2369,35 @@ asmlinkage int vprintk_emit(int facility, int level,
>  	if (other_cpu_in_panic())
>  		return 0;
>  
> +	/* If called from the scheduler, we can not call up(). */
>  	if (level == LOGLEVEL_SCHED) {
>  		level = LOGLEVEL_DEFAULT;
> -		in_sched = true;
> +		do_trylock_unlock = false;
> +		defer_legacy = true;

And this should be:

		defer_legacy = do_trylock_unlock;
		do_trylock_unlock = false;

>  	}
>  
>  	printk_delay(level);
>  
>  	printed_len = vprintk_store(facility, level, dev_info, fmt, args);
>  
> -	/* If called from the scheduler, we can not call up(). */
> -	if (!in_sched && printing_via_unlock) {
> +	if (have_nbcon_console && !have_boot_console) {
> +		nbcon_atomic_flush_pending();

The nbcon consoles will always be flushed directly. At least at this
stage of the patchset. I guess that a later patch will offload it to
the kthread when possible.

> +		/*
> +		 * In panic, the legacy consoles are not allowed to print from
> +		 * the printk calling context unless explicitly allowed. This
> +		 * gives the safe nbcon consoles a chance to print out all the
> +		 * panic messages first. This restriction only applies if
> +		 * there are nbcon consoles registered and they are allowed to
> +		 * flush.
> +		 */
> +		if (this_cpu_in_panic() && !legacy_allow_panic_sync) {
> +			do_trylock_unlock = false;
> +			defer_legacy = false;
> +		}
> +	}
> +
> +	if (do_trylock_unlock) {
>  		/*
>  		 * The caller may be holding system-critical or
>  		 * timing-sensitive locks. Disable preemption during
> @@ -3292,7 +3330,9 @@ void console_flush_on_panic(enum con_flush_mode mode)
>  	if (!have_boot_console)
>  		nbcon_atomic_flush_pending();
>  
> -	console_flush_all(false, &next_seq, &handover);
> +	/* Flush legacy consoles once allowed, even when dangerous. */
> +	if (legacy_allow_panic_sync)
> +		console_flush_all(false, &next_seq, &handover);
>  }

This is a good change. It make the console_flush_on_panic()
API more safe to use. console_flush_all() should not be called
before stopping other CPUs.

Best Regards,
Petr


>  
>  /*
> -- 
> 2.39.2

Re: [PATCH printk v7 29/35] printk: Coordinate direct printing in panic

[Linus Torvalds]

On Tue, 6 Aug 2024 at 02:59, Petr Mladek <pmladek@suse.com> wrote:
>
> I want to be sure that this is acceptable to you.

Just going by the superficial code, it's perfectly fine to delay other
consoles, as long as *some* consoles don't get delayed.

So yes:

> It is a bit different from the buffering during Oops. In this case,
> the new nbcon consoles will still be flushed immediately.

that sounds fine to me.

In general, I think delaying flushing is fine if

 (a) re-entrancy issues with already active (ie non-buffered) printing
on a console

 (b) there are higher-priority consoles that did flush it so that
something made it out

and obviously the whole "nbcon existing" will match that (b) case.

(In a perfect world the (a) case would be per-cpu, ie a "avoid obvious
deadlock", but honestly, I think any oops while the console is already
busy synchronously printing is understandable - and any concurrent
synchronous printing may be the more immediate cause of the oops -
we've certainly seen lots of cases where a bug on one CPU will then
immediately trigger on another one).

            Linus

Re: [PATCH printk v7 24/35] printk: nbcon: Flush new records on device_release()

[John Ogness]

On 2024-08-05, Petr Mladek <pmladek@suse.com> wrote:
>> +	/*
>> +	 * This context must flush any new records added while the console
>> +	 * was locked. The console_srcu_read_lock must be taken to ensure
>> +	 * the console is usable throughout flushing.
>> +	 */
>> +	cookie = console_srcu_read_lock();
>> +	if (console_is_usable(con, console_srcu_read_flags(con)) &&
>> +	    prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
>> +		if (!have_boot_console) {
>> +			__nbcon_atomic_flush_pending_con(con, prb_next_reserve_seq(prb));
>> +		} else if (!is_printk_legacy_deferred()) {
>> +			if (console_trylock())
>> +				console_unlock();
>
> nbcon_device_release() is going to be called in uart_port_unlock*()
> still under the port->lock.
>
> => It smells with a potential deadlock. The console_flush_all() in
>    console_unlock() might want to flush this console under the
>    port->lock as well.
>
>    And it almost happens because nbcon_legacy_emit_next_record()
>    might eventually take con->device_lock() when called in
>    a task context.
>
>    It will not happen here because this code uses console_trylock()
>    which would set @console_may_schedule to false.

Exactly. That is an important point. We must never try to invoke the
write_thread() callback while holding a spinlock.

> Anyway, it would look more safe when the flush was done after releasing
> the port->lock.

Even then we could never invoke the write_thread() callback because the
caller may be holding other spinlocks. If we cannot safely call
console_lock(), we cannot take the device lock. The atomic callback must
be used and that means the port lock cannot be involved in the
console_trylock().

John

Re: [PATCH printk v7 30/35] printk: Add helper for flush type logic

[Petr Mladek]

Hi,

I have spent a lot of time on this patch. I have got lost and rewrote
the comments several times. I hope that they make sense after all.

On Sun 2024-08-04 02:57:33, John Ogness wrote:
> There are many call sites where console flushing occur.
> Depending on the system state and types of consoles, the flush
> methods to use are different. A flush call site generally must
> consider:
> 
> Introduce a new internal struct console_flush_type that
> specifies the flush method(s) that are available for a
> particular call site to use.
> 
> Introduce a helper function to fill out console_flush_type to
> be used for flushing call sites.
> 
> In many system states it is acceptable to flush legacy directly
> via console_unlock or via offload to irq_work. The caller must
> decide which of these methods it prefers.
> 
> Replace the logic of all flushing call sites to use the new
> helper.
> 
> --- a/kernel/printk/internal.h
> +++ b/kernel/printk/internal.h
> @@ -156,6 +157,70 @@ static inline bool console_is_usable(struct console *con, short flags) { return
>  #endif /* CONFIG_PRINTK */
>  
>  extern bool have_boot_console;
> +extern bool have_nbcon_console;
> +extern bool have_legacy_console;
> +
> +/**
> + * struct console_flush_type - Define available console flush methods
> + * @nbcon_atomic:	Flush directly using nbcon_atomic() callback
> + * @legacy_direct:	Call the legacy loop in this context
> + * @legacy_offload:	Offload the legacy loop into IRQ
> + *
> + * Note that the legacy loop also flushes the nbcon consoles.
> + */
> +struct console_flush_type {
> +	bool	nbcon_atomic;
> +	bool	legacy_direct;
> +	bool	legacy_offload;
> +};
> +
> +/*
> + * Identify which console flushing methods are available to the context of
> + * the caller. The caller can then decide which of the available flushing
> + * methods it will use.
> + */
> +static inline void printk_get_console_flush_type(struct console_flush_type *ft)
> +{
> +	memset(ft, 0, sizeof(*ft));
> +
> +	switch (nbcon_get_default_prio()) {
> +	case NBCON_PRIO_NORMAL:
> +		if (have_nbcon_console && !have_boot_console)
> +			ft->nbcon_atomic = true;
> +
> +		if (have_legacy_console || have_boot_console) {
> +			ft->legacy_offload = true;
> +			if (!is_printk_legacy_deferred())
> +				ft->legacy_direct = true;
> +		}

I would suggest to change the semantic and set the _preferred_
flush method instead of an _available_ one. Something like:

		/* Legacy consoles are flushed directly when possible. */
		if (have_legacy_console || have_boot_console) {
			if (!is_printk_legacy_deferred()) {
				ft->legacy_direct = true;
			} else {
				ft->legacy_offload = true;
			}
		}


It would simplify vprintk_emit() in this patch. Also it might keep
nbcon_atomic_flush_pending_con() simple after adding the printk
thread. See below.

IMHO, it will not make any difference in other situations.

I personally prefer the "set preferred flush method" semantic. It
helps to make sure that all callers behave consistently. I mean that
this function would define the flush rules in "all" situations.

That said, I do not resist on it. The code looks reasonable also
with the current semantic. And it is rather easy to review
all callers. Also I might miss something. I still do not see the full
picture (patches adding the kthread).

> +		break;
> +
> +	case NBCON_PRIO_PANIC:
> +		/*
> +		 * In panic, the nbcon consoles will directly print. But
> +		 * only allowed if there are no boot consoles.
> +		 */
> +		if (have_nbcon_console && !have_boot_console)
> +			ft->nbcon_atomic = true;
> +
> +		/*
> +		 * In panic, if nbcon atomic printing occurs, the legacy
> +		 * consoles must remain silent until explicitly allowed.
> +		 * Also, legacy consoles cannot print when deferred. However,
> +		 * console_flush_on_panic() will flush them anyway, even if
> +		 * unsafe.
> +		 */
> +		if ((legacy_allow_panic_sync || !ft->nbcon_atomic) &&
> +		    !is_printk_legacy_deferred()) {
> +			ft->legacy_direct = true;

We should check here whether the legacy loop is needed at all.
I would write something like:

		if (have_legacy_console || have_boot_console) {
			/*
			 * Same decision as in PRIO_NORMAL. But we are interested
			 * only in wheter we could flush directly.
			 */
			if (!is_printk_legacy_deferred()) {
				ft->legacy_direct = true;
			}
			/* PRIO_PANIC specific part: Legacy consoles remain silent... */
			if (ft->nbcon_atomic && !legacy_allow_panic_sync)
				ft->legacy_direct = false;

> +		}
> +		break;
> +
> +	default:
> +		WARN_ON_ONCE(1);
> +		break;
> +	}
> +}
>  
>  extern struct printk_buffers printk_shared_pbufs;
>  
> --- a/kernel/printk/nbcon.c
> +++ b/kernel/printk/nbcon.c
> @@ -1186,8 +1187,16 @@ static void nbcon_atomic_flush_pending_con(struct console *con, u64 stop_seq,
>  	 * other context that will do it.
>  	 */
>  	if (prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
> -		stop_seq = prb_next_reserve_seq(prb);
> -		goto again;
> +		printk_get_console_flush_type(&ft);
> +		/*
> +		 * If nbcon_atomic flushing is not available, this printing
> +		 * must be occurring via the legacy loop. Let that loop be
> +		 * responsible for flushing the new records.
> +		 */

The comment is a bit confusing. It is not clear who is responsible for
calling the legacy loop.

My understanding is that the caller of this function is responsible
for calling the legacy loop. At least this is the case now. The
function is called only from vprintk_emit() and
console_flush_on_panic() and they both call the legacy loop when
needed.

I would rather remove this comment. Instead, I would make it clear
in the function description that: "The caller is responsible for
flushing the console via the legacy loop when needed."

> +		if (ft.nbcon_atomic) {
> +			stop_seq = prb_next_reserve_seq(prb);
> +			goto again;
> +		}

BTW: I wonder how this code would look like after adding the printk
     threads. We should do "goto again" only when ft.nbcon_atomic
     is the preferred (only available) flush type for nbcon consoles.

     IMHO, it is another reason to change the semantic.

>  	}
>  }
>  
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -2369,34 +2364,23 @@ asmlinkage int vprintk_emit(int facility, int level,
>  	if (other_cpu_in_panic())
>  		return 0;
>  
> +	printk_get_console_flush_type(&ft);

It is a nice trick to call printk_get_console_flush_type() this early.
I allows to hack the result when processing the hacky LOGLEVEL_SCHED ;-)

> +
>  	/* If called from the scheduler, we can not call up(). */
>  	if (level == LOGLEVEL_SCHED) {
>  		level = LOGLEVEL_DEFAULT;
>  		do_trylock_unlock = false;
> -		defer_legacy = true;
> +	} else {
> +		do_trylock_unlock = ft.legacy_direct;
>  	}

We could hack the @ft structure directly here:

	if (level == LOGLEVEL_SCHED) {
		level = LOGLEVEL_DEFAULT;
		ft.legacy_offload |= ft.legacy_direct;
		ft.legacy_direct = false;
	}
>  
>  	printk_delay(level);
>  
>  	printed_len = vprintk_store(facility, level, dev_info, fmt, args);
>  
> -	if (have_nbcon_console && !have_boot_console) {
> +	if (ft.nbcon_atomic)
>  		nbcon_atomic_flush_pending();
>  
> -		/*
> -		 * In panic, the legacy consoles are not allowed to print from
> -		 * the printk calling context unless explicitly allowed. This
> -		 * gives the safe nbcon consoles a chance to print out all the
> -		 * panic messages first. This restriction only applies if
> -		 * there are nbcon consoles registered and they are allowed to
> -		 * flush.
> -		 */
> -		if (this_cpu_in_panic() && !legacy_allow_panic_sync) {
> -			do_trylock_unlock = false;
> -			defer_legacy = false;
> -		}
> -	}
> -
>  	if (do_trylock_unlock) {

Then, with the "preferred flush method" semantic, we could use here:

	if (ft.legacy_direct) {

>  		/*
>  		 * The caller may be holding system-critical or
> @@ -2417,7 +2401,7 @@ asmlinkage int vprintk_emit(int facility, int level,
>  		preempt_enable();
>  	}
>  
> -	if (defer_legacy)
> +	if (!do_trylock_unlock && ft.legacy_offload)

and here:

	if (ft.legacy_offload)

>  		defer_console_output();
>  	else
>  		wake_up_klogd();
> @@ -2776,10 +2760,14 @@ void resume_console(void)
>   */
>  static int console_cpu_notify(unsigned int cpu)
>  {
> -	if (!cpuhp_tasks_frozen && printing_via_unlock) {
> -		/* If trylock fails, someone else is doing the printing */
> -		if (console_trylock())
> -			console_unlock();
> +	struct console_flush_type ft;
> +
> +	if (!cpuhp_tasks_frozen) {
> +		printk_get_console_flush_type(&ft);
> +		if (ft.legacy_direct) {
> +			if (console_trylock())
> +				console_unlock();

Why do we actually call only the legacy loop here?
IMHO, we should also do

	if (ft.nbcon_atomic)
 		nbcon_atomic_flush_pending();


> +		}
>  	}
>  	return 0;
>  }
> @@ -3327,7 +3316,8 @@ void console_flush_on_panic(enum con_flush_mode mode)
>  	if (mode == CONSOLE_REPLAY_ALL)
>  		__console_rewind_all();
>  
> -	if (!have_boot_console)
> +	printk_get_console_flush_type(&ft);
> +	if (ft.nbcon_atomic)
>  		nbcon_atomic_flush_pending();

I would use "ft.legacy_direct" also below for the decision about
the legacy loop:

-	if (legacy_allow_panic_sync)
+	if (ft.legacy_direct)
		console_flush_all(false, &next_seq, &handover);

>  
>  	/* Flush legacy consoles once allowed, even when dangerous. */

Best Regards,
Petr

Re: [PATCH printk v7 24/35] printk: nbcon: Flush new records on device_release()

[Petr Mladek]

On Wed 2024-08-07 03:21:57, John Ogness wrote:
> On 2024-08-05, Petr Mladek <pmladek@suse.com> wrote:
> >> +	/*
> >> +	 * This context must flush any new records added while the console
> >> +	 * was locked. The console_srcu_read_lock must be taken to ensure
> >> +	 * the console is usable throughout flushing.
> >> +	 */
> >> +	cookie = console_srcu_read_lock();
> >> +	if (console_is_usable(con, console_srcu_read_flags(con)) &&
> >> +	    prb_read_valid(prb, nbcon_seq_read(con), NULL)) {
> >> +		if (!have_boot_console) {
> >> +			__nbcon_atomic_flush_pending_con(con, prb_next_reserve_seq(prb));
> >> +		} else if (!is_printk_legacy_deferred()) {
> >> +			if (console_trylock())
> >> +				console_unlock();
> >
> > nbcon_device_release() is going to be called in uart_port_unlock*()
> > still under the port->lock.
> >
> > => It smells with a potential deadlock. The console_flush_all() in
> >    console_unlock() might want to flush this console under the
> >    port->lock as well.
> >
> >    And it almost happens because nbcon_legacy_emit_next_record()
> >    might eventually take con->device_lock() when called in
> >    a task context.
> >
> >    It will not happen here because this code uses console_trylock()
> >    which would set @console_may_schedule to false.
> 
> Exactly. That is an important point. We must never try to invoke the
> write_thread() callback while holding a spinlock.
> 
> > Anyway, it would look more safe when the flush was done after releasing
> > the port->lock.
> 
> Even then we could never invoke the write_thread() callback because the
> caller may be holding other spinlocks. If we cannot safely call
> console_lock(), we cannot take the device lock. The atomic callback must
> be used and that means the port lock cannot be involved in the
> console_trylock().

It makes sense. But it is not obvious.

I actually thought about using con->device_lock() lock around
nbcon_legacy_emit_next_record(). It would help to synchronize
the legacy loop against nbcon_device_lock()/nbcon_device_release().
It won't be needed to call the legacy loop in
nbcon_device_release().

But it was bad idea. There are more reasons to avoid taking
con->device_lock() after console_trylock():

  1. con->device_lock() might be a sleeping lock in the future. [*]
     It actually already is a sleeping lock in RT. And
     console_trylock() might be used in printk() in any context.

  2. The less locks we take the more safe printk() will be in
     various contexts and situations, especially in panic().


[*] I still have to wire this into my mental model. I keep forgetting it.

    It would be nice to have a documentation summarizing the main
    ideas and describing the printk() design. I hope that we will
    do it one day.

Best Regards,
Petr

Re: [PATCH printk v7 31/35] printk: nbcon: Implement emergency sections

[Petr Mladek]

On Sun 2024-08-04 02:57:34, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> In emergency situations (something has gone wrong but the
> system continues to operate), usually important information
> (such as a backtrace) is generated via printk(). This
> information should be pushed out to the consoles ASAP.
> 
> Add per-CPU emergency nesting tracking because an emergency
> can arise while in an emergency situation.
> 
> Add functions to mark the beginning and end of emergency
> sections where the urgent messages are generated.
> 
> Perform direct console flushing at the emergency priority if
> the current CPU is in an emergency state and it is safe to do
> so.
> 
> Note that the emergency state is not system-wide. While one CPU
> is in an emergency state, another CPU may attempt to print
> console messages at normal priority.
> 
> Also note that printk() already attempts to flush consoles in
> the caller context for normal priority. However, follow-up
> changes will introduce printing kthreads, in which case the
> normal priority printk() calls will offload to the kthreads.
> 
> Co-developed-by: John Ogness <john.ogness@linutronix.de>
> Signed-off-by: John Ogness <john.ogness@linutronix.de>
> Signed-off-by: Thomas Gleixner (Intel) <tglx@linutronix.de>

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr

Re: [PATCH printk v7 30/35] printk: Add helper for flush type logic

[John Ogness]

On 2024-08-07, Petr Mladek <pmladek@suse.com> wrote:
> I would suggest to change the semantic and set the _preferred_
> flush method instead of an _available_ one.

I will need to evaluate this closely. I worry that the caller needs to
understands how the helper function is choosing the preference. For
example, at the end you make a suggestion that is broken with this
suggested change.

>> +		if (ft.nbcon_atomic) {
>> +			stop_seq = prb_next_reserve_seq(prb);
>> +			goto again;
>> +		}
>
> BTW: I wonder how this code would look like after adding the printk
>      threads. We should do "goto again" only when ft.nbcon_atomic
>      is the preferred (only available) flush type for nbcon consoles.

	if (ft.nbcon_offload) {
		...
	} else if (ft.nbcon_atomic) {
		...
	}

>      IMHO, it is another reason to change the semantic.

The caller does not need to rely on the helper "choosing" the right
one. I understand your point that: It is easier for the caller when we
can blindly rely on the helper to choose for us. But I worry that if we
ever adjust the helper, we might break various call sites that blindly
rely on the helper making a certain choice. If the helper's job is only
to say what is possible, then I would worry less for the future when we
may need to adjust the helper.

>> +	printk_get_console_flush_type(&ft);
>
> It is a nice trick to call printk_get_console_flush_type() this early.
> I allows to hack the result when processing the hacky LOGLEVEL_SCHED ;-)
>
>> +
>>  	/* If called from the scheduler, we can not call up(). */
>>  	if (level == LOGLEVEL_SCHED) {
>>  		level = LOGLEVEL_DEFAULT;
>>  		do_trylock_unlock = false;
>> -		defer_legacy = true;
>> +	} else {
>> +		do_trylock_unlock = ft.legacy_direct;
>>  	}
>
> We could hack the @ft structure directly here:
>
> 	if (level == LOGLEVEL_SCHED) {
> 		level = LOGLEVEL_DEFAULT;
> 		ft.legacy_offload |= ft.legacy_direct;
> 		ft.legacy_direct = false;
> 	}

The hack seems a bit complicated to me. Especially when the helper is
choosing preferred methods. I will think about it.

>> +	if (!cpuhp_tasks_frozen) {
>> +		printk_get_console_flush_type(&ft);
>> +		if (ft.legacy_direct) {
>> +			if (console_trylock())
>> +				console_unlock();
>
> Why do we actually call only the legacy loop here?
> IMHO, we should also do
>
> 	if (ft.nbcon_atomic)
>  		nbcon_atomic_flush_pending();

Atomic consoles do not care if a CPU was online or not. I can add this,
but I expect there is nothing for the atomic console to flush. And when
threading is added, we would need the extra code to avoid atomic
flushing:

	if (!ft.nbcon_offload && ft.nbcon_atomic)                
		nbcon_atomic_flush_pending();

>> @@ -3327,7 +3316,8 @@ void console_flush_on_panic(enum con_flush_mode mode)
>>  	if (mode == CONSOLE_REPLAY_ALL)
>>  		__console_rewind_all();
>>  
>> -	if (!have_boot_console)
>> +	printk_get_console_flush_type(&ft);
>> +	if (ft.nbcon_atomic)
>>  		nbcon_atomic_flush_pending();
>
> I would use "ft.legacy_direct" also below for the decision about
> the legacy loop:
>
> -	if (legacy_allow_panic_sync)
> +	if (ft.legacy_direct)
> 		console_flush_all(false, &next_seq, &handover);

No, because it would mean the console is not flushed if the CPU is in
the deferred state. That is why I added an extra comment in the helper
saying that console_flush_on_panic() will _always_ flush directly.

I thought about adding that extra logic into the helper, but it really
isn't possible. @legacy_allow_panic_sync does not matter if there are no
nbcon consoles. So somehow the helper would need to know that CPU is in
the deferred state, but now it is allowed to do direct printing.

So it seemed more straight forward to have console_flush_on_panic() not
care about what is allowed (for legacy). It is going to flush directly
no matter what.

I will reconsider your suggestions about the helper and also compare the
end result at the call sites (also with threading changes applied) to
see what looks simpler to maintain.

John

Re: [PATCH printk v7 30/35] printk: Add helper for flush type logic

[Petr Mladek]

On Wed 2024-08-07 16:17:51, John Ogness wrote:
> On 2024-08-07, Petr Mladek <pmladek@suse.com> wrote:
> > I would suggest to change the semantic and set the _preferred_
> > flush method instead of an _available_ one.
> 
> I will need to evaluate this closely. I worry that the caller needs to
> understands how the helper function is choosing the preference. For
> example, at the end you make a suggestion that is broken with this
> suggested change.

I see. Well, console_flush_on_panic() is special. It is the last
resort. It actually ignores even the "what is available/allowed"
semantic.

> >> +		if (ft.nbcon_atomic) {
> >> +			stop_seq = prb_next_reserve_seq(prb);
> >> +			goto again;
> >> +		}
> >
> > BTW: I wonder how this code would look like after adding the printk
> >      threads. We should do "goto again" only when ft.nbcon_atomic
> >      is the preferred (only available) flush type for nbcon consoles.
> 
> 	if (ft.nbcon_offload) {
> 		...
> 	} else if (ft.nbcon_atomic) {
> 		...
> 	}
> 
> >      IMHO, it is another reason to change the semantic.
> 
> The caller does not need to rely on the helper "choosing" the right
> one. I understand your point that: It is easier for the caller when we
> can blindly rely on the helper to choose for us. But I worry that if we
> ever adjust the helper, we might break various call sites that blindly
> rely on the helper making a certain choice. If the helper's job is only
> to say what is possible, then I would worry less for the future when we
> may need to adjust the helper.

In the ideal world, the helper should tell the caller what has to be
done to flush pending messages on all consoles. The helper makes
the decision using the global variables where the variables define:

   + type of registered consoles
   + NBCON_PRIO
   + deferred context
   + possible or forced offload to kthreads

IMHO, it depends on how many callers are happy with the proposed
solution.

> >> +	printk_get_console_flush_type(&ft);
> >
> > It is a nice trick to call printk_get_console_flush_type() this early.
> > I allows to hack the result when processing the hacky LOGLEVEL_SCHED ;-)
> >
> >> +
> >>  	/* If called from the scheduler, we can not call up(). */
> >>  	if (level == LOGLEVEL_SCHED) {
> >>  		level = LOGLEVEL_DEFAULT;
> >>  		do_trylock_unlock = false;
> >> -		defer_legacy = true;
> >> +	} else {
> >> +		do_trylock_unlock = ft.legacy_direct;
> >>  	}
> >
> > We could hack the @ft structure directly here:
> >
> > 	if (level == LOGLEVEL_SCHED) {
> > 		level = LOGLEVEL_DEFAULT;
> > 		ft.legacy_offload |= ft.legacy_direct;
> > 		ft.legacy_direct = false;
> > 	}
> 
> The hack seems a bit complicated to me. Especially when the helper is
> choosing preferred methods. I will think about it.

The hack converts legacy_direct -> legacy_offload.

> >> +	if (!cpuhp_tasks_frozen) {
> >> +		printk_get_console_flush_type(&ft);
> >> +		if (ft.legacy_direct) {
> >> +			if (console_trylock())
> >> +				console_unlock();
> >
> > Why do we actually call only the legacy loop here?
> > IMHO, we should also do
> >
> > 	if (ft.nbcon_atomic)
> >  		nbcon_atomic_flush_pending();
> 
> Atomic consoles do not care if a CPU was online or not. I can add this,
> but I expect there is nothing for the atomic console to flush.

console_cpu_notify() has been added by the commit 034260d6779087431
("printk: fix delayed messages from CPU hotplug events"). It
is related to the check

    cpu_online(smp_processor_id()) == 0

which is still called in console_is_usable() even for nbcon consoles.

IMHO, it means that nbcon_atomic_flush_pending() might not be able
to flush the messages when called from vprintk_emit() on CPU which
is just being hot-plugged.


> And when
> threading is added, we would need the extra code to avoid atomic
> flushing:
> 
> 	if (!ft.nbcon_offload && ft.nbcon_atomic)                
> 		nbcon_atomic_flush_pending();

This extra change won't be needed when printk_get_console_flush_type(&ft)
uses the "set the preferred flush type" semantic ;-)


> >> @@ -3327,7 +3316,8 @@ void console_flush_on_panic(enum con_flush_mode mode)
> >>  	if (mode == CONSOLE_REPLAY_ALL)
> >>  		__console_rewind_all();
> >>  
> >> -	if (!have_boot_console)
> >> +	printk_get_console_flush_type(&ft);
> >> +	if (ft.nbcon_atomic)
> >>  		nbcon_atomic_flush_pending();
> >
> > I would use "ft.legacy_direct" also below for the decision about
> > the legacy loop:
> >
> > -	if (legacy_allow_panic_sync)
> > +	if (ft.legacy_direct)
> > 		console_flush_all(false, &next_seq, &handover);
> 
> No, because it would mean the console is not flushed if the CPU is in
> the deferred state. That is why I added an extra comment in the helper
> saying that console_flush_on_panic() will _always_ flush directly.
> 
> I thought about adding that extra logic into the helper, but it really
> isn't possible. @legacy_allow_panic_sync does not matter if there are no
> nbcon consoles. So somehow the helper would need to know that CPU is in
> the deferred state, but now it is allowed to do direct printing.
> 
> So it seemed more straight forward to have console_flush_on_panic() not
> care about what is allowed (for legacy). It is going to flush directly
> no matter what.

Makes sense. console_flush_on_panic() is special. It is the last
resort. And it does things which are not safe/allowed.

> I will reconsider your suggestions about the helper and also compare the
> end result at the call sites (also with threading changes applied) to
> see what looks simpler to maintain.

Sigh, this is a kind of "bike shedding" discussion. It makes some
sense as long as it helps to find bugs and simplify the logic.
Feel free to stop it when you think that it is not longer worth
the effort.

Best Regards,
Petr