[PATCH printk v3 0/7] provide nbcon base

[PATCH printk v3 0/7] provide nbcon base

[John Ogness]

Hi,

This is v3 of a series to introduce the new non-BKL (nbcon)
consoles. v2 is here [0]. For information about the motivation
of the atomic consoles, please read the cover letter of v1 [1].

This series focuses on providing the base functionality of the
nbcon consoles. In particular, it implements the ownership and
priority semantics for nbcon consoles. This series does _not_
include threaded printing, atomic printing regions, or nbcon
drivers. Those features will be added in separate follow-up
series.

The changes since v2:

- rename nbcon_state::hostile_unsafe to
  nbcon_state::unsafe_takeover

- add 2-bit nbcon_state::req_tag to help identify waiters

- remove nbcon_context::unsafe

- remove nbcon_context::hostile

- rename nbcon_context::takeover_unsafe to
  nbcon_context::allow_unsafe_takeover

- callers must now check the current console state to see if it
  acquired the console using the unsafe hostile takeover method

- rename nbcon_write_context::hostile_unsafe to
  nbcon_write_context::unsafe_takeover

- allow direct takeover of owned consoles if higher priority
  and safe

- rename printk_nbcon.c to nbcon.c

- remove most printk.c changes, to be added in a later series
  once all the nbcon functionality is implemented (only init
  and seq updates/checks are kept)

- add nbcon_alloc() to be called early in register_console() to
  allow for simple backout on failure (nbcon_init() changed to
  return void)

- remove my special debugging because it was annoying

- rename __nbcon_seq_to_stored() to __seq_to_nbcon_seq()

- rename __nbcon_seq_to_full() to __nbcon_seq_to_seq()

- implement seq translation as suggested by pmladek

- in nbcon_seq_try_update(), unconditionally attempt cmpxchg to
  allow CONSOLE_REPLAY_ALL to work

- change semantics of all acquire methods to simplify the model

- add nbcon_waiter_matches() to check waiter based on prio and
  req_tag

- add nbcon_context_try_acquire_requested() to implement
  handover part of acquiring after setting request

- in nbcon_context_update_unsafe(), perform handover when
  transitioning to !unsafe and there is a waiter

- in nbcon_emit_next_record(), simplify initializing @wctxt

- in nbcon_emit_next_record(), set the console state to unsafe
  while updating @dropped and @nbcon_seq to avoid data race
  when not in panic

- cleanup comments as suggested by pmladek

John Ogness

[0] https://lore.kernel.org/lkml/20230728000233.50887-1-john.ogness@linutronix.de

[1] https://lore.kernel.org/lkml/20230302195618.156940-6-john.ogness@linutronix.de

Thomas Gleixner (7):
  printk: Add non-BKL (nbcon) console basic infrastructure
  printk: nbcon: Add acquire/release logic
  printk: nbcon: Add buffer management
  printk: nbcon: Add ownership state functions
  printk: nbcon: Add sequence handling
  printk: nbcon: Add emit function and callback function for atomic
    printing
  printk: nbcon: Add functions for drivers to mark unsafe regions

 include/linux/console.h  | 135 ++++++
 kernel/printk/Makefile   |   2 +-
 kernel/printk/internal.h |  29 ++
 kernel/printk/nbcon.c    | 945 +++++++++++++++++++++++++++++++++++++++
 kernel/printk/printk.c   |  60 ++-
 5 files changed, 1152 insertions(+), 19 deletions(-)
 create mode 100644 kernel/printk/nbcon.c


base-commit: cb65d08d735e00cc55ad7700a82a453bb88c93a3
-- 
2.39.2

[PATCH printk v3 1/7] printk: Add non-BKL (nbcon) console basic infrastructure

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

The current console/printk subsystem is protected by a Big Kernel Lock,
(aka console_lock) which has ill defined semantics and is more or less
stateless. This puts severe limitations on the console subsystem and
makes forced takeover and output in emergency and panic situations a
fragile endeavour that is based on try and pray.

The goal of non-BKL (nbcon) consoles is to break out of the console lock
jail and to provide a new infrastructure that avoids the pitfalls and
also allows console drivers to be gradually converted over.

The proposed infrastructure aims for the following properties:

  - Per console locking instead of global locking
  - Per console state that allows to make informed decisions
  - Stateful handover and takeover

As a first step, state is added to struct console. The per console state
is an atomic_t using a 32bit bit field.

Reserve state bits, which will be populated later in the series. Wire
it up into the console register/unregister functionality.

It was decided to use a bitfield because using a plain u32 with
mask/shift operations resulted in uncomprehensible code.

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  | 31 ++++++++++++++++++
 kernel/printk/Makefile   |  2 +-
 kernel/printk/internal.h |  8 +++++
 kernel/printk/nbcon.c    | 70 ++++++++++++++++++++++++++++++++++++++++
 kernel/printk/printk.c   | 13 ++++++--
 5 files changed, 120 insertions(+), 4 deletions(-)
 create mode 100644 kernel/printk/nbcon.c

diff --git a/include/linux/console.h b/include/linux/console.h
index 7de11c763eb3..a2d37a7a98a8 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -156,6 +156,8 @@ static inline int con_debug_leave(void)
  *			/dev/kmesg which requires a larger output buffer.
  * @CON_SUSPENDED:	Indicates if a console is suspended. If true, the
  *			printing callbacks must not be called.
+ * @CON_NBCON:		Console can operate outside of the legacy style console_lock
+ *			constraints.
  */
 enum cons_flags {
 	CON_PRINTBUFFER		= BIT(0),
@@ -166,8 +168,32 @@ enum cons_flags {
 	CON_BRL			= BIT(5),
 	CON_EXTENDED		= BIT(6),
 	CON_SUSPENDED		= BIT(7),
+	CON_NBCON		= BIT(8),
 };
 
+/**
+ * struct nbcon_state - console state for nbcon consoles
+ * @atom:	Compound of the state fields for atomic operations
+ *
+ * To be used for reading and preparing of the value stored in the nbcon
+ * state variable @console::nbcon_state.
+ */
+struct nbcon_state {
+	union {
+		unsigned int	atom;
+		struct {
+		};
+	};
+};
+
+/*
+ * The nbcon_state struct is used to easily create and interpret values that
+ * are stored in the @console::nbcon_state variable. Ensure this struct stays
+ * within the size boundaries of the atomic variable's underlying type in
+ * order to avoid any accidental truncation.
+ */
+static_assert(sizeof(struct nbcon_state) <= sizeof(int));
+
 /**
  * struct console - The console descriptor structure
  * @name:		The name of the console driver
@@ -187,6 +213,8 @@ enum cons_flags {
  * @dropped:		Number of unreported dropped ringbuffer records
  * @data:		Driver private data
  * @node:		hlist node for the console list
+ *
+ * @nbcon_state:	State for nbcon consoles
  */
 struct console {
 	char			name[16];
@@ -206,6 +234,9 @@ struct console {
 	unsigned long		dropped;
 	void			*data;
 	struct hlist_node	node;
+
+	/* nbcon console specific members */
+	atomic_t		__private nbcon_state;
 };
 
 #ifdef CONFIG_LOCKDEP
diff --git a/kernel/printk/Makefile b/kernel/printk/Makefile
index f5b388e810b9..39a2b61c7232 100644
--- a/kernel/printk/Makefile
+++ b/kernel/printk/Makefile
@@ -1,6 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
 obj-y	= printk.o
-obj-$(CONFIG_PRINTK)	+= printk_safe.o
+obj-$(CONFIG_PRINTK)	+= printk_safe.o nbcon.o
 obj-$(CONFIG_A11Y_BRAILLE_CONSOLE)	+= braille.o
 obj-$(CONFIG_PRINTK_INDEX)	+= index.o
 
diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 7d4979d5c3ce..2ca0ab78802c 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -3,6 +3,7 @@
  * internal.h - printk internal definitions
  */
 #include <linux/percpu.h>
+#include <linux/console.h>
 
 #if defined(CONFIG_PRINTK) && defined(CONFIG_SYSCTL)
 void __init printk_sysctl_init(void);
@@ -61,6 +62,10 @@ void defer_console_output(void);
 
 u16 printk_parse_prefix(const char *text, int *level,
 			enum printk_info_flags *flags);
+
+void nbcon_init(struct console *con);
+void nbcon_cleanup(struct console *con);
+
 #else
 
 #define PRINTK_PREFIX_MAX	0
@@ -76,6 +81,9 @@ u16 printk_parse_prefix(const char *text, int *level,
 #define printk_safe_exit_irqrestore(flags) local_irq_restore(flags)
 
 static inline bool printk_percpu_data_ready(void) { return false; }
+static inline void nbcon_init(struct console *con) { }
+static inline void nbcon_cleanup(struct console *con) { }
+
 #endif /* CONFIG_PRINTK */
 
 /**
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
new file mode 100644
index 000000000000..10266d3e7883
--- /dev/null
+++ b/kernel/printk/nbcon.c
@@ -0,0 +1,70 @@
+// SPDX-License-Identifier: GPL-2.0-only
+// Copyright (C) 2022 Linutronix GmbH, John Ogness
+// Copyright (C) 2022 Intel, Thomas Gleixner
+
+#include <linux/kernel.h>
+#include <linux/console.h>
+#include "internal.h"
+/*
+ * Printk console printing implementation for consoles that do not depend on
+ * the legacy style console_lock mechanism.
+ */
+
+/**
+ * nbcon_state_set - Helper function to set the console state
+ * @con:	Console to update
+ * @new:	The new state to write
+ *
+ * Only to be used when the console is not yet or no longer visible in the
+ * system. Otherwise use nbcon_state_try_cmpxchg().
+ */
+static inline void nbcon_state_set(struct console *con, struct nbcon_state *new)
+{
+	atomic_set(&ACCESS_PRIVATE(con, nbcon_state), new->atom);
+}
+
+/**
+ * nbcon_state_read - Helper function to read the console state
+ * @con:	Console to read
+ * @state:	The state to store the result
+ */
+static inline void nbcon_state_read(struct console *con, struct nbcon_state *state)
+{
+	state->atom = atomic_read(&ACCESS_PRIVATE(con, nbcon_state));
+}
+
+/**
+ * nbcon_state_try_cmpxchg() - Helper function for atomic_try_cmpxchg() on console state
+ * @con:	Console to update
+ * @cur:	Old/expected state
+ * @new:	New state
+ *
+ * Return: True on success. False on fail and @cur is updated.
+ */
+static inline bool nbcon_state_try_cmpxchg(struct console *con, struct nbcon_state *cur,
+					   struct nbcon_state *new)
+{
+	return atomic_try_cmpxchg(&ACCESS_PRIVATE(con, nbcon_state), &cur->atom, new->atom);
+}
+
+/**
+ * nbcon_init - Initialize the nbcon console specific data
+ * @con:	Console to initialize
+ */
+void nbcon_init(struct console *con)
+{
+	struct nbcon_state state = { };
+
+	nbcon_state_set(con, &state);
+}
+
+/**
+ * nbcon_cleanup - Cleanup the nbcon console specific data
+ * @con:	Console to cleanup
+ */
+void nbcon_cleanup(struct console *con)
+{
+	struct nbcon_state state = { };
+
+	nbcon_state_set(con, &state);
+}
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 8787d3a72114..c0246093ea41 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3324,9 +3324,10 @@ static void try_enable_default_console(struct console *newcon)
 		newcon->flags |= CON_CONSDEV;
 }
 
-#define con_printk(lvl, con, fmt, ...)			\
-	printk(lvl pr_fmt("%sconsole [%s%d] " fmt),	\
-	       (con->flags & CON_BOOT) ? "boot" : "",	\
+#define con_printk(lvl, con, fmt, ...)				\
+	printk(lvl pr_fmt("%s%sconsole [%s%d] " fmt),		\
+	       (con->flags & CON_NBCON) ? "" : "legacy ",	\
+	       (con->flags & CON_BOOT) ? "boot" : "",		\
 	       con->name, con->index, ##__VA_ARGS__)
 
 static void console_init_seq(struct console *newcon, bool bootcon_registered)
@@ -3486,6 +3487,9 @@ void register_console(struct console *newcon)
 	newcon->dropped = 0;
 	console_init_seq(newcon, bootcon_registered);
 
+	if (newcon->flags & CON_NBCON)
+		nbcon_init(newcon);
+
 	/*
 	 * Put this console in the list - keep the
 	 * preferred driver at the head of the list.
@@ -3577,6 +3581,9 @@ static int unregister_console_locked(struct console *console)
 	 */
 	synchronize_srcu(&console_srcu);
 
+	if (console->flags & CON_NBCON)
+		nbcon_cleanup(console);
+
 	console_sysfs_notify();
 
 	if (console->exit)
-- 
2.39.2

Re: [PATCH printk v3 1/7] printk: Add non-BKL (nbcon) console basic infrastructure

[Petr Mladek]

On Sun 2023-09-03 17:11:33, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> The current console/printk subsystem is protected by a Big Kernel Lock,
> (aka console_lock) which has ill defined semantics and is more or less
> stateless. This puts severe limitations on the console subsystem and
> makes forced takeover and output in emergency and panic situations a
> fragile endeavour that is based on try and pray.
> 
> The goal of non-BKL (nbcon) consoles is to break out of the console lock
> jail and to provide a new infrastructure that avoids the pitfalls and
> also allows console drivers to be gradually converted over.
> 
> The proposed infrastructure aims for the following properties:
> 
>   - Per console locking instead of global locking
>   - Per console state that allows to make informed decisions
>   - Stateful handover and takeover
> 
> As a first step, state is added to struct console. The per console state
> is an atomic_t using a 32bit bit field.
> 
> Reserve state bits, which will be populated later in the series. Wire
> it up into the console register/unregister functionality.
> 
> It was decided to use a bitfield because using a plain u32 with
> mask/shift operations resulted in uncomprehensible code.
> 
> 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>

Looks good to me:

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

Best Regards,
Petr

[PATCH printk v3 2/7] printk: nbcon: Add acquire/release logic

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

Add per console acquire/release functionality. The console 'locked'
state is a combination of multiple state fields:

  - The 'prio' field contains the priority of the context that owns
    the console. This field is used for decisions whether to attempt
    friendly handovers and also prevents takeovers from a less
    severe context, e.g. to protect the panic CPU. A value of 0
    (NBCON_PRIO_NONE) means the console is not locked.

  - The 'cpu' field denotes on which CPU the console is locked.

The acquire mechanism comes with several flavours:

  - Direct acquire when the console is not owned or is owned by a
    lower priority context and the console is in a safe state.

  - Friendly handover mechanism based on a request/grant handshake.
    The requesting context must have a higher priority than the
    current owner.

    The requesting context:

      1) Sets its priority into the 'req_prio' field.

      2) Waits (with a timeout) for the owning context to unlock the
         console.

      3) Sets the 'prio' field and clears the 'req_prio' field.

    The owning context:

      1) Observes the 'req_prio' field set.

      2) Gives up console ownership by clearing the 'prio' field.

  - Unsafe hostile takeover

      The new owner takes the console over without 'req_prio'
      handshake. The new owner must have a higher priority than
      the previous owner.

      This is required when friendly handovers are not possible,
      i.e. the higher priority context interrupted the owning
      context on the same CPU or the owning context is not able
      to make progress on a remote CPU.

The release is the counterpart that either releases the console
directly or yields it gracefully over to a requester.

All operations on @console::nbcon_state are atomic cmpxchg based
to handle concurrency.

The acquire/release functions implement only minimal policies:

  - Preference for higher priority contexts.
  - Protection of the panic CPU.

All other policy decisions have to be made at the call sites:

  - What is marked as an unsafe section.
  - Whether to spin-wait if there is already an owner and the
    console is in an unsafe state.
  - Whether to attempt an unsafe hostile takeover.

The design allows to implement the well known:

    acquire()
    output_one_line()
    release()

algorithm, but also allows to avoid the per line acquire/release for
e.g. panic situations by doing the acquire once and then relying on
the panic CPU protection for the rest.

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 |  62 ++++++
 kernel/printk/nbcon.c   | 458 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 520 insertions(+)

diff --git a/include/linux/console.h b/include/linux/console.h
index a2d37a7a98a8..110e21cf7541 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -175,13 +175,32 @@ enum cons_flags {
  * struct nbcon_state - console state for nbcon consoles
  * @atom:	Compound of the state fields for atomic operations
  *
+ * @req_prio:		The priority of a handover request
+ * @req_tag:		Tagged bits to help identify different requesters
+ *			using the same priority
+ * @prio:		The priority of the current owner
+ * @unsafe:		Console is busy in a non takeover region
+ * @unsafe_takeover:	A hostile takeover in an unsafe state happened in the
+ *			past. The console cannot be safe until re-initialized.
+ * @cpu:		The CPU on which the owner runs
+ *
  * To be used for reading and preparing of the value stored in the nbcon
  * state variable @console::nbcon_state.
+ *
+ * The @prio and @req_prio fields are particularly important to allow
+ * spin-waiting to timeout and give up without the risk of a waiter being
+ * assigned the lock after giving up.
  */
 struct nbcon_state {
 	union {
 		unsigned int	atom;
 		struct {
+			unsigned int prio		:  2;
+			unsigned int req_prio		:  2;
+			unsigned int req_tag		:  2;
+			unsigned int unsafe		:  1;
+			unsigned int unsafe_takeover	:  1;
+			unsigned int cpu		: 24;
 		};
 	};
 };
@@ -194,6 +213,49 @@ struct nbcon_state {
  */
 static_assert(sizeof(struct nbcon_state) <= sizeof(int));
 
+/**
+ * nbcon_prio - console owner priority for nbcon consoles
+ * @NBCON_PRIO_NONE:		Unused
+ * @NBCON_PRIO_NORMAL:		Normal (non-emergency) usage
+ * @NBCON_PRIO_EMERGENCY:	Emergency output (WARN/OOPS...)
+ * @NBCON_PRIO_PANIC:		Panic output
+ * @NBCON_PRIO_MAX:		The number of priority levels
+ *
+ * A context wanting to produce emergency output can carefully takeover the
+ * console, even without consent of the owner. Ideally such a takeover is only
+ * when @nbcon_state::unsafe is not set. However, a context wanting to produce
+ * panic output can ignore the unsafe flag as a last resort. If panic output
+ * is active, no takeover is possible until the panic output releases the
+ * console.
+ */
+enum nbcon_prio {
+	NBCON_PRIO_NONE = 0,
+	NBCON_PRIO_NORMAL,
+	NBCON_PRIO_EMERGENCY,
+	NBCON_PRIO_PANIC,
+	NBCON_PRIO_MAX,
+};
+
+struct console;
+
+/**
+ * struct nbcon_context - Context for console acquire/release
+ * @console:			The associated console
+ * @spinwait_max_us:		Limit for spin-wait acquire
+ * @prio:			Priority of the context
+ * @allow_unsafe_takeover:	Allow performing takeover even if unsafe. Can
+ *				be used only with NBCON_PRIO_PANIC @prio. It
+ *				might cause a system freeze when the console
+ *				is used later.
+ */
+struct nbcon_context {
+	/* members set by caller */
+	struct console		*console;
+	unsigned int		spinwait_max_us;
+	enum nbcon_prio		prio;
+	unsigned int		allow_unsafe_takeover	: 1;
+};
+
 /**
  * struct console - The console descriptor structure
  * @name:		The name of the console driver
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 10266d3e7883..a80567134669 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -4,10 +4,43 @@
 
 #include <linux/kernel.h>
 #include <linux/console.h>
+#include <linux/delay.h>
 #include "internal.h"
 /*
  * Printk console printing implementation for consoles that do not depend on
  * the legacy style console_lock mechanism.
+ *
+ * Console is locked on a CPU when @nbcon_state::prio is set and
+ * @nbcon_state:cpu == current CPU. This is valid for the current execution
+ * context.
+ *
+ * Nesting execution contexts on the same CPU can carefully take over if
+ * the driver allows reentrancy via @nbcon_state::unsafe = false. When the
+ * interrupted context resumes it checks the state before entering an unsafe
+ * region and aborts the operation if it detects a takeover.
+ *
+ * In case of panic the nesting context can take over the console forcefully.
+ *
+ * A concurrent writer on a different CPU with a higher priority can directly
+ * take over if the console is not in an unsafe state by carefully writing
+ * its priority into @nbcon_state::prio.
+ *
+ * If the console is in an unsafe state, the concurrent writer with a higher
+ * priority can request to take over the console by:
+ *
+ *	1) Carefully writing the desired priority into @nbcon_state::req_prio
+ *	   if there is no higher priority request pending.
+ *
+ *	2) Carefully spin on @nbcon_state::prio until it is no longer locked.
+ *
+ *	3) Attempt to lock the console by carefully writing the desired
+ *	   priority to @nbcon_state::prio and carefully removing the desired
+ *	   priority from @nbcon_state::req_prio.
+ *
+ * In case the owner does not react on the request and does not make
+ * observable progress, the waiter will timeout and can then decide to do
+ * an unsafe hostile takeover. Upon unsafe hostile takeover, the console
+ * is kept permanently in the unsafe state.
  */
 
 /**
@@ -47,6 +80,431 @@ static inline bool nbcon_state_try_cmpxchg(struct console *con, struct nbcon_sta
 	return atomic_try_cmpxchg(&ACCESS_PRIVATE(con, nbcon_state), &cur->atom, new->atom);
 }
 
+/**
+ * nbcon_context_try_acquire_direct - Try to acquire directly
+ * @ctxt:	The context of the caller
+ * @cur:	The current console state
+ *
+ * Return:	0 on success and @cur is updated to the new console state.
+ *		Otherwise an error code on failure.
+ *
+ * Errors:
+ *
+ *	-EPERM:		A panic is in progress and this is not the panic CPU
+ *			or this context does not have a priority above the
+ *			current owner or waiter. No acquire method can be
+ *			successful for this context.
+ *
+ *	-EBUSY:		The console is in an unsafe state. The caller should
+ *			try using the handover acquire method.
+ *
+ * The general procedure is to change @nbcon_state::prio from unowned to
+ * owned. Or, if the console is not in the unsafe state, to change
+ * @nbcon_state::prio to a higher priority owner.
+ */
+static int nbcon_context_try_acquire_direct(struct nbcon_context *ctxt,
+					    struct nbcon_state *cur)
+{
+	unsigned int cpu = smp_processor_id();
+	struct console *con = ctxt->console;
+	struct nbcon_state new;
+
+	do {
+		if (other_cpu_in_panic())
+			return -EPERM;
+
+		if (ctxt->prio <= cur->prio || ctxt->prio <= cur->req_prio)
+			return -EPERM;
+
+		if (cur->unsafe)
+			return -EBUSY;
+
+		/*
+		 * Direct acquires should never be attempted if
+		 * an unsafe hostile takeover has ever happened.
+		 */
+		WARN_ON_ONCE(cur->unsafe_takeover);
+
+		new.atom = cur->atom;
+		new.prio	= ctxt->prio;
+		new.req_prio	= NBCON_PRIO_NONE;
+		new.unsafe	= cur->unsafe_takeover;
+		new.cpu		= cpu;
+
+	} while (!nbcon_state_try_cmpxchg(con, cur, &new));
+
+	cur->atom = new.atom;
+
+	return 0;
+}
+
+static bool nbcon_waiter_matches(struct nbcon_state *cur, int expected_tag,
+				int expected_prio)
+{
+	/*
+	 * Since consoles can only be acquired by higher priorities,
+	 * waiting contexts are uniquely identified by @prio. However,
+	 * since owners and waiters can unexpectedly change, it is
+	 * possible that later another waiter appears with the same
+	 * priority. For this reason @req_tag is also needed.
+	 *
+	 * Using the waiting CPU would be better, but there are not enough
+	 * bits in the state variable for this. Since unexpected waiter
+	 * changes are rare and them going unnoticed does not cause fatal
+	 * problems, the tagged bits should be sufficient.
+	 */
+
+	if (cur->req_prio != expected_prio)
+		return false;
+
+	if (cur->req_tag != expected_tag)
+		return false;
+
+	return true;
+}
+
+/**
+ * nbcon_context_try_acquire_requested - Try to acquire after having
+ *					 requested a handover
+ * @ctxt:	The context of the caller
+ * @cur:	The current console state
+ * @req_tag:	The tagged bits used to identify this waiting context
+ *
+ * Return:	0 on success and @cur is updated to the new console state.
+ *		Otherwise an error code on failure.
+ *
+ * Errors:
+ *
+ *	-EPERM:		A panic is in progress and this is not the panic CPU
+ *			or this context is no longer the waiter. For the
+ *			former case, the caller must carefully remove the
+ *			request before aborting the acquire.
+ *
+ *	-EBUSY:		The console is still locked. The caller should
+ *			continue waiting.
+ *
+ * This is a helper function for nbcon_context_try_acquire_handover().
+ *
+ * The use of tagged bits is to partially deal with the situation that while
+ * this waiting context is in udelay(1):
+ *
+ *   1. another context with higher priority directly takes ownership
+ *   2. the higher priority context releases ownership
+ *   3. a lower priority context takes ownership
+ *   4. a context with the same priority as this context requests ownership
+ *   5. this waiting context finishes udelay(1) and thinks it is the waiter
+ *
+ * To address this rare situation, tagged bits are used so that the waiter
+ * can better identify if it is really the waiter. In the above  example, the
+ * original waiter would use a @req_tag value of 1, whereas the follow-up
+ * waiter would use a @req_tag value of 2. This allows the original waiter to
+ * identify in step 5 that it has been replaced by another waiter.
+ */
+static int nbcon_context_try_acquire_requested(struct nbcon_context *ctxt,
+					       struct nbcon_state *cur,
+					       char req_tag)
+{
+	unsigned int cpu = smp_processor_id();
+	struct console *con = ctxt->console;
+	struct nbcon_state new;
+
+	do {
+		/*
+		 * Note: If the acquire is aborted due to a panic CPU,
+		 * the caller must still remove the request!
+		 */
+		if (other_cpu_in_panic())
+			return -EPERM;
+
+		/*
+		 * If an unsafe hostile takeover has occurred, a handover
+		 * is no longer possible.
+		 */
+		if (cur->unsafe_takeover)
+			return -EPERM;
+
+		/* Is this context still the requester? */
+		if (!nbcon_waiter_matches(cur, req_tag, ctxt->prio))
+			return -EPERM;
+
+		/* If still locked, caller should continue waiting. */
+		if (cur->prio != NBCON_PRIO_NONE)
+			return -EBUSY;
+
+		/* Handover acquires should never be attempted if unsafe. */
+		WARN_ON_ONCE(cur->unsafe);
+
+		new.atom = cur->atom;
+		new.prio	= ctxt->prio;
+		new.req_prio	= NBCON_PRIO_NONE;
+		new.unsafe	= cur->unsafe_takeover;
+		new.cpu		= cpu;
+
+	} while (!nbcon_state_try_cmpxchg(con, cur, &new));
+
+	/* Handover success. This context now owns the console. */
+
+	cur->atom = new.atom;
+
+	return 0;
+}
+
+/**
+ * nbcon_context_try_acquire_handover - Try to acquire via handover
+ * @ctxt:	The context of the caller
+ * @cur:	The current console state
+ *
+ * Return:	0 on success and @cur is updated to the new console state.
+ *		Otherwise an error code on failure.
+ *
+ * Errors:
+ *
+ *	-EPERM:		This context is on the same CPU as the current owner
+ *			or the console is permanently in an unsafe state or
+ *			this context is unwilling to wait or a panic is in
+ *			progress and this is not the panic CPU. This is not
+ *			the panic context, so no acquire method can be
+ *			successful for this context.
+ *
+ *	-EBUSY:		The current owner or waiter is such that this context
+ *			is not able to execute a handover. The caller can use
+ *			an unsafe hostile takeover to acquire.
+ *
+ *	-EAGAIN:	@cur is not the current console state. @cur has been
+ *			updated. The caller should retry with direct acquire.
+ *
+ * The general procedure is to set @req_prio and wait until unowned. Then
+ * set @prio (claiming ownership) and clearing @req_prio.
+ *
+ * Note that it is expected that the caller tried
+ * nbcon_context_try_acquire_direct() with @cur before calling this function.
+ */
+static int nbcon_context_try_acquire_handover(struct nbcon_context *ctxt,
+					      struct nbcon_state *cur)
+{
+	unsigned int cpu = smp_processor_id();
+	struct console *con = ctxt->console;
+	struct nbcon_state new;
+	int timeout;
+	int req_tag;
+	int err = 0;
+
+	/*
+	 * Handovers are not possible on the same CPU or when the console
+	 * is permanently in an unsafe state or if the caller is unwilling
+	 * to wait.
+	 */
+	if (cur->cpu == cpu ||
+	    cur->unsafe_takeover ||
+	    ctxt->spinwait_max_us == 0) {
+		goto fail_handover;
+	}
+
+	/*
+	 * Setup a request for handover.
+	 *
+	 * @nbcon_state::req_tag is modified for each handover request. See
+	 * the function description of nbcon_context_try_acquire_requested()
+	 * for details about how @req_tag is used.
+	 */
+	new.atom = cur->atom;
+	new.req_prio = ctxt->prio;
+	new.req_tag++;
+	req_tag = new.req_tag;
+	if (!nbcon_state_try_cmpxchg(con, cur, &new))
+		return -EAGAIN;
+
+	cur->atom = new.atom;
+
+	/* Wait until there is no owner and then acquire directly. */
+	for (timeout = ctxt->spinwait_max_us; timeout >= 0; timeout--) {
+		/* On successful acquire, this request is cleared. */
+		err = nbcon_context_try_acquire_requested(ctxt, cur, req_tag);
+		if (!err)
+			return 0;
+
+		/*
+		 * If the acquire should be aborted, it must be ensured
+		 * that the request is removed before returning to caller.
+		 */
+		if (err == -EPERM)
+			break;
+
+		udelay(1);
+
+		/* Re-read the state because some time has passed. */
+		nbcon_state_read(con, cur);
+	}
+
+	/* Timed out or should abort. Carefully remove handover request. */
+	do {
+		/* No need to remove request if there is a new waiter. */
+		if (!nbcon_waiter_matches(cur, req_tag, ctxt->prio))
+			goto fail_handover;
+
+		/* Unset request for handover. */
+		new.atom = cur->atom;
+		new.req_prio = NBCON_PRIO_NONE;
+		if (nbcon_state_try_cmpxchg(con, cur, &new)) {
+			/*
+			 * Request successfully unset. Report failure of
+			 * acquiring via handover.
+			 */
+			cur->atom = new.atom;
+			goto fail_handover;
+		}
+
+		/*
+		 * Unable to remove request. Try to acquire in case
+		 * the owner has released the lock.
+		 */
+	} while (nbcon_context_try_acquire_requested(ctxt, cur, req_tag));
+
+	/* Acquire at timeout succeeded! */
+	return 0;
+
+fail_handover:
+	/*
+	 * If this is the panic context, the caller can try to acquire using
+	 * the unsafe hostile takeover method.
+	 */
+	if (ctxt->prio == NBCON_PRIO_PANIC &&
+	    ctxt->prio > cur->prio &&
+	    ctxt->prio > cur->req_prio &&
+	    !other_cpu_in_panic()) {
+		return -EBUSY;
+	}
+	return -EPERM;
+}
+
+/**
+ * nbcon_context_try_acquire_hostile - Acquire via unsafe hostile takeover
+ * @ctxt:	The context of the caller
+ * @cur:	The current console state
+ *
+ * @cur is updated to the new console state.
+ *
+ * The general procedure is to set @prio (forcing ownership). This method
+ * must only be used as a final attempt during panic.
+ */
+static void nbcon_context_acquire_hostile(struct nbcon_context *ctxt,
+					  struct nbcon_state *cur)
+{
+	unsigned int cpu = smp_processor_id();
+	struct console *con = ctxt->console;
+	struct nbcon_state new;
+
+	do {
+		new.atom = cur->atom;
+		new.cpu			= cpu;
+		new.prio		= ctxt->prio;
+		new.unsafe		|= cur->unsafe_takeover;
+		new.unsafe_takeover	|= cur->unsafe;
+
+	} while (!nbcon_state_try_cmpxchg(con, cur, &new));
+
+	cur->atom = new.atom;
+}
+
+/**
+ * nbcon_context_try_acquire - Try to acquire nbcon console
+ * @ctxt:	The context of the caller
+ *
+ * Return:	True if the console was acquired. False otherwise.
+ *
+ * If the caller allowed an unsafe hostile takeover, on success the
+ * caller should check the current console state to see if it is
+ * 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)
+{
+	__maybe_unused
+	unsigned int cpu = smp_processor_id();
+	struct console *con = ctxt->console;
+	struct nbcon_state cur;
+	int err;
+
+	/* unsafe hostile takeovers are only allowed for panic */
+	WARN_ON_ONCE(ctxt->allow_unsafe_takeover && (ctxt->prio != NBCON_PRIO_PANIC));
+
+	nbcon_state_read(con, &cur);
+
+	do {
+		err = nbcon_context_try_acquire_direct(ctxt, &cur);
+		if (!err)
+			goto success;
+		else if (err == -EPERM)
+			return false;
+
+		err = nbcon_context_try_acquire_handover(ctxt, &cur);
+		if (!err)
+			goto success;
+		else if (err == -EPERM)
+			return false;
+
+	} while (err == -EAGAIN);
+
+	/* Only attempt unsafe hostile takeover if explicitly requested. */
+	if (!ctxt->allow_unsafe_takeover)
+		return false;
+
+	nbcon_context_acquire_hostile(ctxt, &cur);
+success:
+	return true;
+}
+
+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.
+	 */
+
+	if (cur->prio != expected_prio)
+		return false;
+
+	if (cur->cpu != expected_cpu)
+		return false;
+
+	return true;
+}
+
+/**
+ * nbcon_context_release - Release the console
+ * @ctxt:	The nbcon context from nbcon_context_try_acquire()
+ */
+__maybe_unused
+static void nbcon_context_release(struct nbcon_context *ctxt)
+{
+	unsigned int cpu = smp_processor_id();
+	struct console *con = ctxt->console;
+	struct nbcon_state cur;
+	struct nbcon_state new;
+
+	nbcon_state_read(con, &cur);
+
+	do {
+		if (!nbcon_owner_matches(&cur, cpu, ctxt->prio))
+			return;
+
+		new.atom = cur.atom;
+		new.prio = NBCON_PRIO_NONE;
+
+		/*
+		 * If @unsafe_takeover is set, it is kept set so that
+		 * the state remains permanently unsafe.
+		 */
+		new.unsafe |= cur.unsafe_takeover;
+
+	} while (!nbcon_state_try_cmpxchg(con, &cur, &new));
+}
+
 /**
  * nbcon_init - Initialize the nbcon console specific data
  * @con:	Console to initialize
-- 
2.39.2

Re: [PATCH printk v3 2/7] printk: nbcon: Add acquire/release logic

[kernel test robot]

Hi John,

kernel test robot noticed the following build warnings:

[auto build test WARNING on cb65d08d735e00cc55ad7700a82a453bb88c93a3]

url:    https://github.com/intel-lab-lkp/linux/commits/John-Ogness/printk-Add-non-BKL-nbcon-console-basic-infrastructure/20230903-230708
base:   cb65d08d735e00cc55ad7700a82a453bb88c93a3
patch link:    https://lore.kernel.org/r/20230903150539.245076-3-john.ogness%40linutronix.de
patch subject: [PATCH printk v3 2/7] printk: nbcon: Add acquire/release logic
config: x86_64-defconfig (https://download.01.org/0day-ci/archive/20230904/202309040038.Cn1kd5bN-lkp@intel.com/config)
compiler: gcc-11 (Debian 11.3.0-12) 11.3.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20230904/202309040038.Cn1kd5bN-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202309040038.Cn1kd5bN-lkp@intel.com/

All warnings (new ones prefixed by >>):

>> kernel/printk/nbcon.c:392: warning: expecting prototype for nbcon_context_try_acquire_hostile(). Prototype was for nbcon_context_acquire_hostile() instead


vim +392 kernel/printk/nbcon.c

   379	
   380	/**
   381	 * nbcon_context_try_acquire_hostile - Acquire via unsafe hostile takeover
   382	 * @ctxt:	The context of the caller
   383	 * @cur:	The current console state
   384	 *
   385	 * @cur is updated to the new console state.
   386	 *
   387	 * The general procedure is to set @prio (forcing ownership). This method
   388	 * must only be used as a final attempt during panic.
   389	 */
   390	static void nbcon_context_acquire_hostile(struct nbcon_context *ctxt,
   391						  struct nbcon_state *cur)
 > 392	{
   393		unsigned int cpu = smp_processor_id();
   394		struct console *con = ctxt->console;
   395		struct nbcon_state new;
   396	
   397		do {
   398			new.atom = cur->atom;
   399			new.cpu			= cpu;
   400			new.prio		= ctxt->prio;
   401			new.unsafe		|= cur->unsafe_takeover;
   402			new.unsafe_takeover	|= cur->unsafe;
   403	
   404		} while (!nbcon_state_try_cmpxchg(con, cur, &new));
   405	
   406		cur->atom = new.atom;
   407	}
   408	

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH printk v3 2/7] printk: nbcon: Add acquire/release logic

[John Ogness]

On 2023-09-04, kernel test robot <lkp@intel.com> wrote:
>>> kernel/printk/nbcon.c:392: warning: expecting prototype for nbcon_context_try_acquire_hostile(). Prototype was for nbcon_context_acquire_hostile() instead

Thanks kernel test robot. The kerneldoc will be fixed for v4.

The function was renamed for v3 because it now always succeeds. There is
no need for "try" in the name.

John Ogness

Re: [PATCH printk v3 2/7] printk: nbcon: Add acquire/release logic

[kernel test robot]

Hi John,

kernel test robot noticed the following build warnings:

[auto build test WARNING on cb65d08d735e00cc55ad7700a82a453bb88c93a3]

url:    https://github.com/intel-lab-lkp/linux/commits/John-Ogness/printk-Add-non-BKL-nbcon-console-basic-infrastructure/20230903-230708
base:   cb65d08d735e00cc55ad7700a82a453bb88c93a3
patch link:    https://lore.kernel.org/r/20230903150539.245076-3-john.ogness%40linutronix.de
patch subject: [PATCH printk v3 2/7] printk: nbcon: Add acquire/release logic
config: riscv-allnoconfig (https://download.01.org/0day-ci/archive/20230904/202309040026.yyn4RCk1-lkp@intel.com/config)
compiler: riscv64-linux-gcc (GCC) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20230904/202309040026.yyn4RCk1-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202309040026.yyn4RCk1-lkp@intel.com/

All warnings (new ones prefixed by >>):

>> kernel/printk/nbcon.c:392: warning: expecting prototype for nbcon_context_try_acquire_hostile(). Prototype was for nbcon_context_acquire_hostile() instead


vim +392 kernel/printk/nbcon.c

   379	
   380	/**
   381	 * nbcon_context_try_acquire_hostile - Acquire via unsafe hostile takeover
   382	 * @ctxt:	The context of the caller
   383	 * @cur:	The current console state
   384	 *
   385	 * @cur is updated to the new console state.
   386	 *
   387	 * The general procedure is to set @prio (forcing ownership). This method
   388	 * must only be used as a final attempt during panic.
   389	 */
   390	static void nbcon_context_acquire_hostile(struct nbcon_context *ctxt,
   391						  struct nbcon_state *cur)
 > 392	{
   393		unsigned int cpu = smp_processor_id();
   394		struct console *con = ctxt->console;
   395		struct nbcon_state new;
   396	
   397		do {
   398			new.atom = cur->atom;
   399			new.cpu			= cpu;
   400			new.prio		= ctxt->prio;
   401			new.unsafe		|= cur->unsafe_takeover;
   402			new.unsafe_takeover	|= cur->unsafe;
   403	
   404		} while (!nbcon_state_try_cmpxchg(con, cur, &new));
   405	
   406		cur->atom = new.atom;
   407	}
   408	

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH printk v3 2/7] printk: nbcon: Add acquire/release logic

[Petr Mladek]

On Sun 2023-09-03 17:11:34, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> Add per console acquire/release functionality. The console 'locked'
> state is a combination of multiple state fields:
> 
> --- a/kernel/printk/nbcon.c
> +++ b/kernel/printk/nbcon.c
> @@ -47,6 +80,431 @@ static inline bool nbcon_state_try_cmpxchg(struct console *con, struct nbcon_sta
>  	return atomic_try_cmpxchg(&ACCESS_PRIVATE(con, nbcon_state), &cur->atom, new->atom);
>  }
>  
> +/**
> + * nbcon_context_try_acquire_direct - Try to acquire directly
> + * @ctxt:	The context of the caller
> + * @cur:	The current console state
> + *
> + * Return:	0 on success and @cur is updated to the new console state.
> + *		Otherwise an error code on failure.
> + *
> + * Errors:
> + *
> + *	-EPERM:		A panic is in progress and this is not the panic CPU
> + *			or this context does not have a priority above the
> + *			current owner or waiter. No acquire method can be
> + *			successful for this context.
> + *
> + *	-EBUSY:		The console is in an unsafe state. The caller should
> + *			try using the handover acquire method.
> + *
> + * The general procedure is to change @nbcon_state::prio from unowned to
> + * owned. Or, if the console is not in the unsafe state, to change
> + * @nbcon_state::prio to a higher priority owner.
> + */
> +static int nbcon_context_try_acquire_direct(struct nbcon_context *ctxt,
> +					    struct nbcon_state *cur)
> +{
> +	unsigned int cpu = smp_processor_id();
> +	struct console *con = ctxt->console;
> +	struct nbcon_state new;
> +
> +	do {
> +		if (other_cpu_in_panic())
> +			return -EPERM;
> +
> +		if (ctxt->prio <= cur->prio || ctxt->prio <= cur->req_prio)
> +			return -EPERM;
> +
> +		if (cur->unsafe)
> +			return -EBUSY;
> +
> +		/*
> +		 * Direct acquires should never be attempted if
> +		 * an unsafe hostile takeover has ever happened.
> +		 */
> +		WARN_ON_ONCE(cur->unsafe_takeover);

I was a bit confused by this. My first thought was that
this function should never be called after hostile takeover.
And it did not make sense.

But it means that we should never be here if an unsafe hostile
takeover happened. I would slight change the comment:

		/*
		 * The console should never be safe for a direct acquire
		 * if an unsafe hostile takeover has ever happened.
		 */
		WARN_ON_ONCE(cur->unsafe_takeover);

> +		new.atom = cur->atom;
> +		new.prio	= ctxt->prio;
> +		new.req_prio	= NBCON_PRIO_NONE;
> +		new.unsafe	= cur->unsafe_takeover;
> +		new.cpu		= cpu;
> +
> +	} while (!nbcon_state_try_cmpxchg(con, cur, &new));
> +
> +	cur->atom = new.atom;
> +
> +	return 0;
> +}
> +
> +static bool nbcon_waiter_matches(struct nbcon_state *cur, int expected_tag,
> +				int expected_prio)
> +{
> +	/*
> +	 * Since consoles can only be acquired by higher priorities,
> +	 * waiting contexts are uniquely identified by @prio. However,
> +	 * since owners and waiters can unexpectedly change, it is
> +	 * possible that later another waiter appears with the same
> +	 * priority. For this reason @req_tag is also needed.
> +	 *
> +	 * Using the waiting CPU would be better, but there are not enough
> +	 * bits in the state variable for this. Since unexpected waiter
> +	 * changes are rare and them going unnoticed does not cause fatal
> +	 * problems, the tagged bits should be sufficient.
> +	 */
> +
> +	if (cur->req_prio != expected_prio)
> +		return false;
> +
> +	if (cur->req_tag != expected_tag)
> +		return false;
> +
> +	return true;
> +}
> +
> +/**
> + * nbcon_context_try_acquire_requested - Try to acquire after having
> + *					 requested a handover
> + * @ctxt:	The context of the caller
> + * @cur:	The current console state
> + * @req_tag:	The tagged bits used to identify this waiting context

I would suggest to rename @req_tag to @req_cnt or @req_wait_cnt
and describe it:

 * @req_wait_cnt:	Counter to distinguish waiting contexts of
 *			the same priority.

Later update: I would actually remove the tag completely, see below.

> + * Return:	0 on success and @cur is updated to the new console state.
> + *		Otherwise an error code on failure.
> + *
> + * Errors:
> + *
> + *	-EPERM:		A panic is in progress and this is not the panic CPU
> + *			or this context is no longer the waiter. For the
> + *			former case, the caller must carefully remove the
> + *			request before aborting the acquire.
> + *
> + *	-EBUSY:		The console is still locked. The caller should
> + *			continue waiting.
> + *
> + * This is a helper function for nbcon_context_try_acquire_handover().
> + *
> + * The use of tagged bits is to partially deal with the situation that while
> + * this waiting context is in udelay(1):
> + *
> + *   1. another context with higher priority directly takes ownership
> + *   2. the higher priority context releases ownership
> + *   3. a lower priority context takes ownership
> + *   4. a context with the same priority as this context requests ownership
> + *   5. this waiting context finishes udelay(1) and thinks it is the waiter

I was a bit confused by the above description. First, I though that 1, 2,
3, 4 would match the req_tag number.

> + *
> + * To address this rare situation, tagged bits are used so that the waiter
> + * can better identify if it is really the waiter. In the above  example, the
> + * original waiter would use a @req_tag value of 1, whereas the follow-up
> + * waiter would use a @req_tag value of 2. This allows the original waiter to
> + * identify in step 5 that it has been replaced by another waiter.
> + */

I thought about how to avoid the confusion and update the description
of the @req_tag handling. And I am pretty sure that the race could
not happen in the current implementation.

I suggest to remove @req_tag completely and describe the problem
above nbcon_waiter_matches() implementation. Something like:


/**
 * The request owner is well defined by the @req_prio because
 *
 *   1. Only a context with a higher priority could take over the request.
 *   2. There are only three priorities.
 *   3. Only one CPU is allowed to request PANIC priority.
 *   4. 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
 *	creates a request and starts waiting.
 */
static bool nbcon_waiter_matches(struct nbcon_state *cur,
				 int expected_prio)
{
	return cur->req_prio == expected_prio;
}

> +
> +	if (cur->req_tag != expected_tag)
> +		return false;
> +
> +	return true;
> +}
> +
> +static int nbcon_context_try_acquire_requested(struct nbcon_context *ctxt,
> +					       struct nbcon_state *cur,
> +					       char req_tag)
> +{
> +	unsigned int cpu = smp_processor_id();
> +	struct console *con = ctxt->console;
> +	struct nbcon_state new;
> +
> +	do {
> +		/*
> +		 * Note: If the acquire is aborted due to a panic CPU,
> +		 * the caller must still remove the request!
> +		 */
> +		if (other_cpu_in_panic())
> +			return -EPERM;
> +
> +		/*
> +		 * If an unsafe hostile takeover has occurred, a handover
> +		 * is no longer possible.
> +		 */
> +		if (cur->unsafe_takeover)
> +			return -EPERM;
> +
> +		/* Is this context still the requester? */
> +		if (!nbcon_waiter_matches(cur, req_tag, ctxt->prio))
> +			return -EPERM;
> +
> +		/* If still locked, caller should continue waiting. */
> +		if (cur->prio != NBCON_PRIO_NONE)
> +			return -EBUSY;
> +
> +		/* Handover acquires should never be attempted if unsafe. */
> +		WARN_ON_ONCE(cur->unsafe);

IMHO, it is not completely clear how it is guranteed that cur->unsafe
could never be set at this point. It works because:

  1. The previous owner would never release the lock in unsafe region.

  2. cur->unsafe might be set also because of cur->unsafe_takeover
     but this is checked above.

I would personally use the following comment:

		/*
		 * The lock should have never been released in an unsafe
		 * region.
		 */
		WARN_ON_ONCE(cur->unsafe);

> +
> +		new.atom = cur->atom;
> +		new.prio	= ctxt->prio;
> +		new.req_prio	= NBCON_PRIO_NONE;
> +		new.unsafe	= cur->unsafe_takeover;
> +		new.cpu		= cpu;
> +
> +	} while (!nbcon_state_try_cmpxchg(con, cur, &new));
> +
> +	/* Handover success. This context now owns the console. */
> +
> +	cur->atom = new.atom;
> +
> +	return 0;
> +}

Otherwise, it looks good to me.

Best Regards,
Petr

[PATCH printk v3 3/7] printk: nbcon: Add buffer management

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

In case of hostile takeovers it must be ensured that the previous
owner cannot scribble over the output buffer of the emergency/panic
context. This is achieved by:

 - Adding a global output buffer instance for the panic context.
   This is the only situation where hostile takeovers can occur and
   there is always at most 1 panic context.

 - Allocating an output buffer per console upon console
   registration. This buffer is used by the console owner when not
   in panic context.

 - Choosing the appropriate buffer is handled in the acquire/release
   functions.

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  |  7 +++++++
 kernel/printk/internal.h |  8 ++++++++
 kernel/printk/nbcon.c    | 44 ++++++++++++++++++++++++++++++++++++++--
 kernel/printk/printk.c   | 15 ++++++++------
 4 files changed, 66 insertions(+), 8 deletions(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index 110e21cf7541..d0bf9abd76b8 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -237,6 +237,7 @@ enum nbcon_prio {
 };
 
 struct console;
+struct printk_buffers;
 
 /**
  * struct nbcon_context - Context for console acquire/release
@@ -247,6 +248,7 @@ struct console;
  *				be used only with NBCON_PRIO_PANIC @prio. It
  *				might cause a system freeze when the console
  *				is used later.
+ * @pbufs:			Pointer to the text buffer for this context
  */
 struct nbcon_context {
 	/* members set by caller */
@@ -254,6 +256,9 @@ struct nbcon_context {
 	unsigned int		spinwait_max_us;
 	enum nbcon_prio		prio;
 	unsigned int		allow_unsafe_takeover	: 1;
+
+	/* members set by acquire */
+	struct printk_buffers	*pbufs;
 };
 
 /**
@@ -277,6 +282,7 @@ struct nbcon_context {
  * @node:		hlist node for the console list
  *
  * @nbcon_state:	State for nbcon consoles
+ * @pbufs:		Pointer to nbcon private buffer
  */
 struct console {
 	char			name[16];
@@ -299,6 +305,7 @@ struct console {
 
 	/* nbcon console specific members */
 	atomic_t		__private nbcon_state;
+	struct printk_buffers	*pbufs;
 };
 
 #ifdef CONFIG_LOCKDEP
diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 2ca0ab78802c..cb4ea90f04a5 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -13,6 +13,12 @@ int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write,
 #define printk_sysctl_init() do { } while (0)
 #endif
 
+#define con_printk(lvl, con, fmt, ...)				\
+	printk(lvl pr_fmt("%s%sconsole [%s%d] " fmt),		\
+		(con->flags & CON_NBCON) ? "" : "legacy ",	\
+		(con->flags & CON_BOOT) ? "boot" : "",		\
+		con->name, con->index, ##__VA_ARGS__)
+
 #ifdef CONFIG_PRINTK
 
 #ifdef CONFIG_PRINTK_CALLER
@@ -63,6 +69,7 @@ void defer_console_output(void);
 u16 printk_parse_prefix(const char *text, int *level,
 			enum printk_info_flags *flags);
 
+bool nbcon_alloc(struct console *con);
 void nbcon_init(struct console *con);
 void nbcon_cleanup(struct console *con);
 
@@ -81,6 +88,7 @@ void nbcon_cleanup(struct console *con);
 #define printk_safe_exit_irqrestore(flags) local_irq_restore(flags)
 
 static inline bool printk_percpu_data_ready(void) { return false; }
+static inline bool nbcon_alloc(struct console *con) { return false; }
 static inline void nbcon_init(struct console *con) { }
 static inline void nbcon_cleanup(struct console *con) { }
 
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index a80567134669..c148467b39b8 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -5,6 +5,7 @@
 #include <linux/kernel.h>
 #include <linux/console.h>
 #include <linux/delay.h>
+#include <linux/slab.h>
 #include "internal.h"
 /*
  * Printk console printing implementation for consoles that do not depend on
@@ -20,6 +21,9 @@
  * region and aborts the operation if it detects a takeover.
  *
  * In case of panic the nesting context can take over the console forcefully.
+ * If the interrupted context touches the assigned record buffer after
+ * takeover, it does not cause harm because the interrupting single panic
+ * context is assigned its own panic record buffer.
  *
  * A concurrent writer on a different CPU with a higher priority can directly
  * take over if the console is not in an unsafe state by carefully writing
@@ -406,6 +410,8 @@ static void nbcon_context_acquire_hostile(struct nbcon_context *ctxt,
 	cur->atom = new.atom;
 }
 
+static struct printk_buffers panic_nbcon_pbufs;
+
 /**
  * nbcon_context_try_acquire - Try to acquire nbcon console
  * @ctxt:	The context of the caller
@@ -420,7 +426,6 @@ static void nbcon_context_acquire_hostile(struct nbcon_context *ctxt,
 __maybe_unused
 static bool nbcon_context_try_acquire(struct nbcon_context *ctxt)
 {
-	__maybe_unused
 	unsigned int cpu = smp_processor_id();
 	struct console *con = ctxt->console;
 	struct nbcon_state cur;
@@ -452,6 +457,12 @@ static bool nbcon_context_try_acquire(struct nbcon_context *ctxt)
 
 	nbcon_context_acquire_hostile(ctxt, &cur);
 success:
+	/* Assign the appropriate buffer for this context. */
+	if (atomic_read(&panic_cpu) == cpu)
+		ctxt->pbufs = &panic_nbcon_pbufs;
+	else
+		ctxt->pbufs = con->pbufs;
+
 	return true;
 }
 
@@ -491,7 +502,7 @@ static void nbcon_context_release(struct nbcon_context *ctxt)
 
 	do {
 		if (!nbcon_owner_matches(&cur, cpu, ctxt->prio))
-			return;
+			break;
 
 		new.atom = cur.atom;
 		new.prio = NBCON_PRIO_NONE;
@@ -503,6 +514,30 @@ static void nbcon_context_release(struct nbcon_context *ctxt)
 		new.unsafe |= cur.unsafe_takeover;
 
 	} while (!nbcon_state_try_cmpxchg(con, &cur, &new));
+
+	ctxt->pbufs = NULL;
+}
+
+/**
+ * nbcon_alloc - Allocate buffers needed by the nbcon console
+ * @con:	Console to initialize
+ *
+ * Return:	True on success. False otherwise and the console cannot
+ *		be used.
+ *
+ * This is not part of nbcon_init() because buffer allocation must
+ * be performed earlier in the console registration process.
+ */
+bool nbcon_alloc(struct console *con)
+{
+
+	con->pbufs = kmalloc(sizeof(*con->pbufs), GFP_KERNEL);
+	if (!con->pbufs) {
+		con_printk(KERN_ERR, con, "failed to allocate printing buffer\n");
+		return false;
+	}
+
+	return true;
 }
 
 /**
@@ -513,6 +548,9 @@ void nbcon_init(struct console *con)
 {
 	struct nbcon_state state = { };
 
+	/* nbcon_alloc() must have been called and successful! */
+	BUG_ON(!con->pbufs);
+
 	nbcon_state_set(con, &state);
 }
 
@@ -525,4 +563,6 @@ void nbcon_cleanup(struct console *con)
 	struct nbcon_state state = { };
 
 	nbcon_state_set(con, &state);
+	kfree(con->pbufs);
+	con->pbufs = NULL;
 }
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index c0246093ea41..b340c5a36fe2 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3324,12 +3324,6 @@ static void try_enable_default_console(struct console *newcon)
 		newcon->flags |= CON_CONSDEV;
 }
 
-#define con_printk(lvl, con, fmt, ...)				\
-	printk(lvl pr_fmt("%s%sconsole [%s%d] " fmt),		\
-	       (con->flags & CON_NBCON) ? "" : "legacy ",	\
-	       (con->flags & CON_BOOT) ? "boot" : "",		\
-	       con->name, con->index, ##__VA_ARGS__)
-
 static void console_init_seq(struct console *newcon, bool bootcon_registered)
 {
 	struct console *con;
@@ -3443,6 +3437,15 @@ void register_console(struct console *newcon)
 		goto unlock;
 	}
 
+	if (newcon->flags & CON_NBCON) {
+		/*
+		 * Ensure the nbcon console buffers can be allocated
+		 * before modifying any global data.
+		 */
+		if (!nbcon_alloc(newcon))
+			goto unlock;
+	}
+
 	/*
 	 * See if we want to enable this console driver by default.
 	 *
-- 
2.39.2

Re: [PATCH printk v3 3/7] printk: nbcon: Add buffer management

[Petr Mladek]

On Sun 2023-09-03 17:11:35, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> In case of hostile takeovers it must be ensured that the previous
> owner cannot scribble over the output buffer of the emergency/panic
> context. This is achieved by:
> 
>  - Adding a global output buffer instance for the panic context.
>    This is the only situation where hostile takeovers can occur and
>    there is always at most 1 panic context.
> 
>  - Allocating an output buffer per console upon console
>    registration. This buffer is used by the console owner when not
>    in panic context.
> 
>  - Choosing the appropriate buffer is handled in the acquire/release
>    functions.
> 
> --- a/kernel/printk/nbcon.c
> +++ b/kernel/printk/nbcon.c
> @@ -503,6 +514,30 @@ static void nbcon_context_release(struct nbcon_context *ctxt)
>  		new.unsafe |= cur.unsafe_takeover;
>  
>  	} while (!nbcon_state_try_cmpxchg(con, &cur, &new));
> +
> +	ctxt->pbufs = NULL;
> +}
> +
> +/**
> + * nbcon_alloc - Allocate buffers needed by the nbcon console
> + * @con:	Console to initialize
> + *
> + * Return:	True on success. False otherwise and the console cannot
> + *		be used.
> + *
> + * This is not part of nbcon_init() because buffer allocation must
> + * be performed earlier in the console registration process.
> + */
> +bool nbcon_alloc(struct console *con)
> +{
> +
> +	con->pbufs = kmalloc(sizeof(*con->pbufs), GFP_KERNEL);

We might need to use memblock_alloc() at least for early consoles.

mm_core_init() is called after processing the kernel parameters.

For example, setup_log_buf() or vfs_caches_init_early() use
memblock_alloc() as well.

> +	if (!con->pbufs) {
> +		con_printk(KERN_ERR, con, "failed to allocate printing buffer\n");
> +		return false;
> +	}
> +
> +	return true;
>  }
>  
>  /**
> @@ -525,4 +563,6 @@ void nbcon_cleanup(struct console *con)
>  	struct nbcon_state state = { };
>  
>  	nbcon_state_set(con, &state);
> +	kfree(con->pbufs);
> +	con->pbufs = NULL;

It would be cleaner to create nbcon_free() as a counter part
for nbcon_alloc().

>  }

Best Regards,
Petr

Re: [PATCH printk v3 3/7] printk: nbcon: Add buffer management

[John Ogness]

On 2023-09-06, Petr Mladek <pmladek@suse.com> wrote:
>> +bool nbcon_alloc(struct console *con)
>> +{
>> +
>> +	con->pbufs = kmalloc(sizeof(*con->pbufs), GFP_KERNEL);
>
> We might need to use memblock_alloc() at least for early consoles.

I wasn't planning on addressing early consoles at this stage. The
initial 8250 NBCON implementation will only be for the regular console.

However, since it is planned that NBCON consoles are synchronized with
boot consoles using the console_lock, we might as well establish that
boot consoles (in general) are always synchronized uisng the
console_lock.

This allows us to use the single global pbufs of the legacy consoles
(see console_emit_next_record()) for NBCON boot consoles as well. The
static buffer is much more attractive because it is always available.

For v4, nbcon_alloc() will initialize con->pbufs to use the global
legacy pbufs for con->flags == CON_BOOT.

bool nbcon_alloc(struct console *con)
{
        if (con->flags & CON_BOOT) {
                /*
                 * Boot console printing is synchronized with legacy console
                 * printing, so boot consoles can share the same global printk
                 * buffers.
                 */
                con->pbufs = &printk_shared_pbufs;
        } else {
                con->pbufs = kmalloc(sizeof(*con->pbufs), GFP_KERNEL);
                if (!con->pbufs) {
                        con_printk(KERN_ERR, con, "failed to allocate printing buffer\n");
                        return false;
                }
        }

        return true;
}

John Ogness

[PATCH printk v3 4/7] printk: nbcon: Add ownership state functions

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

Provide functions that are related to the safe handover mechanism
and allow console drivers to dynamically specify unsafe regions:

 - nbcon_context_can_proceed()

   Invoked by a console owner to check whether a handover request
   is pending or whether the console has been taken over by another
   context. If a handover request is pending, this function will
   also perform the handover, thus cancelling its own ownership.

 - nbcon_context_update_unsafe()

   Invoked by a console owner to denote that the driver is about
   to enter or leave a critical region where a take over is unsafe.
   This function is also a cancellation point where loss of
   ownership can occur.

   The unsafe state is stored in the console state and allows a
   new context to make informed decisions whether to attempt a
   takeover of such a console. The unsafe state is also available
   to the driver so that it can make informed decisions about the
   required actions and possibly take a special emergency path.

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/nbcon.c | 107 +++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 106 insertions(+), 1 deletion(-)

diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index c148467b39b8..76e39538bdec 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -490,7 +490,6 @@ static bool nbcon_owner_matches(struct nbcon_state *cur, int expected_cpu,
  * nbcon_context_release - Release the console
  * @ctxt:	The nbcon context from nbcon_context_try_acquire()
  */
-__maybe_unused
 static void nbcon_context_release(struct nbcon_context *ctxt)
 {
 	unsigned int cpu = smp_processor_id();
@@ -518,6 +517,112 @@ static void nbcon_context_release(struct nbcon_context *ctxt)
 	ctxt->pbufs = NULL;
 }
 
+/**
+ * nbcon_context_can_proceed - Check whether ownership can proceed
+ * @ctxt:	The nbcon context from nbcon_context_try_acquire()
+ * @cur:	The current console state
+ *
+ * Return:	True if this context still owns the console. False if
+ *		ownership was handed over or taken.
+ *
+ * Must be invoked after the record was dumped into the assigned buffer
+ * and at appropriate safe places in the driver.
+ *
+ * When this function returns false then the calling context no longer owns
+ * the console and is no longer allowed to go forward. In this case it must
+ * back out immediately and carefully. The buffer content is also no longer
+ * trusted since it no longer belongs to the calling context.
+ */
+static bool nbcon_context_can_proceed(struct nbcon_context *ctxt, struct nbcon_state *cur)
+{
+	unsigned int cpu = smp_processor_id();
+
+	/* Make sure this context still owns the console. */
+	if (!nbcon_owner_matches(cur, cpu, ctxt->prio))
+		return false;
+
+	/* The console owner can proceed if there is no waiter. */
+	if (cur->req_prio == NBCON_PRIO_NONE)
+		return true;
+
+	/*
+	 * A console owner within an unsafe region is always allowed to
+	 * proceed, even if there are waiters. It can perform a handover
+	 * when exiting the unsafe region. Otherwise the waiter will
+	 * need to perform an unsafe hostile takeover.
+	 */
+	if (cur->unsafe)
+		return true;
+
+	/* Waiters always have higher priorities than owners. */
+	WARN_ON_ONCE(cur->req_prio <= cur->prio);
+
+	/*
+	 * Having a safe point for take over and eventually a few
+	 * duplicated characters or a full line is way better than a
+	 * hostile takeover. Post processing can take care of the garbage.
+	 * Release and hand over.
+	 */
+	nbcon_context_release(ctxt);
+
+	/*
+	 * It is not known whether the handover succeeded. The outermost
+	 * callsite has to make the final decision whether printing
+	 * should proceed or not (via reacquire, possibly hostile). The
+	 * console is now unlocked so go back all the way instead of
+	 * trying to implement heuristics in tons of places.
+	 */
+	return false;
+}
+
+/**
+ * nbcon_context_update_unsafe - Update the unsafe bit in @con->nbcon_state
+ * @ctxt:	The nbcon context from nbcon_context_try_acquire()
+ * @unsafe:	The new value for the unsafe bit
+ *
+ * Return:	True if the unsafe state was updated and this context still
+ *		owns the console. Otherwise false if ownership was handed
+ *		over or taken.
+ *
+ * This function allows console owners to modify the unsafe status of the
+ * console.
+ *
+ * When this function returns false then the calling context no longer owns
+ * the console and is no longer allowed to go forward. In this case it must
+ * back out immediately and carefully. The buffer content is also no longer
+ * trusted since it no longer belongs to the calling context.
+ *
+ * Internal helper to avoid duplicated code.
+ */
+__maybe_unused
+static bool nbcon_context_update_unsafe(struct nbcon_context *ctxt, bool unsafe)
+{
+	struct console *con = ctxt->console;
+	struct nbcon_state cur;
+	struct nbcon_state new;
+
+	nbcon_state_read(con, &cur);
+
+	do {
+		/*
+		 * The unsafe bit must not be cleared if an
+		 * unsafe hostile takeover has occurred.
+		 */
+		if (!unsafe && cur.unsafe_takeover)
+			goto out;
+
+		if (!nbcon_context_can_proceed(ctxt, &cur))
+			return false;
+
+		new.atom = cur.atom;
+		new.unsafe = unsafe;
+	} while (!nbcon_state_try_cmpxchg(con, &cur, &new));
+
+	cur.atom = new.atom;
+out:
+	return nbcon_context_can_proceed(ctxt, &cur);
+}
+
 /**
  * nbcon_alloc - Allocate buffers needed by the nbcon console
  * @con:	Console to initialize
-- 
2.39.2

Re: [PATCH printk v3 4/7] printk: nbcon: Add ownership state functions

[Petr Mladek]

On Sun 2023-09-03 17:11:36, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> Provide functions that are related to the safe handover mechanism
> and allow console drivers to dynamically specify unsafe regions:
> 
>  - nbcon_context_can_proceed()
> 
>    Invoked by a console owner to check whether a handover request
>    is pending or whether the console has been taken over by another
>    context. If a handover request is pending, this function will
>    also perform the handover, thus cancelling its own ownership.
> 
>  - nbcon_context_update_unsafe()
> 
>    Invoked by a console owner to denote that the driver is about
>    to enter or leave a critical region where a take over is unsafe.
>    This function is also a cancellation point where loss of
>    ownership can occur.
> 
>    The unsafe state is stored in the console state and allows a
>    new context to make informed decisions whether to attempt a
>    takeover of such a console. The unsafe state is also available
>    to the driver so that it can make informed decisions about the
>    required actions and possibly take a special emergency path.
> 
> 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>

See one comment below.

> --- a/kernel/printk/nbcon.c
> +++ b/kernel/printk/nbcon.c
> @@ -518,6 +517,112 @@ static void nbcon_context_release(struct nbcon_context *ctxt)
>  	ctxt->pbufs = NULL;
>  }
>  
> +/**
> + * nbcon_context_can_proceed - Check whether ownership can proceed
> + * @ctxt:	The nbcon context from nbcon_context_try_acquire()
> + * @cur:	The current console state
> + *
> + * Return:	True if this context still owns the console. False if
> + *		ownership was handed over or taken.
> + *
> + * Must be invoked after the record was dumped into the assigned buffer
> + * and at appropriate safe places in the driver.
> + *
> + * When this function returns false then the calling context no longer owns
> + * the console and is no longer allowed to go forward. In this case it must
> + * back out immediately and carefully. The buffer content is also no longer
> + * trusted since it no longer belongs to the calling context.
> + */
> +static bool nbcon_context_can_proceed(struct nbcon_context *ctxt, struct nbcon_state *cur)
> +{
> +	unsigned int cpu = smp_processor_id();
> +
> +	/* Make sure this context still owns the console. */
> +	if (!nbcon_owner_matches(cur, cpu, ctxt->prio))
> +		return false;
> +
> +	/* The console owner can proceed if there is no waiter. */
> +	if (cur->req_prio == NBCON_PRIO_NONE)
> +		return true;
> +
> +	/*
> +	 * A console owner within an unsafe region is always allowed to
> +	 * proceed, even if there are waiters. It can perform a handover
> +	 * when exiting the unsafe region. Otherwise the waiter will
> +	 * need to perform an unsafe hostile takeover.
> +	 */
> +	if (cur->unsafe)
> +		return true;
> +
> +	/* Waiters always have higher priorities than owners. */
> +	WARN_ON_ONCE(cur->req_prio <= cur->prio);
> +
> +	/*
> +	 * Having a safe point for take over and eventually a few
> +	 * duplicated characters or a full line is way better than a
> +	 * hostile takeover. Post processing can take care of the garbage.
> +	 * Release and hand over.
> +	 */
> +	nbcon_context_release(ctxt);
> +
> +	/*
> +	 * It is not known whether the handover succeeded. The outermost

It was not immediately clear to me what exactly "handover succeeded" did mean.
I would write:

	* It is not clear whether the waiter really took the lock
	* and re-printed the record. The outermost calsite...

> +	 * callsite has to make the final decision whether printing
> +	 * should proceed or not (via reacquire, possibly hostile). The
> +	 * console is now unlocked so go back all the way instead of
> +	 * trying to implement heuristics in tons of places.
> +	 */
> +	return false;
> +}

Best Regards,
Petr

Re: [PATCH printk v3 4/7] printk: nbcon: Add ownership state functions

[John Ogness]

On 2023-09-06, Petr Mladek <pmladek@suse.com> wrote:
>> +static bool nbcon_context_can_proceed(struct nbcon_context *ctxt, struct nbcon_state *cur)
>> +{

[...]

>> +	/*
>> +	 * It is not known whether the handover succeeded. The outermost
>
> It was not immediately clear to me what exactly "handover succeeded" did mean.
> I would write:
>
> 	* It is not clear whether the waiter really took the lock
> 	* and re-printed the record. The outermost calsite...

It is not just about printing. A console is locked by drivers for other
purposes as well. And that is the situation that this comment is mostly
targetting. For v4 I change it to:

        /*
         * It is not clear whether the waiter really took over ownership. The
         * outermost callsite must make the final decision whether console
         * ownership is needed for it to proceed. If yes, it must reacquire
         * ownership (possibly hostile) before carefully proceeding.
         *
         * The calling context no longer owns the console so go back all the
         * way instead of trying to implement reacquire heuristics in tons of
         * places.
         */

John Ogness

[PATCH printk v3 5/7] printk: nbcon: Add sequence handling

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

Add an atomic_long_t field @nbcon_seq to the console struct to
store the sequence number for nbcon consoles. For nbcon consoles
this will be used instead of the non-atomic @seq field. The new
field allows for safe atomic sequence number updates without
requiring any locking.

On 64bit systems the new field stores the full sequence number.
On 32bit systems the new field stores the lower 32 bits of the
sequence number, which are expanded to 64bit as needed by
folding the values based on the sequence numbers available in
the ringbuffer.

For 32bit systems, having a 32bit representation in the console
is sufficient. If a console ever gets more than 2^31 records
behind the ringbuffer then this is the least of the problems.

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 |   7 +++
 kernel/printk/nbcon.c    | 112 +++++++++++++++++++++++++++++++++++++++
 kernel/printk/printk.c   |  31 ++++++++---
 4 files changed, 147 insertions(+), 7 deletions(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index d0bf9abd76b8..8e8dc8d1016b 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -249,6 +249,7 @@ struct printk_buffers;
  *				might cause a system freeze when the console
  *				is used later.
  * @pbufs:			Pointer to the text buffer for this context
+ * @seq:			The sequence number to print for this context
  */
 struct nbcon_context {
 	/* members set by caller */
@@ -259,6 +260,7 @@ struct nbcon_context {
 
 	/* members set by acquire */
 	struct printk_buffers	*pbufs;
+	u64			seq;
 };
 
 /**
@@ -282,6 +284,7 @@ struct nbcon_context {
  * @node:		hlist node for the console list
  *
  * @nbcon_state:	State for nbcon consoles
+ * @nbcon_seq:		Sequence number of the next record for nbcon to print
  * @pbufs:		Pointer to nbcon private buffer
  */
 struct console {
@@ -305,6 +308,7 @@ struct console {
 
 	/* nbcon console specific members */
 	atomic_t		__private nbcon_state;
+	atomic_long_t		__private nbcon_seq;
 	struct printk_buffers	*pbufs;
 };
 
diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index cb4ea90f04a5..956f65308392 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -4,6 +4,7 @@
  */
 #include <linux/percpu.h>
 #include <linux/console.h>
+#include "printk_ringbuffer.h"
 
 #if defined(CONFIG_PRINTK) && defined(CONFIG_SYSCTL)
 void __init printk_sysctl_init(void);
@@ -42,6 +43,8 @@ enum printk_info_flags {
 	LOG_CONT	= 8,	/* text is a fragment of a continuation line */
 };
 
+extern struct printk_ringbuffer *prb;
+
 __printf(4, 0)
 int vprintk_store(int facility, int level,
 		  const struct dev_printk_info *dev_info,
@@ -69,6 +72,8 @@ void defer_console_output(void);
 u16 printk_parse_prefix(const char *text, int *level,
 			enum printk_info_flags *flags);
 
+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_cleanup(struct console *con);
@@ -88,6 +93,8 @@ void nbcon_cleanup(struct console *con);
 #define printk_safe_exit_irqrestore(flags) local_irq_restore(flags)
 
 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_cleanup(struct console *con) { }
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 76e39538bdec..cd8d1e8539af 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -84,6 +84,112 @@ static inline bool nbcon_state_try_cmpxchg(struct console *con, struct nbcon_sta
 	return atomic_try_cmpxchg(&ACCESS_PRIVATE(con, nbcon_state), &cur->atom, new->atom);
 }
 
+#ifdef CONFIG_64BIT
+
+#define __seq_to_nbcon_seq(seq) (seq)
+#define __nbcon_seq_to_seq(seq) (seq)
+
+#else /* CONFIG_64BIT */
+
+#define __seq_to_nbcon_seq(seq) ((u32)seq)
+
+static inline u64 __nbcon_seq_to_seq(u32 nbcon_seq)
+{
+	u64 seq;
+	u64 rb_next_seq;
+
+	/*
+	 * The provided sequence is only the lower 32 bits of the ringbuffer
+	 * sequence. It needs to be expanded to 64bit. Get the next sequence
+	 * number from the ringbuffer and fold it.
+	 *
+	 * Having a 32bit representation in the console is sufficient.
+	 * If a console ever gets more than 2^31 records behind
+	 * the ringbuffer then this is the least of the problems.
+	 *
+	 * Also the access to the ring buffer is always safe.
+	 */
+	rb_next_seq = prb_next_seq(prb);
+	seq = rb_next_seq - ((u32)rb_next_seq - nbcon_seq);
+
+	return seq;
+}
+
+#endif /* CONFIG_64BIT */
+
+/**
+ * nbcon_seq_init - Helper function to initialize the console sequence
+ * @con:	Console to work on
+ *
+ * Set @con->nbcon_seq to the starting record (specified with con->seq).
+ * If the starting record no longer exists, the oldest available record
+ * is chosen. This is especially important on 32bit systems because only
+ * the lower 32 bits of the sequence number are stored. The upper 32 bits
+ * are derived from the sequence numbers available in the ringbuffer.
+ *
+ * For init only. Do not use for runtime updates.
+ */
+static void nbcon_seq_init(struct console *con)
+{
+	u64 seq = max_t(u64, con->seq, prb_first_valid_seq(prb));
+
+	atomic_long_set(&ACCESS_PRIVATE(con, nbcon_seq), __seq_to_nbcon_seq(seq));
+
+	/* Clear con->seq since nbcon consoles use con->nbcon_seq instead. */
+	con->seq = 0;
+}
+
+/**
+ * nbcon_seq_read - Read the current console sequence
+ * @con:	Console to read the sequence of
+ *
+ * Return:	Sequence number of the next record to print on @con.
+ */
+u64 nbcon_seq_read(struct console *con)
+{
+	unsigned long nbcon_seq = atomic_long_read(&ACCESS_PRIVATE(con, nbcon_seq));
+
+	return __nbcon_seq_to_seq(nbcon_seq);
+}
+
+/**
+ * nbcon_seq_force - Force console sequence to a specific value
+ * @con:	Console to work on
+ * @seq:	Sequence number value to set
+ *
+ * Only to be used in extreme situations (such as panic with
+ * CONSOLE_REPLAY_ALL).
+ */
+void nbcon_seq_force(struct console *con, u64 seq)
+{
+	atomic_long_set(&ACCESS_PRIVATE(con, nbcon_seq), __seq_to_nbcon_seq(seq));
+}
+
+/**
+ * nbcon_seq_try_update - Try to update the console sequence number
+ * @ctxt:	Pointer to an acquire context that contains
+ *		all information about the acquire mode
+ * @new_seq:	The new sequence number to set
+ *
+ * @ctxt->seq is updated to the new value of @con::nbcon_seq. This could
+ * be a different value than @new_seq if nbcon_seq_force() was used or the
+ * current context no longer owns the console. In the later case, it will
+ * stop printing anyway.
+ */
+__maybe_unused
+static void nbcon_seq_try_update(struct nbcon_context *ctxt, u64 new_seq)
+{
+	unsigned long nbcon_seq = __seq_to_nbcon_seq(ctxt->seq);
+	struct console *con = ctxt->console;
+
+	if (atomic_long_try_cmpxchg(&ACCESS_PRIVATE(con, nbcon_seq), &nbcon_seq,
+				    __seq_to_nbcon_seq(new_seq))) {
+		ctxt->seq = new_seq;
+	} else {
+		ctxt->seq = atomic_long_read(&ACCESS_PRIVATE(con, nbcon_seq));
+	}
+}
+
 /**
  * nbcon_context_try_acquire_direct - Try to acquire directly
  * @ctxt:	The context of the caller
@@ -463,6 +569,9 @@ static bool nbcon_context_try_acquire(struct nbcon_context *ctxt)
 	else
 		ctxt->pbufs = con->pbufs;
 
+	/* Set the record sequence for this context to print. */
+	ctxt->seq = nbcon_seq_read(ctxt->console);
+
 	return true;
 }
 
@@ -648,6 +757,8 @@ bool nbcon_alloc(struct console *con)
 /**
  * nbcon_init - Initialize the nbcon console specific data
  * @con:	Console to initialize
+ *
+ * This function expects that the legacy @con->seq has been set.
  */
 void nbcon_init(struct console *con)
 {
@@ -656,6 +767,7 @@ void nbcon_init(struct console *con)
 	/* nbcon_alloc() must have been called and successful! */
 	BUG_ON(!con->pbufs);
 
+	nbcon_seq_init(con);
 	nbcon_state_set(con, &state);
 }
 
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index b340c5a36fe2..4e22ff8d065e 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -492,7 +492,7 @@ _DEFINE_PRINTKRB(printk_rb_static, CONFIG_LOG_BUF_SHIFT - PRB_AVGBITS,
 
 static struct printk_ringbuffer printk_rb_dynamic;
 
-static struct printk_ringbuffer *prb = &printk_rb_static;
+struct printk_ringbuffer *prb = &printk_rb_static;
 
 /*
  * We cannot access per-CPU data (e.g. per-CPU flush irq_work) before
@@ -3161,6 +3161,7 @@ void console_flush_on_panic(enum con_flush_mode mode)
 
 	if (mode == CONSOLE_REPLAY_ALL) {
 		struct console *c;
+		short flags;
 		int cookie;
 		u64 seq;
 
@@ -3168,11 +3169,17 @@ void console_flush_on_panic(enum con_flush_mode mode)
 
 		cookie = console_srcu_read_lock();
 		for_each_console_srcu(c) {
-			/*
-			 * This is an unsynchronized assignment, but the
-			 * kernel is in "hope and pray" mode anyway.
-			 */
-			c->seq = seq;
+			flags = console_srcu_read_flags(c);
+
+			if (flags & CON_NBCON) {
+				nbcon_seq_force(c, seq);
+			} else {
+				/*
+				 * This is an unsynchronized assignment. On
+				 * panic legacy consoles are only best effort.
+				 */
+				c->seq = seq;
+			}
 		}
 		console_srcu_read_unlock(cookie);
 	}
@@ -3740,6 +3747,7 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 	struct console *c;
 	u64 last_diff = 0;
 	u64 printk_seq;
+	short flags;
 	int cookie;
 	u64 diff;
 	u64 seq;
@@ -3761,6 +3769,9 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 		for_each_console_srcu(c) {
 			if (con && con != c)
 				continue;
+
+			flags = console_srcu_read_flags(c);
+
 			/*
 			 * If consoles are not usable, it cannot be expected
 			 * that they make forward progress, so only increment
@@ -3768,7 +3779,13 @@ static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progre
 			 */
 			if (!console_is_usable(c))
 				continue;
-			printk_seq = c->seq;
+
+			if (flags & CON_NBCON) {
+				printk_seq = nbcon_seq_read(c);
+			} else {
+				printk_seq = c->seq;
+			}
+
 			if (printk_seq < seq)
 				diff += seq - printk_seq;
 		}
-- 
2.39.2

Re: [PATCH printk v3 5/7] printk: nbcon: Add sequence handling

[Petr Mladek]

On Sun 2023-09-03 17:11:37, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> Add an atomic_long_t field @nbcon_seq to the console struct to
> store the sequence number for nbcon consoles. For nbcon consoles
> this will be used instead of the non-atomic @seq field. The new
> field allows for safe atomic sequence number updates without
> requiring any locking.
> 
> On 64bit systems the new field stores the full sequence number.
> On 32bit systems the new field stores the lower 32 bits of the
> sequence number, which are expanded to 64bit as needed by
> folding the values based on the sequence numbers available in
> the ringbuffer.
> 
> For 32bit systems, having a 32bit representation in the console
> is sufficient. If a console ever gets more than 2^31 records
> behind the ringbuffer then this is the least of the problems.
> 
> --- a/kernel/printk/nbcon.c
> +++ b/kernel/printk/nbcon.c
> +/**
> + * nbcon_seq_try_update - Try to update the console sequence number
> + * @ctxt:	Pointer to an acquire context that contains
> + *		all information about the acquire mode
> + * @new_seq:	The new sequence number to set
> + *
> + * @ctxt->seq is updated to the new value of @con::nbcon_seq. This could
> + * be a different value than @new_seq if nbcon_seq_force() was used or the
> + * current context no longer owns the console. In the later case, it will
> + * stop printing anyway.
> + */
> +__maybe_unused
> +static void nbcon_seq_try_update(struct nbcon_context *ctxt, u64 new_seq)
> +{
> +	unsigned long nbcon_seq = __seq_to_nbcon_seq(ctxt->seq);
> +	struct console *con = ctxt->console;
> +
> +	if (atomic_long_try_cmpxchg(&ACCESS_PRIVATE(con, nbcon_seq), &nbcon_seq,
> +				    __seq_to_nbcon_seq(new_seq))) {
> +		ctxt->seq = new_seq;
> +	} else {
> +		ctxt->seq = atomic_long_read(&ACCESS_PRIVATE(con, nbcon_seq));

This should be:

		ctxt->seq = nbcon_seq_read(con);

so that it extends the 32-bit nbcon_seq to the full 64-bit seq on
32-bit systems.

> +	}
> +}
> +
>  /**
>   * nbcon_context_try_acquire_direct - Try to acquire directly
>   * @ctxt:	The context of the caller

Otherwise, it looks good. With the above fix:

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

Best Regards,
Petr

[PATCH printk v3 6/7] printk: nbcon: Add emit function and callback function for atomic printing

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

Implement an emit function for nbcon consoles to output printk
messages. It utilizes the lockless printk_get_next_message() and
console_prepend_dropped() functions to retrieve/build the output
message. The emit function includes the required safety points to
check for handover/takeover and calls a new write_atomic callback
of the console driver to output the message. It also includes
proper handling for updating the nbcon console sequence number.

A new nbcon_write_context struct is introduced. This is provided
to the write_atomic callback and includes only the information
necessary for performing atomic writes.

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  |  27 +++++++++
 kernel/printk/internal.h |   6 ++
 kernel/printk/nbcon.c    | 124 ++++++++++++++++++++++++++++++++++++++-
 kernel/printk/printk.c   |   9 +--
 4 files changed, 158 insertions(+), 8 deletions(-)

diff --git a/include/linux/console.h b/include/linux/console.h
index 8e8dc8d1016b..8d0a309b5286 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -248,6 +248,7 @@ struct printk_buffers;
  *				be used only with NBCON_PRIO_PANIC @prio. It
  *				might cause a system freeze when the console
  *				is used later.
+ * @backlog:			Ringbuffer has pending records
  * @pbufs:			Pointer to the text buffer for this context
  * @seq:			The sequence number to print for this context
  */
@@ -258,11 +259,28 @@ struct nbcon_context {
 	enum nbcon_prio		prio;
 	unsigned int		allow_unsafe_takeover	: 1;
 
+	/* members set by emit */
+	unsigned int		backlog			: 1;
+
 	/* members set by acquire */
 	struct printk_buffers	*pbufs;
 	u64			seq;
 };
 
+/**
+ * struct nbcon_write_context - Context handed to the nbcon write callbacks
+ * @ctxt:		The core console context
+ * @outbuf:		Pointer to the text buffer for output
+ * @len:		Length to write
+ * @unsafe_takeover:	If a hostile takeover in an unsafe state has occurred
+ */
+struct nbcon_write_context {
+	struct nbcon_context	__private ctxt;
+	char			*outbuf;
+	unsigned int		len;
+	bool			unsafe_takeover;
+};
+
 /**
  * struct console - The console descriptor structure
  * @name:		The name of the console driver
@@ -283,6 +301,7 @@ struct nbcon_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
@@ -307,6 +326,8 @@ struct console {
 	struct hlist_node	node;
 
 	/* nbcon console specific members */
+	bool			(*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;
@@ -436,6 +457,12 @@ static inline bool console_is_registered(const struct console *con)
 	lockdep_assert_console_list_lock_held();			\
 	hlist_for_each_entry(con, &console_list, node)
 
+#ifdef CONFIG_PRINTK
+extern bool nbcon_can_proceed(struct nbcon_write_context *wctxt);
+#else
+static inline bool nbcon_can_proceed(struct nbcon_write_context *wctxt) { return false; }
+#endif
+
 extern int console_set_on_cmdline;
 extern struct console *early_console;
 
diff --git a/kernel/printk/internal.h b/kernel/printk/internal.h
index 956f65308392..4d532623b3cf 100644
--- a/kernel/printk/internal.h
+++ b/kernel/printk/internal.h
@@ -128,3 +128,9 @@ struct printk_message {
 };
 
 bool other_cpu_in_panic(void);
+bool printk_get_next_message(struct printk_message *pmsg, u64 seq,
+			     bool is_extended, bool may_supress);
+
+#ifdef CONFIG_PRINTK
+void console_prepend_dropped(struct printk_message *pmsg, unsigned long dropped);
+#endif
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index cd8d1e8539af..1a3df603111d 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -176,7 +176,6 @@ void nbcon_seq_force(struct console *con, u64 seq)
  * current context no longer owns the console. In the later case, it will
  * stop printing anyway.
  */
-__maybe_unused
 static void nbcon_seq_try_update(struct nbcon_context *ctxt, u64 new_seq)
 {
 	unsigned long nbcon_seq = __seq_to_nbcon_seq(ctxt->seq);
@@ -684,6 +683,32 @@ static bool nbcon_context_can_proceed(struct nbcon_context *ctxt, struct nbcon_s
 	return false;
 }
 
+/**
+ * nbcon_can_proceed - Check whether ownership can proceed
+ * @wctxt:	The write context that was handed to the write function
+ *
+ * Return:	True if this context still owns the console. False if
+ *		ownership was handed over or taken.
+ *
+ * Must be invoked at appropriate safe places in the driver.
+ *
+ * When this function returns false then the calling context no longer owns
+ * the console and is no longer allowed to go forward. In this case it must
+ * back out immediately and carefully. The buffer content is also no longer
+ * trusted since it no longer belongs to the calling context.
+ */
+bool nbcon_can_proceed(struct nbcon_write_context *wctxt)
+{
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(wctxt, ctxt);
+	struct console *con = ctxt->console;
+	struct nbcon_state cur;
+
+	nbcon_state_read(con, &cur);
+
+	return nbcon_context_can_proceed(ctxt, &cur);
+}
+EXPORT_SYMBOL_GPL(nbcon_can_proceed);
+
 /**
  * nbcon_context_update_unsafe - Update the unsafe bit in @con->nbcon_state
  * @ctxt:	The nbcon context from nbcon_context_try_acquire()
@@ -703,7 +728,6 @@ static bool nbcon_context_can_proceed(struct nbcon_context *ctxt, struct nbcon_s
  *
  * Internal helper to avoid duplicated code.
  */
-__maybe_unused
 static bool nbcon_context_update_unsafe(struct nbcon_context *ctxt, bool unsafe)
 {
 	struct console *con = ctxt->console;
@@ -732,6 +756,102 @@ static bool nbcon_context_update_unsafe(struct nbcon_context *ctxt, bool unsafe)
 	return nbcon_context_can_proceed(ctxt, &cur);
 }
 
+/**
+ * nbcon_emit_next_record - Emit a record in the acquired context
+ * @wctxt:	The write context that will be handed to the write function
+ *
+ * Return:	True if this context still owns the console. False if
+ *		ownership was handed over or taken.
+ *
+ * When this function returns false then the calling context no longer owns
+ * the console and is no longer allowed to go forward. In this case it must
+ * back out immediately and carefully. The buffer content is also no longer
+ * trusted since it no longer belongs to the calling context. If the caller
+ * wants to do more it must reacquire the console first.
+ *
+ * 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);
+	struct console *con = ctxt->console;
+	bool is_extended = console_srcu_read_flags(con) & CON_EXTENDED;
+	struct printk_message pmsg = {
+		.pbufs = ctxt->pbufs,
+	};
+	unsigned long con_dropped;
+	struct nbcon_state cur;
+	unsigned long dropped;
+	bool done;
+
+	ctxt->backlog = printk_get_next_message(&pmsg, ctxt->seq, is_extended, true);
+	if (!ctxt->backlog)
+		return true;
+
+	/*
+	 * @con->dropped is not protected in case of an unsafe hostile
+	 * takeover. In that situation the update can be racy so
+	 * annotate it accordingly.
+	 */
+	con_dropped = data_race(READ_ONCE(con->dropped));
+
+	dropped = con_dropped + pmsg.dropped;
+	if (dropped && !is_extended)
+		console_prepend_dropped(&pmsg, dropped);
+
+	/* Safety point. Do not touch state in case of takeover. */
+	nbcon_state_read(con, &cur);
+	if (!nbcon_context_can_proceed(ctxt, &cur))
+		return false;
+
+	/* For skipped records just update seq/dropped in @con. */
+	if (pmsg.outbuf_len == 0)
+		goto update_con;
+
+	/* Set the write context before calling write callback. */
+	wctxt->outbuf = &pmsg.pbufs->outbuf[0];
+	wctxt->len = pmsg.outbuf_len;
+	wctxt->unsafe_takeover = cur.unsafe_takeover;
+
+	if (con->write_atomic) {
+		done = con->write_atomic(con, wctxt);
+	} else {
+		nbcon_context_release(ctxt);
+		WARN_ON_ONCE(1);
+		done = false;
+	}
+
+	/* If not done, the emit was aborted. */
+	if (!done)
+		return false;
+
+	/*
+	 * Since any dropped message was successfully output, reset the
+	 * dropped count for the console.
+	 */
+	dropped = 0;
+update_con:
+	/*
+	 * The dropped count and the sequence number are updated within an
+	 * unsafe section. This limits update races to the panic context and
+	 * allows the panic context to win.
+	 */
+
+	if (!nbcon_context_update_unsafe(ctxt, true))
+		return false;
+
+	if (dropped != con_dropped) {
+		/* Counterpart to the READ_ONCE() above. */
+		WRITE_ONCE(con->dropped, dropped);
+	}
+
+	nbcon_seq_try_update(ctxt, pmsg.seq + 1);
+
+	return nbcon_context_update_unsafe(ctxt, false);
+}
+
 /**
  * nbcon_alloc - Allocate buffers needed by the nbcon console
  * @con:	Console to initialize
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 4e22ff8d065e..f6e1fca2a2cb 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -696,9 +696,6 @@ static ssize_t msg_print_ext_body(char *buf, size_t size,
 	return len;
 }
 
-static bool printk_get_next_message(struct printk_message *pmsg, u64 seq,
-				    bool is_extended, bool may_supress);
-
 /* /dev/kmsg - userspace message inject/listen interface */
 struct devkmsg_user {
 	atomic64_t seq;
@@ -2731,7 +2728,7 @@ static void __console_unlock(void)
  * If @pmsg->pbufs->outbuf is modified, @pmsg->outbuf_len is updated.
  */
 #ifdef CONFIG_PRINTK
-static void console_prepend_dropped(struct printk_message *pmsg, unsigned long dropped)
+void console_prepend_dropped(struct printk_message *pmsg, unsigned long dropped)
 {
 	struct printk_buffers *pbufs = pmsg->pbufs;
 	const size_t scratchbuf_sz = sizeof(pbufs->scratchbuf);
@@ -2785,8 +2782,8 @@ static void console_prepend_dropped(struct printk_message *pmsg, unsigned long d
  * of @pmsg are valid. (See the documentation of struct printk_message
  * for information about the @pmsg fields.)
  */
-static bool printk_get_next_message(struct printk_message *pmsg, u64 seq,
-				    bool is_extended, bool may_suppress)
+bool printk_get_next_message(struct printk_message *pmsg, u64 seq,
+			     bool is_extended, bool may_suppress)
 {
 	static int panic_console_dropped;
 
-- 
2.39.2

Re: [PATCH printk v3 6/7] printk: nbcon: Add emit function and callback function for atomic printing

[Petr Mladek]

On Sun 2023-09-03 17:11:38, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> Implement an emit function for nbcon consoles to output printk
> messages. It utilizes the lockless printk_get_next_message() and
> console_prepend_dropped() functions to retrieve/build the output
> message. The emit function includes the required safety points to
> check for handover/takeover and calls a new write_atomic callback
> of the console driver to output the message. It also includes
> proper handling for updating the nbcon console sequence number.
> 
> A new nbcon_write_context struct is introduced. This is provided
> to the write_atomic callback and includes only the information
> necessary for performing atomic writes.
> 
> 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>

Well, I would prefer one small change, see below.

> --- a/kernel/printk/nbcon.c
> +++ b/kernel/printk/nbcon.c
> @@ -732,6 +756,102 @@ static bool nbcon_context_update_unsafe(struct nbcon_context *ctxt, bool unsafe)
>  	return nbcon_context_can_proceed(ctxt, &cur);
>  }
>  
> +/**
> + * nbcon_emit_next_record - Emit a record in the acquired context
> + * @wctxt:	The write context that will be handed to the write function
> + *
> + * Return:	True if this context still owns the console. False if
> + *		ownership was handed over or taken.
> + *
> + * When this function returns false then the calling context no longer owns
> + * the console and is no longer allowed to go forward. In this case it must
> + * back out immediately and carefully. The buffer content is also no longer
> + * trusted since it no longer belongs to the calling context. If the caller
> + * wants to do more it must reacquire the console first.
> + *
> + * 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);
> +	struct console *con = ctxt->console;
> +	bool is_extended = console_srcu_read_flags(con) & CON_EXTENDED;
> +	struct printk_message pmsg = {
> +		.pbufs = ctxt->pbufs,
> +	};
> +	unsigned long con_dropped;
> +	struct nbcon_state cur;
> +	unsigned long dropped;
> +	bool done;
[...]
> +update_con:
> +	/*
> +	 * The dropped count and the sequence number are updated within an
> +	 * unsafe section. This limits update races to the panic context and
> +	 * allows the panic context to win.
> +	 */
> +
> +	if (!nbcon_context_update_unsafe(ctxt, true))
> +		return false;

I would prefer to use nbcon_enter_unsafe(). It is more self-explaining.
And it will make it easier to see the "all" enter/exit points using cscope.

> +
> +	if (dropped != con_dropped) {
> +		/* Counterpart to the READ_ONCE() above. */
> +		WRITE_ONCE(con->dropped, dropped);
> +	}
> +
> +	nbcon_seq_try_update(ctxt, pmsg.seq + 1);
> +
> +	return nbcon_context_update_unsafe(ctxt, false);
> +}

Best Regards,
Petr

[PATCH printk v3 7/7] printk: nbcon: Add functions for drivers to mark unsafe regions

[John Ogness]

From: Thomas Gleixner <tglx@linutronix.de>

For the write_atomic callback, the console driver may have unsafe
regions that need to be appropriately marked. Provide functions
that accept the nbcon_write_context struct to allow for the driver
to enter and exit unsafe regions.

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/nbcon.c   | 40 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 44 insertions(+)

diff --git a/include/linux/console.h b/include/linux/console.h
index 8d0a309b5286..b3ea9ea66a38 100644
--- a/include/linux/console.h
+++ b/include/linux/console.h
@@ -459,8 +459,12 @@ static inline bool console_is_registered(const struct console *con)
 
 #ifdef CONFIG_PRINTK
 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 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; }
 #endif
 
 extern int console_set_on_cmdline;
diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
index 1a3df603111d..74f9c33201ae 100644
--- a/kernel/printk/nbcon.c
+++ b/kernel/printk/nbcon.c
@@ -756,6 +756,46 @@ static bool nbcon_context_update_unsafe(struct nbcon_context *ctxt, bool unsafe)
 	return nbcon_context_can_proceed(ctxt, &cur);
 }
 
+/**
+ * nbcon_enter_unsafe - Enter an unsafe region in the driver
+ * @wctxt:	The write context that was handed to the write function
+ *
+ * Return:	True if this context still owns the console. False if
+ *		ownership was handed over or taken.
+ *
+ * When this function returns false then the calling context no longer owns
+ * the console and is no longer allowed to go forward. In this case it must
+ * back out immediately and carefully. The buffer content is also no longer
+ * trusted since it no longer belongs to the calling context.
+ */
+bool nbcon_enter_unsafe(struct nbcon_write_context *wctxt)
+{
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(wctxt, ctxt);
+
+	return nbcon_context_update_unsafe(ctxt, true);
+}
+EXPORT_SYMBOL_GPL(nbcon_enter_unsafe);
+
+/**
+ * nbcon_exit_unsafe - Exit an unsafe region in the driver
+ * @wctxt:	The write context that was handed to the write function
+ *
+ * Return:	True if this context still owns the console. False if
+ *		ownership was handed over or taken.
+ *
+ * When this function returns false then the calling context no longer owns
+ * the console and is no longer allowed to go forward. In this case it must
+ * back out immediately and carefully. The buffer content is also no longer
+ * trusted since it no longer belongs to the calling context.
+ */
+bool nbcon_exit_unsafe(struct nbcon_write_context *wctxt)
+{
+	struct nbcon_context *ctxt = &ACCESS_PRIVATE(wctxt, ctxt);
+
+	return nbcon_context_update_unsafe(ctxt, false);
+}
+EXPORT_SYMBOL_GPL(nbcon_exit_unsafe);
+
 /**
  * nbcon_emit_next_record - Emit a record in the acquired context
  * @wctxt:	The write context that will be handed to the write function
-- 
2.39.2

Re: [PATCH printk v3 7/7] printk: nbcon: Add functions for drivers to mark unsafe regions

[Petr Mladek]

On Sun 2023-09-03 17:11:39, John Ogness wrote:
> From: Thomas Gleixner <tglx@linutronix.de>
> 
> For the write_atomic callback, the console driver may have unsafe
> regions that need to be appropriately marked. Provide functions
> that accept the nbcon_write_context struct to allow for the driver
> to enter and exit unsafe regions.
> 
> 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