2019-05-19 15:51:43 +02:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
2016-05-21 02:00:33 +02:00
|
|
|
/*
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
* printk_safe.c - Safe printk for printk-deadlock-prone contexts
|
2016-05-21 02:00:33 +02:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/preempt.h>
|
|
|
|
#include <linux/spinlock.h>
|
2016-05-21 02:00:42 +02:00
|
|
|
#include <linux/debug_locks.h>
|
2016-05-21 02:00:33 +02:00
|
|
|
#include <linux/smp.h>
|
|
|
|
#include <linux/cpumask.h>
|
|
|
|
#include <linux/irq_work.h>
|
|
|
|
#include <linux/printk.h>
|
|
|
|
|
|
|
|
#include "internal.h"
|
|
|
|
|
|
|
|
/*
|
|
|
|
* printk() could not take logbuf_lock in NMI context. Instead,
|
|
|
|
* it uses an alternative implementation that temporary stores
|
|
|
|
* the strings into a per-CPU buffer. The content of the buffer
|
|
|
|
* is later flushed into the main ring buffer via IRQ work.
|
|
|
|
*
|
|
|
|
* The alternative implementation is chosen transparently
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
* by examinig current printk() context mask stored in @printk_context
|
|
|
|
* per-CPU variable.
|
2016-05-21 02:00:33 +02:00
|
|
|
*
|
|
|
|
* The implementation allows to flush the strings also from another CPU.
|
|
|
|
* There are situations when we want to make sure that all buffers
|
|
|
|
* were handled or when IRQs are blocked.
|
|
|
|
*/
|
2017-10-11 18:46:47 +02:00
|
|
|
static int printk_safe_irq_ready __read_mostly;
|
2016-05-21 02:00:33 +02:00
|
|
|
|
2016-12-27 15:16:05 +01:00
|
|
|
#define SAFE_LOG_BUF_LEN ((1 << CONFIG_PRINTK_SAFE_LOG_BUF_SHIFT) - \
|
2016-12-27 15:16:08 +01:00
|
|
|
sizeof(atomic_t) - \
|
|
|
|
sizeof(atomic_t) - \
|
|
|
|
sizeof(struct irq_work))
|
2016-05-21 02:00:33 +02:00
|
|
|
|
2016-12-27 15:16:05 +01:00
|
|
|
struct printk_safe_seq_buf {
|
2016-05-21 02:00:33 +02:00
|
|
|
atomic_t len; /* length of written data */
|
2016-12-27 15:16:08 +01:00
|
|
|
atomic_t message_lost;
|
2016-05-21 02:00:33 +02:00
|
|
|
struct irq_work work; /* IRQ work that flushes the buffer */
|
2016-12-27 15:16:05 +01:00
|
|
|
unsigned char buffer[SAFE_LOG_BUF_LEN];
|
2016-05-21 02:00:33 +02:00
|
|
|
};
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
|
|
|
|
static DEFINE_PER_CPU(struct printk_safe_seq_buf, safe_print_seq);
|
|
|
|
static DEFINE_PER_CPU(int, printk_context);
|
|
|
|
|
|
|
|
#ifdef CONFIG_PRINTK_NMI
|
2016-12-27 15:16:05 +01:00
|
|
|
static DEFINE_PER_CPU(struct printk_safe_seq_buf, nmi_print_seq);
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
#endif
|
2016-05-21 02:00:33 +02:00
|
|
|
|
2016-12-27 15:16:08 +01:00
|
|
|
/* Get flushed in a more safe context. */
|
|
|
|
static void queue_flush_work(struct printk_safe_seq_buf *s)
|
|
|
|
{
|
2017-10-11 18:46:47 +02:00
|
|
|
if (printk_safe_irq_ready)
|
2016-12-27 15:16:08 +01:00
|
|
|
irq_work_queue(&s->work);
|
|
|
|
}
|
|
|
|
|
2016-05-21 02:00:33 +02:00
|
|
|
/*
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
* Add a message to per-CPU context-dependent buffer. NMI and printk-safe
|
|
|
|
* have dedicated buffers, because otherwise printk-safe preempted by
|
|
|
|
* NMI-printk would have overwritten the NMI messages.
|
|
|
|
*
|
2017-10-22 16:30:55 +02:00
|
|
|
* The messages are flushed from irq work (or from panic()), possibly,
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
* from other CPU, concurrently with printk_safe_log_store(). Should this
|
|
|
|
* happen, printk_safe_log_store() will notice the buffer->len mismatch
|
|
|
|
* and repeat the write.
|
2016-05-21 02:00:33 +02:00
|
|
|
*/
|
2017-05-24 07:49:50 +02:00
|
|
|
static __printf(2, 0) int printk_safe_log_store(struct printk_safe_seq_buf *s,
|
|
|
|
const char *fmt, va_list args)
|
2016-05-21 02:00:33 +02:00
|
|
|
{
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
int add;
|
2016-05-21 02:00:33 +02:00
|
|
|
size_t len;
|
2018-05-11 12:54:19 +02:00
|
|
|
va_list ap;
|
2016-05-21 02:00:33 +02:00
|
|
|
|
|
|
|
again:
|
|
|
|
len = atomic_read(&s->len);
|
|
|
|
|
2016-12-13 01:45:40 +01:00
|
|
|
/* The trailing '\0' is not counted into len. */
|
|
|
|
if (len >= sizeof(s->buffer) - 1) {
|
2016-12-27 15:16:08 +01:00
|
|
|
atomic_inc(&s->message_lost);
|
|
|
|
queue_flush_work(s);
|
2016-05-21 02:00:33 +02:00
|
|
|
return 0;
|
2016-05-21 02:00:36 +02:00
|
|
|
}
|
2016-05-21 02:00:33 +02:00
|
|
|
|
|
|
|
/*
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
* Make sure that all old data have been read before the buffer
|
|
|
|
* was reset. This is not needed when we just append data.
|
2016-05-21 02:00:33 +02:00
|
|
|
*/
|
|
|
|
if (!len)
|
|
|
|
smp_rmb();
|
|
|
|
|
2018-05-11 12:54:19 +02:00
|
|
|
va_copy(ap, args);
|
|
|
|
add = vscnprintf(s->buffer + len, sizeof(s->buffer) - len, fmt, ap);
|
|
|
|
va_end(ap);
|
2016-12-27 15:16:08 +01:00
|
|
|
if (!add)
|
|
|
|
return 0;
|
2016-05-21 02:00:33 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Do it once again if the buffer has been flushed in the meantime.
|
|
|
|
* Note that atomic_cmpxchg() is an implicit memory barrier that
|
|
|
|
* makes sure that the data were written before updating s->len.
|
|
|
|
*/
|
|
|
|
if (atomic_cmpxchg(&s->len, len, len + add) != len)
|
|
|
|
goto again;
|
|
|
|
|
2016-12-27 15:16:08 +01:00
|
|
|
queue_flush_work(s);
|
2016-05-21 02:00:33 +02:00
|
|
|
return add;
|
|
|
|
}
|
|
|
|
|
2017-02-06 17:42:53 +01:00
|
|
|
static inline void printk_safe_flush_line(const char *text, int len)
|
2016-05-21 02:00:33 +02:00
|
|
|
{
|
2016-05-21 02:00:42 +02:00
|
|
|
/*
|
2017-02-06 17:42:53 +01:00
|
|
|
* Avoid any console drivers calls from here, because we may be
|
|
|
|
* in NMI or printk_safe context (when in panic). The messages
|
|
|
|
* must go only into the ring buffer at this stage. Consoles will
|
|
|
|
* get explicitly called later when a crashdump is not generated.
|
2016-05-21 02:00:42 +02:00
|
|
|
*/
|
2017-02-06 17:42:53 +01:00
|
|
|
printk_deferred("%.*s", len, text);
|
2016-05-21 02:00:33 +02:00
|
|
|
}
|
|
|
|
|
2016-12-13 01:45:44 +01:00
|
|
|
/* printk part of the temporary buffer line by line */
|
2016-12-27 15:16:05 +01:00
|
|
|
static int printk_safe_flush_buffer(const char *start, size_t len)
|
2016-09-02 01:15:04 +02:00
|
|
|
{
|
2016-12-13 01:45:44 +01:00
|
|
|
const char *c, *end;
|
|
|
|
bool header;
|
|
|
|
|
|
|
|
c = start;
|
|
|
|
end = start + len;
|
|
|
|
header = true;
|
|
|
|
|
|
|
|
/* Print line by line. */
|
|
|
|
while (c < end) {
|
|
|
|
if (*c == '\n') {
|
2016-12-27 15:16:05 +01:00
|
|
|
printk_safe_flush_line(start, c - start + 1);
|
2016-12-13 01:45:44 +01:00
|
|
|
start = ++c;
|
|
|
|
header = true;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Handle continuous lines or missing new line. */
|
|
|
|
if ((c + 1 < end) && printk_get_level(c)) {
|
|
|
|
if (header) {
|
|
|
|
c = printk_skip_level(c);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2016-12-27 15:16:05 +01:00
|
|
|
printk_safe_flush_line(start, c - start);
|
2016-12-13 01:45:44 +01:00
|
|
|
start = c++;
|
|
|
|
header = true;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
header = false;
|
|
|
|
c++;
|
|
|
|
}
|
2016-09-02 01:15:04 +02:00
|
|
|
|
2016-12-13 01:45:44 +01:00
|
|
|
/* Check if there was a partial line. Ignore pure header. */
|
|
|
|
if (start < end && !header) {
|
|
|
|
static const char newline[] = KERN_CONT "\n";
|
|
|
|
|
2016-12-27 15:16:05 +01:00
|
|
|
printk_safe_flush_line(start, end - start);
|
|
|
|
printk_safe_flush_line(newline, strlen(newline));
|
2016-12-13 01:45:44 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
return len;
|
2016-09-02 01:15:04 +02:00
|
|
|
}
|
|
|
|
|
2016-12-27 15:16:08 +01:00
|
|
|
static void report_message_lost(struct printk_safe_seq_buf *s)
|
|
|
|
{
|
|
|
|
int lost = atomic_xchg(&s->message_lost, 0);
|
|
|
|
|
|
|
|
if (lost)
|
|
|
|
printk_deferred("Lost %d message(s)!\n", lost);
|
|
|
|
}
|
|
|
|
|
2016-05-21 02:00:33 +02:00
|
|
|
/*
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
* Flush data from the associated per-CPU buffer. The function
|
2016-05-21 02:00:33 +02:00
|
|
|
* can be called either via IRQ work or independently.
|
|
|
|
*/
|
2016-12-27 15:16:05 +01:00
|
|
|
static void __printk_safe_flush(struct irq_work *work)
|
2016-05-21 02:00:33 +02:00
|
|
|
{
|
|
|
|
static raw_spinlock_t read_lock =
|
|
|
|
__RAW_SPIN_LOCK_INITIALIZER(read_lock);
|
2016-12-27 15:16:05 +01:00
|
|
|
struct printk_safe_seq_buf *s =
|
|
|
|
container_of(work, struct printk_safe_seq_buf, work);
|
2016-05-21 02:00:33 +02:00
|
|
|
unsigned long flags;
|
2016-12-13 01:45:44 +01:00
|
|
|
size_t len;
|
|
|
|
int i;
|
2016-05-21 02:00:33 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The lock has two functions. First, one reader has to flush all
|
|
|
|
* available message to make the lockless synchronization with
|
|
|
|
* writers easier. Second, we do not want to mix messages from
|
|
|
|
* different CPUs. This is especially important when printing
|
|
|
|
* a backtrace.
|
|
|
|
*/
|
|
|
|
raw_spin_lock_irqsave(&read_lock, flags);
|
|
|
|
|
|
|
|
i = 0;
|
|
|
|
more:
|
|
|
|
len = atomic_read(&s->len);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is just a paranoid check that nobody has manipulated
|
|
|
|
* the buffer an unexpected way. If we printed something then
|
2016-12-13 01:45:44 +01:00
|
|
|
* @len must only increase. Also it should never overflow the
|
|
|
|
* buffer size.
|
2016-05-21 02:00:33 +02:00
|
|
|
*/
|
2016-12-13 01:45:44 +01:00
|
|
|
if ((i && i >= len) || len > sizeof(s->buffer)) {
|
2016-12-27 15:16:05 +01:00
|
|
|
const char *msg = "printk_safe_flush: internal error\n";
|
2016-09-02 01:15:04 +02:00
|
|
|
|
2016-12-27 15:16:05 +01:00
|
|
|
printk_safe_flush_line(msg, strlen(msg));
|
2016-12-13 01:45:44 +01:00
|
|
|
len = 0;
|
2016-09-02 01:15:04 +02:00
|
|
|
}
|
2016-05-21 02:00:33 +02:00
|
|
|
|
|
|
|
if (!len)
|
|
|
|
goto out; /* Someone else has already flushed the buffer. */
|
|
|
|
|
|
|
|
/* Make sure that data has been written up to the @len */
|
|
|
|
smp_rmb();
|
2016-12-27 15:16:05 +01:00
|
|
|
i += printk_safe_flush_buffer(s->buffer + i, len - i);
|
2016-05-21 02:00:33 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check that nothing has got added in the meantime and truncate
|
|
|
|
* the buffer. Note that atomic_cmpxchg() is an implicit memory
|
|
|
|
* barrier that makes sure that the data were copied before
|
|
|
|
* updating s->len.
|
|
|
|
*/
|
|
|
|
if (atomic_cmpxchg(&s->len, len, 0) != len)
|
|
|
|
goto more;
|
|
|
|
|
|
|
|
out:
|
2016-12-27 15:16:08 +01:00
|
|
|
report_message_lost(s);
|
2016-05-21 02:00:33 +02:00
|
|
|
raw_spin_unlock_irqrestore(&read_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2016-12-27 15:16:05 +01:00
|
|
|
* printk_safe_flush - flush all per-cpu nmi buffers.
|
2016-05-21 02:00:33 +02:00
|
|
|
*
|
|
|
|
* The buffers are flushed automatically via IRQ work. This function
|
|
|
|
* is useful only when someone wants to be sure that all buffers have
|
|
|
|
* been flushed at some point.
|
|
|
|
*/
|
2016-12-27 15:16:05 +01:00
|
|
|
void printk_safe_flush(void)
|
2016-05-21 02:00:33 +02:00
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
#ifdef CONFIG_PRINTK_NMI
|
2016-12-27 15:16:05 +01:00
|
|
|
__printk_safe_flush(&per_cpu(nmi_print_seq, cpu).work);
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
#endif
|
|
|
|
__printk_safe_flush(&per_cpu(safe_print_seq, cpu).work);
|
|
|
|
}
|
2016-05-21 02:00:33 +02:00
|
|
|
}
|
|
|
|
|
2016-05-21 02:00:42 +02:00
|
|
|
/**
|
2016-12-27 15:16:05 +01:00
|
|
|
* printk_safe_flush_on_panic - flush all per-cpu nmi buffers when the system
|
2016-05-21 02:00:42 +02:00
|
|
|
* goes down.
|
|
|
|
*
|
2016-12-27 15:16:05 +01:00
|
|
|
* Similar to printk_safe_flush() but it can be called even in NMI context when
|
2016-05-21 02:00:42 +02:00
|
|
|
* the system goes down. It does the best effort to get NMI messages into
|
|
|
|
* the main ring buffer.
|
|
|
|
*
|
|
|
|
* Note that it could try harder when there is only one CPU online.
|
|
|
|
*/
|
2016-12-27 15:16:05 +01:00
|
|
|
void printk_safe_flush_on_panic(void)
|
2016-05-21 02:00:42 +02:00
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Make sure that we could access the main ring buffer.
|
|
|
|
* Do not risk a double release when more CPUs are up.
|
|
|
|
*/
|
2018-05-30 09:03:50 +02:00
|
|
|
if (raw_spin_is_locked(&logbuf_lock)) {
|
2016-05-21 02:00:42 +02:00
|
|
|
if (num_online_cpus() > 1)
|
|
|
|
return;
|
|
|
|
|
|
|
|
debug_locks_off();
|
|
|
|
raw_spin_lock_init(&logbuf_lock);
|
|
|
|
}
|
|
|
|
|
2016-12-27 15:16:05 +01:00
|
|
|
printk_safe_flush();
|
2016-05-21 02:00:42 +02:00
|
|
|
}
|
|
|
|
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
#ifdef CONFIG_PRINTK_NMI
|
|
|
|
/*
|
|
|
|
* Safe printk() for NMI context. It uses a per-CPU buffer to
|
|
|
|
* store the message. NMIs are not nested, so there is always only
|
|
|
|
* one writer running. But the buffer might get flushed from another
|
|
|
|
* CPU, so we need to be careful.
|
|
|
|
*/
|
2017-05-24 07:49:50 +02:00
|
|
|
static __printf(1, 0) int vprintk_nmi(const char *fmt, va_list args)
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
{
|
|
|
|
struct printk_safe_seq_buf *s = this_cpu_ptr(&nmi_print_seq);
|
|
|
|
|
|
|
|
return printk_safe_log_store(s, fmt, args);
|
|
|
|
}
|
|
|
|
|
2018-09-05 22:29:49 +02:00
|
|
|
void notrace printk_nmi_enter(void)
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
{
|
2018-06-27 16:20:28 +02:00
|
|
|
this_cpu_or(printk_context, PRINTK_NMI_CONTEXT_MASK);
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
}
|
|
|
|
|
2018-09-05 22:29:49 +02:00
|
|
|
void notrace printk_nmi_exit(void)
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
{
|
2018-06-27 16:20:28 +02:00
|
|
|
this_cpu_and(printk_context, ~PRINTK_NMI_CONTEXT_MASK);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Marks a code that might produce many messages in NMI context
|
|
|
|
* and the risk of losing them is more critical than eventual
|
|
|
|
* reordering.
|
|
|
|
*
|
|
|
|
* It has effect only when called in NMI context. Then printk()
|
|
|
|
* will try to store the messages into the main logbuf directly
|
|
|
|
* and use the per-CPU buffers only as a fallback when the lock
|
|
|
|
* is not available.
|
|
|
|
*/
|
|
|
|
void printk_nmi_direct_enter(void)
|
|
|
|
{
|
|
|
|
if (this_cpu_read(printk_context) & PRINTK_NMI_CONTEXT_MASK)
|
|
|
|
this_cpu_or(printk_context, PRINTK_NMI_DIRECT_CONTEXT_MASK);
|
|
|
|
}
|
|
|
|
|
|
|
|
void printk_nmi_direct_exit(void)
|
|
|
|
{
|
|
|
|
this_cpu_and(printk_context, ~PRINTK_NMI_DIRECT_CONTEXT_MASK);
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
#else
|
|
|
|
|
2017-05-24 07:49:50 +02:00
|
|
|
static __printf(1, 0) int vprintk_nmi(const char *fmt, va_list args)
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* CONFIG_PRINTK_NMI */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lock-less printk(), to avoid deadlocks should the printk() recurse
|
|
|
|
* into itself. It uses a per-CPU buffer to store the message, just like
|
|
|
|
* NMI.
|
|
|
|
*/
|
2017-05-24 07:49:50 +02:00
|
|
|
static __printf(1, 0) int vprintk_safe(const char *fmt, va_list args)
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
{
|
|
|
|
struct printk_safe_seq_buf *s = this_cpu_ptr(&safe_print_seq);
|
|
|
|
|
|
|
|
return printk_safe_log_store(s, fmt, args);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Can be preempted by NMI. */
|
|
|
|
void __printk_safe_enter(void)
|
|
|
|
{
|
|
|
|
this_cpu_inc(printk_context);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Can be preempted by NMI. */
|
|
|
|
void __printk_safe_exit(void)
|
|
|
|
{
|
|
|
|
this_cpu_dec(printk_context);
|
|
|
|
}
|
|
|
|
|
|
|
|
__printf(1, 0) int vprintk_func(const char *fmt, va_list args)
|
|
|
|
{
|
2018-06-27 16:20:28 +02:00
|
|
|
/*
|
|
|
|
* Try to use the main logbuf even in NMI. But avoid calling console
|
|
|
|
* drivers that might have their own locks.
|
|
|
|
*/
|
|
|
|
if ((this_cpu_read(printk_context) & PRINTK_NMI_DIRECT_CONTEXT_MASK) &&
|
|
|
|
raw_spin_trylock(&logbuf_lock)) {
|
|
|
|
int len;
|
|
|
|
|
|
|
|
len = vprintk_store(0, LOGLEVEL_DEFAULT, NULL, 0, fmt, args);
|
|
|
|
raw_spin_unlock(&logbuf_lock);
|
|
|
|
defer_console_output();
|
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
2017-04-20 10:52:31 +02:00
|
|
|
/* Use extra buffer in NMI when logbuf_lock is taken or in safe mode. */
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
if (this_cpu_read(printk_context) & PRINTK_NMI_CONTEXT_MASK)
|
|
|
|
return vprintk_nmi(fmt, args);
|
|
|
|
|
2017-04-20 10:52:31 +02:00
|
|
|
/* Use extra buffer to prevent a recursion deadlock in safe mode. */
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
if (this_cpu_read(printk_context) & PRINTK_SAFE_CONTEXT_MASK)
|
|
|
|
return vprintk_safe(fmt, args);
|
|
|
|
|
2017-04-20 10:52:31 +02:00
|
|
|
/* No obstacles. */
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
return vprintk_default(fmt, args);
|
|
|
|
}
|
|
|
|
|
2016-12-27 15:16:05 +01:00
|
|
|
void __init printk_safe_init(void)
|
2016-05-21 02:00:33 +02:00
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
for_each_possible_cpu(cpu) {
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
struct printk_safe_seq_buf *s;
|
|
|
|
|
|
|
|
s = &per_cpu(safe_print_seq, cpu);
|
|
|
|
init_irq_work(&s->work, __printk_safe_flush);
|
2016-05-21 02:00:33 +02:00
|
|
|
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
#ifdef CONFIG_PRINTK_NMI
|
|
|
|
s = &per_cpu(nmi_print_seq, cpu);
|
2016-12-27 15:16:05 +01:00
|
|
|
init_irq_work(&s->work, __printk_safe_flush);
|
printk: introduce per-cpu safe_print seq buffer
This patch extends the idea of NMI per-cpu buffers to regions
that may cause recursive printk() calls and possible deadlocks.
Namely, printk() can't handle printk calls from schedule code
or printk() calls from lock debugging code (spin_dump() for instance);
because those may be called with `sem->lock' already taken or any
other `critical' locks (p->pi_lock, etc.). An example of deadlock
can be
vprintk_emit()
console_unlock()
up() << raw_spin_lock_irqsave(&sem->lock, flags);
wake_up_process()
try_to_wake_up()
ttwu_queue()
ttwu_activate()
activate_task()
enqueue_task()
enqueue_task_fair()
cfs_rq_of()
task_of()
WARN_ON_ONCE(!entity_is_task(se))
vprintk_emit()
console_trylock()
down_trylock()
raw_spin_lock_irqsave(&sem->lock, flags)
^^^^ deadlock
and some other cases.
Just like in NMI implementation, the solution uses a per-cpu
`printk_func' pointer to 'redirect' printk() calls to a 'safe'
callback, that store messages in a per-cpu buffer and flushes
them back to logbuf buffer later.
Usage example:
printk()
printk_safe_enter_irqsave(flags)
//
// any printk() call from here will endup in vprintk_safe(),
// that stores messages in a special per-CPU buffer.
//
printk_safe_exit_irqrestore(flags)
The 'redirection' mechanism, though, has been reworked, as suggested
by Petr Mladek. Instead of using a per-cpu @print_func callback we now
keep a per-cpu printk-context variable and call either default or nmi
vprintk function depending on its value. printk_nmi_entrer/exit and
printk_safe_enter/exit, thus, just set/celar corresponding bits in
printk-context functions.
The patch only adds printk_safe support, we don't use it yet.
Link: http://lkml.kernel.org/r/20161227141611.940-4-sergey.senozhatsky@gmail.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Calvin Owens <calvinowens@fb.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2016-12-27 15:16:06 +01:00
|
|
|
#endif
|
2016-05-21 02:00:33 +02:00
|
|
|
}
|
|
|
|
|
2017-10-11 18:46:47 +02:00
|
|
|
/*
|
|
|
|
* In the highly unlikely event that a NMI were to trigger at
|
|
|
|
* this moment. Make sure IRQ work is set up before this
|
|
|
|
* variable is set.
|
|
|
|
*/
|
|
|
|
barrier();
|
2016-12-27 15:16:05 +01:00
|
|
|
printk_safe_irq_ready = 1;
|
2016-05-21 02:00:33 +02:00
|
|
|
|
|
|
|
/* Flush pending messages that did not have scheduled IRQ works. */
|
2016-12-27 15:16:05 +01:00
|
|
|
printk_safe_flush();
|
2016-05-21 02:00:33 +02:00
|
|
|
}
|