Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit 4eacdf18374e5d7d21a728b46dfec269ac8ef55c

Authored by Frederic Weisbecker
Committed by Paul E. McKenney
1 parent 90f45e4e72
Exists in smarc-l5.0.0_1.0.0-ga and in 5 other branches imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga

context_tracking: Add comments on interface and internals 

This subsystem lacks many explanations on its purpose and
design. Add these missing comments.

v4: Document function parameter to be more kernel-doc
friendly, as per Namhyung suggestion.

Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Alessio Igor Bogani <abogani@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Geoff Levand <geoff@infradead.org>
Cc: Gilad Ben Yossef <gilad@benyossef.com>
Cc: Hakan Akkan <hakanakkan@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Li Zhong <zhong@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

Showing 1 changed file with 65 additions and 10 deletions Side-by-side Diff

kernel/context_tracking.c
Diff comments   View file @ 4eacdf1
  1 +/*
  2 + * Context tracking: Probe on high level context boundaries such as kernel
  3 + * and userspace. This includes syscalls and exceptions entry/exit.
  4 + *
  5 + * This is used by RCU to remove its dependency on the timer tick while a CPU
  6 + * runs in userspace.
  7 + *
  8 + * Started by Frederic Weisbecker:
  9 + *
  10 + * Copyright (C) 2012 Red Hat, Inc., Frederic Weisbecker <fweisbec@redhat.com>
  11 + *
  12 + * Many thanks to Gilad Ben-Yossef, Paul McKenney, Ingo Molnar, Andrew Morton,
  13 + * Steven Rostedt, Peter Zijlstra for suggestions and improvements.
  14 + *
  15 + */
  16 +
1 17 #include <linux/context_tracking.h>
2 18 #include <linux/rcupdate.h>
3 19 #include <linux/sched.h>
... ... @@ -6,8 +22,8 @@
6 22  
7 23 struct context_tracking {
8 24 /*
9   - * When active is false, hooks are not set to
10   - * minimize overhead: TIF flags are cleared
  25 + * When active is false, probes are unset in order
  26 + * to minimize overhead: TIF flags are cleared
11 27 * and calls to user_enter/exit are ignored. This
12 28 * may be further optimized using static keys.
13 29 */
... ... @@ -24,6 +40,15 @@
24 40 #endif
25 41 };
26 42  
  43 +/**
  44 + * user_enter - Inform the context tracking that the CPU is going to
  45 + * enter userspace mode.
  46 + *
  47 + * This function must be called right before we switch from the kernel
  48 + * to userspace, when it's guaranteed the remaining kernel instructions
  49 + * to execute won't use any RCU read side critical section because this
  50 + * function sets RCU in extended quiescent state.
  51 + */
27 52 void user_enter(void)
28 53 {
29 54 unsigned long flags;
30 55  
31 56  
32 57  
33 58  
34 59  
... ... @@ -39,40 +64,70 @@
39 64 if (in_interrupt())
40 65 return;
41 66  
  67 + /* Kernel threads aren't supposed to go to userspace */
42 68 WARN_ON_ONCE(!current->mm);
43 69  
44 70 local_irq_save(flags);
45 71 if (__this_cpu_read(context_tracking.active) &&
46 72 __this_cpu_read(context_tracking.state) != IN_USER) {
47 73 __this_cpu_write(context_tracking.state, IN_USER);
  74 + /*
  75 + * At this stage, only low level arch entry code remains and
  76 + * then we'll run in userspace. We can assume there won't be
  77 + * any RCU read-side critical section until the next call to
  78 + * user_exit() or rcu_irq_enter(). Let's remove RCU's dependency
  79 + * on the tick.
  80 + */
48 81 rcu_user_enter();
49 82 }
50 83 local_irq_restore(flags);
51 84 }
52 85  
  86 +
  87 +/**
  88 + * user_exit - Inform the context tracking that the CPU is
  89 + * exiting userspace mode and entering the kernel.
  90 + *
  91 + * This function must be called after we entered the kernel from userspace
  92 + * before any use of RCU read side critical section. This potentially include
  93 + * any high level kernel code like syscalls, exceptions, signal handling, etc...
  94 + *
  95 + * This call supports re-entrancy. This way it can be called from any exception
  96 + * handler without needing to know if we came from userspace or not.
  97 + */
53 98 void user_exit(void)
54 99 {
55 100 unsigned long flags;
56 101  
57   - /*
58   - * Some contexts may involve an exception occuring in an irq,
59   - * leading to that nesting:
60   - * rcu_irq_enter() rcu_user_exit() rcu_user_exit() rcu_irq_exit()
61   - * This would mess up the dyntick_nesting count though. And rcu_irq_*()
62   - * helpers are enough to protect RCU uses inside the exception. So
63   - * just return immediately if we detect we are in an IRQ.
64   - */
65 102 if (in_interrupt())
66 103 return;
67 104  
68 105 local_irq_save(flags);
69 106 if (__this_cpu_read(context_tracking.state) == IN_USER) {
70 107 __this_cpu_write(context_tracking.state, IN_KERNEL);
  108 + /*
  109 + * We are going to run code that may use RCU. Inform
  110 + * RCU core about that (ie: we may need the tick again).
  111 + */
71 112 rcu_user_exit();
72 113 }
73 114 local_irq_restore(flags);
74 115 }
75 116  
  117 +
  118 +/**
  119 + * context_tracking_task_switch - context switch the syscall callbacks
  120 + * @prev: the task that is being switched out
  121 + * @next: the task that is being switched in
  122 + *
  123 + * The context tracking uses the syscall slow path to implement its user-kernel
  124 + * boundaries probes on syscalls. This way it doesn't impact the syscall fast
  125 + * path on CPUs that don't do context tracking.
  126 + *
  127 + * But we need to clear the flag on the previous task because it may later
  128 + * migrate to some CPU that doesn't do the context tracking. As such the TIF
  129 + * flag may not be desired there.
  130 + */
76 131 void context_tracking_task_switch(struct task_struct *prev,
77 132 struct task_struct *next)
78 133 {