Commit 6771853b authored by Paul E. McKenney's avatar Paul E. McKenney

doc: Update requirements based on recent changes

These changes include lighter-weight expedited grace periods, the fact
that expedited grace periods and rcu_barrier() no longer block CPU
hotplug, some HTML font fixups, noting that rcu_barrier() need not wait
for a grace period (even if callbacks are posted), the fact that SRCU
read-side critical sections can be used from offline CPUs, and the fact
that SRCU now maintains per-CPU callback lists.
Signed-off-by: default avatarPaul E. McKenney <paulmck@linux.vnet.ibm.com>
parent aa123a74
......@@ -659,8 +659,9 @@ systems with more than one CPU:
In other words, a given instance of <tt>synchronize_rcu()</tt>
can avoid waiting on a given RCU read-side critical section only
if it can prove that <tt>synchronize_rcu()</tt> started first.
</font>
<p>
<p><font color="ffffff">
A related question is &ldquo;When <tt>rcu_read_lock()</tt>
doesn't generate any code, why does it matter how it relates
to a grace period?&rdquo;
......@@ -675,8 +676,9 @@ systems with more than one CPU:
within the critical section, in which case none of the accesses
within the critical section may observe the effects of any
access following the grace period.
</font>
<p>
<p><font color="ffffff">
As of late 2016, mathematical models of RCU take this
viewpoint, for example, see slides&nbsp;62 and&nbsp;63
of the
......@@ -1616,8 +1618,8 @@ CPUs should at least make reasonable forward progress.
In return for its shorter latencies, <tt>synchronize_rcu_expedited()</tt>
is permitted to impose modest degradation of real-time latency
on non-idle online CPUs.
That said, it will likely be necessary to take further steps to reduce this
degradation, hopefully to roughly that of a scheduling-clock interrupt.
Here, &ldquo;modest&rdquo; means roughly the same latency
degradation as a scheduling-clock interrupt.
<p>
There are a number of situations where even
......@@ -1913,12 +1915,9 @@ This requirement is another factor driving batching of grace periods,
but it is also the driving force behind the checks for large numbers
of queued RCU callbacks in the <tt>call_rcu()</tt> code path.
Finally, high update rates should not delay RCU read-side critical
sections, although some read-side delays can occur when using
sections, although some small read-side delays can occur when using
<tt>synchronize_rcu_expedited()</tt>, courtesy of this function's use
of <tt>try_stop_cpus()</tt>.
(In the future, <tt>synchronize_rcu_expedited()</tt> will be
converted to use lighter-weight inter-processor interrupts (IPIs),
but this will still disturb readers, though to a much smaller degree.)
of <tt>smp_call_function_single()</tt>.
<p>
Although all three of these corner cases were understood in the early
......@@ -2192,7 +2191,7 @@ Unfortunately, <tt>synchronize_rcu()</tt> can't do this until all of
its kthreads are spawned, which doesn't happen until some time during
<tt>early_initcalls()</tt> time.
But this is no excuse: RCU is nevertheless required to correctly handle
synchronous grace periods during this time period, which it currently does.
synchronous grace periods during this time period.
Once all of its kthreads are up and running, RCU starts running
normally.
......@@ -2206,8 +2205,10 @@ normally.
<tr><th align="left">Answer:</th></tr>
<tr><td bgcolor="#ffffff"><font color="ffffff">
Very carefully!
</font>
<p>During the &ldquo;dead zone&rdquo; between the time that the
<p><font color="ffffff">
During the &ldquo;dead zone&rdquo; between the time that the
scheduler spawns the first task and the time that all of RCU's
kthreads have been spawned, all synchronous grace periods are
handled by the expedited grace-period mechanism.
......@@ -2220,8 +2221,10 @@ normally.
using workqueues, as is required to avoid problems that would
otherwise occur when a user task received a POSIX signal while
driving an expedited grace period.
</font>
<p>And yes, this does mean that it is unhelpful to send POSIX
<p><font color="ffffff">
And yes, this does mean that it is unhelpful to send POSIX
signals to random tasks between the time that the scheduler
spawns its first kthread and the time that RCU's kthreads
have all been spawned.
......@@ -2308,12 +2311,61 @@ situation, and Dipankar Sarma incorporated <tt>rcu_barrier()</tt> into RCU.
The need for <tt>rcu_barrier()</tt> for module unloading became
apparent later.
<p>
<b>Important note</b>: The <tt>rcu_barrier()</tt> function is not,
repeat, <i>not</i>, obligated to wait for a grace period.
It is instead only required to wait for RCU callbacks that have
already been posted.
Therefore, if there are no RCU callbacks posted anywhere in the system,
<tt>rcu_barrier()</tt> is within its rights to return immediately.
Even if there are callbacks posted, <tt>rcu_barrier()</tt> does not
necessarily need to wait for a grace period.
<table>
<tr><th>&nbsp;</th></tr>
<tr><th align="left">Quick Quiz:</th></tr>
<tr><td>
Wait a minute!
Each RCU callbacks must wait for a grace period to complete,
and <tt>rcu_barrier()</tt> must wait for each pre-existing
callback to be invoked.
Doesn't <tt>rcu_barrier()</tt> therefore need to wait for
a full grace period if there is even one callback posted anywhere
in the system?
</td></tr>
<tr><th align="left">Answer:</th></tr>
<tr><td bgcolor="#ffffff"><font color="ffffff">
Absolutely not!!!
</font>
<p><font color="ffffff">
Yes, each RCU callbacks must wait for a grace period to complete,
but it might well be partly (or even completely) finished waiting
by the time <tt>rcu_barrier()</tt> is invoked.
In that case, <tt>rcu_barrier()</tt> need only wait for the
remaining portion of the grace period to elapse.
So even if there are quite a few callbacks posted,
<tt>rcu_barrier()</tt> might well return quite quickly.
</font>
<p><font color="ffffff">
So if you need to wait for a grace period as well as for all
pre-existing callbacks, you will need to invoke both
<tt>synchronize_rcu()</tt> and <tt>rcu_barrier()</tt>.
If latency is a concern, you can always use workqueues
to invoke them concurrently.
</font></td></tr>
<tr><td>&nbsp;</td></tr>
</table>
<h3><a name="Hotplug CPU">Hotplug CPU</a></h3>
<p>
The Linux kernel supports CPU hotplug, which means that CPUs
can come and go.
It is of course illegal to use any RCU API member from an offline CPU.
It is of course illegal to use any RCU API member from an offline CPU,
with the exception of <a href="#Sleepable RCU">SRCU</a> read-side
critical sections.
This requirement was present from day one in DYNIX/ptx, but
on the other hand, the Linux kernel's CPU-hotplug implementation
is &ldquo;interesting.&rdquo;
......@@ -2323,19 +2375,18 @@ The Linux-kernel CPU-hotplug implementation has notifiers that
are used to allow the various kernel subsystems (including RCU)
to respond appropriately to a given CPU-hotplug operation.
Most RCU operations may be invoked from CPU-hotplug notifiers,
including even normal synchronous grace-period operations
such as <tt>synchronize_rcu()</tt>.
However, expedited grace-period operations such as
<tt>synchronize_rcu_expedited()</tt> are not supported,
due to the fact that current implementations block CPU-hotplug
operations, which could result in deadlock.
including even synchronous grace-period operations such as
<tt>synchronize_rcu()</tt> and <tt>synchronize_rcu_expedited()</tt>.
<p>
In addition, all-callback-wait operations such as
However, all-callback-wait operations such as
<tt>rcu_barrier()</tt> are also not supported, due to the
fact that there are phases of CPU-hotplug operations where
the outgoing CPU's callbacks will not be invoked until after
the CPU-hotplug operation ends, which could also result in deadlock.
Furthermore, <tt>rcu_barrier()</tt> blocks CPU-hotplug operations
during its execution, which results in another type of deadlock
when invoked from a CPU-hotplug notifier.
<h3><a name="Scheduler and RCU">Scheduler and RCU</a></h3>
......@@ -2876,6 +2927,27 @@ It also motivates the <tt>smp_mb__after_srcu_read_unlock()</tt>
API, which, in combination with <tt>srcu_read_unlock()</tt>,
guarantees a full memory barrier.
<p>
Also unlike other RCU flavors, SRCU's callbacks-wait function
<tt>srcu_barrier()</tt> may be invoked from CPU-hotplug notifiers,
though this is not necessarily a good idea.
The reason that this is possible is that SRCU is insensitive
to whether or not a CPU is online, which means that <tt>srcu_barrier()</tt>
need not exclude CPU-hotplug operations.
<p>
As of v4.12, SRCU's callbacks are maintained per-CPU, eliminating
a locking bottleneck present in prior kernel versions.
Although this will allow users to put much heavier stress on
<tt>call_srcu()</tt>, it is important to note that SRCU does not
yet take any special steps to deal with callback flooding.
So if you are posting (say) 10,000 SRCU callbacks per second per CPU,
you are probably totally OK, but if you intend to post (say) 1,000,000
SRCU callbacks per second per CPU, please run some tests first.
SRCU just might need a few adjustment to deal with that sort of load.
Of course, your mileage may vary based on the speed of your CPUs and
the size of your memory.
<p>
The
<a href="https://lwn.net/Articles/609973/#RCU Per-Flavor API Table">SRCU API</a>
......@@ -3034,8 +3106,8 @@ to do some redesign to avoid this scalability problem.
<p>
RCU disables CPU hotplug in a few places, perhaps most notably in the
expedited grace-period and <tt>rcu_barrier()</tt> operations.
If there is a strong reason to use expedited grace periods in CPU-hotplug
<tt>rcu_barrier()</tt> operations.
If there is a strong reason to use <tt>rcu_barrier()</tt> in CPU-hotplug
notifiers, it will be necessary to avoid disabling CPU hotplug.
This would introduce some complexity, so there had better be a <i>very</i>
good reason.
......@@ -3109,9 +3181,5 @@ Andy Lutomirski for their help in rendering
this article human readable, and to Michelle Rankin for her support
of this effort.
Other contributions are acknowledged in the Linux kernel's git archive.
The cartoon is copyright (c) 2013 by Melissa Broussard,
and is provided
under the terms of the Creative Commons Attribution-Share Alike 3.0
United States license.
</body></html>
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment