[lkml]   [2009]   [Sep]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subject[PATCH tip/core/rcu 0/3] rcu: simplify rcu_barrier() interaction with CPU hotplug
This patchset simplifies the interaction of rcu_barrier() with CPU
hotplug operations. The main point of this patchset is to impose an
invariant: offline CPUs never have any RCU callbacks queued. However,
rcutiny doesn't permit CPU hotplug (yes, I have worked in environments
where the last CPU could be offlined, but Linux is thankfully not one of
them), so the patchset goes as follows:

o Replace the rcu_barrier enum with a pointer to the relevant
call_rcu() function, thus eliminating any confusion about
which .h file this enum should reside in.

o Move the rcu_barrier() code to rcutree, and create a lightweight
variant of rcu_barrier() that is suitable for rcutiny. This
lightweight variant is identical to rcutree's implementation of
synchronize_rcu() and friends.

o Create a list in the rcu_state structure that holds RCU
callbacks that were "orphaned" by CPUs that just went offline.
The CPU_DYING notifier moves all RCU callbacks from the outgoing
CPU to the rcu_state lists, and the CPU_DEAD notifier and
_rcu_barrier() function "adopt" these orphans.

b/kernel/rcupdate.c | 33 +++-------
b/kernel/rcutiny.c | 36 ++++++++++
b/kernel/rcutree.c | 120 ++++++++++++++++++++++++++++++++++++
b/kernel/rcutree.h | 11 +++
b/kernel/rcutree_plugin.h | 34 ++++++++++
b/kernel/rcutree_trace.c | 5 -
kernel/rcupdate.c | 120 ------------------------------------
kernel/rcutree.c | 151 ++++++++++++++++++++++++----------------------
8 files changed, 291 insertions(+), 219 deletions(-)

 \ /
  Last update: 2009-09-29 06:53    [W:0.069 / U:1.020 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site