| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <20180626000841.GA15696@linux.vnet.ibm.com>
Date: Mon, 25 Jun 2018 17:08:41 -0700
From: "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To: linux-kernel@...r.kernel.org
Cc: mingo@...nel.org, jiangshanlai@...il.com, dipankar@...ibm.com,
akpm@...ux-foundation.org, mathieu.desnoyers@...icios.com,
josh@...htriplett.org, tglx@...utronix.de, peterz@...radead.org,
rostedt@...dmis.org, dhowells@...hat.com, edumazet@...gle.com,
fweisbec@...il.com, oleg@...hat.com, joel@...lfernandes.org
Subject: [PATCH tip/core/rcu 0/40] RCU ->gp_seq conversion for v4.19
Hello!
This series moves from the two-variable ->gpnum and ->completed to
a single-variable ->gp_seq approach. The key point is that ->gpnum
and ->completed are either identical or ->gpnum is only one greater,
which means that we are using a full unsigned long were a single bit
would do. This series therefore applies the pre-existing grace-period
sequence-number code used by expedited grace periods, _rcu_barrier(),
and SRCU. This change also has the beneficial effect of allowing the
grace-period state to be sampled with a single load, which will be used
to improve the grace-period start code. The individual commits are
as follows:
1. Introduce grace-period sequence numbers (->gp_seq) to normal RCU.
2-17. Make core RCU code use ->gp_seq instead of ->gpnum and ->completed.
18. Move from ->need_future_gp[] to ->gp_seq_needed, which allows
simpler grace-period-already-requested checks.
19. Use the new ->gp_seq_needed to make rcu_nocb_wait_gp() check if
a given grace has already been requested, thus reducing overhead
for no-CBs CPUs.
20-26. Move RCU tracepoints to ->gp_seq.
27. Remove ->gpnum and ->completed.
28. Make simple callback acceleration refer to rdp->gp_seq_needed.
29-31. Update documentation to reflect ->gp_seq.
32. Don't funnel-lock above the leaf rcu_node structure if a grace period
is in progress, thus saving an upper-level rcu_node ->lock
acquisition.
33. Produce the last CleanupMore trace only if the corresponding
grace-period request was quite recent, courtesy of Joel Fernandes.
34. Fix cpustart tracepoint ->gp_seq number, courtesy of Joel Fernandes.
35. Make rcu_start_this_gp() check for grace period already started.
36. Correctly handle grace-period sequence (->gp_seq) wrap in
rcutorture.
37. Regularize resetting of rcu_data wrap indicator.
38-39. Rename grace-period-request variables and parameters, courtesy
of Joel Fernandes.
40. Check current node rather than leaf node for requested grace
period already being in progress.
Thanx, Paul
------------------------------------------------------------------------
Documentation/RCU/Design/Data-Structures/Data-Structures.html | 118 -
Documentation/RCU/Design/Memory-Ordering/Tree-RCU-Memory-Ordering.html | 22
Documentation/RCU/Design/Memory-Ordering/TreeRCU-gp-cleanup.svg | 123 -
Documentation/RCU/Design/Memory-Ordering/TreeRCU-gp-init-1.svg | 16
Documentation/RCU/Design/Memory-Ordering/TreeRCU-gp-init-3.svg | 56
Documentation/RCU/Design/Memory-Ordering/TreeRCU-gp.svg | 237 +--
Documentation/RCU/Design/Memory-Ordering/TreeRCU-qs.svg | 12
Documentation/RCU/stallwarn.txt | 24
include/trace/events/rcu.h | 114 -
kernel/rcu/rcu.h | 70 -
kernel/rcu/rcuperf.c | 44
kernel/rcu/rcutorture.c | 93 -
kernel/rcu/srcutree.c | 5
kernel/rcu/tree.c | 678 ++++------
kernel/rcu/tree.h | 40
kernel/rcu/tree_plugin.h | 54
16 files changed, 832 insertions(+), 874 deletions(-)
Powered by blists - more mailing lists