This patch fixes a long-standing performance bug in classic RCU that
results in massive internal-to-RCU lock contention on systems with
more than a few hundred CPUs. Although this patch creates a separate
flavor of RCU for ease of review and patch maintenance, it is intended
to replace classic RCU.
This patch still handles stress better than does mainline, so I am still
calling it ready for inclusion. This patch is against the -tip tree.
Nevertheless, experience on an actual 1000+ CPU machine would still be
most welcome.
Most of the changes noted below were found while creating an rcutiny
(which should permit ejecting the current rcuclassic) and while doing
detailed line-by-line documentation.
Updates from v9 (http://lkml.org/lkml/2008/12/2/334):
o Fixes from remainder of line-by-line code walkthrough,
including comment spelling, initialization, undesirable
narrowing due to type conversion, removing redundant memory
barriers, removing redundant local-variable initialization,
and removing redundant local variables.
I do not believe that any of these fixes address the CPU-hotplug
issues that Andi Kleen was seeing, but please do give it a whirl
in case the machine is smarter than I am.
A writeup from the walkthrough may be found at the following
URL, in case you are suffering from terminal insomnia or
masochism:
http://www.kernel.org/pub/linux/kernel/people/paulmck/tmp/rcutree-walkthrough.2008.12.16a.pdf
o Made rcutree tracing use seq_file, as suggested some time
ago by Lai Jiangshan.
o Added a .csv variant of the rcudata debugfs trace file, to allow
people having thousands of CPUs to drop the data into
a spreadsheet. Tested with oocalc and gnumeric. Updated
documentation to suit.
Updates from v8 (http://lkml.org/lkml/2008/11/15/139):
o Fix a theoretical race between grace-period initialization and
force_quiescent_state() that could occur if more than three
jiffies were required to carry out the grace-period
initialization. Which it might, if you had enough CPUs.
o Apply Ingo's printk-standardization patch.
o Substitute local variables for repeated accesses to global
variables.
o Fix comment misspellings and redundant (but harmless) increments
of ->n_rcu_pending (this latter after having explicitly added it).
o Apply checkpatch fixes.
Updates from v7 (http://lkml.org/lkml/2008/10/10/291):
o Fixed a number of problems noted by Gautham Shenoy, including
the cpu-stall-detection bug that he was having difficulty
convincing me was real. ;-)
o Changed cpu-stall detection to wait for ten seconds rather than
three in order to reduce false positive, as suggested by Ingo
Molnar.
o Produced a design document (http://lwn.net/Articles/305782/).
The act of writing this document uncovered a number of both
theoretical and "here and now" bugs as noted below.
o Fix dynticks_nesting accounting confusion, simplify WARN_ON()
condition, fix kerneldoc comments, and add memory barriers
in dynticks interface functions.
o Add more data to tracing.
o Remove unused "rcu_barrier" field from rcu_data structure.
o Count calls to rcu_pending() from scheduling-clock interrupt
to use as a surrogate timebase should jiffies stop counting.
o Fix a theoretical race between force_quiescent_state() and
grace-period initialization. Yes, initialization does have to
go on for some jiffies for this race to occur, but given enough
CPUs...
Updates from v6 (http://lkml.org/lkml/2008/9/23/448):
o Fix a number of checkpatch.pl complaints.
o Apply review comments from Ingo Molnar and Lai Jiangshan
on the stall-detection code.
o Fix several bugs in !CONFIG_SMP builds.
o Fix a misspelled config-parameter name so that RCU now announces
at boot time if stall detection is configured.
o Run tests on numerous combinations of configurations parameters,
which after the fixes above, now build and run correctly.
Updates from v5 (http://lkml.org/lkml/2008/9/15/92, bad subject line):
o Fix a compiler error in the !CONFIG_FANOUT_EXACT case (blew a
changeset some time ago, and finally got around to retesting
this option).
o Fix some tracing bugs in rcupreempt that caused incorrect
totals to be printed.
o I now test with a more brutal random-selection online/offline
script (attached). Probably more brutal than it needs to be
on the people reading it as well, but so it goes.
o A number of optimizations and usability improvements:
o Make rcu_pending() ignore the grace-period timeout when
there is no grace period in progress.
o Make force_quiescent_state() avoid going for a global
lock in the case where there is no grace period in
progress.
o Rearrange struct fields to improve struct layout.
o Make call_rcu() initiate a grace period if RCU was
idle, rather than waiting for the next scheduling
clock interrupt.
o Invoke rcu_irq_enter() and rcu_irq_exit() only when
idle, as suggested by Andi Kleen. I still don't
completely trust this change, and might back it out.
o Make CONFIG_RCU_TRACE be the single config variable
manipulated for all forms of RCU, instead of the prior
confusion.
o Document tracing files and formats for both rcupreempt
and rcutree.
Updates from v4 for those missing v5 given its bad subject line:
o Separated dynticks interface so that NMIs and irqs call separate
functions, greatly simplifying it. In particular, this code
no longer requires a proof of correctness. ;-)
o Separated dynticks state out into its own per-CPU structure,
avoiding the duplicated accounting.
o The case where a dynticks-idle CPU runs an irq handler that
invokes call_rcu() is now correctly handled, forcing that CPU
out of dynticks-idle mode.
o Review comments have been applied (thank you all!!!).
For but one example, fixed the dynticks-ordering issue that
Manfred pointed out, saving me much debugging. ;-)
o Adjusted rcuclassic and rcupreempt to handle dynticks changes.
Attached is an updated patch to Classic RCU that applies a hierarchy,
greatly reducing the contention on the top-level lock for large machines.
This passes 10-hour concurrent rcutorture and online-offline testing on
128-CPU ppc64 without dynticks enabled, and exposes some timekeeping
bugs in presence of dynticks (exciting working on a system where
"sleep 1" hangs until interrupted...), which were fixed in the
2.6.27 kernel. It is getting more reliable than mainline by some
measures, so the next version will be against -tip for inclusion.
See also Manfred Spraul's recent patches (or his earlier work from
2004 at http://marc.info/?l=linux-kernel&m=108546384711797&w=2).
We will converge onto a common patch in the fullness of time, but are
currently exploring different regions of the design space. That said,
I have already gratefully stolen quite a few of Manfred's ideas.
This patch provides CONFIG_RCU_FANOUT, which controls the bushiness
of the RCU hierarchy. Defaults to 32 on 32-bit machines and 64 on
64-bit machines. If CONFIG_NR_CPUS is less than CONFIG_RCU_FANOUT,
there is no hierarchy. By default, the RCU initialization code will
adjust CONFIG_RCU_FANOUT to balance the hierarchy, so strongly NUMA
architectures may choose to set CONFIG_RCU_FANOUT_EXACT to disable
this balancing, allowing the hierarchy to be exactly aligned to the
underlying hardware. Up to two levels of hierarchy are permitted
(in addition to the root node), allowing up to 16,384 CPUs on 32-bit
systems and up to 262,144 CPUs on 64-bit systems. I just know that I
am going to regret saying this, but this seems more than sufficient
for the foreseeable future. (Some architectures might wish to set
CONFIG_RCU_FANOUT=4, which would limit such architectures to 64 CPUs.
If this becomes a real problem, additional levels can be added, but I
doubt that it will make a significant difference on real hardware.)
In the common case, a given CPU will manipulate its private rcu_data
structure and the rcu_node structure that it shares with its immediate
neighbors. This can reduce both lock and memory contention by multiple
orders of magnitude, which should eliminate the need for the strange
manipulations that are reported to be required when running Linux on
very large systems.
Some shortcomings:
o More bugs will probably surface as a result of an ongoing
line-by-line code inspection.
Patches will be provided as required.
o There are probably hangs, rcutorture failures, &c. Seems
quite stable on a 128-CPU machine, but that is kind of small
compared to 4096 CPUs. However, seems to do better than
mainline.
Patches will be provided as required.
o The memory footprint of this version is several KB larger
than rcuclassic.
A separate UP-only rcutiny patch will be provided, which will
reduce the memory footprint significantly, even compared
to the old rcuclassic. One such patch passes light testing,
and has a memory footprint smaller even than rcuclassic.
Initial reaction from various embedded guys was "it is not
worth it", so am putting it aside.
Credits:
o Manfred Spraul for ideas, review comments, and bugs spotted,
as well as some good friendly competition. ;-)
o Josh Triplett, Ingo Molnar, Peter Zijlstra, Mathieu Desnoyers,
Lai Jiangshan, Andi Kleen, Andy Whitcroft, and Andrew Morton
for reviews and comments.
o Thomas Gleixner for much-needed help with some timer issues
(see patches below).
o Jon M. Tollefson, Tim Pepper, Andrew Theurer, Jose R. Santos,
Andy Whitcroft, Darrick Wong, Nishanth Aravamudan, Anton
Blanchard, Dave Kleikamp, and Nathan Lynch for keeping machines
alive despite my heavy abuse^Wtesting.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch fixes a comment and clarifies the documentation about the
endianness of descriptors. The current policy is that descriptors will
be little-endian at the API even on big-endian systems; however the
/proc/bus/usb API predates this policy and presents descriptors with
some multibyte fields byte-swapped.
Signed-off-by: Phil Endecott <usb_endian_patch@chezphil.org>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Impact: generalize the sw-IOTLB range checks
Some architectures require special rules to determine whether a range
needs mapping or not. This adds a weak function for architectures to
override.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: generalize phys<->bus<->phys conversions in the swiotlb code
Architectures may need to override these conversions. Implement a
__weak hook point containing the default implementation.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: generalize swiotlb allocation code
Architectures may need to allocate memory specially for use with
the swiotlb. Create the weak function swiotlb_alloc_boot() and
swiotlb_alloc() defaulting to the current behaviour.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch fixes an inconsistency in nfnetlink_conntrack.h that
I introduced myself. The problem is that CTA_NAT_SEQ_UNSPEC is
missing from enum ctattr_natseq. This inconsistency may lead to
problems in the message parsing in userspace (if the message
contains the CTA_NAT_SEQ_* attributes, of course).
This patch breaks backward compatibility, however, the only known
client of this code is libnetfilter_conntrack which indeed crashes
because it assumes the existence of CTA_NAT_SEQ_UNSPEC to do
the parsing.
The CTA_NAT_SEQ_* attributes were introduced in 2.6.25.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
Phonet: keep TX queue disabled when the device is off
SCHED: netem: Correct documentation comment in code.
netfilter: update rwlock initialization for nat_table
netlabel: Compiler warning and NULL pointer dereference fix
e1000e: fix double release of mutex
IA64: HP_SIMETH needs to depend upon NET
netpoll: fix race on poll_list resulting in garbage entry
ipv6: silence log messages for locally generated multicast
sungem: improve ethtool output with internal pcs and serdes
tcp: tcp_vegas cong avoid fix
sungem: Make PCS PHY support partially work again.
Otherwise those using it in transition patches (eg. kvm) can't compile
with CONFIG_SMP=n:
arch/x86/kvm/../../../virt/kvm/kvm_main.c: In function 'make_all_cpus_request':
arch/x86/kvm/../../../virt/kvm/kvm_main.c:380: error: implicit declaration of function 'smp_call_function_many'
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Miles Lane tailing /sys files hit a BUG which Pekka Enberg has tracked
to my 966c8c12dc sprint_symbol(): use
less stack exposing a bug in slub's list_locations() -
kallsyms_lookup() writes a 0 to namebuf[KSYM_NAME_LEN-1], but that was
beyond the end of page provided.
The 100 slop which list_locations() allows at end of page looks roughly
enough for all the other stuff it might print after the symbol before
it checks again: break out KSYM_SYMBOL_LEN earlier than before.
Latencytop and ftrace and are using KSYM_NAME_LEN buffers where they
need KSYM_SYMBOL_LEN buffers, and vmallocinfo a 2*KSYM_NAME_LEN buffer
where it wants a KSYM_SYMBOL_LEN buffer: fix those before anyone copies
them.
[akpm@linux-foundation.org: ftrace.h needs module.h]
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc Miles Lane <miles.lane@gmail.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Steven Rostedt <srostedt@redhat.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Revert
commit 1f7c14c62c
Author: Mingming Cao <cmm@us.ibm.com>
Date: Thu Oct 9 12:50:59 2008 -0400
percpu counter: clean up percpu_counter_sum_and_set()
Before this patch we had the following:
percpu_counter_sum(): return the percpu_counter's value
percpu_counter_sum_and_set(): return the percpu_counter's value, copying
that value into the central value and zeroing the per-cpu counters before
returning.
After this patch, percpu_counter_sum_and_set() has gone, and
percpu_counter_sum() gets the old percpu_counter_sum_and_set()
functionality.
Problem is, as Eric points out, the old percpu_counter_sum_and_set()
functionality was racy and wrong. It zeroes out counters on "other" cpus,
without holding any locks which will prevent races agaist updates from
those other CPUS.
This patch reverts 1f7c14c62c. This means
that percpu_counter_sum_and_set() still has the race, but
percpu_counter_sum() does not.
Note that this is not a simple revert - ext4 has since started using
percpu_counter_sum() for its dirty_blocks counter as well.
Note that this revert patch changes percpu_counter_sum() semantics.
Before the patch, a call to percpu_counter_sum() will bring the counter's
central counter mostly up-to-date, so a following percpu_counter_read()
will return a close value.
After this patch, a call to percpu_counter_sum() will leave the counter's
central accumulator unaltered, so a subsequent call to
percpu_counter_read() can now return a significantly inaccurate result.
If there is any code in the tree which was introduced after
e8ced39d5e was merged, and which depends
upon the new percpu_counter_sum() semantics, that code will break.
Reported-by: Eric Dumazet <dada1@cosmosbay.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mingming Cao <cmm@us.ibm.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A few months back a race was discused between the netpoll napi service
path, and the fast path through net_rx_action:
http://kerneltrap.org/mailarchive/linux-netdev/2007/10/16/345470
A patch was submitted for that bug, but I think we missed a case.
Consider the following scenario:
INITIAL STATE
CPU0 has one napi_struct A on its poll_list
CPU1 is calling netpoll_send_skb and needs to call poll_napi on the same
napi_struct A that CPU0 has on its list
CPU0 CPU1
net_rx_action poll_napi
!list_empty (returns true) locks poll_lock for A
poll_one_napi
napi->poll
netif_rx_complete
__napi_complete
(removes A from poll_list)
list_entry(list->next)
In the above scenario, net_rx_action assumes that the per-cpu poll_list is
exclusive to that cpu. netpoll of course violates that, and because the netpoll
path can dequeue from the poll list, its possible for CPU0 to detect a non-empty
list at the top of the while loop in net_rx_action, but have it become empty by
the time it calls list_entry. Since the poll_list isn't surrounded by any other
structure, the returned data from that list_entry call in this situation is
garbage, and any number of crashes can result based on what exactly that garbage
is.
Given that its not fasible for performance reasons to place exclusive locks
arround each cpus poll list to provide that mutal exclusion, I think the best
solution is modify the netpoll path in such a way that we continue to guarantee
that the poll_list for a cpu is in fact exclusive to that cpu. To do this I've
implemented the patch below. It adds an additional bit to the state field in
the napi_struct. When executing napi->poll from the netpoll_path, this bit will
be set. When a driver calls netif_rx_complete, if that bit is set, it will not
remove the napi_struct from the poll_list. That work will be saved for the next
iteration of net_rx_action.
I've tested this and it seems to work well. About the biggest drawback I can
see to it is the fact that it might result in an extra loop through
net_rx_action in the event that the device is actually contended for (i.e. the
netpoll path actually preforms all the needed work no the device, and the call
to net_rx_action winds up doing nothing, except removing the napi_struct from
the poll_list. However I think this is probably a small price to pay, given
that the alternative is a crash.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'audit.b59' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current:
[PATCH] fix broken timestamps in AVC generated by kernel threads
[patch 1/1] audit: remove excess kernel-doc
[PATCH] asm/generic: fix bug - kernel fails to build when enable some common audit code on Blackfin
[PATCH] return records for fork() both to child and parent
[PATCH] Audit: make audit=0 actually turn off audit
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
tproxy: fixe a possible read from an invalid location in the socket match
zd1211rw: use unaligned safe memcmp() in-place of compare_ether_addr()
mac80211: use unaligned safe memcmp() in-place of compare_ether_addr()
ipw2200: fix netif_*_queue() removal regression
iwlwifi: clean key table in iwl_clear_stations_table function
tcp: tcp_vegas ssthresh bug fix
can: omit received RTR frames for single ID filter lists
ATM: CVE-2008-5079: duplicate listen() on socket corrupts the vcc table
netx-eth: initialize per device spinlock
tcp: make urg+gso work for real this time
enc28j60: Fix sporadic packet loss (corrected again)
hysdn: fix writing outside the field on 64 bits
b1isa: fix b1isa_exit() to really remove registered capi controllers
can: Fix CAN_(EFF|RTR)_FLAG handling in can_filter
Phonet: do not dump addresses from other namespaces
netlabel: Fix a potential NULL pointer dereference
bnx2: Add workaround to handle missed MSI.
xfrm: Fix kernel panic when flush and dump SPD entries
There's no point in having too short SG_IO timeouts, since if the
command does end up timing out, we'll end up through the reset sequence
that is several seconds long in order to abort the command that timed
out.
As a result, shorter timeouts than a few seconds simply do not make
sense, as the recovery would be longer than the timeout itself.
Add a BLK_MIN_SG_TIMEOUT to match the existign BLK_DEFAULT_SG_TIMEOUT.
Suggested-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Jens Axboe <jens.axboe@oracle.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Make sure all FMODE_ constants are documents, and ensure a coherent
style for the already existing comments.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Update FMODE_NDELAY before each ioctl call so that we can kill the
magic FMODE_NDELAY_NOW. It would be even better to do this directly
in setfl(), but for that we'd need to have FMODE_NDELAY for all files,
not just block special files.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
