Michael Wang [Thu, 11 Oct 2012 21:26:42 +0000 (14:26 -0700)]
rcu: Remove old debugfs interfaces and also RCU flavor name
This commit removes the old debugfs interfaces, so that the new
directory-per-RCU-flavor versions remain. Because the RCU flavor is
given by the directory name, there is no need to print it out, so remove
the name from the printout.
Signed-off-by: Michael Wang <wangyun@linux.vnet.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Paul E. McKenney [Fri, 28 Sep 2012 17:49:58 +0000 (10:49 -0700)]
rcu: Fix tracing formatting
The rcu_state structure's ->completed field is unsigned long, so this
commit adjusts show_one_rcugp()'s printf() format to suit. Also add
the required ACCESS_ONCE() directives while we are in this function.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Michael Wang [Thu, 20 Sep 2012 00:51:06 +0000 (08:51 +0800)]
rcu: Optimize the 'rcu_pending' for RCU trace
This patch implements the new 'rcu_pending' interface under each rsp
directory, by using the 'CPU units sequence reading', thus avoiding loss
of tracing data.
Signed-off-by: Michael Wang <wangyun@linux.vnet.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Michael Wang [Thu, 20 Sep 2012 00:51:05 +0000 (08:51 +0800)]
rcu: Optimize the 'rcudata.csv' for RCU trace
This patch implements the new 'rcudata.csv' interface under each rsp
directory, by using the 'CPU units sequence reading', thus avoiding loss
of tracing data.
Signed-off-by: Michael Wang <wangyun@linux.vnet.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Michael Wang [Thu, 20 Sep 2012 00:51:04 +0000 (08:51 +0800)]
rcu: Optimize the 'rcudata' for RCU trace
This patch implements the new 'rcudata' interface under each rsp
directory, by using the 'CPU units sequence reading', thus avoiding loss
of tracing data.
Signed-off-by: Michael Wang <wangyun@linux.vnet.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Michael Wang [Thu, 20 Sep 2012 00:51:03 +0000 (08:51 +0800)]
rcu: Fundamental facility for 'CPU units sequence reading'
This patch add the fundamental facility used by the following patches, so we
can implement the 'CPU units sequence reading' later.
This helps us avoid losing data when there are too many CPUs and too
small of a buffer, since this new approach allows userspace to read out
the data one CPU at a time. Thus, if the buffer is not large enough,
userspace will get whatever CPUs fit, and can then issue another read
for the remainder of the data.
Signed-off-by: Michael Wang <wangyun@linux.vnet.ibm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Paul E. McKenney [Fri, 19 Oct 2012 19:49:17 +0000 (12:49 -0700)]
rcu: Provide RCU CPU stall warnings for tiny RCU
Tiny RCU has historically omitted RCU CPU stall warnings in order to
reduce memory requirements, however, lack of these warnings caused
Thomas Gleixner some debugging pain recently. Therefore, this commit
adds RCU CPU stall warnings to tiny RCU if RCU_TRACE=y. This keeps
the memory footprint small, while still enabling CPU stall warnings
in kernels built to enable them.
Reported-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Paul E. McKenney [Fri, 21 Sep 2012 23:35:25 +0000 (16:35 -0700)]
rcu: Dump number of callbacks in stall warning messages
In theory, if a grace period manages to get started despite there being
no callbacks on any of the CPUs, all CPUs could go into dyntick-idle
mode, so that the grace period would never end. This commit updates
the RCU CPU stall warning messages to detect this condition by summing
up the number of callbacks on all CPUs.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Paul E. McKenney [Wed, 19 Sep 2012 23:58:38 +0000 (16:58 -0700)]
rcu: Print remote CPU's stacks in stall warnings
The RCU CPU stall warnings rely on trigger_all_cpu_backtrace() to
do NMI-based dump of the stack traces of all CPUs. Unfortunately, a
number of architectures do not implement trigger_all_cpu_backtrace(), in
which case RCU falls back to just dumping the stack of the running CPU.
This is unhelpful in the case where the running CPU has detected that
some other CPU has stalled.
This commit therefore makes the running CPU dump the stacks of the
tasks running on the stalled CPUs.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Lai Jiangshan [Fri, 12 Oct 2012 17:14:17 +0000 (01:14 +0800)]
rcutorture: Use DEFINE_STATIC_SRCU()
Use DEFINE_STATIC_SRCU() to simplify the rcutorture.c SRCU test code.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com> Reviewed-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Lai Jiangshan [Fri, 12 Oct 2012 17:14:16 +0000 (01:14 +0800)]
srcu: Add DEFINE_SRCU()
In old days, we had two different API sets for dynamic-allocated per-CPU
data and DEFINE_PER_CPU()-defined per_cpu data, and because SRCU used
dynamic-allocated per-CPU data, its srcu_struct structures cannot be
declared statically. This commit therefore introduces DEFINE_SRCU()
and DEFINE_STATIC_SRCU() to allow statically declared SRCU structures,
using the new static per-CPU interfaces.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
[ paulmck: Updated for __DELAYED_WORK_INITIALIZER() added argument. ]
Lai Jiangshan [Fri, 12 Oct 2012 17:14:15 +0000 (01:14 +0800)]
srcu: Export process_srcu()
Because process_srcu() will be used in DEFINE_SRCU(), which is a macro
that could be expanded pretty much anywhere, it can no longer be static.
Note that process_srcu() is still internal to srcu.h.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Paul E. McKenney [Mon, 15 Oct 2012 15:24:54 +0000 (08:24 -0700)]
rcu: Fix precedence error in cpu_needs_another_gp()
The fix introduced by a10d206e (rcu: Fix day-one dyntick-idle
stall-warning bug) has a C-language precedence error. It turns out
that this error is harmless in that the same result is computed for all
inputs, but the code is nevertheless a potential source of confusion.
This commit therefore introduces parentheses in order to force the
execution of the code to reflect the intent.
Reported-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
rcu: Add a module parameter to force use of expedited RCU primitives
There have been some embedded applications that would benefit from
use of expedited grace-period primitives. In some ways, this is
similar to synchronize_net() doing either a normal or an expedited
grace period depending on lock state, but with control outside of
the kernel.
This commit therefore adds rcu_expedited boot and sysfs parameters
that cause the kernel to substitute expedited primitives for the
normal grace-period primitives.
[ paulmck: Add trace/event/rcu.h to kernel/srcu.c to avoid build error.
Get rid of infinite loop through contention path.]
Signed-off-by: Antti P Miettinen <amiettinen@nvidia.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
projects / karo-tx-linux.git / log