When CPU hotplug is used, some CPUs may be offline at the time a kexec is
performed. The subsequent kernel may expect these CPUs to be already running,
and will declare them stuck. On pseries, there's also a soft-offline (cede)
state that CPUs may be in; this can also cause problems as the kexeced kernel
may ask RTAS if they're online -- and RTAS would say they are. The CPU will
either appear stuck, or will cause a crash as we replace its cede loop beneath
it.
This patch kicks each present offline CPU awake before the kexec, so that
none are forever lost to these assumptions in the subsequent kernel.
Now, the behaviour is that all available CPUs that were offlined are now
online & usable after the kexec. This mimics the behaviour of a full reboot
(on which all CPUs will be restarted).
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Tidies some typos, KERN_INFO-ise an info msg, and add a debug msg showing
when the final sequence starts.
Also adds a comment to kexec_prepare_cpus_wait() to make note of a possible
problem; the need for kexec to deal with CPUs that failed to originally start
up.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently we look pretty stupid when printing out a bunch of things in
prom_init.c. eg.
Max number of cores passed to firmware: 0x0000000000000080
So I've change this to print in decimal:
Max number of cores passed to firmware: 128 (NR_CPUS = 256)
This required adding a prom_print_dec() function and changing some
prom_printk() calls from %x to %lu.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
If a CPU remove is attempted using the 'release' interface on hardware
which supports extended cede, the CPU will be put in the INACTIVE state
rather than the OFFLINE state due to the default preferred_offline_state
in that situation. In the INACTIVE state it will fail to be removed.
This patch changes the preferred offline state to OFFLINE when an CPU is
in the ONLINE state. After cpu_down() is called in dlpar_offline_cpu()
the CPU will be OFFLINE and CPU removal can continue.
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This adds support for the Mac Mini's that were quietly rolled out
in 2005. Work still needs to be done to support suspend and
WakeOnLan.
Signed-off-by: Mark Crichton <crichton@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In testing SMT disable, we have been regularly seeing the following
message:
Querying DEAD? cpu %i (%i) shows %i
This indicates the current delay in pseries_cpu_die where we wait
for the specified CPU to die, is insufficient. Usually, this does
not cause a problem, but we've seen this result in BUG_ON's going
off in the timer code when we try to migrate the timers off the
dead cpu while a timer is still running. Increasing this delay,
as is done in this patch, seems to resolve this issue.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We already defined start_cpu_decrementer() to invoke decrementer for AP as
the following path:
start_secondary() -> secondary_cpu_time_init() -> start_cpu_decrementer()
So remove these incorrect codes introduced from commit:
e7f75ad0 powerpc/47x: Base ppc476 support
And actually we really should not enable decrementer before calling set_dec().
Signed-off-by: Tiejun Chen <tiejun.chen@windriver.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
On PowerPC we should always use generic ISA DMA API implementation
as there is simply no other implementation exist.
Without this patch, the following build error pops up:
sound/built-in.o: In function 'snd_dma_pointer':
(.text+0x74ae): undefined reference to 'dma_spin_lock'
...
make: *** [.tmp_vmlinux1] Error 1
This is PPC_85xx, SMP and some sound drivers set to =y.
Signed-off-by: Anton Vorontsov <avorontsov@mvista.com>
Acked-by: Dave Liu <daveliu@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
While testing cpu offlining, we are regularly seeing the WARN_ON go off
in xics_ipi_dispatch. It can occur when an IPI gets sent to the CPU while
it is going offline. There is already a similar WARN_ON in the handlers
for PPC_MSG_CALL_FUNCTION and PPC_MSG_CALL_FUNC_SINGLE, so the warning
is not needed in that path. The debugger handler handles this case by
simply ignoring IPIs for offline CPUs, so no warning is needed there.
And the reschedule IPI, which is what is occurring in our test environment,
can be safely ignored, so we can simply remove the WARN_ON from xics_ipi_dispatch.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Fix sizes of variables so correct values are exported via /proc.
Cast variable in comparison to avoid compiler error.
Signed-off-by: Matthew McClintock <msm@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
With dynamic PACAs, the kexecing CPU's PACA won't lie within the kernel
static data and there is a chance that something may stomp it when preparing
to kexec. This patch switches this final CPU to a static PACA just before
we pull the switch.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The recent AMCC 405EX Rev D without Security uses a PVR value
that matches the old 405EXr Rev A/B with Security.
The 405EX Rev D without Security would be shown
incorrectly as an 405EXr. The pvr_mask of 0xffff0004
is no longer sufficient to distinguish the 405EX from 405EXr.
This patch replaces 2 entries in the cpu_specs table
and adds 8 more, each using pvr_mask of 0xffff000f
and appropriate pvr_value to distinguish the AMCC
PowerPC 405EX and 405EXr instances.
The cpu_name for these entries now includes the
Rev, in similar fashion to the 440GX.
Signed-off-by: Lee Nipper <lee.nipper@gmail.com>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
UART2 and UART3 on 460EX/GT have incorrect interrupt mappings right now.
UART2 should be 28 (0x1c) and UART3 29 (0x1d). This patch fixes this and
switches to using decimal number instead of hex, since the AppliedMicro
(AMCC) users manuals describe their inerrupt numbers in decimal.
Thanks to Fabien Proriol for pointing this out.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Fabien Proriol <Fabien.Proriol@jdsu.com>
Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
The config options for REDWOOD_[456] were commented out in the powerpc
Kconfig. The ifdefs referencing this options therefore are dead and all
references to this can be removed (Also dependencies in other KConfig
files).
Signed-off-by: Christian Dietrich <qy03fugy@stud.informatik.uni-erlangen.de>
Signed-off-by: Christoph Egger <siccegge@cs.fau.de>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Use the MMU config registers to scan for available direct and
indirect page sizes and print out the result. Will be needed
for future hugetlbfs implementation.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We patch the TLB miss exception vectors to point to alternate
functions when using HW page table on BookE.
However, we were patching in a new branch in the first instruction
of the exception handler instead of the second one, thus overriding
the nop that is in the first instruction.
This cause problems when single stepping as we rely on that nop for
the single step to stop properly within the exception vector range
rather than on the target of the branch.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We use a similar technique to ppc32: We set a thread local flag
to indicate that we are about to enter or have entered the stop
state, and have fixup code in the async interrupt entry code that
reacts to this flag to make us return to a different location
(sets NIP to LINK in our case).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
--
v2. Fix lockdep bug
Re-mask interrupts when coming back from idle
If we are soft disabled and receive a doorbell exception we don't process
it immediately. This means we need to check on the way out of irq restore
if there are any doorbell exceptions to process.
The problem is at that point we don't know what our regs are, and that
in turn makes xmon unhappy. To workaround the problem, instead of checking
for and processing doorbells, we check for any doorbells and if there were
any we send ourselves another.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
include/asm-generic/irq_regs.h declares per-cpu irq_regs variables and
get_irq_regs() and set_irq_regs() helper functions to maintain them.
These can be used to access the proper pt_regs structure related to the
current interrupt entry (if any).
In the powerpc arch code, this is used to maintain irq regs on
decrementer and external interrupt exceptions. However, for the
doorbell exceptions used by the msgsnd/msgrcv IPI mechanism of newer
BookE CPUs, the irq_regs are not kept up to date.
In particular this means that xmon will not work properly on SMP,
because the secondary xmon instances started by IPI will blow up when
they cannot retrieve the irq regs.
This patch fixes the problem by adding calls to maintain the irq regs
across doorbell exceptions.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Note that critical doorbells are an unimplemented stub just like
other critical or machine check handlers, since we haven't done
support for "levelled" exceptions yet.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The decrementer on BookE acts as a level interrupt and doesn't
need to be re-triggered when going negative. It doesn't go
negative anyways (unless programmed to auto-reload with a
negative value) as it stops when reaching 0.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The doorbells use the content of the PIR register to match messages
from other CPUs. This may or may not be the same as our linux CPU
number, so using that as the "target" is no right.
Instead, we sample the PIR register at boot on every processor
and use that value subsequently when sending IPIs.
We also use a per-cpu message mask rather than a global array which
should limit cache line contention.
Note: We could use the CPU number in the device-tree instead of
the PIR register, as they are supposed to be equivalent. This
might prove useful if doorbells are to be used to kick CPUs out
of FW at boot time, thus before we can sample the PIR. This is
however not the case now and using the PIR just works.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Our handling of debug interrupts on Book3E 64-bit is not quite
the way it should be just yet. This is a workaround to let gdb
work at least for now. We ensure that when context switching,
we set the appropriate DBCR0 value for the new task. We also
make sure that we turn off MSR[DE] within the kernel, and set
it as part of the bits that get set when going back to userspace.
In the long run, we will probably set the userspace DBCR0 on the
exception exit code path and ensure we have some proper kernel
value to set on the way into the kernel, a bit like ppc32 does,
but that will take more work.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CONFIG_SMP_750 doesn't exist in Kconfig, therefore removing all
references for it from the source code.
Signed-off-by: Christoph Egger <siccegge@cs.fau.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Use kstrdup when the goal of an allocation is copy a string into the
allocated region.
The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
expression from,to;
expression flag,E1,E2;
statement S;
@@
- to = kmalloc(strlen(from) + 1,flag);
+ to = kstrdup(from, flag);
... when != \(from = E1 \| to = E1 \)
if (to==NULL || ...) S
... when != \(from = E2 \| to = E2 \)
- strcpy(to, from);
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Use kstrdup when the goal of an allocation is copy a string into the
allocated region.
The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
expression from,to;
expression flag,E1,E2;
statement S;
@@
- to = kmalloc(strlen(from) + 1,flag);
+ to = kstrdup(from, flag);
... when != \(from = E1 \| to = E1 \)
if (to==NULL || ...) S
... when != \(from = E2 \| to = E2 \)
- strcpy(to, from);
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Form 1 affinity allows multiple entries in ibm,associativity-reference-points
which represent affinity domains in decreasing order of importance. The
Linux concept of a node is always the first entry, but using the other
values as an input to node_distance() allows the memory allocator to make
better decisions on which node to go first when local memory has been
exhausted.
We keep things simple and create an array indexed by NUMA node, capped at
4 entries. Each time we lookup an associativity property we initialise
the array which is overkill, but since we should only hit this path during
boot it didn't seem worth adding a per node valid bit.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Remove all rcu head inits. We don't care about the RCU head state before
passing it to call_rcu() anyway. Only leave the "on_stack" variants so
debugobjects can keep track of objects on stack.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently the irqs for the i8042, which historically provides keyboard and
mouse (aux) support, is hardwired in the driver rather than parsing the
dts. This patch modifies the powerpc legacy IO code to attempt to parse
the device tree for this information, failing back to the hardcoded values
if it fails.
Signed-off-by: Martyn Welch <martyn.welch@ge.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
At the moment if request_event_sources_irqs() can't allocate or request
the interrupt, it just does a KERN_ERR printk. This may be fine for the
existing RAS code where if we miss an EPOW event it just means that the
event won't be logged and if we miss one of the RAS errors then we could
miss an event that we perhaps should take action on.
But, for the upcoming IO events code that will use event-sources if we
can't allocate or request the interrupt it means we'd potentially miss
an interrupt from the device. So, let's add a WARN_ON() in this error
case so that we're a bit more vocal when something's amiss.
While we're at it, also use pr_err() to neaten the code up a bit.
Signed-off-by: Mark Nelson <markn@au1.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The RAS code has a #define, RAS_VECTOR_OFFSET, that's used in the
check-exception RTAS call for the vector offset of the exception.
We'll be using this same vector offset for the upcoming IO Event interrupts
code (0x500) so let's move it to include/asm/rtas.h and call it
RTAS_VECTOR_EXTERNAL_INTERRUPT.
Signed-off-by: Mark Nelson <markn@au1.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Now we dynamically allocate the paca array, it takes an extra load
whenever we want to access another cpu's paca. One place we do that a lot
is per cpu variables. A simple example:
DEFINE_PER_CPU(unsigned long, vara);
unsigned long test4(int cpu)
{
return per_cpu(vara, cpu);
}
This takes 4 loads, 5 if you include the actual load of the per cpu variable:
ld r11,-32760(r30) # load address of paca pointer
ld r9,-32768(r30) # load link address of percpu variable
sldi r3,r29,9 # get offset into paca (each entry is 512 bytes)
ld r0,0(r11) # load paca pointer
add r3,r0,r3 # paca + offset
ld r11,64(r3) # load paca[cpu].data_offset
ldx r3,r9,r11 # load per cpu variable
If we remove the ppc64 specific per_cpu_offset(), we get the generic one
which indexes into a statically allocated array. This removes one load and
one add:
ld r11,-32760(r30) # load address of __per_cpu_offset
ld r9,-32768(r30) # load link address of percpu variable
sldi r3,r29,3 # get offset into __per_cpu_offset (each entry 8 bytes)
ldx r11,r11,r3 # load __per_cpu_offset[cpu]
ldx r3,r9,r11 # load per cpu variable
Having all the offsets in one array also helps when iterating over a per cpu
variable across a number of cpus, such as in the scheduler. Before we would
need to load one paca cacheline when calculating each per cpu offset. Now we
have 16 (128 / sizeof(long)) per cpu offsets in each cacheline.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Fix smatch warning: constant 0x8000000000000000 is so big it is unsigned long
Signed-off-by: Denis Kirjanov <dkirjanov@kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We need the ability to reset cores for use with kexec/kdump for
SMP systems. Calling this function with the specific core you want
to reset will cause the CPU to spin in reset.
Signed-off-by: Matthew McClintock <msm@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
There are no BATS on BookE - we have the TLBCAM instead. Also correct
the page size information to included extended sizes. We don't actually allow
a 4G page size to be used, so comment on that as well.
Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Use for_each_pci_dev() to simplify the code.
Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Enables support for HMC initiated partition hibernation. This is
a firmware assisted hibernation, since the firmware handles writing
the memory out to disk, along with other partition information,
so we just mimic suspend to ram.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Partition hibernation will use some of the same code as is
currently used for Live Partition Migration. This function
further abstracts this code such that code outside of rtas.c
can utilize it. It also changes the error field in the suspend
me data structure to be an atomic type, since it is set and
checked on different cpus without any barriers or locking.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Since the decrementer and timekeeping code was moved over to using
the generic clockevents and timekeeping infrastructure, several
variables and functions have been obsolete and effectively unused.
This deletes them.
In particular, wakeup_decrementer() is no longer needed since the
generic code reprograms the decrementer as part of the process of
resuming the timekeeping code, which happens during sysdev resume.
Thus the wakeup_decrementer calls in the suspend_enter methods for
52xx platforms have been removed. The call in the powermac cpu
frequency change code has been replaced by set_dec(1), which will
cause a timer interrupt as soon as interrupts are enabled, and the
generic code will then reprogram the decrementer with the correct
value.
This also simplifies the generic_suspend_en/disable_irqs functions
and makes them static since they are not referenced outside time.c.
The preempt_enable/disable calls are removed because the generic
code has disabled all but the boot cpu at the point where these
functions are called, so we can't be moved to another cpu.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently it is possible for userspace to see the result of
gettimeofday() going backwards by 1 microsecond, assuming that
userspace is using the gettimeofday() in the VDSO. The VDSO
gettimeofday() algorithm computes the time in "xsecs", which are
units of 2^-20 seconds, or approximately 0.954 microseconds,
using the algorithm
now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec
and then converts the time in xsecs to seconds and microseconds.
The kernel updates the tb_orig_stamp and stamp_xsec values every
tick in update_vsyscall(). If the length of the tick is not an
integer number of xsecs, then some precision is lost in converting
the current time to xsecs. For example, with CONFIG_HZ=1000, the
tick is 1ms long, which is 1048.576 xsecs. That means that
stamp_xsec will advance by either 1048 or 1049 on each tick.
With the right conditions, it is possible for userspace to get
(timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is
slightly late in updating the vdso_datapage, and then for stamp_xsec
to advance by 1048 when the kernel does update it, and for userspace
to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to
integer truncation. The result is that time appears to go backwards
by 1 microsecond.
To fix this we change the VDSO gettimeofday to use a new field in the
VDSO datapage which stores the nanoseconds part of the time as a
fractional number of seconds in a 0.32 binary fraction format.
(Or put another way, as a 32-bit number in units of 0.23283 ns.)
This is convenient because we can use the mulhwu instruction to
convert it to either microseconds or nanoseconds.
Since it turns out that computing the time of day using this new field
is simpler than either using stamp_xsec (as gettimeofday does) or
stamp_xtime.tv_nsec (as clock_gettime does), this converts both
gettimeofday and clock_gettime to use the new field. The existing
__do_get_tspec function is converted to use the new field and take
a parameter in r7 that indicates the desired resolution, 1,000,000
for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec
function is then unused and is deleted.
The new algorithm is
now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs
+ (stamp_xtime_seconds << 32) + stamp_sec_fraction
with 'now' in units of 2^-32 seconds. That is then converted to
seconds and either microseconds or nanoseconds with
seconds = now >> 32
partseconds = ((now & 0xffffffff) * resolution) >> 32
The 32-bit VDSO code also makes a further simplification: it ignores
the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary
fraction. Doing so gets rid of 4 multiply instructions. Assuming
a timebase frequency of 1GHz or less and an update interval of no
more than 10ms, the upper 32 bits of tb_to_xs will be at least
4503599, so the error from ignoring the low 32 bits will be at most
2.2ns, which is more than an order of magnitude less than the time
taken to do gettimeofday or clock_gettime on our fastest processors,
so there is no possibility of seeing inconsistent values due to this.
This also moves update_gtod() down next to its only caller, and makes
update_vsyscall use the time passed in via the wall_time argument rather
than accessing xtime directly. At present, wall_time always points to
xtime, but that could change in future.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Build of ptrace.h failed for assembly because it
pulls in stdint.h.
Use exportable types (__u32, __u64) to avoid the dependency
on stdint.h.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Andrey Volkov <avolkov@varma-el.com>
Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
crash_kexec_wait_realmode() is defined only if CONFIG_PPC_STD_MMU_64
and CONFIG_SMP, but is called if CONFIG_PPC_STD_MMU_64 even if !CONFIG_SMP.
Fix the conditional compilation around the invocation.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Matt Evans
Michael Neuling
Robert Jennings
Benjamin Herrenschmidt
Tiejun Chen
Anton Vorontsov
Brian King
Neil Horman
Joe Perches
Matthew McClintock
Lee Nipper
Stefan Roese
Christian Dietrich
Michael Ellerman
David Gibson
Christoph Egger
Julia Lawall
Anton Blanchard
Paul E. McKenney
Martyn Welch
Mark Nelson
Denis Kirjanov
Becky Bruce
Kulikov Vasiliy
Paul Mackerras
Sam Ravnborg