Function eeh_pe_state_mark() could possibly have combination of
multiple EEH PE state as its argument. The patch fixes the condition
used to check if EEH_PE_ISOLATED is included.
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
- ibm,rtas-configure-connector should treat the RTAS data as big endian.
- Treat ibm,ppc-interrupt-server#s as big-endian when setting
smp_processor_id during hotplug.
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Acked-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We can use the simpler dump_stack() instead of
show_stack(current, __get_SP())
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Michael points out that __get_SP() is a pretty horrible
function name. Let's give it a better name.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Li Zhong points out an issue with our current __get_SP()
implementation. If ftrace function tracing is enabled (ie -pg
profiling using _mcount) we spill a stack frame on 64bit all the
time.
If a function calls __get_SP() and later calls a function that is
tail call optimised, we will pop the stack frame and the value
returned by __get_SP() is no longer valid. An example from Li can
be found in save_stack_trace -> save_context_stack:
c0000000000432c0 <.save_stack_trace>:
c0000000000432c0: mflr r0
c0000000000432c4: std r0,16(r1)
c0000000000432c8: stdu r1,-128(r1) <-- stack frame for _mcount
c0000000000432cc: std r3,112(r1)
c0000000000432d0: bl <._mcount>
c0000000000432d4: nop
c0000000000432d8: mr r4,r1 <-- __get_SP()
c0000000000432dc: ld r5,632(r13)
c0000000000432e0: ld r3,112(r1)
c0000000000432e4: li r6,1
c0000000000432e8: addi r1,r1,128 <-- pop stack frame
c0000000000432ec: ld r0,16(r1)
c0000000000432f0: mtlr r0
c0000000000432f4: b <.save_context_stack> <-- tail call optimized
save_context_stack ends up with a stack pointer below the current
one, and it is likely to be scribbled over.
Fix this by making __get_SP() a function which returns the
callers stack frame. Also replace inline assembly which grabs
the stack pointer in save_stack_trace and show_stack with
__get_SP().
This also fixes an issue with perf_arch_fetch_caller_regs().
It currently unwinds the stack once, which will skip a
valid stack frame on a leaf function. With the __get_SP() fixes
in this patch, we never need to unwind the stack frame to get
to the first interesting frame.
We have to export __get_SP() because perf_arch_fetch_caller_regs()
(which is used in modules) calls it from a header file.
Reported-by: Li Zhong <zhong@linux.vnet.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We have hit a few customer issues with the topology update code (VPHN
and PRRN). It would be nice to be able to debug the notifications coming
from the hypervisor in both cases to the LPAR, as well as to disable
responding to the notifications at boot-time, to narrow down the source
of the problems. Add a basic level of such functionality, similar to the
numa= command-line parameter. We already have a toggle in
/proc/powerpc/topology_updates that allows run-time enabling/disabling,
so the updates can be started at run-time if desired. But the bugs we've
run into have occured during boot or very shortly after coming to login,
and have resulted in a broken NUMA topology.
Signed-off-by: Nishanth Aravamudan <nacc@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
proc_create can fail, we should check the return value and pass up the
failure.
Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Nishanth Aravamudan <nacc@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Recently we moved HMI handling into Linux kernel instead of taking
HMI directly in OPAL. This new change is dependent on new OPAL call
for HMI recovery which was introduced in newer firmware. While this new
change works fine with latest OPAL firmware, we broke the HMI handling
if we run newer kernel on old OPAL firmware that results in system hang.
This patch fixes this issue by falling back to old HMI behavior on older
OPAL firmware.
This patch introduces a check for opal token OPAL_HANDLE_HMI to see
if we are running on newer firmware or old firmware. On newer firmware
this check would return OPAL_TOKEN_PRESENT, otherwise we are running on
old firmware and fallback to old HMI behavior.
Old firmware: POWER8 System Firmware Release as of today <= SV810_087
Action: Let OPAL handle HMIs
Newer firmware: in development/yet to be released.
Action: Let Linux host handle HMIs.
This patch depends on opal check token patch posted at ppc-devel
https://lists.ozlabs.org/pipermail/linuxppc-dev/2014-August/120224.html
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
[mpe: Minor comment and printk rewording]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
In HMI interrupt handler we don't touch SRR0/SRR1, instead we touch
HSRR0/HSRR1. Hence we don't need to clear MSR_RI bit.
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Declaring sys_call_table as a pointer causes the compiler to generate
the wrong lookup code in arch_syscall_addr().
<arch_syscall_addr>:
lis r9,-16384
rlwinm r3,r3,2,0,29
- lwz r11,30640(r9)
- lwzx r3,r11,r3
+ addi r9,r9,30640
+ lwzx r3,r9,r3
blr
The actual sys_call_table symbol, declared in assembler, is an
array. If we lie about that to the compiler we get the wrong code
generated, as above.
This definition seems only to be used by the syscall tracing code in
kernel/trace/trace_syscalls.c. With this patch I can successfully use
the syscall tracepoints:
bash-3815 [002] .... 333.239082: sys_write -> 0x2
bash-3815 [002] .... 333.239087: sys_dup2(oldfd: a, newfd: 1)
bash-3815 [002] .... 333.239088: sys_dup2 -> 0x1
bash-3815 [002] .... 333.239092: sys_fcntl(fd: a, cmd: 1, arg: 0)
bash-3815 [002] .... 333.239093: sys_fcntl -> 0x1
bash-3815 [002] .... 333.239094: sys_close(fd: a)
bash-3815 [002] .... 333.239094: sys_close -> 0x0
Signed-off-by: Romeo Cane <romeo.cane.ext@coriant.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
If afu_read() returned due to a signal or the AFU file descriptor being
opened non-blocking it would not call finish_wait() before returning,
which could lead to a crash later when something else wakes up the wait
queue.
This patch restructures the wait logic to ensure that the cleanup is
done correctly.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This documentation gives an overview of the hardware architecture, userspace
APIs via /dev/cxl/afuM.N and the syfs files. It also adds a MAINTAINERS file
entry for cxl.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This adds a header file for use by userspace programs wanting to interact with
the kernel cxl driver. It defines structs and magic numbers required for
userspace to interact with devices in /dev/cxl/afuM.N.
Further documentation on this interface is added in a subsequent patch in
Documentation/powerpc/cxl.txt.
It also adds this new userspace header file to Kbuild so it's exported when
doing "make headers_installs".
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This is the core of the cxl driver.
It adds support for using cxl cards in the powernv environment only (ie POWER8
bare metal). It allows access to cxl accelerators by userspace using the
/dev/cxl/afuM.N char devices.
The kernel driver has no knowledge of the function implemented by the
accelerator. It provides services to userspace via the /dev/cxl/afuM.N
devices. When a program opens this device and runs the start work IOCTL, the
accelerator will have coherent access to that processes memory using the same
virtual addresses. That process may mmap the device to access any MMIO space
the accelerator provides. Also, reads on the device will allow interrupts to
be received. These services are further documented in a later patch in
Documentation/powerpc/cxl.txt.
Documentation of the cxl hardware architecture and userspace API is provided in
subsequent patches.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This adds the base cxl support that cannot be built as a module. Specifically
it adds the cxl callbacks that are called from the core powerpc mm code which
must always exist irrespective of if the cxl module is loaded or not. This is
similar to how cell works with CONFIG_SPU_BASE.
This adds a cxl_slbia() call (similar to spu_flush_all_slbs()) which checks if
the cxl module is loaded and in use, returning immediately if it is not. If it
is in use it calls into the cxl SLB invalidation code.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This adds hooks into the core powerpc mm code for cxl.
The core powerpc code sometimes uses local tlbie. Unfortunately this won't
work with the current cxl driver as it relies on snooping tlbie broadcasts.
The cxl hardware can have TLB entries invalidated via MMIO but this is not
currently supported by the driver. In future we can make local tlbie smarter so
that it invalidates cxl contexts via MMIO when it needs to but for now we have
this workaround.
This workaround checks for any active cxl contexts and if so, disables local
tlbie.
This also adds a hook for when SLBs are invalidated. This ensures any
corresponding SLBs in cxl are also invalidated at the same time. This is
required for segment demotion.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This adds the OPAL call to change a PHB into cxl mode.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This adds a new function hash_page_mm() based on the existing hash_page().
This version allows any struct mm to be passed in, rather than assuming
current. This is useful for servicing co-processor faults which are not in the
context of the current running process.
We need to be careful here as the current hash_page() assumes current in a few
places.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This adds a number of functions for allocating IRQs under powernv PCIe for cxl.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This new header adds callbacks and structs needed by the rest of the kernel to
hook into the cxl infrastructure.
This adds the cxl_ctx_in_use() function for use in the mm code to see if any
cxl contexts are currently in use. This is used by the tlbie() to determine if
it can do local TLB invalidations or not. This also adds get/put calls for the
cxl driver module to refcount the active cxl contexts.
cxl_ctx_get/put/in_use are static inlined here as they are called in tlbie
which we want to be fast (mpe's suggestion).
Empty functions are provided when CONFIG_CXL_BASE is not enabled.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Some of the MSI IRQ code in pnv_pci_ioda_msi_setup() is generically useful so
split it out.
This will be used by some of the cxl PCIe code later.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Export mmu_kernel_ssize and mmu_linear_psize. These are needed by the cxl
driver which has it's own MMU. To setup the MMU cxl needs access to these.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Currently msi_bitmap_alloc_hwirqs() will round up any IRQ allocation requests
to the nearest power of 2. eg. ask for 5 IRQs and you'll get 8. This wastes a
lot of IRQs which can be a scarce resource.
For cxl we may require multiple IRQs for every context that is attached to the
accelerator. There may be 1000s of contexts attached, hence we can easily run
out of IRQs, especially if we are needlessly wasting them.
This changes the msi_bitmap_alloc_hwirqs() to allocate only the required number
of IRQs, hence avoiding this wastage. It keeps the natural alignment
requirement though.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This moves spu_flush_all_slbs() into a generic call copro_flush_all_slbs().
This will be useful when we add cxl which also needs a similar SLB flush call.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
__spu_trap_data_seg() currently contains code to determine the VSID and ESID
required for a particular EA and mm struct.
This code is generically useful for other co-processors. This moves the code of
the cell platform so it can be used by other powerpc code. It also adds 1TB
segment handling which Cell didn't support. The new function is called
copro_calculate_slb().
This also moves the internal struct spu_slb to a generic struct copro_slb which
is now used in the Cell and copro code. We use this new struct instead of
passing around esid and vsid parameters.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Currently spu_handle_mm_fault() is in the cell platform.
This code is generically useful for other non-cell co-processors on powerpc.
This patch moves this function out of the cell platform into arch/powerpc/mm so
that others may use it.
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Now that we define these in the KVM code, use these defines when we call
H_SET_MODE. No functional change.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Cody's email address has changed. Update the contact information for
the 24x7 and GPCI counters to the PowerPC developers mailing list.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
catalog_read() implements the read interface for the sysfs file
/sys/bus/event_source/devices/hv_24x7/interface/catalog
It essentially takes a buffer, an offset and count as parameters
to the read() call. It makes a hypervisor call to read a specific
page from the catalog and copy the required bytes into the given
buffer. Each call to catalog_read() returns at most one 4K page.
Given these requirements, we should be able to simplify the
catalog_read().
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Ian pointed out the use of __aligned(4096) caused rather large stack
consumption in single_24x7_request(), so use the kmem_cache
hv_page_cache (which we've already got set up for other allocations)
insead of allocating locally.
CC: Haren Myneni <hbabu@us.ibm.com>
Reported-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Cody P Schafer <dev@codyps.com>
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When reading from the LPC, the OPAL FW calls return the value via pointer
to a uint32_t which is always returned big endian. Our internal inb/outb
implementation byteswaps that fine but our debugfs code is still broken.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: <stable@vger.kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Freescale updates from Scott (27 commits):
"Highlights include DMA32 zone support (SATA, USB, etc now works on 64-bit
FSL kernels), MSI changes, 8xx optimizations and cleanup, t104x board
support, and PrPMC PCI enumeration."
It pulls in more code, including causing us to build a relocatable
kernel, which is good for testing.
The resulting kernel is still usable as a non-crash dump kernel.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Because powernv arrived after these other platforms, the defconfigs
didn't have PPC_POWERNV disabled, and being default y it gets turned on.
If we're going to bother having defconfigs for the specific platforms
then they should only build the code required for those platforms.
The grab bag of everything config is ppc64_defconfig.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
pci_bus_find_capability() is decleared in pci.h, so it is not necessary to do
it again.
This patch removes it.
Signed-off-by: Wei Yang <weiyang@linux.vnet.ibm.com>
Reviewed-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
rtas_call() accepts and returns values in CPU endianness.
The ddw_query_response and ddw_create_response structs members are
defined and treated as BE but as they are passed to rtas_call() as
(u32 *) and they get byteswapped automatically, the data is CPU-endian.
This fixes ddw_query_response and ddw_create_response definitions and use.
of_read_number() is designed to work with device tree cells - it assumes
the input is big-endian and returns data in CPU-endian. However due
to the ddw_create_response struct fix, create.addr_hi/lo are already
CPU-endian so do not byteswap them.
ddw_avail is a pointer to the "ibm,ddw-applicable" property which contains
3 cells which are big-endian as it is a device tree. rtas_call() accepts
a RTAS token in CPU-endian. This makes use of of_property_read_u32_array
to byte swap and avoid the need for a number of be32_to_cpu calls.
Cc: stable@vger.kernel.org # v3.13+
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[aik: folded Anton's patch with of_property_read_u32_array]
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Acked-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Add printk levels to some places in the powerpc port.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Add printk levels to powernv platform code, and convert to
pr_err() etc while here.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There is no need for yet another copy of the command line, just
use boot_command_line like everyone else.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Use pr_fmt to give some context to the error messages in the
module code, and convert open coded debug printk to pr_debug.
Use pr_err for error messages.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Fill in the si_addr_lsb siginfo field so the hwpoison code can
pass to userspace the length of memory that has been corrupted.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
do_page_fault was missing knowledge of HWPOISON, and we would oops
if userspace tried to access a poisoned page:
kernel BUG at arch/powerpc/mm/fault.c:180!
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Exit out early for a kernel fault, avoiding indenting of
most of the function.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Unroll clear_page 8 times. A simple microbenchmark which
allocates and frees a zeroed page:
for (i = 0; i < iterations; i++) {
unsigned long p = __get_free_page(GFP_KERNEL | __GFP_ZERO);
free_page(p);
}
improves 20% on POWER8.
This assumes cacheline sizes won't grow beyond 512 bytes or
page sizes wont drop below 1kB, which is unlikely, but we could
add a runtime check during early init if it makes people nervous.
Michael found that some versions of gcc produce quite bad code
(all multiplies), so we give gcc a hand by using shifts and adds.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
As Michael suggested, the hex prefix for the output of EEH PE
state sysfs entry (/sys/bus/pci/devices/xxx/eeh_pe_state) is
always informative to users.
Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The dma_get_required_mask() function is used by some drivers to
query the platform about what DMA mask is needed to cover all of
memory. This is a bit of a strange semantic when we have to choose
between IOMMU translation or bypass, but essentially what it means
is "what DMA mask will give best performances".
Currently, our IOMMU backend always returns a 32-bit mask here, we
don't do anything special to it when we have bypass available. This
causes some drivers to choose a 32-bit mask, thus losing the ability
to use the bypass window, thinking this is more efficient. The problem
was reported from the driver of following device:
0004:03:00.0 0107: 1000:0087 (rev 05)
0004:03:00.0 Serial Attached SCSI controller: LSI Logic / Symbios \
Logic SAS2308 PCI-Express Fusion-MPT SAS-2 (rev 05)
This patch adds an override of that function in order to, instead,
return a 64-bit mask whenever a bypass window is available in order
for drivers to prefer this configuration.
Reported-by: Murali N. Iyer <mniyer@us.ibm.com>
Suggested-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
It should have been part of commit 1ad7a72c5 ("powerpc/eeh: Report
frozen parent PE prior to child PE"). There are 2 ways to report
EEH errors: proactively polling because of 0xFF's returned from
PCI config or IO read, or interrupt driven event. We missed to
report and handle parent frozen PE prior to child frozen PE for
the later case on PowerNV platform.
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The PEs can be organized as nested. Current implementation doesn't
dump PCI config space for subordinate devices of child PEs. However,
the frozen PE could be caused by those subordinate devices of its
child PEs.
The patch dumps PCI config space for all subordinate devices of the
problematic PE.
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
