| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The MI_DISPLAY_FLIP command needs to be set the same pipe
source image like in pipe source register, e.g source image
size minus one. This fixes screen corrupt issue on Ironlake.
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
On Ironlake plane flip interrupt means flip done event already, the
behavior is not like old chips, and perform like other usual interrupt.
So only need to handle flip done event when receiving that interrupt.
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
When a new page flip is requested, we need to both queue an unpin for
the current framebuffer, and also increment the flip pending count on
the newly submitted buffer.
At flip finish time, we need to unpin the old fb and decrement the flip
pending count on the new buffer.
The old code was conflating the two, and led to hangs when new direct
rendered apps were started, replacing the existing frame buffer. This
patch splits out the buffers and prevents the hangs.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
On 945, we need to avoid entering self-refresh if the compressor is
busy, or we may cause display FIFO underruns leading to ugly flicker.
Fixes fdo bug #24314, kernel bug #15043.
Tested-by: Alexander Lam <lambchop468@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Julien Cristau <jcristau@debian.org> (fd.o #25371)
Cc: stable@kernel.org
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
An untiled framebuffer must be aligned to 64k. This is normally handled
by intel_pin_and_fence_fb_obj(), but the intelfb_create() likes to be
different and do the pinning itself. However, it aligns the buffer
object incorrectly for pre-i965 chipsets causing a PGTBL_ERR when it is
installed onto the output.
Fixes:
KMS error message while initializing modesetting -
render error detected: EIR: 0x10 [i915]
http://bugs.freedesktop.org/show_bug.cgi?id=22936
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
Before changing the status of a buffer with a pending write we will await
upon a new flush for that buffer. So we can take advantage of any flushes
posted whilst the buffer is active and pending processing by the GPU, by
clearing its write_domain and updating its last_rendering_seqno -- thus
saving a potential flush in deep queues and improves flushing behaviour
upon eviction for both GTT space and fences.
In order to reduce the time spent searching the active list for matching
write_domains, we move those to a separate list whose elements are
the buffers belong to the active/flushing list with pending writes.
Orignal patch by Chris Wilson <chris@chris-wilson.co.uk>, forward-ported
by me.
In addition to better performance, this also fixes a real bug. Before
this changes, i915_gem_evict_everything didn't work as advertised. When
the gpu was actually busy and processing request, the flush and subsequent
wait would not move active and dirty buffers to the inactive list, but
just to the flushing list. Which triggered the BUG_ON at the end of this
function. With the more tight dirty buffer tracking, all currently busy and
dirty buffers get moved to the inactive list by one i915_gem_flush operation.
I've left the BUG_ON I've used to prove this in there.
References:
Bug 25911 - 2.10.0 causes kernel oops and system hangs
http://bugs.freedesktop.org/show_bug.cgi?id=25911
Bug 26101 - [i915] xf86-video-intel 2.10.0 (and git) triggers kernel oops
within seconds after login
http://bugs.freedesktop.org/show_bug.cgi?id=26101
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Adam Lantos <hege@playma.org>
Cc: stable@kernel.org
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
The patch that adds cpu_probe_vmbits is erroneously writing to reserved
bit 12. Since we are really only probing high bits, don't write this bit
with a one.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
Cc: Guenter Roeck <guenter.roeck@ericsson.com>
Patchwork: http://patchwork.linux-mips.org/patch/949/
Acked-by: Guenter Roeck <guenter.roeck@ericsson.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
|
|
Test the value that was just allocated rather than the previously tested one.
A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@r@
expression *x;
expression e;
identifier l;
@@
if (x == NULL || ...) {
... when forall
return ...; }
... when != goto l;
when != x = e
when != &x
*x == NULL
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
To: linux-mips@linux-mips.org
To: linux-kernel@vger.kernel.org
To: kernel-janitors@vger.kernel.org
Patchwork: http://patchwork.linux-mips.org/patch/945/
Acked-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
|
|
cpu_cache_init and the things it calls should all be __cpuinit instead
of __devinit.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
Patchwork: http://patchwork.linux-mips.org/patch/938/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
|
|
RTC support was rewritten but the defconfig files were not updated. Enable
IPv6 support which for some folks already is a must have. Assign useful
values to other new options.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
|
|
As reported by Maxime Bizon, the commit "MIPS: PowerTV: Fix support for
timer interrupts with > 64 external IRQs" have broken the r4k timer
since it didn't initialize the cp0_compare_irq_shift variable used in
c0_compare_int_pending() on the architectures whose cpu_has_mips_r2 is
false.
This patch fixes it via initializing the cp0_compare_irq_shift as the
cp0_compare_irq used in the old c0_compare_int_pending().
Reported-by: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: Wu Zhangjin <wuzhangjin@gmail.com>
Cc: David VomLehn <dvomlehn@cisco.com>
Cc: mbizon@freebox.fr
Cc: linux-mips@linux-mips.org
Patchwork: http://patchwork.linux-mips.org/patch/922/
Tested-by: Shane McDonald <mcdonald.shane@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
|
|
Got Ironlake DPLL parameter table, which reflects the hardware
optimized values. So this one trys to list DPLL parameters for
different output types, should potential fix clock issue seen
on new Arrandale CPUs.
This fixes DPLL setting failure on one 1920x1080 dual channel
LVDS for Ironlake. Test has also been made on LVDS panels with
smaller size and CRT/HDMI/DP ports for different monitors on
their all supported modes.
Update:
- Change name of double LVDS to dual LVDS.
- Fix SSC 120M reference clock to use the right range.
Cc: CSJ <changsijay@gmail.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
Revert the following change from commit 6f8372b6 ("RDMA/cm: fix
loopback address support")
The defined behavior of rdma_bind_addr is to associate an RDMA
device with an rdma_cm_id, as long as the user specified a non-
zero address. (ie they weren't just trying to reserve a port)
Currently, if the loopback address is passed to rdma_bind_addr,
no device is associated with the rdma_cm_id. Fix this.
It turns out that important apps such as Open MPI depend on
rdma_bind_addr() NOT associating any RDMA device when binding to a
loopback address. Open MPI is being updated to deal with this, but at
least until a new Open MPI release is available, maintain the previous
behavior: allow rdma_bind_addr() to succeed, but do not bind to a
device.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
|
|
cohd_fin has already been verified not to be NULL, so the argument to
BUG_ON cannot be true.
A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@r@
expression *x;
expression e;
identifier l;
@@
if (x == NULL || ...) {
... when forall
return ...; }
... when != goto l;
when != x = e
when != &x
*x == NULL
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Acked-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
If submitting new buffer failed, a wrong descriptor gets completed and it
doesn't check, if a callback is at all defined, which can lead to an Oops. Fix
these bugs and make ipu_update_channel_buffer() void, because it never fails.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The platform data allocated with kmalloc() will become unreachable once
the init is complete, so it should be freed. The problem was discovered
by kmemleak.
Signed-off-by: Aaro Koskinen <aaro.koskinen@nokia.com>
Acked-by: Adrian Hunter <adrian.hunter@nokia.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
|
|
md ioctls are now handled by the md driver itself, but mdadm
may call RAID_VERSION on other devices as well. Mark the command
as IGNORE_IOCTL so this fails silently rather than printing
an annoying message.
Reported-by: "Michael S. Tsirkin" <m.s.tsirkin@gmail.com>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
This plugs in USB and PCI and other bits for SDK7786.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
This reworks some of the SH7786 PCIe initialization code to dynamically
setup and size the various resource windows, as opposed to the original
code that simply wired in a couple of them statically.
At the same time, we tidy up the initialization code a bit, kill off some
read-only register twiddling that was gleaned from the bus analyzer, and
also propagate the physical slot/channel mapping.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
We need to fall back from logical-flat APIC mode to physical-flat mode
when we have more than 8 CPUs. However, in the presence of CPU
hotplug(with bios listing not enabled but possible cpus as disabled cpus in
MADT), we have to consider the number of possible CPUs rather than
the number of current CPUs; otherwise we may cross the 8-CPU boundary
when CPUs are added later.
32bit apic code can use more cleanups (like the removal of vendor checks in
32bit default_setup_apic_routing()) and more unifications with 64bit code.
Yinghai has some patches in works already. This patch addresses the boot issue
that is reported in the virtualization guest context.
[ hpa: incorporated function annotation feedback from Yinghai Lu ]
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
LKML-Reference: <1265767304.2833.19.camel@sbs-t61.sc.intel.com>
Acked-by: Shaohui Zheng <shaohui.zheng@intel.com>
Reviewed-by: Yinghai Lu <yinghai@kernel.org>
Cc: <stable@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
Commit f71dc176aa06359681c30ba6877ffccab6fba3a6 'Make
hpte_need_flush() correctly mask for multiple page sizes' introduced
bug, which is triggered when a kernel with a 64k base page size is run
on a system whose hardware does not 64k hash PTEs. In this case, we
emulate 64k pages with multiple 4k hash PTEs, however in
hpte_need_flush() we incorrectly only mask the hardware page size from
the address, instead of the logical page size. This causes things to
go wrong when we later attempt to iterate through the hardware
subpages of the logical page.
This patch corrects the error. It has been tested on pSeries bare
metal by Michael Neuling.
Signed-off-by: David Gibson <dwg@au1.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
|
|
======
This fix is related to
http://bugzilla.kernel.org/show_bug.cgi?id=15142
but does not address that exact issue.
======
sysfs does like attributes being removed while they are being accessed
(i.e. read or written) and waits for the access to complete.
As accessing some md attributes takes the same lock that is held while
removing those attributes a deadlock can occur.
This patch addresses 3 issues in md that could lead to this deadlock.
Two relate to calling flush_scheduled_work while the lock is held.
This is probably a bad idea in general and as we use schedule_work to
delete various sysfs objects it is particularly bad.
In one case flush_scheduled_work is called from md_alloc (called by
md_probe) called from do_md_run which holds the lock. This call is
only present to ensure that ->gendisk is set. However we can be sure
that gendisk is always set (though possibly we couldn't when that code
was originally written. This is because do_md_run is called in three
different contexts:
1/ from md_ioctl. This requires that md_open has succeeded, and it
fails if ->gendisk is not set.
2/ from writing a sysfs attribute. This can only happen if the
mddev has been registered in sysfs which happens in md_alloc
after ->gendisk has been set.
3/ from autorun_array which is only called by autorun_devices, which
checks for ->gendisk to be set before calling autorun_array.
So the call to md_probe in do_md_run can be removed, and the check on
->gendisk can also go.
In the other case flush_scheduled_work is being called in do_md_stop,
purportedly to wait for all md_delayed_delete calls (which delete the
component rdevs) to complete. However there really isn't any need to
wait for them - they have already been disconnected in all important
ways.
The third issue is that raid5->stop() removes some attribute names
while the lock is held. There is already some infrastructure in place
to delay attribute removal until after the lock is released (using
schedule_work). So extend that infrastructure to remove the
raid5_attrs_group.
This does not address all lockdep issues related to the sysfs
"s_active" lock. The rest can be address by splitting that lockdep
context between symlinks and non-symlinks which hopefully will happen.
Signed-off-by: NeilBrown <neilb@suse.de>
|
|
This is mandatory for 64-bit processes, and doing it also for 32-bit
processes saves a conditional in the compat case.
This fixes the glibc/nptl/tst-stdio1 test case, as well
as many others, on 64-bit.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Both x86-32 and x86-64 with 32-bit compat use ARCH_DLINFO_IA32,
which defines two saved_auxv entries. But system.h only defines
AT_VECTOR_SIZE_ARCH as 2 for CONFIG_IA32_EMULATION, not for
CONFIG_X86_32. Fix that.
Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
LKML-Reference: <20100209023502.GA15408@us.ibm.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
Unset the bit that indicates that a ctxprog can continue at the end.
Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
|
|
ramfc is zero'ed upon destruction, so it's safer to do things in the right
order.
Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
|
|
- We need to disable pgraph fifo access before checking the current channel,
otherwise we could still hit a running ctxprog.
- The writes to 0x400500 are already handled by pgraph->fifo_access and are
therefore redundant, moreover pgraph fifo access should not be reenabled
before current context is set as invalid. So remove them altogether.
Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
|
|
- In the current situation the padding that is added is dangerous to write
to, userspace could potentially overwrite parts of another bo.
- Depth and stencil buffers are supposed to be large enough in general so
the waste of memory should be acceptable.
- Alternatives are hiding the padding from users or splitting vram into 2
zones.
Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
|
|
Fixes DVI+VGA on my 9400, and likely a lot of other configurations that
got broken by the previos DVI-over-DP fix.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
|
|
With DVI and DP plugged, the DVI clock change interrupts being run can
cause DP link training to fail. This adds a spinlock around init table
parsing to prevent this.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
|
|
On my AMD780V chipset, hda_intel.c can crash the kernel with a divide by
zero
for as-yet unknown reasons. A simple check for zero prevents it, though
the problem that causes it remains. Since the workaround is harmless and
won't affect anyone except victims of this bug, it should be safe;
moreover,
because this crash can be triggered by a user-mode application, there are
denial of service implications on the systems affected by the bug without
the patch.
Signed-off-by: Jody Bruchon <jody@nctritech.com>
Cc: <stable@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
It was recently pointed out that the NFSERR_SERVERFAULT error, which is
designed to inform the user of a serious internal error on the server, was
being mapped to an error value that is internal to the kernel.
This patch maps it to the error EREMOTEIO, which is exported to userland
through errno.h.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
|
|
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Acked-by: David Howells <dhowells@redhat.com>
|
|
Not having an fscache cookie is perfectly valid if the user didn't mount
with the fscache option.
This patch fixes http://bugzilla.kernel.org/show_bug.cgi?id=15234
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Acked-by: David Howells <dhowells@redhat.com>
Cc: stable@kernel.org
|
|
PIT control word (address 0x43) is write-only, reads are undefined.
Cc: stable@kernel.org
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
|
|
Current kvm wallclock does not consider the total_sleep_time which could cause
wrong wallclock in guest after host suspend/resume. This patch solve
this issue by counting total_sleep_time to get the correct host boot time.
Cc: stable@kernel.org
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Glauber Costa <glommer@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
|
|
Export getboottime and monotonic_to_bootbased in order to let them
could be used by following patch.
Cc: stable@kernel.org
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
|
|
Break out sh7723 div4 clocks for SIU and IRDA as
reparent / enable clocks. Similar to the SIU clock
patch for sh7722 by Guennadi.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
This patch updates the shared INTC code to use
set_irq_chained_handler() for intc_redirect_irq().
With this in place request_irq() on a merged irq
which has been redirected will now return -EINVAL
instead of 0 together with a crash. This thanks to
the protection of the IRQ_NOREQUEST flag set for
chained interrupt handlers.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
Merge the SDHI vectors in the sh7724 INTC table
and update the SDHI platform data for Ecovec24,
KFR2R09 and MS7724SE.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
Merge the SDHI vectors in the sh7723 INTC table
and update the SDHI platform data for AP325.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
Merge the SDHI vectors in the sh7722 INTC table
and update the SDHI platform data for Migo-R.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
Extend the shared INTC code with force_enable support to
allow keeping mask bits statically enabled. Needed by
upcoming INTC SDHI patches that mux together a bunch of
vectors to a single linux interrupt which is masked by
a priority register, but needs individual mask bits
constantly enabled.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
This patch updates the INTC code by moving all vectors,
groups and registers from struct intc_desc to struct
intc_hw_desc.
The idea is that INTC tables should go from using the
macro(s) DECLARE_INTC_DESC..() only to using struct
intc_desc with name and hw initialized using the macro
INTC_HW_DESC(). This move makes it easy to initialize
an extended struct intc_desc in the future.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|
Offsets and sizes are wrong for 32 bit.
Got broken with 866ba284 "[S390] cleanup lowcore.h".
Reported-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
|
|
If a CHPID is offline during a device shutdown the ccw_device_halt|clear
may fail and the qdio device stays in state STOPPED until the shutdown is
finished. If an interrupt occurs before the device is set to INACTIVE
the STOPPED state triggers a WARN_ON in the interrupt handler.
Prevent this WARN_ON by catching the STOPPED state in the interrupt
handler.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
|
|
Inbound traffic handling may hang if next buffer to check is in
state ERROR, polling is stopped and the final check for further
available inbound buffers disregards buffers in state ERROR.
This patch includes state ERROR when checking availability of
more inbound buffers.
Cc: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
|
|
Test the value that was just allocated rather than the previously tested one.
A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@r@
expression *x;
expression e;
identifier l;
@@
if (x == NULL || ...) {
... when forall
return ...; }
... when != goto l;
when != x = e
when != &x
*x == NULL
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
During the rc period, patches that are not bugfixes
should be done using the net-next tree.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The action modules have been prefixed with 'act_', but the Kconfig
description was not changed.
Signed-off-by: Jan Luebbe <jluebbe@debian.org>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
Zhenyu Wang
Jesse Barnes
Chris Wilson
Daniel Vetter
David Daney
Julia Lawall
Ralf Baechle
Wu Zhangjin
Sean Hefty
Guennadi Liakhovetski
Aaro Koskinen
Arnd Bergmann
Paul Mundt
Suresh Siddha
David Gibson
NeilBrown
David S. Miller
Serge E. Hallyn
Maarten Maathuis
Ben Skeggs
Jody Bruchon
Trond Myklebust
Marcelo Tosatti
Jason Wang
Magnus Damm
Heiko Carstens
Jan Glauber
Ursula Braun
Joe Perches
Jan Luebbe