The __do_page_fault() fast-path contains a UTLB flush in order to
force an ITLB reload, this isn't needed in practice as the ITLB is
auto-reloaded from the UTLB anyways, which is already displaced by
the manual 'ldtlb' in the update_mmu_cache() path.
This provides a measurable speed up in the TLB miss fast-path.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
ICH8M on apple macbook pro occasionally locks up completely during PCS
initialization if ports other than the first one are enabled. Add a
separate controller ID and only enable the first port.
tj: commit description added and patch updated to fit with the
previous controller ID update.
Signed-off-by: Thomas Rohwer <tr@tng.de>
Signed-off-by: Tejun Heo <htejun@gmail.com>
Move piix_pata_mwdma to top, rename ich9_2port_sata to ich8_2port_sata
for consistency and use automatically incremented values instead of
assigning fixed values to ease adding new controller IDs.
Signed-off-by: Tejun Heo <htejun@gmail.com>
With newer kernels HDD in my old laptop is limited to UDMA 33.
With this patch I get UDMA 100 again.
Signed-off-by: Gabriel Craciunescu <nix.or.die@googlemail.com>
Signed-off-by: Tejun Heo <htejun@gmail.com>
Sebastian Kemper reported that issuing CD/DVD commands under libata is
not fully compatible with ide-scsi. In particular, the
GPCMD_SET_STREAMING was being rejected at the host level in some
instances.
The reason is that libata-scsi insists upon the cmd_len field exactly
matching the SCSI opcode being issued, whereas ide-scsi tolerates
12-byte commands contained within a 16-byte (cmd_len) CDB.
There doesn't seem to be a good reason for us to not be compatible
there, so here is a patch to fix libata-scsi to permit SCSI opcodes so
long as they fit within whatever size CDB is provided.
Signed-off-by: Mark Lord <mlord@pobox.com>
Signed-off-by: Tejun Heo <htejun@gmail.com>
Per Mark's comments, maybe all ATAPI tape drives need ATA_HORKAGE_STUCK_ERR.
This patch applys ATA_HORKAGE_STUCK_ERR for all ATAPI tape drives.
Signed-off-by: Albert Lee <albertcc@tw.ibm.com>
Cc: Mark Lord <liml@rtr.ca>
Signed-off-by: Tejun Heo <htejun@gmail.com>
After an error condition, some ATAPI tape drives set DRQ=1 together
with ERR=1 when asking the host to transfer the CDB of the next packet
command (i.e. request sense). This patch, a revised version of
Alan/Mark's previous patch, adds ATA_HORKAGE_STUCK_ERR to workaround
the problem by ignoring the ERR bit and proceed sending the CDB.
Signed-off-by: Albert Lee <albertcc@tw.ibm.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Mark Lord <liml@rtr.ca>
Signed-off-by: Tejun Heo <htejun@gmail.com>
Satellite U205 has alternate product name where the satellite part is
all capatalized. Add it to the blacklist.
This is reported by Ross Patterson in kernel bugzilla bug #7780.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Ross Patterson <me@rpatterson.net>
Indeed my previous change to alloc_pskb has made it possible
for the TCP header to be misaligned iff the MTU is not a multiple
of 4 (and less than a page). So I suspect the optimised IPsec
MTU calculation is giving you just such an MTU :)
This patch fixes it by changing alloc_pskb to make sure that
the size is at least 32-bit aligned. This does not cause the
problem fixed by the previous patch because max_header is always
32-bit aligned which means that in the SG/NOTSO case this will
be a no-op.
I thought about putting this in the callers but all the current
callers are from TCP. If and when we get a non-TCP caller we
can always create a TCP wrapper for this function and move the
alignment over there.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
It seems that stats of cpu 0 are counted twice, since
for_each_possible_cpu() is looping on all possible cpus, including 0
Before percpu conversion of ip_rt_acct, we should also remove the
assumption that CPU 0 is online (or even possible)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The usb max packet size won't change during the
device's presence. We should store it in a
variable inside rt2x00dev and use that.
This should also fix a division error when the
device is being hot-unplugged while a frame is
being send out.
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The virtio code never hooked through the ->remove callback. Although
noone supports device removal at the moment, this code is already
needed for module unloading.
This of course also revealed bugs in virtio_blk, virtio_net and lguest
unloading paths.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
drivers/acpi/sbs.c: In function acpi_battery_add:
drivers/acpi/sbs.c:811: warning: ignoring return value of device_create_file,
declared with attribute warn_unused_result
Additional cleanups:
* use struct acpi_battery in acpi_battery_remove() to clean up function
calls, just like acpi_battery_add() already does.
* put braces around unregister call, as it depends on dev being not NULL.
* remove unneeded braces
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Alexey Starikovskiy <astarikovskiy@suse.de>
Signed-off-by: Len Brown <len.brown@intel.com>
Hi Trond,
I have discovered that the BUG_ON in nfs_follow_mountpoint:
BUG_ON(IS_ROOT(dentry));
can be triggered by a misbehaving server.
What happens is the client does a lookup and discoveres that the named
directory has a different fsid, so it initiates a mount.
It then performs a GETATTR on the mounted directory and gets a
different fsid again (due to a bug in the NFS server).
This causes nfs_follow_mountpoint to be called on the newly mounted
root, which triggers the BUG_ON.
To duplicate this, have a directory which contains some mountpoints,
and export that directory with the "crossmnt" flag using nfs-utils
1.1.1 (or 1.1.0 I think)
The GETATTR on the root of the mounted filesystem will return the
information for the top exportpoint, while a lookup will return the
correct information. This difference causes the NFS client to BUG.
I think the best way to fix this is to trap this possibility early, so
just before completing the mount in the NFS client, check that it isn't
going to use nfs_mountpoint_inode_operations.
As long as i_op will never change once set (is that true?), this
should be adequately safe.
The following patch shows a possible approach, and it works for me.
i.e. when the NFS server is misbehaving, I get ESTALE on those
mountpoints, while when the NFS server is working correctly, I get
correct behaviour on the client.
NeilBrown
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Pipe messages start out life on a queue on the inode, but when first
read they're moved to the filp's private pointer. So it's possible for
a poll here to return null even though there's a partially read message
available.
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Return an error from gss_import_sec_context_kerberos if the
negotiated context contains encryption or checksum types not
supported by the kernel code.
This fixes an Oops because success was assumed and later code found
no internal_ctx_id.
Signed-off-by: Kevin Coffman <kwc@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Instead of mapping all errors except EACCES to EAGAIN, map all errors
except EAGAIN to EACCES.
An example is user-land negotiating a Kerberos context with an encryption
type that is not supported by the kernel code. (This can happen due to
mis-configuration or a bug in the Kerberos code that does not honor our
request to limit the encryption types negotiated.) This failure is not
transient, and returning EAGAIN causes mount to continuously retry rather
than giving up.
Signed-off-by: Kevin Coffman <kwc@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Since 2.6.18, the superblock sb->s_root has been a dummy dentry with a
dummy inode. This breaks ustat(), which actually uses sb->s_root in a
vfstat() call.
Fix this by making the s_root a dummy alias to the directory inode that was
used when creating the superblock.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-x86:
x86: simplify "make ARCH=x86" and fix kconfig all.config
x86: reboot fixup for wrap2c board
x86: check boundary in count setup resource
x86: fix reboot with no keyboard attached
x86: add hpet sanity checks
x86: on x86_64, correct reading of PC RTC when update in progress in time_64.c
x86: fix freeze in x86_64 RTC update code in time_64.c
ntp: fix typo that makes sync_cmos_clock erratic
Remove x86 merge artifact from top Makefile
x86: fixup cpu_info array conversion
x86: show cpuinfo only for online CPUs
x86: fix cpu-hotplug regression
x86: ignore the sys_getcpu() tcache parameter
x86: voyager use correct header file name
x86: fix smp init sections
x86: fix voyager_cat_init section
x86: fix bogus memcpy in es7000_check_dsdt()
Simplify "make ARCH=x86" and fix kconfig so we again can set 64BIT in
all.config.
For a fix the diffstat is nice:
6 files changed, 3 insertions(+), 36 deletions(-)
The patch reverts these commits:
- 0f855aa64b ("kconfig: add helper to set
config symbol from environment variable")
- 2a113281f5 ("kconfig: use $K64BIT to
set 64BIT with all*config targets")
Roman Zippel pointed out that kconfig supported string compares so
the additional complexity introduced by the above two patches were
not needed.
With this patch we have following behaviour:
# make {allno,allyes,allmod,rand}config [ARCH=...]
option \ host arch | 32bit | 64bit
=====================================================
./. | 32bit | 64bit
ARCH=x86 | 32bit | 32bit
ARCH=i386 | 32bit | 32bit
ARCH=x86_64 | 64bit | 64bit
The general rule are that ARCH= and native architecture takes
precedence over the configuration.
So make ARCH=i386 [whatever] will always build a 32-bit kernel
no matter what the configuration says. The configuration will
be updated to 32-bit if it was configured to 64-bit and the
other way around.
This behaviour is consistent with previous behaviour so no
suprises here.
make ARCH=x86 will per default result in a 32-bit kernel but as
the only ARCH= value x86 allow the user to select between 32-bit
and 64-bit using menuconfig.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andreas Herrmann <aherrman@arcor.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Simplify "make ARCH=x86" and fix kconfig so we again
can set 64BIT in all.config.
For a fix the diffstat is nice:
6 files changed, 3 insertions(+), 36 deletions(-)
The patch reverts these commits:
0f855aa64b
-> kconfig: add helper to set config symbol from environment variable
2a113281f5
-> kconfig: use $K64BIT to set 64BIT with all*config targets
Roman Zippel pointed out that kconfig supported string
compares so the additional complexity introduced by the
above two patches were not needed.
With this patch we have following behaviour:
# make {allno,allyes,allmod,rand}config [ARCH=...]
option \ host arch | 32bit | 64bit
=====================================================
./. | 32bit | 64bit
ARCH=x86 | 32bit | 32bit
ARCH=i386 | 32bit | 32bit
ARCH=x86_64 | 64bit | 64bit
The general rule are that ARCH= and native architecture
takes precedence over the configuration.
So make ARCH=i386 [whatever] will always build a 32-bit
kernel no matter what the configuration says.
The configuration will be updated to 32-bit if it was
configured to 64-bit and the other way around.
This behaviour is consistent with previous behaviour so
no suprises here.
make ARCH=x86 will per default result in a 32-bit kernel
but as the only ARCH= value x86 allow the user to select
between 32-bit and 64-bit using menuconfig.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andreas Herrmann <aherrman@arcor.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Revert this patch:
move the init sections to the end of memory, so that after they
are free, run time memory is all continugous - this should help decrease
memory fragementation. When doing this, we also pack some of the other
sections a little closer together, to make sure we don't waste memory.
To make this happen, we need to rename the .data.init_task section to
.init_task.data, so it doesn't get picked up by the linker script glob.
Since it causes the kernel not to boot up with mtd filesystems.
Signed-off-by: Bernd Schmidt <bernd.schmidt@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
/*
* CPUs often take a performance hit when accessing unaligned memory
* locations. The actual performance hit varies, it can be small if the
* hardware handles it or large if we have to take an exception and fix
* it
* in software.
*
* Since an ethernet header is 14 bytes network drivers often end up
* with
* the IP header at an unaligned offset. The IP header can be aligned by
* shifting the start of the packet by 2 bytes. Drivers should do this
* with:
*
* skb_reserve(NET_IP_ALIGN);
*
* The downside to this alignment of the IP header is that the DMA is
* now
* unaligned. On some architectures the cost of an unaligned DMA is high
* and this cost outweighs the gains made by aligning the IP header.
*
* Since this trade off varies between architectures, we allow
* NET_IP_ALIGN
* to be overridden.
*/
This new function insl_16 allows to read form 32-bit IO and writes to
16-bit aligned memory. This is useful in above described scenario -
In particular with the AXIS AX88180 Gigabit Ethernet MAC.
Once the device is in 32-bit mode, reads from the RX FIFO always
decrements 4bytes.
While on the other side the destination address in SDRAM is always
16-bit aligned.
If we use skb_reserve(0) the receive buffer is 32-bit aligned but later
we hit a unaligned exception in the IP code.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
If you need a 64 bit divide in the kernel, use asm/div64.h.
Revert the addition of udivdi3.
Cc: Bernd Schmidt <bernd.schmidt@analog.com>
Signed-off-by: Mike Frysinger <michael.frysinger@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Needed to make the wireless board, WRAP2C reboot.
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
need to check info->res_num less than PCI_BUS_NUM_RESOURCES, so
info->bus->resource[info->res_num] = res will not beyond of bus resource
array when acpi returns too many resource entries.
Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Gary Hade <gary.hade@us.ibm.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Attempt to fix http://bugzilla.kernel.org/show_bug.cgi?id=8378
Hiroto Shibuya wrote to tell me that he has a VIA EPIA-EK10000 which
suffers from the reboot problem when no keyboard is attached. My first
patch works for him:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=59f4e7d572980a521b7bdba74ab71b21f5995538
But the latest patch does not work for him :
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=8b93789808756bcc1e5c90c99f1b1ef52f839a51
We found that it was necessary to also set the "disable keyboard" flag in
the command byte, as the first patch was doing. The second patch tries to
minimally modify the command byte, but it is not enough.
Please consider this simple one-line patch to help people with low end VIA
motherboards reboot when no keyboard is attached. Hiroto Shibuya has
verified that this works for him (as I no longer have an afflicted
machine).
Additional discussion:
Note that original patch from Truxton DOES
disable keyboard and this has been in main tree since 2.6.14, thus it must have
quite a bit of air time already.
http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.14.y.git;a=commit;h=59f4e7d572980a521b7bdba74ab71b21f5995538
Note that he only mention "System flag" in the description and comment, but
in the code, "disable keyboard" flag is set.
outb(0x14, 0x60); /* set "System flag" */
In 2.6.23, he made a change to read the current byte and then mask the flags,
but along this change, he only set the "System flag" and dropped the setting
of "disable keyboard" flag.
http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.23.y.git;a=commit;h=8b93789808756bcc1e5c90c99f1b1ef52f839a51
outb(cmd | 0x04, 0x60); /* set "System flag" */
So my request is to restore the setting of disable keyboard flag which has been
there since 2.6.14 but disappeared in 2.6.23.
Cc: Lee Garrett <lee-in-berlin@web.de>
Cc: "Hiroto Shibuya" <hiroto.shibuya@gmail.com>
Cc: Natalie Protasevich <protasnb@gmail.com>
Cc: Dmitry Torokhov <dtor@mail.ru>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Aristeu Rozanski <aris@ruivo.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Some BIOSes advertise HPET at 0x0. We really do no want to
allocate a resource there. Check for it and leave early.
Other BIOSes tell us the HPET is at 0xfed0000000000000
instead of 0xfed00000. Add a check and fix it up with a warning
on user request.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Correct potentially unstable PC RTC time register reading in time_64.c
Stop the use of an incorrect technique for reading the standard PC RTC
timer, which is documented to "disconnect" time registers from the bus
while updates are in progress. The use of UIP flag while interrupts
are disabled to protect a 244 microsecond window is one of the
Motorola spec sheet's documented ways to read the RTC time registers
reliably.
tglx: removed locking changes from original patch, as they gain nothing
(read_persistent_clock is only called during boot, suspend, resume - so
no hot path affected) and conflict with the paravirt locking scheme
(see 32bit code), which we do not want to complicate for no benefit.
Signed-off-by: David P. Reed <dpreed@reed.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Fix hard freeze on x86_64 when the ntpd service calls
update_persistent_clock()
A repeatable but randomly timed freeze has been happening in Fedora 6
and 7 for the last year, whenever I run the ntpd service on my AMD64x2
HP Pavilion dv9000z laptop. This freeze is due to the use of
spin_lock(&rtc_lock) under the assumption (per a bad comment) that
set_rtc_mmss is called only with interrupts disabled. The call from
ntp.c to update_persistent_clock is made with interrupts enabled.
Signed-off-by: David P. Reed <dpreed@reed.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Fix a typo in ntp.c that has caused updating of the persistent (RTC)
clock when synced to NTP to behave erratically.
When debugging a freeze that arises on my AMD64 machines when I
run the ntpd service, I added a number of printk's to monitor the
sync_cmos_clock procedure. I discovered that it was not syncing to
cmos RTC every 11 minutes as documented, but instead would keep trying
every second for hours at a time. The reason turned out to be a typo
in sync_cmos_clock, where it attempts to ensure that
update_persistent_clock is called very close to 500 msec. after a 1
second boundary (required by the PC RTC's spec). That typo referred to
"xtime" in one spot, rather than "now", which is derived from "xtime"
but not equal to it. This makes the test erratic, creating a
"coin-flip" that decides when update_persistent_clock is called - when
it is called, which is rarely, it may be at any time during the one
second period, rather than close to 500 msec, so the value written is
needlessly incorrect, too.
Signed-off-by: David P. Reed
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The x86 merge modified the tags target to handle the two separate
source directories. Remove it now that i386/x86_64 are gone completely.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
92cb7612ae sets cpu_info->cpu_index to zero
for no reason. Referencing cpu_info->cpu_index now points always to CPU#0,
which is apparently not what we want.
Remove it.
Spotted-by: Zou Nan hai <nanhai.zou@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Fix regressions introduced with 92cb7612ae.
It can happen that cpuinfo is displayed for CPUs that are not online or
even worse for CPUs not present at all. As an example, following was
shown for a "second" CPU of a single core K8 variant:
processor : 0
vendor_id : unknown
cpu family : 0
model : 0
model name : unknown
stepping : 0
cache size : 0 KB
fpu : yes
fpu_exception : yes
cpuid level : 0
wp : yes
flags :
bogomips : 0.00
clflush size : 0
cache_alignment : 0
address sizes : 0 bits physical, 0 bits virtual
power management:
Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Commit d435d862ba
("cpu hotplug: mce: fix cpu hotplug error handling")
changed the error handling in mce_cpu_callback.
In cases where not all CPUs are brought up during
boot (e.g. using maxcpus and additional_cpus parameters)
mce_cpu_callback now returns NOTFIY_BAD because
for such CPUs cpu_data is not completely filled when
the notifier is called. Thus mce_create_device fails right
at its beginning:
if (!mce_available(&cpu_data[cpu]))
return -EIO;
As a quick fix I suggest to check boot_cpu_data for MCE.
To reproduce this regression:
(1) boot with maxcpus=2 addtional_cpus=2 on a 4 CPU x86-64 system
(2) # echo 1 >/sys/devices/system/cpu/cpu2/online
-bash: echo: write error: Invalid argument
dmesg shows:
_cpu_up: attempt to bring up CPU 2 failed
Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
dont use the vgetcpu tcache - it's causing problems with tasks
migrating, they'll see the old cache up to a jiffy after the
migration, further increasing the costs of the migration.
In the worst case they see a complete bogus information from
the tcache, when a sys_getcpu() call "invalidated" the cache
info by incrementing the jiffies _and_ the cpuid info in the
cache and the following vdso_getcpu() call happens after
vdso_jiffies have been incremented.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Fix header file name for Voyager build.
In file included from arch/x86/kernel/setup_32.c:61:
include/asm-x86/mach-voyager/setup_arch.h:2:26: error: asm/setup_32.h: No such file or directory
make[1]: *** [arch/x86/kernel/setup_32.o] Error 1
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Fix Voyager section mismatch due to using __devinit instead of __cpuinit.
WARNING: vmlinux.o(.text+0xd943): Section mismatch: reference to .init.text:init_gdt (between 'voyager_smp_prepare_boot_cpu' and 'smp_vic_cmn_interrupt')
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
es7000_check_dst() contains a memcpy from 0, which probably should have been
a memset. Remove it and check the retunr value from acpi_get_table_header.
Noticed by: Joe Perches <joe@perches.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Paul Mundt