If we run out of driver tags, we currently treat shared and non-shared
tags the same - both cases hook into the tag waitqueue. This is a bit
more costly than it needs to be on unshared tags, since we have to both
grab the hctx lock, and the waitqueue lock (and disable interrupts).
For the non-shared case, we can simply mark the queue as needing a
restart.

Split blk_mq_dispatch_wait_add() to account for both cases, and
rename it to blk_mq_mark_tag_wait() to better reflect what it
does now.

Without this patch, shared and non-shared performance is about the same
with 4 fio thread hammering on a single null_blk device (~410K, at 75%
sys). With the patch, the shared case is the same, but the non-shared
tags case runs at 431K at 71% sys.
Reviewed-by: NOmar Sandoval <osandov@fb.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Remove unused mutex brd_mutex. It is unused since the commit ff269568
("brd: remove support for BLKFLSBUF").
Signed-off-by: NMikulas Patocka <mpatocka@redhat.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Currently we are inconsistent in when we decide to run the queue. Using
blk_mq_run_hw_queues() we check if the hctx has pending IO before
running it, but we don't do that from the individual queue run function,
blk_mq_run_hw_queue(). This results in a lot of extra and pointless
queue runs, potentially, on flush requests and (much worse) on tag
starvation situations. This is observable just looking at top output,
with lots of kworkers active. For the !async runs, it just adds to the
CPU overhead of blk-mq.

Move the has-pending check into the run function instead of having
callers do it.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

It is possible that the pointer disk can be null and hence
we can get a null pointer deference when accessing disk->flags.
Add a null pointer check to avoid the dereference.

Detected by CoverityScan, CID#1461133 ("Explicit null dereferenced")

Fixes: 8ddcd653 ("block: introduce GENHD_FL_HIDDEN")
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

guard_bio_eod() needs to look at the partition capacity, not just the
capacity of the whole device, when determining if truncation is
necessary.

[   60.268688] attempt to access beyond end of device
[   60.268690] unknown-block(9,1): rw=0, want=67103509, limit=67103506
[   60.268693] buffer_io_error: 2 callbacks suppressed
[   60.268696] Buffer I/O error on dev md1p7, logical block 4524305, async page read

Fixes: 74d46992 ("block: replace bi_bdev with a gendisk pointer and partitions index")
Cc: stable@vger.kernel.org # v4.13
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NGreg Edwards <gedwards@ddn.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Fixes: d004a5e7 ("block: remove __bio_kmap_atomic")
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

We should be exposing the subsystem attributes like 'model' and
'subsysnqn' to sysfs to allow for easier identification of the
subsystem.
Signed-off-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

When creating nvme multipath devices we should populate the 'slaves' and
'holders' directorys properly to aid userspace topology detection.
Signed-off-by: NHannes Reinecke <hare@suse.com>
[hch: split from a larger patch, compile fix for CONFIG_NVME_MULTIPATH=n]
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

When creating nvme multipath devices we should populate the 'slaves' and
'holders' directorys properly to aid userspace topology detection.
Signed-off-by: NHannes Reinecke <hare@suse.com>
[hch: split from a larger patch]
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

We do this by adding a helper that returns the ns_head for a device that
can belong to either the per-controller or per-subsystem block device
nodes, and otherwise reuse all the existing code.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This patch adds native multipath support to the nvme driver.  For each
namespace we create only single block device node, which can be used
to access that namespace through any of the controllers that refer to it.
The gendisk for each controllers path to the name space still exists
inside the kernel, but is hidden from userspace.  The character device
nodes are still available on a per-controller basis.  A new link from
the sysfs directory for the subsystem allows to find all controllers
for a given subsystem.

Currently we will always send I/O to the first available path, this will
be changed once the NVMe Asynchronous Namespace Access (ANA) TP is
ratified and implemented, at which point we will look at the ANA state
for each namespace.  Another possibility that was prototyped is to
use the path that is closes to the submitting NUMA code, which will be
mostly interesting for PCI, but might also be useful for RDMA or FC
transports in the future.  There is not plan to implement round robin
or I/O service time path selectors, as those are not scalable with
the performance rates provided by NVMe.

The multipath device will go away once all paths to it disappear,
any delay to keep it alive needs to be implemented at the controller
level.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Introduce a new struct nvme_ns_head that holds information about an actual
namespace, unlike struct nvme_ns, which only holds the per-controller
namespace information.  For private namespaces there is a 1:1 relation of
the two, but for shared namespaces this lets us discover all the paths to
it.  For now only the identifiers are moved to the new structure, but most
of the information in struct nvme_ns should eventually move over.

To allow lockless path lookup the list of nvme_ns structures per
nvme_ns_head is protected by SRCU, which requires freeing the nvme_ns
structure through call_srcu.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NJavier González <javier@cnexlabs.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This allows us to manage the various uniqueue namespace identifiers
together instead needing various variables and arguments.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This adds a new nvme_subsystem structure so that we can track multiple
controllers that belong to a single subsystem.  For now we only use it
to store the NQN, and to check that we don't have duplicate NQNs unless
the involved subsystems support multiple controllers.

Includes code originally from Hannes Reinecke to expose the subsystems
in sysfs.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Several block layer and NVMe core functions accept a combination
of BLK_MQ_REQ_* flags through the 'flags' argument but there is
no verification at compile time whether the right type of block
layer flags is passed. Make it possible for sparse to verify this.
This patch does not change any functionality.
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Cc: linux-nvme@lists.infradead.org
Cc: Christoph Hellwig <hch@lst.de>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Cc: Ming Lei <ming.lei@redhat.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

The contexts from which a SCSI device can be quiesced or resumed are:
* Writing into /sys/class/scsi_device/*/device/state.
* SCSI parallel (SPI) domain validation.
* The SCSI device power management methods. See also scsi_bus_pm_ops.

It is essential during suspend and resume that neither the filesystem
state nor the filesystem metadata in RAM changes. This is why while
the hibernation image is being written or restored that SCSI devices
are quiesced. The SCSI core quiesces devices through scsi_device_quiesce()
and scsi_device_resume(). In the SDEV_QUIESCE state execution of
non-preempt requests is deferred. This is realized by returning
BLKPREP_DEFER from inside scsi_prep_state_check() for quiesced SCSI
devices. Avoid that a full queue prevents power management requests
to be submitted by deferring allocation of non-preempt requests for
devices in the quiesced state. This patch has been tested by running
the following commands and by verifying that after each resume the
fio job was still running:

for ((i=0; i<10; i++)); do
  (
    cd /sys/block/md0/md &&
    while true; do
      [ "$(<sync_action)" = "idle" ] && echo check > sync_action
      sleep 1
    done
  ) &
  pids=($!)
  for d in /sys/class/block/sd*[a-z]; do
    bdev=${d#/sys/class/block/}
    hcil=$(readlink "$d/device")
    hcil=${hcil#../../../}
    echo 4 > "$d/queue/nr_requests"
    echo 1 > "/sys/class/scsi_device/$hcil/device/queue_depth"
    fio --name="$bdev" --filename="/dev/$bdev" --buffered=0 --bs=512 \
      --rw=randread --ioengine=libaio --numjobs=4 --iodepth=16       \
      --iodepth_batch=1 --thread --loops=$((2**31)) &
    pids+=($!)
  done
  sleep 1
  echo "$(date) Hibernating ..." >>hibernate-test-log.txt
  systemctl hibernate
  sleep 10
  kill "${pids[@]}"
  echo idle > /sys/block/md0/md/sync_action
  wait
  echo "$(date) Done." >>hibernate-test-log.txt
done
Reported-by: NOleksandr Natalenko <oleksandr@natalenko.name>
References: "I/O hangs after resuming from suspend-to-ram" (https://marc.info/?l=linux-block&m=150340235201348).
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NMartin Steigerwald <martin@lichtvoll.de>
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This flag will be used in the next patch to let the block layer
core know whether or not a SCSI request queue has been quiesced.
A quiesced SCSI queue namely only processes RQF_PREEMPT requests.
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NMartin Steigerwald <martin@lichtvoll.de>
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Convert blk_get_request(q, op, __GFP_RECLAIM) into
blk_get_request_flags(q, op, BLK_MQ_PREEMPT). This patch does not
change any functionality.
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Tested-by: NMartin Steigerwald <martin@lichtvoll.de>
Acked-by: David S. Miller <davem@davemloft.net> [ for IDE ]
Acked-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Set RQF_PREEMPT if BLK_MQ_REQ_PREEMPT is passed to
blk_get_request_flags().
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NMartin Steigerwald <martin@lichtvoll.de>
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

A side effect of this patch is that the GFP mask that is passed to
several allocation functions in the legacy block layer is changed
from GFP_KERNEL into __GFP_DIRECT_RECLAIM.
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NMartin Steigerwald <martin@lichtvoll.de>
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This patch makes it possible to pause request allocation for
the legacy block layer by calling blk_mq_freeze_queue() and
blk_mq_unfreeze_queue().
Signed-off-by: NMing Lei <ming.lei@redhat.com>
[ bvanassche: Combined two patches into one, edited a comment and made sure
  REQ_NOWAIT is handled properly in blk_old_get_request() ]
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NMartin Steigerwald <martin@lichtvoll.de>
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Cc: Ming Lei <ming.lei@redhat.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This patch attempts to make the case of hctx re-running on driver tag
failure more robust. Without this patch, it's pretty easy to trigger a
stall condition with shared tags. An example is using null_blk like
this:

modprobe null_blk queue_mode=2 nr_devices=4 shared_tags=1 submit_queues=1 hw_queue_depth=1

which sets up 4 devices, sharing the same tag set with a depth of 1.
Running a fio job ala:

[global]
bs=4k
rw=randread
norandommap
direct=1
ioengine=libaio
iodepth=4

[nullb0]
filename=/dev/nullb0
[nullb1]
filename=/dev/nullb1
[nullb2]
filename=/dev/nullb2
[nullb3]
filename=/dev/nullb3

will inevitably end with one or more threads being stuck waiting for a
scheduler tag. That IO is then stuck forever, until someone else
triggers a run of the queue.

Ensure that we always re-run the hardware queue, if the driver tag we
were waiting for got freed before we added our leftover request entries
back on the dispatch list.
Reviewed-by: NBart Van Assche <bart.vanassche@wdc.com>
Tested-by: NBart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: NMing Lei <ming.lei@redhat.com>
Reviewed-by: NOmar Sandoval <osandov@fb.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Much easier to just opencode this helper.  Also use ARRAY_SIZE instead of
passing the inline bvec array size manually.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NSagi Grimberg <sagi@rimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Currently the NVMe target stores the expexted data length in req->data_len
and uses that for data transfer decisions, but that does not take the
actual transfer length in the SGLs into account.  So this adds a new
transfer_len field, into which the transport drivers store the actual
transfer length.  We then check the two match before actually executing
the command.

The FC transport driver already had such a field, which is removed in
favour of the common one.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

The 'remove_work' may be scheduled to run after nvme_remove()
returns since we can't simply cancel it in nvme_remove() for
avoiding deadlock. Once nvme_remove() returns, this module(nvme)
can be unloaded.

On the other hand, nvme_put_ctrl() calls ctr->ops->free_ctrl
which may point to nvme_pci_free_ctrl() in unloaded module.

This patch avoids this issue by queuing 'remove_work' via 'nvme_wq',
and flush this worqueue in nvme_exit() as suggested by Sagi.
Suggested-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NMing Lei <ming.lei@redhat.com>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This will give udev a chance to observe and handle asynchronous event
notifications and clear the log to unmask future events of the same type.
The driver will create a change uevent of the asyncronuos event result
before submitting the next AEN request to the device if a completed AEN
event is of type error, smart, command set or vendor specific,
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NGuan Junxiong <guanjunxiong@huawei.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Async event work is for core use only and should not be called directly
from drivers.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NGuan Junxiong <guanjunxiong@huawei.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

The driver can handle tracking only one AEN request, so this patch
removes handling for multiple ones.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NJames Smart  <james.smart@broadcom.com>
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This was being saved in a structure, but never used anywhere. The queue
size is obtained through other means, so there's no reason to duplicate
this without a user for it.
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NGuan Junxiong <guanjunxiong@huawei.com>
Reviewed-by: NJames Smart <james.smart@broadcom.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

All the transports were unnecessarilly duplicating the AEN request
accounting. This patch defines everything in one place.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NGuan Junxiong <guanjunxiong@huawei.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

the status is either success or some status id and
we don't need a local variable for it.
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

We already allocated the buffer with kzalloc.
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Avoid that removal of a request queue sporadically triggers the
following warning:

list_del corruption. next->prev should be ffff8807d649b970, but was 6b6b6b6b6b6b6b6b
WARNING: CPU: 3 PID: 342 at lib/list_debug.c:56 __list_del_entry_valid+0x92/0xa0
Call Trace:
 process_one_work+0x11b/0x660
 worker_thread+0x3d/0x3b0
 kthread+0x129/0x140
 ret_from_fork+0x27/0x40
Signed-off-by: NBart Van Assche <bart.vanassche@wdc.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Fix print formatting, but keep the original output to prevent user
breakage as suggested by Joe Perches.
Signed-off-by: NJavier González <javier@cnexlabs.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Copy subnqns using NVMF_NQN_SIZE as it is < 256
Signed-off-by: NJavier González <javier@cnexlabs.com>
Reviewed-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

We have to put the driver tag if dispatch budget can't be got, otherwise
it might cause IO deadlock, especially in case that size of tags is very
small.

Fixes: de148297(blk-mq: introduce .get_budget and .put_budget in blk_mq_ops)
Signed-off-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Use the obvious calling convention.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This helper doesn't buy us much over calling kmap_atomic directly.
In fact in the only caller it does a bit of useless work as the
caller already has the bvec at hand, and said caller would even
buggy for a multi-segment bio due to the use of this helper.

So just remove it.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

There are no users of it anymore.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This reverts commit 358a3a6b.

We have cases that aren't covered 100% in the drivers, so for now
we have to retain the shared tag restart loops.
Signed-off-by: NJens Axboe <axboe@kernel.dk>