blk_mq_unquiesce_queue() is used for unquiescing the
queue explicitly, so replace blk_mq_start_stopped_hw_queues()
with it.
For the scsi part, this patch takes Bart's suggestion to
switch to block quiesce/unquiesce API completely.
Cc: linux-nvme@lists.infradead.org
Cc: linux-scsi@vger.kernel.org
Cc: dm-devel@redhat.com
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
blk_mq_start_stopped_hw_queues() is used implictly
as counterpart of blk_mq_quiesce_queue() for unquiescing queue,
so we introduce blk_mq_unquiesce_queue() and make it
as counterpart of blk_mq_quiesce_queue() explicitly.
This function is for improving the current quiescing mechanism
in the following patches.
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch introduces blk_mq_quiesce_queue_nowait() so
that we can workaround mpt3sas for quiescing its queue.
Once mpt3sas is fixed, we can remove this helper.
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We usually put blk_mq_*() into include/linux/blk-mq.h, so
move this API into there.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
blk_bio_segment_split() makes sure bios have no more than
BIO_MAX_PAGES entries in the bi_io_vec.
This was done because bio_clone_bioset() (when given a
mempool bioset) could not handle larger io_vecs.
No driver uses bio_clone_bioset() any more, they all
use bio_clone_fast() if anything, and bio_clone_fast()
doesn't clone the bi_io_vec.
The main user of of bio_clone_bioset() at this level
is bounce.c, and bouncing now happens before blk_bio_segment_split(),
so that is not of concern.
So remove the big helpful comment and the code.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
bio_clone() is no longer used.
Only bio_clone_bioset() or bio_clone_fast().
This is for the best, as bio_clone() used fs_bio_set,
and filesystems are unlikely to want to use bio_clone().
So remove bio_clone() and all references.
This includes a fix to some incorrect documentation.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This function allocates a bio, then a collection
of pages. It copes with failure.
It currently uses a mempool() to allocate the bio,
but alloc_page() to allocate the pages. These fail
in different ways, so the usage is inconsistent.
Change the bio_clone() to bio_clone_kmalloc()
so that no pool is used either for the bio or the pages.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by : Ming Lei <ming.lei@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
bios that are re-submitted will pass through blk_queue_split() when
blk_queue_bio() is called, and this will split the bio if necessary.
There is no longer any need to do this splitting in xen-blkfront.
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
pblk_submit_read() uses bio_clone_bioset() but doesn't change the
io_vec, so bio_clone_fast() is a better choice.
It also uses fs_bio_set which is intended for filesystems. Using it
in a device driver can deadlock.
So allocate a new bioset, and and use bio_clone_fast().
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Javier González <javier@cnexlabs.com>
Tested-by: Javier González <javier@cnexlabs.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
pktcdvd doesn't change the bi_io_vec of the clone bio,
so it is more efficient to use bio_clone_fast(), and not clone
the bi_io_vec.
This requires providing a bio_set, and it is safest to
provide a dedicated bio_set rather than sharing
fs_bio_set, which filesytems use.
This new bio_set, pkt_bio_set, can also be use for the bio_split()
call as the two allocations (bio_clone_fast, and bio_split) are
independent, neither can block a bio allocated by the other.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drbd does not modify the bi_io_vec of the cloned bio,
so there is no need to clone that part. So bio_clone_fast()
is the better choice.
For bio_clone_fast() we need to specify a bio_set.
We could use fs_bio_set, which bio_clone() uses, or
drbd_md_io_bio_set, which drbd uses for metadata, but it is
generally best to avoid sharing bio_sets unless you can
be certain that there are no interdependencies.
So create a new bio_set, drbd_io_bio_set, and use bio_clone_fast().
Also remove a "XXX cannot fail ???" comment because it definitely
cannot fail - bio_clone_fast() doesn't fail if the GFP flags allow for
sleeping.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
bio_clone() makes a copy of the bi_io_vec, but rbd never changes that,
so there is no need for a copy.
bio_clone_fast() can be used instead, which avoids making the copy.
This requires that we provide a bio_set. bio_clone() uses fs_bio_set,
but it isn't, in general, safe to use the same bio_set at different
levels of the stack, as that can lead to deadlocks. As filesystems
use fs_bio_set, block devices shouldn't.
As rbd never stacks, it is safe to have a single global bio_set for
all rbd devices to use. So allocate that when the module is
initialised, and use it with bio_clone_fast().
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Since commit 23688bf4f8 ("block: ensure to split after potentially
bouncing a bio") blk_queue_bounce() is called *before*
blk_queue_split().
This means that:
1/ the comments blk_queue_split() about bounce buffers are
irrelevant, and
2/ a very large bio (more than BIO_MAX_PAGES) will no longer be
split before it arrives at blk_queue_bounce(), leading to the
possibility that bio_clone_bioset() will fail and a NULL
will be dereferenced.
Separately, blk_queue_bounce() shouldn't use fs_bio_set as the bio
being copied could be from the same set, and this could lead to a
deadlock.
So:
- allocate 2 private biosets for blk_queue_bounce, one for
splitting enormous bios and one for cloning bios.
- add code to split a bio that exceeds BIO_MAX_PAGES.
- Fix up the comments in blk_queue_split()
Credit-to: Ming Lei <tom.leiming@gmail.com> (suggested using single bio_for_each_segment loop)
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A rescuing bioset is only useful if there might be bios from
that same bioset on the bio_list_on_stack queue at a time
when bio_alloc_bioset() is called. This never applies to
q->bio_split.
Allocations from q->bio_split are only ever made from
blk_queue_split() which is only ever called early in each of
various make_request_fn()s. The original bio (call this A)
is then passed to generic_make_request() and is placed on
the bio_list_on_stack queue, and the bio that was allocated
from q->bio_split (B) is processed.
The processing of this may cause other bios to be passed to
generic_make_request() or may even cause the bio B itself to
be passed, possible after some prefix has been split off
(using some other bioset).
generic_make_request() now guarantees that all of these bios
(B and dependants) will be fully processed before the tail
of the original bio A gets handled. None of these early bios
can possible trigger an allocation from the original
q->bio_split as they are either too small to require
splitting or (more likely) are destined for a different queue.
The next time that the original q->bio_split might be used
by this thread is when A is processed again, as it might
still be too big to handle directly. By this time there
cannot be any other bios allocated from q->bio_split in the
generic_make_request() queue. So no rescuing will ever be
needed.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch converts bioset_create() to not create a workqueue by
default, so alloctions will never trigger punt_bios_to_rescuer(). It
also introduces a new flag BIOSET_NEED_RESCUER which tells
bioset_create() to preserve the old behavior.
All callers of bioset_create() that are inside block device drivers,
are given the BIOSET_NEED_RESCUER flag.
biosets used by filesystems or other top-level users do not
need rescuing as the bio can never be queued behind other
bios. This includes fs_bio_set, blkdev_dio_pool,
btrfs_bioset, xfs_ioend_bioset, and one allocated by
target_core_iblock.c.
biosets used by md/raid do not need rescuing as
their usage was recently audited and revised to never
risk deadlock.
It is hoped that most, if not all, of the remaining biosets
can end up being the non-rescued version.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Credit-to: Ming Lei <ming.lei@redhat.com> (minor fixes)
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
"flags" arguments are often seen as good API design as they allow
easy extensibility.
bioset_create_nobvec() is implemented internally as a variation in
flags passed to __bioset_create().
To support future extension, make the internal structure part of the
API.
i.e. add a 'flags' argument to bioset_create() and discard
bioset_create_nobvec().
Note that the bio_split allocations in drivers/md/raid* do not need
the bvec mempool - they should have used bioset_create_nobvec().
Suggested-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
blk_queue_split() is always called with the last arg being q->bio_split,
where 'q' is the first arg.
Also blk_queue_split() sometimes uses the passed-in 'bs' and sometimes uses
q->bio_split.
This is inconsistent and unnecessary. Remove the last arg and always use
q->bio_split inside blk_queue_split()
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Credit-to: Javier González <jg@lightnvm.io> (Noticed that lightnvm was missed)
Reviewed-by: Javier González <javier@cnexlabs.com>
Tested-by: Javier González <javier@cnexlabs.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Move most code into blk_mq_rq_ctx_init, and the rest into
blk_mq_get_request.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch makes sure we always allocate requests in the core blk-mq
code and use a common prepare_request method to initialize them for
both mq I/O schedulers. For Kyber and additional limit_depth method
is added that is called before allocating the request.
Also because none of the intializations can really fail the new method
does not return an error - instead the bfq finish method is hardened
to deal with the no-IOC case.
Last but not least this removes the abuse of RQF_QUEUE by the blk-mq
scheduling code as RQF_ELFPRIV is all that is needed now.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
blk_mq_sched_assign_ioc now only handles the assigned of the ioc if
the schedule needs it (bfq only at the moment). The caller to the
per-request initializer is moved out so that it can be merged with
a similar call for the kyber I/O scheduler.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
icq_to_bic is a container_of operation, so we need to check for NULL
before it. Also move the check outside the spinlock while we're at
it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Merge three functions only tail-called by blk_mq_free_request into
blk_mq_free_request.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Having these as separate helpers in a header really does not help
readability, or my chances to refactor this code sanely.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Having them out of line in blk-mq-sched.c just makes the code flow
unnecessarily complicated.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When a filesystem is mounted from a loop device, writes are
throttled by balance_dirty_pages() twice: once when writing
to the filesystem and once when the loop_handle_cmd() writes
to the backing file. This double-throttling can trigger
positive feedback loops that create significant delays. The
throttling at the lower level is seen by the upper level as
a slow device, so it throttles extra hard.
The PF_LESS_THROTTLE flag was created to handle exactly this
circumstance, though with an NFS filesystem mounted from a
local NFS server. It reduces the throttling on the lower
layer so that it can proceed largely unthrottled.
To demonstrate this, create a filesystem on a loop device
and write (e.g. with dd) several large files which combine
to consume significantly more than the limit set by
/proc/sys/vm/dirty_ratio or dirty_bytes. Measure the total
time taken.
When I do this directly on a device (no loop device) the
total time for several runs (mkfs, mount, write 200 files,
umount) is fairly stable: 28-35 seconds.
When I do this over a loop device the times are much worse
and less stable. 52-460 seconds. Half below 100seconds,
half above.
When I apply this patch, the times become stable again,
though not as fast as the no-loop-back case: 53-72 seconds.
There may be room for further improvement as the total overhead still
seems too high, but this is a big improvement.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ming Lei <tom.leiming@gmail.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Pull NVMe changes for 4.13 from Christoph:
Highlights:
- UUID identifier support from Johannes
- Lots of cleanups from Sagi
- Host Memory Buffer support from me
And lots of cleanups and smaller fixes of course.
Note that the UUID identifier changes are based on top of the uuid tree.
I am the maintainer of that tree and will send it to Linus as soon as
4.12 is released as various other trees depend on it as well (and the
diffstat includes those changes unfortunately)
of_device_ids are not supposed to change at runtime. All functions
working with of_device_ids provided by <linux/of.h> work with const
of_device_ids. So mark the non-const structs as const.
File size before:
text data bss dec hex filename
8908 1096 624 10628 2984 drivers/block/swim3.o
File size after constify swim3_match:
text data bss dec hex filename
9708 296 624 10628 2984 drivers/block/swim3.o
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
This patch fixes two sparse warnings introduced by the "dedicated
error codes for the block layer V3" patch series. These changes
have not been tested.
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
The NVMe 1.3 spec introduces Namespace Optimal IO Boundaries (NOIOB),
which standardizes the stripe mechanism we currently have quirks for.
This patch implements the necessary logic to handle this new feature.
Signed-off-by: Scott Bauer <scott.bauer@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
We don't need to wait for the reset from the delayed work item that
is kicked off when we don't get a keepalive.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
This moves the nvme_reset function from the PCIe driver to common code,
renaming it to nvme_reset_ctrl in the process. Additionally a new
helper nvme_reset_ctrl_sync is added for the case where we want to
wait for the reset. To facilitate that the reset_work work structure is
move to the common nvme_ctrl structure and the ->reset_ctrl method is
removed. For now the drivers initialize the reset_work with their own
callback, but longer term we should move to callouts for specific
parts of the reset process and move even more code to the core.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Now that we get the tagset passed we can have a single implementation for
the I/O and admin queues.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Now that we get the tagset passed we can have a single implementation for
the I/O and admin queues.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Now that we get the tagset passed we can have a single implementation for
the I/O and admin queues.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Now that we get the tagset passed we can have a single implementation for
the I/O and admin queues.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
It only applies to read/write commands, and this way non-PCIe drivers
get the check as well instead of having to duplicate it when adding
metadata support.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
And open code the SHUTDOWN_TIMEOUT macro.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
We accidentally return ERR_PTR(0) which is NULL. The caller isn't
explicitly checking for that but I couldn't immediately spot whether
this would lead to a NULL dereference. Anyway, we can fix add an
error code easily enough.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
To let the host know what happends to the connection establishment,
adjust the behavior of nvmf_log_connect_error to make more connect
specifig error codes human-readble.
Signed-off-by: Guan Junxiong <guanjunxiong@huawei.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Add the new to NVMe 1.3 fields EDSTT, DSTO, FWUG, HCTMA, MNTMT, MXTMT,
and SANICAP into the idenfity controller data structure.
Signed-off-by: Guan Junxiong <guanjunxiong@huawei.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
This was detected by building the nvmet-fc driver with W=1.
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Change the few left over users of ctrl->dev over to using ctrl->device
for logging purposes, so we consistently use the same device.
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Allow overriding the announced NVMe Version of a via configfs.
This is particularly helpful when debugging new features for the host
or target side without bumping the hard coded version (as the target
might not be fully compliant to the announced version yet).
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Guan Junxiong <guanjunxiong@huawei.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Add the UUID field from the NVMe Namespace Identification Descriptor
to the nvmet_ns structure and allow it's population via configfs.
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
A NVMe Identify NS command with a CNS value of '3' is expecting a list
of Namespace Identification Descriptor structures to be returned to
the host for the namespace requested in the namespace identify
command.
This Namespace Identification Descriptor structure consists of the
type of the namespace identifier, the length of the identifier and the
actual identifier.
Valid types are NGUID and UUID which we have saved in our nvme_ns
structure if they have been configured via configfs. If no value has
been assigened to one of these we return an "invalid opcode" back to
the host to maintain backward compatibiliy with older implementations
without Namespace Identify Descriptor list support.
Also as the Namespace Identify Descriptor list is the only mandatory
feature change between 1.2.1 and 1.3 we can bump the advertised
version as well.
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Now that we have a way for getting the UUID from a target, provide it
to userspace as well.
Unfortunately there is already a sysfs attribute called UUID which is
a misnomer as it holds the NGUID value. So instead of creating yet
another wrong name, create a new 'nguid' sysfs attribute for the
NGUID. For the UUID attribute add a check wheter the namespace has a
UUID assigned to it and return this or return the NGUID to maintain
backwards compatibility. This should give userspace a chance to catch
up.
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Sagi Grimberg <sagi@rimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Ming Lei