refill_one() should never be called on a recv struct that
doesn't need a new r_frag allocated. Add a WARN and remove
conditional around r_frag alloc code.
Also, add a comment to explain why r_ibinc may or may not
need refilling.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Instead of splitting up a page into RDS_FRAG_SIZE chunks
ourselves, ask rds_page_remainder_alloc() to do it. While it
is possible PAGE_SIZE > FRAG_SIZE, on x86en it isn't, so having
duplicate "carve up a page into buffers" code seems excessive.
The other modification this spawns is the use of a single
struct scatterlist in rds_page_frag instead of a bare page ptr.
This causes verbosity to increase in some places, and decrease
in others.
Finally, I decided to unify the lifetimes and alloc/free of
rds_page_frag and its page. This is a nice simplification in itself,
but will be extra-nice once we come to adding cmason's recycling
patch.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Currently IB device removal destroys connections which are associated with the
device. This prevents connections from being re-established when replacement
devices are added.
Instead we'll queue shutdown work on the connections as their devices are
removed. When we see that devices are added we triger connection attempts on
all connections that don't currently have a device.
The result is that RDS sockets can resume device-independent work (bcopy, not
RDMA) across IB device removal and restoration.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
A few paths had the same block of code to queue a connection's connect work if
it was in the right state. Let's move this in to a helper function.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB client .remove callback used to free the rds_ibdev for the given
device unconditionally. This could race other users of the struct. This patch
adds refcounting so that we only free the rds_ibdev once all of its users are
done.
Many rds_ibdev users are tied to connections. We give the connection a
reference and change these users to reference the device in the connection
instead of looking it up in the IB client data. The only user of the IB client
data remaining is the first lookup of the device as connections are built up.
Incrementing the reference count of a device found in the IB client data could
race with final freeing so we use an RCU grace period to make sure that freeing
won't happen until those lookups are done.
MRs need the rds_ibdev to get at the pool that they're freed in to. They exist
outside a connection and many MRs can reference different devices from one
socket, so it was natural to have each MR hold a reference. MR refs can be
dropped from interrupt handlers and final device teardown can block so we push
it off to a work struct. Pool teardown had to be fixed to cancel its pending
work instead of deadlocking waiting for all queued work, including itself, to
finish.
MRs get their reference from the global device list, which gets a reference.
It is left unprotected by locks and remains racy. A simple global lock would
be a significant bottleneck. More scalable (complicated) locking should be
done carefully in a later patch.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_ib_xmit_rdma() was calling ib_get_client_data() to get at the rds_ibdevice
just to get the max_sge for the transmit. This patch instead has it get it
directly off the rds_ibdev which is stored on the connection.
The current code won't free the rds_ibdev until all the IB connections that use
it are freed. So it's safe to reference the rds_ibdev this way. In the future
it also makes it easier to support proper reference counting of the rds_ibdev
struct.
As an additional bonus, this gets rid of the performance hit of calling in to
the IB stack to look up the rds_ibdev. The current implementation in the IB
stack acquires an interrupt blocking spinlock to protect the registration of
client callback data.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_ib_cm_handle_connect() could return without unlocking the c_conn_lock if
rds_setup_qp() failed. Rather than adding another imbalanced mutex_unlock() to
this error path we only unlock the mutex once as we exit the function, reducing
the likelyhood of making this same mistake in the future. We remove the
previous mulitple return sites, leaving one unambigious return path.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
This makes sure we have the proper number of references in
rds_ib_xmit_atomic and rds_ib_xmit_rdma. We also consistently
drop references the same way for all message types as the IOs end.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The connection hash was almost entirely RCU ready, this
just makes the final couple of changes to use RCU instead
of spinlocks for everything.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The RDS send_xmit code was trying to get fancy with message
counting and was dropping the final reference on the RDMA messages
too early. This resulted in memory corruption and oopsen.
The fix here is to always add a ref as the parts of the message passes
through rds_send_xmit, and always drop a ref as the parts of the message
go through completion handling.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This is the first in a long line of patches that tries to fix races
between RDS connection shutdown and RDS traffic.
Here we are maintaining a count of active senders to make sure
the connection doesn't go away while they are using it.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The RDS bind lookups are somewhat expensive in terms of CPU
time and locking overhead. This commit changes them into a
faster RCU based hash tree instead of the rbtrees they were using
before.
On large NUMA systems it is a significant improvement.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Allocate send/recv rings in memory that is node-local to the HCA.
This significantly helps performance.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_ib_get_device is called very often as we turn an
ip address into a corresponding device structure. It currently
take a global spinlock as it walks different lists to find active
devices.
This commit changes the lists over to RCU, which isn't very complex
because they are not updated very often at all.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This removes a global waitqueue used to wait for rds messages
and replaces it with a waitqueue inside the rds_message struct.
The global waitqueue turns into a global lock and significantly
bottlenecks operations on large machines.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The bind_lock is almost entirely readonly, but it gets
hammered during normal operations and is a major bottleneck.
This commit changes it to an rwlock, which takes it from 80%
of the system time on a big numa machine down to much lower
numbers.
A better fix would involve RCU, which is done in a later commit
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Update comments to reflect changes in previous commit.
Keeping as separate commits due to different authorship.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_send_xmit is required to loop around after it releases the lock
because someone else could done a trylock, found someone working on the
list and backed off.
But, once we drop our lock, it is possible that someone else does come
in and make progress on the list. We should detect this and not loop
around if another process is actually working on the list.
This patch adds a generation counter that is bumped every time we
get the lock and do some send work. If the retry notices someone else
has bumped the generation counter, it does not need to loop around and
continue working.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
The purpose of the send quota was really to give fairness
when different connections were all using the same
workq thread to send backlogged msgs -- they could only send
so many before another connection could make progress.
Now that each connection is pushing the backlog from its
completion handler, they are all guaranteed to make progress
and the quota isn't needed any longer.
A thread *will* have to send all previously queued data, as well
as any further msgs placed on the queue while while c_send_lock
was held. In a pathological case a single process can get
roped into doing this for long periods while other threads
get off free. But, since it can only do this until the transport
reports full, this is a bounded scenario.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Now that rds_send_xmit() does not block, we can call it directly
instead of going through the helper thread.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_sendmsg() is calling the send worker function to
send the just-queued datagrams, presumably because it wants
the behavior where anything not sent will re-call the send
worker. We now ensure all queued datagrams are sent by retrying
from the send completion handler, so this isn't needed any more.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_message_put() cannot be called with irqs off, so move it after
irqs are re-enabled.
Spinlocks throughout the function do not to use _irqsave because
the lock of c_send_lock at top already disabled irqs.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
This change allows us to call rds_send_xmit() from a tasklet,
which is crucial to our new operating model.
* Change c_send_lock to a spinlock
* Update stats fields "sem_" to "_lock"
* Remove unneeded rds_conn_is_sending()
About locking between shutdown and send -- send checks if the
connection is up. Shutdown puts the connection into
DISCONNECTING. After this, all threads entering send will exit
immediately. However, a thread could be *in* send_xmit(), so
shutdown acquires the c_send_lock to ensure everyone is out
before proceeding with connection shutdown.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Performance is better if we use allocations that don't block
to refill the receive ring. Since the whole reason we were
kicking out to the worker thread was so we could do blocking
allocs, we no longer need to do this.
Remove gfp params from rds_ib_recv_refill(); we always use
GFP_NOWAIT.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
We now ask the transport to give us a rm for the congestion
map, and then we handle it normally. Previously, the
transport defined a function that we would call to send
a congestion map.
Convert TCP and loop transports to new cong map method.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Now that we are signaling send completions much less, we are likely
to have dirty entries in the send queue when the connection is
shut down (on rmmod, for example.) These are cleaned up a little
further down in conn_shutdown, but if we wait on the ring_empty_wait
for them, it'll never happen, and we hand on unload.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Previously, RDS would wait until the final send WR had completed
and then handle cleanup. With silent ops, we do not know
if an atomic, rdma, or data op will be last. This patch
handles any of these cases by keeping a pointer to the last
op in the message in m_last_op.
When the TX completion event fires, rds dispatches to per-op-type
cleanup functions, and then does whole-message cleanup, if the
last op equalled m_last_op.
This patch also moves towards having op-specific functions take
the op struct, instead of the overall rm struct.
rds_ib_connection has a pointer to keep track of a a partially-
completed data send operation. This patch changes it from an
rds_message pointer to the narrower rm_data_op pointer, and
modifies places that use this pointer as needed.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
It hasn't cropped up in the field, but this code ensures it is
impossible to issue operations that pass an rdma cookie (DEST, MAP)
in the same sendmsg call that's actually initiating rdma or atomic
ops.
Disallowing this perverse-but-technically-allowed usage makes silent
RDMA heuristics slightly easier.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Add a flag to the API so users can indicate they want
silent operations. This is needed because silent ops
cannot be used with USE_ONCE MRs, so we can't just
assume silent.
Also, change send_xmit to do atomic op before rdma op if
both are present, and centralize the hairy logic to determine if
we want to attempt silent, or not.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
When dropping ops in the send queue, we notify the client
of failed rdma ops they asked for notifications on, but not
atomic ops. It should be for both.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Do not allocate sgs for data for 0-length datagrams
Set data.op_active in rds_sendmsg() instead of
rds_message_copy_from_user().
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Simplify rds_send_xmit().
Send a congestion map (via xmit_cong_map) without
decrementing send_quota.
Move resetting of conn xmit variables to end of loop.
Update comments.
Implement a special case to turn off sending an rds header
when there is an atomic op and no other data.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
A big changeset, but it's all pretty dumb.
struct rds_rdma_op was already embedded in struct rm_rdma_op.
Remove rds_rdma_op and put its members in rm_rdma_op. Rename
members with "op_" prefix instead of "r_", for consistency.
Of course this breaks a lot, so fixup the code accordingly.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
cmsg_rdma_args just calls rdma_prepare and does a little
arg checking -- not quite enough to justify its existence.
Plus, it is the only caller of rdma_prepare().
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Maybe things worked fine with the flow control code running
even in the non-flow-control case, but making it explicitly
conditional helps the non-fc case be easier to read.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Removed unsignaled_bytes sysctl and code to signal
based on it. I believe unsignaled_wrs is more than
sufficient for our purposes.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Now that the header always goes first, it is possible to
simplify rds_ib_xmit. Instead of having a path to handle 0-byte
dgrams and another path to handle >0, these can both be handled
in one path. This lets us eliminate xmit_populate_wr().
Rename sent to bytes_sent, to differentiate better from other
variable named "send".
Signed-off-by: Andy Grover <andy.grover@oracle.com>
These functions were to cope with differently ordered
sg entries depending on RDS 3.0 or 3.1+. Now that
we've dropped 3.0 compatibility we no longer need them.
Also, modify usage sites for these to refer to sge[0] or [1]
directly. Reorder code to initialize header sgs first.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
RDS 3.0 connections (in OFED 1.3 and earlier) put the
header at the end. 3.1 connections put it at the head.
The code has significant added complexity in order to
handle both configurations. In OFED 1.6 we can
drop this and simplify the code by only supporting
"header-first" configuration.
This patch checks the protocol version, and if prior
to 3.1, does not complete the connection.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
both atomics and rdmas need to convert ib-specific completion codes
into RDS status codes. Rename rds_ib_rdma_send_complete to
rds_ib_send_complete, and have it take a pointer to the function to
call with the new error code.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Instead of using a constant for initiator_depth and
responder_resources, read the per-QP values when the
device is enumerated, and then use these values when creating
the connection.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Implement a CMSG-based interface to do FADD and CSWP ops.
Alter send routines to handle atomic ops.
Add atomic counters to stats.
Add xmit_atomic() to struct rds_transport
Inline rds_ib_send_unmap_rdma into unmap_rm
Signed-off-by: Andy Grover <andy.grover@oracle.com>
The previous code was correct, but made the assumption that
if r_notifier was non-NULL then either r_recverr or r_notify
was true. Valid, but fragile. Changed to explicitly check
r_recverr (shows up in greps for recverr now, too.)
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_message_alloc_sgs() now returns correctly-initialized
sg lists, so calleds need not do this themselves.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
This eliminates a separate memory alloc, although
it is now necessary to add an "r_active" flag, since
it is no longer to use the m_rdma_op pointer as an
indicator of if an rdma op is present.
rdma SGs allocated from rm sg pool.
rds_rm_size also gets bigger. It's a little inefficient to
run through CMSGs twice, but it makes later steps a lot smoother.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
r_m_copy_from_user used to allocate the rm as well as kernel
buffers for the data, and then copy the data in. Now, sendmsg()
allocates the rm, although the data buffer alloc still happens
in r_m_copy_from_user.
SGs are still allocated with rm, but now r_m_alloc_sgs() is
used to reserve them. This allows multiple SG lists to be
allocated from the one rm -- this is important once we also
want to alloc our rdma sgl from this pool.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
First, it looks to me like the atomic_inc is wrong.
We should be decrementing refcount only once here, no? It's
already being done by the mr_put() at the end.
Second, simplify the logic a bit by bailing early (with a warning)
if !mr.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Clearly separate rdma-related variables in rm from data-related ones.
This is in anticipation of adding atomic support.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
This function has been the source of numerous bugs; it's just
too complicated. Simplified to nest spinlocks cleanly within
the second loop body, and kick out early if there are no
rms to drop.
This will be a little slower because conn lock is grabbed for
each entry instead of "caching" the lock across rms, but this
should be entirely irrelevant to fastpath performance.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
On second look at this bug (OFED #2002), it seems that the
collision is not with the retransmission queue (packet acked
by the peer), but with the local send completion. A theoretical
sequence of events (from time t0 to t3) is thought to be as
follows,
Thread #1
t0:
sock_release
rds_release
rds_send_drop_to /* wait on send completion */
t2:
rds_rdma_drop_keys() /* destroy & free all mrs */
Thread #2
t1:
rds_ib_send_cq_comp_handler
rds_ib_send_unmap_rm
rds_message_unmapped /* wake up #1 @ t0 */
t3:
rds_message_put
rds_message_purge
rds_mr_put /* memory corruption detected */
The problem with the rds_rdma_drop_keys() is it could
remove a mr's refcount more than its due (i.e. repeatedly
as long as it still remains in the tree (mr->r_refcount > 0)).
Theoretically it should remove only one reference - reference
by the tree.
/* Release any MRs associated with this socket */
while ((node = rb_first(&rs->rs_rdma_keys))) {
mr = container_of(node, struct rds_mr, r_rb_node);
if (mr->r_trans == rs->rs_transport)
mr->r_invalidate = 0;
rds_mr_put(mr);
}
I think the correct way of doing it is to remove the mr from
the tree and rds_destroy_mr it first, then a rds_mr_put()
to decrement its reference count by one. Whichever thread
holds the last reference will free the mr via rds_mr_put().
Signed-off-by: Tina Yang <tina.yang@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
in_interrupt() is true in softirqs. The BUG_ONs are supposed
to check for if irqs are disabled, so we should use
BUG_ON(irqs_disabled()) instead, duh.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
struct rds_rdma_notify contains a 32 bits hole on 64bit arches,
make sure it is zeroed before copying it to user.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a mutex_unlock missing on the error path. In each case, whenever the
label out is reached from elsewhere in the function, mutex is not locked.
The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
expression E1;
@@
* mutex_lock(E1);
<+... when != E1
if (...) {
... when != E1
* return ...;
}
...+>
* mutex_unlock(E1);
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Reviewed-by: Zach Brown <zach.brown@oracle.com>
Acked-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Also added an explicit break; to avoid
a fallthrough in net/ipv4/tcp_input.c
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the original code, the "goto out" calls "rdma_destroy_id(cm_id);"
That isn't needed here and would cause problems because "cm_id" is an
ERR_PTR. The new code just returns directly.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define a new function to return the waitqueue of a "struct sock".
static inline wait_queue_head_t *sk_sleep(struct sock *sk)
{
return sk->sk_sleep;
}
Change all read occurrences of sk_sleep by a call to this function.
Needed for a future RCU conversion. sk_sleep wont be a field directly
available.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Create per-cpu workqueue threads instead of a single
krdsd thread. This is a step towards better scalability.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
set_page_dirty() unconditionally re-enables interrupts, so
if we call it with irqs off, they will be on after the call,
and that's bad. This patch moves the call after we've re-enabled
interrupts in send_drop_to(), so it's safe.
Also, add BUG_ONs to let us know if we ever do call set_page_dirty
with interrupts off.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the RDMA op has aborted with a remote access error,
in addition to what we already do (tell userspace it has
completed with an error) also unmap it and put() the rm.
Otherwise, hangs may occur on arches that track maps and
will not exit without proper cleanup.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rds_poll_waitq's listeners will be awoken if we receive a congestion
notification. Bad performance may result because *all* polled sockets
contend for this single lock. However, it should not be necessary to
wake pollers when a congestion update arrives if they have never
experienced congestion, and not putting these on the waitq will
hopefully greatly reduce contention.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It seems rds_send_drop_to() called
__rds_rdma_send_complete(rs, rm, RDS_RDMA_CANCELED)
with only rds_sock lock, but not rds_message lock. It raced with
other threads that is attempting to modify the rds_message as well,
such as from within rds_rdma_send_complete().
Signed-off-by: Tina Yang <tina.yang@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RDS's error messages when a connection goes down are a little
extreme. A connection may go down, and it will be re-established,
and everything is fine. This patch links these messages through
rdsdebug(), instead of to printk directly.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
if a machine is shut down without closing sockets properly, and
freeing all MRs, then a BUG_ON will bring it down. This patch
changes these to WARN_ONs -- leaking MRs is not fatal (although
not ideal, and there is more work to do here for a proper fix.)
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix a deadlock between rds_rdma_send_complete() and
rds_send_remove_from_sock() when rds socket lock and
rds message lock are acquired out-of-order.
Signed-off-by: Tina Yang <Tina.Yang@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We have two kinds of loopback: software (via loop transport)
and hardware (via IB). sw is used for 127.0.0.1, and doesn't
support rdma ops. hw is used for sends to local device IPs,
and supports rdma. Both are used in different cases.
For both of these, when there is a congestion map update, we
want to call rds_cong_map_updated() but not actually send
anything -- since loopback local and foreign congestion maps
point to the same spot, they're already in sync.
The old code never called sw loop's xmit_cong_map(),so
rds_cong_map_updated() wasn't being called for it. sw loop
ports would not work right with the congestion monitor.
Fixing that meant that hw loopback now would send congestion maps
to itself. This is also undesirable (racy), so we check for this
case in the ib-specific xmit code.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of waking the send thread whenever any send space is available,
wait until it is at least half empty. This is modeled on how
sock_def_write_space() does it, and may help to minimize context
switches.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Other transports use rds_page_copy_user, which updates our
s_copy_to_user counter. TCP doesn't, so it needs to explicity
call rds_stats_add().
Reported-by: Richard Frank <richard.frank@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most likely cut n paste error - sendmsg() was checking sock_rcvtimeo.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
BUGging on a runtime error code should be avoided. This
patch also eliminates all other BUG()s that have no real
reason to exist.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Joe Perches <joe@perches.com>
Cc: Andy Grover <andy.grover@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (1815 commits)
mac80211: fix reorder buffer release
iwmc3200wifi: Enable wimax core through module parameter
iwmc3200wifi: Add wifi-wimax coexistence mode as a module parameter
iwmc3200wifi: Coex table command does not expect a response
iwmc3200wifi: Update wiwi priority table
iwlwifi: driver version track kernel version
iwlwifi: indicate uCode type when fail dump error/event log
iwl3945: remove duplicated event logging code
b43: fix two warnings
ipw2100: fix rebooting hang with driver loaded
cfg80211: indent regulatory messages with spaces
iwmc3200wifi: fix NULL pointer dereference in pmkid update
mac80211: Fix TX status reporting for injected data frames
ath9k: enable 2GHz band only if the device supports it
airo: Fix integer overflow warning
rt2x00: Fix padding bug on L2PAD devices.
WE: Fix set events not propagated
b43legacy: avoid PPC fault during resume
b43: avoid PPC fault during resume
tcp: fix a timewait refcnt race
...
Fix up conflicts due to sysctl cleanups (dead sysctl_check code and
CTL_UNNUMBERED removed) in
kernel/sysctl_check.c
net/ipv4/sysctl_net_ipv4.c
net/ipv6/addrconf.c
net/sctp/sysctl.c
Not including net/atm/
Compiled tested x86 allyesconfig only
Added a > 80 column line or two, which I ignored.
Existing checkpatch plaints willfully, cheerfully ignored.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RDMA CM is intended to support the use of a loopback address
when establishing a connection; however, the behavior of the CM
when loopback addresses are used is confusing and does not always
work, depending on whether loopback was specified by the server,
the client, or both.
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 rdam_bind_addr,
no device is associated with the rdma_cm_id. Fix this.
If a loopback address is specified by the client as the destination
address for a connection, it will fail to establish a connection.
This is true even if the server is listing across all addresses or
on the loopback address itself. The issue is that the server tries
to translate the IP address carried in the REQ message to a local
net_device address, which fails. The translation is not needed in
this case, since the REQ carries the actual HW address that should
be used.
Finally, cleanup loopback support to be more transport neutral.
Replace separate calls to get/set the sgid and dgid from the
device address to a single call that behaves correctly depending
on the format of the device address. And support both IPv4 and
IPv6 address formats.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
[ Fixed RDS build by s/ib_addr_get/rdma_addr_get/ - Roland ]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
For consistency drop & in front of every proc_handler. Explicity
taking the address is unnecessary and it prevents optimizations
like stubbing the proc_handlers to NULL.
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Joe Perches <joe@perches.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Now that sys_sysctl is a compatiblity wrapper around /proc/sys
all sysctl strategy routines, and all ctl_name and strategy
entries in the sysctl tables are unused, and can be
revmoed.
In addition neigh_sysctl_register has been modified to no longer
take a strategy argument and it's callers have been modified not
to pass one.
Cc: "David Miller" <davem@davemloft.net>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: netdev@vger.kernel.org
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
The generic __sock_create function has a kern argument which allows the
security system to make decisions based on if a socket is being created by
the kernel or by userspace. This patch passes that flag to the
net_proto_family specific create function, so it can do the same thing.
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move receive processing from event handler to a tasklet.
This should help prevent hangcheck timer from going off
when RDS is under heavy load.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This issue was discovered by HP's Pradeep and fixed in OFED
1.3, but not fixed in later versions, since the fix's implementation
was not immediately applyable to the later code. This patch should
do the trick for 1.4+ codebases.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove explicit destruction of passive connection when destroying
active end of the connection. The passive end is also on the
device's connection list, and will thus be cleaned up properly.
Panic was caused by trying to clean it up twice.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
"At rds_ib_recv_refill_one(), it first executes atomic_read(&rds_ib_allocation)
for if-condition checking,
and then executes atomic_inc(&rds_ib_allocation) if the condition was
not satisfied.
However, if any other code which updates rds_ib_allocation executes
between these two atomic operation executions,
it seems that it may result race condition. (especially when
rds_ib_allocation + 1 == rds_ib_sysctl_max_recv_allocation)"
This patch fixes this by using atomic_inc_unless to eliminate the
possibility of allocating more than rds_ib_sysctl_max_recv_allocation
and then decrementing the count if the allocation fails. It also
makes an identical change to the iwarp transport.
Reported-by: Shin Hong <hongshin@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RDS currently supports a GET_MR sockopt to establish a
memory region (MR) for a chunk of memory. However, the fastreg
method ties a MR to a particular destination. The GET_MR_FOR_DEST
sockopt allows the remote machine to be specified, and thus
support for fastreg (aka FRWRs).
Note that this patch does *not* do all of this - it simply
implements the new sockopt in terms of the old one, so applications
can begin to use the new sockopt in preparation for cutover to
FRWRs.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to have better cache layouts of struct sock (separate zones
for rx/tx paths), we need this preliminary patch.
Goal is to transfert fields used at lookup time in the first
read-mostly cache line (inside struct sock_common) and move sk_refcnt
to a separate cache line (only written by rx path)
This patch adds inet_ prefix to daddr, rcv_saddr, dport, num, saddr,
sport and id fields. This allows a future patch to define these
fields as macros, like sk_refcnt, without name clashes.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All usages of structure net_proto_ops should be declared const.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This provides safety against negative optlen at the type
level instead of depending upon (sometimes non-trivial)
checks against this sprinkled all over the the place, in
each and every implementation.
Based upon work done by Arjan van de Ven and feedback
from Linus Torvalds.
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (66 commits)
be2net: fix some cmds to use mccq instead of mbox
atl1e: fix 2.6.31-git4 -- ATL1E 0000:03:00.0: DMA-API: device driver frees DMA
pkt_sched: Fix qstats.qlen updating in dump_stats
ipv6: Log the affected address when DAD failure occurs
wl12xx: Fix print_mac() conversion.
af_iucv: fix race when queueing skbs on the backlog queue
af_iucv: do not call iucv_sock_kill() twice
af_iucv: handle non-accepted sockets after resuming from suspend
af_iucv: fix race in __iucv_sock_wait()
iucv: use correct output register in iucv_query_maxconn()
iucv: fix iucv_buffer_cpumask check when calling IUCV functions
iucv: suspend/resume error msg for left over pathes
wl12xx: switch to %pM to print the mac address
b44: the poll handler b44_poll must not enable IRQ unconditionally
ipv6: Ignore route option with ROUTER_PREF_INVALID
bonding: make ab_arp select active slaves as other modes
cfg80211: fix SME connect
rc80211_minstrel: fix contention window calculation
ssb/sdio: fix printk format warnings
p54usb: add Zcomax XG-705A usbid
...
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu: (46 commits)
powerpc64: convert to dynamic percpu allocator
sparc64: use embedding percpu first chunk allocator
percpu: kill lpage first chunk allocator
x86,percpu: use embedding for 64bit NUMA and page for 32bit NUMA
percpu: update embedding first chunk allocator to handle sparse units
percpu: use group information to allocate vmap areas sparsely
vmalloc: implement pcpu_get_vm_areas()
vmalloc: separate out insert_vmalloc_vm()
percpu: add chunk->base_addr
percpu: add pcpu_unit_offsets[]
percpu: introduce pcpu_alloc_info and pcpu_group_info
percpu: move pcpu_lpage_build_unit_map() and pcpul_lpage_dump_cfg() upward
percpu: add @align to pcpu_fc_alloc_fn_t
percpu: make @dyn_size mandatory for pcpu_setup_first_chunk()
percpu: drop @static_size from first chunk allocators
percpu: generalize first chunk allocator selection
percpu: build first chunk allocators selectively
percpu: rename 4k first chunk allocator to page
percpu: improve boot messages
percpu: fix pcpu_reclaim() locking
...
Fix trivial conflict as by Tejun Heo in kernel/sched.c
Now that RDS transports are no longer compiled-in to RDS core,
there is now the possibility that they will not be loaded. This
adds a helpful suggestion when rds_bind() fails to find a transport.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that transports can be loaded in arbitrary order,
it is important for rds_trans_get_preferred() to look
for them in a particular order, instead of walking the list
until it finds a transport that works for a given address.
Now, each transport registers for a specific transport slot,
and these are ordered so that preferred transports come first,
and then if they are not loaded, other transports are queried.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Enable the building of transports as modules.
Also, improve consistency of Kconfig messages in relation to other
protocols, and move build dependency on IB from the RDS core code
to the rds_rdma module.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that rdma and tcp transports will be modularized,
we need to export a number of functions so they can call them.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This code allows RDS to be tunneled over a TCP connection.
RDMA operations are disabled when using TCP transport,
but this frees RDS from the IB/RDMA stack dependency, and allows
it to be used with standard Ethernet adapters, or in a VM.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Elsewhere the sin_family field holds a value with a name of the form
AF_..., so it seems reasonable to do so here as well. Also the values of
PF_INET and AF_INET are the same.
The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
struct sockaddr_in sip;
@@
(
sip.sin_family ==
- PF_INET
+ AF_INET
|
sip.sin_family !=
- PF_INET
+ AF_INET
|
sip.sin_family =
- PF_INET
+ AF_INET
)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
String literals are constant, and usually, we can also tag the array
of pointers const too, moving it to the .rodata section.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a comment for what's going on. Remove negative logic.
I find this much easier to understand quickly, although
there are a few lines duplicated.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The existing code treated page_shift as a variable, when in fact we
always want to have the fastreg page size be the same as the arch's
page size -- and it is, so this doesn't need to be a variable.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While FMRs allow significant flexibility in what size of pages they can use,
we really just want FMR pages to match CPU page size. Roland says we can
count on this always being supported, so this simplifies things.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Completion or congestion notifications were not being checked
if the socket went to sleep. This patch fixes that.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Backwards compatibility with rds 3.0 causes protocol-
based flow control to be disabled as a side-effect.
I don't want to pull out FC support from the IB transport
but I do want to document and keep the sysctl consistent
if possible.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since RDS 3.0 and 3.1 have different packet formats,
we need to wait until after protocol negotiation
is complete to layout the rx buffers.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Protocol negotiation is logically a property of the
transports, so rds core need not set it.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Of course len is in bytes. Calling it data_len hopefully indicates
a little better what the variable is actually for.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The big differences between RDS 3.0 and 3.1 are protocol-level
flow control, and with 3.1 the header is in front of the data. The header
always ends up in the header buffer, and the data goes in the data page.
In 3.0 our "header" is a trailer, and will end up either in the data
page, the header buffer, or split across the two. Since 3.1 is backwards-
compatible with 3.0, we need to continue to support these cases. This
patch does that -- if using RDS 3.0 wire protocol, it will copy the header
from wherever it ended up into the header buffer.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RDS on IB uses privdata to do protocol version negotiation. Apparently
the IB stack will return a larger privdata buffer than the struct we were
expecting. Just to be extra-sure, this patch adds some checks in this area.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This will be default cause IB connections to failover faster,
but allow a longer retry count to be used if desired.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are a few places where ___cacheline_aligned* is used with
DEFINE_PER_CPU(). Use DEFINE_PER_CPU_SHARED_ALIGNED() instead.
DEFINE_PER_CPU_SHARED_ALIGNED() applies alignment only on SMPs. While
all other converted places used _in_smp variant or only get compiled
for SMP, net/rds used unconditional ____cacheline_aligned. I don't
see any reason these data structures should be aligned on UP and thus
converted together.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Mike Frysinger <vapier@gentoo.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Andy Grover <andy.grover@oracle.com>
In non-SMP mode, the variable section attribute specified by DECLARE_PER_CPU()
does not agree with that specified by DEFINE_PER_CPU(). This means that
architectures that have a small data section references relative to a base
register may throw up linkage errors due to too great a displacement between
where the base register points and the per-CPU variable.
On FRV, the .h declaration says that the variable is in the .sdata section, but
the .c definition says it's actually in the .data section. The linker throws
up the following errors:
kernel/built-in.o: In function `release_task':
kernel/exit.c:78: relocation truncated to fit: R_FRV_GPREL12 against symbol `per_cpu__process_counts' defined in .data section in kernel/built-in.o
kernel/exit.c:78: relocation truncated to fit: R_FRV_GPREL12 against symbol `per_cpu__process_counts' defined in .data section in kernel/built-in.o
To fix this, DECLARE_PER_CPU() should simply apply the same section attribute
as does DEFINE_PER_CPU(). However, this is made slightly more complex by
virtue of the fact that there are several variants on DEFINE, so these need to
be matched by variants on DECLARE.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
rdma_create_id() doesn't return NULL, only ERR_PTR().
Found by smatch (http://repo.or.cz/w/smatch.git). Compile tested.
regards,
dan carpenter
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rdma_create_id() returns ERR_PTR() not null.
Found by smatch (http://repo.or.cz/w/smatch.git). Compile tested.
regards,
dan carpenter
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use kmem_cache_zalloc instead of kmem_cache_alloc/memset.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove unused #include <version.h> in net/rds/af_rds.c.
Signed-off-by: Huang Weiyi <weiyi.huang@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the new function that is simpler and faster.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The first message to a remote node should prompt a new connection.
Even an RDMA op via CMSG. Therefore move CMSG parsing to after
connection establishment.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Putting the constant first is a supposed "best practice" that actually makes
the code harder to read.
Thanks to Roland Dreier for finding a bug in this "simple, obviously correct"
patch.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix hack that restricts the credit advertisement to 127.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RDS_LL_SEND_FULL bit should be set when we stop transmitted due to
flow control. Otherwise the send worker will keep trying as opposed to
sleeping until we unthrottle. Saves CPU.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Had some lingering instances of _iw_ variable names from when
the listen code was centralized into rdma_transport.c
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the recv ring low water mark is 1/4 the depth. Performance
measurements show that this limits iWARP throughput by flow controlling
the rds-stress senders. Setting it to 1/2 seems to max the T3
performance. I tried even higher levels but that didn't help and it
started to increase the rds thread cpu utilization.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Andy Grover
Zach Brown
Chris Mason
Tina Yang
Eric Dumazet
Julia Lawall
Joe Perches
David S. Miller
Dan Carpenter
Tejun Heo
Sherman Pun
Linus Torvalds
Sean Hefty
Eric W. Biederman
Eric Paris
Stephen Hemminger
Alexey Dobriyan
Jan Engelhardt
David Howells
Wei Yongjun
Huang Weiyi
Steve Wise