Right now there's nothing to stop the various paths that use
rs->rs_transport from racing with rmmod and executing freed transport
code. The simple fix is to have binding to a transport also hold a
reference to the transport's module, removing this class of races.
We already had an unused t_owner field which was set for the modular
transports and which wasn't set for the built-in loop transport.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rs_transport is now also used by the rdma paths once the socket is
bound. We don't need this stale comment to tell us what cscope can.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_conn_destroy() can race with all other modifications of the
rds_conn_count but it was modifying the count without locking.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB device list wasn't protected by any locking. Traversal in
both the get_mr and FMR flushing paths could race with additon and
removal.
List manipulation is done with RCU primatives and is protected by the
write side of a rwsem. The list traversal in the get_mr fast path is
protected by a rcu read critical section. The FMR list traversal is
more problematic because it can block while traversing the list. We
protect this with the read side of the rwsem.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
It's nice to not have to go digging in the code to see which event
occurred. It's easy to throw together a quick array that maps the ib
event enums to their strings. I didn't see anything in the stack that
does this translation for us, but I also didn't look very hard.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Flushing FMRs is somewhat expensive, and is currently kicked off when
the interrupt handler notices that we are getting low. The result of
this is that FMR flushing only happens from the interrupt cpus.
This spreads the load more effectively by triggering flushes just before
we allocate a new FMR.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
We're seeing bugs today where IB connection shutdown clears the send
ring while the tasklet is processing completed sends. Implementation
details cause this to dereference a null pointer. Shutdown needs to
wait for send completion to stop before tearing down the connection. We
can't simply wait for the ring to empty because it may contain
unsignaled sends that will never be processed.
This patch tracks the number of signaled sends that we've posted and
waits for them to complete. It also makes sure that the tasklet has
finished executing.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
We are *definitely* counting cycles as closely as DaveM, so
ensure hwcache alignment for our recv ring control structs.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
The recv refill path was leaking fragments because the recv event handler had
marked a ring element as free without freeing its frag. This was happening
because it wasn't processing receives when the conn wasn't marked up or
connecting, as can be the case if it races with rmmod.
Two observations support always processing receives in the callback.
First, buildup should only post receives, thus triggering recv event handler
calls, once it has built up all the state to handle them. Teardown should
destroy the CQ and drain the ring before tearing down the state needed to
process recvs. Both appear to be true today.
Second, this test was fundamentally racy. There is nothing to stop rmmod and
connection destruction from swooping in the moment after the conn state was
sampled but before real receive procesing starts.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
We were seeing very nasty bugs due to fundamental assumption the current code
makes about concurrent work struct processing. The code simpy isn't able to
handle concurrent connection shutdown work function execution today, for
example, which is very much possible once a multi-threaded krdsd was
introduced. The problem compounds as additional work structs are added to the
mix.
krdsd is no longer perforance critical now that send and receive posting and
FMR flushing are done elsewhere, so the safest fix is to move back to the
single threaded krdsd that the current code was built around.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
This patch moves the FMR flushing work in to its own mult-threaded work queue.
This is to maintain performance in preparation for returning the main krdsd
work queue back to a single threaded work queue to avoid deep-rooted
concurrency bugs.
This is also good because it further separates FMRs, which might be removed
some day, from the rest of the code base.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
IB connections were not being destroyed during rmmod.
First, recently IB device removal callback was changed to disconnect
connections that used the removing device rather than destroying them. So
connections with devices during rmmod were not being destroyed.
Second, rds_ib_destroy_nodev_conns() was being called before connections are
disassociated with devices. It would almost never find connections in the
nodev list.
We first get rid of rds_ib_destroy_conns(), which is no longer called, and
refactor the existing caller into the main body of the function and get rid of
the list and lock wrappers.
Then we call rds_ib_destroy_nodev_conns() *after* ib_unregister_client() has
removed the IB device from all the conns and put the conns on the nodev list.
The result is that IB connections are destroyed by rmmod.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB client removal callback can queue work to drop the final reference
to an IB device. We have to make sure that this function has returned before
we complete rmmod or the work threads can try to execute freed code.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Using a delayed work queue helps us make sure a healthy number of FMRs
have queued up over the limit. It makes for a large improvement in RDMA
iops.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
FRM allocation and recycling is performance critical and fairly lock
intensive. The current code has a per connection lock that all
processes bang on and it becomes a major bottleneck on large systems.
This changes things to use a number of cmpxchg based lists instead,
allowing us to go through the whole FMR lifecycle without locking inside
RDS.
Zach Brown pointed out that our usage of cmpxchg for xlist removal is
racey if someone manages to remove and add back an FMR struct into the list
while another CPU can see the FMR's address at the head of the list.
The second CPU might assume the list hasn't changed when in fact any
number of operations might have happened in between the deletion and
reinsertion.
This commit maintains a per cpu count of CPUs that are currently
in xlist removal, and establishes a grace period to make sure that
nobody can see an entry we have just removed from the list.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
rds_send_xmit() was changed to hold an interrupt masking spinlock instead of a
mutex so that it could be called from the IB receive tasklet path. This broke
the TCP transport because its xmit method can block and masks and unmasks
interrupts.
This patch serializes callers to rds_send_xmit() with a simple bit instead of
the current spinlock or previous mutex. This enables rds_send_xmit() to be
called from any context and to call functions which block. Getting rid of the
c_send_lock exposes the bare c_lock acquisitions which are changed to block
interrupts.
A waitqueue is added so that rds_conn_shutdown() can wait for callers to leave
rds_send_xmit() before tearing down partial send state. This lets us get rid
of c_senders.
rds_send_xmit() is changed to check the conn state after acquiring the
RDS_IN_XMIT bit to resolve races with the shutdown path. Previously both
worked with the conn state and then the lock in the same order, allowing them
to race and execute the paths concurrently.
rds_send_reset() isn't racing with rds_send_xmit() now that rds_conn_shutdown()
properly ensures that rds_send_xmit() can't start once the conn state has been
changed. We can remove its previous use of the spinlock.
Finally, c_send_generation is redundant. Callers can race to test the c_flags
bit by simply retrying instead of racing to test the c_send_generation atomic.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
conn->c_lock is acquired in interrupt context. rds_conn_message_info() is
called from user context and was acquiring c_lock without blocking interrupts,
leading to possible deadlocks.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
rds_send_acked_before() wasn't blocking interrupts when acquiring c_lock from
user context but nothing calls it. Rather than fix its use of c_lock we just
remove the function.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
When prefilling the rds frags, we end up doing a lot of allocations.
We're not in atomic context here, and so there's no reason to dip into
atomic reserves. This changes the prefills to use masks that allow
waiting.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This patch is based heavily on an initial patch by Chris Mason.
Instead of freeing slab memory and pages, it keeps them, and
funnels them back to be reused.
The lock minimization strategy uses xchg and cmpxchg atomic ops
for manipulation of pointers to list heads. We anchor the lists with a
pointer to a list_head struct instead of a static list_head struct.
We just have to carefully use the existing primitives with
the difference between a pointer and a static head struct.
For example, 'list_empty()' means that our anchor pointer points to a list with
a single item instead of meaning that our static head element doesn't point to
any list items.
Original patch by Chris, with significant mods and fixes by Andy and Zach.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: Zach Brown <zach.brown@oracle.com>
All it does is call unmap_sg(), so just call that directly.
The comment above unmap_page also may be incorrect, so we
shouldn't hold on to it, either.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
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>
Zach Brown