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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
rds_message_alloc_sgs() now returns correctly-initialized
sg lists, so calleds need not do this themselves.
Signed-off-by: Andy Grover <[email protected]>
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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
Clearly separate rdma-related variables in rm from data-related ones.
This is in anticipation of adding atomic support.
Signed-off-by: Andy Grover <[email protected]>
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 <[email protected]>
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 <[email protected]>
Signed-off-by: Andy Grover <[email protected]>
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 <[email protected]>
Reduce indentation in a couple of places
Add static function ixgbe_psum
Add temporary for adapter->stats
Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Jeff Kirsher <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Did not add #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
because no printk in this module used message prefixing.
Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Jeff Kirsher <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Whitespace cleanups.
Move inline keyword after function type declarations.
Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Jeff Kirsher <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The ordering of operations was messed up in the init and as a result when
VMDQ was enabled we were trying to enable TX rings before setting the VFTE
bits. This resulted in a ring that appeared to fail to enable when in fact
it was blocked because the VFTE bits were cleared after the reset.
Signed-off-by: Alexander Duyck <[email protected]>
Signed-off-by: Jeff Kirsher <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Blackhole routes are used when xfrm_lookup() returns -EREMOTE (error
triggered by IKE for example), hence this kind of route is always
temporary and so we should check if a better route exists for next
packets.
Bug has been introduced by commit d11a4dc18b.
Signed-off-by: Jianzhao Wang <[email protected]>
Signed-off-by: Nicolas Dichtel <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch adds NAPI support to the qeth layer 2 and layer 3
discipline. It is important to understand that we can not enable/disable
IRQs as usual, we have to use the corresponding new QDIO interface.
Also to not overdraw the budget we have to stop and restart buffer
processing at any point during processing a bulk of QDIO buffers.
Having the driver NAPI enabled it is possible to turn on GRO for the
layer 3 discipline.
Signed-off-by: Frank Blaschka <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Since the claw code calls ccwgroup_remove_ccwdev(), we need to make sure
CCWGROUP is enabled when CLAW is enabled. Otherwise we hit fun undefined
references at build time:
ERROR: "ccwgroup_remove_ccwdev" [drivers/s390/net/claw.ko] undefined!
ERROR: "ccwgroup_probe_ccwdev" [drivers/s390/net/claw.ko] undefined!
ERROR: "ccwgroup_driver_register" [drivers/s390/net/claw.ko] undefined!
ERROR: "ccwgroup_driver_unregister" [drivers/s390/net/claw.ko] undefined!
ERROR: "ccwgroup_create_from_string" [drivers/s390/net/claw.ko] undefined!
Signed-off-by: Mike Frysinger <[email protected]>
Signed-off-by: Frank Blaschka <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Extend the qdio API to allow polling in the upper-layer driver. This
is needed by qeth to use NAPI.
To use the new interface the upper-layer driver must specify the
queue_start_poll(). This callback is used to signal the upper-layer
driver that is has initiative and must process the inbound queue by
calling qdio_get_next_buffers(). If the upper-layer driver wants to
stop polling it calls qdio_start_irq().
Since adapter interrupts are not completely stoppable qdio implements
a software bit QDIO_QUEUE_IRQS_DISABLED to safely disable interrupts for an
input queue.
The old interface is preserved and will be used as is by zfcp.
Signed-off-by: Jan Glauber <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
Signed-off-by: Frank Blaschka <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Several users report issues with 32-bit adapters when plugged
into PCI slots in machines with >= 4GB ram. In particular AMD
systems with HyperTransport to PCI bridges seem to trigger the
issue, but it isn't limited to only them.
This issue is not easily reproducible here, yet still continues
to occur in the field. For e1000 on PCI devices, just disable DMA
addresses over the 4GB boundary when in PCI (not PCI-X) mode, to
prevent the issue from continuing to pop up. The performance
impact for this is negligible.
The code was refactored to move the init of the hw struct to its
own function. This allows the init to be called very early in
probe, which then allows using hw-> members for this fix.
A slight refactor to the DMA mask code was done for minor
correctness based on the instructions in DMA-API-HOWTO.
Signed-off-by: Jesse Brandeburg <[email protected]>
Signed-off-by: Jeff Kirsher <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
We use rcu_dereference_check(p, rcu_read_lock_held() ||
lockdep_rtnl_is_held()) several times in network stack.
More usages to come too, so its time to create a helper.
Signed-off-by: Eric Dumazet <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
