| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
[ Upstream commit 6bf536eb5c8ca011d1ff57b5c5f7c57ceac06a37 ]
rmbe_update_limit is used to limit announcing receive
window updating too frequently. RFC7609 request a minimal
increase in the window size of 10% of the receive buffer
space. But current implementation used:
min_t(int, rmbe_size / 10, SOCK_MIN_SNDBUF / 2)
and SOCK_MIN_SNDBUF / 2 == 2304 Bytes, which is almost
always less then 10% of the receive buffer space.
This causes the receiver always sending CDC message to
update its consumer cursor when it consumes more then 2K
of data. And as a result, we may encounter something like
"TCP silly window syndrome" when sending 2.5~8K message.
This patch fixes this using max(rmbe_size / 10, SOCK_MIN_SNDBUF / 2).
With this patch and SMC autocorking enabled, qperf 2K/4K/8K
tcp_bw test shows 45%/75%/40% increase in throughput respectively.
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
commit 4940a1fdf31c39f0806ac831cde333134862030b upstream.
The problem of SMC_CLC_DECL_ERR_REGRMB on the server is very clear.
Based on the fact that whether a new SMC connection can be accepted or
not depends on not only the limit of conn nums, but also the available
entries of rtoken. Since the rtoken release is trigger by peer, while
the conn nums is decrease by local, tons of thing can happen in this
time difference.
This only thing that needs to be mentioned is that now all connection
creations are completely protected by smc_server_lgr_pending lock, it's
enough to check only the available entries in rtokens_used_mask.
Fixes: cd6851f30386 ("smc: remote memory buffers (RMBs)")
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 0537f0a2151375dcf90c1bbfda6a0aaf57164e89 upstream.
The main reason for this unexpected SMC_CLC_DECL_ERR_REGRMB in client
dues to following execution sequence:
Server Conn A: Server Conn B: Client Conn B:
smc_lgr_unregister_conn
smc_lgr_register_conn
smc_clc_send_accept ->
smc_rtoken_add
smcr_buf_unuse
-> Client Conn A:
smc_rtoken_delete
smc_lgr_unregister_conn() makes current link available to assigned to new
incoming connection, while smcr_buf_unuse() has not executed yet, which
means that smc_rtoken_add may fail because of insufficient rtoken_entry,
reversing their execution order will avoid this problem.
Fixes: 3e034725c0d8 ("net/smc: common functions for RMBs and send buffers")
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 56d99e81ecbc997a5f984684d0eeb583992b2072 upstream.
A hung_task is observed when removing SMC-R devices. Suppose that
a link group has two active links(lnk_A, lnk_B) associated with two
different SMC-R devices(dev_A, dev_B). When dev_A is removed, the
link group will be removed from smc_lgr_list and added into
lgr_linkdown_list. lnk_A will be cleared and smcibdev(A)->lnk_cnt
will reach to zero. However, when dev_B is removed then, the link
group can't be found in smc_lgr_list and lnk_B won't be cleared,
making smcibdev->lnk_cnt never reaches zero, which causes a hung_task.
This patch fixes this issue by restoring the implementation of
smc_smcr_terminate_all() to what it was before commit 349d43127dac
("net/smc: fix kernel panic caused by race of smc_sock"). The original
implementation also satisfies the intention that make sure QP destroy
earlier than CQ destroy because we will always wait for smcibdev->lnk_cnt
reaches zero, which guarantees QP has been destroyed.
Fixes: 349d43127dac ("net/smc: fix kernel panic caused by race of smc_sock")
Signed-off-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 349d43127dac00c15231e8ffbcaabd70f7b0e544 ]
A crash occurs when smc_cdc_tx_handler() tries to access smc_sock
but smc_release() has already freed it.
[ 4570.695099] BUG: unable to handle page fault for address: 000000002eae9e88
[ 4570.696048] #PF: supervisor write access in kernel mode
[ 4570.696728] #PF: error_code(0x0002) - not-present page
[ 4570.697401] PGD 0 P4D 0
[ 4570.697716] Oops: 0002 [#1] PREEMPT SMP NOPTI
[ 4570.698228] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.16.0-rc4+ #111
[ 4570.699013] Hardware name: Alibaba Cloud Alibaba Cloud ECS, BIOS 8c24b4c 04/0
[ 4570.699933] RIP: 0010:_raw_spin_lock+0x1a/0x30
<...>
[ 4570.711446] Call Trace:
[ 4570.711746] <IRQ>
[ 4570.711992] smc_cdc_tx_handler+0x41/0xc0
[ 4570.712470] smc_wr_tx_tasklet_fn+0x213/0x560
[ 4570.712981] ? smc_cdc_tx_dismisser+0x10/0x10
[ 4570.713489] tasklet_action_common.isra.17+0x66/0x140
[ 4570.714083] __do_softirq+0x123/0x2f4
[ 4570.714521] irq_exit_rcu+0xc4/0xf0
[ 4570.714934] common_interrupt+0xba/0xe0
Though smc_cdc_tx_handler() checked the existence of smc connection,
smc_release() may have already dismissed and released the smc socket
before smc_cdc_tx_handler() further visits it.
smc_cdc_tx_handler() |smc_release()
if (!conn) |
|
|smc_cdc_tx_dismiss_slots()
| smc_cdc_tx_dismisser()
|
|sock_put(&smc->sk) <- last sock_put,
| smc_sock freed
bh_lock_sock(&smc->sk) (panic) |
To make sure we won't receive any CDC messages after we free the
smc_sock, add a refcount on the smc_connection for inflight CDC
message(posted to the QP but haven't received related CQE), and
don't release the smc_connection until all the inflight CDC messages
haven been done, for both success or failed ones.
Using refcount on CDC messages brings another problem: when the link
is going to be destroyed, smcr_link_clear() will reset the QP, which
then remove all the pending CQEs related to the QP in the CQ. To make
sure all the CQEs will always come back so the refcount on the
smc_connection can always reach 0, smc_ib_modify_qp_reset() was replaced
by smc_ib_modify_qp_error().
And remove the timeout in smc_wr_tx_wait_no_pending_sends() since we
need to wait for all pending WQEs done, or we may encounter use-after-
free when handling CQEs.
For IB device removal routine, we need to wait for all the QPs on that
device been destroyed before we can destroy CQs on the device, or
the refcount on smc_connection won't reach 0 and smc_sock cannot be
released.
Fixes: 5f08318f617b ("smc: connection data control (CDC)")
Reported-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit 90cee52f2e780345d3629e278291aea5ac74f40f ]
We found smc_llc_send_link_delete_all() sometimes wait
for 2s timeout when testing with RDMA link up/down.
It is possible when a smc_link is in ACTIVATING state,
the underlaying QP is still in RESET or RTR state, which
cannot send any messages out.
smc_llc_send_link_delete_all() use smc_link_usable() to
checks whether the link is usable, if the QP is still in
RESET or RTR state, but the smc_link is in ACTIVATING, this
LLC message will always fail without any CQE entering the
CQ, and we will always wait 2s before timeout.
Since we cannot send any messages through the QP before
the QP enter RTS. I add a wrapper smc_link_sendable()
which checks the state of QP along with the link state.
And replace smc_link_usable() with smc_link_sendable()
in all LLC & CDC message sending routine.
Fixes: 5f08318f617b ("smc: connection data control (CDC)")
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit 95f7f3e7dc6bd2e735cb5de11734ea2222b1e05a ]
Commit 8f3d65c16679 ("net/smc: fix wait on already cleared link")
introduced link refcounting to avoid waits on already cleared links.
This patch extents and improves the refcounting to cover all
remaining possible cases for this kind of error situation.
Fixes: 15e1b99aadfb ("net/smc: no WR buffer wait for terminating link group")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
commit 789b6cc2a5f9123b9c549b886fdc47c865cfe0ba upstream.
smc_lgr_cleanup_early() meant to delete the link
group from the link group list, but it deleted
the list head by mistake.
This may cause memory corruption since we didn't
remove the real link group from the list and later
memseted the link group structure.
We got a list corruption panic when testing:
[ 231.277259] list_del corruption. prev->next should be ffff8881398a8000, but was 0000000000000000
[ 231.278222] ------------[ cut here ]------------
[ 231.278726] kernel BUG at lib/list_debug.c:53!
[ 231.279326] invalid opcode: 0000 [#1] SMP NOPTI
[ 231.279803] CPU: 0 PID: 5 Comm: kworker/0:0 Not tainted 5.10.46+ #435
[ 231.280466] Hardware name: Alibaba Cloud ECS, BIOS 8c24b4c 04/01/2014
[ 231.281248] Workqueue: events smc_link_down_work
[ 231.281732] RIP: 0010:__list_del_entry_valid+0x70/0x90
[ 231.282258] Code: 4c 60 82 e8 7d cc 6a 00 0f 0b 48 89 fe 48 c7 c7 88 4c
60 82 e8 6c cc 6a 00 0f 0b 48 89 fe 48 c7 c7 c0 4c 60 82 e8 5b cc 6a 00 <0f>
0b 48 89 fe 48 c7 c7 00 4d 60 82 e8 4a cc 6a 00 0f 0b cc cc cc
[ 231.284146] RSP: 0018:ffffc90000033d58 EFLAGS: 00010292
[ 231.284685] RAX: 0000000000000054 RBX: ffff8881398a8000 RCX: 0000000000000000
[ 231.285415] RDX: 0000000000000001 RSI: ffff88813bc18040 RDI: ffff88813bc18040
[ 231.286141] RBP: ffffffff8305ad40 R08: 0000000000000003 R09: 0000000000000001
[ 231.286873] R10: ffffffff82803da0 R11: ffffc90000033b90 R12: 0000000000000001
[ 231.287606] R13: 0000000000000000 R14: ffff8881398a8000 R15: 0000000000000003
[ 231.288337] FS: 0000000000000000(0000) GS:ffff88813bc00000(0000) knlGS:0000000000000000
[ 231.289160] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 231.289754] CR2: 0000000000e72058 CR3: 000000010fa96006 CR4: 00000000003706f0
[ 231.290485] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 231.291211] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 231.291940] Call Trace:
[ 231.292211] smc_lgr_terminate_sched+0x53/0xa0
[ 231.292677] smc_switch_conns+0x75/0x6b0
[ 231.293085] ? update_load_avg+0x1a6/0x590
[ 231.293517] ? ttwu_do_wakeup+0x17/0x150
[ 231.293907] ? update_load_avg+0x1a6/0x590
[ 231.294317] ? newidle_balance+0xca/0x3d0
[ 231.294716] smcr_link_down+0x50/0x1a0
[ 231.295090] ? __wake_up_common_lock+0x77/0x90
[ 231.295534] smc_link_down_work+0x46/0x60
[ 231.295933] process_one_work+0x18b/0x350
Fixes: a0a62ee15a829 ("net/smc: separate locks for SMCD and SMCR link group lists")
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Acked-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 587acad41f1bc48e16f42bb2aca63bf323380be8 ]
Coverity reports a possible NULL dereferencing problem:
in smc_vlan_by_tcpsk():
6. returned_null: netdev_lower_get_next returns NULL (checked 29 out of 30 times).
7. var_assigned: Assigning: ndev = NULL return value from netdev_lower_get_next.
1623 ndev = (struct net_device *)netdev_lower_get_next(ndev, &lower);
CID 1468509 (#1 of 1): Dereference null return value (NULL_RETURNS)
8. dereference: Dereferencing a pointer that might be NULL ndev when calling is_vlan_dev.
1624 if (is_vlan_dev(ndev)) {
Remove the manual implementation and use netdev_walk_all_lower_dev() to
iterate over the lower devices. While on it remove an obsolete function
parameter comment.
Fixes: cb9d43f67754 ("net/smc: determine vlan_id of stacked net_device")
Suggested-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit cf4f5530bb55ef7d5a91036b26676643b80b1616 ]
The link_id is supposed to be unique, but smcr_next_link_id() doesn't
skip the used link_id as expected. So the patch fixes this.
Fixes: 026c381fb477 ("net/smc: introduce link_idx for link group array")
Signed-off-by: Wen Gu <guwen@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Acked-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit a18cee4791b1123d0a6579a7c89f4b87e48abe03 ]
The abort_work is scheduled when a connection was detected to be
out-of-sync after a link failure. The work calls smc_conn_kill(),
which calls smc_close_active_abort() and that might end up calling
smc_close_cancel_work().
smc_close_cancel_work() cancels any pending close_work and tx_work but
needs to release the sock_lock before and acquires the sock_lock again
afterwards. So when the sock_lock was NOT acquired before then it may
be held after the abort_work completes. Thats why the sock_lock is
acquired before the call to smc_conn_kill() in __smc_lgr_terminate(),
but this is missing in smc_conn_abort_work().
Fix that by acquiring the sock_lock first and release it after the
call to smc_conn_kill().
Fixes: b286a0651e44 ("net/smc: handle incoming CDC validation message")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
With the multi-subnet support of SMC-Dv2 the match for existing link
groups should not include the vlanid of the network device.
Set ini->smcd_version accordingly before the call to smc_conn_create()
and use this value in smc_conn_create() to skip the vlanid check.
Fixes: 5c21c4ccafe8 ("net/smc: determine accepted ISM devices")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The patch that repaired the invalid return code in smcd_new_buf_create()
missed to take care of errno ENOSPC which has a special meaning that no
more DMBEs can be registered on the device. Fix that by keeping this
errno value during the translation of the return code.
Fixes: 6b1bbf94ab36 ("net/smc: fix invalid return code in smcd_new_buf_create()")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Minor conflicts in net/mptcp/protocol.h and
tools/testing/selftests/net/Makefile.
In both cases code was added on both sides in the same place
so just keep both.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
smc_ism_register_dmb() returns error codes set by the ISM driver which
are not guaranteed to be negative or in the errno range. Such values
would not be handled by ERR_PTR() and finally the return code will be
used as a memory address.
Fix that by using a valid negative errno value with ERR_PTR().
Fixes: 72b7f6c48708 ("net/smc: unique reason code for exceeded max dmb count")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The SMCD_DMBE_SIZES should include all valid DMBE buffer sizes, so the
correct value is 6 which means 1MB. With 7 the registration of an ISM
buffer would always fail because of the invalid size requested.
Fix that and set the value to 6.
Fixes: c6ba7c9ba43d ("net/smc: add base infrastructure for SMC-D and ISM")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
SMC Version 2 defines a first contact extension for CLC accept
and CLC confirm. This patch covers sending and receiving of the
CLC first contact extension.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
SMCD Version 2 allows to propose up to 8 additional ISM devices
offered to the peer as candidates for SMCD communication.
This patch covers the server side, i.e. selection of an ISM device
matching one of the proposed ISM devices, that will be used for
CLC accept
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
SMCD Version 2 allows proposing of up to 8 ISM devices in addition
to the native ISM device of SMCD Version 1.
This patch prepares the struct smc_init_info to deal with these
additional 8 ISM devices.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There are 6 types of workers which exist per smc connection. 3 of them
are used for listen and handshake processing, another 2 are used for
close and abort processing and 1 is the tx worker that moves calls to
sleeping functions into a worker.
To prevent flooding of the system work queue when many connections are
opened or closed at the same time (some pattern uperf implements), move
those workers to one of 3 smc-specific work queues. Two work queues are
module-global and used for handshake and close workers. The third work
queue is defined per link group and used by the tx workers that may
sleep waiting for resources of this link group.
And in smc_llc_enqueue() queue the llc_event_work work to the system
prio work queue because its critical that this work is started fast.
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
smc_lgr_cleanup_early() schedules the free worker with delay. DMB
unregistering occurs in this delayed worker increasing the risk
to reach the SMCD SBA limit without need. Terminate the
linkgroup immediately, since termination means early DMB unregistering.
For SMCD the global smc_server_lgr_pending lock is given up early.
A linkgroup to be given up with smc_lgr_cleanup_early() may already
contain more than one connection. Using __smc_lgr_terminate() in
smc_lgr_cleanup_early() covers this.
And consolidate smc_ism_put_vlan() and smc_put_device() into smc_lgr_free()
only.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
smc_clc_send_accept() and smc_clc_send_confirm() are quite similar.
Move common code into a separate function smc_clc_send_confirm_accept().
And introduce separate SMCD and SMCR struct definitions for CLC accept
resp. confirm.
No functional change.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Field names "srv_first_contact" and "cln_first_contact" are misleading,
since they apply to both, server and client. Rename them to
"first_contact_peer" and "first_contact_local".
Rename "ism_gid" by the more precise name "ism_peer_gid".
Rename version constant "SMC_CLC_V1" into "SMC_V1".
No functional change.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When an SMC connection is created, and there is a problem to
create an RMB or DMB, the previously created send buffer is
thrown away as well including buffer descriptor freeing.
Make sure the connection no longer references the freed
buffer descriptor, otherwise bugs like this are possible:
[71556.835148] =============================================================================
[71556.835168] BUG kmalloc-128 (Tainted: G B OE ): Poison overwritten
[71556.835172] -----------------------------------------------------------------------------
[71556.835179] INFO: 0x00000000d20894be-0x00000000aaef63e9 @offset=2724. First byte 0x0 instead of 0x6b
[71556.835215] INFO: Allocated in __smc_buf_create+0x184/0x578 [smc] age=0 cpu=5 pid=46726
[71556.835234] ___slab_alloc+0x5a4/0x690
[71556.835239] __slab_alloc.constprop.0+0x70/0xb0
[71556.835243] kmem_cache_alloc_trace+0x38e/0x3f8
[71556.835250] __smc_buf_create+0x184/0x578 [smc]
[71556.835257] smc_buf_create+0x2e/0xe8 [smc]
[71556.835264] smc_listen_work+0x516/0x6a0 [smc]
[71556.835275] process_one_work+0x280/0x478
[71556.835280] worker_thread+0x66/0x368
[71556.835287] kthread+0x17a/0x1a0
[71556.835294] ret_from_fork+0x28/0x2c
[71556.835301] INFO: Freed in smc_buf_create+0xd8/0xe8 [smc] age=0 cpu=5 pid=46726
[71556.835307] __slab_free+0x246/0x560
[71556.835311] kfree+0x398/0x3f8
[71556.835318] smc_buf_create+0xd8/0xe8 [smc]
[71556.835324] smc_listen_work+0x516/0x6a0 [smc]
[71556.835328] process_one_work+0x280/0x478
[71556.835332] worker_thread+0x66/0x368
[71556.835337] kthread+0x17a/0x1a0
[71556.835344] ret_from_fork+0x28/0x2c
[71556.835348] INFO: Slab 0x00000000a0744551 objects=51 used=51 fp=0x0000000000000000 flags=0x1ffff00000010200
[71556.835352] INFO: Object 0x00000000563480a1 @offset=2688 fp=0x00000000289567b2
[71556.835359] Redzone 000000006783cde2: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835363] Redzone 00000000e35b876e: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835367] Redzone 0000000023074562: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835372] Redzone 00000000b9564b8c: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835376] Redzone 00000000810c6362: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835380] Redzone 0000000065ef52c3: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835384] Redzone 00000000c5dd6984: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835388] Redzone 000000004c480f8f: bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[71556.835392] Object 00000000563480a1: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[71556.835397] Object 000000009c479d06: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[71556.835401] Object 000000006e1dce92: 6b 6b 6b 6b 00 00 00 00 6b 6b 6b 6b 6b 6b 6b 6b kkkk....kkkkkkkk
[71556.835405] Object 00000000227f7cf8: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[71556.835410] Object 000000009a701215: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[71556.835414] Object 000000003731ce76: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[71556.835418] Object 00000000f7085967: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[71556.835422] Object 0000000007f99927: 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b a5 kkkkkkkkkkkkkkk.
[71556.835427] Redzone 00000000579c4913: bb bb bb bb bb bb bb bb ........
[71556.835431] Padding 00000000305aef82: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
[71556.835435] Padding 00000000b1cdd722: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
[71556.835438] Padding 00000000c7568199: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
[71556.835442] Padding 00000000fad4c4d4: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
[71556.835451] CPU: 0 PID: 47939 Comm: kworker/0:15 Tainted: G B OE 5.9.0-rc1uschi+ #54
[71556.835456] Hardware name: IBM 3906 M03 703 (LPAR)
[71556.835464] Workqueue: events smc_listen_work [smc]
[71556.835470] Call Trace:
[71556.835478] [<00000000d5eaeb10>] show_stack+0x90/0xf8
[71556.835493] [<00000000d66fc0f8>] dump_stack+0xa8/0xe8
[71556.835499] [<00000000d61a511c>] check_bytes_and_report+0x104/0x130
[71556.835504] [<00000000d61a57b2>] check_object+0x26a/0x2e0
[71556.835509] [<00000000d61a59bc>] alloc_debug_processing+0x194/0x238
[71556.835514] [<00000000d61a8c14>] ___slab_alloc+0x5a4/0x690
[71556.835519] [<00000000d61a9170>] __slab_alloc.constprop.0+0x70/0xb0
[71556.835524] [<00000000d61aaf66>] kmem_cache_alloc_trace+0x38e/0x3f8
[71556.835530] [<000003ff80549bbc>] __smc_buf_create+0x184/0x578 [smc]
[71556.835538] [<000003ff8054a396>] smc_buf_create+0x2e/0xe8 [smc]
[71556.835545] [<000003ff80540c16>] smc_listen_work+0x516/0x6a0 [smc]
[71556.835549] [<00000000d5f0f448>] process_one_work+0x280/0x478
[71556.835554] [<00000000d5f0f6a6>] worker_thread+0x66/0x368
[71556.835559] [<00000000d5f18692>] kthread+0x17a/0x1a0
[71556.835563] [<00000000d6abf3b8>] ret_from_fork+0x28/0x2c
[71556.835569] INFO: lockdep is turned off.
[71556.835573] FIX kmalloc-128: Restoring 0x00000000d20894be-0x00000000aaef63e9=0x6b
[71556.835577] FIX kmalloc-128: Marking all objects used
Fixes: fd7f3a746582 ("net/smc: remove freed buffer from list")
Reviewed-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
SMC tries to make use of SMCD first. If a problem shows up,
it tries to switch to SMCR. If the SMCD initializing problem shows
up after the SMCD connection has already been initialized, field
rx_off keeps the wrong SMCD value for SMCR, which results in corrupted
data at the receiver.
This patch adds an explicit (re-)setting of field rx_off to zero if the
connection uses SMCR.
Fixes: be244f28d22f ("net/smc: add SMC-D support in data transfer")
Reviewed-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When the maximum dmb buffer limit for an ism device is reached no more
dmb buffers can be registered. When this happens the reason code is set
to SMC_CLC_DECL_MEM indicating out-of-memory. This is the same reason
code that is used when no memory could be allocated for the new dmb
buffer.
This is confusing for users when they see this error but there is more
memory available. To solve this set a separate new reason code when the
maximum dmb limit exceeded.
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There is a current limit of 1920 registered dmb buffers per ISM device
for smc-d. One link group can contain 255 connections, each connection
is using one dmb buffer. When the connection is closed then the
registered buffer is held in a queue and is reused by the next
connection. When a link group is 'full' then another link group is
created and uses an own buffer pool. The link groups are added to a
list using list_add() which puts a new link group to the first position
in the list.
In the situation that many connections are opened (>1920) and a few of
them stay open while others are closed quickly we end up with at least 8
link groups. For a new connection a matching link group is looked up,
iterating over the list of link groups. The trailing 7 link groups
all have registered dmb buffers which could be reused, while the first
link group has only a few dmb buffers and then hit the 1920 limit.
Because the first link group is not full (255 connection limit not
reached) it is chosen and finally the connection falls back to TCP
because there is no dmb buffer available in this link group.
There are multiple ways to fix that: using list_add_tail() allows
to scan older link groups first for free buffers which ensures that
buffers are reused first. This fixes the problem for smc-r link groups
as well. For smc-d there is an even better way to address this problem
because smc-d does not have the 255 connections per link group limit.
So fix the problem for smc-d by allowing large link groups.
Fixes: c6ba7c9ba43d ("net/smc: add base infrastructure for SMC-D and ISM")
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Two buffers are allocated for each SMC connection. Each buffer is
added to a buffer list after creation. When the second buffer
allocation fails, the first buffer is freed but not deleted from
the list. This might result in crashes when another connection picks
up the freed buffer later and starts to work with it.
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Fixes: 6511aad3f039 ("net/smc: change smc_buf_free function parameters")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The dma related ...sync_sg... functions check the link state before the
dma function is actually called. But the check in smc_link_usable()
allows links in ACTIVATING state which are not yet mapped to dma memory.
Under high load it may happen that the sync_sg functions are called for
such a link which results in an debug output like
DMA-API: mlx5_core 0002:00:00.0: device driver tries to sync
DMA memory it has not allocated [device address=0x0000000103370000]
[size=65536 bytes]
To fix that introduce a helper to check for the link state ACTIVE and
use it where appropriate. And move the link state update to ACTIVATING
to the end of smcr_link_init() when most initial setup is done.
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Fixes: d854fcbfaeda ("net/smc: add new link state and related helpers")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When a new ib device is up smc will send an add link invitation to the
peer if needed. This is currently done with rudimentary flow control.
Under high workload these add link invitations can disturb other llc
flows because they arrive unexpected. Fix this by integrating the
invitations into the normal llc event flow and handle them as a flow.
While at it, check for already assigned requests in the flow before
the new add link request is assigned.
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Fixes: 1f90a05d9ff9 ("net/smc: add smcr_port_add() and smcr_link_up() processing")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In a link-down condition we notify the SMC server and expect that the
server will finally trigger the link clear processing on the client
side. This could fail when anything along this notification path goes
wrong. Clear the link as part of SMC client link-down processing to
prevent dangling links.
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Fixes: 541afa10c126 ("net/smc: add smcr_port_err() and smcr_link_down() processing")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The similar smc_ib_devices spinlock has been converted to a mutex.
Protecting the smcd_dev_list by a mutex is possible as well. This
patch converts the smcd_dev_list spinlock to a mutex.
Fixes: c6ba7c9ba43d ("net/smc: add base infrastructure for SMC-D and ISM")
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Tests showed this BUG:
[572555.252867] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:935
[572555.252876] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 131031, name: smcapp
[572555.252879] INFO: lockdep is turned off.
[572555.252883] CPU: 1 PID: 131031 Comm: smcapp Tainted: G O 5.7.0-rc3uschi+ #356
[572555.252885] Hardware name: IBM 3906 M03 703 (LPAR)
[572555.252887] Call Trace:
[572555.252896] [<00000000ac364554>] show_stack+0x94/0xe8
[572555.252901] [<00000000aca1f400>] dump_stack+0xa0/0xe0
[572555.252906] [<00000000ac3c8c10>] ___might_sleep+0x260/0x280
[572555.252910] [<00000000acdc0c98>] __mutex_lock+0x48/0x940
[572555.252912] [<00000000acdc15c2>] mutex_lock_nested+0x32/0x40
[572555.252975] [<000003ff801762d0>] mlx5_lag_get_roce_netdev+0x30/0xc0 [mlx5_core]
[572555.252996] [<000003ff801fb3aa>] mlx5_ib_get_netdev+0x3a/0xe0 [mlx5_ib]
[572555.253007] [<000003ff80063848>] smc_pnet_find_roce_resource+0x1d8/0x310 [smc]
[572555.253011] [<000003ff800602f0>] __smc_connect+0x1f0/0x3e0 [smc]
[572555.253015] [<000003ff80060634>] smc_connect+0x154/0x190 [smc]
[572555.253022] [<00000000acbed8d4>] __sys_connect+0x94/0xd0
[572555.253025] [<00000000acbef620>] __s390x_sys_socketcall+0x170/0x360
[572555.253028] [<00000000acdc6800>] system_call+0x298/0x2b8
[572555.253030] INFO: lockdep is turned off.
Function smc_pnet_find_rdma_dev() might be called from
smc_pnet_find_roce_resource(). It holds the smc_ib_devices list
spinlock while calling infiniband op get_netdev(). At least for mlx5
the get_netdev operation wants mutex serialization, which conflicts
with the smc_ib_devices spinlock.
This patch switches the smc_ib_devices spinlock into a mutex to
allow sleeping when calling get_netdev().
Fixes: a4cf0443c414 ("smc: introduce SMC as an IB-client")
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There might be races in scenarios where both SMC link groups are on the
same system. Prevent that by creating separate wait queues for LLC flows
and messages. Switch to non-interruptable versions of wait_event() and
wake_up() for the llc flow waiter to make sure the waiters get control
sequentially. Fine tune the llc_flow_lock to include the assignment of
the message. Write to system log when an unexpected message was
dropped. And remove an extra indirection and use the existing local
variable lgr in smc_llc_enqueue().
Fixes: 555da9af827d ("net/smc: add event-based llc_flow framework")
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Pre-fetch send buffer for the CDC validation message before entering the
send_lock. Without that the send call might fail with -EBUSY because
there are no free buffers and waiting for buffers is not possible under
send_lock.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Print to system log when SMC links are available or go down, link group
state changes or pnetids are applied to and removed from devices.
The log entries are triggered by either user configuration actions or
adapter activation/deactivation events and are not expected to happen
often. The entries help SMC users to keep track of the SMC link group
status and to detect when actions are needed (like to add replacements
for failed adapters).
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The link_uid of an SMC-R link is exchanged between SMC peers and its
value can be used for debugging purposes. Create a unique link_uid
during link initialization and use it in communication with SMC-R peers.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add helper smcr_lgr_link_deactivate_all() and eliminate duplicate code.
In smc_lgr_free(), clear the smc-r links before smc_lgr_free_bufs() is
called so buffers are already prepared for free. The usage of the soft
parameter in __smc_lgr_terminate() is no longer needed, smc_lgr_free()
can be called directly. smc_lgr_terminate_sched() and
smc_smcd_terminate() set lgr->freeing to indicate that the link group
will be freed soon to avoid unnecessary schedules of the free worker.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Allow to set the reason code for the link group termination, and set
meaningful values before termination processing is triggered. This
reason code is sent to the peer in the final delete link message.
When the LLC request or response layer receives a message type that was
not handled, drop a warning and terminate the link group.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
New connections must not be assigned to asymmetric links. Add asymmetric
link tagging using new link variable link_is_asym. The new helpers
smcr_lgr_set_type() and smcr_lgr_set_type_asym() are called to set the
state of the link group, and tag all links accordingly.
smcr_lgr_conn_assign_link() respects the link tagging and will not
assign new connections to links tagged as asymmetric link.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
For new connections, assign a link from the link group, using some
simple load balancing.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add smc_llc_send_message_wait() which uses smc_wr_tx_send_wait() to send
an LLC message and waits for the message send to complete.
smc_llc_send_link_delete_all() calls the new function to send an
DELETE_LINK,ALL LLC message. The RFC states that the sender of this type
of message needs to wait for the completion event of the message
transmission and can terminate the link afterwards.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Call smc_cdc_msg_validate() when a CDC message with the failover
validation bit enabled was received. Validate that the sequence number
sent with the message is one we already have received. If not, messages
were lost and the connection is terminated using a new abort_work.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When a connection is switched to a new link then a link validation
message must be sent to the peer over the new link, containing the
sequence number of the last CDC message that was sent over the old link.
The peer will validate if this sequence number is the same or lower then
the number he received, and abort the connection if messages were lost.
Add smcr_cdc_msg_send_validation() to send the message validation
message and call it when a connection was switched in
smc_switch_cursor().
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add smc_switch_conns() to sw |
