RDMA/isert: Suppress gcc 7 fall-through complaints

Avoid that gcc 7 reports the following warning when building with W=1:

warning: this statement may fall through [-Wimplicit-fallthrough=]

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/iwcm: Remove a set-but-not-used variable

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/cma: Avoid triggering undefined behavior

According to the C standard the behavior of computations with
integer operands is as follows:
* A computation involving unsigned operands can never overflow,
  because a result that cannot be represented by the resulting
  unsigned integer type is reduced modulo the number that is one
  greater than the largest value that can be represented by the
  resulting type.
* The behavior for signed integer underflow and overflow is
  undefined.

Hence only use unsigned integers when checking for integer
overflow.

This patch is what I came up with after having analyzed the
following smatch warnings:

drivers/infiniband/core/cma.c:3448: cma_resolve_ib_udp() warn: signed overflow undefined. 'offset + conn_param->private_data_len < conn_param->private_data_len'
drivers/infiniband/core/cma.c:3505: cma_connect_ib() warn: signed overflow undefined. 'offset + conn_param->private_data_len < conn_param->private_data_len'

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Acked-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/cm: Suppress gcc 7 fall-through complaints

Avoid that gcc 7 reports the following warning when building with W=1:

warning: this statement may fall through [-Wimplicit-fallthrough=]

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Acked-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/core: Fix endianness annotation in rdma_is_multicast_addr()

Since ipv4_addr is a big endian 32-bit number, annotate it as such.

Fixes: commit be1d325a3358 ("IB/core: Set RoCEv2 MGID according to spec")
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

Merge tag 'mlx5-updates-2017-10-06' of git://git./linux/kernel/git/mellanox/linux into k.o/for-next

Merge Mellanox shared pull request

Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/core: remove redundant check on prot_sg_cnt

prot_sg_cnt cannot be zero as a previous check on ret (from which
prot_sg_cnt is assigned) returns -ENOMEM if is it zero.  Since
it cannot be zero we can simplify the code by removing the non
-zero check on prot_sg_cnt and redundant else statement.

Detected by CoverityScan, COD#1357188 ("Logically dead code")

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/core: Simplify sa_path_set_[sd]lid() calls

Instead of making every caller convert the second argument of
sa_path_set_slid() and sa_path_set_dlid() to big endian format,
make these two functions accept LIDs in CPU endian format.
This patch does not change any functionality.

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com>
Cc: Don Hiatt <don.hiatt@intel.com>
Cc: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

infiniband: add MMU dependency for user_mem

The infiniband subsystem causes a link failure when the umem
driver is built on MMU-less systems:

mm/mmu_notifier.o: In function `do_mmu_notifier_register':
mmu_notifier.c:(.text+0x32): undefined reference to `mm_take_all_locks'
drivers/infiniband/core/umem.o: In function `ib_umem_get':
umem.c:(.text+0x132): undefined reference to `can_do_mlock'
drivers/infiniband/core/umem_odp.o: In function `ib_umem_odp_map_dma_pages':
umem_odp.c:(.text+0x766): undefined reference to `get_user_pages_remote'

This bug has existed for a while but only become apparent in ARM
randconfig builds when the dependency on PCI was lifted, as none
of the ARM-NOMMU targets support PCI at the moment.

We could probably get the umem driver to build by providing an
alternative implementation 'can_do_mlock()' that returns false
on NOMMU-systems, but then we'd still have a problem with the
mmu-notifiers required by CONFIG_INFINIBAND_ON_DEMAND_PAGING,
so simply forbidding umem with NOMMU seems like the simplest
workaround.

Fixes: 931bc0d91639 ("IB: Move PCI dependency from root KConfig to HW's KConfigs")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/rxe: put the pool on allocation failure

If the allocation of elem fails, it is not sufficient to simply check
for NULL and return.  We need to also put our reference on the pool or
else we will leave the pool with a permanent ref count and we will never
be able to free it.

Fixes: 4831ca9e4a8e ("IB/rxe: check for allocation failure on elem")
Suggested-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/rxe: check for allocation failure on elem

The allocation for elem may fail (especially because we're using
GFP_ATOMIC) so best to check for a null return.  This fixes a potential
null pointer dereference when assigning elem->pool.

Detected by CoverityScan CID#1357507 ("Dereference null return value")

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Do not warn on lid conversions for OPA

On OPA devices opa_local_smp_check will receive 32Bit LIDs when the LID
is Extended. In such cases, it is okay to lose the upper 16 bits of the
LID as this information is obtained elsewhere. Do not issue a warning
when calling ib_lid_cpu16() in this case by masking out the upper 16Bits.

[75920.148985] ------------[ cut here ]------------
[75920.154651] WARNING: CPU: 0 PID: 1718 at ./include/rdma/ib_verbs.h:3788 hfi1_process_mad+0x1c1f/0x1c80 [hfi1]
[75920.166192] Modules linked in: ib_ipoib hfi1(E) rdmavt(E) rdma_ucm(E) ib_ucm(E) rdma_cm(E) ib_cm(E) iw_cm(E) ib_umad(E) ib_uverbs(E) ib_core(E) libiscsi scsi_transport_iscsi dm_mirror dm_region_hash dm_log dm_mod dax x86_pkg_temp_thermal intel_powerclamp coretemp kvm irqbypass crct10dif_pclmul crc32_pclmul ghash_clmulni_intel pcbc aesni_intel mei_me ipmi_si iTCO_wdt iTCO_vendor_support crypto_simd ipmi_devintf pcspkr mei sg i2c_i801 glue_helper lpc_ich shpchp ioatdma mfd_core wmi ipmi_msghandler cryptd acpi_power_meter acpi_pad nfsd auth_rpcgss nfs_acl lockd grace sunrpc ip_tables xfs libcrc32c sd_mod mgag200 drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops ttm drm igb ptp ahci libahci pps_core crc32c_intel libata dca i2c_algo_bit i2c_core [last unloaded: ib_core]
[75920.246331] CPU: 0 PID: 1718 Comm: kworker/0:1H Tainted: G        W I E   4.13.0-rc7+ #1
[75920.255907] Hardware name: Intel Corporation S2600WT2/S2600WT2, BIOS SE5C610.86B.01.01.0008.021120151325 02/11/2015
[75920.268158] Workqueue: ib-comp-wq ib_cq_poll_work [ib_core]
[75920.274934] task: ffff88084a718000 task.stack: ffffc9000a424000
[75920.282123] RIP: 0010:hfi1_process_mad+0x1c1f/0x1c80 [hfi1]
[75920.288881] RSP: 0018:ffffc9000a427c38 EFLAGS: 00010206
[75920.295265] RAX: 0000000000010001 RBX: ffff8808361420e8 RCX: ffff880837811d80
[75920.303784] RDX: 0000000000000002 RSI: 0000000000007fff RDI: ffff880837811d80
[75920.312302] RBP: ffffc9000a427d38 R08: 0000000000000000 R09: ffff8808361420e8
[75920.320819] R10: ffff88083841f0e8 R11: ffffc9000a427da8 R12: 0000000000000001
[75920.329335] R13: ffff880837810000 R14: 0000000000000000 R15: ffff88084f1a4800
[75920.337849] FS:  0000000000000000(0000) GS:ffff88085f400000(0000) knlGS:0000000000000000
[75920.347450] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[75920.354405] CR2: 00007f9e4b3d9000 CR3: 0000000001c09000 CR4: 00000000001406f0
[75920.362947] Call Trace:
[75920.366257]  ? ib_mad_recv_done+0x258/0x9b0 [ib_core]
[75920.372457]  ? ib_mad_recv_done+0x258/0x9b0 [ib_core]
[75920.378652]  ? __kmalloc+0x1df/0x210
[75920.383229]  ib_mad_recv_done+0x305/0x9b0 [ib_core]
[75920.389270]  __ib_process_cq+0x5d/0xb0 [ib_core]
[75920.395032]  ib_cq_poll_work+0x20/0x60 [ib_core]
[75920.400777]  process_one_work+0x149/0x360
[75920.405836]  worker_thread+0x4d/0x3c0
[75920.410505]  kthread+0x109/0x140
[75920.414681]  ? rescuer_thread+0x380/0x380
[75920.419731]  ? kthread_park+0x60/0x60
[75920.424406]  ret_from_fork+0x25/0x30
[75920.428972] Code: 4c 89 9d 58 ff ff ff 49 89 45 00 66 b8 00 02 49 89 45 08 e8 44 27 89 e0 4c 8b 9d 58 ff ff ff e9 d8 f6 ff ff 0f ff e9 55 e7 ff ff <0f> ff e9 3b e5 ff ff 0f ff 0f 1f 84 00 00 00 00 00 e9 4b e9 ff
[75921.451269] ---[ end trace cf26df27c9597265 ]---

Fixes: 62ede7779904 ("Add OPA extended LID support")
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Don Hiatt <don.hiatt@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/core: Do not warn on lid conversions for OPA

On OPA devices the user_mad recv_handler can receive 32Bit LIDs
(e.g. OPA_PERMISSIVE_LID) and it is okay to lose the upper 16 bits
of the LID as this information is obtained elsewhere. Do not issue
a warning when calling ib_lid_be16() in this case by masking out
the upper 16Bits.

[75667.310846] ------------[ cut here ]------------
[75667.316447] WARNING: CPU: 0 PID: 1718 at ./include/rdma/ib_verbs.h:3799 recv_handler+0x15a/0x170 [ib_umad]
[75667.327640] Modules linked in: ib_ipoib hfi1(E) rdmavt(E) rdma_ucm(E) ib_ucm(E) rdma_cm(E) ib_cm(E) iw_cm(E) ib_umad(E) ib_uverbs(E) ib_core(E) libiscsi scsi_transport_iscsi dm_mirror dm_region_hash dm_log dm_mod dax x86_pkg_temp_thermal intel_powerclamp coretemp kvm irqbypass crct10dif_pclmul crc32_pclmul ghash_clmulni_intel pcbc aesni_intel mei_me ipmi_si iTCO_wdt iTCO_vendor_support crypto_simd ipmi_devintf pcspkr mei sg i2c_i801 glue_helper lpc_ich shpchp ioatdma mfd_core wmi ipmi_msghandler cryptd acpi_power_meter acpi_pad nfsd auth_rpcgss nfs_acl lockd grace sunrpc ip_tables xfs libcrc32c sd_mod mgag200 drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops ttm drm igb ptp ahci libahci pps_core crc32c_intel libata dca i2c_algo_bit i2c_core [last unloaded: ib_core]
[75667.407704] CPU: 0 PID: 1718 Comm: kworker/0:1H Tainted: G        W I E   4.13.0-rc7+ #1
[75667.417310] Hardware name: Intel Corporation S2600WT2/S2600WT2, BIOS SE5C610.86B.01.01.0008.021120151325 02/11/2015
[75667.429555] Workqueue: ib-comp-wq ib_cq_poll_work [ib_core]
[75667.436360] task: ffff88084a718000 task.stack: ffffc9000a424000
[75667.443549] RIP: 0010:recv_handler+0x15a/0x170 [ib_umad]
[75667.450090] RSP: 0018:ffffc9000a427ce8 EFLAGS: 00010286
[75667.456508] RAX: 00000000ffffffff RBX: ffff88085159ce80 RCX: 0000000000000000
[75667.465094] RDX: ffff88085a47b068 RSI: 0000000000000000 RDI: ffff88085159cf00
[75667.473668] RBP: ffffc9000a427d38 R08: 000000000001efc0 R09: ffff88085159ce80
[75667.482228] R10: ffff88085f007480 R11: ffff88084acf20e8 R12: ffff88085a47b020
[75667.490824] R13: ffff881056842e10 R14: ffff881056840200 R15: ffff88104c8d0800
[75667.499390] FS:  0000000000000000(0000) GS:ffff88085f400000(0000) knlGS:0000000000000000
[75667.509028] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[75667.516080] CR2: 00007f9e4b3d9000 CR3: 0000000001c09000 CR4: 00000000001406f0
[75667.524664] Call Trace:
[75667.528044]  ? find_mad_agent+0x7c/0x1b0 [ib_core]
[75667.534031]  ? ib_mark_mad_done+0x73/0xa0 [ib_core]
[75667.540142]  ib_mad_recv_done+0x423/0x9b0 [ib_core]
[75667.546215]  __ib_process_cq+0x5d/0xb0 [ib_core]
[75667.552007]  ib_cq_poll_work+0x20/0x60 [ib_core]
[75667.557766]  process_one_work+0x149/0x360
[75667.562844]  worker_thread+0x4d/0x3c0
[75667.567529]  kthread+0x109/0x140
[75667.571713]  ? rescuer_thread+0x380/0x380
[75667.576775]  ? kthread_park+0x60/0x60
[75667.581447]  ret_from_fork+0x25/0x30
[75667.586014] Code: 43 4a 0f b6 45 c6 88 43 4b 48 8b 45 b0 48 89 43 4c 48 8b 45 b8 48 89 43 54 8b 45 c0 0f c8 89 43 5c e9 79 ff ff ff e8 16 4e fa e0 <0f> ff e9 42 ff ff ff 66 66 66 66 66 66 2e 0f 1f 84 00 00 00 00
[75667.608323] ---[ end trace cf26df27c9597264 ]---

Fixes: 62ede7779904 ("Add OPA extended LID support")
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Don Hiatt <don.hiatt@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/rdmavt: Correct issues with read-mostly and send size cache lines

The s_ahgpsn was incorrectly placed in the read-mostly section of the QP
and the s_curr_size and s_hdrwords are oversized. The misplaced
s_ahgpsn will cause the read-mostly cachelines to thrash.

Place s_ahgpsn in the send side cache lines and correctly size and
s_hdrwords and s_cur_size to keep the send side cachelines at the same
size.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/core: Use __be32 for LIDs in opa_is_extended_lid

The LIDs passed to opa_extended_lid are in __be32 format,
change function signature accordingly.

This fixes the following sparse warnings:
  drivers/infiniband/core/cm.c:1181:60: warning: incorrect type in
argument 1 (different ba
  drivers/infiniband/core/cm.c:1182:60: warning: incorrect type in
argument 2 (different ba
  drivers/infiniband/core/cm.c:1242:68: warning: incorrect type in
argument 1 (different ba
  drivers/infiniband/core/cm.c:1243:68: warning: incorrect type in
argument 2 (different ba
  drivers/infiniband/core/cm.c:2922:66: warning: incorrect type in
argument 1 (different ba
  drivers/infiniband/core/cm.c:2923:66: warning: incorrect type in
argument 2 (different ba
  include/rdma/opa_addr.h:102:14: warning: cast to restricted __be32

Fixes: e92aa00a5189 ("IB/CM: Add OPA Path record support to CM")
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Don Hiatt <don.hiatt@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Prevent LNI out of sync by resetting host interface version

When the link is disabled and re-enabled, the host version bit is not
set again, so the firmware behaves as though it’s interacting with an
old driver. This causes LNI to get out of sync. The host version bit
needs to be set at load_8051_firmware() and _dc_start(). Currently, it's
only set at load_8051_firmware().

Create a common function to set the bit with the intent to make the code
more maintainable in the future, set the host version bit at _dc_start()
and modify the 8051 command API to prevent a deadlock as _dc_start() is
already holding the dc8051 lock.

Fixes: 913cc67159bc ("IB/hfi1: Always perform offline transition")
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Fix incorrect available receive user context count

The addition of the VNIC contexts to num_rcv_contexts changes the
meaning of the sysfs value nctxts from available user contexts, to
user contexts + reserved VNIC contexts.

User applications that use nctxts are now broken.

Update the calculation so that VNIC contexts are used only if there are
hardware contexts available, and do not silently affect nctxts.

Update code to use the calculated VNIC context number.

Update the sysfs value nctxts to be available user contexts only.

Fixes: 2280740f01ae ("IB/hfi1: Virtual Network Interface Controller (VNIC) HW support")
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Niranjana Vishwanathapura <Niranjana.Vishwanathapura@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Cc: <Stable@vger.kernel.org> #v4.12+
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Fix output trace issues from 16B change

The 16B changes to the output side of the header trace introduced
two issues:

1. An uninitialized field "l4" for 9B packets

   This field needs to be given a value of 0 for 9B
   packets to insure a correct 9B trace.

   The fix adds a new define to insure that there is a dummy
   default for 9B packets to insure the correct string
   is decoded.

2. Use of entry vs. __entry in field references

Fixes: Commit 863cf89d472f ("IB/hfi1: Add 16B trace support")
Reported-by: Kaike Wan <kaike.wan@intel.com>
Reviewed-by: Don Hiatt <don.hiatt@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Add parsing for platform configuration format version 4

Platform configuration format version 4, that didn't use the file
size field, is not parsed by the host driver. Only version 5 is
supported. Add logic in parsing procedure to determine what format
is being used and allow to read data from version 4 files.

Reviewed-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Andrzej Kacprowski <andrzej.kacprowski@intel.com>
Signed-off-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

i40iw: Do not allow posting WR after QP is flushed

A Work Request (WR) posted after QP is flushed will not
get a flush completion.

Correct this problem by not allowing posting of WRs
after a QP is flushed.

Fixes: d37498417947 ("i40iw: add files for iwarp interface")
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

i40iw: Do not generate CQE for RTR on QP flush

If RTR WQE is posted and QP is flushed, a CQE is
incorrectly generated for the RTR WQE. Add code
to look for the RTR and not generate a CQE when
QP is flushed.

Fixes: 280cfc4b74e6 ("i40iw: user kernel shared files")
Signed-off-by: Mustafa Ismail <mustafa.ismail@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

i40iw: Do not retransmit MPA request after it is ACKed

The ACK packets for an MPA request are ignored and
the MPA request is retransmitted if the MPA reply
is late or missing. Fix this by checking ack_rcvd
variable before retransmitting a packet.

Fixes: f27b4746f378 ("i40iw: add connection management code")
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Faisal Latif <faisal.latif@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: return 0 rather than return a garbage status value

For the case where hr_qp->state == IB_QPS_RESET, an uninitialized
value in ret is being returned by function hns_roce_v2_query_qp.
Fix this by setting ret to 0 for this specific return condition.

Detected by CoverityScan, CID#1457203 ("Unitialized scalar variable")

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Fix calltrace for sleeping in atomic

We replace usleep_range that was excessively long anyway with udelay
to avoid using usleep_range function in spin_lock_bh spin region,
thereby avoiding this calltrace:

BUG: scheduling while atomic: insmod/1428/0x00000002
Modules linked in: hns-roce-hw-v2(+) hns_roce rdma_ucm rdma_cm iw_cm ib_uverbs ib_cm ib_core
CPU: 0 PID: 1428 Comm: insmod Not tainted 4.12.0-rc1-00677-g252e8fd-dirty #43
Hardware name: (null) (DT)
Call trace:
[<ffff000008089d20>] dump_backtrace+0x0/0x274
[<ffff00000808a068>] show_stack+0x20/0x28
[<ffff00000844ea58>] dump_stack+0x94/0xb4
[<ffff0000080f975c>] __schedule_bug+0x68/0x84
[<ffff000008a988d4>] __schedule+0x5fc/0x70c
[<ffff000008a98a24>] schedule+0x40/0xa4
[<ffff000008a9c6f0>] schedule_hrtimeout_range_clock+0x98/0xfc
[<ffff000008a9c788>] schedule_hrtimeout_range+0x34/0x40
[<ffff000008a9c098>] usleep_range+0x6c/0x80
[<ffff000000b9ae68>] hns_roce_cmd_send+0xe4/0x264 [hns-roce-hw-v2]
[<ffff000000b9b748>] hns_roce_cmd_query_hw_info+0x40/0x60 [hns-roce-hw-v2]
[<ffff000000b9b790>] hns_roce_v2_profile+0x28/0x668 [hns-roce-hw-v2]
[<ffff000000b6b1f4>] hns_roce_init+0x6c/0x948 [hns-roce-hw-v2]

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Don't unregister a callback we didn't register

The driver doesn't actually register an inetaddr notifier function,
so there is no need to unregister it on shutdown.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Avoid NULL pointer exception

After the loop in hns_roce_v1_mr_free_work_fn function, it is possible that
all qps will have been freed (in which case ne will be 0).  If that
happens, then later in the function when we dereference hr_qp we will
get an exception.  Check ne is not 0 to make sure we actually have an
hr_qp left to work on.

This patch fixes the smatch error as below:
drivers/infiniband/hw/hns/hns_roce_hw_v1.c:1009 hns_roce_v1_mr_free_work_fn()
error: we previously assumed 'hr_qp' could be null

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Set rdma_ah_attr type for querying qp

When querying qp, It needs to return RoCE device ah_attr type
that may be specific to RoCE devices.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Only assign dest_qp if IB_QP_DEST_QPN bit is set

Only when the IB_QP_DEST_QPN flag of attr_mask is set is it valid to
assign the dest_qp_num into the dest_qp field of qp context.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Check return value of kzalloc

When lp_qp_work is NULL, we should return ENOMEM.  In order to do so,
we had to make some upper layer functions return a value instead
of being void type so we can propagate the error up the stack.

This patch fixes the smatch error as below:
drivers/infiniband/hw/hns/hns_roce_hw_v1.c:918 hns_roce_v1_recreate_lp_qp()
error: potential null dereference 'lp_qp_work'.  (kzalloc returns null)

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Refactor code for readability

Put the code for checking the send doorbell status into a
separate function and call it from check_qp_db_process_status
to improve indenting and readability.

It fixes the warning from static checker:
drivers/infiniband/hw/hns/hns_roce_hw_v1.c:3562 check_qp_db_process_status()
warn: inconsistent indenting.

Fixes: 5f110ac4bed8 ("IB/hns: Fix for checkpatch.pl comment style)
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Modify the value with rd&dest_rd of qp_attr

The value of max_rd_atomic and max_dest_rd_atomic in query_qp
are incorrect. It should be assigned by left shifting of
the bit in hip06 SoC.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

iw_cxgb4: add referencing to wait objects

For messages sent from the host to fw that solicit a reply from fw,
the c4iw_wr_wait struct pointer is passed in the host->fw message, and
included in the fw->host fw6_msg reply.  This allows the sender to wait
until the reply is received, and the code processing the ingress reply
to wake up the sender.

If c4iw_wait_for_reply() times out, however, we need to keep the
c4iw_wr_wait object around in case the reply eventually does arrive.
Otherwise we have touch-after-free bugs in the wake_up paths.

This was hit due to a bad kernel driver that blocked ingress processing
of cxgb4 for a long time, causing iw_cxgb4 timeouts, but eventually
resuming ingress processing and thus hitting the touch-after-free bug.

So I want to fix iw_cxgb4 such that we'll at least keep the wait object
around until the reply comes.  If it never comes we leak a small amount of
memory, but if it does come late, we won't potentially crash the system.

So add a kref struct in the c4iw_wr_wait struct, and take a reference
before sending a message to FW that will generate a FW6 reply.  And remove
the reference (and potentially free the wait object) when the reply
is processed.

The ep code also uses the wr_wait for non FW6 CPL messages and doesn't
embed the c4iw_wr_wait object in the message sent to firmware.  So for
those cases we add c4iw_wake_up_noref().

The mr/mw, cq, and qp object create/destroy paths do need this reference
logic.  For these paths, c4iw_ref_send_wait() is introduced to take the
wr_wait reference, send the msg to fw, and then wait for the reply.

So going forward, iw_cxgb4 either uses c4iw_ofld_send(),
c4iw_wait_for_reply() and c4iw_wake_up_noref() like is done in the some
of the endpoint logic, or c4iw_ref_send_wait() and c4iw_wake_up_deref()
(formerly c4iw_wake_up()) when sending messages with the c4iw_wr_wait
object pointer embedded in the message and resulting FW6 reply.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

iw_cxgb4: allocate wait object for each ep object

Remove the embedded c4iw_wr_wait object in preparation for correctly
handling timeouts.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

iw_cxgb4: allocate wait object for each qp object

Remove the local stack allocated c4iw_wr_wait object in preparation for
correctly handling timeouts.

Also cleaned up some error path unwind logic to make it more readable.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

iw_cxgb4: allocate wait object for each cq object

Remove the local stack allocated c4iw_wr_wait object in preparation for
correctly handling timeouts.

Also cleaned up some error path unwind logic to make it more readable.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

iw_cxgb4: allocate wait object for each memory object

Remove the local stack allocated c4iw_wr_wait object in preparation for
correctly handling timeouts.

Also refactored some code to simplify it and make errpath unwinding
more readable.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: remove redundant assignment to variable j

Variable j is being assigned to loop_j and then later being
assigned to a new value in for loops. The first initialization
is therefore redundant and can be removed.

Cleans up clang warning:
warning: Value stored to 'j' is never read

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: make various function static, fixes warnings

The functions hns_roce_table_mhop_get, hns_roce_table_mhop_put,
hns_roce_cleanup_mhop_hem_table,  hns_roce_v1_post_mbox,
hns_roce_cmq_setup_basic_desc, hns_roce_cmq_send,
hns_roce_cmq_query_hw_info are all local to the source and do
not need to be in global scope, so make them static.

Cleans up sparse warnings:
symbol 'hns_roce_table_mhop_get' was not declared. Should it be static?
symbol 'hns_roce_table_mhop_put' was not declared. Should it be static?
symbol 'hns_roce_cleanup_mhop_hem_table' was not declared. Should it be
static?
symbol 'hns_roce_v1_post_mbox' was not declared. Should it be static?
symbol 'hns_roce_cmq_setup_basic_desc' was not declared. Should it be
static?
symbol 'hns_roce_cmq_send' was not declared. Should it be static?
symbol 'hns_roce_cmq_query_hw_info' was not declared. Should it be static?

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

i40iw: delete some stray tabs

These lines were indented too far by mistake.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Acked-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/{ipoib, iser}: Consistent print format of vendor error

Vendor error print should be consistent across protocols to avoid any
confusion.
This patch corrects that.

Suggested-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Reviewed-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Acked-by: Wengang Wang <wen.gang.wang@oracle.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/qib: Use setup_timer and mod_timer

Use setup_timer and mod_timer API instead of structure assignments.

This is done using Coccinelle and semantic patch used
for this as follows:

@@
expression x,y,z,a,b;
@@

-init_timer (&x);
+setup_timer (&x, y, z);
+mod_timer (&a, b);
-x.function = y;
-x.data = z;
-x.expires = b;
-add_timer(&a);

Signed-off-by: Himanshu Jha <himanshujha199640@gmail.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/qedr: Fix rdma_type initialization

Initialize the rdma_type (iWARP or RoCE) which is set according to
device configuration in qed.

Fixes: e6a38c54faf ("RDMA/qedr: Add support for registering an iWARP device")
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

Merge branch 'vnic' into k.o/for-next

Signed-off-by: Doug Ledford <dledford@redhat.com>

Merge branch 'hfi1' into k.o/for-next

Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor reset_ctxt() IOCTL

The IOCTL is a bit unwieldy.  Refactor reset_ctxt() to be a bit more
manageable.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor get_user() IOCTLs

The IOCTL is a bit unwieldy.  Refactor to a common pattern.
Refactor _RECV_CTRL, _POLL_TYPE, _ACK_EVENT and _SET_PKEY
IOCTLs to a common pattern.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor hfi_user_exp_rcv_invalid() IOCTLs

The IOCTL is a bit unwieldy.  Refactor to a common pattern.
Refactor _TID_INVAL_READ IOCTLs.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor hfi_user_exp_rcv_clear() IOCTLs

The IOCTL is a bit unwieldy.  Refactor to a common pattern.
Refactor the _TID_FREE IOCTL.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor hfi_user_exp_rcv_setup() IOCTL

The IOCTL is a bit unwieldy.  Refactor to a common pattern.
Refactor the _TID_UPDATE IOCTL.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor get_base_info

The IOCTL is a bit unwieldy.  Refactor to a common pattern.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Fix parenthesis alignment issues

In preparation to refactoring get_base_info(), cleanup some
checkpatch issues.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor get_ctxt_info

The IOCTL is a bit unwieldy.  Refactor to a common pattern.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Refactor assign_ctxt() IOCTL

The IOCTL is a bit unwieldy.  Refactor to a common pattern.
Refactor the assign_ctxt() IOCTL.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Add a safe wrapper for _rcd_get_by_index

hfi1_rcd_get_by_index assumes that the given index is in the correct
range.  In most cases this is correct because the index is bounded by
a loop.  For these cases, adding a range check to the function is
redundant.

For the use case that is not bounded by the loop range, a _safe wrapper
function is needed to validate the index before accessing the rcd array.

Add a _safe wrapper to _get_by_index to validate the index range.

Update appropriate call sites with the new _safe function.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Extend input hdr tracing for packet type

The etype field from the rhf can have more values than just
IB and BYPASS.

Extend the current tracing to report a symbolic for the etype
field for non-bypass packets.  Bypass packets will continue to
report the l2.

As part of this fix the etype and the l2 are added to the tracing
struct and are available for trigger and filter operations.

Fixes: Commit 863cf89d472f ("IB/hfi1: Add 16B trace support")
Reviewed-by: Don Hiatt <don.hiatt@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Remove unused hfi1_cpulist variables

Following variables: hfi1_cpulist and hfi1_cpulist_count
are unused. Remove them.

Reviewed-by: Harish Chegondi <harish.chegondi@intel.com>
Reviewed-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Inline common calculation

Calculating the offset to a context is done several times throughout
the code.  Create a common inlined function for doing this
calculation.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Remove unnecessary error messages on alloc failures

Per-cpu variables int_counter, rcv_limit, and send_schedule
print unnecessary error messages on failed allocations.
Remove the error messages.

Reviewed-by: Harish Chegondi <harish.chegondi@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Remove unused link_default variable

devdata->link_default is no longer variable

Maintain number of holes by moving dc_shutdown

Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Set default_desc1 just one time

There is no reason to set the default descriptor flag on every SDMA
engine initialization.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Update HFI to use the latest PCI API

The HFI PCI IRQ code uses an obsolete PCI API.  Update the code to use
the new PCI IRQ API and any necessary changes because of the new API.

Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/qib: Update QIB to use the latest PCI API

The QIB PCI IRQ code uses an obsolete PCI API.  Updating the code to use
the new PCI IRQ API and any necessary changes because of the new API.

Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Remove the debug trace message in pin_sdma_pages()

Remove the debug trace statement in pin_sdma_pages() that
gets executed when there is a memory allocation failure as
the trace message doesn't help with debugging the memory
allocation failure.

Cc: Leon Romanovsky <leon@kernel.org>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Convert the macro AHG_HEADER_SET into an inline function

AHG_HEADER_SET macro doesn't conform to the coding standards as it can
affect the control flow. Convert the macro AHG_HEADER_SET into an inline
function ahg_header_set().

Cc: Leon Romanovsky <leon@kernel.org>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/hfi1: Add new state complete decodes for LNI failures

Add state decodes for link width negotiation, verify cap time out
and secure data resolution failures.

Reviewed-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/opa_vnic: Add routing control information

Add protocol specific routing control information in the encapsulation
header as per the configuration.

Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Scott Franco <safranco@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/opa_vnic: Properly set vesw port status

Update eth_link_status and operating status information to
represent the overall status of the virtual Ethernet switch port.

Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/opa_vnic: Properly clear Mac Table Digest

Clear the MAC table digest when the MAC table is freed.

Reviewed-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Scott Franco <safranco@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/opa_vnic: Properly return the total MACs in UC MAC list

Do not include EM specified MAC address in total MACs of the
UC MAC list.

Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/opa_vnic: Allow reset of MAC address

Ensure MAC address is reset while deleting the vesw port.

Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/opa_vnic: Set POD value for Ethernet MTU

Set power on default value of 1500 for eth_mtu.

Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/opa_vnic: Mark unused Ethernet MTU fields as reserved

Per pcp mtu fields are not used, mark them as reserved.

Reviewed-by: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

Merge branches 'hns' and 'misc' into k.o/for-next

Signed-off-by: Doug Ledford <dledford@redhat.com>

iw_cxgb4: change pr_debug to appropriate log level

Error logs of iw_cxgb4 needs to be printed by default. This patch
changes the necessary pr_debug() to appropriate pr_<log level>.

Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

iw_cxgb4: Remove __func__ parameter from pr_debug()

pr_debug() can be enabled to print function names, So removing the
unwanted __func__ parameters from debug logs.
Realign function parameters.

Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/ocrdma: pr_err() strings should end with newlines

pr_err() messages should end with a new-line to avoid other messages
being concatenated.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/mlx5:: pr_err() and mlx5_ib_dbg() strings should end with newlines

pr_err() and mlx5_ib_dbg( messages should terminated with a new-line to
avoid other messages being concatenated.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/ipoib: Remove device when one port fails to init

Call ipoib_remove_one when one of the IPoIB ports fails to initialize in
order not to leave the module in unstable state.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/cxgb3: remove redundant first assignement of sqp

sqp is being initialised when it is being declared and then updated
a little later on making the first initialization redundant. Clean
this up by initializing ptr and sqp at their declaration.

Cleans up warning: "warning: Value stored to 'sqp' during its
initialization is never read"

Fixes: a58e58fafdff ("RDMA/cxgb3: Wrap the software send queue pointer as needed on flush")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/uverbs: clean up INIT_UDATA() macro usage

After changing INIT_UDATA_BUF_OR_NULL() to an inline function,
this does the same change to INIT_UDATA for consistency.
I'm keeping it separate as this part is much larger and
we wouldn't want to backport this to stable kernels if we
ever want to address the gcc warnings by backporting the
first patch.

Again, using an inline function gives us better type
safety here among other issues with macros. I'm using
u64_to_user_ptr() to convert the user pointer to simplify
the logic rather than adding lots of new type casts.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/uverbs: clean up INIT_UDATA_BUF_OR_NULL usage

We get a harmless warning about the fact that we use the result of a
multiplication as a condition:

drivers/infiniband/core/uverbs_main.c: In function 'ib_uverbs_write':
drivers/infiniband/core/uverbs_main.c:787:40: error: '*' in boolean context, suggest '&&' instead [-Werror=int-in-bool-context]
drivers/infiniband/core/uverbs_main.c:787:117: error: '*' in boolean context, suggest '&&' instead [-Werror=int-in-bool-context]
drivers/infiniband/core/uverbs_main.c:790:50: error: '*' in boolean context, suggest '&&' instead [-Werror=int-in-bool-context]
drivers/infiniband/core/uverbs_main.c:790:151: error: '*' in boolean context, suggest '&&' instead [-Werror=int-in-bool-context]

This avoids the problem by using an inline function in place of
the macro.

Fixes: a96e4e2ffe43 ("IB/uverbs: New macro to set pointers to NULL if length is 0 in INIT_UDATA()")
Suggested-by: Christoph Hellwig <hch@infradead.org>
Link: https://patchwork.kernel.org/patch/9940777/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/{cxgb3,cxgb4}: Remove unneeded config dependencies

CHELSIO_T3 already depend on INET
CHELSIO_T4 already depend on (IPV6 || IPV6=n)

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB: Move PCI dependency from root KConfig to HW's KConfigs

No reason to have dependency on PCI for the entire infiniband stack so
move it to KConfig of only the drivers that actually using PCI.

Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

IB/core: fix spelling mistake: "aceess" -> "access"

Trivial fix to spelling mistake in WARN message

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Delete the unnecessary initializing enum to zero

It deletes the unnecessary initializing enum to zero.

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Fix inconsistent warning

This patch fixes smatch inconsistent warning as below:
drivers/infiniband/hw/hns/hns_roce_hem.h:136 hns_roce_hem_first() warn: inconsistent indenting

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Replace condition statement using hardware version information

This patch replaces condition statement to reduce usage of hardware version
information in common driver.

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Add releasing resource operation in error branch

This patch adds releasing resource operation in error branch
of the function named hns_roce_table_get.

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Configure the MTPT in hip08

The MTPT records the attribute of the registered MR. The MTPT format
will be updated in hip08, and the MTPT should be configured.

This patch is to configure the MTPT for the registered MR in hip08.

Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Add support for processing send wr and receive wr

This patch is implementing for posting send request and
receiving request for hip08 RoCE driver. such as
post send verbs and post recv verbs.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Add QP operations support for hip08 SoC

This patch implements QP operations for hip08 RoCE driver and
fixes some checkpatch warning about print message in QP function.
The QP operations includes create QP, query QP, modify QP and
destroy QP.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Add CQ operations support for hip08 RoCE driver

This patch adds CQ relevant operations for hip08 RoCE driver,
such as create CQ, destroy CQ, poll CQ and Request Completion
Notification(req_notify_cq).

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Configure mac&gid and user access region for hip08 RoCE driver

In hip08, the user access region(UAR) pfn is calculated
from pci device memory resource.

This patch mainly sets mac and gid table by configuring
the relevant registers and updates the uar pfn for hip08 SoC.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Support multi hop addressing for PBL in hip08

The block base address in the MR can be retrieved by the block number
which is calculated with the VA in the SGE and MTPT. In hip08, the PBL
supports multi hop addressing to retrieve the block base address by
the block number.

This patch is to add the interfaces in the MR to support multi hop
addressing for the PBL.

Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Split CQE from MTT in hip08

In hip08, the SQWQE/SGE/RQWQE and CQE have different hop num and
page size, so we need to manage the base address table of the
SQWQE/SGE/RQWQE and CQE separately.

This patch is to split CQE from MTT(SQWQE/SGE/RQWQE).

Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Update the interfaces for MTT/CQE multi hop addressing in hip08

The MTT(SQWQE/SGE/RQWQE) and CQE in hip08 can support multi hop
addressing. The address of MTT/CQE can be retrieved by the BT
(Base Address Table) with multi hop addressing.

This patch is to update the interfaces in HEM to support multi hop
addressing for the MTT/CQE.

Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Configure BT BA and BT attribute for the contexts in hip08

BT is used to retrieve the addresses of the contexts(QPC/MPT/CQC/SRQC)
in memory. In order to support multi hop addressing for the contexts,
the BT BA should be configured by mailbox, and the BT attribution will
be set by command.

This patch is to configure the BT BA and BT attribution for the
contexts.

Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Add the interfaces to support multi hop addressing for the contexts in hip08

The contexts (QPC/MTPT/CQC/SRQC) in hip08 can support multi hop
addressing. The address of context can be retrieved by the
BT (Base Address Table) with multi hop addressing. The first hop
BT BA can be retrieved from the RAM in the chip by the bt_idx and
bt_num.

This patch is to add the interfaces in HEM to support multi hop
addressing for the contexts.

Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Add mailbox's implementation for hip08 RoCE driver

In hip08 SoC, the hardware implementation of mailbox command
has changed with hip06 SoC. As a result, it adjusts the
architecture of the command code and implements the interfaces
of mailbox for hip08 SoC.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

RDMA/hns: Add profile support for hip08 driver

The profile's content mainly set some specifications and obtain
some hardware resources by implementing the relative commands.
Because max sge num of send queue is not the same with receive
queue in hip08, we modified the calculation of props->max_sge
in query_device ops.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>