Current code uses many different types when dealing with a port of a RDMA
device: u8, unsigned int and u32. Switch to u32 to clean up the logic.

This allows us to make (at least) the core view consistent and use the
same type. Unfortunately not all places can be converted. Many uverbs
functions expect port to be u8 so keep those places in order not to break
UAPIs.  HW/Spec defined values must also not be changed.

With the switch to u32 we now can support devices with more than 255
ports. U32_MAX is reserved to make control logic a bit easier to deal
with. As a device with U32_MAX ports probably isn't going to happen any
time soon this seems like a non issue.

When a device with more than 255 ports is created uverbs will report the
RDMA device as having 255 ports as this is the max currently supported.

The verbs interface is not changed yet because the IBTA spec limits the
port size in too many places to be u8 and all applications that relies in
verbs won't be able to cope with this change. At this stage, we are
extending the interfaces that are using vendor channel solely

Once the limitation is lifted mlx5 in switchdev mode will be able to have
thousands of SFs created by the device. As the only instance of an RDMA
device that reports more than 255 ports will be a representor device and
it exposes itself as a RAW Ethernet only device CM/MAD/IPoIB and other
ULPs aren't effected by this change and their sysfs/interfaces that are
exposes to userspace can remain unchanged.

While here cleanup some alignment issues and remove unneeded sanity
checks (mainly in rdmavt),

Link: https://lore.kernel.org/r/20210301070420.439400-1-leon@kernel.orgSigned-off-by: NMark Bloch <mbloch@nvidia.com>
Signed-off-by: NLeon Romanovsky <leonro@nvidia.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The HIP09 supports XRC transport service, it greatly saves the number of
QPs required to connect all processes in a large cluster.

Link: https://lore.kernel.org/r/1614826558-35423-1-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

hns_roce_device.h is not specific to hardware, some definitions are only
used for HIP06, they should be moved into hns_roce_hw_v1.h.

Link: https://lore.kernel.org/r/1612517974-31867-9-git-send-email-liweihang@huawei.comSigned-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Currently, the driver updates doorbell looks like this:

post()
{
	wqe.field = 0x111;
	wmb();
	update_wq_db();
}

update_wq_db()
{
	db.field = 0x222;
	__raw_writeq(db, db_reg);
}

writeq() is a better choice than __raw_writeq() because it calls dma_wmb()
to barrier in ARM64, and dma_wmb() is better than wmb() for ROCEE device.

This patch removes all wmb() before updating doorbell of SQ/RQ/CQ/SRQ by
replacing __raw_writeq() with writeq() to improve performence.  The new
process looks like this:

post()
{
	wqe.field = 0x111;
	update_wq_db();
}

update_wq_db()
{
	db.field = 0x222;
	writeq(db, db_reg);
}

Link: https://lore.kernel.org/r/1612517974-31867-8-git-send-email-liweihang@huawei.comSigned-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

This bit should be in type of enum ib_sig_type, or there will be a sparse
warning.

Fixes: bfe86035 ("RDMA/hns: Fix cast from or to restricted __le32 for driver")
Link: https://lore.kernel.org/r/1612517974-31867-3-git-send-email-liweihang@huawei.comReported-by: Nkernel test robot <lkp@intel.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Direct wqe is a mechanism to fill wqe directly into the hardware. In the
case of light load, the wqe will be filled into pcie bar space of the
hardware, this will reduce one memory access operation and therefore
reduce the latency.

Link: https://lore.kernel.org/r/1611997513-27107-1-git-send-email-liweihang@huawei.comSigned-off-by: NYixing Liu <liuyixing1@huawei.com>
Signed-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Reduce parameter numbers of write_srqc() and move some related code into
it from alloc_srqc().

Link: https://lore.kernel.org/r/1611997090-48820-8-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Split the SRQ creation process into multiple steps and encapsulate them
into functions.

Link: https://lore.kernel.org/r/1611997090-48820-7-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Each SRQs contain an reserved WQE, it is inappropriate and should be
removed.

Fixes: c7bcb134 ("RDMA/hns: Add SRQ support for hip08 kernel mode")
Link: https://lore.kernel.org/r/1611997090-48820-6-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

If a user posts WR by wr_list, the head pointer of idx_queue won't be
updated until all wqes are filled, so the judgment of whether head equals
to tail will get a wrong result. Fix above issue and move the head and
tail pointer from the srq structure into the idx_queue structure. After
idx_queue is filled with wqe idx, the head pointer of it will increase.

Fixes: c7bcb134 ("RDMA/hns: Add SRQ support for hip08 kernel mode")
Link: https://lore.kernel.org/r/1611997090-48820-3-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The RQ/SRQ of HIP08 needs one special sge to stop receive reliably. So the
driver needs to allocate at least one SGE when creating RQ/SRQ and ensure
that at least one SGE is filled with the special value during post_recv.

Besides, the kernel driver should only do this for kernel ULP. For
userspace ULP, the userspace driver will allocate the reserved SGE in
buffer, and the kernel driver just needs to pin the corresponding size of
memory based on the userspace driver's requirements.

Link: https://lore.kernel.org/r/1611997090-48820-2-git-send-email-liweihang@huawei.comSigned-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

When creating or re-registering an MR, storing the PDN, access flag and
IOVA information ASAP can simplify the number of parameters passed into
the subsequent process.

Link: https://lore.kernel.org/r/1611395282-991-3-git-send-email-liweihang@huawei.comReported-by: Nkernel test robot <lkp@intel.com>
Signed-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Split the hns_roce_mtr_create() into serval small functions, remove unused
member in 'struct hns_roce_buf_attr' and delete unnecessary MTR page count
check flow to make the MTR creation related codes clearer.

Link: https://lore.kernel.org/r/1611395282-991-2-git-send-email-liweihang@huawei.comSigned-off-by: NXi Wang <wangxi11@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

GFP_KERNEL may cause ida_alloc_range() to sleep, but the spinlock covering
this function is not allowed to sleep, so the spinlock needs to be changed
to mutex.

As there is a certain chance of memory allocation failure, GFP_ATOMIC is
not suitable for QP allocation scenarios.

Fixes: 71586dd2 ("RDMA/hns: Create QP with selected QPN for bank load balance")
Link: https://lore.kernel.org/r/1611048513-28663-1-git-send-email-liweihang@huawei.comSigned-off-by: NYangyang Li <liyangyang20@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

In order to improve performance by balancing the load between different
banks of cache, the CQC cache is desigend to choose one of 4 banks
according to lower 2 bits of CQN. The hns driver needs to count the number
of CQ on each bank and then assigns the CQ being created to the bank with
the minimum load first.

Link: https://lore.kernel.org/r/1610008589-35770-1-git-send-email-liweihang@huawei.comSigned-off-by: NYangyang Li <liyangyang20@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

There is no need to get queue number repeatly for different queues from an
AEQE entity, as they are the same. Furthermore, redefine the AEQE
structure to make the codes more readable.

In addition, HNS_ROCE_EVENT_TYPE_CEQ_OVERFLOW is removed because the
hardware never reports this event.

Link: https://lore.kernel.org/r/1607650657-35992-12-git-send-email-liweihang@huawei.comSigned-off-by: NYixian Liu <liuyixian@huawei.com>
Signed-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Types of some fields, variables and parameters of some functions should be
unsigned.

Link: https://lore.kernel.org/r/1607650657-35992-10-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The high 6 bits of traffic class in GRH is DSCP (Differentiated Services
Codepoint), the driver should shift it before the hardware gets it when
using RoCEv2.

Fixes: 606bf89e ("RDMA/hns: Refactor for hns_roce_v2_modify_qp function")
Fixes: fba429fc ("RDMA/hns: Fix missing fields in address vector")
Link: https://lore.kernel.org/r/1607650657-35992-4-git-send-email-liweihang@huawei.comSigned-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

These flags will be returned to the userspace through ABI, so they should
be defined in hns-abi.h. Furthermore, there is no need to include
hns-abi.h in every source files, it just needs to be included in the
common header file.

Link: https://lore.kernel.org/r/1606872560-17823-1-git-send-email-liweihang@huawei.comReported-by: Nkernel test robot <lkp@intel.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

mlx5 has an ugly flow where it tries to allocate a new MR and replace the
existing MR in the same memory during rereg. This is very complicated and
buggy. Instead of trying to replace in-place inside the driver, provide
support from uverbs to change the entire HW object assigned to a handle
during rereg_mr.

Since destroying a MR is allowed to fail (ie if a MW is pointing at it)
and can't be detected in advance, the algorithm creates a completely new
uobject to hold the new MR and swaps the IDR entries of the two objects.

The old MR in the temporary IDR entry is destroyed, and if it fails
rereg_mr succeeds and destruction is deferred to FD release. This
complexity is why this cannot live in a driver safely.

Link: https://lore.kernel.org/r/20201130075839.278575-4-leon@kernel.orgSigned-off-by: NLeon Romanovsky <leonro@nvidia.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Stash is a mechanism that uses the core information carried by the ARM AXI
bus to access the L3 cache. It can be used to improve the performance by
increasing the hit ratio of L3 cache. CQs need to enable stash by default.

Link: https://lore.kernel.org/r/1606374251-21512-2-git-send-email-liweihang@huawei.comSigned-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

In order to improve performance by balancing the load between different
banks of cache, the QPC cache is desigend to choose one of 8 banks
according to lower 3 bits of QPN. The hns driver needs to count the number
of QP on each bank and then assigns the QP being created to the bank with
the minimum load first.

Link: https://lore.kernel.org/r/1606220649-1465-1-git-send-email-liweihang@huawei.comSigned-off-by: NYangyang Li <liyangyang20@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Information about vlan is stored in GMV(GID/MAC/VLAN) table for HIP09, so
there is no need to copy it to address vector.

Link: https://lore.kernel.org/r/1605526408-6936-2-git-send-email-liweihang@huawei.comSigned-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Add a group of flags to control the 'struct hns_roce_buf' allocation
flow, this is used to support the caller running in atomic context.

Link: https://lore.kernel.org/r/1605347916-15964-1-git-send-email-liweihang@huawei.comSigned-off-by: NXi Wang <wangxi11@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

HIP09 supports to store SGID/SMAC/VLAN together in a table named GMV. The
driver needs to allocate memory for it and tell the information about this
region to hardware.

Link: https://lore.kernel.org/r/1603508836-33054-2-git-send-email-liweihang@huawei.comSigned-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The doorbell needs to store PI information into QPC, so the RoCEE should
wait for the results of storing, that is, it needs two bus operations to
complete a doorbell. When ROCEE is in SDI mode, multiple doorbells may be
interlocked because the RoCEE can only handle bus operations serially. So a
flag to mark if HIP09 is working in SDI mode is added. When the SDI flag is
set, the ROCEE will ignore the PI information of the doorbell, continue to
fetch wqe and verify its validity by it's owner_bit.

Link: https://lore.kernel.org/r/1603195493-22741-1-git-send-email-liweihang@huawei.comSigned-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Some code was removed but the variables were still there, and some
parameters have been changed to be queried from firmware. So the
definitions of them are no longer needed.

Fixes: 2a3d923f ("RDMA/hns: Replace magic numbers with #defines")
Fixes: 82e620d9 ("RDMA/hns: Modify the data structure of hns_roce_av")
Fixes: 82547469 ("IB/hns: Implement the add_gid/del_gid and optimize the GIDs management")
Fixes: 21b97f53 ("RDMA/hns: Fixup qp release bug")
Link: https://lore.kernel.org/r/1601371934-40003-1-git-send-email-liweihang@huawei.comSigned-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

HIP08 supports RC inline up to size of 32 Bytes, and all data should be
put into SQWQE. For HIP09, this capability is extended to 1024 Bytes, if
length of data is longer than 32 Bytes, they will be filled into extended
sge space.

Link: https://lore.kernel.org/r/1599744069-9968-1-git-send-email-liweihang@huawei.comSigned-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

For HIP09, size of SCCC (Soft Congestion Control Context) is increased to
64 Bytes from 32 Bytes. The hardware will get the configuration of SCCC
from driver instead of using a fixed value.

Link: https://lore.kernel.org/r/1600245806-56321-5-git-send-email-liweihang@huawei.comSigned-off-by: NYangyang Li <liyangyang20@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The new version of RoCEE supports using QPC in size of 256B or 512B, so
that HIP09 can supports new congestion control algorithms by using QPC in
larger size.

Link: https://lore.kernel.org/r/1600245806-56321-4-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NYangyang Li <liyangyang20@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The new version of RoCEE supports using CQE in size of 32B or 64B. The
performance of bus can be improved by using larger size of CQE.

Link: https://lore.kernel.org/r/1600245806-56321-3-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The new version of RoCEE supports using CEQE in size of 4B or 64B, AEQE in
size of 16B or 64B. The performance of bus can be improved by using larger
size of EQE.

Link: https://lore.kernel.org/r/1600245806-56321-2-git-send-email-liweihang@huawei.comSigned-off-by: NWenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Move allocation and destruction of memory windows under ib_core
responsibility and clean drivers to ensure that no updates to MW
ib_core structures are done in driver layer.

Link: https://lore.kernel.org/r/20200902081623.746359-2-leon@kernel.orgSigned-off-by: NLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Like any other verbs objects, CQ shouldn't fail during destroy, but
mlx5_ib didn't follow this contract with mixed IB verbs objects with
DEVX. Such mix causes to the situation where FW and kernel are fully
interdependent on the reference counting of each side.

Kernel verbs and drivers that don't have DEVX flows shouldn't fail.

Fixes: e39afe3d ("RDMA: Convert CQ allocations to be under core responsibility")
Link: https://lore.kernel.org/r/20200907120921.476363-7-leon@kernel.orgSigned-off-by: NLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

In similar way to other IB objects, restore the ability to return error on
SRQ destroy. Strictly speaking, this change is not necessary, and provided
here to ensure a symmetrical interface like other destroy functions.

Fixes: 68e326de ("RDMA: Handle SRQ allocations by IB/core")
Link: https://lore.kernel.org/r/20200907120921.476363-5-leon@kernel.orgSigned-off-by: NLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

Like any other IB verbs objects, AH are refcounted by ib_core. The release
of those objects are controlled by ib_core with promise that AH destroy
can't fail.

Being SW object for now, this change makes dealloc_ah() to behave like any
other destroy IB flows.

Fixes: d3456914 ("RDMA: Handle AH allocations by IB/core")
Link: https://lore.kernel.org/r/20200907120921.476363-3-leon@kernel.orgSigned-off-by: NLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The IB verbs objects are counted by the kernel and ib_core ensures that
deallocate PD will success so it will be called once all other objects
that depends on PD will be released. This is achieved by managing various
reference counters on such objects.

The mlx5 driver didn't follow this standard flow when allowed DEVX objects
that are not managed by ib_core to be interleaved with the ones under
ib_core responsibility.

In such interleaved scenarios deallocate command can fail and ib_core will
leave uobject in internal DB and attempt to clean it later to free
resources anyway.

This change partially restores returned value from dealloc_pd() for all
drivers, but keeping in mind that non-DEVX devices and kernel verbs paths
shouldn't fail.

Fixes: 21a428a0 ("RDMA: Handle PD allocations by IB/core")
Link: https://lore.kernel.org/r/20200907120921.476363-2-leon@kernel.orgSigned-off-by: NLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

The UDP source port number in RoCE v2 is used to create entropy for
network routers (ECMP), load balancers and 802.3ad link aggregation
switching that are not aware of RoCE IB headers. Considering that the IB
core has achieved a new interface to get a hashed value of it, the fixed
value of it in QPC and UD WQE in hns driver could be fixed and the port
number is to be set dynamically now.

For QPC of RC, the value could be hashed from flow_lable if the user pass
it in or from remote qpn and local qpn. For WQE of UD, it is set according
to fl or as a random value.

Link: https://lore.kernel.org/r/1598002289-8611-1-git-send-email-liweihang@huawei.comSigned-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

This patch caused some issues on SEND operation, and it should be reverted
to make the drivers work correctly. There will be a better solution that
has been tested carefully to solve the original problem.

This reverts commit 711195e5.

Fixes: 711195e5 ("RDMA/hns: Reserve one sge in order to avoid local length error")
Link: https://lore.kernel.org/r/1597829984-20223-1-git-send-email-liweihang@huawei.comSigned-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>

HIP08_A is an temporary version and all features of it are supported by
HIP08_B. So remove the relevant code.

Link: https://lore.kernel.org/r/1595932941-40613-4-git-send-email-liweihang@huawei.comSigned-off-by: NLang Cheng <chenglang@huawei.com>
Signed-off-by: NYangyang Li <liyangyang20@huawei.com>
Signed-off-by: NWeihang Li <liweihang@huawei.com>
Signed-off-by: NJason Gunthorpe <jgg@nvidia.com>