RDMA/uverbs: Add new relaxed ordering memory region access flag

Add a new relaxed ordering access flag for memory regions.  Using memory
regions with relaxed ordeing set can enhance performance.

This access flag is handled in a best-effort manner, drivers should ignore
if they don't support setting relaxed ordering.

Link: https://lore.kernel.org/r/1578506740-22188-9-git-send-email-yishaih@mellanox.com
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/efa: Allow passing of optional access flags for MR registration

As part of adding a range of optional access flags that drivers need to be
able to accept, mask this range inside efa driver.  This will prevent the
driver from failing when an access flag from that range is passed.

Link: https://lore.kernel.org/r/1578506740-22188-8-git-send-email-yishaih@mellanox.com
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Add optional access flags range

Define a range of access flags that are defined to be optional, both
uverbs and drivers should enable getting them and use if they are
applicable

This will be used, for example, for the relaxed ordering access flag which
unsupporting drivers can ignore.

Link: https://lore.kernel.org/r/1578506740-22188-7-git-send-email-yishaih@mellanox.com
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/uverbs: Verify MR access flags

Verify that MR access flags that are passed from user are all supported
ones, otherwise an error is returned.

Fixes: 4fca03778351 ("IB/uverbs: Move ib_access_flags and ib_read_counters_flags to uapi")
Link: https://lore.kernel.org/r/1578506740-22188-6-git-send-email-yishaih@mellanox.com
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/uverbs: Add ioctl command to get a device context

Allow future extensions of the get context command through the uverbs
ioctl kabi.

Unlike the uverbs version this does not return an async_fd as well, that
has to be done with another command.

Link: https://lore.kernel.org/r/1578506740-22188-5-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Remove ucontext_lock from the uverbs_destry_ufile_hw() path

This lock only serializes ucontext creation. Instead of checking the
ucontext_lock during destruction hold the existing hw_destroy_rwsem during
creation, which is the standard pattern for object creation.

The simplification of locking is needed for the next patch.

Link: https://lore.kernel.org/r/1578506740-22188-4-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Add UVERBS_METHOD_ASYNC_EVENT_ALLOC

Allow the async FD to be allocated separately from the context.

This is necessary to introduce the ioctl to create a context, as an ioctl
should only ever create a single uobject at a time.

If multiple async FDs are created then the first one is used to deliver
affiliated events from any ib_uevent_object, with all subsequent ones will
receive only unaffiliated events.

Link: https://lore.kernel.org/r/1578506740-22188-3-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Merge branch 'mlx5-next' into rdma.git for-next

From the mlx5-next branch at
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux

Merged due to dependencies in the next patches.

* branch 'mlx5-next':
  net/mlx5: Expose relaxed ordering bits
  net/mlx5: Add RoCE accelerator counters

net/mlx5: Expose relaxed ordering bits

Expose relaxed ordering bits in HCA capability and mkey context structs.

Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

net/mlx5: Add RoCE accelerator counters

Add RoCE accelerator definitions.

Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

IB/srp: Never use immediate data if it is disabled by a user

Some SRP targets that do not support specification SRP-2, put the garbage
to the reserved bits of the SRP login response.  The problem was not
detected for a long time because the SRP initiator ignored those bits. But
now one of them is used as SRP_LOGIN_RSP_IMMED_SUPP. And it causes a
critical error on the target when the initiator sends immediate data.

The ib_srp module has a use_imm_date parameter to enable or disable
immediate data manually. But it does not help in the above case, because
use_imm_date is ignored at handling the SRP login response. The problem is
definitely caused by a bug on the target side, but the initiator's
behavior also does not look correct.  The initiator should not use
immediate data if use_imm_date is disabled by a user.

This commit adds an additional checking of use_imm_date at the handling of
SRP login response to avoid unexpected use of immediate data.

Fixes: 882981f4a411 ("RDMA/srp: Add support for immediate data")
Link: https://lore.kernel.org/r/20200115133055.30232-1-sergeygo@mellanox.com
Signed-off-by: Sergey Gorenko <sergeygo@mellanox.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/rxe: Compute the maximum sges and inline size based on the WQE size

The SGE buffer size and max_inline data should be derived from the size of
the WQE. Each value individually sets the WQE size, so compute the actual
sizes based on the actual WQE size and configure the QP with the maximums.

Also fix the missing return of the actual maximum capability to the caller.

Link: https://lore.kernel.org/r/1578962480-17814-3-git-send-email-rao.shoaib@oracle.com
Signed-off-by: Rao Shoaib <rao.shoaib@oracle.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Introduce maximum WQE size to check limits

Introduce maximum WQE size to impose limits on max SGE's and inline data

Link: https://lore.kernel.org/r/1578962480-17814-2-git-send-email-rao.shoaib@oracle.com
Signed-off-by: Rao Shoaib <rao.shoaib@oracle.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/efa: Remove unused ucontext parameter from efa_qp_user_mmap_entries_remove

The ucontext parameter is unused, remove it.

Link: https://lore.kernel.org/r/20200114085706.82229-6-galpress@amazon.com
Reviewed-by: Firas JahJah <firasj@amazon.com>
Reviewed-by: Yossi Leybovich <sleybo@amazon.com>
Signed-off-by: Gal Pressman <galpress@amazon.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/efa: Remove {} brackets from single statement if

The {} brackets are not needed according to the Linux coding style.

Link: https://lore.kernel.org/r/20200114085706.82229-5-galpress@amazon.com
Reviewed-by: Daniel Kranzdorf <dkkranzd@amazon.com>
Reviewed-by: Firas JahJah <firasj@amazon.com>
Signed-off-by: Gal Pressman <galpress@amazon.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/efa: Device definitions documentation updates

Various clarifications and updates to the documentation of the device
definitions.
No functional changes in this patch.

Link: https://lore.kernel.org/r/20200114085706.82229-4-galpress@amazon.com
Reviewed-by: Firas JahJah <firasj@amazon.com>
Reviewed-by: Yossi Leybovich <sleybo@amazon.com>
Signed-off-by: Gal Pressman <galpress@amazon.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Add support for extended atomic in userspace

To support extended atomic operations including cmp & swap and fetch & add
of 8 bytes, 16 bytes, 32 bytes, 64 bytes in userspace, some field in qpc
should be configured.

Link: https://lore.kernel.org/r/1579052546-11746-1-git-send-email-liweihang@huawei.com
Signed-off-by: Jiaran Zhang <zhangjiaran@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Get pf capabilities from firmware

Get pf capabilities from firmware according to different hardwares, if it
fails, all capabilities will be set with a default value.

Link: https://lore.kernel.org/r/1578738761-3176-4-git-send-email-liweihang@huawei.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Add interfaces to get pf capabilities from firmware

pf capabilities are set by default for hip08 previously which should
depends on different types of hardware. So add new interfaces to get them
from firmware.

Link: https://lore.kernel.org/r/1578738761-3176-3-git-send-email-liweihang@huawei.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Remove some redundant variables related to capabilities

In struct hns_roce_caps, max_srq_sg and max_srqwqes is unused, and
max_srqs has the same effect with num_srqs. So remove them from this
structrue.

Link: https://lore.kernel.org/r/1578738761-3176-2-git-send-email-liweihang@huawei.com
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Use READ_ONCE for ib_ufile.async_file

The writer for async_file holds the ucontext_lock, while the readers are
left unlocked. Most readers rely on an implicit locking, either by having
a uobject (which cannot be created before a context) or by holding the
ib_ufile kref.

However ib_uverbs_free_hw_resources() has no implicit lock and has a
possible race. Make this all clear and sane by using READ_ONCE
consistently.

Link: https://lore.kernel.org/r/1578504126-9400-15-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Make ib_uverbs_async_event_file into a uobject

This makes async events aligned with completion events as both are full
uobjects of FD type and use the same uobject lifecycle.

A bunch of duplicate code is consolidated and the general flow between the
two FDs is now very similar.

Link: https://lore.kernel.org/r/1578504126-9400-14-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Remove the ufile arg from rdma_alloc_begin_uobject

Now that all callers provide a non-NULL attrs the ufile is redundant.
Adjust things so that the context handling is done inside alloc_uobj,
and the ib_uverbs_get_ucontext_file() is avoided if we already have the
context.

Link: https://lore.kernel.org/r/1578504126-9400-13-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Simplify type usage for ib_uverbs_async_handler()

This function works on an ib_uverbs_async_file. Accept that as a parameter
instead of the struct ib_uverbs_file.

Consoldiate all the callers working from an ib_uevent_object to a single
function and locate the async_file directly from the struct ib_uobject
instead of using context_ptr.

Link: https://lore.kernel.org/r/1578504126-9400-11-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Do not erase the type of ib_wq.uobject

This is a struct ib_uwq_object pointer, instead of using container_of()
all over the place just store it with its actual type.

Link: https://lore.kernel.org/r/1578504126-9400-10-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Do not erase the type of ib_srq.uobject

This is a struct ib_usrq_object pointer, instead of using container_of()
all over the place just store it with its actual type.

Link: https://lore.kernel.org/r/1578504126-9400-9-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Do not erase the type of ib_qp.uobject

This is a struct ib_uqp_object pointer, instead of using container_of()
all over the place just store it with its actual type.

Link: https://lore.kernel.org/r/1578504126-9400-8-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Do not erase the type of ib_cq.uobject

This is a struct ib_ucq_object pointer, instead of using container_of()
all over the place just store it with its actual type.

Link: https://lore.kernel.org/r/1578504126-9400-7-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Make ib_ucq_object use ib_uevent_object

Any uobject that sends events into the async_event_file should be using
ib_uevent_object so it can use the standard uevent based helper
functions. CQ pushes events into both the async_event and the comp_channel
in an open coded way. Move the async events related stuff to
ib_uevent_object.

Link: https://lore.kernel.org/r/1578504126-9400-6-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Do not allow alloc_commit to fail

This is a left over from an earlier version that creates a lot of
complexity for error unwind, particularly for FD uobjects.

The only reason this was done is so that anon_inode_get_file() could be
called with the final fops and a fully setup uobject. Both need to be
setup since unwinding anon_inode_get_file() via fput will call the
driver's release().

Now that the driver does not provide release, we no longer need to worry
about this complicated sequence, simply create the struct file at the
start and allow the core code's release function to deal with the abort
case.

This allows all the confusing error paths around commit to be removed.

Link: https://lore.kernel.org/r/1578504126-9400-5-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/mlx5: Simplify devx async commands

With the new FD structure the async commands do not need to hold any
references while running. The existing mlx5_cmd_exec_cb() and
mlx5_cmd_cleanup_async_ctx() provide enough synchronization to ensure
that all outstanding commands are completed before the uobject can be
destructed.

Remove the now confusing get_file() and the type erasure of the
devx_async_cmd_event_file.

Link: https://lore.kernel.org/r/1578504126-9400-4-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Simplify destruction of FD uobjects

FD uobjects have a weird split between the struct file and uobject
world. Simplify this to make them pure uobjects and use a generic release
method for all struct file operations.

This fixes the control flow so that mlx5_cmd_cleanup_async_ctx() is always
called before erasing the linked list contents to make the concurrancy
simpler to understand.

For this to work the uobject destruction must fence anything that it is
cleaning up - the design must not rely on struct file lifetime.

Only deliver_event() relies on the struct file to when adding new events
to the queue, add a is_destroyed check under lock to block it.

Link: https://lore.kernel.org/r/1578504126-9400-3-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/mlx5: Use RCU and direct refcounts to keep memory alive

dispatch_event_fd() runs from a notifier with minimal locking, and relies
on RCU and a file refcount to keep the uobject and eventfd alive.

As the next patch wants to remove the file_operations release function
from the drivers, re-organize things so that the devx_event_notifier()
path uses the existing RCU to manage the lifetime of the uobject and
eventfd.

Move the refcount puts to a call_rcu so that the objects are guaranteed to
exist and remove the indirect file refcount.

Link: https://lore.kernel.org/r/1578504126-9400-2-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/uverbs: Remove needs_kfree_rcu from uverbs_obj_type_class

After device disassociation the uapi_objects are destroyed and freed,
however it is still possible that core code can be holding a kref on the
uobject. When it finally goes to uverbs_uobject_free() via the kref_put()
it can trigger a use-after-free on the uapi_object.

Since needs_kfree_rcu is a micro optimization that only benefits file
uobjects, just get rid of it. There is no harm in using kfree_rcu even if
it isn't required, and the number of involved objects is small.

Link: https://lore.kernel.org/r/20200113143306.GA28717@ziepe.ca
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/mlx5: Add mmap support for VAR

Add mmap support for VAR, it uses the 'offset' command mode with
involvement of IB core APIs to find the previously allocated mmap entry.

Link: https://lore.kernel.org/r/20191212110928.334995-6-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/mlx5: Introduce VAR object and its alloc/destroy methods

Introduce VAR object and its alloc/destroy KABI methods. The internal
implementation uses the IB core API to manage mmap/munamp calls.

Link: https://lore.kernel.org/r/20191212110928.334995-5-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/mlx5: Extend caps stage to handle VAR capabilities

Extend caps stage to handle VAR capabilities.

Link: https://lore.kernel.org/r/20191212110928.334995-4-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Merge branch 'mlx5_vdpa' into rdma.git for-next

From the mlx5-next branch at
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux

Merged due to dependencies in the next patches.

* branch 'mlx5_vdpa':
  net/mlx5: Expose vDPA emulation device capabilities
  net/mlx5: Add Virtio Emulation related device capabilities

net/mlx5: Expose vDPA emulation device capabilities

Expose vDPA emulation device capabilities from the core layer.
It includes reading the capabilities from the firmware and exposing
helper functions to access the data.

Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Reviewed-by: Shahaf Shuler <shahafs@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

net/mlx5: Add Virtio Emulation related device capabilities

Add Virtio Emulation related fields to the device capabilities.

It includes a general bit to indicate whether Virtio Emulation is
supported and the capabilities structure itself.

Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Reviewed-by: Shahaf Shuler <shahafs@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

RDMA/core: Remove err in iw_query_port

Since we can return device->ops.query_port directly, so no need to keep
those lines.

Link: https://lore.kernel.org/r/20200109134043.15568-1-guoqing.jiang@cloud.ionos.com
Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Add support for reporting wc as software mode

When hardware is in resetting stage, we may can't poll back all the
expected work completions as the hardware won't generate cqe anymore.

This patch allows the driver to compose the expected wc instead of the
hardware during resetting stage. Once the hardware finished resetting, we
can poll cq from hardware again.

Link: https://lore.kernel.org/r/1578572412-25756-1-git-send-email-liweihang@huawei.com
Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Bugfix for posting a wqe with sge

Driver should first check whether the sge is valid, then fill the valid
sge and the caculated total into hardware, otherwise invalid sges will
cause an error.

Fixes: 52e3b42a2f58 ("RDMA/hns: Filter for zero length of sge in hip08 kernel mode")
Fixes: 7bdee4158b37 ("RDMA/hns: Fill sq wqe context of ud type in hip08")
Link: https://lore.kernel.org/r/1578571852-13704-1-git-send-email-liweihang@huawei.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Add RcvShortLengthErrCnt to hfi1stats

This counter, RxShrErr, is required for error analysis and debug.

Fixes: 7724105686e7 ("IB/hfi1: add driver files")
Link: https://lore.kernel.org/r/20200106134235.119356.29123.stgit@awfm-01.aw.intel.com
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Add software counter for ctxt0 seq drop

All other code paths increment some form of drop counter.

This was missed in the original implementation.

Fixes: 82c2611daaf0 ("staging/rdma/hfi1: Handle packets with invalid RHF on context 0")
Link: https://lore.kernel.org/r/20200106134228.119356.96828.stgit@awfm-01.aw.intel.com
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Return void in packet receiving functions

Packet receiving functions returns int value, and yet the return values
are not used at all.

This patch converts the functions to return void.

Link: https://lore.kernel.org/r/20200106134222.119356.84098.stgit@awfm-01.aw.intel.com
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Grzegorz Andrejczuk <grzegorz.andrejczuk@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Decouple IRQ name from type

IRQ name was connected to IRQ type, this is not sufficient and it would be
better to use name as argument to msix_request_irq instead of assigning it
to variables when function is called.

Index argument was required to generate name and now it can be removed.

To generate name correctly helpers function were added and updated.

Link: https://lore.kernel.org/r/20200106134216.119356.44478.stgit@awfm-01.aw.intel.com
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Grzegorz Andrejczuk <grzegorz.andrejczuk@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Create API for auto activate

Add an auto activate routine for use by the interrupt handler.

Link: https://lore.kernel.org/r/20200106134210.119356.43079.stgit@awfm-01.aw.intel.com
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: IB/hfi1: Add an API to handle special case drop

This patch pushes special case drop logic into an API to be shared by all
interrupt handlers.

Additionally, convert do_drop to a bool.

Link: https://lore.kernel.org/r/20200106134203.119356.36962.stgit@awfm-01.aw.intel.com
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Move common receive IRQ code to function

Tracing interrupts, incrementing interrupt counter and ASPM are part that
will be reused by HFI1 receive IRQ handlers.

Create common function to have shared code in one place.

Link: https://lore.kernel.org/r/20200106134157.119356.32656.stgit@awfm-01.aw.intel.com
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Grzegorz Andrejczuk <grzegorz.andrejczuk@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Add fast and slow handlers for receive context

This patch eliminate special cases by adding a fast_handler member to the
receive context and changes to the fast handler as specified in the new
variable. Initialize the variable as soon as the setting for dma tail is
known when the context is created.

Setting fast path is called every time when any context has entered slow
path. Add function to check if contexts is using fast path and do not set
fast path when it is already done to improve RCD fastpath setting.

Link: https://lore.kernel.org/r/20200106134150.119356.87558.stgit@awfm-01.aw.intel.com
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Grzegorz Andrejczuk <grzegorz.andrejczuk@intel.com>
Signed-off-by: Sadanand Warrier <sadanand.warrier@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Move chip specific functions to chip.c

Move routines and defines associated with hdrq size validation to a chip
specific routine since the limits are specific to the device.

Fix incorrect value for min size 2 -> 32

CSR writes should also be in chip.c.

Create a chip routine to write the hdrq specific CSRs and call as
appropriate.

Link: https://lore.kernel.org/r/20200106134144.119356.74312.stgit@awfm-01.aw.intel.com
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Fix locking in ib_uverbs_event_read

This should not be using ib_dev to test for disassociation, during
disassociation is_closed is set under lock and the waitq is triggered.

Instead check is_closed and be sure to re-obtain the lock to test the
value after the wait_event returns.

Fixes: 036b10635739 ("IB/uverbs: Enable device removal when there are active user space applications")
Link: https://lore.kernel.org/r/1578504126-9400-12-git-send-email-yishaih@mellanox.com
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Reviewed-by: Håkon Bugge <haakon.bugge@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/core: Fix build failure without hugepages

HPAGE_SHIFT is only defined on architectures that support hugepages:

drivers/infiniband/core/umem_odp.c: In function 'ib_umem_odp_get':
drivers/infiniband/core/umem_odp.c:245:26: error: 'HPAGE_SHIFT' undeclared (first use in this function); did you mean 'PAGE_SHIFT'?

Enclose this in an #ifdef.

Fixes: 9ff1b6466a29 ("IB/core: Fix ODP with IB_ACCESS_HUGETLB handling")
Link: https://lore.kernel.org/r/20200109084740.2872079-1-arnd@arndb.de
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/core: Rename event_handler_lock to qp_open_list_lock

This lock is used to protect the qp->open_list linked list. As a side
effect it seems to also globally serialize the qp event_handler, but it
isn't clear if that is a deliberate design.

Link: https://lore.kernel.org/r/20191212113024.336702-5-leon@kernel.org
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/core: Cut down single member ib_cache structure

Given that ib_cache structure has only single member now, merge the cache
lock directly in the ib_device.

Link: https://lore.kernel.org/r/20191212113024.336702-4-leon@kernel.org
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/core: Let IB core distribute cache update events

Currently when the low level driver notifies Pkey, GID, and port change
events they are notified to the registered handlers in the order they are
registered.

IB core and other ULPs such as IPoIB are interested in GID, LID, Pkey
change events.

Since all GID queries done by ULPs are serviced by IB core, and the IB
core deferes cache updates to a work queue, it is possible for other
clients to see stale cache data when they handle their own events.

For example, the below call tree shows how ipoib will call
rdma_query_gid() concurrently with the update to the cache sitting in the
WQ.

mlx5_ib_handle_event()
  ib_dispatch_event()
    ib_cache_event()
       queue_work() -> slow cache update

    [..]
    ipoib_event()
     queue_work()
       [..]
       work handler
         ipoib_ib_dev_flush_light()
           __ipoib_ib_dev_flush()
              ipoib_dev_addr_changed_valid()
                rdma_query_gid() <- Returns old GID, cache not updated.

Move all the event dispatch to a work queue so that the cache update is
always done before any clients are notified.

Fixes: f35faa4ba956 ("IB/core: Simplify ib_query_gid to always refer to cache")
Link: https://lore.kernel.org/r/20191212113024.336702-3-leon@kernel.org
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/mlx5: Do reverse sequence during device removal

When IB device profile initialization completes, device is marked as
active.

However, IB device is not marked inactive, during device removal flow. It
should be the mirror of the add flow.

Hence, mark it inactive during remove sequence.

Link: https://lore.kernel.org/r/20191212113024.336702-2-leon@kernel.org
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Fix coding style issues

Fix some coding style issuses without changing logic of codes, most of the
modification is unreasonable line breaks and alignments.

Link: https://lore.kernel.org/r/1578313276-29080-8-git-send-email-liweihang@huawei.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Lang Cheng <chenglang@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Replace custom macros HNS_ROCE_ALIGN_UP

HNS_ROCE_ALIGN_UP can be replaced by round_up() which is defined in
kernel.h.

Link: https://lore.kernel.org/r/1578313276-29080-7-git-send-email-liweihang@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Remove redundant print information

There are already necessary prints in outer function, prints in
hns_roce_function_clear() may confuse users. So these prints is removed.

Link: https://lore.kernel.org/r/1578313276-29080-6-git-send-email-liweihang@huawei.com
Signed-off-by: Yixing Liu <liuyixing1@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Delete unnessary parameters in hns_roce_v2_qp_modify()

Current state and new state of qp won't be configured when modifying qp,
so these two redundant parameters should be removed.

Link: https://lore.kernel.org/r/1578313276-29080-5-git-send-email-liweihang@huawei.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Update the value of qp type

The values used to represent service type of RC and UD should be
interchanged according to design of hardware. And it's better to define
these types in enumeration than macros.

Link: https://lore.kernel.org/r/1578313276-29080-4-git-send-email-liweihang@huawei.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Remove unused function hns_roce_init_eq_table()

hns_roce_init_eq_table() is an unused function that only retains its
declaration in driver.

Link: https://lore.kernel.org/r/1578313276-29080-3-git-send-email-liweihang@huawei.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Avoid printing address of mtt page

Address of a page shouldn't be printed in case of security issues.

Link: https://lore.kernel.org/r/1578313276-29080-2-git-send-email-liweihang@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Add trace points to follow MR allocation

Track the lifetime of ib_mr objects. Here's sample output from a test run
with NFS/RDMA:

           <...>-361   [009] 79238.772782: mr_alloc:             pd.id=3 mr.id=11 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.772812: mr_alloc:             pd.id=3 mr.id=12 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.772839: mr_alloc:             pd.id=3 mr.id=13 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.772866: mr_alloc:             pd.id=3 mr.id=14 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.772893: mr_alloc:             pd.id=3 mr.id=15 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.772921: mr_alloc:             pd.id=3 mr.id=16 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.772947: mr_alloc:             pd.id=3 mr.id=17 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.772974: mr_alloc:             pd.id=3 mr.id=18 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.773001: mr_alloc:             pd.id=3 mr.id=19 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.773028: mr_alloc:             pd.id=3 mr.id=20 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79238.773055: mr_alloc:             pd.id=3 mr.id=21 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.270942: mr_alloc:             pd.id=3 mr.id=22 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.270975: mr_alloc:             pd.id=3 mr.id=23 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271007: mr_alloc:             pd.id=3 mr.id=24 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271036: mr_alloc:             pd.id=3 mr.id=25 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271067: mr_alloc:             pd.id=3 mr.id=26 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271095: mr_alloc:             pd.id=3 mr.id=27 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271121: mr_alloc:             pd.id=3 mr.id=28 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271153: mr_alloc:             pd.id=3 mr.id=29 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271181: mr_alloc:             pd.id=3 mr.id=30 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271208: mr_alloc:             pd.id=3 mr.id=31 type=MEM_REG max_num_sg=30 rc=0
           <...>-361   [009] 79240.271236: mr_alloc:             pd.id=3 mr.id=32 type=MEM_REG max_num_sg=30 rc=0
           <...>-4351  [001] 79242.299400: mr_dereg:             mr.id=32
           <...>-4351  [001] 79242.299467: mr_dereg:             mr.id=31
           <...>-4351  [001] 79242.299554: mr_dereg:             mr.id=30
           <...>-4351  [001] 79242.299615: mr_dereg:             mr.id=29
           <...>-4351  [001] 79242.299684: mr_dereg:             mr.id=28
           <...>-4351  [001] 79242.299748: mr_dereg:             mr.id=27
           <...>-4351  [001] 79242.299812: mr_dereg:             mr.id=26
           <...>-4351  [001] 79242.299874: mr_dereg:             mr.id=25
           <...>-4351  [001] 79242.299944: mr_dereg:             mr.id=24
           <...>-4351  [001] 79242.300009: mr_dereg:             mr.id=23
           <...>-4351  [001] 79242.300190: mr_dereg:             mr.id=22
           <...>-4351  [001] 79242.300263: mr_dereg:             mr.id=21
           <...>-4351  [001] 79242.300326: mr_dereg:             mr.id=20
           <...>-4351  [001] 79242.300388: mr_dereg:             mr.id=19
           <...>-4351  [001] 79242.300450: mr_dereg:             mr.id=18
           <...>-4351  [001] 79242.300516: mr_dereg:             mr.id=17
           <...>-4351  [001] 79242.300629: mr_dereg:             mr.id=16
           <...>-4351  [001] 79242.300718: mr_dereg:             mr.id=15
           <...>-4351  [001] 79242.300784: mr_dereg:             mr.id=14
           <...>-4351  [001] 79242.300879: mr_dereg:             mr.id=13
           <...>-4351  [001] 79242.300945: mr_dereg:             mr.id=12
           <...>-4351  [001] 79242.301012: mr_dereg:             mr.id=11

Some features of the output:
- The lifetime and owner PD of each MR is clearly visible.
- The type of MR is captured, as is the SGE array size.
- Failing MR allocation can be recorded.

Link: https://lore.kernel.org/r/20191218201820.30584.34636.stgit@manet.1015granger.net
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/core: Trace points for diagnosing completion queue issues

Sample trace events:

   kworker/u29:0-300   [007]   120.042217: cq_alloc:             cq.id=4 nr_cqe=161 comp_vector=2 poll_ctx=WORKQUEUE
          <idle>-0     [002]   120.056292: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   120.056402: cq_process:           cq.id=4 wake-up took 109 [us] from interrupt
    kworker/2:1H-482   [002]   120.056407: cq_poll:              cq.id=4 requested 16, returned 1
          <idle>-0     [002]   120.067503: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   120.067537: cq_process:           cq.id=4 wake-up took 34 [us] from interrupt
    kworker/2:1H-482   [002]   120.067541: cq_poll:              cq.id=4 requested 16, returned 1
          <idle>-0     [002]   120.067657: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   120.067672: cq_process:           cq.id=4 wake-up took 15 [us] from interrupt
    kworker/2:1H-482   [002]   120.067674: cq_poll:              cq.id=4 requested 16, returned 1

 ...

         systemd-1     [002]   122.392653: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   122.392688: cq_process:           cq.id=4 wake-up took 35 [us] from interrupt
    kworker/2:1H-482   [002]   122.392693: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.392836: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.392970: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.393083: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.393195: cq_poll:              cq.id=4 requested 16, returned 3

Several features to note in this output:
 - The WCE count and context type are reported at allocation time
 - The CPU and kworker for each CQ is evident
 - The CQ's restracker ID is tagged on each trace event
 - CQ poll scheduling latency is measured
 - Details about how often single completions occur versus multiple
   completions are evident
 - The cost of the ULP's completion handler is recorded

Link: https://lore.kernel.org/r/20191218201815.30584.3481.stgit@manet.1015granger.net
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/cma: Add trace points in RDMA Connection Manager

Record state transitions as each connection is established. The IP address
of both peers and the Type of Service is reported. These trace points are
not in performance hot paths.

Also, record each cm_event_handler call to ULPs. This eliminates the need
for each ULP to add its own similar trace point in its CM event handler
function.

These new trace points appear in a new trace subsystem called "rdma_cma".

Sample events:

           <...>-220   [004]   121.430733: cm_id_create:         cm.id=0
           <...>-472   [003]   121.430991: cm_event_handler:     cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 ADDR_RESOLVED (0/0)
           <...>-472   [003]   121.430995: cm_event_done:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 result=0
           <...>-472   [003]   121.431172: cm_event_handler:     cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 ROUTE_RESOLVED (2/0)
           <...>-472   [003]   121.431174: cm_event_done:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 result=0
           <...>-220   [004]   121.433480: cm_qp_create:         cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 pd.id=2 qp_type=RC send_wr=4091 recv_wr=256 qp_num=521 rc=0
           <...>-220   [004]   121.433577: cm_send_req:          cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 qp_num=521
     kworker/1:2-973   [001]   121.436190: cm_send_mra:          cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0
     kworker/1:2-973   [001]   121.436340: cm_send_rtu:          cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0
     kworker/1:2-973   [001]   121.436359: cm_event_handler:     cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 ESTABLISHED (9/0)
     kworker/1:2-973   [001]   121.436365: cm_event_done:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 result=0
           <...>-1975  [005]   123.161954: cm_disconnect:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0
           <...>-1975  [005]   123.161974: cm_sent_dreq:         cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0
           <...>-220   [004]   123.162102: cm_disconnect:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0
     kworker/0:1-13    [000]   123.162391: cm_event_handler:     cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 DISCONNECTED (10/0)
     kworker/0:1-13    [000]   123.162393: cm_event_done:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 result=0
           <...>-220   [004]   123.164456: cm_qp_destroy:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0 qp_num=521
           <...>-220   [004]   123.165290: cm_id_destroy:        cm.id=0 src=192.168.2.51:35090 dst=192.168.2.55:20049 tos=0

Some features to note:
- restracker ID of the rdma_cm_id is tagged on each trace event
- The source and destination IP addresses and TOS are reported
- CM event upcalls are shown with decoded event and status
- CM state transitions are reported
- rdma_cm_id lifetime events are captured
- The latency of ULP CM event handlers is reported
- Lifetime events of associated QPs are reported
- Device removal and insertion is reported

This patch is based on previous work by:

Saeed Mahameed <saeedm@mellanox.com>
Mukesh Kacker <mukesh.kacker@oracle.com>
Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Aron Silverton <aron.silverton@oracle.com>
Avinash Repaka <avinash.repaka@oracle.com>
Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>

Link: https://lore.kernel.org/r/20191218201810.30584.3052.stgit@manet.1015granger.net
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/cm: Delete unused CM ARP functions

Clean the code by deleting ARP functions, which are not called anyway.

Link: https://lore.kernel.org/r/20191212093830.316934-46-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/cm: Delete unused CM LAP functions

Clean the code by deleting LAP functions, which are not called anyway.

Link: https://lore.kernel.org/r/20191212093830.316934-43-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/i40iw: fix a potential NULL pointer dereference

A NULL pointer can be returned by in_dev_get(). Thus add a corresponding
check so that a NULL pointer dereference will be avoided at this place.

Fixes: 8e06af711bf2 ("i40iw: add main, hdr, status")
Link: https://lore.kernel.org/r/1577672668-46499-1-git-send-email-xiyuyang19@fudan.edu.cn
Signed-off-by: Xiyu Yang <xiyuyang19@fudan.edu.cn>
Signed-off-by: Xin Tan <tanxin.ctf@gmail.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/rxe: Fix error type of mmap_offset

The type of mmap_offset should be u64 instead of int to match the type of
mminfo.offset. If otherwise, after we create several thousands of CQs, it
will run into overflow issues.

Link: https://lore.kernel.org/r/20191227113613.5020-1-kejiewei.cn@gmail.com
Signed-off-by: Jiewei Ke <kejiewei.cn@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/mlx5: use true,false for bool variable

Fixes coccicheck warning:

drivers/infiniband/hw/mlx5/mr.c:150:2-26: WARNING: Assignment of 0/1 to bool variable
drivers/infiniband/hw/mlx5/mr.c:1455:2-26: WARNING: Assignment of 0/1 to bool variable
drivers/infiniband/hw/mlx5/qp.c:1874:6-20: WARNING: Assignment of 0/1 to bool variable

Link: https://lore.kernel.org/r/1577176812-2238-6-git-send-email-zhengbin13@huawei.com
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhengbin <zhengbin13@huawei.com>
Acked-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/mlx4: use true,false for bool variable

Fixes coccicheck warning:

drivers/infiniband/hw/mlx4/qp.c:852:2-14: WARNING: Assignment of 0/1 to bool variable
drivers/infiniband/hw/mlx4/qp.c:3087:3-10: WARNING: Assignment of 0/1 to bool variable

Link: https://lore.kernel.org/r/1577176812-2238-5-git-send-email-zhengbin13@huawei.com
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhengbin <zhengbin13@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/iser: use true,false for bool variable

Fixes coccicheck warning:

drivers/infiniband/ulp/iser/iser_memory.c:530:2-21: WARNING: Assignment of 0/1 to bool variable
drivers/infiniband/ulp/iser/iser_verbs.c:1096:2-21: WARNING: Assignment of 0/1 to bool variable

Link: https://lore.kernel.org/r/1577176812-2238-4-git-send-email-zhengbin13@huawei.com
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhengbin <zhengbin13@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: use true,false for bool variable

Fixes coccicheck warning:

drivers/infiniband/hw/hfi1/rc.c:2602:1-8: WARNING: Assignment of 0/1 to bool variable

Link: https://lore.kernel.org/r/1577176812-2238-3-git-send-email-zhengbin13@huawei.com
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhengbin <zhengbin13@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/siw: use true,false for bool variable

Fixes coccicheck warning:

drivers/infiniband/sw/siw/siw_cm.c:32:18-41: WARNING: Assignment of 0/1 to bool variable

Link: https://lore.kernel.org/r/1577176812-2238-2-git-send-email-zhengbin13@huawei.com
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhengbin <zhengbin13@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/core: Fix ODP with IB_ACCESS_HUGETLB handling

As VMAs for a given range might not be available as part of the
registration phase in ODP.

ib_init_umem_odp() considered the expected page shift value that was
previously set and initializes its internals accordingly.

If memory isn't backed by physical contiguous pages aligned to a hugepage
boundary an error will be set as part of the page fault flow and come back
to the user as some failed RDMA operation.

Fixes: 0008b84ea9af ("IB/umem: Add support to huge ODP")
Link: https://lore.kernel.org/r/20191222124649.52300-4-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/core: Fix ODP get user pages flow

The nr_pages argument of get_user_pages_remote() should always be in terms
of the system page size, not the MR page size. Use PAGE_SIZE instead of
umem_odp->page_shift.

Fixes: 403cd12e2cf7 ("IB/umem: Add contiguous ODP support")
Link: https://lore.kernel.org/r/20191222124649.52300-3-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/mlx5: Unify ODP MR code paths to allow extra flexibility

Building MR translation table in the ODP case requires additional
flexibility, namely random access to DMA addresses. Make both direct and
indirect ODP MR use same code path, separated from the non-ODP MR code
path.

With the restructuring the correct page_shift is now used around
__mlx5_ib_populate_pas().

Fixes: d2183c6f1958 ("RDMA/umem: Move page_shift from ib_umem to ib_odp_umem")
Link: https://lore.kernel.org/r/20191222124649.52300-2-leon@kernel.org
Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/rdmavt: Correct comments in rdmavt_qp.h header

Comments need to be with the definition of rvt_restart_sge().

Other comments were duplicated in sw/rdmavt/rc.c and were removed.

Fixes: 385156c5f2a6 ("IB/hfi: Move RC functions into a header file")
Link: https://lore.kernel.org/r/20191219211934.58387.88014.stgit@awfm-01.aw.intel.com
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: List all receive contexts from debugfs

The current debugfs output for receive contexts (rcds), stops after the
kernel receive contexts have been displayed.  This is not enough
information to fully diagnose packet drops.

Display all of the receive contexts.

Augment the output with some more context information.

Limit the ring buffer header output to 5 entries to avoid overextending
the sequential file output.

Fixes: bf808b5039c ("IB/hfi1: Add kernel receive context info to debugfs")
Link: https://lore.kernel.org/r/20191219211928.58387.20737.stgit@awfm-01.aw.intel.com
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/hfi1: Add accessor API routines to access context members

This patch adds a set of accessor routines to access context members.

Link: https://lore.kernel.org/r/20191219211922.58387.26548.stgit@awfm-01.aw.intel.com
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/mlx4: Redo TX checksum offload in line with docs

Ingress checksum offload was not working for IPv6 frames because the
conditional expression that checks validation status passed from the
hardware was not matching the algorithm described in the documentation.

This patch defines L4_CSUM flag (which falls inside the badfcs_enc field
in the existing definition of the CQE layout) and replaces the conditional
expression with the one defined in the "ConnectX(r) Family Programmer's
Manual" document.

Link: https://lore.kernel.org/r/20191219134847.413582-1-leon@kernel.org
Signed-off-by: Eugene Crosser <evgenii.cherkashin@profitbricks.com>
Reviewed-by: Jack Wang <jinpu.wang@profitbricks.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/cm: Use RCU synchronization mechanism to protect cm_id_private xa_load()

The RCU mechanism is optimized for read-mostly scenarios and therefore
more suitable to protect the cm_id_private to decrease "cm.lock"
congestion.

This patch replaces the existing spinlock locking mechanism and kfree with
RCU mechanism in places where spinlock(cm.lock) protected xa_load
returning the cm_id_priv

In addition, delete the cm_get_id() function as there is no longer a
distinction if the caller already holds the cm_lock.

Remove an open coded version of cm_get_id().

Link: https://lore.kernel.org/r/20191219134750.413429-1-leon@kernel.org
Signed-off-by: Danit Goldberg <danitg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/srpt: Remove unnecessary assertion in srpt_queue_response

Since ch has already been de-referenced by the time we get to the BUG_ON,
it is useless. The back trace alone is enough to tell what is going on,
delete the redundant BUG_ON.

Link: https://lore.kernel.org/r/20191217194437.25568-1-pakki001@umn.edu
Signed-off-by: Aditya Pakki <pakki001@umn.edu>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/netlink: Do not always generate an ACK for some netlink operations

In rdma_nl_rcv_skb(), the local variable err is assigned the return value
of the supplied callback function, which could be one of
ib_nl_handle_resolve_resp(), ib_nl_handle_set_timeout(), or
ib_nl_handle_ip_res_resp(). These three functions all return skb->len on
success.

rdma_nl_rcv_skb() is merely a copy of netlink_rcv_skb(). The callback
functions used by the latter have the convention: "Returns 0 on success or
a negative error code".

In particular, the statement (equal for both functions):

   if (nlh->nlmsg_flags & NLM_F_ACK || err)

implies that rdma_nl_rcv_skb() always will ack a message, independent of
the NLM_F_ACK being set in nlmsg_flags or not.

The fix could be to change the above statement, but it is better to keep
the two *_rcv_skb() functions equal in this respect and instead change the
three callback functions in the rdma subsystem to the correct convention.

Fixes: 2ca546b92a02 ("IB/sa: Route SA pathrecord query through netlink")
Fixes: ae43f8286730 ("IB/core: Add IP to GID netlink offload")
Link: https://lore.kernel.org/r/20191216120436.3204814-1-haakon.bugge@oracle.com
Suggested-by: Mark Haywood <mark.haywood@oracle.com>
Signed-off-by: Håkon Bugge <haakon.bugge@oracle.com>
Tested-by: Mark Haywood <mark.haywood@oracle.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

IB/mlx5: Fix outstanding_pi index for GSI qps

Commit b0ffeb537f3a ("IB/mlx5: Fix iteration overrun in GSI qps") changed
the way outstanding WRs are tracked for the GSI QP. But the fix did not
cover the case when a call to ib_post_send() fails and updates index to
track outstanding.

Since the prior commmit outstanding_pi should not be bounded otherwise the
loop generate_completions() will fail.

Fixes: b0ffeb537f3a ("IB/mlx5: Fix iteration overrun in GSI qps")
Link: https://lore.kernel.org/r/1576195889-23527-1-git-send-email-psajeepa@purestorage.com
Signed-off-by: Prabhath Sajeepa <psajeepa@purestorage.com>
Acked-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/siw: Simplify QP representation

Change siw_qp to contain ib_qp. Use rdma_is_kernel_res() on contained
ib_qp to distinguish kernel level from user level applications
resources. Apply same mechanism for kernel/user level application
detection to completion queues.

Link: https://lore.kernel.org/r/20191210161729.31598-1-bmt@zurich.ibm.com
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/hns: Simplify the calculation and usage of wqe idx for post verbs

Currently, the wqe idx is calculated repeatly everywhere it is used.  This
patch defines wqe_idx and calculated it only once, then just use it as
needed.

Fixes: 2d40788825ac ("RDMA/hns: Add support for processing send wr and receive wr")
Link: https://lore.kernel.org/r/1575981902-5274-1-git-send-email-liweihang@hisilicon.com
Signed-off-by: Yixian Liu <liuyixian@huawei.com>
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/bnxt_re: Report more number of completion vectors

Report the the data path MSIx vectors allocated by driver as number of
completion vectors. One interrupt vector is used for Control path. So
reporting one less than the total number of MSIx vectors allocated by the
driver.

Link: https://lore.kernel.org/r/1574671174-5064-7-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

RDMA/qedr: Add kernel capability flags for dpm enabled mode

HW/FW support two types of latency enhancement features.  Until now
user-space implemented only edpm (enhanced dpm).  We add kernel capability
flags to differentiate between current FW in kernel that supports both
ldpm and edpm.  Since edpm is not yet supported for iWARP we add different
flags for iWARP + RoCE.  We also fix bad practice of defining sizes in
rdma-core and pass initialization to kernel, for forward compatibility.

The capability flags are added for backward-forward compatibility between
kernel and rdma-core for qedr.

Before this change there was a field called dpm_enabled which could hold
either 0 or 1 value, this indicated whether RoCE edpm was enabled or
not. We modified this field to be dpm_flags, and bit 1 still holds the
same meaning of RoCE edpm being enabled or not.

Link: https://lore.kernel.org/r/20191121112957.25162-1-michal.kalderon@marvell.com
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Linux 5.5-rc4

Merge tag 'riscv/for-v5.5-rc4' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Paul Walmsley:
 "One important fix for RISC-V:

   - Redirect any incoming syscall with an ID less than -1 to
     sys_ni_syscall, rather than allowing them to fall through into the
     syscall handler.

  and two minor build fixes:

   - Export __asm_copy_{from,to}_user() from where they are defined.
     This fixes a build error triggered by some randconfigs.

   - Export flush_icache_all(). I'd resisted this before, since
     historically we didn't want modules to be able to flush the I$
     directly; but apparently everyone else is doing it now"

* tag 'riscv/for-v5.5-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: export flush_icache_all to modules
  riscv: reject invalid syscalls below -1
  riscv: fix compile failure with EXPORT_SYMBOL() & !MMU

Merge tag 'locks-v5.5-1' of git://git./linux/kernel/git/jlayton/linux

Pull /proc/locks formatting fix from Jeff Layton:
 "This is a trivial fix for a _very_ long standing bug in /proc/locks
  formatting. Ordinarily, I'd wait for the merge window for something
  like this, but it is making it difficult to validate some overlayfs
  fixes.

  I've also gone ahead and marked this for stable"

* tag 'locks-v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/jlayton/linux:
  locks: print unsigned ino in /proc/locks

Merge tag '5.5-rc3-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "One performance fix for large directory searches, and one minor style
  cleanup noticed by Clang"

* tag '5.5-rc3-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: Optimize readdir on reparse points
  cifs: Adjust indentation in smb2_open_file

locks: print unsigned ino in /proc/locks

An ino is unsigned, so display it as such in /proc/locks.

Cc: stable@vger.kernel.org
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>

riscv: export flush_icache_all to modules

This is needed by LKDTM (crash dump test module), it calls
flush_icache_range(), which on RISC-V turns into flush_icache_all(). On
other architectures, the actual implementation is exported, so follow
that precedence and export it here too.

Fixes build of CONFIG_LKDTM that fails with:
ERROR: "flush_icache_all" [drivers/misc/lkdtm/lkdtm.ko] undefined!

Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>

riscv: reject invalid syscalls below -1

Running "stress-ng --enosys 4 -t 20 -v" showed a large number of kernel oops
with "Unable to handle kernel paging request at virtual address" message. This
happens when enosys stressor starts testing random non-valid syscalls.

I forgot to redirect any syscall below -1 to sys_ni_syscall.

With the patch kernel oops messages are gone while running stress-ng enosys
stressor.

Signed-off-by: David Abdurachmanov <david.abdurachmanov@sifive.com>
Fixes: 5340627e3fe0 ("riscv: add support for SECCOMP and SECCOMP_FILTER")
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>

riscv: fix compile failure with EXPORT_SYMBOL() & !MMU

When support for !MMU was added, the declaration of
__asm_copy_to_user() & __asm_copy_from_user() were #ifdefed
out hence their EXPORT_SYMBOL() give an error message like:
  .../riscv_ksyms.c:13:15: error: '__asm_copy_to_user' undeclared here
  .../riscv_ksyms.c:14:15: error: '__asm_copy_from_user' undeclared here

Since these symbols are not defined with !MMU it's wrong to export them.
Same for __clear_user() (even though this one is also declared in
include/asm-generic/uaccess.h and thus doesn't give an error message).

Fix this by doing the EXPORT_SYMBOL() directly where these symbols
are defined: inside lib/uaccess.S itself.

Fixes: 6bd33e1ece52 ("riscv: fix compile failure with EXPORT_SYMBOL() & !MMU")
Reported-by: kbuild test robot <lkp@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>