platform/kernel/linux-rpi.git
4 years agoRevert "RDMA/srpt: Postpone HCA removal until after configfs directory removal"
Bart Van Assche [Fri, 1 Nov 2019 20:47:56 +0000 (13:47 -0700)]
Revert "RDMA/srpt: Postpone HCA removal until after configfs directory removal"

Although the mentioned patch fixes a use-after-free bug, it introduces a
hang during shutdown. Since the latter is worse, revert this patch.

Link: https://lore.kernel.org/r/20191101204756.182162-1-bvanassche@acm.org
Reported-by: Honggang Li <honli@redhat.com>
Fixes: 9b64f7d0bb0a ("RDMA/srpt: Postpone HCA removal until after configfs directory removal")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Acked-by: Honggang Li <honli@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/qedr: Remove unsupported modify_port callback
Kamal Heib [Mon, 28 Oct 2019 15:59:31 +0000 (17:59 +0200)]
RDMA/qedr: Remove unsupported modify_port callback

There is no need to return always zero for function which is not
supported.

Fixes: ac1b36e55a51 ("qedr: Add support for user context verbs")
Link: https://lore.kernel.org/r/20191028155931.1114-5-kamalheib1@gmail.com
Signed-off-by: Kamal Heib <kamalheib1@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/ocrdma: Remove unsupported modify_port callback
Kamal Heib [Mon, 28 Oct 2019 15:59:30 +0000 (17:59 +0200)]
RDMA/ocrdma: Remove unsupported modify_port callback

There is no need to return always zero for function which is not
supported.

Fixes: fe2caefcdf58 ("RDMA/ocrdma: Add driver for Emulex OneConnect IBoE RDMA adapter")
Link: https://lore.kernel.org/r/20191028155931.1114-4-kamalheib1@gmail.com
Signed-off-by: Kamal Heib <kamalheib1@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/hns: Remove unsupported modify_port callback
Kamal Heib [Mon, 28 Oct 2019 15:59:29 +0000 (17:59 +0200)]
RDMA/hns: Remove unsupported modify_port callback

There is no need to return always zero for function which is not
supported.

Fixes: 9a4435375cd1 ("IB/hns: Add driver files for hns RoCE driver")
Link: https://lore.kernel.org/r/20191028155931.1114-3-kamalheib1@gmail.com
Signed-off-by: Kamal Heib <kamalheib1@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/core: Fix return code when modify_port isn't supported
Kamal Heib [Mon, 28 Oct 2019 15:59:28 +0000 (17:59 +0200)]
RDMA/core: Fix return code when modify_port isn't supported

Improve return code from ib_modify_port() by doing the following:
 - Use "-EOPNOTSUPP" instead "-ENOSYS" which is the proper return code

 - Allow only fake IB_PORT_CM_SUP manipulation for RoCE providers that
   didn't implement the modify_port callback, otherwise return
   "-EOPNOTSUPP"

Fixes: 61e0962d5221 ("IB: Avoid ib_modify_port() failure for RoCE devices")
Link: https://lore.kernel.org/r/20191028155931.1114-2-kamalheib1@gmail.com
Signed-off-by: Kamal Heib <kamalheib1@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/qedr: Add iWARP doorbell recovery support
Michal Kalderon [Wed, 30 Oct 2019 09:44:17 +0000 (11:44 +0200)]
RDMA/qedr: Add iWARP doorbell recovery support

This patch adds the iWARP specific doorbells to the doorbell recovery
mechanism.

Link: https://lore.kernel.org/r/20191030094417.16866-9-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/qedr: Add doorbell overflow recovery support
Michal Kalderon [Wed, 30 Oct 2019 09:44:16 +0000 (11:44 +0200)]
RDMA/qedr: Add doorbell overflow recovery support

Use the doorbell recovery mechanism to register rdma related doorbells
that will be restored in case there is a doorbell overflow attention.

Link: https://lore.kernel.org/r/20191030094417.16866-8-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/qedr: Use the common mmap API
Michal Kalderon [Wed, 30 Oct 2019 09:44:15 +0000 (11:44 +0200)]
RDMA/qedr: Use the common mmap API

Remove all functions related to mmap from qedr and use the common API.

Link: https://lore.kernel.org/r/20191030094417.16866-7-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/siw: Use the common mmap_xa helpers
Michal Kalderon [Wed, 30 Oct 2019 09:44:14 +0000 (11:44 +0200)]
RDMA/siw: Use the common mmap_xa helpers

Remove the functions related to managing the mmap_xa database.  This code
is now common in ib_core.

Link: https://lore.kernel.org/r/20191030094417.16866-6-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Reviewed-by: Bernard Metzler <bmt@zurich.ibm.com>
Tested-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/efa: Use the common mmap_xa helpers
Michal Kalderon [Wed, 30 Oct 2019 09:44:13 +0000 (11:44 +0200)]
RDMA/efa: Use the common mmap_xa helpers

Remove the functions related to managing the mmap_xa database.  This code
was replaced with common code in ib_core.

Link: https://lore.kernel.org/r/20191030094417.16866-5-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA: Connect between the mmap entry and the umap_priv structure
Michal Kalderon [Wed, 30 Oct 2019 09:44:12 +0000 (11:44 +0200)]
RDMA: Connect between the mmap entry and the umap_priv structure

The rdma_user_mmap_io interface created a common interface for drivers to
correctly map hw resources and zap them once the ucontext is destroyed
enabling the drivers to safely free the hw resources.

However, this meant the drivers need to delay freeing the resource to the
ucontext destroy phase to ensure they were no longer mapped.  The new
mechanism for a common way of handling user/driver address mapping enabled
notifying the driver if all umap_priv mappings were removed, and enabled
freeing the hw resources when they are done with and not delay it until
ucontext destroy.

Since not all drivers use the mechanism, NULL can be sent to the
rdma_user_mmap_io interface to continue working as before.  Drivers that
use the mmap_xa interface can pass the entry being mapped to the
rdma_user_mmap_io function to be linked together.

Link: https://lore.kernel.org/r/20191030094417.16866-4-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
4 years agoRDMA/core: Create mmap database and cookie helper functions
Michal Kalderon [Wed, 30 Oct 2019 09:44:11 +0000 (11:44 +0200)]
RDMA/core: Create mmap database and cookie helper functions

Create some common API's for adding entries to a xa_mmap. Searching for
an entry and freeing one.

The general approach is copied from the EFA driver and improved to be more
general and do more to help the drivers. Integration with the core allows
a reference counted scheme with a free function so that the driver can
know when its mmaps are all gone.

This significant new functionality will be helpful for drivers to have the
correct lifetime model for mmap objects.

Link: https://lore.kernel.org/r/20191030094417.16866-3-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/core: Move core content from ib_uverbs to ib_core
Michal Kalderon [Wed, 30 Oct 2019 09:44:10 +0000 (11:44 +0200)]
RDMA/core: Move core content from ib_uverbs to ib_core

Move functionality that is called by the driver, which is
related to umap, to a new file that will be linked in ib_core.
This is a first step in later enabling ib_uverbs to be optional.
vm_ops is now initialized in ib_uverbs_mmap instead of
priv_init to avoid having to move all the rdma_umap functions
as well.

Link: https://lore.kernel.org/r/20191030094417.16866-2-michal.kalderon@marvell.com
Suggested-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoIB/mlx5: Test write combining support
Michael Guralnik [Mon, 10 Jun 2019 12:21:24 +0000 (15:21 +0300)]
IB/mlx5: Test write combining support

Linux can run in all sorts of physical machines and VMs where write
combining may or may not be supported. Currently there is no way to
reliably tell if the system supports WC, or not. The driver uses WC to
optimize posting work to the HCA, and getting this wrong in either
direction can cause a significant performance loss.

Add a test in mlx5_ib initialization process to test whether
write-combining is supported on the machine.  The test will run as part of
the enable_driver callback to ensure that the test runs after the device
is setup and can create and modify the QP needed, but runs before the
device is exposed to the users.

The test opens UD QP and posts NOP WQEs, the WQE written to the BlueFlame
is different from the WQE in memory, requesting CQE only on the BlueFlame
WQE. By checking whether we received a completion on one of these WQEs we
can know if BlueFlame succeeded and this write-combining must be
supported.

Change reporting of BlueFlame support to be dependent on write-combining
support instead of the FW's guess as to what the machine can do.

Link: https://lore.kernel.org/r/20191027062234.10993-1-leon@kernel.org
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Return proper error value
Leon Romanovsky [Tue, 29 Oct 2019 05:57:21 +0000 (07:57 +0200)]
RDMA/mlx5: Return proper error value

Returned value from mlx5_mr_cache_alloc() is checked to be error or real
pointer. Return proper error code instead of NULL which is not checked
later.

Fixes: 81713d3788d2 ("IB/mlx5: Add implicit MR support")
Link: https://lore.kernel.org/r/20191029055721.7192-1-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>
5 years agoRDMA/hns: Fix build error again
Arnd Bergmann [Mon, 7 Oct 2019 21:18:08 +0000 (23:18 +0200)]
RDMA/hns: Fix build error again

This is not the first attempt to fix building random configurations,
unfortunately the attempt in commit a07fc0bb483e ("RDMA/hns: Fix build
error") caused a new problem when CONFIG_INFINIBAND_HNS_HIP06=m and
CONFIG_INFINIBAND_HNS_HIP08=y:

drivers/infiniband/hw/hns/hns_roce_main.o:(.rodata+0xe60): undefined reference to `__this_module'

Revert commits a07fc0bb483e ("RDMA/hns: Fix build error") and
a3e2d4c7e766 ("RDMA/hns: remove obsolete Kconfig comment") to get back to
the previous state, then fix the issues described there differently, by
adding more specific dependencies: INFINIBAND_HNS can now only be built-in
if at least one of HNS or HNS3 are built-in, and the individual back-ends
are only available if that code is reachable from the main driver.

Fixes: a07fc0bb483e ("RDMA/hns: Fix build error")
Fixes: a3e2d4c7e766 ("RDMA/hns: remove obsolete Kconfig comment")
Fixes: dd74282df573 ("RDMA/hns: Initialize the PCI device for hip08 RoCE")
Fixes: 08805fdbeb2d ("RDMA/hns: Split hw v1 driver from hns roce driver")
Link: https://lore.kernel.org/r/20191007211826.3361202-1-arnd@arndb.de
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoMerge branch 'odp_rework' into rdma.git for-next
Jason Gunthorpe [Mon, 28 Oct 2019 19:44:35 +0000 (16:44 -0300)]
Merge branch 'odp_rework' into rdma.git for-next

Jason Gunthorpe says:

====================
In order to hoist the interval tree code out of the drivers and into the
mmu_notifiers it is necessary for the drivers to not use the interval tree
for other things.

This series replaces the interval tree with an xarray and along the way
re-aligns all the locking to use a sensible SRCU model where the 'update'
step is done by modifying an xarray.

The result is overall much simpler and with less locking in the critical
path. Many functions were reworked for clarity and small details like
using 'imr' to refer to the implicit MR make the entire code flow here
more readable.

This also squashes at least two race bugs on its own, and quite possibily
more that haven't been identified.
====================

Merge conflicts with the odp statistics patch resolved.

* branch 'odp_rework':
  RDMA/odp: Remove broken debugging call to invalidate_range
  RDMA/mlx5: Do not race with mlx5_ib_invalidate_range during create and destroy
  RDMA/mlx5: Do not store implicit children in the odp_mkeys xarray
  RDMA/mlx5: Rework implicit ODP destroy
  RDMA/mlx5: Avoid double lookups on the pagefault path
  RDMA/mlx5: Reduce locking in implicit_mr_get_data()
  RDMA/mlx5: Use an xarray for the children of an implicit ODP
  RDMA/mlx5: Split implicit handling from pagefault_mr
  RDMA/mlx5: Set the HW IOVA of the child MRs to their place in the tree
  RDMA/mlx5: Lift implicit_mr_alloc() into the two routines that call it
  RDMA/mlx5: Rework implicit_mr_get_data
  RDMA/mlx5: Delete struct mlx5_priv->mkey_table
  RDMA/mlx5: Use a dedicated mkey xarray for ODP
  RDMA/mlx5: Split sig_err MR data into its own xarray
  RDMA/mlx5: Use SRCU properly in ODP prefetch

Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/odp: Remove broken debugging call to invalidate_range
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:35 +0000 (13:09 -0300)]
RDMA/odp: Remove broken debugging call to invalidate_range

invalidate_range() also obtains the umem_mutex which is being held at this
point, so if this path were was ever called it would deadlock. Thus
conclude the debugging never triggers and rework it into a simple WARN_ON
and leave things as they are.

While here add a note to explain how we could possibly get inconsistent
page pointers.

Link: https://lore.kernel.org/r/20191009160934.3143-16-jgg@ziepe.ca
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Do not race with mlx5_ib_invalidate_range during create and destroy
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:34 +0000 (13:09 -0300)]
RDMA/mlx5: Do not race with mlx5_ib_invalidate_range during create and destroy

For creation, as soon as the umem_odp is created the notifier can be
called, however the underlying MR may not have been setup yet. This would
cause problems if mlx5_ib_invalidate_range() runs. There is some
confusing/ulocked/racy code that might by trying to solve this, but
without locks it isn't going to work right.

Instead trivially solve the problem by short-circuiting the invalidation
if there are not yet any DMA mapped pages. By definition there is nothing
to invalidate in this case.

The create code will have the umem fully setup before anything is DMA
mapped, and npages is fully locked by the umem_mutex.

For destroy, invalidate the entire MR at the HW to stop DMA then DMA unmap
the pages before destroying the MR. This drives npages to zero and
prevents similar racing with invalidate while the MR is undergoing
destruction.

Arguably it would be better if the umem was created after the MR and
destroyed before, but that would require a big rework of the MR code.

Fixes: 6aec21f6a832 ("IB/mlx5: Page faults handling infrastructure")
Link: https://lore.kernel.org/r/20191009160934.3143-15-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Do not store implicit children in the odp_mkeys xarray
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:33 +0000 (13:09 -0300)]
RDMA/mlx5: Do not store implicit children in the odp_mkeys xarray

These mkeys are entirely internal and are never used by the HW for
page fault. They should also never be used by userspace for prefetch.
Simplify & optimize things by not including them in the xarray.

Since the prefetch path can now never see a child mkey there is no need
for the second synchronize_srcu() during imr destroy.

Link: https://lore.kernel.org/r/20191009160934.3143-14-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Rework implicit ODP destroy
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:32 +0000 (13:09 -0300)]
RDMA/mlx5: Rework implicit ODP destroy

Use SRCU in a sensible way by removing all MRs in the implicit tree from
the two xarrays (the update operation), then a synchronize, followed by a
normal single threaded teardown.

This is only a little unusual from the normal pattern as there can still
be some work pending in the unbound wq that may also require a workqueue
flush. This is tracked with a single atomic, consolidating the redundant
existing atomics and wait queue.

For understand-ability the entire ODP implicit create/destroy flow now
largely exists in a single pair of functions within odp.c, with a few
support functions for tearing down an unused child.

Link: https://lore.kernel.org/r/20191009160934.3143-13-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Avoid double lookups on the pagefault path
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:31 +0000 (13:09 -0300)]
RDMA/mlx5: Avoid double lookups on the pagefault path

Now that the locking is simplified combine pagefault_implicit_mr() with
implicit_mr_get_data() so that we sweep over the idx range only once,
and do the single xlt update at the end, after the child umems are
setup.

This avoids double iteration/xa_loads plus the sketchy failure path if the
xa_load() fails.

Link: https://lore.kernel.org/r/20191009160934.3143-12-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Reduce locking in implicit_mr_get_data()
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:30 +0000 (13:09 -0300)]
RDMA/mlx5: Reduce locking in implicit_mr_get_data()

Now that the child MRs are stored in an xarray we can rely on the SRCU
lock to protect the xa_load and use xa_cmpxchg on the slow allocation path
to resolve races with concurrent page fault.

This reduces the scope of the critical section of umem_mutex for implicit
MRs to only cover mlx5_ib_update_xlt, and avoids taking a lock at all if
the child MR is already in the xarray. This makes it consistent with the
normal ODP MR critical section for umem_lock, and the locking approach
used for destroying an unusued implicit child MR.

The MLX5_IB_UPD_XLT_ATOMIC is no longer needed in implicit_get_child_mr()
since it is no longer called with any locks.

Link: https://lore.kernel.org/r/20191009160934.3143-11-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Use an xarray for the children of an implicit ODP
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:29 +0000 (13:09 -0300)]
RDMA/mlx5: Use an xarray for the children of an implicit ODP

Currently the child leaves are stored in the shared interval tree and
every lookup for a child must be done under the interval tree rwsem.

This is further complicated by dropping the rwsem during iteration (ie the
odp_lookup(), odp_next() pattern), which requires a very tricky an
difficult to understand locking scheme with SRCU.

Instead reserve the interval tree for the exclusive use of the mmu
notifier related code in umem_odp.c and give each implicit MR a xarray
containing all the child MRs.

Since the size of each child is 1GB of VA, a 1 level xarray will index 64G
of VA, and a 2 level will index 2TB, making xarray a much better
data structure choice than an interval tree.

The locking properties of xarray will be used in the next patches to
rework the implicit ODP locking scheme into something simpler.

At this point, the xarray is locked by the implicit MR's umem_mutex, and
read can also be locked by the odp_srcu.

Link: https://lore.kernel.org/r/20191009160934.3143-10-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Split implicit handling from pagefault_mr
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:28 +0000 (13:09 -0300)]
RDMA/mlx5: Split implicit handling from pagefault_mr

The single routine has a very confusing scheme to advance to the next
child MR when working on an implicit parent. This scheme can only be used
when working with an implicit parent and must not be triggered when
working on a normal MR.

Re-arrange things by directly putting all the single-MR stuff into one
function and calling it in a loop for the implicit case. Simplify some of
the error handling in the new pagefault_real_mr() to remove unneeded gotos.

Link: https://lore.kernel.org/r/20191009160934.3143-9-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Set the HW IOVA of the child MRs to their place in the tree
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:27 +0000 (13:09 -0300)]
RDMA/mlx5: Set the HW IOVA of the child MRs to their place in the tree

Instead of rewriting all the IOVA's to 0 as things progress down the tree
make the IOVA of the children equal to placement in the tree. This makes
things easier to understand by keeping mmkey.iova == HW configuration.

Link: https://lore.kernel.org/r/20191009160934.3143-8-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Lift implicit_mr_alloc() into the two routines that call it
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:26 +0000 (13:09 -0300)]
RDMA/mlx5: Lift implicit_mr_alloc() into the two routines that call it

This makes the routines easier to understand, particularly with respect
the locking requirements of the entire sequence. The implicit_mr_alloc()
had a lot of ifs specializing it to each of the callers, and only a very
small amount of code was actually shared.

Following patches will cause the flow in the two functions to diverge
further.

Link: https://lore.kernel.org/r/20191009160934.3143-7-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Rework implicit_mr_get_data
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:25 +0000 (13:09 -0300)]
RDMA/mlx5: Rework implicit_mr_get_data

This function is intended to loop across each MTT chunk in the implicit
parent that intersects the range [io_virt, io_virt+bnct).  But it is has a
confusing construction, so:

- Consistently use imr and odp_imr to refer to the implicit parent
  to avoid confusion with the normal mr and odp of the child
- Directly compute the inclusive start/end indexes by shifting. This is
  clearer to understand the intent and avoids any errors from unaligned
  values of addr
- Iterate directly over the range of MTT indexes, do not make a loop
  out of goto
- Follow 'success oriented flow', with goto error unwind
- Directly calculate the range of idx's that need update_xlt
- Ensure that any leaf MR added to the interval tree always results in an
  update to the XLT

Link: https://lore.kernel.org/r/20191009160934.3143-6-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Delete struct mlx5_priv->mkey_table
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:24 +0000 (13:09 -0300)]
RDMA/mlx5: Delete struct mlx5_priv->mkey_table

No users are left, delete it.

Link: https://lore.kernel.org/r/20191009160934.3143-5-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Use a dedicated mkey xarray for ODP
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:23 +0000 (13:09 -0300)]
RDMA/mlx5: Use a dedicated mkey xarray for ODP

There is a per device xarray storing mkeys that is used to store every
mkey in the system. However, this xarray is now only read by ODP for
certain ODP designated MRs (ODP, implicit ODP, MW, DEVX_INDIRECT).

Create an xarray only for use by ODP, that only contains ODP related
MKeys. This xarray is protected by SRCU and all erases are protected by a
synchronize.

This improves performance:

 - All MRs in the odp_mkeys xarray are ODP MRs, so some tests for is_odp()
   can be deleted. The xarray will also consume fewer nodes.

 - normal MR's are never mixed with ODP MRs in a SRCU data structure so
   performance sucking synchronize_srcu() on every MR destruction is not
   needed.

 - No smp_load_acquire(live) and xa_load() double barrier on read

Due to the SRCU locking scheme care must be taken with the placement of
the xa_store(). Once it completes the MR is immediately visible to other
threads and only through a xa_erase() & synchronize_srcu() cycle could it
be destroyed.

Link: https://lore.kernel.org/r/20191009160934.3143-4-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Split sig_err MR data into its own xarray
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:22 +0000 (13:09 -0300)]
RDMA/mlx5: Split sig_err MR data into its own xarray

The locking model for signature is completely different than ODP, do not
share the same xarray that relies on SRCU locking to support ODP.

Simply store the active mlx5_core_sig_ctx's in an xarray when signature
MRs are created and rely on trivial xarray locking to serialize
everything.

The overhead of storing only a handful of SIG related MRs is going to be
much less than an xarray full of every mkey.

Link: https://lore.kernel.org/r/20191009160934.3143-3-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/mlx5: Use SRCU properly in ODP prefetch
Jason Gunthorpe [Wed, 9 Oct 2019 16:09:21 +0000 (13:09 -0300)]
RDMA/mlx5: Use SRCU properly in ODP prefetch

When working with SRCU protected xarrays the xarray itself should be the
SRCU 'update' point. Instead prefetch is using live as the SRCU update
point and this prevents switching the locking design to use the xarray
instead.

To solve this the prefetch must only read from the xarray once, and hold
on to the actual MR pointer for the duration of the async
operation. Incrementing num_pending_prefetch delays destruction of the MR,
so it is suitable.

Prefetch calls directly to the pagefault_mr using the MR pointer and only
does a single xarray lookup.

All the testing if a MR is prefetchable or not is now done only in the
prefetch code and removed from the pagefault critical path.

Link: https://lore.kernel.org/r/20191009160934.3143-2-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoMerge tag 'v5.4-rc5' into rdma.git for-next
Jason Gunthorpe [Mon, 28 Oct 2019 19:36:29 +0000 (16:36 -0300)]
Merge tag 'v5.4-rc5' into rdma.git for-next

Linux 5.4-rc5

For dependencies in the next patches

Conflict resolved by keeping the delete of the unlock.

Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/vmw_pvrdma: Use resource ids from physical device if available
Bryan Tan [Mon, 28 Oct 2019 18:14:52 +0000 (18:14 +0000)]
RDMA/vmw_pvrdma: Use resource ids from physical device if available

This change allows the RDMA stack to use physical resource numbers if they
are passed up from the device. This is accomplished by separating the
concept of the QP number from the QP handle. Previously, the two were the
same, as the QP number was exposed to the guest and also used to reference
a virtual QP in the device backend.

With physical resource numbers exposed, the QP number given to the guest
is the number assigned from the physical HCA's QP, while the QP handle is
still the internal handle used to reference a virtual QP. Regardless of
whether the device is exposing physical ids, the driver will still try to
pick up the QP handle from the backend if possible. The MR keys exposed to
the guest will also be the MR keys created by the physical HCA, instead of
virtual MR keys. The distinction between handle and keys is already
present for MRs so there is no need to do anything special here.

A new version of the create QP response has been added to the device API
to pass up the QP number and handle. The driver will also report these to
userspace in the udata response if userspace supports it or not create the
queuepair if not. I also had to do a refactor of the destroy qp code to
reuse it if we fail to copy to userspace.

Link: https://lore.kernel.org/r/20191028181444.19448-1-aditr@vmware.com
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Adit Ranadive <aditr@vmware.com>
Signed-off-by: Bryan Tan <bryantan@vmware.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/core: Set DMA parameters correctly
Bart Van Assche [Fri, 25 Oct 2019 22:58:30 +0000 (15:58 -0700)]
RDMA/core: Set DMA parameters correctly

The dma_set_max_seg_size() call in setup_dma_device() does not have any
effect since device->dev.dma_parms is NULL. Fix this by initializing
device->dev.dma_parms first.

Link: https://lore.kernel.org/r/20191025225830.257535-5-bvanassche@acm.org
Fixes: d10bcf947a3e ("RDMA/umem: Combine contiguous PAGE_SIZE regions in SGEs")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/siw: Increase DMA max_segment_size parameter
Bart Van Assche [Fri, 25 Oct 2019 22:58:29 +0000 (15:58 -0700)]
RDMA/siw: Increase DMA max_segment_size parameter

Increase the DMA max_segment_size parameter from 64 KB to 2 GB.

Link: https://lore.kernel.org/r/20191025225830.257535-4-bvanassche@acm.org
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/rxe: Increase DMA max_segment_size parameter
Bart Van Assche [Fri, 25 Oct 2019 22:58:28 +0000 (15:58 -0700)]
RDMA/rxe: Increase DMA max_segment_size parameter

Increase the DMA max_segment_size parameter from 64 KB to 2 GB.

Link: https://lore.kernel.org/r/20191025225830.257535-3-bvanassche@acm.org
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/core: Fix ib_dma_max_seg_size()
Bart Van Assche [Fri, 25 Oct 2019 22:58:27 +0000 (15:58 -0700)]
RDMA/core: Fix ib_dma_max_seg_size()

If dev->dma_device->params == NULL then the maximum DMA segment size is 64
KB. See also the dma_get_max_seg_size() implementation. This patch fixes
the following kernel warning:

  DMA-API: infiniband rxe0: mapping sg segment longer than device claims to support [len=126976] [max=65536]
  WARNING: CPU: 4 PID: 4848 at kernel/dma/debug.c:1220 debug_dma_map_sg+0x3d9/0x450
  RIP: 0010:debug_dma_map_sg+0x3d9/0x450
  Call Trace:
   srp_queuecommand+0x626/0x18d0 [ib_srp]
   scsi_queue_rq+0xd02/0x13e0 [scsi_mod]
   __blk_mq_try_issue_directly+0x2b3/0x3f0
   blk_mq_request_issue_directly+0xac/0xf0
   blk_insert_cloned_request+0xdf/0x170
   dm_mq_queue_rq+0x43d/0x830 [dm_mod]
   __blk_mq_try_issue_directly+0x2b3/0x3f0
   blk_mq_request_issue_directly+0xac/0xf0
   blk_mq_try_issue_list_directly+0xb8/0x170
   blk_mq_sched_insert_requests+0x23c/0x3b0
   blk_mq_flush_plug_list+0x529/0x730
   blk_flush_plug_list+0x21f/0x260
   blk_mq_make_request+0x56b/0xf20
   generic_make_request+0x196/0x660
   submit_bio+0xae/0x290
   blkdev_direct_IO+0x822/0x900
   generic_file_direct_write+0x110/0x200
   __generic_file_write_iter+0x124/0x2a0
   blkdev_write_iter+0x168/0x270
   aio_write+0x1c4/0x310
   io_submit_one+0x971/0x1390
   __x64_sys_io_submit+0x12a/0x390
   do_syscall_64+0x6f/0x2e0
   entry_SYSCALL_64_after_hwframe+0x49/0xbe

Link: https://lore.kernel.org/r/20191025225830.257535-2-bvanassche@acm.org
Cc: <stable@vger.kernel.org>
Fixes: 0b5cb3300ae5 ("RDMA/srp: Increase max_segment_size")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/siw: Fix post_recv QP state locking
Bernard Metzler [Fri, 25 Oct 2019 14:29:03 +0000 (16:29 +0200)]
RDMA/siw: Fix post_recv QP state locking

Do not release qp state lock if not previously acquired.

Fixes: cf049bb31f71 ("RDMA/siw: Fix SQ/RQ drain logic")
Link: https://lore.kernel.org/r/20191025142903.20625-1-bmt@zurich.ibm.com
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/iw_cxgb4: Report correct port speed/width
Potnuri Bharat Teja [Fri, 25 Oct 2019 10:57:02 +0000 (16:27 +0530)]
RDMA/iw_cxgb4: Report correct port speed/width

Query speed/width from corresponding netdev.

Link: https://lore.kernel.org/r/1572001022-4533-1-git-send-email-bharat@chelsio.com
Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/qedr: Fix memory leak in user qp and mr
Michal Kalderon [Sun, 27 Oct 2019 20:04:51 +0000 (22:04 +0200)]
RDMA/qedr: Fix memory leak in user qp and mr

User QPs pbl's weren't freed properly.
MR pbls weren't freed properly.

Fixes: e0290cce6ac0 ("qedr: Add support for memory registeration verbs")
Link: https://lore.kernel.org/r/20191027200451.28187-5-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/qedr: Fix synchronization methods and memory leaks in qedr
Michal Kalderon [Sun, 27 Oct 2019 20:04:50 +0000 (22:04 +0200)]
RDMA/qedr: Fix synchronization methods and memory leaks in qedr

Re-design of the iWARP CM related objects reference counting and
synchronization methods, to ensure operations are synchronized correctly
and that memory allocated for "ep" is properly released. Also makes sure
QP memory is not released before ep is finished accessing it.

Where as the QP object is created/destroyed by external operations, the ep
is created/destroyed by internal operations and represents the tcp
connection associated with the QP.

QP destruction flow:
- needs to wait for ep establishment to complete (either successfully or
  with error)
- needs to wait for ep disconnect to be fully posted to avoid a race
  condition of disconnect being called after reset.
- both the operations above don't always happen, so we use atomic flags to
  indicate whether the qp destruction flow needs to wait for these
  completions or not, if the destroy is called before these operations
  began, the flows will check the flags and not execute them ( connect /
  disconnect).

We use completion structure for waiting for the completions mentioned
above.

The QP refcnt was modified to kref object.  The EP has a kref added to it
to handle additional worker thread accessing it.

Memory Leaks - https://www.spinics.net/lists/linux-rdma/msg83762.html

Concurrency not managed correctly -
https://www.spinics.net/lists/linux-rdma/msg67949.html

Fixes: de0089e692a9 ("RDMA/qedr: Add iWARP connection management qp related callbacks")
Link: https://lore.kernel.org/r/20191027200451.28187-4-michal.kalderon@marvell.com
Reported-by: Chuck Lever <chuck.lever@oracle.com>
Reported-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/qedr: Fix qpids xarray api used
Michal Kalderon [Sun, 27 Oct 2019 20:04:49 +0000 (22:04 +0200)]
RDMA/qedr: Fix qpids xarray api used

The qpids xarray isn't accessed from irq context and therefore there
is no need to use the xa_XXX_irq version of the apis.
Remove the _irq.

Fixes: b6014f9e5f39 ("qedr: Convert qpidr to XArray")
Link: https://lore.kernel.org/r/20191027200451.28187-3-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/qedr: Fix srqs xarray initialization
Michal Kalderon [Sun, 27 Oct 2019 20:04:48 +0000 (22:04 +0200)]
RDMA/qedr: Fix srqs xarray initialization

There was a missing initialization for the srqs xarray.
SRQs xarray can also be called from irq context when searching
for an element and uses the xa_XXX_irq apis, therefore should
be initialized with IRQ flags.

Fixes: 9fd15987ed27 ("qedr: Convert srqidr to XArray")
Link: https://lore.kernel.org/r/20191027200451.28187-2-michal.kalderon@marvell.com
Signed-off-by: Ariel Elior <ariel.elior@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agordma: Remove nes ABI header
Jason Gunthorpe [Thu, 24 Oct 2019 13:51:03 +0000 (13:51 +0000)]
rdma: Remove nes ABI header

This was missed when nes was removed.

Fixes: 2d3c72ed5041 ("rdma: Remove nes")
Link: https://lore.kernel.org/r/20191024135059.GA20084@ziepe.ca
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/hns: Fix memory leak on 'context' on error return path
Colin Ian King [Thu, 24 Oct 2019 13:10:34 +0000 (14:10 +0100)]
RDMA/hns: Fix memory leak on 'context' on error return path

Currently, the error return path when the call to function
dev->dfx->query_cqc_info fails will leak object 'context'. Fix this by
making the error return path via 'err' return return codes rather than
-EMSGSIZE, set ret appropriately for all error return paths and for the
memory leak now return via 'err' rather than just returning without
freeing context.

Link: https://lore.kernel.org/r/20191024131034.19989-1-colin.king@canonical.com
Addresses-Coverity: ("Resource leak")
Fixes: e1c9a0dc2939 ("RDMA/hns: Dump detailed driver-specific CQ")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/hns: Bugfix for qpc/cqc timer configuration
Yangyang Li [Thu, 24 Oct 2019 09:21:57 +0000 (17:21 +0800)]
RDMA/hns: Bugfix for qpc/cqc timer configuration

qpc/cqc timer entry size needs one page, but currently they are fixedly
configured to 4096, which is not appropriate in 64K page scenarios. So
they should be modified to PAGE_SIZE.

Fixes: 0e40dc2f70cd ("RDMA/hns: Add timer allocation support for hip08")
Link: https://lore.kernel.org/r/1571908917-16220-3-git-send-email-liweihang@hisilicon.com
Signed-off-by: Yangyang Li <liyangyang20@huawei.com>
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/hns: Fix to support 64K page for srq
Lijun Ou [Thu, 24 Oct 2019 09:21:56 +0000 (17:21 +0800)]
RDMA/hns: Fix to support 64K page for srq

SRQ's page size configuration of BA and buffer should depend on current
PAGE_SHIFT, or it can't work in scenario of 64K page.

Fixes: c7bcb13442e1 ("RDMA/hns: Add SRQ support for hip08 kernel mode")
Link: https://lore.kernel.org/r/1571908917-16220-2-git-send-email-liweihang@hisilicon.com
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/srpt: Fix TPG creation
Bart Van Assche [Wed, 23 Oct 2019 20:41:06 +0000 (13:41 -0700)]
RDMA/srpt: Fix TPG creation

Unlike the iSCSI target driver, for the SRP target driver it is sufficient
if a single TPG can be associated with each RDMA port name. However, users
started associating multiple TPGs with RDMA port names. Support this by
converting the single TPG in struct srpt_port_id into a list. This patch
fixes the following list corruption issue:

 list_add corruption. prev->next should be next (ffffffffc0a080c0), but was ffffa08a994ce6f0. (prev=ffffa08a994ce6f0).
 WARNING: CPU: 2 PID: 2597 at lib/list_debug.c:28 __list_add_valid+0x6a/0x70
 CPU: 2 PID: 2597 Comm: targetcli Not tainted 5.4.0-rc1.3bfa3c9602a7 #1
 RIP: 0010:__list_add_valid+0x6a/0x70
 Call Trace:
  core_tpg_register+0x116/0x200 [target_core_mod]
  srpt_make_tpg+0x3f/0x60 [ib_srpt]
  target_fabric_make_tpg+0x41/0x290 [target_core_mod]
  configfs_mkdir+0x158/0x3e0
  vfs_mkdir+0x108/0x1a0
  do_mkdirat+0x77/0xe0
  do_syscall_64+0x55/0x1d0
  entry_SYSCALL_64_after_hwframe+0x44/0xa9

Link: https://lore.kernel.org/r/20191023204106.23326-1-bvanassche@acm.org
Reported-by: Honggang LI <honli@redhat.com>
Fixes: a42d985bd5b2 ("ib_srpt: Initial SRP Target merge for v3.3-rc1")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Acked-by: Honggang Li <honli@redhat.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/hns: Delete BITS_PER_BYTE redefinition
Leon Romanovsky [Wed, 23 Oct 2019 05:42:39 +0000 (08:42 +0300)]
RDMA/hns: Delete BITS_PER_BYTE redefinition

HNS redefined available in bits.h define and didn't use it, we can safely
delete it.

Link: https://lore.kernel.org/r/20191023054239.31648-1-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agoRDMA/hns: Prevent undefined behavior in hns_roce_set_user_sq_size()
Jason Gunthorpe [Sat, 8 Jun 2019 09:25:14 +0000 (12:25 +0300)]
RDMA/hns: Prevent undefined behavior in hns_roce_set_user_sq_size()

The "ucmd->log_sq_bb_count" variable is a user controlled variable in the
0-255 range.  If we shift more than then number of bits in an int then
it's undefined behavior (it shift wraps), and potentially the int could
become negative.

Fixes: 9a4435375cd1 ("IB/hns: Add driver files for hns RoCE driver")
Link: https://lore.kernel.org/r/20190608092514.GC28890@mwanda
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>
5 years agoRDMA/cm: Update copyright together with SPDX tag
Leon Romanovsky [Sun, 20 Oct 2019 07:15:56 +0000 (10:15 +0300)]
RDMA/cm: Update copyright together with SPDX tag

Add Mellanox to lust of copyright holders and replace copyright
boilerplate with relevant SPDX tag.

Link: https://lore.kernel.org/r/20191020071559.9743-4-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>
5 years agoRDMA/cm: Use specific keyword to check define
Leon Romanovsky [Sun, 20 Oct 2019 07:15:55 +0000 (10:15 +0300)]
RDMA/cm: Use specific keyword to check define

There is a specific define keyword to check if define exists or not,
let's use it instead of open-coded variant.

Link: https://lore.kernel.org/r/20191020071559.9743-3-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>
5 years agoRDMA/cm: Delete unused cm_is_active_peer function
Leon Romanovsky [Sun, 20 Oct 2019 07:15:54 +0000 (10:15 +0300)]
RDMA/cm: Delete unused cm_is_active_peer function

Function cm_is_active_peer is not used, delete it.

Link: https://lore.kernel.org/r/20191020071559.9743-2-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>
5 years agoLinux 5.4-rc5 v5.4-rc5
Linus Torvalds [Sun, 27 Oct 2019 17:19:19 +0000 (13:19 -0400)]
Linux 5.4-rc5

5 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 27 Oct 2019 11:14:40 +0000 (07:14 -0400)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:
 "Two fixes for the VMWare guest support:

   - Unbreak VMWare platform detection which got wreckaged by converting
     an integer constant to a string constant.

   - Fix the clang build of the VMWAre hypercall by explicitely
     specifying the ouput register for INL instead of using the short
     form"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/cpu/vmware: Fix platform detection VMWARE_PORT macro
  x86/cpu/vmware: Use the full form of INL in VMWARE_HYPERCALL, for clang/llvm

5 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 27 Oct 2019 11:04:22 +0000 (07:04 -0400)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Thomas Gleixner:
 "A small set of fixes for time(keeping):

   - Add a missing include to prevent compiler warnings.

   - Make the VDSO implementation of clock_getres() POSIX compliant
     again. A recent change dropped the NULL pointer guard which is
     required as NULL is a valid pointer value for this function.

   - Fix two function documentation typos"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  posix-cpu-timers: Fix two trivial comments
  timers/sched_clock: Include local timekeeping.h for missing declarations
  lib/vdso: Make clock_getres() POSIX compliant again

5 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 27 Oct 2019 10:59:34 +0000 (06:59 -0400)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Thomas Gleixner:
 "A set of perf fixes:

  kernel:

   - Unbreak the tracking of auxiliary buffer allocations which got
     imbalanced causing recource limit failures.

   - Fix the fallout of splitting of ToPA entries which missed to shift
     the base entry PA correctly.

   - Use the correct context to lookup the AUX event when unmapping the
     associated AUX buffer so the event can be stopped and the buffer
     reference dropped.

  tools:

   - Fix buildiid-cache mode setting in copyfile_mode_ns() when copying
     /proc/kcore

   - Fix freeing id arrays in the event list so the correct event is
     closed.

   - Sync sched.h anc kvm.h headers with the kernel sources.

   - Link jvmti against tools/lib/ctype.o to have weak strlcpy().

   - Fix multiple memory and file descriptor leaks, found by coverity in
     perf annotate.

   - Fix leaks in error handling paths in 'perf c2c', 'perf kmem', found
     by a static analysis tool"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/aux: Fix AUX output stopping
  perf/aux: Fix tracking of auxiliary trace buffer allocation
  perf/x86/intel/pt: Fix base for single entry topa
  perf kmem: Fix memory leak in compact_gfp_flags()
  tools headers UAPI: Sync sched.h with the kernel
  tools headers kvm: Sync kvm.h headers with the kernel sources
  tools headers kvm: Sync kvm headers with the kernel sources
  tools headers kvm: Sync kvm headers with the kernel sources
  perf c2c: Fix memory leak in build_cl_output()
  perf tools: Fix mode setting in copyfile_mode_ns()
  perf annotate: Fix multiple memory and file descriptor leaks
  perf tools: Fix resource leak of closedir() on the error paths
  perf evlist: Fix fix for freed id arrays
  perf jvmti: Link against tools/lib/ctype.h to have weak strlcpy()

5 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 27 Oct 2019 10:55:55 +0000 (06:55 -0400)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Thomas Gleixner:
 "Two fixes for interrupt controller drivers:

   - Skip IRQ_M_EXT entries in the device tree when initializing the
     RISCV PLIC controller to avoid a double init attempt.

   - Use the correct ITS list when issuing the VMOVP synchronization
     command so the operation works only on the ITS instances which are
     associated to a VM"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/sifive-plic: Skip contexts except supervisor in plic_init()
  irqchip/gic-v3-its: Use the exact ITSList for VMOVP

5 years agoMerge tag '5.4-rc5-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sun, 27 Oct 2019 10:41:52 +0000 (06:41 -0400)]
Merge tag '5.4-rc5-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Seven cifs/smb3 fixes, including three for stable"

* tag '5.4-rc5-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: Fix cifsInodeInfo lock_sem deadlock when reconnect occurs
  CIFS: Fix use after free of file info structures
  CIFS: Fix retry mid list corruption on reconnects
  cifs: Fix missed free operations
  CIFS: avoid using MID 0xFFFF
  cifs: clarify comment about timestamp granularity for old servers
  cifs: Handle -EINPROGRESS only when noblockcnt is set

5 years agoMerge tag 'riscv/for-v5.4-rc5-b' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 27 Oct 2019 10:36:57 +0000 (06:36 -0400)]
Merge tag 'riscv/for-v5.4-rc5-b' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Paul Walmsley:
 "Several minor fixes and cleanups for v5.4-rc5:

   - Three build fixes for various SPARSEMEM-related kernel
     configurations

   - Two cleanup patches for the kernel bug and breakpoint trap handler
     code"

* tag 'riscv/for-v5.4-rc5-b' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: cleanup do_trap_break
  riscv: cleanup <asm/bug.h>
  riscv: Fix undefined reference to vmemmap_populate_basepages
  riscv: Fix implicit declaration of 'page_to_section'
  riscv: fix fs/proc/kcore.c compilation with sparsemem enabled

5 years agoMerge tag 'mips_fixes_5.4_3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
Linus Torvalds [Sat, 26 Oct 2019 23:43:12 +0000 (19:43 -0400)]
Merge tag 'mips_fixes_5.4_3' of git://git./linux/kernel/git/mips/linux

Pull MIPS fixes from Paul Burton:
 "A few MIPS fixes:

   - Fix VDSO time-related function behavior for systems where we need
     to fall back to syscalls, but were instead returning bogus results.

   - A fix to TLB exception handlers for Cavium Octeon systems where
     they would inadvertently clobber the $1/$at register.

   - A build fix for bcm63xx configurations.

   - Switch to using my @kernel.org email address"

* tag 'mips_fixes_5.4_3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  MIPS: tlbex: Fix build_restore_pagemask KScratch restore
  MIPS: bmips: mark exception vectors as char arrays
  mips: vdso: Fix __arch_get_hw_counter()
  MAINTAINERS: Use @kernel.org address for Paul Burton

5 years agoMerge tag 'tty-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sat, 26 Oct 2019 20:40:04 +0000 (16:40 -0400)]
Merge tag 'tty-5.4-rc5' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial driver fix from Greg KH:
 "Here is a single tty/serial driver fix for 5.4-rc5 that resolves a
  reported issue.

  It has been in linux-next for a while with no problems"

* tag 'tty-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  8250-men-mcb: fix error checking when get_num_ports returns -ENODEV

5 years agoMerge tag 'staging-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sat, 26 Oct 2019 20:36:47 +0000 (16:36 -0400)]
Merge tag 'staging-5.4-rc5' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fix from Greg KH:
 "Here is a single staging driver fix, for the wlan-ng driver, that
  resolves a reported issue.

  It is been in linux-next for a while with no reported issues"

* tag 'staging-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: wlan-ng: fix exit return when sme->key_idx >= NUM_WEPKEYS

5 years agoMerge tag 'driver-core-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 26 Oct 2019 19:23:08 +0000 (15:23 -0400)]
Merge tag 'driver-core-5.4-rc5' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core fix from Greg KH:
 "Here is a single sysfs fix for 5.4-rc5.

  It resolves an error if you actually try to use the __BIN_ATTR_WO()
  macro, seems I never tested it properly before :(

  This has been in linux-next for a while with no reported issues"

* tag 'driver-core-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  sysfs: Fixes __BIN_ATTR_WO() macro

5 years agoMerge tag 'char-misc-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sat, 26 Oct 2019 19:17:54 +0000 (15:17 -0400)]
Merge tag 'char-misc-5.4-rc5' of git://git./linux/kernel/git/gregkh/char-misc

Pull binder fix from Greg KH:
 "This is a single binder fix to resolve a reported issue by Jann. It's
  been in linux-next for a while with no reported issues"

* tag 'char-misc-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  binder: Don't modify VMA bounds in ->mmap handler

5 years agoMerge tag 'usb-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sat, 26 Oct 2019 19:14:55 +0000 (15:14 -0400)]
Merge tag 'usb-5.4-rc5' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are a number of small USB driver fixes for 5.4-rc5.

  More "fun" with some of the misc USB drivers as found by syzbot, and
  there are a number of other small bugfixes in here for reported
  issues.

  All have been in linux-next for a while with no reported issues"

* tag 'usb-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: cdns3: Error out if USB_DR_MODE_UNKNOWN in cdns3_core_init_role()
  USB: ldusb: fix read info leaks
  USB: serial: ti_usb_3410_5052: clean up serial data access
  USB: serial: ti_usb_3410_5052: fix port-close races
  USB: usblp: fix use-after-free on disconnect
  usb: udc: lpc32xx: fix bad bit shift operation
  usb: cdns3: Fix dequeue implementation.
  USB: legousbtower: fix a signedness bug in tower_probe()
  USB: legousbtower: fix memleak on disconnect
  USB: ldusb: fix memleak on disconnect

5 years agoMerge branch 'i2c/for-current-fixed' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 26 Oct 2019 19:06:58 +0000 (15:06 -0400)]
Merge branch 'i2c/for-current-fixed' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "A few driver fixes for the I2C subsystem"

* 'i2c/for-current-fixed' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: stm32f7: remove warning when compiling with W=1
  i2c: stm32f7: fix a race in slave mode with arbitration loss irq
  i2c: stm32f7: fix first byte to send in slave mode
  i2c: mt65xx: fix NULL ptr dereference
  i2c: aspeed: fix master pending state handling

5 years agoMerge tag 'for-linus-2019-10-26' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 26 Oct 2019 18:59:51 +0000 (14:59 -0400)]
Merge tag 'for-linus-2019-10-26' of git://git.kernel.dk/linux-block

Pull block and io_uring fixes from Jens Axboe:
 "A bit bigger than usual at this point in time, mostly due to some good
  bug hunting work by Pavel that resulted in three io_uring fixes from
  him and two from me. Anyway, this pull request contains:

   - Revert of the submit-and-wait optimization for io_uring, it can't
     always be done safely. It depends on commands always making
     progress on their own, which isn't necessarily the case outside of
     strict file IO. (me)

   - Series of two patches from me and three from Pavel, fixing issues
     with shared data and sequencing for io_uring.

   - Lastly, two timeout sequence fixes for io_uring (zhangyi)

   - Two nbd patches fixing races (Josef)

   - libahci regulator_get_optional() fix (Mark)"

* tag 'for-linus-2019-10-26' of git://git.kernel.dk/linux-block:
  nbd: verify socket is supported during setup
  ata: libahci_platform: Fix regulator_get_optional() misuse
  nbd: handle racing with error'ed out commands
  nbd: protect cmd->status with cmd->lock
  io_uring: fix bad inflight accounting for SETUP_IOPOLL|SETUP_SQTHREAD
  io_uring: used cached copies of sq->dropped and cq->overflow
  io_uring: Fix race for sqes with userspace
  io_uring: Fix broken links with offloading
  io_uring: Fix corrupted user_data
  io_uring: correct timeout req sequence when inserting a new entry
  io_uring : correct timeout req sequence when waiting timeout
  io_uring: revert "io_uring: optimize submit_and_wait API"

5 years agoMerge tag 's390-5.4-5' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Sat, 26 Oct 2019 10:35:46 +0000 (06:35 -0400)]
Merge tag 's390-5.4-5' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Vasily Gorbik:

 - Add R_390_GLOB_DAT relocation type support. This fixes boot problem
   on linux-next.

 - Fix memory leak in zcrypt

* tag 's390-5.4-5' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/kaslr: add support for R_390_GLOB_DAT relocation type
  s390/zcrypt: fix memleak at release

5 years agoMerge tag 'for-linus-5.4-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 26 Oct 2019 10:32:12 +0000 (06:32 -0400)]
Merge tag 'for-linus-5.4-rc5-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixlet from Juergen Gross:
 "Just one patch for issuing a deprecation warning for 32-bit Xen pv
  guests"

* tag 'for-linus-5.4-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen: issue deprecation warning for 32-bit pv guest

5 years agoMerge tag 'dma-mapping-5.4-2' of git://git.infradead.org/users/hch/dma-mapping
Linus Torvalds [Sat, 26 Oct 2019 10:29:04 +0000 (06:29 -0400)]
Merge tag 'dma-mapping-5.4-2' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fix from Christoph Hellwig:
 "Fix a regression in the intel-iommu get_required_mask conversion
  (Arvind Sankar)"

* tag 'dma-mapping-5.4-2' of git://git.infradead.org/users/hch/dma-mapping:
  iommu/vt-d: Return the correct dma mask when we are bypassing the IOMMU

5 years agoMerge tag 'dax-fix-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm...
Linus Torvalds [Sat, 26 Oct 2019 10:26:04 +0000 (06:26 -0400)]
Merge tag 'dax-fix-5.4-rc5' of git://git./linux/kernel/git/nvdimm/nvdimm

Pull dax fix from Dan Williams:
 "Fix a performance regression that followed from a fix to the
  conversion of the fsdax implementation to the xarray. v5.3 users
  report that they stop seeing huge page mappings on an application +
  filesystem layout that was seeing huge pages previously on v5.2"

* tag 'dax-fix-5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  fs/dax: Fix pmd vs pte conflict detection

5 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 26 Oct 2019 00:11:33 +0000 (20:11 -0400)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Nine changes, eight to drivers (qla2xxx, hpsa, lpfc, alua, ch,
  53c710[x2], target) and one core change that tries to close a race
  between sysfs delete and module removal"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: lpfc: remove left-over BUILD_NVME defines
  scsi: core: try to get module before removing device
  scsi: hpsa: add missing hunks in reset-patch
  scsi: target: core: Do not overwrite CDB byte 1
  scsi: ch: Make it possible to open a ch device multiple times again
  scsi: fix kconfig dependency warning related to 53C700_LE_ON_BE
  scsi: sni_53c710: fix compilation error
  scsi: scsi_dh_alua: handle RTPG sense code correctly during state transitions
  scsi: qla2xxx: fix a potential NULL pointer dereference

5 years agoriscv: cleanup do_trap_break
Christoph Hellwig [Thu, 17 Oct 2019 17:37:30 +0000 (19:37 +0200)]
riscv: cleanup do_trap_break

If we always compile the get_break_insn_length inline function we can
remove the ifdefs and let dead code elimination take care of the warn
branch that is now unreadable because the report_bug stub always
returns BUG_TRAP_TYPE_BUG.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
5 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Fri, 25 Oct 2019 21:31:53 +0000 (17:31 -0400)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fix from Dmitry Torokhov:
 "A fix for st1232 driver to properly report coordinates for 2nd and
  subsequent fingers when more than one is on the surface"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: st1232 - fix reporting multitouch coordinates

5 years agonbd: verify socket is supported during setup
Mike Christie [Thu, 17 Oct 2019 21:27:34 +0000 (16:27 -0500)]
nbd: verify socket is supported during setup

nbd requires socket families to support the shutdown method so the nbd
recv workqueue can be woken up from its sock_recvmsg call. If the socket
does not support the callout we will leave recv works running or get hangs
later when the device or module is removed.

This adds a check during socket connection/reconnection to make sure the
socket being passed in supports the needed callout.

Reported-by: syzbot+24c12fa8d218ed26011a@syzkaller.appspotmail.com
Fixes: e9e006f5fcf2 ("nbd: fix max number of supported devs")
Tested-by: Richard W.M. Jones <rjones@redhat.com>
Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoata: libahci_platform: Fix regulator_get_optional() misuse
Mark Brown [Wed, 16 Oct 2019 10:51:05 +0000 (11:51 +0100)]
ata: libahci_platform: Fix regulator_get_optional() misuse

This driver is using regulator_get_optional() to handle all the supplies
that it handles, and only ever enables and disables all supplies en masse
without ever doing any other configuration of the device to handle missing
power. These are clear signs that the API is being misused - it should only
be used for supplies that may be physically absent from the system and in
these cases the hardware usually needs different configuration if the
supply is missing. Instead use normal regualtor_get(), if the supply is
not described in DT then the framework will substitute a dummy regulator in
so no special handling is needed by the consumer driver.

In the case of the PHY regulator the handling in the driver is a hack to
deal with integrated PHYs; the supplies are only optional in the sense
that that there's some confusion in the code about where they're bound to.
From a code point of view they function exactly as normal supplies so can
be treated as such. It'd probably be better to model this by instantiating
a PHY object for integrated PHYs.

Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonbd: handle racing with error'ed out commands
Josef Bacik [Mon, 21 Oct 2019 19:56:28 +0000 (15:56 -0400)]
nbd: handle racing with error'ed out commands

We hit the following warning in production

print_req_error: I/O error, dev nbd0, sector 7213934408 flags 80700
------------[ cut here ]------------
refcount_t: underflow; use-after-free.
WARNING: CPU: 25 PID: 32407 at lib/refcount.c:190 refcount_sub_and_test_checked+0x53/0x60
Workqueue: knbd-recv recv_work [nbd]
RIP: 0010:refcount_sub_and_test_checked+0x53/0x60
Call Trace:
 blk_mq_free_request+0xb7/0xf0
 blk_mq_complete_request+0x62/0xf0
 recv_work+0x29/0xa1 [nbd]
 process_one_work+0x1f5/0x3f0
 worker_thread+0x2d/0x3d0
 ? rescuer_thread+0x340/0x340
 kthread+0x111/0x130
 ? kthread_create_on_node+0x60/0x60
 ret_from_fork+0x1f/0x30
---[ end trace b079c3c67f98bb7c ]---

This was preceded by us timing out everything and shutting down the
sockets for the device.  The problem is we had a request in the queue at
the same time, so we completed the request twice.  This can actually
happen in a lot of cases, we fail to get a ref on our config, we only
have one connection and just error out the command, etc.

Fix this by checking cmd->status in nbd_read_stat.  We only change this
under the cmd->lock, so we are safe to check this here and see if we've
already error'ed this command out, which would indicate that we've
completed it as well.

Reviewed-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonbd: protect cmd->status with cmd->lock
Josef Bacik [Mon, 21 Oct 2019 19:56:27 +0000 (15:56 -0400)]
nbd: protect cmd->status with cmd->lock

We already do this for the most part, except in timeout and clear_req.
For the timeout case we take the lock after we grab a ref on the config,
but that isn't really necessary because we're safe to touch the cmd at
this point, so just move the order around.

For the clear_req cause this is initiated by the user, so again is safe.

Reviewed-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoMerge tag 'modules-for-v5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 25 Oct 2019 20:11:55 +0000 (16:11 -0400)]
Merge tag 'modules-for-v5.4-rc5' of git://git./linux/kernel/git/jeyu/linux

Pull modules fixes from Jessica Yu:

 - Revert __ksymtab_$namespace.$symbol naming scheme back to
   __ksymtab_$symbol, as it was causing issues with depmod.

   Instead, have modpost extract a symbol's namespace from __kstrtabns
   and __ksymtab_strings.

 - Fix `make nsdeps` for out of tree kernel builds (make O=...) caused
   by unescaped '/'.

   Use a different sed delimiter to avoid this problem.

* tag 'modules-for-v5.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/jeyu/linux:
  scripts/nsdeps: use alternative sed delimiter
  symbol namespaces: revert to previous __ksymtab name scheme
  modpost: make updating the symbol namespace explicit
  modpost: delegate updating namespaces to separate function

5 years agoMerge tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
Linus Torvalds [Fri, 25 Oct 2019 20:00:47 +0000 (16:00 -0400)]
Merge tag 'armsoc-fixes' of git://git./linux/kernel/git/soc/soc

Pull ARM SoC fixes from Olof Johansson:
 "A slightly larger set of fixes have accrued in the last two weeks.
  Mostly a collection of the usual smaller fixes:

   - Marvell Armada: USB phy setup issues on Turris Mox

   - Broadcom: GPIO/pinmux DT mapping corrections for Stingray, MMC bus
     width fix for RPi Zero W, GPIO LED removal for RPI CM3. Also some
     maintainer updates.

   - OMAP: Fixlets for display config, interrupt settings for wifi, some
     clock/PM pieces. Also IOMMU regression fix and a ti-sysc
     no-watchdog regression fix.

   - i.MX: A few fixes around PM/settings, some devicetree fixlets and
     catching up with config option changes in DRM

   - Rockchip: RockRro64 misc DT fixups, Hugsun X99 USB-C, Kevin display
     panel settings

  ... and some smaller fixes for Davinci (backlight, McBSP DMA),
  Allwinner (phy regulators, PMU removal on A64, etc)"

* tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc: (42 commits)
  ARM: dts: stm32: relax qspi pins slew-rate for stm32mp157
  MAINTAINERS: Update the Spreadtrum SoC maintainer
  MAINTAINERS: Remove Gregory and Brian for ARCH_BRCMSTB
  ARM: dts: bcm2837-rpi-cm3: Avoid leds-gpio probing issue
  bus: ti-sysc: Fix watchdog quirk handling
  ARM: OMAP2+: Add pdata for OMAP3 ISP IOMMU
  ARM: OMAP2+: Plug in device_enable/idle ops for IOMMUs
  ARM: davinci_all_defconfig: enable GPIO backlight
  ARM: davinci: dm365: Fix McBSP dma_slave_map entry
  ARM: dts: bcm2835-rpi-zero-w: Fix bus-width of sdhci
  ARM: imx_v6_v7_defconfig: Enable CONFIG_DRM_MSM
  arm64: dts: imx8mn: Use correct clock for usdhc's ipg clk
  arm64: dts: imx8mm: Use correct clock for usdhc's ipg clk
  arm64: dts: imx8mq: Use correct clock for usdhc's ipg clk
  ARM: dts: imx7s: Correct GPT's ipg clock source
  ARM: dts: vf610-zii-scu4-aib: Specify 'i2c-mux-idle-disconnect'
  ARM: dts: imx6q-logicpd: Re-Enable SNVS power key
  arm64: dts: lx2160a: Correct CPU core idle state name
  mailmap: Add Simon Arlott (replacement for expired email address)
  arm64: dts: rockchip: Fix override mode for rk3399-kevin panel
  ...

5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 25 Oct 2019 19:52:05 +0000 (15:52 -0400)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fixes from Paolo Bonzini:
 "Bugfixes for ARM, PPC and x86, plus selftest improvements"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: nVMX: Don't leak L1 MMIO regions to L2
  KVM: SVM: Fix potential wrong physical id in avic_handle_ldr_update
  kvm: clear kvmclock MSR on reset
  KVM: x86: fix bugon.cocci warnings
  KVM: VMX: Remove specialized handling of unexpected exit-reasons
  selftests: kvm: fix sync_regs_test with newer gccs
  selftests: kvm: vmx_dirty_log_test: skip the test when VMX is not supported
  selftests: kvm: consolidate VMX support checks
  selftests: kvm: vmx_set_nested_state_test: don't check for VMX support twice
  KVM: Don't shrink/grow vCPU halt_poll_ns if host side polling is disabled
  selftests: kvm: synchronize .gitignore to Makefile
  kvm: x86: Expose RDPID in KVM_GET_SUPPORTED_CPUID
  KVM: arm64: pmu: Reset sample period on overflow handling
  KVM: arm64: pmu: Set the CHAINED attribute before creating the in-kernel event
  arm64: KVM: Handle PMCR_EL0.LC as RES1 on pure AArch64 systems
  KVM: arm64: pmu: Fix cycle counter truncation
  KVM: PPC: Book3S HV: XIVE: Ensure VP isn't already in use

5 years agoMerge tag 'drm-fixes-2019-10-25' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 25 Oct 2019 19:41:14 +0000 (15:41 -0400)]
Merge tag 'drm-fixes-2019-10-25' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Quiet week this week, which I suspect means some people just didn't
  get around to sending me fixes pulls in time. This has 2 komeda and a
  bunch of amdgpu fixes in it:

  komeda:
   - typo fixes
   - flushing pipes fix

  amdgpu:
   - Fix suspend/resume issue related to multi-media engines
   - Fix memory leak in user ptr code related to hmm conversion
   - Fix possible VM faults when allocating page table memory
   - Fix error handling in bo list ioctl"

* tag 'drm-fixes-2019-10-25' of git://anongit.freedesktop.org/drm/drm:
  drm/komeda: Fix typos in komeda_splitter_validate
  drm/komeda: Don't flush inactive pipes
  drm/amdgpu/vce: fix allocation size in enc ring test
  drm/amdgpu: fix error handling in amdgpu_bo_list_create
  drm/amdgpu: fix potential VM faults
  drm/amdgpu: user pages array memory leak fix
  drm/amdgpu/vcn: fix allocation size in enc ring test
  drm/amdgpu/uvd7: fix allocation size in enc ring test (v2)
  drm/amdgpu/uvd6: fix allocation size in enc ring test (v2)

5 years agoMerge tag 'mmc-v5.4-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Fri, 25 Oct 2019 19:25:51 +0000 (15:25 -0400)]
Merge tag 'mmc-v5.4-rc4' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:
 "MMC host fixes:

   - mxs: Fix flags passed to dmaengine_prep_slave_sg

   - cqhci: Add a missing memory barrier

   - sdhci-omap: Fix tuning procedure for temperatures < -20C"

* tag 'mmc-v5.4-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: mxs: fix flags passed to dmaengine_prep_slave_sg
  mmc: cqhci: Commit descriptors before setting the doorbell
  mmc: sdhci-omap: Fix Tuning procedure for temperatures < -20C

5 years agoio_uring: fix bad inflight accounting for SETUP_IOPOLL|SETUP_SQTHREAD
Jens Axboe [Fri, 25 Oct 2019 16:06:15 +0000 (10:06 -0600)]
io_uring: fix bad inflight accounting for SETUP_IOPOLL|SETUP_SQTHREAD

We currently assume that submissions from the sqthread are successful,
and if IO polling is enabled, we use that value for knowing how many
completions to look for. But if we overflowed the CQ ring or some
requests simply got errored and already completed, they won't be
available for polling.

For the case of IO polling and SQTHREAD usage, look at the pending
poll list. If it ever hits empty then we know that we don't have
anymore pollable requests inflight. For that case, simply reset
the inflight count to zero.

Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoio_uring: used cached copies of sq->dropped and cq->overflow
Jens Axboe [Fri, 25 Oct 2019 16:04:25 +0000 (10:04 -0600)]
io_uring: used cached copies of sq->dropped and cq->overflow

We currently use the ring values directly, but that can lead to issues
if the application is malicious and changes these values on our behalf.
Created in-kernel cached versions of them, and just overwrite the user
side when we update them. This is similar to how we treat the sq/cq
ring tail/head updates.

Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoARM: dts: stm32: relax qspi pins slew-rate for stm32mp157
Patrice Chotard [Fri, 25 Oct 2019 13:01:22 +0000 (15:01 +0200)]
ARM: dts: stm32: relax qspi pins slew-rate for stm32mp157

Relax qspi pins slew-rate to minimize peak currents.

Fixes: 844030057339 ("ARM: dts: stm32: add flash nor support on stm32mp157c eval board")

Link: https://lore.kernel.org/r/20191025130122.11407-1-alexandre.torgue@st.com
Signed-off-by: Patrice Chotard <patrice.chotard@st.com>
Signed-off-by: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
5 years agoio_uring: Fix race for sqes with userspace
Pavel Begunkov [Fri, 25 Oct 2019 09:31:31 +0000 (12:31 +0300)]
io_uring: Fix race for sqes with userspace

io_ring_submit() finalises with
1. io_commit_sqring(), which releases sqes to the userspace
2. Then calls to io_queue_link_head(), accessing released head's sqe

Reorder them.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoio_uring: Fix broken links with offloading
Pavel Begunkov [Fri, 25 Oct 2019 09:31:30 +0000 (12:31 +0300)]
io_uring: Fix broken links with offloading

io_sq_thread() processes sqes by 8 without considering links. As a
result, links will be randomely subdivided.

The easiest way to fix it is to call io_get_sqring() inside
io_submit_sqes() as do io_ring_submit().

Downsides:
1. This removes optimisation of not grabbing mm_struct for fixed files
2. It submitting all sqes in one go, without finer-grained sheduling
with cq processing.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoio_uring: Fix corrupted user_data
Pavel Begunkov [Fri, 25 Oct 2019 09:31:29 +0000 (12:31 +0300)]
io_uring: Fix corrupted user_data

There is a bug, where failed linked requests are returned not with
specified @user_data, but with garbage from a kernel stack.

The reason is that io_fail_links() uses req->user_data, which is
uninitialised when called from io_queue_sqe() on fail path.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoxen: issue deprecation warning for 32-bit pv guest
Juergen Gross [Fri, 25 Oct 2019 07:38:58 +0000 (09:38 +0200)]
xen: issue deprecation warning for 32-bit pv guest

Support for the kernel as Xen 32-bit PV guest will soon be removed.
Issue a warning when booted as such.

Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
5 years agoMerge tag 'irqchip-fixes-5.4-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Thomas Gleixner [Fri, 25 Oct 2019 12:25:15 +0000 (14:25 +0200)]
Merge tag 'irqchip-fixes-5.4-2' of git://git./linux/kernel/git/maz/arm-platforms into irq/urgent

Pull the second lot of irqchip updates for 5.4 from Marc Zyngier:

- Sifive PLIC: force driver to skip non-relevant contexts
- GICv4: Don't send VMOVP commands to ITSs that don't have
  this vPE mapped

5 years agoirqchip/sifive-plic: Skip contexts except supervisor in plic_init()
Alan Mikhak [Thu, 24 Oct 2019 16:11:43 +0000 (09:11 -0700)]
irqchip/sifive-plic: Skip contexts except supervisor in plic_init()

Modify plic_init() to skip .dts interrupt contexts other
than supervisor external interrupt.

The .dts entry for plic may specify multiple interrupt contexts.
For example, it may assign two entries IRQ_M_EXT and IRQ_S_EXT,
in that order, to the same interrupt controller. This patch
modifies plic_init() to skip the IRQ_M_EXT context since
IRQ_S_EXT is currently the only supported context.

If IRQ_M_EXT is not skipped, plic_init() will report "handler
already present for context" when it comes across the IRQ_S_EXT
context in the next iteration of its loop.

Without this patch, .dts would have to be edited to replace the
value of IRQ_M_EXT with -1 for it to be skipped.

Signed-off-by: Alan Mikhak <alan.mikhak@sifive.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Paul Walmsley <paul.walmsley@sifive.com> # arch/riscv
Link: https://lkml.kernel.org/r/1571933503-21504-1-git-send-email-alan.mikhak@sifive.com
5 years agoMerge tag 'drm-fixes-5.4-2019-10-23' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Fri, 25 Oct 2019 04:48:53 +0000 (14:48 +1000)]
Merge tag 'drm-fixes-5.4-2019-10-23' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

drm-fixes-5.4-2019-10-23:

amdgpu:
- Fix suspend/resume issue related to multi-media engines
- Fix memory leak in user ptr code related to hmm conversion
- Fix possible VM faults when allocating page table memory
- Fix error handling in bo list ioctl

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191024031809.3155-1-alexander.deucher@amd.com
5 years agoMerge tag 'drm-misc-fixes-2019-10-23' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Fri, 25 Oct 2019 04:45:12 +0000 (14:45 +1000)]
Merge tag 'drm-misc-fixes-2019-10-23' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

Two fixes for komeda, one for typos and one to prevent an hardware issue
when flushing inactive pipes

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20191023112643.evpp6f23mpjwdsn4@gilmour
5 years agocifs: Fix cifsInodeInfo lock_sem deadlock when reconnect occurs
Dave Wysochanski [Wed, 23 Oct 2019 09:02:33 +0000 (05:02 -0400)]
cifs: Fix cifsInodeInfo lock_sem deadlock when reconnect occurs

There's a deadlock that is possible and can easily be seen with
a test where multiple readers open/read/close of the same file
and a disruption occurs causing reconnect.  The deadlock is due
a reader thread inside cifs_strict_readv calling down_read and
obtaining lock_sem, and then after reconnect inside
cifs_reopen_file calling down_read a second time.  If in
between the two down_read calls, a down_write comes from
another process, deadlock occurs.

        CPU0                    CPU1
        ----                    ----
cifs_strict_readv()
 down_read(&cifsi->lock_sem);
                               _cifsFileInfo_put
                                  OR
                               cifs_new_fileinfo
                                down_write(&cifsi->lock_sem);
cifs_reopen_file()
 down_read(&cifsi->lock_sem);

Fix the above by changing all down_write(lock_sem) calls to
down_write_trylock(lock_sem)/msleep() loop, which in turn
makes the second down_read call benign since it will never
block behind the writer while holding lock_sem.

Signed-off-by: Dave Wysochanski <dwysocha@redhat.com>
Suggested-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed--by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
5 years agoCIFS: Fix use after free of file info structures
Pavel Shilovsky [Wed, 23 Oct 2019 22:37:19 +0000 (15:37 -0700)]
CIFS: Fix use after free of file info structures

Currently the code assumes that if a file info entry belongs
to lists of open file handles of an inode and a tcon then
it has non-zero reference. The recent changes broke that
assumption when putting the last reference of the file info.
There may be a situation when a file is being deleted but
nothing prevents another thread to reference it again
and start using it. This happens because we do not hold
the inode list lock while checking the number of references
of the file info structure. Fix this by doing the proper
locking when doing the check.

Fixes: 487317c99477d ("cifs: add spinlock for the openFileList to cifsInodeInfo")
Fixes: cb248819d209d ("cifs: use cifsInodeInfo->open_file_lock while iterating to avoid a panic")
Cc: Stable <stable@vger.kernel.org>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
5 years agoCIFS: Fix retry mid list corruption on reconnects
Pavel Shilovsky [Tue, 22 Oct 2019 15:41:42 +0000 (08:41 -0700)]
CIFS: Fix retry mid list corruption on reconnects

When the client hits reconnect it iterates over the mid
pending queue marking entries for retry and moving them
to a temporary list to issue callbacks later without holding
GlobalMid_Lock. In the same time there is no guarantee that
mids can't be removed from the temporary list or even
freed completely by another thread. It may cause a temporary
list corruption:

[  430.454897] list_del corruption. prev->next should be ffff98d3a8f316c0, but was 2e885cb266355469
[  430.464668] ------------[ cut here ]------------
[  430.466569] kernel BUG at lib/list_debug.c:51!
[  430.468476] invalid opcode: 0000 [#1] SMP PTI
[  430.470286] CPU: 0 PID: 13267 Comm: cifsd Kdump: loaded Not tainted 5.4.0-rc3+ #19
[  430.473472] Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
[  430.475872] RIP: 0010:__list_del_entry_valid.cold+0x31/0x55
...
[  430.510426] Call Trace:
[  430.511500]  cifs_reconnect+0x25e/0x610 [cifs]
[  430.513350]  cifs_readv_from_socket+0x220/0x250 [cifs]
[  430.515464]  cifs_read_from_socket+0x4a/0x70 [cifs]
[  430.517452]  ? try_to_wake_up+0x212/0x650
[  430.519122]  ? cifs_small_buf_get+0x16/0x30 [cifs]
[  430.521086]  ? allocate_buffers+0x66/0x120 [cifs]
[  430.523019]  cifs_demultiplex_thread+0xdc/0xc30 [cifs]
[  430.525116]  kthread+0xfb/0x130
[  430.526421]  ? cifs_handle_standard+0x190/0x190 [cifs]
[  430.528514]  ? kthread_park+0x90/0x90
[  430.530019]  ret_from_fork+0x35/0x40

Fix this by obtaining extra references for mids being retried
and marking them as MID_DELETED which indicates that such a mid
has been dequeued from the pending list.

Also move mid cleanup logic from DeleteMidQEntry to
_cifs_mid_q_entry_release which is called when the last reference
to a particular mid is put. This allows to avoid any use-after-free
of response buffers.

The patch needs to be backported to stable kernels. A stable tag
is not mentioned below because the patch doesn't apply cleanly
to any actively maintained stable kernel.

Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-and-tested-by: David Wysochanski <dwysocha@redhat.com>
Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
5 years agoMerge tag 'devicetree-fixes-for-5.4-2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 24 Oct 2019 22:29:40 +0000 (18:29 -0400)]
Merge tag 'devicetree-fixes-for-5.4-2' of git://git./linux/kernel/git/robh/linux

Pull Devicetree fixes from Rob Herring:
 "A couple more DT fixes for 5.4: fix a ref count, memory leak, and
  Risc-V cpu schema warnings"

* tag 'devicetree-fixes-for-5.4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  of: reserved_mem: add missing of_node_put() for proper ref-counting
  of: unittest: fix memory leak in unittest_data_add
  dt-bindings: riscv: Fix CPU schema errors