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>
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: 6aec21f6a8 ("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>
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>
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>
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>
-----BEGIN PGP SIGNATURE-----
iQFSBAABCAA8FiEEq68RxlopcLEwq+PEeb4+QwBBGIYFAl210Z8eHHRvcnZhbGRz
QGxpbnV4LWZvdW5kYXRpb24ub3JnAAoJEHm+PkMAQRiGv+kIAKRpO7EuDokQL4qp
hxEEaCMJA1T055EMlNU6FVAq/ZbmapzreUyNYiRMpPWKGTWNMkhIcZQfysYeGZz5
y/KRxAiVxlcB+3v3yRmoZd/XoQmhgvJmqD4zhaGI2Utonow4f/SGSEFFZqqs9WND
4HJROjZHgQ4JBxg9Z+QMo0FxbV/DCZpEOUq51N9WJywyyDRb18zotE83stpU+pE2
fjqT7mk0NLrnYXuDRAbFC1Aau9ed4H6LlwLmxaqxq/Pt5Rz7wIKwKL9HIT4Dm/0a
qpani6phhHWL7MwUpa2wkEonFCD03rJFl3DUVJo64Ijh4up5D/jyXQ+GKV2P4WKJ
275Rb5Q=
=WiZZ
-----END PGP SIGNATURE-----
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>
While MR uses live as the SRCU 'update', the MW case uses the xarray
directly, xa_erase() causes the MW to become inaccessible to the pagefault
thread.
Thus whenever a MW is removed from the xarray we must synchronize_srcu()
before freeing it.
This must be done before freeing the mkey as re-use of the mkey while the
pagefault thread is using the stale mkey is undesirable.
Add the missing synchronizes to MW and DEVX indirect mkey and delete the
bogus protection against double destroy in mlx5_core_destroy_mkey()
Fixes: 534fd7aac5 ("IB/mlx5: Manage indirection mkey upon DEVX flow for ODP")
Fixes: 6aec21f6a8 ("IB/mlx5: Page faults handling infrastructure")
Link: https://lore.kernel.org/r/20191001153821.23621-7-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
live is used to signal to the pagefault thread that the MR is initialized
and ready for use. It should be after the umem is assigned and all other
setup is completed. This prevents races (at least) of the form:
CPU0 CPU1
mlx5_ib_alloc_implicit_mr()
implicit_mr_alloc()
live = 1
imr->umem = umem
num_pending_prefetch_inc()
if (live)
atomic_inc(num_pending_prefetch)
atomic_set(num_pending_prefetch,0) // Overwrites other thread's store
Further, live is being used with SRCU as the 'update' in an
acquire/release fashion, so it can not be read and written raw.
Move all live = 1's to after MR initialization is completed and use
smp_store_release/smp_load_acquire() for manipulating it.
Add a missing live = 0 when an implicit MR child is deleted, before
queuing work to do synchronize_srcu().
The barriers in update_odp_mr() were some broken attempt to create a
acquire/release, but were not even applied consistently and missed the
point, delete it as well.
Fixes: 6aec21f6a8 ("IB/mlx5: Page faults handling infrastructure")
Link: https://lore.kernel.org/r/20191001153821.23621-6-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
During destroy setting live = 0 and then synchronize_srcu() prevents
num_pending_prefetch from incrementing, and also, ensures that all work
holding that count is queued on the WQ. Testing before causes races of the
form:
CPU0 CPU1
dereg_mr()
mlx5_ib_advise_mr_prefetch()
srcu_read_lock()
num_pending_prefetch_inc()
if (!live)
live = 0
atomic_read() == 0
// skip flush_workqueue()
atomic_inc()
queue_work();
srcu_read_unlock()
WARN_ON(atomic_read()) // Fails
Swap the order so that the synchronize_srcu() prevents this.
Fixes: a6bc3875f1 ("IB/mlx5: Protect against prefetch of invalid MR")
Link: https://lore.kernel.org/r/20191001153821.23621-5-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
This code is completely broken, the umem of a ODP MR simply cannot be
discarded without a lot more locking, nor can an ODP mkey be blithely
destroyed via destroy_mkey().
Fixes: 6aec21f6a8 ("IB/mlx5: Page faults handling infrastructure")
Link: https://lore.kernel.org/r/20191001153821.23621-2-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
This is more cleanup and consolidation of the hmm APIs and the very
strongly related mmu_notifier interfaces. Many places across the tree
using these interfaces are touched in the process. Beyond that a cleanup
to the page walker API and a few memremap related changes round out the
series:
- General improvement of hmm_range_fault() and related APIs, more
documentation, bug fixes from testing, API simplification &
consolidation, and unused API removal
- Simplify the hmm related kconfigs to HMM_MIRROR and DEVICE_PRIVATE, and
make them internal kconfig selects
- Hoist a lot of code related to mmu notifier attachment out of drivers by
using a refcount get/put attachment idiom and remove the convoluted
mmu_notifier_unregister_no_release() and related APIs.
- General API improvement for the migrate_vma API and revision of its only
user in nouveau
- Annotate mmu_notifiers with lockdep and sleeping region debugging
Two series unrelated to HMM or mmu_notifiers came along due to
dependencies:
- Allow pagemap's memremap_pages family of APIs to work without providing
a struct device
- Make walk_page_range() and related use a constant structure for function
pointers
-----BEGIN PGP SIGNATURE-----
iQIzBAABCgAdFiEEfB7FMLh+8QxL+6i3OG33FX4gmxoFAl1/nnkACgkQOG33FX4g
mxqaRg//c6FqowV1pQlLutvAOAgMdpzfZ9eaaDKngy9RVQxz+k/MmJrdRH/p/mMA
Pq93A1XfwtraGKErHegFXGEDk4XhOustVAVFwvjyXO41dTUdoFVUkti6ftbrl/rS
6CT+X90jlvrwdRY7QBeuo7lxx7z8Qkqbk1O1kc1IOracjKfNJS+y6LTamy6weM3g
tIMHI65PkxpRzN36DV9uCN5dMwFzJ73DWHp1b0acnDIigkl6u5zp6orAJVWRjyQX
nmEd3/IOvdxaubAoAvboNS5CyVb4yS9xshWWMbH6AulKJv3Glca1Aa7QuSpBoN8v
wy4c9+umzqRgzgUJUe1xwN9P49oBNhJpgBSu8MUlgBA4IOc3rDl/Tw0b5KCFVfkH
yHkp8n6MP8VsRrzXTC6Kx0vdjIkAO8SUeylVJczAcVSyHIo6/JUJCVDeFLSTVymh
EGWJ7zX2iRhUbssJ6/izQTTQyCH3YIyZ5QtqByWuX2U7ZrfkqS3/EnBW1Q+j+gPF
Z2yW8iT6k0iENw6s8psE9czexuywa/Lttz94IyNlOQ8rJTiQqB9wLaAvg9hvUk7a
kuspL+JGIZkrL3ouCeO/VA6xnaP+Q7nR8geWBRb8zKGHmtWrb5Gwmt6t+vTnCC2l
olIDebrnnxwfBQhEJ5219W+M1pBpjiTpqK/UdBd92A4+sOOhOD0=
=FRGg
-----END PGP SIGNATURE-----
Merge tag 'for-linus-hmm' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Pull hmm updates from Jason Gunthorpe:
"This is more cleanup and consolidation of the hmm APIs and the very
strongly related mmu_notifier interfaces. Many places across the tree
using these interfaces are touched in the process. Beyond that a
cleanup to the page walker API and a few memremap related changes
round out the series:
- General improvement of hmm_range_fault() and related APIs, more
documentation, bug fixes from testing, API simplification &
consolidation, and unused API removal
- Simplify the hmm related kconfigs to HMM_MIRROR and DEVICE_PRIVATE,
and make them internal kconfig selects
- Hoist a lot of code related to mmu notifier attachment out of
drivers by using a refcount get/put attachment idiom and remove the
convoluted mmu_notifier_unregister_no_release() and related APIs.
- General API improvement for the migrate_vma API and revision of its
only user in nouveau
- Annotate mmu_notifiers with lockdep and sleeping region debugging
Two series unrelated to HMM or mmu_notifiers came along due to
dependencies:
- Allow pagemap's memremap_pages family of APIs to work without
providing a struct device
- Make walk_page_range() and related use a constant structure for
function pointers"
* tag 'for-linus-hmm' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (75 commits)
libnvdimm: Enable unit test infrastructure compile checks
mm, notifier: Catch sleeping/blocking for !blockable
kernel.h: Add non_block_start/end()
drm/radeon: guard against calling an unpaired radeon_mn_unregister()
csky: add missing brackets in a macro for tlb.h
pagewalk: use lockdep_assert_held for locking validation
pagewalk: separate function pointers from iterator data
mm: split out a new pagewalk.h header from mm.h
mm/mmu_notifiers: annotate with might_sleep()
mm/mmu_notifiers: prime lockdep
mm/mmu_notifiers: add a lockdep map for invalidate_range_start/end
mm/mmu_notifiers: remove the __mmu_notifier_invalidate_range_start/end exports
mm/hmm: hmm_range_fault() infinite loop
mm/hmm: hmm_range_fault() NULL pointer bug
mm/hmm: fix hmm_range_fault()'s handling of swapped out pages
mm/mmu_notifiers: remove unregister_no_release
RDMA/odp: remove ib_ucontext from ib_umem
RDMA/odp: use mmu_notifier_get/put for 'struct ib_ucontext_per_mm'
RDMA/mlx5: Use odp instead of mr->umem in pagefault_mr
RDMA/mlx5: Use ib_umem_start instead of umem.address
...
Now that there are allocator APIs that return the ib_umem_odp directly
it should be freed through a umem_odp free'er as well.
Link: https://lore.kernel.org/r/20190819111710.18440-8-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
This is the last creation API that is overloaded for both, there is very
little code sharing and a driver has to be specifically ready for a
umem_odp to be created to use the odp version.
Link: https://lore.kernel.org/r/20190819111710.18440-7-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Implicit ODP umems are special, they don't have any page lists, they don't
exist in the interval tree and they are never DMA mapped.
Instead of trying to guess this based on a zero length use an explicit
flag.
Further, do not allow non-implicit umems to be 0 size.
Link: https://lore.kernel.org/r/20190819111710.18440-4-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
The UMR WQE in the MR re-registration flow requires that
modify_atomic and modify_entity_size capabilities are enabled.
Therefore, check that the these capabilities are present before going to
umr flow and go through slow path if not.
Fixes: c8d75a980f ("IB/mlx5: Respect new UMR capabilities")
Signed-off-by: Moni Shoua <monis@mellanox.com>
Reviewed-by: Guy Levi <guyle@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Link: https://lore.kernel.org/r/20190815083834.9245-8-leon@kernel.org
Signed-off-by: Doug Ledford <dledford@redhat.com>
Introduce helper function to unify various use_umr checks.
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Link: https://lore.kernel.org/r/20190815083834.9245-6-leon@kernel.org
Signed-off-by: Doug Ledford <dledford@redhat.com>
Driver shouldn't allow to use UMR to register a MR when
umr_modify_atomic_disabled is set. Otherwise it will always end up with a
failure in the post send flow which sets the UMR WQE to modify atomic access
right.
Fixes: c8d75a980f ("IB/mlx5: Respect new UMR capabilities")
Signed-off-by: Guy Levi <guyle@mellanox.com>
Reviewed-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Link: https://lore.kernel.org/r/20190731081929.32559-1-leon@kernel.org
Signed-off-by: Doug Ledford <dledford@redhat.com>
Any dma map underlying the MR should only be freed once the MR is fenced
at the hardware.
As of the above we first destroy the MKEY and just after that can safely
call to dma_unmap_single().
Link: https://lore.kernel.org/r/20190723065733.4899-6-leon@kernel.org
Cc: <stable@vger.kernel.org> # 4.3
Fixes: 8a187ee52b ("IB/mlx5: Support the new memory registration API")
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>
Fix unreg_umr to move the MR to a kernel owned PD (i.e. the UMR PD) which
can't be accessed by userspace.
This ensures that nothing can continue to access the MR once it has been
placed in the kernels cache for reuse.
MRs in the cache continue to have their HW state, including DMA tables,
present. Even though the MR has been invalidated, changing the PD provides
an additional layer of protection against use of the MR.
Link: https://lore.kernel.org/r/20190723065733.4899-5-leon@kernel.org
Cc: <stable@vger.kernel.org> # 3.10
Fixes: e126ba97db ("mlx5: Add driver for Mellanox Connect-IB adapters")
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>
Use a direct firmware command to destroy the mkey in case the unreg UMR
operation has failed.
This prevents a case that a mkey will leak out from the cache post a
failure to be destroyed by a UMR WR.
In case the MR cache limit didn't reach a call to add another entry to the
cache instead of the destroyed one is issued.
In addition, replaced a warn message to WARN_ON() as this flow is fatal
and can't happen unless some bug around.
Link: https://lore.kernel.org/r/20190723065733.4899-4-leon@kernel.org
Cc: <stable@vger.kernel.org> # 4.10
Fixes: 49780d42df ("IB/mlx5: Expose MR cache for mlx5_ib")
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>
Fix unreg_umr to ignore the mkey state and do not fail if was freed. This
prevents a case that a user space application already changed the mkey
state to free and then the UMR operation will fail leaving the mkey in an
inappropriate state.
Link: https://lore.kernel.org/r/20190723065733.4899-3-leon@kernel.org
Cc: <stable@vger.kernel.org> # 3.19
Fixes: 968e78dd96 ("IB/mlx5: Enhance UMR support to allow partial page table update")
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>
From git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Required for dependencies in the next patches.
Resolved the conflicts:
- esw_destroy_offloads_acl_tables() use the newer mlx5_esw_for_all_vports()
version
- esw_offloads_steering_init() drop the cap test
- esw_offloads_init() drop the extra function arguments
* branch 'mlx5-next': (39 commits)
net/mlx5: Expose device definitions for object events
net/mlx5: Report EQE data upon CQ completion
net/mlx5: Report a CQ error event only when a handler was set
net/mlx5: mlx5_core_create_cq() enhancements
net/mlx5: Expose the API to register for ANY event
net/mlx5: Use event mask based on device capabilities
net/mlx5: Fix mlx5_core_destroy_cq() error flow
net/mlx5: E-Switch, Handle UC address change in switchdev mode
net/mlx5: E-Switch, Consider host PF for inline mode and vlan pop
net/mlx5: E-Switch, Use iterator for vlan and min-inline setups
net/mlx5: E-Switch, Reg/unreg function changed event at correct stage
net/mlx5: E-Switch, Consolidate eswitch function number of VFs
net/mlx5: E-Switch, Refactor eswitch SR-IOV interface
net/mlx5: Handle host PF vport mac/guid for ECPF
net/mlx5: E-Switch, Use correct flags when configuring vlan
net/mlx5: Reduce dependency on enabled_vfs counter and num_vfs
net/mlx5: Don't handle VF func change if host PF is disabled
net/mlx5: Limit scope of mlx5_get_next_phys_dev() to PCI PF devices
net/mlx5: Move pci status reg access mutex to mlx5_pci_init
net/mlx5: Rename mlx5_pci_dev_type to mlx5_coredev_type
...
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
The lock protecting the data structure does not need to be an rwlock. The
only read access to the lock is in an error path, and if that's limiting
your scalability, you have bigger performance problems.
Eliminate mlx5_mkey_table in favour of using the xarray directly.
reg_mr_callback must use GFP_ATOMIC for allocating XArray nodes as it may
be called in interrupt context.
This also fixes a minor bug where SRCU locking was being used on the radix
tree read side, when RCU was needed too.
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Improve code readability using static helpers for each memory region
type. Re-use the common logic to get smaller functions that are easy
to maintain and reduce code duplication.
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
If possibe, avoid doing a UMR operation to register data and protection
buffers (via MTT/KLM mkeys). Instead, use the local DMA key and map the
SG lists using PA access. This is safe, since the internal key for data
and protection never exposed to the remote server (only signature key
might be exposed). If PA mappings are not possible, perform mapping
using MTT/KLM descriptors.
The setup of the tested benchmark (using iSER ULP):
- 2 servers with 24 cores (1 initiator and 1 target)
- ConnectX-4/ConnectX-5 adapters
- 24 target sessions with 1 LUN each
- ramdisk backstore
- PI active
Performance results running fio (24 jobs, 128 iodepth) using
write_generate=1 and read_verify=1 (w/w.o patch):
bs IOPS(read) IOPS(write)
---- ---------- ----------
512 1266.4K/1262.4K 1720.1K/1732.1K
4k 793139/570902 1129.6K/773982
32k 72660/72086 97229/96164
Using write_generate=0 and read_verify=0 (w/w.o patch):
bs IOPS(read) IOPS(write)
---- ---------- ----------
512 1590.2K/1600.1K 1828.2K/1830.3K
4k 1078.1K/937272 1142.1K/815304
32k 77012/77369 98125/97435
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Suggested-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
In some loads, there is performance degradation when using KLM mkey
instead of MTT mkey. This is because KLM descriptor access is via
indirection that might require more HW resources and cycles.
Using KLM descriptor is not necessary when there are no gaps at the
data/metadata sg lists. As an optimization, use MTT mkey whenever it
is possible. For that matter, allocate internal MTT mkey and choose the
effective pi_mr for in transaction according to the required mapping
scheme.
The setup of the tested benchmark (using iSER ULP):
- 2 servers with 24 cores (1 initiator and 1 target)
- ConnectX-4/ConnectX-5 adapters
- 24 target sessions with 1 LUN each
- ramdisk backstore
- PI active
Performance results running fio (24 jobs, 128 iodepth) using
write_generate=1 and read_verify=1 (w/w.o/baseline):
bs IOPS(read) IOPS(write)
---- ---------- ----------
512 1262.4K/1243.3K/1147.1K 1732.1K/1725.1K/1423.8K
4k 570902/571233/457874 773982/743293/642080
32k 72086/72388/71933 96164/71789/93249
Using write_generate=0 and read_verify=0 (w/w.o patch):
bs IOPS(read) IOPS(write)
---- ---------- ----------
512 1600.1K/1572.1K/1393.3K 1830.3K/1823.5K/1557.2K
4k 937272/921992/762934 815304/753772/646071
32k 77369/75052/72058 97435/73180/94612
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Suggested-by: Max Gurtovoy <maxg@mellanox.com>
Suggested-by: Idan Burstein <idanb@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
IB_WR_REG_SIG_MR is not needed after IB_WR_REG_MR_INTEGRITY
was used.
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
mlx5_ib_map_mr_sg_pi() will map the PI and data dma mapped SG lists to the
mlx5 memory region prior to the registration operation. In the new
API, the mlx5 driver will allocate an internal memory region for the
UMR operation to register both PI and data SG lists. The internal MR
will use KLM mode in order to map 2 (possibly non-contiguous/non-align)
SG lists using 1 memory key. In the new API, each ULP will use 1 memory
region for the signature operation (instead of 3 in the old API). This
memory region will have a key that will be exposed to remote server to
perform RDMA operation. The internal memory key that will map the SG lists
will stay private.
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Update ib_umem_release() to behave similarly to kfree() and allow
submitting NULL pointer as safe input to this function.
Fixes: a52c8e2469 ("RDMA: Clean destroy CQ in drivers do not return errors")
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
This value has always been set to PAGE_SHIFT in the core code, the only
thing that does differently was the ODP path. Move the value into the ODP
struct and still use it for ODP, but change all the non-ODP things to just
use PAGE_SHIFT/PAGE_SIZE/PAGE_MASK directly.
Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
This patch adds support for allocating, deallocating and registering a new
device memory type, STEERING_SW_ICM. This memory can be allocated and
used by a privileged user for direct rule insertion and management of the
device's steering tables.
The type is provided by the user via the dedicated attribute in the
alloc_dm ioctl command.
Signed-off-by: Ariel Levkovich <lariel@mellanox.com>
Reviewed-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
This patch intoruduces a new mlx5_ib driver attribute to the DM allocation
method - the DM type.
In order to allow addition of new types in downstream patches this patch
also refactors the allocation, deallocation and registration handlers to
consider the requested type and perform the necessary actions according to
it.
Since not all future device memory types will be such that are mapped to
user memory, the mandatory page index output attribute is modified to be
optional.
Signed-off-by: Ariel Levkovich <lariel@mellanox.com>
Reviewed-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
In preparation of moving into a model of single IB device multiple ports
move rep to be part of the port structure. We mark a representor device by
setting is_rep, no functional change with this patch.
Signed-off-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
From
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Required for dependencies on the next series
* branch 'mlx5-next':
net/mlx5: E-Switch, add a new prio to be used by the RDMA side
net/mlx5: E-Switch, don't use hardcoded values for FDB prios
net/mlx5: Fix false compilation warning
net/mlx5: Expose MPEIN (Management PCIE INfo) register layout
net/mlx5: Add rate limit print macros
net/mlx5: Add explicit bar address field
net/mlx5: Replace dev_err/warn/info by mlx5_core_err/warn/info
net/mlx5: Use dev->priv.name instead of dev_name
net/mlx5: Make mlx5_core messages independent from mdev->pdev
net/mlx5: Break load_one into three stages
net/mlx5: Function setup/teardown procedures
net/mlx5: Move health and page alloc init to mdev_init
net/mlx5: Split mdev init and pci init
net/mlx5: Remove redundant init functions parameter
net/mlx5: Remove spinlock support from mlx5_write64
net/mlx5: Remove unused MLX5_*_DOORBELL_LOCK macros
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Add bar_addr field to store bar-0 address to avoid calling
pci_resource_start with hard-coded bar-0 as parameter.
Also note that different mlx5 device types will have bar_addr
on different bars.
This patch does not change any functionality.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Signed-off-by: Vu Pham <vuhuong@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The uverbs_attr_bundle with the ucontext is sent down to the drivers ib_x
destroy path as ib_udata. The next patch will use the ib_udata to free the
drivers destroy path from the dependency in 'uobject->context' as we
already did for the create path.
Signed-off-by: Shamir Rabinovitch <shamir.rabinovitch@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
When deferring a prefetch request we need to protect against MR or PD
being destroyed while the request is still enqueued.
The first step is to validate that PD owns the lkey that describes the MR
and that the MR that the lkey refers to is owned by that PD.
The second step is to dequeue all requests when MR is destroyed.
Since PD can't be destroyed while it owns MRs it is guaranteed that when a
worker wakes up the request it refers to is still valid.
Now, it is possible to refrain from taking a reference on the device since
it is assured to be present as pd.
While that, replace the dedicated ordered workqueue with the system
unbound workqueue to reuse an existing resource and improve
performance. This will also fix a bug of queueing to the wrong workqueue.
Fixes: 813e90b1ae ("IB/mlx5: Add advise_mr() support")
Reported-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Yishai Hadas says:
Enable DEVX asynchronous query commands
This series enables querying a DEVX object in an asynchronous mode.
The userspace application won't block when calling the firmware and it will be
able to get the response back once that it will be ready.
To enable the above functionality:
- DEVX asynchronous command completion FD object was introduced.
- The applicable file operations were implemented to enable using it by
the user application.
- Query asynchronous method was added to the DEVX object, it will call the
firmware asynchronously and manages the response on the given input FD.
- Hot unplug support was added for the FD to work properly upon
unbind/disassociate.
- mlx5 core fence for asynchronous commands was implemented and used to
prevent racing upon unbind/disassociate.
This branch is based on mlx5-next & v5.0-rc2 due to dependencies, from
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
* branch 'devx-async':
IB/mlx5: Implement DEVX hot unplug for async command FD
IB/mlx5: Implement the file ops of DEVX async command FD
IB/mlx5: Introduce async DEVX obj query API
IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
When calling debugfs functions, there is no need to ever check the
return value. The function can work or not, but the code logic should
never do something different based on this.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
APIs that have deferred callbacks should have some kind of cleanup
function that callers can use to fence the callbacks. Otherwise things
like module unloading can lead to dangling function pointers, or worse.
The IB MR code is the only place that calls this function and had a
really poor attempt at creating this fence. Provide a good version in
the core code as future patches will add more places that need this
fence.
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
ib_umem_get() can only be called in a method callback, which always has a
udata parameter. This allows ib_umem_get() to derive the ucontext pointer
directly from the udata without requiring the drivers to find it in some
way or another.
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Shamir Rabinovitch <shamir.rabinovitch@oracle.com>
Convert various places to more readable code, which embeds
CONFIG_INFINIBAND_ON_DEMAND_PAGING into the code flow.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Consolidate various checks if MR is ODP backed to one simple helper and
update call sites to use it.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Longer term testing shows this patch didn't play well with MR cache and
caused to call traces during remove_mkeys().
This reverts commit bb7e22a8ab.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
On NVMe offloads connection with many IO queues, EEH takes long time to
recover. The culprit is the synchronize_srcu in the destroy_mkey. The
solution is to use synchronize_srcu only for ODP mkey.
Fixes: b4cfe447d4 ("IB/mlx5: Implement on demand paging by adding support for MMU notifiers")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
The verb advise_mr() is used to give advice to the kernel about an address
range that belongs to a MR. Implement the verb and register it on the
device. The current implementation supports the only known advice to date,
prefetch.
Signed-off-by: Moni Shoua <monis@mellanox.com>
Reviewed-by: Guy Levi <guyle@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Schedule MR cache work only after bucket was initialized.
Cc: <stable@vger.kernel.org> # 4.10
Fixes: 49780d42df ("IB/mlx5: Expose MR cache for mlx5_ib")
Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
