Age | Commit message (Collapse) | Author | Files | Lines |
|
rdma/ib_umem.h and rdma/ib_verbs.h are included by rdma/ib_umem_odp.h.
This patch removes the redundant entries.
Link: https://lore.kernel.org/r/20220823025131.862811-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
|
|
According to the ib spec:
If the CI supports the Base Memory Management Extensions defined in this
specification, the L_Key format must consist of:
24 bit index in the most significant bits of the R_Key, and
8 bit key in the least significant bits of the R_Key
Through a successful Allocate L_Key verb invocation, the CI must let the
consumer own the key portion of the returned R_Key
Therefore, when creating a mkey using DEVX, the consumer is allowed to
change the key part. The kernel should compare only the index part of a
R_Key to determine equality with another R_Key.
Adding capability in order not to break backward compatibility.
Fixes: 534fd7aac56a ("IB/mlx5: Manage indirection mkey upon DEVX flow for ODP")
Link: https://lore.kernel.org/r/3d669aacea85a3a15c3b3b953b3eaba3f80ef9be.1659255945.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
|
|
After replacing the MR cache with an Mkey cache, rename the variables and
functions to fit the new meaning.
Link: https://lore.kernel.org/r/20220726071911.122765-6-michaelgur@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
The pointer imr->umem is assigned twice. Fix this by removing the
redundant one.
Link: https://lore.kernel.org/r/20220518044914.1903125-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Move mlx5_ib_update_mr_pas logic to umr.c, and use
mlx5_umr_post_send_wait() instead of mlx5_ib_post_send_wait().
Since it is the last use of mlx5_ib_post_send_wait(), remove it.
Link: https://lore.kernel.org/r/55a4972f156aba3592a2fc9bcb33e2059acf295f.1649747695.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Move mlx5_ib_update_mr_pas logic to umr.c, and use
mlx5_umr_post_send_wait() instead of mlx5_ib_post_send_wait().
Link: https://lore.kernel.org/r/ed8f2ee6c64804072155d727149abf7105f92536.1649747695.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Move mlx5_ib_can_load_pas_with_umr() and mlx5_ib_can_reconfig_with_umr()
to umr.h and rename them accordingly.
Link: https://lore.kernel.org/r/1b799b0142534a63dfd5bacc5f8ad2256d7777ad.1649747695.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Currently, ent->xlt stores the translation table size. This data should
not be stored in the cache entry but be written directly to the mailbox.
Store ndescs instead, and deduce the translation table size from it
according to the access mode.
Link: https://lore.kernel.org/r/e9dbfaa1f279793a6bd28ee5a31cb4f0f0d70f05.1644947594.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
When allocating a MR from the cache, the driver calls to get_cache_mr(),
and in case of failure, retries with create_cache_mr(). This is the flow
of mlx5_mr_cache_alloc(), so use it instead.
Link: https://lore.kernel.org/r/53c85fcd4de6ec9de0b8e6cbb1bf5d5fe19900c3.1644947594.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Currently IRQs are requested one by one. To balance spreading IRQs
among cpus using such scheme requires remembering cpu mask for the
cpus used for a given device. This complicates the IRQ allocation
scheme in subsequent patch.
Hence, prepare the code for bulk IRQs allocation. This enables
spreading IRQs among cpus in subsequent patch.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Pull rdma updates from Jason Gunthorpe:
"A typical collection of patches this cycle, mostly fixing with a few
new features:
- Fixes from static tools. clang warnings, dead code, unused
variable, coccinelle sweeps, etc
- Driver bug fixes and minor improvements in rxe, bnxt_re, hfi1,
mlx5, irdma, qedr
- rtrs ULP bug fixes an improvments
- Additional counters for bnxt_re
- Support verbs CQ notifications in EFA
- Continued reworking and fixing of rxe
- netlink control to enable/disable optional device counters
- rxe now can use AH objects for its UD path, fixing various bugs in
the process
- Add DMABUF support to EFA"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (103 commits)
RDMA/core: Require the driver to set the IOVA correctly during rereg_mr
RDMA/bnxt_re: Remove unsupported bnxt_re_modify_ah callback
RDMA/irdma: optimize rx path by removing unnecessary copy
RDMA/qed: Use helper function to set GUIDs
RDMA/hns: Use the core code to manage the fixed mmap entries
IB/opa_vnic: Rebranding of OPA VNIC driver to Cornelis Networks
IB/qib: Rebranding of qib driver to Cornelis Networks
IB/hfi1: Rebranding of hfi1 driver to Cornelis Networks
RDMA/bnxt_re: Use helper function to set GUIDs
RDMA/bnxt_re: Fix kernel panic when trying to access bnxt_re_stat_descs
RDMA/qedr: Fix NULL deref for query_qp on the GSI QP
RDMA/hns: Modify the value of MAX_LP_MSG_LEN to meet hardware compatibility
RDMA/hns: Fix initial arm_st of CQ
RDMA/rxe: Make rxe_type_info static const
RDMA/rxe: Use 'bitmap_zalloc()' when applicable
RDMA/rxe: Save a few bytes from struct rxe_pool
RDMA/irdma: Remove the unused variable local_qp
RDMA/core: Fix missed initialization of rdma_hw_stats::lock
RDMA/efa: Add support for dmabuf memory regions
RDMA/umem: Allow pinned dmabuf umem usage
...
|
|
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
A small series to clean up the mlx5 mkey code across the mlx5_core and
InfiniBand.
* branch 'mlx5_mkey':
RDMA/mlx5: Attach ndescs to mlx5_ib_mkey
RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib
RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key
RDMA/mlx5: Remove pd from struct mlx5_core_mkey
RDMA/mlx5: Remove size from struct mlx5_core_mkey
RDMA/mlx5: Remove iova from struct mlx5_core_mkey
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Generalize the use of ndescs by adding it to mlx5_ib_mkey.
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
Move mlx5_core_mkey struct to mlx5_ib, as the mlx5_core doesn't use it
at this point.
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
In mlx5_core and vdpa there is no use of mlx5_core_mkey members except
for the key itself.
As preparation for moving mlx5_core_mkey to mlx5_ib, the occurrences of
struct mlx5_core_mkey in all modules except for mlx5_ib are replaced by
a u32 key.
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
iova is already stored in ibmr->iova, no need to store it here.
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
Control IRQ is the first IRQ vector. This complicates handling of
completion irqs as we need to offset them by one.
in the next patch, there are scenarios where completion and control EQs
will share the same irq. for example: functions with single IRQ. To ease
such scenarios, we shift control IRQ to the end of the irq array.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The error returns for various cases detected by get_prefetchable_mr() get
confused as it flows back to userspace. Properly label each error path and
flow the error code properly back to the system call.
Link: https://lore.kernel.org/r/20210928170846.GA1721590@nvidia.com
Suggested-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Pull rdma updates from Jason Gunthorpe:
"This contains a replacement driver for Intel iWarp hardware. This new
driver supports the old ethernet hardware and also newer chips that
can do ROCE.
Other than that, this contains the typical mix of patches:
- Driver updates and cleanups for bnxt_re, cxgb4, mlx4, and mlx5
- Many static checker driven code clean ups, including a wide
refcount_t conversion
- Several series for the hns driver, more HIP09 HW capabilities,
migration to new HW register manipulators, and code cleanups
- Minor fixes and improvements in srp, rts, and cm
- Improvements throughout for sysfs related code to use
DEVICE_ATTR_*, make the ib_port sysfs first-class, and overall use
sysfs APIs properly
- Intel's new irdma driver replacing i40iw
- rxe general clean ups and Memory Window support"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (211 commits)
RDMA/core: Always release restrack object
RDMA/mlx5: Don't access NULL-cleared mpi pointer
RDMA/irdma: Fix potential overflow expression in irdma_prm_get_pbles
RDMA/irdma: Check contents of user-space irdma_mem_reg_req object
RDMA/rxe: Missing unlock on error in get_srq_wqe()
RDMA/cma: Fix rdma_resolve_route() memory leak
RDMA/core/sa_query: Remove unused argument
RDMA/cma: Fix incorrect Packet Lifetime calculation
RDMA/cma: Protect RMW with qp_mutex
RDMA/cma: Remove unnecessary INIT->INIT transition
RDMA/hns: Add window selection field of congestion control
RDMA/hfi1: Remove use of kmap()
RDMA/irdma: Remove use of kmap()
RDMA/bnxt_re: Fix uninitialized struct bit field rsvd1
IB/isert: Align target max I/O size to initiator size
RDMA/hns: Fix incorrect vlan enable bit in QPC
MAINTAINERS: Update Broadcom RDMA maintainers
RDMA/irdma: Use the queried port attributes
RDMA/rxe: Fix redundant skb_put_zero
RDMA/rxe: Fix extra copy in prepare_ack_packet
...
|
|
Whenever users provided affinity for an EQ creation request, map the
EQ to a matching IRQ.
Matching IRQ=IRQ with the same affinity and type (completion/control) of
the EQ created.
This mapping is being done in agressive dedicated IRQ allocation scheme,
which described bellow.
First, we check whether there is a matching IRQ that his min threshold
is not exhausted.
- min_eqs_threshold = 3 for control EQ.
- min_eqs_threshold = 1 for completion EQ.
In case no matching IRQ was found, try to request a new IRQ.
In case we can't request a new IRQ, reuse least-used matching IRQ.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The users of EQ are running their code on different CPUs and with
various affinity patterns. Move the cpumask setting close to their
actual usage.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Change all the places in the mlx5_ib driver to take the qp type from the
mlx5_ib_qp struct, except the QP initialization flow. It will ensure that
we check the right QP type also for vendor specific QPs.
Link: https://lore.kernel.org/r/b2e16cd65b59cd24fa81c01c7989248da44e58ea.1621413899.git.leonro@nvidia.com
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
The old version of ib_umem_get() need these udata as a parameter but now
they are unnecessary.
Fixes: c320e527e154 ("IB: Allow calls to ib_umem_get from kernel ULPs")
Link: https://lore.kernel.org/r/1620807142-39157-4-git-send-email-liweihang@huawei.com
Signed-off-by: Lang Cheng <chenglang@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Acked-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Variable 'ret' is set to the rerurn value of function
mlx5_mr_cache_alloc() but this value is never read as it is overwritten
with a new value later on, hence it is a redundant assignment and can be
removed
Clean up the following clang-analyzer warning:
drivers/infiniband/hw/mlx5/odp.c:421:2: warning: Value stored to 'ret' is never read [clang-analyzer-deadcode.DeadStores]
Fixes: e6fb246ccafb ("RDMA/mlx5: Consolidate MR destruction to mlx5_ib_dereg_mr()")
Link: https://lore.kernel.org/r/1620296001-120406-1-git-send-email-yang.lee@linux.alibaba.com
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Currently, ODP caps are set during the init stage of mlx5_ib_dev,
regardless of whether the device profile supports ODP or not. There is no
point in setting ODP caps if the device profile doesn't support
ODP. Hence, move setting the ODP caps to the odp_init stage.
Link: https://lore.kernel.org/r/20210318135259.681264-1-leon@kernel.org
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
There is no need to create the ODP EQ if the user doesn't use ODP MRs.
Hence, create it only when the first ODP MR is created. This EQ will be
destroyed only when the device is unloaded.
This will decrease the number of EQs created per device. for example: If
we creates 1K devices (SF/VF/etc'), than we will decrease the num of EQs
by 1K.
Link: https://lore.kernel.org/r/20210314125418.179716-1-leon@kernel.org
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Now that the SRCU stuff has been removed the entire MR destroy logic can
be made a lot simpler. Currently there are many different ways to destroy a
MR and it makes it really hard to do this task correctly. Route all
destruction through mlx5_ib_dereg_mr() and make it work for all
situations.
Since it turns out all the different MR types do basically the same thing
this removes a lot of knowledge of MR internals from ODP and leaves ODP
just exporting an operation to clean up children.
This fixes a few weird corner cases bugs and firmly uses the correct
ordering of the MR destruction:
- Stop parallel access to the mkey via the ODP xarray
- Stop DMA
- Release the umem
- Clean up ODP children
- Free/Recycle the MR
Link: https://lore.kernel.org/r/20210304120745.1090751-4-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
All of the ODP code assumes when it calls mlx5_mr_cache_alloc() the ODP
related fields are zero'd. This is true if the MR was just allocated, but
if the MR is recycled through the cache then the values are never zero'd.
This causes a bug in the odp_stats, they don't reset when the MR is
reallocated, also is_odp_implicit is never 0'd.
So we can use memset on a block of the mlx5_ib_mr reorganize the structure
to put all the data that can be zero'd by the cache at the end.
It is organized as an anonymous struct because the next patch will make
this a union.
Delete the unused smr_info. Don't set the kernel only desc_size on the
user path. No longer any need to zero mr->parent before freeing it, the
memset() will get it now.
Fixes: a3de94e3d61e ("IB/mlx5: Introduce ODP diagnostic counters")
Link: https://lore.kernel.org/r/20210304120745.1090751-2-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
The W=1 allmodconfig build produces the following warning:
drivers/infiniband/hw/mlx5/odp.c:1086: warning: wrong kernel-doc identifier on line:
* Parse a series of data segments for page fault handling.
Fix it by changing /** to be /* as it is written in kernel-doc
documentation.
Fixes: 5e769e444d26 ("RDMA/hw/mlx5/odp: Fix formatting and add missing descriptions in 'pagefault_data_segments()'")
Link: https://lore.kernel.org/r/20210302074214.1054299-2-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Cleanup the synchronize_srcu() from the ODP flow as it was found to be a
very heavy time consumer as part of dereg_mr.
For example de-registration of 10000 ODP MRs each with size of 2M hugepage
took 19.6 sec comparing de-registration of same number of non ODP MRs that
took 172 ms.
The new locking scheme uses the wait_event() mechanism which follows the
use count of the MR instead of using synchronize_srcu().
By that change, the time required for the above test took 95 ms which is
even better than the non ODP flow.
Once fully dropped the srcu usage, had to come with a lock to protect the
XA access.
As part of using the above mechanism we could also clean the
num_deferred_work stuff and follow the use count instead.
Link: https://lore.kernel.org/r/20210202071309.2057998-1-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Fixes the following W=1 kernel build warning(s):
drivers/infiniband/hw/mlx5/odp.c:1062: warning: Function parameter or member 'dev' not described in 'pagefault_data_segments'
drivers/infiniband/hw/mlx5/odp.c:1062: warning: Function parameter or member 'pfault' not described in 'pagefault_data_segments'
drivers/infiniband/hw/mlx5/odp.c:1062: warning: Function parameter or member 'wqe' not described in 'pagefault_data_segments'
drivers/infiniband/hw/mlx5/odp.c:1062: warning: Function parameter or member 'wqe_end' not described in 'pagefault_data_segments'
drivers/infiniband/hw/mlx5/odp.c:1062: warning: Function parameter or member 'bytes_mapped' not described in 'pagefault_data_segments'
drivers/infiniband/hw/mlx5/odp.c:1062: warning: Function parameter or member 'total_wqe_bytes' not described in 'pagefault_data_segments'
drivers/infiniband/hw/mlx5/odp.c:1062: warning: Function parameter or member 'receive_queue' not described in 'pagefault_data_segments'
Link: https://lore.kernel.org/r/20210121094519.2044049-2-lee.jones@linaro.org
Cc: Leon Romanovsky <leon@kernel.org>
Cc: Doug Ledford <dledford@redhat.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: linux-rdma@vger.kernel.org
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Implement the new driver method 'reg_user_mr_dmabuf'. Utilize the core
functions to import dma-buf based memory region and update the mappings.
Add code to handle dma-buf related page fault.
Link: https://lore.kernel.org/r/1608067636-98073-5-git-send-email-jianxin.xiong@intel.com
Signed-off-by: Jianxin Xiong <jianxin.xiong@intel.com>
Reviewed-by: Sean Hefty <sean.hefty@intel.com>
Acked-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Acked-by: Christian Koenig <christian.koenig@amd.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Currently, DM MR registration flow doesn't set the mlx5_ib_dev pointer and
can cause a NULL pointer dereference if userspace dumps the MR via rdma
tool.
Assign the IB device together with the other fields and remove the
redundant reference of mlx5_ib_dev from mlx5_ib_mr.
Cc: stable@vger.kernel.org
Fixes: 6c29f57ea475 ("IB/mlx5: Device memory mr registration support")
Link: https://lore.kernel.org/r/20201203190807.127189-1-leon@kernel.org
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
This function handles an ODP and regular MR flow all mushed together, even
though the two flows are quite different. Split them into two dedicated
functions.
Link: https://lore.kernel.org/r/20201130075839.278575-5-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Sync device with CPU pages upon ODP MR registration. mlx5 already has to
zero the HW's version of the PAS list, may as well deliver a PAS list that
matches the current CPU page tables configuration.
Link: https://lore.kernel.org/r/20200930163828.1336747-5-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Extend advice MR to support non faulting mode, this can improve
performance by increasing the populated page tables in the device.
Link: https://lore.kernel.org/r/20200930163828.1336747-4-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Enable ODP sync without faulting, this improves performance by reducing
the number of page faults in the system.
The gain from this option is that the device page table can be aligned
with the presented pages in the CPU page table without causing page
faults.
As of that, the overhead on data path from hardware point of view to
trigger a fault which end-up by calling the driver to bring the pages
will be dropped.
Link: https://lore.kernel.org/r/20200930163828.1336747-3-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Move to use hmm_range_fault() instead of get_user_pags_remote() to improve
performance in a few aspects:
This includes:
- Dropping the need to allocate and free memory to hold its output
- No need any more to use put_page() to unpin the pages
- The logic to detect contiguous pages is done based on the returned
order, no need to run per page and evaluate.
In addition, moving to use hmm_range_fault() enables to reduce page faults
in the system with it's snapshot mode, this will be introduced in next
patches from this series.
As part of this, cleanup some flows and use the required data structures
to work with hmm_range_fault().
Link: https://lore.kernel.org/r/20200930163828.1336747-2-leon@kernel.org
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Once a mkey is created it can be modified using UMR. This is desirable for
performance reasons. However, different hardware has restrictions on what
modifications are possible using UMR. Make sense of these checks:
- mlx5_ib_can_reconfig_with_umr() returns true if the access flags can be
altered. Most cases create MRs using 0 access flags (now made clear by
consistent use of set_mkc_access_pd_addr_fields()), but the old logic
here was tormented. Make it clear that this is checking if the current
access_flags can be modified using UMR to different access_flags. It is
always OK to use UMR to change flags that all HW supports.
- mlx5_ib_can_load_pas_with_umr() returns true if UMR can be used to
enable and update the PAS/XLT. Enabling requires updating the entity
size, so UMR ends up completely disabled on this old hardware. Make it
clear why it is disabled. FRWR, ODP and cache always requires
mlx5_ib_can_load_pas_with_umr().
- mlx5_ib_pas_fits_in_mr() is used to tell if an existing MR can be
resized to hold a new PAS list. This only works for cached MR's because
we don't store the PAS list size in other cases.
To be very clear, arrange things so any pre-created MR's in the cache
check the newly requested access_flags before allowing the MR to leave the
cache. If UMR cannot set the required access_flags the cache fails to
create the MR.
This in turn means relaxed ordering and atomic are now correctly blocked
early for implicit ODP on older HW.
Link: https://lore.kernel.org/r/20200914112653.345244-6-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Pull rdma updates from Jason Gunthorpe:
"A quiet cycle after the larger 5.8 effort. Substantially cleanup and
driver work with a few smaller features this time.
- Driver updates for hfi1, rxe, mlx5, hns, qedr, usnic, bnxt_re
- Removal of dead or redundant code across the drivers
- RAW resource tracker dumps to include a device specific data blob
for device objects to aide device debugging
- Further advance the IOCTL interface, remove the ability to turn it
off. Add QUERY_CONTEXT, QUERY_MR, and QUERY_PD commands
- Remove stubs related to devices with no pkey table
- A shared CQ scheme to allow multiple ULPs to share the CQ rings of
a device to give higher performance
- Several more static checker, syzkaller and rare crashers fixed"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (121 commits)
RDMA/mlx5: Fix flow destination setting for RDMA TX flow table
RDMA/rxe: Remove pkey table
RDMA/umem: Add a schedule point in ib_umem_get()
RDMA/hns: Fix the unneeded process when getting a general type of CQE error
RDMA/hns: Fix error during modify qp RTS2RTS
RDMA/hns: Delete unnecessary memset when allocating VF resource
RDMA/hns: Remove redundant parameters in set_rc_wqe()
RDMA/hns: Remove support for HIP08_A
RDMA/hns: Refactor hns_roce_v2_set_hem()
RDMA/hns: Remove redundant hardware opcode definitions
RDMA/netlink: Remove CAP_NET_RAW check when dump a raw QP
RDMA/include: Replace license text with SPDX tags
RDMA/rtrs: remove WQ_MEM_RECLAIM for rtrs_wq
RDMA/rtrs-clt: add an additional random 8 seconds before reconnecting
RDMA/cma: Execute rdma_cm destruction from a handler properly
RDMA/cma: Remove unneeded locking for req paths
RDMA/cma: Using the standard locking pattern when delivering the removal event
RDMA/cma: Simplify DEVICE_REMOVAL for internal_id
RDMA/efa: Add EFA 0xefa1 PCI ID
RDMA/efa: User/kernel compatibility handshake mechanism
...
|
|
destroy_prefetch_work() must always be called if the work is not going
to be queued. The num_sge also should have been set to i, not i-1
which avoids the condition where it shouldn't have been called in the
first place.
Cc: stable@vger.kernel.org
Fixes: fb985e278a30 ("RDMA/mlx5: Use SRCU properly in ODP prefetch")
Link: https://lore.kernel.org/r/20200727095712.495652-1-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
According to the locking scheme, mlx5_ib_update_xlt() should be called
with srcu_read_lock(dev->odp->srcu). Prefetch missed this. This fixes the
below WARN from lockdep_assert_held():
WARNING: CPU: 1 PID: 1130 at drivers/infiniband/hw/mlx5/odp.c:132 mlx5_odp_populate_xlt+0x175/0x180 [mlx5_ib]
Modules linked in: xt_conntrack xt_MASQUERADE nf_conntrack_netlink nfnetlink xt_addrtype iptable_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 br_netfilter overlay ib_srp scsi_transport_srp rpcrdma rdma_ucm ib_iser libiscsi scsi_transport_iscsi rdma_cm iw_cm ib_umad ib_ipoib ib_cm mlx5_ib ib_uverbs ib_core mlx5_core mlxfw ptp pps_core
CPU: 1 PID: 1130 Comm: kworker/u16:11 Tainted: G W 5.8.0-rc5_for_upstream_debug_2020_07_13_11_04 #1
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014
Workqueue: events_unbound mlx5_ib_prefetch_mr_work [mlx5_ib]
RIP: 0010:mlx5_odp_populate_xlt+0x175/0x180 [mlx5_ib]
Code: 08 e2 85 c0 0f 84 65 ff ff ff 49 8b 87 60 01 00 00 be ff ff ff ff 48 8d b8 b0 39 00 00 e8 93 e0 50 e1 85 c0 0f 85 45 ff ff ff <0f> 0b e9 3e ff ff ff 0f 0b eb c7 0f 1f 44 00 00 48 8b 87 98 0f 00
RSP: 0018:ffff88840f44fc68 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff88840cc9d000 RCX: ffff88840efcd940
RDX: 0000000000000000 RSI: ffff88844871b9b0 RDI: ffff88840efce100
RBP: ffff88840cc9d040 R08: 0000000000000040 R09: 0000000000000001
R10: ffff88846ced3068 R11: 0000000000000000 R12: 00000000000156ec
R13: 0000000000000004 R14: 0000000000000004 R15: ffff888439941000
FS: 0000000000000000(0000) GS:ffff88846fa80000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f8536d12430 CR3: 0000000437a5e006 CR4: 0000000000360ea0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
mlx5_ib_update_xlt+0x37c/0x7c0 [mlx5_ib]
pagefault_mr+0x315/0x440 [mlx5_ib]
mlx5_ib_prefetch_mr_work+0x56/0xa0 [mlx5_ib]
process_one_work+0x215/0x5c0
worker_thread+0x3c/0x380
? process_one_work+0x5c0/0x5c0
kthread+0x133/0x150
? kthread_park+0x90/0x90
ret_from_fork+0x1f/0x30
Hold the SRCU during prefetch, even though it strictly isn't needed since
prefetch is holding the num_deferred_work it does make it easier to reason
about.
Fixes: 5256edcb98a1 ("RDMA/mlx5: Rework implicit ODP destroy")
Link: https://lore.kernel.org/r/20200719065747.131157-1-leon@kernel.org
Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Prefetch work in mlx5_ib_prefetch_mr_work can be queued and able to run
concurrently with destruction of the implicit MR. The num_deferred_work
was intended to serialize this, but there is a race:
CPU0 CPU1
mlx5_ib_free_implicit_mr()
xa_erase(odp_mkeys)
synchronize_srcu()
__xa_erase(implicit_children)
mlx5_ib_prefetch_mr_work()
pagefault_mr()
pagefault_implicit_mr()
implicit_get_child_mr()
xa_cmpxchg()
atomic_dec_and_test(num_deferred_mr)
wait_event(imr->q_deferred_work)
ib_umem_odp_release(odp_imr)
kfree(odp_imr)
At this point in mlx5_ib_free_implicit_mr() the implicit_children list is
supposed to be empty forever so that destroy_unused_implicit_child_mr()
and related are not and will not be running.
Since it is not empty the destroy_unused_implicit_child_mr() flow ends up
touching deallocated memory as mlx5_ib_free_implicit_mr() already tore down the
imr parent.
The solution is to flush out the prefetch wq by driving num_deferred_work
to zero after creation of new prefetch work is blocked.
Fixes: 5256edcb98a1 ("RDMA/mlx5: Rework implicit ODP destroy")
Link: https://lore.kernel.org/r/20200719065435.130722-1-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
For debugging purpose it will be easier to understand if prefetch works
okay if it has its own counter. Introduce ODP prefetch counter and count
per MR the total number of prefetched pages.
In addition remove comment which is not relevant anymore and anyway not in
the correct place.
Link: https://lore.kernel.org/r/20200621104147.53795-1-leon@kernel.org
Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Reuse newly introduced mlx5_cmd_exec_in() and mlx5_cmd_exec_inout() to
reduce code duplication in mlx5_ib module.
Link: https://lore.kernel.org/r/20200506065513.4668-2-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
|
|
QP type is stored in the IB/core QP struct, but it doesn't have all the
needed information, like internal QP type used in the driver itself.
Update mlx5_ib to have cached QP type which includes both IBTA and
Mellanox specific one.
Such change allows us to make even further cleanup of QP creation flow.
Link: https://lore.kernel.org/r/20200427154636.381474-21-leon@kernel.org
Reviewed-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
|
|
MLX5_QP_FLAG_SIGNATURE is exposed to the users but in the kernel
the create_qp flow treated it differently from other MLX5_QP_FLAG_*s.
Fix it by ditching wq_sig boolean variable and use general flag_en
mechanism.
Link: https://lore.kernel.org/r/20200427154636.381474-16-leon@kernel.org
Reviewed-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
|
|
The mlx5_core doesn't need any functionality coded in qp.c, so move
that file to drivers/infiniband/ be under mlx5_ib responsibility.
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
|
|
Leon Romanovsky says:
====================
This series fixes various corner cases in the mlx5_ib MR cache
implementation, see specific commit messages for more information.
====================
Based on the mlx5-next branch at
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Due to dependencies
* branch 'mlx5_mr-cache':
RDMA/mlx5: Allow MRs to be created in the cache synchronously
RDMA/mlx5: Revise how the hysteresis scheme works for cache filling
RDMA/mlx5: Fix locking in MR cache work queue
RDMA/mlx5: Lock access to ent->available_mrs/limit when doing queue_work
RDMA/mlx5: Fix MR cache size and limit debugfs
RDMA/mlx5: Always remove MRs from the cache before destroying them
RDMA/mlx5: Simplify how the MR cache bucket is located
RDMA/mlx5: Rename the tracking variables for the MR cache
RDMA/mlx5: Replace spinlock protected write with atomic var
{IB,net}/mlx5: Move asynchronous mkey creation to mlx5_ib
{IB,net}/mlx5: Assign mkey variant in mlx5_ib only
{IB,net}/mlx5: Setup mkey variant before mr create command invocation
|