Pull rdma updates from Jason Gunthorpe:
"Seveal fixes scattered across the drivers and a few new features:
- Minor updates and bug fixes to hfi1, efa, iopob, bnxt, hns
- Force disassociate the userspace FD when hns does an async reset
- bnxt new features for optimized modify QP to skip certain stayes,
CQ coalescing, better debug dumping
- mlx5 new data placement ordering feature
- Faster destruction of mlx5 devx HW objects
- Improvements to RDMA CM mad handling"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (51 commits)
RDMA/bnxt_re: Correct the sequence of device suspend
RDMA/bnxt_re: Use the default mode of congestion control
RDMA/bnxt_re: Support different traffic class
IB/cm: Rework sending DREQ when destroying a cm_id
IB/cm: Do not hold reference on cm_id unless needed
IB/cm: Explicitly mark if a response MAD is a retransmission
RDMA/mlx5: Move events notifier registration to be after device registration
RDMA/bnxt_re: Cache MSIx info to a local structure
RDMA/bnxt_re: Refurbish CQ to NQ hash calculation
RDMA/bnxt_re: Refactor NQ allocation
RDMA/bnxt_re: Fail probe early when not enough MSI-x vectors are reserved
RDMA/hns: Fix different dgids mapping to the same dip_idx
RDMA/bnxt_re: Add set_func_resources support for P5/P7 adapters
RDMA/bnxt_re: Enhance RoCE SRIOV resource configuration design
bnxt_en: Add support for RoCE sriov configuration
RDMA/hns: Fix NULL pointer derefernce in hns_roce_map_mr_sg()
RDMA/hns: Fix out-of-order issue of requester when setting FENCE
RDMA/nldev: Add IB device and net device rename events
RDMA/mlx5: Add implementation for ufile_hw_cleanup device operation
RDMA/core: Move ib_uverbs_file struct to uverbs_types.h
...
There are few use cases where CQ create and destroy
is seen before re-creating the CQ, this kind of use
case is disturbing the RR distribution and all the
active CQ getting mapped to only 2 NQ alternatively.
Fixing the CQ to NQ hash calculation by implementing
a quick load sorting mechanism under a mutex.
Using this, if the CQ was allocated and destroyed
before using it, the nq selecting algorithm still
obtains the least loaded CQ. Thus balancing the load
on NQs.
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Link: https://patch.msgid.link/1731577748-1804-4-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Modify QP improvements are for state transitions
from INIT -> RTR and RTR -> RTS.
In order to support the Modify QP Optimization feature,
the driver is expected to check for the feature support
in the CMDQ_QUERY_FUNC and register its support for this
feature with the FW in CMDQ_INITIALIZE_FIRMWARE.
Additionally, the driver is required to specify the new
fields and attribute masks for the transitions as follows:
1. INIT -> RTR:
- New fields: srq_used, type.
- enable srq_used when RC QP is configured to use SRQ.
- set the type based on the QP type.
- Mandatory masks:
- RC: CMDQ_MODIFY_QP_MODIFY_MASK_ACCESS,
CMDQ_MODIFY_QP_MODIFY_MASK_PKEY
- UD QP and QP1: CMDQ_MODIFY_QP_MODIFY_MASK_PKEY,
CMDQ_MODIFY_QP_MODIFY_MASK_QKEY
2. RTR -> RTS:
- New fields: type
- set the type based on the QP type.
- Mandatory masks:
- RC: CMDQ_MODIFY_QP_MODIFY_MASK_ACCESS
- UD QP and QP1: CMDQ_MODIFY_QP_MODIFY_MASK_QKEY
Reviewed-by: Saravanan Vajravel <saravanan.vajravel@broadcom.com>
Reviewed-by: Tushar Rane <tushar.rane@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://patch.msgid.link/1729065346-1364-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Currently the CQ toggle value in the shared page (read by the userlib) is
updated as part of the cqn_handler. There is a potential race of
application calling the CQ ARM doorbell immediately and using the old
toggle value.
Change the sequence of updating CQ toggle value to update in the
bnxt_qplib_service_nq function immediately after reading the toggle value
to be in sync with the HW updated value.
Fixes: e275919d96 ("RDMA/bnxt_re: Share a page to expose per CQ info with userspace")
Link: https://patch.msgid.link/r/1728373302-19530-9-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Chandramohan Akula <chandramohan.akula@broadcom.com>
Reviewed-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
While reporting the completions, SQ Work Queue index is required to
identify the WQE that generated the completions. In variable WQE mode, FW
returns the slot index for Error completions. Driver need to walk through
the shadow queue between the consumer index and producer index and matches
the slot index returned by FW. If a match is found, the next index of the
shadow queue is the WQE index to be considered for remaining poll_cq loop.
Link: https://patch.msgid.link/r/1724042847-1481-3-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Hongguang Gao <hongguang.gao@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Variable size WQE means that each send Work Queue Entry to HW can use
different WQE sizes as opposed to the static WQE size on the current
devices. Set variable WQE mode for Gen P7 devices. Depth of the Queue will
be a multiple of slot which is 16 bytes. The number of slots should be a
multiple of 256 as per the HW requirement.
Initialize the Software shadow queue to hold requests equal to the number
of slots. Also, do not expose the variable size WQE capability until the
last patch in the series.
Link: https://patch.msgid.link/r/1724042847-1481-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Hongguang Gao <hongguang.gao@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
FW reports the HW capability to use PSN table or MSN table and
driver/library need to select it based on this capability.
Use the new capability instead of the older capability check for HW
retransmission while handling the MSN/PSN table. FW report
zero (PSN table) for older adapters to maintain backward compatibility.
Also, Updated the FW interface structures to handle the new fields.
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1716876697-25970-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
The queue index wrap around logic is based on power of 2 size depth.
All queues are created with power of 2 depth. This increases the
memory usage by the driver. This change is required for the next
patches that avoids the power of 2 depth requirement for each of
the queues.
Update the function that increments producer index and consumer
index during wrap around. Also, changes the index handling across
multiple functions.
Signed-off-by: Chandramohan Akula <chandramohan.akula@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1698069803-1787-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Pull rdma updates from Jason Gunthorpe:
"Many small changes across the subystem, some highlights:
- Usual driver cleanups in qedr, siw, erdma, hfi1, mlx4/5, irdma,
mthca, hns, and bnxt_re
- siw now works over tunnel and other netdevs with a MAC address by
removing assumptions about a MAC/GID from the connection manager
- "Doorbell Pacing" for bnxt_re - this is a best effort scheme to
allow userspace to slow down the doorbell rings if the HW gets full
- irdma egress VLAN priority, better QP/WQ sizing
- rxe bug fixes in queue draining and srq resizing
- Support more ethernet speed options in the core layer
- DMABUF support for bnxt_re
- Multi-stage MTT support for erdma to allow much bigger MR
registrations
- A irdma fix with a CVE that came in too late to go to -rc, missing
bounds checking for 0 length MRs"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (87 commits)
IB/hfi1: Reduce printing of errors during driver shut down
RDMA/hfi1: Move user SDMA system memory pinning code to its own file
RDMA/hfi1: Use list_for_each_entry() helper
RDMA/mlx5: Fix trailing */ formatting in block comment
RDMA/rxe: Fix redundant break statement in switch-case.
RDMA/efa: Fix wrong resources deallocation order
RDMA/siw: Call llist_reverse_order in siw_run_sq
RDMA/siw: Correct wrong debug message
RDMA/siw: Balance the reference of cep->kref in the error path
Revert "IB/isert: Fix incorrect release of isert connection"
RDMA/bnxt_re: Fix kernel doc errors
RDMA/irdma: Prevent zero-length STAG registration
RDMA/erdma: Implement hierarchical MTT
RDMA/erdma: Refactor the storage structure of MTT entries
RDMA/erdma: Renaming variable names and field names of struct erdma_mem
RDMA/hns: Support hns HW stats
RDMA/hns: Dump whole QP/CQ/MR resource in raw
RDMA/irdma: Add missing kernel-doc in irdma_setup_umode_qp()
RDMA/mlx4: Copy union directly
RDMA/irdma: Drop unused kernel push code
...
Driver unload hits a hang during stress testing of load/unload.
stack trace snippet -
tasklet_kill at ffffffff9aabb8b2
bnxt_qplib_nq_stop_irq at ffffffffc0a805fb [bnxt_re]
bnxt_qplib_disable_nq at ffffffffc0a80c5b [bnxt_re]
bnxt_re_dev_uninit at ffffffffc0a67d15 [bnxt_re]
bnxt_re_remove_device at ffffffffc0a6af1d [bnxt_re]
tasklet_kill can hang if the tasklet is scheduled after it is disabled.
Modified the sequences to disable the interrupt first and synchronize
irq before disabling the tasklet.
Fixes: 1ac5a40479 ("RDMA/bnxt_re: Add bnxt_re RoCE driver")
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1689322969-25402-3-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
HW may generate completions that indicates QP is destroyed.
Driver should not be scheduling any more completion handlers
for this QP, after the QP is destroyed. Since CQs are active
during the QP destroy, driver may still schedule completion
handlers. This can cause a race where the destroy_cq and poll_cq
running simultaneously.
Snippet of kernel panic while doing bnxt_re driver load unload in loop.
This indicates a poll after the CQ is freed.
[77786.481636] Call Trace:
[77786.481640] <TASK>
[77786.481644] bnxt_re_poll_cq+0x14a/0x620 [bnxt_re]
[77786.481658] ? kvm_clock_read+0x14/0x30
[77786.481693] __ib_process_cq+0x57/0x190 [ib_core]
[77786.481728] ib_cq_poll_work+0x26/0x80 [ib_core]
[77786.481761] process_one_work+0x1e5/0x3f0
[77786.481768] worker_thread+0x50/0x3a0
[77786.481785] ? __pfx_worker_thread+0x10/0x10
[77786.481790] kthread+0xe2/0x110
[77786.481794] ? __pfx_kthread+0x10/0x10
[77786.481797] ret_from_fork+0x2c/0x50
To avoid this, complete all completion handlers before returning the
destroy QP. If free_cq is called soon after destroy_qp, IB stack
will cancel the CQ work before invoking the destroy_cq verb and
this will prevent any race mentioned.
Fixes: 1ac5a40479 ("RDMA/bnxt_re: Add bnxt_re RoCE driver")
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1689322969-25402-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
bnxt_re currently uses the names "bnxt_qplib_creq" and "bnxt_qplib_nq-0"
while registering IRQs. There is no way to distinguish the IRQs of
different device ports when there are multiple IB devices registered.
This could make the scenarios worse where one want to pin IRQs of a device
port to certain CPUs.
Fixed the code to use unique names which has PCI BDF information while
registering interrupts like: "bnxt_re-nq-0@pci:0000:65:00.0" and
"bnxt_re-creq@pci:0000:65:00.1".
Fixes: 1ac5a40479 ("RDMA/bnxt_re: Add bnxt_re RoCE driver")
Link: https://lore.kernel.org/r/1684478897-12247-4-git-send-email-selvin.xavier@broadcom.com
Reviewed-by: Bhargava Chenna Marreddy <bhargava.marreddy@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
When the ulp hook to start the IRQ fails because the rings are not
available, tasklets are not enabled. In this case when the driver is
unloaded, driver calls CREQ tasklet_kill. This causes an indefinite hang
as the tasklet is not enabled.
Driver shouldn't call tasklet_kill if it is not enabled. So using the
creq->requested and nq->requested flags to identify if both tasklets/irqs
are registered. Checking this flag while scheduling the tasklet from
ISR. Also, added a cleanup for disabling tasklet, in case request_irq
fails during start_irq.
Check for return value for bnxt_qplib_rcfw_start_irq and in case the
bnxt_qplib_rcfw_start_irq fails, return bnxt_re_start_irq without
attempting to start NQ IRQs.
Fixes: 1ac5a40479 ("RDMA/bnxt_re: Add bnxt_re RoCE driver")
Link: https://lore.kernel.org/r/1684478897-12247-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
clang with W=1 reports
drivers/infiniband/hw/bnxt_re/qplib_fp.c:303:6: error: variable
'num_srqne_processed' set but not used [-Werror,-Wunused-but-set-variable]
int num_srqne_processed = 0;
^
drivers/infiniband/hw/bnxt_re/qplib_fp.c:304:6: error: variable
'num_cqne_processed' set but not used [-Werror,-Wunused-but-set-variable]
int num_cqne_processed = 0;
^
These variables are not used so remove them.
Signed-off-by: Tom Rix <trix@redhat.com>
Link: https://lore.kernel.org/r/20230325140559.1336056-1-trix@redhat.com
Acked-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Add resize_cq verb support for user space CQs. Resize operation for
kernel CQs are not supported now.
Driver should free the current CQ only after user library polls
for all the completions and switch to new CQ. So after the resize_cq
is returned from the driver, user library polls for existing completions
and store it as temporary data. Once library reaps all completions in the
current CQ, it invokes the ibv_cmd_poll_cq to inform the driver about
the resize_cq completion. Adding a check for user CQs in driver's
poll_cq and complete the resize operation for user CQs.
Updating uverbs_cmd_mask with poll_cq to support this.
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1678868215-23626-1-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
