linux.git/drivers/infiniband, branch master

RDMA/bnxt_re: synchronize the qp-handle table array

2024-10-21T16:28:15+00:00

There is a race between the CREQ tasklet and destroy qp when accessing the
qp-handle table. There is a chance of reading a valid qp-handle in the
CREQ tasklet handler while the QP is already moving ahead with the
destruction.

Fixing this race by implementing a table-lock to synchronize the access.

Fixes: f218d67ef004 ("RDMA/bnxt_re: Allow posting when QPs are in error")
Fixes: 84cf229f4001 ("RDMA/bnxt_re: Fix the qp table indexing")
Link: https://patch.msgid.link/r/1728912975-19346-3-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Kalesh AP 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe

RDMA/bnxt_re: Fix the usage of control path spin locks

2024-10-21T16:28:15+00:00

Control path completion processing always runs in tasklet context. To
synchronize with the posting thread, there is no need to use the irq
variant of spin lock. Use spin_lock_bh instead.

Fixes: 1ac5a4047975 ("RDMA/bnxt_re: Add bnxt_re RoCE driver")
Link: https://patch.msgid.link/r/1728912975-19346-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Kalesh AP 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe

RDMA/mlx5: Round max_rd_atomic/max_dest_rd_atomic up instead of down

2024-10-21T16:25:44+00:00

After the cited commit below max_dest_rd_atomic and max_rd_atomic values
are being rounded down to the next power of 2. As opposed to the old
behavior and mlx4 driver where they used to be rounded up instead.

In order to stay consistent with older code and other drivers, revert to
using fls round function which rounds up to the next power of 2.

Fixes: f18e26af6aba ("RDMA/mlx5: Convert modify QP to use MLX5_SET macros")
Link: https://patch.msgid.link/r/d85515d6ef21a2fa8ef4c8293dce9b58df8a6297.1728550179.git.leon@kernel.org
Signed-off-by: Patrisious Haddad 
Reviewed-by: Maher Sanalla 
Signed-off-by: Leon Romanovsky 
Signed-off-by: Jason Gunthorpe

RDMA/cxgb4: Dump vendor specific QP details

2024-10-21T16:22:33+00:00

Restore the missing functionality to dump vendor specific QP details,
which was mistakenly removed in the commit mentioned in Fixes line.

Fixes: 5cc34116ccec ("RDMA: Add dedicated QP resource tracker function")
Link: https://patch.msgid.link/r/ed9844829135cfdcac7d64285688195a5cd43f82.1728323026.git.leonro@nvidia.com
Reported-by: Dr. David Alan Gilbert 
Closes: https://lore.kernel.org/all/Zv_4qAxuC0dLmgXP@gallifrey
Signed-off-by: Leon Romanovsky 
Signed-off-by: Jason Gunthorpe

RDMA/bnxt_re: Fix the GID table length

2024-10-11T23:49:02+00:00

GID table length is reported by FW. The gid index which is passed to the
driver during modify_qp/create_ah is restricted by the sgid_index field of
struct ib_global_route.  sgid_index is u8 and the max sgid possible is
256.

Each GID entry in HW will have 2 GID entries in the kernel gid table.  So
we can support twice the gid table size reported by FW. Also, restrict the
max GID to 256 also.

Fixes: 847b97887ed4 ("RDMA/bnxt_re: Restrict the max_gids to 256")
Link: https://patch.msgid.link/r/1728373302-19530-11-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Kalesh AP 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe

RDMA/bnxt_re: Fix a bug while setting up Level-2 PBL pages

2024-10-11T23:49:02+00:00

Avoid memory corruption while setting up Level-2 PBL pages for the non MR
resources when num_pages > 256K.

There will be a single PDE page address (contiguous pages in the case of >
PAGE_SIZE), but, current logic assumes multiple pages, leading to invalid
memory access after 256K PBL entries in the PDE.

Fixes: 0c4dcd602817 ("RDMA/bnxt_re: Refactor hardware queue memory allocation")
Link: https://patch.msgid.link/r/1728373302-19530-10-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Bhargava Chenna Marreddy 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe

RDMA/bnxt_re: Change the sequence of updating the CQ toggle value

2024-10-11T23:49:01+00:00

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: e275919d9669 ("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 
Reviewed-by: Selvin Xavier 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe

RDMA/bnxt_re: Fix an error path in bnxt_re_add_device

2024-10-11T23:49:01+00:00

In bnxt_re_add_device(), when register netdev notifier fails, driver is
not unregistering the IB device in the error cleanup path.  Also, removed
the duplicate cleanup in error path of bnxt_re_probe.

Fixes: 94a9dc6ac8f7 ("RDMA/bnxt_re: Group all operations under add_device and remove_device")
Link: https://patch.msgid.link/r/1728373302-19530-8-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Kalesh AP 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe

RDMA/bnxt_re: Avoid CPU lockups due fifo occupancy check loop

2024-10-11T23:49:01+00:00

Driver waits indefinitely for the fifo occupancy to go below a threshold
as soon as the pacing interrupt is received. This can cause soft lockup on
one of the processors, if the rate of DB is very high.

Add a loop count for FPGA and exit the __wait_for_fifo_occupancy_below_th
if the loop is taking more time. Pacing will be continuing until the
occupancy is below the threshold. This is ensured by the checks in
bnxt_re_pacing_timer_exp and further scheduling the work for pacing based
on the fifo occupancy.

Fixes: 2ad4e6303a6d ("RDMA/bnxt_re: Implement doorbell pacing algorithm")
Link: https://patch.msgid.link/r/1728373302-19530-7-git-send-email-selvin.xavier@broadcom.com
Reviewed-by: Kalesh AP 
Reviewed-by: Chandramohan Akula 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe

RDMA/bnxt_re: Fix a possible NULL pointer dereference

2024-10-11T23:49:01+00:00

There is a possibility of a NULL pointer dereference in the failure path
of bnxt_re_add_device().  To address that, moved the update of
"rdev->adev" to bnxt_re_dev_add().

Fixes: dee3da3422d5 ("RDMA/bnxt_re: Change aux driver data to en_info to hold more information")
Link: https://patch.msgid.link/r/1728373302-19530-6-git-send-email-selvin.xavier@broadcom.com
Reported-by: Dan Carpenter 
Closes: https://lore.kernel.org/linux-rdma/CAH-L+nMCwymKGqf5pd8-FZNhxEkDD=kb6AoCaE6fAVi7b3e5Qw@mail.gmail.com/T/#t
Signed-off-by: Kalesh AP 
Signed-off-by: Selvin Xavier 
Signed-off-by: Jason Gunthorpe