Age | Commit message (Collapse) | Author |
|
The TLS TIS object contains the dek/key ID.
By destroying the key first, the TIS would contain an invalid
non-existing key ID.
Reverse the destroy order, this also acheives the desired assymetry
between the destroy and the create flows.
Fixes: d2ead1f360e8 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
After changing the parent_id to be the same for both NICs of same
The cited commit wrongly allow offload of tc redirect flows from
VF to uplink and vice versa when devcies are on different eswitch,
these cases aren't supported by HW.
Disallow the above offloads when devcies are on different eswitch
and VF LAG is not configured.
Fixes: f6dc1264f1c0 ("net/mlx5e: Disallow tc redirect offload cases we don't support")
Signed-off-by: Maor Dickman <maord@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
When driver is reloading during recovery flow, it can't get new commands
till command interface is up again. Otherwise we may get to null pointer
trying to access non initialized command structures.
Add cmdif state to avoid processing commands while cmdif is not ready.
Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
After driver creates (via FW command) an EQ for commands, the driver will
be informed on new commands completion by EQE. However, due to a race in
driver's internal command mode metadata update, some new commands will
still be miss-handled by driver as if we are in polling mode. Such commands
can get two non forced completion, leading to already freed command entry
access.
CREATE_EQ command, that maps EQ to the command queue must be posted to the
command queue while it is empty and no other command should be posted.
Add SW mechanism that once the CREATE_EQ command is about to be executed,
all other commands will return error without being sent to the FW. Allow
sending other commands only after successfully changing the driver's
internal command mode metadata.
We can safely return error to all other commands while creating the command
EQ, as all other commands might be sent from the user/application during
driver load. Application can rerun them later after driver's load was
finished.
Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
When FW response to commands is very slow and all command entries in
use are waiting for completion we can have a race where commands can get
timeout before they get out of the queue and handled. Timeout
completion on uninitialized command will cause releasing command's
buffers before accessing it for initialization and then we will get NULL
pointer exception while trying access it. It may also cause releasing
buffers of another command since we may have timeout completion before
even allocating entry index for this command.
Add entry handling completion to avoid this race.
Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Need to allocate the q counters before init_rx which needs them
when creating the rq.
Fixes: 8520fa57a4e9 ("net/mlx5e: Create q counters on uplink representors")
Signed-off-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Processing commands by cmd_work_handler() while already in Internal
Error State will result in entry leak, since the handler process force
completion without doorbell. Forced completion doesn't release the entry
and event completion will never arrive, so entry should be released.
Fixes: 73dd3a4839c1 ("net/mlx5: Avoid using pending command interface slots")
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
mlx5_cmd_flush() will trigger forced completions to all valid command
entries. Triggered by an asynch event such as fast teardown it can
happen at any stage of the command, including command initialization.
It will trigger forced completion and that can lead to completion on an
uninitialized command entry.
Setting MLX5_CMD_ENT_STATE_PENDING_COMP only after command entry is
initialized will ensure force completion is treated only if command
entry is initialized.
Fixes: 73dd3a4839c1 ("net/mlx5: Avoid using pending command interface slots")
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
In polling mode, set arm_db member to a value that will avoid CQ
event recovery by the HW.
Otherwise we might get event without completion function.
In addition,empty completion function to was added to protect from
unexpected events.
Fixes: 297cccebdc5a ("net/mlx5: DR, Expose an internal API to issue RDMA operations")
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Alex Vesker <valex@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
In cited patch mutex is initialized after its used.
Below call trace is observed.
Fix the order to initialize the mutex early enough.
Similarly follow mirror sequence during cleanup.
kernel: DEBUG_LOCKS_WARN_ON(lock->magic != lock)
kernel: WARNING: CPU: 5 PID: 45916 at kernel/locking/mutex.c:938
__mutex_lock+0x7d6/0x8a0
kernel: Call Trace:
kernel: ? esw_vport_tbl_get+0x3b/0x250 [mlx5_core]
kernel: ? mark_held_locks+0x55/0x70
kernel: ? __slab_free+0x274/0x400
kernel: ? lockdep_hardirqs_on+0x140/0x1d0
kernel: esw_vport_tbl_get+0x3b/0x250 [mlx5_core]
kernel: ? mlx5_esw_chains_create_fdb_prio+0xa57/0xc20 [mlx5_core]
kernel: mlx5_esw_vport_tbl_get+0x88/0xf0 [mlx5_core]
kernel: mlx5_esw_chains_create+0x2f3/0x3e0 [mlx5_core]
kernel: esw_create_offloads_fdb_tables+0x11d/0x580 [mlx5_core]
kernel: esw_offloads_enable+0x26d/0x540 [mlx5_core]
kernel: mlx5_eswitch_enable_locked+0x155/0x860 [mlx5_core]
kernel: mlx5_devlink_eswitch_mode_set+0x1af/0x320 [mlx5_core]
kernel: devlink_nl_cmd_eswitch_set_doit+0x41/0xb0
Fixes: 96e326878fa5 ("net/mlx5e: Eswitch, Use per vport tables for mirroring")
Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
When mlx5_modify_header_alloc() fails, instead of printing the error
value returned, current error log prints 0.
Fix by printing correct error value returned by
mlx5_modify_header_alloc().
Fixes: 6724e66b90ee ("net/mlx5: E-Switch, Get reg_c1 value on miss")
Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Error unwinding is done incorrectly in the cited commit.
When steering init fails, there is no need to perform steering cleanup.
When vport error exists, error cleanup should be mirror of the setup
routine, i.e. to perform steering cleanup before metadata cleanup.
This avoids the call trace in accessing uninitialized objects which are
skipped during steering_init() due to failure in steering_init().
Call trace:
mlx5_cmd_modify_header_alloc:805:(pid 21128): too many modify header
actions 1, max supported 0
E-Switch: Failed to create restore mod header
BUG: kernel NULL pointer dereference, address: 00000000000000d0
[ 677.263079] mlx5_destroy_flow_group+0x13/0x80 [mlx5_core]
[ 677.268921] esw_offloads_steering_cleanup+0x51/0xf0 [mlx5_core]
[ 677.275281] esw_offloads_enable+0x1a5/0x800 [mlx5_core]
[ 677.280949] mlx5_eswitch_enable_locked+0x155/0x860 [mlx5_core]
[ 677.287227] mlx5_devlink_eswitch_mode_set+0x1af/0x320
[ 677.293741] devlink_nl_cmd_eswitch_set_doit+0x41/0xb0
[ 677.299217] genl_rcv_msg+0x1eb/0x430
Fixes: 7983a675ba65 ("net/mlx5: E-Switch, Enable chains only if regs loopback is enabled")
Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
In the switchdev mode, when running "cat
/sys/class/net/NIC/statistics/tx_packets", the ppcnt register is
accessed to get the latest values. But currently this command can
not get the correct values from ppcnt.
From firmware manual, before getting the 802_3 counters, the 802_3
data layout should be set to the ppcnt register.
When the command "cat /sys/class/net/NIC/statistics/tx_packets" is
run, before updating 802_3 data layout with ppcnt register, the
monitor counters are tested. The test result will decide the
802_3 data layout is updated or not.
Actually the monitor counters do not support to monitor rx/tx
stats of 802_3 in switchdev mode. So the rx/tx counters change
will not trigger monitor counters. So the 802_3 data layout will
not be updated in ppcnt register. Finally this command can not get
the latest values from ppcnt register with 802_3 data layout.
Fixes: 5c7e8bbb0257 ("net/mlx5e: Use monitor counters for update stats")
Signed-off-by: Zhu Yanjun <yanjunz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
MLX5_CORE uses the 'imply' keyword to depend on VXLAN, PTP_1588_CLOCK,
MLXFW and PCI_HYPERV_INTERFACE.
This was useful to force vxlan, ptp, etc.. to be reachable to mlx5
regardless of their config states.
Due to the changes in the cited commit below, the semantics of 'imply'
was changed to not force any restriction on the implied config.
As a result of this change, the compilation of MLX5_CORE=y and VXLAN=m
would result in undefined references, as VXLAN now would stay as 'm'.
To fix this we change MLX5_CORE to have a weak dependency on
these modules/configs and make sure they are reachable, by adding:
depend on symbol || !symbol.
For example: VXLAN=m MLX5_CORE=y, this will force MLX5_CORE to m
Fixes: def2fbffe62c ("kconfig: allow symbols implied by y to become m")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: Masahiro Yamada <masahiroy@kernel.org>
Cc: Nicolas Pitre <nico@fluxnic.net>
Cc: Arnd Bergmann <arnd@arndb.de>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
|
|
XSK wakeup function triggers NAPI by posting a NOP WQE to a special XSK
ICOSQ. When the application floods the driver with wakeup requests by
calling sendto() in a certain pattern that ends up in mlx5e_trigger_irq,
the XSK ICOSQ may overflow.
Multiple NOPs are not required and won't accelerate the process, so
avoid posting a second NOP if there is one already on the way. This way
we also avoid increasing the queue size (which might not help anyway).
Fixes: db05815b36cb ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
After allowing parallel tuple insertion, we get the following trace:
[ 5505.142249] ------------[ cut here ]------------
[ 5505.148155] WARNING: CPU: 21 PID: 13313 at lib/radix-tree.c:581 delete_node+0x16c/0x180
[ 5505.295553] CPU: 21 PID: 13313 Comm: kworker/u50:22 Tainted: G OE 5.6.0+ #78
[ 5505.304824] Hardware name: Supermicro Super Server/X10DRT-P, BIOS 2.0b 03/30/2017
[ 5505.313740] Workqueue: nf_flow_table_offload flow_offload_work_handler [nf_flow_table]
[ 5505.323257] RIP: 0010:delete_node+0x16c/0x180
[ 5505.349862] RSP: 0018:ffffb19184eb7b30 EFLAGS: 00010282
[ 5505.356785] RAX: 0000000000000000 RBX: ffff904ac95b86d8 RCX: ffff904b6f938838
[ 5505.365190] RDX: 0000000000000000 RSI: ffff904ac954b908 RDI: ffff904ac954b920
[ 5505.373628] RBP: ffff904b4ac13060 R08: 0000000000000001 R09: 0000000000000000
[ 5505.382155] R10: 0000000000000000 R11: 0000000000000040 R12: 0000000000000000
[ 5505.390527] R13: ffffb19184eb7bfc R14: ffff904b6bef5800 R15: ffff90482c1203c0
[ 5505.399246] FS: 0000000000000000(0000) GS:ffff904c2fc80000(0000) knlGS:0000000000000000
[ 5505.408621] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 5505.415739] CR2: 00007f5d27006010 CR3: 0000000058c10006 CR4: 00000000001626e0
[ 5505.424547] Call Trace:
[ 5505.428429] idr_alloc_u32+0x7b/0xc0
[ 5505.433803] mlx5_tc_ct_entry_add_rule+0xbf/0x950 [mlx5_core]
[ 5505.441354] ? mlx5_fc_create+0x23c/0x370 [mlx5_core]
[ 5505.448225] mlx5_tc_ct_block_flow_offload+0x874/0x10b0 [mlx5_core]
[ 5505.456278] ? mlx5_tc_ct_block_flow_offload+0x63d/0x10b0 [mlx5_core]
[ 5505.464532] nf_flow_offload_tuple.isra.21+0xc5/0x140 [nf_flow_table]
[ 5505.472286] ? __kmalloc+0x217/0x2f0
[ 5505.477093] ? flow_rule_alloc+0x1c/0x30
[ 5505.482117] flow_offload_work_handler+0x1d0/0x290 [nf_flow_table]
[ 5505.489674] ? process_one_work+0x17c/0x580
[ 5505.494922] process_one_work+0x202/0x580
[ 5505.500082] ? process_one_work+0x17c/0x580
[ 5505.505696] worker_thread+0x4c/0x3f0
[ 5505.510458] kthread+0x103/0x140
[ 5505.514989] ? process_one_work+0x580/0x580
[ 5505.520616] ? kthread_bind+0x10/0x10
[ 5505.525837] ret_from_fork+0x3a/0x50
[ 5505.570841] ---[ end trace 07995de9c56d6831 ]---
This happens from parallel deletes/adds to idr, as idr isn't protected.
Fix that by using xarray as the tuple_ids allocator instead of idr.
Fixes: 7da182a998d6 ("netfilter: flowtable: Use work entry per offload command")
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
On s390 FORCE_MAX_ZONEORDER is 9 instead of 11, thus a larger kzalloc()
allocation as done for the firmware tracer will always fail.
Looking at mlx5_fw_tracer_save_trace(), it is actually the driver itself
that copies the debug data into the trace array and there is no need for
the allocation to be contiguous in physical memory. We can therefor use
kvzalloc() instead of kzalloc() and get rid of the large contiguous
allcoation.
Fixes: f53aaa31cce7 ("net/mlx5: FW tracer, implement tracer logic")
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Fixes CT entries list corruption.
After allowing parallel insertion/removals in upper nf flow table
layer, unprotected ct entries list can be corrupted by parallel add/del
on the same flow table.
CT entries list is only used while freeing a ct zone flow table to
go over all the ct entries offloaded on that zone/table, and flush
the table.
As rhashtable already provides an api to go over all the inserted entries,
fix the race by using the rhashtable iteration instead, and remove the list.
Fixes: 7da182a998d6 ("netfilter: flowtable: Use work entry per offload command")
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
In cited commit netdevice is registered after devlink port.
Unregistration flow should be mirror sequence of registration flow.
Hence, unregister netdevice before devlink port.
Fixes: 31e87b39ba9d ("net/mlx5e: Fix devlink port register sequence")
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Cited patch missed to extract PCI pf number accurately for PF and VF
port flavour. It considered PCI device + function number.
Due to this, device having non zero device number shown large pfnum.
Hence, use only PCI function number; to avoid similar errors, derive
pfnum one time for all port flavours.
Fixes: f60f315d339e ("net/mlx5e: Register devlink ports for physical link, PCI PF, VFs")
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
With ct clear action we should not allocate the action in hw
and not release the mod_acts parsed in advance.
It will be done when handling the ct clear action.
Fixes: 1ef3018f5af3 ("net/mlx5e: CT: Support clear action")
Signed-off-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Current value of nest_level, assigned from net_device lower_level value,
does not reflect the actual number of vlan headers, needed to pop.
For ex., if we have untagged ingress traffic sended over vlan devices,
instead of one pop action, driver will perform two pop actions.
To fix that, calculate nest_level as difference between vlan device and
parent device lower_levels.
Fixes: f3b0a18bb6cb ("net: remove unnecessary variables and callback")
Signed-off-by: Dmytro Linkin <dmitrolin@mellanox.com>
Signed-off-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Once driver finishes flashing the firmware image, it should release it.
Fixes: 9c8bca2637b8 ("mlx5: Move firmware flash implementation to devlink")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
When we destroy rules from slow path we need to avoid destroying
termination tables since termination tables are never created in slow
path. By doing so we avoid destroying the termination table created for the
slow path.
Fixes: d8a2034f152a ("net/mlx5: Don't use termination tables in slow path")
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
High frequency of PCI ioread calls during recovery flow may cause the
following trace on powerpc:
[ 248.670288] EEH: 2100000 reads ignored for recovering device at
location=Slot1 driver=mlx5_core pci addr=0000:01:00.1
[ 248.670331] EEH: Might be infinite loop in mlx5_core driver
[ 248.670361] CPU: 2 PID: 35247 Comm: kworker/u192:11 Kdump: loaded
Tainted: G OE ------------ 4.14.0-115.14.1.el7a.ppc64le #1
[ 248.670425] Workqueue: mlx5_health0000:01:00.1 health_recover_work
[mlx5_core]
[ 248.670471] Call Trace:
[ 248.670492] [c00020391c11b960] [c000000000c217ac] dump_stack+0xb0/0xf4
(unreliable)
[ 248.670548] [c00020391c11b9a0] [c000000000045818]
eeh_check_failure+0x5c8/0x630
[ 248.670631] [c00020391c11ba50] [c00000000068fce4]
ioread32be+0x114/0x1c0
[ 248.670692] [c00020391c11bac0] [c00800000dd8b400]
mlx5_error_sw_reset+0x160/0x510 [mlx5_core]
[ 248.670752] [c00020391c11bb60] [c00800000dd75824]
mlx5_disable_device+0x34/0x1d0 [mlx5_core]
[ 248.670822] [c00020391c11bbe0] [c00800000dd8affc]
health_recover_work+0x11c/0x3c0 [mlx5_core]
[ 248.670891] [c00020391c11bc80] [c000000000164fcc]
process_one_work+0x1bc/0x5f0
[ 248.670955] [c00020391c11bd20] [c000000000167f8c]
worker_thread+0xac/0x6b0
[ 248.671015] [c00020391c11bdc0] [c000000000171618] kthread+0x168/0x1b0
[ 248.671067] [c00020391c11be30] [c00000000000b65c]
ret_from_kernel_thread+0x5c/0x80
Reduce the PCI ioread frequency during recovery by using msleep()
instead of cond_resched()
Fixes: 3e5b72ac2f29 ("net/mlx5: Issue SW reset on FW assert")
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Pull rdma updates from Jason Gunthorpe:
"The majority of the patches are cleanups, refactorings and clarity
improvements.
This cycle saw some more activity from Syzkaller, I think we are now
clean on all but one of those bugs, including the long standing and
obnoxious rdma_cm locking design defect. Continue to see many drivers
getting cleanups, with a few new user visible features.
Summary:
- Various driver updates for siw, bnxt_re, rxe, efa, mlx5, hfi1
- Lots of cleanup patches for hns
- Convert more places to use refcount
- Aggressively lock the RDMA CM code that syzkaller says isn't
working
- Work to clarify ib_cm
- Use the new ib_device lifecycle model in bnxt_re
- Fix mlx5's MR cache which seems to be failing more often with the
new ODP code
- mlx5 'dynamic uar' and 'tx steering' user interfaces"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (144 commits)
RDMA/bnxt_re: make bnxt_re_ib_init static
IB/qib: Delete struct qib_ivdev.qp_rnd
RDMA/hns: Fix uninitialized variable bug
RDMA/hns: Modify the mask of QP number for CQE of hip08
RDMA/hns: Reduce the maximum number of extend SGE per WQE
RDMA/hns: Reduce PFC frames in congestion scenarios
RDMA/mlx5: Add support for RDMA TX flow table
net/mlx5: Add support for RDMA TX steering
IB/hfi1: Call kobject_put() when kobject_init_and_add() fails
IB/hfi1: Fix memory leaks in sysfs registration and unregistration
IB/mlx5: Move to fully dynamic UAR mode once user space supports it
IB/mlx5: Limit the scope of struct mlx5_bfreg_info to mlx5_ib
IB/mlx5: Extend QP creation to get uar page index from user space
IB/mlx5: Extend CQ creation to get uar page index from user space
IB/mlx5: Expose UAR object and its alloc/destroy commands
IB/hfi1: Get rid of a warning
RDMA/hns: Remove redundant judgment of qp_type
RDMA/hns: Remove redundant assignment of wc->smac when polling cq
RDMA/hns: Remove redundant qpc setup operations
RDMA/hns: Remove meaningless prints
...
|
|
When health reporter is registered to devlink, devlink will implicitly set
auto recover if and only if the reporter has a recover method. No reason
to explicitly get the auto recover flag from the driver.
Remove this flag from all drivers that called
devlink_health_reporter_create.
All existing health reporters set auto recovery to true if they have a
recover method.
Yet, administrator can unset auto recover via netlink command as prior to
this patch.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
It may be up to the driver (in case ANY HW stats is passed) to select
which type of HW stats he is going to use. Add an infrastructure to
expose this information to user.
$ tc filter add dev enp3s0np1 ingress proto ip handle 1 pref 1 flower dst_ip 192.168.1.1 action drop
$ tc -s filter show dev enp3s0np1 ingress
filter protocol ip pref 1 flower chain 0
filter protocol ip pref 1 flower chain 0 handle 0x1
eth_type ipv4
dst_ip 192.168.1.1
in_hw in_hw_count 2
action order 1: gact action drop
random type none pass val 0
index 1 ref 1 bind 1 installed 10 sec used 10 sec
Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0
used_hw_stats immediate <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add mlx5e_rep_indr_setup_ft_cb to support indr block setup
in FT mode.
Both tc rules and flow table rules are of the same format,
It can re-use tc parsing for that, and move the flow table rules
to their steering domain(the specific chain_index), the indr
block offload in FT also follow this scenario.
Signed-off-by: wenxu <wenxu@ucloud.cn>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Refactor indr setup block for support ft indr setup in the
next patch. The function mlx5e_rep_indr_offload exposes
'flags' in order set additional flag for FT in next patch.
Rename mlx5e_rep_indr_setup_tc_block to mlx5e_rep_indr_setup_block
and add flow_setup_cb_t callback parameters in order set the
specific callback for FT in next patch.
Signed-off-by: wenxu <wenxu@ucloud.cn>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
eswitch_offloads_chains.{c,h} were just introduced this kernel release
cycle, eswitch is in high development demand right now and many
features are planned to be added to it. eswitch deserves its own
directory and here we move these new files to there, in preparation for
upcoming eswitch features and new files.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
|
|
In VF lag mode when remove the bonding module without bring down the
bond device first, we could potentially have circular dependency when we
unload IB devices and also handle fib events:
1. The bond work starts first;
2. The "modprobe -rv bonding" process tries to release the bond device,
with the "pernet_ops_rwsem" lock hold;
3. The bond work blocks in unregister_netdevice_notifier() and waits for
the lock because fib event came right before;
4. The kernel fib module tries to free all the fib entries by broadcasting
the "FIB_EVENT_NH_DEL" event;
5. Upon the fib event this lag_mp module holds the fib lock and queue a
fib work.
So:
bond work -> modprobe task -> kernel fib module -> lag_mp -> bond work
Today we either reload IB devices in roce lag in nic mode or either handle
fib events in switchdev mode, but a new feature could change that we'll
need to reload IB devices also in switchdev mode so this is a future proof
fix as one may not notice this later.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
* 'mlx5-next' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux:
mlx5: Remove uninitialized use of key in mlx5_core_create_mkey
{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
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Leon Romanovsky says:
====================
Those two patches from Michael extends mlx5_core and mlx5_ib flow steering
to support RDMA TX in similar way to already supported RDMA RX.
====================
Based on the mlx5-next branch at
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Due to dependencies
* branch 'mlx5_tx_steering':
RDMA/mlx5: Add support for RDMA TX flow table
net/mlx5: Add support for RDMA TX steering
|
|
Add new RDMA TX flow steering namespace. Flow steering rules in
this namespace are used to filter transmitted RDMA traffic.
Link: https://lore.kernel.org/r/20200324061425.1570190-2-leon@kernel.org
Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Reviewed-by: Maor Gottlieb <maorg@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>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2020-03-25
1) Cleanups from Dan Carpenter and wenxu.
2) Paul and Roi, Some minor updates and fixes to E-Switch to address
issues introduced in the previous reg_c0 updates series.
3) Eli Cohen simplifies and improves flow steering matching group searches
and flow table entries version management.
4) Parav Pandit, improves devlink eswitch mode changes thread safety.
By making devlink rely on driver for thread safety and introducing mlx5
eswitch mode change protection.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently eswitch mode change is occurring from 2 different execution
contexts as below.
1. sriov sysfs enable/disable
2. devlink eswitch set commands
Both of them need to access eswitch related data structures in
synchronized manner.
Without any synchronization below race condition exist.
SR-IOV enable/disable with devlink eswitch mode change:
cpu-0 cpu-1
----- -----
mlx5_device_disable_sriov() mlx5_devlink_eswitch_mode_set()
mlx5_eswitch_disable() esw_offloads_stop()
esw_offloads_disable() mlx5_eswitch_disable()
esw_offloads_disable()
Hence, they are synchronized using a new mode_lock.
eswitch's state_lock is not used as it can lead to a deadlock scenario
below and state_lock is only for vport and fdb exclusive access.
ip link set vf <param>
netlink rcv_msg() - Lock A
rtnl_lock
vfinfo()
esw->state_lock() - Lock B
devlink eswitch_set
devlink_mutex
esw->state_lock() - Lock B
attach_netdev()
register_netdev()
rtnl_lock - Lock A
Alternatives considered:
1. Acquiring rtnl lock before taking esw->state_lock to follow similar
locking sequence as ip link flow during eswitch mode set.
rtnl lock is not good idea for two reasons.
(a) Holding rtnl lock for several hundred device commands is not good
idea.
(b) It leads to below and more similar deadlocks.
devlink eswitch_set
devlink_mutex
rtnl_lock - Lock A
esw->state_lock() - Lock B
eswitch_disable()
reload()
ib_register_device()
ib_cache_setup_one()
rtnl_lock()
2. Exporting devlink lock may lead to undesired use of it in vendor
driver(s) in future.
3. Unloading representors outside of the mode_lock requires
serialization with other process trying to enable the eswitch.
4. Differing the representors life cycle to a different workqueue
requires synchronization with func_change_handler workqueue.
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Bodong Wang <bodong@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Subsequent patch protects eswitch mode changes across sriov and devlink
interfaces. It is desirable for eswitch to provide thread safe eswitch
enable and disable APIs.
Hence, extend eswitch enable API to optionally update num_vfs when
requested.
In subsequent patch, eswitch num_vfs are updated after all the eswitch
users eswitch drops its reference count.
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Bodong Wang <bodong@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
In order to check eswitch state under a lock, prepare code to split
capability check and eswitch state check into two helper functions.
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Bodong Wang <bodong@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
mlx5_unload_one() always returns 0.
Simplify callers of mlx5_unload_one() and remove the dead code.
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
mlx5_register_device() doesn't check for any error and always returns 0.
Simplify mlx5_register_device() to return void and its caller, remove
dead code related to it.
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Group version is used when modifying a rule is allowed
(FLOW_ACT_NO_APPEND is clear) to detect a case where the rule was found
but while the groups where unlocked a new FTE was added. In this case,
the added FTE could be one with identical match value so we need to
attempt again with group lock held.
Change the code so version is retrieved only when FLOW_ACT_NO_APPEND is
cleared. As result, later compare can also be avoided if FLOW_ACT_NO_APPEND
is cleared.
Also improve comments text.
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
FTE version is not used anywhere in the code so avoid incrementing it.
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
When adding a rule to a flow group we need increment the version of the
group. Current code fails to do that and as a result, when trying to add
a rule, we will fail to discover a case where an FTE with the same match
value was added while we scanned the groups of the same match criteria,
thus we may try to add an FTE that was already added.
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Instead of using two different structs for searching groups with the
same match, use a single struct and thus simplify the code, make it more
readable and smaller size which means less code cache misses.
text data bss dec hex
before: 35524 2744 0 38268 957c
after: 35038 2744 0 37782 9396
When testing add 70000 rules, delete all the rules, and repeat three
times taking the average, we get (time in seconds):
Before the change: insert 16.80, delete 11.02
After the change: insert 16.55, delete 10.95
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Reviewed-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
The correct type is u32.
Fixes: d18296ffd9cc ("net/mlx5: E-Switch, Introduce global tables")
Signed-off-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
We allocate a temporary memory but forget to free it.
Fixes: 11b717d61526 ("net/mlx5: E-Switch, Get reg_c0 value on CQE")
Signed-off-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Register c0 loopback is needed to fully support chains and prios.
Enable chains and prio only if loopback (of reg c1 which came together
with c0), is enabled. To be able to check that, move enabling of loopback
before eswitch chains init.
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
Reg c0/c1 matching, rewrite of regs c0/c1, and copy header of regs c1,B
is needed for the restore table to function, might not be supported by
firmware, and creation of the restore table or the copy header will
fail.
Check reg_c1 loopback support, as firmware which supports this,
should have all of the above.
Fixes: 11b717d61526 ("net/mlx5: E-Switch, Get reg_c0 value on CQE")
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|
|
The function mlx5e_rep_setup_ft_cb check chain_index is zero twice.
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
|