When a large number of tasks are issued, the speed of HW processing
mbx will slow down. The standard for judging mbx timeout in the current
firmware is 30ms, and the current timeout standard for the driver is also
30ms.
Considering that firmware scheduling in multi-function scenarios takes a
certain amount of time, this will cause the driver to time out too early
and report a failure before mbx execution times out.
This patch introduces a new mechanism that can set different timeouts for
different cmds and extends the timeout of mbx to 35ms.
Fixes: a04ff739f2 ("RDMA/hns: Add command queue support for hip08 RoCE driver")
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://lore.kernel.org/r/20240710133705.896445-9-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
CEQEs are handled in interrupt handler currently. This may cause the
CPU core staying in interrupt context too long and lead to soft lockup
under heavy load.
Handle CEQEs in BH workqueue and set an upper limit for the number of
CEQE handled by a single call of work handler.
Fixes: a5073d6054 ("RDMA/hns: Add eq support of hip08")
Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://lore.kernel.org/r/20240710133705.896445-3-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Fix the problem of KASAN causing the stack frame size to increase
drivers/infiniband/hw/ocrdma/ocrdma_stats.c:686:16: error: stack frame size (20664) exceeds limit (8192) in 'ocrdma_dbgfs_ops_read' [-Werror,-Wframe-larger-than]
static ssize_t ocrdma_dbgfs_ops_read(struct file *filp, char __user *buffer,
^
Some functions called by ocrdma_dbgfs_ops_read occupy a lot of stack space.
Mark these functions as noinline_for_stack to prevent them from accumulating
in ocrdma_dbgfs_ops_read.
Signed-off-by: Peng Hao <flyingpeng@tencent.com>
Link: https://lore.kernel.org/r/20240710091657.26291-1-flyingpeng@tencent.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
The name_assign_type indicates how the name is provided. Currently
these types are supported:
- RDMA_NAME_ASSIGN_TYPE_UNKNOWN: Unknown or not set;
- RDMA_NAME_ASSIGN_TYPE_USER: Name is provided by the user; The
user-created sub device, rxe and siw device has this type.
When filling nl device info, it is set in the new attribute
RDMA_NLDEV_ATTR_NAME_ASSIGN_TYPE. User-space tools like udev
"rdma_rename" could check this attribute to determine if this
device needs to be renamed or not.
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Link: https://lore.kernel.org/r/522591bef9a369cc8e5dcb77787e017bffee37fe.1719837610.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Reduce nodename string size to fit IB_DEVICE_NODE_DESC_MAX.
drivers/infiniband/hw/qib/qib_verbs.c: In function ‘qib_register_ib_device’:
drivers/infiniband/hw/qib/qib_verbs.c:1554:40: error: ‘%s’ directive output may be truncated writing up to 64 bytes into a region of size 43
[-Werror=format-truncation=]
1554 | "Intel Infiniband HCA %s", init_utsname()->nodename);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/infiniband/hw/qib/qib_verbs.c:1553:9: note: ‘snprintf’ output between 22 and 86 bytes into a destination of size 64
1553 | snprintf(ibdev->node_desc, sizeof(ibdev->node_desc),
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1554 | "Intel Infiniband HCA %s", init_utsname()->nodename);
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
Link: https://lore.kernel.org/r/1fb6393fa2e0702fef995834c3c7db972bbc4d06.1719837715.git.leon@kernel.org
Acked-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
drivers/infiniband/hw/qib/qib_init.c: In function ‘qib_init_one’:
drivers/infiniband/hw/qib/qib_init.c:586:67: error: ‘%d’ directive output may be truncated writing between 1 and 11 bytes into a region of size between 0 and 3 [-Werror=format-truncation=]
586 | snprintf(wq_name, sizeof(wq_name), "qib%d_%d",
| ^~
In function ‘qib_create_workqueues’,
inlined from ‘qib_init_one’ at drivers/infiniband/hw/qib/qib_init.c:1438:8:
drivers/infiniband/hw/qib/qib_init.c:586:60: note: directive argument in the range [-2147483643, 254]
586 | snprintf(wq_name, sizeof(wq_name), "qib%d_%d",
| ^~~~~~~~~~
drivers/infiniband/hw/qib/qib_init.c:586:25: note: ‘snprintf’ output between 7 and 27 bytes into a destination of size 8
586 | snprintf(wq_name, sizeof(wq_name), "qib%d_%d",
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
587 | dd->unit, pidx);
| ~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
Link: https://lore.kernel.org/r/ab5222c414a01e9d2c5129ef26836aace9ee2aa5.1719837715.git.leon@kernel.org
Acked-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
This patch supports driver APIs "add_sub_dev" and "del_sub_dev", to
add and delete a plane device respectively.
A mlx5 plane device is a rdma SMI device; It provides the SMI capability
through user MAD for it's parent, the logical multi-plane aggregated
device. For a plane port:
- It supports QP0 only;
- When adding a plane device, all plane ports are added;
- For some commands like mad_ifc, both plane_index and native portnum
is needed;
- When querying or modifying a plane port context, the native portnum
must be used, as the query/modify_hca_vport_context command doesn't
support plane port.
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Link: https://lore.kernel.org/r/e933cd0562aece181f8657af2ca0f5b387d0f14e.1718553901.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
This patch adds 2 APIs, as well as driver operations to support adding
and deleting an IB sub device, which provides part of functionalities
of it's parent.
A sub device has a type; for a sub device with type "SMI", it provides
the smi capability through umad for its parent, meaning uverb is not
supported.
A sub device cannot live without a parent. So when a parent is
released, all it's sub devices are released as well.
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Link: https://lore.kernel.org/r/44253f7508b21eb2caefea3980c2bc072869116c.1718553901.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
'struct mmu_rb_ops' is not modified in this driver.
Constifying this structure moves some data to a read-only section, so
increase overall security.
On a x86_64, with allmodconfig, as an example:
Before:
======
text data bss dec hex filename
10879 164 0 11043 2b23 drivers/infiniband/hw/hfi1/pin_system.o
After:
=====
text data bss dec hex filename
10907 140 0 11047 2b27 drivers/infiniband/hw/hfi1/pin_system.o
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/b826dd05eefa5f4d6a7a1b4d191eaf37c714ed04.1719259997.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
To avoid leakage for QPs assocoated with SRQ, according to IB spec
(section 10.3.1):
"Note, for QPs that are associated with an SRQ, the Consumer should take
the QP through the Error State before invoking a Destroy QP or a Modify
QP to the Reset State. The Consumer may invoke the Destroy QP without
first performing a Modify QP to the Error State and waiting for the Affiliated
Asynchronous Last WQE Reached Event. However, if the Consumer
does not wait for the Affiliated Asynchronous Last WQE Reached Event,
then WQE and Data Segment leakage may occur. Therefore, it is good
programming practice to teardown a QP that is associated with an SRQ
by using the following process:
- Put the QP in the Error State;
- wait for the Affiliated Asynchronous Last WQE Reached Event;
- either:
- drain the CQ by invoking the Poll CQ verb and either wait for CQ
to be empty or the number of Poll CQ operations has exceeded
CQ capacity size; or
- post another WR that completes on the same CQ and wait for this
WR to return as a WC;
- and then invoke a Destroy QP or Reset QP."
Catch the Last WQE Reached Event in the core layer during drain QP flow.
Signed-off-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Link: https://lore.kernel.org/r/20240619171153.34631-2-mgurtovoy@nvidia.com
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
When creating a QP, one of the attributes is TS format (timestamp).
In some devices, we have a limitation that all QPs should have the same
ts_format. The ts_format is chosen based on the device's capability.
The qp_ts_format cap resides under the RoCE caps table, and the
cap will be 0 when RoCE is disabled. So when RoCE is disabled, the
value that should be queried is sq_ts_format under HCA caps.
Consider the case when the system supports REAL_TIME_TS format (0x2),
some QPs are created with REAL_TIME_TS as ts_format, and afterwards
RoCE gets disabled. When trying to construct a new QP, we can't use
the qp_ts_format, that is queried from the RoCE caps table, Since it
leads to passing 0x0 (FREE_RUNNING_TS) as the value of the qp_ts_format,
which is different than the ts_format of the previously allocated
QPs REAL_TIME_TS format (0x2).
Thus, to resolve this, read the sq_ts_format, which also reflect
the supported ts format for the QP when RoCE is disabled.
Fixes: 4806f1e2fe ("net/mlx5: Set QP timestamp mode to default")
Signed-off-by: Maher Sanalla <msanalla@nvidia.com>
Signed-off-by: Or Har-Toov <ohartoov@nvidia.com>
Link: https://lore.kernel.org/r/32801966eb767c7fd62b8dea3b63991d5fbfe213.1718554199.git.leon@kernel.org
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
drivers/infiniband/hw/mlx4/alias_GUID.c: In function ‘mlx4_ib_init_alias_guid_service’:
drivers/infiniband/hw/mlx4/alias_GUID.c:878:74: error: ‘%d’ directive
output may be truncated writing between 1 and 11 bytes into a region of
size 5 [-Werror=format-truncation=]
878 | snprintf(alias_wq_name, sizeof alias_wq_name, "alias_guid%d", i);
| ^~
drivers/infiniband/hw/mlx4/alias_GUID.c:878:63: note: directive argument in the range [-2147483641, 2147483646]
878 | snprintf(alias_wq_name, sizeof alias_wq_name, "alias_guid%d", i);
| ^~~~~~~~~~~~~~
drivers/infiniband/hw/mlx4/alias_GUID.c:878:17: note: ‘snprintf’ output
between 12 and 22 bytes into a destination of size 15
878 | snprintf(alias_wq_name, sizeof alias_wq_name, "alias_guid%d", i);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
Fixes: a0c64a17ab ("mlx4: Add alias_guid mechanism")
Link: https://lore.kernel.org/r/1951c9500109ca7e36dcd523f8a5f2d0d2a608d1.1718554641.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Increase size of the name array to avoid truncated output warning.
drivers/infiniband/hw/mlx4/mad.c: In function ‘mlx4_ib_alloc_demux_ctx’:
drivers/infiniband/hw/mlx4/mad.c:2197:47: error: ‘%d’ directive output
may be truncated writing between 1 and 11 bytes into a region of size 4
[-Werror=format-truncation=]
2197 | snprintf(name, sizeof(name), "mlx4_ibt%d", port);
| ^~
drivers/infiniband/hw/mlx4/mad.c:2197:38: note: directive argument in
the range [-2147483645, 2147483647]
2197 | snprintf(name, sizeof(name), "mlx4_ibt%d", port);
| ^~~~~~~~~~~~
drivers/infiniband/hw/mlx4/mad.c:2197:9: note: ‘snprintf’ output between
10 and 20 bytes into a destination of size 12
2197 | snprintf(name, sizeof(name), "mlx4_ibt%d", port);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/infiniband/hw/mlx4/mad.c:2205:48: error: ‘%d’ directive output
may be truncated writing between 1 and 11 bytes into a region of size 3
[-Werror=format-truncation=]
2205 | snprintf(name, sizeof(name), "mlx4_ibwi%d", port);
| ^~
drivers/infiniband/hw/mlx4/mad.c:2205:38: note: directive argument in
the range [-2147483645, 2147483647]
2205 | snprintf(name, sizeof(name), "mlx4_ibwi%d", port);
| ^~~~~~~~~~~~~
drivers/infiniband/hw/mlx4/mad.c:2205:9: note: ‘snprintf’ output between
11 and 21 bytes into a destination of size 12
2205 | snprintf(name, sizeof(name), "mlx4_ibwi%d", port);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/infiniband/hw/mlx4/mad.c:2213:48: error: ‘%d’ directive output
may be truncated writing between 1 and 11 bytes into a region of size 3
[-Werror=format-truncation=]
2213 | snprintf(name, sizeof(name), "mlx4_ibud%d", port);
| ^~
drivers/infiniband/hw/mlx4/mad.c:2213:38: note: directive argument in
the range [-2147483645, 2147483647]
2213 | snprintf(name, sizeof(name), "mlx4_ibud%d", port);
| ^~~~~~~~~~~~~
drivers/infiniband/hw/mlx4/mad.c:2213:9: note: ‘snprintf’ output between
11 and 21 bytes into a destination of size 12
2213 | snprintf(name, sizeof(name), "mlx4_ibud%d", port);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
make[6]: *** [scripts/Makefile.build:244: drivers/infiniband/hw/mlx4/mad.o] Error 1
Fixes: fc06573dfa ("IB/mlx4: Initialize SR-IOV IB support for slaves in master context")
Link: https://lore.kernel.org/r/f3798b3ce9a410257d7e1ec7c9e285f1352e256a.1718554569.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The req_transport_retries_exceeded counter shows the number of times
requester detected transport retries exceed error.
The req_rnr_retries_exceeded counter show the number of times the
requester detected RNR NAKs retries exceed error.
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Previously, all IB dev resources are initialized on driver load. As
they are not always used, move the initialization to the time when
they are needed.
To be more specific, move PD (p0) and CQ (c0) initialization to the
time when the first SRQ is created. and move SRQs（s0 and s1)
initialization to the time first QP is created. To avoid concurrent
creations, two new mutexes are also added.
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Link: https://lore.kernel.org/r/98c3e53a8cc0bdfeb6dec6e5bb8b037d78ab00d8.1717409369.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
UMR QP is not used in some cases, so move QP and its CQ creations from
driver load flow to the time first reg_mr occurs, that is when MR
interfaces are first called.
The initialization of dev->umrc.pd and dev->umrc.lock is still done in
driver load because pd is needed for mlx5_mkey_cache_init and the lock
is reused to protect against the concurrent creation.
When testing 4G bytes memory registration latency with rtool [1] and 8
threads in parallel, there is minor performance degradation (<5% for
the max latency) is seen for the first reg_mr with this change.
Link: https://github.com/paravmellanox/rtool [1]
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Link: https://lore.kernel.org/r/55d3c4f8a542fd974d8a4c5816eccfb318a59b38.1717409369.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
From: Leon Romanovsky <leonro@nvidia.com>
Internal mlx5_ib resources are created during mlx5_ib module load. This
behavior is not optimal because it consumes resources that are not
needed when SFs are created. This patch series delays the creation of
mlx5_ib internal resources to the stage when they actually used.
Signed-off-by: Leon Romanovsky <leon@kernel.org>
The test of write combining was added before in mlx5_ib driver. It
opens UD QP and posts NOP WQEs, and uses BlueFlame doorbell. When
BlueFlame is used, WQEs get written directly to a PCI BAR of the
device (in addition to memory) so that the device handles them without
having to access memory.
In this test, the WQEs written in memory are different from the ones
written to the BlueFlame which request CQE update. By checking the
completion reports posted on CQ, we can know if BlueFlame succeeds or
not. The write combining must be supported if BlueFlame succeeds as
its register is written using write combining.
This patch reimplements the test in the same way, but using a pair of
SQ and CQ only. It is moved to mlx5_core as a general feature used by
both mlx5_core and mlx5_ib.
Besides, save write combine test result of the PCI function, so that
its thousands of child functions such as SF can query without paying
the time and resource penalty by itself. The test function is called
only after failing to get the cached result. With this enhancement,
all thousands of SFs of the PF attached to same driver no longer need
to perform WC check explicitly, which is already done in the system.
This saves several commands per SF, thereby speeds up SF creation and
also saves completion EQ creation.
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/4ff5a8cc4c5b5b0d98397baa45a5019bcdbf096e.1717409369.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
