[ Upstream commit a53e215f90 ]
The cited patch tries to ensure no pending works on the mkey cache
workqueue by disabling adding new works and call flush_workqueue().
But this workqueue also has delayed works which might still be pending
the delay time to be queued.
Add cancel_delayed_work() for the delayed works which waits to be queued
and then the flush_workqueue() will flush all works which are already
queued and running.
Fixes: 374012b004 ("RDMA/mlx5: Fix mkey cache possible deadlock on cleanup")
Link: https://lore.kernel.org/r/b8722f14e7ed81452f791764a26d2ed4cfa11478.1698256179.git.leon@kernel.org
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
After the change to use dynamic cache structure, new cache entries
can be added and the mkey allocation can no longer assume that all
mkeys created for the cache have access_flags equal to zero.
Example of a flow that exposes the issue:
A user registers MR with RO on a HCA that cannot UMR RO and the mkey is
created outside of the cache. When the user deregisters the MR, a new
cache entry is created to store mkeys with RO.
Later, the user registers 2 MRs with RO. The first MR is reused from the
new cache entry. When we try to get the second mkey from the cache we see
the entry is empty so we go to the MR cache mkey allocation flow which
would have allocated a mkey with no access flags, resulting the user getting
a MR without RO.
Fixes: dd1b913fb0 ("RDMA/mlx5: Cache all user cacheable mkeys on dereg MR flow")
Reviewed-by: Edward Srouji <edwards@nvidia.com>
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Link: https://lore.kernel.org/r/8a802700b82def3ace3f77cd7a9ad9d734af87e7.1695203958.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
The MR memory allocation requests extra bytes to guarantee that there
is enough space to find the memory aligned to MLX5_UMR_ALIGN.
For power-of-two sizes, the alignment can be guaranteed by kmalloc()
according to commit 59bb47985c ("mm, sl[aou]b: guarantee natural
alignment for kmalloc(power-of-two)").
So if target alignment is power-of-two and adding the extra bytes
crosses a power-of-two boundary, use the next power-of-two as the
allocation size.
Signed-off-by: Yuanyuan Zhong <yzhong@purestorage.com>
Link: https://lore.kernel.org/r/20230629213248.3184245-2-yzhong@purestorage.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
According to PCIe spec, Enable Relaxed Ordering value in the VF's PCI
config space is wired to 0 and PF relaxed ordering (RO) setting should
be applied to the VF. In QEMU (and maybe others), when assigning VFs,
the RO bit in PCI config space is not emulated properly and is always
set to 0.
Therefore, pcie_relaxed_ordering_enabled() always returns 0 for VFs and
VMs and thus MKeys can't be created with RO read even if the PF supports
it.
pcie_relaxed_ordering_enabled() check was added to avoid a syndrome when
creating a MKey with relaxed ordering (RO) enabled when the driver's
relaxed_ordering_read_pci_enabled HCA capability is out of sync with FW.
With the new relaxed_ordering_read capability this can't happen, as it's
set regardless of RO value in PCI config space and thus can't change
during runtime.
Hence, to allow RO read in VFs and VMs, use the new HCA capability
relaxed_ordering_read without checking pcie_relaxed_ordering_enabled().
The old capability checks are kept for backward compatibility with older
FWs.
Allowing RO in VFs and VMs is valuable since it can greatly improve
performance on some setups. For example, testing throughput of a VF on
an AMD EPYC 7763 and ConnectX-6 Dx setup showed roughly 60% performance
improvement.
Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Link: https://lore.kernel.org/r/e7048640d66c341a8fa0465e099926e7989184bc.1681131553.git.leon@kernel.org
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Rename existing HCA capability relaxed_ordering_read to
relaxed_ordering_read_pci_enabled. This is in accordance with recent PRM
change to better describe the capability, as it's set only if both the
device supports relaxed ordering (RO) read and RO is enabled in PCI
config space.
In addition, add new HCA capability relaxed_ordering_read which is set
if the device supports RO read, regardless of RO in PCI config space.
This will be used in the following patch to allow RO in VFs and VMs.
Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Link: https://lore.kernel.org/r/caa0002fd8135086357dfcc368e2f5cc73b08480.1681131553.git.leon@kernel.org
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
pcie_relaxed_ordering_enabled() check was added to avoid a syndrome when
creating a MKey with relaxed ordering (RO) enabled when the driver's
relaxed_ordering_{read,write} HCA capabilities are out of sync with FW.
While this can happen with relaxed_ordering_read, it can't happen with
relaxed_ordering_write as it's set if the device supports RO write,
regardless of RO in PCI config space, and thus can't change during
runtime.
Therefore, drop the pcie_relaxed_ordering_enabled() check for
relaxed_ordering_write while keeping it for relaxed_ordering_read.
Doing so will also allow the usage of RO write in VFs and VMs (where RO
in PCI config space is not reported/emulated properly).
Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Link: https://lore.kernel.org/r/7e8f55e31572c1702d69cae015a395d3a824a38a.1681131553.git.leon@kernel.org
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Block MR cache debugfs creation for IB representor flow as MR cache shouldn't be used
at all in that mode. As part of this change, add missing debugfs cleanup in error path
too.
This change fixes the following debugfs errors:
bond0: (slave enp8s0f1): Enslaving as a backup interface with an up link
mlx5_core 0000:08:00.0: lag map: port 1:1 port 2:1
mlx5_core 0000:08:00.0: shared_fdb:1 mode:queue_affinity
mlx5_core 0000:08:00.0: Operation mode is single FDB
debugfs: Directory '2' with parent '/' already present!
...
debugfs: Directory '22' with parent '/' already present!
Fixes: 73d09b2fe8 ("RDMA/mlx5: Introduce mlx5r_cache_rb_key")
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Link: https://lore.kernel.org/r/482a78c54acbcfa1742a0e06a452546428900ffa.1675328463.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
The non-cache mkeys are stored in the cache only to shorten restarting
application time. Don't store them longer than needed.
Configure cache entries that store non-cache MRs as temporary entries. If
30 seconds have passed and no user reclaimed the temporarily cached mkeys,
an asynchronous work will destroy the mkeys entries.
Link: https://lore.kernel.org/r/20230125222807.6921-7-michaelgur@nvidia.com
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Currently, when dereging an MR, if the mkey doesn't belong to a cache
entry, it will be destroyed. As a result, the restart of applications
with many non-cached mkeys is not efficient since all the mkeys are
destroyed and then recreated. This process takes a long time (for 100,000
MRs, it is ~20 seconds for dereg and ~28 seconds for re-reg).
To shorten the restart runtime, insert all cacheable mkeys to the cache.
If there is no fitting entry to the mkey properties, create a temporary
entry that fits it.
After a predetermined timeout, the cache entries will shrink to the
initial high limit.
The mkeys will still be in the cache when consuming them again after an
application restart. Therefore, the registration will be much faster
(for 100,000 MRs, it is ~4 seconds for dereg and ~5 seconds for re-reg).
The temporary cache entries created to store the non-cache mkeys are not
exposed through sysfs like the default cache entries.
Link: https://lore.kernel.org/r/20230125222807.6921-6-michaelgur@nvidia.com
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Switch from using the mkey order to using the new struct as the key to the
RB tree of cache entries.
The key is all the mkey properties that UMR operations can't modify.
Using this key to define the cache entries and to search and create cache
mkeys.
Link: https://lore.kernel.org/r/20230125222807.6921-5-michaelgur@nvidia.com
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Currently, the cache structure is a static linear array. Therefore, his
size is limited to the number of entries in it and is not expandable. The
entries are dedicated to mkeys of size 2^x and no access_flags. Mkeys with
different properties are not cacheable.
In this patch, we change the cache structure to an RB-tree. This will
allow to extend the cache to support more entries with different mkey
properties.
Link: https://lore.kernel.org/r/20230125222807.6921-4-michaelgur@nvidia.com
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
mkc.log_page_size can be changed using UMR. Therefore, don't treat it as a
cache entry property.
Removing it from struct mlx5_cache_ent.
All cache mkeys will be created with default PAGE_SHIFT, and updated with
the needed page_shift using UMR when passing them to a user.
Link: https://lore.kernel.org/r/20230125222807.6921-2-michaelgur@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
For mlx5 if ATS is enabled in the PCI config then the device will use ATS
requests for only certain DMA operations. This has to be opted in by the
SW side based on the mkey or umem settings.
ATS slows down the PCI performance, so it should only be set in cases when
it is needed. All of these cases revolve around optimizing PCI P2P
transfers and avoiding bad cases where the bus just doesn't work.
Link: https://lore.kernel.org/r/4-v1-bd147097458e+ede-umem_dmabuf_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The Xarray allows us to store the cached mkeys in memory efficient way.
Entries are reserved in the Xarray using xa_cmpxchg before calling to the
upcoming callbacks to avoid allocations in interrupt context. The
xa_cmpxchg can sleep when using GFP_KERNEL, so we call it in a loop to
ensure one reserved entry for each process trying to reserve.
Link: https://lore.kernel.org/r/20220726071911.122765-3-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>
Add support for allocate/deallocate and registering MR of the new type
of ICM area. Support exists only for devices that support sw_owner_v2.
Signed-off-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Pull rdma updates from Jason Gunthorpe:
- Minor bug fixes in mlx5, mthca, pvrdma, rtrs, mlx4, hfi1, hns
- Minor cleanups: coding style, useless includes and documentation
- Reorganize how multicast processing works in rxe
- Replace a red/black tree with xarray in rxe which improves performance
- DSCP support and HW address handle re-use in irdma
- Simplify the mailbox command handling in hns
- Simplify iser now that FMR is eliminated
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (93 commits)
RDMA/nldev: Prevent underflow in nldev_stat_set_counter_dynamic_doit()
IB/iser: Fix error flow in case of registration failure
IB/iser: Generalize map/unmap dma tasks
IB/iser: Use iser_fr_desc as registration context
IB/iser: Remove iser_reg_data_sg helper function
RDMA/rxe: Use standard names for ref counting
RDMA/rxe: Replace red-black trees by xarrays
RDMA/rxe: Shorten pool names in rxe_pool.c
RDMA/rxe: Move max_elem into rxe_type_info
RDMA/rxe: Replace obj by elem in declaration
RDMA/rxe: Delete _locked() APIs for pool objects
RDMA/rxe: Reverse the sense of RXE_POOL_NO_ALLOC
RDMA/rxe: Replace mr by rkey in responder resources
RDMA/rxe: Fix ref error in rxe_av.c
RDMA/hns: Use the reserved loopback QPs to free MR before destroying MPT
RDMA/irdma: Add support for address handle re-use
RDMA/qib: Fix typos in comments
RDMA/mlx5: Fix memory leak in error flow for subscribe event routine
Revert "RDMA/core: Fix ib_qp_usecnt_dec() called when error"
RDMA/rxe: Remove useless argument for update_state()
...
Move the debugfs entry pointers under priv to their own struct.
Add get function for device debugfs root.
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Same as the new mlx5_cmd_do API, report all information to callers and
let them handle the error values and outbox parsing.
The user callback status "work->user_callback(status)" is now similar to
the error rc code returned from the blocking mlx5_cmd_do() version,
and now is defined as follows:
-EREMOTEIO : Command executed by FW, outbox.status != MLX5_CMD_STAT_OK.
Caller must check FW outbox status.
0 : Command execution successful, outbox.status == MLX5_CMD_STAT_OK.
< 0 : Command couldn't execute, FW or driver induced error.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The mkc is the key for the mkey cache, hence, created in each attempt to
get a cache mkey, while pcie_relaxed_ordering_enabled() is called during
the setting of the mkc, but used only for cases where
IB_ACCESS_RELAXED_ORDERING is set.
pcie_relaxed_ordering_enabled() is an expensive call (26 us). Reorder the
code so the driver will call it only when it is needed.
Link: https://lore.kernel.org/r/684be1366cb1d4f05aa3e78986205e4bc410443a.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>
For the case of IB_MR_TYPE_DM the mr does doesn't have a umem, even though
it is a user MR. This causes function mlx5_free_priv_descs() to think that
it is a kernel MR, leading to wrongly accessing mr->descs that will get
wrong values in the union which leads to attempt to release resources that
were not allocated in the first place.
For example:
DMA-API: mlx5_core 0000:08:00.1: device driver tries to free DMA memory it has not allocated [device address=0x0000000000000000] [size=0 bytes]
WARNING: CPU: 8 PID: 1021 at kernel/dma/debug.c:961 check_unmap+0x54f/0x8b0
RIP: 0010:check_unmap+0x54f/0x8b0
Call Trace:
debug_dma_unmap_page+0x57/0x60
mlx5_free_priv_descs+0x57/0x70 [mlx5_ib]
mlx5_ib_dereg_mr+0x1fb/0x3d0 [mlx5_ib]
ib_dereg_mr_user+0x60/0x140 [ib_core]
uverbs_destroy_uobject+0x59/0x210 [ib_uverbs]
uobj_destroy+0x3f/0x80 [ib_uverbs]
ib_uverbs_cmd_verbs+0x435/0xd10 [ib_uverbs]
? uverbs_finalize_object+0x50/0x50 [ib_uverbs]
? lock_acquire+0xc4/0x2e0
? lock_acquired+0x12/0x380
? lock_acquire+0xc4/0x2e0
? lock_acquire+0xc4/0x2e0
? ib_uverbs_ioctl+0x7c/0x140 [ib_uverbs]
? lock_release+0x28a/0x400
ib_uverbs_ioctl+0xc0/0x140 [ib_uverbs]
? ib_uverbs_ioctl+0x7c/0x140 [ib_uverbs]
__x64_sys_ioctl+0x7f/0xb0
do_syscall_64+0x38/0x90
Fix it by reorganizing the dereg flow and mlx5_ib_mr structure:
- Move the ib_umem field into the user MRs structure in the union as it's
applicable only there.
- Function mlx5_ib_dereg_mr() will now call mlx5_free_priv_descs() only
in case there isn't udata, which indicates that this isn't a user MR.
Fixes: f18ec42231 ("RDMA/mlx5: Use a union inside mlx5_ib_mr")
Link: https://lore.kernel.org/r/66bb1dd253c1fd7ceaa9fc411061eefa457b86fb.1637581144.git.leonro@nvidia.com
Signed-off-by: Alaa Hleihel <alaa@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@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>
