RDMA/usnic: fix set-but-not-unused variable 'flags' warning

Remove unused local variable 'flag'
without any logic changes.

Fixes: e3cf00d0a87f ("IB/usnic: Add Cisco VIC low-level hardware driver")
Signed-off-by: Zeng Heng <zengheng4@huawei.com>
Link: https://lore.kernel.org/r/20220929031200.4060891-1-zengheng4@huawei.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/hfi1: Use skb_put_data() instead of skb_put/memcpy pair

Use skb_put_data() instead of skb_put() and memcpy(), which is shorter
and clear. Drop the tmp variable that is not needed any more.

Link: https://lore.kernel.org/r/20220927022919.16902-1-shangxiaojing@huawei.com
Signed-off-by: Shang XiaoJing <shangxiaojing@huawei.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Unified Log Printing Style

The first letter of the log information is changed to lowercase
to keep the same style.

Link: https://lore.kernel.org/r/20220922123315.3732205-13-xuhaoyue1@hisilicon.com
Signed-off-by: Guofeng Yue <yueguofeng@hisilicon.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Replacing magic number with macros in apply_func_caps()

Replacing magic number with macros in function apply_func_caps().

Link: https://lore.kernel.org/r/20220922123315.3732205-12-xuhaoyue1@hisilicon.com
Signed-off-by: Yixing Liu <liuyixing1@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Repacing 'dseg_len' by macros in fill_ext_sge_inl_data()

The sge size is known to be constant, so it's unnecessary to use sizeof to
calculate.

Link: https://lore.kernel.org/r/20220922123315.3732205-11-xuhaoyue1@hisilicon.com
Signed-off-by: Luoyouming <luoyouming@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove redundant 'max_srq_desc_sz' in caps

The max_srq_desc_sz is defined in the code, but never used,
so delete this redundant variable.

Link: https://lore.kernel.org/r/20220922123315.3732205-10-xuhaoyue1@hisilicon.com
Signed-off-by: Yangyang Li <liyangyang20@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove redundant 'num_mtt_segs' and 'max_extend_sg'

The num_mtt_segs and max_extend_sg used to be used for HIP06,
remove them since the HIP06 code has been removed.

Link: https://lore.kernel.org/r/20220922123315.3732205-9-xuhaoyue1@hisilicon.com
Signed-off-by: Yangyang Li <liyangyang20@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove redundant 'phy_addr' in hns_roce_hem_list_find_mtt()

This parameter has never been used. Remove it to simplify the function.

Link: https://lore.kernel.org/r/20220922123315.3732205-8-xuhaoyue1@hisilicon.com
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove redundant 'use_lowmem' argument from hns_roce_init_hem_table()

As hns_roce_init_hem_table() is always called with use_lowmem
being '1', and table->lowmem is set according to that argument,
so remove table->lowmem too.

Also, as the table->lowmem is used to indicate a dma buffer
is allocated with GFP_HIGHUSER or GFP_KERNEL, and calling
dma_alloc_coherent() with GFP_KERNEL seems like a common
pattern.

Link: https://lore.kernel.org/r/20220922123315.3732205-7-xuhaoyue1@hisilicon.com
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove redundant 'bt_level' for hem_list_alloc_item()

The 'bt_level' parameter is not used in hem_list_alloc_item(),
so remove it.

Link: https://lore.kernel.org/r/20220922123315.3732205-6-xuhaoyue1@hisilicon.com
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove redundant 'attr_mask' in modify_qp_init_to_init()

The attr_mask variable is not used in the function,
so remove it.

Link: https://lore.kernel.org/r/20220922123315.3732205-5-xuhaoyue1@hisilicon.com
Signed-off-by: Yixing Liu <liuyixing1@huawei.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove unnecessary brackets when getting point

Delete () when using & to obtain an address.

Link: https://lore.kernel.org/r/20220922123315.3732205-4-xuhaoyue1@hisilicon.com
Signed-off-by: Guofeng Yue <yueguofeng@hisilicon.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove unnecessary braces for single statement blocks

Braces {} are not necessary for single statement blocks.

Link: https://lore.kernel.org/r/20220922123315.3732205-3-xuhaoyue1@hisilicon.com
Signed-off-by: Guofeng Yue <yueguofeng@hisilicon.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Cleanup for a spelling error of Asynchronous

Fixed a spelling error for Asynchronous.

Link: https://lore.kernel.org/r/20220922123315.3732205-2-xuhaoyue1@hisilicon.com
Signed-off-by: Guofeng Yue <yueguofeng@hisilicon.com>
Signed-off-by: Haoyue Xu <xuhaoyue1@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

IB/rdmavt: Add __init/__exit annotations to module init/exit funcs

Add missing __init/__exit annotations to module init/exit funcs.

Fixes: 0194621b2253 ("IB/rdmavt: Create module framework and handle driver registration")
Link: https://lore.kernel.org/r/20220924091457.52446-1-xiujianfeng@huawei.com
Signed-off-by: Xiu Jianfeng <xiujianfeng@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/rxe: Remove redundant num_sge fields

In include/uapi/rdma/rdma_user_rxe.h there are redundant copies of num_sge
in the rxe_send_wr, rxe_recv_wqe, and rxe_dma_info. Only the ones in
rxe_dma_info are actually used by the rxe kernel driver.

The userspace would set these values, but the kernel never read them.

This change has no affect on the current ABI and new or old versions of
rdma-core operate correctly with new or old versions of the kernel rxe
driver.

Link: https://lore.kernel.org/r/20220913222716.18335-1-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/mlx5: Enable ATS support for MRs and umems

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>

RDMA/mlx5: Add support for dmabuf to devx umem

This is modeled after the similar EFA enablement in commit
66f4817b5712 ("RDMA/efa: Add support for dmabuf memory regions").

Like EFA there is no support for revocation so we simply call the
ib_umem_dmabuf_get_pinned() to obtain a umem instead of the normal
ib_umem_get().  Everything else stays the same.

Link: https://lore.kernel.org/r/3-v1-bd147097458e+ede-umem_dmabuf_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/core: Add UVERBS_ATTR_RAW_FD

This uses the same passing protocol as UVERBS_ATTR_FD (eg len = 0 data_s64
= fd), except that the FD is not required to be a uverbs object and the
core code does not covert the FD to an object handle automatically.

Access to the int fd is provided by uverbs_get_raw_fd().

Link: https://lore.kernel.org/r/2-v1-bd147097458e+ede-umem_dmabuf_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

net/mlx5: Add IFC bits for mkey ATS

Allows telling a mkey to use PCI ATS for DMA that flows through it.

Link: https://lore.kernel.org/r/1-v1-bd147097458e+ede-umem_dmabuf_jgg@nvidia.com
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/rxe: Fix resize_finish() in rxe_queue.c

Currently in resize_finish() in rxe_queue.c there is a loop which copies
the entries in the original queue into a newly allocated queue.  The
termination logic for this loop is incorrect. The call to
queue_next_index() updates cons but has no effect on whether the queue is
empty. So if the queue starts out empty nothing is copied but if it is not
then the loop will run forever. This patch changes the loop to compare the
value of cons to the original producer index.

Fixes: ae6e843fe08d0 ("RDMA/rxe: Add memory barriers to kernel queues")
Link: https://lore.kernel.org/r/20220825221446.6512-1-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Reviewed-by: Li Zhijian <lizhijian@fujitsu.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/rxe: Set pd early in mr alloc routines

Move setting of pd in mr objects ahead of any possible errors so that it
will always be set in rxe_mr_cleanup() to avoid seg faults when
rxe_put(mr_pd(mr)) is called.

Fixes: cf40367961d8 ("RDMA/rxe: Move mr cleanup code to rxe_mr_cleanup()")
Link: https://lore.kernel.org/r/20220805183153.32007-2-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Reviewed-by: Li Zhijian <lizhijian@fujitsu.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/rxe: Add send_common_ack() helper

Most code in send_ack() and send_atomic_ack() are duplicate, move them to
a new helper send_common_ack().

In newer IBA spec, some opcodes require acknowledge with a zero-length
read response, with this new helper, we can easily implement it later.

Link: https://lore.kernel.org/r/1659335010-2-1-git-send-email-lizhijian@fujitsu.com
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/core: Clean up a variable name in ib_create_srq_user()

"&srq->pd->usecnt" and "&pd->usecnt" are different names for the same
reference count.  Use "&pd->usecnt" consistently for both the increment
and decrement.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/YyxFe3Pm0uzRuBkQ@kili
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/srp: Support more than 255 rdma ports

Currently ib_srp module does not support devices with more than 256
ports. Switch from u8 to u32 to fix the problem.

Fixes: 1fb7f8973f51 ("RDMA: Support more than 255 rdma ports")
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Mikhael Goikhman <migo@nvidia.com>
Link: https://lore.kernel.org/r/7d80d8844f1abb3a54170b7259f0a02be38080a6.1663747327.git.leonro@nvidia.com
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Use members of generic struct in rxe_mr

rxe_mr and ib_mr have interchangeable members. Remove device specific
members and use ones in the generic struct. Both 'iova' and 'length' are
filled in ib_uverbs or ib_core layer after MR registration.

Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://lore.kernel.org/r/20220921080844.1616883-2-matsuda-daisuke@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB: Set IOVA/LENGTH on IB_MR in core/uverbs layers

Set 'iova' and 'length' on ib_mr in ib_uverbs and ib_core layers to let all
drivers have the members filled. Also, this commit removes redundancy in
the respective drivers.

Previously, commit 04c0a5fcfcf65 ("IB/uverbs: Set IOVA on IB MR in uverbs
layer") changed to set 'iova', but seems to have missed 'length' and the
ib_core layer at that time.

Fixes: 04c0a5fcfcf65 ("IB/uverbs: Set IOVA on IB MR in uverbs layer")
Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://lore.kernel.org/r/20220921080844.1616883-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/cm: Use DLID from inbound/outbound PathRecords as the datapath DLID

In inter-subnet cases, when inbound/outbound PRs are available,
outbound_PR.dlid is used as the requestor's datapath DLID and
inbound_PR.dlid is used as the responder's DLID. The inbound_PR.dlid
is passed to responder side with the "ConnectReq.Primary_Local_Port_LID"
field. With this solution the PERMISSIVE_LID is no longer used in
Primary Local LID field.

Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Link: https://lore.kernel.org/r/b3f6cac685bce9dde37c610be82e2c19d9e51d9e.1662631201.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/cm: Use SLID in the work completion as the DLID in responder side

The responder should always use WC's SLID as the dlid, to follow the
IB SPEC section "13.5.4.2 COMMON RESPONSE ACTIONS":
A responder always takes the following actions in constructing a
response packet:
- The SLID of the received packet is used as the DLID in the response
  packet.

Fixes: ac3a949fb2ff ("IB/CM: Set appropriate slid and dlid when handling CM request")
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Link: https://lore.kernel.org/r/cd17c240231e059d2fc07c17dfe555d548b917eb.1662631201.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/cma: Multiple path records support with netlink channel

Support receiving inbound and outbound IB path records (along with GMP
PathRecord) from user-space service through the RDMA netlink channel.
The LIDs in these 3 PRs can be used in this way:
1. GMP PR: used as the standard local/remote LIDs;
2. DLID of outbound PR: Used as the "dlid" field for outbound traffic;
3. DLID of inbound PR: Used as the "dlid" field for outbound traffic in
   responder side.

This is aimed to support adaptive routing. With current IB routing
solution when a packet goes out it's assigned with a fixed DLID per
target, meaning a fixed router will be used.
The LIDs in inbound/outbound path records can be used to identify group
of routers that allow communication with another subnet's entity. With
them packets from an inter-subnet connection may travel through any
router in the set to reach the target.

As confirmed with Jason, when sending a netlink request, kernel uses
LS_RESOLVE_PATH_USE_ALL so that the service knows kernel supports
multiple PRs.

Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Link: https://lore.kernel.org/r/2fa2b6c93c4c16c8915bac3cfc4f27be1d60519d.1662631201.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/core: Rename rdma_route.num_paths field to num_pri_alt_paths

This fields means the total number of primary and alternative paths,
i.e.,:
  0 - No primary nor alternate path is available;
  1 - Only primary path is available;
  2 - Both primary and alternate path are available.
Rename it to avoid confusion as with follow patches primary path will
support multiple path records.

Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Link: https://lore.kernel.org/r/cbe424de63a56207870d70c5edce7c68e45f429e.1662631201.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/erdma: Support dynamic mtu

Hardware now support jumbo frame for RDMA. So we introduce a new CMDQ
message to support mtu change notification.

Signed-off-by: Cheng Xu <chengyou@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220909093822.33868-5-chengyou@linux.alibaba.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/siw: Fix QP destroy to wait for all references dropped.

Delay QP destroy completion until all siw references to QP are
dropped. The calling RDMA core will free QP structure after
successful return from siw_qp_destroy() call, so siw must not
hold any remaining reference to the QP upon return.
A use-after-free was encountered in xfstest generic/460, while
testing NFSoRDMA. Here, after a TCP connection drop by peer,
the triggered siw_cm_work_handler got delayed until after
QP destroy call, referencing a QP which has already freed.

Fixes: 303ae1cdfdf7 ("rdma/siw: application interface")
Reported-by: Olga Kornievskaia <kolga@netapp.com>
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Link: https://lore.kernel.org/r/20220920082503.224189-1-bmt@zurich.ibm.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/siw: Always consume all skbuf data in sk_data_ready() upcall.

For header and trailer/padding processing, siw did not consume new
skb data until minimum amount present to fill current header or trailer
structure, including potential payload padding. Not consuming any
data during upcall may cause a receive stall, since tcp_read_sock()
is not upcalling again if no new data arrive.
A NFSoRDMA client got stuck at RDMA Write reception of unaligned
payload, if the current skb did contain only the expected 3 padding
bytes, but not the 4 bytes CRC trailer. Expecting 4 more bytes already
arrived in another skb, and not consuming those 3 bytes in the current
upcall left the Write incomplete, waiting for the CRC forever.

Fixes: 8b6a361b8c48 ("rdma/siw: receive path")
Reported-by: Olga Kornievskaia <kolga@netapp.com>
Tested-by: Olga Kornievskaia <kolga@netapp.com>
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Link: https://lore.kernel.org/r/20220920081202.223629-1-bmt@zurich.ibm.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/hfi1: remove rc_only_opcode and uc_only_opcode declarations

rc_only_opcode and uc_only_opcode have been removed since
commit b374e060cc2a ("IB/hfi1: Consolidate pio control masks
into single definition"), so remove them.

Signed-off-by: Gaosheng Cui <cuigaosheng1@huawei.com>
Link: https://lore.kernel.org/r/20220911092325.3216513-1-cuigaosheng1@huawei.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/srpt: Use flex array destination for memcpy()

In preparation for FORTIFY_SOURCE performing run-time destination buffer
bounds checking for memcpy(), specify the destination output buffer
explicitly, instead of asking memcpy() to write past the end of what looked
like a fixed-size object.

Notice that srp_rsp[] is a pointer to a structure that contains
flexible-array member data[]:

struct srp_rsp {
...
__be32 sense_data_len;
__be32 resp_data_len;
u8 data[];
};

link: https://github.com/KSPP/linux/issues/201
Signed-off-by: Hangyu Hua <hbh25y@gmail.com>
Link: https://lore.kernel.org/r/20220909022943.8896-1-hbh25y@gmail.com
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/erdma: Make hardware internal opcodes invisible to driver

Some opcodes are used in hardware internally, and driver does not care
about them. So, we change them to reserved opcodes in driver.

Signed-off-by: Cheng Xu <chengyou@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220909093822.33868-4-chengyou@linux.alibaba.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/erdma: Remove redundant includes

Many of erdma's includes are redundant, because they are already included
indirectly by kernel headers or custom headers. So we remove all the
unnecessary direct-includes. Besides, add linux/pci.h to erdma.h because
it's also used in the file.

Signed-off-by: Cheng Xu <chengyou@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220909093822.33868-3-chengyou@linux.alibaba.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/erdma: Eliminate unnecessary casting for erdma_post_cmd_wait

erdma_post_cmd_wait does not use the 'u64 *req' input parameter directly.
So it is better to define it to 'void *req', and by this we can eliminate
the casting when calling erdma_post_cmd_wait.

Signed-off-by: Cheng Xu <chengyou@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220909093822.33868-2-chengyou@linux.alibaba.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/srp: Fix srp_abort()

Fix the code for converting a SCSI command pointer into an SRP request
pointer.

Cc: Xiao Yang <yangx.jy@fujitsu.com>
Fixes: ad215aaea4f9 ("RDMA/srp: Make struct scsi_cmnd and struct srp_request adjacent")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220908233139.3042628-1-bvanassche@acm.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/qib: fix repeated words in comments

Delete the redundant word 'to'.

Signed-off-by: wangjianli <wangjianli@cdjrlc.com>
Link: https://lore.kernel.org/r/20220908132036.42355-1-wangjianli@cdjrlc.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hfi1: fix repeated words in comments

Delete the redundant word 'to'.

Signed-off-by: wangjianli <wangjianli@cdjrlc.com>
Link: https://lore.kernel.org/r/20220908131824.41106-1-wangjianli@cdjrlc.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/irdma: Validate udata inlen and outlen

Currently ib_copy_from_udata and ib_copy_to_udata could underfill
the request and response buffer if the user-space passes an undersized
value for udata->inlen or udata->outlen respectively [1]
This could lead to undesirable behavior.

Zero initing the buffer only goes as far as preventing using the buffer
uninitialized.

Validate udata->inlen and udata->outlen passed from user-space to ensure
they are at least the required minimum size.

[1] https://lore.kernel.org/linux-rdma/MWHPR11MB0029F37D40D9D4A993F8F549E9D79@MWHPR11MB0029.namprd11.prod.outlook.com/

Fixes: b48c24c2d710 ("RDMA/irdma: Implement device supported verb APIs")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Link: https://lore.kernel.org/r/20220907191324.1173-3-shiraz.saleem@intel.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/irdma: Align AE id codes to correct flush code and event

A number of asynchronous event (AE) ids were not aligned to the
correct flush_code and event_type. Fix these up so that the
correct IBV error and event codes are returned to application.

Also, add handling for new AE ids like IRDMA_AE_INVALID_REQUEST to
return the correct WC error code.

Fixes: 44d9e52977a1 ("RDMA/irdma: Implement device initialization definitions")
Signed-off-by: Sindhu-Devale <sindhu.devale@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Link: https://lore.kernel.org/r/20220907191324.1173-2-shiraz.saleem@intel.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mlx5: Remove duplicate assignment in umr_rereg_pas()

The same value is assigned to 'mr->ibmr.length'. Remove redundant one.

Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://lore.kernel.org/r/20220908083058.3993700-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: convert pr_warn to pr_debug

They could be triggered by user APIs with invalid parameters.

Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Link: https://lore.kernel.org/r/1662518901-2-2-git-send-email-lizhijian@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: use %u to print u32 variables

struct ib_qp_cap {
        u32     max_send_wr;
        u32     max_recv_wr;
        u32     max_send_sge;
        u32     max_recv_sge;
        u32     max_inline_data;
...

To avoid getting a negative value from dmesg:
[410580.579965] rdma_rxe: invalid send sge = 65535 > 32
[410580.583818] rdma_rxe: invalid send wr = -1 > 1048576
[410582.771323] rdma_rxe: invalid recv sge = 65535 > 32
[410582.775310] rdma_rxe: invalid recv wr = -1 > 1048576

Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Link: https://lore.kernel.org/r/1662518901-2-1-git-send-email-lizhijian@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rtrs-clt: Kill xchg_paths

Let's call try_cmpxchg directly for the same purpose.

Signed-off-by: Guoqing Jiang <guoqing.jiang@linux.dev>
Link: https://lore.kernel.org/r/20220903040252.29397-1-guoqing.jiang@linux.dev
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rtrs-clt: Break the loop once one path is connected

No need to iterate all paths after find one connected path.

Signed-off-by: Guoqing Jiang <guoqing.jiang@linux.dev>
Link: https://lore.kernel.org/r/20220902101922.26273-3-guoqing.jiang@linux.dev
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rtrs: Update comments for MAX_SESS_QUEUE_DEPTH

The maximum queue_depth should be 65535 per check_module_params,
also update other relevant comments.

Signed-off-by: Guoqing Jiang <guoqing.jiang@linux.dev>
Link: https://lore.kernel.org/r/20220902101922.26273-2-guoqing.jiang@linux.dev
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hfi1: Remove the unneeded result variable

Return the value set_link_state() directly instead of storing it in
another redundant variable.

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: ye xingchen <ye.xingchen@zte.com.cn>
Link: https://lore.kernel.org/r/20220901074209.313004-1-ye.xingchen@zte.com.cn
Signed-off-by: Leon Romanovsky <leon@kernel.org>

Merge branch 'mlx5-next' into wip/leon-for-next

Perform merge of Mellanox shared branch.

* mlx5-next:
  RDMA/mlx5: Move function mlx5_core_query_ib_ppcnt() to mlx5_ib

RDMA/mlx5: Move function mlx5_core_query_ib_ppcnt() to mlx5_ib

This patch doesn't change any functionality, but move one function
to mlx5_ib because it is not used by mlx5_core.

The actual fix is in the next patch.

Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Chris Mi <cmi@nvidia.com>
Link: https://lore.kernel.org/r/fd47b9138412bd94ed30f838026cbb4cf3878150.1661763871.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/mlx5: Support querying eswitch functions from DEVX

Query eswitch functions returns information of the external host
PF(if it exists). It can be used to check if DEVX is running on ECPF.

Reviewed-by: Erez Shitrit <erezsh@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Bodong Wang <bodong@mellanox.com>
Link: https://lore.kernel.org/r/4265925178ab3224dc1d3e3784bb312d808edca5.1661763785.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/siw: Add missing Kconfig selections

The SoftiWARP Kconfig is missing "select" for CRYPTO and CRYPTO_CRC32C.

In addition, it improperly "depends on" LIBCRC32C, this should be a
"select", similar to net/sctp and others. As a dependency, SIW fails
to appear in generic configurations.

Link: https://lore.kernel.org/r/d366bf02-3271-754f-fc68-1a84016d0e19@talpey.com
Signed-off-by: Tom Talpey <tom@talpey.com>
Acked-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Delete error messages triggered by incoming Read requests

An incoming Read request causes multiple Read responses. If a user MR to
copy data from is unavailable or responder cannot send a reply, then the
error messages can be printed for each response attempt, resulting in
message overflow.

Link: https://lore.kernel.org/r/20220829071218.1639065-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Remove the unused variable obj

The member variable obj in struct rxe_task is not needed.
So remove it to save memory.

Link: https://lore.kernel.org/r/20220822011615.805603-4-yanjun.zhu@linux.dev
Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Reviewed-by: Li Zhijian <lizhijian@fujitsu.com>
Reviewed-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Fix the error caused by qp->sk

When sock_create_kern in the function rxe_qp_init_req fails,
qp->sk is set to NULL.

Then the function rxe_create_qp will call rxe_qp_do_cleanup
to handle allocated resource.

Before handling qp->sk, this variable should be checked.

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Link: https://lore.kernel.org/r/20220822011615.805603-3-yanjun.zhu@linux.dev
Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Reviewed-by: Li Zhijian <lizhijian@fujitsu.com>
Reviewed-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Fix "kernel NULL pointer dereference" error

When rxe_queue_init in the function rxe_qp_init_req fails,
both qp->req.task.func and qp->req.task.arg are not initialized.

Because of creation of qp fails, the function rxe_create_qp will
call rxe_qp_do_cleanup to handle allocated resource.

Before calling __rxe_do_task, both qp->req.task.func and
qp->req.task.arg should be checked.

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Link: https://lore.kernel.org/r/20220822011615.805603-2-yanjun.zhu@linux.dev
Reported-by: syzbot+ab99dc4c6e961eed8b8e@syzkaller.appspotmail.com
Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Reviewed-by: Li Zhijian <lizhijian@fujitsu.com>
Reviewed-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Remove redundant member doorbell_qpn of struct hns_roce_qp

The value of doorbell_qpn is always equal to qpn on current hardware
versions. So remove it.

Link: https://lore.kernel.org/r/20220829105021.1427804-5-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/cm: Refactor cm_insert_listen() and cm_find_listen()

Move the device and service_id match code at the top of
cm_insert_listen() and cm_find_listen() into the final else branch.

Link: https://lore.kernel.org/r/20220819090859.957943-4-markzhang@nvidia.com
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/cm: remove cm_id_priv->id.service_mask and service_mask parameter of cm_init_listen()

The service_mask is always ~cpu_to_be64(0), so the result is always
a NOP when it is &'d with a service_id. Remove it for simplicity.

Link: https://lore.kernel.org/r/20220819090859.957943-3-markzhang@nvidia.com
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/cm: Remove the service_mask parameter from ib_cm_listen()

Remove the service_mask parameter of ib_cm_listen(), as all callers
use 0.

Link: https://lore.kernel.org/r/20220819090859.957943-2-markzhang@nvidia.com
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rtrs: Remove 'dir' argument from rnbd_srv_rdma_ev

Since process_{read,write} already prints direction info if ctx->ops.rdma_ev
fails, no need to pass 'dir'.

Link: https://lore.kernel.org/r/20220826081117.21687-1-guoqing.jiang@linux.dev
Signed-off-by: Guoqing Jiang <guoqing.jiang@linux.dev>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Remove an unused member from struct rxe_mr

Commit 1e75550648da ("Revert "RDMA/rxe: Create duplicate mapping tables for
FMRs"") brought back the member 'va' to struct rxe_mr. However, it is
actually used by nobody and thus can be removed.

Fixes: 1e75550648da ("Revert "RDMA/rxe: Create duplicate mapping tables for FMRs"")
Link: https://lore.kernel.org/r/20220829012335.1212697-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/core: Remove 'device' argument from rdma_build_skb()

'device' argument is never used since rdma_build_skb()
is introduced, so remove it.

Link: https://lore.kernel.org/r/20220826143215.18111-1-linyunsheng@huawei.com
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/srp: Use the attribute group mechanism for sysfs attributes

Simplify the SRP driver by using the attribute group mechanism instead
of calling device_create_file() explicitly.

Link: https://lore.kernel.org/r/20220825213900.864587-5-bvanassche@acm.org
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/srp: Handle dev_set_name() failure

Instead of ignoring dev_set_name() failure, handle dev_set_name()
failure. Convert a device_register() call into device_initialize() and
device_add() calls.

Link: https://lore.kernel.org/r/20220825213900.864587-4-bvanassche@acm.org
Reported-by: Bo Liu <liubo03@inspur.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/srp: Remove the srp_host.released completion

Move the kfree(host) calls into srp_release_dev(). Convert a
device_unregister() call into a device_del() and a device_put() call.
Remove the host->released completion object. This patch prepares for
handling dev_set_name() failure in srp_add_port().

Link: https://lore.kernel.org/r/20220825213900.864587-3-bvanassche@acm.org
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/srp: Rework the srp_add_port() error path

device_register() always calls device_initialize() so calling device_del()
is safe even if device_register() fails. Implement the following advice
from the comment block above device_register(): "NOTE: _Never_ directly free
@dev after calling this function, even if it returned an error! Always use
put_device() to give up the reference initialized in this function instead."
Keep the kfree() call in the error path since srp_release_dev() does not
free the host.

Link: https://lore.kernel.org/r/20220825213900.864587-2-bvanassche@acm.org
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Support MR's restrack raw ops for hns driver

The MR raw restrack attributes come from the queue context maintained by
the ROCEE.

For example:

$ rdma res show mr dev hns_0 mrn 6 -dd -jp -r
[ {
        "ifindex": 4,
        "ifname": "hns_0",
        "data": [ 1,0,0,0,2,0,0,0,0,3,0,0,0,0,2,0,0,0,0,0,32,0,0,0,2,0,0,0,
  2,0,0,0,0,0,0,0 ]
    } ]

Link: https://lore.kernel.org/r/20220822104455.2311053-8-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Support MR's restrack ops for hns driver

The MR restrack attributes come from the queue information maintained by
the driver.

For example:

$ rdma res show mr dev hns_0 mrn 6 -dd -jp
[ {
        "ifindex": 4,
        "ifname": "hns_0",
        "mrn": 6,
        "rkey": "300",
        "lkey": "300",
        "mrlen": 131072,
        "pdn": 8,
        "pid": 1524,
        "comm": "ib_send_bw"
    },
    "drv_pbl_hop_num": 2,
    "drv_ba_pg_shift": 14,
    "drv_buf_pg_shift": 12
}

Link: https://lore.kernel.org/r/20220822104455.2311053-7-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Support QP's restrack raw ops for hns driver

The QP raw restrack attributes come from the queue context maintained by
the ROCEE.

For example:

$ rdma res show qp link hns_0 -jp -dd -r
[ {
        "ifindex": 4,
        "ifname": "hns_0",
        "data": [ 2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,
  5,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,255,156,0,0,63,156,0,0,
  7,0,0,0,1,0,0,0,9,0,0,0,0,0,0,0,2,0,0,0,2,0,0,0,0,0,0,
  0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
  0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,63,156,0,
  0,0,0,0,0 ]
    } ]

Link: https://lore.kernel.org/r/20220822104455.2311053-6-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Support QP's restrack ops for hns driver

The QP restrack attributes come from the queue information maintained by
the driver.

For example:

$ rdma res show qp link hns_0 lqpn 41 -jp -dd
[ {
        "ifindex": 4,
        "ifname": "hns_0",
        "port": 1,
        "lqpn": 41,
        "rqpn": 40,
        "type": "RC",
        "state": "RTR",
        "rq-psn": 12474738,
        "sq-psn": 0,
        "path-mig-state": "ARMED",
        "pdn": 9,
        "pid": 1523,
        "comm": "ib_send_bw"
    },
    "drv_sq_wqe_cnt": 128,
    "drv_sq_max_gs": 1,
    "drv_rq_wqe_cnt": 512,
    "drv_rq_max_gs": 2,
    "drv_ext_sge_sge_cnt": 0
}

Link: https://lore.kernel.org/r/20220822104455.2311053-5-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Support CQ's restrack raw ops for hns driver

The CQ raw restrack attributes come from the queue context maintained by
the ROCEE.

For example:

$ rdma res show cq dev hns_0 cqn 14 -dd -jp -r
[ {
        "ifindex": 4,
        "ifname": "hns_0",
        "data": [ 1,0,0,0,7,0,0,0,0,0,0,0,0,82,6,0,0,82,6,0,0,82,6,0,
  1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,
  6,0,0,0,0,0,0,0 ]
    } ]

Link: https://lore.kernel.org/r/20220822104455.2311053-4-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Add or remove CQ's restrack attributes

Remove the resttrack attributes from the queue context held by ROCEE, and
add the resttrack attributes from the queue information maintained by the
driver.

For example:

$ rdma res show cq dev hns_0 cqn 14 -dd -jp
[ {
        "ifindex": 4,
        "ifname": "hns_0",
        "cqn": 14,
        "cqe": 127,
        "users": 1,
        "adaptive-moderation": false,
        "ctxn": 8,
        "pid": 1524,
        "comm": "ib_send_bw"
    },
    "drv_cq_depth": 128,
    "drv_cons_index": 0,
    "drv_cqe_size": 32,
    "drv_arm_sn": 1
}

Link: https://lore.kernel.org/r/20220822104455.2311053-3-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Remove redundant DFX file and DFX ops structure

There is no need to use a dedicated DXF file and DFX structure to manage
the interface of the query queue context.

Link: https://lore.kernel.org/r/20220822104455.2311053-2-liangwenpeng@huawei.com
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/mlx5: Remove duplicate header inclusion related to ODP

rdma/ib_umem.h and rdma/ib_verbs.h are included by rdma/ib_umem_odp.h.
This patch removes the redundant entries.

Link: https://lore.kernel.org/r/20220823025131.862811-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB: move from strlcpy with unused retval to strscpy

Follow the advice of the below link and prefer 'strscpy' in this
subsystem. Conversion is 1:1 because the return value is not used.
Generated by a coccinelle script.

Link: https://lore.kernel.org/r/CAHk-=wgfRnXz0W3D37d01q3JFkr_i_uTL=V6A6G1oUZcprmknw@mail.gmail.com/
Link: https://lore.kernel.org/r/20220818210018.6841-1-wsa+renesas@sang-engineering.com
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/efa: Support CQ receive entries with source GID

Add a parameter for create CQ admin command to set source address on
receive completion descriptors. Report capability for this feature
through query device verb.

Link: https://lore.kernel.org/r/20220818140449.414-1-mrgolin@amazon.com
Reviewed-by: Firas Jahjah <firasj@amazon.com>
Reviewed-by: Yossi Leybovich <sleybo@amazon.com>
Signed-off-by: Daniel Kranzdorf <dkkranzd@amazon.com>
Signed-off-by: Michael Margolin <mrgolin@amazon.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rtrs-clt: Output sg index when warning on

Output the sg index, so it's a bit easier for debug.

Signed-off-by: Jack Wang <jinpu.wang@ionos.com>
Reviewed-by: Aleksei Marov <aleksei.marov@ionos.com>
Link: https://lore.kernel.org/r/20220818105355.110344-2-haris.iqbal@ionos.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rtrs-srv: Add event tracing support

Add event tracing mechanism for following routines:
- send_io_resp_imm()

How to use:
1. Load the rtrs_server module
2. cd /sys/kernel/debug/tracing
3. If all the events need to be enabled:
        echo 1 > events/rtrs_srv/enable
4. OR only speific routine/event needs to be enabled e.g.
        echo 1 > events/rtrs_srv/send_io_resp_imm/enable
5. cat trace
6. Run some I/O workload which can trigger send_io_resp_imm()

Link: https://lore.kernel.org/r/20220818105240.110234-3-haris.iqbal@ionos.com
Signed-off-by: Santosh Pradhan <santosh.pradhan@ionos.com>
Signed-off-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Md Haris Iqbal <haris.iqbal@ionos.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rtrs-clt: Add event tracing support

Add event tracing mechanism for following routines:
- rtrs_clt_reconnect_work()
- rtrs_clt_close_conns()
- rtrs_rdma_error_recovery()

How to use:
1. Load the rtrs_client module
2. cd /sys/kernel/debug/tracing
3. If all the events need to be enabled:
        echo 1 > events/rtrs_clt/enable
4. OR only speific routine/event needs to be enabled e.g.
        echo 1 > events/rtrs_clt/rtrs_clt_close_conns/enable
5. cat trace
6. Run some workload which can trigger rtrs_clt_close_conns()

Link: https://lore.kernel.org/r/20220818105240.110234-2-haris.iqbal@ionos.com
Signed-off-by: Santosh Pradhan <santosh.pradhan@ionos.com>
Signed-off-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Md Haris Iqbal <haris.iqbal@ionos.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mlx5: Don't compare mkey tags in DEVX indirect mkey

According to the ib spec:
If the CI supports the Base Memory Management Extensions defined in this
specification, the L_Key format must consist of:
24 bit index in the most significant bits of the R_Key, and
8 bit key in the least significant bits of the R_Key
Through a successful Allocate L_Key verb invocation, the CI must let the
consumer own the key portion of the returned R_Key

Therefore, when creating a mkey using DEVX, the consumer is allowed to
change the key part. The kernel should compare only the index part of a
R_Key to determine equality with another R_Key.

Adding capability in order not to break backward compatibility.

Fixes: 534fd7aac56a ("IB/mlx5: Manage indirection mkey upon DEVX flow for ODP")
Link: https://lore.kernel.org/r/3d669aacea85a3a15c3b3b953b3eaba3f80ef9be.1659255945.git.leonro@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/mlx5: Call io_stop_wc() after writing to WC MMIO

This new function is defined only on ARM and serves to guarantee a barrier
in the WC operation. The barrier means that another run of this loop will
not combine with the stores this loop created.

On x86 this is happening implicitly because of the spin_unlock().

Link: https://lore.kernel.org/r/0-v1-c5dade92f363+11-mlx5_io_stop_wc_jgg@nvidia.com
Suggested-by: Pavel Shamis <Pavel.Shamis@arm.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

Linux 6.0-rc1

radix-tree: replace gfp.h inclusion with gfp_types.h

Radix tree header includes gfp.h for __GFP_BITS_SHIFT only. Now we
have gfp_types.h for this.

Fixes powerpc allmodconfig build:

   In file included from include/linux/nodemask.h:97,
                    from include/linux/mmzone.h:17,
                    from include/linux/gfp.h:7,
                    from include/linux/radix-tree.h:12,
                    from include/linux/idr.h:15,
                    from include/linux/kernfs.h:12,
                    from include/linux/sysfs.h:16,
                    from include/linux/kobject.h:20,
                    from include/linux/pci.h:35,
                    from arch/powerpc/kernel/prom_init.c:24:
   include/linux/random.h: In function 'add_latent_entropy':
>> include/linux/random.h:25:46: error: 'latent_entropy' undeclared (first use in this function); did you mean 'add_latent_entropy'?
      25 |         add_device_randomness((const void *)&latent_entropy, sizeof(latent_entropy));
         |                                              ^~~~~~~~~~~~~~
         |                                              add_latent_entropy
   include/linux/random.h:25:46: note: each undeclared identifier is reported only once for each function it appears in

Reported-by: kernel test robot <lkp@intel.com>
CC: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge tag 'pull-fixes' of git://git./linux/kernel/git/viro/vfs

Pull vfs lseek fix from Al Viro:
 "Fix proc_reg_llseek() breakage. Always had been possible if somebody
  left NULL ->proc_lseek, became a practical issue now"

* tag 'pull-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  take care to handle NULL ->proc_lseek()

take care to handle NULL ->proc_lseek()

Easily done now, just by clearing FMODE_LSEEK in ->f_mode
during proc_reg_open() for such entries.

Fixes: 868941b14441 "fs: remove no_llseek"
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Merge tag 'for-linus-6.0-rc1b-tag' of git://git./linux/kernel/git/xen/tip

Pull more xen updates from Juergen Gross:

 - fix the handling of the "persistent grants" feature negotiation
   between Xen blkfront and Xen blkback drivers

 - a cleanup of xen.config and adding xen.config to Xen section in
   MAINTAINERS

 - support HVMOP_set_evtchn_upcall_vector, which is more compliant to
   "normal" interrupt handling than the global callback used up to now

 - further small cleanups

* tag 'for-linus-6.0-rc1b-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  MAINTAINERS: add xen config fragments to XEN HYPERVISOR sections
  xen: remove XEN_SCRUB_PAGES in xen.config
  xen/pciback: Fix comment typo
  xen/xenbus: fix return type in xenbus_file_read()
  xen-blkfront: Apply 'feature_persistent' parameter when connect
  xen-blkback: Apply 'feature_persistent' parameter when connect
  xen-blkback: fix persistent grants negotiation
  x86/xen: Add support for HVMOP_set_evtchn_upcall_vector

Merge tag 'perf-tools-fixes-for-v6.0-2022-08-13' of git://git./linux/kernel/git/acme/linux

Pull more perf tool updates from Arnaldo Carvalho de Melo:

 - 'perf c2c' now supports ARM64, adjust its output to cope with
   differences with what is in x86_64. Now go find false sharing on
   ARM64 (at least Neoverse) as well!

 - Refactor the JSON processing, making the output more compact and thus
   reducing the size of the resulting perf binary

 - Improvements for 'perf offcpu' profiling, including tracking child
   processes

 - Update Intel JSON metrics and events files for broadwellde,
   broadwellx, cascadelakex, haswellx, icelakex, ivytown, jaketown,
   knightslanding, sapphirerapids, skylakex and snowridgex

 - Add 'perf stat' JSON output and a 'perf test' entry for it

 - Ignore memfd and anonymous mmap events if jitdump present

 - Refactor 'perf test' shell tests allowing subdirs

 - Fix an error handling path in 'parse_perf_probe_command()'

 - Fixes for the guest Intel PT tracing patchkit in the 1st batch of
   this merge window

 - Print debuginfod queries if -v option is used, to explain delays in
   processing when debuginfo servers are enabled to fetch DSOs with
   richer symbol tables

 - Improve error message for 'perf record -p not_existing_pid'

 - Fix openssl and libbpf feature detection

 - Add PMU pai_crypto event description for IBM z16 on 'perf list'

 - Fix typos and duplicated words on comments in various places

* tag 'perf-tools-fixes-for-v6.0-2022-08-13' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (81 commits)
  perf test: Refactor shell tests allowing subdirs
  perf vendor events: Update events for snowridgex
  perf vendor events: Update events and metrics for skylakex
  perf vendor events: Update metrics for sapphirerapids
  perf vendor events: Update events for knightslanding
  perf vendor events: Update metrics for jaketown
  perf vendor events: Update metrics for ivytown
  perf vendor events: Update events and metrics for icelakex
  perf vendor events: Update events and metrics for haswellx
  perf vendor events: Update events and metrics for cascadelakex
  perf vendor events: Update events and metrics for broadwellx
  perf vendor events: Update metrics for broadwellde
  perf jevents: Fold strings optimization
  perf jevents: Compress the pmu_events_table
  perf metrics: Copy entire pmu_event in find metric
  perf pmu-events: Hide the pmu_events
  perf pmu-events: Don't assume pmu_event is an array
  perf pmu-events: Move test events/metrics to JSON
  perf test: Use full metric resolution
  perf pmu-events: Hide pmu_events_map
  ...

Merge tag 'powerpc-6.0-2' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Ensure we never emit lwarx with EH=1 on 32-bit, because some 32-bit
   CPUs trap on it rather than ignoring it as they should.

 - Fix ftrace when building with clang, which was broken by some
   refactoring.

 - A couple of other minor fixes.

Thanks to Christophe Leroy, Naveen N.  Rao, Nick Desaulniers, Ondrej
Mosnacek, Pali Rohár, Russell Currey, and Segher Boessenkool.

* tag 'powerpc-6.0-2' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/kexec: Fix build failure from uninitialised variable
  powerpc/ppc-opcode: Fix PPC_RAW_TW()
  powerpc64/ftrace: Fix ftrace for clang builds
  powerpc: Make eh value more explicit when using lwarx
  powerpc: Don't hide eh field of lwarx behind a macro
  powerpc: Fix eh field when calling lwarx on PPC32

Merge tag 'pull-work.misc' of git://git./linux/kernel/git/viro/vfs

Pull /proc/mounts fix from Al Viro:
 "Fix for /proc/mounts escaping - escape the '#' character too"

* tag 'pull-work.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  vfs: escape hash as well

Merge tag '5.20-rc-smb3-client-fixes-part2' of git://git.samba.org/sfrench/cifs-2.6

Pull more cifs updates from Steve French:

 - two fixes for stable, one for a lock length miscalculation, and
   another fixes a lease break timeout bug

 - improvement to handle leases, allows the close timeout to be
   configured more safely

 - five restructuring/cleanup patches

* tag '5.20-rc-smb3-client-fixes-part2' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: Do not access tcon->cfids->cfid directly from is_path_accessible
  cifs: Add constructor/destructors for tcon->cfid
  SMB3: fix lease break timeout when multiple deferred close handles for the same file.
  smb3: allow deferred close timeout to be configurable
  cifs: Do not use tcon->cfid directly, use the cfid we get from open_cached_dir
  cifs: Move cached-dir functions into a separate file
  cifs: Remove {cifs,nfs}_fscache_release_page()
  cifs: fix lock length calculation

afs: Enable multipage folio support

Enable multipage folio support for the afs filesystem.

Support has already been implemented in netfslib, fscache and cachefiles
and in most of afs, but I've waited for Matthew Wilcox's latest folio
changes.

Note that it does require a change to afs_write_begin() to return the
correct subpage.  This is a "temporary" change as we're working on
getting rid of the need for ->write_begin() and ->write_end()
completely, at least as far as network filesystems are concerned - but
it doesn't prevent afs from making use of the capability.

Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Tested-by: kafs-testing@auristor.com
Cc: Marc Dionne <marc.dionne@auristor.com>
Cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/lkml/2274528.1645833226@warthog.procyon.org.uk/
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge tag 'timers-urgent-2022-08-13' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Ingo Molnar:
 "Misc timer fixes:

   - fix a potential use-after-free bug in posix timers

   - correct a prototype

   - address a build warning"

* tag 'timers-urgent-2022-08-13' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  posix-cpu-timers: Cleanup CPU timers before freeing them during exec
  time: Correct the prototype of ns_to_kernel_old_timeval and ns_to_timespec64
  posix-timers: Make do_clock_gettime() static

Merge tag 'x86-urgent-2022-08-13' of git://git./linux/kernel/git/tip/tip

Pull x86 fix from Ingo Molnar:
 "Fix the 'IBPB mitigated RETBleed' mode of operation on AMD CPUs (not
  turned on by default), which also need STIBP enabled (if available) to
  be '100% safe' on even the shortest speculation windows"

* tag 'x86-urgent-2022-08-13' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/bugs: Enable STIBP for IBPB mitigated RETBleed

Merge tag 'i2c-for-5.20-part2' of git://git./linux/kernel/git/wsa/linux

Pull more i2c updates from Wolfram Sang:

 - two driver fixes for issues introduced this cycle

 - one trivial driver improvement regarding ACPI

 - more DTS conversion and additions

 - documentation updates

 - subsystem-wide move from strlcpy to strscpy

* tag 'i2c-for-5.20-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  docs: i2c: i2c-sysfs: fix hyperlinks
  docs: i2c: i2c-sysfs: improve wording
  docs: i2c: instantiating-devices: add syntax coloring to dts and C blocks
  docs: i2c: smbus-protocol: improve DataLow/DataHigh definition
  docs: i2c: i2c-protocol: remove unused legend items
  docs: i2c: i2c-protocol,smbus-protocol: remove nonsense words
  docs: i2c: i2c-protocol: update introductory paragraph
  i2c: move core from strlcpy to strscpy
  i2c: move drivers from strlcpy to strscpy
  i2c: kempld: Support ACPI I2C device declaration
  i2c: mediatek: add i2c compatible for MT8188
  dt-bindings: i2c: update bindings for mt8188 soc
  i2c: microchip-corei2c: fix erroneous late ack send
  dt-bindings: i2c: qcom,i2c-cci: convert to dtschema
  i2c: qcom-geni: Fix GPI DMA buffer sync-back

Merge tag 'ntb-5.20' of https://github.com/jonmason/ntb

Pull NTB updates from Jon Mason:
 "Non-Transparent Bridge updates.

  Fix of heap data and clang warnings, support for a new Intel NTB
  device, and NTB EndPoint Function (EPF) support and the various fixes
  for that"

* tag 'ntb-5.20' of https://github.com/jonmason/ntb:
  MAINTAINERS: add PCI Endpoint NTB drivers to NTB files
  NTB: EPF: Tidy up some bounds checks
  NTB: EPF: Fix error code in epf_ntb_bind()
  PCI: endpoint: pci-epf-vntb: reduce several globals to statics
  PCI: endpoint: pci-epf-vntb: fix error handle in epf_ntb_mw_bar_init()
  PCI: endpoint: Fix Kconfig dependency
  NTB: EPF: set pointer addr to null using NULL rather than 0
  Documentation: PCI: extend subheading underline for "lspci output" section
  Documentation: PCI: Use code-block block for scratchpad registers diagram
  Documentation: PCI: Add specification for the PCI vNTB function device
  PCI: endpoint: Support NTB transfer between RC and EP
  NTB: epf: Allow more flexibility in the memory BAR map method
  PCI: designware-ep: Allow pci_epc_set_bar() update inbound map address
  ntb: intel: add GNR support for Intel PCIe gen5 NTB
  NTB: ntb_tool: uninitialized heap data in tool_fn_write()
  ntb: idt: fix clang -Wformat warnings

Merge tag 'xfs-5.20-merge-8' of git://git./fs/xfs/xfs-linux

Pull more xfs updates from Darrick Wong:
 "There's not a lot this time around, just the usual bug fixes and
  corrections for missing error returns.

   - Return error codes from block device flushes to userspace

   - Fix a deadlock between reclaim and mount time quotacheck

   - Fix an unnecessary ENOSPC return when doing COW on a filesystem
     with severe free space fragmentation

   - Fix a miscalculation in the transaction reservation computations
     for file removal operations"

* tag 'xfs-5.20-merge-8' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: fix inode reservation space for removing transaction
  xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork
  xfs: fix intermittent hang during quotacheck
  xfs: check return codes when flushing block devices