The roce path triggers a work queue that continues to touch the id_priv
but doesn't hold any reference on it. Futher, unlike in the IB case, the
work queue is not fenced during rdma_destroy_id().

This can trigger a use after free if a destroy is triggered in the
incredibly narrow window after the queue_work and the work starting and
obtaining the handler_mutex.

The only purpose of this work queue is to run the ULP event callback from
the standard context, so switch the design to use the existing
cma_work_handler() scheme. This simplifies quite a lot of the flow:

- Use the cma_work_handler() callback to launch the work for roce. This
  requires generating the event synchronously inside the
  rdma_join_multicast(), which in turn means the dummy struct
  ib_sa_multicast can become a simple stack variable.

- cm_work_handler() used the id_priv kref, so we can entirely eliminate
  the kref inside struct cma_multicast. Since the cma_multicast never
  leaks into an unprotected work queue the kfree can be done at the same
  time as for IB.

- Eliminating the general multicast.ib requires using cma_set_mgid() in a
  few places to recompute the mgid.

Fixes: 3c86aa70 ("RDMA/cm: Add RDMA CM support for IBoE devices")
Link: https://lore.kernel.org/r/20200902081122.745412-9-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Two places were open coding this sequence, and also pull in
cma_leave_roce_mc_group() which was called only once.

Link: https://lore.kernel.org/r/20200902081122.745412-8-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

There is no kernel user of RDMA CM multicast so this code managing the
multicast subscription of the kernel-only internal QP is dead. Remove it.

This makes the bug fixes in the next patches much simpler.

Link: https://lore.kernel.org/r/20200902081122.745412-7-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

These are the same thing, except that cma_ndev_work doesn't have a state
transition. Signal no state transition by setting old_state and new_state
== 0.

In all cases the handler function should not be called once
rdma_destroy_id() has progressed passed setting the state.

Link: https://lore.kernel.org/r/20200902081122.745412-6-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

The only place that still uses it is rdma_join_multicast() which is only
doing a sanity check that the caller hasn't done something wrong and
doesn't need the spinlock.

At least in the case of rdma_join_multicast() the information it needs
will remain until the ID is destroyed once it enters these
states. Similarly there is no reason to check for these specific states in
the handler callback, instead use the usual check for a destroyed id under
the handler_mutex.

Link: https://lore.kernel.org/r/20200902081122.745412-5-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

There is a strange unlocked read of the ID state when checking for
reuseaddr. This is because an ID cannot be reusable once it becomes a
listening ID. Instead of using the state to exclude reuse, just clear it
as part of rdma_listen()'s flow to convert reusable into not reusable.

Once a ID goes to listen there is no way back out, and the only use of
reusable is on the bind_list check.

Finally, update the checks under handler_mutex to use READ_ONCE and audit
that once RDMA_CM_LISTEN is observed in a req callback it is stable under
the handler_mutex.

Link: https://lore.kernel.org/r/20200902081122.745412-4-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Re-organize things so the state variable is not read unlocked. The first
attempt to go directly from ADDR_BOUND immediately tells us if the ID is
already bound, if we can't do that then the attempt inside
rdma_bind_addr() to go from IDLE to ADDR_BOUND confirms the ID needs
binding.

Link: https://lore.kernel.org/r/20200902081122.745412-3-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

It is currently a bit confusing, but the design is if the handler_mutex
is held, and the state is in RDMA_CM_CONNECT, then the state cannot leave
RDMA_CM_CONNECT without also serializing with the handler_mutex.

Make this clearer by adding a direct assertion, fixing the usage in
rdma_connect and generally using READ_ONCE to read the state value.

Link: https://lore.kernel.org/r/20200902081122.745412-2-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Use DEFINE_SEQ_ATTRIBUTE macro to simplify the code.

Link: https://lore.kernel.org/r/20200916025022.3992627-1-liushixin2@huawei.comSigned-off-by: Liu Shixin <liushixin2@huawei.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

hw_context stores a pci_device pointer. Use the correct type instead of
void * and avoid typecasts.

Link: https://lore.kernel.org/r/20200914123528.270382-1-parav@mellanox.comSigned-off-by: Parav Pandit <parav@nvidia.com>
Acked-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Implement the XRC specific verbs.  The additional QP type introduced new
logic to the rest of the verbs that now require distinguishing whether a
QP has an "RQ" or an "SQ" or both.

Link: https://lore.kernel.org/r/20200722102339.30104-1-ybason@marvell.comSigned-off-by: Michal Kalderon <mkalderon@marvell.com>
Signed-off-by: Yuval Basson <ybason@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Alignment of parameters was off by one

Link: https://lore.kernel.org/r/20200902165741.8355-9-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

commit 59e8970b ("RDMA/qedr: Return max inline data in QP query
result") changed query_qp max_inline size to return the max roce inline
size.  When iwarp was introduced, this should have been modified to return
the max inline size based on protocol.  This size is cached in the device
attributes

Fixes: 69ad0e7f ("RDMA/qedr: Add support for iWARP in user space")
Link: https://lore.kernel.org/r/20200902165741.8355-8-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Currently iWARP does not support mtu-change. Notify user when MTU changes
that reload of qedr is required for mtu to change.  Display the correct
active mtu.

Fixes: f5b1b177 ("RDMA/qedr: Add iWARP support in existing verbs")
Link: https://lore.kernel.org/r/20200902165741.8355-7-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

MTU change on ethtool is currently not supported for iWARP.  Notify qedr
driver so that appropriate logging can take place.

Link: https://lore.kernel.org/r/20200902165741.8355-6-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

In iWARP, accept could be called after a QP is already destroyed.  In this
case an error should be returned and not success.

Fixes: 82af6d19 ("RDMA/qedr: Fix synchronization methods and memory leaks in qedr")
Link: https://lore.kernel.org/r/20200902165741.8355-5-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

dev->attr.page_size_caps was used uninitialized when setting device
attributes

Fixes: ec72fce4 ("qedr: Add support for RoCE HW init")
Link: https://lore.kernel.org/r/20200902165741.8355-4-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Change the doorbell setting so that the maximum value between the last and
current value is set. This is to avoid doorbells being lost.

Fixes: a7efd777 ("qedr: Add support for PD,PKEY and CQ verbs")
Link: https://lore.kernel.org/r/20200902165741.8355-3-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

The qedr_qp structure wasn't freed when the protocol was RoCE.  kmemleak
output when running basic RoCE scenario.

unreferenced object 0xffff927ad7e22c00 (size 1024):
  comm "ib_send_bw", pid 7082, jiffies 4384133693 (age 274.698s)
  hex dump (first 32 bytes):
    00 b0 cd a2 79 92 ff ff 00 3f a1 a2 79 92 ff ff  ....y....?..y...
    00 ee 5c dd 80 92 ff ff 00 f6 5c dd 80 92 ff ff  ..\.......\.....
  backtrace:
    [<00000000b2ba0f35>] qedr_create_qp+0xb3/0x6c0 [qedr]
    [<00000000e85a43dd>] ib_uverbs_handler_UVERBS_METHOD_QP_CREATE+0x555/0xad0 [ib_uverbs]
    [<00000000fee4d029>] ib_uverbs_cmd_verbs+0xa5a/0xb80 [ib_uverbs]
    [<000000005d622660>] ib_uverbs_ioctl+0xa4/0x110 [ib_uverbs]
    [<00000000eb4cdc71>] ksys_ioctl+0x87/0xc0
    [<00000000abe6b23a>] __x64_sys_ioctl+0x16/0x20
    [<0000000046e7cef4>] do_syscall_64+0x4d/0x90
    [<00000000c6948f76>] entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fixes: 1212767e ("qedr: Add wrapping generic structure for qpidr and adjust idr routines.")
Link: https://lore.kernel.org/r/20200902165741.8355-2-michal.kalderon@marvell.comSigned-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Add work completion opcodes to a new ib_uverbs_wc_opcode enum in
ib_user_verbs.h. This plays the same role as ib_uverbs_wr_opcode
documenting the opcodes in the user space API.

Assigned the IB_WC_XXX opcodes in ib_verbs.h to the IB_UVERBS_WC_XXX
where they are defined. This follows the same pattern as the IB_WR_XXX
opcodes. This fixes an incorrect value for LSO that had crept in but
is not currently being used.

Added a missing IB_WR_BIND_MW opcode in ib_verbs.h.

Link: https://lore.kernel.org/r/20200903224039.437391-2-rpearson@hpe.comSigned-off-by: Bob Pearson <rpearson@hpe.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

This is always the same value as IOVA masked by the page size, just use
that clearer calculation directly.

It is unclear of ocrdma hardware can actually support a true fbo, if so it
could use a different algorithm to compute the best page size.

Link: https://lore.kernel.org/r/17-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

zbva is always false, so fbo is never read.

A 'zero-based-virtual-address' is simply IOVA == 0, and the driver already
supports this.

Link: https://lore.kernel.org/r/16-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.com
Acked-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

For the calls linked to mlx4_ib_umem_calc_optimal_mtt_size() use
ib_umem_num_dma_blocks() inside the function, it is just some weird static
default.

All other places are just using it with PAGE_SIZE, switch to
ib_umem_num_dma_blocks().

As this is the last call site, remove ib_umem_num_count().

Link: https://lore.kernel.org/r/15-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

This driver always uses PAGE_SIZE.

Link: https://lore.kernel.org/r/14-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

This driver always uses a DMA array made up of PAGE_SIZE elements, so just
use ib_umem_num_dma_blocks().

Since rdma_for_each_dma_block() always iterates exactly
ib_umem_num_dma_blocks() there is no need for the early exit check in
build_user_pbes(), delete it.

Link: https://lore.kernel.org/r/13-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

mtr_umem_page_count() does the same thing, replace it with the core code.

Also, ib_umem_find_best_pgsz() should always be called to check that the
umem meets the page_size requirement. If there is a limited set of
page_sizes that work it the pgsz_bitmap should be set to that set. 0 is a
failure and the umem cannot be used.

Lightly tidy the control flow to implement this flow properly.

Link: https://lore.kernel.org/r/12-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comAcked-by: Weihang Li <liweihang@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

ib_umem_page_count() returns the number of 4k entries required for a DMA
map, but bnxt_re already computes a variable page size. The correct API to
determine the size of the page table array is ib_umem_num_dma_blocks().

Fix the overallocation of the page array in fill_umem_pbl_tbl() when
working with larger page sizes by using the right function. Lightly
re-organize this function to make it clearer.

Replace the other calls to ib_umem_num_pages().

Fixes: d8558251 ("RDMA/bnxt_re: Use core helpers to get aligned DMA address")
Link: https://lore.kernel.org/r/11-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comAcked-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

The length of the list populated by qedr_populate_pbls() should be
calculated using ib_umem_num_dma_blocks() with the same size/shift passed
to qedr_populate_pbls().

Link: https://lore.kernel.org/r/10-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comAcked-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

This loop is splitting the DMA SGL into pg_shift sized pages, use the core
code for this directly.

Link: https://lore.kernel.org/r/9-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comAcked-by: Michal Kalderon <michal.kalderon@marvell.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

If ib_umem_find_best_pgsz() returns > PAGE_SIZE then the equation here is
not correct. 'start' should be 'virt'. Change it to use the core code for
page_num and the canonical calculation of page_shift.

Fixes: eb52c033 ("RDMA/i40iw: Use core helpers to get aligned DMA address within a supported page size")
Link: https://lore.kernel.org/r/8-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

If ib_umem_find_best_pgsz() returns > PAGE_SIZE then the equation here is
not correct. 'start' should be 'virt'. Change it to use the core code for
page_num and the canonical calculation of page_shift.

Fixes: 40ddb3f0 ("RDMA/efa: Use API to get contiguous memory blocks aligned to device supported page size")
Link: https://lore.kernel.org/r/7-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comTested-by: Gal Pressman <galpress@amazon.com>
Acked-by: Gal Pressman <galpress@amazon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

ib_umem_num_pages() should only be used by things working with the SGL in
CPU pages directly.

Drivers building DMA lists should use the new ib_num_dma_blocks() which
returns the number of blocks rdma_umem_for_each_block() will return.

To make this general for DMA drivers requires a different implementation.
Computing DMA block count based on umem->address only works if the
requested page size is < PAGE_SIZE and/or the IOVA == umem->address.

Instead the number of DMA pages should be computed in the IOVA address
space, not umem->address. Thus the IOVA has to be stored inside the umem
so it can be used for these calculations.

For now set it to umem->address by default and fix it up if
ib_umem_find_best_pgsz() was called. This allows drivers to be converted
to ib_umem_num_dma_blocks() safely.

Link: https://lore.kernel.org/r/6-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

Generally drivers should be using this core helper to split up the umem
into DMA pages.

These drivers are all probably wrong in some way to pass PAGE_SIZE in as
the HW page size. Either the driver doesn't support other page sizes and
it should use 4096, or the driver does support other page sizes and should
use ib_umem_find_best_pgsz() to select the best HW pages size of the HW
supported set.

The only case it could be correct is if the HW has a global setting for
PAGE_SIZE set at driver initialization time.

Link: https://lore.kernel.org/r/5-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

This helper does the same as rdma_for_each_block(), except it works on a
umem. This simplifies most of the call sites.

Link: https://lore.kernel.org/r/4-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comAcked-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
Acked-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

The calculation in rdma_find_pg_bit() is fairly complicated, and the
function is never called anywhere else. Inline a simpler version into
ib_umem_find_best_pgsz()

Link: https://lore.kernel.org/r/3-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comSigned-off-by: Jason Gunthorpe <jgg@nvidia.com>

rdma_for_each_block() makes assumptions about how the SGL is constructed
that don't work if the block size is below the page size used to to build
the SGL.

The rules for umem SGL construction require that the SG's all be PAGE_SIZE
aligned and we don't encode the actual byte offset of the VA range inside
the SGL using offset and length. So rdma_for_each_block() has no idea
where the actual starting/ending point is to compute the first/last block
boundary if the starting address should be within a SGL.

Fixing the SGL construction turns out to be really hard, and will be the
subject of other patches. For now block smaller pages.

Fixes: 4a353399 ("RDMA/umem: Add API to find best driver supported page size in an MR")
Link: https://lore.kernel.org/r/2-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comReviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

It is possible for a single SGL to span an aligned boundary, eg if the SGL
is

  61440 -> 90112

Then the length is 28672, which currently limits the block size to
32k. With a 32k page size the two covering blocks will be:

  32768->65536 and 65536->98304

However, the correct answer is a 128K block size which will span the whole
28672 bytes in a single block.

Instead of limiting based on length figure out which high IOVA bits don't
change between the start and end addresses. That is the highest useful
page size.

Fixes: 4a353399 ("RDMA/umem: Add API to find best driver supported page size in an MR")
Link: https://lore.kernel.org/r/1-v2-270386b7e60b+28f4-umem_1_jgg@nvidia.comReviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Change counters to return failure like any other verbs destroy, however
this flow shouldn't return error at all.

Link: https://lore.kernel.org/r/20200907120921.476363-10-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Make this interface symmetrical to other destroy paths.

Fixes: a49b1dc7 ("RDMA: Convert destroy_wq to be void")
Link: https://lore.kernel.org/r/20200907120921.476363-9-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Update XRCD destroy flow to allow command failure.

Fixes: 28ad5f65 ("RDMA: Move XRCD to be under ib_core responsibility")
Link: https://lore.kernel.org/r/20200907120921.476363-8-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>