- 24 Aug, 2017 5 commits
-
-
Leon Romanovsky authored
When rdma_cm is initializing a cma_device it checks if this device supports the preferred default GID type. This check was done in a wrong way and therefore sometimes rdma_cm is coming up with default GID type that is not supported by the device. Fix that by checking for supported GID type properly. Fixes: 3c7f67d1 ("IB/cma: Fix default RoCE type setting") Signed-off-by:
Moni Shoua <monis@mellanox.com> Signed-off-by:
Leon Romanovsky <leonro@mellanox.com> Reviewed-by:
Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by:
Doug Ledford <dledford@redhat.com>
-
Doug Ledford authored
Pick up -rc fixes. Signed-off-by: -
Majd Dibbiny authored
CM layer calls ib_modify_port() regardless of the link layer. For the Ethernet ports, qkey violation and Port capabilities are meaningless. Therefore, always return success for ib_modify_port calls on the Ethernet ports. Cc: Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by:
Majd Dibbiny <majd@mellanox.com> Reviewed-by:
Leon Romanovsky <leon@kernel.org> Signed-off-by:
-
Majd Dibbiny authored
In case we have SQ and RQ for Raw Packet QP, the SQ's event handler wasn't assigned. Fixing this by assigning event handler for each WQ after creation. [ 1877.145243] Call Trace: [ 1877.148644] <IRQ> [ 1877.150580] [<ffffffffa07987c5>] ? mlx5_rsc_event+0x105/0x210 [mlx5_core] [ 1877.159581] [<ffffffffa0795bd7>] ? mlx5_cq_event+0x57/0xd0 [mlx5_core] [ 1877.167137] [<ffffffffa079208e>] mlx5_eq_int+0x53e/0x6c0 [mlx5_core] [ 1877.174526] [<ffffffff8101a679>] ? sched_clock+0x9/0x10 [ 1877.180753] [<ffffffff810f717e>] handle_irq_event_percpu+0x3e/0x1e0 [ 1877.188014] [<ffffffff810f735d>] handle_irq_event+0x3d/0x60 [ 1877.194567] [<ffffffff810f9fe7>] handle_edge_irq+0x77/0x130 [ 1877.201129] [<ffffffff81014c3f>] handle_irq+0xbf/0x150 [ 1877.207244] [<ffffffff815ed78a>] ? atomic_notifier_call_chain+0x1a/0x20 [ 1877.214829] [<ffffffff815f434f>] do_IRQ+0x4f/0xf0 [ 1877.220498] [<ffffffff815e94ad>] common_interrupt+0x6d/0x6d [ 1877.227025] <EOI> [ 1877.228967] [<ffffffff814834e2>] ? cpuidle_enter_state+0x52/0xc0 [ 1877.236990] [<ffffffff81483615>] cpuidle_idle_call+0xc5/0x200 [ 1877.243676] [<ffffffff8101bc7e>] arch_cpu_idle+0xe/0x30 [ 1877.249831] [<ffffffff810b4725>] cpu_startup_entry+0xf5/0x290 [ 1877.256513] [<ffffffff815cfee1>] start_secondary+0x265/0x27b [ 1877.263111] Code: Bad RIP value. [ 1877.267296] RIP [< (null)>] (null) [ 1877.273264] RSP <ffff88046fd63df8> [ 1877.277531] CR2: 0000000000000000 Fixes: 19098df2 ("IB/mlx5: Refactor mlx5_ib_qp to accommodate other QP types") Signed-off-by:
Yishai Hadas <yishaih@mellanox.com> Signed-off-by:
-
Noa Osherovich authored
Commit 44c58487 ("IB/core: Define 'ib' and 'roce' rdma_ah_attr types") introduced the concept of type in ah_attr: * During ib_register_device, each port is checked for its type which is stored in ib_device's port_immutable array. * During uverbs' modify_qp, the type is inferred using the port number in ib_uverbs_qp_dest struct (address vector) by accessing the relevant port_immutable array and the type is passed on to providers. IB spec (version 1.3) enforces a valid port value only in Reset to Init. During Init to RTR, the address vector must be valid but port number is not mentioned as a field in the address vector, so its value is not validated, which leads to accesses to a non-allocated memory when inferring the port type. Save the real port number in ib_qp during modify to Init (when the comp_mask indicates that the port number is valid) and use this value to infer the port type. Avoid copying the address vector fields if the matching bit is not set in the attr_mask. Address vector can't be modified before the port, so no valid flow is affected. Fixes: 44c58487 ('IB/core: Define 'ib' and 'roce' rdma_ah_attr types') Signed-off-by:
Noa Osherovich <noaos@mellanox.com> Reviewed-by:
-
- 22 Aug, 2017 35 commits
-
-
Jason Gunthorpe authored
If a message comes in and we do not have the client in the table, then try to load the module supplying that client using MODULE_ALIAS to find it. This duplicates the scheme seen in other netlink muxes (eg nfnetlink). Signed-off-by:
Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Reviewed-by:
-
Jason Gunthorpe authored
Provide a module alias so that if userspace opens a netlink socket for RDMA the kernel support is loaded automatically. Signed-off-by:
-
Li Dongyang authored
We could use kvmalloc_array instead of the kmalloc and __vmalloc combination. After this we don't need to include linux/vmalloc.h Signed-off-by:
Li Dongyang <dongyang.li@anu.edu.au> Reviewed-by:
-
Li Dongyang authored
We observed multiple times on our Lustre OSS servers that when the system memory is fragmented, kmalloc() in create_kernel_qp() could fail order 4/5 allocations while we still have many free pages. Switch to kvmalloc_array() to allow the operation to contine. Signed-off-by:
-
Selvin Xavier authored
ib_get_eth_speed() return 0 on success. Fixing the condition checking and prevent reporting failure for query_port verb. Fixes: d4186194 ("Add generic function to extract IB speed from netdev") Signed-off-by:
Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
-
Yuval Shaia authored
The function pvrdma_idx_ring_is_valid_idx is not in used so let's remove it. Signed-off-by:
Adit Ranadive <aditr@vmware.com> Signed-off-by:
-
Shiraz Saleem authored
The current timeout logic for Control Queue-Pair (CQP) OPs does not take into account whether CQP makes progress but rather blindly waits for a large timeout value, 100000 jiffies for the completion event. Improve this by setting the timeout based on whether the CQP is making progress or not. If the CQP is hung, the timeout will happen sooner, in 5000 jiffies. Each time the CQP progress is detetcted, the timeout extends by 5000 jiffies. Signed-off-by:
Shiraz Saleem <shiraz.saleem@intel.com> Signed-off-by:
Christopher N Bednarz <christopher.n.bednarz@intel.com> Signed-off-by:
Henry Orosco <henry.orosco@intel.com> Signed-off-by:
-
Kaike Wan authored
Reviewed-by:
Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by:
Kaike Wan <kaike.wan@intel.com> Signed-off-by:
Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by:
-
Grzegorz Morys authored
Remove HFI1_VERBS_31BIT_PSN Kconfig option leaving only 31-bit PSNs available. The option was implemented in the early days of the driver and is no longer needed. Reviewed-by:
Grzegorz Morys <grzegorz.morys@intel.com> Signed-off-by:
-
Jakub Byczkowski authored
Do not track physical state separately from host_link_state. Deduce physical state from host_link_state when required. Change cache_physical_state to log_physical_state to make sure host_link_state reflects hardwares physical state properly. Reviewed-by:
Ira Weiny <ira.weiny@intel.com> Signed-off-by:
Jakub Byczkowski <jakub.byczkowski@intel.com> Signed-off-by:
-
Kamenee Arumugame authored
The macro size is valid. This change makes it less ambiguous. Bounds check trap type for better security. Reviewed-by:
Michael J. Ruhl <michael.j.ruhl@intel.com> Signed-off-by:
Kamenee Arumugam <kamenee.arumugam@intel.com> Signed-off-by:
-
Jakub Byczkowski authored
Currently fallback configuration is loaded once per driver instance. With multiple HFI devices in the same system the current code may not load the platform config data for the device. Change fallback platform config data loading to be per device. Reviewed-by:
Easwar Hariharan <easwar.hariharan@intel.com> Reviewed-by:
-
Jakub Byczkowski authored
Add flag in pport data structure to determine when platform config was read from scratch registers. Change conditions in parse_platform_config and get_platform_config_field to use the new flag. Reviewed-by:
-
Dennis Dalessandro authored
A couple bits are used by OPA for link physical state that are not present as part of InfiniBand. Add a short blurb what those states mean and removed an unused state. Cc: Leon Romanovsky <leon@kernel.org> Reviewed-by:
Todd Rimmer <todd.rimmer@intel.com> Reviewed-by:
Brent Rothermel <brent.r.rothermel@intel.com> Signed-off-by:
-
Sebastian Sanchez authored
Check xchg returned value for queuing link down entry to guarantee proper atomic value reads. Fixes: 626c077c ("IB/hfi1: Prevent link down request double queuing") Reviewed-by:
Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by:
-
Colin Ian King authored
Trivial fix to spelling mistake in pr_warn_ratelimited warning message Signed-off-by:
Colin Ian King <colin.king@canonical.com> Acked-by:
-
Dasaratharaman Chandramouli authored
Enabling this bit helps core components query for extended address support using the rdma_cap_opa_ah interface. Reviewed-by:
Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by:
Don Hiatt <don.hiatt@intel.com> Signed-off-by:
-
Don Hiatt authored
PIO/SDMA send logic now uses the hdr_type field to determine the type of packet that has been constructed. Based on the hdr_type, certain things such as PBC flags, padding count and the LT extra trailing bytes are determined. Reviewed-by:
-
Don Hiatt authored
Add 16B bypass packet support for RC/UC traffic types. Reviewed-by:
-
Dasaratharaman Chandramouli authored
Increase lid used in hfi1 driver to 32 bits. qib continues to use 16 bit lids. Reviewed-by:
-
Don Hiatt authored
Add trace support to 16B bypass packets during send and receive. Sample input header trace: <idle>-0 [000] d.h. 271742.509477: input_ibhdr: [0000:05:00.0] (16B) len:24 sc:0 dlid:0xf0000b slid:0x10002 age:0 becn:0 fecn:0 l4:10 rc:0 sc:0 pkey:0x8001 entropy:0x0000 op:0x65,UD_SEND_ONLY_WITH_IMMEDIATE se:0 m:1 pad:3 tver:0 qpn:0xffffff a:0 psn:0x00000001 hlen:248 deth qkey 0x01234567 sqpn 0x000004 Reviewed-by:
-
Don Hiatt authored
Add 16B bypass packet support for UD traffic types. Reviewed-by:
-
Don Hiatt authored
When address handle attributes are initialized, the LIDs are transformed to be in the 32 bit LID space. When constructing the header, hfi1 driver will look at the LID to determine the packet header to be created. Reviewed-by:
-
Don Hiatt authored
Enhance hdr_rcverr() to also handle errors during 16B bypass packet receive. Reviewed-by:
-
Don Hiatt authored
We introduce struct hfi1_opa_header as a union of ib (9B) and 16B headers. Reviewed-by:
-
Don Hiatt authored
We introduce a struct hfi1_16b_header to support 16B headers. 16B bypass packets are received by the driver and processed similar to 9B packets. Add basic support to handle 16B packets. Reviewed-by:
-
Don Hiatt authored
rvt_check_ah() delegates lid verification to underlying driver. Underlying driver uses different conditions to check for dlid depending on whether the device supports extended LIDs Reviewed-by:
-
Michael J. Ruhl authored
There is a mixture of mutex and spinlocks to protect receive context (rcd/uctxt) information. This is not used consistently. Use the mutex to protect device receive context information only. Use the spinlock to protect sub context information only. Protect access to items in the rcd array with a spinlock and reference count. Remove spinlock around dd->rcd array cleanup. Since interrupts are disabled and cleaned up before this point, this lock is not useful. Reviewed-by:
-
Michael J. Ruhl authored
The rcd array can be accessed from user context or during interrupts. Protecting this with a mutex isn't a good idea because the mutex should not be used from an IRQ. Protect the allocation and freeing of rcd array elements with a spinlock. Reviewed-by:
-
Bartlomiej Dudek authored
When DC is shut down (by e.g. disconnecting the cable), the driver should use host_link_state to get port's current physical state. This is due to the fact that physical state is read from DC's CSRs and when DC is shut down and state is changed, its registers are not impacted. Reviewed-by:
Bartlomiej Dudek <bartlomiej.dudek@intel.com> Signed-off-by:
-
Byczkowski, Jakub authored
Do not track logical state separately from host_link_state. Deduce logical state from host_link_state when required. Transitions in set_link_state and goto_offline already make sure host_link_state reflects hardware's logical state properly. Reviewed-by:
-
Sebastian Sanchez authored
The pmtu field doens't have be stored in the QP structure as it can easily be calculated when needed. Reviewed-by:
-
Alex Estrin authored
Current code has some serious flaws. Disarm the flag pending an appropriate patch. Fixes: 53526500 ("IB/hfi1: Permanently enable P_Key checking in HFI") Cc: stable@vger.kernel.org Reviewed-by:
Alex Estrin <alex.estrin@intel.com> Signed-off-by:
-
Amrani, Ram authored
Most user verbs pass user data to the kernel with the inclusion of the ib_uverbs_cmd_hdr structure. This is problematic because the vendor has no ideas if the verb was called by a legacy verb or an extended verb. Also, the incosistency between the verbs is confusing. Fixes: 565197dd ("IB/core: Extend ib_uverbs_create_cq") Signed-off-by:
Ram Amrani <Ram.Amrani@cavium.com> Signed-off-by:
Ariel Elior <Ariel.Elior@cavium.com> Signed-off-by:
-
Romain Perier authored
The PCI pool API is deprecated. This commit replaces the PCI pool old API by the appropriate function with the DMA pool API. Signed-off-by:
Romain Perier <romain.perier@collabora.com> Reviewed-by:
Peter Senna Tschudin <peter.senna@collabora.com> Acked-by:
-
