- 01 Feb, 2018 20 commits
-
-
Alex Estrin authored
On reboot SM can program port pkey table before ipoib registered its event handler, which could result in missing pkey event and leave root interface with initial pkey value from index 0. Since OPA port starts with invalid pkey in index 0, root interface will fail to initialize and stay down with no-carrier flag. For IB ipoib interface may end up with pkey different from value opensm put in pkey table idx 0, resulting in connectivity issues (different mcast groups, for example). Close the window by calling event handler after registration to make sure ipoib pkey is in sync with port pkey table. Reviewed-by:
Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by:
Ira Weiny <ira.weiny@intel.com> Signed-off-by:
Alex Estrin <alex.estrin@intel.com> Signed-off-by:
Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by:
Jason Gunthorpe <jgg@mellanox.com>
-
Mitko Haralanov authored
The routine which shows the fault stats checks the counters to determine whether to show any stats based on the number of transmitted pkts/bytes for a particular opcode. Unfortunately, it only checked the receive counters. As a result, if any packet faults have happened for packets egressing the HFI, those stats would not be shown. In order to fix this, the routine is amended to also check the TX counters. With this change the pkt/byte counts are the sum of both TX and RX counts for the opcode. Fixes: 1b311f89 ("IB/hfi1: Add tx_opcode_stats like the opcode_stats") Reviewed-by:
Don Hiatt <don.hiatt@intel.com> Reviewed-by:
Michael J. Ruhl <michael.j.ruhl@intel.com> Signed-off-by:
Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by:
-
Mike Marciniszyn authored
These values were introduced as part of the 16B code to account for the varying size of the LRH between the differing packet formats. Replace the blind constants with defines based on FIELD_SIZEOF() calls. Fixes: 5b6cabb0 ("IB/hfi1: Add 16B RC/UC support") Reviewed-by:
-
Kamenee Arumugam authored
HFI's counters SendWaitCnt and SendWaitVlCnt are in units of TXE cycle time (at 805MHz). OPA counters PortXmitWait and PortVLXmtWait are in units of flit times. Convert the counter values to flit units using following conversion formula: PortXmitWait = SendWaitCnt * 2 * (4 /link_width) * (25 Gbps /link_speed) PortVLXmitWait = SendWaitVLCnt * 2 * (4 /link_width) * (25 Gbps /link_speed) At link up or downgrade events, the link width can change. To ensure accurate counter calculations, sample the counters after the events, during counter requests, and then aggregate the OPA counters. Reviewed-by:
Kamenee Arumugam <kamenee.arumugam@intel.com> Signed-off-by:
-
Bartlomiej Dudek authored
During PCIe Gen 3 transistion, pcie_pset is read and might be overridden to a default value(i.e. 255) in do_pcie_gen3_transition() routine. If the pcie_pset value is overridden then this new value will be used during initialization of next adapter on a different card. Introducing a new local variable to avoid modification of pcie_pset Reviewed-by:
Bartlomiej Dudek <bartlomiej.dudek@intel.com> Signed-off-by:
Patel Jay P <jay.p.patel@intel.com> Signed-off-by:
-
Sebastian Sanchez authored
The arguments for trace_hfi1_rcvhdr() get computed every time in the hot path regardless of the whether the trace is on or off. This is seen to be costly with a profile. The handling of fault inject isolates the verbs device for all packets regardless of the presence of a RHF_DC_ERR error. Fix the first by computing trace_hfi1_rcvhdr() arguments within the trace itself, so that when the trace is off, the argument data isn't computed. Fix the second by moving the error check to handle_eflags() when an RHF error occurs and by testing for RHF_DC_ERR before executing the reset of handle_eflags(). Reviewed-by:
Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by:
-
Sebastian Sanchez authored
packet->fecn and packet->becn are calculated in the hot path and are never used. Remove these fields as they show to be costly in a profile. Also, remove initialization for becn and fecn in process_ecn() as they're unconditionally assigned in the function and ensure fecn and becn variables use a boolean type. Reviewed-by:
-
Sebastian Sanchez authored
In the receive path, hfi1_ibport is looked up by indexing into an array. A profile shows this to be expensive. The receive context data has a pointer to the ibport data, use that pointer instead. Reviewed-by:
-
Sebastian Sanchez authored
The packet type comparison used to find out if a packet is a bypass packet in the hot path is an expensive operation as seen in a profile. Determine packet's pkey and migration bit through the bypass and 9B code paths instead. Reviewed-by:
-
Sebastian Sanchez authored
In hfi1_rc_rcv(), BTH is computed for all packets received. However, it's only used for packets received with opcodes RDMA_WRITE_LAST and SEND_LAST, and it is a costly operation. Compute BTH only in the RDMA_WRITE_LAST/SEND_LAST code path and let the compiler handle endianness conversion for bitwise operations. Reviewed-by:
-
Mitko Haralanov authored
The s_hdrwords variable was used to indicate whether a packet was already built on a previous iteration of the send engine. This variable assumed the protection of the QP's RVT_S_BUSY flag, which was required since the the QP's s_lock was dropped just prior to the packet being queued on the one of the egress mechanisms. Support for multiple send engine instantiations require that the field not be used due to concurency issues. The ps.txreq signals the "already built" without the potential concurency issues. Fix by getting rid of all s_hdrword usage. A wrapper is added to test for the already built case that used to use s_hdrwords. What used to be stored in s_hdrwords is now in the txreq. The PBC is not counted, but is added in the pio/sdma code paths prior to posting the packet. Reviewed-by:
-
Alex Estrin authored
The dd refcount is speculatively incremented prior to allocating the fd memory with kzalloc(). If that kzalloc() failed the dd refcount leaks. Increment refcount on kzalloc success. Fixes: e11ffbd5 ("IB/hfi1: Do not free hfi1 cdev parent structure early") Reviewed-by:
-
Alex Estrin authored
With IRQF_SHARED flag set and CONFIG_DEBUG_SHIRQ enabled module removal may result in panic in sdma_interrupt() routine if associated sdma context was released before pci_free_irq(); [ 9198.939885] BUG: unable to handle kernel NULL pointer dereference at (null) [ 9198.940514] IP: sdma_make_progress+0xa5/0x450 [hfi1] [ 9198.941114] PGD 170bdc0067 P4D 170bdc0067 PUD 172063e067 PMD 0 [ 9198.941783] Oops: 0000 [#1] SMP ..... [ 9198.958877] CPU: 132 PID: 64173 Comm: rmmod Tainted: G OE 4.14.0-rc4+ #1 [ 9198.961032] Hardware name: Intel Corporation S7200AP/S7200AP, BIOS S72C610.86B.01.02.0118.080620171935 08/06/2017 [ 9198.963323] task: ffff9681397f0000 task.stack: ffffae1647c40000 [ 9198.965695] RIP: 0010:sdma_make_progress+0xa5/0x450 [hfi1] [ 9198.968082] RSP: 0018:ffffae1647c43be8 EFLAGS: 00010046 [ 9198.970503] RAX: 0000000000000000 RBX: ffff9680ce8b5ca8 RCX: 0000000000000000 [ 9198.973006] RDX: 0000000000000000 RSI: 0000000001a00d28 RDI: ffff9680ce8b5ca0 [ 9198.975546] RBP: ffffae1647c43c40 R08: ffff96814325ec00 R09: 00000000ffffffff [ 9198.978142] R10: 000000004325e501 R11: ffff96814325ec00 R12: ffff9680ce8b5c44 [ 9198.980779] R13: ffff9680ce8b5ca0 R14: 0000000000000000 R15: ffff9680ce8b5b00 [ 9198.983462] FS: 00007f31196ba740(0000) GS:ffff96819df00000(0000) knlGS:0000000000000000 [ 9198.986231] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 9198.989036] CR2: 0000000000000000 CR3: 000000170833f000 CR4: 00000000001406e0 [ 9198.991911] Call Trace: [ 9198.994847] sdma_engine_interrupt+0x82/0x100 [hfi1] [ 9198.997852] sdma_interrupt+0x61/0xc0 [hfi1] [ 9199.000852] __free_irq+0x1b3/0x2d0 [ 9199.003873] free_irq+0x35/0x70 [ 9199.006909] pci_free_irq+0x1c/0x30 [ 9199.009999] clean_up_interrupts+0x53/0xf0 [hfi1] [ 9199.013137] hfi1_start_cleanup+0x117/0x190 [hfi1] [ 9199.016315] postinit_cleanup+0x1d/0x270 [hfi1] [ 9199.019529] remove_one+0x1f3/0x210 [hfi1] [ 9199.022738] pci_device_remove+0x39/0xc0 [ 9199.025974] device_release_driver_internal+0x141/0x210 [ 9199.029268] driver_detach+0x3f/0x80 [ 9199.032580] bus_remove_driver+0x55/0xd0 [ 9199.035931] driver_unregister+0x2c/0x50 [ 9199.039321] pci_unregister_driver+0x2a/0xa0 [ 9199.042755] hfi1_mod_cleanup+0x10/0xb50 [hfi1] [ 9199.046196] SyS_delete_module+0x171/0x250 ... Fix by exporting sdma_clean() and removing from sdma_exit(). sdma_exit() now just manipulates the engine state, leaving the memory free to sdma_clean() which is now called just before the dd is freed. Reviewed-by:
-
Michael J. Ruhl authored
The pci_request_irq() interfaces always adds the IRQF_SHARED bit to all IRQ requests. When the kernel is built with CONFIG_DEBUG_SHIRQ config flag, if the IRQF_SHARED bit is set, a call to the IRQ handler is made from the __free_irq() function. This is testing a race condition between the IRQ cleanup and an IRQ racing the cleanup. The HFI driver should be able to handle this race, but does not. This race can cause traces that start with this footprint: BUG: unable to handle kernel NULL pointer dereference at (null) Call Trace: <hfi1 irq handler> ... __free_irq+0x1b3/0x2d0 free_irq+0x35/0x70 pci_free_irq+0x1c/0x30 clean_up_interrupts+0x53/0xf0 [hfi1] hfi1_start_cleanup+0x122/0x190 [hfi1] postinit_cleanup+0x1d/0x280 [hfi1] remove_one+0x233/0x250 [hfi1] pci_device_remove+0x39/0xc0 Export IRQ cleanup function so it can be called from other modules. Using the exported cleanup function: Re-order the driver cleanup code to clean up IRQ resources before other resources, eliminating the race. Re-order error path for init so that the race does not occur. Reduce severity on spurious error message for SDMA IRQs to info. Reviewed-by:
-
Dan Carpenter authored
We should return -ENOMEM if the allocation fails. The current code accidentally returns success. Fixes: bf3c5a93 ("RDMA/nldev: Provide global resource utilization") Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by:
Leon Romanovsky <leonro@mellanox.com> Signed-off-by:
-
oulijun authored
The mtt_table is cleaned up during the err_unmap_cqe label, it is a mistake to duplicate the cleanup during the later unwind labels. Signed-off-by:
Lijun Ou <oulijun@huawei.com> Signed-off-by:
-
oulijun authored
This patch mainly fix some style warings matched with the new checkpatch requirement. The warning as follows: WARNING: function definition argument 'struct hns_roce_cq *' should also have an identifier name Signed-off-by:
-
oulijun authored
The double not-operator is unncessary when used in a boolean context. This patch removes them. Signed-off-by:
-
Markus Elfring authored
Add a jump target so that a bit of exception handling can be better reused at the end of this function. Signed-off-by:
Markus Elfring <elfring@users.sourceforge.net> Acked-by:
Devesh Sharma <devesh.sharma@broadcom.com> Acked-by:
Jonathan Toppins <jtoppins@redhat.com> Signed-off-by:
-
Markus Elfring authored
RDMA/bnxt_re: Delete two error messages for a failed memory allocation in bnxt_qplib_alloc_dpi_tbl() Omit extra messages for a memory allocation failure in this function. This issue was detected by using the Coccinelle software. Signed-off-by:
-
- 31 Jan, 2018 9 commits
-
-
Zhu Yanjun authored
In the function rxe_av_fill_ip_info, the parameter rxe is not used. So it is removed. CC: Srinivas Eeda <srinivas.eeda@oracle.com> CC: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by:
Zhu Yanjun <yanjun.zhu@oracle.com> Reviewed-by:
Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
Doug Ledford <dledford@redhat.com>
-
Zhu Yanjun authored
The function rxe_av_fill_ip_info always returns 0. So the function type is changed to void. CC: Srinivas Eeda <srinivas.eeda@oracle.com> CC: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by:
-
Zhu Yanjun authored
Since the function rxe_av_to_attr always return 0, the function type is changed to void. CC: Srinivas Eeda <srinivas.eeda@oracle.com> CC: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by:
-
Zhu Yanjun authored
In the function rxe_av_to_attr, the parameter rxe is not used. So it is removed. CC: Srinivas Eeda <srinivas.eeda@oracle.com> CC: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by:
-
Zhu Yanjun authored
The function rxe_av_from_attr always return 0. So change the function to void. CC: Srinivas Eeda <srinivas.eeda@oracle.com> CC: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by:
-
Zhu Yanjun authored
In the function rxe_av_from_attr, the parameter rxe is not used. So it is removed. CC: Srinivas Eeda <srinivas.eeda@oracle.com> CC: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by:
-
Leon Romanovsky authored
The nldev was implemented by following devlink implementation, including SET/DEL/NEW commands. However these commands were not implemented and hence don't need to be exposed. Signed-off-by:
-
Dan Carpenter authored
We should return -ENOMEM if the allocation fails. (The current code returns succees). Fixes: 37cb11ac ("RDMA/bnxt_re: Add SRQ support for Broadcom adapters") Signed-off-by:
-
Doug Ledford authored
If there is ever any error while creating srq->umem, we return that error, we don't store it in srq->umem, so any check of srq->umem for IS_ERR is pointless. Further, checking udata is unnecessary as srq->umem is always either NULL or valid, without respect to udata. Reported-by:
-
- 30 Jan, 2018 7 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.gitJason Gunthorpe authored
To resolve conflicts in: drivers/infiniband/hw/mlx5/main.c drivers/infiniband/hw/mlx5/qp.c From patches merged into the -rc cycle. The conflict resolution matches what linux-next has been carrying. Signed-off-by: -
Leon Romanovsky authored
Implement RDMA nldev netlink interface to get detailed information on each QP in the system. This includes the owning process or kernel ULP and detailed information from the qp_attrs. Currently only the dumpit variant is implemented. Reviewed-by:
Mark Bloch <markb@mellanox.com> Signed-off-by:
Steve Wise <swise@opengridcomputing.com> Signed-off-by:
-
Leon Romanovsky authored
Expose through the netlink interface the global per-device utilization of the supported object types. Provide both dumpit and doit callbacks. As an example of possible output from rdmatool for system with 5 mlx5 cards: $ rdma res 1: mlx5_0: qp 4 cq 5 pd 3 2: mlx5_1: qp 4 cq 5 pd 3 3: mlx5_2: qp 4 cq 5 pd 3 4: mlx5_3: qp 2 cq 3 pd 2 5: mlx5_4: qp 4 cq 5 pd 3 Reviewed-by:
-
Leon Romanovsky authored
Track create and destroy operations of PD objects. Reviewed-by:
-
Leon Romanovsky authored
Track create and destroy operations of CQ objects. Reviewed-by:
-
Leon Romanovsky authored
Track create and destroy operations of QP objects. Reviewed-by:
-
Leon Romanovsky authored
The RDMA subsystem has very strict set of objects to work with, but it completely lacks tracking facilities and has no visibility of resource utilization. The following patch adds such infrastructure to keep track of RDMA resources to help with debugging of user space applications. The primary user of this infrastructure is RDMA nldev netlink (following patches), to be exposed to userspace via rdmatool, but it is not limited too that. At this stage, the main three objects (PD, CQ and QP) are added, and more will be added later. Reviewed-by:
-
- 29 Jan, 2018 4 commits
-
-
Leon Romanovsky authored
The KBUILD_MODNAME variable contains the module name and it is known for kernel users during compilation, so let's reuse it to track the owners. Followup patches will store this for resource tracking. Signed-off-by:
-
Leon Romanovsky authored
Each of our modules only allocates a PD in one place, so there isn't any loss in detail, while MODNAME is more useful and recognizable as something to expose to the user. Signed-off-by:
-
Jason Gunthorpe authored
The flags field the enum is used with comes directly from the uapi so it belongs in the uapi headers for clarity and so userspace can use it. Signed-off-by: -
Jason Gunthorpe authored
NET_UDP_TUNNEL is not user selectable, so it should be used as a select in kconfig. CRYPTO_CRC32 is a required library for RDMA_RXE so it should active automatically, as most other CRYPTO_ users do. Signed-off-by:
-
