aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/infiniband
AgeCommit message (Collapse)AuthorFilesLines
2016-11-16IB/mlx5: Limit mkey page size to 2GBMajd Dibbiny6-8/+18
The maximum page size in the mkey context is 2GB. Until today, we didn't enforce this requirement in the code, and therefore, if we got a page size larger than 2GB, we have passed zeros in the log_page_shift instead of the actual value and the registration failed. This patch limits the driver to use compound pages of 2GB for mkeys. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Majd Dibbiny <majd@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Fix reported max SGE calculationEli Cohen2-4/+33
Add the 512 bytes limit of RDMA READ and the size of remote address to the max SGE calculation. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Eli Cohen <eli@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Wait for all async command completions to completeEli Cohen1-0/+28
Wait before continuing unload till all pending mkey async creation requests are done. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Eli Cohen <eli@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Put non zero value in max_ahMaor Gottlieb1-0/+1
We put INT_MAX since this is the max value that can be held. Though there is no hardware limitation, this is practically a large enough number so we can use it. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Reviewed-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Fix atomic cap in indirect UMRMaor Gottlieb1-5/+7
Remove from the driver the limitation imposed by firmware check to not allow change of atomic permissions for indirect UMRs. In order to avoid failures on old firmware, we only ask for change of atomic permissions if atomic operations are supported. Fixes: 968e78dd9644 ('IB/mlx5: Enhance UMR support to allow partial page table update') Signed-off-by: Eli Cohen <eli@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Replace numerical constant with predefined MACROMax Gurtovoy1-1/+1
Replace the pre-defined macro signifying inline umr instead of the numerical constant. Signed-off-by: Max Gurtovoy <maxg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/core: Save QP in ib_flow structureMark Bloch2-2/+3
When we create flow steering rule, we need to save the related QP in the ib_flow struct. this QP is used in destroy flow. Move the QP assignment from ib_uverbs_ex_create_flow into ib_create_flow, this would allow both kernel and userspace consumers to use it. This bug wasn't seen in the wild because there are no kernel consumers currently in the kernel. Fixes: 319a441d1361 ("IB/core: Add receive flow steering support") Signed-off-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/rxe: Update qp state for user queryYonatan Cohen1-0/+1
The method rxe_qp_error() transitions QP to error state and make sure the QP is drained. It did not though update the QP state for user's query. This patch fixes this. Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen <yonatanc@mellanox.com> Reviewed-by: Moni Shoua <monis@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/rxe: Clear queue buffer when modifying QP to resetYonatan Cohen3-0/+12
RXE resets the send-q only once in rxe_qp_init_req() when QP is created, but when the QP is reused after QP reset, the send-q holds previous garbage data. This garbage data wrongly fails CQEs that otherwise should have completed successfully. Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen <yonatanc@mellanox.com> Reviewed-by: Moni Shoua <monis@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/rxe: Fix handling of erroneous WRYonatan Cohen1-8/+13
To correctly handle a erroneous WR this fix does the following 1. Make sure the bad WQE causes a user completion event. 2. Call rxe_completer to handle the erred WQE. Before the fix, when rxe_requester found a bad WQE, it changed its status to IB_WC_LOC_PROT_ERR and exit with 0 for non RC QPs. If this was the 1st WQE then there would be no ACK to invoke the completer and this bad WQE would be stuck in the QP's send-q. On top of that the requester exiting with 0 caused rxe_do_task to endlessly invoke rxe_requester, resulting in a soft-lockup attached below. In case the WQE was not the 1st and rxe_completer did get a chance to handle the bad WQE, it did not cause a complete event since the WQE's IB_SEND_SIGNALED flag was not set. Setting WQE status to IB_SEND_SIGNALED is subject to IBA spec version 1.2.1, section 10.7.3.1 Signaled Completions. NMI watchdog: BUG: soft lockup - CPU#7 stuck for 22s! [<ffffffffa0590145>] ? rxe_pool_get_index+0x35/0xb0 [rdma_rxe] [<ffffffffa05952ec>] lookup_mem+0x3c/0xc0 [rdma_rxe] [<ffffffffa0595534>] copy_data+0x1c4/0x230 [rdma_rxe] [<ffffffffa058c180>] rxe_requester+0x9d0/0x1100 [rdma_rxe] [<ffffffff8158e98a>] ? kfree_skbmem+0x5a/0x60 [<ffffffffa05962c9>] rxe_do_task+0x89/0xf0 [rdma_rxe] [<ffffffffa05963e2>] rxe_run_task+0x12/0x30 [rdma_rxe] [<ffffffffa059110a>] rxe_post_send+0x41a/0x550 [rdma_rxe] [<ffffffff811ef922>] ? __kmalloc+0x182/0x200 [<ffffffff816ba512>] ? down_read+0x12/0x40 [<ffffffffa054bd32>] ib_uverbs_post_send+0x532/0x540 [ib_uverbs] [<ffffffff815f8722>] ? tcp_sendmsg+0x402/0xb80 [<ffffffffa05453dc>] ib_uverbs_write+0x18c/0x3f0 [ib_uverbs] [<ffffffff81623c2e>] ? inet_recvmsg+0x7e/0xb0 [<ffffffff8158764d>] ? sock_recvmsg+0x3d/0x50 [<ffffffff81215b87>] __vfs_write+0x37/0x140 [<ffffffff81216892>] vfs_write+0xb2/0x1b0 [<ffffffff81217ce5>] SyS_write+0x55/0xc0 [<ffffffff816bc672>] entry_SYSCALL_64_fastpath+0x1a/0xa Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen <yonatanc@mellanox.com> Reviewed-by: Moni Shoua <monis@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/rxe: Fix kernel panic in UDP tunnel with GRO and RX checksumYonatan Cohen1-6/+2
Missing initialization of udp_tunnel_sock_cfg causes to following kernel panic, while kernel tries to execute gro_receive(). While being there, we converted udp_port_cfg to use the same initialization scheme as udp_tunnel_sock_cfg. ------------[ cut here ]------------ kernel tried to execute NX-protected page - exploit attempt? (uid: 0) BUG: unable to handle kernel paging request at ffffffffa0588c50 IP: [<ffffffffa0588c50>] __this_module+0x50/0xffffffffffff8400 [ib_rxe] PGD 1c09067 PUD 1c0a063 PMD bb394067 PTE 80000000ad5e8163 Oops: 0011 [#1] SMP Modules linked in: ib_rxe ip6_udp_tunnel udp_tunnel CPU: 5 PID: 0 Comm: swapper/5 Not tainted 4.7.0-rc3+ #2 Hardware name: Red Hat KVM, BIOS Bochs 01/01/2011 task: ffff880235e4e680 ti: ffff880235e68000 task.ti: ffff880235e68000 RIP: 0010:[<ffffffffa0588c50>] [<ffffffffa0588c50>] __this_module+0x50/0xffffffffffff8400 [ib_rxe] RSP: 0018:ffff880237343c80 EFLAGS: 00010282 RAX: 00000000dffe482d RBX: ffff8800ae330900 RCX: 000000002001b712 RDX: ffff8800ae330900 RSI: ffff8800ae102578 RDI: ffff880235589c00 RBP: ffff880237343cb0 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800ae33e262 R13: ffff880235589c00 R14: 0000000000000014 R15: ffff8800ae102578 FS: 0000000000000000(0000) GS:ffff880237340000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffffffffa0588c50 CR3: 0000000001c06000 CR4: 00000000000006e0 Stack: ffffffff8160860e ffff8800ae330900 ffff8800ae102578 0000000000000014 000000000000004e ffff8800ae102578 ffff880237343ce0 ffffffff816088fb 0000000000000000 ffff8800ae330900 0000000000000000 00000000ffad0000 Call Trace: <IRQ> [<ffffffff8160860e>] ? udp_gro_receive+0xde/0x130 [<ffffffff816088fb>] udp4_gro_receive+0x10b/0x2d0 [<ffffffff81611373>] inet_gro_receive+0x1d3/0x270 [<ffffffff81594e29>] dev_gro_receive+0x269/0x3b0 [<ffffffff81595188>] napi_gro_receive+0x38/0x120 [<ffffffffa011caee>] mlx5e_handle_rx_cqe+0x27e/0x340 [mlx5_core] [<ffffffffa011d076>] mlx5e_poll_rx_cq+0x66/0x6d0 [mlx5_core] [<ffffffffa011d7ae>] mlx5e_napi_poll+0x8e/0x400 [mlx5_core] [<ffffffff815949a0>] net_rx_action+0x160/0x380 [<ffffffff816a9197>] __do_softirq+0xd7/0x2c5 [<ffffffff81085c35>] irq_exit+0xf5/0x100 [<ffffffff816a8f16>] do_IRQ+0x56/0xd0 [<ffffffff816a6dcc>] common_interrupt+0x8c/0x8c <EOI> [<ffffffff81061f96>] ? native_safe_halt+0x6/0x10 [<ffffffff81037ade>] default_idle+0x1e/0xd0 [<ffffffff8103828f>] arch_cpu_idle+0xf/0x20 [<ffffffff810c37dc>] default_idle_call+0x3c/0x50 [<ffffffff810c3b13>] cpu_startup_entry+0x323/0x3c0 [<ffffffff81050d8c>] start_secondary+0x15c/0x1a0 RIP [<ffffffffa0588c50>] __this_module+0x50/0xffffffffffff8400 [ib_rxe] RSP <ffff880237343c80> CR2: ffffffffa0588c50 ---[ end trace 489ee31fa7614ac5 ]--- Kernel panic - not syncing: Fatal exception in interrupt Kernel Offset: disabled ---[ end Kernel panic - not syncing: Fatal exception in interrupt ------------[ cut here ]------------ Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen <yonatanc@mellanox.com> Reviewed-by: Moni Shoua <monis@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx4: Fix create CQ error flowMatan Barak1-1/+4
Currently, if ib_copy_to_udata fails, the CQ won't be deleted from the radix tree and the HW (HW2SW). Fixes: 225c7b1feef1 ('IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters') Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Reviewed-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx4: Check gid_index return valueDaniel Jurgens1-1/+4
Check the returned GID index value and return an error if it is invalid. Fixes: 5070cd2239bd ('IB/mlx4: Replace mechanism for RoCE GID management') Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Reviewed-by: Mark Bloch <markb@mellanox.com> Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Fix NULL pointer dereference on debug printEli Cohen1-2/+2
For XRC QP CQs may not exist. Check before attempting dereference. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Eli Cohen <eli@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Reviewed-by: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Fix fatal error dispatchingEli Cohen1-2/+5
When an internal error condition is detected, make sure to set the device inactive after dispatching the event so ULPs can get a notification of this event. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Eli Cohen <eli@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Reviewed-by: Mohamad Haj Yahia <mohamad@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Resolve soft lock on massive reg MRsMoshe Lazer2-1/+7
When calling reg_mr of large MRs (e.g. 4GB) from multiple processes and MR caches can't supply the required amount of MRs the slow-path of MR allocation may be used. In this case we need to serialize the slow-path between the processes to avoid soft lock. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Moshe Lazer <moshel@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Reviewed-by: Eli Cohen <eli@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Use cache line size to select CQE strideDaniel Jurgens1-2/+1
When creating kernel CQs use 128B CQE stride if the cache line size is 128B, 64B otherwise. This prevents multiple CQEs from residing in a 128B cache line, which can cause retries when there are concurrent read and writes in one cache line. Tested with IPoIB on PPC64, saw ~5% throughput improvement. Fixes: e126ba97dba9 ('mlx5: Add driver for Mellanox Connect-IB adapters') Signed-off-by: Daniel Jurgens <danielj@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Validate requested RQT sizeMaor Gottlieb1-0/+8
Validate that the requested size of RQT is supported by firmware. Fixes: c5f9092936fe ('IB/mlx5: Add Receive Work Queue Indirection table operations') Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Reviewed-by: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/mlx5: Fix memory leak in query deviceMajd Dibbiny1-2/+2
We need to free dev->port when we fail to enable RoCE or initialize node data. Fixes: 0837e86a7a34 ('IB/mlx5: Add per port counters') Signed-off-by: Majd Dibbiny <majd@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Reviewed-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/core: Avoid unsigned int overflow in sg_alloc_tableMark Bloch1-1/+1
sg_alloc_table gets unsigned int as parameter while the driver returns it as size_t. Check npages isn't greater than maximum unsigned int. Fixes: eeb8461e36c9 ("IB: Refactor umem to use linear SG table") Signed-off-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/core: Add missing check for addr_resolve callback return valueMark Bloch1-2/+9
When calling rdma_resolve_ip inside rdma_addr_find_l2_eth_by_grh, the return status of the request was ignored in the callback function causing a successful return and an empty dmac. Signed-off-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Alex Vesker <valex@mellanox.com> Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/core: Set routable RoCE gid type for ipv4/ipv6 networksLeon Romanovsky1-1/+20
On Thu, Oct 27, 2016 at 04:36:28PM +0300, Leon Romanovsky wrote: > From: Mark Bloch <markb@mellanox.com> > > If the underlying netowrk type is ipv4 or ipv6 and the device supports > routable RoCE, prefer it so the traffic could cross subnets. > > Signed-off-by: Mark Bloch <markb@mellanox.com> > Signed-off-by: Maor Gottlieb <maorg@mellanox.com> > Signed-off-by: Leon Romanovsky <leon@kernel.org> > --- Hi Doug, Please take the following v1 of this patch where I fixed spelling error from "netowrk" to be "network". Thanks. >From 09f96ba3e9b4442cfb44dca04c6726e55525c9c3 Mon Sep 17 00:00:00 2001 From: Mark Bloch <markb@mellanox.com> Date: Sun, 11 Sep 2016 06:25:10 +0000 Subject: [PATCH rdma-rc v1 3/6] IB/core: Set routable RoCE gid type for ipv4/ipv6 networks If the underlying network type is ipv4 or ipv6 and the device supports routable RoCE, prefer it so the traffic could cross subnets. Signed-off-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/cm: Mark stale CM id's whenever the mad agent was unregisteredMark Bloch1-16/+110
When there is a CM id object that has port assigned to it, it means that the cm-id asked for the specific port that it should go by it, but if that port was removed (hot-unplug event) the cm-id was not updated. In order to fix that the port keeps a list of all the cm-id's that are planning to go by it, whenever the port is removed it marks all of them as invalid. This commit fixes a kernel panic which happens when running traffic between guests and we force reboot a guest mid traffic, it triggers a kernel panic: Call Trace: [<ffffffff815271fa>] ? panic+0xa7/0x16f [<ffffffff8152b534>] ? oops_end+0xe4/0x100 [<ffffffff8104a00b>] ? no_context+0xfb/0x260 [<ffffffff81084db2>] ? del_timer_sync+0x22/0x30 [<ffffffff8104a295>] ? __bad_area_nosemaphore+0x125/0x1e0 [<ffffffff81084240>] ? process_timeout+0x0/0x10 [<ffffffff8104a363>] ? bad_area_nosemaphore+0x13/0x20 [<ffffffff8104aabf>] ? __do_page_fault+0x31f/0x480 [<ffffffff81065df0>] ? default_wake_function+0x0/0x20 [<ffffffffa0752675>] ? free_msg+0x55/0x70 [mlx5_core] [<ffffffffa0753434>] ? cmd_exec+0x124/0x840 [mlx5_core] [<ffffffff8105a924>] ? find_busiest_group+0x244/0x9f0 [<ffffffff8152d45e>] ? do_page_fault+0x3e/0xa0 [<ffffffff8152a815>] ? page_fault+0x25/0x30 [<ffffffffa024da25>] ? cm_alloc_msg+0x35/0xc0 [ib_cm] [<ffffffffa024e821>] ? ib_send_cm_dreq+0xb1/0x1e0 [ib_cm] [<ffffffffa024f836>] ? cm_destroy_id+0x176/0x320 [ib_cm] [<ffffffffa024fb00>] ? ib_destroy_cm_id+0x10/0x20 [ib_cm] [<ffffffffa034f527>] ? ipoib_cm_free_rx_reap_list+0xa7/0x110 [ib_ipoib] [<ffffffffa034f590>] ? ipoib_cm_rx_reap+0x0/0x20 [ib_ipoib] [<ffffffffa034f5a5>] ? ipoib_cm_rx_reap+0x15/0x20 [ib_ipoib] [<ffffffff81094d20>] ? worker_thread+0x170/0x2a0 [<ffffffff8109b2a0>] ? autoremove_wake_function+0x0/0x40 [<ffffffff81094bb0>] ? worker_thread+0x0/0x2a0 [<ffffffff8109aef6>] ? kthread+0x96/0xa0 [<ffffffff8100c20a>] ? child_rip+0xa/0x20 [<ffffffff8109ae60>] ? kthread+0x0/0xa0 [<ffffffff8100c200>] ? child_rip+0x0/0x20 Fixes: a977049dacde ("[PATCH] IB: Add the kernel CM implementation") Signed-off-by: Mark Bloch <markb@mellanox.com> Signed-off-by: Erez Shitrit <erezsh@mellanox.com> Reviewed-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-16IB/uverbs: Fix leak of XRC target QPsTariq Toukan1-5/+2
The real QP is destroyed in case of the ref count reaches zero, but for XRC target QPs this call was missed and caused to QP leaks. Let's call to destroy for all flows. Fixes: 0e0ec7e0638e ('RDMA/core: Export ib_open_qp() to share XRC...') Signed-off-by: Tariq Toukan <tariqt@mellanox.com> Signed-off-by: Noa Osherovich <noaos@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Optimize pio_buf and send_context structsSebastian Sanchez3-28/+28
Both pio_buf and send_context structs have oversized fields and have cachelines that can be optimized. Reduce oversized fields for both structs. Make sure pio_buf struct fits within a cacheline. Move read-only fields to their own cacheline in send_context struct. All of this will avoid cacheline trading as the ring progresses and pio buffers/send contexts are used. Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Get rid of divide in pio buffer allocatorSebastian Sanchez2-4/+8
The div instruction shows costly in profiles. Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Add active channel and backplane support for integrated devicesEaswar Hariharan5-45/+376
Use scratch registers within the HFI1 device to recover signal integrity information that is then used to tune the channel. While there, update error messages to better convey the result of falling back to a backup file. Reviewed-by: Ira Weiny <ira.weiny@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Optimize devdata cachelinesSebastian Sanchez1-53/+55
Profiling shows hot path struct members that need to be in a minimum set of cachelines. Group these struct member in the same cacheline: sc2vl_lock sc2vl rhf_rcv_function_map rcv_limit rhf_offset Group these struct member in the same cacheline: process_pio_send process_dma_send pport rcd int_counter flags num_pports first_user_ctxt Fill holes in struct hfi1_devdata revealed by pahole. Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Unify access to GUID entriesJakub Pawlak6-37/+39
This patch consolidates the node GUIDs and the port GUID handling and unifies access to these items. The knowledge of hfi1 GUIDs' design and their location are kept in accessors to centralize access. Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Reviewed-by: Brian Welty <brian.welty@intel.com> Signed-off-by: Jakub Pawlak <jakub.pawlak@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Optimize pio cachelinesMike Marciniszyn1-5/+5
Move buffers_allocated pcpu pointer to allocator line. Move hw_free pointer to releaser line. Fill other holes revealed by pahole. Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Optimize lkey validation structuresMike Marciniszyn1-5/+5
Profiling shows that the key validation is susceptible to cache line trading when accessing the lkey table. Fix by separating out the read mostly fields from the write fields. In addition the shift amount, which is function of the lkey table size, is precomputed and stored with the table pointer. Since both the shift and table pointer are in the same read mostly cacheline, this saves a cache line in this hot path. Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Inline sdma_txclean() for verbs pioMike Marciniszyn2-12/+18
Short circuit sdma_txclean() by adding an __sdma_txclean() that is only called when the tx has sdma mappings. Convert internal calls to __sdma_txclean(). This removes a call from the critical path. Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Add unique txwait_lock for txreq eventsMike Marciniszyn6-15/+35
Profiling suggests that the read_seqbegin() in the txreq put logic is colliding with other uses of the iowait lock. The packet at a time use of this lock dictates a unique lock to avoid reader/writer collisions when the number of vTxWait events is low. In order to support a unique lock the iowait struct embedded in the QP is extended to remember the lock that protects the queue head. The QP destroy removes that QP from any wait list. It doesn't need to know the head because of the linked list API, but it does need to know the lock required to protect the head. This also opens up the wait logic to have unique per resources locks which needs to be in future refinement. Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Remove incorrect IS_ERR checkDennis Dalessandro1-1/+1
Remove IS_ERR check from caching code as the function being called does not actually return error pointers. Fixes: f19bd643dbde: "IB/hfi1: Prevent NULL pointer deferences in caching code" Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Prevent hardware counter names from being cut offJianxin Xiong1-1/+1
Increase the size of the buffer that is used to construct per-VL and per-SDMA counter names. Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Jianxin Xiong <jianxin.xiong@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Fix ECN processing in prescan_rxqDasaratharaman Chandramouli1-9/+11
When processing ECN via the prescan_rxq path, some fields in the packet structure are passed uninitialized. This can potentially cause NULL pointer exceptions during ECN handling. Reviewed-by: Ira Weiny <ira.weiny@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Fix status error code for unsupported packetsJakub Pawlak2-2/+18
Set the status code BAD_L2 when unsupported type of packet is received and dropped. Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Jakub Pawlak <jakub.pawlak@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Relocate rcvhdrcnt module parameter check.Krzysztof Blaszkowski1-18/+26
Validate the rcvhdrcnt module parameter in a single function at module load time. This allows proper error reporting. Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Krzysztof Blaszkowski <krzysztof.blaszkowski@intel.com> Signed-off-by: Tymoteusz Kielan <tymoteusz.kielan@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Fix rnr_timer additionIra Weiny1-1/+1
The new s_rnr_timeout was not properly being set and the code was incorrectly setting a different timer. Found by code inspection. Cc: <stable@vger.kernel.org> # 4.7.x Fixes: 08279d5c9424 ("staging/rdma/hfi1: use new RNR timer") Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Ira Weiny <ira.weiny@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Delete unused lockEaswar Hariharan2-3/+0
The lock is an unused vestige from qib. Remove it. Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Clean up unused argumentEaswar Hariharan3-5/+3
hfi1_pcie_ddinit takes the PCI device id as an argument but never uses it. Clean it up. Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Remove leftover snoop referencesDennis Dalessandro3-114/+4
A few snoop related variables were missed in the snoop/capture removal to get out of staging. Go back and clean those up too. Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Fix a potential memory leak in hfi1_create_ctxts()Jianxin Xiong1-6/+7
In the function hfi1_create_ctxts the array "dd->rcd" is allocated and then populated with allocated resources in a loop. Previously, if error happened during the loop, only resource allocated in the current iteration would be freed. The array itself would then be freed, leaving the resources that were allocated in previous iterations and referenced by the array elements in limbo. This patch makes sure all allocated resources are freed before freeing the array "dd->rcd". Also the resource allocation now takes account of the numa node the device is attached to. Reviewed-by: Tadeusz Struk <tadeusz.struk@intel.com> Signed-off-by: Jianxin Xiong <jianxin.xiong@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Return ENODEV for unsupported PCI device ids.Krzysztof Blaszkowski1-13/+12
Clean up device type checking. Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Krzysztof Blaszkowski <krzysztof.blaszkowski@intel.com> Signed-off-by: Tymoteusz Kielan <tymoteusz.kielan@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Fix an Oops on pci device force removeTadeusz Struk4-5/+44
This patch fixes an Oops on device unbind, when the device is used by a PSM user process. PSM processes access device resources which are freed on device removal. Similar protection exists in uverbs in ib_core for Verbs clients, but PSM doesn't use ib_uverbs hence a separate protection is required for PSM clients. Cc: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Fix integrity check flags default valuesJakub Pawlak3-40/+32
Prevent setting up integrity check flags when module is loaded with NO_INTEGRITY capability. Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Jakub Pawlak <jakub.pawlak@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/hfi1: Remove redundant sysfs irq affinity entryTadeusz Struk3-101/+0
The IRQ affinity entry is not needed after the irq notifier patch has been added to the hfi1 driver. The irq affinity settings for SDMA engine should be set using the standard /proc/irq/<N>/ interface. Reviewed-by: Jianxin Xiong <jianxin.xiong@intel.com> Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15IB/rdmavt: rdmavt can handle non aligned page mapsDennis Dalessandro1-3/+0
The initial code for rdmavt carried with it a restriction that was a vestige from the qib driver, that to dma map a page it had to be less than a page size. This is not the case on modern hardware, both qib and hfi1 will be just fine with unaligned map requests. This fixes a 4.8 regression where by an IPoIB transfer of > PAGE_SIZE will hang because the dma map page call always fails. This was introduced after commit 5faba5469522 ("IB/ipoib: Report SG feature regardless of HW UD CSUM capability") added the capability to use SG by default. Rather than override this, the HW supports it, so allow SG. Cc: Stable <stable@vger.kernel.org> # 4.8 Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2016-11-15Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/netDavid S. Miller1-26/+28
Several cases of bug fixes in 'net' overlapping other changes in 'net-next-. Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-11infiniband: shut up a maybe-uninitialized warningArnd Bergmann1-26/+28
Some configurations produce this harmless warning when built with gcc -Wmaybe-uninitialized: infiniband/core/cma.c: In function 'cma_get_net_dev': infiniband/core/cma.c:1242:12: warning: 'src_addr_storage.sin_addr.s_addr' may be used uninitialized in this function [-Wmaybe-uninitialized] I previously reported this for the powerpc64 defconfig, but have now reproduced the same thing for x86 as well, using gcc-5 or higher. The code looks correct to me, and this change just rearranges it by making sure we alway initialize the entire address structure to make the warning disappear. My first approach added an initialization at the time of the declaration, which Doug commented may be too costly, so I hope this version doesn't add overhead. Link: http://arm-soc.lixom.net/buildlogs/mainline/v4.7-rc6/buildall.powerpc.ppc64_defconfig.log.passed Link: https://patchwork.kernel.org/patch/9212825/ Acked-by: Haggai Eran <haggaie@mellanox.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Copyright © 1996 – 2026 Jason A. Donenfeld. All Rights Reverse Engineered.