From nobody Thu Dec 18 13:07:57 2025 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 46246C32792 for ; Tue, 23 Aug 2022 09:18:08 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1349300AbiHWJSG (ORCPT ); Tue, 23 Aug 2022 05:18:06 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:47430 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1349336AbiHWJPz (ORCPT ); Tue, 23 Aug 2022 05:15:55 -0400 Received: from ams.source.kernel.org (ams.source.kernel.org [IPv6:2604:1380:4601:e00::1]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 62E7566A79; Tue, 23 Aug 2022 01:32:18 -0700 (PDT) Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ams.source.kernel.org (Postfix) with ESMTPS id 866B3B81C29; Tue, 23 Aug 2022 08:32:17 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id D1EDCC433D6; Tue, 23 Aug 2022 08:32:15 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1661243536; bh=pDZYCUKc7EsLtF1IZaPH9JRuv/2vLuG01EJzQdJBSrM=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=zTElMdIU0LJkckEintWF1VuMjKF6V1XOMoJYFNY1JOldtdmxEFF40meN01hUUoY8o d2PSYj8v0JQxx3z8vGW9jcwW5HSX5eGOq+EHL0SASjLVoDePybW5T8SlMdsrrKJbCz z7dQOWyHH2O0JUpCSrhQppQeJH8aaRNDt8+Z4im8= From: Greg Kroah-Hartman To: linux-kernel@vger.kernel.org Cc: Greg Kroah-Hartman , stable@vger.kernel.org, Md Haris Iqbal , Bob Pearson , Li Zhijian , Leon Romanovsky , Sasha Levin Subject: [PATCH 5.19 308/365] Revert "RDMA/rxe: Create duplicate mapping tables for FMRs" Date: Tue, 23 Aug 2022 10:03:29 +0200 Message-Id: <20220823080131.059195278@linuxfoundation.org> X-Mailer: git-send-email 2.37.2 In-Reply-To: <20220823080118.128342613@linuxfoundation.org> References: <20220823080118.128342613@linuxfoundation.org> User-Agent: quilt/0.67 MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Type: text/plain; charset="utf-8" From: Li Zhijian [ Upstream commit 1e75550648da1fa1cd1969e7597355de8fe8caf6 ] Below 2 commits will be reverted: commit 8ff5f5d9d8cf ("RDMA/rxe: Prevent double freeing rxe_map_set()") commit 647bf13ce944 ("RDMA/rxe: Create duplicate mapping tables for FMRs") The community has a few bug reports which pointed this commit at last. Some proposals are raised up in the meantime but all of them have no follow-up operation. The previous commit led the map_set of FMR to be not available any more if the MR is registered again after invalidating. Although the mentioned patch try to fix a potential race in building/accessing the same table for fast memory regions, it broke rtrs etc ULPs. Since the latter could be worse, revert this patch. With previous commit, it's observed that a same MR in rnbd server will trigger below code path: -> rxe_mr_init_fast() |-> alloc map_set() # map_set is uninitialized |...-> rxe_map_mr_sg() # build the map_set |-> rxe_mr_set_page() |...-> rxe_reg_fast_mr() # mr->state change to VALID from FREE that means # we can access host memory(such rxe_mr_copy) |...-> rxe_invalidate_mr() # mr->state change to FREE from VALID |...-> rxe_reg_fast_mr() # mr->state change to VALID from FREE, # but map_set was not built again |...-> rxe_mr_copy() # kernel crash due to access wild addresses # that lookup from the map_set The backtraces are not always identical. [1st]---------- RIP: 0010:lookup_iova+0x66/0xa0 [rdma_rxe] Code: 00 00 00 48 d3 ee 89 32 c3 4c 8b 18 49 8b 3b 48 8b 47 08 48 39 c6 7= 2 38 48 29 c6 45 31 d2 b8 01 00 00 00 48 63 c8 48 c1 e1 04 <48> 8b 4c 0f 08= 48 39 f1 77 21 83 c0 01 48 29 ce 3d 00 01 00 00 75 RSP: 0018:ffffb7ff80063bf0 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff9b9949d86800 RCX: 0000000000000000 RDX: ffffb7ff80063c00 RSI: 0000000049f6b378 RDI: 002818da00000004 RBP: 0000000000000120 R08: ffffb7ff80063c08 R09: ffffb7ff80063c04 R10: 0000000000000002 R11: ffff9b9916f7eef8 R12: ffff9b99488a0038 R13: ffff9b99488a0038 R14: ffff9b9914fb346a R15: ffff9b990ab27000 FS: 0000000000000000(0000) GS:ffff9b997dc00000(0000) knlGS:0000000000000= 000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007efc33a98ed0 CR3: 0000000014f32004 CR4: 00000000001706f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: rxe_mr_copy.part.0+0x6f/0x140 [rdma_rxe] rxe_responder+0x12ee/0x1b60 [rdma_rxe] ? rxe_icrc_check+0x7e/0x100 [rdma_rxe] ? rxe_rcv+0x1d0/0x780 [rdma_rxe] ? rxe_icrc_hdr.isra.0+0xf6/0x160 [rdma_rxe] rxe_do_task+0x67/0xb0 [rdma_rxe] rxe_xmit_packet+0xc7/0x210 [rdma_rxe] rxe_requester+0x680/0xee0 [rdma_rxe] ? update_load_avg+0x5f/0x690 ? update_load_avg+0x5f/0x690 ? rtrs_clt_recv_done+0x1b/0x30 [rtrs_client] [2nd]---------- RIP: 0010:rxe_mr_copy.part.0+0xa8/0x140 [rdma_rxe] Code: 00 00 49 c1 e7 04 48 8b 00 4c 8d 2c d0 48 8b 44 24 10 4d 03 7d 00 8= 5 ed 7f 10 eb 6c 89 54 24 0c 49 83 c7 10 31 c0 85 ed 7e 5e <49> 8b 3f 8b 14= 24 4c 89 f6 48 01 c7 85 d2 74 06 48 89 fe 4c 89 f7 RSP: 0018:ffffae3580063bf8 EFLAGS: 00010202 RAX: 0000000000018978 RBX: ffff9d7ef7a03600 RCX: 0000000000000008 RDX: 000000000000007c RSI: 000000000000007c RDI: ffff9d7ef7a03600 RBP: 0000000000000120 R08: ffffae3580063c08 R09: ffffae3580063c04 R10: ffff9d7efece0038 R11: ffff9d7ec4b1db00 R12: ffff9d7efece0038 R13: ffff9d7ef4098260 R14: ffff9d7f11e23c6a R15: 4c79500065708144 FS: 0000000000000000(0000) GS:ffff9d7f3dc00000(0000) knlGS:0000000000000= 000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007fce47276c60 CR3: 0000000003f66004 CR4: 00000000001706f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: rxe_responder+0x12ee/0x1b60 [rdma_rxe] ? rxe_icrc_check+0x7e/0x100 [rdma_rxe] ? rxe_rcv+0x1d0/0x780 [rdma_rxe] ? rxe_icrc_hdr.isra.0+0xf6/0x160 [rdma_rxe] rxe_do_task+0x67/0xb0 [rdma_rxe] rxe_xmit_packet+0xc7/0x210 [rdma_rxe] rxe_requester+0x680/0xee0 [rdma_rxe] ? update_load_avg+0x5f/0x690 ? update_load_avg+0x5f/0x690 ? rtrs_clt_recv_done+0x1b/0x30 [rtrs_client] rxe_do_task+0x67/0xb0 [rdma_rxe] tasklet_action_common.constprop.0+0x92/0xc0 __do_softirq+0xe1/0x2d8 run_ksoftirqd+0x21/0x30 smpboot_thread_fn+0x183/0x220 ? sort_range+0x20/0x20 kthread+0xe2/0x110 ? kthread_complete_and_exit+0x20/0x20 ret_from_fork+0x22/0x30 Link: https://lore.kernel.org/r/1658805386-2-1-git-send-email-lizhijian@fuj= itsu.com Link: https://lore.kernel.org/all/20220210073655.42281-1-guoqing.jiang@linu= x.dev/T/ Link: https://www.spinics.net/lists/linux-rdma/msg110836.html Link: https://lore.kernel.org/lkml/94a5ea93-b8bb-3a01-9497-e2021f29598a@lin= ux.dev/t/ Tested-by: Md Haris Iqbal Reviewed-by: Bob Pearson Signed-off-by: Li Zhijian Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin --- drivers/infiniband/sw/rxe/rxe_loc.h | 1 - drivers/infiniband/sw/rxe/rxe_mr.c | 199 +++++++++----------------- drivers/infiniband/sw/rxe/rxe_mw.c | 6 +- drivers/infiniband/sw/rxe/rxe_verbs.c | 39 +++-- drivers/infiniband/sw/rxe/rxe_verbs.h | 21 ++- 5 files changed, 104 insertions(+), 162 deletions(-) diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rx= e/rxe_loc.h index 37484a559d20..d86253c6d6b5 100644 --- a/drivers/infiniband/sw/rxe/rxe_loc.h +++ b/drivers/infiniband/sw/rxe/rxe_loc.h @@ -79,7 +79,6 @@ int mr_check_range(struct rxe_mr *mr, u64 iova, size_t le= ngth); int advance_dma_data(struct rxe_dma_info *dma, unsigned int length); int rxe_invalidate_mr(struct rxe_qp *qp, u32 key); int rxe_reg_fast_mr(struct rxe_qp *qp, struct rxe_send_wqe *wqe); -int rxe_mr_set_page(struct ib_mr *ibmr, u64 addr); int rxe_dereg_mr(struct ib_mr *ibmr, struct ib_udata *udata); void rxe_mr_cleanup(struct rxe_pool_elem *elem); =20 diff --git a/drivers/infiniband/sw/rxe/rxe_mr.c b/drivers/infiniband/sw/rxe= /rxe_mr.c index 3add52129006..c28b18d59a06 100644 --- a/drivers/infiniband/sw/rxe/rxe_mr.c +++ b/drivers/infiniband/sw/rxe/rxe_mr.c @@ -24,7 +24,7 @@ u8 rxe_get_next_key(u32 last_key) =20 int mr_check_range(struct rxe_mr *mr, u64 iova, size_t length) { - struct rxe_map_set *set =3D mr->cur_map_set; + =20 switch (mr->type) { case IB_MR_TYPE_DMA: @@ -32,8 +32,8 @@ int mr_check_range(struct rxe_mr *mr, u64 iova, size_t le= ngth) =20 case IB_MR_TYPE_USER: case IB_MR_TYPE_MEM_REG: - if (iova < set->iova || length > set->length || - iova > set->iova + set->length - length) + if (iova < mr->iova || length > mr->length || + iova > mr->iova + mr->length - length) return -EFAULT; return 0; =20 @@ -65,89 +65,41 @@ static void rxe_mr_init(int access, struct rxe_mr *mr) mr->map_shift =3D ilog2(RXE_BUF_PER_MAP); } =20 -static void rxe_mr_free_map_set(int num_map, struct rxe_map_set *set) -{ - int i; - - for (i =3D 0; i < num_map; i++) - kfree(set->map[i]); - - kfree(set->map); - kfree(set); -} - -static int rxe_mr_alloc_map_set(int num_map, struct rxe_map_set **setp) +static int rxe_mr_alloc(struct rxe_mr *mr, int num_buf) { int i; - struct rxe_map_set *set; + int num_map; + struct rxe_map **map =3D mr->map; =20 - set =3D kmalloc(sizeof(*set), GFP_KERNEL); - if (!set) - goto err_out; + num_map =3D (num_buf + RXE_BUF_PER_MAP - 1) / RXE_BUF_PER_MAP; =20 - set->map =3D kmalloc_array(num_map, sizeof(struct rxe_map *), GFP_KERNEL); - if (!set->map) - goto err_free_set; + mr->map =3D kmalloc_array(num_map, sizeof(*map), GFP_KERNEL); + if (!mr->map) + goto err1; =20 for (i =3D 0; i < num_map; i++) { - set->map[i] =3D kmalloc(sizeof(struct rxe_map), GFP_KERNEL); - if (!set->map[i]) - goto err_free_map; + mr->map[i] =3D kmalloc(sizeof(**map), GFP_KERNEL); + if (!mr->map[i]) + goto err2; } =20 - *setp =3D set; - - return 0; - -err_free_map: - for (i--; i >=3D 0; i--) - kfree(set->map[i]); - - kfree(set->map); -err_free_set: - kfree(set); -err_out: - return -ENOMEM; -} - -/** - * rxe_mr_alloc() - Allocate memory map array(s) for MR - * @mr: Memory region - * @num_buf: Number of buffer descriptors to support - * @both: If non zero allocate both mr->map and mr->next_map - * else just allocate mr->map. Used for fast MRs - * - * Return: 0 on success else an error - */ -static int rxe_mr_alloc(struct rxe_mr *mr, int num_buf, int both) -{ - int ret; - int num_map; - BUILD_BUG_ON(!is_power_of_2(RXE_BUF_PER_MAP)); - num_map =3D (num_buf + RXE_BUF_PER_MAP - 1) / RXE_BUF_PER_MAP; =20 mr->map_shift =3D ilog2(RXE_BUF_PER_MAP); mr->map_mask =3D RXE_BUF_PER_MAP - 1; + mr->num_buf =3D num_buf; - mr->max_buf =3D num_map * RXE_BUF_PER_MAP; mr->num_map =3D num_map; - - ret =3D rxe_mr_alloc_map_set(num_map, &mr->cur_map_set); - if (ret) - return -ENOMEM; - - if (both) { - ret =3D rxe_mr_alloc_map_set(num_map, &mr->next_map_set); - if (ret) - goto err_free; - } + mr->max_buf =3D num_map * RXE_BUF_PER_MAP; =20 return 0; =20 -err_free: - rxe_mr_free_map_set(mr->num_map, mr->cur_map_set); - mr->cur_map_set =3D NULL; +err2: + for (i--; i >=3D 0; i--) + kfree(mr->map[i]); + + kfree(mr->map); +err1: return -ENOMEM; } =20 @@ -164,7 +116,6 @@ void rxe_mr_init_dma(struct rxe_pd *pd, int access, str= uct rxe_mr *mr) int rxe_mr_init_user(struct rxe_pd *pd, u64 start, u64 length, u64 iova, int access, struct rxe_mr *mr) { - struct rxe_map_set *set; struct rxe_map **map; struct rxe_phys_buf *buf =3D NULL; struct ib_umem *umem; @@ -172,6 +123,7 @@ int rxe_mr_init_user(struct rxe_pd *pd, u64 start, u64 = length, u64 iova, int num_buf; void *vaddr; int err; + int i; =20 umem =3D ib_umem_get(pd->ibpd.device, start, length, access); if (IS_ERR(umem)) { @@ -185,20 +137,18 @@ int rxe_mr_init_user(struct rxe_pd *pd, u64 start, u6= 4 length, u64 iova, =20 rxe_mr_init(access, mr); =20 - err =3D rxe_mr_alloc(mr, num_buf, 0); + err =3D rxe_mr_alloc(mr, num_buf); if (err) { pr_warn("%s: Unable to allocate memory for map\n", __func__); goto err_release_umem; } =20 - set =3D mr->cur_map_set; - set->page_shift =3D PAGE_SHIFT; - set->page_mask =3D PAGE_SIZE - 1; - - num_buf =3D 0; - map =3D set->map; + mr->page_shift =3D PAGE_SHIFT; + mr->page_mask =3D PAGE_SIZE - 1; =20 + num_buf =3D 0; + map =3D mr->map; if (length > 0) { buf =3D map[0]->buf; =20 @@ -214,29 +164,33 @@ int rxe_mr_init_user(struct rxe_pd *pd, u64 start, u6= 4 length, u64 iova, pr_warn("%s: Unable to get virtual address\n", __func__); err =3D -ENOMEM; - goto err_release_umem; + goto err_cleanup_map; } =20 buf->addr =3D (uintptr_t)vaddr; buf->size =3D PAGE_SIZE; num_buf++; buf++; + } } =20 mr->ibmr.pd =3D &pd->ibpd; mr->umem =3D umem; mr->access =3D access; + mr->length =3D length; + mr->iova =3D iova; + mr->va =3D start; + mr->offset =3D ib_umem_offset(umem); mr->state =3D RXE_MR_STATE_VALID; mr->type =3D IB_MR_TYPE_USER; =20 - set->length =3D length; - set->iova =3D iova; - set->va =3D start; - set->offset =3D ib_umem_offset(umem); - return 0; =20 +err_cleanup_map: + for (i =3D 0; i < mr->num_map; i++) + kfree(mr->map[i]); + kfree(mr->map); err_release_umem: ib_umem_release(umem); err_out: @@ -250,7 +204,7 @@ int rxe_mr_init_fast(struct rxe_pd *pd, int max_pages, = struct rxe_mr *mr) /* always allow remote access for FMRs */ rxe_mr_init(IB_ACCESS_REMOTE, mr); =20 - err =3D rxe_mr_alloc(mr, max_pages, 1); + err =3D rxe_mr_alloc(mr, max_pages); if (err) goto err1; =20 @@ -268,24 +222,21 @@ int rxe_mr_init_fast(struct rxe_pd *pd, int max_pages= , struct rxe_mr *mr) static void lookup_iova(struct rxe_mr *mr, u64 iova, int *m_out, int *n_ou= t, size_t *offset_out) { - struct rxe_map_set *set =3D mr->cur_map_set; - size_t offset =3D iova - set->iova + set->offset; + size_t offset =3D iova - mr->iova + mr->offset; int map_index; int buf_index; u64 length; - struct rxe_map *map; =20 - if (likely(set->page_shift)) { - *offset_out =3D offset & set->page_mask; - offset >>=3D set->page_shift; + if (likely(mr->page_shift)) { + *offset_out =3D offset & mr->page_mask; + offset >>=3D mr->page_shift; *n_out =3D offset & mr->map_mask; *m_out =3D offset >> mr->map_shift; } else { map_index =3D 0; buf_index =3D 0; =20 - map =3D set->map[map_index]; - length =3D map->buf[buf_index].size; + length =3D mr->map[map_index]->buf[buf_index].size; =20 while (offset >=3D length) { offset -=3D length; @@ -295,8 +246,7 @@ static void lookup_iova(struct rxe_mr *mr, u64 iova, in= t *m_out, int *n_out, map_index++; buf_index =3D 0; } - map =3D set->map[map_index]; - length =3D map->buf[buf_index].size; + length =3D mr->map[map_index]->buf[buf_index].size; } =20 *m_out =3D map_index; @@ -317,7 +267,7 @@ void *iova_to_vaddr(struct rxe_mr *mr, u64 iova, int le= ngth) goto out; } =20 - if (!mr->cur_map_set) { + if (!mr->map) { addr =3D (void *)(uintptr_t)iova; goto out; } @@ -330,13 +280,13 @@ void *iova_to_vaddr(struct rxe_mr *mr, u64 iova, int = length) =20 lookup_iova(mr, iova, &m, &n, &offset); =20 - if (offset + length > mr->cur_map_set->map[m]->buf[n].size) { + if (offset + length > mr->map[m]->buf[n].size) { pr_warn("crosses page boundary\n"); addr =3D NULL; goto out; } =20 - addr =3D (void *)(uintptr_t)mr->cur_map_set->map[m]->buf[n].addr + offset; + addr =3D (void *)(uintptr_t)mr->map[m]->buf[n].addr + offset; =20 out: return addr; @@ -372,7 +322,7 @@ int rxe_mr_copy(struct rxe_mr *mr, u64 iova, void *addr= , int length, return 0; } =20 - WARN_ON_ONCE(!mr->cur_map_set); + WARN_ON_ONCE(!mr->map); =20 err =3D mr_check_range(mr, iova, length); if (err) { @@ -382,7 +332,7 @@ int rxe_mr_copy(struct rxe_mr *mr, u64 iova, void *addr= , int length, =20 lookup_iova(mr, iova, &m, &i, &offset); =20 - map =3D mr->cur_map_set->map + m; + map =3D mr->map + m; buf =3D map[0]->buf + i; =20 while (length > 0) { @@ -628,9 +578,8 @@ int rxe_invalidate_mr(struct rxe_qp *qp, u32 key) int rxe_reg_fast_mr(struct rxe_qp *qp, struct rxe_send_wqe *wqe) { struct rxe_mr *mr =3D to_rmr(wqe->wr.wr.reg.mr); - u32 key =3D wqe->wr.wr.reg.key & 0xff; + u32 key =3D wqe->wr.wr.reg.key; u32 access =3D wqe->wr.wr.reg.access; - struct rxe_map_set *set; =20 /* user can only register MR in free state */ if (unlikely(mr->state !=3D RXE_MR_STATE_FREE)) { @@ -646,36 +595,19 @@ int rxe_reg_fast_mr(struct rxe_qp *qp, struct rxe_sen= d_wqe *wqe) return -EINVAL; } =20 + /* user is only allowed to change key portion of l/rkey */ + if (unlikely((mr->lkey & ~0xff) !=3D (key & ~0xff))) { + pr_warn("%s: key =3D 0x%x has wrong index mr->lkey =3D 0x%x\n", + __func__, key, mr->lkey); + return -EINVAL; + } + mr->access =3D access; - mr->lkey =3D (mr->lkey & ~0xff) | key; - mr->rkey =3D (access & IB_ACCESS_REMOTE) ? mr->lkey : 0; + mr->lkey =3D key; + mr->rkey =3D (access & IB_ACCESS_REMOTE) ? key : 0; + mr->iova =3D wqe->wr.wr.reg.mr->iova; mr->state =3D RXE_MR_STATE_VALID; =20 - set =3D mr->cur_map_set; - mr->cur_map_set =3D mr->next_map_set; - mr->cur_map_set->iova =3D wqe->wr.wr.reg.mr->iova; - mr->next_map_set =3D set; - - return 0; -} - -int rxe_mr_set_page(struct ib_mr *ibmr, u64 addr) -{ - struct rxe_mr *mr =3D to_rmr(ibmr); - struct rxe_map_set *set =3D mr->next_map_set; - struct rxe_map *map; - struct rxe_phys_buf *buf; - - if (unlikely(set->nbuf =3D=3D mr->num_buf)) - return -ENOMEM; - - map =3D set->map[set->nbuf / RXE_BUF_PER_MAP]; - buf =3D &map->buf[set->nbuf % RXE_BUF_PER_MAP]; - - buf->addr =3D addr; - buf->size =3D ibmr->page_size; - set->nbuf++; - return 0; } =20 @@ -695,14 +627,15 @@ int rxe_dereg_mr(struct ib_mr *ibmr, struct ib_udata = *udata) void rxe_mr_cleanup(struct rxe_pool_elem *elem) { struct rxe_mr *mr =3D container_of(elem, typeof(*mr), elem); + int i; =20 rxe_put(mr_pd(mr)); - ib_umem_release(mr->umem); =20 - if (mr->cur_map_set) - rxe_mr_free_map_set(mr->num_map, mr->cur_map_set); + if (mr->map) { + for (i =3D 0; i < mr->num_map; i++) + kfree(mr->map[i]); =20 - if (mr->next_map_set) - rxe_mr_free_map_set(mr->num_map, mr->next_map_set); + kfree(mr->map); + } } diff --git a/drivers/infiniband/sw/rxe/rxe_mw.c b/drivers/infiniband/sw/rxe= /rxe_mw.c index 824739008d5b..6c24bc4318e8 100644 --- a/drivers/infiniband/sw/rxe/rxe_mw.c +++ b/drivers/infiniband/sw/rxe/rxe_mw.c @@ -112,15 +112,15 @@ static int rxe_check_bind_mw(struct rxe_qp *qp, struc= t rxe_send_wqe *wqe, =20 /* C10-75 */ if (mw->access & IB_ZERO_BASED) { - if (unlikely(wqe->wr.wr.mw.length > mr->cur_map_set->length)) { + if (unlikely(wqe->wr.wr.mw.length > mr->length)) { pr_err_once( "attempt to bind a ZB MW outside of the MR\n"); return -EINVAL; } } else { - if (unlikely((wqe->wr.wr.mw.addr < mr->cur_map_set->iova) || + if (unlikely((wqe->wr.wr.mw.addr < mr->iova) || ((wqe->wr.wr.mw.addr + wqe->wr.wr.mw.length) > - (mr->cur_map_set->iova + mr->cur_map_set->length)))) { + (mr->iova + mr->length)))) { pr_err_once( "attempt to bind a VA MW outside of the MR\n"); return -EINVAL; diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/= rxe/rxe_verbs.c index 9d995854a174..d2b4e68402d4 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.c +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c @@ -967,26 +967,41 @@ static struct ib_mr *rxe_alloc_mr(struct ib_pd *ibpd,= enum ib_mr_type mr_type, return ERR_PTR(err); } =20 -/* build next_map_set from scatterlist - * The IB_WR_REG_MR WR will swap map_sets - */ +static int rxe_set_page(struct ib_mr *ibmr, u64 addr) +{ + struct rxe_mr *mr =3D to_rmr(ibmr); + struct rxe_map *map; + struct rxe_phys_buf *buf; + + if (unlikely(mr->nbuf =3D=3D mr->num_buf)) + return -ENOMEM; + + map =3D mr->map[mr->nbuf / RXE_BUF_PER_MAP]; + buf =3D &map->buf[mr->nbuf % RXE_BUF_PER_MAP]; + + buf->addr =3D addr; + buf->size =3D ibmr->page_size; + mr->nbuf++; + + return 0; +} + static int rxe_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents, unsigned int *sg_offset) { struct rxe_mr *mr =3D to_rmr(ibmr); - struct rxe_map_set *set =3D mr->next_map_set; int n; =20 - set->nbuf =3D 0; + mr->nbuf =3D 0; =20 - n =3D ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, rxe_mr_set_page); + n =3D ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, rxe_set_page); =20 - set->va =3D ibmr->iova; - set->iova =3D ibmr->iova; - set->length =3D ibmr->length; - set->page_shift =3D ilog2(ibmr->page_size); - set->page_mask =3D ibmr->page_size - 1; - set->offset =3D set->iova & set->page_mask; + mr->va =3D ibmr->iova; + mr->iova =3D ibmr->iova; + mr->length =3D ibmr->length; + mr->page_shift =3D ilog2(ibmr->page_size); + mr->page_mask =3D ibmr->page_size - 1; + mr->offset =3D mr->iova & mr->page_mask; =20 return n; } diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/= rxe/rxe_verbs.h index 9bdf33346511..3d524238e5c4 100644 --- a/drivers/infiniband/sw/rxe/rxe_verbs.h +++ b/drivers/infiniband/sw/rxe/rxe_verbs.h @@ -289,17 +289,6 @@ struct rxe_map { struct rxe_phys_buf buf[RXE_BUF_PER_MAP]; }; =20 -struct rxe_map_set { - struct rxe_map **map; - u64 va; - u64 iova; - size_t length; - u32 offset; - u32 nbuf; - int page_shift; - int page_mask; -}; - static inline int rkey_is_mw(u32 rkey) { u32 index =3D rkey >> 8; @@ -317,20 +306,26 @@ struct rxe_mr { u32 rkey; enum rxe_mr_state state; enum ib_mr_type type; + u64 va; + u64 iova; + size_t length; + u32 offset; int access; =20 + int page_shift; + int page_mask; int map_shift; int map_mask; =20 u32 num_buf; + u32 nbuf; =20 u32 max_buf; u32 num_map; =20 atomic_t num_mw; =20 - struct rxe_map_set *cur_map_set; - struct rxe_map_set *next_map_set; + struct rxe_map **map; }; =20 enum rxe_mw_state { --=20 2.35.1