From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-alma10-1.taild15c8.ts.net [100.103.45.18]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 1C88447A86F; Tue, 21 Jul 2026 17:53:44 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=100.103.45.18 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784656425; cv=none; b=HqorMfxrlIn1aVFTe8cS07CBn53v2VarrFSDgHLnTtFncpwL2OfX5OecY+rgY+gh0K3iMUvLQ7AHp0vH59nap9YblwqL6YiITCmd/T94Usy7ael+iuUpKPhy2UcgBCEKJtVxklhRFx65WkgaaXl4km1Wci1kdYHcHPkvOK73vLc= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784656425; c=relaxed/simple; bh=GXWRRTtq7UqLe9MH2AYN/L2mlEZ6FHmFsCvv7OPrbEk=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: MIME-Version; b=ehgKbnH1bzyBa9WueLKbyYcra5/kL+yLpfY9MjekOiv3IegxDqsxFGI1YpyJq1QmEk8vKuGzR0JEwLNiaDYN58Bug3TXgrrLe/SACUJN5BxzVOFD80I6v2K1YzUaWJ/fOtus7tZCnzdkKt66GMY0yuiER0Fde18+oyskAAkC2rc= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=linuxfoundation.org header.i=@linuxfoundation.org header.b=J+33ngYf; arc=none smtp.client-ip=100.103.45.18 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=linuxfoundation.org header.i=@linuxfoundation.org header.b="J+33ngYf" Received: by smtp.kernel.org (Postfix) with ESMTPSA id 824661F00A3A; Tue, 21 Jul 2026 17:53:43 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linuxfoundation.org; s=korg; t=1784656424; bh=QsSksQ+eXH2uArfE/EqZffGkslL0op1kRX/YnZxT7Sw=; h=From:To:Cc:Subject:Date:In-Reply-To:References; b=J+33ngYfrzaR4tQ0KoddazGfvjanmiDFeu+Z3Qyt5m3eF0o5gL9GwGW+DNIWIeJnq ZO1OI3JqCLbdFy+TWE91UBEk/pk3mgbraT3/QPeQ6wh2cHMY3vybwOyXiCUpyFzGmm sUGVOaViDMCk5ImLsrwFr1hzOIX4PkV2FKqxXHDg= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Jason Gunthorpe , Sasha Levin Subject: [PATCH 6.18 0383/1611] IB/mlx5: Pull the pdn out of the depths of the umr machinery Date: Tue, 21 Jul 2026 17:08:19 +0200 Message-ID: <20260721152523.819151890@linuxfoundation.org> X-Mailer: git-send-email 2.55.0 In-Reply-To: <20260721152514.750365251@linuxfoundation.org> References: <20260721152514.750365251@linuxfoundation.org> User-Agent: quilt/0.69 X-stable: review X-Patchwork-Hint: ignore Precedence: bulk X-Mailing-List: patches@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit 6.18-stable review patch. If anyone has any objections, please let me know. ------------------ From: Jason Gunthorpe [ Upstream commit f387a9f06ea4f05e607a91b161963c0b19436652 ] Instead of getting the pdn deep inside the umr code, pass it in from the top. to_mpd(mr->ibmr.pd)->pdn is not safe due to the rereg races, so all the call sites need some revision to obtain the pdn in a safe way. Mark them with mlx5_mr_pdn(); following patches will go through and remove these. Cases where the XLT flags are known and do not require the PDN can pass 0, such as for mlx5_ib_dmabuf_invalidate_cb(). Also extract the DMABUF data_direct special case from inside the UMR code and into the only place that needs it, pagefault_dmabuf_mr(). The actual mr was created directly without using the UMR flow. Ultimately this will be moved into mlx5_ib_init_dmabuf_mr(). Link: https://patch.msgid.link/r/6-v1-29ebd2c229b5+fd5-ib_mr_pd_jgg@nvidia.com Assisted-by: Codex:gpt-5-5 Signed-off-by: Jason Gunthorpe Stable-dep-of: 8e0a02a989c1 ("IB/mlx5: Don't mangle the mr->pd inside the rereg callback") Signed-off-by: Sasha Levin --- drivers/infiniband/hw/mlx5/mlx5_ib.h | 9 ++++ drivers/infiniband/hw/mlx5/mr.c | 8 +-- drivers/infiniband/hw/mlx5/odp.c | 12 +++-- drivers/infiniband/hw/mlx5/umr.c | 75 ++++++++++++++-------------- drivers/infiniband/hw/mlx5/umr.h | 6 +-- 5 files changed, 64 insertions(+), 46 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index fbccb0362590bb..ba55e6db3bca79 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -331,6 +331,10 @@ struct mlx5_ib_flow_db { #define MLX5_IB_QPT_DCT IB_QPT_RESERVED4 #define MLX5_IB_WR_UMR IB_WR_RESERVED1 +/* + * A valid pdn is required when flags include MLX5_IB_UPD_XLT_ENABLE, + * MLX5_IB_UPD_XLT_PD or MLX5_IB_UPD_XLT_ACCESS. + */ #define MLX5_IB_UPD_XLT_ZAP BIT(0) #define MLX5_IB_UPD_XLT_ENABLE BIT(1) #define MLX5_IB_UPD_XLT_ATOMIC BIT(2) @@ -1281,6 +1285,11 @@ static inline struct mlx5_ib_pd *to_mpd(struct ib_pd *ibpd) return container_of(ibpd, struct mlx5_ib_pd, ibpd); } +static inline u32 mlx5_mr_pdn(struct mlx5_ib_mr *mr) +{ + return to_mpd(mr->ibmr.pd)->pdn; +} + static inline struct mlx5_ib_srq *to_msrq(struct ib_srq *ibsrq) { return container_of(ibsrq, struct mlx5_ib_srq, ibsrq); diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index c581ac46e0c996..3159f6692a4a5d 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -1505,7 +1505,8 @@ static struct ib_mr *create_real_mr(struct ib_pd *pd, struct ib_umem *umem, * configured properly but left disabled. It is safe to go ahead * and configure it again via UMR while enabling it. */ - err = mlx5r_umr_update_mr_pas(mr, MLX5_IB_UPD_XLT_ENABLE); + err = mlx5r_umr_update_mr_pas(mr, MLX5_IB_UPD_XLT_ENABLE, + to_mpd(pd)->pdn); if (err) { mlx5_ib_dereg_mr(&mr->ibmr, NULL); return ERR_PTR(err); @@ -1614,7 +1615,8 @@ static void mlx5_ib_dmabuf_invalidate_cb(struct dma_buf_attachment *attach) if (!umem_dmabuf->sgt || !mr) return; - mlx5r_umr_update_mr_pas(mr, MLX5_IB_UPD_XLT_ZAP); + /* MLX5_IB_UPD_XLT_ZAP does not change the pdn */ + mlx5r_umr_update_mr_pas(mr, MLX5_IB_UPD_XLT_ZAP, 0); ib_umem_dmabuf_unmap_pages(umem_dmabuf); } @@ -1858,7 +1860,7 @@ static int umr_rereg_pas(struct mlx5_ib_mr *mr, struct ib_pd *pd, mr->ibmr.length = new_umem->length; mr->page_shift = order_base_2(page_size); mr->umem = new_umem; - err = mlx5r_umr_update_mr_pas(mr, upd_flags); + err = mlx5r_umr_update_mr_pas(mr, upd_flags, mlx5_mr_pdn(mr)); if (err) { /* * The MR is revoked at this point so there is no issue to free diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c index 0e8ae85af5a625..47bb42576ad5dc 100644 --- a/drivers/infiniband/hw/mlx5/odp.c +++ b/drivers/infiniband/hw/mlx5/odp.c @@ -836,12 +836,14 @@ static int pagefault_dmabuf_mr(struct mlx5_ib_mr *mr, size_t bcnt, u32 *bytes_mapped, u32 flags) { struct ib_umem_dmabuf *umem_dmabuf = to_ib_umem_dmabuf(mr->umem); + struct mlx5_ib_dev *dev = mr_to_mdev(mr); int access_mode = mr->data_direct ? MLX5_MKC_ACCESS_MODE_KSM : MLX5_MKC_ACCESS_MODE_MTT; unsigned int old_page_shift = mr->page_shift; unsigned int page_shift; unsigned long page_size; u32 xlt_flags = 0; + u32 pdn = 0; int err; if (flags & MLX5_PF_FLAGS_ENABLE) @@ -860,8 +862,12 @@ static int pagefault_dmabuf_mr(struct mlx5_ib_mr *mr, size_t bcnt, err = -EINVAL; } else { page_shift = order_base_2(page_size); + if (mr->data_direct) + pdn = dev->ddr.pdn; + else + pdn = mlx5_mr_pdn(mr); if (page_shift != mr->page_shift && mr->dmabuf_faulted) { - err = mlx5r_umr_dmabuf_update_pgsz(mr, xlt_flags, + err = mlx5r_umr_dmabuf_update_pgsz(mr, xlt_flags, pdn, page_shift); } else { mr->page_shift = page_shift; @@ -869,8 +875,8 @@ static int pagefault_dmabuf_mr(struct mlx5_ib_mr *mr, size_t bcnt, err = mlx5r_umr_update_data_direct_ksm_pas( mr, xlt_flags); else - err = mlx5r_umr_update_mr_pas(mr, - xlt_flags); + err = mlx5r_umr_update_mr_pas(mr, xlt_flags, + pdn); } } dma_resv_unlock(umem_dmabuf->attach->dmabuf->resv); diff --git a/drivers/infiniband/hw/mlx5/umr.c b/drivers/infiniband/hw/mlx5/umr.c index 3e2624cfd17f7d..1452cbb825ed67 100644 --- a/drivers/infiniband/hw/mlx5/umr.c +++ b/drivers/infiniband/hw/mlx5/umr.c @@ -603,11 +603,11 @@ mlx5r_umr_set_update_xlt_ctrl_seg(struct mlx5_wqe_umr_ctrl_seg *ctrl_seg, static void mlx5r_umr_set_update_xlt_mkey_seg(struct mlx5_ib_dev *dev, struct mlx5_mkey_seg *mkey_seg, - struct mlx5_ib_mr *mr, + struct mlx5_ib_mr *mr, u32 pdn, unsigned int page_shift) { mlx5r_umr_set_access_flags(dev, mkey_seg, mr->access_flags); - MLX5_SET(mkc, mkey_seg, pd, to_mpd(mr->ibmr.pd)->pdn); + MLX5_SET(mkc, mkey_seg, pd, pdn); MLX5_SET64(mkc, mkey_seg, start_addr, mr->ibmr.iova); MLX5_SET64(mkc, mkey_seg, len, mr->ibmr.length); MLX5_SET(mkc, mkey_seg, log_page_size, page_shift); @@ -670,23 +670,22 @@ static void mlx5r_umr_final_update_xlt(struct mlx5_ib_dev *dev, wqe->data_seg.byte_count = cpu_to_be32(sg->length); } -static void -_mlx5r_umr_init_wqe(struct mlx5_ib_mr *mr, struct mlx5r_umr_wqe *wqe, - struct ib_sge *sg, unsigned int flags, - unsigned int page_shift, bool dd) +static void _mlx5r_umr_init_wqe(struct mlx5_ib_mr *mr, + struct mlx5r_umr_wqe *wqe, struct ib_sge *sg, + unsigned int flags, u32 pdn, + unsigned int page_shift) { struct mlx5_ib_dev *dev = mr_to_mdev(mr); mlx5r_umr_set_update_xlt_ctrl_seg(&wqe->ctrl_seg, flags, sg); - mlx5r_umr_set_update_xlt_mkey_seg(dev, &wqe->mkey_seg, mr, page_shift); - if (dd) /* Use the data direct internal kernel PD */ - MLX5_SET(mkc, &wqe->mkey_seg, pd, dev->ddr.pdn); + mlx5r_umr_set_update_xlt_mkey_seg(dev, &wqe->mkey_seg, mr, pdn, + page_shift); mlx5r_umr_set_update_xlt_data_seg(&wqe->data_seg, sg); } -static int -_mlx5r_umr_update_mr_pas(struct mlx5_ib_mr *mr, unsigned int flags, bool dd, - size_t start_block, size_t nblocks) +static int _mlx5r_umr_update_mr_pas(struct mlx5_ib_mr *mr, unsigned int flags, + u32 pdn, bool dd, size_t start_block, + size_t nblocks) { size_t ent_size = dd ? sizeof(struct mlx5_ksm) : sizeof(struct mlx5_mtt); struct mlx5_ib_dev *dev = mr_to_mdev(mr); @@ -720,7 +719,7 @@ _mlx5r_umr_update_mr_pas(struct mlx5_ib_mr *mr, unsigned int flags, bool dd, orig_sg_length = sg.length; - _mlx5r_umr_init_wqe(mr, &wqe, &sg, flags, mr->page_shift, dd); + _mlx5r_umr_init_wqe(mr, &wqe, &sg, flags, pdn, mr->page_shift); /* Set initial translation offset to start_block */ offset = (u64)start_block * ent_size; @@ -811,7 +810,8 @@ int mlx5r_umr_update_data_direct_ksm_pas_range(struct mlx5_ib_mr *mr, !(flags & MLX5_IB_UPD_XLT_KEEP_PGSZ))) return -EINVAL; - return _mlx5r_umr_update_mr_pas(mr, flags, true, start_block, nblocks); + return _mlx5r_umr_update_mr_pas(mr, flags, mr_to_mdev(mr)->ddr.pdn, + true, start_block, nblocks); } int mlx5r_umr_update_data_direct_ksm_pas(struct mlx5_ib_mr *mr, @@ -821,12 +821,13 @@ int mlx5r_umr_update_data_direct_ksm_pas(struct mlx5_ib_mr *mr, } int mlx5r_umr_update_mr_pas_range(struct mlx5_ib_mr *mr, unsigned int flags, - size_t start_block, size_t nblocks) + u32 pdn, size_t start_block, size_t nblocks) { if (WARN_ON(mr->umem->is_odp)) return -EINVAL; - return _mlx5r_umr_update_mr_pas(mr, flags, false, start_block, nblocks); + return _mlx5r_umr_update_mr_pas(mr, flags, pdn, false, start_block, + nblocks); } /* @@ -834,9 +835,9 @@ int mlx5r_umr_update_mr_pas_range(struct mlx5_ib_mr *mr, unsigned int flags, * Dmabuf MR is handled in a similar way, except that the MLX5_IB_UPD_XLT_ZAP * flag may be used. */ -int mlx5r_umr_update_mr_pas(struct mlx5_ib_mr *mr, unsigned int flags) +int mlx5r_umr_update_mr_pas(struct mlx5_ib_mr *mr, unsigned int flags, u32 pdn) { - return mlx5r_umr_update_mr_pas_range(mr, flags, 0, 0); + return mlx5r_umr_update_mr_pas_range(mr, flags, pdn, 0, 0); } static bool umr_can_use_indirect_mkey(struct mlx5_ib_dev *dev) @@ -861,6 +862,7 @@ int mlx5r_umr_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages, size_t orig_sg_length; size_t pages_iter; struct ib_sge sg; + u32 pdn = mlx5_mr_pdn(mr); int err = 0; void *xlt; @@ -895,7 +897,8 @@ int mlx5r_umr_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages, } mlx5r_umr_set_update_xlt_ctrl_seg(&wqe.ctrl_seg, flags, &sg); - mlx5r_umr_set_update_xlt_mkey_seg(dev, &wqe.mkey_seg, mr, page_shift); + mlx5r_umr_set_update_xlt_mkey_seg(dev, &wqe.mkey_seg, mr, pdn, + page_shift); mlx5r_umr_set_update_xlt_data_seg(&wqe.data_seg, &sg); for (pages_mapped = 0; @@ -962,17 +965,18 @@ int mlx5r_umr_update_mr_page_shift(struct mlx5_ib_mr *mr, return err; } -static inline int -_mlx5r_dmabuf_umr_update_pas(struct mlx5_ib_mr *mr, unsigned int flags, - size_t start_block, size_t nblocks, bool dd) +static inline int _mlx5r_dmabuf_umr_update_pas(struct mlx5_ib_mr *mr, + unsigned int flags, u32 pdn, + size_t start_block, + size_t nblocks, bool dd) { if (dd) return mlx5r_umr_update_data_direct_ksm_pas_range(mr, flags, start_block, nblocks); else - return mlx5r_umr_update_mr_pas_range(mr, flags, start_block, - nblocks); + return mlx5r_umr_update_mr_pas_range(mr, flags, pdn, + start_block, nblocks); } /** @@ -986,11 +990,9 @@ _mlx5r_dmabuf_umr_update_pas(struct mlx5_ib_mr *mr, unsigned int flags, * Return: On success, returns the number of entries that were zapped. * On error, returns a negative error code. */ -static int _mlx5r_umr_zap_mkey(struct mlx5_ib_mr *mr, - unsigned int flags, - unsigned int page_shift, - size_t *nblocks, - bool dd) +static int _mlx5r_umr_zap_mkey(struct mlx5_ib_mr *mr, unsigned int flags, + unsigned int page_shift, size_t *nblocks, + u32 pdn, bool dd) { unsigned int old_page_shift = mr->page_shift; struct mlx5_ib_dev *dev = mr_to_mdev(mr); @@ -1030,7 +1032,7 @@ static int _mlx5r_umr_zap_mkey(struct mlx5_ib_mr *mr, */ if (*nblocks) mr->page_shift = max_page_shift; - err = _mlx5r_dmabuf_umr_update_pas(mr, flags, 0, *nblocks, dd); + err = _mlx5r_dmabuf_umr_update_pas(mr, flags, pdn, 0, *nblocks, dd); if (err) { mr->page_shift = old_page_shift; return err; @@ -1055,6 +1057,7 @@ static int _mlx5r_umr_zap_mkey(struct mlx5_ib_mr *mr, * entries accordingly * @mr: The memory region to update * @xlt_flags: Translation table update flags + * @pdn: Protection domain number * @page_shift: The new (optimized) page shift to use * * This function updates the page size and mkey translation entries for a DMABUF @@ -1074,7 +1077,7 @@ static int _mlx5r_umr_zap_mkey(struct mlx5_ib_mr *mr, * * Returns 0 on success or a negative error code on failure. */ -int mlx5r_umr_dmabuf_update_pgsz(struct mlx5_ib_mr *mr, u32 xlt_flags, +int mlx5r_umr_dmabuf_update_pgsz(struct mlx5_ib_mr *mr, u32 xlt_flags, u32 pdn, unsigned int page_shift) { unsigned int old_page_shift = mr->page_shift; @@ -1083,7 +1086,7 @@ int mlx5r_umr_dmabuf_update_pgsz(struct mlx5_ib_mr *mr, u32 xlt_flags, int err; err = _mlx5r_umr_zap_mkey(mr, xlt_flags, page_shift, &zapped_blocks, - mr->data_direct); + pdn, mr->data_direct); if (err) return err; @@ -1096,10 +1099,8 @@ int mlx5r_umr_dmabuf_update_pgsz(struct mlx5_ib_mr *mr, u32 xlt_flags, * the page size in the mkey yet. */ err = _mlx5r_dmabuf_umr_update_pas( - mr, - xlt_flags | MLX5_IB_UPD_XLT_KEEP_PGSZ, - zapped_blocks, - total_blocks - zapped_blocks, + mr, xlt_flags | MLX5_IB_UPD_XLT_KEEP_PGSZ, pdn, + zapped_blocks, total_blocks - zapped_blocks, mr->data_direct); if (err) goto err; @@ -1108,7 +1109,7 @@ int mlx5r_umr_dmabuf_update_pgsz(struct mlx5_ib_mr *mr, u32 xlt_flags, err = mlx5r_umr_update_mr_page_shift(mr, mr->page_shift); if (err) goto err; - err = _mlx5r_dmabuf_umr_update_pas(mr, xlt_flags, 0, zapped_blocks, + err = _mlx5r_dmabuf_umr_update_pas(mr, xlt_flags, pdn, 0, zapped_blocks, mr->data_direct); if (err) goto err; diff --git a/drivers/infiniband/hw/mlx5/umr.h b/drivers/infiniband/hw/mlx5/umr.h index 2797dda2b2bef0..1db1662378e773 100644 --- a/drivers/infiniband/hw/mlx5/umr.h +++ b/drivers/infiniband/hw/mlx5/umr.h @@ -100,13 +100,13 @@ int mlx5r_umr_update_data_direct_ksm_pas_range(struct mlx5_ib_mr *mr, size_t nblocks); int mlx5r_umr_update_data_direct_ksm_pas(struct mlx5_ib_mr *mr, unsigned int flags); int mlx5r_umr_update_mr_pas_range(struct mlx5_ib_mr *mr, unsigned int flags, - size_t start_block, size_t nblocks); -int mlx5r_umr_update_mr_pas(struct mlx5_ib_mr *mr, unsigned int flags); + u32 pdn, size_t start_block, size_t nblocks); +int mlx5r_umr_update_mr_pas(struct mlx5_ib_mr *mr, unsigned int flags, u32 pdn); int mlx5r_umr_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages, int page_shift, int flags); int mlx5r_umr_update_mr_page_shift(struct mlx5_ib_mr *mr, unsigned int page_shift); -int mlx5r_umr_dmabuf_update_pgsz(struct mlx5_ib_mr *mr, u32 xlt_flags, +int mlx5r_umr_dmabuf_update_pgsz(struct mlx5_ib_mr *mr, u32 xlt_flags, u32 pdn, unsigned int page_shift); #endif /* _MLX5_IB_UMR_H */ -- 2.53.0