From: Jiri Pirko <jiri@resnulli.us>
To: linux-rdma@vger.kernel.org
Cc: jgg@ziepe.ca, leon@kernel.org, mrgolin@amazon.com,
gal.pressman@linux.dev, sleybo@amazon.com, parav@nvidia.com,
mbloch@nvidia.com, yanjun.zhu@linux.dev,
marco.crivellari@suse.com, roman.gushchin@linux.dev,
phaddad@nvidia.com, lirongqing@baidu.com, ynachum@amazon.com,
huangjunxian6@hisilicon.com, kalesh-anakkur.purayil@broadcom.com,
ohartoov@nvidia.com, michaelgur@nvidia.com, shayd@nvidia.com,
edwards@nvidia.com, sriharsha.basavapatna@broadcom.com,
andrew.gospodarek@broadcom.com, selvin.xavier@broadcom.com
Subject: [PATCH rdma-next v3 15/17] RDMA/mlx5: Use UMEM attribute for CQ doorbell record
Date: Mon, 4 May 2026 15:57:29 +0200 [thread overview]
Message-ID: <20260504135731.2345383-16-jiri@resnulli.us> (raw)
In-Reply-To: <20260504135731.2345383-1-jiri@resnulli.us>
From: Jiri Pirko <jiri@nvidia.com>
Add an optional mlx5 driver-namespace UMEM attribute on CQ
create so userspace can supply the doorbell record umem
explicitly. Resolve it inside mlx5_ib_db_map_user() and use it
as a private DBR page when present; otherwise take the existing
UHW share-or-pin path that preserves per-page DBR sharing
across CQ/QP/SRQ in the same process.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
---
v2->v3:
- moved CQ DBR attr to mlx5 driver namespace
- changed to use ib_umem_get_attr() to get umem
- added page-crossing check
---
drivers/infiniband/hw/mlx5/cq.c | 8 +++-
drivers/infiniband/hw/mlx5/doorbell.c | 51 ++++++++++++++++++++----
drivers/infiniband/hw/mlx5/mlx5_ib.h | 5 ++-
drivers/infiniband/hw/mlx5/qp.c | 4 +-
drivers/infiniband/hw/mlx5/srq.c | 2 +-
include/uapi/rdma/mlx5_user_ioctl_cmds.h | 1 +
6 files changed, 57 insertions(+), 14 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index fb6172a9be57..b69ab49b1b3e 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -760,7 +760,9 @@ static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata,
goto err_umem;
}
- err = mlx5_ib_db_map_user(context, ucmd.db_addr, &cq->db);
+ err = mlx5_ib_db_map_user(context, udata,
+ MLX5_IB_ATTR_CREATE_CQ_DBR_BUF_UMEM,
+ ucmd.db_addr, &cq->db);
if (err)
goto err_umem;
@@ -1519,7 +1521,9 @@ ADD_UVERBS_ATTRIBUTES_SIMPLE(
UVERBS_ATTR_PTR_IN(
MLX5_IB_ATTR_CREATE_CQ_UAR_INDEX,
UVERBS_ATTR_TYPE(u32),
- UA_OPTIONAL));
+ UA_OPTIONAL),
+ UVERBS_ATTR_UMEM(MLX5_IB_ATTR_CREATE_CQ_DBR_BUF_UMEM,
+ UA_OPTIONAL));
const struct uapi_definition mlx5_ib_create_cq_defs[] = {
UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_CQ, &mlx5_ib_cq_create),
diff --git a/drivers/infiniband/hw/mlx5/doorbell.c b/drivers/infiniband/hw/mlx5/doorbell.c
index 020c70328663..c9f9b9179e88 100644
--- a/drivers/infiniband/hw/mlx5/doorbell.c
+++ b/drivers/infiniband/hw/mlx5/doorbell.c
@@ -45,20 +45,56 @@ struct mlx5_ib_user_db_page {
struct mm_struct *mm;
};
-int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context, unsigned long virt,
- struct mlx5_db *db)
+int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context,
+ struct ib_udata *udata, u16 attr_id,
+ unsigned long virt, struct mlx5_db *db)
{
- struct mlx5_ib_user_db_page *page;
+ struct mlx5_ib_user_db_page *page = NULL;
+ unsigned long dma_offset;
int err = 0;
+ if (udata) {
+ struct ib_umem *umem;
+
+ umem = ib_umem_get_attr(context->ibucontext.device, udata,
+ attr_id, sizeof(__be32) * 2, 0);
+ if (IS_ERR(umem))
+ return PTR_ERR(umem);
+ if (umem) {
+ /*
+ * The 8-byte DBR is programmed to the device as one
+ * DMA address, so it must stay within a single page.
+ * An 8-byte range that crosses a page boundary may
+ * be split across two non-contiguous DMA mappings.
+ */
+ if (ib_umem_offset(umem) >
+ PAGE_SIZE - sizeof(__be32) * 2) {
+ ib_umem_release(umem);
+ return -EINVAL;
+ }
+ page = kzalloc_obj(*page);
+ if (!page) {
+ ib_umem_release(umem);
+ return -ENOMEM;
+ }
+ page->umem = umem;
+ dma_offset = ib_umem_offset(umem);
+ }
+ }
+
mutex_lock(&context->db_page_mutex);
+ if (page)
+ goto add_page;
+
+ dma_offset = virt & ~PAGE_MASK;
+
list_for_each_entry(page, &context->db_page_list, list)
if ((current->mm == page->mm) &&
(page->user_virt == (virt & PAGE_MASK)))
goto found;
- page = kmalloc_obj(*page);
+ page = kzalloc_obj(*page);
if (!page) {
err = -ENOMEM;
goto out;
@@ -76,11 +112,11 @@ int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context, unsigned long virt,
mmgrab(current->mm);
page->mm = current->mm;
+add_page:
list_add(&page->list, &context->db_page_list);
found:
- db->dma = sg_dma_address(page->umem->sgt_append.sgt.sgl) +
- (virt & ~PAGE_MASK);
+ db->dma = sg_dma_address(page->umem->sgt_append.sgt.sgl) + dma_offset;
db->u.user_page = page;
++page->refcnt;
@@ -96,7 +132,8 @@ void mlx5_ib_db_unmap_user(struct mlx5_ib_ucontext *context, struct mlx5_db *db)
if (!--db->u.user_page->refcnt) {
list_del(&db->u.user_page->list);
- mmdrop(db->u.user_page->mm);
+ if (db->u.user_page->mm)
+ mmdrop(db->u.user_page->mm);
ib_umem_release(db->u.user_page->umem);
kfree(db->u.user_page);
}
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index e156dc4d7529..45bc8928523a 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -1259,8 +1259,9 @@ to_mmmap(struct rdma_user_mmap_entry *rdma_entry)
int mlx5_ib_dev_res_cq_init(struct mlx5_ib_dev *dev);
int mlx5_ib_dev_res_srq_init(struct mlx5_ib_dev *dev);
-int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context, unsigned long virt,
- struct mlx5_db *db);
+int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context,
+ struct ib_udata *udata, u16 attr_id,
+ unsigned long virt, struct mlx5_db *db);
void mlx5_ib_db_unmap_user(struct mlx5_ib_ucontext *context, struct mlx5_db *db);
void __mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 qpn, struct mlx5_ib_srq *srq);
void mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 qpn, struct mlx5_ib_srq *srq);
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 1b764a573dd7..997ea9bcfc55 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -918,7 +918,7 @@ static int create_user_rq(struct mlx5_ib_dev *dev, struct ib_pd *pd,
ib_umem_num_pages(rwq->umem), page_size, rwq->rq_num_pas,
offset);
- err = mlx5_ib_db_map_user(ucontext, ucmd->db_addr, &rwq->db);
+ err = mlx5_ib_db_map_user(ucontext, NULL, 0, ucmd->db_addr, &rwq->db);
if (err) {
mlx5_ib_dbg(dev, "map failed\n");
goto err_umem;
@@ -1056,7 +1056,7 @@ static int _create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
resp->bfreg_index = MLX5_IB_INVALID_BFREG;
qp->bfregn = bfregn;
- err = mlx5_ib_db_map_user(context, ucmd->db_addr, &qp->db);
+ err = mlx5_ib_db_map_user(context, NULL, 0, ucmd->db_addr, &qp->db);
if (err) {
mlx5_ib_dbg(dev, "map failed\n");
goto err_free;
diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c
index bc22036d7e80..88db0143bc3f 100644
--- a/drivers/infiniband/hw/mlx5/srq.c
+++ b/drivers/infiniband/hw/mlx5/srq.c
@@ -74,7 +74,7 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq,
}
in->umem = srq->umem;
- err = mlx5_ib_db_map_user(ucontext, ucmd.db_addr, &srq->db);
+ err = mlx5_ib_db_map_user(ucontext, NULL, 0, ucmd.db_addr, &srq->db);
if (err) {
mlx5_ib_dbg(dev, "map doorbell failed\n");
goto err_umem;
diff --git a/include/uapi/rdma/mlx5_user_ioctl_cmds.h b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
index 01a2a050e468..b63e75034cda 100644
--- a/include/uapi/rdma/mlx5_user_ioctl_cmds.h
+++ b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
@@ -274,6 +274,7 @@ enum mlx5_ib_device_query_context_attrs {
enum mlx5_ib_create_cq_attrs {
MLX5_IB_ATTR_CREATE_CQ_UAR_INDEX = UVERBS_ID_DRIVER_NS_WITH_UHW,
+ MLX5_IB_ATTR_CREATE_CQ_DBR_BUF_UMEM,
};
enum mlx5_ib_reg_dmabuf_mr_attrs {
--
2.53.0
next prev parent reply other threads:[~2026-05-04 13:58 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-05-04 13:57 [PATCH rdma-next v3 00/17] RDMA: Introduce generic buffer descriptor infrastructure for umem Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 01/17] RDMA/umem: Rename ib_umem_get() to ib_umem_get_va() Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 02/17] RDMA/umem: Split ib_umem_get_va() into a thin wrapper around __ib_umem_get_va() Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 03/17] RDMA/core: Introduce generic buffer descriptor infrastructure for umem Jiri Pirko
2026-05-06 13:37 ` Jason Gunthorpe
2026-05-06 14:14 ` Jiri Pirko
2026-05-12 18:12 ` Jason Gunthorpe
2026-05-04 13:57 ` [PATCH rdma-next v3 04/17] RDMA/umem: Route ib_umem_get_va() through ib_umem_get() Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 05/17] RDMA/uverbs: Inline _uverbs_get_const_{signed,unsigned}() Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 06/17] RDMA/uverbs: Push out CQ buffer umem processing into a helper Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 07/17] RDMA/uverbs: Add CQ buffer UMEM attribute and driver helpers Jiri Pirko
2026-05-06 13:46 ` Jason Gunthorpe
2026-05-06 14:27 ` Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 08/17] RDMA/efa: Use ib_umem_get_cq_buf() for user CQ buffer Jiri Pirko
2026-05-06 13:51 ` Jason Gunthorpe
2026-05-06 14:32 ` Jiri Pirko
2026-05-12 18:18 ` Jason Gunthorpe
2026-05-04 13:57 ` [PATCH rdma-next v3 09/17] RDMA/mlx5: Use ib_umem_get_cq_buf_or_va() " Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 10/17] RDMA/bnxt_re: " Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 11/17] RDMA/mlx4: Use ib_umem_get_cq_buf() " Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 12/17] RDMA/uverbs: Remove legacy umem field from struct ib_cq Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 13/17] RDMA/uverbs: Use UMEM attributes for QP creation Jiri Pirko
2026-05-06 13:43 ` Jason Gunthorpe
2026-05-06 14:17 ` Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 14/17] RDMA/mlx5: Use UMEM attributes for QP buffers in create_qp Jiri Pirko
2026-05-04 13:57 ` Jiri Pirko [this message]
2026-05-04 13:57 ` [PATCH rdma-next v3 16/17] RDMA/mlx5: Use UMEM attribute for QP doorbell record Jiri Pirko
2026-05-04 13:57 ` [PATCH rdma-next v3 17/17] RDMA/uverbs: Track attr consumption and warn on unused attrs Jiri Pirko
2026-05-06 13:56 ` Jason Gunthorpe
2026-05-06 14:22 ` Jiri Pirko
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260504135731.2345383-16-jiri@resnulli.us \
--to=jiri@resnulli.us \
--cc=andrew.gospodarek@broadcom.com \
--cc=edwards@nvidia.com \
--cc=gal.pressman@linux.dev \
--cc=huangjunxian6@hisilicon.com \
--cc=jgg@ziepe.ca \
--cc=kalesh-anakkur.purayil@broadcom.com \
--cc=leon@kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=lirongqing@baidu.com \
--cc=marco.crivellari@suse.com \
--cc=mbloch@nvidia.com \
--cc=michaelgur@nvidia.com \
--cc=mrgolin@amazon.com \
--cc=ohartoov@nvidia.com \
--cc=parav@nvidia.com \
--cc=phaddad@nvidia.com \
--cc=roman.gushchin@linux.dev \
--cc=selvin.xavier@broadcom.com \
--cc=shayd@nvidia.com \
--cc=sleybo@amazon.com \
--cc=sriharsha.basavapatna@broadcom.com \
--cc=yanjun.zhu@linux.dev \
--cc=ynachum@amazon.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox