From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: stable@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
patches@lists.linux.dev, Namjae Jeon <linkinjeon@kernel.org>,
Steve French <smfrench@gmail.com>, Tom Talpey <tom@talpey.com>,
linux-cifs@vger.kernel.org, samba-technical@lists.samba.org,
linux-rdma@vger.kernel.org, Stefan Metzmacher <metze@samba.org>,
Steve French <stfrench@microsoft.com>,
Sasha Levin <sashal@kernel.org>
Subject: [PATCH 6.6 008/196] smb: server: fix IRD/ORD negotiation with the client
Date: Mon, 13 Oct 2025 16:43:19 +0200 [thread overview]
Message-ID: <20251013144315.496078826@linuxfoundation.org> (raw)
In-Reply-To: <20251013144315.184275491@linuxfoundation.org>
6.6-stable review patch. If anyone has any objections, please let me know.
------------------
From: Stefan Metzmacher <metze@samba.org>
[ Upstream commit fad988a2158d743da7971884b93482a73735b25e ]
Already do real negotiation in smb_direct_handle_connect_request()
where we see the requested initiator_depth and responder_resources
from the client.
We should detect legacy iwarp clients using MPA v1
with the custom IRD/ORD negotiation.
We need to send the custom IRD/ORD in big endian,
but we need to try to let clients with broken requests
using little endian (older cifs.ko) to work.
Note the reason why this uses u8 for
initiator_depth and responder_resources is
that the rdma layer also uses it.
Acked-by: Namjae Jeon <linkinjeon@kernel.org>
Cc: Steve French <smfrench@gmail.com>
Cc: Tom Talpey <tom@talpey.com>
Cc: linux-cifs@vger.kernel.org
Cc: samba-technical@lists.samba.org
Cc: linux-rdma@vger.kernel.org
Fixes: 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers")
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
fs/smb/server/transport_rdma.c | 99 +++++++++++++++++++++++++++++-----
1 file changed, 85 insertions(+), 14 deletions(-)
diff --git a/fs/smb/server/transport_rdma.c b/fs/smb/server/transport_rdma.c
index 504e2a1cf33b8..31c1ac256e1be 100644
--- a/fs/smb/server/transport_rdma.c
+++ b/fs/smb/server/transport_rdma.c
@@ -152,6 +152,10 @@ struct smb_direct_transport {
struct work_struct disconnect_work;
bool negotiation_requested;
+
+ bool legacy_iwarp;
+ u8 initiator_depth;
+ u8 responder_resources;
};
#define KSMBD_TRANS(t) ((struct ksmbd_transport *)&((t)->transport))
@@ -346,6 +350,9 @@ static struct smb_direct_transport *alloc_transport(struct rdma_cm_id *cm_id)
t->cm_id = cm_id;
cm_id->context = t;
+ t->initiator_depth = SMB_DIRECT_CM_INITIATOR_DEPTH;
+ t->responder_resources = 1;
+
t->status = SMB_DIRECT_CS_NEW;
init_waitqueue_head(&t->wait_status);
@@ -1623,21 +1630,21 @@ static int smb_direct_send_negotiate_response(struct smb_direct_transport *t,
static int smb_direct_accept_client(struct smb_direct_transport *t)
{
struct rdma_conn_param conn_param;
- struct ib_port_immutable port_immutable;
- u32 ird_ord_hdr[2];
+ __be32 ird_ord_hdr[2];
int ret;
+ /*
+ * smb_direct_handle_connect_request()
+ * already negotiated t->initiator_depth
+ * and t->responder_resources
+ */
memset(&conn_param, 0, sizeof(conn_param));
- conn_param.initiator_depth = min_t(u8, t->cm_id->device->attrs.max_qp_rd_atom,
- SMB_DIRECT_CM_INITIATOR_DEPTH);
- conn_param.responder_resources = 0;
-
- t->cm_id->device->ops.get_port_immutable(t->cm_id->device,
- t->cm_id->port_num,
- &port_immutable);
- if (port_immutable.core_cap_flags & RDMA_CORE_PORT_IWARP) {
- ird_ord_hdr[0] = conn_param.responder_resources;
- ird_ord_hdr[1] = 1;
+ conn_param.initiator_depth = t->initiator_depth;
+ conn_param.responder_resources = t->responder_resources;
+
+ if (t->legacy_iwarp) {
+ ird_ord_hdr[0] = cpu_to_be32(conn_param.responder_resources);
+ ird_ord_hdr[1] = cpu_to_be32(conn_param.initiator_depth);
conn_param.private_data = ird_ord_hdr;
conn_param.private_data_len = sizeof(ird_ord_hdr);
} else {
@@ -2023,10 +2030,13 @@ static bool rdma_frwr_is_supported(struct ib_device_attr *attrs)
return true;
}
-static int smb_direct_handle_connect_request(struct rdma_cm_id *new_cm_id)
+static int smb_direct_handle_connect_request(struct rdma_cm_id *new_cm_id,
+ struct rdma_cm_event *event)
{
struct smb_direct_transport *t;
struct task_struct *handler;
+ u8 peer_initiator_depth;
+ u8 peer_responder_resources;
int ret;
if (!rdma_frwr_is_supported(&new_cm_id->device->attrs)) {
@@ -2040,6 +2050,67 @@ static int smb_direct_handle_connect_request(struct rdma_cm_id *new_cm_id)
if (!t)
return -ENOMEM;
+ peer_initiator_depth = event->param.conn.initiator_depth;
+ peer_responder_resources = event->param.conn.responder_resources;
+ if (rdma_protocol_iwarp(new_cm_id->device, new_cm_id->port_num) &&
+ event->param.conn.private_data_len == 8) {
+ /*
+ * Legacy clients with only iWarp MPA v1 support
+ * need a private blob in order to negotiate
+ * the IRD/ORD values.
+ */
+ const __be32 *ird_ord_hdr = event->param.conn.private_data;
+ u32 ird32 = be32_to_cpu(ird_ord_hdr[0]);
+ u32 ord32 = be32_to_cpu(ird_ord_hdr[1]);
+
+ /*
+ * cifs.ko sends the legacy IRD/ORD negotiation
+ * event if iWarp MPA v2 was used.
+ *
+ * Here we check that the values match and only
+ * mark the client as legacy if they don't match.
+ */
+ if ((u32)event->param.conn.initiator_depth != ird32 ||
+ (u32)event->param.conn.responder_resources != ord32) {
+ /*
+ * There are broken clients (old cifs.ko)
+ * using little endian and also
+ * struct rdma_conn_param only uses u8
+ * for initiator_depth and responder_resources,
+ * so we truncate the value to U8_MAX.
+ *
+ * smb_direct_accept_client() will then
+ * do the real negotiation in order to
+ * select the minimum between client and
+ * server.
+ */
+ ird32 = min_t(u32, ird32, U8_MAX);
+ ord32 = min_t(u32, ord32, U8_MAX);
+
+ t->legacy_iwarp = true;
+ peer_initiator_depth = (u8)ird32;
+ peer_responder_resources = (u8)ord32;
+ }
+ }
+
+ /*
+ * First set what the we as server are able to support
+ */
+ t->initiator_depth = min_t(u8, t->initiator_depth,
+ new_cm_id->device->attrs.max_qp_rd_atom);
+
+ /*
+ * negotiate the value by using the minimum
+ * between client and server if the client provided
+ * non 0 values.
+ */
+ if (peer_initiator_depth != 0)
+ t->initiator_depth = min_t(u8, t->initiator_depth,
+ peer_initiator_depth);
+ if (peer_responder_resources != 0)
+ t->responder_resources = min_t(u8, t->responder_resources,
+ peer_responder_resources);
+
ret = smb_direct_connect(t);
if (ret)
goto out_err;
@@ -2064,7 +2135,7 @@ static int smb_direct_listen_handler(struct rdma_cm_id *cm_id,
{
switch (event->event) {
case RDMA_CM_EVENT_CONNECT_REQUEST: {
- int ret = smb_direct_handle_connect_request(cm_id);
+ int ret = smb_direct_handle_connect_request(cm_id, event);
if (ret) {
pr_err("Can't create transport: %d\n", ret);
--
2.51.0
parent reply other threads:[~2025-10-13 14:58 UTC|newest]
Thread overview: expand[flat|nested] mbox.gz Atom feed
[parent not found: <20251013144315.184275491@linuxfoundation.org>]
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=20251013144315.496078826@linuxfoundation.org \
--to=gregkh@linuxfoundation.org \
--cc=linkinjeon@kernel.org \
--cc=linux-cifs@vger.kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=metze@samba.org \
--cc=patches@lists.linux.dev \
--cc=samba-technical@lists.samba.org \
--cc=sashal@kernel.org \
--cc=smfrench@gmail.com \
--cc=stable@vger.kernel.org \
--cc=stfrench@microsoft.com \
--cc=tom@talpey.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;
as well as URLs for NNTP newsgroup(s).