* [PATCH 0/5] netfilter: nf_flow_table_path: L2 bridge offload
From: Daniel Pawlik @ 2026-06-29 12:32 UTC (permalink / raw)
To: netfilter-devel, netdev
Cc: pablo, fw, phil, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, razor, idosch, matthias.bgg,
angelogioacchino.delregno, bridge, coreteam, linux-mediatek,
linux-arm-kernel, rchen14b, lorenzo, Daniel Pawlik
This series adds L2 bridge offload support to nft_flow_offload, allowing
bridged IPv4/IPv6 flows to be accelerated by the flowtable fast path
without requiring L3 routing.
Background
----------
Hardware flow offload engines (e.g. MediaTek PPE) can accelerate bridged
traffic but require that nft_flow_offload detect and handle bridged flows
differently from routed ones: no routing table lookup, MAC addresses from
the Ethernet header, and VLAN context pre-populated from the bridge port.
Patches
-------
1/5 net: export __dev_fill_forward_path
Refactors dev_fill_forward_path() to expose __dev_fill_forward_path()
which accepts a caller-supplied net_device_path_ctx, needed to
pre-populate VLAN state before the forward path walk.
2/5 net: bridge: add flow offload helpers
Adds br_fdb_has_forwarding_entry_rcu(), br_vlan_get_offload_info_rcu()
and br_vlan_is_enabled_rcu() to expose bridge state to nft_flow_offload
without requiring inclusion of net/bridge/br_private.h.
3/5 netfilter: nf_flow_table_path: add L2 bridge offload
Core of the series. Adds nft_flow_offload_is_bridging() detection,
nft_flow_route_bridging() which avoids nf_route() (fails for
bridged-only subnets), MAC/VLAN pre-population for bridged flows,
and a dst leak fix (allocation references in dsts[] were never
released after nft_default_forward_path() transferred ownership).
nft_flow_route() becomes a thin dispatcher.
4/5 netfilter: nf_flow_table_path: handle DEV_PATH_MTK_WDMA in path info
Fixes zero-source-MAC in PPE entries when a bridged flow traverses
MT7996/MT7915 WiFi WDMA hardware.
5/5 netfilter: nf_flow_table_path: add VLAN passthrough support
Records VLAN encap info for passthrough-mode bridge ports so hardware
offload entries include the correct VLAN tag.
Rebase note
-----------
Originally developed against OpenWrt pending-6.18 patches by Ryan Chen
<rchen14b@gmail.com> and Bo-Cun Chen <bc-bocun.chen@mediatek.com>.
Rebased to current upstream: path discovery infrastructure moved to
nf_flow_table_path.c in commit 93d7a7ed0734 ("netfilter: flowtable: move
path discovery infrastructure to its own file"), so all netfilter changes
now land in that file rather than nft_flow_offload.c.
How to enable bridge offload
-----------------------------
1. Load kmod-br-netfilter so that bridged IP traffic traverses the
netfilter forward chain.
2. Enable netfilter hooks on the bridge:
echo 1 > /sys/class/net/<br>/bridge/nf_call_iptables
echo 1 > /sys/class/net/<br>/bridge/nf_call_ip6tables
3. Register bridge member interfaces in the nft flowtable:
table inet filter {
flowtable f {
hook ingress priority filter
devices = { eth0, wlan0 }
}
chain forward {
type filter hook forward priority filter
meta l4proto { tcp, udp } flow add @f
}
}
Daniel Pawlik (1):
net: bridge: add flow offload helpers
Ryan Chen (4):
net: export __dev_fill_forward_path
netfilter: nf_flow_table_path: add L2 bridge offload
netfilter: nf_flow_table_path: handle DEV_PATH_MTK_WDMA in path info
netfilter: nf_flow_table_path: add VLAN passthrough support
include/linux/if_bridge.h | 23 ++++
include/linux/netdevice.h | 2 +
net/bridge/br_fdb.c | 32 +++++
net/bridge/br_vlan.c | 45 +++++++
net/core/dev.c | 32 +++--
net/netfilter/nf_flow_table_path.c | 201 +++++++++++++++++++++++++++--
6 files changed, 312 insertions(+), 23 deletions(-)
--
2.54.0
^ permalink raw reply
* [PATCH 1/5] net: export __dev_fill_forward_path
From: Daniel Pawlik @ 2026-06-29 12:32 UTC (permalink / raw)
To: netfilter-devel, netdev
Cc: pablo, fw, phil, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, razor, idosch, matthias.bgg,
angelogioacchino.delregno, bridge, coreteam, linux-mediatek,
linux-arm-kernel, rchen14b, lorenzo, Daniel Pawlik
In-Reply-To: <20260629123253.1912621-1-pawlik.dan@gmail.com>
From: Ryan Chen <rchen14b@gmail.com>
Export __dev_fill_forward_path() which accepts a caller-supplied
net_device_path_ctx, allowing callers to pre-populate context (e.g.
VLAN state) before the forward path walk. The existing
dev_fill_forward_path() is refactored to call it.
This is a prerequisite for nft_flow_offload bridge offload, which needs
to supply a pre-populated ctx for bridge port devices.
Signed-off-by: Ryan Chen <rchen14b@gmail.com>
Signed-off-by: Daniel Pawlik <pawlik.dan@gmail.com>
---
include/linux/netdevice.h | 2 ++
net/core/dev.c | 32 ++++++++++++++++++++------------
2 files changed, 22 insertions(+), 12 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 9981d637f8b5..c1d0b897de95 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3422,6 +3422,8 @@ int dev_get_iflink(const struct net_device *dev);
int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
struct net_device_path_stack *stack);
+int __dev_fill_forward_path(struct net_device_path_ctx *ctx, const u8 *daddr,
+ struct net_device_path_stack *stack);
struct net_device *dev_get_by_name(struct net *net, const char *name);
struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
struct net_device *__dev_get_by_name(struct net *net, const char *name);
diff --git a/net/core/dev.c b/net/core/dev.c
index 4b3d5cfdf6e0..62f1d0b64c76 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -750,44 +750,52 @@ static struct net_device_path *dev_fwd_path(struct net_device_path_stack *stack)
return &stack->path[k];
}
-int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
- struct net_device_path_stack *stack)
+int __dev_fill_forward_path(struct net_device_path_ctx *ctx, const u8 *daddr,
+ struct net_device_path_stack *stack)
{
const struct net_device *last_dev;
- struct net_device_path_ctx ctx = {
- .dev = dev,
- };
struct net_device_path *path;
int ret = 0;
- memcpy(ctx.daddr, daddr, sizeof(ctx.daddr));
+ memcpy(ctx->daddr, daddr, sizeof(ctx->daddr));
stack->num_paths = 0;
- while (ctx.dev && ctx.dev->netdev_ops->ndo_fill_forward_path) {
- last_dev = ctx.dev;
+ while (ctx->dev && ctx->dev->netdev_ops->ndo_fill_forward_path) {
+ last_dev = ctx->dev;
path = dev_fwd_path(stack);
if (!path)
return -1;
memset(path, 0, sizeof(struct net_device_path));
- ret = ctx.dev->netdev_ops->ndo_fill_forward_path(&ctx, path);
+ ret = ctx->dev->netdev_ops->ndo_fill_forward_path(ctx, path);
if (ret < 0)
return -1;
- if (WARN_ON_ONCE(last_dev == ctx.dev))
+ if (WARN_ON_ONCE(last_dev == ctx->dev))
return -1;
}
- if (!ctx.dev)
+ if (!ctx->dev)
return ret;
path = dev_fwd_path(stack);
if (!path)
return -1;
path->type = DEV_PATH_ETHERNET;
- path->dev = ctx.dev;
+ path->dev = ctx->dev;
return ret;
}
+EXPORT_SYMBOL_GPL(__dev_fill_forward_path);
+
+int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
+ struct net_device_path_stack *stack)
+{
+ struct net_device_path_ctx ctx = {
+ .dev = dev,
+ };
+
+ return __dev_fill_forward_path(&ctx, daddr, stack);
+}
EXPORT_SYMBOL_GPL(dev_fill_forward_path);
/* must be called under rcu_read_lock(), as we dont take a reference */
--
2.54.0
^ permalink raw reply related
* [PATCH 2/5] net: bridge: add flow offload helpers
From: Daniel Pawlik @ 2026-06-29 12:32 UTC (permalink / raw)
To: netfilter-devel, netdev
Cc: pablo, fw, phil, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, razor, idosch, matthias.bgg,
angelogioacchino.delregno, bridge, coreteam, linux-mediatek,
linux-arm-kernel, rchen14b, lorenzo, Daniel Pawlik
In-Reply-To: <20260629123253.1912621-1-pawlik.dan@gmail.com>
Add three helpers that expose the bridge state needed by nft_flow_offload
without requiring callers to include net/bridge/br_private.h. Each
performs a single br_port_get_rcu() lookup:
- br_fdb_has_forwarding_entry_rcu(): resolves the VLAN id for the packet
(skb tag or PVID when filtering is on, 0 otherwise) then checks whether
the bridge FDB contains a forwarding entry (dst != NULL, non-local) for
the resulting MAC/VLAN pair.
- br_vlan_get_offload_info_rcu(): when VLAN filtering is active, returns
the VLAN id (skb tag or PVID) and writes the bridge VLAN protocol to
*proto in a single port lookup. Returns 0 when filtering is off.
- br_vlan_is_enabled_rcu(): returns true when VLAN filtering is enabled
on the bridge a port device belongs to.
Based on MediaTek SDK patches by Bo-Cun Chen <bc-bocun.chen@mediatek.com>
and the OpenWrt bridge offload series by Ryan Chen <rchen14b@gmail.com>.
Signed-off-by: Daniel Pawlik <pawlik.dan@gmail.com>
---
include/linux/if_bridge.h | 23 ++++++++++++++++++++
net/bridge/br_fdb.c | 32 ++++++++++++++++++++++++++++
net/bridge/br_vlan.c | 45 +++++++++++++++++++++++++++++++++++++++
3 files changed, 100 insertions(+)
diff --git a/include/linux/if_bridge.h b/include/linux/if_bridge.h
index 75673b8bffcb..c1cae54749c5 100644
--- a/include/linux/if_bridge.h
+++ b/include/linux/if_bridge.h
@@ -148,6 +148,9 @@ int br_vlan_get_info(const struct net_device *dev, u16 vid,
struct bridge_vlan_info *p_vinfo);
int br_vlan_get_info_rcu(const struct net_device *dev, u16 vid,
struct bridge_vlan_info *p_vinfo);
+u16 br_vlan_get_offload_info_rcu(const struct net_device *dev,
+ const struct sk_buff *skb, __be16 *proto);
+bool br_vlan_is_enabled_rcu(const struct net_device *dev);
bool br_mst_enabled(const struct net_device *dev);
int br_mst_get_info(const struct net_device *dev, u16 msti, unsigned long *vids);
int br_mst_get_state(const struct net_device *dev, u16 msti, u8 *state);
@@ -184,6 +187,17 @@ static inline int br_vlan_get_info_rcu(const struct net_device *dev, u16 vid,
return -EINVAL;
}
+static inline u16 br_vlan_get_offload_info_rcu(const struct net_device *dev,
+ const struct sk_buff *skb,
+ __be16 *proto)
+{
+ return 0;
+}
+
+static inline bool br_vlan_is_enabled_rcu(const struct net_device *dev)
+{
+ return false;
+}
static inline bool br_mst_enabled(const struct net_device *dev)
{
return false;
@@ -209,6 +223,8 @@ void br_fdb_clear_offload(const struct net_device *dev, u16 vid);
bool br_port_flag_is_set(const struct net_device *dev, unsigned long flag);
u8 br_port_get_stp_state(const struct net_device *dev);
clock_t br_get_ageing_time(const struct net_device *br_dev);
+bool br_fdb_has_forwarding_entry_rcu(const struct net_device *dev,
+ const struct sk_buff *skb, const u8 *addr);
#else
static inline struct net_device *
br_fdb_find_port(const struct net_device *br_dev,
@@ -237,6 +253,13 @@ static inline clock_t br_get_ageing_time(const struct net_device *br_dev)
{
return 0;
}
+
+static inline bool br_fdb_has_forwarding_entry_rcu(const struct net_device *dev,
+ const struct sk_buff *skb,
+ const u8 *addr)
+{
+ return false;
+}
#endif
#endif
diff --git a/net/bridge/br_fdb.c b/net/bridge/br_fdb.c
index e4570bbed854..3161c2689f6a 100644
--- a/net/bridge/br_fdb.c
+++ b/net/bridge/br_fdb.c
@@ -267,6 +267,38 @@ struct net_bridge_fdb_entry *br_fdb_find_rcu(struct net_bridge *br,
return fdb_find_rcu(&br->fdb_hash_tbl, addr, vid);
}
+/**
+ * br_fdb_has_forwarding_entry_rcu - check if a MAC can be forwarded by the bridge
+ * @dev: bridge port network device
+ * @skb: packet buffer (used to determine VLAN id)
+ * @addr: destination MAC address
+ *
+ * Resolves the VLAN id for @skb on @dev (skb VLAN tag when present, PVID
+ * when VLAN filtering is enabled, 0 otherwise) then checks whether the bridge
+ * FDB contains a forwarding entry (dst != NULL, not a local/self entry) for
+ * @addr and that VLAN id. Single br_port_get_rcu() lookup.
+ * Must be called under RCU read lock.
+ */
+bool br_fdb_has_forwarding_entry_rcu(const struct net_device *dev,
+ const struct sk_buff *skb, const u8 *addr)
+{
+ struct net_bridge_port *port = br_port_get_rcu(dev);
+ struct net_bridge_fdb_entry *fdb;
+ u16 vid = 0;
+
+ if (!port)
+ return false;
+ if (br_opt_get(port->br, BROPT_VLAN_ENABLED)) {
+ if (skb_vlan_tag_present(skb))
+ vid = skb_vlan_tag_get_id(skb);
+ else
+ br_vlan_get_pvid_rcu(dev, &vid);
+ }
+ fdb = br_fdb_find_rcu(port->br, addr, vid);
+ return fdb && fdb->dst;
+}
+EXPORT_SYMBOL_GPL(br_fdb_has_forwarding_entry_rcu);
+
/* When a static FDB entry is added, the mac address from the entry is
* added to the bridge private HW address list and all required ports
* are then updated with the new information.
diff --git a/net/bridge/br_vlan.c b/net/bridge/br_vlan.c
index 5560afcaaca3..0b296362adf7 100644
--- a/net/bridge/br_vlan.c
+++ b/net/bridge/br_vlan.c
@@ -1559,6 +1559,51 @@ int br_vlan_get_info_rcu(const struct net_device *dev, u16 vid,
}
EXPORT_SYMBOL_GPL(br_vlan_get_info_rcu);
+/**
+ * br_vlan_get_offload_info_rcu - get VLAN id and protocol for bridge flow offload
+ * @dev: bridge port network device
+ * @skb: packet buffer
+ * @proto: output for the bridge VLAN protocol (set only when return value != 0)
+ *
+ * When VLAN filtering is enabled, resolves the VLAN id for flow offload (skb
+ * VLAN tag id if present, PVID otherwise) and writes the bridge VLAN protocol
+ * to @proto. Returns 0 when filtering is off or @dev is not a bridge port.
+ * Single br_port_get_rcu() lookup. Must be called under RCU read lock.
+ */
+u16 br_vlan_get_offload_info_rcu(const struct net_device *dev,
+ const struct sk_buff *skb, __be16 *proto)
+{
+ struct net_bridge_port *port = br_port_get_rcu(dev);
+ u16 vid = 0;
+
+ if (!port || !br_opt_get(port->br, BROPT_VLAN_ENABLED))
+ return 0;
+ if (skb_vlan_tag_present(skb))
+ vid = skb_vlan_tag_get_id(skb);
+ else
+ br_vlan_get_pvid_rcu(dev, &vid);
+ if (vid)
+ *proto = port->br->vlan_proto;
+ return vid;
+}
+EXPORT_SYMBOL_GPL(br_vlan_get_offload_info_rcu);
+
+/**
+ * br_vlan_is_enabled_rcu - check if VLAN filtering is active on a port's bridge
+ * @dev: bridge port network device
+ *
+ * Returns true if VLAN filtering is enabled on the bridge @dev belongs to.
+ * Returns false when @dev is not a bridge port or filtering is off.
+ * Must be called under RCU read lock.
+ */
+bool br_vlan_is_enabled_rcu(const struct net_device *dev)
+{
+ struct net_bridge_port *port = br_port_get_rcu(dev);
+
+ return port && br_opt_get(port->br, BROPT_VLAN_ENABLED);
+}
+EXPORT_SYMBOL_GPL(br_vlan_is_enabled_rcu);
+
static int br_vlan_is_bind_vlan_dev(const struct net_device *dev)
{
return is_vlan_dev(dev) &&
--
2.54.0
^ permalink raw reply related
* [PATCH 3/5] netfilter: nf_flow_table_path: add L2 bridge offload
From: Daniel Pawlik @ 2026-06-29 12:32 UTC (permalink / raw)
To: netfilter-devel, netdev
Cc: pablo, fw, phil, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, razor, idosch, matthias.bgg,
angelogioacchino.delregno, bridge, coreteam, linux-mediatek,
linux-arm-kernel, rchen14b, lorenzo, Daniel Pawlik
In-Reply-To: <20260629123253.1912621-1-pawlik.dan@gmail.com>
From: Ryan Chen <rchen14b@gmail.com>
Allow nft_flow_offload to accelerate traffic forwarded at layer 2 through
Linux bridge ports.
Detection: nft_flow_offload_is_bridging() identifies bridged flows by
checking that the ingress device is a bridge port and that the destination
MAC appears in the bridge FDB with a forwarding destination port (non-local
entry). VLAN resolution and FDB lookup are combined in a single
br_port_get_rcu() call via br_fdb_has_forwarding_entry_rcu().
Routing: nft_flow_route_bridging() allocates minimal dst entries anchored
to the bridge master device via rt_dst_alloc()/ip6_dst_alloc(). A full
routing table lookup via nf_route() is intentionally avoided: it fails for
prefixes that are only bridged, not routed, through the bridge interface
(e.g. when the bridge has no IP address or the bridged subnet is not in
the routing table).
MAC addresses: for bridged flows, nft_dev_forward_path() copies Ethernet
addresses directly from the packet header instead of going through the
neighbour table. Direction (original vs reply) is resolved against the
conntrack direction so both flow directions receive the correct MAC pair.
VLAN context: nft_br_vlan_dev_fill_forward_path() pre-populates the
net_device_path_ctx with the port VLAN id and protocol before the forward
path walk, enabling VLAN-aware hardware offload entries.
Also:
- info->indev is updated for every path type in nft_dev_path_info() so
the bridge ingress device is correctly tracked regardless of path type.
- nft_flow_route() is now a thin dispatcher that delegates to
nft_flow_route_routing() (routed traffic) or nft_flow_route_bridging()
(bridged traffic); the exported API is unchanged.
Path discovery infrastructure was moved to nf_flow_table_path.c in
commit 93d7a7ed0734 ("netfilter: flowtable: move path discovery
infrastructure to its own file"), so all changes land in that file.
Based on a MediaTek SDK patch by Bo-Cun Chen <bc-bocun.chen@mediatek.com>.
Co-developed-by: Daniel Pawlik <pawlik.dan@gmail.com>
Signed-off-by: Daniel Pawlik <pawlik.dan@gmail.com>
Signed-off-by: Ryan Chen <rchen14b@gmail.com>
---
net/netfilter/nf_flow_table_path.c | 167 +++++++++++++++++++++++++++--
1 file changed, 157 insertions(+), 10 deletions(-)
diff --git a/net/netfilter/nf_flow_table_path.c b/net/netfilter/nf_flow_table_path.c
index 98c03b487f52..6c470854127f 100644
--- a/net/netfilter/nf_flow_table_path.c
+++ b/net/netfilter/nf_flow_table_path.c
@@ -15,6 +15,10 @@
#include <net/netfilter/nf_conntrack_core.h>
#include <net/netfilter/nf_conntrack_extend.h>
#include <net/netfilter/nf_flow_table.h>
+#include <linux/if_bridge.h>
+#include <linux/if_ether.h>
+#include <net/route.h>
+#include <net/ip6_route.h>
static enum flow_offload_xmit_type nft_xmit_type(struct dst_entry *dst)
{
@@ -42,7 +46,25 @@ static bool nft_is_valid_ether_device(const struct net_device *dev)
return true;
}
-static int nft_dev_fill_forward_path(const struct nf_flow_route *route,
+static bool nft_flow_offload_is_bridging(struct sk_buff *skb)
+{
+ bool ret;
+
+ if (!netif_is_bridge_port(skb->dev))
+ return false;
+ if (!skb_mac_header_was_set(skb))
+ return false;
+
+ rcu_read_lock();
+ ret = br_fdb_has_forwarding_entry_rcu(skb->dev, skb,
+ eth_hdr(skb)->h_dest);
+ rcu_read_unlock();
+
+ return ret;
+}
+
+static int nft_dev_fill_forward_path(struct net_device_path_ctx *ctx,
+ const struct nf_flow_route *route,
const struct dst_entry *dst_cache,
const struct nf_conn *ct,
enum ip_conntrack_dir dir, u8 *ha,
@@ -58,6 +80,12 @@ static int nft_dev_fill_forward_path(const struct nf_flow_route *route,
goto out;
}
+ /* Bridging fastpath copies Ethernet addresses into ha; do not replace
+ * them via neighbour lookup on the routed destination device.
+ */
+ if (!is_zero_ether_addr(ha))
+ goto out;
+
n = dst_neigh_lookup(dst_cache, daddr);
if (!n)
return -1;
@@ -72,7 +100,23 @@ static int nft_dev_fill_forward_path(const struct nf_flow_route *route,
return -1;
out:
- return dev_fill_forward_path(dev, ha, stack);
+ return __dev_fill_forward_path(ctx, ha, stack);
+}
+
+static void nft_br_vlan_dev_fill_forward_path(const struct nft_pktinfo *pkt,
+ struct net_device_path_ctx *ctx)
+{
+ __be16 proto = 0;
+ u16 vlan_id;
+
+ rcu_read_lock();
+ vlan_id = br_vlan_get_offload_info_rcu(pkt->skb->dev, pkt->skb, &proto);
+ if (vlan_id) {
+ ctx->num_vlans = 1;
+ ctx->vlan[0].id = vlan_id;
+ ctx->vlan[0].proto = proto;
+ }
+ rcu_read_unlock();
}
struct nft_forward_info {
@@ -103,13 +147,13 @@ static int nft_dev_path_info(const struct net_device_path_stack *stack,
for (i = 0; i < stack->num_paths; i++) {
path = &stack->path[i];
+ info->indev = path->dev;
switch (path->type) {
case DEV_PATH_ETHERNET:
case DEV_PATH_DSA:
case DEV_PATH_VLAN:
case DEV_PATH_PPPOE:
case DEV_PATH_TUN:
- info->indev = path->dev;
if (is_zero_ether_addr(info->h_source))
memcpy(info->h_source, path->dev->dev_addr, ETH_ALEN);
@@ -244,6 +288,7 @@ static int nft_flow_tunnel_update_route(const struct nft_pktinfo *pkt,
}
static int nft_dev_forward_path(const struct nft_pktinfo *pkt,
+ bool is_bridging,
struct nf_flow_route *route,
const struct nf_conn *ct,
enum ip_conntrack_dir dir,
@@ -251,11 +296,33 @@ static int nft_dev_forward_path(const struct nft_pktinfo *pkt,
{
const struct dst_entry *dst = route->tuple[dir].dst;
struct net_device_path_stack stack;
+ struct net_device_path_ctx ctx = {
+ .dev = dst->dev,
+ };
struct nft_forward_info info = {};
+ enum ip_conntrack_info pkt_ctinfo;
+ enum ip_conntrack_dir skb_dir;
+ struct ethhdr *eth;
unsigned char ha[ETH_ALEN];
int i;
- if (nft_dev_fill_forward_path(route, dst, ct, dir, ha, &stack) < 0 ||
+ memset(ha, 0, sizeof(ha));
+
+ if (is_bridging) {
+ nf_ct_get(pkt->skb, &pkt_ctinfo);
+ eth = eth_hdr(pkt->skb);
+ skb_dir = CTINFO2DIR(pkt_ctinfo);
+ if (skb_dir != dir) {
+ memcpy(ha, eth->h_source, ETH_ALEN);
+ memcpy(info.h_source, eth->h_dest, ETH_ALEN);
+ } else {
+ memcpy(ha, eth->h_dest, ETH_ALEN);
+ memcpy(info.h_source, eth->h_source, ETH_ALEN);
+ }
+ nft_br_vlan_dev_fill_forward_path(pkt, &ctx);
+ }
+
+ if (nft_dev_fill_forward_path(&ctx, route, dst, ct, dir, ha, &stack) < 0 ||
nft_dev_path_info(&stack, &info, ha, &ft->data) < 0)
return -ENOENT;
@@ -292,9 +359,11 @@ static int nft_dev_forward_path(const struct nft_pktinfo *pkt,
return 0;
}
-int nft_flow_route(const struct nft_pktinfo *pkt, const struct nf_conn *ct,
- struct nf_flow_route *route, enum ip_conntrack_dir dir,
- struct nft_flowtable *ft)
+static int nft_flow_route_routing(const struct nft_pktinfo *pkt,
+ const struct nf_conn *ct,
+ struct nf_flow_route *route,
+ enum ip_conntrack_dir dir,
+ struct nft_flowtable *ft)
{
struct dst_entry *this_dst = skb_dst(pkt->skb);
struct dst_entry *other_dst = NULL;
@@ -334,12 +403,12 @@ int nft_flow_route(const struct nft_pktinfo *pkt, const struct nf_conn *ct,
nft_default_forward_path(route, this_dst, dir);
nft_default_forward_path(route, other_dst, !dir);
- if (route->tuple[dir].xmit_type == FLOW_OFFLOAD_XMIT_NEIGH &&
- nft_dev_forward_path(pkt, route, ct, dir, ft) < 0)
+ if (route->tuple[dir].xmit_type == FLOW_OFFLOAD_XMIT_NEIGH &&
+ nft_dev_forward_path(pkt, false, route, ct, dir, ft) < 0)
goto err_dst_release;
if (route->tuple[!dir].xmit_type == FLOW_OFFLOAD_XMIT_NEIGH &&
- nft_dev_forward_path(pkt, route, ct, !dir, ft) < 0)
+ nft_dev_forward_path(pkt, false, route, ct, !dir, ft) < 0)
goto err_dst_release;
return 0;
@@ -349,4 +418,82 @@ int nft_flow_route(const struct nft_pktinfo *pkt, const struct nf_conn *ct,
dst_release(route->tuple[!dir].dst);
return -ENOENT;
}
+
+static int nft_flow_route_bridging(const struct nft_pktinfo *pkt,
+ const struct nf_conn *ct,
+ struct nf_flow_route *route,
+ enum ip_conntrack_dir dir,
+ struct nft_flowtable *ft)
+{
+ struct dst_entry *dsts[IP_CT_DIR_MAX] = {};
+ struct net_device *br_dev;
+ int i;
+
+ /* Allocate minimal dsts anchored to the bridge master device to supply
+ * xmit_type and MTU. A full routing lookup via nf_route() is avoided
+ * because it fails for prefixes that are bridged but not routed.
+ */
+ rcu_read_lock();
+ br_dev = netdev_master_upper_dev_get_rcu(pkt->skb->dev);
+ if (!br_dev || !netif_is_bridge_master(br_dev)) {
+ rcu_read_unlock();
+ return -ENOENT;
+ }
+
+ for (i = 0; i < IP_CT_DIR_MAX; i++) {
+ switch (nft_pf(pkt)) {
+ case NFPROTO_IPV4: {
+ struct rtable *rt;
+
+ rt = rt_dst_alloc(br_dev, 0, RTN_UNICAST, true);
+ if (rt)
+ dsts[i] = &rt->dst;
+ break;
+ }
+ case NFPROTO_IPV6: {
+ struct rt6_info *rt;
+
+ rt = ip6_dst_alloc(nft_net(pkt), br_dev, 0);
+ if (rt)
+ dsts[i] = &rt->dst;
+ break;
+ }
+ }
+ }
+ rcu_read_unlock();
+
+ if (!dsts[dir] || !dsts[!dir]) {
+ dst_release(dsts[dir]);
+ dst_release(dsts[!dir]);
+ return -ENOENT;
+ }
+
+ nft_default_forward_path(route, dsts[dir], dir);
+ nft_default_forward_path(route, dsts[!dir], !dir);
+ /* Drop allocation references; route->tuple[*].dst holds the clones. */
+ dst_release(dsts[dir]);
+ dst_release(dsts[!dir]);
+
+ if (route->tuple[dir].xmit_type == FLOW_OFFLOAD_XMIT_NEIGH &&
+ route->tuple[!dir].xmit_type == FLOW_OFFLOAD_XMIT_NEIGH) {
+ if (nft_dev_forward_path(pkt, true, route, ct, dir, ft) ||
+ nft_dev_forward_path(pkt, true, route, ct, !dir, ft)) {
+ dst_release(route->tuple[dir].dst);
+ dst_release(route->tuple[!dir].dst);
+ return -ENOENT;
+ }
+ }
+
+ return 0;
+}
+
+int nft_flow_route(const struct nft_pktinfo *pkt, const struct nf_conn *ct,
+ struct nf_flow_route *route, enum ip_conntrack_dir dir,
+ struct nft_flowtable *ft)
+{
+ if (nft_flow_offload_is_bridging(pkt->skb))
+ return nft_flow_route_bridging(pkt, ct, route, dir, ft);
+
+ return nft_flow_route_routing(pkt, ct, route, dir, ft);
+}
EXPORT_SYMBOL_GPL(nft_flow_route);
--
2.54.0
^ permalink raw reply related
* [PATCH 4/5] netfilter: nf_flow_table_path: handle DEV_PATH_MTK_WDMA in path info
From: Daniel Pawlik @ 2026-06-29 12:32 UTC (permalink / raw)
To: netfilter-devel, netdev
Cc: pablo, fw, phil, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, razor, idosch, matthias.bgg,
angelogioacchino.delregno, bridge, coreteam, linux-mediatek,
linux-arm-kernel, rchen14b, lorenzo, Daniel Pawlik
In-Reply-To: <20260629123253.1912621-1-pawlik.dan@gmail.com>
From: Ryan Chen <rchen14b@gmail.com>
Without this change, nft_dev_path_info() hits the default -ENOENT path
for WiFi bridge offload via WDMA on MT7996. When a bridged flow goes
through the MT7996 WiFi device, the DEV_PATH_MTK_WDMA step does not set
h_source, causing the PPE entry to receive a zero source MAC and packets
to stall in both software fastpath and hardware path.
Based on a MediaTek SDK patch by Bo-Cun Chen <bc-bocun.chen@mediatek.com>.
Signed-off-by: Ryan Chen <rchen14b@gmail.com>
Signed-off-by: Daniel Pawlik <pawlik.dan@gmail.com>
---
net/netfilter/nf_flow_table_path.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/net/netfilter/nf_flow_table_path.c b/net/netfilter/nf_flow_table_path.c
index 6c470854127f..580aa1db3cb4 100644
--- a/net/netfilter/nf_flow_table_path.c
+++ b/net/netfilter/nf_flow_table_path.c
@@ -219,6 +219,10 @@ static int nft_dev_path_info(const struct net_device_path_stack *stack,
}
info->xmit_type = FLOW_OFFLOAD_XMIT_DIRECT;
break;
+ case DEV_PATH_MTK_WDMA:
+ if (is_zero_ether_addr(info->h_source))
+ memcpy(info->h_source, path->dev->dev_addr, ETH_ALEN);
+ break;
default:
return -1;
}
--
2.54.0
^ permalink raw reply related
* [PATCH 5/5] netfilter: nf_flow_table_path: add VLAN passthrough support
From: Daniel Pawlik @ 2026-06-29 12:32 UTC (permalink / raw)
To: netfilter-devel, netdev
Cc: pablo, fw, phil, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, razor, idosch, matthias.bgg,
angelogioacchino.delregno, bridge, coreteam, linux-mediatek,
linux-arm-kernel, rchen14b, lorenzo, Daniel Pawlik
In-Reply-To: <20260629123253.1912621-1-pawlik.dan@gmail.com>
From: Ryan Chen <rchen14b@gmail.com>
VLAN passthrough packets can be offloaded when bridge-nf-filter-vlan-tagged
is enabled. When a packet has a VLAN tag and the bridge does not have VLAN
filtering enabled (passthrough mode), record the VLAN encap info so the
hardware flow offload entry includes the correct VLAN tag.
Without this change, VLAN-tagged bridged traffic cannot be offloaded by PPE
because the VLAN encap information is missing from the flow entry.
Enable with: echo 1 > /proc/sys/net/bridge/bridge-nf-filter-vlan-tagged
Based on a MediaTek SDK patch by Chak-Kei Lam <chak-kei.lam@mediatek.com>.
Signed-off-by: Ryan Chen <rchen14b@gmail.com>
Signed-off-by: Daniel Pawlik <pawlik.dan@gmail.com>
---
net/netfilter/nf_flow_table_path.c | 32 ++++++++++++++++++++++++++++--
1 file changed, 30 insertions(+), 2 deletions(-)
diff --git a/net/netfilter/nf_flow_table_path.c b/net/netfilter/nf_flow_table_path.c
index 580aa1db3cb4..d15c425c88c4 100644
--- a/net/netfilter/nf_flow_table_path.c
+++ b/net/netfilter/nf_flow_table_path.c
@@ -17,6 +17,7 @@
#include <net/netfilter/nf_flow_table.h>
#include <linux/if_bridge.h>
#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
#include <net/route.h>
#include <net/ip6_route.h>
@@ -136,6 +137,29 @@ struct nft_forward_info {
enum flow_offload_xmit_type xmit_type;
};
+static void nft_fill_vlan_passthrough_info(const struct nft_pktinfo *pkt,
+ struct nft_forward_info *info)
+{
+ if (!skb_vlan_tag_present(pkt->skb))
+ return;
+
+ rcu_read_lock();
+ /* when bridge VLAN filtering is enabled, the bridge handles the tag */
+ if (netif_is_bridge_port(pkt->skb->dev) &&
+ !br_vlan_is_enabled_rcu(pkt->skb->dev)) {
+ if (info->num_encaps >= NF_FLOW_TABLE_ENCAP_MAX) {
+ info->indev = NULL;
+ } else {
+ info->encap[info->num_encaps].id =
+ skb_vlan_tag_get_id(pkt->skb);
+ info->encap[info->num_encaps].proto =
+ pkt->skb->vlan_proto;
+ info->num_encaps++;
+ }
+ }
+ rcu_read_unlock();
+}
+
static int nft_dev_path_info(const struct net_device_path_stack *stack,
struct nft_forward_info *info,
unsigned char *ha, struct nf_flowtable *flowtable)
@@ -326,8 +350,12 @@ static int nft_dev_forward_path(const struct nft_pktinfo *pkt,
nft_br_vlan_dev_fill_forward_path(pkt, &ctx);
}
- if (nft_dev_fill_forward_path(&ctx, route, dst, ct, dir, ha, &stack) < 0 ||
- nft_dev_path_info(&stack, &info, ha, &ft->data) < 0)
+ if (nft_dev_fill_forward_path(&ctx, route, dst, ct, dir, ha, &stack) < 0)
+ return -ENOENT;
+
+ nft_fill_vlan_passthrough_info(pkt, &info);
+
+ if (nft_dev_path_info(&stack, &info, ha, &ft->data) < 0)
return -ENOENT;
if (!nft_flowtable_find_dev(info.indev, ft))
--
2.54.0
^ permalink raw reply related
* [PATCH v2] ARM: enable interrupts when unhandled user faults are triggered
From: Xie Yuanbin @ 2026-06-29 12:33 UTC (permalink / raw)
To: linux, bigeasy, clrkwllms, rostedt, rmk+kernel, linusw, arnd
Cc: linux-arm-kernel, linux-kernel, linux-rt-devel, liaohua4,
lilinjie8, Xie Yuanbin
PREEMPT_RT requires interrupts to be enabled when sending signals.
When do_DataAbort()/do_PrefetchAbort() triggers unhandled user faults,
that is `inf->fn()` return a non-zero value, and the interrupts are not
enabled within the hook function, force_sig_fault() will be called
with interrupts disabled.
This can be triggered by user programs executing the bkpt instruction,
with kernel config CONFIG_PERF_EVENTS=n.
Enable interrupts in do_DataAbort()/do_PrefetchAbort() when unhandled
user faults are triggered to fix the issue.
Fixes: c6e61c06d606 ("ARM: 9463/1: Allow to enable RT")
Suggested-by: Russell King <linux@armlinux.org.uk>
Suggested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Xie Yuanbin <xieyuanbin1@huawei.com>
---
v1->v2: https://lore.kernel.org/20260625073522.182503-1-xieyuanbin1@huawei.com
- Move local_irq_enable() from arm_notify_die() to
do_DataAbort()/do_PrefetchAbort()
arch/arm/mm/fault.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c
index e62cc4be5adf..c68677503532 100644
--- a/arch/arm/mm/fault.c
+++ b/arch/arm/mm/fault.c
@@ -633,6 +633,9 @@ do_DataAbort(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
if (!inf->fn(addr, fsr & ~FSR_LNX_PF, regs))
return;
+ if (likely(user_mode(regs)))
+ local_irq_enable();
+
pr_alert("8<--- cut here ---\n");
pr_alert("Unhandled fault: %s (0x%03x) at 0x%08lx\n",
inf->name, fsr, addr);
@@ -663,6 +666,9 @@ do_PrefetchAbort(unsigned long addr, unsigned int ifsr, struct pt_regs *regs)
if (!inf->fn(addr, ifsr | FSR_LNX_PF, regs))
return;
+ if (likely(user_mode(regs)))
+ local_irq_enable();
+
pr_alert("8<--- cut here ---\n");
pr_alert("Unhandled prefetch abort: %s (0x%03x) at 0x%08lx\n",
inf->name, ifsr, addr);
--
2.53.0
^ permalink raw reply related
* RE: [PATCH] MAINTAINERS: Add Sai Krishna Potthuri to Xilinx AMS driver
From: O'Griofa, Conall @ 2026-06-29 12:34 UTC (permalink / raw)
To: Potthuri, Sai Krishna, Jonathan Cameron, Simek, Michal
Cc: linux-iio@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
linux-kernel@vger.kernel.org, Potthuri, Sai Krishna
In-Reply-To: <20260629114754.203415-1-sai.krishna.potthuri@amd.com>
AMD General
Acked-by: O'Griofa, Conall <conall.ogriofa@amd.com>
> -----Original Message-----
> From: Sai Krishna Potthuri <sai.krishna.potthuri@amd.com>
> Sent: 29 June 2026 12:48
> To: O'Griofa, Conall <conall.ogriofa@amd.com>; Jonathan Cameron
> <jic23@kernel.org>; Simek, Michal <michal.simek@amd.com>
> Cc: linux-iio@vger.kernel.org; linux-arm-kernel@lists.infradead.org; linux-
> kernel@vger.kernel.org; Potthuri, Sai Krishna <sai.krishna.potthuri@amd.com>
> Subject: [PATCH] MAINTAINERS: Add Sai Krishna Potthuri to Xilinx AMS driver
>
> Add Sai Krishna Potthuri as a maintainer for the Xilinx AMS driver for continued
> development and maintenance.
>
> Signed-off-by: Sai Krishna Potthuri <sai.krishna.potthuri@amd.com>
> ---
> MAINTAINERS | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 6b4560681b51..924e004a8a5e 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -29460,6 +29460,7 @@ F: include/uapi/linux/fsmap.h
> XILINX AMS DRIVER
> M: Salih Erim <salih.erim@amd.com>
> M: Conall O'Griofa <conall.ogriofa@amd.com>
> +M: Sai Krishna Potthuri <sai.krishna.potthuri@amd.com>
> L: linux-iio@vger.kernel.org
> S: Maintained
> F: Documentation/devicetree/bindings/iio/adc/xlnx,zynqmp-ams.yaml
> --
> 2.25.1
^ permalink raw reply
* [PATCH v2 0/6] clk: sun6i-rtc: Add support for Allwinner A733 SoC
From: Jerome Brunet @ 2026-06-29 12:42 UTC (permalink / raw)
To: Junhui Liu, Alexandre Belloni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Michael Turquette, Stephen Boyd, Maxime Ripard
Cc: linux-rtc, devicetree, linux-arm-kernel, linux-sunxi,
linux-kernel, linux-clk, Jerome Brunet
Add support for the Allwinner A733 RTC and its internal Clock Control
Unit (CCU). Reuse the rtc-sun6i rtc driver while introducing a new
SoC-specific RTC CCU driver to handle the hardware's evolved clock
structure.
The A733 implementation supports hardware detection of three external
crystal frequencies (19.2MHz, 24MHz and 26MHz), which is represented in
the driver via read-only divider operations. Implement logic to derive a
normalized 32kHz reference from these DCXO sources using fixed
pre-dividers. Additionally, provide several new DCXO gate clocks for
peripherals, including SerDes, HDMI, and UFS.
This was tested on a Raxda Cubie A7A.
Changes in v2:
* Changed DT bindings as suggested. Those have changed significantly
since v1 so I did not pick up Rob's review trailer
* Support added in the existing RTC CCU driver rather than a separate driver
* Added DT parsing clean up of the existing driver
* Xtal detection exposed by RO divider rather than a MUX.
* Dropped conversion to aux device for now. This is not strictly related
the a733 support and will submitted again later on.
* Link to v1: https://lore.kernel.org/r/20260121-a733-rtc-v1-0-d359437f23a7@pigmoral.tech
---
Jerome Brunet (5):
dt-bindings: rtc: sun6i: add sun60i-a733 support
clk: sunxi-ng: sun6i-rtc: clean up DT usage
clk: sunxi-ng: div: add read-only operation support
clk: sunxi-ng: sun6i-rtc: split main oscillator div and gate.
clk: sunxi-ng: sun6i-rtc: add a733 support
Junhui Liu (1):
clk: sunxi-ng: sun6i-rtc: Add feature bit for IOSC calibration
.../bindings/rtc/allwinner,sun6i-a31-rtc.yaml | 1 +
drivers/clk/sunxi-ng/ccu-sun6i-rtc.c | 124 +++++++++++++++------
drivers/clk/sunxi-ng/ccu-sun6i-rtc.h | 3 +-
drivers/clk/sunxi-ng/ccu_common.h | 1 +
drivers/clk/sunxi-ng/ccu_div.c | 11 ++
drivers/clk/sunxi-ng/ccu_div.h | 1 +
include/dt-bindings/clock/sun6i-rtc.h | 4 +
7 files changed, 109 insertions(+), 36 deletions(-)
---
base-commit: dc59e4fea9d83f03bad6bddf3fa2e52491777482
change-id: 20251226-a733-rtc-c5167df14e6e
Best regards,
--
Jerome
^ permalink raw reply
* [PATCH v2 1/6] dt-bindings: rtc: sun6i: add sun60i-a733 support
From: Jerome Brunet @ 2026-06-29 12:42 UTC (permalink / raw)
To: Junhui Liu, Alexandre Belloni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Michael Turquette, Stephen Boyd, Maxime Ripard
Cc: linux-rtc, devicetree, linux-arm-kernel, linux-sunxi,
linux-kernel, linux-clk, Jerome Brunet
In-Reply-To: <20260629-a733-rtc-v2-0-7b72112784f8@baylibre.com>
Add a new rtc compatible for the sun60i-a733 SoC and new IDs for the
peripheral oscillator clock gates of this SoC.
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
---
Documentation/devicetree/bindings/rtc/allwinner,sun6i-a31-rtc.yaml | 1 +
include/dt-bindings/clock/sun6i-rtc.h | 4 ++++
2 files changed, 5 insertions(+)
diff --git a/Documentation/devicetree/bindings/rtc/allwinner,sun6i-a31-rtc.yaml b/Documentation/devicetree/bindings/rtc/allwinner,sun6i-a31-rtc.yaml
index 9df5cdb6f63f..197a6f67b180 100644
--- a/Documentation/devicetree/bindings/rtc/allwinner,sun6i-a31-rtc.yaml
+++ b/Documentation/devicetree/bindings/rtc/allwinner,sun6i-a31-rtc.yaml
@@ -33,6 +33,7 @@ properties:
- enum:
- allwinner,sun20i-d1-rtc
- allwinner,sun55i-a523-rtc
+ - allwinner,sun60i-a733-rtc
- const: allwinner,sun50i-r329-rtc
reg:
diff --git a/include/dt-bindings/clock/sun6i-rtc.h b/include/dt-bindings/clock/sun6i-rtc.h
index 3bd3aa3d57ce..5132a393ca4b 100644
--- a/include/dt-bindings/clock/sun6i-rtc.h
+++ b/include/dt-bindings/clock/sun6i-rtc.h
@@ -6,5 +6,9 @@
#define CLK_OSC32K 0
#define CLK_OSC32K_FANOUT 1
#define CLK_IOSC 2
+#define CLK_HOSC_UFS 8
+#define CLK_HOSC_HDMI 9
+#define CLK_HOSC_SERDES0 10
+#define CLK_HOSC_SERDES1 11
#endif /* _DT_BINDINGS_CLK_SUN6I_RTC_H_ */
--
2.47.3
^ permalink raw reply related
* [PATCH v2 3/6] clk: sunxi-ng: sun6i-rtc: Add feature bit for IOSC calibration
From: Jerome Brunet @ 2026-06-29 12:42 UTC (permalink / raw)
To: Junhui Liu, Alexandre Belloni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Michael Turquette, Stephen Boyd, Maxime Ripard
Cc: linux-rtc, devicetree, linux-arm-kernel, linux-sunxi,
linux-kernel, linux-clk, Jerome Brunet
In-Reply-To: <20260629-a733-rtc-v2-0-7b72112784f8@baylibre.com>
From: Junhui Liu <junhui.liu@pigmoral.tech>
The sun6i-rtc CCU driver currently uses a global static variable to
denote whether calibration is supported, which makes IOSC operations
tightly coupled to this file.
Convert this into a feature bit to decouple the logic. This allows the
IOSC clock code to be moved into a shared module for reuse by other SoCs.
Signed-off-by: Junhui Liu <junhui.liu@pigmoral.tech>
Signed-off-by: Jerome Brunet<jbrunet@baylibre.com>
---
drivers/clk/sunxi-ng/ccu-sun6i-rtc.c | 17 +++++++++--------
drivers/clk/sunxi-ng/ccu_common.h | 1 +
2 files changed, 10 insertions(+), 8 deletions(-)
diff --git a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
index 0f528bfaed00..b24c8b196e66 100644
--- a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
+++ b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
@@ -52,8 +52,6 @@ struct sun6i_rtc_match_data {
u8 osc32k_fanout_nparents;
};
-static bool have_iosc_calibration;
-
static int ccu_iosc_enable(struct clk_hw *hw)
{
struct ccu_common *cm = hw_to_ccu_common(hw);
@@ -80,7 +78,7 @@ static unsigned long ccu_iosc_recalc_rate(struct clk_hw *hw,
{
struct ccu_common *cm = hw_to_ccu_common(hw);
- if (have_iosc_calibration) {
+ if (cm->features & CCU_FEATURE_IOSC_CALIBRATION) {
u32 reg = readl(cm->base + IOSC_CLK_CALI_REG);
/*
@@ -119,7 +117,7 @@ static int ccu_iosc_32k_prepare(struct clk_hw *hw)
struct ccu_common *cm = hw_to_ccu_common(hw);
u32 val;
- if (!have_iosc_calibration)
+ if (!(cm->features & CCU_FEATURE_IOSC_CALIBRATION))
return 0;
val = readl(cm->base + IOSC_CLK_CALI_REG);
@@ -134,7 +132,7 @@ static void ccu_iosc_32k_unprepare(struct clk_hw *hw)
struct ccu_common *cm = hw_to_ccu_common(hw);
u32 val;
- if (!have_iosc_calibration)
+ if (!(cm->features & CCU_FEATURE_IOSC_CALIBRATION))
return;
val = readl(cm->base + IOSC_CLK_CALI_REG);
@@ -148,7 +146,7 @@ static unsigned long ccu_iosc_32k_recalc_rate(struct clk_hw *hw,
struct ccu_common *cm = hw_to_ccu_common(hw);
u32 val;
- if (have_iosc_calibration) {
+ if (cm->features & CCU_FEATURE_IOSC_CALIBRATION) {
val = readl(cm->base + IOSC_CLK_CALI_REG);
/* Assume the calibrated 32k clock is accurate. */
@@ -167,7 +165,7 @@ static unsigned long ccu_iosc_32k_recalc_accuracy(struct clk_hw *hw,
struct ccu_common *cm = hw_to_ccu_common(hw);
u32 val;
- if (have_iosc_calibration) {
+ if (cm->features & CCU_FEATURE_IOSC_CALIBRATION) {
val = readl(cm->base + IOSC_CLK_CALI_REG);
/* Assume the calibrated 32k clock is accurate. */
@@ -358,7 +356,10 @@ int sun6i_rtc_ccu_probe(struct device *dev, void __iomem *reg)
return 0;
data = match->data;
- have_iosc_calibration = data->have_iosc_calibration;
+ if (data->have_iosc_calibration) {
+ iosc_clk.features |= CCU_FEATURE_IOSC_CALIBRATION;
+ iosc_32k_clk.features |= CCU_FEATURE_IOSC_CALIBRATION;
+ }
if (!data->have_ext_osc32k) {
/* ext-osc32k-gate is an orphan, so do not register it. */
diff --git a/drivers/clk/sunxi-ng/ccu_common.h b/drivers/clk/sunxi-ng/ccu_common.h
index bbec283b9d99..d9dc24ad5503 100644
--- a/drivers/clk/sunxi-ng/ccu_common.h
+++ b/drivers/clk/sunxi-ng/ccu_common.h
@@ -21,6 +21,7 @@
#define CCU_FEATURE_CLOSEST_RATE BIT(9)
#define CCU_FEATURE_DUAL_DIV BIT(10)
#define CCU_FEATURE_UPDATE_BIT BIT(11)
+#define CCU_FEATURE_IOSC_CALIBRATION BIT(12)
/* MMC timing mode switch bit */
#define CCU_MMC_NEW_TIMING_MODE BIT(30)
--
2.47.3
^ permalink raw reply related
* [PATCH v2 2/6] clk: sunxi-ng: sun6i-rtc: clean up DT usage
From: Jerome Brunet @ 2026-06-29 12:42 UTC (permalink / raw)
To: Junhui Liu, Alexandre Belloni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Michael Turquette, Stephen Boyd, Maxime Ripard
Cc: linux-rtc, devicetree, linux-arm-kernel, linux-sunxi,
linux-kernel, linux-clk, Jerome Brunet
In-Reply-To: <20260629-a733-rtc-v2-0-7b72112784f8@baylibre.com>
With sun6i-rtc compatible devices, the "ext-osc32k" clock input
is optional for the devices that support this input (r329 and onward).
Probably preparing for older SoC support, the driver does something funny
when parsing DT. It check if "ext-osc32k" is present in the clock-names and
if it is not, it uses the first clock as "ext-osc32k". This clock will
actually be the rtc bus clock so what the driver does is wrong.
At the moment, the driver does not support the older SoCs that would have
an external 32k clock provided on index #0 so just remove this quirk.
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
---
drivers/clk/sunxi-ng/ccu-sun6i-rtc.c | 23 +++--------------------
1 file changed, 3 insertions(+), 20 deletions(-)
diff --git a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
index f6bfeba009e8..0f528bfaed00 100644
--- a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
+++ b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
@@ -191,10 +191,8 @@ static struct ccu_common iosc_32k_clk = {
CLK_GET_RATE_NOCACHE),
};
-static const struct clk_hw *ext_osc32k[] = { NULL }; /* updated during probe */
-
-static SUNXI_CCU_GATE_HWS(ext_osc32k_gate_clk, "ext-osc32k-gate",
- ext_osc32k, 0x0, BIT(4), 0);
+static SUNXI_CCU_GATE_FW(ext_osc32k_gate_clk, "ext-osc32k-gate",
+ "ext-osc32k", 0x0, BIT(4), 0);
static const struct clk_hw *osc32k_parents[] = {
&iosc_32k_clk.hw,
@@ -352,7 +350,6 @@ MODULE_DEVICE_TABLE(of, sun6i_rtc_ccu_match);
int sun6i_rtc_ccu_probe(struct device *dev, void __iomem *reg)
{
const struct sun6i_rtc_match_data *data;
- struct clk *ext_osc32k_clk = NULL;
const struct of_device_id *match;
/* This driver is only used for newer variants of the hardware. */
@@ -363,21 +360,7 @@ int sun6i_rtc_ccu_probe(struct device *dev, void __iomem *reg)
data = match->data;
have_iosc_calibration = data->have_iosc_calibration;
- if (data->have_ext_osc32k) {
- const char *fw_name;
-
- /* ext-osc32k was the only input clock in the old binding. */
- fw_name = of_property_present(dev->of_node, "clock-names")
- ? "ext-osc32k" : NULL;
- ext_osc32k_clk = devm_clk_get_optional(dev, fw_name);
- if (IS_ERR(ext_osc32k_clk))
- return PTR_ERR(ext_osc32k_clk);
- }
-
- if (ext_osc32k_clk) {
- /* Link ext-osc32k-gate to its parent. */
- *ext_osc32k = __clk_get_hw(ext_osc32k_clk);
- } else {
+ if (!data->have_ext_osc32k) {
/* ext-osc32k-gate is an orphan, so do not register it. */
sun6i_rtc_ccu_hw_clks.hws[CLK_EXT_OSC32K_GATE] = NULL;
osc32k_init_data.num_parents = 1;
--
2.47.3
^ permalink raw reply related
* [PATCH v2 4/6] clk: sunxi-ng: div: add read-only operation support
From: Jerome Brunet @ 2026-06-29 12:42 UTC (permalink / raw)
To: Junhui Liu, Alexandre Belloni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Michael Turquette, Stephen Boyd, Maxime Ripard
Cc: linux-rtc, devicetree, linux-arm-kernel, linux-sunxi,
linux-kernel, linux-clk, Jerome Brunet
In-Reply-To: <20260629-a733-rtc-v2-0-7b72112784f8@baylibre.com>
Add support for sunxi-ng read-only dividers. This will be
useful to the a733 oscillator detection logic.
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
---
drivers/clk/sunxi-ng/ccu_div.c | 11 +++++++++++
drivers/clk/sunxi-ng/ccu_div.h | 1 +
2 files changed, 12 insertions(+)
diff --git a/drivers/clk/sunxi-ng/ccu_div.c b/drivers/clk/sunxi-ng/ccu_div.c
index 62d680ccb524..0e3d9e6e30fb 100644
--- a/drivers/clk/sunxi-ng/ccu_div.c
+++ b/drivers/clk/sunxi-ng/ccu_div.c
@@ -143,3 +143,14 @@ const struct clk_ops ccu_div_ops = {
.set_rate = ccu_div_set_rate,
};
EXPORT_SYMBOL_NS_GPL(ccu_div_ops, "SUNXI_CCU");
+
+const struct clk_ops ccu_rodiv_ops = {
+ .disable = ccu_div_disable,
+ .enable = ccu_div_enable,
+ .is_enabled = ccu_div_is_enabled,
+
+ .get_parent = ccu_div_get_parent,
+
+ .recalc_rate = ccu_div_recalc_rate,
+};
+EXPORT_SYMBOL_NS_GPL(ccu_rodiv_ops, "SUNXI_CCU");
diff --git a/drivers/clk/sunxi-ng/ccu_div.h b/drivers/clk/sunxi-ng/ccu_div.h
index be00b3277e97..a30a92780a05 100644
--- a/drivers/clk/sunxi-ng/ccu_div.h
+++ b/drivers/clk/sunxi-ng/ccu_div.h
@@ -300,5 +300,6 @@ static inline struct ccu_div *hw_to_ccu_div(struct clk_hw *hw)
}
extern const struct clk_ops ccu_div_ops;
+extern const struct clk_ops ccu_rodiv_ops;
#endif /* _CCU_DIV_H_ */
--
2.47.3
^ permalink raw reply related
* [PATCH v2 6/6] clk: sunxi-ng: sun6i-rtc: add a733 support
From: Jerome Brunet @ 2026-06-29 12:42 UTC (permalink / raw)
To: Junhui Liu, Alexandre Belloni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Michael Turquette, Stephen Boyd, Maxime Ripard
Cc: linux-rtc, devicetree, linux-arm-kernel, linux-sunxi,
linux-kernel, linux-clk, Jerome Brunet
In-Reply-To: <20260629-a733-rtc-v2-0-7b72112784f8@baylibre.com>
Add support for the sun60i a733 CCU RTC.
Compared to the a523, this SoC has a different input oscillator divider
which auto-detects the oscillator rate and select a divider to provide
a fixed 32768Hz clock. It also provides several phy reference clocks
with dedicated clock gates.
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
---
drivers/clk/sunxi-ng/ccu-sun6i-rtc.c | 74 ++++++++++++++++++++++++++++++++++--
drivers/clk/sunxi-ng/ccu-sun6i-rtc.h | 2 +-
2 files changed, 72 insertions(+), 4 deletions(-)
diff --git a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
index 25dd87e78eb7..2ec09e346472 100644
--- a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
+++ b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
@@ -42,11 +42,16 @@
#define LOSC_OUT_GATING_REG 0x60
#define DCXO_CTRL_REG 0x160
+#define DCXO_CTRL_DCXO_EN BIT(1)
#define DCXO_CTRL_CLK16M_RC_EN BIT(0)
+#define DCXO_GATING_REG 0x16c
+
struct sun6i_rtc_match_data {
bool have_ext_osc32k : 1;
bool have_iosc_calibration : 1;
+ bool have_dcxo_status : 1;
+ bool have_phy_ref_gates : 1;
bool rtc_32k_single_parent : 1;
const struct clk_parent_data *osc32k_fanout_parents;
u8 osc32k_fanout_nparents;
@@ -213,7 +218,12 @@ static struct ccu_mux osc32k_clk = {
},
};
-/* This falls back to the global name for fwnodes without a named reference. */
+/*
+ * This falls back to the global name for fwnodes without a named reference.
+ * NOTE: osc24M name might be misleading the oscillator could also be a 26MHz
+ * or a 19.2MHz one starting with the a733. The original name is kept anyway
+ * in case anything is relying on it.
+ */
static const struct clk_parent_data osc24M[] = {
{ .fw_name = "hosc", .name = "osc24M" }
};
@@ -227,8 +237,28 @@ static struct clk_fixed_factor osc24M_32k_div_clk = {
0),
};
-static SUNXI_CCU_GATE_HW(osc24M_32k_clk, "osc24M-32k", &osc24M_32k_div_clk.hw,
- LOSC_OUT_GATING_REG, BIT(16), 0);
+static struct clk_div_table osc24M_32k_div_a733_table[] = {
+ { .val = 0, .div = 732 },
+ { .val = 1, .div = 586 },
+ { .val = 2, .div = 793 },
+ { .val = 3, .div = 732 },
+ { /* Sentinel */ },
+};
+
+static struct ccu_div osc24M_32k_div_a733_clk = {
+ .enable = BIT(1),
+ .div = _SUNXI_CCU_DIV_TABLE(14, 2, osc24M_32k_div_a733_table),
+ .common = {
+ .reg = DCXO_CTRL_REG,
+ .hw.init = CLK_HW_INIT_PARENTS_DATA("osc24M-32k-div",
+ osc24M,
+ &ccu_rodiv_ops,
+ 0),
+ },
+};
+
+static SUNXI_CCU_GATE(osc24M_32k_clk, "osc24M-32k", "osc24M-32k-div",
+ LOSC_OUT_GATING_REG, BIT(16), 0);
static const struct clk_hw *rtc_32k_parents[] = {
&osc32k_clk.common.hw,
@@ -267,6 +297,15 @@ static struct ccu_mux osc32k_fanout_clk = {
},
};
+static SUNXI_CCU_GATE_FW(hosc_serdes1_clk, "hosc-serdes1", "hosc",
+ DCXO_GATING_REG, BIT(5), 0);
+static SUNXI_CCU_GATE_FW(hosc_serdes0_clk, "hosc-serdes0", "hosc",
+ DCXO_GATING_REG, BIT(4), 0);
+static SUNXI_CCU_GATE_FW(hosc_hdmi_clk, "hosc-hdmi", "hosc",
+ DCXO_GATING_REG, BIT(1), 0);
+static SUNXI_CCU_GATE_FW(hosc_ufs_clk, "hosc-ufs", "hosc",
+ DCXO_GATING_REG, BIT(0), 0);
+
static struct ccu_common *sun6i_rtc_ccu_clks[] = {
&iosc_clk,
&iosc_32k_clk,
@@ -275,6 +314,11 @@ static struct ccu_common *sun6i_rtc_ccu_clks[] = {
&osc24M_32k_clk.common,
&rtc_32k_clk.common,
&osc32k_fanout_clk.common,
+ &osc24M_32k_div_a733_clk.common,
+ &hosc_serdes1_clk.common,
+ &hosc_serdes0_clk.common,
+ &hosc_hdmi_clk.common,
+ &hosc_ufs_clk.common,
};
static struct clk_hw_onecell_data sun6i_rtc_ccu_hw_clks = {
@@ -288,6 +332,10 @@ static struct clk_hw_onecell_data sun6i_rtc_ccu_hw_clks = {
[CLK_OSC24M_32K] = &osc24M_32k_clk.common.hw,
[CLK_RTC_32K] = &rtc_32k_clk.common.hw,
[CLK_OSC24M_32K_DIV] = &osc24M_32k_div_clk.hw,
+ [CLK_HOSC_UFS] = &hosc_ufs_clk.common.hw,
+ [CLK_HOSC_HDMI] = &hosc_hdmi_clk.common.hw,
+ [CLK_HOSC_SERDES0] = &hosc_serdes0_clk.common.hw,
+ [CLK_HOSC_SERDES1] = &hosc_serdes1_clk.common.hw,
},
};
@@ -330,6 +378,15 @@ static const struct sun6i_rtc_match_data sun55i_a523_rtc_ccu_data = {
.osc32k_fanout_nparents = ARRAY_SIZE(sun50i_r329_osc32k_fanout_parents),
};
+static const struct sun6i_rtc_match_data sun60i_a733_rtc_ccu_data = {
+ .have_ext_osc32k = true,
+ .have_iosc_calibration = true,
+ .have_dcxo_status = true,
+ .have_phy_ref_gates = true,
+ .osc32k_fanout_parents = sun50i_r329_osc32k_fanout_parents,
+ .osc32k_fanout_nparents = ARRAY_SIZE(sun50i_r329_osc32k_fanout_parents),
+};
+
static const struct of_device_id sun6i_rtc_ccu_match[] = {
{
.compatible = "allwinner,sun50i-h616-rtc",
@@ -343,6 +400,10 @@ static const struct of_device_id sun6i_rtc_ccu_match[] = {
.compatible = "allwinner,sun55i-a523-rtc",
.data = &sun55i_a523_rtc_ccu_data,
},
+ {
+ .compatible = "allwinner,sun60i-a733-rtc",
+ .data = &sun60i_a733_rtc_ccu_data,
+ },
{},
};
MODULE_DEVICE_TABLE(of, sun6i_rtc_ccu_match);
@@ -375,6 +436,13 @@ int sun6i_rtc_ccu_probe(struct device *dev, void __iomem *reg)
osc32k_fanout_init_data.parent_data = data->osc32k_fanout_parents;
osc32k_fanout_init_data.num_parents = data->osc32k_fanout_nparents;
+ if (data->have_dcxo_status)
+ sun6i_rtc_ccu_hw_clks.hws[CLK_OSC24M_32K_DIV] =
+ &osc24M_32k_div_a733_clk.common.hw;
+
+ if (!data->have_phy_ref_gates)
+ sun6i_rtc_ccu_hw_clks.num = CLK_OSC24M_32K_DIV + 1;
+
return devm_sunxi_ccu_probe(dev, reg, &sun6i_rtc_ccu_desc);
}
diff --git a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h
index ab7b92b47f59..4f4f4cb00f1d 100644
--- a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h
+++ b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h
@@ -11,6 +11,6 @@
#define CLK_RTC_32K 6
#define CLK_OSC24M_32K_DIV 7
-#define CLK_NUMBER (CLK_OSC24M_32K_DIV + 1)
+#define CLK_NUMBER (CLK_HOSC_SERDES1 + 1)
#endif /* _CCU_SUN6I_RTC_H */
--
2.47.3
^ permalink raw reply related
* [PATCH v2 5/6] clk: sunxi-ng: sun6i-rtc: split main oscillator div and gate.
From: Jerome Brunet @ 2026-06-29 12:42 UTC (permalink / raw)
To: Junhui Liu, Alexandre Belloni, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Michael Turquette, Stephen Boyd, Maxime Ripard
Cc: linux-rtc, devicetree, linux-arm-kernel, linux-sunxi,
linux-kernel, linux-clk, Jerome Brunet
In-Reply-To: <20260629-a733-rtc-v2-0-7b72112784f8@baylibre.com>
On the a733 the "osc24M-32k" clock has the same gate bits as the previously
supported SoC but a different divider implementation.
Instead of a fixed 750 divider, the divider is selected based on the
rate of the oscillator. It can be seen as a simple read-only divider.
To easily replace the divider part depending the SoC, split the divider
and gate into two separate clock entities.
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
---
drivers/clk/sunxi-ng/ccu-sun6i-rtc.c | 20 +++++++++++---------
drivers/clk/sunxi-ng/ccu-sun6i-rtc.h | 3 ++-
2 files changed, 13 insertions(+), 10 deletions(-)
diff --git a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
index b24c8b196e66..25dd87e78eb7 100644
--- a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
+++ b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.c
@@ -218,17 +218,18 @@ static const struct clk_parent_data osc24M[] = {
{ .fw_name = "hosc", .name = "osc24M" }
};
-static struct ccu_gate osc24M_32k_clk = {
- .enable = BIT(16),
- .common = {
- .reg = LOSC_OUT_GATING_REG,
- .prediv = 750,
- .features = CCU_FEATURE_ALL_PREDIV,
- .hw.init = CLK_HW_INIT_PARENTS_DATA("osc24M-32k", osc24M,
- &ccu_gate_ops, 0),
- },
+static struct clk_fixed_factor osc24M_32k_div_clk = {
+ .mult = 1,
+ .div = 750,
+ .hw.init = CLK_HW_INIT_PARENTS_DATA("osc24M-32k-div",
+ osc24M,
+ &clk_fixed_factor_ops,
+ 0),
};
+static SUNXI_CCU_GATE_HW(osc24M_32k_clk, "osc24M-32k", &osc24M_32k_div_clk.hw,
+ LOSC_OUT_GATING_REG, BIT(16), 0);
+
static const struct clk_hw *rtc_32k_parents[] = {
&osc32k_clk.common.hw,
&osc24M_32k_clk.common.hw
@@ -286,6 +287,7 @@ static struct clk_hw_onecell_data sun6i_rtc_ccu_hw_clks = {
[CLK_EXT_OSC32K_GATE] = &ext_osc32k_gate_clk.common.hw,
[CLK_OSC24M_32K] = &osc24M_32k_clk.common.hw,
[CLK_RTC_32K] = &rtc_32k_clk.common.hw,
+ [CLK_OSC24M_32K_DIV] = &osc24M_32k_div_clk.hw,
},
};
diff --git a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h
index 9ae821fc2599..ab7b92b47f59 100644
--- a/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h
+++ b/drivers/clk/sunxi-ng/ccu-sun6i-rtc.h
@@ -9,7 +9,8 @@
#define CLK_EXT_OSC32K_GATE 4
#define CLK_OSC24M_32K 5
#define CLK_RTC_32K 6
+#define CLK_OSC24M_32K_DIV 7
-#define CLK_NUMBER (CLK_RTC_32K + 1)
+#define CLK_NUMBER (CLK_OSC24M_32K_DIV + 1)
#endif /* _CCU_SUN6I_RTC_H */
--
2.47.3
^ permalink raw reply related
* Re: cleanup the RAID6 P/Q library v3
From: H. Peter Anvin @ 2026-06-29 12:34 UTC (permalink / raw)
To: Christoph Hellwig
Cc: Andrew Morton, Catalin Marinas, Will Deacon, Ard Biesheuvel,
Huacai Chen, WANG Xuerui, Madhavan Srinivasan, Michael Ellerman,
Nicholas Piggin, Christophe Leroy (CS GROUP), Paul Walmsley,
Palmer Dabbelt, Albert Ou, Alexandre Ghiti, Heiko Carstens,
Vasily Gorbik, Alexander Gordeev, Christian Borntraeger,
Sven Schnelle, Thomas Gleixner, Ingo Molnar, Borislav Petkov,
Dave Hansen, x86, Herbert Xu, Dan Williams, Chris Mason,
David Sterba, Arnd Bergmann, Song Liu, Yu Kuai, Li Nan,
linux-kernel, linux-arm-kernel, loongarch, linuxppc-dev,
linux-riscv, linux-s390, linux-crypto, linux-btrfs, linux-arch,
linux-raid
In-Reply-To: <20260629122152.GA21344@lst.de>
On June 29, 2026 5:21:52 AM PDT, Christoph Hellwig <hch@lst.de> wrote:
>On Fri, Jun 26, 2026 at 05:52:18PM -0700, H. Peter Anvin wrote:
>> On 2026-05-19 01:24, Christoph Hellwig wrote:
>> >
>> > raid6: rework registration of optimized algorithms
>> >
>> > - avx2 instead of avx512 is probably the right thing for no
>> > benchmarking, but if it was intentional (it wasn't), that should
>> > be document. So I'll just switch back to the previous version to
>> > keep the state of the art
>>
>> It is unlikely to be the right thing *going forward*, though.
>>
>> The very unfortunate performance inversion is likely model-specific. It is one
>> of those things where you largely would have to have a list of quirks :(
>
>Note that the default is still to benchmark the implementations.
>This is just about the case where the user explicitly disabled that
>benchmark-based selection.
>
The real issue though is that the benchmarking isn't really helpful, either.
It has to do with power across the whole chip, so it perversely enough depends on what *else* is going on...
^ permalink raw reply
* Re: cleanup the RAID6 P/Q library v3
From: Christoph Hellwig @ 2026-06-29 12:48 UTC (permalink / raw)
To: H. Peter Anvin
Cc: Christoph Hellwig, Andrew Morton, Catalin Marinas, Will Deacon,
Ard Biesheuvel, Huacai Chen, WANG Xuerui, Madhavan Srinivasan,
Michael Ellerman, Nicholas Piggin, Christophe Leroy (CS GROUP),
Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti,
Heiko Carstens, Vasily Gorbik, Alexander Gordeev,
Christian Borntraeger, Sven Schnelle, Thomas Gleixner,
Ingo Molnar, Borislav Petkov, Dave Hansen, x86, Herbert Xu,
Dan Williams, Chris Mason, David Sterba, Arnd Bergmann, Song Liu,
Yu Kuai, Li Nan, linux-kernel, linux-arm-kernel, loongarch,
linuxppc-dev, linux-riscv, linux-s390, linux-crypto, linux-btrfs,
linux-arch, linux-raid
In-Reply-To: <8684FEC2-AD56-4BDA-874F-1D3D7170F880@zytor.com>
On Mon, Jun 29, 2026 at 05:34:41AM -0700, H. Peter Anvin wrote:
> The real issue though is that the benchmarking isn't really helpful, either.
>
> It has to do with power across the whole chip, so it perversely enough depends on what *else* is going on...
Yes. And preferences might matter if this is just a file server or
something running a real workload as well. But I don't really have a
good answer to that.
^ permalink raw reply
* Re: [PATCH v2] ARM: enable interrupts when unhandled user faults are triggered
From: Xie Yuanbin @ 2026-06-29 12:48 UTC (permalink / raw)
To: bigeasy, linux, rmk+kernel
Cc: xieyuanbin1, arnd, clrkwllms, liaohua4, lilinjie8, linusw,
linux-arm-kernel, linux-kernel, linux-rt-devel, rostedt
In-Reply-To: <20260629123349.134224-1-xieyuanbin1@huawei.com>
On Mon, 29 Jun 2026 11:40:22 +0200, Sebastian Andrzej Siewior wrote:
> If this is moved to the callers of arm_notify_die() then I don't know
> what to do about baddataabort(). It looks like it gets invoked with
> disabled interrupts, too but I'm not sure. This looks like pre ARM v7.
> Is it reasonable to add such a check there?
Oh, I'm not sure about this place either. Would interrupts here be
disabled? I'm not familiar with the entry assembly code of v4t, so
I don't know how to construct a user-space program that can run to
baddataabort(), and I also don't have a v4t device to test it.
^ permalink raw reply
* Re: [PATCH 6.1 337/522] arm64/mm: Enable batched TLB flush in unmap_hotplug_range()
From: Will Deacon @ 2026-06-29 12:50 UTC (permalink / raw)
To: Anshuman Khandual
Cc: Greg Kroah-Hartman, Ryan Roberts, Ben Hutchings, Catalin Marinas,
David Hildenbrand (Arm), patches, linux-arm-kernel, linux-kernel,
Sasha Levin, stable, mark.rutland
In-Reply-To: <b51f82ed-aa54-4c67-bcef-e59acd10c789@arm.com>
On Thu, Jun 25, 2026 at 07:59:51AM +0530, Anshuman Khandual wrote:
> On 24/06/26 9:59 PM, Greg Kroah-Hartman wrote:
> > On Wed, Jun 24, 2026 at 04:05:01PM +0100, Ryan Roberts wrote:
> >> On 23/06/2026 15:25, Will Deacon wrote:
> >>> On Sun, Jun 21, 2026 at 05:02:27PM +0200, Ben Hutchings wrote:
> >>>>> @@ -949,15 +953,14 @@ static void unmap_hotplug_pmd_range(pud_
> >>>>> WARN_ON(!pmd_present(pmd));
> >>>>> if (pmd_sect(pmd)) {
> >>>>> pmd_clear(pmdp);
> >>>>> -
> >>>>> - /*
> >>>>> - * One TLBI should be sufficient here as the PMD_SIZE
> >>>>> - * range is mapped with a single block entry.
> >>>>> - */
> >>>>> - flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
> >>>>> - if (free_mapped)
> >>>>> + if (free_mapped) {
> >>>>> + /* CONT blocks are not supported in the vmemmap */
> >>>>> + WARN_ON(pmd_cont(pmd));
> >>>>> + flush_tlb_kernel_range(addr, addr + PMD_SIZE);
> >>>>
> >>>> It wasn't clear to me from the commit message why this now adds PMD_SIZE
> >>>> rather than PAGE_SIZE. It seems like this change is fine for Linux
> >>>> 6.13+ with a CPU that supports TLB range flushing, but otherwise results
> >>>> in unnecessarily executing multiple TLB invalidations at intervals of
> >>>> the base page size.
> >>>
> >>> Hmm, the commit message also makes very little sense to me and so I don't
> >>> understand why this patch has us doing multiple TLB invalidations when
> >>> we run into a !cont, block mapping at the PMD level. The old comment
> >>> (which this patch removes) should still apply afaict.
> >>>
> >>> Anshuman, Ryan, any ideas what's going on here?
> >>
> >> I think this change was probably my fault; Given the API is called
> >> flush_tlb_kernel_range() it seemed like an abuse/hack to pretend we are only
> >> flushing the first PAGE_SIZE of the range. But as I understand it, even if the
> >> HW shatters a block mapping into multiple TLB entries, all of the entries
> >> relating to the block mapping will be invalidated if just one of them intersects
> >> the TLBI range/address. So it should be safe to reapply this hack.
> >>
> >> Although ideally I think it would be better if this API took a stride argument;
> >> then intent is clear.
> >>
> >> What's the best way to handle this? Submit a patch for mainline that reverts
> >> this part, then get it backported to stable (implying this current patch will
> >> have been applied to stable)?
> >
> > yes, that's probably the best way.
> Sure, will send out the change as suggested.
In case anybody ends up following the breadcrumbs, the patch is here:
https://lore.kernel.org/r/20260626012845.475959-1-anshuman.khandual@arm.com
Will
^ permalink raw reply
* Re: [PATCH RFC 0/8] clk: sunxi-ng: Add support for Allwinner A733 CCU and PRCM
From: Jerome Brunet @ 2026-06-29 12:52 UTC (permalink / raw)
To: Junhui Liu
Cc: Michael Turquette, Stephen Boyd, Rob Herring, Krzysztof Kozlowski,
Conor Dooley, Chen-Yu Tsai, Jernej Skrabec, Samuel Holland,
Philipp Zabel, Paul Walmsley, Palmer Dabbelt, Albert Ou,
Alexandre Ghiti, Richard Cochran, linux-clk, devicetree,
linux-arm-kernel, linux-sunxi, linux-kernel, linux-riscv, netdev
In-Reply-To: <20260310-a733-clk-v1-0-36b4e9b24457@pigmoral.tech>
On mar. 10 mars 2026 at 16:33, Junhui Liu <junhui.liu@pigmoral.tech> wrote:
> Add support for the main CCU and the PRCM module (R-CCU) found in the
> Allwinner A733 SoC. The clock architecture of the A733 is an evolution
> of the previous A523 design but introduces several significant changes.
>
> One of the key changes is the introduction of a "pll-ref" clock that
> normalizes the physical oscillator frequency (which can be 19.2MHz,
> 24MHz, or 26MHz) into a consistent 24MHz reference for the entire clock
> tree. Additionally, while the A733 inherits many module clock structures
> from the A523, the MCU_CCU has been removed, and the overall clock tree
> has been expanded to support more new functional units.
>
> Also update the sunxi-ng SDM (Sigma-Delta Modulation) helper to support
> a new dual-pattern register design. On the A733, the SDM enable bit has
> been moved from the main PLL register to a second pattern register
> (PATTERN1). The driver is updated to handle this register layout to
> ensure accurate frequency synthesis for "pll-audio0".
>
> This is marked as RFC because the parent clocks for several instances in
> the main CCU are difficult to determine as the user manual provides
> limited information on their specific clock sources. In these cases, the
> implementation follows vendor practices and previous SoC designs,
> generally defaulting to "hosc" where documentation is lacking. In
> contrast, the bus clock gates in the PRCM (R-CCU) are explicitly defined
> based on the Memory Map in the manual, which clearly associates each
> module with its respective bus. Feedback or insights on these specific
> clock parents would be greatly appreciated.
>
> This functionally relies on the RTC series for the A733 SoC [1].
>
> Link: https://lore.kernel.org/all/20260121-a733-rtc-v1-0-d359437f23a7@pigmoral.tech/ [1]
>
Tested-by: Jerome Brunet <jbrunet@baylibre.com>
> ---
> Junhui Liu (8):
> dt-bindings: clk: sun60i-a733-ccu: Add allwinner A733 support
> clk: sunxi-ng: sdm: Add dual patterns support
> clk: sunxi-ng: a733: Add PRCM CCU
> clk: sunxi-ng: a733: Add PLL clocks support
> clk: sunxi-ng: a733: Add bus clocks support
> clk: sunxi-ng: a733: Add mod clocks support
> clk: sunxi-ng: a733: Add bus clock gates
> clk: sunxi-ng: a733: Add reset lines
>
> .../bindings/clock/allwinner,sun60i-a733-ccu.yaml | 107 +
> drivers/clk/sunxi-ng/Kconfig | 10 +
> drivers/clk/sunxi-ng/Makefile | 4 +
> drivers/clk/sunxi-ng/ccu-sun60i-a733-r.c | 276 +++
> drivers/clk/sunxi-ng/ccu-sun60i-a733.c | 2375 ++++++++++++++++++++
> drivers/clk/sunxi-ng/ccu_sdm.c | 51 +-
> drivers/clk/sunxi-ng/ccu_sdm.h | 32 +-
> include/dt-bindings/clock/sun60i-a733-ccu.h | 289 +++
> include/dt-bindings/clock/sun60i-a733-r-ccu.h | 39 +
> include/dt-bindings/reset/sun60i-a733-ccu.h | 131 ++
> include/dt-bindings/reset/sun60i-a733-r-ccu.h | 23 +
> 11 files changed, 3311 insertions(+), 26 deletions(-)
> ---
> base-commit: 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f
> change-id: 20260202-a733-clk-0d4fc00a9f9c
>
> Best regards,
--
Jerome
^ permalink raw reply
* Re: [PATCH 0/5] netfilter: nf_flow_table_path: L2 bridge offload
From: Florian Westphal @ 2026-06-29 12:56 UTC (permalink / raw)
To: Daniel Pawlik
Cc: netfilter-devel, netdev, pablo, phil, davem, edumazet, kuba,
pabeni, horms, andrew+netdev, razor, idosch, matthias.bgg,
angelogioacchino.delregno, bridge, coreteam, linux-mediatek,
linux-arm-kernel, rchen14b, lorenzo
In-Reply-To: <20260629123253.1912621-1-pawlik.dan@gmail.com>
Daniel Pawlik <pawlik.dan@gmail.com> wrote:
> -----------------------------
> 1. Load kmod-br-netfilter so that bridged IP traffic traverses the
> netfilter forward chain.
Ouch. br_netfilter should die. Really. Its a gross hack, never
use this thing...
> 2. Enable netfilter hooks on the bridge:
> echo 1 > /sys/class/net/<br>/bridge/nf_call_iptables
> echo 1 > /sys/class/net/<br>/bridge/nf_call_ip6tables
>
> 3. Register bridge member interfaces in the nft flowtable:
> table inet filter {
> flowtable f {
> hook ingress priority filter
> devices = { eth0, wlan0 }
> }
I think that bridge flowtable should use 'table bridge ...', not
use the br_netfilter compat hacks.
Sorry.
Are you aware of Eric Woudstras bridge flowtable patches?
https://lore.kernel.org/netfilter-devel/20250408142802.96101-5-ericwouds@gmail.com/
^ permalink raw reply
* Re: [PATCH v4 0/4] arm64: cross-CPU NMI via SDEI
From: Kiryl Shutsemau @ 2026-06-29 13:05 UTC (permalink / raw)
To: Catalin Marinas
Cc: Will Deacon, James Morse, Mark Rutland, Marc Zyngier,
Doug Anderson, Petr Mladek, Thomas Gleixner, Andrew Morton,
Baoquan He, Puranjay Mohan, Usama Arif, Breno Leitao,
Julien Thierry, Lecopzer Chen, Sumit Garg, kernel-team, kexec,
linux-arm-kernel, linux-kernel
In-Reply-To: <aj7Pm5H0MN5WwhDK@thinkstation>
On Fri, Jun 26, 2026 at 08:40:57PM +0100, Kiryl Shutsemau wrote:
> But I have not tried calling CPU_OFF directly, without completing the
> event. I assumed it is required. Will give it a try when I have time.
Tried it now, and it doesn't work either -- in a more interesting way.
Calling PSCI CPU_OFF directly from the SDEI handler (event left
uncompleted) reproducibly breaks the kdump capture kernel, and this
reproduces under QEMU's TF-A, not just on Grace -- so it isn't a Grace
firmware quirk.
The test: a CPU wedged with interrupts masked is stopped via the SDEI
rung; its handler calls __cpu_try_die() instead of parking. A/B in QEMU,
changing only that wedged CPU's handling (everything else identical):
- park it (current series): capture kernel boots fully to a shell.
- CPU_OFF from the handler: capture kernel hangs in early boot, around
SDEI re-init, never reaches a shell.
Powering the PE off while its SDEI event is still active leaves EL3's
dispatch state dangling, and the capture kernel trips over it. Completing
the event first and then CPU_OFF -- what I tried originally -- silently
wedges EL3 on Grace instead.
So both routes off fail, and the CPU stays parked. The dump is complete
either way; only re-onlining the stopped CPU in an SMP capture kernel is
lost. It's a cheap QEMU repro now if anyone wants to dig into the EL3
side.
--
Kiryl Shutsemau / Kirill A. Shutemov
^ permalink raw reply
* Re: [PATCH v2] arm64: ptrace: use live x0 for seccomp and audit after ptrace
From: Will Deacon @ 2026-06-29 13:09 UTC (permalink / raw)
To: Yiqi Sun
Cc: catalin.marinas, linux-arm-kernel, linux-kernel, rmk+kernel,
ruanjinjie, kees, mark.rutland
In-Reply-To: <2f435bab0d61d0bf8fbaa54203525aae8e8f5371.1782384161.git.sunyiqixm@gmail.com>
Hi Yiqi,
On Thu, Jun 25, 2026 at 06:45:02PM +0800, Yiqi Sun wrote:
> On arm64, seccomp obtains syscall arguments via
> syscall_get_arguments(), where arg0 is currently read from
> regs->orig_x0. audit_syscall_entry() in syscall_trace_enter() also
> takes arg0 from regs->orig_x0. However, the syscall wrapper consumes
> live arguments from regs->regs[0..5].
>
> A ptracer can modify x0 on syscall-enter stop before seccomp and audit
> run, but cannot update orig_x0 through the native syscall-stop
> interface. This can leave seccomp and audit checking stale arg0 while
> the syscall executes with updated live x0.
>
> Make both paths read arg0 from regs->regs[0], matching the actual
> dispatch arguments and keeping seccomp and audit aligned after ptrace
> updates.
>
> Fixes: f27bb139c387 ("arm64: Miscellaneous library functions")
> Signed-off-by: Yiqi Sun <sunyiqixm@gmail.com>
> ---
> Changes in v2:
> - Also switch the arm64 audit entry path to use live x0
> - Clarify the orig_x0 synchronization comment in syscall_set_arguments()
> ---
> arch/arm64/include/asm/syscall.h | 7 +++----
> arch/arm64/kernel/ptrace.c | 2 +-
> 2 files changed, 4 insertions(+), 5 deletions(-)
Sashiko has pointed out some issues with this patch that look legitimate
to me:
https://sashiko.dev/#/patchset/2f435bab0d61d0bf8fbaa54203525aae8e8f5371.1782384161.git.sunyiqixm@gmail.com
Specifically, we don't appear to handle NO_SYSCALL properly and the
syscall-exit stop is now going to see the return code instead of the
syscall number.
Looking at this more broadly, it looks like orig_x0 is used for three
different cases:
1. syscall restarting:
We restore from orig_x0, which should hold the
original value passed by userspace.
2. syscall_get_arguments():
This must work correctly vs syscall_set_arguments()
(returning the latest set x0) but also
syscall_get_return_value() (so we need to
distinguish the return value and the argument
somehow).
3. syscall_rollback():
Seccomp wants to restore the original values
passed by userspace.
So (1) and (3) look to require the same behaviour, but (2) wants
something different because it needs to reflect changes made via
syscall_set_arguments().
The bodge we have for (2) today is that syscall_set_arguments() updates
orig_x0, but I think that breaks (1) and (2) which is the underlying
problem you're facing here.
I haven't yet figured out the right way to fix this, but I'd be interested
to hear from others. I think the starting point would be removing orig_x0
from syscall_{get,set}_arguments() altogether so that it accurately
represents the initial value passed by userspace.
Will
^ permalink raw reply
* Re: [PATCH] soc: ti: knav_qmss: remove debugfs file on teardown
From: Nishanth Menon @ 2026-06-29 13:12 UTC (permalink / raw)
To: Pengpeng Hou; +Cc: Santosh Shilimkar, linux-kernel, linux-arm-kernel
In-Reply-To: <20260615091251.7129-1-pengpeng@iscas.ac.cn>
On 17:12-20260615, Pengpeng Hou wrote:
> knav_queue_probe() creates the global qmss debugfs file whose show
> callback reads the global kdev state. knav_queue_remove() disables
> runtime PM but leaves the debugfs file and ready state published.
>
> Save the debugfs dentry, remove it during teardown, and clear the global
> ready pointer state.
Please update the commit message once you have rebased.
>
> Signed-off-by: Pengpeng Hou <pengpeng@iscas.ac.cn>
> ---
> drivers/soc/ti/knav_qmss_queue.c | 11 +++++++++--
> 1 file changed, 9 insertions(+), 2 deletions(-)
Please rebase to v7.2-rc1 or the latest linux-next please. The patch no
longer applies.
>
> diff --git a/drivers/soc/ti/knav_qmss_queue.c b/drivers/soc/ti/knav_qmss_queue.c
> index 86d7a9c9ae01..1cc2e4a90c0d 100644
> --- a/drivers/soc/ti/knav_qmss_queue.c
> +++ b/drivers/soc/ti/knav_qmss_queue.c
> @@ -26,6 +26,7 @@
> #include "knav_qmss.h"
>
> static struct knav_device *kdev;
> +static struct dentry *knav_queue_debugfs;
Does this belong to knav_device?
> static DEFINE_MUTEX(knav_dev_lock);
> #define knav_dev_lock_held() \
> lockdep_is_held(&knav_dev_lock)
> @@ -1857,8 +1858,9 @@ static int knav_queue_probe(struct platform_device *pdev)
> goto err;
> }
>
> - debugfs_create_file("qmss", S_IFREG | S_IRUGO, NULL, NULL,
> - &knav_queue_debug_fops);
> + knav_queue_debugfs = debugfs_create_file("qmss", 0444,
I think the 0444 change was triggered by checkpatch.pl? Lets document
the change in commit message?
> + NULL, NULL,
> + &knav_queue_debug_fops);
> device_ready = true;
> return 0;
>
> @@ -1873,9 +1875,14 @@ static int knav_queue_probe(struct platform_device *pdev)
>
> static void knav_queue_remove(struct platform_device *pdev)
> {
> + device_ready = false;
This should already be present.
> + debugfs_remove(knav_queue_debugfs);
> + knav_queue_debugfs = NULL;
> +
> /* TODO: Free resources */
> pm_runtime_put_sync(&pdev->dev);
> pm_runtime_disable(&pdev->dev);
> + kdev = NULL;
I think we renamed this in recent patches. knav_qdev = NULL and should
get the memory released once the remove function returns back?
> }
>
> static struct platform_driver keystone_qmss_driver = {
> --
> 2.50.1 (Apple Git-155)
>
>
--
Regards,
Nishanth Menon
Key (0xDDB5849D1736249D) / Fingerprint: F8A2 8693 54EB 8232 17A3 1A34 DDB5 849D 1736 249D
https://ti.com/opensource
^ permalink raw reply
* Re: [PATCH v2] ARM: enable interrupts when unhandled user faults are triggered
From: Sebastian Andrzej Siewior @ 2026-06-29 13:12 UTC (permalink / raw)
To: Xie Yuanbin
Cc: linux, rmk+kernel, arnd, clrkwllms, liaohua4, lilinjie8, linusw,
linux-arm-kernel, linux-kernel, linux-rt-devel, rostedt
In-Reply-To: <20260629124816.136079-1-xieyuanbin1@huawei.com>
On 2026-06-29 20:48:16 [+0800], Xie Yuanbin wrote:
> On Mon, 29 Jun 2026 11:40:22 +0200, Sebastian Andrzej Siewior wrote:
> > If this is moved to the callers of arm_notify_die() then I don't know
> > what to do about baddataabort(). It looks like it gets invoked with
> > disabled interrupts, too but I'm not sure. This looks like pre ARM v7.
> > Is it reasonable to add such a check there?
>
> Oh, I'm not sure about this place either. Would interrupts here be
> disabled? I'm not familiar with the entry assembly code of v4t, so
> I don't know how to construct a user-space program that can run to
> baddataabort(), and I also don't have a v4t device to test it.
I don't have anything myself but it follows the same pattern.
Sashiko just reported that it is a pre-existing issue :)
Sebastian
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox