* [PATCHES] Networking and Sparc
@ 2017-06-08 3:19 David Miller
2017-06-08 7:09 ` Greg KH
0 siblings, 1 reply; 2+ messages in thread
From: David Miller @ 2017-06-08 3:19 UTC (permalink / raw)
To: stable
[-- Attachment #1: Type: Text/Plain, Size: 110 bytes --]
Please queue up the following sparc and networking bug fixes for 4.9
and 4.11 -stable respectively.
Thanks!
[-- Attachment #2: net_49.mbox --]
[-- Type: Application/Octet-Stream, Size: 22084 bytes --]
From ef35a111ceb6726d163ecd5e6434fd27fe535fa9 Mon Sep 17 00:00:00 2001
From: "Mintz, Yuval" <Yuval.Mintz@cavium.com>
Date: Thu, 1 Jun 2017 15:57:56 +0300
Subject: [PATCH 01/12] bnx2x: Fix Multi-Cos
[ Upstream commit 3968d38917eb9bd0cd391265f6c9c538d9b33ffa ]
Apparently multi-cos isn't working for bnx2x quite some time -
driver implements ndo_select_queue() to allow queue-selection
for FCoE, but the regular L2 flow would cause it to modulo the
fallback's result by the number of queues.
The fallback would return a queue matching the needed tc
[via __skb_tx_hash()], but since the modulo is by the number of TSS
queues where number of TCs is not accounted, transmission would always
be done by a queue configured into using TC0.
Fixes: ada7c19e6d27 ("bnx2x: use XPS if possible for bnx2x_select_queue instead of pure hash")
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
index 0a9108c..0a5ee1d 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
@@ -1931,7 +1931,7 @@ u16 bnx2x_select_queue(struct net_device *dev, struct sk_buff *skb,
}
/* select a non-FCoE queue */
- return fallback(dev, skb) % BNX2X_NUM_ETH_QUEUES(bp);
+ return fallback(dev, skb) % (BNX2X_NUM_ETH_QUEUES(bp) * bp->max_cos);
}
void bnx2x_set_num_queues(struct bnx2x *bp)
--
2.9.4
From dce5f289f163bf27f00c4ce23c4c9585581d8ef4 Mon Sep 17 00:00:00 2001
From: Lance Richardson <lrichard@redhat.com>
Date: Mon, 29 May 2017 13:25:57 -0400
Subject: [PATCH 02/12] vxlan: eliminate cached dst leak
[ Upstream commit 35cf2845563c1aaa01d27bd34d64795c4ae72700 ]
After commit 0c1d70af924b ("net: use dst_cache for vxlan device"),
cached dst entries could be leaked when more than one remote was
present for a given vxlan_fdb entry, causing subsequent netns
operations to block indefinitely and "unregister_netdevice: waiting
for lo to become free." messages to appear in the kernel log.
Fix by properly releasing cached dst and freeing resources in this
case.
Fixes: 0c1d70af924b ("net: use dst_cache for vxlan device")
Signed-off-by: Lance Richardson <lrichard@redhat.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/vxlan.c | 20 +++++++++++++++++---
1 file changed, 17 insertions(+), 3 deletions(-)
diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index 3c4c2cf..3331b5e 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -717,6 +717,22 @@ static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
call_rcu(&f->rcu, vxlan_fdb_free);
}
+static void vxlan_dst_free(struct rcu_head *head)
+{
+ struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
+
+ dst_cache_destroy(&rd->dst_cache);
+ kfree(rd);
+}
+
+static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
+ struct vxlan_rdst *rd)
+{
+ list_del_rcu(&rd->list);
+ vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
+ call_rcu(&rd->rcu, vxlan_dst_free);
+}
+
static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
union vxlan_addr *ip, __be16 *port, __be32 *vni,
u32 *ifindex)
@@ -847,9 +863,7 @@ static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
* otherwise destroy the fdb entry
*/
if (rd && !list_is_singular(&f->remotes)) {
- list_del_rcu(&rd->list);
- vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
- kfree_rcu(rd, rcu);
+ vxlan_fdb_dst_destroy(vxlan, f, rd);
goto out;
}
--
2.9.4
From f2c8c3f5e1998ae0775e8266b2de8bf7e4701d5f Mon Sep 17 00:00:00 2001
From: Ben Hutchings <ben@decadent.org.uk>
Date: Wed, 31 May 2017 13:15:41 +0100
Subject: [PATCH 03/12] ipv6: xfrm: Handle errors reported by
xfrm6_find_1stfragopt()
[ Upstream commit 6e80ac5cc992ab6256c3dae87f7e57db15e1a58c ]
xfrm6_find_1stfragopt() may now return an error code and we must
not treat it as a length.
Fixes: 2423496af35d ("ipv6: Prevent overrun when parsing v6 header options")
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Acked-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv6/xfrm6_mode_ro.c | 2 ++
net/ipv6/xfrm6_mode_transport.c | 2 ++
2 files changed, 4 insertions(+)
diff --git a/net/ipv6/xfrm6_mode_ro.c b/net/ipv6/xfrm6_mode_ro.c
index 0e01590..07d3657 100644
--- a/net/ipv6/xfrm6_mode_ro.c
+++ b/net/ipv6/xfrm6_mode_ro.c
@@ -47,6 +47,8 @@ static int xfrm6_ro_output(struct xfrm_state *x, struct sk_buff *skb)
iph = ipv6_hdr(skb);
hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
+ if (hdr_len < 0)
+ return hdr_len;
skb_set_mac_header(skb, (prevhdr - x->props.header_len) - skb->data);
skb_set_network_header(skb, -x->props.header_len);
skb->transport_header = skb->network_header + hdr_len;
diff --git a/net/ipv6/xfrm6_mode_transport.c b/net/ipv6/xfrm6_mode_transport.c
index 4e34410..1d3bbe6 100644
--- a/net/ipv6/xfrm6_mode_transport.c
+++ b/net/ipv6/xfrm6_mode_transport.c
@@ -28,6 +28,8 @@ static int xfrm6_transport_output(struct xfrm_state *x, struct sk_buff *skb)
iph = ipv6_hdr(skb);
hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
+ if (hdr_len < 0)
+ return hdr_len;
skb_set_mac_header(skb, (prevhdr - x->props.header_len) - skb->data);
skb_set_network_header(skb, -x->props.header_len);
skb->transport_header = skb->network_header + hdr_len;
--
2.9.4
From 3d3dae542c0a1ddf3d97188a0933eac614d8629b Mon Sep 17 00:00:00 2001
From: Ganesh Goudar <ganeshgr@chelsio.com>
Date: Wed, 31 May 2017 18:26:28 +0530
Subject: [PATCH 04/12] cxgb4: avoid enabling napi twice to the same queue
[ Upstream commit e7519f9926f1d0d11c776eb0475eb098c7760f68 ]
Take uld mutex to avoid race between cxgb_up() and
cxgb4_register_uld() to enable napi for the same uld
queue.
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 19dc9e2..f9c2feb 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -2226,10 +2226,14 @@ static int cxgb_up(struct adapter *adap)
if (err)
goto irq_err;
}
+
+ mutex_lock(&uld_mutex);
enable_rx(adap);
t4_sge_start(adap);
t4_intr_enable(adap);
adap->flags |= FULL_INIT_DONE;
+ mutex_unlock(&uld_mutex);
+
notify_ulds(adap, CXGB4_STATE_UP);
#if IS_ENABLED(CONFIG_IPV6)
update_clip(adap);
--
2.9.4
From 280bc882997a3dcb2fe76fa41ce4284ffc42cc39 Mon Sep 17 00:00:00 2001
From: Yuchung Cheng <ycheng@google.com>
Date: Wed, 31 May 2017 11:21:27 -0700
Subject: [PATCH 05/12] tcp: disallow cwnd undo when switching congestion
control
[ Upstream commit 44abafc4cc094214a99f860f778c48ecb23422fc ]
When the sender switches its congestion control during loss
recovery, if the recovery is spurious then it may incorrectly
revert cwnd and ssthresh to the older values set by a previous
congestion control. Consider a congestion control (like BBR)
that does not use ssthresh and keeps it infinite: the connection
may incorrectly revert cwnd to an infinite value when switching
from BBR to another congestion control.
This patch fixes it by disallowing such cwnd undo operation
upon switching congestion control. Note that undo_marker
is not reset s.t. the packets that were incorrectly marked
lost would be corrected. We only avoid undoing the cwnd in
tcp_undo_cwnd_reduction().
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv4/tcp_cong.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/net/ipv4/tcp_cong.c b/net/ipv4/tcp_cong.c
index baea5df..0cdbea9 100644
--- a/net/ipv4/tcp_cong.c
+++ b/net/ipv4/tcp_cong.c
@@ -179,6 +179,7 @@ void tcp_init_congestion_control(struct sock *sk)
{
const struct inet_connection_sock *icsk = inet_csk(sk);
+ tcp_sk(sk)->prior_ssthresh = 0;
if (icsk->icsk_ca_ops->init)
icsk->icsk_ca_ops->init(sk);
if (tcp_ca_needs_ecn(sk))
--
2.9.4
From c4d47f14f988b930ea50be8646fd71a5bdeba55d Mon Sep 17 00:00:00 2001
From: Mark Bloch <markb@mellanox.com>
Date: Fri, 2 Jun 2017 03:24:08 +0300
Subject: [PATCH 06/12] vxlan: fix use-after-free on deletion
[ Upstream commit a53cb29b0af346af44e4abf13d7e59f807fba690 ]
Adding a vxlan interface to a socket isn't symmetrical, while adding
is done in vxlan_open() the deletion is done in vxlan_dellink().
This can cause a use-after-free error when we close the vxlan
interface before deleting it.
We add vxlan_vs_del_dev() to match vxlan_vs_add_dev() and call
it from vxlan_stop() to match the call from vxlan_open().
Fixes: 56ef9c909b40 ("vxlan: Move socket initialization to within rtnl scope")
Acked-by: Jiri Benc <jbenc@redhat.com>
Tested-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Mark Bloch <markb@mellanox.com>
Acked-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/vxlan.c | 19 +++++++++++++------
1 file changed, 13 insertions(+), 6 deletions(-)
diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index 3331b5e..55c4408 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -59,6 +59,8 @@ static const u8 all_zeros_mac[ETH_ALEN + 2];
static int vxlan_sock_add(struct vxlan_dev *vxlan);
+static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
+
/* per-network namespace private data for this module */
struct vxlan_net {
struct list_head vxlan_list;
@@ -1040,6 +1042,8 @@ static void vxlan_sock_release(struct vxlan_dev *vxlan)
rcu_assign_pointer(vxlan->vn4_sock, NULL);
synchronize_net();
+ vxlan_vs_del_dev(vxlan);
+
if (__vxlan_sock_release_prep(sock4)) {
udp_tunnel_sock_release(sock4->sock);
kfree(sock4);
@@ -2300,6 +2304,15 @@ static void vxlan_cleanup(unsigned long arg)
mod_timer(&vxlan->age_timer, next_timer);
}
+static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
+{
+ struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
+
+ spin_lock(&vn->sock_lock);
+ hlist_del_init_rcu(&vxlan->hlist);
+ spin_unlock(&vn->sock_lock);
+}
+
static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
{
struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
@@ -3070,12 +3083,6 @@ static int vxlan_newlink(struct net *src_net, struct net_device *dev,
static void vxlan_dellink(struct net_device *dev, struct list_head *head)
{
struct vxlan_dev *vxlan = netdev_priv(dev);
- struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
-
- spin_lock(&vn->sock_lock);
- if (!hlist_unhashed(&vxlan->hlist))
- hlist_del_rcu(&vxlan->hlist);
- spin_unlock(&vn->sock_lock);
gro_cells_destroy(&vxlan->gro_cells);
list_del(&vxlan->next);
--
2.9.4
From 8991d8b94264b589598d4f1875fcb461980d710f Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Sun, 4 Jun 2017 21:41:10 -0400
Subject: [PATCH 07/12] ipv6: Fix leak in ipv6_gso_segment().
[ Upstream commit e3e86b5119f81e5e2499bea7ea1ebe8ac6aab789 ]
If ip6_find_1stfragopt() fails and we return an error we have to free
up 'segs' because nobody else is going to.
Fixes: 2423496af35d ("ipv6: Prevent overrun when parsing v6 header options")
Reported-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv6/ip6_offload.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/net/ipv6/ip6_offload.c b/net/ipv6/ip6_offload.c
index 013086b..424fbe1 100644
--- a/net/ipv6/ip6_offload.c
+++ b/net/ipv6/ip6_offload.c
@@ -116,8 +116,10 @@ static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
if (udpfrag) {
int err = ip6_find_1stfragopt(skb, &prevhdr);
- if (err < 0)
+ if (err < 0) {
+ kfree_skb_list(segs);
return ERR_PTR(err);
+ }
fptr = (struct frag_hdr *)((u8 *)ipv6h + err);
fptr->frag_off = htons(offset);
if (skb->next)
--
2.9.4
From 7bf821dd602f00c8afd380e457e704b307e060b6 Mon Sep 17 00:00:00 2001
From: Eric Dumazet <edumazet@google.com>
Date: Sat, 3 Jun 2017 09:29:25 -0700
Subject: [PATCH 08/12] net: ping: do not abuse udp_poll()
[ Upstream commit 77d4b1d36926a9b8387c6b53eeba42bcaaffcea3 ]
Alexander reported various KASAN messages triggered in recent kernels
The problem is that ping sockets should not use udp_poll() in the first
place, and recent changes in UDP stack finally exposed this old bug.
Fixes: c319b4d76b9e ("net: ipv4: add IPPROTO_ICMP socket kind")
Fixes: 6d0bfe226116 ("net: ipv6: Add IPv6 support to the ping socket.")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Sasha Levin <alexander.levin@verizon.com>
Cc: Solar Designer <solar@openwall.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Lorenzo Colitti <lorenzo@google.com>
Acked-By: Lorenzo Colitti <lorenzo@google.com>
Tested-By: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
include/net/ipv6.h | 1 +
net/ipv4/af_inet.c | 2 +-
net/ipv6/ping.c | 2 +-
net/ipv6/raw.c | 2 +-
4 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 7f15f95..91afb4a 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1001,6 +1001,7 @@ int inet6_hash_connect(struct inet_timewait_death_row *death_row,
*/
extern const struct proto_ops inet6_stream_ops;
extern const struct proto_ops inet6_dgram_ops;
+extern const struct proto_ops inet6_sockraw_ops;
struct group_source_req;
struct group_filter;
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 971b947..f60fe82 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1015,7 +1015,7 @@ static struct inet_protosw inetsw_array[] =
.type = SOCK_DGRAM,
.protocol = IPPROTO_ICMP,
.prot = &ping_prot,
- .ops = &inet_dgram_ops,
+ .ops = &inet_sockraw_ops,
.flags = INET_PROTOSW_REUSE,
},
diff --git a/net/ipv6/ping.c b/net/ipv6/ping.c
index 66e2d9d..9828681 100644
--- a/net/ipv6/ping.c
+++ b/net/ipv6/ping.c
@@ -198,7 +198,7 @@ static struct inet_protosw pingv6_protosw = {
.type = SOCK_DGRAM,
.protocol = IPPROTO_ICMPV6,
.prot = &pingv6_prot,
- .ops = &inet6_dgram_ops,
+ .ops = &inet6_sockraw_ops,
.flags = INET_PROTOSW_REUSE,
};
diff --git a/net/ipv6/raw.c b/net/ipv6/raw.c
index 1a2fe5c..71ffa526 100644
--- a/net/ipv6/raw.c
+++ b/net/ipv6/raw.c
@@ -1330,7 +1330,7 @@ void raw6_proc_exit(void)
#endif /* CONFIG_PROC_FS */
/* Same as inet6_dgram_ops, sans udp_poll. */
-static const struct proto_ops inet6_sockraw_ops = {
+const struct proto_ops inet6_sockraw_ops = {
.family = PF_INET6,
.owner = THIS_MODULE,
.release = inet6_release,
--
2.9.4
From 714caad3cd0d850e6aeebcf35fb3d5436c75e9ff Mon Sep 17 00:00:00 2001
From: Richard Haines <richard_c_haines@btinternet.com>
Date: Mon, 5 Jun 2017 16:44:40 +0100
Subject: [PATCH 09/12] net/ipv6: Fix CALIPSO causing GPF with datagram support
[ Upstream commit e3ebdb20fddacded2740a333ff66781e0d28b05c ]
When using CALIPSO with IPPROTO_UDP it is possible to trigger a GPF as the
IP header may have moved.
Also update the payload length after adding the CALIPSO option.
Signed-off-by: Richard Haines <richard_c_haines@btinternet.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv6/calipso.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/net/ipv6/calipso.c b/net/ipv6/calipso.c
index 37ac9de..8d772fe 100644
--- a/net/ipv6/calipso.c
+++ b/net/ipv6/calipso.c
@@ -1319,7 +1319,7 @@ static int calipso_skbuff_setattr(struct sk_buff *skb,
struct ipv6hdr *ip6_hdr;
struct ipv6_opt_hdr *hop;
unsigned char buf[CALIPSO_MAX_BUFFER];
- int len_delta, new_end, pad;
+ int len_delta, new_end, pad, payload;
unsigned int start, end;
ip6_hdr = ipv6_hdr(skb);
@@ -1346,6 +1346,8 @@ static int calipso_skbuff_setattr(struct sk_buff *skb,
if (ret_val < 0)
return ret_val;
+ ip6_hdr = ipv6_hdr(skb); /* Reset as skb_cow() may have moved it */
+
if (len_delta) {
if (len_delta > 0)
skb_push(skb, len_delta);
@@ -1355,6 +1357,8 @@ static int calipso_skbuff_setattr(struct sk_buff *skb,
sizeof(*ip6_hdr) + start);
skb_reset_network_header(skb);
ip6_hdr = ipv6_hdr(skb);
+ payload = ntohs(ip6_hdr->payload_len);
+ ip6_hdr->payload_len = htons(payload + len_delta);
}
hop = (struct ipv6_opt_hdr *)(ip6_hdr + 1);
--
2.9.4
From ecf527f30605e14f5cd7f942264fdd26d5a3a0a7 Mon Sep 17 00:00:00 2001
From: Max Filippov <jcmvbkbc@gmail.com>
Date: Mon, 5 Jun 2017 18:31:16 -0700
Subject: [PATCH 10/12] net: ethoc: enable NAPI before poll may be scheduled
[ Upstream commit d220b942a4b6a0640aee78841608f4aa5e8e185e ]
ethoc_reset enables device interrupts, ethoc_interrupt may schedule a
NAPI poll before NAPI is enabled in the ethoc_open, which results in
device being unable to send or receive anything until it's closed and
reopened. In case the device is flooded with ingress packets it may be
unable to recover at all.
Move napi_enable above ethoc_reset in the ethoc_open to fix that.
Fixes: a1702857724f ("net: Add support for the OpenCores 10/100 Mbps Ethernet MAC.")
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Tobias Klauser <tklauser@distanz.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/ethoc.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/ethoc.c b/drivers/net/ethernet/ethoc.c
index c044667..e31199f 100644
--- a/drivers/net/ethernet/ethoc.c
+++ b/drivers/net/ethernet/ethoc.c
@@ -710,6 +710,8 @@ static int ethoc_open(struct net_device *dev)
if (ret)
return ret;
+ napi_enable(&priv->napi);
+
ethoc_init_ring(priv, dev->mem_start);
ethoc_reset(priv);
@@ -722,7 +724,6 @@ static int ethoc_open(struct net_device *dev)
}
phy_start(dev->phydev);
- napi_enable(&priv->napi);
if (netif_msg_ifup(priv)) {
dev_info(&dev->dev, "I/O: %08lx Memory: %08lx-%08lx\n",
--
2.9.4
From ed54b5bf489d59d077de2f16f94298c7cdafb6c8 Mon Sep 17 00:00:00 2001
From: Niklas Cassel <niklas.cassel@axis.com>
Date: Tue, 6 Jun 2017 09:25:00 +0200
Subject: [PATCH 11/12] net: stmmac: fix completely hung TX when using TSO
[ Upstream commit 426849e6611f2092553f8d53372ae310818a6292 ]
stmmac_tso_allocator can fail to set the Last Descriptor bit
on a descriptor that actually was the last descriptor.
This happens when the buffer of the last descriptor ends
up having a size of exactly TSO_MAX_BUFF_SIZE.
When the IP eventually reaches the next last descriptor,
which actually has the bit set, the DMA will hang.
When the DMA hangs, we get a tx timeout, however,
since stmmac does not do a complete reset of the IP
in stmmac_tx_timeout, we end up in a state with
completely hung TX.
Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Acked-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index b2893fb..ef6bff8 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1953,7 +1953,7 @@ static void stmmac_tso_allocator(struct stmmac_priv *priv, unsigned int des,
priv->hw->desc->prepare_tso_tx_desc(desc, 0, buff_size,
0, 1,
- (last_segment) && (buff_size < TSO_MAX_BUFF_SIZE),
+ (last_segment) && (tmp_len <= TSO_MAX_BUFF_SIZE),
0, 0);
tmp_len -= TSO_MAX_BUFF_SIZE;
--
2.9.4
From 1734dbf751a60873b504f248ddb47a79958d61e2 Mon Sep 17 00:00:00 2001
From: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Date: Thu, 1 Jun 2017 18:07:55 +0300
Subject: [PATCH 12/12] net: bridge: start hello timer only if device is up
[ Upstream commit aeb073241fe7a2b932e04e20c60e47718332877f ]
When the transition of NO_STP -> KERNEL_STP was fixed by always calling
mod_timer in br_stp_start, it introduced a new regression which causes
the timer to be armed even when the bridge is down, and since we stop
the timers in its ndo_stop() function, they never get disabled if the
device is destroyed before it's upped.
To reproduce:
$ while :; do ip l add br0 type bridge hello_time 100; brctl stp br0 on;
ip l del br0; done;
CC: Xin Long <lucien.xin@gmail.com>
CC: Ivan Vecera <cera@cera.cz>
CC: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reported-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Fixes: 6d18c732b95c ("bridge: start hello_timer when enabling KERNEL_STP in br_stp_start")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/bridge/br_stp_if.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c
index 5a782f5..16b5aa9 100644
--- a/net/bridge/br_stp_if.c
+++ b/net/bridge/br_stp_if.c
@@ -185,7 +185,8 @@ static void br_stp_start(struct net_bridge *br)
br_debug(br, "using kernel STP\n");
/* To start timers on any ports left in blocking */
- mod_timer(&br->hello_timer, jiffies + br->hello_time);
+ if (br->dev->flags & IFF_UP)
+ mod_timer(&br->hello_timer, jiffies + br->hello_time);
br_port_state_selection(br);
}
--
2.9.4
[-- Attachment #3: net_411.mbox --]
[-- Type: Application/Octet-Stream, Size: 36791 bytes --]
From b1ab18f744ae73510116511fa61a27a2a3efadb4 Mon Sep 17 00:00:00 2001
From: "Mintz, Yuval" <Yuval.Mintz@cavium.com>
Date: Thu, 1 Jun 2017 15:57:56 +0300
Subject: [PATCH 01/19] bnx2x: Fix Multi-Cos
[ Upstream commit 3968d38917eb9bd0cd391265f6c9c538d9b33ffa ]
Apparently multi-cos isn't working for bnx2x quite some time -
driver implements ndo_select_queue() to allow queue-selection
for FCoE, but the regular L2 flow would cause it to modulo the
fallback's result by the number of queues.
The fallback would return a queue matching the needed tc
[via __skb_tx_hash()], but since the modulo is by the number of TSS
queues where number of TCs is not accounted, transmission would always
be done by a queue configured into using TC0.
Fixes: ada7c19e6d27 ("bnx2x: use XPS if possible for bnx2x_select_queue instead of pure hash")
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
index 9e8c061..c2f9a1f 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
@@ -1926,7 +1926,7 @@ u16 bnx2x_select_queue(struct net_device *dev, struct sk_buff *skb,
}
/* select a non-FCoE queue */
- return fallback(dev, skb) % BNX2X_NUM_ETH_QUEUES(bp);
+ return fallback(dev, skb) % (BNX2X_NUM_ETH_QUEUES(bp) * bp->max_cos);
}
void bnx2x_set_num_queues(struct bnx2x *bp)
--
2.9.4
From 30d343e9f1788c3b6756fb1c697f697f76fe81ff Mon Sep 17 00:00:00 2001
From: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Date: Thu, 1 Jun 2017 18:07:55 +0300
Subject: [PATCH 02/19] net: bridge: start hello timer only if device is up
[ Upstream commit aeb073241fe7a2b932e04e20c60e47718332877f ]
When the transition of NO_STP -> KERNEL_STP was fixed by always calling
mod_timer in br_stp_start, it introduced a new regression which causes
the timer to be armed even when the bridge is down, and since we stop
the timers in its ndo_stop() function, they never get disabled if the
device is destroyed before it's upped.
To reproduce:
$ while :; do ip l add br0 type bridge hello_time 100; brctl stp br0 on;
ip l del br0; done;
CC: Xin Long <lucien.xin@gmail.com>
CC: Ivan Vecera <cera@cera.cz>
CC: Sebastian Ott <sebott@linux.vnet.ibm.com>
Reported-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Fixes: 6d18c732b95c ("bridge: start hello_timer when enabling KERNEL_STP in br_stp_start")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/bridge/br_stp_if.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c
index 0db8102..6f12a52 100644
--- a/net/bridge/br_stp_if.c
+++ b/net/bridge/br_stp_if.c
@@ -179,7 +179,8 @@ static void br_stp_start(struct net_bridge *br)
br_debug(br, "using kernel STP\n");
/* To start timers on any ports left in blocking */
- mod_timer(&br->hello_timer, jiffies + br->hello_time);
+ if (br->dev->flags & IFF_UP)
+ mod_timer(&br->hello_timer, jiffies + br->hello_time);
br_port_state_selection(br);
}
--
2.9.4
From 4c383e3582a80e9eaaa044f6a5684814789e1f8e Mon Sep 17 00:00:00 2001
From: Lance Richardson <lrichard@redhat.com>
Date: Mon, 29 May 2017 13:25:57 -0400
Subject: [PATCH 03/19] vxlan: eliminate cached dst leak
[ Upstream commit 35cf2845563c1aaa01d27bd34d64795c4ae72700 ]
After commit 0c1d70af924b ("net: use dst_cache for vxlan device"),
cached dst entries could be leaked when more than one remote was
present for a given vxlan_fdb entry, causing subsequent netns
operations to block indefinitely and "unregister_netdevice: waiting
for lo to become free." messages to appear in the kernel log.
Fix by properly releasing cached dst and freeing resources in this
case.
Fixes: 0c1d70af924b ("net: use dst_cache for vxlan device")
Signed-off-by: Lance Richardson <lrichard@redhat.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/vxlan.c | 20 +++++++++++++++++---
1 file changed, 17 insertions(+), 3 deletions(-)
diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index bdb6ae1..03122a0 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -740,6 +740,22 @@ static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
call_rcu(&f->rcu, vxlan_fdb_free);
}
+static void vxlan_dst_free(struct rcu_head *head)
+{
+ struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
+
+ dst_cache_destroy(&rd->dst_cache);
+ kfree(rd);
+}
+
+static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
+ struct vxlan_rdst *rd)
+{
+ list_del_rcu(&rd->list);
+ vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
+ call_rcu(&rd->rcu, vxlan_dst_free);
+}
+
static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
__be32 *vni, u32 *ifindex)
@@ -864,9 +880,7 @@ static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
* otherwise destroy the fdb entry
*/
if (rd && !list_is_singular(&f->remotes)) {
- list_del_rcu(&rd->list);
- vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
- kfree_rcu(rd, rcu);
+ vxlan_fdb_dst_destroy(vxlan, f, rd);
goto out;
}
--
2.9.4
From 284390398e73531725afd738e62fecb84028d0d3 Mon Sep 17 00:00:00 2001
From: Florian Fainelli <f.fainelli@gmail.com>
Date: Thu, 1 Jun 2017 18:02:39 -0700
Subject: [PATCH 04/19] net: systemport: Fix missing Wake-on-LAN interrupt for
SYSTEMPORT Lite
[ Upstream commit d31353cd753c443ace5723d6878a39f393a0c136 ]
On SYSTEMPORT Lite, since we have the main interrupt source in the first
cell, the second cell is the Wake-on-LAN interrupt, yet the code was not
properly updated to fetch the second cell, and instead looked at the
third and non-existing cell for Wake-on-LAN.
Fixes: 44a4524c54af ("net: systemport: Add support for SYSTEMPORT Lite")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/broadcom/bcmsysport.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/bcmsysport.c b/drivers/net/ethernet/broadcom/bcmsysport.c
index a68d4889..a96916a 100644
--- a/drivers/net/ethernet/broadcom/bcmsysport.c
+++ b/drivers/net/ethernet/broadcom/bcmsysport.c
@@ -1968,9 +1968,12 @@ static int bcm_sysport_probe(struct platform_device *pdev)
priv->num_rx_desc_words = params->num_rx_desc_words;
priv->irq0 = platform_get_irq(pdev, 0);
- if (!priv->is_lite)
+ if (!priv->is_lite) {
priv->irq1 = platform_get_irq(pdev, 1);
- priv->wol_irq = platform_get_irq(pdev, 2);
+ priv->wol_irq = platform_get_irq(pdev, 2);
+ } else {
+ priv->wol_irq = platform_get_irq(pdev, 1);
+ }
if (priv->irq0 <= 0 || (priv->irq1 <= 0 && !priv->is_lite)) {
dev_err(&pdev->dev, "invalid interrupts\n");
ret = -EINVAL;
--
2.9.4
From 804e0f08fc2485c153e9c57f4f135759caf09374 Mon Sep 17 00:00:00 2001
From: Ben Hutchings <ben@decadent.org.uk>
Date: Wed, 31 May 2017 13:15:41 +0100
Subject: [PATCH 05/19] ipv6: xfrm: Handle errors reported by
xfrm6_find_1stfragopt()
[ Upstream commit 6e80ac5cc992ab6256c3dae87f7e57db15e1a58c ]
xfrm6_find_1stfragopt() may now return an error code and we must
not treat it as a length.
Fixes: 2423496af35d ("ipv6: Prevent overrun when parsing v6 header options")
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Acked-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv6/xfrm6_mode_ro.c | 2 ++
net/ipv6/xfrm6_mode_transport.c | 2 ++
2 files changed, 4 insertions(+)
diff --git a/net/ipv6/xfrm6_mode_ro.c b/net/ipv6/xfrm6_mode_ro.c
index 0e01590..07d3657 100644
--- a/net/ipv6/xfrm6_mode_ro.c
+++ b/net/ipv6/xfrm6_mode_ro.c
@@ -47,6 +47,8 @@ static int xfrm6_ro_output(struct xfrm_state *x, struct sk_buff *skb)
iph = ipv6_hdr(skb);
hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
+ if (hdr_len < 0)
+ return hdr_len;
skb_set_mac_header(skb, (prevhdr - x->props.header_len) - skb->data);
skb_set_network_header(skb, -x->props.header_len);
skb->transport_header = skb->network_header + hdr_len;
diff --git a/net/ipv6/xfrm6_mode_transport.c b/net/ipv6/xfrm6_mode_transport.c
index 4439ee4..5e30410 100644
--- a/net/ipv6/xfrm6_mode_transport.c
+++ b/net/ipv6/xfrm6_mode_transport.c
@@ -28,6 +28,8 @@ static int xfrm6_transport_output(struct xfrm_state *x, struct sk_buff *skb)
iph = ipv6_hdr(skb);
hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
+ if (hdr_len < 0)
+ return hdr_len;
skb_set_mac_header(skb, (prevhdr - x->props.header_len) - skb->data);
skb_set_network_header(skb, -x->props.header_len);
skb->transport_header = skb->network_header + hdr_len;
--
2.9.4
From 39fb52a397778458b1b75414b7f9a6ee9a8fba1f Mon Sep 17 00:00:00 2001
From: Ganesh Goudar <ganeshgr@chelsio.com>
Date: Wed, 31 May 2017 18:26:28 +0530
Subject: [PATCH 06/19] cxgb4: avoid enabling napi twice to the same queue
[ Upstream commit e7519f9926f1d0d11c776eb0475eb098c7760f68 ]
Take uld mutex to avoid race between cxgb_up() and
cxgb4_register_uld() to enable napi for the same uld
queue.
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index afb0967..012194b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -2217,10 +2217,14 @@ static int cxgb_up(struct adapter *adap)
if (err)
goto irq_err;
}
+
+ mutex_lock(&uld_mutex);
enable_rx(adap);
t4_sge_start(adap);
t4_intr_enable(adap);
adap->flags |= FULL_INIT_DONE;
+ mutex_unlock(&uld_mutex);
+
notify_ulds(adap, CXGB4_STATE_UP);
#if IS_ENABLED(CONFIG_IPV6)
update_clip(adap);
--
2.9.4
From 91674b884656f9776d4609ec0ef862500e80275f Mon Sep 17 00:00:00 2001
From: Yuchung Cheng <ycheng@google.com>
Date: Wed, 31 May 2017 11:21:27 -0700
Subject: [PATCH 07/19] tcp: disallow cwnd undo when switching congestion
control
[ Upstream commit 44abafc4cc094214a99f860f778c48ecb23422fc ]
When the sender switches its congestion control during loss
recovery, if the recovery is spurious then it may incorrectly
revert cwnd and ssthresh to the older values set by a previous
congestion control. Consider a congestion control (like BBR)
that does not use ssthresh and keeps it infinite: the connection
may incorrectly revert cwnd to an infinite value when switching
from BBR to another congestion control.
This patch fixes it by disallowing such cwnd undo operation
upon switching congestion control. Note that undo_marker
is not reset s.t. the packets that were incorrectly marked
lost would be corrected. We only avoid undoing the cwnd in
tcp_undo_cwnd_reduction().
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv4/tcp_cong.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/net/ipv4/tcp_cong.c b/net/ipv4/tcp_cong.c
index 6e3c512..324c9bc 100644
--- a/net/ipv4/tcp_cong.c
+++ b/net/ipv4/tcp_cong.c
@@ -180,6 +180,7 @@ void tcp_init_congestion_control(struct sock *sk)
{
const struct inet_connection_sock *icsk = inet_csk(sk);
+ tcp_sk(sk)->prior_ssthresh = 0;
if (icsk->icsk_ca_ops->init)
icsk->icsk_ca_ops->init(sk);
if (tcp_ca_needs_ecn(sk))
--
2.9.4
From d6cf264cf40e84fcc6320f3c90a627ac7550548d Mon Sep 17 00:00:00 2001
From: Mark Bloch <markb@mellanox.com>
Date: Fri, 2 Jun 2017 03:24:08 +0300
Subject: [PATCH 08/19] vxlan: fix use-after-free on deletion
[ Upstream commit a53cb29b0af346af44e4abf13d7e59f807fba690 ]
Adding a vxlan interface to a socket isn't symmetrical, while adding
is done in vxlan_open() the deletion is done in vxlan_dellink().
This can cause a use-after-free error when we close the vxlan
interface before deleting it.
We add vxlan_vs_del_dev() to match vxlan_vs_add_dev() and call
it from vxlan_stop() to match the call from vxlan_open().
Fixes: 56ef9c909b40 ("vxlan: Move socket initialization to within rtnl scope")
Acked-by: Jiri Benc <jbenc@redhat.com>
Tested-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Mark Bloch <markb@mellanox.com>
Acked-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/vxlan.c | 19 +++++++++++++------
1 file changed, 13 insertions(+), 6 deletions(-)
diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index 03122a0..70dbd5a 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -59,6 +59,8 @@ static const u8 all_zeros_mac[ETH_ALEN + 2];
static int vxlan_sock_add(struct vxlan_dev *vxlan);
+static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
+
/* per-network namespace private data for this module */
struct vxlan_net {
struct list_head vxlan_list;
@@ -1081,6 +1083,8 @@ static void vxlan_sock_release(struct vxlan_dev *vxlan)
rcu_assign_pointer(vxlan->vn4_sock, NULL);
synchronize_net();
+ vxlan_vs_del_dev(vxlan);
+
if (__vxlan_sock_release_prep(sock4)) {
udp_tunnel_sock_release(sock4->sock);
kfree(sock4);
@@ -2352,6 +2356,15 @@ static void vxlan_cleanup(unsigned long arg)
mod_timer(&vxlan->age_timer, next_timer);
}
+static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
+{
+ struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
+
+ spin_lock(&vn->sock_lock);
+ hlist_del_init_rcu(&vxlan->hlist);
+ spin_unlock(&vn->sock_lock);
+}
+
static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
{
struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
@@ -3289,15 +3302,9 @@ static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
static void vxlan_dellink(struct net_device *dev, struct list_head *head)
{
struct vxlan_dev *vxlan = netdev_priv(dev);
- struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
vxlan_flush(vxlan, true);
- spin_lock(&vn->sock_lock);
- if (!hlist_unhashed(&vxlan->hlist))
- hlist_del_rcu(&vxlan->hlist);
- spin_unlock(&vn->sock_lock);
-
gro_cells_destroy(&vxlan->gro_cells);
list_del(&vxlan->next);
unregister_netdevice_queue(dev, head);
--
2.9.4
From 9191f7cc11ae73198fe3dc753af27ae49a6883ea Mon Sep 17 00:00:00 2001
From: Liam McBirnie <mcbirnie.l@gmail.com>
Date: Thu, 1 Jun 2017 15:36:01 +1000
Subject: [PATCH 09/19] ip6_tunnel: fix traffic class routing for tunnels
[ Upstream commit 5f733ee68f9a4df94775299ac6a7ab260704f6ed ]
ip6_route_output() requires that the flowlabel contains the traffic
class for policy routing.
Commit 0e9a709560db ("ip6_tunnel, ip6_gre: fix setting of DSCP on
encapsulated packets") removed the code which previously added the
traffic class to the flowlabel.
The traffic class is added here because only route lookup needs the
flowlabel to contain the traffic class.
Fixes: 0e9a709560db ("ip6_tunnel, ip6_gre: fix setting of DSCP on encapsulated packets")
Signed-off-by: Liam McBirnie <liam.mcbirnie@boeing.com>
Acked-by: Peter Dawson <peter.a.dawson@boeing.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv6/ip6_tunnel.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
index 15ff339..e2d7867 100644
--- a/net/ipv6/ip6_tunnel.c
+++ b/net/ipv6/ip6_tunnel.c
@@ -1095,6 +1095,9 @@ int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
if (!dst) {
route_lookup:
+ /* add dsfield to flowlabel for route lookup */
+ fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
+
dst = ip6_route_output(net, NULL, fl6);
if (dst->error)
--
2.9.4
From 7ae68a9757d705a590ceb7e92cc1bf592cd6858b Mon Sep 17 00:00:00 2001
From: Soheil Hassas Yeganeh <soheil@google.com>
Date: Fri, 2 Jun 2017 12:38:22 -0400
Subject: [PATCH 10/19] sock: reset sk_err when the error queue is empty
[ Upstream commit 38b257938ac6655d0d6333743303231b9c465ec1 ]
Prior to f5f99309fa74 (sock: do not set sk_err in
sock_dequeue_err_skb), sk_err was reset to the error of
the skb on the head of the error queue.
Applications, most notably ping, are relying on this
behavior to reset sk_err for ICMP packets.
Set sk_err to the ICMP error when there is an ICMP packet
at the head of the error queue.
Fixes: f5f99309fa74 (sock: do not set sk_err in sock_dequeue_err_skb)
Reported-by: Cyril Hrubis <chrubis@suse.cz>
Tested-by: Cyril Hrubis <chrubis@suse.cz>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/core/skbuff.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index f1d0459..ac5059a 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -3755,8 +3755,11 @@ struct sk_buff *sock_dequeue_err_skb(struct sock *sk)
spin_lock_irqsave(&q->lock, flags);
skb = __skb_dequeue(q);
- if (skb && (skb_next = skb_peek(q)))
+ if (skb && (skb_next = skb_peek(q))) {
icmp_next = is_icmp_err_skb(skb_next);
+ if (icmp_next)
+ sk->sk_err = SKB_EXT_ERR(skb_next)->ee.ee_origin;
+ }
spin_unlock_irqrestore(&q->lock, flags);
if (is_icmp_err_skb(skb) && !icmp_next)
--
2.9.4
From 240b1a6df998ed1df98667c20c4daee445a21cf5 Mon Sep 17 00:00:00 2001
From: Eric Garver <e@erig.me>
Date: Fri, 2 Jun 2017 14:54:10 -0400
Subject: [PATCH 11/19] geneve: fix needed_headroom and max_mtu for
collect_metadata
[ Upstream commit 9a1c44d989bff4c992b8b9a112d9fda275ea5515 ]
Since commit 9b4437a5b870 ("geneve: Unify LWT and netdev handling.")
when using COLLECT_METADATA geneve devices are created with too small of
a needed_headroom and too large of a max_mtu. This is because
ip_tunnel_info_af() is not valid with the device level info when using
COLLECT_METADATA and we mistakenly fall into the IPv4 case.
For COLLECT_METADATA, always use the worst case of ipv6 since both
sockets are created.
Fixes: 9b4437a5b870 ("geneve: Unify LWT and netdev handling.")
Signed-off-by: Eric Garver <e@erig.me>
Acked-by: Pravin B Shelar <pshelar@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/geneve.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
index 959fd12..6ebb0f5 100644
--- a/drivers/net/geneve.c
+++ b/drivers/net/geneve.c
@@ -1133,7 +1133,7 @@ static int geneve_configure(struct net *net, struct net_device *dev,
/* make enough headroom for basic scenario */
encap_len = GENEVE_BASE_HLEN + ETH_HLEN;
- if (ip_tunnel_info_af(info) == AF_INET) {
+ if (!metadata && ip_tunnel_info_af(info) == AF_INET) {
encap_len += sizeof(struct iphdr);
dev->max_mtu -= sizeof(struct iphdr);
} else {
--
2.9.4
From 6a76af88ba9e366609b6e30c9114ab99d833b85e Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Sun, 4 Jun 2017 21:41:10 -0400
Subject: [PATCH 12/19] ipv6: Fix leak in ipv6_gso_segment().
[ Upstream commit e3e86b5119f81e5e2499bea7ea1ebe8ac6aab789 ]
If ip6_find_1stfragopt() fails and we return an error we have to free
up 'segs' because nobody else is going to.
Fixes: 2423496af35d ("ipv6: Prevent overrun when parsing v6 header options")
Reported-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv6/ip6_offload.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/net/ipv6/ip6_offload.c b/net/ipv6/ip6_offload.c
index 280268f..cdb3728 100644
--- a/net/ipv6/ip6_offload.c
+++ b/net/ipv6/ip6_offload.c
@@ -116,8 +116,10 @@ static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
if (udpfrag) {
int err = ip6_find_1stfragopt(skb, &prevhdr);
- if (err < 0)
+ if (err < 0) {
+ kfree_skb_list(segs);
return ERR_PTR(err);
+ }
fptr = (struct frag_hdr *)((u8 *)ipv6h + err);
fptr->frag_off = htons(offset);
if (skb->next)
--
2.9.4
From 9ca9a50c58e360633ce3288ce9895558bea97f31 Mon Sep 17 00:00:00 2001
From: Florian Fainelli <f.fainelli@gmail.com>
Date: Fri, 2 Jun 2017 22:05:23 -0700
Subject: [PATCH 13/19] net: dsa: Fix stale cpu_switch reference after unbind
then bind
[ Upstream commit b07ac9894644202614ca87c69f3f45e424a82fef ]
Commit 9520ed8fb841 ("net: dsa: use cpu_switch instead of ds[0]")
replaced the use of dst->ds[0] with dst->cpu_switch since that is
functionally equivalent, however, we can now run into an use after free
scenario after unbinding then rebinding the switch driver.
The use after free happens because we do correctly initialize
dst->cpu_switch the first time we probe in dsa_cpu_parse(), then we
unbind the driver: dsa_dst_unapply() is called, and we rebind again.
dst->cpu_switch now points to a freed "ds" structure, and so when we
finally dereference it in dsa_cpu_port_ethtool_setup(), we oops.
To fix this, simply set dst->cpu_switch to NULL in dsa_dst_unapply()
which guarantees that we always correctly re-assign dst->cpu_switch in
dsa_cpu_parse().
Fixes: 9520ed8fb841 ("net: dsa: use cpu_switch instead of ds[0]")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/dsa/dsa2.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/net/dsa/dsa2.c b/net/dsa/dsa2.c
index 737be64..ffaa4fb 100644
--- a/net/dsa/dsa2.c
+++ b/net/dsa/dsa2.c
@@ -440,8 +440,10 @@ static void dsa_dst_unapply(struct dsa_switch_tree *dst)
dsa_ds_unapply(dst, ds);
}
- if (dst->cpu_switch)
+ if (dst->cpu_switch) {
dsa_cpu_port_ethtool_restore(dst->cpu_switch);
+ dst->cpu_switch = NULL;
+ }
pr_info("DSA: tree %d unapplied\n", dst->tree);
dst->applied = false;
--
2.9.4
From 5e015e16536a62642f9b4100270c3a11f3046c9e Mon Sep 17 00:00:00 2001
From: Eric Dumazet <edumazet@google.com>
Date: Sat, 3 Jun 2017 09:29:25 -0700
Subject: [PATCH 14/19] net: ping: do not abuse udp_poll()
[ Upstream commit 77d4b1d36926a9b8387c6b53eeba42bcaaffcea3 ]
Alexander reported various KASAN messages triggered in recent kernels
The problem is that ping sockets should not use udp_poll() in the first
place, and recent changes in UDP stack finally exposed this old bug.
Fixes: c319b4d76b9e ("net: ipv4: add IPPROTO_ICMP socket kind")
Fixes: 6d0bfe226116 ("net: ipv6: Add IPv6 support to the ping socket.")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Sasha Levin <alexander.levin@verizon.com>
Cc: Solar Designer <solar@openwall.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Lorenzo Colitti <lorenzo@google.com>
Acked-By: Lorenzo Colitti <lorenzo@google.com>
Tested-By: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
include/net/ipv6.h | 1 +
net/ipv4/af_inet.c | 2 +-
net/ipv6/ping.c | 2 +-
net/ipv6/raw.c | 2 +-
4 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index dbf0abb..3e505bb 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1007,6 +1007,7 @@ int inet6_hash_connect(struct inet_timewait_death_row *death_row,
*/
extern const struct proto_ops inet6_stream_ops;
extern const struct proto_ops inet6_dgram_ops;
+extern const struct proto_ops inet6_sockraw_ops;
struct group_source_req;
struct group_filter;
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 13a9a32..b9c1bc5 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1043,7 +1043,7 @@ static struct inet_protosw inetsw_array[] =
.type = SOCK_DGRAM,
.protocol = IPPROTO_ICMP,
.prot = &ping_prot,
- .ops = &inet_dgram_ops,
+ .ops = &inet_sockraw_ops,
.flags = INET_PROTOSW_REUSE,
},
diff --git a/net/ipv6/ping.c b/net/ipv6/ping.c
index 9b522fa..ac826dd 100644
--- a/net/ipv6/ping.c
+++ b/net/ipv6/ping.c
@@ -192,7 +192,7 @@ static struct inet_protosw pingv6_protosw = {
.type = SOCK_DGRAM,
.protocol = IPPROTO_ICMPV6,
.prot = &pingv6_prot,
- .ops = &inet6_dgram_ops,
+ .ops = &inet6_sockraw_ops,
.flags = INET_PROTOSW_REUSE,
};
diff --git a/net/ipv6/raw.c b/net/ipv6/raw.c
index 1f992d9..60be012 100644
--- a/net/ipv6/raw.c
+++ b/net/ipv6/raw.c
@@ -1338,7 +1338,7 @@ void raw6_proc_exit(void)
#endif /* CONFIG_PROC_FS */
/* Same as inet6_dgram_ops, sans udp_poll. */
-static const struct proto_ops inet6_sockraw_ops = {
+const struct proto_ops inet6_sockraw_ops = {
.family = PF_INET6,
.owner = THIS_MODULE,
.release = inet6_release,
--
2.9.4
From 253a9bd8ef54e60e23d26f39eff8fc3d1b4c9aa5 Mon Sep 17 00:00:00 2001
From: Richard Haines <richard_c_haines@btinternet.com>
Date: Mon, 5 Jun 2017 16:44:40 +0100
Subject: [PATCH 15/19] net/ipv6: Fix CALIPSO causing GPF with datagram support
[ Upstream commit e3ebdb20fddacded2740a333ff66781e0d28b05c ]
When using CALIPSO with IPPROTO_UDP it is possible to trigger a GPF as the
IP header may have moved.
Also update the payload length after adding the CALIPSO option.
Signed-off-by: Richard Haines <richard_c_haines@btinternet.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/ipv6/calipso.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/net/ipv6/calipso.c b/net/ipv6/calipso.c
index 37ac9de..8d772fe 100644
--- a/net/ipv6/calipso.c
+++ b/net/ipv6/calipso.c
@@ -1319,7 +1319,7 @@ static int calipso_skbuff_setattr(struct sk_buff *skb,
struct ipv6hdr *ip6_hdr;
struct ipv6_opt_hdr *hop;
unsigned char buf[CALIPSO_MAX_BUFFER];
- int len_delta, new_end, pad;
+ int len_delta, new_end, pad, payload;
unsigned int start, end;
ip6_hdr = ipv6_hdr(skb);
@@ -1346,6 +1346,8 @@ static int calipso_skbuff_setattr(struct sk_buff *skb,
if (ret_val < 0)
return ret_val;
+ ip6_hdr = ipv6_hdr(skb); /* Reset as skb_cow() may have moved it */
+
if (len_delta) {
if (len_delta > 0)
skb_push(skb, len_delta);
@@ -1355,6 +1357,8 @@ static int calipso_skbuff_setattr(struct sk_buff *skb,
sizeof(*ip6_hdr) + start);
skb_reset_network_header(skb);
ip6_hdr = ipv6_hdr(skb);
+ payload = ntohs(ip6_hdr->payload_len);
+ ip6_hdr->payload_len = htons(payload + len_delta);
}
hop = (struct ipv6_opt_hdr *)(ip6_hdr + 1);
--
2.9.4
From 30ba7d79bdc5879550d45d72a5c824f22240cf4d Mon Sep 17 00:00:00 2001
From: Eugeniu Rosca <erosca@de.adit-jv.com>
Date: Tue, 6 Jun 2017 00:08:10 +0200
Subject: [PATCH 16/19] ravb: Fix use-after-free on `ifconfig eth0 down`
[ Upstream commit 79514ef670e9e575a1fe36922268c439d0f0ca8a ]
Commit a47b70ea86bd ("ravb: unmap descriptors when freeing rings") has
introduced the issue seen in [1] reproduced on H3ULCB board.
Fix this by relocating the RX skb ringbuffer free operation, so that
swiotlb page unmapping can be done first. Freeing of aligned TX buffers
is not relevant to the issue seen in [1]. Still, reposition TX free
calls as well, to have all kfree() operations performed consistently
_after_ dma_unmap_*()/dma_free_*().
[1] Console screenshot with the problem reproduced:
salvator-x login: root
root@salvator-x:~# ifconfig eth0 up
Micrel KSZ9031 Gigabit PHY e6800000.ethernet-ffffffff:00: \
attached PHY driver [Micrel KSZ9031 Gigabit PHY] \
(mii_bus:phy_addr=e6800000.ethernet-ffffffff:00, irq=235)
IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
root@salvator-x:~#
root@salvator-x:~# ifconfig eth0 down
==================================================================
BUG: KASAN: use-after-free in swiotlb_tbl_unmap_single+0xc4/0x35c
Write of size 1538 at addr ffff8006d884f780 by task ifconfig/1649
CPU: 0 PID: 1649 Comm: ifconfig Not tainted 4.12.0-rc4-00004-g112eb07287d1 #32
Hardware name: Renesas H3ULCB board based on r8a7795 (DT)
Call trace:
[<ffff20000808f11c>] dump_backtrace+0x0/0x3a4
[<ffff20000808f4d4>] show_stack+0x14/0x1c
[<ffff20000865970c>] dump_stack+0xf8/0x150
[<ffff20000831f8b0>] print_address_description+0x7c/0x330
[<ffff200008320010>] kasan_report+0x2e0/0x2f4
[<ffff20000831eac0>] check_memory_region+0x20/0x14c
[<ffff20000831f054>] memcpy+0x48/0x68
[<ffff20000869ed50>] swiotlb_tbl_unmap_single+0xc4/0x35c
[<ffff20000869fcf4>] unmap_single+0x90/0xa4
[<ffff20000869fd14>] swiotlb_unmap_page+0xc/0x14
[<ffff2000080a2974>] __swiotlb_unmap_page+0xcc/0xe4
[<ffff2000088acdb8>] ravb_ring_free+0x514/0x870
[<ffff2000088b25dc>] ravb_close+0x288/0x36c
[<ffff200008aaf8c4>] __dev_close_many+0x14c/0x174
[<ffff200008aaf9b4>] __dev_close+0xc8/0x144
[<ffff200008ac2100>] __dev_change_flags+0xd8/0x194
[<ffff200008ac221c>] dev_change_flags+0x60/0xb0
[<ffff200008ba2dec>] devinet_ioctl+0x484/0x9d4
[<ffff200008ba7b78>] inet_ioctl+0x190/0x194
[<ffff200008a78c44>] sock_do_ioctl+0x78/0xa8
[<ffff200008a7a128>] sock_ioctl+0x110/0x3c4
[<ffff200008365a70>] vfs_ioctl+0x90/0xa0
[<ffff200008365dbc>] do_vfs_ioctl+0x148/0xc38
[<ffff2000083668f0>] SyS_ioctl+0x44/0x74
[<ffff200008083770>] el0_svc_naked+0x24/0x28
The buggy address belongs to the page:
page:ffff7e001b6213c0 count:0 mapcount:0 mapping: (null) index:0x0
flags: 0x4000000000000000()
raw: 4000000000000000 0000000000000000 0000000000000000 00000000ffffffff
raw: 0000000000000000 ffff7e001b6213e0 0000000000000000 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
ffff8006d884f680: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ffff8006d884f700: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
>ffff8006d884f780: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
^
ffff8006d884f800: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ffff8006d884f880: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
==================================================================
Disabling lock debugging due to kernel taint
root@salvator-x:~#
Fixes: a47b70ea86bd ("ravb: unmap descriptors when freeing rings")
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Acked-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/renesas/ravb_main.c | 24 ++++++++++++------------
1 file changed, 12 insertions(+), 12 deletions(-)
diff --git a/drivers/net/ethernet/renesas/ravb_main.c b/drivers/net/ethernet/renesas/ravb_main.c
index 3cd7989..784782d 100644
--- a/drivers/net/ethernet/renesas/ravb_main.c
+++ b/drivers/net/ethernet/renesas/ravb_main.c
@@ -230,18 +230,6 @@ static void ravb_ring_free(struct net_device *ndev, int q)
int ring_size;
int i;
- /* Free RX skb ringbuffer */
- if (priv->rx_skb[q]) {
- for (i = 0; i < priv->num_rx_ring[q]; i++)
- dev_kfree_skb(priv->rx_skb[q][i]);
- }
- kfree(priv->rx_skb[q]);
- priv->rx_skb[q] = NULL;
-
- /* Free aligned TX buffers */
- kfree(priv->tx_align[q]);
- priv->tx_align[q] = NULL;
-
if (priv->rx_ring[q]) {
for (i = 0; i < priv->num_rx_ring[q]; i++) {
struct ravb_ex_rx_desc *desc = &priv->rx_ring[q][i];
@@ -270,6 +258,18 @@ static void ravb_ring_free(struct net_device *ndev, int q)
priv->tx_ring[q] = NULL;
}
+ /* Free RX skb ringbuffer */
+ if (priv->rx_skb[q]) {
+ for (i = 0; i < priv->num_rx_ring[q]; i++)
+ dev_kfree_skb(priv->rx_skb[q][i]);
+ }
+ kfree(priv->rx_skb[q]);
+ priv->rx_skb[q] = NULL;
+
+ /* Free aligned TX buffers */
+ kfree(priv->tx_align[q]);
+ priv->tx_align[q] = NULL;
+
/* Free TX skb ringbuffer.
* SKBs are freed by ravb_tx_free() call above.
*/
--
2.9.4
From 1c066b033119954bf3c95c7952592e33dd3e5c02 Mon Sep 17 00:00:00 2001
From: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Date: Tue, 6 Jun 2017 01:26:24 +0300
Subject: [PATCH 17/19] net: bridge: fix a null pointer dereference in
br_afspec
[ Upstream commit 1020ce3108cc26fbf09d70550ea2937cb1a211d2 ]
We might call br_afspec() with p == NULL which is a valid use case if
the action is on the bridge device itself, but the bridge tunnel code
dereferences the p pointer without checking, so check if p is null
first.
Reported-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Fixes: efa5356b0d97 ("bridge: per vlan dst_metadata netlink support")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/bridge/br_netlink.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/bridge/br_netlink.c b/net/bridge/br_netlink.c
index 0488c67..8a05a98 100644
--- a/net/bridge/br_netlink.c
+++ b/net/bridge/br_netlink.c
@@ -591,7 +591,7 @@ static int br_afspec(struct net_bridge *br,
err = 0;
switch (nla_type(attr)) {
case IFLA_BRIDGE_VLAN_TUNNEL_INFO:
- if (!(p->flags & BR_VLAN_TUNNEL))
+ if (!p || !(p->flags & BR_VLAN_TUNNEL))
return -EINVAL;
err = br_parse_vlan_tunnel_info(attr, &tinfo_curr);
if (err)
--
2.9.4
From de611ccc852ca22b16328c0bee7b982212e1ef5f Mon Sep 17 00:00:00 2001
From: Max Filippov <jcmvbkbc@gmail.com>
Date: Mon, 5 Jun 2017 18:31:16 -0700
Subject: [PATCH 18/19] net: ethoc: enable NAPI before poll may be scheduled
[ Upstream commit d220b942a4b6a0640aee78841608f4aa5e8e185e ]
ethoc_reset enables device interrupts, ethoc_interrupt may schedule a
NAPI poll before NAPI is enabled in the ethoc_open, which results in
device being unable to send or receive anything until it's closed and
reopened. In case the device is flooded with ingress packets it may be
unable to recover at all.
Move napi_enable above ethoc_reset in the ethoc_open to fix that.
Fixes: a1702857724f ("net: Add support for the OpenCores 10/100 Mbps Ethernet MAC.")
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Tobias Klauser <tklauser@distanz.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/ethoc.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/ethoc.c b/drivers/net/ethernet/ethoc.c
index 23d8274..4f33660 100644
--- a/drivers/net/ethernet/ethoc.c
+++ b/drivers/net/ethernet/ethoc.c
@@ -739,6 +739,8 @@ static int ethoc_open(struct net_device *dev)
if (ret)
return ret;
+ napi_enable(&priv->napi);
+
ethoc_init_ring(priv, dev->mem_start);
ethoc_reset(priv);
@@ -754,7 +756,6 @@ static int ethoc_open(struct net_device *dev)
priv->old_duplex = -1;
phy_start(dev->phydev);
- napi_enable(&priv->napi);
if (netif_msg_ifup(priv)) {
dev_info(&dev->dev, "I/O: %08lx Memory: %08lx-%08lx\n",
--
2.9.4
From f0f33a4c834ab2b1719ce7417a66a34fc8bcba6d Mon Sep 17 00:00:00 2001
From: Niklas Cassel <niklas.cassel@axis.com>
Date: Tue, 6 Jun 2017 09:25:00 +0200
Subject: [PATCH 19/19] net: stmmac: fix completely hung TX when using TSO
[ Upstream commit 426849e6611f2092553f8d53372ae310818a6292 ]
stmmac_tso_allocator can fail to set the Last Descriptor bit
on a descriptor that actually was the last descriptor.
This happens when the buffer of the last descriptor ends
up having a size of exactly TSO_MAX_BUFF_SIZE.
When the IP eventually reaches the next last descriptor,
which actually has the bit set, the DMA will hang.
When the DMA hangs, we get a tx timeout, however,
since stmmac does not do a complete reset of the IP
in stmmac_tx_timeout, we end up in a state with
completely hung TX.
Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Acked-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 4498a38..67e86ac 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1950,7 +1950,7 @@ static void stmmac_tso_allocator(struct stmmac_priv *priv, unsigned int des,
priv->hw->desc->prepare_tso_tx_desc(desc, 0, buff_size,
0, 1,
- (last_segment) && (buff_size < TSO_MAX_BUFF_SIZE),
+ (last_segment) && (tmp_len <= TSO_MAX_BUFF_SIZE),
0, 0);
tmp_len -= TSO_MAX_BUFF_SIZE;
--
2.9.4
[-- Attachment #4: sparc_49.mbox --]
[-- Type: Application/Octet-Stream, Size: 32424 bytes --]
From db8c0ba5e67f9d1044a325fd8da60e5c09f92ab9 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Mon, 5 Jun 2017 11:28:57 -0700
Subject: [PATCH 01/10] sparc64: Add __multi3 for gcc 7.x and later.
[ Upstream commit 1b4af13ff2cc6897557bb0b8d9e2fad4fa4d67aa ]
Reported-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/lib/Makefile | 1 +
arch/sparc/lib/multi3.S | 35 +++++++++++++++++++++++++++++++++++
2 files changed, 36 insertions(+)
create mode 100644 arch/sparc/lib/multi3.S
diff --git a/arch/sparc/lib/Makefile b/arch/sparc/lib/Makefile
index 69912d2..07c03e7 100644
--- a/arch/sparc/lib/Makefile
+++ b/arch/sparc/lib/Makefile
@@ -15,6 +15,7 @@ lib-$(CONFIG_SPARC32) += copy_user.o locks.o
lib-$(CONFIG_SPARC64) += atomic_64.o
lib-$(CONFIG_SPARC32) += lshrdi3.o ashldi3.o
lib-$(CONFIG_SPARC32) += muldi3.o bitext.o cmpdi2.o
+lib-$(CONFIG_SPARC64) += multi3.o
lib-$(CONFIG_SPARC64) += copy_page.o clear_page.o bzero.o
lib-$(CONFIG_SPARC64) += csum_copy.o csum_copy_from_user.o csum_copy_to_user.o
diff --git a/arch/sparc/lib/multi3.S b/arch/sparc/lib/multi3.S
new file mode 100644
index 0000000..d6b6c97
--- /dev/null
+++ b/arch/sparc/lib/multi3.S
@@ -0,0 +1,35 @@
+#include <linux/linkage.h>
+#include <asm/export.h>
+
+ .text
+ .align 4
+ENTRY(__multi3) /* %o0 = u, %o1 = v */
+ mov %o1, %g1
+ srl %o3, 0, %g4
+ mulx %g4, %g1, %o1
+ srlx %g1, 0x20, %g3
+ mulx %g3, %g4, %g5
+ sllx %g5, 0x20, %o5
+ srl %g1, 0, %g4
+ sub %o1, %o5, %o5
+ srlx %o5, 0x20, %o5
+ addcc %g5, %o5, %g5
+ srlx %o3, 0x20, %o5
+ mulx %g4, %o5, %g4
+ mulx %g3, %o5, %o5
+ sethi %hi(0x80000000), %g3
+ addcc %g5, %g4, %g5
+ srlx %g5, 0x20, %g5
+ add %g3, %g3, %g3
+ movcc %xcc, %g0, %g3
+ addcc %o5, %g5, %o5
+ sllx %g4, 0x20, %g4
+ add %o1, %g4, %o1
+ add %o5, %g3, %g2
+ mulx %g1, %o2, %g1
+ add %g1, %g2, %g1
+ mulx %o0, %o3, %o0
+ retl
+ add %g1, %o0, %o0
+ENDPROC(__multi3)
+EXPORT_SYMBOL(__multi3)
--
2.1.2.532.g19b5d50
From 89cc8d783198eda676b52618579e4f4eb7d76d6e Mon Sep 17 00:00:00 2001
From: Mike Kravetz <mike.kravetz@oracle.com>
Date: Fri, 2 Jun 2017 14:51:12 -0700
Subject: [PATCH 02/10] sparc64: mm: fix copy_tsb to correctly copy huge page
TSBs
[ Upstream commit 654f4807624a657f364417c2a7454f0df9961734 ]
When a TSB grows beyond its current capacity, a new TSB is allocated
and copy_tsb is called to copy entries from the old TSB to the new.
A hash shift based on page size is used to calculate the index of an
entry in the TSB. copy_tsb has hard coded PAGE_SHIFT in these
calculations. However, for huge page TSBs the value REAL_HPAGE_SHIFT
should be used. As a result, when copy_tsb is called for a huge page
TSB the entries are placed at the incorrect index in the newly
allocated TSB. When doing hardware table walk, the MMU does not
match these entries and we end up in the TSB miss handling code.
This code will then create and write an entry to the correct index
in the TSB. We take a performance hit for the table walk miss and
recreation of these entries.
Pass a new parameter to copy_tsb that is the page size shift to be
used when copying the TSB.
Suggested-by: Anthony Yznaga <anthony.yznaga@oracle.com>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/kernel/tsb.S | 11 +++++++----
arch/sparc/mm/tsb.c | 7 +++++--
2 files changed, 12 insertions(+), 6 deletions(-)
diff --git a/arch/sparc/kernel/tsb.S b/arch/sparc/kernel/tsb.S
index d568c82..395ec18 100644
--- a/arch/sparc/kernel/tsb.S
+++ b/arch/sparc/kernel/tsb.S
@@ -470,13 +470,16 @@ __tsb_context_switch:
.type copy_tsb,#function
copy_tsb: /* %o0=old_tsb_base, %o1=old_tsb_size
* %o2=new_tsb_base, %o3=new_tsb_size
+ * %o4=page_size_shift
*/
sethi %uhi(TSB_PASS_BITS), %g7
srlx %o3, 4, %o3
- add %o0, %o1, %g1 /* end of old tsb */
+ add %o0, %o1, %o1 /* end of old tsb */
sllx %g7, 32, %g7
sub %o3, 1, %o3 /* %o3 == new tsb hash mask */
+ mov %o4, %g1 /* page_size_shift */
+
661: prefetcha [%o0] ASI_N, #one_read
.section .tsb_phys_patch, "ax"
.word 661b
@@ -501,9 +504,9 @@ copy_tsb: /* %o0=old_tsb_base, %o1=old_tsb_size
/* This can definitely be computed faster... */
srlx %o0, 4, %o5 /* Build index */
and %o5, 511, %o5 /* Mask index */
- sllx %o5, PAGE_SHIFT, %o5 /* Put into vaddr position */
+ sllx %o5, %g1, %o5 /* Put into vaddr position */
or %o4, %o5, %o4 /* Full VADDR. */
- srlx %o4, PAGE_SHIFT, %o4 /* Shift down to create index */
+ srlx %o4, %g1, %o4 /* Shift down to create index */
and %o4, %o3, %o4 /* Mask with new_tsb_nents-1 */
sllx %o4, 4, %o4 /* Shift back up into tsb ent offset */
TSB_STORE(%o2 + %o4, %g2) /* Store TAG */
@@ -511,7 +514,7 @@ copy_tsb: /* %o0=old_tsb_base, %o1=old_tsb_size
TSB_STORE(%o2 + %o4, %g3) /* Store TTE */
80: add %o0, 16, %o0
- cmp %o0, %g1
+ cmp %o0, %o1
bne,pt %xcc, 90b
nop
diff --git a/arch/sparc/mm/tsb.c b/arch/sparc/mm/tsb.c
index e20fbba..84cd593 100644
--- a/arch/sparc/mm/tsb.c
+++ b/arch/sparc/mm/tsb.c
@@ -451,7 +451,8 @@ void tsb_grow(struct mm_struct *mm, unsigned long tsb_index, unsigned long rss)
extern void copy_tsb(unsigned long old_tsb_base,
unsigned long old_tsb_size,
unsigned long new_tsb_base,
- unsigned long new_tsb_size);
+ unsigned long new_tsb_size,
+ unsigned long page_size_shift);
unsigned long old_tsb_base = (unsigned long) old_tsb;
unsigned long new_tsb_base = (unsigned long) new_tsb;
@@ -459,7 +460,9 @@ void tsb_grow(struct mm_struct *mm, unsigned long tsb_index, unsigned long rss)
old_tsb_base = __pa(old_tsb_base);
new_tsb_base = __pa(new_tsb_base);
}
- copy_tsb(old_tsb_base, old_size, new_tsb_base, new_size);
+ copy_tsb(old_tsb_base, old_size, new_tsb_base, new_size,
+ tsb_index == MM_TSB_BASE ?
+ PAGE_SHIFT : REAL_HPAGE_SHIFT);
}
mm->context.tsb_block[tsb_index].tsb = new_tsb;
--
2.1.2.532.g19b5d50
From ba508b33ef0b051c15e990843fb0287c181cd350 Mon Sep 17 00:00:00 2001
From: James Clarke <jrtc27@jrtc27.com>
Date: Mon, 29 May 2017 20:17:56 +0100
Subject: [PATCH 03/10] sparc: Machine description indices can vary
[ Upstream commit c982aa9c304bf0b9a7522fd118fed4afa5a0263c ]
VIO devices were being looked up by their index in the machine
description node block, but this often varies over time as devices are
added and removed. Instead, store the ID and look up using the type,
config handle and ID.
Signed-off-by: James Clarke <jrtc27@jrtc27.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=112541
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/vio.h | 1 +
arch/sparc/kernel/vio.c | 68 +++++++++++++++++++++++++++++++++++++++++---
2 files changed, 65 insertions(+), 4 deletions(-)
diff --git a/arch/sparc/include/asm/vio.h b/arch/sparc/include/asm/vio.h
index 8174f6c..9dca7a8 100644
--- a/arch/sparc/include/asm/vio.h
+++ b/arch/sparc/include/asm/vio.h
@@ -327,6 +327,7 @@ struct vio_dev {
int compat_len;
u64 dev_no;
+ u64 id;
unsigned long channel_id;
diff --git a/arch/sparc/kernel/vio.c b/arch/sparc/kernel/vio.c
index f6bb857..075d389 100644
--- a/arch/sparc/kernel/vio.c
+++ b/arch/sparc/kernel/vio.c
@@ -302,13 +302,16 @@ static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
if (!id) {
dev_set_name(&vdev->dev, "%s", bus_id_name);
vdev->dev_no = ~(u64)0;
+ vdev->id = ~(u64)0;
} else if (!cfg_handle) {
dev_set_name(&vdev->dev, "%s-%llu", bus_id_name, *id);
vdev->dev_no = *id;
+ vdev->id = ~(u64)0;
} else {
dev_set_name(&vdev->dev, "%s-%llu-%llu", bus_id_name,
*cfg_handle, *id);
vdev->dev_no = *cfg_handle;
+ vdev->id = *id;
}
vdev->dev.parent = parent;
@@ -351,27 +354,84 @@ static void vio_add(struct mdesc_handle *hp, u64 node)
(void) vio_create_one(hp, node, &root_vdev->dev);
}
+struct vio_md_node_query {
+ const char *type;
+ u64 dev_no;
+ u64 id;
+};
+
static int vio_md_node_match(struct device *dev, void *arg)
{
+ struct vio_md_node_query *query = (struct vio_md_node_query *) arg;
struct vio_dev *vdev = to_vio_dev(dev);
- if (vdev->mp == (u64) arg)
- return 1;
+ if (vdev->dev_no != query->dev_no)
+ return 0;
+ if (vdev->id != query->id)
+ return 0;
+ if (strcmp(vdev->type, query->type))
+ return 0;
- return 0;
+ return 1;
}
static void vio_remove(struct mdesc_handle *hp, u64 node)
{
+ const char *type;
+ const u64 *id, *cfg_handle;
+ u64 a;
+ struct vio_md_node_query query;
struct device *dev;
- dev = device_find_child(&root_vdev->dev, (void *) node,
+ type = mdesc_get_property(hp, node, "device-type", NULL);
+ if (!type) {
+ type = mdesc_get_property(hp, node, "name", NULL);
+ if (!type)
+ type = mdesc_node_name(hp, node);
+ }
+
+ query.type = type;
+
+ id = mdesc_get_property(hp, node, "id", NULL);
+ cfg_handle = NULL;
+ mdesc_for_each_arc(a, hp, node, MDESC_ARC_TYPE_BACK) {
+ u64 target;
+
+ target = mdesc_arc_target(hp, a);
+ cfg_handle = mdesc_get_property(hp, target,
+ "cfg-handle", NULL);
+ if (cfg_handle)
+ break;
+ }
+
+ if (!id) {
+ query.dev_no = ~(u64)0;
+ query.id = ~(u64)0;
+ } else if (!cfg_handle) {
+ query.dev_no = *id;
+ query.id = ~(u64)0;
+ } else {
+ query.dev_no = *cfg_handle;
+ query.id = *id;
+ }
+
+ dev = device_find_child(&root_vdev->dev, &query,
vio_md_node_match);
if (dev) {
printk(KERN_INFO "VIO: Removing device %s\n", dev_name(dev));
device_unregister(dev);
put_device(dev);
+ } else {
+ if (!id)
+ printk(KERN_ERR "VIO: Removed unknown %s node.\n",
+ type);
+ else if (!cfg_handle)
+ printk(KERN_ERR "VIO: Removed unknown %s node %llu.\n",
+ type, *id);
+ else
+ printk(KERN_ERR "VIO: Removed unknown %s node %llu-%llu.\n",
+ type, *cfg_handle, *id);
}
}
--
2.1.2.532.g19b5d50
From 80d4bfee35913affcb0c8c6e788e17df20b3733d Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:20 -0400
Subject: [PATCH 04/10] sparc64: reset mm cpumask after wrap
[ Upstream commit 588974857359861891f478a070b1dc7ae04a3880 ]
After a wrap (getting a new context version) a process must get a new
context id, which means that we would need to flush the context id from
the TLB before running for the first time with this ID on every CPU. But,
we use mm_cpumask to determine if this process has been running on this CPU
before, and this mask is not reset after a wrap. So, there are two possible
fixes for this issue:
1. Clear mm cpumask whenever mm gets a new context id
2. Unconditionally flush context every time process is running on a CPU
This patch implements the first solution
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/mm/init_64.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index bd7e2aa..f9c0eee 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -710,6 +710,8 @@ void get_new_mmu_context(struct mm_struct *mm)
goto out;
}
}
+ if (mm->context.sparc64_ctx_val)
+ cpumask_clear(mm_cpumask(mm));
mmu_context_bmap[new_ctx>>6] |= (1UL << (new_ctx & 63));
new_ctx |= (tlb_context_cache & CTX_VERSION_MASK);
out:
--
2.1.2.532.g19b5d50
From c543fa27e5f6f74fcca83eccf995ffb2c40ecc17 Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:21 -0400
Subject: [PATCH 05/10] sparc64: combine activate_mm and switch_mm
[ Upstream commit 14d0334c6748ff2aedb3f2f7fdc51ee90a9b54e7 ]
The only difference between these two functions is that in activate_mm we
unconditionally flush context. However, there is no need to keep this
difference after fixing a bug where cpumask was not reset on a wrap. So, in
this patch we combine these.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_context_64.h | 21 +--------------------
1 file changed, 1 insertion(+), 20 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_context_64.h b/arch/sparc/include/asm/mmu_context_64.h
index b84be67..266662a 100644
--- a/arch/sparc/include/asm/mmu_context_64.h
+++ b/arch/sparc/include/asm/mmu_context_64.h
@@ -131,26 +131,7 @@ static inline void switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, str
}
#define deactivate_mm(tsk,mm) do { } while (0)
-
-/* Activate a new MM instance for the current task. */
-static inline void activate_mm(struct mm_struct *active_mm, struct mm_struct *mm)
-{
- unsigned long flags;
- int cpu;
-
- spin_lock_irqsave(&mm->context.lock, flags);
- if (!CTX_VALID(mm->context))
- get_new_mmu_context(mm);
- cpu = smp_processor_id();
- if (!cpumask_test_cpu(cpu, mm_cpumask(mm)))
- cpumask_set_cpu(cpu, mm_cpumask(mm));
-
- load_secondary_context(mm);
- __flush_tlb_mm(CTX_HWBITS(mm->context), SECONDARY_CONTEXT);
- tsb_context_switch(mm);
- spin_unlock_irqrestore(&mm->context.lock, flags);
-}
-
+#define activate_mm(active_mm, mm) switch_mm(active_mm, mm, NULL)
#endif /* !(__ASSEMBLY__) */
#endif /* !(__SPARC64_MMU_CONTEXT_H) */
--
2.1.2.532.g19b5d50
From 372592bcacee13ea42743c5abfaa6cda086156c1 Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:22 -0400
Subject: [PATCH 06/10] sparc64: redefine first version
[ Upstream commit c4415235b2be0cc791572e8e7f7466ab8f73a2bf ]
CTX_FIRST_VERSION defines the first context version, but also it defines
first context. This patch redefines it to only include the first context
version.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_64.h | 2 +-
arch/sparc/mm/init_64.c | 6 +++---
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_64.h b/arch/sparc/include/asm/mmu_64.h
index f7de0db..83b36a5 100644
--- a/arch/sparc/include/asm/mmu_64.h
+++ b/arch/sparc/include/asm/mmu_64.h
@@ -52,7 +52,7 @@
#define CTX_NR_MASK TAG_CONTEXT_BITS
#define CTX_HW_MASK (CTX_NR_MASK | CTX_PGSZ_MASK)
-#define CTX_FIRST_VERSION ((_AC(1,UL) << CTX_VERSION_SHIFT) + _AC(1,UL))
+#define CTX_FIRST_VERSION BIT(CTX_VERSION_SHIFT)
#define CTX_VALID(__ctx) \
(!(((__ctx.sparc64_ctx_val) ^ tlb_context_cache) & CTX_VERSION_MASK))
#define CTX_HWBITS(__ctx) ((__ctx.sparc64_ctx_val) & CTX_HW_MASK)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index f9c0eee..ccbaa34 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -658,7 +658,7 @@ EXPORT_SYMBOL(__flush_dcache_range);
/* get_new_mmu_context() uses "cache + 1". */
DEFINE_SPINLOCK(ctx_alloc_lock);
-unsigned long tlb_context_cache = CTX_FIRST_VERSION - 1;
+unsigned long tlb_context_cache = CTX_FIRST_VERSION;
#define MAX_CTX_NR (1UL << CTX_NR_BITS)
#define CTX_BMAP_SLOTS BITS_TO_LONGS(MAX_CTX_NR)
DECLARE_BITMAP(mmu_context_bmap, MAX_CTX_NR);
@@ -689,9 +689,9 @@ void get_new_mmu_context(struct mm_struct *mm)
if (new_ctx >= ctx) {
int i;
new_ctx = (tlb_context_cache & CTX_VERSION_MASK) +
- CTX_FIRST_VERSION;
+ CTX_FIRST_VERSION + 1;
if (new_ctx == 1)
- new_ctx = CTX_FIRST_VERSION;
+ new_ctx = CTX_FIRST_VERSION + 1;
/* Don't call memset, for 16 entries that's just
* plain silly...
--
2.1.2.532.g19b5d50
From 0d6769b07f387ec457b690c266b8ea68cf2dcb98 Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:23 -0400
Subject: [PATCH 07/10] sparc64: add per-cpu mm of secondary contexts
[ Upstream commit 7a5b4bbf49fe86ce77488a70c5dccfe2d50d7a2d ]
The new wrap is going to use information from this array to figure out
mm's that currently have valid secondary contexts setup.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_context_64.h | 5 +++--
arch/sparc/mm/init_64.c | 1 +
2 files changed, 4 insertions(+), 2 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_context_64.h b/arch/sparc/include/asm/mmu_context_64.h
index 266662a..eb41086 100644
--- a/arch/sparc/include/asm/mmu_context_64.h
+++ b/arch/sparc/include/asm/mmu_context_64.h
@@ -17,6 +17,7 @@ extern spinlock_t ctx_alloc_lock;
extern unsigned long tlb_context_cache;
extern unsigned long mmu_context_bmap[];
+DECLARE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm);
void get_new_mmu_context(struct mm_struct *mm);
#ifdef CONFIG_SMP
void smp_new_mmu_context_version(void);
@@ -74,8 +75,9 @@ void __flush_tlb_mm(unsigned long, unsigned long);
static inline void switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, struct task_struct *tsk)
{
unsigned long ctx_valid, flags;
- int cpu;
+ int cpu = smp_processor_id();
+ per_cpu(per_cpu_secondary_mm, cpu) = mm;
if (unlikely(mm == &init_mm))
return;
@@ -121,7 +123,6 @@ static inline void switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, str
* for the first time, we must flush that context out of the
* local TLB.
*/
- cpu = smp_processor_id();
if (!ctx_valid || !cpumask_test_cpu(cpu, mm_cpumask(mm))) {
cpumask_set_cpu(cpu, mm_cpumask(mm));
__flush_tlb_mm(CTX_HWBITS(mm->context),
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index ccbaa34..d5cebdb 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -662,6 +662,7 @@ unsigned long tlb_context_cache = CTX_FIRST_VERSION;
#define MAX_CTX_NR (1UL << CTX_NR_BITS)
#define CTX_BMAP_SLOTS BITS_TO_LONGS(MAX_CTX_NR)
DECLARE_BITMAP(mmu_context_bmap, MAX_CTX_NR);
+DEFINE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm) = {0};
/* Caller does TLB context flushing on local CPU if necessary.
* The caller also ensures that CTX_VALID(mm->context) is false.
--
2.1.2.532.g19b5d50
From 4cb559160611db28014224b8b521c8fccd22d80b Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:24 -0400
Subject: [PATCH 08/10] sparc64: new context wrap
[ Upstream commit a0582f26ec9dfd5360ea2f35dd9a1b026f8adda0 ]
The current wrap implementation has a race issue: it is called outside of
the ctx_alloc_lock, and also does not wait for all CPUs to complete the
wrap. This means that a thread can get a new context with a new version
and another thread might still be running with the same context. The
problem is especially severe on CPUs with shared TLBs, like sun4v. I used
the following test to very quickly reproduce the problem:
- start over 8K processes (must be more than context IDs)
- write and read values at a memory location in every process.
Very quickly memory corruptions start happening, and what we read back
does not equal what we wrote.
Several approaches were explored before settling on this one:
Approach 1:
Move smp_new_mmu_context_version() inside ctx_alloc_lock, and wait for
every process to complete the wrap. (Note: every CPU must WAIT before
leaving smp_new_mmu_context_version_client() until every one arrives).
This approach ends up with deadlocks, as some threads own locks which other
threads are waiting for, and they never receive softint until these threads
exit smp_new_mmu_context_version_client(). Since we do not allow the exit,
deadlock happens.
Approach 2:
Handle wrap right during mondo interrupt. Use etrap/rtrap to enter into
into C code, and issue new versions to every CPU.
This approach adds some overhead to runtime: in switch_mm() we must add
some checks to make sure that versions have not changed due to wrap while
we were loading the new secondary context. (could be protected by PSTATE_IE
but that degrades performance as on M7 and older CPUs as it takes 50 cycles
for each access). Also, we still need a global per-cpu array of MMs to know
where we need to load new contexts, otherwise we can change context to a
thread that is going way (if we received mondo between switch_mm() and
switch_to() time). Finally, there are some issues with window registers in
rtrap() when context IDs are changed during CPU mondo time.
The approach in this patch is the simplest and has almost no impact on
runtime. We use the array with mm's where last secondary contexts were
loaded onto CPUs and bump their versions to the new generation without
changing context IDs. If a new process comes in to get a context ID, it
will go through get_new_mmu_context() because of version mismatch. But the
running processes do not need to be interrupted. And wrap is quicker as we
do not need to xcall and wait for everyone to receive and complete wrap.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/mm/init_64.c | 81 ++++++++++++++++++++++++++++++++-----------------
1 file changed, 54 insertions(+), 27 deletions(-)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index d5cebdb..57154c6 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -664,6 +664,53 @@ unsigned long tlb_context_cache = CTX_FIRST_VERSION;
DECLARE_BITMAP(mmu_context_bmap, MAX_CTX_NR);
DEFINE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm) = {0};
+static void mmu_context_wrap(void)
+{
+ unsigned long old_ver = tlb_context_cache & CTX_VERSION_MASK;
+ unsigned long new_ver, new_ctx, old_ctx;
+ struct mm_struct *mm;
+ int cpu;
+
+ bitmap_zero(mmu_context_bmap, 1 << CTX_NR_BITS);
+
+ /* Reserve kernel context */
+ set_bit(0, mmu_context_bmap);
+
+ new_ver = (tlb_context_cache & CTX_VERSION_MASK) + CTX_FIRST_VERSION;
+ if (unlikely(new_ver == 0))
+ new_ver = CTX_FIRST_VERSION;
+ tlb_context_cache = new_ver;
+
+ /*
+ * Make sure that any new mm that are added into per_cpu_secondary_mm,
+ * are going to go through get_new_mmu_context() path.
+ */
+ mb();
+
+ /*
+ * Updated versions to current on those CPUs that had valid secondary
+ * contexts
+ */
+ for_each_online_cpu(cpu) {
+ /*
+ * If a new mm is stored after we took this mm from the array,
+ * it will go into get_new_mmu_context() path, because we
+ * already bumped the version in tlb_context_cache.
+ */
+ mm = per_cpu(per_cpu_secondary_mm, cpu);
+
+ if (unlikely(!mm || mm == &init_mm))
+ continue;
+
+ old_ctx = mm->context.sparc64_ctx_val;
+ if (likely((old_ctx & CTX_VERSION_MASK) == old_ver)) {
+ new_ctx = (old_ctx & ~CTX_VERSION_MASK) | new_ver;
+ set_bit(new_ctx & CTX_NR_MASK, mmu_context_bmap);
+ mm->context.sparc64_ctx_val = new_ctx;
+ }
+ }
+}
+
/* Caller does TLB context flushing on local CPU if necessary.
* The caller also ensures that CTX_VALID(mm->context) is false.
*
@@ -678,50 +725,30 @@ void get_new_mmu_context(struct mm_struct *mm)
{
unsigned long ctx, new_ctx;
unsigned long orig_pgsz_bits;
- int new_version;
spin_lock(&ctx_alloc_lock);
+retry:
+ /* wrap might have happened, test again if our context became valid */
+ if (unlikely(CTX_VALID(mm->context)))
+ goto out;
orig_pgsz_bits = (mm->context.sparc64_ctx_val & CTX_PGSZ_MASK);
ctx = (tlb_context_cache + 1) & CTX_NR_MASK;
new_ctx = find_next_zero_bit(mmu_context_bmap, 1 << CTX_NR_BITS, ctx);
- new_version = 0;
if (new_ctx >= (1 << CTX_NR_BITS)) {
new_ctx = find_next_zero_bit(mmu_context_bmap, ctx, 1);
if (new_ctx >= ctx) {
- int i;
- new_ctx = (tlb_context_cache & CTX_VERSION_MASK) +
- CTX_FIRST_VERSION + 1;
- if (new_ctx == 1)
- new_ctx = CTX_FIRST_VERSION + 1;
-
- /* Don't call memset, for 16 entries that's just
- * plain silly...
- */
- mmu_context_bmap[0] = 3;
- mmu_context_bmap[1] = 0;
- mmu_context_bmap[2] = 0;
- mmu_context_bmap[3] = 0;
- for (i = 4; i < CTX_BMAP_SLOTS; i += 4) {
- mmu_context_bmap[i + 0] = 0;
- mmu_context_bmap[i + 1] = 0;
- mmu_context_bmap[i + 2] = 0;
- mmu_context_bmap[i + 3] = 0;
- }
- new_version = 1;
- goto out;
+ mmu_context_wrap();
+ goto retry;
}
}
if (mm->context.sparc64_ctx_val)
cpumask_clear(mm_cpumask(mm));
mmu_context_bmap[new_ctx>>6] |= (1UL << (new_ctx & 63));
new_ctx |= (tlb_context_cache & CTX_VERSION_MASK);
-out:
tlb_context_cache = new_ctx;
mm->context.sparc64_ctx_val = new_ctx | orig_pgsz_bits;
+out:
spin_unlock(&ctx_alloc_lock);
-
- if (unlikely(new_version))
- smp_new_mmu_context_version();
}
static int numa_enabled = 1;
--
2.1.2.532.g19b5d50
From 3e47faa995873df69e049f30ae5eb5281364d5fe Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:25 -0400
Subject: [PATCH 09/10] sparc64: delete old wrap code
[ Upstream commit 0197e41ce70511dc3b71f7fefa1a676e2b5cd60b ]
The old method that is using xcall and softint to get new context id is
deleted, as it is replaced by a method of using per_cpu_secondary_mm
without xcall to perform the context wrap.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_context_64.h | 6 ------
arch/sparc/include/asm/pil.h | 1 -
arch/sparc/kernel/kernel.h | 1 -
arch/sparc/kernel/smp_64.c | 31 -------------------------------
arch/sparc/kernel/ttable_64.S | 2 +-
arch/sparc/mm/ultra.S | 5 -----
6 files changed, 1 insertion(+), 45 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_context_64.h b/arch/sparc/include/asm/mmu_context_64.h
index eb41086..349dd23 100644
--- a/arch/sparc/include/asm/mmu_context_64.h
+++ b/arch/sparc/include/asm/mmu_context_64.h
@@ -19,12 +19,6 @@ extern unsigned long mmu_context_bmap[];
DECLARE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm);
void get_new_mmu_context(struct mm_struct *mm);
-#ifdef CONFIG_SMP
-void smp_new_mmu_context_version(void);
-#else
-#define smp_new_mmu_context_version() do { } while (0)
-#endif
-
int init_new_context(struct task_struct *tsk, struct mm_struct *mm);
void destroy_context(struct mm_struct *mm);
diff --git a/arch/sparc/include/asm/pil.h b/arch/sparc/include/asm/pil.h
index 26693703..522b43d 100644
--- a/arch/sparc/include/asm/pil.h
+++ b/arch/sparc/include/asm/pil.h
@@ -20,7 +20,6 @@
#define PIL_SMP_CALL_FUNC 1
#define PIL_SMP_RECEIVE_SIGNAL 2
#define PIL_SMP_CAPTURE 3
-#define PIL_SMP_CTX_NEW_VERSION 4
#define PIL_DEVICE_IRQ 5
#define PIL_SMP_CALL_FUNC_SNGL 6
#define PIL_DEFERRED_PCR_WORK 7
diff --git a/arch/sparc/kernel/kernel.h b/arch/sparc/kernel/kernel.h
index c9804551..6ae1e77 100644
--- a/arch/sparc/kernel/kernel.h
+++ b/arch/sparc/kernel/kernel.h
@@ -37,7 +37,6 @@ void handle_stdfmna(struct pt_regs *regs, unsigned long sfar, unsigned long sfsr
/* smp_64.c */
void __irq_entry smp_call_function_client(int irq, struct pt_regs *regs);
void __irq_entry smp_call_function_single_client(int irq, struct pt_regs *regs);
-void __irq_entry smp_new_mmu_context_version_client(int irq, struct pt_regs *regs);
void __irq_entry smp_penguin_jailcell(int irq, struct pt_regs *regs);
void __irq_entry smp_receive_signal_client(int irq, struct pt_regs *regs);
diff --git a/arch/sparc/kernel/smp_64.c b/arch/sparc/kernel/smp_64.c
index 8182f7c..d5807d2 100644
--- a/arch/sparc/kernel/smp_64.c
+++ b/arch/sparc/kernel/smp_64.c
@@ -963,37 +963,6 @@ void flush_dcache_page_all(struct mm_struct *mm, struct page *page)
preempt_enable();
}
-void __irq_entry smp_new_mmu_context_version_client(int irq, struct pt_regs *regs)
-{
- struct mm_struct *mm;
- unsigned long flags;
-
- clear_softint(1 << irq);
-
- /* See if we need to allocate a new TLB context because
- * the version of the one we are using is now out of date.
- */
- mm = current->active_mm;
- if (unlikely(!mm || (mm == &init_mm)))
- return;
-
- spin_lock_irqsave(&mm->context.lock, flags);
-
- if (unlikely(!CTX_VALID(mm->context)))
- get_new_mmu_context(mm);
-
- spin_unlock_irqrestore(&mm->context.lock, flags);
-
- load_secondary_context(mm);
- __flush_tlb_mm(CTX_HWBITS(mm->context),
- SECONDARY_CONTEXT);
-}
-
-void smp_new_mmu_context_version(void)
-{
- smp_cross_call(&xcall_new_mmu_context_version, 0, 0, 0);
-}
-
#ifdef CONFIG_KGDB
void kgdb_roundup_cpus(unsigned long flags)
{
diff --git a/arch/sparc/kernel/ttable_64.S b/arch/sparc/kernel/ttable_64.S
index c6dfdaa..170ead6 100644
--- a/arch/sparc/kernel/ttable_64.S
+++ b/arch/sparc/kernel/ttable_64.S
@@ -50,7 +50,7 @@ tl0_resv03e: BTRAP(0x3e) BTRAP(0x3f) BTRAP(0x40)
tl0_irq1: TRAP_IRQ(smp_call_function_client, 1)
tl0_irq2: TRAP_IRQ(smp_receive_signal_client, 2)
tl0_irq3: TRAP_IRQ(smp_penguin_jailcell, 3)
-tl0_irq4: TRAP_IRQ(smp_new_mmu_context_version_client, 4)
+tl0_irq4: BTRAP(0x44)
#else
tl0_irq1: BTRAP(0x41)
tl0_irq2: BTRAP(0x42)
diff --git a/arch/sparc/mm/ultra.S b/arch/sparc/mm/ultra.S
index 5d2fd6c..fcf4d27 100644
--- a/arch/sparc/mm/ultra.S
+++ b/arch/sparc/mm/ultra.S
@@ -971,11 +971,6 @@ xcall_capture:
wr %g0, (1 << PIL_SMP_CAPTURE), %set_softint
retry
- .globl xcall_new_mmu_context_version
-xcall_new_mmu_context_version:
- wr %g0, (1 << PIL_SMP_CTX_NEW_VERSION), %set_softint
- retry
-
#ifdef CONFIG_KGDB
.globl xcall_kgdb_capture
xcall_kgdb_capture:
--
2.1.2.532.g19b5d50
From 4218909feb88bef7a1c3a2305f43cb1a60a56949 Mon Sep 17 00:00:00 2001
From: Jane Chu <jane.chu@oracle.com>
Date: Tue, 6 Jun 2017 14:32:29 -0600
Subject: [PATCH 10/10] arch/sparc: support NR_CPUS = 4096
[ Upstream commit c79a13734d104b5b147d7cb0870276ccdd660dae ]
Linux SPARC64 limits NR_CPUS to 4064 because init_cpu_send_mondo_info()
only allocates a single page for NR_CPUS mondo entries. Thus we cannot
use all 4096 CPUs on some SPARC platforms.
To fix, allocate (2^order) pages where order is set according to the size
of cpu_list for possible cpus. Since cpu_list_pa and cpu_mondo_block_pa
are not used in asm code, there are no imm13 offsets from the base PA
that will break because they can only reach one page.
Orabug: 25505750
Signed-off-by: Jane Chu <jane.chu@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Atish Patra <atish.patra@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/Kconfig | 4 ++--
arch/sparc/kernel/irq_64.c | 17 +++++++++++++----
2 files changed, 15 insertions(+), 6 deletions(-)
diff --git a/arch/sparc/Kconfig b/arch/sparc/Kconfig
index 165ecdd..b27e48e 100644
--- a/arch/sparc/Kconfig
+++ b/arch/sparc/Kconfig
@@ -187,9 +187,9 @@ config NR_CPUS
int "Maximum number of CPUs"
depends on SMP
range 2 32 if SPARC32
- range 2 1024 if SPARC64
+ range 2 4096 if SPARC64
default 32 if SPARC32
- default 64 if SPARC64
+ default 4096 if SPARC64
source kernel/Kconfig.hz
diff --git a/arch/sparc/kernel/irq_64.c b/arch/sparc/kernel/irq_64.c
index 34a7930..e1b1ce6 100644
--- a/arch/sparc/kernel/irq_64.c
+++ b/arch/sparc/kernel/irq_64.c
@@ -1034,17 +1034,26 @@ static void __init init_cpu_send_mondo_info(struct trap_per_cpu *tb)
{
#ifdef CONFIG_SMP
unsigned long page;
+ void *mondo, *p;
- BUILD_BUG_ON((NR_CPUS * sizeof(u16)) > (PAGE_SIZE - 64));
+ BUILD_BUG_ON((NR_CPUS * sizeof(u16)) > PAGE_SIZE);
+
+ /* Make sure mondo block is 64byte aligned */
+ p = kzalloc(127, GFP_KERNEL);
+ if (!p) {
+ prom_printf("SUN4V: Error, cannot allocate mondo block.\n");
+ prom_halt();
+ }
+ mondo = (void *)(((unsigned long)p + 63) & ~0x3f);
+ tb->cpu_mondo_block_pa = __pa(mondo);
page = get_zeroed_page(GFP_KERNEL);
if (!page) {
- prom_printf("SUN4V: Error, cannot allocate cpu mondo page.\n");
+ prom_printf("SUN4V: Error, cannot allocate cpu list page.\n");
prom_halt();
}
- tb->cpu_mondo_block_pa = __pa(page);
- tb->cpu_list_pa = __pa(page + 64);
+ tb->cpu_list_pa = __pa(page);
#endif
}
--
2.1.2.532.g19b5d50
[-- Attachment #5: sparc_411.mbox --]
[-- Type: Application/Octet-Stream, Size: 33524 bytes --]
From d616f3e38d8d3af7a9a4364f3398c5393815e68e Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Mon, 5 Jun 2017 11:28:57 -0700
Subject: [PATCH 01/11] sparc64: Add __multi3 for gcc 7.x and later.
[ Upstream commit 1b4af13ff2cc6897557bb0b8d9e2fad4fa4d67aa ]
Reported-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/lib/Makefile | 1 +
arch/sparc/lib/multi3.S | 35 +++++++++++++++++++++++++++++++++++
2 files changed, 36 insertions(+)
create mode 100644 arch/sparc/lib/multi3.S
diff --git a/arch/sparc/lib/Makefile b/arch/sparc/lib/Makefile
index 69912d2..07c03e7 100644
--- a/arch/sparc/lib/Makefile
+++ b/arch/sparc/lib/Makefile
@@ -15,6 +15,7 @@ lib-$(CONFIG_SPARC32) += copy_user.o locks.o
lib-$(CONFIG_SPARC64) += atomic_64.o
lib-$(CONFIG_SPARC32) += lshrdi3.o ashldi3.o
lib-$(CONFIG_SPARC32) += muldi3.o bitext.o cmpdi2.o
+lib-$(CONFIG_SPARC64) += multi3.o
lib-$(CONFIG_SPARC64) += copy_page.o clear_page.o bzero.o
lib-$(CONFIG_SPARC64) += csum_copy.o csum_copy_from_user.o csum_copy_to_user.o
diff --git a/arch/sparc/lib/multi3.S b/arch/sparc/lib/multi3.S
new file mode 100644
index 0000000..d6b6c97
--- /dev/null
+++ b/arch/sparc/lib/multi3.S
@@ -0,0 +1,35 @@
+#include <linux/linkage.h>
+#include <asm/export.h>
+
+ .text
+ .align 4
+ENTRY(__multi3) /* %o0 = u, %o1 = v */
+ mov %o1, %g1
+ srl %o3, 0, %g4
+ mulx %g4, %g1, %o1
+ srlx %g1, 0x20, %g3
+ mulx %g3, %g4, %g5
+ sllx %g5, 0x20, %o5
+ srl %g1, 0, %g4
+ sub %o1, %o5, %o5
+ srlx %o5, 0x20, %o5
+ addcc %g5, %o5, %g5
+ srlx %o3, 0x20, %o5
+ mulx %g4, %o5, %g4
+ mulx %g3, %o5, %o5
+ sethi %hi(0x80000000), %g3
+ addcc %g5, %g4, %g5
+ srlx %g5, 0x20, %g5
+ add %g3, %g3, %g3
+ movcc %xcc, %g0, %g3
+ addcc %o5, %g5, %o5
+ sllx %g4, 0x20, %g4
+ add %o1, %g4, %o1
+ add %o5, %g3, %g2
+ mulx %g1, %o2, %g1
+ add %g1, %g2, %g1
+ mulx %o0, %o3, %o0
+ retl
+ add %g1, %o0, %o0
+ENDPROC(__multi3)
+EXPORT_SYMBOL(__multi3)
--
2.1.2.532.g19b5d50
From 2c87fae417a6e281c7032f603198b23d9a9eef0c Mon Sep 17 00:00:00 2001
From: Mike Kravetz <mike.kravetz@oracle.com>
Date: Fri, 2 Jun 2017 14:51:12 -0700
Subject: [PATCH 02/11] sparc64: mm: fix copy_tsb to correctly copy huge page
TSBs
[ Upstream commit 654f4807624a657f364417c2a7454f0df9961734 ]
When a TSB grows beyond its current capacity, a new TSB is allocated
and copy_tsb is called to copy entries from the old TSB to the new.
A hash shift based on page size is used to calculate the index of an
entry in the TSB. copy_tsb has hard coded PAGE_SHIFT in these
calculations. However, for huge page TSBs the value REAL_HPAGE_SHIFT
should be used. As a result, when copy_tsb is called for a huge page
TSB the entries are placed at the incorrect index in the newly
allocated TSB. When doing hardware table walk, the MMU does not
match these entries and we end up in the TSB miss handling code.
This code will then create and write an entry to the correct index
in the TSB. We take a performance hit for the table walk miss and
recreation of these entries.
Pass a new parameter to copy_tsb that is the page size shift to be
used when copying the TSB.
Suggested-by: Anthony Yznaga <anthony.yznaga@oracle.com>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/kernel/tsb.S | 11 +++++++----
arch/sparc/mm/tsb.c | 7 +++++--
2 files changed, 12 insertions(+), 6 deletions(-)
diff --git a/arch/sparc/kernel/tsb.S b/arch/sparc/kernel/tsb.S
index 10689cf..07c0df9 100644
--- a/arch/sparc/kernel/tsb.S
+++ b/arch/sparc/kernel/tsb.S
@@ -455,13 +455,16 @@ __tsb_context_switch:
.type copy_tsb,#function
copy_tsb: /* %o0=old_tsb_base, %o1=old_tsb_size
* %o2=new_tsb_base, %o3=new_tsb_size
+ * %o4=page_size_shift
*/
sethi %uhi(TSB_PASS_BITS), %g7
srlx %o3, 4, %o3
- add %o0, %o1, %g1 /* end of old tsb */
+ add %o0, %o1, %o1 /* end of old tsb */
sllx %g7, 32, %g7
sub %o3, 1, %o3 /* %o3 == new tsb hash mask */
+ mov %o4, %g1 /* page_size_shift */
+
661: prefetcha [%o0] ASI_N, #one_read
.section .tsb_phys_patch, "ax"
.word 661b
@@ -486,9 +489,9 @@ copy_tsb: /* %o0=old_tsb_base, %o1=old_tsb_size
/* This can definitely be computed faster... */
srlx %o0, 4, %o5 /* Build index */
and %o5, 511, %o5 /* Mask index */
- sllx %o5, PAGE_SHIFT, %o5 /* Put into vaddr position */
+ sllx %o5, %g1, %o5 /* Put into vaddr position */
or %o4, %o5, %o4 /* Full VADDR. */
- srlx %o4, PAGE_SHIFT, %o4 /* Shift down to create index */
+ srlx %o4, %g1, %o4 /* Shift down to create index */
and %o4, %o3, %o4 /* Mask with new_tsb_nents-1 */
sllx %o4, 4, %o4 /* Shift back up into tsb ent offset */
TSB_STORE(%o2 + %o4, %g2) /* Store TAG */
@@ -496,7 +499,7 @@ copy_tsb: /* %o0=old_tsb_base, %o1=old_tsb_size
TSB_STORE(%o2 + %o4, %g3) /* Store TTE */
80: add %o0, 16, %o0
- cmp %o0, %g1
+ cmp %o0, %o1
bne,pt %xcc, 90b
nop
diff --git a/arch/sparc/mm/tsb.c b/arch/sparc/mm/tsb.c
index bedf08b..0d4b998 100644
--- a/arch/sparc/mm/tsb.c
+++ b/arch/sparc/mm/tsb.c
@@ -496,7 +496,8 @@ void tsb_grow(struct mm_struct *mm, unsigned long tsb_index, unsigned long rss)
extern void copy_tsb(unsigned long old_tsb_base,
unsigned long old_tsb_size,
unsigned long new_tsb_base,
- unsigned long new_tsb_size);
+ unsigned long new_tsb_size,
+ unsigned long page_size_shift);
unsigned long old_tsb_base = (unsigned long) old_tsb;
unsigned long new_tsb_base = (unsigned long) new_tsb;
@@ -504,7 +505,9 @@ void tsb_grow(struct mm_struct *mm, unsigned long tsb_index, unsigned long rss)
old_tsb_base = __pa(old_tsb_base);
new_tsb_base = __pa(new_tsb_base);
}
- copy_tsb(old_tsb_base, old_size, new_tsb_base, new_size);
+ copy_tsb(old_tsb_base, old_size, new_tsb_base, new_size,
+ tsb_index == MM_TSB_BASE ?
+ PAGE_SHIFT : REAL_HPAGE_SHIFT);
}
mm->context.tsb_block[tsb_index].tsb = new_tsb;
--
2.1.2.532.g19b5d50
From 08996c4493c77d95235243e28045b62b26e1d71b Mon Sep 17 00:00:00 2001
From: James Clarke <jrtc27@jrtc27.com>
Date: Mon, 29 May 2017 20:17:56 +0100
Subject: [PATCH 03/11] sparc: Machine description indices can vary
[ Upstream commit c982aa9c304bf0b9a7522fd118fed4afa5a0263c ]
VIO devices were being looked up by their index in the machine
description node block, but this often varies over time as devices are
added and removed. Instead, store the ID and look up using the type,
config handle and ID.
Signed-off-by: James Clarke <jrtc27@jrtc27.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=112541
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/vio.h | 1 +
arch/sparc/kernel/vio.c | 68 +++++++++++++++++++++++++++++++++++++++++---
2 files changed, 65 insertions(+), 4 deletions(-)
diff --git a/arch/sparc/include/asm/vio.h b/arch/sparc/include/asm/vio.h
index 8174f6c..9dca7a8 100644
--- a/arch/sparc/include/asm/vio.h
+++ b/arch/sparc/include/asm/vio.h
@@ -327,6 +327,7 @@ struct vio_dev {
int compat_len;
u64 dev_no;
+ u64 id;
unsigned long channel_id;
diff --git a/arch/sparc/kernel/vio.c b/arch/sparc/kernel/vio.c
index f6bb857..075d389 100644
--- a/arch/sparc/kernel/vio.c
+++ b/arch/sparc/kernel/vio.c
@@ -302,13 +302,16 @@ static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
if (!id) {
dev_set_name(&vdev->dev, "%s", bus_id_name);
vdev->dev_no = ~(u64)0;
+ vdev->id = ~(u64)0;
} else if (!cfg_handle) {
dev_set_name(&vdev->dev, "%s-%llu", bus_id_name, *id);
vdev->dev_no = *id;
+ vdev->id = ~(u64)0;
} else {
dev_set_name(&vdev->dev, "%s-%llu-%llu", bus_id_name,
*cfg_handle, *id);
vdev->dev_no = *cfg_handle;
+ vdev->id = *id;
}
vdev->dev.parent = parent;
@@ -351,27 +354,84 @@ static void vio_add(struct mdesc_handle *hp, u64 node)
(void) vio_create_one(hp, node, &root_vdev->dev);
}
+struct vio_md_node_query {
+ const char *type;
+ u64 dev_no;
+ u64 id;
+};
+
static int vio_md_node_match(struct device *dev, void *arg)
{
+ struct vio_md_node_query *query = (struct vio_md_node_query *) arg;
struct vio_dev *vdev = to_vio_dev(dev);
- if (vdev->mp == (u64) arg)
- return 1;
+ if (vdev->dev_no != query->dev_no)
+ return 0;
+ if (vdev->id != query->id)
+ return 0;
+ if (strcmp(vdev->type, query->type))
+ return 0;
- return 0;
+ return 1;
}
static void vio_remove(struct mdesc_handle *hp, u64 node)
{
+ const char *type;
+ const u64 *id, *cfg_handle;
+ u64 a;
+ struct vio_md_node_query query;
struct device *dev;
- dev = device_find_child(&root_vdev->dev, (void *) node,
+ type = mdesc_get_property(hp, node, "device-type", NULL);
+ if (!type) {
+ type = mdesc_get_property(hp, node, "name", NULL);
+ if (!type)
+ type = mdesc_node_name(hp, node);
+ }
+
+ query.type = type;
+
+ id = mdesc_get_property(hp, node, "id", NULL);
+ cfg_handle = NULL;
+ mdesc_for_each_arc(a, hp, node, MDESC_ARC_TYPE_BACK) {
+ u64 target;
+
+ target = mdesc_arc_target(hp, a);
+ cfg_handle = mdesc_get_property(hp, target,
+ "cfg-handle", NULL);
+ if (cfg_handle)
+ break;
+ }
+
+ if (!id) {
+ query.dev_no = ~(u64)0;
+ query.id = ~(u64)0;
+ } else if (!cfg_handle) {
+ query.dev_no = *id;
+ query.id = ~(u64)0;
+ } else {
+ query.dev_no = *cfg_handle;
+ query.id = *id;
+ }
+
+ dev = device_find_child(&root_vdev->dev, &query,
vio_md_node_match);
if (dev) {
printk(KERN_INFO "VIO: Removing device %s\n", dev_name(dev));
device_unregister(dev);
put_device(dev);
+ } else {
+ if (!id)
+ printk(KERN_ERR "VIO: Removed unknown %s node.\n",
+ type);
+ else if (!cfg_handle)
+ printk(KERN_ERR "VIO: Removed unknown %s node %llu.\n",
+ type, *id);
+ else
+ printk(KERN_ERR "VIO: Removed unknown %s node %llu-%llu.\n",
+ type, *cfg_handle, *id);
}
}
--
2.1.2.532.g19b5d50
From 6243075d665cb9fdc8a37f313142dd76ac876059 Mon Sep 17 00:00:00 2001
From: "Liam R. Howlett" <Liam.Howlett@Oracle.com>
Date: Tue, 30 May 2017 15:45:00 -0400
Subject: [PATCH 04/11] sparc/mm/hugepages: Fix setup_hugepagesz for invalid
values.
[ Upstream commit f322980b74a15e08f8c70a34a5864ecdbf957251 ]
hugetlb_bad_size needs to be called on invalid values. Also change the
pr_warn to a pr_err to better align with other platforms.
Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/mm/init_64.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index 0cda653..7ca1b9d 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -358,7 +358,8 @@ static int __init setup_hugepagesz(char *string)
}
if ((hv_pgsz_mask & cpu_pgsz_mask) == 0U) {
- pr_warn("hugepagesz=%llu not supported by MMU.\n",
+ hugetlb_bad_size();
+ pr_err("hugepagesz=%llu not supported by MMU.\n",
hugepage_size);
goto out;
}
--
2.1.2.532.g19b5d50
From 70e4e3f87743e86738d9629159a816fa10d63b2f Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:20 -0400
Subject: [PATCH 05/11] sparc64: reset mm cpumask after wrap
[ Upstream commit 588974857359861891f478a070b1dc7ae04a3880 ]
After a wrap (getting a new context version) a process must get a new
context id, which means that we would need to flush the context id from
the TLB before running for the first time with this ID on every CPU. But,
we use mm_cpumask to determine if this process has been running on this CPU
before, and this mask is not reset after a wrap. So, there are two possible
fixes for this issue:
1. Clear mm cpumask whenever mm gets a new context id
2. Unconditionally flush context every time process is running on a CPU
This patch implements the first solution
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/mm/init_64.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index 7ca1b9d..f8e30a3 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -759,6 +759,8 @@ void get_new_mmu_context(struct mm_struct *mm)
goto out;
}
}
+ if (mm->context.sparc64_ctx_val)
+ cpumask_clear(mm_cpumask(mm));
mmu_context_bmap[new_ctx>>6] |= (1UL << (new_ctx & 63));
new_ctx |= (tlb_context_cache & CTX_VERSION_MASK);
out:
--
2.1.2.532.g19b5d50
From d95ea5260755f26aad980da478b6cb133157a8fa Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:21 -0400
Subject: [PATCH 06/11] sparc64: combine activate_mm and switch_mm
[ Upstream commit 14d0334c6748ff2aedb3f2f7fdc51ee90a9b54e7 ]
The only difference between these two functions is that in activate_mm we
unconditionally flush context. However, there is no need to keep this
difference after fixing a bug where cpumask was not reset on a wrap. So, in
this patch we combine these.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_context_64.h | 21 +--------------------
1 file changed, 1 insertion(+), 20 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_context_64.h b/arch/sparc/include/asm/mmu_context_64.h
index 22fede6..734a134 100644
--- a/arch/sparc/include/asm/mmu_context_64.h
+++ b/arch/sparc/include/asm/mmu_context_64.h
@@ -133,26 +133,7 @@ static inline void switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, str
}
#define deactivate_mm(tsk,mm) do { } while (0)
-
-/* Activate a new MM instance for the current task. */
-static inline void activate_mm(struct mm_struct *active_mm, struct mm_struct *mm)
-{
- unsigned long flags;
- int cpu;
-
- spin_lock_irqsave(&mm->context.lock, flags);
- if (!CTX_VALID(mm->context))
- get_new_mmu_context(mm);
- cpu = smp_processor_id();
- if (!cpumask_test_cpu(cpu, mm_cpumask(mm)))
- cpumask_set_cpu(cpu, mm_cpumask(mm));
-
- load_secondary_context(mm);
- __flush_tlb_mm(CTX_HWBITS(mm->context), SECONDARY_CONTEXT);
- tsb_context_switch(mm);
- spin_unlock_irqrestore(&mm->context.lock, flags);
-}
-
+#define activate_mm(active_mm, mm) switch_mm(active_mm, mm, NULL)
#endif /* !(__ASSEMBLY__) */
#endif /* !(__SPARC64_MMU_CONTEXT_H) */
--
2.1.2.532.g19b5d50
From c181be17856ccccbf2d147c0818ae30028430a78 Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:22 -0400
Subject: [PATCH 07/11] sparc64: redefine first version
[ Upstream commit c4415235b2be0cc791572e8e7f7466ab8f73a2bf ]
CTX_FIRST_VERSION defines the first context version, but also it defines
first context. This patch redefines it to only include the first context
version.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_64.h | 2 +-
arch/sparc/mm/init_64.c | 6 +++---
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_64.h b/arch/sparc/include/asm/mmu_64.h
index f7de0db..83b36a5 100644
--- a/arch/sparc/include/asm/mmu_64.h
+++ b/arch/sparc/include/asm/mmu_64.h
@@ -52,7 +52,7 @@
#define CTX_NR_MASK TAG_CONTEXT_BITS
#define CTX_HW_MASK (CTX_NR_MASK | CTX_PGSZ_MASK)
-#define CTX_FIRST_VERSION ((_AC(1,UL) << CTX_VERSION_SHIFT) + _AC(1,UL))
+#define CTX_FIRST_VERSION BIT(CTX_VERSION_SHIFT)
#define CTX_VALID(__ctx) \
(!(((__ctx.sparc64_ctx_val) ^ tlb_context_cache) & CTX_VERSION_MASK))
#define CTX_HWBITS(__ctx) ((__ctx.sparc64_ctx_val) & CTX_HW_MASK)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index f8e30a3..63b5044 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -707,7 +707,7 @@ EXPORT_SYMBOL(__flush_dcache_range);
/* get_new_mmu_context() uses "cache + 1". */
DEFINE_SPINLOCK(ctx_alloc_lock);
-unsigned long tlb_context_cache = CTX_FIRST_VERSION - 1;
+unsigned long tlb_context_cache = CTX_FIRST_VERSION;
#define MAX_CTX_NR (1UL << CTX_NR_BITS)
#define CTX_BMAP_SLOTS BITS_TO_LONGS(MAX_CTX_NR)
DECLARE_BITMAP(mmu_context_bmap, MAX_CTX_NR);
@@ -738,9 +738,9 @@ void get_new_mmu_context(struct mm_struct *mm)
if (new_ctx >= ctx) {
int i;
new_ctx = (tlb_context_cache & CTX_VERSION_MASK) +
- CTX_FIRST_VERSION;
+ CTX_FIRST_VERSION + 1;
if (new_ctx == 1)
- new_ctx = CTX_FIRST_VERSION;
+ new_ctx = CTX_FIRST_VERSION + 1;
/* Don't call memset, for 16 entries that's just
* plain silly...
--
2.1.2.532.g19b5d50
From 38b0368264dac99ba6f0294b804fb61d40cb4cc6 Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:23 -0400
Subject: [PATCH 08/11] sparc64: add per-cpu mm of secondary contexts
[ Upstream commit 7a5b4bbf49fe86ce77488a70c5dccfe2d50d7a2d ]
The new wrap is going to use information from this array to figure out
mm's that currently have valid secondary contexts setup.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_context_64.h | 5 +++--
arch/sparc/mm/init_64.c | 1 +
2 files changed, 4 insertions(+), 2 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_context_64.h b/arch/sparc/include/asm/mmu_context_64.h
index 734a134..edb4524 100644
--- a/arch/sparc/include/asm/mmu_context_64.h
+++ b/arch/sparc/include/asm/mmu_context_64.h
@@ -19,6 +19,7 @@ extern spinlock_t ctx_alloc_lock;
extern unsigned long tlb_context_cache;
extern unsigned long mmu_context_bmap[];
+DECLARE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm);
void get_new_mmu_context(struct mm_struct *mm);
#ifdef CONFIG_SMP
void smp_new_mmu_context_version(void);
@@ -76,8 +77,9 @@ void __flush_tlb_mm(unsigned long, unsigned long);
static inline void switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, struct task_struct *tsk)
{
unsigned long ctx_valid, flags;
- int cpu;
+ int cpu = smp_processor_id();
+ per_cpu(per_cpu_secondary_mm, cpu) = mm;
if (unlikely(mm == &init_mm))
return;
@@ -123,7 +125,6 @@ static inline void switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, str
* for the first time, we must flush that context out of the
* local TLB.
*/
- cpu = smp_processor_id();
if (!ctx_valid || !cpumask_test_cpu(cpu, mm_cpumask(mm))) {
cpumask_set_cpu(cpu, mm_cpumask(mm));
__flush_tlb_mm(CTX_HWBITS(mm->context),
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index 63b5044..a4c0bc8 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -711,6 +711,7 @@ unsigned long tlb_context_cache = CTX_FIRST_VERSION;
#define MAX_CTX_NR (1UL << CTX_NR_BITS)
#define CTX_BMAP_SLOTS BITS_TO_LONGS(MAX_CTX_NR)
DECLARE_BITMAP(mmu_context_bmap, MAX_CTX_NR);
+DEFINE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm) = {0};
/* Caller does TLB context flushing on local CPU if necessary.
* The caller also ensures that CTX_VALID(mm->context) is false.
--
2.1.2.532.g19b5d50
From 22a0822c02284c0be544537ce350cf8f6736e39e Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:24 -0400
Subject: [PATCH 09/11] sparc64: new context wrap
[ Upstream commit a0582f26ec9dfd5360ea2f35dd9a1b026f8adda0 ]
The current wrap implementation has a race issue: it is called outside of
the ctx_alloc_lock, and also does not wait for all CPUs to complete the
wrap. This means that a thread can get a new context with a new version
and another thread might still be running with the same context. The
problem is especially severe on CPUs with shared TLBs, like sun4v. I used
the following test to very quickly reproduce the problem:
- start over 8K processes (must be more than context IDs)
- write and read values at a memory location in every process.
Very quickly memory corruptions start happening, and what we read back
does not equal what we wrote.
Several approaches were explored before settling on this one:
Approach 1:
Move smp_new_mmu_context_version() inside ctx_alloc_lock, and wait for
every process to complete the wrap. (Note: every CPU must WAIT before
leaving smp_new_mmu_context_version_client() until every one arrives).
This approach ends up with deadlocks, as some threads own locks which other
threads are waiting for, and they never receive softint until these threads
exit smp_new_mmu_context_version_client(). Since we do not allow the exit,
deadlock happens.
Approach 2:
Handle wrap right during mondo interrupt. Use etrap/rtrap to enter into
into C code, and issue new versions to every CPU.
This approach adds some overhead to runtime: in switch_mm() we must add
some checks to make sure that versions have not changed due to wrap while
we were loading the new secondary context. (could be protected by PSTATE_IE
but that degrades performance as on M7 and older CPUs as it takes 50 cycles
for each access). Also, we still need a global per-cpu array of MMs to know
where we need to load new contexts, otherwise we can change context to a
thread that is going way (if we received mondo between switch_mm() and
switch_to() time). Finally, there are some issues with window registers in
rtrap() when context IDs are changed during CPU mondo time.
The approach in this patch is the simplest and has almost no impact on
runtime. We use the array with mm's where last secondary contexts were
loaded onto CPUs and bump their versions to the new generation without
changing context IDs. If a new process comes in to get a context ID, it
will go through get_new_mmu_context() because of version mismatch. But the
running processes do not need to be interrupted. And wrap is quicker as we
do not need to xcall and wait for everyone to receive and complete wrap.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/mm/init_64.c | 81 ++++++++++++++++++++++++++++++++-----------------
1 file changed, 54 insertions(+), 27 deletions(-)
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index a4c0bc8..3c40ebd 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -713,6 +713,53 @@ unsigned long tlb_context_cache = CTX_FIRST_VERSION;
DECLARE_BITMAP(mmu_context_bmap, MAX_CTX_NR);
DEFINE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm) = {0};
+static void mmu_context_wrap(void)
+{
+ unsigned long old_ver = tlb_context_cache & CTX_VERSION_MASK;
+ unsigned long new_ver, new_ctx, old_ctx;
+ struct mm_struct *mm;
+ int cpu;
+
+ bitmap_zero(mmu_context_bmap, 1 << CTX_NR_BITS);
+
+ /* Reserve kernel context */
+ set_bit(0, mmu_context_bmap);
+
+ new_ver = (tlb_context_cache & CTX_VERSION_MASK) + CTX_FIRST_VERSION;
+ if (unlikely(new_ver == 0))
+ new_ver = CTX_FIRST_VERSION;
+ tlb_context_cache = new_ver;
+
+ /*
+ * Make sure that any new mm that are added into per_cpu_secondary_mm,
+ * are going to go through get_new_mmu_context() path.
+ */
+ mb();
+
+ /*
+ * Updated versions to current on those CPUs that had valid secondary
+ * contexts
+ */
+ for_each_online_cpu(cpu) {
+ /*
+ * If a new mm is stored after we took this mm from the array,
+ * it will go into get_new_mmu_context() path, because we
+ * already bumped the version in tlb_context_cache.
+ */
+ mm = per_cpu(per_cpu_secondary_mm, cpu);
+
+ if (unlikely(!mm || mm == &init_mm))
+ continue;
+
+ old_ctx = mm->context.sparc64_ctx_val;
+ if (likely((old_ctx & CTX_VERSION_MASK) == old_ver)) {
+ new_ctx = (old_ctx & ~CTX_VERSION_MASK) | new_ver;
+ set_bit(new_ctx & CTX_NR_MASK, mmu_context_bmap);
+ mm->context.sparc64_ctx_val = new_ctx;
+ }
+ }
+}
+
/* Caller does TLB context flushing on local CPU if necessary.
* The caller also ensures that CTX_VALID(mm->context) is false.
*
@@ -727,50 +774,30 @@ void get_new_mmu_context(struct mm_struct *mm)
{
unsigned long ctx, new_ctx;
unsigned long orig_pgsz_bits;
- int new_version;
spin_lock(&ctx_alloc_lock);
+retry:
+ /* wrap might have happened, test again if our context became valid */
+ if (unlikely(CTX_VALID(mm->context)))
+ goto out;
orig_pgsz_bits = (mm->context.sparc64_ctx_val & CTX_PGSZ_MASK);
ctx = (tlb_context_cache + 1) & CTX_NR_MASK;
new_ctx = find_next_zero_bit(mmu_context_bmap, 1 << CTX_NR_BITS, ctx);
- new_version = 0;
if (new_ctx >= (1 << CTX_NR_BITS)) {
new_ctx = find_next_zero_bit(mmu_context_bmap, ctx, 1);
if (new_ctx >= ctx) {
- int i;
- new_ctx = (tlb_context_cache & CTX_VERSION_MASK) +
- CTX_FIRST_VERSION + 1;
- if (new_ctx == 1)
- new_ctx = CTX_FIRST_VERSION + 1;
-
- /* Don't call memset, for 16 entries that's just
- * plain silly...
- */
- mmu_context_bmap[0] = 3;
- mmu_context_bmap[1] = 0;
- mmu_context_bmap[2] = 0;
- mmu_context_bmap[3] = 0;
- for (i = 4; i < CTX_BMAP_SLOTS; i += 4) {
- mmu_context_bmap[i + 0] = 0;
- mmu_context_bmap[i + 1] = 0;
- mmu_context_bmap[i + 2] = 0;
- mmu_context_bmap[i + 3] = 0;
- }
- new_version = 1;
- goto out;
+ mmu_context_wrap();
+ goto retry;
}
}
if (mm->context.sparc64_ctx_val)
cpumask_clear(mm_cpumask(mm));
mmu_context_bmap[new_ctx>>6] |= (1UL << (new_ctx & 63));
new_ctx |= (tlb_context_cache & CTX_VERSION_MASK);
-out:
tlb_context_cache = new_ctx;
mm->context.sparc64_ctx_val = new_ctx | orig_pgsz_bits;
+out:
spin_unlock(&ctx_alloc_lock);
-
- if (unlikely(new_version))
- smp_new_mmu_context_version();
}
static int numa_enabled = 1;
--
2.1.2.532.g19b5d50
From 7ac330173e5f7386de4fb7bc2fd9d7d9b6c64142 Mon Sep 17 00:00:00 2001
From: Pavel Tatashin <pasha.tatashin@oracle.com>
Date: Wed, 31 May 2017 11:25:25 -0400
Subject: [PATCH 10/11] sparc64: delete old wrap code
[ Upstream commit 0197e41ce70511dc3b71f7fefa1a676e2b5cd60b ]
The old method that is using xcall and softint to get new context id is
deleted, as it is replaced by a method of using per_cpu_secondary_mm
without xcall to perform the context wrap.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/include/asm/mmu_context_64.h | 6 ------
arch/sparc/include/asm/pil.h | 1 -
arch/sparc/kernel/kernel.h | 1 -
arch/sparc/kernel/smp_64.c | 31 -------------------------------
arch/sparc/kernel/ttable_64.S | 2 +-
arch/sparc/mm/ultra.S | 5 -----
6 files changed, 1 insertion(+), 45 deletions(-)
diff --git a/arch/sparc/include/asm/mmu_context_64.h b/arch/sparc/include/asm/mmu_context_64.h
index edb4524..2cddcda 100644
--- a/arch/sparc/include/asm/mmu_context_64.h
+++ b/arch/sparc/include/asm/mmu_context_64.h
@@ -21,12 +21,6 @@ extern unsigned long mmu_context_bmap[];
DECLARE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm);
void get_new_mmu_context(struct mm_struct *mm);
-#ifdef CONFIG_SMP
-void smp_new_mmu_context_version(void);
-#else
-#define smp_new_mmu_context_version() do { } while (0)
-#endif
-
int init_new_context(struct task_struct *tsk, struct mm_struct *mm);
void destroy_context(struct mm_struct *mm);
diff --git a/arch/sparc/include/asm/pil.h b/arch/sparc/include/asm/pil.h
index 26693703..522b43d 100644
--- a/arch/sparc/include/asm/pil.h
+++ b/arch/sparc/include/asm/pil.h
@@ -20,7 +20,6 @@
#define PIL_SMP_CALL_FUNC 1
#define PIL_SMP_RECEIVE_SIGNAL 2
#define PIL_SMP_CAPTURE 3
-#define PIL_SMP_CTX_NEW_VERSION 4
#define PIL_DEVICE_IRQ 5
#define PIL_SMP_CALL_FUNC_SNGL 6
#define PIL_DEFERRED_PCR_WORK 7
diff --git a/arch/sparc/kernel/kernel.h b/arch/sparc/kernel/kernel.h
index c9804551..6ae1e77 100644
--- a/arch/sparc/kernel/kernel.h
+++ b/arch/sparc/kernel/kernel.h
@@ -37,7 +37,6 @@ void handle_stdfmna(struct pt_regs *regs, unsigned long sfar, unsigned long sfsr
/* smp_64.c */
void __irq_entry smp_call_function_client(int irq, struct pt_regs *regs);
void __irq_entry smp_call_function_single_client(int irq, struct pt_regs *regs);
-void __irq_entry smp_new_mmu_context_version_client(int irq, struct pt_regs *regs);
void __irq_entry smp_penguin_jailcell(int irq, struct pt_regs *regs);
void __irq_entry smp_receive_signal_client(int irq, struct pt_regs *regs);
diff --git a/arch/sparc/kernel/smp_64.c b/arch/sparc/kernel/smp_64.c
index b3bc0ac..fdf3104 100644
--- a/arch/sparc/kernel/smp_64.c
+++ b/arch/sparc/kernel/smp_64.c
@@ -964,37 +964,6 @@ void flush_dcache_page_all(struct mm_struct *mm, struct page *page)
preempt_enable();
}
-void __irq_entry smp_new_mmu_context_version_client(int irq, struct pt_regs *regs)
-{
- struct mm_struct *mm;
- unsigned long flags;
-
- clear_softint(1 << irq);
-
- /* See if we need to allocate a new TLB context because
- * the version of the one we are using is now out of date.
- */
- mm = current->active_mm;
- if (unlikely(!mm || (mm == &init_mm)))
- return;
-
- spin_lock_irqsave(&mm->context.lock, flags);
-
- if (unlikely(!CTX_VALID(mm->context)))
- get_new_mmu_context(mm);
-
- spin_unlock_irqrestore(&mm->context.lock, flags);
-
- load_secondary_context(mm);
- __flush_tlb_mm(CTX_HWBITS(mm->context),
- SECONDARY_CONTEXT);
-}
-
-void smp_new_mmu_context_version(void)
-{
- smp_cross_call(&xcall_new_mmu_context_version, 0, 0, 0);
-}
-
#ifdef CONFIG_KGDB
void kgdb_roundup_cpus(unsigned long flags)
{
diff --git a/arch/sparc/kernel/ttable_64.S b/arch/sparc/kernel/ttable_64.S
index 7bd8f65..efe93ab 100644
--- a/arch/sparc/kernel/ttable_64.S
+++ b/arch/sparc/kernel/ttable_64.S
@@ -50,7 +50,7 @@ tl0_resv03e: BTRAP(0x3e) BTRAP(0x3f) BTRAP(0x40)
tl0_irq1: TRAP_IRQ(smp_call_function_client, 1)
tl0_irq2: TRAP_IRQ(smp_receive_signal_client, 2)
tl0_irq3: TRAP_IRQ(smp_penguin_jailcell, 3)
-tl0_irq4: TRAP_IRQ(smp_new_mmu_context_version_client, 4)
+tl0_irq4: BTRAP(0x44)
#else
tl0_irq1: BTRAP(0x41)
tl0_irq2: BTRAP(0x42)
diff --git a/arch/sparc/mm/ultra.S b/arch/sparc/mm/ultra.S
index 5d2fd6c..fcf4d27 100644
--- a/arch/sparc/mm/ultra.S
+++ b/arch/sparc/mm/ultra.S
@@ -971,11 +971,6 @@ xcall_capture:
wr %g0, (1 << PIL_SMP_CAPTURE), %set_softint
retry
- .globl xcall_new_mmu_context_version
-xcall_new_mmu_context_version:
- wr %g0, (1 << PIL_SMP_CTX_NEW_VERSION), %set_softint
- retry
-
#ifdef CONFIG_KGDB
.globl xcall_kgdb_capture
xcall_kgdb_capture:
--
2.1.2.532.g19b5d50
From 2233a227ab9c6ef561fefcbe3947f98caca066dd Mon Sep 17 00:00:00 2001
From: Jane Chu <jane.chu@oracle.com>
Date: Tue, 6 Jun 2017 14:32:29 -0600
Subject: [PATCH 11/11] arch/sparc: support NR_CPUS = 4096
[ Upstream commit c79a13734d104b5b147d7cb0870276ccdd660dae ]
Linux SPARC64 limits NR_CPUS to 4064 because init_cpu_send_mondo_info()
only allocates a single page for NR_CPUS mondo entries. Thus we cannot
use all 4096 CPUs on some SPARC platforms.
To fix, allocate (2^order) pages where order is set according to the size
of cpu_list for possible cpus. Since cpu_list_pa and cpu_mondo_block_pa
are not used in asm code, there are no imm13 offsets from the base PA
that will break because they can only reach one page.
Orabug: 25505750
Signed-off-by: Jane Chu <jane.chu@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Reviewed-by: Atish Patra <atish.patra@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
arch/sparc/Kconfig | 4 ++--
arch/sparc/kernel/irq_64.c | 17 +++++++++++++----
2 files changed, 15 insertions(+), 6 deletions(-)
diff --git a/arch/sparc/Kconfig b/arch/sparc/Kconfig
index 3db2543..1384d4c 100644
--- a/arch/sparc/Kconfig
+++ b/arch/sparc/Kconfig
@@ -192,9 +192,9 @@ config NR_CPUS
int "Maximum number of CPUs"
depends on SMP
range 2 32 if SPARC32
- range 2 1024 if SPARC64
+ range 2 4096 if SPARC64
default 32 if SPARC32
- default 64 if SPARC64
+ default 4096 if SPARC64
source kernel/Kconfig.hz
diff --git a/arch/sparc/kernel/irq_64.c b/arch/sparc/kernel/irq_64.c
index 4d0248a..99dd133 100644
--- a/arch/sparc/kernel/irq_64.c
+++ b/arch/sparc/kernel/irq_64.c
@@ -1034,17 +1034,26 @@ static void __init init_cpu_send_mondo_info(struct trap_per_cpu *tb)
{
#ifdef CONFIG_SMP
unsigned long page;
+ void *mondo, *p;
- BUILD_BUG_ON((NR_CPUS * sizeof(u16)) > (PAGE_SIZE - 64));
+ BUILD_BUG_ON((NR_CPUS * sizeof(u16)) > PAGE_SIZE);
+
+ /* Make sure mondo block is 64byte aligned */
+ p = kzalloc(127, GFP_KERNEL);
+ if (!p) {
+ prom_printf("SUN4V: Error, cannot allocate mondo block.\n");
+ prom_halt();
+ }
+ mondo = (void *)(((unsigned long)p + 63) & ~0x3f);
+ tb->cpu_mondo_block_pa = __pa(mondo);
page = get_zeroed_page(GFP_KERNEL);
if (!page) {
- prom_printf("SUN4V: Error, cannot allocate cpu mondo page.\n");
+ prom_printf("SUN4V: Error, cannot allocate cpu list page.\n");
prom_halt();
}
- tb->cpu_mondo_block_pa = __pa(page);
- tb->cpu_list_pa = __pa(page + 64);
+ tb->cpu_list_pa = __pa(page);
#endif
}
--
2.1.2.532.g19b5d50
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCHES] Networking and Sparc
2017-06-08 3:19 [PATCHES] Networking and Sparc David Miller
@ 2017-06-08 7:09 ` Greg KH
0 siblings, 0 replies; 2+ messages in thread
From: Greg KH @ 2017-06-08 7:09 UTC (permalink / raw)
To: David Miller; +Cc: stable
On Wed, Jun 07, 2017 at 11:19:25PM -0400, David Miller wrote:
>
> Please queue up the following sparc and networking bug fixes for 4.9
> and 4.11 -stable respectively.
Many thanks for these, all now queued up.
greg k-h
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2017-06-08 7:09 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2017-06-08 3:19 [PATCHES] Networking and Sparc David Miller
2017-06-08 7:09 ` Greg KH
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).