* [RFC][PATCH v4 01/18] Add a new struct for device to manipulate external buffer.
From: xiaohui.xin @ 2010-04-25 9:19 UTC (permalink / raw)
To: netdev, kvm, linux-kernel, mst, mingo, davem, jdike; +Cc: Xin Xiaohui
From: Xin Xiaohui <xiaohui.xin@intel.com>
Signed-off-by: Xin Xiaohui <xiaohui.xin@intel.com>
Signed-off-by: Zhao Yu <yzhao81@gmail.com>
Reviewed-by: Jeff Dike <jdike@linux.intel.com>
---
include/linux/netdevice.h | 19 ++++++++++++++++++-
1 files changed, 18 insertions(+), 1 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index c79a88b..bf79756 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -530,6 +530,22 @@ struct netdev_queue {
unsigned long tx_dropped;
} ____cacheline_aligned_in_smp;
+/* Add a structure in structure net_device, the new field is
+ * named as mp_port. It's for mediate passthru (zero-copy).
+ * It contains the capability for the net device driver,
+ * a socket, and an external buffer creator, external means
+ * skb buffer belongs to the device may not be allocated from
+ * kernel space.
+ */
+struct mpassthru_port {
+ int hdr_len;
+ int data_len;
+ int npages;
+ unsigned flags;
+ struct socket *sock;
+ struct skb_external_page *(*ctor)(struct mpassthru_port *,
+ struct sk_buff *, int);
+};
/*
* This structure defines the management hooks for network devices.
@@ -952,7 +968,8 @@ struct net_device {
struct macvlan_port *macvlan_port;
/* GARP */
struct garp_port *garp_port;
-
+ /* mpassthru */
+ struct mpassthru_port *mp_port;
/* class/net/name entry */
struct device dev;
/* space for optional device, statistics, and wireless sysfs groups */
--
1.5.4.4
^ permalink raw reply related
* [RFC][PATCH v4 18/18] Provides multiple submits and async notifications
From: xiaohui.xin @ 2010-04-25 9:20 UTC (permalink / raw)
To: netdev, kvm, linux-kernel, mst, mingo, davem, jdike; +Cc: Xin Xiaohui
In-Reply-To: <1272187206-18534-17-git-send-email-xiaohui.xin@intel.com>
From: Xin Xiaohui <xiaohui.xin@intel.com>
Provides multiple submits and asynchronous notifications.
The vhost-net backend now only supports synchronous send/recv
operations. The patch provides multiple submits and asynchronous
notifications. This is needed for zero-copy case.
Signed-off-by: Xin Xiaohui <xiaohui.xin@intel.com>
---
drivers/vhost/net.c | 236 +++++++++++++++++++++++++++++++++++++++++++++++-
drivers/vhost/vhost.c | 120 ++++++++++++++-----------
drivers/vhost/vhost.h | 14 +++
3 files changed, 314 insertions(+), 56 deletions(-)
diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
index 38989d1..18f6c41 100644
--- a/drivers/vhost/net.c
+++ b/drivers/vhost/net.c
@@ -23,6 +23,8 @@
#include <linux/if_arp.h>
#include <linux/if_tun.h>
#include <linux/if_macvlan.h>
+#include <linux/mpassthru.h>
+#include <linux/aio.h>
#include <net/sock.h>
@@ -48,6 +50,7 @@ struct vhost_net {
struct vhost_dev dev;
struct vhost_virtqueue vqs[VHOST_NET_VQ_MAX];
struct vhost_poll poll[VHOST_NET_VQ_MAX];
+ struct kmem_cache *cache;
/* Tells us whether we are polling a socket for TX.
* We only do this when socket buffer fills up.
* Protected by tx vq lock. */
@@ -92,11 +95,138 @@ static void tx_poll_start(struct vhost_net *net, struct socket *sock)
net->tx_poll_state = VHOST_NET_POLL_STARTED;
}
+struct kiocb *notify_dequeue(struct vhost_virtqueue *vq)
+{
+ struct kiocb *iocb = NULL;
+ unsigned long flags;
+
+ spin_lock_irqsave(&vq->notify_lock, flags);
+ if (!list_empty(&vq->notifier)) {
+ iocb = list_first_entry(&vq->notifier,
+ struct kiocb, ki_list);
+ list_del(&iocb->ki_list);
+ }
+ spin_unlock_irqrestore(&vq->notify_lock, flags);
+ return iocb;
+}
+
+static void handle_iocb(struct kiocb *iocb)
+{
+ struct vhost_virtqueue *vq = iocb->private;
+ unsigned long flags;
+
+ spin_lock_irqsave(&vq->notify_lock, flags);
+ list_add_tail(&iocb->ki_list, &vq->notifier);
+ spin_unlock_irqrestore(&vq->notify_lock, flags);
+}
+
+static int is_async_vq(struct vhost_virtqueue *vq)
+{
+ return (vq->link_state == VHOST_VQ_LINK_ASYNC);
+}
+
+static void handle_async_rx_events_notify(struct vhost_net *net,
+ struct vhost_virtqueue *vq,
+ struct socket *sock)
+{
+ struct kiocb *iocb = NULL;
+ struct vhost_log *vq_log = NULL;
+ int rx_total_len = 0;
+ unsigned int head, log, in, out;
+ int size;
+
+ if (!is_async_vq(vq))
+ return;
+
+ if (sock->sk->sk_data_ready)
+ sock->sk->sk_data_ready(sock->sk, 0);
+
+ vq_log = unlikely(vhost_has_feature(&net->dev, VHOST_F_LOG_ALL)) ?
+ vq->log : NULL;
+
+ while ((iocb = notify_dequeue(vq)) != NULL) {
+ vhost_add_used_and_signal(&net->dev, vq,
+ iocb->ki_pos, iocb->ki_nbytes);
+ size = iocb->ki_nbytes;
+ head = iocb->ki_pos;
+ rx_total_len += iocb->ki_nbytes;
+
+ if (iocb->ki_dtor)
+ iocb->ki_dtor(iocb);
+ kmem_cache_free(net->cache, iocb);
+
+ /* when log is enabled, recomputing the log info is needed,
+ * since these buffers are in async queue, and may not get
+ * the log info before.
+ */
+ if (unlikely(vq_log)) {
+ if (!log)
+ __vhost_get_vq_desc(&net->dev, vq, vq->iov,
+ ARRAY_SIZE(vq->iov),
+ &out, &in, vq_log,
+ &log, head);
+ vhost_log_write(vq, vq_log, log, size);
+ }
+ if (unlikely(rx_total_len >= VHOST_NET_WEIGHT)) {
+ vhost_poll_queue(&vq->poll);
+ break;
+ }
+ }
+}
+
+static void handle_async_tx_events_notify(struct vhost_net *net,
+ struct vhost_virtqueue *vq)
+{
+ struct kiocb *iocb = NULL;
+ int tx_total_len = 0;
+
+ if (!is_async_vq(vq))
+ return;
+
+ while ((iocb = notify_dequeue(vq)) != NULL) {
+ vhost_add_used_and_signal(&net->dev, vq,
+ iocb->ki_pos, 0);
+ tx_total_len += iocb->ki_nbytes;
+
+ if (iocb->ki_dtor)
+ iocb->ki_dtor(iocb);
+
+ kmem_cache_free(net->cache, iocb);
+ if (unlikely(tx_total_len >= VHOST_NET_WEIGHT)) {
+ vhost_poll_queue(&vq->poll);
+ break;
+ }
+ }
+}
+
+static struct kiocb *create_iocb(struct vhost_net *net,
+ struct vhost_virtqueue *vq,
+ unsigned head)
+{
+ struct kiocb *iocb = NULL;
+
+ if (!is_async_vq(vq))
+ return NULL;
+
+ iocb = kmem_cache_zalloc(net->cache, GFP_KERNEL);
+ if (!iocb)
+ return NULL;
+ iocb->private = vq;
+ iocb->ki_pos = head;
+ iocb->ki_dtor = handle_iocb;
+ if (vq == &net->dev.vqs[VHOST_NET_VQ_RX]) {
+ iocb->ki_user_data = vq->num;
+ iocb->ki_iovec = vq->hdr;
+ }
+ return iocb;
+}
+
/* Expects to be always run from workqueue - which acts as
* read-size critical section for our kind of RCU. */
static void handle_tx(struct vhost_net *net)
{
struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_TX];
+ struct kiocb *iocb = NULL;
unsigned head, out, in, s;
struct msghdr msg = {
.msg_name = NULL,
@@ -129,6 +259,8 @@ static void handle_tx(struct vhost_net *net)
tx_poll_stop(net);
hdr_size = vq->hdr_size;
+ handle_async_tx_events_notify(net, vq);
+
for (;;) {
head = vhost_get_vq_desc(&net->dev, vq, vq->iov,
ARRAY_SIZE(vq->iov),
@@ -156,6 +288,13 @@ static void handle_tx(struct vhost_net *net)
/* Skip header. TODO: support TSO. */
s = move_iovec_hdr(vq->iov, vq->hdr, hdr_size, out);
msg.msg_iovlen = out;
+
+ if (is_async_vq(vq)) {
+ iocb = create_iocb(net, vq, head);
+ if (!iocb)
+ break;
+ }
+
len = iov_length(vq->iov, out);
/* Sanity check */
if (!len) {
@@ -165,12 +304,18 @@ static void handle_tx(struct vhost_net *net)
break;
}
/* TODO: Check specific error and bomb out unless ENOBUFS? */
- err = sock->ops->sendmsg(NULL, sock, &msg, len);
+ err = sock->ops->sendmsg(iocb, sock, &msg, len);
if (unlikely(err < 0)) {
+ if (is_async_vq(vq))
+ kmem_cache_free(net->cache, iocb);
vhost_discard_vq_desc(vq);
tx_poll_start(net, sock);
break;
}
+
+ if (is_async_vq(vq))
+ continue;
+
if (err != len)
pr_err("Truncated TX packet: "
" len %d != %zd\n", err, len);
@@ -182,6 +327,8 @@ static void handle_tx(struct vhost_net *net)
}
}
+ handle_async_tx_events_notify(net, vq);
+
mutex_unlock(&vq->mutex);
unuse_mm(net->dev.mm);
}
@@ -191,6 +338,7 @@ static void handle_tx(struct vhost_net *net)
static void handle_rx(struct vhost_net *net)
{
struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_RX];
+ struct kiocb *iocb = NULL;
unsigned head, out, in, log, s;
struct vhost_log *vq_log;
struct msghdr msg = {
@@ -211,7 +359,8 @@ static void handle_rx(struct vhost_net *net)
int err;
size_t hdr_size;
struct socket *sock = rcu_dereference(vq->private_data);
- if (!sock || skb_queue_empty(&sock->sk->sk_receive_queue))
+ if (!sock || (skb_queue_empty(&sock->sk->sk_receive_queue) &&
+ vq->link_state == VHOST_VQ_LINK_SYNC))
return;
use_mm(net->dev.mm);
@@ -219,9 +368,17 @@ static void handle_rx(struct vhost_net *net)
vhost_disable_notify(vq);
hdr_size = vq->hdr_size;
+ /* In async cases, when write log is enabled, in case the submitted
+ * buffers did not get log info before the log enabling, so we'd
+ * better recompute the log info when needed. We do this in
+ * handle_async_rx_events_notify().
+ */
+
vq_log = unlikely(vhost_has_feature(&net->dev, VHOST_F_LOG_ALL)) ?
vq->log : NULL;
+ handle_async_rx_events_notify(net, vq, sock);
+
for (;;) {
head = vhost_get_vq_desc(&net->dev, vq, vq->iov,
ARRAY_SIZE(vq->iov),
@@ -250,6 +407,13 @@ static void handle_rx(struct vhost_net *net)
s = move_iovec_hdr(vq->iov, vq->hdr, hdr_size, in);
msg.msg_iovlen = in;
len = iov_length(vq->iov, in);
+
+ if (is_async_vq(vq)) {
+ iocb = create_iocb(net, vq, head);
+ if (!iocb)
+ break;
+ }
+
/* Sanity check */
if (!len) {
vq_err(vq, "Unexpected header len for RX: "
@@ -257,13 +421,20 @@ static void handle_rx(struct vhost_net *net)
iov_length(vq->hdr, s), hdr_size);
break;
}
- err = sock->ops->recvmsg(NULL, sock, &msg,
+
+ err = sock->ops->recvmsg(iocb, sock, &msg,
len, MSG_DONTWAIT | MSG_TRUNC);
/* TODO: Check specific error and bomb out unless EAGAIN? */
if (err < 0) {
+ if (is_async_vq(vq))
+ kmem_cache_free(net->cache, iocb);
vhost_discard_vq_desc(vq);
break;
}
+
+ if (is_async_vq(vq))
+ continue;
+
/* TODO: Should check and handle checksum. */
if (err > len) {
pr_err("Discarded truncated rx packet: "
@@ -289,6 +460,8 @@ static void handle_rx(struct vhost_net *net)
}
}
+ handle_async_rx_events_notify(net, vq, sock);
+
mutex_unlock(&vq->mutex);
unuse_mm(net->dev.mm);
}
@@ -342,6 +515,7 @@ static int vhost_net_open(struct inode *inode, struct file *f)
vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, POLLOUT);
vhost_poll_init(n->poll + VHOST_NET_VQ_RX, handle_rx_net, POLLIN);
n->tx_poll_state = VHOST_NET_POLL_DISABLED;
+ n->cache = NULL;
f->private_data = n;
@@ -405,6 +579,18 @@ static void vhost_net_flush(struct vhost_net *n)
vhost_net_flush_vq(n, VHOST_NET_VQ_RX);
}
+static void vhost_async_cleanup(struct vhost_net *n)
+{
+ /* clean the notifier */
+ struct vhost_virtqueue *vq = &n->dev.vqs[VHOST_NET_VQ_RX];
+ struct kiocb *iocb = NULL;
+ if (n->cache) {
+ while ((iocb = notify_dequeue(vq)) != NULL)
+ kmem_cache_free(n->cache, iocb);
+ kmem_cache_destroy(n->cache);
+ }
+}
+
static int vhost_net_release(struct inode *inode, struct file *f)
{
struct vhost_net *n = f->private_data;
@@ -421,6 +607,7 @@ static int vhost_net_release(struct inode *inode, struct file *f)
/* We do an extra flush before freeing memory,
* since jobs can re-queue themselves. */
vhost_net_flush(n);
+ vhost_async_cleanup(n);
kfree(n);
return 0;
}
@@ -472,21 +659,58 @@ static struct socket *get_tap_socket(int fd)
return sock;
}
-static struct socket *get_socket(int fd)
+static struct socket *get_mp_socket(int fd)
+{
+ struct file *file = fget(fd);
+ struct socket *sock;
+ if (!file)
+ return ERR_PTR(-EBADF);
+ sock = mp_get_socket(file);
+ if (IS_ERR(sock))
+ fput(file);
+ return sock;
+}
+
+static struct socket *get_socket(struct vhost_virtqueue *vq, int fd,
+ enum vhost_vq_link_state *state)
{
struct socket *sock;
/* special case to disable backend */
if (fd == -1)
return NULL;
+
+ *state = VHOST_VQ_LINK_SYNC;
+
sock = get_raw_socket(fd);
if (!IS_ERR(sock))
return sock;
sock = get_tap_socket(fd);
if (!IS_ERR(sock))
return sock;
+ sock = get_mp_socket(fd);
+ if (!IS_ERR(sock)) {
+ *state = VHOST_VQ_LINK_ASYNC;
+ return sock;
+ }
return ERR_PTR(-ENOTSOCK);
}
+static void vhost_init_link_state(struct vhost_net *n, int index)
+{
+ struct vhost_virtqueue *vq = n->vqs + index;
+
+ WARN_ON(!mutex_is_locked(&vq->mutex));
+ if (vq->link_state == VHOST_VQ_LINK_ASYNC) {
+ INIT_LIST_HEAD(&vq->notifier);
+ spin_lock_init(&vq->notify_lock);
+ if (!n->cache) {
+ n->cache = kmem_cache_create("vhost_kiocb",
+ sizeof(struct kiocb), 0,
+ SLAB_HWCACHE_ALIGN, NULL);
+ }
+ }
+}
+
static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
{
struct socket *sock, *oldsock;
@@ -510,12 +734,14 @@ static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
r = -EFAULT;
goto err_vq;
}
- sock = get_socket(fd);
+ sock = get_socket(vq, fd, &vq->link_state);
if (IS_ERR(sock)) {
r = PTR_ERR(sock);
goto err_vq;
}
+ vhost_init_link_state(n, index);
+
/* start polling new socket */
oldsock = vq->private_data;
if (sock == oldsock)
diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
index 3f10194..b39e47c 100644
--- a/drivers/vhost/vhost.c
+++ b/drivers/vhost/vhost.c
@@ -860,61 +860,17 @@ static unsigned get_indirect(struct vhost_dev *dev, struct vhost_virtqueue *vq,
return 0;
}
-/* This looks in the virtqueue and for the first available buffer, and converts
- * it to an iovec for convenient access. Since descriptors consist of some
- * number of output then some number of input descriptors, it's actually two
- * iovecs, but we pack them into one and note how many of each there were.
- *
- * This function returns the descriptor number found, or vq->num (which
- * is never a valid descriptor number) if none was found. */
-unsigned vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
- struct iovec iov[], unsigned int iov_size,
- unsigned int *out_num, unsigned int *in_num,
- struct vhost_log *log, unsigned int *log_num)
+/* This computes the log info according to the index of buffer */
+unsigned __vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
+ struct iovec iov[], unsigned int iov_size,
+ unsigned int *out_num, unsigned int *in_num,
+ struct vhost_log *log, unsigned int *log_num,
+ unsigned int head)
{
struct vring_desc desc;
- unsigned int i, head, found = 0;
- u16 last_avail_idx;
+ unsigned int i = head, found = 0;
int ret;
- /* Check it isn't doing very strange things with descriptor numbers. */
- last_avail_idx = vq->last_avail_idx;
- if (get_user(vq->avail_idx, &vq->avail->idx)) {
- vq_err(vq, "Failed to access avail idx at %p\n",
- &vq->avail->idx);
- return vq->num;
- }
-
- if ((u16)(vq->avail_idx - last_avail_idx) > vq->num) {
- vq_err(vq, "Guest moved used index from %u to %u",
- last_avail_idx, vq->avail_idx);
- return vq->num;
- }
-
- /* If there's nothing new since last we looked, return invalid. */
- if (vq->avail_idx == last_avail_idx)
- return vq->num;
-
- /* Only get avail ring entries after they have been exposed by guest. */
- smp_rmb();
-
- /* Grab the next descriptor number they're advertising, and increment
- * the index we've seen. */
- if (get_user(head, &vq->avail->ring[last_avail_idx % vq->num])) {
- vq_err(vq, "Failed to read head: idx %d address %p\n",
- last_avail_idx,
- &vq->avail->ring[last_avail_idx % vq->num]);
- return vq->num;
- }
-
- /* If their number is silly, that's an error. */
- if (head >= vq->num) {
- vq_err(vq, "Guest says index %u > %u is available",
- head, vq->num);
- return vq->num;
- }
-
- /* When we start there are none of either input nor output. */
*out_num = *in_num = 0;
if (unlikely(log))
*log_num = 0;
@@ -978,8 +934,70 @@ unsigned vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
*out_num += ret;
}
} while ((i = next_desc(&desc)) != -1);
+ return head;
+}
+
+/* This looks in the virtqueue and for the first available buffer, and converts
+ * it to an iovec for convenient access. Since descriptors consist of some
+ * number of output then some number of input descriptors, it's actually two
+ * iovecs, but we pack them into one and note how many of each there were.
+ *
+ * This function returns the descriptor number found, or vq->num (which
+ * is never a valid descriptor number) if none was found. */
+unsigned vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
+ struct iovec iov[], unsigned int iov_size,
+ unsigned int *out_num, unsigned int *in_num,
+ struct vhost_log *log, unsigned int *log_num)
+{
+ struct vring_desc desc;
+ unsigned int i, head, found = 0;
+ u16 last_avail_idx;
+ int ret;
+
+ /* Check it isn't doing very strange things with descriptor numbers. */
+ last_avail_idx = vq->last_avail_idx;
+ if (get_user(vq->avail_idx, &vq->avail->idx)) {
+ vq_err(vq, "Failed to access avail idx at %p\n",
+ &vq->avail->idx);
+ return vq->num;
+ }
+
+ if ((u16)(vq->avail_idx - last_avail_idx) > vq->num) {
+ vq_err(vq, "Guest moved used index from %u to %u",
+ last_avail_idx, vq->avail_idx);
+ return vq->num;
+ }
+
+ /* If there's nothing new since last we looked, return invalid. */
+ if (vq->avail_idx == last_avail_idx)
+ return vq->num;
+
+ /* Only get avail ring entries after they have been exposed by guest. */
+ smp_rmb();
+
+ /* Grab the next descriptor number they're advertising, and increment
+ * the index we've seen. */
+ if (get_user(head, &vq->avail->ring[last_avail_idx % vq->num])) {
+ vq_err(vq, "Failed to read head: idx %d address %p\n",
+ last_avail_idx,
+ &vq->avail->ring[last_avail_idx % vq->num]);
+ return vq->num;
+ }
+
+ /* If their number is silly, that's an error. */
+ if (head >= vq->num) {
+ vq_err(vq, "Guest says index %u > %u is available",
+ head, vq->num);
+ return vq->num;
+ }
+
+ ret = __vhost_get_vq_desc(dev, vq, iov, iov_size,
+ out_num, in_num,
+ log, log_num, head);
/* On success, increment avail index. */
+ if (ret == vq->num)
+ return ret;
vq->last_avail_idx++;
return head;
}
diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
index 44591ba..3c9cbce 100644
--- a/drivers/vhost/vhost.h
+++ b/drivers/vhost/vhost.h
@@ -43,6 +43,11 @@ struct vhost_log {
u64 len;
};
+enum vhost_vq_link_state {
+ VHOST_VQ_LINK_SYNC = 0,
+ VHOST_VQ_LINK_ASYNC = 1,
+};
+
/* The virtqueue structure describes a queue attached to a device. */
struct vhost_virtqueue {
struct vhost_dev *dev;
@@ -96,6 +101,10 @@ struct vhost_virtqueue {
/* Log write descriptors */
void __user *log_base;
struct vhost_log log[VHOST_NET_MAX_SG];
+ /* Differiate async socket for 0-copy from normal */
+ enum vhost_vq_link_state link_state;
+ struct list_head notifier;
+ spinlock_t notify_lock;
};
struct vhost_dev {
@@ -124,6 +133,11 @@ unsigned vhost_get_vq_desc(struct vhost_dev *, struct vhost_virtqueue *,
struct iovec iov[], unsigned int iov_count,
unsigned int *out_num, unsigned int *in_num,
struct vhost_log *log, unsigned int *log_num);
+unsigned __vhost_get_vq_desc(struct vhost_dev *, struct vhost_virtqueue *,
+ struct iovec iov[], unsigned int iov_count,
+ unsigned int *out_num, unsigned int *in_num,
+ struct vhost_log *log, unsigned int *log_num,
+ unsigned int head);
void vhost_discard_vq_desc(struct vhost_virtqueue *);
int vhost_add_used(struct vhost_virtqueue *, unsigned int head, int len);
--
1.5.4.4
^ permalink raw reply related
* [RFC][PATCH v4 04/18] Add a function make external buffer owner to query capability.
From: xiaohui.xin @ 2010-04-25 9:19 UTC (permalink / raw)
To: netdev, kvm, linux-kernel, mst, mingo, davem, jdike; +Cc: Xin Xiaohui
In-Reply-To: <1272187206-18534-3-git-send-email-xiaohui.xin@intel.com>
From: Xin Xiaohui <xiaohui.xin@intel.com>
The external buffer owner can use the functions to get
the capability of the underlying NIC driver.
Signed-off-by: Xin Xiaohui <xiaohui.xin@intel.com>
Signed-off-by: Zhao Yu <yzhao81@gmail.com>
Reviewed-by: Jeff Dike <jdike@linux.intel.com>
---
include/linux/netdevice.h | 2 +
net/core/dev.c | 51 +++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 53 insertions(+), 0 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 3a1583b..2f9a4f2 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1599,6 +1599,8 @@ extern gro_result_t napi_gro_frags(struct napi_struct *napi);
extern int netdev_mp_port_attach(struct net_device *dev,
struct mpassthru_port *port);
extern void netdev_mp_port_detach(struct net_device *dev);
+int netdev_mp_port_prep(struct net_device *dev,
+ struct mpassthru_port *port);
static inline void napi_free_frags(struct napi_struct *napi)
{
diff --git a/net/core/dev.c b/net/core/dev.c
index 6a73fc7..4972bc4 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2492,6 +2492,57 @@ void netdev_mp_port_detach(struct net_device *dev)
}
EXPORT_SYMBOL(netdev_mp_port_detach);
+/* To support meidate passthru(zero-copy) with NIC driver,
+ * we'd better query NIC driver for the capability it can
+ * provide, especially for packet split mode, now we only
+ * query for the header size, and the payload a descriptor
+ * may carry. If a driver does not use the API to export,
+ * then we may try to use a default value, currently,
+ * we use the default value from an IGB driver. Now,
+ * it's only called by mpassthru device.
+ */
+#if defined(CONFIG_MEDIATE_PASSTHRU) || defined(CONFIG_MEDIATE_PASSTHRU_MODULE)
+int netdev_mp_port_prep(struct net_device *dev,
+ struct mpassthru_port *port)
+{
+ int rc;
+ int npages, data_len;
+ const struct net_device_ops *ops = dev->netdev_ops;
+
+ /* needed by packet split */
+
+ if (ops->ndo_mp_port_prep) {
+ rc = ops->ndo_mp_port_prep(dev, port);
+ if (rc)
+ return rc;
+ } else {
+ /* If the NIC driver did not report this,
+ * then we try to use default value.
+ */
+ port->hdr_len = 128;
+ port->data_len = 2048;
+ port->npages = 1;
+ }
+
+ if (port->hdr_len <= 0)
+ goto err;
+
+ npages = port->npages;
+ data_len = port->data_len;
+ if (npages <= 0 || npages > MAX_SKB_FRAGS ||
+ (data_len < PAGE_SIZE * (npages - 1) ||
+ data_len > PAGE_SIZE * npages))
+ goto err;
+
+ return 0;
+err:
+ dev_warn(&dev->dev, "invalid page constructor parameters\n");
+
+ return -EINVAL;
+}
+EXPORT_SYMBOL(netdev_mp_port_prep);
+#endif
+
/**
* netif_receive_skb - process receive buffer from network
* @skb: buffer to process
--
1.5.4.4
^ permalink raw reply related
* Re: [PATCH v3] rps: optimize rps_get_cpu()
From: Changli Gao @ 2010-04-25 8:03 UTC (permalink / raw)
To: David Miller; +Cc: therbert, eric.dumazet, netdev
In-Reply-To: <20100425.004842.225645379.davem@davemloft.net>
On Sun, Apr 25, 2010 at 3:48 PM, David Miller <davem@davemloft.net> wrote:
>
> Nevermind, currently we only depend upon the addresses in struct sock
> being 64-bit aligned not the protocol headers.
>
> It shouldn't be hard to make the protocol header addresses 64-bit
> aligned too. Simply setting the default NET_IP_ALIGN to '6' instead
> of '2' ought to be sufficient.
>
> skb->data upon alloc_skb() is 64-bit aligned.
>
> So if we skb_reserve(NET_IP_ALIGN '6'), then we have the ethernet
> header (14 bytes). And since 'saddr' is 12 bytes into struct iphdr it
> will be (6 + 14 + 12) == 32 bytes in from the original 64-bit aligned
> skb->data.
>
> Therefore, since skb->data is 64-bit aligned, skb->data plus a
> multiple of 8 (which 32 is) will also be 64-bit aligned, and that
> means iph->saddr will be 64-bit aligned.
>
But if there is a vlan header, extra 4-bytes are appended to the
ethernet header, so the addresses aren't aligned to 64-bit boundary
when we set NET_IP_ALIGN to 6.
--
Regards，
Changli Gao(xiaosuo@gmail.com)
^ permalink raw reply
* Re: [PATCH net-next-2.6] netns: rename unregister_pernet_subsys parameter
From: David Miller @ 2010-04-25 7:50 UTC (permalink / raw)
To: jpirko; +Cc: netdev, ebiederm
In-Reply-To: <20100425074138.GA2866@psychotron.redhat.com>
From: Jiri Pirko <jpirko@redhat.com>
Date: Sun, 25 Apr 2010 09:41:39 +0200
> Stay consistent with other functions and with comment also and name
> pernet_operations parameter properly.
>
> Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Applied, thanks Jiri.
^ permalink raw reply
* Re: [PATCH] RCU: don't turn off lockdep when find suspicious rcu_dereference_check() usage
From: David Miller @ 2010-04-25 7:49 UTC (permalink / raw)
To: johannes
Cc: paulmck, miles.lane, vgoyal, eparis, laijs, mingo, peterz,
linux-kernel, nauman, eric.dumazet, netdev, jens.axboe,
guijianfeng, lizf
In-Reply-To: <1272181534.3614.1.camel@jlt3.sipsolutions.net>
From: Johannes Berg <johannes@sipsolutions.net>
Date: Sun, 25 Apr 2010 09:45:34 +0200
> The station locking is a tad confusing, but I've added the right
> annotations already, should be coming to a kernel near you soon (i.e.
> are in net-2.6 right now).
Linus took in everything I have so it should be in Linus's tree
by now.
^ permalink raw reply
* Re: [PATCH v3] rps: optimize rps_get_cpu()
From: David Miller @ 2010-04-25 7:48 UTC (permalink / raw)
To: xiaosuo; +Cc: therbert, eric.dumazet, netdev
In-Reply-To: <20100425.003834.147984813.davem@davemloft.net>
From: David Miller <davem@davemloft.net>
Date: Sun, 25 Apr 2010 00:38:34 -0700 (PDT)
> From: Changli Gao <xiaosuo@gmail.com>
> Date: Sun, 25 Apr 2010 14:48:49 +0800
>
>> Are they always aligned to 64-bit boundary? I don't think so.
>
> If not than TCP stack should be crashing for past 15 years.
Nevermind, currently we only depend upon the addresses in struct sock
being 64-bit aligned not the protocol headers.
It shouldn't be hard to make the protocol header addresses 64-bit
aligned too. Simply setting the default NET_IP_ALIGN to '6' instead
of '2' ought to be sufficient.
skb->data upon alloc_skb() is 64-bit aligned.
So if we skb_reserve(NET_IP_ALIGN '6'), then we have the ethernet
header (14 bytes). And since 'saddr' is 12 bytes into struct iphdr it
will be (6 + 14 + 12) == 32 bytes in from the original 64-bit aligned
skb->data.
Therefore, since skb->data is 64-bit aligned, skb->data plus a
multiple of 8 (which 32 is) will also be 64-bit aligned, and that
means iph->saddr will be 64-bit aligned.
^ permalink raw reply
* Re: [PATCH] RCU: don't turn off lockdep when find suspicious rcu_dereference_check() usage
From: Johannes Berg @ 2010-04-25 7:45 UTC (permalink / raw)
To: paulmck
Cc: Miles Lane, Vivek Goyal, Eric Paris, Lai Jiangshan, Ingo Molnar,
Peter Zijlstra, LKML, nauman, eric.dumazet, netdev, Jens Axboe,
Gui Jianfeng, Li Zefan
In-Reply-To: <20100425023455.GM2440@linux.vnet.ibm.com>
On Sat, 2010-04-24 at 19:34 -0700, Paul E. McKenney wrote:
> > [ 51.912282] [ INFO: suspicious rcu_dereference_check() usage. ]
> > [ 51.912285] ---------------------------------------------------
> > [ 51.912289] net/mac80211/sta_info.c:886 invoked
> > rcu_dereference_check() without protection!
> > [ 51.912293]
> > [ 51.912293] other info that might help us debug this:
> > [ 51.912295]
> > [ 51.912298]
> > [ 51.912298] rcu_scheduler_active = 1, debug_locks = 1
> > [ 51.912302] no locks held by wpa_supplicant/3951.
> > [ 51.912305]
> > [ 51.912306] stack backtrace:
> > [ 51.912310] Pid: 3951, comm: wpa_supplicant Not tainted 2.6.34-rc5-git3 #22
> > [ 51.912314] Call Trace:
> > [ 51.912317] <IRQ> [<ffffffff81067fbe>] lockdep_rcu_dereference+0x9d/0xa5
> > [ 51.912345] [<ffffffffa014f9ae>]
> > ieee80211_find_sta_by_hw+0x46/0x10f [mac80211]
> > [ 51.912358] [<ffffffffa014fa8e>] ieee80211_find_sta+0x17/0x19 [mac80211]
> > [ 51.912373] [<ffffffffa01e50f2>] iwl_tx_queue_reclaim+0xdb/0x1b1 [iwlcore]
> > [ 51.912380] [<ffffffff8106842b>] ? mark_lock+0x2d/0x235
> > [ 51.912391] [<ffffffffa0252f1c>] iwl5000_rx_reply_tx+0x4a9/0x556 [iwlagn]
> > [ 51.912399] [<ffffffff8120a353>] ? is_swiotlb_buffer+0x2e/0x3b
> > [ 51.912407] [<ffffffffa024bbf4>] iwl_rx_handle+0x163/0x2b5 [iwlagn]
> > [ 51.912414] [<ffffffff81068904>] ? trace_hardirqs_on_caller+0xfa/0x13f
> > [ 51.912422] [<ffffffffa024c3ac>] iwl_irq_tasklet+0x2bb/0x3c0 [iwlagn]
> > [ 51.912429] [<ffffffff810411f3>] tasklet_action+0xa7/0x10f
> > [ 51.912435] [<ffffffff81042205>] __do_softirq+0x144/0x252
> > [ 51.912442] [<ffffffff81003a8c>] call_softirq+0x1c/0x34
> > [ 51.912447] [<ffffffff810050e4>] do_softirq+0x38/0x80
> > [ 51.912452] [<ffffffff81041cd2>] irq_exit+0x45/0x94
> > [ 51.912457] [<ffffffff81004829>] do_IRQ+0xad/0xc4
> > [ 51.912463] [<ffffffff810cbbd3>] ? might_fault+0x63/0xb3
> > [ 51.912470] [<ffffffff813cfb93>] ret_from_intr+0x0/0xf
> > [ 51.912474] <EOI> [<ffffffff810cbbd3>] ? might_fault+0x63/0xb3
> > [ 51.912484] [<ffffffff8106a75d>] ? lock_release+0x208/0x215
> > [ 51.912490] [<ffffffff810cbc1c>] might_fault+0xac/0xb3
> > [ 51.912495] [<ffffffff810cbbd3>] ? might_fault+0x63/0xb3
> > [ 51.912501] [<ffffffff812025e3>] __clear_user+0x15/0x59
> > [ 51.912508] [<ffffffff8100b2bc>] save_i387_xstate+0x9c/0x1bc
> > [ 51.912515] [<ffffffff81002276>] do_signal+0x240/0x686
> > [ 51.912521] [<ffffffff81002b9c>] ? sysret_check+0x27/0x62
> > [ 51.912527] [<ffffffff8106891e>] ? trace_hardirqs_on_caller+0x114/0x13f
> > [ 51.912533] [<ffffffff813cec80>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> > [ 51.912539] [<ffffffff810026e3>] do_notify_resume+0x27/0x5f
> > [ 51.912545] [<ffffffff813cec80>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> > [ 51.912551] [<ffffffff81002e86>] int_signal+0x12/0x17
>
> This is a repeat from last time that confused me at the time. I could
> do a hacky "fix" by putting an RCU read-side critical section around
> the for_each_sta_info() in ieee80211_find_sta_by_hw(), but I do not
> understand this code well enough to feel comfortable doing so.
>
> Johannes, any enlightenment?
The station locking is a tad confusing, but I've added the right
annotations already, should be coming to a kernel near you soon (i.e.
are in net-2.6 right now).
johannes
^ permalink raw reply
* [PATCH net-next-2.6] netns: rename unregister_pernet_subsys parameter
From: Jiri Pirko @ 2010-04-25 7:41 UTC (permalink / raw)
To: netdev; +Cc: davem, ebiederm
Stay consistent with other functions and with comment also and name
pernet_operations parameter properly.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
diff --git a/net/core/net_namespace.c b/net/core/net_namespace.c
index bd8c471..69a20bf 100644
--- a/net/core/net_namespace.c
+++ b/net/core/net_namespace.c
@@ -469,10 +469,10 @@ EXPORT_SYMBOL_GPL(register_pernet_subsys);
* addition run the exit method for all existing network
* namespaces.
*/
-void unregister_pernet_subsys(struct pernet_operations *module)
+void unregister_pernet_subsys(struct pernet_operations *ops)
{
mutex_lock(&net_mutex);
- unregister_pernet_operations(module);
+ unregister_pernet_operations(ops);
mutex_unlock(&net_mutex);
}
EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
^ permalink raw reply related
* Re: [PATCH v3] rps: optimize rps_get_cpu()
From: David Miller @ 2010-04-25 7:38 UTC (permalink / raw)
To: xiaosuo; +Cc: therbert, eric.dumazet, netdev
In-Reply-To: <z2h412e6f7f1004242348ibd5f96a2i7cd0557c37bb0921@mail.gmail.com>
From: Changli Gao <xiaosuo@gmail.com>
Date: Sun, 25 Apr 2010 14:48:49 +0800
> Are they always aligned to 64-bit boundary? I don't think so.
If not than TCP stack should be crashing for past 15 years.
^ permalink raw reply
* Re: [PATCH v3] rps: optimize rps_get_cpu()
From: Changli Gao @ 2010-04-25 6:48 UTC (permalink / raw)
To: David Miller; +Cc: therbert, eric.dumazet, netdev
In-Reply-To: <20100424.225128.52181685.davem@davemloft.net>
On Sun, Apr 25, 2010 at 1:51 PM, David Miller <davem@davemloft.net> wrote:
> From: Changli Gao <xiaosuo@gmail.com>
> Date: Sat, 24 Apr 2010 23:17:07 +0800
>
>> optimize rps_get_cpu().
>>
>> don't initialize ports when we can get the ports. one memory access for ports
>> than two.
>>
>> Signed-off-by: Changli Gao <xiaosuo@gmail.com>
>
> Applied, thanks.
>
> We can load both addresses in one go on 64-bit btw.
>
Are they always aligned to 64-bit boundary? I don't think so.
--
Regards，
Changli Gao(xiaosuo@gmail.com)
^ permalink raw reply
* Re: [PATCH v3] rps: optimize rps_get_cpu()
From: David Miller @ 2010-04-25 5:51 UTC (permalink / raw)
To: xiaosuo; +Cc: therbert, eric.dumazet, netdev
In-Reply-To: <1272122227-13070-1-git-send-email-xiaosuo@gmail.com>
From: Changli Gao <xiaosuo@gmail.com>
Date: Sat, 24 Apr 2010 23:17:07 +0800
> optimize rps_get_cpu().
>
> don't initialize ports when we can get the ports. one memory access for ports
> than two.
>
> Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Applied, thanks.
We can load both addresses in one go on 64-bit btw.
It seems we're just duplicating, one by one, the optimizations
we already do in INET_COMBINED_PORTS() and INET_ADDR_COOKIE().
:-)
^ permalink raw reply
* Re: [PATCH] e100: Fix the TX workqueue race
From: David Miller @ 2010-04-25 4:10 UTC (permalink / raw)
To: alan; +Cc: e1000-devel, netdev
In-Reply-To: <20100424.195859.193729555.davem@davemloft.net>
From: David Miller <davem@davemloft.net>
Date: Sat, 24 Apr 2010 19:58:59 -0700 (PDT)
> Please generate this patch against net-2.6 so I can apply it, thanks
> Alan.
Nevermind, I took care of this for you.
------------------------------------------------------------------------------
_______________________________________________
E1000-devel mailing list
E1000-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
^ permalink raw reply
* Re: 2.6.33.2 networking regression
From: David Miller @ 2010-04-25 3:19 UTC (permalink / raw)
To: davej; +Cc: netdev, stable
In-Reply-To: <20100425031656.GA27598@redhat.com>
From: Dave Jones <davej@redhat.com>
Date: Sat, 24 Apr 2010 23:16:57 -0400
> There don't appear to be any direct changes to via-velocity.c in the
> diff, so I'm really confused. Any clues ? I'll bisect it, but it
> probably won't be until Monday..
Looks like some x86/PCI/ACPI change causes this, rather than a
networking change.
^ permalink raw reply
* 2.6.33.2 networking regression
From: Dave Jones @ 2010-04-25 3:16 UTC (permalink / raw)
To: netdev; +Cc: stable
Something odd happened when I upgraded my router
from 2.6.33.1 to .33.2. Its internal NIC (a VIA Velocity)
stopped recieving packets.
dmesg was getting flooded with..
[ 188.919957] via-velocity 0000:00:0e.0: BAR 0: set to [io 0xf800-0xf8ff] (PCI address [0xf800-0xf8ff]
[ 188.920002] via-velocity 0000:00:0e.0: BAR 1: set to [mem 0xfdffe000-0xfdffe0ff] (PCI address [0xfdffe000-0xfdffe0ff]
[ 203.913967] via-velocity 0000:00:0e.0: BAR 0: set to [io 0xf800-0xf8ff] (PCI address [0xf800-0xf8ff]
[ 203.914181] via-velocity 0000:00:0e.0: BAR 1: set to [mem 0xfdffe000-0xfdffe0ff] (PCI address [0xfdffe000-0xfdffe0ff]
every so often for some reason.
rebooting back to .1, it works fine.
There don't appear to be any direct changes to via-velocity.c in the
diff, so I'm really confused. Any clues ? I'll bisect it, but it
probably won't be until Monday..
Dave
^ permalink raw reply
* Re: [PATCH 2/2] sky2: add support for receive hashing (v3)
From: David Miller @ 2010-04-25 3:04 UTC (permalink / raw)
To: shemminger; +Cc: jeff, netdev
In-Reply-To: <20100424162239.1aae32e0@nehalam>
From: Stephen Hemminger <shemminger@vyatta.com>
Date: Sat, 24 Apr 2010 16:22:39 -0700
> Subject: sky2: add support for receive hashing
>
> Sky2 hardware supports hardware receive hash calculation.
> Now that Receive Packet Steering is available, add support
> to enable it.
>
> This version does not depend on CONFIG_RPS. Also set_flags rejects
> all values except RXHASH, so driver won't have to change next time
> somebody adds a new one.
>
> Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Applied, thanks Stephen.
^ permalink raw reply
* Re: [PATCH] e100: Fix the TX workqueue race
From: David Miller @ 2010-04-25 3:00 UTC (permalink / raw)
To: alan; +Cc: netdev, e1000-devel
In-Reply-To: <20100424113629.0ad3569b@linux.intel.com>
From: Alan Cox <alan@linux.intel.com>
Date: Sat, 24 Apr 2010 11:36:29 +0100
> Puzzling as it came from a building -next tree. Will see whats
> happened next week if I get time, but I'm afraid net stuff isn't a
> priority - in fact its disappointing that having diagnosed a bug
> months ago (which was the hard bit) and posted a test patch months
> ago the maintainers haven't fixed it.
It's disappointing to me that someone as experienced and skilled
as yourself can't generate a clean patch which is 1) against
the appropriate tree for a bug fix and 2) actually compiles.
Or is this too much to ask? :-)
^ permalink raw reply
* Re: [PATCH] e100: Fix the TX workqueue race
From: David Miller @ 2010-04-25 2:58 UTC (permalink / raw)
To: alan; +Cc: e1000-devel, netdev
In-Reply-To: <20100424121127.084b9766@linux.intel.com>
From: Alan Cox <alan@linux.intel.com>
Date: Sat, 24 Apr 2010 12:11:27 +0100
> No idea why it won't apply - I guess net has diverged from -next in
> this area. Other problem is not typing "stg ref" before "stg export"
It has, the debug print statement above the lines you are changing are
completely different.
Please generate this patch against net-2.6 so I can apply it, thanks
Alan.
------------------------------------------------------------------------------
_______________________________________________
E1000-devel mailing list
E1000-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
^ permalink raw reply
* Re: [PATCH] RCU: don't turn off lockdep when find suspicious rcu_dereference_check() usage
From: Paul E. McKenney @ 2010-04-25 2:36 UTC (permalink / raw)
To: Miles Lane
Cc: Vivek Goyal, Eric Paris, Lai Jiangshan, Ingo Molnar,
Peter Zijlstra, LKML, nauman, eric.dumazet, netdev, Jens Axboe,
Gui Jianfeng, Li Zefan
In-Reply-To: <u2qa44ae5cd1004232235i2e1cd2a0g634fc1d5d8c3f7c2@mail.gmail.com>
On Sat, Apr 24, 2010 at 01:35:01AM -0400, Miles Lane wrote:
> 2.6.34-rc5-git5 with all of your patches applied.
>
> I reconfigured my kernel build options and got the following new issue:
>
> [ 2.686515] [ INFO: suspicious rcu_dereference_check() usage. ]
> [ 2.686519] ---------------------------------------------------
> [ 2.686523] kernel/cgroup.c:4438 invoked rcu_dereference_check()
> without protection!
> [ 2.686526]
> [ 2.686527] other info that might help us debug this:
> [ 2.686529]
> [ 2.686532]
> [ 2.686533] rcu_scheduler_active = 1, debug_locks = 1
> [ 2.686537] 2 locks held by swapper/1:
> [ 2.686540] #0: (mtd_table_mutex){+.+.+.}, at:
> [<ffffffff812d7714>] register_mtd_blktrans+0xa2/0x25e
> [ 2.686555] #1: (&(&blkcg->lock)->rlock){......}, at:
> [<ffffffff811ca7bd>] blkiocg_add_blkio_group+0x29/0x7f
> [ 2.686566]
> [ 2.686567] stack backtrace:
> [ 2.686572] Pid: 1, comm: swapper Not tainted 2.6.34-rc5-git5 #25
> [ 2.686576] Call Trace:
> [ 2.686584] [<ffffffff810642da>] lockdep_rcu_dereference+0x9d/0xa5
> [ 2.686591] [<ffffffff8107af54>] css_id+0x3f/0x52
> [ 2.686597] [<ffffffff811ca7cc>] blkiocg_add_blkio_group+0x38/0x7f
> [ 2.686603] [<ffffffff811cc593>] cfq_init_queue+0xdf/0x2dc
> [ 2.686609] [<ffffffff811bb858>] elevator_init+0xba/0xf5
> [ 2.686616] [<ffffffff812d7046>] ? mtd_blktrans_request+0x0/0x1c
> [ 2.686623] [<ffffffff811c0b62>] blk_init_queue_node+0x12f/0x135
> [ 2.686629] [<ffffffff811c0b74>] blk_init_queue+0xc/0xe
> [ 2.686635] [<ffffffff812d7777>] register_mtd_blktrans+0x105/0x25e
> [ 2.686642] [<ffffffff818c0de9>] ? init_mtdblock+0x0/0x2c
> [ 2.686648] [<ffffffff818c0e13>] init_mtdblock+0x2a/0x2c
> [ 2.686656] [<ffffffff810001ef>] do_one_initcall+0x59/0x14e
> [ 2.686663] [<ffffffff818986a6>] kernel_init+0x160/0x1ea
> [ 2.686669] [<ffffffff81003814>] kernel_thread_helper+0x4/0x10
> [ 2.686677] [<ffffffff8140d77c>] ? restore_args+0x0/0x30
> [ 2.686683] [<ffffffff81898546>] ? kernel_init+0x0/0x1ea
> [ 2.686688] [<ffffffff81003810>] ? kernel_thread_helper+0x0/0x10
> [ 2.687683] mtdoops: mtd device (mtddev=name/number) must be supplied
This should be covered by the patch I sent with my previous email.
And thank you again, Miles, for all the testing!!!
Thanx, Paul
^ permalink raw reply
* Re: [PATCH] RCU: don't turn off lockdep when find suspicious rcu_dereference_check() usage
From: Paul E. McKenney @ 2010-04-25 2:34 UTC (permalink / raw)
To: Miles Lane
Cc: Vivek Goyal, Eric Paris, Lai Jiangshan, Ingo Molnar,
Peter Zijlstra, LKML, nauman, eric.dumazet, netdev, Jens Axboe,
Gui Jianfeng, Li Zefan, Johannes Berg
In-Reply-To: <m2xa44ae5cd1004231559hcf90671asf146a43b4748c2c3@mail.gmail.com>
On Fri, Apr 23, 2010 at 06:59:12PM -0400, Miles Lane wrote:
> On Fri, Apr 23, 2010 at 3:42 PM, Paul E. McKenney
> <paulmck@linux.vnet.ibm.com> wrote:
> > On Fri, Apr 23, 2010 at 08:50:59AM -0400, Miles Lane wrote:
> >> Hi Paul,
> >> There has been a bit of back and forth, and I am not sure what patches
> >> I should test now.
> >> Could you send me a bundle of whatever needs testing now?
> >
> > Hello, Miles,
> >
> > I am posting my set as replies to this message. There are a couple
> > of KVM fixes that are going up via Avi's tree, and a number of networking
> > fixes that are going up via Dave Miller's tree -- a number of these
> > are against quickly changing code, so it didn't make sense for me to
> > keep them separately.
> >
> > I believe that the two splats below are addressed by this patch set
> > carried in the networking tree:
> >
> > https://patchwork.kernel.org/patch/90754/
>
> With your twelve patches and the one linked to above applied to
> 2.6.34-rc5-git3, here are the warnings I see:
>
> [ 0.173969] [ INFO: suspicious rcu_dereference_check() usage. ]
> [ 0.174097] ---------------------------------------------------
> [ 0.174226] include/linux/cgroup.h:534 invoked
> rcu_dereference_check() without protection!
> [ 0.174429]
> [ 0.174430] other info that might help us debug this:
> [ 0.174431]
> [ 0.174792]
> [ 0.174793] rcu_scheduler_active = 1, debug_locks = 1
> [ 0.175037] no locks held by watchdog/0/5.
> [ 0.175162]
> [ 0.175163] stack backtrace:
> [ 0.175405] Pid: 5, comm: watchdog/0 Not tainted 2.6.34-rc5-git3 #22
> [ 0.175534] Call Trace:
> [ 0.175666] [<ffffffff81067fbe>] lockdep_rcu_dereference+0x9d/0xa5
> [ 0.175799] [<ffffffff8102d678>] task_subsys_state+0x59/0x70
> [ 0.175931] [<ffffffff810328fa>] __sched_setscheduler+0x19d/0x300
> [ 0.176064] [<ffffffff8102b477>] ? need_resched+0x1e/0x28
> [ 0.176196] [<ffffffff813cd401>] ? schedule+0x5c3/0x66e
> [ 0.176327] [<ffffffff81091943>] ? watchdog+0x0/0x8c
> [ 0.176457] [<ffffffff81032a78>] sched_setscheduler+0xe/0x10
> [ 0.176587] [<ffffffff8109196d>] watchdog+0x2a/0x8c
> [ 0.176677] [<ffffffff81091943>] ? watchdog+0x0/0x8c
> [ 0.176808] [<ffffffff81057152>] kthread+0x89/0x91
> [ 0.176939] [<ffffffff8106891e>] ? trace_hardirqs_on_caller+0x114/0x13f
> [ 0.177073] [<ffffffff81003994>] kernel_thread_helper+0x4/0x10
> [ 0.177204] [<ffffffff813cfc40>] ? restore_args+0x0/0x30
> [ 0.177334] [<ffffffff810570c9>] ? kthread+0x0/0x91
> [ 0.177463] [<ffffffff81003990>] ? kernel_thread_helper+0x0/0x10
According to Documentation/cgroups/cgroups.txt, we must hold cgroup_mutex,
the task's task_alloc lock, or be in an RCU read-side critical section.
We are in neither of these.
I would argue that sched_setscheduler() should take care of
synchronization, but am not sure which of these three are appropriate
for sched_setscheduler() to acquire. Peter, thoughts?
> [ 3.173419] [ INFO: suspicious rcu_dereference_check() usage. ]
> [ 3.173419] ---------------------------------------------------
> [ 3.173419] kernel/cgroup.c:4438 invoked rcu_dereference_check()
> without protection!
> [ 3.173419]
> [ 3.173419] other info that might help us debug this:
> [ 3.173419]
> [ 3.173419]
> [ 3.173419] rcu_scheduler_active = 1, debug_locks = 1
> [ 3.173419] 2 locks held by async/0/668:
> [ 3.173419] #0: (&shost->scan_mutex){+.+.+.}, at:
> [<ffffffff812df020>] __scsi_add_device+0x83/0xe4
> [ 3.173419] #1: (&(&blkcg->lock)->rlock){......}, at:
> [<ffffffff811f2df9>] blkiocg_add_blkio_group+0x29/0x7f
> [ 3.173419]
> [ 3.173419] stack backtrace:
> [ 3.173419] Pid: 668, comm: async/0 Not tainted 2.6.34-rc5-git3 #22
> [ 3.173419] Call Trace:
> [ 3.173419] [<ffffffff81067fbe>] lockdep_rcu_dereference+0x9d/0xa5
> [ 3.173419] [<ffffffff8107f9ad>] css_id+0x3f/0x51
> [ 3.173419] [<ffffffff811f2e08>] blkiocg_add_blkio_group+0x38/0x7f
> [ 3.173419] [<ffffffff811f4dd0>] cfq_init_queue+0xdf/0x2dc
> [ 3.173419] [<ffffffff811e33b1>] elevator_init+0xba/0xf5
> [ 3.173419] [<ffffffff812dbfaa>] ? scsi_request_fn+0x0/0x451
> [ 3.173419] [<ffffffff811e68d7>] blk_init_queue_node+0x12f/0x135
> [ 3.173419] [<ffffffff811e68e9>] blk_init_queue+0xc/0xe
> [ 3.173419] [<ffffffff812dc41c>] __scsi_alloc_queue+0x21/0x111
> [ 3.173419] [<ffffffff812dc524>] scsi_alloc_queue+0x18/0x64
> [ 3.173419] [<ffffffff812de520>] scsi_alloc_sdev+0x19e/0x256
> [ 3.173419] [<ffffffff812de6be>] scsi_probe_and_add_lun+0xe6/0x9c5
> [ 3.173419] [<ffffffff8106891e>] ? trace_hardirqs_on_caller+0x114/0x13f
> [ 3.173419] [<ffffffff813ce056>] ? __mutex_lock_common+0x3e4/0x43a
> [ 3.173419] [<ffffffff812df020>] ? __scsi_add_device+0x83/0xe4
> [ 3.173419] [<ffffffff812d09dc>] ? transport_setup_classdev+0x0/0x17
> [ 3.173419] [<ffffffff812df020>] ? __scsi_add_device+0x83/0xe4
> [ 3.173419] [<ffffffff812df055>] __scsi_add_device+0xb8/0xe4
> [ 3.173419] [<ffffffff812ea945>] ata_scsi_scan_host+0x74/0x16e
> [ 3.173419] [<ffffffff81057699>] ? autoremove_wake_function+0x0/0x34
> [ 3.173419] [<ffffffff812e8de4>] async_port_probe+0xab/0xb7
> [ 3.173419] [<ffffffff8105e1b1>] ? async_thread+0x0/0x1f4
> [ 3.173419] [<ffffffff8105e2b6>] async_thread+0x105/0x1f4
> [ 3.173419] [<ffffffff81033d8e>] ? default_wake_function+0x0/0xf
> [ 3.173419] [<ffffffff8105e1b1>] ? async_thread+0x0/0x1f4
> [ 3.173419] [<ffffffff81057152>] kthread+0x89/0x91
> [ 3.173419] [<ffffffff8106891e>] ? trace_hardirqs_on_caller+0x114/0x13f
> [ 3.173419] [<ffffffff81003994>] kernel_thread_helper+0x4/0x10
> [ 3.173419] [<ffffffff813cfc40>] ? restore_args+0x0/0x30
> [ 3.173419] [<ffffffff810570c9>] ? kthread+0x0/0x91
> [ 3.173419] [<ffffffff81003990>] ? kernel_thread_helper+0x0/0x10
Please see below for a patch for this based on my earlier conversation
with Vivek Goyal. (Vivek, if you are already pushing a fix elsewhere,
please let me know, and I will drop my patch in favor of yours.)
> [ 32.905446] [ INFO: suspicious rcu_dereference_check() usage. ]
> [ 32.905449] ---------------------------------------------------
> [ 32.905453] net/core/dev.c:1993 invoked rcu_dereference_check()
> without protection!
> [ 32.905456]
> [ 32.905457] other info that might help us debug this:
> [ 32.905458]
> [ 32.905461]
> [ 32.905462] rcu_scheduler_active = 1, debug_locks = 1
> [ 32.905466] 2 locks held by canberra-gtk-pl/4182:
> [ 32.905469] #0: (sk_lock-AF_INET){+.+.+.}, at:
> [<ffffffff81394f7d>] inet_stream_connect+0x3a/0x24d
> [ 32.905483] #1: (rcu_read_lock_bh){.+....}, at:
> [<ffffffff8134a789>] dev_queue_xmit+0x14e/0x4b8
> [ 32.905495]
> [ 32.905496] stack backtrace:
> [ 32.905500] Pid: 4182, comm: canberra-gtk-pl Not tainted 2.6.34-rc5-git3 #22
> [ 32.905504] Call Trace:
> [ 32.905512] [<ffffffff81067fbe>] lockdep_rcu_dereference+0x9d/0xa5
> [ 32.905518] [<ffffffff8134a894>] dev_queue_xmit+0x259/0x4b8
> [ 32.905524] [<ffffffff8134a789>] ? dev_queue_xmit+0x14e/0x4b8
> [ 32.905531] [<ffffffff81041c66>] ? _local_bh_enable_ip+0xcd/0xda
> [ 32.905538] [<ffffffff813536da>] neigh_resolve_output+0x234/0x285
> [ 32.905544] [<ffffffff8136f69f>] ip_finish_output2+0x257/0x28c
> [ 32.905549] [<ffffffff8136f73c>] ip_finish_output+0x68/0x6a
> [ 32.905554] [<ffffffff81370433>] T.866+0x52/0x59
> [ 32.905559] [<ffffffff8137067e>] ip_output+0xaa/0xb4
> [ 32.905565] [<ffffffff8136eb38>] ip_local_out+0x20/0x24
> [ 32.905571] [<ffffffff8136f184>] ip_queue_xmit+0x309/0x368
> [ 32.905578] [<ffffffff810e4226>] ? __kmalloc_track_caller+0x111/0x155
> [ 32.905585] [<ffffffff8138316f>] ? tcp_connect+0x223/0x3d3
> [ 32.905591] [<ffffffff813818f1>] tcp_transmit_skb+0x707/0x745
> [ 32.905597] [<ffffffff813832c2>] tcp_connect+0x376/0x3d3
> [ 32.905604] [<ffffffff81268a43>] ? secure_tcp_sequence_number+0x55/0x6f
> [ 32.905610] [<ffffffff81387270>] tcp_v4_connect+0x3df/0x455
> [ 32.905617] [<ffffffff8133cb59>] ? lock_sock_nested+0xf3/0x102
> [ 32.905623] [<ffffffff81394fe7>] inet_stream_connect+0xa4/0x24d
> [ 32.905629] [<ffffffff8133b398>] sys_connect+0x90/0xd0
> [ 32.905636] [<ffffffff81002b9c>] ? sysret_check+0x27/0x62
> [ 32.905642] [<ffffffff8106891e>] ? trace_hardirqs_on_caller+0x114/0x13f
> [ 32.905649] [<ffffffff813cec80>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> [ 32.905655] [<ffffffff81002b6b>] system_call_fastpath+0x16/0x1b
A fix for the above is already in Dave Miller's tree.
> [ 51.912282] [ INFO: suspicious rcu_dereference_check() usage. ]
> [ 51.912285] ---------------------------------------------------
> [ 51.912289] net/mac80211/sta_info.c:886 invoked
> rcu_dereference_check() without protection!
> [ 51.912293]
> [ 51.912293] other info that might help us debug this:
> [ 51.912295]
> [ 51.912298]
> [ 51.912298] rcu_scheduler_active = 1, debug_locks = 1
> [ 51.912302] no locks held by wpa_supplicant/3951.
> [ 51.912305]
> [ 51.912306] stack backtrace:
> [ 51.912310] Pid: 3951, comm: wpa_supplicant Not tainted 2.6.34-rc5-git3 #22
> [ 51.912314] Call Trace:
> [ 51.912317] <IRQ> [<ffffffff81067fbe>] lockdep_rcu_dereference+0x9d/0xa5
> [ 51.912345] [<ffffffffa014f9ae>]
> ieee80211_find_sta_by_hw+0x46/0x10f [mac80211]
> [ 51.912358] [<ffffffffa014fa8e>] ieee80211_find_sta+0x17/0x19 [mac80211]
> [ 51.912373] [<ffffffffa01e50f2>] iwl_tx_queue_reclaim+0xdb/0x1b1 [iwlcore]
> [ 51.912380] [<ffffffff8106842b>] ? mark_lock+0x2d/0x235
> [ 51.912391] [<ffffffffa0252f1c>] iwl5000_rx_reply_tx+0x4a9/0x556 [iwlagn]
> [ 51.912399] [<ffffffff8120a353>] ? is_swiotlb_buffer+0x2e/0x3b
> [ 51.912407] [<ffffffffa024bbf4>] iwl_rx_handle+0x163/0x2b5 [iwlagn]
> [ 51.912414] [<ffffffff81068904>] ? trace_hardirqs_on_caller+0xfa/0x13f
> [ 51.912422] [<ffffffffa024c3ac>] iwl_irq_tasklet+0x2bb/0x3c0 [iwlagn]
> [ 51.912429] [<ffffffff810411f3>] tasklet_action+0xa7/0x10f
> [ 51.912435] [<ffffffff81042205>] __do_softirq+0x144/0x252
> [ 51.912442] [<ffffffff81003a8c>] call_softirq+0x1c/0x34
> [ 51.912447] [<ffffffff810050e4>] do_softirq+0x38/0x80
> [ 51.912452] [<ffffffff81041cd2>] irq_exit+0x45/0x94
> [ 51.912457] [<ffffffff81004829>] do_IRQ+0xad/0xc4
> [ 51.912463] [<ffffffff810cbbd3>] ? might_fault+0x63/0xb3
> [ 51.912470] [<ffffffff813cfb93>] ret_from_intr+0x0/0xf
> [ 51.912474] <EOI> [<ffffffff810cbbd3>] ? might_fault+0x63/0xb3
> [ 51.912484] [<ffffffff8106a75d>] ? lock_release+0x208/0x215
> [ 51.912490] [<ffffffff810cbc1c>] might_fault+0xac/0xb3
> [ 51.912495] [<ffffffff810cbbd3>] ? might_fault+0x63/0xb3
> [ 51.912501] [<ffffffff812025e3>] __clear_user+0x15/0x59
> [ 51.912508] [<ffffffff8100b2bc>] save_i387_xstate+0x9c/0x1bc
> [ 51.912515] [<ffffffff81002276>] do_signal+0x240/0x686
> [ 51.912521] [<ffffffff81002b9c>] ? sysret_check+0x27/0x62
> [ 51.912527] [<ffffffff8106891e>] ? trace_hardirqs_on_caller+0x114/0x13f
> [ 51.912533] [<ffffffff813cec80>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> [ 51.912539] [<ffffffff810026e3>] do_notify_resume+0x27/0x5f
> [ 51.912545] [<ffffffff813cec80>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> [ 51.912551] [<ffffffff81002e86>] int_signal+0x12/0x17
This is a repeat from last time that confused me at the time. I could
do a hacky "fix" by putting an RCU read-side critical section around
the for_each_sta_info() in ieee80211_find_sta_by_hw(), but I do not
understand this code well enough to feel comfortable doing so.
Johannes, any enlightenment?
> [ 51.929529] [ INFO: suspicious rcu_dereference_check() usage. ]
> [ 51.929532] ---------------------------------------------------
> [ 51.929536] net/mac80211/sta_info.c:886 invoked
> rcu_dereference_check() without protection!
> [ 51.929540]
> [ 51.929541] other info that might help us debug this:
> [ 51.929542]
> [ 51.929545]
> [ 51.929546] rcu_scheduler_active = 1, debug_locks = 1
> [ 51.929550] 1 lock held by Xorg/4013:
> [ 51.929553] #0: (clock-AF_UNIX){++.+..}, at: [<ffffffff8133cebd>]
> sock_def_readable+0x19/0x62
> [ 51.929567]
> [ 51.929568] stack backtrace:
> [ 51.929573] Pid: 4013, comm: Xorg Not tainted 2.6.34-rc5-git3 #22
> [ 51.929576] Call Trace:
> [ 51.929579] <IRQ> [<ffffffff81067fbe>] lockdep_rcu_dereference+0x9d/0xa5
> [ 51.929603] [<ffffffffa014f9fe>]
> ieee80211_find_sta_by_hw+0x96/0x10f [mac80211]
> [ 51.929615] [<ffffffffa014fa8e>] ieee80211_find_sta+0x17/0x19 [mac80211]
> [ 51.929631] [<ffffffffa01e50f2>] iwl_tx_queue_reclaim+0xdb/0x1b1 [iwlcore]
> [ 51.929642] [<ffffffffa0252f1c>] iwl5000_rx_reply_tx+0x4a9/0x556 [iwlagn]
> [ 51.929649] [<ffffffff81068685>] ? mark_held_locks+0x52/0x70
> [ 51.929656] [<ffffffff813cf46c>] ? _raw_spin_unlock_irqrestore+0x3a/0x69
> [ 51.929662] [<ffffffff8120a353>] ? is_swiotlb_buffer+0x2e/0x3b
> [ 51.929671] [<ffffffffa024bbf4>] iwl_rx_handle+0x163/0x2b5 [iwlagn]
> [ 51.929680] [<ffffffffa024c3ac>] iwl_irq_tasklet+0x2bb/0x3c0 [iwlagn]
> [ 51.929687] [<ffffffff810411f3>] tasklet_action+0xa7/0x10f
> [ 51.929693] [<ffffffff81042205>] __do_softirq+0x144/0x252
> [ 51.929700] [<ffffffff81003a8c>] call_softirq+0x1c/0x34
> [ 51.929705] [<ffffffff810050e4>] do_softirq+0x38/0x80
> [ 51.929711] [<ffffffff81041cd2>] irq_exit+0x45/0x94
> [ 51.929717] [<ffffffff81019b10>] smp_apic_timer_interrupt+0x87/0x95
> [ 51.929724] [<ffffffff81003553>] apic_timer_interrupt+0x13/0x20
> [ 51.929727] <EOI> [<ffffffff813cf46e>] ?
> _raw_spin_unlock_irqrestore+0x3c/0x69
> [ 51.929739] [<ffffffff8102d3fb>] __wake_up_sync_key+0x49/0x52
> [ 51.929745] [<ffffffff8133cee7>] sock_def_readable+0x43/0x62
> [ 51.929751] [<ffffffff813b1c61>] unix_stream_sendmsg+0x243/0x2e2
> [ 51.929758] [<ffffffff8133b912>] ? sock_aio_write+0x0/0xcf
> [ 51.929764] [<ffffffff81339342>] __sock_sendmsg+0x59/0x64
> [ 51.929770] [<ffffffff8133b9cd>] sock_aio_write+0xbb/0xcf
> [ 51.929777] [<ffffffff810e9909>] do_sync_readv_writev+0xbc/0xfb
> [ 51.929785] [<ffffffff811c1792>] ? selinux_file_permission+0xa2/0xaf
> [ 51.929790] [<ffffffff810e9690>] ? copy_from_user+0x2a/0x2c
> [ 51.929797] [<ffffffff811baff1>] ? security_file_permission+0x11/0x13
> [ 51.929804] [<ffffffff810ea6a6>] do_readv_writev+0xa2/0x122
> [ 51.929810] [<ffffffff810ead93>] ? fcheck_files+0x8f/0xc9
> [ 51.929816] [<ffffffff810ea764>] vfs_writev+0x3e/0x49
> [ 51.929821] [<ffffffff810ea84a>] sys_writev+0x45/0x8e
> [ 51.929828] [<ffffffff81002b6b>] system_call_fastpath+0x16/0x1b
Ditto.
Thanx, Paul
------------------------------------------------------------------------
commit 0868dd631def762ba00c2f0f397a53c5cdf24ae2
Author: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Date: Sat Apr 24 19:23:30 2010 -0700
block-cgroup: fix RCU-lockdep splat in blkiocg_add_blkio_group()
It is necessary to be in an RCU read-side critical section when invoking
css_id(), so this patch adds one to blkiocg_add_blkio_group(). This is
actually a false positive, because this is called at initialization time,
and hence always refers to the root cgroup, which cannot go away.
Located-by: Miles Lane <miles.lane@gmail.com>
Suggested-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index 5fe03de..55c8c73 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -71,7 +71,9 @@ void blkiocg_add_blkio_group(struct blkio_cgroup *blkcg,
spin_lock_irqsave(&blkcg->lock, flags);
rcu_assign_pointer(blkg->key, key);
+ rcu_read_lock();
blkg->blkcg_id = css_id(&blkcg->css);
+ rcu_read_unlock();
hlist_add_head_rcu(&blkg->blkcg_node, &blkcg->blkg_list);
spin_unlock_irqrestore(&blkcg->lock, flags);
#ifdef CONFIG_DEBUG_BLK_CGROUP
^ permalink raw reply related
* Re: rps perfomance WAS(Re: rps: question
From: Changli Gao @ 2010-04-25 2:31 UTC (permalink / raw)
To: hadi; +Cc: Eric Dumazet, Rick Jones, David Miller, therbert, netdev, robert,
andi
In-Reply-To: <1271938343.4032.30.camel@bigi>
On Thu, Apr 22, 2010 at 8:12 PM, jamal <hadi@cyberus.ca> wrote:
>
>> I see slave/application cpus hit _raw_spin_lock_irqsave() and
>> _raw_spin_unlock_irqrestore().
>>
>> Maybe a ring buffer could help (instead of a double linked queue) for
>> backlog, or the double queue trick, if Changli wants to respin his
>> patch.
>>
>
> Ok, I will have some cycles later today/tommorow or for sure on weekend.
> My setup is still intact - so i can test.
>
I read the code again, and find that we don't use spin_lock_irqsave(),
and we use local_irq_save() and spin_lock() instead, so
_raw_spin_lock_irqsave() and _raw_spin_lock_irqrestore() should not be
related to backlog. the lock maybe sk_receive_queue.lock.
Jamal, did you use a single socket to serve all the clients?
BTW: completion_queue and output_queue in softnet_data both are LIFO
queues. For completion_queue, FIFO is better, as the last used skb is
more likely in cache, and should be used first. Since slab has always
cache the last used memory at the head, we'd better free the skb in
FIFO manner. For output_queue, FIFO is good for fairness among qdiscs.
--
Regards，
Changli Gao(xiaosuo@gmail.com)
^ permalink raw reply
* [PATCH 2/2] sky2: add support for receive hashing (v3)
From: Stephen Hemminger @ 2010-04-24 23:22 UTC (permalink / raw)
To: Jeff Garzik; +Cc: davem, netdev
In-Reply-To: <4BD0F8AA.2060700@garzik.org>
Subject: sky2: add support for receive hashing
Sky2 hardware supports hardware receive hash calculation.
Now that Receive Packet Steering is available, add support
to enable it.
This version does not depend on CONFIG_RPS. Also set_flags rejects
all values except RXHASH, so driver won't have to change next time
somebody adds a new one.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
---
drivers/net/sky2.c | 75 +++++++++++++++++++++++++++++++++++++++++++++++++++--
drivers/net/sky2.h | 23 ++++++++++++++++
2 files changed, 96 insertions(+), 2 deletions(-)
--- a/drivers/net/sky2.c 2010-04-23 08:56:41.444717559 -0700
+++ b/drivers/net/sky2.c 2010-04-23 08:56:51.335959014 -0700
@@ -1193,6 +1193,39 @@ static void rx_set_checksum(struct sky2_
? BMU_ENA_RX_CHKSUM : BMU_DIS_RX_CHKSUM);
}
+/* Enable/disable receive hash calculation (RSS) */
+static void rx_set_rss(struct net_device *dev)
+{
+ struct sky2_port *sky2 = netdev_priv(dev);
+ struct sky2_hw *hw = sky2->hw;
+ int i, nkeys = 4;
+
+ /* Supports IPv6 and other modes */
+ if (hw->flags & SKY2_HW_NEW_LE) {
+ nkeys = 10;
+ sky2_write32(hw, SK_REG(sky2->port, RSS_CFG), HASH_ALL);
+ }
+
+ /* Program RSS initial values */
+ if (dev->features & NETIF_F_RXHASH) {
+ u32 key[nkeys];
+
+ get_random_bytes(key, nkeys * sizeof(u32));
+ for (i = 0; i < nkeys; i++)
+ sky2_write32(hw, SK_REG(sky2->port, RSS_KEY + i * 4),
+ key[i]);
+
+ /* Need to turn on (undocumented) flag to make hashing work */
+ sky2_write32(hw, SK_REG(sky2->port, RX_GMF_CTRL_T),
+ RX_STFW_ENA);
+
+ sky2_write32(hw, Q_ADDR(rxqaddr[sky2->port], Q_CSR),
+ BMU_ENA_RX_RSS_HASH);
+ } else
+ sky2_write32(hw, Q_ADDR(rxqaddr[sky2->port], Q_CSR),
+ BMU_DIS_RX_RSS_HASH);
+}
+
/*
* The RX Stop command will not work for Yukon-2 if the BMU does not
* reach the end of packet and since we can't make sure that we have
@@ -1425,6 +1458,9 @@ static void sky2_rx_start(struct sky2_po
if (!(hw->flags & SKY2_HW_NEW_LE))
rx_set_checksum(sky2);
+ if (!(hw->flags & SKY2_HW_RSS_BROKEN))
+ rx_set_rss(sky2->netdev);
+
/* submit Rx ring */
for (i = 0; i < sky2->rx_pending; i++) {
re = sky2->rx_ring + i;
@@ -2534,6 +2570,14 @@ static void sky2_rx_checksum(struct sky2
}
}
+static void sky2_rx_hash(struct sky2_port *sky2, u32 status)
+{
+ struct sk_buff *skb;
+
+ skb = sky2->rx_ring[sky2->rx_next].skb;
+ skb->rxhash = le32_to_cpu(status);
+}
+
/* Process status response ring */
static int sky2_status_intr(struct sky2_hw *hw, int to_do, u16 idx)
{
@@ -2606,6 +2650,10 @@ static int sky2_status_intr(struct sky2_
sky2_rx_checksum(sky2, status);
break;
+ case OP_RSS_HASH:
+ sky2_rx_hash(sky2, status);
+ break;
+
case OP_TXINDEXLE:
/* TX index reports status for both ports */
sky2_tx_done(hw->dev[0], status & 0xfff);
@@ -2960,6 +3008,8 @@ static int __devinit sky2_init(struct sk
switch(hw->chip_id) {
case CHIP_ID_YUKON_XL:
hw->flags = SKY2_HW_GIGABIT | SKY2_HW_NEWER_PHY;
+ if (hw->chip_rev < CHIP_REV_YU_XL_A2)
+ hw->flags |= SKY2_HW_RSS_BROKEN;
break;
case CHIP_ID_YUKON_EC_U:
@@ -2985,10 +3035,11 @@ static int __devinit sky2_init(struct sk
dev_err(&hw->pdev->dev, "unsupported revision Yukon-EC rev A1\n");
return -EOPNOTSUPP;
}
- hw->flags = SKY2_HW_GIGABIT;
+ hw->flags = SKY2_HW_GIGABIT | SKY2_HW_RSS_BROKEN;
break;
case CHIP_ID_YUKON_FE:
+ hw->flags = SKY2_HW_RSS_BROKEN;
break;
case CHIP_ID_YUKON_FE_P:
@@ -4112,6 +4163,25 @@ static int sky2_set_eeprom(struct net_de
return sky2_vpd_write(sky2->hw, cap, data, eeprom->offset, eeprom->len);
}
+static int sky2_set_flags(struct net_device *dev, u32 data)
+{
+ struct sky2_port *sky2 = netdev_priv(dev);
+
+ if (data & ~ETH_FLAG_RXHASH)
+ return -EOPNOTSUPP;
+
+ if (data & ETH_FLAG_RXHASH) {
+ if (sky2->hw->flags & SKY2_HW_RSS_BROKEN)
+ return -EINVAL;
+
+ dev->features |= NETIF_F_RXHASH;
+ } else
+ dev->features &= ~NETIF_F_RXHASH;
+
+ rx_set_rss(dev);
+
+ return 0;
+}
static const struct ethtool_ops sky2_ethtool_ops = {
.get_settings = sky2_get_settings,
@@ -4143,6 +4213,7 @@ static const struct ethtool_ops sky2_eth
.phys_id = sky2_phys_id,
.get_sset_count = sky2_get_sset_count,
.get_ethtool_stats = sky2_get_ethtool_stats,
+ .set_flags = sky2_set_flags,
};
#ifdef CONFIG_SKY2_DEBUG
@@ -4496,6 +4567,10 @@ static __devinit struct net_device *sky2
if (highmem)
dev->features |= NETIF_F_HIGHDMA;
+ /* Enable receive hashing unless hardware is known broken */
+ if (!(hw->flags & SKY2_HW_RSS_BROKEN))
+ dev->features |= NETIF_F_RXHASH;
+
#ifdef SKY2_VLAN_TAG_USED
/* The workaround for FE+ status conflicts with VLAN tag detection. */
if (!(sky2->hw->chip_id == CHIP_ID_YUKON_FE_P &&
@@ -4692,7 +4767,7 @@ static int __devinit sky2_probe(struct p
goto err_out_iounmap;
/* ring for status responses */
- hw->st_size = hw->ports * roundup_pow_of_two(2*RX_MAX_PENDING + TX_MAX_PENDING);
+ hw->st_size = hw->ports * roundup_pow_of_two(3*RX_MAX_PENDING + TX_MAX_PENDING);
hw->st_le = pci_alloc_consistent(pdev, hw->st_size * sizeof(struct sky2_status_le),
&hw->st_dma);
if (!hw->st_le)
--- a/drivers/net/sky2.h 2010-04-23 08:56:41.424736307 -0700
+++ b/drivers/net/sky2.h 2010-04-23 08:56:51.335959014 -0700
@@ -694,8 +694,21 @@ enum {
TXA_CTRL = 0x0210,/* 8 bit Tx Arbiter Control Register */
TXA_TEST = 0x0211,/* 8 bit Tx Arbiter Test Register */
TXA_STAT = 0x0212,/* 8 bit Tx Arbiter Status Register */
+
+ RSS_KEY = 0x0220, /* RSS Key setup */
+ RSS_CFG = 0x0248, /* RSS Configuration */
};
+enum {
+ HASH_TCP_IPV6_EX_CTRL = 1<<5,
+ HASH_IPV6_EX_CTRL = 1<<4,
+ HASH_TCP_IPV6_CTRL = 1<<3,
+ HASH_IPV6_CTRL = 1<<2,
+ HASH_TCP_IPV4_CTRL = 1<<1,
+ HASH_IPV4_CTRL = 1<<0,
+
+ HASH_ALL = 0x3f,
+};
enum {
B6_EXT_REG = 0x0300,/* External registers (GENESIS only) */
@@ -2261,6 +2274,7 @@ struct sky2_hw {
#define SKY2_HW_NEW_LE 0x00000020 /* new LSOv2 format */
#define SKY2_HW_AUTO_TX_SUM 0x00000040 /* new IP decode for Tx */
#define SKY2_HW_ADV_POWER_CTL 0x00000080 /* additional PHY power regs */
+#define SKY2_HW_RSS_BROKEN 0x00000100
u8 chip_id;
u8 chip_rev;
^ permalink raw reply
* [PATCH kernel 2.6.34-rc5] smc91c92_cs: spin_unlock_irqrestore before calling smc_interrupt()
From: Ken Kawasaki @ 2010-04-24 20:37 UTC (permalink / raw)
To: netdev
In-Reply-To: <20100411075014.d3e80847.ken_kawasaki@spring.nifty.jp>
smc91c92_cs:
* spin_unlock_irqrestore before calling smc_interrupt() in media_check()
to avoid lockup.
* use spin_lock_irqsave for ethtool function.
Signed-off-by: Ken Kawasaki <ken_kawasaki@spring.nifty.jp>
---
--- linux-2.6.34-rc5/drivers/net/pcmcia/smc91c92_cs.c.orig 2010-04-24 07:33:50.000000000 +0900
+++ linux-2.6.34-rc5/drivers/net/pcmcia/smc91c92_cs.c 2010-04-24 07:42:44.000000000 +0900
@@ -1804,23 +1804,30 @@ static void media_check(u_long arg)
SMC_SELECT_BANK(1);
media |= (inw(ioaddr + CONFIG) & CFG_AUI_SELECT) ? 2 : 1;
+ SMC_SELECT_BANK(saved_bank);
+ spin_unlock_irqrestore(&smc->lock, flags);
+
/* Check for pending interrupt with watchdog flag set: with
this, we can limp along even if the interrupt is blocked */
if (smc->watchdog++ && ((i>>8) & i)) {
if (!smc->fast_poll)
printk(KERN_INFO "%s: interrupt(s) dropped!\n", dev->name);
+ local_irq_save(flags);
smc_interrupt(dev->irq, dev);
+ local_irq_restore(flags);
smc->fast_poll = HZ;
}
if (smc->fast_poll) {
smc->fast_poll--;
smc->media.expires = jiffies + HZ/100;
add_timer(&smc->media);
- SMC_SELECT_BANK(saved_bank);
- spin_unlock_irqrestore(&smc->lock, flags);
return;
}
+ spin_lock_irqsave(&smc->lock, flags);
+
+ saved_bank = inw(ioaddr + BANK_SELECT);
+
if (smc->cfg & CFG_MII_SELECT) {
if (smc->mii_if.phy_id < 0)
goto reschedule;
@@ -1978,15 +1985,16 @@ static int smc_get_settings(struct net_d
unsigned int ioaddr = dev->base_addr;
u16 saved_bank = inw(ioaddr + BANK_SELECT);
int ret;
+ unsigned long flags;
- spin_lock_irq(&smc->lock);
+ spin_lock_irqsave(&smc->lock, flags);
SMC_SELECT_BANK(3);
if (smc->cfg & CFG_MII_SELECT)
ret = mii_ethtool_gset(&smc->mii_if, ecmd);
else
ret = smc_netdev_get_ecmd(dev, ecmd);
SMC_SELECT_BANK(saved_bank);
- spin_unlock_irq(&smc->lock);
+ spin_unlock_irqrestore(&smc->lock, flags);
return ret;
}
@@ -1996,15 +2004,16 @@ static int smc_set_settings(struct net_d
unsigned int ioaddr = dev->base_addr;
u16 saved_bank = inw(ioaddr + BANK_SELECT);
int ret;
+ unsigned long flags;
- spin_lock_irq(&smc->lock);
+ spin_lock_irqsave(&smc->lock, flags);
SMC_SELECT_BANK(3);
if (smc->cfg & CFG_MII_SELECT)
ret = mii_ethtool_sset(&smc->mii_if, ecmd);
else
ret = smc_netdev_set_ecmd(dev, ecmd);
SMC_SELECT_BANK(saved_bank);
- spin_unlock_irq(&smc->lock);
+ spin_unlock_irqrestore(&smc->lock, flags);
return ret;
}
@@ -2014,12 +2023,13 @@ static u32 smc_get_link(struct net_devic
unsigned int ioaddr = dev->base_addr;
u16 saved_bank = inw(ioaddr + BANK_SELECT);
u32 ret;
+ unsigned long flags;
- spin_lock_irq(&smc->lock);
+ spin_lock_irqsave(&smc->lock, flags);
SMC_SELECT_BANK(3);
ret = smc_link_ok(dev);
SMC_SELECT_BANK(saved_bank);
- spin_unlock_irq(&smc->lock);
+ spin_unlock_irqrestore(&smc->lock, flags);
return ret;
}
@@ -2056,16 +2066,17 @@ static int smc_ioctl (struct net_device
int rc = 0;
u16 saved_bank;
unsigned int ioaddr = dev->base_addr;
+ unsigned long flags;
if (!netif_running(dev))
return -EINVAL;
- spin_lock_irq(&smc->lock);
+ spin_lock_irqsave(&smc->lock, flags);
saved_bank = inw(ioaddr + BANK_SELECT);
SMC_SELECT_BANK(3);
rc = generic_mii_ioctl(&smc->mii_if, mii, cmd, NULL);
SMC_SELECT_BANK(saved_bank);
- spin_unlock_irq(&smc->lock);
+ spin_unlock_irqrestore(&smc->lock, flags);
return rc;
}
^ permalink raw reply
* Re: DDoS attack causing bad effect on conntrack searches
From: Eric Dumazet @ 2010-04-24 20:11 UTC (permalink / raw)
To: Jesper Dangaard Brouer
Cc: paulmck, Patrick McHardy, Changli Gao, hawk,
Linux Kernel Network Hackers, Netfilter Developers
In-Reply-To: <Pine.LNX.4.64.1004241219280.32071@ask.diku.dk>
Le samedi 24 avril 2010 à 13:11 +0200, Jesper Dangaard Brouer a écrit :
> On Fri, 23 Apr 2010, Eric Dumazet wrote:
>
> > Le jeudi 22 avril 2010 à 22:38 +0200, Jesper Dangaard Brouer a écrit :
> >
> >>
> >> I think its plausable, there is a lot of modification going on.
> >> Approx 40.000 deletes/sec and 40.000 inserts/sec.
> >> The hash bucket size is 300032, and with 80000 modifications/sec, we are
> >> (potentially) changing 26.6% of the hash chains each second.
> >>
> >> As can be seen from the graphs:
> >> http://people.netfilter.org/hawk/DDoS/2010-04-12__001/list.html
> >>
> >> Notice that primarily CPU2 is doing the 40k deletes/sec, while CPU1 is
> >> caught searching...
> >>
> >>
> >>> maybe hash table has one slot :)
> >>
> >> Guess I have to reproduce the DoS attack in a testlab (I will first have
> >> time Tuesday). So we can determine if its bad hashing or restart of the
> >> search loop.
> >>
> >>
> >> The traffic pattern was fairly simple:
> >>
> >> 200 bytes UDP packets, comming from approx 60 source IPs, going to one
> >> destination IP. The UDP destination port number was varied in the range
> >> of 1 to 6000. The source UDP port was varied a bit more, some ranging
> >> from 32768 to 61000, and some from 1028 to 5000.
> >>
> >>
> >
> > Re-reading this, I am not sure there is a real problem on RCU as you
> > pointed out.
> >
> > With 800.000 entries, in a 300.032 buckets hash table, each lookup hit
> > about 3 entries (aka searches in conntrack stats)
> >
> > 300.000 packets/second -> 900.000 'searches' per second.
>
> The machine is not getting 300.000 pps, its only getting 40.000 pps (the
> rest is stopped by the NIC by sending Ethernet flowcontrol pause frames)
>
> http://people.netfilter.org/hawk/DDoS/2010-04-12__001/eth0-rx.png
>
> We are doing 700.000 'searches' per second, with 40.000 pps, thus on
> average the list lenght (in each hash bucket) just need to be 17.5
> elements. Is this an acceptable has distribution, with 900.000 elements
> in a 300.032 buckets hash table?
>
>
> > If you have four cpus all trying to insert/delete entries in //, they
> > all hit the central conntrack lock.
>
> This machine only have two CPUs, or rather one physical CPU and one
> hyperthreaded. The server is a old HP380 G4, with an old Xeon type CPU
> 3.4 GHz (1MB cache). If remember correctly its based on Pentium-4
> technologi, which had a pretty bad hyperthreading.
>
>
> > On a DDOS scenario, every packet needs to take this lock twice,
> > once to free an old conntrack (early drop), once to insert a new entry.
>
> I was worried about if the "early drop" e.g. free an old conntrack would
> disturbe the conntrack searching?
>
>
> > To scale this, only way would be to have an array of locks, like we have
> > for TCP/UDP hash tables.
> >
> > I did some tests here, with a multiqueue card, flooded with 300.000
> > pack/second, 65.536 source IP, millions of flows, and nothing wrong
> > happened (but packets drops, of course)
>
> A small hint when testing, use Haralds tool 'lnstat' to see the stats on
> the command line, thus you don't need to RRDtool graphe every thing:
>
> Command:
> lnstat -f nf_conntrack -i 1 -c 1000
>
> I don't have a multiqueue NIC in this old machine.
>
> I also ran some tests on my 10G testlab, but it didn't go wrong.
> Tweeking the pktgen DDoS I could get the system to do 4.500.000 'searches'
> per sec with a 1.500.000 packets/sec. (Have not reloaded the kernel with
> the new failed lookup stats). Guess my 10G machines are too fast to hit
> the issues.
>
>
> > My two cpus were busy 100%, after tweaking smp_affinities, because on
> > first try, irqbalance put "01" mask on both queues, so only one ksoftirq
> > was working, other cpu was idle :(
>
> Think my machine is some what slower than yours, perhaps its simply not
> fast enough for this kind of workload (pretty sure that the cache is the
> CPU is getting f*ked in this case).
>
> On my machine one CPU in stuck in softirq:
> http://people.netfilter.org/hawk/DDoS/2010-04-12__001/cpu_softirq001.png
>
> And another observation is that the CPUs are disturbing each other on the
> RX softirq code path.
> http://people.netfilter.org/hawk/DDoS/2010-04-12__001/softnet_time_squeeze_rx-softirq001.png
> (/proc/net/softnet_stat column 3)
>
> Monday or Tuesdag I'll do a test setup with some old HP380 G4 machines to
> see if I can reproduce the DDoS attack senario. And see if I can get
> it into to lookup loop.
Theorically a loop is very unlikely, given a single retry is very
unlikly too.
Unless a cpu gets in its cache a corrupted value of a 'next' pointer.
Maybe a hardware problem ?
My test machine is a fairly low end one, an AMD Athlon Dual core 5050e,
2.6 GHz
I used an igb card for ease of setup, and to make sure my two cores
would handle packets in parallel, without RPS.
With same hash bucket size (300.032) and max conntracks (800.000), and
after more than 10 hours of test, not a single lookup was restarted
because of a nulls with wrong value.
I can setup a test on a 16 cpu machine, multiqueue card too.
Hmm, I forgot to say I am using net-next-2.6, not your kernel version...
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* Re: [RFC][PATCH v3 2/3] Provides multiple submits and asynchronous notifications.
From: Michael S. Tsirkin @ 2010-04-24 19:32 UTC (permalink / raw)
To: xiaohui.xin; +Cc: arnd, netdev, kvm, linux-kernel, mingo, davem, jdike
In-Reply-To: <1272006513-5188-1-git-send-email-xiaohui.xin@intel.com>
On Fri, Apr 23, 2010 at 03:08:33PM +0800, xiaohui.xin@intel.com wrote:
> From: Xin Xiaohui <xiaohui.xin@intel.com>
>
> The vhost-net backend now only supports synchronous send/recv
> operations. The patch provides multiple submits and asynchronous
> notifications. This is needed for zero-copy case.
>
> Signed-off-by: Xin Xiaohui <xiaohui.xin@intel.com>
> ---
>
> Michael,
> >>>Can't vhost supply a kiocb completion callback that will handle the list?
> >>Yes, thanks. And with it I also remove the vq->receivr finally.
> >>Thanks
> >>Xiaohui
>
> >Nice progress. I commented on some minor issues below.
> >Thanks!
>
> The updated patch addressed your comments on the minor issues.
> Thanks!
>
> Thanks
> Xiaohui
>
> drivers/vhost/net.c | 236 +++++++++++++++++++++++++++++++++++++++++++++++-
> drivers/vhost/vhost.c | 120 ++++++++++++++-----------
> drivers/vhost/vhost.h | 14 +++
> 3 files changed, 314 insertions(+), 56 deletions(-)
>
> diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
> index 38989d1..18f6c41 100644
> --- a/drivers/vhost/net.c
> +++ b/drivers/vhost/net.c
> @@ -23,6 +23,8 @@
> #include <linux/if_arp.h>
> #include <linux/if_tun.h>
> #include <linux/if_macvlan.h>
> +#include <linux/mpassthru.h>
> +#include <linux/aio.h>
>
> #include <net/sock.h>
>
> @@ -48,6 +50,7 @@ struct vhost_net {
> struct vhost_dev dev;
> struct vhost_virtqueue vqs[VHOST_NET_VQ_MAX];
> struct vhost_poll poll[VHOST_NET_VQ_MAX];
> + struct kmem_cache *cache;
> /* Tells us whether we are polling a socket for TX.
> * We only do this when socket buffer fills up.
> * Protected by tx vq lock. */
> @@ -92,11 +95,138 @@ static void tx_poll_start(struct vhost_net *net, struct socket *sock)
> net->tx_poll_state = VHOST_NET_POLL_STARTED;
> }
>
> +struct kiocb *notify_dequeue(struct vhost_virtqueue *vq)
> +{
> + struct kiocb *iocb = NULL;
> + unsigned long flags;
> +
> + spin_lock_irqsave(&vq->notify_lock, flags);
> + if (!list_empty(&vq->notifier)) {
> + iocb = list_first_entry(&vq->notifier,
> + struct kiocb, ki_list);
> + list_del(&iocb->ki_list);
> + }
> + spin_unlock_irqrestore(&vq->notify_lock, flags);
> + return iocb;
> +}
> +
> +static void handle_iocb(struct kiocb *iocb)
> +{
> + struct vhost_virtqueue *vq = iocb->private;
> + unsigned long flags;
> +
> + spin_lock_irqsave(&vq->notify_lock, flags);
> + list_add_tail(&iocb->ki_list, &vq->notifier);
> + spin_unlock_irqrestore(&vq->notify_lock, flags);
Don't we need to wake up the wq as well?
> +}
> +
> +static int is_async_vq(struct vhost_virtqueue *vq)
> +{
> + return (vq->link_state == VHOST_VQ_LINK_ASYNC);
() not needed
> +}
> +
> +static void handle_async_rx_events_notify(struct vhost_net *net,
> + struct vhost_virtqueue *vq,
> + struct socket *sock)
> +{
> + struct kiocb *iocb = NULL;
> + struct vhost_log *vq_log = NULL;
> + int rx_total_len = 0;
> + unsigned int head, log, in, out;
> + int size;
> +
> + if (!is_async_vq(vq))
> + return;
> +
> + if (sock->sk->sk_data_ready)
> + sock->sk->sk_data_ready(sock->sk, 0);
> +
> + vq_log = unlikely(vhost_has_feature(&net->dev, VHOST_F_LOG_ALL)) ?
> + vq->log : NULL;
> +
> + while ((iocb = notify_dequeue(vq)) != NULL) {
> + vhost_add_used_and_signal(&net->dev, vq,
> + iocb->ki_pos, iocb->ki_nbytes);
> + size = iocb->ki_nbytes;
> + head = iocb->ki_pos;
> + rx_total_len += iocb->ki_nbytes;
> +
> + if (iocb->ki_dtor)
> + iocb->ki_dtor(iocb);
I am confused by the above. Isn't ki_dtor handle_iocb?
Why is it called here?
> + kmem_cache_free(net->cache, iocb);
> +
> + /* when log is enabled, recomputing the log info is needed,
> + * since these buffers are in async queue, and may not get
> + * the log info before.
> + */
> + if (unlikely(vq_log)) {
> + if (!log)
log is uninitialized now?
> + __vhost_get_vq_desc(&net->dev, vq, vq->iov,
> + ARRAY_SIZE(vq->iov),
> + &out, &in, vq_log,
> + &log, head);
> + vhost_log_write(vq, vq_log, log, size);
> + }
> + if (unlikely(rx_total_len >= VHOST_NET_WEIGHT)) {
> + vhost_poll_queue(&vq->poll);
> + break;
> + }
> + }
> +}
> +
> +static void handle_async_tx_events_notify(struct vhost_net *net,
> + struct vhost_virtqueue *vq)
> +{
> + struct kiocb *iocb = NULL;
> + int tx_total_len = 0;
> +
> + if (!is_async_vq(vq))
> + return;
> +
> + while ((iocb = notify_dequeue(vq)) != NULL) {
Please just write this as while (((iocb = notify_dequeue(vq)))
above as well
> + vhost_add_used_and_signal(&net->dev, vq,
> + iocb->ki_pos, 0);
pls indent continuation lines to the roght of (
above as well
> + tx_total_len += iocb->ki_nbytes;
> +
> + if (iocb->ki_dtor)
> + iocb->ki_dtor(iocb);
same question as above
> +
> + kmem_cache_free(net->cache, iocb);
> + if (unlikely(tx_total_len >= VHOST_NET_WEIGHT)) {
> + vhost_poll_queue(&vq->poll);
> + break;
> + }
> + }
> +}
> +
> +static struct kiocb *create_iocb(struct vhost_net *net,
> + struct vhost_virtqueue *vq,
> + unsigned head)
> +{
> + struct kiocb *iocb = NULL;
> +
> + if (!is_async_vq(vq))
> + return NULL;
> +
> + iocb = kmem_cache_zalloc(net->cache, GFP_KERNEL);
> + if (!iocb)
> + return NULL;
> + iocb->private = vq;
> + iocb->ki_pos = head;
> + iocb->ki_dtor = handle_iocb;
So, dtor calls handle_iocb, but what causes vhost
to wake-up is really poll, right?
> + if (vq == &net->dev.vqs[VHOST_NET_VQ_RX]) {
> + iocb->ki_user_data = vq->num;
Is the above used?
> + iocb->ki_iovec = vq->hdr;
> + }
> + return iocb;
> +}
> +
> /* Expects to be always run from workqueue - which acts as
> * read-size critical section for our kind of RCU. */
> static void handle_tx(struct vhost_net *net)
> {
> struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_TX];
> + struct kiocb *iocb = NULL;
Why do we need to init iocb to NULL?
> unsigned head, out, in, s;
> struct msghdr msg = {
> .msg_name = NULL,
> @@ -129,6 +259,8 @@ static void handle_tx(struct vhost_net *net)
> tx_poll_stop(net);
> hdr_size = vq->hdr_size;
>
> + handle_async_tx_events_notify(net, vq);
> +
> for (;;) {
> head = vhost_get_vq_desc(&net->dev, vq, vq->iov,
> ARRAY_SIZE(vq->iov),
> @@ -156,6 +288,13 @@ static void handle_tx(struct vhost_net *net)
> /* Skip header. TODO: support TSO. */
> s = move_iovec_hdr(vq->iov, vq->hdr, hdr_size, out);
> msg.msg_iovlen = out;
> +
> + if (is_async_vq(vq)) {
> + iocb = create_iocb(net, vq, head);
> + if (!iocb)
> + break;
> + }
> +
> len = iov_length(vq->iov, out);
> /* Sanity check */
> if (!len) {
> @@ -165,12 +304,18 @@ static void handle_tx(struct vhost_net *net)
> break;
> }
> /* TODO: Check specific error and bomb out unless ENOBUFS? */
> - err = sock->ops->sendmsg(NULL, sock, &msg, len);
> + err = sock->ops->sendmsg(iocb, sock, &msg, len);
> if (unlikely(err < 0)) {
> + if (is_async_vq(vq))
> + kmem_cache_free(net->cache, iocb);
> vhost_discard_vq_desc(vq);
> tx_poll_start(net, sock);
> break;
> }
> +
> + if (is_async_vq(vq))
> + continue;
> +
> if (err != len)
> pr_err("Truncated TX packet: "
> " len %d != %zd\n", err, len);
> @@ -182,6 +327,8 @@ static void handle_tx(struct vhost_net *net)
> }
> }
>
> + handle_async_tx_events_notify(net, vq);
> +
> mutex_unlock(&vq->mutex);
> unuse_mm(net->dev.mm);
> }
> @@ -191,6 +338,7 @@ static void handle_tx(struct vhost_net *net)
> static void handle_rx(struct vhost_net *net)
> {
> struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_RX];
> + struct kiocb *iocb = NULL;
> unsigned head, out, in, log, s;
> struct vhost_log *vq_log;
> struct msghdr msg = {
> @@ -211,7 +359,8 @@ static void handle_rx(struct vhost_net *net)
> int err;
> size_t hdr_size;
> struct socket *sock = rcu_dereference(vq->private_data);
> - if (!sock || skb_queue_empty(&sock->sk->sk_receive_queue))
> + if (!sock || (skb_queue_empty(&sock->sk->sk_receive_queue) &&
> + vq->link_state == VHOST_VQ_LINK_SYNC))
> return;
>
> use_mm(net->dev.mm);
> @@ -219,9 +368,17 @@ static void handle_rx(struct vhost_net *net)
> vhost_disable_notify(vq);
> hdr_size = vq->hdr_size;
>
> + /* In async cases, when write log is enabled, in case the submitted
> + * buffers did not get log info before the log enabling, so we'd
> + * better recompute the log info when needed. We do this in
> + * handle_async_rx_events_notify().
> + */
> +
> vq_log = unlikely(vhost_has_feature(&net->dev, VHOST_F_LOG_ALL)) ?
> vq->log : NULL;
>
> + handle_async_rx_events_notify(net, vq, sock);
> +
> for (;;) {
> head = vhost_get_vq_desc(&net->dev, vq, vq->iov,
> ARRAY_SIZE(vq->iov),
> @@ -250,6 +407,13 @@ static void handle_rx(struct vhost_net *net)
> s = move_iovec_hdr(vq->iov, vq->hdr, hdr_size, in);
> msg.msg_iovlen = in;
> len = iov_length(vq->iov, in);
> +
> + if (is_async_vq(vq)) {
> + iocb = create_iocb(net, vq, head);
> + if (!iocb)
> + break;
> + }
> +
> /* Sanity check */
> if (!len) {
> vq_err(vq, "Unexpected header len for RX: "
> @@ -257,13 +421,20 @@ static void handle_rx(struct vhost_net *net)
> iov_length(vq->hdr, s), hdr_size);
> break;
> }
> - err = sock->ops->recvmsg(NULL, sock, &msg,
> +
> + err = sock->ops->recvmsg(iocb, sock, &msg,
> len, MSG_DONTWAIT | MSG_TRUNC);
> /* TODO: Check specific error and bomb out unless EAGAIN? */
> if (err < 0) {
> + if (is_async_vq(vq))
> + kmem_cache_free(net->cache, iocb);
> vhost_discard_vq_desc(vq);
> break;
> }
> +
> + if (is_async_vq(vq))
> + continue;
> +
> /* TODO: Should check and handle checksum. */
> if (err > len) {
> pr_err("Discarded truncated rx packet: "
> @@ -289,6 +460,8 @@ static void handle_rx(struct vhost_net *net)
> }
> }
>
> + handle_async_rx_events_notify(net, vq, sock);
> +
> mutex_unlock(&vq->mutex);
> unuse_mm(net->dev.mm);
> }
> @@ -342,6 +515,7 @@ static int vhost_net_open(struct inode *inode, struct file *f)
> vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, POLLOUT);
> vhost_poll_init(n->poll + VHOST_NET_VQ_RX, handle_rx_net, POLLIN);
> n->tx_poll_state = VHOST_NET_POLL_DISABLED;
> + n->cache = NULL;
>
> f->private_data = n;
>
> @@ -405,6 +579,18 @@ static void vhost_net_flush(struct vhost_net *n)
> vhost_net_flush_vq(n, VHOST_NET_VQ_RX);
> }
>
> +static void vhost_async_cleanup(struct vhost_net *n)
> +{
> + /* clean the notifier */
> + struct vhost_virtqueue *vq = &n->dev.vqs[VHOST_NET_VQ_RX];
> + struct kiocb *iocb = NULL;
> + if (n->cache) {
> + while ((iocb = notify_dequeue(vq)) != NULL)
> + kmem_cache_free(n->cache, iocb);
> + kmem_cache_destroy(n->cache);
> + }
> +}
> +
> static int vhost_net_release(struct inode *inode, struct file *f)
> {
> struct vhost_net *n = f->private_data;
> @@ -421,6 +607,7 @@ static int vhost_net_release(struct inode *inode, struct file *f)
> /* We do an extra flush before freeing memory,
> * since jobs can re-queue themselves. */
> vhost_net_flush(n);
> + vhost_async_cleanup(n);
> kfree(n);
> return 0;
> }
> @@ -472,21 +659,58 @@ static struct socket *get_tap_socket(int fd)
> return sock;
> }
>
> -static struct socket *get_socket(int fd)
> +static struct socket *get_mp_socket(int fd)
> +{
> + struct file *file = fget(fd);
> + struct socket *sock;
> + if (!file)
> + return ERR_PTR(-EBADF);
> + sock = mp_get_socket(file);
> + if (IS_ERR(sock))
> + fput(file);
> + return sock;
> +}
> +
> +static struct socket *get_socket(struct vhost_virtqueue *vq, int fd,
> + enum vhost_vq_link_state *state)
> {
> struct socket *sock;
> /* special case to disable backend */
> if (fd == -1)
> return NULL;
> +
> + *state = VHOST_VQ_LINK_SYNC;
> +
> sock = get_raw_socket(fd);
> if (!IS_ERR(sock))
> return sock;
> sock = get_tap_socket(fd);
> if (!IS_ERR(sock))
> return sock;
> + sock = get_mp_socket(fd);
> + if (!IS_ERR(sock)) {
> + *state = VHOST_VQ_LINK_ASYNC;
> + return sock;
> + }
> return ERR_PTR(-ENOTSOCK);
> }
>
> +static void vhost_init_link_state(struct vhost_net *n, int index)
so let's pass link state as parameter, and set it in this function.
And maybe pass in vq, no need for index tricks.
> +{
> + struct vhost_virtqueue *vq = n->vqs + index;
> +
> + WARN_ON(!mutex_is_locked(&vq->mutex));
there's a single place of call, I don't think we need this check.
> + if (vq->link_state == VHOST_VQ_LINK_ASYNC) {
> + INIT_LIST_HEAD(&vq->notifier);
> + spin_lock_init(&vq->notify_lock);
> + if (!n->cache) {
> + n->cache = kmem_cache_create("vhost_kiocb",
vhost_net_kiocb a better name
> + sizeof(struct kiocb), 0,
> + SLAB_HWCACHE_ALIGN, NULL);
> + }
no need for {} for single statement if.
> + }
> +}
> +
> static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
> {
> struct socket *sock, *oldsock;
> @@ -510,12 +734,14 @@ static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
> r = -EFAULT;
> goto err_vq;
> }
> - sock = get_socket(fd);
> + sock = get_socket(vq, fd, &vq->link_state);
> if (IS_ERR(sock)) {
> r = PTR_ERR(sock);
> goto err_vq;
> }
>
> + vhost_init_link_state(n, index);
> +
> /* start polling new socket */
> oldsock = vq->private_data;
> if (sock == oldsock)
> diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> index 3f10194..add77d3 100644
> --- a/drivers/vhost/vhost.c
> +++ b/drivers/vhost/vhost.c
> @@ -860,61 +860,17 @@ static unsigned get_indirect(struct vhost_dev *dev, struct vhost_virtqueue *vq,
> return 0;
> }
>
> -/* This looks in the virtqueue and for the first available buffer, and converts
> - * it to an iovec for convenient access. Since descriptors consist of some
> - * number of output then some number of input descriptors, it's actually two
> - * iovecs, but we pack them into one and note how many of each there were.
> - *
> - * This function returns the descriptor number found, or vq->num (which
> - * is never a valid descriptor number) if none was found. */
> -unsigned vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
> - struct iovec iov[], unsigned int iov_size,
> - unsigned int *out_num, unsigned int *in_num,
> - struct vhost_log *log, unsigned int *log_num)
> +/* This computes the log info according to the index of buffer */
> +unsigned __vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
> + struct iovec iov[], unsigned int iov_size,
> + unsigned int *out_num, unsigned int *in_num,
> + struct vhost_log *log, unsigned int *log_num,
> + unsigned int head)
> {
> struct vring_desc desc;
> unsigned int i, head, found = 0;
> - u16 last_avail_idx;
> - int ret;
> -
> - /* Check it isn't doing very strange things with descriptor numbers. */
> - last_avail_idx = vq->last_avail_idx;
> - if (get_user(vq->avail_idx, &vq->avail->idx)) {
> - vq_err(vq, "Failed to access avail idx at %p\n",
> - &vq->avail->idx);
> - return vq->num;
> - }
> -
> - if ((u16)(vq->avail_idx - last_avail_idx) > vq->num) {
> - vq_err(vq, "Guest moved used index from %u to %u",
> - last_avail_idx, vq->avail_idx);
> - return vq->num;
> - }
> -
> - /* If there's nothing new since last we looked, return invalid. */
> - if (vq->avail_idx == last_avail_idx)
> - return vq->num;
> + unsigned int ret;
>
> - /* Only get avail ring entries after they have been exposed by guest. */
> - smp_rmb();
> -
> - /* Grab the next descriptor number they're advertising, and increment
> - * the index we've seen. */
> - if (get_user(head, &vq->avail->ring[last_avail_idx % vq->num])) {
> - vq_err(vq, "Failed to read head: idx %d address %p\n",
> - last_avail_idx,
> - &vq->avail->ring[last_avail_idx % vq->num]);
> - return vq->num;
> - }
> -
> - /* If their number is silly, that's an error. */
> - if (head >= vq->num) {
> - vq_err(vq, "Guest says index %u > %u is available",
> - head, vq->num);
> - return vq->num;
> - }
> -
> - /* When we start there are none of either input nor output. */
> *out_num = *in_num = 0;
> if (unlikely(log))
> *log_num = 0;
> @@ -978,8 +934,70 @@ unsigned vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
> *out_num += ret;
> }
> } while ((i = next_desc(&desc)) != -1);
> + return head;
> +}
> +
> +/* This looks in the virtqueue and for the first available buffer, and converts
> + * it to an iovec for convenient access. Since descriptors consist of some
> + * number of output then some number of input descriptors, it's actually two
> + * iovecs, but we pack them into one and note how many of each there were.
> + *
> + * This function returns the descriptor number found, or vq->num (which
> + * is never a valid descriptor number) if none was found. */
> +unsigned vhost_get_vq_desc(struct vhost_dev *dev, struct vhost_virtqueue *vq,
> + struct iovec iov[], unsigned int iov_size,
> + unsigned int *out_num, unsigned int *in_num,
> + struct vhost_log *log, unsigned int *log_num)
> +{
> + struct vring_desc desc;
> + unsigned int i, head, found = 0;
> + u16 last_avail_idx;
> + unsigned int ret;
> +
> + /* Check it isn't doing very strange things with descriptor numbers. */
> + last_avail_idx = vq->last_avail_idx;
> + if (get_user(vq->avail_idx, &vq->avail->idx)) {
> + vq_err(vq, "Failed to access avail idx at %p\n",
> + &vq->avail->idx);
> + return vq->num;
> + }
> +
> + if ((u16)(vq->avail_idx - last_avail_idx) > vq->num) {
> + vq_err(vq, "Guest moved used index from %u to %u",
> + last_avail_idx, vq->avail_idx);
> + return vq->num;
> + }
> +
> + /* If there's nothing new since last we looked, return invalid. */
> + if (vq->avail_idx == last_avail_idx)
> + return vq->num;
> +
> + /* Only get avail ring entries after they have been exposed by guest. */
> + rmb();
> +
> + /* Grab the next descriptor number they're advertising, and increment
> + * the index we've seen. */
> + if (get_user(head, &vq->avail->ring[last_avail_idx % vq->num])) {
> + vq_err(vq, "Failed to read head: idx %d address %p\n",
> + last_avail_idx,
> + &vq->avail->ring[last_avail_idx % vq->num]);
> + return vq->num;
> + }
> +
> + /* If their number is silly, that's an error. */
> + if (head >= vq->num) {
> + vq_err(vq, "Guest says index %u > %u is available",
> + head, vq->num);
> + return vq->num;
> + }
> +
> + ret = __vhost_get_vq_desc(dev, vq, iov, iov_size,
> + out_num, in_num,
> + log, log_num, head);
>
> /* On success, increment avail index. */
> + if (ret == vq->num)
> + return ret;
> vq->last_avail_idx++;
> return head;
> }
> diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
> index 44591ba..3c9cbce 100644
> --- a/drivers/vhost/vhost.h
> +++ b/drivers/vhost/vhost.h
> @@ -43,6 +43,11 @@ struct vhost_log {
> u64 len;
> };
>
> +enum vhost_vq_link_state {
> + VHOST_VQ_LINK_SYNC = 0,
> + VHOST_VQ_LINK_ASYNC = 1,
> +};
> +
> /* The virtqueue structure describes a queue attached to a device. */
> struct vhost_virtqueue {
> struct vhost_dev *dev;
> @@ -96,6 +101,10 @@ struct vhost_virtqueue {
> /* Log write descriptors */
> void __user *log_base;
> struct vhost_log log[VHOST_NET_MAX_SG];
> + /* Differiate async socket for 0-copy from normal */
> + enum vhost_vq_link_state link_state;
> + struct list_head notifier;
> + spinlock_t notify_lock;
> };
>
> struct vhost_dev {
> @@ -124,6 +133,11 @@ unsigned vhost_get_vq_desc(struct vhost_dev *, struct vhost_virtqueue *,
> struct iovec iov[], unsigned int iov_count,
> unsigned int *out_num, unsigned int *in_num,
> struct vhost_log *log, unsigned int *log_num);
> +unsigned __vhost_get_vq_desc(struct vhost_dev *, struct vhost_virtqueue *,
> + struct iovec iov[], unsigned int iov_count,
> + unsigned int *out_num, unsigned int *in_num,
> + struct vhost_log *log, unsigned int *log_num,
> + unsigned int head);
> void vhost_discard_vq_desc(struct vhost_virtqueue *);
>
> int vhost_add_used(struct vhost_virtqueue *, unsigned int head, int len);
> --
> 1.5.4.4
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox