* [PATCHv8 1/3] tun: export underlying socket
From: Michael S. Tsirkin @ 2009-11-04 15:55 UTC (permalink / raw)
To: netdev, virtualization, kvm, linux-kernel, mingo, linux-mm, akpm
In-Reply-To: <cover.1257349249.git.mst@redhat.com>
Tun device looks similar to a packet socket
in that both pass complete frames from/to userspace.
This patch fills in enough fields in the socket underlying tun driver
to support sendmsg/recvmsg operations, and message flags
MSG_TRUNC and MSG_DONTWAIT, and exports access to this socket
to modules. Regular read/write behaviour is unchanged.
This way, code using raw sockets to inject packets
into a physical device, can support injecting
packets into host network stack almost without modification.
First user of this interface will be vhost virtualization
accelerator.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: David S. Miller <davem@davemloft.net>
---
DaveM agreed that this will go in through Rusty's tree together
with vhost patch, because vhost is the first user of the new
interface.
drivers/net/tun.c | 101 +++++++++++++++++++++++++++++++++++++++---------
include/linux/if_tun.h | 14 +++++++
2 files changed, 96 insertions(+), 19 deletions(-)
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 4fdfa2a..18f8876 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -144,6 +144,7 @@ static int tun_attach(struct tun_struct *tun, struct file *file)
err = 0;
tfile->tun = tun;
tun->tfile = tfile;
+ tun->socket.file = file;
dev_hold(tun->dev);
sock_hold(tun->socket.sk);
atomic_inc(&tfile->count);
@@ -158,6 +159,7 @@ static void __tun_detach(struct tun_struct *tun)
/* Detach from net device */
netif_tx_lock_bh(tun->dev);
tun->tfile = NULL;
+ tun->socket.file = NULL;
netif_tx_unlock_bh(tun->dev);
/* Drop read queue */
@@ -387,7 +389,8 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
/* Notify and wake up reader process */
if (tun->flags & TUN_FASYNC)
kill_fasync(&tun->fasync, SIGIO, POLL_IN);
- wake_up_interruptible(&tun->socket.wait);
+ wake_up_interruptible_poll(&tun->socket.wait, POLLIN |
+ POLLRDNORM | POLLRDBAND);
return NETDEV_TX_OK;
drop:
@@ -743,7 +746,7 @@ static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
len = min_t(int, skb->len, len);
skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
- total += len;
+ total += skb->len;
tun->dev->stats.tx_packets++;
tun->dev->stats.tx_bytes += len;
@@ -751,34 +754,23 @@ static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
return total;
}
-static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
- unsigned long count, loff_t pos)
+static ssize_t tun_do_read(struct tun_struct *tun,
+ struct kiocb *iocb, const struct iovec *iv,
+ ssize_t len, int noblock)
{
- struct file *file = iocb->ki_filp;
- struct tun_file *tfile = file->private_data;
- struct tun_struct *tun = __tun_get(tfile);
DECLARE_WAITQUEUE(wait, current);
struct sk_buff *skb;
- ssize_t len, ret = 0;
-
- if (!tun)
- return -EBADFD;
+ ssize_t ret = 0;
DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
- len = iov_length(iv, count);
- if (len < 0) {
- ret = -EINVAL;
- goto out;
- }
-
add_wait_queue(&tun->socket.wait, &wait);
while (len) {
current->state = TASK_INTERRUPTIBLE;
/* Read frames from the queue */
if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
- if (file->f_flags & O_NONBLOCK) {
+ if (noblock) {
ret = -EAGAIN;
break;
}
@@ -805,6 +797,27 @@ static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
current->state = TASK_RUNNING;
remove_wait_queue(&tun->socket.wait, &wait);
+ return ret;
+}
+
+static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
+ unsigned long count, loff_t pos)
+{
+ struct file *file = iocb->ki_filp;
+ struct tun_file *tfile = file->private_data;
+ struct tun_struct *tun = __tun_get(tfile);
+ ssize_t len, ret;
+
+ if (!tun)
+ return -EBADFD;
+ len = iov_length(iv, count);
+ if (len < 0) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
+ ret = min_t(ssize_t, ret, len);
out:
tun_put(tun);
return ret;
@@ -847,7 +860,8 @@ static void tun_sock_write_space(struct sock *sk)
return;
if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
- wake_up_interruptible_sync(sk->sk_sleep);
+ wake_up_interruptible_sync_poll(sk->sk_sleep, POLLOUT |
+ POLLWRNORM | POLLWRBAND);
tun = container_of(sk, struct tun_sock, sk)->tun;
kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
@@ -858,6 +872,37 @@ static void tun_sock_destruct(struct sock *sk)
free_netdev(container_of(sk, struct tun_sock, sk)->tun->dev);
}
+static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
+ struct msghdr *m, size_t total_len)
+{
+ struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
+ return tun_get_user(tun, m->msg_iov, total_len,
+ m->msg_flags & MSG_DONTWAIT);
+}
+
+static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
+ struct msghdr *m, size_t total_len,
+ int flags)
+{
+ struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
+ int ret;
+ if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
+ return -EINVAL;
+ ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
+ flags & MSG_DONTWAIT);
+ if (ret > total_len) {
+ m->msg_flags |= MSG_TRUNC;
+ ret = flags & MSG_TRUNC ? ret : total_len;
+ }
+ return ret;
+}
+
+/* Ops structure to mimic raw sockets with tun */
+static const struct proto_ops tun_socket_ops = {
+ .sendmsg = tun_sendmsg,
+ .recvmsg = tun_recvmsg,
+};
+
static struct proto tun_proto = {
.name = "tun",
.owner = THIS_MODULE,
@@ -986,6 +1031,7 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
goto err_free_dev;
init_waitqueue_head(&tun->socket.wait);
+ tun->socket.ops = &tun_socket_ops;
sock_init_data(&tun->socket, sk);
sk->sk_write_space = tun_sock_write_space;
sk->sk_sndbuf = INT_MAX;
@@ -1489,6 +1535,23 @@ static void tun_cleanup(void)
rtnl_link_unregister(&tun_link_ops);
}
+/* Get an underlying socket object from tun file. Returns error unless file is
+ * attached to a device. The returned object works like a packet socket, it
+ * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
+ * holding a reference to the file for as long as the socket is in use. */
+struct socket *tun_get_socket(struct file *file)
+{
+ struct tun_struct *tun;
+ if (file->f_op != &tun_fops)
+ return ERR_PTR(-EINVAL);
+ tun = tun_get(file);
+ if (!tun)
+ return ERR_PTR(-EBADFD);
+ tun_put(tun);
+ return &tun->socket;
+}
+EXPORT_SYMBOL_GPL(tun_get_socket);
+
module_init(tun_init);
module_exit(tun_cleanup);
MODULE_DESCRIPTION(DRV_DESCRIPTION);
diff --git a/include/linux/if_tun.h b/include/linux/if_tun.h
index 3f5fd52..404abe0 100644
--- a/include/linux/if_tun.h
+++ b/include/linux/if_tun.h
@@ -86,4 +86,18 @@ struct tun_filter {
__u8 addr[0][ETH_ALEN];
};
+#ifdef __KERNEL__
+#if defined(CONFIG_TUN) || defined(CONFIG_TUN_MODULE)
+struct socket *tun_get_socket(struct file *);
+#else
+#include <linux/err.h>
+#include <linux/errno.h>
+struct file;
+struct socket;
+static inline struct socket *tun_get_socket(struct file *f)
+{
+ return ERR_PTR(-EINVAL);
+}
+#endif /* CONFIG_TUN */
+#endif /* __KERNEL__ */
#endif /* __IF_TUN_H */
--
1.6.5.2.143.g8cc62
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply related
* [PATCHv8 0/3] vhost: a kernel-level virtio server
From: Michael S. Tsirkin @ 2009-11-04 15:52 UTC (permalink / raw)
To: netdev, virtualization, kvm, linux-kernel, mingo, linux-mm, akpm
Ok, I think I've addressed all comments so far here.
Rusty, I'd like this to go into linux-next, through your tree, and
hopefully 2.6.33. What do you think?
---
This implements vhost: a kernel-level backend for virtio,
The main motivation for this work is to reduce virtualization
overhead for virtio by removing system calls on data path,
without guest changes. For virtio-net, this removes up to
4 system calls per packet: vm exit for kick, reentry for kick,
iothread wakeup for packet, interrupt injection for packet.
This driver is pretty minimal, but it's fully functional (including
migration support interfaces), and already shows performance (especially
latency) improvement over userspace.
Some more detailed description attached to the patch itself.
The patches apply to both 2.6.32-rc6 and kvm.git. I'd like them to go
into linux-next if possible. Please comment.
Changelog from v7:
- Add note on RCU usage, mirroring this in vhost/vhost.h
- Fix locking typo noted by Eric Dumazet
- Fix warnings on 32 bit
Changelog from v6:
- review comments by Daniel Walker addressed
- checkpatch cleanup
- fix build on 32 bit
- maintainers entry corrected
Changelog from v5:
- tun support
- backends with virtio net header support (enables GSO, checksum etc)
- 32 bit compat fixed
- support indirect buffers, tx exit mitigation,
tx interrupt mitigation
- support write logging (allows migration without virtio ring code in userspace)
Changelog from v4:
- disable rx notification when have rx buffers
- addressed all comments from Rusty's review
- copy bugfixes from lguest commits:
ebf9a5a99c1a464afe0b4dfa64416fc8b273bc5c
e606490c440900e50ccf73a54f6fc6150ff40815
Changelog from v3:
- checkpatch fixes
Changelog from v2:
- Comments on RCU usage
- Compat ioctl support
- Make variable static
- Copied more idiomatic english from Rusty
Changes from v1:
- Move use_mm/unuse_mm from fs/aio.c to mm instead of copying.
- Reorder code to avoid need for forward declarations
- Kill a couple of debugging printks
Michael S. Tsirkin (3):
tun: export underlying socket
mm: export use_mm/unuse_mm to modules
vhost_net: a kernel-level virtio server
MAINTAINERS | 9 +
arch/x86/kvm/Kconfig | 1 +
drivers/Makefile | 1 +
drivers/net/tun.c | 101 ++++-
drivers/vhost/Kconfig | 11 +
drivers/vhost/Makefile | 2 +
drivers/vhost/net.c | 633 +++++++++++++++++++++++++++++
drivers/vhost/vhost.c | 970 ++++++++++++++++++++++++++++++++++++++++++++
drivers/vhost/vhost.h | 158 +++++++
include/linux/Kbuild | 1 +
include/linux/if_tun.h | 14 +
include/linux/miscdevice.h | 1 +
include/linux/vhost.h | 126 ++++++
mm/mmu_context.c | 3 +
14 files changed, 2012 insertions(+), 19 deletions(-)
create mode 100644 drivers/vhost/Kconfig
create mode 100644 drivers/vhost/Makefile
create mode 100644 drivers/vhost/net.c
create mode 100644 drivers/vhost/vhost.c
create mode 100644 drivers/vhost/vhost.h
create mode 100644 include/linux/vhost.h
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply
* [PATCH] ems_usb: Fix byte order issues on big endian machines
From: Sebastian Haas @ 2009-11-04 15:48 UTC (permalink / raw)
To: netdev; +Cc: socketcan-core
CPC-USB is using a ARM7 core with little endian byte order. The "id" field
in can_msg needs byte order conversion from/to CPU byte order.
Signed-off-by: Sebastian Haas <haas@ems-wuensche.com>
---
drivers/net/can/usb/ems_usb.c | 5 ++++-
1 files changed, 4 insertions(+), 1 deletions(-)
diff --git a/drivers/net/can/usb/ems_usb.c b/drivers/net/can/usb/ems_usb.c
index 9012e0a..abdbd9c 100644
--- a/drivers/net/can/usb/ems_usb.c
+++ b/drivers/net/can/usb/ems_usb.c
@@ -319,7 +319,7 @@ static void ems_usb_rx_can_msg(struct ems_usb *dev, struct ems_cpc_msg *msg)
cf = (struct can_frame *)skb_put(skb, sizeof(struct can_frame));
- cf->can_id = msg->msg.can_msg.id;
+ cf->can_id = le32_to_cpu(msg->msg.can_msg.id);
cf->can_dlc = min_t(u8, msg->msg.can_msg.length, 8);
if (msg->type == CPC_MSG_TYPE_EXT_CAN_FRAME
@@ -813,6 +813,9 @@ static netdev_tx_t ems_usb_start_xmit(struct sk_buff *skb, struct net_device *ne
msg->length = CPC_CAN_MSG_MIN_SIZE + cf->can_dlc;
}
+ /* Respect byte order */
+ msg->msg.can_msg.id = cpu_to_le32(msg->msg.can_msg.id);
+
for (i = 0; i < MAX_TX_URBS; i++) {
if (dev->tx_contexts[i].echo_index == MAX_TX_URBS) {
context = &dev->tx_contexts[i];
--
EMS Dr. Thomas Wuensche e.K.
Sonnenhang 3
85304 Ilmmuenster
HRA Neuburg a.d. Donau, HR-Nr. 70.106
Phone: +49-8441-490260
Fax : +49-8441-81860
http://www.ems-wuensche.com
^ permalink raw reply related
* [PATCH] ems_usb: Fix byte order issues on big endian machines
From: Sebastian Haas @ 2009-11-04 15:48 UTC (permalink / raw)
To: netdev-u79uwXL29TY76Z2rM5mHXA; +Cc: socketcan-core-0fE9KPoRgkgATYTw5x5z8w
CPC-USB is using a ARM7 core with little endian byte order. The "id" field
in can_msg needs byte order conversion from/to CPU byte order.
Signed-off-by: Sebastian Haas <haas-zsNKPWJ8Pib6hrUXjxyGrA@public.gmane.org>
---
drivers/net/can/usb/ems_usb.c | 5 ++++-
1 files changed, 4 insertions(+), 1 deletions(-)
diff --git a/drivers/net/can/usb/ems_usb.c b/drivers/net/can/usb/ems_usb.c
index 9012e0a..abdbd9c 100644
--- a/drivers/net/can/usb/ems_usb.c
+++ b/drivers/net/can/usb/ems_usb.c
@@ -319,7 +319,7 @@ static void ems_usb_rx_can_msg(struct ems_usb *dev, struct ems_cpc_msg *msg)
cf = (struct can_frame *)skb_put(skb, sizeof(struct can_frame));
- cf->can_id = msg->msg.can_msg.id;
+ cf->can_id = le32_to_cpu(msg->msg.can_msg.id);
cf->can_dlc = min_t(u8, msg->msg.can_msg.length, 8);
if (msg->type == CPC_MSG_TYPE_EXT_CAN_FRAME
@@ -813,6 +813,9 @@ static netdev_tx_t ems_usb_start_xmit(struct sk_buff *skb, struct net_device *ne
msg->length = CPC_CAN_MSG_MIN_SIZE + cf->can_dlc;
}
+ /* Respect byte order */
+ msg->msg.can_msg.id = cpu_to_le32(msg->msg.can_msg.id);
+
for (i = 0; i < MAX_TX_URBS; i++) {
if (dev->tx_contexts[i].echo_index == MAX_TX_URBS) {
context = &dev->tx_contexts[i];
--
EMS Dr. Thomas Wuensche e.K.
Sonnenhang 3
85304 Ilmmuenster
HRA Neuburg a.d. Donau, HR-Nr. 70.106
Phone: +49-8441-490260
Fax : +49-8441-81860
http://www.ems-wuensche.com
^ permalink raw reply related
* Re: [RFC,PATCH] mutex: mutex_is_owner() helper
From: Ingo Molnar @ 2009-11-04 15:40 UTC (permalink / raw)
To: Eric Dumazet, Peter Zijlstra, Linus Torvalds
Cc: David S. Miller, Linux Netdev List, linux kernel
In-Reply-To: <4AF19D06.3060401@gmail.com>
* Eric Dumazet <eric.dumazet@gmail.com> wrote:
> mutex_is_locked() is called most of the time to check if mutex is locked by current
> thread. But it's a lazy check, because mutex might be locked by another thread.
>
> Adds a new mutex_is_owned_by() helper, that can check ownership if CONFIG_SMP or
> CONFIG_DEBUG_MUTEXES are set.
>
> Returns are
> 0 if mutex is unlocked.
> 1 if locked
> -1 if not locked by designated thread.
>
> Last return value is possible only if CONFIG_SMP=y or CONFIG_DEBUG_MUTEXES=y
>
> Example of use :
>
> int rtnl_is_locked(void)
> {
> return mutex_is_locked(&rtnl_mutex);
> }
> ->
> int rtnl_is_locked(void)
> {
> return mutex_is_owned_by(&rtnl_mutex, current_thread_info()) == 1;
> }
To make sure this does not extend mutexes to be used a recursive
mutexes, mind naming it more clearly, like debug_mutex_is_owned(), and
adding a comment that says that this shouldnt be taken?
Also, it's somewhat imprecise: on !SMP && !DEBUG_MUTEXES we might return
a false '1'. Which happens to work for the rtnl usecase - but might not
in other cases.
Ingo
^ permalink raw reply
* [PATCH 25/25] mlx4_en: Loopback enable
From: Yevgeny Petrilin @ 2009-11-04 15:32 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
All packets are subject to loopback.
To enable loopback we write the dest mac to the wqe.
Signed-off-by: Yevgeny Petrilin <yevgenyp-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/en_netdev.c | 2 +-
drivers/net/mlx4/en_tx.c | 14 ++++++++++++++
drivers/net/mlx4/mlx4_en.h | 1 +
3 files changed, 16 insertions(+), 1 deletions(-)
diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index 8406fbb..947fb2f 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -108,7 +108,7 @@ static void mlx4_en_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
mutex_unlock(&mdev->state_lock);
}
-static u64 mlx4_en_mac_to_u64(u8 *addr)
+u64 mlx4_en_mac_to_u64(u8 *addr)
{
u64 mac = 0;
int i;
diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index a893733..cfbe361 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -599,6 +599,7 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
struct mlx4_wqe_data_seg *data;
struct skb_frag_struct *frag;
struct mlx4_en_tx_info *tx_info;
+ struct ethhdr *ethh;
int tx_ind = 0;
int nr_txbb;
int desc_size;
@@ -610,6 +611,8 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
int i;
int lso_header_size;
void *fragptr;
+ u64 mac;
+ u32 mac_l, mac_h;
real_size = get_real_size(skb, dev, &lso_header_size);
if (unlikely(!real_size))
@@ -675,6 +678,17 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
priv->port_stats.tx_chksum_offload++;
}
+ /* Copy dst mac address to wqe */
+ skb_reset_mac_header(skb);
+ ethh = eth_hdr(skb);
+ if (ethh && ethh->h_dest) {
+ mac = mlx4_en_mac_to_u64(ethh->h_dest);
+ mac_h = (u32) ((mac & 0xffff00000000) >> 16);
+ mac_l = (u32) (mac & 0xffffffff);
+ tx_desc->ctrl.srcrb_flags |= cpu_to_be32(mac_h);
+ tx_desc->ctrl.imm = cpu_to_be32(mac_l);
+ }
+
/* Handle LSO (TSO) packets */
if (lso_header_size) {
/* Mark opcode as LSO */
diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index b4731d5..02d07e9 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -564,6 +564,7 @@ int mlx4_SET_PORT_qpn_calc(struct mlx4_dev *dev, u8 port, u32 base_qpn,
int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev *mdev, u8 port, u8 reset);
+u64 mlx4_en_mac_to_u64(u8 *addr);
/*
* Globals
*/
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 21/25] mlx4: Adding protocol fields to multicast group
From: Yevgeny Petrilin @ 2009-11-04 15:32 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
The multicast attachment mechanism will be used both by
IB and Ethernet, so we need to specify for each multicast
address (whether it is gid or mac) its protocol.
Signed-off-by: Yevgeny Petrilin <yevgenyp-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/infiniband/hw/mlx4/main.c | 6 +++-
drivers/net/mlx4/mcg.c | 41 +++++++++++++++++++++---------------
include/linux/mlx4/device.h | 12 +++++++++-
3 files changed, 38 insertions(+), 21 deletions(-)
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 3cb3f47..5b67a31 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -450,13 +450,15 @@ static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
return mlx4_multicast_attach(to_mdev(ibqp->device)->dev,
&to_mqp(ibqp)->mqp, gid->raw,
!!(to_mqp(ibqp)->flags &
- MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK));
+ MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK),
+ MLX4_PROT_IB_IPV6);
}
static int mlx4_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
{
return mlx4_multicast_detach(to_mdev(ibqp->device)->dev,
- &to_mqp(ibqp)->mqp, gid->raw);
+ &to_mqp(ibqp)->mqp, gid->raw,
+ MLX4_PROT_IB_IPV6);
}
static int init_node_data(struct mlx4_ib_dev *dev)
diff --git a/drivers/net/mlx4/mcg.c b/drivers/net/mlx4/mcg.c
index daa08f1..c4f928a 100644
--- a/drivers/net/mlx4/mcg.c
+++ b/drivers/net/mlx4/mcg.c
@@ -96,7 +96,8 @@ static int mlx4_MGID_HASH(struct mlx4_dev *dev, struct mlx4_cmd_mailbox *mailbox
* entry in hash chain and *mgm holds end of hash chain.
*/
static int find_mgm(struct mlx4_dev *dev,
- u8 *gid, struct mlx4_cmd_mailbox *mgm_mailbox,
+ u8 *gid, enum mlx4_protocol prot,
+ struct mlx4_cmd_mailbox *mgm_mailbox,
u16 *hash, int *prev, int *index)
{
struct mlx4_cmd_mailbox *mailbox;
@@ -135,8 +136,9 @@ static int find_mgm(struct mlx4_dev *dev,
return err;
}
- if (!memcmp(mgm->gid, gid, 16))
- return err;
+ if (!memcmp(mgm->gid, gid, 16) &&
+ (prot == be32_to_cpu(mgm->members_count) >> 30))
+ return err;
*prev = *index;
*index = be32_to_cpu(mgm->next_gid_index) >> 6;
@@ -154,14 +156,17 @@ int mlx4_MCAST_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
qp.qpn = vhcr->in_modifier & 0xffffff;
if (vhcr->op_modifier)
- return mlx4_multicast_attach(dev, &qp, inbox->buf, vhcr->in_modifier >> 31);
+ return mlx4_multicast_attach(dev, &qp, inbox->buf,
+ vhcr->in_modifier >> 31,
+ (vhcr->in_modifier >> 28) & 0x7);
else
- return mlx4_multicast_detach(dev, &qp, inbox->buf);
+ return mlx4_multicast_detach(dev, &qp, inbox->buf,
+ (vhcr->in_modifier >> 28) & 0x7);
}
static int mlx4_MCAST(struct mlx4_dev *dev, struct mlx4_qp *qp,
- u8 gid[16], u8 attach,
- u8 block_loopback)
+ u8 gid[16], u8 attach, u8 block_loopback,
+ enum mlx4_protocol prot)
{
struct mlx4_cmd_mailbox *mailbox;
int err;
@@ -176,6 +181,7 @@ static int mlx4_MCAST(struct mlx4_dev *dev, struct mlx4_qp *qp,
memcpy(mailbox->buf, gid, 16);
qpn = qp->qpn;
+ qpn |= (prot << 28);
if (attach && block_loopback)
qpn |= (1 << 31);
@@ -186,7 +192,7 @@ static int mlx4_MCAST(struct mlx4_dev *dev, struct mlx4_qp *qp,
}
int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
- int block_mcast_loopback)
+ int block_mcast_loopback, enum mlx4_protocol prot)
{
struct mlx4_priv *priv = mlx4_priv(dev);
struct mlx4_cmd_mailbox *mailbox;
@@ -199,7 +205,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
int err;
if (mlx4_is_slave(dev))
- return mlx4_MCAST(dev, qp, gid, 1, block_mcast_loopback);
+ return mlx4_MCAST(dev, qp, gid, 1, block_mcast_loopback, prot);
mailbox = mlx4_alloc_cmd_mailbox(dev);
if (IS_ERR(mailbox))
@@ -208,7 +214,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
mutex_lock(&priv->mcg_table.mutex);
- err = find_mgm(dev, gid, mailbox, &hash, &prev, &index);
+ err = find_mgm(dev, gid, prot, mailbox, &hash, &prev, &index);
if (err)
goto out;
@@ -230,7 +236,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
memcpy(mgm->gid, gid, 16);
}
- members_count = be32_to_cpu(mgm->members_count);
+ members_count = be32_to_cpu(mgm->members_count) & 0xffffff;
if (members_count == MLX4_QP_PER_MGM) {
mlx4_err(dev, "MGM at index %x is full.\n", index);
err = -ENOMEM;
@@ -250,7 +256,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
else
mgm->qp[members_count++] = cpu_to_be32(qp->qpn & MGM_QPN_MASK);
- mgm->members_count = cpu_to_be32(members_count);
+ mgm->members_count = cpu_to_be32(members_count | ((u32) prot << 30));
err = mlx4_WRITE_MCG(dev, index, mailbox);
if (err)
@@ -285,7 +291,8 @@ out:
}
EXPORT_SYMBOL_GPL(mlx4_multicast_attach);
-int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
+int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
+ enum mlx4_protocol prot)
{
struct mlx4_priv *priv = mlx4_priv(dev);
struct mlx4_cmd_mailbox *mailbox;
@@ -297,7 +304,7 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
int err;
if (mlx4_is_slave(dev))
- return mlx4_MCAST(dev, qp, gid, 0, 0);
+ return mlx4_MCAST(dev, qp, gid, 0, 0, prot);
mailbox = mlx4_alloc_cmd_mailbox(dev);
if (IS_ERR(mailbox))
@@ -306,7 +313,7 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
mutex_lock(&priv->mcg_table.mutex);
- err = find_mgm(dev, gid, mailbox, &hash, &prev, &index);
+ err = find_mgm(dev, gid, prot, mailbox, &hash, &prev, &index);
if (err)
goto out;
@@ -316,7 +323,7 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
goto out;
}
- members_count = be32_to_cpu(mgm->members_count);
+ members_count = be32_to_cpu(mgm->members_count) & 0xffffff;
for (loc = -1, i = 0; i < members_count; ++i)
if ((be32_to_cpu(mgm->qp[i]) & MGM_QPN_MASK) == qp->qpn)
loc = i;
@@ -328,7 +335,7 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
}
- mgm->members_count = cpu_to_be32(--members_count);
+ mgm->members_count = cpu_to_be32(--members_count | ((u32) prot << 30));
mgm->qp[loc] = mgm->qp[i - 1];
mgm->qp[i - 1] = 0;
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index 246b7bc..3d74198 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -171,6 +171,13 @@ enum mlx4_special_vlan_idx {
MLX4_VLAN_REGULAR
};
+enum mlx4_protocol {
+ MLX4_PROT_IB_IPV6 = 0,
+ MLX4_PROT_ETH,
+ MLX4_PROT_IB_IPV4,
+ MLX4_PROT_FCOE
+};
+
enum {
MLX4_NUM_FEXCH = 64 * 1024,
};
@@ -488,8 +495,9 @@ int mlx4_INIT_PORT(struct mlx4_dev *dev, int port);
int mlx4_CLOSE_PORT(struct mlx4_dev *dev, int port);
int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
- int block_mcast_loopback);
-int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16]);
+ int block_mcast_loopback, enum mlx4_protocol prot);
+int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
+ enum mlx4_protocol prot);
int mlx4_register_mac(struct mlx4_dev *dev, u8 port, u64 mac, int *qpn);
void mlx4_unregister_mac(struct mlx4_dev *dev, u8 port, int qpn);
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 19/25] mlx4: Randomizing mac addresses for slaves
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Signed-off-by: Yevgeny Petrilin <yevgenyp-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/fw.c | 6 ++++++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/drivers/net/mlx4/fw.c b/drivers/net/mlx4/fw.c
index 60889d3..9028592 100644
--- a/drivers/net/mlx4/fw.c
+++ b/drivers/net/mlx4/fw.c
@@ -32,6 +32,7 @@
* SOFTWARE.
*/
+#include <linux/etherdevice.h>
#include <linux/mlx4/cmd.h>
#include <linux/cache.h>
@@ -148,6 +149,7 @@ int mlx4_QUERY_SLAVE_CAP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vh
struct mlx4_cmd_mailbox *outbox)
{
struct mlx4_caps *caps = outbox->buf;
+ u8 rand_mac[6];
int i;
memcpy(caps, &dev->caps, sizeof *caps);
@@ -165,6 +167,10 @@ int mlx4_QUERY_SLAVE_CAP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vh
for (i = 1; i <= dev->caps.num_ports; ++i) {
caps->gid_table_len[i] = 1;
caps->pkey_table_len[i] = 1;
+#define MLX4_MAC_HEAD 0x2c9000000ULL
+ random_ether_addr(rand_mac);
+ caps->def_mac[i] = MLX4_MAC_HEAD | rand_mac[0] |
+ ((u64)(rand_mac[1]) << 8) | ((u64)(rand_mac[2]) << 16);
}
} else {
caps->sqp_demux = dev->num_slaves;
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 18/25] mlx4_en: Use reasonable resources for slaves.
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Each slave gets a single MSI-X vector when in multi function mode.
In this case each slave would allocate less resources (which
normally depend on number of completion vectors).
We set a minimum number of RX and TX queues to achieve an
acceptable performance in multi function mode.
Signed-off-by: Yevgeny Petrilin <yevgenyp-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/en_main.c | 12 +++++++-----
drivers/net/mlx4/en_netdev.c | 3 ++-
drivers/net/mlx4/en_tx.c | 2 +-
drivers/net/mlx4/mlx4_en.h | 4 +++-
4 files changed, 13 insertions(+), 8 deletions(-)
diff --git a/drivers/net/mlx4/en_main.c b/drivers/net/mlx4/en_main.c
index 507e11f..3883071 100644
--- a/drivers/net/mlx4/en_main.c
+++ b/drivers/net/mlx4/en_main.c
@@ -93,8 +93,6 @@ static int mlx4_en_get_profile(struct mlx4_en_dev *mdev)
params->prof[i].tx_ppp = pfctx;
params->prof[i].tx_ring_size = MLX4_EN_DEF_TX_RING_SIZE;
params->prof[i].rx_ring_size = MLX4_EN_DEF_RX_RING_SIZE;
- params->prof[i].tx_ring_num = MLX4_EN_NUM_TX_RINGS +
- (!!pfcrx) * MLX4_EN_NUM_PPP_RINGS;
}
return 0;
@@ -216,11 +214,15 @@ static void *mlx4_en_add(struct mlx4_dev *dev)
/* If we did not receive an explicit number of Rx rings, default to
* the number of completion vectors populated by the mlx4_core */
mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
+ mdev->profile.prof[i].tx_ring_num = max_t(int,
+ min_t(int, roundup_pow_of_two(dev->caps.num_comp_vectors),
+ NUM_TX_RINGS), MIN_TX_RINGS) +
+ (!!pfcrx) * MLX4_EN_NUM_PPP_RINGS;
mlx4_info(mdev, "Using %d tx rings for port:%d\n",
mdev->profile.prof[i].tx_ring_num, i);
- mdev->profile.prof[i].rx_ring_num = min_t(int,
- roundup_pow_of_two(dev->caps.num_comp_vectors),
- MAX_RX_RINGS);
+ mdev->profile.prof[i].rx_ring_num = max_t(int,
+ min_t(int, roundup_pow_of_two(dev->caps.num_comp_vectors),
+ MAX_RX_RINGS), MIN_RX_RINGS);
mlx4_info(mdev, "Defaulting to %d rx rings for port:%d\n",
mdev->profile.prof[i].rx_ring_num, i);
}
diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index fd96078..488f967 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -1024,7 +1024,8 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
*/
dev->netdev_ops = &mlx4_netdev_ops;
dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
- dev->real_num_tx_queues = MLX4_EN_NUM_TX_RINGS;
+ dev->real_num_tx_queues = priv->tx_ring_num -
+ !!(priv->prof->rx_ppp) * MLX4_EN_NUM_PPP_RINGS;
SET_ETHTOOL_OPS(dev, &mlx4_en_ethtool_ops);
diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
index 8c72799..a893733 100644
--- a/drivers/net/mlx4/en_tx.c
+++ b/drivers/net/mlx4/en_tx.c
@@ -583,7 +583,7 @@ u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb)
*/
if (priv->prof->rx_ppp && priv->vlgrp && vlan_tx_tag_present(skb)) {
vlan_tag = vlan_tx_tag_get(skb);
- return MLX4_EN_NUM_TX_RINGS + (vlan_tag >> 13);
+ return priv->tx_ring_num - MLX4_EN_NUM_PPP_RINGS + (vlan_tag >> 13);
}
return skb_tx_hash(dev, skb);
diff --git a/drivers/net/mlx4/mlx4_en.h b/drivers/net/mlx4/mlx4_en.h
index 4376147..6e0c9ed 100644
--- a/drivers/net/mlx4/mlx4_en.h
+++ b/drivers/net/mlx4/mlx4_en.h
@@ -95,6 +95,9 @@
#define MLX4_EN_PAGE_SIZE (1 << MLX4_EN_PAGE_SHIFT)
#define MAX_TX_RINGS 16
#define MAX_RX_RINGS 16
+#define MIN_RX_RINGS 4
+#define NUM_TX_RINGS 8
+#define MIN_TX_RINGS 4
#define TXBB_SIZE 64
#define HEADROOM (2048 / TXBB_SIZE + 1)
#define STAMP_STRIDE 64
@@ -137,7 +140,6 @@ enum {
#define MLX4_EN_MIN_TX_SIZE (4096 / TXBB_SIZE)
#define MLX4_EN_SMALL_PKT_SIZE 64
-#define MLX4_EN_NUM_TX_RINGS 8
#define MLX4_EN_NUM_PPP_RINGS 8
#define MLX4_EN_DEF_TX_RING_SIZE 512
#define MLX4_EN_DEF_RX_RING_SIZE 1024
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 17/25] mlx4: Changed Mac management
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
RX qp is given from a pool of previously reserved range of QPs.
Activating Mac steering now.
Giving an option to change mac without the need to change the RX QP.
All the mac management is done by the master, added wrappers for the
mac management.
Signed-off-by: Yevgeny Petrilin <yevgenyp-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/cmd.c | 19 ++++++++
drivers/net/mlx4/en_netdev.c | 32 +++++++-------
drivers/net/mlx4/en_port.c | 1 +
drivers/net/mlx4/en_port.h | 8 +++-
drivers/net/mlx4/en_rx.c | 11 +----
drivers/net/mlx4/main.c | 11 +++--
drivers/net/mlx4/mlx4.h | 7 ++-
drivers/net/mlx4/port.c | 96 ++++++++++++++++++++++++++++++++----------
include/linux/mlx4/cmd.h | 9 ++--
include/linux/mlx4/device.h | 5 +-
10 files changed, 137 insertions(+), 62 deletions(-)
diff --git a/drivers/net/mlx4/cmd.c b/drivers/net/mlx4/cmd.c
index 7afbfef..183ba8d 100644
--- a/drivers/net/mlx4/cmd.c
+++ b/drivers/net/mlx4/cmd.c
@@ -507,6 +507,24 @@ static int mlx4_RESOURCE_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vh
} else
mlx4_free_mtt_range(dev, param1 /* first */, param2 /* order */);
break;
+ case RES_MAC:
+ switch (vhcr->op) {
+ case MLX4_CMD_ALLOC_RES:
+ ret = mlx4_register_mac(dev, vhcr->op_modifier,
+ vhcr->in_param, (int *) &vhcr->out_param);
+ vhcr->errno = ret;
+ break;
+ case MLX4_CMD_FREE_RES:
+ mlx4_unregister_mac(dev, vhcr->op_modifier, vhcr->in_param);
+ break;
+ case MLX4_CMD_REPLACE_RES:
+ ret = mlx4_replace_mac(dev, vhcr->op_modifier,
+ vhcr->out_param, vhcr->in_param);
+ vhcr->errno = ret;
+ break;
+ default:
+ vhcr->errno = -EINVAL;
+ }
default:
vhcr->errno = -EINVAL;
}
@@ -538,6 +556,7 @@ static struct mlx4_cmd_info {
{MLX4_CMD_NOP, 0, 0, 0, NULL, NULL},
{MLX4_CMD_ALLOC_RES, 0, 0, 1, NULL, mlx4_RESOURCE_wrapper},
{MLX4_CMD_FREE_RES, 0, 0, 0, NULL, mlx4_RESOURCE_wrapper},
+ {MLX4_CMD_REPLACE_RES, 0, 0, 1, NULL, mlx4_RESOURCE_wrapper},
{MLX4_CMD_GET_EVENT, 0, 0, 1, NULL, mlx4_GET_EVENT_wrapper},
{MLX4_CMD_SW2HW_MPT, 1, 0, 0, NULL, NULL}, /* need verifier */
diff --git a/drivers/net/mlx4/en_netdev.c b/drivers/net/mlx4/en_netdev.c
index c48b0f4..fd96078 100644
--- a/drivers/net/mlx4/en_netdev.c
+++ b/drivers/net/mlx4/en_netdev.c
@@ -145,9 +145,8 @@ static void mlx4_en_do_set_mac(struct work_struct *work)
mutex_lock(&mdev->state_lock);
if (priv->port_up) {
/* Remove old MAC and insert the new one */
- mlx4_unregister_mac(mdev->dev, priv->port, priv->mac_index);
- err = mlx4_register_mac(mdev->dev, priv->port,
- priv->mac, &priv->mac_index);
+ err = mlx4_replace_mac(mdev->dev, priv->port,
+ priv->base_qpn, priv->mac);
if (err)
en_err(priv, "Failed changing HW MAC address\n");
} else
@@ -596,10 +595,19 @@ int mlx4_en_start_port(struct net_device *dev)
++rx_index;
}
+ /* Set port mac number */
+ en_dbg(DRV, priv, "Setting mac for port %d\n", priv->port);
+ err = mlx4_register_mac(mdev->dev, priv->port,
+ priv->mac, &priv->base_qpn);
+ if (err) {
+ en_err(priv, "Failed setting port mac\n");
+ goto cq_err;
+ }
+
err = mlx4_en_config_rss_steer(priv);
if (err) {
en_err(priv, "Failed configuring rss steering\n");
- goto cq_err;
+ goto mac_err;
}
/* Configure tx cq's and rings */
@@ -652,21 +660,13 @@ int mlx4_en_start_port(struct net_device *dev)
en_err(priv, "Failed setting default qp numbers\n");
goto tx_err;
}
- /* Set port mac number */
- en_dbg(DRV, priv, "Setting mac for port %d\n", priv->port);
- err = mlx4_register_mac(mdev->dev, priv->port,
- priv->mac, &priv->mac_index);
- if (err) {
- en_err(priv, "Failed setting port mac\n");
- goto tx_err;
- }
/* Init port */
en_dbg(HW, priv, "Initializing port\n");
err = mlx4_INIT_PORT(mdev->dev, priv->port);
if (err) {
en_err(priv, "Failed Initializing port\n");
- goto mac_err;
+ goto tx_err;
}
/* Schedule multicast task to populate multicast list */
@@ -676,8 +676,6 @@ int mlx4_en_start_port(struct net_device *dev)
netif_tx_start_all_queues(dev);
return 0;
-mac_err:
- mlx4_unregister_mac(mdev->dev, priv->port, priv->mac_index);
tx_err:
while (tx_index--) {
mlx4_en_deactivate_tx_ring(priv, &priv->tx_ring[tx_index]);
@@ -685,6 +683,8 @@ tx_err:
}
mlx4_en_release_rss_steer(priv);
+mac_err:
+ mlx4_unregister_mac(mdev->dev, priv->port, priv->base_qpn);
cq_err:
while (rx_index--)
mlx4_en_deactivate_cq(priv, &priv->rx_cq[rx_index]);
@@ -716,7 +716,7 @@ void mlx4_en_stop_port(struct net_device *dev)
mlx4_CLOSE_PORT(mdev->dev, priv->port);
/* Unregister Mac address for the port */
- mlx4_unregister_mac(mdev->dev, priv->port, priv->mac_index);
+ mlx4_unregister_mac(mdev->dev, priv->port, priv->base_qpn);
/* Free TX Rings */
for (i = 0; i < priv->tx_ring_num; i++) {
diff --git a/drivers/net/mlx4/en_port.c b/drivers/net/mlx4/en_port.c
index a29abe8..c099cb4 100644
--- a/drivers/net/mlx4/en_port.c
+++ b/drivers/net/mlx4/en_port.c
@@ -127,6 +127,7 @@ int mlx4_SET_PORT_qpn_calc(struct mlx4_dev *dev, u8 port, u32 base_qpn,
memset(context, 0, sizeof *context);
context->base_qpn = cpu_to_be32(base_qpn);
+ context->n_mac = 0x7;
context->promisc = cpu_to_be32(promisc << SET_PORT_PROMISC_SHIFT | base_qpn);
context->mcast = cpu_to_be32(1 << SET_PORT_PROMISC_SHIFT | base_qpn);
context->intra_no_vlan = 0;
diff --git a/drivers/net/mlx4/en_port.h b/drivers/net/mlx4/en_port.h
index e6477f1..674a804 100644
--- a/drivers/net/mlx4/en_port.h
+++ b/drivers/net/mlx4/en_port.h
@@ -59,14 +59,18 @@ struct mlx4_set_port_general_context {
struct mlx4_set_port_rqp_calc_context {
__be32 base_qpn;
+ u8 rererved;
+ u8 n_mac;
+ u8 n_vlan;
+ u8 n_prio;
__be32 flags;
- u8 reserved[3];
+ u8 reserved2[3];
u8 mac_miss;
u8 intra_no_vlan;
u8 no_vlan;
u8 intra_vlan_miss;
u8 vlan_miss;
- u8 reserved2[3];
+ u8 reserved3[3];
u8 no_vlan_prio;
__be32 promisc;
__be32 mcast;
diff --git a/drivers/net/mlx4/en_rx.c b/drivers/net/mlx4/en_rx.c
index 03b781a..d3884ce 100644
--- a/drivers/net/mlx4/en_rx.c
+++ b/drivers/net/mlx4/en_rx.c
@@ -865,16 +865,10 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
}
/* Configure RSS indirection qp */
- err = mlx4_qp_reserve_range(mdev->dev, 1, 1, &priv->base_qpn);
- if (err) {
- en_err(priv, "Failed to reserve range for RSS "
- "indirection qp\n");
- goto rss_err;
- }
err = mlx4_qp_alloc(mdev->dev, priv->base_qpn, &rss_map->indir_qp);
if (err) {
en_err(priv, "Failed to allocate RSS indirection QP\n");
- goto reserve_err;
+ goto rss_err;
}
rss_map->indir_qp.event = mlx4_en_sqp_event;
mlx4_en_fill_qp_context(priv, 0, 0, 0, 1, priv->base_qpn,
@@ -900,8 +894,6 @@ indir_err:
MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->indir_qp);
mlx4_qp_remove(mdev->dev, &rss_map->indir_qp);
mlx4_qp_free(mdev->dev, &rss_map->indir_qp);
-reserve_err:
- mlx4_qp_release_range(mdev->dev, priv->base_qpn, 1);
rss_err:
for (i = 0; i < good_qps; i++) {
mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i],
@@ -923,7 +915,6 @@ void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv)
MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->indir_qp);
mlx4_qp_remove(mdev->dev, &rss_map->indir_qp);
mlx4_qp_free(mdev->dev, &rss_map->indir_qp);
- mlx4_qp_release_range(mdev->dev, priv->base_qpn, 1);
for (i = 0; i < priv->rx_ring_num; i++) {
mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i],
diff --git a/drivers/net/mlx4/main.c b/drivers/net/mlx4/main.c
index c2b6b35..c5ee288 100644
--- a/drivers/net/mlx4/main.c
+++ b/drivers/net/mlx4/main.c
@@ -99,7 +99,7 @@ static struct mlx4_profile default_profile = {
.num_mtt = 1 << 20,
};
-static int log_num_mac = 2;
+static int log_num_mac = 7;
module_param_named(log_num_mac, log_num_mac, int, 0444);
MODULE_PARM_DESC(log_num_mac, "Log2 max number of MACs per ETH port (1-7)");
@@ -1150,9 +1150,12 @@ static int mlx4_init_port_info(struct mlx4_dev *dev, int port)
info->dev = dev;
info->port = port;
- mlx4_init_mac_table(dev, &info->mac_table);
- mlx4_init_vlan_table(dev, &info->vlan_table);
-
+ if (!mlx4_is_slave(dev)) {
+ mlx4_init_mac_table(dev, &info->mac_table);
+ mlx4_init_vlan_table(dev, &info->vlan_table);
+ info->base_qpn = dev->caps.reserved_qps_base[MLX4_QP_REGION_ETH_ADDR] +
+ (port - 1) * (1 << log_num_mac);
+ }
sprintf(info->dev_name, "mlx4_port%d", port);
info->port_attr.attr.name = info->dev_name;
info->port_attr.attr.mode = S_IRUGO | S_IWUSR;
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index 516e375..3e434e4 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -101,7 +101,10 @@ enum mlx4_resource {
RES_CQ,
RES_SRQ,
RES_MPT,
- RES_MTT
+ RES_MTT,
+ RES_MAC,
+ RES_VLAN,
+ RES_MCAST
};
enum mlx4_alloc_mode {
@@ -340,7 +343,6 @@ struct mlx4_catas_err {
struct mlx4_mac_table {
__be64 entries[MLX4_MAX_MAC_NUM];
- int refs[MLX4_MAX_MAC_NUM];
struct mutex mutex;
int total;
int max;
@@ -365,6 +367,7 @@ struct mlx4_port_info {
enum mlx4_port_type tmp_type;
struct mlx4_mac_table mac_table;
struct mlx4_vlan_table vlan_table;
+ int base_qpn;
};
struct mlx4_sense {
diff --git a/drivers/net/mlx4/port.c b/drivers/net/mlx4/port.c
index 67f0751..6e729dd 100644
--- a/drivers/net/mlx4/port.c
+++ b/drivers/net/mlx4/port.c
@@ -48,10 +48,8 @@ void mlx4_init_mac_table(struct mlx4_dev *dev, struct mlx4_mac_table *table)
int i;
mutex_init(&table->mutex);
- for (i = 0; i < MLX4_MAX_MAC_NUM; i++) {
+ for (i = 0; i < MLX4_MAX_MAC_NUM; i++)
table->entries[i] = 0;
- table->refs[i] = 0;
- }
table->max = 1 << dev->caps.log_num_macs;
table->total = 0;
}
@@ -90,24 +88,33 @@ static int mlx4_set_port_mac_table(struct mlx4_dev *dev, u8 port,
return err;
}
-int mlx4_register_mac(struct mlx4_dev *dev, u8 port, u64 mac, int *index)
+int mlx4_register_mac(struct mlx4_dev *dev, u8 port, u64 mac, int *qpn)
{
- struct mlx4_mac_table *table = &mlx4_priv(dev)->port[port].mac_table;
+ struct mlx4_port_info *info = &mlx4_priv(dev)->port[port];
+ struct mlx4_mac_table *table = &info->mac_table;
+ u64 out_param;
int i, err = 0;
int free = -1;
+ if (mlx4_is_slave(dev)) {
+ err = mlx4_cmd_imm(dev, mac, &out_param, RES_MAC, port,
+ MLX4_CMD_ALLOC_RES, MLX4_CMD_TIME_CLASS_A);
+ if (!err)
+ *qpn = out_param;
+ return err;
+ }
+
mlx4_dbg(dev, "Registering MAC: 0x%llx\n", (unsigned long long) mac);
mutex_lock(&table->mutex);
for (i = 0; i < MLX4_MAX_MAC_NUM - 1; i++) {
- if (free < 0 && !table->refs[i]) {
+ if (free < 0 && !table->entries[i]) {
free = i;
continue;
}
if (mac == (MLX4_MAC_MASK & be64_to_cpu(table->entries[i]))) {
- /* MAC already registered, increase refernce count */
- *index = i;
- ++table->refs[i];
+ /* MAC already registered, Must not have duplicates */
+ err = -EEXIST;
goto out;
}
}
@@ -120,18 +127,16 @@ int mlx4_register_mac(struct mlx4_dev *dev, u8 port, u64 mac, int *index)
}
/* Register new MAC */
- table->refs[free] = 1;
table->entries[free] = cpu_to_be64(mac | MLX4_MAC_VALID);
err = mlx4_set_port_mac_table(dev, port, table->entries);
if (unlikely(err)) {
mlx4_err(dev, "Failed adding MAC: 0x%llx\n", (unsigned long long) mac);
- table->refs[free] = 0;
table->entries[free] = 0;
goto out;
}
- *index = free;
+ *qpn = info->base_qpn + free;
++table->total;
out:
mutex_unlock(&table->mutex);
@@ -139,20 +144,35 @@ out:
}
EXPORT_SYMBOL_GPL(mlx4_register_mac);
-void mlx4_unregister_mac(struct mlx4_dev *dev, u8 port, int index)
+static int validate_index(struct mlx4_dev *dev,
+ struct mlx4_mac_table *table, int index)
{
- struct mlx4_mac_table *table = &mlx4_priv(dev)->port[port].mac_table;
+ int err = 0;
- mutex_lock(&table->mutex);
- if (!table->refs[index]) {
- mlx4_warn(dev, "No MAC entry for index %d\n", index);
- goto out;
+ if (index < 0 || index >= table->max || !table->entries[index]) {
+ mlx4_warn(dev, "No valid Mac entry for the given index\n");
+ err = -EINVAL;
}
- if (--table->refs[index]) {
- mlx4_warn(dev, "Have more references for index %d,"
- "no need to modify MAC table\n", index);
- goto out;
+ return err;
+}
+
+void mlx4_unregister_mac(struct mlx4_dev *dev, u8 port, int qpn)
+{
+ struct mlx4_port_info *info = &mlx4_priv(dev)->port[port];
+ struct mlx4_mac_table *table = &info->mac_table;
+ int index = qpn - info->base_qpn;
+
+ if (mlx4_is_slave(dev)) {
+ mlx4_cmd(dev, qpn, RES_MAC, port,
+ MLX4_CMD_FREE_RES, MLX4_CMD_TIME_CLASS_A);
+ return;
}
+
+ mutex_lock(&table->mutex);
+
+ if (validate_index(dev, table, index))
+ goto out;
+
table->entries[index] = 0;
mlx4_set_port_mac_table(dev, port, table->entries);
--table->total;
@@ -161,6 +181,38 @@ out:
}
EXPORT_SYMBOL_GPL(mlx4_unregister_mac);
+int mlx4_replace_mac(struct mlx4_dev *dev, u8 port, int qpn, u64 new_mac)
+{
+ struct mlx4_port_info *info = &mlx4_priv(dev)->port[port];
+ struct mlx4_mac_table *table = &info->mac_table;
+ int index = qpn - info->base_qpn;
+ int err;
+
+ if (mlx4_is_slave(dev)) {
+ err = mlx4_cmd_imm(dev, new_mac, (u64 *) &qpn, RES_MAC, port,
+ MLX4_CMD_REPLACE_RES, MLX4_CMD_TIME_CLASS_A);
+ return err;
+ }
+
+ mutex_lock(&table->mutex);
+
+ err = validate_index(dev, table, index);
+ if (err)
+ goto out;
+
+ table->entries[index] = cpu_to_be64(new_mac | MLX4_MAC_VALID);
+
+ err = mlx4_set_port_mac_table(dev, port, table->entries);
+ if (unlikely(err)) {
+ mlx4_err(dev, "Failed adding MAC: 0x%llx\n", (unsigned long long) new_mac);
+ table->entries[index] = 0;
+ }
+out:
+ mutex_unlock(&table->mutex);
+ return err;
+}
+EXPORT_SYMBOL_GPL(mlx4_replace_mac);
+
static int mlx4_set_port_vlan_table(struct mlx4_dev *dev, u8 port,
__be32 *entries)
{
diff --git a/include/linux/mlx4/cmd.h b/include/linux/mlx4/cmd.h
index f265091..f729c82 100644
--- a/include/linux/mlx4/cmd.h
+++ b/include/linux/mlx4/cmd.h
@@ -122,10 +122,11 @@ enum {
/* virtual commands */
MLX4_CMD_ALLOC_RES = 0x50,
MLX4_CMD_FREE_RES = 0x51,
- MLX4_CMD_GET_EVENT = 0x52,
- MLX4_CMD_QUERY_SLAVE_CAP = 0x53,
- MLX4_CMD_MCAST_ATTACH = 0x54,
- MLX4_CMD_GET_SLAVE_SQP = 0x55,
+ MLX4_CMD_REPLACE_RES = 0x52,
+ MLX4_CMD_GET_EVENT = 0x53,
+ MLX4_CMD_QUERY_SLAVE_CAP = 0x54,
+ MLX4_CMD_MCAST_ATTACH = 0x55,
+ MLX4_CMD_GET_SLAVE_SQP = 0x56,
/* debug commands */
MLX4_CMD_QUERY_DEBUG_MSG = 0x2a,
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index bbd398b..246b7bc 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -491,8 +491,9 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
int block_mcast_loopback);
int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16]);
-int mlx4_register_mac(struct mlx4_dev *dev, u8 port, u64 mac, int *index);
-void mlx4_unregister_mac(struct mlx4_dev *dev, u8 port, int index);
+int mlx4_register_mac(struct mlx4_dev *dev, u8 port, u64 mac, int *qpn);
+void mlx4_unregister_mac(struct mlx4_dev *dev, u8 port, int qpn);
+int mlx4_replace_mac(struct mlx4_dev *dev, u8 port, int qpn, u64 new_mac);
int mlx4_register_vlan(struct mlx4_dev *dev, u8 port, u16 vlan, int *index);
void mlx4_unregister_vlan(struct mlx4_dev *dev, u8 port, int index);
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 16/25] mlx4_core: boot sriov
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Add virtual function device ids.
Establish comm channel when sriov is enabled, and boot vfs through it.
Each slave gets one msi-X for completions, the master also gets one for
async events.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
Signed-off-by: Yevgeny Petrilin <yevgenyp-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/eq.c | 21 ++--
drivers/net/mlx4/main.c | 315 +++++++++++++++++++++++++++++++++++++----------
2 files changed, 260 insertions(+), 76 deletions(-)
diff --git a/drivers/net/mlx4/eq.c b/drivers/net/mlx4/eq.c
index b0c8022..29b2cd6 100644
--- a/drivers/net/mlx4/eq.c
+++ b/drivers/net/mlx4/eq.c
@@ -770,16 +770,18 @@ int mlx4_init_eq_table(struct mlx4_dev *dev)
goto err_out_unmap;
}
- err = mlx4_create_eq(dev, MLX4_NUM_ASYNC_EQE + MLX4_NUM_SPARE_EQE,
- (dev->flags & MLX4_FLAG_MSI_X) ? dev->caps.num_comp_vectors : 0,
- &priv->eq_table.eq[dev->caps.num_comp_vectors]);
- if (err)
- goto err_out_comp;
+ if (!mlx4_is_slave(dev)) {
+ err = mlx4_create_eq(dev, MLX4_NUM_ASYNC_EQE + MLX4_NUM_SPARE_EQE,
+ (dev->flags & MLX4_FLAG_MSI_X) ? dev->caps.num_comp_vectors : 0,
+ &priv->eq_table.eq[dev->caps.num_comp_vectors]);
+ if (err)
+ goto err_out_comp;
+ }
if (dev->flags & MLX4_FLAG_MSI_X) {
const char *eq_name;
- for (i = 0; i < dev->caps.num_comp_vectors + 1; ++i) {
+ for (i = 0; i < dev->caps.num_comp_vectors + !(mlx4_is_slave(dev)); ++i) {
if (i < dev->caps.num_comp_vectors) {
snprintf(priv->eq_table.irq_names +
i * MLX4_IRQNAME_SIZE,
@@ -825,13 +827,14 @@ int mlx4_init_eq_table(struct mlx4_dev *dev)
priv->eq_table.eq[dev->caps.num_comp_vectors].eqn, err);
}
- for (i = 0; i < dev->caps.num_comp_vectors + 1; ++i)
+ for (i = 0; i < dev->caps.num_comp_vectors + !(mlx4_is_slave(dev)); ++i)
eq_set_ci(&priv->eq_table.eq[i], 1);
return 0;
err_out_async:
- mlx4_free_eq(dev, &priv->eq_table.eq[dev->caps.num_comp_vectors]);
+ if (!mlx4_is_slave(dev))
+ mlx4_free_eq(dev, &priv->eq_table.eq[dev->caps.num_comp_vectors]);
err_out_comp:
i = dev->caps.num_comp_vectors;
@@ -866,7 +869,7 @@ void mlx4_cleanup_eq_table(struct mlx4_dev *dev)
mlx4_free_irqs(dev);
- for (i = 0; i < dev->caps.num_comp_vectors + 1; ++i)
+ for (i = 0; i < dev->caps.num_comp_vectors + !mlx4_is_slave(dev); ++i)
mlx4_free_eq(dev, &priv->eq_table.eq[i]);
if (!mlx4_is_slave(dev))
diff --git a/drivers/net/mlx4/main.c b/drivers/net/mlx4/main.c
index 2852445..c2b6b35 100644
--- a/drivers/net/mlx4/main.c
+++ b/drivers/net/mlx4/main.c
@@ -73,6 +73,18 @@ MODULE_PARM_DESC(msi_x, "attempt to use MSI-X if nonzero");
#endif /* CONFIG_PCI_MSI */
+#ifdef CONFIG_PCI_IOV
+
+static int sr_iov;
+module_param(sr_iov, int, 0444);
+MODULE_PARM_DESC(sr_iov, "enable #sr_iov functions if sr_iov > 0");
+
+static int probe_vf;
+module_param(probe_vf, int, 0444);
+MODULE_PARM_DESC(probe_vf, "number of vfs to probe by pf driver (sr_iov > 0)");
+
+#endif /* CONFIG_PCI_IOV */
+
static char mlx4_version[] __devinitdata =
DRV_NAME ": Mellanox ConnectX core driver v"
DRV_VERSION " (" DRV_RELDATE ")\n";
@@ -779,12 +791,56 @@ static void mlx4_free_icms(struct mlx4_dev *dev)
mlx4_free_icm(dev, priv->fw.aux_icm, 0);
}
+static void mlx4_slave_exit(struct mlx4_dev *dev)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+
+ down(&priv->cmd.poll_sem);
+ if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_RESET, 0, MLX4_COMM_TIME))
+ mlx4_warn(dev, "Failed to close slave function.\n");
+ up(&priv->cmd.poll_sem);
+}
+
static void mlx4_close_hca(struct mlx4_dev *dev)
{
- mlx4_CLOSE_HCA(dev, 0);
- mlx4_free_icms(dev);
- mlx4_UNMAP_FA(dev);
- mlx4_free_icm(dev, mlx4_priv(dev)->fw.fw_icm, 0);
+ if (mlx4_is_slave(dev))
+ mlx4_slave_exit(dev);
+ else {
+ mlx4_CLOSE_HCA(dev, 0);
+ mlx4_free_icms(dev);
+ mlx4_UNMAP_FA(dev);
+ mlx4_free_icm(dev, mlx4_priv(dev)->fw.fw_icm, 0);
+ }
+}
+
+static int mlx4_init_slave(struct mlx4_dev *dev)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ u64 dma = (u64) priv->mfunc.vhcr_dma;
+
+ down(&priv->cmd.poll_sem);
+ mlx4_warn(dev, "Sending reset\n");
+ if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_RESET, 0, MLX4_COMM_TIME))
+ goto err;
+ mlx4_warn(dev, "Sending vhcr0\n");
+ if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR0, dma >> 48,
+ MLX4_COMM_TIME))
+ goto err;
+ if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR1, dma >> 32,
+ MLX4_COMM_TIME))
+ goto err;
+ if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR2, dma >> 16,
+ MLX4_COMM_TIME))
+ goto err;
+ if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR_EN, dma, MLX4_COMM_TIME))
+ goto err;
+ up(&priv->cmd.poll_sem);
+ return 0;
+
+err:
+ mlx4_comm_cmd(dev, MLX4_COMM_CMD_RESET, 0, 0);
+ up(&priv->cmd.poll_sem);
+ return -EIO;
}
static int mlx4_init_hca(struct mlx4_dev *dev)
@@ -798,51 +854,67 @@ static int mlx4_init_hca(struct mlx4_dev *dev)
u64 icm_size;
int err;
- err = mlx4_QUERY_FW(dev);
- if (err) {
- if (err == -EACCES)
- mlx4_info(dev, "non-primary physical function, skipping.\n");
- else
- mlx4_err(dev, "QUERY_FW command failed, aborting.\n");
- return err;
- }
+ if (!mlx4_is_slave(dev)) {
+ err = mlx4_QUERY_FW(dev); /* TODO: verify FW version in slaves as well */
+ if (err) {
+ if (err == -EACCES)
+ mlx4_info(dev, "non-primary physical function, skipping.\n");
+ else {
+ mlx4_err(dev, "QUERY_FW command failed, aborting.\n");
+ if (mlx4_is_master(dev))
+ mlx4_err(dev, "Are you using SRIOV-enabled firmware?\n");
+ }
+ return err;
+ }
- err = mlx4_load_fw(dev);
- if (err) {
- mlx4_err(dev, "Failed to start FW, aborting.\n");
- return err;
- }
+ err = mlx4_load_fw(dev);
+ if (err) {
+ mlx4_err(dev, "Failed to start FW, aborting.\n");
+ return err;
+ }
- mlx4_cfg.log_pg_sz_m = 1;
- mlx4_cfg.log_pg_sz = 0;
- err = mlx4_MOD_STAT_CFG(dev, &mlx4_cfg);
- if (err)
- mlx4_warn(dev, "Failed to override log_pg_sz parameter\n");
+ mlx4_cfg.log_pg_sz_m = 1;
+ mlx4_cfg.log_pg_sz = 0;
+ err = mlx4_MOD_STAT_CFG(dev, &mlx4_cfg);
+ if (err)
+ mlx4_warn(dev, "Failed to override log_pg_sz parameter\n");
- err = mlx4_dev_cap(dev, &dev_cap);
- if (err) {
- mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
- goto err_stop_fw;
- }
+ err = mlx4_dev_cap(dev, &dev_cap);
+ if (err) {
+ mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
+ goto err_stop_fw;
+ }
- profile = default_profile;
+ profile = default_profile;
+ icm_size = mlx4_make_profile(dev, &profile, &dev_cap, &init_hca);
+ if ((long long) icm_size < 0) {
+ err = icm_size;
+ goto err_stop_fw;
+ }
- icm_size = mlx4_make_profile(dev, &profile, &dev_cap, &init_hca);
- if ((long long) icm_size < 0) {
- err = icm_size;
- goto err_stop_fw;
- }
+ init_hca.log_uar_sz = ilog2(dev->caps.num_uars);
- init_hca.log_uar_sz = ilog2(dev->caps.num_uars);
+ err = mlx4_init_icm(dev, &dev_cap, &init_hca, icm_size);
+ if (err)
+ goto err_stop_fw;
- err = mlx4_init_icm(dev, &dev_cap, &init_hca, icm_size);
- if (err)
- goto err_stop_fw;
+ err = mlx4_INIT_HCA(dev, &init_hca);
+ if (err) {
+ mlx4_err(dev, "INIT_HCA command failed, aborting.\n");
+ goto err_free_icm;
+ }
+ } else {
+ err = mlx4_init_slave(dev);
+ if (err) {
+ mlx4_err(dev, "Failed to initialize slave\n");
+ return err;
+ }
- err = mlx4_INIT_HCA(dev, &init_hca);
- if (err) {
- mlx4_err(dev, "INIT_HCA command failed, aborting.\n");
- goto err_free_icm;
+ err = mlx4_slave_cap(dev);
+ if (err) {
+ mlx4_err(dev, "Failed to obtain slave caps\n");
+ goto err_close;
+ }
}
err = mlx4_QUERY_ADAPTER(dev, &adapter);
@@ -857,15 +929,17 @@ static int mlx4_init_hca(struct mlx4_dev *dev)
return 0;
err_close:
- mlx4_CLOSE_HCA(dev, 0);
+ mlx4_close_hca(dev);
err_free_icm:
- mlx4_free_icms(dev);
+ if (!mlx4_is_slave(dev))
+ mlx4_free_icms(dev);
err_stop_fw:
- mlx4_UNMAP_FA(dev);
- mlx4_free_icm(dev, priv->fw.fw_icm, 0);
-
+ if (!mlx4_is_slave(dev)) {
+ mlx4_UNMAP_FA(dev);
+ mlx4_free_icm(dev, priv->fw.fw_icm, 0);
+ }
return err;
}
@@ -1019,8 +1093,13 @@ static void mlx4_enable_msi_x(struct mlx4_dev *dev)
int i;
if (msi_x) {
- nreq = min_t(int, dev->caps.num_eqs - dev->caps.reserved_eqs,
- num_possible_cpus() + 1);
+ /* The master only uses en event EQ,
+ * Each one of the slaves have 1 completion eq */
+ if (mlx4_is_mfunc(dev))
+ nreq = 1;
+ else
+ nreq = min_t(int, dev->caps.num_eqs - dev->caps.reserved_eqs,
+ num_possible_cpus() + 1);
entries = kcalloc(nreq, sizeof *entries, GFP_KERNEL);
if (!entries)
goto no_msi;
@@ -1029,8 +1108,10 @@ static void mlx4_enable_msi_x(struct mlx4_dev *dev)
entries[i].entry = i;
retry:
+ printk("Requesting %d MSIX vectors\n", nreq);
err = pci_enable_msix(dev->pdev, entries, nreq);
if (err) {
+ printk("pci_enable_msix failed\n");
/* Try again if at least 2 vectors are available */
if (err > 1) {
mlx4_info(dev, "Requested %d vectors, "
@@ -1043,7 +1124,7 @@ static void mlx4_enable_msi_x(struct mlx4_dev *dev)
goto no_msi;
}
- dev->caps.num_comp_vectors = nreq - 1;
+ dev->caps.num_comp_vectors = nreq - !mlx4_is_slave(dev);
for (i = 0; i < nreq; ++i)
priv->eq_table.eq[i].irq = entries[i].vector;
@@ -1054,7 +1135,9 @@ static void mlx4_enable_msi_x(struct mlx4_dev *dev)
}
no_msi:
- dev->caps.num_comp_vectors = 1;
+ /* CX1: no completion vectors needed for the master since it doesn't
+ * have interfaces */
+ dev->caps.num_comp_vectors = mlx4_is_master(dev) ? 0 : 1;
for (i = 0; i < 2; ++i)
priv->eq_table.eq[i].irq = dev->pdev->irq;
@@ -1111,10 +1194,9 @@ static int __mlx4_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
}
/*
- * Check for BARs. We expect 0: 1MB
+ * Check for BARs.
*/
- if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM) ||
- pci_resource_len(pdev, 0) != 1 << 20) {
+ if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
dev_err(&pdev->dev, "Missing DCS, aborting.\n");
err = -ENODEV;
goto err_disable_pdev;
@@ -1172,34 +1254,98 @@ static int __mlx4_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
INIT_LIST_HEAD(&priv->pgdir_list);
mutex_init(&priv->pgdir_mutex);
- /*
- * Now reset the HCA before we touch the PCI capabilities or
- * attempt a firmware command, since a boot ROM may have left
- * the HCA in an undefined state.
- */
- err = mlx4_reset(dev);
- if (err) {
- mlx4_err(dev, "Failed to reset HCA, aborting.\n");
- goto err_free_dev;
+ /* Detect if this device is a virtual function */
+ switch (id->device) {
+ case 0x6341:
+ case 0x634b:
+ case 0x6733:
+ case 0x673d:
+ case 0x6369:
+ case 0x6751:
+ case 0x6765:
+#ifdef CONFIG_PCI_IOV
+ /* When acting as pf, we normally skip vfs unless explicitly
+ * requested to probe them.
+ * TODO: add ARI support */
+ if (sr_iov && PCI_FUNC(pdev->devfn) > probe_vf) {
+ mlx4_warn(dev, "Skipping virtual function:%d\n",
+ PCI_FUNC(pdev->devfn));
+ err = -ENODEV;
+ goto err_free_dev;
+ }
+#endif /* CONFIG_PCI_IOV */
+ mlx4_warn(dev, "Detected virtual function - running in slave mode\n");
+ dev->flags |= MLX4_FLAG_SLAVE;
+ break;
+ default:
+ ;
+ }
+
+ /* We reset the device and enable SRIOV only for physical devices */
+ if (!mlx4_is_slave(dev)) {
+ /*
+ * Now reset the HCA before we touch the PCI capabilities or
+ * attempt a firmware command, since a boot ROM may have left
+ * the HCA in an undefined state.
+ */
+ err = mlx4_reset(dev);
+ if (err) {
+ mlx4_err(dev, "Failed to reset HCA, aborting.\n");
+ goto err_free_dev;
+ }
+#ifdef CONFIG_PCI_IOV
+ if (sr_iov) {
+ mlx4_warn(dev, "Enabling sriov with:%d vfs\n", sr_iov);
+ if (pci_enable_sriov(pdev, sr_iov)) {
+ mlx4_err(dev, "Failed to enable sriov, aborting.\n");
+ goto err_free_dev;
+ }
+ mlx4_warn(dev, "Running in master mode\n");
+ dev->flags |= MLX4_FLAG_SRIOV | MLX4_FLAG_MASTER;
+ dev->num_slaves = sr_iov;
+ }
+#endif /* CONFIG_PCI_IOV */
}
if (mlx4_cmd_init(dev)) {
mlx4_err(dev, "Failed to init command interface, aborting.\n");
- goto err_free_dev;
+ goto err_sriov;
+ }
+
+ /* In slave functions, the communication channel must be initialized before
+ * posting commands */
+ if (mlx4_is_slave(dev)) {
+ if (mlx4_multi_func_init(dev)) {
+ mlx4_err(dev, "Failed to init slave mfunc interface, aborting.\n");
+ goto err_cmd;
+ }
}
err = mlx4_init_hca(dev);
if (err)
goto err_cmd;
+ /* In master functions, the communication channel must be initialized after obtaining
+ * its address from fw */
+ if (mlx4_is_master(dev)) {
+ if (mlx4_multi_func_init(dev)) {
+ mlx4_err(dev, "Failed to init master mfunc interface, aborting.\n");
+ goto err_close;
+ }
+ }
+
err = mlx4_alloc_eq_table(dev);
if (err)
goto err_close;
mlx4_enable_msi_x(dev);
+ if (mlx4_is_slave(dev) && !(dev->flags & MLX4_FLAG_MSI_X)) {
+ mlx4_err(dev, "INTx is not supported in slave mode, aborting.\n");
+ goto err_free_eq;
+ }
err = mlx4_setup_hca(dev);
- if (err == -EBUSY && (dev->flags & MLX4_FLAG_MSI_X)) {
+ if (err == -EBUSY && (dev->flags & MLX4_FLAG_MSI_X) && !mlx4_is_slave(dev)) {
dev->flags &= ~MLX4_FLAG_MSI_X;
pci_disable_msix(pdev);
err = mlx4_setup_hca(dev);
@@ -1223,6 +1369,12 @@ static int __mlx4_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
pci_set_drvdata(pdev, dev);
+ /* Start serving comm channel:
+ * - In master function: poll for commands
+ * - in slave functions: poll for events
+ * TODO - enable comm channel interrupts */
+ if (mlx4_is_master(dev) || mlx4_is_slave(dev))
+ queue_delayed_work(priv->mfunc.comm_wq, &priv->mfunc.comm_work, 0);
return 0;
err_port:
@@ -1233,7 +1385,8 @@ err_port:
mlx4_cleanup_qp_table(dev);
mlx4_cleanup_srq_table(dev);
mlx4_cleanup_cq_table(dev);
- mlx4_cmd_use_polling(dev);
+ if (!mlx4_is_master(dev) && !mlx4_is_slave(dev))
+ mlx4_cmd_use_polling(dev);
mlx4_cleanup_eq_table(dev);
mlx4_cleanup_mr_table(dev);
mlx4_cleanup_pd_table(dev);
@@ -1251,6 +1404,14 @@ err_close:
err_cmd:
mlx4_cmd_cleanup(dev);
+err_sriov:
+ if (mlx4_is_master(dev) || mlx4_is_slave(dev))
+ mlx4_multi_func_cleanup(dev);
+#ifdef CONFIG_PCI_IOV
+ if (dev->flags & MLX4_FLAG_SRIOV)
+ pci_disable_sriov(pdev);
+#endif /* CONFIG_PCI_IOV */
+
err_free_dev:
kfree(priv);
@@ -1283,19 +1444,24 @@ static void mlx4_remove_one(struct pci_dev *pdev)
int p;
if (dev) {
+ /* Stop serving commands and events over comm channel */
+ if (mlx4_is_master(dev) || mlx4_is_slave(dev))
+ cancel_delayed_work_sync(&priv->mfunc.comm_work);
mlx4_stop_sense(dev);
mlx4_unregister_device(dev);
for (p = 1; p <= dev->caps.num_ports; p++) {
mlx4_cleanup_port_info(&priv->port[p]);
- mlx4_CLOSE_PORT(dev, p);
+ if (!mlx4_is_slave(dev))
+ mlx4_CLOSE_PORT(dev, p);
}
mlx4_cleanup_mcg_table(dev);
mlx4_cleanup_qp_table(dev);
mlx4_cleanup_srq_table(dev);
mlx4_cleanup_cq_table(dev);
- mlx4_cmd_use_polling(dev);
+ if (!mlx4_is_master(dev) && !mlx4_is_slave(dev))
+ mlx4_cmd_use_polling(dev);
mlx4_cleanup_eq_table(dev);
mlx4_cleanup_mr_table(dev);
mlx4_cleanup_pd_table(dev);
@@ -1303,10 +1469,18 @@ static void mlx4_remove_one(struct pci_dev *pdev)
mlx4_cleanup_uar_table(dev);
mlx4_free_eq_table(dev);
mlx4_close_hca(dev);
+ if (mlx4_is_master(dev) || mlx4_is_slave(dev))
+ mlx4_multi_func_cleanup(dev);
mlx4_cmd_cleanup(dev);
if (dev->flags & MLX4_FLAG_MSI_X)
pci_disable_msix(pdev);
+#ifdef CONFIG_PCI_IOV
+ if (dev->flags & MLX4_FLAG_SRIOV) {
+ mlx4_warn(dev, "Disabling sriov\n");
+ pci_disable_sriov(pdev);
+ }
+#endif /* CONFIG_PCI_IOV */
kfree(priv);
pci_release_regions(pdev);
@@ -1321,14 +1495,21 @@ int mlx4_restart_one(struct pci_dev *pdev)
return __mlx4_init_one(pdev, NULL);
}
+
static struct pci_device_id mlx4_pci_table[] = {
{ PCI_VDEVICE(MELLANOX, 0x6340) }, /* MT25408 "Hermon" SDR */
+ { PCI_VDEVICE(MELLANOX, 0x6341) }, /* MT25408 "Hermon" SDR VF */
{ PCI_VDEVICE(MELLANOX, 0x634a) }, /* MT25408 "Hermon" DDR */
+ { PCI_VDEVICE(MELLANOX, 0x634b) }, /* MT25408 "Hermon" DDR VF */
{ PCI_VDEVICE(MELLANOX, 0x6354) }, /* MT25408 "Hermon" QDR */
{ PCI_VDEVICE(MELLANOX, 0x6732) }, /* MT25408 "Hermon" DDR PCIe gen2 */
+ { PCI_VDEVICE(MELLANOX, 0x6733) }, /* MT25408 "Hermon" DDR PCIe gen2 VF */
{ PCI_VDEVICE(MELLANOX, 0x673c) }, /* MT25408 "Hermon" QDR PCIe gen2 */
+ { PCI_VDEVICE(MELLANOX, 0x673d) }, /* MT25408 "Hermon" QDR PCIe gen2 VF */
{ PCI_VDEVICE(MELLANOX, 0x6368) }, /* MT25408 "Hermon" EN 10GigE */
+ { PCI_VDEVICE(MELLANOX, 0x6369) }, /* MT25408 "Hermon" EN 10GigE VF */
{ PCI_VDEVICE(MELLANOX, 0x6750) }, /* MT25408 "Hermon" EN 10GigE PCIe gen2 */
+ { PCI_VDEVICE(MELLANOX, 0x6751) }, /* MT25408 "Hermon" EN 10GigE PCIe gen2 VF */
{ PCI_VDEVICE(MELLANOX, 0x6372) }, /* MT25458 ConnectX EN 10GBASE-T 10GigE */
{ PCI_VDEVICE(MELLANOX, 0x675a) }, /* MT25458 ConnectX EN 10GBASE-T+Gen2 10GigE */
{ PCI_VDEVICE(MELLANOX, 0x6764) }, /* MT26468 ConnectX EN 10GigE PCIe gen2*/
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 15/25] mlx4_core: multi-function interface setup
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Master verifies hw healthiness.
Interface drivers are only supported on slaves.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/intf.c | 21 ++++++++++++++-------
1 files changed, 14 insertions(+), 7 deletions(-)
diff --git a/drivers/net/mlx4/intf.c b/drivers/net/mlx4/intf.c
index 0e7eb10..f5f4523 100644
--- a/drivers/net/mlx4/intf.c
+++ b/drivers/net/mlx4/intf.c
@@ -88,8 +88,11 @@ int mlx4_register_interface(struct mlx4_interface *intf)
mutex_lock(&intf_mutex);
list_add_tail(&intf->list, &intf_list);
- list_for_each_entry(priv, &dev_list, dev_list)
- mlx4_add_device(intf, priv);
+ list_for_each_entry(priv, &dev_list, dev_list) {
+ /* CX1: master cannot run interfaces */
+ if (!mlx4_is_master(&priv->dev))
+ mlx4_add_device(intf, priv);
+ }
mutex_unlock(&intf_mutex);
@@ -135,11 +138,14 @@ int mlx4_register_device(struct mlx4_dev *dev)
mutex_lock(&intf_mutex);
list_add_tail(&priv->dev_list, &dev_list);
- list_for_each_entry(intf, &intf_list, list)
- mlx4_add_device(intf, priv);
-
+ /* CX1: master cannot run interfaces */
+ if (!mlx4_is_master(dev)) {
+ list_for_each_entry(intf, &intf_list, list)
+ mlx4_add_device(intf, priv);
+ }
mutex_unlock(&intf_mutex);
- mlx4_start_catas_poll(dev);
+ if (!mlx4_is_slave(dev))
+ mlx4_start_catas_poll(dev);
return 0;
}
@@ -149,7 +155,8 @@ void mlx4_unregister_device(struct mlx4_dev *dev)
struct mlx4_priv *priv = mlx4_priv(dev);
struct mlx4_interface *intf;
- mlx4_stop_catas_poll(dev);
+ if (!mlx4_is_slave(dev))
+ mlx4_stop_catas_poll(dev);
mutex_lock(&intf_mutex);
list_for_each_entry(intf, &intf_list, list)
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 13/25] mlx4_core: remove unused driver/kernel uar (kar)
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/main.c | 25 +------------------------
drivers/net/mlx4/mlx4.h | 2 --
2 files changed, 1 insertions(+), 26 deletions(-)
diff --git a/drivers/net/mlx4/main.c b/drivers/net/mlx4/main.c
index 4edc8d7..d8143c6 100644
--- a/drivers/net/mlx4/main.c
+++ b/drivers/net/mlx4/main.c
@@ -883,26 +883,11 @@ static int mlx4_setup_hca(struct mlx4_dev *dev)
return err;
}
- err = mlx4_uar_alloc(dev, &priv->driver_uar);
- if (err) {
- mlx4_err(dev, "Failed to allocate driver access region, "
- "aborting.\n");
- goto err_uar_table_free;
- }
-
- priv->kar = ioremap(priv->driver_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
- if (!priv->kar) {
- mlx4_err(dev, "Couldn't map kernel access region, "
- "aborting.\n");
- err = -ENOMEM;
- goto err_uar_free;
- }
-
err = mlx4_init_pd_table(dev);
if (err) {
mlx4_err(dev, "Failed to initialize "
"protection domain table, aborting.\n");
- goto err_kar_unmap;
+ goto err_uar_table_free;
}
err = mlx4_init_mr_table(dev);
@@ -1015,12 +1000,6 @@ err_mr_table_free:
err_pd_table_free:
mlx4_cleanup_pd_table(dev);
-err_kar_unmap:
- iounmap(priv->kar);
-
-err_uar_free:
- mlx4_uar_free(dev, &priv->driver_uar);
-
err_uar_table_free:
mlx4_cleanup_uar_table(dev);
return err;
@@ -1316,8 +1295,6 @@ static void mlx4_remove_one(struct pci_dev *pdev)
mlx4_cleanup_mr_table(dev);
mlx4_cleanup_pd_table(dev);
- iounmap(priv->kar);
- mlx4_uar_free(dev, &priv->driver_uar);
mlx4_cleanup_uar_table(dev);
mlx4_free_eq_table(dev);
mlx4_close_hca(dev);
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index b8b30ca..516e375 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -401,8 +401,6 @@ struct mlx4_priv {
void __iomem *clr_base;
- struct mlx4_uar driver_uar;
- void __iomem *kar;
struct mlx4_port_info port[MLX4_MAX_PORTS + 1];
struct mlx4_sense sense;
struct mutex port_mutex;
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 12/25] mlx4_core: associate resources with specific functions
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Resources are associated with functions by means of pds - the upper 7 msbs
specify the function id. Since cqs do not refer to a specific pd, the function
id is passed explicitly to hw in the cq context.
For uars, which are statically partitioned between slaves, we only need to adjust
their offsets with respect to the associated function bar.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/cq.c | 4 +++-
drivers/net/mlx4/eq.c | 13 ++++++++++++-
drivers/net/mlx4/pd.c | 12 +++++++++---
3 files changed, 24 insertions(+), 5 deletions(-)
diff --git a/drivers/net/mlx4/cq.c b/drivers/net/mlx4/cq.c
index 3fb9f7f..9f9f246 100644
--- a/drivers/net/mlx4/cq.c
+++ b/drivers/net/mlx4/cq.c
@@ -59,7 +59,8 @@ struct mlx4_cq_context {
__be32 solicit_producer_index;
__be32 consumer_index;
__be32 producer_index;
- u32 reserved4[2];
+ __be32 pd; /* for sriov guest id */
+ u32 reserved4;
__be64 db_rec_addr;
};
@@ -293,6 +294,7 @@ int mlx4_cq_alloc(struct mlx4_dev *dev, int nent, struct mlx4_mtt *mtt,
mtt_addr = mlx4_mtt_addr(dev, mtt);
cq_context->mtt_base_addr_h = mtt_addr >> 32;
cq_context->mtt_base_addr_l = cpu_to_be32(mtt_addr & 0xffffffff);
+ cq_context->pd = cpu_to_be32(dev->caps.pd_base << dev->caps.slave_pd_shift);
cq_context->db_rec_addr = cpu_to_be64(db_rec);
err = mlx4_SW2HW_CQ(dev, mailbox, cq->cqn);
diff --git a/drivers/net/mlx4/eq.c b/drivers/net/mlx4/eq.c
index a27e1c4..46064cc 100644
--- a/drivers/net/mlx4/eq.c
+++ b/drivers/net/mlx4/eq.c
@@ -486,9 +486,20 @@ static void __iomem *mlx4_get_eq_uar(struct mlx4_dev *dev, struct mlx4_eq *eq)
{
struct mlx4_priv *priv = mlx4_priv(dev);
int index;
+ int offset;
- index = eq->eqn / 4 - dev->caps.reserved_eqs / 4;
+ /* CX1: slave EQ DBs are located in the comm channel page */
+ if (mlx4_is_slave(dev) || mlx4_is_master(dev)) {
+ if (eq->eqn - dev->caps.reserved_eqs >= MLX4_MFUNC_EQ_NUM) {
+ mlx4_err(dev, "eqn:%d doorbell out of range (reserved:%d)\n",
+ eq->eqn, dev->caps.reserved_eqs);
+ return NULL;
+ }
+ offset = 0x800 + (eq->eqn - dev->caps.reserved_eqs) * 8;
+ return ((void *) priv->mfunc.comm) + offset;
+ }
+ index = eq->eqn / 4 - dev->caps.reserved_eqs / 4;
if (!priv->eq_table.uar_map[index]) {
priv->eq_table.uar_map[index] =
ioremap(pci_resource_start(dev->pdev, 2) +
diff --git a/drivers/net/mlx4/pd.c b/drivers/net/mlx4/pd.c
index c4988d6..08eaf08 100644
--- a/drivers/net/mlx4/pd.c
+++ b/drivers/net/mlx4/pd.c
@@ -45,7 +45,7 @@ int mlx4_pd_alloc(struct mlx4_dev *dev, u32 *pdn)
*pdn = mlx4_bitmap_alloc(&priv->pd_bitmap);
if (*pdn == -1)
return -ENOMEM;
-
+ *pdn |= dev->caps.pd_base << dev->caps.slave_pd_shift;
return 0;
}
EXPORT_SYMBOL_GPL(mlx4_pd_alloc);
@@ -72,12 +72,18 @@ void mlx4_cleanup_pd_table(struct mlx4_dev *dev)
int mlx4_uar_alloc(struct mlx4_dev *dev, struct mlx4_uar *uar)
{
+ int offset;
+
uar->index = mlx4_bitmap_alloc(&mlx4_priv(dev)->uar_table.bitmap);
if (uar->index == -1)
return -ENOMEM;
- uar->pfn = (pci_resource_start(dev->pdev, 2) >> PAGE_SHIFT) + uar->index;
-
+ if (mlx4_is_slave(dev))
+ offset = uar->index % ((int) pci_resource_len(dev->pdev, 2) /
+ dev->caps.uar_page_size);
+ else
+ offset = uar->index;
+ uar->pfn = (pci_resource_start(dev->pdev, 2) >> PAGE_SHIFT) + offset;
return 0;
}
EXPORT_SYMBOL_GPL(mlx4_uar_alloc);
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 11/25] mlx4_core: per-function capabilities
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
The master function builds a HW profile, and manages all resources.
Other functions query the master for function-specific capabilities.
EQs, MSI-X vectors, and UARs are statically divided among all functions,
while other resources are dynamically assigned later upon request.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/cmd.c | 2 +
drivers/net/mlx4/fw.c | 78 +++++++++++++++++++++++++++++++++++++++++-
drivers/net/mlx4/fw.h | 4 ++
drivers/net/mlx4/main.c | 70 ++++++++++++++++++++++++++++++++++++++-
drivers/net/mlx4/mlx4.h | 4 ++-
drivers/net/mlx4/profile.c | 26 ++++++++++++--
include/linux/mlx4/cmd.h | 1 +
include/linux/mlx4/device.h | 3 ++
8 files changed, 180 insertions(+), 8 deletions(-)
diff --git a/drivers/net/mlx4/cmd.c b/drivers/net/mlx4/cmd.c
index 29074a0..7afbfef 100644
--- a/drivers/net/mlx4/cmd.c
+++ b/drivers/net/mlx4/cmd.c
@@ -39,6 +39,7 @@
#include <asm/io.h>
#include "mlx4.h"
+#include "fw.h"
#include "en_port.h"
#define CMD_POLL_TOKEN 0xffff
@@ -524,6 +525,7 @@ static struct mlx4_cmd_info {
struct mlx4_cmd_mailbox *outbox);
} cmd_info[] = {
{MLX4_CMD_QUERY_FW, 0, 1, 0, NULL, NULL},
+ {MLX4_CMD_QUERY_SLAVE_CAP, 0, 1, 0, NULL, mlx4_QUERY_SLAVE_CAP_wrapper},
{MLX4_CMD_QUERY_ADAPTER, 0, 1, 0, NULL, NULL},
{MLX4_CMD_GET_SLAVE_SQP, 0, 1, 0, NULL, mlx4_GET_SLAVE_SQP_wrapper},
diff --git a/drivers/net/mlx4/fw.c b/drivers/net/mlx4/fw.c
index 87eb2d4..60889d3 100644
--- a/drivers/net/mlx4/fw.c
+++ b/drivers/net/mlx4/fw.c
@@ -143,6 +143,76 @@ int mlx4_QUERY_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *v
MLX4_CMD_TIME_CLASS_B);
}
+int mlx4_QUERY_SLAVE_CAP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ struct mlx4_caps *caps = outbox->buf;
+ int i;
+
+ memcpy(caps, &dev->caps, sizeof *caps);
+
+ /* CX1: blue-flame not supported in vfs */
+ caps->bf_reg_size = 0;
+ caps->bf_regs_per_page = 0;
+
+ /* CX1:
+ * On ingress, slave0 gets full ownership over qp1, and demultiplexes GSI mads
+ * on behalf of other slaves.
+ * On egress, slave>0 must tunnel their GSI mads for validation by the qp1 owner */
+ if (slave) {
+ caps->sqp_demux = 0;
+ for (i = 1; i <= dev->caps.num_ports; ++i) {
+ caps->gid_table_len[i] = 1;
+ caps->pkey_table_len[i] = 1;
+ }
+ } else {
+ caps->sqp_demux = dev->num_slaves;
+ for (i = 1; i <= dev->caps.num_ports; ++i) {
+ caps->gid_table_len[i] = dev->num_slaves;
+ caps->pkey_table_len[i] = 1;
+ }
+ }
+
+ /* Slave functions allocate themselves EQs, UARs, and PDs.
+ * - num is the maximum resource index
+ * - reserved is the minimum resource index */
+ caps->num_eqs = dev->caps.reserved_eqs + MLX4_MFUNC_EQ_NUM * (slave + 2);
+ caps->reserved_eqs = dev->caps.reserved_eqs + MLX4_MFUNC_EQ_NUM * (slave + 1);
+ caps->num_uars = dev->caps.num_uars * (slave + 1);
+ caps->reserved_uars = max_t(int, dev->caps.num_uars * slave, dev->caps.reserved_uars);
+
+ /* PDs have the same range in every guest; the distinction is in the msbs,
+ * which contains the guest ID (vf + 1) */
+ caps->pd_base = slave + 1;
+
+ /* All other resources are allocated by the master, but we still report
+ * 'num' and 'reserved' capabilities as follows:
+ * - num remains the maximum resource index
+ * - 'num - reserved' is the total available objects of a resource, but
+ * resource indices may be less than 'reserved'
+ * TODO: set per-resource quotas */
+ return 0;
+}
+
+int mlx4_QUERY_SLAVE_CAP(struct mlx4_dev *dev, struct mlx4_caps *caps)
+{
+ struct mlx4_cmd_mailbox *mailbox;
+ int err;
+
+ mailbox = mlx4_alloc_cmd_mailbox(dev);
+ if (IS_ERR(mailbox))
+ return PTR_ERR(mailbox);
+
+ err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0, MLX4_CMD_QUERY_SLAVE_CAP,
+ MLX4_CMD_TIME_CLASS_A);
+ if (!err)
+ memcpy(caps, mailbox->buf, sizeof *caps);
+
+ mlx4_free_cmd_mailbox(dev, mailbox);
+ return err;
+}
+
int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
{
struct mlx4_cmd_mailbox *mailbox;
@@ -405,9 +475,13 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
* Each UAR has 4 EQ doorbells; so if a UAR is reserved, then
* we can't use any EQs whose doorbell falls on that page,
* even if the EQ itself isn't reserved.
+ * CX1: in multi-function mode, EQ doorbells are virtualized so the whole EQ
+ * range is available: the master and each of the slaves get 4 EQ doorbells
+ * starting from the first non-reserved EQ.
*/
- dev_cap->reserved_eqs = max(dev_cap->reserved_uars * 4,
- dev_cap->reserved_eqs);
+ if (!mlx4_is_mfunc(dev))
+ dev_cap->reserved_eqs = max(dev_cap->reserved_uars * 4,
+ dev_cap->reserved_eqs);
mlx4_dbg(dev, "Max ICM size %lld MB\n",
(unsigned long long) dev_cap->max_icm_sz >> 20);
diff --git a/drivers/net/mlx4/fw.h b/drivers/net/mlx4/fw.h
index 526d7f3..d066c69 100644
--- a/drivers/net/mlx4/fw.h
+++ b/drivers/net/mlx4/fw.h
@@ -160,6 +160,10 @@ struct mlx4_set_ib_param {
};
int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap);
+int mlx4_QUERY_SLAVE_CAP(struct mlx4_dev *dev, struct mlx4_caps *caps);
+int mlx4_QUERY_SLAVE_CAP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
int mlx4_MAP_FA(struct mlx4_dev *dev, struct mlx4_icm *icm);
int mlx4_UNMAP_FA(struct mlx4_dev *dev);
int mlx4_RUN_FW(struct mlx4_dev *dev);
diff --git a/drivers/net/mlx4/main.c b/drivers/net/mlx4/main.c
index 291a505..4edc8d7 100644
--- a/drivers/net/mlx4/main.c
+++ b/drivers/net/mlx4/main.c
@@ -141,6 +141,7 @@ static void mlx4_set_port_mask(struct mlx4_dev *dev)
if (dev->caps.port_type[i] == MLX4_PORT_TYPE_IB)
dev->caps.port_mask |= 1 << (i - 1);
}
+
static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
{
int err;
@@ -185,6 +186,7 @@ static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
dev->caps.supported_type[i] = dev_cap->supported_port_types[i];
}
+ dev->caps.uar_page_size = PAGE_SIZE;
dev->caps.num_uars = dev_cap->uar_size / PAGE_SIZE;
dev->caps.local_ca_ack_delay = dev_cap->local_ca_ack_delay;
dev->caps.bf_reg_size = dev_cap->bf_reg_size;
@@ -211,7 +213,9 @@ static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
dev->caps.reserved_mtts = DIV_ROUND_UP(dev_cap->reserved_mtts,
dev->caps.mtts_per_seg);
dev->caps.reserved_mrws = dev_cap->reserved_mrws;
- dev->caps.reserved_uars = dev_cap->reserved_uars;
+
+ /* The first 128 UARs are used for EQ doorbells */
+ dev->caps.reserved_uars = max_t(int, 128, dev_cap->reserved_uars);
dev->caps.reserved_pds = dev_cap->reserved_pds;
dev->caps.mtt_entry_sz = dev->caps.mtts_per_seg * dev_cap->mtt_entry_sz;
dev->caps.max_msg_sz = dev_cap->max_msg_sz;
@@ -265,6 +269,70 @@ static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FC_ADDR] +
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FC_EXCH];
+ /* CX1: vf0 takes role of GSI multiplexing instead of master */
+ dev->caps.sqp_demux = 0;
+ return 0;
+}
+
+static int mlx4_slave_cap(struct mlx4_dev *dev)
+{
+ int err;
+ u32 page_size;
+
+ err = mlx4_QUERY_SLAVE_CAP(dev, &dev->caps);
+ if (err)
+ return err;
+
+ page_size = ~dev->caps.page_size_cap + 1;
+ mlx4_warn(dev, "HCA minimum page size:%d\n", page_size);
+ if (page_size > PAGE_SIZE) {
+ mlx4_err(dev, "HCA minimum page size of %d bigger than "
+ "kernel PAGE_SIZE of %ld, aborting.\n",
+ page_size, PAGE_SIZE);
+ return -ENODEV;
+ }
+
+ /* TODO: relax this assumption */
+ if (dev->caps.uar_page_size != PAGE_SIZE) {
+ mlx4_err(dev, "UAR size:%d != kernel PAGE_SIZE of %ld\n",
+ dev->caps.uar_page_size, PAGE_SIZE);
+ return -ENODEV;
+ }
+
+ if (dev->caps.num_ports > MLX4_MAX_PORTS) {
+ mlx4_err(dev, "HCA has %d ports, but we only support %d, "
+ "aborting.\n", dev->caps.num_ports, MLX4_MAX_PORTS);
+ return -ENODEV;
+ }
+
+ if (dev->caps.uar_page_size * (dev->caps.num_uars -
+ dev->caps.reserved_uars) >
+ pci_resource_len(dev->pdev, 2)) {
+ mlx4_err(dev, "HCA reported UAR region size of 0x%x bigger than "
+ "PCI resource 2 size of 0x%llx, aborting.\n",
+ dev->caps.uar_page_size * dev->caps.num_uars,
+ (unsigned long long) pci_resource_len(dev->pdev, 2));
+ return -ENODEV;
+ }
+
+ /* Adjust eq number */
+ if (dev->caps.num_eqs - dev->caps.reserved_eqs > num_possible_cpus() + 1)
+ dev->caps.num_eqs = dev->caps.reserved_eqs + num_possible_cpus() + 1;
+
+#if 0
+ mlx4_warn(dev, "sqp_demux:%d\n", dev->caps.sqp_demux);
+ mlx4_warn(dev, "num_uars:%d reserved_uars:%d uar region:0x%x bar2:0x%llx\n",
+ dev->caps.num_uars, dev->caps.reserved_uars,
+ dev->caps.uar_page_size * dev->caps.num_uars,
+ pci_resource_len(dev->pdev, 2));
+ mlx4_warn(dev, "num_eqs:%d reserved_eqs:%d\n", dev->caps.num_eqs,
+ dev->caps.reserved_eqs);
+ mlx4_warn(dev, "num_pds:%d reserved_pds:%d slave_pd_shift:%d pd_base:%d\n",
+ dev->caps.num_pds,
+ dev->caps.reserved_pds,
+ dev->caps.slave_pd_shift,
+ dev->caps.pd_base);
+#endif
return 0;
}
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index b01820f..b8b30ca 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -69,7 +69,8 @@ enum {
};
enum {
- MLX4_NUM_PDS = 1 << 15
+ MLX4_NUM_PDS = 1 << 15,
+ MLX4_SLAVE_PD_SHIFT = 17, /* the 7 msbs encode the slave id */
};
enum {
@@ -112,6 +113,7 @@ enum mlx4_alloc_mode {
enum {
MLX4_MFUNC_MAX = 64,
+ MLX4_MFUNC_EQ_NUM = 4,
MLX4_MFUNC_MAX_EQES = 8,
MLX4_MFUNC_EQE_MASK = (MLX4_MFUNC_MAX_EQES - 1)
};
diff --git a/drivers/net/mlx4/profile.c b/drivers/net/mlx4/profile.c
index ca25b9d..f6197e9 100644
--- a/drivers/net/mlx4/profile.c
+++ b/drivers/net/mlx4/profile.c
@@ -105,9 +105,19 @@ u64 mlx4_make_profile(struct mlx4_dev *dev,
profile[MLX4_RES_AUXC].num = request->num_qp;
profile[MLX4_RES_SRQ].num = request->num_srq;
profile[MLX4_RES_CQ].num = request->num_cq;
- profile[MLX4_RES_EQ].num = min_t(unsigned, dev_cap->max_eqs,
- dev_cap->reserved_eqs +
- num_possible_cpus() + 1);
+ if (mlx4_is_master(dev)) {
+ profile[MLX4_RES_EQ].num = dev_cap->reserved_eqs +
+ MLX4_MFUNC_EQ_NUM *
+ (dev->num_slaves + 1);
+ if (profile[MLX4_RES_EQ].num > dev_cap->max_eqs) {
+ mlx4_warn(dev, "Not enough eqs for:%ld slave functions\n", dev->num_slaves);
+ kfree(profile);
+ return -ENOMEM;
+ }
+ } else
+ profile[MLX4_RES_EQ].num = min_t(unsigned, dev_cap->max_eqs,
+ dev_cap->reserved_eqs +
+ num_possible_cpus() + 1);
profile[MLX4_RES_DMPT].num = request->num_mpt;
profile[MLX4_RES_CMPT].num = MLX4_NUM_CMPTS;
profile[MLX4_RES_MTT].num = request->num_mtt;
@@ -196,7 +206,13 @@ u64 mlx4_make_profile(struct mlx4_dev *dev,
init_hca->log_num_cqs = profile[i].log_num;
break;
case MLX4_RES_EQ:
- dev->caps.num_eqs = profile[i].num;
+ if (mlx4_is_master(dev)) {
+ dev->caps.num_eqs = dev_cap->reserved_eqs +
+ min_t(unsigned,
+ MLX4_MFUNC_EQ_NUM,
+ num_possible_cpus() + 1);
+ } else
+ dev->caps.num_eqs = profile[i].num;
init_hca->eqc_base = profile[i].start;
init_hca->log_num_eqs = profile[i].log_num;
break;
@@ -232,6 +248,8 @@ u64 mlx4_make_profile(struct mlx4_dev *dev,
* of the HCA profile anyway.
*/
dev->caps.num_pds = MLX4_NUM_PDS;
+ dev->caps.slave_pd_shift = MLX4_SLAVE_PD_SHIFT;
+ dev->caps.pd_base = 0;
kfree(profile);
return total_size;
diff --git a/include/linux/mlx4/cmd.h b/include/linux/mlx4/cmd.h
index f1659f2..f265091 100644
--- a/include/linux/mlx4/cmd.h
+++ b/include/linux/mlx4/cmd.h
@@ -123,6 +123,7 @@ enum {
MLX4_CMD_ALLOC_RES = 0x50,
MLX4_CMD_FREE_RES = 0x51,
MLX4_CMD_GET_EVENT = 0x52,
+ MLX4_CMD_QUERY_SLAVE_CAP = 0x53,
MLX4_CMD_MCAST_ATTACH = 0x54,
MLX4_CMD_GET_SLAVE_SQP = 0x55,
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index b20c8d8..4b9091a 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -192,6 +192,7 @@ struct mlx4_caps {
int pkey_table_len[MLX4_MAX_PORTS + 1];
int local_ca_ack_delay;
int num_uars;
+ int uar_page_size;
int bf_reg_size;
int bf_regs_per_page;
int max_sq_sg;
@@ -226,6 +227,8 @@ struct mlx4_caps {
int num_qp_per_mgm;
int num_pds;
int reserved_pds;
+ int slave_pd_shift;
+ int pd_base;
int mtt_entry_sz;
u32 max_msg_sz;
u32 page_size_cap;
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 10/25] mlx4_core: track slave special qps
From: Yevgeny Petrilin @ 2009-11-04 15:31 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
A single function can own the actual Infiniband hw special qps.
This function demuxes inbound mads and protects outbound mads from spoofing.
To achieve this, it needs to know the special qps used by other functions.
This patch tracks special qp registration and notifies the owning function
on any changes.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/cmd.c | 2 +
drivers/net/mlx4/eq.c | 15 +++++++++
drivers/net/mlx4/mlx4.h | 11 +++++++
drivers/net/mlx4/qp.c | 67 +++++++++++++++++++++++++++++++++++++++++++
include/linux/mlx4/cmd.h | 1 +
include/linux/mlx4/device.h | 3 ++
6 files changed, 99 insertions(+), 0 deletions(-)
diff --git a/drivers/net/mlx4/cmd.c b/drivers/net/mlx4/cmd.c
index 0a7c9c0..29074a0 100644
--- a/drivers/net/mlx4/cmd.c
+++ b/drivers/net/mlx4/cmd.c
@@ -525,6 +525,7 @@ static struct mlx4_cmd_info {
} cmd_info[] = {
{MLX4_CMD_QUERY_FW, 0, 1, 0, NULL, NULL},
{MLX4_CMD_QUERY_ADAPTER, 0, 1, 0, NULL, NULL},
+ {MLX4_CMD_GET_SLAVE_SQP, 0, 1, 0, NULL, mlx4_GET_SLAVE_SQP_wrapper},
{MLX4_CMD_INIT_PORT, 0, 0, 0, NULL, mlx4_INIT_PORT_wrapper},
{MLX4_CMD_CLOSE_PORT, 0, 0, 0, NULL, mlx4_CLOSE_PORT_wrapper},
@@ -568,6 +569,7 @@ static struct mlx4_cmd_info {
{MLX4_CMD_INIT2INIT_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
{MLX4_CMD_SUSPEND_QP, 0, 0, 0, NULL, NULL}, /* need verifier */
{MLX4_CMD_UNSUSPEND_QP, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_CONF_SPECIAL_QP, 0, 0, 0, NULL, mlx4_CONF_SPECIAL_QP_wrapper},
{MLX4_CMD_MAD_IFC, 1, 1, 0, NULL, NULL}, /* need verifier */
/* Native multicast commands are not available for guests */
diff --git a/drivers/net/mlx4/eq.c b/drivers/net/mlx4/eq.c
index 1e8b62d..a27e1c4 100644
--- a/drivers/net/mlx4/eq.c
+++ b/drivers/net/mlx4/eq.c
@@ -358,6 +358,17 @@ static int mlx4_eq_int(struct mlx4_dev *dev, struct mlx4_eq *eq)
return eqes_found;
}
+static void mlx4_update_sqp(struct mlx4_dev *dev)
+{
+ if (!dev->caps.sqp_demux) {
+ mlx4_warn(dev, "unexpected update_sqp event\n");
+ return;
+ }
+ if (mlx4_GET_SLAVE_SQP(dev, mlx4_priv(dev)->mfunc.demux_sqp,
+ dev->caps.sqp_demux))
+ mlx4_warn(dev, "couldn't update sqp\n");
+}
+
void mlx4_slave_async_eq_poll(struct work_struct *work)
{
struct delayed_work *delay = container_of(work, struct delayed_work, work);
@@ -402,6 +413,10 @@ void mlx4_slave_async_eq_poll(struct work_struct *work)
mlx4_warn(dev, "slave async EQ overrun\n");
break;
+ case MLX4_EVENT_TYPE_SQP_UPDATE:
+ mlx4_update_sqp(dev);
+ break;
+
default:
mlx4_warn(dev, "Unhandled event:%02x\n", eqe.type);
}
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index fb1c6f0..b01820f 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -111,6 +111,7 @@ enum mlx4_alloc_mode {
};
enum {
+ MLX4_MFUNC_MAX = 64,
MLX4_MFUNC_MAX_EQES = 8,
MLX4_MFUNC_EQE_MASK = (MLX4_MFUNC_MAX_EQES - 1)
};
@@ -219,6 +220,7 @@ struct mlx4_slave_state {
struct mlx4_slave_eqe eq[MLX4_MFUNC_MAX_EQES];
u16 eq_pi;
u16 eq_ci;
+ int sqp_start;
spinlock_t lock;
};
@@ -246,6 +248,7 @@ struct mlx4_mfunc {
dma_addr_t vhcr_dma;
struct mlx4_mfunc_master_ctx master;
+ u32 demux_sqp[MLX4_MFUNC_MAX];
};
struct mlx4_cmd {
@@ -531,6 +534,14 @@ int mlx4_QUERY_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *v
struct mlx4_cmd_mailbox *outbox);
int mlx4_get_port_ib_caps(struct mlx4_dev *dev, u8 port, __be32 *caps);
+int mlx4_CONF_SPECIAL_QP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
+int mlx4_GET_SLAVE_SQP(struct mlx4_dev *dev, u32 *sqp, int num);
+int mlx4_GET_SLAVE_SQP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
+
int mlx4_MCAST_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
struct mlx4_cmd_mailbox *inbox,
struct mlx4_cmd_mailbox *outbox);
diff --git a/drivers/net/mlx4/qp.c b/drivers/net/mlx4/qp.c
index 065c7fc..99b9ded 100644
--- a/drivers/net/mlx4/qp.c
+++ b/drivers/net/mlx4/qp.c
@@ -145,6 +145,54 @@ int mlx4_qp_modify(struct mlx4_dev *dev, struct mlx4_mtt *mtt,
}
EXPORT_SYMBOL_GPL(mlx4_qp_modify);
+u32 mlx4_get_slave_sqp(struct mlx4_dev *dev, int slave)
+{
+ if (mlx4_is_master(dev) && slave < dev->num_slaves) {
+ return mlx4_priv(dev)->mfunc.master.slave_state[slave].sqp_start;
+ }
+ if (mlx4_is_slave(dev) && slave < dev->caps.sqp_demux) {
+ return mlx4_priv(dev)->mfunc.demux_sqp[slave];
+ }
+ return 0;
+}
+EXPORT_SYMBOL_GPL(mlx4_get_slave_sqp);
+
+int mlx4_GET_SLAVE_SQP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ u32 *slave_sqp = outbox->buf;
+ int i;
+
+ /* CX1: special qp demultiplexing is done by slave0 */
+ if (slave) {
+ mlx4_warn(dev, "Denying slave_sqp request from slave:%d\n", slave);
+ return -EINVAL;
+ }
+ for (i = 0; i < 64; i++)
+ slave_sqp[i] = mlx4_get_slave_sqp(dev, i);
+ return 0;
+}
+
+int mlx4_GET_SLAVE_SQP(struct mlx4_dev *dev, u32 *sqp, int num)
+{
+ struct mlx4_cmd_mailbox *mailbox;
+ int err;
+
+ mailbox = mlx4_alloc_cmd_mailbox(dev);
+ if (IS_ERR(mailbox))
+ return PTR_ERR(mailbox);
+
+ err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0, MLX4_CMD_GET_SLAVE_SQP,
+ MLX4_CMD_TIME_CLASS_A);
+ if (!err)
+ memcpy(sqp, mailbox->buf, sizeof (u32) * num);
+
+ mlx4_free_cmd_mailbox(dev, mailbox);
+ return err;
+}
+EXPORT_SYMBOL_GPL(mlx4_GET_SLAVE_SQP);
+
int mlx4_qp_reserve_range(struct mlx4_dev *dev, int cnt, int align, int *base)
{
struct mlx4_priv *priv = mlx4_priv(dev);
@@ -331,6 +379,25 @@ static int mlx4_CONF_SPECIAL_QP(struct mlx4_dev *dev, u32 base_qpn)
MLX4_CMD_TIME_CLASS_B);
}
+int mlx4_CONF_SPECIAL_QP_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ int ret;
+
+ priv->mfunc.master.slave_state[slave].sqp_start = vhcr->in_modifier & 0xffffff;
+ if (!slave) {
+ /* CX1: slave0 owns real special QPs */
+ ret = mlx4_CONF_SPECIAL_QP(dev, priv->mfunc.master.slave_state[slave].sqp_start);
+ if (ret)
+ return ret;
+ }
+ /* Notify slave0 that an SQP change occured */
+ mlx4_slave_event(dev, 0, MLX4_EVENT_TYPE_SQP_UPDATE, 0, 0);
+ return 0;
+}
+
int mlx4_init_qp_table(struct mlx4_dev *dev)
{
struct mlx4_qp_table *qp_table = &mlx4_priv(dev)->qp_table;
diff --git a/include/linux/mlx4/cmd.h b/include/linux/mlx4/cmd.h
index 8265843..f1659f2 100644
--- a/include/linux/mlx4/cmd.h
+++ b/include/linux/mlx4/cmd.h
@@ -124,6 +124,7 @@ enum {
MLX4_CMD_FREE_RES = 0x51,
MLX4_CMD_GET_EVENT = 0x52,
MLX4_CMD_MCAST_ATTACH = 0x54,
+ MLX4_CMD_GET_SLAVE_SQP = 0x55,
/* debug commands */
MLX4_CMD_QUERY_DEBUG_MSG = 0x2a,
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index 9735f40..b20c8d8 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -99,6 +99,7 @@ enum mlx4_event {
MLX4_EVENT_TYPE_EQ_OVERFLOW = 0x0f,
MLX4_EVENT_TYPE_ECC_DETECT = 0x0e,
MLX4_EVENT_TYPE_CMD = 0x0a,
+ MLX4_EVENT_TYPE_SQP_UPDATE = 0xfe,
MLX4_EVENT_TYPE_NONE = 0xff,
};
@@ -242,6 +243,7 @@ struct mlx4_caps {
int log_num_prios;
enum mlx4_port_type port_type[MLX4_MAX_PORTS + 1];
u8 supported_type[MLX4_MAX_PORTS + 1];
+ u8 sqp_demux;
u32 port_mask;
enum mlx4_port_type possible_type[MLX4_MAX_PORTS + 1];
};
@@ -470,6 +472,7 @@ void mlx4_qp_release_range(struct mlx4_dev *dev, int base_qpn, int cnt);
int mlx4_qp_alloc(struct mlx4_dev *dev, int qpn, struct mlx4_qp *qp);
void mlx4_qp_free(struct mlx4_dev *dev, struct mlx4_qp *qp);
+u32 mlx4_get_slave_sqp(struct mlx4_dev *dev, int vf);
int mlx4_srq_alloc(struct mlx4_dev *dev, u32 pdn, struct mlx4_mtt *mtt,
u64 db_rec, struct mlx4_srq *srq);
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 08/25] mlx4_core: fix buggy parsing of reserved eq cap
From: Yevgeny Petrilin @ 2009-11-04 15:30 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/fw.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/drivers/net/mlx4/fw.c b/drivers/net/mlx4/fw.c
index a63f84a..87eb2d4 100644
--- a/drivers/net/mlx4/fw.c
+++ b/drivers/net/mlx4/fw.c
@@ -243,7 +243,7 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_MPT_OFFSET);
dev_cap->max_mpts = 1 << (field & 0x3f);
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_EQ_OFFSET);
- dev_cap->reserved_eqs = 1 << (field & 0xf);
+ dev_cap->reserved_eqs = field & 0xf;
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_EQ_OFFSET);
dev_cap->max_eqs = 1 << (field & 0xf);
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_MTT_OFFSET);
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 07/25] mlx4_core: add port para-virtualization
From: Yevgeny Petrilin @ 2009-11-04 15:30 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Ports are a shared resource among functions, so special behavior is needed here:
- Bring up ports if at least one function has done so.
- Bring down ports if all functions have done so.
- Aggregate IB port capabilities
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/cmd.c | 5 +++
drivers/net/mlx4/fw.c | 55 ++++++++++++++++++++++++++++++++++
drivers/net/mlx4/mlx4.h | 16 ++++++++++
drivers/net/mlx4/port.c | 76 +++++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 152 insertions(+), 0 deletions(-)
diff --git a/drivers/net/mlx4/cmd.c b/drivers/net/mlx4/cmd.c
index dd23aba..d2646d1 100644
--- a/drivers/net/mlx4/cmd.c
+++ b/drivers/net/mlx4/cmd.c
@@ -526,6 +526,11 @@ static struct mlx4_cmd_info {
{MLX4_CMD_QUERY_FW, 0, 1, 0, NULL, NULL},
{MLX4_CMD_QUERY_ADAPTER, 0, 1, 0, NULL, NULL},
+ {MLX4_CMD_INIT_PORT, 0, 0, 0, NULL, mlx4_INIT_PORT_wrapper},
+ {MLX4_CMD_CLOSE_PORT, 0, 0, 0, NULL, mlx4_CLOSE_PORT_wrapper},
+ {MLX4_CMD_QUERY_PORT, 0, 1, 0, NULL, mlx4_QUERY_PORT_wrapper},
+ {MLX4_CMD_SET_PORT, 1, 0, 0, NULL, mlx4_SET_PORT_wrapper},
+
{MLX4_CMD_SW2HW_EQ, 1, 0, 0, NULL, NULL}, /* need verifier */
{MLX4_CMD_NOP, 0, 0, 0, NULL, NULL},
{MLX4_CMD_ALLOC_RES, 0, 0, 1, NULL, mlx4_RESOURCE_wrapper},
diff --git a/drivers/net/mlx4/fw.c b/drivers/net/mlx4/fw.c
index 4ca8060..a63f84a 100644
--- a/drivers/net/mlx4/fw.c
+++ b/drivers/net/mlx4/fw.c
@@ -135,6 +135,14 @@ int mlx4_MOD_STAT_CFG(struct mlx4_dev *dev, struct mlx4_mod_stat_cfg *cfg)
return err;
}
+int mlx4_QUERY_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ return mlx4_cmd_box(dev, 0, outbox->dma, vhcr->in_modifier, 0, MLX4_CMD_QUERY_PORT,
+ MLX4_CMD_TIME_CLASS_B);
+}
+
int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
{
struct mlx4_cmd_mailbox *mailbox;
@@ -809,6 +817,29 @@ int mlx4_INIT_HCA(struct mlx4_dev *dev, struct mlx4_init_hca_param *param)
return err;
}
+int mlx4_INIT_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ int port;
+ int err;
+
+ port = vhcr->in_modifier;
+ if (priv->mfunc.master.slave_state[slave].init_port_mask & (1 << port))
+ return 0;
+
+ /* Enable port only if it was previously disabled */
+ if (!priv->mfunc.master.init_port_ref[port]) {
+ err = mlx4_INIT_PORT(dev, port);
+ if (err)
+ return err;
+ }
+ ++priv->mfunc.master.init_port_ref[port];
+ priv->mfunc.master.slave_state[slave].init_port_mask |= (1 << port);
+ return 0;
+}
+
int mlx4_INIT_PORT(struct mlx4_dev *dev, int port)
{
struct mlx4_cmd_mailbox *mailbox;
@@ -863,6 +894,30 @@ int mlx4_INIT_PORT(struct mlx4_dev *dev, int port)
}
EXPORT_SYMBOL_GPL(mlx4_INIT_PORT);
+int mlx4_CLOSE_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ int port;
+ int err;
+
+ port = vhcr->in_modifier;
+ if (!(priv->mfunc.master.slave_state[slave].init_port_mask & (1 << port)))
+ return 0;
+
+ /* CX1: master doesn't have interfaces - close port if this slave is
+ * the last user */
+ if (priv->mfunc.master.init_port_ref[port] == 1) {
+ err = mlx4_CLOSE_PORT(dev, port);
+ if (err)
+ return err;
+ }
+ --priv->mfunc.master.init_port_ref[port];
+ priv->mfunc.master.slave_state[slave].init_port_mask &= ~(1 << port);
+ return 0;
+}
+
int mlx4_CLOSE_PORT(struct mlx4_dev *dev, int port)
{
return mlx4_cmd(dev, 0, port, 0, MLX4_CMD_CLOSE_PORT, 1000);
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index fafa293..8a7a041 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -46,6 +46,7 @@
#include <linux/mlx4/driver.h>
#include <linux/mlx4/doorbell.h>
#include <linux/mlx4/cmd.h>
+#include <rdma/ib_verbs.h>
#define DRV_NAME "mlx4_core"
#define PFX DRV_NAME ": "
@@ -212,11 +213,14 @@ struct mlx4_slave_eqe {
struct mlx4_slave_state {
u8 comm_toggle;
u8 last_cmd;
+ u8 init_port_mask;
dma_addr_t vhcr_dma;
+ __be32 ib_cap_mask[MLX4_MAX_PORTS + 1];
};
struct mlx4_mfunc_master_ctx {
struct mlx4_slave_state *slave_state;
+ int init_port_ref[MLX4_MAX_PORTS + 1];
};
struct mlx4_vhcr {
@@ -505,6 +509,18 @@ void mlx4_init_mac_table(struct mlx4_dev *dev, struct mlx4_mac_table *table);
void mlx4_init_vlan_table(struct mlx4_dev *dev, struct mlx4_vlan_table *table);
int mlx4_SET_PORT(struct mlx4_dev *dev, u8 port);
+int mlx4_SET_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
+int mlx4_INIT_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
+int mlx4_CLOSE_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
+int mlx4_QUERY_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
int mlx4_get_port_ib_caps(struct mlx4_dev *dev, u8 port, __be32 *caps);
int mlx4_MCAST_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
diff --git a/drivers/net/mlx4/port.c b/drivers/net/mlx4/port.c
index 606aa58..67f0751 100644
--- a/drivers/net/mlx4/port.c
+++ b/drivers/net/mlx4/port.c
@@ -294,6 +294,82 @@ int mlx4_get_port_ib_caps(struct mlx4_dev *dev, u8 port, __be32 *caps)
return err;
}
+int mlx4_SET_PORT_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ int reset_qkey_viols;
+ int port;
+ int is_eth;
+ int err;
+ int i;
+ __be32 agg_cap_mask;
+ __be32 slave_cap_mask;
+ __be32 new_cap_mask;
+
+ port = vhcr->in_modifier & 0xff;
+ is_eth = vhcr->op_modifier;
+
+ /* For Ethernet, we currently support only slave0.
+ * TODO: add multi-vf support */
+ if (is_eth) {
+ if (slave)
+ return -EINVAL;
+ return mlx4_cmd(dev, inbox->dma, vhcr->in_modifier,
+ vhcr->op_modifier,
+ MLX4_CMD_SET_PORT,
+ MLX4_CMD_TIME_CLASS_B);
+ }
+
+ /* For IB, we only consider:
+ * - The capability mask, which is set to the aggregate of all slave frunction
+ * capabilities
+ * - The QKey violatin counter - reset according to each request.
+ */
+
+ if (dev->flags & MLX4_FLAG_OLD_PORT_CMDS) {
+ reset_qkey_viols = (*(u8 *) inbox->buf) & 0x40;
+ new_cap_mask = ((__be32 *) inbox->buf)[2];
+ } else {
+ reset_qkey_viols = ((u8 *) inbox->buf)[3] & 0x1;
+ new_cap_mask = ((__be32 *) inbox->buf)[1];
+ }
+
+ /* CX1: only slave0 has access to qp0 */
+ if (slave && (new_cap_mask & cpu_to_be32(IB_PORT_SM))) {
+ mlx4_warn(dev, "denying sm port capability for slave:%d\n", slave);
+ return -EINVAL;
+ }
+
+ agg_cap_mask = 0;
+ slave_cap_mask = priv->mfunc.master.slave_state[slave].ib_cap_mask[port];
+ priv->mfunc.master.slave_state[slave].ib_cap_mask[port] = new_cap_mask;
+ for (i = 0; i < dev->num_slaves; i++)
+ agg_cap_mask |= priv->mfunc.master.slave_state[slave].ib_cap_mask[port];
+
+#if 0
+ mlx4_warn(dev, "old_slave_cap:0x%x slave_cap:0x%x cap:0x%x qkey_reset:%d\n",
+ slave_cap_mask, priv->mfunc.master.slave_state[slave].ib_cap_mask[port],
+ agg_cap_mask, reset_qkey_viols);
+#endif
+
+ memset(inbox->buf, 0, 256);
+ if (dev->flags & MLX4_FLAG_OLD_PORT_CMDS) {
+ *(u8 *) inbox->buf = !!reset_qkey_viols << 6;
+ ((__be32 *) inbox->buf)[2] = agg_cap_mask;
+ } else {
+ ((u8 *) inbox->buf)[3] = !!reset_qkey_viols;
+ ((__be32 *) inbox->buf)[1] = agg_cap_mask;
+ }
+
+ err = mlx4_cmd(dev, inbox->dma, port, is_eth, MLX4_CMD_SET_PORT,
+ MLX4_CMD_TIME_CLASS_B);
+ if (err)
+ priv->mfunc.master.slave_state[slave].ib_cap_mask[port] = slave_cap_mask;
+ return err;
+}
+
int mlx4_SET_PORT(struct mlx4_dev *dev, u8 port)
{
struct mlx4_cmd_mailbox *mailbox;
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 06/25] mlx4_core: slave multicast support
From: Yevgeny Petrilin @ 2009-11-04 15:30 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Multicast table processing requires multiple related commands.
To keep things simple, low-level multicast handling is done only by the master;
a new virtual command is added to allow slaves to attach/detach QPs to mulitcast
groups at a higher abstraction level.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/cmd.c | 1 +
drivers/net/mlx4/mcg.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
drivers/net/mlx4/mlx4.h | 4 ++++
include/linux/mlx4/cmd.h | 1 +
4 files changed, 51 insertions(+), 0 deletions(-)
diff --git a/drivers/net/mlx4/cmd.c b/drivers/net/mlx4/cmd.c
index b1701d2..dd23aba 100644
--- a/drivers/net/mlx4/cmd.c
+++ b/drivers/net/mlx4/cmd.c
@@ -565,6 +565,7 @@ static struct mlx4_cmd_info {
{MLX4_CMD_MAD_IFC, 1, 1, 0, NULL, NULL}, /* need verifier */
/* Native multicast commands are not available for guests */
+ {MLX4_CMD_MCAST_ATTACH, 1, 0, 0, NULL, mlx4_MCAST_wrapper},
{MLX4_CMD_DIAG_RPRT, 0, 1, 0, NULL, NULL}, /* need verifier */
/* Ethernet specific commands */
diff --git a/drivers/net/mlx4/mcg.c b/drivers/net/mlx4/mcg.c
index 5ccbce9..96f09ed 100644
--- a/drivers/net/mlx4/mcg.c
+++ b/drivers/net/mlx4/mcg.c
@@ -146,6 +146,45 @@ static int find_mgm(struct mlx4_dev *dev,
return err;
}
+int mlx4_MCAST_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox)
+{
+ struct mlx4_qp qp; /* dummy for calling attach/detach */
+
+ qp.qpn = vhcr->in_modifier & 0xffffff;
+ if (vhcr->op_modifier)
+ return mlx4_multicast_attach(dev, &qp, inbox->buf, vhcr->in_modifier >> 31);
+ else
+ return mlx4_multicast_detach(dev, &qp, inbox->buf);
+}
+
+static int mlx4_MCAST(struct mlx4_dev *dev, struct mlx4_qp *qp,
+ u8 gid[16], u8 attach,
+ u8 block_loopback)
+{
+ struct mlx4_cmd_mailbox *mailbox;
+ int err;
+ int qpn;
+
+ if (!mlx4_is_slave(dev))
+ return -EBADF;
+
+ mailbox = mlx4_alloc_cmd_mailbox(dev);
+ if (IS_ERR(mailbox))
+ return PTR_ERR(mailbox);
+
+ memcpy(mailbox->buf, gid, 16);
+ qpn = qp->qpn;
+ if (attach && block_loopback)
+ qpn |= (1 << 31);
+
+ err = mlx4_cmd(dev, mailbox->dma, qpn, attach, MLX4_CMD_MCAST_ATTACH,
+ MLX4_CMD_TIME_CLASS_A);
+ mlx4_free_cmd_mailbox(dev, mailbox);
+ return err;
+}
+
int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
int block_mcast_loopback)
{
@@ -159,6 +198,9 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
int i;
int err;
+ if (mlx4_is_slave(dev))
+ return mlx4_MCAST(dev, qp, gid, 1, block_mcast_loopback);
+
mailbox = mlx4_alloc_cmd_mailbox(dev);
if (IS_ERR(mailbox))
return PTR_ERR(mailbox);
@@ -254,6 +296,9 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
int i, loc;
int err;
+ if (mlx4_is_slave(dev))
+ return mlx4_MCAST(dev, qp, gid, 0, 0);
+
mailbox = mlx4_alloc_cmd_mailbox(dev);
if (IS_ERR(mailbox))
return PTR_ERR(mailbox);
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index 91f1215..fafa293 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -507,4 +507,8 @@ void mlx4_init_vlan_table(struct mlx4_dev *dev, struct mlx4_vlan_table *table);
int mlx4_SET_PORT(struct mlx4_dev *dev, u8 port);
int mlx4_get_port_ib_caps(struct mlx4_dev *dev, u8 port, __be32 *caps);
+int mlx4_MCAST_wrapper(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
+
#endif /* MLX4_H */
diff --git a/include/linux/mlx4/cmd.h b/include/linux/mlx4/cmd.h
index dceb93f..8265843 100644
--- a/include/linux/mlx4/cmd.h
+++ b/include/linux/mlx4/cmd.h
@@ -123,6 +123,7 @@ enum {
MLX4_CMD_ALLOC_RES = 0x50,
MLX4_CMD_FREE_RES = 0x51,
MLX4_CMD_GET_EVENT = 0x52,
+ MLX4_CMD_MCAST_ATTACH = 0x54,
/* debug commands */
MLX4_CMD_QUERY_DEBUG_MSG = 0x2a,
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 03/25] mlx4_core: add multi-function communication channel
From: Yevgeny Petrilin @ 2009-11-04 15:30 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
The communication channel consists of 2 registers per vf (a slave function)
that are shared with the pf (the master function), as well as a new command for
inter-function memory copying (only exposed to the master).
The communication channel is used to establish a Virtual HCA Command Register
(vHCR) in each slave function, which allows it to pass FW commands to the master
function for execution.
The slave also uses the vHCR to pull slave-specific events from the master.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/cmd.c | 483 ++++++++++++++++++++++++++++++++++++++++++-
drivers/net/mlx4/eq.c | 67 ++++++
drivers/net/mlx4/fw.c | 8 +
drivers/net/mlx4/mlx4.h | 72 +++++++-
include/linux/mlx4/cmd.h | 7 +-
include/linux/mlx4/device.h | 3 +-
6 files changed, 627 insertions(+), 13 deletions(-)
diff --git a/drivers/net/mlx4/cmd.c b/drivers/net/mlx4/cmd.c
index 65ec77d..03dfdfc 100644
--- a/drivers/net/mlx4/cmd.c
+++ b/drivers/net/mlx4/cmd.c
@@ -41,6 +41,7 @@
#include <asm/io.h>
#include "mlx4.h"
+#include "en_port.h"
#define CMD_POLL_TOKEN 0xffff
@@ -140,6 +141,46 @@ static int mlx4_status_to_errno(u8 status)
return trans_table[status];
}
+static int comm_pending(struct mlx4_dev *dev)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ __be32 status = readl(&priv->mfunc.comm->slave_read);
+
+ return (be32_to_cpu(status) >> 30) != priv->cmd.comm_toggle;
+}
+
+int mlx4_comm_cmd(struct mlx4_dev *dev, u8 cmd, u16 param, unsigned long timeout)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ unsigned long end;
+ u32 val;
+
+ /* First, verify that the master reports correct status */
+ if (comm_pending(dev)) {
+ mlx4_warn(dev, "Communication channel is not idle\n");
+ return -EAGAIN;
+ }
+
+ /* Write command */
+ if (cmd == MLX4_COMM_CMD_RESET)
+ priv->cmd.comm_toggle = 0;
+ else if (++priv->cmd.comm_toggle > 2)
+ priv->cmd.comm_toggle = 1;
+ val = param | (cmd << 16) | (priv->cmd.comm_toggle << 30);
+ __raw_writel((__force u32) cpu_to_be32(val), &priv->mfunc.comm->slave_write);
+ wmb();
+
+ end = msecs_to_jiffies(timeout) + jiffies;
+ while (comm_pending(dev) && time_before(jiffies, end))
+ cond_resched();
+
+ if (comm_pending(dev)) {
+ mlx4_warn(dev, "Communication channel timed out\n");
+ return -ETIMEDOUT;
+ }
+ return 0;
+}
+
static int cmd_pending(struct mlx4_dev *dev)
{
u32 status = readl(mlx4_priv(dev)->cmd.hcr + HCR_STATUS_OFFSET);
@@ -207,6 +248,33 @@ out:
return ret;
}
+static int mlx4_slave_cmd_poll(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
+ int out_is_imm, u32 in_modifier, u8 op_modifier,
+ u16 op, unsigned long timeout)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ struct mlx4_vhcr *vhcr = priv->mfunc.vhcr;
+ int ret;
+
+ down(&priv->cmd.poll_sem);
+ vhcr->in_param = in_param;
+ vhcr->out_param = out_param ? *out_param : 0;
+ vhcr->in_modifier = in_modifier;
+ vhcr->timeout = timeout;
+ vhcr->op = op;
+ vhcr->token = CMD_POLL_TOKEN;
+ vhcr->op_modifier = op_modifier;
+ vhcr->errno = 0;
+ ret = mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR_POST, 0, MLX4_COMM_TIME + timeout);
+ if (!ret) {
+ if (out_is_imm)
+ *out_param = vhcr->out_param;
+ ret = vhcr->errno;
+ }
+ up(&priv->cmd.poll_sem);
+ return ret;
+}
+
static int mlx4_cmd_poll(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
int out_is_imm, u32 in_modifier, u8 op_modifier,
u16 op, unsigned long timeout)
@@ -314,12 +382,380 @@ int __mlx4_cmd(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
if (mlx4_priv(dev)->cmd.use_events)
return mlx4_cmd_wait(dev, in_param, out_param, out_is_imm,
in_modifier, op_modifier, op, timeout);
+
+ if (mlx4_is_slave(dev))
+ return mlx4_slave_cmd_poll(dev, in_param, out_param, out_is_imm,
+ in_modifier, op_modifier, op, timeout);
else
return mlx4_cmd_poll(dev, in_param, out_param, out_is_imm,
in_modifier, op_modifier, op, timeout);
}
EXPORT_SYMBOL_GPL(__mlx4_cmd);
+int mlx4_ACCESS_MEM(struct mlx4_dev *dev, u64 master_addr, int slave, u64 slave_addr,
+ int size, int is_read)
+{
+ u64 in_param;
+ u64 out_param;
+ int slave_id = slave + 1; /* index 0 is reserved for the master */
+
+ if ((slave_addr & 0xfff) | (master_addr & 0xfff) |
+ (slave_id & ~0x7f) | (size & 0xff)) {
+ mlx4_err(dev, "Bad access mem params - slave_addr:0x%llx "
+ "master_addr:0x%llx slave_id:%d size:%d\n",
+ slave_addr, master_addr, slave_id, size);
+ return -EINVAL;
+ }
+
+ if (is_read) {
+ in_param = (u64) slave_id | slave_addr;
+ out_param = master_addr;
+ } else {
+ in_param = master_addr;
+ out_param = (u64) slave_id | slave_addr;
+ }
+
+ return mlx4_cmd_imm(dev, in_param, &out_param, size, 0,
+ MLX4_CMD_ACCESS_MEM,
+ MLX4_CMD_TIME_CLASS_A);
+}
+
+static struct mlx4_cmd_info {
+ u8 opcode;
+ u8 has_inbox;
+ u8 has_outbox;
+ u8 out_is_imm;
+ int (*verify)(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox);
+ int (*wrapper)(struct mlx4_dev *dev, int slave, struct mlx4_vhcr *vhcr,
+ struct mlx4_cmd_mailbox *inbox,
+ struct mlx4_cmd_mailbox *outbox);
+} cmd_info[] = {
+ {MLX4_CMD_QUERY_FW, 0, 1, 0, NULL, NULL},
+ {MLX4_CMD_QUERY_ADAPTER, 0, 1, 0, NULL, NULL},
+
+ {MLX4_CMD_SW2HW_EQ, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_NOP, 0, 0, 0, NULL, NULL},
+
+ {MLX4_CMD_SW2HW_MPT, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_QUERY_MPT, 0, 1, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_HW2SW_MPT, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_READ_MTT, 0, 1, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_SYNC_TPT, 1, 0, 0, NULL, NULL}, /* need verifier */
+
+ {MLX4_CMD_HW2SW_EQ, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_QUERY_EQ, 0, 1, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_SW2HW_CQ, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_HW2SW_CQ, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_QUERY_CQ, 0, 1, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_MODIFY_CQ, 1, 0, 1, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_SW2HW_SRQ, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_HW2SW_SRQ, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_QUERY_SRQ, 0, 1, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_ARM_SRQ, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_RST2INIT_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_INIT2RTR_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_RTR2RTS_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_RTS2RTS_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_SQERR2RTS_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_2ERR_QP, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_RTS2SQD_QP, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_SQD2SQD_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_SQD2RTS_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_2RST_QP, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_QUERY_QP, 0, 1, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_INIT2INIT_QP, 1, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_SUSPEND_QP, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_UNSUSPEND_QP, 0, 0, 0, NULL, NULL}, /* need verifier */
+ {MLX4_CMD_MAD_IFC, 1, 1, 0, NULL, NULL}, /* need verifier */
+
+ /* Native multicast commands are not available for guests */
+ {MLX4_CMD_DIAG_RPRT, 0, 1, 0, NULL, NULL}, /* need verifier */
+
+ /* Ethernet specific commands */
+ {MLX4_CMD_SET_VLAN_FLTR, 1, 0, 0, NULL, NULL}, /* need wrapper */
+ {MLX4_CMD_SET_MCAST_FLTR, 0, 0, 0, NULL, NULL}, /* need wrapper */
+ {MLX4_CMD_DUMP_ETH_STATS, 0, 1, 0, NULL, NULL}, /* need wrapper */
+};
+
+static int mlx4_master_process_vhcr(struct mlx4_dev *dev, int slave)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ struct mlx4_cmd_info *cmd = NULL;
+ struct mlx4_vhcr *vhcr = priv->mfunc.vhcr;
+ struct mlx4_cmd_mailbox *inbox = NULL;
+ struct mlx4_cmd_mailbox *outbox = NULL;
+ u64 in_param;
+ u64 out_param;
+ int ret;
+ int i;
+
+ /* DMA in the vHCR */
+ ret = mlx4_ACCESS_MEM(dev, priv->mfunc.vhcr_dma, slave,
+ priv->mfunc.master.slave_state[slave].vhcr_dma,
+ ALIGN(sizeof(struct mlx4_vhcr),
+ MLX4_ACCESS_MEM_ALIGN), 1);
+ if (ret) {
+ mlx4_err(dev, "Failed reading vhcr\n");
+ return ret;
+ }
+
+ /* Lookup command */
+ for (i = 0; i < ARRAY_SIZE(cmd_info); ++i) {
+ if (vhcr->op == cmd_info[i].opcode) {
+ cmd = &cmd_info[i];
+ break;
+ }
+ }
+ if (!cmd) {
+ mlx4_err(dev, "Unknown command:0x%x accepted from slave:%d\n",
+ vhcr->op, slave);
+ vhcr->errno = -EINVAL;
+ goto out_status;
+ }
+
+ /* Read inbox */
+ if (cmd->has_inbox) {
+ inbox = mlx4_alloc_cmd_mailbox(dev);
+ if (IS_ERR(inbox)) {
+ ret = PTR_ERR(inbox);
+ inbox = NULL;
+ goto out;
+ }
+
+ /* FIXME: add mailbox size per-command */
+ ret = mlx4_ACCESS_MEM(dev, inbox->dma, slave,
+ vhcr->in_param,
+ MLX4_MAILBOX_SIZE, 1);
+ if (ret) {
+ mlx4_err(dev, "Failed reading inbox\n");
+ goto out;
+ }
+ }
+
+ /* Apply permission and bound checks if applicable */
+ if (cmd->verify && cmd->verify(dev, slave, vhcr, inbox)) {
+ mlx4_warn(dev, "Command:0x%x failed protection checks\n", vhcr->op);
+ vhcr->errno = -EPERM;
+ goto out_status;
+ }
+
+ /* Allocate outbox */
+ if (cmd->has_outbox) {
+ outbox = mlx4_alloc_cmd_mailbox(dev);
+ if (IS_ERR(outbox)) {
+ ret = PTR_ERR(outbox);
+ outbox = NULL;
+ goto out;
+ }
+ }
+
+ /* Execute the command! */
+ if (cmd->wrapper)
+ vhcr->errno = cmd->wrapper(dev, slave, vhcr, inbox, outbox);
+ else {
+ in_param = cmd->has_inbox ? (u64) inbox->dma : vhcr->in_param;
+ out_param = cmd->has_outbox ? (u64) outbox->dma : vhcr->out_param;
+ vhcr->errno = __mlx4_cmd(dev, in_param, &out_param,
+ cmd->out_is_imm,
+ vhcr->in_modifier,
+ vhcr->op_modifier,
+ vhcr->op,
+ vhcr->timeout);
+ if (cmd->out_is_imm)
+ vhcr->out_param = out_param;
+ }
+
+ /* Write outbox if command completed successfully */
+ if (cmd->has_outbox && !vhcr->errno) {
+ ret = mlx4_ACCESS_MEM(dev, outbox->dma, slave,
+ vhcr->out_param,
+ MLX4_MAILBOX_SIZE, 0);
+ if (ret) {
+ mlx4_err(dev, "Failed writing outbox\n");
+ goto out;
+ }
+ }
+
+out_status:
+ /* DMA back vhcr result */
+ ret = mlx4_ACCESS_MEM(dev, priv->mfunc.vhcr_dma, slave,
+ priv->mfunc.master.slave_state[slave].vhcr_dma,
+ ALIGN(sizeof(struct mlx4_vhcr),
+ MLX4_ACCESS_MEM_ALIGN), 0);
+ if (ret)
+ mlx4_err(dev, "Failed writing vhcr result\n");
+
+ if (vhcr->errno)
+ mlx4_warn(dev, "vhcr command:0x%x slave:%d failed with error:%d\n",
+ vhcr->op, slave, vhcr->errno);
+ /* Fall through... */
+
+out:
+ mlx4_free_cmd_mailbox(dev, inbox);
+ mlx4_free_cmd_mailbox(dev, outbox);
+ return ret;
+}
+
+static void mlx4_master_do_cmd(struct mlx4_dev *dev, int slave, u8 cmd, u16 param, u8 toggle)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ struct mlx4_slave_state *slave_state = priv->mfunc.master.slave_state;
+ u8 toggle_next;
+ u32 reply;
+
+ if (cmd == MLX4_COMM_CMD_RESET) {
+ mlx4_warn(dev, "Received reset from slave:%d\n", slave);
+ goto reset_slave;
+ }
+
+ /* Increment next toggle token */
+ toggle_next = slave_state[slave].comm_toggle + 1;
+ if (toggle_next > 2)
+ toggle_next = 1;
+ if (toggle != toggle_next) {
+ mlx4_warn(dev, "Incorrect token:%d from slave:%d expected:%d\n",
+ toggle, toggle_next, slave);
+ goto reset_slave;
+ }
+
+ switch (cmd) {
+ case MLX4_COMM_CMD_VHCR0:
+ if (slave_state[slave].last_cmd != MLX4_COMM_CMD_RESET)
+ goto reset_slave;
+ slave_state[slave].vhcr_dma = ((u64) param) << 48;
+ break;
+ case MLX4_COMM_CMD_VHCR1:
+ if (slave_state[slave].last_cmd != MLX4_COMM_CMD_VHCR0)
+ goto reset_slave;
+ slave_state[slave].vhcr_dma |= ((u64) param) << 32;
+ break;
+ case MLX4_COMM_CMD_VHCR2:
+ if (slave_state[slave].last_cmd != MLX4_COMM_CMD_VHCR1)
+ goto reset_slave;
+ slave_state[slave].vhcr_dma |= ((u64) param) << 16;
+ break;
+ case MLX4_COMM_CMD_VHCR_EN:
+ if (slave_state[slave].last_cmd != MLX4_COMM_CMD_VHCR2)
+ goto reset_slave;
+ slave_state[slave].vhcr_dma |= param;
+ break;
+ case MLX4_COMM_CMD_VHCR_POST:
+ if ((slave_state[slave].last_cmd != MLX4_COMM_CMD_VHCR_EN) &&
+ (slave_state[slave].last_cmd != MLX4_COMM_CMD_VHCR_POST))
+ goto reset_slave;
+ if (mlx4_master_process_vhcr(dev, slave)) {
+ mlx4_err(dev, "Failed processing vhcr for slave:%d, reseting slave.\n", slave);
+ goto reset_slave;
+ }
+ break;
+ default:
+ mlx4_warn(dev, "Bad comm cmd:%d from slave:%d\n", cmd, slave);
+ goto reset_slave;
+ }
+
+ slave_state[slave].last_cmd = cmd;
+ slave_state[slave].comm_toggle = toggle_next;
+ reply = (u32) toggle_next << 30;
+ __raw_writel((__force u32) cpu_to_be32(reply),
+ &priv->mfunc.comm[slave].slave_read);
+ wmb();
+ return;
+
+reset_slave:
+ /* FIXME: cleanup any slave resources */
+ slave_state[slave].last_cmd = MLX4_COMM_CMD_RESET;
+ slave_state[slave].comm_toggle = 0;
+ __raw_writel((__force u32) 0, &priv->mfunc.comm[slave].slave_write);
+ __raw_writel((__force u32) 0, &priv->mfunc.comm[slave].slave_read);
+ wmb();
+}
+
+/* master command processing */
+static void mlx4_master_poll_comm(struct work_struct *work)
+{
+ struct delayed_work *delay = container_of(work, struct delayed_work, work);
+ struct mlx4_mfunc *mfunc = container_of(delay, struct mlx4_mfunc, comm_work);
+ struct mlx4_priv *priv = container_of(mfunc, struct mlx4_priv, mfunc);
+ struct mlx4_dev *dev = &priv->dev;
+ u32 comm_cmd;
+ int polled = 0;
+ int i;
+
+ /* Give each slave a chance for one command */
+ for (i = 0; i < dev->num_slaves; i++) {
+ comm_cmd = be32_to_cpu(readl(&priv->mfunc.comm[i].slave_write));
+ if (comm_cmd >> 30 != priv->mfunc.master.slave_state[i].comm_toggle) {
+ mlx4_master_do_cmd(dev, i, comm_cmd >> 16, comm_cmd, comm_cmd >> 30);
+ polled = 1;
+ }
+ }
+ queue_delayed_work(priv->mfunc.comm_wq, &priv->mfunc.comm_work,
+ polled ? HZ / 1000 : HZ / 10);
+}
+
+int mlx4_multi_func_init(struct mlx4_dev *dev)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+ int i;
+
+ priv->mfunc.vhcr = dma_alloc_coherent(&(dev->pdev->dev), PAGE_SIZE,
+ &priv->mfunc.vhcr_dma,
+ GFP_KERNEL);
+ if (!priv->mfunc.vhcr) {
+ mlx4_err(dev, "Couldn't allocate vhcr.\n");
+ return -ENOMEM;
+ }
+
+ if (mlx4_is_master(dev))
+ priv->mfunc.comm = ioremap(pci_resource_start(dev->pdev,
+ priv->fw.comm_bar) +
+ priv->fw.comm_base,
+ MLX4_COMM_PAGESIZE);
+ else
+ priv->mfunc.comm = ioremap(pci_resource_start(dev->pdev, 0) +
+ MLX4_SLAVE_COMM_BASE,
+ MLX4_COMM_PAGESIZE);
+ if (!priv->mfunc.comm) {
+ mlx4_err(dev, "Couldn't map communication vector.");
+ goto err_vhcr;
+ }
+
+ if (mlx4_is_master(dev)) {
+ priv->mfunc.master.slave_state = kzalloc(dev->num_slaves *
+ sizeof(struct mlx4_slave_state),
+ GFP_KERNEL);
+ if (!priv->mfunc.master.slave_state)
+ goto err_comm;
+
+ for (i = 0; i < dev->num_slaves; ++i)
+ priv->mfunc.master.slave_state[i].last_cmd = MLX4_COMM_CMD_RESET;
+
+ INIT_DELAYED_WORK(&priv->mfunc.comm_work, mlx4_master_poll_comm);
+ priv->mfunc.comm_wq = create_singlethread_workqueue("mlx4_comm");
+ if (!priv->mfunc.comm_wq) {
+ kfree(priv->mfunc.master.slave_state);
+ goto err_comm;
+ }
+ } else {
+ priv->cmd.comm_toggle = 0;
+ INIT_DELAYED_WORK(&priv->mfunc.comm_work, mlx4_slave_async_eq_poll);
+ priv->mfunc.comm_wq = create_singlethread_workqueue("mlx4_event");
+ if (!priv->mfunc.comm_wq)
+ goto err_comm;
+ }
+ return 0;
+
+err_comm:
+ iounmap(priv->mfunc.comm);
+err_vhcr:
+ dma_free_coherent(&(dev->pdev->dev), PAGE_SIZE,
+ priv->mfunc.vhcr,
+ priv->mfunc.vhcr_dma);
+ priv->mfunc.vhcr = NULL;
+ return -ENOMEM;
+}
+
int mlx4_cmd_init(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
@@ -329,22 +765,47 @@ int mlx4_cmd_init(struct mlx4_dev *dev)
priv->cmd.use_events = 0;
priv->cmd.toggle = 1;
- priv->cmd.hcr = ioremap(pci_resource_start(dev->pdev, 0) + MLX4_HCR_BASE,
- MLX4_HCR_SIZE);
- if (!priv->cmd.hcr) {
- mlx4_err(dev, "Couldn't map command register.");
- return -ENOMEM;
+ priv->cmd.hcr = priv->mfunc.vhcr = NULL;
+
+ if (!mlx4_is_slave(dev)) {
+ priv->cmd.hcr = ioremap(pci_resource_start(dev->pdev, 0) +
+ ((dev->flags & MLX4_FLAG_SRIOV) ?
+ MLX4_HCR_SRIOV_BASE :
+ MLX4_HCR_BASE),
+ MLX4_HCR_SIZE);
+ if (!priv->cmd.hcr) {
+ mlx4_err(dev, "Couldn't map command register.");
+ return -ENOMEM;
+ }
}
priv->cmd.pool = pci_pool_create("mlx4_cmd", dev->pdev,
MLX4_MAILBOX_SIZE,
MLX4_MAILBOX_SIZE, 0);
- if (!priv->cmd.pool) {
- iounmap(priv->cmd.hcr);
- return -ENOMEM;
- }
+ if (!priv->cmd.pool)
+ goto err_hcr;
return 0;
+
+err_hcr:
+ if (!mlx4_is_slave(dev))
+ iounmap(priv->cmd.hcr);
+ return -ENOMEM;
+}
+
+void mlx4_multi_func_cleanup(struct mlx4_dev *dev)
+{
+ struct mlx4_priv *priv = mlx4_priv(dev);
+
+ if (priv->mfunc.vhcr) {
+ destroy_workqueue(priv->mfunc.comm_wq);
+ kfree(priv->mfunc.master.slave_state);
+ iounmap(priv->mfunc.comm);
+ dma_free_coherent(&(dev->pdev->dev), PAGE_SIZE,
+ priv->mfunc.vhcr,
+ priv->mfunc.vhcr_dma);
+ priv->mfunc.vhcr = NULL;
+ }
}
void mlx4_cmd_cleanup(struct mlx4_dev *dev)
@@ -352,7 +813,9 @@ void mlx4_cmd_cleanup(struct mlx4_dev *dev)
struct mlx4_priv *priv = mlx4_priv(dev);
pci_pool_destroy(priv->cmd.pool);
- iounmap(priv->cmd.hcr);
+
+ if (!mlx4_is_slave(dev))
+ iounmap(priv->cmd.hcr);
}
/*
diff --git a/drivers/net/mlx4/eq.c b/drivers/net/mlx4/eq.c
index bffb799..70c16d4 100644
--- a/drivers/net/mlx4/eq.c
+++ b/drivers/net/mlx4/eq.c
@@ -160,6 +160,22 @@ static struct mlx4_eqe *next_eqe_sw(struct mlx4_eq *eq)
return !!(eqe->owner & 0x80) ^ !!(eq->cons_index & eq->nent) ? NULL : eqe;
}
+static int mlx4_GET_EVENT(struct mlx4_dev *dev, struct mlx4_slave_eqe *eqe)
+{
+ int ret;
+ u64 out_param;
+
+ ret = mlx4_cmd_imm(dev, 0, &out_param, 0, 0, MLX4_CMD_GET_EVENT,
+ MLX4_CMD_TIME_CLASS_A);
+ if (!ret) {
+ eqe->type = out_param & 0xff;
+ eqe->port = (out_param >> 8) & 0xff;
+ eqe->param = out_param >> 32;
+ } else
+ mlx4_err(dev, "Failed retrieving event\n");
+ return ret;
+}
+
static int mlx4_eq_int(struct mlx4_dev *dev, struct mlx4_eq *eq)
{
struct mlx4_eqe *eqe;
@@ -262,6 +278,57 @@ static int mlx4_eq_int(struct mlx4_dev *dev, struct mlx4_eq *eq)
return eqes_found;
}
+void mlx4_slave_async_eq_poll(struct work_struct *work)
+{
+ struct delayed_work *delay = container_of(work, struct delayed_work, work);
+ struct mlx4_mfunc *mfunc = container_of(delay, struct mlx4_mfunc, comm_work);
+ struct mlx4_priv *priv = container_of(mfunc, struct mlx4_priv, mfunc);
+ struct mlx4_dev *dev = &priv->dev;
+ struct mlx4_slave_eqe eqe;
+ int ret;
+ int i;
+
+ for (i = 0; i < MLX4_MFUNC_MAX_EQES; i++) {
+ ret = mlx4_GET_EVENT(dev, &eqe);
+ if (ret || eqe.type == MLX4_EVENT_TYPE_NONE)
+ break;
+
+ switch (eqe.type) {
+ case MLX4_EVENT_TYPE_PATH_MIG:
+ case MLX4_EVENT_TYPE_COMM_EST:
+ case MLX4_EVENT_TYPE_SQ_DRAINED:
+ case MLX4_EVENT_TYPE_SRQ_QP_LAST_WQE:
+ case MLX4_EVENT_TYPE_WQ_CATAS_ERROR:
+ case MLX4_EVENT_TYPE_PATH_MIG_FAILED:
+ case MLX4_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
+ case MLX4_EVENT_TYPE_WQ_ACCESS_ERROR:
+ mlx4_qp_event(dev, eqe.param, eqe.type);
+ break;
+
+ case MLX4_EVENT_TYPE_SRQ_LIMIT:
+ case MLX4_EVENT_TYPE_SRQ_CATAS_ERROR:
+ mlx4_srq_event(dev, eqe.param, eqe.type);
+ break;
+
+ case MLX4_EVENT_TYPE_PORT_CHANGE:
+ mlx4_dispatch_event(dev, eqe.param, eqe.port);
+ break;
+
+ case MLX4_EVENT_TYPE_CQ_ERROR:
+ mlx4_cq_event(dev, eqe.param, eqe.type);
+ break;
+
+ case MLX4_EVENT_TYPE_EQ_OVERFLOW:
+ mlx4_warn(dev, "slave async EQ overrun\n");
+ break;
+
+ default:
+ mlx4_warn(dev, "Unhandled event:%02x\n", eqe.type);
+ }
+ }
+ queue_delayed_work(priv->mfunc.comm_wq, &priv->mfunc.comm_work, HZ);
+}
+
static irqreturn_t mlx4_interrupt(int irq, void *dev_ptr)
{
struct mlx4_dev *dev = dev_ptr;
diff --git a/drivers/net/mlx4/fw.c b/drivers/net/mlx4/fw.c
index 3c16602..4ca8060 100644
--- a/drivers/net/mlx4/fw.c
+++ b/drivers/net/mlx4/fw.c
@@ -554,6 +554,9 @@ int mlx4_QUERY_FW(struct mlx4_dev *dev)
#define QUERY_FW_CLR_INT_BASE_OFFSET 0x20
#define QUERY_FW_CLR_INT_BAR_OFFSET 0x28
+#define QUERY_FW_COMM_BASE_OFFSET 0x40
+#define QUERY_FW_COMM_BAR_OFFSET 0x48
+
mailbox = mlx4_alloc_cmd_mailbox(dev);
if (IS_ERR(mailbox))
return PTR_ERR(mailbox);
@@ -614,6 +617,11 @@ int mlx4_QUERY_FW(struct mlx4_dev *dev)
MLX4_GET(fw->clr_int_bar, outbox, QUERY_FW_CLR_INT_BAR_OFFSET);
fw->clr_int_bar = (fw->clr_int_bar >> 6) * 2;
+ MLX4_GET(fw->comm_base, outbox, QUERY_FW_COMM_BASE_OFFSET);
+ MLX4_GET(fw->comm_bar, outbox, QUERY_FW_COMM_BAR_OFFSET);
+ fw->comm_bar = (fw->comm_bar >> 6) * 2;
+ mlx4_dbg(dev, "Communication vector bar:%d offset:0x%llx\n", fw->comm_bar,
+ fw->comm_base);
mlx4_dbg(dev, "FW size %d KB\n", fw->fw_pages >> 2);
/*
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index 5836c94..88b4ff4 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -53,8 +53,11 @@
enum {
MLX4_HCR_BASE = 0x80680,
+ MLX4_HCR_SRIOV_BASE = 0x4080680, /* good for SRIOV FW ony */
MLX4_HCR_SIZE = 0x0001c,
- MLX4_CLR_INT_SIZE = 0x00008
+ MLX4_CLR_INT_SIZE = 0x00008,
+ MLX4_SLAVE_COMM_BASE = 0x0,
+ MLX4_COMM_PAGESIZE = 0x1000
};
enum {
@@ -80,6 +83,21 @@ enum {
MLX4_NUM_CMPTS = MLX4_CMPT_NUM_TYPE << MLX4_CMPT_SHIFT
};
+#define MLX4_COMM_TIME 10000
+enum {
+ MLX4_COMM_CMD_RESET,
+ MLX4_COMM_CMD_VHCR0,
+ MLX4_COMM_CMD_VHCR1,
+ MLX4_COMM_CMD_VHCR2,
+ MLX4_COMM_CMD_VHCR_EN,
+ MLX4_COMM_CMD_VHCR_POST
+};
+
+enum {
+ MLX4_MFUNC_MAX_EQES = 8,
+ MLX4_MFUNC_EQE_MASK = (MLX4_MFUNC_MAX_EQES - 1)
+};
+
#ifdef CONFIG_MLX4_DEBUG
extern int mlx4_debug_level;
#else /* CONFIG_MLX4_DEBUG */
@@ -154,12 +172,56 @@ struct mlx4_profile {
struct mlx4_fw {
u64 clr_int_base;
u64 catas_offset;
+ u64 comm_base;
struct mlx4_icm *fw_icm;
struct mlx4_icm *aux_icm;
u32 catas_size;
u16 fw_pages;
u8 clr_int_bar;
u8 catas_bar;
+ u8 comm_bar;
+};
+
+struct mlx4_comm {
+ u32 slave_write;
+ u32 slave_read;
+};
+
+struct mlx4_slave_eqe {
+ u8 type;
+ u8 port;
+ u32 param;
+};
+
+struct mlx4_slave_state {
+ u8 comm_toggle;
+ u8 last_cmd;
+ dma_addr_t vhcr_dma;
+};
+
+struct mlx4_mfunc_master_ctx {
+ struct mlx4_slave_state *slave_state;
+};
+
+struct mlx4_vhcr {
+ u64 in_param;
+ u64 out_param;
+ u32 in_modifier;
+ u32 timeout;
+ u16 op;
+ u16 token;
+ u8 op_modifier;
+ int errno;
+};
+
+struct mlx4_mfunc {
+ struct mlx4_comm __iomem *comm;
+ struct workqueue_struct *comm_wq;
+ struct delayed_work comm_work;
+ struct mlx4_vhcr *vhcr;
+ dma_addr_t vhcr_dma;
+
+ struct mlx4_mfunc_master_ctx master;
};
struct mlx4_cmd {
@@ -175,6 +237,7 @@ struct mlx4_cmd {
u16 token_mask;
u8 use_events;
u8 toggle;
+ u8 comm_toggle;
};
struct mlx4_uar_table {
@@ -294,6 +357,7 @@ struct mlx4_priv {
struct mlx4_fw fw;
struct mlx4_cmd cmd;
+ struct mlx4_mfunc mfunc;
struct mlx4_bitmap pd_bitmap;
struct mlx4_uar_table uar_table;
@@ -372,13 +436,19 @@ u64 mlx4_make_profile(struct mlx4_dev *dev,
struct mlx4_profile *request,
struct mlx4_dev_cap *dev_cap,
struct mlx4_init_hca_param *init_hca);
+void mlx4_slave_async_eq_poll(struct work_struct *work);
int mlx4_cmd_init(struct mlx4_dev *dev);
void mlx4_cmd_cleanup(struct mlx4_dev *dev);
+int mlx4_multi_func_init(struct mlx4_dev *dev);
+void mlx4_multi_func_cleanup(struct mlx4_dev *dev);
void mlx4_cmd_event(struct mlx4_dev *dev, u16 token, u8 status, u64 out_param);
int mlx4_cmd_use_events(struct mlx4_dev *dev);
void mlx4_cmd_use_polling(struct mlx4_dev *dev);
+int mlx4_comm_cmd(struct mlx4_dev *dev, u8 cmd, u16 param, unsigned long timeout);
+
+
void mlx4_cq_completion(struct mlx4_dev *dev, u32 cqn);
void mlx4_cq_event(struct mlx4_dev *dev, u32 cqn, int event_type);
diff --git a/include/linux/mlx4/cmd.h b/include/linux/mlx4/cmd.h
index 0f82293..dda1762 100644
--- a/include/linux/mlx4/cmd.h
+++ b/include/linux/mlx4/cmd.h
@@ -117,6 +117,10 @@ enum {
/* miscellaneous commands */
MLX4_CMD_DIAG_RPRT = 0x30,
MLX4_CMD_NOP = 0x31,
+ MLX4_CMD_ACCESS_MEM = 0x2e,
+
+ /* virtual commands */
+ MLX4_CMD_GET_EVENT = 0x52,
/* debug commands */
MLX4_CMD_QUERY_DEBUG_MSG = 0x2a,
@@ -130,7 +134,8 @@ enum {
};
enum {
- MLX4_MAILBOX_SIZE = 4096
+ MLX4_MAILBOX_SIZE = 4096,
+ MLX4_ACCESS_MEM_ALIGN = 256,
};
enum {
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index f35703d..9735f40 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -98,7 +98,8 @@ enum mlx4_event {
MLX4_EVENT_TYPE_PORT_CHANGE = 0x09,
MLX4_EVENT_TYPE_EQ_OVERFLOW = 0x0f,
MLX4_EVENT_TYPE_ECC_DETECT = 0x0e,
- MLX4_EVENT_TYPE_CMD = 0x0a
+ MLX4_EVENT_TYPE_CMD = 0x0a,
+ MLX4_EVENT_TYPE_NONE = 0xff,
};
enum {
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 02/25] mlx4_core: add support for arbitrary bitmap sizes
From: Yevgeny Petrilin @ 2009-11-04 15:30 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
drivers/net/mlx4/alloc.c | 10 ++++++++++
drivers/net/mlx4/mlx4.h | 2 ++
2 files changed, 12 insertions(+), 0 deletions(-)
diff --git a/drivers/net/mlx4/alloc.c b/drivers/net/mlx4/alloc.c
index ad95d5f..bc68cc3 100644
--- a/drivers/net/mlx4/alloc.c
+++ b/drivers/net/mlx4/alloc.c
@@ -177,6 +177,16 @@ int mlx4_bitmap_init(struct mlx4_bitmap *bitmap, u32 num, u32 mask,
return 0;
}
+/* Like bitmap_init, but doesn't require 'num' to be a power of 2 or
+ * a non-trivial mask */
+int mlx4_bitmap_init_no_mask(struct mlx4_bitmap *bitmap, u32 num,
+ u32 reserved_bot, u32 reserved_top)
+{
+ u32 num_rounded = roundup_pow_of_two(num);
+ return mlx4_bitmap_init(bitmap, num_rounded, num_rounded - 1,
+ reserved_bot, num_rounded - num + reserved_top);
+}
+
void mlx4_bitmap_cleanup(struct mlx4_bitmap *bitmap)
{
kfree(bitmap->table);
diff --git a/drivers/net/mlx4/mlx4.h b/drivers/net/mlx4/mlx4.h
index bc72d6e..5836c94 100644
--- a/drivers/net/mlx4/mlx4.h
+++ b/drivers/net/mlx4/mlx4.h
@@ -330,6 +330,8 @@ u32 mlx4_bitmap_alloc_range(struct mlx4_bitmap *bitmap, int cnt, int align);
void mlx4_bitmap_free_range(struct mlx4_bitmap *bitmap, u32 obj, int cnt);
int mlx4_bitmap_init(struct mlx4_bitmap *bitmap, u32 num, u32 mask,
u32 reserved_bot, u32 resetrved_top);
+int mlx4_bitmap_init_no_mask(struct mlx4_bitmap *bitmap, u32 num,
+ u32 reserved_bot, u32 reserved_top);
void mlx4_bitmap_cleanup(struct mlx4_bitmap *bitmap);
int mlx4_reset(struct mlx4_dev *dev);
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 01/25] mlx4_core: identify function as pf or vf
From: Yevgeny Petrilin @ 2009-11-04 15:30 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
---
include/linux/mlx4/device.h | 19 +++++++++++++++++++
1 files changed, 19 insertions(+), 0 deletions(-)
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index ce7cc6c..f35703d 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -42,6 +42,9 @@
enum {
MLX4_FLAG_MSI_X = 1 << 0,
MLX4_FLAG_OLD_PORT_CMDS = 1 << 1,
+ MLX4_FLAG_MASTER = 1 << 2,
+ MLX4_FLAG_SLAVE = 1 << 3,
+ MLX4_FLAG_SRIOV = 1 << 4,
};
enum {
@@ -375,6 +378,7 @@ struct mlx4_av {
struct mlx4_dev {
struct pci_dev *pdev;
unsigned long flags;
+ unsigned long num_slaves;
struct mlx4_caps caps;
struct radix_tree_root qp_table_tree;
u32 rev_id;
@@ -400,6 +404,21 @@ struct mlx4_init_port_param {
if (((type) == MLX4_PORT_TYPE_IB ? (dev)->caps.port_mask : \
~(dev)->caps.port_mask) & 1 << ((port) - 1))
+static inline int mlx4_is_slave(struct mlx4_dev *dev)
+{
+ return dev->flags & MLX4_FLAG_SLAVE;
+}
+
+static inline int mlx4_is_master(struct mlx4_dev *dev)
+{
+ return dev->flags & MLX4_FLAG_MASTER;
+}
+
+static inline int mlx4_is_mfunc(struct mlx4_dev *dev)
+{
+ return dev->flags & (MLX4_FLAG_MASTER | MLX4_FLAG_SLAVE);
+}
+
int mlx4_buf_alloc(struct mlx4_dev *dev, int size, int max_direct,
struct mlx4_buf *buf);
void mlx4_buf_free(struct mlx4_dev *dev, int size, struct mlx4_buf *buf);
--
1.6.1.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [PATCH 00/25] mlx4: multi-function framework and Ethernet sriov.
From: Yevgeny Petrilin @ 2009-11-04 15:30 UTC (permalink / raw)
To: rdreier-FYB4Gu1CFyUAvxtiuMwx3w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
liranl-VPRAkNaXOzVS1MOuV/RT9w, tziporet-VPRAkNaXOzVS1MOuV/RT9w,
yevgenyp-VPRAkNaXOzVS1MOuV/RT9w
In ConnectX, nearly all device resources can be associated with any function.
Once allocated, these resources can be manipulated exclusively by their associated
functions in the same manner as in the single-function case.
The device continues to use a single HCA Command Register (HCR), which is owned by the pf.
Thus, the pf is given the role of allocating resources to vfs, and controlling shared
resource (such as ports).
A dedicated HW communication channel is used to transfer resource allocation and
modification requests from vfs to the pf, and events from the pf to the vfs.
Consequently, no hypervisor-specific services are required for pf-vf communication.
This patch set sets the ground for multi-function operation and enables Ethernet interfaces on vfs.
Signed-off-by: Liran Liss <liranl-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
Signed-off-by: Yevgeny Petrilin <yevgenyp-VPRAkNaXOzVS1MOuV/RT9w@public.gmane.org>
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* [RFC,PATCH] mutex: mutex_is_owner() helper
From: Eric Dumazet @ 2009-11-04 15:25 UTC (permalink / raw)
To: David S. Miller, Ingo Molnar; +Cc: Linux Netdev List, linux kernel
mutex_is_locked() is called most of the time to check if mutex is locked by current
thread. But it's a lazy check, because mutex might be locked by another thread.
Adds a new mutex_is_owned_by() helper, that can check ownership if CONFIG_SMP or
CONFIG_DEBUG_MUTEXES are set.
Returns are
0 if mutex is unlocked.
1 if locked
-1 if not locked by designated thread.
Last return value is possible only if CONFIG_SMP=y or CONFIG_DEBUG_MUTEXES=y
Example of use :
int rtnl_is_locked(void)
{
return mutex_is_locked(&rtnl_mutex);
}
->
int rtnl_is_locked(void)
{
return mutex_is_owned_by(&rtnl_mutex, current_thread_info()) == 1;
}
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
Documentation/mutex-design.txt | 1 +
include/linux/mutex.h | 20 ++++++++++++++++++++
2 files changed, 21 insertions(+)
diff --git a/Documentation/mutex-design.txt b/Documentation/mutex-design.txt
index aa60d1f..521607c 100644
--- a/Documentation/mutex-design.txt
+++ b/Documentation/mutex-design.txt
@@ -133,6 +133,7 @@ the APIs of 'struct mutex' have been streamlined:
int mutex_trylock(struct mutex *lock);
void mutex_unlock(struct mutex *lock);
int mutex_is_locked(struct mutex *lock);
+ int mutex_is_owned_by(struct mutex *lock, struct thread_info *ti);
void mutex_lock_nested(struct mutex *lock, unsigned int subclass);
int mutex_lock_interruptible_nested(struct mutex *lock,
unsigned int subclass);
diff --git a/include/linux/mutex.h b/include/linux/mutex.h
index 878cab4..95a8c5b 100644
--- a/include/linux/mutex.h
+++ b/include/linux/mutex.h
@@ -118,6 +118,26 @@ static inline int mutex_is_locked(struct mutex *lock)
return atomic_read(&lock->count) != 1;
}
+/**
+ * mutex_is_owned_by - check mutex ownership
+ * @lock: the mutex to be queried
+ * @ti: thread_info pointer
+ *
+ * Returns: 0 if mutex is unlocked.
+ * 1 if locked
+ * -1 if not locked by designated thread.
+ */
+static inline int mutex_is_owned_by(struct mutex *lock, struct thread_info *ti)
+{
+ if (atomic_read(&lock->count) == 1)
+ return 0;
+#if defined(CONFIG_DEBUG_MUTEXES) || defined(CONFIG_SMP)
+ if (lock->owner != ti)
+ return -1;
+#endif
+ return 1;
+}
+
/*
* See kernel/mutex.c for detailed documentation of these APIs.
* Also see Documentation/mutex-design.txt.
^ permalink raw reply related
* Re: [announce] new rt2800 drivers for Ralink wireless & project tree
From: John W. Linville @ 2009-11-04 15:15 UTC (permalink / raw)
To: Gertjan van Wingerde
Cc: Bartlomiej Zolnierkiewicz, linux-wireless, linux-kernel, netdev,
Ivo van Doorn, Randy Dunlap, Luis Correia, Ingo Molnar,
Johannes Berg, Jarek Poplawski, Pekka Enberg, David Miller
In-Reply-To: <4AF09A15.3060504@gmail.com>
On Tue, Nov 03, 2009 at 10:01:09PM +0100, Gertjan van Wingerde wrote:
> However, it is good to see that you have put effort in providing a patch series
> for these issues.
Yes, indeed. Thanks to Bart for making patches available!
> Now, since I believe that it is better to work with people, rather than against
> them, would it be possible to post this patch series somewhere as a set of
> separate patches, so that they can be reviewed as such?
I must echo this request.
Thanks!
John
--
John W. Linville Someday the world will need a hero, and you
linville@tuxdriver.com might be all we have. Be ready.
^ 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