* Re: linux-next: build failure after merge of the virtio tree
From: Rusty Russell @ 2012-12-10 22:41 UTC (permalink / raw)
To: Stephen Rothwell
Cc: linux-next, linux-kernel, Jason Wang, David Miller, netdev
In-Reply-To: <20121210133725.cb4a76545691b099c73c40ea@canb.auug.org.au>
Stephen Rothwell <sfr@canb.auug.org.au> writes:
> Hi Rusty,
>
> After merging the virtio tree, today's linux-next build (x86_64
> allmodconfig) failed like this:
>
> drivers/net/virtio_net.c: In function 'vq2txq':
> drivers/net/virtio_net.c:150:2: error: implicit declaration of function 'virtqueue_get_queue_index' [-Werror=implicit-function-declaration]
>
> Caused by commit 986a4f4d452d ("virtio_net: multiqueue support") from the
> net-next tree interacting with commit 105e892960e1 ("virtio: move
> queue_index and num_free fields into core struct virtqueue") from the
> virtio tree.
>
> I applied the patch below and can carry it as necessary.
Thanks for this, your fix is correct.
Cheers,
Rusty.
^ permalink raw reply
* Re: [PATCHv7] virtio-spec: virtio network device multiqueue support
From: Rusty Russell @ 2012-12-10 23:50 UTC (permalink / raw)
To: Michael S. Tsirkin; +Cc: bhutchings, netdev, kvm, virtualization
In-Reply-To: <20121210104025.GA26640@redhat.com>
"Michael S. Tsirkin" <mst@redhat.com> writes:
> Add multiqueue support to virtio network device.
> Add a new feature flag VIRTIO_NET_F_MQ for this feature, a new
> configuration field max_virtqueue_pairs to detect supported number of
> virtqueues as well as a new command VIRTIO_NET_CTRL_MQ to program
> packet steering for unidirectional protocols.
>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Thanks, applied.
Cheers,
Rusty.
^ permalink raw reply
* [PATCH net-next] net: gro: dev_gro_receive() cleanup
From: Eric Dumazet @ 2012-12-10 23:28 UTC (permalink / raw)
To: David Miller; +Cc: netdev
From: Eric Dumazet <edumazet@google.com>
__napi_gro_receive() is inlined from two call sites for no good reason.
Lets move the prep stuff in a function of its own, called only if/when
needed. This saves 300 bytes on x86 :
# size net/core/dev.o.after net/core/dev.o.before
text data bss dec hex filename
51968 1238 1040 54246 d3e6 net/core/dev.o.before
51664 1238 1040 53942 d2b6 net/core/dev.o.after
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
net/core/dev.c | 52 +++++++++++++++++++++++------------------------
1 file changed, 26 insertions(+), 26 deletions(-)
diff --git a/net/core/dev.c b/net/core/dev.c
index a4c4a1b..4783850 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3603,6 +3603,28 @@ void napi_gro_flush(struct napi_struct *napi, bool flush_old)
}
EXPORT_SYMBOL(napi_gro_flush);
+static void gro_list_prepare(struct napi_struct *napi, struct sk_buff *skb)
+{
+ struct sk_buff *p;
+ unsigned int maclen = skb->dev->hard_header_len;
+
+ for (p = napi->gro_list; p; p = p->next) {
+ unsigned long diffs;
+
+ diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev;
+ diffs |= p->vlan_tci ^ skb->vlan_tci;
+ if (maclen == ETH_HLEN)
+ diffs |= compare_ether_header(skb_mac_header(p),
+ skb_gro_mac_header(skb));
+ else if (!diffs)
+ diffs = memcmp(skb_mac_header(p),
+ skb_gro_mac_header(skb),
+ maclen);
+ NAPI_GRO_CB(p)->same_flow = !diffs;
+ NAPI_GRO_CB(p)->flush = 0;
+ }
+}
+
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
{
struct sk_buff **pp = NULL;
@@ -3619,6 +3641,8 @@ static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff
if (skb_is_gso(skb) || skb_has_frag_list(skb))
goto normal;
+ gro_list_prepare(napi, skb);
+
rcu_read_lock();
list_for_each_entry_rcu(ptype, head, list) {
if (ptype->type != type || !ptype->callbacks.gro_receive)
@@ -3695,30 +3719,6 @@ normal:
goto pull;
}
-static inline gro_result_t
-__napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
-{
- struct sk_buff *p;
- unsigned int maclen = skb->dev->hard_header_len;
-
- for (p = napi->gro_list; p; p = p->next) {
- unsigned long diffs;
-
- diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev;
- diffs |= p->vlan_tci ^ skb->vlan_tci;
- if (maclen == ETH_HLEN)
- diffs |= compare_ether_header(skb_mac_header(p),
- skb_gro_mac_header(skb));
- else if (!diffs)
- diffs = memcmp(skb_mac_header(p),
- skb_gro_mac_header(skb),
- maclen);
- NAPI_GRO_CB(p)->same_flow = !diffs;
- NAPI_GRO_CB(p)->flush = 0;
- }
-
- return dev_gro_receive(napi, skb);
-}
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
{
@@ -3768,7 +3768,7 @@ gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
{
skb_gro_reset_offset(skb);
- return napi_skb_finish(__napi_gro_receive(napi, skb), skb);
+ return napi_skb_finish(dev_gro_receive(napi, skb), skb);
}
EXPORT_SYMBOL(napi_gro_receive);
@@ -3866,7 +3866,7 @@ gro_result_t napi_gro_frags(struct napi_struct *napi)
if (!skb)
return GRO_DROP;
- return napi_frags_finish(napi, skb, __napi_gro_receive(napi, skb));
+ return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
}
EXPORT_SYMBOL(napi_gro_frags);
^ permalink raw reply related
* Re: [PATCH] smsc75xx: only set mac address once on bind
From: Dan Williams @ 2012-12-10 23:00 UTC (permalink / raw)
To: Steve Glendinning; +Cc: netdev, Bjorn Mork
In-Reply-To: <1355137279-2695-1-git-send-email-steve.glendinning@shawell.net>
On Mon, 2012-12-10 at 11:01 +0000, Steve Glendinning wrote:
> This patch changes when we decide what the device's MAC address
> is from per ifconfig up to once when the device is connected.
>
> Without this patch, a manually forced device MAC is overwritten
> on ifconfig down/up. Also devices that have no EEPROM are
> assigned a new random address on ifconfig down/up instead of
> persisting the same one.
Does this mean that on devices without EEPROM, ifconfig XXX
down/ifconfig XXX up will generate a *new* random address? That seems a
bit odd; why wouldn't the first random address generated for the device
persist until either (a) changed by ifconfig or (b) device was
disconnected?
Dan
> Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
> Reported-by: Robert Cunningham <rcunningham@nsmsurveillance.com>
> Cc: Bjorn Mork <bjorn@mork.no>
> Cc: Dan Williams <dcbw@redhat.com>
> ---
> drivers/net/usb/smsc75xx.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/usb/smsc75xx.c b/drivers/net/usb/smsc75xx.c
> index 1cbd936..251a335 100644
> --- a/drivers/net/usb/smsc75xx.c
> +++ b/drivers/net/usb/smsc75xx.c
> @@ -1054,8 +1054,6 @@ static int smsc75xx_reset(struct usbnet *dev)
>
> netif_dbg(dev, ifup, dev->net, "PHY reset complete\n");
>
> - smsc75xx_init_mac_address(dev);
> -
> ret = smsc75xx_set_mac_address(dev);
> if (ret < 0) {
> netdev_warn(dev->net, "Failed to set mac address\n");
> @@ -1422,6 +1420,14 @@ static int smsc75xx_bind(struct usbnet *dev, struct usb_interface *intf)
> dev->net->hw_features = NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
> NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 | NETIF_F_RXCSUM;
>
> + ret = smsc75xx_wait_ready(dev, 0);
> + if (ret < 0) {
> + netdev_warn(dev->net, "device not ready in smsc75xx_bind\n");
> + return ret;
> + }
> +
> + smsc75xx_init_mac_address(dev);
> +
> /* Init all registers */
> ret = smsc75xx_reset(dev);
> if (ret < 0) {
^ permalink raw reply
* RE: [PATCH net-next] bnx2x: use netdev_alloc_frag()
From: Dmitry Kravkov @ 2012-12-10 22:57 UTC (permalink / raw)
To: Eric Dumazet, David Miller; +Cc: netdev, Eilon Greenstein
In-Reply-To: <1355177766.27891.68.camel@edumazet-glaptop>
> -----Original Message-----
> From: Eric Dumazet [mailto:eric.dumazet@gmail.com]
> Sent: Tuesday, December 11, 2012 12:16 AM
> To: David Miller
> Cc: netdev; Eilon Greenstein; Dmitry Kravkov
> Subject: [PATCH net-next] bnx2x: use netdev_alloc_frag()
>
> From: Eric Dumazet <edumazet@google.com>
>
> Using netdev_alloc_frag() instead of kmalloc() permits better GRO or
> TCP coalescing behavior, as skb_gro_receive() doesn't have to fallback
> to frag_list overhead.
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Dmitry Kravkov <dmitry@broadcom.com>
> Cc: Eilon Greenstein <eilong@broadcom.com>
Thanks a lot, Eric!
Acked-by: Dmitry Kravkov <dmitry@broadcom.com>
^ permalink raw reply
* RE: [PATCH] net: fix a race in gro_cell_poll()
From: Dmitry Kravkov @ 2012-12-10 22:46 UTC (permalink / raw)
To: Eric Dumazet, David Miller; +Cc: netdev@vger.kernel.org
In-Reply-To: <1355178723.27891.85.camel@edumazet-glaptop>
> -----Original Message-----
> From: Eric Dumazet [mailto:eric.dumazet@gmail.com]
> Sent: Tuesday, December 11, 2012 12:32 AM
> To: Dmitry Kravkov; David Miller
> Cc: netdev@vger.kernel.org
> Subject: [PATCH] net: fix a race in gro_cell_poll()
>
> From: Eric Dumazet <edumazet@google.com>
>
> Dmitry Kravkov reported packet drops for GRE packets since GRO support
> was added.
>
> There is a race in gro_cell_poll() because we call napi_complete()
> without any synchronization with a concurrent gro_cells_receive()
>
> Once bug was triggered, we queued packets but did not schedule NAPI
> poll.
>
> We can fix this issue using the spinlock protected the napi_skbs queue,
> as we have to hold it to perform skb dequeue anyway.
>
> As we open-code skb_dequeue(), we no longer need to mask IRQS, as both
> producer and consumer run under BH context.
>
> Bug added in commit c9e6bc644e (net: add gro_cells infrastructure)
>
> Reported-by: Dmitry Kravkov <dmitry@broadcom.com>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> ---
>
> David: I could reproduce the bug Dmitry reported, and have
> verified this patch fixes the issue.
>
> include/net/gro_cells.h | 14 +++++++++-----
> 1 file changed, 9 insertions(+), 5 deletions(-)
>
> diff --git a/include/net/gro_cells.h b/include/net/gro_cells.h
> index 4fd8a4b..e5062c9 100644
> --- a/include/net/gro_cells.h
> +++ b/include/net/gro_cells.h
> @@ -17,7 +17,6 @@ struct gro_cells {
>
> static inline void gro_cells_receive(struct gro_cells *gcells, struct sk_buff *skb)
> {
> - unsigned long flags;
> struct gro_cell *cell = gcells->cells;
> struct net_device *dev = skb->dev;
>
> @@ -35,32 +34,37 @@ static inline void gro_cells_receive(struct gro_cells
> *gcells, struct sk_buff *s
> return;
> }
>
> - spin_lock_irqsave(&cell->napi_skbs.lock, flags);
> + /* We run in BH context */
> + spin_lock(&cell->napi_skbs.lock);
>
> __skb_queue_tail(&cell->napi_skbs, skb);
> if (skb_queue_len(&cell->napi_skbs) == 1)
> napi_schedule(&cell->napi);
>
> - spin_unlock_irqrestore(&cell->napi_skbs.lock, flags);
> + spin_unlock(&cell->napi_skbs.lock);
> }
>
> +/* called unser BH context */
> static inline int gro_cell_poll(struct napi_struct *napi, int budget)
> {
> struct gro_cell *cell = container_of(napi, struct gro_cell, napi);
> struct sk_buff *skb;
> int work_done = 0;
>
> + spin_lock(&cell->napi_skbs.lock);
> while (work_done < budget) {
> - skb = skb_dequeue(&cell->napi_skbs);
> + skb = __skb_dequeue(&cell->napi_skbs);
> if (!skb)
> break;
> -
> + spin_unlock(&cell->napi_skbs.lock);
> napi_gro_receive(napi, skb);
> work_done++;
> + spin_lock(&cell->napi_skbs.lock);
> }
>
> if (work_done < budget)
> napi_complete(napi);
> + spin_unlock(&cell->napi_skbs.lock);
> return work_done;
> }
>
>
My scenario is working, now
Thanks Eric.
Tested-by: Dmitry Kravkov <dmitry@broadcom.com>
^ permalink raw reply
* Re: [RFC PATCH v2 3/3] tun: fix LSM/SELinux labeling of tun/tap devices
From: Paul Moore @ 2012-12-10 22:43 UTC (permalink / raw)
To: Michael S. Tsirkin; +Cc: netdev, linux-security-module, selinux, jasowang
In-Reply-To: <20121210175035.GA31856@redhat.com>
On Monday, December 10, 2012 07:50:35 PM Michael S. Tsirkin wrote:
> On Mon, Dec 10, 2012 at 12:33:49PM -0500, Paul Moore wrote:
> > On Monday, December 10, 2012 07:26:56 PM Michael S. Tsirkin wrote:
> > > On Mon, Dec 10, 2012 at 12:04:35PM -0500, Paul Moore wrote:
> > > > On Friday, December 07, 2012 02:25:16 PM Michael S. Tsirkin wrote:
> > > > > On Thu, Dec 06, 2012 at 04:09:51PM -0500, Paul Moore wrote:
> > > > > > On Thursday, December 06, 2012 10:57:16 PM Michael S. Tsirkin
wrote:
> > > > > > > On Thu, Dec 06, 2012 at 11:56:45AM -0500, Paul Moore wrote:
> > > > > > > > The SETQUEUE/tun_socket:create_queue permissions do not yet
> > > > > > > > exist
> > > > > > > > in any released SELinux policy as we are just now adding them
> > > > > > > > with
> > > > > > > > this patchset. With current policies loaded into a kernel with
> > > > > > > > this patchset applied the SETQUEUE/tun_socket:create_queue
> > > > > > > > permission would be treated according to the policy's unknown
> > > > > > > > permission setting.
> > > > > > >
> > > > > > > OK I think we need to rethink what we are doing here: what you
> > > > > > > sent
> > > > > > > addresses the problem as stated but I think we mis-stated it.
> > > > > > > Let
> > > > > > > me try to restate the problem: it is not just selinux problem.
> > > > > > > Let's
> > > > > > > assume qemu wants to use tun, I (libvirt) don't want to run it
> > > > > > > as
> > > > > > > root.
> > > > > > >
> > > > > > > 1. TUNSETIFF: I can open tun, attach an fd and pass it to qemu.
> > > > > > > Now, qemu does not invoke TUNSETIFF so it can run without
> > > > > > > kernel priveledges.
> > > > > >
> > > > > > Correct me if I'm wrong, but I believe libvirt does this while
> > > > > > running
> > > > > > as root. Assuming that is the case, why not simply
> > > > > > setuid()/setgid()
> > > > > > to the same credentials as the QEMU instance before creating the
> > > > > > TUN
> > > > > > device? You can always (re)configure the device afterwards while
> > > > > > running as root/CAP_NET_ADMIN.
> > > > >
> > > > > We want isolation between qemu instances.
> > > >
> > > > Understood, I agree.
> > > >
> > > > Achieving separation via SELinux is easily done, with libvirt/sVirt
> > > > already doing this for us automatically in most cases; the only thing
> > > > we
> > > > will want to do is make sure the SELinux policy is aware of the new
> > > > permission.
> > > >
> > > > Achieving separation via DAC should also be easily done, simply run
> > > > each
> > > > QEMU instance with a separate UID and/or GID.
> > > >
> > > > > Giving qemu right to open tun and SETIFF would give it rights
> > > > > to access any tun device.
> > > >
> > > > I'm quickly looked at tun_chr_open() again and I don't see any special
> > > > rights/privileges required, the same for tun_chr_ioctl() and
> > > > __tun_chr_ioctl(). Looking at tun_set_queue() I see we call
> > > > tun_not_capable() which does a simple DAC check; it must have the same
> > > > UID/GID or have CAP_NET_ADMIN.
> > > >
> > > > I'm having a hard time seeing the problem you are describing; help me
> > > > understand.
> > >
> > > The issue is guest controls the number of queues in use.
> > > So qemu would be required to be allowed to call tun_set_queue.
> > > If we allow this we have a problem as one qemu will be
> > > able to access any tun.
> >
> > QEMU can call tun_set_queue() as long as it satisfies tun_not_capable(),
> > which from a practical point of view means that the TUN device was
> > created with the same UID/GID as the QEMU instance. If you want TUN
> > device separation between QEMU instances using DAC you need to run each
> > QEMU instance with a different UID/GID (which you should be doing anyway
> > if you want DAC enforced general separation).
> >
> > I believe I've stated this point several times now and I don't feel you've
> > addressed it properly.
>
> Look at how it works at the moment:
> a priveledged libvirt server calls tun_set_iff
> and passes the fd to qemu which is not priveledged.
>
> The result is isolation between qemu instances without
> need to create uid per qemu instance.
Okay, good. That is my understanding.
> How do we create multiple queues? It makes sense to
> follow this model and pass in fds for individual queues.
Okay.
> However they need to be disabled initially
> so libvirt can not do tun_set_queue for us.
Unrelated question: why do the queues need to be disabled initially? Is this
to prevent traffic from being queued up? Some other reason? I'm just curious
as to the reason ...
> When qemu later calls tun_set_queue it will fail which means we
> can't utilize multiqueue.
I still don't understand why in the multiqueue case libvirt doesn't just
change it's effective UID/GID when creating the TUN device, or just use the
TUNSETOWNER/TUNSETGROUP commands. This would solve the problem you describe
above and - at least to me - seems like a better solution conceptually.
Help me understand why you believe that will not work.
Do you not want to give ownership of the TUN device to QEMU? That would be
the only reason I can think of, but all of your comments that I can recall
have been about isolation between QEMU instances and not access control
between a QEMU instance and its assigned TUN device.
> My solution is an unpriveledged variant
> of tun_set_queue that only enables/disables
> a queue without attach/detach.
--
paul moore
security and virtualization @ redhat
^ permalink raw reply
* [PATCH] net: fix a race in gro_cell_poll()
From: Eric Dumazet @ 2012-12-10 22:32 UTC (permalink / raw)
To: Dmitry Kravkov, David Miller; +Cc: netdev@vger.kernel.org
In-Reply-To: <1355176551.27891.57.camel@edumazet-glaptop>
From: Eric Dumazet <edumazet@google.com>
Dmitry Kravkov reported packet drops for GRE packets since GRO support
was added.
There is a race in gro_cell_poll() because we call napi_complete()
without any synchronization with a concurrent gro_cells_receive()
Once bug was triggered, we queued packets but did not schedule NAPI
poll.
We can fix this issue using the spinlock protected the napi_skbs queue,
as we have to hold it to perform skb dequeue anyway.
As we open-code skb_dequeue(), we no longer need to mask IRQS, as both
producer and consumer run under BH context.
Bug added in commit c9e6bc644e (net: add gro_cells infrastructure)
Reported-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
David: I could reproduce the bug Dmitry reported, and have
verified this patch fixes the issue.
include/net/gro_cells.h | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)
diff --git a/include/net/gro_cells.h b/include/net/gro_cells.h
index 4fd8a4b..e5062c9 100644
--- a/include/net/gro_cells.h
+++ b/include/net/gro_cells.h
@@ -17,7 +17,6 @@ struct gro_cells {
static inline void gro_cells_receive(struct gro_cells *gcells, struct sk_buff *skb)
{
- unsigned long flags;
struct gro_cell *cell = gcells->cells;
struct net_device *dev = skb->dev;
@@ -35,32 +34,37 @@ static inline void gro_cells_receive(struct gro_cells *gcells, struct sk_buff *s
return;
}
- spin_lock_irqsave(&cell->napi_skbs.lock, flags);
+ /* We run in BH context */
+ spin_lock(&cell->napi_skbs.lock);
__skb_queue_tail(&cell->napi_skbs, skb);
if (skb_queue_len(&cell->napi_skbs) == 1)
napi_schedule(&cell->napi);
- spin_unlock_irqrestore(&cell->napi_skbs.lock, flags);
+ spin_unlock(&cell->napi_skbs.lock);
}
+/* called unser BH context */
static inline int gro_cell_poll(struct napi_struct *napi, int budget)
{
struct gro_cell *cell = container_of(napi, struct gro_cell, napi);
struct sk_buff *skb;
int work_done = 0;
+ spin_lock(&cell->napi_skbs.lock);
while (work_done < budget) {
- skb = skb_dequeue(&cell->napi_skbs);
+ skb = __skb_dequeue(&cell->napi_skbs);
if (!skb)
break;
-
+ spin_unlock(&cell->napi_skbs.lock);
napi_gro_receive(napi, skb);
work_done++;
+ spin_lock(&cell->napi_skbs.lock);
}
if (work_done < budget)
napi_complete(napi);
+ spin_unlock(&cell->napi_skbs.lock);
return work_done;
}
^ permalink raw reply related
* [GIT] Networking
From: David Miller @ 2012-12-10 22:31 UTC (permalink / raw)
To: torvalds; +Cc: akpm, netdev, linux-kernel
1) Netlink socket dumping had several missing verifications and
checks.
In particular, address comparisons in the request byte code
interpreter could access past the end of the address in the
inet_request_sock.
Also, address family and address prefix lengths were not
validated properly at all.
This means arbitrary applications can read past the end of
certain kernel data structures.
Fixes from Neal Cardwell.
2) ip_check_defrag() operates in contexts where we're in the
process of, or about to, input the packet into the real
protocols (specifically macvlan and AF_PACKET snooping).
Unfortunately, it does a pskb_may_pull() which can modify
the backing packet data which is not legal if the SKB is
shared. It very much can be shared in this context.
Deal with the possibility that the SKB is segmented by
using skb_copy_bits().
Fix from Johannes Berg based upon a report by Eric Leblond.
Please pull, thanks a lot!
The following changes since commit ed23ec4f0a510528e0ffe415f9394107418ae854:
mm: vmscan: fix inappropriate zone congestion clearing (2012-12-08 08:41:18 -0800)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git master
for you to fetch changes up to 1bf3751ec90cc3174e01f0d701e8449ce163d113:
ipv4: ip_check_defrag must not modify skb before unsharing (2012-12-10 13:51:44 -0500)
----------------------------------------------------------------
Johannes Berg (1):
ipv4: ip_check_defrag must not modify skb before unsharing
Neal Cardwell (4):
inet_diag: fix oops for IPv4 AF_INET6 TCP SYN-RECV state
inet_diag: validate byte code to prevent oops in inet_diag_bc_run()
inet_diag: avoid unsafe and nonsensical prefix matches in inet_diag_bc_run()
inet_diag: validate port comparison byte code to prevent unsafe reads
net/ipv4/inet_diag.c | 154 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---------------------
net/ipv4/ip_fragment.c | 19 ++++++-------
2 files changed, 131 insertions(+), 42 deletions(-)
^ permalink raw reply
* Re: [RFC PATCH v2 3/3] tun: fix LSM/SELinux labeling of tun/tap devices
From: Paul Moore @ 2012-12-10 22:21 UTC (permalink / raw)
To: Eric Paris
Cc: Michael S. Tsirkin, Linux Netdev List, LSM List, SE-Linux,
jasowang
In-Reply-To: <CACLa4pu6UCpzKfscVoEPzLySHitta1yTqPa7cA0d=xUj5ws6HA@mail.gmail.com>
On Monday, December 10, 2012 01:42:12 PM Eric Paris wrote:
> Let me abstract a little here Paul. Lets say user A starts an
> unclassified process and a top secret process. SELinux policy darn
> well better be able to enforce that they can not attach to the same
> tun.
>
> Am I missing something here?
Relax, all the SELinux enforced separation still exists, and works. We're
just fixing the LSM/SELinux stuff that was broken with the multiqueue addition
and adding a new SELinux permission to control access to the new queue
command.
What we are currently discussing is DAC only. While Michael have different
opinions on how to solve the DAC issues, we agree that SELinux works
correctly.
--
paul moore
security and virtualization @ redhat
^ permalink raw reply
* [PATCH net-next] bnx2x: use netdev_alloc_frag()
From: Eric Dumazet @ 2012-12-10 22:16 UTC (permalink / raw)
To: David Miller; +Cc: netdev, Eilon Greenstein, Dmitry Kravkov
From: Eric Dumazet <edumazet@google.com>
Using netdev_alloc_frag() instead of kmalloc() permits better GRO or
TCP coalescing behavior, as skb_gro_receive() doesn't have to fallback
to frag_list overhead.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Dmitry Kravkov <dmitry@broadcom.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
---
drivers/net/ethernet/broadcom/bnx2x/bnx2x.h | 2
drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c | 43 ++++++++++----
2 files changed, 32 insertions(+), 13 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x.h b/drivers/net/ethernet/broadcom/bnx2x/bnx2x.h
index 9a3b81e..e8d4db1 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x.h
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x.h
@@ -489,7 +489,7 @@ struct bnx2x_fastpath {
u32 ustorm_rx_prods_offset;
u32 rx_buf_size;
-
+ u32 rx_frag_size; /* 0 if kmalloced(), or rx_buf_size + NET_SKB_PAD */
dma_addr_t status_blk_mapping;
enum bnx2x_tpa_mode_t mode;
diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
index 67baddd..c9f8ae0 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
@@ -552,6 +552,23 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
return 0;
}
+static void bnx2x_frag_free(const struct bnx2x_fastpath *fp, void *data)
+{
+ if (fp->rx_frag_size)
+ put_page(virt_to_head_page(data));
+ else
+ kfree(data);
+}
+
+static void *bnx2x_frag_alloc(const struct bnx2x_fastpath *fp)
+{
+ if (fp->rx_frag_size)
+ return netdev_alloc_frag(fp->rx_frag_size);
+
+ return kmalloc(fp->rx_buf_size + NET_SKB_PAD, GFP_ATOMIC);
+}
+
+
static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
struct bnx2x_agg_info *tpa_info,
u16 pages,
@@ -574,15 +591,14 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
goto drop;
/* Try to allocate the new data */
- new_data = kmalloc(fp->rx_buf_size + NET_SKB_PAD, GFP_ATOMIC);
-
+ new_data = bnx2x_frag_alloc(fp);
/* Unmap skb in the pool anyway, as we are going to change
pool entry status to BNX2X_TPA_STOP even if new skb allocation
fails. */
dma_unmap_single(&bp->pdev->dev, dma_unmap_addr(rx_buf, mapping),
fp->rx_buf_size, DMA_FROM_DEVICE);
if (likely(new_data))
- skb = build_skb(data, 0);
+ skb = build_skb(data, fp->rx_frag_size);
if (likely(skb)) {
#ifdef BNX2X_STOP_ON_ERROR
@@ -619,7 +635,7 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
return;
}
- kfree(new_data);
+ bnx2x_frag_free(fp, new_data);
drop:
/* drop the packet and keep the buffer in the bin */
DP(NETIF_MSG_RX_STATUS,
@@ -635,7 +651,7 @@ static int bnx2x_alloc_rx_data(struct bnx2x *bp,
struct eth_rx_bd *rx_bd = &fp->rx_desc_ring[index];
dma_addr_t mapping;
- data = kmalloc(fp->rx_buf_size + NET_SKB_PAD, GFP_ATOMIC);
+ data = bnx2x_frag_alloc(fp);
if (unlikely(data == NULL))
return -ENOMEM;
@@ -643,7 +659,7 @@ static int bnx2x_alloc_rx_data(struct bnx2x *bp,
fp->rx_buf_size,
DMA_FROM_DEVICE);
if (unlikely(dma_mapping_error(&bp->pdev->dev, mapping))) {
- kfree(data);
+ bnx2x_frag_free(fp, data);
BNX2X_ERR("Can't map rx data\n");
return -ENOMEM;
}
@@ -845,9 +861,9 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
dma_unmap_addr(rx_buf, mapping),
fp->rx_buf_size,
DMA_FROM_DEVICE);
- skb = build_skb(data, 0);
+ skb = build_skb(data, fp->rx_frag_size);
if (unlikely(!skb)) {
- kfree(data);
+ bnx2x_frag_free(fp, data);
bnx2x_fp_qstats(bp, fp)->
rx_skb_alloc_failed++;
goto next_rx;
@@ -1145,7 +1161,7 @@ static void bnx2x_free_tpa_pool(struct bnx2x *bp,
dma_unmap_single(&bp->pdev->dev,
dma_unmap_addr(first_buf, mapping),
fp->rx_buf_size, DMA_FROM_DEVICE);
- kfree(data);
+ bnx2x_frag_free(fp, data);
first_buf->data = NULL;
}
}
@@ -1190,8 +1206,7 @@ void bnx2x_init_rx_rings(struct bnx2x *bp)
struct sw_rx_bd *first_buf =
&tpa_info->first_buf;
- first_buf->data = kmalloc(fp->rx_buf_size + NET_SKB_PAD,
- GFP_ATOMIC);
+ first_buf->data = bnx2x_frag_alloc(fp);
if (!first_buf->data) {
BNX2X_ERR("Failed to allocate TPA skb pool for queue[%d] - disabling TPA on this queue!\n",
j);
@@ -1323,7 +1338,7 @@ static void bnx2x_free_rx_bds(struct bnx2x_fastpath *fp)
fp->rx_buf_size, DMA_FROM_DEVICE);
rx_buf->data = NULL;
- kfree(data);
+ bnx2x_frag_free(fp, data);
}
}
@@ -1782,6 +1797,10 @@ static void bnx2x_set_rx_buf_size(struct bnx2x *bp)
mtu +
BNX2X_FW_RX_ALIGN_END;
/* Note : rx_buf_size doesnt take into account NET_SKB_PAD */
+ if (fp->rx_buf_size + NET_SKB_PAD <= PAGE_SIZE)
+ fp->rx_frag_size = fp->rx_buf_size + NET_SKB_PAD;
+ else
+ fp->rx_frag_size = 0;
}
}
^ permalink raw reply related
* Re: [PATCH net-next] doc: Tighten-up and clarify description of tcp_fin_timeout
From: David Miller @ 2012-12-10 22:14 UTC (permalink / raw)
To: raj; +Cc: netdev
In-Reply-To: <20121210213300.9D6BB29003EA@tardy>
From: raj@tardy.usa.hp.com (Rick Jones)
Date: Mon, 10 Dec 2012 13:33:00 -0800 (PST)
> From: Rick Jones <rick.jones2@hp.com>
>
> The description for tcp_fin_timeout should be tigher and more clear.
>
> In addition to being tighter, we should make the spelling of the
> state name consistent with what utilities report, remove the now
> dated reference to 2.2 and put the default in the consistent place.
>
> Signed-off-by: Rick Jones <rick.jones2@hp.com>
Applied, thanks Rick.
^ permalink raw reply
* Re: [net-next 1/7] bna: Code Cleanup and Enhancements
From: David Miller @ 2012-12-10 22:13 UTC (permalink / raw)
To: rmody; +Cc: netdev, bhutchings, David.Laight, adapter_linux_open_src_team
In-Reply-To: <1355175725-19202-2-git-send-email-rmody@brocade.com>
From: Rasesh Mody <rmody@brocade.com>
Date: Mon, 10 Dec 2012 13:41:59 -0800
> - skb = unmap_array[unmap_cons].skb;
> - BUG_ON(!(skb));
> - unmap_array[unmap_cons].skb = NULL;
> + curr_ua = &unmap_array[unmap_cons];
> +
> + skb = curr_ua->skb;
> + BUG_ON(!(skb));\
^^^
Really?
Please carefully review your own work before submitting it for
inclusion. When you have erroneous things like this in the
very first patch it reflects very poorly upon the quality of
your work.
^ permalink raw reply
* RE: ipgre rss is broken since gro
From: Eric Dumazet @ 2012-12-10 21:55 UTC (permalink / raw)
To: Dmitry Kravkov; +Cc: netdev@vger.kernel.org
In-Reply-To: <504C9EFCA2D0054393414C9CB605C37F1BFC32A1@SJEXCHMB06.corp.ad.broadcom.com>
On Mon, 2012-12-10 at 19:20 +0000, Dmitry Kravkov wrote:
> Yep, this resolved the issue - Interface is functional after 3 and 100 TCP connections. Thanks
>
>
I guess its only lowering probability of the race.
Could you try instead the following patch ?
I'll address the queue_mapping separately for net-next
Thanks
diff --git a/include/net/gro_cells.h b/include/net/gro_cells.h
index 4fd8a4b..e5062c9 100644
--- a/include/net/gro_cells.h
+++ b/include/net/gro_cells.h
@@ -17,7 +17,6 @@ struct gro_cells {
static inline void gro_cells_receive(struct gro_cells *gcells, struct sk_buff *skb)
{
- unsigned long flags;
struct gro_cell *cell = gcells->cells;
struct net_device *dev = skb->dev;
@@ -35,32 +34,37 @@ static inline void gro_cells_receive(struct gro_cells *gcells, struct sk_buff *s
return;
}
- spin_lock_irqsave(&cell->napi_skbs.lock, flags);
+ /* We run in BH context */
+ spin_lock(&cell->napi_skbs.lock);
__skb_queue_tail(&cell->napi_skbs, skb);
if (skb_queue_len(&cell->napi_skbs) == 1)
napi_schedule(&cell->napi);
- spin_unlock_irqrestore(&cell->napi_skbs.lock, flags);
+ spin_unlock(&cell->napi_skbs.lock);
}
+/* called unser BH context */
static inline int gro_cell_poll(struct napi_struct *napi, int budget)
{
struct gro_cell *cell = container_of(napi, struct gro_cell, napi);
struct sk_buff *skb;
int work_done = 0;
+ spin_lock(&cell->napi_skbs.lock);
while (work_done < budget) {
- skb = skb_dequeue(&cell->napi_skbs);
+ skb = __skb_dequeue(&cell->napi_skbs);
if (!skb)
break;
-
+ spin_unlock(&cell->napi_skbs.lock);
napi_gro_receive(napi, skb);
work_done++;
+ spin_lock(&cell->napi_skbs.lock);
}
if (work_done < budget)
napi_complete(napi);
+ spin_unlock(&cell->napi_skbs.lock);
return work_done;
}
^ permalink raw reply related
* [net-next 6/7] bna: Firmware update
From: Rasesh Mody @ 2012-12-10 21:42 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
In-Reply-To: <1355175725-19202-1-git-send-email-rmody@brocade.com>
Change Details:
- Added Stats clear counter to the bfi_enet_stats_mac structure and
ethtool stats
- Modified the firmware naming convention to contain the firmware image
version (3.1.0.0). The new convention is
<firmware-image>-<firmware-version>.bin The change will enforce loading
only compatible firmware with this driver and also avoid over-writing
the old firmware image in-order to load new version driver as the
firmware names used to be the same.
Signed-off-by: Rasesh Mody <rmody@brocade.com>
---
drivers/net/ethernet/brocade/bna/bfi_enet.h | 1 +
drivers/net/ethernet/brocade/bna/bnad_ethtool.c | 1 +
drivers/net/ethernet/brocade/bna/cna.h | 4 ++--
3 files changed, 4 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/brocade/bna/bfi_enet.h b/drivers/net/ethernet/brocade/bna/bfi_enet.h
index eef6e1f..7d10e33 100644
--- a/drivers/net/ethernet/brocade/bna/bfi_enet.h
+++ b/drivers/net/ethernet/brocade/bna/bfi_enet.h
@@ -787,6 +787,7 @@ struct bfi_enet_stats_bpc {
/* MAC Rx Statistics */
struct bfi_enet_stats_mac {
+ u64 stats_clr_cnt; /* times this stats cleared */
u64 frame_64; /* both rx and tx counter */
u64 frame_65_127; /* both rx and tx counter */
u64 frame_128_255; /* both rx and tx counter */
diff --git a/drivers/net/ethernet/brocade/bna/bnad_ethtool.c b/drivers/net/ethernet/brocade/bna/bnad_ethtool.c
index 40e1e84..455b5a2 100644
--- a/drivers/net/ethernet/brocade/bna/bnad_ethtool.c
+++ b/drivers/net/ethernet/brocade/bna/bnad_ethtool.c
@@ -102,6 +102,7 @@ static const char *bnad_net_stats_strings[BNAD_ETHTOOL_STATS_NUM] = {
"rx_unmap_q_alloc_failed",
"rxbuf_alloc_failed",
+ "mac_stats_clr_cnt",
"mac_frame_64",
"mac_frame_65_127",
"mac_frame_128_255",
diff --git a/drivers/net/ethernet/brocade/bna/cna.h b/drivers/net/ethernet/brocade/bna/cna.h
index 32e8f17..14ca931 100644
--- a/drivers/net/ethernet/brocade/bna/cna.h
+++ b/drivers/net/ethernet/brocade/bna/cna.h
@@ -37,8 +37,8 @@
extern char bfa_version[];
-#define CNA_FW_FILE_CT "ctfw.bin"
-#define CNA_FW_FILE_CT2 "ct2fw.bin"
+#define CNA_FW_FILE_CT "ctfw-3.1.0.0.bin"
+#define CNA_FW_FILE_CT2 "ct2fw-3.1.0.0.bin"
#define FC_SYMNAME_MAX 256 /*!< max name server symbolic name size */
#pragma pack(1)
--
1.7.1
^ permalink raw reply related
* [net-next 4/7] bna: Rx Page Based Allocation
From: Rasesh Mody @ 2012-12-10 21:42 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
In-Reply-To: <1355175725-19202-1-git-send-email-rmody@brocade.com>
Change Details:
Enhanced support for GRO. Page-base allocation method for Rx buffers is
used in GRO. Skb allocation has been removed in Rx path to use always warm-cache
skbs provided by napi_get_frags.
Signed-off-by: Rasesh Mody <rmody@brocade.com>
---
drivers/net/ethernet/brocade/bna/bnad.c | 318 ++++++++++++++++++++++++------
drivers/net/ethernet/brocade/bna/bnad.h | 19 ++
2 files changed, 273 insertions(+), 64 deletions(-)
diff --git a/drivers/net/ethernet/brocade/bna/bnad.c b/drivers/net/ethernet/brocade/bna/bnad.c
index da5470a..22d52ef 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.c
+++ b/drivers/net/ethernet/brocade/bna/bnad.c
@@ -266,53 +266,181 @@ bnad_msix_tx(int irq, void *data)
return IRQ_HANDLED;
}
+static inline void
+bnad_rxq_alloc_uninit(struct bnad *bnad, struct bna_rcb *rcb)
+{
+ struct bnad_rx_unmap_q *unmap_q = rcb->unmap_q;
+
+ unmap_q->reuse_pi = -1;
+ unmap_q->alloc_order = -1;
+ unmap_q->map_size = 0;
+ unmap_q->type = BNAD_RXBUF_NONE;
+}
+
+/* Default is page-based allocation. Multi-buffer support - TBD */
+static int
+bnad_rxq_alloc_init(struct bnad *bnad, struct bna_rcb *rcb)
+{
+ struct bnad_rx_unmap_q *unmap_q = rcb->unmap_q;
+ int mtu, order;
+
+ bnad_rxq_alloc_uninit(bnad, rcb);
+
+ mtu = bna_enet_mtu_get(&bnad->bna.enet);
+ order = get_order(mtu);
+
+ if (bna_is_small_rxq(rcb->id)) {
+ unmap_q->alloc_order = 0;
+ unmap_q->map_size = rcb->rxq->buffer_size;
+ } else {
+ unmap_q->alloc_order = order;
+ unmap_q->map_size =
+ (rcb->rxq->buffer_size > 2048) ?
+ PAGE_SIZE << order : 2048;
+ }
+
+ BUG_ON(((PAGE_SIZE << order) % unmap_q->map_size));
+
+ unmap_q->type = BNAD_RXBUF_PAGE;
+
+ return 0;
+}
+
+static inline void
+bnad_rxq_cleanup_page(struct bnad *bnad, struct bnad_rx_unmap *unmap)
+{
+ if (!unmap->page)
+ return;
+
+ dma_unmap_page(&bnad->pcidev->dev,
+ dma_unmap_addr(&unmap->vector, dma_addr),
+ unmap->vector.len, DMA_FROM_DEVICE);
+ put_page(unmap->page);
+ unmap->page = NULL;
+ dma_unmap_addr_set(&unmap->vector, dma_addr, 0);
+ unmap->vector.len = 0;
+}
+
+static inline void
+bnad_rxq_cleanup_skb(struct bnad *bnad, struct bnad_rx_unmap *unmap)
+{
+ if (!unmap->skb)
+ return;
+
+ dma_unmap_single(&bnad->pcidev->dev,
+ dma_unmap_addr(&unmap->vector, dma_addr),
+ unmap->vector.len, DMA_FROM_DEVICE);
+ dev_kfree_skb_any(unmap->skb);
+ unmap->skb = NULL;
+ dma_unmap_addr_set(&unmap->vector, dma_addr, 0);
+ unmap->vector.len = 0;
+}
+
static void
bnad_rxq_cleanup(struct bnad *bnad, struct bna_rcb *rcb)
{
- struct bnad_rx_unmap *unmap_q = rcb->unmap_q;
- struct sk_buff *skb;
+ struct bnad_rx_unmap_q *unmap_q = rcb->unmap_q;
int i;
for (i = 0; i < rcb->q_depth; i++) {
- struct bnad_rx_unmap *unmap = &unmap_q[i];
+ struct bnad_rx_unmap *unmap = &unmap_q->unmap[i];
- skb = unmap->skb;
- if (!skb)
- continue;
+ if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
+ bnad_rxq_cleanup_page(bnad, unmap);
+ else
+ bnad_rxq_cleanup_skb(bnad, unmap);
+ }
+ bnad_rxq_alloc_uninit(bnad, rcb);
+}
- unmap->skb = NULL;
- dma_unmap_single(&bnad->pcidev->dev,
- dma_unmap_addr(&unmap->vector, dma_addr),
- unmap->vector.len, DMA_FROM_DEVICE);
- dma_unmap_addr_set(&unmap->vector, dma_addr, 0);
- unmap->vector.len = 0;
- dev_kfree_skb_any(skb);
+static u32
+bnad_rxq_refill_page(struct bnad *bnad, struct bna_rcb *rcb, u32 nalloc)
+{
+ u32 alloced, prod, q_depth;
+ struct bnad_rx_unmap_q *unmap_q = rcb->unmap_q;
+ struct bnad_rx_unmap *unmap, *prev;
+ struct bna_rxq_entry *rxent;
+ struct page *page;
+ u32 page_offset, alloc_size;
+ dma_addr_t dma_addr;
+
+ prod = rcb->producer_index;
+ q_depth = rcb->q_depth;
+
+ alloc_size = PAGE_SIZE << unmap_q->alloc_order;
+ alloced = 0;
+
+ while (nalloc--) {
+ unmap = &unmap_q->unmap[prod];
+
+ if (unmap_q->reuse_pi < 0) {
+ page = alloc_pages(GFP_ATOMIC | __GFP_COMP,
+ unmap_q->alloc_order);
+ page_offset = 0;
+ } else {
+ prev = &unmap_q->unmap[unmap_q->reuse_pi];
+ page = prev->page;
+ page_offset = prev->page_offset + unmap_q->map_size;
+ get_page(page);
+ }
+
+ if (unlikely(!page)) {
+ BNAD_UPDATE_CTR(bnad, rxbuf_alloc_failed);
+ rcb->rxq->rxbuf_alloc_failed++;
+ goto finishing;
+ }
+
+ dma_addr = dma_map_page(&bnad->pcidev->dev, page, page_offset,
+ unmap_q->map_size, DMA_FROM_DEVICE);
+
+ unmap->page = page;
+ unmap->page_offset = page_offset;
+ dma_unmap_addr_set(&unmap->vector, dma_addr, dma_addr);
+ unmap->vector.len = unmap_q->map_size;
+ page_offset += unmap_q->map_size;
+
+ if (page_offset < alloc_size)
+ unmap_q->reuse_pi = prod;
+ else
+ unmap_q->reuse_pi = -1;
+
+ rxent = &((struct bna_rxq_entry *)rcb->sw_q)[prod];
+ BNA_SET_DMA_ADDR(dma_addr, &rxent->host_addr);
+ BNA_QE_INDX_INC(prod, q_depth);
+ alloced++;
}
+
+finishing:
+ if (likely(alloced)) {
+ rcb->producer_index = prod;
+ smp_mb();
+ if (likely(test_bit(BNAD_RXQ_POST_OK, &rcb->flags)))
+ bna_rxq_prod_indx_doorbell(rcb);
+ }
+
+ return alloced;
}
-/* Allocate and post BNAD_RXQ_REFILL_THRESHOLD_SHIFT buffers at a time */
-static void
-bnad_rxq_post(struct bnad *bnad, struct bna_rcb *rcb)
+static u32
+bnad_rxq_refill_skb(struct bnad *bnad, struct bna_rcb *rcb, u32 nalloc)
{
- u32 to_alloc, alloced, prod, q_depth, buff_sz;
- struct bnad_rx_unmap *unmap_q = rcb->unmap_q;
+ u32 alloced, prod, q_depth, buff_sz;
+ struct bnad_rx_unmap_q *unmap_q = rcb->unmap_q;
struct bnad_rx_unmap *unmap;
struct bna_rxq_entry *rxent;
struct sk_buff *skb;
dma_addr_t dma_addr;
buff_sz = rcb->rxq->buffer_size;
- alloced = 0;
- to_alloc = BNA_QE_FREE_CNT(rcb, rcb->q_depth);
- if (!(to_alloc >> BNAD_RXQ_REFILL_THRESHOLD_SHIFT))
- return;
-
prod = rcb->producer_index;
q_depth = rcb->q_depth;
- while (to_alloc--) {
- skb = netdev_alloc_skb_ip_align(bnad->netdev,
- buff_sz);
+ alloced = 0;
+ while (nalloc--) {
+ unmap = &unmap_q->unmap[prod];
+
+ skb = netdev_alloc_skb_ip_align(bnad->netdev, buff_sz);
+
if (unlikely(!skb)) {
BNAD_UPDATE_CTR(bnad, rxbuf_alloc_failed);
rcb->rxq->rxbuf_alloc_failed++;
@@ -320,13 +448,13 @@ bnad_rxq_post(struct bnad *bnad, struct bna_rcb *rcb)
}
dma_addr = dma_map_single(&bnad->pcidev->dev, skb->data,
buff_sz, DMA_FROM_DEVICE);
- rxent = &((struct bna_rxq_entry *)rcb->sw_q)[prod];
- BNA_SET_DMA_ADDR(dma_addr, &rxent->host_addr);
- unmap = &unmap_q[prod];
unmap->skb = skb;
dma_unmap_addr_set(&unmap->vector, dma_addr, dma_addr);
unmap->vector.len = buff_sz;
+
+ rxent = &((struct bna_rxq_entry *)rcb->sw_q)[prod];
+ BNA_SET_DMA_ADDR(dma_addr, &rxent->host_addr);
BNA_QE_INDX_INC(prod, q_depth);
alloced++;
}
@@ -338,6 +466,24 @@ finishing:
if (likely(test_bit(BNAD_RXQ_POST_OK, &rcb->flags)))
bna_rxq_prod_indx_doorbell(rcb);
}
+
+ return alloced;
+}
+
+static inline void
+bnad_rxq_post(struct bnad *bnad, struct bna_rcb *rcb)
+{
+ struct bnad_rx_unmap_q *unmap_q = rcb->unmap_q;
+ u32 to_alloc;
+
+ to_alloc = BNA_QE_FREE_CNT(rcb, rcb->q_depth);
+ if (!(to_alloc >> BNAD_RXQ_REFILL_THRESHOLD_SHIFT))
+ return;
+
+ if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
+ bnad_rxq_refill_page(bnad, rcb, to_alloc);
+ else
+ bnad_rxq_refill_skb(bnad, rcb, to_alloc);
}
#define flags_cksum_prot_mask (BNA_CQ_EF_IPV4 | BNA_CQ_EF_L3_CKSUM_OK | \
@@ -354,17 +500,62 @@ finishing:
#define flags_udp6 (BNA_CQ_EF_IPV6 | \
BNA_CQ_EF_UDP | BNA_CQ_EF_L4_CKSUM_OK)
+static inline struct sk_buff *
+bnad_cq_prepare_skb(struct bnad_rx_ctrl *rx_ctrl,
+ struct bnad_rx_unmap_q *unmap_q,
+ struct bnad_rx_unmap *unmap,
+ u32 length, u32 flags)
+{
+ struct bnad *bnad = rx_ctrl->bnad;
+ struct sk_buff *skb;
+
+ if (BNAD_RXBUF_IS_PAGE(unmap_q->type)) {
+ skb = napi_get_frags(&rx_ctrl->napi);
+ if (unlikely(!skb))
+ return NULL;
+
+ dma_unmap_page(&bnad->pcidev->dev,
+ dma_unmap_addr(&unmap->vector, dma_addr),
+ unmap->vector.len, DMA_FROM_DEVICE);
+ skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
+ unmap->page, unmap->page_offset, length);
+ skb->len += length;
+ skb->data_len += length;
+ skb->truesize += length;
+
+ unmap->page = NULL;
+ unmap->vector.len = 0;
+
+ return skb;
+ }
+
+ skb = unmap->skb;
+ BUG_ON(!skb);
+
+ dma_unmap_single(&bnad->pcidev->dev,
+ dma_unmap_addr(&unmap->vector, dma_addr),
+ unmap->vector.len, DMA_FROM_DEVICE);
+
+ skb_put(skb, length);
+
+ skb->protocol = eth_type_trans(skb, bnad->netdev);
+
+ unmap->skb = NULL;
+ unmap->vector.len = 0;
+ return skb;
+}
+
static u32
bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
{
- struct bna_cq_entry *cq, *cmpl, *next_cmpl;
+ struct bna_cq_entry *cq, *cmpl;
struct bna_rcb *rcb = NULL;
- struct bnad_rx_unmap *unmap_q, *unmap;
- unsigned int packets = 0;
+ struct bnad_rx_unmap_q *unmap_q;
+ struct bnad_rx_unmap *unmap;
struct sk_buff *skb;
- u32 flags, masked_flags;
struct bna_pkt_rate *pkt_rt = &ccb->pkt_rate;
- struct bnad_rx_ctrl *rx_ctrl = (struct bnad_rx_ctrl *)(ccb->ctrl);
+ struct bnad_rx_ctrl *rx_ctrl = ccb->ctrl;
+ u32 packets = 0, length = 0, flags, masked_flags;
prefetch(bnad->netdev);
@@ -373,6 +564,8 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
while (cmpl->valid && (packets < budget)) {
packets++;
+ flags = ntohl(cmpl->flags);
+ length = ntohs(cmpl->length);
BNA_UPDATE_PKT_CNT(pkt_rt, ntohs(cmpl->length));
if (bna_is_small_rxq(cmpl->rxq_id))
@@ -381,32 +574,25 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
rcb = ccb->rcb[0];
unmap_q = rcb->unmap_q;
- unmap = &unmap_q[rcb->consumer_index];
+ unmap = &unmap_q->unmap[rcb->consumer_index];
- skb = unmap->skb;
- BUG_ON(!(skb));
- unmap->skb = NULL;
- dma_unmap_single(&bnad->pcidev->dev,
- dma_unmap_addr(&unmap->vector, dma_addr),
- unmap->vector.len, DMA_FROM_DEVICE);
- unmap->vector.len = 0;
- BNA_QE_INDX_INC(rcb->consumer_index, rcb->q_depth);
- BNA_QE_INDX_INC(ccb->producer_index, ccb->q_depth);
- next_cmpl = &cq[ccb->producer_index];
+ if (unlikely(flags & (BNA_CQ_EF_MAC_ERROR |
+ BNA_CQ_EF_FCS_ERROR |
+ BNA_CQ_EF_TOO_LONG))) {
+ if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
+ bnad_rxq_cleanup_page(bnad, unmap);
+ else
+ bnad_rxq_cleanup_skb(bnad, unmap);
- prefetch(next_cmpl);
-
- flags = ntohl(cmpl->flags);
- if (unlikely
- (flags &
- (BNA_CQ_EF_MAC_ERROR | BNA_CQ_EF_FCS_ERROR |
- BNA_CQ_EF_TOO_LONG))) {
- dev_kfree_skb_any(skb);
rcb->rxq->rx_packets_with_error++;
goto next;
}
- skb_put(skb, ntohs(cmpl->length));
+ skb = bnad_cq_prepare_skb(ccb->ctrl, unmap_q, unmap,
+ length, flags);
+
+ if (unlikely(!skb))
+ break;
masked_flags = flags & flags_cksum_prot_mask;
@@ -421,22 +607,24 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
skb_checksum_none_assert(skb);
rcb->rxq->rx_packets++;
- rcb->rxq->rx_bytes += skb->len;
- skb->protocol = eth_type_trans(skb, bnad->netdev);
+ rcb->rxq->rx_bytes += length;
if (flags & BNA_CQ_EF_VLAN)
__vlan_hwaccel_put_tag(skb, ntohs(cmpl->vlan_tag));
- if (skb->ip_summed == CHECKSUM_UNNECESSARY)
- napi_gro_receive(&rx_ctrl->napi, skb);
+ if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
+ napi_gro_frags(&rx_ctrl->napi);
else
netif_receive_skb(skb);
next:
cmpl->valid = 0;
- cmpl = next_cmpl;
+ BNA_QE_INDX_INC(rcb->consumer_index, rcb->q_depth);
+ BNA_QE_INDX_INC(ccb->producer_index, ccb->q_depth);
+ cmpl = &cq[ccb->producer_index];
}
+ napi_gro_flush(&rx_ctrl->napi, false);
if (likely(test_bit(BNAD_RXQ_STARTED, &ccb->rcb[0]->flags)))
bna_ib_ack_disable_irq(ccb->i_dbell, packets);
@@ -956,8 +1144,7 @@ bnad_cb_rx_post(struct bnad *bnad, struct bna_rx *rx)
struct bna_ccb *ccb;
struct bna_rcb *rcb;
struct bnad_rx_ctrl *rx_ctrl;
- int i;
- int j;
+ int i, j;
for (i = 0; i < BNAD_MAX_RXP_PER_RX; i++) {
rx_ctrl = &rx_info->rx_ctrl[i];
@@ -972,6 +1159,7 @@ bnad_cb_rx_post(struct bnad *bnad, struct bna_rx *rx)
if (!rcb)
continue;
+ bnad_rxq_alloc_init(bnad, rcb);
set_bit(BNAD_RXQ_STARTED, &rcb->flags);
set_bit(BNAD_RXQ_POST_OK, &rcb->flags);
bnad_rxq_post(bnad, rcb);
@@ -1861,9 +2049,11 @@ bnad_setup_rx(struct bnad *bnad, u32 rx_id)
/* Fill Unmap Q memory requirements */
BNAD_FILL_UNMAPQ_MEM_REQ(&res_info[BNA_RX_RES_MEM_T_UNMAPQ],
- rx_config->num_paths + ((rx_config->rxp_type == BNA_RXP_SINGLE)
- ? 0 : rx_config->num_paths), (bnad->rxq_depth *
- sizeof(struct bnad_rx_unmap)));
+ rx_config->num_paths +
+ ((rx_config->rxp_type == BNA_RXP_SINGLE) ?
+ 0 : rx_config->num_paths),
+ ((bnad->rxq_depth * sizeof(struct bnad_rx_unmap)) +
+ sizeof(struct bnad_rx_unmap_q)));
/* Allocate resource */
err = bnad_rx_res_alloc(bnad, res_info, rx_id);
diff --git a/drivers/net/ethernet/brocade/bna/bnad.h b/drivers/net/ethernet/brocade/bna/bnad.h
index db132c9..134d534 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.h
+++ b/drivers/net/ethernet/brocade/bna/bnad.h
@@ -233,10 +233,29 @@ struct bnad_rx_vector {
};
struct bnad_rx_unmap {
+ struct page *page;
+ u32 page_offset;
struct sk_buff *skb;
struct bnad_rx_vector vector;
};
+enum bnad_rxbuf_type {
+ BNAD_RXBUF_NONE = 0,
+ BNAD_RXBUF_SKB = 1,
+ BNAD_RXBUF_PAGE = 2,
+ BNAD_RXBUF_MULTI = 3
+};
+
+#define BNAD_RXBUF_IS_PAGE(_type) ((_type) == BNAD_RXBUF_PAGE)
+
+struct bnad_rx_unmap_q {
+ int reuse_pi;
+ int alloc_order;
+ u32 map_size;
+ enum bnad_rxbuf_type type;
+ struct bnad_rx_unmap unmap[0];
+};
+
/* Bit mask values for bnad->cfg_flags */
#define BNAD_CF_DIM_ENABLED 0x01 /* DIM */
#define BNAD_CF_PROMISC 0x02
--
1.7.1
^ permalink raw reply related
* [net-next 5/7] bna: Add RX State
From: Rasesh Mody @ 2012-12-10 21:42 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
In-Reply-To: <1355175725-19202-1-git-send-email-rmody@brocade.com>
Change Details:
- BNA state machine for Rx in start_wait state moves it to stop_wait on
receipt of RX_E_STOP. In Rx stop_wait state, on receipt of
RX_E_STARTED event does enet stop
RX_E_STOPPED event does rx_cleanup_cbfn
rx_cleanup_cbfn in this case is called without post_cbfn. post_cbfn
happens only after RX_E_STARTED event is received in start_wait. Without
doing post_cbfn, NAPI remains disabled and in cleanup we try to disable
again causing endless wait. ifconfig process and other workers can thus
get stuck.
- Introducing start_stop_wait state for Rx. This state handles the case of
if post_cbfn is not done simply do stop without the cleanup.
Signed-off-by: Rasesh Mody <rmody@brocade.com>
---
drivers/net/ethernet/brocade/bna/bna_tx_rx.c | 27 +++++++++++++++++++++++++-
1 files changed, 26 insertions(+), 1 deletions(-)
diff --git a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
index 4df6d4b..ea6f4a0 100644
--- a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
+++ b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
@@ -1355,6 +1355,8 @@ bfa_fsm_state_decl(bna_rx, stopped,
struct bna_rx, enum bna_rx_event);
bfa_fsm_state_decl(bna_rx, start_wait,
struct bna_rx, enum bna_rx_event);
+bfa_fsm_state_decl(bna_rx, start_stop_wait,
+ struct bna_rx, enum bna_rx_event);
bfa_fsm_state_decl(bna_rx, rxf_start_wait,
struct bna_rx, enum bna_rx_event);
bfa_fsm_state_decl(bna_rx, started,
@@ -1432,7 +1434,7 @@ static void bna_rx_sm_start_wait(struct bna_rx *rx,
{
switch (event) {
case RX_E_STOP:
- bfa_fsm_set_state(rx, bna_rx_sm_stop_wait);
+ bfa_fsm_set_state(rx, bna_rx_sm_start_stop_wait);
break;
case RX_E_FAIL:
@@ -1488,6 +1490,29 @@ bna_rx_sm_rxf_stop_wait(struct bna_rx *rx, enum bna_rx_event event)
}
+static void
+bna_rx_sm_start_stop_wait_entry(struct bna_rx *rx)
+{
+}
+
+static void
+bna_rx_sm_start_stop_wait(struct bna_rx *rx, enum bna_rx_event event)
+{
+ switch (event) {
+ case RX_E_FAIL:
+ case RX_E_STOPPED:
+ bfa_fsm_set_state(rx, bna_rx_sm_stopped);
+ break;
+
+ case RX_E_STARTED:
+ bna_rx_enet_stop(rx);
+ break;
+
+ default:
+ bfa_sm_fault(event);
+ }
+}
+
void
bna_rx_sm_started_entry(struct bna_rx *rx)
{
--
1.7.1
^ permalink raw reply related
* [net-next 7/7] bna: Driver Version Updated to 3.1.2.1
From: Rasesh Mody @ 2012-12-10 21:42 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
In-Reply-To: <1355175725-19202-1-git-send-email-rmody@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
---
drivers/net/ethernet/brocade/bna/bnad.h | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/drivers/net/ethernet/brocade/bna/bnad.h b/drivers/net/ethernet/brocade/bna/bnad.h
index 134d534..72ba586 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.h
+++ b/drivers/net/ethernet/brocade/bna/bnad.h
@@ -71,7 +71,7 @@ struct bnad_rx_ctrl {
#define BNAD_NAME "bna"
#define BNAD_NAME_LEN 64
-#define BNAD_VERSION "3.0.23.0"
+#define BNAD_VERSION "3.1.2.1"
#define BNAD_MAILBOX_MSIX_INDEX 0
#define BNAD_MAILBOX_MSIX_VECTORS 1
--
1.7.1
^ permalink raw reply related
* [net-next 2/7] bna: Tx and Rx Optimizations
From: Rasesh Mody @ 2012-12-10 21:42 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
In-Reply-To: <1355175725-19202-1-git-send-email-rmody@brocade.com>
Change details:
- Have contiguous queue pages for TxQ, RxQ and CQ. Data structure and
QPT changes related to contiguous queue pages
- Optimized Tx and Rx unmap structures. Tx and Rx fast path changes due to
unmap data structure changes
- Re-factored Tx and Rx fastpath routines as per the new queue data structures
- Implemented bnad_txq_wi_prepare() to program the opcode, flags, frame_len
and num_vectors in the work item
- Reduced Max TxQ and RxQ depth to 2048 while default value for Tx/Rx queue
depth is unaltered (512)
Signed-off-by: Rasesh Mody <rmody@brocade.com>
---
drivers/net/ethernet/brocade/bna/bna.h | 2 +
drivers/net/ethernet/brocade/bna/bna_tx_rx.c | 109 +++--
drivers/net/ethernet/brocade/bna/bna_types.h | 9 +-
drivers/net/ethernet/brocade/bna/bnad.c | 690 +++++++++++---------------
drivers/net/ethernet/brocade/bna/bnad.h | 41 +-
5 files changed, 381 insertions(+), 470 deletions(-)
diff --git a/drivers/net/ethernet/brocade/bna/bna.h b/drivers/net/ethernet/brocade/bna/bna.h
index ede532b..25dae75 100644
--- a/drivers/net/ethernet/brocade/bna/bna.h
+++ b/drivers/net/ethernet/brocade/bna/bna.h
@@ -138,6 +138,8 @@ do { \
#define BNA_QE_INDX_ADD(_qe_idx, _qe_num, _q_depth) \
((_qe_idx) = ((_qe_idx) + (_qe_num)) & ((_q_depth) - 1))
+#define BNA_QE_INDX_INC(_idx, _q_depth) BNA_QE_INDX_ADD(_idx, 1, _q_depth)
+
#define BNA_Q_INDEX_CHANGE(_old_idx, _updated_idx, _q_depth) \
(((_updated_idx) - (_old_idx)) & ((_q_depth) - 1))
diff --git a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
index 71144b3..bb5467b 100644
--- a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
+++ b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
@@ -1908,6 +1908,9 @@ bna_rxq_qpt_setup(struct bna_rxq *rxq,
struct bna_mem_descr *swqpt_mem,
struct bna_mem_descr *page_mem)
{
+ u8 *kva;
+ u64 dma;
+ struct bna_dma_addr bna_dma;
int i;
rxq->qpt.hw_qpt_ptr.lsb = qpt_mem->dma.lsb;
@@ -1917,13 +1920,21 @@ bna_rxq_qpt_setup(struct bna_rxq *rxq,
rxq->qpt.page_size = page_size;
rxq->rcb->sw_qpt = (void **) swqpt_mem->kva;
+ rxq->rcb->sw_q = page_mem->kva;
+
+ kva = page_mem->kva;
+ BNA_GET_DMA_ADDR(&page_mem->dma, dma);
for (i = 0; i < rxq->qpt.page_count; i++) {
- rxq->rcb->sw_qpt[i] = page_mem[i].kva;
+ rxq->rcb->sw_qpt[i] = kva;
+ kva += PAGE_SIZE;
+
+ BNA_SET_DMA_ADDR(dma, &bna_dma);
((struct bna_dma_addr *)rxq->qpt.kv_qpt_ptr)[i].lsb =
- page_mem[i].dma.lsb;
+ bna_dma.lsb;
((struct bna_dma_addr *)rxq->qpt.kv_qpt_ptr)[i].msb =
- page_mem[i].dma.msb;
+ bna_dma.msb;
+ dma += PAGE_SIZE;
}
}
@@ -1935,6 +1946,9 @@ bna_rxp_cqpt_setup(struct bna_rxp *rxp,
struct bna_mem_descr *swqpt_mem,
struct bna_mem_descr *page_mem)
{
+ u8 *kva;
+ u64 dma;
+ struct bna_dma_addr bna_dma;
int i;
rxp->cq.qpt.hw_qpt_ptr.lsb = qpt_mem->dma.lsb;
@@ -1944,14 +1958,21 @@ bna_rxp_cqpt_setup(struct bna_rxp *rxp,
rxp->cq.qpt.page_size = page_size;
rxp->cq.ccb->sw_qpt = (void **) swqpt_mem->kva;
+ rxp->cq.ccb->sw_q = page_mem->kva;
+
+ kva = page_mem->kva;
+ BNA_GET_DMA_ADDR(&page_mem->dma, dma);
for (i = 0; i < rxp->cq.qpt.page_count; i++) {
- rxp->cq.ccb->sw_qpt[i] = page_mem[i].kva;
+ rxp->cq.ccb->sw_qpt[i] = kva;
+ kva += PAGE_SIZE;
+ BNA_SET_DMA_ADDR(dma, &bna_dma);
((struct bna_dma_addr *)rxp->cq.qpt.kv_qpt_ptr)[i].lsb =
- page_mem[i].dma.lsb;
+ bna_dma.lsb;
((struct bna_dma_addr *)rxp->cq.qpt.kv_qpt_ptr)[i].msb =
- page_mem[i].dma.msb;
+ bna_dma.msb;
+ dma += PAGE_SIZE;
}
}
@@ -2250,8 +2271,8 @@ bna_rx_res_req(struct bna_rx_config *q_cfg, struct bna_res_info *res_info)
res_info[BNA_RX_RES_MEM_T_CQPT_PAGE].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_RX_RES_MEM_T_CQPT_PAGE].res_u.mem_info;
mem_info->mem_type = BNA_MEM_T_DMA;
- mem_info->len = PAGE_SIZE;
- mem_info->num = cpage_count * q_cfg->num_paths;
+ mem_info->len = PAGE_SIZE * cpage_count;
+ mem_info->num = q_cfg->num_paths;
res_info[BNA_RX_RES_MEM_T_DQPT].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_RX_RES_MEM_T_DQPT].res_u.mem_info;
@@ -2268,8 +2289,8 @@ bna_rx_res_req(struct bna_rx_config *q_cfg, struct bna_res_info *res_info)
res_info[BNA_RX_RES_MEM_T_DPAGE].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_RX_RES_MEM_T_DPAGE].res_u.mem_info;
mem_info->mem_type = BNA_MEM_T_DMA;
- mem_info->len = PAGE_SIZE;
- mem_info->num = dpage_count * q_cfg->num_paths;
+ mem_info->len = PAGE_SIZE * dpage_count;
+ mem_info->num = q_cfg->num_paths;
res_info[BNA_RX_RES_MEM_T_HQPT].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_RX_RES_MEM_T_HQPT].res_u.mem_info;
@@ -2286,8 +2307,8 @@ bna_rx_res_req(struct bna_rx_config *q_cfg, struct bna_res_info *res_info)
res_info[BNA_RX_RES_MEM_T_HPAGE].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_RX_RES_MEM_T_HPAGE].res_u.mem_info;
mem_info->mem_type = BNA_MEM_T_DMA;
- mem_info->len = (hpage_count ? PAGE_SIZE : 0);
- mem_info->num = (hpage_count ? (hpage_count * q_cfg->num_paths) : 0);
+ mem_info->len = PAGE_SIZE * hpage_count;
+ mem_info->num = (hpage_count ? q_cfg->num_paths : 0);
res_info[BNA_RX_RES_MEM_T_IBIDX].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_RX_RES_MEM_T_IBIDX].res_u.mem_info;
@@ -2332,7 +2353,7 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
struct bna_mem_descr *dsqpt_mem;
struct bna_mem_descr *hpage_mem;
struct bna_mem_descr *dpage_mem;
- int i, cpage_idx = 0, dpage_idx = 0, hpage_idx = 0;
+ int i;
int dpage_count, hpage_count, rcb_idx;
if (!bna_rx_res_check(rx_mod, rx_cfg))
@@ -2352,14 +2373,14 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
hpage_mem = &res_info[BNA_RX_RES_MEM_T_HPAGE].res_u.mem_info.mdl[0];
dpage_mem = &res_info[BNA_RX_RES_MEM_T_DPAGE].res_u.mem_info.mdl[0];
- page_count = res_info[BNA_RX_RES_MEM_T_CQPT_PAGE].res_u.mem_info.num /
- rx_cfg->num_paths;
+ page_count = res_info[BNA_RX_RES_MEM_T_CQPT_PAGE].res_u.mem_info.len /
+ PAGE_SIZE;
- dpage_count = res_info[BNA_RX_RES_MEM_T_DPAGE].res_u.mem_info.num /
- rx_cfg->num_paths;
+ dpage_count = res_info[BNA_RX_RES_MEM_T_DPAGE].res_u.mem_info.len /
+ PAGE_SIZE;
- hpage_count = res_info[BNA_RX_RES_MEM_T_HPAGE].res_u.mem_info.num /
- rx_cfg->num_paths;
+ hpage_count = res_info[BNA_RX_RES_MEM_T_HPAGE].res_u.mem_info.len /
+ PAGE_SIZE;
rx = bna_rx_get(rx_mod, rx_cfg->rx_type);
rx->bna = bna;
@@ -2446,10 +2467,7 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
q0->rx_packets_with_error = q0->rxbuf_alloc_failed = 0;
bna_rxq_qpt_setup(q0, rxp, dpage_count, PAGE_SIZE,
- &dqpt_mem[i], &dsqpt_mem[i], &dpage_mem[dpage_idx]);
- q0->rcb->page_idx = dpage_idx;
- q0->rcb->page_count = dpage_count;
- dpage_idx += dpage_count;
+ &dqpt_mem[i], &dsqpt_mem[i], &dpage_mem[i]);
if (rx->rcb_setup_cbfn)
rx->rcb_setup_cbfn(bnad, q0->rcb);
@@ -2475,10 +2493,7 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
bna_rxq_qpt_setup(q1, rxp, hpage_count, PAGE_SIZE,
&hqpt_mem[i], &hsqpt_mem[i],
- &hpage_mem[hpage_idx]);
- q1->rcb->page_idx = hpage_idx;
- q1->rcb->page_count = hpage_count;
- hpage_idx += hpage_count;
+ &hpage_mem[i]);
if (rx->rcb_setup_cbfn)
rx->rcb_setup_cbfn(bnad, q1->rcb);
@@ -2510,10 +2525,7 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
rxp->cq.ccb->id = i;
bna_rxp_cqpt_setup(rxp, page_count, PAGE_SIZE,
- &cqpt_mem[i], &cswqpt_mem[i], &cpage_mem[cpage_idx]);
- rxp->cq.ccb->page_idx = cpage_idx;
- rxp->cq.ccb->page_count = page_count;
- cpage_idx += page_count;
+ &cqpt_mem[i], &cswqpt_mem[i], &cpage_mem[i]);
if (rx->ccb_setup_cbfn)
rx->ccb_setup_cbfn(bnad, rxp->cq.ccb);
@@ -3230,6 +3242,9 @@ bna_txq_qpt_setup(struct bna_txq *txq, int page_count, int page_size,
struct bna_mem_descr *swqpt_mem,
struct bna_mem_descr *page_mem)
{
+ u8 *kva;
+ u64 dma;
+ struct bna_dma_addr bna_dma;
int i;
txq->qpt.hw_qpt_ptr.lsb = qpt_mem->dma.lsb;
@@ -3239,14 +3254,21 @@ bna_txq_qpt_setup(struct bna_txq *txq, int page_count, int page_size,
txq->qpt.page_size = page_size;
txq->tcb->sw_qpt = (void **) swqpt_mem->kva;
+ txq->tcb->sw_q = page_mem->kva;
+
+ kva = page_mem->kva;
+ BNA_GET_DMA_ADDR(&page_mem->dma, dma);
for (i = 0; i < page_count; i++) {
- txq->tcb->sw_qpt[i] = page_mem[i].kva;
+ txq->tcb->sw_qpt[i] = kva;
+ kva += PAGE_SIZE;
+ BNA_SET_DMA_ADDR(dma, &bna_dma);
((struct bna_dma_addr *)txq->qpt.kv_qpt_ptr)[i].lsb =
- page_mem[i].dma.lsb;
+ bna_dma.lsb;
((struct bna_dma_addr *)txq->qpt.kv_qpt_ptr)[i].msb =
- page_mem[i].dma.msb;
+ bna_dma.msb;
+ dma += PAGE_SIZE;
}
}
@@ -3430,8 +3452,8 @@ bna_tx_res_req(int num_txq, int txq_depth, struct bna_res_info *res_info)
res_info[BNA_TX_RES_MEM_T_PAGE].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_TX_RES_MEM_T_PAGE].res_u.mem_info;
mem_info->mem_type = BNA_MEM_T_DMA;
- mem_info->len = PAGE_SIZE;
- mem_info->num = num_txq * page_count;
+ mem_info->len = PAGE_SIZE * page_count;
+ mem_info->num = num_txq;
res_info[BNA_TX_RES_MEM_T_IBIDX].res_type = BNA_RES_T_MEM;
mem_info = &res_info[BNA_TX_RES_MEM_T_IBIDX].res_u.mem_info;
@@ -3457,14 +3479,11 @@ bna_tx_create(struct bna *bna, struct bnad *bnad,
struct bna_txq *txq;
struct list_head *qe;
int page_count;
- int page_size;
- int page_idx;
int i;
intr_info = &res_info[BNA_TX_RES_INTR_T_TXCMPL].res_u.intr_info;
- page_count = (res_info[BNA_TX_RES_MEM_T_PAGE].res_u.mem_info.num) /
- tx_cfg->num_txq;
- page_size = res_info[BNA_TX_RES_MEM_T_PAGE].res_u.mem_info.len;
+ page_count = (res_info[BNA_TX_RES_MEM_T_PAGE].res_u.mem_info.len) /
+ PAGE_SIZE;
/**
* Get resources
@@ -3529,7 +3548,6 @@ bna_tx_create(struct bna *bna, struct bnad *bnad,
/* TxQ */
i = 0;
- page_idx = 0;
list_for_each(qe, &tx->txq_q) {
txq = (struct bna_txq *)qe;
txq->tcb = (struct bna_tcb *)
@@ -3569,14 +3587,11 @@ bna_tx_create(struct bna *bna, struct bnad *bnad,
txq->tcb->id = i;
/* QPT, SWQPT, Pages */
- bna_txq_qpt_setup(txq, page_count, page_size,
+ bna_txq_qpt_setup(txq, page_count, PAGE_SIZE,
&res_info[BNA_TX_RES_MEM_T_QPT].res_u.mem_info.mdl[i],
&res_info[BNA_TX_RES_MEM_T_SWQPT].res_u.mem_info.mdl[i],
&res_info[BNA_TX_RES_MEM_T_PAGE].
- res_u.mem_info.mdl[page_idx]);
- txq->tcb->page_idx = page_idx;
- txq->tcb->page_count = page_count;
- page_idx += page_count;
+ res_u.mem_info.mdl[i]);
/* Callback to bnad for setting up TCB */
if (tx->tcb_setup_cbfn)
diff --git a/drivers/net/ethernet/brocade/bna/bna_types.h b/drivers/net/ethernet/brocade/bna/bna_types.h
index d3eb8bd..dc50f78 100644
--- a/drivers/net/ethernet/brocade/bna/bna_types.h
+++ b/drivers/net/ethernet/brocade/bna/bna_types.h
@@ -430,6 +430,7 @@ struct bna_ib {
struct bna_tcb {
/* Fast path */
void **sw_qpt;
+ void *sw_q;
void *unmap_q;
u32 producer_index;
u32 consumer_index;
@@ -437,8 +438,6 @@ struct bna_tcb {
u32 q_depth;
void __iomem *q_dbell;
struct bna_ib_dbell *i_dbell;
- int page_idx;
- int page_count;
/* Control path */
struct bna_txq *txq;
struct bnad *bnad;
@@ -563,13 +562,12 @@ struct bna_tx_mod {
struct bna_rcb {
/* Fast path */
void **sw_qpt;
+ void *sw_q;
void *unmap_q;
u32 producer_index;
u32 consumer_index;
u32 q_depth;
void __iomem *q_dbell;
- int page_idx;
- int page_count;
/* Control path */
struct bna_rxq *rxq;
struct bna_ccb *ccb;
@@ -626,6 +624,7 @@ struct bna_pkt_rate {
struct bna_ccb {
/* Fast path */
void **sw_qpt;
+ void *sw_q;
u32 producer_index;
volatile u32 *hw_producer_index;
u32 q_depth;
@@ -633,8 +632,6 @@ struct bna_ccb {
struct bna_rcb *rcb[2];
void *ctrl; /* For bnad */
struct bna_pkt_rate pkt_rate;
- int page_idx;
- int page_count;
/* Control path */
struct bna_cq *cq;
diff --git a/drivers/net/ethernet/brocade/bna/bnad.c b/drivers/net/ethernet/brocade/bna/bnad.c
index b7fb391..da5470a 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.c
+++ b/drivers/net/ethernet/brocade/bna/bnad.c
@@ -61,23 +61,17 @@ static const u8 bnad_bcast_addr[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
/*
* Local MACROS
*/
-#define BNAD_TX_UNMAPQ_DEPTH (bnad->txq_depth * 2)
-
-#define BNAD_RX_UNMAPQ_DEPTH (bnad->rxq_depth)
-
#define BNAD_GET_MBOX_IRQ(_bnad) \
(((_bnad)->cfg_flags & BNAD_CF_MSIX) ? \
((_bnad)->msix_table[BNAD_MAILBOX_MSIX_INDEX].vector) : \
((_bnad)->pcidev->irq))
-#define BNAD_FILL_UNMAPQ_MEM_REQ(_res_info, _num, _depth) \
+#define BNAD_FILL_UNMAPQ_MEM_REQ(_res_info, _num, _size) \
do { \
(_res_info)->res_type = BNA_RES_T_MEM; \
(_res_info)->res_u.mem_info.mem_type = BNA_MEM_T_KVA; \
(_res_info)->res_u.mem_info.num = (_num); \
- (_res_info)->res_u.mem_info.len = \
- sizeof(struct bnad_unmap_q) + \
- (sizeof(struct bnad_skb_unmap) * ((_depth) - 1)); \
+ (_res_info)->res_u.mem_info.len = (_size); \
} while (0)
static void
@@ -103,48 +97,58 @@ bnad_remove_from_list(struct bnad *bnad)
static void
bnad_cq_cleanup(struct bnad *bnad, struct bna_ccb *ccb)
{
- struct bna_cq_entry *cmpl, *next_cmpl;
- unsigned int wi_range, wis = 0, ccb_prod = 0;
+ struct bna_cq_entry *cmpl;
int i;
- BNA_CQ_QPGE_PTR_GET(ccb_prod, ccb->sw_qpt, cmpl,
- wi_range);
-
for (i = 0; i < ccb->q_depth; i++) {
- wis++;
- if (likely(--wi_range))
- next_cmpl = cmpl + 1;
- else {
- BNA_QE_INDX_ADD(ccb_prod, wis, ccb->q_depth);
- wis = 0;
- BNA_CQ_QPGE_PTR_GET(ccb_prod, ccb->sw_qpt,
- next_cmpl, wi_range);
- }
+ cmpl = &((struct bna_cq_entry *)ccb->sw_q)[i];
cmpl->valid = 0;
- cmpl = next_cmpl;
}
}
+/* Tx Datapath functions */
+
+
+/* Caller should ensure that the entry at unmap_q[index] is valid */
static u32
-bnad_pci_unmap_skb(struct device *pdev, struct bnad_skb_unmap *array,
- u32 index, u32 depth, struct sk_buff *skb, u32 frag)
+bnad_tx_buff_unmap(struct bnad *bnad,
+ struct bnad_tx_unmap *unmap_q,
+ u32 q_depth, u32 index)
{
- int j;
- array[index].skb = NULL;
-
- dma_unmap_single(pdev, dma_unmap_addr(&array[index], dma_addr),
- skb_headlen(skb), DMA_TO_DEVICE);
- dma_unmap_addr_set(&array[index], dma_addr, 0);
- BNA_QE_INDX_ADD(index, 1, depth);
+ struct bnad_tx_unmap *unmap;
+ struct sk_buff *skb;
+ int vector, nvecs;
+
+ unmap = &unmap_q[index];
+ nvecs = unmap->nvecs;
+
+ skb = unmap->skb;
+ unmap->skb = NULL;
+ unmap->nvecs = 0;
+ dma_unmap_single(&bnad->pcidev->dev,
+ dma_unmap_addr(&unmap->vectors[0], dma_addr),
+ skb_headlen(skb), DMA_TO_DEVICE);
+ dma_unmap_addr_set(&unmap->vectors[0], dma_addr, 0);
+ nvecs--;
+
+ vector = 0;
+ while (nvecs) {
+ vector++;
+ if (vector == BFI_TX_MAX_VECTORS_PER_WI) {
+ vector = 0;
+ BNA_QE_INDX_INC(index, q_depth);
+ unmap = &unmap_q[index];
+ }
- for (j = 0; j < frag; j++) {
- dma_unmap_page(pdev, dma_unmap_addr(&array[index], dma_addr),
- skb_frag_size(&skb_shinfo(skb)->frags[j]),
- DMA_TO_DEVICE);
- dma_unmap_addr_set(&array[index], dma_addr, 0);
- BNA_QE_INDX_ADD(index, 1, depth);
+ dma_unmap_page(&bnad->pcidev->dev,
+ dma_unmap_addr(&unmap->vectors[vector], dma_addr),
+ skb_shinfo(skb)->frags[nvecs].size, DMA_TO_DEVICE);
+ dma_unmap_addr_set(&unmap->vectors[vector], dma_addr, 0);
+ nvecs--;
}
+ BNA_QE_INDX_INC(index, q_depth);
+
return index;
}
@@ -154,79 +158,64 @@ bnad_pci_unmap_skb(struct device *pdev, struct bnad_skb_unmap *array,
* so DMA unmap & freeing is fine.
*/
static void
-bnad_txq_cleanup(struct bnad *bnad,
- struct bna_tcb *tcb)
+bnad_txq_cleanup(struct bnad *bnad, struct bna_tcb *tcb)
{
- u32 unmap_cons;
- struct bnad_unmap_q *unmap_q = tcb->unmap_q;
- struct bnad_skb_unmap *unmap_array;
- struct sk_buff *skb = NULL;
- int q;
-
- unmap_array = unmap_q->unmap_array;
+ struct bnad_tx_unmap *unmap_q = tcb->unmap_q;
+ struct sk_buff *skb;
+ int i;
- for (q = 0; q < unmap_q->q_depth; q++) {
- skb = unmap_array[q].skb;
+ for (i = 0; i < tcb->q_depth; i++) {
+ skb = unmap_q[i].skb;
if (!skb)
continue;
-
- unmap_cons = q;
- unmap_cons = bnad_pci_unmap_skb(&bnad->pcidev->dev, unmap_array,
- unmap_cons, unmap_q->q_depth, skb,
- skb_shinfo(skb)->nr_frags);
+ bnad_tx_buff_unmap(bnad, unmap_q, tcb->q_depth, i);
dev_kfree_skb_any(skb);
}
}
-/* Data Path Handlers */
-
/*
* bnad_txcmpl_process : Frees the Tx bufs on Tx completion
* Can be called in a) Interrupt context
* b) Sending context
*/
static u32
-bnad_txcmpl_process(struct bnad *bnad,
- struct bna_tcb *tcb)
+bnad_txcmpl_process(struct bnad *bnad, struct bna_tcb *tcb)
{
- u32 unmap_cons, sent_packets = 0, sent_bytes = 0;
- u16 wis, updated_hw_cons;
- struct bnad_unmap_q *unmap_q = tcb->unmap_q;
- struct bnad_skb_unmap *unmap_array;
- struct sk_buff *skb;
+ u32 sent_packets = 0, sent_bytes = 0;
+ u32 wis, unmap_wis, hw_cons, cons, q_depth;
+ struct bnad_tx_unmap *unmap_q = tcb->unmap_q;
+ struct bnad_tx_unmap *unmap;
+ struct sk_buff *skb;
/* Just return if TX is stopped */
if (!test_bit(BNAD_TXQ_TX_STARTED, &tcb->flags))
return 0;
- updated_hw_cons = *(tcb->hw_consumer_index);
-
- wis = BNA_Q_INDEX_CHANGE(tcb->consumer_index,
- updated_hw_cons, tcb->q_depth);
+ hw_cons = *(tcb->hw_consumer_index);
+ cons = tcb->consumer_index;
+ q_depth = tcb->q_depth;
+ wis = BNA_Q_INDEX_CHANGE(cons, hw_cons, q_depth);
BUG_ON(!(wis <= BNA_QE_IN_USE_CNT(tcb, tcb->q_depth)));
- unmap_array = unmap_q->unmap_array;
- unmap_cons = unmap_q->consumer_index;
-
while (wis) {
- skb = unmap_array[unmap_cons].skb;
+ unmap = &unmap_q[cons];
+
+ skb = unmap->skb;
sent_packets++;
sent_bytes += skb->len;
- wis -= BNA_TXQ_WI_NEEDED(1 + skb_shinfo(skb)->nr_frags);
- unmap_cons = bnad_pci_unmap_skb(&bnad->pcidev->dev, unmap_array,
- unmap_cons, unmap_q->q_depth, skb,
- skb_shinfo(skb)->nr_frags);
+ unmap_wis = BNA_TXQ_WI_NEEDED(unmap->nvecs);
+ wis -= unmap_wis;
+ cons = bnad_tx_buff_unmap(bnad, unmap_q, q_depth, cons);
dev_kfree_skb_any(skb);
}
/* Update consumer pointers. */
- tcb->consumer_index = updated_hw_cons;
- unmap_q->consumer_index = unmap_cons;
+ tcb->consumer_index = hw_cons;
tcb->txq->tx_packets += sent_packets;
tcb->txq->tx_bytes += sent_bytes;
@@ -278,110 +267,79 @@ bnad_msix_tx(int irq, void *data)
}
static void
-bnad_rcb_cleanup(struct bnad *bnad, struct bna_rcb *rcb)
-{
- struct bnad_unmap_q *unmap_q = rcb->unmap_q;
-
- rcb->producer_index = 0;
- rcb->consumer_index = 0;
-
- unmap_q->producer_index = 0;
- unmap_q->consumer_index = 0;
-}
-
-static void
bnad_rxq_cleanup(struct bnad *bnad, struct bna_rcb *rcb)
{
- struct bnad_unmap_q *unmap_q;
- struct bnad_skb_unmap *unmap_array;
+ struct bnad_rx_unmap *unmap_q = rcb->unmap_q;
struct sk_buff *skb;
- int unmap_cons;
+ int i;
+
+ for (i = 0; i < rcb->q_depth; i++) {
+ struct bnad_rx_unmap *unmap = &unmap_q[i];
- unmap_q = rcb->unmap_q;
- unmap_array = unmap_q->unmap_array;
- for (unmap_cons = 0; unmap_cons < unmap_q->q_depth; unmap_cons++) {
- skb = unmap_array[unmap_cons].skb;
+ skb = unmap->skb;
if (!skb)
continue;
- unmap_array[unmap_cons].skb = NULL;
+
+ unmap->skb = NULL;
dma_unmap_single(&bnad->pcidev->dev,
- dma_unmap_addr(&unmap_array[unmap_cons],
- dma_addr),
- rcb->rxq->buffer_size,
- DMA_FROM_DEVICE);
- dev_kfree_skb(skb);
+ dma_unmap_addr(&unmap->vector, dma_addr),
+ unmap->vector.len, DMA_FROM_DEVICE);
+ dma_unmap_addr_set(&unmap->vector, dma_addr, 0);
+ unmap->vector.len = 0;
+ dev_kfree_skb_any(skb);
}
- bnad_rcb_cleanup(bnad, rcb);
}
+/* Allocate and post BNAD_RXQ_REFILL_THRESHOLD_SHIFT buffers at a time */
static void
bnad_rxq_post(struct bnad *bnad, struct bna_rcb *rcb)
{
- u16 to_alloc, alloced, unmap_prod, wi_range;
- struct bnad_unmap_q *unmap_q = rcb->unmap_q;
- struct bnad_skb_unmap *unmap_array;
+ u32 to_alloc, alloced, prod, q_depth, buff_sz;
+ struct bnad_rx_unmap *unmap_q = rcb->unmap_q;
+ struct bnad_rx_unmap *unmap;
struct bna_rxq_entry *rxent;
struct sk_buff *skb;
dma_addr_t dma_addr;
+ buff_sz = rcb->rxq->buffer_size;
alloced = 0;
- to_alloc =
- BNA_QE_FREE_CNT(unmap_q, unmap_q->q_depth);
-
- unmap_array = unmap_q->unmap_array;
- unmap_prod = unmap_q->producer_index;
+ to_alloc = BNA_QE_FREE_CNT(rcb, rcb->q_depth);
+ if (!(to_alloc >> BNAD_RXQ_REFILL_THRESHOLD_SHIFT))
+ return;
- BNA_RXQ_QPGE_PTR_GET(unmap_prod, rcb->sw_qpt, rxent, wi_range);
+ prod = rcb->producer_index;
+ q_depth = rcb->q_depth;
while (to_alloc--) {
- if (!wi_range)
- BNA_RXQ_QPGE_PTR_GET(unmap_prod, rcb->sw_qpt, rxent,
- wi_range);
skb = netdev_alloc_skb_ip_align(bnad->netdev,
- rcb->rxq->buffer_size);
+ buff_sz);
if (unlikely(!skb)) {
BNAD_UPDATE_CTR(bnad, rxbuf_alloc_failed);
rcb->rxq->rxbuf_alloc_failed++;
goto finishing;
}
- unmap_array[unmap_prod].skb = skb;
dma_addr = dma_map_single(&bnad->pcidev->dev, skb->data,
- rcb->rxq->buffer_size,
- DMA_FROM_DEVICE);
- dma_unmap_addr_set(&unmap_array[unmap_prod], dma_addr,
- dma_addr);
- BNA_SET_DMA_ADDR(dma_addr, &rxent->host_addr);
- BNA_QE_INDX_ADD(unmap_prod, 1, unmap_q->q_depth);
+ buff_sz, DMA_FROM_DEVICE);
+ rxent = &((struct bna_rxq_entry *)rcb->sw_q)[prod];
- rxent++;
- wi_range--;
+ BNA_SET_DMA_ADDR(dma_addr, &rxent->host_addr);
+ unmap = &unmap_q[prod];
+ unmap->skb = skb;
+ dma_unmap_addr_set(&unmap->vector, dma_addr, dma_addr);
+ unmap->vector.len = buff_sz;
+ BNA_QE_INDX_INC(prod, q_depth);
alloced++;
}
finishing:
if (likely(alloced)) {
- unmap_q->producer_index = unmap_prod;
- rcb->producer_index = unmap_prod;
+ rcb->producer_index = prod;
smp_mb();
if (likely(test_bit(BNAD_RXQ_POST_OK, &rcb->flags)))
bna_rxq_prod_indx_doorbell(rcb);
}
}
-static inline void
-bnad_refill_rxq(struct bnad *bnad, struct bna_rcb *rcb)
-{
- struct bnad_unmap_q *unmap_q = rcb->unmap_q;
-
- if (!test_and_set_bit(BNAD_RXQ_REFILL, &rcb->flags)) {
- if (BNA_QE_FREE_CNT(unmap_q, unmap_q->q_depth)
- >> BNAD_RXQ_REFILL_THRESHOLD_SHIFT)
- bnad_rxq_post(bnad, rcb);
- smp_mb__before_clear_bit();
- clear_bit(BNAD_RXQ_REFILL, &rcb->flags);
- }
-}
-
#define flags_cksum_prot_mask (BNA_CQ_EF_IPV4 | BNA_CQ_EF_L3_CKSUM_OK | \
BNA_CQ_EF_IPV6 | \
BNA_CQ_EF_TCP | BNA_CQ_EF_UDP | \
@@ -399,21 +357,21 @@ bnad_refill_rxq(struct bnad *bnad, struct bna_rcb *rcb)
static u32
bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
{
- struct bna_cq_entry *cmpl, *next_cmpl;
+ struct bna_cq_entry *cq, *cmpl, *next_cmpl;
struct bna_rcb *rcb = NULL;
- unsigned int wi_range, packets = 0, wis = 0;
- struct bnad_unmap_q *unmap_q;
- struct bnad_skb_unmap *unmap_array, *curr_ua;
+ struct bnad_rx_unmap *unmap_q, *unmap;
+ unsigned int packets = 0;
struct sk_buff *skb;
- u32 flags, unmap_cons, masked_flags;
+ u32 flags, masked_flags;
struct bna_pkt_rate *pkt_rt = &ccb->pkt_rate;
struct bnad_rx_ctrl *rx_ctrl = (struct bnad_rx_ctrl *)(ccb->ctrl);
prefetch(bnad->netdev);
- BNA_CQ_QPGE_PTR_GET(ccb->producer_index, ccb->sw_qpt, cmpl,
- wi_range);
- BUG_ON(!(wi_range <= ccb->q_depth));
- while (cmpl->valid && packets < budget) {
+
+ cq = ccb->sw_q;
+ cmpl = &cq[ccb->producer_index];
+
+ while (cmpl->valid && (packets < budget)) {
packets++;
BNA_UPDATE_PKT_CNT(pkt_rt, ntohs(cmpl->length));
@@ -423,33 +381,19 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
rcb = ccb->rcb[0];
unmap_q = rcb->unmap_q;
- unmap_array = unmap_q->unmap_array;
- unmap_cons = unmap_q->consumer_index;
+ unmap = &unmap_q[rcb->consumer_index];
- curr_ua = &unmap_array[unmap_cons];
-
- skb = curr_ua->skb;
- BUG_ON(!(skb));\
- curr_ua->skb = NULL;
+ skb = unmap->skb;
+ BUG_ON(!(skb));
+ unmap->skb = NULL;
dma_unmap_single(&bnad->pcidev->dev,
- dma_unmap_addr(curr_ua, dma_addr),
- rcb->rxq->buffer_size,
- DMA_FROM_DEVICE);
- BNA_QE_INDX_ADD(unmap_q->consumer_index, 1, unmap_q->q_depth);
-
- /* Should be more efficient ? Performance ? */
- BNA_QE_INDX_ADD(rcb->consumer_index, 1, rcb->q_depth);
-
- wis++;
- if (likely(--wi_range))
- next_cmpl = cmpl + 1;
- else {
- BNA_QE_INDX_ADD(ccb->producer_index, wis, ccb->q_depth);
- wis = 0;
- BNA_CQ_QPGE_PTR_GET(ccb->producer_index, ccb->sw_qpt,
- next_cmpl, wi_range);
- BUG_ON(!(wi_range <= ccb->q_depth));
- }
+ dma_unmap_addr(&unmap->vector, dma_addr),
+ unmap->vector.len, DMA_FROM_DEVICE);
+ unmap->vector.len = 0;
+ BNA_QE_INDX_INC(rcb->consumer_index, rcb->q_depth);
+ BNA_QE_INDX_INC(ccb->producer_index, ccb->q_depth);
+ next_cmpl = &cq[ccb->producer_index];
+
prefetch(next_cmpl);
flags = ntohl(cmpl->flags);
@@ -493,16 +437,12 @@ next:
cmpl = next_cmpl;
}
- BNA_QE_INDX_ADD(ccb->producer_index, wis, ccb->q_depth);
-
if (likely(test_bit(BNAD_RXQ_STARTED, &ccb->rcb[0]->flags)))
bna_ib_ack_disable_irq(ccb->i_dbell, packets);
- bnad_refill_rxq(bnad, ccb->rcb[0]);
+ bnad_rxq_post(bnad, ccb->rcb[0]);
if (ccb->rcb[1])
- bnad_refill_rxq(bnad, ccb->rcb[1]);
-
- clear_bit(BNAD_FP_IN_RX_PATH, &rx_ctrl->flags);
+ bnad_rxq_post(bnad, ccb->rcb[1]);
return packets;
}
@@ -777,12 +717,9 @@ bnad_cb_tcb_setup(struct bnad *bnad, struct bna_tcb *tcb)
{
struct bnad_tx_info *tx_info =
(struct bnad_tx_info *)tcb->txq->tx->priv;
- struct bnad_unmap_q *unmap_q = tcb->unmap_q;
+ tcb->priv = tcb;
tx_info->tcb[tcb->id] = tcb;
- unmap_q->producer_index = 0;
- unmap_q->consumer_index = 0;
- unmap_q->q_depth = BNAD_TX_UNMAPQ_DEPTH;
}
static void
@@ -796,16 +733,6 @@ bnad_cb_tcb_destroy(struct bnad *bnad, struct bna_tcb *tcb)
}
static void
-bnad_cb_rcb_setup(struct bnad *bnad, struct bna_rcb *rcb)
-{
- struct bnad_unmap_q *unmap_q = rcb->unmap_q;
-
- unmap_q->producer_index = 0;
- unmap_q->consumer_index = 0;
- unmap_q->q_depth = BNAD_RX_UNMAPQ_DEPTH;
-}
-
-static void
bnad_cb_ccb_setup(struct bnad *bnad, struct bna_ccb *ccb)
{
struct bnad_rx_info *rx_info =
@@ -891,10 +818,9 @@ bnad_tx_cleanup(struct delayed_work *work)
struct bnad_tx_info *tx_info =
container_of(work, struct bnad_tx_info, tx_cleanup_work);
struct bnad *bnad = NULL;
- struct bnad_unmap_q *unmap_q;
struct bna_tcb *tcb;
unsigned long flags;
- uint32_t i, pending = 0;
+ u32 i, pending = 0;
for (i = 0; i < BNAD_MAX_TXQ_PER_TX; i++) {
tcb = tx_info->tcb[i];
@@ -910,10 +836,6 @@ bnad_tx_cleanup(struct delayed_work *work)
bnad_txq_cleanup(bnad, tcb);
- unmap_q = tcb->unmap_q;
- unmap_q->producer_index = 0;
- unmap_q->consumer_index = 0;
-
smp_mb__before_clear_bit();
clear_bit(BNAD_TXQ_FREE_SENT, &tcb->flags);
}
@@ -929,7 +851,6 @@ bnad_tx_cleanup(struct delayed_work *work)
spin_unlock_irqrestore(&bnad->bna_lock, flags);
}
-
static void
bnad_cb_tx_cleanup(struct bnad *bnad, struct bna_tx *tx)
{
@@ -978,7 +899,7 @@ bnad_rx_cleanup(void *work)
struct bnad_rx_ctrl *rx_ctrl;
struct bnad *bnad = NULL;
unsigned long flags;
- uint32_t i;
+ u32 i;
for (i = 0; i < BNAD_MAX_RXP_PER_RX; i++) {
rx_ctrl = &rx_info->rx_ctrl[i];
@@ -1035,7 +956,6 @@ bnad_cb_rx_post(struct bnad *bnad, struct bna_rx *rx)
struct bna_ccb *ccb;
struct bna_rcb *rcb;
struct bnad_rx_ctrl *rx_ctrl;
- struct bnad_unmap_q *unmap_q;
int i;
int j;
@@ -1054,17 +974,7 @@ bnad_cb_rx_post(struct bnad *bnad, struct bna_rx *rx)
set_bit(BNAD_RXQ_STARTED, &rcb->flags);
set_bit(BNAD_RXQ_POST_OK, &rcb->flags);
- unmap_q = rcb->unmap_q;
-
- /* Now allocate & post buffers for this RCB */
- /* !!Allocation in callback context */
- if (!test_and_set_bit(BNAD_RXQ_REFILL, &rcb->flags)) {
- if (BNA_QE_FREE_CNT(unmap_q, unmap_q->q_depth)
- >> BNAD_RXQ_REFILL_THRESHOLD_SHIFT)
- bnad_rxq_post(bnad, rcb);
- smp_mb__before_clear_bit();
- clear_bit(BNAD_RXQ_REFILL, &rcb->flags);
- }
+ bnad_rxq_post(bnad, rcb);
}
}
}
@@ -1788,10 +1698,9 @@ bnad_setup_tx(struct bnad *bnad, u32 tx_id)
spin_unlock_irqrestore(&bnad->bna_lock, flags);
/* Fill Unmap Q memory requirements */
- BNAD_FILL_UNMAPQ_MEM_REQ(
- &res_info[BNA_TX_RES_MEM_T_UNMAPQ],
- bnad->num_txq_per_tx,
- BNAD_TX_UNMAPQ_DEPTH);
+ BNAD_FILL_UNMAPQ_MEM_REQ(&res_info[BNA_TX_RES_MEM_T_UNMAPQ],
+ bnad->num_txq_per_tx, (sizeof(struct bnad_tx_unmap) *
+ bnad->txq_depth));
/* Allocate resources */
err = bnad_tx_res_alloc(bnad, res_info, tx_id);
@@ -1929,7 +1838,7 @@ bnad_setup_rx(struct bnad *bnad, u32 rx_id)
&res_info[BNA_RX_RES_T_INTR].res_u.intr_info;
struct bna_rx_config *rx_config = &bnad->rx_config[rx_id];
static const struct bna_rx_event_cbfn rx_cbfn = {
- .rcb_setup_cbfn = bnad_cb_rcb_setup,
+ .rcb_setup_cbfn = NULL,
.rcb_destroy_cbfn = NULL,
.ccb_setup_cbfn = bnad_cb_ccb_setup,
.ccb_destroy_cbfn = bnad_cb_ccb_destroy,
@@ -1951,11 +1860,10 @@ bnad_setup_rx(struct bnad *bnad, u32 rx_id)
spin_unlock_irqrestore(&bnad->bna_lock, flags);
/* Fill Unmap Q memory requirements */
- BNAD_FILL_UNMAPQ_MEM_REQ(
- &res_info[BNA_RX_RES_MEM_T_UNMAPQ],
- rx_config->num_paths +
- ((rx_config->rxp_type == BNA_RXP_SINGLE) ? 0 :
- rx_config->num_paths), BNAD_RX_UNMAPQ_DEPTH);
+ BNAD_FILL_UNMAPQ_MEM_REQ(&res_info[BNA_RX_RES_MEM_T_UNMAPQ],
+ rx_config->num_paths + ((rx_config->rxp_type == BNA_RXP_SINGLE)
+ ? 0 : rx_config->num_paths), (bnad->rxq_depth *
+ sizeof(struct bnad_rx_unmap)));
/* Allocate resource */
err = bnad_rx_res_alloc(bnad, res_info, rx_id);
@@ -2536,125 +2444,34 @@ bnad_stop(struct net_device *netdev)
}
/* TX */
-/*
- * bnad_start_xmit : Netdev entry point for Transmit
- * Called under lock held by net_device
- */
-static netdev_tx_t
-bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
+/* Returns 0 for success */
+static int
+bnad_txq_wi_prepare(struct bnad *bnad, struct bna_tcb *tcb,
+ struct sk_buff *skb, struct bna_txq_entry *txqent)
{
- struct bnad *bnad = netdev_priv(netdev);
- u32 txq_id = 0;
- struct bna_tcb *tcb = bnad->tx_info[0].tcb[txq_id];
-
- u16 txq_prod, vlan_tag = 0;
- u32 unmap_prod, wis, wis_used, wi_range;
- u32 vectors, vect_id, i, acked;
- int err;
- unsigned int len;
- u32 gso_size;
-
- struct bnad_unmap_q *unmap_q = tcb->unmap_q;
- dma_addr_t dma_addr;
- struct bna_txq_entry *txqent;
- u16 flags;
-
- if (unlikely(skb->len <= ETH_HLEN)) {
- dev_kfree_skb(skb);
- BNAD_UPDATE_CTR(bnad, tx_skb_too_short);
- return NETDEV_TX_OK;
- }
- if (unlikely(skb_headlen(skb) > BFI_TX_MAX_DATA_PER_VECTOR)) {
- dev_kfree_skb(skb);
- BNAD_UPDATE_CTR(bnad, tx_skb_headlen_too_long);
- return NETDEV_TX_OK;
- }
- if (unlikely(skb_headlen(skb) == 0)) {
- dev_kfree_skb(skb);
- BNAD_UPDATE_CTR(bnad, tx_skb_headlen_zero);
- return NETDEV_TX_OK;
- }
-
- /*
- * Takes care of the Tx that is scheduled between clearing the flag
- * and the netif_tx_stop_all_queues() call.
- */
- if (unlikely(!test_bit(BNAD_TXQ_TX_STARTED, &tcb->flags))) {
- dev_kfree_skb(skb);
- BNAD_UPDATE_CTR(bnad, tx_skb_stopping);
- return NETDEV_TX_OK;
- }
-
- vectors = 1 + skb_shinfo(skb)->nr_frags;
- if (unlikely(vectors > BFI_TX_MAX_VECTORS_PER_PKT)) {
- dev_kfree_skb(skb);
- BNAD_UPDATE_CTR(bnad, tx_skb_max_vectors);
- return NETDEV_TX_OK;
- }
- wis = BNA_TXQ_WI_NEEDED(vectors); /* 4 vectors per work item */
- acked = 0;
- if (unlikely(wis > BNA_QE_FREE_CNT(tcb, tcb->q_depth) ||
- vectors > BNA_QE_FREE_CNT(unmap_q, unmap_q->q_depth))) {
- if ((u16) (*tcb->hw_consumer_index) !=
- tcb->consumer_index &&
- !test_and_set_bit(BNAD_TXQ_FREE_SENT, &tcb->flags)) {
- acked = bnad_txcmpl_process(bnad, tcb);
- if (likely(test_bit(BNAD_TXQ_TX_STARTED, &tcb->flags)))
- bna_ib_ack(tcb->i_dbell, acked);
- smp_mb__before_clear_bit();
- clear_bit(BNAD_TXQ_FREE_SENT, &tcb->flags);
- } else {
- netif_stop_queue(netdev);
- BNAD_UPDATE_CTR(bnad, netif_queue_stop);
- }
-
- smp_mb();
- /*
- * Check again to deal with race condition between
- * netif_stop_queue here, and netif_wake_queue in
- * interrupt handler which is not inside netif tx lock.
- */
- if (likely
- (wis > BNA_QE_FREE_CNT(tcb, tcb->q_depth) ||
- vectors > BNA_QE_FREE_CNT(unmap_q, unmap_q->q_depth))) {
- BNAD_UPDATE_CTR(bnad, netif_queue_stop);
- return NETDEV_TX_BUSY;
- } else {
- netif_wake_queue(netdev);
- BNAD_UPDATE_CTR(bnad, netif_queue_wakeup);
- }
- }
-
- unmap_prod = unmap_q->producer_index;
- flags = 0;
-
- txq_prod = tcb->producer_index;
- BNA_TXQ_QPGE_PTR_GET(txq_prod, tcb->sw_qpt, txqent, wi_range);
- txqent->hdr.wi.reserved = 0;
- txqent->hdr.wi.num_vectors = vectors;
+ u16 flags = 0;
+ u32 gso_size;
+ u16 vlan_tag = 0;
if (vlan_tx_tag_present(skb)) {
- vlan_tag = (u16) vlan_tx_tag_get(skb);
+ vlan_tag = (u16)vlan_tx_tag_get(skb);
flags |= (BNA_TXQ_WI_CF_INS_PRIO | BNA_TXQ_WI_CF_INS_VLAN);
}
if (test_bit(BNAD_RF_CEE_RUNNING, &bnad->run_flags)) {
- vlan_tag =
- (tcb->priority & 0x7) << 13 | (vlan_tag & 0x1fff);
+ vlan_tag = ((tcb->priority & 0x7) << VLAN_PRIO_SHIFT)
+ | (vlan_tag & 0x1fff);
flags |= (BNA_TXQ_WI_CF_INS_PRIO | BNA_TXQ_WI_CF_INS_VLAN);
}
-
txqent->hdr.wi.vlan_tag = htons(vlan_tag);
if (skb_is_gso(skb)) {
gso_size = skb_shinfo(skb)->gso_size;
-
- if (unlikely(gso_size > netdev->mtu)) {
- dev_kfree_skb(skb);
+ if (unlikely(gso_size > bnad->netdev->mtu)) {
BNAD_UPDATE_CTR(bnad, tx_skb_mss_too_long);
- return NETDEV_TX_OK;
+ return -EINVAL;
}
if (unlikely((gso_size + skb_transport_offset(skb) +
- tcp_hdrlen(skb)) >= skb->len)) {
+ tcp_hdrlen(skb)) >= skb->len)) {
txqent->hdr.wi.opcode =
__constant_htons(BNA_TXQ_WI_SEND);
txqent->hdr.wi.lso_mss = 0;
@@ -2665,25 +2482,22 @@ bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
txqent->hdr.wi.lso_mss = htons(gso_size);
}
- err = bnad_tso_prepare(bnad, skb);
- if (unlikely(err)) {
- dev_kfree_skb(skb);
+ if (bnad_tso_prepare(bnad, skb)) {
BNAD_UPDATE_CTR(bnad, tx_skb_tso_prepare);
- return NETDEV_TX_OK;
+ return -EINVAL;
}
+
flags |= (BNA_TXQ_WI_CF_IP_CKSUM | BNA_TXQ_WI_CF_TCP_CKSUM);
txqent->hdr.wi.l4_hdr_size_n_offset =
- htons(BNA_TXQ_WI_L4_HDR_N_OFFSET
- (tcp_hdrlen(skb) >> 2,
- skb_transport_offset(skb)));
- } else {
+ htons(BNA_TXQ_WI_L4_HDR_N_OFFSET(
+ tcp_hdrlen(skb) >> 2, skb_transport_offset(skb)));
+ } else {
txqent->hdr.wi.opcode = __constant_htons(BNA_TXQ_WI_SEND);
txqent->hdr.wi.lso_mss = 0;
- if (unlikely(skb->len > (netdev->mtu + ETH_HLEN))) {
- dev_kfree_skb(skb);
+ if (unlikely(skb->len > (bnad->netdev->mtu + ETH_HLEN))) {
BNAD_UPDATE_CTR(bnad, tx_skb_non_tso_too_long);
- return NETDEV_TX_OK;
+ return -EINVAL;
}
if (skb->ip_summed == CHECKSUM_PARTIAL) {
@@ -2691,11 +2505,13 @@ bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
if (skb->protocol == __constant_htons(ETH_P_IP))
proto = ip_hdr(skb)->protocol;
+#ifdef NETIF_F_IPV6_CSUM
else if (skb->protocol ==
__constant_htons(ETH_P_IPV6)) {
/* nexthdr may not be TCP immediately. */
proto = ipv6_hdr(skb)->nexthdr;
}
+#endif
if (proto == IPPROTO_TCP) {
flags |= BNA_TXQ_WI_CF_TCP_CKSUM;
txqent->hdr.wi.l4_hdr_size_n_offset =
@@ -2705,12 +2521,11 @@ bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
BNAD_UPDATE_CTR(bnad, tcpcsum_offload);
if (unlikely(skb_headlen(skb) <
- skb_transport_offset(skb) + tcp_hdrlen(skb))) {
- dev_kfree_skb(skb);
+ skb_transport_offset(skb) +
+ tcp_hdrlen(skb))) {
BNAD_UPDATE_CTR(bnad, tx_skb_tcp_hdr);
- return NETDEV_TX_OK;
+ return -EINVAL;
}
-
} else if (proto == IPPROTO_UDP) {
flags |= BNA_TXQ_WI_CF_UDP_CKSUM;
txqent->hdr.wi.l4_hdr_size_n_offset =
@@ -2719,51 +2534,149 @@ bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
BNAD_UPDATE_CTR(bnad, udpcsum_offload);
if (unlikely(skb_headlen(skb) <
- skb_transport_offset(skb) +
+ skb_transport_offset(skb) +
sizeof(struct udphdr))) {
- dev_kfree_skb(skb);
BNAD_UPDATE_CTR(bnad, tx_skb_udp_hdr);
- return NETDEV_TX_OK;
+ return -EINVAL;
}
} else {
- dev_kfree_skb(skb);
+
BNAD_UPDATE_CTR(bnad, tx_skb_csum_err);
- return NETDEV_TX_OK;
+ return -EINVAL;
}
- } else {
+ } else
txqent->hdr.wi.l4_hdr_size_n_offset = 0;
- }
}
txqent->hdr.wi.flags = htons(flags);
-
txqent->hdr.wi.frame_length = htonl(skb->len);
- unmap_q->unmap_array[unmap_prod].skb = skb;
+ return 0;
+}
+
+/*
+ * bnad_start_xmit : Netdev entry point for Transmit
+ * Called under lock held by net_device
+ */
+static netdev_tx_t
+bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
+{
+ struct bnad *bnad = netdev_priv(netdev);
+ u32 txq_id = 0;
+ struct bna_tcb *tcb = NULL;
+ struct bnad_tx_unmap *unmap_q, *unmap, *head_unmap;
+ u32 prod, q_depth, vect_id;
+ u32 wis, vectors, len;
+ int i;
+ dma_addr_t dma_addr;
+ struct bna_txq_entry *txqent;
+
len = skb_headlen(skb);
- txqent->vector[0].length = htons(len);
- dma_addr = dma_map_single(&bnad->pcidev->dev, skb->data,
- skb_headlen(skb), DMA_TO_DEVICE);
- dma_unmap_addr_set(&unmap_q->unmap_array[unmap_prod], dma_addr,
- dma_addr);
- BNA_SET_DMA_ADDR(dma_addr, &txqent->vector[0].host_addr);
- BNA_QE_INDX_ADD(unmap_prod, 1, unmap_q->q_depth);
+ /* Sanity checks for the skb */
+
+ if (unlikely(skb->len <= ETH_HLEN)) {
+ dev_kfree_skb(skb);
+ BNAD_UPDATE_CTR(bnad, tx_skb_too_short);
+ return NETDEV_TX_OK;
+ }
+ if (unlikely(len > BFI_TX_MAX_DATA_PER_VECTOR)) {
+ dev_kfree_skb(skb);
+ BNAD_UPDATE_CTR(bnad, tx_skb_headlen_zero);
+ return NETDEV_TX_OK;
+ }
+ if (unlikely(len == 0)) {
+ dev_kfree_skb(skb);
+ BNAD_UPDATE_CTR(bnad, tx_skb_headlen_zero);
+ return NETDEV_TX_OK;
+ }
+
+ tcb = bnad->tx_info[0].tcb[txq_id];
+ q_depth = tcb->q_depth;
+ prod = tcb->producer_index;
- vect_id = 0;
- wis_used = 1;
+ unmap_q = tcb->unmap_q;
- for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+ /*
+ * Takes care of the Tx that is scheduled between clearing the flag
+ * and the netif_tx_stop_all_queues() call.
+ */
+ if (unlikely(!test_bit(BNAD_TXQ_TX_STARTED, &tcb->flags))) {
+ dev_kfree_skb(skb);
+ BNAD_UPDATE_CTR(bnad, tx_skb_stopping);
+ return NETDEV_TX_OK;
+ }
+
+ vectors = 1 + skb_shinfo(skb)->nr_frags;
+ wis = BNA_TXQ_WI_NEEDED(vectors); /* 4 vectors per work item */
+
+ if (unlikely(vectors > BFI_TX_MAX_VECTORS_PER_PKT)) {
+ dev_kfree_skb(skb);
+ BNAD_UPDATE_CTR(bnad, tx_skb_max_vectors);
+ return NETDEV_TX_OK;
+ }
+
+ /* Check for available TxQ resources */
+ if (unlikely(wis > BNA_QE_FREE_CNT(tcb, q_depth))) {
+ if ((*tcb->hw_consumer_index != tcb->consumer_index) &&
+ !test_and_set_bit(BNAD_TXQ_FREE_SENT, &tcb->flags)) {
+ u32 sent;
+ sent = bnad_txcmpl_process(bnad, tcb);
+ if (likely(test_bit(BNAD_TXQ_TX_STARTED, &tcb->flags)))
+ bna_ib_ack(tcb->i_dbell, sent);
+ smp_mb__before_clear_bit();
+ clear_bit(BNAD_TXQ_FREE_SENT, &tcb->flags);
+ } else {
+ netif_stop_queue(netdev);
+ BNAD_UPDATE_CTR(bnad, netif_queue_stop);
+ }
+
+ smp_mb();
+ /*
+ * Check again to deal with race condition between
+ * netif_stop_queue here, and netif_wake_queue in
+ * interrupt handler which is not inside netif tx lock.
+ */
+ if (likely(wis > BNA_QE_FREE_CNT(tcb, q_depth))) {
+ BNAD_UPDATE_CTR(bnad, netif_queue_stop);
+ return NETDEV_TX_BUSY;
+ } else {
+ netif_wake_queue(netdev);
+ BNAD_UPDATE_CTR(bnad, netif_queue_wakeup);
+ }
+ }
+
+ txqent = &((struct bna_txq_entry *)tcb->sw_q)[prod];
+ head_unmap = &unmap_q[prod];
+
+ /* Program the opcode, flags, frame_len, num_vectors in WI */
+ if (bnad_txq_wi_prepare(bnad, tcb, skb, txqent)) {
+ dev_kfree_skb(skb);
+ return NETDEV_TX_OK;
+ }
+ txqent->hdr.wi.reserved = 0;
+ txqent->hdr.wi.num_vectors = vectors;
+
+ head_unmap->skb = skb;
+ head_unmap->nvecs = 0;
+
+ /* Program the vectors */
+ unmap = head_unmap;
+ dma_addr = dma_map_single(&bnad->pcidev->dev, skb->data,
+ len, DMA_TO_DEVICE);
+ BNA_SET_DMA_ADDR(dma_addr, &txqent->vector[0].host_addr);
+ txqent->vector[0].length = htons(len);
+ dma_unmap_addr_set(&unmap->vectors[0], dma_addr, dma_addr);
+ head_unmap->nvecs++;
+
+ for (i = 0, vect_id = 0; i < vectors - 1; i++) {
const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
u16 size = skb_frag_size(frag);
if (unlikely(size == 0)) {
- unmap_prod = unmap_q->producer_index;
-
- unmap_prod = bnad_pci_unmap_skb(&bnad->pcidev->dev,
- unmap_q->unmap_array,
- unmap_prod, unmap_q->q_depth, skb,
- i);
+ /* Undo the changes starting at tcb->producer_index */
+ bnad_tx_buff_unmap(bnad, unmap_q, q_depth,
+ tcb->producer_index);
dev_kfree_skb(skb);
BNAD_UPDATE_CTR(bnad, tx_skb_frag_zero);
return NETDEV_TX_OK;
@@ -2771,47 +2684,35 @@ bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
len += size;
- if (++vect_id == BFI_TX_MAX_VECTORS_PER_WI) {
+ vect_id++;
+ if (vect_id == BFI_TX_MAX_VECTORS_PER_WI) {
vect_id = 0;
- if (--wi_range)
- txqent++;
- else {
- BNA_QE_INDX_ADD(txq_prod, wis_used,
- tcb->q_depth);
- wis_used = 0;
- BNA_TXQ_QPGE_PTR_GET(txq_prod, tcb->sw_qpt,
- txqent, wi_range);
- }
- wis_used++;
+ BNA_QE_INDX_INC(prod, q_depth);
+ txqent = &((struct bna_txq_entry *)tcb->sw_q)[prod];
txqent->hdr.wi_ext.opcode =
__constant_htons(BNA_TXQ_WI_EXTENSION);
+ unmap = &unmap_q[prod];
}
- BUG_ON(!(size <= BFI_TX_MAX_DATA_PER_VECTOR));
- txqent->vector[vect_id].length = htons(size);
dma_addr = skb_frag_dma_map(&bnad->pcidev->dev, frag,
0, size, DMA_TO_DEVICE);
- dma_unmap_addr_set(&unmap_q->unmap_array[unmap_prod], dma_addr,
- dma_addr);
BNA_SET_DMA_ADDR(dma_addr, &txqent->vector[vect_id].host_addr);
- BNA_QE_INDX_ADD(unmap_prod, 1, unmap_q->q_depth);
+ txqent->vector[vect_id].length = htons(size);
+ dma_unmap_addr_set(&unmap->vectors[vect_id], dma_addr,
+ dma_addr);
+ head_unmap->nvecs++;
}
if (unlikely(len != skb->len)) {
- unmap_prod = unmap_q->producer_index;
-
- unmap_prod = bnad_pci_unmap_skb(&bnad->pcidev->dev,
- unmap_q->unmap_array, unmap_prod,
- unmap_q->q_depth, skb,
- skb_shinfo(skb)->nr_frags);
+ /* Undo the changes starting at tcb->producer_index */
+ bnad_tx_buff_unmap(bnad, unmap_q, q_depth, tcb->producer_index);
dev_kfree_skb(skb);
BNAD_UPDATE_CTR(bnad, tx_skb_len_mismatch);
return NETDEV_TX_OK;
}
- unmap_q->producer_index = unmap_prod;
- BNA_QE_INDX_ADD(txq_prod, wis_used, tcb->q_depth);
- tcb->producer_index = txq_prod;
+ BNA_QE_INDX_INC(prod, q_depth);
+ tcb->producer_index = prod;
smp_mb();
@@ -3332,7 +3233,6 @@ bnad_pci_probe(struct pci_dev *pdev,
if (err)
goto res_free;
-
/* Set up timers */
setup_timer(&bnad->bna.ioceth.ioc.ioc_timer, bnad_ioc_timeout,
((unsigned long)bnad));
diff --git a/drivers/net/ethernet/brocade/bna/bnad.h b/drivers/net/ethernet/brocade/bna/bnad.h
index 65fe74e..db132c9 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.h
+++ b/drivers/net/ethernet/brocade/bna/bnad.h
@@ -83,12 +83,9 @@ struct bnad_rx_ctrl {
#define BNAD_IOCETH_TIMEOUT 10000
-#define BNAD_MAX_Q_DEPTH 0x10000
-#define BNAD_MIN_Q_DEPTH 0x200
-
-#define BNAD_MAX_RXQ_DEPTH (BNAD_MAX_Q_DEPTH / bnad_rxqs_per_cq)
-/* keeping MAX TX and RX Q depth equal */
-#define BNAD_MAX_TXQ_DEPTH BNAD_MAX_RXQ_DEPTH
+#define BNAD_MIN_Q_DEPTH 512
+#define BNAD_MAX_RXQ_DEPTH 2048
+#define BNAD_MAX_TXQ_DEPTH 2048
#define BNAD_JUMBO_MTU 9000
@@ -101,9 +98,8 @@ struct bnad_rx_ctrl {
#define BNAD_TXQ_TX_STARTED 1
/* Bit positions for rcb->flags */
-#define BNAD_RXQ_REFILL 0
-#define BNAD_RXQ_STARTED 1
-#define BNAD_RXQ_POST_OK 2
+#define BNAD_RXQ_STARTED 0
+#define BNAD_RXQ_POST_OK 1
/* Resource limits */
#define BNAD_NUM_TXQ (bnad->num_tx * bnad->num_txq_per_tx)
@@ -221,18 +217,24 @@ struct bnad_rx_info {
struct work_struct rx_cleanup_work;
} ____cacheline_aligned;
-/* Unmap queues for Tx / Rx cleanup */
-struct bnad_skb_unmap {
+struct bnad_tx_vector {
+ DEFINE_DMA_UNMAP_ADDR(dma_addr);
+};
+
+struct bnad_tx_unmap {
struct sk_buff *skb;
+ u32 nvecs;
+ struct bnad_tx_vector vectors[BFI_TX_MAX_VECTORS_PER_WI];
+};
+
+struct bnad_rx_vector {
DEFINE_DMA_UNMAP_ADDR(dma_addr);
+ u32 len;
};
-struct bnad_unmap_q {
- u32 producer_index;
- u32 consumer_index;
- u32 q_depth;
- /* This should be the last one */
- struct bnad_skb_unmap unmap_array[1];
+struct bnad_rx_unmap {
+ struct sk_buff *skb;
+ struct bnad_rx_vector vector;
};
/* Bit mask values for bnad->cfg_flags */
@@ -252,11 +254,6 @@ struct bnad_unmap_q {
#define BNAD_RF_STATS_TIMER_RUNNING 5
#define BNAD_RF_TX_PRIO_SET 6
-
-/* Define for Fast Path flags */
-/* Defined as bit positions */
-#define BNAD_FP_IN_RX_PATH 0
-
struct bnad {
struct net_device *netdev;
u32 id;
--
1.7.1
^ permalink raw reply related
* [net-next 3/7] bna: TX Intr Coalescing Fix
From: Rasesh Mody @ 2012-12-10 21:42 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
In-Reply-To: <1355175725-19202-1-git-send-email-rmody@brocade.com>
Change Details:
For Tx IB, IPM was enabled with inter_pkt_timeo of 0. This caused the
Tx IB not to generate interrupt till inter_pkt_count of packets have been
received. Correct definition for BFI_TX_INTERPKT_TIMEO & BFI_TX_INTERPKT_COUNT
Signed-off-by: Rasesh Mody <rmody@brocade.com>
---
drivers/net/ethernet/brocade/bna/bna_hw_defs.h | 3 ++-
drivers/net/ethernet/brocade/bna/bna_tx_rx.c | 2 +-
2 files changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/brocade/bna/bna_hw_defs.h b/drivers/net/ethernet/brocade/bna/bna_hw_defs.h
index b8c4e21..af3f7bb 100644
--- a/drivers/net/ethernet/brocade/bna/bna_hw_defs.h
+++ b/drivers/net/ethernet/brocade/bna/bna_hw_defs.h
@@ -46,7 +46,8 @@
#define BFI_MAX_INTERPKT_COUNT 0xFF
#define BFI_MAX_INTERPKT_TIMEO 0xF /* in 0.5us units */
#define BFI_TX_COALESCING_TIMEO 20 /* 20 * 5 = 100us */
-#define BFI_TX_INTERPKT_COUNT 32
+#define BFI_TX_INTERPKT_COUNT 12 /* Pkt Cnt = 12 */
+#define BFI_TX_INTERPKT_TIMEO 15 /* 15 * 0.5 = 7.5us */
#define BFI_RX_COALESCING_TIMEO 12 /* 12 * 5 = 60us */
#define BFI_RX_INTERPKT_COUNT 6 /* Pkt Cnt = 6 */
#define BFI_RX_INTERPKT_TIMEO 3 /* 3 * 0.5 = 1.5us */
diff --git a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
index bb5467b..4df6d4b 100644
--- a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
+++ b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
@@ -3569,7 +3569,7 @@ bna_tx_create(struct bna *bna, struct bnad *bnad,
if (intr_info->intr_type == BNA_INTR_T_INTX)
txq->ib.intr_vector = (1 << txq->ib.intr_vector);
txq->ib.coalescing_timeo = tx_cfg->coalescing_timeo;
- txq->ib.interpkt_timeo = 0; /* Not used */
+ txq->ib.interpkt_timeo = BFI_TX_INTERPKT_TIMEO;
txq->ib.interpkt_count = BFI_TX_INTERPKT_COUNT;
/* TCB */
--
1.7.1
^ permalink raw reply related
* [net-next 1/7] bna: Code Cleanup and Enhancements
From: Rasesh Mody @ 2012-12-10 21:41 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
In-Reply-To: <1355175725-19202-1-git-send-email-rmody@brocade.com>
Change details:
- Remove unnecessary prefetch
- Simplify checking & comparison of CQ flags
- Dereference & store unmap_array, unmap_cons & current unmap_array
element only once
- Make structures tx_config & rx_config cache line aligned.
Signed-off-by: Rasesh Mody <rmody@brocade.com>
---
drivers/net/ethernet/brocade/bna/bnad.c | 45 ++++++++++++++++++++-----------
drivers/net/ethernet/brocade/bna/bnad.h | 4 +-
2 files changed, 31 insertions(+), 18 deletions(-)
diff --git a/drivers/net/ethernet/brocade/bna/bnad.c b/drivers/net/ethernet/brocade/bna/bnad.c
index b441f33..b7fb391 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.c
+++ b/drivers/net/ethernet/brocade/bna/bnad.c
@@ -210,7 +210,6 @@ bnad_txcmpl_process(struct bnad *bnad,
unmap_array = unmap_q->unmap_array;
unmap_cons = unmap_q->consumer_index;
- prefetch(&unmap_array[unmap_cons + 1]);
while (wis) {
skb = unmap_array[unmap_cons].skb;
@@ -383,6 +382,20 @@ bnad_refill_rxq(struct bnad *bnad, struct bna_rcb *rcb)
}
}
+#define flags_cksum_prot_mask (BNA_CQ_EF_IPV4 | BNA_CQ_EF_L3_CKSUM_OK | \
+ BNA_CQ_EF_IPV6 | \
+ BNA_CQ_EF_TCP | BNA_CQ_EF_UDP | \
+ BNA_CQ_EF_L4_CKSUM_OK)
+
+#define flags_tcp4 (BNA_CQ_EF_IPV4 | BNA_CQ_EF_L3_CKSUM_OK | \
+ BNA_CQ_EF_TCP | BNA_CQ_EF_L4_CKSUM_OK)
+#define flags_tcp6 (BNA_CQ_EF_IPV6 | \
+ BNA_CQ_EF_TCP | BNA_CQ_EF_L4_CKSUM_OK)
+#define flags_udp4 (BNA_CQ_EF_IPV4 | BNA_CQ_EF_L3_CKSUM_OK | \
+ BNA_CQ_EF_UDP | BNA_CQ_EF_L4_CKSUM_OK)
+#define flags_udp6 (BNA_CQ_EF_IPV6 | \
+ BNA_CQ_EF_UDP | BNA_CQ_EF_L4_CKSUM_OK)
+
static u32
bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
{
@@ -390,15 +403,12 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
struct bna_rcb *rcb = NULL;
unsigned int wi_range, packets = 0, wis = 0;
struct bnad_unmap_q *unmap_q;
- struct bnad_skb_unmap *unmap_array;
+ struct bnad_skb_unmap *unmap_array, *curr_ua;
struct sk_buff *skb;
- u32 flags, unmap_cons;
+ u32 flags, unmap_cons, masked_flags;
struct bna_pkt_rate *pkt_rt = &ccb->pkt_rate;
struct bnad_rx_ctrl *rx_ctrl = (struct bnad_rx_ctrl *)(ccb->ctrl);
- if (!test_bit(BNAD_RXQ_STARTED, &ccb->rcb[0]->flags))
- return 0;
-
prefetch(bnad->netdev);
BNA_CQ_QPGE_PTR_GET(ccb->producer_index, ccb->sw_qpt, cmpl,
wi_range);
@@ -416,12 +426,13 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
unmap_array = unmap_q->unmap_array;
unmap_cons = unmap_q->consumer_index;
- skb = unmap_array[unmap_cons].skb;
- BUG_ON(!(skb));
- unmap_array[unmap_cons].skb = NULL;
+ curr_ua = &unmap_array[unmap_cons];
+
+ skb = curr_ua->skb;
+ BUG_ON(!(skb));\
+ curr_ua->skb = NULL;
dma_unmap_single(&bnad->pcidev->dev,
- dma_unmap_addr(&unmap_array[unmap_cons],
- dma_addr),
+ dma_unmap_addr(curr_ua, dma_addr),
rcb->rxq->buffer_size,
DMA_FROM_DEVICE);
BNA_QE_INDX_ADD(unmap_q->consumer_index, 1, unmap_q->q_depth);
@@ -452,13 +463,15 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
}
skb_put(skb, ntohs(cmpl->length));
+
+ masked_flags = flags & flags_cksum_prot_mask;
+
if (likely
((bnad->netdev->features & NETIF_F_RXCSUM) &&
- (((flags & BNA_CQ_EF_IPV4) &&
- (flags & BNA_CQ_EF_L3_CKSUM_OK)) ||
- (flags & BNA_CQ_EF_IPV6)) &&
- (flags & (BNA_CQ_EF_TCP | BNA_CQ_EF_UDP)) &&
- (flags & BNA_CQ_EF_L4_CKSUM_OK)))
+ ((masked_flags == flags_tcp4) ||
+ (masked_flags == flags_udp4) ||
+ (masked_flags == flags_tcp6) ||
+ (masked_flags == flags_udp6))))
skb->ip_summed = CHECKSUM_UNNECESSARY;
else
skb_checksum_none_assert(skb);
diff --git a/drivers/net/ethernet/brocade/bna/bnad.h b/drivers/net/ethernet/brocade/bna/bnad.h
index d783392..65fe74e 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.h
+++ b/drivers/net/ethernet/brocade/bna/bnad.h
@@ -284,8 +284,8 @@ struct bnad {
u8 tx_coalescing_timeo;
u8 rx_coalescing_timeo;
- struct bna_rx_config rx_config[BNAD_MAX_RX];
- struct bna_tx_config tx_config[BNAD_MAX_TX];
+ struct bna_rx_config rx_config[BNAD_MAX_RX]____cacheline_aligned;
+ struct bna_tx_config tx_config[BNAD_MAX_TX]____cacheline_aligned;
void __iomem *bar0; /* BAR0 address */
--
1.7.1
^ permalink raw reply related
* [net-next 0/7] bna: Driver Version Updated to 3.1.2.1
From: Rasesh Mody @ 2012-12-10 21:41 UTC (permalink / raw)
To: davem, netdev
Cc: bhutchings, David.Laight, adapter_linux_open_src_team,
Rasesh Mody
Hi Dave,
Resubmitting the patch set, it addresses the review feedbacks.
The following patch-set includes Tx Rx changes, bug fixes, firmware
update, code cleanup and enhancements.
This also updates the BNA driver to v3.1.2.1.
The patches have been compiled and tested against 3.7.0-rc3.
Thanks,
Rasesh
Rasesh Mody (7):
bna: Code Cleanup and Enhancements
bna: Tx and Rx Optimizations
bna: TX Intr Coalescing Fix
bna: Rx Page Based Allocation
bna: Add RX State
bna: Firmware update
bna: Driver Version Updated to 3.1.2.1
drivers/net/ethernet/brocade/bna/bfi_enet.h | 1 +
drivers/net/ethernet/brocade/bna/bna.h | 2 +
drivers/net/ethernet/brocade/bna/bna_hw_defs.h | 3 +-
drivers/net/ethernet/brocade/bna/bna_tx_rx.c | 138 +++--
drivers/net/ethernet/brocade/bna/bna_types.h | 9 +-
drivers/net/ethernet/brocade/bna/bnad.c | 937 +++++++++++++----------
drivers/net/ethernet/brocade/bna/bnad.h | 66 +-
drivers/net/ethernet/brocade/bna/bnad_ethtool.c | 1 +
drivers/net/ethernet/brocade/bna/cna.h | 4 +-
9 files changed, 661 insertions(+), 500 deletions(-)
^ permalink raw reply
* [PATCH net-next] doc: Tighten-up and clarify description of tcp_fin_timeout
From: Rick Jones @ 2012-12-10 21:33 UTC (permalink / raw)
To: netdev, davem
From: Rick Jones <rick.jones2@hp.com>
The description for tcp_fin_timeout should be tigher and more clear.
In addition to being tighter, we should make the spelling of the
state name consistent with what utilities report, remove the now
dated reference to 2.2 and put the default in the consistent place.
Signed-off-by: Rick Jones <rick.jones2@hp.com>
diff --git a/Documentation/networking/ip-sysctl.txt b/Documentation/networking/ip-sysctl.txt
index 1b830ca..f486c86 100644
--- a/Documentation/networking/ip-sysctl.txt
+++ b/Documentation/networking/ip-sysctl.txt
@@ -224,15 +224,14 @@ tcp_fack - BOOLEAN
The value is not used, if tcp_sack is not enabled.
tcp_fin_timeout - INTEGER
- Time to hold socket in state FIN-WAIT-2, if it was closed
- by our side. Peer can be broken and never close its side,
- or even died unexpectedly. Default value is 60sec.
- Usual value used in 2.2 was 180 seconds, you may restore
- it, but remember that if your machine is even underloaded WEB server,
- you risk to overflow memory with kilotons of dead sockets,
- FIN-WAIT-2 sockets are less dangerous than FIN-WAIT-1,
- because they eat maximum 1.5K of memory, but they tend
- to live longer. Cf. tcp_max_orphans.
+ The length of time an orphaned (no longer referenced by any
+ application) connection will remain in the FIN_WAIT_2 state
+ before it is aborted at the local end. While a perfectly
+ valid "receive only" state for an un-orphaned connection, an
+ orphaned connection in FIN_WAIT_2 state could otherwise wait
+ forever for the remote to close its end of the connection.
+ Cf. tcp_max_orphans
+ Default: 60 seconds
tcp_frto - INTEGER
Enables Forward RTO-Recovery (F-RTO) defined in RFC4138.
^ permalink raw reply related
* netconsole fun
From: Peter Hurley @ 2012-12-10 14:17 UTC (permalink / raw)
To: netdev
Now that netpoll has been disabled for slaved devices, is there a
recommended method of running netconsole on a machine that has a slaved
device?
I ask because it seems pretty difficult to get netconsole running even
on a machine that has multiple physical interfaces, only one of which is
slaved.
I scoured the documentation but didn't find anything relevant (does
Documentation/networking/netconsole.txt need a patch?). AFAICT, the last
discussion ended here back in June '11
http://lkml.indiana.edu/hypermail/linux/kernel/1106.1/03185.html with
"Looks like we need to teach the tun-tap driver to support netpoll that
shouldn't be too hard. I'll send up a patch shortly
Neil"
Any help or pointers provided would be much appreciated.
Regards,
Peter Hurley
^ permalink raw reply
* Re: [RFC PATCH net-next 0/3 V4] net-tcp: TCP/IP stack bypass for loopback connections
From: David Miller @ 2012-12-10 21:02 UTC (permalink / raw)
To: wpan; +Cc: netdev, brutus
In-Reply-To: <cover.1354674151.git.wpan@redhat.com>
From: Weiping Pan <wpan@redhat.com>
Date: Wed, 5 Dec 2012 10:54:16 +0800
> Friends VS AF__UNIX
> Their call path are almost the same, but AF_UNIX uses its own send/recv codes
> with proper locks,
> so AF_UNIX's performance is much better than Friends.
While I understand the other portions of your analysis, this one
mystifies me.
In both cases, the sender has to queue the SKB onto the receiver's
queue. And in both cases, the sender takes the lock on that queue.
So the locking contention really ought to be similar if not identical.
The only difference is that AF_UNIX takes the unix_sk()->lock of the
remote socket around these operations.
If that is enough of a synchronizer to "fix" the contention or reduce
it, then this would be very easy to test by adding a friend lock to
tcp_sk().
^ 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