openwrt/target/linux/generic/patches-3.7/131-atm-fixes.patch

commit d7d3d8f1ee4435e32bc6c93187798b7e2e3170a9
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Thu Nov 29 23:28:30 2012 +0000

    solos-pci: remove list_vccs() debugging function
    
    No idea why we've gone so long dumping a list of VCCs with vci==0 on
    every ->open() call...
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit c93967bfd3a3dffa759a3f28370167bf3cdbc3d0
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Thu Nov 29 23:27:20 2012 +0000

    solos-pci: use GFP_KERNEL where possible, not GFP_ATOMIC
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit ad6999e17ae4f7b99f6d28f425ae970acb115347
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Thu Nov 29 23:15:30 2012 +0000

    solos-pci: clean up pclose() function
    
     - Flush pending TX skbs from the queue rather than waiting for them all to
       complete (suggested by Krzysztof Mazur <krzysiek@podlesie.net>).
     - Clear ATM_VF_ADDR only when the PKT_PCLOSE packet has been submitted.
     - Don't clear ATM_VF_READY at all — vcc_destroy_socket() does that for us.
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit 2547e97f29d6d69d567947d5ef90b6b4fbcaf565
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Wed Nov 28 10:15:05 2012 +0000

    pppoatm: optimise PPP channel wakeups after sock_owned_by_user()
    
    We don't need to schedule the wakeup tasklet on *every* unlock; only if we
    actually blocked the channel in the first place.
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
    Acked-by: Krzysztof Mazur <krzysiek@podlesie.net>

commit 990b0884a2e9668c08e9daa4d70a54d65329cb6f
Author: Krzysztof Mazur <krzysiek@podlesie.net>
Date:   Wed Nov 28 09:08:04 2012 +0100

    br2684: allow assign only on a connected socket
    
    The br2684 does not check if used vcc is in connected state,
    causing potential Oops in pppoatm_send() when vcc->send() is called
    on not fully connected socket.
    
    Now br2684 can be assigned only on connected sockets; otherwise
    -EINVAL error is returned.
    
    Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit f49b6da01f0abebb17f6241473d53018d923f6b0
Author: Nathan Williams <nathan@traverse.com.au>
Date:   Tue Nov 27 17:34:09 2012 +1100

    solos-pci: Fix leak of skb received for unknown vcc
    
    ... and ensure that the next skb is set up for RX in the DMA case.
    
    Signed-off-by: Nathan Williams <nathan@traverse.com.au>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit a61d37ff4a555886c4ebe31d2c6d893afb6f4d3c
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Wed Nov 28 00:46:45 2012 +0000

    br2684: fix module_put() race
    
    The br2684 code used module_put() during unassignment from vcc with
    hope that we have BKL. This assumption is no longer true.
    
    Now owner field in atmvcc is used to move this module_put()
    to vcc_destroy_socket().
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
    Acked-by: Krzysztof Mazur <krzysiek@podlesie.net>

commit c52f40629884ddc62c7af445fd5d620fdb466fb2
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Wed Nov 28 00:05:52 2012 +0000

    pppoatm: fix missing wakeup in pppoatm_send()
    
    Now that we can return zero from pppoatm_send() for reasons *other* than
    the queue being full, that means we can't depend on a subsequent call to
    pppoatm_pop() waking the queue, and we might leave it stalled
    indefinitely.
    
    Use the ->release_cb() callback to wake the queue after the sock is
    unlocked.
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
    Acked-by: Krzysztof Mazur <krzysiek@podlesie.net>

commit 35826e7372fe39b7db7930eda0267c82d68d1a4c
Author: David Woodhouse <dwmw2@infradead.org>
Date:   Tue Nov 27 23:28:36 2012 +0000

    br2684: don't send frames on not-ready vcc
    
    Avoid submitting packets to a vcc which is being closed. Things go badly
    wrong when the ->pop method gets later called after everything's been
    torn down.
    
    Use the ATM socket lock for synchronisation with vcc_destroy_socket(),
    which clears the ATM_VF_READY bit under the same lock. Otherwise, we
    could end up submitting a packet to the device driver even after its
    ->ops->close method has been called. And it could call the vcc's ->pop
    method after the protocol has been shut down. Which leads to a panic.
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
    Acked-by: Krzysztof Mazur <krzysiek@podlesie.net>

commit 7f940dde65de4a707f3dd723bb6ce9de90ca1eab
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Wed Nov 28 00:03:11 2012 +0000

    atm: add release_cb() callback to vcc
    
    The immediate use case for this is that it will allow us to ensure that a
    pppoatm queue is woken after it has to drop a packet due to the sock being
    locked.
    
    Note that 'release_cb' is called when the socket is *unlocked*. This is
    not to be confused with vcc_release() — which probably ought to be called
    vcc_close().
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
    Acked-by: Krzysztof Mazur <krzysiek@podlesie.net>

commit def1b2f9083f84d0a77730e537c76429914d17c1
Author: David Woodhouse <David.Woodhouse@intel.com>
Date:   Tue Nov 27 23:49:24 2012 +0000

    solos-pci: wait for pending TX to complete when releasing vcc
    
    We should no longer be calling the old pop routine for the vcc, after
    vcc_release() has completed. Make sure we wait for any pending TX skbs
    to complete, by waiting for our own PKT_PCLOSE control skb to be sent.
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit 397ff16dce53888ec693b3718640be2560204751
Author: Krzysztof Mazur <krzysiek@podlesie.net>
Date:   Tue Nov 6 23:17:02 2012 +0100

    pppoatm: do not inline pppoatm_may_send()
    
    The pppoatm_may_send() is quite heavy and it's called three times
    in pppoatm_send() and inlining costs more than 200 bytes of code
    (more than 10% of total pppoatm driver code size).
    
    add/remove: 1/0 grow/shrink: 0/1 up/down: 132/-367 (-235)
    function                                     old     new   delta
    pppoatm_may_send                               -     132    +132
    pppoatm_send                                 900     533    -367
    
    Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit 071d93931a75dc1f82f0baa9959613af81c5a032
Author: Krzysztof Mazur <krzysiek@podlesie.net>
Date:   Sat Nov 10 23:33:19 2012 +0100

    pppoatm: drop frames to not-ready vcc
    
    The vcc_destroy_socket() closes vcc before the protocol is detached
    from vcc by calling vcc->push() with NULL skb. This leaves some time
    window, where the protocol may call vcc->send() on closed vcc
    and crash.
    
    Now pppoatm_send(), like vcc_sendmsg(), checks for vcc flags that
    indicate that vcc is not ready. If the vcc is not ready we just
    drop frame. Queueing frames is much more complicated because we
    don't have callbacks that inform us about vcc flags changes.
    
    Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit 3ac108006fd7f20cb8fc8ea2287f1497bcda00a1
Author: Krzysztof Mazur <krzysiek@podlesie.net>
Date:   Tue Nov 6 23:17:00 2012 +0100

    pppoatm: take ATM socket lock in pppoatm_send()
    
    The pppoatm_send() does not take any lock that will prevent concurrent
    vcc_sendmsg(). This causes two problems:
    
    	- there is no locking between checking the send queue size
    	  with atm_may_send() and incrementing sk_wmem_alloc,
    	  and the real queue size can be a little higher than sk_sndbuf
    
    	- the vcc->sendmsg() can be called concurrently. I'm not sure
    	  if it's allowed. Some drivers (eni, nicstar, ...) seem
    	  to assume it will never happen.
    
    Now pppoatm_send() takes ATM socket lock, the same that is used
    in vcc_sendmsg() and other ATM socket functions. The pppoatm_send()
    is called with BH disabled, so bh_lock_sock() is used instead
    of lock_sock().
    
    Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Cc: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit e41faed9cde1acce657f75a0b19a1787e9850d3f
Author: Krzysztof Mazur <krzysiek@podlesie.net>
Date:   Tue Nov 6 23:16:59 2012 +0100

    pppoatm: fix module_put() race
    
    The pppoatm used module_put() during unassignment from vcc with
    hope that we have BKL. This assumption is no longer true.
    
    Now owner field in atmvcc is used to move this module_put()
    to vcc_destroy_socket().
    
    Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit 3b1a914595f3f9beb9e38ff3ddc7bdafa092ba22
Author: Krzysztof Mazur <krzysiek@podlesie.net>
Date:   Tue Nov 6 23:16:58 2012 +0100

    pppoatm: allow assign only on a connected socket
    
    The pppoatm does not check if used vcc is in connected state,
    causing an Oops in pppoatm_send() when vcc->send() is called
    on not fully connected socket.
    
    Now pppoatm can be assigned only on connected sockets; otherwise
    -EINVAL error is returned.
    
    Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Cc: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

commit ec809bd817dfa1905283468e4c813684ed4efe78
Author: Krzysztof Mazur <krzysiek@podlesie.net>
Date:   Tue Nov 6 23:16:57 2012 +0100

    atm: add owner of push() callback to atmvcc
    
    The atm is using atmvcc->push(vcc, NULL) callback to notify protocol
    that vcc will be closed and protocol must detach from it. This callback
    is usually used by protocol to decrement module usage count by module_put(),
    but it leaves small window then module is still used after module_put().
    
    Now the owner of push() callback is kept in atmvcc and
    module_put(atmvcc->owner) is called after the protocol is detached from vcc.
    
    Signed-off-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
    Acked-by: Chas Williams <chas@cmf.nrl.navy.mil>

commit ae088d663beebb3cad0e7abaac67ee61a7c578d5
Author: David Woodhouse <dwmw2@infradead.org>
Date:   Sun Nov 25 12:06:52 2012 +0000

    atm: br2684: Fix excessive queue bloat
    
    There's really no excuse for an additional wmem_default of buffering
    between the netdev queue and the ATM device. Two packets (one in-flight,
    and one ready to send) ought to be fine. It's not as if it should take
    long to get another from the netdev queue when we need it.
    
    If necessary we can make the queue space configurable later, but I don't
    think it's likely to be necessary.
    
    cf. commit 9d02daf754238adac48fa075ee79e7edd3d79ed3 (pppoatm: Fix
    excessive queue bloat) which did something very similar for PPPoATM.
    
    Note that there is a tremendously unlikely race condition which may
    result in qspace temporarily going negative. If a CPU running the
    br2684_pop() function goes off into the weeds for a long period of time
    after incrementing qspace to 1, but before calling netdev_wake_queue()...
    and another CPU ends up calling br2684_start_xmit() and *stopping* the
    queue again before the first CPU comes back, the netdev queue could
    end up being woken when qspace has already reached zero.
    
    An alternative approach to coping with this race would be to check in
    br2684_start_xmit() for qspace==0 and return NETDEV_TX_BUSY, but just
    using '> 0' and '< 1' for comparison instead of '== 0' and '!= 0' is
    simpler. It just warranted a mention of *why* we do it that way...
    
    Move the call to atmvcc->send() to happen *after* the accounting and
    potentially stopping the netdev queue, in br2684_xmit_vcc(). This matters
    if the ->send() call suffers an immediate failure, because it'll call
    br2684_pop() with the offending skb before returning. We want that to
    happen *after* we've done the initial accounting for the packet in
    question. Also make it return an appropriate success/failure indication
    while we're at it.
    
    Tested by running 'ping -l 1000 bottomless.aaisp.net.uk' from within my
    network, with only a single PPPoE-over-BR2684 link running. And after
    setting txqueuelen on the nas0 interface to something low (5, in fact).
    Before the patch, we'd see about 15 packets being queued and a resulting
    latency of ~56ms being reached. After the patch, we see only about 8,
    which is fairly much what we expect. And a max latency of ~36ms. On this
    OpenWRT box, wmem_default is 163840.
    
    Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
    Reviewed-by: Krzysztof Mazur <krzysiek@podlesie.net>
    Signed-off-by: David S. Miller <davem@davemloft.net>

--- a/drivers/atm/solos-pci.c
+++ b/drivers/atm/solos-pci.c
@@ -92,6 +92,7 @@ struct pkt_hdr {
 };
 
 struct solos_skb_cb {
+	struct completion c;
 	struct atm_vcc *vcc;
 	uint32_t dma_addr;
 };
@@ -164,7 +165,6 @@ static void fpga_queue(struct solos_card
 static uint32_t fpga_tx(struct solos_card *);
 static irqreturn_t solos_irq(int irq, void *dev_id);
 static struct atm_vcc* find_vcc(struct atm_dev *dev, short vpi, int vci);
-static int list_vccs(int vci);
 static int atm_init(struct solos_card *, struct device *);
 static void atm_remove(struct solos_card *);
 static int send_command(struct solos_card *card, int dev, const char *buf, size_t size);
@@ -710,7 +710,8 @@ void solos_bh(unsigned long card_arg)
 						dev_warn(&card->dev->dev, "Received packet for unknown VPI.VCI %d.%d on port %d\n",
 							 le16_to_cpu(header->vpi), le16_to_cpu(header->vci),
 							 port);
-					continue;
+					dev_kfree_skb_any(skb);
+					break;
 				}
 				atm_charge(vcc, skb->truesize);
 				vcc->push(vcc, skb);
@@ -790,44 +791,6 @@ static struct atm_vcc *find_vcc(struct a
 	return vcc;
 }
 
-static int list_vccs(int vci)
-{
-	struct hlist_head *head;
-	struct atm_vcc *vcc;
-	struct hlist_node *node;
-	struct sock *s;
-	int num_found = 0;
-	int i;
-
-	read_lock(&vcc_sklist_lock);
-	if (vci != 0){
-		head = &vcc_hash[vci & (VCC_HTABLE_SIZE -1)];
-		sk_for_each(s, node, head) {
-			num_found ++;
-			vcc = atm_sk(s);
-			printk(KERN_DEBUG "Device: %d Vpi: %d Vci: %d\n",
-			       vcc->dev->number,
-			       vcc->vpi,
-			       vcc->vci);
-		}
-	} else {
-		for(i = 0; i < VCC_HTABLE_SIZE; i++){
-			head = &vcc_hash[i];
-			sk_for_each(s, node, head) {
-				num_found ++;
-				vcc = atm_sk(s);
-				printk(KERN_DEBUG "Device: %d Vpi: %d Vci: %d\n",
-				       vcc->dev->number,
-				       vcc->vpi,
-				       vcc->vci);
-			}
-		}
-	}
-	read_unlock(&vcc_sklist_lock);
-	return num_found;
-}
-
-
 static int popen(struct atm_vcc *vcc)
 {
 	struct solos_card *card = vcc->dev->dev_data;
@@ -840,7 +803,7 @@ static int popen(struct atm_vcc *vcc)
 		return -EINVAL;
 	}
 
-	skb = alloc_skb(sizeof(*header), GFP_ATOMIC);
+	skb = alloc_skb(sizeof(*header), GFP_KERNEL);
 	if (!skb) {
 		if (net_ratelimit())
 			dev_warn(&card->dev->dev, "Failed to allocate sk_buff in popen()\n");
@@ -857,8 +820,6 @@ static int popen(struct atm_vcc *vcc)
 
 	set_bit(ATM_VF_ADDR, &vcc->flags);
 	set_bit(ATM_VF_READY, &vcc->flags);
-	list_vccs(0);
-
 
 	return 0;
 }
@@ -866,10 +827,21 @@ static int popen(struct atm_vcc *vcc)
 static void pclose(struct atm_vcc *vcc)
 {
 	struct solos_card *card = vcc->dev->dev_data;
-	struct sk_buff *skb;
+	unsigned char port = SOLOS_CHAN(vcc->dev);
+	struct sk_buff *skb, *tmpskb;
 	struct pkt_hdr *header;
 
-	skb = alloc_skb(sizeof(*header), GFP_ATOMIC);
+	/* Remove any yet-to-be-transmitted packets from the pending queue */
+	spin_lock(&card->tx_queue_lock);
+	skb_queue_walk_safe(&card->tx_queue[port], skb, tmpskb) {
+		if (SKB_CB(skb)->vcc == vcc) {
+			skb_unlink(skb, &card->tx_queue[port]);
+			solos_pop(vcc, skb);
+		}
+	}
+	spin_unlock(&card->tx_queue_lock);
+
+	skb = alloc_skb(sizeof(*header), GFP_KERNEL);
 	if (!skb) {
 		dev_warn(&card->dev->dev, "Failed to allocate sk_buff in pclose()\n");
 		return;
@@ -881,15 +853,21 @@ static void pclose(struct atm_vcc *vcc)
 	header->vci = cpu_to_le16(vcc->vci);
 	header->type = cpu_to_le16(PKT_PCLOSE);
 
-	fpga_queue(card, SOLOS_CHAN(vcc->dev), skb, NULL);
+	init_completion(&SKB_CB(skb)->c);
 
-	clear_bit(ATM_VF_ADDR, &vcc->flags);
-	clear_bit(ATM_VF_READY, &vcc->flags);
+	fpga_queue(card, port, skb, NULL);
+
+	if (!wait_for_completion_timeout(&SKB_CB(skb)->c, 5 * HZ))
+		dev_warn(&card->dev->dev, "Timeout waiting for VCC close on port %d\n",
+			 port);
 
 	/* Hold up vcc_destroy_socket() (our caller) until solos_bh() in the
 	   tasklet has finished processing any incoming packets (and, more to
 	   the point, using the vcc pointer). */
 	tasklet_unlock_wait(&card->tlet);
+
+	clear_bit(ATM_VF_ADDR, &vcc->flags);
+
 	return;
 }
 
@@ -1011,9 +989,12 @@ static uint32_t fpga_tx(struct solos_car
 			if (vcc) {
 				atomic_inc(&vcc->stats->tx);
 				solos_pop(vcc, oldskb);
-			} else
+			} else {
+				struct pkt_hdr *header = (void *)oldskb->data;
+				if (le16_to_cpu(header->type) == PKT_PCLOSE)
+					complete(&SKB_CB(oldskb)->c);
 				dev_kfree_skb_irq(oldskb);
-
+			}
 		}
 	}
 	/* For non-DMA TX, write the 'TX start' bit for all four ports simultaneously */
@@ -1248,7 +1229,7 @@ static int atm_init(struct solos_card *c
 		card->atmdev[i]->phy_data = (void *)(unsigned long)i;
 		atm_dev_signal_change(card->atmdev[i], ATM_PHY_SIG_FOUND);
 
-		skb = alloc_skb(sizeof(*header), GFP_ATOMIC);
+		skb = alloc_skb(sizeof(*header), GFP_KERNEL);
 		if (!skb) {
 			dev_warn(&card->dev->dev, "Failed to allocate sk_buff in atm_init()\n");
 			continue;
@@ -1345,6 +1326,8 @@ static struct pci_driver fpga_driver = {
 
 static int __init solos_pci_init(void)
 {
+	BUILD_BUG_ON(sizeof(struct solos_skb_cb) > sizeof(((struct sk_buff *)0)->cb));
+
 	printk(KERN_INFO "Solos PCI Driver Version %s\n", VERSION);
 	return pci_register_driver(&fpga_driver);
 }
--- a/include/linux/atmdev.h
+++ b/include/linux/atmdev.h
@@ -99,6 +99,7 @@ struct atm_vcc {
 	struct atm_dev	*dev;		/* device back pointer */
 	struct atm_qos	qos;		/* QOS */
 	struct atm_sap	sap;		/* SAP */
+	void (*release_cb)(struct atm_vcc *vcc); /* release_sock callback */
 	void (*push)(struct atm_vcc *vcc,struct sk_buff *skb);
 	void (*pop)(struct atm_vcc *vcc,struct sk_buff *skb); /* optional */
 	int (*push_oam)(struct atm_vcc *vcc,void *cell);
@@ -106,6 +107,7 @@ struct atm_vcc {
 	void		*dev_data;	/* per-device data */
 	void		*proto_data;	/* per-protocol data */
 	struct k_atm_aal_stats *stats;	/* pointer to AAL stats group */
+	struct module *owner;		/* owner of ->push function */
 	/* SVC part --- may move later ------------------------------------- */
 	short		itf;		/* interface number */
 	struct sockaddr_atmsvc local;
--- a/net/atm/br2684.c
+++ b/net/atm/br2684.c
@@ -68,12 +68,15 @@ struct br2684_vcc {
 	/* keep old push, pop functions for chaining */
 	void (*old_push)(struct atm_vcc *vcc, struct sk_buff *skb);
 	void (*old_pop)(struct atm_vcc *vcc, struct sk_buff *skb);
+	void (*old_release_cb)(struct atm_vcc *vcc);
+	struct module *old_owner;
 	enum br2684_encaps encaps;
 	struct list_head brvccs;
 #ifdef CONFIG_ATM_BR2684_IPFILTER
 	struct br2684_filter filter;
 #endif /* CONFIG_ATM_BR2684_IPFILTER */
 	unsigned int copies_needed, copies_failed;
+	atomic_t qspace;
 };
 
 struct br2684_dev {
@@ -181,18 +184,15 @@ static struct notifier_block atm_dev_not
 static void br2684_pop(struct atm_vcc *vcc, struct sk_buff *skb)
 {
 	struct br2684_vcc *brvcc = BR2684_VCC(vcc);
-	struct net_device *net_dev = skb->dev;
 
-	pr_debug("(vcc %p ; net_dev %p )\n", vcc, net_dev);
+	pr_debug("(vcc %p ; net_dev %p )\n", vcc, brvcc->device);
 	brvcc->old_pop(vcc, skb);
 
-	if (!net_dev)
-		return;
-
-	if (atm_may_send(vcc, 0))
-		netif_wake_queue(net_dev);
-
+	/* If the queue space just went up from zero, wake */
+	if (atomic_inc_return(&brvcc->qspace) == 1)
+		netif_wake_queue(brvcc->device);
 }
+
 /*
  * Send a packet out a particular vcc.  Not to useful right now, but paves
  * the way for multiple vcc's per itf.  Returns true if we can send,
@@ -256,16 +256,30 @@ static int br2684_xmit_vcc(struct sk_buf
 	ATM_SKB(skb)->atm_options = atmvcc->atm_options;
 	dev->stats.tx_packets++;
 	dev->stats.tx_bytes += skb->len;
-	atmvcc->send(atmvcc, skb);
 
-	if (!atm_may_send(atmvcc, 0)) {
+	if (atomic_dec_return(&brvcc->qspace) < 1) {
+		/* No more please! */
 		netif_stop_queue(brvcc->device);
-		/*check for race with br2684_pop*/
-		if (atm_may_send(atmvcc, 0))
-			netif_start_queue(brvcc->device);
+		/* We might have raced with br2684_pop() */
+		if (unlikely(atomic_read(&brvcc->qspace) > 0))
+			netif_wake_queue(brvcc->device);
 	}
 
-	return 1;
+	/* If this fails immediately, the skb will be freed and br2684_pop()
+	   will wake the queue if appropriate. Just return an error so that
+	   the stats are updated correctly */
+	return !atmvcc->send(atmvcc, skb);
+}
+
+static void br2684_release_cb(struct atm_vcc *atmvcc)
+{
+	struct br2684_vcc *brvcc = BR2684_VCC(atmvcc);
+
+	if (atomic_read(&brvcc->qspace) > 0)
+		netif_wake_queue(brvcc->device);
+
+	if (brvcc->old_release_cb)
+		brvcc->old_release_cb(atmvcc);
 }
 
 static inline struct br2684_vcc *pick_outgoing_vcc(const struct sk_buff *skb,
@@ -279,6 +293,8 @@ static netdev_tx_t br2684_start_xmit(str
 {
 	struct br2684_dev *brdev = BRPRIV(dev);
 	struct br2684_vcc *brvcc;
+	struct atm_vcc *atmvcc;
+	netdev_tx_t ret = NETDEV_TX_OK;
 
 	pr_debug("skb_dst(skb)=%p\n", skb_dst(skb));
 	read_lock(&devs_lock);
@@ -289,9 +305,26 @@ static netdev_tx_t br2684_start_xmit(str
 		dev->stats.tx_carrier_errors++;
 		/* netif_stop_queue(dev); */
 		dev_kfree_skb(skb);
-		read_unlock(&devs_lock);
-		return NETDEV_TX_OK;
+		goto out_devs;
 	}
+	atmvcc = brvcc->atmvcc;
+
+	bh_lock_sock(sk_atm(atmvcc));
+
+	if (test_bit(ATM_VF_RELEASED, &atmvcc->flags) ||
+	    test_bit(ATM_VF_CLOSE, &atmvcc->flags) ||
+	    !test_bit(ATM_VF_READY, &atmvcc->flags)) {
+		dev->stats.tx_dropped++;
+		dev_kfree_skb(skb);
+		goto out;
+	}
+
+	if (sock_owned_by_user(sk_atm(atmvcc))) {
+		netif_stop_queue(brvcc->device);
+		ret = NETDEV_TX_BUSY;
+		goto out;
+	}
+
 	if (!br2684_xmit_vcc(skb, dev, brvcc)) {
 		/*
 		 * We should probably use netif_*_queue() here, but that
@@ -303,8 +336,11 @@ static netdev_tx_t br2684_start_xmit(str
 		dev->stats.tx_errors++;
 		dev->stats.tx_fifo_errors++;
 	}
+ out:
+	bh_unlock_sock(sk_atm(atmvcc));
+ out_devs:
 	read_unlock(&devs_lock);
-	return NETDEV_TX_OK;
+	return ret;
 }
 
 /*
@@ -377,9 +413,10 @@ static void br2684_close_vcc(struct br26
 	list_del(&brvcc->brvccs);
 	write_unlock_irq(&devs_lock);
 	brvcc->atmvcc->user_back = NULL;	/* what about vcc->recvq ??? */
+	brvcc->atmvcc->release_cb = brvcc->old_release_cb;
 	brvcc->old_push(brvcc->atmvcc, NULL);	/* pass on the bad news */
+	module_put(brvcc->old_owner);
 	kfree(brvcc);
-	module_put(THIS_MODULE);
 }
 
 /* when AAL5 PDU comes in: */
@@ -504,6 +541,13 @@ static int br2684_regvcc(struct atm_vcc
 	brvcc = kzalloc(sizeof(struct br2684_vcc), GFP_KERNEL);
 	if (!brvcc)
 		return -ENOMEM;
+	/*
+	 * Allow two packets in the ATM queue. One actually being sent, and one
+	 * for the ATM 'TX done' handler to send. It shouldn't take long to get
+	 * the next one from the netdev queue, when we need it. More than that
+	 * would be bufferbloat.
+	 */
+	atomic_set(&brvcc->qspace, 2);
 	write_lock_irq(&devs_lock);
 	net_dev = br2684_find_dev(&be.ifspec);
 	if (net_dev == NULL) {
@@ -546,9 +590,13 @@ static int br2684_regvcc(struct atm_vcc
 	brvcc->encaps = (enum br2684_encaps)be.encaps;
 	brvcc->old_push = atmvcc->push;
 	brvcc->old_pop = atmvcc->pop;
+	brvcc->old_release_cb = atmvcc->release_cb;
+	brvcc->old_owner = atmvcc->owner;
 	barrier();
 	atmvcc->push = br2684_push;
 	atmvcc->pop = br2684_pop;
+	atmvcc->release_cb = br2684_release_cb;
+	atmvcc->owner = THIS_MODULE;
 
 	/* initialize netdev carrier state */
 	if (atmvcc->dev->signal == ATM_PHY_SIG_LOST)
@@ -687,10 +735,13 @@ static int br2684_ioctl(struct socket *s
 			return -ENOIOCTLCMD;
 		if (!capable(CAP_NET_ADMIN))
 			return -EPERM;
-		if (cmd == ATM_SETBACKEND)
+		if (cmd == ATM_SETBACKEND) {
+			if (sock->state != SS_CONNECTED)
+				return -EINVAL;
 			return br2684_regvcc(atmvcc, argp);
-		else
+		} else {
 			return br2684_create(argp);
+		}
 #ifdef CONFIG_ATM_BR2684_IPFILTER
 	case BR2684_SETFILT:
 		if (atmvcc->push != br2684_push)
--- a/net/atm/common.c
+++ b/net/atm/common.c
@@ -126,10 +126,19 @@ static void vcc_write_space(struct sock
 	rcu_read_unlock();
 }
 
+static void vcc_release_cb(struct sock *sk)
+{
+	struct atm_vcc *vcc = atm_sk(sk);
+
+	if (vcc->release_cb)
+		vcc->release_cb(vcc);
+}
+
 static struct proto vcc_proto = {
 	.name	  = "VCC",
 	.owner	  = THIS_MODULE,
 	.obj_size = sizeof(struct atm_vcc),
+	.release_cb = vcc_release_cb,
 };
 
 int vcc_create(struct net *net, struct socket *sock, int protocol, int family)
@@ -156,7 +165,9 @@ int vcc_create(struct net *net, struct s
 	atomic_set(&sk->sk_rmem_alloc, 0);
 	vcc->push = NULL;
 	vcc->pop = NULL;
+	vcc->owner = NULL;
 	vcc->push_oam = NULL;
+	vcc->release_cb = NULL;
 	vcc->vpi = vcc->vci = 0; /* no VCI/VPI yet */
 	vcc->atm_options = vcc->aal_options = 0;
 	sk->sk_destruct = vcc_sock_destruct;
@@ -175,6 +186,7 @@ static void vcc_destroy_socket(struct so
 			vcc->dev->ops->close(vcc);
 		if (vcc->push)
 			vcc->push(vcc, NULL); /* atmarpd has no push */
+		module_put(vcc->owner);
 
 		while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
 			atm_return(vcc, skb->truesize);
--- a/net/atm/pppoatm.c
+++ b/net/atm/pppoatm.c
@@ -60,6 +60,8 @@ struct pppoatm_vcc {
 	struct atm_vcc	*atmvcc;	/* VCC descriptor */
 	void (*old_push)(struct atm_vcc *, struct sk_buff *);
 	void (*old_pop)(struct atm_vcc *, struct sk_buff *);
+	void (*old_release_cb)(struct atm_vcc *);
+	struct module *old_owner;
 					/* keep old push/pop for detaching */
 	enum pppoatm_encaps encaps;
 	atomic_t inflight;
@@ -107,6 +109,24 @@ static void pppoatm_wakeup_sender(unsign
 	ppp_output_wakeup((struct ppp_channel *) arg);
 }
 
+static void pppoatm_release_cb(struct atm_vcc *atmvcc)
+{
+	struct pppoatm_vcc *pvcc = atmvcc_to_pvcc(atmvcc);
+
+	/*
+	 * As in pppoatm_pop(), it's safe to clear the BLOCKED bit here because
+	 * the wakeup *can't* race with pppoatm_send(). They both hold the PPP
+	 * channel's ->downl lock. And the potential race with *setting* it,
+	 * which leads to the double-check dance in pppoatm_may_send(), doesn't
+	 * exist here. In the sock_owned_by_user() case in pppoatm_send(), we
+	 * set the BLOCKED bit while the socket is still locked. We know that
+	 * ->release_cb() can't be called until that's done.
+	 */
+	if (test_and_clear_bit(BLOCKED, &pvcc->blocked))
+		tasklet_schedule(&pvcc->wakeup_tasklet);
+	if (pvcc->old_release_cb)
+		pvcc->old_release_cb(atmvcc);
+}
 /*
  * This gets called every time the ATM card has finished sending our
  * skb.  The ->old_pop will take care up normal atm flow control,
@@ -151,12 +171,11 @@ static void pppoatm_unassign_vcc(struct
 	pvcc = atmvcc_to_pvcc(atmvcc);
 	atmvcc->push = pvcc->old_push;
 	atmvcc->pop = pvcc->old_pop;
+	atmvcc->release_cb = pvcc->old_release_cb;
 	tasklet_kill(&pvcc->wakeup_tasklet);
 	ppp_unregister_channel(&pvcc->chan);
 	atmvcc->user_back = NULL;
 	kfree(pvcc);
-	/* Gee, I hope we have the big kernel lock here... */
-	module_put(THIS_MODULE);
 }
 
 /* Called when an AAL5 PDU comes in */
@@ -165,9 +184,13 @@ static void pppoatm_push(struct atm_vcc
 	struct pppoatm_vcc *pvcc = atmvcc_to_pvcc(atmvcc);
 	pr_debug("\n");
 	if (skb == NULL) {			/* VCC was closed */
+		struct module *module;
+
 		pr_debug("removing ATMPPP VCC %p\n", pvcc);
+		module = pvcc->old_owner;
 		pppoatm_unassign_vcc(atmvcc);
 		atmvcc->push(atmvcc, NULL);	/* Pass along bad news */
+		module_put(module);
 		return;
 	}
 	atm_return(atmvcc, skb->truesize);
@@ -211,7 +234,7 @@ error:
 	ppp_input_error(&pvcc->chan, 0);
 }
 
-static inline int pppoatm_may_send(struct pppoatm_vcc *pvcc, int size)
+static int pppoatm_may_send(struct pppoatm_vcc *pvcc, int size)
 {
 	/*
 	 * It's not clear that we need to bother with using atm_may_send()
@@ -269,10 +292,33 @@ static inline int pppoatm_may_send(struc
 static int pppoatm_send(struct ppp_channel *chan, struct sk_buff *skb)
 {
 	struct pppoatm_vcc *pvcc = chan_to_pvcc(chan);
+	struct atm_vcc *vcc;
+	int ret;
+
 	ATM_SKB(skb)->vcc = pvcc->atmvcc;
 	pr_debug("(skb=0x%p, vcc=0x%p)\n", skb, pvcc->atmvcc);
 	if (skb->data[0] == '\0' && (pvcc->flags & SC_COMP_PROT))
 		(void) skb_pull(skb, 1);
+
+	vcc = ATM_SKB(skb)->vcc;
+	bh_lock_sock(sk_atm(vcc));
+	if (sock_owned_by_user(sk_atm(vcc))) {
+		/*
+		 * Needs to happen (and be flushed, hence test_and_) before we unlock
+		 * the socket. It needs to be seen by the time our ->release_cb gets
+		 * called.
+		 */
+		test_and_set_bit(BLOCKED, &pvcc->blocked);
+		goto nospace;
+	}
+	if (test_bit(ATM_VF_RELEASED, &vcc->flags) ||
+	    test_bit(ATM_VF_CLOSE, &vcc->flags) ||
+	    !test_bit(ATM_VF_READY, &vcc->flags)) {
+		bh_unlock_sock(sk_atm(vcc));
+		kfree_skb(skb);
+		return DROP_PACKET;
+	}
+
 	switch (pvcc->encaps) {		/* LLC encapsulation needed */
 	case e_llc:
 		if (skb_headroom(skb) < LLC_LEN) {
@@ -285,8 +331,10 @@ static int pppoatm_send(struct ppp_chann
 			}
 			consume_skb(skb);
 			skb = n;
-			if (skb == NULL)
+			if (skb == NULL) {
+				bh_unlock_sock(sk_atm(vcc));
 				return DROP_PACKET;
+			}
 		} else if (!pppoatm_may_send(pvcc, skb->truesize))
 			goto nospace;
 		memcpy(skb_push(skb, LLC_LEN), pppllc, LLC_LEN);
@@ -296,6 +344,7 @@ static int pppoatm_send(struct ppp_chann
 			goto nospace;
 		break;
 	case e_autodetect:
+		bh_unlock_sock(sk_atm(vcc));
 		pr_debug("Trying to send without setting encaps!\n");
 		kfree_skb(skb);
 		return 1;
@@ -305,9 +354,12 @@ static int pppoatm_send(struct ppp_chann
 	ATM_SKB(skb)->atm_options = ATM_SKB(skb)->vcc->atm_options;
 	pr_debug("atm_skb(%p)->vcc(%p)->dev(%p)\n",
 		 skb, ATM_SKB(skb)->vcc, ATM_SKB(skb)->vcc->dev);
-	return ATM_SKB(skb)->vcc->send(ATM_SKB(skb)->vcc, skb)
+	ret = ATM_SKB(skb)->vcc->send(ATM_SKB(skb)->vcc, skb)
 	    ? DROP_PACKET : 1;
+	bh_unlock_sock(sk_atm(vcc));
+	return ret;
 nospace:
+	bh_unlock_sock(sk_atm(vcc));
 	/*
 	 * We don't have space to send this SKB now, but we might have
 	 * already applied SC_COMP_PROT compression, so may need to undo
@@ -362,6 +414,8 @@ static int pppoatm_assign_vcc(struct atm
 	atomic_set(&pvcc->inflight, NONE_INFLIGHT);
 	pvcc->old_push = atmvcc->push;
 	pvcc->old_pop = atmvcc->pop;
+	pvcc->old_owner = atmvcc->owner;
+	pvcc->old_release_cb = atmvcc->release_cb;
 	pvcc->encaps = (enum pppoatm_encaps) be.encaps;
 	pvcc->chan.private = pvcc;
 	pvcc->chan.ops = &pppoatm_ops;
@@ -377,7 +431,9 @@ static int pppoatm_assign_vcc(struct atm
 	atmvcc->user_back = pvcc;
 	atmvcc->push = pppoatm_push;
 	atmvcc->pop = pppoatm_pop;
+	atmvcc->release_cb = pppoatm_release_cb;
 	__module_get(THIS_MODULE);
+	atmvcc->owner = THIS_MODULE;
 
 	/* re-process everything received between connection setup and
 	   backend setup */
@@ -406,6 +462,8 @@ static int pppoatm_ioctl(struct socket *
 			return -ENOIOCTLCMD;
 		if (!capable(CAP_NET_ADMIN))
 			return -EPERM;
+		if (sock->state != SS_CONNECTED)
+			return -EINVAL;
 		return pppoatm_assign_vcc(atmvcc, argp);
 		}
 	case PPPIOCGCHAN: