* Re: [PATCH 2/3] b43: Fix oops if firmware is not available
From: Johannes Berg @ 2014-01-12 8:40 UTC (permalink / raw)
To: Ben Hutchings
Cc: Larry Finger, linville-2XuSBdqkA4R54TAoqtyWWQ,
linux-wireless-u79uwXL29TY76Z2rM5mHXA,
netdev-u79uwXL29TY76Z2rM5mHXA, Stable
In-Reply-To: <1389500647.3720.51.camel-nDn/Rdv9kqW9Jme8/bJn5UCKIB8iOfG2tUK59QYPAWc@public.gmane.org>
On Sun, 2014-01-12 at 04:24 +0000, Ben Hutchings wrote:
> You could switch back to synchronous firmware loading soon, as it's not
> going to support a usermode helper any more.
>
> But until then, the proper fix for this is going to be to cancel the
> waiter earlier in teardown.
I don't think we found a way when we looked at this, and instead made
the module unload wait for the request_firmware callback to come back
(see all users of the request_firmware_complete struct member in
iwlwifi/iwl-drv.c)
johannes
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* [PATCH 1/5] net: mvneta: increase the 64-bit rx/tx stats out of the hot path
From: Willy Tarreau @ 2014-01-12 9:31 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389519069-1619-1-git-send-email-w@1wt.eu>
Better count packets and bytes in the stack and on 32 bit then
accumulate them at the end for once. This saves two memory writes
and two memory barriers per packet. The incoming packet rate was
increased by 4.7% on the Openblocks AX3 thanks to this.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 15 +++++++++++----
1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index d5f0d72..baa85af 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -1391,6 +1391,8 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
{
struct net_device *dev = pp->dev;
int rx_done, rx_filled;
+ u32 rcvd_pkts = 0;
+ u32 rcvd_bytes = 0;
/* Get number of received packets */
rx_done = mvneta_rxq_busy_desc_num_get(pp, rxq);
@@ -1428,10 +1430,8 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
rx_bytes = rx_desc->data_size -
(ETH_FCS_LEN + MVNETA_MH_SIZE);
- u64_stats_update_begin(&pp->rx_stats.syncp);
- pp->rx_stats.packets++;
- pp->rx_stats.bytes += rx_bytes;
- u64_stats_update_end(&pp->rx_stats.syncp);
+ rcvd_pkts++;
+ rcvd_bytes += rx_bytes;
/* Linux processing */
skb_reserve(skb, MVNETA_MH_SIZE);
@@ -1452,6 +1452,13 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
}
}
+ if (rcvd_pkts) {
+ u64_stats_update_begin(&pp->rx_stats.syncp);
+ pp->rx_stats.packets += rcvd_pkts;
+ pp->rx_stats.bytes += rcvd_bytes;
+ u64_stats_update_end(&pp->rx_stats.syncp);
+ }
+
/* Update rxq management counters */
mvneta_rxq_desc_num_update(pp, rxq, rx_done, rx_filled);
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH 0/5] Assorted mvneta fixes
From: Willy Tarreau @ 2014-01-12 9:31 UTC (permalink / raw)
To: davem
Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT,
Arnaud Ebalard, Eric Dumazet
Hi,
this series provides some fixes for a number of issues met with the
mvneta driver :
- driver lockup when reading stats while sending traffic from multiple
CPUs : this obviously only happens on SMP and is the result of missing
locking on the driver. The problem was present since the introduction
of the driver in 3.8. The first patch performs some changes that are
needed for the second one which actually fixes the issue by using
per-cpu counters. It could make sense to backport this to the relevant
stable versions.
- mvneta_tx_timeout calls various functions to reset the NIC, and these
functions sleep, which is not allowed here, resulting in a panic.
Better completely disable this Tx timeout handler for now since it is
never called. The problem was encountered while developing some new
features, it's uncertain whether it's possible to reproduce it with
regular usage, so maybe a backport to stable is not needed.
- replace the Tx timer with a real Tx IRQ. As first reported by Arnaud
Ebalard and explained by Eric Dumazet, there is no way this driver
can work correctly if it uses a driver to recycle the Tx descriptors.
If too many packets are sent at once, the driver quickly ends up with
no descriptors (which happens twice as easily in GSO) and has to wait
10ms for recycling its descriptors and being able to send again. Eric
has worked around this in the core GSO code. But still when routing
traffic or sending UDP packets, the limitation is very visible. Using
Tx IRQs allows Tx descriptors to be recycled when sent. The coalesce
value is still configurable using ethtool. This fix turns the UDP
send bitrate from 134 Mbps to 987 Mbps (ie: line rate). It's made of
two patches, one to add the relevant bits from the original Marvell's
driver, and another one to implement the change. I don't know if it
should be backported to stable, as the bug only causes poor performance.
Thanks,
Willy
---
Willy Tarreau (5):
net: mvneta: increase the 64-bit rx/tx stats out of the hot path
net: mvneta: use per_cpu stats to fix an SMP lock up
net: mvneta: do not schedule in mvneta_tx_timeout
net: mvneta: add missing bit descriptions for interrupt masks and causes
net: mvneta: replace Tx timer with a real interrupt
drivers/net/ethernet/marvell/mvneta.c | 217 ++++++++++++++++++----------------
1 file changed, 116 insertions(+), 101 deletions(-)
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply
* [PATCH 3/5] net: mvneta: do not schedule in mvneta_tx_timeout
From: Willy Tarreau @ 2014-01-12 9:31 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389519069-1619-1-git-send-email-w@1wt.eu>
If a queue timeout is reported, we can oops because of some
schedules while the caller is atomic, as shown below :
mvneta d0070000.ethernet eth0: tx timeout
BUG: scheduling while atomic: bash/1528/0x00000100
Modules linked in: slhttp_ethdiv(C) [last unloaded: slhttp_ethdiv]
CPU: 2 PID: 1528 Comm: bash Tainted: G WC 3.13.0-rc4-mvebu-nf #180
[<c0011bd9>] (unwind_backtrace+0x1/0x98) from [<c000f1ab>] (show_stack+0xb/0xc)
[<c000f1ab>] (show_stack+0xb/0xc) from [<c02ad323>] (dump_stack+0x4f/0x64)
[<c02ad323>] (dump_stack+0x4f/0x64) from [<c02abe67>] (__schedule_bug+0x37/0x4c)
[<c02abe67>] (__schedule_bug+0x37/0x4c) from [<c02ae261>] (__schedule+0x325/0x3ec)
[<c02ae261>] (__schedule+0x325/0x3ec) from [<c02adb97>] (schedule_timeout+0xb7/0x118)
[<c02adb97>] (schedule_timeout+0xb7/0x118) from [<c0020a67>] (msleep+0xf/0x14)
[<c0020a67>] (msleep+0xf/0x14) from [<c01dcbe5>] (mvneta_stop_dev+0x21/0x194)
[<c01dcbe5>] (mvneta_stop_dev+0x21/0x194) from [<c01dcfe9>] (mvneta_tx_timeout+0x19/0x24)
[<c01dcfe9>] (mvneta_tx_timeout+0x19/0x24) from [<c024afc7>] (dev_watchdog+0x18b/0x1c4)
[<c024afc7>] (dev_watchdog+0x18b/0x1c4) from [<c0020b53>] (call_timer_fn.isra.27+0x17/0x5c)
[<c0020b53>] (call_timer_fn.isra.27+0x17/0x5c) from [<c0020cad>] (run_timer_softirq+0x115/0x170)
[<c0020cad>] (run_timer_softirq+0x115/0x170) from [<c001ccb9>] (__do_softirq+0xbd/0x1a8)
[<c001ccb9>] (__do_softirq+0xbd/0x1a8) from [<c001cfad>] (irq_exit+0x61/0x98)
[<c001cfad>] (irq_exit+0x61/0x98) from [<c000d4bf>] (handle_IRQ+0x27/0x60)
[<c000d4bf>] (handle_IRQ+0x27/0x60) from [<c000843b>] (armada_370_xp_handle_irq+0x33/0xc8)
[<c000843b>] (armada_370_xp_handle_irq+0x33/0xc8) from [<c000fba9>] (__irq_usr+0x49/0x60)
So for now, let's simply ignore these timeouts generally caused by bugs
only.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 11 -----------
1 file changed, 11 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 40d3e8b..84220c1 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -2244,16 +2244,6 @@ static void mvneta_stop_dev(struct mvneta_port *pp)
mvneta_rx_reset(pp);
}
-/* tx timeout callback - display a message and stop/start the network device */
-static void mvneta_tx_timeout(struct net_device *dev)
-{
- struct mvneta_port *pp = netdev_priv(dev);
-
- netdev_info(dev, "tx timeout\n");
- mvneta_stop_dev(pp);
- mvneta_start_dev(pp);
-}
-
/* Return positive if MTU is valid */
static int mvneta_check_mtu_valid(struct net_device *dev, int mtu)
{
@@ -2634,7 +2624,6 @@ static const struct net_device_ops mvneta_netdev_ops = {
.ndo_set_rx_mode = mvneta_set_rx_mode,
.ndo_set_mac_address = mvneta_set_mac_addr,
.ndo_change_mtu = mvneta_change_mtu,
- .ndo_tx_timeout = mvneta_tx_timeout,
.ndo_get_stats64 = mvneta_get_stats64,
.ndo_do_ioctl = mvneta_ioctl,
};
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH 5/5] net: mvneta: replace Tx timer with a real interrupt
From: Willy Tarreau @ 2014-01-12 9:31 UTC (permalink / raw)
To: davem
Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT,
Arnaud Ebalard, Eric Dumazet
In-Reply-To: <1389519069-1619-1-git-send-email-w@1wt.eu>
Right now the mvneta driver doesn't handle Tx IRQ, and relies on two
mechanisms to flush Tx descriptors : a flush at the end of mvneta_tx()
and a timer. If a burst of packets is emitted faster than the device
can send them, then the queue is stopped until next wake-up of the
timer 10ms later. This causes jerky output traffic with bursts and
pauses, making it difficult to reach line rate with very few streams.
A test on UDP traffic shows that it's not possible to go beyond 134
Mbps / 12 kpps of outgoing traffic with 1500-bytes IP packets. Routed
traffic tends to observe pauses as well if the traffic is bursty,
making it even burstier after the wake-up.
It seems that this feature was inherited from the original driver but
nothing there mentions any reason for not using the interrupt instead,
which the chip supports.
Thus, this patch enables Tx interrupts and removes the timer. It does
the two at once because it's not really possible to make the two
mechanisms coexist, so a split patch doesn't make sense.
First tests performed on a Mirabox (Armada 370) show that less CPU
seems to be used when sending traffic. One reason might be that we now
call the mvneta_tx_done_gbe() with a mask indicating which queues have
been done instead of looping over all of them.
The same UDP test above now happily reaches 987 Mbps / 87.7 kpps.
Single-stream TCP traffic can now more easily reach line rate. HTTP
transfers of 1 MB objects over a single connection went from 730 to
840 Mbps. It is even possible to go significantly higher (>900 Mbps)
by tweaking tcp_tso_win_divisor.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Cc: Arnaud Ebalard <arno@natisbad.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 71 ++++++-----------------------------
1 file changed, 12 insertions(+), 59 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index defda6f..df75a23 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -216,9 +216,6 @@
#define MVNETA_RX_COAL_PKTS 32
#define MVNETA_RX_COAL_USEC 100
-/* Timer */
-#define MVNETA_TX_DONE_TIMER_PERIOD 10
-
/* Napi polling weight */
#define MVNETA_RX_POLL_WEIGHT 64
@@ -274,16 +271,11 @@ struct mvneta_port {
void __iomem *base;
struct mvneta_rx_queue *rxqs;
struct mvneta_tx_queue *txqs;
- struct timer_list tx_done_timer;
struct net_device *dev;
u32 cause_rx_tx;
struct napi_struct napi;
- /* Flags */
- unsigned long flags;
-#define MVNETA_F_TX_DONE_TIMER_BIT 0
-
/* Napi weight */
int weight;
@@ -1149,17 +1141,6 @@ static void mvneta_tx_done_pkts_coal_set(struct mvneta_port *pp,
txq->done_pkts_coal = value;
}
-/* Trigger tx done timer in MVNETA_TX_DONE_TIMER_PERIOD msecs */
-static void mvneta_add_tx_done_timer(struct mvneta_port *pp)
-{
- if (test_and_set_bit(MVNETA_F_TX_DONE_TIMER_BIT, &pp->flags) == 0) {
- pp->tx_done_timer.expires = jiffies +
- msecs_to_jiffies(MVNETA_TX_DONE_TIMER_PERIOD);
- add_timer(&pp->tx_done_timer);
- }
-}
-
-
/* Handle rx descriptor fill by setting buf_cookie and buf_phys_addr */
static void mvneta_rx_desc_fill(struct mvneta_rx_desc *rx_desc,
u32 phys_addr, u32 cookie)
@@ -1651,15 +1632,6 @@ out:
dev_kfree_skb_any(skb);
}
- if (txq->count >= MVNETA_TXDONE_COAL_PKTS)
- mvneta_txq_done(pp, txq);
-
- /* If after calling mvneta_txq_done, count equals
- * frags, we need to set the timer
- */
- if (txq->count == frags && frags > 0)
- mvneta_add_tx_done_timer(pp);
-
return NETDEV_TX_OK;
}
@@ -1935,14 +1907,22 @@ static int mvneta_poll(struct napi_struct *napi, int budget)
/* Read cause register */
cause_rx_tx = mvreg_read(pp, MVNETA_INTR_NEW_CAUSE) &
- MVNETA_RX_INTR_MASK(rxq_number);
+ (MVNETA_RX_INTR_MASK(rxq_number) | MVNETA_TX_INTR_MASK(txq_number));
+
+ /* Release Tx descriptors */
+ if (cause_rx_tx & MVNETA_TX_INTR_MASK_ALL) {
+ int tx_todo = 0;
+
+ mvneta_tx_done_gbe(pp, (cause_rx_tx & MVNETA_TX_INTR_MASK_ALL), &tx_todo);
+ cause_rx_tx &= ~MVNETA_TX_INTR_MASK_ALL;
+ }
/* For the case where the last mvneta_poll did not process all
* RX packets
*/
cause_rx_tx |= pp->cause_rx_tx;
if (rxq_number > 1) {
- while ((cause_rx_tx != 0) && (budget > 0)) {
+ while ((cause_rx_tx & MVNETA_RX_INTR_MASK_ALL) && (budget > 0)) {
int count;
struct mvneta_rx_queue *rxq;
/* get rx queue number from cause_rx_tx */
@@ -1974,7 +1954,7 @@ static int mvneta_poll(struct napi_struct *napi, int budget)
napi_complete(napi);
local_irq_save(flags);
mvreg_write(pp, MVNETA_INTR_NEW_MASK,
- MVNETA_RX_INTR_MASK(rxq_number));
+ MVNETA_RX_INTR_MASK(rxq_number) | MVNETA_TX_INTR_MASK(txq_number));
local_irq_restore(flags);
}
@@ -1982,26 +1962,6 @@ static int mvneta_poll(struct napi_struct *napi, int budget)
return rx_done;
}
-/* tx done timer callback */
-static void mvneta_tx_done_timer_callback(unsigned long data)
-{
- struct net_device *dev = (struct net_device *)data;
- struct mvneta_port *pp = netdev_priv(dev);
- int tx_done = 0, tx_todo = 0;
-
- if (!netif_running(dev))
- return ;
-
- clear_bit(MVNETA_F_TX_DONE_TIMER_BIT, &pp->flags);
-
- tx_done = mvneta_tx_done_gbe(pp,
- (((1 << txq_number) - 1) &
- MVNETA_CAUSE_TXQ_SENT_DESC_ALL_MASK),
- &tx_todo);
- if (tx_todo > 0)
- mvneta_add_tx_done_timer(pp);
-}
-
/* Handle rxq fill: allocates rxq skbs; called when initializing a port */
static int mvneta_rxq_fill(struct mvneta_port *pp, struct mvneta_rx_queue *rxq,
int num)
@@ -2251,7 +2211,7 @@ static void mvneta_start_dev(struct mvneta_port *pp)
/* Unmask interrupts */
mvreg_write(pp, MVNETA_INTR_NEW_MASK,
- MVNETA_RX_INTR_MASK(rxq_number));
+ MVNETA_RX_INTR_MASK(rxq_number) | MVNETA_TX_INTR_MASK(txq_number));
phy_start(pp->phy_dev);
netif_tx_start_all_queues(pp->dev);
@@ -2527,8 +2487,6 @@ static int mvneta_stop(struct net_device *dev)
free_irq(dev->irq, pp);
mvneta_cleanup_rxqs(pp);
mvneta_cleanup_txqs(pp);
- del_timer(&pp->tx_done_timer);
- clear_bit(MVNETA_F_TX_DONE_TIMER_BIT, &pp->flags);
return 0;
}
@@ -2887,11 +2845,6 @@ static int mvneta_probe(struct platform_device *pdev)
}
}
- pp->tx_done_timer.data = (unsigned long)dev;
- pp->tx_done_timer.function = mvneta_tx_done_timer_callback;
- init_timer(&pp->tx_done_timer);
- clear_bit(MVNETA_F_TX_DONE_TIMER_BIT, &pp->flags);
-
pp->tx_ring_size = MVNETA_MAX_TXD;
pp->rx_ring_size = MVNETA_MAX_RXD;
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH 2/5] net: mvneta: use per_cpu stats to fix an SMP lock up
From: Willy Tarreau @ 2014-01-12 9:31 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389519069-1619-1-git-send-email-w@1wt.eu>
Stats writers are mvneta_rx() and mvneta_tx(). They don't lock anything
when they update the stats, and as a result, it randomly happens that
the stats freeze on SMP if two updates happen during stats retrieval.
This is very easily reproducible by starting two HTTP servers and binding
each of them to a different CPU, then consulting /proc/net/dev in loops
during transfers, the interface should immediately lock up. This issue
also randomly happens upon link state changes during transfers, because
the stats are collected in this situation, but it takes more attempts to
reproduce it.
The comments in netdevice.h suggest using per_cpu stats instead to get
rid of this issue.
This patch implements this. It merges both rx_stats and tx_stats into
a single "stats" member with a single syncp. Both mvneta_rx() and
mvneta_rx() now only update the a single CPU's counters.
In turn, mvneta_get_stats64() does the summing by iterating over all CPUs
to get their respective stats.
With this change, stats are still correct and no more lockup is encountered.
Note that this bug was present since the first import of the mvneta
driver. It might make sense to backport it to some stable trees. If
so, it depends on "d33dc73 net: mvneta: increase the 64-bit rx/tx stats
out of the hot path".
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 84 +++++++++++++++++++++++------------
1 file changed, 55 insertions(+), 29 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index baa85af..40d3e8b 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -221,10 +221,12 @@
#define MVNETA_RX_BUF_SIZE(pkt_size) ((pkt_size) + NET_SKB_PAD)
-struct mvneta_stats {
+struct mvneta_pcpu_stats {
struct u64_stats_sync syncp;
- u64 packets;
- u64 bytes;
+ u64 rx_packets;
+ u64 rx_bytes;
+ u64 tx_packets;
+ u64 tx_bytes;
};
struct mvneta_port {
@@ -250,8 +252,7 @@ struct mvneta_port {
u8 mcast_count[256];
u16 tx_ring_size;
u16 rx_ring_size;
- struct mvneta_stats tx_stats;
- struct mvneta_stats rx_stats;
+ struct mvneta_pcpu_stats *stats;
struct mii_bus *mii_bus;
struct phy_device *phy_dev;
@@ -461,21 +462,29 @@ struct rtnl_link_stats64 *mvneta_get_stats64(struct net_device *dev,
{
struct mvneta_port *pp = netdev_priv(dev);
unsigned int start;
+ int cpu;
- memset(stats, 0, sizeof(struct rtnl_link_stats64));
-
- do {
- start = u64_stats_fetch_begin_bh(&pp->rx_stats.syncp);
- stats->rx_packets = pp->rx_stats.packets;
- stats->rx_bytes = pp->rx_stats.bytes;
- } while (u64_stats_fetch_retry_bh(&pp->rx_stats.syncp, start));
+ for_each_possible_cpu(cpu) {
+ struct mvneta_pcpu_stats *cpu_stats;
+ u64 rx_packets;
+ u64 rx_bytes;
+ u64 tx_packets;
+ u64 tx_bytes;
+ cpu_stats = per_cpu_ptr(pp->stats, cpu);
+ do {
+ start = u64_stats_fetch_begin_bh(&cpu_stats->syncp);
+ rx_packets = cpu_stats->rx_packets;
+ rx_bytes = cpu_stats->rx_bytes;
+ tx_packets = cpu_stats->tx_packets;
+ tx_bytes = cpu_stats->tx_bytes;
+ } while (u64_stats_fetch_retry_bh(&cpu_stats->syncp, start));
- do {
- start = u64_stats_fetch_begin_bh(&pp->tx_stats.syncp);
- stats->tx_packets = pp->tx_stats.packets;
- stats->tx_bytes = pp->tx_stats.bytes;
- } while (u64_stats_fetch_retry_bh(&pp->tx_stats.syncp, start));
+ stats->rx_packets += rx_packets;
+ stats->rx_bytes += rx_bytes;
+ stats->tx_packets += tx_packets;
+ stats->tx_bytes += tx_bytes;
+ }
stats->rx_errors = dev->stats.rx_errors;
stats->rx_dropped = dev->stats.rx_dropped;
@@ -1453,10 +1462,12 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
}
if (rcvd_pkts) {
- u64_stats_update_begin(&pp->rx_stats.syncp);
- pp->rx_stats.packets += rcvd_pkts;
- pp->rx_stats.bytes += rcvd_bytes;
- u64_stats_update_end(&pp->rx_stats.syncp);
+ struct mvneta_pcpu_stats *stats = this_cpu_ptr(pp->stats);
+
+ u64_stats_update_begin(&stats->syncp);
+ stats->rx_packets += rcvd_pkts;
+ stats->rx_bytes += rcvd_bytes;
+ u64_stats_update_end(&stats->syncp);
}
/* Update rxq management counters */
@@ -1589,11 +1600,12 @@ static int mvneta_tx(struct sk_buff *skb, struct net_device *dev)
out:
if (frags > 0) {
- u64_stats_update_begin(&pp->tx_stats.syncp);
- pp->tx_stats.packets++;
- pp->tx_stats.bytes += skb->len;
- u64_stats_update_end(&pp->tx_stats.syncp);
+ struct mvneta_pcpu_stats *stats = this_cpu_ptr(pp->stats);
+ u64_stats_update_begin(&stats->syncp);
+ stats->tx_packets++;
+ stats->tx_bytes += skb->len;
+ u64_stats_update_end(&stats->syncp);
} else {
dev->stats.tx_dropped++;
dev_kfree_skb_any(skb);
@@ -2758,6 +2770,7 @@ static int mvneta_probe(struct platform_device *pdev)
const char *mac_from;
int phy_mode;
int err;
+ int cpu;
/* Our multiqueue support is not complete, so for now, only
* allow the usage of the first RX queue
@@ -2799,9 +2812,6 @@ static int mvneta_probe(struct platform_device *pdev)
pp = netdev_priv(dev);
- u64_stats_init(&pp->tx_stats.syncp);
- u64_stats_init(&pp->rx_stats.syncp);
-
pp->weight = MVNETA_RX_POLL_WEIGHT;
pp->phy_node = phy_node;
pp->phy_interface = phy_mode;
@@ -2820,6 +2830,19 @@ static int mvneta_probe(struct platform_device *pdev)
goto err_clk;
}
+ /* Alloc per-cpu stats */
+ pp->stats = alloc_percpu(struct mvneta_pcpu_stats);
+ if (!pp->stats) {
+ err = -ENOMEM;
+ goto err_unmap;
+ }
+
+ for_each_possible_cpu(cpu) {
+ struct mvneta_pcpu_stats *stats;
+ stats = per_cpu_ptr(pp->stats, cpu);
+ u64_stats_init(&stats->syncp);
+ }
+
dt_mac_addr = of_get_mac_address(dn);
if (dt_mac_addr) {
mac_from = "device tree";
@@ -2849,7 +2872,7 @@ static int mvneta_probe(struct platform_device *pdev)
err = mvneta_init(pp, phy_addr);
if (err < 0) {
dev_err(&pdev->dev, "can't init eth hal\n");
- goto err_unmap;
+ goto err_free_stats;
}
mvneta_port_power_up(pp, phy_mode);
@@ -2879,6 +2902,8 @@ static int mvneta_probe(struct platform_device *pdev)
err_deinit:
mvneta_deinit(pp);
+err_free_stats:
+ free_percpu(pp->stats);
err_unmap:
iounmap(pp->base);
err_clk:
@@ -2899,6 +2924,7 @@ static int mvneta_remove(struct platform_device *pdev)
unregister_netdev(dev);
mvneta_deinit(pp);
clk_disable_unprepare(pp->clk);
+ free_percpu(pp->stats);
iounmap(pp->base);
irq_dispose_mapping(dev->irq);
free_netdev(dev);
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH 4/5] net: mvneta: add missing bit descriptions for interrupt masks and causes
From: Willy Tarreau @ 2014-01-12 9:31 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389519069-1619-1-git-send-email-w@1wt.eu>
Marvell has not published the chip's datasheet yet, so it's very hard
to find the relevant bits to manipulate to change the IRQ behaviour.
Fortunately, these bits are described in the proprietary LSP patch set
which is publicly available here :
http://www.plugcomputer.org/downloads/mirabox/
So let's put them back in the driver in order to reduce the burden of
current and future maintenance.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 44 +++++++++++++++++++++++++++++++++--
1 file changed, 42 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 84220c1..defda6f 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -101,16 +101,56 @@
#define MVNETA_CPU_RXQ_ACCESS_ALL_MASK 0x000000ff
#define MVNETA_CPU_TXQ_ACCESS_ALL_MASK 0x0000ff00
#define MVNETA_RXQ_TIME_COAL_REG(q) (0x2580 + ((q) << 2))
+
+/* Exception Interrupt Port/Queue Cause register */
+
#define MVNETA_INTR_NEW_CAUSE 0x25a0
-#define MVNETA_RX_INTR_MASK(nr_rxqs) (((1 << nr_rxqs) - 1) << 8)
#define MVNETA_INTR_NEW_MASK 0x25a4
+
+/* bits 0..7 = TXQ SENT, one bit per queue.
+ * bits 8..15 = RXQ OCCUP, one bit per queue.
+ * bits 16..23 = RXQ FREE, one bit per queue.
+ * bit 29 = OLD_REG_SUM, see old reg ?
+ * bit 30 = TX_ERR_SUM, one bit for 4 ports
+ * bit 31 = MISC_SUM, one bit for 4 ports
+ */
+#define MVNETA_TX_INTR_MASK(nr_txqs) (((1 << nr_txqs) - 1) << 0)
+#define MVNETA_TX_INTR_MASK_ALL (0xff << 0)
+#define MVNETA_RX_INTR_MASK(nr_rxqs) (((1 << nr_rxqs) - 1) << 8)
+#define MVNETA_RX_INTR_MASK_ALL (0xff << 8)
+
#define MVNETA_INTR_OLD_CAUSE 0x25a8
#define MVNETA_INTR_OLD_MASK 0x25ac
+
+/* Data Path Port/Queue Cause Register */
#define MVNETA_INTR_MISC_CAUSE 0x25b0
#define MVNETA_INTR_MISC_MASK 0x25b4
+
+#define MVNETA_CAUSE_PHY_STATUS_CHANGE BIT(0)
+#define MVNETA_CAUSE_LINK_CHANGE BIT(1)
+#define MVNETA_CAUSE_PTP BIT(4)
+
+#define MVNETA_CAUSE_INTERNAL_ADDR_ERR BIT(7)
+#define MVNETA_CAUSE_RX_OVERRUN BIT(8)
+#define MVNETA_CAUSE_RX_CRC_ERROR BIT(9)
+#define MVNETA_CAUSE_RX_LARGE_PKT BIT(10)
+#define MVNETA_CAUSE_TX_UNDERUN BIT(11)
+#define MVNETA_CAUSE_PRBS_ERR BIT(12)
+#define MVNETA_CAUSE_PSC_SYNC_CHANGE BIT(13)
+#define MVNETA_CAUSE_SERDES_SYNC_ERR BIT(14)
+
+#define MVNETA_CAUSE_BMU_ALLOC_ERR_SHIFT 16
+#define MVNETA_CAUSE_BMU_ALLOC_ERR_ALL_MASK (0xF << MVNETA_CAUSE_BMU_ALLOC_ERR_SHIFT)
+#define MVNETA_CAUSE_BMU_ALLOC_ERR_MASK(pool) (1 << (MVNETA_CAUSE_BMU_ALLOC_ERR_SHIFT + (pool)))
+
+#define MVNETA_CAUSE_TXQ_ERROR_SHIFT 24
+#define MVNETA_CAUSE_TXQ_ERROR_ALL_MASK (0xFF << MVNETA_CAUSE_TXQ_ERROR_SHIFT)
+#define MVNETA_CAUSE_TXQ_ERROR_MASK(q) (1 << (MVNETA_CAUSE_TXQ_ERROR_SHIFT + (q)))
+
#define MVNETA_INTR_ENABLE 0x25b8
#define MVNETA_TXQ_INTR_ENABLE_ALL_MASK 0x0000ff00
-#define MVNETA_RXQ_INTR_ENABLE_ALL_MASK 0xff000000
+#define MVNETA_RXQ_INTR_ENABLE_ALL_MASK 0xff000000 // note: neta says it's 0x000000FF
+
#define MVNETA_RXQ_CMD 0x2680
#define MVNETA_RXQ_DISABLE_SHIFT 8
#define MVNETA_RXQ_ENABLE_MASK 0x000000ff
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* Re: [PATCH 1/1] When timestamping is enabled, stmmac_tx_clean will call stmmac_get_tx_hwtstamp to get tx TS. It's possible that skb is NULL because there are other network frames that use several descriptors. So we must return immediately in stmmac_get_tx_hwtstamp if skb is NULL to avoid system crash.
From: Bruce Liu @ 2014-01-12 9:39 UTC (permalink / raw)
To: peppe.cavallaro; +Cc: netdev, linux-kernel, damuzi000
In-Reply-To: <1389492095-6589-1-git-send-email-damuzi000@gmail.com>
When timestamping is enabled, stmmac_tx_clean will call stmmac_get_tx_hwtstamp to get tx TS.
It's possible that skb is NULL because there are other network frames that use several descriptors.
So we must return immediately in stmmac_get_tx_hwtstamp if skb is NULL to avoid system crash.
Signed-off-by: Bruce Liu <damuzi000@gmail.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 797b56a..47f2287 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -332,7 +332,7 @@ static void stmmac_get_tx_hwtstamp(struct stmmac_priv *priv,
return;
/* exit if skb doesn't support hw tstamp */
- if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)))
+ if (likely(!skb || !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)))
return;
if (priv->adv_ts)
--
1.7.9.5
^ permalink raw reply related
* [PATCH] ipv4: delete the statements limit icmp message in ip_error()
From: Duan Jiong @ 2014-01-12 9:52 UTC (permalink / raw)
To: David Miller; +Cc: netdev
Because after the icmp_send() is called, the icmpv4_xrlim_allow()
will be called to limit the icmp message too.
Signed-off-by: Duan Jiong <duanj.fnst@cn.fujitsu.com>
---
net/ipv4/route.c | 21 +--------------------
1 file changed, 1 insertion(+), 20 deletions(-)
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index f8da282..e521de1 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -900,10 +900,7 @@ static int ip_error(struct sk_buff *skb)
{
struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
struct rtable *rt = skb_rtable(skb);
- struct inet_peer *peer;
- unsigned long now;
struct net *net;
- bool send;
int code;
net = dev_net(rt->dst.dev);
@@ -936,23 +933,7 @@ static int ip_error(struct sk_buff *skb)
break;
}
- peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1);
-
- send = true;
- if (peer) {
- now = jiffies;
- peer->rate_tokens += now - peer->rate_last;
- if (peer->rate_tokens > ip_rt_error_burst)
- peer->rate_tokens = ip_rt_error_burst;
- peer->rate_last = now;
- if (peer->rate_tokens >= ip_rt_error_cost)
- peer->rate_tokens -= ip_rt_error_cost;
- else
- send = false;
- inet_putpeer(peer);
- }
- if (send)
- icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
+ icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
out: kfree_skb(skb);
return 0;
--
1.8.3.1
^ permalink raw reply related
* [PATCH net-next] ipv4: remove the useless argument from ip_tunnel_hash()
From: Duan Jiong @ 2014-01-12 10:04 UTC (permalink / raw)
To: David Miller; +Cc: netdev
Since commit c544193214 introduced function ip_tunnel_hash(), the argument
itn is no longer in use, so remove it.
Signed-off-by: Duan Jiong <duanj.fnst@cn.fujitsu.com>
---
net/ipv4/ip_tunnel.c | 9 ++++-----
1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/net/ipv4/ip_tunnel.c b/net/ipv4/ip_tunnel.c
index d3929a6..6f5a347 100644
--- a/net/ipv4/ip_tunnel.c
+++ b/net/ipv4/ip_tunnel.c
@@ -61,8 +61,7 @@
#include <net/ip6_route.h>
#endif
-static unsigned int ip_tunnel_hash(struct ip_tunnel_net *itn,
- __be32 key, __be32 remote)
+static unsigned int ip_tunnel_hash(__be32 key, __be32 remote)
{
return hash_32((__force u32)key ^ (__force u32)remote,
IP_TNL_HASH_BITS);
@@ -204,7 +203,7 @@ struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
struct ip_tunnel *t, *cand = NULL;
struct hlist_head *head;
- hash = ip_tunnel_hash(itn, key, remote);
+ hash = ip_tunnel_hash(key, remote);
head = &itn->tunnels[hash];
hlist_for_each_entry_rcu(t, head, hash_node) {
@@ -236,7 +235,7 @@ struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
cand = t;
}
- hash = ip_tunnel_hash(itn, key, 0);
+ hash = ip_tunnel_hash(key, 0);
head = &itn->tunnels[hash];
hlist_for_each_entry_rcu(t, head, hash_node) {
@@ -292,7 +291,7 @@ static struct hlist_head *ip_bucket(struct ip_tunnel_net *itn,
else
remote = 0;
- h = ip_tunnel_hash(itn, parms->i_key, remote);
+ h = ip_tunnel_hash(parms->i_key, remote);
return &itn->tunnels[h];
}
--
1.8.3.1
^ permalink raw reply related
* Re: [ipv6_vti.c] INFO: trying to register non-static key.
From: Fengguang Wu @ 2014-01-12 10:18 UTC (permalink / raw)
To: David Miller; +Cc: roy.qing.li, netdev, linux-kernel
In-Reply-To: <20140111.194502.2043728861122940485.davem@davemloft.net>
On Sat, Jan 11, 2014 at 07:45:02PM -0800, David Miller wrote:
> From: Fengguang Wu <fengguang.wu@intel.com>
> Date: Sun, 12 Jan 2014 07:10:23 +0800
>
> > Here is a warning introduced by commit 469bdcefd ("ipv6: fix the use
> > of pcpu_tstats in ip6_vti.c"):
>
> It isn't fixed by the following, which is also in Linus's tree?
Yes.. Sorry it's a duplicate report. I searched the full commit sha1
in my outbox before this report but the previous report email only
mentioned the abbreviated commit id..
Cheers,
Fengguang
> >From 657e5d19657542631461e72fdc375b1e83e72070 Mon Sep 17 00:00:00 2001
> From: Li RongQing <roy.qing.li@gmail.com>
> Date: Tue, 7 Jan 2014 15:39:43 +0800
> Subject: [PATCH] ipv6: pcpu_tstats.syncp should be initialised in ip6_vti.c
>
> initialise pcpu_tstats.syncp to kill the calltrace
> [ 11.973950] Call Trace:
> [ 11.973950] [<819bbaff>] dump_stack+0x48/0x60
> [ 11.973950] [<819bbaff>] dump_stack+0x48/0x60
> [ 11.973950] [<81078dcf>] __lock_acquire.isra.22+0x1bf/0xc10
> [ 11.973950] [<81078dcf>] __lock_acquire.isra.22+0x1bf/0xc10
> [ 11.973950] [<81079fa7>] lock_acquire+0x77/0xa0
> [ 11.973950] [<81079fa7>] lock_acquire+0x77/0xa0
> [ 11.973950] [<817ca7ab>] ? dev_get_stats+0xcb/0x130
> [ 11.973950] [<817ca7ab>] ? dev_get_stats+0xcb/0x130
> [ 11.973950] [<8183862d>] ip_tunnel_get_stats64+0x6d/0x230
> [ 11.973950] [<8183862d>] ip_tunnel_get_stats64+0x6d/0x230
> [ 11.973950] [<817ca7ab>] ? dev_get_stats+0xcb/0x130
> [ 11.973950] [<817ca7ab>] ? dev_get_stats+0xcb/0x130
> [ 11.973950] [<811cf8c1>] ? __nla_reserve+0x21/0xd0
> [ 11.973950] [<811cf8c1>] ? __nla_reserve+0x21/0xd0
> [ 11.973950] [<817ca7ab>] dev_get_stats+0xcb/0x130
> [ 11.973950] [<817ca7ab>] dev_get_stats+0xcb/0x130
> [ 11.973950] [<817d5409>] rtnl_fill_ifinfo+0x569/0xe20
> [ 11.973950] [<817d5409>] rtnl_fill_ifinfo+0x569/0xe20
> [ 11.973950] [<810352e0>] ? kvm_clock_read+0x20/0x30
> [ 11.973950] [<810352e0>] ? kvm_clock_read+0x20/0x30
> [ 11.973950] [<81008e38>] ? sched_clock+0x8/0x10
> [ 11.973950] [<81008e38>] ? sched_clock+0x8/0x10
> [ 11.973950] [<8106ba45>] ? sched_clock_local+0x25/0x170
> [ 11.973950] [<8106ba45>] ? sched_clock_local+0x25/0x170
> [ 11.973950] [<810da6bd>] ? __kmalloc+0x3d/0x90
> [ 11.973950] [<810da6bd>] ? __kmalloc+0x3d/0x90
> [ 11.973950] [<817b8c10>] ? __kmalloc_reserve.isra.41+0x20/0x70
> [ 11.973950] [<817b8c10>] ? __kmalloc_reserve.isra.41+0x20/0x70
> [ 11.973950] [<810da81a>] ? slob_alloc_node+0x2a/0x60
> [ 11.973950] [<810da81a>] ? slob_alloc_node+0x2a/0x60
> [ 11.973950] [<817b919a>] ? __alloc_skb+0x6a/0x2b0
> [ 11.973950] [<817b919a>] ? __alloc_skb+0x6a/0x2b0
> [ 11.973950] [<817d8795>] rtmsg_ifinfo+0x65/0xe0
> [ 11.973950] [<817d8795>] rtmsg_ifinfo+0x65/0xe0
> [ 11.973950] [<817cbd31>] register_netdevice+0x531/0x5a0
> [ 11.973950] [<817cbd31>] register_netdevice+0x531/0x5a0
> [ 11.973950] [<81892b87>] ? ip6_tnl_get_cap+0x27/0x90
> [ 11.973950] [<81892b87>] ? ip6_tnl_get_cap+0x27/0x90
> [ 11.973950] [<817cbdb6>] register_netdev+0x16/0x30
> [ 11.973950] [<817cbdb6>] register_netdev+0x16/0x30
> [ 11.973950] [<81f574a6>] vti6_init_net+0x1c4/0x1d4
> [ 11.973950] [<81f574a6>] vti6_init_net+0x1c4/0x1d4
> [ 11.973950] [<81f573af>] ? vti6_init_net+0xcd/0x1d4
> [ 11.973950] [<81f573af>] ? vti6_init_net+0xcd/0x1d4
> [ 11.973950] [<817c16df>] ops_init.constprop.11+0x17f/0x1c0
> [ 11.973950] [<817c16df>] ops_init.constprop.11+0x17f/0x1c0
> [ 11.973950] [<817c1779>] register_pernet_operations.isra.9+0x59/0x90
> [ 11.973950] [<817c1779>] register_pernet_operations.isra.9+0x59/0x90
> [ 11.973950] [<817c18d1>] register_pernet_device+0x21/0x60
> [ 11.973950] [<817c18d1>] register_pernet_device+0x21/0x60
> [ 11.973950] [<81f574b6>] ? vti6_init_net+0x1d4/0x1d4
> [ 11.973950] [<81f574b6>] ? vti6_init_net+0x1d4/0x1d4
> [ 11.973950] [<81f574c7>] vti6_tunnel_init+0x11/0x68
> [ 11.973950] [<81f574c7>] vti6_tunnel_init+0x11/0x68
> [ 11.973950] [<81f572a1>] ? mip6_init+0x73/0xb4
> [ 11.973950] [<81f572a1>] ? mip6_init+0x73/0xb4
> [ 11.973950] [<81f0cba4>] do_one_initcall+0xbb/0x15b
> [ 11.973950] [<81f0cba4>] do_one_initcall+0xbb/0x15b
> [ 11.973950] [<811a00d8>] ? sha_transform+0x528/0x1150
> [ 11.973950] [<811a00d8>] ? sha_transform+0x528/0x1150
> [ 11.973950] [<81f0c544>] ? repair_env_string+0x12/0x51
> [ 11.973950] [<81f0c544>] ? repair_env_string+0x12/0x51
> [ 11.973950] [<8105c30d>] ? parse_args+0x2ad/0x440
> [ 11.973950] [<8105c30d>] ? parse_args+0x2ad/0x440
> [ 11.973950] [<810546be>] ? __usermodehelper_set_disable_depth+0x3e/0x50
> [ 11.973950] [<810546be>] ? __usermodehelper_set_disable_depth+0x3e/0x50
> [ 11.973950] [<81f0cd27>] kernel_init_freeable+0xe3/0x182
> [ 11.973950] [<81f0cd27>] kernel_init_freeable+0xe3/0x182
> [ 11.973950] [<81f0c532>] ? do_early_param+0x7a/0x7a
> [ 11.973950] [<81f0c532>] ? do_early_param+0x7a/0x7a
> [ 11.973950] [<819b5b1b>] kernel_init+0xb/0x100
> [ 11.973950] [<819b5b1b>] kernel_init+0xb/0x100
> [ 11.973950] [<819cebf7>] ret_from_kernel_thread+0x1b/0x28
> [ 11.973950] [<819cebf7>] ret_from_kernel_thread+0x1b/0x28
> [ 11.973950] [<819b5b10>] ? rest_init+0xc0/0xc0
> [ 11.973950] [<819b5b10>] ? rest_init+0xc0/0xc0
>
> Before 469bdcefdc ("ipv6: fix the use of pcpu_tstats in ip6_vti.c"),
> the pcpu_tstats.syncp is not used to pretect the 64bit elements of
> pcpu_tstats, so not appear this calltrace.
>
> Reported-by: Fengguang Wu <fengguang.wu@intel.com>
> Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
> Signed-off-by: David S. Miller <davem@davemloft.net>
> ---
> net/ipv6/ip6_vti.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/net/ipv6/ip6_vti.c b/net/ipv6/ip6_vti.c
> index a4564b0..7b42d5e 100644
> --- a/net/ipv6/ip6_vti.c
> +++ b/net/ipv6/ip6_vti.c
> @@ -732,12 +732,18 @@ static void vti6_dev_setup(struct net_device *dev)
> static inline int vti6_dev_init_gen(struct net_device *dev)
> {
> struct ip6_tnl *t = netdev_priv(dev);
> + int i;
>
> t->dev = dev;
> t->net = dev_net(dev);
> dev->tstats = alloc_percpu(struct pcpu_tstats);
> if (!dev->tstats)
> return -ENOMEM;
> + for_each_possible_cpu(i) {
> + struct pcpu_tstats *stats;
> + stats = per_cpu_ptr(dev->tstats, i);
> + u64_stats_init(&stats->syncp);
> + }
> return 0;
> }
>
> --
> 1.7.11.7
^ permalink raw reply
* [PATCH net-next 0/6] mvneta driver performance improvements
From: Willy Tarreau @ 2014-01-12 11:24 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau
Hi,
this patch series implements several performance improvements on the
mvneta driver.
- The first 3 patches are essentially cleanups, code deduplication
and minor optimizations for not re-fetching a value we already have
(status).
- patch 4 changes the prefetch of Rx descriptor from current one to
next one. In benchmarks, it results in about 1% general performance
increase on HTTP traffic, probably because prefetching the current
descriptor does not leave enough time between the start of prefetch
and its usage.
- patch 5 implements support for build_skb() on Rx path. The driver
now preallocates frags instead of skbs and builds an skb just before
delivering it. This results in a 2% performance increase on HTTP
traffic, and up to 5% on small packet Rx rate.
- patch 6 implements rx_copybreak for small packets (256 bytes). It
avoids a dma_map_single()/dma_unmap_single() and increases the Rx
rate by 16.4%, from 486kpps to 573kpps. Further improvements up to
711kpps are possible depending how the DMA is used.
This patch series depends on the previous series of fixes and is only
for the net-next tree.
Thanks!
Willy
---
Willy Tarreau (6):
net: mvneta: remove tests for impossible cases in the tx_done path
net: mvneta: factor rx refilling code
net: mvneta: simplify access to the rx descriptor status
net: mvneta: prefetch next rx descriptor instead of current one
net: mvneta: convert to build_skb()
net: mvneta: implement rx_copybreak
drivers/net/ethernet/marvell/mvneta.c | 152 +++++++++++++++++++++-------------
1 file changed, 95 insertions(+), 57 deletions(-)
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply
* [PATCH net-next 1/6] net: mvneta: remove tests for impossible cases in the tx_done path
From: Willy Tarreau @ 2014-01-12 11:24 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389525848-1814-1-git-send-email-w@1wt.eu>
Currently, mvneta_txq_bufs_free() calls mvneta_tx_done_policy() with
a non-null cause to retrieve the pointer to the next queue to process.
There are useless tests on the return queue number and on the pointer,
all of which are well defined within a known limited set. This code
path is fast, although not critical. Removing 3 tests here that the
compiler could not optimize (verified) is always desirable.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 15 +++++++++------
1 file changed, 9 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index df75a23..2fb9559 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -1276,13 +1276,16 @@ static void mvneta_rx_csum(struct mvneta_port *pp,
skb->ip_summed = CHECKSUM_NONE;
}
-/* Return tx queue pointer (find last set bit) according to causeTxDone reg */
+/* Return tx queue pointer (find last set bit) according to <cause> returned
+ * form tx_done reg. <cause> must not be null. The return value is always a
+ * valid queue for matching the first one found in <cause>.
+ */
static struct mvneta_tx_queue *mvneta_tx_done_policy(struct mvneta_port *pp,
u32 cause)
{
int queue = fls(cause) - 1;
- return (queue < 0 || queue >= txq_number) ? NULL : &pp->txqs[queue];
+ return &pp->txqs[queue];
}
/* Free tx queue skbuffs */
@@ -1651,7 +1654,9 @@ static void mvneta_txq_done_force(struct mvneta_port *pp,
txq->txq_get_index = 0;
}
-/* handle tx done - called from tx done timer callback */
+/* Handle tx done - called in softirq context. The <cause_tx_done> argument
+ * must be a valid cause according to MVNETA_TXQ_INTR_MASK_ALL.
+ */
static u32 mvneta_tx_done_gbe(struct mvneta_port *pp, u32 cause_tx_done,
int *tx_todo)
{
@@ -1660,10 +1665,8 @@ static u32 mvneta_tx_done_gbe(struct mvneta_port *pp, u32 cause_tx_done,
struct netdev_queue *nq;
*tx_todo = 0;
- while (cause_tx_done != 0) {
+ while (cause_tx_done) {
txq = mvneta_tx_done_policy(pp, cause_tx_done);
- if (!txq)
- break;
nq = netdev_get_tx_queue(pp->dev, txq->id);
__netif_tx_lock(nq, smp_processor_id());
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH net-next 3/6] net: mvneta: simplify access to the rx descriptor status
From: Willy Tarreau @ 2014-01-12 11:24 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389525848-1814-1-git-send-email-w@1wt.eu>
At several places, we already know the value of the rx status but
we call functions which dereference the pointer again to get it
and don't need the descriptor for anything else. Simplify this
task by replacing the rx desc pointer by the status word itself.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 25 ++++++++++++-------------
1 file changed, 12 insertions(+), 13 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index eccafd1..aa3a4f7 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -528,14 +528,14 @@ struct rtnl_link_stats64 *mvneta_get_stats64(struct net_device *dev,
/* Rx descriptors helper methods */
-/* Checks whether the given RX descriptor is both the first and the
- * last descriptor for the RX packet. Each RX packet is currently
+/* Checks whether the RX descriptor having this status is both the first
+ * and the last descriptor for the RX packet. Each RX packet is currently
* received through a single RX descriptor, so not having each RX
* descriptor with its first and last bits set is an error
*/
-static int mvneta_rxq_desc_is_first_last(struct mvneta_rx_desc *desc)
+static int mvneta_rxq_desc_is_first_last(u32 status)
{
- return (desc->status & MVNETA_RXD_FIRST_LAST_DESC) ==
+ return (status & MVNETA_RXD_FIRST_LAST_DESC) ==
MVNETA_RXD_FIRST_LAST_DESC;
}
@@ -1234,10 +1234,10 @@ static void mvneta_rx_error(struct mvneta_port *pp,
{
u32 status = rx_desc->status;
- if (!mvneta_rxq_desc_is_first_last(rx_desc)) {
+ if (!mvneta_rxq_desc_is_first_last(status)) {
netdev_err(pp->dev,
"bad rx status %08x (buffer oversize), size=%d\n",
- rx_desc->status, rx_desc->data_size);
+ status, rx_desc->data_size);
return;
}
@@ -1261,13 +1261,12 @@ static void mvneta_rx_error(struct mvneta_port *pp,
}
}
-/* Handle RX checksum offload */
-static void mvneta_rx_csum(struct mvneta_port *pp,
- struct mvneta_rx_desc *rx_desc,
+/* Handle RX checksum offload based on the descriptor's status */
+static void mvneta_rx_csum(struct mvneta_port *pp, u32 status,
struct sk_buff *skb)
{
- if ((rx_desc->status & MVNETA_RXD_L3_IP4) &&
- (rx_desc->status & MVNETA_RXD_L4_CSUM_OK)) {
+ if ((status & MVNETA_RXD_L3_IP4) &&
+ (status & MVNETA_RXD_L4_CSUM_OK)) {
skb->csum = 0;
skb->ip_summed = CHECKSUM_UNNECESSARY;
return;
@@ -1449,7 +1448,7 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
rx_status = rx_desc->status;
skb = (struct sk_buff *)rx_desc->buf_cookie;
- if (!mvneta_rxq_desc_is_first_last(rx_desc) ||
+ if (!mvneta_rxq_desc_is_first_last(rx_status) ||
(rx_status & MVNETA_RXD_ERR_SUMMARY)) {
dev->stats.rx_errors++;
mvneta_rx_error(pp, rx_desc);
@@ -1472,7 +1471,7 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
skb->protocol = eth_type_trans(skb, dev);
- mvneta_rx_csum(pp, rx_desc, skb);
+ mvneta_rx_csum(pp, rx_status, skb);
napi_gro_receive(&pp->napi, skb);
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH net-next 5/6] net: mvneta: convert to build_skb()
From: Willy Tarreau @ 2014-01-12 11:24 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389525848-1814-1-git-send-email-w@1wt.eu>
Make use of build_skb() to allocate frags on the RX path. When frag size
is lower than a page size, we can use netdev_alloc_frag(), and we fall back
to kmalloc() for larger sizes. The frag size is stored into the mvneta_port
struct. The alloc/free functions check the frag size to decide what alloc/
free method to use. MTU changes are safe because the MTU change function
stops the device and clears the queues before applying the change.
With this patch, I observed a reproducible 2% performance improvement on
HTTP-based benchmarks, and 5% on small packet RX rate.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 49 +++++++++++++++++++++++++----------
1 file changed, 35 insertions(+), 14 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index c7b37e0..726a8d2 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -268,6 +268,7 @@ struct mvneta_pcpu_stats {
struct mvneta_port {
int pkt_size;
+ unsigned int frag_size;
void __iomem *base;
struct mvneta_rx_queue *rxqs;
struct mvneta_tx_queue *txqs;
@@ -1332,28 +1333,43 @@ static int mvneta_txq_done(struct mvneta_port *pp,
return tx_done;
}
+static void *mvneta_frag_alloc(const struct mvneta_port *pp)
+{
+ if (likely(pp->frag_size <= PAGE_SIZE))
+ return netdev_alloc_frag(pp->frag_size);
+ else
+ return kmalloc(pp->frag_size, GFP_ATOMIC);
+}
+
+static void mvneta_frag_free(const struct mvneta_port *pp, void *data)
+{
+ if (likely(pp->frag_size <= PAGE_SIZE))
+ put_page(virt_to_head_page(data));
+ else
+ kfree(data);
+}
+
/* Refill processing */
static int mvneta_rx_refill(struct mvneta_port *pp,
struct mvneta_rx_desc *rx_desc)
{
dma_addr_t phys_addr;
- struct sk_buff *skb;
+ void *data;
- skb = netdev_alloc_skb(pp->dev, pp->pkt_size);
- if (!skb)
+ data = mvneta_frag_alloc(pp);
+ if (!data)
return -ENOMEM;
- phys_addr = dma_map_single(pp->dev->dev.parent, skb->head,
+ phys_addr = dma_map_single(pp->dev->dev.parent, data,
MVNETA_RX_BUF_SIZE(pp->pkt_size),
DMA_FROM_DEVICE);
if (unlikely(dma_mapping_error(pp->dev->dev.parent, phys_addr))) {
- dev_kfree_skb(skb);
+ mvneta_frag_free(pp, data);
return -ENOMEM;
}
- mvneta_rx_desc_fill(rx_desc, phys_addr, (u32)skb);
-
+ mvneta_rx_desc_fill(rx_desc, phys_addr, (u32)data);
return 0;
}
@@ -1407,9 +1423,9 @@ static void mvneta_rxq_drop_pkts(struct mvneta_port *pp,
rx_done = mvneta_rxq_busy_desc_num_get(pp, rxq);
for (i = 0; i < rxq->size; i++) {
struct mvneta_rx_desc *rx_desc = rxq->descs + i;
- struct sk_buff *skb = (struct sk_buff *)rx_desc->buf_cookie;
+ void *data = (void *)rx_desc->buf_cookie;
- dev_kfree_skb_any(skb);
+ mvneta_frag_free(pp, data);
dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
}
@@ -1440,20 +1456,21 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
while (rx_done < rx_todo) {
struct mvneta_rx_desc *rx_desc = mvneta_rxq_next_desc_get(rxq);
struct sk_buff *skb;
+ unsigned char *data;
u32 rx_status;
int rx_bytes, err;
rx_done++;
rx_filled++;
rx_status = rx_desc->status;
- skb = (struct sk_buff *)rx_desc->buf_cookie;
+ data = (unsigned char *)rx_desc->buf_cookie;
if (!mvneta_rxq_desc_is_first_last(rx_status) ||
- (rx_status & MVNETA_RXD_ERR_SUMMARY)) {
+ (rx_status & MVNETA_RXD_ERR_SUMMARY) ||
+ !(skb = build_skb(data, pp->frag_size > PAGE_SIZE ? 0 : pp->frag_size))) {
dev->stats.rx_errors++;
mvneta_rx_error(pp, rx_desc);
- mvneta_rx_desc_fill(rx_desc, rx_desc->buf_phys_addr,
- (u32)skb);
+ /* leave the descriptor untouched */
continue;
}
@@ -1466,7 +1483,7 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
rcvd_bytes += rx_bytes;
/* Linux processing */
- skb_reserve(skb, MVNETA_MH_SIZE);
+ skb_reserve(skb, MVNETA_MH_SIZE + NET_SKB_PAD);
skb_put(skb, rx_bytes);
skb->protocol = eth_type_trans(skb, dev);
@@ -2276,6 +2293,8 @@ static int mvneta_change_mtu(struct net_device *dev, int mtu)
mvneta_cleanup_rxqs(pp);
pp->pkt_size = MVNETA_RX_PKT_SIZE(pp->dev->mtu);
+ pp->frag_size = SKB_DATA_ALIGN(MVNETA_RX_BUF_SIZE(pp->pkt_size)) +
+ SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
ret = mvneta_setup_rxqs(pp);
if (ret) {
@@ -2423,6 +2442,8 @@ static int mvneta_open(struct net_device *dev)
mvneta_mac_addr_set(pp, dev->dev_addr, rxq_def);
pp->pkt_size = MVNETA_RX_PKT_SIZE(pp->dev->mtu);
+ pp->frag_size = SKB_DATA_ALIGN(MVNETA_RX_BUF_SIZE(pp->pkt_size)) +
+ SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
ret = mvneta_setup_rxqs(pp);
if (ret)
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH net-next 2/6] net: mvneta: factor rx refilling code
From: Willy Tarreau @ 2014-01-12 11:24 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389525848-1814-1-git-send-email-w@1wt.eu>
Make mvneta_rxq_fill() use mvneta_rx_refill() instead of using
duplicate code.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 23 +++--------------------
1 file changed, 3 insertions(+), 20 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 2fb9559..eccafd1 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -1969,32 +1969,15 @@ static int mvneta_poll(struct napi_struct *napi, int budget)
static int mvneta_rxq_fill(struct mvneta_port *pp, struct mvneta_rx_queue *rxq,
int num)
{
- struct net_device *dev = pp->dev;
int i;
for (i = 0; i < num; i++) {
- struct sk_buff *skb;
- struct mvneta_rx_desc *rx_desc;
- unsigned long phys_addr;
-
- skb = dev_alloc_skb(pp->pkt_size);
- if (!skb) {
- netdev_err(dev, "%s:rxq %d, %d of %d buffs filled\n",
+ memset(rxq->descs + i, 0, sizeof(struct mvneta_rx_desc));
+ if (mvneta_rx_refill(pp, rxq->descs + i) != 0) {
+ netdev_err(pp->dev, "%s:rxq %d, %d of %d buffs filled\n",
__func__, rxq->id, i, num);
break;
}
-
- rx_desc = rxq->descs + i;
- memset(rx_desc, 0, sizeof(struct mvneta_rx_desc));
- phys_addr = dma_map_single(dev->dev.parent, skb->head,
- MVNETA_RX_BUF_SIZE(pp->pkt_size),
- DMA_FROM_DEVICE);
- if (unlikely(dma_mapping_error(dev->dev.parent, phys_addr))) {
- dev_kfree_skb(skb);
- break;
- }
-
- mvneta_rx_desc_fill(rx_desc, phys_addr, (u32)skb);
}
/* Add this number of RX descriptors as non occupied (ready to
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH net-next 4/6] net: mvneta: prefetch next rx descriptor instead of current one
From: Willy Tarreau @ 2014-01-12 11:24 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389525848-1814-1-git-send-email-w@1wt.eu>
Currently, the mvneta driver tries to prefetch the current Rx
descriptor during read. Tests have shown that prefetching the
next one instead increases general performance by about 1% on
HTTP traffic.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index aa3a4f7..c7b37e0 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -611,6 +611,7 @@ mvneta_rxq_next_desc_get(struct mvneta_rx_queue *rxq)
int rx_desc = rxq->next_desc_to_proc;
rxq->next_desc_to_proc = MVNETA_QUEUE_NEXT_DESC(rxq, rx_desc);
+ prefetch(rxq->descs + rxq->next_desc_to_proc);
return rxq->descs + rx_desc;
}
@@ -1442,7 +1443,6 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
u32 rx_status;
int rx_bytes, err;
- prefetch(rx_desc);
rx_done++;
rx_filled++;
rx_status = rx_desc->status;
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH net-next 6/6] net: mvneta: implement rx_copybreak
From: Willy Tarreau @ 2014-01-12 11:24 UTC (permalink / raw)
To: davem; +Cc: netdev, Willy Tarreau, Thomas Petazzoni, Gregory CLEMENT
In-Reply-To: <1389525848-1814-1-git-send-email-w@1wt.eu>
calling dma_map_single()/dma_unmap_single() is quite expensive compared
to copying a small packet. So let's copy short frames and keep the buffers
mapped. We set the limit to 256 bytes which seems to give good results both
on the XP-GP board and on the AX3/4.
The Rx small packet rate increased by 16.4% doing this, from 486kpps to
573kpps. It is worth noting that even the call to the function
dma_sync_single_range_for_cpu() is expensive (300 ns) although less
than dma_unmap_single(). Without it, the packet rate raises to 711kpps
(+24% more). Thus on systems where coherency from device to CPU is
guaranteed by a snoop control unit, this patch should provide even more
gains, and probably rx_copybreak could be increased.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
---
drivers/net/ethernet/marvell/mvneta.c | 44 ++++++++++++++++++++++++++++++-----
1 file changed, 38 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 726a8d2..f5fc7a2 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -444,6 +444,8 @@ static int txq_number = 8;
static int rxq_def;
+static int rx_copybreak __read_mostly = 256;
+
#define MVNETA_DRIVER_NAME "mvneta"
#define MVNETA_DRIVER_VERSION "1.0"
@@ -1463,22 +1465,51 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
rx_done++;
rx_filled++;
rx_status = rx_desc->status;
+ rx_bytes = rx_desc->data_size - (ETH_FCS_LEN + MVNETA_MH_SIZE);
data = (unsigned char *)rx_desc->buf_cookie;
if (!mvneta_rxq_desc_is_first_last(rx_status) ||
- (rx_status & MVNETA_RXD_ERR_SUMMARY) ||
- !(skb = build_skb(data, pp->frag_size > PAGE_SIZE ? 0 : pp->frag_size))) {
+ (rx_status & MVNETA_RXD_ERR_SUMMARY)) {
+ err_drop_frame:
dev->stats.rx_errors++;
mvneta_rx_error(pp, rx_desc);
/* leave the descriptor untouched */
continue;
}
- dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
+ if (rx_bytes <= rx_copybreak) {
+ /* better copy a small frame and not unmap the DMA region */
+ skb = netdev_alloc_skb_ip_align(dev, rx_bytes);
+ if (unlikely(!skb))
+ goto err_drop_frame;
+
+ dma_sync_single_range_for_cpu(dev->dev.parent,
+ rx_desc->buf_phys_addr,
+ MVNETA_MH_SIZE + NET_SKB_PAD,
+ rx_bytes,
+ DMA_FROM_DEVICE);
+ memcpy(skb_put(skb, rx_bytes),
+ data + MVNETA_MH_SIZE + NET_SKB_PAD,
+ rx_bytes);
+
+ skb->protocol = eth_type_trans(skb, dev);
+ mvneta_rx_csum(pp, rx_status, skb);
+ napi_gro_receive(&pp->napi, skb);
+
+ rcvd_pkts++;
+ rcvd_bytes += rx_bytes;
+
+ /* leave the descriptor and buffer untouched */
+ continue;
+ }
+
+ skb = build_skb(data, pp->frag_size > PAGE_SIZE ? 0 : pp->frag_size);
+ if (!skb)
+ goto err_drop_frame;
+
+ dma_unmap_single(dev->dev.parent, rx_desc->buf_phys_addr,
MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
- rx_bytes = rx_desc->data_size -
- (ETH_FCS_LEN + MVNETA_MH_SIZE);
rcvd_pkts++;
rcvd_bytes += rx_bytes;
@@ -1495,7 +1526,7 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
/* Refill processing */
err = mvneta_rx_refill(pp, rx_desc);
if (err) {
- netdev_err(pp->dev, "Linux processing - Can't refill\n");
+ netdev_err(dev, "Linux processing - Can't refill\n");
rxq->missed++;
rx_filled--;
}
@@ -2945,3 +2976,4 @@ module_param(rxq_number, int, S_IRUGO);
module_param(txq_number, int, S_IRUGO);
module_param(rxq_def, int, S_IRUGO);
+module_param(rx_copybreak, int, S_IRUGO | S_IWUSR);
--
1.7.12.2.21.g234cd45.dirty
^ permalink raw reply related
* [PATCH] MAINTAINERS: add virtio-dev ML for virtio
From: Michael S. Tsirkin @ 2014-01-12 11:37 UTC (permalink / raw)
To: linux-kernel
Cc: Rusty Russell, Andrew Morton, Joe Perches, Greg Kroah-Hartman,
Mauro Carvalho Chehab, David S. Miller, amit.shah, Jason Wang,
Wanlong Gao, virtio-dev, netdev, virtualization
Since virtio is an OASIS standard draft now, virtio implementation
discussions are taking place on the virtio-dev OASIS mailing list.
Update MAINTAINERS.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
MAINTAINERS | 3 +
drivers/net/virtio_net.c | 355 +++++++++++++++++++++++++----------------------
2 files changed, 191 insertions(+), 167 deletions(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index ffcaf97..75202af 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -9088,6 +9088,7 @@ F: include/media/videobuf2-*
VIRTIO CONSOLE DRIVER
M: Amit Shah <amit.shah@redhat.com>
+L: virtio-dev@lists.oasis-open.org
L: virtualization@lists.linux-foundation.org
S: Maintained
F: drivers/char/virtio_console.c
@@ -9097,6 +9098,7 @@ F: include/uapi/linux/virtio_console.h
VIRTIO CORE, NET AND BLOCK DRIVERS
M: Rusty Russell <rusty@rustcorp.com.au>
M: "Michael S. Tsirkin" <mst@redhat.com>
+L: virtio-dev@lists.oasis-open.org
L: virtualization@lists.linux-foundation.org
S: Maintained
F: drivers/virtio/
@@ -9109,6 +9111,7 @@ F: include/uapi/linux/virtio_*.h
VIRTIO HOST (VHOST)
M: "Michael S. Tsirkin" <mst@redhat.com>
L: kvm@vger.kernel.org
+L: virtio-dev@lists.oasis-open.org
L: virtualization@lists.linux-foundation.org
L: netdev@vger.kernel.org
S: Maintained
--
MST
^ permalink raw reply related
* Re: [PATCH 1/1] When timestamping is enabled, stmmac_tx_clean will call stmmac_get_tx_hwtstamp to get tx TS. It's possible that skb is NULL because there are other network frames that use several descriptors. So we must return immediately in stmmac_get_tx_hwtstamp if skb is NULL to avoid system crash.
From: Daniel Borkmann @ 2014-01-12 11:38 UTC (permalink / raw)
To: Bruce Liu; +Cc: peppe.cavallaro, netdev, linux-kernel
In-Reply-To: <20140112093951.GA3743@gmail.com>
On 01/12/2014 10:39 AM, Bruce Liu wrote:
> When timestamping is enabled, stmmac_tx_clean will call stmmac_get_tx_hwtstamp to get tx TS.
> It's possible that skb is NULL because there are other network frames that use several descriptors.
> So we must return immediately in stmmac_get_tx_hwtstamp if skb is NULL to avoid system crash.
>
>
> Signed-off-by: Bruce Liu <damuzi000@gmail.com>
Please see Documentation/SubmittingPatches +489
You subject line is way too long and should just be something like:
[PATCH net-next] net: stmmac: fix NULL pointer dereference in stmmac_get_tx_hwtstamp
Don't indent your actual commit message with whitespaces as prefix,
and do a line break after around 70 chars.
Btw, I mentioned net-next in the subject since merge window will
open soon anyway.
> ---
> drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> index 797b56a..47f2287 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> @@ -332,7 +332,7 @@ static void stmmac_get_tx_hwtstamp(struct stmmac_priv *priv,
> return;
>
> /* exit if skb doesn't support hw tstamp */
> - if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)))
> + if (likely(!skb || !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)))
> return;
>
> if (priv->adv_ts)
>
^ permalink raw reply
* Re: [RFC net-next 0/3] bonding: new option API
From: Nikolay Aleksandrov @ 2014-01-12 12:09 UTC (permalink / raw)
To: Scott Feldman
Cc: Netdev, Andy Gospodarek, Jay Vosburgh, Veaceslav Falico,
David S. Miller
In-Reply-To: <AAA51DDF-74AF-4A70-8072-62F2593F8619@cumulusnetworks.com>
On 01/10/2014 09:32 PM, Scott Feldman wrote:
>
> On Jan 10, 2014, at 5:11 AM, Nikolay Aleksandrov <nikolay@redhat.com> wrote:
>
>> Hi,
>> This patchset aims to introduce a new option API that can be easily
>> extended if necessary and which attempts to remove some common problems
>> and code. In the beginning there was support for inter-option dependencies,
>> but that turned out to be unnecessary as the only 2 options that _enforce_
>> another option to be set prior to setting are up/down delay and they can be
>> easily re-worked to not require miimon to be set, so we can spare ourselves
>> 100+ lines of checks, dealing with complex dependency errors and such.
>> In case this becomes necessary I've kept the old version of this patch-set
>> which has it, and can easily re-work it at any time.
>> There're still a lot of things to fix/clean but I've done some limited testing
>> with the options that are converted and it seems to work.
>> The main exported functions (as can be seen) are:
>> __bond_opt_set() - to be used when a string is passed which needs to be
>> converted in the case of BOND_OPTVAL_INTEGER. (sysfs)
>> __bond_opt_intset() - to be used when a value is passed to
>> BOND_OPTVAL_INTEGER (netlink), this function can't
>> be used for BOND_OPTVAL_STRING options
>> These two can be used from inside other options to stop them (e.g., arp_interval
>> stopping miimon and vice versa).
>> I've also added bond_opt_tryset_rtnl() mostly for sysfs use.
>> See the description of patch 01 and the comments inside for more information.
>>
>> Value tables of converted options are no longer exported, and can be accessed
>> through the API (bond_opt_get_val() & bond_opt_get_flags).
>> Another good side-effect is that the error codes are standard for all options
>> for the common errors at least.
>
> Nice!
>
>> When/if this patchset is posted for inclusion, I'll have all options converted.
>> I actually had them before but while on vacation during December a lot of code
>> went in changing the bonding options and have to re-work most of the patches.
>
> Oops, sorry about that ;)
>
>> Some of the future plans for this are:
>> Verbose outputting of dependencies (done, just have to polish it)
>> Automatic sysfs generation from the bond_opts[].
>
> I had a patch in my queue to do something similar, but yours is so much nicer.
>
> For sysfs nodes, there is a file permission. I wonder if bond_opts should have a sense of RO, RW, or WO? Then automatic sysfs generation is even easier. Hmmm, actually I think the answer is no. Nevermind.
>
>> Use of the API in bond_check_params() and thus cleaning it up
>> Structure for accessor fn parameter passing so we can implement get/set
>> in a more general manner
>>
>> Sending it with 2 options converted which illustrate the use of different
>> features of the API. I've tested them via sysfs.
>>
>> Any thoughts, comments and suggestions are very welcome.
>
> Nice job Nik, well done.
>
> -scott
>
Hi Scott,
Thank you for the review, I'll take care of the comments for the first version.
I saw your first comment about generalizing this to more than just the
bonding, I'll look into that because I know I'm reinventing the wheel here
as there are already network drivers which have similar APIs, but such
project would have much larger set of requirements (e.g. a much better and
more descriptive inter-dependency checks, custom errors, custom
permissions, more opaque objects etc.). Maybe we can leave it as the next
step, I have to give it some more thought :-)
Cheers,
Nik
^ permalink raw reply
* Re: [RFC Patch net-next 4/4] net_sched: make ingress qdisc lockless
From: Jamal Hadi Salim @ 2014-01-12 12:30 UTC (permalink / raw)
To: John Fastabend, Cong Wang
Cc: Stephen Hemminger, Eric Dumazet, Linux Kernel Network Developers,
John Fastabend, David S. Miller
In-Reply-To: <52CF4795.106@intel.com>
On 01/09/14 20:06, John Fastabend wrote:
> Just to re-iterate you need to go through each and every qdisc,
> classifier, action and verify it is safe to run in parallel. Take
> a look at how the skb lists are managed in the qdiscs. If we want
> to do this we need to make these changes in some coherent way
> because it touches lots of pieces.
>
Indeed. Everything assumes the global qdisc lock is protecting them.
Actually actions are probably the best at the moment because
the lock is very fine grained to just the action instance
and it protects both control and data paths.
But filters have stuff littered everywhere. Egress qdiscs
as you mention have queues at multi levels etc.
> Also your stats are going to get hosed none of the bstats, qstats
> supports this.
Stats are probably the easiest to "fix".
Didnt Eric (or somebody else) fix netdev level stats to use seq counts?
Would that idea not be applicable here?
>I'll send out the classifier set later tonight
> if you want. I got stalled going through the actions.
>
The thing to note is:
actions can be shared across filters, netdevices and cpus.
By default they are not shared across filters and netdevices that is
a config option. You still have to worry about sharing across cpus
which will happen because a flow can be shared across cpus.
You could probably get rid of the lock if you can show
that you can make data and control path mutually exclusive
(rtnl will protect control path).
> Finally any global state in those qdiscs is going to drive performance
> down so many of them would likely need to be redesigned.
>
I feel like per-cpu qdiscs is the best surgery at the moment.
cheers,
jamal
^ permalink raw reply
* Re: [Patch net-next 0/7] net_sched: some more cleanup and improvements
From: Jamal Hadi Salim @ 2014-01-12 12:32 UTC (permalink / raw)
To: Cong Wang, netdev; +Cc: David S. Miller
In-Reply-To: <1389312845-10304-1-git-send-email-xiyou.wangcong@gmail.com>
On 01/09/14 19:13, Cong Wang wrote:
> This patchset collects the previous patches I sent which Jamal doesn't object.
> They are still some cleanup and improvements for tc actions and filters.
>
>
Cong - I dont have time to test at the moment and like you said
they dont look controversial; i will review them, please try to test
them to the best you can.
cheers,
jamal
^ permalink raw reply
* Re: [Patch net-next 1/7] net_sched: act: move idx_gen into struct tcf_hashinfo
From: Jamal Hadi Salim @ 2014-01-12 12:34 UTC (permalink / raw)
To: Cong Wang, netdev; +Cc: David S. Miller
In-Reply-To: <1389312845-10304-2-git-send-email-xiyou.wangcong@gmail.com>
On 01/09/14 19:13, Cong Wang wrote:
> There is no need to store the index separatedly
> since tcf_hashinfo is allocated statically too.
>
> Cc: Jamal Hadi Salim <jhs@mojatatu.com>
> Cc: David S. Miller <davem@davemloft.net>
> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
An improvement for sure. Thanks Cong!
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
^ permalink raw reply
* Re: [Patch net-next 2/7] net_sched: act: clean up notification functions
From: Jamal Hadi Salim @ 2014-01-12 12:38 UTC (permalink / raw)
To: Cong Wang, netdev; +Cc: David S. Miller
In-Reply-To: <1389312845-10304-3-git-send-email-xiyou.wangcong@gmail.com>
On 01/09/14 19:14, Cong Wang wrote:
> Refactor tcf_add_notify() and factor out tcf_del_notify().
>
> Cc: Jamal Hadi Salim <jhs@mojatatu.com>
> Cc: David S. Miller <davem@davemloft.net>
> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Thanks Cong!
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
^ 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