* [PATCH net-next 1/4] net: allow large number of rx queues
From: Pankaj Gupta @ 2014-11-18 16:22 UTC (permalink / raw)
To: linux-kernel, netdev
Cc: davem, jasowang, mst, dgibson, vfalico, edumazet, vyasevic, hkchu,
wuzhy, xemul, therbert, bhutchings, xii, stephen, jiri,
sergei.shtylyov, Pankaj Gupta
In-Reply-To: <1416327778-17716-1-git-send-email-pagupta@redhat.com>
netif_alloc_rx_queues() uses kcalloc() to allocate memory
for "struct netdev_queue *_rx" array.
If we are doing large rx queue allocation kcalloc() might
fail, so this patch does a fallback to vzalloc().
Similar implementation is done for tx queue allocation in
netif_alloc_netdev_queues().
We avoid failure of high order memory allocation
with the help of vzalloc(), this allows us to do large
rx and tx queue allocation which in turn helps us to
increase the number of queues in tun.
As vmalloc() adds overhead on a critical network path,
__GFP_REPEAT flag is used with kzalloc() to do this fallback
only when really needed.
Signed-off-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: David Gibson <dgibson@redhat.com>
---
net/core/dev.c | 19 +++++++++++++------
1 file changed, 13 insertions(+), 6 deletions(-)
diff --git a/net/core/dev.c b/net/core/dev.c
index e916ba8..abe9560 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6059,17 +6059,25 @@ void netif_stacked_transfer_operstate(const struct net_device *rootdev,
EXPORT_SYMBOL(netif_stacked_transfer_operstate);
#ifdef CONFIG_SYSFS
+static void netif_free_rx_queues(struct net_device *dev)
+{
+ kvfree(dev->_rx);
+}
+
static int netif_alloc_rx_queues(struct net_device *dev)
{
unsigned int i, count = dev->num_rx_queues;
struct netdev_rx_queue *rx;
+ size_t sz = count * sizeof(*rx);
BUG_ON(count < 1);
- rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
- if (!rx)
- return -ENOMEM;
-
+ rx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
+ if (!rx) {
+ rx = vzalloc(sz);
+ if (!rx)
+ return -ENOMEM;
+ }
dev->_rx = rx;
for (i = 0; i < count; i++)
@@ -6698,9 +6706,8 @@ void free_netdev(struct net_device *dev)
netif_free_tx_queues(dev);
#ifdef CONFIG_SYSFS
- kfree(dev->_rx);
+ netif_free_rx_queues(dev);
#endif
-
kfree(rcu_dereference_protected(dev->ingress_queue, 1));
/* Flush device addresses */
--
1.8.3.1
^ permalink raw reply related
* [PATCH v4 3/4] r8169: Use coherent_rmb() and coherent_wmb() for DescOwn checks
From: Alexander Duyck @ 2014-11-18 17:29 UTC (permalink / raw)
To: linux-arch, netdev, linux-kernel
Cc: mathieu.desnoyers, peterz, benh, heiko.carstens, mingo, mikey,
linux, donald.c.skidmore, matthew.vick, geert, jeffrey.t.kirsher,
romieu, paulmck, nic_swsd, will.deacon, michael, tony.luck,
torvalds, oleg, schwidefsky, fweisbec, davem
In-Reply-To: <20141118172644.26303.37688.stgit@ahduyck-server>
The r8169 use a pair of wmb() calls when setting up the descriptor rings.
The first is to synchronize the descriptor data with the descriptor status,
and the second is to synchronize the descriptor status with the use of the
MMIO doorbell to notify the device that descriptors are ready. This can
come at a heavy price on some systems, and is not really necessary on
systems such as x86 as a simple barrier() would suffice to order store/store
accesses. As such we can replace the first memory barrier with
coherent_wmb() to reduce the cost for these accesses.
In addition the r8169 uses a rmb() to prevent compiler optimization in the
cleanup paths, however by moving the barrier down a few lines and replacing
it with a coherent_rmb() we should be able to use it to guarantee
descriptor accesses do not occur until the device has updated the DescOwn
bit from its end.
One last change I made is to move the update of cur_tx in the xmit path to
after the wmb. This way we can guarantee the device and all CPUs should
see the DescOwn update before they see the cur_tx value update.
Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>
Cc: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
---
drivers/net/ethernet/realtek/r8169.c | 29 +++++++++++++++++++++--------
1 file changed, 21 insertions(+), 8 deletions(-)
diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
index cf154f7..2886c4a 100644
--- a/drivers/net/ethernet/realtek/r8169.c
+++ b/drivers/net/ethernet/realtek/r8169.c
@@ -6601,6 +6601,9 @@ static inline void rtl8169_mark_to_asic(struct RxDesc *desc, u32 rx_buf_sz)
{
u32 eor = le32_to_cpu(desc->opts1) & RingEnd;
+ /* Force memory writes to complete before releasing descriptor */
+ coherent_wmb();
+
desc->opts1 = cpu_to_le32(DescOwn | eor | rx_buf_sz);
}
@@ -6608,7 +6611,6 @@ static inline void rtl8169_map_to_asic(struct RxDesc *desc, dma_addr_t mapping,
u32 rx_buf_sz)
{
desc->addr = cpu_to_le64(mapping);
- wmb();
rtl8169_mark_to_asic(desc, rx_buf_sz);
}
@@ -7077,16 +7079,18 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
skb_tx_timestamp(skb);
- wmb();
+ /* Force memory writes to complete before releasing descriptor */
+ coherent_wmb();
/* Anti gcc 2.95.3 bugware (sic) */
status = opts[0] | len | (RingEnd * !((entry + 1) % NUM_TX_DESC));
txd->opts1 = cpu_to_le32(status);
- tp->cur_tx += frags + 1;
-
+ /* Force all memory writes to complete before notifying device */
wmb();
+ tp->cur_tx += frags + 1;
+
RTL_W8(TxPoll, NPQ);
mmiowb();
@@ -7185,11 +7189,16 @@ static void rtl_tx(struct net_device *dev, struct rtl8169_private *tp)
struct ring_info *tx_skb = tp->tx_skb + entry;
u32 status;
- rmb();
status = le32_to_cpu(tp->TxDescArray[entry].opts1);
if (status & DescOwn)
break;
+ /* This barrier is needed to keep us from reading
+ * any other fields out of the Tx descriptor until
+ * we know the status of DescOwn
+ */
+ coherent_rmb();
+
rtl8169_unmap_tx_skb(&tp->pci_dev->dev, tx_skb,
tp->TxDescArray + entry);
if (status & LastFrag) {
@@ -7284,11 +7293,16 @@ static int rtl_rx(struct net_device *dev, struct rtl8169_private *tp, u32 budget
struct RxDesc *desc = tp->RxDescArray + entry;
u32 status;
- rmb();
status = le32_to_cpu(desc->opts1) & tp->opts1_mask;
-
if (status & DescOwn)
break;
+
+ /* This barrier is needed to keep us from reading
+ * any other fields out of the Rx descriptor until
+ * we know the status of DescOwn
+ */
+ coherent_rmb();
+
if (unlikely(status & RxRES)) {
netif_info(tp, rx_err, dev, "Rx ERROR. status = %08x\n",
status);
@@ -7350,7 +7364,6 @@ process_pkt:
}
release_descriptor:
desc->opts2 = 0;
- wmb();
rtl8169_mark_to_asic(desc, rx_buf_sz);
}
^ permalink raw reply related
* [PATCH v4 4/4] fm10k/igb/ixgbe: Use coherent_rmb on Rx descriptor reads
From: Alexander Duyck @ 2014-11-18 17:29 UTC (permalink / raw)
To: linux-arch, netdev, linux-kernel
Cc: mathieu.desnoyers, peterz, benh, heiko.carstens, mingo, mikey,
linux, donald.c.skidmore, matthew.vick, geert, jeffrey.t.kirsher,
romieu, paulmck, nic_swsd, will.deacon, michael, tony.luck,
torvalds, oleg, schwidefsky, fweisbec, davem
In-Reply-To: <20141118172644.26303.37688.stgit@ahduyck-server>
This change makes it so that coherent_rmb is used when reading the Rx
descriptor. The advantage of coherent_rmb is that it allows for a much
lower cost barrier on x86, powerpc, arm, and arm64 architectures than a
traditional memory barrier when dealing with reads that only have to
synchronize to coherent memory.
In addition I have updated the code so that it just checks to see if any
bits have been set instead of just the DD bit since the DD bit will always
be set as a part of a descriptor write-back so we just need to check for a
non-zero value being present at that memory location rather than just
checking for any specific bit. This allows the code itself to appear much
cleaner and allows the compiler more room to optimize.
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Matthew Vick <matthew.vick@intel.com>
Cc: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
---
drivers/net/ethernet/intel/fm10k/fm10k_main.c | 6 +++---
drivers/net/ethernet/intel/igb/igb_main.c | 6 +++---
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 9 ++++-----
3 files changed, 10 insertions(+), 11 deletions(-)
diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_main.c b/drivers/net/ethernet/intel/fm10k/fm10k_main.c
index e645af4..1819311 100644
--- a/drivers/net/ethernet/intel/fm10k/fm10k_main.c
+++ b/drivers/net/ethernet/intel/fm10k/fm10k_main.c
@@ -620,14 +620,14 @@ static bool fm10k_clean_rx_irq(struct fm10k_q_vector *q_vector,
rx_desc = FM10K_RX_DESC(rx_ring, rx_ring->next_to_clean);
- if (!fm10k_test_staterr(rx_desc, FM10K_RXD_STATUS_DD))
+ if (!rx_desc->d.staterr)
break;
/* This memory barrier is needed to keep us from reading
* any other fields out of the rx_desc until we know the
- * RXD_STATUS_DD bit is set
+ * descriptor has been written back
*/
- rmb();
+ coherent_rmb();
/* retrieve a buffer from the ring */
skb = fm10k_fetch_rx_buffer(rx_ring, rx_desc, skb);
diff --git a/drivers/net/ethernet/intel/igb/igb_main.c b/drivers/net/ethernet/intel/igb/igb_main.c
index a2d72a8..51cda23 100644
--- a/drivers/net/ethernet/intel/igb/igb_main.c
+++ b/drivers/net/ethernet/intel/igb/igb_main.c
@@ -6918,14 +6918,14 @@ static bool igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
- if (!igb_test_staterr(rx_desc, E1000_RXD_STAT_DD))
+ if (!rx_desc->wb.upper.status_error)
break;
/* This memory barrier is needed to keep us from reading
* any other fields out of the rx_desc until we know the
- * RXD_STAT_DD bit is set
+ * descriptor has been written back
*/
- rmb();
+ coherent_rmb();
/* retrieve a buffer from the ring */
skb = igb_fetch_rx_buffer(rx_ring, rx_desc, skb);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index d2df4e3..fdb49c0 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -2003,15 +2003,14 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
- if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_DD))
+ if (!rx_desc->wb.upper.status_error)
break;
- /*
- * This memory barrier is needed to keep us from reading
+ /* This memory barrier is needed to keep us from reading
* any other fields out of the rx_desc until we know the
- * RXD_STAT_DD bit is set
+ * descriptor has been written back
*/
- rmb();
+ coherent_rmb();
/* retrieve a buffer from the ring */
skb = ixgbe_fetch_rx_buffer(rx_ring, rx_desc);
^ permalink raw reply related
* [PATCH v4 2/4] arch: Add lightweight memory barriers coherent_rmb() and coherent_wmb()
From: Alexander Duyck @ 2014-11-18 17:29 UTC (permalink / raw)
To: linux-arch, netdev, linux-kernel
Cc: mathieu.desnoyers, peterz, benh, heiko.carstens, mingo, mikey,
linux, donald.c.skidmore, matthew.vick, geert, jeffrey.t.kirsher,
romieu, paulmck, nic_swsd, will.deacon, michael, tony.luck,
torvalds, oleg, schwidefsky, fweisbec, davem
In-Reply-To: <20141118172644.26303.37688.stgit@ahduyck-server>
There are a number of situations where the mandatory barriers rmb() and
wmb() are used to order memory/memory operations in the device drivers
and those barriers are much heavier than they actually need to be. For
example in the case of PowerPC wmb() calls the heavy-weight sync
instruction when for coherent memory operations all that is really needed
is an lsync or eieio instruction.
This commit adds a coherent only version of the mandatory memory barriers
rmb() and wmb(). In most cases this should result in the barrier being the
same as the SMP barriers for the SMP case, however in some cases we use a
barrier that is somewhere in between rmb() and smp_rmb(). For example on
ARM the rmb barriers break down as follows:
Barrier Call Explanation
-------------- -------- ----------------------------------
rmb() dsb() Data synchronization barrier - system
coherent_rmb() dmb(osh) data memory barrier - outer sharable
smp_rmb() dmb(ish) data memory barrier - inner sharable
These new barriers are not as safe as the standard rmb() and wmb().
Specifically they do not guarantee ordering between coherent and incoherent
memories. The primary use case for these would be to enforce ordering of
reads and writes when accessing coherent memory that is shared between the
CPU and a device.
It may also be noted that there is no coherent_mb(). Most architectures
don't provide a good mechanism for performing a coherent only full barrier
without resorting to the same mechanism used in mb(). As such there isn't
much to be gained in trying to define such a function.
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Michael Ellerman <michael@ellerman.id.au>
Cc: Michael Neuling <mikey@neuling.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
---
Documentation/memory-barriers.txt | 41 +++++++++++++++++++++++++++++++++++
arch/arm/include/asm/barrier.h | 4 +++
arch/arm64/include/asm/barrier.h | 3 +++
arch/ia64/include/asm/barrier.h | 3 +++
arch/metag/include/asm/barrier.h | 14 ++++++------
arch/mips/include/asm/barrier.h | 9 ++++----
arch/powerpc/include/asm/barrier.h | 17 +++++++++------
arch/s390/include/asm/barrier.h | 2 ++
arch/sparc/include/asm/barrier_64.h | 3 +++
arch/x86/include/asm/barrier.h | 11 ++++++---
arch/x86/um/asm/barrier.h | 13 ++++++-----
include/asm-generic/barrier.h | 8 +++++++
12 files changed, 100 insertions(+), 28 deletions(-)
diff --git a/Documentation/memory-barriers.txt b/Documentation/memory-barriers.txt
index 22a969c..d86cdc2 100644
--- a/Documentation/memory-barriers.txt
+++ b/Documentation/memory-barriers.txt
@@ -1615,6 +1615,47 @@ There are some more advanced barrier functions:
operations" subsection for information on where to use these.
+ (*) coherent_wmb();
+ (*) coherent_rmb();
+
+ These are for use with memory based device I/O to guarantee the ordering
+ of cache coherent writes or reads with respect to other writes or reads
+ to cache coherent memory.
+
+ For example, consider a device driver that shares memory with a device
+ and uses a descriptor status value to indicate if the descriptor belongs
+ to the device or the CPU, and a doorbell to notify it when new
+ descriptors are available:
+
+ if (desc->status != DEVICE_OWN) {
+ /* do not read data until we own descriptor */
+ coherent_rmb();
+
+ /* read/modify data */
+ read_data = desc->data;
+ desc->data = write_data;
+
+ /* flush modifications before status update */
+ coherent_wmb();
+
+ /* assign ownership */
+ desc->status = DEVICE_OWN;
+
+ /* force memory to sync before notifying device via MMIO */
+ wmb();
+
+ /* notify device of new descriptors */
+ writel(DESC_NOTIFY, doorbell);
+ }
+
+ The coherent_rmb() allows us guarantee the device has released ownership
+ before we read the data from the descriptor, and he coherent_wmb() allows
+ us to guarantee the data is written to the descriptor before the device
+ can see it now has ownership. The wmb() is needed to guarantee that the
+ cache coherent memory writes have completed before attempting a write to
+ the cache incoherent MMIO region.
+
+
MMIO WRITE BARRIER
------------------
diff --git a/arch/arm/include/asm/barrier.h b/arch/arm/include/asm/barrier.h
index c6a3e73..8d33c9e 100644
--- a/arch/arm/include/asm/barrier.h
+++ b/arch/arm/include/asm/barrier.h
@@ -43,10 +43,14 @@
#define mb() do { dsb(); outer_sync(); } while (0)
#define rmb() dsb()
#define wmb() do { dsb(st); outer_sync(); } while (0)
+#define coherent_rmb() dmb(osh)
+#define coherent_wmb() dmb(oshst)
#else
#define mb() barrier()
#define rmb() barrier()
#define wmb() barrier()
+#define coherent_rmb() barrier()
+#define coherent_wmb() barrier()
#endif
#ifndef CONFIG_SMP
diff --git a/arch/arm64/include/asm/barrier.h b/arch/arm64/include/asm/barrier.h
index 6389d60..d250ac4 100644
--- a/arch/arm64/include/asm/barrier.h
+++ b/arch/arm64/include/asm/barrier.h
@@ -32,6 +32,9 @@
#define rmb() dsb(ld)
#define wmb() dsb(st)
+#define coherent_rmb() dmb(oshld)
+#define coherent_wmb() dmb(oshst)
+
#ifndef CONFIG_SMP
#define smp_mb() barrier()
#define smp_rmb() barrier()
diff --git a/arch/ia64/include/asm/barrier.h b/arch/ia64/include/asm/barrier.h
index e8fffb0..ec4e652 100644
--- a/arch/ia64/include/asm/barrier.h
+++ b/arch/ia64/include/asm/barrier.h
@@ -39,6 +39,9 @@
#define rmb() mb()
#define wmb() mb()
+#define coherent_rmb() mb()
+#define coherent_wmb() mb()
+
#ifdef CONFIG_SMP
# define smp_mb() mb()
#else
diff --git a/arch/metag/include/asm/barrier.h b/arch/metag/include/asm/barrier.h
index 6d8b8c9..56999f3 100644
--- a/arch/metag/include/asm/barrier.h
+++ b/arch/metag/include/asm/barrier.h
@@ -4,8 +4,6 @@
#include <asm/metag_mem.h>
#define nop() asm volatile ("NOP")
-#define mb() wmb()
-#define rmb() barrier()
#ifdef CONFIG_METAG_META21
@@ -41,11 +39,13 @@ static inline void wr_fence(void)
#endif /* !CONFIG_METAG_META21 */
-static inline void wmb(void)
-{
- /* flush writes through the write combiner */
- wr_fence();
-}
+/* flush writes through the write combiner */
+#define mb() wr_fence()
+#define rmb() barrier()
+#define wmb() mb()
+
+#define coherent_rmb() rmb()
+#define coherent_wmb() wmb()
#ifndef CONFIG_SMP
#define fence() do { } while (0)
diff --git a/arch/mips/include/asm/barrier.h b/arch/mips/include/asm/barrier.h
index 3d69aa8..2d54df7 100644
--- a/arch/mips/include/asm/barrier.h
+++ b/arch/mips/include/asm/barrier.h
@@ -75,20 +75,21 @@
#include <asm/wbflush.h>
-#define wmb() fast_wmb()
-#define rmb() fast_rmb()
#define mb() wbflush()
#define iob() wbflush()
#else /* !CONFIG_CPU_HAS_WB */
-#define wmb() fast_wmb()
-#define rmb() fast_rmb()
#define mb() fast_mb()
#define iob() fast_iob()
#endif /* !CONFIG_CPU_HAS_WB */
+#define wmb() fast_wmb()
+#define rmb() fast_rmb()
+#define coherent_wmb() fast_wmb()
+#define coherent_rmb() fast_rmb()
+
#if defined(CONFIG_WEAK_ORDERING) && defined(CONFIG_SMP)
# ifdef CONFIG_CPU_CAVIUM_OCTEON
# define smp_mb() __sync()
diff --git a/arch/powerpc/include/asm/barrier.h b/arch/powerpc/include/asm/barrier.h
index cb6d66c..40c668b 100644
--- a/arch/powerpc/include/asm/barrier.h
+++ b/arch/powerpc/include/asm/barrier.h
@@ -36,8 +36,6 @@
#define set_mb(var, value) do { var = value; mb(); } while (0)
-#ifdef CONFIG_SMP
-
#ifdef __SUBARCH_HAS_LWSYNC
# define SMPWMB LWSYNC
#else
@@ -45,12 +43,17 @@
#endif
#define __lwsync() __asm__ __volatile__ (stringify_in_c(LWSYNC) : : :"memory")
+#define coherent_rmb() __lwsync()
+#define coherent_wmb() __asm__ __volatile__ (stringify_in_c(SMPWMB) : : :"memory")
+
+#ifdef CONFIG_SMP
+#define smp_lwsync() __lwsync()
#define smp_mb() mb()
-#define smp_rmb() __lwsync()
-#define smp_wmb() __asm__ __volatile__ (stringify_in_c(SMPWMB) : : :"memory")
+#define smp_rmb() coherent_rmb()
+#define smp_wmb() coherent_wmb()
#else
-#define __lwsync() barrier()
+#define smp_lwsync() barrier()
#define smp_mb() barrier()
#define smp_rmb() barrier()
@@ -72,7 +75,7 @@
#define smp_store_release(p, v) \
do { \
compiletime_assert_atomic_type(*p); \
- __lwsync(); \
+ smp_lwsync(); \
ACCESS_ONCE(*p) = (v); \
} while (0)
@@ -80,7 +83,7 @@ do { \
({ \
typeof(*p) ___p1 = ACCESS_ONCE(*p); \
compiletime_assert_atomic_type(*p); \
- __lwsync(); \
+ smp_lwsync(); \
___p1; \
})
diff --git a/arch/s390/include/asm/barrier.h b/arch/s390/include/asm/barrier.h
index 33d191d..a50117d 100644
--- a/arch/s390/include/asm/barrier.h
+++ b/arch/s390/include/asm/barrier.h
@@ -24,6 +24,8 @@
#define rmb() mb()
#define wmb() mb()
+#define coherent_rmb() rmb()
+#define coherent_wmb() wmb()
#define smp_mb() mb()
#define smp_rmb() rmb()
#define smp_wmb() wmb()
diff --git a/arch/sparc/include/asm/barrier_64.h b/arch/sparc/include/asm/barrier_64.h
index 6c974c0..97bdcd4 100644
--- a/arch/sparc/include/asm/barrier_64.h
+++ b/arch/sparc/include/asm/barrier_64.h
@@ -37,6 +37,9 @@ do { __asm__ __volatile__("ba,pt %%xcc, 1f\n\t" \
#define rmb() __asm__ __volatile__("":::"memory")
#define wmb() __asm__ __volatile__("":::"memory")
+#define coherent_rmb() rmb()
+#define coherent_wmb() wmb()
+
#define set_mb(__var, __value) \
do { __var = __value; membar_safe("#StoreLoad"); } while(0)
diff --git a/arch/x86/include/asm/barrier.h b/arch/x86/include/asm/barrier.h
index 5238000..340a7fe 100644
--- a/arch/x86/include/asm/barrier.h
+++ b/arch/x86/include/asm/barrier.h
@@ -24,13 +24,16 @@
#define wmb() asm volatile("sfence" ::: "memory")
#endif
-#ifdef CONFIG_SMP
-#define smp_mb() mb()
#ifdef CONFIG_X86_PPRO_FENCE
-# define smp_rmb() rmb()
+#define coherent_rmb() rmb()
#else
-# define smp_rmb() barrier()
+#define coherent_rmb() barrier()
#endif
+#define coherent_wmb() barrier()
+
+#ifdef CONFIG_SMP
+#define smp_mb() mb()
+#define smp_rmb() coherent_rmb()
#define smp_wmb() barrier()
#define set_mb(var, value) do { (void)xchg(&var, value); } while (0)
#else /* !SMP */
diff --git a/arch/x86/um/asm/barrier.h b/arch/x86/um/asm/barrier.h
index d6511d9..d4952e0 100644
--- a/arch/x86/um/asm/barrier.h
+++ b/arch/x86/um/asm/barrier.h
@@ -29,17 +29,18 @@
#endif /* CONFIG_X86_32 */
-#ifdef CONFIG_SMP
-
-#define smp_mb() mb()
#ifdef CONFIG_X86_PPRO_FENCE
-#define smp_rmb() rmb()
+#define coherent_rmb() rmb()
#else /* CONFIG_X86_PPRO_FENCE */
-#define smp_rmb() barrier()
+#define coherent_rmb() barrier()
#endif /* CONFIG_X86_PPRO_FENCE */
+#define coherent_wmb() barrier()
-#define smp_wmb() barrier()
+#ifdef CONFIG_SMP
+#define smp_mb() mb()
+#define smp_rmb() coherent_rmb()
+#define smp_wmb() coherent_wmb()
#define set_mb(var, value) do { (void)xchg(&var, value); } while (0)
#else /* CONFIG_SMP */
diff --git a/include/asm-generic/barrier.h b/include/asm-generic/barrier.h
index 1402fa8..52a9a71 100644
--- a/include/asm-generic/barrier.h
+++ b/include/asm-generic/barrier.h
@@ -42,6 +42,14 @@
#define wmb() mb()
#endif
+#ifndef coherent_rmb
+#define coherent_rmb() rmb()
+#endif
+
+#ifndef coherent_wmb
+#define coherent_wmb() wmb()
+#endif
+
#ifndef read_barrier_depends
#define read_barrier_depends() do { } while (0)
#endif
^ permalink raw reply related
* [PATCH v4 1/4] arch: Cleanup read_barrier_depends() and comments
From: Alexander Duyck @ 2014-11-18 17:29 UTC (permalink / raw)
To: linux-arch, netdev, linux-kernel
Cc: mathieu.desnoyers, peterz, benh, heiko.carstens, mingo, mikey,
linux, donald.c.skidmore, matthew.vick, geert, jeffrey.t.kirsher,
romieu, paulmck, nic_swsd, will.deacon, michael, tony.luck,
torvalds, oleg, schwidefsky, fweisbec, davem
In-Reply-To: <20141118172644.26303.37688.stgit@ahduyck-server>
This patch is meant to cleanup the handling of read_barrier_depends and
smp_read_barrier_depends. In multiple spots in the kernel headers
read_barrier_depends is defined as "do {} while (0)", however we then go
into the SMP vs non-SMP sections and have the SMP version reference
read_barrier_depends, and the non-SMP define it as yet another empty
do/while.
With this commit I went through and cleaned out the duplicate definitions
and reduced the number of definitions down to 2 per header. In addition I
moved the 50 line comments for the macro from the x86 and mips headers that
defined it as an empty do/while to those that were actually defining the
macro, alpha and blackfin.
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
---
arch/alpha/include/asm/barrier.h | 51 ++++++++++++++++++++++++++++++
arch/blackfin/include/asm/barrier.h | 51 ++++++++++++++++++++++++++++++
arch/ia64/include/asm/barrier.h | 22 +++++--------
arch/metag/include/asm/barrier.h | 7 ++--
arch/mips/include/asm/barrier.h | 52 -------------------------------
arch/powerpc/include/asm/barrier.h | 6 ++--
arch/s390/include/asm/barrier.h | 5 ++-
arch/sparc/include/asm/barrier_64.h | 4 +-
arch/x86/include/asm/barrier.h | 59 ++---------------------------------
arch/x86/um/asm/barrier.h | 7 ++--
10 files changed, 129 insertions(+), 135 deletions(-)
diff --git a/arch/alpha/include/asm/barrier.h b/arch/alpha/include/asm/barrier.h
index 3832bdb..77516c8 100644
--- a/arch/alpha/include/asm/barrier.h
+++ b/arch/alpha/include/asm/barrier.h
@@ -7,6 +7,57 @@
#define rmb() __asm__ __volatile__("mb": : :"memory")
#define wmb() __asm__ __volatile__("wmb": : :"memory")
+/**
+ * read_barrier_depends - Flush all pending reads that subsequents reads
+ * depend on.
+ *
+ * No data-dependent reads from memory-like regions are ever reordered
+ * over this barrier. All reads preceding this primitive are guaranteed
+ * to access memory (but not necessarily other CPUs' caches) before any
+ * reads following this primitive that depend on the data return by
+ * any of the preceding reads. This primitive is much lighter weight than
+ * rmb() on most CPUs, and is never heavier weight than is
+ * rmb().
+ *
+ * These ordering constraints are respected by both the local CPU
+ * and the compiler.
+ *
+ * Ordering is not guaranteed by anything other than these primitives,
+ * not even by data dependencies. See the documentation for
+ * memory_barrier() for examples and URLs to more information.
+ *
+ * For example, the following code would force ordering (the initial
+ * value of "a" is zero, "b" is one, and "p" is "&a"):
+ *
+ * <programlisting>
+ * CPU 0 CPU 1
+ *
+ * b = 2;
+ * memory_barrier();
+ * p = &b; q = p;
+ * read_barrier_depends();
+ * d = *q;
+ * </programlisting>
+ *
+ * because the read of "*q" depends on the read of "p" and these
+ * two reads are separated by a read_barrier_depends(). However,
+ * the following code, with the same initial values for "a" and "b":
+ *
+ * <programlisting>
+ * CPU 0 CPU 1
+ *
+ * a = 2;
+ * memory_barrier();
+ * b = 3; y = b;
+ * read_barrier_depends();
+ * x = a;
+ * </programlisting>
+ *
+ * does not enforce ordering, since there is no data dependency between
+ * the read of "a" and the read of "b". Therefore, on some CPUs, such
+ * as Alpha, "y" could be set to 3 and "x" to 0. Use rmb()
+ * in cases like this where there are no data dependencies.
+ */
#define read_barrier_depends() __asm__ __volatile__("mb": : :"memory")
#ifdef CONFIG_SMP
diff --git a/arch/blackfin/include/asm/barrier.h b/arch/blackfin/include/asm/barrier.h
index 4200068..dfb66fe 100644
--- a/arch/blackfin/include/asm/barrier.h
+++ b/arch/blackfin/include/asm/barrier.h
@@ -22,6 +22,57 @@
# define mb() do { barrier(); smp_check_barrier(); smp_mark_barrier(); } while (0)
# define rmb() do { barrier(); smp_check_barrier(); } while (0)
# define wmb() do { barrier(); smp_mark_barrier(); } while (0)
+/*
+ * read_barrier_depends - Flush all pending reads that subsequents reads
+ * depend on.
+ *
+ * No data-dependent reads from memory-like regions are ever reordered
+ * over this barrier. All reads preceding this primitive are guaranteed
+ * to access memory (but not necessarily other CPUs' caches) before any
+ * reads following this primitive that depend on the data return by
+ * any of the preceding reads. This primitive is much lighter weight than
+ * rmb() on most CPUs, and is never heavier weight than is
+ * rmb().
+ *
+ * These ordering constraints are respected by both the local CPU
+ * and the compiler.
+ *
+ * Ordering is not guaranteed by anything other than these primitives,
+ * not even by data dependencies. See the documentation for
+ * memory_barrier() for examples and URLs to more information.
+ *
+ * For example, the following code would force ordering (the initial
+ * value of "a" is zero, "b" is one, and "p" is "&a"):
+ *
+ * <programlisting>
+ * CPU 0 CPU 1
+ *
+ * b = 2;
+ * memory_barrier();
+ * p = &b; q = p;
+ * read_barrier_depends();
+ * d = *q;
+ * </programlisting>
+ *
+ * because the read of "*q" depends on the read of "p" and these
+ * two reads are separated by a read_barrier_depends(). However,
+ * the following code, with the same initial values for "a" and "b":
+ *
+ * <programlisting>
+ * CPU 0 CPU 1
+ *
+ * a = 2;
+ * memory_barrier();
+ * b = 3; y = b;
+ * read_barrier_depends();
+ * x = a;
+ * </programlisting>
+ *
+ * does not enforce ordering, since there is no data dependency between
+ * the read of "a" and the read of "b". Therefore, on some CPUs, such
+ * as Alpha, "y" could be set to 3 and "x" to 0. Use rmb()
+ * in cases like this where there are no data dependencies.
+ */
# define read_barrier_depends() do { barrier(); smp_check_barrier(); } while (0)
#endif
diff --git a/arch/ia64/include/asm/barrier.h b/arch/ia64/include/asm/barrier.h
index a48957c..e8fffb0 100644
--- a/arch/ia64/include/asm/barrier.h
+++ b/arch/ia64/include/asm/barrier.h
@@ -35,26 +35,22 @@
* it's (presumably) much slower than mf and (b) mf.a is supported for
* sequential memory pages only.
*/
-#define mb() ia64_mf()
-#define rmb() mb()
-#define wmb() mb()
-#define read_barrier_depends() do { } while(0)
+#define mb() ia64_mf()
+#define rmb() mb()
+#define wmb() mb()
#ifdef CONFIG_SMP
# define smp_mb() mb()
-# define smp_rmb() rmb()
-# define smp_wmb() wmb()
-# define smp_read_barrier_depends() read_barrier_depends()
-
#else
-
# define smp_mb() barrier()
-# define smp_rmb() barrier()
-# define smp_wmb() barrier()
-# define smp_read_barrier_depends() do { } while(0)
-
#endif
+#define smp_rmb() smp_mb()
+#define smp_wmb() smp_mb()
+
+#define read_barrier_depends() do { } while (0)
+#define smp_read_barrier_depends() do { } while (0)
+
#define smp_mb__before_atomic() barrier()
#define smp_mb__after_atomic() barrier()
diff --git a/arch/metag/include/asm/barrier.h b/arch/metag/include/asm/barrier.h
index c7591e8..6d8b8c9 100644
--- a/arch/metag/include/asm/barrier.h
+++ b/arch/metag/include/asm/barrier.h
@@ -47,8 +47,6 @@ static inline void wmb(void)
wr_fence();
}
-#define read_barrier_depends() do { } while (0)
-
#ifndef CONFIG_SMP
#define fence() do { } while (0)
#define smp_mb() barrier()
@@ -82,7 +80,10 @@ static inline void fence(void)
#define smp_wmb() barrier()
#endif
#endif
-#define smp_read_barrier_depends() do { } while (0)
+
+#define read_barrier_depends() do { } while (0)
+#define smp_read_barrier_depends() do { } while (0)
+
#define set_mb(var, value) do { var = value; smp_mb(); } while (0)
#define smp_store_release(p, v) \
diff --git a/arch/mips/include/asm/barrier.h b/arch/mips/include/asm/barrier.h
index d0101dd..3d69aa8 100644
--- a/arch/mips/include/asm/barrier.h
+++ b/arch/mips/include/asm/barrier.h
@@ -10,58 +10,6 @@
#include <asm/addrspace.h>
-/*
- * read_barrier_depends - Flush all pending reads that subsequents reads
- * depend on.
- *
- * No data-dependent reads from memory-like regions are ever reordered
- * over this barrier. All reads preceding this primitive are guaranteed
- * to access memory (but not necessarily other CPUs' caches) before any
- * reads following this primitive that depend on the data return by
- * any of the preceding reads. This primitive is much lighter weight than
- * rmb() on most CPUs, and is never heavier weight than is
- * rmb().
- *
- * These ordering constraints are respected by both the local CPU
- * and the compiler.
- *
- * Ordering is not guaranteed by anything other than these primitives,
- * not even by data dependencies. See the documentation for
- * memory_barrier() for examples and URLs to more information.
- *
- * For example, the following code would force ordering (the initial
- * value of "a" is zero, "b" is one, and "p" is "&a"):
- *
- * <programlisting>
- * CPU 0 CPU 1
- *
- * b = 2;
- * memory_barrier();
- * p = &b; q = p;
- * read_barrier_depends();
- * d = *q;
- * </programlisting>
- *
- * because the read of "*q" depends on the read of "p" and these
- * two reads are separated by a read_barrier_depends(). However,
- * the following code, with the same initial values for "a" and "b":
- *
- * <programlisting>
- * CPU 0 CPU 1
- *
- * a = 2;
- * memory_barrier();
- * b = 3; y = b;
- * read_barrier_depends();
- * x = a;
- * </programlisting>
- *
- * does not enforce ordering, since there is no data dependency between
- * the read of "a" and the read of "b". Therefore, on some CPUs, such
- * as Alpha, "y" could be set to 3 and "x" to 0. Use rmb()
- * in cases like this where there are no data dependencies.
- */
-
#define read_barrier_depends() do { } while(0)
#define smp_read_barrier_depends() do { } while(0)
diff --git a/arch/powerpc/include/asm/barrier.h b/arch/powerpc/include/asm/barrier.h
index bab79a1..cb6d66c 100644
--- a/arch/powerpc/include/asm/barrier.h
+++ b/arch/powerpc/include/asm/barrier.h
@@ -33,7 +33,6 @@
#define mb() __asm__ __volatile__ ("sync" : : : "memory")
#define rmb() __asm__ __volatile__ ("sync" : : : "memory")
#define wmb() __asm__ __volatile__ ("sync" : : : "memory")
-#define read_barrier_depends() do { } while(0)
#define set_mb(var, value) do { var = value; mb(); } while (0)
@@ -50,16 +49,17 @@
#define smp_mb() mb()
#define smp_rmb() __lwsync()
#define smp_wmb() __asm__ __volatile__ (stringify_in_c(SMPWMB) : : :"memory")
-#define smp_read_barrier_depends() read_barrier_depends()
#else
#define __lwsync() barrier()
#define smp_mb() barrier()
#define smp_rmb() barrier()
#define smp_wmb() barrier()
-#define smp_read_barrier_depends() do { } while(0)
#endif /* CONFIG_SMP */
+#define read_barrier_depends() do { } while (0)
+#define smp_read_barrier_depends() do { } while (0)
+
/*
* This is a barrier which prevents following instructions from being
* started until the value of the argument x is known. For example, if
diff --git a/arch/s390/include/asm/barrier.h b/arch/s390/include/asm/barrier.h
index b5dce65..33d191d 100644
--- a/arch/s390/include/asm/barrier.h
+++ b/arch/s390/include/asm/barrier.h
@@ -24,11 +24,12 @@
#define rmb() mb()
#define wmb() mb()
-#define read_barrier_depends() do { } while(0)
#define smp_mb() mb()
#define smp_rmb() rmb()
#define smp_wmb() wmb()
-#define smp_read_barrier_depends() read_barrier_depends()
+
+#define read_barrier_depends() do { } while (0)
+#define smp_read_barrier_depends() do { } while (0)
#define smp_mb__before_atomic() smp_mb()
#define smp_mb__after_atomic() smp_mb()
diff --git a/arch/sparc/include/asm/barrier_64.h b/arch/sparc/include/asm/barrier_64.h
index 305dcc3..6c974c0 100644
--- a/arch/sparc/include/asm/barrier_64.h
+++ b/arch/sparc/include/asm/barrier_64.h
@@ -37,7 +37,6 @@ do { __asm__ __volatile__("ba,pt %%xcc, 1f\n\t" \
#define rmb() __asm__ __volatile__("":::"memory")
#define wmb() __asm__ __volatile__("":::"memory")
-#define read_barrier_depends() do { } while(0)
#define set_mb(__var, __value) \
do { __var = __value; membar_safe("#StoreLoad"); } while(0)
@@ -51,7 +50,8 @@ do { __asm__ __volatile__("ba,pt %%xcc, 1f\n\t" \
#define smp_wmb() __asm__ __volatile__("":::"memory")
#endif
-#define smp_read_barrier_depends() do { } while(0)
+#define read_barrier_depends() do { } while (0)
+#define smp_read_barrier_depends() do { } while (0)
#define smp_store_release(p, v) \
do { \
diff --git a/arch/x86/include/asm/barrier.h b/arch/x86/include/asm/barrier.h
index 0f4460b..5238000 100644
--- a/arch/x86/include/asm/barrier.h
+++ b/arch/x86/include/asm/barrier.h
@@ -24,60 +24,6 @@
#define wmb() asm volatile("sfence" ::: "memory")
#endif
-/**
- * read_barrier_depends - Flush all pending reads that subsequents reads
- * depend on.
- *
- * No data-dependent reads from memory-like regions are ever reordered
- * over this barrier. All reads preceding this primitive are guaranteed
- * to access memory (but not necessarily other CPUs' caches) before any
- * reads following this primitive that depend on the data return by
- * any of the preceding reads. This primitive is much lighter weight than
- * rmb() on most CPUs, and is never heavier weight than is
- * rmb().
- *
- * These ordering constraints are respected by both the local CPU
- * and the compiler.
- *
- * Ordering is not guaranteed by anything other than these primitives,
- * not even by data dependencies. See the documentation for
- * memory_barrier() for examples and URLs to more information.
- *
- * For example, the following code would force ordering (the initial
- * value of "a" is zero, "b" is one, and "p" is "&a"):
- *
- * <programlisting>
- * CPU 0 CPU 1
- *
- * b = 2;
- * memory_barrier();
- * p = &b; q = p;
- * read_barrier_depends();
- * d = *q;
- * </programlisting>
- *
- * because the read of "*q" depends on the read of "p" and these
- * two reads are separated by a read_barrier_depends(). However,
- * the following code, with the same initial values for "a" and "b":
- *
- * <programlisting>
- * CPU 0 CPU 1
- *
- * a = 2;
- * memory_barrier();
- * b = 3; y = b;
- * read_barrier_depends();
- * x = a;
- * </programlisting>
- *
- * does not enforce ordering, since there is no data dependency between
- * the read of "a" and the read of "b". Therefore, on some CPUs, such
- * as Alpha, "y" could be set to 3 and "x" to 0. Use rmb()
- * in cases like this where there are no data dependencies.
- **/
-
-#define read_barrier_depends() do { } while (0)
-
#ifdef CONFIG_SMP
#define smp_mb() mb()
#ifdef CONFIG_X86_PPRO_FENCE
@@ -86,16 +32,17 @@
# define smp_rmb() barrier()
#endif
#define smp_wmb() barrier()
-#define smp_read_barrier_depends() read_barrier_depends()
#define set_mb(var, value) do { (void)xchg(&var, value); } while (0)
#else /* !SMP */
#define smp_mb() barrier()
#define smp_rmb() barrier()
#define smp_wmb() barrier()
-#define smp_read_barrier_depends() do { } while (0)
#define set_mb(var, value) do { var = value; barrier(); } while (0)
#endif /* SMP */
+#define read_barrier_depends() do { } while (0)
+#define smp_read_barrier_depends() do { } while (0)
+
#if defined(CONFIG_X86_PPRO_FENCE)
/*
diff --git a/arch/x86/um/asm/barrier.h b/arch/x86/um/asm/barrier.h
index cc04e67..d6511d9 100644
--- a/arch/x86/um/asm/barrier.h
+++ b/arch/x86/um/asm/barrier.h
@@ -29,8 +29,6 @@
#endif /* CONFIG_X86_32 */
-#define read_barrier_depends() do { } while (0)
-
#ifdef CONFIG_SMP
#define smp_mb() mb()
@@ -42,7 +40,6 @@
#define smp_wmb() barrier()
-#define smp_read_barrier_depends() read_barrier_depends()
#define set_mb(var, value) do { (void)xchg(&var, value); } while (0)
#else /* CONFIG_SMP */
@@ -50,11 +47,13 @@
#define smp_mb() barrier()
#define smp_rmb() barrier()
#define smp_wmb() barrier()
-#define smp_read_barrier_depends() do { } while (0)
#define set_mb(var, value) do { var = value; barrier(); } while (0)
#endif /* CONFIG_SMP */
+#define read_barrier_depends() do { } while (0)
+#define smp_read_barrier_depends() do { } while (0)
+
/*
* Stop RDTSC speculation. This is needed when you need to use RDTSC
* (or get_cycles or vread that possibly accesses the TSC) in a defined
^ permalink raw reply related
* [PATCH v4 0/4] Add lightweight memory barriers for coherent memory access
From: Alexander Duyck @ 2014-11-18 17:28 UTC (permalink / raw)
To: linux-arch, netdev, linux-kernel
Cc: mathieu.desnoyers, peterz, benh, heiko.carstens, mingo, mikey,
linux, donald.c.skidmore, matthew.vick, geert, jeffrey.t.kirsher,
romieu, paulmck, nic_swsd, will.deacon, michael, tony.luck,
torvalds, oleg, schwidefsky, fweisbec, davem
These patches introduce two new primitives for synchronizing cache coherent
memory writes and reads. These two new primitives are:
coherent_rmb()
coherent_wmb()
The first patch cleans up some unnecessary overhead related to the
definition of read_barrier_depends, smp_read_barrier_depends, and comments
related to the barrier.
The second patch adds the primitives for the applicable architectures and
asm-generic.
The third patch adds the barriers to r8169 which turns out to be a good
example of where the new barriers might be useful as they have full
rmb()/wmb() barriers ordering accesses to the descriptors and the DescOwn
bit.
The fourth patch adds support for coherent_rmb() to the Intel fm10k, igb,
and ixgbe drivers. Testing with the ixgbe driver has shown a processing
time reduction of at least 7ns per 64B frame on a Core i7-4930K.
This patch series is essentially the v4 for:
v3: Add lightweight memory barriers fast_rmb() and fast_wmb()
v2: Introduce load_acquire() and store_release()
v1: Introduce read_acquire()
The key changes in this patch series versus the earlier patches are:
v4:
- Updated ARM barrier domains to outer shareable
- Renamed barriers coherent_rmb and coherent_wmb
- Added smp_lwsync for use in smp_load_acquire/smp_store_release
v3:
- Moved away from acquire()/store() and instead focused on barriers
- Added cleanup of read_barrier_depends
- Added change in r8169 to fix cur_tx/DescOwn ordering
- Simplified changes to just replacing/moving barriers in r8169
- Added update to documentation with code example
v2:
- Renamed read_acquire() to be consistent with smp_load_acquire()
- Changed barrier used to be consistent with smp_load_acquire()
- Updated PowerPC code to use __lwsync based on IBM article
- Added store_release() as this is a viable use case for drivers
- Added r8169 patch which is able to fully use primitives
- Added fm10k/igb/ixgbe patch which is able to test performance
---
Alexander Duyck (4):
arch: Cleanup read_barrier_depends() and comments
arch: Add lightweight memory barriers coherent_rmb() and coherent_wmb()
r8169: Use coherent_rmb() and coherent_wmb() for DescOwn checks
fm10k/igb/ixgbe: Use coherent_rmb on Rx descriptor reads
Documentation/memory-barriers.txt | 41 +++++++++++++++
arch/alpha/include/asm/barrier.h | 51 ++++++++++++++++++
arch/arm/include/asm/barrier.h | 4 +
arch/arm64/include/asm/barrier.h | 3 +
arch/blackfin/include/asm/barrier.h | 51 ++++++++++++++++++
arch/ia64/include/asm/barrier.h | 25 ++++-----
arch/metag/include/asm/barrier.h | 19 ++++---
arch/mips/include/asm/barrier.h | 61 ++--------------------
arch/powerpc/include/asm/barrier.h | 23 +++++---
arch/s390/include/asm/barrier.h | 7 ++-
arch/sparc/include/asm/barrier_64.h | 7 ++-
arch/x86/include/asm/barrier.h | 70 ++++---------------------
arch/x86/um/asm/barrier.h | 20 ++++---
drivers/net/ethernet/intel/fm10k/fm10k_main.c | 6 +-
drivers/net/ethernet/intel/igb/igb_main.c | 6 +-
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 9 +--
drivers/net/ethernet/realtek/r8169.c | 29 ++++++++--
include/asm-generic/barrier.h | 8 +++
18 files changed, 259 insertions(+), 181 deletions(-)
--
^ permalink raw reply
* Re: [PATCH] [net]ip_tunnel: the lack of vti_link_ops' dellink() cause kernel panic
From: Nicolas Dichtel @ 2014-11-18 17:23 UTC (permalink / raw)
To: Xin Long, network dev; +Cc: Steffen Klassert
In-Reply-To: <1416307059-32732-1-git-send-email-lucien.xin@gmail.com>
Le 18/11/2014 11:37, Xin Long a écrit :
> Now the vti_link_ops do not point the .dellink, rtnl_newlink will invoke
> the default function, unregister_netdevice_queue, which will cause the
> dev to unregister later. so when we delete a vti device, the net_device
> will be removed, but the tunnel still in the tunnel list. then, we add
> a new vti, in ip_tunnel_find():
>
> hlist_for_each_entry_rcu(t, head, hash_node) {
> if (local == t->parms.iph.saddr &&
> remote == t->parms.iph.daddr &&
> link == t->parms.link &&
> type == t->dev->type &&
> ip_tunnel_key_match(&t->parms, flags, key))
> break;
> }
>
> the dev of ip_tunnel *t may be null because of unregister_netdevice_queue
> motioned above. so the panic will happen:
>
> [ 3835.072977] IP: [<ffffffffa04103fd>] ip_tunnel_find+0x9d/0xc0 [ip_tunnel]
> [ 3835.073008] PGD b2c21067 PUD b7277067 PMD 0
> [ 3835.073008] Oops: 0000 [#1] SMP
> .....
> [ 3835.073008] Stack:
> [ 3835.073008] ffff8800b72d77f0 ffffffffa0411924 ffff8800bb956000 ffff8800b72d78e0
> [ 3835.073008] ffff8800b72d78a0 0000000000000000 ffffffffa040d100 ffff8800b72d7858
> [ 3835.073008] ffffffffa040b2e3 0000000000000000 0000000000000000 0000000000000000
> [ 3835.073008] Call Trace:
> [ 3835.073008] [<ffffffffa0411924>] ip_tunnel_newlink+0x64/0x160 [ip_tunnel]
> [ 3835.073008] [<ffffffffa040b2e3>] vti_newlink+0x43/0x70 [ip_vti]
> [ 3835.073008] [<ffffffff8150d4da>] rtnl_newlink+0x4fa/0x5f0
> [ 3835.073008] [<ffffffff812f68bb>] ? nla_strlcpy+0x5b/0x70
> [ 3835.073008] [<ffffffff81508fb0>] ? rtnl_link_ops_get+0x40/0x60
> [ 3835.073008] [<ffffffff8150d11f>] ? rtnl_newlink+0x13f/0x5f0
> [ 3835.073008] [<ffffffff81509cf4>] rtnetlink_rcv_msg+0xa4/0x270
> [ 3835.073008] [<ffffffff8126adf5>] ? sock_has_perm+0x75/0x90
> [ 3835.073008] [<ffffffff81509c50>] ? rtnetlink_rcv+0x30/0x30
> [ 3835.073008] [<ffffffff81529e39>] netlink_rcv_skb+0xa9/0xc0
> [ 3835.073008] [<ffffffff81509c48>] rtnetlink_rcv+0x28/0x30
> ....
>
> the reproduction can be like this:
>
> modprobe ip_vti
> ip link del ip_vti0 type vti
> ip link add ip_vti0 type vti
> rmmod ip_vti
>
> do that one or more time, kernel will panic.
>
> Signed-off-by: Xin Long <lucien.xin@gmail.com>
A quick look at the ipv6 side seems to show that there is the same problem. Can
you provide the IPv6 patch too?
Note also that the maintainer of this module is Steffen Klassert, please don't
forget to CC him.
> ---
> net/ipv4/ip_vti.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/net/ipv4/ip_vti.c b/net/ipv4/ip_vti.c
> index 3e86101..1a7e979 100644
> --- a/net/ipv4/ip_vti.c
> +++ b/net/ipv4/ip_vti.c
> @@ -528,6 +528,7 @@ static struct rtnl_link_ops vti_link_ops __read_mostly = {
> .validate = vti_tunnel_validate,
> .newlink = vti_newlink,
> .changelink = vti_changelink,
> + .dellink = ip_tunnel_dellink,
Nitpicking: other lines into this struct uses tabs to align the '=', but
the one you add uses spaces.
Thank you,
Nicolas
^ permalink raw reply
* Re: [PATCH net-next] tun: return NET_XMIT_DROP for dropped packets
From: Amos Kong @ 2014-11-18 16:53 UTC (permalink / raw)
To: Jason Wang; +Cc: davem, netdev, linux-kernel, Michael S. Tsirkin
In-Reply-To: <1416288041-30921-1-git-send-email-jasowang@redhat.com>
[-- Attachment #1: Type: text/plain, Size: 1416 bytes --]
On Tue, Nov 18, 2014 at 01:20:41PM +0800, Jason Wang wrote:
> After commit 5d097109257c03a71845729f8db6b5770c4bbedc
> ("tun: only queue packets on device"), NETDEV_TX_OK was returned for
> dropped packets. This will confuse pktgen since dropped packets were
> counted as sent ones.
>
> Fixing this by returning NET_XMIT_DROP to let pktgen count it as error
> packet.
>
> Cc: Michael S. Tsirkin <mst@redhat.com>
> Signed-off-by: Jason Wang <jasowang@redhat.com>
> ---
> drivers/net/tun.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index e3fa65a..ac53a73 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -819,7 +819,7 @@ drop:
> skb_tx_error(skb);
> kfree_skb(skb);
> rcu_read_unlock();
> - return NETDEV_TX_OK;
> + return NET_XMIT_DROP;
Quoted from linux/drivers/firewire/net.c:
/*
* FIXME: According to a patch from 2003-02-26, "returning non-zero
* causes serious problems" here, allegedly. Before that patch,
* -ERRNO was returned which is not appropriate under Linux 2.6.
* Perhaps more needs to be done? Stop the queue in serious
* conditions and restart it elsewhere?
*/
I saw many drivers return NETDEV_TX_OK in xmit for drop packets, eg: virtio_net.c
> }
>
> static void tun_net_mclist(struct net_device *dev)
> --
> 1.9.1
--
Amos.
[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply
* Re: [PATCH 2/4] arch: Add lightweight memory barriers fast_rmb() and fast_wmb()
From: Will Deacon @ 2014-11-18 16:48 UTC (permalink / raw)
To: Alexander Duyck
Cc: Benjamin Herrenschmidt, Alexander Duyck,
linux-arch@vger.kernel.org, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org, mathieu.desnoyers@polymtl.ca,
peterz@infradead.org, heiko.carstens@de.ibm.com, mingo@kernel.org,
mikey@neuling.org, linux@arm.linux.org.uk,
donald.c.skidmore@intel.com, matthew.vick@intel.com,
geert@linux-m68k.org, jeffrey.t.kirsher@intel.com,
romieu@fr.zoreil.com, paulmck@linux.vnet.ibm.com
In-Reply-To: <546B71DE.4050506@redhat.com>
On Tue, Nov 18, 2014 at 04:20:46PM +0000, Alexander Duyck wrote:
> On 11/18/2014 03:58 AM, Will Deacon wrote:
> > So actually, this is an interesting case where the barrier would like to
> > know whether the memory returned by dma_alloc_coherent is h/w coherent
> > (normal, cacheable) or s/w coherent (normal, non-cacheable). I think Ben
> > is thinking of the h/w coherent case (i.e. actual snooping into the CPU
> > caches by the DMA master).
> >
> > For the former, we could use inner-shareable barriers. For the latter, we'd
> > need to use outer-shareable barriers.
> >
> > If we can't tell, then these should be dmb(osh), which will work for both.
> >
>
> Okay, so I will update the ARM portion of my patches to use osh and
> oshst then since it sounds like I was using too strong of barriers.
Sounds good. Another reason this is interesting is because the native
acquire/release instructions on ARMv8 actually take into account the
shareability domain of the virtual address, so using them would give you
the shareability domain you want but slightly stronger ordering guarantees
within that domain.
Still, either of them will be a damn sight better than the dsb we currently
have courtesy of the mandatory barriers.
Will
^ permalink raw reply
* Re: [PATCH 2/2] net: can: comparison of unsigned variable
From: Michal Simek @ 2014-11-18 16:47 UTC (permalink / raw)
To: Sudip Mukherjee, Wolfgang Grandegger, Marc Kleine-Budde,
Michal Simek, Sören Brinkmann
Cc: netdev, linux-kernel, linux-arm-kernel, linux-can
In-Reply-To: <1416318427-28676-2-git-send-email-sudipm.mukherjee@gmail.com>
On 11/18/2014 02:47 PM, Sudip Mukherjee wrote:
> err was of the type u32. it was being compared with < 0, and being
> an unsigned variable the comparison would have been always false.
>
> moreover, err was getting the return value from set_reset_mode()
> and xcan_set_bittiming(), and both are returning int.
>
> Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> ---
> drivers/net/can/xilinx_can.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/can/xilinx_can.c b/drivers/net/can/xilinx_can.c
> index 72fe96f..67c2dd4 100644
> --- a/drivers/net/can/xilinx_can.c
> +++ b/drivers/net/can/xilinx_can.c
> @@ -300,7 +300,8 @@ static int xcan_set_bittiming(struct net_device *ndev)
> static int xcan_chip_start(struct net_device *ndev)
> {
> struct xcan_priv *priv = netdev_priv(ndev);
> - u32 err, reg_msr, reg_sr_mask;
> + u32 reg_msr, reg_sr_mask;
> + int err;
> unsigned long timeout;
>
> /* Check if it is in reset mode */
>
Reviewed-by: Michal Simek <michal.simek@xilinx.com>
Thanks,
Michal
^ permalink raw reply
* Re: [PATCH 2/4] arch: Add lightweight memory barriers fast_rmb() and fast_wmb()
From: Alexander Duyck @ 2014-11-18 16:20 UTC (permalink / raw)
To: Will Deacon
Cc: Benjamin Herrenschmidt, Alexander Duyck,
linux-arch@vger.kernel.org, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org, mathieu.desnoyers@polymtl.ca,
peterz@infradead.org, heiko.carstens@de.ibm.com, mingo@kernel.org,
mikey@neuling.org, linux@arm.linux.org.uk,
donald.c.skidmore@intel.com, matthew.vick@intel.com,
geert@linux-m68k.org, jeffrey.t.kirsher@intel.com,
romieu@fr.zoreil.com, paulmck@linux.vnet.ibm.com
In-Reply-To: <20141118115836.GL18842@arm.com>
On 11/18/2014 03:58 AM, Will Deacon wrote:
> On Tue, Nov 18, 2014 at 03:13:29AM +0000, Alexander Duyck wrote:
>> On 11/17/2014 04:39 PM, Benjamin Herrenschmidt wrote:
>>> On Mon, 2014-11-17 at 12:24 -0800, Alexander Duyck wrote:
>>>> Yes and no. So for example on ARM I used the dmb() operation, however
>>>> I
>>>> have to use the barrier at the system level instead of just the inner
>>>> shared domain. However on many other architectures they are just the
>>>> same as the smp_* variants.
>>>>
>>>> Basically the resultant code is somewhere between the smp and non-smp
>>>> barriers in terms of what they cover.
>>> There I don't quite follow you. You need to explain better especially in
>>> the documentation because otherwise people will get it wrong...
>>>
>>> If it's ordering in the coherent domain, I fail to see how a DMA agent
>>> is different than another processor when it comes to barriers, so I fail
>>> to see the difference with smp_*
>>>
>>> I understand the MMIO vs. memory issue, we do have the same on powerpc,
>>> but that other aspect eludes me.
>>>
>> ARM adds some funky things. They have two different types of
>> primitives, a dmb() which is a data memory barrier, and a dsb() which is
>> a data synchronization barrier. Then with each of those they have the
>> "domains" the barriers are effective within.
>>
>> So for example on ARM a rmb() is dsb(sy) which means it is a system wide
>> synchronization barrier which stops execution on the CPU core until the
>> read completes. However the smp_rmb() is a dmb(ish) which means it is
>> only a barrier as far as the inner shareable domain which I believe only
>> goes as far as the local shared cache hierarchy and only guarantees read
>> ordering without necessarily halting the CPU or stopping in-order
>> speculative reads. So what a coherent_rmb() would be in my setup is
>> dmb(sy) which means the barrier runs all the way out to memory, and it
>> is allowed to speculative read as long as it does it in order.
>>
>> If it is still unclear you might check out Will Deacon's talk on the
>> topic at https://www.youtube.com/watch?v=6ORn6_35kKo, at about 7:00 in
>> he explains the whole domains thing, and at 13:30 he explains dmb()/dsb().
> So actually, this is an interesting case where the barrier would like to
> know whether the memory returned by dma_alloc_coherent is h/w coherent
> (normal, cacheable) or s/w coherent (normal, non-cacheable). I think Ben
> is thinking of the h/w coherent case (i.e. actual snooping into the CPU
> caches by the DMA master).
>
> For the former, we could use inner-shareable barriers. For the latter, we'd
> need to use outer-shareable barriers.
>
> If we can't tell, then these should be dmb(osh), which will work for both.
>
> Will
Okay, so I will update the ARM portion of my patches to use osh and
oshst then since it sounds like I was using too strong of barriers.
- Alex
^ permalink raw reply
* [PATCH v2] usbnet: rtl8150: remove unused variable
From: Sudip Mukherjee @ 2014-11-18 16:25 UTC (permalink / raw)
To: Petko Manolov; +Cc: Sudip Mukherjee, linux-usb, netdev, linux-kernel
remove unused variable
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
---
change in v2: changed the commit message
drivers/net/usb/rtl8150.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/net/usb/rtl8150.c b/drivers/net/usb/rtl8150.c
index 6e87e57..d37b7dc 100644
--- a/drivers/net/usb/rtl8150.c
+++ b/drivers/net/usb/rtl8150.c
@@ -753,14 +753,13 @@ static int rtl8150_open(struct net_device *netdev)
static int rtl8150_close(struct net_device *netdev)
{
rtl8150_t *dev = netdev_priv(netdev);
- int res = 0;
netif_stop_queue(netdev);
if (!test_bit(RTL8150_UNPLUG, &dev->flags))
disable_net_traffic(dev);
unlink_all_urbs(dev);
- return res;
+ return 0;
}
static void rtl8150_get_drvinfo(struct net_device *netdev, struct ethtool_drvinfo *info)
--
1.8.1.2
^ permalink raw reply related
* [PATCH net-next 4/4] tuntap: Increase the number of queues in tun.
From: Pankaj Gupta @ 2014-11-18 16:22 UTC (permalink / raw)
To: linux-kernel, netdev
Cc: davem, jasowang, mst, dgibson, vfalico, edumazet, vyasevic, hkchu,
wuzhy, xemul, therbert, bhutchings, xii, stephen, jiri,
sergei.shtylyov, Pankaj Gupta
In-Reply-To: <1416327778-17716-1-git-send-email-pagupta@redhat.com>
Networking under kvm works best if we allocate a per-vCPU RX and TX
queue in a virtual NIC. This requires a per-vCPU queue on the host side.
It is now safe to increase the maximum number of queues.
Preceding patches:
net: allow large number of rx queues
tuntap: Reduce the size of tun_struct by using flex array
tuntap: Accepts tuntap max queue length as net sysctl parameter
made sure this won't cause failures due to high order memory
allocations. Increase it to 256: this is the max number of vCPUs
KVM supports.
Signed-off-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: David Gibson <dgibson@redhat.com>
---
drivers/net/tun.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index a9b3eb4..2fb31b7 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -115,10 +115,11 @@ struct tap_filter {
unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
};
-/* DEFAULT_MAX_NUM_RSS_QUEUES were chosen to let the rx/tx queues allocated for
- * the netdevice to be fit in one page. So we can make sure the success of
- * memory allocation. TODO: increase the limit. */
-#define MAX_TAP_QUEUES DEFAULT_MAX_NUM_RSS_QUEUES
+/* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal
+ * to max number of vCPUS in guest. Also, we are making sure here
+ * queue memory allocation do not fail.
+ */
+#define MAX_TAP_QUEUES 256
#define MIN_TAP_QUEUES 1
#define MAX_TAP_FLOWS 4096
--
1.8.3.1
^ permalink raw reply related
* [PATCH net-next 3/4] tuntap: reduce the size of tun_struct by using flex array.
From: Pankaj Gupta @ 2014-11-18 16:22 UTC (permalink / raw)
To: linux-kernel, netdev
Cc: davem, jasowang, mst, dgibson, vfalico, edumazet, vyasevic, hkchu,
wuzhy, xemul, therbert, bhutchings, xii, stephen, jiri,
sergei.shtylyov, Pankaj Gupta
In-Reply-To: <1416327778-17716-1-git-send-email-pagupta@redhat.com>
This patch switches to flex array to implement the flow caches, it brings
several advantages:
- Reduce the size of the tun_struct structure, which allows us to increase the
upper limit of queues in future.
- Avoid higher order memory allocation. It will be useful when switching to
pure hashing in flow cache which may demand a larger size array in future.
After this patch, the size of tun_struct on x86_64 reduced from 8512 to
328
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: David Gibson <dgibson@redhat.com>
---
drivers/net/tun.c | 49 +++++++++++++++++++++++++++++++++++++------------
1 file changed, 37 insertions(+), 12 deletions(-)
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index e3fa65a..bd07a6d 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -65,6 +65,7 @@
#include <linux/nsproxy.h>
#include <linux/virtio_net.h>
#include <linux/rcupdate.h>
+#include <linux/flex_array.h>
#include <net/ipv6.h>
#include <net/net_namespace.h>
#include <net/netns/generic.h>
@@ -188,7 +189,7 @@ struct tun_struct {
int debug;
#endif
spinlock_t lock;
- struct hlist_head flows[TUN_NUM_FLOW_ENTRIES];
+ struct flex_array *flows;
struct timer_list flow_gc_timer;
unsigned long ageing_time;
unsigned int numdisabled;
@@ -249,10 +250,11 @@ static void tun_flow_flush(struct tun_struct *tun)
spin_lock_bh(&tun->lock);
for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
+ struct hlist_head *h = flex_array_get(tun->flows, i);
struct tun_flow_entry *e;
struct hlist_node *n;
- hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link)
+ hlist_for_each_entry_safe(e, n, h, hash_link)
tun_flow_delete(tun, e);
}
spin_unlock_bh(&tun->lock);
@@ -264,10 +266,11 @@ static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index)
spin_lock_bh(&tun->lock);
for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
+ struct hlist_head *h = flex_array_get(tun->flows, i);
struct tun_flow_entry *e;
struct hlist_node *n;
- hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
+ hlist_for_each_entry_safe(e, n, h, hash_link) {
if (e->queue_index == queue_index)
tun_flow_delete(tun, e);
}
@@ -287,10 +290,11 @@ static void tun_flow_cleanup(unsigned long data)
spin_lock_bh(&tun->lock);
for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
+ struct hlist_head *h = flex_array_get(tun->flows, i);
struct tun_flow_entry *e;
struct hlist_node *n;
- hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
+ hlist_for_each_entry_safe(e, n, h, hash_link) {
unsigned long this_timer;
count++;
this_timer = e->updated + delay;
@@ -317,7 +321,7 @@ static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
if (!rxhash)
return;
else
- head = &tun->flows[tun_hashfn(rxhash)];
+ head = flex_array_get(tun->flows, tun_hashfn(rxhash));
rcu_read_lock();
@@ -380,7 +384,8 @@ static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb,
txq = skb_get_hash(skb);
if (txq) {
- e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq);
+ e = tun_flow_find(flex_array_get(tun->flows,
+ tun_hashfn(txq)), txq);
if (e) {
tun_flow_save_rps_rxhash(e, txq);
txq = e->queue_index;
@@ -760,8 +765,8 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
rxhash = skb_get_hash(skb);
if (rxhash) {
struct tun_flow_entry *e;
- e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)],
- rxhash);
+ e = tun_flow_find(flex_array_get(tun->flows,
+ tun_hashfn(rxhash)), rxhash);
if (e)
tun_flow_save_rps_rxhash(e, rxhash);
}
@@ -896,23 +901,40 @@ static const struct net_device_ops tap_netdev_ops = {
#endif
};
-static void tun_flow_init(struct tun_struct *tun)
+static int tun_flow_init(struct tun_struct *tun)
{
- int i;
+ struct flex_array *buckets;
+ int i, err;
+
+ buckets = flex_array_alloc(sizeof(struct hlist_head),
+ TUN_NUM_FLOW_ENTRIES, GFP_KERNEL);
+ if (!buckets)
+ return -ENOMEM;
+
+ err = flex_array_prealloc(buckets, 0, TUN_NUM_FLOW_ENTRIES, GFP_KERNEL);
+ if (err) {
+ flex_array_free(buckets);
+ return -ENOMEM;
+ }
+ tun->flows = buckets;
for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++)
- INIT_HLIST_HEAD(&tun->flows[i]);
+ INIT_HLIST_HEAD((struct hlist_head *)
+ flex_array_get(buckets, i));
tun->ageing_time = TUN_FLOW_EXPIRE;
setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun);
mod_timer(&tun->flow_gc_timer,
round_jiffies_up(jiffies + tun->ageing_time));
+
+ return 0;
}
static void tun_flow_uninit(struct tun_struct *tun)
{
del_timer_sync(&tun->flow_gc_timer);
tun_flow_flush(tun);
+ flex_array_free(tun->flows);
}
/* Initialize net device. */
@@ -1674,7 +1696,10 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
goto err_free_dev;
tun_net_init(dev);
- tun_flow_init(tun);
+
+ err = tun_flow_init(tun);
+ if (err < 0)
+ goto err_free_dev;
dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
--
1.8.3.1
^ permalink raw reply related
* [PATCH net-next 2/4] tuntap: Accept tuntap maximum number of queues as sysctl
From: Pankaj Gupta @ 2014-11-18 16:22 UTC (permalink / raw)
To: linux-kernel, netdev
Cc: davem, jasowang, mst, dgibson, vfalico, edumazet, vyasevic, hkchu,
wuzhy, xemul, therbert, bhutchings, xii, stephen, jiri,
sergei.shtylyov, Pankaj Gupta
In-Reply-To: <1416327778-17716-1-git-send-email-pagupta@redhat.com>
This patch accepts maximum number of tun/tap queues allocated as
sysctl entry which a user space application like libvirt
can make use of to limit maximum number of tuntap queues.
Value of sysctl entry is writable dynamically.
If no value is set for sysctl entry 'net.tuntap.max_queues'
a default value 256 is used which is equal to maximum number
of vCPUS allowed by KVM.
Signed-off-by: Pankaj Gupta <pagupta@redhat.com>
---
drivers/net/tun.c | 33 +++++++++++++++++++++++++++++++--
1 file changed, 31 insertions(+), 2 deletions(-)
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index e3fa65a..b03a745 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -71,6 +71,7 @@
#include <net/rtnetlink.h>
#include <net/sock.h>
#include <linux/seq_file.h>
+#include <linux/sysctl.h>
#include <linux/uio.h>
#include <asm/uaccess.h>
@@ -117,10 +118,16 @@ struct tap_filter {
* the netdevice to be fit in one page. So we can make sure the success of
* memory allocation. TODO: increase the limit. */
#define MAX_TAP_QUEUES DEFAULT_MAX_NUM_RSS_QUEUES
+#define MIN_TAP_QUEUES 1
#define MAX_TAP_FLOWS 4096
#define TUN_FLOW_EXPIRE (3 * HZ)
+static struct ctl_table_header *tun_sysctl_header;
+static int tun_queues = MAX_TAP_QUEUES;
+static int min_queues = MIN_TAP_QUEUES;
+static int max_queues = MAX_TAP_QUEUES;
+
/* A tun_file connects an open character device to a tuntap netdevice. It
* also contains all socket related structures (except sock_fprog and tap_filter)
* to serve as one transmit queue for tuntap device. The sock_fprog and
@@ -197,6 +204,19 @@ struct tun_struct {
u32 flow_count;
};
+static struct ctl_table tun_ctl_table[] = {
+ {
+ .procname = "tun_max_queues",
+ .data = &tun_queues,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = proc_dointvec_minmax,
+ .extra1 = &min_queues,
+ .extra2 = &max_queues
+ },
+ { }
+};
+
static inline u32 tun_hashfn(u32 rxhash)
{
return rxhash & 0x3ff;
@@ -547,7 +567,7 @@ static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filte
err = -E2BIG;
if (!tfile->detached &&
- tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES)
+ tun->numqueues + tun->numdisabled == tun_queues)
goto out;
err = 0;
@@ -1624,7 +1644,7 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
char *name;
unsigned long flags = 0;
int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ?
- MAX_TAP_QUEUES : 1;
+ tun_queues : 1;
if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
return -EPERM;
@@ -2335,6 +2355,13 @@ static int __init tun_init(void)
pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
pr_info("%s\n", DRV_COPYRIGHT);
+ tun_sysctl_header = register_net_sysctl(&init_net, "net/tuntap",
+ tun_ctl_table);
+
+ if (!tun_sysctl_header)
+ pr_err("Can't register tun_ctl_table. Tun device queue"
+ "setting to default value : %d queues.\n", tun_queues);
+
ret = rtnl_link_register(&tun_link_ops);
if (ret) {
pr_err("Can't register link_ops\n");
@@ -2357,6 +2384,8 @@ static void tun_cleanup(void)
{
misc_deregister(&tun_miscdev);
rtnl_link_unregister(&tun_link_ops);
+ if (tun_sysctl_header)
+ unregister_net_sysctl_table(tun_sysctl_header);
}
/* Get an underlying socket object from tun file. Returns error unless file is
--
1.8.3.1
^ permalink raw reply related
* [PATCH net-net 0/4] Increase the limit of tuntap queues
From: Pankaj Gupta @ 2014-11-18 16:22 UTC (permalink / raw)
To: linux-kernel, netdev
Cc: davem, jasowang, mst, dgibson, vfalico, edumazet, vyasevic, hkchu,
wuzhy, xemul, therbert, bhutchings, xii, stephen, jiri,
sergei.shtylyov, Pankaj Gupta
This patch series is followup to the RFC posted as:
https://lkml.org/lkml/2014/8/18/392
Changes from RFC are:
PATCH 1: Sergei Shtylyov - Add an empty line after declarations.
PATCH 2: Jiri Pirko - Do not introduce new module paramaters.
Michael.S.Tsirkin - We can use sysctl for limiting max number
of queues.
Networking under KVM works best if we allocate a per-vCPU rx and tx
queue in a virtual NIC. This requires a per-vCPU queue on the host side.
Modern physical NICs have multiqueue support for large number of queues.
To scale vNIC to run multiple queues parallel to maximum number of vCPU's
we need to increase number of queues support in tuntap.
This series is to increase the limit of tuntap queues. Original work is being
done by 'jasowang@redhat.com'. I am taking this 'https://lkml.org/lkml/2013/6/19/29'
patch series as a reference. As per discussion in the patch series:
There were two reasons which prevented us from increasing number of tun queues:
- The netdev_queue array in netdevice were allocated through kmalloc, which may
cause a high order memory allocation too when we have several queues.
E.g. sizeof(netdev_queue) is 320, which means a high order allocation would
happens when the device has more than 16 queues.
- We store the hash buckets in tun_struct which results a very large size of
tun_struct, this high order memory allocation fail easily when the memory is
fragmented.
The patch 60877a32bce00041528576e6b8df5abe9251fa73 increases the number of tx
queues. Memory allocation fallback to vzalloc() when kmalloc() fails.
This series tries to address following issues:
- Increase the number of netdev_queue queues for rx similarly its done for tx
queues by falling back to vzalloc() when memory allocation with kmalloc() fails.
- Switches to use flex array to implement the flow caches to avoid higher order
allocations.
- Accept maximum number of queues as sysctl param so that any user space
application like libvirt can use this value to limit number of queues. Also
Administrators can specify maximum number of queues by updating this sysctl
entry.
- Increase number of queues to 256, maximum number is equal to maximum number
of vCPUS allowed in a guest.
I have done some testing to find out any regression and with sample program
which creates tun/tap for single queue / multiqueue device and it seems to be
working fine. I will also post the performance numbers.
tuntap: Increase the number of queues in tun
tuntap: Reduce the size of tun_struct by using flex array
tuntap: Accept tuntap max queue length as sysctl entry
net: allow large number of rx queues
drivers/net/tun.c | 91 +++++++++++++++++++++++++++++++++++---------
net/core/dev.c | 19 ++++++---
2 files changed, 86 insertions(+), 24 deletions(-)
^ permalink raw reply
* Re: [PATCH 1/1 net-next] wireless: remove unnecessary sizeof(u8)
From: Larry Finger @ 2014-11-18 16:12 UTC (permalink / raw)
To: Gheorhios, John W. Linville
Cc: Fabian Frederick, Emmanuel Grumbach, Stefano Brivio,
Johannes Berg, Julian Calaby, linux-wireless,
linux-kernel@vger.kernel.org, Intel Linux Wireless, Chaoming Li,
b43-dev, netdev
In-Reply-To: <CAFpvzpxYrjR_tSF99iosMGZTAmt=_6o_aA+iYjNhxRGJ9aJ6oA@mail.gmail.com>
On 11/18/2014 07:34 AM, Gheorhios wrote:
> Anyone could gently send the link for downloading B43 linux drivers for
> this procedure?
>
>
>
> Then transfer it over to your Ubuntu box.
>
> Now in your Ubuntu Box [computer] please make your way to your Home folder.
>
> Once you are at your home folder right click on your home folder and make a
> new folder and call it wireless.
>
> Now that you have made a new folder called wireless in your home directory,
> it is time to move the downloaded file into the new folder called wireless.
>
> Move The Wireless Folder To The Firmware Directory
>
> sudo cp -r ~/wireless/* /lib/firmware/
>
> Now let's double check to make sure the download made it to the firmware
> directory. To do this type this into the terminal:
>
> ls /lib/firmware
>
> Ok so now that the download is in the firmware directory we need to go to
> that directory. To go there open your terminal and type in:
>
> cd /lib/firmware
>
> Now that you have changed directories let's double check to make sure you
> are in the right directory, this next code tells us where we are in the
> computer file directory. This next code stands for "print working
> directory".
>
> pwd
>
> Are you at /lib/firmware if so good if not go back one step.
>
> Now that we are in the firmware directory. We have to extract the download,
> to do this type in:
>
> sudo -s
>
> Then enter your password then:
>
> tar xvf b43-all-fw.tar_.gz
>
> Now is the firmware extracted properly? check by typing:
>
> ls /lib/firmware/b43
>
> or:
>
> ls /lib/firmware/b43legacy
>
> Do you see the ucode files? if so then delete the gz file:
>
> sudo rm *.gz
>
> Then:
>
> exit
>
> Reboot
No, I do not know where that file is found. Even if I knew of such a file,
Broadcom has expressly declined to provide that firmware for redistribution.
Posting such a file could invite legal action. If someone else has violated
Broadcom's directive, I would not facilitate that violation.
Rather than doing that, the link at
http://wireless.kernel.org/en/users/Drivers/b43#firmwareinstallation shows what
to do for Ubuntu installations. Follow those instructions - they refer to a
legal way to get the firmware. I would have thought that asking this question on
an Ubuntu Forum would have been more productive.
By the way, piggybacking your request on this thread is very bad netiquette. You
should not have done a "reply-to".
Larry
^ permalink raw reply
* Re: [PATCH] usbnet: rtl8150: remove unused variable
From: Petko Manolov @ 2014-11-18 15:30 UTC (permalink / raw)
To: Sudip Mukherjee; +Cc: linux-usb, netdev, linux-kernel
In-Reply-To: <1416315599-16939-1-git-send-email-sudipm.mukherjee@gmail.com>
On 14-11-18 18:29:59, Sudip Mukherjee wrote:
> we were just returning the initial value of res, instead now
> we are returning the value directly.
Looks OK, but could you please fix the wording of the commit message to something like:
Remove unused variable.
This is a tiny patch and the code speaks for itself.
Petko
> Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> ---
> drivers/net/usb/rtl8150.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/drivers/net/usb/rtl8150.c b/drivers/net/usb/rtl8150.c
> index 6e87e57..d37b7dc 100644
> --- a/drivers/net/usb/rtl8150.c
> +++ b/drivers/net/usb/rtl8150.c
> @@ -753,14 +753,13 @@ static int rtl8150_open(struct net_device *netdev)
> static int rtl8150_close(struct net_device *netdev)
> {
> rtl8150_t *dev = netdev_priv(netdev);
> - int res = 0;
>
> netif_stop_queue(netdev);
> if (!test_bit(RTL8150_UNPLUG, &dev->flags))
> disable_net_traffic(dev);
> unlink_all_urbs(dev);
>
> - return res;
> + return 0;
> }
>
> static void rtl8150_get_drvinfo(struct net_device *netdev, struct ethtool_drvinfo *info)
> --
> 1.8.1.2
>
^ permalink raw reply
* [PATCH net] net/mlx4_en: Add VXLAN ndo calls to the PF net device ops too
From: Or Gerlitz @ 2014-11-18 15:51 UTC (permalink / raw)
To: David S. Miller
Cc: netdev, Florian Westphal, Amir Vadai, Saeed Mahameed, Or Gerlitz
This is currently missing, which results in a crash when one attempts
to set VXLAN tunnel over the mlx4_en when acting as PF.
[ 2408.785472] BUG: unable to handle kernel NULL pointer dereference at (null)
[...]
[ 2408.994104] Call Trace:
[ 2408.996584] [<ffffffffa021f7f5>] ? vxlan_get_rx_port+0xd6/0x103 [vxlan]
[ 2409.003316] [<ffffffffa021f71f>] ? vxlan_lowerdev_event+0xf2/0xf2 [vxlan]
[ 2409.010225] [<ffffffffa0630358>] mlx4_en_start_port+0x862/0x96a [mlx4_en]
[ 2409.017132] [<ffffffffa063070f>] mlx4_en_open+0x17f/0x1b8 [mlx4_en]
While here, make sure to invoke vxlan_get_rx_port() only when VXLAN
offloads are actually enabled and not when they are only supported.
Reported-by: Ido Shamay <idos@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 7 ++++++-
1 files changed, 6 insertions(+), 1 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
index c5fcc56..4d69e38 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
@@ -1693,7 +1693,7 @@ int mlx4_en_start_port(struct net_device *dev)
mlx4_set_stats_bitmap(mdev->dev, &priv->stats_bitmap);
#ifdef CONFIG_MLX4_EN_VXLAN
- if (priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS)
+ if (priv->mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
vxlan_get_rx_port(dev);
#endif
priv->port_up = true;
@@ -2422,6 +2422,11 @@ static const struct net_device_ops mlx4_netdev_ops_master = {
.ndo_rx_flow_steer = mlx4_en_filter_rfs,
#endif
.ndo_get_phys_port_id = mlx4_en_get_phys_port_id,
+#ifdef CONFIG_MLX4_EN_VXLAN
+ .ndo_add_vxlan_port = mlx4_en_add_vxlan_port,
+ .ndo_del_vxlan_port = mlx4_en_del_vxlan_port,
+ .ndo_gso_check = mlx4_en_gso_check,
+#endif
};
int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
--
1.7.1
^ permalink raw reply related
* Re: [PATCH 0/4] Add lightweight memory barriers fast_rmb() and fast_wmb()
From: Alexander Duyck @ 2014-11-18 15:44 UTC (permalink / raw)
To: David Laight, linux-arch@vger.kernel.org, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org
Cc: mathieu.desnoyers@polymtl.ca, peterz@infradead.org,
benh@kernel.crashing.org, heiko.carstens@de.ibm.com,
mingo@kernel.org, mikey@neuling.org, linux@arm.linux.org.uk,
donald.c.skidmore@intel.com, matthew.vick@intel.com,
geert@linux-m68k.org, jeffrey.t.kirsher@intel.com,
romieu@fr.zoreil.com, paulmck@linux.vnet.ibm.com,
nic_swsd@realtek.com, will.deacon@arm.com, michael@ellerman.id.au,
tony.luck@intel.com, "torvalds@li
In-Reply-To: <063D6719AE5E284EB5DD2968C1650D6D1C9F2930@AcuExch.aculab.com>
On 11/18/2014 01:57 AM, David Laight wrote:
> From: Alexander Duyck
>> These patches introduce two new primitives for synchronizing cache-enabled
>> memory writes and reads. These two new primitives are:
>>
>> fast_rmb()
>> fast_wmb()
> Not sure I like the names.
> If the aim is to sync data into the local cache so that hardware
> that is doing cache-snooping accesses sees the data then maybe
> local_rmb() and local_wmb()
Yeah, that is the general consensus. I am planning to change them to
coherent_rmb() and coherent_wmb().
> IIRC read_barrier_depends() is a nop on everything except alpha.
> Maybe add the default if it isn't defined by the MD file?
>
> David
>
From my patch the only two I saw define it were alpha and blackfin. It
is already defined in asm-generic, the rest is just clean-up since I
suspect some of the arch tree barrier.h calls just borrowed from
asm-generic without sorting out what became redundancies.
Thanks,
Alex
^ permalink raw reply
* Re: [PATCH net-next] PPC: bpf_jit_comp: Unify BPF_MOD | BPF_X and BPF_DIV | BPF_X
From: Alexei Starovoitov @ 2014-11-18 15:37 UTC (permalink / raw)
To: Denis Kirjanov
Cc: Michael Ellerman, netdev@vger.kernel.org, Philippe Bergheaud,
linuxppc-dev, Daniel Borkmann
In-Reply-To: <CAOJe8K0_u51QkKmVoVmFuGaDtKfcHP6tSWx_fgJM_erRRoQfUA@mail.gmail.com>
On Mon, Nov 17, 2014 at 10:58 PM, Denis Kirjanov <kda@linux-powerpc.org> wrote:
> Hi Michael,
>
> This patch added no new functionality so I haven't put the test
> results (of course I ran the test suite to check the patch).
>
> The output :
> [ 650.198958] test_bpf: Summary: 60 PASSED, 0 FAILED
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
btw, please don't top post.
^ permalink raw reply
* Re: [PATCH net] bonding: fix curr_active_slave/carrier with loadbalance arp monitoring
From: Andy Gospodarek @ 2014-11-18 15:28 UTC (permalink / raw)
To: Veaceslav Falico
Cc: Nikolay Aleksandrov, netdev, davem, Jay Vosburgh, Andy Gospodarek,
Ding Tianhong
In-Reply-To: <20141118143727.GA2643@raspberrypi>
On Tue, Nov 18, 2014 at 03:37:27PM +0100, Veaceslav Falico wrote:
> On Tue, Nov 18, 2014 at 03:14:44PM +0100, Nikolay Aleksandrov wrote:
> >Since commit 6fde8f037e60 ("bonding: fix locking in
> >bond_loadbalance_arp_mon()") we can have a stale bond carrier state and
> >stale curr_active_slave when using arp monitoring in loadbalance modes. The
> >reason is that in bond_loadbalance_arp_mon() we can't have
> >do_failover == true but slave_state_changed == false, whenever do_failover
> >is true then slave_state_changed is also true. Then the following piece
> >from bond_loadbalance_arp_mon():
> > if (slave_state_changed) {
> > bond_slave_state_change(bond);
> > if (BOND_MODE(bond) == BOND_MODE_XOR)
> > bond_update_slave_arr(bond, NULL);
> > } else if (do_failover) {
>
> Ouch, must have been a big PITA to track :).
Agreed!
>
> > block_netpoll_tx();
> > bond_select_active_slave(bond);
> > unblock_netpoll_tx();
> > }
> >
> >will execute only the first branch, always and regardless of do_failover.
> >Since these two events aren't related in such way, we need to decouple and
> >consider them separately.
> >
> >For example this issue could lead to the following result:
> >Bonding Mode: load balancing (round-robin)
> >*MII Status: down*
> >MII Polling Interval (ms): 0
> >Up Delay (ms): 0
> >Down Delay (ms): 0
> >ARP Polling Interval (ms): 100
> >ARP IP target/s (n.n.n.n form): 192.168.9.2
> >
> >Slave Interface: ens12
> >*MII Status: up*
> >Speed: 10000 Mbps
> >Duplex: full
> >Link Failure Count: 2
> >Permanent HW addr: 00:0f:53:01:42:2c
> >Slave queue ID: 0
> >
> >Slave Interface: eth1
> >*MII Status: up*
> >Speed: Unknown
> >Duplex: Unknown
> >Link Failure Count: 70
> >Permanent HW addr: 52:54:00:2f:0f:8e
> >Slave queue ID: 0
> >
> >Since some interfaces are up, then the status of the bond should also be
> >up, but it will never change unless something invokes bond_set_carrier()
> >(i.e. enslave, bond_select_active_slave etc). Now, if I force the
> >calling of bond_select_active_slave via for example changing
> >primary_reselect (it can change in any mode), then the MII status goes to
> >"up" because it calls bond_select_active_slave() which should've been done
> >from bond_loadbalance_arp_mon() itself.
> >
> >CC: Veaceslav Falico <vfalico@gmail.com>
>
> Acked-by: Veaceslav Falico <vfalico@gmail.com>
Acked-by: Andy Gospodarek <gospo@cumulusnetworks.com>
>
> >CC: Jay Vosburgh <j.vosburgh@gmail.com>
> >CC: Andy Gospodarek <andy@greyhouse.net>
> >CC: Ding Tianhong <dingtianhong@huawei.com>
> >
> >Fixes: 6fde8f037e60 ("bonding: fix locking in bond_loadbalance_arp_mon()")
> >Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
> >---
> >Note: I left the parent if() the same even though we can shorten it. I think
> > it's better this way since it shows that any of the two events can cause
> > it to enter even though currently we can't have do_failover without
> > slave_state_changed, that may also change in the future.
> >
> >drivers/net/bonding/bond_main.c | 3 ++-
> >1 file changed, 2 insertions(+), 1 deletion(-)
> >
> >diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> >index c9ac06cfe6b7..a5115fb7cf33 100644
> >--- a/drivers/net/bonding/bond_main.c
> >+++ b/drivers/net/bonding/bond_main.c
> >@@ -2471,7 +2471,8 @@ static void bond_loadbalance_arp_mon(struct work_struct *work)
> > bond_slave_state_change(bond);
> > if (BOND_MODE(bond) == BOND_MODE_XOR)
> > bond_update_slave_arr(bond, NULL);
> >- } else if (do_failover) {
> >+ }
> >+ if (do_failover) {
> > block_netpoll_tx();
> > bond_select_active_slave(bond);
> > unblock_netpoll_tx();
> >--
> >1.9.3
> >
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* Re: [PATCH net] bonding: fix curr_active_slave/carrier with loadbalance arp monitoring
From: Veaceslav Falico @ 2014-11-18 14:37 UTC (permalink / raw)
To: Nikolay Aleksandrov
Cc: netdev, davem, Jay Vosburgh, Andy Gospodarek, Ding Tianhong
In-Reply-To: <1416320084-25339-1-git-send-email-nikolay@redhat.com>
On Tue, Nov 18, 2014 at 03:14:44PM +0100, Nikolay Aleksandrov wrote:
>Since commit 6fde8f037e60 ("bonding: fix locking in
>bond_loadbalance_arp_mon()") we can have a stale bond carrier state and
>stale curr_active_slave when using arp monitoring in loadbalance modes. The
>reason is that in bond_loadbalance_arp_mon() we can't have
>do_failover == true but slave_state_changed == false, whenever do_failover
>is true then slave_state_changed is also true. Then the following piece
>from bond_loadbalance_arp_mon():
> if (slave_state_changed) {
> bond_slave_state_change(bond);
> if (BOND_MODE(bond) == BOND_MODE_XOR)
> bond_update_slave_arr(bond, NULL);
> } else if (do_failover) {
Ouch, must have been a big PITA to track :).
> block_netpoll_tx();
> bond_select_active_slave(bond);
> unblock_netpoll_tx();
> }
>
>will execute only the first branch, always and regardless of do_failover.
>Since these two events aren't related in such way, we need to decouple and
>consider them separately.
>
>For example this issue could lead to the following result:
>Bonding Mode: load balancing (round-robin)
>*MII Status: down*
>MII Polling Interval (ms): 0
>Up Delay (ms): 0
>Down Delay (ms): 0
>ARP Polling Interval (ms): 100
>ARP IP target/s (n.n.n.n form): 192.168.9.2
>
>Slave Interface: ens12
>*MII Status: up*
>Speed: 10000 Mbps
>Duplex: full
>Link Failure Count: 2
>Permanent HW addr: 00:0f:53:01:42:2c
>Slave queue ID: 0
>
>Slave Interface: eth1
>*MII Status: up*
>Speed: Unknown
>Duplex: Unknown
>Link Failure Count: 70
>Permanent HW addr: 52:54:00:2f:0f:8e
>Slave queue ID: 0
>
>Since some interfaces are up, then the status of the bond should also be
>up, but it will never change unless something invokes bond_set_carrier()
>(i.e. enslave, bond_select_active_slave etc). Now, if I force the
>calling of bond_select_active_slave via for example changing
>primary_reselect (it can change in any mode), then the MII status goes to
>"up" because it calls bond_select_active_slave() which should've been done
>from bond_loadbalance_arp_mon() itself.
>
>CC: Veaceslav Falico <vfalico@gmail.com>
Acked-by: Veaceslav Falico <vfalico@gmail.com>
>CC: Jay Vosburgh <j.vosburgh@gmail.com>
>CC: Andy Gospodarek <andy@greyhouse.net>
>CC: Ding Tianhong <dingtianhong@huawei.com>
>
>Fixes: 6fde8f037e60 ("bonding: fix locking in bond_loadbalance_arp_mon()")
>Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
>---
>Note: I left the parent if() the same even though we can shorten it. I think
> it's better this way since it shows that any of the two events can cause
> it to enter even though currently we can't have do_failover without
> slave_state_changed, that may also change in the future.
>
> drivers/net/bonding/bond_main.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
>diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
>index c9ac06cfe6b7..a5115fb7cf33 100644
>--- a/drivers/net/bonding/bond_main.c
>+++ b/drivers/net/bonding/bond_main.c
>@@ -2471,7 +2471,8 @@ static void bond_loadbalance_arp_mon(struct work_struct *work)
> bond_slave_state_change(bond);
> if (BOND_MODE(bond) == BOND_MODE_XOR)
> bond_update_slave_arr(bond, NULL);
>- } else if (do_failover) {
>+ }
>+ if (do_failover) {
> block_netpoll_tx();
> bond_select_active_slave(bond);
> unblock_netpoll_tx();
>--
>1.9.3
>
^ permalink raw reply
* Re: [net-next 03/12] i40e: Handle a single mss packet with more than 8 frags
From: Eric Dumazet @ 2014-11-18 14:33 UTC (permalink / raw)
To: David Laight
Cc: 'Nelson, Shannon', Kirsher, Jeffrey T,
davem@davemloft.net, Kong, Serey, netdev@vger.kernel.org,
nhorman@redhat.com, sassmann@redhat.com, jogreene@redhat.com
In-Reply-To: <063D6719AE5E284EB5DD2968C1650D6D1C9F290B@AcuExch.aculab.com>
On Tue, 2014-11-18 at 09:46 +0000, David Laight wrote:
> That isn't entirely clear from the patch context.
gso_segs = 1 was not clear ???
gso_segs = 1;
if (skb_shinfo(skb)->nr_frags >= I40E_MAX_BUFFER_TXD)
skb_linearize(skb);
Nelson, what happens if a packet has 2 segments and uses 17 fragments ?
Is the hardware OK with that ?
netperf -t TCP_SENDFILE -- -m 171
I did the test and it seems fine on my host, but maybe it was pure luck.
While the following indeed froze the card :
netperf -H 7.7.7.27 -t TCP_SENDFILE -- -m 30
[3257268.702689] i40e 0000:84:00.0: requesting a pf reset
[3257268.831786] i40e 0000:84:00.0: i40e_ptp_init: added PHC on eth2
[3257268.912798] i40e 0000:84:00.0 eth2: NIC Link is Up 40 Gbps Full Duplex, Flow Control: RX/TX
[3257338.895753] i40e 0000:84:00.0: Detected Tx Unit Hang
[3257338.895753] VSI <518>
[3257338.895753] Tx Queue <16>
[3257338.895753] next_to_use <17c>
[3257338.895753] next_to_clean <b2>
[3257338.895757] i40e 0000:84:00.0: tx_bi[next_to_clean]
[3257338.895757] time_stamp <1c20f09c9>
[3257338.895757] jiffies <1c20f0eac>
[3257338.895759] i40e 0000:84:00.0: tx hang detected on queue 16, resetting adapter
[3257338.895769] i40e 0000:84:00.0 eth2: tx_timeout recovery level 1
[3257338.895926] i40e 0000:84:00.0: i40e_vsi_control_tx: VSI seid 518 Tx ring 0 disable timeout
[3257338.949922] i40e 0000:84:00.0: i40e_vsi_control_tx: VSI seid 520 Tx ring 64 disable timeout
[3257339.410705] i40e 0000:84:00.0: PF reset failed, -15
>
> For a non-TSO packet the skb_serialize() is less likely to fail
> since it doesn't need contiguous pages.
Any memory allocation can fail, regardless of the size.
We do not want a memory stress being able to crash the host,
likely or not. It is not worth discussing this, really.
^ permalink raw reply
* [PATCH net] bonding: fix curr_active_slave/carrier with loadbalance arp monitoring
From: Nikolay Aleksandrov @ 2014-11-18 14:14 UTC (permalink / raw)
To: netdev
Cc: davem, Nikolay Aleksandrov, Veaceslav Falico, Jay Vosburgh,
Andy Gospodarek, Ding Tianhong
Since commit 6fde8f037e60 ("bonding: fix locking in
bond_loadbalance_arp_mon()") we can have a stale bond carrier state and
stale curr_active_slave when using arp monitoring in loadbalance modes. The
reason is that in bond_loadbalance_arp_mon() we can't have
do_failover == true but slave_state_changed == false, whenever do_failover
is true then slave_state_changed is also true. Then the following piece
from bond_loadbalance_arp_mon():
if (slave_state_changed) {
bond_slave_state_change(bond);
if (BOND_MODE(bond) == BOND_MODE_XOR)
bond_update_slave_arr(bond, NULL);
} else if (do_failover) {
block_netpoll_tx();
bond_select_active_slave(bond);
unblock_netpoll_tx();
}
will execute only the first branch, always and regardless of do_failover.
Since these two events aren't related in such way, we need to decouple and
consider them separately.
For example this issue could lead to the following result:
Bonding Mode: load balancing (round-robin)
*MII Status: down*
MII Polling Interval (ms): 0
Up Delay (ms): 0
Down Delay (ms): 0
ARP Polling Interval (ms): 100
ARP IP target/s (n.n.n.n form): 192.168.9.2
Slave Interface: ens12
*MII Status: up*
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 2
Permanent HW addr: 00:0f:53:01:42:2c
Slave queue ID: 0
Slave Interface: eth1
*MII Status: up*
Speed: Unknown
Duplex: Unknown
Link Failure Count: 70
Permanent HW addr: 52:54:00:2f:0f:8e
Slave queue ID: 0
Since some interfaces are up, then the status of the bond should also be
up, but it will never change unless something invokes bond_set_carrier()
(i.e. enslave, bond_select_active_slave etc). Now, if I force the
calling of bond_select_active_slave via for example changing
primary_reselect (it can change in any mode), then the MII status goes to
"up" because it calls bond_select_active_slave() which should've been done
from bond_loadbalance_arp_mon() itself.
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: Ding Tianhong <dingtianhong@huawei.com>
Fixes: 6fde8f037e60 ("bonding: fix locking in bond_loadbalance_arp_mon()")
Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
---
Note: I left the parent if() the same even though we can shorten it. I think
it's better this way since it shows that any of the two events can cause
it to enter even though currently we can't have do_failover without
slave_state_changed, that may also change in the future.
drivers/net/bonding/bond_main.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index c9ac06cfe6b7..a5115fb7cf33 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -2471,7 +2471,8 @@ static void bond_loadbalance_arp_mon(struct work_struct *work)
bond_slave_state_change(bond);
if (BOND_MODE(bond) == BOND_MODE_XOR)
bond_update_slave_arr(bond, NULL);
- } else if (do_failover) {
+ }
+ if (do_failover) {
block_netpoll_tx();
bond_select_active_slave(bond);
unblock_netpoll_tx();
--
1.9.3
^ permalink raw reply related
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