On Mon, 28 Sep 2015 11:30:00 -0500 (CDT) Christoph Lameter
wrote:
> On Mon, 28 Sep 2015, Jesper Dangaard Brouer wrote:
>
> > Not knowing SLUB as well as you, it took me several hours to realize
> > init_object() didn't overwrite the freepointer in the object. Thus, I
> >
On Tue, Sep 29, 2015 at 03:27:03AM +0300, Pravin Shelar wrote:
> On Mon, Sep 28, 2015 at 2:46 AM, Andrew Vagin wrote:
> > Hi,
> >
> > With this patch, I can't connect two local tcp ipv6 sockets.
> >
> > [root@fc22-vm criu]# strace -e network python ipv6.py
> > socket(PF_INET6,
RUC_Soft_Sec writes:
> Current code assigns 0 to variable 'retval', which makes ezusb_probe() to
> return success even if alloc_orinocodev() fails.
>
> The related code snippets in mantis_dma_init() is as following.
>
> 1573 static int ezusb_probe(struct usb_interface
Hello,
On Mon, 28 Sep 2015, Eric W. Biederman wrote:
> Julian Anastasov writes:
>
> > On Mon, 28 Sep 2015, Anand Gurram wrote:
> >
> >> I am currently using kernel version 3.16.7 on a linux switch.
> >> While creating and destroying network namespaces I am observing below
On Mon, 2015-09-28 at 12:37 -0700, Tom Herbert wrote:
> I think it's easier to just call skb_checksum_help from the driver
> when the packet is actually sent to the device (should be no cost for
> late binding).
That's true for checksum. Not for things like TSO though, and I wonder
if it's worth
> The mwifiex_dbg macro has two tests that could be consolidated
> into a function reducing overall object size ~10KB (~4%).
>
> So convert the macro into a function.
>
> $ size drivers/net/wireless/mwifiex/built-in.o* (x86-64 defconfig)
>text data bss dec hex filename
Hi Paul,
On Mon, 28 Sep 2015, Paul Gortmaker wrote:
> On 28/09/2015 (Mon 23:09) Geert Uytterhoeven wrote:
>
> > Hi Paul,
...
> >
> > Why did you choose this approach?
> > What about changing the "bool"s to "tristate"s in Kconfig instead?
>
> Long answer is here:
>
>
Anand Gurram writes:
>>If the message just spits out a few times and then goes away it simply
>>means that something is taking a while to cleanup and drop it's
>>reference.
>
> The message just spits out few times and then goes away, I am trying
> to debug why cleanup
On Tue, Sep 29, 2015 at 12:59:35AM +0300, Or Gerlitz wrote:
> On Tue, Sep 29, 2015 at 12:04 AM, Christoffer Dall
> wrote:
> > Hi,
> >
> > In doing some performance experiments I found that using a 10G Mellanox
> > MX354A Dual port FDR CX3 device on a server running
> > >> > +struct qede_rx_queue {
> > >> > + __le16 *hw_cons_ptr;
> > >>
> > >> The __ variants of constants should be reserved for use in user
> > >> visible API's
> > >
> > > Really? If so, this needs to be fixed not only here but in lots of
> > > places in the series [e.g.,
> IW_AUTH_ALG_OPEN_SYSTEM is ambiguous in set_auth for WEP as
> wpa_supplicant uses it for both no encryption and WEP open system.
> Cache the last mode set (only of these two) and use it here.
>
> This allows wpa_supplicant to work with unencrypted APs.
>
> Signed-off-by: Ondrej Zary
Hi Daniel,
[auto build test results on v4.3-rc3 -- if it's inappropriate base, please
ignore]
config: m68k-sun3_defconfig (attached as .config)
reproduce:
wget
https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross
-O ~/bin/make.cross
chmod +x
Thanks Julian, I will check if these two commits work for me.
>I think, they will appear in other stable versions too...
Yes, I saw them in other versions, the fix which is suggested in those
branches didn't work for me.
Hope the above two commits help.
Regards,
Anand
On Tue, Sep 29, 2015 at
>If the message just spits out a few times and then goes away it simply
>means that something is taking a while to cleanup and drop it's
>reference.
The message just spits out few times and then goes away, I am trying
to debug why cleanup is taking long,
and where it is still referenced. Any
On Mon, 2015-09-28 at 20:04 -0700, Tom Herbert wrote:
>
> > I've been pondering a bit of a redesign in this space. I think the
> > skb struct should be explicit in its instructions to hardware for
> > which offloads to do for each packet.
> >
> > In this way, the stack would be *directly*
On Mon, 28 Sep 2015 11:28:15 -0500 (CDT)
Christoph Lameter wrote:
> On Mon, 28 Sep 2015, Jesper Dangaard Brouer wrote:
>
> > > Do you really need separate parameters for freelist_head? If you just want
> > > to deal with one object pass it as freelist_head and set cnt = 1?
> >
>
Hi David, Paul,
This patch series makes the remaining m68k Ethernet drivers modular.
It's an alternative to the last 3 patches of Paul Gortmaker's series
"[PATCH net-next 0/6] make non-modular code explicitly non-modular".
Note that "[PATCH 5/5] net: macmace: Allow modular build" depends
Signed-off-by: Geert Uytterhoeven
---
drivers/net/ethernet/amd/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/amd/Kconfig b/drivers/net/ethernet/amd/Kconfig
index afc62ea804fc35d4..0038709fd317d83c 100644
---
Signed-off-by: Geert Uytterhoeven
---
This depends on "[PATCH 4/5] m68k/mac: Export Peripheral System
Controller (PSC) base address to modules".
---
drivers/net/ethernet/apple/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
If CONFIG_HPLANCE=m and CONFIG_NET_POLL_CONTROLLER=y:
ERROR: "lance_poll" [drivers/net/ethernet/amd/hplance.ko] undefined!
Add the missing export to fix this.
Signed-off-by: Geert Uytterhoeven
---
drivers/net/ethernet/amd/7990.c | 1 +
1 file changed, 1 insertion(+)
If CONFIG_MACMACE=m:
ERROR: psc [drivers/net/ethernet/apple/macmace.ko] undefined!
Add the missing export to fix this.
Signed-off-by: Geert Uytterhoeven
---
I'm OK with this going in through the netdev tree, as "net: macmace:
Allow modular build" depends on it.
---
The modular driver supports only one card, just like the built-in
driver.
Note that this limitation is a problem which affects all Nubus card
drivers, because they have to do all their own bus matching, because
Nubus still lacks the necessary driver model support.
Suggested-by: Finn Thain
On Mon, 28 Sep 2015 06:48:09 -0700
roopa wrote:
> On 9/28/15, 1:57 AM, Matthew Dupre wrote:
> > Hi,
> >
> > I'm interested in the Linux kernel's support for per-flow IPv4 ECMP
> > (i.e. consistent path selection based on a hash of the connection
> > tuple). I'd been
From: Peter Nørlund
> Sent: 29 September 2015 12:29
...
> As for using L4 hashing with anycast, CloudFlare apparently does L4
> hashing - they could have disabled it, but they didn't. Besides,
> analysis of my own load balancers showed that only one in every
> 500,000,000 packets is fragmented.
[Re: [PATCH net-next 0/6] make non-modular code explicitly non-modular] On
29/09/2015 (Tue 16:32) Finn Thain wrote:
>
> Hi Paul,
>
> On Mon, 28 Sep 2015, Paul Gortmaker wrote:
>
> > On 28/09/2015 (Mon 23:09) Geert Uytterhoeven wrote:
> >
> > > Hi Paul,
> ...
> > >
> > > Why did you choose
On Mon, 28 Sep 2015 19:55:41 -0700 (PDT)
David Miller wrote:
> From: David Miller
> Date: Mon, 28 Sep 2015 19:33:55 -0700 (PDT)
>
> > From: Peter Nørlund
> > Date: Wed, 23 Sep 2015 21:49:35 +0200
> >
> >> When the routing cache was
Run the NF_INET_LOCAL_SOCKET_IN netfilter chain rules after the
destination socket for IPv4 unicast and multicast ports have been
looked up.
Signed-off-by: Daniel Mack
---
net/ipv4/netfilter/nf_tables_ipv4.c | 10 +-
net/ipv4/tcp_ipv4.c | 8
Run the NF_INET_LOCAL_SOCKET_IN netfilter chain rules after the
destination socket for IPv6 unicast and multicast ports have been
looked up.
Signed-off-by: Daniel Mack
---
net/ipv6/netfilter/nf_tables_ipv6.c | 14 --
net/ipv6/tcp_ipv6.c | 8
Add a new chain type NF_INET_LOCAL_SOCKET_IN which is ran after the
input demux is complete and the final destination socket (if any)
has been determined.
This helps filtering packets based on information stored in the
destination socket, such as cgroup controller supplied net class IDs.
Note
Run the NF_INET_LOCAL_SOCKET_IN netfilter chain rules after the
destination socket for DCCP packets have been looked up.
Signed-off-by: Daniel Mack
---
net/dccp/ipv4.c | 14 +-
net/dccp/ipv6.c | 14 +-
2 files changed, 26 insertions(+), 2 deletions(-)
Here is a patch set that enables for full support for match rules
that take into account information about the local receiver socket.
Such rules allow administrators to implement per-application or
per-container firewalls which filter any type of network traffic
directed to or originated from a
pkt->sk is set to whatever was passed to nh_hook() by the caller,
and for post demux chains, this is the one that should be looked
at, as skb->sk is still NULL at this point in time.
Signed-off-by: Daniel Mack
---
net/netfilter/nft_meta.c | 7 ---
1 file changed, 4
The high-level netfilter hook API already enables users to pass a socket,
but that information is lost when the chains are walked.
In order to let internal eval callbacks use the passed filter rather than
skb->sk, add a pointer of type 'struct sock' to 'struct nft_pktinfo' and
set that field via
Run the NF_INET_LOCAL_SOCKET_IN netfilter chain rules after the
destination socket for SCTP packets have been looked up.
Signed-off-by: Daniel Mack
---
net/sctp/input.c | 11 ++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/net/sctp/input.c
On Tue, 29 Sep 2015 02:35:04 +0200,
Pravin Shelar wrote:
>
> On Mon, Sep 28, 2015 at 6:12 AM, Takashi Iwai wrote:
> > [I resent this since the previous mail didn't go out properly, as it
> > seems; apologies if you already read it, please disregard]
> >
> > Hi,
> >
> > I noticed
Allow to change the replay threshold (XFRMA_REPLAY_THRESH) and expiry
timer (XFRMA_ETIMER_THRESH) of a state without having to set other
attributes like replay counter and byte lifetime. Changing these other
values while traffic flows will break the state.
Signed-off-by: Michael Rossberg
Similar to the notifier_call callback of a notifier_block, change the
function signature of switchdev dump operation to:
int switchdev_port_obj_dump(struct net_device *dev,
enum switchdev_obj_id id, void *obj,
int (*cb)(void
This patchset changes switchdev add, del, dump operations from this:
int (*switchdev_port_obj_add)(struct net_device *dev,
struct switchdev_obj *obj,
struct switchdev_trans *trans);
int
This patch series makes changes to TCP/DCCP stacks so that
we can switch listener code to lockless mode.
This is done by marking const the listener socket in all
appropriate paths.
FastOpen code had to be changed to not dynamically allocate
a very small structure to make code simpler for
Factorize code to get tcp header from skb. It makes no sense
to duplicate code in callers.
Signed-off-by: Eric Dumazet
---
include/net/tcp.h| 3 +--
net/ipv4/tcp_input.c | 4 ++--
net/ipv4/tcp_ipv4.c | 2 +-
net/ipv4/tcp_minisocks.c | 2 +-
Before changing dccp_v6_request_recv_sock() sock argument
to const, we need to get rid of security_sk_classify_flow(),
and it seems doable by reusing inet6_csk_route_req() helper.
We need to add a proto parameter to inet6_csk_route_req(),
not assume it is TCP.
Signed-off-by: Eric Dumazet
While auditing TCP stack for upcoming 'lockless' listener changes,
I found I had to change fastopen_init_queue() to properly init the object
before publishing it.
Otherwise an other cpu could try to lock the spinlock before it gets
properly initialized.
Instead of adding appropriate barriers,
tcp_syn_flood_action() will soon be called with unlocked socket.
In order to avoid SYN flood warning being emitted multiple times,
use xchg().
Extend max_qlen_log and synflood_warned fields in struct listen_sock
to u32
Signed-off-by: Eric Dumazet
---
Some common IPv4/IPv6 code can be factorized.
Also constify cookie_init_sequence() socket argument.
Signed-off-by: Eric Dumazet
---
include/net/tcp.h | 19 ++-
net/ipv4/syncookies.c | 6 +-
net/ipv6/syncookies.c | 5 +
3 files changed, 12
This method does not touch the listener socket.
Signed-off-by: Eric Dumazet
---
include/net/tcp.h| 2 +-
net/ipv4/tcp_minisocks.c | 4 +++-
2 files changed, 4 insertions(+), 2 deletions(-)
diff --git a/include/net/tcp.h b/include/net/tcp.h
index
We'll soon no longer hold listener socket lock, these
functions do not modify the socket in any way.
Signed-off-by: Eric Dumazet
---
include/net/inet_connection_sock.h | 2 +-
include/net/tcp.h | 2 +-
net/dccp/dccp.h| 2 +-
These functions do not change the listener socket.
Goal is to make sure tcp_conn_request() is not messing with
listener in a racy way.
Signed-off-by: Eric Dumazet
---
include/net/tcp.h | 2 +-
net/ipv4/tcp_ipv4.c | 3 ++-
net/ipv6/tcp_ipv6.c | 3 ++-
3 files changed, 5
The FDB dump callback requires the related net_device so move it to the
struct switchdev_fdb_dump superset instead of using a callback param.
With this done, it'll be simpler to change the dump function signature.
Signed-off-by: Vivien Didelot
---
On 28/09/15 14:05, Alexander Stein wrote:
Otherwise 4294967295 (MBit/s) (-1) will be printed when there is no link.
Documentation/ABI/testing/sysfs-class-net does not state if this shall be
signed or unsigned.
Also remove the now unused variable fmt_udec.
[...]
- ret =
The net_device associated to a dump operation does not have to be passed
to the callback. switchdev stores it in a superset struct, if needed.
Also some drivers (such as DSA drivers) may not have easy access to it.
This will simplify pushing the callback function down to the drivers.
The static switchdev_port_vlan_dump_put function don't need the
net_device parameter, so remove it.
Signed-off-by: Vivien Didelot
---
net/switchdev/switchdev.c | 11 +--
1 file changed, 5 insertions(+), 6 deletions(-)
diff --git
socket is not touched, make it const.
Signed-off-by: Eric Dumazet
---
include/net/inet_hashtables.h | 2 +-
net/ipv4/inet_hashtables.c| 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/include/net/inet_hashtables.h b/include/net/inet_hashtables.h
The socket points to the (shared) listener.
Signed-off-by: Eric Dumazet
---
include/net/inet_connection_sock.h | 3 ++-
net/ipv4/inet_connection_sock.c| 2 +-
2 files changed, 3 insertions(+), 2 deletions(-)
diff --git a/include/net/inet_connection_sock.h
socket no longer needs to be read/write
Signed-off-by: Eric Dumazet
---
net/dccp/dccp.h | 2 +-
net/dccp/minisocks.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/dccp/dccp.h b/net/dccp/dccp.h
index 8ed1df2771bd..2409619b7043 100644
---
Signed-off-by: Eric Dumazet
---
include/net/sock.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index 94dff7f566f5..dfe2eb8e1132 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -759,7 +759,7 @@ static
Once we realize tcp_rcv_synsent_state_process() does not use
its 'len' argument and we get rid of it, then it becomes clear
this argument is no longer used in tcp_rcv_state_process()
Signed-off-by: Eric Dumazet
---
include/net/tcp.h| 2 +-
net/ipv4/tcp_input.c |
None of these functions need to change the socket, make it
const.
Signed-off-by: Eric Dumazet
---
include/net/request_sock.h | 4 ++--
net/dccp/dccp.h| 2 +-
net/dccp/ipv4.c| 2 +-
net/dccp/ipv6.c| 2 +-
net/dccp/minisocks.c |
Now that switchdev and its drivers directly use specific switchdev_obj_*
structures, move them out of the switchdev_obj union and get rif of this
outer structure.
Signed-off-by: Vivien Didelot
---
include/net/switchdev.h | 53
Both new_steering_entry() and existing_steering_entry() return values
based on their success or failure, but currently they fall through
silently. This can make troubleshooting difficult, as we were unable
to tell which one of these two functions returned errors or
specifically what code was
hi all,
in drivers/net/wireless/mwifiex/sta_ioctl.c the following functions
mwifiex_set_wpa_ie_helper
mwifiex_set_wapi_ie
mwifiex_set_wps_ie
can truncate the incoming ie_len argument from u16 to u8 when it gets
stored in mwifiex_private.wpa_ie_len,
Similar to the notifier_call callback of a notifier_block, change the
function signature of switchdev add and del operations to:
int switchdev_port_obj_add/del(struct net_device *dev,
enum switchdev_obj_id id, void *obj);
This allows the caller to pass a
Similar to the notifier_call callback of a notifier_block, change the
function signature of switchdev add and del operations to:
int switchdev_port_obj_add/del(struct net_device *dev,
enum switchdev_obj_id id, void *obj);
This allows the caller to pass a
On Mon, 28 Sep 2015 15:20:33 -0400, John W. Linville wrote:
> > To be really useful, geneve should open both IPv4 and IPv6 socket when
> > it's metadata based. Take a look at my recent patchset that does this
> > for vxlan: http://thread.gmane.org/gmane.linux.network/379282
>
> OK, that seems
On Tue, Sep 29, 2015 at 12:38:36PM -0400, Vivien Didelot wrote:
> Because of the default 0 value of ret in dsa_slave_port_attr_set, a
> driver may return -EOPNOTSUPP from the commit phase of a STP state,
> which triggers a WARN() from switchdev.
>
> This happened on a 6185 switch which does not
On 09/29/2015 08:48 AM, Jesper Dangaard Brouer wrote:
Make it possible to free a freelist with several objects by adjusting
API of slab_free() and __slab_free() to have head, tail and an objects
counter (cnt).
Tail being NULL indicate single object free of head object. This
allow compiler
The #ifdef of CONFIG_SLUB_DEBUG is located very far from
the associated #else. For readability mark it with a comment.
Signed-off-by: Jesper Dangaard Brouer
Acked-by: Christoph Lameter
---
mm/slub.c |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
Most important part of this patchset is the introducing of what I call
detached freelist, for improving SLUB performance of object freeing in
the "slowpath" of kmem_cache_free_bulk.
Tagging patchset with "V4" to avoid confusion with "V2":
(V2) http://thread.gmane.org/gmane.linux.kernel.mm/137469
The FDB dump callback requires the related net_device so move it to the
struct switchdev_fdb_dump superset instead of using a callback param.
With this done, it'll be simpler to change the dump function signature.
Signed-off-by: Vivien Didelot
---
Because of the default 0 value of ret in dsa_slave_port_attr_set, a
driver may return -EOPNOTSUPP from the commit phase of a STP state,
which triggers a WARN() from switchdev.
This happened on a 6185 switch which does not support hardware bridging.
Reported-by: Andrew Lunn
The static switchdev_port_vlan_dump_put function does not need the
net_device parameter, so remove it.
Signed-off-by: Vivien Didelot
---
net/switchdev/switchdev.c | 11 +--
1 file changed, 5 insertions(+), 6 deletions(-)
diff --git
Similar to the notifier_call callback of a notifier_block, change the
function signature of switchdev dump operation to:
int switchdev_port_obj_dump(struct net_device *dev,
enum switchdev_obj_id id, void *obj,
int (*cb)(void
Hi Julian and Eric
I tried both the patches which you have suggested, the issue is still
seen, I am observing same warning message thrown on the console
"unregister_netdevice: waiting for lo to become free. Usage count = 1".
>Sometimes people have addressed this class of issue with code
Hi Vivien,
[auto build test results on next-20150929 -- if it's inappropriate base, please
ignore]
config: i386-randconfig-s1-201539 (attached as .config)
reproduce:
git checkout b215cce51157820c4fb92ecfdc72f281a4286676
# save the attached .config to linux build tree
make ARCH=i386
All
When compiling Documentation/ptp/testptp.c the following compiler
warnings are printed out:
Documentation/ptp/testptp.c: In function ‘main’:
Documentation/ptp/testptp.c:367:11: warning: format ‘%lld’ expects argument
of type ‘long long int’, but argument 3 has type ‘__s64’ [-Wformat=]
This change focus on improving the speed of object freeing in the
"slowpath" of kmem_cache_free_bulk.
The calls slab_free (fastpath) and __slab_free (slowpath) have been
extended with support for bulk free, which amortize the overhead of
the (locked) cmpxchg_double.
To use the new bulking
Now that switchdev and its drivers directly use specific switchdev_obj_*
structures, move them out of the switchdev_obj union and get rif of this
outer structure.
Signed-off-by: Vivien Didelot
---
include/net/switchdev.h | 53
On Tue, 29 Sep 2015 09:38:30 -0700
Alexander Duyck wrote:
> On 09/29/2015 08:48 AM, Jesper Dangaard Brouer wrote:
> > Make it possible to free a freelist with several objects by adjusting
> > API of slab_free() and __slab_free() to have head, tail and an objects
> >
Begin forwarded message:
Date: Tue, 29 Sep 2015 07:19:32 +
From: "bugzilla-dae...@bugzilla.kernel.org"
To: "shemmin...@linux-foundation.org"
Subject: [Bug 105221] New: system panics under load on mlx4_en interfaces
Implement a basic approach of bulking in the SLAB allocator. Simply
use local_irq_{disable,enable} and call single alloc/free in a loop.
This simple implementation approach is surprising fast.
Notice the normal SLAB fastpath is: 96 cycles (24.119 ns). Below table
show that single object bulking
With kernels vmlinuz-4.3.0-rc2-00228-gd4a748a and earlier it is no
problem for me to stream HD-videos (700-800 Kbyte/s) from YouTube.
With the same video material and kernels
vmlinuz-4.3.0-rc2-00438-gd8cc397 and later I only reach 70-80 KByte/s.
That's a one-tenth than before.
The merges between
This patchset changes switchdev add, del, dump operations from this:
int (*switchdev_port_obj_add)(struct net_device *dev,
struct switchdev_obj *obj,
struct switchdev_trans *trans);
int
The net_device associated to a dump operation does not have to be passed
to the callback. switchdev stores it in a superset struct, if needed.
Also some drivers (such as DSA drivers) may not have easy access to it.
This will simplify pushing the callback function down to the drivers.
Add support for filtering neighbor dumps by master device by adding
the NDA_MASTER attribute to the dump request. A new netlink flag,
NLM_F_DUMP_FILTERED, is added to indicate the kernel supports the
request and output is filtered as requested.
Signed-off-by: David Ahern
Make it possible to free a freelist with several objects by adjusting
API of slab_free() and __slab_free() to have head, tail and an objects
counter (cnt).
Tail being NULL indicate single object free of head object. This
allow compiler inline constant propagation in slab_free() and
From: Christoph Lameter
NOTICE: Accepted by AKPM
http://ozlabs.org/~akpm/mmots/broken-out/slub-avoid-irqoff-on-in-bulk-allocation.patch
Use the new function that can do allocation while
interrupts are disabled. Avoids irq on/off sequences.
Signed-off-by: Christoph Lameter
From: Christoph Lameter
NOTICE: Accepted by AKPM
http://ozlabs.org/~akpm/mmots/broken-out/slub-create-new-___slab_alloc-function-that-can-be-called-with-irqs-disabled.patch
Bulk alloc needs a function like that because it enables interrupts before
calling __slab_alloc which
Hi Dave,
> This patch adds missing inline wrappers for nla_get_le32 and
> nla_get_le64. The 802.15.4 MAC byteorder is little endian and we keep
> the byteorder for fields like address configuration in the same
> byteorder as it comes from the MAC layer.
>
> To provide these fields for nl802154
From: David Ahern
Date: Mon, 28 Sep 2015 10:12:13 -0700
> Wolfgang reported that IPv6 stack is ignoring oif in output route lookups:
...
> The stack does consider the oif but a mismatch in rt6_device_match is not
> considered fatal because RT6_LOOKUP_F_IFACE is not set
From: Liviu Dudau
Date: Mon, 28 Sep 2015 17:51:51 +0100
> On some embedded systems the EEPROM does not contain a valid MAC address.
> In that case it is better to fallback to a generated mac address and
> let init scripts fix the value later.
>
> Reported-by: Liviu Dudau
This patch set is meant to replace the calls to napi_schedule with
napi_schedule_irqoff as this should help to reduce the interrupt overhead
slightly by removing the unneeded call to local_irq_save and
local_irq_restore.
---
Alexander Duyck (3):
ixgbe/ixgbevf: use napi_schedule_irqoff()
The i40e_intr and i40e/i40evf_msix_clean_rings functions run from hard
interrupt context or with interrupts already disabled in netpoll.
They can use napi_schedule_irqoff() instead of napi_schedule()
Signed-off-by: Alexander Duyck
---
The fm10k_msix_clean_rings function runs from hard interrupt context or
with interrupts already disabled in netpoll.
It can use napi_schedule_irqoff() instead of napi_schedule()
Signed-off-by: Alexander Duyck
---
drivers/net/ethernet/intel/fm10k/fm10k_pci.c |2 +-
1
From: Pravin B Shelar
Date: Mon, 28 Sep 2015 17:24:25 -0700
> Earlier patch 6ae459bda tried to detect void ckecksum partial
> skb by comparing pull length to checksum offset. But it does
> not work for all cases since checksum-offset depends on
> updates to skb->data.
>
>
On 9/29/15 5:23 PM, David Miller wrote:
From: David Ahern
Date: Mon, 28 Sep 2015 10:16:50 -0700
v2
- rebased to top of net-next
- addressed Niks comments (checking master, removing extra lines, and
flipping the order of patches 1 and 2)
This still needs some
On Fri, Sep 25, 2015 at 3:35 PM, Thomas F Herbert
wrote:
> Pravin,
>
> Another comment and question. Please seen inline below.
>
> Thanks,
>
> --Tom
>
> On 9/24/15 7:42 PM, Pravin Shelar wrote:
>>
>> On Thu, Sep 24, 2015 at 10:58 AM, Thomas F Herbert
>>
Hi,
I'm seeing really poor IPv6 performance compared to IPv4. I've
checked using two different ARM platforms - an iMX6 platform using
the FEC driver, and an Armada 38x using mvneta.
The following was captured using iperf between the target system
and my laptop. The problem only occurs one-way.
From: Nikolay Aleksandrov
Date: Fri, 25 Sep 2015 19:00:11 +0200
> This patch changes the bridge vlan implementation to use rhashtables
> instead of bitmaps.
This seems to be taking the code in a good direction, and I'm kinda
happy to see more rhashtable users in the tree as
From: Alexander Stein
Date: Mon, 28 Sep 2015 15:05:33 +0200
> Otherwise 4294967295 (MBit/s) (-1) will be printed when there is no link.
> Documentation/ABI/testing/sysfs-class-net does not state if this shall be
> signed or unsigned.
> Also remove the now
In the current implementation of ILA, LWT is used to perform
translation on both the input and output paths. This is functional,
however there is a big performance hit in the receive path. Early
demux occurs before the routing lookup (a hit actually obviates the
route lookup). Therefore the stack
Add the rhashtable_replace_fast function. This replaces one object in
the table with another atomically. The hashes of the new and old objects
must be equal.
Signed-off-by: Tom Herbert
---
include/linux/rhashtable.h | 82 ++
1
1 - 100 of 208 matches
Mail list logo