This allows using OpenCores ethernet controller attached to its host in
native-endian mode with bi-endian CPUs. Example of such system is Xtensa
XTFPGA platform.
Signed-off-by: Max Filippov
---
Changes v1->v2:
- expand changelog with motivation for the change.
drivers/net/ethernet/ethoc.c | 14
On Tue, Sep 22, 2015 at 12:10:56PM -0400, Tejun Heo wrote:
>
> That's a pentium pro era errata. Virtually no working machine is
> affected by that anymore and nobody builds kernel with that option.
> In most cases, store_release and load_acquire are cheaper as they're
> more specific. On x86, sto
On 09/22/2015 03:55 PM, David Ahern wrote:
ip_route_input_slow is a maze of gotos (9 of them!) making it error
prone and difficult to read. This patchset refactors it, removing all
but 2 of the labels. The brd_input label for broadcast path requires
too many inputs to make a reasonble helper out
On 9/22/15, Christophe Leroy wrote:
> In several architectures, ip_fast_csum() is inlined
> There are functions like ip_send_check() which do nothing
> much more than calling ip_fast_csum().
> Inlining ip_fast_csum() allows the compiler to optimise better
Hi Christophe,
I did try it and see no di
Hi johannes,
Could you please tell me which kernel version will merge this patch?
Thanks,
Zhonghui
On 2015/9/22 22:05, Johannes Berg wrote:
> On Sat, 2015-09-19 at 10:40 +0800, Fu, Zhonghui wrote:
>> Now, PM core supports asynchronous suspend/resume mode for devices
>> during system suspend/
Currently error log messages in ip6gre_err are printed at 'warn'
level. This is different to most other tunnel types which don't
print any messages. These log messages don't provide any information
that couldn't be deduced with networking tools. Also it can be annoying
to have one end of the tunnel
Jiri Benc writes:
> One of the selling points of lwtunnel was the ability to specify the tunnel
> destination using routes. However, this doesn't really work currently, as
> ARP and ndisc replies are not handled correctly. ARP and ndisc replies won't
> have tunnel metadata attached, thus they wil
From: Wilson Kok
dump_rules returns skb length and not error.
But when family == AF_UNSPEC, the caller of dump_rules
assumes that it returns an error. Hence, when family == AF_UNSPEC,
we continue trying to dump on -EMSGSIZE errors resulting in
incorrect dump idx carried between skbs belonging to
From: Eric Dumazet
Neal suggested to move sk_txhash init into tcp_create_openreq_child(),
called both from IPv4 and IPv6.
This opportunity was missed in commit 58d607d3e52f ("tcp: provide
skb->hash to synack packets")
Signed-off-by: Eric Dumazet
Signed-off-by: Neal Cardwell
---
v2: respin
n
On Tue, 2015-09-22 at 17:31 -0700, David Miller wrote:
> This doesn't apply to net-next, please respin.
Sure.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info
On 9/22/15 8:33 PM, Alexander Duyck wrote:
Why pass loopback_dev instead of just passing the net pointer? Seems
like in the path below there are cases where you end up not needing to
dereference it and it might be worth the effort to avoid dereferencing
it if you don't need to.
Same thing for n
On 9/22/15 8:15 PM, Eric W. Biederman wrote:
David Ahern writes:
e_nobufs has 1 user. Move setting err to -ENOBUFS for the 1 user and
use the goto out label instead of e_nobufs. Stepping stone patch; next
one moves rth code into a helper function.
Ick you are pessimizing the code.
You will
On 09/22/2015 03:55 PM, David Ahern wrote:
Move the rth lookup and allocation from ip_route_input_slow into a
helper function. Code move only; no operational change intended.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 104 ---
1 file
David Ahern writes:
> e_nobufs has 1 user. Move setting err to -ENOBUFS for the 1 user and
> use the goto out label instead of e_nobufs. Stepping stone patch; next
> one moves rth code into a helper function.
Ick you are pessimizing the code.
You will almost certainly have better code generatio
Simon Horman writes:
> On Mon, Sep 21, 2015 at 01:01:39PM -0500, Eric W. Biederman wrote:
>> Instead store ipvs in extra2 so that proc_do_defense_mode can easily
>> find the ipvs that it's value is associated with.
>>
>> Signed-off-by: "Eric W. Biederman"
>
> I am wondering if this fix should b
On 09/22/2015 06:04 PM, Alexander Duyck wrote:
On 09/22/2015 03:55 PM, David Ahern wrote:
err is initialized to -EINVAL when it is declared. It is not reset until
fib_lookup which is well after the 3 users of the martian_source jump. So
resetting err to -EINVAL at martian_source label is not nee
* Neil Armstrong [150922 02:01]:
> In case the DaVinci Emac is directly connected to a
> non-mdio PHY/device, it should be possible to provide
> a fixed link configuration in the DT.
>
> Signed-off-by: Neil Armstrong
Ethernet works for me with this patch:
Tested-by: Tony Lindgren
> ---
> dr
On Mon, Sep 21, 2015 at 01:01:39PM -0500, Eric W. Biederman wrote:
> Instead store ipvs in extra2 so that proc_do_defense_mode can easily
> find the ipvs that it's value is associated with.
>
> Signed-off-by: "Eric W. Biederman"
I am wondering if this fix should be included in v4.3 and stable.
C
On 09/22/2015 03:55 PM, David Ahern wrote:
err is initialized to -EINVAL when it is declared. It is not reset until
fib_lookup which is well after the 3 users of the martian_source jump. So
resetting err to -EINVAL at martian_source label is not needed.
Removing that line obviates the need for t
From: David Daney
When the Cavium mdio-octeon devices appear in the Thunder family of
arm64 based SoCs, they show up as PCI devices. Add PCI driver
wrapping so the driver is bound in the standard PCI device scan.
When in this form, a single PCI device may have more than a single
bus, we call th
From: Neil Armstrong
Date: Tue, 22 Sep 2015 11:28:14 +0200
> The Marvell Egress rx trailer check must be fixed to
> correctly detect bad bits in the third byte of the
> Eggress trailer as described in the Table 28 of the
> 88E6060 datasheet.
> The current code incorrectly omits to check the third
From: Dmitry Vyukov
Date: Tue, 22 Sep 2015 10:51:52 +0200
> rhashtable_rehash_one() uses complex logic to update entry->next field,
> after INIT_RHT_NULLS_HEAD and NULLS_MARKER expansion:
>
> entry->next = 1 | ((base + off) << 1)
>
> This can be compiled along the lines of:
>
> entry->next = b
From: Tobias Klauser
Date: Tue, 22 Sep 2015 09:29:49 +0200
> Converts the ch9200 driver to use the module_usb_driver() macro which
> makes the code smaller and a bit simpler.
>
> Signed-off-by: Tobias Klauser
Applied, thanks.
--
To unsubscribe from this list: send the line "unsubscribe netdev"
From: Jesse Gross
Date: Mon, 21 Sep 2015 20:21:20 -0700
> When support for megaflows was introduced, OVS needed to start
> installing flows with a mask applied to them. Since masking is an
> expensive operation, OVS also had an optimization that would only
> take the parts of the flow keys that w
From: Eric Dumazet
Date: Mon, 21 Sep 2015 16:28:27 -0700
> From: Eric Dumazet
>
> Neal suggested to move sk_txhash init into tcp_create_openreq_child(),
> called both from IPv4 and IPv6.
>
> This opportunity was missed in commit 58d607d3e52f ("tcp: provide
> skb->hash to synack packets")
>
>
From: Michael Grzeschik
Date: Tue, 22 Sep 2015 11:01:41 +0200
> git://git.pengutronix.de/git/mgr/linux.git tags/arcnet-cleanup-v4.3-rc2
Please build this tree against net-next, thanks.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vg
From: Russell King
Date: Mon, 21 Sep 2015 21:42:59 +0100
> Commit 54d792f257c6 ("net: dsa: Centralise global and port setup
> code into mv88e6xxx.") merged in the 4.2 merge window broke the link
> speed forcing for the CPU port of Marvell DSA switches. The original
> code was:
>
> /* MA
On Tue, Sep 22, 2015 at 10:50:41AM +0200, Pablo Neira Ayuso wrote:
> On Tue, Sep 22, 2015 at 10:22:13AM +0300, Julian Anastasov wrote:
> [...]
> >
> > v2 looks good to me,
> >
> > Acked-by: Julian Anastasov
>
> Thanks a lot for reviewing Julian.
>
> Simon, please let me know how you want t
From: Eric Dumazet
After a good amount of debugging, I found bnx2x was byte swaping
the 40 bytes of rss_key.
If we byte swap the key, then bnx2x generates hashes matching
MSDN specs as documented in (Verifying the RSS Hash Calculation)
https://msdn.microsoft.com/en-us/library/windows/hardware/f
fw filter uses tp->root==NULL to check if it is the old method,
so it doesn't need allocation at all in this case. This patch
reverts the offending commit and adds some comments for old
method to make it obvious.
Fixes: 33f8b9ecdb15 ("net_sched: move tp->root allocation into fw_init()")
Reported-b
On Mon, Sep 21, 2015 at 8:21 PM, Jesse Gross wrote:
> When support for megaflows was introduced, OVS needed to start
> installing flows with a mask applied to them. Since masking is an
> expensive operation, OVS also had an optimization that would only
> take the parts of the flow keys that were c
From: Jacob Keller
This patch fixes a corner case issue with the PF/VF mailbox code.
Currently, fm10k_mbx_reset_work clears various state about the mailbox.
However, it does not clear the Tx FIFO head/tail pointers. We can't
simply clear these pointers as we unintentionally drop untransmitted
mes
From: Francois Romieu
The device probe method e1000_probe calls e1000_init_eeprom_params
itself so there's no reason to call it again from e1000_do_write_eeprom
or e1000_do_read_eeprom.
The sentence above assumes that e1000_init_eeprom_params is effective.
e1000_init_eeprom_params depends mostly
From: Don Skidmore
X550 has HW support for SCTP flow director filters SCTP mask. This
patch adds it like we do for UDP and TCP.
Signed-off-by: Donald C Skidmore
Tested-by: Krishneil Singh
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c | 10 ++
drivers/
This series contains updates to e1000, e1000e, igbvf, ixgbe, ixgbevf and
fm10k.
Jacob provides several updates for fm10k, which cleans up comments and
most notably a fix for a corner case issue with the PF/VF mailbox code.
The issue being fm10k_mbx_reset_work clears various states about the
mailbo
From: Don Skidmore
This patch is part of the future enablement of X550 SFP+ support. This
HW uses different SDP so the interrupts need to be set up accordingly.
Signed-off-by: Donald C Skidmore
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 12 +---
1
From: Mark Rustad
When an SFP not present error is returned by the reset_hw method,
accept it and go on, since an SFP can still be inserted. Previously
it was only accepted for 82598 devices.
Signed-off-by: Mark Rustad
Tested-by: Phil Schmitt
Signed-off-by: Jeff Kirsher
---
drivers/net/ether
From: Toshiaki Makita
Setting ndo_features_check to passthru_features_check allows the driver
to skip the check for multiple tagged TSO packets and enables stacked
VLAN TSO.
Tested with I217-LM.
Signed-off-by: Toshiaki Makita
Tested-by: Aaron Brown
Signed-off-by: Jeff Kirsher
---
drivers/net
From: Mark Rustad
Add support for the SFP insertion interrupt on X550EM devices with
SFPs.
Signed-off-by: Mark Rustad
Tested-by: Phil Schmitt
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/net/ethern
From: Toshiaki Makita
Setting ndo_features_check to passthru_features_check allows the driver
to skip the check for multiple tagged TSO packets and enables stacked
VLAN TSO.
Tested with I350.
Signed-off-by: Toshiaki Makita
Tested-by: Aaron Brown
Signed-off-by: Jeff Kirsher
---
drivers/net/et
From: Mark Rustad
Provide I2C combined operations on X550EM, not X550 devices.
Signed-off-by: Mark Rustad
Tested-by: Phil Schmitt
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel/ixgbe/ixgbe_x550.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/
From: Mark Rustad
Most I2C accesses take and release semaphores for each access. Now
there is a reason to perform multiple I2C operations under the same
holding of the semaphore, so provide unlocked I2C methods for that
purpose.
Signed-off-by: Mark Rustad
Tested-by: Phil Schmitt
Signed-off-by:
From: Jacob Keller
This comment is no longer true due to a couple of mailbox locking
refactors, and we now don't actually do any rtnl protected operations
directly in the mailbox path. Remove this comment as it is factually
incorrect and confusing.
Signed-off-by: Jacob Keller
Tested-by: Krishne
From: Toshiaki Makita
Setting ndo_features_check to passthru_features_check allows the driver
to skip the check for multiple tagged TSO packets and enables stacked
VLAN TSO.
Tested with 82599ES.
Signed-off-by: Toshiaki Makita
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel/ixgbevf/
From: Mark Rustad
The lan_id is being set after a previous I2C eeprom access which
makes no sense because it needs to be set before any access. Move
the setting to before the access.
Signed-off-by: Mark Rustad
Tested-by: Phil Schmitt
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel
From: Mark Rustad
Set the bit banging mode in the hardware when performing bit banging
I2C operations on X550. Also control the output enable on both the
clock and data lines.
Signed-off-by: Mark Rustad
Tested-by: Phil Schmitt
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel/ixgbe/
From: Jacob Keller
Signed-off-by: Jacob Keller
Tested-by: Krishneil Singh
Signed-off-by: Jeff Kirsher
---
drivers/net/ethernet/intel/fm10k/fm10k_mbx.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_mbx.c
b/drivers/net/ethernet/i
From: Jacob Keller
Add a private ethtool flag to enable display of these statistics, which
are generally less useful. However, sometimes it can be useful for
debugging purposes. The most useful portion is the ability to see what
the PF thinks the VF mailboxes look like.
Signed-off-by: Jacob Kell
From: Jacob Keller
When we connect to the mailbox, we insert a fake disconnect header so
that the code does not see an invalid header and thus instantly error
every time we bring up the mailbox. However, we incorrectly record the
tail and head from the local perspective. Since the remote end shou
From: "John W. Linville"
Date: Mon, 21 Sep 2015 10:29:09 -0400
> Partially due to a pre-exising "thinko", the new metadata-based tx/rx
> paths were handling ECN propagation differently than the traditional
> tx/rx paths. This patch removes the "thinko" (involving multiple
> ip_hdr assignments) o
From: David Woodhouse
Date: Mon, 21 Sep 2015 15:02:07 +0100
> From: David Woodhouse
>
> If an RX interrupt was already received but NAPI has not yet run when
> the RX timeout happens, we end up in cp_tx_timeout() with RX interrupts
> already disabled. Blindly re-enabling them will cause an IRQ
From: David Woodhouse
Date: Mon, 21 Sep 2015 15:01:49 +0100
> From: David Woodhouse
>
> The TX timeout handling has been observed to trigger RX IRQ storms. And
> since cp_interrupt() just keeps saying that it handled the interrupt,
> the machine then dies. Fix the return value from cp_interrupt
On Tue, 2015-09-22 at 16:06 -0700, David Miller wrote:
> From: David Miller
> Date: Tue, 22 Sep 2015 14:59:35 -0700 (PDT)
>
> > From: Eric Dumazet
> > Date: Fri, 18 Sep 2015 15:21:52 -0700
> >
> >> From: Eric Dumazet
> >>
> >> While trying to understand why bnx2x RSS hash was not matching
> >
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:59 +0200
> skb->len is always non-negative.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
>
> [1]: http://permalink.gmane.org/gmane.linux.kernel/2038576
>
> Signed
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:53 +0200
> To avoid underflows signed variables should be used in expression.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
>
> [1]: http://permalink.gmane.org/gman
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:54 +0200
> Thresholds uses -1 to indicate that default value should be used.
> Since thresholds are unsigned sign checking makes no sense.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_t
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:51 +0200
> Variable can store negative values.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
>
> [1]: http://permalink.gmane.org/gmane.linux.kernel/2038576
>
> Sig
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:55 +0200
> Thresholds uses -1 to indicate that default value should be used.
> Since thresholds are unsigned sign checking makes no sense.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_t
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:49 +0200
> Difference of unsigned values is also unsigned so it does not make
> sense to check its sign.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
>
> [1]: http
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:52 +0200
> Unsigned minus constant is still unsigned so checking its sign makes no
> sense.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
>
> [1]: http://permalink.
From: Andrzej Hajda
Date: Mon, 21 Sep 2015 15:33:50 +0200
> phy_mode can be negative.
>
> The problem has been detected using proposed semantic patch
> scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
>
> [1]: http://permalink.gmane.org/gmane.linux.kernel/2038576
>
> Signed-off-by
From: David Miller
Date: Tue, 22 Sep 2015 14:59:35 -0700 (PDT)
> From: Eric Dumazet
> Date: Fri, 18 Sep 2015 15:21:52 -0700
>
>> From: Eric Dumazet
>>
>> While trying to understand why bnx2x RSS hash was not matching
>> Toeplitz specifications, I implemented driver code to support
>> ethtool
e_inval has 1 explicit user and 1 fallthrough user -- martian_destination.
Move setting err to EINVAL before the 2 users and use the goto out label
instead of e_inval.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 7 +++
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/net/i
Move stat and logging for martian_destination error into helper function
similar to ip_handle_martian_source. Code move only; no functional change.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 24 +++-
1 file changed, 15 insertions(+), 9 deletions(-)
diff --git a/net/ip
err is initialized to -EINVAL when it is declared. It is not reset until
fib_lookup which is well after the 3 users of the martian_source jump. So
resetting err to -EINVAL at martian_source label is not needed.
Removing that line obviates the need for the martian_source_keep_err label
so delete it
Move the rth lookup and allocation from ip_route_input_slow into a
helper function. Code move only; no operational change intended.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 104 ---
1 file changed, 60 insertions(+), 44 deletions(-)
di
Move no_route code into helper. Add call to helper at jump sites and
remove goto label.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 40 ++--
1 file changed, 26 insertions(+), 14 deletions(-)
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 454c38b
Move call to ip_handle_martian_source to various jump sites. The
function only increments a counter and possibly does additional
logging if CONFIG_IP_ROUTE_VERBOSE is enabled.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 38 --
1 file changed, 20 insertio
e_nobufs has 1 user. Move setting err to -ENOBUFS for the 1 user and
use the goto out label instead of e_nobufs. Stepping stone patch; next
one moves rth code into a helper function.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 10 --
1 file changed, 4 insertions(+), 6 deletions(-)
Move call to ip_handle_martian_dest to various jump sites. The
function only increments a counter and possibly does additional
logging if CONFIG_IP_ROUTE_VERBOSE is enabled.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 24 +---
1 file changed, 13 insertions(+), 11 deleti
ip_route_input_slow is a maze of gotos (9 of them!) making it error
prone and difficult to read. This patchset refactors it, removing all
but 2 of the labels. The brd_input label for broadcast path requires
too many inputs to make a reasonble helper out of it so I left it as is.
None of these patc
Move calls to ip_route_local_input to jump sites and remove
local_input label.
Signed-off-by: David Ahern
---
net/ipv4/route.c | 15 +++
1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index efc8e58c1964..454c38bd23a8 100644
--- a/ne
On Tue, 22 Sep 2015 15:16:27 -0700
Joe Stringer wrote:
> On 22 September 2015 at 15:05, Steven Rostedt wrote:
> >
> > The attached config gives the following error:
> >
> > net/built-in.o: In function `handle_fragments':
> > net/openvswitch/conntrack.c:316: undefined reference to `nf_ct_frag6_ga
Signed-off-by: Guillaume Gomez
---
include/linux/clockchips.h | 2 +-
include/linux/crypto.h | 2 +-
include/linux/device.h | 2 +-
include/linux/netdevice.h | 2 +-
include/linux/percpu-refcount.h | 2 +-
include/linux/timekeeping.h | 2 +-
kernel/time/clocksourc
On 22 September 2015 at 15:05, Steven Rostedt wrote:
>
> The attached config gives the following error:
>
> net/built-in.o: In function `handle_fragments':
> net/openvswitch/conntrack.c:316: undefined reference to `nf_ct_frag6_gather'
> Makefile:916: recipe for target 'vmlinux' failed
>
> -- Steve
On Tue, 22 Sep 2015 16:01:36 -0400
Neil Horman wrote:
> + clear_bit(NAPI_STATE_NPSVC, &n->state);
> +
> }
why introduce extra line here?
> + /*
> + * If we set this bit but see that it has already been set,
> + * that indicates that napi has been disabled and we need
> +
The attached config gives the following error:
net/built-in.o: In function `handle_fragments':
net/openvswitch/conntrack.c:316: undefined reference to `nf_ct_frag6_gather'
Makefile:916: recipe for target 'vmlinux' failed
-- Steve
config.gz
Description: application/gzip
David Woodhouse :
[...]
> What do you think of finally enabling hw csum and TSO by default, btw?
Your router can actively use it and you'll have to keep the pieces
together if it breaks, right ? So I'd go for it.
--
Ueimor
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
From: Eric Dumazet
Date: Fri, 18 Sep 2015 15:21:52 -0700
> From: Eric Dumazet
>
> While trying to understand why bnx2x RSS hash was not matching
> Toeplitz specifications, I implemented driver code to support
> ethtool -x|-X to try various RSS key combinations easily.
>
> Then, after some debu
For some reason we were carrying the budget value around between the
various calls to napi->poll. If for example one of the drivers called had
a bug in which it returned a non-zero value for work this could result in
the budget value becoming negative.
Rather than carry around a value of budget t
This patch corrects an issue in which the polling routine would increase
the budget for Rx to at least 1 per queue if multiple queues were present.
This would result in Rx packets being processed when the budget was 0 which
is meant to indicate that no Rx can be handled.
Signed-off-by: Alexander D
While reviewing some other patches related to netpoll I started looking at
how the drivers were handling their budgets. It turns out that the ixgbe,
fm10k, and i40e drivers contained cases where they would increase the
budget from 0 and as a result it was possible for them to clean Rx frames
insid
This patch corrects an issue in which the polling routine would increase
the budget for Rx to at least 1 per queue if multiple queues were present.
This would result in Rx packets being processed when the budget was 0 which
is meant to indicate that no Rx can be handled.
Signed-off-by: Alexander D
The polling routine for i40e was rounding up the budget for Rx cleanup to
1. This is incorrect as netpoll will call is expecting no Rx to be
processed as the budget passed was 0.
Signed-off-by: Alexander Duyck
---
drivers/net/ethernet/intel/i40e/i40e_txrx.c |5 +
1 file changed, 5 inser
On Tue, Sep 22, 2015 at 12:57 PM, Pravin B Shelar wrote:
> VXLAN device can receive skb with checksum partial. But the checksum
> offset could be in outer header which is pulled on receive. This results
> in negative checksum offset for the skb. Such skb can cause the assert
> failure in skb_check
On Tue, Sep 22, 2015 at 1:36 PM, Bjørn Mork wrote:
>
> http://download.intel.com/design/archives/processors/pro/docs/24268935.pdf
>
> Says "NoFix" for erratas 66 and 92.
Yeah, 66 and 92 do look like they could cause the apparent ordering of
accesses to be violated. That said, both of them seem to
Linus Torvalds writes:
> The PPRO_FENCE_BUG thing is rather questionable. I'm not sure it's
> even documented, but I can't find the official PPro errata now. I
> think I discussed it with Andy Glew back when Intel was starting to
> document the memory ordering rules.
>
> I'd be willing to get rid
From: Pablo Neira Ayuso
Date: Tue, 22 Sep 2015 11:13:50 +0200
> The following patchset contains Netfilter/IPVS updates for your net-next tree
> in this 4.4 development cycle, they are:
...
Pulled, thanks Pablo.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of
On Tue, Sep 22, 2015 at 2:02 PM, Neal Cardwell wrote:
> More generally, my sense is that we should tweak the is_cwnd_limited code to
> shift from saying "set is_cwnd_limited to true iff the cwnd is known to be
> limiting transmits" to saying "set is_cwnd_limited to true iff the packets in
> flight
Drivers might call napi_disable while not holding the napi instance poll_lock.
In those instances, its possible for a race condition to exist between
poll_one_napi and napi_disable. That is to say, poll_one_napi only tests the
NAPI_STATE_SCHED bit to see if there is work to do during a poll, and a
On Tue, Sep 22, 2015 at 12:50 PM, Tejun Heo wrote:
>
> Hmmm? We're doing that. PPRO_FENCE makes both acquire and release
> use smp_mb().
Oh, right you are. I just grepped for rmb, and missed it because as
you say, it's a full mb.
The PPRO_FENCE_BUG thing is rather questionable. I'm not sure it
VXLAN device can receive skb with checksum partial. But the checksum
offset could be in outer header which is pulled on receive. This results
in negative checksum offset for the skb. Such skb can cause the assert
failure in skb_checksum_help(). Following patch fixes the bug by setting
checksum-none
On Tue, Sep 22, 2015 at 12:46:13PM -0700, Alexander Duyck wrote:
> On 09/22/2015 12:01 PM, Neil Horman wrote:
> >On Tue, Sep 22, 2015 at 02:00:23PM -0400, Neil Horman wrote:
> >>Drivers might call napi_disable while not holding the napi instance
> >>poll_lock.
> >>In those instances, its possible
Hello,
On Tue, Sep 22, 2015 at 12:28:45PM -0700, Linus Torvalds wrote:
> Both acquire and smp_rmb() are equally free on x86.
Was this always like this? Ah, okay, from 2007. Was thinking it's
still doing an actual rmb. Talk about being confused.
> It appears that we don't do the X86_PPRO_FENCE
On 09/22/2015 12:01 PM, Neil Horman wrote:
On Tue, Sep 22, 2015 at 02:00:23PM -0400, Neil Horman wrote:
Drivers might call napi_disable while not holding the napi instance poll_lock.
In those instances, its possible for a race condition to exist between
poll_one_napi and napi_disable. That is t
David Miller writes:
> From: Robert Jarzmik
> Date: Fri, 18 Sep 2015 18:36:56 +0200
>
>> Which brings me to wonder which is the more correct :
>> (a) replace to reproduce the same calculation
>> Previously mtt was compared to a difference of 76ns steps (as 307ns / 4
>> =
>> 76ns):
>>
On Tue, Sep 22, 2015 at 11:53 AM, Tejun Heo wrote:
>
> I see. The write path here is cold so the competition is between rmb
> and acquire. Unless some significant archs completely screwed it up,
> acquire still seems like the better option. It's essentially free on
> x86 after all.
Both acquir
On Tue, Sep 22, 2015 at 02:00:23PM -0400, Neil Horman wrote:
> Drivers might call napi_disable while not holding the napi instance poll_lock.
> In those instances, its possible for a race condition to exist between
> poll_one_napi and napi_disable. That is to say, poll_one_napi only tests the
> NA
Hello, Linus.
On Tue, Sep 22, 2015 at 11:42:33AM -0700, Linus Torvalds wrote:
...
> smp_rmb() should generally be about the same cost as an acquire. It
> can go either way.
>
> So *if* the algorithm is amenable to smp_wmb()/smp_rmb() kind of
> barriers, that's actually quite possibly better than
On Tue, Sep 22, 2015 at 9:10 AM, Tejun Heo wrote:
>
> That's a pentium pro era errata. Virtually no working machine is
> affected by that anymore and nobody builds kernel with that option.
> In most cases, store_release and load_acquire are cheaper as they're
> more specific. On x86, store_relea
1 - 100 of 246 matches
Mail list logo
