Increase the default TX ring size (from 512 to 1024) to match
the RX ring size.
This gives the XDP TX ring a better chance to keep up with the
rate of its RX ring in case of a high load of XDP_TX actions.
Tested:
Ethtool counter rx_xdp_tx_full used to increase, after applying this
patch it
Several small performance improvements in RX datapath,
including:
- Compiler branch predictor hints.
- Replace a multiplication with a shift operation.
- Minimize variables scope.
- Write-prefetch for packet header.
- Avoid trinary-operator ("?") when value can be preset in a matching
branch.
-
From: Saeed Mahameed
Avoid touching RX QP RSS context when loading with only
one RX ring, to allow optimized A0 RX steering.
Enable by:
- loading mlx4_core with module param: log_num_mgm_entry_size = -6.
- then: ethtool -L rx 1
Performance tests:
Tested on ConnectX3Pro,
Instead of having their own NAPIs, XDP TX completion queues get
polled within the corresponding RX NAPI.
This prevents any possible race on TX ring prod/cons indices,
between the context that issues the transmits (RX NAPI) and the
context that handles the completions (was previously done in
a
Several small performance improvements in TX CQ polling,
including:
- Compiler branch predictor hints.
- Minimize variables scope.
- More proper check of cq type.
- Use boolean instead of int for a binary indication.
Performance tests:
Tested on ConnectX3Pro, Intel(R) Xeon(R) CPU E5-2680 v3 @
Hi Dave,
This patchset contains data-path improvements, mainly for XDP_DROP
and XDP_TX cases.
Main patches:
* Patch 2 by Saeed allows enabling optimized A0 RX steering (in HW) when
setting a single RX ring.
With this configuration, HW packet-rate dramatically improves,
reaching 28.1 Mpps
On 15/06/17 00:51, Julien Gomes wrote:
> Hi Nikolay,
>
> On 06/14/2017 05:04 AM, Nikolay Aleksandrov wrote:
>
>> This has been on our todo list and I'm definitely interested in the
>> implementation.
>> A few things that need careful consideration from my POV. First are the
>> security
>>
I would need your partnership in a transaction and details will
disclose to you once i receive your reply.
Thanks
Saif.
> -Original Message-
> From: netdev-ow...@vger.kernel.org [mailto:netdev-ow...@vger.kernel.org]
> On Behalf Of Johannes Berg
> Sent: Thursday, June 15, 2017 5:28 PM
> To: netdev@vger.kernel.org
> Cc: Johannes Berg
> Subject: [RFC v2] networking: convert many more places to skb_put_zero()
In sctp_for_each_transport, pos is used to save how many objs it has
dumped. Now it gets the last obj by sctp_transport_get_idx, then gets
the next obj by sctp_transport_get_next.
The issue is that in the meanwhile if some objs in transport hashtable
are removed and the objs nums are less than
On Tue, Jun 13, 2017 at 10:56:08AM -0400, Jon Mason wrote:
> The of_mdio_parse_addr() helper function is useful to other code, but
> the module dependency chain causes issues. To work around this, we can
> move of_mdio_parse_addr() to be an inline function in the header file.
> This gets rid of
From: Johannes Berg
There were many places that my previous spatch didn't find,
as pointed out by yuan linyu in various patches.
The following spatch found many more and also removes the
now unnecessary casts:
@@
identifier p, p2;
expression len;
debugfs_remove already check mwifiex_dfs_dir, so remove it.
Signed-off-by: Shawn Lin
---
drivers/net/wireless/marvell/mwifiex/debugfs.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/net/wireless/marvell/mwifiex/debugfs.c
Now when starting the dad work in addrconf_mod_dad_work, if the dad work
is idle and queued, it needs to hold ifa.
The problem is there's one gap in [1], during which if the pending dad work
is removed elsewhere. It will miss to hold ifa, but the dad word is still
idea and queue.
if
On Wed, 14 Jun 2017 13:38:07 -0700, Stephen Hemminger wrote:
> Does this change the default? Before your patches what was the checksum
> setting for the new tunnel.
Yes, it does. See the kernel patches.
I realize it's a user visible change; however, given that the real
reason act_tunnel_key was
Ok, understand
> -Original Message-
> From: Johannes Berg [mailto:johan...@sipsolutions.net]
> Sent: Thursday, June 15, 2017 3:45 PM
> To: YUAN Linyu; netdev@vger.kernel.org
> Subject: Re: [RFC] networking: convert many more places to skb_put_zero()
>
> On Thu, 2017-06-15 at 07:20
On Thu, 2017-06-15 at 07:20 +, YUAN Linyu wrote:
> >
> In my opinion if spatch can do it even it found one place, keep it.
> Only leave difficult places like ndisc.c to me.
It's not so simple - I'd have to tailor the spatch to it pretty much I
guess, spending far more time on the spatch than
Le 14/06/2017 à 22:31, Stephen Hemminger a écrit :
> On Wed, 14 Jun 2017 18:45:42 +0200
> Nicolas Dichtel wrote:
>
>> This option is documented in gre6 help, but was not supported.
>>
>> Fixes: af89576d7a8c ("iproute2: GRE over IPv6 tunnel support.")
>> Signed-off-by:
> -Original Message-
> From: Johannes Berg [mailto:johan...@sipsolutions.net]
> Sent: Thursday, June 15, 2017 3:12 PM
> To: YUAN Linyu; netdev@vger.kernel.org
> Subject: Re: [RFC] networking: convert many more places to skb_put_zero()
>
> On Thu, 2017-06-15 at 07:05 +, YUAN Linyu
Count types of operation done by HW.
Signed-off-by: Harsh Jain
---
drivers/crypto/chelsio/chcr_algo.c | 16 +-
drivers/crypto/chelsio/chcr_core.c | 2 ++
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 1 +
GCM and CBC mode of operation requires Last Cipher Block.
This patch set lcb bit in WR header when required.
Signed-off-by: Harsh Jain
---
drivers/crypto/chelsio/chcr_algo.c | 18 +++---
drivers/crypto/chelsio/chcr_algo.h | 4 ++--
2 files changed, 13
This series is based on cryptodev2.6 tree and includes bug fix ,ctr(aes),
rfc3686(ctr(aes)) algo.
Harsh Jain (7):
crypto: chcr - Pass lcb bit setting to firmware
crypto: chcr - Set fallback key
crypto: chcr - Return correct error code
crypto: chcr - Avoid changing request structure
It send multiple WRs to H/W to handle large sg lists. Adds ctr(aes)
and rfc(ctr(aes)) modes.
Signed-off-by: Harsh Jain
---
drivers/crypto/chelsio/chcr_algo.c | 786 ---
drivers/crypto/chelsio/chcr_algo.h | 26 +-
Do not update assoclen received in aead_request.
Signed-off-by: Harsh Jain
---
drivers/crypto/chelsio/chcr_algo.c | 37 ++---
1 file changed, 14 insertions(+), 23 deletions(-)
diff --git a/drivers/crypto/chelsio/chcr_algo.c
Allocate new sg list in case received destination sg list has entry
greater that 2k.
Signed-off-by: Harsh Jain
---
drivers/crypto/chelsio/chcr_algo.c | 153 +++
drivers/crypto/chelsio/chcr_crypto.h | 6 ++
2 files changed, 142
Return correct error instead of EINVAL.
Signed-off-by: Harsh Jain
---
drivers/crypto/chelsio/chcr_algo.c | 76 +-
1 file changed, 42 insertions(+), 34 deletions(-)
diff --git a/drivers/crypto/chelsio/chcr_algo.c
Add myself as maintainer for chcr.
Signed-off-by: Harsh Jain
---
MAINTAINERS | 7 +++
1 file changed, 7 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index 1f20176..504dc65 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3706,6 +3706,13 @@ S: Supported
F:
When multiple devices are present in system select device
in round-robin fashion for crypto operations
Signed-off-by: Atul Gupta
Reviewed-by: Ganesh Goudar
---
drivers/crypto/chelsio/chcr_algo.c | 8 ++--
Set key of fallback tfm for rfc4309.
Signed-off-by: Harsh Jain
---
drivers/crypto/chelsio/chcr_algo.c | 12 +++-
1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/drivers/crypto/chelsio/chcr_algo.c
b/drivers/crypto/chelsio/chcr_algo.c
index
On Thu, 2017-06-15 at 07:05 +, YUAN Linyu wrote:
> > @@
> > type t;
> > expression skb, len;
> > identifier p;
> > @@
> > t *p
> > - = skb_put(skb, len);
> > + = skb_put_zero(skb, len);
> > -memset(p, 0, len);
> >
> > and it can't figure out that it should remove the variable, without
> >
> -Original Message-
> From: Johannes Berg [mailto:johan...@sipsolutions.net]
> Sent: Thursday, June 15, 2017 2:58 PM
> To: YUAN Linyu; netdev@vger.kernel.org
> Subject: Re: [RFC] networking: convert many more places to skb_put_zero()
>
> On Thu, 2017-06-15 at 00:23 +, YUAN Linyu
On Wed, Jun 14, 2017 at 11:15 PM, Arnd Bergmann wrote:
> We get a lot of very large stack frames using gcc-7.0.1 with the default
> -fsanitize-address-use-after-scope --param asan-stack=1 options, which
> can easily cause an overflow of the kernel stack, e.g.
>
>
On Wed, 2017-06-14 at 23:45 +, YUAN Linyu wrote:
> >
> > But still a few more, I think that you didn't find, e.g.
> > in mac80211.
>
> Yes, I thought you already done in mac80211
So did I, but I also only applied my spatch :)
johannes
On Thu, 2017-06-15 at 00:23 +, YUAN Linyu wrote:
> Hi,
> Indeed, it find more.
> Compare with my patch, still lost pattern like below,
> 1. sctp and openvswitch
> --- a/net/sctp/output.c
> +++ b/net/sctp/output.c
> @@ -463,7 +463,7 @@ static int sctp_packet_pack(struct sctp_packet
> *packet,
Hi Andrew,
On Wed, Jun 14, 2017 at 06:35:37PM +0200, Andrew Lunn wrote:
> On Wed, Jun 14, 2017 at 05:49:10PM +0200, Antoine Tenart wrote:
> >
> > The Ethernet controllers of the Marvel Kirkwood, Dove, Orion5x,
> > -MV78xx0, Armada 370 and Armada XP have an identical unit that provides
> > -an
These patches are used to support new chips.
Hayes Wang (3):
r8152: support new chip 8050
r8152: support RTL8153B
r8152: add byte_enable for ocp_read_word function
drivers/net/usb/r8152.c | 687 ++--
1 file changed, 671 insertions(+), 16
The settings of the new chip are the same with RTL8152, except that
its product ID is 0x8050.
Signed-off-by: Hayes Wang
---
drivers/net/usb/r8152.c | 10 ++
1 file changed, 10 insertions(+)
diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c
index
Add byte_enable for ocp_read_word() to replace reading 4
bytes data with reading the desired 2 bytes data.
This is used to avoid the issue which is described in
commit b4d99def0938 ("r8152: remove sram_read"). The
original method always reads 4 bytes data, and it may
have problem when reading the
This patch supports two new chips for RTL8153B.
Signed-off-by: Hayes Wang
---
drivers/net/usb/r8152.c | 673 ++--
1 file changed, 658 insertions(+), 15 deletions(-)
diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c
From: Pablo Cascón
Add general use per-vNIC mailbox area and use it for VLAN filtering
support. Initially proto is hardcoded to 802.1q.
Signed-off-by: Pablo Cascón
Signed-off-by: Jakub Kicinski
---
201 - 240 of 240 matches
Mail list logo