On 2021-03-23 19:56, Jay Vosburgh wrote:
Maxim Mikityanskiy wrote:
On 2021-03-22 16:09, Leon Romanovsky wrote:
On Mon, Mar 22, 2021 at 02:38:46PM +0200, Maxim Mikityanskiy wrote:
After lockdep gets triggered for the first time, it gets disabled, and
lockdep_enabled() will return false. It
On 2021-03-22 16:09, Leon Romanovsky wrote:
On Mon, Mar 22, 2021 at 02:38:46PM +0200, Maxim Mikityanskiy wrote:
After lockdep gets triggered for the first time, it gets disabled, and
lockdep_enabled() will return false. It will affect lockdep_is_held(),
which will start returning true all the
On 2021-03-22 16:21, Lv Yunlong wrote:
My static analyzer tool reported a potential uaf in
mlx5e_ktls_del_rx. In this function, if the condition
cancel_work_sync(&resync->work) is true, and then
priv_rx could be freed. But priv_rx is used later.
I'm unfamiliar with how this function works. Maybe
t use
xmit_hash")
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
While this patch works around the issue, I would like to discuss better
options. Another straightforward approach is to extend lockdep API with
lockdep_is_not_held(), which will be basically !lockdep_is_held() wh
On 2021-03-15 10:38, Antoine Tenart wrote:
Quoting Saeed Mahameed (2021-03-12 21:54:18)
On Fri, 2021-03-12 at 16:04 +0100, Antoine Tenart wrote:
netif_set_xps_queue must be called with the rtnl lock taken, and this
is
now enforced using ASSERT_RTNL(). mlx5e_attach_netdev was taking the
lock con
Hierarchical QoS hardware offload")
Suggested-by: Eric Dumazet
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
net/sched/sch_htb.c | 15 +--
1 file changed, 9 insertions(+), 6 deletions(-)
diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
index b23203159996
The HTB offload feature introduced a few bugs in HTB. One affects the
non-offload mode, preventing attaching qdiscs to HTB classes, and the
other affects the error flow, when the netdev doesn't support the
offload, but it was requested. This short series fixes them.
Maxim Mikityansk
pspotmail.com
Fixes: d03b195b5aa0 ("sch_htb: Hierarchical QoS hardware offload")
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
net/sched/sch_htb.c | 4
1 file changed, 4 insertions(+)
diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
index dff3adf5a915..b23203
On 2021-03-10 19:03, Eric Dumazet wrote:
On 3/10/21 3:54 PM, Maxim Mikityanskiy wrote:
On 2021-03-09 17:20, Eric Dumazet wrote:
On 3/9/21 4:13 PM, syzbot wrote:
Hello,
syzbot found the following issue on:
HEAD commit: 38b5133a octeontx2-pf: Fix otx2_get_fecparam()
git tree: net
: Maxim Mikityanskiy
Date: Tue Jan 19 12:08:13 2021 +
sch_htb: Hierarchical QoS hardware offload
bisection log: https://syzkaller.appspot.com/x/bisect.txt?x=13ab12ecd0
final oops: https://syzkaller.appspot.com/x/report.txt?x=106b12ecd0
console output: https
uot;,"handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo&qu
On 2021-01-29 18:05, David Ahern wrote:
On 12/15/20 12:42 AM, Maxim Mikityanskiy wrote:
This commit adds support for configuring HTB in offload mode. HTB
offload eliminates the single qdisc lock in the datapath and offloads
the algorithm to the NIC. The new 'offload' parameter i
lasses are created as usual, but filters should be moved to clsact for
lock-free classification (filters attached to HTB itself are not
supported in the offload mode):
# tc filter add dev eth0 egress protocol ip flower dst_port 80
action skbedit priority 1:10
Signed-off-by: Maxim Mikityanskiy
R
On 2021-01-19 17:50, Björn Töpel wrote:
This series extends bind() for XDP sockets, so that the bound socket
is added to the netdev_rx_queue _rx array in the netdevice. We call
this to register the socket. To redirect packets to the registered
socket, a new BPF helper is used: bpf_redirect_xsk().
This commit adds support for statistics of offloaded HTB. Bytes and
packets counters for leaf and inner nodes are supported, the values are
taken from per-queue qdiscs, and the numbers that the user sees should
have the same behavior as the software (non-offloaded) HTB.
Signed-off-by: Maxim
t enabled by default, a new parameter is used to enable it:
# tc qdisc replace dev eth0 root handle 1: htb offload
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
include/linux/netdevice.h| 1 +
include/net/pkt_cls.h| 36 ++
include/ua
), and * is the counter name
(the counters are the same as for the normal SQs).
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
.../net/ethernet/mellanox/mlx5/core/Makefile | 6 +-
drivers/net/ethernet/mellanox/mlx5/core/en.h | 27 +-
.../ethernet/mellanox/mlx5/core/en/params.h
In a following commit, sch_htb will start using extack in the delete
class operation to pass hardware errors in offload mode. This commit
prepares for that by adding the extack parameter to this callback and
converting usage of the existing qdiscs.
Signed-off-by: Maxim Mikityanskiy
Reviewed-by
explanations of how it works in the commit message. Remove %hu
from format strings. Add resiliency when leaf_del_last fails to create a
new leaf node.
Maxim Mikityanskiy (5):
net: sched: Add multi-queue support to sch_tree_lock
net: sched: Add extack to Qdisc_class_ops.delete
sch_htb: Hierarchical
qdisc itself if it's the MQROOT.
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
include/net/sch_generic.h | 14 ++
1 file changed, 10 insertions(+), 4 deletions(-)
diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 639e465a108f..9448e8cf1ee6
d
TX queues, regardless of real_num_*_queues, still taking into
consideration num_*_queues to avoid overflows.
Fixes: 1c1efc2af158 ("xsk: Create and free buffer pool independently from umem")
Fixes: a41b4f3c58dd ("xsk: simplify xdp_clear_umem_at_qid implementation")
Signed-off-by: M
On 2020-12-22 03:17, Jakub Kicinski wrote:
On Tue, 15 Dec 2020 09:42:11 +0200 Maxim Mikityanskiy wrote:
+ q->offload = nla_get_flag(tb[TCA_HTB_OFFLOAD]);
+
+ if (q->offload) {
+ if (!tc_can_offload(dev) || !dev->netdev_ops->ndo_setup_tc)
+
On 2020-12-17 17:09, Jamal Hadi Salim wrote:
On 2020-12-16 6:47 a.m., Maxim Mikityanskiy wrote:
On 2020-12-15 18:37, Jamal Hadi Salim wrote:
[..]
Same question above:
Is there a limit to the number of classes that can be created?
Yes, the commit message of the mlx5 patch lists the
On 2020-12-16 21:01, Cong Wang wrote:
On Mon, Dec 14, 2020 at 12:30 PM Maxim Mikityanskiy wrote:
On 2020-12-14 21:35, Cong Wang wrote:
On Mon, Dec 14, 2020 at 7:13 AM Maxim Mikityanskiy wrote:
On 2020-12-11 21:16, Cong Wang wrote:
On Fri, Dec 11, 2020 at 7:26 AM Maxim Mikityanskiy wrote
On 2020-12-15 18:37, Jamal Hadi Salim wrote:
On 2020-12-14 3:30 p.m., Maxim Mikityanskiy wrote:
On 2020-12-14 21:35, Cong Wang wrote:
On Mon, Dec 14, 2020 at 7:13 AM Maxim Mikityanskiy
wrote:
On 2020-12-11 21:16, Cong Wang wrote:
On Fri, Dec 11, 2020 at 7:26 AM Maxim Mikityanskiy
wrote
ted, the
granularity of rate and ceil parameters may be different, etc. - so, the
offload is not enabled by default, a new parameter is used to enable it:
# tc qdisc replace dev eth0 root handle 1: htb offload
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
inclu
), and * is the counter name
(the counters are the same as for the normal SQs).
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
.../net/ethernet/mellanox/mlx5/core/Makefile | 6 +-
drivers/net/ethernet/mellanox/mlx5/core/en.h | 27 +-
.../ethernet/mellanox/mlx5/core/en/params.h
This commit adds support for statistics of offloaded HTB. Bytes and
packets counters for leaf and inner nodes are supported, the values are
taken from per-queue qdiscs, and the numbers that the user sees should
have the same behavior as the software (non-offloaded) HTB.
Signed-off-by: Maxim
qdisc itself if it's the MQROOT.
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
include/net/sch_generic.h | 14 ++
1 file changed, 10 insertions(+), 4 deletions(-)
diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 162ed6249e8b..cfac2b0b5cc7
to 80 chars per
line.
v3 changes:
Fixed the CI failure on parisc with 16-bit xchg by replacing it with
WRITE_ONCE. Fixed the capability bits in mlx5_ifc.h and the value of
MLX5E_QOS_MAX_LEAF_NODES.
Maxim Mikityanskiy (4):
net: sched: Add multi-queue support to sch_tree_lock
sch_htb
lasses are created as usual, but filters should be moved to clsact for
lock-free classification (filters attached to HTB itself are not
supported in the offload mode):
# tc filter add dev eth0 egress protocol ip flower dst_port 80
action skbedit priority 1:10
Signed-off-by: Maxim Mi
On 2020-12-14 21:35, Cong Wang wrote:
On Mon, Dec 14, 2020 at 7:13 AM Maxim Mikityanskiy wrote:
On 2020-12-11 21:16, Cong Wang wrote:
On Fri, Dec 11, 2020 at 7:26 AM Maxim Mikityanskiy wrote:
HTB doesn't scale well because of contention on a single lock, and it
also consumes CPU.
On 2020-12-11 21:16, Cong Wang wrote:
On Fri, Dec 11, 2020 at 7:26 AM Maxim Mikityanskiy wrote:
HTB doesn't scale well because of contention on a single lock, and it
also consumes CPU. This patch adds support for offloading HTB to
hardware that supports hierarchical rate limiting.
), and * is the counter name
(the counters are the same as for the normal SQs).
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
.../net/ethernet/mellanox/mlx5/core/Makefile | 6 +-
drivers/net/ethernet/mellanox/mlx5/core/en.h | 27 +-
.../ethernet/mellanox/mlx5/core/en/params.h
to 80 chars per
line.
Maxim Mikityanskiy (4):
net: sched: Add multi-queue support to sch_tree_lock
sch_htb: Hierarchical QoS hardware offload
sch_htb: Stats for offloaded HTB
net/mlx5e: Support HTB offload
.../net/ethernet/mellanox/mlx5/core/Makefile | 6 +-
drivers/net/ethernet
This commit adds support for statistics of offloaded HTB. Bytes and
packets counters for leaf and inner nodes are supported, the values are
taken from per-queue qdiscs, and the numbers that the user sees should
have the same behavior as the software (non-offloaded) HTB.
Signed-off-by: Maxim
ted, the
granularity of rate and ceil parameters may be different, etc. - so, the
offload is not enabled by default, a new parameter is used to enable it:
# tc qdisc replace dev eth0 root handle 1: htb offload
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
inclu
qdisc itself if it's the MQROOT.
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
include/net/sch_generic.h | 14 ++
1 file changed, 10 insertions(+), 4 deletions(-)
diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 162ed6249e8b..cfac2b0b5cc7
On 2020-12-11 10:41, Dan Carpenter wrote:
On Thu, Dec 10, 2020 at 05:07:28PM +0200, Maxim Mikityanskiy wrote:
On 2020-12-10 10:28, Dan Carpenter wrote:
Hi Maxim,
url:
https://github.com/0day-ci/linux/commits/Maxim-Mikityanskiy/HTB-offload/20201210-000703
base:
https://git.kernel.org/pub
On 2020-12-10 10:28, Dan Carpenter wrote:
Hi Maxim,
url:
https://github.com/0day-ci/linux/commits/Maxim-Mikityanskiy/HTB-offload/20201210-000703
base: https://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next.git
afae3cc2da100ead3cd6ef4bb1fb8bc9d4b817c5
config: i386-randconfig-m021
This commit adds support for statistics of offloaded HTB. Bytes and
packets counters for leaf and inner nodes are supported, the values are
taken from per-queue qdiscs, and the numbers that the user sees should
have the same behavior as the software (non-offloaded) HTB.
Signed-off-by: Maxim
), and * is the counter name
(the counters are the same as for the normal SQs).
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
.../net/ethernet/mellanox/mlx5/core/Makefile | 6 +-
drivers/net/ethernet/mellanox/mlx5/core/en.h | 27 +-
.../ethernet/mellanox/mlx5/core/en/params.h
ted, the
granularity of rate and ceil parameters may be different, etc. - so, the
offload is not enabled by default, a new parameter is used to enable it:
# tc qdisc replace dev eth0 root handle 1: htb offload
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
inclu
14.2 times in the UDP
multi-stream test, compared to software HTB implementation (more details
in the mlx5 patch).
[1]: https://www.spinics.net/lists/netdev/msg628422.html
[2]: https://www.spinics.net/lists/netdev/msg663548.html
Maxim Mikityanskiy (4):
net: sched: Add multi-queue support to
qdisc itself if it's the MQROOT.
Signed-off-by: Maxim Mikityanskiy
Reviewed-by: Tariq Toukan
---
include/net/sch_generic.h | 14 ++
1 file changed, 10 insertions(+), 4 deletions(-)
diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 162ed6249e8b..cfac2b0b5cc7
program that drops the buffer.
This change simply puts the handling of MEM_TYPE_XSK_BUFF_POOL back
into xdp_return_buff().
Reported-by: Maxim Mikityanskiy
Fixes: 82c41671ca4f ("xdp: Simplify xdp_return_{frame, frame_rx_napi, buff}")
Signed-off-by: Björn Töpel
Thanks for addre
On 2020-11-05 00:59, Jakub Kicinski wrote:
On Tue, 3 Nov 2020 11:18:25 -0800 Saeed Mahameed wrote:
From: Maxim Mikityanskiy
On resync, the driver calls inet_lookup_established
(__inet6_lookup_established) that increases sk_refcnt of the socket. To
decrease it, the driver set skb->destruc
On 2020-09-04 16:59, Björn Töpel wrote:
On 2020-09-04 15:53, Björn Töpel wrote:
This series addresses a problem that arises when AF_XDP zero-copy is
enabled, and the kernel softirq Rx processing and userland process
is running on the same core.
[...]
@Maxim I'm not well versed in Mellanox
On 2020-09-04 18:05, Willem de Bruijn wrote:
On Thu, Sep 3, 2020 at 11:01 PM Saeed Mahameed wrote:
From: Maxim Mikityanskiy
A constant for the number of DS in an empty WQE (i.e. a WQE without data
segments) is needed in multiple places (normal TX data path, MPWQE in
XDP), but currently we
On 2020-09-04 18:27, Willem de Bruijn wrote:
On Thu, Sep 3, 2020 at 11:00 PM Saeed Mahameed wrote:
From: Maxim Mikityanskiy
A huge function mlx5e_sq_xmit was split into several to achieve multiple
goals:
1. Reuse the code in IPoIB.
2. Better intergrate with TLS, IPSEC, GENEVE and checksum
On 2020-09-04 01:46, Jakub Kicinski wrote:
On Thu, 3 Sep 2020 14:00:18 -0700 Saeed Mahameed wrote:
+static inline void mlx5e_tx_wi_consume_fifo_skbs(struct mlx5e_txqsq *sq,
+struct mlx5e_tx_wqe_info *wi,
+
On 2020-09-04 18:06, Willem de Bruijn wrote:
On Thu, Sep 3, 2020 at 11:01 PM Saeed Mahameed wrote:
From: Maxim Mikityanskiy
mlx5e_txwqe_complete performs some actions that can be taken to separate
functions:
1. Update the flags needed for hardware timestamping.
2. Stop the TX queue if
d all bits got to wanted.
Signed-off-by: Maxim Mikityanskiy
---
netlink/features.c | 11 ++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/netlink/features.c b/netlink/features.c
index 133529d..7622594 100644
--- a/netlink/features.c
+++ b/netlink/features.c
@@ -243,6 +
-netlink always return a reply when requested,
and it still avoids unnecessary calls to __netdev_update_features if the
wanted features haven't changed.
Fixes: 0980bfcd6954 ("ethtool: set netdev features with FEATURES_SET request")
Signed-off-by: Maxim Mikityanskiy
---
net/ethtool/featu
tive.
Fixes: 0980bfcd6954 ("ethtool: set netdev features with FEATURES_SET request")
Signed-off-by: Maxim Mikityanskiy
---
net/ethtool/features.c | 11 +++
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/net/ethtool/features.c b/net/ethtool/features.c
index 4e632dc
-tx-offload: on [fixed]
Fitler out dev->hw_features from req_wanted to fix it and to resemble
the legacy ethtool behavior.
Fixes: 0980bfcd6954 ("ethtool: set netdev features with FEATURES_SET request")
Signed-off-by: Maxim Mikityanskiy
---
net/ethtool/features.c | 3 ++-
1
userspace tool.
v2 changes: Added Fixes tags.
Maxim Mikityanskiy (3):
ethtool: Fix preserving of wanted feature bits in netlink interface
ethtool: Account for hw_features in netlink interface
ethtool: Don't omit the netlink reply if no features were changed
net/ethtool/features.c
d all bits got to wanted.
Signed-off-by: Maxim Mikityanskiy
---
netlink/features.c | 25 ++---
1 file changed, 18 insertions(+), 7 deletions(-)
diff --git a/netlink/features.c b/netlink/features.c
index 133529d..4f63fa2 100644
--- a/netlink/features.c
+++ b/netlink/featu
This series contains fixes for ethtool-netlink to make the behavior
similar to the legacy ethtool and avoid regressions caused by
differences in behavior between the old and the new tool.
Please also see the sibling series for the kernel.
Maxim Mikityanskiy (2):
netlink: Fix the condition for
from active_mask to avoid
reacting on bit changes that we asked for. The new condition now
matches the ifs in the loop above that print "[requested on/off]" and
"[not requested]".
Signed-off-by: Maxim Mikityanskiy
---
netlink/features.c | 2 +-
1 file changed, 1 insertion
-netlink always return a reply when requested,
and it still avoids unnecessary calls to __netdev_update_features if the
wanted features haven't changed.
Signed-off-by: Maxim Mikityanskiy
---
net/ethtool/features.c | 11 ---
1 file changed, 4 insertions(+), 7 deletions(-)
diff --git a/net/
is commit fixes it by changing the formula to:
(req_wanted & req_mask) | (old_wanted & ~req_mask),
where old_active was replaced by old_wanted to account for the wanted
bits.
The shortcut condition for the case where nothing was changed now
compares wanted bitmasks, instead of wanted to act
-tx-offload: on [fixed]
Fitler out dev->hw_features from req_wanted to fix it and to resemble
the legacy ethtool behavior.
Signed-off-by: Maxim Mikityanskiy
---
net/ethtool/features.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/net/ethtool/features.c b/net/e
userspace tool.
Maxim Mikityanskiy (3):
ethtool: Fix preserving of wanted feature bits in netlink interface
ethtool: Account for hw_features in netlink interface
ethtool: Don't omit the netlink reply if no features were changed
net/ethtool/features.c | 19 ++-
1 file change
On 2020-07-08 19:38, YU, Xiangning wrote:
Lockless Token Bucket (LTB) is a qdisc implementation that controls the
use of outbound bandwidth on a shared link. With the help of lockless
qdisc, and by decoupling rate limiting and bandwidth sharing, LTB is
designed to scale in the cloud data centers.
On 2020-07-21 08:03, Magnus Karlsson wrote:
Move queue_id, dev, and need_wakeup from the umem to the
buffer pool. This so that we in a later commit can share the umem
between multiple HW queues. There is one buffer pool per dev and
queue id, so these variables should belong to the buffer pool, no
On 2020-07-21 08:04, Magnus Karlsson wrote:
Enable the sharing of dma mappings by moving them out from the buffer
pool. Instead we put each dma mapped umem region in a list in the umem
structure. If dma has already been mapped for this umem and device, it
is not mapped again and the existing dma
On 2020-07-10 17:16, Magnus Karlsson wrote:
Move the fill and completion rings from the umem to the buffer
pool. This so that we in a later commit can share the umem
between multiple HW queue ids. In this case, we need one fill and
completion ring per queue id. As the buffer pool is per queue id
On 2020-07-08 09:44, Cong Wang wrote:
On Fri, Jun 26, 2020 at 3:46 AM Maxim Mikityanskiy wrote:
HTB doesn't scale well because of contention on a single lock, and it
also consumes CPU. Mellanox hardware supports hierarchical rate limiting
that can be leveraged by offloading the function
On 2020-07-10 17:16, Magnus Karlsson wrote:
Add support to share a umem between different devices. This mode
can be invoked with the XDP_SHARED_UMEM bind flag. Previously,
sharing was only supported within the same device. Note that when
sharing a umem between devices, just as in the case of shar
On 2020-07-02 15:19, Magnus Karlsson wrote:
Add support to share a umem between different devices. This mode
can be invoked with the XDP_SHARED_UMEM bind flag. Previously,
sharing was only supported within the same device. Note that when
sharing a umem between devices, just as in the case of shar
Signed-off-by: Maxim Mikityanskiy
---
Magnus, here are my minor changes to the mlx5 part (only mlx5 is
included in these two patches). I renamed pool to xsk_pool when used
outside of en/xsk and renamed umem.{c,h} to pool.{c,h}.
.../net/ethernet/mellanox/mlx5/core/Makefile | 2 +-
drivers/net
-off-by: Maxim Mikityanskiy
---
drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c | 4 ++--
.../net/ethernet/mellanox/mlx5/core/en/xsk/pool.c| 12 ++--
drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h | 8
drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.c | 10
On 2020-07-02 15:18, Magnus Karlsson wrote:
This patch set adds support to share a umem between AF_XDP sockets
bound to different queue ids on the same device or even between
devices. It has already been possible to do this by registering the
umem multiple times, but this wastes a lot of memory.
On 2020-07-02 15:19, Magnus Karlsson wrote:
Replace the explicit umem reference passed to the driver in
AF_XDP zero-copy mode with the buffer pool instead. This in
preparation for extending the functionality of the zero-copy mode
so that umems can be shared between queues on the same netdev and
a
On 2020-07-02 15:19, Magnus Karlsson wrote:
Create and free the buffer pool independently from the umem. Move
these operations that are performed on the buffer pool from the
umem create and destroy functions to new create and destroy
functions just for the buffer pool. This so that in later commi
Comments are very welcome before I move to the full implementation.
On 2020-06-26 13:46, Maxim Mikityanskiy wrote:
This patch is a follow-up for the RFC posted earlier [1]. You can find a
detailed description of the motivation there, and this patch is RFC code
that shows possible implementation
On 2020-07-06 21:39, Daniel Borkmann wrote:
On 7/2/20 2:18 PM, Magnus Karlsson wrote:
This patch set adds support to share a umem between AF_XDP sockets
bound to different queue ids on the same device or even between
devices. It has already been possible to do this by registering the
umem multip
Thanks a lot for your reply! It was really helpful. I have a few
comments, please see below.
On 2020-06-24 23:21, Samudrala, Sridhar wrote:
>
>
> On 6/17/2020 6:15 AM, Maxim Mikityanskiy wrote:
>> Hi,
>>
>> I discovered Intel ADQ feature [1] that allows to boost
eciated.
Thanks,
Max
[1]: https://www.spinics.net/lists/netdev/msg628422.html
Signed-off-by: Maxim Mikityanskiy
---
drivers/net/ethernet/mellanox/mlx5/core/en.h | 1 +
.../net/ethernet/mellanox/mlx5/core/en_main.c | 38 ++-
.../net/ethernet/mellanox/mlx5/core/en_tx.c | 12 +
Hi,
I discovered Intel ADQ feature [1] that allows to boost performance by
picking dedicated queues for application traffic. We did some research,
and I got some level of understanding how it works, but I have some
questions, and I hope you could answer them.
1. SO_INCOMING_NAPI_ID usage. In
On 2020-05-08 16:14, Björn Töpel wrote:
On Fri, 8 May 2020 at 15:08, Björn Töpel wrote:
On Fri, 8 May 2020 at 15:01, Maxim Mikityanskiy wrote:
[]
All zeros hints that you're probably putting in the wrong DMA address somewhere.
Hmm, I can't see that you're using xsk_b
On 2020-05-08 15:27, Björn Töpel wrote:
On 2020-05-08 13:55, Maxim Mikityanskiy wrote:
On 2020-05-07 13:42, Björn Töpel wrote:
From: Björn Töpel
Use the new MEM_TYPE_XSK_BUFF_POOL API in lieu of MEM_TYPE_ZERO_COPY in
mlx5e. It allows to drop a lot of code from the driver (which is now
common
ter. (Maxim/Björn)
Signed-off-by: Maxim Mikityanskiy
Signed-off-by: Björn Töpel
Björn's signoff should go first (as the original author).
---
include/net/xdp.h | 4 +-
include/net/xdp_sock.h | 2 +
include/net/xdp_sock_drv.h | 152
include/net/xsk_buf
slightly improve performance.
rfc->v1: Put back the sanity check for XSK params, use XSK API to get
the total headroom size. (Maxim)
Signed-off-by: Björn Töpel
Signed-off-by: Maxim Mikityanskiy
I did some functional and performance tests.
Unfortunately, something is wrong with the traf
On 2020-05-04 14:37, Björn Töpel wrote:
From: Björn Töpel
In order to simplify AF_XDP zero-copy enablement for NIC driver
developers, a new AF_XDP buffer allocation API is added. The
implementation is based on a single core (single producer/consumer)
buffer pool for the AF_XDP UMEM.
A buffer i
On 2020-05-04 14:37, Björn Töpel wrote:
From: Björn Töpel
In order to simplify AF_XDP zero-copy enablement for NIC driver
developers, a new AF_XDP buffer allocation API is added. The
implementation is based on a single core (single producer/consumer)
buffer pool for the AF_XDP UMEM.
A buffer i
Hi Sean,
I was analyzing this code and have some concerns. Please see the comments below.
On Wed, Oct 24, 2018 at 1:04 AM Sean Tranchetti wrote:
>
> Current handling of CHECKSUM_COMPLETE packets by the UDP stack is
> incorrect for any packet that has an incorrect checksum value.
>
> udp4/6_csum_
On Thu, Sep 5, 2019 at 4:11 AM Kevin Laatz wrote:
>
> Currently, we don't add headroom to the handle in i40e_zca_free,
> i40e_alloc_buffer_slow_zc and i40e_alloc_buffer_zc. The addition of the
> headroom to the handle was removed in
> commit 2f86c806a8a8 ("i40e: modify driver for handling offsets"
On 2019-08-22 04:44, Kevin Laatz wrote:
> Currently, addresses are chunk size aligned. This means, we are very
> restricted in terms of where we can place chunk within the umem. For
> example, if we have a chunk size of 2k, then our chunks can only be placed
> at 0,2k,4k,6k,8k... and so on (ie. eve
On 2019-08-01 13:05, Maxim Mikityanskiy wrote:
> On 2019-07-30 11:53, Kevin Laatz wrote:
>> With the addition of the unaligned chunks option, we need to make sure we
>> handle the offsets accordingly based on the mode we are currently running
>> in. This patch modifies the dr
On 2019-08-19 09:35, Magnus Karlsson wrote:
> On Sat, Aug 17, 2019 at 12:02 AM Jonathan Lemon
> wrote:
>>
>>
>>
>> On 16 Aug 2019, at 8:37, Yonghong Song wrote:
>>
>>> On 8/16/19 3:26 AM, Magnus Karlsson wrote:
The zc is not used in the xsk part of libbpf, so let us remove it.
Not
g
some heavy operations on XDP_SETUP_PROG,
so they all have to handle these cases internally to return early if
they happen. This patch puts this check into the kernel code, so that
all drivers will benefit from it.
Signed-off-by: Maxim Mikityanskiy
---
v2 changes: Cover the case when the program
On 2019-07-30 11:53, Kevin Laatz wrote:
> With the addition of the unaligned chunks option, we need to make sure we
> handle the offsets accordingly based on the mode we are currently running
> in. This patch modifies the driver to appropriately mask the address for
> each case.
>
> Signed-off-by:
On 2019-07-24 08:10, Kevin Laatz wrote:
> With the addition of the unaligned chunks option, we need to make sure we
> handle the offsets accordingly based on the mode we are currently running
> in. This patch modifies the driver to appropriately mask the address for
> each case.
>
> Signed-off-by:
On 2019-07-24 08:10, Kevin Laatz wrote:
> Currently, addresses are chunk size aligned. This means, we are very
> restricted in terms of where we can place chunk within the umem. For
> example, if we have a chunk size of 2k, then our chunks can only be placed
> at 0,2k,4k,6k,8k... and so on (ie. eve
On 2019-07-24 08:10, Kevin Laatz wrote:
> This patch adds support for the unaligned chunks mode. The addition of the
> unaligned chunks option will allow users to run the application with more
> relaxed chunk placement in the XDP umem.
>
> Unaligned chunks mode can be used with the '-u' or '--unal
On 2019-07-24 08:10, Kevin Laatz wrote:
> Currently, addresses are chunk size aligned. This means, we are very
> restricted in terms of where we can place chunk within the umem. For
> example, if we have a chunk size of 2k, then our chunks can only be placed
> at 0,2k,4k,6k,8k... and so on (ie. eve
() instead.
Fixes: a038e9794541 ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Arnd Bergmann
Signed-off-by: Maxim Mikityanskiy
---
v2 changes: use kvzalloc/kvfree and fix a memory leak.
.../mellanox/mlx5/core/en/xsk/setup.c | 27 ---
1 file changed,
On 2019-06-12 19:14, Maxim Mikityanskiy wrote:
> dev_change_xdp_fd doesn't perform any checks in case it uninstalls an
> XDP program. It means that the driver's ndo_bpf can be called with
> XDP_SETUP_PROG asking to set it to NULL even if it's already NULL. This
> case h
1 - 100 of 242 matches
Mail list logo
