Skip to content

L
linux
Switch branch/tag
  1. 20 Oct, 2021 7 commits
    • Yunsheng Lin's avatar
      net: hns3: schedule the polling again when allocation fails · 68752b24
      Yunsheng Lin authored 
      Currently when there is a rx page allocation failure, it is
possible that polling may be stopped if there is no more packet
to be reveiced, which may cause queue stall problem under memory
pressure.

This patch makes sure polling is scheduled again when there is
any rx page allocation failure, and polling will try to allocate
receive buffers until it succeeds.

Now the allocation retry is added, it is unnecessary to do the rx
page allocation at the end of rx cleaning, so remove it. And reset
the unused_count to zero after calling hns3_nic_alloc_rx_buffers()
to avoid calling hns3_nic_alloc_rx_buffers() repeatedly under
memory pressure.

Fixes: 76ad4f0e ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: default avatarYunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: default avatarGuangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      68752b24
    • Yunsheng Lin's avatar
      net: hns3: fix for miscalculation of rx unused desc · 9f9f0f19
      Yunsheng Lin authored 
      rx unused desc is the desc that need attatching new buffer
before refilling to hw to receive new packet, the number of
desc need attatching new buffer is calculated using next_to_use
and next_to_clean. when next_to_use == next_to_clean, currently
hns3 driver assumes that all the desc has the buffer attatched,
but 'next_to_use == next_to_clean' also means all the desc need
attatching new buffer if hw has comsumed all the desc and the
driver has not attatched any buffer to the desc yet.

This patch adds 'refill' in desc_cb to indicate whether a new
buffer has been refilled to a desc.

Fixes: 76ad4f0e ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: default avatarYunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: default avatarGuangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      9f9f0f19
    • Yunsheng Lin's avatar
      net: hns3: fix the max tx size according to user manual · adfb7b49
      Yunsheng Lin authored 
      Currently the max tx size supported by the hw is calculated by
using the max BD num supported by the hw. According to the hw
user manual, the max tx size is fixed value for both non-TSO and
TSO skb.

This patch updates the max tx size according to the manual.

Fixes: 8ae10cfb("net: hns3: support tx-scatter-gather-fraglist feature")
Signed-off-by: default avatarYunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: default avatarGuangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      adfb7b49
    • Guangbin Huang's avatar
      net: hns3: add limit ets dwrr bandwidth cannot be 0 · 731797fd
      Guangbin Huang authored 
      If ets dwrr bandwidth of tc is set to 0, the hardware will switch to SP
mode. In this case, this tc may occupy all the tx bandwidth if it has
huge traffic, so it violates the purpose of the user setting.

To fix this problem, limit the ets dwrr bandwidth must greater than 0.

Fixes: cacde272 ("net: hns3: Add hclge_dcb module for the support of DCB feature")
Signed-off-by: default avatarGuangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      731797fd
    • Guangbin Huang's avatar
      net: hns3: reset DWRR of unused tc to zero · b63fcaab
      Guangbin Huang authored 
      Currently, DWRR of tc will be initialized to a fixed value when this tc
is enabled, but it is not been reset to 0 when this tc is disabled. It
cause a problem that the DWRR of unused tc is not 0 after using tc tool
to add and delete multi-tc parameters.

For examples, after enabling 4 TCs and restoring to 1 TC by follow
tc commands:

$ tc qdisc add dev eth0 root mqprio num_tc 4 map 0 1 2 3 0 1 2 3 queues \
  8@0 8@8 8@16 8@24 hw 1 mode channel
$ tc qdisc del dev eth0 root

Now there is just one TC is enabled for eth0, but the tc info querying by
debugfs is shown as follow:

$ cat /mnt/hns3/0000:7d:00.0/tm/tc_sch_info
enabled tc number: 1
weight_offset: 14
TC    MODE  WEIGHT
0     dwrr    100
1     dwrr    100
2     dwrr    100
3     dwrr    100
4     dwrr      0
5     dwrr      0
6     dwrr      0
7     dwrr      0

This patch fixes it by resetting DWRR of tc to 0 when tc is disabled.

Fixes: 84844054 ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: default avatarGuangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      b63fcaab
    • Jiaran Zhang's avatar
      net: hns3: Add configuration of TM QCN error event · 60484103
      Jiaran Zhang authored 
      Add configuration of interrupt type and fifo interrupt enable of TM QCN
error event if enabled, otherwise this event will not be reported when
there is error.

Fixes: d914971d ("net: hns3: remove redundant query in hclge_config_tm_hw_err_int()")
Signed-off-by: default avatarJiaran Zhang <zhangjiaran@huawei.com>
Signed-off-by: default avatarGuangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      60484103
    • Eugene Crosser's avatar
      vrf: Revert "Reset skb conntrack connection..." · 55161e67
      Eugene Crosser authored 
      This reverts commit 09e856d5.

When an interface is enslaved in a VRF, prerouting conntrack hook is
called twice: once in the context of the original input interface, and
once in the context of the VRF interface. If no special precausions are
taken, this leads to creation of two conntrack entries instead of one,
and breaks SNAT.

Commit above was intended to avoid creation of extra conntrack entries
when input interface is enslaved in a VRF. It did so by resetting
conntrack related data associated with the skb when it enters VRF context.

However it breaks netfilter operation. Imagine a use case when conntrack
zone must be assigned based on the original input interface, rather than
VRF interface (that would make original interfaces indistinguishable). One
could create netfilter rules similar to these:

        chain rawprerouting {
                type filter hook prerouting priority raw;
                iif realiface1 ct zone set 1 return
                iif realiface2 ct zone set 2 return
        }

This works before the mentioned commit, but not after: zone assignment
is "forgotten", and any subsequent NAT or filtering that is dependent
on the conntrack zone does not work.

Here is a reproducer script that demonstrates the difference in behaviour.

==========
#!/bin/sh

# This script demonstrates unexpected change of nftables behaviour
# caused by commit 09e856d5 ""vrf: Reset skb conntrack
# connection on VRF rcv"
# https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=09e856d54bda5f288ef8437a90ab2b9b3eab83d1
#
# Before the commit, it was possible to assign conntrack zone to a
# packet (or mark it for `notracking`) in the prerouting chanin, raw
# priority, based on the `iif` (interface from which the packet
# arrived).
# After the change, # if the interface is enslaved in a VRF, such
# assignment is lost. Instead, assignment based on the `iif` matching
# the VRF master interface is honored. Thus it is impossible to
# distinguish packets based on the original interface.
#
# This script demonstrates this change of behaviour: conntrack zone 1
# or 2 is assigned depending on the match with the original interface
# or the vrf master interface. It can be observed that conntrack entry
# appears in different zone in the kernel versions before and after
# the commit.

IPIN=172.30.30.1
IPOUT=172.30.30.2
PFXL=30

ip li sh vein >/dev/null 2>&1 && ip li del vein
ip li sh tvrf >/dev/null 2>&1 && ip li del tvrf
nft list table testct >/dev/null 2>&1 && nft delete table testct

ip li add vein type veth peer veout
ip li add tvrf type vrf table 9876
ip li set veout master tvrf
ip li set vein up
ip li set veout up
ip li set tvrf up
/sbin/sysctl -w net.ipv4.conf.veout.accept_local=1
/sbin/sysctl -w net.ipv4.conf.veout.rp_filter=0
ip addr add $IPIN/$PFXL dev vein
ip addr add $IPOUT/$PFXL dev veout

nft -f - <<__END__
table testct {
	chain rawpre {
		type filter hook prerouting priority raw;
		iif { veout, tvrf } meta nftrace set 1
		iif veout ct zone set 1 return
		iif tvrf ct zone set 2 return
		notrack
	}
	chain rawout {
		type filter hook output priority raw;
		notrack
	}
}
__END__

uname -rv
conntrack -F
ping -W 1 -c 1 -I vein $IPOUT
conntrack -L
Signed-off-by: default avatarEugene Crosser <crosser@average.org>
Acked-by: default avatarDavid Ahern <dsahern@kernel.org>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      55161e67
  3. 19 Oct, 2021 5 commits
  5. 18 Oct, 2021 12 commits
  7. 17 Oct, 2021 11 commits
  9. 16 Oct, 2021 3 commits
    • Nikolay Aleksandrov's avatar
      net: bridge: mcast: use multicast_membership_interval for IGMPv3 · fac3cb82
      Nikolay Aleksandrov authored 
      When I added IGMPv3 support I decided to follow the RFC for computing
the GMI dynamically:
" 8.4. Group Membership Interval

   The Group Membership Interval is the amount of time that must pass
   before a multicast router decides there are no more members of a
   group or a particular source on a network.

   This value MUST be ((the Robustness Variable) times (the Query
   Interval)) plus (one Query Response Interval)."

But that actually is inconsistent with how the bridge used to compute it
for IGMPv2, where it was user-configurable that has a correct default value
but it is up to user-space to maintain it. This would make it consistent
with the other timer values which are also maintained correct by the user
instead of being dynamically computed. It also changes back to the previous
user-expected GMI behaviour for IGMPv3 queries which were supported before
IGMPv3 was added. Note that to properly compute it dynamically we would
need to add support for "Robustness Variable" which is currently missing.
Reported-by: default avatarHangbin Liu <liuhangbin@gmail.com>
Fixes: 0436862e ("net: bridge: mcast: support for IGMPv3/MLDv2 ALLOW_NEW_SOURCES report")
Signed-off-by: default avatarNikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      fac3cb82
    • Stefano Garzarella's avatar
      vsock_diag_test: remove free_sock_stat() call in test_no_sockets · ba95a622
      Stefano Garzarella authored 
      In `test_no_sockets` we don't expect any sockets, indeed
check_no_sockets() prints an error and exits if `sockets` list is
not empty, so free_sock_stat() call is unnecessary since it would
only be called when the `sockets` list is empty.

This was discovered by a strange warning printed by gcc v11.2.1:
  In file included from ../../include/linux/list.h:7,
                   from vsock_diag_test.c:18:
  vsock_diag_test.c: In function ‘test_no_sockets’:
  ../../include/linux/kernel.h:35:45: error: array subscript ‘struct vsock_stat[0]’ is partly outside array bound
  s of ‘struct list_head[1]’ [-Werror=array-bounds]
     35 |         const typeof(((type *)0)->member) * __mptr = (ptr);     \
        |                                             ^~~~~~
  ../../include/linux/list.h:352:9: note: in expansion of macro ‘container_of’
    352 |         container_of(ptr, type, member)
        |         ^~~~~~~~~~~~
  ../../include/linux/list.h:393:9: note: in expansion of macro ‘list_entry’
    393 |         list_entry((pos)->member.next, typeof(*(pos)), member)
        |         ^~~~~~~~~~
  ../../include/linux/list.h:522:21: note: in expansion of macro ‘list_next_entry’
    522 |                 n = list_next_entry(pos, member);                       \
        |                     ^~~~~~~~~~~~~~~
  vsock_diag_test.c:325:9: note: in expansion of macro ‘list_for_each_entry_safe’
    325 |         list_for_each_entry_safe(st, next, sockets, list) {
        |         ^~~~~~~~~~~~~~~~~~~~~~~~
  In file included from vsock_diag_test.c:18:
  vsock_diag_test.c:333:19: note: while referencing ‘sockets’
    333 |         LIST_HEAD(sockets);
        |                   ^~~~~~~
  ../../include/linux/list.h:23:26: note: in definition of macro ‘LIST_HEAD’
     23 |         struct list_head name = LIST_HEAD_INIT(name)

It seems related to some compiler optimization and assumption
about the empty `sockets` list, since this warning is printed
only with -02 or -O3. Also removing `exit(1)` from
check_no_sockets() makes the warning disappear since in that
case free_sock_stat() can be reached also when the list is
not empty.
Reported-by: default avatarMarc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: default avatarStefano Garzarella <sgarzare@redhat.com>
Link: https://lore.kernel.org/r/20211014152045.173872-1-sgarzare@redhat.comSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      ba95a622
    • Jakub Kicinski's avatar
      Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue · 2151135a
      Jakub Kicinski authored 
      Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2021-10-14

Brett ensures RDMA nodes are removed during release and rebuild. He also
corrects fw.mgmt.api to include the patch number for proper
identification.

Dave stops ida_free() being called when an IDA has not been allocated.

Michal corrects the order of parameters being provided and the number of
entries skipped for UDP tunnels.
====================

Link: https://lore.kernel.org/r/20211014181953.3538330-1-anthony.l.nguyen@intel.comSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      2151135a
  11. 15 Oct, 2021 2 commits
    • Stephen Suryaputra's avatar
      ipv6: When forwarding count rx stats on the orig netdev · 0857d6f8
      Stephen Suryaputra authored 
      Commit bdb7cc64 ("ipv6: Count interface receive statistics on the
ingress netdev") does not work when ip6_forward() executes on the skbs
with vrf-enslaved netdev. Use IP6CB(skb)->iif to get to the right one.

Add a selftest script to verify.

Fixes: bdb7cc64 ("ipv6: Count interface receive statistics on the ingress netdev")
Signed-off-by: default avatarStephen Suryaputra <ssuryaextr@gmail.com>
Reviewed-by: default avatarDavid Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20211014130845.410602-1-ssuryaextr@gmail.comSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      0857d6f8
    • David S. Miller's avatar
      Merge branch 'tcp-md5-vrf-fix' · 4884ddba
      David S. Miller authored 
      Leonard Crestez says:

====================
tcp: md5: Fix overlap between vrf and non-vrf keys

With net.ipv4.tcp_l3mdev_accept=1 it is possible for a listen socket to
accept connection from the same client address in different VRFs. It is
also possible to set different MD5 keys for these clients which differ only
in the tcpm_l3index field.

This appears to work when distinguishing between different VRFs but not
between non-VRF and VRF connections. In particular:

* tcp_md5_do_lookup_exact will match a non-vrf key against a vrf key. This
means that adding a key with l3index != 0 after a key with l3index == 0
will cause the earlier key to be deleted. Both keys can be present if the
non-vrf key is added later.
* _tcp_md5_do_lookup can match a non-vrf key before a vrf key. This casues
failures if the passwords differ.

This can be fixed by making tcp_md5_do_lookup_exact perform an actual exact
comparison on l3index and by making  __tcp_md5_do_lookup perfer vrf-bound
keys above other considerations like prefixlen.

The fact that keys with l3index==0 affect VRF connections is usually not
desirable, VRFs are meant to be completely independent. This behavior needs
to preserved for backwards compatibility. Also, applications can just bind
listen sockets to VRF and never specify TCP_MD5SIG_FLAG_IFINDEX at all.

So far the combination of TCP_MD5SIG_FLAG_IFINDEX with tcpm_ifindex == 0
was an error, accept this to mean "key only applies to default VRF". This
is what applications using VRFs for traffic separation want.

This also contains tests for the second part. It does not contain tests for
overlapping keys, that would require more changes in nettest to add
multiple keys. These scenarios are also covered by my tests for TCP-AO,
especially around this area:
https://github.com/cdleonard/tcp-authopt-test/blob/main/tcp_authopt_test/test_vrf_bind.py

Changes since V2:
* Rename --do-bind-key-ifindex to --force-bind-key-ifindex
* Fix referencing TCP_MD5SIG_FLAG_IFINDEX as TCP_MD5SIG_IFINDEX
Link to v2: https://lore.kernel.org/netdev/cover.1634107317.git.cdleonard@gmail.com/

Changes since V1:
* Accept (TCP_MD5SIG_IFINDEX with tcpm_ifindex == 0)
* Add flags for explicitly including or excluding TCP_MD5SIG_FLAG_IFINDEX
to nettest
* Add few more tests in fcnal-test.sh.
Link to v1: https://lore.kernel.org/netdev/3d8387d499f053dba5cd9184c0f7b8445c4470c6.1633542093.git.cdleonard@gmail.com/
====================
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      4884ddba
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7