platform/kernel/linux-starfive.git
2 years agoMerge branch 'mlx5e-use-tls-tx-pool-to-improve-connection-rate'
Jakub Kicinski [Fri, 29 Jul 2022 04:50:57 +0000 (21:50 -0700)]
Merge branch 'mlx5e-use-tls-tx-pool-to-improve-connection-rate'

Tariq Toukan says:

====================
mlx5e use TLS TX pool to improve connection rate

To offload encryption operations, the mlx5 device maintains state and
keeps track of every kTLS device-offloaded connection.  Two HW objects
are used per TX context of a kTLS offloaded connection: a. Transport
interface send (TIS) object, to reach the HW context.  b. Data Encryption
Key (DEK) to perform the crypto operations.

These two objects are created and destroyed per TLS TX context, via FW
commands.  In total, 4 FW commands are issued per TLS TX context, which
seriously limits the connection rate.

In this series, we aim to save creation and destroy of TIS objects by
recycling them.  Upon recycling of a TIS, the HW still needs to be
notified for the re-mapping between a TIS and a context. This is done by
posting WQEs via an SQ, significantly faster API than the FW command
interface.

A pool is used for recycling. The pool dynamically interacts to the load
and connection rate, growing and shrinking accordingly.

Saving the TIS FW commands per context increases connection rate by ~42%,
from 11.6K to 16.5K connections per sec.

Connection rate is still limited by FW bottleneck due to the remaining
per context FW commands (DEK create/destroy). This will soon be addressed
in a followup series.  By combining the two series, the FW bottleneck
will be released, and a significantly higher (about 100K connections per
sec) kTLS TX device-offloaded connection rate is reached.
====================

Link: https://lore.kernel.org/r/20220727094346.10540-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/mlx5e: kTLS, Dynamically re-size TX recycling pool
Tariq Toukan [Wed, 27 Jul 2022 09:43:46 +0000 (12:43 +0300)]
net/mlx5e: kTLS, Dynamically re-size TX recycling pool

Let the TLS TX recycle pool be more flexible in size, by continuously
and dynamically allocating and releasing HW resources in response to
changes in the connections rate and load.

Allocate and release pool entries in bulks (16). Use a workqueue to
release/allocate in the background. Allocate a new bulk when the pool
size goes lower than the low threshold (1K). Symmetric operation is done
when the pool size gets greater than the upper threshold (4K).

Every idle pool entry holds: 1 TIS, 1 DEK (HW resources), in addition to
~100 bytes in host memory.

Start with an empty pool to minimize memory and HW resources waste for
non-TLS users that have the device-offload TLS enabled.

Upon a new request, in case the pool is empty, do not wait for a whole bulk
allocation to complete.  Instead, trigger an instant allocation of a single
resource to reduce latency.

Performance tests:
Before: 11,684 CPS
After:  16,556 CPS

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/mlx5e: kTLS, Recycle objects of device-offloaded TLS TX connections
Tariq Toukan [Wed, 27 Jul 2022 09:43:45 +0000 (12:43 +0300)]
net/mlx5e: kTLS, Recycle objects of device-offloaded TLS TX connections

The transport interface send (TIS) object is responsible for performing
all transport related operations of the transmit side.  The ConnectX HW
uses a TIS object to save and access the TLS crypto information and state
of an offloaded TX kTLS connection.

Before this patch, we used to create a new TIS per connection and destroy
it once it’s closed. Every create and destroy of a TIS is a FW command.

Same applies for the private TLS context, where we used to dynamically
allocate and free it per connection.

Resources recycling reduce the impact of the allocation/free operations
and helps speeding up the connection rate.

In this feature we maintain a pool of TX objects and use it to recycle
the resources instead of re-creating them per connection.

A cached TIS popped from the pool is updated to serve the new connection
via the fast-path HW interface, updating the tls static and progress
params. This is a very fast operation, significantly faster than FW
commands.

On recycling, a WQE fence is required after the context params change.
This guarantees that the data is sent after the context has been
successfully updated in hardware, and that the context modification
doesn't interfere with existing traffic.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/mlx5e: kTLS, Take stats out of OOO handler
Tariq Toukan [Wed, 27 Jul 2022 09:43:44 +0000 (12:43 +0300)]
net/mlx5e: kTLS, Take stats out of OOO handler

Let the caller of mlx5e_ktls_tx_handle_ooo() take care of updating the
stats, according to the returned value.  As the switch/case blocks are
already there, this change saves unnecessary branches in the handler.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/mlx5e: kTLS, Introduce TLS-specific create TIS
Tariq Toukan [Wed, 27 Jul 2022 09:43:43 +0000 (12:43 +0300)]
net/mlx5e: kTLS, Introduce TLS-specific create TIS

TLS TIS objects have a defined role in mapping and reaching the HW TLS
contexts.  Some standard TIS attributes (like LAG port affinity) are
not relevant for them.

Use a dedicated TLS TIS create function instead of the generic
mlx5e_create_tis.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/tls: Multi-threaded calls to TX tls_dev_del
Tariq Toukan [Wed, 27 Jul 2022 09:43:42 +0000 (12:43 +0300)]
net/tls: Multi-threaded calls to TX tls_dev_del

Multiple TLS device-offloaded contexts can be added in parallel via
concurrent calls to .tls_dev_add, while calls to .tls_dev_del are
sequential in tls_device_gc_task.

This is not a sustainable behavior. This creates a rate gap between add
and del operations (addition rate outperforms the deletion rate).  When
running for enough time, the TLS device resources could get exhausted,
failing to offload new connections.

Replace the single-threaded garbage collector work with a per-context
alternative, so they can be handled on several cores in parallel. Use
a new dedicated destruct workqueue for this.

Tested with mlx5 device:
Before: 22141 add/sec,   103 del/sec
After:  11684 add/sec, 11684 del/sec

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/tls: Perform immediate device ctx cleanup when possible
Tariq Toukan [Wed, 27 Jul 2022 09:43:41 +0000 (12:43 +0300)]
net/tls: Perform immediate device ctx cleanup when possible

TLS context destructor can be run in atomic context. Cleanup operations
for device-offloaded contexts could require access and interaction with
the device callbacks, which might sleep. Hence, the cleanup of such
contexts must be deferred and completed inside an async work.

For all others, this is not necessary, as cleanup is atomic. Invoke
cleanup immediately for them, avoiding queueing redundant gc work.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: Fix unsigned comparison with less than zero
Yang Li [Thu, 28 Jul 2022 03:10:19 +0000 (11:10 +0800)]
tls: rx: Fix unsigned comparison with less than zero

The return from the call to tls_rx_msg_size() is int, it can be
a negative error code, however this is being assigned to an
unsigned long variable 'sz', so making 'sz' an int.

Eliminate the following coccicheck warning:
./net/tls/tls_strp.c:211:6-8: WARNING: Unsigned expression compared with zero: sz < 0

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220728031019.32838-1-yang.lee@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'tls-rx-follow-ups-to-rx-work'
Jakub Kicinski [Fri, 29 Jul 2022 04:50:02 +0000 (21:50 -0700)]
Merge branch 'tls-rx-follow-ups-to-rx-work'

Jakub Kicinski says:

====================
tls: rx: follow ups to rx work

A selection of unrelated changes. First some selftest polishing.
Next a change to rcvtimeo handling for locking based on an exchange
with Eric. Follow up to Paolo's comments from yesterday. Last but
not least a fix to a false positive warning, turns out I've been
testing with DEBUG_NET=n this whole time.
====================

Link: https://lore.kernel.org/r/20220727031524.358216-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: fix the false positive warning
Jakub Kicinski [Wed, 27 Jul 2022 03:15:24 +0000 (20:15 -0700)]
tls: rx: fix the false positive warning

I went too far in the accessor conversion, we can't use tls_strp_msg()
after decryption because the message may not be ready. What we care
about on this path is that the output skb is detached, i.e. we didn't
somehow just turn around and used the input skb with its TCP data
still attached. So look at the anchor directly.

Fixes: 84c61fe1a75b ("tls: rx: do not use the standard strparser")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: strp: rename and multithread the workqueue
Jakub Kicinski [Wed, 27 Jul 2022 03:15:23 +0000 (20:15 -0700)]
tls: strp: rename and multithread the workqueue

Paolo points out that there seems to be no strong reason strparser
users a single threaded workqueue. Perhaps there were some performance
or pinning considerations? Since we don't know (and it's the slow path)
let's default to the most natural, multi-threaded choice.

Also rename the workqueue to "tls-".

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: don't consider sock_rcvtimeo() cumulative
Jakub Kicinski [Wed, 27 Jul 2022 03:15:22 +0000 (20:15 -0700)]
tls: rx: don't consider sock_rcvtimeo() cumulative

Eric indicates that restarting rcvtimeo on every wait may be fine.
I thought that we should consider it cumulative, and made
tls_rx_reader_lock() return the remaining timeo after acquiring
the reader lock.

tls_rx_rec_wait() gets its timeout passed in by value so it
does not keep track of time previously spent.

Make the lock waiting consistent with tls_rx_rec_wait() - don't
keep track of time spent.

Read the timeo fresh in tls_rx_rec_wait().
It's unclear to me why callers are supposed to cache the value.

Link: https://lore.kernel.org/all/CANn89iKcmSfWgvZjzNGbsrndmCch2HC_EPZ7qmGboDNaWoviNQ@mail.gmail.com/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: tls: handful of memrnd() and length checks
Jakub Kicinski [Wed, 27 Jul 2022 03:15:21 +0000 (20:15 -0700)]
selftests: tls: handful of memrnd() and length checks

Add a handful of memory randomizations and precise length checks.
Nothing is really broken here, I did this to increase confidence
when debugging. It does fix a GCC warning, tho. Apparently GCC
recognizes that memory needs to be initialized for send() but
does not recognize that for write().

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: usb: delete extra space and tab in blank line
Xie Shaowen [Wed, 27 Jul 2022 08:12:53 +0000 (16:12 +0800)]
net: usb: delete extra space and tab in blank line

delete extra space and tab in blank line, there is no functional change.

Signed-off-by: Xie Shaowen <studentxswpy@163.com>
Link: https://lore.kernel.org/r/20220727081253.3043941-1-studentxswpy@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Jakub Kicinski [Fri, 29 Jul 2022 01:21:16 +0000 (18:21 -0700)]
Merge git://git./linux/kernel/git/netdev/net

No conflicts.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'net-5.19-final' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 28 Jul 2022 18:54:59 +0000 (11:54 -0700)]
Merge tag 'net-5.19-final' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from bluetooth and netfilter, no known blockers for
  the release.

  Current release - regressions:

   - wifi: mac80211: do not abuse fq.lock in ieee80211_do_stop(), fix
     taking the lock before its initialized

   - Bluetooth: mgmt: fix double free on error path

  Current release - new code bugs:

   - eth: ice: fix tunnel checksum offload with fragmented traffic

  Previous releases - regressions:

   - tcp: md5: fix IPv4-mapped support after refactoring, don't take the
     pure v6 path

   - Revert "tcp: change pingpong threshold to 3", improving detection
     of interactive sessions

   - mld: fix netdev refcount leak in mld_{query | report}_work() due to
     a race

   - Bluetooth:
      - always set event mask on suspend, avoid early wake ups
      - L2CAP: fix use-after-free caused by l2cap_chan_put

   - bridge: do not send empty IFLA_AF_SPEC attribute

  Previous releases - always broken:

   - ping6: fix memleak in ipv6_renew_options()

   - sctp: prevent null-deref caused by over-eager error paths

   - virtio-net: fix the race between refill work and close, resulting
     in NAPI scheduled after close and a BUG()

   - macsec:
      - fix three netlink parsing bugs
      - avoid breaking the device state on invalid change requests
      - fix a memleak in another error path

  Misc:

   - dt-bindings: net: ethernet-controller: rework 'fixed-link' schema

   - two more batches of sysctl data race adornment"

* tag 'net-5.19-final' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (67 commits)
  stmmac: dwmac-mediatek: fix resource leak in probe
  ipv6/addrconf: fix a null-ptr-deref bug for ip6_ptr
  net: ping6: Fix memleak in ipv6_renew_options().
  net/funeth: Fix fun_xdp_tx() and XDP packet reclaim
  sctp: leave the err path free in sctp_stream_init to sctp_stream_free
  sfc: disable softirqs for ptp TX
  ptp: ocp: Select CRC16 in the Kconfig.
  tcp: md5: fix IPv4-mapped support
  virtio-net: fix the race between refill work and close
  mptcp: Do not return EINPROGRESS when subflow creation succeeds
  Bluetooth: L2CAP: Fix use-after-free caused by l2cap_chan_put
  Bluetooth: Always set event mask on suspend
  Bluetooth: mgmt: Fix double free on error path
  wifi: mac80211: do not abuse fq.lock in ieee80211_do_stop()
  ice: do not setup vlan for loopback VSI
  ice: check (DD | EOF) bits on Rx descriptor rather than (EOP | RS)
  ice: Fix VSIs unable to share unicast MAC
  ice: Fix tunnel checksum offload with fragmented traffic
  ice: Fix max VLANs available for VF
  netfilter: nft_queue: only allow supported familes and hooks
  ...

2 years agostmmac: dwmac-mediatek: fix resource leak in probe
Dan Carpenter [Thu, 28 Jul 2022 11:52:09 +0000 (14:52 +0300)]
stmmac: dwmac-mediatek: fix resource leak in probe

If mediatek_dwmac_clks_config() fails, then call stmmac_remove_config_dt()
before returning.  Otherwise it is a resource leak.

Fixes: fa4b3ca60e80 ("stmmac: dwmac-mediatek: fix clock issue")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/YuJ4aZyMUlG6yGGa@kili
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoipv6/addrconf: fix a null-ptr-deref bug for ip6_ptr
Ziyang Xuan [Thu, 28 Jul 2022 01:33:07 +0000 (09:33 +0800)]
ipv6/addrconf: fix a null-ptr-deref bug for ip6_ptr

Change net device's MTU to smaller than IPV6_MIN_MTU or unregister
device while matching route. That may trigger null-ptr-deref bug
for ip6_ptr probability as following.

=========================================================
BUG: KASAN: null-ptr-deref in find_match.part.0+0x70/0x134
Read of size 4 at addr 0000000000000308 by task ping6/263

CPU: 2 PID: 263 Comm: ping6 Not tainted 5.19.0-rc7+ #14
Call trace:
 dump_backtrace+0x1a8/0x230
 show_stack+0x20/0x70
 dump_stack_lvl+0x68/0x84
 print_report+0xc4/0x120
 kasan_report+0x84/0x120
 __asan_load4+0x94/0xd0
 find_match.part.0+0x70/0x134
 __find_rr_leaf+0x408/0x470
 fib6_table_lookup+0x264/0x540
 ip6_pol_route+0xf4/0x260
 ip6_pol_route_output+0x58/0x70
 fib6_rule_lookup+0x1a8/0x330
 ip6_route_output_flags_noref+0xd8/0x1a0
 ip6_route_output_flags+0x58/0x160
 ip6_dst_lookup_tail+0x5b4/0x85c
 ip6_dst_lookup_flow+0x98/0x120
 rawv6_sendmsg+0x49c/0xc70
 inet_sendmsg+0x68/0x94

Reproducer as following:
Firstly, prepare conditions:
$ip netns add ns1
$ip netns add ns2
$ip link add veth1 type veth peer name veth2
$ip link set veth1 netns ns1
$ip link set veth2 netns ns2
$ip netns exec ns1 ip -6 addr add 2001:0db8:0:f101::1/64 dev veth1
$ip netns exec ns2 ip -6 addr add 2001:0db8:0:f101::2/64 dev veth2
$ip netns exec ns1 ifconfig veth1 up
$ip netns exec ns2 ifconfig veth2 up
$ip netns exec ns1 ip -6 route add 2000::/64 dev veth1 metric 1
$ip netns exec ns2 ip -6 route add 2001::/64 dev veth2 metric 1

Secondly, execute the following two commands in two ssh windows
respectively:
$ip netns exec ns1 sh
$while true; do ip -6 addr add 2001:0db8:0:f101::1/64 dev veth1; ip -6 route add 2000::/64 dev veth1 metric 1; ping6 2000::2; done

$ip netns exec ns1 sh
$while true; do ip link set veth1 mtu 1000; ip link set veth1 mtu 1500; sleep 5; done

It is because ip6_ptr has been assigned to NULL in addrconf_ifdown() firstly,
then ip6_ignore_linkdown() accesses ip6_ptr directly without NULL check.

cpu0 cpu1
fib6_table_lookup
__find_rr_leaf
addrconf_notify [ NETDEV_CHANGEMTU ]
addrconf_ifdown
RCU_INIT_POINTER(dev->ip6_ptr, NULL)
find_match
ip6_ignore_linkdown

So we can add NULL check for ip6_ptr before using in ip6_ignore_linkdown() to
fix the null-ptr-deref bug.

Fixes: dcd1f572954f ("net/ipv6: Remove fib6_idev")
Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20220728013307.656257-1-william.xuanziyang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: ping6: Fix memleak in ipv6_renew_options().
Kuniyuki Iwashima [Thu, 28 Jul 2022 01:22:20 +0000 (18:22 -0700)]
net: ping6: Fix memleak in ipv6_renew_options().

When we close ping6 sockets, some resources are left unfreed because
pingv6_prot is missing sk->sk_prot->destroy().  As reported by
syzbot [0], just three syscalls leak 96 bytes and easily cause OOM.

    struct ipv6_sr_hdr *hdr;
    char data[24] = {0};
    int fd;

    hdr = (struct ipv6_sr_hdr *)data;
    hdr->hdrlen = 2;
    hdr->type = IPV6_SRCRT_TYPE_4;

    fd = socket(AF_INET6, SOCK_DGRAM, NEXTHDR_ICMP);
    setsockopt(fd, IPPROTO_IPV6, IPV6_RTHDR, data, 24);
    close(fd);

To fix memory leaks, let's add a destroy function.

Note the socket() syscall checks if the GID is within the range of
net.ipv4.ping_group_range.  The default value is [1, 0] so that no
GID meets the condition (1 <= GID <= 0).  Thus, the local DoS does
not succeed until we change the default value.  However, at least
Ubuntu/Fedora/RHEL loosen it.

    $ cat /usr/lib/sysctl.d/50-default.conf
    ...
    -net.ipv4.ping_group_range = 0 2147483647

Also, there could be another path reported with these options, and
some of them require CAP_NET_RAW.

  setsockopt
      IPV6_ADDRFORM (inet6_sk(sk)->pktoptions)
      IPV6_RECVPATHMTU (inet6_sk(sk)->rxpmtu)
      IPV6_HOPOPTS (inet6_sk(sk)->opt)
      IPV6_RTHDRDSTOPTS (inet6_sk(sk)->opt)
      IPV6_RTHDR (inet6_sk(sk)->opt)
      IPV6_DSTOPTS (inet6_sk(sk)->opt)
      IPV6_2292PKTOPTIONS (inet6_sk(sk)->opt)

  getsockopt
      IPV6_FLOWLABEL_MGR (inet6_sk(sk)->ipv6_fl_list)

For the record, I left a different splat with syzbot's one.

  unreferenced object 0xffff888006270c60 (size 96):
    comm "repro2", pid 231, jiffies 4294696626 (age 13.118s)
    hex dump (first 32 bytes):
      01 00 00 00 44 00 00 00 00 00 00 00 00 00 00 00  ....D...........
      00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    backtrace:
      [<00000000f6bc7ea9>] sock_kmalloc (net/core/sock.c:2564 net/core/sock.c:2554)
      [<000000006d699550>] do_ipv6_setsockopt.constprop.0 (net/ipv6/ipv6_sockglue.c:715)
      [<00000000c3c3b1f5>] ipv6_setsockopt (net/ipv6/ipv6_sockglue.c:1024)
      [<000000007096a025>] __sys_setsockopt (net/socket.c:2254)
      [<000000003a8ff47b>] __x64_sys_setsockopt (net/socket.c:2265 net/socket.c:2262 net/socket.c:2262)
      [<000000007c409dcb>] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80)
      [<00000000e939c4a9>] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120)

[0]: https://syzkaller.appspot.com/bug?extid=a8430774139ec3ab7176

Fixes: 6d0bfe226116 ("net: ipv6: Add IPv6 support to the ping socket.")
Reported-by: syzbot+a8430774139ec3ab7176@syzkaller.appspotmail.com
Reported-by: Ayushman Dutta <ayudutta@amazon.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220728012220.46918-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowatch_queue: Fix missing locking in add_watch_to_object()
Linus Torvalds [Thu, 28 Jul 2022 09:31:12 +0000 (10:31 +0100)]
watch_queue: Fix missing locking in add_watch_to_object()

If a watch is being added to a queue, it needs to guard against
interference from addition of a new watch, manual removal of a watch and
removal of a watch due to some other queue being destroyed.

KEYCTL_WATCH_KEY guards against this for the same {key,queue} pair by
holding the key->sem writelocked and by holding refs on both the key and
the queue - but that doesn't prevent interaction from other {key,queue}
pairs.

While add_watch_to_object() does take the spinlock on the event queue,
it doesn't take the lock on the source's watch list.  The assumption was
that the caller would prevent that (say by taking key->sem) - but that
doesn't prevent interference from the destruction of another queue.

Fix this by locking the watcher list in add_watch_to_object().

Fixes: c73be61cede5 ("pipe: Add general notification queue support")
Reported-by: syzbot+03d7b43290037d1f87ca@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: keyrings@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agowatch_queue: Fix missing rcu annotation
David Howells [Thu, 28 Jul 2022 09:31:06 +0000 (10:31 +0100)]
watch_queue: Fix missing rcu annotation

Since __post_watch_notification() walks wlist->watchers with only the
RCU read lock held, we need to use RCU methods to add to the list (we
already use RCU methods to remove from the list).

Fix add_watch_to_object() to use hlist_add_head_rcu() instead of
hlist_add_head() for that list.

Fixes: c73be61cede5 ("pipe: Add general notification queue support")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agonet: cdns,macb: use correct xlnx prefix for Xilinx
Krzysztof Kozlowski [Tue, 26 Jul 2022 07:08:02 +0000 (09:08 +0200)]
net: cdns,macb: use correct xlnx prefix for Xilinx

Use correct vendor for Xilinx versions of Cadence MACB/GEM Ethernet
controller.  The Versal compatible was not released, so it can be
changed.  Zynq-7xxx and Ultrascale+ has to be kept in new and deprecated
form.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Acked-by: Harini Katakam <harini.katakam@amd.com>
Link: https://lore.kernel.org/r/20220726070802.26579-2-krzysztof.kozlowski@linaro.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agodt-bindings: net: cdns,macb: use correct xlnx prefix for Xilinx
Krzysztof Kozlowski [Tue, 26 Jul 2022 07:08:01 +0000 (09:08 +0200)]
dt-bindings: net: cdns,macb: use correct xlnx prefix for Xilinx

Use correct vendor for Xilinx versions of Cadence MACB/GEM Ethernet
controller.  The Versal compatible was not released, so it can be
changed.  Zynq-7xxx and Ultrascale+ has to be kept in new and deprecated
form.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220726070802.26579-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet/funeth: Fix fun_xdp_tx() and XDP packet reclaim
Dimitris Michailidis [Tue, 26 Jul 2022 21:59:23 +0000 (14:59 -0700)]
net/funeth: Fix fun_xdp_tx() and XDP packet reclaim

The current implementation of fun_xdp_tx(), used for XPD_TX, is
incorrect in that it takes an address/length pair and later releases it
with page_frag_free(). It is OK for XDP_TX but the same code is used by
ndo_xdp_xmit. In that case it loses the XDP memory type and releases the
packet incorrectly for some of the types. Assorted breakage follows.

Change fun_xdp_tx() to take xdp_frame and rely on xdp_return_frame() in
reclaim.

Fixes: db37bc177dae ("net/funeth: add the data path")
Signed-off-by: Dimitris Michailidis <dmichail@fungible.com>
Link: https://lore.kernel.org/r/20220726215923.7887-1-dmichail@fungible.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next...
Paolo Abeni [Thu, 28 Jul 2022 09:54:56 +0000 (11:54 +0200)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
ice: PPPoE offload support

Marcin Szycik says:

Add support for dissecting PPPoE and PPP-specific fields in flow dissector:
PPPoE session id and PPP protocol type. Add support for those fields in
tc-flower and support offloading PPPoE. Finally, add support for hardware
offload of PPPoE packets in switchdev mode in ice driver.

Example filter:
tc filter add dev $PF1 ingress protocol ppp_ses prio 1 flower pppoe_sid \
    1234 ppp_proto ip skip_sw action mirred egress redirect dev $VF1_PR

Changes in iproute2 are required to use the new fields (will be submitted
soon).

ICE COMMS DDP package is required to create a filter in ice.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  ice: Add support for PPPoE hardware offload
  flow_offload: Introduce flow_match_pppoe
  net/sched: flower: Add PPPoE filter
  flow_dissector: Add PPPoE dissectors
====================

Link: https://lore.kernel.org/r/20220726203133.2171332-1-anthony.l.nguyen@intel.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoadd missing includes and forward declarations to networking includes under linux/
Jakub Kicinski [Tue, 26 Jul 2022 21:56:52 +0000 (14:56 -0700)]
add missing includes and forward declarations to networking includes under linux/

Similarly to a recent include/net/ cleanup, this patch adds
missing includes to networking headers under include/linux.
All these problems are currently masked by the existing users
including the missing dependency before the broken header.

Link: https://lore.kernel.org/all/20220723045755.2676857-1-kuba@kernel.org/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20220726215652.158167-1-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoRevert "Merge branch 'octeontx2-minor-tc-fixes'"
Paolo Abeni [Tue, 26 Jul 2022 11:09:27 +0000 (13:09 +0200)]
Revert "Merge branch 'octeontx2-minor-tc-fixes'"

This reverts commit 35d099da41967f114c6472b838e12014706c26e7, reversing
changes made to 58d8bcd47ecc55f1ab92320fe36c31ff4d83cc0c.

I wrongly applied that to the net-next tree instead of the intended
target tree (net). Reverting it on net-next.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: mv88e6xxx: fix speed setting for CPU/DSA ports
Marcin Wojtas [Tue, 26 Jul 2022 23:09:18 +0000 (01:09 +0200)]
net: dsa: mv88e6xxx: fix speed setting for CPU/DSA ports

Commit 3c783b83bd0f ("net: dsa: mv88e6xxx: get rid of SPEED_MAX setting")
stopped relying on SPEED_MAX constant and hardcoded speed settings
for the switch ports and rely on phylink configuration.

It turned out, however, that when the relevant code is called,
the mac_capabilites of CPU/DSA port remain unset.
mv88e6xxx_setup_port() is called via mv88e6xxx_setup() in
dsa_tree_setup_switches(), which precedes setting the caps in
phylink_get_caps down in the chain of dsa_tree_setup_ports().

As a result the mac_capabilites are 0 and the default speed for CPU/DSA
port is 10M at the start. To fix that, execute mv88e6xxx_get_caps()
and obtain the capabilities driectly.

Fixes: 3c783b83bd0f ("net: dsa: mv88e6xxx: get rid of SPEED_MAX setting")
Signed-off-by: Marcin Wojtas <mw@semihalf.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20220726230918.2772378-1-mw@semihalf.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net...
Jakub Kicinski [Thu, 28 Jul 2022 02:56:28 +0000 (19:56 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-07-26

This series contains updates to ice driver only.

Przemyslaw corrects accounting for VF VLANs to allow for correct number
of VLANs for untrusted VF. He also correct issue with checksum offload
on VXLAN tunnels.

Ani allows for two VSIs to share the same MAC address.

Maciej corrects checked bits for descriptor completion of loopback

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  ice: do not setup vlan for loopback VSI
  ice: check (DD | EOF) bits on Rx descriptor rather than (EOP | RS)
  ice: Fix VSIs unable to share unicast MAC
  ice: Fix tunnel checksum offload with fragmented traffic
  ice: Fix max VLANs available for VF
====================

Link: https://lore.kernel.org/r/20220726204646.2171589-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: devlink: remove redundant net_eq() check from sb_pool_get_dumpit()
Jiri Pirko [Wed, 27 Jul 2022 05:59:12 +0000 (07:59 +0200)]
net: devlink: remove redundant net_eq() check from sb_pool_get_dumpit()

The net_eq() check is already performed inside
devlinks_xa_for_each_registered_get() helper, so remove the redundant
appearance.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20220727055912.568391-1-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agodt-bindings: net: hirschmann,hellcreek: use absolute path to other schema
Krzysztof Kozlowski [Tue, 26 Jul 2022 11:56:50 +0000 (13:56 +0200)]
dt-bindings: net: hirschmann,hellcreek: use absolute path to other schema

Absolute path to other DT schema is preferred over relative one.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Acked-by: Kurt Kanzenbach <kurt@linutronix.de>
Acked-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220726115650.100726-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/sched: sch_cbq: change the type of cbq_set_lss to void
Zhengchao Shao [Tue, 26 Jul 2022 03:07:48 +0000 (11:07 +0800)]
net/sched: sch_cbq: change the type of cbq_set_lss to void

Change the type of cbq_set_lss to void.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Link: https://lore.kernel.org/r/20220726030748.243505-1-shaozhengchao@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosctp: leave the err path free in sctp_stream_init to sctp_stream_free
Xin Long [Mon, 25 Jul 2022 22:11:06 +0000 (18:11 -0400)]
sctp: leave the err path free in sctp_stream_init to sctp_stream_free

A NULL pointer dereference was reported by Wei Chen:

  BUG: kernel NULL pointer dereference, address: 0000000000000000
  RIP: 0010:__list_del_entry_valid+0x26/0x80
  Call Trace:
   <TASK>
   sctp_sched_dequeue_common+0x1c/0x90
   sctp_sched_prio_dequeue+0x67/0x80
   __sctp_outq_teardown+0x299/0x380
   sctp_outq_free+0x15/0x20
   sctp_association_free+0xc3/0x440
   sctp_do_sm+0x1ca7/0x2210
   sctp_assoc_bh_rcv+0x1f6/0x340

This happens when calling sctp_sendmsg without connecting to server first.
In this case, a data chunk already queues up in send queue of client side
when processing the INIT_ACK from server in sctp_process_init() where it
calls sctp_stream_init() to alloc stream_in. If it fails to alloc stream_in
all stream_out will be freed in sctp_stream_init's err path. Then in the
asoc freeing it will crash when dequeuing this data chunk as stream_out
is missing.

As we can't free stream out before dequeuing all data from send queue, and
this patch is to fix it by moving the err path stream_out/in freeing in
sctp_stream_init() to sctp_stream_free() which is eventually called when
freeing the asoc in sctp_association_free(). This fix also makes the code
in sctp_process_init() more clear.

Note that in sctp_association_init() when it fails in sctp_stream_init(),
sctp_association_free() will not be called, and in that case it should
go to 'stream_free' err path to free stream instead of 'fail_init'.

Fixes: 5bbbbe32a431 ("sctp: introduce stream scheduler foundations")
Reported-by: Wei Chen <harperchen1110@gmail.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/831a3dc100c4908ff76e5bcc363be97f2778bc0b.1658787066.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: disable softirqs for ptp TX
Alejandro Lucero [Tue, 26 Jul 2022 06:45:04 +0000 (08:45 +0200)]
sfc: disable softirqs for ptp TX

Sending a PTP packet can imply to use the normal TX driver datapath but
invoked from the driver's ptp worker. The kernel generic TX code
disables softirqs and preemption before calling specific driver TX code,
but the ptp worker does not. Although current ptp driver functionality
does not require it, there are several reasons for doing so:

   1) The invoked code is always executed with softirqs disabled for non
      PTP packets.
   2) Better if a ptp packet transmission is not interrupted by softirq
      handling which could lead to high latencies.
   3) netdev_xmit_more used by the TX code requires preemption to be
      disabled.

Indeed a solution for dealing with kernel preemption state based on static
kernel configuration is not possible since the introduction of dynamic
preemption level configuration at boot time using the static calls
functionality.

Fixes: f79c957a0b537 ("drivers: net: sfc: use netdev_xmit_more helper")
Signed-off-by: Alejandro Lucero <alejandro.lucero-palau@amd.com>
Link: https://lore.kernel.org/r/20220726064504.49613-1-alejandro.lucero-palau@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoptp: ocp: Select CRC16 in the Kconfig.
Jonathan Lemon [Tue, 26 Jul 2022 22:06:04 +0000 (15:06 -0700)]
ptp: ocp: Select CRC16 in the Kconfig.

The crc16() function is used to check the firmware validity, but
the library was not explicitly selected.

Fixes: 3c3673bde50c ("ptp: ocp: Add firmware header checks")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Vadim Fedorenko <vadfed@fb.com>
Link: https://lore.kernel.org/r/20220726220604.1339972-1-jonathan.lemon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: md5: fix IPv4-mapped support
Eric Dumazet [Tue, 26 Jul 2022 11:57:43 +0000 (11:57 +0000)]
tcp: md5: fix IPv4-mapped support

After the blamed commit, IPv4 SYN packets handled
by a dual stack IPv6 socket are dropped, even if
perfectly valid.

$ nstat | grep MD5
TcpExtTCPMD5Failure             5                  0.0

For a dual stack listener, an incoming IPv4 SYN packet
would call tcp_inbound_md5_hash() with @family == AF_INET,
while tp->af_specific is pointing to tcp_sock_ipv6_specific.

Only later when an IPv4-mapped child is created, tp->af_specific
is changed to tcp_sock_ipv6_mapped_specific.

Fixes: 7bbb765b7349 ("net/tcp: Merge TCP-MD5 inbound callbacks")
Reported-by: Brian Vazquez <brianvv@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Dmitry Safonov <dima@arista.com>
Tested-by: Leonard Crestez <cdleonard@gmail.com>
Link: https://lore.kernel.org/r/20220726115743.2759832-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'asm-generic-fixes-5.19-2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 27 Jul 2022 16:50:18 +0000 (09:50 -0700)]
Merge tag 'asm-generic-fixes-5.19-2' of git://git./linux/kernel/git/arnd/asm-generic

Pull asm-generic fixes from Arnd Bergmann:
 "Two more bug fixes for asm-generic, one addressing an incorrect
  Kconfig symbol reference and another one fixing a build failure for
  the perf tool on mips and possibly others"

* tag 'asm-generic-fixes-5.19-2' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic:
  asm-generic: remove a broken and needless ifdef conditional
  tools: Fixed MIPS builds due to struct flock re-definition

2 years agoMerge tag 'soc-fixes-5.19-4' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
Linus Torvalds [Wed, 27 Jul 2022 16:43:07 +0000 (09:43 -0700)]
Merge tag 'soc-fixes-5.19-4' of git://git./linux/kernel/git/soc/soc

Pull ARM SoC fixes from Arnd Bergmann:
 "One last set of changes for the soc tree:

   - fix clock frequency on lan966x

   - fix incorrect GPIO numbers on some pxa machines

   - update Baolin's email address"

* tag 'soc-fixes-5.19-4' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
  ARM: pxa2xx: Fix GPIO descriptor tables
  mailmap: update Baolin Wang's email
  ARM: dts: lan966x: fix sys_clk frequency

2 years agoMerge branch 'smc-updates'
David S. Miller [Wed, 27 Jul 2022 12:24:43 +0000 (13:24 +0100)]
Merge branch 'smc-updates'

Wenjia Zhang says:

====================
net/smc: updates 2022-07-25

please apply the following patches to netdev's net-next tree.

These patches do some preparation to make ISM available for uses beyond
SMC-D, and a bunch of cleanups.

v2: add "Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>"
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/smc: Enable module load on netlink usage
Stefan Raspl [Mon, 25 Jul 2022 14:10:00 +0000 (16:10 +0200)]
net/smc: Enable module load on netlink usage

Previously, the smc and smc_diag modules were automatically loaded as
dependencies of the ism module whenever an ISM device was present.
With the pending rework of the ISM API, the smc module will no longer
automatically be loaded in presence of an ISM device. Usage of an AF_SMC
socket will still trigger loading of the smc modules, but usage of a
netlink socket will not.
This is addressed by setting the correct module aliases.

Signed-off-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Wenjia Zhang < wenjia@linux.ibm.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/smc: Pass on DMBE bit mask in IRQ handler
Stefan Raspl [Mon, 25 Jul 2022 14:09:59 +0000 (16:09 +0200)]
net/smc: Pass on DMBE bit mask in IRQ handler

Make the DMBE bits, which are passed on individually in ism_move() as
parameter idx, available to the receiver.

Signed-off-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Wenjia Zhang < wenjia@linux.ibm.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agos390/ism: Cleanups
Stefan Raspl [Mon, 25 Jul 2022 14:09:58 +0000 (16:09 +0200)]
s390/ism: Cleanups

Reworked signature of the function to retrieve the system EID: No plausible
reason to use a double pointer. And neither to pass in the device as an
argument, as this identifier is by definition per system, not per device.
Plus some minor consistency edits.

Signed-off-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Wenjia Zhang < wenjia@linux.ibm.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/smc: Eliminate struct smc_ism_position
Heiko Carstens [Mon, 25 Jul 2022 14:09:57 +0000 (16:09 +0200)]
net/smc: Eliminate struct smc_ism_position

This struct is used in a single place only, and its usage generates
inefficient code. Time to clean up!

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-and-tested-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Wenjia Zhang < wenjia@linux.ibm.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agovirtio-net: fix the race between refill work and close
Jason Wang [Mon, 25 Jul 2022 07:21:59 +0000 (15:21 +0800)]
virtio-net: fix the race between refill work and close

We try using cancel_delayed_work_sync() to prevent the work from
enabling NAPI. This is insufficient since we don't disable the source
of the refill work scheduling. This means an NAPI poll callback after
cancel_delayed_work_sync() can schedule the refill work then can
re-enable the NAPI that leads to use-after-free [1].

Since the work can enable NAPI, we can't simply disable NAPI before
calling cancel_delayed_work_sync(). So fix this by introducing a
dedicated boolean to control whether or not the work could be
scheduled from NAPI.

[1]
==================================================================
BUG: KASAN: use-after-free in refill_work+0x43/0xd4
Read of size 2 at addr ffff88810562c92e by task kworker/2:1/42

CPU: 2 PID: 42 Comm: kworker/2:1 Not tainted 5.19.0-rc1+ #480
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
Workqueue: events refill_work
Call Trace:
 <TASK>
 dump_stack_lvl+0x34/0x44
 print_report.cold+0xbb/0x6ac
 ? _printk+0xad/0xde
 ? refill_work+0x43/0xd4
 kasan_report+0xa8/0x130
 ? refill_work+0x43/0xd4
 refill_work+0x43/0xd4
 process_one_work+0x43d/0x780
 worker_thread+0x2a0/0x6f0
 ? process_one_work+0x780/0x780
 kthread+0x167/0x1a0
 ? kthread_exit+0x50/0x50
 ret_from_fork+0x22/0x30
 </TASK>
...

Fixes: b2baed69e605c ("virtio_net: set/cancel work on ndo_open/ndo_stop")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'dsa-microchip-phylink-mac-config'
David S. Miller [Wed, 27 Jul 2022 08:39:17 +0000 (09:39 +0100)]
Merge branch 'dsa-microchip-phylink-mac-config'

Arun Ramadoss says:

====================
net: dsa: microchip: add support for phylink mac config and link up

This patch series add support common phylink mac config and link up for the ksz
series switches. At present, ksz8795 and ksz9477 doesn't implement the phylink
mac config and link up. It configures the mac interface in the port setup hook.
ksz8830 series switch does not mac link configuration. For lan937x switches, in
the part support patch series has support only for MII and RMII configuration.
Some group of switches have some register address and bit fields common and
others are different. So, this patch aims to have common phylink implementation
which configures the register based on the chip id.

Changes in v2
- combined the modification of duplex, tx_pause and rx_pause into single
  function.

Changes in v1
- Squash the reading rgmii value from dt to patch which apply the rgmii value
- Created the new function ksz_port_set_xmii_speed
- Seperated the namespace values for xmii_ctrl_0 and xmii_ctrl_1 register
- Applied the rgmii delay value based on the rx/tx-internal-delay-ps
====================

Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: add support for phylink mac config
Arun Ramadoss [Sun, 24 Jul 2022 09:38:12 +0000 (15:08 +0530)]
net: dsa: microchip: add support for phylink mac config

This patch add support for phylink mac config for ksz series of
switches. All the files ksz8795, ksz9477 and lan937x uses the ksz common
xmii function. Instead of calling from the individual files, it is moved
to the ksz common phylink mac config function.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: ksz8795: use common xmii function
Arun Ramadoss [Sun, 24 Jul 2022 09:28:22 +0000 (14:58 +0530)]
net: dsa: microchip: ksz8795: use common xmii function

This patch updates the ksz8795 cpu configuration to use the ksz common
xmii set functions.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: ksz9477: use common xmii function
Arun Ramadoss [Sun, 24 Jul 2022 09:28:21 +0000 (14:58 +0530)]
net: dsa: microchip: ksz9477: use common xmii function

In ksz9477.c file, configuring the xmii register is performed based on
the flag NEW_XMII. The flag is reset for ksz9893 switch and set for
other switch. This patch uses the ksz common xmii set and get function.
The bit values are configured based on the chip id.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: apply rgmii tx and rx delay in phylink mac config
Arun Ramadoss [Sun, 24 Jul 2022 09:28:20 +0000 (14:58 +0530)]
net: dsa: microchip: apply rgmii tx and rx delay in phylink mac config

This patch read the rgmii tx and rx delay from device tree and stored it
in the ksz_port.  It applies the rgmii delay to the xmii tune adjust
register based on the interface selected in phylink mac config. There
are two rgmii port in LAN937x and value to be loaded in the register
vary depends on the port selected.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: lan937x: add support for configuing xMII register
Arun Ramadoss [Sun, 24 Jul 2022 09:28:19 +0000 (14:58 +0530)]
net: dsa: microchip: lan937x: add support for configuing xMII register

This patch add the common ksz_set_xmii function for ksz series switch
and update the lan937x code phylink mac config. The register address for
the ksz8795 is Port 5 Interface control 6 and for all other switch is
xMII Control 1.
The bit value for selecting the interface is same for
KSZ8795 and KSZ9893 are same. The bit values for KSZ9477 and lan973x are
same. So, this patch add the bit value for each switches in
ksz_chip_data and configure the registers based on the chip id.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: add support for common phylink mac link up
Arun Ramadoss [Sun, 24 Jul 2022 09:28:18 +0000 (14:58 +0530)]
net: dsa: microchip: add support for common phylink mac link up

This patch add the support for common phylink mac link up for the ksz
series switch. The register address, bit position and values are
configured based on the chip id to the dev->info structure.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: add common duplex and flow control function
Arun Ramadoss [Sun, 24 Jul 2022 09:28:17 +0000 (14:58 +0530)]
net: dsa: microchip: add common duplex and flow control function

This patch add common function for configuring the Full/Half duplex and
transmit/receive flow control. KSZ8795 uses the Global control register
4 for configuring the duplex and flow control, whereas all other KSZ9477
based switch uses the xMII Control 0 register.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: add common ksz port xmii speed selection function
Arun Ramadoss [Sun, 24 Jul 2022 09:28:16 +0000 (14:58 +0530)]
net: dsa: microchip: add common ksz port xmii speed selection function

This patch adds the function for configuring the 100/10Mbps speed
selection for the ksz switches. KSZ8795 switch uses Global control 4
register 0x06 bit 4 for choosing 100/10Mpbs. Other switches uses xMII
control 1 0xN300 for it.
For KSZ8795, if the bit is set then 10Mbps is chosen and if bit is
clear then 100Mbps chosen. For all other switches it is other way
around, if the bit is set then 100Mbps is chosen.
So, this patch add the generic function for ksz switch to select the
100/10Mbps speed selection. While configuring, first it disables the
gigabit functionality and then configure the respective speed.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: microchip: add common gigabit set and get function
Arun Ramadoss [Sun, 24 Jul 2022 09:28:15 +0000 (14:58 +0530)]
net: dsa: microchip: add common gigabit set and get function

This patch add helper function for setting and getting the gigabit
enable for the ksz series switch. KSZ8795 switch has different register
address compared to all other ksz switches. KSZ8795 series uses the Port
5 Interface control 6 Bit 6 for configuring the 1Gbps or 100/10Mbps
speed selection. All other switches uses the xMII control 1 0xN301
register Bit6 for gigabit.
Further, for KSZ8795 & KSZ9893 switches if bit 1 then 1Gbps is chosen
and if bit 0 then 100/10Mbps is chosen. It is other way around for
other switches bit 0 is for 1Gbps. So, this patch implements the common
function for configuring the gigabit set and get capability.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: net: Fix typo 'the the' in comment
Slark Xiao [Mon, 25 Jul 2022 02:01:24 +0000 (10:01 +0800)]
selftests: net: Fix typo 'the the' in comment

Replace 'the the' with 'the' in the comment.

Signed-off-by: Slark Xiao <slark_xiao@163.com>
Link: https://lore.kernel.org/r/20220725020124.5760-1-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoip6mr: remove stray rcu_read_unlock() from ip6_mr_forward()
Eric Dumazet [Mon, 25 Jul 2022 20:05:54 +0000 (13:05 -0700)]
ip6mr: remove stray rcu_read_unlock() from ip6_mr_forward()

One rcu_read_unlock() should have been removed in blamed commit.

Fixes: 9b1c21d898fd ("ip6mr: do not acquire mrt_lock while calling ip6_mr_forward()")
Reported-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20220725200554.2563581-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomptcp: Do not return EINPROGRESS when subflow creation succeeds
Mat Martineau [Mon, 25 Jul 2022 20:52:31 +0000 (13:52 -0700)]
mptcp: Do not return EINPROGRESS when subflow creation succeeds

New subflows are created within the kernel using O_NONBLOCK, so
EINPROGRESS is the expected return value from kernel_connect().
__mptcp_subflow_connect() has the correct logic to consider EINPROGRESS
to be a successful case, but it has also used that error code as its
return value.

Before v5.19 this was benign: all the callers ignored the return
value. Starting in v5.19 there is a MPTCP_PM_CMD_SUBFLOW_CREATE generic
netlink command that does use the return value, so the EINPROGRESS gets
propagated to userspace.

Make __mptcp_subflow_connect() always return 0 on success instead.

Fixes: ec3edaa7ca6c ("mptcp: Add handling of outgoing MP_JOIN requests")
Fixes: 702c2f646d42 ("mptcp: netlink: allow userspace-driven subflow establishment")
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Link: https://lore.kernel.org/r/20220725205231.87529-1-mathew.j.martineau@linux.intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Jakub Kicinski [Wed, 27 Jul 2022 02:53:09 +0000 (19:53 -0700)]
Merge git://git./linux/kernel/git/netfilter/nf

Florian Westphal says:

====================
netfilter updates for net

Three late fixes for netfilter:

1) If nf_queue user requests packet truncation below size of l3 header,
   we corrupt the skb, then crash.  Reject such requests.

2) add cond_resched() calls when doing cycle detection in the
   nf_tables graph.  This avoids softlockup warning with certain
   rulesets.

3) Reject rulesets that use nftables 'queue' expression in family/chain
   combinations other than those that are supported.  Currently the ruleset
   will load, but when userspace attempts to reinject you get WARN splat +
   packet drops.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nft_queue: only allow supported familes and hooks
  netfilter: nf_tables: add rescheduling points during loop detection walks
  netfilter: nf_queue: do not allow packet truncation below transport header offset
====================

Link: https://lore.kernel.org/r/20220726192056.13497-1-fw@strlen.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'for-net-2022-07-26' of git://git.kernel.org/pub/scm/linux/kernel/git/bluet...
Jakub Kicinski [Wed, 27 Jul 2022 02:48:24 +0000 (19:48 -0700)]
Merge tag 'for-net-2022-07-26' of git://git./linux/kernel/git/bluetooth/bluetooth

Luiz Augusto von Dentz says:

====================
bluetooth pull request for net:

 - Fix early wakeup after suspend
 - Fix double free on error
 - Fix use-after-free on l2cap_chan_put

* tag 'for-net-2022-07-26' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth:
  Bluetooth: L2CAP: Fix use-after-free caused by l2cap_chan_put
  Bluetooth: Always set event mask on suspend
  Bluetooth: mgmt: Fix double free on error path
====================

Link: https://lore.kernel.org/r/20220726221328.423714-1-luiz.dentz@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'mm-hotfixes-stable-2022-07-26' of git://git.kernel.org/pub/scm/linux/kerne...
Linus Torvalds [Wed, 27 Jul 2022 02:38:46 +0000 (19:38 -0700)]
Merge tag 'mm-hotfixes-stable-2022-07-26' of git://git./linux/kernel/git/akpm/mm

Pull misc fixes from Andrew Morton:
 "Thirteen hotfixes.

  Eight are cc:stable and the remainder are for post-5.18 issues or are
  too minor to warrant backporting"

* tag 'mm-hotfixes-stable-2022-07-26' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
  mailmap: update Gao Xiang's email addresses
  userfaultfd: provide properly masked address for huge-pages
  Revert "ocfs2: mount shared volume without ha stack"
  hugetlb: fix memoryleak in hugetlb_mcopy_atomic_pte
  fs: sendfile handles O_NONBLOCK of out_fd
  ntfs: fix use-after-free in ntfs_ucsncmp()
  secretmem: fix unhandled fault in truncate
  mm/hugetlb: separate path for hwpoison entry in copy_hugetlb_page_range()
  mm: fix missing wake-up event for FSDAX pages
  mm: fix page leak with multiple threads mapping the same page
  mailmap: update Seth Forshee's email address
  tmpfs: fix the issue that the mount and remount results are inconsistent.
  mm: kfence: apply kmemleak_ignore_phys on early allocated pool

2 years agomailmap: update Gao Xiang's email addresses
Gao Xiang [Tue, 19 Jul 2022 15:42:46 +0000 (23:42 +0800)]
mailmap: update Gao Xiang's email addresses

I've been in Alibaba Cloud for more than one year, mainly to address
cloud-native challenges (such as high-performance container images) for
open source communities.

Update my email addresses on behalf of my current employer (Alibaba Cloud)
to support all my (team) work in this area.  Also add an outdated
@redhat.com address of me.

Link: https://lkml.kernel.org/r/20220719154246.62970-1-xiang@kernel.org
Signed-off-by: Gao Xiang <xiang@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2 years agouserfaultfd: provide properly masked address for huge-pages
Nadav Amit [Mon, 11 Jul 2022 16:59:06 +0000 (09:59 -0700)]
userfaultfd: provide properly masked address for huge-pages

Commit 824ddc601adc ("userfaultfd: provide unmasked address on
page-fault") was introduced to fix an old bug, in which the offset in the
address of a page-fault was masked.  Concerns were raised - although were
never backed by actual code - that some userspace code might break because
the bug has been around for quite a while.  To address these concerns a
new flag was introduced, and only when this flag is set by the user,
userfaultfd provides the exact address of the page-fault.

The commit however had a bug, and if the flag is unset, the offset was
always masked based on a base-page granularity.  Yet, for huge-pages, the
behavior prior to the commit was that the address is masked to the
huge-page granulrity.

While there are no reports on real breakage, fix this issue.  If the flag
is unset, use the address with the masking that was done before.

Link: https://lkml.kernel.org/r/20220711165906.2682-1-namit@vmware.com
Fixes: 824ddc601adc ("userfaultfd: provide unmasked address on page-fault")
Signed-off-by: Nadav Amit <namit@vmware.com>
Reported-by: James Houghton <jthoughton@google.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: James Houghton <jthoughton@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2 years agoMerge branch 'tls-rx-decrypt-from-the-tcp-queue'
Jakub Kicinski [Tue, 26 Jul 2022 21:38:53 +0000 (14:38 -0700)]
Merge branch 'tls-rx-decrypt-from-the-tcp-queue'

Jakub Kicinski says:

====================
tls: rx: decrypt from the TCP queue

This is the final part of my TLS Rx rework. It switches from
strparser to decrypting data from skbs queued in TCP. We don't
need the full strparser for TLS, its needs are very basic.
This set gives us a small but measurable (6%) performance
improvement (continuous stream).
====================

Link: https://lore.kernel.org/r/20220722235033.2594446-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: do not use the standard strparser
Jakub Kicinski [Fri, 22 Jul 2022 23:50:33 +0000 (16:50 -0700)]
tls: rx: do not use the standard strparser

TLS is a relatively poor fit for strparser. We pause the input
every time a message is received, wait for a read which will
decrypt the message, start the parser, repeat. strparser is
built to delineate the messages, wrap them in individual skbs
and let them float off into the stack or a different socket.
TLS wants the data pages and nothing else. There's no need
for TLS to keep cloning (and occasionally skb_unclone()'ing)
the TCP rx queue.

This patch uses a pre-allocated skb and attaches the skbs
from the TCP rx queue to it as frags. TLS is careful never
to modify the input skb without CoW'ing / detaching it first.

Since we call TCP rx queue cleanup directly we also get back
the benefit of skb deferred free.

Overall this results in a 6% gain in my benchmarks.

Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: device: add input CoW helper
Jakub Kicinski [Fri, 22 Jul 2022 23:50:32 +0000 (16:50 -0700)]
tls: rx: device: add input CoW helper

Wrap the remaining skb_cow_data() into a helper, so it's easier
to replace down the lane. The new version will change the skb
so make sure relevant pointers get reloaded after the call.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: allow tls to decrypt directly from the tcp rcv queue
Jakub Kicinski [Fri, 22 Jul 2022 23:50:31 +0000 (16:50 -0700)]
tcp: allow tls to decrypt directly from the tcp rcv queue

Expose TCP rx queue accessor and cleanup, so that TLS can
decrypt directly from the TCP queue. The expectation
is that the caller can access the skb returned from
tcp_recv_skb() and up to inq bytes worth of data (some
of which may be in ->next skbs) and then call
tcp_read_done() when data has been consumed.
The socket lock must be held continuously across
those two operations.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: device: keep the zero copy status with offload
Jakub Kicinski [Fri, 22 Jul 2022 23:50:30 +0000 (16:50 -0700)]
tls: rx: device: keep the zero copy status with offload

The non-zero-copy path assumes a full skb with decrypted contents.
This means the device offload would have to CoW the data. Try
to keep the zero-copy status instead, copy the data to user space.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: don't free the output in case of zero-copy
Jakub Kicinski [Fri, 22 Jul 2022 23:50:29 +0000 (16:50 -0700)]
tls: rx: don't free the output in case of zero-copy

In the future we'll want to reuse the input skb in case of
zero-copy so we shouldn't always free darg.skb. Move the
freeing of darg.skb into the non-zc cases. All cases will
now free ctx->recv_pkt (inside let tls_rx_rec_done()).

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: factor SW handling out of tls_rx_one_record()
Jakub Kicinski [Fri, 22 Jul 2022 23:50:28 +0000 (16:50 -0700)]
tls: rx: factor SW handling out of tls_rx_one_record()

After recent changes the SW side of tls_rx_one_record() can
be nicely encapsulated in its own function. Move the pad handling
as well. This will be useful for ->zc handling in tls_decrypt_device().

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotls: rx: wrap recv_pkt accesses in helpers
Jakub Kicinski [Fri, 22 Jul 2022 23:50:27 +0000 (16:50 -0700)]
tls: rx: wrap recv_pkt accesses in helpers

To allow for the logic to change later wrap accesses
which interrogate the input skb in helper functions.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'implement-dev-info-and-dev-flash-for-line-cards'
Jakub Kicinski [Tue, 26 Jul 2022 20:54:58 +0000 (13:54 -0700)]
Merge branch 'implement-dev-info-and-dev-flash-for-line-cards'

Jiri Pirko says:

====================
Implement dev info and dev flash for line cards

This patchset implements two features:
1) "devlink dev info" is exposed for line card (patches 6-9)
2) "devlink dev flash" is implemented for line card gearbox
   flashing (patch 10)

For every line card, "a nested" auxiliary device is created which
allows to bind the features mentioned above (patch 4).

The relationship between line card and its auxiliary dev devlink
is carried over extra line card netlink attribute (patches 3 and 5).

The first patch removes devlink_mutex from devlink_register/unregister()
which eliminates possible deadlock during devlink reload command. The
second patchset follows up with putting net pointer check into new
helper.

Examples:

$ devlink lc show pci/0000:01:00.0 lc 1
pci/0000:01:00.0:
  lc 1 state active type 16x100G nested_devlink auxiliary/mlxsw_core.lc.0
    supported_types:
       16x100G

$ devlink dev show auxiliary/mlxsw_core.lc.0
auxiliary/mlxsw_core.lc.0

$ devlink dev info auxiliary/mlxsw_core.lc.0
auxiliary/mlxsw_core.lc.0:
  versions:
      fixed:
        hw.revision 0
        fw.psid MT_0000000749
      running:
        ini.version 4
        fw 19.2010.1312

$ devlink dev flash auxiliary/mlxsw_core.lc.0 file mellanox/fw-AGB-rel-19_2010_1312-022-EVB.mfa2
====================

Link: https://lore.kernel.org/r/20220725082925.366455-1-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: mlxsw: Check line card info on activated line card
Jiri Pirko [Mon, 25 Jul 2022 08:29:25 +0000 (10:29 +0200)]
selftests: mlxsw: Check line card info on activated line card

Once line card is activated, check the FW version and PSID are exposed.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: mlxsw: Check line card info on provisioned line card
Jiri Pirko [Mon, 25 Jul 2022 08:29:24 +0000 (10:29 +0200)]
selftests: mlxsw: Check line card info on provisioned line card

Once line card is provisioned, check if HW revision and INI version
are exposed on associated nested auxiliary device.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: core_linecards: Implement line card device flashing
Jiri Pirko [Mon, 25 Jul 2022 08:29:23 +0000 (10:29 +0200)]
mlxsw: core_linecards: Implement line card device flashing

Implement flash_update() devlink op for the line card devlink instance
to allow user to update line card gearbox FW using MDDT register
and mlxfw.

Example:
$ devlink dev flash auxiliary/mlxsw_core.lc.0 file mellanox/fw-AGB-rel-19_2010_1312-022-EVB.mfa2

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: core_linecards: Expose device PSID over device info
Jiri Pirko [Mon, 25 Jul 2022 08:29:22 +0000 (10:29 +0200)]
mlxsw: core_linecards: Expose device PSID over device info

Use tunneled MGIR to obtain PSID of line card device and extend
device_info_get() op to fill up the info with that.

Example:

$ devlink dev info auxiliary/mlxsw_core.lc.0
auxiliary/mlxsw_core.lc.0:
  versions:
      fixed:
        hw.revision 0
        fw.psid MT_0000000749
      running:
        ini.version 4
        fw 19.2010.1312

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: reg: Add Management DownStream Device Tunneling Register
Jiri Pirko [Mon, 25 Jul 2022 08:29:21 +0000 (10:29 +0200)]
mlxsw: reg: Add Management DownStream Device Tunneling Register

The MDDT register allows to deliver query and request messages (PRM
registers, commands) to a DownStream device.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: core_linecards: Probe active line cards for devices and expose FW version
Jiri Pirko [Mon, 25 Jul 2022 08:29:20 +0000 (10:29 +0200)]
mlxsw: core_linecards: Probe active line cards for devices and expose FW version

In case the line card is active, go over all possible existing
devices (gearboxes) on it and expose FW version of the flashable one.

Example:

$ devlink dev info auxiliary/mlxsw_core.lc.0
auxiliary/mlxsw_core.lc.0:
  versions:
      fixed:
        hw.revision 0
      running:
        ini.version 4
        fw 19.2010.1312

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: reg: Extend MDDQ by device_info
Jiri Pirko [Mon, 25 Jul 2022 08:29:19 +0000 (10:29 +0200)]
mlxsw: reg: Extend MDDQ by device_info

Extend existing MDDQ register by possibility to query information about
devices residing on a line card.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: core_linecards: Expose HW revision and INI version
Jiri Pirko [Mon, 25 Jul 2022 08:29:18 +0000 (10:29 +0200)]
mlxsw: core_linecards: Expose HW revision and INI version

Implement info_get() to expose HW revision of a linecard and loaded INI
version.

Example:

$ devlink dev info auxiliary/mlxsw_core.lc.0
auxiliary/mlxsw_core.lc.0:
  versions:
      fixed:
        hw.revision 0
      running:
        ini.version 4

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: core_linecards: Introduce per line card auxiliary device
Jiri Pirko [Mon, 25 Jul 2022 08:29:17 +0000 (10:29 +0200)]
mlxsw: core_linecards: Introduce per line card auxiliary device

In order to be eventually able to expose line card gearbox version and
possibility to flash FW, model the line card as a separate device on
auxiliary bus.

Add the auxiliary device for provisioned line card in order to be able
to expose provisioned line card info over devlink dev info. When the
line card becomes active, there may be other additional info added to
the output.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: devlink: introduce nested devlink entity for line card
Jiri Pirko [Mon, 25 Jul 2022 08:29:16 +0000 (10:29 +0200)]
net: devlink: introduce nested devlink entity for line card

For the purpose of exposing device info and allow flash update which is
going to be implemented in follow-up patches, introduce a possibility
for a line card to expose relation to nested devlink entity. The nested
devlink entity represents the line card.

Example:

$ devlink lc show pci/0000:01:00.0 lc 1
pci/0000:01:00.0:
  lc 1 state active type 16x100G nested_devlink auxiliary/mlxsw_core.lc.0
    supported_types:
       16x100G
$ devlink dev show auxiliary/mlxsw_core.lc.0
auxiliary/mlxsw_core.lc.0

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: devlink: move net check into devlinks_xa_for_each_registered_get()
Jiri Pirko [Mon, 25 Jul 2022 08:29:15 +0000 (10:29 +0200)]
net: devlink: move net check into devlinks_xa_for_each_registered_get()

Benefit from having devlinks iterator helper
devlinks_xa_for_each_registered_get() and move the net pointer
check inside.

Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: devlink: make sure that devlink_try_get() works with valid pointer during xarray...
Jiri Pirko [Mon, 25 Jul 2022 08:29:14 +0000 (10:29 +0200)]
net: devlink: make sure that devlink_try_get() works with valid pointer during xarray iteration

Remove dependency on devlink_mutex during devlinks xarray iteration.

The reason is that devlink_register/unregister() functions taking
devlink_mutex would deadlock during devlink reload operation of devlink
instance which registers/unregisters nested devlink instances.

The devlinks xarray consistency is ensured internally by xarray.
There is a reference taken when working with devlink using
devlink_try_get(). But there is no guarantee that devlink pointer
picked during xarray iteration is not freed before devlink_try_get()
is called.

Make sure that devlink_try_get() works with valid pointer.
Achieve it by:
1) Splitting devlink_put() so the completion is sent only
   after grace period. Completion unblocks the devlink_unregister()
   routine, which is followed-up by devlink_free()
2) During devlinks xa_array iteration, get devlink pointer from xa_array
   holding RCU read lock and taking reference using devlink_try_get()
   before unlock.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoBluetooth: L2CAP: Fix use-after-free caused by l2cap_chan_put
Luiz Augusto von Dentz [Thu, 21 Jul 2022 16:10:50 +0000 (09:10 -0700)]
Bluetooth: L2CAP: Fix use-after-free caused by l2cap_chan_put

This fixes the following trace which is caused by hci_rx_work starting up
*after* the final channel reference has been put() during sock_close() but
*before* the references to the channel have been destroyed, so instead
the code now rely on kref_get_unless_zero/l2cap_chan_hold_unless_zero to
prevent referencing a channel that is about to be destroyed.

  refcount_t: increment on 0; use-after-free.
  BUG: KASAN: use-after-free in refcount_dec_and_test+0x20/0xd0
  Read of size 4 at addr ffffffc114f5bf18 by task kworker/u17:14/705

  CPU: 4 PID: 705 Comm: kworker/u17:14 Tainted: G S      W
  4.14.234-00003-g1fb6d0bd49a4-dirty #28
  Hardware name: Qualcomm Technologies, Inc. SM8150 V2 PM8150
  Google Inc. MSM sm8150 Flame DVT (DT)
  Workqueue: hci0 hci_rx_work
  Call trace:
   dump_backtrace+0x0/0x378
   show_stack+0x20/0x2c
   dump_stack+0x124/0x148
   print_address_description+0x80/0x2e8
   __kasan_report+0x168/0x188
   kasan_report+0x10/0x18
   __asan_load4+0x84/0x8c
   refcount_dec_and_test+0x20/0xd0
   l2cap_chan_put+0x48/0x12c
   l2cap_recv_frame+0x4770/0x6550
   l2cap_recv_acldata+0x44c/0x7a4
   hci_acldata_packet+0x100/0x188
   hci_rx_work+0x178/0x23c
   process_one_work+0x35c/0x95c
   worker_thread+0x4cc/0x960
   kthread+0x1a8/0x1c4
   ret_from_fork+0x10/0x18

Cc: stable@kernel.org
Reported-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Tested-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
2 years agoBluetooth: Always set event mask on suspend
Abhishek Pandit-Subedi [Mon, 25 Jul 2022 22:34:21 +0000 (15:34 -0700)]
Bluetooth: Always set event mask on suspend

When suspending, always set the event mask once disconnects are
successful. Otherwise, if wakeup is disallowed, the event mask is not
set before suspend continues and can result in an early wakeup.

Fixes: 182ee45da083 ("Bluetooth: hci_sync: Rework hci_suspend_notifier")
Cc: stable@vger.kernel.org
Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
2 years agoBluetooth: mgmt: Fix double free on error path
Dan Carpenter [Tue, 26 Jul 2022 15:13:44 +0000 (18:13 +0300)]
Bluetooth: mgmt: Fix double free on error path

Don't call mgmt_pending_remove() twice (double free).

Fixes: 6b88eff43704 ("Bluetooth: hci_sync: Refactor remove Adv Monitor")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
2 years agowifi: mac80211: do not abuse fq.lock in ieee80211_do_stop()
Tetsuo Handa [Sun, 17 Jul 2022 12:21:52 +0000 (21:21 +0900)]
wifi: mac80211: do not abuse fq.lock in ieee80211_do_stop()

lockdep complains use of uninitialized spinlock at ieee80211_do_stop() [1],
for commit f856373e2f31ffd3 ("wifi: mac80211: do not wake queues on a vif
that is being stopped") guards clear_bit() using fq.lock even before
fq_init() from ieee80211_txq_setup_flows() initializes this spinlock.

According to discussion [2], Toke was not happy with expanding usage of
fq.lock. Since __ieee80211_wake_txqs() is called under RCU read lock, we
can instead use synchronize_rcu() for flushing ieee80211_wake_txqs().

Link: https://syzkaller.appspot.com/bug?extid=eceab52db7c4b961e9d6
Link: https://lkml.kernel.org/r/874k0zowh2.fsf@toke.dk
Reported-by: syzbot <syzbot+eceab52db7c4b961e9d6@syzkaller.appspotmail.com>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fixes: f856373e2f31ffd3 ("wifi: mac80211: do not wake queues on a vif that is being stopped")
Tested-by: syzbot <syzbot+eceab52db7c4b961e9d6@syzkaller.appspotmail.com>
Acked-by: Toke Høiland-Jørgensen <toke@kernel.org>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/9cc9b81d-75a3-3925-b612-9d0ad3cab82b@I-love.SAKURA.ne.jp
[ pick up commit 3598cb6e1862 ("wifi: mac80211: do not abuse fq.lock in ieee80211_do_stop()") from -next]
Link: https://lore.kernel.org/all/87o7xcq6qt.fsf@kernel.org/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoice: do not setup vlan for loopback VSI
Maciej Fijalkowski [Thu, 7 Jul 2022 10:20:43 +0000 (12:20 +0200)]
ice: do not setup vlan for loopback VSI

Currently loopback test is failiing due to the error returned from
ice_vsi_vlan_setup(). Skip calling it when preparing loopback VSI.

Fixes: 0e674aeb0b77 ("ice: Add handler for ethtool selftest")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: George Kuruvinakunnel <george.kuruvinakunnel@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: check (DD | EOF) bits on Rx descriptor rather than (EOP | RS)
Maciej Fijalkowski [Thu, 7 Jul 2022 10:20:42 +0000 (12:20 +0200)]
ice: check (DD | EOF) bits on Rx descriptor rather than (EOP | RS)

Tx side sets EOP and RS bits on descriptors to indicate that a
particular descriptor is the last one and needs to generate an irq when
it was sent. These bits should not be checked on completion path
regardless whether it's the Tx or the Rx. DD bit serves this purpose and
it indicates that a particular descriptor is either for Rx or was
successfully Txed. EOF is also set as loopback test does not xmit
fragmented frames.

Look at (DD | EOF) bits setting in ice_lbtest_receive_frames() instead
of EOP and RS pair.

Fixes: 0e674aeb0b77 ("ice: Add handler for ethtool selftest")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: George Kuruvinakunnel <george.kuruvinakunnel@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: Fix VSIs unable to share unicast MAC
Anirudh Venkataramanan [Thu, 21 Jul 2022 08:03:09 +0000 (10:03 +0200)]
ice: Fix VSIs unable to share unicast MAC

The driver currently does not allow two VSIs in the same PF domain
to have the same unicast MAC address. This is incorrect in the sense
that a policy decision is being made in the driver when it must be
left to the user. This approach was causing issues when rebooting
the system with VFs spawned not being able to change their MAC addresses.
Such errors were present in dmesg:

[ 7921.068237] ice 0000:b6:00.2 ens2f2: Unicast MAC 6a:0d:e4:70:ca:d1 already
exists on this PF. Preventing setting VF 7 unicast MAC address to 6a:0d:e4:70:ca:d1

Fix that by removing this restriction. Doing this also allows
us to remove some additional code that's checking if a unicast MAC
filter already exists.

Fixes: 47ebc7b02485 ("ice: Check if unicast MAC exists before setting VF MAC")
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Sylwester Dziedziuch <sylwesterx.dziedziuch@intel.com>
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
Signed-off-by: Jedrzej Jagielski <jedrzej.jagielski@intel.com>
Tested-by: Marek Szlosek <marek.szlosek@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: Fix tunnel checksum offload with fragmented traffic
Przemyslaw Patynowski [Mon, 18 Jul 2022 11:34:27 +0000 (13:34 +0200)]
ice: Fix tunnel checksum offload with fragmented traffic

Fix checksum offload on VXLAN tunnels.
In case, when mpls protocol is not used, set l4 header to transport
header of skb. This fixes case, when user tries to offload checksums
of VXLAN tunneled traffic.

Steps for reproduction (requires link partner with tunnels):
ip l s enp130s0f0 up
ip a f enp130s0f0
ip a a 10.10.110.2/24 dev enp130s0f0
ip l s enp130s0f0 mtu 1600
ip link add vxlan12_sut type vxlan id 12 group 238.168.100.100 dev enp130s0f0 dstport 4789
ip l s vxlan12_sut up
ip a a 20.10.110.2/24 dev vxlan12_sut
iperf3 -c 20.10.110.1 #should connect

Offload params: td_offset, cd_tunnel_params were
corrupted, due to l4 header pointing wrong address. NIC would then drop
those packets internally, due to incorrect TX descriptor data,
which increased GLV_TEPC register.

Fixes: 69e66c04c672 ("ice: Add mpls+tso support")
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
Signed-off-by: Jedrzej Jagielski <jedrzej.jagielski@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: Fix max VLANs available for VF
Przemyslaw Patynowski [Mon, 4 Jul 2022 13:46:12 +0000 (15:46 +0200)]
ice: Fix max VLANs available for VF

Legacy VLAN implementation allows for untrusted VF to have 8 VLAN
filters, not counting VLAN 0 filters. Current VLAN_V2 implementation
lowers available filters for VF, by counting in VLAN 0 filter for both
TPIDs.
Fix this by counting only non zero VLAN filters.
Without this patch, untrusted VF would not be able to access 8 VLAN
filters.

Fixes: cc71de8fa133 ("ice: Add support for VIRTCHNL_VF_OFFLOAD_VLAN_V2")
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
Tested-by: Marek Szlosek <marek.szlosek@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agonetfilter: nft_queue: only allow supported familes and hooks
Florian Westphal [Tue, 26 Jul 2022 17:49:00 +0000 (19:49 +0200)]
netfilter: nft_queue: only allow supported familes and hooks

Trying to use 'queue' statement in ingress (for example)
triggers a splat on reinject:

WARNING: CPU: 3 PID: 1345 at net/netfilter/nf_queue.c:291

... because nf_reinject cannot find the ruleset head.

The netdev family doesn't support async resume at the moment anyway,
so disallow loading such rulesets with a more appropriate
error message.

v2: add 'validate' callback and also check hook points, v1 did
allow ingress use in 'table inet', but that doesn't work either. (Pablo)

Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agonetfilter: nf_tables: add rescheduling points during loop detection walks
Florian Westphal [Tue, 26 Jul 2022 10:44:35 +0000 (12:44 +0200)]
netfilter: nf_tables: add rescheduling points during loop detection walks

Add explicit rescheduling points during ruleset walk.

Switching to a faster algorithm is possible but this is a much
smaller change, suitable for nf tree.

Link: https://bugzilla.netfilter.org/show_bug.cgi?id=1460
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agonetfilter: nf_queue: do not allow packet truncation below transport header offset
Florian Westphal [Tue, 26 Jul 2022 10:42:06 +0000 (12:42 +0200)]
netfilter: nf_queue: do not allow packet truncation below transport header offset

Domingo Dirutigliano and Nicola Guerrera report kernel panic when
sending nf_queue verdict with 1-byte nfta_payload attribute.

The IP/IPv6 stack pulls the IP(v6) header from the packet after the
input hook.

If user truncates the packet below the header size, this skb_pull() will
result in a malformed skb (skb->len < 0).

Fixes: 7af4cc3fa158 ("[NETFILTER]: Add "nfnetlink_queue" netfilter queue handler over nfnetlink")
Reported-by: Domingo Dirutigliano <pwnzer0tt1@proton.me>
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agoice: Add support for PPPoE hardware offload
Marcin Szycik [Mon, 18 Jul 2022 12:18:13 +0000 (14:18 +0200)]
ice: Add support for PPPoE hardware offload

Add support for creating PPPoE filters in switchdev mode. Add support
for parsing PPPoE and PPP-specific tc options: pppoe_sid and ppp_proto.

Example filter:
tc filter add dev $PF1 ingress protocol ppp_ses prio 1 flower pppoe_sid \
    1234 ppp_proto ip skip_sw action mirred egress redirect dev $VF1_PR

Changes in iproute2 are required to use the new fields.

ICE COMMS DDP package is required to create a filter as it contains PPPoE
profiles. Added a warning message when loaded DDP package does not contain
required profiles.

Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoflow_offload: Introduce flow_match_pppoe
Wojciech Drewek [Mon, 18 Jul 2022 12:18:12 +0000 (14:18 +0200)]
flow_offload: Introduce flow_match_pppoe

Allow to offload PPPoE filters by adding flow_rule_match_pppoe.
Drivers can extract PPPoE specific fields from now on.

Signed-off-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agonet/sched: flower: Add PPPoE filter
Wojciech Drewek [Mon, 18 Jul 2022 12:18:11 +0000 (14:18 +0200)]
net/sched: flower: Add PPPoE filter

Add support for PPPoE specific fields for tc-flower.
Those fields can be provided only when protocol was set
to ETH_P_PPP_SES. Defines, dump, load and set are being done here.

Overwrite basic.n_proto only in case of PPP_IP and PPP_IPV6,
otherwise leave it as ETH_P_PPP_SES.

Signed-off-by: Wojciech Drewek <wojciech.drewek@intel.com>
Acked-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoMerge tag 's390-5.19-7' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Tue, 26 Jul 2022 17:03:53 +0000 (10:03 -0700)]
Merge tag 's390-5.19-7' of git://git./linux/kernel/git/s390/linux

Pull s390 fix from Alexander GordeevL

 - Prevent relatively slow PRNO TRNG random number operation from being
   called from interrupt context. That could for example cause some
   network loads to timeout.

* tag 's390-5.19-7' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/archrandom: prevent CPACF trng invocations in interrupt context

2 years agoflow_dissector: Add PPPoE dissectors
Wojciech Drewek [Mon, 18 Jul 2022 12:18:10 +0000 (14:18 +0200)]
flow_dissector: Add PPPoE dissectors

Allow to dissect PPPoE specific fields which are:
- session ID (16 bits)
- ppp protocol (16 bits)
- type (16 bits) - this is PPPoE ethertype, for now only
  ETH_P_PPP_SES is supported, possible ETH_P_PPP_DISC
  in the future

The goal is to make the following TC command possible:

  # tc filter add dev ens6f0 ingress prio 1 protocol ppp_ses \
      flower \
        pppoe_sid 12 \
        ppp_proto ip \
      action drop

Note that only PPPoE Session is supported.

Signed-off-by: Wojciech Drewek <wojciech.drewek@intel.com>
Acked-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>