platform/kernel/linux-starfive.git
5 years agoMerge branch 'bnxt_en-Bug-fixes'
David S. Miller [Sun, 18 Aug 2019 20:05:01 +0000 (13:05 -0700)]
Merge branch 'bnxt_en-Bug-fixes'

Michael Chan says:

====================
bnxt_en: Bug fixes.

2 Bug fixes related to 57500 shutdown sequence and doorbell sequence,
2 TC Flower bug fixes related to the setting of the flow direction,
1 NVRAM update bug fix, and a minor fix to suppress an unnecessary
error message.  Please queue for -stable as well.  Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Fix to include flow direction in L2 key
Somnath Kotur [Sat, 17 Aug 2019 21:04:52 +0000 (17:04 -0400)]
bnxt_en: Fix to include flow direction in L2 key

FW expects the driver to provide unique flow reference handles
for Tx or Rx flows. When a Tx flow and an Rx flow end up sharing
a reference handle, flow offload does not seem to work.
This could happen in the case of 2 flows having their L2 fields
wildcarded but in different direction.
Fix to incorporate the flow direction as part of the L2 key

v2: Move the dir field to the end of the bnxt_tc_l2_key struct to
fix the warning reported by kbuild test robot <lkp@intel.com>.
There is existing code that initializes the structure using
nested initializer and will warn with the new u8 field added to
the beginning.  The structure also packs nicer when this new u8 is
added to the end of the structure [MChan].

Fixes: abd43a13525d ("bnxt_en: Support for 64-bit flow handle.")
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Use correct src_fid to determine direction of the flow
Venkat Duvvuru [Sat, 17 Aug 2019 21:04:51 +0000 (17:04 -0400)]
bnxt_en: Use correct src_fid to determine direction of the flow

Direction of the flow is determined using src_fid. For an RX flow,
src_fid is PF's fid and for TX flow, src_fid is VF's fid. Direction
of the flow must be specified, when getting statistics for that flow.
Currently, for DECAP flow, direction is determined incorrectly, i.e.,
direction is initialized as TX for DECAP flow, instead of RX. Because
of which, stats are not reported for this DECAP flow, though it is
offloaded and there is traffic for that flow, resulting in flow age out.

This patch fixes the problem by determining the DECAP flow's direction
using correct fid.  Set the flow direction in all cases for consistency
even if 64-bit flow handle is not used.

Fixes: abd43a13525d ("bnxt_en: Support for 64-bit flow handle.")
Signed-off-by: Venkat Duvvuru <venkatkumar.duvvuru@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Suppress HWRM errors for HWRM_NVM_GET_VARIABLE command
Vasundhara Volam [Sat, 17 Aug 2019 21:04:50 +0000 (17:04 -0400)]
bnxt_en: Suppress HWRM errors for HWRM_NVM_GET_VARIABLE command

For newly added NVM parameters, older firmware may not have the support.
Suppress the error message to avoid the unncessary error message which is
triggered when devlink calls the driver during initialization.

Fixes: 782a624d00fa ("bnxt_en: Add bnxt_en initial params table and register it.")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Fix handling FRAG_ERR when NVM_INSTALL_UPDATE cmd fails
Vasundhara Volam [Sat, 17 Aug 2019 21:04:49 +0000 (17:04 -0400)]
bnxt_en: Fix handling FRAG_ERR when NVM_INSTALL_UPDATE cmd fails

If FW returns FRAG_ERR in response error code, driver is resending the
command only when HWRM command returns success. Fix the code to resend
NVM_INSTALL_UPDATE command with DEFRAG install flags, if FW returns
FRAG_ERR in its response error code.

Fixes: cb4d1d626145 ("bnxt_en: Retry failed NVM_INSTALL_UPDATE with defragmentation flag enabled.")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Improve RX doorbell sequence.
Michael Chan [Sat, 17 Aug 2019 21:04:48 +0000 (17:04 -0400)]
bnxt_en: Improve RX doorbell sequence.

When both RX buffers and RX aggregation buffers have to be
replenished at the end of NAPI, post the RX aggregation buffers first
before RX buffers.  Otherwise, we may run into a situation where
there are only RX buffers without RX aggregation buffers for a split
second.  This will cause the hardware to abort the RX packet and
report buffer errors, which will cause unnecessary cleanup by the
driver.

Ringing the Aggregation ring doorbell first before the RX ring doorbell
will prevent some of these buffer errors.  Use the same sequence during
ring initialization as well.

Fixes: 697197e5a173 ("bnxt_en: Re-structure doorbells.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Fix VNIC clearing logic for 57500 chips.
Michael Chan [Sat, 17 Aug 2019 21:04:47 +0000 (17:04 -0400)]
bnxt_en: Fix VNIC clearing logic for 57500 chips.

During device shutdown, the VNIC clearing sequence needs to be modified
to free the VNIC first before freeing the RSS contexts.  The current
code is doing the reverse and we can get mis-directed RX completions
to CP ring ID 0 when the RSS contexts are freed and zeroed.  The clearing
of RSS contexts is not required with the new sequence.

Refactor the VNIC clearing logic into a new function bnxt_clear_vnic()
and do the chip specific VNIC clearing sequence.

Fixes: 7b3af4f75b81 ("bnxt_en: Add RSS support for 57500 chips.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: kalmia: fix memory leaks
Wenwen Wang [Wed, 14 Aug 2019 18:56:43 +0000 (13:56 -0500)]
net: kalmia: fix memory leaks

In kalmia_init_and_get_ethernet_addr(), 'usb_buf' is allocated through
kmalloc(). In the following execution, if the 'status' returned by
kalmia_send_init_packet() is not 0, 'usb_buf' is not deallocated, leading
to memory leaks. To fix this issue, add the 'out' label to free 'usb_buf'.

Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agocx82310_eth: fix a memory leak bug
Wenwen Wang [Wed, 14 Aug 2019 18:03:38 +0000 (13:03 -0500)]
cx82310_eth: fix a memory leak bug

In cx82310_bind(), 'dev->partial_data' is allocated through kmalloc().
Then, the execution waits for the firmware to become ready. If the firmware
is not ready in time, the execution is terminated. However, the allocated
'dev->partial_data' is not deallocated on this path, leading to a memory
leak bug. To fix this issue, free 'dev->partial_data' before returning the
error.

Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnx2x: Fix VF's VLAN reconfiguration in reload.
Manish Chopra [Sun, 18 Aug 2019 14:25:48 +0000 (07:25 -0700)]
bnx2x: Fix VF's VLAN reconfiguration in reload.

Commit 04f05230c5c13 ("bnx2x: Remove configured vlans as
part of unload sequence."), introduced a regression in driver
that as a part of VF's reload flow, VLANs created on the VF
doesn't get re-configured in hardware as vlan metadata/info
was not getting cleared for the VFs which causes vlan PING to stop.

This patch clears the vlan metadata/info so that VLANs gets
re-configured back in the hardware in VF's reload flow and
PING/traffic continues for VLANs created over the VFs.

Fixes: 04f05230c5c13 ("bnx2x: Remove configured vlans as part of unload sequence.")
Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Sudarsana Kalluru <skalluru@marvell.com>
Signed-off-by: Shahed Shaikh <shshaikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Sat, 17 Aug 2019 19:37:47 +0000 (12:37 -0700)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth

Johan Hedberg says:

====================
pull request: bluetooth 2019-08-17

Here's a set of Bluetooth fixes for the 5.3-rc series:

 - Multiple fixes for Qualcomm (btqca & hci_qca) drivers
 - Minimum encryption key size debugfs setting (this is required for
   Bluetooth Qualification)
 - Fix hidp_send_message() to have a meaningful return value
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoBluetooth: Add debug setting for changing minimum encryption key size
Marcel Holtmann [Tue, 16 Jul 2019 18:34:41 +0000 (20:34 +0200)]
Bluetooth: Add debug setting for changing minimum encryption key size

For testing and qualification purposes it is useful to allow changing
the minimum encryption key size value that the host stack is going to
enforce. This adds a new debugfs setting min_encrypt_key_size to achieve
this functionality.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
5 years agotipc: fix false detection of retransmit failures
Tuong Lien [Thu, 15 Aug 2019 03:24:08 +0000 (10:24 +0700)]
tipc: fix false detection of retransmit failures

This commit eliminates the use of the link 'stale_limit' & 'prev_from'
(besides the already removed - 'stale_cnt') variables in the detection
of repeated retransmit failures as there is no proper way to initialize
them to avoid a false detection, i.e. it is not really a retransmission
failure but due to a garbage values in the variables.

Instead, a jiffies variable will be added to individual skbs (like the
way we restrict the skb retransmissions) in order to mark the first skb
retransmit time. Later on, at the next retransmissions, the timestamp
will be checked to see if the skb in the link transmq is "too stale",
that is, the link tolerance time has passed, so that a link reset will
be ordered. Note, just checking on the first skb in the queue is fine
enough since it must be the oldest one.
A counter is also added to keep track the actual skb retransmissions'
number for later checking when the failure happens.

The downside of this approach is that the skb->cb[] buffer is about to
be exhausted, however it is always able to allocate another memory area
and keep a reference to it when needed.

Fixes: 77cf8edbc0e7 ("tipc: simplify stale link failure criteria")
Reported-by: Hoang Le <hoang.h.le@dektech.com.au>
Acked-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agolan78xx: Fix memory leaks
Wenwen Wang [Wed, 14 Aug 2019 16:23:13 +0000 (11:23 -0500)]
lan78xx: Fix memory leaks

In lan78xx_probe(), a new urb is allocated through usb_alloc_urb() and
saved to 'dev->urb_intr'. However, in the following execution, if an error
occurs, 'dev->urb_intr' is not deallocated, leading to memory leaks. To fix
this issue, invoke usb_free_urb() to free the allocated urb before
returning from the function.

Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMAINTAINERS: r8169: Update path to the driver
Denis Efremov [Wed, 14 Aug 2019 12:12:09 +0000 (15:12 +0300)]
MAINTAINERS: r8169: Update path to the driver

Update MAINTAINERS record to reflect the filename change.
The file was moved in commit 25e992a4603c ("r8169: rename
r8169.c to r8169_main.c")

Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: nic_swsd@realtek.com
Cc: David S. Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Denis Efremov <efremov@linux.com>
Reviewed-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMAINTAINERS: PHY LIBRARY: Update files in the record
Denis Efremov [Wed, 14 Aug 2019 12:58:00 +0000 (15:58 +0300)]
MAINTAINERS: PHY LIBRARY: Update files in the record

Update MAINTAINERS to reflect that sysfs-bus-mdio was removed in
commit a6cd0d2d493a ("Documentation: net-sysfs: Remove duplicate
PHY device documentation") and sysfs-class-net-phydev was added in
commit 86f22d04dfb5 ("net: sysfs: Document PHY device sysfs
attributes").

Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Denis Efremov <efremov@linux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'rxrpc-fixes-20190814' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Thu, 15 Aug 2019 23:33:22 +0000 (16:33 -0700)]
Merge tag 'rxrpc-fixes-20190814' of git://git./linux/kernel/git/dhowells/linux-fs

David Howells says:

====================
rxrpc: Fix local endpoint handling

Here's a pair of patches that fix two issues in the handling of local
endpoints (rxrpc_local structs):

 (1) Use list_replace_init() rather than list_replace() if we're going to
     unconditionally delete the replaced item later, lest the list get
     corrupted.

 (2) Don't access the rxrpc_local object after passing our ref to the
     workqueue, not even to illuminate tracepoints, as the work function
     may cause the object to be freed.  We have to cache the information
     beforehand.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Thu, 15 Aug 2019 21:01:14 +0000 (14:01 -0700)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

This patchset contains Netfilter fixes for net:

1) Extend selftest to cover flowtable with ipsec, from Florian Westphal.

2) Fix interaction of ipsec with flowtable, also from Florian.

3) User-after-free with bound set to rule that fails to load.

4) Adjust state and timeout for flows that expire.

5) Timeout update race with flows in teardown state.

6) Ensure conntrack id hash calculation use invariants as input,
   from Dirk Morris.

7) Do not push flows into flowtable for TCP fin/rst packets.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/packet: fix race in tpacket_snd()
Eric Dumazet [Wed, 14 Aug 2019 09:11:57 +0000 (02:11 -0700)]
net/packet: fix race in tpacket_snd()

packet_sendmsg() checks tx_ring.pg_vec to decide
if it must call tpacket_snd().

Problem is that the check is lockless, meaning another thread
can issue a concurrent setsockopt(PACKET_TX_RING ) to flip
tx_ring.pg_vec back to NULL.

Given that tpacket_snd() grabs pg_vec_lock mutex, we can
perform the check again to solve the race.

syzbot reported :

kasan: CONFIG_KASAN_INLINE enabled
kasan: GPF could be caused by NULL-ptr deref or user memory access
general protection fault: 0000 [#1] PREEMPT SMP KASAN
CPU: 1 PID: 11429 Comm: syz-executor394 Not tainted 5.3.0-rc4+ #101
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
RIP: 0010:packet_lookup_frame+0x8d/0x270 net/packet/af_packet.c:474
Code: c1 ee 03 f7 73 0c 80 3c 0e 00 0f 85 cb 01 00 00 48 8b 0b 89 c0 4c 8d 24 c1 48 b8 00 00 00 00 00 fc ff df 4c 89 e1 48 c1 e9 03 <80> 3c 01 00 0f 85 94 01 00 00 48 8d 7b 10 4d 8b 3c 24 48 b8 00 00
RSP: 0018:ffff88809f82f7b8 EFLAGS: 00010246
RAX: dffffc0000000000 RBX: ffff8880a45c7030 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 1ffff110148b8e06 RDI: ffff8880a45c703c
RBP: ffff88809f82f7e8 R08: ffff888087aea200 R09: fffffbfff134ae50
R10: fffffbfff134ae4f R11: ffffffff89a5727f R12: 0000000000000000
R13: 0000000000000001 R14: ffff8880a45c6ac0 R15: 0000000000000000
FS:  00007fa04716f700(0000) GS:ffff8880ae900000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fa04716edb8 CR3: 0000000091eb4000 CR4: 00000000001406e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 packet_current_frame net/packet/af_packet.c:487 [inline]
 tpacket_snd net/packet/af_packet.c:2667 [inline]
 packet_sendmsg+0x590/0x6250 net/packet/af_packet.c:2975
 sock_sendmsg_nosec net/socket.c:637 [inline]
 sock_sendmsg+0xd7/0x130 net/socket.c:657
 ___sys_sendmsg+0x3e2/0x920 net/socket.c:2311
 __sys_sendmmsg+0x1bf/0x4d0 net/socket.c:2413
 __do_sys_sendmmsg net/socket.c:2442 [inline]
 __se_sys_sendmmsg net/socket.c:2439 [inline]
 __x64_sys_sendmmsg+0x9d/0x100 net/socket.c:2439
 do_syscall_64+0xfd/0x6a0 arch/x86/entry/common.c:296
 entry_SYSCALL_64_after_hwframe+0x49/0xbe

Fixes: 69e3c75f4d54 ("net: TX_RING and packet mmap")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: myri10ge: fix memory leaks
Wenwen Wang [Wed, 14 Aug 2019 06:38:39 +0000 (01:38 -0500)]
net: myri10ge: fix memory leaks

In myri10ge_probe(), myri10ge_alloc_slices() is invoked to allocate slices
related structures. Later on, myri10ge_request_irq() is used to get an irq.
However, if this process fails, the allocated slices related structures are
not deallocated, leading to memory leaks. To fix this issue, revise the
target label of the goto statement to 'abort_with_slices'.

Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: tls, fix sk_write_space NULL write when tx disabled
John Fastabend [Wed, 14 Aug 2019 05:31:54 +0000 (05:31 +0000)]
net: tls, fix sk_write_space NULL write when tx disabled

The ctx->sk_write_space pointer is only set when TLS tx mode is enabled.
When running without TX mode its a null pointer but we still set the
sk sk_write_space pointer on close().

Fix the close path to only overwrite sk->sk_write_space when the current
pointer is to the tls_write_space function indicating the tls module should
clean it up properly as well.

Reported-by: Hillf Danton <hdanton@sina.com>
Cc: Ying Xue <ying.xue@windriver.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Fixes: 57c722e932cfb ("net/tls: swap sk_write_space on close")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoliquidio: add cleanup in octeon_setup_iq()
Wenwen Wang [Wed, 14 Aug 2019 05:14:49 +0000 (00:14 -0500)]
liquidio: add cleanup in octeon_setup_iq()

If oct->fn_list.enable_io_queues() fails, no cleanup is executed, leading
to memory/resource leaks. To fix this issue, invoke
octeon_delete_instr_queue() before returning from the function.

Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mlx5-fixes-2019-08-15' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Thu, 15 Aug 2019 19:09:39 +0000 (12:09 -0700)]
Merge tag 'mlx5-fixes-2019-08-15' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
Mellanox, mlx5 fixes 2019-08-15

This series introduces two fixes to mlx5 driver.

1) Eran fixes a compatibility issue with ethtool flash.
2) Maxim fixes a race in XSK wakeup flow.

Please pull and let me know if there is any problem.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx5e: Fix compatibility issue with ethtool flash device
Eran Ben Elisha [Thu, 1 Aug 2019 11:27:30 +0000 (14:27 +0300)]
net/mlx5e: Fix compatibility issue with ethtool flash device

Cited patch deleted ethtool flash device support, as ethtool core can
fallback into devlink flash callback. However, this is supported only if
there is a devlink port registered over the corresponding netdevice.

As mlx5e do not have devlink port support over native netdevice, it broke
the ability to flash device via ethtool.

This patch re-add the ethtool callback to avoid user functionality breakage
when trying to flash device via ethtool.

Fixes: 9c8bca2637b8 ("mlx5: Move firmware flash implementation to devlink")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Fix a race with XSKICOSQ in XSK wakeup flow
Maxim Mikityanskiy [Fri, 2 Aug 2019 11:42:14 +0000 (14:42 +0300)]
net/mlx5e: Fix a race with XSKICOSQ in XSK wakeup flow

Add a missing spinlock around XSKICOSQ usage at the activation stage,
because there is a race between a configuration change and the
application calling sendto().

Fixes: db05815b36cb ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agoselftests: net: tcp_fastopen_backup_key.sh: fix shellcheck issue
Anders Roxell [Thu, 15 Aug 2019 07:58:26 +0000 (09:58 +0200)]
selftests: net: tcp_fastopen_backup_key.sh: fix shellcheck issue

When running tcp_fastopen_backup_key.sh the following issue was seen in
a busybox environment.
./tcp_fastopen_backup_key.sh: line 33: [: -ne: unary operator expected

Shellcheck showed the following issue.
$ shellcheck tools/testing/selftests/net/tcp_fastopen_backup_key.sh

In tools/testing/selftests/net/tcp_fastopen_backup_key.sh line 33:
        if [ $val -ne 0 ]; then
             ^-- SC2086: Double quote to prevent globbing and word splitting.

Rework to do a string comparison instead.

Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agocxgb4: fix a memory leak bug
Wenwen Wang [Tue, 13 Aug 2019 09:18:52 +0000 (04:18 -0500)]
cxgb4: fix a memory leak bug

In blocked_fl_write(), 't' is not deallocated if bitmap_parse_user() fails,
leading to a memory leak bug. To fix this issue, free t before returning
the error.

Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoBluetooth: hci_qca: Skip 1 error print in device_want_to_sleep()
Rocky Liao [Wed, 14 Aug 2019 07:42:39 +0000 (15:42 +0800)]
Bluetooth: hci_qca: Skip 1 error print in device_want_to_sleep()

Don't fall through to print error message when receive sleep indication
in HCI_IBS_RX_ASLEEP state, this is allowed behavior.

Signed-off-by: Rocky Liao <rjliao@codeaurora.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agorxrpc: Fix read-after-free in rxrpc_queue_local()
David Howells [Tue, 13 Aug 2019 21:26:36 +0000 (22:26 +0100)]
rxrpc: Fix read-after-free in rxrpc_queue_local()

rxrpc_queue_local() attempts to queue the local endpoint it is given and
then, if successful, prints a trace line.  The trace line includes the
current usage count - but we're not allowed to look at the local endpoint
at this point as we passed our ref on it to the workqueue.

Fix this by reading the usage count before queuing the work item.

Also fix the reading of local->debug_id for trace lines, which must be done
with the same consideration as reading the usage count.

Fixes: 09d2bf595db4 ("rxrpc: Add a tracepoint to track rxrpc_local refcounting")
Reported-by: syzbot+78e71c5bab4f76a6a719@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agorxrpc: Fix local endpoint replacement
David Howells [Mon, 12 Aug 2019 22:30:06 +0000 (23:30 +0100)]
rxrpc: Fix local endpoint replacement

When a local endpoint (struct rxrpc_local) ceases to be in use by any
AF_RXRPC sockets, it starts the process of being destroyed, but this
doesn't cause it to be removed from the namespace endpoint list immediately
as tearing it down isn't trivial and can't be done in softirq context, so
it gets deferred.

If a new socket comes along that wants to bind to the same endpoint, a new
rxrpc_local object will be allocated and rxrpc_lookup_local() will use
list_replace() to substitute the new one for the old.

Then, when the dying object gets to rxrpc_local_destroyer(), it is removed
unconditionally from whatever list it is on by calling list_del_init().

However, list_replace() doesn't reset the pointers in the replaced
list_head and so the list_del_init() will likely corrupt the local
endpoints list.

Fix this by using list_replace_init() instead.

Fixes: 730c5fd42c1e ("rxrpc: Fix local endpoint refcounting")
Reported-by: syzbot+193e29e9387ea5837f1d@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agonetfilter: nft_flow_offload: skip tcp rst and fin packets
Pablo Neira Ayuso [Tue, 13 Aug 2019 15:41:13 +0000 (17:41 +0200)]
netfilter: nft_flow_offload: skip tcp rst and fin packets

TCP rst and fin packets do not qualify to place a flow into the
flowtable. Most likely there will be no more packets after connection
closure. Without this patch, this flow entry expires and connection
tracking picks up the entry in ESTABLISHED state using the fixup
timeout, which makes this look inconsistent to the user for a connection
that is actually already closed.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
5 years agosctp: fix memleak in sctp_send_reset_streams
zhengbin [Tue, 13 Aug 2019 14:05:50 +0000 (22:05 +0800)]
sctp: fix memleak in sctp_send_reset_streams

If the stream outq is not empty, need to kfree nstr_list.

Fixes: d570a59c5b5f ("sctp: only allow the out stream reset when the stream outq is empty")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: zhengbin <zhengbin13@huawei.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
5 years agonetlink: Fix nlmsg_parse as a wrapper for strict message parsing
David Ahern [Mon, 12 Aug 2019 20:07:07 +0000 (13:07 -0700)]
netlink: Fix nlmsg_parse as a wrapper for strict message parsing

Eric reported a syzbot warning:

BUG: KMSAN: uninit-value in nh_valid_get_del_req+0x6f1/0x8c0 net/ipv4/nexthop.c:1510
CPU: 0 PID: 11812 Comm: syz-executor444 Not tainted 5.3.0-rc3+ #17
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x191/0x1f0 lib/dump_stack.c:113
 kmsan_report+0x162/0x2d0 mm/kmsan/kmsan_report.c:109
 __msan_warning+0x75/0xe0 mm/kmsan/kmsan_instr.c:294
 nh_valid_get_del_req+0x6f1/0x8c0 net/ipv4/nexthop.c:1510
 rtm_del_nexthop+0x1b1/0x610 net/ipv4/nexthop.c:1543
 rtnetlink_rcv_msg+0x115a/0x1580 net/core/rtnetlink.c:5223
 netlink_rcv_skb+0x431/0x620 net/netlink/af_netlink.c:2477
 rtnetlink_rcv+0x50/0x60 net/core/rtnetlink.c:5241
 netlink_unicast_kernel net/netlink/af_netlink.c:1302 [inline]
 netlink_unicast+0xf6c/0x1050 net/netlink/af_netlink.c:1328
 netlink_sendmsg+0x110f/0x1330 net/netlink/af_netlink.c:1917
 sock_sendmsg_nosec net/socket.c:637 [inline]
 sock_sendmsg net/socket.c:657 [inline]
 ___sys_sendmsg+0x14ff/0x1590 net/socket.c:2311
 __sys_sendmmsg+0x53a/0xae0 net/socket.c:2413
 __do_sys_sendmmsg net/socket.c:2442 [inline]
 __se_sys_sendmmsg+0xbd/0xe0 net/socket.c:2439
 __x64_sys_sendmmsg+0x56/0x70 net/socket.c:2439
 do_syscall_64+0xbc/0xf0 arch/x86/entry/common.c:297
 entry_SYSCALL_64_after_hwframe+0x63/0xe7

The root cause is nlmsg_parse calling __nla_parse which means the
header struct size is not checked.

nlmsg_parse should be a wrapper around __nlmsg_parse with
NL_VALIDATE_STRICT for the validate argument very much like
nlmsg_parse_deprecated is for NL_VALIDATE_LIBERAL.

Fixes: 3de6440354465 ("netlink: re-add parse/validate functions in strict mode")
Reported-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
5 years agonet: phy: consider AN_RESTART status when reading link status
Heiner Kallweit [Mon, 12 Aug 2019 19:20:02 +0000 (21:20 +0200)]
net: phy: consider AN_RESTART status when reading link status

After configuring and restarting aneg we immediately try to read the
link status. On some systems the PHY may not yet have cleared the
"aneg complete" and "link up" bits, resulting in a false link-up
signal. See [0] for a report.
Clause 22 and 45 both require the PHY to keep the AN_RESTART
bit set until the PHY actually starts auto-negotiation.
Let's consider this in the generic functions for reading link status.
The commit marked as fixed is the first one where the patch applies
cleanly.

[0] https://marc.info/?t=156518400300003&r=1&w=2

Fixes: c1164bb1a631 ("net: phy: check PMAPMD link status only in genphy_c45_read_link")
Tested-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
5 years agonet/mlx4_en: fix a memory leak bug
Wenwen Wang [Mon, 12 Aug 2019 19:11:35 +0000 (14:11 -0500)]
net/mlx4_en: fix a memory leak bug

In mlx4_en_config_rss_steer(), 'rss_map->indir_qp' is allocated through
kzalloc(). After that, mlx4_qp_alloc() is invoked to configure RSS
indirection. However, if mlx4_qp_alloc() fails, the allocated
'rss_map->indir_qp' is not deallocated, leading to a memory leak bug.

To fix the above issue, add the 'qp_alloc_err' label to free
'rss_map->indir_qp'.

Fixes: 4931c6ef04b4 ("net/mlx4_en: Optimized single ring steering")
Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
5 years agoibmveth: Convert multicast list size for little-endian system
Thomas Falcon [Mon, 12 Aug 2019 21:13:06 +0000 (16:13 -0500)]
ibmveth: Convert multicast list size for little-endian system

The ibm,mac-address-filters property defines the maximum number of
addresses the hypervisor's multicast filter list can support. It is
encoded as a big-endian integer in the OF device tree, but the virtual
ethernet driver does not convert it for use by little-endian systems.
As a result, the driver is not behaving as it should on affected systems
when a large number of multicast addresses are assigned to the device.

Reported-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
5 years agos390/qeth: serialize cmd reply with concurrent timeout
Julian Wiedmann [Mon, 12 Aug 2019 14:44:35 +0000 (16:44 +0200)]
s390/qeth: serialize cmd reply with concurrent timeout

Callbacks for a cmd reply run outside the protection of card->lock, to
allow for additional cmds to be issued & enqueued in parallel.

When qeth_send_control_data() bails out for a cmd without having
received a reply (eg. due to timeout), its callback may concurrently be
processing a reply that just arrived. In this case, the callback
potentially accesses a stale reply->reply_param area that eg. was
on-stack and has already been released.

To avoid this race, add some locking so that qeth_send_control_data()
can (1) wait for a concurrently running callback, and (2) zap any
pending callback that still wants to run.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
5 years agosctp: fix the transport error_count check
Xin Long [Mon, 12 Aug 2019 12:49:12 +0000 (20:49 +0800)]
sctp: fix the transport error_count check

As the annotation says in sctp_do_8_2_transport_strike():

  "If the transport error count is greater than the pf_retrans
   threshold, and less than pathmaxrtx ..."

It should be transport->error_count checked with pathmaxrxt,
instead of asoc->pf_retrans.

Fixes: 5aa93bcf66f4 ("sctp: Implement quick failover draft from tsvwg")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
5 years agonetfilter: conntrack: Use consistent ct id hash calculation
Dirk Morris [Thu, 8 Aug 2019 20:57:51 +0000 (13:57 -0700)]
netfilter: conntrack: Use consistent ct id hash calculation

Change ct id hash calculation to only use invariants.

Currently the ct id hash calculation is based on some fields that can
change in the lifetime on a conntrack entry in some corner cases. The
current hash uses the whole tuple which contains an hlist pointer which
will change when the conntrack is placed on the dying list resulting in
a ct id change.

This patch also removes the reply-side tuple and extension pointer from
the hash calculation so that the ct id will will not change from
initialization until confirmation.

Fixes: 3c79107631db1f7 ("netfilter: ctnetlink: don't use conntrack/expect object addresses as id")
Signed-off-by: Dirk Morris <dmorris@metaloft.com>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
5 years agonet: phy: at803x: stop switching phy delay config needlessly
André Draszik [Fri, 9 Aug 2019 11:20:25 +0000 (12:20 +0100)]
net: phy: at803x: stop switching phy delay config needlessly

This driver does a funny dance disabling and re-enabling
RX and/or TX delays. In any of the RGMII-ID modes, it first
disables the delays, just to re-enable them again right
away. This looks like a needless exercise.

Just enable the respective delays when in any of the
relevant 'id' modes, and disable them otherwise.

Also, remove comments which don't add anything that can't be
seen by looking at the code.

Signed-off-by: André Draszik <git@andred.net>
CC: Andrew Lunn <andrew@lunn.ch>
CC: Florian Fainelli <f.fainelli@gmail.com>
CC: Heiner Kallweit <hkallweit1@gmail.com>
CC: "David S. Miller" <davem@davemloft.net>
CC: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoBluetooth: btqca: Reset download type to default
Balakrishna Godavarthi [Thu, 8 Aug 2019 08:56:08 +0000 (14:26 +0530)]
Bluetooth: btqca: Reset download type to default

This patch will reset the download flag to default value
before retrieving the download mode type.

Fixes: 32646db8cc28 ("Bluetooth: btqca: inject command complete event during fw download")
Signed-off-by: Balakrishna Godavarthi <bgodavar@codeaurora.org>
Tested-by: Claire Chang <tientzu@chromium.org>
Reviewed-by: Claire Chang <tientzu@chromium.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: btqca: release_firmware after qca_inject_cmd_complete_event
Claire Chang [Tue, 6 Aug 2019 09:56:29 +0000 (17:56 +0800)]
Bluetooth: btqca: release_firmware after qca_inject_cmd_complete_event

commit 32646db8cc28 ("Bluetooth: btqca: inject command complete event
during fw download") added qca_inject_cmd_complete_event() for certain
qualcomm chips. However, qca_download_firmware() will return without
calling release_firmware() in this case.

This leads to a memory leak like the following found by kmemleak:

unreferenced object 0xfffffff3868a5880 (size 128):
  comm "kworker/u17:5", pid 347, jiffies 4294676481 (age 312.157s)
  hex dump (first 32 bytes):
    ac fd 00 00 00 00 00 00 00 d0 7e 17 80 ff ff ff  ..........~.....
    00 00 00 00 00 00 00 00 00 59 8a 86 f3 ff ff ff  .........Y......
  backtrace:
    [<00000000978ce31d>] kmem_cache_alloc_trace+0x194/0x298
    [<000000006ea0398c>] _request_firmware+0x74/0x4e4
    [<000000004da31ca0>] request_firmware+0x44/0x64
    [<0000000094572996>] qca_download_firmware+0x74/0x6e4 [btqca]
    [<00000000b24d615a>] qca_uart_setup+0xc0/0x2b0 [btqca]
    [<00000000364a6d5a>] qca_setup+0x204/0x570 [hci_uart]
    [<000000006be1a544>] hci_uart_setup+0xa8/0x148 [hci_uart]
    [<00000000d64c0f4f>] hci_dev_do_open+0x144/0x530 [bluetooth]
    [<00000000f69f5110>] hci_power_on+0x84/0x288 [bluetooth]
    [<00000000d4151583>] process_one_work+0x210/0x420
    [<000000003cf3dcfb>] worker_thread+0x2c4/0x3e4
    [<000000007ccaf055>] kthread+0x124/0x134
    [<00000000bef1f723>] ret_from_fork+0x10/0x18
    [<00000000c36ee3dd>] 0xffffffffffffffff
unreferenced object 0xfffffff37b16de00 (size 128):
  comm "kworker/u17:5", pid 347, jiffies 4294676873 (age 311.766s)
  hex dump (first 32 bytes):
    da 07 00 00 00 00 00 00 00 50 ff 0b 80 ff ff ff  .........P......
    00 00 00 00 00 00 00 00 00 dd 16 7b f3 ff ff ff  ...........{....
  backtrace:
    [<00000000978ce31d>] kmem_cache_alloc_trace+0x194/0x298
    [<000000006ea0398c>] _request_firmware+0x74/0x4e4
    [<000000004da31ca0>] request_firmware+0x44/0x64
    [<0000000094572996>] qca_download_firmware+0x74/0x6e4 [btqca]
    [<000000000cde20a9>] qca_uart_setup+0x144/0x2b0 [btqca]
    [<00000000364a6d5a>] qca_setup+0x204/0x570 [hci_uart]
    [<000000006be1a544>] hci_uart_setup+0xa8/0x148 [hci_uart]
    [<00000000d64c0f4f>] hci_dev_do_open+0x144/0x530 [bluetooth]
    [<00000000f69f5110>] hci_power_on+0x84/0x288 [bluetooth]
    [<00000000d4151583>] process_one_work+0x210/0x420
    [<000000003cf3dcfb>] worker_thread+0x2c4/0x3e4
    [<000000007ccaf055>] kthread+0x124/0x134
    [<00000000bef1f723>] ret_from_fork+0x10/0x18
    [<00000000c36ee3dd>] 0xffffffffffffffff

Make sure release_firmware() is called aftre
qca_inject_cmd_complete_event() to avoid the memory leak.

Fixes: 32646db8cc28 ("Bluetooth: btqca: inject command complete event during fw download")
Signed-off-by: Claire Chang <tientzu@chromium.org>
Reviewed-by: Balakrishna Godavarthi <bgodavar@codeaurora.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hidp: Let hidp_send_message return number of queued bytes
Fabian Henneke [Mon, 15 Jul 2019 17:40:56 +0000 (19:40 +0200)]
Bluetooth: hidp: Let hidp_send_message return number of queued bytes

Let hidp_send_message return the number of successfully queued bytes
instead of an unconditional 0.

With the return value fixed to 0, other drivers relying on hidp, such as
hidraw, can not return meaningful values from their respective
implementations of write(). In particular, with the current behavior, a
hidraw device's write() will have different return values depending on
whether the device is connected via USB or Bluetooth, which makes it
harder to abstract away the transport layer.

Signed-off-by: Fabian Henneke <fabian.henneke@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_qca: Send VS pre shutdown command.
Harish Bandi [Fri, 12 Jul 2019 05:09:40 +0000 (10:39 +0530)]
Bluetooth: hci_qca: Send VS pre shutdown command.

WCN399x chips are coex chips, it needs a VS pre shutdown
command while turning off the BT. So that chip can inform
BT is OFF to other active clients.

Signed-off-by: Harish Bandi <c-hbandi@codeaurora.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: btqca: Use correct byte format for opcode of injected command
Matthias Kaehlcke [Mon, 8 Jul 2019 21:57:42 +0000 (14:57 -0700)]
Bluetooth: btqca: Use correct byte format for opcode of injected command

The opcode of the command injected by commit 32646db8cc28 ("Bluetooth:
btqca: inject command complete event during fw download") uses the CPU
byte format, however it should always be little endian. In practice it
shouldn't really matter, since all we need is an opcode != 0, but still
let's do things correctly and keep sparse happy.

Fixes: 32646db8cc28 ("Bluetooth: btqca: inject command complete event during fw download")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_qca: Use kfree_skb() instead of kfree()
Wei Yongjun [Tue, 9 Jul 2019 01:35:30 +0000 (01:35 +0000)]
Bluetooth: hci_qca: Use kfree_skb() instead of kfree()

Use kfree_skb() instead of kfree() to free sk_buff.

Fixes: 2faa3f15fa2f ("Bluetooth: hci_qca: wcn3990: Drop baudrate change vendor event")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: btqca: Add a short delay before downloading the NVM
Matthias Kaehlcke [Tue, 9 Jul 2019 22:44:50 +0000 (15:44 -0700)]
Bluetooth: btqca: Add a short delay before downloading the NVM

On WCN3990 downloading the NVM sometimes fails with a "TLV response
size mismatch" error:

[  174.949955] Bluetooth: btqca.c:qca_download_firmware() hci0: QCA Downloading qca/crnv21.bin
[  174.958718] Bluetooth: btqca.c:qca_tlv_send_segment() hci0: QCA TLV response size mismatch

It seems the controller needs a short time after downloading the
firmware before it is ready for the NVM. A delay as short as 1 ms
seems sufficient, make it 10 ms just in case. No event is received
during the delay, hence we don't just silently drop an extra event.

Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: btusb: Fix error return code in btusb_mtk_setup_firmware()
Wei Yongjun [Wed, 10 Jul 2019 06:12:22 +0000 (06:12 +0000)]
Bluetooth: btusb: Fix error return code in btusb_mtk_setup_firmware()

Fix to return error code -EINVAL from the error handling
case instead of 0, as done elsewhere in this function.

Fixes: a1c49c434e15 ("Bluetooth: btusb: Add protocol support for MediaTek MT7668U USB devices")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agonet: tc35815: Explicitly check NET_IP_ALIGN is not zero in tc35815_rx
Nathan Chancellor [Mon, 12 Aug 2019 03:13:45 +0000 (20:13 -0700)]
net: tc35815: Explicitly check NET_IP_ALIGN is not zero in tc35815_rx

clang warns:

drivers/net/ethernet/toshiba/tc35815.c:1507:30: warning: use of logical
'&&' with constant operand [-Wconstant-logical-operand]
                        if (!HAVE_DMA_RXALIGN(lp) && NET_IP_ALIGN)
                                                  ^  ~~~~~~~~~~~~
drivers/net/ethernet/toshiba/tc35815.c:1507:30: note: use '&' for a
bitwise operation
                        if (!HAVE_DMA_RXALIGN(lp) && NET_IP_ALIGN)
                                                  ^~
                                                  &
drivers/net/ethernet/toshiba/tc35815.c:1507:30: note: remove constant to
silence this warning
                        if (!HAVE_DMA_RXALIGN(lp) && NET_IP_ALIGN)
                                                 ~^~~~~~~~~~~~~~~
1 warning generated.

Explicitly check that NET_IP_ALIGN is not zero, which matches how this
is checked in other parts of the tree. Because NET_IP_ALIGN is a build
time constant, this check will be constant folded away during
optimization.

Fixes: 82a9928db560 ("tc35815: Enable StripCRC feature")
Link: https://github.com/ClangBuiltLinux/linux/issues/608
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: initialise addr_trail_end when setting node addresses
Chris Packham [Sun, 11 Aug 2019 20:18:25 +0000 (08:18 +1200)]
tipc: initialise addr_trail_end when setting node addresses

We set the field 'addr_trial_end' to 'jiffies', instead of the current
value 0, at the moment the node address is initialized. This guarantees
we don't inadvertently enter an address trial period when the node
address is explicitly set by the user.

Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Check existence of .port_mdb_add callback before calling it
Chen-Yu Tsai [Sun, 11 Aug 2019 14:18:25 +0000 (22:18 +0800)]
net: dsa: Check existence of .port_mdb_add callback before calling it

The dsa framework has optional .port_mdb_{prepare,add,del} callback fields
for drivers to handle multicast database entries. When adding an entry, the
framework goes through a prepare phase, then a commit phase. Drivers not
providing these callbacks should be detected in the prepare phase.

DSA core may still bypass the bridge layer and call the dsa_port_mdb_add
function directly with no prepare phase or no switchdev trans object,
and the framework ends up calling an undefined .port_mdb_add callback.
This results in a NULL pointer dereference, as shown in the log below.

The other functions seem to be properly guarded. Do the same for
.port_mdb_add in dsa_switch_mdb_add_bitmap() as well.

    8<--- cut here ---
    Unable to handle kernel NULL pointer dereference at virtual address 00000000
    pgd = (ptrval)
    [00000000] *pgd=00000000
    Internal error: Oops: 80000005 [#1] SMP ARM
    Modules linked in: rtl8xxxu rtl8192cu rtl_usb rtl8192c_common rtlwifi mac80211 cfg80211
    CPU: 1 PID: 134 Comm: kworker/1:2 Not tainted 5.3.0-rc1-00247-gd3519030752a #1
    Hardware name: Allwinner sun7i (A20) Family
    Workqueue: events switchdev_deferred_process_work
    PC is at 0x0
    LR is at dsa_switch_event+0x570/0x620
    pc : [<00000000>]    lr : [<c08533ec>]    psr: 80070013
    sp : ee871db8  ip : 00000000  fp : ee98d0a4
    r10: 0000000c  r9 : 00000008  r8 : ee89f710
    r7 : ee98d040  r6 : ee98d088  r5 : c0f04c48  r4 : ee98d04c
    r3 : 00000000  r2 : ee89f710  r1 : 00000008  r0 : ee98d040
    Flags: Nzcv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment none
    Control: 10c5387d  Table: 6deb406a  DAC: 00000051
    Process kworker/1:2 (pid: 134, stack limit = 0x(ptrval))
    Stack: (0xee871db8 to 0xee872000)
    1da0:                                                       ee871e14 103ace2d
    1dc0: 00000000 ffffffff 00000000 ee871e14 00000005 00000000 c08524a0 00000000
    1de0: ffffe000 c014bdfc c0f04c48 ee871e98 c0f04c48 ee9e5000 c0851120 c014bef0
    1e00: 00000000 b643aea2 ee9b4068 c08509a8 ee2bf940 ee89f710 ee871ecb 00000000
    1e20: 00000008 103ace2d 00000000 c087e248 ee29c868 103ace2d 00000001 ffffffff
    1e40: 00000000 ee871e98 00000006 00000000 c0fb2a50 c087e2d0 ffffffff c08523c4
    1e60: ffffffff c014bdfc 00000006 c0fad2d0 ee871e98 ee89f710 00000000 c014c500
    1e80: 00000000 ee89f3c0 c0f04c48 00000000 ee9e5000 c087dfb4 ee9e5000 00000000
    1ea0: ee89f710 ee871ecb 00000001 103ace2d 00000000 c0f04c48 00000000 c087e0a8
    1ec0: 00000000 efd9a3e0 0089f3c0 103ace2d ee89f700 ee89f710 ee9e5000 00000122
    1ee0: 00000100 c087e130 ee89f700 c0fad2c8 c1003ef0 c087de4c 2e928000 c0fad2ec
    1f00: c0fad2ec ee839580 ef7a62c0 ef7a9400 00000000 c087def8 c0fad2ec c01447dc
    1f20: ef315640 ef7a62c0 00000008 ee839580 ee839594 ef7a62c0 00000008 c0f03d00
    1f40: ef7a62d8 ef7a62c0 ffffe000 c0145b84 ffffe000 c0fb2420 c0bfaa8c 00000000
    1f60: ffffe000 ee84b600 ee84b5c0 00000000 ee870000 ee839580 c0145b40 ef0e5ea4
    1f80: ee84b61c c014a6f8 00000001 ee84b5c0 c014a5b0 00000000 00000000 00000000
    1fa0: 00000000 00000000 00000000 c01010e8 00000000 00000000 00000000 00000000
    1fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
    1fe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000
    [<c08533ec>] (dsa_switch_event) from [<c014bdfc>] (notifier_call_chain+0x48/0x84)
    [<c014bdfc>] (notifier_call_chain) from [<c014bef0>] (raw_notifier_call_chain+0x18/0x20)
    [<c014bef0>] (raw_notifier_call_chain) from [<c08509a8>] (dsa_port_mdb_add+0x48/0x74)
    [<c08509a8>] (dsa_port_mdb_add) from [<c087e248>] (__switchdev_handle_port_obj_add+0x54/0xd4)
    [<c087e248>] (__switchdev_handle_port_obj_add) from [<c087e2d0>] (switchdev_handle_port_obj_add+0x8/0x14)
    [<c087e2d0>] (switchdev_handle_port_obj_add) from [<c08523c4>] (dsa_slave_switchdev_blocking_event+0x94/0xa4)
    [<c08523c4>] (dsa_slave_switchdev_blocking_event) from [<c014bdfc>] (notifier_call_chain+0x48/0x84)
    [<c014bdfc>] (notifier_call_chain) from [<c014c500>] (blocking_notifier_call_chain+0x50/0x68)
    [<c014c500>] (blocking_notifier_call_chain) from [<c087dfb4>] (switchdev_port_obj_notify+0x44/0xa8)
    [<c087dfb4>] (switchdev_port_obj_notify) from [<c087e0a8>] (switchdev_port_obj_add_now+0x90/0x104)
    [<c087e0a8>] (switchdev_port_obj_add_now) from [<c087e130>] (switchdev_port_obj_add_deferred+0x14/0x5c)
    [<c087e130>] (switchdev_port_obj_add_deferred) from [<c087de4c>] (switchdev_deferred_process+0x64/0x104)
    [<c087de4c>] (switchdev_deferred_process) from [<c087def8>] (switchdev_deferred_process_work+0xc/0x14)
    [<c087def8>] (switchdev_deferred_process_work) from [<c01447dc>] (process_one_work+0x218/0x50c)
    [<c01447dc>] (process_one_work) from [<c0145b84>] (worker_thread+0x44/0x5bc)
    [<c0145b84>] (worker_thread) from [<c014a6f8>] (kthread+0x148/0x150)
    [<c014a6f8>] (kthread) from [<c01010e8>] (ret_from_fork+0x14/0x2c)
    Exception stack(0xee871fb0 to 0xee871ff8)
    1fa0:                                     00000000 00000000 00000000 00000000
    1fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
    1fe0: 00000000 00000000 00000000 00000000 00000013 00000000
    Code: bad PC value
    ---[ end trace 1292c61abd17b130 ]---

    [<c08533ec>] (dsa_switch_event) from [<c014bdfc>] (notifier_call_chain+0x48/0x84)
    corresponds to

$ arm-linux-gnueabihf-addr2line -C -i -e vmlinux c08533ec

linux/net/dsa/switch.c:156
linux/net/dsa/switch.c:178
linux/net/dsa/switch.c:328

Fixes: e6db98db8a95 ("net: dsa: add switch mdb bitmap functions")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_ptp: Keep unmatched entries in a linked list
Petr Machata [Sun, 11 Aug 2019 07:48:37 +0000 (10:48 +0300)]
mlxsw: spectrum_ptp: Keep unmatched entries in a linked list

To identify timestamps for matching with their packets, Spectrum-1 uses a
five-tuple of (port, direction, domain number, message type, sequence ID).
If there are several clients from the same domain behind a single port
sending Delay_Req's, the only thing differentiating these packets, as far
as Spectrum-1 is concerned, is the sequence ID. Should sequence IDs between
individual clients be similar, conflicts may arise. That is not a problem
to hardware, which will simply deliver timestamps on a first comes, first
served basis.

However the driver uses a simple hash table to store the unmatched pieces.
When a new conflicting piece arrives, it pushes out the previously stored
one, which if it is a packet, is delivered without timestamp. Later on as
the corresponding timestamps arrive, the first one is mismatched to the
second packet, and the second one is never matched and eventually is GCd.

To correct this issue, instead of using a simple rhashtable, use rhltable
to keep the unmatched entries.

Previously, a found unmatched entry would always be removed from the hash
table. That is not the case anymore--an incompatible entry is left in the
hash table. Therefore removal from the hash table cannot be used to confirm
the validity of the looked-up pointer, instead the lookup would simply need
to be redone. Therefore move it inside the critical section. This
simplifies a lot of the code.

Fixes: 8748642751ed ("mlxsw: spectrum: PTP: Support SIOCGHWTSTAMP, SIOCSHWTSTAMP ioctls")
Reported-by: Alex Veber <alexve@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: nps_enet: Fix function names in doc comments
Jonathan Neuschäfer [Sat, 10 Aug 2019 11:11:56 +0000 (13:11 +0200)]
net: nps_enet: Fix function names in doc comments

Adjust the function names in two doc comments to match the corresponding
functions.

Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agorxrpc: Fix local refcounting
David Howells [Fri, 9 Aug 2019 21:47:47 +0000 (22:47 +0100)]
rxrpc: Fix local refcounting

Fix rxrpc_unuse_local() to handle a NULL local pointer as it can be called
on an unbound socket on which rx->local is not yet set.

The following reproduced (includes omitted):

int main(void)
{
socket(AF_RXRPC, SOCK_DGRAM, AF_INET);
return 0;
}

causes the following oops to occur:

BUG: kernel NULL pointer dereference, address: 0000000000000010
...
RIP: 0010:rxrpc_unuse_local+0x8/0x1b
...
Call Trace:
 rxrpc_release+0x2b5/0x338
 __sock_release+0x37/0xa1
 sock_close+0x14/0x17
 __fput+0x115/0x1e9
 task_work_run+0x72/0x98
 do_exit+0x51b/0xa7a
 ? __context_tracking_exit+0x4e/0x10e
 do_group_exit+0xab/0xab
 __x64_sys_exit_group+0x14/0x17
 do_syscall_64+0x89/0x1d4
 entry_SYSCALL_64_after_hwframe+0x49/0xbe

Reported-by: syzbot+20dee719a2e090427b5f@syzkaller.appspotmail.com
Fixes: 730c5fd42c1e ("rxrpc: Fix local endpoint refcounting")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeffrey Altman <jaltman@auristor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetdevsim: Restore per-network namespace accounting for fib entries
David Ahern [Tue, 6 Aug 2019 19:15:17 +0000 (12:15 -0700)]
netdevsim: Restore per-network namespace accounting for fib entries

Prior to the commit in the fixes tag, the resource controller in netdevsim
tracked fib entries and rules per network namespace. Restore that behavior.

Fixes: 5fc494225c1e ("netdevsim: create devlink instance per netdevsim instance")
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Sun, 11 Aug 2019 21:49:34 +0000 (14:49 -0700)]
Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2019-08-11

The following pull-request contains BPF updates for your *net* tree.

The main changes are:

1) x64 JIT code generation fix for backward-jumps to 1st insn, from Alexei.

2) Fix buggy multi-closing of BTF file descriptor in libbpf, from Andrii.

3) Fix libbpf_num_possible_cpus() to make it thread safe, from Takshak.

4) Fix bpftool to dump an error if pinning fails, from Jakub.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: swap sk_write_space on close
Jakub Kicinski [Sat, 10 Aug 2019 01:36:23 +0000 (18:36 -0700)]
net/tls: swap sk_write_space on close

Now that we swap the original proto and clear the ULP pointer
on close we have to make sure no callback will try to access
the freed state. sk_write_space is not part of sk_prot, remember
to swap it.

Reported-by: syzbot+dcdc9deefaec44785f32@syzkaller.appspotmail.com
Fixes: 95fa145479fb ("bpf: sockmap/tls, close can race with map free")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agohv_netvsc: Fix a warning of suspicious RCU usage
Dexuan Cui [Fri, 9 Aug 2019 01:58:08 +0000 (01:58 +0000)]
hv_netvsc: Fix a warning of suspicious RCU usage

This fixes a warning of "suspicious rcu_dereference_check() usage"
when nload runs.

Fixes: 776e726bfb34 ("netvsc: fix RCU warning in get_stats")
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mlx5-fixes-2019-08-08' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Fri, 9 Aug 2019 20:23:33 +0000 (13:23 -0700)]
Merge tag 'mlx5-fixes-2019-08-08' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
Mellanox, mlx5 fixes 2019-08-08

This series introduces some fixes to mlx5 driver.

Highlights:
1) From Tariq, Critical mlx5 kTLS fixes to better align with hw specs.
2) From Aya, Fixes to mlx5 tx devlink health reporter.
3) From Maxim, aRFs parsing to use flow dissector to avoid relying on
invalid skb fields.

Please pull and let me know if there is any problem.

For -stable v4.3
 ('net/mlx5e: Only support tx/rx pause setting for port owner')
For -stable v4.9
 ('net/mlx5e: Use flow keys dissector to parse packets for ARFS')
For -stable v5.1
 ('net/mlx5e: Fix false negative indication on tx reporter CQE recovery')
 ('net/mlx5e: Remove redundant check in CQE recovery flow of tx reporter')
 ('net/mlx5e: ethtool, Avoid setting speed to 56GBASE when autoneg off')

Note: when merged with net-next this minor conflict will pop up:
++<<<<<<< (net-next)
 +      if (is_eswitch_flow) {
 +              flow->esw_attr->match_level = match_level;
 +              flow->esw_attr->tunnel_match_level = tunnel_match_level;
++=======
+       if (flow->flags & MLX5E_TC_FLOW_ESWITCH) {
+               flow->esw_attr->inner_match_level = inner_match_level;
+               flow->esw_attr->outer_match_level = outer_match_level;
++>>>>>>> (net)

To resolve, use hunks from net (2nd) and replace:
if (flow->flags & MLX5E_TC_FLOW_ESWITCH)
with
if (is_eswitch_flow)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoixgbe: fix possible deadlock in ixgbe_service_task()
Taehee Yoo [Thu, 8 Aug 2019 16:37:56 +0000 (09:37 -0700)]
ixgbe: fix possible deadlock in ixgbe_service_task()

ixgbe_service_task() calls unregister_netdev() under rtnl_lock().
But unregister_netdev() internally calls rtnl_lock().
So deadlock would occur.

Fixes: 59dd45d550c5 ("ixgbe: firmware recovery mode")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'Fix-collisions-in-socket-cookie-generation'
David S. Miller [Fri, 9 Aug 2019 20:14:46 +0000 (13:14 -0700)]
Merge branch 'Fix-collisions-in-socket-cookie-generation'

Daniel Borkmann says:

====================
Fix collisions in socket cookie generation

This change makes the socket cookie generator as a global counter
instead of per netns in order to fix cookie collisions for BPF use
cases we ran into. See main patch #1 for more details.

Given the change is small/trivial and fixes an issue we're seeing
my preference would be net tree (though it cleanly applies to
net-next as well). Went for net tree instead of bpf tree here given
the main change is in net/core/sock_diag.c, but either way would be
fine with me.

v1 -> v2:
  - Fix up commit description in patch #1, thanks Eric!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobpf: sync bpf.h to tools infrastructure
Daniel Borkmann [Thu, 8 Aug 2019 11:57:26 +0000 (13:57 +0200)]
bpf: sync bpf.h to tools infrastructure

Pull in updates in BPF helper function description.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosock: make cookie generation global instead of per netns
Daniel Borkmann [Thu, 8 Aug 2019 11:57:25 +0000 (13:57 +0200)]
sock: make cookie generation global instead of per netns

Generating and retrieving socket cookies are a useful feature that is
exposed to BPF for various program types through bpf_get_socket_cookie()
helper.

The fact that the cookie counter is per netns is quite a limitation
for BPF in practice in particular for programs in host namespace that
use socket cookies as part of a map lookup key since they will be
causing socket cookie collisions e.g. when attached to BPF cgroup hooks
or cls_bpf on tc egress in host namespace handling container traffic
from veth or ipvlan devices with peer in different netns. Change the
counter to be global instead.

Socket cookie consumers must assume the value as opqaue in any case.
Not every socket must have a cookie generated and knowledge of the
counter value itself does not provide much value either way hence
conversion to global is fine.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Martynas Pumputis <m@lambda.lt>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'rxrpc-fixes-20190809' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Fri, 9 Aug 2019 18:27:17 +0000 (11:27 -0700)]
Merge tag 'rxrpc-fixes-20190809' of git://git./linux/kernel/git/dhowells/linux-fs

David Howells says:

====================
Here's a couple of fixes for rxrpc:

 (1) Fix refcounting of the local endpoint.

 (2) Don't calculate or report packet skew information.  This has been
     obsolete since AFS 3.1 and so is a waste of resources.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'bpf-bpftool-pinning-error-msg'
Daniel Borkmann [Fri, 9 Aug 2019 15:38:53 +0000 (17:38 +0200)]
Merge branch 'bpf-bpftool-pinning-error-msg'

Jakub Kicinski says:

====================
First make sure we don't use "prog" in error messages because
the pinning operation could be performed on a map. Second add
back missing error message if pin syscall failed.
====================

Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
5 years agotools: bpftool: add error message on pin failure
Jakub Kicinski [Wed, 7 Aug 2019 00:19:23 +0000 (17:19 -0700)]
tools: bpftool: add error message on pin failure

No error message is currently printed if the pin syscall
itself fails. It got lost in the loadall refactoring.

Fixes: 77380998d91d ("bpftool: add loadall command")
Reported-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
5 years agotools: bpftool: fix error message (prog -> object)
Jakub Kicinski [Wed, 7 Aug 2019 00:19:22 +0000 (17:19 -0700)]
tools: bpftool: fix error message (prog -> object)

Change an error message to work for any object being
pinned not just programs.

Fixes: 71bb428fe2c1 ("tools: bpf: add bpftool")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
5 years agorxrpc: Don't bother generating maxSkew in the ACK packet
David Howells [Fri, 9 Aug 2019 14:20:41 +0000 (15:20 +0100)]
rxrpc: Don't bother generating maxSkew in the ACK packet

Don't bother generating maxSkew in the ACK packet as it has been obsolete
since AFS 3.1.

Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Jeffrey Altman <jaltman@auristor.com>
5 years agorxrpc: Fix local endpoint refcounting
David Howells [Fri, 9 Aug 2019 14:20:41 +0000 (15:20 +0100)]
rxrpc: Fix local endpoint refcounting

The object lifetime management on the rxrpc_local struct is broken in that
the rxrpc_local_processor() function is expected to clean up and remove an
object - but it may get requeued by packets coming in on the backing UDP
socket once it starts running.

This may result in the assertion in rxrpc_local_rcu() firing because the
memory has been scheduled for RCU destruction whilst still queued:

rxrpc: Assertion failed
------------[ cut here ]------------
kernel BUG at net/rxrpc/local_object.c:468!

Note that if the processor comes around before the RCU free function, it
will just do nothing because ->dead is true.

Fix this by adding a separate refcount to count active users of the
endpoint that causes the endpoint to be destroyed when it reaches 0.

The original refcount can then be used to refcount objects through the work
processor and cause the memory to be rcu freed when that reaches 0.

Fixes: 4f95dd78a77e ("rxrpc: Rework local endpoint management")
Reported-by: syzbot+1e0edc4b8b7494c28450@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agonetfilter: nf_flow_table: teardown flow timeout race
Pablo Neira Ayuso [Fri, 9 Aug 2019 09:01:35 +0000 (11:01 +0200)]
netfilter: nf_flow_table: teardown flow timeout race

Flows that are in teardown state (due to RST / FIN TCP packet) still
have their offload flag set on. Hence, the conntrack garbage collector
may race to undo the timeout adjustment that the fixup routine performs,
leaving the conntrack entry in place with the internal offload timeout
(one day).

Update teardown flow state to ESTABLISHED and set tracking to liberal,
then once the offload bit is cleared, adjust timeout if it is more than
the default fixup timeout (conntrack might already have set a lower
timeout from the packet path).

Fixes: da5984e51063 ("netfilter: nf_flow_table: add support for sending flows back to the slow path")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
5 years agonetfilter: nf_flow_table: conntrack picks up expired flows
Pablo Neira Ayuso [Fri, 9 Aug 2019 09:01:33 +0000 (11:01 +0200)]
netfilter: nf_flow_table: conntrack picks up expired flows

Update conntrack entry to pick up expired flows, otherwise the conntrack
entry gets stuck with the internal offload timeout (one day). The TCP
state also needs to be adjusted to ESTABLISHED state and tracking is set
to liberal mode in order to give conntrack a chance to pick up the
expired flow.

Fixes: ac2a66665e23 ("netfilter: add generic flow table infrastructure")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
5 years agonetfilter: nf_tables: use-after-free in failing rule with bound set
Pablo Neira Ayuso [Fri, 9 Aug 2019 09:01:27 +0000 (11:01 +0200)]
netfilter: nf_tables: use-after-free in failing rule with bound set

If a rule that has already a bound anonymous set fails to be added, the
preparation phase releases the rule and the bound set. However, the
transaction object from the abort path still has a reference to the set
object that is stale, leading to a use-after-free when checking for the
set->bound field. Add a new field to the transaction that specifies if
the set is bound, so the abort path can skip releasing it since the rule
command owns it and it takes care of releasing it. After this update,
the set->bound field is removed.

[   24.649883] Unable to handle kernel paging request at virtual address 0000000000040434
[   24.657858] Mem abort info:
[   24.660686]   ESR = 0x96000004
[   24.663769]   Exception class = DABT (current EL), IL = 32 bits
[   24.669725]   SET = 0, FnV = 0
[   24.672804]   EA = 0, S1PTW = 0
[   24.675975] Data abort info:
[   24.678880]   ISV = 0, ISS = 0x00000004
[   24.682743]   CM = 0, WnR = 0
[   24.685723] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000428952000
[   24.692207] [0000000000040434] pgd=0000000000000000
[   24.697119] Internal error: Oops: 96000004 [#1] SMP
[...]
[   24.889414] Call trace:
[   24.891870]  __nf_tables_abort+0x3f0/0x7a0
[   24.895984]  nf_tables_abort+0x20/0x40
[   24.899750]  nfnetlink_rcv_batch+0x17c/0x588
[   24.904037]  nfnetlink_rcv+0x13c/0x190
[   24.907803]  netlink_unicast+0x18c/0x208
[   24.911742]  netlink_sendmsg+0x1b0/0x350
[   24.915682]  sock_sendmsg+0x4c/0x68
[   24.919185]  ___sys_sendmsg+0x288/0x2c8
[   24.923037]  __sys_sendmsg+0x7c/0xd0
[   24.926628]  __arm64_sys_sendmsg+0x2c/0x38
[   24.930744]  el0_svc_common.constprop.0+0x94/0x158
[   24.935556]  el0_svc_handler+0x34/0x90
[   24.939322]  el0_svc+0x8/0xc
[   24.942216] Code: 37280300 f9404023 91014262 aa1703e0 (f9401863)
[   24.948336] ---[ end trace cebbb9dcbed3b56f ]---

Fixes: f6ac85858976 ("netfilter: nf_tables: unbind set in rule from commit path")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
5 years agonet: tundra: tsi108: use spin_lock_irqsave instead of spin_lock_irq in IRQ context
Fuqian Huang [Fri, 9 Aug 2019 05:35:39 +0000 (13:35 +0800)]
net: tundra: tsi108: use spin_lock_irqsave instead of spin_lock_irq in IRQ context

As spin_unlock_irq will enable interrupts.
Function tsi108_stat_carry is called from interrupt handler tsi108_irq.
Interrupts are enabled in interrupt handler.
Use spin_lock_irqsave/spin_unlock_irqrestore instead of spin_(un)lock_irq
in IRQ context to avoid this.

Signed-off-by: Fuqian Huang <huangfq.daxian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoteam: Add vlan tx offload to hw_enc_features
YueHaibing [Thu, 8 Aug 2019 06:22:47 +0000 (14:22 +0800)]
team: Add vlan tx offload to hw_enc_features

We should also enable team's vlan tx offload in hw_enc_features,
pass the vlan packets to the slave devices with vlan tci, let the
slave handle vlan tunneling offload implementation.

Fixes: 3268e5cb494d ("team: Advertise tunneling offload features")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: prevent skb_orphan() from leaking TLS plain text with offload
Jakub Kicinski [Thu, 8 Aug 2019 00:03:59 +0000 (17:03 -0700)]
net/tls: prevent skb_orphan() from leaking TLS plain text with offload

sk_validate_xmit_skb() and drivers depend on the sk member of
struct sk_buff to identify segments requiring encryption.
Any operation which removes or does not preserve the original TLS
socket such as skb_orphan() or skb_clone() will cause clear text
leaks.

Make the TCP socket underlying an offloaded TLS connection
mark all skbs as decrypted, if TLS TX is in offload mode.
Then in sk_validate_xmit_skb() catch skbs which have no socket
(or a socket with no validation) and decrypted flag set.

Note that CONFIG_SOCK_VALIDATE_XMIT, CONFIG_TLS_DEVICE and
sk->sk_validate_xmit_skb are slightly interchangeable right now,
they all imply TLS offload. The new checks are guarded by
CONFIG_TLS_DEVICE because that's the option guarding the
sk_buff->decrypted member.

Second, smaller issue with orphaning is that it breaks
the guarantee that packets will be delivered to device
queues in-order. All TLS offload drivers depend on that
scheduling property. This means skb_orphan_partial()'s
trick of preserving partial socket references will cause
issues in the drivers. We need a full orphan, and as a
result netem delay/throttling will cause all TLS offload
skbs to be dropped.

Reusing the sk_buff->decrypted flag also protects from
leaking clear text when incoming, decrypted skb is redirected
(e.g. by TC).

See commit 0608c69c9a80 ("bpf: sk_msg, sock{map|hash} redirect
through ULP") for justification why the internal flag is safe.
The only location which could leak the flag in is tcp_bpf_sendmsg(),
which is taken care of by clearing the previously unused bit.

v2:
 - remove superfluous decrypted mark copy (Willem);
 - remove the stale doc entry (Boris);
 - rely entirely on EOR marking to prevent coalescing (Boris);
 - use an internal sendpages flag instead of marking the socket
   (Boris).
v3 (Willem):
 - reorganize the can_skb_orphan_partial() condition;
 - fix the flag leak-in through tcp_bpf_sendmsg.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'skbedit-batch-fixes'
David S. Miller [Fri, 9 Aug 2019 05:37:06 +0000 (22:37 -0700)]
Merge branch 'skbedit-batch-fixes'

Roman Mashak says:

====================
Fix batched event generation for skbedit action

When adding or deleting a batch of entries, the kernel sends up to
TCA_ACT_MAX_PRIO (defined to 32 in kernel) entries in an event to user
space. However it does not consider that the action sizes may vary and
require different skb sizes.

For example, consider the following script adding 32 entries with all
supported skbedit parameters and cookie (in order to maximize netlink
messages size):

% cat tc-batch.sh
TC="sudo /mnt/iproute2.git/tc/tc"

$TC actions flush action skbedit
for i in `seq 1 $1`;
do
   cmd="action skbedit queue_mapping 2 priority 10 mark 7/0xaabbccdd \
               ptype host inheritdsfield \
               index $i cookie aabbccddeeff112233445566778800a1 "
   args=$args$cmd
done
$TC actions add $args
%
% ./tc-batch.sh 32
Error: Failed to fill netlink attributes while adding TC action.
We have an error talking to the kernel
%

patch 1 adds callback in tc_action_ops of skbedit action, which calculates
the action size, and passes size to tcf_add_notify()/tcf_del_notify().

patch 2 updates the TDC test suite with relevant skbedit test cases.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotc-testing: updated skbedit action tests with batch create/delete
Roman Mashak [Wed, 7 Aug 2019 19:57:29 +0000 (15:57 -0400)]
tc-testing: updated skbedit action tests with batch create/delete

Update TDC tests with cases varifying ability of TC to install or delete
batches of skbedit actions.

Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet sched: update skbedit action for batched events operations
Roman Mashak [Wed, 7 Aug 2019 19:57:28 +0000 (15:57 -0400)]
net sched: update skbedit action for batched events operations

Add get_fill_size() routine used to calculate the action size
when building a batch of events.

Fixes: ca9b0e27e ("pkt_action: add new action skbedit")
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: sja1105: remove set but not used variables 'tx_vid' and 'rx_vid'
YueHaibing [Wed, 7 Aug 2019 13:08:56 +0000 (21:08 +0800)]
net: dsa: sja1105: remove set but not used variables 'tx_vid' and 'rx_vid'

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/dsa/sja1105/sja1105_main.c: In function sja1105_fdb_dump:
drivers/net/dsa/sja1105/sja1105_main.c:1226:14: warning:
 variable tx_vid set but not used [-Wunused-but-set-variable]
drivers/net/dsa/sja1105/sja1105_main.c:1226:6: warning:
 variable rx_vid set but not used [-Wunused-but-set-variable]

They are not used since commit 6d7c7d948a2e ("net: dsa:
sja1105: Fix broken learning with vlan_filtering disabled")

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobonding: Add vlan tx offload to hw_enc_features
YueHaibing [Wed, 7 Aug 2019 02:19:59 +0000 (10:19 +0800)]
bonding: Add vlan tx offload to hw_enc_features

As commit 30d8177e8ac7 ("bonding: Always enable vlan tx offload")
said, we should always enable bonding's vlan tx offload, pass the
vlan packets to the slave devices with vlan tci, let them to handle
vlan implementation.

Now if encapsulation protocols like VXLAN is used, skb->encapsulation
may be set, then the packet is passed to vlan device which based on
bonding device. However in netif_skb_features(), the check of
hw_enc_features:

 if (skb->encapsulation)
                 features &= dev->hw_enc_features;

clears NETIF_F_HW_VLAN_CTAG_TX/NETIF_F_HW_VLAN_STAG_TX. This results
in same issue in commit 30d8177e8ac7 like this:

vlan_dev_hard_start_xmit
  -->dev_queue_xmit
    -->validate_xmit_skb
      -->netif_skb_features //NETIF_F_HW_VLAN_CTAG_TX is cleared
      -->validate_xmit_vlan
        -->__vlan_hwaccel_push_inside //skb->tci is cleared
...
 --> bond_start_xmit
   --> bond_xmit_hash //BOND_XMIT_POLICY_ENCAP34
     --> __skb_flow_dissect // nhoff point to IP header
        -->  case htons(ETH_P_8021Q)
             // skb_vlan_tag_present is false, so
             vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
             //vlan point to ip header wrongly

Fixes: b2a103e6d0af ("bonding: convert to ndo_fix_features")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: sch_taprio: fix memleak in error path for sched list parse
Ivan Khoronzhuk [Tue, 6 Aug 2019 22:45:40 +0000 (01:45 +0300)]
net: sched: sch_taprio: fix memleak in error path for sched list parse

In error case, all entries should be freed from the sched list
before deleting it. For simplicity use rcu way.

Fixes: 5a781ccbd19e46 ("tc: Add support for configuring the taprio scheduler")
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: docs: replace IPX in tuntap documentation
Stephen Hemminger [Mon, 5 Aug 2019 22:30:03 +0000 (15:30 -0700)]
net: docs: replace IPX in tuntap documentation

IPX is no longer supported, but the example in the documentation
might useful. Replace it with IPv6.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodocs: admin-guide: remove references to IPX and token-ring
Stephen Hemminger [Mon, 5 Aug 2019 22:30:02 +0000 (15:30 -0700)]
docs: admin-guide: remove references to IPX and token-ring

Both IPX and TR have not been supported for a while now.
Remove them from the /proc/sys/net documentation.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoxen/netback: Reset nr_frags before freeing skb
Ross Lagerwall [Mon, 5 Aug 2019 15:34:34 +0000 (16:34 +0100)]
xen/netback: Reset nr_frags before freeing skb

At this point nr_frags has been incremented but the frag does not yet
have a page assigned so freeing the skb results in a crash. Reset
nr_frags before freeing the skb to prevent this.

Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoinet: frags: re-introduce skb coalescing for local delivery
Guillaume Nault [Fri, 2 Aug 2019 15:15:03 +0000 (17:15 +0200)]
inet: frags: re-introduce skb coalescing for local delivery

Before commit d4289fcc9b16 ("net: IP6 defrag: use rbtrees for IPv6
defrag"), a netperf UDP_STREAM test[0] using big IPv6 datagrams (thus
generating many fragments) and running over an IPsec tunnel, reported
more than 6Gbps throughput. After that patch, the same test gets only
9Mbps when receiving on a be2net nic (driver can make a big difference
here, for example, ixgbe doesn't seem to be affected).

By reusing the IPv4 defragmentation code, IPv6 lost fragment coalescing
(IPv4 fragment coalescing was dropped by commit 14fe22e33462 ("Revert
"ipv4: use skb coalescing in defragmentation"")).

Without fragment coalescing, be2net runs out of Rx ring entries and
starts to drop frames (ethtool reports rx_drops_no_frags errors). Since
the netperf traffic is only composed of UDP fragments, any lost packet
prevents reassembly of the full datagram. Therefore, fragments which
have no possibility to ever get reassembled pile up in the reassembly
queue, until the memory accounting exeeds the threshold. At that point
no fragment is accepted anymore, which effectively discards all
netperf traffic.

When reassembly timeout expires, some stale fragments are removed from
the reassembly queue, so a few packets can be received, reassembled
and delivered to the netperf receiver. But the nic still drops frames
and soon the reassembly queue gets filled again with stale fragments.
These long time frames where no datagram can be received explain why
the performance drop is so significant.

Re-introducing fragment coalescing is enough to get the initial
performances again (6.6Gbps with be2net): driver doesn't drop frames
anymore (no more rx_drops_no_frags errors) and the reassembly engine
works at full speed.

This patch is quite conservative and only coalesces skbs for local
IPv4 and IPv6 delivery (in order to avoid changing skb geometry when
forwarding). Coalescing could be extended in the future if need be, as
more scenarios would probably benefit from it.

[0]: Test configuration
Sender:
ip xfrm policy flush
ip xfrm state flush
ip xfrm state add src fc00:1::1 dst fc00:2::1 proto esp spi 0x1000 aead 'rfc4106(gcm(aes))' 0x0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b 96 mode transport sel src fc00:1::1 dst fc00:2::1
ip xfrm policy add src fc00:1::1 dst fc00:2::1 dir in tmpl src fc00:1::1 dst fc00:2::1 proto esp mode transport action allow
ip xfrm state add src fc00:2::1 dst fc00:1::1 proto esp spi 0x1001 aead 'rfc4106(gcm(aes))' 0x0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b 96 mode transport sel src fc00:2::1 dst fc00:1::1
ip xfrm policy add src fc00:2::1 dst fc00:1::1 dir out tmpl src fc00:2::1 dst fc00:1::1 proto esp mode transport action allow
netserver -D -L fc00:2::1

Receiver:
ip xfrm policy flush
ip xfrm state flush
ip xfrm state add src fc00:2::1 dst fc00:1::1 proto esp spi 0x1001 aead 'rfc4106(gcm(aes))' 0x0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b 96 mode transport sel src fc00:2::1 dst fc00:1::1
ip xfrm policy add src fc00:2::1 dst fc00:1::1 dir in tmpl src fc00:2::1 dst fc00:1::1 proto esp mode transport action allow
ip xfrm state add src fc00:1::1 dst fc00:2::1 proto esp spi 0x1000 aead 'rfc4106(gcm(aes))' 0x0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b 96 mode transport sel src fc00:1::1 dst fc00:2::1
ip xfrm policy add src fc00:1::1 dst fc00:2::1 dir out tmpl src fc00:1::1 dst fc00:2::1 proto esp mode transport action allow
netperf -H fc00:2::1 -f k -P 0 -L fc00:1::1 -l 60 -t UDP_STREAM -I 99,5 -i 5,5 -T5,5 -6

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx5e: Remove redundant check in CQE recovery flow of tx reporter
Aya Levin [Thu, 8 Aug 2019 12:55:48 +0000 (15:55 +0300)]
net/mlx5e: Remove redundant check in CQE recovery flow of tx reporter

Remove check of recovery bit, in the beginning of the CQE recovery
function. This test is already performed right before the reporter
is invoked, when CQE error is detected.

Fixes: de8650a82071 ("net/mlx5e: Add tx reporter support")
Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Fix error flow of CQE recovery on tx reporter
Aya Levin [Tue, 6 Aug 2019 12:19:19 +0000 (15:19 +0300)]
net/mlx5e: Fix error flow of CQE recovery on tx reporter

CQE recovery function begins with test and set of recovery bit. Add an
error flow which ensures clearing of this bit when leaving the recovery
function, to allow further recoveries to take place. This allows removal
of clearing recovery bit on sq activate.

Fixes: de8650a82071 ("net/mlx5e: Add tx reporter support")
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Fix false negative indication on tx reporter CQE recovery
Aya Levin [Wed, 7 Aug 2019 12:59:06 +0000 (15:59 +0300)]
net/mlx5e: Fix false negative indication on tx reporter CQE recovery

Remove wrong error return value when SQ is not in error state.
CQE recovery on TX reporter queries the sq state. If the sq is not in
error state, the sq is either in ready or reset state. Ready state is
good state which doesn't require recovery and reset state is a temporal
state which ends in ready state. With this patch, CQE recovery in this
scenario is successful.

Fixes: de8650a82071 ("net/mlx5e: Add tx reporter support")
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: kTLS, Fix tisn field placement
Tariq Toukan [Tue, 30 Jul 2019 10:45:42 +0000 (13:45 +0300)]
net/mlx5e: kTLS, Fix tisn field placement

Shift the tisn field in the WQE control segment, per the
HW specification.

Fixes: d2ead1f360e8 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: kTLS, Fix tisn field name
Tariq Toukan [Thu, 8 Aug 2019 09:26:57 +0000 (12:26 +0300)]
net/mlx5e: kTLS, Fix tisn field name

Use the proper tisn field name from the union in struct mlx5_wqe_ctrl_seg.

Fixes: d2ead1f360e8 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: kTLS, Fix progress params context WQE layout
Tariq Toukan [Tue, 30 Jul 2019 08:55:25 +0000 (11:55 +0300)]
net/mlx5e: kTLS, Fix progress params context WQE layout

The TLS progress params context WQE should not include an
Eth segment, drop it.
In addition, align the tls_progress_params layout with the
HW specification document:
- fix the tisn field name.
- remove the valid bit.

Fixes: a12ff35e0fb7 ("net/mlx5: Introduce TLS TX offload hardware bits and structures")
Fixes: d2ead1f360e8 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: kTLS, Fix wrong TIS opmod constants
Tariq Toukan [Sun, 21 Jul 2019 11:18:42 +0000 (14:18 +0300)]
net/mlx5: kTLS, Fix wrong TIS opmod constants

Fix the used constants for TLS TIS opmods, per the HW specification.

Fixes: a12ff35e0fb7 ("net/mlx5: Introduce TLS TX offload hardware bits and structures")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: crypto, Fix wrong offset in encryption key command
Tariq Toukan [Sun, 21 Jul 2019 11:13:15 +0000 (14:13 +0300)]
net/mlx5: crypto, Fix wrong offset in encryption key command

Fix the 128b key offset in key encryption key creation command,
per the HW specification.

Fixes: 45d3b55dc665 ("net/mlx5: Add crypto library to support create/destroy encryption key")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: ethtool, Avoid setting speed to 56GBASE when autoneg off
Mohamad Heib [Tue, 23 Apr 2019 18:13:48 +0000 (21:13 +0300)]
net/mlx5e: ethtool, Avoid setting speed to 56GBASE when autoneg off

Setting speed to 56GBASE is allowed only with auto-negotiation enabled.

This patch prevent setting speed to 56GBASE when auto-negotiation disabled.

Fixes: f62b8bb8f2d3 ("net/mlx5: Extend mlx5_core to support ConnectX-4 Ethernet functionality")
Signed-off-by: Mohamad Heib <mohamadh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Only support tx/rx pause setting for port owner
Huy Nguyen [Thu, 1 Aug 2019 16:10:19 +0000 (11:10 -0500)]
net/mlx5e: Only support tx/rx pause setting for port owner

Only support changing tx/rx pause frame setting if the net device
is the vport group manager.

Fixes: 3c2d18ef22df ("net/mlx5e: Support ethtool get/set_pauseparam")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: Support inner header match criteria for non decap flow action
Huy Nguyen [Wed, 17 Jul 2019 19:04:31 +0000 (14:04 -0500)]
net/mlx5: Support inner header match criteria for non decap flow action

We have an issue that OVS application creates an offloaded drop rule
that drops VXLAN traffic with both inner and outer header match
criteria. mlx5_core driver detects correctly the inner and outer
header match criteria but does not enable the inner header match criteria
due to an incorrect assumption in mlx5_eswitch_add_offloaded_rule that
only decap rule needs inner header criteria.

Solution:
Remove mlx5_esw_flow_attr's match_level and tunnel_match_level and add
two new members: inner_match_level and outer_match_level.
inner/outer_match_level is set to NONE if the inner/outer match criteria
is not specified in the tc rule creation request. The decap assumption is
removed and the code just needs to check for inner/outer_match_level to
enable the corresponding bit in firmware's match_criteria_enable value.

Fixes: 6363651d6dd7 ("net/mlx5e: Properly set steering match levels for offloaded TC decap rules")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Use flow keys dissector to parse packets for ARFS
Maxim Mikityanskiy [Fri, 5 Jul 2019 14:59:28 +0000 (17:59 +0300)]
net/mlx5e: Use flow keys dissector to parse packets for ARFS

The current ARFS code relies on certain fields to be set in the SKB
(e.g. transport_header) and extracts IP addresses and ports by custom
code that parses the packet. The necessary SKB fields, however, are not
always set at that point, which leads to an out-of-bounds access. Use
skb_flow_dissect_flow_keys() to get the necessary information reliably,
fix the out-of-bounds access and reuse the code.

Fixes: 18c908e477dc ("net/mlx5e: Add accelerated RFS support")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agoMerge tag 'batadv-net-for-davem-20190808' of git://git.open-mesh.org/linux-merge
David S. Miller [Thu, 8 Aug 2019 18:25:39 +0000 (11:25 -0700)]
Merge tag 'batadv-net-for-davem-20190808' of git://git.open-mesh.org/linux-merge

Simon Wunderlich says:

====================
Here are some batman-adv bugfixes:

 - Fix netlink dumping of all mcast_flags buckets, by Sven Eckelmann

 - Fix deletion of RTR(4|6) mcast list entries, by Sven Eckelmann
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Wed, 7 Aug 2019 00:11:59 +0000 (17:11 -0700)]
Merge git://git./linux/kernel/git/netdev/net

Pull networking fixes from David Miller:
 "Yeah I should have sent a pull request last week, so there is a lot
  more here than usual:

   1) Fix memory leak in ebtables compat code, from Wenwen Wang.

   2) Several kTLS bug fixes from Jakub Kicinski (circular close on
      disconnect etc.)

   3) Force slave speed check on link state recovery in bonding 802.3ad
      mode, from Thomas Falcon.

   4) Clear RX descriptor bits before assigning buffers to them in
      stmmac, from Jose Abreu.

   5) Several missing of_node_put() calls, mostly wrt. for_each_*() OF
      loops, from Nishka Dasgupta.

   6) Double kfree_skb() in peak_usb can driver, from Stephane Grosjean.

   7) Need to hold sock across skb->destructor invocation, from Cong
      Wang.

   8) IP header length needs to be validated in ipip tunnel xmit, from
      Haishuang Yan.

   9) Use after free in ip6 tunnel driver, also from Haishuang Yan.

  10) Do not use MSI interrupts on r8169 chips before RTL8168d, from
      Heiner Kallweit.

  11) Upon bridge device init failure, we need to delete the local fdb.
      From Nikolay Aleksandrov.

  12) Handle erros from of_get_mac_address() properly in stmmac, from
      Martin Blumenstingl.

  13) Handle concurrent rename vs. dump in netfilter ipset, from Jozsef
      Kadlecsik.

  14) Setting NETIF_F_LLTX on mac80211 causes complete breakage with
      some devices, so revert. From Johannes Berg.

  15) Fix deadlock in rxrpc, from David Howells.

  16) Fix Kconfig deps of enetc driver, we must have PHYLIB. From Yue
      Haibing.

  17) Fix mvpp2 crash on module removal, from Matteo Croce.

  18) Fix race in genphy_update_link, from Heiner Kallweit.

  19) bpf_xdp_adjust_head() stopped working with generic XDP when we
      fixes generic XDP to support stacked devices properly, fix from
      Jesper Dangaard Brouer.

  20) Unbalanced RCU locking in rt6_update_exception_stamp_rt(), from
      David Ahern.

  21) Several memory leaks in new sja1105 driver, from Vladimir Oltean"

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (214 commits)
  net: dsa: sja1105: Fix memory leak on meta state machine error path
  net: dsa: sja1105: Fix memory leak on meta state machine normal path
  net: dsa: sja1105: Really fix panic on unregistering PTP clock
  net: dsa: sja1105: Use the LOCKEDS bit for SJA1105 E/T as well
  net: dsa: sja1105: Fix broken learning with vlan_filtering disabled
  net: dsa: qca8k: Add of_node_put() in qca8k_setup_mdio_bus()
  net: sched: sample: allow accessing psample_group with rtnl
  net: sched: police: allow accessing police->params with rtnl
  net: hisilicon: Fix dma_map_single failed on arm64
  net: hisilicon: fix hip04-xmit never return TX_BUSY
  net: hisilicon: make hip04_tx_reclaim non-reentrant
  tc-testing: updated vlan action tests with batch create/delete
  net sched: update vlan action for batched events operations
  net: stmmac: tc: Do not return a fragment entry
  net: stmmac: Fix issues when number of Queues >= 4
  net: stmmac: xgmac: Fix XGMAC selftests
  be2net: disable bh with spin_lock in be_process_mcc
  net: cxgb3_main: Fix a resource leak in a error path in 'init_one()'
  net: ethernet: sun4i-emac: Support phy-handle property for finding PHYs
  net: bridge: move default pvid init/deinit to NETDEV_REGISTER/UNREGISTER
  ...

5 years agoMerge branch 'sja1105-fixes'
David S. Miller [Tue, 6 Aug 2019 21:37:02 +0000 (14:37 -0700)]
Merge branch 'sja1105-fixes'

Vladimir Oltean says:

====================
Fixes for SJA1105 DSA: FDBs, Learning and PTP

This is an assortment of functional fixes for the sja1105 switch driver
targeted for the "net" tree (although they apply on net-next just as
well).

Patch 1/5 ("net: dsa: sja1105: Fix broken learning with vlan_filtering
disabled") repairs a breakage introduced in the early development stages
of the driver: support for traffic from the CPU has broken "normal"
frame forwarding (based on DMAC) - there is connectivity through the
switch only because all frames are flooded.
I debated whether this patch qualifies as a fix, since it puts the
switch into a mode it has never operated in before (aka SVL). But
"normal" forwarding did use to work before the "Traffic support for
SJA1105 DSA driver" patchset, and arguably this patch should have been
part of that.
Also, it would be strange for this feature to be broken in the 5.2 LTS.

Patch 2/5 ("net: dsa: sja1105: Use the LOCKEDS bit for SJA1105 E/T as
well") is a simplification of a previous FDB-related patch that is
currently in the 5.3 rc's.

Patches 3/5 - 5/5 fix various crashes found while running linuxptp over the
switch ports for extended periods of time, or in conjunction with other
error conditions. The fixed-up commits were all introduced in 5.2.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>