Skip to content

K
Kernel

openeuler / Kernel
接近 2 年 前同步成功

通知 8
Star 0
Fork 0
K
Kernel
切换分支/标签
  1. 27 5月, 2020 5 次提交
  3. 26 5月, 2020 2 次提交
    • Q
      qlcnic: fix missing release in qlcnic_83xx_interrupt_test. · 15c97385
      由 Qiushi Wu 提交于 
      In function qlcnic_83xx_interrupt_test(), function
qlcnic_83xx_diag_alloc_res() is not handled by function
qlcnic_83xx_diag_free_res() after a call of the function
qlcnic_alloc_mbx_args() failed. Fix this issue by adding
a jump target "fail_mbx_args", and jump to this new target
when qlcnic_alloc_mbx_args() failed.

Fixes: b6b4316c ("qlcnic: Handle qlcnic_alloc_mbx_args() failure")
Signed-off-by: NQiushi Wu <wu000273@umn.edu>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      15c97385
    • V
      dpaa_eth: fix usage as DSA master, try 3 · 5d14c304
      由 Vladimir Oltean 提交于 
      The dpaa-eth driver probes on compatible string for the MAC node, and
the fman/mac.c driver allocates a dpaa-ethernet platform device that
triggers the probing of the dpaa-eth net device driver.

All of this is fine, but the problem is that the struct device of the
dpaa_eth net_device is 2 parents away from the MAC which can be
referenced via of_node. So of_find_net_device_by_node can't find it, and
DSA switches won't be able to probe on top of FMan ports.

It would be a bit silly to modify a core function
(of_find_net_device_by_node) to look for dev->parent->parent->of_node
just for one driver. We're just 1 step away from implementing full
recursion.

Actually there have already been at least 2 previous attempts to make
this work:
- Commit a1a50c8e ("fsl/man: Inherit parent device and of_node")
- One or more of the patches in "[v3,0/6] adapt DPAA drivers for DSA":
  https://patchwork.ozlabs.org/project/netdev/cover/1508178970-28945-1-git-send-email-madalin.bucur@nxp.com/
  (I couldn't really figure out which one was supposed to solve the
  problem and how).

Point being, it looks like this is still pretty much a problem today.
On T1040, the /sys/class/net/eth0 symlink currently points to

../../devices/platform/ffe000000.soc/ffe400000.fman/ffe4e6000.ethernet/dpaa-ethernet.0/net/eth0

which pretty much illustrates the problem. The closest of_node we've got
is the "fsl,fman-memac" at /soc@ffe000000/fman@400000/ethernet@e6000,
which is what we'd like to be able to reference from DSA as host port.

For of_find_net_device_by_node to find the eth0 port, we would need the
parent of the eth0 net_device to not be the "dpaa-ethernet" platform
device, but to point 1 level higher, aka the "fsl,fman-memac" node
directly. The new sysfs path would look like this:

../../devices/platform/ffe000000.soc/ffe400000.fman/ffe4e6000.ethernet/net/eth0

And this is exactly what SET_NETDEV_DEV does. It sets the parent of the
net_device. The new parent has an of_node associated with it, and
of_dev_node_match already checks for the of_node of the device or of its
parent.

Fixes: a1a50c8e ("fsl/man: Inherit parent device and of_node")
Fixes: c6e26ea8 ("dpaa_eth: change device used")
Signed-off-by: NVladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      5d14c304
  5. 24 5月, 2020 6 次提交
  7. 23 5月, 2020 22 次提交
    • S
      net/mlx5: Fix error flow in case of function_setup failure · 4f7400d5
      由 Shay Drory 提交于 
      Currently, if an error occurred during mlx5_function_setup(), we
keep dev->state as DEVICE_STATE_UP.
Fixing it by adding a goto label.

Fixes: e161105e ("net/mlx5: Function setup/teardown procedures")
Signed-off-by: NShay Drory <shayd@mellanox.com>
Reviewed-by: NMoshe Shemesh <moshe@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      4f7400d5
    • R
      net/mlx5e: CT: Correctly get flow rule · d37bd5e8
      由 Roi Dayan 提交于 
      The correct way is to us the flow_cls_offload_flow_rule() wrapper
instead of f->rule directly.

Fixes: 4c3844d9 ("net/mlx5e: CT: Introduce connection tracking")
Signed-off-by: NRoi Dayan <roid@mellanox.com>
Reviewed-by: NOz Shlomo <ozsh@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      d37bd5e8
    • M
      net/mlx5e: Update netdev txq on completions during closure · 5e911e2c
      由 Moshe Shemesh 提交于 
      On sq closure when we free its descriptors, we should also update netdev
txq on completions which would not arrive. Otherwise if we reopen sqs
and attach them back, for example on fw fatal recovery flow, we may get
tx timeout.

Fixes: 29429f33 ("net/mlx5e: Timeout if SQ doesn't flush during close")
Signed-off-by: NMoshe Shemesh <moshe@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      5e911e2c
    • R
      net/mlx5: Annotate mutex destroy for root ns · 9ca41539
      由 Roi Dayan 提交于 
      Invoke mutex_destroy() to catch any errors.

Fixes: 2cc43b49 ("net/mlx5_core: Managing root flow table")
Signed-off-by: NRoi Dayan <roid@mellanox.com>
Reviewed-by: NMark Bloch <markb@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      9ca41539
    • R
      net/mlx5: Don't maintain a case of del_sw_func being null · 6eb7a268
      由 Roi Dayan 提交于 
      Add del_sw_func cb for root ns. Now there is no need to
maintain a case of del_sw_func being null when freeing the node.

Fixes: 2cc43b49 ("net/mlx5_core: Managing root flow table")
Signed-off-by: NRoi Dayan <roid@mellanox.com>
Reviewed-by: NMark Bloch <markb@mellanox.com>
Reviewed-by: NPaul Blakey <paulb@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      6eb7a268
    • R
      net/mlx5: Fix cleaning unmanaged flow tables · aee37f3d
      由 Roi Dayan 提交于 
      Unmanaged flow tables doesn't have a parent and tree_put_node()
assume there is always a parent if cleaning is needed. fix that.

Fixes: 5281a0c9 ("net/mlx5: fs_core: Introduce unmanaged flow tables")
Signed-off-by: NRoi Dayan <roid@mellanox.com>
Reviewed-by: NMark Bloch <markb@mellanox.com>
Reviewed-by: NPaul Blakey <paulb@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      aee37f3d
    • M
      net/mlx5: Fix memory leak in mlx5_events_init · df14ad1e
      由 Moshe Shemesh 提交于 
      Fix memory leak in mlx5_events_init(), in case
create_single_thread_workqueue() fails, events
struct should be freed.

Fixes: 5d3c537f ("net/mlx5: Handle event of power detection in the PCIE slot")
Signed-off-by: NMoshe Shemesh <moshe@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      df14ad1e
    • R
      net/mlx5e: Fix inner tirs handling · a16b8e0d
      由 Roi Dayan 提交于 
      In the cited commit inner_tirs argument was added to create and destroy
inner tirs, and no indication was added to mlx5e_modify_tirs_hash()
function. In order to have a consistent handling, use
inner_indir_tir[0].tirn in tirs destroy/modify function as an indication
to whether inner tirs are created.
Inner tirs are not created for representors and before this commit,
a call to mlx5e_modify_tirs_hash() was sending HW commands to
modify non-existent inner tirs.

Fixes: 46dc933c ("net/mlx5e: Provide explicit directive if to create inner indirect tirs")
Signed-off-by: NRoi Dayan <roid@mellanox.com>
Reviewed-by: NVlad Buslov <vladbu@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      a16b8e0d
    • T
      net/mlx5e: kTLS, Destroy key object after destroying the TIS · 16736e11
      由 Tariq Toukan 提交于 
      The TLS TIS object contains the dek/key ID.
By destroying the key first, the TIS would contain an invalid
non-existing key ID.
Reverse the destroy order, this also acheives the desired assymetry
between the destroy and the create flows.

Fixes: d2ead1f3 ("net/mlx5e: Add kTLS TX HW offload support")
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Reviewed-by: NBoris Pismenny <borisp@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      16736e11
    • M
      net/mlx5e: Fix allowed tc redirect merged eswitch offload cases · 32134847
      由 Maor Dickman 提交于 
      After changing the parent_id to be the same for both NICs of same
The cited commit wrongly allow offload of tc redirect flows from
VF to uplink and vice versa when devcies are on different eswitch,
these cases aren't supported by HW.

Disallow the above offloads when devcies are on different eswitch
and VF LAG is not configured.

Fixes: f6dc1264 ("net/mlx5e: Disallow tc redirect offload cases we don't support")
Signed-off-by: NMaor Dickman <maord@mellanox.com>
Reviewed-by: NRoi Dayan <roid@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      32134847
    • E
      net/mlx5: Avoid processing commands before cmdif is ready · f7936ddd
      由 Eran Ben Elisha 提交于 
      When driver is reloading during recovery flow, it can't get new commands
till command interface is up again. Otherwise we may get to null pointer
trying to access non initialized command structures.

Add cmdif state to avoid processing commands while cmdif is not ready.

Fixes: e126ba97 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NMoshe Shemesh <moshe@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      f7936ddd
    • E
      net/mlx5: Fix a race when moving command interface to events mode · d43b7007
      由 Eran Ben Elisha 提交于 
      After driver creates (via FW command) an EQ for commands, the driver will
be informed on new commands completion by EQE. However, due to a race in
driver's internal command mode metadata update, some new commands will
still be miss-handled by driver as if we are in polling mode. Such commands
can get two non forced completion, leading to already freed command entry
access.

CREATE_EQ command, that maps EQ to the command queue must be posted to the
command queue while it is empty and no other command should be posted.

Add SW mechanism that once the CREATE_EQ command is about to be executed,
all other commands will return error without being sent to the FW. Allow
sending other commands only after successfully changing the driver's
internal command mode metadata.
We can safely return error to all other commands while creating the command
EQ, as all other commands might be sent from the user/application during
driver load. Application can rerun them later after driver's load was
finished.

Fixes: e126ba97 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NMoshe Shemesh <moshe@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      d43b7007
    • M
      net/mlx5: Add command entry handling completion · 17d00e83
      由 Moshe Shemesh 提交于 
      When FW response to commands is very slow and all command entries in
use are waiting for completion we can have a race where commands can get
timeout before they get out of the queue and handled. Timeout
completion on uninitialized command will cause releasing command's
buffers before accessing it for initialization and then we will get NULL
pointer exception while trying access it. It may also cause releasing
buffers of another command since we may have timeout completion before
even allocating entry index for this command.
Add entry handling completion to avoid this race.

Fixes: e126ba97 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: NMoshe Shemesh <moshe@mellanox.com>
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
      17d00e83
    • Q
      net: sun: fix missing release regions in cas_init_one(). · 5a730153
      由 Qiushi Wu 提交于 
      In cas_init_one(), "pdev" is requested by "pci_request_regions", but it
was not released after a call of the function “pci_write_config_byte”
failed. Thus replace the jump target “err_write_cacheline” by
"err_out_free_res".

Fixes: 1f26dac3 ("[NET]: Add Sun Cassini driver.")
Signed-off-by: NQiushi Wu <wu000273@umn.edu>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      5a730153
    • V
      net: mscc: ocelot: fix address ageing time (again) · bf655ba2
      由 Vladimir Oltean 提交于 
      ocelot_set_ageing_time has 2 callers:
 - felix_set_ageing_time: from drivers/net/dsa/ocelot/felix.c
 - ocelot_port_attr_ageing_set: from drivers/net/ethernet/mscc/ocelot.c

The issue described in the fixed commit below actually happened for the
felix_set_ageing_time code path only, since ocelot_port_attr_ageing_set
was already dividing by 1000. So to make both paths symmetrical (and to
fix addresses getting aged way too fast on Ocelot), stop dividing by
1000 at caller side altogether.

Fixes: c0d7eccb ("net: mscc: ocelot: ANA_AUTOAGE_AGE_PERIOD holds a value in seconds, not ms")
Signed-off-by: NVladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      bf655ba2
    • H
      r8169: fix OCP access on RTL8117 · 561535b0
      由 Heiner Kallweit 提交于 
      According to r8168 vendor driver DASHv3 chips like RTL8168fp/RTL8117
need a special addressing for OCP access.
Fix is compile-tested only due to missing test hardware.

Fixes: 1287723a ("r8169: add support for RTL8117")
Signed-off-by: NHeiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      561535b0
    • J
      mlxsw: spectrum: Fix use-after-free of split/unsplit/type_set in case reload fails · 4340f42f
      由 Jiri Pirko 提交于 
      In case of reload fail, the mlxsw_sp->ports contains a pointer to a
freed memory (either by reload_down() or reload_up() error path).
Fix this by initializing the pointer to NULL and checking it before
dereferencing in split/unsplit/type_set callpaths.

Fixes: 24cc68ad ("mlxsw: core: Add support for reload")
Reported-by: NDanielle Ratson <danieller@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      4340f42f
    • J
      net: ethernet: stmmac: Enable interface clocks on probe for IPQ806x · a96ac8a0
      由 Jonathan McDowell 提交于 
      The ipq806x_gmac_probe() function enables the PTP clock but not the
appropriate interface clocks. This means that if the bootloader hasn't
done so attempting to bring up the interface will fail with an error
like:

[   59.028131] ipq806x-gmac-dwmac 37600000.ethernet: Failed to reset the dma
[   59.028196] ipq806x-gmac-dwmac 37600000.ethernet eth1: stmmac_hw_setup: DMA engine initialization failed
[   59.034056] ipq806x-gmac-dwmac 37600000.ethernet eth1: stmmac_open: Hw setup failed

This patch, a slightly cleaned up version of one posted by Sergey
Sergeev in:

https://forum.openwrt.org/t/support-for-mikrotik-rb3011uias-rm/4064/257

correctly enables the clock; we have already configured the source just
before this.

Tested on a MikroTik RB3011.
Signed-off-by: NJonathan McDowell <noodles@earth.li>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      a96ac8a0
    • I
      netdevsim: Ensure policer drop counter always increases · be43224f
      由 Ido Schimmel 提交于 
      In case the policer drop counter is retrieved when the jiffies value is
a multiple of 64, the counter will not be incremented.

This randomly breaks a selftest [1] the reads the counter twice and
checks that it was incremented:

```
TEST: Trap policer                                                  [FAIL]
	Policer drop counter was not incremented
```

Fix by always incrementing the counter by 1.

[1] tools/testing/selftests/drivers/net/netdevsim/devlink_trap.sh

Fixes: ad188458 ("netdevsim: Add devlink-trap policer support")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Reviewed-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      be43224f
    • V
      net/ethernet/freescale: rework quiesce/activate for ucc_geth · 79dde73c
      由 Valentin Longchamp 提交于 
      ugeth_quiesce/activate are used to halt the controller when there is a
link change that requires to reconfigure the mac.

The previous implementation called netif_device_detach(). This however
causes the initial activation of the netdevice to fail precisely because
it's detached. For details, see [1].

A possible workaround was the revert of commit
net: linkwatch: add check for netdevice being present to linkwatch_do_dev
However, the check introduced in the above commit is correct and shall be
kept.

The netif_device_detach() is thus replaced with
netif_tx_stop_all_queues() that prevents any tranmission. This allows to
perform mac config change required by the link change, without detaching
the corresponding netdevice and thus not preventing its initial
activation.

[1] https://lists.openwall.net/netdev/2020/01/08/201Signed-off-by: NValentin Longchamp <valentin@longchamp.me>
Acked-by: NMatteo Ghidoni <matteo.ghidoni@ch.abb.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      79dde73c
    • R
      net: mvpp2: fix RX hashing for non-10G ports · 3138a07c
      由 Russell King 提交于 
      When rxhash is enabled on any ethernet port except the first in each CP
block, traffic flow is prevented.  The analysis is below:

I've been investigating this afternoon, and what I've found, comparing
a kernel without 895586d5 and with 895586d5 applied is:

- The table programmed into the hardware via mvpp22_rss_fill_table()
  appears to be identical with or without the commit.

- When rxhash is enabled on eth2, mvpp2_rss_port_c2_enable() reports
  that c2.attr[0] and c2.attr[2] are written back containing:

   - with 895586d5, failing:    00200000 40000000
   - without 895586d5, working: 04000000 40000000

- When disabling rxhash, c2.attr[0] and c2.attr[2] are written back as:

   04000000 00000000

The second value represents the MVPP22_CLS_C2_ATTR2_RSS_EN bit, the
first value is the queue number, which comprises two fields. The high
5 bits are 24:29 and the low three are 21:23 inclusive. This comes
from:

       c2.attr[0] = MVPP22_CLS_C2_ATTR0_QHIGH(qh) |
                     MVPP22_CLS_C2_ATTR0_QLOW(ql);

So, the working case gives eth2 a queue id of 4.0, or 32 as per
port->first_rxq, and the non-working case a queue id of 0.1, or 1.
The allocation of queue IDs seems to be in mvpp2_port_probe():

        if (priv->hw_version == MVPP21)
                port->first_rxq = port->id * port->nrxqs;
        else
                port->first_rxq = port->id * priv->max_port_rxqs;

Where:

        if (priv->hw_version == MVPP21)
                priv->max_port_rxqs = 8;
        else
                priv->max_port_rxqs = 32;

Making the port 0 (eth0 / eth1) have port->first_rxq = 0, and port 1
(eth2) be 32. It seems the idea is that the first 32 queues belong to
port 0, the second 32 queues belong to port 1, etc.

mvpp2_rss_port_c2_enable() gets the queue number from it's parameter,
'ctx', which comes from mvpp22_rss_ctx(port, 0). This returns
port->rss_ctx[0].

mvpp22_rss_context_create() is responsible for allocating that, which
it does by looking for an unallocated priv->rss_tables[] pointer. This
table is shared amongst all ports on the CP silicon.

When we write the tables in mvpp22_rss_fill_table(), the RSS table
entry is defined by:

                u32 sel = MVPP22_RSS_INDEX_TABLE(rss_ctx) |
                          MVPP22_RSS_INDEX_TABLE_ENTRY(i);

where rss_ctx is the context ID (queue number) and i is the index in
the table.

If we look at what is written:

- The first table to be written has "sel" values of 00000000..0000001f,
  containing values 0..3. This appears to be for eth1. This is table 0,
  RX queue number 0.
- The second table has "sel" values of 00000100..0000011f, and appears
  to be for eth2.  These contain values 0x20..0x23. This is table 1,
  RX queue number 0.
- The third table has "sel" values of 00000200..0000021f, and appears
  to be for eth3.  These contain values 0x40..0x43. This is table 2,
  RX queue number 0.

How do queue numbers translate to the RSS table?  There is another
table - the RXQ2RSS table, indexed by the MVPP22_RSS_INDEX_QUEUE field
of MVPP22_RSS_INDEX and accessed through the MVPP22_RXQ2RSS_TABLE
register. Before 895586d5, it was:

       mvpp2_write(priv, MVPP22_RSS_INDEX,
                   MVPP22_RSS_INDEX_QUEUE(port->first_rxq));
       mvpp2_write(priv, MVPP22_RXQ2RSS_TABLE,
                   MVPP22_RSS_TABLE_POINTER(port->id));

and after:

       mvpp2_write(priv, MVPP22_RSS_INDEX, MVPP22_RSS_INDEX_QUEUE(ctx));
       mvpp2_write(priv, MVPP22_RXQ2RSS_TABLE, MVPP22_RSS_TABLE_POINTER(ctx));

Before the commit, for eth2, that would've contained '32' for the
index and '1' for the table pointer - mapping queue 32 to table 1.
Remember that this is queue-high.queue-low of 4.0.

After the commit, we appear to map queue 1 to table 1. That again
looks fine on the face of it.

Section 9.3.1 of the A8040 manual seems indicate the reason that the
queue number is separated. queue-low seems to always come from the
classifier, whereas queue-high can be from the ingress physical port
number or the classifier depending on the MVPP2_CLS_SWFWD_PCTRL_REG.

We set the port bit in MVPP2_CLS_SWFWD_PCTRL_REG, meaning that queue-high
comes from the MVPP2_CLS_SWFWD_P2HQ_REG() register... and this seems to
be where our bug comes from.

mvpp2_cls_oversize_rxq_set() sets this up as:

        mvpp2_write(port->priv, MVPP2_CLS_SWFWD_P2HQ_REG(port->id),
                    (port->first_rxq >> MVPP2_CLS_OVERSIZE_RXQ_LOW_BITS));

        val = mvpp2_read(port->priv, MVPP2_CLS_SWFWD_PCTRL_REG);
        val |= MVPP2_CLS_SWFWD_PCTRL_MASK(port->id);
        mvpp2_write(port->priv, MVPP2_CLS_SWFWD_PCTRL_REG, val);

Setting the MVPP2_CLS_SWFWD_PCTRL_MASK bit means that the queue-high
for eth2 is _always_ 4, so only queues 32 through 39 inclusive are
available to eth2. Yet, we're trying to tell the classifier to set
queue-high, which will be ignored, to zero. Hence, the queue-high
field (MVPP22_CLS_C2_ATTR0_QHIGH()) from the classifier will be
ignored.

This means we end up directing traffic from eth2 not to queue 1, but
to queue 33, and then we tell it to look up queue 33 in the RSS table.
However, RSS table has not been programmed for queue 33, and so it ends
up (presumably) dropping the packets.

It seems that mvpp22_rss_context_create() doesn't take account of the
fact that the upper 5 bits of the queue ID can't actually be changed
due to the settings in mvpp2_cls_oversize_rxq_set(), _or_ it seems that
mvpp2_cls_oversize_rxq_set() has been missed in this commit. Either
way, these two functions mutually disagree with what queue number
should be used.

Looking deeper into what mvpp2_cls_oversize_rxq_set() and the MTU
validation is doing, it seems that MVPP2_CLS_SWFWD_P2HQ_REG() is used
for over-sized packets attempting to egress through this port. With
the classifier having had RSS enabled and directing eth2 traffic to
queue 1, we may still have packets appearing on queue 32 for this port.

However, the only way we may end up with over-sized packets attempting
to egress through eth2 - is if the A8040 forwards frames between its
ports. From what I can see, we don't support that feature, and the
kernel restricts the egress packet size to the MTU. In any case, if we
were to attempt to transmit an oversized packet, we have no support in
the kernel to deal with that appearing in the port's receive queue.

So, this patch attempts to solve the issue by clearing the
MVPP2_CLS_SWFWD_PCTRL_MASK() bit, allowing MVPP22_CLS_C2_ATTR0_QHIGH()
from the classifier to define the queue-high field of the queue number.

My testing seems to confirm my findings above - clearing this bit
means that if I enable rxhash on eth2, the interface can then pass
traffic, as we are now directing traffic to RX queue 1 rather than
queue 33. Traffic still seems to work with rxhash off as well.
Reported-by: NMatteo Croce <mcroce@redhat.com>
Tested-by: NMatteo Croce <mcroce@redhat.com>
Fixes: 895586d5 ("net: mvpp2: cls: Use RSS contexts to handle RSS tables")
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      3138a07c
    • C
      felix: Fix initialization of ioremap resources · b4024c9e
      由 Claudiu Manoil 提交于 
      The caller of devm_ioremap_resource(), either accidentally
or by wrong assumption, is writing back derived resource data
to global static resource initialization tables that should
have been constant.  Meaning that after it computes the final
physical start address it saves the address for no reason
in the static tables.  This doesn't affect the first driver
probing after reboot, but it breaks consecutive driver reloads
(i.e. driver unbind & bind) because the initialization tables
no longer have the correct initial values.  So the next probe()
will map the device registers to wrong physical addresses,
causing ARM SError async exceptions.
This patch fixes all of the above.

Fixes: 56051948 ("net: dsa: ocelot: add driver for Felix switch family")
Signed-off-by: NClaudiu Manoil <claudiu.manoil@nxp.com>
Reviewed-by: NVladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: NVladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      b4024c9e
  9. 22 5月, 2020 3 次提交
  11. 21 5月, 2020 2 次提交
    • J
      wireguard: noise: separate receive counter from send counter · a9e90d99
      由 Jason A. Donenfeld 提交于 
      In "wireguard: queueing: preserve flow hash across packet scrubbing", we
were required to slightly increase the size of the receive replay
counter to something still fairly small, but an increase nonetheless.
It turns out that we can recoup some of the additional memory overhead
by splitting up the prior union type into two distinct types. Before, we
used the same "noise_counter" union for both sending and receiving, with
sending just using a simple atomic64_t, while receiving used the full
replay counter checker. This meant that most of the memory being
allocated for the sending counter was being wasted. Since the old
"noise_counter" type increased in size in the prior commit, now is a
good time to split up that union type into a distinct "noise_replay_
counter" for receiving and a boring atomic64_t for sending, each using
neither more nor less memory than required.

Also, since sometimes the replay counter is accessed without
necessitating additional accesses to the bitmap, we can reduce cache
misses by hoisting the always-necessary lock above the bitmap in the
struct layout. We also change a "noise_replay_counter" stack allocation
to kmalloc in a -DDEBUG selftest so that KASAN doesn't trigger a stack
frame warning.

All and all, removing a bit of abstraction in this commit makes the code
simpler and smaller, in addition to the motivating memory usage
recuperation. For example, passing around raw "noise_symmetric_key"
structs is something that really only makes sense within noise.c, in the
one place where the sending and receiving keys can safely be thought of
as the same type of object; subsequent to that, it's important that we
uniformly access these through keypair->{sending,receiving}, where their
distinct roles are always made explicit. So this patch allows us to draw
that distinction clearly as well.

Fixes: e7096c13 ("net: WireGuard secure network tunnel")
Signed-off-by: NJason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      a9e90d99
    • J
      wireguard: queueing: preserve flow hash across packet scrubbing · c78a0b4a
      由 Jason A. Donenfeld 提交于 
      It's important that we clear most header fields during encapsulation and
decapsulation, because the packet is substantially changed, and we don't
want any info leak or logic bug due to an accidental correlation. But,
for encapsulation, it's wrong to clear skb->hash, since it's used by
fq_codel and flow dissection in general. Without it, classification does
not proceed as usual. This change might make it easier to estimate the
number of innerflows by examining clustering of out of order packets,
but this shouldn't open up anything that can't already be inferred
otherwise (e.g. syn packet size inference), and fq_codel can be disabled
anyway.

Furthermore, it might be the case that the hash isn't used or queried at
all until after wireguard transmits the encrypted UDP packet, which
means skb->hash might still be zero at this point, and thus no hash
taken over the inner packet data. In order to address this situation, we
force a calculation of skb->hash before encrypting packet data.

Of course this means that fq_codel might transmit packets slightly more
out of order than usual. Toke did some testing on beefy machines with
high quantities of parallel flows and found that increasing the
reply-attack counter to 8192 takes care of the most pathological cases
pretty well.
Reported-by: NDave Taht <dave.taht@gmail.com>
Reviewed-and-tested-by: NToke Høiland-Jørgensen <toke@toke.dk>
Fixes: e7096c13 ("net: WireGuard secure network tunnel")
Signed-off-by: NJason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      c78a0b4a