====================
100GbE Intel Wired LAN Driver Updates 2020-10-07
This series contains updates to ice driver only.
Andy Shevchenko changes usage to %*phD format to print small buffer as hex
string.
Bruce removes repeated words reported by checkpatch.
Ani changes ice_info_get_dsn() to return void as it always returns
success.
Jake adds devlink reporting of fw.app.bundle_id. Moves devlink_port
structure to ice_vsi to resolve issues with cleanup. Adds additional
debug info for firmware updates.
Bixuan Cui resolves -Wpointer-to-int-cast warnings.
Dan adds additional packet type masks and checks to prevent overwriting
existing Flow Director rules.
====================
Dan Nowlin [Wed, 7 Oct 2020 17:54:47 +0000 (10:54 -0700)]
ice: fix adding IP4 IP6 Flow Director rules
A subsequent addition of an IP4 or IP6 rule after other rules would
overwrite any existing TCAM entries of related L4 protocols(ex: tcp4 or
udp6). This was due to the mask including too many TCAM entries. Add new
packet type masks with bits properly excluded so rules are not overwritten.
Signed-off-by: Dan Nowlin <dan.nowlin@intel.com> Signed-off-by: Henry Tieman <henry.w.tieman@intel.com> Tested-by: Brijesh Behera <brijeshx.behera@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Bixuan Cui [Wed, 7 Oct 2020 17:54:46 +0000 (10:54 -0700)]
ice: Fix pointer cast warnings
pointers should be casted to unsigned long to avoid
-Wpointer-to-int-cast warnings:
drivers/net/ethernet/intel/ice/ice_flow.h:197:33: warning:
cast from pointer to integer of different size
drivers/net/ethernet/intel/ice/ice_flow.h:198:32: warning:
cast to pointer from integer of different size
Signed-off-by: Bixuan Cui <cuibixuan@huawei.com> Tested-by: Aaron Brown <aaron.f.brown@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jacob Keller [Wed, 7 Oct 2020 17:54:45 +0000 (10:54 -0700)]
ice: add additional debug logging for firmware update
While debugging a recent failure to update the flash of an ice device,
I found it helpful to add additional logging which helped determine the
root cause of the problem being a timeout issue.
Add some extra dev_dbg() logging messages which can be enabled using the
dynamic debug facility, including one for ice_aq_wait_for_event that
will use jiffies to capture a rough estimate of how long we waited for
the completion of a firmware command.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Brijesh Behera <brijeshx.behera@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jacob Keller [Wed, 7 Oct 2020 17:54:44 +0000 (10:54 -0700)]
ice: refactor devlink_port to be per-VSI
Currently, the devlink_port structure is stored within the ice_pf. This
made sense because we create a single devlink_port for each PF. This
setup does not mesh with the abstractions in the driver very well, and
led to a flow where we accidentally call devlink_port_unregister twice
during error cleanup.
In particular, if devlink_port_register or devlink_port_unregister are
called twice, this leads to a kernel panic. This appears to occur during
some possible flows while cleaning up from a failure during driver
probe.
If register_netdev fails, then we will call devlink_port_unregister in
ice_cfg_netdev as it cleans up. Later, we again call
devlink_port_unregister since we assume that we must cleanup the port
that is associated with the PF structure.
This occurs because we cleanup the devlink_port for the main PF even
though it was not allocated. We allocated the port within a per-VSI
function for managing the main netdev, but did not release the port when
cleaning up that VSI, the allocation and destruction are not aligned.
Instead of attempting to manage the devlink_port as part of the PF
structure, manage it as part of the PF VSI. Doing this has advantages,
as we can match the de-allocation of the devlink_port with the
unregister_netdev associated with the main PF VSI.
Moving the port to the VSI is preferable as it paves the way for
handling devlink ports allocated for other purposes such as SR-IOV VFs.
Since we're changing up how we allocate the devlink_port, also change
the indexing. Originally, we indexed the port using the PF id number.
This came from an old goal of sharing a devlink for each physical
function. Managing devlink instances across multiple function drivers is
not workable. Instead, lets set the port number to the logical port
number returned by firmware and set the index using the VSI index
(sometimes referred to as VSI handle).
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Aaron Brown <aaron.f.brown@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jacob Keller [Wed, 7 Oct 2020 17:54:43 +0000 (10:54 -0700)]
ice: add the DDP Track ID to devlink info
Add "fw.app.bundle_id" to display the DDP Track ID of the active DDP
package. This id is similar to "fw.bundle_id" and is a unique identifier
for the DDP package that is loaded in the device. Each new DDP has
a unique Track ID generated for it, and the ID can be used to identify
and track the DDP package.
Add documentation for the new devlink info version.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Aaron Brown <aaron.f.brown@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Bruce Allan [Wed, 7 Oct 2020 17:54:41 +0000 (10:54 -0700)]
ice: remove repeated words
A new test in checkpatch detects repeated words; cleanup all pre-existing
occurrences of those now.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com> Tested-by: Aaron Brown <aaron.f.brown@intel.com> Co-developed-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Andy Shevchenko [Wed, 7 Oct 2020 17:54:40 +0000 (10:54 -0700)]
ice: devlink: use %*phD to print small buffer
Use %*phD format to print small buffer as hex string.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Jacob Keller <jacob.e.keller@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net: dsa: microchip: add ksz9563 to ksz9477 I2C driver
Add support for the KSZ9563 3-Port Gigabit Ethernet Switch to the
ksz9477 driver. The KSZ9563 supports both SPI (already in) and I2C. The
ksz9563 is already in the device tree binding documentation.
Signed-off-by: Christian Eggers <ceggers@arri.de> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Manjunath Patil [Tue, 6 Oct 2020 21:31:37 +0000 (14:31 -0700)]
net/rds: suppress page allocation failure error in recv buffer refill
RDS/IB tries to refill the recv buffer in softirq context using
GFP_NOWAIT flag. However alloc failure is handled by queueing a work to
refill the recv buffer with GFP_KERNEL flag. This means failure to
allocate with GFP_NOWAIT isn't fatal. Do not print the PAF warnings if
softirq context fails to refill the recv buffer. We will see the PAF
warnings when worker also fails to allocate.
====================
Add devlink reload action and limit options
Introduce new options on devlink reload API to enable the user to select
the reload action required and constrains limits on these actions that he
may want to ensure. Complete support for reload actions in mlx5.
The following reload actions are supported:
driver_reinit: driver entities re-initialization, applying devlink-param
and devlink-resource values.
fw_activate: firmware activate.
The uAPI is backward compatible, if the reload action option is omitted
from the reload command, the driver reinit action will be used.
Note that when required to do firmware activation some drivers may need
to reload the driver. On the other hand some drivers may need to reset
the firmware to reinitialize the driver entities. Therefore, the devlink
reload command returns the actions which were actually performed.
By default reload actions are not limited and driver implementation may
include reset or downtime as needed to perform the actions.
However, if reload limit is selected, the driver should perform only if
it can do it while keeping the limit constraints.
Reload limit added:
no_reset: No reset allowed, no down time allowed, no link flap and no
configuration is lost.
Each driver which supports devlink reload command should expose the
reload actions and limits supported.
Add reload stats to hold the history per reload action per limit.
For example, the number of times fw_activate has been done on this
device since the driver module was added or if the firmware activation
was done with or without reset.
Patch 1 changes devlink_reload_supported() param type to enable using
it before allocating devlink.
Patch 2-3 add the new API reload action and reload limit options to
devlink reload.
Patch 4-5 add reload stats and remote reload stats. These stats are
exposed through devlink dev get.
Patches 6-11 add support on mlx5 for devlink reload action fw_activate
and handle the firmware reset events.
Patches 12-13 add devlink enable remote dev reset parameter and use it
in mlx5.
Patches 14-15 mlx5 add devlink reload limit no_reset support for
fw_activate reload action.
Patch 16 adds documentation file devlink-reload.rst
====================
Moshe Shemesh [Wed, 7 Oct 2020 06:00:56 +0000 (09:00 +0300)]
net/mlx5: Add support for devlink reload limit no reset
Add support for devlink reload action fw_activate with reload limit
no_reset which does firmware live patching, updating the firmware image
without reset, no downtime and no configuration lose. The driver checks
if the firmware is capable of handling the pending firmware changes as a
live patch. If it is then it triggers firmware live patching flow.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:55 +0000 (09:00 +0300)]
net/mlx5: Add support for fw live patch event
Firmware live patch event notifies the driver that the firmware was just
updated using live patch. In such case the driver should not reload or
re-initiate entities, part to updating the firmware version and
re-initiate the firmware tracer which can be updated by live patch with
new strings database to help debugging an issue.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:54 +0000 (09:00 +0300)]
net/mlx5: Add devlink param enable_remote_dev_reset support
The enable_remote_dev_reset devlink param flags that the host admin
allows resets by other hosts. In case it is cleared mlx5 host PF driver
will send NACK on pci sync for firmware update reset request and the
command will fail.
By default enable_remote_dev_reset parameter is true, so pci sync for
firmware update reset is enabled.
The enable_remote_dev_reset devlink param flags that the host admin
allows device resets that can be initiated by other hosts. This
parameter is useful for setups where a device is shared by different
hosts, such as multi-host setup. Once the user set this parameter to
false, the driver should NACK any attempt to reset the device while the
driver is loaded.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:52 +0000 (09:00 +0300)]
net/mlx5: Add support for devlink reload action fw activate
Add support for devlink reload action fw_activate. To activate firmware
image the mlx5 driver resets the firmware and reloads it from flash. If
a new image was stored on flash it will be loaded. Once this reload
command is executed the driver initiates fw sync reset flow, where the
firmware synchronizes all PFs on coming reset and driver reload.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:50 +0000 (09:00 +0300)]
net/mlx5: Handle sync reset now event
On sync_reset_now event the driver does reload and PCI link toggle to
activate firmware upgrade reset. When the firmware sends this event it
syncs the event on all PFs, so all PFs will do PCI link toggle at once.
To do PCI link toggle, the driver ensures that no other device ID under
the same bridge by checking that all the PF functions under the same PCI
bridge have same device ID. If no other device it uses PCI bridge link
control to turn link down and up.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:49 +0000 (09:00 +0300)]
net/mlx5: Handle sync reset request event
Once the driver gets sync_reset_request from firmware it prepares for the
coming reset and sends acknowledge.
After getting this event the driver expects device reset, either it will
trigger PCI reset on sync_reset_now event or such PCI reset will be
triggered by another PF of the same device. So it moves to reset
requested mode and if it gets PCI reset triggered by the other PF it
detect the reset and reloads.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:46 +0000 (09:00 +0300)]
devlink: Add remote reload stats
Add remote reload stats to hold the history of actions performed due
devlink reload commands initiated by remote host. For example, in case
firmware activation with reset finished successfully but was initiated
by remote host.
The function devlink_remote_reload_actions_performed() is exported to
enable drivers update on remote reload actions performed as it was not
initiated by their own devlink instance.
Expose devlink remote reload stats to the user through devlink dev get
command.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:45 +0000 (09:00 +0300)]
devlink: Add reload stats
Add reload stats to hold the history per reload action type and limit.
For example, the number of times fw_activate has been performed on this
device since the driver module was added or if the firmware activation
was performed with or without reset.
Add devlink notification on stats update.
Expose devlink reload stats to the user through devlink dev get command.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:44 +0000 (09:00 +0300)]
devlink: Add devlink reload limit option
Add reload limit to demand restrictions on reload actions.
Reload limits supported:
no_reset: No reset allowed, no down time allowed, no link flap and no
configuration is lost.
By default reload limit is unspecified and so no constraints on reload
actions are required.
Some combinations of action and limit are invalid. For example, driver
can not reinitialize its entities without any downtime.
The no_reset reload limit will have usecase in this patchset to
implement restricted fw_activate on mlx5.
Have the uapi parameter of reload limit ready for future support of
multiselection.
Moshe Shemesh [Wed, 7 Oct 2020 06:00:43 +0000 (09:00 +0300)]
devlink: Add reload action option to devlink reload command
Add devlink reload action to allow the user to request a specific reload
action. The action parameter is optional, if not specified then devlink
driver re-init action is used (backward compatible).
Note that when required to do firmware activation some drivers may need
to reload the driver. On the other hand some drivers may need to reset
the firmware to reinitialize the driver entities. Therefore, the devlink
reload command returns the actions which were actually performed.
Reload actions supported are:
driver_reinit: driver entities re-initialization, applying devlink-param
and devlink-resource values.
fw_activate: firmware activate.
command examples:
$devlink dev reload pci/0000:82:00.0 action driver_reinit
reload_actions_performed:
driver_reinit
$devlink dev reload pci/0000:82:00.0 action fw_activate
reload_actions_performed:
driver_reinit fw_activate
Signed-off-by: Moshe Shemesh <moshe@mellanox.com> Reviewed-by: Jakub Kicinski <kuba@kernel.org> Reviewed-by: Jacob Keller <jacob.e.keller@intel.com> Reviewed-by: Jiri Pirko <jiri@nvidia.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Moshe Shemesh [Wed, 7 Oct 2020 06:00:42 +0000 (09:00 +0300)]
devlink: Change devlink_reload_supported() param type
Change devlink_reload_supported() function to get devlink_ops pointer
param instead of devlink pointer param.
This change will be used in the next patch to check if devlink reload is
supported before devlink instance is allocated.
Signed-off-by: Moshe Shemesh <moshe@mellanox.com> Reviewed-by: Jakub Kicinski <kuba@kernel.org> Reviewed-by: Jiri Pirko <jiri@nvidia.com> Reviewed-by: Jacob Keller <jacob.e.keller@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet [Wed, 7 Oct 2020 16:51:11 +0000 (09:51 -0700)]
net/sched: get rid of qdisc->padded
kmalloc() of sufficiently big portion of memory is cache-aligned
in regular conditions. If some debugging options are used,
there is no reason qdisc structures would need 64-byte alignment
if most other kernel structures are not aligned.
This get rid of QDISC_ALIGN and QDISC_ALIGNTO.
Addition of privdata field will help implementing
the reverse of qdisc_priv() and documents where
the private data is.
Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Allen Pais <allen.lkml@gmail.com> Acked-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Matthieu Baerts [Tue, 6 Oct 2020 16:06:30 +0000 (18:06 +0200)]
selftests: mptcp: interpret \n as a new line
In case of errors, this message was printed:
(...)
balanced bwidth with unbalanced delay 5233 max 5005 [ fail ]
client exit code 0, server 0
\nnetns ns3-0-EwnkPH socket stat for 10003:
(...)
Obviously, the idea was to add a new line before the socket stat and not
print "\nnetns".
The commit 7d7df4bf8e27 ("selftests: mptcp: interpret \n as a new line")
is very similar to this one. But the modification in simult_flows.sh was
missed because this commit above was done in parallel to one here below.
Fixes: ab564bb667af ("mptcp: simult flow self-tests") Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net> Acked-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Paolo Abeni [Tue, 6 Oct 2020 06:27:34 +0000 (08:27 +0200)]
mptcp: fix infinite loop on recvmsg()/worker() race.
If recvmsg() and the workqueue race to dequeue the data
pending on some subflow, the current mapping for such
subflow covers several skbs and some of them have not
reached yet the received, either the worker or recvmsg()
can find a subflow with the data_avail flag set - since
the current mapping is valid and in sequence - but no
skbs in the receive queue - since the other entity just
processed them.
The above will lead to an unbounded loop in __mptcp_move_skbs()
and a subsequent hang of any task trying to acquiring the msk
socket lock.
This change addresses the issue stopping the __mptcp_move_skbs()
loop as soon as we detect the above race (empty receive queue
with data_avail set).
Reported-and-tested-by: syzbot+fcf8ca5817d6e92c6567@syzkaller.appspotmail.com Fixes: 00a56791f4a8 ("mptcp: move ooo skbs into msk out of order queue.") Signed-off-by: Paolo Abeni <pabeni@redhat.com> Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Allen Pais [Tue, 6 Oct 2020 05:53:32 +0000 (11:23 +0530)]
cxgb4: convert tasklets to use new tasklet_setup() API
In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.
Signed-off-by: Romain Perier <romain.perier@gmail.com> Signed-off-by: Allen Pais <apais@linux.microsoft.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Johannes Berg [Wed, 7 Oct 2020 10:53:51 +0000 (12:53 +0200)]
ethtool: correct policy for ETHTOOL_MSG_CHANNELS_SET
This accidentally got wired up to the *get* policy instead
of the *set* policy, causing operations to be rejected. Fix
it by wiring up the correct policy instead.
Fixes: 5d1a13bcdb96 ("ethtool: wire up set policies to ops") Reported-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Johannes Berg <johannes.berg@intel.com> Tested-by: Leon Romanovsky <leonro@nvidia.com> Reviewed-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The ETHTOOL_A_STRSET_COUNTS_ONLY flag attribute was previously
not allowed to be used, but now due to the policy size reduction
we would access the tb[] array out of bounds since we tried to
check for the attribute despite it not being accepted.
Fix both issues by adding it correctly to the appropriate policy.
Fixes: f815c6bb01dc ("ethtool: trim policy tables") Fixes: d7bef0af95e3 ("ethtool: provide string sets with STRSET_GET request") Reported-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Johannes Berg <johannes.berg@intel.com> Tested-by: Leon Romanovsky <leonro@nvidia.com> Reviewed-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Linus Torvalds [Thu, 8 Oct 2020 21:25:46 +0000 (14:25 -0700)]
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Pull vhost fixes from Michael Tsirkin:
"Some last minute vhost,vdpa fixes.
The last two of them haven't been in next but they do seem kind of
obvious, very small and safe, fix bugs reported in the field, and they
are both in a new mlx5 vdpa driver, so it's not like we can introduce
regressions"
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
vdpa/mlx5: Fix dependency on MLX5_CORE
vdpa/mlx5: should keep avail_index despite device status
vhost-vdpa: fix page pinning leakage in error path
vhost-vdpa: fix vhost_vdpa_map() on error condition
vhost: Don't call log_access_ok() when using IOTLB
vhost: Use vhost_get_used_size() in vhost_vring_set_addr()
vhost: Don't call access_ok() when using IOTLB
vhost vdpa: fix vhost_vdpa_open error handling
Pull networking fixes from Jakub Kicinski:
"One more set of fixes from the networking tree:
- add missing input validation in nl80211_del_key(), preventing
out-of-bounds access
- last minute fix / improvement of a MRP netlink (uAPI) interface
introduced in 5.9 (current) release
- fix "unresolved symbol" build error under CONFIG_NET w/o
CONFIG_INET due to missing tcp_timewait_sock and inet_timewait_sock
BTF.
- fix 32 bit sub-register bounds tracking in the bpf verifier for OR
case
- tcp: fix receive window update in tcp_add_backlog()
- openvswitch: handle DNAT tuple collision in conntrack-related code
- r8169: wait for potential PHY reset to finish after applying a FW
file, avoiding unexpected PHY behaviour and failures later on
- mscc: fix tail dropping watermarks for Ocelot switches
- avoid use-after-free in macsec code after a call to the GRO layer
- avoid use-after-free in sctp error paths
- add a device id for Cellient MPL200 WWAN card
- rxrpc fixes:
- fix the xdr encoding of the contents read from an rxrpc key
- fix a BUG() for a unsupported encoding type.
- fix missing _bh lock annotations.
- fix acceptance handling for an incoming call where the incoming
call is encrypted.
- the server token keyring isn't network namespaced - it belongs
to the server, so there's no need. Namespacing it means that
request_key() fails to find it.
- fix a leak of the server keyring"
* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (21 commits)
net: usb: qmi_wwan: add Cellient MPL200 card
macsec: avoid use-after-free in macsec_handle_frame()
r8169: consider that PHY reset may still be in progress after applying firmware
openvswitch: handle DNAT tuple collision
sctp: fix sctp_auth_init_hmacs() error path
bridge: Netlink interface fix.
net: wireless: nl80211: fix out-of-bounds access in nl80211_del_key()
bpf: Fix scalar32_min_max_or bounds tracking
tcp: fix receive window update in tcp_add_backlog()
net: usb: rtl8150: set random MAC address when set_ethernet_addr() fails
mptcp: more DATA FIN fixes
net: mscc: ocelot: warn when encoding an out-of-bounds watermark value
net: mscc: ocelot: divide watermark value by 60 when writing to SYS_ATOP
net: qrtr: ns: Fix the incorrect usage of rcu_read_lock()
rxrpc: Fix server keyring leak
rxrpc: The server keyring isn't network-namespaced
rxrpc: Fix accept on a connection that need securing
rxrpc: Fix some missing _bh annotations on locking conn->state_lock
rxrpc: Downgrade the BUG() for unsupported token type in rxrpc_read()
rxrpc: Fix rxkad token xdr encoding
...
We should allow userspace emulating the virtio device be
able to get to vq's avail_index, regardless of vDPA device
status. Save the index that was last seen when virtq was
stopped, so that userspace doesn't complain.
Eric Dumazet [Wed, 7 Oct 2020 08:42:46 +0000 (01:42 -0700)]
macsec: avoid use-after-free in macsec_handle_frame()
De-referencing skb after call to gro_cells_receive() is not allowed.
We need to fetch skb->len earlier.
Fixes: d77b70bcf14a ("macsec: enable GRO and RPS on macsec devices") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Paolo Abeni <pabeni@redhat.com> Acked-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Heiner Kallweit [Wed, 7 Oct 2020 11:34:51 +0000 (13:34 +0200)]
r8169: consider that PHY reset may still be in progress after applying firmware
Some firmware files trigger a PHY soft reset and don't wait for it to
be finished. PHY register writes directly after applying the firmware
may fail or provide unexpected results therefore. Fix this by waiting
for bit BMCR_RESET to be cleared after applying firmware.
There's nothing wrong with the referenced change, it's just that the
fix will apply cleanly only after this change.
Fixes: d5a63c9987fb ("r8169: fix firmware not resetting tp->ocp_base") Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Dumitru Ceara [Wed, 7 Oct 2020 15:48:03 +0000 (17:48 +0200)]
openvswitch: handle DNAT tuple collision
With multiple DNAT rules it's possible that after destination
translation the resulting tuples collide.
For example, two openvswitch flows:
nw_dst=10.0.0.10,tp_dst=10, actions=ct(commit,table=2,nat(dst=20.0.0.1:20))
nw_dst=10.0.0.20,tp_dst=10, actions=ct(commit,table=2,nat(dst=20.0.0.1:20))
Assuming two TCP clients initiating the following connections:
10.0.0.10:5000->10.0.0.10:10
10.0.0.10:5000->10.0.0.20:10
Both tuples would translate to 10.0.0.10:5000->20.0.0.1:20 causing
nf_conntrack_confirm() to fail because of tuple collision.
Netfilter handles this case by allocating a null binding for SNAT at
egress by default. Perform the same operation in openvswitch for DNAT
if no explicit SNAT is requested by the user and allocate a null binding
for SNAT for packets in the "original" direction.
Reported-at: https://bugzilla.redhat.com/1877128 Suggested-by: Florian Westphal <fw@strlen.de> Fixes: 12c019f65748 ("openvswitch: Interface with NAT.") Signed-off-by: Dumitru Ceara <dceara@redhat.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet [Thu, 8 Oct 2020 08:38:31 +0000 (01:38 -0700)]
sctp: fix sctp_auth_init_hmacs() error path
After freeing ep->auth_hmacs we have to clear the pointer
or risk use-after-free as reported by syzbot:
BUG: KASAN: use-after-free in sctp_auth_destroy_hmacs net/sctp/auth.c:509 [inline]
BUG: KASAN: use-after-free in sctp_auth_destroy_hmacs net/sctp/auth.c:501 [inline]
BUG: KASAN: use-after-free in sctp_auth_free+0x17e/0x1d0 net/sctp/auth.c:1070
Read of size 8 at addr ffff8880a8ff52c0 by task syz-executor941/6874
1) Fix "unresolved symbol" build error under CONFIG_NET w/o CONFIG_INET due
to missing tcp_timewait_sock and inet_timewait_sock BTF, from Yonghong Song.
2) Fix 32 bit sub-register bounds tracking for OR case, from Daniel Borkmann.
====================
Linus Torvalds [Thu, 8 Oct 2020 18:14:17 +0000 (11:14 -0700)]
Merge tag 'drm-fixes-2020-10-08' of git://anongit.freedesktop.org/drm/drm
Pull drm nouveau fixes from Dave Airlie:
"Karol found two last minute nouveau fixes, they both fix crashes, the
TTM one follows what other drivers do already, and the other is for
bailing on load on unrecognised chipsets.
- fix crash in TTM alloc fail path
- return error earlier for unknown chipsets"
* tag 'drm-fixes-2020-10-08' of git://anongit.freedesktop.org/drm/drm:
drm/nouveau/mem: guard against NULL pointer access in mem_del
drm/nouveau/device: return error for unknown chipsets
Linus Torvalds [Thu, 8 Oct 2020 18:10:13 +0000 (11:10 -0700)]
Merge tag 'exfat-for-5.9-rc9' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/exfat
Pull exfat fixes from Namjae Jeon:
- Fix use of uninitialized spinlock on error path
- Fix missing err assignment in exfat_build_inode()
* tag 'exfat-for-5.9-rc9' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/exfat:
exfat: fix use of uninitialized spinlock on error path
exfat: fix pointer error checking
Linus Torvalds [Thu, 8 Oct 2020 18:01:53 +0000 (11:01 -0700)]
Merge tag 'for-linus-5.9b-rc9-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip
Pull xen fix from Juergen Gross:
"One fix for a regression when booting as a Xen guest on ARM64
introduced probably during the 5.9 cycle. It is very low risk as it is
modifying Xen specific code only.
The exact commit introducing the bug hasn't been identified yet, but
everything was fine in 5.8 and only in 5.9 some configurations started
to fail"
* tag 'for-linus-5.9b-rc9-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
arm/arm64: xen: Fix to convert percpu address to gfn correctly
David Howells [Wed, 7 Oct 2020 13:22:12 +0000 (14:22 +0100)]
afs: Fix deadlock between writeback and truncate
The afs filesystem has a lock[*] that it uses to serialise I/O operations
going to the server (vnode->io_lock), as the server will only perform one
modification operation at a time on any given file or directory. This
prevents the the filesystem from filling up all the call slots to a server
with calls that aren't going to be executed in parallel anyway, thereby
allowing operations on other files to obtain slots.
[*] Note that is probably redundant for directories at least since
i_rwsem is used to serialise directory modifications and
lookup/reading vs modification. The server does allow parallel
non-modification ops, however.
When a file truncation op completes, we truncate the in-memory copy of the
file to match - but we do it whilst still holding the io_lock, the idea
being to prevent races with other operations.
However, if writeback starts in a worker thread simultaneously with
truncation (whilst notify_change() is called with i_rwsem locked, writeback
pays it no heed), it may manage to set PG_writeback bits on the pages that
will get truncated before afs_setattr_success() manages to call
truncate_pagecache(). Truncate will then wait for those pages - whilst
still inside io_lock:
Note that whilst afs_setattr() calls filemap_write_and_wait(), the fact
that the caller is holding i_rwsem doesn't preclude more pages being
dirtied through an mmap'd region.
Fix this by:
(1) Use the vnode validate_lock to mediate access between afs_setattr()
and afs_writepages():
(a) Exclusively lock validate_lock in afs_setattr() around the whole
RPC operation.
(b) If WB_SYNC_ALL isn't set on entry to afs_writepages(), trying to
shared-lock validate_lock and returning immediately if we couldn't
get it.
(c) If WB_SYNC_ALL is set, wait for the lock.
The validate_lock is also used to validate a file and to zap its cache
if the file was altered by a third party, so it's probably a good fit
for this.
(2) Move the truncation outside of the io_lock in setattr, using the same
hook as is used for local directory editing.
This requires the old i_size to be retained in the operation record as
we commit the revised status to the inode members inside the io_lock
still, but we still need to know if we reduced the file size.
Fixes: b7b4753fe8b8 ("afs: Overhaul volume and server record caching and fileserver rotation") Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm: avoid early COW write protect games during fork()
In commit b67d7efe955a ("mm: Do early cow for pinned pages during fork()
for ptes") we write-protected the PTE before doing the page pinning
check, in order to avoid a race with concurrent fast-GUP pinning (which
doesn't take the mm semaphore or the page table lock).
That trick doesn't actually work - it doesn't handle memory ordering
properly, and doing so would be prohibitively expensive.
It also isn't really needed. While we're moving in the direction of
allowing and supporting page pinning without marking the pinned area
with MADV_DONTFORK, the fact is that we've never really supported this
kind of odd "concurrent fork() and page pinning", and doing the
serialization on a pte level is just wrong.
We can add serialization with a per-mm sequence counter, so we know how
to solve that race properly, but we'll do that at a more appropriate
time. Right now this just removes the write protect games.
It also turns out that the write protect games actually break on Power,
as reported by Aneesh Kumar:
"Architecture like ppc64 expects set_pte_at to be not used for updating
a valid pte. This is further explained in commit 39989ef0d34b ("mm:
Use ptep/pmdp_set_numa() for updating _PAGE_NUMA bit")"
net: wireless: nl80211: fix out-of-bounds access in nl80211_del_key()
In nl80211_parse_key(), key.idx is first initialized as -1.
If this value of key.idx remains unmodified and gets returned, and
nl80211_key_allowed() also returns 0, then rdev_del_key() gets called
with key.idx = -1.
This causes an out-of-bounds array access.
Handle this issue by checking if the value of key.idx after
nl80211_parse_key() is called and return -EINVAL if key.idx < 0.
Daniel Borkmann [Wed, 7 Oct 2020 13:48:58 +0000 (15:48 +0200)]
bpf: Fix scalar32_min_max_or bounds tracking
Simon reported an issue with the current scalar32_min_max_or() implementation.
That is, compared to the other 32 bit subreg tracking functions, the code in
scalar32_min_max_or() stands out that it's using the 64 bit registers instead
of 32 bit ones. This leads to bounds tracking issues, for example:
The bound tests on the map value force the upper unsigned bound to be 25769803777
in 64 bit (0b11000000000000000000000000000000001) and then lower one to be 1. By
using OR they are truncated and thus result in the range [1,1] for the 32 bit reg
tracker. This is incorrect given the only thing we know is that the value must be
positive and thus 2147483647 (0b1111111111111111111111111111111) at max for the
subregs. Fix it by using the {u,s}32_{min,max}_value vars instead. This also makes
sense, for example, for the case where we update dst_reg->s32_{min,max}_value in
the else branch we need to use the newly computed dst_reg->u32_{min,max}_value as
we know that these are positive. Previously, in the else branch the 64 bit values
of umin_value=1 and umax_value=32212254719 were used and latter got truncated to
be 1 as upper bound there. After the fix the subreg range is now correct:
Karol Herbst [Tue, 6 Oct 2020 22:05:27 +0000 (00:05 +0200)]
drm/nouveau/device: return error for unknown chipsets
Previously the code relied on device->pri to be NULL and to fail probing
later. We really should just return an error inside nvkm_device_ctor for
unsupported GPUs.
Fixes: ebc222585998 ("drm/nouveau/device: rework mmio mapping code to get rid of second map") Signed-off-by: Karol Herbst <kherbst@redhat.com> Cc: dann frazier <dann.frazier@canonical.com> Cc: dri-devel <dri-devel@lists.freedesktop.org> Cc: Dave Airlie <airlied@redhat.com> Cc: stable@vger.kernel.org Reviewed-by: Jeremy Cline <jcline@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com> Link: https://patchwork.freedesktop.org/patch/msgid/20201006220528.13925-1-kherbst@redhat.com
If exfat_read_root() returns an error, spinlock is used in
exfat_evict_inode() without initialization. This patch combines
exfat_cache_init_inode() with exfat_inode_init_once() to initialize
spinlock by slab constructor.
arm/arm64: xen: Fix to convert percpu address to gfn correctly
Use per_cpu_ptr_to_phys() instead of virt_to_phys() for per-cpu
address conversion.
In xen_starting_cpu(), per-cpu xen_vcpu_info address is converted
to gfn by virt_to_gfn() macro. However, since the virt_to_gfn(v)
assumes the given virtual address is in linear mapped kernel memory
area, it can not convert the per-cpu memory if it is allocated on
vmalloc area.
This depends on CONFIG_NEED_PER_CPU_EMBED_FIRST_CHUNK.
If it is enabled, the first chunk of percpu memory is linear mapped.
In the other case, that is allocated from vmalloc area. Moreover,
if the first chunk of percpu has run out until allocating
xen_vcpu_info, it will be allocated on the 2nd chunk, which is
based on kernel memory or vmalloc memory (depends on
CONFIG_NEED_PER_CPU_KM).
Without this fix and kernel configured to use vmalloc area for
the percpu memory, the Dom0 kernel will fail to boot with following
errors.
Linus Torvalds [Tue, 6 Oct 2020 19:09:29 +0000 (12:09 -0700)]
Merge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Pull arm64 fix from Catalin Marinas:
"Fix a kernel panic in the AES crypto code caused by a BR tail call not
matching the target BTI instruction (when branch target identification
is enabled)"
* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
crypto: arm64: Use x16 with indirect branch to bti_c
Linus Torvalds [Tue, 6 Oct 2020 19:00:52 +0000 (12:00 -0700)]
Merge tag 'platform-drivers-x86-v5.9-3' of git://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86
Pull another x86 platform driver fix from Hans de Goede:
"One final pdx86 fix for Tablet Mode reporting regressions (which make
the keyboard and touchpad unusable) on various Asus notebooks.
These regressions were caused by the asus-nb-wmi and the intel-vbtn
drivers both receiving recent patches to start reporting Tablet Mode /
to report it on more models.
Due to a miscommunication between Andy and me, Andy's earlier pull-req
only contained the fix for the intel-vbtn driver and not the fix for
the asus-nb-wmi code.
This fix has been tested as a downstream patch in Fedora kernels for
approx two weeks with no problems being reported"
* tag 'platform-drivers-x86-v5.9-3' of git://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86:
platform/x86: asus-wmi: Fix SW_TABLET_MODE always reporting 1 on many different models
Linus Torvalds [Tue, 6 Oct 2020 18:05:44 +0000 (11:05 -0700)]
Merge tag 'drm-fixes-2020-10-06-1' of git://anongit.freedesktop.org/drm/drm
Pull drm fixes from Dave Airlie:
"Daniel queued these up last week and I took a long weekend so didn't
get them out, but fixing the OOB access on get font seems like
something we should land and it's cc'ed stable as well.
The other big change is a partial revert for a regression on android
on the clcd fbdev driver, and one other docs fix.
fbdev:
- Re-add FB_ARMCLCD for android
- Fix global-out-of-bounds read in fbcon_get_font()
core:
- Small doc fix"
* tag 'drm-fixes-2020-10-06-1' of git://anongit.freedesktop.org/drm/drm:
drm: drm_dsc.h: fix a kernel-doc markup
Partially revert "video: fbdev: amba-clcd: Retire elder CLCD driver"
fbcon: Fix global-out-of-bounds read in fbcon_get_font()
Fonts: Support FONT_EXTRA_WORDS macros for built-in fonts
fbdev, newport_con: Move FONT_EXTRA_WORDS macros into linux/font.h
Linus Torvalds [Mon, 5 Oct 2020 17:56:22 +0000 (10:56 -0700)]
usermodehelper: reset umask to default before executing user process
Kernel threads intentionally do CLONE_FS in order to follow any changes
that 'init' does to set up the root directory (or cwd).
It is admittedly a bit odd, but it avoids the situation where 'init'
does some extensive setup to initialize the system environment, and then
we execute a usermode helper program, and it uses the original FS setup
from boot time that may be very limited and incomplete.
[ Both Al Viro and Eric Biederman point out that 'pivot_root()' will
follow the root regardless, since it fixes up other users of root (see
chroot_fs_refs() for details), but overmounting root and doing a
chroot() would not. ]
However, Vegard Nossum noticed that the CLONE_FS not only means that we
follow the root and current working directories, it also means we share
umask with whatever init changed it to. That wasn't intentional.
Just reset umask to the original default (0022) before actually starting
the usermode helper program.
Reported-by: Vegard Nossum <vegard.nossum@oracle.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Acked-by: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Mon, 5 Oct 2020 18:26:27 +0000 (11:26 -0700)]
splice: teach splice pipe reading about empty pipe buffers
Tetsuo Handa reports that splice() can return 0 before the real EOF, if
the data in the splice source pipe is an empty pipe buffer. That empty
pipe buffer case doesn't happen in any normal situation, but you can
trigger it by doing a write to a pipe that fails due to a page fault.
Jeremy Linton [Tue, 6 Oct 2020 16:33:26 +0000 (11:33 -0500)]
crypto: arm64: Use x16 with indirect branch to bti_c
The AES code uses a 'br x7' as part of a function called by
a macro. That branch needs a bti_j as a target. This results
in a panic as seen below. Using x16 (or x17) with an indirect
branch keeps the target bti_c.
Bad mode in Synchronous Abort handler detected on CPU1, code 0x34000003 -- BTI
CPU: 1 PID: 265 Comm: cryptomgr_test Not tainted 5.8.11-300.fc33.aarch64 #1
pstate: 20400c05 (nzCv daif +PAN -UAO BTYPE=j-)
pc : aesbs_encrypt8+0x0/0x5f0 [aes_neon_bs]
lr : aesbs_xts_encrypt+0x48/0xe0 [aes_neon_bs]
sp : ffff80001052b730
Fixes: 1b78860cc495 ("crypto: arm64 - Use modern annotations for assembly functions") Cc: <stable@vger.kernel.org> # 5.6.x- Signed-off-by: Jeremy Linton <jeremy.linton@arm.com> Suggested-by: Dave P Martin <Dave.Martin@arm.com> Reviewed-by: Ard Biesheuvel <ardb@kernel.org> Reviewed-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20201006163326.2780619-1-jeremy.linton@arm.com Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
====================
ethtool: allow dumping policies to user space
This series wires up ethtool policies to ops, so they can be
dumped to user space for feature discovery.
First patch wires up GET commands, and second patch wires up SETs.
The policy tables are trimmed to save space and LoC.
Next - take care of linking up nested policies for the header
(which is the policy what we actually care about). And once header
policy is linked make sure that attribute range validation for flags
is done by policy, not a conditions in the code. New type of policy
is needed to validate masks (patch 6).
Netlink as always staying a step ahead of all the other kernel
API interfaces :)
Jakub Kicinski [Mon, 5 Oct 2020 22:07:39 +0000 (15:07 -0700)]
ethtool: specify which header flags are supported per command
Perform header flags validation through the policy.
Only pause command supports ETHTOOL_FLAG_STATS. Create a separate
policy to be able to express that in policy dumps to user space.
Note that even though the core will validate the header policy,
it cannot record multiple layers of attributes and we have to
re-parse header sub-attrs. When doing so we could skip attribute
validation, or use most permissive policy. Opt for the former.
We will no longer return the extack cookie for flags but since
we only added first new flag in this release it's not expected
that any user space had a chance to make use of it.
v2: - remove the re-validation in ethnl_parse_header_dev_get()
Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski [Mon, 5 Oct 2020 22:07:38 +0000 (15:07 -0700)]
netlink: add mask validation
We don't have good validation policy for existing unsigned int attrs
which serve as flags (for new ones we could use NLA_BITFIELD32).
With increased use of policy dumping having the validation be
expressed as part of the policy is important. Add validation
policy in form of a mask of supported/valid bits.
Support u64 in the uAPI to be future-proof, but really for now
the embedded mask member can only hold 32 bits, so anything with
bit 32+ set will always fail validation.
Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski [Mon, 5 Oct 2020 22:07:36 +0000 (15:07 -0700)]
ethtool: link up ethnl_header_policy as a nested policy
To get the most out of parsing by the core, and to allow dumping
full policies we need to specify which policy applies to nested
attrs. For headers it's ethnl_header_policy.
$ sed -i 's@\(ETHTOOL_A_.*HEADER\].*=\) { .type = NLA_NESTED },@\1\n\t\tNLA_POLICY_NESTED(ethnl_header_policy),@' net/ethtool/*
Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski [Mon, 5 Oct 2020 22:07:35 +0000 (15:07 -0700)]
ethtool: trim policy tables
Since ethtool uses strict attribute validation there's no need
to initialize all attributes in policy tables. 0 is NLA_UNSPEC
which is going to be rejected. Remove the NLA_REJECTs.
Similarly attributes above maxattrs are rejected, so there's
no need to always size the policy tables to ETHTOOL_A_..._MAX.
v2: - new patch
Suggested-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski [Mon, 5 Oct 2020 22:07:33 +0000 (15:07 -0700)]
ethtool: wire up get policies to ops
Wire up policies for get commands in struct nla_policy of the ethtool
family. Make use of genetlink code attr validation and parsing, as well
as allow dumping policies to user space.
For every ETHTOOL_MSG_*_GET:
- add 'ethnl_' prefix to policy name
- add extern declaration in net/ethtool/netlink.h
- wire up the policy & attr in ethtool_genl_ops[].
- remove .request_policy and .max_attr from ethnl_request_ops.
Obviously core only records the first "layer" of parsed attrs
so we still need to parse the sub-attrs of the nested header
attribute.
v2:
- merge of patches 1 and 2 from v1
- remove stray empty lines in ops
- also remove .max_attr
Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
This small patchset creates netstats addition dev_sw_netstats_rx_add()
based on dev_lstats_add() and replaces some open coding
in both drivers/net and net branches.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Fabian Frederick <fabf@skynet.be> Acked-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Tue, 6 Oct 2020 13:18:20 +0000 (06:18 -0700)]
Merge tag 'rxrpc-fixes-20201005' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs
David Howells says:
====================
rxrpc: Miscellaneous fixes
Here are some miscellaneous rxrpc fixes:
(1) Fix the xdr encoding of the contents read from an rxrpc key.
(2) Fix a BUG() for a unsupported encoding type.
(3) Fix missing _bh lock annotations.
(4) Fix acceptance handling for an incoming call where the incoming call
is encrypted.
(5) The server token keyring isn't network namespaced - it belongs to the
server, so there's no need. Namespacing it means that request_key()
fails to find it.
(6) Fix a leak of the server keyring.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
====================
net: atlantic: phy tunables from mac driver
This series implements phy tunables settings via MAC driver callbacks.
AQC 10G devices use integrated MAC+PHY solution, where PHY is fully controlled
by MAC firmware. Therefore, it is not possible to implement separate phy driver
for these.
We use ethtool ops callbacks to implement downshift and EDPC tunables.
v3: fixed flaw in EDPD logic, from Andrew
v2: comments from Andrew
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean [Mon, 5 Oct 2020 14:48:38 +0000 (17:48 +0300)]
net: always dump full packets with skb_dump
Currently skb_dump has a restriction to only dump full packet for the
first 5 socket buffers, then only headers will be printed. Remove this
arbitrary and confusing restriction, which is only documented vaguely
("up to") in the comments above the prototype.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Mon, 5 Oct 2020 13:48:13 +0000 (06:48 -0700)]
tcp: fix receive window update in tcp_add_backlog()
We got reports from GKE customers flows being reset by netfilter
conntrack unless nf_conntrack_tcp_be_liberal is set to 1.
Traces seemed to suggest ACK packet being dropped by the
packet capture, or more likely that ACK were received in the
wrong order.
wscale=7, SYN and SYNACK not shown here.
This ACK allows the sender to send 1871*128 bytes from seq 51359321 :
New right edge of the window -> 51359321+1871*128=51598809
09:17:23.389210 IP A > B: Flags [.], ack 51359321, win 1871, options [nop,nop,TS val 10 ecr 999], length 0
09:17:23.389212 IP B > A: Flags [.], seq 51422681:51424089, ack 1577, win 268, options [nop,nop,TS val 999 ecr 10], length 1408
09:17:23.389214 IP A > B: Flags [.], ack 51422681, win 1376, options [nop,nop,TS val 10 ecr 999], length 0
09:17:23.389253 IP B > A: Flags [.], seq 51424089:51488857, ack 1577, win 268, options [nop,nop,TS val 999 ecr 10], length 64768
09:17:23.389272 IP A > B: Flags [.], ack 51488857, win 859, options [nop,nop,TS val 10 ecr 999], length 0
09:17:23.389275 IP B > A: Flags [.], seq 51488857:51521241, ack 1577, win 268, options [nop,nop,TS val 999 ecr 10], length 32384
Receiver now allows to send 606*128=77568 from seq 51521241 :
New right edge of the window -> 51521241+606*128=51598809
09:17:23.389296 IP A > B: Flags [.], ack 51521241, win 606, options [nop,nop,TS val 10 ecr 999], length 0
09:17:23.389308 IP B > A: Flags [.], seq 51521241:51553625, ack 1577, win 268, options [nop,nop,TS val 999 ecr 10], length 32384
It seems the sender exceeds RWIN allowance, since 51611353 > 51598809
09:17:23.389346 IP B > A: Flags [.], seq 51553625:51611353, ack 1577, win 268, options [nop,nop,TS val 999 ecr 10], length 57728
09:17:23.389356 IP B > A: Flags [.], seq 51611353:51618393, ack 1577, win 268, options [nop,nop,TS val 999 ecr 10], length 7040
09:17:23.389367 IP A > B: Flags [.], ack 51611353, win 0, options [nop,nop,TS val 10 ecr 999], length 0
netfilter conntrack is not happy and sends RST
09:17:23.389389 IP A > B: Flags [R], seq 92176528, win 0, length 0
09:17:23.389488 IP B > A: Flags [R], seq 174478967, win 0, length 0
Now imagine ACK were delivered out of order and tcp_add_backlog() sets window based on wrong packet.
New right edge of the window -> 51521241+859*128=51631193
Normally TCP stack handles OOO packets just fine, but it
turns out tcp_add_backlog() does not. It can update the window
field of the aggregated packet even if the ACK sequence
of the last received packet is too old.
Many thanks to Alexandre Ferrieux for independently reporting the issue
and suggesting a fix.
Fixes: f6ff999993ab ("tcp: implement coalescing on backlog queue") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Alexandre Ferrieux <alexandre.ferrieux@orange.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Acked-by: Neal Cardwell <ncardwell@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
net: usb: rtl8150: set random MAC address when set_ethernet_addr() fails
When get_registers() fails in set_ethernet_addr(),the uninitialized
value of node_id gets copied over as the address.
So, check the return value of get_registers().
If get_registers() executed successfully (i.e., it returns
sizeof(node_id)), copy over the MAC address using ether_addr_copy()
(instead of using memcpy()).
Else, if get_registers() failed instead, a randomly generated MAC
address is set as the MAC address instead.
Reported-by: syzbot+abbc768b560c84d92fd3@syzkaller.appspotmail.com Tested-by: syzbot+abbc768b560c84d92fd3@syzkaller.appspotmail.com Acked-by: Petko Manolov <petkan@nucleusys.com> Signed-off-by: Anant Thazhemadam <anant.thazhemadam@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Paolo Abeni [Mon, 5 Oct 2020 10:36:44 +0000 (12:36 +0200)]
mptcp: don't skip needed ack
Currently we skip calling tcp_cleanup_rbuf() when packets
are moved into the OoO queue or simply dropped. In both
cases we still increment tp->copied_seq, and we should
ask the TCP stack to check for ack.
Fixes: 0c1dca43441d ("mptcp: call tcp_cleanup_rbuf on subflows") Signed-off-by: Paolo Abeni <pabeni@redhat.com> Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Paolo Abeni [Mon, 5 Oct 2020 10:01:06 +0000 (12:01 +0200)]
mptcp: more DATA FIN fixes
Currently data fin on data packet are not handled properly:
the 'rcv_data_fin_seq' field is interpreted as the last
sequence number carrying a valid data, but for data fin
packet with valid maps we currently store map_seq + map_len,
that is, the next value.
The 'write_seq' fields carries instead the value subseguent
to the last valid byte, so in mptcp_write_data_fin() we
never detect correctly the last DSS map.
Fixes: df82c6e19f5c ("mptcp: Use MPTCP-level flag for sending DATA_FIN") Fixes: d3d5656b0fd5 ("mptcp: Handle incoming 32-bit DATA_FIN values") Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean [Mon, 5 Oct 2020 09:09:12 +0000 (12:09 +0300)]
net: mscc: ocelot: warn when encoding an out-of-bounds watermark value
There is an upper bound to the value that a watermark may hold. That
upper bound is not immediately obvious during configuration, and it
might be possible to have accidental truncation.
Actually this has happened already, add a warning to prevent it from
happening again.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean [Mon, 5 Oct 2020 09:09:11 +0000 (12:09 +0300)]
net: mscc: ocelot: divide watermark value by 60 when writing to SYS_ATOP
Tail dropping is enabled for a port when:
1. A source port consumes more packet buffers than the watermark encoded
in SYS:PORT:ATOP_CFG.ATOP.
AND
2. Total memory use exceeds the consumption watermark encoded in
SYS:PAUSE_CFG:ATOP_TOT_CFG.
The unit of these watermarks is a 60 byte memory cell. That unit is
programmed properly into ATOP_TOT_CFG, but not into ATOP. Actually when
written into ATOP, it would get truncated and wrap around.
Fixes: 1bed22a67906 ("net: mscc: Add initial Ocelot switch support") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
net: qrtr: ns: Fix the incorrect usage of rcu_read_lock()
The rcu_read_lock() is not supposed to lock the kernel_sendmsg() API
since it has the lock_sock() in qrtr_sendmsg() which will sleep. Hence,
fix it by excluding the locking for kernel_sendmsg().
While at it, let's also use radix_tree_deref_retry() to confirm the
validity of the pointer returned by radix_tree_deref_slot() and use
radix_tree_iter_resume() to resume iterating the tree properly before
releasing the lock as suggested by Doug.
Fixes: 943718fb61b6 ("net: qrtr: ns: Protect radix_tree_deref_slot() using rcu read locks") Reported-by: Douglas Anderson <dianders@chromium.org> Reviewed-by: Douglas Anderson <dianders@chromium.org> Tested-by: Douglas Anderson <dianders@chromium.org> Tested-by: Alex Elder <elder@linaro.org> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Hans de Goede [Wed, 16 Sep 2020 14:14:39 +0000 (16:14 +0200)]
platform/x86: asus-wmi: Fix SW_TABLET_MODE always reporting 1 on many different models
Commit 3c36ceada8a4 ("platform/x86: asus-wmi: Add support for
SW_TABLET_MODE") added support for reporting SW_TABLET_MODE using the
Asus 0x00120063 WMI-device-id to see if various transformer models were
docked into their keyboard-dock (SW_TABLET_MODE=0) or if they were
being used as a tablet.
The new SW_TABLET_MODE support (naively?) assumed that non Transformer
devices would either not support the 0x00120063 WMI-device-id at all,
or would NOT set ASUS_WMI_DSTS_PRESENCE_BIT in their reply when querying
the device-id.
Unfortunately this is not true and we have received many bug reports about
this change causing the asus-wmi driver to always report SW_TABLET_MODE=1
on non Transformer devices. This causes libinput to think that these are
360 degree hinges style 2-in-1s folded into tablet-mode. Making libinput
suppress keyboard and touchpad events from the builtin keyboard and
touchpad. So effectively this causes the keyboard and touchpad to not work
on many non Transformer Asus models.
This commit fixes this by using the existing DMI based quirk mechanism in
asus-nb-wmi.c to allow using the 0x00120063 device-id for reporting
SW_TABLET_MODE on Transformer models and ignoring it on all other models.
projects / kernel.git / log