linux-stable.git/net/core/devlink.c, branch v5.5.7

devlink: report 0 after hitting end in region read

2020-02-11T12:37:22+00:00

[ Upstream commit d5b90e99e1d51b7b5d2b74fbc4c2db236a510913 ]

commit fdd41ec21e15 ("devlink: Return right error code in case of errors
for region read") modified the region read code to report errors
properly in unexpected cases.

In the case where the start_offset and ret_offset match, it unilaterally
converted this into an error. This causes an issue for the "dump"
version of the command. In this case, the devlink region dump will
always report an invalid argument:

000000000000ffd0 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
000000000000ffe0 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
devlink answers: Invalid argument
000000000000fff0 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff

This occurs because the expected flow for the dump is to return 0 after
there is no further data.

The simplest fix would be to stop converting the error code to -EINVAL
if start_offset == ret_offset. However, avoid unnecessary work by
checking for when start_offset is larger than the region size and
returning 0 upfront.

Fixes: fdd41ec21e15 ("devlink: Return right error code in case of errors for region read")
Signed-off-by: Jacob Keller 
Acked-by: Jiri Pirko 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

devlink: correct misspelling of snapshot

2020-01-11T22:30:24+00:00

The function to obtain a unique snapshot id was mistakenly typo'd as
devlink_region_shapshot_id_get. Fix this typo by renaming the function
and all of its users.

Signed-off-by: Jacob Keller 
Acked-by: Jiri Pirko 
Signed-off-by: David S. Miller

devlink: Wait longer before warning about unset port type

2020-01-11T00:45:40+00:00

The commit cited below causes devlink to emit a warning if a type was
not set on a devlink port for longer than 30 seconds to "prevent
misbehavior of drivers". This proved to be problematic when
unregistering the backing netdev. The flow is always:

devlink_port_type_clear()	// schedules the warning
unregister_netdev()		// blocking
devlink_port_unregister()	// cancels the warning

The call to unregister_netdev() can block for long periods of time for
various reasons: RTNL lock is contended, large amounts of configuration
to unroll following dismantle of the netdev, etc. This results in
devlink emitting a warning despite the driver behaving correctly.

In emulated environments (of future hardware) which are usually very
slow, the warning can also be emitted during port creation as more than
30 seconds can pass between the time the devlink port is registered and
when its type is set.

In addition, syzbot has hit this warning [1] 1974 times since 07/11/19
without being able to produce a reproducer. Probably because
reproduction depends on the load or other bugs (e.g., RTNL not being
released).

To prevent bogus warnings, increase the timeout to 1 hour.

[1] https://syzkaller.appspot.com/bug?id=e99b59e9c024a666c9f7450dc162a4b74d09d9cb

Fixes: 136bf27fc0e9 ("devlink: add warning in case driver does not set port type")
Signed-off-by: Ido Schimmel 
Reported-by: syzbot+b0a18ed7b08b735d2f41@syzkaller.appspotmail.com
Reported-by: Alex Veber 
Tested-by: Alex Veber 
Acked-by: Jiri Pirko 
Signed-off-by: David S. Miller

Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

2019-11-17T05:51:42+00:00

Lots of overlapping changes and parallel additions, stuff
like that.

Signed-off-by: David S. Miller

Merge branch 'mlx5-next' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux

2019-11-13T22:24:58+00:00

1) New generic devlink param "enable_roce", for downstream devlink
   reload support

2) Do vport ACL configuration on per vport basis when
   enabling/disabling a vport. This enables to have vports enabled/disabled
   outside of eswitch config for future

3) Split the code for legacy vs offloads mode and make it clear

4) Tide up vport locking and workqueue usage

5) Fix metadata enablement for ECPF

6) Make explicit use of VF property to publish IB_DEVICE_VIRTUAL_FUNCTION

7) E-Switch and flow steering core low level support and refactoring for
   netfilter flowtables offload

Signed-off-by: Saeed Mahameed

devlink: Allow large formatted message of binary output

2019-11-12T19:25:44+00:00

Devlink supports pair output of name and value. When the value is
binary, it must be presented in an array. If the length of the binary
value exceeds fmsg limitation, break the value into chunks internally.

Signed-off-by: Aya Levin 
Acked-by: Jiri Pirko 
Signed-off-by: David S. Miller

devlink: Add method for time-stamp on reporter's dump

2019-11-12T00:04:21+00:00

When setting the dump's time-stamp, use ktime_get_real in addition to
jiffies. This simplifies the user space implementation and bypasses
some inconsistent behavior with translating jiffies to current time.
The time taken is transformed into nsec, to comply with y2038 issue.

Fixes: c8e1da0bf923 ("devlink: Add health report functionality")
Signed-off-by: Aya Levin 
Acked-by: Jiri Pirko 
Acked-by: Arnd Bergmann 
Signed-off-by: David S. Miller

devlink: Add new "enable_roce" generic device param

2019-11-11T20:15:29+00:00

New device parameter to enable/disable handling of RoCE traffic in the
device.

Signed-off-by: Michael Guralnik 
Acked-by: Jiri Pirko 
Reviewed-by: Maor Gottlieb 
Signed-off-by: Saeed Mahameed

devlink: disallow reload operation during device cleanup

2019-11-10T03:38:36+00:00

There is a race between driver code that does setup/cleanup of device
and devlink reload operation that in some drivers works with the same
code. Use after free could we easily obtained by running:

while true; do
        echo "0000:00:10.0" >/sys/bus/pci/drivers/mlxsw_spectrum2/bind
        devlink dev reload pci/0000:00:10.0 &
        echo "0000:00:10.0" >/sys/bus/pci/drivers/mlxsw_spectrum2/unbind
done

Fix this by enabling reload only after setup of device is complete and
disabling it at the beginning of the cleanup process.

Reported-by: Ido Schimmel 
Fixes: 2d8dc5bbf4e7 ("devlink: Add support for reload")
Signed-off-by: Jiri Pirko 
Signed-off-by: David S. Miller

devlink: disallow reload operation during device cleanup

2019-11-08T22:11:47+00:00

There is a race between driver code that does setup/cleanup of device
and devlink reload operation that in some drivers works with the same
code. Use after free could we easily obtained by running:

while true; do
        echo 10 > /sys/bus/netdevsim/new_device
        devlink dev reload netdevsim/netdevsim10 &
        echo 10 > /sys/bus/netdevsim/del_device
done

Fix this by enabling reload only after setup of device is complete and
disabling it at the beginning of the cleanup process.

Reported-by: Ido Schimmel 
Fixes: 2d8dc5bbf4e7 ("devlink: Add support for reload")
Signed-off-by: Jiri Pirko 
Acked-by: Jakub Kicinski 
Signed-off-by: David S. Miller