Jeff Kirsher says:
====================
Intel Wired LAN Driver Updates 2017-10-26
This series contains fixes to e1000, igb, ixgbe and i40e.
Vincenzo Maffione fixes a potential race condition which would result in
the interface being up but transmits are disabled in the hardware.
Colin Ian King fixes a possible NULL pointer dereference in e1000, which
was found by Coverity.
Jean-Philippe Brucker fixes a possible kernel panic when a driver cannot
map a transmit buffer, which is caused by an erroneous test.
Alex provides a fix for ixgbe, which is a partial revert of the commit
ffed21bcee ("ixgbe: Don't bother clearing buffer memory for descriptor rings")
because the previous commit messed up the exception handling path by
adding the count back in when we did not need to. Also fixed a typo,
where the transmit ITR setting was being used to determine if we were
using adaptive receive interrupt moderation or not. Lastly, fixed a
memory leak by including programming descriptors in the cleaned count.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
According to DWMAC databook the first queue operating mode
must always be in DCB.
As MTL_QUEUE_DCB = 1, we need to always set the first queue
operating mode to DCB otherwise driver will think that queue
is in AVB mode (because MTL_QUEUE_AVB = 0).
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
According to DT bindings documentation we are expecting a
property called "snps,read-requests" but we are parsing
instead a property called "read,read-requests".
This is clearly a typo. Fix it.
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Saeed Mahameed says:
====================
Mellanox, mlx5 fixes 2017-10-26
The series includes some misc fixes for mlx5 core and etherent driver.
Please pull and let me know if there's any problem.
For -Stable:
net/mlx5e: Properly deal with encap flows add/del under neigh update (kernels >= 4.12)
net/mlx5: Fix health work queue spin lock to IRQ safe (kernels >= 4.13)
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch updates the i40e driver to include programming descriptors in
the cleaned_count. Without this change it becomes possible for us to leak
memory as we don't trigger a large enough allocation when the time comes to
allocate new buffers and we end up overwriting a number of rx_buffers equal
to the number of programming descriptors we encountered.
Fixes: 0e626ff7cc ("i40e: Fix support for flow director programming status")
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Anders K. Pedersen <akp@cohaesio.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
It looks like there was either a copy/paste error or just a typo that
resulted in the Tx ITR setting being used to determine if we were using
adaptive Rx interrupt moderation or not.
This patch fixes the typo.
Fixes: 65e87c0398 ("i40evf: support queue-specific settings for interrupt moderation")
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch is a partial revert of "ixgbe: Don't bother clearing buffer
memory for descriptor rings". Specifically I messed up the exception
handling path a bit and this resulted in us incorrectly adding the count
back in when we didn't need to.
In order to make this simpler I am reverting most of the exception handling
path change and instead just replacing the bit that was handled by the
unmap_and_free call.
Fixes: ffed21bcee ("ixgbe: Don't bother clearing buffer memory for descriptor rings")
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When the driver cannot map a TX buffer, instead of rolling back
gracefully and retrying later, we currently get a panic:
[ 159.885994] igb 0000:00:00.0: TX DMA map failed
[ 159.886588] Unable to handle kernel paging request at virtual address ffff00000a08c7a8
...
[ 159.897031] PC is at igb_xmit_frame_ring+0x9c8/0xcb8
Fix the erroneous test that leads to this situation.
Signed-off-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Currently if the stat type is invalid then data[i] is being set
either by dereferencing a null pointer p, or it is reading from
an incorrect previous location if we had a valid stat type
previously. Fix this by skipping over the read of p on an invalid
stat type.
Detected by CoverityScan, CID#113385 ("Explicit null dereferenced")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch fixes a race condition that can result into the interface being
up and carrier on, but with transmits disabled in the hardware.
The bug may show up by repeatedly IFF_DOWN+IFF_UP the interface, which
allows e1000_watchdog() interleave with e1000_down().
CPU x CPU y
--------------------------------------------------------------------
e1000_down():
netif_carrier_off()
e1000_watchdog():
if (carrier == off) {
netif_carrier_on();
enable_hw_transmit();
}
disable_hw_transmit();
e1000_watchdog():
/* carrier on, do nothing */
Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch replaces GFP_KERNEL by GFP_ATOMIC to avoid sleeping in the
ndo_set_rx_mode() call which is called with BH disabled.
Fixes: 3f518509de ("ethernet: Add new driver for Marvell Armada 375 network unit")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When calling mvpp2_prs_mac_multi_set() from mvpp2_prs_mac_init(), two
parameters (the port index and the table index) are inverted. Fixes
this.
Fixes: 3f518509de ("ethernet: Add new driver for Marvell Armada 375 network unit")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a typo in the mvpp2_prs_tcam_data_cmp() function, as
the shift value is inverted with the data.
Fixes: 3f518509de ("ethernet: Add new driver for Marvell Armada 375 network unit")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previously, tc with ets type and zero bandwidth is not accepted
by driver. This behavior does not follow the IEEE802.1qaz spec.
If there are tcs with ets type and zero bandwidth, these tcs are
assigned to the lowest priority tc_group #0. We equally distribute
100% bw of the tc_group #0 to these zero bandwidth ets tcs.
Also, the non zero bandwidth ets tcs are assigned to tc_group #1.
If there is no zero bandwidth ets tc, the non zero bandwidth ets tcs
are assigned to tc_group #0.
Fixes: cdcf11212b ("net/mlx5e: Validate BW weight values of ETS")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Currently, the encap action offload is handled in the actions parse
function and not in mlx5e_tc_add_fdb_flow() where we deal with all
the other aspects of offloading actions (vlan, modify header) and
the rule itself.
When the neigh update code (mlx5e_tc_encap_flows_add()) recreates the
encap entry and offloads the related flows, we wrongly call again into
mlx5e_tc_add_fdb_flow(), this for itself would cause us to handle
again the offloading of vlans and header re-write which puts things
in non consistent state and step on freed memory (e.g the modify
header parse buffer which is already freed).
Since on error, mlx5e_tc_add_fdb_flow() detaches and may release the
encap entry, it causes a corruption at the neigh update code which goes
over the list of flows associated with this encap entry, or double free
when the tc flow is later deleted by user-space.
When neigh update (mlx5e_tc_encap_flows_del()) unoffloads the flows related
to an encap entry which is now invalid, we do a partial repeat of the eswitch
flow removal code which is wrong too.
To fix things up we do the following:
(1) handle the encap action offload in the eswitch flow add function
mlx5e_tc_add_fdb_flow() as done for the other actions and the rule itself.
(2) modify the neigh update code (mlx5e_tc_encap_flows_add/del) to only
deal with the encap entry and rules delete/add and not with any of
the other offloaded actions.
Fixes: 232c001398 ('net/mlx5e: Add support to neighbour update flow')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
mlx5_ib_add is called during mlx5_pci_resume after a pci error.
Before mlx5_ib_add completes, there are multiple events which trigger
function mlx5_ib_event. This cause kernel panic because mlx5_ib_event
accesses unitialized resources.
The fix is to extend Erez Shitrit's patch <97834eba7c19>
("net/mlx5: Delay events till ib registration ends") to cover
the pci resume code path.
Trace:
mlx5_core 0001:01:00.6: mlx5_pci_resume was called
mlx5_core 0001:01:00.6: firmware version: 16.20.1011
mlx5_core 0001:01:00.6: mlx5_attach_interface:164:(pid 779):
mlx5_ib_event:2996:(pid 34777): warning: event on port 1
mlx5_ib_event:2996:(pid 34782): warning: event on port 1
Unable to handle kernel paging request for data at address 0x0001c104
Faulting instruction address: 0xd000000008f411fc
Oops: Kernel access of bad area, sig: 11 [#1]
...
...
Call Trace:
[c000000fff77bb70] [d000000008f4119c] mlx5_ib_event+0x64/0x470 [mlx5_ib] (unreliable)
[c000000fff77bc60] [d000000008e67130] mlx5_core_event+0xb8/0x210 [mlx5_core]
[c000000fff77bd10] [d000000008e4bd00] mlx5_eq_int+0x528/0x860[mlx5_core]
Fixes: 97834eba7c ("net/mlx5: Delay events till ib registration ends")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
spin_lock/unlock of health->wq_lock should be IRQ safe.
It was changed to spin_lock_irqsave since adding commit 0179720d6b
("net/mlx5: Introduce trigger_health_work function") which uses
spin_lock from asynchronous event (IRQ) context.
Thus, all spin_lock/unlock of health->wq_lock should have been moved
to IRQ safe mode.
However, one occurrence on new code using this lock missed that
change, resulting in possible deadlock:
kernel: Possible unsafe locking scenario:
kernel: CPU0
kernel: ----
kernel: lock(&(&health->wq_lock)->rlock);
kernel: <Interrupt>
kernel: lock(&(&health->wq_lock)->rlock);
kernel: #012 *** DEADLOCK ***
Fixes: 2a0165a034 ("net/mlx5: Cancel delayed recovery work when unloading the driver")
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Previously we did not ensure that a netdev is a representative netdev
before dereferencing its private data. This can occur when an upper netdev
is created on a representative netdev. This patch corrects this by first
ensuring that the netdev is a representative netdev before using it.
Checking only switchdev_port_same_parent_id is not sufficient to ensure
that we can safely use the netdev. Failing to check that the netdev is also
a representative netdev would result in incorrect dereferencing.
Fixes: 1a1e586f54 ("nfp: add basic action capabilities to flower offloads")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When Tx IRQs are used, txq_bufs_free() can be called from both the Tx
path and from NAPI poll(). This led to CPU stalls as if these two tasks
(Tx and Poll) are scheduled on two CPUs at the same time, DMA unmapping
operations are done on the same txq buffers.
This patch adds a check not to call txq_done() from the Tx path if Tx
interrupts are used as it does not make sense to do so.
Fixes: edc660fa09 ("net: mvpp2: replace TX coalescing interrupts with hrtimer")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The TSO header buffers are coming from a per cpu pool and should not
be unmapped as they are reused. The PPv2 driver was unmapping all
descriptors buffers unconditionally. This patch fixes this by checking
the buffers dma addresses before unmapping them, and by not unmapping
those who are located in the TSO header pool.
Fixes: 186cd4d4e4 ("net: mvpp2: software tso support")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TSO headers are managed with txq index and therefore should be aligned
with the txq size, not with the aggregated txq size.
Fixes: 186cd4d4e4 ("net: mvpp2: software tso support")
Reported-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Yan Markman <ymarkman@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the possible reason for different hard to reproduce
problems on my ARMv7-SMP test system.
The symptoms are in recent kernels imprecise external aborts,
and in older kernels various kinds of network stalls and
unexpected page allocation failures.
My testing indicates that the trouble started between v4.5 and v4.6
and prevails up to v4.14.
Using the dirty_tx before acquiring the spin lock is clearly
wrong and was first introduced with v4.6.
Fixes: e3ad57c967 ("stmmac: review RX/TX ring management")
Signed-off-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prevent infinite loop by correctly setting the loop condition to
break when i == 10.
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When using GMAC4 the valid timestamp is from CTX next desc but
we are passing the previous desc to get_rx_timestamp_status()
callback.
Fix this and while at it rework a little bit the function logic.
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When RX HW timestamp is enabled and a frame is discarded we are
not freeing the skb but instead only setting to NULL the entry.
Add a call to dev_kfree_skb_any() so that skb entry is correctly
freed.
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Spectrum tunnels do not default to ttl of "inherit" like the Linux ones
do. Configure TIGCR on router init so that the TTL of tunnel packets is
copied from the overlay packets.
Fixes: ee954d1a91 ("mlxsw: spectrum_router: Support GRE tunnels")
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The TIGCR register is used for setting up the IPinIP Tunnel
configuration.
Fixes: ee954d1a91 ("mlxsw: spectrum_router: Support GRE tunnels")
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Default Tx rates cause very long ISR delays on Tx.
0xff is 510us delay, giving only ~ 2000 interrupts per seconds for
Tx rings cleanup. With these settings udp tx rate was never higher than
~800Mbps on a single stream. Changing min delay to 0xF makes it
way better with ~6Gbps
TCP stream performance is almost unaffected by this change, since LSO
optimizations play important role.
CPU load is affected insignificantly by this change.
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Aquantia NIC allows both TX and RX interrupt throttle rate (ITR)
management, but this was used in a very limited way via predefined
values. This patch allows to setup ITR default values via module
command line arguments and via standard ethtool coalescing settings.
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
That may lead to mmio resource leakage.
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no much practical use from having MSIX vectors more that number
of cpus, thus cap this first with preconfigured limit, then with number
of cpus online.
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When doing ifconfig down/up, driver did not reported carrier_off neither
in nic_stop nor in nic_start. That caused link to be visible as "up"
during couple of seconds immediately after "ifconfig up".
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Queue stat strings are cleaned up, duplicate stat name strings removed,
queue restarts counter added
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Internal statistics system on chip never gets reset until hardware
reboot. This is quite inconvenient in terms of ethtool statistics usage.
This patch implements incremental statistics update inside of
service callback.
Upon nic initialization, first request is done to fetch
initial stat data, current collected stat data gets cleared.
Internal statistics mailbox readout is improved to save space and
increase readability
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ethtool ena_get_channels() expose the max number of queues as the max
number of queues ENA supports (128 queues) and not the actual number
of created queues.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Decrease log level of checksum errors as these messages can be
triggered remotely by bad packets.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch correctly sets the number of additional header descriptors
that will be sent in an indirect SCRQ entry.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When an EMAD is transmitted, a timeout work item is scheduled with a
delay of 200ms, so that another EMAD will be retried until a maximum of
five retries.
In certain situations, it's possible for the function waiting on the
EMAD to be associated with a work item that is queued on the same
workqueue (`mlxsw_core`) as the timeout work item. This results in
flushing a work item on the same workqueue.
According to commit e159489baa ("workqueue: relax lockdep annotation
on flush_work()") the above may lead to a deadlock in case the workqueue
has only one worker active or if the system in under memory pressure and
the rescue worker is in use. The latter explains the very rare and
random nature of the lockdep splats we have been seeing:
[ 52.730240] ============================================
[ 52.736179] WARNING: possible recursive locking detected
[ 52.742119] 4.14.0-rc3jiri+ #4 Not tainted
[ 52.746697] --------------------------------------------
[ 52.752635] kworker/1:3/599 is trying to acquire lock:
[ 52.758378] (mlxsw_core_driver_name){+.+.}, at: [<ffffffff811c4fa4>] flush_work+0x3a4/0x5e0
[ 52.767837]
but task is already holding lock:
[ 52.774360] (mlxsw_core_driver_name){+.+.}, at: [<ffffffff811c65c4>] process_one_work+0x7d4/0x12f0
[ 52.784495]
other info that might help us debug this:
[ 52.791794] Possible unsafe locking scenario:
[ 52.798413] CPU0
[ 52.801144] ----
[ 52.803875] lock(mlxsw_core_driver_name);
[ 52.808556] lock(mlxsw_core_driver_name);
[ 52.813236]
*** DEADLOCK ***
[ 52.819857] May be due to missing lock nesting notation
[ 52.827450] 3 locks held by kworker/1:3/599:
[ 52.832221] #0: (mlxsw_core_driver_name){+.+.}, at: [<ffffffff811c65c4>] process_one_work+0x7d4/0x12f0
[ 52.842846] #1: ((&(&bridge->fdb_notify.dw)->work)){+.+.}, at: [<ffffffff811c65c4>] process_one_work+0x7d4/0x12f0
[ 52.854537] #2: (rtnl_mutex){+.+.}, at: [<ffffffff822ad8e7>] rtnl_lock+0x17/0x20
[ 52.863021]
stack backtrace:
[ 52.867890] CPU: 1 PID: 599 Comm: kworker/1:3 Not tainted 4.14.0-rc3jiri+ #4
[ 52.875773] Hardware name: Mellanox Technologies Ltd. "MSN2100-CB2F"/"SA001017", BIOS 5.6.5 06/07/2016
[ 52.886267] Workqueue: mlxsw_core mlxsw_sp_fdb_notify_work [mlxsw_spectrum]
[ 52.894060] Call Trace:
[ 52.909122] __lock_acquire+0xf6f/0x2a10
[ 53.025412] lock_acquire+0x158/0x440
[ 53.047557] flush_work+0x3c4/0x5e0
[ 53.087571] __cancel_work_timer+0x3ca/0x5e0
[ 53.177051] cancel_delayed_work_sync+0x13/0x20
[ 53.182142] mlxsw_reg_trans_bulk_wait+0x12d/0x7a0 [mlxsw_core]
[ 53.194571] mlxsw_core_reg_access+0x586/0x990 [mlxsw_core]
[ 53.225365] mlxsw_reg_query+0x10/0x20 [mlxsw_core]
[ 53.230882] mlxsw_sp_fdb_notify_work+0x2a3/0x9d0 [mlxsw_spectrum]
[ 53.237801] process_one_work+0x8f1/0x12f0
[ 53.321804] worker_thread+0x1fd/0x10c0
[ 53.435158] kthread+0x28e/0x370
[ 53.448703] ret_from_fork+0x2a/0x40
[ 53.453017] mlxsw_spectrum 0000:01:00.0: EMAD retries (2/5) (tid=bf4549b100000774)
[ 53.453119] mlxsw_spectrum 0000:01:00.0: EMAD retries (5/5) (tid=bf4549b100000770)
[ 53.453132] mlxsw_spectrum 0000:01:00.0: EMAD reg access failed (tid=bf4549b100000770,reg_id=200b(sfn),type=query,status=0(operation performed))
[ 53.453143] mlxsw_spectrum 0000:01:00.0: Failed to get FDB notifications
Fix this by creating another workqueue for EMAD timeouts, thereby
preventing the situation of a work item trying to flush a work item
queued on the same workqueue.
Fixes: caf7297e7a ("mlxsw: core: Introduce support for asynchronous EMAD register access")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reported-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
hwrm_send_message() is replaced with _hwrm_send_message(), and
hwrm_cmd_lock mutex lock is grabbed for the whole period of
firmware call until the firmware DCB parameters have been copied.
This will prevent possible corruption of the firmware data.
Fixes: 7df4ae9fe8 ("bnxt_en: Implement DCBNL to support host-based DCBX.")
Signed-off-by: Sankar Patchineelam <sankar.patchineelam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In bnxt_find_nvram_item(), it is copying firmware response data after
releasing the mutex. This can cause the firmware response data
to be corrupted if the next firmware response overwrites the response
buffer. The rare problem shows up when running ethtool -i repeatedly.
Fix it by calling the new variant _hwrm_send_message_silent() that requires
the caller to take the mutex and to release it after the response data has
been copied.
Fixes: 3ebf6f0a09 ("bnxt_en: Add installed-package version reporting via Ethtool GDRVINFO")
Reported-by: Sarveswara Rao Mygapula <sarveswararao.mygapula@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In bnxt_sriov_enable(), we calculate to see if we have enough hardware
resources to enable the requested number of VFs. The logic to check
for minimum completion rings and statistics contexts is missing. Add
the required checks so that VF configuration won't fail.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
PCIE PCIE_EP_REG_LINK_STATUS_CONTROL register is only defined in PF
config space, so we must read it from the PF.
Fixes: 90c4f788f6 ("bnxt_en: Report PCIe link speed and width during driver load")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As a further improvement to the PF/VF link change logic, use a private
mutex instead of the rtnl lock to protect link change logic. With the
new mutex, we don't have to take the rtnl lock in the workqueue when
we have to handle link related functions. If the VF and PF drivers
are running on the same host and both take the rtnl lock and one is
waiting for the other, it will cause timeout. This patch fixes these
timeouts.
Fixes: 90c694bb71 ("bnxt_en: Fix RTNL lock usage on bnxt_update_link().")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Link status query firmware messages originating from the VFs are forwarded
to the PF. The driver handles these interactions in a workqueue for the
VF and PF. The VF driver waits for the response from the PF in the
workqueue. If the PF and VF driver are running on the same host and the
work for both PF and VF are queued on the same workqueue, the VF driver
may not get the response if the PF work item is queued behind it on the
same workqueue. This will lead to the VF link query message timing out.
To prevent this, we create a private workqueue for PFs instead of using
the common workqueue. The VF query and PF response will never be on
the same workqueue.
Fixes: c0c050c58d ("bnxt_en: New Broadcom ethernet driver.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The DMA reset timeout, used in read_poll_timeout, is
ten times shorter than the sleep time.
This patch fixes these values interchanging them, as it was
before the read_poll_timeout introduction.
Fixes: 8a70aeca80 ("net: stmmac: Use readl_poll_timeout")
Signed-off-by: Emiliano Ingrassia <ingrassia@epigenesys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While experimenting with changes to the timekeeping code, I
ran into a build error in the liquidio driver:
drivers/net/ethernet/cavium/liquidio/lio_main.c: In function 'liquidio_ptp_settime':
drivers/net/ethernet/cavium/liquidio/lio_main.c:1850:22: error: passing argument 1 of 'timespec_to_ns' from incompatible pointer type [-Werror=incompatible-pointer-types]
The driver had a type mismatch since it was first merged, but
this never caused problems because it is only built on 64-bit
architectures that define timespec and timespec64 to the same
type.
If we ever want to compile-test the driver on 32-bit or change
the way that 64-bit timespec64 is defined, we need to fix it,
so let's just do it now.
Fixes: f21fb3ed36 ("Add support of Cavium Liquidio ethernet adapters")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rtl_init_one() currently enables PCI wakeups if the ethernet device
is found to be WOL-capable. There is no need to do this when
rtl8169_set_wol() will correctly enable or disable the same wakeup flag
when WOL is activated/deactivated.
This works around an ACPI DSDT bug which prevents the Acer laptop models
Aspire ES1-533, Aspire ES1-732, PackardBell ENTE69AP and Gateway NE533
from entering S3 suspend - even when no ethernet cable is connected.
On these platforms, the DSDT says that GPE08 is a wakeup source for
ethernet, but this GPE fires as soon as the system goes into suspend,
waking the system up immediately. Having the wakeup normally disabled
avoids this issue in the default case.
With this change, WOL will continue to be unusable on these platforms
(it will instantly wake up if WOL is later enabled by the user) but we
do not expect this to be a commonly used feature on these consumer
laptops. We have separately determined that WOL works fine without any
ACPI GPEs enabled during sleep, so a DSDT fix or override would be
possible to make WOL work.
Signed-off-by: Daniel Drake <drake@endlessm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
page_address() does not handle NULL argument gracefully,
make sure we NULL-check the page pointer before passing it
to page_address().
Fixes: ecd63a0217 ("nfp: add XDP support in the driver")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The while loop fetching 64 bit ethtool statistics may have
to retry multiple times, it shouldn't modify the outside state.
Fixes: 4c3523623d ("net: add driver for Netronome NFP4000/NFP6000 NIC VFs")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller