Age | Commit message (Collapse) | Author |
|
[ Upstream commit b2663a4f30e85ec606b806f5135413e6d5c78d1e ]
When the IP stack passes SKBs the sfc driver puts them in 2 different TX
queues (called partners), one for checksummed and one for not checksummed.
If the SKB has xmit_more set the driver will delay pushing the work to the
NIC.
When later it does decide to push the buffers this patch ensures it also
pushes the partner queue, if that also has any delayed work. Before this
fix the work in the partner queue would be left for a long time and cause
a netdev watchdog.
Fixes: 70b33fb ("sfc: add support for skb->xmit_more")
Reported-by: Jianlin Shi <jishi@redhat.com>
Signed-off-by: Martin Habets <mhabets@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit e6dbe1eb2db0d7a14991c06278dd3030c45fb825 ]
priv->hwts_*_en indicate if timestamping is enabled/disabled at run
time. But priv->dma_cap.time_stamp and priv->dma_cap.atime_stamp
indicates HW is support for PTPv1/PTPv2.
Signed-off-by: Phil Reid <preid@electromag.com.au>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit c02b05011fadf8e409e41910217ca689f2fc9d91 ]
When doing memcpy/memset of EQEs, we should use sizeof struct
mlx4_eqe as the base size and not caps.eqe_size which could be bigger.
If caps.eqe_size is bigger than the struct mlx4_eqe then we corrupt
data in the master context.
When using a 64 byte stride, the memcpy copied over 63 bytes to the
slave_eq structure. This resulted in copying over the entire eqe of
interest, including its ownership bit -- and also 31 bytes of garbage
into the next WQE in the slave EQ -- which did NOT include the ownership
bit (and therefore had no impact).
However, once the stride is increased to 128, we are overwriting the
ownership bits of *three* eqes in the slave_eq struct. This results
in an incorrect ownership bit for those eqes, which causes the eq to
seem to be full. The issue therefore surfaced only once 128-byte EQEs
started being used in SRIOV and (overarchitectures that have 128/256
byte cache-lines such as PPC) - e.g after commit 77507aa249ae
"net/mlx4_core: Enable CQE/EQE stride support".
Fixes: 08ff32352d6f ('mlx4: 64-byte CQE/EQE support')
Signed-off-by: Carol L Soto <clsoto@linux.vnet.ibm.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 1d20a16062e771b6e26b843c0cde3b17c1146e00 ]
drivers/net/ethernet/sfc/selftest.c: In function ‘efx_iterate_state’:
drivers/net/ethernet/sfc/selftest.c:388:9: warning: passing argument 1 of ‘memcpy’ discards ‘const’ qualifier from pointer target type [-Wdiscarded-array-qualifiers]
This is because the msg[] member of struct efx_loopback_payload
is marked as 'const'. Remove that.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 6423fc34160939142d72ffeaa2db6408317f54df ]
During driver probing the following code path is triggered.
igb_probe
->igb_sw_init
->igb_probe_vfs
->igb_pci_enable_sriov
->igb_sriov_reinit
Doing the SR-IOV re-init is not necessary during probing since we're
starting from scratch. Here we can call igb_enable_sriov() right away.
Running igb_sriov_reinit() during igb_probe() also seems to cause
occasional packet loss on some onboard 82576 NICs. Reproduced on
Dell and HP servers with onboard 82576 NICs.
Example:
Intel Corporation 82576 Gigabit Network Connection [8086:10c9] (rev 01)
Subsystem: Dell Device [1028:0481]
Signed-off-by: Stefan Assmann <sassmann@kpanic.de>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 4548a697e4969d695047cebd6d9af5e2f6cc728e ]
tse_poll() calls __napi_complete() with irq enabled. This leads napi
poll_list corruption and may stop all napi drivers working.
Use napi_complete() instead of __napi_complete().
Signed-off-by: Atsushi Nemoto <nemoto@toshiba-tops.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 72ddef0506da852dc82f078f37ced8ef4d74a2bf ]
When initializing igb driver (e.g. 82576, I350), IGB_FLAG_QUEUE_PAIRS is
set if adapter->rss_queues exceeds half of max_rss_queues in
igb_init_queue_configuration().
On the other hand, IGB_FLAG_QUEUE_PAIRS is not set even if the number of
queues exceeds half of max_combined in igb_set_channels() when changing
the number of queues by "ethtool -L".
In this case, if numvecs is larger than MAX_MSIX_ENTRIES (10), the size
of adapter->msix_entries[], an overflow can occur in
igb_set_interrupt_capability(), which in turn leads to an oops.
Fix this problem as follows:
- When changing the number of queues by "ethtool -L", set
IGB_FLAG_QUEUE_PAIRS in the same way as initializing igb driver.
- When increasing the size of q_vector, reallocate it appropriately.
(With IGB_FLAG_QUEUE_PAIRS set, the size of q_vector gets larger.)
Another possible way to fix this problem is to cap the queues at its
initial number, which is the number of the initial online cpus. But this
is not the optimal way because we cannot increase queues when another
cpu becomes online.
Note that before commit cd14ef54d25b ("igb: Change to use statically
allocated array for MSIx entries"), this problem did not cause oops
but just made the number of queues become 1 because of entering msi_only
mode in igb_set_interrupt_capability().
Fixes: 907b7835799f ("igb: Add ethtool support to configure number of channels")
CC: stable <stable@vger.kernel.org>
Signed-off-by: Shota Suzuki <suzuki_shota_t3@lab.ntt.co.jp>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit ade4dc3e616e33c80d7e62855fe1b6f9895bc7c3 ]
The commit "e29aa33 bna: Enable Multi Buffer RX" moved packets counter
increment from the beginning of the NAPI processing loop after the check
for erroneous packets so they are never accounted. This counter is used
to inform firmware about number of processed completions (packets).
As these packets are never acked the firmware fires IRQs for them again
and again.
Fixes: e29aa33 ("bna: Enable Multi Buffer RX")
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Acked-by: Rasesh Mody <rasesh.mody@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit d3d11fe08ccc9bff174fc958722b5661f0932486 ]
The temperature registers appear to report values in degrees Celsius
while the hwmon API mandates values to be exposed in millidegrees
Celsius. Do the conversion so that the values reported by "sensors"
are correct.
Fixes: aed93e0bf493 ("tg3: Add hwmon support for temperature")
Signed-off-by: Jean Delvare <jdelvare@suse.de>
Cc: Prashant Sreedharan <prashant@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Cc: stable@vger.kernel.org [v3.6+]
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 1c1bf34951e8d17941bf708d1901c47e81b15d55 ]
The port-change event processing in procedure mlx4_eq_int() uses "slave"
as the vf_oper array index. Since the value of "slave" is the PF function
index, the result is that the PF link state is used for deciding to
propagate the event for all the VFs. The VF link state should be used,
so the VF function index should be used here.
Fixes: 948e306d7d64 ('net/mlx4: Add VF link state support')
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit d53c66a5b80698620f7c9ba2372fff4017e987b8 ]
Michel reported following lockdep splat
[ 44.718117] INFO: trying to register non-static key.
[ 44.723081] the code is fine but needs lockdep annotation.
[ 44.728559] turning off the locking correctness validator.
[ 44.734036] CPU: 8 PID: 5483 Comm: ethtool Not tainted 4.1.0
[ 44.770289] Call Trace:
[ 44.772741] [<ffffffff816eb1cd>] dump_stack+0x4c/0x65
[ 44.777879] [<ffffffff8111d921>] ? console_unlock+0x1f1/0x510
[ 44.783708] [<ffffffff811121f5>] __lock_acquire+0x1d05/0x1f10
[ 44.789538] [<ffffffff8111370a>] ? mark_held_locks+0x6a/0x90
[ 44.795276] [<ffffffff81113835>] ? trace_hardirqs_on_caller+0x105/0x1d0
[ 44.801967] [<ffffffff8111390d>] ? trace_hardirqs_on+0xd/0x10
[ 44.807793] [<ffffffff811330fa>] ? hrtimer_try_to_cancel+0x4a/0x250
[ 44.814142] [<ffffffff81112ba6>] lock_acquire+0xb6/0x290
[ 44.819537] [<ffffffff810d6675>] ? flush_work+0x5/0x280
[ 44.824844] [<ffffffff810d66ad>] flush_work+0x3d/0x280
[ 44.830061] [<ffffffff810d6675>] ? flush_work+0x5/0x280
[ 44.835366] [<ffffffff816f3c43>] ? schedule_hrtimeout_range+0x13/0x20
[ 44.841889] [<ffffffff8112ec9b>] ? usleep_range+0x4b/0x50
[ 44.847365] [<ffffffff8111370a>] ? mark_held_locks+0x6a/0x90
[ 44.853102] [<ffffffff810d8585>] ? __cancel_work_timer+0x105/0x1c0
[ 44.859359] [<ffffffff81113835>] ? trace_hardirqs_on_caller+0x105/0x1d0
[ 44.866045] [<ffffffff810d851f>] __cancel_work_timer+0x9f/0x1c0
[ 44.872048] [<ffffffffa0010982>] ? bnx2x_func_stop+0x42/0x90 [bnx2x]
[ 44.878481] [<ffffffff810d8670>] cancel_work_sync+0x10/0x20
[ 44.884134] [<ffffffffa00259e5>] bnx2x_chip_cleanup+0x245/0x730 [bnx2x]
[ 44.890829] [<ffffffff8110ce02>] ? up+0x32/0x50
[ 44.895439] [<ffffffff811306b5>] ? del_timer_sync+0x5/0xd0
[ 44.901005] [<ffffffffa005596d>] bnx2x_nic_unload+0x20d/0x8e0 [bnx2x]
[ 44.907527] [<ffffffff811f1aef>] ? might_fault+0x5f/0xb0
[ 44.912921] [<ffffffffa005851c>] bnx2x_reload_if_running+0x2c/0x50 [bnx2x]
[ 44.919879] [<ffffffffa005a3c5>] bnx2x_set_ringparam+0x2b5/0x460 [bnx2x]
[ 44.926664] [<ffffffff815d498b>] dev_ethtool+0x55b/0x1c40
[ 44.932148] [<ffffffff815dfdc7>] ? rtnl_lock+0x17/0x20
[ 44.937364] [<ffffffff815e7f8b>] dev_ioctl+0x17b/0x630
[ 44.942582] [<ffffffff815abf8d>] sock_do_ioctl+0x5d/0x70
[ 44.947972] [<ffffffff815ac013>] sock_ioctl+0x73/0x280
[ 44.953192] [<ffffffff8124c1c8>] do_vfs_ioctl+0x88/0x5b0
[ 44.958587] [<ffffffff8110d0b3>] ? up_read+0x23/0x40
[ 44.963631] [<ffffffff812584cc>] ? __fget_light+0x6c/0xa0
[ 44.969105] [<ffffffff8124c781>] SyS_ioctl+0x91/0xb0
[ 44.974149] [<ffffffff816f4dd7>] system_call_fastpath+0x12/0x6f
As bnx2x_init_ptp() is only called if bp->flags contains PTP_SUPPORTED,
we also need to guard bnx2x_stop_ptp() with same condition, otherwise
ptp_task workqueue is not initialized and kernel barfs on
cancel_work_sync()
Fixes: eeed018cbfa30 ("bnx2x: Add timestamping and PTP hardware clock support")
Reported-by: Michel Lespinasse <walken@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Michal Kalderon <Michal.Kalderon@qlogic.com>
Cc: Ariel Elior <Ariel.Elior@qlogic.com>
Cc: Yuval Mintz <Yuval.Mintz@qlogic.com>
Cc: David Decotigny <decot@google.com>
Acked-by: Sony Chacko <sony.chacko@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 488a9b48e398b157703766e2cd91ea45ac6997c5 ]
Indication of a single completed packet, marked by txbbs_skipped
being bigger then zero, in not enough in order to wake up a
stopped TX queue. The completed packet may contain a single TXBB,
while next packet to be sent (after the wake up) may have multiple
TXBBs (LSO/TSO packets for example), causing overflow in queue followed
by WQE corruption and TX queue timeout.
Instead, wake the stopped queue only when there's enough room for the
worst case (maximum sized WQE) packet that we should need to handle after
the queue is opened again.
Also created an helper routine - mlx4_en_is_tx_ring_full, which checks
if the current TX ring is full or not. It provides better code readability
and removes code duplication.
Signed-off-by: Ido Shamay <idos@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit b65657fc240ae6c1d2a1e62db9a0e61ac9631d7a ]
The Ethernet controller found in the Armada 370, 380 and 385 SoCs don't
support TCP/IP checksumming with frame sizes larger than 1600 bytes.
This patch fixes the issue by disabling the features NETIF_F_IP_CSUM and
NETIF_F_TSO for the Armada 370 and compatibles SoCs when the MTU is set
to a value greater than 1600 bytes.
Signed-off-by: Simon Guinot <simon.guinot@sequanux.org>
Fixes: c5aff18204da ("net: mvneta: driver for Marvell Armada 370/XP network unit")
Cc: <stable@vger.kernel.org> # v3.8+
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit f522a975a8101895a85354b9c143f41b8248e71a ]
The mvneta driver supports the Ethernet IP found in the Armada 370, XP,
380 and 385 SoCs. Since at least one more hardware feature is available
for the Armada XP SoCs then a way to identify them is needed.
This patch introduces a new compatible string "marvell,armada-xp-neta".
Signed-off-by: Simon Guinot <simon.guinot@sequanux.org>
Fixes: c5aff18204da ("net: mvneta: driver for Marvell Armada 370/XP network unit")
Cc: <stable@vger.kernel.org> # v3.8+
Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit f1590670ce069eefeb93916391a67643e6ad1630 ]
Current implementation of descriptor init procedure only takes
care about setting/clearing ownership flag in "des0"/"des1"
fields while it is perfectly possible to get unexpected bits
set because of the following factors:
[1] On driver probe underlying memory allocated with
dma_alloc_coherent() might not be zeroed and so
it will be filled with garbage.
[2] During driver operation some bits could be set by SD/MMC
controller (for example error flags etc).
And unexpected and/or randomly set flags in "des0"/"des1"
fields may lead to unpredictable behavior of GMAC DMA block.
This change addresses both items above with:
[1] Use of dma_zalloc_coherent() instead of simple
dma_alloc_coherent() to make sure allocated memory is
zeroed. That shouldn't affect performance because
this allocation only happens once on driver probe.
[2] Do explicit zeroing of both "des0" and "des1" fields
of all buffer descriptors during initialization of
DMA transfer.
And while at it fixed identation of dma_free_coherent()
counterpart as well.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: arc-linux-dev@synopsys.com
Cc: linux-kernel@vger.kernel.org
Cc: stable@vger.kernel.org
Cc: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
architectures
[ Upstream commit 59d2d18cc4e9ba30b370db18d0e02d792699da96 ]
If CONFIG_ARCH_DMA_ADDR_T_64BIT enabled for x86 systems and physical
memory is more than 4GB, dma_map_page may return a valid memory
address which greater than 0xffffffff. As a result, the mlx5 device page
allocator RB tree will be initialized with valid addresses greater than
0xfffffff.
However, (addr & PAGE_MASK) set the high four bytes to zeros. So, it's
impossible for the function, free_4k, to release the pages whose
addresses greater than 4GB. Memory leaks. And mlx5_ib module can't
release the pages when user try to remove the module, as a result,
system hang.
[root@rdma05 root]# dmesg | grep addr | head
addr = 3fe384000
addr & PAGE_MASK = fe384000
[root@rdma05 root]# rmmod mlx5_ib <---- hang on
---------------------- cosnole log -----------------
mlx5_ib 0000:04:00.0: irq 138 for MSI/MSI-X
alloc irq_desc for 139 on node -1
alloc kstat_irqs on node -1
mlx5_ib 0000:04:00.0: irq 139 for MSI/MSI-X
0000:04:00.0:free_4k:221:(pid 1519): page not found
0000:04:00.0:free_4k:221:(pid 1519): page not found
0000:04:00.0:free_4k:221:(pid 1519): page not found
0000:04:00.0:free_4k:221:(pid 1519): page not found
---------------------- cosnole log -----------------
Fixes: bf0bf77f6519 ('mlx5: Support communicating arbitrary host page size to firmware')
Signed-off-by: Honggang Li <honli@redhat.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit e51000db4c880165eab06ec0990605f24e75203f ]
There are several places in the driver (all in control paths) where
coherent dma memory is being allocated using either dma_alloc_coherent()
or the deprecated pci_alloc_consistent(). All these calls should be
changed to use dma_zalloc_coherent() to avoid uninitialized fields in
data structures backed by this memory.
Reported-by: Joerg Roedel <jroedel@suse.de>
Tested-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@avagotech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 08e8331654d1d7b2c58045e549005bc356aa7810 ]
There is a race condition between e1000_change_mtu's cleanups and
netpoll, when we change the MTU across jumbo size:
Changing MTU frees all the rx buffers:
e1000_change_mtu -> e1000_down -> e1000_clean_all_rx_rings ->
e1000_clean_rx_ring
Then, close to the end of e1000_change_mtu:
pr_info -> ... -> netpoll_poll_dev -> e1000_clean ->
e1000_clean_rx_irq -> e1000_alloc_rx_buffers -> e1000_alloc_frag
And when we come back to do the rest of the MTU change:
e1000_up -> e1000_configure -> e1000_configure_rx ->
e1000_alloc_jumbo_rx_buffers
alloc_jumbo finds the buffers already != NULL, since data (shared with
page in e1000_rx_buffer->rxbuf) has been re-alloc'd, but it's garbage,
or at least not what is expected when in jumbo state.
This results in an unusable adapter (packets don't get through), and a
NULL pointer dereference on the next call to e1000_clean_rx_ring
(other mtu change, link down, shutdown):
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [<ffffffff81194d6e>] put_compound_page+0x7e/0x330
[...]
Call Trace:
[<ffffffff81195445>] put_page+0x55/0x60
[<ffffffff815d9f44>] e1000_clean_rx_ring+0x134/0x200
[<ffffffff815da055>] e1000_clean_all_rx_rings+0x45/0x60
[<ffffffff815df5e0>] e1000_down+0x1c0/0x1d0
[<ffffffff811e2260>] ? deactivate_slab+0x7f0/0x840
[<ffffffff815e21bc>] e1000_change_mtu+0xdc/0x170
[<ffffffff81647050>] dev_set_mtu+0xa0/0x140
[<ffffffff81664218>] do_setlink+0x218/0xac0
[<ffffffff814459e9>] ? nla_parse+0xb9/0x120
[<ffffffff816652d0>] rtnl_newlink+0x6d0/0x890
[<ffffffff8104f000>] ? kvm_clock_read+0x20/0x40
[<ffffffff810a2068>] ? sched_clock_cpu+0xa8/0x100
[<ffffffff81663802>] rtnetlink_rcv_msg+0x92/0x260
By setting the allocator to a dummy version, netpoll can't mess up our
rx buffers. The allocator is set back to a sane value in
e1000_configure_rx.
Fixes: edbbb3ca1077 ("e1000: implement jumbo receive with partial descriptors")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 07841f9d94c11afe00c0498cf242edf4075729f4 ]
When system is out of memory, refilling of RX buffers fails while
the driver continue to pass the received packets to the kernel stack.
At some point, when all RX buffers deplete, driver may fall into a
sleep, and not recover when memory for new RX buffers is once again
availible. This is because hardware does not have valid descriptors,
so no interrupt will be generated for the driver to return to work
in napi context. Fix it by schedule the napi poll function from
stats_task delayed workqueue, as long as the allocations fail.
Signed-off-by: Ido Shamay <idos@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 42eab005a5dd5d7ea2b0328aecc4d6cc0c23c9c2 ]
By default, the number of tx queues is limited by the number of online cpus
in mlx4_en_get_profile(). However, this limit no longer holds after the
ethtool .set_channels method has been called. In that situation, the driver
may access invalid bits of certain cpumask variables when queue_index >=
nr_cpu_ids.
Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Acked-by: Ido Shamay <idos@mellanox.com>
Fixes: d03a68f ("net/mlx4_en: Configure the XPS queue mapping on driver load")
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit d0af71a3573f1217b140c60b66f1a9b335fb058b ]
tg3_init_one() calls tg3_halt() without tp->lock despite its assumption
and causes deadlock.
If lockdep is enabled, a warning like this shows up before the stall:
[ BUG: bad unlock balance detected! ]
3.19.0test #3 Tainted: G E
-------------------------------------
insmod/369 is trying to release lock (&(&tp->lock)->rlock) at:
[<ffffffffa02d5a1d>] tg3_chip_reset+0x14d/0x780 [tg3]
but there are no more locks to release!
tg3_init_one() doesn't call tg3_halt() under normal situation but
during kexec kdump I hit this problem.
Fixes: 932f19de ("tg3: Release tp->lock before invoking synchronize_irq()")
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 62749e2cb3c4a7da3eaa5c01a7e787aebeff8536 ]
Name fits better. Plus there's going to be introduced
__vlan_insert_tag later on.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 074975d0374333f656c48487aa046a21a9b9d7a1 ]
Commit 9a2620c877454 ("bnx2x: prevent WARN during driver unload")
switched the napi/busy_lock locking mechanism from spin_lock() into
spin_lock_bh(), breaking inter-operability with netconsole, as netpoll
disables interrupts prior to calling our napi mechanism.
This switches the driver into using atomic assignments instead of the
spinlock mechanisms previously employed.
Based on initial patch from Yuval Mintz & Ariel Elior
I basically added softirq starvation avoidance, and mixture
of atomic operations, plain writes and barriers.
Note this slightly reduces the overhead for this driver when no
busy_poll sockets are in use.
Fixes: 9a2620c877454 ("bnx2x: prevent WARN during driver unload")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit fde913e25496761a4e2a4c81230c913aba6289a2 ]
Commit 1daa4303b4ca ("net/mlx4_core: Deprecate error message at
ConnectX-2 cards startup to debug") did the deprecation only for port 1
of the card. Need to deprecate for port 2 as well.
Fixes: 1daa4303b4ca ("net/mlx4_core: Deprecate error message at ConnectX-2 cards startup to debug")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit e5eda89d97ec256ba14e7e861387cc0468259c18 ]
Netdevice registration should be performed a the end of the driver
initialization flow. If we don't do that, after calling register_netdevice,
device callbacks may be issued by higher layers of the stack before
final configuration of the device is done.
For example (VXLAN configuration race), mlx4_SET_PORT_VXLAN was issued
after the register_netdev command. System network scripts may configure
the interface (UP) right after the registration, which also attach
unicast VXLAN steering rule, before mlx4_SET_PORT_VXLAN was called,
causing the firmware to fail the rule attachment.
Fixes: 837052d0ccc5 ("net/mlx4_en: Add netdev support for TCP/IP offloads of vxlan tunneling")
Signed-off-by: Ido Shamay <idos@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 87f966d97b89774162df04d2106c6350c8fe4cb3 ]
On a MIPS Malta board, tons of fifo underflow errors have been observed
when using u-boot as bootloader instead of YAMON. The reason for that
is that YAMON used to set the pcnet device to SRAM mode but u-boot does
not. As a result, the default Tx threshold (64 bytes) is now too small to
keep the fifo relatively used and it can result to Tx fifo underflow errors.
As a result of which, it's best to setup the SRAM on supported controllers
so we can always use the NOUFLO bit.
Cc: <netdev@vger.kernel.org>
Cc: <stable@vger.kernel.org>
Cc: <linux-kernel@vger.kernel.org>
Cc: Don Fry <pcnet32@frontier.com>
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit da293700568ed3d96fcf062ac15d7d7c41377f11 ]
EEH recovery for bnx2x based adapters is not reliable on all Power
systems using the default hot reset, which can result in an
unrecoverable EEH error. Forcing the use of fundamental reset
during EEH recovery fixes this.
Cc: stable<stable@vger.kernel.org>
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit af5cbc9822f6bbe399925760a4d5ee82c21f258c ]
The current driver support receive VLAN CTAG HW acceleration feature
(NETIF_F_HW_VLAN_CTAG_RX) through software simulation. There calls the
api .skb_copy_to_linear_data_offset() to skip the VLAN tag, but there
have overlap between the two memory data point range. The patch just fix
the issue.
V2:
Michael Grzeschik suggest to use memmove() instead of skb_copy_to_linear_data_offset().
Reported-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Fixes: 1b7bde6d659d ("net: fec: implement rx_copybreak to improve rx performance")
Signed-off-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 4092e6acf5cb16f56154e2dd22d647023dc3d646 ]
This patch adds bcmgenet_tx_poll for the tx_rings. This can reduce the
interrupt load and send xmit in network stack on time. This also
separated for the completion of tx_ring16 from bcmgenet_poll.
The bcmgenet_tx_reclaim of tx_ring[{0,1,2,3}] operative by an interrupt
is to be not more than a certain number TxBDs. It is caused by too
slowly reclaiming the transmitted skb. Therefore, performance
degradation of xmit after 605ad7f ("tcp: refine TSO autosizing").
Signed-off-by: Jaedon Shin <jaedon.shin@gmail.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
This reverts commit 1e918876853aa85435e0f17fd8b4a92dcfff53d6.
Revert BQL support in r8169 driver as several regressions
point to this commit and we cannot figure out the real
cause yet.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
|
|
[ Upstream commit 24e579c8898aa641ede3149234906982290934e5 ]
With the commit d75b1ade567ffab ("net: less interrupt masking in NAPI") napi
repoll is done only when work_done == budget. When in busy_poll is we return 0
in napi_poll. We should return budget.
Signed-off-by: Govindarajulu Varadarajan <_govind@gmx.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 6088beef3f7517717bd21d90b379714dd0837079 ]
NAPI poll logic now enforces that a poller returns exactly the budget
when it wants to be called again.
If a driver limits TX completion, it has to return budget as well when
the limit is hit, not the number of received packets.
Reported-and-tested-by: Mike Galbraith <umgwanakikbuti@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: d75b1ade567f ("net: less interrupt masking in NAPI")
Cc: Manish Chopra <manish.chopra@qlogic.com>
Acked-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 02a54164c52ed6eca3089a0d402170fbf34d6cf5 upstream.
In Dual EMAC, the default VLANs are used to segregate Rx packets between
the ports, so adding the same default VLAN to the switch will affect the
normal packet transfers. So returning error on addition of dual EMAC
default VLANs.
Even if EMAC 0 default port VLAN is added to EMAC 1, it will lead to
break dual EMAC port separations.
Fixes: d9ba8f9e6298 (driver: net: ethernet: cpsw: dual emac interface implementation)
Reported-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 25906052d953d3fbdb7e19480b9de5e6bb949f3f upstream.
Since ALE table is a common resource for both the interfaces in Dual EMAC
mode and while bringing up the second interface in cpsw_ndo_set_rx_mode()
all the multicast entries added by the first interface is flushed out and
only second interface multicast addresses are added. Fixing this by
flushing multicast addresses based on dual EMAC port vlans which will not
affect the other emac port multicast addresses.
Fixes: d9ba8f9 (driver: net: ethernet: cpsw: dual emac interface implementation)
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 776d4e9f5c0229037707f692b386b1f2a5bac054 upstream.
Adds FCoE config option I40E_FCOE, so that FCoE can be enabled
as needed but otherwise have it disabled by default.
This also eliminate multiple FCoE config checks, instead now just
one config check for CONFIG_I40E_FCOE.
The I40E FCoE was added with 3.17 kernel and therefore this patch
shall be applied to stable 3.17 kernel also.
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Tested-by: Jim Young <jamesx.m.young@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 7ce67a38f799d1fb332f672b117efbadedaa5352 upstream.
The CPSW IP implements pulse-signaled interrupts. Due to
that we must write a correct, pre-defined value to the
CPDMA_MACEOIVECTOR register so the controller generates
a pulse on the correct IRQ line to signal the End Of
Interrupt.
The way the driver is written today, all four IRQ lines
are requested using the same IRQ handler and, because of
that, we could fall into situations where a TX IRQ fires
but we tell the controller that we ended an RX IRQ (or
vice-versa). This situation triggers an IRQ storm on the
reserved IRQ 127 of INTC which will in turn call ack_bad_irq()
which will, then, print a ton of:
unexpected IRQ trap at vector 00
In order to fix the problem, we are moving all calls to
cpdma_ctlr_eoi() inside the IRQ handler and making sure
we *always* write the correct value to the CPDMA_MACEOIVECTOR
register. Note that the algorithm assumes that IRQ numbers and
value-to-be-written-to-EOI are proportional, meaning that a
write of value 0 would trigger an EOI pulse for the RX_THRESHOLD
Interrupt and that's the IRQ number sitting in the 0-th index
of our irqs_table array.
This, however, is safe at least for current implementations of
CPSW so we will refrain from making the check smarter (and, as
a side-effect, slower) until we actually have a platform where
IRQ lines are swapped.
This patch has been tested for several days with AM335x- and
AM437x-based platforms. AM57x was left out because there are
still pending patches to enable ethernet in mainline for that
platform. A read of the TRM confirms the statement on previous
paragraph.
Reported-by: Yegor Yefremov <yegorslists@googlemail.com>
Fixes: 510a1e7 (drivers: net: davinci_cpdma: acknowledge interrupt properly)
Signed-off-by: Felipe Balbi <balbi@ti.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 17e96834fd35997ca7cdfbf15413bcd5a36ad448 ]
Hardware always provides compliment of IP pseudo checksum. Stack expects
whole packet checksum without pseudo checksum if CHECKSUM_COMPLETE is set.
This causes checksum error in nf & ovs.
kernel: qg-19546f09-f2: hw csum failure
kernel: CPU: 9 PID: 0 Comm: swapper/9 Tainted: GF O-------------- 3.10.0-123.8.1.el7.x86_64 #1
kernel: Hardware name: Cisco Systems Inc UCSB-B200-M3/UCSB-B200-M3, BIOS B200M3.2.2.3.0.080820141339 08/08/2014
kernel: ffff881218f40000 df68243feb35e3a8 ffff881237a43ab8 ffffffff815e237b
kernel: ffff881237a43ad0 ffffffff814cd4ca ffff8829ec71eb00 ffff881237a43af0
kernel: ffffffff814c6232 0000000000000286 ffff8829ec71eb00 ffff881237a43b00
kernel: Call Trace:
kernel: <IRQ> [<ffffffff815e237b>] dump_stack+0x19/0x1b
kernel: [<ffffffff814cd4ca>] netdev_rx_csum_fault+0x3a/0x40
kernel: [<ffffffff814c6232>] __skb_checksum_complete_head+0x62/0x70
kernel: [<ffffffff814c6251>] __skb_checksum_complete+0x11/0x20
kernel: [<ffffffff8155a20c>] nf_ip_checksum+0xcc/0x100
kernel: [<ffffffffa049edc7>] icmp_error+0x1f7/0x35c [nf_conntrack_ipv4]
kernel: [<ffffffff814cf419>] ? netif_rx+0xb9/0x1d0
kernel: [<ffffffffa040eb7b>] ? internal_dev_recv+0xdb/0x130 [openvswitch]
kernel: [<ffffffffa04c8330>] nf_conntrack_in+0xf0/0xa80 [nf_conntrack]
kernel: [<ffffffff81509380>] ? inet_del_offload+0x40/0x40
kernel: [<ffffffffa049e302>] ipv4_conntrack_in+0x22/0x30 [nf_conntrack_ipv4]
kernel: [<ffffffff815005ca>] nf_iterate+0xaa/0xc0
kernel: [<ffffffff81509380>] ? inet_del_offload+0x40/0x40
kernel: [<ffffffff81500664>] nf_hook_slow+0x84/0x140
kernel: [<ffffffff81509380>] ? inet_del_offload+0x40/0x40
kernel: [<ffffffff81509dd4>] ip_rcv+0x344/0x380
Hardware verifies IP & tcp/udp header checksum but does not provide payload
checksum, use CHECKSUM_UNNECESSARY. Set it only if its valid IP tcp/udp packet.
Cc: Jiri Benc <jbenc@redhat.com>
Cc: Stefan Assmann <sassmann@redhat.com>
Reported-by: Sunil Choudhary <schoudha@redhat.com>
Signed-off-by: Govindarajulu Varadarajan <_govind@gmx.com>
Reviewed-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 7a05dc64e2e4c611d89007b125b20c0d2a4d31a5 ]
Commit d75b1ade567f ("net: less interrupt masking in NAPI") uncovered
wrong alx_poll() behavior.
A NAPI poll() handler is supposed to return exactly the budget when/if
napi_complete() has not been called.
It is also supposed to return number of frames that were received, so
that netdev_budget can have a meaning.
Also, in case of TX pressure, we still have to dequeue received
packets : alx_clean_rx_irq() has to be called even if
alx_clean_tx_irq(alx) returns false, otherwise device is half duplex.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: d75b1ade567f ("net: less interrupt masking in NAPI")
Reported-by: Oded Gabbay <oded.gabbay@amd.com>
Bisected-by: Oded Gabbay <oded.gabbay@amd.com>
Tested-by: Oded Gabbay <oded.gabbay@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit a51e0df4c1e06afd7aba84496c14238e6b363caa ]
Previously, mlx4_mt_rereg_write filled the MPT's entity_size with the
old MTT's page shift, which could result in using an incorrect offset.
Fix the initialization to be after we calculate the new MTT offset.
In addition, assign mtt order to -1 after calling mlx4_mtt_cleanup. This
is necessary in order to mark the MTT as invalid and avoid freeing it later.
Fixes: e630664 ('mlx4_core: Add helper functions to support MR re-registration')
Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 5f35227ea34bb616c436d9da47fc325866c428f3 ]
GSO isn't the only offload feature with restrictions that
potentially can't be expressed with the current features mechanism.
Checksum is another although it's a general issue that could in
theory apply to anything. Even if it may be possible to
implement these restrictions in other ways, it can result in
duplicate code or inefficient per-packet behavior.
This generalizes ndo_gso_check so that drivers can remove any
features that don't make sense for a given packet, similar to
netif_skb_features(). It also converts existing driver
restrictions to the new format, completing the work that was
done to support tunnel protocols since the issues apply to
checksums as well.
By actually removing features from the set that are used to do
offloading, it solves another problem with the existing
interface. In these cases, GSO would run with the original set
of features and not do anything because it appears that
segmentation is not required.
CC: Tom Herbert <therbert@google.com>
CC: Joe Stringer <joestringer@nicira.com>
CC: Eric Dumazet <edumazet@google.com>
CC: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Acked-by: Tom Herbert <therbert@google.com>
Fixes: 04ffcb255f22 ("net: Add ndo_gso_check")
Tested-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 492f5add4be84652bbe13da8a250d60c6856a5c5 ]
iowrite32() will byteswap it's argument on big endian archs.
iowrite32be() will byteswap on little endian archs.
Since we don't want to do this unnecessary byteswap on the fast path,
doorbell is stored in the NIC's native endianness. Using the right
iowrite() according to the arch endianness.
CC: Wei Yang <weiyang@linux.vnet.ibm.com>
CC: David Laight <david.laight@aculab.com>
Fixes: 6a4e812 ("net/mlx4_en: Avoid calling bswap in tx fast path")
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 05b0aa579397b734f127af58e401a30784a1e315 ]
During driver load in tg3_init_one, if the driver detects DMA activity before
intializing the chip tg3_halt is called. As part of tg3_halt interrupts are
disabled using routine tg3_disable_ints. This routine was using mailbox value
which was not initialized (default value is 0). As a result driver was writing
0x00000001 to pci config space register 0, which is the vendor id / device id.
This driver bug was exposed because of the commit a7877b17a667 (PCI: Check only
the Vendor ID to identify Configuration Request Retry). Also this issue is only
seen in older generation chipsets like 5722 because config space write to offset
0 from driver is possible. The newer generation chips ignore writes to offset 0.
Also without commit a7877b17a667, for these older chips when a GRC reset is
issued the Bootcode would reprogram the vendor id/device id, which is the reason
this bug was masked earlier.
Fixed by initializing the interrupt mailbox registers before calling tg3_halt.
Please queue for -stable.
Reported-by: Nils Holland <nholland@tisys.org>
Reported-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Prashant Sreedharan <prashant@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit c3f2511feac088030055012cc8f64ebd84c87dbc ]
This commit contains 2 fixes for the 128B CQE/EQE stride feaure.
Wei found that mlx4_QUERY_HCA function marked the wrong capability
in flags (64B CQE/EQE), when CQE/EQE stride feature was enabled.
Also added small fix in initial CQE ownership bit assignment, when CQE
is size is not default 32B.
Fixes: 77507aa24 (net/mlx4: Enable CQE/EQE stride support)
Signed-off-by: Wei Yang <weiyang@linux.vnet.ibm.com>
Signed-off-by: Ido Shamay <idos@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 5f478b41033606d325e420df693162e2524c2b94 ]
mvneta_tx() dereferences skb to get skb->len too late,
as hardware might have completed the transmit and TX completion
could have freed the skb from another cpu.
Fixes: 71f6d1b31fb1 ("net: mvneta: replace Tx timer with a real interrupt")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit aebea2ba0f7495e1a1c9ea5e753d146cb2f6b845 ]
The mvneta driver sets the amount of Tx coalesce packets to 16 by
default. Normally that does not cause any trouble since the driver
uses a much larger Tx ring size (532 packets). But some sockets
might run with very small buffers, much smaller than the equivalent
of 16 packets. This is what ping is doing for example, by setting
SNDBUF to 324 bytes rounded up to 2kB by the kernel.
The problem is that there is no documented method to force a specific
packet to emit an interrupt (eg: the last of the ring) nor is it
possible to make the NIC emit an interrupt after a given delay.
In this case, it causes trouble, because when ping sends packets over
its raw socket, the few first packets leave the system, and the first
15 packets will be emitted without an IRQ being generated, so without
the skbs being freed. And since the socket's buffer is small, there's
no way to reach that amount of packets, and the ping ends up with
"send: no buffer available" after sending 6 packets. Running with 3
instances of ping in parallel is enough to hide the problem, because
with 6 packets per instance, that's 18 packets total, which is enough
to grant a Tx interrupt before all are sent.
The original driver in the LSP kernel worked around this design flaw
by using a software timer to clean up the Tx descriptors. This timer
was slow and caused terrible network performance on some Tx-bound
workloads (such as routing) but was enough to make tools like ping
work correctly.
Instead here, we simply set the packet counts before interrupt to 1.
This ensures that each packet sent will produce an interrupt. NAPI
takes care of coalescing interrupts since the interrupt is disabled
once generated.
No measurable performance impact nor CPU usage were observed on small
nor large packets, including when saturating the link on Tx, and this
fixes tools like ping which rely on too small a send buffer. If one
wants to increase this value for certain workloads where it is safe
to do so, "ethtool -C $dev tx-frames" will override this default
setting.
This fix needs to be applied to stable kernels starting with 3.10.
Tested-By: Maggie Mae Roxas <maggie.mae.roxas@gmail.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This resolves the following bug which can be reproduced by building the
kernel with CONFIG_DEBUG_ATOMIC_SLEEP=y and reading network statistics
while the network interface is down.
e.g.:
ifconfig eth0 down
cat /sys/class/net/eth0/statistics/tx_errors
----
[ 1238.161349] BUG: sleeping function called from invalid context at drivers/base/power/runtime.c:952
[ 1238.188279] in_atomic(): 1, irqs_disabled(): 0, pid: 1388, name: cat
[ 1238.207425] CPU: 0 PID: 1388 Comm: cat Not tainted 3.10.31-ltsi-00046-gefa0b46 #1087
[ 1238.230737] Backtrace:
[ 1238.238123] [<c0012e64>] (dump_backtrace+0x0/0x10c) from [<c0013000>] (show_stack+0x18/0x1c)
[ 1238.263499] r6:000003b8 r5:c06160c0 r4:c0669e00 r3:00404000
[ 1238.280583] [<c0012fe8>] (show_stack+0x0/0x1c) from [<c04515a4>] (dump_stack+0x20/0x28)
[ 1238.304631] [<c0451584>] (dump_stack+0x0/0x28) from [<c004970c>] (__might_sleep+0xf8/0x118)
[ 1238.329734] [<c0049614>] (__might_sleep+0x0/0x118) from [<c02465ac>] (__pm_runtime_resume+0x38/0x90)
[ 1238.357170] r7:d616f000 r6:c049c458 r5:00000004 r4:d6a17210
[ 1238.374251] [<c0246574>] (__pm_runtime_resume+0x0/0x90) from [<c029b1c4>] (sh_eth_get_stats+0x44/0x280)
[ 1238.402468] r7:d616f000 r6:c049c458 r5:d5c21000 r4:d5c21000
[ 1238.419552] [<c029b180>] (sh_eth_get_stats+0x0/0x280) from [<c03ae39c>] (dev_get_stats+0x54/0x88)
[ 1238.446204] r5:d5c21000 r4:d5ed7e08
[ 1238.456980] [<c03ae348>] (dev_get_stats+0x0/0x88) from [<c03c677c>] (netstat_show.isra.15+0x54/0x9c)
[ 1238.484413] r6:d5c21000 r5:d5c21238 r4:00000028 r3:00000001
[ 1238.501495] [<c03c6728>] (netstat_show.isra.15+0x0/0x9c) from [<c03c69b8>] (show_tx_errors+0x18/0x1c)
[ 1238.529196] r7:d5f945d8 r6:d5f945c0 r5:c049716c r4:c0650e7c
[ 1238.546279] [<c03c69a0>] (show_tx_errors+0x0/0x1c) from [<c023963c>] (dev_attr_show+0x24/0x50)
[ 1238.572157] [<c0239618>] (dev_attr_show+0x0/0x50) from [<c010c148>] (sysfs_read_file+0xb0/0x140)
[ 1238.598554] r5:c049716c r4:d5c21240
[ 1238.609326] [<c010c098>] (sysfs_read_file+0x0/0x140) from [<c00b9ee4>] (vfs_read+0xb0/0x13c)
[ 1238.634679] [<c00b9e34>] (vfs_read+0x0/0x13c) from [<c00ba0ac>] (SyS_read+0x44/0x74)
[ 1238.657944] r8:bef45bf0 r7:00000000 r6:d6ac0600 r5:00000000 r4:00000000
[ 1238.678172] [<c00ba068>] (SyS_read+0x0/0x74) from [<c000eec0>] (ret_fast_syscall+0x0/0x30)
----
Signed-off-by: Mitsuhiro Kimura <mitsuhiro.kimura.kc@renesas.com>
Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Original code only check/alloc plat_dat for the CONFIG_OF case, this
patch check/alloc it earlier and unconditionally to avoid kernel build
warnings:
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:275
stmmac_pltfr_probe() warn: variable dereferenced before check 'plat_dat'
V2: Fix coding style.
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In the current driver, allocation size of skb does not care the alignment
adjust after allocation.
And also, in the current implementation, buffer alignment method by
sh_eth_set_receive_align function has a bug that this function displace
buffer start address forcedly when the alignment is corrected.
In the result, tail of the skb will exceed allocated area and kernel panic
will be occurred.
This patch fix this issue.
Signed-off-by: Mitsuhiro Kimura <mitsuhiro.kimura.kc@renesas.com>
Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
ndo_bridge_setlink() is currently only called on the slave if
IFLA_AF_SPEC is set but this is a very fragile assumption and may
change in the future.
Cc: Ajit Khaparde <ajit.khaparde@emulex.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|