Since this variable is now part of a structure and not allocated dynamically,
this test is irrelevant now.
Signed-off-by: Dotan Barak <dotanb@dev.mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RX rings were cleaned while there was still possible RX traffic completion
handling.
Change the sequance of events so that the port is closed and the QPs are being
stopped before RX cleanup.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The port vlan table size is 126 (used for IBoE) so after 126 we will
not have space and the user need to see it only in debug print and not
error.
Signed-off-by: Aviad Yehezkel <aviadye@mellanox.com>
Reviewed-by: Yevgeny Petrilin <yevgenyp@mellanox.com>
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To avoid a race between the open function and everything that happens after
register_netdev() move it to be the last operation called.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are no counters allocated to the eth device when the port is down, so
this query is meaningless at that time.
It also leads to querying incorrect counters (since the counter_index is not
valid when the device port is down).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Eric Dumazet <edumazet@google.com>
mlx4 exclusively uses order-2 allocations in RX path, which are
likely to fail under memory pressure.
We therefore drop frames more than needed.
This patch tries order-3, order-2, order-1 and finally order-0
allocations to keep good performance, yet allow allocations if/when
memory gets fragmented.
By using larger pages, and avoiding unnecessary get_page()/put_page()
on compound pages, this patch improves performance as well, lowering
false sharing on struct page.
Also use GFP_KERNEL allocations in initialization path, as allocating 12
MB (390 order-3 pages) can easily fail with GFP_ATOMIC.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Amir Vadai <amirv@mellanox.com>
Acked-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
get user pages might fail partially in macvtap zero copy
mode. To recover we need to put all pages that we got,
but code used a wrong index resulting in double-free
errors.
Reported-by: Brad Hubbard <bhubbard@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
get user pages might fail partially in tun zero copy
mode. To recover we need to put all pages that we got,
but code used a wrong index resulting in double-free
errors.
Reported-by: Brad Hubbard <bhubbard@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ben Hutchings says:
====================
1. Make EEH recovery work when using legacy interrupts, from Alexandre
Rames.
2. Enable accelerated RFS for VLAN-tagged flows, from Andy Lutomirski.
3. Improve performance for non-TCP (and particularly UDP) traffic, which
regressed in 3.10 when we switched to always allocating paged RX
buffers. Partly by Jon Cooper.
4. Some minor bug fixes to IOMMU detection, timestamping capabilities,
and IRQ cleanup on the probe failure path.
I've dropped the RX skb cache, which improved some benchmarks but
perhaps needs some reworking to be more generally useful.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The zero MAC entry in the fdb is used as default destination. With
multiple default destinations it is possible to use vxlan in
environments that disable multicast on the infrastructure level, e.g.
public clouds.
Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Fix following sparse warnings.
drivers/net/vxlan.c:238:44: warning: incorrect type in argument 3 (different base types)
drivers/net/vxlan.c:238:44: expected restricted __be32 [usertype] value
drivers/net/vxlan.c:238:44: got unsigned int const [unsigned] [usertype] remote_vni
drivers/net/vxlan.c:1735:18: warning: incorrect type in initializer (different signedness)
drivers/net/vxlan.c:1735:18: expected int *id
drivers/net/vxlan.c:1735:18: got unsigned int static [toplevel] *<noident>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
To improve power consumption in idle associated mode FW may lower
RX power. This low linearity mode is acceptable for listening low rate
RX such as beacons and groupcast. The driver enables LPRX only if PM
is enabled and associated AP's beacon TX rate is 1Mbps or 6Mbps.
LPRX RSSI threshold is used to limit a range where LPRX is applied.
Signed-off-by: Alexander Bondar <alexander.bondar@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
A few places use 'pcie_trans' which is a bit non-standard,
use 'trans_pcie' there as well.
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The PCIe code has an array of buffer descriptors (RXBs) that have pages
and DMA mappings attached. In regular use, the array isn't used and the
buffers are either on the hardware receive queue or the rx_free/rx_used
lists for recycling.
Occasionally, during module unload, we'd see a warning from this:
WARNING: at lib/list_debug.c:32 __list_add+0x91/0xa0()
list_add corruption. prev->next should be next (c31c98cc), but was c31c80bc. (prev=c31c80bc).
Pid: 519, comm: rmmod Tainted: G W O 3.4.24-dev #3
Call Trace:
[<c10335b2>] warn_slowpath_common+0x72/0xa0
[<c1033683>] warn_slowpath_fmt+0x33/0x40
[<c12e31d1>] __list_add+0x91/0xa0
[<fdf2083c>] iwl_pcie_rxq_free_rbs+0xcc/0xe0 [iwlwifi]
[<fdf21b3f>] iwl_pcie_rx_free+0x3f/0x210 [iwlwifi]
[<fdf2dd7a>] iwl_trans_pcie_free+0x2a/0x90 [iwlwifi]
The reason for this seems to be that in iwl_pcie_rxq_free_rbs() we use
the array to free all buffers (the hardware receive queue isn't in use
any more at this point). The function also adds all buffers to rx_used
because it's also used during initialisation (when no freeing happens.)
This can cause the warning because it may add entries to the list that
are already on it. Luckily, this is harmless because it can only happen
when the entire data structure is freed anyway, since during init both
lists are initialized from scratch.
Disentangle this code and treat init and free separately. During init
we just need to put them onto the list after freeing all buffers (for
switching between 4k/8k buffers); during free no list manipulations
are necessary at all.
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
In case that an AP/GO interface is started while there is a
station/P2P client associated, need to make sure that the AP/GO
beacon time is far enough from the station's one in oder to allow
the station to receive the DTIM beacons and the following traffic
etc.
To resolve this, when the AP is started, check if there is an
active station interface, and guarantee that the AP/GO TBTT is far
enough from the station one.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Add prints visible to the user when entering and exiting
thrermal throttling, because so users can tell that the
NIC is getting too hot (and throughput will decrease.)
Signed-off-by: eytan lifshitz <eytan.lifshitz@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Currently, there is no good possibility to debug netlink traffic that
is being exchanged between kernel and user space. Therefore, this patch
implements a netlink virtual device, so that netlink messages will be
made visible to PF_PACKET sockets. Once there was an approach with a
similar idea [1], but it got forgotten somehow.
I think it makes most sense to accept the "overhead" of an extra netlink
net device over implementing the same functionality from PF_PACKET
sockets once again into netlink sockets. We have BPF filters that can
already be easily applied which even have netlink extensions, we have
RX_RING zero-copy between kernel- and user space that can be reused,
and much more features. So instead of re-implementing all of this, we
simply pass the skb to a given PF_PACKET socket for further analysis.
Another nice benefit that comes from that is that no code needs to be
changed in user space packet analyzers (maybe adding a dissector, but
not more), thus out of the box, we can already capture pcap files of
netlink traffic to debug/troubleshoot netlink problems.
Also thanks goes to Thomas Graf, Flavio Leitner, Jesper Dangaard Brouer.
[1] http://marc.info/?l=linux-netdev&m=113813401516110
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This restores commits:
c573972c111a5904342cda2e2c2149
which initially accidently went into 'net', were
reverted there, and then properly placed into 'net-next'.
But the next net --> net-next merge accidently wiped them
out again.
Reported-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The device::iommu_group field may be set even if no IOMMU is in use.
iommu_present() is still a better indicator, although it doesn't tell
us whether *our* device is affected.
Reported-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
The lifetime of an irq_cpu_rmap is odd: we have to allocate it before
installing IRQ handlers and free it before removing the IRQ handlers.
As a result of this asymmetry, it was omitted from some failure paths.
On another failure path, we could try to remove IRQ handlers we
had not yet installed.
Move the irq_cpu_rmap allocation and freeing alongside IRQ handler
installation and removal, in efx_nic_{init,fini}_interrupts().
Count the number of IRQ handlers successfully installed and only
remove those on the failure path.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
GRO can handle non-TCP packets and pass them up without coalescing,
but it has to do some extra work to parse the packet which we can
bypass using the hardware parse result. (This condition yields a
false negative for TCP/IPv6 packets received by Falcon, but its
performance is already poor in that case due to lack of checksum
offload.)
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
As far as I know, the hardware doesn't support matching on both IP
fields and vlan tag, but it can at least match on the IP fields.
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
The kernel can generate software receive timestamps and we should
report those for all ports regardless of hardware capabilities.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
PCI legacy interrupts are level-triggered, and we cannot mask them up
on an isolated device. Instead, disable the IRQ at the controller
until we have recovered.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Map BIT 9 in TX DMA DWARD 0 as HW write back option.
We must turn on this option in the last TX descriptor,
this is required for old HW compatability.
This option indicate to HW that WB is required for this descriptor.
Signed-off-by: Kirshenbaum Erez <erezk@wilocity.com>
Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The vring index (MAC queue id) must be set in all TX descriptors
otherwise HW will fail to release descriptors for a specific vring
(disconnect or vring switch flows).
This is normally occurs when fragmentation required, if vring index
will not be the same for all SKB descriptors HW will fail to flush
this MAC queue.
Signed-off-by: Kirshenbaum Erez <erezk@wilocity.com>
Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Use common names instead of chip specific ones.
The patch contains no functional changes, but
it makes it easier to add support for further
descriptor sizes.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Different chipsets may use different TXWI descriptor
size. Instead of using a hardcoded value, use the
'queue->winfo_size' which holds the correct value for
a given device.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The current code uses the same index value both
for the channel information array and for the TX
power table. The index starts from 14, however the
index of the TX power table must start from zero.
Fix it, in order to get the correct TX power value
for a given channel.
The changes in rt61pci.c and rt73usb.c are compile
tested only.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: stable@vger.kernel.org
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
