The SAP ref counter gets decremented twice when deleting a socket,
although for all but the first socket of a SAP the SAP ref counter was
incremented only once.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the cases where a lot of interfaces are used in conjunction with a
lot of LLC sockets bound to the same SAP, the iteration of the socket
list becomes prohibitively expensive.
Replacing the list with a a local address based hash significantly
improves the bind and listener lookup operations as well as the
datagram delivery.
Connected sockets delivery is also improved, but this patch does not
address the case where we have lots of sockets with the same local
address connected to different remote addresses.
In order to keep the socket sanity checks alive and fast a socket
counter was added to the SAP structure.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds a per SAP device based hash table to solve the
multicast delivery scalability issue when we have large number of
interfaces and a large number of sockets bound to the same SAP.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Optimize multicast delivery by doing the actual delivery without
holding the lock. Based on the same approach used in UDP code.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the reclamation phase we use the SLAB_DESTROY_BY_RCU mechanism,
which require some extra checks in the lookup code:
a) If the current socket was released, reallocated & inserted in
another list it will short circuit the iteration for the current list,
thus we need to restart the lookup.
b) If the current socket was released, reallocated & inserted in the
same list we just need to recheck it matches the look-up criteria and
if not we can skip to the next element.
In this case there is no need to restart the lookup, since sockets are
inserted at the start of the list and the worst that will happen is
that we will iterate throught some of the list elements more then
once.
Note that the /proc and multicast delivery was not yet converted to
RCU, it still uses spinlocks for protection.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using bind(MAC address) with LLC sockets has O(n) complexity, where n
is the number of interfaces. To overcome this, we add support for
SO_BINDTODEVICE which drops the complexity to O(1).
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using dev_hard_header allows us to use LLC with VLANs and potentially
other Ethernet/TokernRing specific encapsulations. It also removes code
duplication between LLC and Ethernet/TokenRing core code.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A few lines earlier we assume that best->slave could be either null or non-null so
we should check it here as well.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit f001fde5ea changed
net_device.dev_addr from a 32-byte array to a pointer.
I found 4 ethernet drivers which rely on sizeof(dev_addr), which are now
only copying 4 bytes of the address information on 32bit systems.
Fix them to use ETH_ALEN.
Signed-off-by: Daniel Drake <dsd@laptop.org>
Reviewed-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The gfar_select_queue() function was used to set queue mapping
only for forwarding/bridging applications and the condition
for locally generated packets was completely ignored.
The solution is to remove the gfar_select_queue() function and
use skb_record_rx_queue to set queue mapping for
forwarding/bridging applications. This will ensure that in case of
forwarding/bridging applications txq = rxq will be selected and
skb_tx_hash will be used to pick up a txq for locally generated packets.
Signed-off-by: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using macro tun_sk is more clear and shorter. However tun.c has tun_sk,
but doesn't use it.
Signed-off-by: Vitaliy Gusev <vgusev@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If PHY doesn't have an IRQ, phylib would poll for link changes, and
would call adjust_link() every second. In that case we disable and
enable the controller every second.
Let's better check if there is actually anything changed, and, if so,
change the MAC settings.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 864fdf884e ("ucc_geth:
Fix hangs after switching from full to half duplex") ucc_geth driver
disables the controller during MAC configuration changes. Though,
disabling the controller might take quite awhile, and so the netdev
watchdog might get upset:
NETDEV WATCHDOG: eth2 (ucc_geth): transmit queue 0 timed out
------------[ cut here ]------------
Badness at c02729a8 [verbose debug info unavailable]
NIP: c02729a8 LR: c02729a8 CTR: c01b6088
REGS: c0451c40 TRAP: 0700 Not tainted (2.6.32-trunk-8360e)
[...]
NIP [c02729a8] dev_watchdog+0x280/0x290
LR [c02729a8] dev_watchdog+0x280/0x290
Call Trace:
[c0451cf0] [c02729a8] dev_watchdog+0x280/0x290 (unreliable)
[c0451d50] [c00377c4] run_timer_softirq+0x164/0x224
[c0451da0] [c0032a38] __do_softirq+0xb8/0x13c
[c0451df0] [c00065cc] do_softirq+0xa0/0xac
[c0451e00] [c003280c] irq_exit+0x7c/0x9c
[c0451e10] [c00640c4] __ipipe_sync_stage+0x248/0x24c
[...]
This patch fixes the issue by detaching the netdev during the
time we change the configuration.
Reported-by: Lennart Sorensen <lsorense@csclub.uwaterloo.ca>
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Tested-by: Lennart Sorensen <lsorense@csclub.uwaterloo.ca>
Cc: Stable <stable@vger.kernel.org> [2.6.32]
Signed-off-by: David S. Miller <davem@davemloft.net>
Following oops was seen with the ucc_geth driver:
Unable to handle kernel paging request for data at address 0x00000058
Faulting instruction address: 0xc024f2fc
Oops: Kernel access of bad area, sig: 11 [#1]
[...]
NIP [c024f2fc] skb_recycle_check+0x14/0x100
LR [e30aa0a4] ucc_geth_poll+0xd8/0x4e0 [ucc_geth_driver]
Call Trace:
[df857d50] [c000b03c] __ipipe_grab_irq+0x3c/0xa4 (unreliable)
[df857d60] [e30aa0a4] ucc_geth_poll+0xd8/0x4e0 [ucc_geth_driver]
[df857dd0] [c0258cf8] net_rx_action+0xf8/0x1b8
[df857e10] [c0032a38] __do_softirq+0xb8/0x13c
[df857e60] [c00065cc] do_softirq+0xa0/0xac
[...]
This is because ucc_geth_tx() tries to process an empty queue when
queues are logically stopped. Stopping the queues doesn't disable
polling, and since nowadays ucc_geth_tx() is actually called from
the polling routine, the oops above might pop up.
Fix this by removing 'netif_queue_stopped() == 0' check.
Reported-by: Lennart Sorensen <lsorense@csclub.uwaterloo.ca>
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Tested-by: Lennart Sorensen <lsorense@csclub.uwaterloo.ca>
Cc: Stable <stable@vger.kernel.org> [2.6.32]
Signed-off-by: David S. Miller <davem@davemloft.net>
when using policy routing and the skb mark:
there are cases where a back path validation requires us
to use a different routing table for src ip validation than
the one used for mapping ingress dst ip.
One such a case is transparent proxying where we pretend to be
the destination system and therefore the local table
is used for incoming packets but possibly a main table would
be used on outbound.
Make the default behavior to allow the above and if users
need to turn on the symmetry via sysctl src_valid_mark
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the device is reset during MTU change, ring size change, or self
test, etc, the cnic status block needs to be properly initialized if
cnic is registered.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Apparently some awk versions choke on C-style comments -- who knew? :-)
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This updates pktgen so that it does not decrement skb->users
when it receives valid NET_XMIT_xxx values. These are now
valid return values from ndo_start_xmit in net-next-2.6.
They also indicate the skb has been consumed.
This fixes pktgen to work correctly with vlan devices.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are BUGs "scheduling while atomic" triggered by the timer
rhine_tx_timeout(). They are caused by calling napi_disable() (with
msleep()). This patch fixes it by moving most of the timer content to
the workqueue function (similarly to other drivers, like tg3), with
spin_lock() changed to BH version.
Additionally, there is spin_lock_irq() moved in rhine_close() to
exclude napi_disable() etc., also tg3's way.
Reported-by: Andrey Rahmatullin <wrar@altlinux.org>
Tested-by: Andrey Rahmatullin <wrar@altlinux.org>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds a pci_save_state() call in ixgbe_resume() after
pci_restore_state(). A similar change was made in ixgbe_io_slot_reset()
that accommodates pci_restore_state() new behavior. This change makes
pci_restore_state() clear the saved_state flag This is necessary due
to a resent kernel change to pci_restore_state() so that it now clears
the saved_state flag of the device right after the device.s standard
configuration registers have been poplulated with the previously saved
values.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some misc changes to cleanup whitespace issues and fix/remove some #define
HW defintions.
1) fix some whitespace issues
2) more verbose err msg when resources aren't available to configure vnic
3) remove unused #define
4) fix RSS #define rss hash types
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Scott Feldman <scofeldm@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only rx_usec and tx_usec options for ethtool -C are settable as those
are the only settings that make sense to HW. Adds driver reporting of
intr coalescing timer value in usec units rather than HW units.
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Scott Feldman <scofeldm@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previous driver was aligning ring sizes to 16 descs, but hardware actually
wants desc ring sizes to be aligned to 32 descs.
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Scott Feldman <scofeldm@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In set_multicast(), only push pkt filter changes down to firmware if
pkt filter actually changes.
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Scott Feldman <scofeldm@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During dev->open(), make sure we get at least one skb on the Rx ring.
Otherwise abort the interface load. Also, if we get skb allocation
failures in NAPI poll while trying to replenish the ring, try again
later so we don't end up starving out the Rx ring completely.
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Scott Feldman <scofeldm@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix dev->stop shutdown bug where driver was stopping xmit queue and then
disabling intrs. Fix is to disable intrs first and then stop the xmit
queue, otherwise an interrupt could cause the queue to be rewoken. Also,
no need to explicitly do queue servicing because queues are cleaned and
reset back to initial state at end of dev->stop. Servicing queues also
had the side-effect of also rewakening the xmit queue, which is not what
we want.
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Scott Feldman <scofeldm@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Changes to return correct values for transceiver and supported in
ethtool get_settings function.
Signed-off-by: Sarveshwar Bandi <sarveshwarb@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NIC controller has to be set to an appropriate mode before doing a loopback
test. Test will fail otherwise.
Signed-off-by: Sarveshwar Bandi <sarveshwarb@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change ensures that loopback test command gives up after 4 seconds when
the hardware is not responsive. This could happen if the ports are connected
properly in loopback mode.
Signed-off-by: Sarveshwar Bandi <sarveshwarb@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ethtool code for enabling Wake on Lan was not correctly checking the
status register bits so as a result ports 0 and 2 were both being allowed
to set WOL to enabled even though it is only supported on the first port
for our adapters.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The watermark values for the 82575 were not being set correctly. As a
result the high and low watermark values were set to the same value which
can lead to excess xon/xoff packets being generated.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change resolves an issue seen in some configurations where the link
may drop to 100Mb/s even though the link itself supports 1000Mb/s. The
issue was root caused to the fact that we were only trying the link once.
Now instead we will try up to 5 attempts on a faulty cable before
downshifting to 100Mb/s.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We were forcing the PCS link up in error when we are in KX mode. We should
only be disabling autoneg, not forcing the link up.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This hardware watchdog can misfire, so it does more harm than good.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some cables have EEPROMs that conflict with the PHY's on-board EEPROM
so it cannot load firmware.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The PHY's firmware very occasionally appears to lock up very early, but
with the heartbeat update still running. Rebooting the microcontroller
core seems to be sufficient to recover.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The PHY boots in a mode which is not necessarily optimal. This change
switches it to self-configure mode (except when in loopback, which
won't work in that mode if an SFP+ module is not present) by rebooting
the PHY's microcontroller, and replicating the sequence of configuration
writes from the boot EEPROM with the appropriate changes.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If we see the PHY remaining stuck in a link-down state due to PCS being
down while PMA/PMD is up, we briefly switch to PMA/PMD loopback and back,
which usually unsticks it.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need buffer->len to remain valid to work out the correct address to
be unmapped. We therefore need to clear buffer->len after the unmap
operation.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The XGXS block may not get a link immediately in XGXS or XAUI loopback
modes, so we still need to check it. Split falcon_xaui_link_ok() into
falcon_xgxs_link_ok(), which checks only the Falcon XGXS block, and
falcon_xmac_link_ok(), which checks one or both sides of the link as
appropriate. Also rename falcon_check_xaui_link() to
falcon_xmac_link_ok_retry().
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This prevents efx->link_advertising from being blatted during
a reset.
The phy_short_reach sysfs node is now destroyed later in the
port shutdown process, so check for STATE_RUNNING after
acquiring the rtnl_lock (just like in set_phy_flash_cfg).
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. remove address argument since we use same address IWMC_SDIO_DATA_ADDR in
all cases
2. add __iwmct_tx - non locking tx function for already locked contexts
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. add TRACE level
2. use TRACE where needed to reduce the noise
3 don't INFOEX from driver
4. add DUMP level for packets dumps
5. use correct context for the log messages
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. removed redundant NULL-pointers checks in iwmct_fw_load
as release_firmware and kfree are NULL pointer friendly
2. remove redundant memset of the parser since the structure
is fully initialized in iwmct_fw_parser_init function
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Octavian Purdila
Ralf Baechle
Dan Carpenter
Daniel Drake
Sandeep Gopalpet
Vitaliy Gusev
Anton Vorontsov
Denis Kirjanov
Jamal Hadi Salim
Michael Chan
John W. Linville
David S. Miller
John Fastabend
Jarek Poplawski
Don Skidmore
Scott Feldman
Sarveshwar Bandi
Williams, Mitch A
Alexander Duyck
Ben Hutchings
Matthew Slattery
Neil Turton
Steve Hodgson
Tomas Winkler