skb_dma_unmap() is quite expensive for small packets,
because we use two different cache lines from skb_shared_info.
One to access nr_frags, one to access dma_maps[0]
Instead of dma_maps being an array of MAX_SKB_FRAGS + 1 elements,
let dma_head alone in a new dma_head field, close to nr_frags,
to reduce cache lines misses.
Tested on my dev machine (bnx2 & tg3 adapters), nice speedup !
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With bi-directional stress traffic, the receiver could hang causing the
hardware to stop and a "Detected Tx Unit Hang" message dumped to the system
logfile. Temporarily workaround this issue by disabling Tx flow control by
default. The issue is currently being investigated and a follow-on patch
will be provided to revert this when it is resolved.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch provides support for the next generation Intel desktop
and mobile gigabit ethernet LOM adapters. These adapters are the
follow-on parts to the LOMs tied to the prior ICH chipsets and are
comprised of a MAC in the PCH chipset and an external PHY (82577 for
mobile and 82578 for desktop versions). New features consist of PHY
wakeup to save power by completely turning off the MAC while in Sx
state, and 4K jumbo frames.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
By putting the maximum frame size supported by the hardware into the
adapter structure, the change_mtu entry point function can be cleaned
up of checks for all the different max frame sizes supported by
Signed-off-by: David S. Miller <davem@davemloft.net>
The flow control thresholds, i.e. high and low watermarks of the Rx
FIFO for when the hardware should transmit PAUSE frames (XON and XOFF,
respectively), need to be tuned for more efficient use of the FIFO.
The logic to set the thresholds for parts that support early-receive
(ERT) was also wrong in that it should check whether jumbo frames are
in use.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During reset, the driver was attempting to disable the Smart Powerdown
feature even if the part does not support Smart Powerdown. Check for
support before attempting to disable the feature.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CRC stripping should be enabled by default but was not if it was not
specified as a module parameter.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Second round of drivers for Gb cards (and NIU one I forgot in the 10GB round)
Now that core network takes care of trans_start updates, dont do it
in drivers themselves, if possible. Drivers can avoid one cache miss
(on dev->trans_start) in their start_xmit() handler.
Exceptions are NETIF_F_LLTX drivers
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The phy port status register has the MDI-X status bit on bit 11, not bit 3
as is currently setup in the define. This patch corrects that so the
correct bit is checked on igp PHY types.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
after the recent changes to wired drivers to use only
netif_carrier_off the driver can have outstanding tx work to
complete that will never complete once link is down. Since the
intel hardware will hold this tx work forever, the driver
notices a tx timeout condition internally and might try
to instigate printk and reset of the part with a
netif_stop_queue, which doesn't work because link is down.
Don't bother arming to tx hang detection when link is down.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It was pointed out that the Intel wired ethernet drivers do not need to
wake the tx queue since netif_carrier_on/off will take care of the qdisc
management in order to guarantee the correct handling of the transmit
routine enable state.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As reported by Andrew Lutomirski <amluto@gmail.com>
All the intel wired ethernet drivers were calling netif_carrier_off
and netif_stop_queue (or variants) before calling register_netdevice
This is incorrect behavior as was pointed out by davem, and causes
ifconfig and friends to report a strange state before first link
after the driver was loaded, since without a netif_carrier_off, the stack
assumes carrier_on, but before register_netdev, netlink messages are not
sent out telling link state.
This apparently confused *some* versions of networkmanager.
Andy tested this for e1000e and confirmed it was working for him.
see thread: http://marc.info/?l=linux-netdev&m=123946479705636&w=2
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andy Lutomirski <amluto@gmail.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the e1000e transmit cleanup inner loop exited early, then
cleaned might not be true. This could cause tx hangs or other
badness. Use count to track the total number of descriptors
cleaned instead of basing a tx queue restart off of a temporary
working state variable.
This code now makes the flow the same for e1000/e1000e/igb/ixgbe
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prevent e1000e from putting the adapter into D3 during shutdown except when
we're going to power off the system, since doing that may generally cause
problems with kexec to happen (such problems were observed for igb and
forcedeth). For this purpose seperate e1000e_shutdown() from e1000e_suspend()
and use the appropriate PCI PM callbacks in both of them.
Signed-off-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
e1000/e1000e compile report a possible unused variable, fix
that for now. Shortly after this a small refactor and bug
fix will follow in the same code.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32)
Signed-off-by: Yang Hongyang<yanghy@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Replace all DMA_64BIT_MASK macro with DMA_BIT_MASK(64)
Signed-off-by: Yang Hongyang<yanghy@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As noticed by Alan Cox, it is possible for e1000e to exit its interrupt
handler or NAPI with interrupts enabled even when the driver is unloading or
being configured administratively down.
fix related to fix for: http://bugzilla.kernel.org/show_bug.cgi?id=12876
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
CC: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
e1000e (and e1000, igb, ixgbe, ixgb) all do a series of operations each
time a multicast address is added. The flow goes something like
1) stack adds one multicast address
2) stack passes whole current list of unicast and multicast addresses to
driver
3) driver clears entire list in hardware
4) driver programs each multicast address using iomem in a loop
This was causing multicast packets to be lost during the reprogramming
process.
reference with test program:
http://kerneltrap.org/mailarchive/linux-netdev/2009/3/14/5160514/thread
Thanks to Dave Boutcher for his report and test program.
This driver fix prepares an array all at once in memory and programs it in
one shot to the hardware, not requiring an "erase" cycle. It would still
be possible for packets to be dropped while the receiver is off during
reprogramming.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
CC: Dave Boutcher <daveboutcher@gmail.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change updates the e1000e tx cleanup routine to more closely match
what already exists in igb and e1000.
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>
Add device ID for a new variant of the 82574 adapter.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Acked-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When merging into Jeff's tree:
commit 5f66f20806
Author: Arthur Jones <ajones@riverbed.com>
Date: Thu Mar 19 01:13:08 2009 +0000
e1000e: allow tx of pre-formatted vlan tagged packets
We lost one line, this fixes that missing
piece...
Signed-off-by: Arthur Jones <ajones@riverbed.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As with igb, when the e1000e driver is fed 802.1q
packets with hardware checksum on, it chokes with an
error of the form:
checksum_partial proto=81!
As the logic there was not smart enough to look into
the vlan header to pick out the encapsulated protocol.
There are times when we'd like to send these packets
out without having to configure a vlan on the interface.
Here we check for the vlan tag and allow the packet to
go out wiht the correct hardware checksum.
Thanks to Kand Ly <kand@riverbed.com> for discovering the
issue and the coming up with a solution. This patch is
based upon his work.
Fixups from Stephen Hemminger <shemminger@vyatta.com> and
Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Arthur Jones <ajones@riverbed.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Stephen Hemminger <shemminger@vyatta.com>
CC: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There were a few issues I noticed in e1000e. These include a double free
of the skb if mapping fails, and the fact that context descriptors appear
to be left in the descriptor ring after the failure.
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>
Add device ID and related support for 82583 mac.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Radheka Godse <radheka.godse@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is in reference to https://bugzilla.redhat.com/show_bug.cgi?id=484494
Also addresses issue show in kerneloops
The e1000e transmit code was calling pci_unmap_page on dma handles that it
might have called pci_map_single on.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During the move of support for PCIe devices from e1000 to e1000e, this
workaround necessary only for older non-PCIe devices was mistakenly
copied into e1000e. Remove it.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Impact: Make symbol static.
Fix this sparse warning:
drivers/net/e1000e/82571.c:1229:5: warning: symbol 'e1000_check_for_serdes_link_82571' was not declared. Should it be static?
Signed-off-by: Hannes Eder <hannes@hanneseder.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Single-thread access must be ensured for ICH8 NVM and PHY operations.
This synchronization is provided by the nvm_mutex. To assist in
understanding the contexts from which this code could be reached,
a WARN was output if the mutex was not going to be immediately
acquirable (if !mutex_trylock()). The code has now been optimized,
and we have verified that the few remaining mutex contentions are
reasonable and non-blocking, and it is time to remove the
mutex_trylock() and WARN messages.
Signed-off-by: dave graham <david.graham@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch addresses an issue where we did not restart auto-negotiation on
serdes links when the link partner was disabled and re-enabled. It includes
reworking the serdes link detect mechanism to be a state machine for
82571 and 82572 parts only.
Signed-off-by: dave graham <david.graham@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RXSEQ interrupts were used to force link state interrogation of serdes
links, as the Si was not guaranteed to report LSC interrupts when the
link changed state. On some bladeservers this resulted in false link up
reports if no link partner was connected. The RXSEQ treatment is
not necessary, as the link can be monitored from the watchdog timer, and
the false link indications cease.
Signed-off-by: dave graham <david.graham@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
82571 and 82572 Errata #13 documents that the Si feature DMA Dynamic
Clock Gating should be disabled, and identifies the workaround of
disabling the feature by EEPROM setting. EEPROM versions that do not
include the recommended workaround have been found in the field, and so
some customers remain at risk. Because the feature DMA Dynamic clock
Gating can be disabled by directly setting the appropriate bit in the
E1000_CTRL_EXT register, this patch overrides the EEPROM setting, and
force-disables the feature.
Signed-off-by: dave graham <david.graham@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
most if not all of the devices supported by e1000e support
AER (Advanced Error Reporting) so we attempt to register
with the OS that we know how to reset ourselves after
a fatal error.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cosmetic change to use struct e1000_mac_info.serdes_has_link
consistently as the 'bool' that it's declared as.
No functional change.
Signed-off-by: Alex Chiang <achiang@hp.com>
Acked-by: Jeff Kirsher <Jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Base versions handle constant folding now.
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a hardware errata in some revisions of the 82574 that needs
to be worked around in the driver by setting a register bit at init.
If this bit is not set A0 versions of the 82574 can generate
tx hangs.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
LLTX is deprecated and complicated, don't use it. It was observed by Don Ash
<donash4@gmail.com> that e1000e was acquiring this lock in the NAPI cleanup
path. This is obviously a bug, as this is a leftover from when e1000
supported multiple tx queues and fake netdevs.
another user reported this to us and tested routing with the 2.6.27 kernel and
this patch and reported a 3.5 % improvement in packets forwarded in a
multi-port test on 82571 parts.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that VLAN has GRO support as well, we can call its GRO handler
as well.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Following the removal of the unused struct net_device * parameter from
the NAPI functions named *netif_rx_* in commit 908a7a1, they are
exactly equivalent to the corresponding *napi_* functions and are
therefore redundant.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adds process name of the current mutex holder to the WARN message output
when the e1000e driver attempts to acquire the nvm_mutex and finds that
it is already being held. With this patch the WARN message indicates
both the process name of the current mutex holder and the process name of
the attempted acquisition, which together will help to identify the
contending codepaths.
Signed-off-by: David Graham <david.graham@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Device drivers that use pci_request_regions() (and similar APIs) have a
reasonable expectation that they are the only ones accessing their device.
As part of the e1000e hunt, we were afraid that some userland (X or some
bootsplash stuff) was mapping the MMIO region that the driver thought it
had exclusively via /dev/mem or via various sysfs resource mappings.
This patch adds the option for device drivers to cause their reserved
regions to the "banned from /dev/mem use" list, so now both kernel memory
and device-exclusive MMIO regions are banned.
NOTE: This is only active when CONFIG_STRICT_DEVMEM is set.
In addition to the config option, a kernel parameter iomem=relaxed is
provided for the cases where developers want to diagnose, in the field,
drivers issues from userspace.
Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Fix this sparse warnings:
drivers/net/e1000e/es2lan.c:1265:5: warning: symbol 'e1000_read_kmrn_reg_80003es2lan' was not declared. Should it be static?
drivers/net/e1000e/es2lan.c:1298:5: warning: symbol 'e1000_write_kmrn_reg_80003es2lan' was not declared. Should it be static?
Signed-off-by: Hannes Eder <hannes@hanneseder.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the napi api was changed to separate its 1:1 binding to the net_device
struct, the netif_rx_[prep|schedule|complete] api failed to remove the now
vestigual net_device structure parameter. This patch cleans up that api by
properly removing it..
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds GRO support to e1000e by making it invoke napi_gro_receive
instead of netif_receive_skb.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
During a reset, releasing the swflag after it failed to be acquired would
cause a double unlock of the mutex. Instead, test whether acquisition of
the swflag was successful and if not, do not release the swflag. The reset
must still be done to bring the device to a quiescent state.
This resolves [BUG 12200] BUG: bad unlock balance detected! e1000e
http://bugzilla.kernel.org/show_bug.cgi?id=12200
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
