So that it is also an offset from skb->head, reduces its size from 8 to 4 bytes
on 64bit architectures, allowing us to combine the 4 bytes hole left by the
layer headers conversion, reducing struct sk_buff size to 256 bytes, i.e. 4
64byte cachelines, and since the sk_buff slab cache is SLAB_HWCACHE_ALIGN...
:-)
Many calculations that previously required that skb->{transport,network,
mac}_header be first converted to a pointer now can be done directly, being
meaningful as offsets or pointers.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ip_hdrlen() buddy, created to reduce the number of skb->h.th-> uses and to
avoid the longer, open coded equivalent.
Ditched a no-op in bnx2 in the process.
I wonder if we should have a BUG_ON(skb->h.th->doff < 5) in tcp_optlen()...
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the quite common 'skb->h.raw - skb->data' sequence.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now the skb->nh union has just one member, .raw, i.e. it is just like the
skb->mac union, strange, no? I'm just leaving it like that till the transport
layer is done with, when we'll rename skb->mac.raw to skb->mac_header (or
->mac_header_offset?), ditto for ->{h,nh}.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the quite common 'skb->nh.raw - skb->data' sequence.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 60cba200f1. It's been
linked to lockups of the e1000 hardware, see for example
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=229603
but it's likely that the commit itself is not really introducing the
bug, but just allowing an unrelated problem to rear its ugly head (ie
one current working theory is that the code exposes us to a hardware
race condition by decreasing the amount of time we spend in each NAPI
poll cycle).
We'll revert it until root cause is known. Intel has a repeatable
reproduction on two different machines and bus traces of the hardware
doing something bad.
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@suse.de>
Cc: Dave Jones <davej@redhat.com>
Cc: Auke Kok <auke-jan.h.kok@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch splits the vlan_group struct into a multi-allocated struct. On
x86_64, the size of the original struct is a little more than 32KB, causing
a 4-order allocation, which is prune to problems caused by buddy-system
external fragmentation conditions.
I couldn't just use vmalloc() because vfree() cannot be called in the
softirq context of the RCU callback.
Signed-off-by: Dan Aloni <da-x@monatomic.org>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit d2ed16356f.
As Thomas Gleixner reports:
"e1000 is not working anymore. ifup fails permanentely.
ADDRCONF(NETDEV_UP): eth0: link is not ready
nothing else"
The broken commit was identified with "git bisect".
Auke Kok says:
"I think we need to drop this now. The report that says that this
*fixes* something might have been on regular interrupts only. I
currently suspect that it breaks all MSI interrupts, which would make
sense if I look a the code. Very bad indeed."
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Auke Kok <auke-jan.h.kok@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now that 2.6.19 provides a proper implementation that saves MSI, PCI-E
config space, we can have the e1000 driver use those instead of it's
custom implementation.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Remove the NETIF_F_TSO #ifdef-ery in drivers/net; this was
for old-old-2.4 compat (even current 2.4 has NETIF_F_TSO)
but it's time to get rid of it by now.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
The driver was still mis-calculating the number of bytes sent during
transmit, now the driver computes what appears to be exactly 100%
correct byte counts (not including CRC) when figuring out how many
bytes and frames were sent during the current transmit packet.
Since the driver sets the IP checksum insertion bit (IXSM in Status
field) in transmit context descriptors, it should clear the IP checksum
bits of any garbage so as not to confuse the hardware.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Print RX/TX flow control setting at link up time to display the
actual link FC properties instead of the advertised values.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
This fix attempts to solve a customer (IBM) reported issue with NAPI
enabled e1000 having bad performance when transmitting simultaneously
on four ports. The issue comes down to an interaction between NAPI,
hardware interrupt balancing, and the driver rescheduling poll on
the same processor. Try to fix by allowing the driver to re-enable
interrupts sooner instead of polling one more time, when there was
recently all the work completed in cleanup.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Unfortunately the read-free MSI interrupt handler needs to flush write
the icr register and thus we can't be read-free. Our MSI irq routine
thus becomes a lot more simpler since we don't need to track link state
anymore.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
This reverts commit 72f3ab7462, which was
superceded by commit 683a2aa339
("e1000: Do not truncate TSO TCP header with 82544 workaround"), which
fixed the real problem.
Signed-off-by: Jeff Garzik <jeff@garzik.org>
The e1000 driver has a workaround for 82544 on PCI-X where if the
terminating byte of a buffer is at addresses 0-3 mod 8, then 4 bytes
are shaved off it and defered to a new segment. This is due to an
erratum that could otherwise cause TX hangs.
Unfortunately this breaks TSO because it may cause the TCP header to
be split over two segments which itself causes TX hangs. The solution
is to pull 4 bytes of data up from the next segment rather than pushing
4 bytes off. This ensures the TCP header remains in one piece and
works around the PCI-X hang.
This patch is based on one from Jesse Brandeburg.
This bug has been trigered by both CONFIG_DEBUG_SLAB as well as Xen.
Note that the only reason we don't see this normally is because the
TCP stack starts writing from the end, i.e., it writes the TCP header
first then slaps on the IP header, etc. So the end of the TCP header
(skb->tail - 1 here) is always aligned correctly.
Had we made the start of the IP header (e.g., IPv6) 8-byte aligned
instead, this would happen for normal TCP traffic as well.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
--
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
--
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Currently after an interface up, the link state is detected 2 seconds later
when the first watchdog timer runs. This patch changes that by triggering
the hardware to generate a link-change interrupt from the up() function
instead. This has the result that the link state gets detected immediately
and without races. This has the potential to speed up booting since a normal
distribution boot process waits for a link before DHCP is attempted.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Add 3 extra packet redirect counters for tracking purposes to make sure
we can test that all packets arrive properly.
Originally from Jesse Brandeburg <jesse.brandeburg@intel.com>,
rewritten to use feature flags by me.
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Allow the user to vary the size that copybreak works. Currently cb is enabled
for packets < 256 bytes, but various tests indicate that this should be
configurable for specific use cases. In addition, this parameter allows us
to force never/always during testing to get full and predictable coverage of
both code paths.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Assign the PBA to be large enough to contain at least 2 jumbo frames on
all adapters. This dramatically increases performance on several adapters
and fixes TX performance degradation issues where the PBA was misallocated
in the old algorithm.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
the driver has (ancient) code for messing with TIPG from the 82542 days.
Unfortunately this code was running on our current adapters and setting
TIPG for fiber to be +1 over the copper value. This caused 1.45Mpps
to be sent instead of 1.487Mpps.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
This bugfix makes sure that the driver data reflects the full new situation
before the adapter is reinitialized.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
In rare occasions, ESB2 systems would end up started without the RX
unit being turned on. Add a check that runs post-init to work around
this issue.
Originally from Jesse Brandeburg <jesse.brandeburg@intel.com>,
rewritten to use feature flags by me.
Signed-off-by: Jeff Garzik <jeff@garzik.org>
CONFIG_DEBUG_SLAB changes alignments of the data structures the slab
allocators return. These break certain workarounds for TSO on the 82544.
Since DEBUG_SLAB is relatively rare and not used for performance sensitive
cases, the simplest fix is to disable TSO in this special situation.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
If the user has forced gigabit speed, phy power management must be disabled;
otherwise the NIC would try to negotiate to a linkspeed of 10/100 mbit on
shutdown, which would lead to a total loss of link. This loss of link breaks
Wake-on-Lan and IPMI.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Several bugs existed in how we handle manageability issues all
over the driver. This patch consolidates all the managability
release and init code in two single functions and call them from
appropriate locations. This fixes several BMC packet redirect issues
and powerup/down hiccups.
Originally from Jesse Brandeburg <jesse.brandeburg@intel.com>, rewritten
to use feature flags by me.
Signed-off-by: Jeff Garzik <jeff@garzik.org>
The 82543 chip does not count tx_carrier_errors properly in FD mode;
report zeros instead of garbage.
Originally from Jesse Brandeburg <jesse.brandeburg@intel.com>, rewritten
to use feature flags by me.
Signed-off-by: Jeff Garzik <jeff@garzik.org>
The dynamic interrupt rate control patches omitted proper counting
for jumbo's and TSO resulting in suboptimal interrupt mitigation strategies.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
... into anonymous union of __wsum and __u32 (csum and csum_offset resp.)
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
ia64:
drivers/built-in.o(.text+0xd9a72): In function `e1000_xmit_frame':
: undefined reference to `csum_ipv6_magic'
Cc: Auke Kok <auke-jan.h.kok@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Add a new dynamic itr algorithm, with 2 modes, and make it the default
operation mode. This greatly reduces latency and increases small packet
performance, at the "cost" of some CPU utilization. Bulk traffic
throughput is unaffected.
The driver can limit the amount of interrupts per second that the
adapter will generate for incoming packets. It does this by writing a
value to the adapter that is based on the maximum amount of interrupts
that the adapter will generate per second.
Setting InterruptThrottleRate to a value greater or equal to 100 will
program the adapter to send out a maximum of that many interrupts per
second, even if more packets have come in. This reduces interrupt
load on the system and can lower CPU utilization under heavy load,
but will increase latency as packets are not processed as quickly.
The default behaviour of the driver previously assumed a static
InterruptThrottleRate value of 8000, providing a good fallback value
for all traffic types,but lacking in small packet performance and
latency. The hardware can handle many more small packets per second
however, and for this reason an adaptive interrupt moderation algorithm
was implemented.
Since 7.3.x, the driver has two adaptive modes (setting 1 or 3) in
which it dynamically adjusts the InterruptThrottleRate value based on
the traffic that it receives. After determining the type of incoming
traffic in the last timeframe, it will adjust the InterruptThrottleRate
to an appropriate value for that traffic.
The algorithm classifies the incoming traffic every interval into
classes. Once the class is determined, the InterruptThrottleRate
value is adjusted to suit that traffic type the best. There are
three classes defined: "Bulk traffic", for large amounts of packets
of normal size; "Low latency", for small amounts of traffic and/or
a significant percentage of small packets; and "Lowest latency",
for almost completely small packets or minimal traffic.
In dynamic conservative mode, the InterruptThrottleRate value is
set to 4000 for traffic that falls in class "Bulk traffic". If
traffic falls in the "Low latency" or "Lowest latency" class, the
InterruptThrottleRate is increased stepwise to 20000. This default
mode is suitable for most applications.
For situations where low latency is vital such as cluster or
grid computing, the algorithm can reduce latency even more when
InterruptThrottleRate is set to mode 1. In this mode, which operates
the same as mode 3, the InterruptThrottleRate will be increased
stepwise to 70000 for traffic in class "Lowest latency".
Setting InterruptThrottleRate to 0 turns off any interrupt moderation
and may improve small packet latency, but is generally not suitable
for bulk throughput traffic.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Cc: Rick Jones <rick.jones2@hp.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Add a generic MSI interrupt routine that is IO read-free, speeding up
MSI interrupt handling.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Spec fix: don't set IDE unless we are actually setting the tx
int delay time.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Add a netif_wake/start_queue counter to the ethtool statistics to indicated
to the user that their transmit ring could be too small for their workload.
Signed-off-by: Jesse brandeburg <jesse.brandeburg@intel.com>
Cc: Jamal Hadi <hadi@cyberus.ca>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Add support for a Low Profile quad-port PCI-E adapter and 2 variants
of the ICH8 systems' onboard NIC's.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
This memsetting was added in a paranoid rage debugging TX hangs, but
are no longer of importance. We can beef up the performance quite a
bit removing them. Make sure to fill in next_to_watch to allow this.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
IA64 SMP systems were seeing TX issues with multiple cpu's attempting
to write tail registers unordered. This mmiowb() fixes the issue.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Enable early receives on 82573 for jumbo frame performance. Jumbo's
are only supported on 82573L with ASPM disabled.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Enable TSO for IPV6. All e1000 hardware supports it. This reduces CPU
utilizations by 50% when transmitting IPv6 frames.
Fix symbol naming enabling ipv6 TSO. Turn off TSO6 for 10/100.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Remove debugging code disabling MULR (multiple reads). It's not usable
for a wide audience and there are no known problems with MULR right
now.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Pass the work_struct pointer to the work function rather than context data.
The work function can use container_of() to work out the data.
For the cases where the container of the work_struct may go away the moment the
pending bit is cleared, it is made possible to defer the release of the
structure by deferring the clearing of the pending bit.
To make this work, an extra flag is introduced into the management side of the
work_struct. This governs auto-release of the structure upon execution.
Ordinarily, the work queue executor would release the work_struct for further
scheduling or deallocation by clearing the pending bit prior to jumping to the
work function. This means that, unless the driver makes some guarantee itself
that the work_struct won't go away, the work function may not access anything
else in the work_struct or its container lest they be deallocated.. This is a
problem if the auxiliary data is taken away (as done by the last patch).
However, if the pending bit is *not* cleared before jumping to the work
function, then the work function *may* access the work_struct and its container
with no problems. But then the work function must itself release the
work_struct by calling work_release().
In most cases, automatic release is fine, so this is the default. Special
initiators exist for the non-auto-release case (ending in _NAR).
Signed-Off-By: David Howells <dhowells@redhat.com>
Arnaldo Carvalho de Melo