Fix transport header size
Fix the transpoert header size for UDP packets.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I was recently shown that vmxnet3 devices on transmit, will drop very small udp
frames consistently. This is due to a regression introduced by commit
39d4a96fd7. This commit attempts to introduce an
optimization to the tx path, indicating that the underlying hardware behaves
optimally when at least 54 bytes of header data are available for direct access.
This causes problems however, if the entire frame is less than 54 bytes long.
The subsequent pskb_may_pull in vmxnet3_parse_and_copy_hdr fails, causing an
error return code, which leads to vmxnet3_tq_xmit dropping the frame.
Fix it by placing a cap on the copy length. For frames longer than 54 bytes, we
do the pull as we normally would. If the frame is shorter than that, copy the
whole frame, but no more. This ensures that we still get the optimization for
qualifying frames, but don't do any damange for frames that are too short.
Also, since I'm unable to do this, it wuold be great if vmware could follow up
this patch with some additional code commentary as to why 54 bytes is an optimal
pull length for a virtual NIC driver. The comment that introduced this was
vague on that. Thanks!
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Max Matveev <mmatveev@redhat.com>
CC: Max Matveev <mmatveev@redhat.com>
CC: "David S. Miller" <davem@davemloft.net>
CC: Shreyas Bhatewara <sbhatewara@vmware.com>
CC: "VMware, Inc." <pv-drivers@vmware.com>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All tables of function pointers should be const to make hacks
more difficult. Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All drivers that support modification of the RX flow hash indirection
table initialise it in the same way: RX rings are assigned to table
entries in rotation. Make that default policy explicit by having them
call a ethtool_rxfh_indir_default() function.
In the ethtool core, add support for a zero size value for
ETHTOOL_SRXFHINDIR, which resets the table to this default.
Partly-suggested-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a new ethtool operation (get_rxfh_indir_size) to get the
indirectional table size. Use this to validate the user buffer size
before calling get_rxfh_indir or set_rxfh_indir. Use get_rxnfc to get
the number of RX rings, and validate the contents of the new
indirection table before calling set_rxfh_indir. Remove this
validation from drivers.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Let caller know the result of adding/removing vlan id to/from vlan
filter.
In some drivers I make those functions to just return 0. But in those
where there is able to see if hw setup went correctly, return value is
set appropriately.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Per discussion with Ben Hutchings and David Miller, go through and
remove assignments of "N/A" to fw_version in various drivers'
.get_drvinfo routines. While there clean-up some use of bare
constants and such.
Signed-off-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
v2: add couple missing conversions in drivers
split unexporting netdev_fix_features()
implemented %pNF
convert sock::sk_route_(no?)caps
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
The device.h header was including module.h, making it present for
most of these drivers. But we want to clean that up. Call out the
include of module.h in the modular network drivers.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
To ease skb->truesize sanitization, its better to be able to localize
all references to skb frags size.
Define accessors : skb_frag_size() to fetch frag size, and
skb_frag_size_{set|add|sub}() to manipulate it.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
vmxnet3 allocates a page per skb fragment. We must account
PAGE_SIZE increments on skb->truesize, not the actual frag length.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When I converted some drivers from pci_map_page to skb_frag_dma_map I
neglected to convert PCI_DMA_xDEVICE into DMA_x_DEVICE and
pci_dma_mapping_error into dma_mapping_error.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Shreyas Bhatewara <sbhatewara@vmware.com>
Cc: "VMware, Inc." <pv-drivers@vmware.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Correct the description of ethtool_rxnfc::rule_locs; it is an array
of currently used locations, not all possible valid locations.
Add note that drivers must not use ethtool_rxnfc::rule_locs.
The rule_locs argument to ethtool_ops::get_rxnfc is either NULL or a
pointer to an array of u32, so change the parameter type accordingly.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The vmxnet3 driver enables vlan filters if filtering is enabled for
any vlan. In promiscuous mode the filter table is cleared to in
order to disable filtering. However, if a vlan device is subsequently
created that vlan will be added to the filter, re-engaging it. As a
result, not only do we not see all the vlans in promiscuous mode, we
don't even see vlans for which a filter was previously created.
CC: Scott J. Goldman <scottjg@vmware.com>
CC: Shreyas Bhatewara <sbhatewara@vmware.com>
CC: VMware PV-Drivers <pv-drivers@vmware.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NETIF_F_HIGHDMA is being disabled even when dma64 is true. This patch fixes it.
CC: Michal Miroslaw <mirq-linux@rere.qmqm.pl>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Parent device for netdev should be set before netdev_info() can be called
otherwise there is a NULL pointer dereference and probe() fails.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: Scott J. Goldman <scottjg@vmware.com>--
Signed-off-by: David S. Miller <davem@davemloft.net>
vmxnet3 device supports only power-of-two number of queues. The driver
therefore needs to check this and rounds down the number of queues to the
nearest power of two.
Signed-off-by: Yong Wang <yongwang@vmware.com>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
If the rx ring is completely empty, then the device may never fire an rx
interrupt. Unfortunately, the rx interrupt is what triggers populating the
rx ring with fresh buffers, so this will cause networking to lock up.
This patch replenishes the skb in recv descriptor as soon as it is
peeled off while processing rx completions. If the skb/buffer
allocation fails, existing one is recycled and the packet in hand is
dropped. This way none of the RX desc is ever left empty, thus avoiding
starvation
Signed-off-by: Scott J. Goldman <scottjg@vmware.com>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When receiving packets from another guest on the same hypervisor, it's
generally possible to receive large packets because no segmentation is
necessary and these packets are handled by LRO. However, when doing
routing or bridging we must disable LRO and lose this benefit. In
these cases GRO can still be used and it is very effective because the
packets which are segmented in the hypervisor are received very close
together and can easily be merged.
CC: Shreyas Bhatewara <sbhatewara@vmware.com>
CC: Scott Goldman <scottjg@vmware.com>
CC: VMware PV-Drivers <pv-drivers@vmware.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Scott J. Goldman <scottjg@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This converts the vmxnet3 driver to use the new vlan model. In doing so
it fixes missing tags in tcpdump and failure to do checksum offload when
tx vlan offload is disabled.
CC: Shreyas Bhatewara <sbhatewara@vmware.com>
CC: VMware PV-Drivers <pv-drivers@vmware.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Scott J. Goldman <scottjg@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While doing some backporting I noticed that vmxnet3 had a variable that was set
but never used. Get rid of it, and stop the compiler from griping
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: "VMware, Inc." <pv-drivers@vmware.com>
CC: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: David S. Miller <davem@conan.davemloft.net>
Convert vmxnet3 driver to 64 bit statistics interface.
This driver was already counting packet per queue in a 64 bit value so not
a huge change. Eliminate unused old net_device_stats structure.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Scott J. Goldman <scottjg@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During initialization of vmxnet3, the state of LRO
gets out of sync with netdev->features.
This leads to very poor TCP performance in a IP forwarding
setup and is hitting many VMware users.
Simplified call sequence:
1. vmxnet3_declare_features() initializes "adapter->lro" to true.
2. The kernel automatically disables LRO if IP forwarding is enabled,
so vmxnet3_set_flags() gets called. This also updates netdev->features.
3. Now vmxnet3_setup_driver_shared() is called. "adapter->lro" is still
set to true and LRO gets enabled again, even though
netdev->features shows it's disabled.
Fix it by updating "adapter->lro", too.
The private vmxnet3 adapter flags are scheduled for removal
in net-next, see commit a0d2730c95
"net: vmxnet3: convert to hw_features".
Patch applies to 2.6.37 / 2.6.38 and 2.6.39-rc6.
Please CC: comments.
Signed-off-by: Thomas Jarosch <thomas.jarosch@intra2net.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Resending this patch with few changes.
Avoid multiple queues when MSI or MSI-X not available
Limit number of Tx queues to 1 if MSI/MSI-X support is not configured in
the kernel. This will make number of tx and rx queues equal when MSI/X
is not configured thus providing better performance.
Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This updates the network drivers so that they don't access the
ethtool_cmd::speed field directly, but use ethtool_cmd_speed()
instead.
For most of the drivers, these changes are purely cosmetic and don't
fix any problem, such as for those 1GbE/10GbE drivers that indirectly
call their own ethtool get_settings()/mii_ethtool_gset(). The changes
are meant to enforce code consistency and provide robustness with
future larger throughputs, at the expense of a few CPU cycles for each
ethtool operation.
All drivers compiled with make allyesconfig ion x86_64 have been
updated.
Tested: make allyesconfig on x86_64 + e1000e/bnx2x work
Signed-off-by: David Decotigny <decot@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This also removes private feature flags that were always set to true.
You may want to move vmxnet3_set_features() to vmxnet3_drv.c as a following
cleanup.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit d5dbda2380 "ethtool: Add
support for vlan accleration.", drivers that have NETIF_F_HW_VLAN_TX,
and/or NETIF_F_HW_VLAN_RX feature, but do not allow enable/disable vlan
acceleration via ethtool set_flags, always return -EINVAL from that
function. Fix by returning -EINVAL only if requested features do not
match current settings and can not be changed by driver.
Change any driver that define ethtool->set_flags to use
ethtool_invalid_flags() to avoid similar problems in the future
(also on drivers that do not have the problem).
Tested with modified (to reproduce this bug) myri10ge driver.
Cc: stable@kernel.org # 2.6.37+
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of single tx and rx queues, three MSI-x vectors are allocated instead
of two. This patch fixes that.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Access to cmd register is racey, especially in smp environments. Protect
it using a spinlock.
Signed-off-by: Matthieu Bucchianeri <matthieu@vmware.com>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a small possibility of a race where the suspend routine gets
called, while a napi callback is still pending and when that comes up,
it enables interrupts which just got disabled in the suspend routine.
This change adds napi disable call in suspend and enable in resume to
avoid race.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Acked-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Show per-queue stats in ethtool -S output for vmxnet3 interface. Register dump
of ethtool should dump registers for all tx and rx queues.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a performance enhancement fix. vmxnet3 device performs better when
provided with at least 54 bytes (ethernet 14 + IP 20+ TCP 20) in the first SG
buffer. For UDP packets driver provides lesser than that in first sg. This
change fixes the same. Also avoid the redundant pskb_may_pull() call.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make hw vlan tag stripping as enabled by default. Thereby remove
the code to conditionally enable it later.
Signed-off-by: Guolin Yang <gyang@vmware.com>
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While activating the device get it's MAC address from netdev. This will allow
the MAC address configured using ifconfig to persist through the reset.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix a bug while changing ring size when MTU is changed.
Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Acked-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using static const generally increases object text and decreases data size.
It also generally decreases overall object size.
Signed-off-by: Joe Perches <joe@perches.com>
There were several paths that didn't release their locks.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some drivers are using skb_transport_offset(skb) instead of skb->csum_start
for NETIF_F_HW_CSUM offload. This does not matter now, but if someone
implements checksumming of encapsulated packets then this will break silently.
TSO output paths are left as they are, since they are for IP+TCP only
(might be worth converting though).
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
flush_scheduled_work() is on its way out. This patch contains simple
conversions to replace flush_scheduled_work() usage with direct
cancels and flushes.
Directly cancel the used works on driver detach and flush them in
other cases.
The conversions are mostly straight forward and the only dangers are,
* Forgetting to cancel/flush one or more used works.
* Cancelling when a work should be flushed (ie. the work must be
executed once scheduled whether the driver is detaching or not).
I've gone over the changes multiple times but it would be much
appreciated if you can review with the above points in mind.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jay Cliburn <jcliburn@gmail.com>
Cc: Michael Chan <mchan@broadcom.com>
Cc: Divy Le Ray <divy@chelsio.com>
Cc: e1000-devel@lists.sourceforge.net
Cc: Vasanthy Kolluri <vkolluri@cisco.com>
Cc: Samuel Ortiz <samuel@sortiz.org>
Cc: Lennert Buytenhek <buytenh@wantstofly.org>
Cc: Andrew Gallatin <gallatin@myri.com>
Cc: Francois Romieu <romieu@fr.zoreil.com>
Cc: Ramkrishna Vepa <ramkrishna.vepa@exar.com>
Cc: Matt Carlson <mcarlson@broadcom.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Cc: Shreyas Bhatewara <sbhatewara@vmware.com>
Cc: netdev@vger.kernel.org
If RSS is disabled, we can ifdef out some RSS specific code. This fixes
the compile error found by Randy Dunlap.
Signed-off-by: Scott J. Goldman <scottjg@vmware.com>
Reviewed-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add multiqueue support to vmxnet3 driver
This change adds multiqueue and thus receive side scaling support
to vmxnet3 device driver. Number of rx queues is limited to 1 in cases
where MSI is not configured or one MSIx vector is not available per rx
queue
Signed-off-by: Shreyas Bhatewara <sbhatewara@vmware.com>
Reviewed-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
