dpdk.git
9 years agoi40e/base: rename err to perrno
Helin Zhang [Thu, 30 Apr 2015 15:03:10 +0000 (23:03 +0800)]
i40e/base: rename err to perrno

To be consistent with the original base driver, the variable name
of 'err' should be renamed to 'perrno'.

Test report: http://www.dpdk.org/ml/archives/dev/2015-May/017384.html

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
Tested-by: Min Cao <min.cao@intel.com>
9 years agoi40e/base: adjust register definitions and relevant
Helin Zhang [Thu, 30 Apr 2015 15:03:09 +0000 (23:03 +0800)]
i40e/base: adjust register definitions and relevant

Some macros of register definitions or relevant are added, modified
or deleted. In detail, they are as follows.
- I40E_PRTDCB_RUPTQ
- I40E_GLGEN_GPIO_CTL
- I40E_GLGEN_MDIO_CTRL
- I40E_GLGEN_RSTENA_EMP
- I40E_GLPCI_LATCT
- I40E_GLTPH_CTRL
- I40E_GLPRT_BPRCH
- I40E_GLPRT_TDPC
- I40E_GLSCD_QUANTA
Also reading the register of I40E_GLPRT_TDPC is deleted as its
definition is deleted.

Test report: http://www.dpdk.org/ml/archives/dev/2015-May/017384.html

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
Tested-by: Min Cao <min.cao@intel.com>
9 years agoi40e: update copyright
Helin Zhang [Thu, 30 Apr 2015 15:03:07 +0000 (23:03 +0800)]
i40e: update copyright

Copyright is updated.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoi40e: disable setting of PHY configuration
Helin Zhang [Thu, 30 Apr 2015 15:03:08 +0000 (23:03 +0800)]
i40e: disable setting of PHY configuration

There was a known link issue on 40G ports on NVM version (FVL3E),
when setting phy configuration. As a workaround, setting of phy
configuration should be disabled. The impact is that the link cannot
be forcedly configured, which doesn't affect any feature functions.
The workaround can be removed when a formal fix is ready later.

Test report: http://www.dpdk.org/ml/archives/dev/2015-May/017384.html

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
Tested-by: Min Cao <min.cao@intel.com>
9 years agomk: introduce EXTRA_LDLIBS variable
Keith Wiles [Thu, 14 May 2015 14:21:32 +0000 (09:21 -0500)]
mk: introduce EXTRA_LDLIBS variable

Signed-off-by: Keith Wiles <keith.wiles@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agomk: simplify conditions for libs to link
Keith Wiles [Thu, 14 May 2015 14:21:31 +0000 (09:21 -0500)]
mk: simplify conditions for libs to link

Simplify the ifdefs in rte.app.mk to make the code more
readable and maintainable by introducing a internal
_LDLIBS-y variable to build up the LDLIBS variable.

The new internal variable _LDLIBS-y should not be
used outside of the rte.app.mk file.

Signed-off-by: Keith Wiles <keith.wiles@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agomaintainers: claim responsibility for docs
John McNamara [Wed, 13 May 2015 13:08:28 +0000 (14:08 +0100)]
maintainers: claim responsibility for docs

Signed-off-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agovhost: fix virtio freeze due to missed interrupt
Huawei Xie [Wed, 29 Apr 2015 11:11:24 +0000 (19:11 +0800)]
vhost: fix virtio freeze due to missed interrupt

Update of used->idx and read of avail->flags could be reordered.
Memory fence should be used to ensure the order, otherwise guest could see
a stale used->idx value after it toggles the interrupt suppression flag.
After guest sets the interrupt suppression flag, it will check if there
is more buffer to process through used->idx. If it sees a stale value,
it will exit the processing while host won't send interrupt to guest.

Signed-off-by: Huawei Xie <huawei.xie@intel.com>
Reviewed-by: Luke Gorrie <luke@snabb.co>
9 years agoeal/linux: fix irq handling with igb_uio
Stephen Hemminger [Tue, 28 Apr 2015 16:36:38 +0000 (09:36 -0700)]
eal/linux: fix irq handling with igb_uio

The introduction of uio_pci_generic broke interrupt handling with
igb_uio. The igb_uio device uses the kernel read/write method to
enable disable IRQ's; the uio_pci_generic has to use PCI intx
config read/write to enable disable interrupts.

Since igb_uio uses MSI-X the PCI intx config read/write won't
work.

Fixes: c112df6875a5 ("eal/linux: toggle interrupt for uio_pci_generic")

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agomk: remove NO_AUTOLIBS option
Keith Wiles [Fri, 1 May 2015 14:29:47 +0000 (09:29 -0500)]
mk: remove NO_AUTOLIBS option

NO_AUTOLIBS is not required as it was not used or defined in the config files.

Signed-off-by: Keith Wiles <keith.wiles@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agoscripts: fix build in a symlinked directory
Olivier Matz [Tue, 5 May 2015 09:00:09 +0000 (11:00 +0200)]
scripts: fix build in a symlinked directory

The script relpath.sh returns the relative path of the first directory
from the second directory. It is used to generate relative symlinks,
which can be useful if the build directory is embedded in the dpdk
directory: the whole dpdk can be moved without breaking the links,
which is helpful for an installation.

In case the build directory is a symlink, the script was not generating
the proper relative path. Fix this by calling "readlink -f" on the
arguments.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Tested-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
9 years agopcap: fix storage of name and type in queues
Konstantin Ananyev [Thu, 23 Apr 2015 14:32:48 +0000 (15:32 +0100)]
pcap: fix storage of name and type in queues

pcap_rx_queue/pcap_tx_queue should store it's own copy of name/type values,
not the pointer to temporary allocated space.

Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: John Mcnamara <john.mcnamara@intel.com>
9 years agoexamples/vhost: fix build with debug enabled
Ouyang Changchun [Mon, 4 May 2015 07:15:19 +0000 (15:15 +0800)]
examples/vhost: fix build with debug enabled

Minor fix for the referring of a pointer when debug and dump is enabled.

Fixes: 72ec8d77ac68 ("examples/vhost: rework duplicated code")

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
9 years agovfio: set eventfd non-blocking and not inherited
Stephen Hemminger [Wed, 29 Apr 2015 15:15:45 +0000 (08:15 -0700)]
vfio: set eventfd non-blocking and not inherited

Set internal event file descriptor to be non-block and not
inherited across exec.  This prevents accidental hangs and
passing in another thread.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
9 years agoapps: fix default mbuf size
Konstantin Ananyev [Wed, 29 Apr 2015 23:31:51 +0000 (00:31 +0100)]
apps: fix default mbuf size

Latest mbuf changes (priv_size addition and related fixes)
exposed small problem with testpmd and few other sample apps:
when mbuf size is exaclty 2KB or less, that causes
ixgbe PMD to select scattered RX even for configs with 'normal'
max packet length (max_rx_pkt_len == ETHER_MAX_LEN).
To overcome that problem and unify the code, new macro was created
to represent recommended minimal buffer length for mbuf.
When appropriate, samples are updated to use that macro.

Fixes: dfb03bbe2b ("app/testpmd: use standard functions to initialize
mbufs and mbuf pool")

Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agoeal/bsd: fix build on FreeBSD
Bruce Richardson [Tue, 5 May 2015 12:47:34 +0000 (13:47 +0100)]
eal/bsd: fix build on FreeBSD

Compilation on FreeBSD with clang was broken, giving the error message:

lib/librte_eal/bsdapp/eal/eal_pci.c:438:16: fatal error: assigning to
  'struct rte_pci_id *' from 'const struct rte_pci_id *' discards qualifiers
  [-Wincompatible-pointer-types-discards-qualifiers]
    for (id_table = dr->id_table ; id_table->vendor_id != 0; id_table++) {
                  ^ ~~~~~~~~~~~~

This patch fixes the issue by adding "const" to the type of id_table.

Fixes: 6065355a03fc ("pci: make device id tables const")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agokni: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:52 +0000 (14:03 +0100)]
kni: remove inclusion of mbuf header

The kni header file includes the mbuf header file, but it does not
need to do so as it only uses pointers to the struct rte_mbuf type, and
does not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agopipeline: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:54 +0000 (14:03 +0100)]
pipeline: remove inclusion of mbuf header

The pipeline header file includes the mbuf header file, but it does not
need to do so as it only uses pointers to the struct rte_mbuf type, and
does not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agotable: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:56 +0000 (14:03 +0100)]
table: remove inclusion of mbuf header

The rte_table header file includes the mbuf header file, but it does not
need to do so as it only uses pointers to the struct rte_mbuf type, and
does not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agodistributor: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:50 +0000 (14:03 +0100)]
distributor: remove inclusion of mbuf header

The distributor header file includes the mbuf header file, but it does not
need to do so as it only uses pointers to the struct rte_mbuf type, and
does not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agoip_frag: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:53 +0000 (14:03 +0100)]
ip_frag: remove inclusion of mbuf header

The ip_frag header file includes the mbuf header file, but it does not
need to do so as it only uses pointers to the struct rte_mbuf type, and
does not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agovhost: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:57 +0000 (14:03 +0100)]
vhost: remove inclusion of mbuf header

The virtio_net header file includes the mbuf header file, but it does not
need to do so as it only uses pointers to the struct rte_mbuf type, and
does not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agovirtio: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:55 +0000 (14:03 +0100)]
virtio: remove inclusion of mbuf header

The virtqueue header files include the mbuf header file, but they do not
need to do so as they only use pointers to the struct rte_mbuf type, and
do not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agoethdev: remove inclusion of mbuf header
Bruce Richardson [Thu, 23 Apr 2015 13:03:51 +0000 (14:03 +0100)]
ethdev: remove inclusion of mbuf header

The ethdev header file includes the mbuf header file, but it does not
need to do so as it only uses pointers to the struct rte_mbuf type, and
does not use any of the mbuf internals, nor any of the mbuf functions or
macros. Therefore the inclusion is unnecessary, and can be replaced by a
forward declaration of the mbuf type.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agoenic: add device ids
Sujith Sankar [Thu, 7 May 2015 03:49:09 +0000 (09:19 +0530)]
enic: add device ids

This patch adds support for enic in the nic_uio driver so that enic
could be used on FreeBSD.

Signed-off-by: Sujith Sankar <ssujith@cisco.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
9 years agokni: fix vhost build with kernels 3.19 and 4.0
Pablo de Lara [Tue, 5 May 2015 14:08:00 +0000 (15:08 +0100)]
kni: fix vhost build with kernels 3.19 and 4.0

Due to commit c0371da6 in kernel 3.19, which removed msg_iov
and msg_iovlen from struct msghdr, DPDK would not build.
Also, functions memcpy_toiovecend and memcpy_fromiovecend
were removed in commits ba7438ae and 57dd8a07, being substituted by
copy_from_iter and copy_to_iter.

This patch makes use of struct iov_iter, which has references
to msg_iov and msg_iovln, and makes use of copy_from_iter
and copy_to_iter.

Reported-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agokni: fix build with kernel 4.0
Pablo de Lara [Tue, 28 Apr 2015 17:37:42 +0000 (18:37 +0100)]
kni: fix build with kernel 4.0

Due to API changes in function pointer ndo_bridge_setlink
(commit ad41faa8) and the rename of functions vlan_tx_*
(commit df8a39de) in kernel 4.0, DPDK would not build.

This patch adds the properly checks to fix the compilation.

Reported-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoixgbe: use scattered Rx with bulk allocation
Vlad Zolotarov [Wed, 29 Apr 2015 08:38:17 +0000 (11:38 +0300)]
ixgbe: use scattered Rx with bulk allocation

Simply initialze rx_pkt_burst callback to ixgbe_recv_pkts_lro_bulk_alloc()
if the conditions are right.

This is possible because work against HW in LRO and scattered cases is
exactly the same and LRO callback already supports the bulk allocation.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoixgbe: remove simple scalar scattered Rx method
Vlad Zolotarov [Wed, 29 Apr 2015 08:38:16 +0000 (11:38 +0300)]
ixgbe: remove simple scalar scattered Rx method

Kill ixgbe_recv_scattered_pkts() - use ixgbe_recv_pkts_lro_single_alloc()
instead.

Work against HW queues in LRO and scattered Rx cases is exactly the same.
Therefore we may drop the inferior callback.

This patch also changes the sw_rsc_ring allocation in the
ixgbe_dev_rx_queue_setup() to always allocate sw_rsc_ring instead of
explicitly allocating it in all possible cases when it may be needed:
LRO and/or scattered Rx.

This will only impose sizeof(void*) * IXGBE_MAX_RING_DESC = 32KB overhead
per Rx queue as a price for a much simpler code, which seems reasonable.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoixgbe: rename rsc to sc/scattered_rx
Vlad Zolotarov [Wed, 29 Apr 2015 08:38:15 +0000 (11:38 +0300)]
ixgbe: rename rsc to sc/scattered_rx

   - ixgbe_rsc_entry -> ixgbe_scattered_rx_entry
   - sw_rsc_ring -> sw_sc_ring
   - ixgbe_free_rsc_cluster() -> ixgbe_free_sc_cluster()
   - In local variables: xx_rsc_yy -> xx_sc_yy

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoixgbe: remove unused field for RSC
Vlad Zolotarov [Wed, 29 Apr 2015 08:38:14 +0000 (11:38 +0300)]
ixgbe: remove unused field for RSC

Fixes: 8eecb3295aed ("ixgbe: add LRO support")

Reported-by: Shiweixian <shiweixian@huawei.com>
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoixgbe: move PMD specific fields out of base driver
Vlad Zolotarov [Wed, 29 Apr 2015 08:38:13 +0000 (11:38 +0300)]
ixgbe: move PMD specific fields out of base driver

Move rx_bulk_alloc_allowed and rx_vec_allowed from ixgbe_hw to ixgbe_adapter.

Fixes: 01fa1d6215fa ("ixgbe: unify Rx setup")

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agodoc: link doxygen api to source code
John McNamara [Wed, 22 Apr 2015 16:10:02 +0000 (17:10 +0100)]
doc: link doxygen api to source code

Enabled Doxygen option to add links to the source code
in documented entities.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Siobhan Butler <siobhan.a.butler@intel.com>
9 years agodoc: disable doxygen member sorting
John McNamara [Wed, 29 Apr 2015 10:47:47 +0000 (11:47 +0100)]
doc: disable doxygen member sorting

Disabled the doxygen option to sort member data so that functions
and struct memebers are listed in order of definition in the
brief and main sections.

Previously they were sorted in the brief section and were in
definition order in the main section.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Siobhan Butler <siobhan.a.butler@intel.com>
9 years agodoc: fix spellings and typos
John McNamara [Fri, 24 Apr 2015 12:58:11 +0000 (13:58 +0100)]
doc: fix spellings and typos

Fixed several typos and spelling errors in guide docs.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
9 years agoigb: handle VF LPE mailbox message
Sergio Gonzalez Monroy [Wed, 18 Mar 2015 12:01:31 +0000 (12:01 +0000)]
igb: handle VF LPE mailbox message

This patch adds the handle function for the LPE mailbox message (VF to
PF) to set maximum packet size, which can be used to enable jumbo
frame support.

Signed-off-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoeal/bsd: fix virtio on FreeBSD
Raz Amir [Thu, 16 Apr 2015 11:52:07 +0000 (14:52 +0300)]
eal/bsd: fix virtio on FreeBSD

Closing /dev/io fd causes SIGBUS in inb/outb instructions
as the process loses the IOPL privileges once the fd is closed:
(gdb) bt
0  0x0000000000492f2c in outb (port=49170, data=0 '\000')
at /usr/include/machine/cpufunc.h:244
1  0x0000000000492f7a in outb_p (data=0 '\000', port=49170)
at /dpdk/dpdk-2.0.0/lib/librte_pmd_virtio/virtio_pci.h:211
2  0x000000000049328d in vtpci_set_status (hw=0x80331f380, status=0 '\000')
at /dpdk/dpdk-2.0.0/lib/librte_pmd_virtio/virtio_pci.c:130
3  0x00000000004931fe in vtpci_reset (hw=0x80331f380)
at /dpdk/dpdk-2.0.0/lib/librte_pmd_virtio/virtio_pci.c:108
4  0x00000000004a175e in eth_virtio_dev_init (eth_dev=0x831b80 <rte_eth_devices>)
at /dpdk/dpdk-2.0.0/lib/librte_pmd_virtio/virtio_ethdev.c:1150
5  0x0000000000462c09 in rte_eth_dev_init (pci_drv=0x79d1a0 <rte_virtio_pmd>,
pci_dev=0x802417560) at /dpdk/dpdk-2.0.0/lib/librte_ether/rte_ethdev.c:326
6  0x000000000046f03f in rte_eal_pci_probe_one_driver (dr=0x79d1a0 <rte_virtio_pmd>,
dev=0x802417560) at /dpdk/dpdk-2.0.0/lib/librte_eal/bsdapp/eal/eal_pci.c:487
7  0x0000000000475b06 in pci_probe_all_drivers (dev=0x802417560)
at /dpdk/dpdk-2.0.0/lib/librte_eal/common/eal_common_pci.c:116
8  0x0000000000475bb9 in rte_eal_pci_probe ()
at /dpdk/dpdk-2.0.0/lib/librte_eal/common/eal_common_pci.c:246
9  0x000000000046cd63 in rte_eal_init (argc=5, argv=0x7fffffffeaf0)
at /dpdk/dpdk-2.0.0/lib/librte_eal/bsdapp/eal/eal.c:554
10 0x0000000000404544 in main ()

Fixes: 8a312224bcde ("eal/bsd: fix fd leak")

Signed-off-by: Raz Amir <razamir22@gmail.com>
Reviewed-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agolog: fix log level check
Jean Dao [Mon, 6 Apr 2015 19:14:20 +0000 (21:14 +0200)]
log: fix log level check

According to the api, rte_log() / rte_vlog() are supposed to check the log level
and type but they were not doing so. This check was only done in the RTE_LOG
macro while this macro is only there to remove log messages at build time.

rte_log() always calls rte_vlog(), so do the check in rte_vlog() only.

Signed-off-by: Jean Dao <jean.dao@6wind.com>
Signed-off-by: David Marchand <david.marchand@6wind.com>
9 years agoenic: remove use of old flow director struct
Sujith Sankar [Tue, 21 Apr 2015 05:51:28 +0000 (11:21 +0530)]
enic: remove use of old flow director struct

This patch removes the use of rte_fdir_filter from enic_clsf.
This also takes care of modifying the version and copyright string.

Signed-off-by: Sujith Sankar <ssujith@cisco.com>
9 years agoacl: use setjmp/longjmp to handle alloc failures at build phase
Konstantin Ananyev [Thu, 19 Feb 2015 17:44:19 +0000 (17:44 +0000)]
acl: use setjmp/longjmp to handle alloc failures at build phase

During build phase ACL doing quite a lot of memory allocations
for relatively small temporary structures.
In theory each of such allocation can fail, so we need to handle
all these possible failures.
That adds a lot of extra checks and makes the code harder to read and follow.
To simplify the process, made changes to handle all such failures
in one place.
Note, that all that memory for temporary structures
is freed at one go at the end of build phase.

Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoapp/test: check cloning with different priv sizes
Olivier Matz [Wed, 22 Apr 2015 09:57:30 +0000 (11:57 +0200)]
app/test: check cloning with different priv sizes

Verify that we can attach a mbuf to another one that does not have
the same data room size and priv_size.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoapp/test: check cloning a clone
Olivier Matz [Wed, 22 Apr 2015 09:57:29 +0000 (11:57 +0200)]
app/test: check cloning a clone

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoapp/test: enhance mbuf refcnt check
Olivier Matz [Wed, 22 Apr 2015 09:57:28 +0000 (11:57 +0200)]
app/test: enhance mbuf refcnt check

Check that the data in the cloned mbuf is the same than in the
reference mbuf.
Check that the reference counter is incremented for each segment.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoapp/test: rename mbuf variable
Olivier Matz [Wed, 22 Apr 2015 09:57:27 +0000 (11:57 +0200)]
app/test: rename mbuf variable

It's better to name the mbuf 'm' instead of 'mc' as it's not a clone.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agombuf: allow to clone an indirect mbuf
Olivier Matz [Wed, 22 Apr 2015 09:57:26 +0000 (11:57 +0200)]
mbuf: allow to clone an indirect mbuf

Remove one limitation of rte_pktmbuf_attach(): "mbuf we're attaching to
must be direct".

Now, when we attach to an indirect mbuf:
- copy the all relevant fields (addr, len, offload, ...) as before
- get the pointer to the mbuf that embeds the data buffer (direct mbuf),
  and increase the reference counter of this one.

When detaching the mbuf, we can retrieve this direct mbuf as the pointer
is determined from the buffer address.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agombuf: fix cloning with private mbuf data
Olivier Matz [Wed, 22 Apr 2015 09:57:25 +0000 (11:57 +0200)]
mbuf: fix cloning with private mbuf data

Add a new priv_size field in mbuf structure that should
be initialized at mbuf pool creation. This field contains the
size of the application private data in mbufs.

Introduce new static inline functions rte_mbuf_from_indirect()
and rte_mbuf_to_baddr() to replace the existing macros, which
take the private size in account when attaching and detaching
mbufs.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Reviewed-by: Zoltan Kiss <zoltan.kiss@linaro.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoapps: use helper to create mbuf pools
Olivier Matz [Wed, 22 Apr 2015 09:57:24 +0000 (11:57 +0200)]
apps: use helper to create mbuf pools

When it's possible, use the new helper to create the mbuf pools.
Most of the patch is trivial, except for the following files that
have some specifics (indirect mbufs):
- ip_fragmentation
- ip_pipeline
- ipv4_multicast
- vhost

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agombuf: introduce a new helper to create a pool
Olivier Matz [Wed, 22 Apr 2015 09:57:23 +0000 (11:57 +0200)]
mbuf: introduce a new helper to create a pool

Add a new wrapper to rte_mempool_create() to simplify the creation
of a packet mbuf pool.

This wrapper can be used if there is no specific mempool flags, and
no specific mbuf or pool constructor function, which is most of the
use cases.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoapp/testpmd: use standard functions to initialize mbufs and mbuf pool
Olivier Matz [Wed, 22 Apr 2015 09:57:22 +0000 (11:57 +0200)]
app/testpmd: use standard functions to initialize mbufs and mbuf pool

The rte_pktmbuf_pool_init() and rte_pktmbuf_init() functions now
support to have a non-hardcoded buffer length. We can remove the
specific functions used in testpmd and replace them by the standard
ones.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agombuf: fix init when private size is not zero
Olivier Matz [Wed, 22 Apr 2015 09:57:21 +0000 (11:57 +0200)]
mbuf: fix init when private size is not zero

Allow the user to use the default rte_pktmbuf_init() function even
if the mbuf private size is not 0.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agombuf: add accessors for data room and private size
Olivier Matz [Wed, 22 Apr 2015 09:57:20 +0000 (11:57 +0200)]
mbuf: add accessors for data room and private size

This code retrieving the pool private area is duplicated in many
places, we can use of function for it.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoexamples: always initialize mbuf pool private area
Olivier Matz [Wed, 22 Apr 2015 09:57:19 +0000 (11:57 +0200)]
examples: always initialize mbuf pool private area

The mbuf pool private area must always be populated in a mbuf pool.
The applications or drivers may expect that for a mbuf pool, the mbuf
pool private area (mbuf_data_room_size and mbuf_priv_size) are
properly filled.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agombuf: fix data room size calculation in pool init
Olivier Matz [Wed, 22 Apr 2015 09:57:18 +0000 (11:57 +0200)]
mbuf: fix data room size calculation in pool init

Deduct the mbuf data room size from mempool->elt_size and priv_size,
instead of using an hardcoded value that is not related to the real
buffer size.

To use rte_pktmbuf_pool_init(), the user can either:
- give a NULL parameter to rte_pktmbuf_pool_init(): in this case, the
  private size is assumed to be 0, and the room size is
  mp->elt_size - sizeof(struct rte_mbuf).
- give the rte_pktmbuf_pool_private filled with appropriate
  data_room_size and priv_size values.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoixgbe/base: update readme
Ouyang Changchun [Thu, 26 Feb 2015 02:08:47 +0000 (10:08 +0800)]
ixgbe/base: update readme

Update the share code version into cid-10g-shared-code.2015.02.03.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: more constants
Ouyang Changchun [Thu, 12 Feb 2015 12:00:47 +0000 (20:00 +0800)]
ixgbe/base: more constants

Update macros in ixgbe_type header files.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: command for flow director
Ouyang Changchun [Thu, 12 Feb 2015 12:01:01 +0000 (20:01 +0800)]
ixgbe/base: command for flow director

Add flow director command into argument to hold the current value of FDIRCMD register;
handle flow director mask for non-cloud mode.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: hook hardware address declaration
Ouyang Changchun [Thu, 12 Feb 2015 12:00:46 +0000 (20:00 +0800)]
ixgbe/base: hook hardware address declaration

Use IOMEM for hw addr.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: add timeout to host interface command
Ouyang Changchun [Thu, 12 Feb 2015 12:00:43 +0000 (20:00 +0800)]
ixgbe/base: add timeout to host interface command

Add timeout as new argument in host interface command.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: get host interface command status
Ouyang Changchun [Thu, 12 Feb 2015 12:00:42 +0000 (20:00 +0800)]
ixgbe/base: get host interface command status

Add new function ixgbe_get_hi_status to get host interface command status.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: rework host interface command
Ouyang Changchun [Thu, 12 Feb 2015 12:00:50 +0000 (20:00 +0800)]
ixgbe/base: rework host interface command

Request and response command have different struct.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: new X550 device ids
Ouyang Changchun [Thu, 12 Feb 2015 12:01:06 +0000 (20:01 +0800)]
ixgbe/base: new X550 device ids

The old device id: IXGBE_DEV_ID_X550EM_X is split into 2 new device id:
IXGBE_DEV_ID_X550EM_X_10G_T and IXGBE_DEV_ID_X550EM_X_1G_T

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: calculate X550 checksum
Ouyang Changchun [Thu, 12 Feb 2015 12:01:00 +0000 (20:01 +0800)]
ixgbe/base: calculate X550 checksum

Add function to calculate checksum for X550; and add buffer into argument list
to hold the eeprom image.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: support kx4 phy
Ouyang Changchun [Thu, 12 Feb 2015 12:01:04 +0000 (20:01 +0800)]
ixgbe/base: support kx4 phy

Setup kx4 phy for x550em;

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: new X557 phy
Ouyang Changchun [Thu, 12 Feb 2015 12:00:48 +0000 (20:00 +0800)]
ixgbe/base: new X557 phy

Add new phy ID: X557_PHY_ID, and implement its internal setup function and external init function.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: remove lan id from phy struct
Ouyang Changchun [Thu, 12 Feb 2015 12:00:44 +0000 (20:00 +0800)]
ixgbe/base: remove lan id from phy struct

Remove lan_id from struct ixgbe_phy_info.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: set X540 phy power
Ouyang Changchun [Thu, 12 Feb 2015 12:00:40 +0000 (20:00 +0800)]
ixgbe/base: set X540 phy power

Define the function of setting phy power;
Set copper phy power for x540.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: extract management capability check
Ouyang Changchun [Thu, 12 Feb 2015 12:00:39 +0000 (20:00 +0800)]
ixgbe/base: extract management capability check

Define function ixgbe_mng_present to check if management capability is present or not.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: get X550 bus info
Ouyang Changchun [Thu, 12 Feb 2015 12:00:49 +0000 (20:00 +0800)]
ixgbe/base: get X550 bus info

New function to get bus information for x550em;

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix X550 check
Ouyang Changchun [Thu, 12 Feb 2015 12:01:05 +0000 (20:01 +0800)]
ixgbe/base: fix X550 check

Update the macro to read/write iosf sb stat.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix X550 CS4227 address
Ouyang Changchun [Thu, 12 Feb 2015 12:00:38 +0000 (20:00 +0800)]
ixgbe/base: fix X550 CS4227 address

Update the address of IXGBE_CS4227.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix X550 init early return
Ouyang Changchun [Thu, 12 Feb 2015 12:00:37 +0000 (20:00 +0800)]
ixgbe/base: fix X550 init early return

It could early return according to read status.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix X550 link speed
Ouyang Changchun [Thu, 12 Feb 2015 12:00:36 +0000 (20:00 +0800)]
ixgbe/base: fix X550 link speed

Remove unnecessary '|' for the link speed.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: rework link speed branches
Ouyang Changchun [Thu, 12 Feb 2015 12:00:57 +0000 (20:00 +0800)]
ixgbe/base: rework link speed branches

Use switch-case statement to replace if-else statement.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: support 5G link speed
Ouyang Changchun [Thu, 12 Feb 2015 12:00:56 +0000 (20:00 +0800)]
ixgbe/base: support 5G link speed

Support 5G link speed for X550.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix mac type checks
Ouyang Changchun [Thu, 12 Feb 2015 12:00:51 +0000 (20:00 +0800)]
ixgbe/base: fix mac type checks

It needs '>=' rather than '=='.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix Tx pending clearing
Ouyang Changchun [Thu, 12 Feb 2015 12:00:45 +0000 (20:00 +0800)]
ixgbe/base: fix Tx pending clearing

Wait for a last completion before clearing buffers,
and before proceeding, it needs make sure that the PCIe block does not have
transactions pending.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix bus type overwrite
Ouyang Changchun [Thu, 12 Feb 2015 12:00:35 +0000 (20:00 +0800)]
ixgbe/base: fix bus type overwrite

It needs check if the bus type is the unknown type or not.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: fix SFP probing
Ouyang Changchun [Thu, 12 Feb 2015 12:00:59 +0000 (20:00 +0800)]
ixgbe/base: fix SFP probing

LAN ID is needed for i2c access, so move it before reading I2C eeprom.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: set SFP probing retries
Ouyang Changchun [Thu, 12 Feb 2015 12:00:58 +0000 (20:00 +0800)]
ixgbe/base: set SFP probing retries

Check if SFP is detected or not.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: i2c bit-bang mode
Ouyang Changchun [Thu, 12 Feb 2015 12:01:03 +0000 (20:01 +0800)]
ixgbe/base: i2c bit-bang mode

Support the bit-bang mode on X550;
And negate I2C output enable;

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: i2c combined functions
Ouyang Changchun [Thu, 12 Feb 2015 12:00:54 +0000 (20:00 +0800)]
ixgbe/base: i2c combined functions

New API to perform I2C read combined operation.
New API to perform I2C write combined operation.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: use another log macro
Ouyang Changchun [Thu, 12 Feb 2015 12:00:34 +0000 (20:00 +0800)]
ixgbe/base: use another log macro

Use DEBUGOUT to replace ERROR_REPORT1.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agoixgbe/base: minor changes
Ouyang Changchun [Thu, 12 Feb 2015 12:00:33 +0000 (20:00 +0800)]
ixgbe/base: minor changes

Remove '&' before function name;
Adjust the spaces etc.

Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
9 years agonet: remove unneeded include
Bruce Richardson [Thu, 23 Apr 2015 12:49:19 +0000 (13:49 +0100)]
net: remove unneeded include

There are no memcpy functions in rte_ip.h so there is no need to include
rte_memcpy.h in that file.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agopci: allow const pci address parameter
Stephen Hemminger [Thu, 16 Apr 2015 23:23:40 +0000 (16:23 -0700)]
pci: allow const pci address parameter

probe and close both don't modify the rte_pci_addr structure
that is passed.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
9 years agopci: make device id tables const
Stephen Hemminger [Thu, 16 Apr 2015 23:23:39 +0000 (16:23 -0700)]
pci: make device id tables const

The PCI device id table is immutable and should be made const
in all drivers. The pseudo drivers can initialize their local
copy as necessary.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
9 years agoethdev: remove unnecessary paren on return
Stephen Hemminger [Thu, 9 Apr 2015 21:29:42 +0000 (14:29 -0700)]
ethdev: remove unnecessary paren on return

The Linux style is not to put extra useless paren's around
the expression passed to return statement.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoethdev: make tables const
Stephen Hemminger [Thu, 9 Apr 2015 21:29:41 +0000 (14:29 -0700)]
ethdev: make tables const

The statistics tables and null mac address should be immutable.
Fix up get_addr routines to accept const args.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agoethdev: remove extra inline
Stephen Hemminger [Thu, 9 Apr 2015 21:29:39 +0000 (14:29 -0700)]
ethdev: remove extra inline

There is no reason to inline functions that are not in the critical
path.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agoenic: migrate flow director filtering to new API
Sujith Sankar [Thu, 9 Apr 2015 09:29:32 +0000 (14:59 +0530)]
enic: migrate flow director filtering to new API

This patch helps enic migrate to the new flow-director API.

It takes care of the following.
1.  The change in fdir_filter structure and stats structure
2.  DPDK interface functions in enic_ethdev.c
3.  ENIC driver functions that deal with the VIC adapter

Signed-off-by: Sujith Sankar <ssujith@cisco.com>
9 years agoenic: fix port id in received mbufs
Adrien Mazarguil [Tue, 14 Apr 2015 14:23:55 +0000 (16:23 +0200)]
enic: fix port id in received mbufs

This field is not supposed to contain the RX queue index. Applications can
rely on it to determine the port a given mbuf comes from.

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Acked-by: Sujith Sankar <ssujith@cisco.com>
9 years agouse simple zero initializers
Thomas Monjalon [Wed, 15 Apr 2015 17:09:28 +0000 (19:09 +0200)]
use simple zero initializers

To initialize a structure with zeros, one field was explicitly set
to avoid "missing initializer" bug with old GCC (e.g. 4.4).
This warning is now disabled (commit <insertlater>) for old versions of GCC,
so the workarounds may be removed.

These initializers should not be needed for static variables but they
are still used to workaround an ICC bug (see commit b2595c4aa92d).

There is one remaining exception where {0} initializer doesn't work cleanly,
even with recent GCC:
lib/librte_pmd_ixgbe/ixgbe_rxtx_vec.c:735:9:
error: missing braces around initializer [-Werror=missing-braces]
  struct rte_mbuf mb_def = {0}; /* zeroed mbuf */

Tested with gcc-4.4.7 (CentOS), gcc-4.7.2 (Debian), gcc-4.9.2 (Arch),
clang-3.6.0 and icc-13.1.1.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Tested-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Tested-by: John McNamara <john.mcnamara@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Vlad Zolotarov <vladz@cloudius-systems.com>
9 years agomk: fix build with gcc 4.4 and clang
Thomas Monjalon [Tue, 14 Apr 2015 09:23:36 +0000 (11:23 +0200)]
mk: fix build with gcc 4.4 and clang

With GCC 4.4.7 from CentOS 6.5, the following errors arise:

lib/librte_pmd_ixgbe/ixgbe_rxtx.c: In function 'ixgbe_dev_rx_queue_setup':
lib/librte_pmd_ixgbe/ixgbe_rxtx.c:2509: error: missing initializer
lib/librte_pmd_ixgbe/ixgbe_rxtx.c:2509: error: (near initialization for 'dev_info.driver_name')

lib/librte_pmd_ixgbe/ixgbe_rxtx.c: In function 'ixgbe_set_rsc':
lib/librte_pmd_ixgbe/ixgbe_rxtx.c:4072: error: missing initializer
lib/librte_pmd_ixgbe/ixgbe_rxtx.c:4072: error: (near initialization for 'dev_info.driver_name')

lib/librte_pmd_ixgbe/ixgbe_rxtx.c: In function 'ixgbe_recv_pkts_lro_single_alloc':
lib/librte_pmd_ixgbe/ixgbe_rxtx.c:1479: error: 'next_rsc_entry' may be used uninitialized in this function
lib/librte_pmd_ixgbe/ixgbe_rxtx.c:1480: error: 'next_rxe' may be used uninitialized in this function

The "missing initializer" warning is a GCC bug which seems fixed in 4.7.
The same warning is thrown by clang.
The "may be used uninitialized" warning is another GCC bug which seems fixed in 4.7.

Fixes: 8eecb3295aed ("ixgbe: add LRO support")

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Vlad Zolotarov <vladz@cloudius-systems.com>
9 years agodoc: fix vhost guide
Igor Ryzhov [Wed, 8 Apr 2015 16:31:07 +0000 (19:31 +0300)]
doc: fix vhost guide

Guide says that a configure parameter to choose between vhost cuse
and vhost user will be introduced in the future, but it’s already
added by commit 28a1ccca41bf.

Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
Acked-by: Changchun Ouyang <changchun.ouyang@intel.com>
9 years agodoc: convert prog guide glossary to definition list
John McNamara [Fri, 10 Apr 2015 15:39:26 +0000 (16:39 +0100)]
doc: convert prog guide glossary to definition list

Converted the Glossary table in the Programmer's Guide
to a definition list to improve rendering.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
9 years agodoc: fix verbatim sections for vmxnet3
John McNamara [Fri, 10 Apr 2015 14:45:39 +0000 (15:45 +0100)]
doc: fix verbatim sections for vmxnet3

Fixed two verbatim text sections in vmxnet3 docs.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Yong Wang <yongwang@vmware.com>
9 years agoixgbe: add LRO support
Vlad Zolotarov [Mon, 30 Mar 2015 19:21:11 +0000 (22:21 +0300)]
ixgbe: add LRO support

    - Only x540 and 82599 devices support LRO.
    - Add the appropriate HW configuration.
    - Add RSC aware rx_pkt_burst() handlers:
       - Implemented bulk allocation and non-bulk allocation versions.
       - Add LRO-specific fields to rte_eth_rxmode, to rte_eth_dev_data
         and to ixgbe_rx_queue.
       - Use the appropriate handler when LRO is requested.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoixgbe: code refactoring
Vlad Zolotarov [Mon, 30 Mar 2015 19:21:10 +0000 (22:21 +0300)]
ixgbe: code refactoring

   - ixgbe_rx_alloc_bufs():
      - Reset the rte_mbuf fields only when requested.
      - Take the RDT update out of the function.
      - Add the stub when RTE_LIBRTE_IXGBE_RX_ALLOW_BULK_ALLOC is not defined.
   - ixgbe_recv_scattered_pkts():
      - Take the code that updates the fields of the cluster's HEAD buffer into
        the inline function.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoixgbe: cleanups
Vlad Zolotarov [Mon, 30 Mar 2015 19:21:09 +0000 (22:21 +0300)]
ixgbe: cleanups

   - Removed the not needed casting.
   - ixgbe_dev_rx_init(): shorten the lines by defining a local alias variable to access
                          &dev->data->dev_conf.rxmode.

Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoethdev: make dev_ops const
Stephen Hemminger [Tue, 7 Apr 2015 21:21:03 +0000 (14:21 -0700)]
ethdev: make dev_ops const

The ethernet device ops function table should be made const for
safety and security.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>