oerrors was txdgpc - hw_stats->gptc,
txdgpc is the number of packets DMA'ed by the host
and was being reset on every call to read stats so it could be < gptc.
Because we currently have no way to add txdgpc to struct hw_stats so
that we can maintain a persistent value per port oerrors has now been
set to 0. References to txdgpc is now removed as we don't use it. This
patch also removes rxnfgpc as it's not used anywhere.
Fixes: afebc86be134 ("ixgbe: refactor stats register reads") Signed-off-by: Maryam Tahhan <maryam.tahhan@intel.com> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
For 2.1 release, in attempt to minimize number of RX routines to support,
ixgbe scatter and ixgbe LRO RX routines were merged into one
that can handle both cases.
Though I completely missed the fact, that while LRO could only be used
when HW CRC strip is enabled, scatter RX should work for both cases
(HW CRC strip on/off).
That patch restores missed functionality.
Fixes: 9d8a92628f21 ("ixgbe: remove simple scalar scattered Rx method") Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Xuelin Shi [Wed, 29 Jul 2015 06:38:20 +0000 (14:38 +0800)]
ixgbe: fix data access on big endian cpu
1. cpu use data owned by ixgbe must use rte_le_to_cpu_xx(...)
2. cpu fill data to ixgbe must use rte_cpu_to_le_xx(...)
3. checking pci status with converted constant
Signed-off-by: Xuelin Shi <xuelin.shi@freescale.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
This patch fixes the issue:
Testpmd crashed with Segmentation fault when setup tx queues on vf
Steps for reproduce:
- create one vf device from i40e driver
- bind vf device to igb_uio and start testpmd
With debugging tools, we saw the struct i40e_vf is cleared after
memcpy(&dev->data->dev_conf, dev_conf, sizeof(dev->data->dev_conf)) in
rte_eth_dev_configure, which should not happen, and the pointer to
i40e_vf isn't in the range of i40e_adapter.
The root cause is the dev_private_size in i40e virtual function driver
struct rte_i40evf_pmd was set incorrectly.
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com> Tested-by: Marvin Liu <yong.liu@intel.com>
Build log:
lib/librte_eal/common/eal_common_pci.c:188:4: error:
implicit declaration of function pci_config_space_set
The function rte_eal_pci_probe_one_driver, which calls
pci_config_space_set, was moved to eal_common_pci.c,
but pci_config_space_set was left in eal_pci.c with static specifier.
Fixes: 4d4ebca4 ("pci: merge probing and closing functions for linux and bsd") Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Thomas Monjalon [Tue, 28 Jul 2015 16:22:39 +0000 (18:22 +0200)]
config: disable bnx2x driver
This driver has too many issues:
- too big
- bad coding style
- no git history (dropped in 2 patches)
- no documentation
- no BSD support
- no maintainer
And the biggest one, constraining this disabling:
- many build issues
If the last 4 issues are not fixed in the next release 2.2,
the driver must be removed.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
When adding with cuckoo hash a key which was already inserted
a new slot is dequeued and then enqueued back, but the enqueue
operation was not done properly.
Thomas Monjalon [Tue, 28 Jul 2015 08:29:51 +0000 (10:29 +0200)]
eal/bsd: fix build
The BSD function for contigmem init and attach must now use the same name
as Linux (hugepage prefix) to avoid code duplication.
The attach function was renamed but the init function was forgotten.
Fixes: d12b6da14bfa ("eal: deduplicate memory initialization") Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Jijiang Liu [Fri, 3 Jul 2015 06:58:17 +0000 (14:58 +0800)]
doc: add a VXLAN sample guide
Add a VXLAN sample guide in the sample_app_ug directory.
It includes:
- Add the overlay networking picture with svg format.
- Add the TEP termination framework picture with svg format.
- Add the tep_termination.rst file
- Change the index.rst file for the above pictures index.
Signed-off-by: Jijiang Liu <jijiang.liu@intel.com> Signed-off-by: Thomas Long <thomas.long@intel.com> Acked-by: Marvin Liu <yong.liu@intel.com>
examples/l3fwd-power: enable one-shot Rx interrupt and polling switch
The patch demonstrates how to handle per rx queue interrupt in a NAPI-like
implementation in userspace. The working thread mainly runs in polling mode
and switch to interrupt mode only if there is no packet received in recent polls.
The working thread returns to polling mode immediately once it receives an
interrupt notification caused by the incoming packets.
The sample keeps running in polling mode if the binding PMD hasn't supported
the rx interrupt yet. Now only ixgbe(pf/vf) and igb support it.
Signed-off-by: Danny Zhou <danny.zhou@intel.com> Signed-off-by: Cunming Liang <cunming.liang@intel.com>
The patch does below for igb PF:
- Setup NIC to generate MSI-X interrupts
- Set the IVAR register to map interrupt causes to vectors
- Implement interrupt enable/disable functions
Signed-off-by: Danny Zhou <danny.zhou@intel.com> Signed-off-by: Cunming Liang <cunming.liang@intel.com>
The patch does below things for ixgbe PF and VF:
- Setup NIC to generate MSI-X interrupts
- Set the IVAR register to map interrupt causes to vectors
- Implement interrupt enable/disable functions
Signed-off-by: Danny Zhou <danny.zhou@intel.com> Signed-off-by: Yong Liu <yong.liu@intel.com> Signed-off-by: Cunming Liang <cunming.liang@intel.com>
The patch adds two dev_ops functions to enable and disable rx queue
interrupts.
In addition, it adds rte_eth_dev_rx_intr_ctl/rx_intr_q to support
per port or per queue rx intr event set.
Signed-off-by: Danny Zhou <danny.zhou@intel.com> Signed-off-by: Cunming Liang <cunming.liang@intel.com>
The patch exposes intr event fd create and release for PMD.
The device driver can assign the number of event associated with interrupt vector.
It also provides misc functions to check 1) allows other slowpath intr(e.g. lsc);
2) intr event on fastpath is enabled or not.
eal/linux: fix link status interrupt with uio_pci_generic
The intr handle type(RTE_INTR_HANDLE_UIO_INTX) was introduced by UIO pci generic.
When turning on the lsc interrupt, it complains fd read error.
The patch uses the correct read size in the case of RTE_INTR_HANDLE_UIO_INTX.
Fixes: 3f313bef3467 ("eal/linux: fix irq handling with igb_uio") Reported-by: Yong Liu <yong.liu@intel.com> Signed-off-by: Cunming Liang <cunming.liang@intel.com>
The patch adds 'rte_epoll_wait' and 'rte_epoll_ctl' for async event wakeup.
It defines 'struct rte_epoll_event' as the event param.
When the event fds add to a specified epoll instance, 'eptrs' will hold
the rte_epoll_event object pointer.
The 'op' uses the same enum as epoll_wait/ctl does.
The epoll event support to carry a raw user data and to register a callback
which is executed during wakeup.
The patch adds interrupt vectors support in rte_intr_handle.
'vec_en' is set when interrupt vectors are detected and associated
event fds are set. Those event fds are stored in efds[].
'intr_vec' is reserved for device driver to initialize the vector
mapping table.
Signed-off-by: Danny Zhou <danny.zhou@intel.com> Signed-off-by: Cunming Liang <cunming.liang@intel.com>
When the bonded port is started it also starts the slave port,
but the slave port status is not set. A slave_flag has been
added to struct rte_port to resolve this issue.
Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com>
Pablo de Lara [Thu, 23 Jul 2015 16:12:07 +0000 (17:12 +0100)]
examples/l3fwd: increase lookup burst size to 8
With the new hash implementation, the minimum lookup burst size
to get good performance is 8, since its internal pipeline
consists of 4 stages of 2 entries each, so to avoid
duplication, burst size should be 8 or more entries.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Maciej Gajdzica [Tue, 21 Jul 2015 14:39:00 +0000 (16:39 +0200)]
examples/ip_pipeline: fix cpu cores parsing
This patch fixes parsing value of core variable in pipeline config.
Before not every combination of cores (c), sockets (s) and
hyperthreading (h) was parsed correctly.
Signed-off-by: Maciej Gajdzica <maciejx.t.gajdzica@intel.com> Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
As Steve pointed out, the commit 11b220c6498d ("ixgbe: fix release queue mbufs")
is not complete.
As at queue stop we don't reset vector related rx queue fields to their
initial values.
Fixes: c95584dc2b18 ("ixgbe: new vectorized functions for Rx/Tx") Reported-by: Cunming Liang <cunming.liang@intel.com> Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Jan Viktorin [Fri, 24 Jul 2015 17:11:44 +0000 (19:11 +0200)]
igb_uio: fix build
The missing header prevents to build with linux v3.18.
The problem is with kzalloc and kfree which are undefined in the igb_uio
driver.
HOSTCC="/usr/bin/gcc" HOSTCFLAGS="" ARCH=arm
RTE_TARGET=arm-armv7-a-linuxapp-gcc
CROSS=arm-none-linux-gnueabi-
lib/librte_eal/linuxapp/igb_uio/igb_uio.c:
error: implicit declaration of function 'kzalloc'
error: implicit declaration of function 'kfree'
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The logical OR operator evaluates to 1 if any of its operands is
different than 0.
Error showed later while initializing PMD:
EAL: TSC frequency is ~0 KHz
<snip>
PMD: eth_ixgbe_dev_init(): Hardware Initialization Failure: -30
EAL: Error - exiting with code: 1
Cause: Requested device 0000:84:00.0 cannot be used
Fixes: 040cf8a41187 ("eal: deduplicate timer functions") Signed-off-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Ravi Kerur [Sat, 25 Jul 2015 19:36:28 +0000 (12:36 -0700)]
eal: deduplicate timer functions
Move common functions from BSD/Linux to eal_common_timer.c.
BSD uses sysctl and Linux uses CLOCK_MONOTIC_RAW to calibrate TSC.
HPET is specific to Linux and not integrated in the common init.
Signed-off-by: Ravi Kerur <rkerur@gmail.com> Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
David Marchand [Thu, 23 Jul 2015 07:15:32 +0000 (09:15 +0200)]
eal: remove useless PCI id header inclusions
Signed-off-by: David Marchand <david.marchand@6wind.com>
[Thomas: move inclusion used by ixgbe bypass] Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Bruce Richardson [Wed, 22 Jul 2015 09:13:30 +0000 (10:13 +0100)]
ixgbe: fix check for split packets
The check for split packets to be reassembled in the vector ixgbe PMD
was incorrectly only checking the first 16 elements of the array instead
of all 32. This is fixed by changing the uint32_t values to be uint64_t
instead.
Fixes: cf4b4708a88a ("ixgbe: improve slow-path perf with vector scattered Rx") Reported-by: Zoltan Kiss <zoltan.kiss@linaro.org> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Pablo de Lara [Thu, 23 Jul 2015 14:29:41 +0000 (15:29 +0100)]
ixgbe: remove Rx bulk allocation option
RTE_LIBRTE_IXGBE_RX_ALLOW_BULK_ALLOC config option is not really
necessary, as bulk alloc rx function can be used anyway, as long as the
necessary conditions are satisfied, which are checked already
in the library.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
The vector/SSE pmd used a different element type for the tx queue sw_ring
entries. This led to lots of typecasts in the code which required specific
use of bracketing, leading to subtle errors.
For example, in the original code:
txe = (struct ixgbe_tx_entry_v *)&txq->sw_ring[i];
instead needs to be written as:
txe = &((struct ixgbe_tx_entry_v *)txq->sw_ring)[i];
We can eliminate this problem, by having two software ring pointers in the
structure for the two different element types.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
The function inside the vector/SSE poll-mode driver for releasing
the mbufs on the TX queues had the same name as another function
inside the regular PMD. To keep consistency and avoid confusion,
rename the vector PMD version to have a "_vec" suffix.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Helin Zhang <helin.zhang@intel.com>
The calculations of what mbufs were valid in the RX and TX queues were
incorrect when freeing the mbufs for the vector PMD. This led to crashes
due to invalid reference counts when mbuf debugging was turned on, and
possibly other more subtle problems (such as mbufs being freed when in use)
in other cases.
To fix this, the following changes were made:
* correct counts and post-loop values in the TX release function for the
vector code.
* create a new separate RX release function for the RX vector code, since the
tracking of what mbufs are valid or not is different for that code path
Fixes: c95584dc2b18 ("ixgbe: new vectorized functions for Rx/Tx") Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Bug fix to remove (rxnfgpc - hw_stats->gprc) which does not account
for phy errors.
hw_stats->gprc is the Number of good (non-erred) Rx packets (from the
network) that pass L2 filtering and has a legal length as defined by
LongPacketEnable. While rxnfgpc is the Number of good (non-erred with
legal length) Rx packets (from the network) regardless of packet
filtering and receive enablement. Thus hw_stats->gprc can be > rxnfgpc
and this calculation should be removed from the calculation of ierrors.
Validated with testpmd by sending packets to the interface without
forwarding enabled - packets should be dropped and the error counters
incremented.
Fixes: f6bf669b9900 ("ixgbe: account more Rx errors") Signed-off-by: Maryam Tahhan <maryam.tahhan@intel.com> Acked-by: Balazs Nemeth <balazs.nemeth@intel.com> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com> Tested-by: Marvin Liu <yong.liu@intel.com>
This fixes cases in KNI where kernel allocation function return value
is needlessly casted.
Detected with coccinelle:
lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c:3181:25-28:
WARNING: casting value returned by memory allocation function to (u32 *) is useless.
lib/librte_eal/linuxapp/kni/kni_vhost.c:690:9-28:
WARNING: casting value returned by memory allocation function to (struct rte_kni_fifo *) is useless.
lib/librte_eal/linuxapp/kni/kni_vhost.c:684:13-27:
WARNING: casting value returned by memory allocation function to (struct sk_buff *) is useless
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Update cxgbe documentation to include support for FreeBSD:
1. Add instructions on how to compile CXGBE PMD in FreeBSD.
2. Add instructions on how to flash firmware image on Chelsio T5 cards in
FreeBSD.
3. Add sample application usage for FreeBSD.
4. Add an extra step to reload kernel module in Linux in order for the new
firmware to be loaded.
5. Typo fixes.
Fix "MACRO redefined" and "function redefined" compilation errors in FreeBSD
by adding CXGBE prefix to them. Also remove reference to a linux header
linux/if_ether.h and use DPDK macros directly. Finally, enable CXGBE PMD
for FreeBSD.
nic_uio requires the pci ids to be present in rte_pci_dev_ids.h in order to
bind the devices to nic_uio. However, it's better to remove this whitelist
of pci ids, and instead rely on hw.nic_uio.bdfs kenv parameter to allow
binding any device to nic_uio.
Suggested-by: David Marchand <david.marchand@6wind.com> Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com> Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com> Acked-by: David Marchand <david.marchand@6wind.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
The fm10k driver was reading the interrupt cause register but then
using the interrupt mask register defines to look at the bits.
The result is that if a fault happens, the driver would never clear
the fault and would get into an infinite cycle of interrupts.
Note: I don't work for Intel or have the hardware manuals (probably
requires NDA anyway), but this looks logical and matches how the
known working Linux driver handles these bits.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Jing Chen <jing.d.chen@intel.com>
Pablo de Lara [Wed, 22 Jul 2015 09:50:11 +0000 (10:50 +0100)]
ixgbe: fix build with bulk alloc disabled
ixgbe_recv_pkts_lro uses field rx_free_trigger
in structure ixgbe_rx_queue, but that field is only defined
if IXGBE_RX_ALLOW_BULK_ALLOC is enabled, so even though
that field is not used when it is disabled,
compiler complains about it.
Therefore, the lines of code that use that field
have been ifdef.
Fixes: 8eecb329 ("ixgbe: add LRO support") Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Pablo de Lara [Tue, 21 Jul 2015 14:10:19 +0000 (15:10 +0100)]
hash: move struct field to keep ABI stable
In order to keep the ABI consistent with the old hash library,
hash_func_init_val field has been moved, so it remains
at the same offset as previously, since hash_func and
hash_func_init_val are fields accessed by the public function
rte_hash_hash and must keep the same offset as older versions.
Fixes: 48a399119619 ("hash: replace with cuckoo hash implementation") Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
The virtual queue ring size and the number of slots actually usable
are separate parameters. In the most common environment (QEMU)
the virtual queue ring size is 256, but some environments the
ring maybe much larger.
The ring size comes from the host and the driver must use the
actual size passed.
The number of descriptors can be either zero to use the whole
available ring, or some value smaller. This is used to limit
the number of mbufs allocated for the receive ring. If more
descriptors are requested than available the size is silently
truncated.
Note: the ring size (from host) must be a power of two, but
the number of descriptors used can be any size from 1 to the
size of the virtual ring.
Fixes: d78deadae4dc ("virtio: fix ring size negotiation") Reported-by: Changchun Ouyang <changchun.ouyang@intel.com> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Changchun Ouyang <changchun.ouyang@intel.com>
Malloc was moved to the EAL and dummy malloc library was left
to not break apps that had a librte_malloc.so dependency.
Note that the dummy library will be removed in the next release.
When building a combined library, all objects are copied to the same
directory before creating the library itself.
There are a few issues:
- CONFIG_RTE_LIBRTE_MALLOC is not a valid option anymore resulting
in wrong syntax and a compilation failure. Fix it by replacing it
with CONFIG_RTE_LIBRTE_EAL.
- As we kept a dummy library, there are now two objects with the
same name. This means that the proper rte_malloc.o object in eal gets
overwritten by an empty rte_malloc.o object from the dummy malloc lib.
Fix it by changing the name of rte_malloc.o object in the dummy
library.
- Update the copyright year.
This problem was discovered when passing invalid PCI id to the
blacklist API in devargs.
Any failures in rte_devargs_add would cause a core dump because
it would call rte_log() before the the EAL log environment was
initialized. Rather than try and log just remove the messages
and leave it up to the caller to check the return value.
Most of the other failure possibilities are when malloc() fails, and if
that happens any logging that used malloc() would also fail.
This failure was not caught by the standalone tests to devargs
because the tests are run after calling rte_eal_init (which is not
how devargs is intended to be used).
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Cleanup the code in bonding that checks ports.
* Use standard rte_eth_dev_is_valid_port
* Change name of driver string to avoid variable namespace conflicts
* Get rid of unnecessary string comparison stuff. A simple pointer
check is enough here.
* Get rid of unnecessary assignment of driver_name, it is already
done by common code.
* Don't generate unnecessary log messages on error.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Declan Doherty <declan.doherty@intel.com>
The function rte_eth_dev_is_valid_port is good way to have all
drivers using same function and solves several hotplug related
bugs from drivers not checking attached flag.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Change the log level of startup messages. Anything that is
just normal activity (like getting virtual areas) is changed
to debug level. Anything that is a failure should be NOTICE
or ERR severity.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
i40evf: fix RSS with less Rx queues than Tx queues
I40e VF driver uses the num_queue_pairs in vf structure to construct
queue index look up table. When the nb_rx_queue is less than nb_tx_queue,
num_queue_pairs is equal to nb_tx_queue. It will make the table use
invalid queue index, then application cannot poll packets on these queues.
This patch also moves the inline function i40e_align_floor from
i40e_ethdev.c to i40e_ethdev.h.
Test report: http://dpdk.org/ml/archives/dev/2015-July/021838.html
Due to the NIC's firmware update, the input set of sctp flow is changed
to source IP, destination IP, source port, destination port and
Verification-Tag. This patch adds the sport and dport in the programming
packet of flow director.
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com> Tested-by: Marvin Liu <yong.liu@intel.com>
git.droids-corp.org / dpdk.git / log