Broadcast RARP packet by injecting it to receiving mbuf array at
rte_vhost_dequeue_burst().
Commit 33226236a35e ("vhost: handle request to send RARP") iterates
all host interfaces and then broadcast it by all of them. It did
notify the switches about the new location of the migrated VM, however,
the mac learning table in the target host is wrong (at least in my
test with OVS):
$ ovs-appctl fdb/show ovsbr0
port VLAN MAC Age
1 0 b6:3c:72:71:cd:4d 10
LOCAL 0 b6:3c:72:71:cd:4e 10
LOCAL 0 52:54:00:12:34:68 9
1 0 56:f6:64:2c:bc:c0 1
Where 52:54:00:12:34:68 is the mac of the VM. As you can see from the
above, the port learned is "LOCAL", which is the "ovsbr0" port. That
is reasonable, since we indeed send the pkt by the "ovsbr0" interface.
The wrong mac table lead all the packets to the VM go to the "ovsbr0"
in the end, which ends up with all packets being lost, until the guest
send a ARP quest (or reply) to refresh the mac learning table.
Jianfeng then came up with a solution I have thought of firstly but NAKed
by myself, concerning it has potential issues [0]. The solution is as title
stated: broadcast the RARP packet by injecting it to the receiving mbuf
arrays at rte_vhost_dequeue_burst(). The re-bring of that idea made me
think it twice; it looked like a false concern to me then. And I had done
a rough verification: it worked as expected.
[0]: http://dpdk.org/ml/archives/dev/2016-February/033527.html
Another note is that while preparing this version, I found that DPDK has
some ARP related structures and macros defined. So, use them instead of
the one from standard header files here.
Cc: Thibaut Collet <thibaut.collet@6wind.com>
Suggested-by: Jianfeng Tan <jianfeng.tan@intel.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
rte_get_log_type and rte_get_log_level functions has been available
for many versions. But they are missing from the shared library map
and therefore do not get exported correctly.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
This is useful when sections have duplicate names.
Signed-off-by: Rich Lane <rich.lane@bigswitch.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
This patch adds new function rte_jobstats_abort.
It marks *job* as finished and time of this work will be add to management
time instead of execution time.
This function should be used instead of rte_jobstats_finish if condition
occurs, condition is defined by the application for example when receiving
n>0 packets.
Example of usage is added to the example l2fwd-jobstats.
At maximum load do-while loop inside Idle job will be execute once because
one or more jobs waiting to be executed, so this time should not be include
as the execution time by calling rte_jobstats_abort().
Signed-off-by: Marcin Kerlin <marcinx.kerlin@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
User should be able to configure ethdev with zero rx/tx queues,
but both should not be zero.
After above change, rte_eth_dev_tx_queue_config,
rte_eth_dev_rx_queue_config should allocate memory for rx/tx queues only
when number of rx/tx queues are nonzero.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Macro RTE_PROC_PRIMARY_OR_ERR_RET blocking the secondary process from
API usage. API access should be given to both secondary and primary.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Macros RTE_PROC_PRIMARY_OR_ERR_RET and RTE_PROC_PRIMARY_OR_RET
are blocking the secondary process from using the APIs.
API access should be given to both secondary and primary.
Reported-by: Sean Harte <sean.harte@intel.com>
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Include vfio map/rd/wr support for pci ioport.
Signed-off-by: Santosh Shukla <sshukla@mvista.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
vfio_pci_mmap() try to map all pci bars. ioport region are not mapped in
vfio/kernel so ignore mmaping for ioport.
Signed-off-by: Santosh Shukla <sshukla@mvista.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
iopl() syscall not supported in linux-arm/arm64 so always return 0 value.
Suggested-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Santosh Shukla <sshukla@mvista.com>
Acked-by: Jan Viktorin <viktorin@rehivetech.com>
Acked-by: David Marchand <david.marchand@6wind.com>
Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
When built in a C++ application, the include fails for 2 reasons:
rte_mbuf_offload.h:128:24: error:
invalid conversion from ‘void*’ to ‘rte_pktmbuf_offload_pool_private*’ [-fpermissive]
rte_mempool_get_priv(mpool);
^
The cast must be explicit for C++.
rte_mbuf_offload.h:304:1: error: expected declaration before ‘}’ token
There was a closing brace for __cplusplus but not an opening one.
Fixes: 78c8709b5ddb ("mbuf_offload: introduce library to attach offloads to mbuf")
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
When built in a C++ application, the jhash include fails:
rte_jhash.h:123:22: error:
invalid conversion from ‘const void*’ to ‘const uint32_t*’ [-fpermissive]
const uint32_t *k = key;
^
The cast must be explicit for C++.
Fixes: 8718219a8737 ("hash: add new jhash functions")
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
When built in a C++ application, the keepalive include fails:
rte_keepalive.h:142:41: error: ‘ALIVE’ was not declared in this scope
keepcfg->state_flags[rte_lcore_id()] = ALIVE;
^
C++ requires to use a scope operator to access an enum inside a struct.
There was also a namespace issue for the values (no RTE prefix).
The solution is to move the struct and related code out of the header file.
Fixes: 75583b0d1efd ("eal: add keep alive monitoring")
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Remy Horton <remy.horton@intel.com>
Malfunctioning virtio clients may not send VHOST_USER_SET_MEM_TABLE for
some reason. This causes NULL dereference in qva_to_vva().
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
The vhost_net_device_ops indirection is unnecessary because there is only
one implementation of the vhost common code.
Removing it makes the code more readable.
Signed-off-by: Rich Lane <rich.lane@bigswitch.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Since commits ff909fe21f and 4e32101f9b, it is now possible to free
memzones and rings.
The rte_mempool_create() should be modified to take advantage of this
and not leak memory when an allocation fails.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
The common vhost code only supported a single mmap per device. vhost-user
worked around this by saving the address/length/fd of each mmap after the end
of the rte_virtio_memory struct. This only works if the vhost-user code frees
dev->mem, since the common code is unaware of the extra info. The
VHOST_USER_RESET_OWNER message is one situation where the common code frees
dev->mem and leaks the fds and mappings. This happens every time I shut down a
VM.
The new code calls back into the implementation (vhost-user or vhost-cuse) to
clean up these resources.
The vhost-cuse changes are only compile tested.
Signed-off-by: Rich Lane <rich.lane@bigswitch.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
To claim that we support vhost-user live migration support:
SET_LOG_BASE request will be send only when this feature flag
is set.
Besides this flag, we actually need another feature flag set
to make vhost-user live migration work: VHOST_F_LOG_ALL.
Which, however, has been enabled long time ago.
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Tested-by: Pavel Fedin <p.fedin@samsung.com>
While in former patch we enabled GUEST_ANNOUNCE feature, so that the
guest OS will broadcast a GARP message after migration to notify the
switch about the new location of migrated VM, the thing is that
GUEST_ANNOUNCE is enabled since kernel v3.5 only. For older kernel,
VHOST_USER_SEND_RARP request comes to rescue.
The payload of this new request is the mac address of the migrated VM,
with that, we could construct a RARP message, and then broadcast it
to host interfaces.
That's how this patch works:
- list all interfaces, with the help of SIOCGIFCONF ioctl command
- construct an RARP message and broadcast it
Cc: Thibaut Collet <thibaut.collet@6wind.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
It's actually a feature already enabled in Linux kernel (since v3.5).
What we need to do is simply to claim that we support such feature,
and nothing else.
With that, the guest will send an ARP message after live migration
to notify the switches about the new location of migrated VM.
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Tested-by: Pavel Fedin <p.fedin@samsung.com>
Every time we copy a buf to vring desc, we need to log it.
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Victor Kaplansky <victork@redhat.com>
Tested-by: Pavel Fedin <p.fedin@samsung.com>
Introduce vhost_log_write() helper function to log the dirty pages we
touched. Page size is harded code to 4096 (VHOST_LOG_PAGE), and each
log is presented by 1 bit.
Therefore, vhost_log_write() simply finds the right bit for related
page we are gonna change, and set it to 1. dev->log_base denotes the
start of the dirty page bitmap.
Every time we update virtio used ring, we need to log it. And it's
been done by a new vhost_log_write() wrapper, vhost_log_used_vring().
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Victor Kaplansky <victork@redhat.com>
Tested-by: Pavel Fedin <p.fedin@samsung.com>
VHOST_USER_SET_LOG_BASE request is used to tell the backend (dpdk
vhost-user) where we should log dirty pages, and how big the log
buffer is.
This request introduces a new payload:
typedef struct VhostUserLog {
uint64_t mmap_size;
uint64_t mmap_offset;
} VhostUserLog;
Also, a fd is delivered from QEMU by ancillary data.
With those info given, an area of memory is mmaped, assigned
to dev->log_base, for logging dirty pages.
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Victor Kaplansky <victork@redhat.com>
Tested-by: Pavel Fedin <p.fedin@samsung.com>
Commit d0cf91303d73 added dependency on librte_net headers to vhost
but did not add this to the Makefile, which makes builds
non-deterministic. Curiously it is non-parallel build that is
consistently broken by this missing dependency, usually it's the other
way around, but trying to build without -j(n) fails with:
lib/librte_vhost/vhost_rxtx.c:41:20:
fatal error: rte_ip.h: No such file or directory
Fixes: d0cf91303d73 ("vhost: add Tx offload capabilities")
Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Add guest offload setting in vhost lib.
Virtio 1.0 spec (5.1.6.4 Processing of Incoming Packets) says:
1. If the VIRTIO_NET_F_GUEST_CSUM feature was negotiated, the
VIRTIO_NET_HDR_F_NEEDS_CSUM bit in flags can be set: if so,
the packet checksum at offset csum_offset from csum_start
and any preceding checksums have been validated. The checksum
on the packet is incomplete and csum_start and csum_offset
indicate how to calculate it (see Packet Transmission point 1).
2. If the VIRTIO_NET_F_GUEST_TSO4, TSO6 or UFO options were
negotiated, then gso_type MAY be something other than
VIRTIO_NET_HDR_GSO_NONE, and gso_size field indicates the
desired MSS (see Packet Transmission point 2).
In order to support these features, the following changes are added,
1. Extend 'VHOST_SUPPORTED_FEATURES' macro to add the offload features negotiation.
2. Enqueue these offloads: convert some fields in mbuf to the fields in virtio_net_hdr.
There are more explanations for the implementation.
For VM2VM case, there is no need to do checksum, for we think the
data should be reliable enough, and setting VIRTIO_NET_HDR_F_NEEDS_CSUM
at RX side will let the TCP layer to bypass the checksum validation,
so that the RX side could receive the packet in the end.
In terms of us-vhost, at vhost RX side, the offload information is
inherited from mbuf, which is in turn inherited from TX side. If we
can still get those info at RX side, it means the packet is from
another VM at same host. So, it's safe to set the
VIRTIO_NET_HDR_F_NEEDS_CSUM, to skip checksum validation.
Signed-off-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Add vhost TX offload (CSUM and TSO) support capabilities in vhost lib.
In order to support these features, and the following changes are added,
1. Extend 'VHOST_SUPPORTED_FEATURES' macro to add the offload features
negotiation.
2. Dequeue TX offload: convert the fileds in virtio_net_hdr to the
related fileds in mbuf.
Signed-off-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Most of the code is inspired on virtio driver.
rte_pci_ioport structure is filled at map time with anything needed for later
read / write calls.
At the moment, base field is used to store a x86 ioport (uint16_t) and will
be reused for other arches.
Signed-off-by: David Marchand <david.marchand@6wind.com>
Tested-by: Santosh Shukla <sshukla@mvista.com>
Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
The compiler optimization was disabled a long time ago
without describing what was the exact issue.
Maybe it does not apply anymore.
As it looks unneeded, let's remove this strange pragma.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
The structure feature_entry does not need leaf/subleaf
which were copied from x86 CPUID implementation.
On x86, a valid flag is detected with the non-zero leaf value.
This check is replaced by a check with a dummy "none" register.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
The structure feature_entry does not need leaf/subleaf
which were copied from x86 CPUID implementation.
On x86, a valid flag is detected with the non-zero leaf value.
This check is replaced by a check with a dummy "none" register.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Tested-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
The patch c344eab3ee has moved the hardware definition of CPU flags.
Now the functions checking these hardware flags are also moved.
The function rte_cpu_get_flag_enabled() is no more inline.
The benefits are:
- remove rte_cpu_feature_table from the ABI (recently added)
- hide hardware details from the API
- allow to adapt structures per arch (done in next patch)
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
The new function rte_cpu_get_flag_name() is added to the EAL API.
It is implemented (duplicated) in each arch because the next patch
will remove the public exposure of the feature tables.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
non-temporal/transient/stream version of rte_prefetch0()
The non-temporal prefetch is intended as a prefetch hint that processor
will use the prefetched data only once or short period,
unlike the rte_prefetch0() function which imply that
prefetched data to use repeatedly.
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Jan Viktorin <viktorin@rehivetech.com>
slow-path data structures need not be 128-byte cache aligned.
Reduce the alignment to 64-byte to save the memory.
No behavior change for 64-byte cache aligned systems as minimum
cache line size as 64.
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
existing rte_bitmap library implementation optimally configured to run on
64-bytes cache line, extending to 128-bytes cache line targets.
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
No need to split mbuf structure to two cache lines for 128-byte cache
line size targets as it can fit on a single 128-byte cache line.
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
- RTE_CACHE_LINE_MIN_SIZE(Supported minimum cache line size)
- __rte_cache_min_aligned(Force minimum cache line alignment)
- RTE_CACHE_LINE_SIZE_LOG2(Express cache line size in terms of log2)
Suggested-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
by default, all the targets will be configured with the 64-byte cache line
size, targets which have different cache line size can be overridden
through target specific config file.
Selected ThunderX and power8 as CONFIG_RTE_CACHE_LINE_SIZE=128 targets
based on existing configuration.
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
The file rte_config.h is automatically generated and included.
No need to #include it.
The example performance-thread needs a makefile fix to avoid
overwriting the default cflags.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
As discussed on list, switch numbering scheme to be based on year/month.
Release 2.3 then becomes 16.04.
Ref: http://dpdk.org/ml/archives/dev/2015-December/030336.html
Also, added zero padding to the month so that it appear as 16.04 and
not 16.4 in "make showversion" and rte_version().
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
It was requested by Intel, more than one year ago, to replace the name
"Intel DPDK" by "DPDK".
Some references to the old name were still in some docs and code comments,
leading to confusion.
Fixes: ac8ada004c12 ("doc: remove Intel references from release notes")
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
fix the error reported by checkpatch:
"ERROR: return is not a function, parentheses are not required"
remove parentheses in return like:
"return (logical expressions)"
remove parentheses in return a function like:
"return (rte_mempool_lookup(...))"
Fixes: 6307b909b8e0 ("lib: remove extra parenthesis after return")
Signed-off-by: Huawei Xie <huawei.xie@intel.com>
Currently rte_eal_check_module() detects Linux kernel modules via reading
/proc/modules. Built-in ones aren't listed there and therefore they are not
being found.
Add support for checking built-in modules with parsing the sysfs files
This commit obsoletes the /proc/modules parsing approach.
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Acked-by: David Marchand <david.marchand@6wind.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
When configuring RTE_MACHINE to "default", rte_memcpy implementation
is the default one (old AVX).
In this code, clang raises a warning thanks to -Wsometimes-uninitialized:
rte_memcpy.h:838:6: error:
variable 'srcofs' is used uninitialized whenever 'if' condition is false
if (dstofss > 0) {
^~~~~~~~~~~
rte_memcpy.h:849:6: note: uninitialized use occurs here
if (srcofs == 0) {
^~~~~~
It is fixed by moving srcofs initialization out of the condition.
Also dstofss calculation is corrected.
Fixes: 1ae817f9f887 ("eal/x86: tune memcpy for platforms without AVX512")
Reported-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Normally we could set RTE_PCI_DRV_NEED_MAPPING flag so that eal will
invoke pci_map_device internally for us. From that point view, there
is no need to export pci_map_device.
However, for virtio pmd driver, which is designed to work without
binding UIO (or something similar first), pci_map_device() will fail,
which ends up with virtio pmd driver being skipped. Therefore, we can
not set RTE_PCI_DRV_NEED_MAPPING blindly at virtio pmd driver.
Therefore, this patch exports pci_map_device, and let virtio pmd call
it when necessary.
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Tested-by: Santosh Shukla <sshukla@mvista.com>
Tested-by: Qian Xu <qian.q.xu@intel.com>
Reviewed-by: Tetsuya Mukawa <mukawa@igel.co.jp>
Tested-by: Tetsuya Mukawa <mukawa@igel.co.jp>
Acked-by: David Marchand <david.marchand@6wind.com>
Acked-by: Huawei Xie <huawei.xie@intel.com>
Move cpu_feature_table array from arch specific rte_cpuflags.h files to
new arch specific rte_cpuflags.c files.
Main motivation is to escape from static variable declarations in
header files. cpu_feature_table has many copies in final binary, even
exist in some object files that does not use this variable at all.
And this can be a sample to create architecture specific source files
and move some functions which are not performance sensitive from
architecture header files to source files.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
This commit is adding a generic mechanism to support multiple IOMMU
types. For now, it's only type 1 (x86 IOMMU) and no-IOMMU (a special
VFIO mode that doesn't use IOMMU at all), but it's easily extended
by adding necessary definitions to eal_vfio.h, and DMA mapping
functions to eal_pci_vfio.c.
Since type 1 IOMMU module is no longer necessary to have VFIO,
we fix the module check to check for vfio-pci instead. It's not
ideal and triggers VFIO checks more often (and thus produces more
error output, which was the reason behind the module check in the
first place), so we compensate for that by providing more verbose
logging, indicating whether VFIO initialization has succeeded or
failed.
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Santosh Shukla <sshukla@mvista.com>
Tested-by: Santosh Shukla <sshukla@mvista.com>
In fedora 22 with GCC version 5.3.1, when compile,
will result an error:
include/rte_memcpy.h:309:7: error: "RTE_MACHINE_CPUFLAG_AVX2"
is not defined [-Werror=undef]
#elif RTE_MACHINE_CPUFLAG_AVX2
Fixes: 9484092baad3 ("eal/x86: optimize memcpy for AVX512 platforms")
Signed-off-by: Michael Qiu <michael.qiu@intel.com>
Acked-by: Zhihong Wang <zhihong.wang@intel.com>
