Jianfeng Tan [Mon, 12 Feb 2018 03:20:27 +0000 (03:20 +0000)]
net/virtio-user: fix start with kernel vhost
After reset owner in below patch, we failed to set owner before
sending further vhost messages. It is OK with vhost user implemented
DPDK/VPP/Contrail, but it sees "Operation not permitted" error when
used with vhost kernel.
We fix this by setting owner every time the device is started.
Fixes: 0d6a8752ac9d ("net/virtio-user: fix crash as features change") Cc: stable@dpdk.org Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Jianfeng Tan [Mon, 12 Feb 2018 03:20:26 +0000 (03:20 +0000)]
net/vhost: fix log messages on create/destroy
The original words are not accurate. For example, as destroy_device
callback gets called, it does not necessarily mean that the connection
is closed.
Fixes: ee584e9710b9 ("vhost: add driver on top of the library") Cc: stable@dpdk.org Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Maxime Coquelin [Mon, 12 Feb 2018 15:46:12 +0000 (16:46 +0100)]
vhost: do not take lock on owner reset
A deadlock happens when handling VHOST_USER_RESET_OWNER request
for the same reason the lock is not taken for
VHOST_USER_GET_VRING_BASE.
It is safe not to take the lock, as the queues are no more used
by the application when the virtqueues and the device are reset.
Fixes: a3688046995f ("vhost: protect active rings from async ring changes") Cc: stable@dpdk.org Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Tiwei Bie <tiwei.bie@intel.com> Reviewed-by: Jianfeng Tan <jianfeng.tan@intel.com>
Maxime Coquelin [Mon, 12 Feb 2018 15:46:11 +0000 (16:46 +0100)]
net/virtio: fix resuming port with Rx vector path
Since commit efc83a1e7fc3 ("net/virtio: fix queue setup consistency"),
when resuming a virtio port, the rx rings are refilled with new mbufs
until they are full (vq->vq_free_cnt == 0). This is done without
ensuring that the descriptor index remains a multiple of
RTE_VIRTIO_VPMD_RX_REARM_THRESH, which is a prerequisite when using the
vector mode. This can cause an out of bound access in the rx ring.
This commit changes the vector refill method from
virtqueue_enqueue_recv_refill_simple() to virtio_rxq_rearm_vec(), which
properly checks that the refill is done by batch of
RTE_VIRTIO_VPMD_RX_REARM_THRESH.
As virtqueue_enqueue_recv_refill_simple() is no more used, this
patch also removes the function.
Olivier Matz [Mon, 12 Feb 2018 15:46:10 +0000 (16:46 +0100)]
net/virtio: fix mbuf data offset for simple Rx
The mbuf->data_off was was not properly set for the first received
mbufs. Fix this by setting it in virtqueue_enqueue_recv_refill_simple(),
which is used to enqueue the first mbuf in the ring.
The function virtio_rxq_rearm_vec(), which is used to rearm the ring
with new mbufs, is valid and does not need to be updated.
Fixes: cab0461234e7 ("virtio: fill Rx avail ring with blank mbufs") Cc: stable@dpdk.org Signed-off-by: Olivier Matz <olivier.matz@6wind.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Matan Azrad [Sun, 11 Feb 2018 17:27:34 +0000 (17:27 +0000)]
net/failsafe: fix reconfiguration
Fail-safe PMD manages the states of its sub-devices gradually:
DEV_UNDEFINED, DEV_PARSED, DEV_PROBED, DEV_ACTIVE, DEV_STARTED.
When the sub-device arguments successfully was parsed, the state is
raised from DEV_UNDEFINED to DEV_PARSED.
When the sub-device successfully was probed, the state is raised from
DEV_PARSED to DEV_PROBED.
When the sub-device successfully was configured by
rte_eth_dev_configure(), the state is raised from DEV_PROBED to
DEV_ACTIVE.
When the sub-device successfully was started by rte_eth_dev_start(), the
state is raised from DEV_ACTIVE to DEV_STARTED.
When the sub-device successfully was stopped by rte_eth_dev_stop(), the
state is degraded from DEV_STARTED to DEV_ACTIVE.
When the sub-device successfully was closed by rte_eth_dev_close(), the
state is degraded from DEV_ACTIVE to DEV_PROBED.
When the sub-device successfully was removed by
rte_eal_hotplug_remove(), the state is degraded from DEV_PROBED to
DEV_UNDEFINED.
Fail-safe dev_configure() operation calls to its sub-devices
dev_configure() operation, but only for sub-devices which are in
DEV_PROBED state, means that sub-devices which are in DEV_ACTIVE state
because the application triggered dev_stop() operation cannot be
reconfigured again by dev_configure() operation which is really
problematic when application wants to reconfigure its ports.
Actually, the application may get success report when some of the
sub-devices are not in the wanted configuration.
The current behavior of fail-safe dev_configure() is correct only for
the first time dev_configure() is triggered by the application or for
sub-device synchronization in plug-in event, but it ignores the option
for reconfiguration from application side.
Allow calling to sub-devices dev_configure() operations also in
DEV_ACTIVE state when the call was triggered by the application.
Matan Azrad [Mon, 12 Feb 2018 20:51:42 +0000 (20:51 +0000)]
net/failsafe: fix hotplug races
Fail-safe uses a periodic alarm mechanism, running from the host
thread, to manage the hot-plug events of its sub-devices. This
management requires a lot of sub-devices PMDs operations
(stop, close, start, configure, etc.).
While the hot-plug alarm runs in the host thread, the application may
call fail-safe operations, which directly trigger the sub-devices PMDs
operations as well. This call may occur from any thread decided by the
application (probably the master thread).
Thus, more than one operation can be executed to a sub-device at the
same time. This can initiate a lot of races in the sub-PMDs.
Moreover, some control operations update the fail-safe internal
databases, which can be used by the alarm mechanism at the same time.
This can also initiate races and crashes.
Fail-safe is the owner of its sub-devices and must synchronize their
use according to the ETHDEV ownership rules.
Synchronize hot-plug management by a new lock mechanism uses a mutex to
atomically defend each critical section in the fail-safe hot-plug
mechanism and control operations to prevent any races between them.
Matan Azrad [Mon, 12 Feb 2018 20:51:41 +0000 (20:51 +0000)]
net/failsafe: fix removal scope
The fail-safe PMD uses a per sub-device flag called "remove" to
indicate the scope where the sub-device was removed physically and
whether its software resources should be released.
This flag is set when the fail-safe receives an RMV notification
about the physical removal of the sub-device, and should be unset when
all the sub-device resources are released.
The previous code wrongly unsets the flag in dev_configure(), instead
of when the software resources release is completed.
Change the remove flag unsetting to take action in the end of the
software resources release.
Matan Azrad [Mon, 12 Feb 2018 20:51:40 +0000 (20:51 +0000)]
net/failsafe: fix hotplug alarm cancel
The hot-plug alarm mechanism of fail-safe PMD is responsible for
handling removed devices during a plug-out event and to restore them
back to activity following a plug-in event.
Fail-safe sets a flag called "pending_alarm" to validate that only one
alarm callback is pending at any time. While this flag is required to
avoid simultaneous initiations of the alarm thread - it should not be
considered during alarm thread cancellation.
So, when failsafe_hotplug_alarm_cancel() was called while the alarm
callback was being executed the alarm mechanism was not stopped.
Skip checking the "pending_alarm" flag to allow alarm thread
cancellation all the times.
Ophir Munk [Tue, 13 Feb 2018 08:14:25 +0000 (08:14 +0000)]
net/tap: add CRC stripping capability
CRC stripping is executed in the kernel outside of TAP PMD scope.
There is no prevention that the TAP PMD will report on Rx CRC
stripping capability.
In the corrupted code, TAP PMD did not report on this capability.
The fix enables TAP PMD to report that Rx CRC stripping is supported.
Shahaf Shuler [Tue, 13 Feb 2018 12:31:55 +0000 (14:31 +0200)]
net/mlx5: revert multicast rule verbs flow type
This is to revert the following commits:
commit da646bd93888 ("net/mlx5: fix all multi verification code position")
commit 0a40a1363a4d ("net/mlx5: fix flow type for allmulti rules")
The last one introduced a bug in the following diff:
@ -1262,6 +1274,7 @@ struct ibv_spec_header {
eth.val.ether_type &= eth.mask.ether_type;
}
mlx5_flow_create_copy(parser, ð, eth_size);
+ parser->allmulti = eth.val.dst_mac[0] & 1;
return 0;
}
As broadcast rules will be considered of type allmulti as well.
The patch was originally intended to enable VF to receive all multicast
traffic by using the IBV_FLOW_ATTR_MC_DEFAULT flow type.
Since the support was removed from the kernel there is no point with
fixing this issue, hence the revert.
Fixes: da646bd93888 ("net/mlx5: fix all multi verification code position") Fixes: 0a40a1363a4d ("net/mlx5: fix flow type for allmulti rules") Cc: stable@dpdk.org Signed-off-by: Shahaf Shuler <shahafs@mellanox.com> Acked-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Martin Klozik [Tue, 13 Feb 2018 15:03:47 +0000 (15:03 +0000)]
ethdev: increase log level of port allocation failure
DPDK API does not propagate the reason of device allocation failure
from rte_eth_dev_allocate() up to the DPDK application (e.g. Open
vSwitch).
Log level of associated log entries was changed to warning. So user
can find additional details in log files also in production systems,
where debug messages cannot be turned on.
Signed-off-by: Martin Klozik <martinx.klozik@intel.com> Acked-by: Thomas Monjalon <thomas@monjalon.net>
Jerin Jacob [Mon, 12 Feb 2018 13:13:43 +0000 (18:43 +0530)]
ethdev: fix data alignment
The struct rte_eth_dev_data is used in ethdev fastpath routines
and it not aligned to cache line size. This patch fixes the ethdev
data alignment.
The alignment was broken from the "first public release" changeset
where ethdev data address was aligned only to the first port.
Remaining ports alignment was defined by the size of the struct
(rte_eth_dev_data). This scheme is not guaranteed to be cache line
aligned all the time.
"ethdev: add port ownership" change set introduced a
rte_eth_dev_shared_data container for port ownership change,
This resulted in rte_eth_dev->data memory for the first port also
as cache unaligned.
Added a compiler alignment attribute to make sure
rte_eth_dev->data always cache aligned so that CPU/compiler
1) Avoid sharing the element with another cache line
2) Can load/store the elements in struct rte_eth_dev_data as
naturally aligned.
Some platform like thunderX could see performance regression of 1%
at "ethdev: add port ownership" change set with
1 port/1 queue l3fwd application and this patch fixes that regression.
example command:
sudo ./examples/l3fwd/build/l3fwd -c 0xff00 -- -p 0x1 --config="(0,0,9)"
Fixes: af75078fece3 ("first public release") Fixes: 5b7ba31148a8 ("ethdev: add port ownership") Cc: stable@dpdk.org Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com> Acked-by: Thomas Monjalon <thomas@monjalon.net>
Lee Daly [Tue, 23 Jan 2018 16:45:29 +0000 (16:45 +0000)]
examples/l3fwd: convert Altivec files to SPDX license tag
This updates the Intel and IBM license on files in examples to be
the standard BSD-3-Clause license used for the rest of DPDK,
bringing the files in compliance with the DPDK licensing policy.
Lee Daly [Thu, 8 Feb 2018 14:03:54 +0000 (14:03 +0000)]
examples/performance-thread: convert license to SPDX tag
This updates the Intel and Cavium license on files in examples to be
the standard BSD-3-Clause license used for the rest of DPDK, bringing
the files in compliance with the DPDK licensing policy.
Signed-off-by: Lee Daly <lee.daly@intel.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
If build fails because of failed experimental check and stdout is
redirected to /dev/null, it is absolutely unclear why build fails.
Fixes: a4bcd61de82d ("buildtools: add script to check experimental API exports") Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com> Acked-by: Neil Horman <nhorman@tuxdriver.com>
This reverts commit 15692396fd68932b6a81f00f12d4b0da12baa7d3
(eal/ppc64: implement arch-specific TSC freq query).
We intended to derive pkt/sec estimation with cpu clock frequency.
As timebase register serves the timer purpose, we need to stick with it
for calculating pkt/sec, hence reverting the change.
Ferruh Yigit [Thu, 8 Feb 2018 12:08:03 +0000 (12:08 +0000)]
app/testpmd: enable CRC strip without capability check
Some hardware doesn't support disabling CRC strip. In techboard it has
been decided to enable CRC strip always.
The testpmd update in commit 8b9bd0efe0b6, enables CRC strip only if
PMD reports CRC strip capability. Not all PMDs updated to report CRC
strip.
For the PMDs not reporting CRC strip testpmd behavior changed and
disabling CRC strip for them. And this may generate error for PMDs that
doesn't support disabling CRC strip.
Removing capability check for this release. In long term there can be
option to remove CRC strip flag completely or adding a new flag to let
PMD say disabling is not supported.
Fixes: 8b9bd0efe0b6 ("app/testpmd: disable Rx VLAN offloads by default") Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Olivier Matz [Thu, 8 Feb 2018 08:24:17 +0000 (09:24 +0100)]
net/bnxt: fix packet type
The hw flags are not read correctly: the defines
RX_PKT_CMPL_FLAGS_ITYPE_* are not bits but values, so the should not be
tested with if (value & X) but with if ((value & MASK) == X).
This was resulting in a wrong packet type.
For instance, an IPv4/ICMP packet was returning a value of 7 for
the layer 4, which is undefined.
This patch rework the way packet types are processed, to ensure
that only valid packet types will be advertised.
Olivier Matz [Tue, 6 Feb 2018 13:39:31 +0000 (14:39 +0100)]
net/bnxt: fix Rx checksum flags
Fix the Rx offload flags when the IP or L4 checksum is seen as incorrect
by the hardware. In this case, the proper value is PKT_RX_IP_CKSUM_BAD.
PKT_RX_IP_CKSUM_NONE means that the checksum may be incorrect in the
packet headers, but the integrity of the IP header is verified. This is
mostly useful for virtual drivers.
Fixes: 7ec39d8c524b ("net/bnxt: update status of Rx IP/L4 CKSUM") Cc: stable@dpdk.org Signed-off-by: Olivier Matz <olivier.matz@6wind.com> Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Moti Haimovsky [Mon, 5 Feb 2018 16:17:20 +0000 (18:17 +0200)]
net/tap: fix cleanup on allocation failure
This patch complements the partial cleanup done inside
eth_dev_tap_create when the routine failed.
Such a failure left a non-functional device attached to the system.
Fixes: 050fe6e9ff97 ("drivers/net: use ethdev allocation helper for vdev") Cc: stable@dpdk.org Signed-off-by: Moti Haimovsky <motih@mellanox.com> Acked-by: Pascal Mazon <pascal.mazon@6wind.com>
Xiaoyun Li [Thu, 8 Feb 2018 05:45:15 +0000 (13:45 +0800)]
net/ixgbe: fix CRC stripping capability
testpmd needs to get CRC strip offload from rx_offload_capa with a
recent change. This patch adds CRC strip flag to rx_offload_capa
for ixgbe and ixgbevf, as they have the capability.
Fixes: 8b9bd0efe0b6 ("app/testpmd: disable Rx VLAN offloads by default") Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Neil Horman [Thu, 1 Feb 2018 12:19:11 +0000 (07:19 -0500)]
compat: relicense some files
Received a note the other day from the Linux Foundation governance board
for DPDK indicating that several files I have copyright on need to be
relicensed to be compliant with the DPDK licensing guidelines. I have
some concerns with some parts of the request, but am not opposed to
other parts. So, for those pieces that we are in consensus on, I'm
proposing that we change their license from BSD 2 clause to 3 clause.
I'm also updating the files to use the SPDX licensing scheme
Signed-off-by: Neil Horman <nhorman@tuxdriver.com> Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Vipin Varghese [Mon, 5 Feb 2018 17:32:07 +0000 (23:02 +0530)]
doc: update minimum requirement of l2fwd-crypto
Added note section to update information for use cases working with
multiple crypto devices.
Signed-off-by: Vipin Varghese <vipin.varghese@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Nikhil Rao [Sun, 4 Feb 2018 18:18:31 +0000 (23:48 +0530)]
eventdev: check error in default Rx conf callback
The default adapter configuration callback is invoked when a Rx
queue is added to the adapter and the adapter detects that a SW
service is needed. The adapter needs to re-configure the device
with an additional port and to do do, it needs to stop the
device and restart it after it is done reconfiguring it. This
patch adds code to check the return code of
rte_event_dev_start() for both when the reconfiguration fails
and when it succeeds and introduces a new error code (-EIO)
for the first case.
Beilei Xing [Tue, 6 Feb 2018 05:33:31 +0000 (13:33 +0800)]
net/i40e: fix Rx interrupt
This patch fixes interval error and corrects macros when
enabling Rx interrupt mode. The patch also fixes a logical
error during supporting multiple drivers.
Should be not possible to load conflicting DDP profiles. Only DDP
profiles of the same group (not 0) can be loaded together. If DDP
profile group is 0, it is exclusive, i.e. it cannot be loaded with
any other DDP profile. If DDP profile groups are different, these
profiles cannot be loaded together.
Fixes: b319712f53c8 ("net/i40e: extended list of operations for DDP processing") Cc: stable@dpdk.org Signed-off-by: Kirill Rybalchenko <kirill.rybalchenko@intel.com> Acked-by: Andrey Chilikin <andrey.chilikin@intel.com> Acked-by: Beilei Xing <beilei.xing@intel.com>
Bao-Long Tran [Wed, 31 Jan 2018 08:48:16 +0000 (15:48 +0700)]
examples/ip_pipeline: fix timer period unit
The timer_period option specified by users via config file
should have unit of 1 millisecond. However timer_period is
internally converted to unit of 10 millisecond.
When querying either the name or the value of a stat using the xstats
APIs, a check is done to see if the regular stats API or the xstats APIs
for the driver need to be used. However, the id of the stat requested is
checked to see if it is greater than the number of basic stats, rather
than checking for greater-or-equal, meaning that the xstat with the lowest
id gets incorrectly treated as a basic stat.
This problem manifests itself when you call proc_info using "--xstats-id"
for the first xstat, you get no name of the stat printed, and a random(ish)
stat value.
Fixes: 4773152f850b ("ethdev: optimize xstats by ids APIs") Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Jerin Jacob [Tue, 6 Feb 2018 17:14:55 +0000 (22:44 +0530)]
bus/fslmc: fix build with latest glibc
Fix the following build error with latest glibc-headers(
Part of aarch64-linux-gnu-gcc-7.3.0 toolchain)
In file included from drivers/bus/fslmc/mc/fsl_mc_sys.h:30:0,
from drivers/bus/fslmc/mc/dpmng.c:7:
/usr/aarch64-linux-gnu/usr/include/libio.h:21:2: error:
#warning "<libio.h> is deprecated; use <stdio.h> instead." [-Werror=cpp]
Fixes: 3af733ba8da8 ("bus/fslmc: introduce MC object functions") Cc: stable@dpdk.org Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Matan Azrad [Mon, 5 Feb 2018 14:09:22 +0000 (14:09 +0000)]
app/testpmd: fix port topology in RSS forward config
The testpmd user can configure port topology mode to define the port
topology between the testpmd forward ports(paired, chained and loop).
When multi-queue ports are configured by the user, the testpmd
streams are created by rss_fwd_config_setup() function, this function
doesn't take into account the chained topology mode and configures the
forward streams with paired topology mode in this case.
Configure the stream Tx port by dedicated function which calculates
a valid Tx port index as a function of the topology mode and the Rx
port index.
Matan Azrad [Mon, 5 Feb 2018 14:09:21 +0000 (14:09 +0000)]
app/testpmd: fix port index in RSS forward config
When multi-queue ports are configured by the user, the testpmd streams
are created by rss_fwd_config_setup() function.
This function may configure to the streams either invalid Rx ports or
invalid Tx ports.
An invalid Tx port is configured when the number of ports is odd.
In this case, the last Tx port will be always invalid.
An invalid Rx port is configured when NUMA support is configured by the
user and the number of forward ports is much smaller than the number of
all ports. In this case, also the Tx port is invalid.
Shreyansh Jain [Tue, 6 Feb 2018 06:43:38 +0000 (12:13 +0530)]
raw/skeleton: fix device start test
Device can only be started if firmware is loaded, as per Skeleton
rawdev driver semantics. This patch fixes original implementation
which attempted to start the device without loading firmware.
Herbert Guan [Wed, 31 Jan 2018 07:39:19 +0000 (15:39 +0800)]
build: add more implementers IDs and PNs for ARM
1) Add native PN option '-march=native' to allow automatic detection.
Set 'arm_force_native_march' to 'true' in config/arm/meson.build
to use native PN option.
2) Add implementer_pn option for part num selection in cross compile
3) Add known Arm cortex PN support
4) Add known implementers' IDs (use generic flags/archs by default)
5) Sync build options with config/common_armv8a_linuxapp
Signed-off-by: Herbert Guan <herbert.guan@arm.com> Acked-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Since rdma-core glue libraries are intrinsically tied to their respective
PMDs and used as internal plug-ins, their presence in the default search
path among other system libraries for the dynamic linker is not necessarily
desired.
This commit enables their installation and subsequent look-up at run time
in RTE_EAL_PMD_PATH if configured to a nonempty string. This path can also
be overridden by environment variables MLX[45]_GLUE_PATH.
When built as separate objects, these libraries do not have unique names.
Since they do not maintain a stable ABI, loading an incompatible library
may result in a crash (e.g. in case multiple versions are installed).
This patch addresses the above by versioning glue libraries, both on the
file system (version suffix) and by comparing a dedicated version field
member in glue structures.
David Harton [Sat, 3 Feb 2018 14:55:23 +0000 (09:55 -0500)]
net/virtio: fix crash while freeing mbufs
virtio_dev_free_mbufs was recently modified to free the
virtqueues but failed to check whether the array was
allocated. Added a check to ensure vqs was non-null.
Fixes: bdb32afbb610 ("net/virtio: rationalize queue flushing") Signed-off-by: David Harton <dharton@cisco.com> Reviewed-by: Olivier Matz <olivier.matz@6wind.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Ophir Munk [Mon, 5 Feb 2018 14:40:42 +0000 (14:40 +0000)]
net/tap: fix eBPF handling of non-RSS flows
The eBPF classifier (section "cls_q" in tap_bpf_program.c) is tracing
marked packets in which skb->cb[1] contains an RSS queue number, and
redirects those packets to the matched queue.
It is expected that skb->cb[1] has been previously set with a valid RSS
queue number during an eBPF action (section "l3_l4" in tap_bpf_program.c).
However, for non-RSS flows, skb->cb[1] may contain a random unset value,
which could falsely be interpreted as a valid RSS queue.
To avoid this potential error, tap_bpf_program.c has been updated as
follows:
1. After calculating the RSS queue number, it is added a unique offset in
order to uniquely identify it as a valid RSS queue number.
2. After matching an RSS queue to a packet, skb->cb[1] is set to 0.
Ophir Munk [Mon, 5 Feb 2018 10:59:08 +0000 (10:59 +0000)]
net/tap: fix multi segments capability
TAP device is supporting multi segments Tx, however this capability is
not reported when querying the TAP device.
This commit adds this capability report.
Fixes: 818fe14a9891 ("net/tap: use new Tx offloads API") Cc: stable@dpdk.org Signed-off-by: Ophir Munk <ophirmu@mellanox.com> Acked-by: Pascal Mazon <pascal.mazon@6wind.com>
Maxime Coquelin [Mon, 5 Feb 2018 15:04:56 +0000 (16:04 +0100)]
vhost: fix IOTLB pool out-of-memory handling
In the unlikely case the IOTLB memory pool runs out of memory,
an issue may happen if all entries are used by the IOTLB cache,
and an IOTLB miss happen. If the iotlb pending list is empty,
then no memory is freed and allocation fails a second time.
This patch fixes this by doing an IOTLB cache random evict if
the IOTLB pending list is empty, ensuring the second allocation
try will succeed.
In the same spirit, the opposite is done when inserting an
IOTLB entry in the IOTLB cache fails due to out of memory. In
this case, the IOTLB pending is flushed if the IOTLB cache is
empty to ensure the new entry can be inserted.
Fixes: d012d1f293f4 ("vhost: add IOTLB helper functions") Fixes: f72c2ad63aeb ("vhost: add pending IOTLB miss request list and helpers") Cc: stable@dpdk.org Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
git.droids-corp.org / dpdk.git / log