> From: Adrien Mazarguil
>
> When NUMA is enabled, the per-core fwd_lcore structure mbuf pool pointer
> is not set, causing a crash.
>
> Initialize fwd_lcore after allocating NUMA memory pools.
>
> Signed-off-by: Adrien Mazarguil
Acked-by: Thomas Monjalon
Applied for version 1.6.0r2.
> From: Maxime Leroy
>
> vdev ethdev can not be allocated on a numa socket that is not socket 0.
> The reason comes from rte_eth_dev_allocate() which uses rte_socket_id() to
> identify the socket on which vdev driver data should be allocated.
> However, at this initialization step,
On Thu, May 01, 2014 at 08:53:02AM +0200, Thomas Monjalon wrote:
> 2014-04-30 11:22, Neil Horman:
> > On Wed, Apr 30, 2014 at 01:09:38PM +0200, Thomas Monjalon wrote:
> > > The 4 spec files are used to build 4 different git trees with their own
> > > versioning:
> > > http://dpdk.org/browse
> >
This patchset adds support for using VFIO instead of IGB_UIO to
map the device BARs.
VFIO is a kernel 3.6+ driver allowing secure DMA from userspace
by means of using IOMMU instead of working directly with physical
memory like igb_uio does.
Short summary:
* Adding support for VFIO in EAL PCI
Currently, EAL does not distinguish between actual failures and
expected initialization errors. E.g. sometimes the driver fails to
initialize because it was not supposed to be initialized in the
first place, such as device not being managed by said driver.
This patch makes EAL fail on actual
Rename the RTE_PCI_DRV_NEED_IGB_UIO to be more generic,
retain old macro for backwards compatibility. Probably should
be removed in one of the next releases.
Signed-off-by: Anatoly Burakov
diff --git a/app/test/test_pci.c b/app/test/test_pci.c
index 445cf57..e13f635 100644
---
In order to make the code a bit more clean while using multiple
drivers, IGB_UIO mapping has been separated into its own file.
Signed-off-by: Anatoly Burakov
create mode 100644 lib/librte_eal/linuxapp/eal/eal_pci_uio.c
create mode 100644 lib/librte_eal/linuxapp/eal/include/eal_pci_init.h
Moving interrupt type enum out of igb_uio and renaming it to be more
generic. Such a strange header naming and separation is done mostly to
make coming virtio patches easier to port to dpdk.org tree.
Signed-off-by: Anatoly Burakov
create mode 100644
Make VFIO compilation optional for all configs.
Signed-off-by: Anatoly Burakov
diff --git a/config/defconfig_i686-default-linuxapp-gcc
b/config/defconfig_i686-default-linuxapp-gcc
index ea90f12..5410f57 100644
--- a/config/defconfig_i686-default-linuxapp-gcc
+++
Currently, igb_uio is always compiled. Since no Linux distribution
would want to include igb_uio by default or even with DPDK, we need
to make sure that igb_uio compilation can be optional (i.e. some
distributions could ship DPDK with VFIO support only).
Signed-off-by: Anatoly Burakov
diff
Unlike igb_uio, VFIO interrupt type is not set by kernel module
parameters but is set up via ioctl() calls at runtime. This warrants
a new EAL command-line parameter. It will have no effect if VFIO is
not compiled, but will set VFIO interrupt type to either "legacy" or
"msix" if VFIO support is
Note that since igb_uio no longer has a PCI ID list, it can now be
bound to any device, not just those explicitly supported by DPDK. In
other words, it now behaves similar to PCI stub, VFIO and other generic
PCI drivers.
Therefore to bind a new device to igb_uio, the user will now have to
first
This makes it possible to run DPDK without hugepage memory when VFIO
is used, as VFIO uses virtual addresses to set up DMA mappings.
Signed-off-by: Anatoly Burakov
diff --git a/lib/librte_eal/linuxapp/eal/eal_memory.c
b/lib/librte_eal/linuxapp/eal/eal_memory.c
index 73a6394..b075145 100644
---
Creating code to handle VFIO interrupts in EAL interrupts, and also
adding a header eal_vfio.h.
This header checks two things:
* checks if CONFIG_RTE_EAL_VFIO was enabled during build time
* checks that kernel version is 3.6+ so that DPDK would still compile
on older kernels despite VFIO
Adding unit tests for VFIO interrupt type command-line parameter.
We don't know if VFIO is compiled (eal_vfio.h header is internal
to Linuxapp EAL), we check this flag regardless.
Also, quick-fixed a bug in base_virtaddr parsing that prevented it
from working (making unit test fail), will later
Since igb_uio no longer has a PCI ID list, the script will no
longer distinguish between supported and unsupported NICs.
There's a weird behaviour of sysfs when a new device ID is added to
new_id. Subsequent writing to "bind" will result in IOError on
closing the file. This error is harmless but
Support for loading/unloading VFIO drivers, binding/unbinding
devices to/from VFIO, also setting up correct userspace permissions.
Signed-off-by: Anatoly Burakov
diff --git a/tools/setup.sh b/tools/setup.sh
index 39be8fc..2ffa55a 100755
--- a/tools/setup.sh
+++ b/tools/setup.sh
@@ -187,6
On Wed, Apr 30, 2014 at 02:46:41AM +0200, Thomas Monjalon wrote:
> The goal of this patch serie is to be able to package DPDK
> for RPM-based distributions.
>
> The file naming currently doesn't allow to install different DPDK versions.
> But the packaging naming should be ready to manage
Hi David,
> Looking at bsd implementation, we can see that there is a potential mem
> leak in linux implementation. Fix this.
>
> Signed-off-by: David Marchand
> ---
> lib/librte_eal/linuxapp/eal/eal_pci.c |1 +
> 1 file changed, 1 insertion(+)
>
> diff --git
2014-05-01 06:28, Neil Horman:
> On Thu, May 01, 2014 at 08:53:02AM +0200, Thomas Monjalon wrote:
> > 2014-04-30 11:22, Neil Horman:
> > > On Wed, Apr 30, 2014 at 01:09:38PM +0200, Thomas Monjalon wrote:
> > > > The 4 spec files are used to build 4 different git trees with their
> > > > own
> > >
> -Original Message-
> From: Konrad Rzeszutek Wilk [mailto:konrad.wilk at oracle.com]
> Sent: Thursday, May 01, 2014 6:07 PM
> To: Samuel Monderer
> Cc: xen-devel at lists.xen.org; Shimon Zadok; Eyal Ben Saadon; Andrew
> Sergeev
> Subject: Re: [Xen-devel] dpdk kernel module when calling
> From: Didier Pallard
>
> In current version, coremask can only be specified using a bitmask.
> It will now be possible to specify core masks in 2 different ways:
> - Using a bitmask (-c 0xnnn): bitmask must be in hex format and start with 0x
> - Using a core set description in following
Hi David, Didier,
> + /* Set memory amount per socket; round up to be sure that
> sum of all */
> + /* sockets allocation is greater than requested memory size
> */
> + for (socket_id=0 ; socket_id socket_id++) {
> +
On Thu, 1 May 2014 11:05:38 +
"Burakov, Anatoly" wrote:
> This patchset adds support for using VFIO instead of IGB_UIO to
> map the device BARs.
>
> VFIO is a kernel 3.6+ driver allowing secure DMA from userspace
> by means of using IOMMU instead of working directly with physical
> memory
On Wed, Apr 30, 2014 at 10:57:00PM +0200, Thomas Monjalon wrote:
> If some DPDK libraries are installed on the system, the linker was trying
> to use them before searching in -L path.
> The obscure reason is that we were prefixing -L with -Wl, to pass it
> directly to the linker.
> But -L is also
On Thu, 1 May 2014 11:06:12 +
"Burakov, Anatoly" wrote:
>
> #
> +# Unloads VFIO modules.
> +#
> +remove_vfio_module()
> +{
> + echo "Unloading any existing VFIO module"
> + /sbin/lsmod | grep -s vfio > /dev/null
> + if [ $? -eq 0 ] ; then
> + sudo /sbin/rmmod
2014-04-30 11:22, Neil Horman:
> On Wed, Apr 30, 2014 at 01:09:38PM +0200, Thomas Monjalon wrote:
> > The 4 spec files are used to build 4 different git trees with their own
> > versioning:
> > http://dpdk.org/browse
> >
> > So I think it's saner to keep them in their repository.
[...]
>
>
27 matches
Mail list logo
