These lists can be initialized once and for all at build time.
With this, those lists are only manipulated in a common place
(and we could even make them private).
A nice side effect is that pci drivers can now register in constructors.
Signed-off-by: David Marchand <david.marchand@6wind.com>
Reviewed-by: Jan Viktorin <viktorin@rehivetech.com>
Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
rte_eal_dev_init is declared in both eal_private.h and rte_dev.h since its
introduction.
This function has been exported in ABI, so remove it from eal_private.h
Fixes: e57f20e051 ("eal: make vdev init path generic for both virtual and pci devices")
Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Reviewed-by: Jan Viktorin <viktorin@rehivetech.com>
An application might be linked to DPDK but not really use it,
so move the cpu flag check to the EAL initialization instead.
Signed-off-by: Flavio Leitner <fbl@sysclose.org>
Acked-by: Aaron Conole <aconole@redhat.com>
In ASLR-enabled system, it is possible that selected
virtual space is occupied by program segments. Therefore,
error path should not blindly unmap all memmory segments
but only those already mapped.
Steps that lead to crash:
1. memeseg 0 in secondary process overlaps with libc.so
2. mmap of /dev/zero fails for virtual space of memseg 0
3. munmap of memseg 0 leads to unmapping libc.so itself
4. app gets SIGSEGV after returning from syscall to libc
Fixes: ea329d7f8e ("mem: fix leak after mapping failure")
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
RTE_EAL_SINGLE_FILE_SEGMENTS was introduced with ivshmem integration.
Now that ivshmem was removed (commit c711ccb309)
and a simple git grep shows no one else references it;
I think we can now remove it.
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
When running single-core, some drivers tend to call rte_delay_us for a
long time, and that is causing packet drops.
To avoid this, rte_delay_us can be replaced with user-defined delay
function with:
void rte_delay_us_callback_register(void(*userfunc)(unsigned));
When userfunc==rte_delay_us_block build-in blocking delay function is
restored.
Signed-off-by: Jozef Martiniak <jozmarti@cisco.com>
Compile error:
.../lib/librte_eal/linuxapp/kni/kni_net.c:
In function ‘kni_net_rx_lo_fifo’:
.../lib/librte_eal/linuxapp/kni/kni_net.c:331:1:
error: the frame size of 1056 bytes is larger than 1024 bytes
[-Werror=frame-larger-than=]
This compile error seen with some compiler / kernel combinations.
Moved some local variables to the kni_dev struct.
Fixes: 8451269e6d ("kni: remove continuous memory restriction")
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Use mempool buf_addr and buf_physaddr fields for address translation.
Since each mbuf address calculated separately, the restriction of all
mbufs should come from a continuous memory restriction is no more valid.
mbuf related FIFO's content changed, rx_q and alloc_q now carries
physical address of mbufs. tx_q and free_q content not changed, they
still carries virtual address of mbufs.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Linux kernel v4.8 removes macro DEFINE_PCI_DEVICE_TABLE
Linux: 7e9321599011 ("treewide: remove references to the now unnecessary
DEFINE_PCI_DEVICE_TABLE")
Replaced macro with its value in kni ethtool drivers.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Compile error:
CC [M] .../build/lib/librte_eal/linuxapp/kni/igb_main.o
.../build/lib/librte_eal/linuxapp/kni/igb_main.c:
In function ‘igb_check_swap_media’:
.../build/lib/librte_eal/linuxapp/kni/igb_main.c:1556:7:
error: variable ‘link’ set but not used [-Werror=unused-but-set-variable]
bool link;
^
With Linux kernel >= v3.0 this warning disabled:
Linux: 8417da6f2128 ("kbuild: Fix passing -Wno-* options to gcc 4.4+")
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Add support for RHEL 7.3, which uses kernel 3.10,
but backported features from newer kernels.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Fix build error with Linux kernel >= v4.7
Fix compile error because of Linux API change, 'trans_start' field
removed from 'struct net_device'.
Linux: 9b36627acecd ("net: remove dev->trans_start")
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
There is no need for the page files to be readable (and executable) by
other users. This can be exploited by non-privileged users to access the
working memory of a DPDK app.
Open the files with 0600.
Signed-off-by: Robin Jarry <robin.jarry@6wind.com>
Exported header files used by applications should allow the strictest
compiler flags. Language extensions used in many places must be explicitly
marked to avoid warnings and compilation failures.
Unnamed structs/unions are allowed since C11, however many compiler
versions do not use this mode by default.
This commit prevents the following errors:
error: ISO C99 doesn't support unnamed structs/unions
error: struct has no named members
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Exported header files used by applications should allow the strictest
compiler flags. Language extensions used in many places must be explicitly
marked or removed to avoid warnings and compilation failures.
This commit prevents the following errors:
error: type of bit-field `[...]' is a GCC extension
Note: the standard does not require implementations to issue a diagnostic
message with these, and such errors do not occur with recent GCC or clang
versions. However, GCC 4.7 is still common and using the extension keyword
is easier than checking compiler version.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Exported header files used by applications should allow the strictest
compiler flags. Language extensions used in many places must be explicitly
marked or removed to avoid warnings and compilation failures.
The extension keyword is used whenever the C99 syntax cannot do it.
This commit prevents the following errors:
error: ISO C forbids zero-size array `[...]'
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Fix pernet calls when HAVE_SIMPLIFIED_PERNET_OPERATIONS is not set.
Fixes: e6734d21b4 ("kni: fix build with kernel 2.6.32")
Signed-off-by: Vincent Guo <guopengfei160@163.com>
Acked-by Ferruh Yigit <ferruh.yigit@intel.com>
Removing KNI interface that has no PCI driver for ethtool support cause
kernel crash.
Fixes: 109febfe58 ("net/igb: move PCI device IDs from EAL")
Fixes: 221fba3b98 ("net/ixgbe: move PCI device IDs from EAL")
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
License information is already in LICENSE.GPL.
Remove two extra copies and change referred filename in the files.
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
PCI device ids moved from common header into igb driver itself.
KNI starts using pci_device_id from kni/ethtool/igb driver, this is only
for KNI ethtool support, KNI data path is not affected.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
PCI device ids moved from common header into ixgbe driver itself.
KNI starts using pci_device_id from kni/ethtool/ixgbe driver, this is
only for KNI ethtool support, KNI data path is not affected.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Following discussions on the mailing list [1] and since nobody stood up to
implement the necessary cleanups, here is the ivshmem integration removal.
There is not much to say about this patch, a lot of code is being removed.
The default configuration file for packet_ordering example is replaced with
the "native" x86 file.
The only tricky part is in eal_memory with the memseg index stuff.
More cleanups can be done after this but will come in subsequent patchsets.
[1]: http://dpdk.org/ml/archives/dev/2016-June/040844.html
Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
The log history feature was deprecated in 16.07.
The remaining empty functions are removed in 16.11.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: David Marchand <david.marchand@6wind.com>
In rte_mem_virt2phy: Value returned from a function and indicating the
number of bytes was ignored. This could cause a wrong pfn (page frame
number) mask read from pagemap file.
When read returns less than the number of sizeof(uint64_t) bytes,
function rte_mem_virt2phy returns error.
Coverity issue: 13212
Fixes: 40b966a211 ("ivshmem: library changes for mmaping using ivshmem")
Signed-off-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
The offset of the 2nd mmap() when mapping the region after msix_bar
needs to take region address into consideration as mmap() takes
address that is resource-relative instead of bar-relative. This is
exposed when binding vmxnet3 to vfio-pci.
Fixes: 90a1633b23 ("eal/linux: allow to map BARs with MSI-X tables")
Signed-off-by: Yong Wang <yongwang@vmware.com>
Signed-off-by: Ronghua Zhang <rzhang@vmware.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
Having constructor function in the header file is generally
a bad idea, as it will eventually be implanted to 3rd party
library.
In this case it causes linking issues with 3rd party libraries
when an application is not linked to dpdk, due to missing
symbol called by constructor.
Fixes: ba7468997e ("spinlock: add HTM lock elision for x86")
Signed-off-by: Damjan Marion <damarion@cisco.com>
Reviewed-by: Jan Viktorin <viktorin@rehivetech.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
When using Xen Dom0, it looks that /proc/self/pagemap returns 0.
This breaks the creation of mbufs pool.
We can workaround this in rte_mem_virt2phy() by browsing the dpdk memory
segments. This only works for dpdk memory, but it's enough to fix the
mempool creation.
Fixes: c042ba2067 ("mempool: rework support of Xen dom0")
Fixes: 3097de6e6b ("mem: get physical address of any pointer")
Reported-by: Huilong Xu <huilongx.xu@intel.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
When building as shared library, the compiler complains for
undefined reference to `rte_xen_mem_phy2mch'
The symbol rte_xen_mem_phy2mch was introduced in DPDK 2.2
and has been called in mempool recently via rte_mem_phy2mch.
Fixes: c042ba2067 ("mempool: rework support of Xen dom0")
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Fix the compilation with CONFIG_RTE_LIBRTE_XEN_DOM0=y, by correcting the
typo in variable names.
Fixes: 8dab483701 ("xen: return machine address without knowing memseg id")
Reported-by: Huilong Xu <huilongx.xu@intel.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
We can now just OR the vfio_enabled sequentially and so adding new VFIO
subsystems (vfio_platform) is possible.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The module eal_pci_vfio_mp_sync is quite generic so it shouldn't contain the
"pci" string in its name. The internal functions don't need the pci_* prefix.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The vfio_cfg is a module-global variable and so together with this
variable, it is necessary to move functions:
* pci_vfio_get_group_fd
- renamed to vfio_get_group_fd
- pci_* version removed (no other call in EAL)
* pci_vfio_setup_device
- renamed as vfio_setup_device
* pci_vfio_enable
- renamed as vfio_enable
- generalized to check for a specific vfio driver presence
- pci_* specialization preserved as a wrapper
* pci_vfio_is_enabled
- renamed as vfio_is_enabled
- generalized to check for a specific vfio driver presence
to preserve the semantics of VFIO + PCI
- pci_* specialization preserved as a wrapper
* clear_current_group
- private function, just moved
To stop GCC complaining about "defined but not used", the private
function pci_vfio_get_group_no has been removed entirely.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The setup logic access the global vfio_cfg variable that will be moved in the
following commits. We need to separate all accesses to this variable to a
general code.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
The pci_vfio_set_iommu_type is not PCI-specific and it is a private function
of the eal_pci_vfio.c. We just rename the function and make it available even
for non-PCI devices.
The pci_vfio_has_supported_extensions is not PCI-specific and it is a private
function of the eal_pci_vfio.c. We just rename the function and make it
available even for non-PCI devices.
The pci_vfio_get_container_fd is not PCI-specific. Move the implementation to
the eal_vfio.c as vfio_get_container_fd. No other code seems to call this
function.
Generalize the pci_vfio_get_group_no to not be PCI-specific. Move the general
implementation to the eal_vfio.c as vfio_get_group_no and leave the original
pci_vfio_get_group_no being a wrapper around this to preserve compilation
issues. The pci_vfio_get_group_no function will be removed later.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
We make the iommu_types public temporarily here until the depending stuff is
refactored. The iommu_types and dma_map functions will be changed to be private
inside the eal_vfio module later.
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
mmap the iomem range of the PCI device fails for kernels that
enabled CONFIG_IO_STRICT_DEVMEM option:
EAL: pci_map_resource():
cannot mmap(39, 0x7f1c51800000, 0x100000, 0x0):
Invalid argument (0xffffffffffffffff)
CONFIG_IO_STRICT_DEVMEM is introduced in Linux v4.5 and not enabled
by default:
Linux commit: 90a545e restrict /dev/mem to idle io memory ranges
As a workaround igb_uio can stop reserving PCI memory resources, from
kernel point of view iomem region looks like idle and mmap works
again. This matches uio_pci_generic usage.
With this update device iomem range is not protected against any
other kernel drivers or userspace access. But this shouldn't
be a problem for dpdk usage module since purpose of the igb_uio
module is to provide userspace access.
Fixes: af75078fec ("first public release")
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Current code does not munmap 'hugepage' mapping (hugepage info file) on
function exit, leaking resources.
Coverity issue: 97920
Fixes: b6a468ad41 ("memory: add --socket-mem option")
Signed-off-by: Daniel Mrzyglod <danielx.t.mrzyglod@intel.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
This reverts commit 593a084afc.
Since recently [1], it is not possible to run the dpdk with
non-root privileges and the --no-huge option. This is because the eal
layer tries to lock the memory. Using locked memory is mandatory for
physical devices because they reference physical addresses.
But a user may want to start the dpdk without locked memory, because he
does not have the permission to do so, and/or does not have this need,
for instance because he uses virtual drivers.
So this commit reverts the use of MAP_LOCKED in mmap() flags.
[1] http://www.dpdk.org/ml/archives/dev/2016-May/039404.html
Fixes: 593a084afc ("mem: lock pages when not using hugepages")
Reported-by: Panu Matilainen <pmatilai@redhat.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
EAL memory init allocates all free hugepages of the whole system,
which seen from sysfs, even when applications do not ask so many.
When there is a limitation on how many hugepages an application can
use (such as cgroup.hugetlb), or hugetlbfs is specified with an
option of size (exceeding the quota of the fs), it just fails to
start even there are enough hugepages allocated.
To fix above issue, this patch:
- Changes the logic to continue memory init to see if hugetlb
requirement of application can be addressed by already allocated
hugepages.
- To make sure each hugepage is allocated successfully, we add a
recover mechanism, which relies on a mem access to fault-in
hugepages, and if it fails with SIGBUS, recover to previously
saved stack environment with siglongjmp().
For the case of CONFIG_RTE_EAL_SINGLE_FILE_SEGMENTS (enabled by
default when compiling IVSHMEM target), it's indispensable to
mapp all free hugepages in the system. Under this case, it fails
to start when allocating fails.
Test example:
a. cgcreate -g hugetlb:/test-subgroup
b. cgset -r hugetlb.1GB.limit_in_bytes=2147483648 test-subgroup
c. cgexec -g hugetlb:test-subgroup \
./examples/helloworld/build/helloworld -c 0x2 -n 4
Fixes: af75078fec ("first public release")
Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Tested-by: Yulong Pei <yulong.pei@intel.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Using gcc 6.1, in some cases, kni fails to compile
because of unused variables:
lib/librte_eal/linuxapp/kni/ixgbe_main.c:82:19:
error: ‘ixgbe_copyright’
defined but not used [-Werror=unused-const-variable=]
lib/librte_eal/linuxapp/kni/ixgbe_main.c:62:19:
error: ‘ixgbe_driver_string’
defined but not used [-Werror=unused-const-variable=]
Fixes: 3fc5ca2f63 ("kni: initial import")
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Following compile error observed with CentOS 6.8, which uses kernel
kernel-devel-2.6.32-642.el6.x86_64:
In function 'igbuio_msix_mask_irq':
error: 'PCI_MSIX_ENTRY_CTRL_MASKBIT' undeclared
Reported-by: Thiago Martins <thiagocmartinsc@gmail.com>
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
