drivers: replace packed attributes

MSVC struct packing is not compatible with GCC. Replace macro
__rte_packed with __rte_packed_begin to push existing pack value
and set packing to 1-byte and macro

drivers: replace packed attributes

MSVC struct packing is not compatible with GCC. Replace macro
__rte_packed with __rte_packed_begin to push existing pack value
and set packing to 1-byte and macro __rte_packed_end to restore
the pack value prior to the push.

Macro __rte_packed_end is deliberately utilized to trigger a
MSVC compiler warning if no existing packing has been pushed allowing
easy identification of locations where the __rte_packed_begin is
missing.

Signed-off-by: Andre Muezerie <andremue@linux.microsoft.com>
Reviewed-by: Tyler Retzlaff <roretzla@linux.microsoft.com>

show more ...

drivers: use portable variadic macros

Many places are using a GCC extension related to variadic macros,
where a name prepends the ellipsis. This results in a warning like
the one below when compilin

drivers: use portable variadic macros

Many places are using a GCC extension related to variadic macros,
where a name prepends the ellipsis. This results in a warning like
the one below when compiling the code with MSVC:

app\test-pmd\testpmd.h(1314): error C2608:
invalid token '...' in macro parameter list

Variadic macros became a standard part of the C language with C99.
GCC, Clang and MSVC handle them properly.

The fix is to remove the prefix name (args... becomes ...) and use
__VA_ARGS__.

Signed-off-by: Andre Muezerie <andremue@linux.microsoft.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>

show more ...

bus/fslmc: make IOVA mode configuration dynamic

IOVA mode should not be configured with CFLAGS because
1) User can perform "--iova-mode" to configure IOVA.
2) IOVA mode is determined by negotiation

bus/fslmc: make IOVA mode configuration dynamic

IOVA mode should not be configured with CFLAGS because
1) User can perform "--iova-mode" to configure IOVA.
2) IOVA mode is determined by negotiation between multiple devices.
Eal is in VA mode only when all devices support VA mode.

Hence:
1) Remove RTE_LIBRTE_DPAA2_USE_PHYS_IOVA cflags.
Instead, use rte_eal_iova_mode API to identify VA or PA mode.
2) Support memory IOMMU mapping and I/O IOMMU mapping(PCI space).
3) For memory IOMMU, in VA mode, IOVA:VA = 1:1;
in PA mode, IOVA:VA = PA:VA. The mapping policy is determined by
EAL memory driver.
4) For I/O IOMMU, IOVA:VA is up to I/O driver configuration.
In general, it's aligned with memory IOMMU mapping.
5) Memory and I/O IOVA tables are created and update when DMA
mapping is setup, which takes place of dpaax IOVA table.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Vanshika Shukla <vanshika.shukla@nxp.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

show more ...

dma/dpaa: add option for error checks

add user configurable DMA error checks.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

dma/dpaa: support scatter-gather

Support copy_sg operation for scatter gather.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

dma/dpaa: add QDMA stall workaround for ERR050265

Non-prefetchable read setting in the source descriptor may be
required for targets other than local memory. Prefetchable read
setting will offer bet

dma/dpaa: add QDMA stall workaround for ERR050265

Non-prefetchable read setting in the source descriptor may be
required for targets other than local memory. Prefetchable read
setting will offer better performance for misaligned transfers
in the form of fewer transactions and should be set if possible.
This patch also fixes QDMA stall issue due to unaligned
transactions.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

show more ...

dma/dpaa: add workaround for ERR050757

ERR050757 on LS104x indicates:

For outbound PCIe read transactions, a completion buffer is used
to store the PCIe completions till the data is passed back to

dma/dpaa: add workaround for ERR050757

ERR050757 on LS104x indicates:

For outbound PCIe read transactions, a completion buffer is used
to store the PCIe completions till the data is passed back to the
initiator. At most 16 outstanding transactions are allowed and
maximum read request is 256 bytes. The completion buffer size
inside the controller needs to be at least 4KB, but the PCIe
controller has 3 KB of buffer. In case the size of pending
outbound read transactions of more than 3KB, the PCIe controller
may drop the incoming completions without notifying the initiator
of the transaction, leaving transactions unfinished. All
subsequent outbound reads to PCIe are blocked permanently.
To avoid qDMA hang as it keeps waiting for data that was silently
dropped, set stride mode for qDMA.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

show more ...

dma/dpaa: support silent mode

add silent mode support.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

dma/dpaa: support burst capacity API

This patch improves the dpaa qdma driver and
adds dpaa_qdma_burst_capacity API which returns the
remaining space in the descriptor ring.

Signed-off-by: Jun Yang

dma/dpaa: support burst capacity API

This patch improves the dpaa qdma driver and
adds dpaa_qdma_burst_capacity API which returns the
remaining space in the descriptor ring.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

show more ...

dma/dpaa: refactor driver

This patch refactor the DPAA DMA driver code with changes:
- HW descriptors rename and update with details.
- update qdma engine and queue structures
- using rte_ring AP

dma/dpaa: refactor driver

This patch refactor the DPAA DMA driver code with changes:
- HW descriptors rename and update with details.
- update qdma engine and queue structures
- using rte_ring APIs for enqueue and dequeue.

Signed-off-by: Jun Yang <jun.yang@nxp.com>
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

show more ...

dma/dpaa2: move QDMA header to common driver

Include rte_pmd_dpaax_qdma.h instead of rte_pmd_dpaa2_qdma.h
and change code accordingly.

Signed-off-by: Jun Yang <jun.yang@nxp.com>

dma/dpaa2: change DMA copy return value

The return value of DMA copy/sg copy should be index of
descriptor copied in success.

Signed-off-by: Jun Yang <jun.yang@nxp.com>

dma/dpaa2: limit max descriptor number

For non-SG format, the index is saved in FD with
DPAA2_QDMA_FD_ATT_TYPE_OFFSET(13) bits width.

The max descriptor number of ring is power of 2, so the
eventua

dma/dpaa2: limit max descriptor number

For non-SG format, the index is saved in FD with
DPAA2_QDMA_FD_ATT_TYPE_OFFSET(13) bits width.

The max descriptor number of ring is power of 2, so the
eventual max number is:
((1 << DPAA2_QDMA_FD_ATT_TYPE_OFFSET) / 2)

Signed-off-by: Jun Yang <jun.yang@nxp.com>

show more ...

dma/dpaa2: support short FD

Short FD can be used for single transfer scenario which shows higher
performance than FLE.
1) Save index context in FD att field for short and FLE (NonSG).
2) Identify FD

dma/dpaa2: support short FD

Short FD can be used for single transfer scenario which shows higher
performance than FLE.
1) Save index context in FD att field for short and FLE (NonSG).
2) Identify FD type by att of FD.
3) Force 48 bits address for source address and fle according to spec.

Signed-off-by: Jun Yang <jun.yang@nxp.com>

show more ...

bus/fslmc: enhance QBMAN DQ storage logic

Multiple DQ storages are used among multiple cores, the single dq
storage of first union is leak if multiple storages are allocated.
It does not make sense

bus/fslmc: enhance QBMAN DQ storage logic

Multiple DQ storages are used among multiple cores, the single dq
storage of first union is leak if multiple storages are allocated.
It does not make sense to keep the single dq storage of union,
remove it and reuse the first dq storage of multiple storages
for this case.

Signed-off-by: Jun Yang <jun.yang@nxp.com>

show more ...

dma/dpaa2: refactor driver

refactor the driver code with changes in:
- multiple HW queues
- SMA single copy and SG copy
- silent mode

Signed-off-by: Jun Yang <jun.yang@nxp.com>

dma/dpaa2: configure route by port by PCIe port param

struct {
uint64_t coreid : 4; /**--rbp.sportid / rbp.dportid*/
uint64_t pfid : 8; /**--rbp.spfid / rbp.dpfid*/
uint64_t vfen : 1; /**--rbp.sv

dma/dpaa2: configure route by port by PCIe port param

struct {
uint64_t coreid : 4; /**--rbp.sportid / rbp.dportid*/
uint64_t pfid : 8; /**--rbp.spfid / rbp.dpfid*/
uint64_t vfen : 1; /**--rbp.svfa / rbp.dvfa*/
uint64_t vfid : 16; /**--rbp.svfid / rbp.dvfid*/
.....
} pcie;

Signed-off-by: Jun Yang <jun.yang@nxp.com>

show more ...

drivers: use bitops API instead of compiler builtins

Stop using directly __builtin_ bit operations,
prefer existing DPDK wrappers.

Note: this is a brute sed all over drivers (skipping base drivers)

drivers: use bitops API instead of compiler builtins

Stop using directly __builtin_ bit operations,
prefer existing DPDK wrappers.

Note: this is a brute sed all over drivers (skipping base drivers)
for __builtin_* that have a direct replacement in EAL bitops.
There is more work to do, like adding some missing macros inspired from
kernel (FIELD_*) macros but this is left for later.

Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

show more ...

dma/cnxk: support queue priority configuration

Allow configuration of DPI DMA queue priority through
mailbox request.

Signed-off-by: Vamsi Attunuru <vattunuru@marvell.com>

dma/idxd: set GRPCFG traffic class for better performance

Set GRPCFG traffic class to value of 1 for best performance on current
generation of accelerators. Applicable to gen1 and gen2 devices.

Sig

dma/idxd: set GRPCFG traffic class for better performance

Set GRPCFG traffic class to value of 1 for best performance on current
generation of accelerators. Applicable to gen1 and gen2 devices.

Signed-off-by: Shaiq Wani <shaiq.wani@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

show more ...

dma/idxd: fix free function mismatch in device probe

The data structure is allocated with rte_malloc and incorrectly
freed in cleanup logic using free.

Bugzilla ID: 1549
Fixes: 9449330a8458 ("dma/i

dma/idxd: fix free function mismatch in device probe

The data structure is allocated with rte_malloc and incorrectly
freed in cleanup logic using free.

Bugzilla ID: 1549
Fixes: 9449330a8458 ("dma/idxd: create dmadev instances on PCI probe")
Cc: stable@dpdk.org

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@huawei.com>
Acked-by: Wathsala Vithanage <wathsala.vithanage@arm.com>

show more ...

use C linkage where appropriate in headers

Assure that 'extern "C" { /../ }' do not cover files included from a
particular header file, and address minor issues resulting from this
change of order.

use C linkage where appropriate in headers

Assure that 'extern "C" { /../ }' do not cover files included from a
particular header file, and address minor issues resulting from this
change of order.

Dealing with C++ should delegate to the individual include file level,
rather than being imposed by the user of that file. For example,
forcing C linkage prevents __Generic macros being replaced with
overloaded static inline functions in C++ translation units.

Eliminate 'extern "C"' from files which do not declare any symbols
(e.g., only macros or struct types).

On the other hand, the headers check is too naive in assuming that all
headers must contain a 'extern "C"'. Such a check was added in commit
1ee492bdc4ff ("buildtools/chkincs: check missing C++ guards").
Since this current change results in many headers not containing such
a token, remove the check for 'extern "C"' until we have a better
implementation.

Signed-off-by: Mattias Rönnblom <mattias.ronnblom@ericsson.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
Signed-off-by: David Marchand <david.marchand@redhat.com>

show more ...

drivers: use per line logging in helpers

Use RTE_LOG(_DP)?_LINE(_PREFIX)? in existing macros that append a \n.

Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chengwen Feng <fen

drivers: use per line logging in helpers

Use RTE_LOG(_DP)?_LINE(_PREFIX)? in existing macros that append a \n.

Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chengwen Feng <fengchengwen@huawei.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Thomas Monjalon <thomas@monjalon.net>

show more ...

drivers: remove redundant newline from logs

Fix places where two newline characters may be logged.

Cc: stable@dpdk.org

Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chengwen

drivers: remove redundant newline from logs

Fix places where two newline characters may be logged.

Cc: stable@dpdk.org

Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chengwen Feng <fengchengwen@huawei.com>

show more ...

drivers: replace logging helpers

Prepare for the conversion to RTE_LOG_LINE.
Some drivers currently add no \n in the logging macro and
have the callers do it.

Let's invert this logic but again, thi

drivers: replace logging helpers

Prepare for the conversion to RTE_LOG_LINE.
Some drivers currently add no \n in the logging macro and
have the callers do it.

Let's invert this logic but again, thinking about backports, introduce
the change in new macros.

Signed-off-by: David Marchand <david.marchand@redhat.com>

show more ...