dma/idxd: move config script from raw driver

Move the example script for configuring IDXD devices bound to the IDXD
kernel driver from raw to dma, and create a symlink to still allow use from
raw.

dma/idxd: move config script from raw driver

Move the example script for configuring IDXD devices bound to the IDXD
kernel driver from raw to dma, and create a symlink to still allow use from
raw.

Signed-off-by: Conor Walsh <[email protected]>
Signed-off-by: Kevin Laatz <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: build only if dmadev not present

Only build the rawdev IDXD/IOAT drivers if the dmadev drivers are not
present.

This change requires the dependencies to be reordered in
drivers/meson.buil

raw/ioat: build only if dmadev not present

Only build the rawdev IDXD/IOAT drivers if the dmadev drivers are not
present.

This change requires the dependencies to be reordered in
drivers/meson.build so that rawdev can use the "RTE_DMA_* build macros to
check for the presence of the equivalent dmadev driver.

A note is also added to the documentation to inform users of this change.

Signed-off-by: Bruce Richardson <[email protected]>
Signed-off-by: Kevin Laatz <[email protected]>
Reviewed-by: Conor Walsh <[email protected]>

show more ...

version: 21.11-rc0

Start a new release cycle with empty release notes.

The ABI version becomes 22.0.
The map files are updated to the new ABI major number (22).
The ABI exceptions are dropped and C

version: 21.11-rc0

Start a new release cycle with empty release notes.

The ABI version becomes 22.0.
The map files are updated to the new ABI major number (22).
The ABI exceptions are dropped and CI ABI checks are disabled because
compatibility is not preserved.

Signed-off-by: Thomas Monjalon <[email protected]>
Acked-by: Ferruh Yigit <[email protected]>
Acked-by: David Marchand <[email protected]>

show more ...

raw/ioat: fix termination descriptor for batch

When batch_size == 1, idxd has to add a dummy termination descriptor
to satisfy HW requirements.
Right now it uses NOP descriptor with FENCE flag.
This

raw/ioat: fix termination descriptor for batch

When batch_size == 1, idxd has to add a dummy termination descriptor
to satisfy HW requirements.
Right now it uses NOP descriptor with FENCE flag.
This is excessive and fencing can slowdown things quite significantly.
The patch removes FENCE flag from termination dummy descriptor.
That helps to improve performance for no-burst scenarios.

Fixes: 245efe544d8e ("raw/ioat: report status of completed jobs")

Signed-off-by: Konstantin Ananyev <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: fix config script queue size calculation

The queue size calculation is currently based on "max_tokens" rather than
"max_work_queues_size". This is resulting in the queue size being
incorre

raw/ioat: fix config script queue size calculation

The queue size calculation is currently based on "max_tokens" rather than
"max_work_queues_size". This is resulting in the queue size being
incorrectly configured when using the script to configure devices bound to
the IDXD kernel driver.
This patch fixes this miscalculation so devices are configured with
appropriate queue size.

Fixes: 01863b9d2354 ("raw/ioat: include example configuration script")
Cc: [email protected]

Reported-by: Konstantin Ananyev <[email protected]>
Signed-off-by: Kevin Laatz <[email protected]>
Acked-by: Konstantin Ananyev <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: fix missing ring pointer reset

In the event of a device reconfigure, "hdls_avail" is not being reset. This
can lead to miscalculations in rte_ioat_completed_ops(), causing the
function to

raw/ioat: fix missing ring pointer reset

In the event of a device reconfigure, "hdls_avail" is not being reset. This
can lead to miscalculations in rte_ioat_completed_ops(), causing the
function to report an incorrect amount of completed operations. This patch
fixes the issue by resetting "hdls_avail" during the device configure.

Fixes: 74464005a2af ("raw/ioat: rework SW ring layout")
Cc: [email protected]

Signed-off-by: Kevin Laatz <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: fix memory leak in device configure

During device configure, memory is allocated for "hdl_ring_flags". In the
event of another call to the device configure function (reconfigure), a
memory

raw/ioat: fix memory leak in device configure

During device configure, memory is allocated for "hdl_ring_flags". In the
event of another call to the device configure function (reconfigure), a
memory leak would occur. This patch fixes the memory leak by free'ing the
memory before reallocating it.

Fixes: 245efe544d8e ("raw/ioat: report status of completed jobs")
Cc: [email protected]

Signed-off-by: Kevin Laatz <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: add device reset to configuration script

Currently once a device is configured, the user does not have the ability
to reset the device via the script.

This patch adds a device reset optio

raw/ioat: add device reset to configuration script

Currently once a device is configured, the user does not have the ability
to reset the device via the script.

This patch adds a device reset option to the script. For example
"$dpdk_idxd_cfg.py 0 --reset" would reset device 0.

Signed-off-by: Kevin Laatz <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: handle PCI address in configuration script

Currently the user needs to find the DSA instance number for any DSA device
they would like to configure using this script, which can be cumberso

raw/ioat: handle PCI address in configuration script

Currently the user needs to find the DSA instance number for any DSA device
they would like to configure using this script, which can be cumbersome and
error-prone since the instance numbering may change when changing the
binding of the devices between vfio-pci and idxd.

This patch improves the usability of the script by adding the ability to
specify the DSA device to configure using the device's PCI address instead
of the DSA instance number. For example, "$dpdk_idxd_cfg.py 0" and
"$dpdk_idxd_cfg.py 6a:01.0" are both valid references to the same device
(assuming the numbering).

Signed-off-by: Kevin Laatz <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: fix missing device name in idxd bus scan

The device name is not being initialized during the idxd bus scan which
will cause segmentation faults when an appliation tries to access this
info

raw/ioat: fix missing device name in idxd bus scan

The device name is not being initialized during the idxd bus scan which
will cause segmentation faults when an appliation tries to access this
information.

This patch adds the required initialization of the device name so that it
can be read without issues.

Fixes: b7aaf417f936 ("raw/ioat: add bus driver for device scanning automatically")
Cc: [email protected]

Reported-by: Sunil Pai G <[email protected]>
Signed-off-by: Kevin Laatz <[email protected]>
Acked-by: Bruce Richardson <[email protected]>
Tested-by: Sunil Pai G <[email protected]>

show more ...

raw/ioat: remove special case for no status reporting

The special fast-path for returning completed descriptors without
reporting status or user-handles returns the number of completed ring
slots us

raw/ioat: remove special case for no status reporting

The special fast-path for returning completed descriptors without
reporting status or user-handles returns the number of completed ring
slots used, rather than the number of actual user-submitted jobs. This
means that the counts returned are too high, as the batch descriptor
slots would be included in the total. Therefore remove this special
case, and use the normal status-processing path so that the returned
count is correct in all cases.

Fixes: 245efe544d8e ("raw/ioat: report status of completed jobs")

Reported-by: Jiayu Hu <[email protected]>
Signed-off-by: Bruce Richardson <[email protected]>
Acked-by: Kevin Laatz <[email protected]>

show more ...

raw/ioat: fix ring space checks

When enqueuing a descriptor, when checking that there is at least one
slot free for the current descriptor and a later batch descriptor, we
need to test for both two

raw/ioat: fix ring space checks

When enqueuing a descriptor, when checking that there is at least one
slot free for the current descriptor and a later batch descriptor, we
need to test for both two free and one free, in case the last write
was a batch descriptor which is allowed to use the "spare" slot.

Similarly, when computing the free space in the ring to return to the
user, we need to take account of the same condition, so that we do not
return a "-1" ring space value, by blindly subtracting "2".

Fixes: 245efe544d8e ("raw/ioat: report status of completed jobs")

Reported-by: Jiayu Hu <[email protected]>
Signed-off-by: Bruce Richardson <[email protected]>
Acked-by: Kevin Laatz <[email protected]>

show more ...

raw/ioat: fix parameter shadow warning

In the function __idxd_completed_ops() we have a parameter shadow warning
due to a local variable having the same name as one of the function
parameters. This

raw/ioat: fix parameter shadow warning

In the function __idxd_completed_ops() we have a parameter shadow warning
due to a local variable having the same name as one of the function
parameters. This issue is fixed by simply renaming the local variable.

This warning was caught when additions were made to the OVS codebase,
which include adding calls the IOAT APIs. The OVS build passes the
-Wshadow flag by default, allowing the warning to be seen when building
OVS with DPDK 21.05-rc2.

Fixes: 245efe544d8e ("raw/ioat: report status of completed jobs")

Reported-by: Sunil Pai G <[email protected]>
Signed-off-by: Kevin Laatz <[email protected]>
Tested-by: Sunil Pai G <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

log: register with standardized names

Let's try to enforce the convention where most drivers use a pmd. logtype
with their class reflected in it, and libraries use a lib. logtype.

Introduce two new

log: register with standardized names

Let's try to enforce the convention where most drivers use a pmd. logtype
with their class reflected in it, and libraries use a lib. logtype.

Introduce two new macros:
- RTE_LOG_REGISTER_DEFAULT can be used when a single logtype is
used in a component. It is associated to the default name provided
by the build system,
- RTE_LOG_REGISTER_SUFFIX can be used when multiple logtypes are used,
and then the passed name is appended to the default name,

RTE_LOG_REGISTER is left untouched for existing external users
and for components that do not comply with the convention.

There is a new Meson variable log_prefix to adapt the default name
for baseband (pmd.bb.), bus (no pmd.) and mempool (no pmd.) classes.

Note: achieved with below commands + reverted change on net/bonding +
edits on crypto/virtio, compress/mlx5, regex/mlx5

$ git grep -l RTE_LOG_REGISTER drivers/ |
while read file; do
pattern=${file##drivers/};
class=${pattern%%/*};
pattern=${pattern#$class/};
drv=${pattern%%/*};
case "$class" in
baseband) pattern=pmd.bb.$drv;;
bus) pattern=bus.$drv;;
mempool) pattern=mempool.$drv;;
*) pattern=pmd.$class.$drv;;
esac
sed -i -e 's/RTE_LOG_REGISTER(\(.*\), '$pattern',/RTE_LOG_REGISTER_DEFAULT(\1,/' $file;
sed -i -e 's/RTE_LOG_REGISTER(\(.*\), '$pattern'\.\(.*\),/RTE_LOG_REGISTER_SUFFIX(\1, \2,/' $file;
done

$ git grep -l RTE_LOG_REGISTER lib/ |
while read file; do
pattern=${file##lib/};
pattern=lib.${pattern%%/*};
sed -i -e 's/RTE_LOG_REGISTER(\(.*\), '$pattern',/RTE_LOG_REGISTER_DEFAULT(\1,/' $file;
sed -i -e 's/RTE_LOG_REGISTER(\(.*\), '$pattern'\.\(.*\),/RTE_LOG_REGISTER_SUFFIX(\1, \2,/' $file;
done

Signed-off-by: David Marchand <[email protected]>
Signed-off-by: Thomas Monjalon <[email protected]>
Acked-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: skip VA requirement for bus without device

If after a bus scan, there are no devices using a particular bus, then
that bus should not be taken into account when deciding whether DPDK
shoul

raw/ioat: skip VA requirement for bus without device

If after a bus scan, there are no devices using a particular bus, then
that bus should not be taken into account when deciding whether DPDK
should be run in VA or PA addressing mode. This becomes an issue when
the DSA bus driver code is used on a system without an IOMMU. The PCI
bus correctly reports that it only works in PA mode, while the DSA bus -
also correctly - reports that it works only in VA mode. The difference
is that there will be no devices found in a scan for the DSA bus, since
the kernel driver can only present those to userspace in the presence of
an IOMMU.

While we could change DSA instance to always report that it does not
care about the addressing mode, this would imply that it could be used
with DPDK in PA mode which is not the case. Therefore, this patch
changes the driver to report DC (don't care) in the case where no
devices are present, and VA otherwise.

NOTE: this addressing mode use of VA-only applies only in the case of
using DSA through the idxd kernel driver. The use of DSA though vfio-pci
is unaffected and works as with other PCI devices.

Fixes: b7aaf417f936 ("raw/ioat: add bus driver for device scanning automatically")

Reported-by: Harry van Haaren <[email protected]>
Signed-off-by: Bruce Richardson <[email protected]>
Tested-by: Anatoly Burakov <[email protected]>
Tested-by: Harry van Haaren <[email protected]>
Tested-by: Conor Walsh <[email protected]>

show more ...

raw/ioat: fix directory handle leak

When reading the /dev directory as part of the bus scan for DSA devices,
the directory handle from opendir was not freed on function return,
leading to a resource

raw/ioat: fix directory handle leak

When reading the /dev directory as part of the bus scan for DSA devices,
the directory handle from opendir was not freed on function return,
leading to a resource leak.

Coverity issue: 370588
Fixes: b7aaf417f936 ("raw/ioat: add bus driver for device scanning automatically")

Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: report status of completed jobs

Add improved error handling to rte_ioat_completed_ops(). This patch adds
new parameters to the function to enable the user to track the completion
status of

raw/ioat: report status of completed jobs

Add improved error handling to rte_ioat_completed_ops(). This patch adds
new parameters to the function to enable the user to track the completion
status of each individual operation in a batch. With this addition, the
function can help the user to determine firstly, how many operations may
have failed or been skipped and then secondly, which specific operations
did not complete successfully.

Signed-off-by: Kevin Laatz <[email protected]>
Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: add API to query remaining ring space

Add a new API to query remaining descriptor ring capacity. This API is
useful, for example, when an application needs to enqueue a fragmented
packet a

raw/ioat: add API to query remaining ring space

Add a new API to query remaining descriptor ring capacity. This API is
useful, for example, when an application needs to enqueue a fragmented
packet and wants to ensure that all segments of the packet will be enqueued
together.

Signed-off-by: Kevin Laatz <[email protected]>
Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: rework SW ring layout

The ring management in the idxd part of the driver is more complex than
it needs to be, tracking individual batches in a ring and having null
descriptors as padding t

raw/ioat: rework SW ring layout

The ring management in the idxd part of the driver is more complex than
it needs to be, tracking individual batches in a ring and having null
descriptors as padding to avoid having single-operation batches. This can
be simplified by using a regular ring-based layout, with additional
overflow at the end to ensure that the one does not need to wrap within a
batch.

Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: move idxd functions to separate file

Split the rte_ioat_rawdev_fns.h file into two separate headers, so that
the data structures for the original ioat devices and the newer idxd
ones can b

raw/ioat: move idxd functions to separate file

Split the rte_ioat_rawdev_fns.h file into two separate headers, so that
the data structures for the original ioat devices and the newer idxd
ones can be kept separate from each other. This makes code management
and rework easier.

Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: add bus driver for device scanning automatically

Rather than using a vdev with args, DPDK can scan and initialize the
devices automatically using a bus-type driver. This bus does not need

raw/ioat: add bus driver for device scanning automatically

Rather than using a vdev with args, DPDK can scan and initialize the
devices automatically using a bus-type driver. This bus does not need to
worry about registering device drivers, rather it can initialize the
devices directly on probe.

The device instances (queues) to use are detected from /dev with the
additional info about them got from /sys.

Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: allow perform operations function to return error

Change the return type for the rte_ioat_perform_ops() function from void to
int to allow the possibility of returning an error code in fut

raw/ioat: allow perform operations function to return error

Change the return type for the rte_ioat_perform_ops() function from void to
int to allow the possibility of returning an error code in future, should
it be necessary.

Signed-off-by: Kevin Laatz <[email protected]>

show more ...

raw/ioat: make workqueue name configurable in script

Add a "--name-prefix" parameter to the quick configuration script for
DSA. This allows the queues configured on a DSA instance to be made
availab

raw/ioat: make workqueue name configurable in script

Add a "--name-prefix" parameter to the quick configuration script for
DSA. This allows the queues configured on a DSA instance to be made
available to only one DPDK process in a setup with multiple DPDK process
instances.

Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: fix script for configuring small number of queues

The dpdk_idxd_cfg.py script included with the driver for convenience did
not work properly where the number of queues to be configured was

raw/ioat: fix script for configuring small number of queues

The dpdk_idxd_cfg.py script included with the driver for convenience did
not work properly where the number of queues to be configured was
less than the number of groups or engines. This was because there would
be configured groups/engines not assigned to queues. Fix this by
limiting the engine and group counts to be no bigger than the number of
queues.

Fixes: 01863b9d2354 ("raw/ioat: include example configuration script")
Cc: [email protected]

Signed-off-by: Bruce Richardson <[email protected]>

show more ...

raw/ioat: expand descriptor struct to full 64 bytes

Although it's unused by the driver, add the interrupt handle field in
the descriptor to the descriptor structure for completeness, and
explicitly

raw/ioat: expand descriptor struct to full 64 bytes

Although it's unused by the driver, add the interrupt handle field in
the descriptor to the descriptor structure for completeness, and
explicitly add the reserved padding field on the end of the structure
too. This means that when a descriptor is defined on the stack, or
initialized by the compiler, the unused/reserved space will be zeroed
appropriately.

Signed-off-by: Bruce Richardson <[email protected]>

show more ...