This patch refactors the mempool cache to fix two bugs:
1. When a mempool is created with a per-lcore cache size of N objects, the
per-lcore caches were actually created with a size of 1.5 * N objects.
2. The mempool cache field names did not reflect their purpose;
the "flushthresh" field held the
> -Original Message-
> From: Stephen Hemminger
> Sent: Tuesday, October 1, 2024 11:44 PM
> To: Kusztal, ArkadiuszX
> Cc: dev@dpdk.org; ferruh.yi...@amd.com; Ji, Kai ; Dooley,
> Brian
> Subject: Re: [PATCH v2 1/3] net: add thread-safe crc api
>
> On Tue, 1 Oct 2024 19:11:48 +0100
> A
> -Original Message-
> From: shaib...@amazon.com
> Sent: Wednesday, October 2, 2024 11:22 AM
> To: tho...@monjalon.net
> Cc: dev@dpdk.org; Brandes, Shai
> Subject: [PATCH] maintainers: update ena maintainers list
>
> From: Shai Brandes
>
> Added a new ena team member.
>
> Signed-off-b
> -Original Message-
> From: David Marchand
> Sent: Wednesday, October 2, 2024 9:42 AM
> To: Kusztal, ArkadiuszX
> Cc: dev@dpdk.org; ferruh.yi...@amd.com; Ji, Kai ; Dooley,
> Brian
> Subject: Re: [PATCH v2 1/3] net: add thread-safe crc api
>
> On Tue, Oct 1, 2024 at 9:27 PM Arkadiusz
On Tue, Oct 1, 2024 at 6:39 PM Stephen Hemminger
wrote:
> diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
> index eec0400dad..595cadd5b8 100644
> --- a/lib/eal/include/rte_common.h
> +++ b/lib/eal/include/rte_common.h
> @@ -228,6 +228,40 @@ typedef uint16_t unaligned_uint1
On 9/27/24 02:31, Nicolas Chautru wrote:
v4: updated to remove device error logging part which may use
different mechanism in the future, possibly rte trace point.
v3: updated based on Maxime comments related to passing string
through function prototype. Thanks
v2: updated with comments from
On Wed, Oct 02, 2024 at 12:59:58AM +0300, Isaac Boukris wrote:
> On Tue, Oct 1, 2024 at 11:01 PM Bruce Richardson
> wrote:
> >
> > On Tue, Oct 01, 2024 at 03:22:51AM +0300, Isaac Boukris wrote:
> > > If the tsc_known_freq cpu flag is missing, it means the kernel doesn't
> > > trust it and calculat
From: Shai Brandes
Added a new ena team member.
Signed-off-by: Shai Brandes
---
MAINTAINERS | 1 +
1 file changed, 1 insertion(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index 812463fe9f..01b1c2cb7f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -663,6 +663,7 @@ M: Shai Brandes
M: Evgeny Sche
Add support for using DPDK source tree directory as well as DPDK
tarball with the pre-build directory that can user specify and
type of location, it can be stored in the local filesystem or SUT
node. Additionally, this can be set up with the config file or
cmd arguments/environment variables.
Sign
Since the DPDK may already be built, some more general name
is needed that includes both the DPDK location and the build
config (if we are going to build).
Signed-off-by: Tomáš Ďurovec
---
dts/conf.yaml | 2 +-
dts/framework/config/__init__.py | 26 ++---
When the DPDK build can be already pre-build, there is
not a need for defining multiple build targets. To make
it cleaner we decide to use one DPDK build wheater can
be pre-build or DTS will build it.
Signed-off-by: Tomáš Ďurovec
---
dts/conf.yaml | 14 +--
dts/fram
Before, the remote session did't allows to copy directories,
only files. This feature will be used in future commit.
Signed-off-by: Tomáš Ďurovec
---
dts/framework/testbed_model/os_session.py| 120 ++-
dts/framework/testbed_model/posix_session.py | 88 +-
dts/fra
The OSSession (and its subclasses) should accept PurePaths
for remote paths to translate from OS-unaware (PurePath)
to OS-aware (Path) only on the remote side. For local paths,
they should accept Paths, as Python is OS-aware locally.
Signed-off-by: Tomáš Ďurovec
---
.../remote_session/remote_ses
By adding support for external build, we extend the
argument documentation for supported options.
Signed-off-by: Tomáš Ďurovec
---
doc/guides/tools/dts.rst | 9 ++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/doc/guides/tools/dts.rst b/doc/guides/tools/dts.rst
index 65cce
In the previous commits we're adding the support for copying
the whole local DPDK tree directory and git-ref option was
meant to do the same thing.
Signed-off-by: Tomáš Ďurovec
---
doc/guides/tools/dts.rst | 9 ---
dts/framework/settings.py | 51 +++-
dts/framework/utils.py|
On Wed, Oct 02, 2024 at 07:56:36PM +0300, Isaac Boukris wrote:
> The CPU value is often not accurate, allow overriding it based
> on info from the host OS.
>
> On Linux X86, if the tsc_known_freq cpu flag is missing, it means
> the kernel doesn't trust it and calculates its own. We should do
> the
On Wed, Oct 02, 2024 at 07:56:34PM +0300, Isaac Boukris wrote:
> Changes in V4:
> - lower ronding on Windows too
> - on freebsd prefer system tsc_hz
>
> Isaac Boukris (2):
> timer: lower rounding of TSC estimation to 100KHz
> timer: allow platform to override cpu TSC frequency
>
> lib/eal/
If compiler detection of use-after-free is enabled then this drivers
debug messages will cause warnings. Change to move debug message
before the object is freed.
Bugzilla ID: 1551
Fixes: 55c1238246d5 ("net/sfc: add more debug messages to transfer flows")
Signed-off-by: Stephen Hemminger
Reviewed-
With proper annotation, GCC discovers that this driver is
calling rte_free() on an object that was not allocated
(it is part of array in another object).
In function ‘cpfl_flow_js_mr_layout’,
inlined from ‘cpfl_flow_js_mr_action’ at
../drivers/net/cpfl/cpfl_flow_parser.c:848:9,
inlined fr
If conversion of cBF to eBPF fails then an object allocated with
rte_malloc() would be passed to free().
[908/3201] Compiling C object lib/librte_bpf.a.p/bpf_bpf_convert.c.o
../lib/bpf/bpf_convert.c: In function ‘rte_bpf_convert’:
../lib/bpf/bpf_convert.c:559:17: warning: ‘free’ called on pointer
The driver would refer to the mempool object after it was freed.
Bugzilla ID: 1554
Fixes: 7ea187184a51 ("common/cnxk: support 1-N pool-aura per NIX LF")
Cc: rbhans...@marvell.com
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
---
drivers/net/cnxk/cnxk_ethdev_sec.c | 2 +-
1 file changed,
The driver cleanup code was freeing the filter object
then dereferencing it.
Bugzilla ID: 1550
Fixes: 6a4d050e2855 ("net/igb: flush all the filter")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
---
drivers/net/e1000/igb_ethdev.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
Calling rte_free twice on same object will corrupt the heap.
Warning is:
In function 'nfp_pre_tun_table_check_del',
inlined from 'nfp_flow_destroy' at
../drivers/net/nfp/flower/nfp_flower_flow.c:5143:9:
../drivers/net/nfp/flower/nfp_flower_flow.c:3830:9: error: pointer 'entry' used
after 'rte_fre
The TAILQ_FOREACH() macro would refer to info after it
had been freed. Fix by introducing TAILQ_FOREACH_SAFE here.
Fixes: 4a19f89104f8 ("raw/ifpga/base: support multiple cards")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
---
drivers/raw/ifpga/base/opae_intel_max10.c | 11 +--
It is possible that the info pointer (hp) could get freed twice.
Fix by nulling after free.
In function 'setup_la12xx_dev',
inlined from 'la12xx_bbdev_create' at
../drivers/baseband/la12xx/bbdev_la12xx.c:1029:8,
inlined from 'la12xx_bbdev_probe' at
../drivers/baseband/la12xx/bbdev_la12xx.c:1075:
The macro in this driver was redefining LIST_FOR_EACH_ENTRY_SAFE
as a simple LIST_FOR_EACH macro. But they are not the same
the _SAFE variant guarantees that there will not be use after free.
Fixes: fb4ac04e9bfa ("common/idpf: introduce common library")
Cc: junfeng@intel.com
Signed-off-by: Ste
Checking return value of rte_memzone_free() is pointless
and if it failed then it was because the pointer was null.
Fixes: 7b1374b1e6e7 ("common/qat: limit configuration to primary process")
Cc: arkadiuszx.kusz...@intel.com
Signed-off-by: Stephen Hemminger
---
drivers/common/qat/qat_device.c |
The raw ifpga driver redefines malloc to be opae_malloc
and free to be opae_free; which is a bad idea.
This leads to case where interrupt efd array is allocated
with calloc() and then passed to rte_free. The workaround
is to allocate the array with rte_calloc() instead.
Fixes: d61138d4f0e2 ("driv
On 2024-10-02 20:26, David Marchand wrote:
Hello Mattias,
On Mon, Sep 9, 2024 at 9:28 PM Mattias Rönnblom
wrote:
@@ -46,9 +49,21 @@ testsuite_teardown(void)
static int32_t dummy_cb(void *args)
{
RTE_SET_USED(args);
- service_tick++;
+
+ service_calls++;
+
+ switc
On 10/1/2024 5:25 PM, Stephen Hemminger wrote:
On Tue, 1 Oct 2024 14:25:46 +0200
David Marchand wrote:
On Tue, Oct 1, 2024 at 2:21 PM Burakov, Anatoly
wrote:
+
+/**
+ * Frees the memory space pointed to by the provided pointer.
+ *
+ * This pointer must have been returned by a previous call
On Wed, Oct 2, 2024 at 10:42 AM Kusztal, ArkadiuszX
wrote:
> > Exporting internals but not using them out of the library makes no sense.
> >
> So there must have been a misunderstanding on my part, the initial idea was
> to prevent the user from calling these functions.
net_crc.h is an internal
Eliminate unreachable code and have dummy service function perform a
short delay again, mimicking previous service core test behavior.
Fixes: a37e053b2364 ("service: extend service function call statistics")
Signed-off-by: Mattias Rönnblom
---
app/test/test_service_cores.c | 12 +---
1
On 2024-10-02 21:08, David Marchand wrote:
On Wed, Oct 2, 2024 at 8:57 PM Mattias Rönnblom wrote:
Coverity flagged this patch with issue #445158.
rte_delay_ms() is now unreachable.
I suppose this delay is not that important for the unit test and we
can remove it, but as I am not sure I'll let
On Sun, Sep 29, 2024 at 8:25 PM Ajit Khaparde
wrote:
> On Sat, Sep 28, 2024 at 12:18 PM Thomas Monjalon wrote:
> > > This email address is bouncing remove from MAINTAINERS.
> > >
> > > Signed-off-by: Stephen Hemminger
> Acked-by: Ajit Khaparde
Applied, thanks.
--
David Marchand
On Tue, Oct 1, 2024 at 9:27 PM Arkadiusz Kusztal
wrote:
>
> The current net CRC API is not thread-safe, this patch
> solves this by adding another, thread-safe API functions.
> This API is also safe to use across multiple processes,
> yet with limitations on max-simd-bitwidth, which will be checke
> -Original Message-
> From: David Marchand
> Sent: Wednesday, October 2, 2024 11:02 AM
> To: Kusztal, ArkadiuszX
> Cc: dev@dpdk.org; ferruh.yi...@amd.com; Ji, Kai ; Dooley,
> Brian
> Subject: Re: [PATCH v2 1/3] net: add thread-safe crc api
>
> On Wed, Oct 2, 2024 at 10:42 AM Kusztal,
Using the freed value for tracing is not a good idea.
Although it is harmless for tracing, it will cause analyzers to flag
this as unsafe.
Signed-off-by: Stephen Hemminger
Acked-by: Chengwen Feng
Acked-by: Anatoly Burakov
---
lib/eal/common/eal_common_memzone.c | 3 ++-
1 file changed, 2 inser
The device structure is allocated with rte_malloc() and
then incorrectly freed with free(). This will lead to
corrupt malloc pool.
Bugzilla ID: 1552
Fixes: c8e79da7c676 ("crypto/bcmfs: introduce BCMFS driver")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
Acked-by: Ajit Khaparde
---
dri
Recent versions of GCC have some additional function attributes that can
help with DPDK performance and stability.
The alloc_align attribute tells the compiler what the alignment
of the allocation will be, and the optimizer can use this to produce
better code (especially memcpy and structure copie
The pointer bdev is allocated with rte_zmalloc() and then
incorrectly freed with free() which will lead pool corruption.
Bugzilla ID: 1553
Fixes: c19beb3f38cd ("examples/vhost_blk: introduce vhost storage sample")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
Acked-by: Chengwen Feng
---
If conversion of cBF to eBPF fails then an object allocated with
rte_malloc() would be passed to free().
[908/3201] Compiling C object lib/librte_bpf.a.p/bpf_bpf_convert.c.o
../lib/bpf/bpf_convert.c: In function ‘rte_bpf_convert’:
../lib/bpf/bpf_convert.c:559:17: warning: ‘free’ called on pointer
The driver would refer to the mempool object after it was freed.
Bugzilla ID: 1554
Fixes: 7ea187184a51 ("common/cnxk: support 1-N pool-aura per NIX LF")
Cc: rbhans...@marvell.com
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
---
drivers/net/cnxk/cnxk_ethdev_sec.c | 2 +-
1 file changed,
The code to cleanup in case of error was passing incorrect
value to rte_free. The ports[] entry was allocated with
rte_malloc and that should be used instead of the offset
in that object.
Fixes: 97a05c1fe634 ("event/cnxk: add port config")
Cc: sthot...@marvell.com
Cc: sta...@dpdk.org
Signed-off-b
If compiler detection of use-after-free is enabled then this drivers
debug messages will cause warnings. Change to move debug message
before the object is freed.
Bugzilla ID: 1551
Fixes: 55c1238246d5 ("net/sfc: add more debug messages to transfer flows")
Signed-off-by: Stephen Hemminger
Reviewed-
The driver cleanup code was freeing the filter object
then dereferencing it.
Bugzilla ID: 1550
Fixes: 6a4d050e2855 ("net/igb: flush all the filter")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
---
drivers/net/e1000/igb_ethdev.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
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: kevin.la...@intel.com
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
Reviewed-by: Bruce Richardson
With proper annotation, GCC discovers that this driver is
calling rte_free() on an object that was not allocated
(it is part of array in another object).
In function ‘cpfl_flow_js_mr_layout’,
inlined from ‘cpfl_flow_js_mr_action’ at
../drivers/net/cpfl/cpfl_flow_parser.c:848:9,
inlined fr
Calling rte_free twice on same object will corrupt the heap.
Warning is:
In function 'nfp_pre_tun_table_check_del',
inlined from 'nfp_flow_destroy' at
../drivers/net/nfp/flower/nfp_flower_flow.c:5143:9:
../drivers/net/nfp/flower/nfp_flower_flow.c:3830:9: error: pointer 'entry' used
after 'rte_fre
The TAILQ_FOREACH() macro would refer to info after it
had been freed. Fix by introducing TAILQ_FOREACH_SAFE here.
Fixes: 4a19f89104f8 ("raw/ifpga/base: support multiple cards")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
---
drivers/raw/ifpga/base/opae_intel_max10.c | 11 +--
From: Fidaullah Noonari
This patch adds noiommu option to dpdk-devbind.
If the no IOMMU is detected, then if noiommu flag is set
the vfio-pci unsafe_noiommu_mode flag is set.
Signed-off-by: Fidaullah Noonari
Signed-off-by: Stephen Hemminger
---
v2 - rebase patch and fix some minor stuff
r
Hello Mattias,
On Mon, Sep 9, 2024 at 9:28 PM Mattias Rönnblom
wrote:
> @@ -46,9 +49,21 @@ testsuite_teardown(void)
> static int32_t dummy_cb(void *args)
> {
> RTE_SET_USED(args);
> - service_tick++;
> +
> + service_calls++;
> +
> + switch (rte_rand_max(3)) {
> +
On Tue, Sep 17, 2024 at 10:20 AM Maxime Coquelin
wrote:
> On 9/16/24 14:30, David Marchand wrote:
> > At the moment, if VFIO is not available at DPDK init, it won't be
> > available unless a subsequent rte_vfio_enable() is done.
> >
> > Yet, even if rte_vfio_enable() is called again in primary and
On Wed, Oct 2, 2024 at 9:45 AM Wathsala Wathawana Vithanage
wrote:
>
>
>
> > -Original Message-
> > Subject: [PATCH v6 17/17] eal: add function attributes for allocation
> > functions
> >
> > The allocation functions take a alignment argument that can be useful to
> > hint
> > the compil
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: kevin.la...@intel.com
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
Reviewed-by: Bruce Richardson
Recent versions of GCC have some additional function attributes that can
help with DPDK performance and stability.
The alloc_align attribute tells the compiler what the alignment
of the allocation will be, and the optimizer can use this to produce
better code (especially memcpy and structure copie
The pointer bdev is allocated with rte_zmalloc() and then
incorrectly freed with free() which will lead pool corruption.
Bugzilla ID: 1553
Fixes: c19beb3f38cd ("examples/vhost_blk: introduce vhost storage sample")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
Acked-by: Chengwen Feng
---
The code to cleanup in case of error was passing incorrect
value to rte_free. The ports[] entry was allocated with
rte_malloc and that should be used instead of the offset
in that object.
Fixes: 97a05c1fe634 ("event/cnxk: add port config")
Cc: sthot...@marvell.com
Cc: sta...@dpdk.org
Signed-off-b
Using the freed value for tracing is not a good idea.
Although it is harmless for tracing, it will cause analyzers to flag
this as unsafe.
Signed-off-by: Stephen Hemminger
Acked-by: Chengwen Feng
Acked-by: Anatoly Burakov
---
lib/eal/common/eal_common_memzone.c | 3 ++-
1 file changed, 2 inser
The device structure is allocated with rte_malloc() and
then incorrectly freed with free(). This will lead to
corrupt malloc pool.
Bugzilla ID: 1552
Fixes: c8e79da7c676 ("crypto/bcmfs: introduce BCMFS driver")
Cc: sta...@dpdk.org
Signed-off-by: Stephen Hemminger
Acked-by: Ajit Khaparde
---
dri
On Wed, Oct 2, 2024 at 8:57 PM Mattias Rönnblom wrote:
> > Coverity flagged this patch with issue #445158.
> > rte_delay_ms() is now unreachable.
> >
> > I suppose this delay is not that important for the unit test and we
> > can remove it, but as I am not sure I'll let you have a look and send
>
On Wed, Oct 2, 2024 at 8:11 PM Bruce Richardson
wrote:
>
> On Wed, Oct 02, 2024 at 07:56:36PM +0300, Isaac Boukris wrote:
> > The CPU value is often not accurate, allow overriding it based
> > on info from the host OS.
> >
> > On Linux X86, if the tsc_known_freq cpu flag is missing, it means
> > t
On Wed, Oct 2, 2024 at 6:27 PM Bruce Richardson
wrote:
>
> On Wed, Oct 02, 2024 at 05:57:08PM +0200, David Marchand wrote:
> > While invoking telemetry commands (which may happen at any time,
> > out of control of the application), an application thread may
> > concurrently add/remove ports.
> > T
David Marchand, Oct 02, 2024 at 21:06:
On Wed, Oct 2, 2024 at 6:27 PM Bruce Richardson
wrote:
On Wed, Oct 02, 2024 at 05:57:08PM +0200, David Marchand wrote:
> While invoking telemetry commands (which may happen at any time,
> out of control of the application), an application thread may
> con
On Wed, Oct 2, 2024 at 9:09 PM Robin Jarry wrote:
> >> An alternative to this macro-fu, is to just define a single ethdev
> >> telemetry function, and within that, take the lock and then dispatch to the
> >> appropriate subfunction based upon the actual command coming in. The
> >> dispatch may be
David Marchand, Oct 02, 2024 at 21:18:
On Wed, Oct 2, 2024 at 9:09 PM Robin Jarry wrote:
I was going to suggest adding a rte_spinlock_t* parameter to a new
telemetry register function that would need to be held while the
callback is invoked. Or if we want to keep doors open to other kinds of
lo
Following a discussion we had during the summit, here is one series to
fix a race between an application thread and the telemetry thread
handling requests on ethdev ports.
The problem may be generic to other device classes providing telemetry
callbacks, but for now, this series goes with a simple a
On Tue, Oct 1, 2024 at 9:45 PM Morten Brørup wrote:
>
> > From: Jerin Jacob [mailto:jerinjac...@gmail.com]
> > Sent: Tuesday, 1 October 2024 18.06
> >
> > On Tue, Oct 1, 2024 at 9:32 PM Morten Brørup
> > wrote:
> > >
> > > > From: Jerin Jacob [mailto:jerinjac...@gmail.com]
> > > > Sent: Tuesday,
On Wed, Oct 02, 2024 at 05:57:08PM +0200, David Marchand wrote:
> While invoking telemetry commands (which may happen at any time,
> out of control of the application), an application thread may
> concurrently add/remove ports.
> The telemetry callbacks may then access partially
> initialised/unini
> -Original Message-
> Subject: [PATCH v6 17/17] eal: add function attributes for allocation
> functions
>
> The allocation functions take a alignment argument that can be useful to hint
> the compiler optimizer. This is supported by GCC and Clang but only useful
> with
> GCC because
On Tue, 21 Jun 2022 14:29:23 +0200
Stanislaw Kardach wrote:
> Since the riscv64_linux_gcc was in fact a Ubuntu-specific cross-file,
> rename it.
>
> Signed-off-by: Stanislaw Kardach
> ---
Patch is out dated, and needs to be rebased.
Best to avoid OS variant specific cross files.
The CPU value is often not accurate, allow overriding it based
on info from the host OS.
On Linux X86, if the tsc_known_freq cpu flag is missing, it means
the kernel doesn't trust it and calculates its own. We should do
the same to avoid drift.
On freebsd we have access to the kernel tsc_hz value
In practice, the frequency is often not a nice round number, while
the estimation results are rather accurate, just a couple of KHz
away from the kernel's tsc_khz value, so it should suffice.
Rounding to 10MHz can cause a significant drift from real time,
up to a second per 10 minutes.
See also b
Changes in V4:
- lower ronding on Windows too
- on freebsd prefer system tsc_hz
Isaac Boukris (2):
timer: lower rounding of TSC estimation to 100KHz
timer: allow platform to override cpu TSC frequency
lib/eal/common/eal_common_timer.c | 3 +-
lib/eal/common/eal_private.h | 2 +-
lib
This patch refactors the mempool cache to fix two bugs:
1. When a mempool is created with a per-lcore cache size of N objects, the
per-lcore caches were actually created with a size of 1.5 * N objects.
2. The mempool cache field names did not reflect their purpose;
the "flushthresh" field held the
In order to avoid breaking Vhost-user live-migration, we want the
rte_vhost_driver_set_max_queue_num API to only be effective with
VDUSE.
Furthermore, this API is only really needed for VDUSE where the
device number of queues is defined by the backend. For Vhost-user,
this is defined by the fronte
This patch refactors the mempool cache to fix two bugs:
1. When a mempool is created with a per-lcore cache size of N objects, the
per-lcore caches were actually created with a size of 1.5 * N objects.
2. The mempool cache field names did not reflect their purpose;
the "flushthresh" field held the
On Tue, Oct 01, 2024 at 06:26:03PM +0200, Mattias Rönnblom wrote:
> Calling rte_exit() from a worker lcore thread causes a deadlock in
> rte_service_finalize().
>
> This patch makes rte_service_finalize() deadlock-free by avoiding the
> need to synchronize with service lcore threads, which in turn
The allocation functions take a alignment argument that
can be useful to hint the compiler optimizer. This is supported
by GCC and Clang but only useful with GCC because Clang gives
warning if alignment is 0.
Newer versions of GCC have a malloc attribute that can be used to find
mismatches between
The macro in this driver was redefining LIST_FOR_EACH_ENTRY_SAFE
as a simple LIST_FOR_EACH macro. But they are not the same
the _SAFE variant guarantees that there will not be use after free.
Fixes: fb4ac04e9bfa ("common/idpf: introduce common library")
Cc: junfeng@intel.com
Signed-off-by: Ste
The raw ifpga driver redefines malloc to be opae_malloc
and free to be opae_free; which is a bad idea.
This leads to case where interrupt efd array is allocated
with calloc() and then passed to rte_free. The workaround
is to allocate the array with rte_calloc() instead.
Fixes: d61138d4f0e2 ("driv
It is possible that the info pointer (hp) could get freed twice.
Fix by nulling after free.
In function 'setup_la12xx_dev',
inlined from 'la12xx_bbdev_create' at
../drivers/baseband/la12xx/bbdev_la12xx.c:1029:8,
inlined from 'la12xx_bbdev_probe' at
../drivers/baseband/la12xx/bbdev_la12xx.c:1075:
Checking return value of rte_memzone_free() is pointless
and if it failed then it was because the pointer was null.
Fixes: 7b1374b1e6e7 ("common/qat: limit configuration to primary process")
Cc: arkadiuszx.kusz...@intel.com
Signed-off-by: Stephen Hemminger
---
drivers/common/qat/qat_device.c |
A next commit will protect all telemetry commands.
Prefer exposing all commands regardless of OS, and return an error
when invoked on Windows.
Signed-off-by: David Marchand
---
lib/ethdev/rte_ethdev_telemetry.c | 12 +---
1 file changed, 9 insertions(+), 3 deletions(-)
diff --git a/lib
While invoking telemetry commands (which may happen at any time,
out of control of the application), an application thread may
concurrently add/remove ports.
The telemetry callbacks may then access partially
initialised/uninitialised ethdev data.
Reuse the ethdev lock that protects port allocation
Hi Gowri,
Can you fix the CI issues reported.
http://mails.dpdk.org/archives/test-report/2024-September/798079.html
openssl asym autotest is also failing.
http://mails.dpdk.org/archives/test-report/2024-September/798209.html
> Subject: [PATCH v3 6/6] app/crypto-perf: support EDDSA
>
> Added
On Wed, Oct 2, 2024 at 9:40 PM Mattias Rönnblom
wrote:
>
> Eliminate unreachable code and have dummy service function perform a
> short delay again, mimicking previous service core test behavior.
>
Coverity issue: 445158
> Fixes: a37e053b2364 ("service: extend service function call statistics")
>
On Wed, Oct 2, 2024 at 4:42 PM Maxime Coquelin
wrote:
>
> In order to avoid breaking Vhost-user live-migration, we want the
> rte_vhost_driver_set_max_queue_num API to only be effective with
> VDUSE.
>
> Furthermore, this API is only really needed for VDUSE where the
> device number of queues is d
From: Mattias Rönnblom
Calling rte_exit() from a worker lcore thread causes a deadlock in
rte_service_finalize().
This patch makes rte_service_finalize() deadlock-free by avoiding the
need to synchronize with service lcore threads, which in turn is
achieved by moving service and per-lcore state
> > From: Pavan Nikhilesh
> >
> > Event pre-scheduling improves scheduling performance by assigning events
> > to event ports in advance when dequeues are issued.
> > This series introduces various types and levels of pre-scheduling to the
> > eventdev library.
>
> Any numbers on how much this im
> The code to cleanup in case of error was passing incorrect
> value to rte_free. The ports[] entry was allocated with
> rte_malloc and that should be used instead of the offset
> in that object.
>
> Fixes: 97a05c1fe634 ("event/cnxk: add port config")
> Cc: sthot...@marvell.com
> Cc: sta...@dpdk.o
90 matches
Mail list logo
