On 11/21/2014 04:38 PM, Alexandru Badicioiu wrote:
On 21 November 2014 16:25, Taras Kondratiuk taras.kondrat...@linaro.org
7. Two queues one of them ORDERED, several threads. Verify that buffers
scheduled from ORDERED queue are enqueue into the second queue in
correct order
On 11/21/2014 06:16 PM, Ciprian Barbu wrote:
On Fri, Nov 21, 2014 at 5:54 PM, Bala Manoharan
bala.manoha...@linaro.org wrote:
Few points,
* Inorder to check ordered state of buffers from second queue they should be
dequeued by a single thread
as scheduler will despatch the buffers from
On 11/19/2014 06:12 PM, Maxim Uvarov wrote:
Fix macro logic. Application has to be terminated in any case.
Signed-off-by: Maxim Uvarov maxim.uva...@linaro.org
---
platform/linux-generic/include/api/odp_debug.h | 11 +++
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git
On 11/20/2014 10:15 AM, Maxim Uvarov wrote:
On 11/20/2014 01:49 AM, Anders Roxell wrote:
On 2014-11-19 16:23, Mike Holmes wrote:
On 19 November 2014 16:13, Maxim Uvarov maxim.uva...@linaro.org wrote:
On 11/19/2014 11:21 PM, Mike Holmes wrote:
On 19 November 2014 10:36, Yan Songming
On 11/20/2014 03:18 PM, Ciprian Barbu wrote:
On Wed, Nov 19, 2014 at 3:39 PM, Taras Kondratiuk
taras.kondrat...@linaro.org wrote:
diff --git a/test/cunit/buffer/Makefile.am b/test/cunit/buffer/Makefile.am
new file mode 100644
index 000..071a701
--- /dev/null
+++ b/test/cunit/buffer
On 11/20/2014 04:22 PM, Maxim Uvarov wrote:
Applied with minimal fixes:
1.
sed -i s/ODP_ERR\(/fprintf\(stderr,\ /g
0001-cunit-add-tests-for-crypto-APIs.patch
2. remove setting:
-Wno-error
3. in main()
no need to set up pool_base to NULL.
4. added binary to .gitignore
Current master is
ODP application may want to override default ODP logging behaviour
and use custom logging function. Add a weak odp_override_log() function
for this purpose instead of default fprintf().
Signed-off-by: Taras Kondratiuk taras.kondrat...@linaro.org
---
platform/linux-generic/include/api/odp_debug.h
-November/004987.html
*** BLURB HERE ***
Taras Kondratiuk (2):
platform: debug: replace fprintf() with odp_override_log()
platform: debug: Simplify ODP_LOG() macro
platform/linux-generic/include/api/odp_debug.h | 45 ++--
.../linux-generic/include/odp_debug_internal.h
On 11/18/2014 09:23 PM, Bill Fischofer wrote:
ODP_SAY()?
A few more:
LOG_APP
LOG_USR
Normally odp_*_print() functions will be used for debugging.
Maybe they can be printed with LOG_DBG level?
___
lng-odp mailing list
lng-odp@lists.linaro.org
On 11/18/2014 11:45 AM, Maxim Uvarov wrote:
On 11/18/2014 10:45 AM, Alexandru Badicioiu wrote:
Wouldn't be better to do this:
odp_crypto_CFLAGS = $(AM_CFLAGS) -Wno-error -I./crypto
I suppose at some point CUnit will move to 2.1-3.
Thanks,
Alex
Yes, that is good. Please add also comment why
On 11/14/2014 04:18 AM, Bill Fischofer wrote:
diff --git a/platform/linux-generic/include/api/odp_impl_types.h
b/platform/linux-generic/include/api/odp_impl_types.h
new file mode 100644
index 000..f4deedb
--- /dev/null
+++ b/platform/linux-generic/include/api/odp_impl_types.h
@@ -0,0 +1,72
On 11/14/2014 08:29 PM, Mike Holmes wrote:
On 14 November 2014 13:21, Anders Roxell anders.rox...@linaro.org
mailto:anders.rox...@linaro.org wrote:
On 2014-11-14 15:16, Alexandru Badicioiu wrote:
I was told that the CUnit version is 2.1-3, this is why I have used it.
CI is on
On 11/17/2014 12:00 PM, Maxim Uvarov wrote:
On 11/17/2014 12:01 PM, Taras Kondratiuk wrote:
On 11/15/2014 12:01 PM, Maxim Uvarov wrote:
Merged!
Guys, what about commit message? How one should understand what was the
reason for this patch by looking into git log?
yes, agree maybe we need
On 11/14/2014 11:49 AM, Alexandru Badicioiu wrote:
Hi,
for build errors - which compiler/toolchain do you use? I did not modify
in any way compilation flags and I have no problem with building with a
PPC toolchain and also with the native gcc on my development machine? I
use CUnit-2.1-3.
On 11/13/2014 01:21 AM, Bill Fischofer wrote:
*** Buffer/Packet API Patch v3 ***
v3 corrects some errors in the packaging of v2 and also changes
odp_buffer_alloc() and odp_packet_alloc() to by default allocate
buffers/packets of the default size set when the buffer pool was
created.
Bill
On 11/11/2014 08:14 PM, Mike Holmes wrote:
On 11 November 2014 11:33, Taras Kondratiuk taras.kondrat...@linaro.org
mailto:taras.kondrat...@linaro.org wrote:
On 11/11/2014 06:22 PM, Anders Roxell wrote:
On 11 November 2014 17:12, Taras Kondratiuk
taras.kondrat
On 11/12/2014 02:59 AM, Bill Fischofer wrote:
[snip]
+static inline odp_buffer_hdr_t *validate_buf(odp_buffer_t buf)
+{
+ odp_buffer_bits_t handle;
+ odp_buffer_hdr_t *buf_hdr;
+ handle.u32 = buf;
+
+ /* For buffer handles, segment index must be 0 */
+ if
On 11/10/2014 08:07 PM, Bill Fischofer wrote:
Thanks. See inline comments:
On Mon, Nov 10, 2014 at 11:37 AM, Taras Kondratiuk
taras.kondrat...@linaro.org mailto:taras.kondrat...@linaro.org wrote:
On 11/10/2014 07:10 PM, Bill Fischofer wrote:
Signed-off-by: Bill Fischofer
On 11/11/2014 06:03 AM, Bill Fischofer wrote:
*** BLURB HERE ***
This patch series upgrades the main ODP repository to the full v1.0 level
for the Buffer/Packet APIs. After discussion with Mike and Anders, it was
agreed to break this into multiple parts to facilitate review. These parts,
);
+ pool-s.bufcount = 0;
+
/* Form block freelist for pool */
uint8_t *blk = pool-s.pool_base_addr + pool-s.pool_size -
pool-s.seg_size;
Bill
On Tue, Nov 11, 2014 at 2:26 AM, Taras Kondratiuk
taras.kondrat
On 11/11/2014 03:47 PM, Bill Fischofer wrote:
v3 patch posted to correct 2nd problem (ret_buf returning to
blk_freelist instead of buf_freelist).
The pool bufcount is decremented as part of ret_buf processing. It
reflects the number of allocated buffers and will return to 0 naturally
when the
On 11/11/2014 03:44 PM, Bill Fischofer wrote:
v3 incorporates bugfix: ret_buf must return to buf_freelist, not blk_freelist
A few more issues:
- odp_buffer_pool_next() implementation is missing.
- ODP_BUFFER_POOL_NULL is mentioned in documentation, but not defined
in the header file.
Sorry
On 11/11/2014 04:18 PM, Bill Fischofer wrote:
Actually, it's easier to deal with these one by one. I have no problem
with that.
odp_buffer_pool_next() is in odp_buffer_pool.c starting at line 294. Do
you not see it?
Oh. It is there since v7, but I'm still on v6. Too many versions per
day :)
On 11/11/2014 04:13 PM, Bill Fischofer wrote:
v4 fixes bufcount initialization during pool creation
Signed-off-by: Bill Fischofer bill.fischo...@linaro.org
diff --git a/platform/linux-generic/odp_buffer_pool.c
b/platform/linux-generic/odp_buffer_pool.c
index a48d7d6..3e37d60 100644
---
On 11/11/2014 04:50 PM, Anders Roxell wrote:
On 2014-11-11 08:24, Bill Fischofer wrote:
So what you're saying is that patches aren't always the practical solution?
:)
I like patches better... however, different taste =)
I spoke with Anders earlier and will get setup on people.linaro.org
On 11/11/2014 04:13 PM, Bill Fischofer wrote:
v4 fixes bufcount initialization during pool creation
Signed-off-by: Bill Fischofer bill.fischo...@linaro.org
---
platform/linux-generic/odp_buffer.c | 263 ++-
platform/linux-generic/odp_buffer_pool.c | 664 +++-
On 11/11/2014 06:22 PM, Anders Roxell wrote:
On 11 November 2014 17:12, Taras Kondratiuk taras.kondrat...@linaro.org wrote:
On 11/11/2014 04:50 PM, Anders Roxell wrote:
On 2014-11-11 08:24, Bill Fischofer wrote:
So what you're saying is that patches aren't always the practical
solution
On 11/10/2014 10:22 AM, Taras Kondratiuk wrote:
On 11/08/2014 01:17 AM, Bill Fischofer wrote:
Signed-off-by: Bill Fischofer bill.fischo...@linaro.org
---
configure.ac| 11 +++
platform/linux-generic/include/odp_debug_internal.h | 11
On 11/10/2014 02:29 PM, Bill Fischofer wrote:
That's interesting. The default makefiles for GCC on linux have all the
pendantic warnings turned on and I'm not seeing those errors. I assume
you're getting them with the TI compiler? Are those the only routines
that have this problem?
Strange
On 11/10/2014 02:26 PM, Bill Fischofer wrote:
The main use case for anonymous bit fields is when mapping HW registers
that have HW-reserved bits that should not be touched by SW. Another is
for architected packet formats with explicitly reserved fields that
again SW should not try to repurpose.
Do you plan to add tests for async out of place?
Please check comments below.
On 11/10/2014 05:31 PM, Alexandru Badicioiu wrote:
Ping.
On 5 November 2014 11:00, alexandru.badici...@linaro.org
mailto:alexandru.badici...@linaro.org wrote:
From: Alexandru Badicioiu
On 11/10/2014 02:57 PM, Bill Fischofer wrote:
I'm using gcc 4.8.1 on Ubuntu for my development. I'll test with your
changes and add them to the next revision since I assume they'll be
benign for my compiler.
C casting has always been a painful area.
I assume on x86 unaligned access is not an
On 11/10/2014 07:10 PM, Bill Fischofer wrote:
Signed-off-by: Bill Fischofer bill.fischo...@linaro.org
---
Versions of this patch
==
v1 - Original (missing files)
v2 - Added missing odp_buffer_inlines.h
v3 - Added missing odph_tcp.h
v4 - (mispublished as v3): Allow NULL arguments to
Please add 'cunit:' or 'test: cunit:' prefix to the subject.
On 11/06/2014 05:08 PM, Yan Songming wrote:
From: yan.songming yan.songm...@linaro.org
Add the cunit test for none syne queue. Test the base queue function
Signed-off-by: yan.songming yan.songm...@linaro.org
---
Fix spelling problem
On 11/04/2014 07:35 PM, Bill Fischofer wrote:
If the ODP prototypes simply all said 'static inline' then would that
permit the implementation to choose later which were actually static inline?
I.e., if the common odp_xxx.h says
static inline void *odp_foo(...);
Then can the implementation can
On 11/04/2014 07:51 PM, Ciprian Barbu wrote:
On Tue, Nov 4, 2014 at 7:27 PM, Taras Kondratiuk
taras.kondrat...@linaro.org wrote:
Current ODP headers contains following parts in each file:
1. Platform-specific
1.a. Handle typedefs: e.g. odp_queue_t
1.b. Macros, enums: e.g
On 11/04/2014 10:48 AM, Shmulik Ladkani wrote:
Hi,
I don't have a big picture of what you are trying to achieve, but ODP
is not a framework for thread-local initialization of 3rd party
libraries.
I completely agree.
However my usecase is an exception, as I was trying to create a C++
binding
On 11/03/2014 11:50 AM, Shmulik Ladkani wrote:
Hi,
I'd like to suggest the following addition to odp_init.h:
diff --git a/platform/linux-generic/include/api/odp_init.h
b/platform/linux-generic/include/api/odp_init.h
index 8e018cbdba..9b75609fc7 100644
---
On 11/03/2014 12:22 PM, Shmulik Ladkani wrote:
Hi Taras,
On Mon, 03 Nov 2014 12:00:33 +0200 Taras Kondratiuk
taras.kondrat...@linaro.org wrote:
Could you please elaborate it? What is a usecase?
Think of a library extending the ODP library.
Specifically, I was playing around creating a C
.
On 3 July 2014 13:14, Mike Holmes mike.hol...@linaro.org
mailto:mike.hol...@linaro.org wrote:
Let me take a look, I will ping you offline to make sure I
understand and replicate the issue correctly
On 3 July 2014 06:26, Taras Kondratiuk taras.kondrat...@linaro.org
On 11/03/2014 03:02 PM, Shmulik Ladkani wrote:
On Mon, 03 Nov 2014 14:09:02 +0200 Taras Kondratiuk
taras.kondrat...@linaro.org wrote:
Won't it be simpler to add a new local initialization API in your
library instead of piggy-backing on ODP local init call?
Something like:
stl_odp_local_init
Since commit e968af6 Remove pktio type awareness ipsec application doesn't
accept 'f' and 't' options. getopt_long() function is configured to stop
processing at first non-option, so 'm' option at the end is not processed.
Remove these redundant options.
Signed-off-by: Taras Kondratiuk
On 10/15/2014 09:14 PM, Mike Holmes wrote:
Unit testing replicates a situation similar to an NFV case where the workload
is varied without rebooting the HW.
In this use case there have to be hooks for the application to inform the
implementation that HW needs to be released.
odp_term_global
On 10/13/2014 08:29 PM, Rosenboim, Leonid wrote:
Taras,
I am not worried too much about implementation at this stage, as I am sure we
can figure it out,
but I am more concerned about semantics.
If the request to attach a queue to a certain packet output port is a separate
request from queue
On 10/10/2014 09:28 PM, Rosenboim, Leonid wrote:
Stuart,
The declaration of a queue as PKTIN, PKTOUT during creation time is NOT an
implementation detail, and was in my opinion deliberate and changing this will
have serious consequences for our ability to support this API on Octeon SoCs.
On 10/09/2014 03:08 PM, Savolainen, Petri (NSN - FI/Espoo) wrote:
Google docs is not optimal for having long discussions or recording the
decisions.
- It's hard to follow the discussion, when it happens all over the document
(potentially same subject on multiple places)
- Changes in a doc or
On 10/06/2014 03:09 PM, Maxim Uvarov wrote:
On 10/06/2014 02:19 PM, Venkatesh Vivekanandan wrote:
On 6 October 2014 15:44, Ciprian Barbu ciprian.ba...@linaro.org
mailto:ciprian.ba...@linaro.org wrote:
Hi,
Do we have a definite convention for sending patches against platform
On 09/29/2014 09:24 PM, Mike Holmes wrote:
Signed-off-by: Mike Holmes mike.hol...@linaro.org
---
release.dox | 54 ++
1 file changed, 54 insertions(+)
create mode 100644 release.dox
diff --git a/release.dox b/release.dox
new file mode
branch
when branch maintainer have updated it to the same API level as linux-generic in
release branch.
This approach allows to unblock and speed up linux-generic development.
As before patches to any branch should go through ML review process.
--
Taras Kondratiuk
Lähettäjä: ext Magnus Karlsson
Päivämäärä:19.09.2014 10.25 (GMT-08:00)
Saaja: Taras Kondratiuk
Kopio: lng-odp@lists.linaro.org
Aihe: Re: [lng-odp] Relaxing merge process
Taras,
Sounds good as long as we substitute the word branches with
repositories. Branches do not make much sense IMHO
Yes something like this.
On 09/19/2014 12:07 PM, Mike Holmes wrote:
As repos, something like this ?
On 19 September 2014 14:55, Taras Kondratiuk
taras.kondrat...@linaro.org mailto:taras.kondrat...@linaro.org wrote:
Both approaches have pros and cons.
I prefer branches, so all
On 09/19/2014 02:10 PM, Mike Holmes wrote:
My 2 cents
Repos:
* A private SoCs implementation such as Axxia, Cavium, Freescale are
already separate repos, why should public ones be different ?
* When you pull you only get the implementation you want, you don't
get others for
On 09/19/2014 02:28 PM, Jacob, Jerin wrote:
How about git submodules, A master git odp repo and platforms specific
sub repos within it.Best of both worlds.
http://git-scm.com/book/en/Git-Tools-Submodules
I had very unpleasant experience with git submodules.
Better to go with a repo tool and
On 09/13/2014 02:27 PM, Maxim Uvarov wrote:
On 09/13/2014 12:50 PM, Taras Kondratiuk wrote:
On 09/13/2014 12:01 AM, Maxim Uvarov wrote:
Patch was merged, thanks Robbie and everybody!
Great!
Could you please consider merging Keystone2 Crypto series to fix broken
IPsec build for Keystone?
I
On 09/10/2014 12:45 PM, Maxim Uvarov wrote:
Patch set breaks odp compilation for ks2 and dpdk. Taras, Venky are you
ready to send update?
Yes. KS2 crypto patch series is in ML already. I'll post an updated
version soon. DPDK can just reuse linux-generic's implementation of
crypto.
On 09/10/2014 05:35 PM, Ola Liljedahl wrote:
I think we should separate between ODP implementations and the ODP API.
Users (applications) are primarily interested in the API. Any API
changes can be separated into backwards and not backwards compatible
changes. I think a major.minor designation
to 16k.
Taras Kondratiuk (3):
Add platform-specific Makefile.inc for applications
linux-keystone2: Switch to McSDK libraries
linux-keystone2: Add intial crypto implementation
example/Makefile.inc |1 +
platform/linux-keystone2/Makefile.am
Platform-specific Makefile.inc is useful when application need to be
built with platform-specific configuration: SDK search path,
configurations, etc.
Signed-off-by: Taras Kondratiuk taras.kondrat...@linaro.org
Signed-off-by: Taras Kondratiuk ta...@ti.com
---
example/Makefile.inc |1 +
test
The patch implements asynchronous out-of-place ODP crypto operations:
results are stored in a new buffer.
Implementation based on NWAL library.
Signed-off-by: Taras Kondratiuk taras.kondrat...@linaro.org
Signed-off-by: Taras Kondratiuk ta...@ti.com
---
platform/linux-keystone2/Makefile.am
Hi
API and header names will evolve and change continuously. Some changes
may be not backward compatible. Application will need some mechanism to
support at least a few recent revisions of ODP API. For this we can use
simple versioning.
#include odp_version.h
#if ODP_VERSION = 100
/* Call ODP
Hi
We have agreed that 'free' API is needed for all ODP resources like
queues, pools, crypto sessions, etc. Application have to free all
resources before exit.
I think additionally we need corresponding 'close' API for odp_init_*()
functions to free resources allocated by implementation during
On 09/04/2014 05:22 PM, Ola Liljedahl wrote:
On 4 September 2014 15:45, Taras Kondratiuk taras.kondrat...@linaro.org
mailto:taras.kondrat...@linaro.org wrote:
On 09/04/2014 02:55 PM, Ola Liljedahl wrote:
I am fine with the application having the responsibility
On 09/02/2014 08:59 PM, Ciprian Barbu wrote:
Hello Taras,
I have updated my openvswitch_odp to the latest ODP version and I
would like some help testing it on keystone.
I haven't compiled ODP for KS2 yet, so maybe you can help me with that.
Let's make a HO. Are you available today?
The
On 09/03/2014 08:18 PM, Mike Holmes wrote:
Do you mean https://cards.linaro.org/browse/LNG-483 that was scheduled
behind OVS work ?
Exactly this one.
___
lng-odp mailing list
lng-odp@lists.linaro.org
http://lists.linaro.org/mailman/listinfo/lng-odp
Platform-specific Makefile.inc is useful when application need to be
built with platform-specific configuration: SDK search path,
configurations, etc.
Signed-off-by: Taras Kondratiuk taras.kondrat...@linaro.org
Signed-off-by: Taras Kondratiuk ta...@ti.com
---
example/Makefile.inc |1 +
test
license boilerplates, updated README
- Increased max number of hardware descriptors to 16k.
Taras Kondratiuk (3):
Add platform-specific Makefile.inc for applications
linux-keystone2: Switch to McSDK libraries
linux-keystone2: Add intial crypto implementation
example/Makefile.inc
The patch implements asynchronous out-of-place ODP crypto operations:
results are stored in a new buffer.
Implementation based on NWAL library.
Signed-off-by: Taras Kondratiuk taras.kondrat...@linaro.org
Signed-off-by: Taras Kondratiuk ta...@ti.com
---
platform/linux-keystone2/Makefile.am
On 08/26/2014 07:03 PM, Mike Holmes wrote:
Signed-off-by: Mike Holmes mike.hol...@linaro.org
---
v2:
Factor out odp_platform_init so that odp_global_init is only defined
for linux_generic and resued.
Hi Mike
I doubt it is worth to reuse odp_init_global() from linux-generic.
Initialization
On 08/27/2014 03:16 PM, Taras Kondratiuk wrote:
On 08/26/2014 07:03 PM, Mike Holmes wrote:
Signed-off-by: Mike Holmes mike.hol...@linaro.org
---
v2:
Factor out odp_platform_init so that odp_global_init is only defined
for linux_generic and resued.
Hi Mike
I doubt it is worth to reuse
On 08/25/2014 05:37 PM, Maxim Uvarov wrote:
I pushed here some raw code:
https://git.linaro.org/people/maxim.uvarov/odp.git/shortlog/refs/heads/odp_ipc2
Example here app here:
https://git.linaro.org/people/maxim.uvarov/odp.git/blob/refs/heads/odp_ipc2:/example/fork/odp_fork.c
On 08/26/2014 02:01 PM, Gilad Ben Yossef wrote:
Hi Bill and everyone,
As this is my first post (a few patches not withstanding), I'd like to
start by saying thank you for the work you and everyone are putting into
ODP.
Now for the topic at hand - I know I am a bit of a Johnny-come-lately to
On 08/20/2014 02:59 PM, Robbie King (robking) wrote:
Hey Taras, a while back we tabled the completion event discussion until we
linux-generic crypto implementation got into the repository. If I remember
correctly, you had a proposal for dealing with the completion event issue
that you
On 08/21/2014 12:43 AM, Mike Holmes wrote:
On 20 August 2014 17:31, Taras Kondratiuk taras.kondrat...@linaro.org
mailto:taras.kondrat...@linaro.org wrote:
On 08/18/2014 04:03 PM, Mike Holmes wrote:
Signed-off-by: Mike Holmes mike.hol...@linaro.org
mailto:mike.hol...@linaro.org
This series adds initial accelerated Crypto API implementation. To do this
Keystone2 ODP implementation migrates to TI Multicore SDK instead of OpenEM
helper libraries.
Taras Kondratiuk (3):
Add platform specific Makefile.inc
linux-keystone2: Switch to McSDK libraries
linux-keystone2: Add
301 - 374 of 374 matches
Mail list logo