From: Hemant Agrawal <hemant.agrawal@nxp.com>
To: Olivier MATZ <olivier.matz@6wind.com>
Cc: <dev@dpdk.org>, <thomas.monjalon@6wind.com>,
<shreyansh.jain@nxp.com>, <ferruh.yigit@intel.com>
Subject: Re: [dpdk-dev] [PATCH v2] mempool/dpaa2: add DPAA2 hardware offloaded mempool
Date: Tue, 11 Apr 2017 11:28:26 +0530 [thread overview]
Message-ID: <9977cf0c-e1cf-b235-e199-c5242daa75c2@nxp.com> (raw)
In-Reply-To: <20170410215847.704092d2@neon>
Hi Olivier,
On 4/11/2017 1:28 AM, Olivier MATZ wrote:
> Hi Hemant,
>
> On Sun, 9 Apr 2017 13:29:46 +0530
> Hemant Agrawal <hemant.agrawal@nxp.com> wrote:
>
>> DPAA2 Hardware Mempool handlers allow enqueue/dequeue from NXP's
>> QBMAN hardware block.
>> CONFIG_RTE_MBUF_DEFAULT_MEMPOOL_OPS is set to 'dpaa2', if the pool
>> is enabled.
>>
>> This memory pool currently supports packet mbuf type blocks only.
>>
>> Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
>
> [...]
>
>
>> --- a/drivers/bus/Makefile
>> +++ b/drivers/bus/Makefile
>> @@ -33,6 +33,10 @@ include $(RTE_SDK)/mk/rte.vars.mk
>>
>> core-libs := librte_eal librte_mbuf librte_mempool librte_ring librte_ether
>>
>> +ifeq ($(CONFIG_RTE_LIBRTE_DPAA2_MEMPOOL),y)
>> +CONFIG_RTE_LIBRTE_FSLMC_BUS = $(CONFIG_RTE_LIBRTE_DPAA2_MEMPOOL)
>> +endif
>> +
>> DIRS-$(CONFIG_RTE_LIBRTE_FSLMC_BUS) += fslmc
>> DEPDIRS-fslmc = ${core-libs}
>>
>
> What's the purpose of this?
> Not sure we are allowed to modify the configs in the Makefiles.
DPAA2_MEMPOOL will not work without the DPAA2 mempool hw instance
detected on FSLMC_BUS.
So, it is required that if you are enabling DPAA2_MEMPOOL, FSLMC_BUS is
to be enabled.
Currently the config structure do not provide such dependency definitions.
This was done based on the suggestions on the initial patches from
Ferruh and Jerin.
>> + ret = dpbp_get_attributes(&avail_dpbp->dpbp, CMD_PRI_LOW,
>> + avail_dpbp->token, &dpbp_attr);
>> + if (ret != 0) {
>> + PMD_INIT_LOG(ERR, "Resource read failure with"
>> + " err code: %d\n", ret);
>> + p_ret = ret;
>> + ret = dpbp_disable(&avail_dpbp->dpbp, CMD_PRI_LOW,
>> + avail_dpbp->token);
>> + return p_ret;
>> + }
>> +
>> + /* Allocate the bp_list which will be added into global_bp_list */
>> + bp_list = (struct dpaa2_bp_list *)malloc(sizeof(struct dpaa2_bp_list));
>> + if (!bp_list) {
>> + PMD_INIT_LOG(ERR, "No heap memory available");
>> + return -ENOMEM;
>> + }
>> +
>
> I think the cast is not needed.
> Are you sure you want to use malloc() and not rte_malloc()? It would be in
> hugepages.
>
Yes! you are right, cast is not required and rte_malloc will be better
than malloc.
> [...]
>
>>
>>
>
>
> I still have some concerns about the fact that the mempool handler assumes that
> the objects are necessarily mbufs. I guess for this reason it does not pass
> mempool autotests?
>
Based on some of your previous suggestion, I have some thoughts to
address it, but it will be longer term.
1. add some kind of capability APIs in mempool
2. Or, indicate to mempool to differentiate between mbuf pool vs
non-mbuf pool.
I will initiate discussions on these topics.
> We should probably move forward and let it go in 17.05, but this is something
> that should be enhanced in my opinion.
>
Thanks for your valued suggestions and reviews. Yes! Once the basic
stuff in in the 17.05, we will start working on making it more generic.
> Regards,
> Olivier
>
>
next prev parent reply other threads:[~2017-04-11 5:58 UTC|newest]
Thread overview: 54+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-03-17 12:47 [dpdk-dev] [PATCH v1] NXP DPAA2 External Mempool Driver Hemant Agrawal
2017-03-17 12:47 ` [dpdk-dev] [PATCH v1] mempool/dpaa2: add DPAA2 hardware offloaded mempool Hemant Agrawal
2017-03-22 6:09 ` Jianbo Liu
2017-03-23 16:57 ` Hemant Agrawal
2017-03-24 14:57 ` Ferruh Yigit
2017-03-24 15:59 ` Ferruh Yigit
2017-03-24 16:31 ` Olivier Matz
2017-03-24 16:38 ` Ferruh Yigit
2017-03-24 16:42 ` Olivier Matz
2017-03-27 16:30 ` Olivier Matz
2017-03-28 9:45 ` Hemant Agrawal
2017-03-30 11:29 ` Ferruh Yigit
2017-03-30 12:50 ` Hemant Agrawal
2017-03-30 13:52 ` Thomas Monjalon
2017-04-03 14:50 ` Ferruh Yigit
2017-04-09 7:59 ` [dpdk-dev] [PATCH v2] NXP DPAA2 External Mempool Driver Hemant Agrawal
2017-04-09 7:59 ` [dpdk-dev] [PATCH v2] mempool/dpaa2: add DPAA2 hardware offloaded mempool Hemant Agrawal
2017-04-10 19:58 ` Olivier MATZ
2017-04-11 5:58 ` Hemant Agrawal [this message]
2017-04-11 7:50 ` Thomas Monjalon
2017-04-11 8:39 ` Ferruh Yigit
2017-04-11 12:50 ` Thomas Monjalon
2017-04-11 12:56 ` Olivier MATZ
2017-04-11 13:54 ` Hemant Agrawal
2017-04-11 13:42 ` [dpdk-dev] [PATCH v3] NXP DPAA2 External Mempool Driver Hemant Agrawal
2017-04-11 13:42 ` [dpdk-dev] [PATCH v3] mempool/dpaa2: add DPAA2 hardware offloaded mempool Hemant Agrawal
2017-04-12 11:31 ` [dpdk-dev] [PATCH v3] NXP DPAA2 External Mempool Driver Ferruh Yigit
2017-04-12 13:50 ` Ferruh Yigit
2017-03-17 13:42 ` [dpdk-dev] [PATCH v1] " Thomas Monjalon
2017-03-17 17:12 ` Hemant Agrawal
2017-03-17 17:22 ` Olivier Matz
2017-03-20 10:08 ` Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 00/21] NXP DPAA2 FSLMC Bus driver Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 01/21] mk/dpaa2: add the crc support to the machine type Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 02/21] bus/fslmc: introducing fsl-mc bus driver Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 03/21] bus/fslmc: add QBMAN driver to bus Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 04/21] bus/fslmc: introduce MC object functions Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 05/21] bus/fslmc: add mc dpio object support Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 06/21] bus/fslmc: add mc dpbp " Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 07/21] eal/vfio: adding vfio utility functions in map file Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 08/21] bus/fslmc: add vfio support Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 09/21] bus/fslmc: scan for net and sec device Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 10/21] bus/fslmc: add debug log support Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 11/21] bus/fslmc: dpio portal driver Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 12/21] bus/fslmc: introduce support for hardware mempool object Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 13/21] bus/fslmc: affine dpio to crypto threads Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 14/21] bus/fslmc: define queues for DPAA2 devices Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 15/21] bus/fslmc: define hardware annotation area size Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 16/21] bus/fslmc: introduce true and false macros Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 17/21] bus/fslmc: define VLAN header length Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 18/21] bus/fslmc: add packet FLE definitions Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 19/21] bus/fslmc: add physical-virtual address translation helpers Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 20/21] bus/fslmc: add support for DMA mapping for ARM SMMU Hemant Agrawal
2017-04-09 7:50 ` [dpdk-dev] [PATCH v3 21/21] bus/fslmc: frame queue based dq storage alloc Hemant Agrawal
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=9977cf0c-e1cf-b235-e199-c5242daa75c2@nxp.com \
--to=hemant.agrawal@nxp.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=olivier.matz@6wind.com \
--cc=shreyansh.jain@nxp.com \
--cc=thomas.monjalon@6wind.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).