From: Ferruh Yigit <ferruh.yigit@xilinx.com>
To: "lihuisong (C)" <lihuisong@huawei.com>,
Xiaoyun Li <xiaoyun.li@intel.com>,
Aman Singh <aman.deep.singh@intel.com>,
Yuying Zhang <yuying.zhang@intel.com>,
Helin Zhang <helin.zhang@intel.com>,
Jingjing Wu <jingjing.wu@intel.com>,
Thomas Monjalon <thomas.monjalon@6wind.com>
Cc: <dev@dpdk.org>, Thomas Monjalon <thomas@monjalon.net>,
Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>,
<stable@dpdk.org>
Subject: Re: [PATCH 1/3] app/testpmd: fix displaying RSS info
Date: Mon, 30 May 2022 14:02:12 +0100 [thread overview]
Message-ID: <e739ca7a-11c1-6b6c-22db-019c788cdf18@xilinx.com> (raw)
In-Reply-To: <f43997dd-1071-61ca-0ed2-07c33e113d86@huawei.com>
On 5/30/2022 1:32 PM, lihuisong (C) wrote:
> CAUTION: This message has originated from an External Source. Please use
> proper judgment and caution when opening attachments, clicking links, or
> responding to this email.
>
>
> 在 2022/5/30 18:43, Ferruh Yigit 写道:
>> On 5/27/2022 3:30 AM, lihuisong (C) wrote:
>>> CAUTION: This message has originated from an External Source. Please
>>> use proper judgment and caution when opening attachments, clicking
>>> links, or responding to this email.
>>>
>>>
>>> 在 2022/5/26 1:37, Ferruh Yigit 写道:
>>>> When supported RSS offload flow types are printed via 'show port
>>>> info #'
>>>> command, flow names are get from flow type array which is wrong and
>>>> causing some RSS flow types not being displayed.
>>>>
>>>> Instead RSS flow type array should be used. Also helper functions added
>>>> and existing code updated to use helpers.
>>>>
>>>> Fixes: b12964f621dc ("ethdev: unification of RSS offload types")
>>>> Cc: stable@dpdk.org
>>>>
>>>> Signed-off-by: Ferruh Yigit <ferruh.yigit@xilinx.com>
>>>> ---
>>>> Cc: helin.zhang@intel.com
>>>>
>>>> Note:
>>>> In ethdev, flow type macros 'RTE_ETH_FLOW_*' and RSS type macros
>>>> 'RTE_ETH_RSS_*' are related, buy they seems diverged a little, may need
>>>> to check that too.
>>>> ---
>>>> app/test-pmd/config.c | 92
>>>> ++++++++++++++++++++++++++++++-------------
>>>> 1 file changed, 64 insertions(+), 28 deletions(-)
>>>>
>>>> diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
>>>> index 72d2606d19d5..c353d224ef06 100644
>>>> --- a/app/test-pmd/config.c
>>>> +++ b/app/test-pmd/config.c
>>>> @@ -147,6 +147,32 @@ const struct rss_type_info rss_type_table[] = {
>>>> { NULL, 0 },
>>>> };
>>>>
>>>> +static const char *
>>>> +rsstype_to_str(uint64_t rss_type)
>>>> +{
>>>> + int i;
>>>> +
>>>> + for (i = 0; rss_type_table[i].str != NULL; i++) {
>>>> + if (rss_type_table[i].rss_type == rss_type)
>>>> + return rss_type_table[i].str;
>>>> + }
>>>> +
>>>> + return NULL;
>>>> +}
>>>> +
>>>> +static uint64_t
>>>> +str_to_rsstype(const char *str)
>>>> +{
>>>> + int i;
>>>> +
>>>> + for (i = 0; rss_type_table[i].str != NULL; i++) {
>>>> + if (!strcmp(rss_type_table[i].str, str))
>>>> + return rss_type_table[i].rss_type;
>>>> + }
>>>> +
>>>> + return 0;
>>>> +}
>>>> +
>>>> static const struct {
>>>> enum rte_eth_fec_mode mode;
>>>> const char *name;
>>>> @@ -779,19 +805,21 @@ port_infos_display(portid_t port_id)
>>>> if (!dev_info.flow_type_rss_offloads)
>>>> printf("No RSS offload flow type is supported.\n");
>>>> else {
>>>> + uint64_t rss_types = dev_info.flow_type_rss_offloads;
>>>> uint16_t i;
>>>> - char *p;
>>>>
>>>> printf("Supported RSS offload flow types:\n");
>>>> - for (i = RTE_ETH_FLOW_UNKNOWN + 1;
>>>> - i < sizeof(dev_info.flow_type_rss_offloads) *
>>>> CHAR_BIT; i++) {
>>>> - if (!(dev_info.flow_type_rss_offloads & (1ULL
>>>> << i)))
>>>> - continue;
>>>> - p = flowtype_to_str(i);
>>>> - if (p)
>>>> - printf(" %s\n", p);
>>>> - else
>>>> - printf(" user defined %d\n", i);
>>>> + for (i = 0; rss_types != 0; i++) {
>>>> + if (rss_types & 1) {
>>>> + uint64_t rss_type = 1ULL << i;
>>>> + const char *p = rsstype_to_str(rss_type);
>>>> +
>>>> + if (p)
>>>> + printf(" %s\n", p);
>>>> + else
>>>> + printf(" user defined
>>>> 0x%"PRIx64"\n", rss_type);
> Maybe we need to add a mapping table between RSS offload and name for
> 'flow_type_rss_offloads'.
>>>> + }
>>>> + rss_types >>= 1;
>>>> }
>>>> }
>>>>
>>>> @@ -1547,6 +1575,7 @@ port_flow_complain(struct rte_flow_error *error)
>>>> static void
>>>> rss_config_display(struct rte_flow_action_rss *rss_conf)
>>>> {
>>>> + uint64_t rss_types;
>>>> uint8_t i;
>>>>
>>>> if (rss_conf == NULL) {
>>>> @@ -1582,16 +1611,23 @@ rss_config_display(struct
>>>> rte_flow_action_rss *rss_conf)
>>>> }
>>>>
>>>> printf(" types:\n");
>>>> - if (rss_conf->types == 0) {
>>>> + rss_types = rss_conf->types;
>>>> + if (rss_types == 0) {
>>>> printf(" none\n");
>>>> return;
>>>> }
>>>> - for (i = 0; rss_type_table[i].str; i++) {
>>>> - if ((rss_conf->types &
>>>> - rss_type_table[i].rss_type) ==
>>>> - rss_type_table[i].rss_type &&
>>>> - rss_type_table[i].rss_type != 0)
>>>> - printf(" %s\n", rss_type_table[i].str);
>>>> +
>>>> + for (i = 0; rss_types != 0; i++) {
>>>> + if (rss_types & 1) {
>>>> + uint64_t rss_type = 1ULL << i;
>>>> + const char *p = rsstype_to_str(rss_type);
>>> It seems that we can't use one bit to get rss type name.
>>> Because part of name in rss_type_table[] consist of multiple bits.
>>> Maybe it's better to use the original method to display RSS type name.
>>>
>>
>> Right, it doesn't cover 'all' case, but thinking twice how useful
>> 'all' case is, it doesn't really cover all options, and for user it is
>> hard to know which functions are set when it displays 'all'.
>> What about to remove 'all' item from 'rss_type_table[]'?
> It seems that remove 'all' item isn't the way to resolve this issue.
> There are other iterms in 'rss_type_table[], such as, 'ip', 'udp',
> 'tcp' and so on. The 'rss_type_table[]' should primarily serve
> the "show port rss xxx" and "port confg rss xxx" commands.
> They all need these.
>
You are right, it is not just 'all'.
>>
>>> On the other hand, RSS type name are printed in many places as this
>>> patch modified.
>>> It is recommended that you encapsulate a funcion to display RSS type.
>>
>> Yes but each are formatting slightly different, so to let various
>> formatting possible, common function returns string instead of
>> printing them. As 3/3 of this set does some formatting.
>
> We can unify this display if we add the mapping table mentioned above.
>
If it can be unified, agree that this simplifies the code [2].
Both above two changes already exists in your set [1], if you don't mind
to adapt 2/3 & 3/3 of this set into yours, we can continue with your
set, is this OK for you?
[1]
https://patches.dpdk.org/project/dpdk/list/?series=22735
[2]
In your 'show_rss_types()', it can get additional parameter for
formatting, like after how many items to break to next line, this can
cover existing formatting options, but not sure how elegant a solution
it is.
>>
>>>> +
>>>> + if (p)
>>>> + printf(" %s\n", p);
>>>> + else
>>>> + printf(" user defined 0x%"PRIx64"\n",
>>>> rss_type);
>>>> + }
>>>> + rss_types >>= 1;
>>>> }
>>>> }
>>>>
>>>> @@ -3823,11 +3859,16 @@ port_rss_hash_conf_show(portid_t port_id,
>>>> int show_rss_key)
>>>> return;
>>>> }
>>>> printf("RSS functions:\n ");
>>>> - for (i = 0; rss_type_table[i].str; i++) {
>>>> - if (rss_type_table[i].rss_type == 0)
>>>> - continue;
>>>> - if ((rss_hf & rss_type_table[i].rss_type) ==
>>>> rss_type_table[i].rss_type)
>>>> - printf("%s ", rss_type_table[i].str);
>>>> + for (i = 0; rss_hf != 0; i++) {
>>>> + if (rss_hf & 1) {
>>>> + uint64_t rss_type = 1ULL << i;
>>>> + const char *p = rsstype_to_str(rss_type);
>>>> + if (p)
>>>> + printf("%s ", p);
>>>> + else
>>>> + printf("0x%"PRIx64" ", rss_type);
>>>> + }
>>>> + rss_hf >>= 1;
>>>> }
>>>> printf("\n");
>>>> if (!show_rss_key)
>>>> @@ -3844,15 +3885,10 @@ port_rss_hash_key_update(portid_t port_id,
>>>> char rss_type[], uint8_t *hash_key,
>>>> {
>>>> struct rte_eth_rss_conf rss_conf;
>>>> int diag;
>>>> - unsigned int i;
>>>>
>>>> rss_conf.rss_key = NULL;
>>>> rss_conf.rss_key_len = 0;
>>>> - rss_conf.rss_hf = 0;
>>>> - for (i = 0; rss_type_table[i].str; i++) {
>>>> - if (!strcmp(rss_type_table[i].str, rss_type))
>>>> - rss_conf.rss_hf = rss_type_table[i].rss_type;
>>>> - }
>>>> + rss_conf.rss_hf = str_to_rsstype(rss_type);
>>>> diag = rte_eth_dev_rss_hash_conf_get(port_id, &rss_conf);
>>>> if (diag == 0) {
>>>> rss_conf.rss_key = hash_key;
>>
>> .
next prev parent reply other threads:[~2022-05-30 13:02 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-25 17:37 Ferruh Yigit
2022-05-26 3:52 ` lihuisong (C)
2022-05-26 8:56 ` Ferruh Yigit
2022-05-27 2:48 ` lihuisong (C)
2022-05-27 2:30 ` lihuisong (C)
2022-05-30 10:43 ` Ferruh Yigit
2022-05-30 12:32 ` lihuisong (C)
2022-05-30 13:02 ` Ferruh Yigit [this message]
2022-05-31 2:07 ` lihuisong (C)
2022-05-31 16:35 ` Andrew Rybchenko
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=e739ca7a-11c1-6b6c-22db-019c788cdf18@xilinx.com \
--to=ferruh.yigit@xilinx.com \
--cc=aman.deep.singh@intel.com \
--cc=andrew.rybchenko@oktetlabs.ru \
--cc=dev@dpdk.org \
--cc=helin.zhang@intel.com \
--cc=jingjing.wu@intel.com \
--cc=lihuisong@huawei.com \
--cc=stable@dpdk.org \
--cc=thomas.monjalon@6wind.com \
--cc=thomas@monjalon.net \
--cc=xiaoyun.li@intel.com \
--cc=yuying.zhang@intel.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).