[v7,02/25] ethdev: add a link status text representation

  This commit add function which treat link status structure
and format it to text representation.

Signed-off-by: Ivan Dyukov <i.dyukov@samsung.com>
---
 MAINTAINERS                              |   1 +
 app/test/Makefile                        |   3 +
 app/test/meson.build                     |   2 +
 app/test/test_ethdev_link.c              | 299 +++++++++++++++++++++++
 lib/librte_ethdev/rte_ethdev.c           | 174 +++++++++++++
 lib/librte_ethdev/rte_ethdev.h           |  54 ++++
 lib/librte_ethdev/rte_ethdev_version.map |   4 +
 7 files changed, 537 insertions(+)
 create mode 100644 app/test/test_ethdev_link.c
  

On 7/10/2020 8:02 AM, Ivan Dyukov wrote:
> This commit add function which treat link status structure
> and format it to text representation.
> 
> Signed-off-by: Ivan Dyukov <i.dyukov@samsung.com>

<...>

> +static int
> +rte_eth_link_strf_parser(char *str, size_t len, const char *const fmt,
> +			   const struct rte_eth_link *link)
> +{
> +	size_t offset = 0;
> +	const char *fmt_cur = fmt;
> +	char *str_cur = str;
> +	double gbits = (double)link->link_speed / 1000.;
> +	static const char autoneg_str[]       = "Autoneg";
> +	static const char fixed_str[]         = "Fixed";
> +	static const char fdx_str[]           = "FDX";
> +	static const char hdx_str[]           = "HDX";
> +	static const char unknown_str[]       = "Unknown";
> +	static const char up_str[]            = "Up";
> +	static const char down_str[]          = "Down";
> +	char gbits_str[20];
> +	char mbits_str[20];
> +
> +	/* preformat complex formatting to easily concatinate it further */
> +	snprintf(mbits_str, sizeof(mbits_str), "%u", link->link_speed);
> +	snprintf(gbits_str, sizeof(gbits_str), "%.1f", gbits);
> +	/* init str before formatting */
> +	str[0] = 0;
> +	while (*fmt_cur) {
> +		/* check str bounds */
> +		if (offset > (len - 1)) {
> +			str[len - 1] = '\0';
> +			return -1;
> +		}
> +		if (*fmt_cur == '%') {
> +			/* set null terminator to current position,
> +			 * it's required for strlcat
> +			 */
> +			*str_cur = '\0';
> +			switch (*++fmt_cur) {
> +			/* Speed in Mbits/s */
> +			case 'M':
> +				if (link->link_speed ==
> +				    ETH_SPEED_NUM_UNKNOWN)
> +					offset = strlcat(str, unknown_str,
> +							 len);
> +				else
> +					offset = strlcat(str, mbits_str, len);
> +				break;
> +			/* Speed in Gbits/s */
> +			case 'G':
> +				if (link->link_speed ==
> +				    ETH_SPEED_NUM_UNKNOWN)
> +					offset = strlcat(str, unknown_str,
> +							 len);
> +				else
> +					offset = strlcat(str, gbits_str, len);
> +				break;
> +			/* Link status */
> +			case 'S':
> +				offset = strlcat(str, link->link_status ?
> +					up_str : down_str, len);
> +				break;
> +			/* Link autoneg */
> +			case 'A':
> +				offset = strlcat(str, link->link_autoneg ?
> +					autoneg_str : fixed_str, len);
> +				break;
> +			/* Link duplex */
> +			case 'D':
> +				offset = strlcat(str, link->link_duplex ?
> +					fdx_str : hdx_str, len);
> +				break;
> +			/* ignore unknown specifier */
> +			default:
> +				*str_cur = '%';
> +				offset++;
> +				fmt_cur--;
> +				break;

What do you think ignoring the unknown specifiers and keep continue
processing the string, instead of break? Just keep unknown specifier
as it is in the output string.

> +
> +			}
> +			if (offset > (len - 1))
> +				return -1;

For me "offset >= len" is simpler than "offset > (len - 1)", also it prevents any possible error when "len == 0" ('len' is unsigned) although I can see you have check for it.
Anyway both deos same thing, up to you.

> +
> +			str_cur = str + offset;
> +		} else {
> +			*str_cur++ = *fmt_cur;
> +			offset++;

Why keeping both offset and the pointer ('str_cur'), just offset should be enough "str[offset++] = *fmt_cur;", I think this simplifies a little bit.

> +		}
> +		fmt_cur++;
> +	}
> +	*str_cur = '\0';
> +	return offset;
> +}
> +
> +int
> +rte_eth_link_printf(const char *const fmt,
> +		    const struct rte_eth_link *link)
> +{
> +	char text[200];
> +	int ret;
> +
> +	ret = rte_eth_link_strf(text, 200, fmt, link);
> +	if (ret > 0)
> +		printf("%s", text);
> +	return ret;
> +}
> +
> +int
> +rte_eth_link_strf(char *str, size_t len, const char *const fmt,
> +		    const struct rte_eth_link *link)
> +{
> +	size_t offset = 0;
> +	double gbits = (double)link->link_speed / 1000.;
> +	char speed_gbits_str[20];
> +	char speed_mbits_str[20];
> +	/* TBD: make it international? */
> +	static const char link_down_str[]     = "Link down\n";
> +	static const char link_up_str[]       = "Link up at ";
> +	static const char unknown_speed_str[] = "Unknown speed ";
> +	static const char mbits_str[]	      = "Mbit/s";
> +	static const char gbits_str[]	      = "Gbit/s";
> +	static const char fdx_str[]           = "FDX ";
> +	static const char hdx_str[]           = "HDX ";
> +	/* autoneg is latest param in default string, so add '\n' */
> +	static const char autoneg_str[]       = "Autoneg\n";
> +	static const char fixed_str[]         = "Fixed\n";

Please put an empty line after variables.

<...>

> +/**
> + * Format link status to textual representation. This function threats all
> + * special values like ETH_SPEED_NUM_UNKNOWN, ETH_LINK_DOWN etc. and convert
> + * them to textual representation.
> + *
> + * @param str
> + *   A pointer to a string to be filled with textual representation of
> + *   device status.
> + * @param len
> + *   Length of available memory at 'str' string.
> + * @param fmt
> + *   Format string which allow to format link status. If NULL is provided
> + *   , default formatting will be applied.
> + *   Following specifiers are available:
> + *    - '%M' link speed in Mbits/s
> + *    - '%G' link speed in Gbits/s
> + *    - '%S' link status. e.g. Up or Down
> + *    - '%A' link autonegotiation state
> + *    - '%D' link duplex state
> + * @param link

It should be 'eth_link' otherwise doxygen is complaining.

> + *   Link status provided by rte_eth_link_get function
> + * @return
> + *   - Number of bytes written to str array. In case of error, -1 is returned.
> + *
> + */
> +__rte_experimental
> +int rte_eth_link_strf(char *str, size_t len, const char *const fmt,
> +			const struct rte_eth_link *eth_link);
> +

For the API name, I guess 'strf' is for 'stringify' but what do you think about
'rte_eth_link_to_str()', I think more clear and simpler.
  

10/07/2020 09:02, Ivan Dyukov:
> This commit add function which treat link status structure
> and format it to text representation.

It is missing an explanation about why it is required.
Which problem is it solving?

> Signed-off-by: Ivan Dyukov <i.dyukov@samsung.com>

I'm surprised there is not so much review on this patch.

[...]
> --- a/lib/librte_ethdev/rte_ethdev.h
> +++ b/lib/librte_ethdev/rte_ethdev.h
> +/**
> + * print formatted link status to stdout. This function threats all
> + * special values like ETH_SPEED_NUM_UNKNOWN, ETH_LINK_DOWN etc. and convert
> + * them to textual representation.

I don't understand the need for this function.
If needed, the application can send the output of rte_eth_link_strf()
to fprintf or a log function.

> + *
> + * @param fmt
> + *   Format string which allow to format link status. If NULL is provided
> + *   , default formatting will be applied.
> + *   Following specifiers are available:
> + *    - '%M' link speed in Mbits/s
> + *    - '%G' link speed in Gbits/s
> + *    - '%S' link status. e.g. Up or Down
> + *    - '%A' link autonegotiation state
> + *    - '%D' link duplex state
> + * @param link
> + *   Link status provided by rte_eth_link_get function
> + * @return
> + *   - Number of bytes written to stdout. In case of error, -1 is returned.
> + *
> + */
> +__rte_experimental
> +int rte_eth_link_printf(const char *const fmt,
> +			const struct rte_eth_link *link);
> +
> +/**
> + * Format link status to textual representation. This function threats all

not "threats"

> + * special values like ETH_SPEED_NUM_UNKNOWN, ETH_LINK_DOWN etc. and convert
> + * them to textual representation.
> + *
> + * @param str
> + *   A pointer to a string to be filled with textual representation of
> + *   device status.
> + * @param len
> + *   Length of available memory at 'str' string.
> + * @param fmt
> + *   Format string which allow to format link status. If NULL is provided
> + *   , default formatting will be applied.

Please do not start a line with a comma,
it should be ending the previous line, like here ;)
Even better: start the new sentence on a new line.

> + *   Following specifiers are available:
> + *    - '%M' link speed in Mbits/s
> + *    - '%G' link speed in Gbits/s
> + *    - '%S' link status. e.g. Up or Down
> + *    - '%A' link autonegotiation state
> + *    - '%D' link duplex state

These specifiers look OK.

> + * @param link
> + *   Link status provided by rte_eth_link_get function
> + * @return
> + *   - Number of bytes written to str array. In case of error, -1 is returned.

Better to have error case on a separate line.

> + *
> + */
> +__rte_experimental
> +int rte_eth_link_strf(char *str, size_t len, const char *const fmt,
> +			const struct rte_eth_link *eth_link);
  

On Fri, 10 Jul 2020 13:06:02 +0000
"Yigit, Ferruh" <ferruh.yigit@intel.com> wrote:

> > +			/* ignore unknown specifier */
> > +			default:
> > +				*str_cur = '%';
> > +				offset++;
> > +				fmt_cur--;
> > +				break;  
> 
> What do you think ignoring the unknown specifiers and keep continue
> processing the string, instead of break? Just keep unknown specifier
> as it is in the output string.

My comment was that the function should behave the same as
snprintf() does when handed unknown specifier. What snprintf
does is duplicate the string in the output buffer (like any
other non format character).
  

> On Fri, 10 Jul 2020 13:06:02 +0000
> "Yigit, Ferruh" <ferruh.yigit@intel.com> wrote:
> 
> > > +			/* ignore unknown specifier */
> > > +			default:
> > > +				*str_cur = '%';
> > > +				offset++;
> > > +				fmt_cur--;
> > > +				break;
> >
> > What do you think ignoring the unknown specifiers and keep continue
> > processing the string, instead of break? Just keep unknown specifier
> > as it is in the output string.
> 
> My comment was that the function should behave the same as
> snprintf() does when handed unknown specifier. What snprintf does is
> duplicate the string in the output buffer (like any other non format
> character).

+1
  

10.07.2020 16:06, Yigit, Ferruh пишет:
> On 7/10/2020 8:02 AM, Ivan Dyukov wrote:
>> This commit add function which treat link status structure
>> and format it to text representation.
>>
>> Signed-off-by: Ivan Dyukov <i.dyukov@samsung.com>
> <...>
>
>> +static int
>> +rte_eth_link_strf_parser(char *str, size_t len, const char *const fmt,
>> +			   const struct rte_eth_link *link)
>> +{
>> +	size_t offset = 0;
>> +	const char *fmt_cur = fmt;
>> +	char *str_cur = str;
>> +	double gbits = (double)link->link_speed / 1000.;
>> +	static const char autoneg_str[]       = "Autoneg";
>> +	static const char fixed_str[]         = "Fixed";
>> +	static const char fdx_str[]           = "FDX";
>> +	static const char hdx_str[]           = "HDX";
>> +	static const char unknown_str[]       = "Unknown";
>> +	static const char up_str[]            = "Up";
>> +	static const char down_str[]          = "Down";
>> +	char gbits_str[20];
>> +	char mbits_str[20];
>> +
>> +	/* preformat complex formatting to easily concatinate it further */
>> +	snprintf(mbits_str, sizeof(mbits_str), "%u", link->link_speed);
>> +	snprintf(gbits_str, sizeof(gbits_str), "%.1f", gbits);
>> +	/* init str before formatting */
>> +	str[0] = 0;
>> +	while (*fmt_cur) {
>> +		/* check str bounds */
>> +		if (offset > (len - 1)) {
>> +			str[len - 1] = '\0';
>> +			return -1;
>> +		}
>> +		if (*fmt_cur == '%') {
>> +			/* set null terminator to current position,
>> +			 * it's required for strlcat
>> +			 */
>> +			*str_cur = '\0';
>> +			switch (*++fmt_cur) {
>> +			/* Speed in Mbits/s */
>> +			case 'M':
>> +				if (link->link_speed ==
>> +				    ETH_SPEED_NUM_UNKNOWN)
>> +					offset = strlcat(str, unknown_str,
>> +							 len);
>> +				else
>> +					offset = strlcat(str, mbits_str, len);
>> +				break;
>> +			/* Speed in Gbits/s */
>> +			case 'G':
>> +				if (link->link_speed ==
>> +				    ETH_SPEED_NUM_UNKNOWN)
>> +					offset = strlcat(str, unknown_str,
>> +							 len);
>> +				else
>> +					offset = strlcat(str, gbits_str, len);
>> +				break;
>> +			/* Link status */
>> +			case 'S':
>> +				offset = strlcat(str, link->link_status ?
>> +					up_str : down_str, len);
>> +				break;
>> +			/* Link autoneg */
>> +			case 'A':
>> +				offset = strlcat(str, link->link_autoneg ?
>> +					autoneg_str : fixed_str, len);
>> +				break;
>> +			/* Link duplex */
>> +			case 'D':
>> +				offset = strlcat(str, link->link_duplex ?
>> +					fdx_str : hdx_str, len);
>> +				break;
>> +			/* ignore unknown specifier */
>> +			default:
>> +				*str_cur = '%';
>> +				offset++;
>> +				fmt_cur--;
>> +				break;
> What do you think ignoring the unknown specifiers and keep continue
> processing the string, instead of break? Just keep unknown specifier
> as it is in the output string.
yep. it's exactly what code do.  break exit from the switch but not from 
string processing.  I have unit tests for this case. They  work fine. 
Please review unit tests and send me more cases if they need to be tested.
>
>> +
>> +			}
>> +			if (offset > (len - 1))
>> +				return -1;
> For me "offset >= len" is simpler than "offset > (len - 1)", also it prevents any possible error when "len == 0" ('len' is unsigned) although I can see you have check for it.
> Anyway both deos same thing, up to you.
>
>> +
>> +			str_cur = str + offset;
>> +		} else {
>> +			*str_cur++ = *fmt_cur;
>> +			offset++;
> Why keeping both offset and the pointer ('str_cur'), just offset should be enough "str[offset++] = *fmt_cur;", I think this simplifies a little bit.
>
>> +		}
>> +		fmt_cur++;
>> +	}
>> +	*str_cur = '\0';
>> +	return offset;
>> +}
>> +
>> +int
>> +rte_eth_link_printf(const char *const fmt,
>> +		    const struct rte_eth_link *link)
>> +{
>> +	char text[200];
>> +	int ret;
>> +
>> +	ret = rte_eth_link_strf(text, 200, fmt, link);
>> +	if (ret > 0)
>> +		printf("%s", text);
>> +	return ret;
>> +}
>> +
>> +int
>> +rte_eth_link_strf(char *str, size_t len, const char *const fmt,
>> +		    const struct rte_eth_link *link)
>> +{
>> +	size_t offset = 0;
>> +	double gbits = (double)link->link_speed / 1000.;
>> +	char speed_gbits_str[20];
>> +	char speed_mbits_str[20];
>> +	/* TBD: make it international? */
>> +	static const char link_down_str[]     = "Link down\n";
>> +	static const char link_up_str[]       = "Link up at ";
>> +	static const char unknown_speed_str[] = "Unknown speed ";
>> +	static const char mbits_str[]	      = "Mbit/s";
>> +	static const char gbits_str[]	      = "Gbit/s";
>> +	static const char fdx_str[]           = "FDX ";
>> +	static const char hdx_str[]           = "HDX ";
>> +	/* autoneg is latest param in default string, so add '\n' */
>> +	static const char autoneg_str[]       = "Autoneg\n";
>> +	static const char fixed_str[]         = "Fixed\n";
> Please put an empty line after variables.
>
> <...>
>
>> +/**
>> + * Format link status to textual representation. This function threats all
>> + * special values like ETH_SPEED_NUM_UNKNOWN, ETH_LINK_DOWN etc. and convert
>> + * them to textual representation.
>> + *
>> + * @param str
>> + *   A pointer to a string to be filled with textual representation of
>> + *   device status.
>> + * @param len
>> + *   Length of available memory at 'str' string.
>> + * @param fmt
>> + *   Format string which allow to format link status. If NULL is provided
>> + *   , default formatting will be applied.
>> + *   Following specifiers are available:
>> + *    - '%M' link speed in Mbits/s
>> + *    - '%G' link speed in Gbits/s
>> + *    - '%S' link status. e.g. Up or Down
>> + *    - '%A' link autonegotiation state
>> + *    - '%D' link duplex state
>> + * @param link
> It should be 'eth_link' otherwise doxygen is complaining.
>
>> + *   Link status provided by rte_eth_link_get function
>> + * @return
>> + *   - Number of bytes written to str array. In case of error, -1 is returned.
>> + *
>> + */
>> +__rte_experimental
>> +int rte_eth_link_strf(char *str, size_t len, const char *const fmt,
>> +			const struct rte_eth_link *eth_link);
>> +
> For the API name, I guess 'strf' is for 'stringify' but what do you think about
> 'rte_eth_link_to_str()', I think more clear and simpler.
  

10.07.2020 18:22, Stephen Hemminger пишет:
> On Fri, 10 Jul 2020 13:06:02 +0000
> "Yigit, Ferruh" <ferruh.yigit@intel.com> wrote:
>
>>> +			/* ignore unknown specifier */
>>> +			default:
>>> +				*str_cur = '%';
>>> +				offset++;
>>> +				fmt_cur--;
>>> +				break;
>> What do you think ignoring the unknown specifiers and keep continue
>> processing the string, instead of break? Just keep unknown specifier
>> as it is in the output string.
> My comment was that the function should behave the same as
> snprintf() does when handed unknown specifier. What snprintf
> does is duplicate the string in the output buffer (like any
> other non format character).
Looks like I need to rewrite comment "ignore unkown specifier". It's not 
what code do. This function work like snprintf.
  

> Subject: Re: [PATCH v7 02/25] ethdev: add a link status text representation
> 
> 10.07.2020 16:06, Yigit, Ferruh пишет:
> > On 7/10/2020 8:02 AM, Ivan Dyukov wrote:
> >> This commit add function which treat link status structure and format
> >> it to text representation.
> >>
> >> Signed-off-by: Ivan Dyukov <i.dyukov@samsung.com>
> > <...>
> >
> >> +static int
> >> +rte_eth_link_strf_parser(char *str, size_t len, const char *const fmt,
> >> +			   const struct rte_eth_link *link) {
> >> +	size_t offset = 0;
> >> +	const char *fmt_cur = fmt;
> >> +	char *str_cur = str;
> >> +	double gbits = (double)link->link_speed / 1000.;
> >> +	static const char autoneg_str[]       = "Autoneg";
> >> +	static const char fixed_str[]         = "Fixed";
> >> +	static const char fdx_str[]           = "FDX";
> >> +	static const char hdx_str[]           = "HDX";
> >> +	static const char unknown_str[]       = "Unknown";
> >> +	static const char up_str[]            = "Up";
> >> +	static const char down_str[]          = "Down";
> >> +	char gbits_str[20];
> >> +	char mbits_str[20];
> >> +
> >> +	/* preformat complex formatting to easily concatinate it further */
> >> +	snprintf(mbits_str, sizeof(mbits_str), "%u", link->link_speed);
> >> +	snprintf(gbits_str, sizeof(gbits_str), "%.1f", gbits);
> >> +	/* init str before formatting */
> >> +	str[0] = 0;
> >> +	while (*fmt_cur) {
> >> +		/* check str bounds */
> >> +		if (offset > (len - 1)) {
> >> +			str[len - 1] = '\0';
> >> +			return -1;
> >> +		}
> >> +		if (*fmt_cur == '%') {
> >> +			/* set null terminator to current position,
> >> +			 * it's required for strlcat
> >> +			 */
> >> +			*str_cur = '\0';
> >> +			switch (*++fmt_cur) {
> >> +			/* Speed in Mbits/s */
> >> +			case 'M':
> >> +				if (link->link_speed ==
> >> +				    ETH_SPEED_NUM_UNKNOWN)
> >> +					offset = strlcat(str, unknown_str,
> >> +							 len);
> >> +				else
> >> +					offset = strlcat(str, mbits_str, len);
> >> +				break;
> >> +			/* Speed in Gbits/s */
> >> +			case 'G':
> >> +				if (link->link_speed ==
> >> +				    ETH_SPEED_NUM_UNKNOWN)
> >> +					offset = strlcat(str, unknown_str,
> >> +							 len);
> >> +				else
> >> +					offset = strlcat(str, gbits_str, len);
> >> +				break;
> >> +			/* Link status */
> >> +			case 'S':
> >> +				offset = strlcat(str, link->link_status ?
> >> +					up_str : down_str, len);
> >> +				break;
> >> +			/* Link autoneg */
> >> +			case 'A':
> >> +				offset = strlcat(str, link->link_autoneg ?
> >> +					autoneg_str : fixed_str, len);
> >> +				break;
> >> +			/* Link duplex */
> >> +			case 'D':
> >> +				offset = strlcat(str, link->link_duplex ?
> >> +					fdx_str : hdx_str, len);
> >> +				break;
> >> +			/* ignore unknown specifier */
> >> +			default:
> >> +				*str_cur = '%';
> >> +				offset++;
> >> +				fmt_cur--;
> >> +				break;
> > What do you think ignoring the unknown specifiers and keep continue
> > processing the string, instead of break? Just keep unknown specifier
> > as it is in the output string.
> yep. it's exactly what code do.  break exit from the switch but not from string
> processing.  I have unit tests for this case. They  work fine.
> Please review unit tests and send me more cases if they need to be tested.

yes it does as you explained, I miss read it, so it is good, please ignore this comment.