Re: [ndctl PATCH] cxl/event_trace: parse arrays separately from strings

["Verma, Vishal L" <vishal.l.verma@intel.com>]

On Thu, 2024-02-15 at 22:06 -0800, alison.schofield@intel.com wrote:
> From: Alison Schofield <alison.schofield@intel.com>
> 
> Arrays are being parsed as strings based on a flag that seems like
> it would be the differentiator, ARRAY and STRING, but it is not.
> 
> libtraceevent sets the flags for arrays and strings like this:
> array:  TEP_FIELD_IS_[ARRAY | STRING]
> string: TEP_FIELD_IS_[ARRAY | STRING | DYNAMIC]
> 
> Use TEP_FIELD_IS_DYNAMIC to discover the field type, otherwise arrays
> get parsed as strings and 'cxl monitor' returns gobbledygook in the
> array type fields.
> 
> This fixes the "data" field of cxl_generic_events and the "uuid"
> field
> of cxl_poison.
> 
> Before:
> {"system":"cxl","event":"cxl_generic_event","timestamp":3469041387470
> ,"memdev":"mem0","host":"cxl_mem.0","log":0,"hdr_uuid":"ba5eba11-
> abcd-efeb-a55a-
> a55aa5a55aa5","serial":0,"hdr_flags":8,"hdr_handle":1,"hdr_related_ha
> ndle":42422,"hdr_timestamp":0,"hdr_length":128,"hdr_maint_op_class":0
> ,"data":"Þ­¾ï"}

When applying, b4 complains of these in the commit message as
"suspicious unicode control characters". I'm also not a huge fan of the
super long lines, perhaps we can just drop the before/after examples?

> 
> After:
> {"system":"cxl","event":"cxl_generic_event","timestamp":312851657810,
> "memdev":"mem0","host":"cxl_mem.0","log":0,"hdr_uuid":"ba5eba11-abcd-
> efeb-a55a-
> a55aa5a55aa5","serial":0,"hdr_flags":8,"hdr_handle":1,"hdr_related_ha
> ndle":42422,"hdr_timestamp":0,"hdr_length":128,"hdr_maint_op_class":0
> ,"data":[222,173,190,239,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
> 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
> ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]}
> 
> Before:
> {"system":"cxl","event":"cxl_poison","timestamp":3292418311609,"memde
> v":"mem1","host":"cxl_mem.1","serial":1,"trace_type":2,"region":"regi
> on5","overflow_ts":0,"hpa":1035355557888,"dpa":1073741824,"dpa_length
> ":64,"uuid":"�Fe�c�CI�����2�]","source":0,"flags":0}
> 
> After:
> {"system":"cxl","event":"cxl_poison","timestamp":94600531271,"memdev"
> :"mem1","host":"cxl_mem.1","serial":1,"trace_type":2,"region":"region
> 5","overflow_ts":0,"hpa":1035355557888,"dpa":1073741824,"dpa_length":
> 64,"uuid":[139,200,184,22,236,103,76,121,157,243,47,110,243,11,158,62
> ],"source":0,"flags":0}
> 
> That cxl_poison uuid format can be further improved by using the
> trace
> type (__field_struct uuid_t) in the CXL kernel driver. The parser
> will
> automatically pick up that new type, as illustrated in the "hdr_uuid"
> of cxl_generic_media event trace above.
> 
> Signed-off-by: Alison Schofield <alison.schofield@intel.com>
> ---
>  cxl/event_trace.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/cxl/event_trace.c b/cxl/event_trace.c
> index db8cc85f0b6f..1b5aa09de8b2 100644
> --- a/cxl/event_trace.c
> +++ b/cxl/event_trace.c
> @@ -109,7 +109,13 @@ static int cxl_event_to_json(struct tep_event
> *event, struct tep_record *record,
>  		struct tep_format_field *f = fields[i];
>  		int len;
>  
> -		if (f->flags & TEP_FIELD_IS_STRING) {
> +		/*
> +		 * libtraceevent differentiates arrays and strings
> like this:
> +		 * array:  TEP_FIELD_IS_[ARRAY | STRING]
> +		 * string: TEP_FIELD_IS_[ARRAY | STRING | DYNAMIC]
> +		 */
> +		if ((f->flags & TEP_FIELD_IS_STRING) &&
> +		    ((f->flags & TEP_FIELD_IS_DYNAMIC))) {
>  			char *str;
>  
>  			str = tep_get_field_raw(NULL, event, f-
> >name, record, &len, 0);
> 
> base-commit: a871e6153b11fe63780b37cdcb1eb347b296095c