Deferred probe times

Deferred probe times

[Federico Giovanardi]

[-- Attachment #1: Type: text/plain, Size: 328 bytes --]

Hello,

I was poking around the boot data database while I noticed that the 
deferred_probe initicalls are considered as a giant block;

There is a patch that looks never went upstream to measure them:  
https://lkml.iu.edu/1707.3/02349.html

And another one from me to print the result in the .svg.


Have a nice day
Federico


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: update_bootgraph_deferred_init.patch --]
[-- Type: text/x-diff; name=update_bootgraph_deferred_init.patch, Size: 961 bytes --]

commit c8d97bad2987b33b2ce804805071218fd7f68bbd
Author: Federico Giovanardi <federico.giovanardi@cnhind.com>
Date:   Fri Oct 4 14:13:30 2024 +0200

    Update bootgraph to handle deferred initcalls

diff --git a/scripts/bootgraph.pl b/scripts/bootgraph.pl
index 79c903292ae8..8057d9f474df 100755
--- a/scripts/bootgraph.pl
+++ b/scripts/bootgraph.pl
@@ -67,7 +67,8 @@ my $cyheader = 0;
 
 while (<>) {
 	my $line = $_;
-	if ($line =~ /([0-9\.]+)\] calling  ([a-zA-Z0-9\_\.]+)\+/) {
+	if ($line =~ /([0-9\.]+)\] calling  ([a-zA-Z0-9\_\.]+)\+/ or
+		$line =~ /([0-9\.]+)\] deferred probe ([a-zA-Z0-9\_\.]+) @/) {
 		my $func = $2;
 		if ($done == 0) {
 			$start{$func} = $1;
@@ -109,6 +110,12 @@ while (<>) {
 			$maxtime = $1;
 		}
 	}
+	if ($line =~ /([0-9\.]+)\] deferred probe ([a-zA-Z0-9\_\.]+).*returned/) {
+		if ($done == 0) {
+			$end{$2} = $1;
+			$maxtime = $1;
+		}
+	}
 
 	if ($line =~ /([0-9\.]+)\] async_continuing @ ([0-9]+)/) {
 		my $pid = $2;

RE: Deferred probe times

[Bird, Tim]

> -----Original Message-----
> From: Federico Giovanardi <federico@giovanardi.dev>
> Hello,
> 
> I was poking around the boot data database while I noticed that the
> deferred_probe initicalls are considered as a giant block;

Hmm.  I didn't think anyone was looking at this data besides myself.
The site is really still under construction, so I apologize for missing
and mis-formatted data.  But thanks for looking at
the data and finding something that needs attention upstream!!

> 
> There is a patch that looks never went upstream to measure them:
> https://lkml.iu.edu/1707.3/02349.html

Wow, that's interesting.  Here's the lore link for this thread:
https://lore.kernel.org/lkml/20170725233159.89543-1-toddpoynor@google.com/#t

It doesn't look like there were (after discussion) objections to accepting upstream.
I wonder what happened.

...
It looks like it was superceded by this patch:
https://lore.kernel.org/lkml/20180620023903.242758-1-toddpoynor@gmail.com/
which may provide some of the data necessary.

The thing missing there are some (IMHO) better name lookups for the probe function.
I wrote a patch for that, but didn't persist in trying to upstream it.  Maybe I should
dust that off and send it again.

> 
> And another one from me to print the result in the .svg.
> 
> 
> Have a nice day
> Federico

Re: Deferred probe times

[Federico Giovanardi]

[-- Attachment #1: Type: text/plain, Size: 1641 bytes --]


Well, the comment about the giant block was not a complaint, of course.

With both patches, I used to get the attached image, so the 
deferred_probe slot it's separated into its constituents.


Thanks

Federico


On 2025-05-24 02:16, Bird, Tim wrote:
>> -----Original Message-----
>> From: Federico Giovanardi <federico@giovanardi.dev>
>> Hello,
>> 
>> I was poking around the boot data database while I noticed that the
>> deferred_probe initicalls are considered as a giant block;
> 
> Hmm.  I didn't think anyone was looking at this data besides myself.
> The site is really still under construction, so I apologize for missing
> and mis-formatted data.  But thanks for looking at
> the data and finding something that needs attention upstream!!
> 
>> 
>> There is a patch that looks never went upstream to measure them:
>> https://lkml.iu.edu/1707.3/02349.html
> 
> Wow, that's interesting.  Here's the lore link for this thread:
> https://lore.kernel.org/lkml/20170725233159.89543-1-toddpoynor@google.com/#t
> 
> It doesn't look like there were (after discussion) objections to 
> accepting upstream.
> I wonder what happened.
> 
> ...
> It looks like it was superceded by this patch:
> https://lore.kernel.org/lkml/20180620023903.242758-1-toddpoynor@gmail.com/
> which may provide some of the data necessary.
> 
> The thing missing there are some (IMHO) better name lookups for the 
> probe function.
> I wrote a patch for that, but didn't persist in trying to upstream it.  
> Maybe I should
> dust that off and send it again.
> 
>> 
>> And another one from me to print the result in the .svg.
>> 
>> 
>> Have a nice day
>> Federico

[-- Attachment #2: image.png --]
[-- Type: image/png, Size: 73235 bytes --]