linux-kernel - Re: [PATCH 5/6] eventfs: get rid of dentry pointers without refcounts

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Tue, 30 Jan 2024 13:52:05 -0800
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Steven Rostedt <rostedt@...dmis.org>
Cc: Masami Hiramatsu <mhiramat@...nel.org>, linux-kernel@...r.kernel.org, 
	linux-trace-kernel@...r.kernel.org
Subject: Re: [PATCH 5/6] eventfs: get rid of dentry pointers without refcounts

On Tue, 30 Jan 2024 at 12:55, Steven Rostedt <rostedt@...dmis.org> wrote:
>
> I'm going to be putting back the ei->name pointer as the above actually
> adds more memory usage.

I did it mainly because I hate having multiple different allocation
sites that then have to do that kref_init() etc individually, and once
there was a single site the "name" thing really looked lik ean obvious
simplification.

That said, I think you're confused about the memory usage.

Sure, 'kstrdup_const()' optimizes away the allocation for static
constant strings, but what it does *not* do is to optimize away the
pointer.

In contrast, allocating them together gets rid of the pointer itself,
because now the name is just an offset in the structure.

And the pointer is 8 bytes right there.

So allocating the string _with_ the ei will always save at least 8 bytes.

So whenever the string is less than that in length it's *always* a win.

And even when it's not an absolute win, it will take advantage of the
kmalloc() quantized sizes too, and generally not be a loss even with
longer names.

So I'm pretty sure you are simply entirely wrong on the memory usage.
Not counting the size of the pointer is overlooking a big piece of the
puzzle.

Btw, you can look at name lengths in tracefs with something stupid like this:

    find . | sed 's:^.*/::' | tr -c '\n' . | sort | uniq -c

and you will see that (at least in my case) names of lengths 1..7 are
dominating it all:

      1 .
   2189 ..
     34 ...
   2229 ....
    207 .....
   6833 ......
   2211 .......

with the rest being insignificant in comparison.

The reason? There's a *lot* of those 'filter' and 'enable' and 'id'
files. All of which are better off without a 'const char *name' taking
8 bytes.

              Linus