linux-kernel - Re: [PATCH v2] drop_caches: re-enable message after disabling

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] [day] [month] [year] [list]

Message-Id: <20250314122803.14568-1-rwchen404@gmail.com>
Date: Fri, 14 Mar 2025 20:28:03 +0800
From: Ruiwu Chen <rwchen404@...il.com>
To: joel.granados@...nel.org,
	mcgrof@...nel.org
Cc: corbet@....net,
	keescook@...omium.org,
	linux-doc@...r.kernel.org,
	linux-fsdevel@...r.kernel.org,
	linux-kernel@...r.kernel.org,
	rwchen404@...il.com,
	viro@...iv.linux.org.uk,
	zachwade.k@...il.com
Subject: Re: [PATCH v2] drop_caches: re-enable message after disabling

> On Wed, Mar 12, 2025 at 03:55:22PM -0700, Luis Chamberlain wrote:
> > On Mon, Mar 10, 2025 at 02:51:11PM +0100, Joel Granados wrote:
> > > On Sat, Mar 08, 2025 at 04:05:49PM +0800, Ruiwu Chen wrote:
> > > > >> When 'echo 4 > /proc/sys/vm/drop_caches' the message is disabled,
> > > > >> but there is no interface to enable the message, only by restarting
> > > > >> the way, so add the 'echo 0 > /proc/sys/vm/drop_caches' way to
> > > > >> enabled the message again.
> > > > >> 
> > > > >> Signed-off-by: Ruiwu Chen <rwchen404@...il.com>
> > > > >
> > > > > You are overcomplicating things, if you just want to re-enable messages
> > > > > you can just use:
> > > > >
> > > > > -		stfu |= sysctl_drop_caches & 4;
> > > > > +		stfu = sysctl_drop_caches & 4;
> > > > >
> > > > > The bool is there as 4 is intended as a bit flag, you can can figure
> > > > > out what values you want and just append 4 to it to get the expected
> > > > > result.
> > > > >
> > > > >  Luis
> > > >
> > > > Is that what you mean ?
> > > >
> > > > -               stfu |= sysctl_drop_caches & 4;
> > > > +               stfu ^= sysctl_drop_caches & 4;
> > > >
> > > > 'echo 4 > /sys/kernel/vm/drop_caches' can disable or open messages,
> > > > This is what I originally thought, but there is uncertainty that when different operators execute the command,
> > > > It is not possible to determine whether this time is enabled or turned on unless you operate it twice.
> > >
> > > So can you use ^= or not?
> > 
> > No,  ^= does not work, see a boolean truth table.

I don't quite agree with you, you change this, 
echo {1,2,3} will have the meaning of enable message

The initial logic:
echo 1: free pagecache
echo 2: free slab
echo 3: free pagecache and slab
echo 4: disable message

If you change it to something like this:
stfu = sysctl_drop_caches & 4;
echo 1: free pagecache  and enable message
echo 2: free slab       and enable message
echo 3: free pagecache  and enable message
echo 4: disable message

echo 4 becomes meaningless, when echo 4 only the next message can be disabled
Unable to continuously disable echo{1,2,3}

echo {1,2,3} always enabled the message
echo {1,2,3} should not have the meaning of enabling messages

My thoughts:
stfu ^= !!(sysctl_drop_caches & 4);
echo 1: free pagecache
echo 2: free slab
echo 3: free pagecache
echo 4: disable message(odd-numbered operation), enable message(even-numbered operation)

{1, 2, 3} & 4 = 0
stfu ^ 0 = stfu
when echo{1, 2, 3} the stfu is not affected

0 ^ 1 = 1 echo 4: disable message(odd-numbered operation)
1 ^ 1 = 0 echo 4: enable message(even-numbered operation)
stfu ^ 1 = !stfu

when echo 4
stfu(0) -> stfu(1) -> stfu(0) -> stfu(1) -> stfu(0) -> ...
 
> >
> > > And what does operate it twice mean?

echo 4 can:
stfu = 1 # turn off
stfu = 0 # turn on
stfu = 1 # turn off
stfu = 0 # turn on
...

> >
> > I think the reporter meant an "sysadmin", say two folks admining a system.
> > Since we this as a flag to enable disabling it easily we can just
> > always check for the flag as I suggested:
> >
> > stfu = sysctl_drop_caches & 4
> I sent out a new version of this patch. Its a bit late to push it though
> the next merge window, so it is in sysctl-testing until the next cycle
> 
> Thx again
> 
> Best
> 
> -- 
> 
> Joel Granados

Ruiwu