linux-kernel - Re: [PATCH v2 1/3] writeback: Let trace_balance_dirty

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20250304103957.08c79da0@gandalf.local.home>
Date: Tue, 4 Mar 2025 10:39:57 -0500
From: Steven Rostedt <rostedt@...dmis.org>
To: Tang Yizhou <yizhou.tang@...pee.com>
Cc: tj@...nel.org, jack@...e.cz, brauner@...nel.org, willy@...radead.org,
 akpm@...ux-foundation.org, mhiramat@...nel.org, ast@...nel.org,
 linux-mm@...ck.org, linux-fsdevel@...r.kernel.org,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 1/3] writeback: Let trace_balance_dirty_pages() take
 struct dtc as parameter

On Tue,  4 Mar 2025 19:03:16 +0800
Tang Yizhou <yizhou.tang@...pee.com> wrote:

> @@ -664,16 +660,16 @@ TRACE_EVENT(balance_dirty_pages,
>  	),
>  
>  	TP_fast_assign(
> -		unsigned long freerun = (thresh + bg_thresh) / 2;
> +		unsigned long freerun = (dtc->thresh + dtc->bg_thresh) / 2;
>  		strscpy_pad(__entry->bdi, bdi_dev_name(wb->bdi), 32);
>  
>  		__entry->limit		= global_wb_domain.dirty_limit;
>  		__entry->setpoint	= (global_wb_domain.dirty_limit +
>  						freerun) / 2;
> -		__entry->dirty		= dirty;
> +		__entry->dirty		= dtc->dirty;
>  		__entry->bdi_setpoint	= __entry->setpoint *
> -						bdi_thresh / (thresh + 1);
> -		__entry->bdi_dirty	= bdi_dirty;
> +						dtc->wb_thresh / (dtc->thresh + 1);
> +		__entry->bdi_dirty	= dtc->wb_dirty;
>  		__entry->dirty_ratelimit = KBps(dirty_ratelimit);
>  		__entry->task_ratelimit	= KBps(task_ratelimit);
>  		__entry->dirtied	= dirtied;

I don't know how much of a fast path these tracepoints are in, but instead
of doing the divisions above, why not just save the values in the ring
buffer, and do the divisions in the TP_printk() section, which is done when
the user reads it and not when the code is executing?

-- Steve