[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAHk-=whetxWT58ACKO3aGPZQAb_CCpono6U_siynPuvDs2J6rg@mail.gmail.com>
Date: Tue, 4 Feb 2025 08:49:57 -0800
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Christian Brauner <brauner@...nel.org>
Cc: Mateusz Guzik <mjguzik@...il.com>, Oleg Nesterov <oleg@...hat.com>,
David Howells <dhowells@...hat.com>, "Gautham R. Shenoy" <gautham.shenoy@....com>,
K Prateek Nayak <kprateek.nayak@....com>, Neeraj Upadhyay <Neeraj.Upadhyay@....com>,
Oliver Sang <oliver.sang@...el.com>, Swapnil Sapkal <swapnil.sapkal@....com>,
WangYuli <wangyuli@...ontech.com>, linux-fsdevel@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] pipe: don't update {a,c,m}time for anonymous pipes
On Tue, 4 Feb 2025 at 08:34, Christian Brauner <brauner@...nel.org> wrote:
>
> So really that mnt_get_write_access() should be pointless for
> anonymous pipes. In other words, couldn't this also just be:
I bet you'll find that even just inode_update_times() isn't all that cheap.
Yes, the silly "get writability" thing is pointless for pipes, but the
"get current time" really isn't a no-op either. There's all the crazy
"coarse time vs fine time" crud, there's justv a *lot* of overhead.
That's not really noticeable on a "real" file where you never access
just one byte anyway, and where any write costs are elsewhere. But
pipes can be very cheap, so the silly overheads really show up.
Linus
Powered by blists - more mailing lists