[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAGudoHHD3ZpYhx+wLSCeQV_bEkBysLahAO-ovXhbPDc4vWeO5A@mail.gmail.com>
Date: Tue, 24 Jan 2023 07:48:51 +0100
From: Mateusz Guzik <mjguzik@...il.com>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Al Viro <viro@...iv.linux.org.uk>, Uros Bizjak <ubizjak@...il.com>,
David Laight <David.Laight@...lab.com>,
Andrew Morton <akpm@...ux-foundation.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] lib/genalloc: use try_cmpxchg in {set,clear}_bits_ll
On 1/24/23, Linus Torvalds <torvalds@...ux-foundation.org> wrote:
> On Mon, Jan 23, 2023 at 4:11 PM Mateusz Guzik <mjguzik@...il.com> wrote:
>>
>> On another point how to end up dealing with lockref less, I have to
>> note glibc switched fstat(fd, buf) to use newfstatat(fd, "", buf,
>> AT_EMPTY_PATH) internally.
>
> Yeah, that's just plain stupid.
>
> I don't know what the thinking (or rather, lack-thereof) was inside
> glibc for that one.
>
To my reading the fstat system call is now legacy so they switched to
something else. There is a lot to say about the stat situation (going
past the fd vs lookup issue), which I'm going to do on fsdevel (perhaps
next week).
All flaming aside, I think Uros is still waiting for an answer what
should be done in cmpxchg loops and someone else will have to provide
it.
I stated my case for the best course of action being to do nothing
(for x86-64 anyway), but I'm not an authority figure.
--
Mateusz Guzik <mjguzik gmail.com>
Powered by blists - more mailing lists