[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20251005153008.GF2441659@ZenIV>
Date: Sun, 5 Oct 2025 16:30:08 +0100
From: Al Viro <viro@...iv.linux.org.uk>
To: Oleg Nesterov <oleg@...hat.com>
Cc: Boqun Feng <boqun.feng@...il.com>, David Howells <dhowells@...hat.com>,
Ingo Molnar <mingo@...hat.com>, Li RongQing <lirongqing@...du.com>,
Linus Torvalds <torvalds@...ux-foundation.org>,
Peter Zijlstra <peterz@...radead.org>,
Waiman Long <longman@...hat.com>, Will Deacon <will@...nel.org>,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 0/5] seqlock: introduce SEQLOCK_READ_SECTION()
On Sun, Oct 05, 2025 at 04:49:29PM +0200, Oleg Nesterov wrote:
> OK, let me name it SEQLOCK_READ_SECTION().
>
> Al, could you look at 5/5? Please nack it if you think it makes no sense
> or wrong. I abused __dentry_path() to add another example. To simplify the
> review, this is how it looks after the patch:
>
> static char *__dentry_path(const struct dentry *d, struct prepend_buffer *p)
> {
> const struct dentry *dentry;
> struct prepend_buffer b;
>
> rcu_read_lock();
> __SEQLOCK_READ_SECTION(&rename_lock, lockless, seq, NULL) {
> dentry = d;
> b = *p;
> while (!IS_ROOT(dentry)) {
> const struct dentry *parent = dentry->d_parent;
>
> prefetch(parent);
> if (!prepend_name(&b, &dentry->d_name))
> break;
> dentry = parent;
> }
> if (lockless)
> rcu_read_unlock();
<cringe>
> }
>
> if (b.len == p->len)
> prepend_char(&b, '/');
> return extract_string(&b);
> }
>
>
> TODO: add another trivial helper
>
> static inline int need_seqretry_or_lock(seqlock_t *lock, int *seq)
> {
> int ret = !(*seq & 1) && read_seqretry(lock, *seq);
>
> if (ret)
> *seq = 1; /* make this counter odd */
>
> return ret;
> }
>
> which can be used when the read section is more complex. Say, d_walk().
prepend_path() would be interesting to look at...
Powered by blists - more mailing lists