Messages in this thread | | | Date | Mon, 9 Sep 2013 10:30:28 +0200 | From | Peter Zijlstra <> | Subject | Re: [PATCH v7 1/4] spinlock: A new lockref structure for lockless update of refcount |
| |
On Sun, Sep 01, 2013 at 01:13:06AM +0100, Al Viro wrote:
> +static noinline_for_stack > +char *dentry_name(char *buf, char *end, const struct dentry *d, struct printf_spec spec, > + int depth) > +{ > + int i, n = 0; > + const char *s; > + char *p = buf; > + const struct dentry *array[4]; > + char c; > + > + if (depth < 0) { > + depth = 1; > + WARN_ON(1); > + } > + if (depth > 4) { > + depth = 4; > + WARN_ON(1); > + } > + > + rcu_read_lock(); > + for (i = 0; i < depth; i++) { > + struct dentry *p = ACCESS_ONCE(d->d_parent); > + array[i] = d; > + if (d == p) > + break; > + d = p; > + } > + if (!i) { /* root dentry has a bloody inconvenient name */ > + i++; > + goto do_name; > + } > + if (i == depth) > + goto do_name; > + while (i && n != spec.precision) { > + if (buf < end) > + *buf = '/'; > + buf++; > + n++; > +do_name: > + s = ACCESS_ONCE(array[--i]->d_name.name); > + while (n != spec.precision && (c = *s++) != '\0') { > + if (buf < end) > + *buf = c; > + buf++; > + n++; > + } > + } > + rcu_read_unlock();
Should one or both of those ACCESS_ONCE()s be an rcu_dereference()?
| |