Christoph Hellwig wrote:
The cond_resched_lock here is not safe here, because the pointer you
are going to dereference in list_for_each_entry might not be valid
anymore. This should look more like:
void d_drop_negative_children(struct dentry *parent)
{
struct dentry *dentry;
again:
spin_lock(&dcache_lock);
list_for_each_entry(dentry, &parent->d_subdirs, d_u.d_child) {
if !(dentry->d_inode)
continue;
spin_lock(&dentry->d_lock);
__d_drop(dentry);
spin_unlock(&dentry->d_lock);
if (need_resched()) {
spin_unlock(&dcache_lock);
cond_resched();
goto again;
}
}
spin_unlock(&dcache_lock);
}
Yes, we have been bitten by the same issue.
Instead of need_resched(), it may be better if you do:
if (cond_resched_lock(&dcache_lock))
goto again;
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
