Re: [patch] epoll use a single inode ...

[Linus Torvalds]

On Wed, 7 Mar 2007, Eric Dumazet wrote:
> 
> I was thinking about being able to cache the name into the dentry, do you 
> think it's worth the pain ? (its not SMP safe for example...)

Actually, it *can* be SMP-safe, if you do it right. Something like

	len = dentry->d_name.len;

	if (!len) {
		len = snprintf(dentry->d_iname,
			DNAME_INLINE_LEN_MIN,
			"pipe:[%lu]",
			dentry->d_inode->i_ino);
		smp_wmb();
		dentry->d_name.len = len;
	}
	if (len >= buflen)
		len = buflen-1;
	memcpy(buffer, dentry->d_iname, len);
	buffer[len] = 0;
	return buffer;

should work, although it depends on the fact that our snprintf() 
implementation should be "stable" (ie if snprintf() modifies the buffer 
temporarily as it goes along, that would break, but I think our 
vsnprintf is good in that respect).

So you could have two different CPU's doing the snprintf() on the same 
buffer at the same time (and assigning the length at the same time), but 
since they'll write the same thing, you don't really care.

It's a bit subtle, though. And probably not really worth it.

		Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/