Re: [PATCH 1/3] PCI: vpd handle longer delays in access

[Matthew Wilcox <matthew@wil.cx>]

On Thu, Sep 04, 2008 at 01:56:37PM -0700, Stephen Hemminger wrote:
> -		udelay(10);
> +		if (signal_pending(current))
> +			return -EINTR;

If you're going to use _killable instead of _interruptible, this needs
to be fatal_signal_pending().  Otherwise the one who owns the lock can
be interrupted by _any_ signal while those waiting for the lock can only
be interrupted by fatal signals.  Which seems daft to me.

> -	spin_lock_irq(&vpd->lock);
> +	ret = mutex_lock_killable(&vpd->lock);
> +	if (ret)
> +		return ret;

What's wrong with the shorter:

	if (mutex_lock_killable(&vpd->lock))
		return -EINTR;
?

The actual error is irrelevant here since userspace will never consume it.

(I agree with Peter about use of yield())

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."