netdev
[Top] [All Lists]

Re: R[PATCH 2.6.10-rc1 8/15] wireless/orinoco: Refactor spinlocks so we

To: Dan Williams <dcbw@xxxxxxxxxx>
Subject: Re: R[PATCH 2.6.10-rc1 8/15] wireless/orinoco: Refactor spinlocks so we don't necessarily have to disable interrupts
From: Jeff Garzik <jgarzik@xxxxxxxxx>
Date: Tue, 26 Oct 2004 15:44:21 -0400
Cc: netdev@xxxxxxxxxxx, jgarzik@xxxxxxxxxx, hermes@xxxxxxxxxxxxxxxxxxxxx
In-reply-to: <1098817472.3663.66.camel@dcbw.boston.redhat.com>
References: <1098814320.3663.24.camel@dcbw.boston.redhat.com> <1098817472.3663.66.camel@dcbw.boston.redhat.com>
Sender: netdev-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.3) Gecko/20040922
Dan Williams wrote:
Update in-kernel orinoco wireless drivers to upstream CVS.
None of this is original code by Dan Williams, simply a
broken down patch set split-out from upstream orinoco CVS.

o Refactor spinlocks so we don't necessarily have to disable interrupts

Signed-off-by: Dan Williams <dcbw@xxxxxxxxxx>

--- a/drivers/net/wireless/orinoco.h.8-orinoco-spinlock 2004-10-26 10:44:41.445687264 -0400
+++ b/drivers/net/wireless/orinoco.h 2004-10-26 10:45:39.296892544 -0400
@@ -71,6 +71,8 @@
u16 channel_mask;
int broken_disableport;
+ unsigned int irq_no_disable:1;
+
/* Configuration paramaters */
u32 iw_mode;
int prefer_port3;
@@ -129,11 +131,17 @@
extern inline int orinoco_lock(struct orinoco_private *priv,
unsigned long *flags)
{
- spin_lock_irqsave(&priv->lock, *flags);
+ if (priv->irq_no_disable)
+ spin_lock_bh(&priv->lock);
+ else
+ spin_lock_irqsave(&priv->lock, *flags);
if (priv->hw_unavailable) {
- printk(KERN_DEBUG "orinoco_lock() called with hw_unavailable (dev=%p)\n",
+ DEBUG(1, "orinoco_lock() called with hw_unavailable (dev=%p)\n",
priv->ndev);
- spin_unlock_irqrestore(&priv->lock, *flags);
+ if (priv->irq_no_disable)
+ spin_unlock_bh(&priv->lock);
+ else
+ spin_unlock_irqrestore(&priv->lock, *flags);
return -EBUSY;

This entire area has problems.

Orinoco doesn't need to invent its own locking primitives, nor does it need to be inventing functions that take *flags as an argument.

        Jeff




<Prev in Thread] Current Thread [Next in Thread>