linux-kernel - Re: Re: [PATCH v2 06/11] i2c: nomadik: support short xfer timeouts using waitqueue & hrtimer

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <stofbh7be43it734k5icmjpo7oya2x4maevl2xirnf7spfkrzv@h27rc5riemao>
Date: Mon, 4 Mar 2024 16:09:38 +0100
From: Andi Shyti <andi.shyti@...nel.org>
To: Théo Lebrun <theo.lebrun@...tlin.com>
Cc: Linus Walleij <linus.walleij@...aro.org>, 
	Rob Herring <robh+dt@...nel.org>, Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>, 
	Conor Dooley <conor+dt@...nel.org>, Thomas Bogendoerfer <tsbogend@...ha.franken.de>, 
	linux-arm-kernel@...ts.infradead.org, linux-i2c@...r.kernel.org, devicetree@...r.kernel.org, 
	linux-kernel@...r.kernel.org, linux-mips@...r.kernel.org, 
	Gregory Clement <gregory.clement@...tlin.com>, Vladimir Kondratiev <vladimir.kondratiev@...ileye.com>, 
	Thomas Petazzoni <thomas.petazzoni@...tlin.com>, Tawfik Bayouk <tawfik.bayouk@...ileye.com>, 
	Andi Shyti <andi.shyti@...nel.org>
Subject: Re: Re: [PATCH v2 06/11] i2c: nomadik: support short xfer timeouts
 using waitqueue & hrtimer

Hi Theo,

On Mon, Mar 04, 2024 at 03:32:38PM +0100, Théo Lebrun wrote:
> On Mon Mar 4, 2024 at 2:54 PM CET, Andi Shyti wrote:
> > > +static bool nmk_i2c_wait_xfer_done(struct nmk_i2c_dev *priv)
> > > +{
> > > +	if (priv->timeout_usecs < jiffies_to_usecs(1)) {
> > > +		unsigned long timeout_usecs = priv->timeout_usecs;
> > > +		ktime_t timeout = ktime_set(0, timeout_usecs * NSEC_PER_USEC);
> > > +
> > > +		wait_event_hrtimeout(priv->xfer_wq, priv->xfer_done, timeout);
> > > +	} else {
> > > +		unsigned long timeout = usecs_to_jiffies(priv->timeout_usecs);
> > > +
> > > +		wait_event_timeout(priv->xfer_wq, priv->xfer_done, timeout);
> > > +	}
> > > +
> > > +	return priv->xfer_done;
> >
> > You could eventually write this as
> >
> >   static bool nmk_i2c_wait_xfer_done(struct nmk_i2c_dev *priv)
> >   {
> > 	if (priv->timeout_usecs < jiffies_to_usecs(1)) {
> > 		...
> >
> > 		return !wait_event_hrtimeout(...);
> > 	}
> >
> > 	...
> > 	return wait_event_timeout(...);
> >   }
> >
> > It looks a bit cleaner to me... your choice.
> 
> The full block would become:
> 
> static bool nmk_i2c_wait_xfer_done(struct nmk_i2c_dev *priv)
> {
> 	if (priv->timeout_usecs < jiffies_to_usecs(1)) {
> 		unsigned long timeout_usecs = priv->timeout_usecs;
> 		ktime_t timeout = ktime_set(0, timeout_usecs * NSEC_PER_USEC);
> 
> 		return !wait_event_hrtimeout(priv->xfer_wq, priv->xfer_done,
> 					     timeout);
> 	}
> 
> 	return wait_event_timeout(priv->xfer_wq, priv->xfer_done,
> 				  usecs_to_jiffies(priv->timeout_usecs));
> }
> 
> Three things:
> 
>  - Deindenting the jiffy timeout case means no variable declaration
>    after the if-block. This is fine from my point-of-view.
> 
>  - It means we depend on the half-mess that are return values from
>    wait_event_*timeout() macros. I wanted to avoid that because it
>    looks like an error when you read the above code and see one is
>    negated while the other is not.
> 
>  - Also, I'm not confident in casting either return value to bool; what
>    happens if either macro returns an error? This is a theoretical case
>    that shouldn't happen, but behavior might change at some point or
>    bugs could occur. We know priv->xfer_done will give us the right
>    answer.
> 
> My preference still goes to the original version, but I'm happy we are
> having a discussion about this code block.

sure... it's not a binding comment.

Andi