Re: [ANNOUNCE] desc.c -- dump the i386 descriptor tables

From: William Lee Irwin III <wli@holomorphy.com>
To: Chuck Ebbert <76306.1226@compuserve.com>
Cc: linux-kernel <linux-kernel@vger.kernel.org>
Subject: Re: [ANNOUNCE] desc.c -- dump the i386 descriptor tables
Date: Thu, 24 Apr 2003 02:12:50 -0700	[thread overview]
Message-ID: <20030424091250.GO8978@holomorphy.com> (raw)
In-Reply-To: <200304240509_MC3-1-35CF-A2DD@compuserve.com>

wli wrote:
>> Spiffy; this should help debug various things.

On Thu, Apr 24, 2003 at 05:06:58AM -0400, Chuck Ebbert wrote:
>   I forgot to mention: try comparing 2.2, 2.4 and 2.5.
>   Also, this is what I've been using to look at interrupt entry
> point alignment.  On 2.4 for sure, and probably on 2.5 you have
> a 1-in-8 chance of getting a pathologically badly aligned timer
> handler on 32-byte cacheline machines every time you compile
> (IRQ 0 entry address & 0x1f == 0x1c.)  OTOH the pagefault handler
> has come up 8-byte aligned every time but I didn't look at the
> source to see if it's coded that way.

I had more in mind using it to verify the correctness of per-node IDT
stuff than checking optimality in general, but it does look very handy.

-- wli