Date: Thu, 24 Apr 2008 10:03:57 -0700 From: Marcel Moolenaar <xcllnt@mac.com> To: Marius Strobl <marius@alchemy.franken.de> Cc: cvs-src@freebsd.org, src-committers@freebsd.org, cvs-all@freebsd.org, Peter Wemm <peter@wemm.org>, John Baldwin <jhb@freebsd.org> Subject: Re: cvs commit: src/sys/amd64/amd64 machdep.c Message-ID: <D29CB5CD-99AC-4C4F-998B-B9FA62AABBA4@mac.com> In-Reply-To: <20080424162722.GB66545@alchemy.franken.de> References: <200804190725.m3J7Pvie056329@repoman.freebsd.org> <200804230858.31200.jhb@freebsd.org> <e7db6d980804231740p4821e2a9m61f5e9bac954b463@mail.gmail.com> <8764A4AA-DE86-43A6-B161-3159DE7E5AB8@mac.com> <20080424162722.GB66545@alchemy.franken.de>
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 24, 2008, at 9:27 AM, Marius Strobl wrote:
> On Wed, Apr 23, 2008 at 06:26:14PM -0700, Marcel Moolenaar wrote:
>>
>> On Apr 23, 2008, at 5:40 PM, Peter Wemm wrote:
>>
>>> On Wed, Apr 23, 2008 at 5:58 AM, John Baldwin <jhb@freebsd.org>
>>> wrote:
>>>> On Saturday 19 April 2008 03:25:57 am Peter Wemm wrote:
>>>>> peter 2008-04-19 07:25:57 UTC
>>>>>
>>>>> FreeBSD src repository
>>>>>
>>>>> Modified files:
>>>>> sys/amd64/amd64 machdep.c
>>>>> Log:
>>>>> Put in a real isa_irq_pending() stub in order to remove two lines
>>>>> of
>>>>> dmesg noise from sio per unit. sio likes to probe if interrupts
>>>>> are
>>>>> configured correctly by looking at the pending bits of the atpic
>>>>> in order
>>>>> to put a non-fatal warning on the console. I think I'd rather
>>>>> read the
>>>>> pending bits from the apics, but I'm not sure its worth the
>>>>> hassle.
>>>>
>>>> Actually, the x86 interrupt sources have a pending method so this
>>>> can be
>>>> replaced. Could probably easily write something like this:
>>>>
>>>> int
>>>> intr_pending(u_int irq)
>>>> {
>>>> struct intsrc *isrc;
>>>>
>>>> isc = intr_lookup_source(irq);
>>>> if (isrc == NULL)
>>>> panic("bizarre");
>>>> return (isrc->is_pic->pic_pending(isrc));
>>>> }
>>>>
>>>> For intr_machdep.c and use this in sio:
>>>>
>>>> #if defined(__i386__) || defined(__amd64__)
>>>> foo = intr_pending(rman_get_start(irq_resource));
>>>> #else
>>>> foo = isa_irq_pending() & (1 << rman_get_start(irq_resource));
>>>> #endif
>>>>
>>>> or some such. I'd really prefer to kill isa_irq_pending().
>>>
>>> Let's just add intr_pending() to all MD backends that currently
>>> provide isa_irq_pending() for sio's benenfit. Either as a simple
>>> wrapper around the now-static local isa_irq_pending() (ia64), or by
>>> simplifying and converting isa_irq_pending() into intr_pending()
>>> (sparc64).
>>
>> sio() is only for i386 and amd64, isa_irq_pending() should be
>> removed from all MD code, except from i386 and amd64.
>
> How likely is it that another driver starts using
> isa_irq_pending() or a intr_pending()?
I think it's Very unlikely.
There's no portable use of these functions possible. The best
way to check if IRQ settings are correct is by registering an
ISR, raise an interrupt and check if the ISR gets called.
This is because PICs get programmed as part of setting up ISRs.
Thus, interrupts without an ISR have undefined behaviour.
Of course, registering an ISR and raising an interrupt to check
if the settings are correct has the same side-effects as
unchecked interrupt use: storms and/or stray interrupts.
--
Marcel Moolenaar
xcllnt@mac.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?D29CB5CD-99AC-4C4F-998B-B9FA62AABBA4>