Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 11 Sep 2010 07:55:36 -0400
From:      John Baldwin <jhb@FreeBSD.org>
To:        perryh@pluto.rain.com
Cc:        freebsd-stable@freebsd.org
Subject:   Re: MSIX failure
Message-ID:  <4C8B6E38.7030303@FreeBSD.org>
In-Reply-To: <4c8b3777.A5l69BLBx5dsBnNl%perryh@pluto.rain.com>
References:  <AANLkTikvd09fw3fUhxsXDhMoWE-kvMoPjXPSTfzYY73T@mail.gmail.com> <20100909150012.GA27370@lordcow.org> <4c89d653.UyUz1Sv9zPckJnig%perryh@pluto.rain.com> <201009100757.49777.jhb@freebsd.org> <4c8b3777.A5l69BLBx5dsBnNl%perryh@pluto.rain.com>

next in thread | previous in thread | raw e-mail | index | archive | help
perryh@pluto.rain.com wrote:
> John Baldwin <jhb@freebsd.org> wrote:
>> On Friday, September 10, 2010 2:55:15 am perryh@pluto.rain.com wrote:
>>> ...
>>> It is arguably a bug to open O_RDWR when only examining things.
>> You have to have RDWR permission to issue the ioctl to read config
>> registers which pciconf does when examining capabilities.
> 
> So much for avoiding a reboot for <bsd@lordcow.org> (or whatever the
> correct address is -- that one bounced), but now it is beginning to
> look as if there may be a POLA violation at a lower level.  Unless
> there are devices out there whose state can be changed by merely
> _reading_ (not writing) their configuration registers, I would not
> expect to need RDWR permission just to read them.

Umm, not all config registers are standardized and devices are free to 
define device-specific registers with device-specific behavior (many 
do).  Given that, it is quite possible for a device to implement a 
register that takes action when read (e.g. resets to 0 or clears status 
bits, etc.).

pciconf -l itself does not require RDWR permissions, only -c which does 
a good bit more work.

-- 
John Baldwin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4C8B6E38.7030303>