From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 4168 invoked by alias); 18 Mar 2003 19:13:20 -0000 Mailing-List: contact ecos-maintainers-help@sources.redhat.com; run by ezmlm Precedence: bulk List-Subscribe: List-Post: List-Help: , Sender: ecos-maintainers-owner@sources.redhat.com Received: (qmail 4063 invoked from network); 18 Mar 2003 19:13:19 -0000 Subject: Re: Patches for the eCos 2.0 branch From: Gary Thomas To: Nick Garnett Cc: John Dallaway , eCos Maintainers In-Reply-To: References: <200303180848.24682.jld@ecoscentric.com> <1047995318.7459.2925.camel@hermes.chez-thomas.org> <1048011920.9579.3582.camel@hermes.chez-thomas.org> Content-Type: text/plain Content-Transfer-Encoding: 7bit X-Mailer: Ximian Evolution 1.0.3 (1.0.3-4) Date: Tue, 18 Mar 2003 19:13:00 -0000 Message-Id: <1048014798.7462.3684.camel@hermes.chez-thomas.org> Mime-Version: 1.0 X-SW-Source: 2003-03/txt/msg00019.txt.bz2 On Tue, 2003-03-18 at 12:05, Nick Garnett wrote: > Gary Thomas writes: > > > > > Actually, I tried to make it such that a dependency does not exist > > (note the most recent version which this patch reflects does try > > and work properly even without an updated RedBoot). I may have > > missed the mark, but I did try. > > > > It certainly caused me a bit of head scratching when I tried running > on an Adder board with an old RedBoot yesterday. Fortunately I > remembered your change and updating the Adder's RedBoot fixed it. > > The problem appeared to be that *(CYGHWR_HAL_VSR_TABLE + 0x1F0) > contained 0xFFFF, resulting in _mpc8xx_allocBd() starting again from > QUICC_BD_BASE and overwriting RedBoot's stuff for the serial device. > > The effect I saw was that halfway through initializing the rx buffer > descriptors, the serial device went haywire. > > I'm not sure how it is possible to make this code backward compatible, > since without the value in (CYGHWR_HAL_VSR_TABLE + 0x1F0), eCos has no > way of knowing where in the DPRAM it can make allocations. The drivers are supposed to be flexible in this case. If the eCos program changes the serial port setup (allocates different buffer descriptors), the driver in RedBoot should be OK with it since it uses only the data from the CPM after initialization. I could see a problem though - RedBoot could choose some region for the buffers for the serial port and then, because of the lack of communication, eCos might choose those same locations for the network. If you were running an eCos program with networking in it, I could see this happening. Alas, it seems that updating RedBoot is a good idea anyway :-) -- ------------------------------------------------------------ Gary Thomas | MLB Associates | Consulting for the +1 (970) 229-1963 | Embedded world http://www.mlbassoc.com/ | email: | gpg: http://www.chez-thomas.org/gary/gpg_key.asc ------------------------------------------------------------