From mboxrd@z Thu Jan 1 00:00:00 1970 From: Darius Augulis Date: Mon, 22 Nov 2010 20:30:15 +0200 Subject: [U-Boot] [PATCH] ARM: S3C64XX: fix timer broken by relocation In-Reply-To: <20101113181301.4ABCA150ADD@gemini.denx.de> References: <20101101204832.3822.79056.stgit@darius-desktop> <4CDD84E0.4080609@gmail.com> <4CDD9495.9010700@free.fr> <4CDEC32B.7050309@gmail.com> <20101113181301.4ABCA150ADD@gemini.denx.de> Message-ID: <4CEAB6B7.5060406@gmail.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Hi Wolfgang, On 11/13/2010 08:13 PM, Wolfgang Denk wrote: > Dear Darius Augulis, > > In message<4CDEC32B.7050309@gmail.com> you wrote: >> >>> FWIW, there already is a solution based on statics and post-relocation >>> initialization for orion5x. The principle there is that the timer is not >>> used before calling board_init_r, so we don't need initializing >>> timestamp before relocation. > > STOP!! I don't think we want this. > >> in this case it seems like timer must be initialised after relocation. >> But I don't know if it could be correct for all ARM architectures. Maybe >> some of them use timer before relocation. > > Indeed. Drivers and other code may want to implement timeouts and the > like, and need at least basic timer services like udelay() and such. > >> Global data is good place to store important static variables because >> they are valid before and after relocation and it could be common for >> all architectures. Since we have automatic size calculation of global >> data structure there should not be a problem to add several additional >> bytes specific to every CPU. > > We should keep the gd as small as possible, but adinng one or two > integers here is indeed probably the best approach. could you please merge it or point me what is missing? thanks, Darius. > > > Best regards, > > Wolfgang Denk >