From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sudeep KarkadaNagesha Subject: Re: [RFC PATCH 02/11] ARM: DT/kernel: define ARM specific arch_of_get_cpu_node Date: Tue, 16 Jul 2013 10:03:21 +0100 Message-ID: <51E50C59.50900@arm.com> References: <1373883732-26303-1-git-send-email-Sudeep.KarkadaNagesha@arm.com> <1373883732-26303-3-git-send-email-Sudeep.KarkadaNagesha@arm.com> <51E44943.7000600@gmail.com> Mime-Version: 1.0 Content-Transfer-Encoding: 8BIT Return-path: In-Reply-To: <51E44943.7000600@gmail.com> Sender: linux-pm-owner@vger.kernel.org List-ID: Content-Type: text/plain; charset="us-ascii" To: Rob Herring Cc: Sudeep KarkadaNagesha , "linux-pm@vger.kernel.org" , "linux-kernel@vger.kernel.org" , "linux-arm-kernel@lists.infradead.org" , Lorenzo Pieralisi , "linux@arm.linux.org.uk" , "arnd@arndb.de" , "viresh.kumar@linaro.org" , "rob.herring@calxeda.com" , "rjw@sisk.pl" , "grant.likely@linaro.org" , "cpufreq@vger.kernel.org" , "kernel@pengutronix.de" , "gregkh@linuxfoundation.org" , "olof@lixom.net" , "gregory.clement@free-electrons.com" , "shawn.guo@linaro.org" <> On 15/07/13 20:10, Rob Herring wrote: > On 07/15/2013 05:22 AM, Sudeep.KarkadaNagesha@arm.com wrote: >> From: Sudeep KarkadaNagesha >> >> CPU subsystem now provides architecture specific hook to retrieve the >> of_node. Most of the cpu DT node parsing and initialisation is contained >> in devtree.c. It's better to contain all CPU device node parsing there. >> >> arch_of_get_cpu_node is mainly used to assign cpu->of_node when CPUs get >> registered. This patch overrides the defination of the same. It can also >> act as the helper function in pre-SMP/early initialisation stages to >> retrieve CPU device node pointers in logical ordering. >> >> This mainly helps to avoid replication of the code doing CPU node parsing >> and physical(MPIDR) to logical mapping. >> >> Signed-off-by: Sudeep KarkadaNagesha > > [snip] > >> +struct device_node * __init arch_of_get_cpu_node(int cpu) >> +{ >> + struct device_node *cpun, *cpus; >> + const u32 *cell; >> + u64 hwid; >> + int ac; >> + >> + cpus = of_find_node_by_path("/cpus"); >> + if (WARN(!cpus, "Missing cpus node, bailing out\n")) >> + return NULL; >> + >> + if (WARN_ON(of_property_read_u32(cpus, "#address-cells", &ac))) >> + ac = of_n_addr_cells(cpus); >> + >> + for_each_child_of_node(cpus, cpun) { >> + if (of_node_cmp(cpun->type, "cpu")) >> + continue; >> + cell = of_get_property(cpun, "reg", NULL); >> + if (WARN(!cell, "%s: missing reg property\n", cpun->full_name)) >> + continue; >> + >> + hwid = of_read_number(cell, ac); >> + if ((hwid & MPIDR_HWID_BITMASK) == cpu_logical_map(cpu)) > > Most of this function is not ARM specific, so it would be nice if we > could shrink the arch specific part down to just this match. A default > match of reg == logical cpu number might be useful. > I completely agree, in fact that was my initial idea too. But when I had a look at powerpc implementation of "of_get_cpu_node" in arch/powerpc/kernel/prom.c, it looked like PPC is using some compatibles(e.g. ibm,ppc-interrupt-server#s) which are not specified in ePAPR. I am not sure is that's allowed or not, if allowed then we can't have generic of_get_cpu_node with just arch specific hwid matching function. Let me know how would you prefer me to proceed on this. Regards, Sudeep