On Sun, May 24, 2009 at 23:28, gyang wrote:
> On Sun, 2009-05-24 at 01:31 -0400, Mike Frysinger wrote:
>> On Fri, Nov 7, 2008 at 06:51,  <[email protected]> wrote:
>> > Modified: trunk/arch/blackfin/mm/init.c (5600 => 5601)
>> >
>> > +void __cpuinit reserve_pda(void)
>> > +{
>> > +   printk(KERN_INFO "PDA for CPU%u reserved at %p\n", smp_processor_id(),
>> > +                                   &cpu_pda[smp_processor_id()]);
>> > +}
>> > +
>> >  void __init mem_init(void)
>> >  {
>> >     unsigned int codek = 0, datak = 0, initk = 0;
>> > @@ -141,21 +183,13 @@
>> >
>> >  static int __init sram_init(void)
>> >  {
>> > -   unsigned long tmp;
>> > -
>> >     /* Initialize the blackfin L1 Memory. */
>> >     bfin_sram_init();
>> >
>> > -   /* Allocate this once; never free it.  We assume this gives us a
>> > -      pointer to the start of L1 scratchpad memory; panic if it
>> > -      doesn't.  */
>> > -   tmp = (unsigned long)l1sram_alloc(sizeof(struct l1_scratch_task_info));
>> > -   if (tmp != (unsigned long)L1_SCRATCH_TASK_INFO) {
>> > -           printk(KERN_EMERG "mem_init(): Did not get the right address 
>> > from
>> > l1sram_alloc: %08lx != %08lx\n",
>> > -                   tmp, (unsigned long)L1_SCRATCH_TASK_INFO);
>> > -           panic("No L1, time to give up\n");
>> > -   }
>> > -
>> > +   /* Reserve the PDA space for the boot CPU right after we
>> > +    * initialized the scratch memory allocator.
>> > +    */
>> > +   reserve_pda();
>> >     return 0;
>> >  }
>> >  pure_initcall(sram_init);
>>
>> maybe i'm missing something, but this reserve_pda() change looks
>> wrong.  you changed code that allocated the start of scratchpad for L1
>> scratch task info to a call to reserve_pda() ... except that this new
>> reserve_pda() call merely calls printk() and does nothing else.  so
>> where exactly is this reservation now taking place ?  why does
>
> I must had lost allocating L1_SCRATCH_TASK_INFO from l1 scratchpad.
>
>> reserve_pda() exist at all ?  we might as well drop sram_init()
>
> reserve_pda() now is not used, so I will remove it.
>
>> completely and make bfin_sram_init() the entry point for
>> pure_initcall() while we're at it.
>
> Is it appropriate if just allocating L1_SCRATCH_TASK_INFO in
> bfin_sram_init()?

that should be OK
-mike

_______________________________________________
Linux-kernel-commits mailing list
[email protected]
https://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits

Reply via email to