[U-Boot] u-boot on beagleboard finds incorrect memory size.

[Robert Nelson]

On Fri, Jan 27, 2012 at 10:25 AM, Tom Rini <tom.rini at gmail.com> wrote:
> On Fri, Jan 27, 2012 at 9:23 AM, Robert Nelson <robertcnelson at gmail.com> wrote:
>> On Fri, Jan 27, 2012 at 9:58 AM, Tom Rini <tom.rini at gmail.com> wrote:
>>> On Fri, Jan 27, 2012 at 8:50 AM, wilsonjonathan
>>> <piercing_male at hotmail.com> wrote:
>>>> I think the area of code that needs investigating is in
>>>> arch/arm/cpu/armv7/omap3/sdrc.c
>>>>
>>>> line 82 onwards :-
>>>>
>>>>
>>>>> u32 get_sdr_cs_size(u32 cs)
>>>>> {
>>>>>       u32 size;
>>>>>
>>>>>       /* get ram size field */
>>>>>       size = readl(&sdrc_base->cs[cs].mcfg) >> 8;
>>>>>       size &= 0x3FF;          /* remove unwanted bits */
>>>>>       size <<= 21;            /* multiply by 2 MiB to find size in MB */
>>>>>       return size;
>>>>> }
>>>>>
>>>>
>>>> Which I think is set in arc/arm/include/asm/arch-omap3/cpu.h but I don't
>>>> understand how the -> stuff works.
>>>
>>> This is all correct.  However, what's going wrong is we're programming
>>> the size of memory found incorrectly and we don't yet have a safe
>>> method to probe how much memory is really there (aside:
>>> get_ram_size(...) goes off into the weeds on omap3 and it's on my TODO
>>> list, and rising fast, to get my flyswatter out, hooked up and see
>>> what's really happening).  So what I really need to know is which of
>>> the cases in beagle.c this board is falling into (I suspect it's the
>>> top half of that else, but it's 2x128MB not 1x256).  Thanks again!
>>
>>
>> Okay, little bit more information is shown with printf's..
>>
>> U-Boot SPL 2011.12-00001-g35bbe6c-dirty (Jan 27 2012 - 10:15:33)
>> Texas Instruments Revision detection unimplemented
>> case REVISION_C4:
>> actual pop_id: 0xBA
>> actual pop_mfr: 0x2C
>> case REVISION_C4: Neither option was selected.
>> OMAP SD/MMC: 0
>> reading u-boot.img
>> reading u-boot.img
>>
>> using:
>>
>> diff --git a/board/ti/beagle/beagle.c b/board/ti/beagle/beagle.c
>> index 9aa05d4..4c7bf94 100644
>> --- a/board/ti/beagle/beagle.c
>> +++ b/board/ti/beagle/beagle.c
>> @@ -158,7 +158,11 @@ void get_board_mem_timings(u32 *mcfg, u32 *ctrla,
>> u32 *ctrlb, u32 *rfr_ctrl,
>>        *mr = MICRON_V_MR_165;
>>        switch (get_board_revision()) {
>>        case REVISION_C4:
>> +       printf("case REVISION_C4:\n");
>> +       printf("actual pop_id: 0x%02X \n", pop_id);
>> +       printf("actual pop_mfr: 0x%02X \n", pop_mfr);
>>                if (pop_mfr == NAND_MFR_STMICRO && pop_id == 0xba) {
>> +                       printf("pop_id == 0xba (C4)\n");
>>                        /* Beagleboard Rev C4, 256MB DDR */
>>                        *mcfg = NUMONYX_V_MCFG_165(256 << 20);
>>                        *ctrla = NUMONYX_V_ACTIMA_165;
>> @@ -166,6 +170,7 @@ void get_board_mem_timings(u32 *mcfg, u32 *ctrla,
>> u32 *ctrlb, u32 *rfr_ctrl,
>>                        *rfr_ctrl = SDP_3430_SDRC_RFR_CTRL_165MHz;
>>                        break;
>>                } else if (pop_mfr == NAND_MFR_MICRON && pop_id == 0xbc) {
>> +                       printf("pop_id == 0xbc (c5)\n");
>>                        /* Beagleboard Rev C5, 256MB DDR */
>>                        *mcfg = MICRON_V_MCFG_200(256 << 20);
>>                        *ctrla = MICRON_V_ACTIMA_200;
>> @@ -173,6 +178,7 @@ void get_board_mem_timings(u32 *mcfg, u32 *ctrla,
>> u32 *ctrlb, u32 *rfr_ctrl,
>>                        *rfr_ctrl = SDP_3430_SDRC_RFR_CTRL_200MHz;
>>                        break;
>>                }
>> +       printf("case REVISION_C4: Neither option was selected.\n");
>>        case REVISION_XM_A:
>>        case REVISION_XM_B:
>>        case REVISION_XM_C:
>>
>> On my c4, it looks like neither option is selected for it..
>>
>> ./include/linux/mtd/nand.h:#define NAND_MFR_STMICRO     0x20
>> ./include/linux/mtd/nand.h:#define NAND_MFR_MICRON              0x2c
>>
>> looks like the RAM/NAND options might have been reversed for C4 and C5
>> in the IF statement.. but i dont' have a C5 to test..
>
> I've got a C5 which has been OK.  I've got to step out for a few but
> I'll have a patch later today for this I hope, if you don't beat me to
> it :)

Ah, now i see the problem..  There actually isn't support in u-boot
for the regular C4.. ;)

When the memory options where transfered from x-loader. The special
'white label' 512Mb DDR C4 from Special Computing became the "regular
C4"..

see:
http://gitorious.org/x-loader/x-loader/blobs/master/board/omap3530beagle/omap3530beagle.c#line576

and

http://gitorious.org/x-loader/x-loader/commit/1dac1a7c188e79edabe211ccaadf58cd83b278dc


So first we need to document the "512MB DDR" something like...

diff --git a/board/ti/beagle/beagle.c b/board/ti/beagle/beagle.c
index 6a457cb..cf55c79 100644
--- a/board/ti/beagle/beagle.c
+++ b/board/ti/beagle/beagle.c
@@ -159,7 +159,7 @@ void get_board_mem_timings(u32 *mcfg, u32 *ctrla,
u32 *ctrlb, u32 *rfr_ctrl,
        switch (get_board_revision()) {
        case REVISION_C4:
                if (pop_mfr == NAND_MFR_STMICRO && pop_id == 0xba) {
-                       /* 512MB DDR */
+                       /* Beagle Rev C4 from Special Computing, 512MB DDR */
                        *mcfg = NUMONYX_V_MCFG_165(512 << 20);
                        *ctrla = NUMONYX_V_ACTIMA_165;
                        *ctrlb = NUMONYX_V_ACTIMB_165;


and then readd c4 memory timings..

Regards,

-- 
Robert Nelson
http://www.rcn-ee.com/