[PATCH 02/14] omap: Map only available sram memory

[Shilimkar, Santosh]

> -----Original Message-----
> From: Grazvydas Ignotas [mailto:notasas at gmail.com]
> Sent: Monday, October 04, 2010 2:34 PM
> To: Shilimkar, Santosh
> Cc: Tony Lindgren; linux-omap at vger.kernel.org; linux-arm-
> kernel at lists.infradead.org; Russell King - ARM Linux
> Subject: Re: [PATCH 02/14] omap: Map only available sram memory
> 
> >> >
> >> > This hangs OMAP3 pandora:
> >> >
> >> > [    0.000000] Linux version
> >> > 2.6.36-rc6-next-20101001-00002-ge76bb53-dirty (notaz at pixelinis) (gcc
> >> > version 4.3.3 (Sourcery G++ Lite 2009q1-20
> >> > [    0.000000] CPU: ARMv7 Processor [411fc082] revision 2 (ARMv7),
> >> cr=10c53c7f
> >> > [    0.000000] CPU: VIPT nonaliasing data cache, VIPT nonaliasing
> >> > instruction cache
> >> > [    0.000000] Machine: Pandora Handheld Console
> >> > [    0.000000] Ignoring unrecognised tag 0x54410008
> >> > [    0.000000] bootconsole [earlycon0] enabled
> >> > [    0.000000] Reserving 6422528 bytes SDRAM for VRAM
> >> > [    0.000000] Memory policy: ECC disabled, Data cache writeback
> >> > [    0.000000] OMAP3430/3530 ES2.1 (l2cache iva sgx neon isp )
> >> > [    0.000000] SRAM: Mapped pa 0x40200000 to va 0xfe400000 size:
> 0x10000
> >> > (stuck here)
> >> >
> >> > reverting this fixes the problem.
> >>
> >> Hmm, boots fine here with overo. Any idea what in this patch breaks
> >> pandora?
> >>
> > The change in this patch is not board dependent really. Have tested this
> > on 3430SDP. Pandora is OMAP3 based, right ?
> 
> OMAP3530 ES2.1, also tried on friend's beagleboard b5 (also ES2.1) and
> it has the same problem. Maybe it's because of older Cortex A8 used
> there, or I'm missing some errata workaround in defconfig.
> 
> BTW, hacking the size to 1M on top of your patch fixes the problem too:
> 
>         base = omap_sram_start;
>         base = ROUND_DOWN(base, PAGE_SIZE);
>         omap_sram_io_desc[0].pfn = __phys_to_pfn(base);
> -       omap_sram_io_desc[0].length = ROUND_DOWN(omap_sram_size,
> PAGE_SIZE);
> +       omap_sram_io_desc[0].length = 0x100000;
This is the exact reason this patch is created :)
So that you map only available memory instead of 1 MB
>         iotable_init(omap_sram_io_desc, ARRAY_SIZE(omap_sram_io_desc));