Dear Stefan Roese,
In message 200905111531.11474.sr@denx.de you wrote:
On Monday 11 May 2009 15:20:10 Wolfgang Denk wrote:
And what happens when the TLB is not correctly set? Will we get a useful error message then?
No. SDRAM needs to work correctly so that this RAM-booting version can function. And for this the MMU needs to be configured and the SDRAM controller setup correctly. If this is not the case and SDRAM is not working from the debugger then we have no chance to load this image without problems (verify will most likely fail) or run it.
Hm... if the image cannot run anyway without initialized SDRAM TLB, then why do you scan for it? My understanding is that the result wouldbe the same even if you omitted this test: without initialized SDRAM TLB the system would just hang / crash without error message. So this test is dead code.
Or am I missing something?
I'm "scanning" for the SDRAM TLB, because it can be on any TLB location. In my first implementation I assumed that the SDRAM TLB was TLB #0 (configured via BDI2000). This worked just fine. But then I loaded this RAM-booting image via U-Boot and started it here (via go ...). And it crashed since U-Boot uses a different TLB #. So now I'm scanning/detecting the SDRAM TLB to be independent from the originator of the TLB.
Ah, I see. Could you please put this information in the commit message, too?
Best regards,
Wolfgang Denk