Re: [U-Boot] [PATCH 8/8] powerpc/85xx: Add memory test feature for mpc85xx.
Wolfgang,
As Timur pointed out, the post framework doesn't work for us. After U-boot relocate itself to RAM, we have only 2GB memory to test. The best place is before relocation. Many other boards do that. Following your idea of reusing code, I can only reuse the test pattern generator. I am open to suggestions.
York Sun ----- Original Message ----- From:"Wolfgang Denk" wd@denx.de To:"York Sun" yorksun@freescale.com Cc:"Timur Tabi" timur.tabi@gmail.com, "Kumar Gala" galak@kernel.crashing.org, "u-boot@lists.denx.de" u-boot@lists.denx.de Sent:7/28/2010 4:50 PM Subject:Re: [U-Boot] [PATCH 8/8] powerpc/85xx: Add memory test feature for mpc85xx.
Dear York Sun,
In message 1280351179.8571.63.camel@oslab-l1 you wrote:
We already have too many different implementations of a memory test in U-Boot, and I will not accept adding yet another one.
I can reuse your testing code but have to move the desired code out of memory.c file to avoid the need for CONFIG_POST and CONFIG_SYS_POST_MEMORY. I also add a progress indicator. My testing
NAK, and NAK.
Please integrate your code into the existing POST framework instead, as a number of other boards already did.
A progress indicator may be a nice little toy, but how does it integrate into the POST framework?
target is 2GB at a time, up to physically memory size which is easily over 8GB. Without progress indicator, it feels hung when it is actually running.
Yes, memory testing takes time. In the context of a power-on self test (and this is what you are doing, right?) we should take care to fit it into the existing framework, though.
Please take a look at the patch below.
Instead of integrating your needs into an existing framework you invent yet another one. I don't want to have this, sorry.
Best regards,
Wolfgang Denk
Dear sun york-R58495,
In message 516c01cb2eac$1d67cdf1$0c23400a@fsl.freescale.net you wrote:
As Timur pointed out, the post framework doesn't work for us. After
And I explained that if it doesn't work as is please try to adapt / extend it to match your needs instead of reinventing the wheel.
U-boot relocate itself to RAM, we have only 2GB memory to test. The best place is before relocation. Many other boards do that. Following your
I don't understand what you are saying. You sound as if you assumed that the POST code would run after relocation only? This is NOT the case. THere are tests that run before relocation (incluing the RAM test), and others, that run after relocation.
idea of reusing code, I can only reuse the test pattern generator. I am open to suggestions.
I suggest to check the code again, or to explain in a little more detail why you think you can neither use nor extend the existing framework.
Best regards,
Wolfgang Denk
Wolfgang,
Please help me to understand the post_word_write() and post_word_load(). I see some implementations which don't use NVRAM. Is it OK to use gd? Does the post_word_load() require the data survive reset? How is the data validated? I don't see any code to validate the value.
Regards,
York
Dear York Sun,
In message 1280524829.25910.58.camel@oslab-l1 you wrote:
Please help me to understand the post_word_write() and post_word_load(). I see some implementations which don't use NVRAM. Is it OK to use gd?
No.
Does the post_word_load() require the data survive reset? How is the
Yes.
data validated? I don't see any code to validate the value.
It checks against BOOTMODE_MAGIC.
Best regards,
Wolfgang Denk
Dear Wolfgang,
Thanks for the quick respond.
On Fri, 2010-07-30 at 23:29 +0200, Wolfgang Denk wrote:
Dear York Sun,
In message 1280524829.25910.58.camel@oslab-l1 you wrote:
Please help me to understand the post_word_write() and post_word_load(). I see some implementations which don't use NVRAM. Is it OK to use gd?
No.
Does the post_word_load() require the data survive reset? How is the
Yes.
data validated? I don't see any code to validate the value.
It checks against BOOTMODE_MAGIC.
What's if I don't want to use flash and don't have NVRAM? Can I use gd but let it fail the validation in case of data corruption?
York
Dear York Sun,
In message 1280525648.25910.61.camel@oslab-l1 you wrote:
data validated? I don't see any code to validate the value.
It checks against BOOTMODE_MAGIC.
What's if I don't want to use flash and don't have NVRAM? Can I use gd but let it fail the validation in case of data corruption?
I don't understand the problem. All we need is some register or memory location that is guaranteed to keep it's value during a hard reset of the processor. This can be NVRAM, or On-Chip-Memory, or some register. I am pretty sure you know this CPU much better than me, and I am optimistic that you can find such a register.
Best regards,
Wolfgang Denk
participants (3)
-
sun york-R58495 -
Wolfgang Denk -
York Sun