Am 14/04/2012 22:13, schrieb Wolfgang Denk: > Dear Stefano, > > In message <4f898ca5.8070...@denx.de> you wrote: >> >> The start address CONFIG_SYS_MEMTEST_START is the lowest address of the >> SDRAM that can be accessed, that is MMDC0_ARB_BASE_ADDR (0x10000000) for >> i.MX6. > > But it _should_ be the lowest address that can _safely_ be used for > such tests, i. e. the range MEMTEST_START - MEMTEST_START must neither > contain the exception vectors, not U-Boot's code, stack or malloc arena. > >> dynamically, also the last RAM address that can be tested should be >> computed at run time and not at compile time with CONFIG_SYS_MEMTEST_END. > > CONFIG_SYS_MEMTEST_END is NOT the end of physical RAM, but the end of > the range that can be tested without crashing U-Boot...
Well, but why do we have to define at compile time this value, that in most cases can be wrong ? Why should "mtest" use a default value if not all parameters are passed ? IMHO it is better if mtest returns with an error if end address is not provided. > > Such a patch would be wrong. Really I think it is better to get rid completely of CONFIG_SYS_MEMTEST_START and CONFIG_SYS_MEMTEST_END without trying to make some (maybe wrong) assumptions, and let mtest doing its job Best regards, Stefano Babic -- ===================================================================== DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-0 Fax: +49-8142-66989-80 Email: off...@denx.de ===================================================================== _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
