On Wed, Dec 26, 2012 at 10:56:53AM -0800, Simon Glass wrote:
This series aims to get all the memory functions running correctly on sandbox.
There was some discussion about this a while ago, and a commit was added to show a possible approach:
355a8357 sandbox: Change md command to use map_physmem
This commit was subsequently reverted because it used map_physmem() instead of the NOP that most architectures need for the memory functions.
This series introduces map_sysmem(), a NOP on all architectures except sandbox. It allows us to use a ram buffer to which all U-Boot addresses are relative. The memory commands (including hashing) are updated to use this so that sandbox can now use those commands.
Half of the mtest code is behind #ifdefs and there is duplication of some functions in both versions of the memory test. Several patches here clean this up a bit and get it working on sandbox.
The numeric setenv_ulong() function is a useful way of avoiding a 'char buf[17]; sprintf(buf, "%ld", ...); setenv("...", buf)' sequence. There is also setenv_addr(). What is missing is setenv_hex() which sets a ulong in hex format. Add this function and then make use of it in the main places: common/ drivers/ and net/.
The recently added and very basic hash instructure can help reduce code duplication in some cases. Redo the crc32 command to use this, and make it available through the 'hash' command. Also a few bugs were found in hashing with verify disabled - the arg count was not checked and a variable declaration was missing.
To permit the memory tester to run on sandbox, we need ctrl-C to work. To achieve this, add a proper implementation of sandbox's tstc(), with a simple FIFO for character input. An os_usleep() is added to ensure that U-Boot does not consume infinite CPU when setting at the command prompt.
With all of this it is possible to use the memory commands in sandbox, as well as crc32 and the other hashing commands.
So, aside from a few posted comments: Reviewed-by: Tom Rini trini@ti.com
And pending the answer to if you plan to v2 the hash command part (and since everyone sets CONFIG_CMD_CRC32, we do want to be careful), I'm OK with applying this bundle.