Hello Mudit, On Sunday 28 of August 2016 09:27:16 Mudit Jain wrote: > Hi, > > That's great news Pavel. > Should I start merging all your findings and push it as a patch ? > What would be the next steps ?
I suggest to invest the time into SD part of the project now. It needs to use DMA as well. As I have seen, there is/or has been copy of DMA support in libbsd code. It is necessary to decide where DMA implementation would sit. There should be only single implementation to make system maintainable. Even if two independent APIs are used then channels (resources) allocation has to be managed by single code. Try to push SD part code to the state where it reads sectors through device. If you compile shell into RTEMS application you can use something like hexdump -n 512 /dev/sd-device to read actual SDcard content. You need to config RTEMS with #define CONFIGURE_SHELL_COMMANDS_INIT #define CONFIGURE_SHELL_COMMANDS_ALL #define CONFIGURE_SHELL_MOUNT_MSDOS #define SHELL_TASK_PRIORITY 50 and start shell from Init by call like rtems_shell_init("SHLL",RTEMS_MINIMUM_STACK_SIZE+0x1000, SHELL_TASK_PRIORITY,"/dev/console",1,0, NULL); As for the actual DMA API code. It can be considered as initial base for experiments but it is far from mergeable state. It has to be cleaned, may it be, it worth to think about some generalized DMA API which can be used by multiple targets and then rewrite code to that concept. So you can take my branch, reorganize and merge patches to have initial DMA support introduction patch done right way and functional. You can even send it for review to document the work and get feedback but expect that most of the code would require discussion and rewrite. Good thing is, that evolution can be incremental and can be tested that it works still after each iteration. Best wishes, Pavel _______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel