|
I reserved such memory as cma range, and Im sure it can be
allocated via cma, such as dma_alloc_writecombine. I want to hard
code the physical address because I have to make sure such
address can be access by another ipu cores of the soc platform.
I think this can be reserved via a generic way: Just customize
page allocation for a shm object. (I have a look at the kerrnel
source, every omap_bo_object was bind a shm file, so, i only
need allocate pages by my way will resolve all the further problem.
otherwise, i would encounter many other unkown issues for me). On 06.07.2017 18:01, Kamil Konieczny
wrote:
On 06.07.2017 11:43, kipade wrote:In fact, I want to allocate a continous of pages from a given physical address for a shm object in shmem_file_setup, might i should write a customized shmem_file_setup2 for this?I use arm 32bit cpu, with kernel 3.14why do want this specific address 0x90000000 ? can it be 0x40000000 ? or 0x50000000 ? or 0x68880000 ? is this 0x90000000 located in hardware registers ? some memory regions are not accessible directly for allocation, because they are used for communication with hardware SoC, and so they are used by kernel drivers or directly by kernel kernel driver may or may not expose that memory region for user space, but this is driver work maybe you should ask on platform forum, you did not answer my question on what platform you are workingOn 06.07.2017 05:08, kipade wrote:Here I want to allocate a block of shared memory from a given physical address, such as 0x90000000, 6M. However, alloc_pages can't fit this point. I had reserved enugh pages via cmem, there also have enough continous pages for such allocation anywhere else. What should I do? |
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@xxxxxxxxxxxxxxxxx https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
