I am doing some development for a project which requires data
to be present inside the kernel as quickly as possible. The system
is very time critical. Right now I am doing a disk read from the disk
to the user space and then using "memcpy_fromfs" to copy data into the kernel.
I would prefer to eliminate the extra copying that takes place
from user to kernel space.
Can I read from the disk directly into kernel space?
is this really broken (I mean, I shoudn't even be thinking about it?? :( ).
I did try it out, but there didn't seem to be and marked improvements. I
have no clue why. Maybe I did something wrong.
Is there any other way how I can eliminate the copying? I
was thinking of reading the data from the disk to the user space and
share memory segment with the kernel memory. I am not sure how to do this.
This will surely be a good idea but breaks the clear boundary from
the user space to kernel space ( which could be harmful ).
Any help in this area is appreciated.
-- Furquan Ansari fansari@tisl.ukans.edu http://www.tisl.ukans.edu/~fansari/