Searched hist:ff0ab8af9c3f36e7b6f716c3b9e8811a4202eec6 (Results 1 – 2 of 2) sorted by relevance
| /openbmc/linux/arch/parisc/kernel/ |
| H A D | process.c | diff ff0ab8af9c3f36e7b6f716c3b9e8811a4202eec6 Fri Oct 05 17:55:57 CDT 2012 Al Viro <viro@zeniv.linux.org.uk> parisc: optimizations in copy_thread() and friends
* in user thread case the registers had been copied as part of task_struct
already; no need to do it in copy_thread().
* no need to store kernel stack pointer into regs->r21; we know its offset
anyway.
* no need to clobber r3 in sys_fork_wrapper and friends - r28 will do just
as well and *it* will be overwritten anyway.
* no need to mess with storing the return address for child - it should just
use syscall_exit.
* no need to bother with separate stack frame for sys_clone() - just branch
there and be done with that.
* no need to bother with wrapper_exit - we need it only on the child_return,
so let's just do it there.
* use the same ksp for kernel threads and userland ones, while we are at it,
and let ret_from_kernel_execve() go through the normal syscall_exit. More
straightforward is better here...
[fixes from jejb folded]
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
|
| H A D | entry.S | diff ff0ab8af9c3f36e7b6f716c3b9e8811a4202eec6 Fri Oct 05 17:55:57 CDT 2012 Al Viro <viro@zeniv.linux.org.uk> parisc: optimizations in copy_thread() and friends
* in user thread case the registers had been copied as part of task_struct
already; no need to do it in copy_thread().
* no need to store kernel stack pointer into regs->r21; we know its offset
anyway.
* no need to clobber r3 in sys_fork_wrapper and friends - r28 will do just
as well and *it* will be overwritten anyway.
* no need to mess with storing the return address for child - it should just
use syscall_exit.
* no need to bother with separate stack frame for sys_clone() - just branch
there and be done with that.
* no need to bother with wrapper_exit - we need it only on the child_return,
so let's just do it there.
* use the same ksp for kernel threads and userland ones, while we are at it,
and let ret_from_kernel_execve() go through the normal syscall_exit. More
straightforward is better here...
[fixes from jejb folded]
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
|