| Rev |
Age |
Author |
Path |
Log message |
Diff |
| 2619 |
6194 d 20 h |
jermar |
/trunk/ |
Modify ipc_answer_*() to make use of all six syscall arguments. The recommended
means of answering calls is via the ipc_answer_m() macros (where m denotes the
number of return arguments) that automatically decide between the fast register
version or the slow universal version of ipc_answer(). |
|
| 2618 |
6195 d 8 h |
jermar |
/trunk/ |
Modify asynchronous IPC to make use of all six syscall arguments. The preferred
means of asynchronous communication is now via the set of ipc_call_async_m()
macros, where m is the number of payload arguments passed to the kernel. These
macros will automatically decide between the fast and the universal slow version
of ipc_call_async. |
|
| 2617 |
6196 d 1 h |
jermar |
/trunk/kernel/generic/src/ipc/ |
STRUCT_TO_USPACE may fail in sys_ipc_call_sync_fast. |
|
| 2615 |
6196 d 5 h |
jermar |
/trunk/ |
Modify synchronous IPC to make use of all six syscall arguments. The preferred
means of synchronous communication is now via the set of ipc_call_sync_m_n()
macros, where m is the number of payload arguments passed to the kernel and n is
the number of return values. These macros will automatically decide between the
fast and the universal slow version of ipc_call_sync. |
|
| 2601 |
6204 d 0 h |
jermar |
/trunk/kernel/generic/src/ipc/ |
Fix and improve two IPC related comments. |
|
| 2557 |
6247 d 5 h |
jermar |
/trunk/kernel/generic/src/ipc/ |
Enable forwarding of IPC_M_AS_AREA_SEND, IPC_M_AS_AREA_RECV, IPC_M_DATA_SEND
calls. In order to prevent the forwarder from cloberring the call data (i.e.
source and destination address, and size) by treating these three methods as
immutable on forward. This feature is experimental, but has huge benefits in
that it can significantly reduce the amount of data sharing (the middle man need
not modify its address space mappings) or the amount of data copying (the middle
man need not receive the data from the sender and then resend them to the next
recipient). As a result, it can reduce N such calls for a communication channel
with N tasks along the way to 1 such call. |
|
| 2556 |
6247 d 6 h |
jermar |
/trunk/kernel/generic/ |
Rename as_get_size() to as_area_get_size() and add a doxygen comment. |
|
| 2494 |
6326 d 21 h |
jermar |
/trunk/kernel/generic/ |
Implement a new mechanism for copying data between two address spaces based on
IPC. New system method IPC_M_DATA_SEND is introduced. The implementation is
generic and does not make use of any possible architecture optimizations such as
direct copy from one address space to another. |
|
| 2472 |
6350 d 22 h |
jermar |
/trunk/kernel/generic/src/ipc/ |
Some left-over stuff from the last commit (IPC cleanup and improved
comments.) |
|
| 2471 |
6350 d 22 h |
jermar |
/trunk/ |
Greatly improve comments in the IPC layer.
Now I think I finally start to understand our IPC internals :-) |
|
| 2359 |
6381 d 7 h |
jermar |
/trunk/ |
Improve comments for the IPC subsystem.
Fix formatting and indentation. |
|
| 2310 |
6391 d 22 h |
jermar |
/trunk/kernel/generic/ |
Cleanup the waitq_wakeup() interface.
Replace numeric constants (i.e. 0)
and boolean constants (i.e. false) with
more readable WAKEUP_FIRST. Also change the
type of the second argument of waitq_wakeup()
to a newly introduced type wakeup_mode_t.
Fix behaviour of waitq_wakeup() in case
that WAKEUP_ALL semantics is required
but no threads are sleeping in the wait
queue. This is a similar fix to that of
Jan Hudecek committed in the RCU branch,
but, IMHO, is more straightforward and
also doesn't eat up previous missed
wakeups. |
|
| 2216 |
6421 d 21 h |
decky |
/trunk/ |
make thread ID 64 bit (task ID is 64 bit already)
cleanup thread syscalls |
|
| 2183 |
6424 d 1 h |
jermar |
/trunk/kernel/generic/ |
Continue to de-oversynchronize the kernel.
- replace as->refcount with an atomic counter; accesses to this
reference counter are not to be done when the as->lock mutex is held;
this gets us rid of mutex_lock_active();
Remove the possibility of a deadlock between TLB shootdown and asidlock.
- get rid of mutex_lock_active() on as->lock
- when locking the asidlock spinlock, always do it conditionally and with
preemption disabled; in the unsuccessful case, enable interrupts and try again
- there should be no deadlock between TLB shootdown and the as->lock mutexes
- PLEASE REVIEW !!!
Add DEADLOCK_PROBE's to places where we have spinlock_trylock() loops. |
|
| 2098 |
6489 d 23 h |
decky |
/trunk/kernel/generic/ |
avoid conflict with potential keyword |
|
| 2071 |
6501 d 21 h |
jermar |
/trunk/ |
(c) versus (C) |
|
| 2064 |
6515 d 20 h |
vana |
/trunk/ |
SIMICS added as a simulator kernel runs on |
|
| 1933 |
6593 d 20 h |
jermar |
/trunk/kernel/generic/ |
Implement efficient IPC notification cleanup. |
|
| 1932 |
6593 d 21 h |
jermar |
/trunk/kernel/ |
Do not copy notification config out of the IRQ structure.
Add the notify member, that toggles notifications on and off instead.
This is good for preventing other tasks from registering the notification
while the notification config is stored outside the IRQ structure.
It should also help to implement the cleanup code. |
|
| 1923 |
6594 d 20 h |
jermar |
/trunk/ |
Replace the old IRQ dispatcher and IPC notifier with new implementation.
Note that all architectures except for sparc64 are now broken
and don't even compile. |
|