- removed (%d) Sleeping print from contended kernel mutexes.
- removed (%d) Waiting print from WAIT_EVENT used by the pager for suspending tasks.
- removed prints from the mutex_control syscall and user mutex test.
- Fixed a wrong instruction in mutex.S user library
- Added support for blocking lock/unlock
- Divided waiting into wait_on_prepare and wait_on_prepared_wait
so that mutex_control lock is released after getting in the waitqueue.
- Declaring waitqueue on the stack should be done outside wait_on_prepare
Issues:
- Tests can be simplified for atomic data access instead of producer/consumer.
- kmalloc variable sized memory caches are not freed properly. Currently only the
last slot can be freed, occupied correctly. it should be done in any slot, i.e.
1, 2, 3, 4 instead of just 5.
- Need to add a mutex to kmalloc.
mmap() with MAP_ANON | MAP_SHARED is used for creating shared
memory mappings that can be shared among a process and its descendants
(think fork())
Currently shmget/at creates shared virtual files on the fly and calls
do_mmap to create anonymous shared mappings. Now sys_mmap uses this
interface to create virtual files on the fly and call do_mmap on those
files.
In the future a common shared virtual file creation method should be
added and commonly used by both interfaces.
- Mutex test added. Forked tasks demonstrate produce/consumer using a
shared mmap'ed page.
- Added l4lib assembler syscall
- Added forgotten SWI to mutex control offset in syscall page.
- Added mutex head initialization
- Contended child successfully sleeps in a waitqueue.
Issues:
- Somehow the child's produced page buffer is altered at about [4020] offset.
Parent fails to validate buffer therefore.
- Need to add syncing to test so that parent does not unlock and lock again
before child has a chance to lock buffer and produce.
- Compiles and Codezero runs as normal without touching mutex implementation
- Mutex implementation needs testing.
The mutex control syscall allows userspace programs to declare any virtual
address as a mutex lock and ask for help from the kernel syscall
for resolving locking contentions.
Previously during ipc copy, only the currently active task flags were
checked. This means the flags of whoever doing the actual copy was used
in the ipc. Now flags are stored in the ktcb and checked by the copy routine.
Current use of the flags is to determine short/full/extended ipc.
Benefits & Facts:
- Messages up to 2 kilobytes may be sent.
- Both parties may use non-disjoint user buffers. E.g. any userspace address.
- Userspace buffers can page fault.
- Page faults punish timeslice of only the faulting thread.
- Any number of extended ipcs can take place at any one time, since
only ktcbs of ipc parties are engaged. No global buffer is used.
- This also provides smp-safety benefit.
Disadvantages:
- There is triple copying penalty. This has to be done:
- Sender buffer to sender ktcb
- Sender ktcb to receiver ktcb
- Receiver ktcb to receiver buffer.
This is due to the fact that buffers can be on non-disjoint userspace addresses.
If you want to avoid disadvantages and lose some of the benefits,
(e.g. address freedom, shorter copy size) use FULL IPC.
Upon fork, child was created in a new space but as a copy of any
cloned thread in the parent space. This was due to the search of forker thread
by its space id (which is shared among many cloned threads).
Now fixed.
modified: src/api/thread.c
modified: tasks/mm0/src/task.c
- Test0 has a full ipc mr read/write test.
- A full ipc occurs for definite only if both parties use the FULL IPC flag.
Otherwise the thread that makes the ipc copy rules on whether it was a short
or a full copy.
- Added a full ipc send/recv test
- Removed non-zero value checking in r2 for ipc that was there
to catch inadvertent full ipc calls.
- Added correct hanlding for read/write mrs for current status of utcb.
TODO:
- Add mapping of every utcb to every task for privileged access so that
the kernel can access every utcb without switching spaces.
- Removal of same mappings
- Upon thread creation need to copy page tables accordingly i.e.
each task will have its own utcb mapped with USER access, but every
other utcb as kernel access only. Need to handle this case upon page
table copying.
- Added ARM register ipc usage explanation to glue/arm/message.h
- In the current design, the unused r2 register is a system register
that kernel checks for ipc flags such as:
- IPC type: e.g. full or extended.
- In extended IPC, MR index containing message buffer ptr.
- In extended IPC, message size
- During execve/map_stack when setting up the environment and argument stack,
the stack map size was in bytes instead of pfns.
- Can now execve 16 forked tasks with success.
- Only the topmost parent prints pass messages. Any failed child can print fail message.
- Added testing of 16 forked threads doing file create/read/write/close and
16 forked x 4 cloned = 64 threads spawning/exiting
It is possible that in a mult-threaded address space write-faults
(or faults in general) on the same address can occur. This is because
threads may become runnable during the handling of the first fault.
This causes duplicate faults on the same private page in the same address space.
On the case of write faulted private page, this causes a spurious page allocation
Currently this case is detected and handled, but in the future we need
a generic way of detecting duplicate faults (of any kind) and cease duplicate
IPC at a very early stage. This is not done yet as it requires knowledge of the
PTEs of physical pages in the pager (like reverse mapping in Linux).
- KIP's pointer to UTCB seems to work with existing l4lib ipc functions.
- Works up to clone()
- In clone we mmap() the same UTCB on each new thread - excessive.
- Generally during page fault handling, cloned threads may fault on the same page
multiple times even though a single handling would be enough for all of them.
Need to detect and handle this.
Added setting of utcb address to l4_thread_control.
This is going to be moved to exchange_registers() since we need to pass
both the utcb physical and virtual address and exregs fits such context
modification better than thread_control.
- Now libl4 has no references to utcb page or shmat etc.
- Pager does not deal with special case utcb page allocation.
It instead allocates a shared page from shm memory pool.
- All tasks working to original standard.
Next:
- Add per-thread utcb allocation from the kernel
- Add larger register file for standard ipc
- Add long ipc (up to 1Kb)
Previously a so-called utcb shared page was used for transfering
data between posix services. This was a special shmat/get/dt case
allocating from its own virtual pool. Now the term utcb is renamed
as a shared page and integrated with the shm* handling routines.
Generic l4 threads will use long-ipc and not this method. Posix
services will continue to communicate on a shared page for now.
modified: tasks/libl4/include/l4lib/ipcdefs.h
modified: tasks/libl4/src/init.c
new file: tasks/libposix/init.c
modified: tasks/mm0/include/shm.h
modified: tasks/mm0/include/task.h
deleted: tasks/mm0/include/utcb.h
modified: tasks/mm0/main.c
modified: tasks/mm0/src/boot.c
modified: tasks/mm0/src/clone.c
modified: tasks/mm0/src/execve.c
modified: tasks/mm0/src/exit.c
modified: tasks/mm0/src/init.c
modified: tasks/mm0/src/shm.c
modified: tasks/mm0/src/task.c
deleted: tasks/mm0/src/utcb.c
deleted: tools/l4-qemu
- Directory creation, file read/write is OK.
- Cannot reuse old task's fds. They are not recycled for some reason.
- Problems with fork/clone/exit. They fail for a reason.
