We now have a working periodic wall checker, enabled by USE_CHECKING_WALL.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9459 a95241bf-73f2-0310-859d-f6bbb57e9c96
list (but not the structure itself).
Now calls heap_init_post_thread(), heap_init_post_sem() was called ..._postsem() before.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9457 a95241bf-73f2-0310-859d-f6bbb57e9c96
they are called.
Fixed some return types (mostly from int to status_t).
Some minor other cleanup.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9440 a95241bf-73f2-0310-859d-f6bbb57e9c96
they are called.
Introduced a cpu_init_post_vm() that will now call arch_init_post_vm() instead
of letting main() doing it.
Fixed some return types (mostly from int to status_t).
Some minor other cleanup.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9439 a95241bf-73f2-0310-859d-f6bbb57e9c96
three code locations that didn't handle them properly (when iterating over all
areas).
That and the fact that unmap_and_free_physical_pages() is no longer broken now
leads to the boot region being unmapped and freed properly as intended before.
That revealed another bug: the boot loader region was disposed too early, we
need a few other components pick up boot loader resources first; it now happens
in vm_init_post_sem().
allocate_kernel_args() was also broken, and actually didn't try to allocate
kernel args areas, but tried to reallocate all the virtual allocated ranges
of the boot loader...
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9437 a95241bf-73f2-0310-859d-f6bbb57e9c96
boot loaders passed in memory completely wrong (for example, the premapped
graphics buffer could be mapped to the same region as some thread stacks,
just because the VM did not use its knowledge about that mapping): the boot
loader regions are now reserved so that they cannot be overwritten
with other regions anymore. The kernel_args arguments are now tracked, and
areas are created for them (they are not yet freed, though). arch_vm_init_end()
must now call the new vm_free_unused_boot_loader_range() to free any remaining
parts of the memory not yet mapped in by any areas.
Renamed a lot of init functions (ie. postsem to post_sem, init2 to post_area),
also updated because of the vm_translation_map*() arch_vm_translation_map*() move.
B_ALREADY_WIRED can now only be used during kernel startup (it's a private flag).
Replaced PAGE_SIZE with B_PAGE_SIZE.
Some minor cleanups.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9429 a95241bf-73f2-0310-859d-f6bbb57e9c96
correctly by calling arch_vm_translation_map_init_kernel_map_post_sem().
dump_aspace() now behaves like all debugging functions should behave: it
now accepts decimal numbers as well as hexadecimal numbers as IDs.
vm_translation_map_create() was renamed to arch_translation_map_init_map().
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9427 a95241bf-73f2-0310-859d-f6bbb57e9c96
Moved some functions around, so that the public functions come last.
Renamed the internal vm_translation_map_quick_query() to early_query(),
renamed vm_translation_map_quick_map() to arch_vm_translation_map_early_map().
Removed the _module part of the initialization functions.
Added a function that completes the initialization of the kernel map (it
creates the map's lock, which was just not done before).
arch_vm_translation_map_init_map() will now fail with an appropriate
error code if the map's lock couldn't be created for anything that
is not the kernel map.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9426 a95241bf-73f2-0310-859d-f6bbb57e9c96
loader preallocated, arch_vm_init_end() (formerly called _endvm())
now calls the new vm_free_unused_boot_loader_range() with the correct
parameters.
Moved the creation of the DMA area from 0-0xa0000 into arch_vm_init_post_area().
Changed return types to status_t rather than just having "int" stand
for a status code.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9425 a95241bf-73f2-0310-859d-f6bbb57e9c96
USE_WALL feature can now be used.
Started implementing a USE_CHECKING_WALLS feature, that has a kernel
daemon running that periodically checks the walls of *all* memory
allocations - does not work yet (because of some initialization order
problems).
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9396 a95241bf-73f2-0310-859d-f6bbb57e9c96
team and process group IDs as well.
Added a ToDo item that process groups do only work well as long as
their group leader is still active.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9394 a95241bf-73f2-0310-859d-f6bbb57e9c96
ignored completely (so that zero and below become valid).
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9393 a95241bf-73f2-0310-859d-f6bbb57e9c96
no longer pop %fs from the iframe in that case, too.
When we restore a custom stack, however, we'll now pop %fs from the iframe
as such strange threads may even do their own business with it.
Added some comments.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9369 a95241bf-73f2-0310-859d-f6bbb57e9c96
We also should disable interrupts in arch_restore_fork_frame().
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9368 a95241bf-73f2-0310-859d-f6bbb57e9c96
always copies the arguments and environment arrays.
No need for the "owns_arrays" in team_arg anymore - it now always owns
them.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9363 a95241bf-73f2-0310-859d-f6bbb57e9c96
Switched to load_image() instad of using the old team_create_team() to start "init".
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9361 a95241bf-73f2-0310-859d-f6bbb57e9c96
The path name is now always taken from the provided argument list (except for
_user_exec() which now replaces args[0] with the path argument).
The team_arg structure know nows if it must delete its arguments or not - this
also fixes a memory leak in case team_create_team() failed when invoked from
userland.
load_image() is now also exported for the kernel.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9360 a95241bf-73f2-0310-859d-f6bbb57e9c96
process ID (in the current session).
get_death_entry() could call free() with interrupts disabled; now follows
the same mechanism as chosen with the groups (ie. it returns a pointer
to a death_entry to be freed by the calling function).
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9354 a95241bf-73f2-0310-859d-f6bbb57e9c96
the specified process group (by passing in a negative pid, as in BeOS).
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9351 a95241bf-73f2-0310-859d-f6bbb57e9c96
- we now have a working process group/session model. Implemented related syscalls
and the kernel backend, some POSIX process functions are now available in the
kernel as well.
- dying teams now monitor their exitus by leaving death_entry structures in
their parent team.
- Implemented wait_for_child() for pid == -1 and pid > 0; IOW waiting for process
groups is missing yet. Hasn't yet been tested intensively.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9350 a95241bf-73f2-0310-859d-f6bbb57e9c96
may be renamed):
- B_RELEASE_ALL: the semaphore count is set to 0, all waiting threads are released
(the "count" argument of release_etc_sem() is ignored then)
- B_RELEASE_IF_WAITING_ONLY: the semaphore count is only decreased if there
are any waiting threads; ie. the semaphore is signaled
Together, they will make the pthread_cond_*() functions easy to implement, and
they come in handy at other places, too.
Removed release_threads variable in release_sem_etc(), as it was useless.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9330 a95241bf-73f2-0310-859d-f6bbb57e9c96
Also removed syscall for that function (shouldn't have been there in the
first place, anyway).
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9325 a95241bf-73f2-0310-859d-f6bbb57e9c96
no longer abuse the semaphore mechanism to carry an extra status value,
instead, wait_for_thread() registers a death_entry with the thread that
will be filled upon exit.
Removed delete_sem_etc() as it's no longer of use or available.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9323 a95241bf-73f2-0310-859d-f6bbb57e9c96
flag set.
Fixes _user_wait_for_thread() which returned an invalid return code in case
of error.
Some minor cosmetics.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9321 a95241bf-73f2-0310-859d-f6bbb57e9c96
not be restored from the iframe; it will be set by the scheduling code.
This is only of concern on SMP machines and could prevent TLS from working.
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9310 a95241bf-73f2-0310-859d-f6bbb57e9c96
same mechanism as NewOS, ie. using the iframe "stack" pointer (using the
new i386_get_current_iframe() function).
Implemented fork() support functions arch_{store|restore}_fork_frame().
git-svn-id: file:///srv/svn/repos/haiku/trunk/current@9292 a95241bf-73f2-0310-859d-f6bbb57e9c96
Axel Dörfler