PHYSICAL_PAGE_{NO,CAN}_WAIT into an actual flag
PHYSICAL_PAGE_DONT_WAIT.
* Pass the flags through to the chunk mapper callback.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27979 a95241bf-73f2-0310-859d-f6bbb57e9c96
* unload settings when ref_count is zero and boot device is available
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27969 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Use the new VMCache::Read() flags parameter to directly read into the
physical page in the page fault handler instead of mapping it first.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27966 a95241bf-73f2-0310-859d-f6bbb57e9c96
in struct stat.
* Instead, I followed Marcus' great idea and added a compatibility check in
the runtime loader: now, R5 binaries (also shared libraries) are detected,
and they get special versions for stat(), fstat(), and lstat() that return
the smaller stat struct.
* However, I've disabled (in src/system/libroot/posix/sys/stat.c) using the
larger stat field for now, as this breaks some of our optional packages.
So until we rebuild them all, this shouldn't be enabled.
* This should now also be used for BeOS compatibility in libnetwork.so.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27961 a95241bf-73f2-0310-859d-f6bbb57e9c96
* memset() is now available through the commpage.
* CPU modules can provide a model-optimized memset().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27952 a95241bf-73f2-0310-859d-f6bbb57e9c96
subdirectories. Also moved the x86 kernel arch_string.S there.
* Moved memcpy.c from src/system/libroot/posix/string into the
arch/generic subdirectory.
* Dealt with the consequences of moving things around. Affected are also
the boot loader and runtime loader builds.
Adjust the m68k and ppc parts, too, but only the x86 build is tested.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27947 a95241bf-73f2-0310-859d-f6bbb57e9c96
- try to accomodate for low ram systems by making the first kernel heap allocation smaller
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27943 a95241bf-73f2-0310-859d-f6bbb57e9c96
arch_vm_aspace_swap().
* The x86 implementation does now maintain a bit mask per
vm_translation_map_arch_info indicating on which CPUs the address
space is active. This allows flush_tmap() to avoid ICI for user
address spaces when the team isn't currently running on any other CPU.
In this context ICI is relatively expensive, particularly since we map
most pages via vm_map_page() and therefore invoke flush_tmap() pretty
much for every single page.
This optimization speeds up a "hello world" compilation about 20% on
my machine (KDEBUG turned off, freshly booted), but interestingly it
has virtually no effect on the "-j2" haiku build time.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27912 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Added smp_send_multicast_ici(), which sends the message to all CPUs
specified via a mask.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27910 a95241bf-73f2-0310-859d-f6bbb57e9c96
timed out thread has a higher priority than the currently running one.
Maybe we should even restrict this behavior to realtime threads.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27897 a95241bf-73f2-0310-859d-f6bbb57e9c96
* get_nth_symbol() did not correctly iterate over the symbol hash, causing it
to return the same symbols more than once, and omit others.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27896 a95241bf-73f2-0310-859d-f6bbb57e9c96
need to cast explicitly before the multiplication/shift, since the
former is 32 bit and the latter 64 bit. The worst instance was in
swap_file_add(), where the page count was int32, so that swap file
sizes between 2 and 4 GB resulted in a negative available swap space
size. Fixes bug #2721.
* Fixed and added optional debug output.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27877 a95241bf-73f2-0310-859d-f6bbb57e9c96
ignore every other cache (starting from the first).
* The consequence of this was that no blocks were written back automatically
for those caches, and their transactions were never idle, causing bug #2781.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27831 a95241bf-73f2-0310-859d-f6bbb57e9c96
Currently it only contains KDEBUG and the block cache debugging macros.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27816 a95241bf-73f2-0310-859d-f6bbb57e9c96
buffer format, the buffer can now also contain other events than just
stack traces. ATM these are only references to the image events
(created/deleted). Therefore we no longer have to flush the profiling
buffer after such an event, since the debugger can exactly match the
samples. Since we couldn't flush when the profiling timer hit while the
thread was in the kernel, that wasn't working that well anyway.
"profile -f" fails to translate stack trace addresses only very rarely,
now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27775 a95241bf-73f2-0310-859d-f6bbb57e9c96
exec*()ed, the debugger never got notified that a runtime loader image
was created.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27773 a95241bf-73f2-0310-859d-f6bbb57e9c96
using it for something else screws stack traces while being in such a
function.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27772 a95241bf-73f2-0310-859d-f6bbb57e9c96
up module image paths in module_init_post_boot_device(). Not sure whether
it also could cause #2776.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27760 a95241bf-73f2-0310-859d-f6bbb57e9c96
* No longer send a SIGKILL when a page fault occurred, there was no
signal handler, and the debugger told us to continue as usual. Instead
we send a SIGSEGV. Instead in handle_signal() when not in the main
thread and there's no handler for the signal, we first send the main
thread a SIGKILL before letting the thread die.
So in cases where the main thread caused an unhandled page fault, the
team will die from the SIGSEGV, now. This fixes bug #2773.
* For the other hardware-caused signals we do now do the same as in case
of page faults, i.e. we first check whether the thread has a handler
for the signal in question. If so, we don't notify the debugger, but
send the signal right away.
* B_GENERAL_PROTECTION_FAULT is translated to SIGILL now. Seems better
than SIGKILL.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27757 a95241bf-73f2-0310-859d-f6bbb57e9c96
letting the boot loader provide full paths for the pre-loaded images.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27753 a95241bf-73f2-0310-859d-f6bbb57e9c96
been mounted, before anyone could try to load any modules from it.
Also pass it a flag whether the boot volume is where the boot loader
pre-loaded the modules from.
* module_init_post_boot_device() changes the pre-loaded module image
paths to normalized boot volume paths, now. Got rid of the code in
register_preloaded_module_image() which tried something like this.
* Changed module image ref counting. A referenced module has single
reference to its image, which is released when the module becomes
unreferenced.
* get_module() for a reference module will not try to re-get and re-set
the module's image anymore. That could lead to a similar module (from
different paths) being loaded at the same time. A module from a new
file can only be loaded when the old one has been put completely.
* Simplified B_KEEP_ALIVE module handling a bit. When the module is
initialized, we add another reference, which we'll never free. Thus
the module remains loaded without special handling. Removed
module_image::keep_loaded. A B_KEEP_ALIVE module remains referenced
and thus its image remains referenced, too.
* Removed module::file, a cached path to the module's image. An
optimization that wouldn't work with multiple root directories for
modules (/boot/beos/..., /boot/common/...) or when module files were
moved. get_module() does now always search the image file, when the
module is still unreferenced. This should be a bit slower than before,
but I didn't notice any difference in VMware at least. If it turns out
to be a problem we could introduce a more intelligent cache that stays
up to date by using node monitoring.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27752 a95241bf-73f2-0310-859d-f6bbb57e9c96
function (normalize_path()), now. There was some code duplication
before.
* Added "bool traverseLink" parameter to vfs_normalize_path(). When
true and the leaf component is a symlink, it will be resolved.
* KPath:
- Added similar leaf link traversal parameter to SetTo() and
SetPath().
- Added Normalize().
- Added DetachBuffer(), which returns the object's current buffer and
unsets itself.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27751 a95241bf-73f2-0310-859d-f6bbb57e9c96
code (like found in Firefox) that relies on get_next_image_info() to
return it first.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27732 a95241bf-73f2-0310-859d-f6bbb57e9c96
elf_add_memory_image_symbol(). The former creates and registers a new
image that has not been loaded from a file. The latter adds a symbol to
its symbol table. This is mainly a debug feature, allowing to name code
or data in memory regions that aren't associated with loaded ELF
objects.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27721 a95241bf-73f2-0310-859d-f6bbb57e9c96
added FUNCTION_END() calls for a good deal of functions. The respective
ELF symbols do now have a correct size.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27718 a95241bf-73f2-0310-859d-f6bbb57e9c96
team does have a runtime loader image. The "profile" tool can thus
translate addresses in the runtime loader correctly.
Note that this change will break code that assumes that the application
image is the first image returned by get_next_image_info().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27715 a95241bf-73f2-0310-859d-f6bbb57e9c96
been called for a team, and fail installing the default debugger if it
is set. This makes disable_debugger() actually work. Fixes bug #2763.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27713 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Extended the profiling API by an option to record a variable number of
samples per tick. The stack depth is used as a maximum.
* Added new option "-f" to the "profile" tool. When specified it
increments the hit counts of all symbols in the full available caller
stack. I.e. the resulting hit counts will approximate the total time
spent in each function or any function directly or indirectly called
by it. Thus "_start" and "main" will usually get 100% and leaf
functions only what time has actually been spent in them.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27699 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Reduced the kernel_args array size from 32 to 16 (7 are used on a normal
build).
* Cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27690 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Fixed hash_remove_current(): It didn't update "lastElement" and thus
always also removed all elements in the same bucket preceding the one
to be removed. Also got rid of the useless "for" loop.
Fixes#2757.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27689 a95241bf-73f2-0310-859d-f6bbb57e9c96
passed to the kernel with "boot" instead of the volume name; the kernel
mounts the boot volume always as "/boot".
* This should fix#2757.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27685 a95241bf-73f2-0310-859d-f6bbb57e9c96
1 ms to 1 us. The long minimum timeout seriously screwed profiling,
since by default it works with 1 ms ticks, stopping the timer when the
thread is unscheduled and restarting it with the remaining time when it
is scheduled again.
I could also imagine that this had a negative effect on latencies and
the precision of thread wakeup times.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27672 a95241bf-73f2-0310-859d-f6bbb57e9c96
flags on the wrong thread (the current one instead of the newly created
one). This would cause the thread not to be debugged as it should, e.g.
profiling wouldn't work correctly.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27671 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Adjusted used files systems accordingly.
* BFS::Stream::GetName() was broken. It accessed the small data region
which wasn't loaded, since BFS::Stream derived from bfs_inode, which
is a variably-sized structure with the small data region at the end.
Changed that to a ref-counted, shared member instead.
* Implemented RootFileSystem::GetName().
* Added Directory::GetPath() to get a full path of the directory or an
entry.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27664 a95241bf-73f2-0310-859d-f6bbb57e9c96
buffer during a timer event that interrupted a kernel function: We do
now flush the buffer as soon as it is 70% full, *if* we didn't interrupt
a kernel function. When the buffer runs full and we still haven't hit a
user function, we drop the tick. The number of dropped ticks is recorded
and sent to the debugger with the next update message.
Reverted the previous partial solution (the temporary disabling of
profiling while in debugger support code).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27658 a95241bf-73f2-0310-859d-f6bbb57e9c96
is created or deleted (or exec*() has been invoked). The counter is
sent with several debugger messages.
* Track the image event counter that is used when samples are added to
the profiling buffer. If the current team counter differs, we flush
the buffer first (sending an update message to the debugger), so that
the debugger has a chance to match the addresses to the correct images.
* Disable profiling for a thread while it runs in the debugger support
code. This fixes potential deadlocks which could occur when a
profiling timer event occurred that would require the buffer to be
flushed while the thread was just sending something to the debugger or
waiting for a command. As it turns out, this is not sufficient either,
since we should never try to flush the buffer when the timer event
occurred in the kernel, since the thread might hold a lock that the
debugger thread could try to acquire. Will implement a more general
solution later.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27656 a95241bf-73f2-0310-859d-f6bbb57e9c96
* We can't enable interrupts in an interrupt handler. Instead we use the
newly introduced callback feature, which notifies the debugger right
before returning from the interrupt.
* We didn't indicate that the profiling buffer was full and that the
thread shouldn't be profiled ATM. Therefore it could happen that it
was profiled while trying to notify the debugger that the profiling
buffer was full, resulting in a deadlock. Introduce a respective flag
in the thread debug structure.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27649 a95241bf-73f2-0310-859d-f6bbb57e9c96
interrupt handler and will be executed right before returning from the
interrupt.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27648 a95241bf-73f2-0310-859d-f6bbb57e9c96
be tracked to the kernel, which then counts the hits, an area is
passed to kernel in which the hits are recorded. When the area is
full, the debugger is notified. For some reason that part doesn't work
yet -- the whole system freezes when waiting for a reply.
* Reorganized the profile tool code a bit. For one with respect to the
changed API, but also to prepare tracking of image creation/deletion.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27640 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Added global hash table for images.
* Improved a bit of code by using the new image hash table. E.g.
_get_image_info() can return infos for images of any team, now.
* Fixed remove_images() comment: The function must not be invoked with
the team lock being held.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27606 a95241bf-73f2-0310-859d-f6bbb57e9c96
creation), we didn't release the team debug info spinlock and reenabled
interrupts.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27602 a95241bf-73f2-0310-859d-f6bbb57e9c96
various system information.
* Implemented retrieving some VM stats via this call.
* The VM now maintains a page fault counter, and sets system_info::page_faults
accordingly.
* Added a (pretty simple) "vmstat" command line app.
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27597 a95241bf-73f2-0310-859d-f6bbb57e9c96
messages easily hit the previous limit. Maybe another solution should be
sought.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27591 a95241bf-73f2-0310-859d-f6bbb57e9c96
still pretty much work in progress.
* Introduced init_thread_debug_info() which is used instead of
clear_thread_debug_info() when the thread is created. The latter
requires former initialization.
* user_debug_thread_deleted() is now already invoked in thread_exit(),
not in the undertaker.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27531 a95241bf-73f2-0310-859d-f6bbb57e9c96
disabled, as long as a fault handler is installed for the thread. It
allows user_memcpy() to be invoked with interrupts disabled -- in this
case it will simply fail, when the address is valid, but not mapped yet.
This is a more desirable behavior for debug facilities, though in all
other cases it is likely a bug. We should probably introduce a thread
flag to discriminate these situations.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27530 a95241bf-73f2-0310-859d-f6bbb57e9c96
many iframes are supposed to be skipped before recording the stack
trace. Currently implemented for x86 only.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27529 a95241bf-73f2-0310-859d-f6bbb57e9c96
When specified, the message is read but not removed from the port.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27527 a95241bf-73f2-0310-859d-f6bbb57e9c96
as we cannot do so with interrupts turned off (ACPI needs to allocate memory
dynamically).
* Turn off interrupts right before going to sleep (_GTS), this at least works
in VMware, maybe it also works on real hardware.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27500 a95241bf-73f2-0310-859d-f6bbb57e9c96
* CID 314 and CID 315: Same as above just for WriteLockPartition.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27476 a95241bf-73f2-0310-859d-f6bbb57e9c96
into two seperate checks instead and return B_BAD_VALUE on a passed NULL path.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27475 a95241bf-73f2-0310-859d-f6bbb57e9c96
need for a null check. Or the null check needs to be moved to after last is
initialized. But if the user_memcpy of cookie succeeds, last should not be null
since it is just a copy of cookie. Probably.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27429 a95241bf-73f2-0310-859d-f6bbb57e9c96
