The BOOT_GDT_SEGMENT_COUNT was based on USER_DATA_SEGMENT on both
x86 and x86_64. However, on x86_64 the order of the segments is
different, leading to a too small gBootGDT array. Move the define to
the arch specific headers so they can be setup correctly in either case.
Also add a STATIC_ASSERT() to check that the descriptors fit into the
array.
Pointed out by CID 1210898.
This patch adds user_access() which can be used to gracefully handle
page faults that may happen when accessing user memory. It is used
by arch_cpu_user{memcpy, memset, strlcpy}() to allow using optimized
functions from the standard library.
Currently only x64 uses this, but nothing really is arch specific here.
Signed-off-by: Paweł Dziepak <pdziepak@quarnos.org>
The kernel is allowed to use fpu anywhere so we must make sure that
user state is not clobbered by saving fpu state at interrupt entry.
There is no need to do that in case of system calls since all fpu
data registers are caller saved.
We do not need, though, to save the whole fpu state at task swich
(again, thanks to calling convention). Only status and control
registers are preserved. This patch actually adds xmm0-15 register
to clobber list of task swich code, but the only reason of that is
to make sure that nothing bad happens inside the function that
executes that task swich. Inspection of the generated code shows
that no xmm registers are actually saved.
Signed-off-by: Paweł Dziepak <pdziepak@quarnos.org>
Enable SSE as a part of the "preparation of the environment to run any
C or C++ code" in the entry points of stage2 bootloader.
SSE2 is going to be used by memset() and memcpy().
Signed-off-by: Paweł Dziepak <pdziepak@quarnos.org>
The possibility to specify custom memcpy and memset implementations
in cpu modules is currently unused and there is generally no point
in such feature.
There are only 2 x86 vendors that really matter and there isn't
very big difference in performance of the generic optmized versions
of these funcions across different models. Even if we wanted different
versions of memset and memcpy depending on the processor model or
features much better solution would be to use STT_GNU_IFUNC and save
one indirect call.
Long story short, we don't really benefit in any way from
get_optimized_functions and the feature it implements and it only adds
unnecessary complexity to the code.
Signed-off-by: Paweł Dziepak <pdziepak@quarnos.org>
* Removes default mapping of a portion of the RAM (will be done
as needed)
* Passes on the page directory area to kernel, so on early vm init
the kernel can use the area for pagetable allocation.
* Leaves it to the platform to pass in physical memory range(s). This
will ultimately come from FDT.
* Fix long standing issue with allocation of the heap, potentially
causing other part of the bootloader to overwrite the heap.
* Implements pagetable allocator in kernel for early vm mapping.
This fixes the first PANIC seen, we now just get the same one later
on when the VM is up... more to come...
This reverts commit 3fbb24680c.
As I mentioned in #11131, this fix is not correct, and works around
the problem. The real reason was that arch_debug_call_with_fault_handler
was not working properly, so the fault handler went crazy.
With commit eb92810 that is fixed so this can be reverted.
If GCC knows what these functions are actually doing the resulting
code can be optimized better what is especially noticeable in case of
invocations of atomic_{or,and}() that ignore the result. Obviously,
everything is inlined what also improves performance.
Signed-off-by: Paweł Dziepak <pdziepak@quarnos.org>
When an ARMv7 CPU is detected, immediately turn on the FPU. This allows
us to use vsnprintf in the TRACE call in that function, as our libc is
compiled with floating point support and will trigger a fault if the FPU
is not available.
This lets the boot go further, and crash in mmu_init. Next steps:
* Find why mmu_init is crashing
* Setup some fault handlers, otherwise we call uboot ones, and they are
not very helpful. They will also probably not work once the mmu is
enabledvery helpful. They will also probably not work once the mmu is
enabledvery helpful. They will also probably not work once the mmu is
enabled...
This patch makes it possible to inline rdmsr and wrmsr instruction. The
performance impact shouldn't be significant since they are used relatively
rarely and wrmsr is usually a serializing instruction, but there is no reason
not to do so.
The goal of this patch is to amortize the cost of context switch by making
the compiler aware that context switch clobbers all registers. Because all
register need to be saved anyway there is no additional cost of using
callee saved register in the function that does the context switch.
Similarly to previous patch regarding GDT this is mostly a rewrite of
IDT handling code from C to C++. Thanks to constexpr IDT is now entirely
generated at compile-time.
Virtually no functional change, just rewriting the code from
"C in *.cpp files" to C++. Use of constexpr may be advantageous but
that code is not performance critical anyway.
* Instead of forcing the hash-table to use a copy of the key,
introduce and use TypeOperation template to avoid taking a
reference of a reference type (which gcc2 doesn't allow).
For potential boot volumes with older packages states the respective
item in the boot volume menu now has a sub menu for selecting a state.
The boot loader functionality for this feature is complete -- i.e. the
respective kernel is loaded and the name of the old state is added to
the kernel args -- but kernel packagefs and package daemon support is
still missing.
After load_image() the child thread is suspended and the parent is
expected to resume it later. However, it is possible that the parent
attempts to resume its child after it has been notified that the image
had been loaded but before the child managed to suspend itself. In such
case the child would suspends itself after that wake up attempt and,
consequently will not be ever resumed.
To mitigate that problem flag Thread::going_to_suspend has been added
which helps synchronizing thread suspension and continuation in a similar
way that "traditional" thread blocking is performed. This means that
the child should behave in a following manner: set its going_to_suspend flag,
notify the parent (i.e. any thread that may want to resume it), acquire
its scheduler_lock and suspend itself if the going_to_suspend flag is set.
The parent should follow pattern: clear going_to_suspend flag of the thread
that is about to be resumed, acquire that thread scheduler_lock and enqueue
it in a run queue if it is suspended.
Thanks Oliver for reporting the bug and identifying what causes it.
Most of the actual UserEvent work is done in DPC so that we don't have
to care about the limitations of the context in which UserEvent::Fire()
is invoked. This requires appropriate management of lifetime of UserEvent
instances to make sure that DoDPC() method is always called on a valid
object.
* Add isb just because.
* pdziepak pointed out that ARMv5 and before
had different barrier support.
* pdziepak also mentioned that dsb was too strong
for __sync_synchronize
* On ARMv6 or older, we do a simulated dsb.
* Move __sync_synchronize into thread.c in libroot
and use the new arch_atomic.h dsb/dmb defines.
* Gets arm @bootstrap-raw to end of bootstrap.
* Don't assume verdex as it isn't clear this was
occurring.
* Make an educated guess on HAIKU_BOOT_PLATFORM
based on provided board (but still allow it to
be overridden)
* Error out if user doesn't populate
HAIKU_BOOT_PLATFORM or enters an unknown board
name.
* You need to add "-sHAIKU_BOOT_BOARD=xxx" to
your jam to build for the proper ARM device.
* Rename beagle to beagleboneblk as per the
documentation.
* Use atomic_get_and_set for return value
* Atomics are no longer volatile
* Add missing arch_cpu_pause stub
* Move arch_cpu_idle to arch_cpu header to match
other architectures
For non-US keyboards, the extra 102th/105th key is used to reach \. But,
we also need it to report | when shifted (this is the key left to
"enter").
This affects only USB keyboards. Thanks to gordoncjp for reporting!
UserEvent can be fired from scheduler_reschedule() i.e. while holding current
thread scheduler_lock. If the current thread goes sleep and during reschedule
one of its timers sends a signel to it, then scheduler_enqueue_in_run_queue()
attempts to acquire again its scheduler_lock resulting in a deadlock.
There was also a minor issue with both scheduler_reschedule() and
scheduler_enqueue_in_run_queue() acquiring current CPU scheduler mode lock.
* Set max cpu to 1 for PPC until atomic functions are finished
* We have atomic functions inline in the kernel and assembly
code in libroot post-scheduler merge... isn't that a lot of
duplication?
Add boot loader debug menu option "Save syslog from previous session
during boot". If enabled (defaults to true), the previous session's
debug syslog data is copy to a separate buffer and passed to the
kernel, which writes it back to the file /var/log/previous_syslog.
As long as Haiku still boots, this should now be the most convenient way
to retrieve the output from a kernel crash.
Previous implementation based on the actual load of each core and share
each thread has in that load turned up to be very problematic when
balancing load on very heavily loaded systems (i.e. more threads
consuming all available CPU time than there is logical CPUs).
The new approach is to estimate how much load would a thread produce
if it had all CPU time only for itself. Summing such load estimations
of each thread assigned to a given core we get a rank that contains
much more information than just simple actual core load.
This field forces kernel to track each CPU load all the time. It is not
a problem with the current scheduler on a multicore systems, but on
single core machnies or with any other future scheduler this field may
become just an unnecessary burden. It isn't difficult for an application
to compute CPU load by itself when it needs it.
atomic_{get, set}64() are problematic on architectures without 64 bit
compare and swap.
Also, using sequential lock instead of atomic access ensures that
any reads from cpu_ent::active_time won't require any writes to shared
memory.
The client code is not supposed to change the topology info.
It would be also nice if cpu_topology_node::children was an array of
pointers to const but that would require several const_casts in the
topology tree generation code so it's probably not worth it.
Apparently, reading from dr3 is slower than reading from memory
with cache hit.
Also, depending on hypervisor configuration, accessing dr3 may cause
a VM exit (and, at least on kvm, it does), what makes it much slower
than a memory access even when there is a cache miss.
Add get_safemode_option_early() and get_safemode_boolean_early() to get
safemode options before the kernel heap has been initialized. They use a
simplified parser.
* VMTranslationMap:
- Add DebugPrintMappingInfo(): Given a virtual address it is supposed
to print the paging structure information for that address. To be
implemented by derived classes.
- Add DebugGetReverseMappingInfo(): Given a physical addresss it is
supposed to find all virtual addresses mapped to it. To be
implemented by derived classes.
* X86VMTranslationMapPAE: Implement the new methods
DebugPrintMappingInfo() and DebugGetReverseMappingInfo().
* Add KDL command "mapping". It supports both virtual address lookups
and reverse lookups.
* VMAddressSpace: Add randomizingEnabled property.
* VMUserAddressSpace: Randomize addresses only when randomizingEnabled
property is set.
* create_team_arg(): Check, if the team's environment contains
"DISABLE_ASLR=1". Set the team's address space property
randomizingEnabled accordingly in load_image_internal() and
exec_team().
* Create new interface for cpuidle modules (similar to the cpufreq
interface)
* Generic cpuidle module is no longer needed
* Fix and update Intel C-State module
It's a browser for the system package content, where entries can be
selected to blacklist them. The selected entries are removed from the
packagefs instance in the boot loader, so that e.g. selected drivers
won't be picked up. The paths are also added to the safe mode driver
settings and will be interpreted when the system packagefs instance is
mounted by the kernel.
* Make Menu and MenuItem polymorphic.
* MenuItem:
- Make SetMarked() virtual, so it can be overridden.
- Add SetSubmenu() and Supermenu().
- Delete the submenu in the destructor.
* Menu:
- Add Entered()/Exited() hooks. They frame the time the user navigates
the menu or any of its submenus. The hooks allow for subclasses
populating their item list dynamically.
- Add SortItems().
* Update boot loader menu copyright text to include 2013, now that it is
over soon. :-)
* pin idle threads to their specific CPUs
* allow scheduler to implement SMP_MSG_RESCHEDULE handler
* scheduler_set_thread_priority() reworked
* at reschedule: enqueue old thread after dequeueing the new one
* Thread::scheduler_lock protects thread state, priority, etc.
* sThreadCreationLock protects thread creation and removal and list of
threads in team.
* Team::signal_lock and Team::time_lock protect list of threads in team
as well.
* Scheduler uses its own internal locking.
* The UNMAP command is theoretically much faster, as it can get many block
ranges instead of just a single range.
* Furthermore, the ATA TRIM command resembles it much better.
* Therefore, fs_trim_data now gets an array of ranges, and we use SCSI UNMAP
to trim.
* Updated BFS code to collect array ranges to fully support the new
fs_trim_data possibilities.
* No need for the atomically changed variables to be declared as
volatile.
* Drop support for atomically getting and setting unaligned data.
* Introduce atomic_get_and_set[64]() which works the same as
atomic_set[64]() used to. atomic_set[64]() does not return the
previous value anymore.
The flag main purpose is to avoid race conditions between event handler
and cancel_timer(). However, cancel_timer() is safe even without
using gSchedulerLock.
If the event is scheduled to happen on other CPU than the CPU that
invokes cancel_timer() then cancel_timer() either disables the event
before its handler starts executing or waits until the event handler
is done.
If the event is scheduled on the same CPU that calls cancel_timer()
then, since cancel_timer() disables interrupts, the event is either
executed before cancel_timer() or when the timer interrupt handler
starts running the event is already disabled.
* Replace ports list mutex with R/W-lock.
* Move team port list protection to separate array of mutexes.
Relieve contention on sPortsLock by removing Team::port_list from its
protected items. With this, set_port_owner() only needs to acquire the
sPortsLock for reading.
* Add another hash table holding the ports by name. Used by find_port()
so it doesn't have to iterate over the list anymore.
* Use slab-based memory allocator for port messages. sPortQuotaLock was
acquired on every message send or receive and was thus another point
of contention. The lock is not necessary anymore.
* Lock for port hashes and Port::lock are no longer locked in a nested
fashion to reduce chances of blocking other threads.
* Make operations concurrency-safe by adding an atomically accessed
Port::state which provides linearization points to port creation and
deletion. Both operations are now divided into logical and physical
parts, the logical part just updating the state and the physical part
adding/remove it to/from the port hash and team port list.
* set_port_owner() is the only remaining function which still locks
Port::lock and one or two of sTeamListLock[] in a nested fashion.
Since it needs to move the port from one team list to another and
change Port::owner, there's no way around.
* Ports are now reference counted to make accesses to already-deleted
ports safe.
* Should fix#8007.
Simple scheduler behaves exactly the same as affine scheduler with a
single core. Obviously, affine scheduler is more complicated thus
introduces greater overhead but quite a lot of multicore logic has been
disabled on single core systems in the previous commit.
There is a global heap of cores, where the key is the highest priority
of threads running on that core. Moreover, for each core there is
a heap of logical processors on this core where the key is the priority
of currently running thread.
The per-core heap is used for load balancing among logical processors
on that core. The global heap is used in initial decision where to put
the thread (note that the algorithm that makes this decision is not
complete yet).
Simple scheduler is used when we do not have to worry about cache affinity
(i.e. single core with or without SMT, multicore with all cache levels
shared).
When we replace gSchedulerLock with more fine grained locking affine
scheduler should also be chosen when logical CPU count is high (regardless
of cache).
In SMP systems simple scheduler will be used only when all logical
processors share all levels of cache and the number of CPUs is low.
In such systems we do not have to care about cache affinity and
the contention on the lock protecting shared run queue is low. Single
run queue makes load balancing very simple.
Kernel support for yielding to all (including lower priority) threads
has been removed. POSIX sched_yield() remains unchanged.
If a thread really needs to yield to everyone it can reduce its priority
to the lowest possible and then yield (it will then need to manually
return to its prvious priority upon continuing).
Each thread has its minimal priority that depends on the static priority.
However, it is still able to starve threads with even lower priority
(e.g. CPU bound threads with lower static priority). To prevent this
another penalty is introduced. When the minimal priority is reached
penalty (count mod minimal_priority) is added, where count is the number
of time slices since the thread reached its minimal priority. This prevents
starvation of lower priorirt threads (since all CPU bound threads may have
their priority temporaily reduced to 1) but preserves relation between
static priorities - when there are two CPU bound threads the one with
higher static priority would get more CPU time.
This adds the -mapcs-frame compiler flag for ARM to have "stable"
stack frames, adds support to the kernel for dumping stack crawls,
and initial support for iframes. There' much more functionality
to unlock in KDL, but this makes debugging already a lot more
comfortable.....
Since both platforms can boot the same kernel we must accept either
arg, so we make sure they are identical for now.
TODO: use a union or KMessage maybe?
* Mostly useful for virtualization at the moment. Works in QEmu.
* Can be enabled by safemode settings/menu.
* Please note that x2APIC normally requires use of VT-d interrupt remapping feature
on real hardware, which we don't support yet.
- Instead of implicitly registering and unregistering a service
instance on construction/destruction, DefaultNotificationService
now exports explicit Register()/Unregister() calls, which subclasses
are expected to call when they're ready.
- Adjust all implementing subclasses. Resolves an issue with deadlocks
when booting a DEBUG=1 build.
* Add "bool kernel" parameter to vfs_entry_ref_to_path(), so it can be
specified for which I/O context the entry ref shall be translated.
* _user_entry_ref_to_path(): Use the calling team's I/O context instead
of the kernel's. Fixes the bug that in a chroot the syscall would
return a path for outside the chroot.
Currently there are two generators. The fast one is the same one the scheduler
is using. The standard one is the same algorithm libroot's rand() uses. Should
there be a need for more cryptographically PRNG MD4 or MD5 might be a good
candidates.
Placing commpage and team user data somewhere at the top of the user accessible
virtual address space prevents these areas from conflicting with elf images
that require to be mapped at exact address (in most cases: runtime_loader).
This patch introduces randomization of commpage position. From now on commpage
table contains offsets from begining to of the commpage to the particular
commpage entry. Similary addresses of symbols in ELF memory image "commpage"
are just offsets from the begining of the commpage.
This patch also updates KDL so that commpage entries are recognized and shown
correctly in stack trace. An update of Debugger is yet to be done.
Set execute disable bit for any page that belongs to area with neither
B_EXECUTE_AREA nor B_KERNEL_EXECUTE_AREA set.
In order to take advanage of NX bit in 32 bit protected mode PAE must be
enabled. Thus, from now on it is also enabled when the CPU supports NX bit.
vm_page_fault() takes additional argument which indicates whether page fault
was caused by an illegal instruction fetch.
The physical memory map area was not included in the kernel virtual
address space range (it was below KERNEL_BASE). This caused problems
if an I/O operation took place on physical memory mapped there (the
bad address error seen in #9547 was occurring in lock_memory_etc()).
Changed KERNEL_BASE and KERNEL_SIZE to cover the area and add a null
area that covers all of it. Also changed X86VMTranslationMap64Bit to
handle large pages in Query(), as the physical map area uses large
pages.
* Added the aforementioned functions.
* create_area_etc() now takes a guard size parameter.
* The thread_info::stack_base/end range now refers to the usable range
only.
Since we're using multi-part uImage format, we can add the FDT as
a seperate "blob" in the uImage, if the used U-Boot version is not
"FDT enabled".
This is used for example for our Verdex target. Currently I've got
a local hack in the platform/u-boot/Jamfile, looking into pulling
in the FDT files and a proper Jam setup to do that properly...
This detects everything up to ARMv6 right now. Need to check more
recent ARM ARMs for ARMv7 detection.
The detected details get passed on to the kernel, which can use
the pre-detected info for selecting right pagetable format and such.
Copyright removal of Axel done after agreement with Axel @ BeGeistert
that for files that were copy/pasted from x86 arch and then fully
replaced the implementation, removal of original copyright holder is
allowed, since their actual code is gone ;)
This is to make sure all ARM platforms will benefit from planned work on this
MMU/CPU code. The less code duplicated, the better.
Compile-tested for all supported ARM platforms
This also implements the fault handler correctly now, and cleans up the
exception handling. Seems a lot more stable now, no unexpected panics or
faults happening anymore.
* The only implementation that would accept more than 2 TB was the one in
scsi_disk. But even that one was limited to 63 TB.
* Now there is a new utility function devfs_compute_geometry_size() which
does it correctly for sizes up to 2^64 which should be good enough for
quite some time :-)
* This fixes bug #8992.
* For now let's include the same fields in platform_kernel_args
than in the OF version.
* This allows linking the kernel.
Later on we should allow supporting more than a single boot platform,
to have a single kernel per arch.
The lowest 4 bits of the MSR serves as a hint to the hardware to
favor performance or energy saving. 0 means a hint preference for
highest performance while 15 corresponds to the maximum energy
savings. A value of 7 translates into a hint to balance performance
with energy savings.
The default reset value of the MSR is 0. If BIOS doesn't intialize
the MSR, the hardware will run in performance state. This patch
initialize the MSR with value of 7 for balance between performance
and energy savings
Signed-off-by: Fredrik Holmqvist <fredrik.holmqvist@gmail.com>
Renamed {32,64}/int.cpp to {32,64}/descriptors.cpp, which now contain
functions for GDT and TSS setup that were previously in arch_cpu.cpp,
as well as the IDT setup code. These get called from the init functions
in arch_cpu.cpp, rather than having a bunch of ifdef'd chunks of code
for 32/64.
Reused x86 arch_user_debugger.cpp, with a few minor changes to make
the code work for both 32 and 64 bit. Something isn't quite working
right, if a breakpoint is hit the kernel will hang. Other than that
everything appears to work correctly.
* Changed IS_USER_ADDRESS to check an address using USER_BASE and
USER_SIZE, rather than just !IS_KERNEL_ADDRESS. The old check would
allow user buffers to point into the physical memory map area.
* Added an unmapped hole at the end of the bottom half of the address
space which catches buffers that cross into the uncanonical address
region. This also removes the need to check for uncanonical return
addresses in the syscall handler, it is no longer possible for the
return address to be uncanonical under normal circumstances. All
cases in which the return address might be changed by the kernel
are still handled via the IRET path.
The cookie is used to store the base address of the area that was just
visited. On 64-bit systems, int32 is not sufficient. Therefore, changed
to ssize_t which retains compatibility on x86 while expanding to a
sufficient size on x86_64.
Userland switch is implemented, as is basic system call support (using
SYSCALL/SYSRET). The system call handler is not yet complete: it doesn't
handle more than 6 arguments, and does not perform all the necessary kernel
entry/exit work (neither does the interrupt handler). However, this is
sufficient for runtime_loader to start and print some debug output.
Since this argument may be used to pass pointers, uint32 is not
correct for 64-bit. Effectively no change on 32-bit targets, both
size_t and uint32 are unsigned long there.
No major changes to the kernel: just compiled in arch_smp.cpp and fixed the
IDT load in arch_cpu_init_percpu to use the correct limit for x86_64 (uses
sizeof(interrupt_descriptor)). In the boot loader, changed smp_boot_other_cpus
to construct a temporary GDT and get the page directory address from CR3, as
what's in kernel_args will be 64-bit stuff and will not work to switch the
CPUs into 32-bit mode in the trampoline code. Refactored 64-bit kernel entry
code to not use the stack after disabling paging, as the secondary CPUs are
given a 32-bit virtual stack address by the SMP trampoline code which will
no longer work.
A proper page fault handler was required for areas that were not locked
into the kernel address space. This enables the boot process to get
up to the point of trying to find the boot volume.
* Thread creation and switching is working fine, however threads do not yet
get interrupted because I've not implemented hardware interrupt handling
yet (I'll do that next).
* I've made some changes to struct iframe: I've removed the e/r prefixes
from the member names for both 32/64, so now they're just named ip, ax,
bp, etc. This makes it easier to write code that works with both 32/64
without having to deal with different iframe member names.
This has been done by adding typedefs in elf_common.h to the correct ELF
structures for the architecture, and changing all Elf32_* uses to those
types. I don't know whether image loading works as I cannot test it yet,
there may be some 64-bit safety issues around. However, symbol lookup for
the kernel is working correctly.
* Uses 64-bit multiplication, special handling for CPUs clocked < 1 GHz
in system_time_nsecs() not required like on x86.
* Tested against a straight conversion of the x86 version, noticably
faster with a large number of system_time() calls.
* Added empty source files for all the 64-bit paging method code, and a
stub implementation of X86PagingMethod64Bit.
* arch_vm_translation_map.cpp has been modified to use X86PagingMethod64Bit
on x86_64.
* Some things are currently ifndef'd out completely for x86_64 because
they aren't implemented, there's a few other ifdef's to handle x86_64
differences but most of the code works unchanged.
* Renamed some i386_* functions to x86_*.
* Added a temporary method for setting the current thread on x86_64
(a global variable, not SMP safe). This will be changed to be done
via the GS segment but I've not implemented that yet.
