| Rev |
Age |
Author |
Path |
Log message |
Diff |
Changes |
| 3189 |
5855 d 20 h |
jermar |
/trunk/kernel/arch/ppc32/src/ |
Implement memcpy_to_uspace_failover_address for ppc32.
This fixes kernel side of ticket #59. |
|
/trunk/kernel/arch/ppc32/src/asm.S
|
| 3188 |
5857 d 19 h |
jermar |
/trunk/kernel/generic/src/mm/ |
Finer grained waiting for memory with consideration of the number
of available memory frames. |
|
/trunk/kernel/generic/src/mm/frame.c
|
| 3187 |
5857 d 23 h |
jermar |
/trunk/kernel/generic/src/mm/ |
Get rid of the infamous "Sleep not implemented" panic during low memory
condition. Use a condition variable to implement the sleep. As of now, the
condition is rather coarse and can cause that threads may sleep unnecessarily.
Needs to be fine tuned and based on the amount of memory available. |
|
/trunk/kernel/generic/src/mm/frame.c
|
| 3186 |
5857 d 23 h |
jermar |
/trunk/kernel/ |
Support for active mutexes. Active mutexes implement busy waiting, pretty much
in the same way as spinlocks, but can be passed to condition variables, which is
the motivation for this enhancement. |
|
/trunk/kernel/genarch/src/mm/as_ht.c
/trunk/kernel/generic/include/synch/mutex.h
/trunk/kernel/generic/src/ipc/ipc.c
/trunk/kernel/generic/src/mm/as.c
/trunk/kernel/generic/src/proc/task.c
/trunk/kernel/generic/src/synch/condvar.c
/trunk/kernel/generic/src/synch/mutex.c
/trunk/kernel/generic/src/synch/rwlock.c
/trunk/kernel/test/mm/slab2.c
|
| 3185 |
5858 d 22 h |
jermar |
/trunk/kernel/generic/src/mm/ |
cstyle for frame.c |
|
/trunk/kernel/generic/src/mm/frame.c
|
| 3184 |
5859 d 2 h |
jermar |
/trunk/kernel/generic/src/ipc/ |
Fix a scary two-year old bug: do not attempt to initialize a call structure if
its allocation failed. |
|
/trunk/kernel/generic/src/ipc/ipc.c
|
| 3183 |
5859 d 3 h |
jermar |
/trunk/kernel/generic/src/mm/ |
Avoid easy deadlock while printing slab cache statistics. |
|
/trunk/kernel/generic/src/mm/slab.c
|
| 3182 |
5859 d 3 h |
jermar |
/trunk/kernel/generic/include/mm/ |
cstyle for slab.h |
|
/trunk/kernel/generic/include/mm/slab.h
|
| 3181 |
5860 d 7 h |
jermar |
/trunk/kernel/generic/src/ |
cstyle for kinit.c and scheduler.c |
|
/trunk/kernel/generic/src/main/kinit.c
/trunk/kernel/generic/src/proc/scheduler.c
|
| 3180 |
5860 d 9 h |
jermar |
/trunk/kernel/generic/src/mm/ |
cstyle for slab.c |
|
/trunk/kernel/generic/src/mm/slab.c
|
| 3179 |
5861 d 0 h |
jermar |
/trunk/kernel/ |
Introduce the notion of FPU_NO_CFLAGS and separate building of test/fpu
objects and building the rest of kernel objects. Now, amd64 and ia32 can
enable things such as SSE selectively for the test/fpu/*. |
|
/trunk/kernel/Makefile
/trunk/kernel/arch/amd64/Makefile.inc
/trunk/kernel/arch/ia32/Makefile.inc
|
| 3178 |
5861 d 22 h |
jermar |
/trunk/boot/arch/ia64/loader/gefi/ |
Remove the ia64 loader's dependency on libgcc. The dependecny would break
builds with the new toolchain. |
|
/trunk/boot/arch/ia64/loader/gefi/HelenOS/Makefile
/trunk/boot/arch/ia64/loader/gefi/apps/Makefile
|
| 3177 |
5862 d 0 h |
jermar |
/trunk/contrib/toolchain/ |
Upgrade the toolchain to gcc 4.3.1.
This revealed several bugs fixed in a couple of previous commits.
The ia64 toolchain no longer requires the extra includes.
The ppc32 architecture feels a little bit broken because the console
service doesn't start for some reason. |
|
/trunk/contrib/toolchain/toolchain.amd64.sh
/trunk/contrib/toolchain/toolchain.arm32.sh
/trunk/contrib/toolchain/toolchain.ia32.sh
/trunk/contrib/toolchain/toolchain.ia64.sh
/trunk/contrib/toolchain/toolchain.mipseb32.sh
/trunk/contrib/toolchain/toolchain.mipsel32.sh
/trunk/contrib/toolchain/toolchain.ppc32.sh
/trunk/contrib/toolchain/toolchain.ppc64.sh
/trunk/contrib/toolchain/toolchain.sparc64.sh
|
| 3176 |
5862 d 0 h |
jermar |
/trunk/contrib/conf/ |
Remove outdated sample configuration files or configuration files that
are strongly inspired by Simics configuration files. |
|
/trunk/contrib/conf/HelenOS.amd64.simics
/trunk/contrib/conf/HelenOS.ppc32.simics
/trunk/contrib/conf/HelenOS.sparc64.simics
/trunk/contrib/conf/SPMIPS.simics
/trunk/contrib/conf/dot.bochsrc
/trunk/contrib/conf/simics.conf
/trunk/contrib/conf/spmips.conf
/trunk/contrib/conf/vmware.conf
|
| 3175 |
5862 d 5 h |
jermar |
/trunk/kernel/generic/src/main/ |
Don't print anything before the kernel initializes the first THE
structure. This fixes a regression introduced in r3061. If anything
worked from r3061 until now, it was only due to luck. arm32 was not
that lucky when built with gcc 4.3.1. |
|
/trunk/kernel/generic/src/main/main.c
|
| 3173 |
5862 d 20 h |
jermar |
/trunk/kernel/arch/ia32/ |
Don't build the ia32 kernel with MMX/SSE/3DNOW enabled. This was causing
problems with gcc 4.3.1 which generated MMX aided code in a context that
didn't allow this. |
|
/trunk/kernel/arch/ia32/Makefile.inc
|
| 3172 |
5862 d 20 h |
jermar |
/trunk/kernel/ |
Frame pointers will make debugging much easier. |
|
/trunk/kernel/Makefile
|
| 3167 |
5863 d 8 h |
svoboda |
/trunk/kernel/arch/mips32/include/ |
Fix mips32 atomic_add() which would break build unless a constant was propagated into the second argument. (Specifically at -O0) |
|
/trunk/kernel/arch/mips32/include/atomic.h
|
| 3165 |
5863 d 20 h |
jermar |
/trunk/kernel/test/avltree/ |
Fix an off-by-one bug in avltree1 test and also declare an unused static
function as used. These two bugs were breaking building by gcc 4.3.1. |
|
/trunk/kernel/test/avltree/avltree1.c
|
| 3164 |
5864 d 20 h |
jermar |
/trunk/ |
First argument of atomic functions is read-write (ia32). |
|
/trunk/kernel/arch/ia32/include/atomic.h
/trunk/uspace/lib/libc/arch/ia32/include/atomic.h
|
| 3163 |
5864 d 20 h |
jermar |
/trunk/ |
First argument to atomic functions is read-write (amd64). |
|
/trunk/kernel/arch/amd64/include/atomic.h
/trunk/uspace/lib/libc/arch/amd64/include/atomic.h
|
| 3162 |
5864 d 20 h |
jermar |
/trunk/ |
The first argument to atomic_add() is read-write. (mips32) |
|
/trunk/kernel/arch/mips32/include/atomic.h
/trunk/uspace/lib/libc/arch/mips32/include/atomic.h
|
| 3147 |
5866 d 6 h |
jermar |
/trunk/boot/arch/ |
Add TODO comments to arm32 and mips32 loaders saying that caches must be put
into the coherent state before passing control to the kernel. |
|
/trunk/boot/arch/arm32/loader/boot.S
/trunk/boot/arch/mips32/loader/asm.S
|
| 3146 |
5866 d 7 h |
jermar |
/trunk/boot/arch/sparc64/loader/ |
Don't make any assumptions about the contents of the I-cache in sparc64
loader and before passing control to the kernel, invalidate the I-cache. |
|
/trunk/boot/arch/sparc64/loader/asm.S
/trunk/boot/arch/sparc64/loader/ofwarch.c
|
| 3145 |
5866 d 9 h |
jermar |
/trunk/kernel/arch/sparc64/include/ |
On sparc64, when the operand to the FLUSH instruction doesn't matter, the
instruction's semantics is to flush the pipeline. |
|
/trunk/kernel/arch/sparc64/include/barrier.h
/trunk/kernel/arch/sparc64/include/mm/tlb.h
|
| 3144 |
5866 d 21 h |
jermar |
/trunk/boot/arch/sparc64/loader/ |
Issue a sparc64 write barrier before passing control from the loader to
the kernel. |
|
/trunk/boot/arch/sparc64/loader/asm.S
|
| 3143 |
5867 d 1 h |
svoboda |
/trunk/kernel/ |
Implement smc_coherence[_block]() for ppc32 |
|
/trunk/kernel/arch/ppc32/include/barrier.h
/trunk/kernel/generic/src/mm/backend_elf.c
|
| 3142 |
5867 d 2 h |
svoboda |
/trunk/kernel/generic/src/mm/ |
Fix build when smc_coherence_block() is defined as an empty macro |
|
/trunk/kernel/generic/src/mm/backend_elf.c
|
| 3141 |
5867 d 5 h |
jermar |
/trunk/kernel/generic/src/mm/ |
Maintain cache coherence when the ELF backend duplicates a frame from an
executable segment. |
|
/trunk/kernel/generic/src/mm/backend_elf.c
|
| 3140 |
5867 d 5 h |
jermar |
/trunk/kernel/arch/ia64/include/ |
Fix operands for fc_i(). |
|
/trunk/kernel/arch/ia64/include/barrier.h
|
| 3139 |
5867 d 5 h |
jermar |
/trunk/kernel/arch/mips32/src/ |
Maintain cache coherence when setting breakpoints in kconsole on mips32. |
|
/trunk/kernel/arch/mips32/src/debugger.c
|
| 3138 |
5867 d 6 h |
jermar |
/trunk/kernel/arch/amd64/src/ |
Formatting fixes for amd64 debugger. |
|
/trunk/kernel/arch/amd64/src/debugger.c
|
| 3137 |
5867 d 6 h |
jermar |
/trunk/kernel/generic/src/proc/ |
Maintain cache coherence after sys_task_spawn() copies ELF image from user space. |
|
/trunk/kernel/generic/src/proc/task.c
|
| 3136 |
5867 d 6 h |
jermar |
/trunk/kernel/arch/mips32/src/ |
Maintain cache coherence after mips32 install exception handlers. |
|
/trunk/kernel/arch/mips32/src/mips32.c
|
| 3135 |
5867 d 6 h |
jermar |
/trunk/kernel/arch/arm32/src/ |
Maintain cache coherence after arm32 installs exception handlers. |
|
/trunk/kernel/arch/arm32/src/exception.c
|
| 3134 |
5867 d 7 h |
jermar |
/trunk/kernel/arch/ |
Add smc_coherence_block(). |
|
/trunk/kernel/arch/arm32/include/barrier.h
/trunk/kernel/arch/ia32/include/barrier.h
/trunk/kernel/arch/ia64/include/barrier.h
/trunk/kernel/arch/mips32/include/barrier.h
/trunk/kernel/arch/ppc32/include/barrier.h
/trunk/kernel/arch/ppc64/include/barrier.h
/trunk/kernel/arch/sparc64/include/barrier.h
|
| 3133 |
5867 d 21 h |
jermar |
/trunk/kernel/arch/ |
Add smc_coherence() macro to all architectures.
So far, only amd64, ia32, ia64 and sparc64 are implemented. |
|
/trunk/kernel/arch/sparc64/include/mm/cache_spec.h
/trunk/kernel/arch/arm32/include/barrier.h
/trunk/kernel/arch/ia32/include/barrier.h
/trunk/kernel/arch/ia64/include/barrier.h
/trunk/kernel/arch/mips32/include/barrier.h
/trunk/kernel/arch/ppc32/include/barrier.h
/trunk/kernel/arch/ppc64/include/barrier.h
/trunk/kernel/arch/sparc64/include/barrier.h
/trunk/kernel/arch/sparc64/include/mm/tlb.h
/trunk/kernel/arch/sparc64/src/mm/cache.S
|
| 3130 |
5869 d 10 h |
jermar |
/trunk/kernel/generic/src/mm/ |
Add a comment explaining the reason for a memory barrier after page table update in detail. |
|
/trunk/kernel/generic/src/mm/page.c
|
| 3128 |
5869 d 23 h |
jermar |
/trunk/kernel/generic/src/mm/ |
Issue a memory barrier after modifying a page table mapping in order to
prevent memory prefetching. |
|
/trunk/kernel/generic/src/mm/page.c
|
| 3120 |
5874 d 2 h |
jermar |
/trunk/ |
Simple support for cscope. |
|
/trunk/Makefile
|