Brian Anderson
edf6e1ec0e
rt: Zero the bottom frame's return address and base pointer
...
My reading of libunwind leads me to believe this is expected.
Closes #1322
2011-12-19 18:47:39 -08:00
Brian Anderson
1bfc4e2e6c
rt: Save and restore %rax/%eax in __morestack
...
This doesn't matter now since we use an out pointer for return
values but it's sure to show up mysteriously someday.
2011-12-19 18:16:21 -08:00
Brian Anderson
d2e7336a01
rt: Remove unused saves from __morestack
2011-12-19 16:24:10 -08:00
Brian Anderson
50dc80f617
rt: Give upcall_new_stack the same convention as other upcalls
2011-12-19 16:24:10 -08:00
Brian Anderson
bedcaad97e
rt: Give upcall_del_stack the same convention as other upcalls
2011-12-19 16:24:10 -08:00
Brian Anderson
1a1fdf34b1
rt: Add lots of documentation to __morestack
2011-12-18 02:02:35 -08:00
Brian Anderson
471b4d6e58
rt: Use a DWARF CFI scheme that works on mac in __morestack
2011-12-17 14:10:44 -08:00
Brian Anderson
7d51961b6f
rt: Fix saved address of unwinding %rsp in __morestack
...
I haven't thought too deeply about this, but I think I was telling the
unwinder to use the stack pointer for the wrong frame when unwinding. Not sure
how that could have worked at all, but this results in the correct alignment
for cleanups.
2011-12-16 18:18:43 -08:00
Brian Anderson
121c420184
rt: Insert stack alignment checks into upcalls
2011-12-16 18:18:43 -08:00
Brian Anderson
1c1bc2f1cd
rt: Remove some duplicated code from 64-bit __morestack
2011-12-12 16:58:21 -08:00
Brian Anderson
887d970fee
rt: Make unwinding through __morestack work on mac
...
Had to bump the min stack size in some of the tests from 256 bytes
to 1024 bytes. Not sure why yet.
2011-12-12 16:15:29 -08:00
Brian Anderson
3e32f6623c
rt: Write CFI instructions that (might) work on mac in __morestack
...
The DW_CFA_val_offset_sf instruction doesn't seem to work on mac,
even after implementing it in the llvm-mc assembler, so now
I'm looking for a different way to communicate the stack pointer
adjustment to the unwinder.
2011-12-12 16:14:17 -08:00
Brian Anderson
7a8e73662d
rt: Make unwinding through __morestack work on mac
...
Had to bump the min stack size in some of the tests from 256 bytes
to 1024 bytes. Not sure why yet.
2011-12-11 18:49:04 -08:00
Brian Anderson
7bc34f63d8
rt: Write CFI instructions that (might) work on mac in __morestack
...
The DW_CFA_val_offset_sf instruction doesn't seem to work on mac,
even after implementing it in the llvm-mc assembler, so now
I'm looking for a different way to communicate the stack pointer
adjustment to the unwinder.
2011-12-11 16:32:36 -08:00
Brian Anderson
5cd08586f6
rt: Add DWARF CFI info to asm_call_on_stack
...
This will allow the unwinder to unwind this function so we can call
upcall_fail on the C stack.
2011-12-11 14:35:09 -08:00
Brian Anderson
4699b6865f
build: Use llvm-mc as our assembler
...
LLVM's assembler understands .cfi pseudo-ops on the mac, which we need in
order to generate frames that can be DWARF-unwound
2011-12-10 18:51:21 -08:00
Brian Anderson
9a738fd61d
rt: Various tweaks to make __morestack unwinding work on linux
...
When unwinding through __morestack the stack limit in the TLS is invalidated
and must be reset. Instead of actually landing at __morestack we're
just going to make all our Rust landing pads call upcall_reset_stack_limit,
which will find the stack segment that corresponds to the current stack
pointer and put the limit in the TLS.
Also massively expand the stack segment red zone to make more room for the
dynamic linker. Will fix in the future.
2011-12-06 16:32:41 -08:00
Austin Seipp
b513a5a500
Make valgrind usage more consistent and less error prone.
...
I was still having issues with the build system somehow getting confused
as to which set of valgrind headers to use when compiling rt.
This commit moves all the valgrind headers into their own directory
under rt and makes the usage more consistent. The compiler is now passed
the -DNVALGRIND flag when valgrind is not installed, as opposed to
passing -DHAVE_VALGRIND.
We also pass -I src/rt to the compiler when building rt so you can more
easily import what you want. I also cleaned up some erroneous #includes
along the way.
It should be safe to always just import the local valgrind headers and use
them without question. NVALGRIND turns the operations to no-ops when it
is active, and the build and tests run cleanly with or without.
2011-12-06 01:15:29 -06:00
Brian Anderson
3b8bfaf534
rt: Use an out pointer for rust_new_stack
...
upcall_call_shim_on_c_stack does not handle return values
2011-12-05 17:11:35 -08:00
Brian Anderson
c254936f32
rt: Fix __morestack prologue on x86_64 mac
2011-12-02 15:32:51 -08:00
Brian Anderson
691b517fb9
rt: Fix 64-bit linux __morestack
2011-12-01 20:56:05 -08:00
Brian Anderson
3e5b314c0b
rt: Implement __morestack for 64-bit mac. Doesn't unwind yet
2011-12-01 19:44:14 -08:00
Brian Anderson
8bf3d85b9d
rt: Implement record_sp for 64-bit mac
2011-12-01 19:44:13 -08:00
Brian Anderson
f2de42c39a
rt: Reorganize stack growth code
2011-11-30 20:31:33 -08:00
Brian Anderson
dd48a1af58
rt: Remove redundant defs of RUST_DEL_STACK
2011-11-30 20:31:33 -08:00
Brian Anderson
e6ef4d929c
rt: Remove the stack pointer field of stk_seg
2011-11-30 20:31:33 -08:00
Brian Anderson
408d4ec0ef
rt: Remove upcall_alloc_c_stack/call_c_stack, et. al.
...
We are using upcall_call_shim_on_c_stack now
2011-11-30 20:31:32 -08:00
Brian Anderson
8cbba79208
rt: Fix the name of the linux macro in x86_64/morestack.S
2011-11-29 13:50:34 -08:00
Brian Anderson
afb8f01741
rt: Fix calculation of stack args location in x86_64/morestack.S
2011-11-29 13:47:54 -08:00
Brian Anderson
a9967423f8
rt: Make 64-bit __morestack grow and shrink properly
2011-11-28 21:57:52 -08:00
Brian Anderson
5461c61af2
rt: Remove a FIXME frome x86_64/morestack.S
2011-11-28 20:14:12 -08:00
Brian Anderson
ef20afc36c
rt: Get 64-bit Linux __morestack allocating new stacks
2011-11-28 18:39:18 -08:00
Brian Anderson
3a80d7b0e3
rt: Set hidden visibility on __morestack
...
Each crate will have its own copy of this function, which should not be
exported.
2011-11-28 18:01:39 -08:00
Brian Anderson
cd75c9ce11
rt: Implement part of the 64-bit __morestack
2011-11-28 16:29:52 -08:00
Brian Anderson
020d8f12a0
rt: Implement record_sp for 64-bit linux
2011-11-27 19:00:26 -08:00
Brian Anderson
0580d4e77a
rt: Add x86_64 record_sp stub
2011-11-22 18:02:12 -08:00
Niko Matsakis
9fa44a41e6
get pure wrappers approach running
2011-11-18 16:32:19 -08:00
Niko Matsakis
f9231e932c
add an extra underscore
2011-11-17 12:02:15 -08:00
Brian Anderson
a92699fe58
rt: Make morestack link on x86_64. Still utterly broken
2011-11-17 11:12:59 -08:00
Brian Anderson
d2199e8716
Various fixes for x86_64 on linux
...
Configure LLVM correctly, use the right data layout, add the readlink
function back, fix C constants, etc.
2011-11-16 15:24:56 -08:00
Niko Matsakis
59e2a0a9c1
add shim upcall func
2011-11-16 15:16:40 -08:00
Niko Matsakis
7baf68b0bd
hack around the problem that x86_64 expects first few args in regs.
...
call on c-stack expects all data to be delivered on the stack.
2011-11-09 20:05:49 -08:00
Niko Matsakis
fc064f4410
use uint64_t when adjusting stack ptr
2011-11-09 20:05:49 -08:00
Niko Matsakis
ff5b319ce5
correct stack alignment
2011-11-08 21:14:09 -08:00
Niko Matsakis
1a28ad4fa4
fix bug in loading argument regs
2011-11-08 21:14:09 -08:00
Niko Matsakis
b30a76ed1b
correct calling convention for x86_64
2011-11-08 21:14:09 -08:00
Niko Matsakis
0668eed81e
fix alignment of xmm register storage
2011-11-08 21:11:47 -08:00
Niko Matsakis
b578ed9b2a
add required symbols
2011-11-02 14:14:20 -07:00
Niko Matsakis
0982c7f78c
hastily port so we don't fail to build
2011-11-02 14:14:20 -07:00
Niko Matsakis
20946e6970
start going back to an i386 build
2011-11-02 14:13:22 -07:00
