23c6aa9a54d0a220fb2557197c555b2684c93876 |
27-Oct-2014 |
Christopher Ferris <cferris@google.com> |
Fix aarch64 handling of error code. In the is_signal_frame call, there is an error return path that can return a non-zero value. Unfortunately, only zero is considered an error, so this causes the frame to be handled incorrectly. Once this was fixed, I also modified the code so that if the treating the unwind as in a signal frame fails, we can fall back and try a couple of different paths. It turns out this is_signal_frame check was broken on arm too. Bug: 18100846 Change-Id: I519aea7594352be60df665457eb2986ab4f0f513
is_signal_frame.c
step.c
|
1e549bb0cef88f170cbbbd273212f45cf8f9af43 |
17-Sep-2014 |
Elliott Hughes <enh@google.com> |
Correctly identify the aarch64 kernel's signal trampoline. This was misfiring because it was seeing a "corrected" PC that pointed to the nop before the mov that's the first instruction in the trampoline. Bug: 17436734 (cherry picked from commit 6d90028ee35b7028004e5a4f39ffa55dd0fbd6de) Change-Id: I86abd5c49668e3f6694837b9a780d537c4c7a2f1
is_signal_frame.c
|
60c0e104182996ce3739e3377d1f8f3f706853ca |
09-Jun-2014 |
Christopher Ferris <cferris@google.com> |
Fix ip adjust on aarch64. I didn't add the ip adjust when aarch64 support was added. This meant that the ip pointed to the wrong instruction when doing unwinds. Change-Id: Ie60322b1bd0f4dae95852cc93d163b850595b0b5
step.c
|
a3112726a13e9b4694821314a4cbbd742764c817 |
04-Apr-2014 |
Christopher Ferris <cferris@google.com> |
Fix aarch64 map_local_is_* calls. Also fix one formatting issue in ia64/Ginit.c. Change-Id: Idbb112a438a5c59a0fef2949abb07a48c96e0e94
init.c
|
f4a8df5f4f338f1a12c25213227e98b34b42447f |
08-Mar-2014 |
Christopher Ferris <cferris@google.com> |
Fix maps caching for local processes. It is possible for the map to change while running libunwind. For example, if lots of threads are doing local unwinds at the same time. Allow the cached map to regenerate when it detects this case. Included in this change is a refactor of all of the os code so that it can also be used to do caching. This is a prelude to being able to attempt to push the code upstream. Also, this moves the code back closer to the original upstream code. Hopefully, this will allow me to upstream all of these changes. Change-Id: Ia219fa61e16e36416133bc95b1dd2161bd5b8ff7
init.c
init_local.c
|
1df3245fb2c4d437e930e226331e02e218451db4 |
27-Feb-2014 |
Christopher Ferris <cferris@google.com> |
AARCH64: Change Debug calls to use %lx instead of %x. The attached patch simply modifies the Debug statements for aarch64 to use %lx instead of %x. Tested by compiling for aarch64 using a cross-compiler. Change-Id: I1f05ed759cef2ff4704c955752f0c19fc22dbb40 Signed-off-by: Christopher Ferris <cferris@google.com>
init.c
|
0996405de25a3aab5045d20803c8f0a5441d0c57 |
01-Feb-2014 |
Christopher Ferris <cferris@google.com> |
Fix memory corruption bug. There are two problems, some of the init code needs the map created before the init function since it does a verify check during init. The second problem is that the local addr space seems to change for reasons I don't understand, so it's necessary to init the map list when doing a unw_cursor init. Bug: 12852152 Change-Id: I32246d7c5adf7e24192868cb32a8c2fdc53c12f0
init_local.c
|
16b95a68caaa7e021209e2cd6a877ae1e558f740 |
23-Jan-2014 |
Christopher Ferris <cferris@google.com> |
Add speed-ups to libunwind. Add caching of /proc/self/maps data structures. Add caching of elf image maps attached to the maps data structure. Add a set of new map api functions to handle creation/destruction of the maps by external code. Remove the creation of the maps data structure in the local address space init. Change-Id: I0adae397d433fc3902f83a6f377f6889161937e5
init.c
|
cdf9ee587b78148c5d48dae1b5ea72ec8df64c96 |
23-Nov-2013 |
Christopher Ferris <cferris@google.com> |
Only check the map for local unwinds. Change-Id: I7b7eed29a1b7ddad8e0d852598260809d3eb98d1
init.c
|
7d46a21e0a2cb561e4cad57b101a7137e01023dc |
14-Nov-2013 |
Christopher Ferris <cferris@google.com> |
For current process, check memory accesses. When attempting to unwind the current process, the library simple reads and writes to memory for memory accesses. Unfortunately, if an unwind winds up with invalid data, then the process can segfault. Add a guard to all memory accesses that verifies that reads/writes are valid before attempting them. For now, this guard is a nop on all systems except linux systems. Also, modify the adjust_ip function to use the access_mem function so that all reads go through the right function. Bug: 11652982 Change-Id: I9d1feacae585374f6bf1d899f8a006ec9ecb47c4
init.c
|
efb75a0b84fed700173700a2e4fee153ba491b32 |
01-Oct-2013 |
Christopher Ferris <cferris@google.com> |
Update external/libunwind to build in android. Bug: 8410085 Change-Id: Ideb46602d1e1cf2df5f858ef11d498055c02b969
nit.h
|
ac6c0a6535975f1dc2da6e4e2766614baac2a14a |
11-May-2013 |
Yvan Roux <yvan.roux@linaro.org> |
[PATCH] AArch64 port.
create_addr_space.c
get_proc_info.c
get_save_loc.c
global.c
init.c
init_local.c
init_remote.c
is_signal_frame.c
regs.c
resume.c
step.c
create_addr_space.c
get_proc_info.c
get_save_loc.c
global.c
init.c
init_local.c
init_remote.c
is_signal_frame.c
regs.c
resume.c
step.c
en-offsets.c
nit.h
s_fpreg.c
ffsets.h
egname.c
iglongjmp.S
nwind_i.h
|