Add MIPS support

Add MIPS support

Signed-off-by: Chris Dearman <chris@mips.com>

[Ben Chan]

some drive-by comments

http://breakpad.appspot.com/381002/diff/1/src/client/linux/handler/exception_...
File src/client/linux/handler/exception_handler_unittest.cc (right):

http://breakpad.appspot.com/381002/diff/1/src/client/linux/handler/exception_...
src/client/linux/handler/exception_handler_unittest.cc:183:
__builtin___clear_cache(memory + kOffset, memory + kOffset +
sizeof(instructions));
nit: line break after comma

http://breakpad.appspot.com/381002/diff/1/src/client/linux/handler/exception_...
src/client/linux/handler/exception_handler_unittest.cc:320:
__builtin___clear_cache(memory + kOffset, memory + kOffset +
sizeof(instructions));
nit: line break after comma

http://breakpad.appspot.com/381002/diff/1/src/client/linux/handler/exception_...
src/client/linux/handler/exception_handler_unittest.cc:457:
__builtin___clear_cache(memory + kOffset, memory + kOffset +
sizeof(instructions));
nit: line break after comma

http://breakpad.appspot.com/381002/diff/1/src/google_breakpad/common/minidump...
File src/google_breakpad/common/minidump_cpu_mips.h (right):

http://breakpad.appspot.com/381002/diff/1/src/google_breakpad/common/minidump...
src/google_breakpad/common/minidump_cpu_mips.h:1: /* Copyright (c) 2009, Google
Inc.
2012

http://breakpad.appspot.com/381002/diff/1/src/google_breakpad/common/minidump...
File src/google_breakpad/common/minidump_format.h (right):

http://breakpad.appspot.com/381002/diff/1/src/google_breakpad/common/minidump...
src/google_breakpad/common/minidump_format.h:100: //#define MD_CONTEXT_MIPS 
0x00010000  /* CONTEXT_R4000 (same value as x86?) */
perhaps remove this line instead of commenting it out

[Ted Mielczarek]

A couple of other places you probably want to add MIPS support, off the top of
my head:
http://code.google.com/p/google-breakpad/source/browse/trunk/src/processor/mi...
http://code.google.com/p/google-breakpad/source/browse/trunk/src/processor/mi...

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/handler/excepti...
File src/client/linux/handler/exception_handler_unittest.cc (right):

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/handler/excepti...
src/client/linux/handler/exception_handler_unittest.cc:159: #endif
Might make sense to define this as a file-global constant since you have to
repeat this ifdef over and over.

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/handler/excepti...
src/client/linux/handler/exception_handler_unittest.cc:185: #endif
It's probably true that we need to flush the icache on all platforms. Maybe
there's some quirk that makes it work on x86/x86-64. (I confess to not knowing
the specifics of CPU caches.)

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/handler/excepti...
src/client/linux/handler/exception_handler_unittest.cc:257: }
We could probably factor this out into a helper function. (Actually I think I
even have a patch somewhere that added this to the Minidump APIs.)

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
File src/client/linux/minidump_writer/linux_core_dumper.cc (right):

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
src/client/linux/minidump_writer/linux_core_dumper.cc:103: stack_pointer =
(uint8_t *)info->regs.regs[MD_CONTEXT_MIPS_REG_SP];
Use a C++-style cast.

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
File src/client/linux/minidump_writer/linux_ptrace_dumper.cc (right):

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
src/client/linux/minidump_writer/linux_ptrace_dumper.cc:188: #if
!defined(__ANDROID__) || defined(__mips__)
I recall seeing someone with an Android/MIPS port, so I'm a little wary of
mixing these defines.

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
src/client/linux/minidump_writer/linux_ptrace_dumper.cc:200: #endif
This block could stand a comment, it's a bit inscrutable to me.

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
src/client/linux/minidump_writer/linux_ptrace_dumper.cc:228: stack_pointer =
(uint8_t*)info->regs.regs[MD_CONTEXT_MIPS_REG_SP];
C++-style cast. I'm guessing that MIPS pointers are always 64-bit, so this cast
always works? Still might be nice to stick with the style used on other
architectures.

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
File src/client/linux/minidump_writer/linux_ptrace_dumper_unittest.cc (right):

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
src/client/linux/minidump_writer/linux_ptrace_dumper_unittest.cc:283: #if
!defined(__ARM_EABI__) && !defined(__mips__)
I wonder if we shouldn't stick a "not x86" define somewhere in a header, since
you wound up taking all the same #ifdefs as the ARM code.

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
File src/client/linux/minidump_writer/minidump_writer.cc (right):

http://breakpad.appspot.com/381002/diff/4001/src/client/linux/minidump_writer...
src/client/linux/minidump_writer/minidump_writer.cc:419: out->float_save.fir =
uc->uc_mcontext.fpc_eir; // Unused
Two spaces after the ; before the //. (I get nitpicked for that all the time.)

http://breakpad.appspot.com/381002/diff/4001/src/google_breakpad/common/minid...
File src/google_breakpad/common/minidump_cpu_mips.h (right):

http://breakpad.appspot.com/381002/diff/4001/src/google_breakpad/common/minid...
src/google_breakpad/common/minidump_cpu_mips.h:31: * portions of DbgHelp.h from
the Windows Platform SDK.
This particular bit of the comment is probably not true, unless you're actually
replicating structure definitions you found in Microsoft headers.

http://breakpad.appspot.com/381002/diff/4001/src/google_breakpad/common/minid...
src/google_breakpad/common/minidump_cpu_mips.h:85: #endif
I don't really understand the purpose of this block (without having read the
rest of the patch). Generally these files are supposed to contain only
platform-independent structure definitions.

http://breakpad.appspot.com/381002/diff/4001/src/google_breakpad/common/minid...
src/google_breakpad/common/minidump_cpu_mips.h:94: * are not exactly minidumps.
Mark always tells me that comments in Google source are supposed to read in
third-person, FWIW.

http://breakpad.appspot.com/381002/diff/4001/src/processor/minidump.cc
File src/processor/minidump.cc (right):

http://breakpad.appspot.com/381002/diff/4001/src/processor/minidump.cc#newcod...
src/processor/minidump.cc:716: BPLOG(ERROR) << "MinidumpContext MIPS not match
system info";
This error message doesn't actually read right. Maybe it's wrong elsewhere, but
could you fix it? (Feel free to fix it elsewhere as well.)

[deepthi]

I guess there is one more place that needs porting. File : minidump_writer.cc
where DSO info is written to dump. Now it reads DT_DEBUG tag to get the linklist
of shared libs which loader filled. But for MIPS specifically, this DT_DEBUG tag
is not used. Instead DT_MIPS_RLD_MAP is used. We wont be getting any useful info
about shared libs for MIPS if this porting is not done.

[deepthi]

The below modification in Function : Dump() File :
src/client/linux/minidump_writer/minidump_writer.cc will work for above
mentioned issue.


#ifdef __mips__
        if (dyn.d_tag == DT_MIPS_RLD_MAP) {
            r_debug = (struct r_debug*)(*(ElfW(Addr)*)(dyn.d_un.d_ptr));
            continue;
        }
#else
        if (dyn.d_tag == DT_DEBUG) {
          r_debug = (struct r_debug*)dyn.d_un.d_ptr;
          continue;
        }
#endif