Check dtrace invop earlier, shotcut if it is dtrace fbt exception.

[howard0su_gmail.com]

Summary

Check invop of dtrace inside exception.S to reduce overhead to
call into C code. also check this earlier also reduce unnecessary
overhead.

[andrew]

You will also need to adjust the number of frames to skip in DTrace when running a stack trace as there is not one fewer.

sys/arm/arm/exception.S

374 ↗

(On Diff #10468)

This should use PSR_USR32_MODE

378–385 ↗

(On Diff #10468)

In l41BSD we moved this to the C handler

388 ↗

(On Diff #10468)

Move FBT_PATCHVAL from sys/cddl/dev/fbt/arm/fbt_isa.c to trap.h and use this here. You should also rename it to something like FBT_BREAKPOINT. We need to be able to tell the difference between DTrace and DDB breakpoint instructions.

[howard0su_gmail.com]

Check dtrace invop earlier, shotcut if it is dtrace fbt exception.

[howard0su_gmail.com]

it can enable fbt under VERSATILEPB. But still has issue when I try to sysctl kern.kdb.enter=1, i hit kernel stack overflow.

sys/arm/arm/exception.S

378–385 ↗

(On Diff #10468)

we have to do here since I want to check the INSN in assembly.

[howard0su_gmail.com]

Check dtrace invop earlier, shotcut if it is dtrace fbt exception.

[andrew]

sys/arm/arm/exception.S

397 ↗

(On Diff #10477)

This is only needed in the non-KDTRACE_HOOKS case.

sys/arm/include/trap.h

10 ↗

(On Diff #10477)

It should be named FBT_BREAKPOINT, and there is no need for the comment.

[howard0su_gmail.com]

fix armv7

[howard0su_gmail.com]

Fix armv7 test under BBB

[howard0su_gmail.com]

reduce stack to skip to 2

[howard0su_gmail.com]

Fix 5th parameter add Dtrace support to trap-v6 code.

[howard0su_gmail.com]

FAULT_TYPE_MASK is not needed

[andrew]

Unless someone has any objections I'm planning on committing this tomorrow.

[howard0su_gmail.com]

Rename FBT_PATCHVAL to FBT_BREAKPOINT

[howard0su_gmail.com]

Rename FBT_PATCHVAL to FBT_BREAKPOINT