Commit graph

6 commits

Author SHA1 Message Date
Jon Medhurst
ad684dce87 ARM: 8179/1: kprobes-test: Fix compile error "bad immediate value for offset"
When compiling kprobes-test-arm.c the following error has been observed

/tmp/ccoT403o.s:21439: Error: bad immediate value for offset (4168)

This is caused by the compiler spilling it's literal pool too far away
from the site which is trying to reference it with a PC relative load.
This arises because the compiler is underestimating the size of the
inline assembler code present, which apparently it approximates as 4
bytes per line or instruction.

We fix this problem by moving the operations which generate more than
4 bytes out of the text section. Specifically, moving the .ascii
directives to the .rodata section.

Signed-off-by: Jon Medhurst <tixy@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-09-30 16:55:24 +01:00
Taras Kondratiuk
41b5368f31 ARM: kprobes-test: Workaround GAS .align bug
By default if no fill symbol is given to .align directive in a code
section it fills gap with NOPs. If previous fragment is not
instruction-aligned, additional pre-alignment is done by zero bytes
before NOPs. These zero bytes are marked as data by special symbol $d in
symbol table. Unfortunately GAS assumes that there is only code in the
code section so it "puts back" code symbol $a at the end of this
pre-alignment. So if there is some data after alignment it will be
interpreted as code and will be swapped back to LE for BE8 system during
a final linking.

If explicit fill value is given to .align, the NOP-padding code is
skipped and symbol table does not get messed-up.

So the workaround for this issue:
Use explicit fill value if data should be aligned in the code section.

Acked-by: Ben Dooks <ben.dooks@codethink.co.uk>
Acked-by: Jon Medhurst <tixy@linaro.org>
Signed-off-by: Taras Kondratiuk <taras.kondratiuk@linaro.org>
2014-04-01 16:48:27 +03:00
Jon Medhurst
46589e2922 ARM: 7174/1: Fix build error in kprobes test code on Thumb2 kernels
When compiling kprobes-test-thumb.c an error like below may occur:

/tmp/ccKcuJcG.s:19179: Error: offset out of range

This is caused by the compiler underestimating the size of the inline
assembler instructions containing ".space 0x1000" and failing to spill
the literal pool in time to prevent the generation of PC relative load
instruction with invalid offsets.

The fix implemented by this patch is to replace a single large .space
directive by a number of 4 byte .space's. This requires splitting the
macros which generate test cases for branch instructions into two forms:
one with, and one without support for inserting extra code between
branch and target.

Acked-by: Nicolas Pitre <nicolas.pitre@linaro.org>
Signed-off-by: Jon Medhurst <jon.medhurst@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-11-26 21:58:53 +00:00
Jon Medhurst
c0cc6df163 ARM: kprobes: Add ARM instruction simulation test cases
Signed-off-by: Jon Medhurst <tixy@yxit.co.uk>
Acked-by: Nicolas Pitre <nicolas.pitre@linaro.org>
2011-09-20 18:17:43 +00:00
Jon Medhurst
c7054aad53 ARM: kprobes: Add Thumb instruction simulation test cases
Signed-off-by: Jon Medhurst <tixy@yxit.co.uk>
Acked-by: Nicolas Pitre <nicolas.pitre@linaro.org>
2011-09-20 18:17:43 +00:00
Jon Medhurst
a43bc69b39 ARM: kprobes: Framework for instruction set test cases
On ARM we have to simulate/emulate CPU instructions in order to
singlestep them. This patch adds a framework which can be used to
construct test cases for different instruction forms. It is described in
detail in the in-source comments of kprobes-test.c

Signed-off-by: Jon Medhurst <tixy@yxit.co.uk>
Acked-by: Nicolas Pitre <nicolas.pitre@linaro.org>
2011-09-20 18:17:43 +00:00