[#4223] JUMP in cplb error handler may be too short

Submitted By: Luke Yang

Open Date

2008-07-06 03:28:13 Close Date

2008-07-18 16:36:50

Priority:

Medium Assignee:

Michael Hennerich

Status:

Closed Fixed In Release:

N/A

Found In Release:

2008R1-RC8 Release:

Category:

Kernel Functions Board:

N/A

Processor:

ALL Silicon Revision:

Is this bug repeatable?:

Yes Resolution:

Fixed

Uboot version or rev.:

Toolchain version or rev.:

App binary format:

N/A

Summary: JUMP in cplb error handler may be too short

Details:

In blackfin/kernel/cplb_nompu/cplbhdr.S

.Lcplb_error:

R1 = sp;

SP += -12;

call _panic_cplb_error;

SP += 12;

JUMP _handle_bad_cplb;

When _handle_bad_cplb is a long jump, you get error:

arch/blackfin/kernel/cplb-nompu/built-in.o: In function `_cplb_hdr':

(.text+0xf8): relocation truncated to fit: R_pcrel12_jump_s against symbol `handle_bad_cplb' defined in .text section in arch/blackfin/mach-common/built-in.o

Better use an indirect jump here.

Follow-ups

--- Robin Getz 2008-07-07 00:15:10

Luke:

Unless you have modified the kernel - this should be all in the same section,

and short jump will work fine.

Can you give more details on how to reproduce?

--- Sonic Zhang 2008-07-09 00:17:59

Luke deselects the kernel optimization options to move some routines into L1

Sram. He suggests to use jump.l instead of jump.

--- Michael Hennerich 2008-07-09 03:14:43

I have a fix allredy in my tree - assign to me.

-Michael

--- Robin Getz 2008-07-09 07:13:42

Sonic:

If Luke found a problem - then Luke needs to follow up when we ask for more

information.

I would like to understand what options he had selected - can he post is kernel

.config?

Thanks

--- Michael Hennerich 2008-07-09 07:26:30

The same bug triggered while working on the hibernate stuff.

We jump from one object file cplbhdr.o into entry.o:handle_bad_cplb since we

don't know how far they are linked apart from each other, it's more fail save to

use a long jump here

Fixed.

-Michael

--- Sonic Zhang 2008-07-09 22:43:13

Luke is a bit lazy to update the webpage. He told me his steps. He deselects all

options in "Blackfin Kernel Optimizations". He said the second option

"Locate entire ASM lowlevel exception/interrupt - syscall and CPLB handler

in L1 Memory".

--- Robin Getz 2008-07-18 16:36:45

Since Michael fixed - closing.

In the future - if people (eg - Luke) are too lazy to monitor the forums - we

should not work on things they report.

Files

Changes

Commits

Dependencies

Duplicates

Associations

Partner Forums

Partner Libraries

[#4223] JUMP in cplb error handler may be too short