Re: [PATCH] ARM: fix alignement of __bug_table section entries

From: Robert Jarzmik
Date: Sat Sep 05 2015 - 13:16:13 EST


Russell King - ARM Linux <linux@xxxxxxxxxxxxxxxx> writes:

> On Sat, Sep 05, 2015 at 03:48:38PM +0200, Robert Jarzmik wrote:
>> This time I took my JTAG to have a look at the flow, in arch/arm/mm/alignment.c,
>> where I added the small chunk in [2], which gave in my case :
>> RJK: fault=4 instr=0x00000000 instrptr=0xc02b37c8 thumb_mode=0 tinstr=0x0000
>
> Right, so as fault is nonzero, this means that we were unable to read the
> instruction. That seems mad though - the instruction pointer is certainly
> valid, and as we're using probe_kernel_address(), that switches to the
> kernel "segment" before trying to read kernel addresses. That should
> mean that __copy_from_user_inatomic() is able to read the instruction.
>
> I think this is the root cause of the issue.

And there is more madness to come : I tried to "reread" the instruction [1] a
second time if the first result was 4 :
RJK: fault=4 instr=0x00000000(@c385d72c) instrptr=0xc02b39e8 thumb_mode=0 tinstr=0x0000
RJK: reread instruction: [0xc02b39e8] = 0x10c650b2: 0

Guess what, the second probe_kernel_address() with the same parameters returns
0, and everything works. It's insane.

>> Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none
> It seems you have SW_DOMAIN_PAN enabled.
That's the default arch/arm/Kconfig implies.
And ... this is what also _is_ the cause of this behavior : removing
SW_DOMAIN_PAN makes all my pxa boards work again !!!

Moreover, this is consistent with the fact that this commit is in linux-next but
not in v4.1 :
a5e090acbf54 ("ARM: software-based priviledged-no-access support")

So the issue is around this SW_DOMAIN_PAN, at least on PXA.

--
Robert

[1]
@@ -787,6 +798,15 @@ do_alignment(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
instr = __mem_to_opcode_arm(instr);
}

+ pr_info("RJK: fault=%d instr=0x%08lx(@%p) instrptr=0x%08lx thumb_mode=%lu tinstr=0x%04x\n",
+ fault, instr, &instr, instrptr, thumb_mode(regs), tinstr);
+ if (fault == 4 && !thumb_mode(regs)) {
+ fault = probe_kernel_address(instrptr, instr);
+ pr_info("RJK: reread instruction: [0x%08lx] = 0x%08lx: %u\n",
+ instrptr, instr, fault);
+ rjk_debug_point(instrptr);
+ }
+
if (fault) {
type = TYPE_FAULT;
goto bad_or_fault;

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/