Crash on ARM: 8148/1 flush TLS and thumbee register

[Stefan Agner]

Hi all,

While working on Vybrid (vf610) Cortex-M4 support I also hit this crash
observed by Joachim:

...
Freeing unused kernel memory: 16K (8800a000 - 8800e000)

Unhandled exception: IPSR = 00000005 LR = fffffff1
CPU: 0 PID: 1 Comm: swapper Not tainted 3.18.0-rc1-00041-ga30465a #216
task: 8b838000 ti: 8b82a000 task.ti: 8b82a000
PC is at flush_thread+0x32/0x40
LR is at flush_thread+0x21/0x40
pc : [<8f00157a>]    lr : [<8f001569>]    psr: 4100000b
sp : 8b82be20  ip : 00000000  fp : 8b83c000
r10: 00000001  r9 : 88018c84  r8 : 8bb85000
r7 : 8b838000  r6 : 00000000  r5 : 8bb77400  r4 : 8b82a000
r3 : ffff0ff0  r2 : 8b82a000  r1 : 00000000  r0 : 88020354
xPSR: 4100000b
CPU: 0 PID: 1 Comm: swapper Not tainted 3.18.0-rc1-00041-ga30465a #216
[<8f002bc1>] (unwind_backtrace) from [<8f002033>] (show_stack+0xb/0xc)
[<8f002033>] (show_stack) from [<8f00265b>] (__invalid_entry+0x4b/0x4c)

Can reproduce it on 3.18-rc1, as well as 3.17.

Am 2014-09-26 23:27, schrieb Nathan Lynch:
<snip>
> Thanks for the report -- I see r3 has ffff0ff0 so I'm guessing this is
> set_tls attempting to clear the tls location in the kuser helper page,
> which I suppose isn't appropriate on MMU-less ARM?

I dug a bit deeper and found out that CONFIG_KUSER_HELPERS was enabled
on my build. Without CONFIG_KUSER_HELPERS, the kernel did not crash with
this patch (which makes sense, looking the code, the crashing code is
not executed in the !CONFIG_KUSER_HELPERS case)...

However, wouldn't a "depends on MMU" be required for
CONFIG_KUSER_HELPERS anyway? As far as I understood, this maps stuff
into the user space tasks address space, which is only possible with
MMU...

Also added Uwe Kleine-König as he worked on !MMU Cortex-M3 support.

--
Stefan