[PATCH v3] ARMv7M: Fix exception handler for supporting FPU

[Daniel Gutson]

On Mon, Aug 31, 2015 at 2:34 PM, Gedare Bloom <gedare at rtems.org> wrote:
> I'd approve 2 patches in case you want to give credit. First patch
> with Sudarshan's fix, and Martin's improvement second.

+1

>
> On Mon, Aug 31, 2015 at 1:28 PM, Daniel Gutson
> <daniel.gutson at tallertechnologies.com> wrote:
>> Side question to Joel and others: is this enough to credit Sudarshan?
>> He did an amazing job for finding the bug and proposing an initial
>> solution.
>>
>> On Mon, Aug 31, 2015 at 11:37 AM, Martin Galvan
>> <martin.galvan at tallertechnologies.com> wrote:
>>> On exception entry, _ARMV7M_Exception_default stores the previous Stack Pointer
>>> in a CPU_Exception_frame. The SP can be MSP or PSP, depending on the mode
>>> in which the exception was taken. To know this, we must check the value of LR.
>>>
>>> Right now the code checks whether it should store MSP or PSP by comparing LR to
>>> -3. However, this doesn't work if we're using an FPU. This patch fixes that
>>> by checking bit 2 of LR, which indicates which SP we were using.
>>>
>>> Thanks to Sudarshan Rajagopalan for finding the bug and proposing a first
>>> version of the fix.
>>>
>>> Closes #2401.
>>>
>>> ---
>>>  cpukit/score/cpu/arm/armv7m-exception-default.c | 27 +++++++++++++++++--------
>>>  1 file changed, 19 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/cpukit/score/cpu/arm/armv7m-exception-default.c b/cpukit/score/cpu/arm/armv7m-exception-default.c
>>> index e890cdf..d04dbea 100644
>>> --- a/cpukit/score/cpu/arm/armv7m-exception-default.c
>>> +++ b/cpukit/score/cpu/arm/armv7m-exception-default.c
>>> @@ -22,16 +22,27 @@
>>>
>>>  void __attribute__((naked)) _ARMV7M_Exception_default( void )
>>>  {
>>> +    /* On exception entry, ARMv7M saves context state onto a stack pointed to
>>> +     * by either MSP or PSP. The value stored in LR indicates whether we were
>>> +     * in Thread or Handler mode, whether we were using the FPU (if any),
>>> +     * and which stack pointer we were using.
>>> +     * In particular, bit 2 of LR will be 0 if we were using MSP.
>>> +     *
>>> +     * For a more detailed explanation, see the Exception Entry Behavior
>>> +     * section of the ARMv7M Architecture Reference Manual.
>>> +     */
>>> +
>>> +    /* As we're in Handler mode here, we'll always operate on MSP.
>>> +     * However, we need to store the right SP in our CPU_Exception_frame.
>>> +     */
>>>    __asm__ volatile (
>>> -    "sub sp, %[cpufsz]\n"
>>> +    "sub sp, %[cpufsz]\n" /* Allocate space for a CPU_Exception_frame. */
>>>      "stm sp, {r0-r12}\n"
>>> -    "mov r2, lr\n"
>>> -    "mrs r1, msp\n"
>>> -    "mrs r0, psp\n"
>>> -    "cmn r2, #3\n"
>>> -    "itt ne\n"
>>> -    "movne r0, r1\n"
>>> -    "addne r0, %[cpufsz]\n"
>>> +    "tst lr, #4\n"       /* Check if bit 2 of LR is zero. (If so, PSR.Z = 1) */
>>> +    "itte eq\n"          /* IF bit 2 of LR is zero... (PSR.Z == 1) */
>>> +    "mrseq r0, msp\n"    /* THEN we were using MSP. */
>>> +    "addeq r0, %[cpufsz]\n" /* THEN, set r0 = old MSP. */
>>> +    "mrsne r0, psp\n"    /* ELSE it's not zero; we were using PSP. */
>>>      "add r2, r0, %[v7mlroff]\n"
>>>      "add r1, sp, %[cpulroff]\n"
>>>      "ldm r2, {r3-r5}\n"
>>> --
>>> 1.9.1
>>
>>
>>
>> --
>>
>> Daniel F. Gutson
>> Chief Engineering Officer, SPD
>>
>> San Lorenzo 47, 3rd Floor, Office 5
>> Córdoba, Argentina
>>
>> Phone:   +54 351 4217888 / +54 351 4218211
>> Skype:    dgutson
>> LinkedIn: http://ar.linkedin.com/in/danielgutson



-- 

Daniel F. Gutson
Chief Engineering Officer, SPD

San Lorenzo 47, 3rd Floor, Office 5
Córdoba, Argentina

Phone:   +54 351 4217888 / +54 351 4218211
Skype:    dgutson
LinkedIn: http://ar.linkedin.com/in/danielgutson