On 8/23/19 1:00 PM, David Hildenbrand wrote:
> On 22.08.19 13:11, Janosch Frank wrote:
>> By adding a load reset routine to cstart.S we can also test the clear
>> reset done by subcode 0, as we now can restore our registers again.
>>
>> Signed-off-by: Janosch Frank <frankja@xxxxxxxxxxxxx>
>> ---
>> I managed to extract this from another bigger test, so let's add it to the bunch.
>> I'd be very happy about assembly review :-)
>> ---
>> s390x/cstart64.S | 27 +++++++++++++++++++++++++++
>> s390x/diag308.c | 31 ++++++++++---------------------
>> 2 files changed, 37 insertions(+), 21 deletions(-)
>>
>> diff --git a/s390x/cstart64.S b/s390x/cstart64.S
>> index dedfe80..47045e1 100644
>> --- a/s390x/cstart64.S
>> +++ b/s390x/cstart64.S
>> @@ -145,6 +145,33 @@ memsetxc:
>> .endm
>>
>> .section .text
>> +/*
>> + * load_reset calling convention:
>> + * %r2 subcode (0 or 1)
>> + */
>> +.globl load_reset
>> +load_reset:
>> + SAVE_REGS
>> + /* Save the first PSW word to the IPL PSW */
>> + epsw %r0, %r1
>> + st %r0, 0
>> + /* Store the address and the bit for 31 bit addressing */
>> + larl %r0, 0f
>> + oilh %r0, 0x8000
>> + st %r0, 0x4
>> + /* Do the reset */
>> + diag %r0,%r2,0x308
>> + /* Failure path */
>> + xgr %r2, %r2
>> + br %r14
>> + /* Success path */
>> + /* We lost cr0 due to the reset */
>> +0: larl %r1, initial_cr0
>> + lctlg %c0, %c0, 0(%r1)
>> + RESTORE_REGS
>> + lhi %r2, 1
>> + br %r14
>> +
>> pgm_int:
>> SAVE_REGS
>> brasl %r14, handle_pgm_int
>> diff --git a/s390x/diag308.c b/s390x/diag308.c
>> index f085b1a..baf9fd3 100644
>> --- a/s390x/diag308.c
>> +++ b/s390x/diag308.c
>> @@ -21,32 +21,20 @@ static void test_priv(void)
>> check_pgm_int_code(PGM_INT_CODE_PRIVILEGED_OPERATION);
>> }
>>
>> +
>> /*
>> - * Check that diag308 with subcode 1 loads the PSW at address 0, i.e.
>> + * Check that diag308 with subcode 0 and 1 loads the PSW at address 0, i.e.
>> * that we can put a pointer into address 4 which then gets executed.
>> */
>> +extern int load_reset(u64);
>> +static void test_subcode0(void)
>> +{
>> + report("load modified clear done", load_reset(0));
>> +}
>> +
>> static void test_subcode1(void)
>> {
>> - uint64_t saved_psw = *(uint64_t *)0;
>> - long subcode = 1;
>> - long ret, tmp;
>> -
>> - asm volatile (
>> - " epsw %0,%1\n"
>> - " st %0,0\n"
>> - " larl %0,0f\n"
>> - " oilh %0,0x8000\n"
>> - " st %0,4\n"
>> - " diag 0,%2,0x308\n"
>> - " lghi %0,0\n"
>> - " j 1f\n"
>> - "0: lghi %0,1\n"
>> - "1:"
>> - : "=&d"(ret), "=&d"(tmp) : "d"(subcode) : "memory");
>> -
>> - *(uint64_t *)0 = saved_psw;
>> -
>> - report("load normal reset done", ret == 1);
>> + report("load normal reset done", load_reset(1));
>> }
>>
>> /* Expect a specification exception when using an uneven register */
>> @@ -107,6 +95,7 @@ static struct {
>> void (*func)(void);
>> } tests[] = {
>> { "privileged", test_priv },
>> + { "subcode 0", test_subcode0 },
>> { "subcode 1", test_subcode1 },
>> { "subcode 5", test_subcode5 },
>> { "subcode 6", test_subcode6 },
>>
>
> So, in general I am wondering if we should restore the original IPL_PSW
> after we used it - is there any chance we might require the old value
> again (I guess we're fine with cpu resets)?
I currently don't see a need, but we could cache it in the restart old
psw address. Or we just store back the two word constant.
Attachment:
signature.asc
Description: OpenPGP digital signature
