On 12/11/2019 14.42, Janosch Frank wrote:
> On 11/12/19 1:09 PM, David Hildenbrand wrote:
>> On 11.11.19 16:33, Janosch Frank wrote:
>>> On a diag308 subcode 0 CRs will be reset, so we need a PSW mask
>>> without DAT. Also we need to set the short psw indication to be
>>> compliant with the architecture.
>>>
>>> Let's therefore define a reset PSW mask with 64 bit addressing and
>>> short PSW indication that is compliant with architecture and use it.
>>>
>>> Signed-off-by: Janosch Frank <frankja@xxxxxxxxxxxxx>
>>> ---
>>> lib/s390x/asm-offsets.c | 1 +
>>> lib/s390x/asm/arch_def.h | 3 ++-
>>> s390x/cstart64.S | 24 +++++++++++++++++-------
>>> 3 files changed, 20 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/lib/s390x/asm-offsets.c b/lib/s390x/asm-offsets.c
>>> index 4b213f8..61d2658 100644
>>> --- a/lib/s390x/asm-offsets.c
>>> +++ b/lib/s390x/asm-offsets.c
>>> @@ -58,6 +58,7 @@ int main(void)
>>> OFFSET(GEN_LC_SW_INT_FPRS, lowcore, sw_int_fprs);
>>> OFFSET(GEN_LC_SW_INT_FPC, lowcore, sw_int_fpc);
>>> OFFSET(GEN_LC_SW_INT_CRS, lowcore, sw_int_crs);
>>> + OFFSET(GEN_LC_SW_INT_PSW, lowcore, sw_int_psw);
>>> OFFSET(GEN_LC_MCCK_EXT_SA_ADDR, lowcore, mcck_ext_sa_addr);
>>> OFFSET(GEN_LC_FPRS_SA, lowcore, fprs_sa);
>>> OFFSET(GEN_LC_GRS_SA, lowcore, grs_sa);
>>> diff --git a/lib/s390x/asm/arch_def.h b/lib/s390x/asm/arch_def.h
>>> index 07d4e5e..7d25e4f 100644
>>> --- a/lib/s390x/asm/arch_def.h
>>> +++ b/lib/s390x/asm/arch_def.h
>>> @@ -79,7 +79,8 @@ struct lowcore {
>>> uint32_t sw_int_fpc; /* 0x0300 */
>>> uint8_t pad_0x0304[0x0308 - 0x0304]; /* 0x0304 */
>>> uint64_t sw_int_crs[16]; /* 0x0308 */
>>> - uint8_t pad_0x0310[0x11b0 - 0x0388]; /* 0x0388 */
>>> + struct psw sw_int_psw; /* 0x0388 */
>>> + uint8_t pad_0x0310[0x11b0 - 0x0390]; /* 0x0390 */
>>> uint64_t mcck_ext_sa_addr; /* 0x11b0 */
>>> uint8_t pad_0x11b8[0x1200 - 0x11b8]; /* 0x11b8 */
>>> uint64_t fprs_sa[16]; /* 0x1200 */
[...]
>> This patch breaks the smp test under TCG (no clue and no time to look
>> into the details :) ):
>
> I forgot to fixup the offset calculation at the top of the patch once
> again...
Maybe add a
_Static_assert(sizeof(struct lowcore) == xyz)
after the struct definitions, to avoid that this happens again?
Thomas
