s390x/tcg: Implement MONITOR CALL
PATCH v2 1/1 s390x/tcg: Implement MONITOR CALL
Recent upstream Linux uses the MONITOR CALL instruction for things like BUG_ON() and WARN_ON(). We currently inject an operation exception when we hit a MONITOR CALL instruction - which is wrong, as the instruction is not glued to specific CPU features.
Doing a simple WARN_ON_ONCE() currently results in a panic:
[ 18.162801] illegal operation: 0001 ilc:2 [#1] SMP [ 18.162889] Modules linked in: [...] [ 18.165476] Kernel panic - not syncing: Fatal exception: panic_on_oops
With a proper implementation, we now get:
[ 18.242754] ------------[ cut here ]------------ [ 18.242855] WARNING: CPU: 7 PID: 1 at init/main.c:1534 [...] [ 18.242919] Modules linked in: [...] [ 18.246262] ---[ end trace a420477d71dc97b4 ]--- [ 18.259014] Freeing unused kernel memory: 4220K
---
v1 -> v2:
- Simplify by not using the tb flags, always calling the helper.
I looked into monitor-event counting, which looks easy at first glance
- but proper DAT/access exception handling is tricky. Leaving that for a cold winter evening :)
--- target/s390x/excp_helper.c | 23 +++++++++++++++++++++++ target/s390x/helper.h | 1 + target/s390x/insn-data.def | 3 +++ target/s390x/translate.c | 21 +++++++++++++++++++++ 4 files changed, 48 insertions(+)
target/s390x/excp_helper.c +23 -0
@@ -610,4 +610,27 @@ void s390x_cpu_do_unaligned_access(CPUState *cs, vaddr addr,tcg_s390_program_interrupt (env,PGM_SPECIFICATION , retaddr); } +static void QEMU_NORETURN monitor_event(CPUS390XState *env, +uint64_t monitor_code, +uint8_t monitor_class,uintptr_t ra) +{ +/* Store the Monitor Code and the Monitor Class Number into the lowcore */ +stq_phys (env_cpu (env)->as, + env->psa + offsetof(LowCore , monitor_code), monitor_code); +stw_phys (env_cpu (env)->as, + env->psa + offsetof(LowCore , mon_class_num), monitor_class); + +tcg_s390_program_interrupt (env,PGM_MONITOR , ra); +} + +void HELPER (monitor_call )(CPUS390XState *env,uint64_t monitor_code, +uint32_t monitor_class) +{ +g_assert (monitor_class <=0xff ); + +if (env->cregs[8 ] & (0x8000 >> monitor_class)) { +monitor_event (env, monitor_code, monitor_class,GETPC ()); + } +} +#endif /* CONFIG_USER_ONLY */
target/s390x/helper.h +1 -0
@@ -349,4 +349,5 @@ DEF_HELPER_3(sic, void, env, i64, i64) DEF_HELPER_3(rpcit, void, env, i32, i32) DEF_HELPER_5(pcistb, void, env, i32, i32, i64, i32) DEF_HELPER_4(mpcifc, void, env, i32, i64, i32) +DEF_HELPER_3(monitor_call, void, env, i64, i32)#endif
target/s390x/insn-data.def +3 -0
@@ -617,6 +617,9 @@ C(0x9a00, LAM, RS_a, Z, 0, a2, 0, 0, lam, 0) C(0xeb9a, LAMY, RSY_a, LD, 0, a2, 0, 0, lam, 0) +/* MONITOR CALL */ + C(0xaf00, MC, SI, Z, la1, 0, 0, 0, mc, 0) + /* MOVE */ C(0xd200, MVC, SS_a, Z, la1, a2, 0, 0, mvc, 0) C(0xe544, MVHHI, SIL, GIE, la1, i2, 0, m1_16, mov2, 0)
target/s390x/translate.c +21 -0
@@ -3302,6 +3302,27 @@ static DisasJumpType op_lcbb(DisasContext *s, DisasOps *o)return DISAS_NEXT ; } +static DisasJumpType op_mc (DisasContext *s,DisasOps *o) +{ +#if !defined(CONFIG_USER_ONLY ) +TCGv_i32 i2; +#endif +const uint16_t monitor_class =get_field (s, i2); + +if (monitor_class &0xff00 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +#if !defined(CONFIG_USER_ONLY ) + i2 =tcg_const_i32 (monitor_class); +gen_helper_monitor_call (cpu_env, o->addr1, i2); +tcg_temp_free_i32 (i2); +#endif +/* Defaults to a NOP. */ +return DISAS_NEXT ; +} +static DisasJumpType op_mov2 (DisasContext *s,DisasOps *o) { o->out = o->in2;
Signature · 1 line
2.26.2
Quoted 19 lines · David Hildenbrand
Recent upstream Linux uses the MONITOR CALL instruction for things like
BUG_ON()andWARN_ON(). We currently inject an operation exception when we hit a MONITOR CALL instruction - which is wrong, as the instruction is not glued to specific CPU features.Doing a simple
WARN_ON_ONCE()currently results in a panic:[ 18.162801] illegal operation: 0001 ilc:2 [#1] SMP [ 18.162889] Modules linked in: [...] [ 18.165476] Kernel panic - not syncing: Fatal exception: panic_on_oopsWith a proper implementation, we now get:
[ 18.242754] ------------[ cut here ]------------ [ 18.242855] WARNING: CPU: 7 PID: 1 at init/main.c:1534 [...] [ 18.242919] Modules linked in: [...] [ 18.246262] ---[ end trace a420477d71dc97b4 ]--- [ 18.259014] Freeing unused kernel memory: 4220K
Quoted 3 lines
Reported-by Christian Borntraeger <[email protected]>Signed-off-by David Hildenbrand <[email protected]>---
r~
Cornelia Huck
AppliedQuoted 21 lines · On Fri, 18 Sep 2020 10:51:22 +0200 David Hildenbrand <[email protected]> wrote:
Recent upstream Linux uses the MONITOR CALL instruction for things like
BUG_ON()andWARN_ON(). We currently inject an operation exception when we hit a MONITOR CALL instruction - which is wrong, as the instruction is not glued to specific CPU features.Doing a simple
WARN_ON_ONCE()currently results in a panic:[ 18.162801] illegal operation: 0001 ilc:2 [#1] SMP [ 18.162889] Modules linked in: [...] [ 18.165476] Kernel panic - not syncing: Fatal exception: panic_on_oopsWith a proper implementation, we now get:
[ 18.242754] ------------[ cut here ]------------ [ 18.242855] WARNING: CPU: 7 PID: 1 at init/main.c:1534 [...] [ 18.242919] Modules linked in: [...] [ 18.246262] ---[ end trace a420477d71dc97b4 ]--- [ 18.259014] Freeing unused kernel memory: 4220KReported-by Christian Borntraeger <[email protected]>Signed-off-by David Hildenbrand <[email protected]>
Quoted 8 lines
---
v1 -> v2:
- Simplify by not using the tb flags, always calling the helper.I looked into monitor-event counting, which looks easy at first glance
- but proper DAT/access exception handling is tricky. Leaving that for a cold winter evening :)
Fortunately, this is covered by a facility bit, so no need to do this now :)
Quoted 7 lines
--- target/s390x/excp_helper.c | 23 +++++++++++++++++++++++ target/s390x/helper.h | 1 + target/s390x/insn-data.def | 3 +++ target/s390x/translate.c | 21 +++++++++++++++++++++ 4 files changed, 48 insertions(+)
Thanks, applied.