Skip to content

Commit

Permalink
[S390] idle time accounting vs. machine checks
Browse files Browse the repository at this point in the history
A machine check can interrupt the i/o and external interrupt handler
anytime. If the machine check occurs while the interrupt handler is
waking up from idle vtime_start_cpu can get executed a second time
and the int_clock / async_enter_timer values in the lowcore get
clobbered. This can confuse the cpu time accounting.
To fix this problem two changes are needed. First the machine check
handler has to use its own copies of int_clock and async_enter_timer,
named mcck_clock and mcck_enter_timer. Second the nested execution
of vtime_start_cpu has to be prevented. This is done in s390_idle_check
by checking the wait bit in the program status word.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
  • Loading branch information
Martin Schwidefsky authored and Martin Schwidefsky committed May 17, 2010
1 parent 6a2df3a commit 6377981
Show file tree
Hide file tree
Showing 9 changed files with 125 additions and 116 deletions.
9 changes: 5 additions & 4 deletions arch/s390/include/asm/cputime.h
Original file line number Diff line number Diff line change
Expand Up @@ -188,15 +188,16 @@ struct s390_idle_data {

DECLARE_PER_CPU(struct s390_idle_data, s390_idle);

void vtime_start_cpu(void);
void vtime_start_cpu(__u64 int_clock, __u64 enter_timer);
cputime64_t s390_get_idle_time(int cpu);

#define arch_idle_time(cpu) s390_get_idle_time(cpu)

static inline void s390_idle_check(void)
static inline void s390_idle_check(struct pt_regs *regs, __u64 int_clock,
__u64 enter_timer)
{
if ((&__get_cpu_var(s390_idle))->idle_enter != 0ULL)
vtime_start_cpu();
if (regs->psw.mask & PSW_MASK_WAIT)
vtime_start_cpu(int_clock, enter_timer);
}

static inline int s390_nohz_delay(int cpu)
Expand Down
98 changes: 51 additions & 47 deletions arch/s390/include/asm/lowcore.h
Original file line number Diff line number Diff line change
Expand Up @@ -104,37 +104,39 @@ struct _lowcore {
/* CPU time accounting values */
__u64 sync_enter_timer; /* 0x0250 */
__u64 async_enter_timer; /* 0x0258 */
__u64 exit_timer; /* 0x0260 */
__u64 user_timer; /* 0x0268 */
__u64 system_timer; /* 0x0270 */
__u64 steal_timer; /* 0x0278 */
__u64 last_update_timer; /* 0x0280 */
__u64 last_update_clock; /* 0x0288 */
__u64 mcck_enter_timer; /* 0x0260 */
__u64 exit_timer; /* 0x0268 */
__u64 user_timer; /* 0x0270 */
__u64 system_timer; /* 0x0278 */
__u64 steal_timer; /* 0x0280 */
__u64 last_update_timer; /* 0x0288 */
__u64 last_update_clock; /* 0x0290 */

/* Current process. */
__u32 current_task; /* 0x0290 */
__u32 thread_info; /* 0x0294 */
__u32 kernel_stack; /* 0x0298 */
__u32 current_task; /* 0x0298 */
__u32 thread_info; /* 0x029c */
__u32 kernel_stack; /* 0x02a0 */

/* Interrupt and panic stack. */
__u32 async_stack; /* 0x029c */
__u32 panic_stack; /* 0x02a0 */
__u32 async_stack; /* 0x02a4 */
__u32 panic_stack; /* 0x02a8 */

/* Address space pointer. */
__u32 kernel_asce; /* 0x02a4 */
__u32 user_asce; /* 0x02a8 */
__u32 user_exec_asce; /* 0x02ac */
__u32 kernel_asce; /* 0x02ac */
__u32 user_asce; /* 0x02b0 */
__u32 user_exec_asce; /* 0x02b4 */

/* SMP info area */
__u32 cpu_nr; /* 0x02b0 */
__u32 softirq_pending; /* 0x02b4 */
__u32 percpu_offset; /* 0x02b8 */
__u32 ext_call_fast; /* 0x02bc */
__u64 int_clock; /* 0x02c0 */
__u64 clock_comparator; /* 0x02c8 */
__u32 machine_flags; /* 0x02d0 */
__u32 ftrace_func; /* 0x02d4 */
__u8 pad_0x02d8[0x0300-0x02d8]; /* 0x02d8 */
__u32 cpu_nr; /* 0x02b8 */
__u32 softirq_pending; /* 0x02bc */
__u32 percpu_offset; /* 0x02c0 */
__u32 ext_call_fast; /* 0x02c4 */
__u64 int_clock; /* 0x02c8 */
__u64 mcck_clock; /* 0x02d0 */
__u64 clock_comparator; /* 0x02d8 */
__u32 machine_flags; /* 0x02e0 */
__u32 ftrace_func; /* 0x02e4 */
__u8 pad_0x02e8[0x0300-0x02e8]; /* 0x02e8 */

/* Interrupt response block */
__u8 irb[64]; /* 0x0300 */
Expand Down Expand Up @@ -232,38 +234,40 @@ struct _lowcore {
/* CPU accounting and timing values. */
__u64 sync_enter_timer; /* 0x02a0 */
__u64 async_enter_timer; /* 0x02a8 */
__u64 exit_timer; /* 0x02b0 */
__u64 user_timer; /* 0x02b8 */
__u64 system_timer; /* 0x02c0 */
__u64 steal_timer; /* 0x02c8 */
__u64 last_update_timer; /* 0x02d0 */
__u64 last_update_clock; /* 0x02d8 */
__u64 mcck_enter_timer; /* 0x02b0 */
__u64 exit_timer; /* 0x02b8 */
__u64 user_timer; /* 0x02c0 */
__u64 system_timer; /* 0x02c8 */
__u64 steal_timer; /* 0x02d0 */
__u64 last_update_timer; /* 0x02d8 */
__u64 last_update_clock; /* 0x02e0 */

/* Current process. */
__u64 current_task; /* 0x02e0 */
__u64 thread_info; /* 0x02e8 */
__u64 kernel_stack; /* 0x02f0 */
__u64 current_task; /* 0x02e8 */
__u64 thread_info; /* 0x02f0 */
__u64 kernel_stack; /* 0x02f8 */

/* Interrupt and panic stack. */
__u64 async_stack; /* 0x02f8 */
__u64 panic_stack; /* 0x0300 */
__u64 async_stack; /* 0x0300 */
__u64 panic_stack; /* 0x0308 */

/* Address space pointer. */
__u64 kernel_asce; /* 0x0308 */
__u64 user_asce; /* 0x0310 */
__u64 user_exec_asce; /* 0x0318 */
__u64 kernel_asce; /* 0x0310 */
__u64 user_asce; /* 0x0318 */
__u64 user_exec_asce; /* 0x0320 */

/* SMP info area */
__u32 cpu_nr; /* 0x0320 */
__u32 softirq_pending; /* 0x0324 */
__u64 percpu_offset; /* 0x0328 */
__u64 ext_call_fast; /* 0x0330 */
__u64 int_clock; /* 0x0338 */
__u64 clock_comparator; /* 0x0340 */
__u64 vdso_per_cpu_data; /* 0x0348 */
__u64 machine_flags; /* 0x0350 */
__u64 ftrace_func; /* 0x0358 */
__u8 pad_0x0368[0x0380-0x0360]; /* 0x0360 */
__u32 cpu_nr; /* 0x0328 */
__u32 softirq_pending; /* 0x032c */
__u64 percpu_offset; /* 0x0330 */
__u64 ext_call_fast; /* 0x0338 */
__u64 int_clock; /* 0x0340 */
__u64 mcck_clock; /* 0x0348 */
__u64 clock_comparator; /* 0x0350 */
__u64 vdso_per_cpu_data; /* 0x0358 */
__u64 machine_flags; /* 0x0360 */
__u64 ftrace_func; /* 0x0368 */
__u8 pad_0x0370[0x0380-0x0370]; /* 0x0370 */

/* Interrupt response block. */
__u8 irb[64]; /* 0x0380 */
Expand Down
2 changes: 2 additions & 0 deletions arch/s390/kernel/asm-offsets.c
Original file line number Diff line number Diff line change
Expand Up @@ -112,6 +112,7 @@ int main(void)
DEFINE(__LC_RETURN_MCCK_PSW, offsetof(struct _lowcore, return_mcck_psw));
DEFINE(__LC_SYNC_ENTER_TIMER, offsetof(struct _lowcore, sync_enter_timer));
DEFINE(__LC_ASYNC_ENTER_TIMER, offsetof(struct _lowcore, async_enter_timer));
DEFINE(__LC_MCCK_ENTER_TIMER, offsetof(struct _lowcore, mcck_enter_timer));
DEFINE(__LC_EXIT_TIMER, offsetof(struct _lowcore, exit_timer));
DEFINE(__LC_USER_TIMER, offsetof(struct _lowcore, user_timer));
DEFINE(__LC_SYSTEM_TIMER, offsetof(struct _lowcore, system_timer));
Expand All @@ -127,6 +128,7 @@ int main(void)
DEFINE(__LC_USER_ASCE, offsetof(struct _lowcore, user_asce));
DEFINE(__LC_USER_EXEC_ASCE, offsetof(struct _lowcore, user_exec_asce));
DEFINE(__LC_INT_CLOCK, offsetof(struct _lowcore, int_clock));
DEFINE(__LC_MCCK_CLOCK, offsetof(struct _lowcore, mcck_clock));
DEFINE(__LC_MACHINE_FLAGS, offsetof(struct _lowcore, machine_flags));
DEFINE(__LC_FTRACE_FUNC, offsetof(struct _lowcore, ftrace_func));
DEFINE(__LC_IRB, offsetof(struct _lowcore, irb));
Expand Down
54 changes: 26 additions & 28 deletions arch/s390/kernel/entry.S
Original file line number Diff line number Diff line change
Expand Up @@ -742,15 +742,14 @@ __critical_end:

.globl mcck_int_handler
mcck_int_handler:
stck __LC_INT_CLOCK
stck __LC_MCCK_CLOCK
spt __LC_CPU_TIMER_SAVE_AREA # revalidate cpu timer
lm %r0,%r15,__LC_GPREGS_SAVE_AREA # revalidate gprs
SAVE_ALL_BASE __LC_SAVE_AREA+32
la %r12,__LC_MCK_OLD_PSW
tm __LC_MCCK_CODE,0x80 # system damage?
bo BASED(mcck_int_main) # yes -> rest of mcck code invalid
mvc __LC_SAVE_AREA+52(8),__LC_ASYNC_ENTER_TIMER
mvc __LC_ASYNC_ENTER_TIMER(8),__LC_CPU_TIMER_SAVE_AREA
mvc __LC_MCCK_ENTER_TIMER(8),__LC_CPU_TIMER_SAVE_AREA
tm __LC_MCCK_CODE+5,0x02 # stored cpu timer value valid?
bo BASED(1f)
la %r14,__LC_SYNC_ENTER_TIMER
Expand All @@ -764,7 +763,7 @@ mcck_int_handler:
bl BASED(0f)
la %r14,__LC_LAST_UPDATE_TIMER
0: spt 0(%r14)
mvc __LC_ASYNC_ENTER_TIMER(8),0(%r14)
mvc __LC_MCCK_ENTER_TIMER(8),0(%r14)
1: tm __LC_MCCK_CODE+2,0x09 # mwp + ia of old psw valid?
bno BASED(mcck_int_main) # no -> skip cleanup critical
tm __LC_MCK_OLD_PSW+1,0x01 # test problem state bit
Expand All @@ -786,9 +785,9 @@ mcck_int_main:
bno BASED(mcck_no_vtime) # no -> skip cleanup critical
tm SP_PSW+1(%r15),0x01 # interrupting from user ?
bz BASED(mcck_no_vtime)
UPDATE_VTIME __LC_EXIT_TIMER,__LC_ASYNC_ENTER_TIMER,__LC_USER_TIMER
UPDATE_VTIME __LC_EXIT_TIMER,__LC_MCCK_ENTER_TIMER,__LC_USER_TIMER
UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
mvc __LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
mvc __LC_LAST_UPDATE_TIMER(8),__LC_MCCK_ENTER_TIMER
mcck_no_vtime:
l %r9,__LC_THREAD_INFO # load pointer to thread_info struct
la %r2,SP_PTREGS(%r15) # load pt_regs
Expand All @@ -811,7 +810,6 @@ mcck_no_vtime:
mcck_return:
mvc __LC_RETURN_MCCK_PSW(8),SP_PSW(%r15) # move return PSW
ni __LC_RETURN_MCCK_PSW+1,0xfd # clear wait state bit
mvc __LC_ASYNC_ENTER_TIMER(8),__LC_SAVE_AREA+52
tm __LC_RETURN_MCCK_PSW+1,0x01 # returning to user ?
bno BASED(0f)
lm %r0,%r15,SP_R0(%r15) # load gprs 0-15
Expand Down Expand Up @@ -934,15 +932,16 @@ cleanup_critical:

cleanup_system_call:
mvc __LC_RETURN_PSW(8),0(%r12)
c %r12,BASED(.Lmck_old_psw)
be BASED(0f)
la %r12,__LC_SAVE_AREA+16
b BASED(1f)
0: la %r12,__LC_SAVE_AREA+32
1:
clc __LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+4)
bh BASED(0f)
mvc __LC_SYNC_ENTER_TIMER(8),__LC_MCCK_ENTER_TIMER
c %r12,BASED(.Lmck_old_psw)
be BASED(0f)
mvc __LC_SYNC_ENTER_TIMER(8),__LC_ASYNC_ENTER_TIMER
0: c %r12,BASED(.Lmck_old_psw)
la %r12,__LC_SAVE_AREA+32
be BASED(0f)
la %r12,__LC_SAVE_AREA+16
0: clc __LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+8)
bhe BASED(cleanup_vtime)
clc __LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn)
Expand Down Expand Up @@ -984,16 +983,19 @@ cleanup_sysc_tif:
cleanup_sysc_restore:
clc 4(4,%r12),BASED(cleanup_sysc_restore_insn)
be BASED(2f)
mvc __LC_EXIT_TIMER(8),__LC_MCCK_ENTER_TIMER
c %r12,BASED(.Lmck_old_psw)
be BASED(0f)
mvc __LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
clc 4(4,%r12),BASED(cleanup_sysc_restore_insn+4)
0: clc 4(4,%r12),BASED(cleanup_sysc_restore_insn+4)
be BASED(2f)
mvc __LC_RETURN_PSW(8),SP_PSW(%r15)
c %r12,BASED(.Lmck_old_psw)
bne BASED(0f)
mvc __LC_SAVE_AREA+32(16),SP_R12(%r15)
b BASED(1f)
0: mvc __LC_SAVE_AREA+16(16),SP_R12(%r15)
1: lm %r0,%r11,SP_R0(%r15)
la %r12,__LC_SAVE_AREA+32
be BASED(1f)
la %r12,__LC_SAVE_AREA+16
1: mvc 0(16,%r12),SP_R12(%r15)
lm %r0,%r11,SP_R0(%r15)
l %r15,SP_R15(%r15)
2: la %r12,__LC_RETURN_PSW
br %r14
Expand All @@ -1009,19 +1011,15 @@ cleanup_io_tif:

cleanup_io_restore:
clc 4(4,%r12),BASED(cleanup_io_restore_insn)
be BASED(2f)
mvc __LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
be BASED(1f)
mvc __LC_EXIT_TIMER(8),__LC_MCCK_ENTER_TIMER
clc 4(4,%r12),BASED(cleanup_io_restore_insn+4)
be BASED(2f)
be BASED(1f)
mvc __LC_RETURN_PSW(8),SP_PSW(%r15)
c %r12,BASED(.Lmck_old_psw)
bne BASED(0f)
mvc __LC_SAVE_AREA+32(16),SP_R12(%r15)
b BASED(1f)
0: mvc __LC_SAVE_AREA+16(16),SP_R12(%r15)
1: lm %r0,%r11,SP_R0(%r15)
lm %r0,%r11,SP_R0(%r15)
l %r15,SP_R15(%r15)
2: la %r12,__LC_RETURN_PSW
1: la %r12,__LC_RETURN_PSW
br %r14
cleanup_io_restore_insn:
.long io_done - 4 + 0x80000000
Expand Down
Loading

0 comments on commit 6377981

Please sign in to comment.