powerpc: More fixes for lazy IRQ vs. idle

Looks like we still have issues with pSeries and Cell idle code
vs. the lazy irq state. In fact, the reset fixes that went upstream
are exposing the problem more by causing BUG_ON() to trigger (which
this patch turns into a WARN_ON instead).

We need to be careful when using a variant of low power state that
has the side effect of turning interrupts back on, to properly set
all the SW & lazy state to look as if everything is enabled before
we enter the low power state with MSR:EE off as we will return with
MSR:EE on. If not, we have a discrepancy of state which can cause
things to go very wrong later on.

This patch moves the logic into a helper and uses it from the
pseries and cell idle code. The power4/970 idle code already got
things right (in assembly even !) so I'm not touching it. The power7
"bare metal" idle code is subtly different and correct. Remains PA6T
and some hypervisor based Cell platforms which have questionable
code in there, but they are mostly dead platforms so I'll fix them
when I manage to get final answers from the respective maintainers
about how the low power state actually works on them.
Signed-off-by: default avatarBenjamin Herrenschmidt <benh@kernel.crashing.org>
CC: stable@vger.kernel.org [v3.4]
parent 2437fccf
...@@ -125,6 +125,8 @@ static inline bool arch_irq_disabled_regs(struct pt_regs *regs) ...@@ -125,6 +125,8 @@ static inline bool arch_irq_disabled_regs(struct pt_regs *regs)
return !regs->softe; return !regs->softe;
} }
extern bool prep_irq_for_idle(void);
#else /* CONFIG_PPC64 */ #else /* CONFIG_PPC64 */
#define SET_MSR_EE(x) mtmsr(x) #define SET_MSR_EE(x) mtmsr(x)
......
...@@ -286,6 +286,52 @@ void notrace restore_interrupts(void) ...@@ -286,6 +286,52 @@ void notrace restore_interrupts(void)
__hard_irq_enable(); __hard_irq_enable();
} }
/*
* This is a helper to use when about to go into idle low-power
* when the latter has the side effect of re-enabling interrupts
* (such as calling H_CEDE under pHyp).
*
* You call this function with interrupts soft-disabled (this is
* already the case when ppc_md.power_save is called). The function
* will return whether to enter power save or just return.
*
* In the former case, it will have notified lockdep of interrupts
* being re-enabled and generally sanitized the lazy irq state,
* and in the latter case it will leave with interrupts hard
* disabled and marked as such, so the local_irq_enable() call
* in cpu_idle() will properly re-enable everything.
*/
bool prep_irq_for_idle(void)
{
/*
* First we need to hard disable to ensure no interrupt
* occurs before we effectively enter the low power state
*/
hard_irq_disable();
/*
* If anything happened while we were soft-disabled,
* we return now and do not enter the low power state.
*/
if (lazy_irq_pending())
return false;
/* Tell lockdep we are about to re-enable */
trace_hardirqs_on();
/*
* Mark interrupts as soft-enabled and clear the
* PACA_IRQ_HARD_DIS from the pending mask since we
* are about to hard enable as well as a side effect
* of entering the low power state.
*/
local_paca->irq_happened &= ~PACA_IRQ_HARD_DIS;
local_paca->soft_enabled = 1;
/* Tell the caller to enter the low power state */
return true;
}
#endif /* CONFIG_PPC64 */ #endif /* CONFIG_PPC64 */
int arch_show_interrupts(struct seq_file *p, int prec) int arch_show_interrupts(struct seq_file *p, int prec)
......
...@@ -42,11 +42,9 @@ static void cbe_power_save(void) ...@@ -42,11 +42,9 @@ static void cbe_power_save(void)
{ {
unsigned long ctrl, thread_switch_control; unsigned long ctrl, thread_switch_control;
/* /* Ensure our interrupt state is properly tracked */
* We need to hard disable interrupts, the local_irq_enable() done by if (!prep_irq_for_idle())
* our caller upon return will hard re-enable. return;
*/
hard_irq_disable();
ctrl = mfspr(SPRN_CTRLF); ctrl = mfspr(SPRN_CTRLF);
...@@ -81,6 +79,9 @@ static void cbe_power_save(void) ...@@ -81,6 +79,9 @@ static void cbe_power_save(void)
*/ */
ctrl &= ~(CTRL_RUNLATCH | CTRL_TE); ctrl &= ~(CTRL_RUNLATCH | CTRL_TE);
mtspr(SPRN_CTRLT, ctrl); mtspr(SPRN_CTRLT, ctrl);
/* Re-enable interrupts in MSR */
__hard_irq_enable();
} }
static int cbe_system_reset_exception(struct pt_regs *regs) static int cbe_system_reset_exception(struct pt_regs *regs)
......
...@@ -99,15 +99,18 @@ static int snooze_loop(struct cpuidle_device *dev, ...@@ -99,15 +99,18 @@ static int snooze_loop(struct cpuidle_device *dev,
static void check_and_cede_processor(void) static void check_and_cede_processor(void)
{ {
/* /*
* Interrupts are soft-disabled at this point, * Ensure our interrupt state is properly tracked,
* but not hard disabled. So an interrupt might have * also checks if no interrupt has occurred while we
* occurred before entering NAP, and would be potentially * were soft-disabled
* lost (edge events, decrementer events, etc...) unless
* we first hard disable then check.
*/ */
hard_irq_disable(); if (prep_irq_for_idle()) {
if (!lazy_irq_pending())
cede_processor(); cede_processor();
#ifdef CONFIG_TRACE_IRQFLAGS
/* Ensure that H_CEDE returns with IRQs on */
if (WARN_ON(!(mfmsr() & MSR_EE)))
__hard_irq_enable();
#endif
}
} }
static int dedicated_cede_loop(struct cpuidle_device *dev, static int dedicated_cede_loop(struct cpuidle_device *dev,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment