[PATCH v4 2/2] powerpc/rtas: Implement reentrant rtas call

From: Leonardo Bras
Date: Thu May 14 2020 - 19:53:05 EST


Implement rtas_call_reentrant() for reentrant rtas-calls:
"ibm,int-on", "ibm,int-off",ibm,get-xive" and "ibm,set-xive".

On LoPAPR Version 1.1 (March 24, 2016), from 7.3.10.1 to 7.3.10.4,
items 2 and 3 say:

2 - For the PowerPC External Interrupt option: The * call must be
reentrant to the number of processors on the platform.
3 - For the PowerPC External Interrupt option: The * argument call
buffer for each simultaneous call must be physically unique.

So, these rtas-calls can be called in a lockless way, if using
a different buffer for each cpu doing such rtas call.

For this, it was suggested to add the buffer (struct rtas_args)
in the PACA struct, so each cpu can have it's own buffer.

Reentrant rtas calls are useful to avoid deadlocks in crashing,
where rtas-calls are needed, but some other thread crashed holding
the rtas.lock.

This is a backtrace of a deadlock from a kdump testing environment:

#0 arch_spin_lock
#1 lock_rtas ()
#2 rtas_call (token=8204, nargs=1, nret=1, outputs=0x0)
#3 ics_rtas_mask_real_irq (hw_irq=4100)
#4 machine_kexec_mask_interrupts
#5 default_machine_crash_shutdown
#6 machine_crash_shutdown
#7 __crash_kexec
#8 crash_kexec
#9 oops_end

Signed-off-by: Leonardo Bras <leobras.c@xxxxxxxxx>
---
arch/powerpc/include/asm/paca.h | 2 ++
arch/powerpc/include/asm/rtas.h | 1 +
arch/powerpc/kernel/rtas.c | 52 +++++++++++++++++++++++++++++
arch/powerpc/sysdev/xics/ics-rtas.c | 22 ++++++------
4 files changed, 66 insertions(+), 11 deletions(-)

diff --git a/arch/powerpc/include/asm/paca.h b/arch/powerpc/include/asm/paca.h
index e3cc9eb9204d..5a76ba50b40f 100644
--- a/arch/powerpc/include/asm/paca.h
+++ b/arch/powerpc/include/asm/paca.h
@@ -29,6 +29,7 @@
#include <asm/hmi.h>
#include <asm/cpuidle.h>
#include <asm/atomic.h>
+#include <asm/rtas-types.h>

#include <asm-generic/mmiowb_types.h>

@@ -270,6 +271,7 @@ struct paca_struct {
#ifdef CONFIG_MMIOWB
struct mmiowb_state mmiowb_state;
#endif
+ struct rtas_args reentrant_args;
} ____cacheline_aligned;

extern void copy_mm_to_paca(struct mm_struct *mm);
diff --git a/arch/powerpc/include/asm/rtas.h b/arch/powerpc/include/asm/rtas.h
index c35c5350b7e4..fa7509c85881 100644
--- a/arch/powerpc/include/asm/rtas.h
+++ b/arch/powerpc/include/asm/rtas.h
@@ -236,6 +236,7 @@ extern struct rtas_t rtas;
extern int rtas_token(const char *service);
extern int rtas_service_present(const char *service);
extern int rtas_call(int token, int, int, int *, ...);
+int rtas_call_reentrant(int token, int nargs, int nret, int *outputs, ...);
void rtas_call_unlocked(struct rtas_args *args, int token, int nargs,
int nret, ...);
extern void __noreturn rtas_restart(char *cmd);
diff --git a/arch/powerpc/kernel/rtas.c b/arch/powerpc/kernel/rtas.c
index c5fa251b8950..31710b358f44 100644
--- a/arch/powerpc/kernel/rtas.c
+++ b/arch/powerpc/kernel/rtas.c
@@ -41,6 +41,7 @@
#include <asm/time.h>
#include <asm/mmu.h>
#include <asm/topology.h>
+#include <asm/paca.h>

/* This is here deliberately so it's only used in this file */
void enter_rtas(unsigned long);
@@ -483,6 +484,57 @@ int rtas_call(int token, int nargs, int nret, int *outputs, ...)
}
EXPORT_SYMBOL(rtas_call);

+/**
+ * rtas_call_reentrant() - Used for reentrant rtas calls
+ * @token: Token for desired reentrant RTAS call
+ * @nargs: Number of Input Parameters
+ * @nret: Number of Output Parameters
+ * @outputs: Array of outputs
+ * @...: Inputs for desired RTAS call
+ *
+ * According to LoPAR documentation, only "ibm,int-on", "ibm,int-off",
+ * "ibm,get-xive" and "ibm,set-xive" are currently reentrant.
+ * Reentrant calls need their own rtas_args buffer, so not using rtas.args, but
+ * PACA one instead.
+ *
+ * Return: -1 on error,
+ * First output value of RTAS call if (nret > 0),
+ * 0 otherwise,
+ */
+
+int rtas_call_reentrant(int token, int nargs, int nret, int *outputs, ...)
+{
+ va_list list;
+ struct rtas_args *args;
+ unsigned long flags;
+ int i, ret = 0;
+
+ if (!rtas.entry || token == RTAS_UNKNOWN_SERVICE)
+ return -1;
+
+ local_irq_save(flags);
+ preempt_disable();
+
+ /* We use the per-cpu (PACA) rtas args buffer */
+ args = &local_paca->reentrant_args;
+
+ va_start(list, outputs);
+ va_rtas_call_unlocked(args, token, nargs, nret, list);
+ va_end(list);
+
+ if (nret > 1 && outputs)
+ for (i = 0; i < nret - 1; ++i)
+ outputs[i] = be32_to_cpu(args->rets[i + 1]);
+
+ if (nret > 0)
+ ret = be32_to_cpu(args->rets[0]);
+
+ local_irq_restore(flags);
+ preempt_enable();
+
+ return ret;
+}
+
/* For RTAS_BUSY (-2), delay for 1 millisecond. For an extended busy status
* code of 990n, perform the hinted delay of 10^n (last digit) milliseconds.
*/
diff --git a/arch/powerpc/sysdev/xics/ics-rtas.c b/arch/powerpc/sysdev/xics/ics-rtas.c
index 6aabc74688a6..4cf18000f07c 100644
--- a/arch/powerpc/sysdev/xics/ics-rtas.c
+++ b/arch/powerpc/sysdev/xics/ics-rtas.c
@@ -50,8 +50,8 @@ static void ics_rtas_unmask_irq(struct irq_data *d)

server = xics_get_irq_server(d->irq, irq_data_get_affinity_mask(d), 0);

- call_status = rtas_call(ibm_set_xive, 3, 1, NULL, hw_irq, server,
- DEFAULT_PRIORITY);
+ call_status = rtas_call_reentrant(ibm_set_xive, 3, 1, NULL, hw_irq,
+ server, DEFAULT_PRIORITY);
if (call_status != 0) {
printk(KERN_ERR
"%s: ibm_set_xive irq %u server %x returned %d\n",
@@ -60,7 +60,7 @@ static void ics_rtas_unmask_irq(struct irq_data *d)
}

/* Now unmask the interrupt (often a no-op) */
- call_status = rtas_call(ibm_int_on, 1, 1, NULL, hw_irq);
+ call_status = rtas_call_reentrant(ibm_int_on, 1, 1, NULL, hw_irq);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_int_on irq=%u returned %d\n",
__func__, hw_irq, call_status);
@@ -91,7 +91,7 @@ static void ics_rtas_mask_real_irq(unsigned int hw_irq)
if (hw_irq == XICS_IPI)
return;

- call_status = rtas_call(ibm_int_off, 1, 1, NULL, hw_irq);
+ call_status = rtas_call_reentrant(ibm_int_off, 1, 1, NULL, hw_irq);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_int_off irq=%u returned %d\n",
__func__, hw_irq, call_status);
@@ -99,8 +99,8 @@ static void ics_rtas_mask_real_irq(unsigned int hw_irq)
}

/* Have to set XIVE to 0xff to be able to remove a slot */
- call_status = rtas_call(ibm_set_xive, 3, 1, NULL, hw_irq,
- xics_default_server, 0xff);
+ call_status = rtas_call_reentrant(ibm_set_xive, 3, 1, NULL, hw_irq,
+ xics_default_server, 0xff);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_set_xive(0xff) irq=%u returned %d\n",
__func__, hw_irq, call_status);
@@ -131,7 +131,7 @@ static int ics_rtas_set_affinity(struct irq_data *d,
if (hw_irq == XICS_IPI || hw_irq == XICS_IRQ_SPURIOUS)
return -1;

- status = rtas_call(ibm_get_xive, 1, 3, xics_status, hw_irq);
+ status = rtas_call_reentrant(ibm_get_xive, 1, 3, xics_status, hw_irq);

if (status) {
printk(KERN_ERR "%s: ibm,get-xive irq=%u returns %d\n",
@@ -146,8 +146,8 @@ static int ics_rtas_set_affinity(struct irq_data *d,
return -1;
}

- status = rtas_call(ibm_set_xive, 3, 1, NULL,
- hw_irq, irq_server, xics_status[1]);
+ status = rtas_call_reentrant(ibm_set_xive, 3, 1, NULL,
+ hw_irq, irq_server, xics_status[1]);

if (status) {
printk(KERN_ERR "%s: ibm,set-xive irq=%u returns %d\n",
@@ -179,7 +179,7 @@ static int ics_rtas_map(struct ics *ics, unsigned int virq)
return -EINVAL;

/* Check if RTAS knows about this interrupt */
- rc = rtas_call(ibm_get_xive, 1, 3, status, hw_irq);
+ rc = rtas_call_reentrant(ibm_get_xive, 1, 3, status, hw_irq);
if (rc)
return -ENXIO;

@@ -198,7 +198,7 @@ static long ics_rtas_get_server(struct ics *ics, unsigned long vec)
{
int rc, status[2];

- rc = rtas_call(ibm_get_xive, 1, 3, status, vec);
+ rc = rtas_call_reentrant(ibm_get_xive, 1, 3, status, vec);
if (rc)
return -1;
return status[0];
--
2.25.4