* [PATCH 0/3] xics mask_irq fix and other patches
@ 2011-03-21 18:12 Milton Miller
2011-03-21 18:12 ` [PATCH] powerpc: pseries/smp: query-cpu-stopped-state support won't change Milton Miller
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Milton Miller @ 2011-03-21 18:12 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev
Hi Ben.
I've submitted the fix for the xics mask real irq I mentioned.
During testing I realized current code is passing rtas a linux irq,
which means no interrupt (at least those above NR_IRQs) is being disabled.
I also created a patch to rename irq to hwirq in xics to try to make
the distiction more clear. For your consideration, I don't care if you
don't merge it especially with the upcoming rewrite.
Also, I am sending a printk to printk_once patch that affects BML.
Thanks,
milton
MAIL FROM: <miltonm@bga•com>
RCPT TO: <miltonm@bga•com>
RCPT TO: <linuxppc-dev@lists•ozlabs.org>
RCPT TO: <buytenh@secretlab•ca>
DATA
From: Milton Miller <miltonm@bga•com>
Subject: [PATCH 1/2] powerpc: xics: fix numberspace mismatch from irq_desc conversion
To: linuxppc-dev@lists•ozlabs.org
Cc: Lennert Buytenhek <buytenh@secretlab•ca>
Message-id: <xics-mask-irq-desc-fix@mdm•bga.com>
In-Reply-To: <patches-for-39-rc1@mdm•bga.com>
References: <patches-for-39-rc1@mdm•bga.com>
commit 79f26c268ebad29bd75d078cfc09d3d82b30ccbd (powerpc:
platforms/pseries irq_data conversion) pushed irq_desc down into many
functions, dererencing the descriptor irq field as late as possible.
But it incorrectly passed a linix virtural irq number to RTAS,
resulting in the interrupt not being disabled and possibly
other bad things, such as another interrupt being disabled and/or
a checkstop.
In addition this missed the point of xics_mask_unknown_vec and
the seperation of xics_mask_real_irq from xics_mask_irq. When
xics_mask_unknown_vec is called it's because the hardware delivered an
irq source for which we have no linux irq allocated, and thefore we can
not have an irq_desc allocated.
Revert xics_mask_real_irq to its prior version, naming the argument
hwirq to highlight the difference.
Signed-off-by: Milton Miller <miltonm@bga•com>
--
Lennert can you please review the other patches for similar problems?
Any reference to irq_map[x].hwirq is in a different number domain.
I initially saw the problem pushing irq_desc through unknown_vec
found this due to my knowledge of the split (I added that function)
and only realized the wrong number space when preparing to test the fix.
In fact I tried without the patch and my firmware checkstops the machine
(hardware stops executing instructions)!
diff --git a/arch/powerpc/platforms/pseries/xics.c b/arch/powerpc/platforms/pseries/xics.c
index 01fea46..5686db9 100644
--- a/arch/powerpc/platforms/pseries/xics.c
+++ b/arch/powerpc/platforms/pseries/xics.c
@@ -250,26 +250,26 @@ static unsigned int xics_startup(struct irq_data *d)
return 0;
}
-static void xics_mask_real_irq(struct irq_data *d)
+static void xics_mask_real_irq(unsigned int hwirq)
{
int call_status;
- if (d->irq == XICS_IPI)
+ if (hwirq == XICS_IPI)
return;
- call_status = rtas_call(ibm_int_off, 1, 1, NULL, d->irq);
+ call_status = rtas_call(ibm_int_off, 1, 1, NULL, hwirq);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_int_off irq=%u returned %d\n",
- __func__, d->irq, call_status);
+ __func__, hwirq, call_status);
return;
}
/* Have to set XIVE to 0xff to be able to remove a slot */
- call_status = rtas_call(ibm_set_xive, 3, 1, NULL, d->irq,
+ call_status = rtas_call(ibm_set_xive, 3, 1, NULL, hwirq,
default_server, 0xff);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_set_xive(0xff) irq=%u returned %d\n",
- __func__, d->irq, call_status);
+ __func__, hwirq, call_status);
return;
}
}
@@ -283,13 +283,13 @@ static void xics_mask_irq(struct irq_data *d)
irq = (unsigned int)irq_map[d->irq].hwirq;
if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
return;
- xics_mask_real_irq(d);
+ xics_mask_real_irq(irq);
}
static void xics_mask_unknown_vec(unsigned int vec)
{
printk(KERN_ERR "Interrupt %u (real) is invalid, disabling it.\n", vec);
- xics_mask_real_irq(irq_get_irq_data(vec));
+ xics_mask_real_irq(vec);
}
static inline unsigned int xics_xirr_vector(unsigned int xirr)
^ permalink raw reply related [flat|nested] 4+ messages in thread
* [PATCH] powerpc: pseries/smp: query-cpu-stopped-state support won't change
2011-03-21 18:12 [PATCH 0/3] xics mask_irq fix and other patches Milton Miller
@ 2011-03-21 18:12 ` Milton Miller
2011-03-21 18:12 ` [PATCH 2/2] powerpc: xics: use hwirq for xics domain irq number Milton Miller
2011-03-21 21:38 ` [PATCH 1/2] powerpc: xics: fix numberspace mismatch from irq_desc conversion Milton Miller
2 siblings, 0 replies; 4+ messages in thread
From: Milton Miller @ 2011-03-21 18:12 UTC (permalink / raw)
To: linuxppc-dev
If a given firmware doesn't have a token to support query-cpu-stopped-state,
its not likely to change during the lifetime of the kernel.
Only print this information once, not once per secondary thread.
While here, make the line wrap grep friendly.
Signed-off-by: Milton Miller <miltonm@bga•com>
Index: work.git/arch/powerpc/platforms/pseries/smp.c
===================================================================
--- work.git.orig/arch/powerpc/platforms/pseries/smp.c 2011-03-21 11:27:54.000000000 -0500
+++ work.git/arch/powerpc/platforms/pseries/smp.c 2011-03-21 11:29:51.000000000 -0500
@@ -64,8 +64,8 @@ int smp_query_cpu_stopped(unsigned int p
int qcss_tok = rtas_token("query-cpu-stopped-state");
if (qcss_tok == RTAS_UNKNOWN_SERVICE) {
- printk(KERN_INFO "Firmware doesn't support "
- "query-cpu-stopped-state\n");
+ printk_once(KERN_INFO
+ "Firmware doesn't support query-cpu-stopped-state\n");
return QCSS_HARDWARE_ERROR;
}
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH 2/2] powerpc: xics: use hwirq for xics domain irq number
2011-03-21 18:12 [PATCH 0/3] xics mask_irq fix and other patches Milton Miller
2011-03-21 18:12 ` [PATCH] powerpc: pseries/smp: query-cpu-stopped-state support won't change Milton Miller
@ 2011-03-21 18:12 ` Milton Miller
2011-03-21 21:38 ` [PATCH 1/2] powerpc: xics: fix numberspace mismatch from irq_desc conversion Milton Miller
2 siblings, 0 replies; 4+ messages in thread
From: Milton Miller @ 2011-03-21 18:12 UTC (permalink / raw)
To: linuxppc-dev
To try to avoid future confusion, rename irq to hwirq when it refers
to a xics domain number instead of a linux irq number.
Signed-off-by: Milton Miller <miltonm@bga•com>
Index: next.git/arch/powerpc/platforms/pseries/xics.c
===================================================================
--- next.git.orig/arch/powerpc/platforms/pseries/xics.c 2011-03-19 23:21:21.775085058 -0500
+++ next.git/arch/powerpc/platforms/pseries/xics.c 2011-03-20 00:57:33.144560092 -0500
@@ -204,33 +204,33 @@ static int get_irq_server(unsigned int v
static void xics_unmask_irq(struct irq_data *d)
{
- unsigned int irq;
+ unsigned int hwirq;
int call_status;
int server;
pr_devel("xics: unmask virq %d\n", d->irq);
- irq = (unsigned int)irq_map[d->irq].hwirq;
- pr_devel(" -> map to hwirq 0x%x\n", irq);
- if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+ hwirq = (unsigned int)irq_map[d->irq].hwirq;
+ pr_devel(" -> map to hwirq 0x%x\n", hwirq);
+ if (hwirq == XICS_IPI || hwirq == XICS_IRQ_SPURIOUS)
return;
server = get_irq_server(d->irq, d->affinity, 0);
- call_status = rtas_call(ibm_set_xive, 3, 1, NULL, irq, server,
+ call_status = rtas_call(ibm_set_xive, 3, 1, NULL, hwirq, server,
DEFAULT_PRIORITY);
if (call_status != 0) {
printk(KERN_ERR
"%s: ibm_set_xive irq %u server %x returned %d\n",
- __func__, irq, server, call_status);
+ __func__, hwirq, server, call_status);
return;
}
/* Now unmask the interrupt (often a no-op) */
- call_status = rtas_call(ibm_int_on, 1, 1, NULL, irq);
+ call_status = rtas_call(ibm_int_on, 1, 1, NULL, hwirq);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_int_on irq=%u returned %d\n",
- __func__, irq, call_status);
+ __func__, hwirq, call_status);
return;
}
}
@@ -276,14 +276,14 @@ static void xics_mask_real_irq(unsigned
static void xics_mask_irq(struct irq_data *d)
{
- unsigned int irq;
+ unsigned int hwirq;
pr_devel("xics: mask virq %d\n", d->irq);
- irq = (unsigned int)irq_map[d->irq].hwirq;
- if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+ hwirq = (unsigned int)irq_map[d->irq].hwirq;
+ if (hwirq == XICS_IPI || hwirq == XICS_IRQ_SPURIOUS)
return;
- xics_mask_real_irq(irq);
+ xics_mask_real_irq(hwirq);
}
static void xics_mask_unknown_vec(unsigned int vec)
@@ -373,37 +373,37 @@ static unsigned char pop_cppr(void)
static void xics_eoi_direct(struct irq_data *d)
{
- unsigned int irq = (unsigned int)irq_map[d->irq].hwirq;
+ unsigned int hwirq = (unsigned int)irq_map[d->irq].hwirq;
iosync();
- direct_xirr_info_set((pop_cppr() << 24) | irq);
+ direct_xirr_info_set((pop_cppr() << 24) | hwirq);
}
static void xics_eoi_lpar(struct irq_data *d)
{
- unsigned int irq = (unsigned int)irq_map[d->irq].hwirq;
+ unsigned int hwirq = (unsigned int)irq_map[d->irq].hwirq;
iosync();
- lpar_xirr_info_set((pop_cppr() << 24) | irq);
+ lpar_xirr_info_set((pop_cppr() << 24) | hwirq);
}
static int
xics_set_affinity(struct irq_data *d, const struct cpumask *cpumask, bool force)
{
- unsigned int irq;
+ unsigned int hwirq;
int status;
int xics_status[2];
int irq_server;
- irq = (unsigned int)irq_map[d->irq].hwirq;
- if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+ hwirq = (unsigned int)irq_map[d->irq].hwirq;
+ if (hwirq == XICS_IPI || hwirq == XICS_IRQ_SPURIOUS)
return -1;
- status = rtas_call(ibm_get_xive, 1, 3, xics_status, irq);
+ status = rtas_call(ibm_get_xive, 1, 3, xics_status, hwirq);
if (status) {
printk(KERN_ERR "%s: ibm,get-xive irq=%u returns %d\n",
- __func__, irq, status);
+ __func__, hwirq, status);
return -1;
}
@@ -418,11 +418,11 @@ xics_set_affinity(struct irq_data *d, co
}
status = rtas_call(ibm_set_xive, 3, 1, NULL,
- irq, irq_server, xics_status[1]);
+ hwirq, irq_server, xics_status[1]);
if (status) {
printk(KERN_ERR "%s: ibm,set-xive irq=%u returns %d\n",
- __func__, irq, status);
+ __func__, hwirq, status);
return -1;
}
@@ -874,7 +874,7 @@ void xics_kexec_teardown_cpu(int seconda
void xics_migrate_irqs_away(void)
{
int cpu = smp_processor_id(), hw_cpu = hard_smp_processor_id();
- unsigned int irq, virq;
+ int virq;
/* If we used to be the default server, move to the new "boot_cpuid" */
if (hw_cpu == default_server)
@@ -892,6 +892,7 @@ void xics_migrate_irqs_away(void)
for_each_irq(virq) {
struct irq_desc *desc;
struct irq_chip *chip;
+ unsigned int hwirq;
int xics_status[2];
int status;
unsigned long flags;
@@ -901,9 +902,9 @@ void xics_migrate_irqs_away(void)
continue;
if (irq_map[virq].host != xics_host)
continue;
- irq = (unsigned int)irq_map[virq].hwirq;
+ hwirq = (unsigned int)irq_map[virq].hwirq;
/* We need to get IPIs still. */
- if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+ if (hwirq == XICS_IPI || hwirq == XICS_IRQ_SPURIOUS)
continue;
desc = irq_to_desc(virq);
@@ -918,10 +919,10 @@ void xics_migrate_irqs_away(void)
raw_spin_lock_irqsave(&desc->lock, flags);
- status = rtas_call(ibm_get_xive, 1, 3, xics_status, irq);
+ status = rtas_call(ibm_get_xive, 1, 3, xics_status, hwirq);
if (status) {
printk(KERN_ERR "%s: ibm,get-xive irq=%u returns %d\n",
- __func__, irq, status);
+ __func__, hwirq, status);
goto unlock;
}
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH 1/2] powerpc: xics: fix numberspace mismatch from irq_desc conversion
2011-03-21 18:12 [PATCH 0/3] xics mask_irq fix and other patches Milton Miller
2011-03-21 18:12 ` [PATCH] powerpc: pseries/smp: query-cpu-stopped-state support won't change Milton Miller
2011-03-21 18:12 ` [PATCH 2/2] powerpc: xics: use hwirq for xics domain irq number Milton Miller
@ 2011-03-21 21:38 ` Milton Miller
2 siblings, 0 replies; 4+ messages in thread
From: Milton Miller @ 2011-03-21 21:38 UTC (permalink / raw)
To: linuxppc-dev; +Cc: Lennert Buytenhek
commit 79f26c268ebad29bd75d078cfc09d3d82b30ccbd (powerpc:
platforms/pseries irq_data conversion) pushed irq_desc down into many
functions, dererencing the descriptor irq field as late as possible.
But it incorrectly passed a linix virtural irq number to RTAS,
resulting in the interrupt not being disabled and possibly
other bad things, such as another interrupt being disabled and/or
a checkstop.
In addition this missed the point of xics_mask_unknown_vec and
the seperation of xics_mask_real_irq from xics_mask_irq. When
xics_mask_unknown_vec is called it's because the hardware delivered an
irq source for which we have no linux irq allocated, and thefore we can
not have an irq_desc allocated.
Revert xics_mask_real_irq to its prior version, naming the argument
hwirq to highlight the difference.
Signed-off-by: Milton Miller <miltonm@bga•com>
--
Lennert can you please review the other patches for similar problems?
Any reference to irq_map[x].hwirq is in a different number domain.
I initially saw the problem pushing irq_desc through unknown_vec
found this due to my knowledge of the split (I added that function)
and only realized the wrong number space when preparing to test the fix.
In fact I tried without the patch and my firmware checkstops the machine
(hardware stops executing instructions)!
diff --git a/arch/powerpc/platforms/pseries/xics.c b/arch/powerpc/platforms/pseries/xics.c
index 01fea46..5686db9 100644
--- a/arch/powerpc/platforms/pseries/xics.c
+++ b/arch/powerpc/platforms/pseries/xics.c
@@ -250,26 +250,26 @@ static unsigned int xics_startup(struct irq_data *d)
return 0;
}
-static void xics_mask_real_irq(struct irq_data *d)
+static void xics_mask_real_irq(unsigned int hwirq)
{
int call_status;
- if (d->irq == XICS_IPI)
+ if (hwirq == XICS_IPI)
return;
- call_status = rtas_call(ibm_int_off, 1, 1, NULL, d->irq);
+ call_status = rtas_call(ibm_int_off, 1, 1, NULL, hwirq);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_int_off irq=%u returned %d\n",
- __func__, d->irq, call_status);
+ __func__, hwirq, call_status);
return;
}
/* Have to set XIVE to 0xff to be able to remove a slot */
- call_status = rtas_call(ibm_set_xive, 3, 1, NULL, d->irq,
+ call_status = rtas_call(ibm_set_xive, 3, 1, NULL, hwirq,
default_server, 0xff);
if (call_status != 0) {
printk(KERN_ERR "%s: ibm_set_xive(0xff) irq=%u returned %d\n",
- __func__, d->irq, call_status);
+ __func__, hwirq, call_status);
return;
}
}
@@ -283,13 +283,13 @@ static void xics_mask_irq(struct irq_data *d)
irq = (unsigned int)irq_map[d->irq].hwirq;
if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
return;
- xics_mask_real_irq(d);
+ xics_mask_real_irq(irq);
}
static void xics_mask_unknown_vec(unsigned int vec)
{
printk(KERN_ERR "Interrupt %u (real) is invalid, disabling it.\n", vec);
- xics_mask_real_irq(irq_get_irq_data(vec));
+ xics_mask_real_irq(vec);
}
static inline unsigned int xics_xirr_vector(unsigned int xirr)
^ permalink raw reply related [flat|nested] 4+ messages in thread
end of thread, other threads:[~2011-03-21 21:38 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-03-21 18:12 [PATCH 0/3] xics mask_irq fix and other patches Milton Miller
2011-03-21 18:12 ` [PATCH] powerpc: pseries/smp: query-cpu-stopped-state support won't change Milton Miller
2011-03-21 18:12 ` [PATCH 2/2] powerpc: xics: use hwirq for xics domain irq number Milton Miller
2011-03-21 21:38 ` [PATCH 1/2] powerpc: xics: fix numberspace mismatch from irq_desc conversion Milton Miller
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox