LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* hpet_late_init hang
@ 2008-01-16  3:32 Yinghai Lu
  2008-01-16  6:51 ` Ingo Molnar
  0 siblings, 1 reply; 6+ messages in thread
From: Yinghai Lu @ 2008-01-16  3:32 UTC (permalink / raw)
  To: Ingo Molnar; +Cc: LKML

"
commit e5ed385fa0d6f35406e3e3ed75e5eb9adeb811df
Author: Balaji Rao <balajirrao@gmail.com>
Date:   Tue Jan 15 16:53:29 2008 +0100

    Assign IRQs to HPET Timers
"
in x86.git

cause my servers hang
after
Calling initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100()

after reverting that I got:

initcall 0xffffffff80b947d1 ran for 19 msecs: pci_iommu_init+0x0/0x13()
Calling initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100()
hpet0: at MMIO 0xfed00000, IRQs 2, 8, 31
hpet0: 3 32-bit timers, 25000000 Hz
initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100() returned 0.
initcall 0xffffffff80b9a465 ran for 7 msecs: hpet_late_init+0x0/0x100()

           CPU0       CPU1       CPU2       CPU3       CPU4       CPU5
      CPU6       CPU7
  0:         86          0          0          0          0          0
         1          0   IO-APIC-edge      timer
  4:          0          0          0          0          0          0
         1        838   IO-APIC-edge      serial
  7:          1          0          0          0          0          0
         0          0   IO-APIC-edge
  8:          0          0          0          0          0          0
         0          0   IO-APIC-edge      rtc0

for mcp55, it should already route hpet to ioapic pin2 or the irq0.

YH

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: hpet_late_init hang
  2008-01-16  3:32 hpet_late_init hang Yinghai Lu
@ 2008-01-16  6:51 ` Ingo Molnar
  2008-01-16  8:34   ` Balaji Rao
  0 siblings, 1 reply; 6+ messages in thread
From: Ingo Molnar @ 2008-01-16  6:51 UTC (permalink / raw)
  To: Yinghai Lu; +Cc: LKML, Balaji Rao


(Balaji Cc:-ed)

* Yinghai Lu <yhlu.kernel@gmail.com> wrote:

> "
> commit e5ed385fa0d6f35406e3e3ed75e5eb9adeb811df
> Author: Balaji Rao <balajirrao@gmail.com>
> Date:   Tue Jan 15 16:53:29 2008 +0100
> 
>     Assign IRQs to HPET Timers
> "
> in x86.git
> 
> cause my servers hang
> after
> Calling initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100()

i'm wondering, where does it hang exactly and why?

> after reverting that I got:
> 
> initcall 0xffffffff80b947d1 ran for 19 msecs: pci_iommu_init+0x0/0x13()
> Calling initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100()
> hpet0: at MMIO 0xfed00000, IRQs 2, 8, 31
> hpet0: 3 32-bit timers, 25000000 Hz
> initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100() returned 0.
> initcall 0xffffffff80b9a465 ran for 7 msecs: hpet_late_init+0x0/0x100()
> 
>            CPU0       CPU1       CPU2       CPU3       CPU4       CPU5
>       CPU6       CPU7
>   0:         86          0          0          0          0          0
>          1          0   IO-APIC-edge      timer
>   4:          0          0          0          0          0          0
>          1        838   IO-APIC-edge      serial
>   7:          1          0          0          0          0          0
>          0          0   IO-APIC-edge
>   8:          0          0          0          0          0          0
>          0          0   IO-APIC-edge      rtc0
> 
> for mcp55, it should already route hpet to ioapic pin2 or the irq0.

hm, these new bits:

+       /* Assign IRQs statically for legacy devices */
+       hpetp->hp_dev[0].hd_hdwirq = hdp->hd_irq[0];
+       hpetp->hp_dev[1].hd_hdwirq = hdp->hd_irq[1];

seem to be different from where we came from:

-       for (i = 2; i < nrtimers; timer++, i++)
-               hd.hd_irq[i] = (timer->hpet_config & Tn_INT_ROUTE_CNF_MASK) >>
-                       Tn_INT_ROUTE_CNF_SHIFT;

	Ingo

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: hpet_late_init hang
  2008-01-16  6:51 ` Ingo Molnar
@ 2008-01-16  8:34   ` Balaji Rao
  2008-01-16 21:39     ` Yinghai Lu
  0 siblings, 1 reply; 6+ messages in thread
From: Balaji Rao @ 2008-01-16  8:34 UTC (permalink / raw)
  To: Ingo Molnar; +Cc: Yinghai Lu, LKML

On Wednesday 16 January 2008 12:21:33 pm Ingo Molnar wrote:
Hi Ingo,
> * Yinghai Lu <yhlu.kernel@gmail.com> wrote:
> > "
> > commit e5ed385fa0d6f35406e3e3ed75e5eb9adeb811df
> > Author: Balaji Rao <balajirrao@gmail.com>
> > Date:   Tue Jan 15 16:53:29 2008 +0100
> >
> >     Assign IRQs to HPET Timers
> > "
> > in x86.git
> >
> > cause my servers hang
> > after
> > Calling initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100()
>
> i'm wondering, where does it hang exactly and why?
>
> > after reverting that I got:
> >
> > initcall 0xffffffff80b947d1 ran for 19 msecs: pci_iommu_init+0x0/0x13()
> > Calling initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100()
> > hpet0: at MMIO 0xfed00000, IRQs 2, 8, 31
> > hpet0: 3 32-bit timers, 25000000 Hz
> > initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100() returned 0.
> > initcall 0xffffffff80b9a465 ran for 7 msecs: hpet_late_init+0x0/0x100()
> >
Looks like IRQ 31 is assigned to timer 3, even without the patch! I wonder who wrote the number 31. But the manual says 
that it is zero by default.

I think we should check whether the timer has been allocated an IRQ before proceeding to assign one to it.
Here is a patch that does this.

Yinghai, could you please apply this on top of my patch and check ?

---
Index: linux-2.6/arch/x86/kernel/hpet.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/hpet.c
+++ linux-2.6/arch/x86/kernel/hpet.c
@@ -116,7 +116,8 @@ int is_hpet_enabled(void)
 static void hpet_reserve_platform_timers(unsigned long id)
 {
 	struct hpet __iomem *hpet = hpet_virt_address;
-	unsigned int nrtimers;
+	struct hpet_timer __iomem *timer = &hpet->hpet_timers[2];
+	unsigned int nrtimers, i;
 	struct hpet_data hd;
 
 	nrtimers = ((id & HPET_ID_NUMBER) >> HPET_ID_NUMBER_SHIFT) + 1;
@@ -134,10 +135,9 @@ static void hpet_reserve_platform_timers
 	hd.hd_irq[0] = HPET_LEGACY_8254;
 	hd.hd_irq[1] = HPET_LEGACY_RTC;
 
-	/*
-	 * IRQs for the other timers are assigned dynamically
-	 * in hpet_alloc
-	 */
+       for (i = 2; i < nrtimers; timer++, i++)
+	       hd.hd_irq[i] = (timer->hpet_config & Tn_INT_ROUTE_CNF_MASK) >>
+		       Tn_INT_ROUTE_CNF_SHIFT;
 	hpet_alloc(&hd);
 }
 #else
Index: linux-2.6/drivers/char/hpet.c
===================================================================
--- linux-2.6.orig/drivers/char/hpet.c
+++ linux-2.6/drivers/char/hpet.c
@@ -852,6 +852,12 @@ int hpet_alloc(struct hpet_data *hdp)
 
 		timer = &hpet->hpet_timers[devp - hpetp->hp_dev];
 
+		/* Check if there's already an IRQ assigned to the timer */
+		if (hdp->hd_irq[i]) {
+			hpetp->hp_dev[i].hd_hdwirq = hdp->hd_irq[i];
+			continue;
+		}
+
 		hpet_config = readq(&timer->hpet_config);
 		irq_bitmap = (hpet_config & Tn_INT_ROUTE_CAP_MASK)
 			>> Tn_INT_ROUTE_CAP_SHIFT;

---
regards,
balaji rao

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: hpet_late_init hang
  2008-01-16  8:34   ` Balaji Rao
@ 2008-01-16 21:39     ` Yinghai Lu
  2008-01-17 16:12       ` Balaji Rao
  0 siblings, 1 reply; 6+ messages in thread
From: Yinghai Lu @ 2008-01-16 21:39 UTC (permalink / raw)
  To: Balaji Rao; +Cc: Ingo Molnar, LKML

On Jan 16, 2008 12:34 AM, Balaji Rao <balajirrao@gmail.com> wrote:
> On Wednesday 16 January 2008 12:21:33 pm Ingo Molnar wrote:
> Hi Ingo,
> > * Yinghai Lu <yhlu.kernel@gmail.com> wrote:
> > > "
> > > commit e5ed385fa0d6f35406e3e3ed75e5eb9adeb811df
> > > Author: Balaji Rao <balajirrao@gmail.com>
> > > Date:   Tue Jan 15 16:53:29 2008 +0100
> > >
> > >     Assign IRQs to HPET Timers
> > > "
> > > in x86.git
> > >
> > > cause my servers hang
> > > after
> > > Calling initcall 0xffffffff80b9a465: hpet_late_init+0x0/0x100()
..
> Looks like IRQ 31 is assigned to timer 3, even without the patch! I wonder who wrote the number 31. But the manual says
> that it is zero by default.
>
> I think we should check whether the timer has been allocated an IRQ before proceeding to assign one to it.
> Here is a patch that does this.
>
> Yinghai, could you please apply this on top of my patch and check ?
>
> ---
> Index: linux-2.6/arch/x86/kernel/hpet.c
> ===================================================================
> --- linux-2.6.orig/arch/x86/kernel/hpet.c
...
>

it works

YH

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: hpet_late_init hang
  2008-01-16 21:39     ` Yinghai Lu
@ 2008-01-17 16:12       ` Balaji Rao
  2008-01-18 12:41         ` Ingo Molnar
  0 siblings, 1 reply; 6+ messages in thread
From: Balaji Rao @ 2008-01-17 16:12 UTC (permalink / raw)
  To: Ingo Molnar; +Cc: Yinghai Lu, LKML

On Thursday 17 January 2008 03:09:42 am Yinghai Lu wrote:
> > Looks like IRQ 31 is assigned to timer 3, even without the patch! I wonder who wrote the number 31. But the manual says
> > that it is zero by default.
> >
> > I think we should check whether the timer has been allocated an IRQ before proceeding to assign one to it.
> > Here is a patch that does this.
> >
> > Yinghai, could you please apply this on top of my patch and check ?
> >
> > ---
> > Index: linux-2.6/arch/x86/kernel/hpet.c
> > ===================================================================
> > --- linux-2.6.orig/arch/x86/kernel/hpet.c
> ...
> >
> 
> it works
> 
Cool!

Ingo, here's the patch with the SOB. Please apply.

regards,
balaji rao

Check if there's an IRQ assigned to the timer before assigning one.

Signed-off-by: Balaji Rao <balajirrao@gmail.com>

Index: linux-2.6/arch/x86/kernel/hpet.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/hpet.c
+++ linux-2.6/arch/x86/kernel/hpet.c
@@ -116,7 +116,8 @@ int is_hpet_enabled(void)
 static void hpet_reserve_platform_timers(unsigned long id)
 {
        struct hpet __iomem *hpet = hpet_virt_address;
-       unsigned int nrtimers;
+       struct hpet_timer __iomem *timer = &hpet->hpet_timers[2];
+       unsigned int nrtimers, i;
        struct hpet_data hd;
 
        nrtimers = ((id & HPET_ID_NUMBER) >> HPET_ID_NUMBER_SHIFT) + 1;
@@ -134,10 +135,9 @@ static void hpet_reserve_platform_timers
        hd.hd_irq[0] = HPET_LEGACY_8254;
        hd.hd_irq[1] = HPET_LEGACY_RTC;
 
-       /*
-        * IRQs for the other timers are assigned dynamically
-        * in hpet_alloc
-        */
+       for (i = 2; i < nrtimers; timer++, i++)
+              hd.hd_irq[i] = (timer->hpet_config & Tn_INT_ROUTE_CNF_MASK) >>
+                      Tn_INT_ROUTE_CNF_SHIFT;
        hpet_alloc(&hd);
 }
 #else
Index: linux-2.6/drivers/char/hpet.c
===================================================================
--- linux-2.6.orig/drivers/char/hpet.c
+++ linux-2.6/drivers/char/hpet.c
@@ -852,6 +852,12 @@ int hpet_alloc(struct hpet_data *hdp)
 
                timer = &hpet->hpet_timers[devp - hpetp->hp_dev];
 
+               /* Check if there's already an IRQ assigned to the timer */
+               if (hdp->hd_irq[i]) {
+                       hpetp->hp_dev[i].hd_hdwirq = hdp->hd_irq[i];
+                       continue;
+               }
+
                hpet_config = readq(&timer->hpet_config);
                irq_bitmap = (hpet_config & Tn_INT_ROUTE_CAP_MASK)
                        >> Tn_INT_ROUTE_CAP_SHIFT;

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: hpet_late_init hang
  2008-01-17 16:12       ` Balaji Rao
@ 2008-01-18 12:41         ` Ingo Molnar
  0 siblings, 0 replies; 6+ messages in thread
From: Ingo Molnar @ 2008-01-18 12:41 UTC (permalink / raw)
  To: Balaji Rao; +Cc: Yinghai Lu, LKML


* Balaji Rao <balajirrao@gmail.com> wrote:

> Ingo, here's the patch with the SOB. Please apply.

thanks - reinstated your original patch and added your fix too. I added:

 Signed-off-by: Balaji Rao <balajirrao@gmail.com>
 Tested-by: Yinghai Lu <yhlu.kernel@gmail.com>

	Ingo

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2008-01-18 12:41 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-01-16  3:32 hpet_late_init hang Yinghai Lu
2008-01-16  6:51 ` Ingo Molnar
2008-01-16  8:34   ` Balaji Rao
2008-01-16 21:39     ` Yinghai Lu
2008-01-17 16:12       ` Balaji Rao
2008-01-18 12:41         ` Ingo Molnar

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).