Linux MIPS Architecture development
 help / color / mirror / Atom feed
* [PATCH v2] Loongson: Lemote-2F: Fixup of _rdmsr and _wrmsr
@ 2010-03-11  2:57 Wu Zhangjin
  2010-03-11  3:25 ` Wu Zhangjin
  0 siblings, 1 reply; 2+ messages in thread
From: Wu Zhangjin @ 2010-03-11  2:57 UTC (permalink / raw)
  To: Ralf Baechle; +Cc: linux-mips, David Daney, Wu Zhangjin

From: Wu Zhangjin <wuzhangjin@gmail.com>

Changes from the old version:
  o Using raw spinlock instead of spinlock as suggested by David Daney.

The _rdmsr, _wrmsr operation must be atomic to ensure the accessing to msr
address is what we want.

Without this patch, in some cases, the reboot operation(fs2f_reboot) defined in
arch/mips/loongson/lemote-2f/reset.c may fail for it called _rdmsr, _wrmsr but
may be interrupted/preempted by the other related operations and make the
_rdmsr get the wrong value or make the _wrmsr write a wrong value to an
unexpected target.

Signed-off-by: Wu Zhangjin <wuzhangjin@gmail.com>
---
 arch/mips/pci/ops-loongson2.c |   10 ++++++++++
 1 files changed, 10 insertions(+), 0 deletions(-)

diff --git a/arch/mips/pci/ops-loongson2.c b/arch/mips/pci/ops-loongson2.c
index 2bb4057..dc434ec 100644
--- a/arch/mips/pci/ops-loongson2.c
+++ b/arch/mips/pci/ops-loongson2.c
@@ -180,15 +180,21 @@ struct pci_ops loongson_pci_ops = {
 };
 
 #ifdef CONFIG_CS5536
+DEFINE_RAW_SPINLOCK(msr_lock);
+
 void _rdmsr(u32 msr, u32 *hi, u32 *lo)
 {
 	struct pci_bus bus = {
 		.number = PCI_BUS_CS5536
 	};
 	u32 devfn = PCI_DEVFN(PCI_IDSEL_CS5536, 0);
+	unsigned long flags;
+
+	spin_lock_irqsave(&msr_lock, flags);
 	loongson_pcibios_write(&bus, devfn, PCI_MSR_ADDR, 4, msr);
 	loongson_pcibios_read(&bus, devfn, PCI_MSR_DATA_LO, 4, lo);
 	loongson_pcibios_read(&bus, devfn, PCI_MSR_DATA_HI, 4, hi);
+	spin_unlock_irqrestore(&msr_lock, flags);
 }
 EXPORT_SYMBOL(_rdmsr);
 
@@ -198,9 +204,13 @@ void _wrmsr(u32 msr, u32 hi, u32 lo)
 		.number = PCI_BUS_CS5536
 	};
 	u32 devfn = PCI_DEVFN(PCI_IDSEL_CS5536, 0);
+	unsigned long flags;
+
+	spin_lock_irqsave(&msr_lock, flags);
 	loongson_pcibios_write(&bus, devfn, PCI_MSR_ADDR, 4, msr);
 	loongson_pcibios_write(&bus, devfn, PCI_MSR_DATA_LO, 4, lo);
 	loongson_pcibios_write(&bus, devfn, PCI_MSR_DATA_HI, 4, hi);
+	spin_unlock_irqrestore(&msr_lock, flags);
 }
 EXPORT_SYMBOL(_wrmsr);
 #endif
-- 
1.7.0.1

^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH v2] Loongson: Lemote-2F: Fixup of _rdmsr and _wrmsr
  2010-03-11  2:57 [PATCH v2] Loongson: Lemote-2F: Fixup of _rdmsr and _wrmsr Wu Zhangjin
@ 2010-03-11  3:25 ` Wu Zhangjin
  0 siblings, 0 replies; 2+ messages in thread
From: Wu Zhangjin @ 2010-03-11  3:25 UTC (permalink / raw)
  To: Ralf Baechle; +Cc: linux-mips, David Daney

On Thu, 2010-03-11 at 10:57 +0800, Wu Zhangjin wrote:
> From: Wu Zhangjin <wuzhangjin@gmail.com>
> 
> Changes from the old version:
>   o Using raw spinlock instead of spinlock as suggested by David Daney.
[...]
> --- a/arch/mips/pci/ops-loongson2.c
> +++ b/arch/mips/pci/ops-loongson2.c
> @@ -180,15 +180,21 @@ struct pci_ops loongson_pci_ops = {
>  };
>  
>  #ifdef CONFIG_CS5536
> +DEFINE_RAW_SPINLOCK(msr_lock);
> +
>  void _rdmsr(u32 msr, u32 *hi, u32 *lo)
>  {
>  	struct pci_bus bus = {
>  		.number = PCI_BUS_CS5536
>  	};
>  	u32 devfn = PCI_DEVFN(PCI_IDSEL_CS5536, 0);
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&msr_lock, flags);

Please ignore this patch for the operations should be raw_ too.

Regards,
	Wu Zhangjin

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2010-03-11  3:32 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-03-11  2:57 [PATCH v2] Loongson: Lemote-2F: Fixup of _rdmsr and _wrmsr Wu Zhangjin
2010-03-11  3:25 ` Wu Zhangjin

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox