2011-05-16 18:59:12

by Suresh Siddha

Subject: [patch 1/4] x86, ioapic: fix potential resume deadlock

From: Daniel J Blueman <[email protected]>
Subject: x86, ioapic: fix potential resume deadlock

Fix a potential deadlock when resuming; here the calling function
has disabled interrupts, so we cannot sleep.

Change the memory allocation flag from GFP_KERNEL to GFP_ATOMIC.

TODO: We can do away with this memory allocation during resume by
reusing the ioapic suspend/resume code that uses boot time allocated
buffers.

Signed-off-by: Daniel J Blueman <[email protected]>
Signed-off-by: Suresh Siddha <[email protected]>
Cc: [email protected] [v2.6.39]
---
arch/x86/kernel/apic/io_apic.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

Index: linux-2.6-tip/arch/x86/kernel/apic/io_apic.c
===================================================================
--- linux-2.6-tip.orig/arch/x86/kernel/apic/io_apic.c
+++ linux-2.6-tip/arch/x86/kernel/apic/io_apic.c
@@ -621,14 +621,14 @@ struct IO_APIC_route_entry **alloc_ioapi
struct IO_APIC_route_entry **ioapic_entries;

ioapic_entries = kzalloc(sizeof(*ioapic_entries) * nr_ioapics,
- GFP_KERNEL);
+ GFP_ATOMIC);
if (!ioapic_entries)
return 0;

for (apic = 0; apic < nr_ioapics; apic++) {
ioapic_entries[apic] =
kzalloc(sizeof(struct IO_APIC_route_entry) *
- nr_ioapic_registers[apic], GFP_KERNEL);
+ nr_ioapic_registers[apic], GFP_ATOMIC);
if (!ioapic_entries[apic])
goto nomem;
}

2011-05-16 19:53:14

by Daniel J Blueman

Subject: Re: [patch 1/4] x86, ioapic: fix potential resume deadlock

On 16 May 2011 19:56, Suresh Siddha <[email protected]> wrote:
> From: Daniel J Blueman <[email protected]>
> Subject: x86, ioapic: fix potential resume deadlock
>
> Fix a potential deadlock when resuming; here the calling function
> has disabled interrupts, so we cannot sleep.
>
> Change the memory allocation flag from GFP_KERNEL to GFP_ATOMIC.
>
> TODO: We can do away with this memory allocation during resume by
> reusing the ioapic suspend/resume code that uses boot time allocated
> buffers.
>
> Signed-off-by: Daniel J Blueman <[email protected]>
> Signed-off-by: Suresh Siddha <[email protected]>
> Cc: [email protected] ? [v2.6.39]
> ---
> ?arch/x86/kernel/apic/io_apic.c | ? ?4 ++--
> ?1 file changed, 2 insertions(+), 2 deletions(-)
>
> Index: linux-2.6-tip/arch/x86/kernel/apic/io_apic.c
> ===================================================================
> --- linux-2.6-tip.orig/arch/x86/kernel/apic/io_apic.c
> +++ linux-2.6-tip/arch/x86/kernel/apic/io_apic.c
> @@ -621,14 +621,14 @@ struct IO_APIC_route_entry **alloc_ioapi
> ? ? ? ?struct IO_APIC_route_entry **ioapic_entries;
>
> ? ? ? ?ioapic_entries = kzalloc(sizeof(*ioapic_entries) * nr_ioapics,
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? GFP_KERNEL);
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? GFP_ATOMIC);
> ? ? ? ?if (!ioapic_entries)
> ? ? ? ? ? ? ? ?return 0;
>
> ? ? ? ?for (apic = 0; apic < nr_ioapics; apic++) {
> ? ? ? ? ? ? ? ?ioapic_entries[apic] =
> ? ? ? ? ? ? ? ? ? ? ? ?kzalloc(sizeof(struct IO_APIC_route_entry) *
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? nr_ioapic_registers[apic], GFP_KERNEL);
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? nr_ioapic_registers[apic], GFP_ATOMIC);
> ? ? ? ? ? ? ? ?if (!ioapic_entries[apic])
> ? ? ? ? ? ? ? ? ? ? ? ?goto nomem;
> ? ? ? ?}

I just tested this patch series (1-3) and it does addresses the resume
allocation failures I was consistently seeing (triggering your
warning); it passes nicely with full lock, allocation etc debugging,
so looks all good. Tested-by: Daniel J Blueman
<[email protected]>

Thanks, Suresh!

Daniel
--
Daniel J Blueman