On Wed, 16 Aug 2006, Len Brown wrote:
> > acpi_pci_link_set() allocates with GFP_ATOMIC. On resume from suspend,
> > this is called with interrupts off, otherwise GFP_KERNEL is safe.
> So you are suggesting this?
> diff --git a/drivers/acpi/pci_link.c b/drivers/acpi/pci_link.c
> index 7f3e7e7..d53bd98 100644
> --- a/drivers/acpi/pci_link.c
> +++ b/drivers/acpi/pci_link.c
> @@ -307,7 +307,7 @@ static int acpi_pci_link_set(struct acpi
> if (!link || !irq)
> return -EINVAL;
>
> - resource = kmalloc(sizeof(*resource) + 1, GFP_ATOMIC);
> + resource = kmalloc(sizeof(*resource) + 1, irqs_disabled() ? GFP_ATOMIC: GFP_KERNEL);
> if (!resource)
> return -ENOMEM;
>
>
Yes, exactly.
> > acpi_pci_link_resume -> acpi_pci_link_set -> acpi_set_current_resources ->
> > acpi_rs_set_srs_method_data -> acpi_ut_create_internal_object_dbg ->
> > acpi_ut_allocate_object_desc_dbg -> acpi_os_acquire_object ->
> > kmem_cache_alloc(GFP_KERNEL) flag.
> I don't understand this comment.
> acpi_os_acquire_object is implemented in aclinux.h:
> static inline void *acpi_os_acquire_object(acpi_cache_t * cache) {
> return kmem_cache_zalloc(cache, irqs_disabled() ? GFP_ATOMIC : GFP_KERNEL);
> }
This happened in some -rc of 2.6.18 (can be seen at
http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fv2.6%2Ftesting%2Fpatch-2.6.18-rc4.bz2;z=2727)
- the acpi_os_acquire_object() was renamed to acpi_os_validate_interface()
... which allocates solely with GFP_KERNEL, but this may be OK, I haven't
checked.
Anyway ...
Unfortunately, looking at the refactorized ACPI code in
2.6.18-rc4, there are still issues with sleeping functions called with
disabled interrupts (during resume), in ACPI code.
Two random examples:
- when acpi_pci_link_set() is called during resume (local irqs off), the
following callchain happens, which is bad: acpi_pci_link_resume ->
acpi_pci_link_set -> acpi_set_current_resources ->
acpi_rs_set_srs_method_data -> acpi_ns_evaluate -> acpi_ns_get_node ..
here the mutex is acquired. Not good.
- device_power_up -> sysdev_resume -> __sysdev_resume -> cpufreq_resume ->
blocking_notifier_call_chain -> down on semaphore. Not good.
Is there any general idea for solution?
Thanks,
--
JiKos.
-
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
