Re: [PATCH] aerdrv: Move cper_print_pcie() out of interrupt context

[Borislav Petkov <bp@alien8.de>]

On Thu, May 09, 2013 at 12:01:21AM +0200, Rafael J. Wysocki wrote:
> On Wednesday, May 08, 2013 11:15:19 AM Lance Ortiz wrote:
> > The following warning was seen on 3.9 when a corrected PCIe error was being
> > handled by the AER subsystem.
> > 
> > WARNING: at .../drivers/pci/search.c:214 pci_get_dev_by_id+0x8a/0x90()
> > 
> > This occurred because code was added to the function cper_print_pcie() that
> > calls the pci_get_domain_bus_and_slot() function.
> 
> Do you know which commit added that code?

1d5210008bd3a26daf4b06aed9d6c330dd4c83e2

> > cper_print_pcie() is called
> > in an interrupt context and pci_get* functions are not supposed to be called
> > in that context hence the warning.
> > 
> > The solution is to move the call to cper_print_aer() out of the interrupt
> > context and into aer_recover_queue() to avoid any warnings when calling
> > pci_get* functions.
> 
> The way the changes are described here isn't particularly clear to me.  I'd say
> something like
> 
>   If cper_print_aer() is called by aer_recover_work_func(), there won't be any
>   reason to call it from cper_print_pcie() any more, in which case all of the
>   problematic code needed only to prepare for the cper_print_aer() call,
>   including the invocation of pci_get_domain_bus_and_slot() causing the warning
>   to be printed, may be removed from there.  Make that happen."
> 
> Also, since aer_recover_work_func() is going to be the only existing caller of
> cper_print_aer() after this change, as far as I can say, and it doesn't use the
> function's first argument, that argument should be dropped entirely.

Hmm, that needs more diddling: AFAICT __ghes_print_estatus() figures out
what the prefix is depending on the ->error_severity coming from the
acpi_hest_generic_status thing so it probably needs to be handed down or
similar...

Fun.

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--