On Wed, Feb 23, 2011 at 11:43 AM, Tony Luck <tony.luck@xxxxxxxxx> wrote:
> On Tue, Feb 22, 2011 at 5:53 PM, Mike Waychison <mikew@xxxxxxxxxx> wrote:
>> +static void __init dmi_sysfs_register_handle(const struct dmi_header *dh,
>> + void *_ret)
>> +{
>
> I'm seeing some unaligned references from this function
> on ia64. Code is loading a 2-byte value from an odd
> address. We are dereferencing "dh", so looking at the
> dmi_header definition, I'd have thought that we must
> be touching dh->handle:
>
> struct dmi_header {
> u8 type;
> u8 length;
> u16 handle;
> };
>
> I don't see any code in this function that does this,
> so I assume there is some inlining or macro stuff
> happening.
Ya, I don't think there is anything that guarantees that dmi_header is
aligned. The access is probably the struct copy I did here:
/* Set the key */
entry->dh = *dh;
What if we changed that guy to use memcpy? We'd probably need to
memcpy the dmi_system_event_log as well. Patch attached (applies on
top of the entire patchset).
Attachment:
patch
Description: Binary data
