[openib-general] CM misuse of in_atomic/irqs_disabled

[Hal Rosenstock]

Hi Sean,

Should the following in the CM be changed:

int ib_cm_establish(struct ib_cm_id *cm_id)
{
...
        work = kmalloc(sizeof *work, (in_atomic() || irqs_disabled()) ?
                                      GFP_ATOMIC : GFP_KERNEL);
to just
        work = kmalloc(sizeof *work, GFP_ATOMIC);

similar to the other core changes for this issue ?

-- Hal