machine_check_early() is taken as an NMI, so nmi_enter() is used
there. machine_check_exception() is no longer taken as an NMI (it's
invoked via irq_work in the case a machine check hits in kernel mode),
so remove the nmi_enter() from that case.
In NMI context, hash faults don't try to refill the hash table, which
can lead to crashes accessing non-pinned kernel pages. System reset
still has this potential problem.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Drop change in show_regs() which breaks Book3E]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200508043408.886394-12-npiggin@gmail.com
long machine_check_early(struct pt_regs *regs)
{
long handled = 0;
+ bool nested = in_nmi();
+ if (!nested)
+ nmi_enter();
hv_nmi_check_nonrecoverable(regs);
*/
if (ppc_md.machine_check_early)
handled = ppc_md.machine_check_early(regs);
+
+ if (!nested)
+ nmi_exit();
+
return handled;
}
void machine_check_exception(struct pt_regs *regs)
{
int recover = 0;
- bool nested = in_nmi();
+ bool nested;
+
+ /*
+ * BOOK3S_64 does not call this handler as a non-maskable interrupt
+ * (it uses its own early real-mode handler to handle the MCE proper
+ * and then raises irq_work to call this handler when interrupts are
+ * enabled). Set nested = true for this case, which just makes it avoid
+ * the nmi_enter/exit.
+ */
+ if (IS_ENABLED(CONFIG_PPC_BOOK3S_64) || in_nmi())
+ nested = true;
+ else
+ nested = false;
if (!nested)
nmi_enter();