sparc64: Fix lost interrupts on sun4u. (d0cac39e) · Commits · 戴 / test

Based upon a report by Meelis Roos. Sparc64 SBUS and PCI controllers use a combination of IMAP and ICLR registers to manage device interrupts. The IMAP register contains the "valid" enable bit as well as CPU targetting information. Whereas the ICLR register is written with zero at the end of handling an interrupt to reset the state machine for that interrupt to IDLE so it can be sent again. For PCI slot and SBUS slot devices we can have multiple interrupts sharing the same IMAP register. There are individual ICLR registers but only one IMAP register for managing those. We represent each shared case with individual virtual IRQs so the generic IRQ layer thinks there is only one user of the IRQ instance. In such shared IMAP cases this is wrong, so if there are multiple active users then a free_irq() call will prematurely turn off the interrupt by clearing the Valid bit in the IMAP register even though there are other active users. Fix this by simply doing nothing in sun4u_disable_irq() and checking IRQF_DISABLED during IRQ dispatch. This situation doesn't exist in the hypervisor sun4v cases, so I left those alone. Tested-by:

Meelis Roos <mroos@linux.ee> Signed-off-by:

David S. Miller <davem@davemloft.net>

arch/sparc/kernel/irq_64.c

+19 −10

Original line number	Diff line number	Diff line
		@@ -323,17 +323,25 @@ static void sun4u_set_affinity(unsigned int virt_irq,
		sun4u_irq_enable(virt_irq);
		}

		/* Don't do anything. The desc->status check for IRQ_DISABLED in
		* handler_irq() will skip the handler call and that will leave the
		* interrupt in the sent state. The next ->enable() call will hit the
		* ICLR register to reset the state machine.
		*
		* This scheme is necessary, instead of clearing the Valid bit in the
		* IMAP register, to handle the case of IMAP registers being shared by
		* multiple INOs (and thus ICLR registers). Since we use a different
		* virtual IRQ for each shared IMAP instance, the generic code thinks
		* there is only one user so it prematurely calls ->disable() on
		* free_irq().
		*
		* We have to provide an explicit ->disable() method instead of using
		* NULL to get the default. The reason is that if the generic code
		* sees that, it also hooks up a default ->shutdown method which
		* invokes ->mask() which we do not want. See irq_chip_set_defaults().
		*/
		static void sun4u_irq_disable(unsigned int virt_irq)
		{
		struct irq_handler_data *data = get_irq_chip_data(virt_irq);

		if (likely(data)) {
		unsigned long imap = data->imap;
		unsigned long tmp = upa_readq(imap);

		tmp &= ~IMAP_VALID;
		upa_writeq(tmp, imap);
		}
		}

		static void sun4u_irq_eoi(unsigned int virt_irq)
		@@ -746,6 +754,7 @@ void handler_irq(int irq, struct pt_regs *regs)

		desc = irq_desc + virt_irq;

		if (!(desc->status & IRQ_DISABLED))
		desc->handle_irq(virt_irq, desc);

		bucket_pa = next_pa;

Admin message