Re: [PATCH] e1000: fix IRQx nobody cared for shared irq with INTx

[Jeff Garzik <jeff@garzik.org>]

Yinghai Lu wrote:
> when try to kexec one latest kernel from kernel.org from RHEL 5.1 got
> 
> ACPI: PCI Interrupt 0000:02:00.0[A] -> Link [LNKA] -> GSI 19 (level, low) -> IRQ 19
> acpi->mptable 2 : Int: type 0, pol 1, trig 1, bus 02, IRQ 00, APIC ID 0, APIC INT 13
> PCI: Setting latency timer of device 0000:02:00.0 to 64
> PCI: Enabling Mem-Wr-Inval for device 0000:02:00.0
> scsi0 :  on PCI bus 02 device 00 irq 19
> irq 19: nobody cared (try booting with the "irqpoll" option)
> Pid: 1, comm: swapper Not tainted 2.6.24-smp-07682-g551e4fb-dirty #19
> 
> Call Trace:
>  <IRQ>  [<ffffffff8026a3eb>] __report_bad_irq+0x30/0x72
>  [<ffffffff8026a651>] note_interrupt+0x224/0x26f
>  [<ffffffff8026ae78>] handle_fasteoi_irq+0xa5/0xc8
>  [<ffffffff8021ffdc>] call_softirq+0x1c/0x28
>  [<ffffffff802218e2>] do_IRQ+0xf1/0x15f
>  [<ffffffff8021f361>] ret_from_intr+0x0/0xa
>  <EOI>  [<ffffffff80785bd6>] pci_mmcfg_write+0x0/0xb0
>  [<ffffffff80224dd9>] native_read_tsc+0xd/0x1d
>  [<ffffffff804681e7>] __delay+0x17/0x22
>  [<ffffffff805cb1a9>] lpfc_sli_brdrestart+0x14c/0x16b
>  [<ffffffff805cb264>] lpfc_do_config_port+0x9c/0x3e4
>  [<ffffffff802d77a9>] sysfs_link_sibling+0x17/0x31
>  [<ffffffff805cb674>] lpfc_sli_hba_setup+0xc8/0x4a2
>  [<ffffffff80825397>] lpfc_pci_probe_one+0x750/0x914
>  [<ffffffff804728f3>] pci_device_probe+0xb3/0xfb
>  [<ffffffff804d958c>] driver_probe_device+0xb5/0x132
>  [<ffffffff804d96ab>] __driver_attach+0x0/0x93
>  [<ffffffff804d9705>] __driver_attach+0x5a/0x93
>  [<ffffffff804d89bc>] bus_for_each_dev+0x44/0x6f
>  [<ffffffff804d91b9>] bus_add_driver+0xae/0x1f5
>  [<ffffffff804d990e>] driver_register+0x59/0xce
>  [<ffffffff80472b56>] __pci_register_driver+0x4a/0x7c
>  [<ffffffff80c2d78a>] lpfc_init+0x98/0xba
>  [<ffffffff80c0d6e7>] kernel_init+0x175/0x2e1
>  [<ffffffff8021fc68>] child_rip+0xa/0x12
>  [<ffffffff80c0d572>] kernel_init+0x0/0x2e1
>  [<ffffffff8021fc5e>] child_rip+0x0/0x12
> 
> handlers:
> [<ffffffff805cdbce>] (lpfc_intr_handler+0x0/0x4c6)
> Disabling IRQ #19
> 
> root caused that there is one Intel card that shared io apic pin and irq with
>  lpfc
> 
> e1000_probe path only use pci_enable_device to setup irq entry but masked, and
> will use e1000_open to use request_irq/setup_irq to install action and
> enable/unmask that io apic entry.
> 
> but lpfc driver will call it's probe and request_irq/setup_irq. so it
> enable/umask that io apic entry. and only lpfc's action the lpfc_intr_handler
> is installed.
> 
> and some case, the e1000 sent out irq (hw bug or first kernel doesn't call
> e1000_irq_disable?)
> that irq will confuse the hanlder ... it is not for lpfc_intr_handler...
> 
> So try to call pci_intx(dev, 0) in e1000_probe,
> and later call pci_intx(dev, 1) after request_irq in e1000_open patch, if the
> irq is using INTx
> 
> even e1000 is using MSI, still need this patch. Because even pci_enable_msi in
> e1000_open path will call pci_intx(dev, 0), that is too late. when we have lpfc
> driver loaded before use ifconfig to set network connection.
> 
> othe drivers may need to be updated in the same way, if they have same problem
> like nobody cared irq with shared INTx irq.
> 
> Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
> 
> Index: linux-2.6/drivers/net/e1000/e1000_main.c
> ===================================================================
> --- linux-2.6.orig/drivers/net/e1000/e1000_main.c
> +++ linux-2.6/drivers/net/e1000/e1000_main.c
> @@ -324,6 +324,9 @@ static int e1000_request_irq(struct e100
>  			pci_disable_msi(adapter->pdev);
>  		DPRINTK(PROBE, ERR,
>  		        "Unable to allocate interrupt Error: %d\n", err);
> +	} else if (!adapter->have_msi) {
> +		/* enable INTx before if not using MSI */
> +		pci_intx(adapter->pdev, 1);
>  	}
>  
>  	return err;
> @@ -934,6 +937,8 @@ e1000_probe(struct pci_dev *pdev,
>  	uint16_t eeprom_apme_mask = E1000_EEPROM_APME;
>  	DECLARE_MAC_BUF(mac);
>  
> +	/* disable INTx at first */
> +	pci_intx(pdev, 0);
>  	if ((err = pci_enable_device(pdev)))
>  		return err;
>  
> Index: linux-2.6/drivers/net/e1000e/netdev.c
> ===================================================================
> --- linux-2.6.orig/drivers/net/e1000e/netdev.c
> +++ linux-2.6/drivers/net/e1000e/netdev.c
> @@ -960,6 +960,9 @@ static int e1000_request_irq(struct e100
>  			err);
>  		if (adapter->flags & FLAG_MSI_ENABLED)
>  			pci_disable_msi(adapter->pdev);
> +	} else if (!(adapter->flags & FLAG_MSI_ENABLED)) {
> +		/* enable INTx before if not using MSI */
> +		pci_intx(adapter->pdev, 1);
>  	}
>  
>  	return err;

These seem sane.


> @@ -3726,6 +3729,8 @@ static int __devinit e1000_probe(struct 
>  	u16 eeprom_apme_mask = E1000_EEPROM_APME;
>  
>  	e1000e_disable_l1aspm(pdev);
> +	/* disable INTx at first */
> +	pci_intx(pdev, 0);
>  	err = pci_enable_device(pdev);
>  	if (err)
>  		return err;

Any pci_* call before pci_enable_device() is questionable.  I would put 
it after pci_enable_device(), unless there is a _strong_ reason.

PCI devices are not considered available, with resources assigned, until 
pci_enable_device()

I am also curious what irq events are being raised?  That seems like 
another problem area to address, since pci_intx() is just a band-aid 
hiding that behavior.

	Jeff