Hi,
On Mon, Feb 27, 2012 at 07:03:49PM -0800, Kuninori Morimoto wrote:
> current renesas_usbhs host driver will get below WARNING when host was removed
>
> WARNING: at /opt/usr/src/WORK/morimoto/gitlinux/linux-2.6/kernel/irq/manage.c:1)
> Trying to free already-free IRQ 0
> Modules linked in:
> [<c000e25c>] (unwind_backtrace+0x0/0xe4) from [<c001695c>] (warn_slowpath_commo)
> [<c001695c>] (warn_slowpath_common+0x4c/0x64) from [<c00169f4>] (warn_slowpath_)
> [<c00169f4>] (warn_slowpath_fmt+0x2c/0x3c) from [<c004978c>] (__free_irq+0x88/0)
> [<c004978c>] (__free_irq+0x88/0x148) from [<c00498a0>] (free_irq+0x54/0x78)
> [<c00498a0>] (free_irq+0x54/0x78) from [<c01b023c>] (usb_remove_hcd+0x13c/0x170)
> [<c01b023c>] (usb_remove_hcd+0x13c/0x170) from [<c01c7b1c>] (usbhsh_stop+0x44/0)
> ...
>
> renesas_usbhs host hcd->irq should be -1 since
> this driver doesn't have hcd->driver->irq.
>
> But, from 68d07f64b8a11a852d48d1b05b724c3e20c0d94b,
> (USB: Don't fail USB3 probe on missing legacy PCI IRQ.)
> current renesas_usbhs didn't call usb_hcd_request_irqs()
renesas is not supposed to call that. usb_add_hcd() is calling that,
which makes me wonder if this patch is really needed.
> which will set hcd->irq on usb_add_hcd().
> Because the "irqnum" was 0.
>
> This patch use irqnum = -1 for usb_add_hcd().
>
> Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@xxxxxxxxxxx>
> ---
>
> >> Felipe
>
> This patch is for linus/master
>
>
> drivers/usb/renesas_usbhs/mod_host.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/usb/renesas_usbhs/mod_host.c b/drivers/usb/renesas_usbhs/mod_host.c
> index 1834cf5..f9c668b 100644
> --- a/drivers/usb/renesas_usbhs/mod_host.c
> +++ b/drivers/usb/renesas_usbhs/mod_host.c
> @@ -1441,7 +1441,7 @@ static int usbhsh_start(struct usbhs_priv *priv)
> int ret;
>
> /* add hcd */
> - ret = usb_add_hcd(hcd, 0, 0);
> + ret = usb_add_hcd(hcd, -1, 0);
what will happen here is that usb_add_hcd() will call
usb_hcd_request_irqs() which will fall into the else branch:
2325 static int usb_hcd_request_irqs(struct usb_hcd *hcd,
2326 unsigned int irqnum, unsigned long irqflags)
2327 {
2328 int retval;
2329
2330 if (hcd->driver->irq) {
2331
2332 /* IRQF_DISABLED doesn't work as advertised when used together
2333 * with IRQF_SHARED. As usb_hcd_irq() will always disable
2334 * interrupts we can remove it here.
2335 */
2336 if (irqflags & IRQF_SHARED)
2337 irqflags &= ~IRQF_DISABLED;
2338
2339 snprintf(hcd->irq_descr, sizeof(hcd->irq_descr), "%s:usb%d",
2340 hcd->driver->description, hcd->self.busnum);
2341 retval = request_irq(irqnum, &usb_hcd_irq, irqflags,
2342 hcd->irq_descr, hcd);
2343 if (retval != 0) {
2344 dev_err(hcd->self.controller,
2345 "request interrupt %d failed\n",
2346 irqnum);
2347 return retval;
2348 }
2349 hcd->irq = irqnum;
2350 dev_info(hcd->self.controller, "irq %d, %s 0x%08llx\n", irqnum,
2351 (hcd->driver->flags & HCD_MEMORY) ?
2352 "io mem" : "io base",
2353 (unsigned long long)hcd->rsrc_start);
2354 } else {
2355 hcd->irq = -1;
2356 if (hcd->rsrc_start)
2357 dev_info(hcd->self.controller, "%s 0x%08llx\n",
2358 (hcd->driver->flags & HCD_MEMORY) ?
2359 "io mem" : "io base",
2360 (unsigned long long)hcd->rsrc_start);
2361 }
2362 return 0;
2363 }
Are you sure you never initialize driver->irq by mistake ??
--
balbi
Attachment:
signature.asc
Description: Digital signature
