| Message ID | 1386076560-1886-1-git-send-email-nschichan@freebox.fr (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Dec 03, 2013 at 02:16:00PM +0100, Nicolas Schichan wrote: > If a timer interrupt kicks right after the wait_event_timeout() call, > we may endup a reporting timeout before the timeout really occured. Please see my reply to Jason Cooper about this in the thread "Spurious timeouts in mvmdio".
diff --git a/drivers/net/ethernet/marvell/mvmdio.c b/drivers/net/ethernet/marvell/mvmdio.c index 7354960..2814bbc 100644 --- a/drivers/net/ethernet/marvell/mvmdio.c +++ b/drivers/net/ethernet/marvell/mvmdio.c @@ -76,9 +76,19 @@ static int orion_mdio_wait_ready(struct mii_bus *bus) { struct orion_mdio_dev *dev = bus->priv; unsigned long timeout = usecs_to_jiffies(MVMDIO_SMI_TIMEOUT); - unsigned long end = jiffies + timeout; + unsigned long end; int timedout = 0; + /* + * make sure that we wait at least more than one + * jiffy. wait_event_timeout() with a timeout parameter of 1 + * jiffy may return before the SMI access is done if a timer + * interrupt kicks immediately after. + */ + if (timeout < 2) + timeout = 2; + end = jiffies + timeout; + while (1) { if (orion_mdio_smi_is_done(dev)) return 0;
If a timer interrupt kicks right after the wait_event_timeout() call, we may endup a reporting timeout before the timeout really occured. Signed-off-by: Nicolas Schichan <nschichan@freebox.fr> --- drivers/net/ethernet/marvell/mvmdio.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)