| Message ID | 1432032711-7020-1-git-send-email-hofrat@osadl.org |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
On Tue, 2015-05-19 at 12:51 +0200, Nicholas Mc Guire wrote: > Converting milliseconds to jiffies by "val * HZ / 1000" is technically > is not a clean solution as it does not handle all corner cases correctly. > By changing the conversion to use msecs_to_jiffies(val) conversion is > correct in all cases. > > in the current code: > mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); > for HZ < 100 (e.g. CONFIG_HZ == 64|32 in alpha) this effectively results > in no delay at all. [] > diff --git a/drivers/net/irda/irda-usb.c b/drivers/net/irda/irda-usb.c [] > @@ -848,7 +848,9 @@ static void irda_usb_receive(struct urb *urb) > * Jean II */ > self->rx_defer_timer.function = irda_usb_rx_defer_expired; > self->rx_defer_timer.data = (unsigned long) urb; > - mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); > + mod_timer(&self->rx_defer_timer, > + jiffies + (msecs_to_jiffies(10))); The unnecessary () around msecs_to_jiffies could be removed. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
From: Joe Perches <joe@perches.com> Date: Tue, 19 May 2015 07:32:15 -0700 > On Tue, 2015-05-19 at 12:51 +0200, Nicholas Mc Guire wrote: >> Converting milliseconds to jiffies by "val * HZ / 1000" is technically >> is not a clean solution as it does not handle all corner cases correctly. >> By changing the conversion to use msecs_to_jiffies(val) conversion is >> correct in all cases. >> >> in the current code: >> mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); >> for HZ < 100 (e.g. CONFIG_HZ == 64|32 in alpha) this effectively results >> in no delay at all. > [] >> diff --git a/drivers/net/irda/irda-usb.c b/drivers/net/irda/irda-usb.c > [] >> @@ -848,7 +848,9 @@ static void irda_usb_receive(struct urb *urb) >> * Jean II */ >> self->rx_defer_timer.function = irda_usb_rx_defer_expired; >> self->rx_defer_timer.data = (unsigned long) urb; >> - mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); >> + mod_timer(&self->rx_defer_timer, >> + jiffies + (msecs_to_jiffies(10))); > > The unnecessary () around msecs_to_jiffies could be removed. Agreed. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, 2015-05-19 at 16:45 -0400, David Miller wrote: > From: Joe Perches <joe@perches.com> > Date: Tue, 19 May 2015 07:32:15 -0700 > > > On Tue, 2015-05-19 at 12:51 +0200, Nicholas Mc Guire wrote: > >> Converting milliseconds to jiffies by "val * HZ / 1000" is technically > >> is not a clean solution as it does not handle all corner cases correctly. > >> By changing the conversion to use msecs_to_jiffies(val) conversion is > >> correct in all cases. > >> > >> in the current code: > >> mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); > >> for HZ < 100 (e.g. CONFIG_HZ == 64|32 in alpha) this effectively results > >> in no delay at all. > > [] > >> diff --git a/drivers/net/irda/irda-usb.c b/drivers/net/irda/irda-usb.c > > [] > >> @@ -848,7 +848,9 @@ static void irda_usb_receive(struct urb *urb) > >> * Jean II */ > >> self->rx_defer_timer.function = irda_usb_rx_defer_expired; > >> self->rx_defer_timer.data = (unsigned long) urb; > >> - mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); > >> + mod_timer(&self->rx_defer_timer, > >> + jiffies + (msecs_to_jiffies(10))); > > > > The unnecessary () around msecs_to_jiffies could be removed. The other thing that could be done is to use max(1ul, msecs_to_jiffies()) so that there's always some delay even if HZ <= 50 -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, May 19, 2015 at 02:09:07PM -0700, Joe Perches wrote: > On Tue, 2015-05-19 at 16:45 -0400, David Miller wrote: > > From: Joe Perches <joe@perches.com> > > Date: Tue, 19 May 2015 07:32:15 -0700 > > > > > On Tue, 2015-05-19 at 12:51 +0200, Nicholas Mc Guire wrote: > > >> Converting milliseconds to jiffies by "val * HZ / 1000" is technically > > >> is not a clean solution as it does not handle all corner cases correctly. > > >> By changing the conversion to use msecs_to_jiffies(val) conversion is > > >> correct in all cases. > > >> > > >> in the current code: > > >> mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); > > >> for HZ < 100 (e.g. CONFIG_HZ == 64|32 in alpha) this effectively results > > >> in no delay at all. > > > [] > > >> diff --git a/drivers/net/irda/irda-usb.c b/drivers/net/irda/irda-usb.c > > > [] > > >> @@ -848,7 +848,9 @@ static void irda_usb_receive(struct urb *urb) > > >> * Jean II */ > > >> self->rx_defer_timer.function = irda_usb_rx_defer_expired; > > >> self->rx_defer_timer.data = (unsigned long) urb; > > >> - mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); > > >> + mod_timer(&self->rx_defer_timer, > > >> + jiffies + (msecs_to_jiffies(10))); > > > > > > The unnecessary () around msecs_to_jiffies could be removed. > > The other thing that could be done is to use > max(1ul, msecs_to_jiffies()) > so that there's always some delay even if HZ <= 50 > I may be mistaken, but I am quite sure that msecs_to_jiffies() always returns at least 1 in such situations. Otherwise there would be lots of converstions to 0 in the kernel. Guenter -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, 2015-05-19 at 18:44 -0700, Guenter Roeck wrote: > On Tue, May 19, 2015 at 02:09:07PM -0700, Joe Perches wrote: > > The other thing that could be done is to use > > max(1ul, msecs_to_jiffies()) > > so that there's always some delay even if HZ <= 50 > > > I may be mistaken, but I am quite sure that msecs_to_jiffies() always returns > at least 1 in such situations. Otherwise there would be lots of conversions > to 0 in the kernel. Thanks Guenter Nope, you're not mistaken. Non 0 constants always return a value > 0 The existing uses like max(1ul, msecs_to_jiffies(<variable>)) just guard against <variable> being 0. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/net/irda/irda-usb.c b/drivers/net/irda/irda-usb.c index f6c9163..f3e4563 100644 --- a/drivers/net/irda/irda-usb.c +++ b/drivers/net/irda/irda-usb.c @@ -848,7 +848,9 @@ static void irda_usb_receive(struct urb *urb) * Jean II */ self->rx_defer_timer.function = irda_usb_rx_defer_expired; self->rx_defer_timer.data = (unsigned long) urb; - mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); + mod_timer(&self->rx_defer_timer, + jiffies + (msecs_to_jiffies(10))); + return; }
API compliance scanning with coccinelle flagged: Converting milliseconds to jiffies by "val * HZ / 1000" is technically is not a clean solution as it does not handle all corner cases correctly. By changing the conversion to use msecs_to_jiffies(val) conversion is correct in all cases. in the current code: mod_timer(&self->rx_defer_timer, jiffies + (10 * HZ / 1000)); for HZ < 100 (e.g. CONFIG_HZ == 64|32 in alpha) this effectively results in no delay at all. Patch was compile tested for x86_64_defconfig (implies CONFIG_USB_IRDA=m) Patch is against 4.1-rc4 (localversion-next is -next-20150519) Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org> --- drivers/net/irda/irda-usb.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)