| Message ID | 1425388897-5434-4-git-send-email-mort@bork.org (mailing list archive) |
|---|---|
| State | Rejected |
| Delegated to: | Herbert Xu |
| Headers | show |
On Tue, 3 Mar 2015 08:21:35 -0500 Martin Hicks <mort@bork.org> wrote: > The submission count was off by one. > > Signed-off-by: Martin Hicks <mort@bork.org> > --- sadly, this directly contradicts: commit 4b24ea971a93f5d0bec34bf7bfd0939f70cfaae6 Author: Vishnu Suresh <Vishnu@freescale.com> Date: Mon Oct 20 21:06:18 2008 +0800 crypto: talitos - Preempt overflow interrupts off-by-one fix My guess is your request submission pattern differs from that of Vishnu's (probably IPSec and/or tcrypt), or later h/w versions have gotten better about dealing with channel near-overflow conditions. Either way, I'd prefer we not do this: it might break others, and I'm guessing doesn't improve performance _that_ much? If it does, we could risk it and restrict it to SEC versions 3.3 and above maybe? Not sure what to do here exactly, barring digging up and old 2.x SEC and testing. Kim p.s. I checked, Vishnu isn't with Freescale anymore, so I can't cc him. -- To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Ok, I'm fine dropping this patch. I'm sure it doesn't affect performance in a measurable way. mh On Tue, Mar 3, 2015 at 7:35 PM, Kim Phillips <kim.phillips@freescale.com> wrote: > On Tue, 3 Mar 2015 08:21:35 -0500 > Martin Hicks <mort@bork.org> wrote: > >> The submission count was off by one. >> >> Signed-off-by: Martin Hicks <mort@bork.org> >> --- > sadly, this directly contradicts: > > commit 4b24ea971a93f5d0bec34bf7bfd0939f70cfaae6 > Author: Vishnu Suresh <Vishnu@freescale.com> > Date: Mon Oct 20 21:06:18 2008 +0800 > > crypto: talitos - Preempt overflow interrupts off-by-one fix > > My guess is your request submission pattern differs from that of > Vishnu's (probably IPSec and/or tcrypt), or later h/w versions have > gotten better about dealing with channel near-overflow conditions. > Either way, I'd prefer we not do this: it might break others, and > I'm guessing doesn't improve performance _that_ much? > > If it does, we could risk it and restrict it to SEC versions 3.3 and > above maybe? Not sure what to do here exactly, barring digging up > and old 2.x SEC and testing. > > Kim > > p.s. I checked, Vishnu isn't with Freescale anymore, so I can't > cc him.
diff --git a/drivers/crypto/talitos.c b/drivers/crypto/talitos.c index 89cf4d5..7709805 100644 --- a/drivers/crypto/talitos.c +++ b/drivers/crypto/talitos.c @@ -2722,8 +2722,7 @@ static int talitos_probe(struct platform_device *ofdev) goto err_out; } - atomic_set(&priv->chan[i].submit_count, - -(priv->chfifo_len - 1)); + atomic_set(&priv->chan[i].submit_count, -priv->chfifo_len); } dma_set_mask(dev, DMA_BIT_MASK(36));
The submission count was off by one. Signed-off-by: Martin Hicks <mort@bork.org> --- drivers/crypto/talitos.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)