2011-04-12 00:15:24

by Kim Phillips

Subject: [PATCH 1/4] crypto: caam - handle interrupt lines shared across rings

- add IRQF_SHARED to request_irq flags to support parts such as
the p1023 that has one IRQ line per couple of rings.

- resetting a job ring triggers an interrupt, so move request_irq
prior to jr_reset to avoid 'got IRQ but nobody cared' messages.

- disable IRQs in h/w to avoid contention between reset and
interrupt status

- delete invalid comment - if there were incomplete jobs,
module would be in use, preventing an unload.

Signed-off-by: Kim Phillips <[email protected]>
---
this, and the remaining patches in this series, tested on p1023, p3041,
p4080, and 32- and 64-bit p5020.

drivers/crypto/caam/jr.c | 46 ++++++++++++++++++++--------------------------
1 files changed, 20 insertions(+), 26 deletions(-)

diff --git a/drivers/crypto/caam/jr.c b/drivers/crypto/caam/jr.c
index 68cb9af..340fa32 100644
--- a/drivers/crypto/caam/jr.c
+++ b/drivers/crypto/caam/jr.c
@@ -292,10 +292,10 @@ static int caam_reset_hw_jr(struct device *dev)
unsigned int timeout = 100000;

/*
- * FIXME: disabling IRQs here inhibits proper job completion
- * and error propagation
+ * mask interrupts since we are going to poll
+ * for reset completion status
*/
- disable_irq(jrp->irq);
+ setbits32(&jrp->rregs->rconfig_lo, JRCFG_IMSK);

/* initiate flush (required prior to reset) */
wr_reg32(&jrp->rregs->jrcommand, JRCR_RESET);
@@ -320,7 +320,8 @@ static int caam_reset_hw_jr(struct device *dev)
return -EIO;
}

- enable_irq(jrp->irq);
+ /* unmask interrupts */
+ clrbits32(&jrp->rregs->rconfig_lo, JRCFG_IMSK);

return 0;
}
@@ -336,6 +337,21 @@ static int caam_jr_init(struct device *dev)

jrp = dev_get_drvdata(dev);

+ /* Connect job ring interrupt handler. */
+ for_each_possible_cpu(i)
+ tasklet_init(&jrp->irqtask[i], caam_jr_dequeue,
+ (unsigned long)dev);
+
+ error = request_irq(jrp->irq, caam_jr_interrupt, IRQF_SHARED,
+ "caam-jobr", dev);
+ if (error) {
+ dev_err(dev, "can't connect JobR %d interrupt (%d)\n",
+ jrp->ridx, jrp->irq);
+ irq_dispose_mapping(jrp->irq);
+ jrp->irq = 0;
+ return -EINVAL;
+ }
+
error = caam_reset_hw_jr(dev);
if (error)
return error;
@@ -404,28 +420,6 @@ static int caam_jr_init(struct device *dev)
(JOBR_INTC_COUNT_THLD << JRCFG_ICDCT_SHIFT) |
(JOBR_INTC_TIME_THLD << JRCFG_ICTT_SHIFT));

- /* Connect job ring interrupt handler. */
- for_each_possible_cpu(i)
- tasklet_init(&jrp->irqtask[i], caam_jr_dequeue,
- (unsigned long)dev);
-
- error = request_irq(jrp->irq, caam_jr_interrupt, 0,
- "caam-jobr", dev);
- if (error) {
- dev_err(dev, "can't connect JobR %d interrupt (%d)\n",
- jrp->ridx, jrp->irq);
- irq_dispose_mapping(jrp->irq);
- jrp->irq = 0;
- dma_unmap_single(dev, inpbusaddr, sizeof(u32 *) * JOBR_DEPTH,
- DMA_BIDIRECTIONAL);
- dma_unmap_single(dev, outbusaddr, sizeof(u32 *) * JOBR_DEPTH,
- DMA_BIDIRECTIONAL);
- kfree(jrp->inpring);
- kfree(jrp->outring);
- kfree(jrp->entinfo);
- return -EINVAL;
- }
-
jrp->assign = JOBR_UNASSIGNED;
return 0;
}
--
1.7.4.3.dirty

2011-04-15 09:50:56

by Herbert Xu

Subject: Re: [PATCH 1/4] crypto: caam - handle interrupt lines shared across rings

On Mon, Apr 11, 2011 at 07:15:16PM -0500, Kim Phillips wrote:
> - add IRQF_SHARED to request_irq flags to support parts such as
> the p1023 that has one IRQ line per couple of rings.
>
> - resetting a job ring triggers an interrupt, so move request_irq
> prior to jr_reset to avoid 'got IRQ but nobody cared' messages.
>
> - disable IRQs in h/w to avoid contention between reset and
> interrupt status
>
> - delete invalid comment - if there were incomplete jobs,
> module would be in use, preventing an unload.
>
> Signed-off-by: Kim Phillips <[email protected]>
> ---
> this, and the remaining patches in this series, tested on p1023, p3041,
> p4080, and 32- and 64-bit p5020.

All four patches applied. Thanks Kim!
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

2011-05-02 23:27:33

by Kim Phillips

Subject: Re: [PATCH 1/4] crypto: caam - handle interrupt lines shared across rings

On Fri, 15 Apr 2011 17:50:49 +0800
Herbert Xu <[email protected]> wrote:

> On Mon, Apr 11, 2011 at 07:15:16PM -0500, Kim Phillips wrote:
> > - add IRQF_SHARED to request_irq flags to support parts such as
> > the p1023 that has one IRQ line per couple of rings.
> >
> > - resetting a job ring triggers an interrupt, so move request_irq
> > prior to jr_reset to avoid 'got IRQ but nobody cared' messages.
> >
> > - disable IRQs in h/w to avoid contention between reset and
> > interrupt status
> >
> > - delete invalid comment - if there were incomplete jobs,
> > module would be in use, preventing an unload.
> >
> > Signed-off-by: Kim Phillips <[email protected]>
> > ---
> > this, and the remaining patches in this series, tested on p1023, p3041,
> > p4080, and 32- and 64-bit p5020.
>
> All four patches applied. Thanks Kim!

I don't see them in cryptodev-2.6 - did you forget to push them?

Thanks,

Kim

2011-05-02 23:56:17

by Herbert Xu

Subject: Re: [PATCH 1/4] crypto: caam - handle interrupt lines shared across rings

On Mon, May 02, 2011 at 06:27:27PM -0500, Kim Phillips wrote:
>
> I don't see them in cryptodev-2.6 - did you forget to push them?

Indeed, they were stuck on my other machine. They should be on
master now and should show up on the public mirror soon.

Thanks,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt