This change is required for post SEC-5.0 devices which have RNG4. Setting RDB
in security configuration register allows CAAM to use the "Random Data Buffer"
to be filled by a single request. The Random Data Buffer is large enough for
ten packets to get their IVs from a single request. If the Random Data Buffer
is not enabled, then each IV causes a separate request, and RNG4 hardware
cannot keep up resulting in lower IPSEC throughput.
Signed-off-by: Vakul Garg <[email protected]>
---
drivers/crypto/caam/ctrl.c | 3 +++
drivers/crypto/caam/regs.h | 4 +++-
2 files changed, 6 insertions(+), 1 deletions(-)
diff --git a/drivers/crypto/caam/ctrl.c b/drivers/crypto/caam/ctrl.c
index b3fecfa..53dd54e 100644
--- a/drivers/crypto/caam/ctrl.c
+++ b/drivers/crypto/caam/ctrl.c
@@ -234,6 +234,9 @@ static int caam_probe(struct platform_device *pdev)
setbits32(&topregs->ctrl.mcr, MCFGR_WDENABLE |
(sizeof(dma_addr_t) == sizeof(u64) ? MCFGR_LONG_PTR : 0));
+ if (of_device_is_compatible(nprop, "fsl,sec-v5.0"))
+ setbits32(&topregs->ctrl.scfgr, SCFGR_RDBENABLE);
+
if (sizeof(dma_addr_t) == sizeof(u64))
if (of_device_is_compatible(nprop, "fsl,sec-v5.0"))
dma_set_mask(dev, DMA_BIT_MASK(40));
diff --git a/drivers/crypto/caam/regs.h b/drivers/crypto/caam/regs.h
index 3223fc6..cd6feda 100644
--- a/drivers/crypto/caam/regs.h
+++ b/drivers/crypto/caam/regs.h
@@ -252,7 +252,8 @@ struct caam_ctrl {
/* Read/Writable */
u32 rsvd1;
u32 mcr; /* MCFG Master Config Register */
- u32 rsvd2[2];
+ u32 rsvd2;
+ u32 scfgr; /* SCFGR, Security Config Register */
/* Bus Access Configuration Section 010-11f */
/* Read/Writable */
@@ -299,6 +300,7 @@ struct caam_ctrl {
#define MCFGR_WDFAIL 0x20000000 /* DECO watchdog force-fail */
#define MCFGR_DMA_RESET 0x10000000
#define MCFGR_LONG_PTR 0x00010000 /* Use >32-bit desc addressing */
+#define SCFGR_RDBENABLE 0x00000400
/* AXI read cache control */
#define MCFGR_ARCACHE_SHIFT 12
--
1.7.7.6
On Wed, 6 Feb 2013 17:17:46 +0530
Vakul Garg <[email protected]> wrote:
> This change is required for post SEC-5.0 devices which have RNG4. Setting RDB
wrap your commit message text to 75 chars
> in security configuration register allows CAAM to use the "Random Data Buffer"
> to be filled by a single request. The Random Data Buffer is large enough for
> ten packets to get their IVs from a single request. If the Random Data Buffer
> is not enabled, then each IV causes a separate request, and RNG4 hardware
> cannot keep up resulting in lower IPSEC throughput.
Linux kernel IPSec or another IPSEC stack? how much lower?
> + if (of_device_is_compatible(nprop, "fsl,sec-v5.0"))
> + setbits32(&topregs->ctrl.scfgr, SCFGR_RDBENABLE);
> +
this belongs further down - at the end of the RNG4 initialization
section.
Kim
> -----Original Message-----
> From: Phillips Kim-R1AAHA
> Sent: Thursday, February 07, 2013 7:03 AM
> To: Garg Vakul-B16394
> Cc: [email protected]; Herbert Xu; David S. Miller; Liu
> Shengzhou-B36685; Porosanu Alexandru-B06830; [email protected]
> Subject: Re: [PATCH] crypto: caam - set RDB bit in security configuration
> register
>
> On Wed, 6 Feb 2013 17:17:46 +0530
> Vakul Garg <[email protected]> wrote:
>
> > This change is required for post SEC-5.0 devices which have RNG4.
> > Setting RDB
>
> wrap your commit message text to 75 chars
Ok
>
> > in security configuration register allows CAAM to use the "Random Data
> Buffer"
> > to be filled by a single request. The Random Data Buffer is large
> > enough for ten packets to get their IVs from a single request. If the
> > Random Data Buffer is not enabled, then each IV causes a separate
> > request, and RNG4 hardware cannot keep up resulting in lower IPSEC
> throughput.
>
> Linux kernel IPSec or another IPSEC stack? how much lower?
We measured USDPAA IPSEC throughput to be 1.9 times better on T4240 with this change.
>
> > + if (of_device_is_compatible(nprop, "fsl,sec-v5.0"))
> > + setbits32(&topregs->ctrl.scfgr, SCFGR_RDBENABLE);
> > +
>
> this belongs further down - at the end of the RNG4 initialization
> section.
Ok
>
> Kim