This patch implements a UIO interface for the SMX Cryptengine.
The Cryptengine found on the Nias Digital SMX board is best suited
for a UIO interface. It is not wired in to the cryptographic API
as the engine handles it's own keys, algorithms, everything. All
that we know about is that if there's room in the buffer, you can
write data to it and when there's data ready, you read it out again.
There isn't necessarily even any direct correlation between data
going in and data coming out again, the engine may consume or
generate data all on its own.
This driver is for proprietary hardware but we're always told to
submit the drivers anyway; here you are. :-)
Signed-Off-By: Ben Nizette <[email protected]>
---
diff --git a/MAINTAINERS b/MAINTAINERS
index 2cdb591..480fd44 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3554,6 +3554,11 @@ M: [email protected]
L: [email protected]
S: Maintained
+SMX UIO Interface
+P: Ben Nizette
+M: [email protected]
+S: Maintained
+
SOFTMAC LAYER (IEEE 802.11)
P: Daniel Drake
M: [email protected]
diff --git a/drivers/uio/Kconfig b/drivers/uio/Kconfig
index b778ed7..e1ab43b 100644
--- a/drivers/uio/Kconfig
+++ b/drivers/uio/Kconfig
@@ -26,4 +26,12 @@ config UIO_CIF
To compile this driver as a module, choose M here: the module
will be called uio_cif.
+config UIO_SMX
+ tristate "SMX cryptengine UIO interface"
+ depends on UIO
+ default n
+ help
+ Userspace IO interface to the Cryptography engine found on the
+ Nias Digital SMX boards. If you compile this as a module, it
+ will be called uio_smx.
endmenu
diff --git a/drivers/uio/Makefile b/drivers/uio/Makefile
index 7fecfb4..18c4566 100644
--- a/drivers/uio/Makefile
+++ b/drivers/uio/Makefile
@@ -1,2 +1,3 @@
obj-$(CONFIG_UIO) += uio.o
obj-$(CONFIG_UIO_CIF) += uio_cif.o
+obj-$(CONFIG_UIO_SMX) += uio_smx.o
diff --git a/drivers/uio/uio_smx.c b/drivers/uio/uio_smx.c
new file mode 100644
index 0000000..e56b37d
--- /dev/null
+++ b/drivers/uio/uio_smx.c
@@ -0,0 +1,114 @@
+/*
+ * UIO SMX Cryptengine driver.
+ *
+ * (C) 2008 Nias Digital P/L <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/device.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/uio_driver.h>
+
+#include <asm/io.h>
+
+#define SMX_CSR 0x00000000
+#define SMX_EnD 0x00000001
+#define SMX_RUN 0x00000002
+#define SMX_DRDY 0x00000004
+#define SMX_ERR 0x00000008
+
+static irqreturn_t smx_handler(int irq, struct uio_info *dev_info)
+{
+ void __iomem *csr = dev_info->mem[0].internal_addr + SMX_CSR;
+
+ u32 status = ioread32(csr);
+
+ /* Disable interrupt */
+ iowrite8(status & ~SMX_DRDY, csr);
+ return IRQ_HANDLED;
+}
+
+static int __devinit smx_ce_probe(struct platform_device *dev)
+{
+ struct uio_info *info;
+ struct resource *regs;
+
+ info = kzalloc(sizeof(struct uio_info), GFP_KERNEL);
+ if (!info)
+ return -ENOMEM;
+
+ regs = platform_get_resource(dev, IORESOURCE_MEM, 0);
+ if (!regs)
+ goto out_free;
+
+ info->mem[0].addr = regs->start;
+ if (!info->mem[0].addr)
+ goto out_free;
+ info->mem[0].internal_addr = ioremap(regs->start, regs->end -
regs->start);
+ if (!info->mem[0].internal_addr)
+ goto out_free;
+
+ info->mem[0].size = regs->end - regs->start;
+ info->mem[0].memtype = UIO_MEM_PHYS;
+
+ info->name = "smx-ce";
+ info->version = "0.03";
+ info->irq = platform_get_irq(dev, 0);
+ info->irq_flags = 0;
+ info->handler = smx_handler;
+
+ platform_set_drvdata(dev, info);
+
+ if (uio_register_device(&dev->dev, info))
+ goto out_unmap;
+
+ return 0;
+out_unmap:
+ iounmap(info->mem[0].internal_addr);
+out_free:
+ kfree(info);
+
+ return -ENODEV;
+}
+
+static int __devexit smx_ce_remove(struct platform_device *dev)
+{
+ struct uio_info *info = platform_get_drvdata(dev);
+
+ uio_unregister_device(info);
+ platform_set_drvdata(dev, NULL);
+ iounmap(info->mem[0].internal_addr);
+
+ kfree (info);
+
+ return 0;
+}
+
+static struct platform_driver smx_ce_driver = {
+ .probe = smx_ce_probe,
+ .remove = __devexit_p(smx_ce_remove),
+ .driver = {
+ .name = "smx-ce",
+ .owner = THIS_MODULE,
+ },
+};
+
+static int __init smx_ce_init_module(void)
+{
+ return platform_driver_register(&smx_ce_driver);
+}
+module_init(smx_ce_init_module);
+
+static void __exit smx_ce_exit_module(void)
+{
+ platform_driver_unregister(&smx_ce_driver);
+}
+module_exit(smx_ce_exit_module);
+
+MODULE_LICENSE("GPL V2");
+MODULE_AUTHOR("Ben Nizette <[email protected]>");
Am Tue, 11 Mar 2008 15:57:10 +1100
schrieb Ben Nizette <[email protected]>:
Hi Ben,
> This patch implements a UIO interface for the SMX Cryptengine.
>
> The Cryptengine found on the Nias Digital SMX board is best suited
> for a UIO interface. It is not wired in to the cryptographic API
> as the engine handles it's own keys, algorithms, everything. All
> that we know about is that if there's room in the buffer, you can
> write data to it and when there's data ready, you read it out again.
>
> There isn't necessarily even any direct correlation between data
> going in and data coming out again, the engine may consume or
> generate data all on its own.
>
> This driver is for proprietary hardware but we're always told to
> submit the drivers anyway; here you are. :-)
I've got no problems with that, but I'd like to know if that board can
be bought somewhere (the niasdigital.com homepage doesn't mention it)
and if and under which license the userspace part of the driver is
available. The homepage says you're in favor of the GPL, so if it can
be checked out or downloaded somewhere, please mention it, maybe in
the Kconfig help text. If the userspace part is proprietary, please
mention that, too.
There are some minor problems with your patch, see below.
Note: testing your patches with checkpatch.pl would help...
>
> Signed-Off-By: Ben Nizette <[email protected]>
that should be signed-off-by (lower-case)
> ---
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 2cdb591..480fd44 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -3554,6 +3554,11 @@ M: [email protected]
> L: [email protected]
> S: Maintained
>
> +SMX UIO Interface
> +P: Ben Nizette
> +M: [email protected]
> +S: Maintained
> +
> SOFTMAC LAYER (IEEE 802.11)
> P: Daniel Drake
> M: [email protected]
> diff --git a/drivers/uio/Kconfig b/drivers/uio/Kconfig
> index b778ed7..e1ab43b 100644
> --- a/drivers/uio/Kconfig
> +++ b/drivers/uio/Kconfig
> @@ -26,4 +26,12 @@ config UIO_CIF
> To compile this driver as a module, choose M here: the
> module will be called uio_cif.
>
> +config UIO_SMX
> + tristate "SMX cryptengine UIO interface"
> + depends on UIO
> + default n
> + help
> + Userspace IO interface to the Cryptography engine found on
> the
> + Nias Digital SMX boards. If you compile this as a module,
> it
> + will be called uio_smx.
Please give a hint that a userspace part is needed, and how it can be
obtained.
> endmenu
> diff --git a/drivers/uio/Makefile b/drivers/uio/Makefile
> index 7fecfb4..18c4566 100644
> --- a/drivers/uio/Makefile
> +++ b/drivers/uio/Makefile
> @@ -1,2 +1,3 @@
> obj-$(CONFIG_UIO) += uio.o
> obj-$(CONFIG_UIO_CIF) += uio_cif.o
> +obj-$(CONFIG_UIO_SMX) += uio_smx.o
> diff --git a/drivers/uio/uio_smx.c b/drivers/uio/uio_smx.c
> new file mode 100644
> index 0000000..e56b37d
> --- /dev/null
> +++ b/drivers/uio/uio_smx.c
> @@ -0,0 +1,114 @@
> +/*
> + * UIO SMX Cryptengine driver.
> + *
> + * (C) 2008 Nias Digital P/L <[email protected]>
> + *
> + * This program is free software; you can redistribute it and/or
> modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + *
> + */
> +
> +#include <linux/device.h>
> +#include <linux/module.h>
> +#include <linux/platform_device.h>
> +#include <linux/uio_driver.h>
> +
> +#include <asm/io.h>
> +
> +#define SMX_CSR 0x00000000
> +#define SMX_EnD 0x00000001
> +#define SMX_RUN 0x00000002
> +#define SMX_DRDY 0x00000004
> +#define SMX_ERR 0x00000008
> +
> +static irqreturn_t smx_handler(int irq, struct uio_info *dev_info)
> +{
> + void __iomem *csr = dev_info->mem[0].internal_addr + SMX_CSR;
> +
> + u32 status = ioread32(csr);
> +
> + /* Disable interrupt */
> + iowrite8(status & ~SMX_DRDY, csr);
> + return IRQ_HANDLED;
> +}
You don't support shared interrupts. Is that intentional?
> +
> +static int __devinit smx_ce_probe(struct platform_device *dev)
> +{
> + struct uio_info *info;
> + struct resource *regs;
> +
> + info = kzalloc(sizeof(struct uio_info), GFP_KERNEL);
> + if (!info)
> + return -ENOMEM;
> +
> + regs = platform_get_resource(dev, IORESOURCE_MEM, 0);
> + if (!regs)
> + goto out_free;
> +
> + info->mem[0].addr = regs->start;
> + if (!info->mem[0].addr)
> + goto out_free;
> + info->mem[0].internal_addr = ioremap(regs->start, regs->end -
> regs->start);
Word wrap - please fix your mail client
> + if (!info->mem[0].internal_addr)
> + goto out_free;
> +
> + info->mem[0].size = regs->end - regs->start;
> + info->mem[0].memtype = UIO_MEM_PHYS;
> +
> + info->name = "smx-ce";
> + info->version = "0.03";
> + info->irq = platform_get_irq(dev, 0);
> + info->irq_flags = 0;
> + info->handler = smx_handler;
> +
> + platform_set_drvdata(dev, info);
> +
> + if (uio_register_device(&dev->dev, info))
> + goto out_unmap;
> +
> + return 0;
> +out_unmap:
> + iounmap(info->mem[0].internal_addr);
> +out_free:
> + kfree(info);
> +
> + return -ENODEV;
> +}
> +
> +static int __devexit smx_ce_remove(struct platform_device *dev)
> +{
> + struct uio_info *info = platform_get_drvdata(dev);
> +
> + uio_unregister_device(info);
> + platform_set_drvdata(dev, NULL);
> + iounmap(info->mem[0].internal_addr);
> +
> + kfree (info);
No space before parenthesis
> +
> + return 0;
> +}
> +
> +static struct platform_driver smx_ce_driver = {
> + .probe = smx_ce_probe,
> + .remove = __devexit_p(smx_ce_remove),
> + .driver = {
> + .name = "smx-ce",
> + .owner = THIS_MODULE,
> + },
> +};
> +
> +static int __init smx_ce_init_module(void)
> +{
> + return platform_driver_register(&smx_ce_driver);
> +}
> +module_init(smx_ce_init_module);
> +
> +static void __exit smx_ce_exit_module(void)
> +{
> + platform_driver_unregister(&smx_ce_driver);
> +}
> +module_exit(smx_ce_exit_module);
> +
> +MODULE_LICENSE("GPL V2");
this needs to be "GPL v2" (lowercase 'v'), otherwise it won't build the
module (at least on 2.6.25-rc5).
> +MODULE_AUTHOR("Ben Nizette <[email protected]>");
On Tue, 2008-03-11 at 09:15 +0100, Hans-Jürgen Koch wrote:
> Am Tue, 11 Mar 2008 15:57:10 +1100
> schrieb Ben Nizette <[email protected]>:
> > This driver is for proprietary hardware but we're always told to
> > submit the drivers anyway; here you are. :-)
>
> I've got no problems with that, but I'd like to know if that board can
> be bought somewhere (the niasdigital.com homepage doesn't mention it)
> and if and under which license the userspace part of the driver is
> available. The homepage says you're in favor of the GPL, so if it can
> be checked out or downloaded somewhere, please mention it, maybe in
> the Kconfig help text. If the userspace part is proprietary, please
> mention that, too.
The userspace part will be released under the GPL along with the
hardware itself, Q4 '08. Have changed the Kconfig text to reflect this.
Of course this means for the next few months I will be the only direct
user of this driver, but I'm also submitting it in the hope that I can
convince others that it's a good thing to do. I'm an active member of
AVRFreaks, an AVR32 support forum, and I spend far too much of my life
trying to convince people that polling the peripheral and
mmap'ing /dev/mem is not a good way to go. Fingers crossed that this
driver at least serves to help educate until the SMX hardware is
released and it can serve it's main purpose :-)
<snip>
> > +static irqreturn_t smx_handler(int irq, struct uio_info *dev_info)
> > +{
> > + void __iomem *csr = dev_info->mem[0].internal_addr + SMX_CSR;
> > +
> > + u32 status = ioread32(csr);
> > +
> > + /* Disable interrupt */
> > + iowrite8(status & ~SMX_DRDY, csr);
> > + return IRQ_HANDLED;
> > +}
>
> You don't support shared interrupts. Is that intentional?
>
On the hardware we have there is no possibility that the interrupt is
shared. That said, it's little effort to allow the sharing and makes
the whole thing more future-proof, changed.
<snip>
> > +
> > +MODULE_LICENSE("GPL V2");
>
> this needs to be "GPL v2" (lowercase 'v'), otherwise it won't build the
> module (at least on 2.6.25-rc5).
>
Ah yep, thanks.
Thanks for the review, shall send a rejigged patch presently.
--Ben.
On Tue, Mar 11, 2008 at 03:57:10PM +1100, Ben Nizette wrote:
> + regs = platform_get_resource(dev, IORESOURCE_MEM, 0);
> + if (!regs)
> + goto out_free;
> +
> + info->mem[0].addr = regs->start;
> + if (!info->mem[0].addr)
> + goto out_free;
> + info->mem[0].internal_addr = ioremap(regs->start, regs->end -
> regs->start);
> + if (!info->mem[0].internal_addr)
> + goto out_free;
> +
> + info->mem[0].size = regs->end - regs->start;
> + info->mem[0].memtype = UIO_MEM_PHYS;
> +
You have an off-by-1 in the resource size. ioremap and struct resource
are not equal in their expectations. You probably want to do something
like:
info->mem[0].size = regs->end - regs->start + 1;
info->mem[0].internal_addr = ioremap(regs->start, info->mem[0].size);
...
and make sure that your resource end doesn't overlap with a start address
of an unrelated resource in your definition. /proc/iomem tends to be
useful for this.
On Wed, 2008-03-12 at 19:08 +0900, Paul Mundt wrote:
> On Tue, Mar 11, 2008 at 03:57:10PM +1100, Ben Nizette wrote:
> > +
> > + info->mem[0].size = regs->end - regs->start;
> > + info->mem[0].memtype = UIO_MEM_PHYS;
> > +
> You have an off-by-1 in the resource size. ioremap and struct resource
> are not equal in their expectations. You probably want to do something
> like:
>
> info->mem[0].size = regs->end - regs->start + 1;
> info->mem[0].internal_addr = ioremap(regs->start, info->mem[0].size);
> ...
>
You're right, thanks. I hadn't noticed since we don't actually access
all the way to the end of the region in normal use.
> and make sure that your resource end doesn't overlap with a start address
> of an unrelated resource in your definition. /proc/iomem tends to be
> useful for this.
Yup. In our case this is fairly easy as the engine just takes up a
whole Chip Select's worth of address space on an AVR32 AT32AP7000.
--Ben.