* Re: [PATCH 4/5] libata/sff: Use ops->bmdma_stop instead of ata_bmdma_stop()
From: Jeff Garzik @ 2009-12-03 7:35 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: tj, linux-ide, linuxppc-dev
In-Reply-To: <20091202003629.4342710080C@ozlabs.org>
On 12/01/2009 07:36 PM, Benjamin Herrenschmidt wrote:
> In libata-sff, ata_sff_post_internal_cmd() directly calls ata_bmdma_stop()
> instead of ap->ops->bmdma_stop(). This can be a problem for controllers
> that use their own bmdma_stop for which the generic sff one isn't suitable
>
> Signed-off-by: Benjamin Herrenschmidt<benh@kernel.crashing.org>
> ---
>
> drivers/ata/libata-sff.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> --- linux-work.orig/drivers/ata/libata-sff.c 2009-12-01 17:48:27.000000000 +1100
> +++ linux-work/drivers/ata/libata-sff.c 2009-12-01 17:48:48.000000000 +1100
> @@ -2384,7 +2384,7 @@ void ata_sff_post_internal_cmd(struct at
> ap->hsm_task_state = HSM_ST_IDLE;
>
> if (ap->ioaddr.bmdma_addr)
> - ata_bmdma_stop(qc);
> + ap->ops->bmdma_stop(qc);
>
> spin_unlock_irqrestore(ap->lock, flags);
applied
As we discussed, feel free to carry this in your tree as well. git
should be able to sort it out.
Acked-by: Jeff Garzik <jgarzik@redhat.com>
^ permalink raw reply
* Re: [PATCH 5/5] libata/drivers: Add pata_macio, driver Apple PowerMac/PowerBook IDE controller
From: Jeff Garzik @ 2009-12-03 8:12 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: tj, linux-ide, linuxppc-dev
In-Reply-To: <20091202003630.29919100851@ozlabs.org>
On 12/01/2009 07:36 PM, Benjamin Herrenschmidt wrote:
> This is a libata driver for the "macio" IDE controller used on most Apple
> PowerMac and PowerBooks. It's a libata equivalent of drivers/ide/ppc/pmac.c
>
> It supports all the features of its predecessor, including mediabay hotplug
> and suspend/resume. It should also support module load/unload.
>
> The timing calculations have been simplified to use pre-calculated tables
> compared to drivers/ide/pmac.c and it uses the new mediabay interface
> provided by a previous patch.
>
> Signed-off-by: Benjamin Herrenschmidt<benh@kernel.crashing.org>
> ---
>
> v2. Better tested now, seems to be reasonably solid.
>
> Addressed Tejun comments and made remove more robust vs. media-bay,
> should also fix Andreas problem.
>
>
> drivers/ata/Kconfig | 10
> drivers/ata/Makefile | 1
> drivers/ata/pata_macio.c | 1427 +++++++++++++++++++++++++++++++++++++++++++++++
> 3 files changed, 1438 insertions(+)
Looks fine to me. Two minor comments, which might perhaps be ignored if
that is your taste:
* prefer enums to #define's, for constants
* prefer direct function call to "ap->ops->foo_bar()", because
ap->ops->foo_bar() is guaranteed to be a constant value known to the
driver. The driver is the entity responsible for the function pointer.
Maybe saves a cycle or two. Not terribly important, but hey, calling
ap->ops->sff_exec_command() from pata_macio_bmdma_setup() is a hot path.
Jeff
^ permalink raw reply
* Re: [PATCH] [SCSI] mpt fusion: Fix 32 bit platforms with 64 bit resources.
From: Wolfgang Denk @ 2009-12-03 8:56 UTC (permalink / raw)
To: Pravin Bathija, Benjamin Herrenschmidt, Desai, Kashyap
Cc: linuxppc-dev, Eric.Moore, linux-scsi
In-Reply-To: <1259805106-23636-1-git-send-email-pbathija@amcc.com>
Dear Pravin Bathija,
In message <1259805106-23636-1-git-send-email-pbathija@amcc.com> you wrote:
> Powerpc 44x uses 36 bit real address while the real address defined
> in MPT Fusion driver is of type 32 bit. This causes ioremap to fail and driver
> fails to initialize. This fix changes the data types representing the real
> address from unsigned long 32-bit types to resource_size_t which is 64-bit. The
> driver has been tested, the disks get discovered correctly and can do IO.
...
> --- a/drivers/message/fusion/mptbase.c
> +++ b/drivers/message/fusion/mptbase.c
> @@ -1511,7 +1511,7 @@ mpt_mapresources(MPT_ADAPTER *ioc)
> {
> u8 __iomem *mem;
> int ii;
> - unsigned long mem_phys;
> + resource_size_t mem_phys;
> unsigned long port;
> u32 msize;
> u32 psize;
I'm not sure if this one-liner really covers all the related issues.
We submitted a similar (but apparently more complete) patch more than
a year ago. Dunno why it has never been picked up. See
http://thread.gmane.org/gmane.linux.scsi/46082 for reference.
> From: Yuri Tikhonov <yur@emcraft.com>
To: linux-scsi@vger.kernel.org
Subject: [PATCH] mptbase: use resource_size_t instead of unsigned long and u32
Date: Thu, 13 Nov 2008 11:33:16 +0300
Hello,
The following patch adds using resource_size_t for the
pci_resource_start()/pci_resource_len() return values. This
makes mptbase driver work correctly on 32 bit systems with
64 bit resources (e.g. PPC440SPe).
Do some minor cleanups in mpt_mapresources() as well.
Signed-off-by: Yuri Tikhonov <yur@emcraft.com>
Signed-off-by: Ilya Yanok <yanok@emcraft.com>
---
drivers/message/fusion/mptbase.c | 38 ++++++++++++++++++++++++++------------
drivers/message/fusion/mptbase.h | 5 +++--
2 files changed, 29 insertions(+), 14 deletions(-)
diff --git a/drivers/message/fusion/mptbase.c b/drivers/message/fusion/mptbase.c
index d6a0074..9daf844 100644
--- a/drivers/message/fusion/mptbase.c
+++ b/drivers/message/fusion/mptbase.c
@@ -1488,11 +1488,12 @@ static int
mpt_mapresources(MPT_ADAPTER *ioc)
{
u8 __iomem *mem;
+ u8 __iomem *port;
int ii;
- unsigned long mem_phys;
- unsigned long port;
- u32 msize;
- u32 psize;
+ resource_size_t mem_phys;
+ resource_size_t port_phys;
+ resource_size_t msize;
+ resource_size_t psize;
u8 revision;
int r = -ENODEV;
struct pci_dev *pdev;
@@ -1530,13 +1531,13 @@ mpt_mapresources(MPT_ADAPTER *ioc)
}
mem_phys = msize = 0;
- port = psize = 0;
+ port_phys = psize = 0;
for (ii = 0; ii < DEVICE_COUNT_RESOURCE; ii++) {
if (pci_resource_flags(pdev, ii) & PCI_BASE_ADDRESS_SPACE_IO) {
if (psize)
continue;
/* Get I/O space! */
- port = pci_resource_start(pdev, ii);
+ port_phys = pci_resource_start(pdev, ii);
psize = pci_resource_len(pdev, ii);
} else {
if (msize)
@@ -1546,11 +1547,8 @@ mpt_mapresources(MPT_ADAPTER *ioc)
msize = pci_resource_len(pdev, ii);
}
}
- ioc->mem_size = msize;
- mem = NULL;
/* Get logical ptr for PciMem0 space */
- /*mem = ioremap(mem_phys, msize);*/
mem = ioremap(mem_phys, msize);
if (mem == NULL) {
printk(MYIOC_s_ERR_FMT ": ERROR - Unable to map adapter"
@@ -1558,14 +1556,24 @@ mpt_mapresources(MPT_ADAPTER *ioc)
return -EINVAL;
}
ioc->memmap = mem;
- dinitprintk(ioc, printk(MYIOC_s_INFO_FMT "mem = %p, mem_phys = %lx\n",
- ioc->name, mem, mem_phys));
+ dinitprintk(ioc, printk(MYIOC_s_INFO_FMT "mem=%p, mem_phys=%llx\n",
+ ioc->name, mem, (u64)mem_phys));
ioc->mem_phys = mem_phys;
ioc->chip = (SYSIF_REGS __iomem *)mem;
/* Save Port IO values in case we need to do downloadboot */
- ioc->pio_mem_phys = port;
+ port = ioremap(port_phys, psize);
+ if (port == NULL) {
+ printk(MYIOC_s_ERR_FMT ": ERROR - Unable to map adapter"
+ " port!\n", ioc->name);
+ return -EINVAL;
+ }
+ ioc->portmap = port;
+ dinitprintk(ioc, printk(MYIOC_s_INFO_FMT "port=%p, port_phys=%llx\n",
+ ioc->name, port, (u64)port_phys));
+
+ ioc->pio_mem_phys = port_phys;
ioc->pio_chip = (SYSIF_REGS __iomem *)port;
return 0;
@@ -1790,6 +1798,7 @@ mpt_attach(struct pci_dev *pdev, const struct pci_device_id *id)
list_del(&ioc->list);
if (ioc->alt_ioc)
ioc->alt_ioc->alt_ioc = NULL;
+ iounmap(ioc->portmap);
iounmap(ioc->memmap);
if (r != -5)
pci_release_selected_regions(pdev, ioc->bars);
@@ -2547,6 +2556,11 @@ mpt_adapter_dispose(MPT_ADAPTER *ioc)
ioc->pci_irq = -1;
}
+ if (ioc->portmap != NULL) {
+ iounmap(ioc->portmap);
+ ioc->portmap = NULL;
+ }
+
if (ioc->memmap != NULL) {
iounmap(ioc->memmap);
ioc->memmap = NULL;
diff --git a/drivers/message/fusion/mptbase.h b/drivers/message/fusion/mptbase.h
index dff048c..17826b3 100644
--- a/drivers/message/fusion/mptbase.h
+++ b/drivers/message/fusion/mptbase.h
@@ -584,8 +584,8 @@ typedef struct _MPT_ADAPTER
SYSIF_REGS __iomem *chip; /* == c8817000 (mmap) */
SYSIF_REGS __iomem *pio_chip; /* Programmed IO (downloadboot) */
u8 bus_type;
- u32 mem_phys; /* == f4020000 (mmap) */
- u32 pio_mem_phys; /* Programmed IO (downloadboot) */
+ resource_size_t mem_phys; /* == f4020000 (mmap) */
+ resource_size_t pio_mem_phys; /* Programmed IO (downloadboot) */
int mem_size; /* mmap memory size */
int number_of_buses;
int devices_per_bus;
@@ -635,6 +635,7 @@ typedef struct _MPT_ADAPTER
int bars; /* bitmask of BAR's that must be configured */
int msi_enable;
u8 __iomem *memmap; /* mmap address */
+ u8 __iomem *portmap; /* mmap address */
struct Scsi_Host *sh; /* Scsi Host pointer */
SpiCfgData spi_data; /* Scsi config. data */
RaidCfgData raid_data; /* Raid config. data */
--
1.5.6.1
--
Yuri Tikhonov, Senior Software Engineer
Emcraft Systems, www.emcraft.com
Best regards,
Wolfgang Denk
--
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
"I haven't lost my mind - it's backed up on tape somewhere."
^ permalink raw reply related
* Re: Recommended functions for accessing internal registers
From: Fortini Matteo @ 2009-12-03 11:16 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev@lists.ozlabs.org
In-Reply-To: <1259787475.2076.1160.camel@pasglop>
I'm on an embedded system, so every resource counts.
One of the biggest impacts is when writing to a communication/memory
access FIFO or reading/writing configurations.
In these cases, I'd just need to make sure that there's no I/O
reordering and/or subsequent r/w are not optimized away, I believe.
Should I switch to the deprecated "volatile" attribute?
Thank you.
Cheers,
Matteo
Il 02/12/2009 21.57, Benjamin Herrenschmidt ha scritto:
> On Tue, 2009-12-01 at 17:44 +0100, Fortini Matteo wrote:
>
>> I see that throughout the kernel source, internal PPC registers are
>> accessed through [in|out]_be[32|16|8]() functions. However, they are
>> translated into 3 inline assembly instructions, one of which is an
>> isync, which has a huge performance hit.
>> I tried using readl_be() which seems to be the right function according
>> to the Documentation/ dir, but it is translated directly to in_be32(),
>> so no luck.
>>
>> Is it really necessary to use all those instructions? I know I could use
>> a (volatile u32 *) variable to avoid subsequent read/writes to be
>> optimized out, but it seems to be a deprecated use.
>>
> There are good reasons why the accessors contain those barriers. What
> are you doing that would be performance critical enough for those to be
> a problem ?
>
> Cheers,
> Ben.
>
>
>
^ permalink raw reply
* Re: [PATCH] ppc64: re-enable kexec to allow module loads with CONFIG_MODVERSIONS and CONFIG_RELOCATABLE turned on
From: Neil Horman @ 2009-12-03 15:04 UTC (permalink / raw)
To: linuxppc-dev; +Cc: rusty, paulus
In-Reply-To: <20091119195216.GC11414@hmsreliant.think-freely.org>
>>> Paul, Ben, given that Rusty hasn't come back with any opinion on this patch,
>>do you
>>> feel comfortable merging it via the ppc tree? Currently the earlyinit
>routine
>>> is only compiled in and used for your arch, so I think its fairly benign.
>>
>>Sorry, I *did* track down the archives for linuxppc-dev, then find your post,
>>then read your patch. But I didn't actually reply.
>>
>>Other than minor issues, there's one significant one: you shouldn't be trying
>>to change rodata. It might work on PPC today, but it's poor form at least.
>>
>>How's this? Untested on ppc.
>>
>I'll try grab a ppc64 system and test this soon.
>
>>Other changes:
>>1) I also changed reloc_start to an array; this is a good idea for any
>> linker-defined symbols so the compiler can't make assumptions about size.
>Ok, I'll certainly trust your linker skill over mine :)
>
>>2) local.h? How about module.h?
>Seems good to me
>>3) I don't think the extra ". = 0" is necessary.
>Its not, I was just trying to be clear about where reloc_start was to be placed.
>
>>4) ARCH_USES_RELOC_ENTRIES isn't clear enough for me; I prefer
>> ARCH_RELOCATE_KCRCTAB.
>Sounds good to me.
>
>
>I'll test this as soon as I'm able
>Thanks!
>Neil
>
Just finished testing your patch Rusty, it all works quite well, Thanks!
Will this be going in via your tree, or the ppc tree?
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Neil
^ permalink raw reply
* Re: [PATCH v2] i2c-mpc: generate START condition after STOP caused by read i2c_msg
From: Michael Lawnick @ 2009-12-03 15:09 UTC (permalink / raw)
To: Ben Dooks; +Cc: linuxppc-dev, Esben Haabendal, Esben Haabendal, linux-i2c
In-Reply-To: <20090526213351.GG23114@fluff.org.uk>
Ben Dooks said the following:
> On Tue, May 26, 2009 at 01:30:21PM +0200, Esben Haabendal wrote:
>> On Tue, May 19, 2009 at 7:22 AM, Esben Haabendal <eha@doredevelopment.dk> wrote:
>> > This fixes MAL (arbitration lost) bug caused by illegal use of
>> > RSTA (repeated START) after STOP condition generated after last byte
>> > of reads. With this patch, it is possible to do an i2c_transfer() with
>> > additional i2c_msg's following the I2C_M_RD messages.
>> >
>> > It still needs to be resolved if it is possible to fix this issue
>> > by removing the STOP condition after reads in a robust way.
>> >
>> > Signed-off-by: Esben Haabendal <eha@doredevelopment.dk>
>> > ---
>> > ?drivers/i2c/busses/i2c-mpc.c | ? ?9 +++++++--
>> > ?1 files changed, 7 insertions(+), 2 deletions(-)
>>
>> Any blockers to get this accepted?
>
> It would be nice to get an ack from someone who can actually test
> the driver before getting this merged.
>
What is the state of this patch?
Shouldn't we attack the problem on a more general way by inventing a
Flag I2C_M_RESTART (or better I2C_M_NO_RESTART for backward compatibility)?
This way the client driver is able to decide what it needs. If we do the
choice within adapter, chance is about 50% to be wrong.
Just my 2 Cents.
--
Michael
^ permalink raw reply
* Re: [PATCH 5/5] libata/drivers: Add pata_macio, driver Apple PowerMac/PowerBook IDE controller
From: Mikael Pettersson @ 2009-12-03 15:10 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: tj, linux-ide, linuxppc-dev, jeff
In-Reply-To: <20091202003630.29919100851@ozlabs.org>
Benjamin Herrenschmidt writes:
> This is a libata driver for the "macio" IDE controller used on most Apple
> PowerMac and PowerBooks. It's a libata equivalent of drivers/ide/ppc/pmac.c
>
> It supports all the features of its predecessor, including mediabay hotplug
> and suspend/resume. It should also support module load/unload.
>
> The timing calculations have been simplified to use pre-calculated tables
> compared to drivers/ide/pmac.c and it uses the new mediabay interface
> provided by a previous patch.
>
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> ---
>
> v2. Better tested now, seems to be reasonably solid.
>
> Addressed Tejun comments and made remove more robust vs. media-bay,
> should also fix Andreas problem.
Tested in 2.6.32-rc8 and 2.6.32 final on a G4 eMac.
Works fine, except for the yaboot problem:
> /sbin/installkernel 2.6.32 vmlinux System.map
ofpath: Driver: is not supported
ybin: Unable to find OpenFirmware path for boot=/dev/sda2
ybin: Please add ofboot=<path> where <path> is the OpenFirmware path to /dev/sda2 to /etc/yaboot.conf
ofpath: Driver: is not supported
ybin: Unable to determine OpenFirmware path for macosx=/dev/sda4
ybin: Try specifying the real OpenFirmware path for macosx=/dev/sda4 in /etc/yaboot.conf
(userspace is YDL 6.2)
/Mikael
^ permalink raw reply
* Re: [PATCH v2] i2c-mpc: generate START condition after STOP caused by read i2c_msg
From: Ben Dooks @ 2009-12-03 15:29 UTC (permalink / raw)
To: Michael Lawnick
Cc: linuxppc-dev, Esben Haabendal, Esben Haabendal, Ben Dooks,
linux-i2c
In-Reply-To: <4B17D4C5.3070100@gmx.de>
On Thu, Dec 03, 2009 at 04:09:57PM +0100, Michael Lawnick wrote:
> Ben Dooks said the following:
> > On Tue, May 26, 2009 at 01:30:21PM +0200, Esben Haabendal wrote:
> >> On Tue, May 19, 2009 at 7:22 AM, Esben Haabendal <eha@doredevelopment.dk> wrote:
> >> > This fixes MAL (arbitration lost) bug caused by illegal use of
> >> > RSTA (repeated START) after STOP condition generated after last byte
> >> > of reads. With this patch, it is possible to do an i2c_transfer() with
> >> > additional i2c_msg's following the I2C_M_RD messages.
> >> >
> >> > It still needs to be resolved if it is possible to fix this issue
> >> > by removing the STOP condition after reads in a robust way.
> >> >
> >> > Signed-off-by: Esben Haabendal <eha@doredevelopment.dk>
> >> > ---
> >> > ?drivers/i2c/busses/i2c-mpc.c | ? ?9 +++++++--
> >> > ?1 files changed, 7 insertions(+), 2 deletions(-)
> >>
> >> Any blockers to get this accepted?
> >
> > It would be nice to get an ack from someone who can actually test
> > the driver before getting this merged.
> >
> What is the state of this patch?
> Shouldn't we attack the problem on a more general way by inventing a
> Flag I2C_M_RESTART (or better I2C_M_NO_RESTART for backward compatibility)?
> This way the client driver is able to decide what it needs. If we do the
> choice within adapter, chance is about 50% to be wrong.
The documentation for 'struct i2c_msg' already says the STOP should only
be generated for the last message of the transfer. If STOP is being
generated for a message that isn't the last in the transfer than this
is incorrect behaviour.
Unless otherwise indicated, I'll put the patch in the series that I'll
send to Linus early next week.
--
Ben
Q: What's a light-year?
A: One-third less calories than a regular year.
^ permalink raw reply
* Re: [PATCH v2] i2c-mpc: generate START condition after STOP caused by read i2c_msg
From: Joakim Tjernlund @ 2009-12-03 15:42 UTC (permalink / raw)
To: Michael Lawnick
Cc: linuxppc-dev, Esben Haabendal, Esben Haabendal, Ben Dooks,
linux-i2c
In-Reply-To: <4B17D4C5.3070100@gmx.de>
>
> Ben Dooks said the following:
> > On Tue, May 26, 2009 at 01:30:21PM +0200, Esben Haabendal wrote:
> >> On Tue, May 19, 2009 at 7:22 AM, Esben Haabendal <eha@doredevelopment.dk> wrote:
> >> > This fixes MAL (arbitration lost) bug caused by illegal use of
> >> > RSTA (repeated START) after STOP condition generated after last byte
> >> > of reads. With this patch, it is possible to do an i2c_transfer() with
> >> > additional i2c_msg's following the I2C_M_RD messages.
> >> >
> >> > It still needs to be resolved if it is possible to fix this issue
> >> > by removing the STOP condition after reads in a robust way.
> >> >
> >> > Signed-off-by: Esben Haabendal <eha@doredevelopment.dk>
> >> > ---
> >> > ?drivers/i2c/busses/i2c-mpc.c | ? ?9 +++++++--
> >> > ?1 files changed, 7 insertions(+), 2 deletions(-)
> >>
> >> Any blockers to get this accepted?
> >
> > It would be nice to get an ack from someone who can actually test
> > the driver before getting this merged.
> >
> What is the state of this patch?
> Shouldn't we attack the problem on a more general way by inventing a
> Flag I2C_M_RESTART (or better I2C_M_NO_RESTART for backward compatibility)?
> This way the client driver is able to decide what it needs. If we do the
> choice within adapter, chance is about 50% to be wrong.
I have sent another patch (which I think is in Linus tree already) that
fixes the problem properly.
Can't remember if the driver handles M_RESTART or not. That is
a separate issue though.
Jocke
^ permalink raw reply
* Re: [PATCH v2] i2c-mpc: generate START condition after STOP caused by read i2c_msg
From: Joakim Tjernlund @ 2009-12-03 15:49 UTC (permalink / raw)
To: Ben Dooks
Cc: Esben Haabendal, Michael Lawnick, linuxppc-dev, Esben Haabendal,
Ben Dooks, linux-i2c
In-Reply-To: <20091203152916.GC23152@trinity.fluff.org>
>
> On Thu, Dec 03, 2009 at 04:09:57PM +0100, Michael Lawnick wrote:
> > Ben Dooks said the following:
> > > On Tue, May 26, 2009 at 01:30:21PM +0200, Esben Haabendal wrote:
> > >> On Tue, May 19, 2009 at 7:22 AM, Esben Haabendal <eha@doredevelopment.dk> wrote:
> > >> > This fixes MAL (arbitration lost) bug caused by illegal use of
> > >> > RSTA (repeated START) after STOP condition generated after last byte
> > >> > of reads. With this patch, it is possible to do an i2c_transfer() with
> > >> > additional i2c_msg's following the I2C_M_RD messages.
> > >> >
> > >> > It still needs to be resolved if it is possible to fix this issue
> > >> > by removing the STOP condition after reads in a robust way.
> > >> >
> > >> > Signed-off-by: Esben Haabendal <eha@doredevelopment.dk>
> > >> > ---
> > >> > ?drivers/i2c/busses/i2c-mpc.c | ? ?9 +++++++--
> > >> > ?1 files changed, 7 insertions(+), 2 deletions(-)
> > >>
> > >> Any blockers to get this accepted?
> > >
> > > It would be nice to get an ack from someone who can actually test
> > > the driver before getting this merged.
> > >
> > What is the state of this patch?
> > Shouldn't we attack the problem on a more general way by inventing a
> > Flag I2C_M_RESTART (or better I2C_M_NO_RESTART for backward compatibility)?
> > This way the client driver is able to decide what it needs. If we do the
> > choice within adapter, chance is about 50% to be wrong.
>
> The documentation for 'struct i2c_msg' already says the STOP should only
> be generated for the last message of the transfer. If STOP is being
> generated for a message that isn't the last in the transfer than this
> is incorrect behaviour.
>
> Unless otherwise indicated, I'll put the patch in the series that I'll
> send to Linus early next week.
Don't, already fixed properly and in Linus tree already.
^ permalink raw reply
* [PATCH v3] ppc440spe-adma: adds updated ppc440spe adma driver
From: Anatolij Gustschin @ 2009-12-03 16:36 UTC (permalink / raw)
To: linux-raid
Cc: wd, dzu, Yuri Tikhonov, linuxppc-dev, dan.j.williams,
Anatolij Gustschin
This patch adds new version of the PPC440SPe ADMA driver.
Signed-off-by: Yuri Tikhonov <yur@emcraft.com>
Signed-off-by: Anatolij Gustschin <agust@denx.de>
---
Changes since v2:
- Correct Sign-offs order.
- Place the driver in 'drivers/dma/ppc4xx' directory
as there are wishes to extend the driver to support
other CPUs (PPC460EX) later.
- Don't duplicate directory name in the driver file names.
- Perform checks for 'len' > PAGE_SIZE cases while
looking for suitable channel for async operations
so that there will be a fallback to synchronous path
if a user of the api should ever submit pq_val requests
with 'len' > PAGE_SIZE.
.../powerpc/dts-bindings/4xx/ppc440spe-adma.txt | 93 +
arch/powerpc/boot/dts/katmai.dts | 52 +-
arch/powerpc/include/asm/async_tx.h | 47 +
arch/powerpc/include/asm/dcr-regs.h | 23 +
drivers/dma/Kconfig | 11 +
drivers/dma/Makefile | 1 +
drivers/dma/ppc4xx/Makefile | 1 +
drivers/dma/ppc4xx/adma.c | 5027 ++++++++++++++++++++
drivers/dma/ppc4xx/adma.h | 195 +
drivers/dma/ppc4xx/dma.h | 223 +
drivers/dma/ppc4xx/xor.h | 110 +
11 files changed, 5782 insertions(+), 1 deletions(-)
create mode 100644 Documentation/powerpc/dts-bindings/4xx/ppc440spe-adma.txt
create mode 100644 arch/powerpc/include/asm/async_tx.h
create mode 100644 drivers/dma/ppc4xx/Makefile
create mode 100644 drivers/dma/ppc4xx/adma.c
create mode 100644 drivers/dma/ppc4xx/adma.h
create mode 100644 drivers/dma/ppc4xx/dma.h
create mode 100644 drivers/dma/ppc4xx/xor.h
diff --git a/Documentation/powerpc/dts-bindings/4xx/ppc440spe-adma.txt b/Documentation/powerpc/dts-bindings/4xx/ppc440spe-adma.txt
new file mode 100644
index 0000000..515ebcf
--- /dev/null
+++ b/Documentation/powerpc/dts-bindings/4xx/ppc440spe-adma.txt
@@ -0,0 +1,93 @@
+PPC440SPe DMA/XOR (DMA Controller and XOR Accelerator)
+
+Device nodes needed for operation of the ppc440spe-adma driver
+are specified hereby. These are I2O/DMA, DMA and XOR nodes
+for DMA engines and Memory Queue Module node. The latter is used
+by ADMA driver for configuration of RAID-6 H/W capabilities of
+the PPC440SPe. In addition to the nodes and properties described
+below, the ranges property of PLB node must specify ranges for
+DMA devices.
+
+ i) The I2O node
+
+ Required properties:
+
+ - compatible : "ibm,i2o-440spe";
+ - reg : <registers mapping>
+ - dcr-reg : <DCR registers range>
+
+ Example:
+
+ I2O: i2o@400100000 {
+ compatible = "ibm,i2o-440spe";
+ reg = <0x00000004 0x00100000 0x100>;
+ dcr-reg = <0x060 0x020>;
+ };
+
+
+ ii) The DMA node
+
+ Required properties:
+
+ - compatible : "ibm,dma-440spe";
+ - cell-index : 1 cell, hardware index of the DMA engine
+ (typically 0x0 and 0x1 for DMA0 and DMA1)
+ - reg : <registers mapping>
+ - dcr-reg : <DCR registers range>
+ - interrupts : <interrupt mapping for DMA0/1 interrupts sources:
+ 2 sources: DMAx CS FIFO Needs Service IRQ (on UIC0)
+ and DMA Error IRQ (on UIC1). The latter is common
+ for both DMA engines>.
+ - interrupt-parent : needed for interrupt mapping
+
+ Example:
+
+ DMA0: dma0@400100100 {
+ compatible = "ibm,dma-440spe";
+ cell-index = <0>;
+ reg = <0x00000004 0x00100100 0x100>;
+ dcr-reg = <0x060 0x020>;
+ interrupt-parent = <&DMA0>;
+ interrupts = <0 1>;
+ #interrupt-cells = <1>;
+ #address-cells = <0>;
+ #size-cells = <0>;
+ interrupt-map = <
+ 0 &UIC0 0x14 4
+ 1 &UIC1 0x16 4>;
+ };
+
+
+ iii) XOR Accelerator node
+
+ Required properties:
+
+ - compatible : "amcc,xor-accelerator";
+ - reg : <registers mapping>
+ - interrupts : <interrupt mapping for XOR interrupt source>
+ - interrupt-parent : for interrupt mapping
+
+ Example:
+
+ xor-accel@400200000 {
+ compatible = "amcc,xor-accelerator";
+ reg = <0x00000004 0x00200000 0x400>;
+ interrupt-parent = <&UIC1>;
+ interrupts = <0x1f 4>;
+ };
+
+
+ iv) Memory Queue Module node
+
+ Required properties:
+
+ - compatible : "ibm,mq-440spe";
+ - dcr-reg : <DCR registers range>
+
+ Example:
+
+ MQ0: mq {
+ compatible = "ibm,mq-440spe";
+ dcr-reg = <0x040 0x020>;
+ };
+
diff --git a/arch/powerpc/boot/dts/katmai.dts b/arch/powerpc/boot/dts/katmai.dts
index b8cd97c..033a134 100644
--- a/arch/powerpc/boot/dts/katmai.dts
+++ b/arch/powerpc/boot/dts/katmai.dts
@@ -108,12 +108,19 @@
dcr-reg = <0x00c 0x002>;
};
+ MQ0: mq {
+ compatible = "ibm,mq-440spe";
+ dcr-reg = <0x040 0x020>;
+ };
+
plb {
compatible = "ibm,plb-440spe", "ibm,plb-440gp", "ibm,plb4";
#address-cells = <2>;
#size-cells = <1>;
/* addr-child addr-parent size */
- ranges = <0x4 0xe0000000 0x4 0xe0000000 0x20000000
+ ranges = <0x4 0x00100000 0x4 0x00100000 0x00001000
+ 0x4 0x00200000 0x4 0x00200000 0x00000400
+ 0x4 0xe0000000 0x4 0xe0000000 0x20000000
0xc 0x00000000 0xc 0x00000000 0x20000000
0xd 0x00000000 0xd 0x00000000 0x80000000
0xd 0x80000000 0xd 0x80000000 0x80000000
@@ -400,6 +407,49 @@
0x0 0x0 0x0 0x3 &UIC3 0xa 0x4 /* swizzled int C */
0x0 0x0 0x0 0x4 &UIC3 0xb 0x4 /* swizzled int D */>;
};
+
+ I2O: i2o@400100000 {
+ compatible = "ibm,i2o-440spe";
+ reg = <0x00000004 0x00100000 0x100>;
+ dcr-reg = <0x060 0x020>;
+ };
+
+ DMA0: dma0@400100100 {
+ compatible = "ibm,dma-440spe";
+ cell-index = <0>;
+ reg = <0x00000004 0x00100100 0x100>;
+ dcr-reg = <0x060 0x020>;
+ interrupt-parent = <&DMA0>;
+ interrupts = <0 1>;
+ #interrupt-cells = <1>;
+ #address-cells = <0>;
+ #size-cells = <0>;
+ interrupt-map = <
+ 0 &UIC0 0x14 4
+ 1 &UIC1 0x16 4>;
+ };
+
+ DMA1: dma1@400100200 {
+ compatible = "ibm,dma-440spe";
+ cell-index = <1>;
+ reg = <0x00000004 0x00100200 0x100>;
+ dcr-reg = <0x060 0x020>;
+ interrupt-parent = <&DMA1>;
+ interrupts = <0 1>;
+ #interrupt-cells = <1>;
+ #address-cells = <0>;
+ #size-cells = <0>;
+ interrupt-map = <
+ 0 &UIC0 0x16 4
+ 1 &UIC1 0x16 4>;
+ };
+
+ xor-accel@400200000 {
+ compatible = "amcc,xor-accelerator";
+ reg = <0x00000004 0x00200000 0x400>;
+ interrupt-parent = <&UIC1>;
+ interrupts = <0x1f 4>;
+ };
};
chosen {
diff --git a/arch/powerpc/include/asm/async_tx.h b/arch/powerpc/include/asm/async_tx.h
new file mode 100644
index 0000000..8b2dc55
--- /dev/null
+++ b/arch/powerpc/include/asm/async_tx.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (C) 2008-2009 DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * The full GNU General Public License is included in this distribution in the
+ * file called COPYING.
+ */
+#ifndef _ASM_POWERPC_ASYNC_TX_H_
+#define _ASM_POWERPC_ASYNC_TX_H_
+
+#if defined(CONFIG_440SPe) || defined(CONFIG_440SP)
+extern struct dma_chan *
+ppc440spe_async_tx_find_best_channel(enum dma_transaction_type cap,
+ struct page **dst_lst, int dst_cnt, struct page **src_lst,
+ int src_cnt, size_t src_sz);
+
+#define async_tx_find_channel(dep, cap, dst_lst, dst_cnt, src_lst, \
+ src_cnt, src_sz) \
+ ppc440spe_async_tx_find_best_channel(cap, dst_lst, dst_cnt, src_lst, \
+ src_cnt, src_sz)
+#else
+
+#define async_tx_find_channel(dep, type, dst, dst_count, src, src_count, len) \
+ __async_tx_find_channel(dep, type)
+
+struct dma_chan *
+__async_tx_find_channel(struct async_submit_ctl *submit,
+ enum dma_transaction_type tx_type);
+
+#endif
+
+#endif
diff --git a/arch/powerpc/include/asm/dcr-regs.h b/arch/powerpc/include/asm/dcr-regs.h
index 828e3aa..380274d 100644
--- a/arch/powerpc/include/asm/dcr-regs.h
+++ b/arch/powerpc/include/asm/dcr-regs.h
@@ -157,4 +157,27 @@
#define L2C_SNP_SSR_32G 0x0000f000
#define L2C_SNP_ESR 0x00000800
+/*
+ * DCR register offsets for 440SP/440SPe I2O/DMA controller.
+ * The base address is configured in the device tree.
+ */
+#define DCRN_I2O0_IBAL 0x006
+#define DCRN_I2O0_IBAH 0x007
+#define I2O_REG_ENABLE 0x00000001 /* Enable I2O/DMA access */
+
+/* 440SP/440SPe Software Reset DCR */
+#define DCRN_SDR0_SRST 0x0200
+#define DCRN_SDR0_SRST_I2ODMA (0x80000000 >> 15) /* Reset I2O/DMA */
+
+/* 440SP/440SPe Memory Queue DCR offsets */
+#define DCRN_MQ0_XORBA 0x04
+#define DCRN_MQ0_CF2H 0x06
+#define DCRN_MQ0_CFBHL 0x0f
+#define DCRN_MQ0_BAUH 0x10
+
+/* HB/LL Paths Configuration Register */
+#define MQ0_CFBHL_TPLM 28
+#define MQ0_CFBHL_HBCL 23
+#define MQ0_CFBHL_POLY 15
+
#endif /* __DCR_REGS_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index b401dad..af43aca 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -111,6 +111,17 @@ config SH_DMAE
help
Enable support for the Renesas SuperH DMA controllers.
+config AMCC_PPC440SPE_ADMA
+ tristate "AMCC PPC440SPe ADMA support"
+ depends on 440SPe || 440SP
+ select DMA_ENGINE
+ select ARCH_HAS_ASYNC_TX_FIND_CHANNEL
+ help
+ Enable support for the AMCC PPC440SPe RAID engines.
+
+config ARCH_HAS_ASYNC_TX_FIND_CHANNEL
+ bool
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index eca71ba..ab697ae 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -10,3 +10,4 @@ obj-$(CONFIG_AT_HDMAC) += at_hdmac.o
obj-$(CONFIG_MX3_IPU) += ipu/
obj-$(CONFIG_TXX9_DMAC) += txx9dmac.o
obj-$(CONFIG_SH_DMAE) += shdma.o
+obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += ppc4xx/
diff --git a/drivers/dma/ppc4xx/Makefile b/drivers/dma/ppc4xx/Makefile
new file mode 100644
index 0000000..b3d259b
--- /dev/null
+++ b/drivers/dma/ppc4xx/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += adma.o
diff --git a/drivers/dma/ppc4xx/adma.c b/drivers/dma/ppc4xx/adma.c
new file mode 100644
index 0000000..0a3478e
--- /dev/null
+++ b/drivers/dma/ppc4xx/adma.c
@@ -0,0 +1,5027 @@
+/*
+ * Copyright (C) 2006-2009 DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * Further porting to arch/powerpc by
+ * Anatolij Gustschin <agust@denx.de>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * The full GNU General Public License is included in this distribution in the
+ * file called COPYING.
+ */
+
+/*
+ * This driver supports the asynchrounous DMA copy and RAID engines available
+ * on the AMCC PPC440SPe Processors.
+ * Based on the Intel Xscale(R) family of I/O Processors (IOP 32x, 33x, 134x)
+ * ADMA driver written by D.Williams.
+ */
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/async_tx.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+#include <linux/uaccess.h>
+#include <linux/proc_fs.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <asm/dcr.h>
+#include <asm/dcr-regs.h>
+#include "adma.h"
+
+enum ppc_adma_init_code {
+ PPC_ADMA_INIT_OK = 0,
+ PPC_ADMA_INIT_MEMRES,
+ PPC_ADMA_INIT_MEMREG,
+ PPC_ADMA_INIT_ALLOC,
+ PPC_ADMA_INIT_COHERENT,
+ PPC_ADMA_INIT_CHANNEL,
+ PPC_ADMA_INIT_IRQ1,
+ PPC_ADMA_INIT_IRQ2,
+ PPC_ADMA_INIT_REGISTER
+};
+
+static char *ppc_adma_errors[] = {
+ [PPC_ADMA_INIT_OK] = "ok",
+ [PPC_ADMA_INIT_MEMRES] = "failed to get memory resource",
+ [PPC_ADMA_INIT_MEMREG] = "failed to request memory region",
+ [PPC_ADMA_INIT_ALLOC] = "failed to allocate memory for adev "
+ "structure",
+ [PPC_ADMA_INIT_COHERENT] = "failed to allocate coherent memory for "
+ "hardware descriptors",
+ [PPC_ADMA_INIT_CHANNEL] = "failed to allocate memory for channel",
+ [PPC_ADMA_INIT_IRQ1] = "failed to request first irq",
+ [PPC_ADMA_INIT_IRQ2] = "failed to request second irq",
+ [PPC_ADMA_INIT_REGISTER] = "failed to register dma async device",
+};
+
+static enum ppc_adma_init_code
+ppc440spe_adma_devices[PPC440SPE_ADMA_ENGINES_NUM];
+
+struct ppc_dma_chan_ref {
+ struct dma_chan *chan;
+ struct list_head node;
+};
+
+/* The list of channels exported by ppc440spe ADMA */
+struct list_head
+ppc440spe_adma_chan_list = LIST_HEAD_INIT(ppc440spe_adma_chan_list);
+
+/* This flag is set when want to refetch the xor chain in the interrupt
+ * handler
+ */
+static u32 do_xor_refetch;
+
+/* Pointer to DMA0, DMA1 CP/CS FIFO */
+static void *ppc440spe_dma_fifo_buf;
+
+/* Pointers to last submitted to DMA0, DMA1 CDBs */
+static struct ppc440spe_adma_desc_slot *chan_last_sub[3];
+static struct ppc440spe_adma_desc_slot *chan_first_cdb[3];
+
+/* Pointer to last linked and submitted xor CB */
+static struct ppc440spe_adma_desc_slot *xor_last_linked;
+static struct ppc440spe_adma_desc_slot *xor_last_submit;
+
+/* This array is used in data-check operations for storing a pattern */
+static char ppc440spe_qword[16];
+
+static atomic_t ppc440spe_adma_err_irq_ref;
+static dcr_host_t ppc440spe_mq_dcr_host;
+static unsigned int ppc440spe_mq_dcr_len;
+
+/* Since RXOR operations use the common register (MQ0_CF2H) for setting-up
+ * the block size in transactions, then we do not allow to activate more than
+ * only one RXOR transactions simultaneously. So use this var to store
+ * the information about is RXOR currently active (PPC440SPE_RXOR_RUN bit is
+ * set) or not (PPC440SPE_RXOR_RUN is clear).
+ */
+static unsigned long ppc440spe_rxor_state;
+
+/* These are used in enable & check routines
+ */
+static u32 ppc440spe_r6_enabled;
+static struct ppc440spe_adma_chan *ppc440spe_r6_tchan;
+static struct completion ppc440spe_r6_test_comp;
+
+static int ppc440spe_adma_dma2rxor_prep_src(
+ struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_rxor *cursor, int index,
+ int src_cnt, u32 addr);
+static void ppc440spe_adma_dma2rxor_set_src(
+ struct ppc440spe_adma_desc_slot *desc,
+ int index, dma_addr_t addr);
+static void ppc440spe_adma_dma2rxor_set_mult(
+ struct ppc440spe_adma_desc_slot *desc,
+ int index, u8 mult);
+
+#ifdef ADMA_LL_DEBUG
+#define ADMA_LL_DBG(x) ({ if (1) x; 0; })
+#else
+#define ADMA_LL_DBG(x) ({ if (0) x; 0; })
+#endif
+
+static void print_cb(struct ppc440spe_adma_chan *chan, void *block)
+{
+ struct dma_cdb *cdb;
+ struct xor_cb *cb;
+ int i;
+
+ switch (chan->device->id) {
+ case 0:
+ case 1:
+ cdb = block;
+
+ pr_debug("CDB at %p [%d]:\n"
+ "\t attr 0x%02x opc 0x%02x cnt 0x%08x\n"
+ "\t sg1u 0x%08x sg1l 0x%08x\n"
+ "\t sg2u 0x%08x sg2l 0x%08x\n"
+ "\t sg3u 0x%08x sg3l 0x%08x\n",
+ cdb, chan->device->id,
+ cdb->attr, cdb->opc, le32_to_cpu(cdb->cnt),
+ le32_to_cpu(cdb->sg1u), le32_to_cpu(cdb->sg1l),
+ le32_to_cpu(cdb->sg2u), le32_to_cpu(cdb->sg2l),
+ le32_to_cpu(cdb->sg3u), le32_to_cpu(cdb->sg3l)
+ );
+ break;
+ case 2:
+ cb = block;
+
+ pr_debug("CB at %p [%d]:\n"
+ "\t cbc 0x%08x cbbc 0x%08x cbs 0x%08x\n"
+ "\t cbtah 0x%08x cbtal 0x%08x\n"
+ "\t cblah 0x%08x cblal 0x%08x\n",
+ cb, chan->device->id,
+ cb->cbc, cb->cbbc, cb->cbs,
+ cb->cbtah, cb->cbtal,
+ cb->cblah, cb->cblal);
+ for (i = 0; i < 16; i++) {
+ if (i && !cb->ops[i].h && !cb->ops[i].l)
+ continue;
+ pr_debug("\t ops[%2d]: h 0x%08x l 0x%08x\n",
+ i, cb->ops[i].h, cb->ops[i].l);
+ }
+ break;
+ }
+}
+
+static void print_cb_list(struct ppc440spe_adma_chan *chan,
+ struct ppc440spe_adma_desc_slot *iter)
+{
+ for (; iter; iter = iter->hw_next)
+ print_cb(chan, iter->hw_desc);
+}
+
+static void prep_dma_xor_dbg(int id, dma_addr_t dst, dma_addr_t *src,
+ unsigned int src_cnt)
+{
+ int i;
+
+ pr_debug("\n%s(%d):\nsrc: ", __func__, id);
+ for (i = 0; i < src_cnt; i++)
+ pr_debug("\t0x%016llx ", src[i]);
+ pr_debug("dst:\n\t0x%016llx\n", dst);
+}
+
+static void prep_dma_pq_dbg(int id, dma_addr_t *dst, dma_addr_t *src,
+ unsigned int src_cnt)
+{
+ int i;
+
+ pr_debug("\n%s(%d):\nsrc: ", __func__, id);
+ for (i = 0; i < src_cnt; i++)
+ pr_debug("\t0x%016llx ", src[i]);
+ pr_debug("dst: ");
+ for (i = 0; i < 2; i++)
+ pr_debug("\t0x%016llx ", dst[i]);
+}
+
+static void prep_dma_pqzero_sum_dbg(int id, dma_addr_t *src,
+ unsigned int src_cnt,
+ const unsigned char *scf)
+{
+ int i;
+
+ pr_debug("\n%s(%d):\nsrc(coef): ", __func__, id);
+ if (scf) {
+ for (i = 0; i < src_cnt; i++)
+ pr_debug("\t0x%016llx(0x%02x) ", src[i], scf[i]);
+ } else {
+ for (i = 0; i < src_cnt; i++)
+ pr_debug("\t0x%016llx(no) ", src[i]);
+ }
+
+ pr_debug("dst: ");
+ for (i = 0; i < 2; i++)
+ pr_debug("\t0x%016llx ", src[src_cnt + i]);
+}
+
+/******************************************************************************
+ * Command (Descriptor) Blocks low-level routines
+ ******************************************************************************/
+/**
+ * ppc440spe_desc_init_interrupt - initialize the descriptor for INTERRUPT
+ * pseudo operation
+ */
+static void ppc440spe_desc_init_interrupt(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan)
+{
+ struct xor_cb *p;
+
+ switch (chan->device->id) {
+ case PPC440SPE_XOR_ID:
+ p = desc->hw_desc;
+ memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+ /* NOP with Command Block Complete Enable */
+ p->cbc = XOR_CBCR_CBCE_BIT;
+ break;
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+ /* NOP with interrupt */
+ set_bit(PPC440SPE_DESC_INT, &desc->flags);
+ break;
+ default:
+ printk(KERN_ERR "Unsupported id %d in %s\n", chan->device->id,
+ __func__);
+ break;
+ }
+}
+
+/**
+ * ppc440spe_desc_init_null_xor - initialize the descriptor for NULL XOR
+ * pseudo operation
+ */
+static void ppc440spe_desc_init_null_xor(struct ppc440spe_adma_desc_slot *desc)
+{
+ memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+ desc->hw_next = NULL;
+ desc->src_cnt = 0;
+ desc->dst_cnt = 1;
+}
+
+/**
+ * ppc440spe_desc_init_xor - initialize the descriptor for XOR operation
+ */
+static void ppc440spe_desc_init_xor(struct ppc440spe_adma_desc_slot *desc,
+ int src_cnt, unsigned long flags)
+{
+ struct xor_cb *hw_desc = desc->hw_desc;
+
+ memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+ desc->hw_next = NULL;
+ desc->src_cnt = src_cnt;
+ desc->dst_cnt = 1;
+
+ hw_desc->cbc = XOR_CBCR_TGT_BIT | src_cnt;
+ if (flags & DMA_PREP_INTERRUPT)
+ /* Enable interrupt on completion */
+ hw_desc->cbc |= XOR_CBCR_CBCE_BIT;
+}
+
+/**
+ * ppc440spe_desc_init_dma2pq - initialize the descriptor for PQ
+ * operation in DMA2 controller
+ */
+static void ppc440spe_desc_init_dma2pq(struct ppc440spe_adma_desc_slot *desc,
+ int dst_cnt, int src_cnt, unsigned long flags)
+{
+ struct xor_cb *hw_desc = desc->hw_desc;
+
+ memset(desc->hw_desc, 0, sizeof(struct xor_cb));
+ desc->hw_next = NULL;
+ desc->src_cnt = src_cnt;
+ desc->dst_cnt = dst_cnt;
+ memset(desc->reverse_flags, 0, sizeof(desc->reverse_flags));
+ desc->descs_per_op = 0;
+
+ hw_desc->cbc = XOR_CBCR_TGT_BIT;
+ if (flags & DMA_PREP_INTERRUPT)
+ /* Enable interrupt on completion */
+ hw_desc->cbc |= XOR_CBCR_CBCE_BIT;
+}
+
+#define DMA_CTRL_FLAGS_LAST DMA_PREP_FENCE
+#define DMA_PREP_ZERO_P (DMA_CTRL_FLAGS_LAST << 1)
+#define DMA_PREP_ZERO_Q (DMA_PREP_ZERO_P << 1)
+
+/**
+ * ppc440spe_desc_init_dma01pq - initialize the descriptors for PQ operation
+ * with DMA0/1
+ */
+static void ppc440spe_desc_init_dma01pq(struct ppc440spe_adma_desc_slot *desc,
+ int dst_cnt, int src_cnt, unsigned long flags,
+ unsigned long op)
+{
+ struct dma_cdb *hw_desc;
+ struct ppc440spe_adma_desc_slot *iter;
+ u8 dopc;
+
+ /* Common initialization of a PQ descriptors chain */
+ set_bits(op, &desc->flags);
+ desc->src_cnt = src_cnt;
+ desc->dst_cnt = dst_cnt;
+
+ /* WXOR MULTICAST if both P and Q are being computed
+ * MV_SG1_SG2 if Q only
+ */
+ dopc = (desc->dst_cnt == DMA_DEST_MAX_NUM) ?
+ DMA_CDB_OPC_MULTICAST : DMA_CDB_OPC_MV_SG1_SG2;
+
+ list_for_each_entry(iter, &desc->group_list, chain_node) {
+ hw_desc = iter->hw_desc;
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+
+ if (likely(!list_is_last(&iter->chain_node,
+ &desc->group_list))) {
+ /* set 'next' pointer */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot, chain_node);
+ clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+ } else {
+ /* this is the last descriptor.
+ * this slot will be pasted from ADMA level
+ * each time it wants to configure parameters
+ * of the transaction (src, dst, ...)
+ */
+ iter->hw_next = NULL;
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC440SPE_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+ }
+ }
+
+ /* Set OPS depending on WXOR/RXOR type of operation */
+ if (!test_bit(PPC440SPE_DESC_RXOR, &desc->flags)) {
+ /* This is a WXOR only chain:
+ * - first descriptors are for zeroing destinations
+ * if PPC440SPE_ZERO_P/Q set;
+ * - descriptors remained are for GF-XOR operations.
+ */
+ iter = list_first_entry(&desc->group_list,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+
+ if (test_bit(PPC440SPE_ZERO_P, &desc->flags)) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ }
+
+ if (test_bit(PPC440SPE_ZERO_Q, &desc->flags)) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ }
+
+ list_for_each_entry_from(iter, &desc->group_list, chain_node) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = dopc;
+ }
+ } else {
+ /* This is either RXOR-only or mixed RXOR/WXOR */
+
+ /* The first 1 or 2 slots in chain are always RXOR,
+ * if need to calculate P & Q, then there are two
+ * RXOR slots; if only P or only Q, then there is one
+ */
+ iter = list_first_entry(&desc->group_list,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+
+ if (desc->dst_cnt == DMA_DEST_MAX_NUM) {
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ }
+
+ /* The remaining descs (if any) are WXORs */
+ if (test_bit(PPC440SPE_DESC_WXOR, &desc->flags)) {
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ list_for_each_entry_from(iter, &desc->group_list,
+ chain_node) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = dopc;
+ }
+ }
+ }
+}
+
+/**
+ * ppc440spe_desc_init_dma01pqzero_sum - initialize the descriptor
+ * for PQ_ZERO_SUM operation
+ */
+static void ppc440spe_desc_init_dma01pqzero_sum(
+ struct ppc440spe_adma_desc_slot *desc,
+ int dst_cnt, int src_cnt)
+{
+ struct dma_cdb *hw_desc;
+ struct ppc440spe_adma_desc_slot *iter;
+ int i = 0;
+ u8 dopc = (dst_cnt == 2) ? DMA_CDB_OPC_MULTICAST :
+ DMA_CDB_OPC_MV_SG1_SG2;
+ /*
+ * Initialize starting from 2nd or 3rd descriptor dependent
+ * on dst_cnt. First one or two slots are for cloning P
+ * and/or Q to chan->pdest and/or chan->qdest as we have
+ * to preserve original P/Q.
+ */
+ iter = list_first_entry(&desc->group_list,
+ struct ppc440spe_adma_desc_slot, chain_node);
+ iter = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot, chain_node);
+
+ if (dst_cnt > 1) {
+ iter = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot, chain_node);
+ }
+ /* initialize each source descriptor in chain */
+ list_for_each_entry_from(iter, &desc->group_list, chain_node) {
+ hw_desc = iter->hw_desc;
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->src_cnt = 0;
+ iter->dst_cnt = 0;
+
+ /* This is a ZERO_SUM operation:
+ * - <src_cnt> descriptors starting from 2nd or 3rd
+ * descriptor are for GF-XOR operations;
+ * - remaining <dst_cnt> descriptors are for checking the result
+ */
+ if (i++ < src_cnt)
+ /* MV_SG1_SG2 if only Q is being verified
+ * MULTICAST if both P and Q are being verified
+ */
+ hw_desc->opc = dopc;
+ else
+ /* DMA_CDB_OPC_DCHECK128 operation */
+ hw_desc->opc = DMA_CDB_OPC_DCHECK128;
+
+ if (likely(!list_is_last(&iter->chain_node,
+ &desc->group_list))) {
+ /* set 'next' pointer */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ } else {
+ /* this is the last descriptor.
+ * this slot will be pasted from ADMA level
+ * each time it wants to configure parameters
+ * of the transaction (src, dst, ...)
+ */
+ iter->hw_next = NULL;
+ /* always enable interrupt generation since we get
+ * the status of pqzero from the handler
+ */
+ set_bit(PPC440SPE_DESC_INT, &iter->flags);
+ }
+ }
+ desc->src_cnt = src_cnt;
+ desc->dst_cnt = dst_cnt;
+}
+
+/**
+ * ppc440spe_desc_init_memcpy - initialize the descriptor for MEMCPY operation
+ */
+static void ppc440spe_desc_init_memcpy(struct ppc440spe_adma_desc_slot *desc,
+ unsigned long flags)
+{
+ struct dma_cdb *hw_desc = desc->hw_desc;
+
+ memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+ desc->hw_next = NULL;
+ desc->src_cnt = 1;
+ desc->dst_cnt = 1;
+
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC440SPE_DESC_INT, &desc->flags);
+ else
+ clear_bit(PPC440SPE_DESC_INT, &desc->flags);
+
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+}
+
+/**
+ * ppc440spe_desc_init_memset - initialize the descriptor for MEMSET operation
+ */
+static void ppc440spe_desc_init_memset(struct ppc440spe_adma_desc_slot *desc,
+ int value, unsigned long flags)
+{
+ struct dma_cdb *hw_desc = desc->hw_desc;
+
+ memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+ desc->hw_next = NULL;
+ desc->src_cnt = 1;
+ desc->dst_cnt = 1;
+
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC440SPE_DESC_INT, &desc->flags);
+ else
+ clear_bit(PPC440SPE_DESC_INT, &desc->flags);
+
+ hw_desc->sg1u = hw_desc->sg1l = cpu_to_le32((u32)value);
+ hw_desc->sg3u = hw_desc->sg3l = cpu_to_le32((u32)value);
+ hw_desc->opc = DMA_CDB_OPC_DFILL128;
+}
+
+/**
+ * ppc440spe_desc_set_src_addr - set source address into the descriptor
+ */
+static void ppc440spe_desc_set_src_addr(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan,
+ int src_idx, dma_addr_t addrh,
+ dma_addr_t addrl)
+{
+ struct dma_cdb *dma_hw_desc;
+ struct xor_cb *xor_hw_desc;
+ phys_addr_t addr64, tmplow, tmphi;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ if (!addrh) {
+ addr64 = addrl;
+ tmphi = (addr64 >> 32);
+ tmplow = (addr64 & 0xFFFFFFFF);
+ } else {
+ tmphi = addrh;
+ tmplow = addrl;
+ }
+ dma_hw_desc = desc->hw_desc;
+ dma_hw_desc->sg1l = cpu_to_le32((u32)tmplow);
+ dma_hw_desc->sg1u |= cpu_to_le32((u32)tmphi);
+ break;
+ case PPC440SPE_XOR_ID:
+ xor_hw_desc = desc->hw_desc;
+ xor_hw_desc->ops[src_idx].l = addrl;
+ xor_hw_desc->ops[src_idx].h |= addrh;
+ break;
+ }
+}
+
+/**
+ * ppc440spe_desc_set_src_mult - set source address mult into the descriptor
+ */
+static void ppc440spe_desc_set_src_mult(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan, u32 mult_index,
+ int sg_index, unsigned char mult_value)
+{
+ struct dma_cdb *dma_hw_desc;
+ struct xor_cb *xor_hw_desc;
+ u32 *psgu;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_hw_desc = desc->hw_desc;
+
+ switch (sg_index) {
+ /* for RXOR operations set multiplier
+ * into source cued address
+ */
+ case DMA_CDB_SG_SRC:
+ psgu = &dma_hw_desc->sg1u;
+ break;
+ /* for WXOR operations set multiplier
+ * into destination cued address(es)
+ */
+ case DMA_CDB_SG_DST1:
+ psgu = &dma_hw_desc->sg2u;
+ break;
+ case DMA_CDB_SG_DST2:
+ psgu = &dma_hw_desc->sg3u;
+ break;
+ default:
+ BUG();
+ }
+
+ *psgu |= cpu_to_le32(mult_value << mult_index);
+ break;
+ case PPC440SPE_XOR_ID:
+ xor_hw_desc = desc->hw_desc;
+ break;
+ default:
+ BUG();
+ }
+}
+
+/**
+ * ppc440spe_desc_set_dest_addr - set destination address into the descriptor
+ */
+static void ppc440spe_desc_set_dest_addr(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan,
+ dma_addr_t addrh, dma_addr_t addrl,
+ u32 dst_idx)
+{
+ struct dma_cdb *dma_hw_desc;
+ struct xor_cb *xor_hw_desc;
+ phys_addr_t addr64, tmphi, tmplow;
+ u32 *psgu, *psgl;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ if (!addrh) {
+ addr64 = addrl;
+ tmphi = (addr64 >> 32);
+ tmplow = (addr64 & 0xFFFFFFFF);
+ } else {
+ tmphi = addrh;
+ tmplow = addrl;
+ }
+ dma_hw_desc = desc->hw_desc;
+
+ psgu = dst_idx ? &dma_hw_desc->sg3u : &dma_hw_desc->sg2u;
+ psgl = dst_idx ? &dma_hw_desc->sg3l : &dma_hw_desc->sg2l;
+
+ *psgl = cpu_to_le32((u32)tmplow);
+ *psgu |= cpu_to_le32((u32)tmphi);
+ break;
+ case PPC440SPE_XOR_ID:
+ xor_hw_desc = desc->hw_desc;
+ xor_hw_desc->cbtal = addrl;
+ xor_hw_desc->cbtah |= addrh;
+ break;
+ }
+}
+
+/**
+ * ppc440spe_desc_set_byte_count - set number of data bytes involved
+ * into the operation
+ */
+static void ppc440spe_desc_set_byte_count(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan,
+ u32 byte_count)
+{
+ struct dma_cdb *dma_hw_desc;
+ struct xor_cb *xor_hw_desc;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_hw_desc = desc->hw_desc;
+ dma_hw_desc->cnt = cpu_to_le32(byte_count);
+ break;
+ case PPC440SPE_XOR_ID:
+ xor_hw_desc = desc->hw_desc;
+ xor_hw_desc->cbbc = byte_count;
+ break;
+ }
+}
+
+/**
+ * ppc440spe_desc_set_rxor_block_size - set RXOR block size
+ */
+static inline void ppc440spe_desc_set_rxor_block_size(u32 byte_count)
+{
+ /* assume that byte_count is aligned on the 512-boundary;
+ * thus write it directly to the register (bits 23:31 are
+ * reserved there).
+ */
+ dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_CF2H, byte_count);
+}
+
+/**
+ * ppc440spe_desc_set_dcheck - set CHECK pattern
+ */
+static void ppc440spe_desc_set_dcheck(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan, u8 *qword)
+{
+ struct dma_cdb *dma_hw_desc;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_hw_desc = desc->hw_desc;
+ iowrite32(qword[0], &dma_hw_desc->sg3l);
+ iowrite32(qword[4], &dma_hw_desc->sg3u);
+ iowrite32(qword[8], &dma_hw_desc->sg2l);
+ iowrite32(qword[12], &dma_hw_desc->sg2u);
+ break;
+ default:
+ BUG();
+ }
+}
+
+/**
+ * ppc440spe_xor_set_link - set link address in xor CB
+ */
+static void ppc440spe_xor_set_link(struct ppc440spe_adma_desc_slot *prev_desc,
+ struct ppc440spe_adma_desc_slot *next_desc)
+{
+ struct xor_cb *xor_hw_desc = prev_desc->hw_desc;
+
+ if (unlikely(!next_desc || !(next_desc->phys))) {
+ printk(KERN_ERR "%s: next_desc=0x%p; next_desc->phys=0x%llx\n",
+ __func__, next_desc,
+ next_desc ? next_desc->phys : 0);
+ BUG();
+ }
+
+ xor_hw_desc->cbs = 0;
+ xor_hw_desc->cblal = next_desc->phys;
+ xor_hw_desc->cblah = 0;
+ xor_hw_desc->cbc |= XOR_CBCR_LNK_BIT;
+}
+
+/**
+ * ppc440spe_desc_set_link - set the address of descriptor following this
+ * descriptor in chain
+ */
+static void ppc440spe_desc_set_link(struct ppc440spe_adma_chan *chan,
+ struct ppc440spe_adma_desc_slot *prev_desc,
+ struct ppc440spe_adma_desc_slot *next_desc)
+{
+ unsigned long flags;
+ struct ppc440spe_adma_desc_slot *tail = next_desc;
+
+ if (unlikely(!prev_desc || !next_desc ||
+ (prev_desc->hw_next && prev_desc->hw_next != next_desc))) {
+ /* If previous next is overwritten something is wrong.
+ * though we may refetch from append to initiate list
+ * processing; in this case - it's ok.
+ */
+ printk(KERN_ERR "%s: prev_desc=0x%p; next_desc=0x%p; "
+ "prev->hw_next=0x%p\n", __func__, prev_desc,
+ next_desc, prev_desc ? prev_desc->hw_next : 0);
+ BUG();
+ }
+
+ local_irq_save(flags);
+
+ /* do s/w chaining both for DMA and XOR descriptors */
+ prev_desc->hw_next = next_desc;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ break;
+ case PPC440SPE_XOR_ID:
+ /* bind descriptor to the chain */
+ while (tail->hw_next)
+ tail = tail->hw_next;
+ xor_last_linked = tail;
+
+ if (prev_desc == xor_last_submit)
+ /* do not link to the last submitted CB */
+ break;
+ ppc440spe_xor_set_link(prev_desc, next_desc);
+ break;
+ }
+
+ local_irq_restore(flags);
+}
+
+/**
+ * ppc440spe_desc_get_src_addr - extract the source address from the descriptor
+ */
+static u32 ppc440spe_desc_get_src_addr(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan, int src_idx)
+{
+ struct dma_cdb *dma_hw_desc;
+ struct xor_cb *xor_hw_desc;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_hw_desc = desc->hw_desc;
+ /* May have 0, 1, 2, or 3 sources */
+ switch (dma_hw_desc->opc) {
+ case DMA_CDB_OPC_NO_OP:
+ case DMA_CDB_OPC_DFILL128:
+ return 0;
+ case DMA_CDB_OPC_DCHECK128:
+ if (unlikely(src_idx)) {
+ printk(KERN_ERR "%s: try to get %d source for"
+ " DCHECK128\n", __func__, src_idx);
+ BUG();
+ }
+ return le32_to_cpu(dma_hw_desc->sg1l);
+ case DMA_CDB_OPC_MULTICAST:
+ case DMA_CDB_OPC_MV_SG1_SG2:
+ if (unlikely(src_idx > 2)) {
+ printk(KERN_ERR "%s: try to get %d source from"
+ " DMA descr\n", __func__, src_idx);
+ BUG();
+ }
+ if (src_idx) {
+ if (le32_to_cpu(dma_hw_desc->sg1u) &
+ DMA_CUED_XOR_WIN_MSK) {
+ u8 region;
+
+ if (src_idx == 1)
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ desc->unmap_len;
+
+ region = (le32_to_cpu(
+ dma_hw_desc->sg1u)) >>
+ DMA_CUED_REGION_OFF;
+
+ region &= DMA_CUED_REGION_MSK;
+ switch (region) {
+ case DMA_RXOR123:
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ (desc->unmap_len << 1);
+ case DMA_RXOR124:
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ (desc->unmap_len * 3);
+ case DMA_RXOR125:
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ (desc->unmap_len << 2);
+ default:
+ printk(KERN_ERR
+ "%s: try to"
+ " get src3 for region %02x"
+ "PPC440SPE_DESC_RXOR12?\n",
+ __func__, region);
+ BUG();
+ }
+ } else {
+ printk(KERN_ERR
+ "%s: try to get %d"
+ " source for non-cued descr\n",
+ __func__, src_idx);
+ BUG();
+ }
+ }
+ return le32_to_cpu(dma_hw_desc->sg1l);
+ default:
+ printk(KERN_ERR "%s: unknown OPC 0x%02x\n",
+ __func__, dma_hw_desc->opc);
+ BUG();
+ }
+ return le32_to_cpu(dma_hw_desc->sg1l);
+ case PPC440SPE_XOR_ID:
+ /* May have up to 16 sources */
+ xor_hw_desc = desc->hw_desc;
+ return xor_hw_desc->ops[src_idx].l;
+ }
+ return 0;
+}
+
+/**
+ * ppc440spe_desc_get_dest_addr - extract the destination address from the
+ * descriptor
+ */
+static u32 ppc440spe_desc_get_dest_addr(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan, int idx)
+{
+ struct dma_cdb *dma_hw_desc;
+ struct xor_cb *xor_hw_desc;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_hw_desc = desc->hw_desc;
+
+ if (likely(!idx))
+ return le32_to_cpu(dma_hw_desc->sg2l);
+ return le32_to_cpu(dma_hw_desc->sg3l);
+ case PPC440SPE_XOR_ID:
+ xor_hw_desc = desc->hw_desc;
+ return xor_hw_desc->cbtal;
+ }
+ return 0;
+}
+
+/**
+ * ppc440spe_desc_get_src_num - extract the number of source addresses from
+ * the descriptor
+ */
+static u32 ppc440spe_desc_get_src_num(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan)
+{
+ struct dma_cdb *dma_hw_desc;
+ struct xor_cb *xor_hw_desc;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_hw_desc = desc->hw_desc;
+
+ switch (dma_hw_desc->opc) {
+ case DMA_CDB_OPC_NO_OP:
+ case DMA_CDB_OPC_DFILL128:
+ return 0;
+ case DMA_CDB_OPC_DCHECK128:
+ return 1;
+ case DMA_CDB_OPC_MV_SG1_SG2:
+ case DMA_CDB_OPC_MULTICAST:
+ /*
+ * Only for RXOR operations we have more than
+ * one source
+ */
+ if (le32_to_cpu(dma_hw_desc->sg1u) &
+ DMA_CUED_XOR_WIN_MSK) {
+ /* RXOR op, there are 2 or 3 sources */
+ if (((le32_to_cpu(dma_hw_desc->sg1u) >>
+ DMA_CUED_REGION_OFF) &
+ DMA_CUED_REGION_MSK) == DMA_RXOR12) {
+ /* RXOR 1-2 */
+ return 2;
+ } else {
+ /* RXOR 1-2-3/1-2-4/1-2-5 */
+ return 3;
+ }
+ }
+ return 1;
+ default:
+ printk(KERN_ERR "%s: unknown OPC 0x%02x\n",
+ __func__, dma_hw_desc->opc);
+ BUG();
+ }
+ case PPC440SPE_XOR_ID:
+ /* up to 16 sources */
+ xor_hw_desc = desc->hw_desc;
+ return xor_hw_desc->cbc & XOR_CDCR_OAC_MSK;
+ default:
+ BUG();
+ }
+ return 0;
+}
+
+/**
+ * ppc440spe_desc_get_dst_num - get the number of destination addresses in
+ * this descriptor
+ */
+static u32 ppc440spe_desc_get_dst_num(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan)
+{
+ struct dma_cdb *dma_hw_desc;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ /* May be 1 or 2 destinations */
+ dma_hw_desc = desc->hw_desc;
+ switch (dma_hw_desc->opc) {
+ case DMA_CDB_OPC_NO_OP:
+ case DMA_CDB_OPC_DCHECK128:
+ return 0;
+ case DMA_CDB_OPC_MV_SG1_SG2:
+ case DMA_CDB_OPC_DFILL128:
+ return 1;
+ case DMA_CDB_OPC_MULTICAST:
+ if (desc->dst_cnt == 2)
+ return 2;
+ else
+ return 1;
+ default:
+ printk(KERN_ERR "%s: unknown OPC 0x%02x\n",
+ __func__, dma_hw_desc->opc);
+ BUG();
+ }
+ case PPC440SPE_XOR_ID:
+ /* Always only 1 destination */
+ return 1;
+ default:
+ BUG();
+ }
+ return 0;
+}
+
+/**
+ * ppc440spe_desc_get_link - get the address of the descriptor that
+ * follows this one
+ */
+static inline u32 ppc440spe_desc_get_link(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan)
+{
+ if (!desc->hw_next)
+ return 0;
+
+ return desc->hw_next->phys;
+}
+
+/**
+ * ppc440spe_desc_is_aligned - check alignment
+ */
+static inline int ppc440spe_desc_is_aligned(
+ struct ppc440spe_adma_desc_slot *desc, int num_slots)
+{
+ return (desc->idx & (num_slots - 1)) ? 0 : 1;
+}
+
+/**
+ * ppc440spe_chan_xor_slot_count - get the number of slots necessary for
+ * XOR operation
+ */
+static int ppc440spe_chan_xor_slot_count(size_t len, int src_cnt,
+ int *slots_per_op)
+{
+ int slot_cnt;
+
+ /* each XOR descriptor provides up to 16 source operands */
+ slot_cnt = *slots_per_op = (src_cnt + XOR_MAX_OPS - 1)/XOR_MAX_OPS;
+
+ if (likely(len <= PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT))
+ return slot_cnt;
+
+ printk(KERN_ERR "%s: len %d > max %d !!\n",
+ __func__, len, PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT);
+ BUG();
+ return slot_cnt;
+}
+
+/**
+ * ppc440spe_dma2_pq_slot_count - get the number of slots necessary for
+ * DMA2 PQ operation
+ */
+static int ppc440spe_dma2_pq_slot_count(dma_addr_t *srcs,
+ int src_cnt, size_t len)
+{
+ signed long long order = 0;
+ int state = 0;
+ int addr_count = 0;
+ int i;
+ for (i = 1; i < src_cnt; i++) {
+ dma_addr_t cur_addr = srcs[i];
+ dma_addr_t old_addr = srcs[i-1];
+ switch (state) {
+ case 0:
+ if (cur_addr == old_addr + len) {
+ /* direct RXOR */
+ order = 1;
+ state = 1;
+ if (i == src_cnt-1)
+ addr_count++;
+ } else if (old_addr == cur_addr + len) {
+ /* reverse RXOR */
+ order = -1;
+ state = 1;
+ if (i == src_cnt-1)
+ addr_count++;
+ } else {
+ state = 3;
+ }
+ break;
+ case 1:
+ if (i == src_cnt-2 || (order == -1
+ && cur_addr != old_addr - len)) {
+ order = 0;
+ state = 0;
+ addr_count++;
+ } else if (cur_addr == old_addr + len*order) {
+ state = 2;
+ if (i == src_cnt-1)
+ addr_count++;
+ } else if (cur_addr == old_addr + 2*len) {
+ state = 2;
+ if (i == src_cnt-1)
+ addr_count++;
+ } else if (cur_addr == old_addr + 3*len) {
+ state = 2;
+ if (i == src_cnt-1)
+ addr_count++;
+ } else {
+ order = 0;
+ state = 0;
+ addr_count++;
+ }
+ break;
+ case 2:
+ order = 0;
+ state = 0;
+ addr_count++;
+ break;
+ }
+ if (state == 3)
+ break;
+ }
+ if (src_cnt <= 1 || (state != 1 && state != 2)) {
+ pr_err("%s: src_cnt=%d, state=%d, addr_count=%d, order=%lld\n",
+ __func__, src_cnt, state, addr_count, order);
+ for (i = 0; i < src_cnt; i++)
+ pr_err("\t[%d] 0x%llx \n", i, srcs[i]);
+ BUG();
+ }
+
+ return (addr_count + XOR_MAX_OPS - 1) / XOR_MAX_OPS;
+}
+
+
+/******************************************************************************
+ * ADMA channel low-level routines
+ ******************************************************************************/
+
+static u32
+ppc440spe_chan_get_current_descriptor(struct ppc440spe_adma_chan *chan);
+static void ppc440spe_chan_append(struct ppc440spe_adma_chan *chan);
+
+/**
+ * ppc440spe_adma_device_clear_eot_status - interrupt ack to XOR or DMA engine
+ */
+static void ppc440spe_adma_device_clear_eot_status(
+ struct ppc440spe_adma_chan *chan)
+{
+ struct dma_regs *dma_reg;
+ struct xor_regs *xor_reg;
+ u8 *p = chan->device->dma_desc_pool_virt;
+ struct dma_cdb *cdb;
+ u32 rv, i;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ /* read FIFO to ack */
+ dma_reg = chan->device->dma_reg;
+ while ((rv = ioread32(&dma_reg->csfpl))) {
+ i = rv & DMA_CDB_ADDR_MSK;
+ cdb = (struct dma_cdb *)&p[i -
+ (u32)chan->device->dma_desc_pool];
+
+ /* Clear opcode to ack. This is necessary for
+ * ZeroSum operations only
+ */
+ cdb->opc = 0;
+
+ if (test_bit(PPC440SPE_RXOR_RUN,
+ &ppc440spe_rxor_state)) {
+ /* probably this is a completed RXOR op,
+ * get pointer to CDB using the fact that
+ * physical and virtual addresses of CDB
+ * in pools have the same offsets
+ */
+ if (le32_to_cpu(cdb->sg1u) &
+ DMA_CUED_XOR_BASE) {
+ /* this is a RXOR */
+ clear_bit(PPC440SPE_RXOR_RUN,
+ &ppc440spe_rxor_state);
+ }
+ }
+
+ if (rv & DMA_CDB_STATUS_MSK) {
+ /* ZeroSum check failed
+ */
+ struct ppc440spe_adma_desc_slot *iter;
+ dma_addr_t phys = rv & ~DMA_CDB_MSK;
+
+ /*
+ * Update the status of corresponding
+ * descriptor.
+ */
+ list_for_each_entry(iter, &chan->chain,
+ chain_node) {
+ if (iter->phys == phys)
+ break;
+ }
+ /*
+ * if cannot find the corresponding
+ * slot it's a bug
+ */
+ BUG_ON(&iter->chain_node == &chan->chain);
+
+ if (iter->xor_check_result) {
+ if (test_bit(PPC440SPE_DESC_PCHECK,
+ &iter->flags)) {
+ *iter->xor_check_result |=
+ SUM_CHECK_P_RESULT;
+ } else
+ if (test_bit(PPC440SPE_DESC_QCHECK,
+ &iter->flags)) {
+ *iter->xor_check_result |=
+ SUM_CHECK_Q_RESULT;
+ } else
+ BUG();
+ }
+ }
+ }
+
+ rv = ioread32(&dma_reg->dsts);
+ if (rv) {
+ pr_err("DMA%d err status: 0x%x\n",
+ chan->device->id, rv);
+ /* write back to clear */
+ iowrite32(rv, &dma_reg->dsts);
+ }
+ break;
+ case PPC440SPE_XOR_ID:
+ /* reset status bits to ack */
+ xor_reg = chan->device->xor_reg;
+ rv = ioread32be(&xor_reg->sr);
+ iowrite32be(rv, &xor_reg->sr);
+
+ if (rv & (XOR_IE_ICBIE_BIT|XOR_IE_ICIE_BIT|XOR_IE_RPTIE_BIT)) {
+ if (rv & XOR_IE_RPTIE_BIT) {
+ /* Read PLB Timeout Error.
+ * Try to resubmit the CB
+ */
+ u32 val = ioread32be(&xor_reg->ccbalr);
+
+ iowrite32be(val, &xor_reg->cblalr);
+
+ val = ioread32be(&xor_reg->crsr);
+ iowrite32be(val | XOR_CRSR_XAE_BIT,
+ &xor_reg->crsr);
+ } else
+ pr_err("XOR ERR 0x%x status\n", rv);
+ break;
+ }
+
+ /* if the XORcore is idle, but there are unprocessed CBs
+ * then refetch the s/w chain here
+ */
+ if (!(ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT) &&
+ do_xor_refetch)
+ ppc440spe_chan_append(chan);
+ break;
+ }
+}
+
+/**
+ * ppc440spe_chan_is_busy - get the channel status
+ */
+static int ppc440spe_chan_is_busy(struct ppc440spe_adma_chan *chan)
+{
+ struct dma_regs *dma_reg;
+ struct xor_regs *xor_reg;
+ int busy = 0;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_reg = chan->device->dma_reg;
+ /* if command FIFO's head and tail pointers are equal and
+ * status tail is the same as command, then channel is free
+ */
+ if (ioread16(&dma_reg->cpfhp) != ioread16(&dma_reg->cpftp) ||
+ ioread16(&dma_reg->cpftp) != ioread16(&dma_reg->csftp))
+ busy = 1;
+ break;
+ case PPC440SPE_XOR_ID:
+ /* use the special status bit for the XORcore
+ */
+ xor_reg = chan->device->xor_reg;
+ busy = (ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT) ? 1 : 0;
+ break;
+ }
+
+ return busy;
+}
+
+/**
+ * ppc440spe_chan_set_first_xor_descriptor - init XORcore chain
+ */
+static void ppc440spe_chan_set_first_xor_descriptor(
+ struct ppc440spe_adma_chan *chan,
+ struct ppc440spe_adma_desc_slot *next_desc)
+{
+ struct xor_regs *xor_reg = chan->device->xor_reg;
+
+ if (ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT)
+ printk(KERN_INFO "%s: Warn: XORcore is running "
+ "when try to set the first CDB!\n",
+ __func__);
+
+ xor_last_submit = xor_last_linked = next_desc;
+
+ iowrite32be(XOR_CRSR_64BA_BIT, &xor_reg->crsr);
+
+ iowrite32be(next_desc->phys, &xor_reg->cblalr);
+ iowrite32be(0, &xor_reg->cblahr);
+ iowrite32be(ioread32be(&xor_reg->cbcr) | XOR_CBCR_LNK_BIT,
+ &xor_reg->cbcr);
+
+ chan->hw_chain_inited = 1;
+}
+
+/**
+ * ppc440spe_dma_put_desc - put DMA0,1 descriptor to FIFO.
+ * called with irqs disabled
+ */
+static void ppc440spe_dma_put_desc(struct ppc440spe_adma_chan *chan,
+ struct ppc440spe_adma_desc_slot *desc)
+{
+ u32 pcdb;
+ struct dma_regs *dma_reg = chan->device->dma_reg;
+
+ pcdb = desc->phys;
+ if (!test_bit(PPC440SPE_DESC_INT, &desc->flags))
+ pcdb |= DMA_CDB_NO_INT;
+
+ chan_last_sub[chan->device->id] = desc;
+
+ ADMA_LL_DBG(print_cb(chan, desc->hw_desc));
+
+ iowrite32(pcdb, &dma_reg->cpfpl);
+}
+
+/**
+ * ppc440spe_chan_append - update the h/w chain in the channel
+ */
+static void ppc440spe_chan_append(struct ppc440spe_adma_chan *chan)
+{
+ struct xor_regs *xor_reg;
+ struct ppc440spe_adma_desc_slot *iter;
+ struct xor_cb *xcb;
+ u32 cur_desc;
+ unsigned long flags;
+
+ local_irq_save(flags);
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ cur_desc = ppc440spe_chan_get_current_descriptor(chan);
+
+ if (likely(cur_desc)) {
+ iter = chan_last_sub[chan->device->id];
+ BUG_ON(!iter);
+ } else {
+ /* first peer */
+ iter = chan_first_cdb[chan->device->id];
+ BUG_ON(!iter);
+ ppc440spe_dma_put_desc(chan, iter);
+ chan->hw_chain_inited = 1;
+ }
+
+ /* is there something new to append */
+ if (!iter->hw_next)
+ break;
+
+ /* flush descriptors from the s/w queue to fifo */
+ list_for_each_entry_continue(iter, &chan->chain, chain_node) {
+ ppc440spe_dma_put_desc(chan, iter);
+ if (!iter->hw_next)
+ break;
+ }
+ break;
+ case PPC440SPE_XOR_ID:
+ /* update h/w links and refetch */
+ if (!xor_last_submit->hw_next)
+ break;
+
+ xor_reg = chan->device->xor_reg;
+ /* the last linked CDB has to generate an interrupt
+ * that we'd be able to append the next lists to h/w
+ * regardless of the XOR engine state at the moment of
+ * appending of these next lists
+ */
+ xcb = xor_last_linked->hw_desc;
+ xcb->cbc |= XOR_CBCR_CBCE_BIT;
+
+ if (!(ioread32be(&xor_reg->sr) & XOR_SR_XCP_BIT)) {
+ /* XORcore is idle. Refetch now */
+ do_xor_refetch = 0;
+ ppc440spe_xor_set_link(xor_last_submit,
+ xor_last_submit->hw_next);
+
+ ADMA_LL_DBG(print_cb_list(chan,
+ xor_last_submit->hw_next));
+
+ xor_last_submit = xor_last_linked;
+ iowrite32be(ioread32be(&xor_reg->crsr) |
+ XOR_CRSR_RCBE_BIT | XOR_CRSR_64BA_BIT,
+ &xor_reg->crsr);
+ } else {
+ /* XORcore is running. Refetch later in the handler */
+ do_xor_refetch = 1;
+ }
+
+ break;
+ }
+
+ local_irq_restore(flags);
+}
+
+/**
+ * ppc440spe_chan_get_current_descriptor - get the currently executed descriptor
+ */
+static u32
+ppc440spe_chan_get_current_descriptor(struct ppc440spe_adma_chan *chan)
+{
+ struct dma_regs *dma_reg;
+ struct xor_regs *xor_reg;
+
+ if (unlikely(!chan->hw_chain_inited))
+ /* h/w descriptor chain is not initialized yet */
+ return 0;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_reg = chan->device->dma_reg;
+ return ioread32(&dma_reg->acpl) & (~DMA_CDB_MSK);
+ case PPC440SPE_XOR_ID:
+ xor_reg = chan->device->xor_reg;
+ return ioread32be(&xor_reg->ccbalr);
+ }
+ return 0;
+}
+
+/**
+ * ppc440spe_chan_run - enable the channel
+ */
+static void ppc440spe_chan_run(struct ppc440spe_adma_chan *chan)
+{
+ struct xor_regs *xor_reg;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ /* DMAs are always enabled, do nothing */
+ break;
+ case PPC440SPE_XOR_ID:
+ /* drain write buffer */
+ xor_reg = chan->device->xor_reg;
+
+ /* fetch descriptor pointed to in <link> */
+ iowrite32be(XOR_CRSR_64BA_BIT | XOR_CRSR_XAE_BIT,
+ &xor_reg->crsr);
+ break;
+ }
+}
+
+/******************************************************************************
+ * ADMA device level
+ ******************************************************************************/
+
+static void ppc440spe_chan_start_null_xor(struct ppc440spe_adma_chan *chan);
+static int ppc440spe_adma_alloc_chan_resources(struct dma_chan *chan);
+
+static dma_cookie_t
+ppc440spe_adma_tx_submit(struct dma_async_tx_descriptor *tx);
+
+static void ppc440spe_adma_set_dest(struct ppc440spe_adma_desc_slot *tx,
+ dma_addr_t addr, int index);
+static void
+ppc440spe_adma_memcpy_xor_set_src(struct ppc440spe_adma_desc_slot *tx,
+ dma_addr_t addr, int index);
+
+static void
+ppc440spe_adma_pq_set_dest(struct ppc440spe_adma_desc_slot *tx,
+ dma_addr_t *paddr, unsigned long flags);
+static void
+ppc440spe_adma_pq_set_src(struct ppc440spe_adma_desc_slot *tx,
+ dma_addr_t addr, int index);
+static void
+ppc440spe_adma_pq_set_src_mult(struct ppc440spe_adma_desc_slot *tx,
+ unsigned char mult, int index, int dst_pos);
+static void
+ppc440spe_adma_pqzero_sum_set_dest(struct ppc440spe_adma_desc_slot *tx,
+ dma_addr_t paddr, dma_addr_t qaddr);
+
+static struct page *ppc440spe_rxor_srcs[32];
+
+/**
+ * ppc440spe_can_rxor - check if the operands may be processed with RXOR
+ */
+static int ppc440spe_can_rxor(struct page **srcs, int src_cnt, size_t len)
+{
+ int i, order = 0, state = 0;
+ int idx = 0;
+
+ if (unlikely(!(src_cnt > 1)))
+ return 0;
+
+ BUG_ON(src_cnt > ARRAY_SIZE(ppc440spe_rxor_srcs));
+
+ /* Skip holes in the source list before checking */
+ for (i = 0; i < src_cnt; i++) {
+ if (!srcs[i])
+ continue;
+ ppc440spe_rxor_srcs[idx++] = srcs[i];
+ }
+ src_cnt = idx;
+
+ for (i = 1; i < src_cnt; i++) {
+ char *cur_addr = page_address(ppc440spe_rxor_srcs[i]);
+ char *old_addr = page_address(ppc440spe_rxor_srcs[i - 1]);
+
+ switch (state) {
+ case 0:
+ if (cur_addr == old_addr + len) {
+ /* direct RXOR */
+ order = 1;
+ state = 1;
+ } else if (old_addr == cur_addr + len) {
+ /* reverse RXOR */
+ order = -1;
+ state = 1;
+ } else
+ goto out;
+ break;
+ case 1:
+ if ((i == src_cnt - 2) ||
+ (order == -1 && cur_addr != old_addr - len)) {
+ order = 0;
+ state = 0;
+ } else if ((cur_addr == old_addr + len * order) ||
+ (cur_addr == old_addr + 2 * len) ||
+ (cur_addr == old_addr + 3 * len)) {
+ state = 2;
+ } else {
+ order = 0;
+ state = 0;
+ }
+ break;
+ case 2:
+ order = 0;
+ state = 0;
+ break;
+ }
+ }
+
+out:
+ if (state == 1 || state == 2)
+ return 1;
+
+ return 0;
+}
+
+/**
+ * ppc440spe_adma_device_estimate - estimate the efficiency of processing
+ * the operation given on this channel. It's assumed that 'chan' is
+ * capable to process 'cap' type of operation.
+ * @chan: channel to use
+ * @cap: type of transaction
+ * @dst_lst: array of destination pointers
+ * @dst_cnt: number of destination operands
+ * @src_lst: array of source pointers
+ * @src_cnt: number of source operands
+ * @src_sz: size of each source operand
+ */
+static int ppc440spe_adma_estimate(struct dma_chan *chan,
+ enum dma_transaction_type cap, struct page **dst_lst, int dst_cnt,
+ struct page **src_lst, int src_cnt, size_t src_sz)
+{
+ int ef = 1;
+
+ if (cap == DMA_PQ || cap == DMA_PQ_VAL) {
+ /* If RAID-6 capabilities were not activated don't try
+ * to use them
+ */
+ if (unlikely(!ppc440spe_r6_enabled))
+ return -1;
+ }
+ /* In the current implementation of ppc440spe ADMA driver it
+ * makes sense to pick out only pq case, because it may be
+ * processed:
+ * (1) either using Biskup method on DMA2;
+ * (2) or on DMA0/1.
+ * Thus we give a favour to (1) if the sources are suitable;
+ * else let it be processed on one of the DMA0/1 engines.
+ * In the sum_product case where destination is also the
+ * source process it on DMA0/1 only.
+ */
+ if (cap == DMA_PQ && chan->chan_id == PPC440SPE_XOR_ID) {
+
+ if (dst_cnt == 1 && src_cnt == 2 && dst_lst[0] == src_lst[1])
+ ef = 0; /* sum_product case, process on DMA0/1 */
+ else if (ppc440spe_can_rxor(src_lst, src_cnt, src_sz))
+ ef = 3; /* override (DMA0/1 + idle) */
+ else
+ ef = 0; /* can't process on DMA2 if !rxor */
+ }
+
+ /* channel idleness increases the priority */
+ if (likely(ef) &&
+ !ppc440spe_chan_is_busy(to_ppc440spe_adma_chan(chan)))
+ ef++;
+
+ return ef;
+}
+
+struct dma_chan *
+ppc440spe_async_tx_find_best_channel(enum dma_transaction_type cap,
+ struct page **dst_lst, int dst_cnt, struct page **src_lst,
+ int src_cnt, size_t src_sz)
+{
+ struct dma_chan *best_chan = NULL;
+ struct ppc_dma_chan_ref *ref;
+ int best_rank = -1;
+
+ if (unlikely(!src_sz))
+ return NULL;
+ if (src_sz > PAGE_SIZE) {
+ /*
+ * should a user of the api ever pass > PAGE_SIZE requests
+ * we sort out cases where temporary page-sized buffers
+ * are used.
+ */
+ switch (cap) {
+ case DMA_PQ:
+ if (src_cnt == 1 && dst_lst[1] == src_lst[0])
+ return NULL;
+ if (src_cnt == 2 && dst_lst[1] == src_lst[1])
+ return NULL;
+ break;
+ case DMA_PQ_VAL:
+ case DMA_XOR_VAL:
+ return NULL;
+ default:
+ break;
+ }
+ }
+
+ list_for_each_entry(ref, &ppc440spe_adma_chan_list, node) {
+ if (dma_has_cap(cap, ref->chan->device->cap_mask)) {
+ int rank;
+
+ rank = ppc440spe_adma_estimate(ref->chan, cap, dst_lst,
+ dst_cnt, src_lst, src_cnt, src_sz);
+ if (rank > best_rank) {
+ best_rank = rank;
+ best_chan = ref->chan;
+ }
+ }
+ }
+
+ return best_chan;
+}
+EXPORT_SYMBOL_GPL(ppc440spe_async_tx_find_best_channel);
+
+/**
+ * ppc440spe_get_group_entry - get group entry with index idx
+ * @tdesc: is the last allocated slot in the group.
+ */
+static struct ppc440spe_adma_desc_slot *
+ppc440spe_get_group_entry(struct ppc440spe_adma_desc_slot *tdesc, u32 entry_idx)
+{
+ struct ppc440spe_adma_desc_slot *iter = tdesc->group_head;
+ int i = 0;
+
+ if (entry_idx < 0 || entry_idx >= (tdesc->src_cnt + tdesc->dst_cnt)) {
+ printk("%s: entry_idx %d, src_cnt %d, dst_cnt %d\n",
+ __func__, entry_idx, tdesc->src_cnt, tdesc->dst_cnt);
+ BUG();
+ }
+
+ list_for_each_entry(iter, &tdesc->group_list, chain_node) {
+ if (i++ == entry_idx)
+ break;
+ }
+ return iter;
+}
+
+/**
+ * ppc440spe_adma_free_slots - flags descriptor slots for reuse
+ * @slot: Slot to free
+ * Caller must hold &ppc440spe_chan->lock while calling this function
+ */
+static void ppc440spe_adma_free_slots(struct ppc440spe_adma_desc_slot *slot,
+ struct ppc440spe_adma_chan *chan)
+{
+ int stride = slot->slots_per_op;
+
+ while (stride--) {
+ slot->slots_per_op = 0;
+ slot = list_entry(slot->slot_node.next,
+ struct ppc440spe_adma_desc_slot,
+ slot_node);
+ }
+}
+
+static void ppc440spe_adma_unmap(struct ppc440spe_adma_chan *chan,
+ struct ppc440spe_adma_desc_slot *desc)
+{
+ u32 src_cnt, dst_cnt;
+ dma_addr_t addr;
+
+ /*
+ * get the number of sources & destination
+ * included in this descriptor and unmap
+ * them all
+ */
+ src_cnt = ppc440spe_desc_get_src_num(desc, chan);
+ dst_cnt = ppc440spe_desc_get_dst_num(desc, chan);
+
+ /* unmap destinations */
+ if (!(desc->async_tx.flags & DMA_COMPL_SKIP_DEST_UNMAP)) {
+ while (dst_cnt--) {
+ addr = ppc440spe_desc_get_dest_addr(
+ desc, chan, dst_cnt);
+ dma_unmap_page(chan->device->dev,
+ addr, desc->unmap_len,
+ DMA_FROM_DEVICE);
+ }
+ }
+
+ /* unmap sources */
+ if (!(desc->async_tx.flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
+ while (src_cnt--) {
+ addr = ppc440spe_desc_get_src_addr(
+ desc, chan, src_cnt);
+ dma_unmap_page(chan->device->dev,
+ addr, desc->unmap_len,
+ DMA_TO_DEVICE);
+ }
+ }
+}
+
+/**
+ * ppc440spe_adma_run_tx_complete_actions - call functions to be called
+ * upon completion
+ */
+static dma_cookie_t ppc440spe_adma_run_tx_complete_actions(
+ struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan,
+ dma_cookie_t cookie)
+{
+ int i;
+
+ BUG_ON(desc->async_tx.cookie < 0);
+ if (desc->async_tx.cookie > 0) {
+ cookie = desc->async_tx.cookie;
+ desc->async_tx.cookie = 0;
+
+ /* call the callback (must not sleep or submit new
+ * operations to this channel)
+ */
+ if (desc->async_tx.callback)
+ desc->async_tx.callback(
+ desc->async_tx.callback_param);
+
+ /* unmap dma addresses
+ * (unmap_single vs unmap_page?)
+ *
+ * actually, ppc's dma_unmap_page() functions are empty, so
+ * the following code is just for the sake of completeness
+ */
+ if (chan && chan->needs_unmap && desc->group_head &&
+ desc->unmap_len) {
+ struct ppc440spe_adma_desc_slot *unmap =
+ desc->group_head;
+ /* assume 1 slot per op always */
+ u32 slot_count = unmap->slot_cnt;
+
+ /* Run through the group list and unmap addresses */
+ for (i = 0; i < slot_count; i++) {
+ BUG_ON(!unmap);
+ ppc440spe_adma_unmap(chan, unmap);
+ unmap = unmap->hw_next;
+ }
+ }
+ }
+
+ /* run dependent operations */
+ dma_run_dependencies(&desc->async_tx);
+
+ return cookie;
+}
+
+/**
+ * ppc440spe_adma_clean_slot - clean up CDB slot (if ack is set)
+ */
+static int ppc440spe_adma_clean_slot(struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_adma_chan *chan)
+{
+ /* the client is allowed to attach dependent operations
+ * until 'ack' is set
+ */
+ if (!async_tx_test_ack(&desc->async_tx))
+ return 0;
+
+ /* leave the last descriptor in the chain
+ * so we can append to it
+ */
+ if (list_is_last(&desc->chain_node, &chan->chain) ||
+ desc->phys == ppc440spe_chan_get_current_descriptor(chan))
+ return 1;
+
+ if (chan->device->id != PPC440SPE_XOR_ID) {
+ /* our DMA interrupt handler clears opc field of
+ * each processed descriptor. For all types of
+ * operations except for ZeroSum we do not actually
+ * need ack from the interrupt handler. ZeroSum is a
+ * special case since the result of this operation
+ * is available from the handler only, so if we see
+ * such type of descriptor (which is unprocessed yet)
+ * then leave it in chain.
+ */
+ struct dma_cdb *cdb = desc->hw_desc;
+ if (cdb->opc == DMA_CDB_OPC_DCHECK128)
+ return 1;
+ }
+
+ dev_dbg(chan->device->common.dev, "\tfree slot %llx: %d stride: %d\n",
+ desc->phys, desc->idx, desc->slots_per_op);
+
+ list_del(&desc->chain_node);
+ ppc440spe_adma_free_slots(desc, chan);
+ return 0;
+}
+
+/**
+ * __ppc440spe_adma_slot_cleanup - this is the common clean-up routine
+ * which runs through the channel CDBs list until reach the descriptor
+ * currently processed. When routine determines that all CDBs of group
+ * are completed then corresponding callbacks (if any) are called and slots
+ * are freed.
+ */
+static void __ppc440spe_adma_slot_cleanup(struct ppc440spe_adma_chan *chan)
+{
+ struct ppc440spe_adma_desc_slot *iter, *_iter, *group_start = NULL;
+ dma_cookie_t cookie = 0;
+ u32 current_desc = ppc440spe_chan_get_current_descriptor(chan);
+ int busy = ppc440spe_chan_is_busy(chan);
+ int seen_current = 0, slot_cnt = 0, slots_per_op = 0;
+
+ dev_dbg(chan->device->common.dev, "ppc440spe adma%d: %s\n",
+ chan->device->id, __func__);
+
+ if (!current_desc) {
+ /* There were no transactions yet, so
+ * nothing to clean
+ */
+ return;
+ }
+
+ /* free completed slots from the chain starting with
+ * the oldest descriptor
+ */
+ list_for_each_entry_safe(iter, _iter, &chan->chain,
+ chain_node) {
+ dev_dbg(chan->device->common.dev, "\tcookie: %d slot: %d "
+ "busy: %d this_desc: %#llx next_desc: %#x "
+ "cur: %#x ack: %d\n",
+ iter->async_tx.cookie, iter->idx, busy, iter->phys,
+ ppc440spe_desc_get_link(iter, chan), current_desc,
+ async_tx_test_ack(&iter->async_tx));
+ prefetch(_iter);
+ prefetch(&_iter->async_tx);
+
+ /* do not advance past the current descriptor loaded into the
+ * hardware channel,subsequent descriptors are either in process
+ * or have not been submitted
+ */
+ if (seen_current)
+ break;
+
+ /* stop the search if we reach the current descriptor and the
+ * channel is busy, or if it appears that the current descriptor
+ * needs to be re-read (i.e. has been appended to)
+ */
+ if (iter->phys == current_desc) {
+ BUG_ON(seen_current++);
+ if (busy || ppc440spe_desc_get_link(iter, chan)) {
+ /* not all descriptors of the group have
+ * been completed; exit.
+ */
+ break;
+ }
+ }
+
+ /* detect the start of a group transaction */
+ if (!slot_cnt && !slots_per_op) {
+ slot_cnt = iter->slot_cnt;
+ slots_per_op = iter->slots_per_op;
+ if (slot_cnt <= slots_per_op) {
+ slot_cnt = 0;
+ slots_per_op = 0;
+ }
+ }
+
+ if (slot_cnt) {
+ if (!group_start)
+ group_start = iter;
+ slot_cnt -= slots_per_op;
+ }
+
+ /* all the members of a group are complete */
+ if (slots_per_op != 0 && slot_cnt == 0) {
+ struct ppc440spe_adma_desc_slot *grp_iter, *_grp_iter;
+ int end_of_chain = 0;
+
+ /* clean up the group */
+ slot_cnt = group_start->slot_cnt;
+ grp_iter = group_start;
+ list_for_each_entry_safe_from(grp_iter, _grp_iter,
+ &chan->chain, chain_node) {
+
+ cookie = ppc440spe_adma_run_tx_complete_actions(
+ grp_iter, chan, cookie);
+
+ slot_cnt -= slots_per_op;
+ end_of_chain = ppc440spe_adma_clean_slot(
+ grp_iter, chan);
+ if (end_of_chain && slot_cnt) {
+ /* Should wait for ZeroSum completion */
+ if (cookie > 0)
+ chan->completed_cookie = cookie;
+ return;
+ }
+
+ if (slot_cnt == 0 || end_of_chain)
+ break;
+ }
+
+ /* the group should be complete at this point */
+ BUG_ON(slot_cnt);
+
+ slots_per_op = 0;
+ group_start = NULL;
+ if (end_of_chain)
+ break;
+ else
+ continue;
+ } else if (slots_per_op) /* wait for group completion */
+ continue;
+
+ cookie = ppc440spe_adma_run_tx_complete_actions(iter, chan,
+ cookie);
+
+ if (ppc440spe_adma_clean_slot(iter, chan))
+ break;
+ }
+
+ BUG_ON(!seen_current);
+
+ if (cookie > 0) {
+ chan->completed_cookie = cookie;
+ pr_debug("\tcompleted cookie %d\n", cookie);
+ }
+
+}
+
+/**
+ * ppc440spe_adma_tasklet - clean up watch-dog initiator
+ */
+static void ppc440spe_adma_tasklet(unsigned long data)
+{
+ struct ppc440spe_adma_chan *chan = (struct ppc440spe_adma_chan *) data;
+
+ spin_lock_nested(&chan->lock, SINGLE_DEPTH_NESTING);
+ __ppc440spe_adma_slot_cleanup(chan);
+ spin_unlock(&chan->lock);
+}
+
+/**
+ * ppc440spe_adma_slot_cleanup - clean up scheduled initiator
+ */
+static void ppc440spe_adma_slot_cleanup(struct ppc440spe_adma_chan *chan)
+{
+ spin_lock_bh(&chan->lock);
+ __ppc440spe_adma_slot_cleanup(chan);
+ spin_unlock_bh(&chan->lock);
+}
+
+/**
+ * ppc440spe_adma_alloc_slots - allocate free slots (if any)
+ */
+static struct ppc440spe_adma_desc_slot *ppc440spe_adma_alloc_slots(
+ struct ppc440spe_adma_chan *chan, int num_slots,
+ int slots_per_op)
+{
+ struct ppc440spe_adma_desc_slot *iter = NULL, *_iter;
+ struct ppc440spe_adma_desc_slot *alloc_start = NULL;
+ struct list_head chain = LIST_HEAD_INIT(chain);
+ int slots_found, retry = 0;
+
+
+ BUG_ON(!num_slots || !slots_per_op);
+ /* start search from the last allocated descrtiptor
+ * if a contiguous allocation can not be found start searching
+ * from the beginning of the list
+ */
+retry:
+ slots_found = 0;
+ if (retry == 0)
+ iter = chan->last_used;
+ else
+ iter = list_entry(&chan->all_slots,
+ struct ppc440spe_adma_desc_slot,
+ slot_node);
+ list_for_each_entry_safe_continue(iter, _iter, &chan->all_slots,
+ slot_node) {
+ prefetch(_iter);
+ prefetch(&_iter->async_tx);
+ if (iter->slots_per_op) {
+ slots_found = 0;
+ continue;
+ }
+
+ /* start the allocation if the slot is correctly aligned */
+ if (!slots_found++)
+ alloc_start = iter;
+
+ if (slots_found == num_slots) {
+ struct ppc440spe_adma_desc_slot *alloc_tail = NULL;
+ struct ppc440spe_adma_desc_slot *last_used = NULL;
+
+ iter = alloc_start;
+ while (num_slots) {
+ int i;
+ /* pre-ack all but the last descriptor */
+ if (num_slots != slots_per_op)
+ async_tx_ack(&iter->async_tx);
+
+ list_add_tail(&iter->chain_node, &chain);
+ alloc_tail = iter;
+ iter->async_tx.cookie = 0;
+ iter->hw_next = NULL;
+ iter->flags = 0;
+ iter->slot_cnt = num_slots;
+ iter->xor_check_result = NULL;
+ for (i = 0; i < slots_per_op; i++) {
+ iter->slots_per_op = slots_per_op - i;
+ last_used = iter;
+ iter = list_entry(iter->slot_node.next,
+ struct ppc440spe_adma_desc_slot,
+ slot_node);
+ }
+ num_slots -= slots_per_op;
+ }
+ alloc_tail->group_head = alloc_start;
+ alloc_tail->async_tx.cookie = -EBUSY;
+ list_splice(&chain, &alloc_tail->group_list);
+ chan->last_used = last_used;
+ return alloc_tail;
+ }
+ }
+ if (!retry++)
+ goto retry;
+
+ /* try to free some slots if the allocation fails */
+ tasklet_schedule(&chan->irq_tasklet);
+ return NULL;
+}
+
+/**
+ * ppc440spe_adma_alloc_chan_resources - allocate pools for CDB slots
+ */
+static int ppc440spe_adma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *slot = NULL;
+ char *hw_desc;
+ int i, db_sz;
+ int init;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+ init = ppc440spe_chan->slots_allocated ? 0 : 1;
+ chan->chan_id = ppc440spe_chan->device->id;
+
+ /* Allocate descriptor slots */
+ i = ppc440spe_chan->slots_allocated;
+ if (ppc440spe_chan->device->id != PPC440SPE_XOR_ID)
+ db_sz = sizeof(struct dma_cdb);
+ else
+ db_sz = sizeof(struct xor_cb);
+
+ for (; i < (ppc440spe_chan->device->pool_size / db_sz); i++) {
+ slot = kzalloc(sizeof(struct ppc440spe_adma_desc_slot),
+ GFP_KERNEL);
+ if (!slot) {
+ printk(KERN_INFO "SPE ADMA Channel only initialized"
+ " %d descriptor slots", i--);
+ break;
+ }
+
+ hw_desc = (char *) ppc440spe_chan->device->dma_desc_pool_virt;
+ slot->hw_desc = (void *) &hw_desc[i * db_sz];
+ dma_async_tx_descriptor_init(&slot->async_tx, chan);
+ slot->async_tx.tx_submit = ppc440spe_adma_tx_submit;
+ INIT_LIST_HEAD(&slot->chain_node);
+ INIT_LIST_HEAD(&slot->slot_node);
+ INIT_LIST_HEAD(&slot->group_list);
+ slot->phys = ppc440spe_chan->device->dma_desc_pool + i * db_sz;
+ slot->idx = i;
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+ ppc440spe_chan->slots_allocated++;
+ list_add_tail(&slot->slot_node, &ppc440spe_chan->all_slots);
+ spin_unlock_bh(&ppc440spe_chan->lock);
+ }
+
+ if (i && !ppc440spe_chan->last_used) {
+ ppc440spe_chan->last_used =
+ list_entry(ppc440spe_chan->all_slots.next,
+ struct ppc440spe_adma_desc_slot,
+ slot_node);
+ }
+
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d: allocated %d descriptor slots\n",
+ ppc440spe_chan->device->id, i);
+
+ /* initialize the channel and the chain with a null operation */
+ if (init) {
+ switch (ppc440spe_chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ ppc440spe_chan->hw_chain_inited = 0;
+ /* Use WXOR for self-testing */
+ if (!ppc440spe_r6_tchan)
+ ppc440spe_r6_tchan = ppc440spe_chan;
+ break;
+ case PPC440SPE_XOR_ID:
+ ppc440spe_chan_start_null_xor(ppc440spe_chan);
+ break;
+ default:
+ BUG();
+ }
+ ppc440spe_chan->needs_unmap = 1;
+ }
+
+ return (i > 0) ? i : -ENOMEM;
+}
+
+/**
+ * ppc440spe_desc_assign_cookie - assign a cookie
+ */
+static dma_cookie_t ppc440spe_desc_assign_cookie(
+ struct ppc440spe_adma_chan *chan,
+ struct ppc440spe_adma_desc_slot *desc)
+{
+ dma_cookie_t cookie = chan->common.cookie;
+
+ cookie++;
+ if (cookie < 0)
+ cookie = 1;
+ chan->common.cookie = desc->async_tx.cookie = cookie;
+ return cookie;
+}
+
+/**
+ * ppc440spe_rxor_set_region_data -
+ */
+static void ppc440spe_rxor_set_region(struct ppc440spe_adma_desc_slot *desc,
+ u8 xor_arg_no, u32 mask)
+{
+ struct xor_cb *xcb = desc->hw_desc;
+
+ xcb->ops[xor_arg_no].h |= mask;
+}
+
+/**
+ * ppc440spe_rxor_set_src -
+ */
+static void ppc440spe_rxor_set_src(struct ppc440spe_adma_desc_slot *desc,
+ u8 xor_arg_no, dma_addr_t addr)
+{
+ struct xor_cb *xcb = desc->hw_desc;
+
+ xcb->ops[xor_arg_no].h |= DMA_CUED_XOR_BASE;
+ xcb->ops[xor_arg_no].l = addr;
+}
+
+/**
+ * ppc440spe_rxor_set_mult -
+ */
+static void ppc440spe_rxor_set_mult(struct ppc440spe_adma_desc_slot *desc,
+ u8 xor_arg_no, u8 idx, u8 mult)
+{
+ struct xor_cb *xcb = desc->hw_desc;
+
+ xcb->ops[xor_arg_no].h |= mult << (DMA_CUED_MULT1_OFF + idx * 8);
+}
+
+/**
+ * ppc440spe_adma_check_threshold - append CDBs to h/w chain if threshold
+ * has been achieved
+ */
+static void ppc440spe_adma_check_threshold(struct ppc440spe_adma_chan *chan)
+{
+ dev_dbg(chan->device->common.dev, "ppc440spe adma%d: pending: %d\n",
+ chan->device->id, chan->pending);
+
+ if (chan->pending >= PPC440SPE_ADMA_THRESHOLD) {
+ chan->pending = 0;
+ ppc440spe_chan_append(chan);
+ }
+}
+
+/**
+ * ppc440spe_adma_tx_submit - submit new descriptor group to the channel
+ * (it's not necessary that descriptors will be submitted to the h/w
+ * chains too right now)
+ */
+static dma_cookie_t ppc440spe_adma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct ppc440spe_adma_desc_slot *sw_desc;
+ struct ppc440spe_adma_chan *chan = to_ppc440spe_adma_chan(tx->chan);
+ struct ppc440spe_adma_desc_slot *group_start, *old_chain_tail;
+ int slot_cnt;
+ int slots_per_op;
+ dma_cookie_t cookie;
+
+ sw_desc = tx_to_ppc440spe_adma_slot(tx);
+
+ group_start = sw_desc->group_head;
+ slot_cnt = group_start->slot_cnt;
+ slots_per_op = group_start->slots_per_op;
+
+ spin_lock_bh(&chan->lock);
+
+ cookie = ppc440spe_desc_assign_cookie(chan, sw_desc);
+
+ if (unlikely(list_empty(&chan->chain))) {
+ /* first peer */
+ list_splice_init(&sw_desc->group_list, &chan->chain);
+ chan_first_cdb[chan->device->id] = group_start;
+ } else {
+ /* isn't first peer, bind CDBs to chain */
+ old_chain_tail = list_entry(chan->chain.prev,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ list_splice_init(&sw_desc->group_list,
+ &old_chain_tail->chain_node);
+ /* fix up the hardware chain */
+ ppc440spe_desc_set_link(chan, old_chain_tail, group_start);
+ }
+
+ /* increment the pending count by the number of operations */
+ chan->pending += slot_cnt / slots_per_op;
+ ppc440spe_adma_check_threshold(chan);
+ spin_unlock_bh(&chan->lock);
+
+ dev_dbg(chan->device->common.dev,
+ "ppc440spe adma%d: %s cookie: %d slot: %d tx %p\n",
+ chan->device->id, __func__,
+ sw_desc->async_tx.cookie, sw_desc->idx, sw_desc);
+
+ return cookie;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_interrupt - prepare CDB for a pseudo DMA operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_interrupt(
+ struct dma_chan *chan, unsigned long flags)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+ int slot_cnt, slots_per_op;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d: %s\n", ppc440spe_chan->device->id,
+ __func__);
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+ slot_cnt = slots_per_op = 1;
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ ppc440spe_desc_init_interrupt(group_start, ppc440spe_chan);
+ group_start->unmap_len = 0;
+ sw_desc->async_tx.flags = flags;
+ }
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_memcpy - prepare CDB for a MEMCPY operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_memcpy(
+ struct dma_chan *chan, dma_addr_t dma_dest,
+ dma_addr_t dma_src, size_t len, unsigned long flags)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+ int slot_cnt, slots_per_op;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+ if (unlikely(!len))
+ return NULL;
+
+ BUG_ON(unlikely(len > PPC440SPE_ADMA_DMA_MAX_BYTE_COUNT));
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d: %s len: %u int_en %d\n",
+ ppc440spe_chan->device->id, __func__, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+ slot_cnt = slots_per_op = 1;
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ ppc440spe_desc_init_memcpy(group_start, flags);
+ ppc440spe_adma_set_dest(group_start, dma_dest, 0);
+ ppc440spe_adma_memcpy_xor_set_src(group_start, dma_src, 0);
+ ppc440spe_desc_set_byte_count(group_start, ppc440spe_chan, len);
+ sw_desc->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_memset - prepare CDB for a MEMSET operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_memset(
+ struct dma_chan *chan, dma_addr_t dma_dest, int value,
+ size_t len, unsigned long flags)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+ int slot_cnt, slots_per_op;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+ if (unlikely(!len))
+ return NULL;
+
+ BUG_ON(unlikely(len > PPC440SPE_ADMA_DMA_MAX_BYTE_COUNT));
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d: %s cal: %u len: %u int_en %d\n",
+ ppc440spe_chan->device->id, __func__, value, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ slot_cnt = slots_per_op = 1;
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ ppc440spe_desc_init_memset(group_start, value, flags);
+ ppc440spe_adma_set_dest(group_start, dma_dest, 0);
+ ppc440spe_desc_set_byte_count(group_start, ppc440spe_chan, len);
+ sw_desc->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_xor - prepare CDB for a XOR operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_xor(
+ struct dma_chan *chan, dma_addr_t dma_dest,
+ dma_addr_t *dma_src, u32 src_cnt, size_t len,
+ unsigned long flags)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+ int slot_cnt, slots_per_op;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+ ADMA_LL_DBG(prep_dma_xor_dbg(ppc440spe_chan->device->id,
+ dma_dest, dma_src, src_cnt));
+ if (unlikely(!len))
+ return NULL;
+ BUG_ON(unlikely(len > PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT));
+
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d: %s src_cnt: %d len: %u int_en: %d\n",
+ ppc440spe_chan->device->id, __func__, src_cnt, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+ slot_cnt = ppc440spe_chan_xor_slot_count(len, src_cnt, &slots_per_op);
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ ppc440spe_desc_init_xor(group_start, src_cnt, flags);
+ ppc440spe_adma_set_dest(group_start, dma_dest, 0);
+ while (src_cnt--)
+ ppc440spe_adma_memcpy_xor_set_src(group_start,
+ dma_src[src_cnt], src_cnt);
+ ppc440spe_desc_set_byte_count(group_start, ppc440spe_chan, len);
+ sw_desc->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+static inline void
+ppc440spe_desc_set_xor_src_cnt(struct ppc440spe_adma_desc_slot *desc,
+ int src_cnt);
+static void ppc440spe_init_rxor_cursor(struct ppc440spe_rxor *cursor);
+
+/**
+ * ppc440spe_adma_init_dma2rxor_slot -
+ */
+static void ppc440spe_adma_init_dma2rxor_slot(
+ struct ppc440spe_adma_desc_slot *desc,
+ dma_addr_t *src, int src_cnt)
+{
+ int i;
+
+ /* initialize CDB */
+ for (i = 0; i < src_cnt; i++) {
+ ppc440spe_adma_dma2rxor_prep_src(desc, &desc->rxor_cursor, i,
+ desc->src_cnt, (u32)src[i]);
+ }
+}
+
+/**
+ * ppc440spe_dma01_prep_mult -
+ * for Q operation where destination is also the source
+ */
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma01_prep_mult(
+ struct ppc440spe_adma_chan *ppc440spe_chan,
+ dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+ const unsigned char *scf, size_t len, unsigned long flags)
+{
+ struct ppc440spe_adma_desc_slot *sw_desc = NULL;
+ unsigned long op = 0;
+ int slot_cnt;
+
+ set_bit(PPC440SPE_DESC_WXOR, &op);
+ slot_cnt = 2;
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+
+ /* use WXOR, each descriptor occupies one slot */
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+ if (sw_desc) {
+ struct ppc440spe_adma_chan *chan;
+ struct ppc440spe_adma_desc_slot *iter;
+ struct dma_cdb *hw_desc;
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+ set_bits(op, &sw_desc->flags);
+ sw_desc->src_cnt = src_cnt;
+ sw_desc->dst_cnt = dst_cnt;
+ /* First descriptor, zero data in the destination and copy it
+ * to q page using MULTICAST transfer.
+ */
+ iter = list_first_entry(&sw_desc->group_list,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ /* set 'next' pointer */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MULTICAST;
+
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, dst[0], 0);
+ ppc440spe_desc_set_dest_addr(iter, chan, 0, dst[1], 1);
+ ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ src[0]);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+ iter->unmap_len = len;
+
+ /*
+ * Second descriptor, multiply data from the q page
+ * and store the result in real destination.
+ */
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = NULL;
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC440SPE_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ ppc440spe_desc_set_src_addr(iter, chan, 0,
+ DMA_CUED_XOR_HB, dst[1]);
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, dst[0], 0);
+
+ ppc440spe_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+ DMA_CDB_SG_DST1, scf[0]);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+ iter->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ return sw_desc;
+}
+
+/**
+ * ppc440spe_dma01_prep_sum_product -
+ * Dx = A*(P+Pxy) + B*(Q+Qxy) operation where destination is also
+ * the source.
+ */
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma01_prep_sum_product(
+ struct ppc440spe_adma_chan *ppc440spe_chan,
+ dma_addr_t *dst, dma_addr_t *src, int src_cnt,
+ const unsigned char *scf, size_t len, unsigned long flags)
+{
+ struct ppc440spe_adma_desc_slot *sw_desc = NULL;
+ unsigned long op = 0;
+ int slot_cnt;
+
+ set_bit(PPC440SPE_DESC_WXOR, &op);
+ slot_cnt = 3;
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+
+ /* WXOR, each descriptor occupies one slot */
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+ if (sw_desc) {
+ struct ppc440spe_adma_chan *chan;
+ struct ppc440spe_adma_desc_slot *iter;
+ struct dma_cdb *hw_desc;
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+ set_bits(op, &sw_desc->flags);
+ sw_desc->src_cnt = src_cnt;
+ sw_desc->dst_cnt = 1;
+ /* 1st descriptor, src[1] data to q page and zero destination */
+ iter = list_first_entry(&sw_desc->group_list,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MULTICAST;
+
+ ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+ *dst, 0);
+ ppc440spe_desc_set_dest_addr(iter, chan, 0,
+ ppc440spe_chan->qdest, 1);
+ ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ src[1]);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+ iter->unmap_len = len;
+
+ /* 2nd descriptor, multiply src[1] data and store the
+ * result in destination */
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ /* set 'next' pointer */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC440SPE_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ ppc440spe_chan->qdest);
+ ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+ *dst, 0);
+ ppc440spe_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+ DMA_CDB_SG_DST1, scf[1]);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+ iter->unmap_len = len;
+
+ /*
+ * 3rd descriptor, multiply src[0] data and xor it
+ * with destination
+ */
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = NULL;
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC440SPE_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC440SPE_DESC_INT, &iter->flags);
+
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ src[0]);
+ ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+ *dst, 0);
+ ppc440spe_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+ DMA_CDB_SG_DST1, scf[0]);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan, len);
+ iter->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ return sw_desc;
+}
+
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma01_prep_pq(
+ struct ppc440spe_adma_chan *ppc440spe_chan,
+ dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+ const unsigned char *scf, size_t len, unsigned long flags)
+{
+ int slot_cnt;
+ struct ppc440spe_adma_desc_slot *sw_desc = NULL, *iter;
+ unsigned long op = 0;
+ unsigned char mult = 1;
+
+ pr_debug("%s: dst_cnt %d, src_cnt %d, len %d\n",
+ __func__, dst_cnt, src_cnt, len);
+ /* select operations WXOR/RXOR depending on the
+ * source addresses of operators and the number
+ * of destinations (RXOR support only Q-parity calculations)
+ */
+ set_bit(PPC440SPE_DESC_WXOR, &op);
+ if (!test_and_set_bit(PPC440SPE_RXOR_RUN, &ppc440spe_rxor_state)) {
+ /* no active RXOR;
+ * do RXOR if:
+ * - there are more than 1 source,
+ * - len is aligned on 512-byte boundary,
+ * - source addresses fit to one of 4 possible regions.
+ */
+ if (src_cnt > 1 &&
+ !(len & MQ0_CF2H_RXOR_BS_MASK) &&
+ (src[0] + len) == src[1]) {
+ /* may do RXOR R1 R2 */
+ set_bit(PPC440SPE_DESC_RXOR, &op);
+ if (src_cnt != 2) {
+ /* may try to enhance region of RXOR */
+ if ((src[1] + len) == src[2]) {
+ /* do RXOR R1 R2 R3 */
+ set_bit(PPC440SPE_DESC_RXOR123,
+ &op);
+ } else if ((src[1] + len * 2) == src[2]) {
+ /* do RXOR R1 R2 R4 */
+ set_bit(PPC440SPE_DESC_RXOR124, &op);
+ } else if ((src[1] + len * 3) == src[2]) {
+ /* do RXOR R1 R2 R5 */
+ set_bit(PPC440SPE_DESC_RXOR125,
+ &op);
+ } else {
+ /* do RXOR R1 R2 */
+ set_bit(PPC440SPE_DESC_RXOR12,
+ &op);
+ }
+ } else {
+ /* do RXOR R1 R2 */
+ set_bit(PPC440SPE_DESC_RXOR12, &op);
+ }
+ }
+
+ if (!test_bit(PPC440SPE_DESC_RXOR, &op)) {
+ /* can not do this operation with RXOR */
+ clear_bit(PPC440SPE_RXOR_RUN,
+ &ppc440spe_rxor_state);
+ } else {
+ /* can do; set block size right now */
+ ppc440spe_desc_set_rxor_block_size(len);
+ }
+ }
+
+ /* Number of necessary slots depends on operation type selected */
+ if (!test_bit(PPC440SPE_DESC_RXOR, &op)) {
+ /* This is a WXOR only chain. Need descriptors for each
+ * source to GF-XOR them with WXOR, and need descriptors
+ * for each destination to zero them with WXOR
+ */
+ slot_cnt = src_cnt;
+
+ if (flags & DMA_PREP_ZERO_P) {
+ slot_cnt++;
+ set_bit(PPC440SPE_ZERO_P, &op);
+ }
+ if (flags & DMA_PREP_ZERO_Q) {
+ slot_cnt++;
+ set_bit(PPC440SPE_ZERO_Q, &op);
+ }
+ } else {
+ /* Need 1/2 descriptor for RXOR operation, and
+ * need (src_cnt - (2 or 3)) for WXOR of sources
+ * remained (if any)
+ */
+ slot_cnt = dst_cnt;
+
+ if (flags & DMA_PREP_ZERO_P)
+ set_bit(PPC440SPE_ZERO_P, &op);
+ if (flags & DMA_PREP_ZERO_Q)
+ set_bit(PPC440SPE_ZERO_Q, &op);
+
+ if (test_bit(PPC440SPE_DESC_RXOR12, &op))
+ slot_cnt += src_cnt - 2;
+ else
+ slot_cnt += src_cnt - 3;
+
+ /* Thus we have either RXOR only chain or
+ * mixed RXOR/WXOR
+ */
+ if (slot_cnt == dst_cnt)
+ /* RXOR only chain */
+ clear_bit(PPC440SPE_DESC_WXOR, &op);
+ }
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+ /* for both RXOR/WXOR each descriptor occupies one slot */
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+ if (sw_desc) {
+ ppc440spe_desc_init_dma01pq(sw_desc, dst_cnt, src_cnt,
+ flags, op);
+
+ /* setup dst/src/mult */
+ pr_debug("%s: set dst descriptor 0, 1: 0x%016llx, 0x%016llx\n",
+ __func__, dst[0], dst[1]);
+ ppc440spe_adma_pq_set_dest(sw_desc, dst, flags);
+ while (src_cnt--) {
+ ppc440spe_adma_pq_set_src(sw_desc, src[src_cnt],
+ src_cnt);
+
+ /* NOTE: "Multi = 0 is equivalent to = 1" as it
+ * stated in 440SPSPe_RAID6_Addendum_UM_1_17.pdf
+ * doesn't work for RXOR with DMA0/1! Instead, multi=0
+ * leads to zeroing source data after RXOR.
+ * So, for P case set-up mult=1 explicitly.
+ */
+ if (!(flags & DMA_PREP_PQ_DISABLE_Q))
+ mult = scf[src_cnt];
+ ppc440spe_adma_pq_set_src_mult(sw_desc,
+ mult, src_cnt, dst_cnt - 1);
+ }
+
+ /* Setup byte count foreach slot just allocated */
+ sw_desc->async_tx.flags = flags;
+ list_for_each_entry(iter, &sw_desc->group_list,
+ chain_node) {
+ ppc440spe_desc_set_byte_count(iter,
+ ppc440spe_chan, len);
+ iter->unmap_len = len;
+ }
+ }
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ return sw_desc;
+}
+
+static struct ppc440spe_adma_desc_slot *ppc440spe_dma2_prep_pq(
+ struct ppc440spe_adma_chan *ppc440spe_chan,
+ dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+ const unsigned char *scf, size_t len, unsigned long flags)
+{
+ int slot_cnt, descs_per_op;
+ struct ppc440spe_adma_desc_slot *sw_desc = NULL, *iter;
+ unsigned long op = 0;
+ unsigned char mult = 1;
+
+ BUG_ON(!dst_cnt);
+ /*pr_debug("%s: dst_cnt %d, src_cnt %d, len %d\n",
+ __func__, dst_cnt, src_cnt, len);*/
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+ descs_per_op = ppc440spe_dma2_pq_slot_count(src, src_cnt, len);
+ if (descs_per_op < 0) {
+ spin_unlock_bh(&ppc440spe_chan->lock);
+ return NULL;
+ }
+
+ /* depending on number of sources we have 1 or 2 RXOR chains */
+ slot_cnt = descs_per_op * dst_cnt;
+
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt, 1);
+ if (sw_desc) {
+ op = slot_cnt;
+ sw_desc->async_tx.flags = flags;
+ list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+ ppc440spe_desc_init_dma2pq(iter, dst_cnt, src_cnt,
+ --op ? 0 : flags);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+ len);
+ iter->unmap_len = len;
+
+ ppc440spe_init_rxor_cursor(&(iter->rxor_cursor));
+ iter->rxor_cursor.len = len;
+ iter->descs_per_op = descs_per_op;
+ }
+ op = 0;
+ list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+ op++;
+ if (op % descs_per_op == 0)
+ ppc440spe_adma_init_dma2rxor_slot(iter, src,
+ src_cnt);
+ if (likely(!list_is_last(&iter->chain_node,
+ &sw_desc->group_list))) {
+ /* set 'next' pointer */
+ iter->hw_next =
+ list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ ppc440spe_xor_set_link(iter, iter->hw_next);
+ } else {
+ /* this is the last descriptor. */
+ iter->hw_next = NULL;
+ }
+ }
+
+ /* fixup head descriptor */
+ sw_desc->dst_cnt = dst_cnt;
+ if (flags & DMA_PREP_ZERO_P)
+ set_bit(PPC440SPE_ZERO_P, &sw_desc->flags);
+ if (flags & DMA_PREP_ZERO_Q)
+ set_bit(PPC440SPE_ZERO_Q, &sw_desc->flags);
+
+ /* setup dst/src/mult */
+ ppc440spe_adma_pq_set_dest(sw_desc, dst, flags);
+
+ while (src_cnt--) {
+ /* handle descriptors (if dst_cnt == 2) inside
+ * the ppc440spe_adma_pq_set_srcxxx() functions
+ */
+ ppc440spe_adma_pq_set_src(sw_desc, src[src_cnt],
+ src_cnt);
+ if (!(flags & DMA_PREP_PQ_DISABLE_Q))
+ mult = scf[src_cnt];
+ ppc440spe_adma_pq_set_src_mult(sw_desc,
+ mult, src_cnt, dst_cnt - 1);
+ }
+ }
+ spin_unlock_bh(&ppc440spe_chan->lock);
+ ppc440spe_desc_set_rxor_block_size(len);
+ return sw_desc;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_pq - prepare CDB (group) for a GF-XOR operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_pq(
+ struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
+ unsigned int src_cnt, const unsigned char *scf,
+ size_t len, unsigned long flags)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *sw_desc = NULL;
+ int dst_cnt = 0;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+ ADMA_LL_DBG(prep_dma_pq_dbg(ppc440spe_chan->device->id,
+ dst, src, src_cnt));
+ BUG_ON(!len);
+ BUG_ON(unlikely(len > PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT));
+ BUG_ON(!src_cnt);
+
+ if (src_cnt == 1 && dst[1] == src[0]) {
+ dma_addr_t dest[2];
+
+ /* dst[1] is real destination (Q) */
+ dest[0] = dst[1];
+ /* this is the page to multicast source data to */
+ dest[1] = ppc440spe_chan->qdest;
+ sw_desc = ppc440spe_dma01_prep_mult(ppc440spe_chan,
+ dest, 2, src, src_cnt, scf, len, flags);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+ }
+
+ if (src_cnt == 2 && dst[1] == src[1]) {
+ sw_desc = ppc440spe_dma01_prep_sum_product(ppc440spe_chan,
+ &dst[1], src, 2, scf, len, flags);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+ }
+
+ if (!(flags & DMA_PREP_PQ_DISABLE_P)) {
+ BUG_ON(!dst[0]);
+ dst_cnt++;
+ flags |= DMA_PREP_ZERO_P;
+ }
+
+ if (!(flags & DMA_PREP_PQ_DISABLE_Q)) {
+ BUG_ON(!dst[1]);
+ dst_cnt++;
+ flags |= DMA_PREP_ZERO_Q;
+ }
+
+ BUG_ON(!dst_cnt);
+
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d: %s src_cnt: %d len: %u int_en: %d\n",
+ ppc440spe_chan->device->id, __func__, src_cnt, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ switch (ppc440spe_chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ sw_desc = ppc440spe_dma01_prep_pq(ppc440spe_chan,
+ dst, dst_cnt, src, src_cnt, scf,
+ len, flags);
+ break;
+
+ case PPC440SPE_XOR_ID:
+ sw_desc = ppc440spe_dma2_prep_pq(ppc440spe_chan,
+ dst, dst_cnt, src, src_cnt, scf,
+ len, flags);
+ break;
+ }
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_pqzero_sum - prepare CDB group for
+ * a PQ_ZERO_SUM operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_pqzero_sum(
+ struct dma_chan *chan, dma_addr_t *pq, dma_addr_t *src,
+ unsigned int src_cnt, const unsigned char *scf, size_t len,
+ enum sum_check_flags *pqres, unsigned long flags)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *sw_desc, *iter;
+ dma_addr_t pdest, qdest;
+ int slot_cnt, slots_per_op, idst, dst_cnt;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+
+ if (flags & DMA_PREP_PQ_DISABLE_P)
+ pdest = 0;
+ else
+ pdest = pq[0];
+
+ if (flags & DMA_PREP_PQ_DISABLE_Q)
+ qdest = 0;
+ else
+ qdest = pq[1];
+
+ ADMA_LL_DBG(prep_dma_pqzero_sum_dbg(ppc440spe_chan->device->id,
+ src, src_cnt, scf));
+
+ /* Always use WXOR for P/Q calculations (two destinations).
+ * Need 1 or 2 extra slots to verify results are zero.
+ */
+ idst = dst_cnt = (pdest && qdest) ? 2 : 1;
+
+ /* One additional slot per destination to clone P/Q
+ * before calculation (we have to preserve destinations).
+ */
+ slot_cnt = src_cnt + dst_cnt * 2;
+ slots_per_op = 1;
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+ sw_desc = ppc440spe_adma_alloc_slots(ppc440spe_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ ppc440spe_desc_init_dma01pqzero_sum(sw_desc, dst_cnt, src_cnt);
+
+ /* Setup byte count for each slot just allocated */
+ sw_desc->async_tx.flags = flags;
+ list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+ len);
+ iter->unmap_len = len;
+ }
+
+ if (pdest) {
+ struct dma_cdb *hw_desc;
+ struct ppc440spe_adma_chan *chan;
+
+ iter = sw_desc->group_head;
+ chan = to_ppc440spe_adma_chan(iter->async_tx.chan);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter->src_cnt = 0;
+ iter->dst_cnt = 0;
+ ppc440spe_desc_set_dest_addr(iter, chan, 0,
+ ppc440spe_chan->pdest, 0);
+ ppc440spe_desc_set_src_addr(iter, chan, 0, 0, pdest);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+ len);
+ iter->unmap_len = 0;
+ /* override pdest to preserve original P */
+ pdest = ppc440spe_chan->pdest;
+ }
+ if (qdest) {
+ struct dma_cdb *hw_desc;
+ struct ppc440spe_adma_chan *chan;
+
+ iter = list_first_entry(&sw_desc->group_list,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ chan = to_ppc440spe_adma_chan(iter->async_tx.chan);
+
+ if (pdest) {
+ iter = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ }
+
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter->src_cnt = 0;
+ iter->dst_cnt = 0;
+ ppc440spe_desc_set_dest_addr(iter, chan, 0,
+ ppc440spe_chan->qdest, 0);
+ ppc440spe_desc_set_src_addr(iter, chan, 0, 0, qdest);
+ ppc440spe_desc_set_byte_count(iter, ppc440spe_chan,
+ len);
+ iter->unmap_len = 0;
+ /* override qdest to preserve original Q */
+ qdest = ppc440spe_chan->qdest;
+ }
+
+ /* Setup destinations for P/Q ops */
+ ppc440spe_adma_pqzero_sum_set_dest(sw_desc, pdest, qdest);
+
+ /* Setup zero QWORDs into DCHECK CDBs */
+ idst = dst_cnt;
+ list_for_each_entry_reverse(iter, &sw_desc->group_list,
+ chain_node) {
+ /*
+ * The last CDB corresponds to Q-parity check,
+ * the one before last CDB corresponds
+ * P-parity check
+ */
+ if (idst == DMA_DEST_MAX_NUM) {
+ if (idst == dst_cnt) {
+ set_bit(PPC440SPE_DESC_QCHECK,
+ &iter->flags);
+ } else {
+ set_bit(PPC440SPE_DESC_PCHECK,
+ &iter->flags);
+ }
+ } else {
+ if (qdest) {
+ set_bit(PPC440SPE_DESC_QCHECK,
+ &iter->flags);
+ } else {
+ set_bit(PPC440SPE_DESC_PCHECK,
+ &iter->flags);
+ }
+ }
+ iter->xor_check_result = pqres;
+
+ /*
+ * set it to zero, if check fail then result will
+ * be updated
+ */
+ *iter->xor_check_result = 0;
+ ppc440spe_desc_set_dcheck(iter, ppc440spe_chan,
+ ppc440spe_qword);
+
+ if (!(--dst_cnt))
+ break;
+ }
+
+ /* Setup sources and mults for P/Q ops */
+ list_for_each_entry_continue_reverse(iter, &sw_desc->group_list,
+ chain_node) {
+ struct ppc440spe_adma_chan *chan;
+ u32 mult_dst;
+
+ chan = to_ppc440spe_adma_chan(iter->async_tx.chan);
+ ppc440spe_desc_set_src_addr(iter, chan, 0,
+ DMA_CUED_XOR_HB,
+ src[src_cnt - 1]);
+ if (qdest) {
+ mult_dst = (dst_cnt - 1) ? DMA_CDB_SG_DST2 :
+ DMA_CDB_SG_DST1;
+ ppc440spe_desc_set_src_mult(iter, chan,
+ DMA_CUED_MULT1_OFF,
+ mult_dst,
+ scf[src_cnt - 1]);
+ }
+ if (!(--src_cnt))
+ break;
+ }
+ }
+ spin_unlock_bh(&ppc440spe_chan->lock);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/**
+ * ppc440spe_adma_prep_dma_xor_zero_sum - prepare CDB group for
+ * XOR ZERO_SUM operation
+ */
+static struct dma_async_tx_descriptor *ppc440spe_adma_prep_dma_xor_zero_sum(
+ struct dma_chan *chan, dma_addr_t *src, unsigned int src_cnt,
+ size_t len, enum sum_check_flags *result, unsigned long flags)
+{
+ struct dma_async_tx_descriptor *tx;
+ dma_addr_t pq[2];
+
+ /* validate P, disable Q */
+ pq[0] = src[0];
+ pq[1] = 0;
+ flags |= DMA_PREP_PQ_DISABLE_Q;
+
+ tx = ppc440spe_adma_prep_dma_pqzero_sum(chan, pq, &src[1],
+ src_cnt - 1, 0, len,
+ result, flags);
+ return tx;
+}
+
+/**
+ * ppc440spe_adma_set_dest - set destination address into descriptor
+ */
+static void ppc440spe_adma_set_dest(struct ppc440spe_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index)
+{
+ struct ppc440spe_adma_chan *chan;
+
+ BUG_ON(index >= sw_desc->dst_cnt);
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ /* to do: support transfers lengths >
+ * PPC440SPE_ADMA_DMA/XOR_MAX_BYTE_COUNT
+ */
+ ppc440spe_desc_set_dest_addr(sw_desc->group_head,
+ chan, 0, addr, index);
+ break;
+ case PPC440SPE_XOR_ID:
+ sw_desc = ppc440spe_get_group_entry(sw_desc, index);
+ ppc440spe_desc_set_dest_addr(sw_desc,
+ chan, 0, addr, index);
+ break;
+ }
+}
+
+static void ppc440spe_adma_pq_zero_op(struct ppc440spe_adma_desc_slot *iter,
+ struct ppc440spe_adma_chan *chan, dma_addr_t addr)
+{
+ /* To clear destinations update the descriptor
+ * (P or Q depending on index) as follows:
+ * addr is destination (0 corresponds to SG2):
+ */
+ ppc440spe_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE, addr, 0);
+
+ /* ... and the addr is source: */
+ ppc440spe_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB, addr);
+
+ /* addr is always SG2 then the mult is always DST1 */
+ ppc440spe_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+ DMA_CDB_SG_DST1, 1);
+}
+
+/**
+ * ppc440spe_adma_pq_set_dest - set destination address into descriptor
+ * for the PQXOR operation
+ */
+static void ppc440spe_adma_pq_set_dest(struct ppc440spe_adma_desc_slot *sw_desc,
+ dma_addr_t *addrs, unsigned long flags)
+{
+ struct ppc440spe_adma_desc_slot *iter;
+ struct ppc440spe_adma_chan *chan;
+ dma_addr_t paddr, qaddr;
+ dma_addr_t addr = 0, ppath, qpath;
+ int index = 0, i;
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+ if (flags & DMA_PREP_PQ_DISABLE_P)
+ paddr = 0;
+ else
+ paddr = addrs[0];
+
+ if (flags & DMA_PREP_PQ_DISABLE_Q)
+ qaddr = 0;
+ else
+ qaddr = addrs[1];
+
+ if (!paddr || !qaddr)
+ addr = paddr ? paddr : qaddr;
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ /* walk through the WXOR source list and set P/Q-destinations
+ * for each slot:
+ */
+ if (!test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags)) {
+ /* This is WXOR-only chain; may have 1/2 zero descs */
+ if (test_bit(PPC440SPE_ZERO_P, &sw_desc->flags))
+ index++;
+ if (test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags))
+ index++;
+
+ iter = ppc440spe_get_group_entry(sw_desc, index);
+ if (addr) {
+ /* one destination */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list, chain_node)
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, addr, 0);
+ } else {
+ /* two destinations */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list, chain_node) {
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, paddr, 0);
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, qaddr, 1);
+ }
+ }
+
+ if (index) {
+ /* To clear destinations update the descriptor
+ * (1st,2nd, or both depending on flags)
+ */
+ index = 0;
+ if (test_bit(PPC440SPE_ZERO_P,
+ &sw_desc->flags)) {
+ iter = ppc440spe_get_group_entry(
+ sw_desc, index++);
+ ppc440spe_adma_pq_zero_op(iter, chan,
+ paddr);
+ }
+
+ if (test_bit(PPC440SPE_ZERO_Q,
+ &sw_desc->flags)) {
+ iter = ppc440spe_get_group_entry(
+ sw_desc, index++);
+ ppc440spe_adma_pq_zero_op(iter, chan,
+ qaddr);
+ }
+
+ return;
+ }
+ } else {
+ /* This is RXOR-only or RXOR/WXOR mixed chain */
+
+ /* If we want to include destination into calculations,
+ * then make dest addresses cued with mult=1 (XOR).
+ */
+ ppath = test_bit(PPC440SPE_ZERO_P, &sw_desc->flags) ?
+ DMA_CUED_XOR_HB :
+ DMA_CUED_XOR_BASE |
+ (1 << DMA_CUED_MULT1_OFF);
+ qpath = test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags) ?
+ DMA_CUED_XOR_HB :
+ DMA_CUED_XOR_BASE |
+ (1 << DMA_CUED_MULT1_OFF);
+
+ /* Setup destination(s) in RXOR slot(s) */
+ iter = ppc440spe_get_group_entry(sw_desc, index++);
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ paddr ? ppath : qpath,
+ paddr ? paddr : qaddr, 0);
+ if (!addr) {
+ /* two destinations */
+ iter = ppc440spe_get_group_entry(sw_desc,
+ index++);
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ qpath, qaddr, 0);
+ }
+
+ if (test_bit(PPC440SPE_DESC_WXOR, &sw_desc->flags)) {
+ /* Setup destination(s) in remaining WXOR
+ * slots
+ */
+ iter = ppc440spe_get_group_entry(sw_desc,
+ index);
+ if (addr) {
+ /* one destination */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list,
+ chain_node)
+ ppc440spe_desc_set_dest_addr(
+ iter, chan,
+ DMA_CUED_XOR_BASE,
+ addr, 0);
+
+ } else {
+ /* two destinations */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list,
+ chain_node) {
+ ppc440spe_desc_set_dest_addr(
+ iter, chan,
+ DMA_CUED_XOR_BASE,
+ paddr, 0);
+ ppc440spe_desc_set_dest_addr(
+ iter, chan,
+ DMA_CUED_XOR_BASE,
+ qaddr, 1);
+ }
+ }
+ }
+
+ }
+ break;
+
+ case PPC440SPE_XOR_ID:
+ /* DMA2 descriptors have only 1 destination, so there are
+ * two chains - one for each dest.
+ * If we want to include destination into calculations,
+ * then make dest addresses cued with mult=1 (XOR).
+ */
+ ppath = test_bit(PPC440SPE_ZERO_P, &sw_desc->flags) ?
+ DMA_CUED_XOR_HB :
+ DMA_CUED_XOR_BASE |
+ (1 << DMA_CUED_MULT1_OFF);
+
+ qpath = test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags) ?
+ DMA_CUED_XOR_HB :
+ DMA_CUED_XOR_BASE |
+ (1 << DMA_CUED_MULT1_OFF);
+
+ iter = ppc440spe_get_group_entry(sw_desc, 0);
+ for (i = 0; i < sw_desc->descs_per_op; i++) {
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ paddr ? ppath : qpath,
+ paddr ? paddr : qaddr, 0);
+ iter = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ }
+
+ if (!addr) {
+ /* Two destinations; setup Q here */
+ iter = ppc440spe_get_group_entry(sw_desc,
+ sw_desc->descs_per_op);
+ for (i = 0; i < sw_desc->descs_per_op; i++) {
+ ppc440spe_desc_set_dest_addr(iter,
+ chan, qpath, qaddr, 0);
+ iter = list_entry(iter->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ }
+ }
+
+ break;
+ }
+}
+
+/**
+ * ppc440spe_adma_pq_zero_sum_set_dest - set destination address into descriptor
+ * for the PQ_ZERO_SUM operation
+ */
+static void ppc440spe_adma_pqzero_sum_set_dest(
+ struct ppc440spe_adma_desc_slot *sw_desc,
+ dma_addr_t paddr, dma_addr_t qaddr)
+{
+ struct ppc440spe_adma_desc_slot *iter, *end;
+ struct ppc440spe_adma_chan *chan;
+ dma_addr_t addr = 0;
+ int idx;
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+ /* walk through the WXOR source list and set P/Q-destinations
+ * for each slot
+ */
+ idx = (paddr && qaddr) ? 2 : 1;
+ /* set end */
+ list_for_each_entry_reverse(end, &sw_desc->group_list,
+ chain_node) {
+ if (!(--idx))
+ break;
+ }
+ /* set start */
+ idx = (paddr && qaddr) ? 2 : 1;
+ iter = ppc440spe_get_group_entry(sw_desc, idx);
+
+ if (paddr && qaddr) {
+ /* two destinations */
+ list_for_each_entry_from(iter, &sw_desc->group_list,
+ chain_node) {
+ if (unlikely(iter == end))
+ break;
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, paddr, 0);
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, qaddr, 1);
+ }
+ } else {
+ /* one destination */
+ addr = paddr ? paddr : qaddr;
+ list_for_each_entry_from(iter, &sw_desc->group_list,
+ chain_node) {
+ if (unlikely(iter == end))
+ break;
+ ppc440spe_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, addr, 0);
+ }
+ }
+
+ /* The remaining descriptors are DATACHECK. These have no need in
+ * destination. Actually, these destinations are used there
+ * as sources for check operation. So, set addr as source.
+ */
+ ppc440spe_desc_set_src_addr(end, chan, 0, 0, addr ? addr : paddr);
+
+ if (!addr) {
+ end = list_entry(end->chain_node.next,
+ struct ppc440spe_adma_desc_slot, chain_node);
+ ppc440spe_desc_set_src_addr(end, chan, 0, 0, qaddr);
+ }
+}
+
+/**
+ * ppc440spe_desc_set_xor_src_cnt - set source count into descriptor
+ */
+static inline void ppc440spe_desc_set_xor_src_cnt(
+ struct ppc440spe_adma_desc_slot *desc,
+ int src_cnt)
+{
+ struct xor_cb *hw_desc = desc->hw_desc;
+
+ hw_desc->cbc &= ~XOR_CDCR_OAC_MSK;
+ hw_desc->cbc |= src_cnt;
+}
+
+/**
+ * ppc440spe_adma_pq_set_src - set source address into descriptor
+ */
+static void ppc440spe_adma_pq_set_src(struct ppc440spe_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index)
+{
+ struct ppc440spe_adma_chan *chan;
+ dma_addr_t haddr = 0;
+ struct ppc440spe_adma_desc_slot *iter = NULL;
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ /* DMA0,1 may do: WXOR, RXOR, RXOR+WXORs chain
+ */
+ if (test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags)) {
+ /* RXOR-only or RXOR/WXOR operation */
+ int iskip = test_bit(PPC440SPE_DESC_RXOR12,
+ &sw_desc->flags) ? 2 : 3;
+
+ if (index == 0) {
+ /* 1st slot (RXOR) */
+ /* setup sources region (R1-2-3, R1-2-4,
+ * or R1-2-5)
+ */
+ if (test_bit(PPC440SPE_DESC_RXOR12,
+ &sw_desc->flags))
+ haddr = DMA_RXOR12 <<
+ DMA_CUED_REGION_OFF;
+ else if (test_bit(PPC440SPE_DESC_RXOR123,
+ &sw_desc->flags))
+ haddr = DMA_RXOR123 <<
+ DMA_CUED_REGION_OFF;
+ else if (test_bit(PPC440SPE_DESC_RXOR124,
+ &sw_desc->flags))
+ haddr = DMA_RXOR124 <<
+ DMA_CUED_REGION_OFF;
+ else if (test_bit(PPC440SPE_DESC_RXOR125,
+ &sw_desc->flags))
+ haddr = DMA_RXOR125 <<
+ DMA_CUED_REGION_OFF;
+ else
+ BUG();
+ haddr |= DMA_CUED_XOR_BASE;
+ iter = ppc440spe_get_group_entry(sw_desc, 0);
+ } else if (index < iskip) {
+ /* 1st slot (RXOR)
+ * shall actually set source address only once
+ * instead of first <iskip>
+ */
+ iter = NULL;
+ } else {
+ /* 2nd/3d and next slots (WXOR);
+ * skip first slot with RXOR
+ */
+ haddr = DMA_CUED_XOR_HB;
+ iter = ppc440spe_get_group_entry(sw_desc,
+ index - iskip + sw_desc->dst_cnt);
+ }
+ } else {
+ int znum = 0;
+
+ /* WXOR-only operation; skip first slots with
+ * zeroing destinations
+ */
+ if (test_bit(PPC440SPE_ZERO_P, &sw_desc->flags))
+ znum++;
+ if (test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags))
+ znum++;
+
+ haddr = DMA_CUED_XOR_HB;
+ iter = ppc440spe_get_group_entry(sw_desc,
+ index + znum);
+ }
+
+ if (likely(iter)) {
+ ppc440spe_desc_set_src_addr(iter, chan, 0, haddr, addr);
+
+ if (!index &&
+ test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags) &&
+ sw_desc->dst_cnt == 2) {
+ /* if we have two destinations for RXOR, then
+ * setup source in the second descr too
+ */
+ iter = ppc440spe_get_group_entry(sw_desc, 1);
+ ppc440spe_desc_set_src_addr(iter, chan, 0,
+ haddr, addr);
+ }
+ }
+ break;
+
+ case PPC440SPE_XOR_ID:
+ /* DMA2 may do Biskup */
+ iter = sw_desc->group_head;
+ if (iter->dst_cnt == 2) {
+ /* both P & Q calculations required; set P src here */
+ ppc440spe_adma_dma2rxor_set_src(iter, index, addr);
+
+ /* this is for Q */
+ iter = ppc440spe_get_group_entry(sw_desc,
+ sw_desc->descs_per_op);
+ }
+ ppc440spe_adma_dma2rxor_set_src(iter, index, addr);
+ break;
+ }
+}
+
+/**
+ * ppc440spe_adma_memcpy_xor_set_src - set source address into descriptor
+ */
+static void ppc440spe_adma_memcpy_xor_set_src(
+ struct ppc440spe_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index)
+{
+ struct ppc440spe_adma_chan *chan;
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+ sw_desc = sw_desc->group_head;
+
+ if (likely(sw_desc))
+ ppc440spe_desc_set_src_addr(sw_desc, chan, index, 0, addr);
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_inc_addr -
+ */
+static void ppc440spe_adma_dma2rxor_inc_addr(
+ struct ppc440spe_adma_desc_slot *desc,
+ struct ppc440spe_rxor *cursor, int index, int src_cnt)
+{
+ cursor->addr_count++;
+ if (index == src_cnt - 1) {
+ ppc440spe_desc_set_xor_src_cnt(desc, cursor->addr_count);
+ } else if (cursor->addr_count == XOR_MAX_OPS) {
+ ppc440spe_desc_set_xor_src_cnt(desc, cursor->addr_count);
+ cursor->addr_count = 0;
+ cursor->desc_count++;
+ }
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_prep_src - setup RXOR types in DMA2 CDB
+ */
+static int ppc440spe_adma_dma2rxor_prep_src(
+ struct ppc440spe_adma_desc_slot *hdesc,
+ struct ppc440spe_rxor *cursor, int index,
+ int src_cnt, u32 addr)
+{
+ int rval = 0;
+ u32 sign;
+ struct ppc440spe_adma_desc_slot *desc = hdesc;
+ int i;
+
+ for (i = 0; i < cursor->desc_count; i++) {
+ desc = list_entry(hdesc->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ }
+
+ switch (cursor->state) {
+ case 0:
+ if (addr == cursor->addrl + cursor->len) {
+ /* direct RXOR */
+ cursor->state = 1;
+ cursor->xor_count++;
+ if (index == src_cnt-1) {
+ ppc440spe_rxor_set_region(desc,
+ cursor->addr_count,
+ DMA_RXOR12 << DMA_CUED_REGION_OFF);
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ }
+ } else if (cursor->addrl == addr + cursor->len) {
+ /* reverse RXOR */
+ cursor->state = 1;
+ cursor->xor_count++;
+ set_bit(cursor->addr_count, &desc->reverse_flags[0]);
+ if (index == src_cnt-1) {
+ ppc440spe_rxor_set_region(desc,
+ cursor->addr_count,
+ DMA_RXOR12 << DMA_CUED_REGION_OFF);
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ }
+ } else {
+ printk(KERN_ERR "Cannot build "
+ "DMA2 RXOR command block.\n");
+ BUG();
+ }
+ break;
+ case 1:
+ sign = test_bit(cursor->addr_count,
+ desc->reverse_flags)
+ ? -1 : 1;
+ if (index == src_cnt-2 || (sign == -1
+ && addr != cursor->addrl - 2*cursor->len)) {
+ cursor->state = 0;
+ cursor->xor_count = 1;
+ cursor->addrl = addr;
+ ppc440spe_rxor_set_region(desc,
+ cursor->addr_count,
+ DMA_RXOR12 << DMA_CUED_REGION_OFF);
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ } else if (addr == cursor->addrl + 2*sign*cursor->len) {
+ cursor->state = 2;
+ cursor->xor_count = 0;
+ ppc440spe_rxor_set_region(desc,
+ cursor->addr_count,
+ DMA_RXOR123 << DMA_CUED_REGION_OFF);
+ if (index == src_cnt-1) {
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ }
+ } else if (addr == cursor->addrl + 3*cursor->len) {
+ cursor->state = 2;
+ cursor->xor_count = 0;
+ ppc440spe_rxor_set_region(desc,
+ cursor->addr_count,
+ DMA_RXOR124 << DMA_CUED_REGION_OFF);
+ if (index == src_cnt-1) {
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ }
+ } else if (addr == cursor->addrl + 4*cursor->len) {
+ cursor->state = 2;
+ cursor->xor_count = 0;
+ ppc440spe_rxor_set_region(desc,
+ cursor->addr_count,
+ DMA_RXOR125 << DMA_CUED_REGION_OFF);
+ if (index == src_cnt-1) {
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ }
+ } else {
+ cursor->state = 0;
+ cursor->xor_count = 1;
+ cursor->addrl = addr;
+ ppc440spe_rxor_set_region(desc,
+ cursor->addr_count,
+ DMA_RXOR12 << DMA_CUED_REGION_OFF);
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ }
+ break;
+ case 2:
+ cursor->state = 0;
+ cursor->addrl = addr;
+ cursor->xor_count++;
+ if (index) {
+ ppc440spe_adma_dma2rxor_inc_addr(
+ desc, cursor, index, src_cnt);
+ }
+ break;
+ }
+
+ return rval;
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_set_src - set RXOR source address; it's assumed that
+ * ppc440spe_adma_dma2rxor_prep_src() has already done prior this call
+ */
+static void ppc440spe_adma_dma2rxor_set_src(
+ struct ppc440spe_adma_desc_slot *desc,
+ int index, dma_addr_t addr)
+{
+ struct xor_cb *xcb = desc->hw_desc;
+ int k = 0, op = 0, lop = 0;
+
+ /* get the RXOR operand which corresponds to index addr */
+ while (op <= index) {
+ lop = op;
+ if (k == XOR_MAX_OPS) {
+ k = 0;
+ desc = list_entry(desc->chain_node.next,
+ struct ppc440spe_adma_desc_slot, chain_node);
+ xcb = desc->hw_desc;
+
+ }
+ if ((xcb->ops[k++].h & (DMA_RXOR12 << DMA_CUED_REGION_OFF)) ==
+ (DMA_RXOR12 << DMA_CUED_REGION_OFF))
+ op += 2;
+ else
+ op += 3;
+ }
+
+ BUG_ON(k < 1);
+
+ if (test_bit(k-1, desc->reverse_flags)) {
+ /* reverse operand order; put last op in RXOR group */
+ if (index == op - 1)
+ ppc440spe_rxor_set_src(desc, k - 1, addr);
+ } else {
+ /* direct operand order; put first op in RXOR group */
+ if (index == lop)
+ ppc440spe_rxor_set_src(desc, k - 1, addr);
+ }
+}
+
+/**
+ * ppc440spe_adma_dma2rxor_set_mult - set RXOR multipliers; it's assumed that
+ * ppc440spe_adma_dma2rxor_prep_src() has already done prior this call
+ */
+static void ppc440spe_adma_dma2rxor_set_mult(
+ struct ppc440spe_adma_desc_slot *desc,
+ int index, u8 mult)
+{
+ struct xor_cb *xcb = desc->hw_desc;
+ int k = 0, op = 0, lop = 0;
+
+ /* get the RXOR operand which corresponds to index mult */
+ while (op <= index) {
+ lop = op;
+ if (k == XOR_MAX_OPS) {
+ k = 0;
+ desc = list_entry(desc->chain_node.next,
+ struct ppc440spe_adma_desc_slot,
+ chain_node);
+ xcb = desc->hw_desc;
+
+ }
+ if ((xcb->ops[k++].h & (DMA_RXOR12 << DMA_CUED_REGION_OFF)) ==
+ (DMA_RXOR12 << DMA_CUED_REGION_OFF))
+ op += 2;
+ else
+ op += 3;
+ }
+
+ BUG_ON(k < 1);
+ if (test_bit(k-1, desc->reverse_flags)) {
+ /* reverse order */
+ ppc440spe_rxor_set_mult(desc, k - 1, op - index - 1, mult);
+ } else {
+ /* direct order */
+ ppc440spe_rxor_set_mult(desc, k - 1, index - lop, mult);
+ }
+}
+
+/**
+ * ppc440spe_init_rxor_cursor -
+ */
+static void ppc440spe_init_rxor_cursor(struct ppc440spe_rxor *cursor)
+{
+ memset(cursor, 0, sizeof(struct ppc440spe_rxor));
+ cursor->state = 2;
+}
+
+/**
+ * ppc440spe_adma_pq_set_src_mult - set multiplication coefficient into
+ * descriptor for the PQXOR operation
+ */
+static void ppc440spe_adma_pq_set_src_mult(
+ struct ppc440spe_adma_desc_slot *sw_desc,
+ unsigned char mult, int index, int dst_pos)
+{
+ struct ppc440spe_adma_chan *chan;
+ u32 mult_idx, mult_dst;
+ struct ppc440spe_adma_desc_slot *iter = NULL, *iter1 = NULL;
+
+ chan = to_ppc440spe_adma_chan(sw_desc->async_tx.chan);
+
+ switch (chan->device->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ if (test_bit(PPC440SPE_DESC_RXOR, &sw_desc->flags)) {
+ int region = test_bit(PPC440SPE_DESC_RXOR12,
+ &sw_desc->flags) ? 2 : 3;
+
+ if (index < region) {
+ /* RXOR multipliers */
+ iter = ppc440spe_get_group_entry(sw_desc,
+ sw_desc->dst_cnt - 1);
+ if (sw_desc->dst_cnt == 2)
+ iter1 = ppc440spe_get_group_entry(
+ sw_desc, 0);
+
+ mult_idx = DMA_CUED_MULT1_OFF + (index << 3);
+ mult_dst = DMA_CDB_SG_SRC;
+ } else {
+ /* WXOR multiplier */
+ iter = ppc440spe_get_group_entry(sw_desc,
+ index - region +
+ sw_desc->dst_cnt);
+ mult_idx = DMA_CUED_MULT1_OFF;
+ mult_dst = dst_pos ? DMA_CDB_SG_DST2 :
+ DMA_CDB_SG_DST1;
+ }
+ } else {
+ int znum = 0;
+
+ /* WXOR-only;
+ * skip first slots with destinations (if ZERO_DST has
+ * place)
+ */
+ if (test_bit(PPC440SPE_ZERO_P, &sw_desc->flags))
+ znum++;
+ if (test_bit(PPC440SPE_ZERO_Q, &sw_desc->flags))
+ znum++;
+
+ iter = ppc440spe_get_group_entry(sw_desc, index + znum);
+ mult_idx = DMA_CUED_MULT1_OFF;
+ mult_dst = dst_pos ? DMA_CDB_SG_DST2 : DMA_CDB_SG_DST1;
+ }
+
+ if (likely(iter)) {
+ ppc440spe_desc_set_src_mult(iter, chan,
+ mult_idx, mult_dst, mult);
+
+ if (unlikely(iter1)) {
+ /* if we have two destinations for RXOR, then
+ * we've just set Q mult. Set-up P now.
+ */
+ ppc440spe_desc_set_src_mult(iter1, chan,
+ mult_idx, mult_dst, 1);
+ }
+
+ }
+ break;
+
+ case PPC440SPE_XOR_ID:
+ iter = sw_desc->group_head;
+ if (sw_desc->dst_cnt == 2) {
+ /* both P & Q calculations required; set P mult here */
+ ppc440spe_adma_dma2rxor_set_mult(iter, index, 1);
+
+ /* and then set Q mult */
+ iter = ppc440spe_get_group_entry(sw_desc,
+ sw_desc->descs_per_op);
+ }
+ ppc440spe_adma_dma2rxor_set_mult(iter, index, mult);
+ break;
+ }
+}
+
+/**
+ * ppc440spe_adma_free_chan_resources - free the resources allocated
+ */
+static void ppc440spe_adma_free_chan_resources(struct dma_chan *chan)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ struct ppc440spe_adma_desc_slot *iter, *_iter;
+ int in_use_descs = 0;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+ ppc440spe_adma_slot_cleanup(ppc440spe_chan);
+
+ spin_lock_bh(&ppc440spe_chan->lock);
+ list_for_each_entry_safe(iter, _iter, &ppc440spe_chan->chain,
+ chain_node) {
+ in_use_descs++;
+ list_del(&iter->chain_node);
+ }
+ list_for_each_entry_safe_reverse(iter, _iter,
+ &ppc440spe_chan->all_slots, slot_node) {
+ list_del(&iter->slot_node);
+ kfree(iter);
+ ppc440spe_chan->slots_allocated--;
+ }
+ ppc440spe_chan->last_used = NULL;
+
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d %s slots_allocated %d\n",
+ ppc440spe_chan->device->id,
+ __func__, ppc440spe_chan->slots_allocated);
+ spin_unlock_bh(&ppc440spe_chan->lock);
+
+ /* one is ok since we left it on there on purpose */
+ if (in_use_descs > 1)
+ printk(KERN_ERR "SPE: Freeing %d in use descriptors!\n",
+ in_use_descs - 1);
+}
+
+/**
+ * ppc440spe_adma_is_complete - poll the status of an ADMA transaction
+ * @chan: ADMA channel handle
+ * @cookie: ADMA transaction identifier
+ */
+static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
+ dma_cookie_t cookie, dma_cookie_t *done, dma_cookie_t *used)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+ dma_cookie_t last_used;
+ dma_cookie_t last_complete;
+ enum dma_status ret;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+ last_used = chan->cookie;
+ last_complete = ppc440spe_chan->completed_cookie;
+
+ if (done)
+ *done = last_complete;
+ if (used)
+ *used = last_used;
+
+ ret = dma_async_is_complete(cookie, last_complete, last_used);
+ if (ret == DMA_SUCCESS)
+ return ret;
+
+ ppc440spe_adma_slot_cleanup(ppc440spe_chan);
+
+ last_used = chan->cookie;
+ last_complete = ppc440spe_chan->completed_cookie;
+
+ if (done)
+ *done = last_complete;
+ if (used)
+ *used = last_used;
+
+ return dma_async_is_complete(cookie, last_complete, last_used);
+}
+
+/**
+ * ppc440spe_adma_eot_handler - end of transfer interrupt handler
+ */
+static irqreturn_t ppc440spe_adma_eot_handler(int irq, void *data)
+{
+ struct ppc440spe_adma_chan *chan = data;
+
+ dev_dbg(chan->device->common.dev,
+ "ppc440spe adma%d: %s\n", chan->device->id, __func__);
+
+ tasklet_schedule(&chan->irq_tasklet);
+ ppc440spe_adma_device_clear_eot_status(chan);
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * ppc440spe_adma_err_handler - DMA error interrupt handler;
+ * do the same things as a eot handler
+ */
+static irqreturn_t ppc440spe_adma_err_handler(int irq, void *data)
+{
+ struct ppc440spe_adma_chan *chan = data;
+
+ dev_dbg(chan->device->common.dev,
+ "ppc440spe adma%d: %s\n", chan->device->id, __func__);
+
+ tasklet_schedule(&chan->irq_tasklet);
+ ppc440spe_adma_device_clear_eot_status(chan);
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * ppc440spe_test_callback - called when test operation has been done
+ */
+static void ppc440spe_test_callback(void *unused)
+{
+ complete(&ppc440spe_r6_test_comp);
+}
+
+/**
+ * ppc440spe_adma_issue_pending - flush all pending descriptors to h/w
+ */
+static void ppc440spe_adma_issue_pending(struct dma_chan *chan)
+{
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+ dev_dbg(ppc440spe_chan->device->common.dev,
+ "ppc440spe adma%d: %s %d \n", ppc440spe_chan->device->id,
+ __func__, ppc440spe_chan->pending);
+
+ if (ppc440spe_chan->pending) {
+ ppc440spe_chan->pending = 0;
+ ppc440spe_chan_append(ppc440spe_chan);
+ }
+}
+
+/**
+ * ppc440spe_chan_start_null_xor - initiate the first XOR operation (DMA engines
+ * use FIFOs (as opposite to chains used in XOR) so this is a XOR
+ * specific operation)
+ */
+static void ppc440spe_chan_start_null_xor(struct ppc440spe_adma_chan *chan)
+{
+ struct ppc440spe_adma_desc_slot *sw_desc, *group_start;
+ dma_cookie_t cookie;
+ int slot_cnt, slots_per_op;
+
+ dev_dbg(chan->device->common.dev,
+ "ppc440spe adma%d: %s\n", chan->device->id, __func__);
+
+ spin_lock_bh(&chan->lock);
+ slot_cnt = ppc440spe_chan_xor_slot_count(0, 2, &slots_per_op);
+ sw_desc = ppc440spe_adma_alloc_slots(chan, slot_cnt, slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ list_splice_init(&sw_desc->group_list, &chan->chain);
+ async_tx_ack(&sw_desc->async_tx);
+ ppc440spe_desc_init_null_xor(group_start);
+
+ cookie = chan->common.cookie;
+ cookie++;
+ if (cookie <= 1)
+ cookie = 2;
+
+ /* initialize the completed cookie to be less than
+ * the most recently used cookie
+ */
+ chan->completed_cookie = cookie - 1;
+ chan->common.cookie = sw_desc->async_tx.cookie = cookie;
+
+ /* channel should not be busy */
+ BUG_ON(ppc440spe_chan_is_busy(chan));
+
+ /* set the descriptor address */
+ ppc440spe_chan_set_first_xor_descriptor(chan, sw_desc);
+
+ /* run the descriptor */
+ ppc440spe_chan_run(chan);
+ } else
+ printk(KERN_ERR "ppc440spe adma%d"
+ " failed to allocate null descriptor\n",
+ chan->device->id);
+ spin_unlock_bh(&chan->lock);
+}
+
+/**
+ * ppc440spe_test_raid6 - test are RAID-6 capabilities enabled successfully.
+ * For this we just perform one WXOR operation with the same source
+ * and destination addresses, the GF-multiplier is 1; so if RAID-6
+ * capabilities are enabled then we'll get src/dst filled with zero.
+ */
+static int ppc440spe_test_raid6(struct ppc440spe_adma_chan *chan)
+{
+ struct ppc440spe_adma_desc_slot *sw_desc, *iter;
+ struct page *pg;
+ char *a;
+ dma_addr_t dma_addr, addrs[2];
+ unsigned long op = 0;
+ int rval = 0;
+
+ set_bit(PPC440SPE_DESC_WXOR, &op);
+
+ pg = alloc_page(GFP_KERNEL);
+ if (!pg)
+ return -ENOMEM;
+
+ spin_lock_bh(&chan->lock);
+ sw_desc = ppc440spe_adma_alloc_slots(chan, 1, 1);
+ if (sw_desc) {
+ /* 1 src, 1 dsr, int_ena, WXOR */
+ ppc440spe_desc_init_dma01pq(sw_desc, 1, 1, 1, op);
+ list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+ ppc440spe_desc_set_byte_count(iter, chan, PAGE_SIZE);
+ iter->unmap_len = PAGE_SIZE;
+ }
+ } else {
+ rval = -EFAULT;
+ spin_unlock_bh(&chan->lock);
+ goto exit;
+ }
+ spin_unlock_bh(&chan->lock);
+
+ /* Fill the test page with ones */
+ memset(page_address(pg), 0xFF, PAGE_SIZE);
+ dma_addr = dma_map_page(chan->device->dev, pg, 0,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+
+ /* Setup addresses */
+ ppc440spe_adma_pq_set_src(sw_desc, dma_addr, 0);
+ ppc440spe_adma_pq_set_src_mult(sw_desc, 1, 0, 0);
+ addrs[0] = dma_addr;
+ addrs[1] = 0;
+ ppc440spe_adma_pq_set_dest(sw_desc, addrs, DMA_PREP_PQ_DISABLE_Q);
+
+ async_tx_ack(&sw_desc->async_tx);
+ sw_desc->async_tx.callback = ppc440spe_test_callback;
+ sw_desc->async_tx.callback_param = NULL;
+
+ init_completion(&ppc440spe_r6_test_comp);
+
+ ppc440spe_adma_tx_submit(&sw_desc->async_tx);
+ ppc440spe_adma_issue_pending(&chan->common);
+
+ wait_for_completion(&ppc440spe_r6_test_comp);
+
+ /* Now check if the test page is zeroed */
+ a = page_address(pg);
+ if ((*(u32 *)a) == 0 && memcmp(a, a+4, PAGE_SIZE-4) == 0) {
+ /* page is zero - RAID-6 enabled */
+ rval = 0;
+ } else {
+ /* RAID-6 was not enabled */
+ rval = -EINVAL;
+ }
+exit:
+ __free_page(pg);
+ return rval;
+}
+
+static void ppc440spe_adma_init_capabilities(struct ppc440spe_adma_device *adev)
+{
+ switch (adev->id) {
+ case PPC440SPE_DMA0_ID:
+ case PPC440SPE_DMA1_ID:
+ dma_cap_set(DMA_MEMCPY, adev->common.cap_mask);
+ dma_cap_set(DMA_INTERRUPT, adev->common.cap_mask);
+ dma_cap_set(DMA_MEMSET, adev->common.cap_mask);
+ dma_cap_set(DMA_PQ, adev->common.cap_mask);
+ dma_cap_set(DMA_PQ_VAL, adev->common.cap_mask);
+ dma_cap_set(DMA_XOR_VAL, adev->common.cap_mask);
+ break;
+ case PPC440SPE_XOR_ID:
+ dma_cap_set(DMA_XOR, adev->common.cap_mask);
+ dma_cap_set(DMA_PQ, adev->common.cap_mask);
+ dma_cap_set(DMA_INTERRUPT, adev->common.cap_mask);
+ adev->common.cap_mask = adev->common.cap_mask;
+ break;
+ }
+
+ /* Set base routines */
+ adev->common.device_alloc_chan_resources =
+ ppc440spe_adma_alloc_chan_resources;
+ adev->common.device_free_chan_resources =
+ ppc440spe_adma_free_chan_resources;
+ adev->common.device_is_tx_complete = ppc440spe_adma_is_complete;
+ adev->common.device_issue_pending = ppc440spe_adma_issue_pending;
+
+ /* Set prep routines based on capability */
+ if (dma_has_cap(DMA_MEMCPY, adev->common.cap_mask)) {
+ adev->common.device_prep_dma_memcpy =
+ ppc440spe_adma_prep_dma_memcpy;
+ }
+ if (dma_has_cap(DMA_MEMSET, adev->common.cap_mask)) {
+ adev->common.device_prep_dma_memset =
+ ppc440spe_adma_prep_dma_memset;
+ }
+ if (dma_has_cap(DMA_XOR, adev->common.cap_mask)) {
+ adev->common.max_xor = XOR_MAX_OPS;
+ adev->common.device_prep_dma_xor =
+ ppc440spe_adma_prep_dma_xor;
+ }
+ if (dma_has_cap(DMA_PQ, adev->common.cap_mask)) {
+ switch (adev->id) {
+ case PPC440SPE_DMA0_ID:
+ dma_set_maxpq(&adev->common,
+ DMA0_FIFO_SIZE / sizeof(struct dma_cdb), 0);
+ break;
+ case PPC440SPE_DMA1_ID:
+ dma_set_maxpq(&adev->common,
+ DMA1_FIFO_SIZE / sizeof(struct dma_cdb), 0);
+ break;
+ case PPC440SPE_XOR_ID:
+ adev->common.max_pq = XOR_MAX_OPS * 3;
+ break;
+ }
+ adev->common.device_prep_dma_pq =
+ ppc440spe_adma_prep_dma_pq;
+ }
+ if (dma_has_cap(DMA_PQ_VAL, adev->common.cap_mask)) {
+ switch (adev->id) {
+ case PPC440SPE_DMA0_ID:
+ adev->common.max_pq = DMA0_FIFO_SIZE /
+ sizeof(struct dma_cdb);
+ break;
+ case PPC440SPE_DMA1_ID:
+ adev->common.max_pq = DMA1_FIFO_SIZE /
+ sizeof(struct dma_cdb);
+ break;
+ }
+ adev->common.device_prep_dma_pq_val =
+ ppc440spe_adma_prep_dma_pqzero_sum;
+ }
+ if (dma_has_cap(DMA_XOR_VAL, adev->common.cap_mask)) {
+ switch (adev->id) {
+ case PPC440SPE_DMA0_ID:
+ adev->common.max_xor = DMA0_FIFO_SIZE /
+ sizeof(struct dma_cdb);
+ break;
+ case PPC440SPE_DMA1_ID:
+ adev->common.max_xor = DMA1_FIFO_SIZE /
+ sizeof(struct dma_cdb);
+ break;
+ }
+ adev->common.device_prep_dma_xor_val =
+ ppc440spe_adma_prep_dma_xor_zero_sum;
+ }
+ if (dma_has_cap(DMA_INTERRUPT, adev->common.cap_mask)) {
+ adev->common.device_prep_dma_interrupt =
+ ppc440spe_adma_prep_dma_interrupt;
+ }
+ pr_info("%s: AMCC(R) PPC440SP(E) ADMA Engine: "
+ "( %s%s%s%s%s%s%s)\n",
+ dev_name(adev->dev),
+ dma_has_cap(DMA_PQ, adev->common.cap_mask) ? "pq " : "",
+ dma_has_cap(DMA_PQ_VAL, adev->common.cap_mask) ? "pq_val " : "",
+ dma_has_cap(DMA_XOR, adev->common.cap_mask) ? "xor " : "",
+ dma_has_cap(DMA_XOR_VAL, adev->common.cap_mask) ? "xor_val " : "",
+ dma_has_cap(DMA_MEMCPY, adev->common.cap_mask) ? "memcpy " : "",
+ dma_has_cap(DMA_MEMSET, adev->common.cap_mask) ? "memset " : "",
+ dma_has_cap(DMA_INTERRUPT, adev->common.cap_mask) ? "intr " : "");
+}
+
+static int ppc440spe_adma_setup_irqs(struct ppc440spe_adma_device *adev,
+ struct ppc440spe_adma_chan *chan,
+ int *initcode)
+{
+ struct device_node *np;
+ int ret;
+
+ np = container_of(adev->dev, struct of_device, dev)->node;
+ if (adev->id != PPC440SPE_XOR_ID) {
+ adev->err_irq = irq_of_parse_and_map(np, 1);
+ if (adev->err_irq == NO_IRQ) {
+ dev_warn(adev->dev, "no err irq resource?\n");
+ *initcode = PPC_ADMA_INIT_IRQ2;
+ adev->err_irq = -ENXIO;
+ } else
+ atomic_inc(&ppc440spe_adma_err_irq_ref);
+ } else {
+ adev->err_irq = -ENXIO;
+ }
+
+ adev->irq = irq_of_parse_and_map(np, 0);
+ if (adev->irq == NO_IRQ) {
+ dev_err(adev->dev, "no irq resource\n");
+ *initcode = PPC_ADMA_INIT_IRQ1;
+ ret = -ENXIO;
+ goto err_irq_map;
+ }
+ dev_dbg(adev->dev, "irq %d, err irq %d\n",
+ adev->irq, adev->err_irq);
+
+ ret = request_irq(adev->irq, ppc440spe_adma_eot_handler,
+ 0, dev_driver_string(adev->dev), chan);
+ if (ret) {
+ dev_err(adev->dev, "can't request irq %d\n",
+ adev->irq);
+ *initcode = PPC_ADMA_INIT_IRQ1;
+ ret = -EIO;
+ goto err_req1;
+ }
+
+ /* only DMA engines have a separate error IRQ
+ * so it's Ok if err_irq < 0 in XOR engine case.
+ */
+ if (adev->err_irq > 0) {
+ /* both DMA engines share common error IRQ */
+ ret = request_irq(adev->err_irq,
+ ppc440spe_adma_err_handler,
+ IRQF_SHARED,
+ dev_driver_string(adev->dev),
+ chan);
+ if (ret) {
+ dev_err(adev->dev, "can't request irq %d\n",
+ adev->err_irq);
+ *initcode = PPC_ADMA_INIT_IRQ2;
+ ret = -EIO;
+ goto err_req2;
+ }
+ }
+
+ if (adev->id == PPC440SPE_XOR_ID) {
+ /* enable XOR engine interrupts */
+ iowrite32be(XOR_IE_CBCIE_BIT | XOR_IE_ICBIE_BIT |
+ XOR_IE_ICIE_BIT | XOR_IE_RPTIE_BIT,
+ &adev->xor_reg->ier);
+ } else {
+ u32 mask, enable;
+
+ np = of_find_compatible_node(NULL, NULL, "ibm,i2o-440spe");
+ if (!np) {
+ pr_err("%s: can't find I2O device tree node\n",
+ __func__);
+ ret = -ENODEV;
+ goto err_req2;
+ }
+ adev->i2o_reg = of_iomap(np, 0);
+ if (!adev->i2o_reg) {
+ pr_err("%s: failed to map I2O registers\n", __func__);
+ of_node_put(np);
+ ret = -EINVAL;
+ goto err_req2;
+ }
+ of_node_put(np);
+ /* Unmask 'CS FIFO Attention' interrupts and
+ * enable generating interrupts on errors
+ */
+ enable = (adev->id == PPC440SPE_DMA0_ID) ?
+ ~(I2O_IOPIM_P0SNE | I2O_IOPIM_P0EM) :
+ ~(I2O_IOPIM_P1SNE | I2O_IOPIM_P1EM);
+ mask = ioread32(&adev->i2o_reg->iopim) & enable;
+ iowrite32(mask, &adev->i2o_reg->iopim);
+ }
+ return 0;
+
+err_req2:
+ free_irq(adev->irq, chan);
+err_req1:
+ irq_dispose_mapping(adev->irq);
+err_irq_map:
+ if (adev->err_irq > 0) {
+ if (atomic_dec_and_test(&ppc440spe_adma_err_irq_ref))
+ irq_dispose_mapping(adev->err_irq);
+ }
+ return ret;
+}
+
+static void ppc440spe_adma_release_irqs(struct ppc440spe_adma_device *adev,
+ struct ppc440spe_adma_chan *chan)
+{
+ u32 mask, disable;
+
+ if (adev->id == PPC440SPE_XOR_ID) {
+ /* disable XOR engine interrupts */
+ mask = ioread32be(&adev->xor_reg->ier);
+ mask &= ~(XOR_IE_CBCIE_BIT | XOR_IE_ICBIE_BIT |
+ XOR_IE_ICIE_BIT | XOR_IE_RPTIE_BIT);
+ iowrite32be(mask, &adev->xor_reg->ier);
+ } else {
+ /* disable DMAx engine interrupts */
+ disable = (adev->id == PPC440SPE_DMA0_ID) ?
+ (I2O_IOPIM_P0SNE | I2O_IOPIM_P0EM) :
+ (I2O_IOPIM_P1SNE | I2O_IOPIM_P1EM);
+ mask = ioread32(&adev->i2o_reg->iopim) | disable;
+ iowrite32(mask, &adev->i2o_reg->iopim);
+ }
+ free_irq(adev->irq, chan);
+ irq_dispose_mapping(adev->irq);
+ if (adev->err_irq > 0) {
+ free_irq(adev->err_irq, chan);
+ if (atomic_dec_and_test(&ppc440spe_adma_err_irq_ref)) {
+ irq_dispose_mapping(adev->err_irq);
+ iounmap(adev->i2o_reg);
+ }
+ }
+}
+
+/**
+ * ppc440spe_adma_probe - probe the asynch device
+ */
+static int __devinit ppc440spe_adma_probe(struct of_device *ofdev,
+ const struct of_device_id *match)
+{
+ struct device_node *np = ofdev->node;
+ struct resource res;
+ struct ppc440spe_adma_device *adev;
+ struct ppc440spe_adma_chan *chan;
+ struct ppc_dma_chan_ref *ref, *_ref;
+ int ret = 0, initcode = PPC_ADMA_INIT_OK;
+ const u32 *idx;
+ int len;
+ void *regs;
+ u32 id, pool_size;
+
+ if (of_device_is_compatible(np, "amcc,xor-accelerator")) {
+ id = PPC440SPE_XOR_ID;
+ /* As far as the XOR engine is concerned, it does not
+ * use FIFOs but uses linked list. So there is no dependency
+ * between pool size to allocate and the engine configuration.
+ */
+ pool_size = PAGE_SIZE << 1;
+ } else {
+ /* it is DMA0 or DMA1 */
+ idx = of_get_property(np, "cell-index", &len);
+ if (!idx || (len != sizeof(u32))) {
+ dev_err(&ofdev->dev, "Device node %s has missing "
+ "or invalid cell-index property\n",
+ np->full_name);
+ return -EINVAL;
+ }
+ id = *idx;
+ /* DMA0,1 engines use FIFO to maintain CDBs, so we
+ * should allocate the pool accordingly to size of this
+ * FIFO. Thus, the pool size depends on the FIFO depth:
+ * how much CDBs pointers the FIFO may contain then so
+ * much CDBs we should provide in the pool.
+ * That is
+ * CDB size = 32B;
+ * CDBs number = (DMA0_FIFO_SIZE >> 3);
+ * Pool size = CDBs number * CDB size =
+ * = (DMA0_FIFO_SIZE >> 3) << 5 = DMA0_FIFO_SIZE << 2.
+ */
+ pool_size = (id == PPC440SPE_DMA0_ID) ?
+ DMA0_FIFO_SIZE : DMA1_FIFO_SIZE;
+ pool_size <<= 2;
+ }
+
+ if (of_address_to_resource(np, 0, &res)) {
+ dev_err(&ofdev->dev, "failed to get memory resource\n");
+ initcode = PPC_ADMA_INIT_MEMRES;
+ ret = -ENODEV;
+ goto out;
+ }
+
+ if (!request_mem_region(res.start, resource_size(&res),
+ dev_driver_string(&ofdev->dev))) {
+ dev_err(&ofdev->dev, "failed to request memory region "
+ "(0x%016llx-0x%016llx)\n",
+ (u64)res.start, (u64)res.end);
+ initcode = PPC_ADMA_INIT_MEMREG;
+ ret = -EBUSY;
+ goto out;
+ }
+
+ /* create a device */
+ adev = kzalloc(sizeof(*adev), GFP_KERNEL);
+ if (!adev) {
+ dev_err(&ofdev->dev, "failed to allocate device\n");
+ initcode = PPC_ADMA_INIT_ALLOC;
+ ret = -ENOMEM;
+ goto err_adev_alloc;
+ }
+
+ adev->id = id;
+ adev->pool_size = pool_size;
+ /* allocate coherent memory for hardware descriptors */
+ adev->dma_desc_pool_virt = dma_alloc_coherent(&ofdev->dev,
+ adev->pool_size, &adev->dma_desc_pool,
+ GFP_KERNEL);
+ if (adev->dma_desc_pool_virt == NULL) {
+ dev_err(&ofdev->dev, "failed to allocate %d bytes of coherent "
+ "memory for hardware descriptors\n",
+ adev->pool_size);
+ initcode = PPC_ADMA_INIT_COHERENT;
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+ dev_dbg(&ofdev->dev, "allocted descriptor pool virt 0x%p phys 0x%llx\n",
+ adev->dma_desc_pool_virt, (u64)adev->dma_desc_pool);
+
+ regs = ioremap(res.start, resource_size(&res));
+ if (!regs) {
+ dev_err(&ofdev->dev, "failed to ioremap regs!\n");
+ goto err_regs_alloc;
+ }
+
+ if (adev->id == PPC440SPE_XOR_ID) {
+ adev->xor_reg = regs;
+ /* Reset XOR */
+ iowrite32be(XOR_CRSR_XASR_BIT, &adev->xor_reg->crsr);
+ iowrite32be(XOR_CRSR_64BA_BIT, &adev->xor_reg->crrr);
+ } else {
+ size_t fifo_size = (adev->id == PPC440SPE_DMA0_ID) ?
+ DMA0_FIFO_SIZE : DMA1_FIFO_SIZE;
+ adev->dma_reg = regs;
+ /* DMAx_FIFO_SIZE is defined in bytes,
+ * <fsiz> - is defined in number of CDB pointers (8byte).
+ * DMA FIFO Length = CSlength + CPlength, where
+ * CSlength = CPlength = (fsiz + 1) * 8.
+ */
+ iowrite32(DMA_FIFO_ENABLE | ((fifo_size >> 3) - 2),
+ &adev->dma_reg->fsiz);
+ /* Configure DMA engine */
+ iowrite32(DMA_CFG_DXEPR_HP | DMA_CFG_DFMPP_HP | DMA_CFG_FALGN,
+ &adev->dma_reg->cfg);
+ /* Clear Status */
+ iowrite32(~0, &adev->dma_reg->dsts);
+ }
+
+ adev->dev = &ofdev->dev;
+ adev->common.dev = &ofdev->dev;
+ INIT_LIST_HEAD(&adev->common.channels);
+ dev_set_drvdata(&ofdev->dev, adev);
+
+ /* create a channel */
+ chan = kzalloc(sizeof(*chan), GFP_KERNEL);
+ if (!chan) {
+ dev_err(&ofdev->dev, "can't allocate channel structure\n");
+ initcode = PPC_ADMA_INIT_CHANNEL;
+ ret = -ENOMEM;
+ goto err_chan_alloc;
+ }
+
+ spin_lock_init(&chan->lock);
+ INIT_LIST_HEAD(&chan->chain);
+ INIT_LIST_HEAD(&chan->all_slots);
+ chan->device = adev;
+ chan->common.device = &adev->common;
+ list_add_tail(&chan->common.device_node, &adev->common.channels);
+ tasklet_init(&chan->irq_tasklet, ppc440spe_adma_tasklet,
+ (unsigned long)chan);
+
+ /* allocate and map helper pages for async validation or
+ * async_mult/async_sum_product operations on DMA0/1.
+ */
+ if (adev->id != PPC440SPE_XOR_ID) {
+ chan->pdest_page = alloc_page(GFP_KERNEL);
+ chan->qdest_page = alloc_page(GFP_KERNEL);
+ if (!chan->pdest_page ||
+ !chan->qdest_page) {
+ if (chan->pdest_page)
+ __free_page(chan->pdest_page);
+ if (chan->qdest_page)
+ __free_page(chan->qdest_page);
+ ret = -ENOMEM;
+ goto err_page_alloc;
+ }
+ chan->pdest = dma_map_page(&ofdev->dev, chan->pdest_page, 0,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ chan->qdest = dma_map_page(&ofdev->dev, chan->qdest_page, 0,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ }
+
+ ref = kmalloc(sizeof(*ref), GFP_KERNEL);
+ if (ref) {
+ ref->chan = &chan->common;
+ INIT_LIST_HEAD(&ref->node);
+ list_add_tail(&ref->node, &ppc440spe_adma_chan_list);
+ } else {
+ dev_err(&ofdev->dev, "failed to allocate channel reference!\n");
+ ret = -ENOMEM;
+ goto err_ref_alloc;
+ }
+
+ ret = ppc440spe_adma_setup_irqs(adev, chan, &initcode);
+ if (ret)
+ goto err_irq;
+
+ ppc440spe_adma_init_capabilities(adev);
+
+ ret = dma_async_device_register(&adev->common);
+ if (ret) {
+ initcode = PPC_ADMA_INIT_REGISTER;
+ dev_err(&ofdev->dev, "failed to register dma device\n");
+ goto err_dev_reg;
+ }
+
+ goto out;
+
+err_dev_reg:
+ ppc440spe_adma_release_irqs(adev, chan);
+err_irq:
+ list_for_each_entry_safe(ref, _ref, &ppc440spe_adma_chan_list, node) {
+ if (chan == to_ppc440spe_adma_chan(ref->chan)) {
+ list_del(&ref->node);
+ kfree(ref);
+ }
+ }
+err_ref_alloc:
+ if (adev->id != PPC440SPE_XOR_ID) {
+ dma_unmap_page(&ofdev->dev, chan->pdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ dma_unmap_page(&ofdev->dev, chan->qdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ __free_page(chan->pdest_page);
+ __free_page(chan->qdest_page);
+ }
+err_page_alloc:
+ kfree(chan);
+err_chan_alloc:
+ if (adev->id == PPC440SPE_XOR_ID)
+ iounmap(adev->xor_reg);
+ else
+ iounmap(adev->dma_reg);
+err_regs_alloc:
+ dma_free_coherent(adev->dev, adev->pool_size,
+ adev->dma_desc_pool_virt,
+ adev->dma_desc_pool);
+err_dma_alloc:
+ kfree(adev);
+err_adev_alloc:
+ release_mem_region(res.start, resource_size(&res));
+out:
+ if (id < PPC440SPE_ADMA_ENGINES_NUM)
+ ppc440spe_adma_devices[id] = initcode;
+
+ return ret;
+}
+
+/**
+ * ppc440spe_adma_remove - remove the asynch device
+ */
+static int __devexit ppc440spe_adma_remove(struct of_device *ofdev)
+{
+ struct ppc440spe_adma_device *adev = dev_get_drvdata(&ofdev->dev);
+ struct device_node *np = ofdev->node;
+ struct resource res;
+ struct dma_chan *chan, *_chan;
+ struct ppc_dma_chan_ref *ref, *_ref;
+ struct ppc440spe_adma_chan *ppc440spe_chan;
+
+ dev_set_drvdata(&ofdev->dev, NULL);
+ if (adev->id < PPC440SPE_ADMA_ENGINES_NUM)
+ ppc440spe_adma_devices[adev->id] = -1;
+
+ dma_async_device_unregister(&adev->common);
+
+ list_for_each_entry_safe(chan, _chan, &adev->common.channels,
+ device_node) {
+ ppc440spe_chan = to_ppc440spe_adma_chan(chan);
+ ppc440spe_adma_release_irqs(adev, ppc440spe_chan);
+ tasklet_kill(&ppc440spe_chan->irq_tasklet);
+ if (adev->id != PPC440SPE_XOR_ID) {
+ dma_unmap_page(&ofdev->dev, ppc440spe_chan->pdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ dma_unmap_page(&ofdev->dev, ppc440spe_chan->qdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ __free_page(ppc440spe_chan->pdest_page);
+ __free_page(ppc440spe_chan->qdest_page);
+ }
+ list_for_each_entry_safe(ref, _ref, &ppc440spe_adma_chan_list,
+ node) {
+ if (ppc440spe_chan ==
+ to_ppc440spe_adma_chan(ref->chan)) {
+ list_del(&ref->node);
+ kfree(ref);
+ }
+ }
+ list_del(&chan->device_node);
+ kfree(ppc440spe_chan);
+ }
+
+ dma_free_coherent(adev->dev, adev->pool_size,
+ adev->dma_desc_pool_virt, adev->dma_desc_pool);
+ if (adev->id == PPC440SPE_XOR_ID)
+ iounmap(adev->xor_reg);
+ else
+ iounmap(adev->dma_reg);
+ of_address_to_resource(np, 0, &res);
+ release_mem_region(res.start, resource_size(&res));
+ kfree(adev);
+ return 0;
+}
+
+/*
+ * /sys driver interface to enable h/w RAID-6 capabilities
+ * Files created in e.g. /sys/devices/plb.0/400100100.dma0/driver/
+ * directory are "devices", "enable" and "poly".
+ * "devices" shows available engines.
+ * "enable" is used to enable RAID-6 capabilities or to check
+ * whether these has been activated.
+ * "poly" allows setting/checking used polynomial (for PPC440SPe only).
+ */
+
+static ssize_t show_ppc440spe_devices(struct device_driver *dev, char *buf)
+{
+ ssize_t size = 0;
+ int i;
+
+ for (i = 0; i < PPC440SPE_ADMA_ENGINES_NUM; i++) {
+ if (ppc440spe_adma_devices[i] == -1)
+ continue;
+ size += snprintf(buf + size, PAGE_SIZE - size,
+ "PPC440SP(E)-ADMA.%d: %s\n", i,
+ ppc_adma_errors[ppc440spe_adma_devices[i]]);
+ }
+ return size;
+}
+
+static ssize_t show_ppc440spe_r6enable(struct device_driver *dev, char *buf)
+{
+ return snprintf(buf, PAGE_SIZE,
+ "PPC440SP(e) RAID-6 capabilities are %sABLED.\n",
+ ppc440spe_r6_enabled ? "EN" : "DIS");
+}
+
+static ssize_t store_ppc440spe_r6enable(struct device_driver *dev,
+ const char *buf, size_t count)
+{
+ unsigned long val;
+
+ if (!count || count > 11)
+ return -EINVAL;
+
+ if (!ppc440spe_r6_tchan)
+ return -EFAULT;
+
+ /* Write a key */
+ sscanf(buf, "%lx", &val);
+ dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_XORBA, val);
+ isync();
+
+ /* Verify whether it really works now */
+ if (ppc440spe_test_raid6(ppc440spe_r6_tchan) == 0) {
+ pr_info("PPC440SP(e) RAID-6 has been activated "
+ "successfully\n");
+ ppc440spe_r6_enabled = 1;
+ } else {
+ pr_info("PPC440SP(e) RAID-6 hasn't been activated!"
+ " Error key ?\n");
+ ppc440spe_r6_enabled = 0;
+ }
+ return count;
+}
+
+static ssize_t show_ppc440spe_r6poly(struct device_driver *dev, char *buf)
+{
+ ssize_t size = 0;
+ u32 reg;
+
+#ifdef CONFIG_440SP
+ /* 440SP has fixed polynomial */
+ reg = 0x4d;
+#else
+ reg = dcr_read(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL);
+ reg >>= MQ0_CFBHL_POLY;
+ reg &= 0xFF;
+#endif
+
+ size = snprintf(buf, PAGE_SIZE, "PPC440SP(e) RAID-6 driver "
+ "uses 0x1%02x polynomial.\n", reg);
+ return size;
+}
+
+static ssize_t store_ppc440spe_r6poly(struct device_driver *dev,
+ const char *buf, size_t count)
+{
+ unsigned long reg, val;
+
+#ifdef CONFIG_440SP
+ /* 440SP uses default 0x14D polynomial only */
+ return -EINVAL;
+#endif
+
+ if (!count || count > 6)
+ return -EINVAL;
+
+ /* e.g., 0x14D or 0x11D */
+ sscanf(buf, "%lx", &val);
+
+ if (val & ~0x1FF)
+ return -EINVAL;
+
+ val &= 0xFF;
+ reg = dcr_read(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL);
+ reg &= ~(0xFF << MQ0_CFBHL_POLY);
+ reg |= val << MQ0_CFBHL_POLY;
+ dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL, reg);
+
+ return count;
+}
+
+static DRIVER_ATTR(devices, S_IRUGO, show_ppc440spe_devices, NULL);
+static DRIVER_ATTR(enable, S_IRUGO | S_IWUSR, show_ppc440spe_r6enable,
+ store_ppc440spe_r6enable);
+static DRIVER_ATTR(poly, S_IRUGO | S_IWUSR, show_ppc440spe_r6poly,
+ store_ppc440spe_r6poly);
+
+/*
+ * Common initialisation for RAID engines; allocate memory for
+ * DMAx FIFOs, perform configuration common for all DMA engines.
+ * Further DMA engine specific configuration is done at probe time.
+ */
+static int ppc440spe_configure_raid_devices(void)
+{
+ struct device_node *np;
+ struct resource i2o_res;
+ struct i2o_regs __iomem *i2o_reg;
+ dcr_host_t i2o_dcr_host;
+ unsigned int dcr_base, dcr_len;
+ int i, ret;
+
+ np = of_find_compatible_node(NULL, NULL, "ibm,i2o-440spe");
+ if (!np) {
+ pr_err("%s: can't find I2O device tree node\n",
+ __func__);
+ return -ENODEV;
+ }
+
+ if (of_address_to_resource(np, 0, &i2o_res)) {
+ of_node_put(np);
+ return -EINVAL;
+ }
+
+ i2o_reg = of_iomap(np, 0);
+ if (!i2o_reg) {
+ pr_err("%s: failed to map I2O registers\n", __func__);
+ of_node_put(np);
+ return -EINVAL;
+ }
+
+ /* Get I2O DCRs base */
+ dcr_base = dcr_resource_start(np, 0);
+ dcr_len = dcr_resource_len(np, 0);
+ if (!dcr_base && !dcr_len) {
+ pr_err("%s: can't get DCR registers base/len!\n",
+ np->full_name);
+ of_node_put(np);
+ iounmap(i2o_reg);
+ return -ENODEV;
+ }
+
+ i2o_dcr_host = dcr_map(np, dcr_base, dcr_len);
+ if (!DCR_MAP_OK(i2o_dcr_host)) {
+ pr_err("%s: failed to map DCRs!\n", np->full_name);
+ of_node_put(np);
+ iounmap(i2o_reg);
+ return -ENODEV;
+ }
+ of_node_put(np);
+
+ /* Provide memory regions for DMA's FIFOs: I2O, DMA0 and DMA1 share
+ * the base address of FIFO memory space.
+ * Actually we need twice more physical memory than programmed in the
+ * <fsiz> register (because there are two FIFOs for each DMA: CP and CS)
+ */
+ ppc440spe_dma_fifo_buf = kmalloc((DMA0_FIFO_SIZE + DMA1_FIFO_SIZE) << 1,
+ GFP_KERNEL);
+ if (!ppc440spe_dma_fifo_buf) {
+ pr_err("%s: DMA FIFO buffer allocation failed.\n", __func__);
+ iounmap(i2o_reg);
+ dcr_unmap(i2o_dcr_host, dcr_len);
+ return -ENOMEM;
+ }
+
+ /*
+ * Configure h/w
+ */
+ /* Reset I2O/DMA */
+ mtdcri(SDR0, DCRN_SDR0_SRST, DCRN_SDR0_SRST_I2ODMA);
+ mtdcri(SDR0, DCRN_SDR0_SRST, 0);
+
+ /* Setup the base address of mmaped registers */
+ dcr_write(i2o_dcr_host, DCRN_I2O0_IBAH, (u32)(i2o_res.start >> 32));
+ dcr_write(i2o_dcr_host, DCRN_I2O0_IBAL, (u32)(i2o_res.start) |
+ I2O_REG_ENABLE);
+ dcr_unmap(i2o_dcr_host, dcr_len);
+
+ /* Setup FIFO memory space base address */
+ iowrite32(0, &i2o_reg->ifbah);
+ iowrite32(((u32)__pa(ppc440spe_dma_fifo_buf)), &i2o_reg->ifbal);
+
+ /* set zero FIFO size for I2O, so the whole
+ * ppc440spe_dma_fifo_buf is used by DMAs.
+ * DMAx_FIFOs will be configured while probe.
+ */
+ iowrite32(0, &i2o_reg->ifsiz);
+ iounmap(i2o_reg);
+
+ /* To prepare WXOR/RXOR functionality we need access to
+ * Memory Queue Module DCRs (finally it will be enabled
+ * via /sys interface of the ppc440spe ADMA driver).
+ */
+ np = of_find_compatible_node(NULL, NULL, "ibm,mq-440spe");
+ if (!np) {
+ pr_err("%s: can't find MQ device tree node\n",
+ __func__);
+ ret = -ENODEV;
+ goto out_free;
+ }
+
+ /* Get MQ DCRs base */
+ dcr_base = dcr_resource_start(np, 0);
+ dcr_len = dcr_resource_len(np, 0);
+ if (!dcr_base && !dcr_len) {
+ pr_err("%s: can't get DCR registers base/len!\n",
+ np->full_name);
+ ret = -ENODEV;
+ goto out_mq;
+ }
+
+ ppc440spe_mq_dcr_host = dcr_map(np, dcr_base, dcr_len);
+ if (!DCR_MAP_OK(ppc440spe_mq_dcr_host)) {
+ pr_err("%s: failed to map DCRs!\n", np->full_name);
+ ret = -ENODEV;
+ goto out_mq;
+ }
+ of_node_put(np);
+ ppc440spe_mq_dcr_len = dcr_len;
+
+ /* Set HB alias */
+ dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_BAUH, DMA_CUED_XOR_HB);
+
+ /* Set:
+ * - LL transaction passing limit to 1;
+ * - Memory controller cycle limit to 1;
+ * - Galois Polynomial to 0x14d (default)
+ */
+ dcr_write(ppc440spe_mq_dcr_host, DCRN_MQ0_CFBHL,
+ (1 << MQ0_CFBHL_TPLM) | (1 << MQ0_CFBHL_HBCL) |
+ (PPC440SPE_DEFAULT_POLY << MQ0_CFBHL_POLY));
+
+ atomic_set(&ppc440spe_adma_err_irq_ref, 0);
+ for (i = 0; i < PPC440SPE_ADMA_ENGINES_NUM; i++)
+ ppc440spe_adma_devices[i] = -1;
+
+ return 0;
+
+out_mq:
+ of_node_put(np);
+out_free:
+ kfree(ppc440spe_dma_fifo_buf);
+ return ret;
+}
+
+static struct of_device_id __devinitdata ppc440spe_adma_of_match[] = {
+ { .compatible = "ibm,dma-440spe", },
+ { .compatible = "amcc,xor-accelerator", },
+ {},
+};
+MODULE_DEVICE_TABLE(of, ppc440spe_adma_of_match);
+
+static struct of_platform_driver ppc440spe_adma_driver = {
+ .match_table = ppc440spe_adma_of_match,
+ .probe = ppc440spe_adma_probe,
+ .remove = __devexit_p(ppc440spe_adma_remove),
+ .driver = {
+ .name = "PPC440SP(E)-ADMA",
+ .owner = THIS_MODULE,
+ },
+};
+
+static __init int ppc440spe_adma_init(void)
+{
+ int ret;
+
+ ret = ppc440spe_configure_raid_devices();
+ if (ret)
+ return ret;
+
+ ret = of_register_platform_driver(&ppc440spe_adma_driver);
+ if (ret) {
+ pr_err("%s: failed to register platform driver\n",
+ __func__);
+ goto out_reg;
+ }
+
+ /* Initialization status */
+ ret = driver_create_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_devices);
+ if (ret)
+ goto out_dev;
+
+ /* RAID-6 h/w enable entry */
+ ret = driver_create_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_enable);
+ if (ret)
+ goto out_en;
+
+ /* GF polynomial to use */
+ ret = driver_create_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_poly);
+ if (!ret)
+ return ret;
+
+ driver_remove_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_enable);
+out_en:
+ driver_remove_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_devices);
+out_dev:
+ /* User will not be able to enable h/w RAID-6 */
+ pr_err("%s: failed to create RAID-6 driver interface\n",
+ __func__);
+ of_unregister_platform_driver(&ppc440spe_adma_driver);
+out_reg:
+ dcr_unmap(ppc440spe_mq_dcr_host, ppc440spe_mq_dcr_len);
+ kfree(ppc440spe_dma_fifo_buf);
+ return ret;
+}
+
+static void __exit ppc440spe_adma_exit(void)
+{
+ driver_remove_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_poly);
+ driver_remove_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_enable);
+ driver_remove_file(&ppc440spe_adma_driver.driver,
+ &driver_attr_devices);
+ of_unregister_platform_driver(&ppc440spe_adma_driver);
+ dcr_unmap(ppc440spe_mq_dcr_host, ppc440spe_mq_dcr_len);
+ kfree(ppc440spe_dma_fifo_buf);
+}
+
+arch_initcall(ppc440spe_adma_init);
+module_exit(ppc440spe_adma_exit);
+
+MODULE_AUTHOR("Yuri Tikhonov <yur@emcraft.com>");
+MODULE_DESCRIPTION("PPC440SPE ADMA Engine Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/dma/ppc4xx/adma.h b/drivers/dma/ppc4xx/adma.h
new file mode 100644
index 0000000..8ada5a8
--- /dev/null
+++ b/drivers/dma/ppc4xx/adma.h
@@ -0,0 +1,195 @@
+/*
+ * 2006-2009 (C) DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * This file is licensed under the terms of the GNU General Public License
+ * version 2. This program is licensed "as is" without any warranty of
+ * any kind, whether express or implied.
+ */
+
+#ifndef _PPC440SPE_ADMA_H
+#define _PPC440SPE_ADMA_H
+
+#include <linux/types.h>
+#include "dma.h"
+#include "xor.h"
+
+#define to_ppc440spe_adma_chan(chan) \
+ container_of(chan, struct ppc440spe_adma_chan, common)
+#define to_ppc440spe_adma_device(dev) \
+ container_of(dev, struct ppc440spe_adma_device, common)
+#define tx_to_ppc440spe_adma_slot(tx) \
+ container_of(tx, struct ppc440spe_adma_desc_slot, async_tx)
+
+/* Default polynomial (for 440SP is only available) */
+#define PPC440SPE_DEFAULT_POLY 0x4d
+
+#define PPC440SPE_ADMA_ENGINES_NUM (XOR_ENGINES_NUM + DMA_ENGINES_NUM)
+
+#define PPC440SPE_ADMA_WATCHDOG_MSEC 3
+#define PPC440SPE_ADMA_THRESHOLD 1
+
+#define PPC440SPE_DMA0_ID 0
+#define PPC440SPE_DMA1_ID 1
+#define PPC440SPE_XOR_ID 2
+
+#define PPC440SPE_ADMA_DMA_MAX_BYTE_COUNT 0xFFFFFFUL
+/* this is the XOR_CBBCR width */
+#define PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT (1 << 31)
+#define PPC440SPE_ADMA_ZERO_SUM_MAX_BYTE_COUNT PPC440SPE_ADMA_XOR_MAX_BYTE_COUNT
+
+#define PPC440SPE_RXOR_RUN 0
+
+#define MQ0_CF2H_RXOR_BS_MASK 0x1FF
+
+#undef ADMA_LL_DEBUG
+
+/**
+ * struct ppc440spe_adma_device - internal representation of an ADMA device
+ * @dev: device
+ * @dma_reg: base for DMAx register access
+ * @xor_reg: base for XOR register access
+ * @i2o_reg: base for I2O register access
+ * @id: HW ADMA Device selector
+ * @dma_desc_pool_virt: base of DMA descriptor region (CPU address)
+ * @dma_desc_pool: base of DMA descriptor region (DMA address)
+ * @pool_size: size of the pool
+ * @irq: DMAx or XOR irq number
+ * @err_irq: DMAx error irq number
+ * @common: embedded struct dma_device
+ */
+struct ppc440spe_adma_device {
+ struct device *dev;
+ struct dma_regs __iomem *dma_reg;
+ struct xor_regs __iomem *xor_reg;
+ struct i2o_regs __iomem *i2o_reg;
+ int id;
+ void *dma_desc_pool_virt;
+ dma_addr_t dma_desc_pool;
+ size_t pool_size;
+ int irq;
+ int err_irq;
+ struct dma_device common;
+};
+
+/**
+ * struct ppc440spe_adma_chan - internal representation of an ADMA channel
+ * @lock: serializes enqueue/dequeue operations to the slot pool
+ * @device: parent device
+ * @chain: device chain view of the descriptors
+ * @common: common dmaengine channel object members
+ * @all_slots: complete domain of slots usable by the channel
+ * @pending: allows batching of hardware operations
+ * @completed_cookie: identifier for the most recently completed operation
+ * @slots_allocated: records the actual size of the descriptor slot pool
+ * @hw_chain_inited: h/w descriptor chain initialization flag
+ * @irq_tasklet: bottom half where ppc440spe_adma_slot_cleanup runs
+ * @needs_unmap: if buffers should not be unmapped upon final processing
+ * @pdest_page: P destination page for async validate operation
+ * @qdest_page: Q destination page for async validate operation
+ * @pdest: P dma addr for async validate operation
+ * @qdest: Q dma addr for async validate operation
+ */
+struct ppc440spe_adma_chan {
+ spinlock_t lock;
+ struct ppc440spe_adma_device *device;
+ struct list_head chain;
+ struct dma_chan common;
+ struct list_head all_slots;
+ struct ppc440spe_adma_desc_slot *last_used;
+ int pending;
+ dma_cookie_t completed_cookie;
+ int slots_allocated;
+ int hw_chain_inited;
+ struct tasklet_struct irq_tasklet;
+ u8 needs_unmap;
+ struct page *pdest_page;
+ struct page *qdest_page;
+ dma_addr_t pdest;
+ dma_addr_t qdest;
+};
+
+struct ppc440spe_rxor {
+ u32 addrl;
+ u32 addrh;
+ int len;
+ int xor_count;
+ int addr_count;
+ int desc_count;
+ int state;
+};
+
+/**
+ * struct ppc440spe_adma_desc_slot - PPC440SPE-ADMA software descriptor
+ * @phys: hardware address of the hardware descriptor chain
+ * @group_head: first operation in a transaction
+ * @hw_next: pointer to the next descriptor in chain
+ * @async_tx: support for the async_tx api
+ * @slot_node: node on the iop_adma_chan.all_slots list
+ * @chain_node: node on the op_adma_chan.chain list
+ * @group_list: list of slots that make up a multi-descriptor transaction
+ * for example transfer lengths larger than the supported hw max
+ * @unmap_len: transaction bytecount
+ * @hw_desc: virtual address of the hardware descriptor chain
+ * @stride: currently chained or not
+ * @idx: pool index
+ * @slot_cnt: total slots used in an transaction (group of operations)
+ * @src_cnt: number of sources set in this descriptor
+ * @dst_cnt: number of destinations set in the descriptor
+ * @slots_per_op: number of slots per operation
+ * @descs_per_op: number of slot per P/Q operation see comment
+ * for ppc440spe_prep_dma_pqxor function
+ * @flags: desc state/type
+ * @reverse_flags: 1 if a corresponding rxor address uses reversed address order
+ * @xor_check_result: result of zero sum
+ * @crc32_result: result crc calculation
+ */
+struct ppc440spe_adma_desc_slot {
+ dma_addr_t phys;
+ struct ppc440spe_adma_desc_slot *group_head;
+ struct ppc440spe_adma_desc_slot *hw_next;
+ struct dma_async_tx_descriptor async_tx;
+ struct list_head slot_node;
+ struct list_head chain_node; /* node in channel ops list */
+ struct list_head group_list; /* list */
+ unsigned int unmap_len;
+ void *hw_desc;
+ u16 stride;
+ u16 idx;
+ u16 slot_cnt;
+ u8 src_cnt;
+ u8 dst_cnt;
+ u8 slots_per_op;
+ u8 descs_per_op;
+ unsigned long flags;
+ unsigned long reverse_flags[8];
+
+#define PPC440SPE_DESC_INT 0 /* generate interrupt on complete */
+#define PPC440SPE_ZERO_P 1 /* clear P destionaion */
+#define PPC440SPE_ZERO_Q 2 /* clear Q destination */
+#define PPC440SPE_COHERENT 3 /* src/dst are coherent */
+
+#define PPC440SPE_DESC_WXOR 4 /* WXORs are in chain */
+#define PPC440SPE_DESC_RXOR 5 /* RXOR is in chain */
+
+#define PPC440SPE_DESC_RXOR123 8 /* CDB for RXOR123 operation */
+#define PPC440SPE_DESC_RXOR124 9 /* CDB for RXOR124 operation */
+#define PPC440SPE_DESC_RXOR125 10 /* CDB for RXOR125 operation */
+#define PPC440SPE_DESC_RXOR12 11 /* CDB for RXOR12 operation */
+#define PPC440SPE_DESC_RXOR_REV 12 /* CDB has srcs in reversed order */
+
+#define PPC440SPE_DESC_PCHECK 13
+#define PPC440SPE_DESC_QCHECK 14
+
+#define PPC440SPE_DESC_RXOR_MSK 0x3
+
+ struct ppc440spe_rxor rxor_cursor;
+
+ union {
+ u32 *xor_check_result;
+ u32 *crc32_result;
+ };
+};
+
+#endif /* _PPC440SPE_ADMA_H */
diff --git a/drivers/dma/ppc4xx/dma.h b/drivers/dma/ppc4xx/dma.h
new file mode 100644
index 0000000..bcde2df
--- /dev/null
+++ b/drivers/dma/ppc4xx/dma.h
@@ -0,0 +1,223 @@
+/*
+ * 440SPe's DMA engines support header file
+ *
+ * 2006-2009 (C) DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * This file is licensed under the term of the GNU General Public License
+ * version 2. The program licensed "as is" without any warranty of any
+ * kind, whether express or implied.
+ */
+
+#ifndef _PPC440SPE_DMA_H
+#define _PPC440SPE_DMA_H
+
+#include <linux/types.h>
+
+/* Number of elements in the array with statical CDBs */
+#define MAX_STAT_DMA_CDBS 16
+/* Number of DMA engines available on the contoller */
+#define DMA_ENGINES_NUM 2
+
+/* Maximum h/w supported number of destinations */
+#define DMA_DEST_MAX_NUM 2
+
+/* FIFO's params */
+#define DMA0_FIFO_SIZE 0x1000
+#define DMA1_FIFO_SIZE 0x1000
+#define DMA_FIFO_ENABLE (1<<12)
+
+/* DMA Configuration Register. Data Transfer Engine PLB Priority: */
+#define DMA_CFG_DXEPR_LP (0<<26)
+#define DMA_CFG_DXEPR_HP (3<<26)
+#define DMA_CFG_DXEPR_HHP (2<<26)
+#define DMA_CFG_DXEPR_HHHP (1<<26)
+
+/* DMA Configuration Register. DMA FIFO Manager PLB Priority: */
+#define DMA_CFG_DFMPP_LP (0<<23)
+#define DMA_CFG_DFMPP_HP (3<<23)
+#define DMA_CFG_DFMPP_HHP (2<<23)
+#define DMA_CFG_DFMPP_HHHP (1<<23)
+
+/* DMA Configuration Register. Force 64-byte Alignment */
+#define DMA_CFG_FALGN (1 << 19)
+
+/*UIC0:*/
+#define D0CPF_INT (1<<12)
+#define D0CSF_INT (1<<11)
+#define D1CPF_INT (1<<10)
+#define D1CSF_INT (1<<9)
+/*UIC1:*/
+#define DMAE_INT (1<<9)
+
+/* I2O IOP Interrupt Mask Register */
+#define I2O_IOPIM_P0SNE (1<<3)
+#define I2O_IOPIM_P0EM (1<<5)
+#define I2O_IOPIM_P1SNE (1<<6)
+#define I2O_IOPIM_P1EM (1<<8)
+
+/* DMA CDB fields */
+#define DMA_CDB_MSK (0xF)
+#define DMA_CDB_64B_ADDR (1<<2)
+#define DMA_CDB_NO_INT (1<<3)
+#define DMA_CDB_STATUS_MSK (0x3)
+#define DMA_CDB_ADDR_MSK (0xFFFFFFF0)
+
+/* DMA CDB OpCodes */
+#define DMA_CDB_OPC_NO_OP (0x00)
+#define DMA_CDB_OPC_MV_SG1_SG2 (0x01)
+#define DMA_CDB_OPC_MULTICAST (0x05)
+#define DMA_CDB_OPC_DFILL128 (0x24)
+#define DMA_CDB_OPC_DCHECK128 (0x23)
+
+#define DMA_CUED_XOR_BASE (0x10000000)
+#define DMA_CUED_XOR_HB (0x00000008)
+
+#ifdef CONFIG_440SP
+#define DMA_CUED_MULT1_OFF 0
+#define DMA_CUED_MULT2_OFF 8
+#define DMA_CUED_MULT3_OFF 16
+#define DMA_CUED_REGION_OFF 24
+#define DMA_CUED_XOR_WIN_MSK (0xFC000000)
+#else
+#define DMA_CUED_MULT1_OFF 2
+#define DMA_CUED_MULT2_OFF 10
+#define DMA_CUED_MULT3_OFF 18
+#define DMA_CUED_REGION_OFF 26
+#define DMA_CUED_XOR_WIN_MSK (0xF0000000)
+#endif
+
+#define DMA_CUED_REGION_MSK 0x3
+#define DMA_RXOR123 0x0
+#define DMA_RXOR124 0x1
+#define DMA_RXOR125 0x2
+#define DMA_RXOR12 0x3
+
+/* S/G addresses */
+#define DMA_CDB_SG_SRC 1
+#define DMA_CDB_SG_DST1 2
+#define DMA_CDB_SG_DST2 3
+
+/*
+ * DMAx engines Command Descriptor Block Type
+ */
+struct dma_cdb {
+ /*
+ * Basic CDB structure (Table 20-17, p.499, 440spe_um_1_22.pdf)
+ */
+ u8 pad0[2]; /* reserved */
+ u8 attr; /* attributes */
+ u8 opc; /* opcode */
+ u32 sg1u; /* upper SG1 address */
+ u32 sg1l; /* lower SG1 address */
+ u32 cnt; /* SG count, 3B used */
+ u32 sg2u; /* upper SG2 address */
+ u32 sg2l; /* lower SG2 address */
+ u32 sg3u; /* upper SG3 address */
+ u32 sg3l; /* lower SG3 address */
+};
+
+/*
+ * DMAx hardware registers (p.515 in 440SPe UM 1.22)
+ */
+struct dma_regs {
+ u32 cpfpl;
+ u32 cpfph;
+ u32 csfpl;
+ u32 csfph;
+ u32 dsts;
+ u32 cfg;
+ u8 pad0[0x8];
+ u16 cpfhp;
+ u16 cpftp;
+ u16 csfhp;
+ u16 csftp;
+ u8 pad1[0x8];
+ u32 acpl;
+ u32 acph;
+ u32 s1bpl;
+ u32 s1bph;
+ u32 s2bpl;
+ u32 s2bph;
+ u32 s3bpl;
+ u32 s3bph;
+ u8 pad2[0x10];
+ u32 earl;
+ u32 earh;
+ u8 pad3[0x8];
+ u32 seat;
+ u32 sead;
+ u32 op;
+ u32 fsiz;
+};
+
+/*
+ * I2O hardware registers (p.528 in 440SPe UM 1.22)
+ */
+struct i2o_regs {
+ u32 ists;
+ u32 iseat;
+ u32 isead;
+ u8 pad0[0x14];
+ u32 idbel;
+ u8 pad1[0xc];
+ u32 ihis;
+ u32 ihim;
+ u8 pad2[0x8];
+ u32 ihiq;
+ u32 ihoq;
+ u8 pad3[0x8];
+ u32 iopis;
+ u32 iopim;
+ u32 iopiq;
+ u8 iopoq;
+ u8 pad4[3];
+ u16 iiflh;
+ u16 iiflt;
+ u16 iiplh;
+ u16 iiplt;
+ u16 ioflh;
+ u16 ioflt;
+ u16 ioplh;
+ u16 ioplt;
+ u32 iidc;
+ u32 ictl;
+ u32 ifcpp;
+ u8 pad5[0x4];
+ u16 mfac0;
+ u16 mfac1;
+ u16 mfac2;
+ u16 mfac3;
+ u16 mfac4;
+ u16 mfac5;
+ u16 mfac6;
+ u16 mfac7;
+ u16 ifcfh;
+ u16 ifcht;
+ u8 pad6[0x4];
+ u32 iifmc;
+ u32 iodb;
+ u32 iodbc;
+ u32 ifbal;
+ u32 ifbah;
+ u32 ifsiz;
+ u32 ispd0;
+ u32 ispd1;
+ u32 ispd2;
+ u32 ispd3;
+ u32 ihipl;
+ u32 ihiph;
+ u32 ihopl;
+ u32 ihoph;
+ u32 iiipl;
+ u32 iiiph;
+ u32 iiopl;
+ u32 iioph;
+ u32 ifcpl;
+ u32 ifcph;
+ u8 pad7[0x8];
+ u32 iopt;
+};
+
+#endif /* _PPC440SPE_DMA_H */
diff --git a/drivers/dma/ppc4xx/xor.h b/drivers/dma/ppc4xx/xor.h
new file mode 100644
index 0000000..daed738
--- /dev/null
+++ b/drivers/dma/ppc4xx/xor.h
@@ -0,0 +1,110 @@
+/*
+ * 440SPe's XOR engines support header file
+ *
+ * 2006-2009 (C) DENX Software Engineering.
+ *
+ * Author: Yuri Tikhonov <yur@emcraft.com>
+ *
+ * This file is licensed under the term of the GNU General Public License
+ * version 2. The program licensed "as is" without any warranty of any
+ * kind, whether express or implied.
+ */
+
+#ifndef _PPC440SPE_XOR_H
+#define _PPC440SPE_XOR_H
+
+#include <linux/types.h>
+
+/* Number of XOR engines available on the contoller */
+#define XOR_ENGINES_NUM 1
+
+/* Number of operands supported in the h/w */
+#define XOR_MAX_OPS 16
+
+/*
+ * XOR Command Block Control Register bits
+ */
+#define XOR_CBCR_LNK_BIT (1<<31) /* link present */
+#define XOR_CBCR_TGT_BIT (1<<30) /* target present */
+#define XOR_CBCR_CBCE_BIT (1<<29) /* command block compete enable */
+#define XOR_CBCR_RNZE_BIT (1<<28) /* result not zero enable */
+#define XOR_CBCR_XNOR_BIT (1<<15) /* XOR/XNOR */
+#define XOR_CDCR_OAC_MSK (0x7F) /* operand address count */
+
+/*
+ * XORCore Status Register bits
+ */
+#define XOR_SR_XCP_BIT (1<<31) /* core processing */
+#define XOR_SR_ICB_BIT (1<<17) /* invalid CB */
+#define XOR_SR_IC_BIT (1<<16) /* invalid command */
+#define XOR_SR_IPE_BIT (1<<15) /* internal parity error */
+#define XOR_SR_RNZ_BIT (1<<2) /* result not Zero */
+#define XOR_SR_CBC_BIT (1<<1) /* CB complete */
+#define XOR_SR_CBLC_BIT (1<<0) /* CB list complete */
+
+/*
+ * XORCore Control Set and Reset Register bits
+ */
+#define XOR_CRSR_XASR_BIT (1<<31) /* soft reset */
+#define XOR_CRSR_XAE_BIT (1<<30) /* enable */
+#define XOR_CRSR_RCBE_BIT (1<<29) /* refetch CB enable */
+#define XOR_CRSR_PAUS_BIT (1<<28) /* pause */
+#define XOR_CRSR_64BA_BIT (1<<27) /* 64/32 CB format */
+#define XOR_CRSR_CLP_BIT (1<<25) /* continue list processing */
+
+/*
+ * XORCore Interrupt Enable Register
+ */
+#define XOR_IE_ICBIE_BIT (1<<17) /* Invalid Command Block IRQ Enable */
+#define XOR_IE_ICIE_BIT (1<<16) /* Invalid Command IRQ Enable */
+#define XOR_IE_RPTIE_BIT (1<<14) /* Read PLB Timeout Error IRQ Enable */
+#define XOR_IE_CBCIE_BIT (1<<1) /* CB complete interrupt enable */
+#define XOR_IE_CBLCI_BIT (1<<0) /* CB list complete interrupt enable */
+
+/*
+ * XOR Accelerator engine Command Block Type
+ */
+struct xor_cb {
+ /*
+ * Basic 64-bit format XOR CB (Table 19-1, p.463, 440spe_um_1_22.pdf)
+ */
+ u32 cbc; /* control */
+ u32 cbbc; /* byte count */
+ u32 cbs; /* status */
+ u8 pad0[4]; /* reserved */
+ u32 cbtah; /* target address high */
+ u32 cbtal; /* target address low */
+ u32 cblah; /* link address high */
+ u32 cblal; /* link address low */
+ struct {
+ u32 h;
+ u32 l;
+ } __attribute__ ((packed)) ops[16];
+} __attribute__ ((packed));
+
+/*
+ * XOR hardware registers Table 19-3, UM 1.22
+ */
+struct xor_regs {
+ u32 op_ar[16][2]; /* operand address[0]-high,[1]-low registers */
+ u8 pad0[352]; /* reserved */
+ u32 cbcr; /* CB control register */
+ u32 cbbcr; /* CB byte count register */
+ u32 cbsr; /* CB status register */
+ u8 pad1[4]; /* reserved */
+ u32 cbtahr; /* operand target address high register */
+ u32 cbtalr; /* operand target address low register */
+ u32 cblahr; /* CB link address high register */
+ u32 cblalr; /* CB link address low register */
+ u32 crsr; /* control set register */
+ u32 crrr; /* control reset register */
+ u32 ccbahr; /* current CB address high register */
+ u32 ccbalr; /* current CB address low register */
+ u32 plbr; /* PLB configuration register */
+ u32 ier; /* interrupt enable register */
+ u32 pecr; /* parity error count register */
+ u32 sr; /* status register */
+ u32 revidr; /* revision ID register */
+};
+
+#endif /* _PPC440SPE_XOR_H */
--
1.6.2.5
^ permalink raw reply related
* Re: PCI interrupt question
From: David Hawkins @ 2009-12-03 17:11 UTC (permalink / raw)
To: Jeff Hane; +Cc: linuxppc-dev@ozlabs.org
In-Reply-To: <1259821316.12651.26.camel@qu102.quarc.com>
Hi Jeff,
> I'm trying to get interrupts working for my PCI device on a 460ex and
> am having problem. My ISR never triggers.
>
> I'm new to PCI(and ppc) and LDD said that I could read the config reg
> INTURRUPT_LINE to get the interrupt assigned to my PCI device. Well,
> this always reads zero.
>
> After reading through the code it appears that the interrupt is being
> assigned after reading some information out of the device tree and then
> filling in the irg in the pci_dev structure.
>
> I'm just looking for confirmation that I should be calling request_irq
> with the irq that I found in the pci_dev struct.
Can you clarify:
1. 460EX is your PCI host CPU?
2. You have some PCI device - what? Have you tested it works in a
standard PCI bus? Eg. ran lspci from an x86 host.
3. When your host processor boots, it assigns the PCI resources.
On an x86, its the BIOS that assigns the PCI addresses and IRQs
on the PCI devices. For your 460EX, it could be your bootloader,
or it could be your host OS (Linux).
4. If you boot Linux on your 460EX and run lspci, and the device
configuration space register for the IRQ line is 0, then your
host has not setup the PCI interface on your device correctly.
5. Once you have the PCI IRQ assigned to the device correctly,
you need to know which IRQ line on the host this corresponds to.
I'm pretty sure that needs to be described in the DTS as
commented by Stefan.
These comments might not be 100% correct, but the list of things
to check should be close enough for you to track down your
problem.
Cheers,
Dave
^ permalink raw reply
* BUG: Bad page map in process
From: Sean MacLennan @ 2009-12-03 19:18 UTC (permalink / raw)
To: linuxppc-dev
With 2.6.32 I sometimes get lots and lots of "Bad page map" errors as
shown below. I believe these started in 2.6.32-rc8 or possibly
2.6.32-rc7. Pika just switched to 2.6.31 so I have been concentrating
on that release, and not really testing the 2.6.32 stream.
I wish I could give more info, but I don't really even have the time to
write this email :( I am hoping somebody will have a eureka moment.
Cheers,
Sean
[ 147.410448] BUG: Bad page map in process udevd pte:ffffffffffffffff pmd:cd883000
[ 147.417989] addr:48003000 vm_flags:08000875 anon_vma:(null) mapping:cf4150ac index:3
[ 147.425837] vma->vm_ops->fault: filemap_fault+0x0/0x410
[ 147.431101] vma->vm_file->f_op->mmap: nfs_file_mmap+0x0/0x94
[ 147.436795] Call Trace:
[ 147.439268] [cf2dbd80] [c00065a4] show_stack+0x48/0x168 (unreliable)
[ 147.445708] [cf2dbdb0] [c0064ac8] print_bad_pte+0x14c/0x204
[ 147.451337] [cf2dbde0] [c0064cd4] vm_normal_page+0x90/0x98
[ 147.456875] [cf2dbdf0] [c0064ee8] unmap_vmas+0x20c/0x638
[ 147.462239] [cf2dbe80] [c0069194] exit_mmap+0xc8/0x170
[ 147.467441] [cf2dbea0] [c001caa4] mmput+0x50/0xf4
[ 147.472195] [cf2dbeb0] [c0020808] exit_mm+0x100/0x138
[ 147.477299] [cf2dbee0] [c00211b0] do_exit+0xb0/0x580
[ 147.482315] [cf2dbf20] [c00216c4] do_group_exit+0x44/0x9c
[ 147.487766] [cf2dbf30] [c0021730] sys_exit_group+0x14/0x28
[ 147.493308] [cf2dbf40] [c000dde4] ret_from_syscall+0x0/0x3c
^ permalink raw reply
* Re: PCI interrupt question
From: Jeff Hane @ 2009-12-03 19:39 UTC (permalink / raw)
To: linuxppc-dev@ozlabs.org
In-Reply-To: <4B17F14A.1060001@ovro.caltech.edu>
On Thu, 2009-12-03 at 09:11 -0800, David Hawkins wrote:
> Hi Jeff,
>
> > I'm trying to get interrupts working for my PCI device on a 460ex and
> > am having problem. My ISR never triggers.
> >
> > I'm new to PCI(and ppc) and LDD said that I could read the config reg
> > INTURRUPT_LINE to get the interrupt assigned to my PCI device. Well,
> > this always reads zero.
> >
> > After reading through the code it appears that the interrupt is being
> > assigned after reading some information out of the device tree and then
> > filling in the irg in the pci_dev structure.
> >
> > I'm just looking for confirmation that I should be calling request_irq
> > with the irq that I found in the pci_dev struct.
>
> Can you clarify:
>
> 1. 460EX is your PCI host CPU?
yes. We are using a canyonlands board with 460ex.
>
> 2. You have some PCI device - what? Have you tested it works in a
> standard PCI bus? Eg. ran lspci from an x86 host.
It is a FPGA board that we are using to develop a video codec/SOC. We
have not connected it an x86; however, we know the the PCI interface is
working because we can talk to the FPGA board.
>
> 3. When your host processor boots, it assigns the PCI resources.
> On an x86, its the BIOS that assigns the PCI addresses and IRQs
> on the PCI devices. For your 460EX, it could be your bootloader,
> or it could be your host OS (Linux).
We are using the vanilla versions of u-boot and linux for the
canyonlands boards and have made few changes and definitely not to the
PCI setup.
>
> 4. If you boot Linux on your 460EX and run lspci, and the device
> configuration space register for the IRQ line is 0, then your
> host has not setup the PCI interface on your device correctly.
>
> 5. Once you have the PCI IRQ assigned to the device correctly,
> you need to know which IRQ line on the host this corresponds to.
> I'm pretty sure that needs to be described in the DTS as
> commented by Stefan.
>
So are you saying linux should be writing the irq number to the
INTERRUPT_LINE config reg? This is what I expected but I do not see
it.
I believe the DTS is being parsed properly and the connection is made
to the correct interrupt line on the device. But somebody still needs
to assign and IRQ number, right? This is the part that is not clear,
there is an irq field in pci_dev structure which is filled in after
looking at the DTS and I just want to be sure this is the irq number to
be used when calling request_irq.
thanks,
jeff
> These comments might not be 100% correct, but the list of things
> to check should be close enough for you to track down your
> problem.
>
> Cheers,
> Dave
>
^ permalink raw reply
* Re: PCI interrupt question
From: David Hawkins @ 2009-12-03 19:52 UTC (permalink / raw)
To: Jeff Hane; +Cc: linuxppc-dev@ozlabs.org
In-Reply-To: <1259869140.18190.52.camel@qu102.quarc.com>
Hi Jeff,
> yes. We are using a canyonlands board with 460ex.
Ok.
>> 2. You have some PCI device - what? Have you tested it works in a
>> standard PCI bus? Eg. ran lspci from an x86 host.
>
> It is a FPGA board that we are using to develop a video codec/SOC. We
> have not connected it an x86; however, we know the the PCI interface is
> working because we can talk to the FPGA board.
Ok, so that means the BARs are setup.
>> 3. When your host processor boots, it assigns the PCI resources.
>> On an x86, its the BIOS that assigns the PCI addresses and IRQs
>> on the PCI devices. For your 460EX, it could be your bootloader,
>> or it could be your host OS (Linux).
>
> We are using the vanilla versions of u-boot and linux for the
> canyonlands boards and have made few changes and definitely not to the
> PCI setup.
>
>> 4. If you boot Linux on your 460EX and run lspci, and the device
>> configuration space register for the IRQ line is 0, then your
>> host has not setup the PCI interface on your device correctly.
So what does the output of lspci look like?
>> 5. Once you have the PCI IRQ assigned to the device correctly,
>> you need to know which IRQ line on the host this corresponds to.
>> I'm pretty sure that needs to be described in the DTS as
>> commented by Stefan.
>>
>
> So are you saying linux should be writing the irq number to the
> INTERRUPT_LINE config reg? This is what I expected but I do not see
> it.
Yes. U-Boot or Linux needs to configure the PCI interface.
The output of lspci, or a logic analyzer trace of the PCI
activity on boot would help tell you what is happening.
> I believe the DTS is being parsed properly and the connection is made
> to the correct interrupt line on the device.
Right, so if someone asserted the interrupt the host would
understand it was from the PCI system.
> But somebody still needs to assign and IRQ number, right?
Yep. It might require enabling a PCI subsystem in U-Boot.
I recall playing with a Yosemite board and a PCI card,
but I forget whether I had to do anything to enable
PCI configuration space setup. Post a question to the
U-Boot list.
> This is the part that is not clear, there is an irq field in
> pci_dev structure which is filled in after looking at the
> DTS and I just want to be sure this is the irq number to
> be used when calling request_irq.
Really? I thought the pci_dev structures represent the
devices found on the PCI bus, and the IRQ line in
that structure was merely copied from the configuration
space registers. When you request that interrupt, the
code would have to remap it to a host IRQ line, which
it would use the DTS for. (But I'm just speculating).
Bottom line is; if the IRQ field of lspci is 0, then you
need to figure out that problem first :)
Cheers,
Dave
^ permalink raw reply
* Re: PCI interrupt question
From: Ira W. Snyder @ 2009-12-03 19:59 UTC (permalink / raw)
To: David Hawkins; +Cc: Jeff Hane, linuxppc-dev@ozlabs.org
In-Reply-To: <4B1816F9.1020601@ovro.caltech.edu>
On Thu, Dec 03, 2009 at 11:52:25AM -0800, David Hawkins wrote:
> Hi Jeff,
>
> > yes. We are using a canyonlands board with 460ex.
>
> Ok.
>
> >> 2. You have some PCI device - what? Have you tested it works in a
> >> standard PCI bus? Eg. ran lspci from an x86 host.
> >
> > It is a FPGA board that we are using to develop a video codec/SOC. We
> > have not connected it an x86; however, we know the the PCI interface is
> > working because we can talk to the FPGA board.
>
> Ok, so that means the BARs are setup.
>
> >> 3. When your host processor boots, it assigns the PCI resources.
> >> On an x86, its the BIOS that assigns the PCI addresses and IRQs
> >> on the PCI devices. For your 460EX, it could be your bootloader,
> >> or it could be your host OS (Linux).
> >
> > We are using the vanilla versions of u-boot and linux for the
> > canyonlands boards and have made few changes and definitely not to the
> > PCI setup.
> >
> >> 4. If you boot Linux on your 460EX and run lspci, and the device
> >> configuration space register for the IRQ line is 0, then your
> >> host has not setup the PCI interface on your device correctly.
>
> So what does the output of lspci look like?
>
> >> 5. Once you have the PCI IRQ assigned to the device correctly,
> >> you need to know which IRQ line on the host this corresponds to.
> >> I'm pretty sure that needs to be described in the DTS as
> >> commented by Stefan.
> >>
> >
> > So are you saying linux should be writing the irq number to the
> > INTERRUPT_LINE config reg? This is what I expected but I do not see
> > it.
>
> Yes. U-Boot or Linux needs to configure the PCI interface.
> The output of lspci, or a logic analyzer trace of the PCI
> activity on boot would help tell you what is happening.
>
> > I believe the DTS is being parsed properly and the connection is made
> > to the correct interrupt line on the device.
>
> Right, so if someone asserted the interrupt the host would
> understand it was from the PCI system.
>
> > But somebody still needs to assign and IRQ number, right?
>
> Yep. It might require enabling a PCI subsystem in U-Boot.
> I recall playing with a Yosemite board and a PCI card,
> but I forget whether I had to do anything to enable
> PCI configuration space setup. Post a question to the
> U-Boot list.
>
> > This is the part that is not clear, there is an irq field in
> > pci_dev structure which is filled in after looking at the
> > DTS and I just want to be sure this is the irq number to
> > be used when calling request_irq.
>
> Really? I thought the pci_dev structures represent the
> devices found on the PCI bus, and the IRQ line in
> that structure was merely copied from the configuration
> space registers. When you request that interrupt, the
> code would have to remap it to a host IRQ line, which
> it would use the DTS for. (But I'm just speculating).
>
The DTS shouldn't need to specify PCI cards, only the PCI controller.
Linux can probe the PCI bus to figure out what is plugged in, just like
it does on x86.
When Linux is assigning BARs and IRQs, it writes the BAR addresses to
the PCI cards, but does not write the IRQ. This was my experience with
PCI hotplug.
> Bottom line is; if the IRQ field of lspci is 0, then you
> need to figure out that problem first :)
>
In the PCI hotplug case mentioned above, lspci would show the IRQ as 0
until I called request_irq() in the kernel. Afterwards, the correct IRQ
showed up in the lspci output.
Try request_irq(pdev->irq, ....) and see if it works.
Ira
^ permalink raw reply
* Re: PCI interrupt question
From: David Hawkins @ 2009-12-03 20:03 UTC (permalink / raw)
To: Ira W. Snyder; +Cc: Jeff Hane, linuxppc-dev@ozlabs.org
In-Reply-To: <20091203195908.GA17100@ovro.caltech.edu>
Hi Ira,
> When Linux is assigning BARs and IRQs, it writes the BAR addresses to
> the PCI cards, but does not write the IRQ. This was my experience with
> PCI hotplug.
Ooh, subtle.
>> Bottom line is; if the IRQ field of lspci is 0, then you
>> need to figure out that problem first :)
>>
>
> In the PCI hotplug case mentioned above, lspci would show the IRQ as 0
> until I called request_irq() in the kernel. Afterwards, the correct IRQ
> showed up in the lspci output.
>
> Try request_irq(pdev->irq, ....) and see if it works.
You can also try lscpi -xxx, and setpci for writing the
registers. So try cheating and writing to the IRQ register
in the PCI interface.
Another possible problem, you may have your FPGA PCI core
configured in host mode (which can still be a master or target),
and then the IRQ register can not be written.
Cheers,
Dave
^ permalink raw reply
* Re: BUG: Bad page map in process
From: Jake Magee @ 2009-12-03 20:06 UTC (permalink / raw)
To: Sean MacLennan; +Cc: linuxppc-dev
In-Reply-To: <20091203141853.3eafba47@lappy.seanm.ca>
Have you tried the following patches? I'm not sure if they have made
it into the newer releases.
http://www.mail-archive.com/linuxppc-dev@lists.ozlabs.org/msg37188.html
Also for reference... http://patchwork.ozlabs.org/patch/34047/
On Thu, Dec 3, 2009 at 1:18 PM, Sean MacLennan <smaclennan@pikatech.com> wr=
ote:
> With 2.6.32 I sometimes get lots and lots of "Bad page map" errors as
> shown below. I believe these started in 2.6.32-rc8 or possibly
> 2.6.32-rc7. Pika just switched to 2.6.31 so I have been concentrating
> on that release, and not really testing the 2.6.32 stream.
>
> I wish I could give more info, but I don't really even have the time to
> write this email :( I am hoping somebody will have a eureka moment.
>
> Cheers,
> =A0 Sean
>
>
> [ =A0147.410448] BUG: Bad page map in process udevd =A0pte:ffffffffffffff=
ff pmd:cd883000
> [ =A0147.417989] addr:48003000 vm_flags:08000875 anon_vma:(null) mapping:=
cf4150ac index:3
> [ =A0147.425837] vma->vm_ops->fault: filemap_fault+0x0/0x410
> [ =A0147.431101] vma->vm_file->f_op->mmap: nfs_file_mmap+0x0/0x94
> [ =A0147.436795] Call Trace:
> [ =A0147.439268] [cf2dbd80] [c00065a4] show_stack+0x48/0x168 (unreliable)
> [ =A0147.445708] [cf2dbdb0] [c0064ac8] print_bad_pte+0x14c/0x204
> [ =A0147.451337] [cf2dbde0] [c0064cd4] vm_normal_page+0x90/0x98
> [ =A0147.456875] [cf2dbdf0] [c0064ee8] unmap_vmas+0x20c/0x638
> [ =A0147.462239] [cf2dbe80] [c0069194] exit_mmap+0xc8/0x170
> [ =A0147.467441] [cf2dbea0] [c001caa4] mmput+0x50/0xf4
> [ =A0147.472195] [cf2dbeb0] [c0020808] exit_mm+0x100/0x138
> [ =A0147.477299] [cf2dbee0] [c00211b0] do_exit+0xb0/0x580
> [ =A0147.482315] [cf2dbf20] [c00216c4] do_group_exit+0x44/0x9c
> [ =A0147.487766] [cf2dbf30] [c0021730] sys_exit_group+0x14/0x28
> [ =A0147.493308] [cf2dbf40] [c000dde4] ret_from_syscall+0x0/0x3c
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/linuxppc-dev
>
^ permalink raw reply
* Re: PCI interrupt question
From: Ira W. Snyder @ 2009-12-03 19:44 UTC (permalink / raw)
To: Jeff Hane; +Cc: linuxppc-dev@ozlabs.org
In-Reply-To: <1259869140.18190.52.camel@qu102.quarc.com>
On Thu, Dec 03, 2009 at 11:39:00AM -0800, Jeff Hane wrote:
> On Thu, 2009-12-03 at 09:11 -0800, David Hawkins wrote:
> > Hi Jeff,
> >
> > > I'm trying to get interrupts working for my PCI device on a 460ex and
> > > am having problem. My ISR never triggers.
> > >
> > > I'm new to PCI(and ppc) and LDD said that I could read the config reg
> > > INTURRUPT_LINE to get the interrupt assigned to my PCI device. Well,
> > > this always reads zero.
> > >
> > > After reading through the code it appears that the interrupt is being
> > > assigned after reading some information out of the device tree and then
> > > filling in the irg in the pci_dev structure.
> > >
> > > I'm just looking for confirmation that I should be calling request_irq
> > > with the irq that I found in the pci_dev struct.
> >
> > Can you clarify:
> >
> > 1. 460EX is your PCI host CPU?
>
> yes. We are using a canyonlands board with 460ex.
>
> >
> > 2. You have some PCI device - what? Have you tested it works in a
> > standard PCI bus? Eg. ran lspci from an x86 host.
>
> It is a FPGA board that we are using to develop a video codec/SOC. We
> have not connected it an x86; however, we know the the PCI interface is
> working because we can talk to the FPGA board.
>
> >
> > 3. When your host processor boots, it assigns the PCI resources.
> > On an x86, its the BIOS that assigns the PCI addresses and IRQs
> > on the PCI devices. For your 460EX, it could be your bootloader,
> > or it could be your host OS (Linux).
>
> We are using the vanilla versions of u-boot and linux for the
> canyonlands boards and have made few changes and definitely not to the
> PCI setup.
>
> >
> > 4. If you boot Linux on your 460EX and run lspci, and the device
> > configuration space register for the IRQ line is 0, then your
> > host has not setup the PCI interface on your device correctly.
> >
> > 5. Once you have the PCI IRQ assigned to the device correctly,
> > you need to know which IRQ line on the host this corresponds to.
> > I'm pretty sure that needs to be described in the DTS as
> > commented by Stefan.
> >
>
> So are you saying linux should be writing the irq number to the
> INTERRUPT_LINE config reg? This is what I expected but I do not see
> it.
> I believe the DTS is being parsed properly and the connection is made
> to the correct interrupt line on the device. But somebody still needs
> to assign and IRQ number, right? This is the part that is not clear,
> there is an irq field in pci_dev structure which is filled in after
> looking at the DTS and I just want to be sure this is the irq number to
> be used when calling request_irq.
>
Last time I checked, Linux did not write the assigned IRQ to the card's
PCI configuration space. However, the struct pci_dev irq member
contained the correct value. You should be using the value from the
structure for request_irq(). That's how all PCI drivers I've ever seen
work.
Ira
^ permalink raw reply
* [RFC PATCH v4 1/2] powerpc: reserve fixmap entries for early debug
From: Albert Herranz @ 2009-12-03 20:34 UTC (permalink / raw)
To: linuxppc-dev; +Cc: Albert Herranz
In-Reply-To: <cover.1259871725.git.albert_herranz@yahoo.es>
Add a set of entries to the fixmap table to allow usage of known
reserved virtual address space by early debug code.
The address space reserved is the top 128K of the 32-bit address
space. This allows, if required, the use of a BAT to do the mappings.
Signed-off-by: Albert Herranz <albert_herranz@yahoo.es>
---
arch/powerpc/include/asm/fixmap.h | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/arch/powerpc/include/asm/fixmap.h b/arch/powerpc/include/asm/fixmap.h
index f1f4e23..5c2c023 100644
--- a/arch/powerpc/include/asm/fixmap.h
+++ b/arch/powerpc/include/asm/fixmap.h
@@ -44,6 +44,9 @@
*/
enum fixed_addresses {
FIX_HOLE,
+ /* reserve the top 128K for early debugging purposes */
+ FIX_EARLY_DEBUG_TOP = FIX_HOLE,
+ FIX_EARLY_DEBUG_BASE = FIX_EARLY_DEBUG_TOP+((128*1024)/PAGE_SIZE)-1,
#ifdef CONFIG_HIGHMEM
FIX_KMAP_BEGIN, /* reserved pte's for temporary kernel mappings */
FIX_KMAP_END = FIX_KMAP_BEGIN+(KM_TYPE_NR*NR_CPUS)-1,
--
1.6.3.3
^ permalink raw reply related
* [RFC PATCH v4 0/2] rework of usbgecko-based early debug
From: Albert Herranz @ 2009-12-03 20:34 UTC (permalink / raw)
To: linuxppc-dev; +Cc: Albert Herranz
This is v4 of the "early debugging using usbgecko" patch.
It tries to implement the ideas suggested by Benjamin Herrenschmidt
regarding the use of a set of fixmap entries to provide safe
virtual address space for an early mapping.
Albert Herranz (2):
powerpc: reserve fixmap entries for early debug
powerpc: gamecube/wii: early debugging using usbgecko
arch/powerpc/Kconfig.debug | 8 +++
arch/powerpc/include/asm/fixmap.h | 3 +
arch/powerpc/include/asm/udbg.h | 1 +
arch/powerpc/kernel/head_32.S | 25 +++++++++
arch/powerpc/kernel/udbg.c | 2 +
arch/powerpc/platforms/embedded6xx/usbgecko_udbg.c | 56 ++++++++++++++++++++
arch/powerpc/platforms/embedded6xx/usbgecko_udbg.h | 2 +
7 files changed, 97 insertions(+), 0 deletions(-)
^ permalink raw reply
* [RFC PATCH v4 2/2] powerpc: gamecube/wii: early debugging using usbgecko
From: Albert Herranz @ 2009-12-03 20:35 UTC (permalink / raw)
To: linuxppc-dev; +Cc: Albert Herranz
In-Reply-To: <cover.1259871725.git.albert_herranz@yahoo.es>
Add support for using the USB Gecko adapter as an early debugging
console on the Nintendo GameCube and Wii video game consoles.
The USB Gecko is a 3rd party memory card interface adapter that provides
a EXI (External Interface) to USB serial converter.
Signed-off-by: Albert Herranz <albert_herranz@yahoo.es>
---
arch/powerpc/Kconfig.debug | 8 +++
arch/powerpc/include/asm/udbg.h | 1 +
arch/powerpc/kernel/head_32.S | 25 +++++++++
arch/powerpc/kernel/udbg.c | 2 +
arch/powerpc/platforms/embedded6xx/usbgecko_udbg.c | 56 ++++++++++++++++++++
arch/powerpc/platforms/embedded6xx/usbgecko_udbg.h | 2 +
6 files changed, 94 insertions(+), 0 deletions(-)
diff --git a/arch/powerpc/Kconfig.debug b/arch/powerpc/Kconfig.debug
index 3b10051..11e385b 100644
--- a/arch/powerpc/Kconfig.debug
+++ b/arch/powerpc/Kconfig.debug
@@ -254,6 +254,14 @@ config PPC_EARLY_DEBUG_CPM
using a CPM-based serial port. This assumes that the bootwrapper
has run, and set up the CPM in a particular way.
+config PPC_EARLY_DEBUG_USBGECKO
+ bool "Early debugging through the USB Gecko adapter"
+ depends on GAMECUBE_COMMON
+ select USBGECKO_UDBG
+ help
+ Select this to enable early debugging for Nintendo GameCube/Wii
+ consoles via an external USB Gecko adapter.
+
endchoice
config PPC_EARLY_DEBUG_44x_PHYSLOW
diff --git a/arch/powerpc/include/asm/udbg.h b/arch/powerpc/include/asm/udbg.h
index cd21e5e..11ae699 100644
--- a/arch/powerpc/include/asm/udbg.h
+++ b/arch/powerpc/include/asm/udbg.h
@@ -51,6 +51,7 @@ extern void __init udbg_init_btext(void);
extern void __init udbg_init_44x_as1(void);
extern void __init udbg_init_40x_realmode(void);
extern void __init udbg_init_cpm(void);
+extern void __init udbg_init_usbgecko(void);
#endif /* __KERNEL__ */
#endif /* _ASM_POWERPC_UDBG_H */
diff --git a/arch/powerpc/kernel/head_32.S b/arch/powerpc/kernel/head_32.S
index 829c3fe..00b89c0 100644
--- a/arch/powerpc/kernel/head_32.S
+++ b/arch/powerpc/kernel/head_32.S
@@ -164,6 +164,9 @@ __after_mmu_off:
#ifdef CONFIG_PPC_EARLY_DEBUG_CPM
bl setup_cpm_bat
#endif
+#ifdef CONFIG_PPC_EARLY_DEBUG_USBGECKO
+ bl setup_usbgecko_bat
+#endif
/*
* Call setup_cpu for CPU 0 and initialize 6xx Idle
@@ -1203,6 +1206,28 @@ setup_cpm_bat:
blr
#endif
+#ifdef CONFIG_PPC_EARLY_DEBUG_USBGECKO
+setup_usbgecko_bat:
+ /* prepare a BAT for early io */
+#if defined(CONFIG_GAMECUBE)
+ lis r8, 0x0c00
+#elif defined(CONFIG_WII)
+ lis r8, 0x0d00
+#else
+#error Invalid platform for USB Gecko based early debugging.
+#endif
+ /*
+ * The virtual address used must match the virtual address
+ * associated to the fixmap entry FIX_EARLY_DEBUG_BASE.
+ */
+ lis r11, 0xfffe /* top 128K */
+ ori r8, r8, 0x002a /* uncached, guarded ,rw */
+ ori r11, r11, 0x3 /* 128K */
+ mtspr SPRN_DBAT1L, r8
+ mtspr SPRN_DBAT1U, r11
+ blr
+#endif
+
#ifdef CONFIG_8260
/* Jump into the system reset for the rom.
* We first disable the MMU, and then jump to the ROM reset address.
diff --git a/arch/powerpc/kernel/udbg.c b/arch/powerpc/kernel/udbg.c
index fc9af47..e39cad8 100644
--- a/arch/powerpc/kernel/udbg.c
+++ b/arch/powerpc/kernel/udbg.c
@@ -60,6 +60,8 @@ void __init udbg_early_init(void)
udbg_init_40x_realmode();
#elif defined(CONFIG_PPC_EARLY_DEBUG_CPM)
udbg_init_cpm();
+#elif defined(CONFIG_PPC_EARLY_DEBUG_USBGECKO)
+ udbg_init_usbgecko();
#endif
#ifdef CONFIG_PPC_EARLY_DEBUG
diff --git a/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.c b/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.c
index ba4c7cc..edc956c 100644
--- a/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.c
+++ b/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.c
@@ -17,6 +17,7 @@
#include <asm/io.h>
#include <asm/prom.h>
#include <asm/udbg.h>
+#include <asm/fixmap.h>
#include "usbgecko_udbg.h"
@@ -270,3 +271,58 @@ done:
of_node_put(np);
return;
}
+
+#ifdef CONFIG_PPC_EARLY_DEBUG_USBGECKO
+
+static phys_addr_t __init ug_early_grab_io_addr(void)
+{
+#if defined(CONFIG_GAMECUBE)
+ return 0x0c000000;
+#elif defined(CONFIG_WII)
+ return 0x0d000000;
+#else
+#error Invalid platform for USB Gecko based early debugging.
+#endif
+}
+
+/*
+ * USB Gecko early debug support initialization for udbg.
+ */
+void __init udbg_init_usbgecko(void)
+{
+ void __iomem *early_debug_area;
+ void __iomem *exi_io_base;
+
+ /*
+ * At this point we have a BAT already setup that enables I/O
+ * to the EXI hardware.
+ *
+ * The BAT uses a virtual address range reserved at the fixmap.
+ * This must match the virtual address configured in
+ * head_32.S:setup_usbgecko_bat().
+ */
+ early_debug_area = (void __iomem *)__fix_to_virt(FIX_EARLY_DEBUG_BASE);
+ exi_io_base = early_debug_area + 0x00006800;
+
+ /* try to detect a USB Gecko */
+ if (!ug_udbg_probe(exi_io_base))
+ return;
+
+ /* we found a USB Gecko, load udbg hooks */
+ udbg_putc = ug_udbg_putc;
+ udbg_getc = ug_udbg_getc;
+ udbg_getc_poll = ug_udbg_getc_poll;
+
+ /*
+ * Prepare again the same BAT for MMU_init.
+ * This allows udbg I/O to continue working after the MMU is
+ * turned on for real.
+ * It is safe to continue using the same virtual address as it is
+ * a reserved fixmap area.
+ */
+ setbat(1, (unsigned long)early_debug_area,
+ ug_early_grab_io_addr(), 128*1024, PAGE_KERNEL_NCG);
+}
+
+#endif /* CONFIG_PPC_EARLY_DEBUG_USBGECKO */
+
diff --git a/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.h b/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.h
index 3929de3..bb6cde4 100644
--- a/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.h
+++ b/arch/powerpc/platforms/embedded6xx/usbgecko_udbg.h
@@ -27,4 +27,6 @@ static inline void __init ug_udbg_init(void)
#endif /* CONFIG_USBGECKO_UDBG */
+void __init udbg_init_usbgecko(void);
+
#endif /* __USBGECKO_UDBG_H */
--
1.6.3.3
^ permalink raw reply related
* Re: Recommended functions for accessing internal registers
From: Benjamin Herrenschmidt @ 2009-12-03 21:10 UTC (permalink / raw)
To: Fortini Matteo; +Cc: linuxppc-dev@lists.ozlabs.org
In-Reply-To: <4B179DF5.90600@mta.it>
On Thu, 2009-12-03 at 12:16 +0100, Fortini Matteo wrote:
> I'm on an embedded system, so every resource counts.
> One of the biggest impacts is when writing to a communication/memory
> access FIFO or reading/writing configurations.
> In these cases, I'd just need to make sure that there's no I/O
> reordering and/or subsequent r/w are not optimized away, I believe.
>
> Should I switch to the deprecated "volatile" attribute?
If it's a single register fifo, you can use the _outs/_ins functions or
the iomap.h variants which are cleaner.
If it's a linear region, look at memcpy_to_io...
You can always go directly poking at it but you need appropriate memory
barriers in and around your accesses.
The reason there's a twi/isync pair inside in_* is to ensure that the
read is actually performed immediately. Without this, it could be
delayed until the CPU decides to "consume" the data which could cause
timing issues when a read is followed by a delay for example. There's a
few other reasons why it's a good idea to do so.
Cheers,
Ben.
> Thank you.
>
> Cheers,
> Matteo
>
> Il 02/12/2009 21.57, Benjamin Herrenschmidt ha scritto:
> > On Tue, 2009-12-01 at 17:44 +0100, Fortini Matteo wrote:
> >
> >> I see that throughout the kernel source, internal PPC registers are
> >> accessed through [in|out]_be[32|16|8]() functions. However, they are
> >> translated into 3 inline assembly instructions, one of which is an
> >> isync, which has a huge performance hit.
> >> I tried using readl_be() which seems to be the right function according
> >> to the Documentation/ dir, but it is translated directly to in_be32(),
> >> so no luck.
> >>
> >> Is it really necessary to use all those instructions? I know I could use
> >> a (volatile u32 *) variable to avoid subsequent read/writes to be
> >> optimized out, but it seems to be a deprecated use.
> >>
> > There are good reasons why the accessors contain those barriers. What
> > are you doing that would be performance critical enough for those to be
> > a problem ?
> >
> > Cheers,
> > Ben.
> >
> >
> >
^ permalink raw reply
* Re: PCI interrupt question
From: Wolfgang Denk @ 2009-12-03 21:28 UTC (permalink / raw)
To: David Hawkins; +Cc: Jeff Hane, linuxppc-dev@ozlabs.org
In-Reply-To: <4B1816F9.1020601@ovro.caltech.edu>
Dear David Hawkins,
In message <4B1816F9.1020601@ovro.caltech.edu> you wrote:
>
> Yep. It might require enabling a PCI subsystem in U-Boot.
> I recall playing with a Yosemite board and a PCI card,
> but I forget whether I had to do anything to enable
> PCI configuration space setup. Post a question to the
> U-Boot list.
PCI is enabled by default on the Canyonlands board, both in U-Boot and
Linux.
Best regards,
Wolfgang Denk
--
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
"In the long run, every program becomes rococo, and then rubble."
- Alan Perlis
^ permalink raw reply
* Re: PCI interrupt question
From: David Hawkins @ 2009-12-03 21:38 UTC (permalink / raw)
To: Wolfgang Denk; +Cc: Jeff Hane, linuxppc-dev@ozlabs.org
In-Reply-To: <20091203212821.463D5E6D391@gemini.denx.de>
Hi Wolfgang,
> In message <4B1816F9.1020601@ovro.caltech.edu> you wrote:
>> Yep. It might require enabling a PCI subsystem in U-Boot.
>> I recall playing with a Yosemite board and a PCI card,
>> but I forget whether I had to do anything to enable
>> PCI configuration space setup. Post a question to the
>> U-Boot list.
>
> PCI is enabled by default on the Canyonlands board, both in U-Boot and
> Linux.
Great, thanks!
So Jeff, check out whether or not you can write to the
IRQ register in the PCI configuration space of the device
(using setpci from your host).
If you cannot write to the register, then I'd check your
FPGA core settings.
Cheers,
Dave
PS. Here's the output of lspci for a couple of machines I
have in the lab:
----------------------------------------------------------------------
PLX PCI-9054 master/target interface, with Trenton x86 host board
-----------------------------------------------------------------
[dwh@labslcor6 ~]$ lspci -s 03:0c.0 -v
03:0c.0 Bridge: PLX Technology, Inc. PCI9054 32-bit 33MHz PCI <-> IOBus
Bridge (
rev 0a)
Subsystem: PLX Technology, Inc. PCI9054 32-bit 33MHz PCI <->
IOBus Bridg
e
Flags: bus master, medium devsel, latency 64, IRQ 169
Memory at fe9ff000 (32-bit, non-prefetchable) [size=256]
Memory at fb000000 (32-bit, prefetchable) [size=8M]
Capabilities: <access denied>
[dwh@labslcor6 ~]$ lspci -s 03:0c.0 -x
03:0c.0 Bridge: PLX Technology, Inc. PCI9054 32-bit 33MHz PCI <-> IOBus
Bridge (
rev 0a)
00: b5 10 54 90 17 01 90 02 0a 00 80 06 08 40 00 00
10: 00 f0 9f fe 08 00 00 fb 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 b5 10 54 90
30: 00 00 00 00 40 00 00 00 00 00 00 00 0b 01 00 00
0x3c = IRQ 11, whereas the lspci output indicates IRQ 169,
so there is an IRQ mapping going on there.
----------------------------------------------------------------------
MPC8349EA-based board, with Force Computers x86 host CPU
--------------------------------------------------------
[dwh@labslcor4 ~]$ lspci -s 01:0b.0 -v
01:0b.0 Power PC: Freescale Semiconductor Inc MPC8349E (rev 30)
Flags: 66MHz, fast devsel, IRQ 9
Memory at f9100000 (32-bit, non-prefetchable) [size=1M]
Memory at f9600000 (32-bit, non-prefetchable) [size=4K]
Memory at f9500000 (32-bit, non-prefetchable) [size=1M]
Memory at f9400000 (64-bit, non-prefetchable) [size=1M]
Capabilities: <access denied>
[dwh@labslcor4 ~]$ lspci -s 01:0b.0 -x
01:0b.0 Power PC: Freescale Semiconductor Inc MPC8349E (rev 30)
00: 57 19 80 00 02 01 b0 00 30 00 20 0b 08 40 00 00
10: 00 00 10 f9 00 00 60 f9 00 00 50 f9 00 00 00 00
20: 04 00 40 f9 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 48 00 00 00 00 00 00 00 09 01 00 00
0x3c = IRQ 9, and that matches the lspci output, so there is
no extra mapping here.
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox