[PATCH] rapidio/tsi721: modify mport name assignment

Modify mport device name assignment to provide clear reference to devices
in systems with multiple Tsi721 bridges.

This patch is applicable to kernel versions starting from v3.2.

Signed-off-by: Alexandre Bounine <redacted>
Cc: Matt Porter <mporter@kernel.crashing.org>
---
 drivers/rapidio/devices/tsi721.c |    3 ++-
 include/linux/rio.h              |    3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/rapidio/devices/tsi721.c b/drivers/rapidio/devices/tsi721.c
index 5d44252..8533f36 100644
--- a/drivers/rapidio/devices/tsi721.c
+++ b/drivers/rapidio/devices/tsi721.c

@@ -2165,7 +2165,8 @@ static int __devinit tsi721_setup_mport(struct tsi721_device *priv)
 	rio_init_dbell_res(&mport->riores[RIO_DOORBELL_RESOURCE], 0, 0xffff);
 	rio_init_mbox_res(&mport->riores[RIO_INB_MBOX_RESOURCE], 0, 3);
 	rio_init_mbox_res(&mport->riores[RIO_OUTB_MBOX_RESOURCE], 0, 3);
-	strcpy(mport->name, "Tsi721 mport");
+	snprintf(mport->name, RIO_MAX_MPORT_NAME, "%s(%s)",
+		 dev_driver_string(&pdev->dev), dev_name(&pdev->dev));
 
 	/* Hook up interrupt handler */
 

diff --git a/include/linux/rio.h b/include/linux/rio.h
index a90ebad..1a7b6c7 100644
--- a/include/linux/rio.h
+++ b/include/linux/rio.h

@@ -30,6 +30,7 @@
 #define RIO_MAX_MPORTS		8
 #define RIO_MAX_MPORT_RESOURCES	16
 #define RIO_MAX_DEV_RESOURCES	16
+#define RIO_MAX_MPORT_NAME	40
 
 #define RIO_GLOBAL_TABLE	0xff	/* Indicates access of a switch's
 					   global routing table if it

@@ -255,7 +256,7 @@ struct rio_mport {
 				 */
 	enum rio_phy_type phy_type;	/* RapidIO phy type */
 	u32 phys_efptr;
-	unsigned char name[40];
+	unsigned char name[RIO_MAX_MPORT_NAME];
 	void *priv;		/* Master port private data */
 #ifdef CONFIG_RAPIDIO_DMA_ENGINE
 	struct dma_device	dma;

-- 
1.7.8.4

Modify RIO enumeration to apply RX/TX enable operations only to active
switch ports. This will leave inactive ports in condition consistent with
their state.

This patch is applicable to kernel versions starting from v2.6.35.

Signed-off-by: Alexandre Bounine <redacted>
Cc: Matt Porter <mporter@kernel.crashing.org>
---
 drivers/rapidio/rio-scan.c |    9 +++++----
 1 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/rapidio/rio-scan.c b/drivers/rapidio/rio-scan.c
index 2bebd79..02e686b 100644
--- a/drivers/rapidio/rio-scan.c
+++ b/drivers/rapidio/rio-scan.c

@@ -839,12 +839,10 @@ static int __devinit rio_enum_peer(struct rio_net *net, struct rio_mport *port,
 		for (port_num = 0;
 		     port_num < RIO_GET_TOTAL_PORTS(rdev->swpinfo);
 		     port_num++) {
-			/*Enable Input Output Port (transmitter reviever)*/
-			rio_enable_rx_tx_port(port, 0,
+			if (sw_inport == port_num) {
+				rio_enable_rx_tx_port(port, 0,
 					      RIO_ANY_DESTID(port->sys_size),
 					      hopcount, port_num);
-
-			if (sw_inport == port_num) {
 				rdev->rswitch->port_ok |= (1 << port_num);
 				continue;
 			}

@@ -857,6 +855,9 @@ static int __devinit rio_enum_peer(struct rio_net *net, struct rio_mport *port,
 				pr_debug(
 				    "RIO: scanning device on port %d\n",
 				    port_num);
+				rio_enable_rx_tx_port(port, 0,
+					      RIO_ANY_DESTID(port->sys_size),
+					      hopcount, port_num);
 				rdev->rswitch->port_ok |= (1 << port_num);
 				rio_route_add_entry(rdev, RIO_GLOBAL_TABLE,
 						RIO_ANY_DESTID(port->sys_size),

-- 
1.7.8.4

On Tue, 21 Aug 2012 10:23:54 -0400
Alexandre Bounine [off-list ref] wrote:

Modify mport device name assignment to provide clear reference to devices
in systems with multiple Tsi721 bridges.

This patch is applicable to kernel versions starting from v3.2.

This seems to imply that you think the patch should be backported into
earlier kernels.  But no reason for doing this was provided.

And it's a non-back-compatible change to a userspace-visible interface!
As such we'd need extraordinary justification to merge it into
*future* kernels, let alone backport it.

Please, do provide much better changelogging than this.

On Tue, 21 Aug 2012 10:23:55 -0400
Alexandre Bounine [off-list ref] wrote:

Modify RIO enumeration to apply RX/TX enable operations only to active
switch ports. This will leave inactive ports in condition consistent with
their state.

It's unclear (to me) what the effects of this are.  Does it fix some
user-visible malfunction?  Or is it just some nice-to-have thing?  Or
what?

This patch is applicable to kernel versions starting from v2.6.35.

Is that a recommendation that such a backport be performed?  If so,
please provide the reasoning.

On Friday, August 24, 2012 5:04 PM
Andrew Morton [off-list ref]

=20
On Tue, 21 Aug 2012 10:23:55 -0400
Alexandre Bounine [off-list ref] wrote:
=20

quoted

Modify RIO enumeration to apply RX/TX enable operations only to
active switch ports. This will leave inactive ports in condition consist=

ent

quoted

with their state.

=20
It's unclear (to me) what the effects of this are.  Does it fix some
user-visible malfunction?  Or is it just some nice-to-have thing?  Or
what?
=20

This patch was intended to keep inactive switch ports with inbound and
outbound packet transfers disabled to block unexpected packets during
hot insertion event. While it does not fix any visible malfunction it
was intended to prevent such events in future.  =20

Since this patch was published it was brought to my attention that some
existing switches may require inbound and outbound ports enabled for
inactive ports as well (not spec compliant).
Therefore, please remove this patch from your tree.    =20

quoted

This patch is applicable to kernel versions starting from v2.6.35.

=20
Is that a recommendation that such a backport be performed?  If so,
please provide the reasoning.

On Friday, August 24, 2012 5:02 PM
Andrew Morton [off-list ref]

=20
On Tue, 21 Aug 2012 10:23:54 -0400
Alexandre Bounine [off-list ref] wrote:
=20

quoted

Modify mport device name assignment to provide clear reference to device=

s

quoted

in systems with multiple Tsi721 bridges.

This patch is applicable to kernel versions starting from v3.2.

=20
This seems to imply that you think the patch should be backported into
earlier kernels.  But no reason for doing this was provided.

We cannot prevent users of earlier versions from using more than one Tsi721
device in their systems. In that case the old name form will not provide
definitive reference to a particular mport. Please, see comment below.

=20

quoted

--- a/drivers/rapidio/devices/tsi721.c
+++ b/drivers/rapidio/devices/tsi721.c

@@ -2165,7 +2165,8 @@ static int __devinit tsi721_setup_mport(struct

tsi721_device *priv)

quoted

 	rio_init_dbell_res(&mport->riores[RIO_DOORBELL_RESOURCE], 0,

0xffff);

quoted

 	rio_init_mbox_res(&mport->riores[RIO_INB_MBOX_RESOURCE], 0, 3);
 	rio_init_mbox_res(&mport->riores[RIO_OUTB_MBOX_RESOURCE], 0, 3);
-	strcpy(mport->name, "Tsi721 mport");
+	snprintf(mport->name, RIO_MAX_MPORT_NAME, "%s(%s)",
+		 dev_driver_string(&pdev->dev), dev_name(&pdev->dev));

=20
And it's a non-back-compatible change to a userspace-visible interface!
As such we'd need extraordinary justification to merge it into
*future* kernels, let alone backport it.

An mport device do not provide a userspace-visible interface at this point.
This name is used internally in RapidIO subsystem for resource tracking and
debugging. For Tsi721 as a PCIe peripheral, having the mport name that may =
be
easily associated with its PCIe base helps with tracking of assigned PCIe
resources. This option looks to me better than adding simple indexing
component into existing name format, especially for systems that use add-on
cards with PCIe-to-SRIO bridges.

=20
Please, do provide much better changelogging than this.

Modify RapidIO mport device name assignment to include device name of PCIe
side of Tsi721 bridge. The new name format is intended to provide definitiv=
e
reference between RapidIO and PCIe sides of the bridge in systems with
multiple Tsi721 bridges.