[PATCH] libata transport attributes

[PATCH] libata transport attributes

[Justin Cormack]

[-- Attachment #1: Type: text/plain, Size: 949 bytes --]

Here is a rough patch to add libata transport attributes, along the
lines of fibre channel and parallel scsi. I wrote it as it seemed to be
the cleanest way of extracting ata-specific information (currently drive
model, serial number and firmware revision) from sysfs (in
/sys/class/libata_transport/). There are a few issues, in particular:

1. it wont compile modular, as libata depends on scsi_transport_libata
and vice-versa at the moment. I am not sure how you are supposed to get
around this (and there arent any significant number of drivers in tree
using the transport modules yet).

2. It would be nice if the device directory in sysfs had a symlink to
the transport attributes directory, not just the other way round.

3. I couldnt work out what scsi_transport_template.size was the size of,
as I couldnt see where it was used anywhere...

patch against 2.6.5-rc3-libata1 but probably applies against most recent
kernels.

Justin Cormack


[-- Attachment #2: patch-libata-transport --]
[-- Type: text/x-patch, Size: 7735 bytes --]

diff -urN linux-2.6.5-rc3-libata1-orig/drivers/scsi/Kconfig linux-2.6.5-rc3-libata1/drivers/scsi/Kconfig
--- linux-2.6.5-rc3-libata1-orig/drivers/scsi/Kconfig	2004-03-31 09:41:24.000000000 +0100
+++ linux-2.6.5-rc3-libata1/drivers/scsi/Kconfig	2004-03-31 09:44:39.024199360 +0100
@@ -214,6 +214,13 @@
 	  each attached FiberChannel device to sysfs, say Y.
 	  Otherwise, say N.
 
+config SCSI_LIBATA_ATTRS
+        tristate "Libata SCSI Transport Attributes"
+        depends on SCSI
+        help
+          If you wish to export transport-specific information about
+          each attached SATA device to sysfs, say Y.  Otherwise, say N.
+
 endmenu
 
 menu "SCSI low-level drivers"
diff -urN linux-2.6.5-rc3-libata1-orig/drivers/scsi/libata-core.c linux-2.6.5-rc3-libata1/drivers/scsi/libata-core.c
--- linux-2.6.5-rc3-libata1-orig/drivers/scsi/libata-core.c	2004-03-31 09:41:24.000000000 +0100
+++ linux-2.6.5-rc3-libata1/drivers/scsi/libata-core.c	2004-03-31 09:46:23.574305328 +0100
@@ -44,6 +44,8 @@
 
 #include "libata.h"
 
+#include <scsi/scsi_transport_libata.h>
+
 static void atapi_cdb_send(struct ata_port *ap);
 static unsigned int ata_busy_sleep (struct ata_port *ap,
 				    unsigned long tmout_pat,
@@ -2938,6 +2940,7 @@
 	host->max_channel = 1;
 	host->unique_id = ata_unique_id++;
 	host->max_cmd_len = 12;
+        host->transportt = &libata_transport_template;
 	scsi_set_device(host, &ent->pdev->dev);
 	scsi_assign_lock(host, &host_set->lock);
 
@@ -3532,3 +3535,4 @@
 EXPORT_SYMBOL_GPL(ata_scsi_slave_config);
 EXPORT_SYMBOL_GPL(ata_scsi_release);
 EXPORT_SYMBOL_GPL(ata_host_intr);
+EXPORT_SYMBOL_GPL(ata_dev_id_string);
diff -urN linux-2.6.5-rc3-libata1-orig/drivers/scsi/Makefile linux-2.6.5-rc3-libata1/drivers/scsi/Makefile
--- linux-2.6.5-rc3-libata1-orig/drivers/scsi/Makefile	2004-03-31 09:41:24.000000000 +0100
+++ linux-2.6.5-rc3-libata1/drivers/scsi/Makefile	2004-03-31 09:43:56.033734904 +0100
@@ -28,6 +28,7 @@
 # --------------------------
 obj-$(CONFIG_SCSI_SPI_ATTRS)	+= scsi_transport_spi.o
 obj-$(CONFIG_SCSI_FC_ATTRS) 	+= scsi_transport_fc.o
+obj-$(CONFIG_SCSI_LIBATA_ATTRS) += scsi_transport_libata.o
 
 
 obj-$(CONFIG_SCSI_AMIGA7XX)	+= amiga7xx.o	53c7xx.o
diff -urN linux-2.6.5-rc3-libata1-orig/drivers/scsi/scsi_transport_libata.c linux-2.6.5-rc3-libata1/drivers/scsi/scsi_transport_libata.c
--- linux-2.6.5-rc3-libata1-orig/drivers/scsi/scsi_transport_libata.c	1970-01-01 01:00:00.000000000 +0100
+++ linux-2.6.5-rc3-libata1/drivers/scsi/scsi_transport_libata.c	2004-03-31 09:42:42.784870424 +0100
@@ -0,0 +1,116 @@
+/* 
+   Libata transport specific attributes exported to sysfs.
+
+   Copyright (c) 2004 Justin Cormack
+ 
+   The contents of this file are subject to the Open
+   Software License version 1.1 that can be found at
+   http://www.opensource.org/licenses/osl-1.1.txt and is included herein
+   by reference.
+
+   Alternatively, the contents of this file may be used under the terms
+   of the GNU General Public License version 2 (the "GPL") as distributed
+   in the kernel source COPYING file, in which case the provisions of
+   the GPL are applicable instead of the above.  If you wish to allow
+   the use of your version of this file only under the terms of the
+   GPL and not to allow others to use your version of this file under
+   the OSL, indicate your decision by deleting the provisions above and
+   replace them with the notice and other provisions required by the GPL.
+   If you do not delete the provisions above, a recipient may use your
+   version of this file under either the OSL or the GPL.
+
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/init.h>
+#include <linux/ata.h>
+#include <scsi/scsi.h>
+#include "scsi.h"
+#include "hosts.h"
+#include <linux/libata.h>
+#include "libata.h"
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_transport.h>
+
+static void transport_class_release(struct class_device *class_dev);
+
+struct class libata_transport_class = {
+	.name = "libata_transport",
+	.release = transport_class_release,
+};
+
+static __init int libata_transport_init(void)
+{
+	return class_register(&libata_transport_class);
+}
+
+static void __exit libata_transport_exit(void)
+{
+	class_unregister(&libata_transport_class);
+}
+
+static void transport_class_release(struct class_device *class_dev)
+{
+	struct scsi_device *sdev = transport_class_to_sdev(class_dev);
+	put_device(&sdev->sdev_gendev);
+}
+
+static ssize_t show_libata_transport_string(struct class_device *cdev, char *buf, int offset, int len)
+{
+        struct scsi_device *sdev = transport_class_to_sdev(cdev);
+	struct ata_port *ap = (struct ata_port *) &sdev->host->hostdata[0];
+	struct ata_device *dev = &ap->device[sdev->id];
+
+	ata_dev_id_string(dev, buf, offset, len);
+	buf[len + 1] = '\n';
+	buf[len + 2] = 0;
+
+	return len + 2;
+}
+
+static ssize_t show_libata_transport_model(struct class_device *cdev, char *buf)
+{
+        return show_libata_transport_string(cdev, buf, ATA_ID_PROD_OFS, ATA_PROD_LEN);
+}
+
+static ssize_t show_libata_transport_firmware(struct class_device *cdev, char *buf)
+{
+        return show_libata_transport_string(cdev, buf, ATA_ID_FW_OFS, ATA_FW_LEN);
+}
+
+static ssize_t show_libata_transport_serial(struct class_device *cdev, char *buf)
+{
+        return show_libata_transport_string(cdev, buf, ATA_ID_SERNO_OFS, ATA_SERNO_LEN);
+}
+
+static CLASS_DEVICE_ATTR(model, S_IRUGO, show_libata_transport_model, NULL);
+static CLASS_DEVICE_ATTR(firmware, S_IRUGO, show_libata_transport_firmware, NULL);
+static CLASS_DEVICE_ATTR(serial, S_IRUGO, show_libata_transport_serial, NULL);
+
+struct class_device_attribute *libata_transport_attrs[] = {
+	&class_device_attr_model,
+	&class_device_attr_firmware,
+	&class_device_attr_serial,
+	NULL
+};
+
+struct scsi_transport_template libata_transport_template = {
+        .attrs = libata_transport_attrs,
+	.class = &libata_transport_class,
+	.setup = NULL,
+	.cleanup = NULL,
+	/* what is this the size of? */
+	.size = sizeof(libata_transport_attrs) - sizeof(unsigned long),
+};
+
+EXPORT_SYMBOL(libata_transport_template);
+
+MODULE_AUTHOR("Justin Cormack");
+MODULE_DESCRIPTION("Libata Transport Attributes");
+MODULE_LICENSE("GPL");
+
+module_init(libata_transport_init);
+module_exit(libata_transport_exit);
diff -urN linux-2.6.5-rc3-libata1-orig/include/linux/ata.h linux-2.6.5-rc3-libata1/include/linux/ata.h
--- linux-2.6.5-rc3-libata1-orig/include/linux/ata.h	2004-03-31 09:41:24.000000000 +0100
+++ linux-2.6.5-rc3-libata1/include/linux/ata.h	2004-03-31 09:48:13.945526352 +0100
@@ -39,13 +39,16 @@
 	ATA_ID_WORDS		= 256,
 	ATA_ID_PROD_OFS		= 27,
 	ATA_ID_SERNO_OFS	= 10,
+        ATA_ID_FW_OFS           = 23,
 	ATA_ID_MAJOR_VER	= 80,
 	ATA_ID_PIO_MODES	= 64,
 	ATA_ID_UDMA_MODES	= 88,
 	ATA_ID_PIO4		= (1 << 1),
 
 	ATA_PCI_CTL_OFS		= 2,
+        ATA_PROD_LEN            = 40,
 	ATA_SERNO_LEN		= 20,
+        ATA_FW_LEN              = 8,
 	ATA_UDMA0		= (1 << 0),
 	ATA_UDMA1		= ATA_UDMA0 | (1 << 1),
 	ATA_UDMA2		= ATA_UDMA1 | (1 << 2),
diff -urN linux-2.6.5-rc3-libata1-orig/include/scsi/scsi_transport_libata.h linux-2.6.5-rc3-libata1/include/scsi/scsi_transport_libata.h
--- linux-2.6.5-rc3-libata1-orig/include/scsi/scsi_transport_libata.h	1970-01-01 01:00:00.000000000 +0100
+++ linux-2.6.5-rc3-libata1/include/scsi/scsi_transport_libata.h	2004-03-31 09:43:10.491658352 +0100
@@ -0,0 +1,12 @@
+/*
+    Libata transport specific attributes exported to sysfs.
+ */
+
+#ifndef SCSI_TRANSPORT_LIBATA_H
+#define SCSI_TRANSPORT_LIBATA_H
+
+struct scsi_transport_template;
+
+extern struct scsi_transport_template libata_transport_template;
+
+#endif /* SCSI_TRANSPORT_LIBATA_H */

Re: [PATCH] libata transport attributes

[Jeff Garzik]

Justin Cormack wrote:
> Here is a rough patch to add libata transport attributes, along the
> lines of fibre channel and parallel scsi. I wrote it as it seemed to be
> the cleanest way of extracting ata-specific information (currently drive
> model, serial number and firmware revision) from sysfs (in
> /sys/class/libata_transport/). There are a few issues, in particular:
> 
> 1. it wont compile modular, as libata depends on scsi_transport_libata
> and vice-versa at the moment. I am not sure how you are supposed to get
> around this (and there arent any significant number of drivers in tree
> using the transport modules yet).
> 
> 2. It would be nice if the device directory in sysfs had a symlink to
> the transport attributes directory, not just the other way round.
> 
> 3. I couldnt work out what scsi_transport_template.size was the size of,
> as I couldnt see where it was used anywhere...
> 
> patch against 2.6.5-rc3-libata1 but probably applies against most recent
> kernels.


Did you see the comments I posted WRT mort's patch?

Since libata is leaving SCSI in 2.7, I would rather not add superfluous 
stuff like this at all.

Further, you can already retrieve the information you export with _zero_ 
new code.

	Jeff

Re: [PATCH] libata transport attributes

[Justin Cormack]

On Wed, 2004-03-31 at 22:07, Jeff Garzik wrote:

> Did you see the comments I posted WRT mort's patch?
> 

oops, no I missed his patch and your comments until now.

> Since libata is leaving SCSI in 2.7, I would rather not add superfluous 
> stuff like this at all.
> 

I didnt know this.

> Further, you can already retrieve the information you export with _zero_ 
> new code.

How? Sorry to be dumb...

Justin

Re: [PATCH] libata transport attributes

[Jeff Garzik]

Justin Cormack wrote:
> On Wed, 2004-03-31 at 22:07, Jeff Garzik wrote:
> 
> 
>>Did you see the comments I posted WRT mort's patch?
>>
> 
> 
> oops, no I missed his patch and your comments until now.
> 
> 
>>Since libata is leaving SCSI in 2.7, I would rather not add superfluous 
>>stuff like this at all.
>>
> 
> 
> I didnt know this.
> 
> 
>>Further, you can already retrieve the information you export with _zero_ 
>>new code.
> 
> 
> How? Sorry to be dumb...

Obtain the SCSI inquiry info from userspace...

	Jeff

Re: [PATCH] libata transport attributes

[Justin Cormack]

On Thu, 2004-04-01 at 19:10, Jeff Garzik wrote:
> Justin Cormack wrote:
> > On Wed, 2004-03-31 at 22:07, Jeff Garzik wrote:
> > 
> > 
> >>Did you see the comments I posted WRT mort's patch?
> >>
> > 
> > 
> > oops, no I missed his patch and your comments until now.
> > 
> > 
> >>Since libata is leaving SCSI in 2.7, I would rather not add superfluous 
> >>stuff like this at all.
> >>
> > 
> > 
> > I didnt know this.
> > 
> > 
> >>Further, you can already retrieve the information you export with _zero_ 
> >>new code.
> > 
> > 
> > How? Sorry to be dumb...
> 
> Obtain the SCSI inquiry info from userspace...

I looked at doing this, but unfortunately this doesnt work due to
incompatibilities between the SCSI and ATA way of doing things. The ATA
product ids are truncated from 20 chars to 16 (there is no separate
vendor ID so both are effectively combined into one field with arbitrary
layout). The product revision level is only 4 bytes in scsi, unlike the
8 byte firmware revision in ATA. You can get the serial number though.

Currently you use ATA as vendor id, I will quite happily patch libata to
stick the whole ATA identify into some EVPD scsi inquiry pages as vendor
specific info and write a userspace tool to parse it if you would
rather. Thats much easier to junk when you move out of scsi-world and
presumably start to export native sysfs info.

Justin

Re: [PATCH] libata transport attributes

[Dieter Stueken]

Jeff Garzik wrote:

> Since libata is leaving SCSI in 2.7, I would rather not add superfluous stuff like this at all.
> Further, you can already retrieve the information you export with _zero_ new code.

does this include informations about the drives SMART status, too?

The smartmotools (http://smartmontools.sourceforge.net/) won't work
with libata until some "passthrough" command gets implemented :-(
Will this be available soon (or even working now), or do I have to
wait until 2.7?

This is the main reason, that prevents me from switching to SATA now!
Just when writing these lines, I'm analyzing some prefail indications
from one of my 300G PATA disks via SMART. I have no problem, to replace
that disk now, but when running those disks blindly without SMART over
SATA, I certainly would loose those 300G very soon...

So, can you give us an idea of a time line about those things, please?

Dieter.
-- 
Dieter Stüken, con terra GmbH, Münster
     stueken@conterra.de
     http://www.conterra.de/
     (0)251-7474-501

Re: [PATCH] libata transport attributes

[Jeff Garzik]

Dieter Stueken wrote:
> Jeff Garzik wrote:
> 
>> Since libata is leaving SCSI in 2.7, I would rather not add 
>> superfluous stuff like this at all.
>> Further, you can already retrieve the information you export with 
>> _zero_ new code.
> 
> 
> does this include informations about the drives SMART status, too?

No.  SMART would not be appropriate for a transport attribute either.


> The smartmotools (http://smartmontools.sourceforge.net/) won't work
> with libata until some "passthrough" command gets implemented :-(
> Will this be available soon (or even working now), or do I have to
> wait until 2.7?

Soon.

	Jeff