[PATCH] get the kernel to warn about deprecated SCSI ioctls

[PATCH] get the kernel to warn about deprecated SCSI ioctls

[James Bottomley]

The non unique SCSI ioctls:

SCSI_IOCTL_SEND_COMMAND
SCSI_IOCTL_TEST_UNIT_READY
SCSI_IOCTL_BENCHMARK_COMMAND
SCSI_IOCTL_SYNC
SCSI_IOCTL_START_UNIT
SCSI_IOCTL_STOP_UNIT

Have been deprecated for a while now.  We should make the kernel
complain when a program actually uses one of them, so users have had
adequate warning before they are removed.

James

===== drivers/scsi/scsi_ioctl.c 1.28 vs edited =====
--- 1.28/drivers/scsi/scsi_ioctl.c	2004-06-26 17:40:24 -05:00
+++ edited/drivers/scsi/scsi_ioctl.c	2004-08-19 09:45:22 -05:00
@@ -391,6 +391,21 @@
 	if (!scsi_block_when_processing_errors(sdev))
 		return -ENODEV;
 
+	/* Check for deprecated ioctls ... all the ioctls which don't
+	 * follow the new unique numbering scheme are deprecated */
+	switch (cmd) {
+	case SCSI_IOCTL_SEND_COMMAND:
+	case SCSI_IOCTL_TEST_UNIT_READY:
+	case SCSI_IOCTL_BENCHMARK_COMMAND:
+	case SCSI_IOCTL_SYNC:
+	case SCSI_IOCTL_START_UNIT:
+	case SCSI_IOCTL_STOP_UNIT:
+		printk(KERN_WARNING "program %s is using a deprecated SCSI ioctl, please convert it to SG_IO\n", current->comm);
+		break;
+	default:
+		break;
+	}
+
 	switch (cmd) {
 	case SCSI_IOCTL_GET_IDLUN:
 		if (verify_area(VERIFY_WRITE, arg, sizeof(struct scsi_idlun)))

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[Christoph Hellwig]

On Thu, Aug 19, 2004 at 12:19:01PM -0400, James Bottomley wrote:
> The non unique SCSI ioctls:
> 
> SCSI_IOCTL_SEND_COMMAND
> SCSI_IOCTL_TEST_UNIT_READY
> SCSI_IOCTL_BENCHMARK_COMMAND
> SCSI_IOCTL_SYNC
> SCSI_IOCTL_START_UNIT
> SCSI_IOCTL_STOP_UNIT
> 
> Have been deprecated for a while now.  We should make the kernel
> complain when a program actually uses one of them, so users have had
> adequate warning before they are removed.

drivers/block/scsi_ioctl.c also implements SCSI_IOCTL_SEND_COMMAND, so
you'll have to warn there aswell.

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[James Bottomley]

On Thu, 2004-08-19 at 12:21, Christoph Hellwig wrote:
> drivers/block/scsi_ioctl.c also implements SCSI_IOCTL_SEND_COMMAND, so
> you'll have to warn there aswell.

Done.

No other places I've missed?

James

===== drivers/block/scsi_ioctl.c 1.53 vs edited =====
--- 1.53/drivers/block/scsi_ioctl.c	2004-08-14 05:24:49 -05:00
+++ edited/drivers/block/scsi_ioctl.c	2004-08-19 15:51:44 -05:00
@@ -510,6 +510,7 @@
 		 * old junk scsi send command ioctl
 		 */
 		case SCSI_IOCTL_SEND_COMMAND:
+			printk(KERN_WARNING "program %s is using a deprecated SCSI ioctl, please convert it to SG_IO\n", current->comm);
 			err = -EINVAL;
 			if (!arg)
 				break;
===== drivers/scsi/scsi_ioctl.c 1.28 vs edited =====
--- 1.28/drivers/scsi/scsi_ioctl.c	2004-06-26 17:40:24 -05:00
+++ edited/drivers/scsi/scsi_ioctl.c	2004-08-19 09:45:22 -05:00
@@ -391,6 +391,21 @@
 	if (!scsi_block_when_processing_errors(sdev))
 		return -ENODEV;
 
+	/* Check for deprecated ioctls ... all the ioctls which don't
+	 * follow the new unique numbering scheme are deprecated */
+	switch (cmd) {
+	case SCSI_IOCTL_SEND_COMMAND:
+	case SCSI_IOCTL_TEST_UNIT_READY:
+	case SCSI_IOCTL_BENCHMARK_COMMAND:
+	case SCSI_IOCTL_SYNC:
+	case SCSI_IOCTL_START_UNIT:
+	case SCSI_IOCTL_STOP_UNIT:
+		printk(KERN_WARNING "program %s is using a deprecated SCSI ioctl, please convert it to SG_IO\n", current->comm);
+		break;
+	default:
+		break;
+	}
+
 	switch (cmd) {
 	case SCSI_IOCTL_GET_IDLUN:
 		if (verify_area(VERIFY_WRITE, arg, sizeof(struct scsi_idlun)))

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[Jeff Garzik]

On Thu, Aug 19, 2004 at 12:19:01PM -0400, James Bottomley wrote:
> The non unique SCSI ioctls:
> 
> SCSI_IOCTL_SEND_COMMAND
> SCSI_IOCTL_TEST_UNIT_READY
> SCSI_IOCTL_BENCHMARK_COMMAND
> SCSI_IOCTL_SYNC
> SCSI_IOCTL_START_UNIT
> SCSI_IOCTL_STOP_UNIT
> 
> Have been deprecated for a while now.  We should make the kernel
> complain when a program actually uses one of them, so users have had
> adequate warning before they are removed.

any chance of deprecating the doorlock ioctls too?

	Jeff

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[James Bottomley]

On Thu, 2004-08-19 at 12:23, Jeff Garzik wrote:
> On Thu, Aug 19, 2004 at 12:19:01PM -0400, James Bottomley wrote:
> > The non unique SCSI ioctls:
> > 
> > SCSI_IOCTL_SEND_COMMAND
> > SCSI_IOCTL_TEST_UNIT_READY
> > SCSI_IOCTL_BENCHMARK_COMMAND
> > SCSI_IOCTL_SYNC
> > SCSI_IOCTL_START_UNIT
> > SCSI_IOCTL_STOP_UNIT
> > 
> > Have been deprecated for a while now.  We should make the kernel
> > complain when a program actually uses one of them, so users have had
> > adequate warning before they are removed.
> 
> any chance of deprecating the doorlock ioctls too?

The doorlock ones are unique, and are used by some of the CD tools.

We really need to know when an app locks the door rather than having to
scan the commands in SG_IO.

James

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[Jeff Garzik]

James Bottomley wrote:
> On Thu, 2004-08-19 at 12:23, Jeff Garzik wrote:
> 
>>On Thu, Aug 19, 2004 at 12:19:01PM -0400, James Bottomley wrote:
>>
>>>The non unique SCSI ioctls:
>>>
>>>SCSI_IOCTL_SEND_COMMAND
>>>SCSI_IOCTL_TEST_UNIT_READY
>>>SCSI_IOCTL_BENCHMARK_COMMAND
>>>SCSI_IOCTL_SYNC
>>>SCSI_IOCTL_START_UNIT
>>>SCSI_IOCTL_STOP_UNIT
>>>
>>>Have been deprecated for a while now.  We should make the kernel
>>>complain when a program actually uses one of them, so users have had
>>>adequate warning before they are removed.
>>
>>any chance of deprecating the doorlock ioctls too?
> 
> 
> The doorlock ones are unique, and are used by some of the CD tools.
> 
> We really need to know when an app locks the door rather than having to
> scan the commands in SG_IO.


You need to scan the commands in SG_IO anyway, otherwise you do not know 
in all cases when an app locks the door :)

	Jeff

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[Douglas Gilbert]

James Bottomley wrote:
> The non unique SCSI ioctls:
> 
> SCSI_IOCTL_SEND_COMMAND
> SCSI_IOCTL_TEST_UNIT_READY
> SCSI_IOCTL_BENCHMARK_COMMAND
> SCSI_IOCTL_SYNC
> SCSI_IOCTL_START_UNIT
> SCSI_IOCTL_STOP_UNIT
> 
> Have been deprecated for a while now.  We should make the kernel
> complain when a program actually uses one of them, so users have had
> adequate warning before they are removed.

James,
[Wearing my smartmontools maintainer hat] such a change is going
to cause a lot of noise in the log from the smartd daemon.
The code to use SG_IO ioctl in smartmontools is in CVS but
won't be released for a month or so.

If they are not going to be dropped in the lk 2.6 series
why not wait until the lk 2.7 development starts (or late
lk 2.6 series)?

Doug Gilbert

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[James Bottomley]

On Thu, 2004-08-19 at 18:18, Douglas Gilbert wrote:
> [Wearing my smartmontools maintainer hat] such a change is going
> to cause a lot of noise in the log from the smartd daemon.
> The code to use SG_IO ioctl in smartmontools is in CVS but
> won't be released for a month or so.
> 
> If they are not going to be dropped in the lk 2.6 series
> why not wait until the lk 2.7 development starts (or late
> lk 2.6 series)?

The current issue is that some of them have no compat code, so they
don't actually work 32 on 64.

If the noise becomes unbearable, we can rate limit it, but I'd like to
get the message out.

James

Re: [PATCH] get the kernel to warn about deprecated SCSI ioctls

[Pat LaVarre]

Yes, please, which SCSI pass thru ioctl should I use in Linux.

> SCSI_IOCTL_SEND_COMMAND ...
> please convert ... to SG_IO

Ouch, I am instead choosing the mostly undocumented ioctl
CDROM_SEND_PACKET, opened O_NONBLOCK, because I can't see how else to
reach 2.4 Linux /dev/hd$v thru cdrom ide-cd.

> SCSI_IOCTL_SEND_COMMAND ... deprecated ...

Me the newbie, I missed and I still do not see how I should know this
SCSI pass thru ioctl is deprecated, unless we add the printk you
suggest.

Pat LaVarre
http://marc.theaimsgroup.com/?l=linux-scsi&m=109278095826988
http://linux-pel.blog-city.com/read/756175.htm
http://linux-pel.blog-city.com/read/771746.htm

P.S. I'm resorting to SCSI pass thru for such purposes as toggling the
invisible MMC PWP write protect switch found in DVD-RAM discs, Iomega
REV discs, etc.