Re: ide-floppy.c vs devfs

Re: ide-floppy.c vs devfs

[Borsenkow Andrej]

>> Try the devfs test version that I just uploaded to 
>> 
>> http://paulbristow.net/linux/idefloppy.html 
>> 
>> This is early days, and I'm not sure what the best approach is... 
>> 
>> Feedback is greatly appreaciated. 
>

On this page you mention that Mandrake includes this patch, but your
driver (link on the above patch) differs from Mandrake one. In patch I
sent to Mandrake I just fake drive capacity in case
CAPACITY_NO_CARTRIDGE is returned.

>
> I haven't had time to look at this closely, but I question why you're 
> trying to prevent grok_partitions() from doing it's thing. There's 
> supposed to be a flag set for removable media which ensures media 
> revalidation on scanning the parent directory or looking up an entry. 
> I'd rather see that mechanism fixed. 

Currently grok_partitions() silently ignores any media with size == 0,
irrespectively of whether removable is set or not. It means that *no*
entries for drive (either disc or part?) are created at all and there is
nothing that later on triggers revalidation.

I still believe grok_partitions() should create disc node for removables
even if current capacity == 0 (that for removable just means there is no
media inserted). Then it would work the way you describe and faking
drive capacity for empty drives could be removed (IIRC it happens for
other removables like SCSI disks as well).


-andrej

Re: ide-floppy.c vs devfs

[Paul Bristow]

The problem is one of making it user-comprehensible.

If I just use grok_partitions I get no disc nodes created in devfs.  If 
I do my thing I get a floppy node created, regardless of wether a disk 
is inserted at load time.  This gives something for you to attempt to 
mount after inserting a disk, that causes a revalidate, and gets the 
correct disk nodes created.

This is made somewhat more complicated by the fact that ide-floppy disks 
can use either the whole disk, with no partition table or, more 
commonly, partition4.  So a user-friendly solution would be to create a 
floppy node that pointed to the partition, if it existed, or the whole 
disk if it didn't.  With appropriate code to handle that fact that 
anyone can partition these disks in any way they like.

Note this doesn't take account of the nice ATAPI command that sets the 
disk into "ignore track 0" mode, making a partition 4 look like an 
entire floppy with 1 less track.

Anyone up to telling me how this is handled in the SCSI layer?

Borsenkow Andrej wrote:

>>>Try the devfs test version that I just uploaded to 
>>>
>>>http://paulbristow.net/linux/idefloppy.html 
>>>
>>>This is early days, and I'm not sure what the best approach is... 
>>>
>>>Feedback is greatly appreaciated. 
>>>
> 
> On this page you mention that Mandrake includes this patch, but your
> driver (link on the above patch) differs from Mandrake one. In patch I
> sent to Mandrake I just fake drive capacity in case
> CAPACITY_NO_CARTRIDGE is returned.


Sorry, not trying to detract from anyone's work.  That line refers to 

mandrake having the clik support way before it was included in the stock 

kernel.  I'll clarify the web page.
 
>>I haven't had time to look at this closely, but I question why you're 
>>trying to prevent grok_partitions() from doing it's thing. There's 
>>supposed to be a flag set for removable media which ensures media 
>>revalidation on scanning the parent directory or looking up an entry. 
>>I'd rather see that mechanism fixed. 
>>
> 
> Currently grok_partitions() silently ignores any media with size == 0,
> irrespectively of whether removable is set or not. It means that *no*
> entries for drive (either disc or part?) are created at all and there is
> nothing that later on triggers revalidation.
> 
> I still believe grok_partitions() should create disc node for removables
> even if current capacity == 0 (that for removable just means there is no
> media inserted). Then it would work the way you describe and faking
> drive capacity for empty drives could be removed (IIRC it happens for
> other removables like SCSI disks as well).
> 



-- 

Paul

Email: 
paul@paulbristow.net
Web: 
http://paulbristow.net
ICQ: 
11965223

RE: ide-floppy.c vs devfs

[Borsenkow Andrej]

> 
> This is made somewhat more complicated by the fact that ide-floppy
disks
> can use either the whole disk, with no partition table or, more
> commonly, partition4.  So a user-friendly solution would be to create
a
> floppy node that pointed to the partition, if it existed, or the whole
> disk if it didn't.  With appropriate code to handle that fact that
> anyone can partition these disks in any way they like.
> 

Where's the problem? Use .../disc for whole disc or .../part4 for
"normal" access. (Or /dev/hdc and /dev/hdc4 if you prefer) It is nice if
partition code can detect it but it is not ide-floppy driver problem.

> Note this doesn't take account of the nice ATAPI command that sets the
> disk into "ignore track 0" mode, making a partition 4 look like an
> entire floppy with 1 less track.
> 

Why complicate things more than needed?

> Anyone up to telling me how this is handled in the SCSI layer?
> 

When I boot without media in Jaz drive I get something like "no media
inserted, assuming 1GB 512B per sector". Actually I modeled my patch
from this - use some default values reported by drive when no media
currently exists.

-andrej

Re: ide-floppy.c vs devfs

[Paul Bristow]

Borsenkow Andrej wrote:

>>This is made somewhat more complicated by the fact that ide-floppy
>>
> disks
> 
>>can use either the whole disk, with no partition table or, more
>>commonly, partition4.  So a user-friendly solution would be to create
>>
> a
> 
>>floppy node that pointed to the partition, if it existed, or the whole
>>disk if it didn't.  With appropriate code to handle that fact that
>>anyone can partition these disks in any way they like.
>>
>>
> 
> Where's the problem? Use .../disc for whole disc or .../part4 for
> "normal" access. (Or /dev/hdc and /dev/hdc4 if you prefer) It is nice if
> partition code can detect it but it is not ide-floppy driver problem.


Just wondering if we should be clever for the users here.  Maybe I 
should leave that to user-space tools?  Or is there anything in devfs 
that can take care of this?  The nice solution for end-users might be
a /dev/idefloppy that is a symlink to the relevant node in the 
/dev/ide... tree.


>>Note this doesn't take account of the nice ATAPI command that sets the
>>disk into "ignore track 0" mode, making a partition 4 look like an
>>entire floppy with 1 less track.
>>
>>
> 
> Why complicate things more than needed?

Because you can boot from a zip or ls-120 drive, with the BIOS setting 
it to this mode.  There are disks out there that are unreadable unless 
you ignore track zero, by formatting them in a PC like this.


>>Anyone up to telling me how this is handled in the SCSI layer?
>>
>>
> 
> When I boot without media in Jaz drive I get something like "no media
> inserted, assuming 1GB 512B per sector". Actually I modeled my patch
> from this - use some default values reported by drive when no media
> currently exists.


OK.  This makes the most sense here.  I'm happy to go with this.

I'll dig out your patch - discovered I was on holiday when you 
originally submitted it - and code and test something over the next day 
or so.  Thanks for the help


> -andrej


-- 

Paul

Email: 
paul@paulbristow.net
Web: 
http://paulbristow.net
ICQ: 
11965223

Re: ide-floppy.c vs devfs

[Mike Fedyk]

On Tue, Nov 27, 2001 at 06:47:03PM +0100, Paul Bristow wrote:
> 
> 
> Borsenkow Andrej wrote:
> 
> >>This is made somewhat more complicated by the fact that ide-floppy
> >>
> >disks
> >
> >>can use either the whole disk, with no partition table or, more
> >>commonly, partition4.  So a user-friendly solution would be to create
> >>
> >a
> >
> >>floppy node that pointed to the partition, if it existed, or the whole
> >>disk if it didn't.  With appropriate code to handle that fact that
> >>anyone can partition these disks in any way they like.
> >>
> >>
> >
> >Where's the problem? Use .../disc for whole disc or .../part4 for
> >"normal" access. (Or /dev/hdc and /dev/hdc4 if you prefer) It is nice if
> >partition code can detect it but it is not ide-floppy driver problem.
> 
> 
> Just wondering if we should be clever for the users here.  Maybe I 
> should leave that to user-space tools?  Or is there anything in devfs 
> that can take care of this?  The nice solution for end-users might be
> a /dev/idefloppy that is a symlink to the relevant node in the 
> /dev/ide... tree.
>

Can't devfsd be configured to do something like this?

> >Why complicate things more than needed?
> 
> Because you can boot from a zip or ls-120 drive, with the BIOS setting 
> it to this mode.  There are disks out there that are unreadable unless 
> you ignore track zero, by formatting them in a PC like this.
>

If that's what is required to boot from ls-120 and zip, then it probably
needs to be in kernel space.

MF

[PATCH] ide-floppy real devfs support for testing

[Paul Bristow]

[-- Attachment #1: Type: text/plain, Size: 727 bytes --]

Right :-),


the following patch depends on the previous via fix patch, which is also 
available at http://paulbristow.net/linux/idefloppy.html

Both patches can be applied against 2.4.14-16

This patch implements devfs support in the ide-floppy driver, as it is 
documented in Richards FAQ.

If there is a disk in the drive, a disc node and any partitions are 
created in
   /dev/ide/hostx/busx/targetx/lunx/

and a symlink is created in
   /dev/ide/fd/cxbxtxux to the disc node

If there is no disk in the drive, you just get no partition information, 
a disc node is still created so you have something to revalidate

against.

Cheers,

-- 

Paul

Email: 
paul@paulbristow.net
Web: 
http://paulbristow.net
ICQ: 
11965223




[-- Attachment #2: ide-floppy-devfs-patch --]
[-- Type: text/plain, Size: 5683 bytes --]

diff -burN -X dontdiff linux-2.4.16-clean/drivers/ide/ide-floppy.c linux-2.4.16/drivers/ide/ide-floppy.c
--- linux-2.4.16-clean/drivers/ide/ide-floppy.c	Wed Nov 28 19:25:34 2001
+++ linux-2.4.16/drivers/ide/ide-floppy.c	Wed Nov 28 19:30:36 2001
@@ -1,10 +1,12 @@
 /*
- * linux/drivers/ide/ide-floppy.c	Version 0.97.sv	Jan 14 2001
+ * linux/drivers/ide/ide-floppy.c	
  *
  * Copyright (C) 1996 - 1999 Gadi Oxman <gadio@netvision.net.il>
  * Copyright (C) 2000 - 2001 Paul Bristow <paul@paulbristow.net>
  */
 
+#define IDEFLOPPY_VERSION "0.99-devfs-test4"
+
 /*
  * IDE ATAPI floppy driver.
  *
@@ -82,9 +84,9 @@
  *		msec works on my system. The variable ticks is exposed 
  *		in /proc/ide/hdx/settings. Each tick is 10 msec. If ticks is
  *		set to zero, the driver reverts to the old algorithm. --Skip
+ * Ver 0.99   Nov 22 01  Add devfs support
  */
 
-#define IDEFLOPPY_VERSION "0.98"
 
 #include <linux/config.h>
 #include <linux/module.h>
@@ -101,6 +103,7 @@
 #include <linux/slab.h>
 #include <linux/cdrom.h>
 #include <linux/ide.h>
+#include <linux/devfs_fs_kernel.h>
 
 #include <asm/byteorder.h>
 #include <asm/irq.h>
@@ -275,6 +278,7 @@
  */
 typedef struct {
 	ide_drive_t *drive;
+	devfs_handle_t de;		/* devfs entry */
 
 	idefloppy_pc_t *pc;			/* Current packet command */
 	idefloppy_pc_t *failed_pc; 		/* Last failed packet command */
@@ -673,6 +677,8 @@
 #define IDEFLOPPY_MIN(a,b)	((a)<(b) ? (a):(b))
 #define	IDEFLOPPY_MAX(a,b)	((a)>(b) ? (a):(b))
 
+extern devfs_handle_t ide_devfs_handle; /* Hook into ide devfs chain */
+
 /*
  *	Too bad. The drive wants to send us data which we are not ready to accept.
  *	Just throw it away.
@@ -1097,6 +1103,7 @@
 	return ide_started;
 }
 
+
 /*
  *	Issue a packet command
  */
@@ -1483,9 +1490,14 @@
                         if (memcmp (descriptor, &floppy->capacity, sizeof (idefloppy_capacity_descriptor_t)))
                                 printk (KERN_INFO "%s: %dkB, %d blocks, %d sector size\n", drive->name, blocks * length / 1024, blocks, length);
                         floppy->capacity = *descriptor;
-                        if (!length || length % 512)
+			                  if (!length || length % 512) {
                                 printk (KERN_NOTICE "%s: %d bytes block size not supported\n", drive->name, length);
-                        else {
+		                    } else if (!i && descriptor->dc == CAPACITY_NO_CARTRIDGE                                   && drive->removable && !(length % 512)) {
+			                    /* Set these two so that idefloppy_capacity returns a 
+			                       positive value, needed for devfs registration. */
+			                    floppy->blocks = blocks;
+			                    floppy->bs_factor = length / 512;
+			                  } else {
                                 floppy->blocks = blocks;
                                 floppy->block_size = length;
                                 if ((floppy->bs_factor = length / 512) != 1)
@@ -2027,6 +2039,7 @@
 	ide_add_setting(drive,	"ticks",		SETTING_RW,	-1,		-1,		TYPE_BYTE,	0,	255,		1,	1,	&floppy->ticks,			NULL);
 }
 
+
 /*
  *	Driver initialization.
  */
@@ -2035,6 +2048,9 @@
 	struct idefloppy_id_gcw gcw;
 	int major = HWIF(drive)->major, i;
 	int minor = drive->select.b.unit << PARTN_BITS;
+	char fname[64],iname[64]; /* for devfs */
+ 	devfs_handle_t idefloppy_devfs_handle; /* for devfs */
+ 	ide_hwif_t *hwif = HWIF(drive);
 
 	*((unsigned short *) &gcw) = drive->id->config;
 	drive->driver_data = floppy;
@@ -2080,26 +2096,46 @@
 		set_bit(IDEFLOPPY_CLIK_DRIVE, &floppy->flags);
 	}
 
-
 	(void) idefloppy_get_capacity (drive);
 	idefloppy_add_settings(drive);
+	
 	for (i = 0; i < MAX_DRIVES; ++i) {
-		ide_hwif_t *hwif = HWIF(drive);
 
 		if (drive != &hwif->drives[i]) continue;
 		hwif->gd->de_arr[i] = drive->de;
+
 		if (drive->removable)
 			hwif->gd->flags[i] |= GENHD_FL_REMOVABLE;
 		break;
 	}
+  
+  /* Always register drive with devfs */
+	floppy->de = devfs_register (drive->de, "disc", DEVFS_FL_REMOVABLE,
+	       			 major, minor,
+				       S_IFBLK | S_IRUGO | S_IWUGO,
+				       ide_fops, NULL);
+	/* Create ide/fd entry in devfs */			       
+  idefloppy_devfs_handle = devfs_mk_dir(ide_devfs_handle,"fd",NULL);
+	
+	sprintf (fname, "c%db%dt%du%d",
+		 (hwif->channel && hwif->mate) ? hwif->mate->index : hwif->index,
+  		hwif->channel, i, hwif->drives[i].lun);
+	sprintf (iname, "../host%d/bus%d/target%d/lun%d/disc",
+		 (hwif->channel && hwif->mate) ? hwif->mate->index : hwif->index,
+  		hwif->channel, i, hwif->drives[i].lun);
+  devfs_mk_symlink(idefloppy_devfs_handle, fname, DEVFS_FL_REMOVABLE, iname, NULL, NULL);
 }
 
+
 static int idefloppy_cleanup (ide_drive_t *drive)
 {
 	idefloppy_floppy_t *floppy = drive->driver_data;
 
 	if (ide_unregister_subdriver (drive))
 		return 1;
+	
+	devfs_unregister(floppy->de);
+
 	drive->driver_data = NULL;
 	kfree (floppy);
 	return 0;
@@ -2212,6 +2248,7 @@
 
 module_init(idefloppy_init);
 module_exit(idefloppy_exit);
+
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Maintainer: Paul Bristow <paul@paulbristow.net>");
 MODULE_DESCRIPTION("ATAPI Floppy Driver V"IDEFLOPPY_VERSION);
diff -burN -X dontdiff linux-2.4.16-clean/drivers/ide/ide-probe.c linux-2.4.16/drivers/ide/ide-probe.c
--- linux-2.4.16-clean/drivers/ide/ide-probe.c	Mon Nov 26 14:29:17 2001
+++ linux-2.4.16/drivers/ide/ide-probe.c	Wed Nov 28 18:49:38 2001
@@ -121,6 +121,7 @@
 					if (!strstr(id->model, "oppy") && !strstr(id->model, "poyp") && !strstr(id->model, "ZIP"))
 						printk("cdrom or floppy?, assuming ");
 					if (drive->media != ide_cdrom) {
+				    drive->removable = 1;
 						printk ("FLOPPY");
 						break;
 					}

Re: ide-floppy.c vs devfs

[Borsenkow Andrej]

The fix for it is included in current Mandrake cooker. As I already
said, I do not like it (the fix is to pretend we have media when we have
none) but it is the only possible way given current partition layer
design it seems.

I would prefer that partition layer registered .../disc for removable
media irrespectively of whether media is inserted or not, but well ...
it's up to other people to decide.

-andrej

Sorry, replying off web archive.

ide-floppy.c vs devfs

[Jack Howarth]

Paul,
   Is ide-floppy supposed to be devfs friendly? I ask
because on Linux 2.4.15-pre2 (and previous kernels)
on the Powermac (Debian ppc sid) I find that the
ide-floppy driver only creates device for my ide 
zip if a zip disk is inserted at boot time. Otherwise
the /dev/ide/host1/bus0/target1/lun0 directory doesn't
contain a device node for the zip whereas the ide-cdrom
driver creates a /dev/ide/host1/bus0/target0/lun0/cd
fine without a cdrom disk inserted at boot time. I find
that the only way I can get the zip device created at
boot time without media inserted is to use ide-scsi 
emulation and turn off ide-floppy when I configure the
kernel. Is this issue a known problem and does it
exist for other arches than powerpc (Powermac)? Thanks
in advance for any advice.
                      Jack
------------------------------------------------------------------------------
Jack W. Howarth, Ph.D.                                    231 Albert Sabin Way
NMR Facility Director                              Cincinnati, Ohio 45267-0524
Dept. of Molecular Genetics                              phone: (513) 558-4420
Univ. of Cincinnati College of Medicine                    fax: (513) 558-8474

Re: ide-floppy.c vs devfs

[Richard Gooch]

Jack Howarth writes:
> Paul,
>    Is ide-floppy supposed to be devfs friendly? I ask
> because on Linux 2.4.15-pre2 (and previous kernels)
> on the Powermac (Debian ppc sid) I find that the
> ide-floppy driver only creates device for my ide 
> zip if a zip disk is inserted at boot time. Otherwise
> the /dev/ide/host1/bus0/target1/lun0 directory doesn't
> contain a device node for the zip whereas the ide-cdrom
> driver creates a /dev/ide/host1/bus0/target0/lun0/cd
> fine without a cdrom disk inserted at boot time. I find
> that the only way I can get the zip device created at
> boot time without media inserted is to use ide-scsi 
> emulation and turn off ide-floppy when I configure the
> kernel. Is this issue a known problem and does it
> exist for other arches than powerpc (Powermac)? Thanks
> in advance for any advice.

Is this an LS-120 drive? If so, I'm soon getting one, so I can finally
play with this.

				Regards,

					Richard....
Permanent: rgooch@atnf.csiro.au
Current:   rgooch@ras.ucalgary.ca

Re: ide-floppy.c vs devfs

[Paul Bristow]

Hi Jack,

Try the devfs test version that I just uploaded to

http://paulbristow.net/linux/idefloppy.html

This is early days, and I'm not sure what the best approach is...

Feedback is greatly appreaciated.

Jack Howarth wrote:

> Paul,
>    Is ide-floppy supposed to be devfs friendly? I ask
> because on Linux 2.4.15-pre2 (and previous kernels)
> on the Powermac (Debian ppc sid) I find that the
> ide-floppy driver only creates device for my ide 
> zip if a zip disk is inserted at boot time. Otherwise
> the /dev/ide/host1/bus0/target1/lun0 directory doesn't
> contain a device node for the zip whereas the ide-cdrom
> driver creates a /dev/ide/host1/bus0/target0/lun0/cd
> fine without a cdrom disk inserted at boot time. I find
> that the only way I can get the zip device created at
> boot time without media inserted is to use ide-scsi 
> emulation and turn off ide-floppy when I configure the
> kernel. Is this issue a known problem and does it
> exist for other arches than powerpc (Powermac)? Thanks
> in advance for any advice.
>                       Jack
> ------------------------------------------------------------------------------
> Jack W. Howarth, Ph.D.                                    231 Albert Sabin Way
> NMR Facility Director                              Cincinnati, Ohio 45267-0524
> Dept. of Molecular Genetics                              phone: (513) 558-4420
> Univ. of Cincinnati College of Medicine                    fax: (513) 558-8474
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 


-- 

Paul

Email: 
paul@paulbristow.net
Web: 
http://paulbristow.net
ICQ: 
11965223

Re: ide-floppy.c vs devfs

[Richard Gooch]

Paul Bristow writes:
> Hi Jack,
> 
> Try the devfs test version that I just uploaded to
> 
> http://paulbristow.net/linux/idefloppy.html
> 
> This is early days, and I'm not sure what the best approach is...
> 
> Feedback is greatly appreaciated.

I haven't had time to look at this closely, but I question why you're
trying to prevent grok_partitions() from doing it's thing. There's
supposed to be a flag set for removable media which ensures media
revalidation on scanning the parent directory or looking up an entry.
I'd rather see that mechanism fixed.

				Regards,

					Richard....
Permanent: rgooch@atnf.csiro.au
Current:   rgooch@ras.ucalgary.ca