Skip to contentSkip to conversation

Let's get rid of ide-scsi

DISCUSSION Alex Davis 6 participants 12 messages Feb 10, 2006

lore
Share & tools
lore

Alex Davis

author
#0 lore raw

I think we should get rid of ide-scsi.

Reasons:
1) It's broken.
2) It's unmaintained.
3) It's unneeded.

I'll submit a patch if people agree.

I code, therefore I am

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

#1 lore raw
Quoted 10 lines · Alex Davis

I think we should get rid of ide-scsi.

Reasons:
1) It's broken.
2) It's unmaintained.
3) It's unneeded.

I'll submit a patch if people agree.

I code, therefore I am

I personally do not agree with this. I worked on at boot disk(floppy) which contained the kernel and modules to find a cdrom (or usb disk) and use it as my 2nd stage. If I had to use ide-cd, I would not beable to do my first stage loader on a single floppy (I support ide and scsi cdroms via sr-mod).

ide-cd.ko is > than sr-mod.ko + ide-scsi.ko

I am aware that scsi_mod.ko is larger than those 3 combined and I still need it regardless for usb.

My personal vote would be to drop the entire ide subsystem which would thus drop ide-scsi. The SCSI layer has been a general block device layer for more than true scsi devices. USB, Firewire, and SATA use the scsi layer. And as I understand it, libata is starting to handle PATA devices. Once it can handle PATA fine, the ide code would pretty much be useless.

I am also against the seperate USB block layer, I personally saw no use in it.

Signature · 2 lines
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???
#2 lore raw
Quoted 3 lines · Wakko Warner

I am also against the seperate USB block layer, I personally saw no use in it.

What "seperate USB block layer"?

thanks,

greg k-h

#3 lore raw
Quoted 11 lines · Greg KH
Quoted 2 lines · Wakko Warner

I am also against the seperate USB block layer, I personally saw no use in it.

What "seperate USB block layer"?

thanks,

greg k-h

I assume he means CONFIG_BLK_DEV_UB.

Signature · 1 line
Joshua Kwan
author
#8 lore raw
In reply to Wakko Warner
Quoted 22 lines · --- Wakko Warner
Quoted 10 lines · Alex Davis

I think we should get rid of ide-scsi.

Reasons:
1) It's broken.
2) It's unmaintained.
3) It's unneeded.

I'll submit a patch if people agree.

I code, therefore I am

I personally do not agree with this. I worked on at boot disk(floppy) which contained the kernel and modules to find a cdrom (or usb disk) and use it as my 2nd stage. If I had to use ide-cd, I would not beable to do my first stage loader on a single floppy (I support ide and scsi cdroms via sr-mod).

ide-cd.ko is > than sr-mod.ko + ide-scsi.ko

I am aware that scsi_mod.ko is larger than those 3 combined and I still need it regardless for usb.

Quoted 13 lines

My personal vote would be to drop the entire ide subsystem which would thus drop ide-scsi. The SCSI layer has been a general block device layer for more than true scsi devices. USB, Firewire, and SATA use the scsi layer. And as I understand it, libata is starting to handle PATA devices. Once it can handle PATA fine, the ide code would pretty much be useless.

I am also against the seperate USB block layer, I personally saw no use in it.

Signature · 2 lines
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???

Wakko:

Modules can be compressed: On a 2.6.15 kernel doing a 'gzip -9 idecd.ko' reduced its size from 43616 bytes to 19234 bytes. The only additional step is modifying 'modules.dep' and changing idecd.ko to idecd.ko.gz. You now have a fully functional ide cdrom driver.

I code, therefore I am

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

#9 lore raw
Quoted 25 lines · Alex Davis
Quoted 10 lines · --- Wakko Warner

I personally do not agree with this. I worked on at boot disk(floppy) which contained the kernel and modules to find a cdrom (or usb disk) and use it as my 2nd stage. If I had to use ide-cd, I would not beable to do my first stage loader on a single floppy (I support ide and scsi cdroms via sr-mod).

ide-cd.ko is > than sr-mod.ko + ide-scsi.ko

I am aware that scsi_mod.ko is larger than those 3 combined and I still need it regardless for usb.

Quoted 13 lines

My personal vote would be to drop the entire ide subsystem which would thus drop ide-scsi. The SCSI layer has been a general block device layer for more than true scsi devices. USB, Firewire, and SATA use the scsi layer. And as I understand it, libata is starting to handle PATA devices. Once it can handle PATA fine, the ide code would pretty much be useless.

I am also against the seperate USB block layer, I personally saw no use in it.

Signature · 2 lines
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???
Quoted 5 lines

Wakko:

Modules can be compressed: On a 2.6.15 kernel doing a 'gzip -9 idecd.ko' reduced its size from 43616 bytes to 19234 bytes. The only additional step is modifying 'modules.dep' and changing idecd.ko to idecd.ko.gz. You now have a fully functional ide cdrom driver.

This I did not know. I'm not sure if it will really matter or not. The initramfs is already gzip -9'd. I have a list of modules that are required for stage 1 which pulls in the dependancies for those modules. It does currently fit on a single floppy. I'm using a upx compressed kernel, a gzip -9'd initramfs, kernel is compiled with -Os and I'm using a -Os compiled busybox statically compiled with uclibc. When it's all said and done, I have less than 10kb available on a floppy. I thought it was quite an acomplishment getting all that one 1 floppy.

Signature · 2 lines
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???
#4 lore raw
In reply to Greg KH
Quoted 6 lines · Greg KH
Quoted 3 lines · Wakko Warner

I am also against the seperate USB block layer, I personally saw no use in it.

What "seperate USB block layer"?

Maybe not a "block layer", but there was this Under drivers/block devices in the config:
Low Performance USB Block driver

Signature · 2 lines
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???
#10 lore raw
In reply to Alex Davis
Quoted 6 lines · Alex Davis

I think we should get rid of ide-scsi.

Reasons:
1) It's broken.
2) It's unmaintained.
3) It's unneeded.

#1 is half wrong
#2 is half wrong
#3 is totally wrong

There are devices such as multichangers that need it.

Please wait for the longer term cure - killing off drivers/ide 8).

Alan

#5 lore raw
In reply to Wakko Warner
Quoted 11 lines · Wakko Warner
Quoted 6 lines · Greg KH
Quoted 3 lines · Wakko Warner

I am also against the seperate USB block layer, I personally saw no use in it.

What "seperate USB block layer"?

Maybe not a "block layer", but there was this Under drivers/block devices in the config:
Low Performance USB Block driver

What is your objection to this driver? It fills a real need for people who do not want the whole scsi stack in their kernels (embedded, memory constraints, closed systems, etc.), and probably is not even considered "Low Performance" anymore.

thanks,

greg k-h

#6 lore raw
Quoted 4 lines · Greg KH
Quoted 3 lines · Wakko Warner
Quoted 1 line · Greg KH

What "seperate USB block layer"?

Quoted 8 lines
Quoted 3 lines

Maybe not a "block layer", but there was this Under drivers/block devices in the config:
Low Performance USB Block driver

What is your objection to this driver? It fills a real need for people who do not want the whole scsi stack in their kernels (embedded, memory constraints, closed systems, etc.), and probably is not even considered "Low Performance" anymore.

Ok, now this I did not know which is why I personally objected to it. I saw no reason to have it with usb-storage since both did something similar. Now that I know what it's purpose is, I don't see a problem with it as far as availability to the ones who are low memory, embedded, etc, but I won't need it myself. I normally use systtems with scsi controllers and need the full scsi layer.

If/When libata takes over ide in general, how many of these machine will then require the scsi layer? I would think all systems would except ones without internal disks (non-usb/firewire).

I do appreciate the info, thanks.

Signature · 2 lines
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???
#7 lore raw
Quoted 3 lines · Wakko Warner

If/When libata takes over ide in general, how many of these machine will then require the scsi layer? I would think all systems would except ones without internal disks (non-usb/firewire).

You'll want libata (but not eventually all of the scsi layer) for just about anything at that point. On the bright side you won't have scsi loaded for your USB devices and drivers/ide loaded for your IDE disks so for most cases I suspect it will be neutral or save memory.

If you are really tighht on memory and just using CF then its probably worth writing a simple CF driver for embedded use. Its probably a matter of 10K of code if that for the subset in question.

#11 lore raw
In reply to Alex Davis
Quoted 8 lines · Alex Davis

I think we should get rid of ide-scsi.

Reasons:
1) It's broken.
2) It's unmaintained.
3) It's unneeded.

I'll submit a patch if people agree.

People don't agree, could we not have another war like cdrecord? Please?

Go back and read the old posts, particularly discussing tape, MO drives, and some ide-floppy devices.

It's there, it's maintained as needed, and it works far better than the alternatives for some devices.

Signature · 3 lines
bill davidsen <[email protected]>
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979