OSDev.org

Brendan wrote:

Kevin wrote:

Nobody has managed to define a successful common specification yet (even when large companies like SCO, Sun, Intel and IBM collaborate they just ended up with "stillborn" specs); so I doubt a lone hobbyist like me (who doesn't care about common device driver specifications to begin with) will succeed where everyone else failed.

Define "success". For me using a common interface was a success as soon as I ran the first driver that wasn't primarily written for tyndur. It saved me some work that I could better invest in other parts of the OS.

You did a huge amount of work to support UDI, and after all of that work you've managed to use one driver that provides nothing more than basic functionality? Awesome.

Hello there,

Tyndur does not use UDI. They use the simple CDI(Common Driver Interface), which was only developed to share some basic drivers for hobby OSs IIRC. So apparently Kevin and the other developers of tyndur did not spend a huge amount of work to implement CDI.

Hi,

Combuster wrote:

For my OS; it's a micro-kernel. If one process (e.g. file system code) establishes a connection to another process (e.g. "mounts" a storage device driver) and sends a message saying "I want 1234 bytes at offset 12345678", then that message goes directly to the other process. It is possible to have a middle-man (e.g. something in between the file system code and the storage device driver that does encryption), but this increases the communication overheads (task switching between processes, etc).

Even so, if you would include a copy of the same source code that does the encryption in every driver, you end up with bugs and updates equally repeated among all of them. And since you also want the option of doing encryption on a laptop hard drive, you will still put that code in one place and include it in every driver.

More like I'd put the encryption in the kernel for all software to use.

Combuster wrote:Your code to the UDI environment is in one place, and it's included in every driver.

In the "driver driver" shim thing that wasn't supposed to be necessary if the OS uses UDI as its native driver interface?

Combuster wrote:

Of course my example was not for my OS and was not for any one specific OS. If there are 1000 very different OSs and 5% of those OS decide to put encryption in the storage device driver, then 5% of OSs are screwed.

Of course they aren't. Any driver without encryption can be turned into a driver with encryption because it's not hardware related at all.

Sure; but which side of the "driver interface" abstraction is encryption on? On the driver's side (where it has to be if the hardware supports encryption), or in the file system/s where it can't work, or in layer/s of unnecessary bloat between the file system and driver?

Combuster wrote:

Encryption has no business in hardware drivers, except if the device does encryption in hardware.

If one specific device's hardware supports encryption, then the OS's standard interface used by all of its device drivers should include (e.g.) a "set_encryption_key()" function; and if one specific device's hardware doesn't support encryption then the OS's standard interface used by all of its device drivers should not include (e.g.) a "set_encryption_key()" function?

Did you deliberately put that contradiction in there as an attempt to turn a statement into something completely unrelated, or was that a coincidental interpretation error?

It's a deliberate contradiction. Either the driver always takes care of encryption (via. hardware, a library, a kernel API, or with its own code), or the driver never takes care of encryption (and its done by other layers that can't take advantage of any encryption built into the device), or it's a bad/inconsistent abstraction.

Combuster wrote:

Some OSs may be so stupid and broken that they put I/O scheduling in the block layer (where it's unable to make decisions that take into account device specific things, like where the disk drive's heads happen to be)

If they're stupid enough to suffer from general dementia, then perhaps. But most schedulers will at least make an attempt to remember what the last written block was.

Remembering where the last written block was is fairly useless for generic code (on the wrong side of the abstraction) that doesn't know if the device is SSD or mechanical disk or hybrid; or (for mechanical disk) if "block number 12345" is on the same cylinder as "block number 12346" or not.

Combuster wrote:

Do you even know what a device driver interface is? It's supposed to be a clean abstraction that covers a super-set of features that all devices provide.

Therefore all video cards must provide an audio interface because it might have an HDMI port? Therefore all sound cards have to provide a storage interface because old soundblasters came with their CD bus controller when ATAPI wasn't a thing yet? Or a simple harddisk has to know proper responses to anything a blu-ray burner might be doing?

Here's the part you deliberately ignored for your misguided and juvenile attempt to take my words out of context: "Of course there are many different categories for devices (e.g. storage device, sound, printer, network, video, etc) and therefore (to keep the abstraction for a specific category clean) there's a set of different device driver interfaces (one for each category).

Basically; one interface for storage devices (that includes things like read, write, media change, secure erase, trim, etc); one interface for sound (where nothing used for storage devices makes any sense; that includes things like sending/receiving streams of digitised sound and/or MIDI, volume and balance controls, mixers, effects like echo, etc); one interface for printers (where nothing used for storage devices or sound cards makes any sense), etc.

Combuster wrote:This is why in practice, interfaces actually are a subset rather than a superset, and they offer extensions when they are relevant for that device.

SCSI is not a clean abstraction

Are you suggesting that because a bus protocol is too generic, you're not allowed to write drivers for both the host controller, and by extension having no option but not to support it's child devices as well?

I'm suggesting that the SCSI command set is not a clean abstraction for any of the different categories of devices. Furthermore, I will suggest that UDI doesn't provide adequate abstractions for any of the different categories of devices; except for SCSI controllers and USB controllers.


Cheers,

Brendan

Brendan wrote:More like I'd put the encryption in the kernel for all software to use.

I thought you were suggesting it would be in the device driver?

Brendan wrote:

Combuster wrote:Your code to the UDI environment is in one place, and it's included in every driver.

In the "driver driver" shim thing that wasn't supposed to be necessary if the OS uses UDI as its native driver interface?

What does this mean? If you have a native driver interface of any kind, you'll need to have code on the OS side to talk to the drivers and probably provide support routines (e.g. resource allocation) as well. That's what UDI calls the "environment". It's the only code which is custom to the operating system, and where policy is set (e.g. single-threaded vs. multitasking, synchronous vs. asynchronous, locking, etc.) -- and unfortunately it's the most difficult part of UDI, as I understand it.

Brendan wrote:Sure; but which side of the "driver interface" abstraction is encryption on? On the driver's side (where it has to be if the hardware supports encryption), or in the file system/s where it can't work, or in layer/s of unnecessary bloat between the file system and driver?

UDI does support chaining drivers together -- I'm not familiar with the specifics, but it seems to me that the "UDI way" would probably be to make an "encryption driver" to provide the functionality, then configure the OS to chain it with the storage driver. (It might require a "metalanguage" for encryption, which probably doesn't exist at this time.) I don't see that as "bloat" if you're only adding the layer if and when you want encryption added.

Brendan wrote:It's a deliberate contradiction. Either the driver always takes care of encryption (via. hardware, a library, a kernel API, or with its own code), or the driver never takes care of encryption (and its done by other layers that can't take advantage of any encryption built into the device), or it's a bad/inconsistent abstraction.

Perhaps there are other options you haven't thought of?

Brendan wrote:

Combuster wrote:

Some OSs may be so stupid and broken that they put I/O scheduling in the block layer (where it's unable to make decisions that take into account device specific things, like where the disk drive's heads happen to be)

If they're stupid enough to suffer from general dementia, then perhaps. But most schedulers will at least make an attempt to remember what the last written block was.

Remembering where the last written block was is fairly useless for generic code (on the wrong side of the abstraction) that doesn't know if the device is SSD or mechanical disk or hybrid; or (for mechanical disk) if "block number 12345" is on the same cylinder as "block number 12346" or not.

Even for mechanical disks, you never know if two blocks are really on the same cylinder. Sector reallocation is always a possibility, and I believe it's "transparent" even to the operating system. (But surely slower!)

Brendan wrote:

Combuster wrote:Are you suggesting that because a bus protocol is too generic, you're not allowed to write drivers for both the host controller, and by extension having no option but not to support it's child devices as well?

I'm suggesting that the SCSI command set is not a clean abstraction for any of the different categories of devices. Furthermore, I will suggest that UDI doesn't provide adequate abstractions for any of the different categories of devices; except for SCSI controllers and USB controllers.

If the SCSI command set is so poor, why did Linux change from direct IDE support to pretending IDE drives are actually SCSI?

Deven

Hi,

Did you know this post was been dead for almost a year before you posted?

Deven wrote:

Brendan wrote:More like I'd put the encryption in the kernel for all software to use.

I thought you were suggesting it would be in the device driver?

You're confusing interface and implementation. The standard interface for a device driver might say "device driver is responsible for providing encryption", but that doesn't mean the kernel can't provide a default implementation that device drivers can use to fulfil that responsibility (if the device doesn't support it in hardware).

Deven wrote:

Brendan wrote:

Combuster wrote:Your code to the UDI environment is in one place, and it's included in every driver.

In the "driver driver" shim thing that wasn't supposed to be necessary if the OS uses UDI as its native driver interface?

What does this mean? If you have a native driver interface of any kind, you'll need to have code on the OS side to talk to the drivers and probably provide support routines (e.g. resource allocation) as well. That's what UDI calls the "environment". It's the only code which is custom to the operating system, and where policy is set (e.g. single-threaded vs. multitasking, synchronous vs. asynchronous, locking, etc.) -- and unfortunately it's the most difficult part of UDI, as I understand it.

I don't know what "OS side" is (typically "OS" is kernel and drivers and file systems and GUI and utilities and boot code and installer and standard fonts and ...). You don't need code on the kernel side to talk to drivers (e.g. virtually every micro-kernel that's ever existed).

Deven wrote:

Brendan wrote:It's a deliberate contradiction. Either the driver always takes care of encryption (via. hardware, a library, a kernel API, or with its own code), or the driver never takes care of encryption (and its done by other layers that can't take advantage of any encryption built into the device), or it's a bad/inconsistent abstraction.

Perhaps there are other options you haven't thought of?

And perhaps if there are options I didn't think of someone might suggest an alternative option?

Deven wrote:

Brendan wrote:

Combuster wrote:If they're stupid enough to suffer from general dementia, then perhaps. But most schedulers will at least make an attempt to remember what the last written block was.

Remembering where the last written block was is fairly useless for generic code (on the wrong side of the abstraction) that doesn't know if the device is SSD or mechanical disk or hybrid; or (for mechanical disk) if "block number 12345" is on the same cylinder as "block number 12346" or not.

Even for mechanical disks, you never know if two blocks are really on the same cylinder. Sector reallocation is always a possibility, and I believe it's "transparent" even to the operating system. (But surely slower!)

Maybe, but it's a corner-case that misses the point.

Deven wrote:

Brendan wrote:

Combuster wrote:Are you suggesting that because a bus protocol is too generic, you're not allowed to write drivers for both the host controller, and by extension having no option but not to support it's child devices as well?

I'm suggesting that the SCSI command set is not a clean abstraction for any of the different categories of devices. Furthermore, I will suggest that UDI doesn't provide adequate abstractions for any of the different categories of devices; except for SCSI controllers and USB controllers.

If the SCSI command set is so poor, why did Linux change from direct IDE support to pretending IDE drives are actually SCSI?

I don't know why Linux developers consistently make bad choices. I tried to find out, but all I found is this quote from Linus himself:

"ide-scsi has always been broken. You should not use it, and indeed there was never any good reason for it existing AT ALL. But because of a broken interface to cdrecord, cdrecord historically only wanted to touch SCSI devices. Ergo, a silly emulation layer that wasn't really worth it."


Cheers,

Brendan